145 files changed, 74587 insertions, 13791 deletions

diff --git a/drivers/md/.gitignore b/drivers/md/.gitignore
deleted file mode 100644
index a7afec6b19c..00000000000
--- a/drivers/md/.gitignore
+++ /dev/null
@@ -1,4 +0,0 @@
-mktables
-raid6altivec*.c
-raid6int*.c
-raid6tables.c
diff --git a/drivers/md/Kconfig b/drivers/md/Kconfig
index 07d92c11b5d..5bdedf6df15 100644
--- a/drivers/md/Kconfig
+++ b/drivers/md/Kconfig
@@ -30,6 +30,20 @@ config BLK_DEV_MD
 
 	  If unsure, say N.
 
+config MD_AUTODETECT
+	bool "Autodetect RAID arrays during kernel boot"
+	depends on BLK_DEV_MD=y
+	default y
+	---help---
+	  If you say Y here, then the kernel will try to autodetect raid
+	  arrays as part of its boot process. 
+
+	  If you don't use raid and say Y, this autodetection can cause 
+	  a several-second delay in the boot time due to various
+	  synchronisation steps that are part of this step.
+
+	  If unsure, say Y.
+
 config MD_LINEAR
 	tristate "Linear (append) mode"
 	depends on BLK_DEV_MD
@@ -86,8 +100,8 @@ config MD_RAID1
 	  If unsure, say Y.
 
 config MD_RAID10
-	tristate "RAID-10 (mirrored striping) mode (EXPERIMENTAL)"
-	depends on BLK_DEV_MD && EXPERIMENTAL
+	tristate "RAID-10 (mirrored striping) mode"
+	depends on BLK_DEV_MD
 	---help---
 	  RAID-10 provides a combination of striping (RAID-0) and
 	  mirroring (RAID-1) with easier configuration and more flexible
@@ -107,8 +121,11 @@ config MD_RAID10
 config MD_RAID456
 	tristate "RAID-4/RAID-5/RAID-6 mode"
 	depends on BLK_DEV_MD
+	select RAID6_PQ
 	select ASYNC_MEMCPY
 	select ASYNC_XOR
+	select ASYNC_PQ
+	select ASYNC_RAID6_RECOV
 	---help---
 	  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
@@ -137,44 +154,14 @@ config MD_RAID456
 
 	  If unsure, say Y.
 
-config MD_RAID5_RESHAPE
-	bool "Support adding drives to a raid-5 array"
-	depends on MD_RAID456
-	default y
-	---help---
-	  A RAID-5 set can be expanded by adding extra drives. This
-	  requires "restriping" the array which means (almost) every
-	  block must be written to a different place.
-
-          This option allows such restriping to be done while the array
-	  is online.
-
-	  You will need mdadm version 2.4.1 or later to use this
-	  feature safely.  During the early stage of reshape there is
-	  a critical section where live data is being over-written.  A
-	  crash during this time needs extra care for recovery.  The
-	  newer mdadm takes a copy of the data in the critical section
-	  and will restore it, if necessary, after a crash.
-
-	  The mdadm usage is e.g.
-	       mdadm --grow /dev/md1 --raid-disks=6
-	  to grow '/dev/md1' to having 6 disks.
-
-	  Note: The array can only be expanded, not contracted.
-	  There should be enough spares already present to make the new
-	  array workable.
-
-	  If unsure, say Y.
-
 config MD_MULTIPATH
 	tristate "Multipath I/O support"
 	depends on BLK_DEV_MD
 	help
-	  Multipath-IO is the ability of certain devices to address the same
-	  physical disk over multiple 'IO paths'. The code ensures that such
-	  paths can be defined and handled at runtime, and ensures that a
-	  transparent failover to the backup path(s) happens if a IO errors
-	  arrives on the primary path.
+	  MD_MULTIPATH provides a simple multi-path personality for use
+	  the MD framework.  It is not under active development.  New
+	  projects should consider using DM_MULTIPATH which has more
+	  features and more testing.
 
 	  If unsure, say N.
 
@@ -187,8 +174,14 @@ config MD_FAULTY
 
 	  In unsure, say N.
 
+source "drivers/md/bcache/Kconfig"
+
+config BLK_DEV_DM_BUILTIN
+	boolean
+
 config BLK_DEV_DM
 	tristate "Device mapper support"
+	select BLK_DEV_DM_BUILTIN
 	---help---
 	  Device-mapper is a low level volume manager.  It works by allowing
 	  people to specify mappings for ranges of logical sectors.  Various
@@ -210,6 +203,23 @@ config DM_DEBUG
 
 	  If unsure, say N.
 
+config DM_BUFIO
+       tristate
+       depends on BLK_DEV_DM
+       ---help---
+	 This interface allows you to do buffered I/O on a device and acts
+	 as a cache, holding recently-read blocks in memory and performing
+	 delayed writes.
+
+config DM_BIO_PRISON
+       tristate
+       depends on BLK_DEV_DM
+       ---help---
+	 Some bio locking schemes used by other device-mapper targets
+	 including thin provisioning.
+
+source "drivers/md/persistent-data/Kconfig"
+
 config DM_CRYPT
 	tristate "Crypt target support"
 	depends on BLK_DEV_DM
@@ -232,9 +242,60 @@ config DM_CRYPT
 config DM_SNAPSHOT
        tristate "Snapshot target"
        depends on BLK_DEV_DM
+       select DM_BUFIO
        ---help---
          Allow volume managers to take writable snapshots of a device.
 
+config DM_THIN_PROVISIONING
+       tristate "Thin provisioning target"
+       depends on BLK_DEV_DM
+       select DM_PERSISTENT_DATA
+       select DM_BIO_PRISON
+       ---help---
+         Provides thin provisioning and snapshots that share a data store.
+
+config DM_CACHE
+       tristate "Cache target (EXPERIMENTAL)"
+       depends on BLK_DEV_DM
+       default n
+       select DM_PERSISTENT_DATA
+       select DM_BIO_PRISON
+       ---help---
+         dm-cache attempts to improve performance of a block device by
+         moving frequently used data to a smaller, higher performance
+         device.  Different 'policy' plugins can be used to change the
+         algorithms used to select which blocks are promoted, demoted,
+         cleaned etc.  It supports writeback and writethrough modes.
+
+config DM_CACHE_MQ
+       tristate "MQ Cache Policy (EXPERIMENTAL)"
+       depends on DM_CACHE
+       default y
+       ---help---
+         A cache policy that uses a multiqueue ordered by recent hit
+         count to select which blocks should be promoted and demoted.
+         This is meant to be a general purpose policy.  It prioritises
+         reads over writes.
+
+config DM_CACHE_CLEANER
+       tristate "Cleaner Cache Policy (EXPERIMENTAL)"
+       depends on DM_CACHE
+       default y
+       ---help---
+         A simple cache policy that writes back all data to the
+         origin.  Used when decommissioning a dm-cache.
+
+config DM_ERA
+       tristate "Era target (EXPERIMENTAL)"
+       depends on BLK_DEV_DM
+       default n
+       select DM_PERSISTENT_DATA
+       select DM_BIO_PRISON
+       ---help---
+         dm-era tracks which parts of a block device are written to
+         over time.  Useful for maintaining cache coherency when using
+         vendor snapshots.
+
 config DM_MIRROR
        tristate "Mirror target"
        depends on BLK_DEV_DM
@@ -242,6 +303,43 @@ config DM_MIRROR
          Allow volume managers to mirror logical volumes, also
          needed for live data migration tools such as 'pvmove'.
 
+config DM_LOG_USERSPACE
+	tristate "Mirror userspace logging"
+	depends on DM_MIRROR && NET
+	select CONNECTOR
+	---help---
+	  The userspace logging module provides a mechanism for
+	  relaying the dm-dirty-log API to userspace.  Log designs
+	  which are more suited to userspace implementation (e.g.
+	  shared storage logs) or experimental logs can be implemented
+	  by leveraging this framework.
+
+config DM_RAID
+       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, 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
+	 of a single drive. For a given sector (row) number, (N - 1) drives
+	 contain data sectors, and one drive contains the parity protection.
+	 For a RAID-4 set, the parity blocks are present on a single drive,
+	 while a RAID-5 set distributes the parity across the drives in one
+	 of the available parity distribution methods.
+
+	 A RAID-6 set of N drives with a capacity of C MB per drive
+	 provides the capacity of C * (N - 2) MB, and protects
+	 against a failure of any two drives. For a given sector
+	 (row) number, (N - 2) drives contain data sectors, and two
+	 drives contains two independent redundancy syndromes.  Like
+	 RAID-5, RAID-6 distributes the syndromes across the drives
+	 in one of the available parity distribution methods.
+
 config DM_ZERO
 	tristate "Zero target"
 	depends on BLK_DEV_DM
@@ -260,9 +358,28 @@ config DM_MULTIPATH
 	---help---
 	  Allow volume managers to support multipath hardware.
 
+config DM_MULTIPATH_QL
+	tristate "I/O Path Selector based on the number of in-flight I/Os"
+	depends on DM_MULTIPATH
+	---help---
+	  This path selector is a dynamic load balancer which selects
+	  the path with the least number of in-flight I/Os.
+
+	  If unsure, say N.
+
+config DM_MULTIPATH_ST
+	tristate "I/O Path Selector based on the service time"
+	depends on DM_MULTIPATH
+	---help---
+	  This path selector is a dynamic load balancer which selects
+	  the path expected to complete the incoming I/O in the shortest
+	  time.
+
+	  If unsure, say N.
+
 config DM_DELAY
-	tristate "I/O delaying target (EXPERIMENTAL)"
-	depends on BLK_DEV_DM && EXPERIMENTAL
+	tristate "I/O delaying target"
+	depends on BLK_DEV_DM
 	---help---
 	A target that delays reads and/or writes and can send
 	them to different devices.  Useful for testing.
@@ -270,9 +387,49 @@ config DM_DELAY
 	If unsure, say N.
 
 config DM_UEVENT
-	bool "DM uevents (EXPERIMENTAL)"
-	depends on BLK_DEV_DM && EXPERIMENTAL
+	bool "DM uevents"
+	depends on BLK_DEV_DM
 	---help---
 	Generate udev events for DM events.
 
+config DM_FLAKEY
+       tristate "Flakey target"
+       depends on BLK_DEV_DM
+       ---help---
+         A target that intermittently fails I/O for debugging purposes.
+
+config DM_VERITY
+	tristate "Verity target support"
+	depends on BLK_DEV_DM
+	select CRYPTO
+	select CRYPTO_HASH
+	select DM_BUFIO
+	---help---
+	  This device-mapper target creates a read-only device that
+	  transparently validates the data on one underlying device against
+	  a pre-generated tree of cryptographic checksums stored on a second
+	  device.
+
+	  You'll need to activate the digests you're going to use in the
+	  cryptoapi configuration.
+
+	  To compile this code as a module, choose M here: the module will
+	  be called dm-verity.
+
+	  If unsure, say N.
+
+config DM_SWITCH
+	tristate "Switch target support (EXPERIMENTAL)"
+	depends on BLK_DEV_DM
+	---help---
+	  This device-mapper target creates a device that supports an arbitrary
+	  mapping of fixed-size regions of I/O across a fixed set of paths.
+	  The path used for any specific region can be switched dynamically
+	  by sending the target a message.
+
+	  To compile this code as a module, choose M here: the module will
+	  be called dm-switch.
+
+	  If unsure, say N.
+
 endif # MD
diff --git a/drivers/md/Makefile b/drivers/md/Makefile
index f1ef33dfd8c..a2da532b1c2 100644
--- a/drivers/md/Makefile
+++ b/drivers/md/Makefile
@@ -2,19 +2,21 @@
 # Makefile for the kernel software RAID and LVM drivers.
 #
 
-dm-mod-objs	:= dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \
-		   dm-ioctl.o dm-io.o dm-kcopyd.o
-dm-multipath-objs := dm-path-selector.o dm-mpath.o
-dm-snapshot-objs := dm-snap.o dm-exception-store.o
-dm-mirror-objs	:= dm-raid1.o
-md-mod-objs     := md.o bitmap.o
-raid456-objs	:= raid5.o raid6algos.o raid6recov.o raid6tables.o \
-		   raid6int1.o raid6int2.o raid6int4.o \
-		   raid6int8.o raid6int16.o raid6int32.o \
-		   raid6altivec1.o raid6altivec2.o raid6altivec4.o \
-		   raid6altivec8.o \
-		   raid6mmx.o raid6sse1.o raid6sse2.o
-hostprogs-y	:= mktables
+dm-mod-y	+= dm.o dm-table.o dm-target.o dm-linear.o dm-stripe.o \
+		   dm-ioctl.o dm-io.o dm-kcopyd.o dm-sysfs.o dm-stats.o
+dm-multipath-y	+= dm-path-selector.o dm-mpath.o
+dm-snapshot-y	+= dm-snap.o dm-exception-store.o dm-snap-transient.o \
+		    dm-snap-persistent.o
+dm-mirror-y	+= dm-raid1.o
+dm-log-userspace-y \
+		+= dm-log-userspace-base.o dm-log-userspace-transfer.o
+dm-thin-pool-y	+= dm-thin.o dm-thin-metadata.o
+dm-cache-y	+= dm-cache-target.o dm-cache-metadata.o dm-cache-policy.o
+dm-cache-mq-y   += dm-cache-policy-mq.o
+dm-cache-cleaner-y += dm-cache-policy-cleaner.o
+dm-era-y	+= dm-era-target.o
+md-mod-y	+= md.o bitmap.o
+raid456-y	+= raid5.o
 
 # Note: link order is important.  All raid personalities
 # and must come before md.o, as they each initialise 
@@ -28,84 +30,32 @@ obj-$(CONFIG_MD_RAID10)		+= raid10.o
 obj-$(CONFIG_MD_RAID456)	+= raid456.o
 obj-$(CONFIG_MD_MULTIPATH)	+= multipath.o
 obj-$(CONFIG_MD_FAULTY)		+= faulty.o
+obj-$(CONFIG_BCACHE)		+= bcache/
 obj-$(CONFIG_BLK_DEV_MD)	+= md-mod.o
 obj-$(CONFIG_BLK_DEV_DM)	+= dm-mod.o
+obj-$(CONFIG_BLK_DEV_DM_BUILTIN) += dm-builtin.o
+obj-$(CONFIG_DM_BUFIO)		+= dm-bufio.o
+obj-$(CONFIG_DM_BIO_PRISON)	+= dm-bio-prison.o
 obj-$(CONFIG_DM_CRYPT)		+= dm-crypt.o
 obj-$(CONFIG_DM_DELAY)		+= dm-delay.o
+obj-$(CONFIG_DM_FLAKEY)		+= dm-flakey.o
 obj-$(CONFIG_DM_MULTIPATH)	+= dm-multipath.o dm-round-robin.o
+obj-$(CONFIG_DM_MULTIPATH_QL)	+= dm-queue-length.o
+obj-$(CONFIG_DM_MULTIPATH_ST)	+= dm-service-time.o
+obj-$(CONFIG_DM_SWITCH)		+= dm-switch.o
 obj-$(CONFIG_DM_SNAPSHOT)	+= dm-snapshot.o
-obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o dm-log.o
+obj-$(CONFIG_DM_PERSISTENT_DATA)	+= persistent-data/
+obj-$(CONFIG_DM_MIRROR)		+= dm-mirror.o dm-log.o dm-region-hash.o
+obj-$(CONFIG_DM_LOG_USERSPACE)	+= dm-log-userspace.o
 obj-$(CONFIG_DM_ZERO)		+= dm-zero.o
-
-quiet_cmd_unroll = UNROLL  $@
-      cmd_unroll = $(PERL) $(srctree)/$(src)/unroll.pl $(UNROLL) \
-                   < $< > $@ || ( rm -f $@ && exit 1 )
-
-ifeq ($(CONFIG_ALTIVEC),y)
-altivec_flags := -maltivec -mabi=altivec
-endif
+obj-$(CONFIG_DM_RAID)	+= dm-raid.o
+obj-$(CONFIG_DM_THIN_PROVISIONING)	+= dm-thin-pool.o
+obj-$(CONFIG_DM_VERITY)		+= dm-verity.o
+obj-$(CONFIG_DM_CACHE)		+= dm-cache.o
+obj-$(CONFIG_DM_CACHE_MQ)	+= dm-cache-mq.o
+obj-$(CONFIG_DM_CACHE_CLEANER)	+= dm-cache-cleaner.o
+obj-$(CONFIG_DM_ERA)		+= dm-era.o
 
 ifeq ($(CONFIG_DM_UEVENT),y)
 dm-mod-objs			+= dm-uevent.o
 endif
-
-targets += raid6int1.c
-$(obj)/raid6int1.c:   UNROLL := 1
-$(obj)/raid6int1.c:   $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-targets += raid6int2.c
-$(obj)/raid6int2.c:   UNROLL := 2
-$(obj)/raid6int2.c:   $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-targets += raid6int4.c
-$(obj)/raid6int4.c:   UNROLL := 4
-$(obj)/raid6int4.c:   $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-targets += raid6int8.c
-$(obj)/raid6int8.c:   UNROLL := 8
-$(obj)/raid6int8.c:   $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-targets += raid6int16.c
-$(obj)/raid6int16.c:  UNROLL := 16
-$(obj)/raid6int16.c:  $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-targets += raid6int32.c
-$(obj)/raid6int32.c:  UNROLL := 32
-$(obj)/raid6int32.c:  $(src)/raid6int.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-CFLAGS_raid6altivec1.o += $(altivec_flags)
-targets += raid6altivec1.c
-$(obj)/raid6altivec1.c:   UNROLL := 1
-$(obj)/raid6altivec1.c:   $(src)/raid6altivec.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-CFLAGS_raid6altivec2.o += $(altivec_flags)
-targets += raid6altivec2.c
-$(obj)/raid6altivec2.c:   UNROLL := 2
-$(obj)/raid6altivec2.c:   $(src)/raid6altivec.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-CFLAGS_raid6altivec4.o += $(altivec_flags)
-targets += raid6altivec4.c
-$(obj)/raid6altivec4.c:   UNROLL := 4
-$(obj)/raid6altivec4.c:   $(src)/raid6altivec.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-CFLAGS_raid6altivec8.o += $(altivec_flags)
-targets += raid6altivec8.c
-$(obj)/raid6altivec8.c:   UNROLL := 8
-$(obj)/raid6altivec8.c:   $(src)/raid6altivec.uc $(src)/unroll.pl FORCE
-	$(call if_changed,unroll)
-
-quiet_cmd_mktable = TABLE   $@
-      cmd_mktable = $(obj)/mktables > $@ || ( rm -f $@ && exit 1 )
-
-targets += raid6tables.c
-$(obj)/raid6tables.c: $(obj)/mktables FORCE
-	$(call if_changed,mktable)
diff --git a/drivers/md/bcache/Kconfig b/drivers/md/bcache/Kconfig
new file mode 100644
index 00000000000..4d200883c50
--- /dev/null
+++ b/drivers/md/bcache/Kconfig
@@ -0,0 +1,26 @@
+
+config BCACHE
+	tristate "Block device as cache"
+	---help---
+	Allows a block device to be used as cache for other devices; uses
+	a btree for indexing and the layout is optimized for SSDs.
+
+	See Documentation/bcache.txt for details.
+
+config BCACHE_DEBUG
+	bool "Bcache debugging"
+	depends on BCACHE
+	---help---
+	Don't select this option unless you're a developer
+
+	Enables extra debugging tools, allows expensive runtime checks to be
+	turned on.
+
+config BCACHE_CLOSURES_DEBUG
+	bool "Debug closures"
+	depends on BCACHE
+	select DEBUG_FS
+	---help---
+	Keeps all active closures in a linked list and provides a debugfs
+	interface to list them, which makes it possible to see asynchronous
+	operations that get stuck.
diff --git a/drivers/md/bcache/Makefile b/drivers/md/bcache/Makefile
new file mode 100644
index 00000000000..c488b846f83
--- /dev/null
+++ b/drivers/md/bcache/Makefile
@@ -0,0 +1,8 @@
+
+obj-$(CONFIG_BCACHE)	+= bcache.o
+
+bcache-y		:= alloc.o bset.o btree.o closure.o debug.o extents.o\
+	io.o journal.o movinggc.o request.o stats.o super.o sysfs.o trace.o\
+	util.o writeback.o
+
+CFLAGS_request.o	+= -Iblock
diff --git a/drivers/md/bcache/alloc.c b/drivers/md/bcache/alloc.c
new file mode 100644
index 00000000000..443d03fbac4
--- /dev/null
+++ b/drivers/md/bcache/alloc.c
@@ -0,0 +1,696 @@
+/*
+ * Primary bucket allocation code
+ *
+ * Copyright 2012 Google, Inc.
+ *
+ * Allocation in bcache is done in terms of buckets:
+ *
+ * Each bucket has associated an 8 bit gen; this gen corresponds to the gen in
+ * btree pointers - they must match for the pointer to be considered valid.
+ *
+ * Thus (assuming a bucket has no dirty data or metadata in it) we can reuse a
+ * bucket simply by incrementing its gen.
+ *
+ * The gens (along with the priorities; it's really the gens are important but
+ * the code is named as if it's the priorities) are written in an arbitrary list
+ * of buckets on disk, with a pointer to them in the journal header.
+ *
+ * When we invalidate a bucket, we have to write its new gen to disk and wait
+ * for that write to complete before we use it - otherwise after a crash we
+ * could have pointers that appeared to be good but pointed to data that had
+ * been overwritten.
+ *
+ * Since the gens and priorities are all stored contiguously on disk, we can
+ * batch this up: We fill up the free_inc list with freshly invalidated buckets,
+ * call prio_write(), and when prio_write() finishes we pull buckets off the
+ * free_inc list and optionally discard them.
+ *
+ * free_inc isn't the only freelist - if it was, we'd often to sleep while
+ * priorities and gens were being written before we could allocate. c->free is a
+ * smaller freelist, and buckets on that list are always ready to be used.
+ *
+ * If we've got discards enabled, that happens when a bucket moves from the
+ * free_inc list to the free list.
+ *
+ * There is another freelist, because sometimes we have buckets that we know
+ * have nothing pointing into them - these we can reuse without waiting for
+ * priorities to be rewritten. These come from freed btree nodes and buckets
+ * that garbage collection discovered no longer had valid keys pointing into
+ * them (because they were overwritten). That's the unused list - buckets on the
+ * unused list move to the free list, optionally being discarded in the process.
+ *
+ * It's also important to ensure that gens don't wrap around - with respect to
+ * either the oldest gen in the btree or the gen on disk. This is quite
+ * difficult to do in practice, but we explicitly guard against it anyways - if
+ * a bucket is in danger of wrapping around we simply skip invalidating it that
+ * time around, and we garbage collect or rewrite the priorities sooner than we
+ * would have otherwise.
+ *
+ * bch_bucket_alloc() allocates a single bucket from a specific cache.
+ *
+ * bch_bucket_alloc_set() allocates one or more buckets from different caches
+ * out of a cache set.
+ *
+ * free_some_buckets() drives all the processes described above. It's called
+ * from bch_bucket_alloc() and a few other places that need to make sure free
+ * buckets are ready.
+ *
+ * invalidate_buckets_(lru|fifo)() find buckets that are available to be
+ * invalidated, and then invalidate them and stick them on the free_inc list -
+ * in either lru or fifo order.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+
+#include <linux/blkdev.h>
+#include <linux/freezer.h>
+#include <linux/kthread.h>
+#include <linux/random.h>
+#include <trace/events/bcache.h>
+
+/* Bucket heap / gen */
+
+uint8_t bch_inc_gen(struct cache *ca, struct bucket *b)
+{
+	uint8_t ret = ++b->gen;
+
+	ca->set->need_gc = max(ca->set->need_gc, bucket_gc_gen(b));
+	WARN_ON_ONCE(ca->set->need_gc > BUCKET_GC_GEN_MAX);
+
+	return ret;
+}
+
+void bch_rescale_priorities(struct cache_set *c, int sectors)
+{
+	struct cache *ca;
+	struct bucket *b;
+	unsigned next = c->nbuckets * c->sb.bucket_size / 1024;
+	unsigned i;
+	int r;
+
+	atomic_sub(sectors, &c->rescale);
+
+	do {
+		r = atomic_read(&c->rescale);
+
+		if (r >= 0)
+			return;
+	} while (atomic_cmpxchg(&c->rescale, r, r + next) != r);
+
+	mutex_lock(&c->bucket_lock);
+
+	c->min_prio = USHRT_MAX;
+
+	for_each_cache(ca, c, i)
+		for_each_bucket(b, ca)
+			if (b->prio &&
+			    b->prio != BTREE_PRIO &&
+			    !atomic_read(&b->pin)) {
+				b->prio--;
+				c->min_prio = min(c->min_prio, b->prio);
+			}
+
+	mutex_unlock(&c->bucket_lock);
+}
+
+/*
+ * Background allocation thread: scans for buckets to be invalidated,
+ * invalidates them, rewrites prios/gens (marking them as invalidated on disk),
+ * then optionally issues discard commands to the newly free buckets, then puts
+ * them on the various freelists.
+ */
+
+static inline bool can_inc_bucket_gen(struct bucket *b)
+{
+	return bucket_gc_gen(b) < BUCKET_GC_GEN_MAX;
+}
+
+bool bch_can_invalidate_bucket(struct cache *ca, struct bucket *b)
+{
+	BUG_ON(!ca->set->gc_mark_valid);
+
+	return (!GC_MARK(b) ||
+		GC_MARK(b) == GC_MARK_RECLAIMABLE) &&
+		!atomic_read(&b->pin) &&
+		can_inc_bucket_gen(b);
+}
+
+void __bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
+{
+	lockdep_assert_held(&ca->set->bucket_lock);
+	BUG_ON(GC_MARK(b) && GC_MARK(b) != GC_MARK_RECLAIMABLE);
+
+	if (GC_SECTORS_USED(b))
+		trace_bcache_invalidate(ca, b - ca->buckets);
+
+	bch_inc_gen(ca, b);
+	b->prio = INITIAL_PRIO;
+	atomic_inc(&b->pin);
+}
+
+static void bch_invalidate_one_bucket(struct cache *ca, struct bucket *b)
+{
+	__bch_invalidate_one_bucket(ca, b);
+
+	fifo_push(&ca->free_inc, b - ca->buckets);
+}
+
+/*
+ * Determines what order we're going to reuse buckets, smallest bucket_prio()
+ * first: we also take into account the number of sectors of live data in that
+ * bucket, and in order for that multiply to make sense we have to scale bucket
+ *
+ * Thus, we scale the bucket priorities so that the bucket with the smallest
+ * prio is worth 1/8th of what INITIAL_PRIO is worth.
+ */
+
+#define bucket_prio(b)							\
+({									\
+	unsigned min_prio = (INITIAL_PRIO - ca->set->min_prio) / 8;	\
+									\
+	(b->prio - ca->set->min_prio + min_prio) * GC_SECTORS_USED(b);	\
+})
+
+#define bucket_max_cmp(l, r)	(bucket_prio(l) < bucket_prio(r))
+#define bucket_min_cmp(l, r)	(bucket_prio(l) > bucket_prio(r))
+
+static void invalidate_buckets_lru(struct cache *ca)
+{
+	struct bucket *b;
+	ssize_t i;
+
+	ca->heap.used = 0;
+
+	for_each_bucket(b, ca) {
+		if (!bch_can_invalidate_bucket(ca, b))
+			continue;
+
+		if (!heap_full(&ca->heap))
+			heap_add(&ca->heap, b, bucket_max_cmp);
+		else if (bucket_max_cmp(b, heap_peek(&ca->heap))) {
+			ca->heap.data[0] = b;
+			heap_sift(&ca->heap, 0, bucket_max_cmp);
+		}
+	}
+
+	for (i = ca->heap.used / 2 - 1; i >= 0; --i)
+		heap_sift(&ca->heap, i, bucket_min_cmp);
+
+	while (!fifo_full(&ca->free_inc)) {
+		if (!heap_pop(&ca->heap, b, bucket_min_cmp)) {
+			/*
+			 * We don't want to be calling invalidate_buckets()
+			 * multiple times when it can't do anything
+			 */
+			ca->invalidate_needs_gc = 1;
+			wake_up_gc(ca->set);
+			return;
+		}
+
+		bch_invalidate_one_bucket(ca, b);
+	}
+}
+
+static void invalidate_buckets_fifo(struct cache *ca)
+{
+	struct bucket *b;
+	size_t checked = 0;
+
+	while (!fifo_full(&ca->free_inc)) {
+		if (ca->fifo_last_bucket <  ca->sb.first_bucket ||
+		    ca->fifo_last_bucket >= ca->sb.nbuckets)
+			ca->fifo_last_bucket = ca->sb.first_bucket;
+
+		b = ca->buckets + ca->fifo_last_bucket++;
+
+		if (bch_can_invalidate_bucket(ca, b))
+			bch_invalidate_one_bucket(ca, b);
+
+		if (++checked >= ca->sb.nbuckets) {
+			ca->invalidate_needs_gc = 1;
+			wake_up_gc(ca->set);
+			return;
+		}
+	}
+}
+
+static void invalidate_buckets_random(struct cache *ca)
+{
+	struct bucket *b;
+	size_t checked = 0;
+
+	while (!fifo_full(&ca->free_inc)) {
+		size_t n;
+		get_random_bytes(&n, sizeof(n));
+
+		n %= (size_t) (ca->sb.nbuckets - ca->sb.first_bucket);
+		n += ca->sb.first_bucket;
+
+		b = ca->buckets + n;
+
+		if (bch_can_invalidate_bucket(ca, b))
+			bch_invalidate_one_bucket(ca, b);
+
+		if (++checked >= ca->sb.nbuckets / 2) {
+			ca->invalidate_needs_gc = 1;
+			wake_up_gc(ca->set);
+			return;
+		}
+	}
+}
+
+static void invalidate_buckets(struct cache *ca)
+{
+	BUG_ON(ca->invalidate_needs_gc);
+
+	switch (CACHE_REPLACEMENT(&ca->sb)) {
+	case CACHE_REPLACEMENT_LRU:
+		invalidate_buckets_lru(ca);
+		break;
+	case CACHE_REPLACEMENT_FIFO:
+		invalidate_buckets_fifo(ca);
+		break;
+	case CACHE_REPLACEMENT_RANDOM:
+		invalidate_buckets_random(ca);
+		break;
+	}
+}
+
+#define allocator_wait(ca, cond)					\
+do {									\
+	while (1) {							\
+		set_current_state(TASK_INTERRUPTIBLE);			\
+		if (cond)						\
+			break;						\
+									\
+		mutex_unlock(&(ca)->set->bucket_lock);			\
+		if (kthread_should_stop())				\
+			return 0;					\
+									\
+		try_to_freeze();					\
+		schedule();						\
+		mutex_lock(&(ca)->set->bucket_lock);			\
+	}								\
+	__set_current_state(TASK_RUNNING);				\
+} while (0)
+
+static int bch_allocator_push(struct cache *ca, long bucket)
+{
+	unsigned i;
+
+	/* Prios/gens are actually the most important reserve */
+	if (fifo_push(&ca->free[RESERVE_PRIO], bucket))
+		return true;
+
+	for (i = 0; i < RESERVE_NR; i++)
+		if (fifo_push(&ca->free[i], bucket))
+			return true;
+
+	return false;
+}
+
+static int bch_allocator_thread(void *arg)
+{
+	struct cache *ca = arg;
+
+	mutex_lock(&ca->set->bucket_lock);
+
+	while (1) {
+		/*
+		 * First, we pull buckets off of the unused and free_inc lists,
+		 * possibly issue discards to them, then we add the bucket to
+		 * the free list:
+		 */
+		while (!fifo_empty(&ca->free_inc)) {
+			long bucket;
+
+			fifo_pop(&ca->free_inc, bucket);
+
+			if (ca->discard) {
+				mutex_unlock(&ca->set->bucket_lock);
+				blkdev_issue_discard(ca->bdev,
+					bucket_to_sector(ca->set, bucket),
+					ca->sb.block_size, GFP_KERNEL, 0);
+				mutex_lock(&ca->set->bucket_lock);
+			}
+
+			allocator_wait(ca, bch_allocator_push(ca, bucket));
+			wake_up(&ca->set->btree_cache_wait);
+			wake_up(&ca->set->bucket_wait);
+		}
+
+		/*
+		 * We've run out of free buckets, we need to find some buckets
+		 * we can invalidate. First, invalidate them in memory and add
+		 * them to the free_inc list:
+		 */
+
+retry_invalidate:
+		allocator_wait(ca, ca->set->gc_mark_valid &&
+			       !ca->invalidate_needs_gc);
+		invalidate_buckets(ca);
+
+		/*
+		 * Now, we write their new gens to disk so we can start writing
+		 * new stuff to them:
+		 */
+		allocator_wait(ca, !atomic_read(&ca->set->prio_blocked));
+		if (CACHE_SYNC(&ca->set->sb)) {
+			/*
+			 * This could deadlock if an allocation with a btree
+			 * node locked ever blocked - having the btree node
+			 * locked would block garbage collection, but here we're
+			 * waiting on garbage collection before we invalidate
+			 * and free anything.
+			 *
+			 * But this should be safe since the btree code always
+			 * uses btree_check_reserve() before allocating now, and
+			 * if it fails it blocks without btree nodes locked.
+			 */
+			if (!fifo_full(&ca->free_inc))
+				goto retry_invalidate;
+
+			bch_prio_write(ca);
+		}
+	}
+}
+
+/* Allocation */
+
+long bch_bucket_alloc(struct cache *ca, unsigned reserve, bool wait)
+{
+	DEFINE_WAIT(w);
+	struct bucket *b;
+	long r;
+
+	/* fastpath */
+	if (fifo_pop(&ca->free[RESERVE_NONE], r) ||
+	    fifo_pop(&ca->free[reserve], r))
+		goto out;
+
+	if (!wait) {
+		trace_bcache_alloc_fail(ca, reserve);
+		return -1;
+	}
+
+	do {
+		prepare_to_wait(&ca->set->bucket_wait, &w,
+				TASK_UNINTERRUPTIBLE);
+
+		mutex_unlock(&ca->set->bucket_lock);
+		schedule();
+		mutex_lock(&ca->set->bucket_lock);
+	} while (!fifo_pop(&ca->free[RESERVE_NONE], r) &&
+		 !fifo_pop(&ca->free[reserve], r));
+
+	finish_wait(&ca->set->bucket_wait, &w);
+out:
+	wake_up_process(ca->alloc_thread);
+
+	trace_bcache_alloc(ca, reserve);
+
+	if (expensive_debug_checks(ca->set)) {
+		size_t iter;
+		long i;
+		unsigned j;
+
+		for (iter = 0; iter < prio_buckets(ca) * 2; iter++)
+			BUG_ON(ca->prio_buckets[iter] == (uint64_t) r);
+
+		for (j = 0; j < RESERVE_NR; j++)
+			fifo_for_each(i, &ca->free[j], iter)
+				BUG_ON(i == r);
+		fifo_for_each(i, &ca->free_inc, iter)
+			BUG_ON(i == r);
+	}
+
+	b = ca->buckets + r;
+
+	BUG_ON(atomic_read(&b->pin) != 1);
+
+	SET_GC_SECTORS_USED(b, ca->sb.bucket_size);
+
+	if (reserve <= RESERVE_PRIO) {
+		SET_GC_MARK(b, GC_MARK_METADATA);
+		SET_GC_MOVE(b, 0);
+		b->prio = BTREE_PRIO;
+	} else {
+		SET_GC_MARK(b, GC_MARK_RECLAIMABLE);
+		SET_GC_MOVE(b, 0);
+		b->prio = INITIAL_PRIO;
+	}
+
+	return r;
+}
+
+void __bch_bucket_free(struct cache *ca, struct bucket *b)
+{
+	SET_GC_MARK(b, 0);
+	SET_GC_SECTORS_USED(b, 0);
+}
+
+void bch_bucket_free(struct cache_set *c, struct bkey *k)
+{
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		__bch_bucket_free(PTR_CACHE(c, k, i),
+				  PTR_BUCKET(c, k, i));
+}
+
+int __bch_bucket_alloc_set(struct cache_set *c, unsigned reserve,
+			   struct bkey *k, int n, bool wait)
+{
+	int i;
+
+	lockdep_assert_held(&c->bucket_lock);
+	BUG_ON(!n || n > c->caches_loaded || n > 8);
+
+	bkey_init(k);
+
+	/* sort by free space/prio of oldest data in caches */
+
+	for (i = 0; i < n; i++) {
+		struct cache *ca = c->cache_by_alloc[i];
+		long b = bch_bucket_alloc(ca, reserve, wait);
+
+		if (b == -1)
+			goto err;
+
+		k->ptr[i] = PTR(ca->buckets[b].gen,
+				bucket_to_sector(c, b),
+				ca->sb.nr_this_dev);
+
+		SET_KEY_PTRS(k, i + 1);
+	}
+
+	return 0;
+err:
+	bch_bucket_free(c, k);
+	bkey_put(c, k);
+	return -1;
+}
+
+int bch_bucket_alloc_set(struct cache_set *c, unsigned reserve,
+			 struct bkey *k, int n, bool wait)
+{
+	int ret;
+	mutex_lock(&c->bucket_lock);
+	ret = __bch_bucket_alloc_set(c, reserve, k, n, wait);
+	mutex_unlock(&c->bucket_lock);
+	return ret;
+}
+
+/* Sector allocator */
+
+struct open_bucket {
+	struct list_head	list;
+	unsigned		last_write_point;
+	unsigned		sectors_free;
+	BKEY_PADDED(key);
+};
+
+/*
+ * We keep multiple buckets open for writes, and try to segregate different
+ * write streams for better cache utilization: first we look for a bucket where
+ * the last write to it was sequential with the current write, and failing that
+ * we look for a bucket that was last used by the same task.
+ *
+ * The ideas is if you've got multiple tasks pulling data into the cache at the
+ * same time, you'll get better cache utilization if you try to segregate their
+ * data and preserve locality.
+ *
+ * For example, say you've starting Firefox at the same time you're copying a
+ * bunch of files. Firefox will likely end up being fairly hot and stay in the
+ * cache awhile, but the data you copied might not be; if you wrote all that
+ * data to the same buckets it'd get invalidated at the same time.
+ *
+ * Both of those tasks will be doing fairly random IO so we can't rely on
+ * detecting sequential IO to segregate their data, but going off of the task
+ * should be a sane heuristic.
+ */
+static struct open_bucket *pick_data_bucket(struct cache_set *c,
+					    const struct bkey *search,
+					    unsigned write_point,
+					    struct bkey *alloc)
+{
+	struct open_bucket *ret, *ret_task = NULL;
+
+	list_for_each_entry_reverse(ret, &c->data_buckets, list)
+		if (!bkey_cmp(&ret->key, search))
+			goto found;
+		else if (ret->last_write_point == write_point)
+			ret_task = ret;
+
+	ret = ret_task ?: list_first_entry(&c->data_buckets,
+					   struct open_bucket, list);
+found:
+	if (!ret->sectors_free && KEY_PTRS(alloc)) {
+		ret->sectors_free = c->sb.bucket_size;
+		bkey_copy(&ret->key, alloc);
+		bkey_init(alloc);
+	}
+
+	if (!ret->sectors_free)
+		ret = NULL;
+
+	return ret;
+}
+
+/*
+ * Allocates some space in the cache to write to, and k to point to the newly
+ * allocated space, and updates KEY_SIZE(k) and KEY_OFFSET(k) (to point to the
+ * end of the newly allocated space).
+ *
+ * May allocate fewer sectors than @sectors, KEY_SIZE(k) indicates how many
+ * sectors were actually allocated.
+ *
+ * If s->writeback is true, will not fail.
+ */
+bool bch_alloc_sectors(struct cache_set *c, struct bkey *k, unsigned sectors,
+		       unsigned write_point, unsigned write_prio, bool wait)
+{
+	struct open_bucket *b;
+	BKEY_PADDED(key) alloc;
+	unsigned i;
+
+	/*
+	 * We might have to allocate a new bucket, which we can't do with a
+	 * spinlock held. So if we have to allocate, we drop the lock, allocate
+	 * and then retry. KEY_PTRS() indicates whether alloc points to
+	 * allocated bucket(s).
+	 */
+
+	bkey_init(&alloc.key);
+	spin_lock(&c->data_bucket_lock);
+
+	while (!(b = pick_data_bucket(c, k, write_point, &alloc.key))) {
+		unsigned watermark = write_prio
+			? RESERVE_MOVINGGC
+			: RESERVE_NONE;
+
+		spin_unlock(&c->data_bucket_lock);
+
+		if (bch_bucket_alloc_set(c, watermark, &alloc.key, 1, wait))
+			return false;
+
+		spin_lock(&c->data_bucket_lock);
+	}
+
+	/*
+	 * If we had to allocate, we might race and not need to allocate the
+	 * second time we call find_data_bucket(). If we allocated a bucket but
+	 * didn't use it, drop the refcount bch_bucket_alloc_set() took:
+	 */
+	if (KEY_PTRS(&alloc.key))
+		bkey_put(c, &alloc.key);
+
+	for (i = 0; i < KEY_PTRS(&b->key); i++)
+		EBUG_ON(ptr_stale(c, &b->key, i));
+
+	/* Set up the pointer to the space we're allocating: */
+
+	for (i = 0; i < KEY_PTRS(&b->key); i++)
+		k->ptr[i] = b->key.ptr[i];
+
+	sectors = min(sectors, b->sectors_free);
+
+	SET_KEY_OFFSET(k, KEY_OFFSET(k) + sectors);
+	SET_KEY_SIZE(k, sectors);
+	SET_KEY_PTRS(k, KEY_PTRS(&b->key));
+
+	/*
+	 * Move b to the end of the lru, and keep track of what this bucket was
+	 * last used for:
+	 */
+	list_move_tail(&b->list, &c->data_buckets);
+	bkey_copy_key(&b->key, k);
+	b->last_write_point = write_point;
+
+	b->sectors_free	-= sectors;
+
+	for (i = 0; i < KEY_PTRS(&b->key); i++) {
+		SET_PTR_OFFSET(&b->key, i, PTR_OFFSET(&b->key, i) + sectors);
+
+		atomic_long_add(sectors,
+				&PTR_CACHE(c, &b->key, i)->sectors_written);
+	}
+
+	if (b->sectors_free < c->sb.block_size)
+		b->sectors_free = 0;
+
+	/*
+	 * k takes refcounts on the buckets it points to until it's inserted
+	 * into the btree, but if we're done with this bucket we just transfer
+	 * get_data_bucket()'s refcount.
+	 */
+	if (b->sectors_free)
+		for (i = 0; i < KEY_PTRS(&b->key); i++)
+			atomic_inc(&PTR_BUCKET(c, &b->key, i)->pin);
+
+	spin_unlock(&c->data_bucket_lock);
+	return true;
+}
+
+/* Init */
+
+void bch_open_buckets_free(struct cache_set *c)
+{
+	struct open_bucket *b;
+
+	while (!list_empty(&c->data_buckets)) {
+		b = list_first_entry(&c->data_buckets,
+				     struct open_bucket, list);
+		list_del(&b->list);
+		kfree(b);
+	}
+}
+
+int bch_open_buckets_alloc(struct cache_set *c)
+{
+	int i;
+
+	spin_lock_init(&c->data_bucket_lock);
+
+	for (i = 0; i < 6; i++) {
+		struct open_bucket *b = kzalloc(sizeof(*b), GFP_KERNEL);
+		if (!b)
+			return -ENOMEM;
+
+		list_add(&b->list, &c->data_buckets);
+	}
+
+	return 0;
+}
+
+int bch_cache_allocator_start(struct cache *ca)
+{
+	struct task_struct *k = kthread_run(bch_allocator_thread,
+					    ca, "bcache_allocator");
+	if (IS_ERR(k))
+		return PTR_ERR(k);
+
+	ca->alloc_thread = k;
+	return 0;
+}
diff --git a/drivers/md/bcache/bcache.h b/drivers/md/bcache/bcache.h
new file mode 100644
index 00000000000..d2ebcf32309
--- /dev/null
+++ b/drivers/md/bcache/bcache.h
@@ -0,0 +1,942 @@
+#ifndef _BCACHE_H
+#define _BCACHE_H
+
+/*
+ * SOME HIGH LEVEL CODE DOCUMENTATION:
+ *
+ * Bcache mostly works with cache sets, cache devices, and backing devices.
+ *
+ * Support for multiple cache devices hasn't quite been finished off yet, but
+ * it's about 95% plumbed through. A cache set and its cache devices is sort of
+ * like a md raid array and its component devices. Most of the code doesn't care
+ * about individual cache devices, the main abstraction is the cache set.
+ *
+ * Multiple cache devices is intended to give us the ability to mirror dirty
+ * cached data and metadata, without mirroring clean cached data.
+ *
+ * Backing devices are different, in that they have a lifetime independent of a
+ * cache set. When you register a newly formatted backing device it'll come up
+ * in passthrough mode, and then you can attach and detach a backing device from
+ * a cache set at runtime - while it's mounted and in use. Detaching implicitly
+ * invalidates any cached data for that backing device.
+ *
+ * A cache set can have multiple (many) backing devices attached to it.
+ *
+ * There's also flash only volumes - this is the reason for the distinction
+ * between struct cached_dev and struct bcache_device. A flash only volume
+ * works much like a bcache device that has a backing device, except the
+ * "cached" data is always dirty. The end result is that we get thin
+ * provisioning with very little additional code.
+ *
+ * Flash only volumes work but they're not production ready because the moving
+ * garbage collector needs more work. More on that later.
+ *
+ * BUCKETS/ALLOCATION:
+ *
+ * Bcache is primarily designed for caching, which means that in normal
+ * operation all of our available space will be allocated. Thus, we need an
+ * efficient way of deleting things from the cache so we can write new things to
+ * it.
+ *
+ * To do this, we first divide the cache device up into buckets. A bucket is the
+ * unit of allocation; they're typically around 1 mb - anywhere from 128k to 2M+
+ * works efficiently.
+ *
+ * Each bucket has a 16 bit priority, and an 8 bit generation associated with
+ * it. The gens and priorities for all the buckets are stored contiguously and
+ * packed on disk (in a linked list of buckets - aside from the superblock, all
+ * of bcache's metadata is stored in buckets).
+ *
+ * The priority is used to implement an LRU. We reset a bucket's priority when
+ * we allocate it or on cache it, and every so often we decrement the priority
+ * of each bucket. It could be used to implement something more sophisticated,
+ * if anyone ever gets around to it.
+ *
+ * The generation is used for invalidating buckets. Each pointer also has an 8
+ * bit generation embedded in it; for a pointer to be considered valid, its gen
+ * must match the gen of the bucket it points into.  Thus, to reuse a bucket all
+ * we have to do is increment its gen (and write its new gen to disk; we batch
+ * this up).
+ *
+ * Bcache is entirely COW - we never write twice to a bucket, even buckets that
+ * contain metadata (including btree nodes).
+ *
+ * THE BTREE:
+ *
+ * Bcache is in large part design around the btree.
+ *
+ * At a high level, the btree is just an index of key -> ptr tuples.
+ *
+ * Keys represent extents, and thus have a size field. Keys also have a variable
+ * number of pointers attached to them (potentially zero, which is handy for
+ * invalidating the cache).
+ *
+ * The key itself is an inode:offset pair. The inode number corresponds to a
+ * backing device or a flash only volume. The offset is the ending offset of the
+ * extent within the inode - not the starting offset; this makes lookups
+ * slightly more convenient.
+ *
+ * Pointers contain the cache device id, the offset on that device, and an 8 bit
+ * generation number. More on the gen later.
+ *
+ * Index lookups are not fully abstracted - cache lookups in particular are
+ * still somewhat mixed in with the btree code, but things are headed in that
+ * direction.
+ *
+ * Updates are fairly well abstracted, though. There are two different ways of
+ * updating the btree; insert and replace.
+ *
+ * BTREE_INSERT will just take a list of keys and insert them into the btree -
+ * overwriting (possibly only partially) any extents they overlap with. This is
+ * used to update the index after a write.
+ *
+ * BTREE_REPLACE is really cmpxchg(); it inserts a key into the btree iff it is
+ * overwriting a key that matches another given key. This is used for inserting
+ * data into the cache after a cache miss, and for background writeback, and for
+ * the moving garbage collector.
+ *
+ * There is no "delete" operation; deleting things from the index is
+ * accomplished by either by invalidating pointers (by incrementing a bucket's
+ * gen) or by inserting a key with 0 pointers - which will overwrite anything
+ * previously present at that location in the index.
+ *
+ * This means that there are always stale/invalid keys in the btree. They're
+ * filtered out by the code that iterates through a btree node, and removed when
+ * a btree node is rewritten.
+ *
+ * BTREE NODES:
+ *
+ * Our unit of allocation is a bucket, and we we can't arbitrarily allocate and
+ * free smaller than a bucket - so, that's how big our btree nodes are.
+ *
+ * (If buckets are really big we'll only use part of the bucket for a btree node
+ * - no less than 1/4th - but a bucket still contains no more than a single
+ * btree node. I'd actually like to change this, but for now we rely on the
+ * bucket's gen for deleting btree nodes when we rewrite/split a node.)
+ *
+ * Anyways, btree nodes are big - big enough to be inefficient with a textbook
+ * btree implementation.
+ *
+ * The way this is solved is that btree nodes are internally log structured; we
+ * can append new keys to an existing btree node without rewriting it. This
+ * means each set of keys we write is sorted, but the node is not.
+ *
+ * We maintain this log structure in memory - keeping 1Mb of keys sorted would
+ * be expensive, and we have to distinguish between the keys we have written and
+ * the keys we haven't. So to do a lookup in a btree node, we have to search
+ * each sorted set. But we do merge written sets together lazily, so the cost of
+ * these extra searches is quite low (normally most of the keys in a btree node
+ * will be in one big set, and then there'll be one or two sets that are much
+ * smaller).
+ *
+ * This log structure makes bcache's btree more of a hybrid between a
+ * conventional btree and a compacting data structure, with some of the
+ * advantages of both.
+ *
+ * GARBAGE COLLECTION:
+ *
+ * We can't just invalidate any bucket - it might contain dirty data or
+ * metadata. If it once contained dirty data, other writes might overwrite it
+ * later, leaving no valid pointers into that bucket in the index.
+ *
+ * Thus, the primary purpose of garbage collection is to find buckets to reuse.
+ * It also counts how much valid data it each bucket currently contains, so that
+ * allocation can reuse buckets sooner when they've been mostly overwritten.
+ *
+ * It also does some things that are really internal to the btree
+ * implementation. If a btree node contains pointers that are stale by more than
+ * some threshold, it rewrites the btree node to avoid the bucket's generation
+ * wrapping around. It also merges adjacent btree nodes if they're empty enough.
+ *
+ * THE JOURNAL:
+ *
+ * Bcache's journal is not necessary for consistency; we always strictly
+ * order metadata writes so that the btree and everything else is consistent on
+ * disk in the event of an unclean shutdown, and in fact bcache had writeback
+ * caching (with recovery from unclean shutdown) before journalling was
+ * implemented.
+ *
+ * Rather, the journal is purely a performance optimization; we can't complete a
+ * write until we've updated the index on disk, otherwise the cache would be
+ * inconsistent in the event of an unclean shutdown. This means that without the
+ * journal, on random write workloads we constantly have to update all the leaf
+ * nodes in the btree, and those writes will be mostly empty (appending at most
+ * a few keys each) - highly inefficient in terms of amount of metadata writes,
+ * and it puts more strain on the various btree resorting/compacting code.
+ *
+ * The journal is just a log of keys we've inserted; on startup we just reinsert
+ * all the keys in the open journal entries. That means that when we're updating
+ * a node in the btree, we can wait until a 4k block of keys fills up before
+ * writing them out.
+ *
+ * For simplicity, we only journal updates to leaf nodes; updates to parent
+ * nodes are rare enough (since our leaf nodes are huge) that it wasn't worth
+ * the complexity to deal with journalling them (in particular, journal replay)
+ * - updates to non leaf nodes just happen synchronously (see btree_split()).
+ */
+
+#define pr_fmt(fmt) "bcache: %s() " fmt "\n", __func__
+
+#include <linux/bcache.h>
+#include <linux/bio.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/mutex.h>
+#include <linux/rbtree.h>
+#include <linux/rwsem.h>
+#include <linux/types.h>
+#include <linux/workqueue.h>
+
+#include "bset.h"
+#include "util.h"
+#include "closure.h"
+
+struct bucket {
+	atomic_t	pin;
+	uint16_t	prio;
+	uint8_t		gen;
+	uint8_t		last_gc; /* Most out of date gen in the btree */
+	uint16_t	gc_mark; /* Bitfield used by GC. See below for field */
+};
+
+/*
+ * I'd use bitfields for these, but I don't trust the compiler not to screw me
+ * as multiple threads touch struct bucket without locking
+ */
+
+BITMASK(GC_MARK,	 struct bucket, gc_mark, 0, 2);
+#define GC_MARK_RECLAIMABLE	1
+#define GC_MARK_DIRTY		2
+#define GC_MARK_METADATA	3
+#define GC_SECTORS_USED_SIZE	13
+#define MAX_GC_SECTORS_USED	(~(~0ULL << GC_SECTORS_USED_SIZE))
+BITMASK(GC_SECTORS_USED, struct bucket, gc_mark, 2, GC_SECTORS_USED_SIZE);
+BITMASK(GC_MOVE, struct bucket, gc_mark, 15, 1);
+
+#include "journal.h"
+#include "stats.h"
+struct search;
+struct btree;
+struct keybuf;
+
+struct keybuf_key {
+	struct rb_node		node;
+	BKEY_PADDED(key);
+	void			*private;
+};
+
+struct keybuf {
+	struct bkey		last_scanned;
+	spinlock_t		lock;
+
+	/*
+	 * Beginning and end of range in rb tree - so that we can skip taking
+	 * lock and checking the rb tree when we need to check for overlapping
+	 * keys.
+	 */
+	struct bkey		start;
+	struct bkey		end;
+
+	struct rb_root		keys;
+
+#define KEYBUF_NR		500
+	DECLARE_ARRAY_ALLOCATOR(struct keybuf_key, freelist, KEYBUF_NR);
+};
+
+struct bio_split_pool {
+	struct bio_set		*bio_split;
+	mempool_t		*bio_split_hook;
+};
+
+struct bio_split_hook {
+	struct closure		cl;
+	struct bio_split_pool	*p;
+	struct bio		*bio;
+	bio_end_io_t		*bi_end_io;
+	void			*bi_private;
+};
+
+struct bcache_device {
+	struct closure		cl;
+
+	struct kobject		kobj;
+
+	struct cache_set	*c;
+	unsigned		id;
+#define BCACHEDEVNAME_SIZE	12
+	char			name[BCACHEDEVNAME_SIZE];
+
+	struct gendisk		*disk;
+
+	unsigned long		flags;
+#define BCACHE_DEV_CLOSING	0
+#define BCACHE_DEV_DETACHING	1
+#define BCACHE_DEV_UNLINK_DONE	2
+
+	unsigned		nr_stripes;
+	unsigned		stripe_size;
+	atomic_t		*stripe_sectors_dirty;
+	unsigned long		*full_dirty_stripes;
+
+	unsigned long		sectors_dirty_last;
+	long			sectors_dirty_derivative;
+
+	struct bio_set		*bio_split;
+
+	unsigned		data_csum:1;
+
+	int (*cache_miss)(struct btree *, struct search *,
+			  struct bio *, unsigned);
+	int (*ioctl) (struct bcache_device *, fmode_t, unsigned, unsigned long);
+
+	struct bio_split_pool	bio_split_hook;
+};
+
+struct io {
+	/* Used to track sequential IO so it can be skipped */
+	struct hlist_node	hash;
+	struct list_head	lru;
+
+	unsigned long		jiffies;
+	unsigned		sequential;
+	sector_t		last;
+};
+
+struct cached_dev {
+	struct list_head	list;
+	struct bcache_device	disk;
+	struct block_device	*bdev;
+
+	struct cache_sb		sb;
+	struct bio		sb_bio;
+	struct bio_vec		sb_bv[1];
+	struct closure		sb_write;
+	struct semaphore	sb_write_mutex;
+
+	/* Refcount on the cache set. Always nonzero when we're caching. */
+	atomic_t		count;
+	struct work_struct	detach;
+
+	/*
+	 * Device might not be running if it's dirty and the cache set hasn't
+	 * showed up yet.
+	 */
+	atomic_t		running;
+
+	/*
+	 * Writes take a shared lock from start to finish; scanning for dirty
+	 * data to refill the rb tree requires an exclusive lock.
+	 */
+	struct rw_semaphore	writeback_lock;
+
+	/*
+	 * Nonzero, and writeback has a refcount (d->count), iff there is dirty
+	 * data in the cache. Protected by writeback_lock; must have an
+	 * shared lock to set and exclusive lock to clear.
+	 */
+	atomic_t		has_dirty;
+
+	struct bch_ratelimit	writeback_rate;
+	struct delayed_work	writeback_rate_update;
+
+	/*
+	 * Internal to the writeback code, so read_dirty() can keep track of
+	 * where it's at.
+	 */
+	sector_t		last_read;
+
+	/* Limit number of writeback bios in flight */
+	struct semaphore	in_flight;
+	struct task_struct	*writeback_thread;
+
+	struct keybuf		writeback_keys;
+
+	/* For tracking sequential IO */
+#define RECENT_IO_BITS	7
+#define RECENT_IO	(1 << RECENT_IO_BITS)
+	struct io		io[RECENT_IO];
+	struct hlist_head	io_hash[RECENT_IO + 1];
+	struct list_head	io_lru;
+	spinlock_t		io_lock;
+
+	struct cache_accounting	accounting;
+
+	/* The rest of this all shows up in sysfs */
+	unsigned		sequential_cutoff;
+	unsigned		readahead;
+
+	unsigned		verify:1;
+	unsigned		bypass_torture_test:1;
+
+	unsigned		partial_stripes_expensive:1;
+	unsigned		writeback_metadata:1;
+	unsigned		writeback_running:1;
+	unsigned char		writeback_percent;
+	unsigned		writeback_delay;
+
+	uint64_t		writeback_rate_target;
+	int64_t			writeback_rate_proportional;
+	int64_t			writeback_rate_derivative;
+	int64_t			writeback_rate_change;
+
+	unsigned		writeback_rate_update_seconds;
+	unsigned		writeback_rate_d_term;
+	unsigned		writeback_rate_p_term_inverse;
+};
+
+enum alloc_reserve {
+	RESERVE_BTREE,
+	RESERVE_PRIO,
+	RESERVE_MOVINGGC,
+	RESERVE_NONE,
+	RESERVE_NR,
+};
+
+struct cache {
+	struct cache_set	*set;
+	struct cache_sb		sb;
+	struct bio		sb_bio;
+	struct bio_vec		sb_bv[1];
+
+	struct kobject		kobj;
+	struct block_device	*bdev;
+
+	struct task_struct	*alloc_thread;
+
+	struct closure		prio;
+	struct prio_set		*disk_buckets;
+
+	/*
+	 * When allocating new buckets, prio_write() gets first dibs - since we
+	 * may not be allocate at all without writing priorities and gens.
+	 * prio_buckets[] contains the last buckets we wrote priorities to (so
+	 * gc can mark them as metadata), prio_next[] contains the buckets
+	 * allocated for the next prio write.
+	 */
+	uint64_t		*prio_buckets;
+	uint64_t		*prio_last_buckets;
+
+	/*
+	 * free: Buckets that are ready to be used
+	 *
+	 * free_inc: Incoming buckets - these are buckets that currently have
+	 * cached data in them, and we can't reuse them until after we write
+	 * their new gen to disk. After prio_write() finishes writing the new
+	 * gens/prios, they'll be moved to the free list (and possibly discarded
+	 * in the process)
+	 */
+	DECLARE_FIFO(long, free)[RESERVE_NR];
+	DECLARE_FIFO(long, free_inc);
+
+	size_t			fifo_last_bucket;
+
+	/* Allocation stuff: */
+	struct bucket		*buckets;
+
+	DECLARE_HEAP(struct bucket *, heap);
+
+	/*
+	 * If nonzero, we know we aren't going to find any buckets to invalidate
+	 * until a gc finishes - otherwise we could pointlessly burn a ton of
+	 * cpu
+	 */
+	unsigned		invalidate_needs_gc:1;
+
+	bool			discard; /* Get rid of? */
+
+	struct journal_device	journal;
+
+	/* The rest of this all shows up in sysfs */
+#define IO_ERROR_SHIFT		20
+	atomic_t		io_errors;
+	atomic_t		io_count;
+
+	atomic_long_t		meta_sectors_written;
+	atomic_long_t		btree_sectors_written;
+	atomic_long_t		sectors_written;
+
+	struct bio_split_pool	bio_split_hook;
+};
+
+struct gc_stat {
+	size_t			nodes;
+	size_t			key_bytes;
+
+	size_t			nkeys;
+	uint64_t		data;	/* sectors */
+	unsigned		in_use; /* percent */
+};
+
+/*
+ * Flag bits, for how the cache set is shutting down, and what phase it's at:
+ *
+ * CACHE_SET_UNREGISTERING means we're not just shutting down, we're detaching
+ * all the backing devices first (their cached data gets invalidated, and they
+ * won't automatically reattach).
+ *
+ * CACHE_SET_STOPPING always gets set first when we're closing down a cache set;
+ * we'll continue to run normally for awhile with CACHE_SET_STOPPING set (i.e.
+ * flushing dirty data).
+ */
+#define CACHE_SET_UNREGISTERING		0
+#define	CACHE_SET_STOPPING		1
+
+struct cache_set {
+	struct closure		cl;
+
+	struct list_head	list;
+	struct kobject		kobj;
+	struct kobject		internal;
+	struct dentry		*debug;
+	struct cache_accounting accounting;
+
+	unsigned long		flags;
+
+	struct cache_sb		sb;
+
+	struct cache		*cache[MAX_CACHES_PER_SET];
+	struct cache		*cache_by_alloc[MAX_CACHES_PER_SET];
+	int			caches_loaded;
+
+	struct bcache_device	**devices;
+	struct list_head	cached_devs;
+	uint64_t		cached_dev_sectors;
+	struct closure		caching;
+
+	struct closure		sb_write;
+	struct semaphore	sb_write_mutex;
+
+	mempool_t		*search;
+	mempool_t		*bio_meta;
+	struct bio_set		*bio_split;
+
+	/* For the btree cache */
+	struct shrinker		shrink;
+
+	/* For the btree cache and anything allocation related */
+	struct mutex		bucket_lock;
+
+	/* log2(bucket_size), in sectors */
+	unsigned short		bucket_bits;
+
+	/* log2(block_size), in sectors */
+	unsigned short		block_bits;
+
+	/*
+	 * Default number of pages for a new btree node - may be less than a
+	 * full bucket
+	 */
+	unsigned		btree_pages;
+
+	/*
+	 * Lists of struct btrees; lru is the list for structs that have memory
+	 * allocated for actual btree node, freed is for structs that do not.
+	 *
+	 * We never free a struct btree, except on shutdown - we just put it on
+	 * the btree_cache_freed list and reuse it later. This simplifies the
+	 * code, and it doesn't cost us much memory as the memory usage is
+	 * dominated by buffers that hold the actual btree node data and those
+	 * can be freed - and the number of struct btrees allocated is
+	 * effectively bounded.
+	 *
+	 * btree_cache_freeable effectively is a small cache - we use it because
+	 * high order page allocations can be rather expensive, and it's quite
+	 * common to delete and allocate btree nodes in quick succession. It
+	 * should never grow past ~2-3 nodes in practice.
+	 */
+	struct list_head	btree_cache;
+	struct list_head	btree_cache_freeable;
+	struct list_head	btree_cache_freed;
+
+	/* Number of elements in btree_cache + btree_cache_freeable lists */
+	unsigned		btree_cache_used;
+
+	/*
+	 * If we need to allocate memory for a new btree node and that
+	 * allocation fails, we can cannibalize another node in the btree cache
+	 * to satisfy the allocation - lock to guarantee only one thread does
+	 * this at a time:
+	 */
+	wait_queue_head_t	btree_cache_wait;
+	struct task_struct	*btree_cache_alloc_lock;
+
+	/*
+	 * When we free a btree node, we increment the gen of the bucket the
+	 * node is in - but we can't rewrite the prios and gens until we
+	 * finished whatever it is we were doing, otherwise after a crash the
+	 * btree node would be freed but for say a split, we might not have the
+	 * pointers to the new nodes inserted into the btree yet.
+	 *
+	 * This is a refcount that blocks prio_write() until the new keys are
+	 * written.
+	 */
+	atomic_t		prio_blocked;
+	wait_queue_head_t	bucket_wait;
+
+	/*
+	 * For any bio we don't skip we subtract the number of sectors from
+	 * rescale; when it hits 0 we rescale all the bucket priorities.
+	 */
+	atomic_t		rescale;
+	/*
+	 * When we invalidate buckets, we use both the priority and the amount
+	 * of good data to determine which buckets to reuse first - to weight
+	 * those together consistently we keep track of the smallest nonzero
+	 * priority of any bucket.
+	 */
+	uint16_t		min_prio;
+
+	/*
+	 * max(gen - last_gc) for all buckets. When it gets too big we have to gc
+	 * to keep gens from wrapping around.
+	 */
+	uint8_t			need_gc;
+	struct gc_stat		gc_stats;
+	size_t			nbuckets;
+
+	struct task_struct	*gc_thread;
+	/* Where in the btree gc currently is */
+	struct bkey		gc_done;
+
+	/*
+	 * The allocation code needs gc_mark in struct bucket to be correct, but
+	 * it's not while a gc is in progress. Protected by bucket_lock.
+	 */
+	int			gc_mark_valid;
+
+	/* Counts how many sectors bio_insert has added to the cache */
+	atomic_t		sectors_to_gc;
+
+	wait_queue_head_t	moving_gc_wait;
+	struct keybuf		moving_gc_keys;
+	/* Number of moving GC bios in flight */
+	struct semaphore	moving_in_flight;
+
+	struct workqueue_struct	*moving_gc_wq;
+
+	struct btree		*root;
+
+#ifdef CONFIG_BCACHE_DEBUG
+	struct btree		*verify_data;
+	struct bset		*verify_ondisk;
+	struct mutex		verify_lock;
+#endif
+
+	unsigned		nr_uuids;
+	struct uuid_entry	*uuids;
+	BKEY_PADDED(uuid_bucket);
+	struct closure		uuid_write;
+	struct semaphore	uuid_write_mutex;
+
+	/*
+	 * A btree node on disk could have too many bsets for an iterator to fit
+	 * on the stack - have to dynamically allocate them
+	 */
+	mempool_t		*fill_iter;
+
+	struct bset_sort_state	sort;
+
+	/* List of buckets we're currently writing data to */
+	struct list_head	data_buckets;
+	spinlock_t		data_bucket_lock;
+
+	struct journal		journal;
+
+#define CONGESTED_MAX		1024
+	unsigned		congested_last_us;
+	atomic_t		congested;
+
+	/* The rest of this all shows up in sysfs */
+	unsigned		congested_read_threshold_us;
+	unsigned		congested_write_threshold_us;
+
+	struct time_stats	btree_gc_time;
+	struct time_stats	btree_split_time;
+	struct time_stats	btree_read_time;
+
+	atomic_long_t		cache_read_races;
+	atomic_long_t		writeback_keys_done;
+	atomic_long_t		writeback_keys_failed;
+
+	enum			{
+		ON_ERROR_UNREGISTER,
+		ON_ERROR_PANIC,
+	}			on_error;
+	unsigned		error_limit;
+	unsigned		error_decay;
+
+	unsigned short		journal_delay_ms;
+	bool			expensive_debug_checks;
+	unsigned		verify:1;
+	unsigned		key_merging_disabled:1;
+	unsigned		gc_always_rewrite:1;
+	unsigned		shrinker_disabled:1;
+	unsigned		copy_gc_enabled:1;
+
+#define BUCKET_HASH_BITS	12
+	struct hlist_head	bucket_hash[1 << BUCKET_HASH_BITS];
+};
+
+struct bbio {
+	unsigned		submit_time_us;
+	union {
+		struct bkey	key;
+		uint64_t	_pad[3];
+		/*
+		 * We only need pad = 3 here because we only ever carry around a
+		 * single pointer - i.e. the pointer we're doing io to/from.
+		 */
+	};
+	struct bio		bio;
+};
+
+#define BTREE_PRIO		USHRT_MAX
+#define INITIAL_PRIO		32768U
+
+#define btree_bytes(c)		((c)->btree_pages * PAGE_SIZE)
+#define btree_blocks(b)							\
+	((unsigned) (KEY_SIZE(&b->key) >> (b)->c->block_bits))
+
+#define btree_default_blocks(c)						\
+	((unsigned) ((PAGE_SECTORS * (c)->btree_pages) >> (c)->block_bits))
+
+#define bucket_pages(c)		((c)->sb.bucket_size / PAGE_SECTORS)
+#define bucket_bytes(c)		((c)->sb.bucket_size << 9)
+#define block_bytes(c)		((c)->sb.block_size << 9)
+
+#define prios_per_bucket(c)				\
+	((bucket_bytes(c) - sizeof(struct prio_set)) /	\
+	 sizeof(struct bucket_disk))
+#define prio_buckets(c)					\
+	DIV_ROUND_UP((size_t) (c)->sb.nbuckets, prios_per_bucket(c))
+
+static inline size_t sector_to_bucket(struct cache_set *c, sector_t s)
+{
+	return s >> c->bucket_bits;
+}
+
+static inline sector_t bucket_to_sector(struct cache_set *c, size_t b)
+{
+	return ((sector_t) b) << c->bucket_bits;
+}
+
+static inline sector_t bucket_remainder(struct cache_set *c, sector_t s)
+{
+	return s & (c->sb.bucket_size - 1);
+}
+
+static inline struct cache *PTR_CACHE(struct cache_set *c,
+				      const struct bkey *k,
+				      unsigned ptr)
+{
+	return c->cache[PTR_DEV(k, ptr)];
+}
+
+static inline size_t PTR_BUCKET_NR(struct cache_set *c,
+				   const struct bkey *k,
+				   unsigned ptr)
+{
+	return sector_to_bucket(c, PTR_OFFSET(k, ptr));
+}
+
+static inline struct bucket *PTR_BUCKET(struct cache_set *c,
+					const struct bkey *k,
+					unsigned ptr)
+{
+	return PTR_CACHE(c, k, ptr)->buckets + PTR_BUCKET_NR(c, k, ptr);
+}
+
+static inline uint8_t gen_after(uint8_t a, uint8_t b)
+{
+	uint8_t r = a - b;
+	return r > 128U ? 0 : r;
+}
+
+static inline uint8_t ptr_stale(struct cache_set *c, const struct bkey *k,
+				unsigned i)
+{
+	return gen_after(PTR_BUCKET(c, k, i)->gen, PTR_GEN(k, i));
+}
+
+static inline bool ptr_available(struct cache_set *c, const struct bkey *k,
+				 unsigned i)
+{
+	return (PTR_DEV(k, i) < MAX_CACHES_PER_SET) && PTR_CACHE(c, k, i);
+}
+
+/* Btree key macros */
+
+/*
+ * This is used for various on disk data structures - cache_sb, prio_set, bset,
+ * jset: The checksum is _always_ the first 8 bytes of these structs
+ */
+#define csum_set(i)							\
+	bch_crc64(((void *) (i)) + sizeof(uint64_t),			\
+		  ((void *) bset_bkey_last(i)) -			\
+		  (((void *) (i)) + sizeof(uint64_t)))
+
+/* Error handling macros */
+
+#define btree_bug(b, ...)						\
+do {									\
+	if (bch_cache_set_error((b)->c, __VA_ARGS__))			\
+		dump_stack();						\
+} while (0)
+
+#define cache_bug(c, ...)						\
+do {									\
+	if (bch_cache_set_error(c, __VA_ARGS__))			\
+		dump_stack();						\
+} while (0)
+
+#define btree_bug_on(cond, b, ...)					\
+do {									\
+	if (cond)							\
+		btree_bug(b, __VA_ARGS__);				\
+} while (0)
+
+#define cache_bug_on(cond, c, ...)					\
+do {									\
+	if (cond)							\
+		cache_bug(c, __VA_ARGS__);				\
+} while (0)
+
+#define cache_set_err_on(cond, c, ...)					\
+do {									\
+	if (cond)							\
+		bch_cache_set_error(c, __VA_ARGS__);			\
+} while (0)
+
+/* Looping macros */
+
+#define for_each_cache(ca, cs, iter)					\
+	for (iter = 0; ca = cs->cache[iter], iter < (cs)->sb.nr_in_set; iter++)
+
+#define for_each_bucket(b, ca)						\
+	for (b = (ca)->buckets + (ca)->sb.first_bucket;			\
+	     b < (ca)->buckets + (ca)->sb.nbuckets; b++)
+
+static inline void cached_dev_put(struct cached_dev *dc)
+{
+	if (atomic_dec_and_test(&dc->count))
+		schedule_work(&dc->detach);
+}
+
+static inline bool cached_dev_get(struct cached_dev *dc)
+{
+	if (!atomic_inc_not_zero(&dc->count))
+		return false;
+
+	/* Paired with the mb in cached_dev_attach */
+	smp_mb__after_atomic();
+	return true;
+}
+
+/*
+ * bucket_gc_gen() returns the difference between the bucket's current gen and
+ * the oldest gen of any pointer into that bucket in the btree (last_gc).
+ */
+
+static inline uint8_t bucket_gc_gen(struct bucket *b)
+{
+	return b->gen - b->last_gc;
+}
+
+#define BUCKET_GC_GEN_MAX	96U
+
+#define kobj_attribute_write(n, fn)					\
+	static struct kobj_attribute ksysfs_##n = __ATTR(n, S_IWUSR, NULL, fn)
+
+#define kobj_attribute_rw(n, show, store)				\
+	static struct kobj_attribute ksysfs_##n =			\
+		__ATTR(n, S_IWUSR|S_IRUSR, show, store)
+
+static inline void wake_up_allocators(struct cache_set *c)
+{
+	struct cache *ca;
+	unsigned i;
+
+	for_each_cache(ca, c, i)
+		wake_up_process(ca->alloc_thread);
+}
+
+/* Forward declarations */
+
+void bch_count_io_errors(struct cache *, int, const char *);
+void bch_bbio_count_io_errors(struct cache_set *, struct bio *,
+			      int, const char *);
+void bch_bbio_endio(struct cache_set *, struct bio *, int, const char *);
+void bch_bbio_free(struct bio *, struct cache_set *);
+struct bio *bch_bbio_alloc(struct cache_set *);
+
+void bch_generic_make_request(struct bio *, struct bio_split_pool *);
+void __bch_submit_bbio(struct bio *, struct cache_set *);
+void bch_submit_bbio(struct bio *, struct cache_set *, struct bkey *, unsigned);
+
+uint8_t bch_inc_gen(struct cache *, struct bucket *);
+void bch_rescale_priorities(struct cache_set *, int);
+
+bool bch_can_invalidate_bucket(struct cache *, struct bucket *);
+void __bch_invalidate_one_bucket(struct cache *, struct bucket *);
+
+void __bch_bucket_free(struct cache *, struct bucket *);
+void bch_bucket_free(struct cache_set *, struct bkey *);
+
+long bch_bucket_alloc(struct cache *, unsigned, bool);
+int __bch_bucket_alloc_set(struct cache_set *, unsigned,
+			   struct bkey *, int, bool);
+int bch_bucket_alloc_set(struct cache_set *, unsigned,
+			 struct bkey *, int, bool);
+bool bch_alloc_sectors(struct cache_set *, struct bkey *, unsigned,
+		       unsigned, unsigned, bool);
+
+__printf(2, 3)
+bool bch_cache_set_error(struct cache_set *, const char *, ...);
+
+void bch_prio_write(struct cache *);
+void bch_write_bdev_super(struct cached_dev *, struct closure *);
+
+extern struct workqueue_struct *bcache_wq;
+extern const char * const bch_cache_modes[];
+extern struct mutex bch_register_lock;
+extern struct list_head bch_cache_sets;
+
+extern struct kobj_type bch_cached_dev_ktype;
+extern struct kobj_type bch_flash_dev_ktype;
+extern struct kobj_type bch_cache_set_ktype;
+extern struct kobj_type bch_cache_set_internal_ktype;
+extern struct kobj_type bch_cache_ktype;
+
+void bch_cached_dev_release(struct kobject *);
+void bch_flash_dev_release(struct kobject *);
+void bch_cache_set_release(struct kobject *);
+void bch_cache_release(struct kobject *);
+
+int bch_uuid_write(struct cache_set *);
+void bcache_write_super(struct cache_set *);
+
+int bch_flash_dev_create(struct cache_set *c, uint64_t size);
+
+int bch_cached_dev_attach(struct cached_dev *, struct cache_set *);
+void bch_cached_dev_detach(struct cached_dev *);
+void bch_cached_dev_run(struct cached_dev *);
+void bcache_device_stop(struct bcache_device *);
+
+void bch_cache_set_unregister(struct cache_set *);
+void bch_cache_set_stop(struct cache_set *);
+
+struct cache_set *bch_cache_set_alloc(struct cache_sb *);
+void bch_btree_cache_free(struct cache_set *);
+int bch_btree_cache_alloc(struct cache_set *);
+void bch_moving_init_cache_set(struct cache_set *);
+int bch_open_buckets_alloc(struct cache_set *);
+void bch_open_buckets_free(struct cache_set *);
+
+int bch_cache_allocator_start(struct cache *ca);
+
+void bch_debug_exit(void);
+int bch_debug_init(struct kobject *);
+void bch_request_exit(void);
+int bch_request_init(void);
+
+#endif /* _BCACHE_H */
diff --git a/drivers/md/bcache/bset.c b/drivers/md/bcache/bset.c
new file mode 100644
index 00000000000..54541641530
--- /dev/null
+++ b/drivers/md/bcache/bset.c
@@ -0,0 +1,1331 @@
+/*
+ * Code for working with individual keys, and sorted sets of keys with in a
+ * btree node
+ *
+ * Copyright 2012 Google, Inc.
+ */
+
+#define pr_fmt(fmt) "bcache: %s() " fmt "\n", __func__
+
+#include "util.h"
+#include "bset.h"
+
+#include <linux/console.h>
+#include <linux/random.h>
+#include <linux/prefetch.h>
+
+#ifdef CONFIG_BCACHE_DEBUG
+
+void bch_dump_bset(struct btree_keys *b, struct bset *i, unsigned set)
+{
+	struct bkey *k, *next;
+
+	for (k = i->start; k < bset_bkey_last(i); k = next) {
+		next = bkey_next(k);
+
+		printk(KERN_ERR "block %u key %u/%u: ", set,
+		       (unsigned) ((u64 *) k - i->d), i->keys);
+
+		if (b->ops->key_dump)
+			b->ops->key_dump(b, k);
+		else
+			printk("%llu:%llu\n", KEY_INODE(k), KEY_OFFSET(k));
+
+		if (next < bset_bkey_last(i) &&
+		    bkey_cmp(k, b->ops->is_extents ?
+			     &START_KEY(next) : next) > 0)
+			printk(KERN_ERR "Key skipped backwards\n");
+	}
+}
+
+void bch_dump_bucket(struct btree_keys *b)
+{
+	unsigned i;
+
+	console_lock();
+	for (i = 0; i <= b->nsets; i++)
+		bch_dump_bset(b, b->set[i].data,
+			      bset_sector_offset(b, b->set[i].data));
+	console_unlock();
+}
+
+int __bch_count_data(struct btree_keys *b)
+{
+	unsigned ret = 0;
+	struct btree_iter iter;
+	struct bkey *k;
+
+	if (b->ops->is_extents)
+		for_each_key(b, k, &iter)
+			ret += KEY_SIZE(k);
+	return ret;
+}
+
+void __bch_check_keys(struct btree_keys *b, const char *fmt, ...)
+{
+	va_list args;
+	struct bkey *k, *p = NULL;
+	struct btree_iter iter;
+	const char *err;
+
+	for_each_key(b, k, &iter) {
+		if (b->ops->is_extents) {
+			err = "Keys out of order";
+			if (p && bkey_cmp(&START_KEY(p), &START_KEY(k)) > 0)
+				goto bug;
+
+			if (bch_ptr_invalid(b, k))
+				continue;
+
+			err =  "Overlapping keys";
+			if (p && bkey_cmp(p, &START_KEY(k)) > 0)
+				goto bug;
+		} else {
+			if (bch_ptr_bad(b, k))
+				continue;
+
+			err = "Duplicate keys";
+			if (p && !bkey_cmp(p, k))
+				goto bug;
+		}
+		p = k;
+	}
+#if 0
+	err = "Key larger than btree node key";
+	if (p && bkey_cmp(p, &b->key) > 0)
+		goto bug;
+#endif
+	return;
+bug:
+	bch_dump_bucket(b);
+
+	va_start(args, fmt);
+	vprintk(fmt, args);
+	va_end(args);
+
+	panic("bch_check_keys error:  %s:\n", err);
+}
+
+static void bch_btree_iter_next_check(struct btree_iter *iter)
+{
+	struct bkey *k = iter->data->k, *next = bkey_next(k);
+
+	if (next < iter->data->end &&
+	    bkey_cmp(k, iter->b->ops->is_extents ?
+		     &START_KEY(next) : next) > 0) {
+		bch_dump_bucket(iter->b);
+		panic("Key skipped backwards\n");
+	}
+}
+
+#else
+
+static inline void bch_btree_iter_next_check(struct btree_iter *iter) {}
+
+#endif
+
+/* Keylists */
+
+int __bch_keylist_realloc(struct keylist *l, unsigned u64s)
+{
+	size_t oldsize = bch_keylist_nkeys(l);
+	size_t newsize = oldsize + u64s;
+	uint64_t *old_keys = l->keys_p == l->inline_keys ? NULL : l->keys_p;
+	uint64_t *new_keys;
+
+	newsize = roundup_pow_of_two(newsize);
+
+	if (newsize <= KEYLIST_INLINE ||
+	    roundup_pow_of_two(oldsize) == newsize)
+		return 0;
+
+	new_keys = krealloc(old_keys, sizeof(uint64_t) * newsize, GFP_NOIO);
+
+	if (!new_keys)
+		return -ENOMEM;
+
+	if (!old_keys)
+		memcpy(new_keys, l->inline_keys, sizeof(uint64_t) * oldsize);
+
+	l->keys_p = new_keys;
+	l->top_p = new_keys + oldsize;
+
+	return 0;
+}
+
+struct bkey *bch_keylist_pop(struct keylist *l)
+{
+	struct bkey *k = l->keys;
+
+	if (k == l->top)
+		return NULL;
+
+	while (bkey_next(k) != l->top)
+		k = bkey_next(k);
+
+	return l->top = k;
+}
+
+void bch_keylist_pop_front(struct keylist *l)
+{
+	l->top_p -= bkey_u64s(l->keys);
+
+	memmove(l->keys,
+		bkey_next(l->keys),
+		bch_keylist_bytes(l));
+}
+
+/* Key/pointer manipulation */
+
+void bch_bkey_copy_single_ptr(struct bkey *dest, const struct bkey *src,
+			      unsigned i)
+{
+	BUG_ON(i > KEY_PTRS(src));
+
+	/* Only copy the header, key, and one pointer. */
+	memcpy(dest, src, 2 * sizeof(uint64_t));
+	dest->ptr[0] = src->ptr[i];
+	SET_KEY_PTRS(dest, 1);
+	/* We didn't copy the checksum so clear that bit. */
+	SET_KEY_CSUM(dest, 0);
+}
+
+bool __bch_cut_front(const struct bkey *where, struct bkey *k)
+{
+	unsigned i, len = 0;
+
+	if (bkey_cmp(where, &START_KEY(k)) <= 0)
+		return false;
+
+	if (bkey_cmp(where, k) < 0)
+		len = KEY_OFFSET(k) - KEY_OFFSET(where);
+	else
+		bkey_copy_key(k, where);
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		SET_PTR_OFFSET(k, i, PTR_OFFSET(k, i) + KEY_SIZE(k) - len);
+
+	BUG_ON(len > KEY_SIZE(k));
+	SET_KEY_SIZE(k, len);
+	return true;
+}
+
+bool __bch_cut_back(const struct bkey *where, struct bkey *k)
+{
+	unsigned len = 0;
+
+	if (bkey_cmp(where, k) >= 0)
+		return false;
+
+	BUG_ON(KEY_INODE(where) != KEY_INODE(k));
+
+	if (bkey_cmp(where, &START_KEY(k)) > 0)
+		len = KEY_OFFSET(where) - KEY_START(k);
+
+	bkey_copy_key(k, where);
+
+	BUG_ON(len > KEY_SIZE(k));
+	SET_KEY_SIZE(k, len);
+	return true;
+}
+
+/* Auxiliary search trees */
+
+/* 32 bits total: */
+#define BKEY_MID_BITS		3
+#define BKEY_EXPONENT_BITS	7
+#define BKEY_MANTISSA_BITS	(32 - BKEY_MID_BITS - BKEY_EXPONENT_BITS)
+#define BKEY_MANTISSA_MASK	((1 << BKEY_MANTISSA_BITS) - 1)
+
+struct bkey_float {
+	unsigned	exponent:BKEY_EXPONENT_BITS;
+	unsigned	m:BKEY_MID_BITS;
+	unsigned	mantissa:BKEY_MANTISSA_BITS;
+} __packed;
+
+/*
+ * BSET_CACHELINE was originally intended to match the hardware cacheline size -
+ * it used to be 64, but I realized the lookup code would touch slightly less
+ * memory if it was 128.
+ *
+ * It definites the number of bytes (in struct bset) per struct bkey_float in
+ * the auxiliar search tree - when we're done searching the bset_float tree we
+ * have this many bytes left that we do a linear search over.
+ *
+ * Since (after level 5) every level of the bset_tree is on a new cacheline,
+ * we're touching one fewer cacheline in the bset tree in exchange for one more
+ * cacheline in the linear search - but the linear search might stop before it
+ * gets to the second cacheline.
+ */
+
+#define BSET_CACHELINE		128
+
+/* Space required for the btree node keys */
+static inline size_t btree_keys_bytes(struct btree_keys *b)
+{
+	return PAGE_SIZE << b->page_order;
+}
+
+static inline size_t btree_keys_cachelines(struct btree_keys *b)
+{
+	return btree_keys_bytes(b) / BSET_CACHELINE;
+}
+
+/* Space required for the auxiliary search trees */
+static inline size_t bset_tree_bytes(struct btree_keys *b)
+{
+	return btree_keys_cachelines(b) * sizeof(struct bkey_float);
+}
+
+/* Space required for the prev pointers */
+static inline size_t bset_prev_bytes(struct btree_keys *b)
+{
+	return btree_keys_cachelines(b) * sizeof(uint8_t);
+}
+
+/* Memory allocation */
+
+void bch_btree_keys_free(struct btree_keys *b)
+{
+	struct bset_tree *t = b->set;
+
+	if (bset_prev_bytes(b) < PAGE_SIZE)
+		kfree(t->prev);
+	else
+		free_pages((unsigned long) t->prev,
+			   get_order(bset_prev_bytes(b)));
+
+	if (bset_tree_bytes(b) < PAGE_SIZE)
+		kfree(t->tree);
+	else
+		free_pages((unsigned long) t->tree,
+			   get_order(bset_tree_bytes(b)));
+
+	free_pages((unsigned long) t->data, b->page_order);
+
+	t->prev = NULL;
+	t->tree = NULL;
+	t->data = NULL;
+}
+EXPORT_SYMBOL(bch_btree_keys_free);
+
+int bch_btree_keys_alloc(struct btree_keys *b, unsigned page_order, gfp_t gfp)
+{
+	struct bset_tree *t = b->set;
+
+	BUG_ON(t->data);
+
+	b->page_order = page_order;
+
+	t->data = (void *) __get_free_pages(gfp, b->page_order);
+	if (!t->data)
+		goto err;
+
+	t->tree = bset_tree_bytes(b) < PAGE_SIZE
+		? kmalloc(bset_tree_bytes(b), gfp)
+		: (void *) __get_free_pages(gfp, get_order(bset_tree_bytes(b)));
+	if (!t->tree)
+		goto err;
+
+	t->prev = bset_prev_bytes(b) < PAGE_SIZE
+		? kmalloc(bset_prev_bytes(b), gfp)
+		: (void *) __get_free_pages(gfp, get_order(bset_prev_bytes(b)));
+	if (!t->prev)
+		goto err;
+
+	return 0;
+err:
+	bch_btree_keys_free(b);
+	return -ENOMEM;
+}
+EXPORT_SYMBOL(bch_btree_keys_alloc);
+
+void bch_btree_keys_init(struct btree_keys *b, const struct btree_keys_ops *ops,
+			 bool *expensive_debug_checks)
+{
+	unsigned i;
+
+	b->ops = ops;
+	b->expensive_debug_checks = expensive_debug_checks;
+	b->nsets = 0;
+	b->last_set_unwritten = 0;
+
+	/* XXX: shouldn't be needed */
+	for (i = 0; i < MAX_BSETS; i++)
+		b->set[i].size = 0;
+	/*
+	 * Second loop starts at 1 because b->keys[0]->data is the memory we
+	 * allocated
+	 */
+	for (i = 1; i < MAX_BSETS; i++)
+		b->set[i].data = NULL;
+}
+EXPORT_SYMBOL(bch_btree_keys_init);
+
+/* Binary tree stuff for auxiliary search trees */
+
+static unsigned inorder_next(unsigned j, unsigned size)
+{
+	if (j * 2 + 1 < size) {
+		j = j * 2 + 1;
+
+		while (j * 2 < size)
+			j *= 2;
+	} else
+		j >>= ffz(j) + 1;
+
+	return j;
+}
+
+static unsigned inorder_prev(unsigned j, unsigned size)
+{
+	if (j * 2 < size) {
+		j = j * 2;
+
+		while (j * 2 + 1 < size)
+			j = j * 2 + 1;
+	} else
+		j >>= ffs(j);
+
+	return j;
+}
+
+/* I have no idea why this code works... and I'm the one who wrote it
+ *
+ * However, I do know what it does:
+ * Given a binary tree constructed in an array (i.e. how you normally implement
+ * a heap), it converts a node in the tree - referenced by array index - to the
+ * index it would have if you did an inorder traversal.
+ *
+ * Also tested for every j, size up to size somewhere around 6 million.
+ *
+ * The binary tree starts at array index 1, not 0
+ * extra is a function of size:
+ *   extra = (size - rounddown_pow_of_two(size - 1)) << 1;
+ */
+static unsigned __to_inorder(unsigned j, unsigned size, unsigned extra)
+{
+	unsigned b = fls(j);
+	unsigned shift = fls(size - 1) - b;
+
+	j  ^= 1U << (b - 1);
+	j <<= 1;
+	j  |= 1;
+	j <<= shift;
+
+	if (j > extra)
+		j -= (j - extra) >> 1;
+
+	return j;
+}
+
+static unsigned to_inorder(unsigned j, struct bset_tree *t)
+{
+	return __to_inorder(j, t->size, t->extra);
+}
+
+static unsigned __inorder_to_tree(unsigned j, unsigned size, unsigned extra)
+{
+	unsigned shift;
+
+	if (j > extra)
+		j += j - extra;
+
+	shift = ffs(j);
+
+	j >>= shift;
+	j  |= roundup_pow_of_two(size) >> shift;
+
+	return j;
+}
+
+static unsigned inorder_to_tree(unsigned j, struct bset_tree *t)
+{
+	return __inorder_to_tree(j, t->size, t->extra);
+}
+
+#if 0
+void inorder_test(void)
+{
+	unsigned long done = 0;
+	ktime_t start = ktime_get();
+
+	for (unsigned size = 2;
+	     size < 65536000;
+	     size++) {
+		unsigned extra = (size - rounddown_pow_of_two(size - 1)) << 1;
+		unsigned i = 1, j = rounddown_pow_of_two(size - 1);
+
+		if (!(size % 4096))
+			printk(KERN_NOTICE "loop %u, %llu per us\n", size,
+			       done / ktime_us_delta(ktime_get(), start));
+
+		while (1) {
+			if (__inorder_to_tree(i, size, extra) != j)
+				panic("size %10u j %10u i %10u", size, j, i);
+
+			if (__to_inorder(j, size, extra) != i)
+				panic("size %10u j %10u i %10u", size, j, i);
+
+			if (j == rounddown_pow_of_two(size) - 1)
+				break;
+
+			BUG_ON(inorder_prev(inorder_next(j, size), size) != j);
+
+			j = inorder_next(j, size);
+			i++;
+		}
+
+		done += size - 1;
+	}
+}
+#endif
+
+/*
+ * Cacheline/offset <-> bkey pointer arithmetic:
+ *
+ * t->tree is a binary search tree in an array; each node corresponds to a key
+ * in one cacheline in t->set (BSET_CACHELINE bytes).
+ *
+ * This means we don't have to store the full index of the key that a node in
+ * the binary tree points to; to_inorder() gives us the cacheline, and then
+ * bkey_float->m gives us the offset within that cacheline, in units of 8 bytes.
+ *
+ * cacheline_to_bkey() and friends abstract out all the pointer arithmetic to
+ * make this work.
+ *
+ * To construct the bfloat for an arbitrary key we need to know what the key
+ * immediately preceding it is: we have to check if the two keys differ in the
+ * bits we're going to store in bkey_float->mantissa. t->prev[j] stores the size
+ * of the previous key so we can walk backwards to it from t->tree[j]'s key.
+ */
+
+static struct bkey *cacheline_to_bkey(struct bset_tree *t, unsigned cacheline,
+				      unsigned offset)
+{
+	return ((void *) t->data) + cacheline * BSET_CACHELINE + offset * 8;
+}
+
+static unsigned bkey_to_cacheline(struct bset_tree *t, struct bkey *k)
+{
+	return ((void *) k - (void *) t->data) / BSET_CACHELINE;
+}
+
+static unsigned bkey_to_cacheline_offset(struct bset_tree *t,
+					 unsigned cacheline,
+					 struct bkey *k)
+{
+	return (u64 *) k - (u64 *) cacheline_to_bkey(t, cacheline, 0);
+}
+
+static struct bkey *tree_to_bkey(struct bset_tree *t, unsigned j)
+{
+	return cacheline_to_bkey(t, to_inorder(j, t), t->tree[j].m);
+}
+
+static struct bkey *tree_to_prev_bkey(struct bset_tree *t, unsigned j)
+{
+	return (void *) (((uint64_t *) tree_to_bkey(t, j)) - t->prev[j]);
+}
+
+/*
+ * For the write set - the one we're currently inserting keys into - we don't
+ * maintain a full search tree, we just keep a simple lookup table in t->prev.
+ */
+static struct bkey *table_to_bkey(struct bset_tree *t, unsigned cacheline)
+{
+	return cacheline_to_bkey(t, cacheline, t->prev[cacheline]);
+}
+
+static inline uint64_t shrd128(uint64_t high, uint64_t low, uint8_t shift)
+{
+	low >>= shift;
+	low  |= (high << 1) << (63U - shift);
+	return low;
+}
+
+static inline unsigned bfloat_mantissa(const struct bkey *k,
+				       struct bkey_float *f)
+{
+	const uint64_t *p = &k->low - (f->exponent >> 6);
+	return shrd128(p[-1], p[0], f->exponent & 63) & BKEY_MANTISSA_MASK;
+}
+
+static void make_bfloat(struct bset_tree *t, unsigned j)
+{
+	struct bkey_float *f = &t->tree[j];
+	struct bkey *m = tree_to_bkey(t, j);
+	struct bkey *p = tree_to_prev_bkey(t, j);
+
+	struct bkey *l = is_power_of_2(j)
+		? t->data->start
+		: tree_to_prev_bkey(t, j >> ffs(j));
+
+	struct bkey *r = is_power_of_2(j + 1)
+		? bset_bkey_idx(t->data, t->data->keys - bkey_u64s(&t->end))
+		: tree_to_bkey(t, j >> (ffz(j) + 1));
+
+	BUG_ON(m < l || m > r);
+	BUG_ON(bkey_next(p) != m);
+
+	if (KEY_INODE(l) != KEY_INODE(r))
+		f->exponent = fls64(KEY_INODE(r) ^ KEY_INODE(l)) + 64;
+	else
+		f->exponent = fls64(r->low ^ l->low);
+
+	f->exponent = max_t(int, f->exponent - BKEY_MANTISSA_BITS, 0);
+
+	/*
+	 * Setting f->exponent = 127 flags this node as failed, and causes the
+	 * lookup code to fall back to comparing against the original key.
+	 */
+
+	if (bfloat_mantissa(m, f) != bfloat_mantissa(p, f))
+		f->mantissa = bfloat_mantissa(m, f) - 1;
+	else
+		f->exponent = 127;
+}
+
+static void bset_alloc_tree(struct btree_keys *b, struct bset_tree *t)
+{
+	if (t != b->set) {
+		unsigned j = roundup(t[-1].size,
+				     64 / sizeof(struct bkey_float));
+
+		t->tree = t[-1].tree + j;
+		t->prev = t[-1].prev + j;
+	}
+
+	while (t < b->set + MAX_BSETS)
+		t++->size = 0;
+}
+
+static void bch_bset_build_unwritten_tree(struct btree_keys *b)
+{
+	struct bset_tree *t = bset_tree_last(b);
+
+	BUG_ON(b->last_set_unwritten);
+	b->last_set_unwritten = 1;
+
+	bset_alloc_tree(b, t);
+
+	if (t->tree != b->set->tree + btree_keys_cachelines(b)) {
+		t->prev[0] = bkey_to_cacheline_offset(t, 0, t->data->start);
+		t->size = 1;
+	}
+}
+
+void bch_bset_init_next(struct btree_keys *b, struct bset *i, uint64_t magic)
+{
+	if (i != b->set->data) {
+		b->set[++b->nsets].data = i;
+		i->seq = b->set->data->seq;
+	} else
+		get_random_bytes(&i->seq, sizeof(uint64_t));
+
+	i->magic	= magic;
+	i->version	= 0;
+	i->keys		= 0;
+
+	bch_bset_build_unwritten_tree(b);
+}
+EXPORT_SYMBOL(bch_bset_init_next);
+
+void bch_bset_build_written_tree(struct btree_keys *b)
+{
+	struct bset_tree *t = bset_tree_last(b);
+	struct bkey *prev = NULL, *k = t->data->start;
+	unsigned j, cacheline = 1;
+
+	b->last_set_unwritten = 0;
+
+	bset_alloc_tree(b, t);
+
+	t->size = min_t(unsigned,
+			bkey_to_cacheline(t, bset_bkey_last(t->data)),
+			b->set->tree + btree_keys_cachelines(b) - t->tree);
+
+	if (t->size < 2) {
+		t->size = 0;
+		return;
+	}
+
+	t->extra = (t->size - rounddown_pow_of_two(t->size - 1)) << 1;
+
+	/* First we figure out where the first key in each cacheline is */
+	for (j = inorder_next(0, t->size);
+	     j;
+	     j = inorder_next(j, t->size)) {
+		while (bkey_to_cacheline(t, k) < cacheline)
+			prev = k, k = bkey_next(k);
+
+		t->prev[j] = bkey_u64s(prev);
+		t->tree[j].m = bkey_to_cacheline_offset(t, cacheline++, k);
+	}
+
+	while (bkey_next(k) != bset_bkey_last(t->data))
+		k = bkey_next(k);
+
+	t->end = *k;
+
+	/* Then we build the tree */
+	for (j = inorder_next(0, t->size);
+	     j;
+	     j = inorder_next(j, t->size))
+		make_bfloat(t, j);
+}
+EXPORT_SYMBOL(bch_bset_build_written_tree);
+
+/* Insert */
+
+void bch_bset_fix_invalidated_key(struct btree_keys *b, struct bkey *k)
+{
+	struct bset_tree *t;
+	unsigned inorder, j = 1;
+
+	for (t = b->set; t <= bset_tree_last(b); t++)
+		if (k < bset_bkey_last(t->data))
+			goto found_set;
+
+	BUG();
+found_set:
+	if (!t->size || !bset_written(b, t))
+		return;
+
+	inorder = bkey_to_cacheline(t, k);
+
+	if (k == t->data->start)
+		goto fix_left;
+
+	if (bkey_next(k) == bset_bkey_last(t->data)) {
+		t->end = *k;
+		goto fix_right;
+	}
+
+	j = inorder_to_tree(inorder, t);
+
+	if (j &&
+	    j < t->size &&
+	    k == tree_to_bkey(t, j))
+fix_left:	do {
+			make_bfloat(t, j);
+			j = j * 2;
+		} while (j < t->size);
+
+	j = inorder_to_tree(inorder + 1, t);
+
+	if (j &&
+	    j < t->size &&
+	    k == tree_to_prev_bkey(t, j))
+fix_right:	do {
+			make_bfloat(t, j);
+			j = j * 2 + 1;
+		} while (j < t->size);
+}
+EXPORT_SYMBOL(bch_bset_fix_invalidated_key);
+
+static void bch_bset_fix_lookup_table(struct btree_keys *b,
+				      struct bset_tree *t,
+				      struct bkey *k)
+{
+	unsigned shift = bkey_u64s(k);
+	unsigned j = bkey_to_cacheline(t, k);
+
+	/* We're getting called from btree_split() or btree_gc, just bail out */
+	if (!t->size)
+		return;
+
+	/* k is the key we just inserted; we need to find the entry in the
+	 * lookup table for the first key that is strictly greater than k:
+	 * it's either k's cacheline or the next one
+	 */
+	while (j < t->size &&
+	       table_to_bkey(t, j) <= k)
+		j++;
+
+	/* Adjust all the lookup table entries, and find a new key for any that
+	 * have gotten too big
+	 */
+	for (; j < t->size; j++) {
+		t->prev[j] += shift;
+
+		if (t->prev[j] > 7) {
+			k = table_to_bkey(t, j - 1);
+
+			while (k < cacheline_to_bkey(t, j, 0))
+				k = bkey_next(k);
+
+			t->prev[j] = bkey_to_cacheline_offset(t, j, k);
+		}
+	}
+
+	if (t->size == b->set->tree + btree_keys_cachelines(b) - t->tree)
+		return;
+
+	/* Possibly add a new entry to the end of the lookup table */
+
+	for (k = table_to_bkey(t, t->size - 1);
+	     k != bset_bkey_last(t->data);
+	     k = bkey_next(k))
+		if (t->size == bkey_to_cacheline(t, k)) {
+			t->prev[t->size] = bkey_to_cacheline_offset(t, t->size, k);
+			t->size++;
+		}
+}
+
+/*
+ * Tries to merge l and r: l should be lower than r
+ * Returns true if we were able to merge. If we did merge, l will be the merged
+ * key, r will be untouched.
+ */
+bool bch_bkey_try_merge(struct btree_keys *b, struct bkey *l, struct bkey *r)
+{
+	if (!b->ops->key_merge)
+		return false;
+
+	/*
+	 * Generic header checks
+	 * Assumes left and right are in order
+	 * Left and right must be exactly aligned
+	 */
+	if (!bch_bkey_equal_header(l, r) ||
+	     bkey_cmp(l, &START_KEY(r)))
+		return false;
+
+	return b->ops->key_merge(b, l, r);
+}
+EXPORT_SYMBOL(bch_bkey_try_merge);
+
+void bch_bset_insert(struct btree_keys *b, struct bkey *where,
+		     struct bkey *insert)
+{
+	struct bset_tree *t = bset_tree_last(b);
+
+	BUG_ON(!b->last_set_unwritten);
+	BUG_ON(bset_byte_offset(b, t->data) +
+	       __set_bytes(t->data, t->data->keys + bkey_u64s(insert)) >
+	       PAGE_SIZE << b->page_order);
+
+	memmove((uint64_t *) where + bkey_u64s(insert),
+		where,
+		(void *) bset_bkey_last(t->data) - (void *) where);
+
+	t->data->keys += bkey_u64s(insert);
+	bkey_copy(where, insert);
+	bch_bset_fix_lookup_table(b, t, where);
+}
+EXPORT_SYMBOL(bch_bset_insert);
+
+unsigned bch_btree_insert_key(struct btree_keys *b, struct bkey *k,
+			      struct bkey *replace_key)
+{
+	unsigned status = BTREE_INSERT_STATUS_NO_INSERT;
+	struct bset *i = bset_tree_last(b)->data;
+	struct bkey *m, *prev = NULL;
+	struct btree_iter iter;
+
+	BUG_ON(b->ops->is_extents && !KEY_SIZE(k));
+
+	m = bch_btree_iter_init(b, &iter, b->ops->is_extents
+				? PRECEDING_KEY(&START_KEY(k))
+				: PRECEDING_KEY(k));
+
+	if (b->ops->insert_fixup(b, k, &iter, replace_key))
+		return status;
+
+	status = BTREE_INSERT_STATUS_INSERT;
+
+	while (m != bset_bkey_last(i) &&
+	       bkey_cmp(k, b->ops->is_extents ? &START_KEY(m) : m) > 0)
+		prev = m, m = bkey_next(m);
+
+	/* prev is in the tree, if we merge we're done */
+	status = BTREE_INSERT_STATUS_BACK_MERGE;
+	if (prev &&
+	    bch_bkey_try_merge(b, prev, k))
+		goto merged;
+#if 0
+	status = BTREE_INSERT_STATUS_OVERWROTE;
+	if (m != bset_bkey_last(i) &&
+	    KEY_PTRS(m) == KEY_PTRS(k) && !KEY_SIZE(m))
+		goto copy;
+#endif
+	status = BTREE_INSERT_STATUS_FRONT_MERGE;
+	if (m != bset_bkey_last(i) &&
+	    bch_bkey_try_merge(b, k, m))
+		goto copy;
+
+	bch_bset_insert(b, m, k);
+copy:	bkey_copy(m, k);
+merged:
+	return status;
+}
+EXPORT_SYMBOL(bch_btree_insert_key);
+
+/* Lookup */
+
+struct bset_search_iter {
+	struct bkey *l, *r;
+};
+
+static struct bset_search_iter bset_search_write_set(struct bset_tree *t,
+						     const struct bkey *search)
+{
+	unsigned li = 0, ri = t->size;
+
+	while (li + 1 != ri) {
+		unsigned m = (li + ri) >> 1;
+
+		if (bkey_cmp(table_to_bkey(t, m), search) > 0)
+			ri = m;
+		else
+			li = m;
+	}
+
+	return (struct bset_search_iter) {
+		table_to_bkey(t, li),
+		ri < t->size ? table_to_bkey(t, ri) : bset_bkey_last(t->data)
+	};
+}
+
+static struct bset_search_iter bset_search_tree(struct bset_tree *t,
+						const struct bkey *search)
+{
+	struct bkey *l, *r;
+	struct bkey_float *f;
+	unsigned inorder, j, n = 1;
+
+	do {
+		unsigned p = n << 4;
+		p &= ((int) (p - t->size)) >> 31;
+
+		prefetch(&t->tree[p]);
+
+		j = n;
+		f = &t->tree[j];
+
+		/*
+		 * n = (f->mantissa > bfloat_mantissa())
+		 *	? j * 2
+		 *	: j * 2 + 1;
+		 *
+		 * We need to subtract 1 from f->mantissa for the sign bit trick
+		 * to work  - that's done in make_bfloat()
+		 */
+		if (likely(f->exponent != 127))
+			n = j * 2 + (((unsigned)
+				      (f->mantissa -
+				       bfloat_mantissa(search, f))) >> 31);
+		else
+			n = (bkey_cmp(tree_to_bkey(t, j), search) > 0)
+				? j * 2
+				: j * 2 + 1;
+	} while (n < t->size);
+
+	inorder = to_inorder(j, t);
+
+	/*
+	 * n would have been the node we recursed to - the low bit tells us if
+	 * we recursed left or recursed right.
+	 */
+	if (n & 1) {
+		l = cacheline_to_bkey(t, inorder, f->m);
+
+		if (++inorder != t->size) {
+			f = &t->tree[inorder_next(j, t->size)];
+			r = cacheline_to_bkey(t, inorder, f->m);
+		} else
+			r = bset_bkey_last(t->data);
+	} else {
+		r = cacheline_to_bkey(t, inorder, f->m);
+
+		if (--inorder) {
+			f = &t->tree[inorder_prev(j, t->size)];
+			l = cacheline_to_bkey(t, inorder, f->m);
+		} else
+			l = t->data->start;
+	}
+
+	return (struct bset_search_iter) {l, r};
+}
+
+struct bkey *__bch_bset_search(struct btree_keys *b, struct bset_tree *t,
+			       const struct bkey *search)
+{
+	struct bset_search_iter i;
+
+	/*
+	 * First, we search for a cacheline, then lastly we do a linear search
+	 * within that cacheline.
+	 *
+	 * To search for the cacheline, there's three different possibilities:
+	 *  * The set is too small to have a search tree, so we just do a linear
+	 *    search over the whole set.
+	 *  * The set is the one we're currently inserting into; keeping a full
+	 *    auxiliary search tree up to date would be too expensive, so we
+	 *    use a much simpler lookup table to do a binary search -
+	 *    bset_search_write_set().
+	 *  * Or we use the auxiliary search tree we constructed earlier -
+	 *    bset_search_tree()
+	 */
+
+	if (unlikely(!t->size)) {
+		i.l = t->data->start;
+		i.r = bset_bkey_last(t->data);
+	} else if (bset_written(b, t)) {
+		/*
+		 * Each node in the auxiliary search tree covers a certain range
+		 * of bits, and keys above and below the set it covers might
+		 * differ outside those bits - so we have to special case the
+		 * start and end - handle that here:
+		 */
+
+		if (unlikely(bkey_cmp(search, &t->end) >= 0))
+			return bset_bkey_last(t->data);
+
+		if (unlikely(bkey_cmp(search, t->data->start) < 0))
+			return t->data->start;
+
+		i = bset_search_tree(t, search);
+	} else {
+		BUG_ON(!b->nsets &&
+		       t->size < bkey_to_cacheline(t, bset_bkey_last(t->data)));
+
+		i = bset_search_write_set(t, search);
+	}
+
+	if (btree_keys_expensive_checks(b)) {
+		BUG_ON(bset_written(b, t) &&
+		       i.l != t->data->start &&
+		       bkey_cmp(tree_to_prev_bkey(t,
+			  inorder_to_tree(bkey_to_cacheline(t, i.l), t)),
+				search) > 0);
+
+		BUG_ON(i.r != bset_bkey_last(t->data) &&
+		       bkey_cmp(i.r, search) <= 0);
+	}
+
+	while (likely(i.l != i.r) &&
+	       bkey_cmp(i.l, search) <= 0)
+		i.l = bkey_next(i.l);
+
+	return i.l;
+}
+EXPORT_SYMBOL(__bch_bset_search);
+
+/* Btree iterator */
+
+typedef bool (btree_iter_cmp_fn)(struct btree_iter_set,
+				 struct btree_iter_set);
+
+static inline bool btree_iter_cmp(struct btree_iter_set l,
+				  struct btree_iter_set r)
+{
+	return bkey_cmp(l.k, r.k) > 0;
+}
+
+static inline bool btree_iter_end(struct btree_iter *iter)
+{
+	return !iter->used;
+}
+
+void bch_btree_iter_push(struct btree_iter *iter, struct bkey *k,
+			 struct bkey *end)
+{
+	if (k != end)
+		BUG_ON(!heap_add(iter,
+				 ((struct btree_iter_set) { k, end }),
+				 btree_iter_cmp));
+}
+
+static struct bkey *__bch_btree_iter_init(struct btree_keys *b,
+					  struct btree_iter *iter,
+					  struct bkey *search,
+					  struct bset_tree *start)
+{
+	struct bkey *ret = NULL;
+	iter->size = ARRAY_SIZE(iter->data);
+	iter->used = 0;
+
+#ifdef CONFIG_BCACHE_DEBUG
+	iter->b = b;
+#endif
+
+	for (; start <= bset_tree_last(b); start++) {
+		ret = bch_bset_search(b, start, search);
+		bch_btree_iter_push(iter, ret, bset_bkey_last(start->data));
+	}
+
+	return ret;
+}
+
+struct bkey *bch_btree_iter_init(struct btree_keys *b,
+				 struct btree_iter *iter,
+				 struct bkey *search)
+{
+	return __bch_btree_iter_init(b, iter, search, b->set);
+}
+EXPORT_SYMBOL(bch_btree_iter_init);
+
+static inline struct bkey *__bch_btree_iter_next(struct btree_iter *iter,
+						 btree_iter_cmp_fn *cmp)
+{
+	struct btree_iter_set unused;
+	struct bkey *ret = NULL;
+
+	if (!btree_iter_end(iter)) {
+		bch_btree_iter_next_check(iter);
+
+		ret = iter->data->k;
+		iter->data->k = bkey_next(iter->data->k);
+
+		if (iter->data->k > iter->data->end) {
+			WARN_ONCE(1, "bset was corrupt!\n");
+			iter->data->k = iter->data->end;
+		}
+
+		if (iter->data->k == iter->data->end)
+			heap_pop(iter, unused, cmp);
+		else
+			heap_sift(iter, 0, cmp);
+	}
+
+	return ret;
+}
+
+struct bkey *bch_btree_iter_next(struct btree_iter *iter)
+{
+	return __bch_btree_iter_next(iter, btree_iter_cmp);
+
+}
+EXPORT_SYMBOL(bch_btree_iter_next);
+
+struct bkey *bch_btree_iter_next_filter(struct btree_iter *iter,
+					struct btree_keys *b, ptr_filter_fn fn)
+{
+	struct bkey *ret;
+
+	do {
+		ret = bch_btree_iter_next(iter);
+	} while (ret && fn(b, ret));
+
+	return ret;
+}
+
+/* Mergesort */
+
+void bch_bset_sort_state_free(struct bset_sort_state *state)
+{
+	if (state->pool)
+		mempool_destroy(state->pool);
+}
+
+int bch_bset_sort_state_init(struct bset_sort_state *state, unsigned page_order)
+{
+	spin_lock_init(&state->time.lock);
+
+	state->page_order = page_order;
+	state->crit_factor = int_sqrt(1 << page_order);
+
+	state->pool = mempool_create_page_pool(1, page_order);
+	if (!state->pool)
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL(bch_bset_sort_state_init);
+
+static void btree_mergesort(struct btree_keys *b, struct bset *out,
+			    struct btree_iter *iter,
+			    bool fixup, bool remove_stale)
+{
+	int i;
+	struct bkey *k, *last = NULL;
+	BKEY_PADDED(k) tmp;
+	bool (*bad)(struct btree_keys *, const struct bkey *) = remove_stale
+		? bch_ptr_bad
+		: bch_ptr_invalid;
+
+	/* Heapify the iterator, using our comparison function */
+	for (i = iter->used / 2 - 1; i >= 0; --i)
+		heap_sift(iter, i, b->ops->sort_cmp);
+
+	while (!btree_iter_end(iter)) {
+		if (b->ops->sort_fixup && fixup)
+			k = b->ops->sort_fixup(iter, &tmp.k);
+		else
+			k = NULL;
+
+		if (!k)
+			k = __bch_btree_iter_next(iter, b->ops->sort_cmp);
+
+		if (bad(b, k))
+			continue;
+
+		if (!last) {
+			last = out->start;
+			bkey_copy(last, k);
+		} else if (!bch_bkey_try_merge(b, last, k)) {
+			last = bkey_next(last);
+			bkey_copy(last, k);
+		}
+	}
+
+	out->keys = last ? (uint64_t *) bkey_next(last) - out->d : 0;
+
+	pr_debug("sorted %i keys", out->keys);
+}
+
+static void __btree_sort(struct btree_keys *b, struct btree_iter *iter,
+			 unsigned start, unsigned order, bool fixup,
+			 struct bset_sort_state *state)
+{
+	uint64_t start_time;
+	bool used_mempool = false;
+	struct bset *out = (void *) __get_free_pages(__GFP_NOWARN|GFP_NOIO,
+						     order);
+	if (!out) {
+		struct page *outp;
+
+		BUG_ON(order > state->page_order);
+
+		outp = mempool_alloc(state->pool, GFP_NOIO);
+		out = page_address(outp);
+		used_mempool = true;
+		order = state->page_order;
+	}
+
+	start_time = local_clock();
+
+	btree_mergesort(b, out, iter, fixup, false);
+	b->nsets = start;
+
+	if (!start && order == b->page_order) {
+		/*
+		 * Our temporary buffer is the same size as the btree node's
+		 * buffer, we can just swap buffers instead of doing a big
+		 * memcpy()
+		 */
+
+		out->magic	= b->set->data->magic;
+		out->seq	= b->set->data->seq;
+		out->version	= b->set->data->version;
+		swap(out, b->set->data);
+	} else {
+		b->set[start].data->keys = out->keys;
+		memcpy(b->set[start].data->start, out->start,
+		       (void *) bset_bkey_last(out) - (void *) out->start);
+	}
+
+	if (used_mempool)
+		mempool_free(virt_to_page(out), state->pool);
+	else
+		free_pages((unsigned long) out, order);
+
+	bch_bset_build_written_tree(b);
+
+	if (!start)
+		bch_time_stats_update(&state->time, start_time);
+}
+
+void bch_btree_sort_partial(struct btree_keys *b, unsigned start,
+			    struct bset_sort_state *state)
+{
+	size_t order = b->page_order, keys = 0;
+	struct btree_iter iter;
+	int oldsize = bch_count_data(b);
+
+	__bch_btree_iter_init(b, &iter, NULL, &b->set[start]);
+
+	if (start) {
+		unsigned i;
+
+		for (i = start; i <= b->nsets; i++)
+			keys += b->set[i].data->keys;
+
+		order = get_order(__set_bytes(b->set->data, keys));
+	}
+
+	__btree_sort(b, &iter, start, order, false, state);
+
+	EBUG_ON(oldsize >= 0 && bch_count_data(b) != oldsize);
+}
+EXPORT_SYMBOL(bch_btree_sort_partial);
+
+void bch_btree_sort_and_fix_extents(struct btree_keys *b,
+				    struct btree_iter *iter,
+				    struct bset_sort_state *state)
+{
+	__btree_sort(b, iter, 0, b->page_order, true, state);
+}
+
+void bch_btree_sort_into(struct btree_keys *b, struct btree_keys *new,
+			 struct bset_sort_state *state)
+{
+	uint64_t start_time = local_clock();
+
+	struct btree_iter iter;
+	bch_btree_iter_init(b, &iter, NULL);
+
+	btree_mergesort(b, new->set->data, &iter, false, true);
+
+	bch_time_stats_update(&state->time, start_time);
+
+	new->set->size = 0; // XXX: why?
+}
+
+#define SORT_CRIT	(4096 / sizeof(uint64_t))
+
+void bch_btree_sort_lazy(struct btree_keys *b, struct bset_sort_state *state)
+{
+	unsigned crit = SORT_CRIT;
+	int i;
+
+	/* Don't sort if nothing to do */
+	if (!b->nsets)
+		goto out;
+
+	for (i = b->nsets - 1; i >= 0; --i) {
+		crit *= state->crit_factor;
+
+		if (b->set[i].data->keys < crit) {
+			bch_btree_sort_partial(b, i, state);
+			return;
+		}
+	}
+
+	/* Sort if we'd overflow */
+	if (b->nsets + 1 == MAX_BSETS) {
+		bch_btree_sort(b, state);
+		return;
+	}
+
+out:
+	bch_bset_build_written_tree(b);
+}
+EXPORT_SYMBOL(bch_btree_sort_lazy);
+
+void bch_btree_keys_stats(struct btree_keys *b, struct bset_stats *stats)
+{
+	unsigned i;
+
+	for (i = 0; i <= b->nsets; i++) {
+		struct bset_tree *t = &b->set[i];
+		size_t bytes = t->data->keys * sizeof(uint64_t);
+		size_t j;
+
+		if (bset_written(b, t)) {
+			stats->sets_written++;
+			stats->bytes_written += bytes;
+
+			stats->floats += t->size - 1;
+
+			for (j = 1; j < t->size; j++)
+				if (t->tree[j].exponent == 127)
+					stats->failed++;
+		} else {
+			stats->sets_unwritten++;
+			stats->bytes_unwritten += bytes;
+		}
+	}
+}
diff --git a/drivers/md/bcache/bset.h b/drivers/md/bcache/bset.h
new file mode 100644
index 00000000000..5f6728d5d4d
--- /dev/null
+++ b/drivers/md/bcache/bset.h
@@ -0,0 +1,566 @@
+#ifndef _BCACHE_BSET_H
+#define _BCACHE_BSET_H
+
+#include <linux/bcache.h>
+#include <linux/kernel.h>
+#include <linux/types.h>
+
+#include "util.h" /* for time_stats */
+
+/*
+ * BKEYS:
+ *
+ * A bkey contains a key, a size field, a variable number of pointers, and some
+ * ancillary flag bits.
+ *
+ * We use two different functions for validating bkeys, bch_ptr_invalid and
+ * bch_ptr_bad().
+ *
+ * bch_ptr_invalid() primarily filters out keys and pointers that would be
+ * invalid due to some sort of bug, whereas bch_ptr_bad() filters out keys and
+ * pointer that occur in normal practice but don't point to real data.
+ *
+ * The one exception to the rule that ptr_invalid() filters out invalid keys is
+ * that it also filters out keys of size 0 - these are keys that have been
+ * completely overwritten. It'd be safe to delete these in memory while leaving
+ * them on disk, just unnecessary work - so we filter them out when resorting
+ * instead.
+ *
+ * We can't filter out stale keys when we're resorting, because garbage
+ * collection needs to find them to ensure bucket gens don't wrap around -
+ * unless we're rewriting the btree node those stale keys still exist on disk.
+ *
+ * We also implement functions here for removing some number of sectors from the
+ * front or the back of a bkey - this is mainly used for fixing overlapping
+ * extents, by removing the overlapping sectors from the older key.
+ *
+ * BSETS:
+ *
+ * A bset is an array of bkeys laid out contiguously in memory in sorted order,
+ * along with a header. A btree node is made up of a number of these, written at
+ * different times.
+ *
+ * There could be many of them on disk, but we never allow there to be more than
+ * 4 in memory - we lazily resort as needed.
+ *
+ * We implement code here for creating and maintaining auxiliary search trees
+ * (described below) for searching an individial bset, and on top of that we
+ * implement a btree iterator.
+ *
+ * BTREE ITERATOR:
+ *
+ * Most of the code in bcache doesn't care about an individual bset - it needs
+ * to search entire btree nodes and iterate over them in sorted order.
+ *
+ * The btree iterator code serves both functions; it iterates through the keys
+ * in a btree node in sorted order, starting from either keys after a specific
+ * point (if you pass it a search key) or the start of the btree node.
+ *
+ * AUXILIARY SEARCH TREES:
+ *
+ * Since keys are variable length, we can't use a binary search on a bset - we
+ * wouldn't be able to find the start of the next key. But binary searches are
+ * slow anyways, due to terrible cache behaviour; bcache originally used binary
+ * searches and that code topped out at under 50k lookups/second.
+ *
+ * So we need to construct some sort of lookup table. Since we only insert keys
+ * into the last (unwritten) set, most of the keys within a given btree node are
+ * usually in sets that are mostly constant. We use two different types of
+ * lookup tables to take advantage of this.
+ *
+ * Both lookup tables share in common that they don't index every key in the
+ * set; they index one key every BSET_CACHELINE bytes, and then a linear search
+ * is used for the rest.
+ *
+ * For sets that have been written to disk and are no longer being inserted
+ * into, we construct a binary search tree in an array - traversing a binary
+ * search tree in an array gives excellent locality of reference and is very
+ * fast, since both children of any node are adjacent to each other in memory
+ * (and their grandchildren, and great grandchildren...) - this means
+ * prefetching can be used to great effect.
+ *
+ * It's quite useful performance wise to keep these nodes small - not just
+ * because they're more likely to be in L2, but also because we can prefetch
+ * more nodes on a single cacheline and thus prefetch more iterations in advance
+ * when traversing this tree.
+ *
+ * Nodes in the auxiliary search tree must contain both a key to compare against
+ * (we don't want to fetch the key from the set, that would defeat the purpose),
+ * and a pointer to the key. We use a few tricks to compress both of these.
+ *
+ * To compress the pointer, we take advantage of the fact that one node in the
+ * search tree corresponds to precisely BSET_CACHELINE bytes in the set. We have
+ * a function (to_inorder()) that takes the index of a node in a binary tree and
+ * returns what its index would be in an inorder traversal, so we only have to
+ * store the low bits of the offset.
+ *
+ * The key is 84 bits (KEY_DEV + key->key, the offset on the device). To
+ * compress that,  we take advantage of the fact that when we're traversing the
+ * search tree at every iteration we know that both our search key and the key
+ * we're looking for lie within some range - bounded by our previous
+ * comparisons. (We special case the start of a search so that this is true even
+ * at the root of the tree).
+ *
+ * So we know the key we're looking for is between a and b, and a and b don't
+ * differ higher than bit 50, we don't need to check anything higher than bit
+ * 50.
+ *
+ * We don't usually need the rest of the bits, either; we only need enough bits
+ * to partition the key range we're currently checking.  Consider key n - the
+ * key our auxiliary search tree node corresponds to, and key p, the key
+ * immediately preceding n.  The lowest bit we need to store in the auxiliary
+ * search tree is the highest bit that differs between n and p.
+ *
+ * Note that this could be bit 0 - we might sometimes need all 80 bits to do the
+ * comparison. But we'd really like our nodes in the auxiliary search tree to be
+ * of fixed size.
+ *
+ * The solution is to make them fixed size, and when we're constructing a node
+ * check if p and n differed in the bits we needed them to. If they don't we
+ * flag that node, and when doing lookups we fallback to comparing against the
+ * real key. As long as this doesn't happen to often (and it seems to reliably
+ * happen a bit less than 1% of the time), we win - even on failures, that key
+ * is then more likely to be in cache than if we were doing binary searches all
+ * the way, since we're touching so much less memory.
+ *
+ * The keys in the auxiliary search tree are stored in (software) floating
+ * point, with an exponent and a mantissa. The exponent needs to be big enough
+ * to address all the bits in the original key, but the number of bits in the
+ * mantissa is somewhat arbitrary; more bits just gets us fewer failures.
+ *
+ * We need 7 bits for the exponent and 3 bits for the key's offset (since keys
+ * are 8 byte aligned); using 22 bits for the mantissa means a node is 4 bytes.
+ * We need one node per 128 bytes in the btree node, which means the auxiliary
+ * search trees take up 3% as much memory as the btree itself.
+ *
+ * Constructing these auxiliary search trees is moderately expensive, and we
+ * don't want to be constantly rebuilding the search tree for the last set
+ * whenever we insert another key into it. For the unwritten set, we use a much
+ * simpler lookup table - it's just a flat array, so index i in the lookup table
+ * corresponds to the i range of BSET_CACHELINE bytes in the set. Indexing
+ * within each byte range works the same as with the auxiliary search trees.
+ *
+ * These are much easier to keep up to date when we insert a key - we do it
+ * somewhat lazily; when we shift a key up we usually just increment the pointer
+ * to it, only when it would overflow do we go to the trouble of finding the
+ * first key in that range of bytes again.
+ */
+
+struct btree_keys;
+struct btree_iter;
+struct btree_iter_set;
+struct bkey_float;
+
+#define MAX_BSETS		4U
+
+struct bset_tree {
+	/*
+	 * We construct a binary tree in an array as if the array
+	 * started at 1, so that things line up on the same cachelines
+	 * better: see comments in bset.c at cacheline_to_bkey() for
+	 * details
+	 */
+
+	/* size of the binary tree and prev array */
+	unsigned		size;
+
+	/* function of size - precalculated for to_inorder() */
+	unsigned		extra;
+
+	/* copy of the last key in the set */
+	struct bkey		end;
+	struct bkey_float	*tree;
+
+	/*
+	 * The nodes in the bset tree point to specific keys - this
+	 * array holds the sizes of the previous key.
+	 *
+	 * Conceptually it's a member of struct bkey_float, but we want
+	 * to keep bkey_float to 4 bytes and prev isn't used in the fast
+	 * path.
+	 */
+	uint8_t			*prev;
+
+	/* The actual btree node, with pointers to each sorted set */
+	struct bset		*data;
+};
+
+struct btree_keys_ops {
+	bool		(*sort_cmp)(struct btree_iter_set,
+				    struct btree_iter_set);
+	struct bkey	*(*sort_fixup)(struct btree_iter *, struct bkey *);
+	bool		(*insert_fixup)(struct btree_keys *, struct bkey *,
+					struct btree_iter *, struct bkey *);
+	bool		(*key_invalid)(struct btree_keys *,
+				       const struct bkey *);
+	bool		(*key_bad)(struct btree_keys *, const struct bkey *);
+	bool		(*key_merge)(struct btree_keys *,
+				     struct bkey *, struct bkey *);
+	void		(*key_to_text)(char *, size_t, const struct bkey *);
+	void		(*key_dump)(struct btree_keys *, const struct bkey *);
+
+	/*
+	 * Only used for deciding whether to use START_KEY(k) or just the key
+	 * itself in a couple places
+	 */
+	bool		is_extents;
+};
+
+struct btree_keys {
+	const struct btree_keys_ops	*ops;
+	uint8_t			page_order;
+	uint8_t			nsets;
+	unsigned		last_set_unwritten:1;
+	bool			*expensive_debug_checks;
+
+	/*
+	 * Sets of sorted keys - the real btree node - plus a binary search tree
+	 *
+	 * set[0] is special; set[0]->tree, set[0]->prev and set[0]->data point
+	 * to the memory we have allocated for this btree node. Additionally,
+	 * set[0]->data points to the entire btree node as it exists on disk.
+	 */
+	struct bset_tree	set[MAX_BSETS];
+};
+
+static inline struct bset_tree *bset_tree_last(struct btree_keys *b)
+{
+	return b->set + b->nsets;
+}
+
+static inline bool bset_written(struct btree_keys *b, struct bset_tree *t)
+{
+	return t <= b->set + b->nsets - b->last_set_unwritten;
+}
+
+static inline bool bkey_written(struct btree_keys *b, struct bkey *k)
+{
+	return !b->last_set_unwritten || k < b->set[b->nsets].data->start;
+}
+
+static inline unsigned bset_byte_offset(struct btree_keys *b, struct bset *i)
+{
+	return ((size_t) i) - ((size_t) b->set->data);
+}
+
+static inline unsigned bset_sector_offset(struct btree_keys *b, struct bset *i)
+{
+	return bset_byte_offset(b, i) >> 9;
+}
+
+#define __set_bytes(i, k)	(sizeof(*(i)) + (k) * sizeof(uint64_t))
+#define set_bytes(i)		__set_bytes(i, i->keys)
+
+#define __set_blocks(i, k, block_bytes)				\
+	DIV_ROUND_UP(__set_bytes(i, k), block_bytes)
+#define set_blocks(i, block_bytes)				\
+	__set_blocks(i, (i)->keys, block_bytes)
+
+static inline size_t bch_btree_keys_u64s_remaining(struct btree_keys *b)
+{
+	struct bset_tree *t = bset_tree_last(b);
+
+	BUG_ON((PAGE_SIZE << b->page_order) <
+	       (bset_byte_offset(b, t->data) + set_bytes(t->data)));
+
+	if (!b->last_set_unwritten)
+		return 0;
+
+	return ((PAGE_SIZE << b->page_order) -
+		(bset_byte_offset(b, t->data) + set_bytes(t->data))) /
+		sizeof(u64);
+}
+
+static inline struct bset *bset_next_set(struct btree_keys *b,
+					 unsigned block_bytes)
+{
+	struct bset *i = bset_tree_last(b)->data;
+
+	return ((void *) i) + roundup(set_bytes(i), block_bytes);
+}
+
+void bch_btree_keys_free(struct btree_keys *);
+int bch_btree_keys_alloc(struct btree_keys *, unsigned, gfp_t);
+void bch_btree_keys_init(struct btree_keys *, const struct btree_keys_ops *,
+			 bool *);
+
+void bch_bset_init_next(struct btree_keys *, struct bset *, uint64_t);
+void bch_bset_build_written_tree(struct btree_keys *);
+void bch_bset_fix_invalidated_key(struct btree_keys *, struct bkey *);
+bool bch_bkey_try_merge(struct btree_keys *, struct bkey *, struct bkey *);
+void bch_bset_insert(struct btree_keys *, struct bkey *, struct bkey *);
+unsigned bch_btree_insert_key(struct btree_keys *, struct bkey *,
+			      struct bkey *);
+
+enum {
+	BTREE_INSERT_STATUS_NO_INSERT = 0,
+	BTREE_INSERT_STATUS_INSERT,
+	BTREE_INSERT_STATUS_BACK_MERGE,
+	BTREE_INSERT_STATUS_OVERWROTE,
+	BTREE_INSERT_STATUS_FRONT_MERGE,
+};
+
+/* Btree key iteration */
+
+struct btree_iter {
+	size_t size, used;
+#ifdef CONFIG_BCACHE_DEBUG
+	struct btree_keys *b;
+#endif
+	struct btree_iter_set {
+		struct bkey *k, *end;
+	} data[MAX_BSETS];
+};
+
+typedef bool (*ptr_filter_fn)(struct btree_keys *, const struct bkey *);
+
+struct bkey *bch_btree_iter_next(struct btree_iter *);
+struct bkey *bch_btree_iter_next_filter(struct btree_iter *,
+					struct btree_keys *, ptr_filter_fn);
+
+void bch_btree_iter_push(struct btree_iter *, struct bkey *, struct bkey *);
+struct bkey *bch_btree_iter_init(struct btree_keys *, struct btree_iter *,
+				 struct bkey *);
+
+struct bkey *__bch_bset_search(struct btree_keys *, struct bset_tree *,
+			       const struct bkey *);
+
+/*
+ * Returns the first key that is strictly greater than search
+ */
+static inline struct bkey *bch_bset_search(struct btree_keys *b,
+					   struct bset_tree *t,
+					   const struct bkey *search)
+{
+	return search ? __bch_bset_search(b, t, search) : t->data->start;
+}
+
+#define for_each_key_filter(b, k, iter, filter)				\
+	for (bch_btree_iter_init((b), (iter), NULL);			\
+	     ((k) = bch_btree_iter_next_filter((iter), (b), filter));)
+
+#define for_each_key(b, k, iter)					\
+	for (bch_btree_iter_init((b), (iter), NULL);			\
+	     ((k) = bch_btree_iter_next(iter));)
+
+/* Sorting */
+
+struct bset_sort_state {
+	mempool_t		*pool;
+
+	unsigned		page_order;
+	unsigned		crit_factor;
+
+	struct time_stats	time;
+};
+
+void bch_bset_sort_state_free(struct bset_sort_state *);
+int bch_bset_sort_state_init(struct bset_sort_state *, unsigned);
+void bch_btree_sort_lazy(struct btree_keys *, struct bset_sort_state *);
+void bch_btree_sort_into(struct btree_keys *, struct btree_keys *,
+			 struct bset_sort_state *);
+void bch_btree_sort_and_fix_extents(struct btree_keys *, struct btree_iter *,
+				    struct bset_sort_state *);
+void bch_btree_sort_partial(struct btree_keys *, unsigned,
+			    struct bset_sort_state *);
+
+static inline void bch_btree_sort(struct btree_keys *b,
+				  struct bset_sort_state *state)
+{
+	bch_btree_sort_partial(b, 0, state);
+}
+
+struct bset_stats {
+	size_t sets_written, sets_unwritten;
+	size_t bytes_written, bytes_unwritten;
+	size_t floats, failed;
+};
+
+void bch_btree_keys_stats(struct btree_keys *, struct bset_stats *);
+
+/* Bkey utility code */
+
+#define bset_bkey_last(i)	bkey_idx((struct bkey *) (i)->d, (i)->keys)
+
+static inline struct bkey *bset_bkey_idx(struct bset *i, unsigned idx)
+{
+	return bkey_idx(i->start, idx);
+}
+
+static inline void bkey_init(struct bkey *k)
+{
+	*k = ZERO_KEY;
+}
+
+static __always_inline int64_t bkey_cmp(const struct bkey *l,
+					const struct bkey *r)
+{
+	return unlikely(KEY_INODE(l) != KEY_INODE(r))
+		? (int64_t) KEY_INODE(l) - (int64_t) KEY_INODE(r)
+		: (int64_t) KEY_OFFSET(l) - (int64_t) KEY_OFFSET(r);
+}
+
+void bch_bkey_copy_single_ptr(struct bkey *, const struct bkey *,
+			      unsigned);
+bool __bch_cut_front(const struct bkey *, struct bkey *);
+bool __bch_cut_back(const struct bkey *, struct bkey *);
+
+static inline bool bch_cut_front(const struct bkey *where, struct bkey *k)
+{
+	BUG_ON(bkey_cmp(where, k) > 0);
+	return __bch_cut_front(where, k);
+}
+
+static inline bool bch_cut_back(const struct bkey *where, struct bkey *k)
+{
+	BUG_ON(bkey_cmp(where, &START_KEY(k)) < 0);
+	return __bch_cut_back(where, k);
+}
+
+#define PRECEDING_KEY(_k)					\
+({								\
+	struct bkey *_ret = NULL;				\
+								\
+	if (KEY_INODE(_k) || KEY_OFFSET(_k)) {			\
+		_ret = &KEY(KEY_INODE(_k), KEY_OFFSET(_k), 0);	\
+								\
+		if (!_ret->low)					\
+			_ret->high--;				\
+		_ret->low--;					\
+	}							\
+								\
+	_ret;							\
+})
+
+static inline bool bch_ptr_invalid(struct btree_keys *b, const struct bkey *k)
+{
+	return b->ops->key_invalid(b, k);
+}
+
+static inline bool bch_ptr_bad(struct btree_keys *b, const struct bkey *k)
+{
+	return b->ops->key_bad(b, k);
+}
+
+static inline void bch_bkey_to_text(struct btree_keys *b, char *buf,
+				    size_t size, const struct bkey *k)
+{
+	return b->ops->key_to_text(buf, size, k);
+}
+
+static inline bool bch_bkey_equal_header(const struct bkey *l,
+					 const struct bkey *r)
+{
+	return (KEY_DIRTY(l) == KEY_DIRTY(r) &&
+		KEY_PTRS(l) == KEY_PTRS(r) &&
+		KEY_CSUM(l) == KEY_CSUM(l));
+}
+
+/* Keylists */
+
+struct keylist {
+	union {
+		struct bkey		*keys;
+		uint64_t		*keys_p;
+	};
+	union {
+		struct bkey		*top;
+		uint64_t		*top_p;
+	};
+
+	/* Enough room for btree_split's keys without realloc */
+#define KEYLIST_INLINE		16
+	uint64_t		inline_keys[KEYLIST_INLINE];
+};
+
+static inline void bch_keylist_init(struct keylist *l)
+{
+	l->top_p = l->keys_p = l->inline_keys;
+}
+
+static inline void bch_keylist_init_single(struct keylist *l, struct bkey *k)
+{
+	l->keys = k;
+	l->top = bkey_next(k);
+}
+
+static inline void bch_keylist_push(struct keylist *l)
+{
+	l->top = bkey_next(l->top);
+}
+
+static inline void bch_keylist_add(struct keylist *l, struct bkey *k)
+{
+	bkey_copy(l->top, k);
+	bch_keylist_push(l);
+}
+
+static inline bool bch_keylist_empty(struct keylist *l)
+{
+	return l->top == l->keys;
+}
+
+static inline void bch_keylist_reset(struct keylist *l)
+{
+	l->top = l->keys;
+}
+
+static inline void bch_keylist_free(struct keylist *l)
+{
+	if (l->keys_p != l->inline_keys)
+		kfree(l->keys_p);
+}
+
+static inline size_t bch_keylist_nkeys(struct keylist *l)
+{
+	return l->top_p - l->keys_p;
+}
+
+static inline size_t bch_keylist_bytes(struct keylist *l)
+{
+	return bch_keylist_nkeys(l) * sizeof(uint64_t);
+}
+
+struct bkey *bch_keylist_pop(struct keylist *);
+void bch_keylist_pop_front(struct keylist *);
+int __bch_keylist_realloc(struct keylist *, unsigned);
+
+/* Debug stuff */
+
+#ifdef CONFIG_BCACHE_DEBUG
+
+int __bch_count_data(struct btree_keys *);
+void __bch_check_keys(struct btree_keys *, const char *, ...);
+void bch_dump_bset(struct btree_keys *, struct bset *, unsigned);
+void bch_dump_bucket(struct btree_keys *);
+
+#else
+
+static inline int __bch_count_data(struct btree_keys *b) { return -1; }
+static inline void __bch_check_keys(struct btree_keys *b, const char *fmt, ...) {}
+static inline void bch_dump_bucket(struct btree_keys *b) {}
+void bch_dump_bset(struct btree_keys *, struct bset *, unsigned);
+
+#endif
+
+static inline bool btree_keys_expensive_checks(struct btree_keys *b)
+{
+#ifdef CONFIG_BCACHE_DEBUG
+	return *b->expensive_debug_checks;
+#else
+	return false;
+#endif
+}
+
+static inline int bch_count_data(struct btree_keys *b)
+{
+	return btree_keys_expensive_checks(b) ? __bch_count_data(b) : -1;
+}
+
+#define bch_check_keys(b, ...)						\
+do {									\
+	if (btree_keys_expensive_checks(b))				\
+		__bch_check_keys(b, __VA_ARGS__);			\
+} while (0)
+
+#endif
diff --git a/drivers/md/bcache/btree.c b/drivers/md/bcache/btree.c
new file mode 100644
index 00000000000..7347b610096
--- /dev/null
+++ b/drivers/md/bcache/btree.c
@@ -0,0 +1,2518 @@
+/*
+ * Copyright (C) 2010 Kent Overstreet <kent.overstreet@gmail.com>
+ *
+ * Uses a block device as cache for other block devices; optimized for SSDs.
+ * All allocation is done in buckets, which should match the erase block size
+ * of the device.
+ *
+ * Buckets containing cached data are kept on a heap sorted by priority;
+ * bucket priority is increased on cache hit, and periodically all the buckets
+ * on the heap have their priority scaled down. This currently is just used as
+ * an LRU but in the future should allow for more intelligent heuristics.
+ *
+ * Buckets have an 8 bit counter; freeing is accomplished by incrementing the
+ * counter. Garbage collection is used to remove stale pointers.
+ *
+ * Indexing is done via a btree; nodes are not necessarily fully sorted, rather
+ * as keys are inserted we only sort the pages that have not yet been written.
+ * When garbage collection is run, we resort the entire node.
+ *
+ * All configuration is done via sysfs; see Documentation/bcache.txt.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "extents.h"
+
+#include <linux/slab.h>
+#include <linux/bitops.h>
+#include <linux/freezer.h>
+#include <linux/hash.h>
+#include <linux/kthread.h>
+#include <linux/prefetch.h>
+#include <linux/random.h>
+#include <linux/rcupdate.h>
+#include <trace/events/bcache.h>
+
+/*
+ * Todo:
+ * register_bcache: Return errors out to userspace correctly
+ *
+ * Writeback: don't undirty key until after a cache flush
+ *
+ * Create an iterator for key pointers
+ *
+ * On btree write error, mark bucket such that it won't be freed from the cache
+ *
+ * Journalling:
+ *   Check for bad keys in replay
+ *   Propagate barriers
+ *   Refcount journal entries in journal_replay
+ *
+ * Garbage collection:
+ *   Finish incremental gc
+ *   Gc should free old UUIDs, data for invalid UUIDs
+ *
+ * Provide a way to list backing device UUIDs we have data cached for, and
+ * probably how long it's been since we've seen them, and a way to invalidate
+ * dirty data for devices that will never be attached again
+ *
+ * Keep 1 min/5 min/15 min statistics of how busy a block device has been, so
+ * that based on that and how much dirty data we have we can keep writeback
+ * from being starved
+ *
+ * Add a tracepoint or somesuch to watch for writeback starvation
+ *
+ * When btree depth > 1 and splitting an interior node, we have to make sure
+ * alloc_bucket() cannot fail. This should be true but is not completely
+ * obvious.
+ *
+ * Plugging?
+ *
+ * If data write is less than hard sector size of ssd, round up offset in open
+ * bucket to the next whole sector
+ *
+ * Superblock needs to be fleshed out for multiple cache devices
+ *
+ * Add a sysfs tunable for the number of writeback IOs in flight
+ *
+ * Add a sysfs tunable for the number of open data buckets
+ *
+ * IO tracking: Can we track when one process is doing io on behalf of another?
+ * IO tracking: Don't use just an average, weigh more recent stuff higher
+ *
+ * Test module load/unload
+ */
+
+#define MAX_NEED_GC		64
+#define MAX_SAVE_PRIO		72
+
+#define PTR_DIRTY_BIT		(((uint64_t) 1 << 36))
+
+#define PTR_HASH(c, k)							\
+	(((k)->ptr[0] >> c->bucket_bits) | PTR_GEN(k, 0))
+
+#define insert_lock(s, b)	((b)->level <= (s)->lock)
+
+/*
+ * These macros are for recursing down the btree - they handle the details of
+ * locking and looking up nodes in the cache for you. They're best treated as
+ * mere syntax when reading code that uses them.
+ *
+ * op->lock determines whether we take a read or a write lock at a given depth.
+ * If you've got a read lock and find that you need a write lock (i.e. you're
+ * going to have to split), set op->lock and return -EINTR; btree_root() will
+ * call you again and you'll have the correct lock.
+ */
+
+/**
+ * btree - recurse down the btree on a specified key
+ * @fn:		function to call, which will be passed the child node
+ * @key:	key to recurse on
+ * @b:		parent btree node
+ * @op:		pointer to struct btree_op
+ */
+#define btree(fn, key, b, op, ...)					\
+({									\
+	int _r, l = (b)->level - 1;					\
+	bool _w = l <= (op)->lock;					\
+	struct btree *_child = bch_btree_node_get((b)->c, op, key, l, _w);\
+	if (!IS_ERR(_child)) {						\
+		_child->parent = (b);					\
+		_r = bch_btree_ ## fn(_child, op, ##__VA_ARGS__);	\
+		rw_unlock(_w, _child);					\
+	} else								\
+		_r = PTR_ERR(_child);					\
+	_r;								\
+})
+
+/**
+ * btree_root - call a function on the root of the btree
+ * @fn:		function to call, which will be passed the child node
+ * @c:		cache set
+ * @op:		pointer to struct btree_op
+ */
+#define btree_root(fn, c, op, ...)					\
+({									\
+	int _r = -EINTR;						\
+	do {								\
+		struct btree *_b = (c)->root;				\
+		bool _w = insert_lock(op, _b);				\
+		rw_lock(_w, _b, _b->level);				\
+		if (_b == (c)->root &&					\
+		    _w == insert_lock(op, _b)) {			\
+			_b->parent = NULL;				\
+			_r = bch_btree_ ## fn(_b, op, ##__VA_ARGS__);	\
+		}							\
+		rw_unlock(_w, _b);					\
+		bch_cannibalize_unlock(c);				\
+		if (_r == -EINTR)					\
+			schedule();					\
+	} while (_r == -EINTR);						\
+									\
+	finish_wait(&(c)->btree_cache_wait, &(op)->wait);		\
+	_r;								\
+})
+
+static inline struct bset *write_block(struct btree *b)
+{
+	return ((void *) btree_bset_first(b)) + b->written * block_bytes(b->c);
+}
+
+static void bch_btree_init_next(struct btree *b)
+{
+	/* If not a leaf node, always sort */
+	if (b->level && b->keys.nsets)
+		bch_btree_sort(&b->keys, &b->c->sort);
+	else
+		bch_btree_sort_lazy(&b->keys, &b->c->sort);
+
+	if (b->written < btree_blocks(b))
+		bch_bset_init_next(&b->keys, write_block(b),
+				   bset_magic(&b->c->sb));
+
+}
+
+/* Btree key manipulation */
+
+void bkey_put(struct cache_set *c, struct bkey *k)
+{
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (ptr_available(c, k, i))
+			atomic_dec_bug(&PTR_BUCKET(c, k, i)->pin);
+}
+
+/* Btree IO */
+
+static uint64_t btree_csum_set(struct btree *b, struct bset *i)
+{
+	uint64_t crc = b->key.ptr[0];
+	void *data = (void *) i + 8, *end = bset_bkey_last(i);
+
+	crc = bch_crc64_update(crc, data, end - data);
+	return crc ^ 0xffffffffffffffffULL;
+}
+
+void bch_btree_node_read_done(struct btree *b)
+{
+	const char *err = "bad btree header";
+	struct bset *i = btree_bset_first(b);
+	struct btree_iter *iter;
+
+	iter = mempool_alloc(b->c->fill_iter, GFP_NOWAIT);
+	iter->size = b->c->sb.bucket_size / b->c->sb.block_size;
+	iter->used = 0;
+
+#ifdef CONFIG_BCACHE_DEBUG
+	iter->b = &b->keys;
+#endif
+
+	if (!i->seq)
+		goto err;
+
+	for (;
+	     b->written < btree_blocks(b) && i->seq == b->keys.set[0].data->seq;
+	     i = write_block(b)) {
+		err = "unsupported bset version";
+		if (i->version > BCACHE_BSET_VERSION)
+			goto err;
+
+		err = "bad btree header";
+		if (b->written + set_blocks(i, block_bytes(b->c)) >
+		    btree_blocks(b))
+			goto err;
+
+		err = "bad magic";
+		if (i->magic != bset_magic(&b->c->sb))
+			goto err;
+
+		err = "bad checksum";
+		switch (i->version) {
+		case 0:
+			if (i->csum != csum_set(i))
+				goto err;
+			break;
+		case BCACHE_BSET_VERSION:
+			if (i->csum != btree_csum_set(b, i))
+				goto err;
+			break;
+		}
+
+		err = "empty set";
+		if (i != b->keys.set[0].data && !i->keys)
+			goto err;
+
+		bch_btree_iter_push(iter, i->start, bset_bkey_last(i));
+
+		b->written += set_blocks(i, block_bytes(b->c));
+	}
+
+	err = "corrupted btree";
+	for (i = write_block(b);
+	     bset_sector_offset(&b->keys, i) < KEY_SIZE(&b->key);
+	     i = ((void *) i) + block_bytes(b->c))
+		if (i->seq == b->keys.set[0].data->seq)
+			goto err;
+
+	bch_btree_sort_and_fix_extents(&b->keys, iter, &b->c->sort);
+
+	i = b->keys.set[0].data;
+	err = "short btree key";
+	if (b->keys.set[0].size &&
+	    bkey_cmp(&b->key, &b->keys.set[0].end) < 0)
+		goto err;
+
+	if (b->written < btree_blocks(b))
+		bch_bset_init_next(&b->keys, write_block(b),
+				   bset_magic(&b->c->sb));
+out:
+	mempool_free(iter, b->c->fill_iter);
+	return;
+err:
+	set_btree_node_io_error(b);
+	bch_cache_set_error(b->c, "%s at bucket %zu, block %u, %u keys",
+			    err, PTR_BUCKET_NR(b->c, &b->key, 0),
+			    bset_block_offset(b, i), i->keys);
+	goto out;
+}
+
+static void btree_node_read_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	closure_put(cl);
+}
+
+static void bch_btree_node_read(struct btree *b)
+{
+	uint64_t start_time = local_clock();
+	struct closure cl;
+	struct bio *bio;
+
+	trace_bcache_btree_read(b);
+
+	closure_init_stack(&cl);
+
+	bio = bch_bbio_alloc(b->c);
+	bio->bi_rw	= REQ_META|READ_SYNC;
+	bio->bi_iter.bi_size = KEY_SIZE(&b->key) << 9;
+	bio->bi_end_io	= btree_node_read_endio;
+	bio->bi_private	= &cl;
+
+	bch_bio_map(bio, b->keys.set[0].data);
+
+	bch_submit_bbio(bio, b->c, &b->key, 0);
+	closure_sync(&cl);
+
+	if (!test_bit(BIO_UPTODATE, &bio->bi_flags))
+		set_btree_node_io_error(b);
+
+	bch_bbio_free(bio, b->c);
+
+	if (btree_node_io_error(b))
+		goto err;
+
+	bch_btree_node_read_done(b);
+	bch_time_stats_update(&b->c->btree_read_time, start_time);
+
+	return;
+err:
+	bch_cache_set_error(b->c, "io error reading bucket %zu",
+			    PTR_BUCKET_NR(b->c, &b->key, 0));
+}
+
+static void btree_complete_write(struct btree *b, struct btree_write *w)
+{
+	if (w->prio_blocked &&
+	    !atomic_sub_return(w->prio_blocked, &b->c->prio_blocked))
+		wake_up_allocators(b->c);
+
+	if (w->journal) {
+		atomic_dec_bug(w->journal);
+		__closure_wake_up(&b->c->journal.wait);
+	}
+
+	w->prio_blocked	= 0;
+	w->journal	= NULL;
+}
+
+static void btree_node_write_unlock(struct closure *cl)
+{
+	struct btree *b = container_of(cl, struct btree, io);
+
+	up(&b->io_mutex);
+}
+
+static void __btree_node_write_done(struct closure *cl)
+{
+	struct btree *b = container_of(cl, struct btree, io);
+	struct btree_write *w = btree_prev_write(b);
+
+	bch_bbio_free(b->bio, b->c);
+	b->bio = NULL;
+	btree_complete_write(b, w);
+
+	if (btree_node_dirty(b))
+		schedule_delayed_work(&b->work, 30 * HZ);
+
+	closure_return_with_destructor(cl, btree_node_write_unlock);
+}
+
+static void btree_node_write_done(struct closure *cl)
+{
+	struct btree *b = container_of(cl, struct btree, io);
+	struct bio_vec *bv;
+	int n;
+
+	bio_for_each_segment_all(bv, b->bio, n)
+		__free_page(bv->bv_page);
+
+	__btree_node_write_done(cl);
+}
+
+static void btree_node_write_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	struct btree *b = container_of(cl, struct btree, io);
+
+	if (error)
+		set_btree_node_io_error(b);
+
+	bch_bbio_count_io_errors(b->c, bio, error, "writing btree");
+	closure_put(cl);
+}
+
+static void do_btree_node_write(struct btree *b)
+{
+	struct closure *cl = &b->io;
+	struct bset *i = btree_bset_last(b);
+	BKEY_PADDED(key) k;
+
+	i->version	= BCACHE_BSET_VERSION;
+	i->csum		= btree_csum_set(b, i);
+
+	BUG_ON(b->bio);
+	b->bio = bch_bbio_alloc(b->c);
+
+	b->bio->bi_end_io	= btree_node_write_endio;
+	b->bio->bi_private	= cl;
+	b->bio->bi_rw		= REQ_META|WRITE_SYNC|REQ_FUA;
+	b->bio->bi_iter.bi_size	= roundup(set_bytes(i), block_bytes(b->c));
+	bch_bio_map(b->bio, i);
+
+	/*
+	 * If we're appending to a leaf node, we don't technically need FUA -
+	 * this write just needs to be persisted before the next journal write,
+	 * which will be marked FLUSH|FUA.
+	 *
+	 * Similarly if we're writing a new btree root - the pointer is going to
+	 * be in the next journal entry.
+	 *
+	 * But if we're writing a new btree node (that isn't a root) or
+	 * appending to a non leaf btree node, we need either FUA or a flush
+	 * when we write the parent with the new pointer. FUA is cheaper than a
+	 * flush, and writes appending to leaf nodes aren't blocking anything so
+	 * just make all btree node writes FUA to keep things sane.
+	 */
+
+	bkey_copy(&k.key, &b->key);
+	SET_PTR_OFFSET(&k.key, 0, PTR_OFFSET(&k.key, 0) +
+		       bset_sector_offset(&b->keys, i));
+
+	if (!bio_alloc_pages(b->bio, GFP_NOIO)) {
+		int j;
+		struct bio_vec *bv;
+		void *base = (void *) ((unsigned long) i & ~(PAGE_SIZE - 1));
+
+		bio_for_each_segment_all(bv, b->bio, j)
+			memcpy(page_address(bv->bv_page),
+			       base + j * PAGE_SIZE, PAGE_SIZE);
+
+		bch_submit_bbio(b->bio, b->c, &k.key, 0);
+
+		continue_at(cl, btree_node_write_done, NULL);
+	} else {
+		b->bio->bi_vcnt = 0;
+		bch_bio_map(b->bio, i);
+
+		bch_submit_bbio(b->bio, b->c, &k.key, 0);
+
+		closure_sync(cl);
+		continue_at_nobarrier(cl, __btree_node_write_done, NULL);
+	}
+}
+
+void __bch_btree_node_write(struct btree *b, struct closure *parent)
+{
+	struct bset *i = btree_bset_last(b);
+
+	lockdep_assert_held(&b->write_lock);
+
+	trace_bcache_btree_write(b);
+
+	BUG_ON(current->bio_list);
+	BUG_ON(b->written >= btree_blocks(b));
+	BUG_ON(b->written && !i->keys);
+	BUG_ON(btree_bset_first(b)->seq != i->seq);
+	bch_check_keys(&b->keys, "writing");
+
+	cancel_delayed_work(&b->work);
+
+	/* If caller isn't waiting for write, parent refcount is cache set */
+	down(&b->io_mutex);
+	closure_init(&b->io, parent ?: &b->c->cl);
+
+	clear_bit(BTREE_NODE_dirty,	 &b->flags);
+	change_bit(BTREE_NODE_write_idx, &b->flags);
+
+	do_btree_node_write(b);
+
+	atomic_long_add(set_blocks(i, block_bytes(b->c)) * b->c->sb.block_size,
+			&PTR_CACHE(b->c, &b->key, 0)->btree_sectors_written);
+
+	b->written += set_blocks(i, block_bytes(b->c));
+}
+
+void bch_btree_node_write(struct btree *b, struct closure *parent)
+{
+	unsigned nsets = b->keys.nsets;
+
+	lockdep_assert_held(&b->lock);
+
+	__bch_btree_node_write(b, parent);
+
+	/*
+	 * do verify if there was more than one set initially (i.e. we did a
+	 * sort) and we sorted down to a single set:
+	 */
+	if (nsets && !b->keys.nsets)
+		bch_btree_verify(b);
+
+	bch_btree_init_next(b);
+}
+
+static void bch_btree_node_write_sync(struct btree *b)
+{
+	struct closure cl;
+
+	closure_init_stack(&cl);
+
+	mutex_lock(&b->write_lock);
+	bch_btree_node_write(b, &cl);
+	mutex_unlock(&b->write_lock);
+
+	closure_sync(&cl);
+}
+
+static void btree_node_write_work(struct work_struct *w)
+{
+	struct btree *b = container_of(to_delayed_work(w), struct btree, work);
+
+	mutex_lock(&b->write_lock);
+	if (btree_node_dirty(b))
+		__bch_btree_node_write(b, NULL);
+	mutex_unlock(&b->write_lock);
+}
+
+static void bch_btree_leaf_dirty(struct btree *b, atomic_t *journal_ref)
+{
+	struct bset *i = btree_bset_last(b);
+	struct btree_write *w = btree_current_write(b);
+
+	lockdep_assert_held(&b->write_lock);
+
+	BUG_ON(!b->written);
+	BUG_ON(!i->keys);
+
+	if (!btree_node_dirty(b))
+		schedule_delayed_work(&b->work, 30 * HZ);
+
+	set_btree_node_dirty(b);
+
+	if (journal_ref) {
+		if (w->journal &&
+		    journal_pin_cmp(b->c, w->journal, journal_ref)) {
+			atomic_dec_bug(w->journal);
+			w->journal = NULL;
+		}
+
+		if (!w->journal) {
+			w->journal = journal_ref;
+			atomic_inc(w->journal);
+		}
+	}
+
+	/* Force write if set is too big */
+	if (set_bytes(i) > PAGE_SIZE - 48 &&
+	    !current->bio_list)
+		bch_btree_node_write(b, NULL);
+}
+
+/*
+ * Btree in memory cache - allocation/freeing
+ * mca -> memory cache
+ */
+
+#define mca_reserve(c)	(((c->root && c->root->level)		\
+			  ? c->root->level : 1) * 8 + 16)
+#define mca_can_free(c)						\
+	max_t(int, 0, c->btree_cache_used - mca_reserve(c))
+
+static void mca_data_free(struct btree *b)
+{
+	BUG_ON(b->io_mutex.count != 1);
+
+	bch_btree_keys_free(&b->keys);
+
+	b->c->btree_cache_used--;
+	list_move(&b->list, &b->c->btree_cache_freed);
+}
+
+static void mca_bucket_free(struct btree *b)
+{
+	BUG_ON(btree_node_dirty(b));
+
+	b->key.ptr[0] = 0;
+	hlist_del_init_rcu(&b->hash);
+	list_move(&b->list, &b->c->btree_cache_freeable);
+}
+
+static unsigned btree_order(struct bkey *k)
+{
+	return ilog2(KEY_SIZE(k) / PAGE_SECTORS ?: 1);
+}
+
+static void mca_data_alloc(struct btree *b, struct bkey *k, gfp_t gfp)
+{
+	if (!bch_btree_keys_alloc(&b->keys,
+				  max_t(unsigned,
+					ilog2(b->c->btree_pages),
+					btree_order(k)),
+				  gfp)) {
+		b->c->btree_cache_used++;
+		list_move(&b->list, &b->c->btree_cache);
+	} else {
+		list_move(&b->list, &b->c->btree_cache_freed);
+	}
+}
+
+static struct btree *mca_bucket_alloc(struct cache_set *c,
+				      struct bkey *k, gfp_t gfp)
+{
+	struct btree *b = kzalloc(sizeof(struct btree), gfp);
+	if (!b)
+		return NULL;
+
+	init_rwsem(&b->lock);
+	lockdep_set_novalidate_class(&b->lock);
+	mutex_init(&b->write_lock);
+	lockdep_set_novalidate_class(&b->write_lock);
+	INIT_LIST_HEAD(&b->list);
+	INIT_DELAYED_WORK(&b->work, btree_node_write_work);
+	b->c = c;
+	sema_init(&b->io_mutex, 1);
+
+	mca_data_alloc(b, k, gfp);
+	return b;
+}
+
+static int mca_reap(struct btree *b, unsigned min_order, bool flush)
+{
+	struct closure cl;
+
+	closure_init_stack(&cl);
+	lockdep_assert_held(&b->c->bucket_lock);
+
+	if (!down_write_trylock(&b->lock))
+		return -ENOMEM;
+
+	BUG_ON(btree_node_dirty(b) && !b->keys.set[0].data);
+
+	if (b->keys.page_order < min_order)
+		goto out_unlock;
+
+	if (!flush) {
+		if (btree_node_dirty(b))
+			goto out_unlock;
+
+		if (down_trylock(&b->io_mutex))
+			goto out_unlock;
+		up(&b->io_mutex);
+	}
+
+	mutex_lock(&b->write_lock);
+	if (btree_node_dirty(b))
+		__bch_btree_node_write(b, &cl);
+	mutex_unlock(&b->write_lock);
+
+	closure_sync(&cl);
+
+	/* wait for any in flight btree write */
+	down(&b->io_mutex);
+	up(&b->io_mutex);
+
+	return 0;
+out_unlock:
+	rw_unlock(true, b);
+	return -ENOMEM;
+}
+
+static unsigned long bch_mca_scan(struct shrinker *shrink,
+				  struct shrink_control *sc)
+{
+	struct cache_set *c = container_of(shrink, struct cache_set, shrink);
+	struct btree *b, *t;
+	unsigned long i, nr = sc->nr_to_scan;
+	unsigned long freed = 0;
+
+	if (c->shrinker_disabled)
+		return SHRINK_STOP;
+
+	if (c->btree_cache_alloc_lock)
+		return SHRINK_STOP;
+
+	/* Return -1 if we can't do anything right now */
+	if (sc->gfp_mask & __GFP_IO)
+		mutex_lock(&c->bucket_lock);
+	else if (!mutex_trylock(&c->bucket_lock))
+		return -1;
+
+	/*
+	 * It's _really_ critical that we don't free too many btree nodes - we
+	 * have to always leave ourselves a reserve. The reserve is how we
+	 * guarantee that allocating memory for a new btree node can always
+	 * succeed, so that inserting keys into the btree can always succeed and
+	 * IO can always make forward progress:
+	 */
+	nr /= c->btree_pages;
+	nr = min_t(unsigned long, nr, mca_can_free(c));
+
+	i = 0;
+	list_for_each_entry_safe(b, t, &c->btree_cache_freeable, list) {
+		if (freed >= nr)
+			break;
+
+		if (++i > 3 &&
+		    !mca_reap(b, 0, false)) {
+			mca_data_free(b);
+			rw_unlock(true, b);
+			freed++;
+		}
+	}
+
+	for (i = 0; (nr--) && i < c->btree_cache_used; i++) {
+		if (list_empty(&c->btree_cache))
+			goto out;
+
+		b = list_first_entry(&c->btree_cache, struct btree, list);
+		list_rotate_left(&c->btree_cache);
+
+		if (!b->accessed &&
+		    !mca_reap(b, 0, false)) {
+			mca_bucket_free(b);
+			mca_data_free(b);
+			rw_unlock(true, b);
+			freed++;
+		} else
+			b->accessed = 0;
+	}
+out:
+	mutex_unlock(&c->bucket_lock);
+	return freed;
+}
+
+static unsigned long bch_mca_count(struct shrinker *shrink,
+				   struct shrink_control *sc)
+{
+	struct cache_set *c = container_of(shrink, struct cache_set, shrink);
+
+	if (c->shrinker_disabled)
+		return 0;
+
+	if (c->btree_cache_alloc_lock)
+		return 0;
+
+	return mca_can_free(c) * c->btree_pages;
+}
+
+void bch_btree_cache_free(struct cache_set *c)
+{
+	struct btree *b;
+	struct closure cl;
+	closure_init_stack(&cl);
+
+	if (c->shrink.list.next)
+		unregister_shrinker(&c->shrink);
+
+	mutex_lock(&c->bucket_lock);
+
+#ifdef CONFIG_BCACHE_DEBUG
+	if (c->verify_data)
+		list_move(&c->verify_data->list, &c->btree_cache);
+
+	free_pages((unsigned long) c->verify_ondisk, ilog2(bucket_pages(c)));
+#endif
+
+	list_splice(&c->btree_cache_freeable,
+		    &c->btree_cache);
+
+	while (!list_empty(&c->btree_cache)) {
+		b = list_first_entry(&c->btree_cache, struct btree, list);
+
+		if (btree_node_dirty(b))
+			btree_complete_write(b, btree_current_write(b));
+		clear_bit(BTREE_NODE_dirty, &b->flags);
+
+		mca_data_free(b);
+	}
+
+	while (!list_empty(&c->btree_cache_freed)) {
+		b = list_first_entry(&c->btree_cache_freed,
+				     struct btree, list);
+		list_del(&b->list);
+		cancel_delayed_work_sync(&b->work);
+		kfree(b);
+	}
+
+	mutex_unlock(&c->bucket_lock);
+}
+
+int bch_btree_cache_alloc(struct cache_set *c)
+{
+	unsigned i;
+
+	for (i = 0; i < mca_reserve(c); i++)
+		if (!mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL))
+			return -ENOMEM;
+
+	list_splice_init(&c->btree_cache,
+			 &c->btree_cache_freeable);
+
+#ifdef CONFIG_BCACHE_DEBUG
+	mutex_init(&c->verify_lock);
+
+	c->verify_ondisk = (void *)
+		__get_free_pages(GFP_KERNEL, ilog2(bucket_pages(c)));
+
+	c->verify_data = mca_bucket_alloc(c, &ZERO_KEY, GFP_KERNEL);
+
+	if (c->verify_data &&
+	    c->verify_data->keys.set->data)
+		list_del_init(&c->verify_data->list);
+	else
+		c->verify_data = NULL;
+#endif
+
+	c->shrink.count_objects = bch_mca_count;
+	c->shrink.scan_objects = bch_mca_scan;
+	c->shrink.seeks = 4;
+	c->shrink.batch = c->btree_pages * 2;
+	register_shrinker(&c->shrink);
+
+	return 0;
+}
+
+/* Btree in memory cache - hash table */
+
+static struct hlist_head *mca_hash(struct cache_set *c, struct bkey *k)
+{
+	return &c->bucket_hash[hash_32(PTR_HASH(c, k), BUCKET_HASH_BITS)];
+}
+
+static struct btree *mca_find(struct cache_set *c, struct bkey *k)
+{
+	struct btree *b;
+
+	rcu_read_lock();
+	hlist_for_each_entry_rcu(b, mca_hash(c, k), hash)
+		if (PTR_HASH(c, &b->key) == PTR_HASH(c, k))
+			goto out;
+	b = NULL;
+out:
+	rcu_read_unlock();
+	return b;
+}
+
+static int mca_cannibalize_lock(struct cache_set *c, struct btree_op *op)
+{
+	struct task_struct *old;
+
+	old = cmpxchg(&c->btree_cache_alloc_lock, NULL, current);
+	if (old && old != current) {
+		if (op)
+			prepare_to_wait(&c->btree_cache_wait, &op->wait,
+					TASK_UNINTERRUPTIBLE);
+		return -EINTR;
+	}
+
+	return 0;
+}
+
+static struct btree *mca_cannibalize(struct cache_set *c, struct btree_op *op,
+				     struct bkey *k)
+{
+	struct btree *b;
+
+	trace_bcache_btree_cache_cannibalize(c);
+
+	if (mca_cannibalize_lock(c, op))
+		return ERR_PTR(-EINTR);
+
+	list_for_each_entry_reverse(b, &c->btree_cache, list)
+		if (!mca_reap(b, btree_order(k), false))
+			return b;
+
+	list_for_each_entry_reverse(b, &c->btree_cache, list)
+		if (!mca_reap(b, btree_order(k), true))
+			return b;
+
+	WARN(1, "btree cache cannibalize failed\n");
+	return ERR_PTR(-ENOMEM);
+}
+
+/*
+ * We can only have one thread cannibalizing other cached btree nodes at a time,
+ * or we'll deadlock. We use an open coded mutex to ensure that, which a
+ * cannibalize_bucket() will take. This means every time we unlock the root of
+ * the btree, we need to release this lock if we have it held.
+ */
+static void bch_cannibalize_unlock(struct cache_set *c)
+{
+	if (c->btree_cache_alloc_lock == current) {
+		c->btree_cache_alloc_lock = NULL;
+		wake_up(&c->btree_cache_wait);
+	}
+}
+
+static struct btree *mca_alloc(struct cache_set *c, struct btree_op *op,
+			       struct bkey *k, int level)
+{
+	struct btree *b;
+
+	BUG_ON(current->bio_list);
+
+	lockdep_assert_held(&c->bucket_lock);
+
+	if (mca_find(c, k))
+		return NULL;
+
+	/* btree_free() doesn't free memory; it sticks the node on the end of
+	 * the list. Check if there's any freed nodes there:
+	 */
+	list_for_each_entry(b, &c->btree_cache_freeable, list)
+		if (!mca_reap(b, btree_order(k), false))
+			goto out;
+
+	/* We never free struct btree itself, just the memory that holds the on
+	 * disk node. Check the freed list before allocating a new one:
+	 */
+	list_for_each_entry(b, &c->btree_cache_freed, list)
+		if (!mca_reap(b, 0, false)) {
+			mca_data_alloc(b, k, __GFP_NOWARN|GFP_NOIO);
+			if (!b->keys.set[0].data)
+				goto err;
+			else
+				goto out;
+		}
+
+	b = mca_bucket_alloc(c, k, __GFP_NOWARN|GFP_NOIO);
+	if (!b)
+		goto err;
+
+	BUG_ON(!down_write_trylock(&b->lock));
+	if (!b->keys.set->data)
+		goto err;
+out:
+	BUG_ON(b->io_mutex.count != 1);
+
+	bkey_copy(&b->key, k);
+	list_move(&b->list, &c->btree_cache);
+	hlist_del_init_rcu(&b->hash);
+	hlist_add_head_rcu(&b->hash, mca_hash(c, k));
+
+	lock_set_subclass(&b->lock.dep_map, level + 1, _THIS_IP_);
+	b->parent	= (void *) ~0UL;
+	b->flags	= 0;
+	b->written	= 0;
+	b->level	= level;
+
+	if (!b->level)
+		bch_btree_keys_init(&b->keys, &bch_extent_keys_ops,
+				    &b->c->expensive_debug_checks);
+	else
+		bch_btree_keys_init(&b->keys, &bch_btree_keys_ops,
+				    &b->c->expensive_debug_checks);
+
+	return b;
+err:
+	if (b)
+		rw_unlock(true, b);
+
+	b = mca_cannibalize(c, op, k);
+	if (!IS_ERR(b))
+		goto out;
+
+	return b;
+}
+
+/**
+ * bch_btree_node_get - find a btree node in the cache and lock it, reading it
+ * in from disk if necessary.
+ *
+ * If IO is necessary and running under generic_make_request, returns -EAGAIN.
+ *
+ * The btree node will have either a read or a write lock held, depending on
+ * level and op->lock.
+ */
+struct btree *bch_btree_node_get(struct cache_set *c, struct btree_op *op,
+				 struct bkey *k, int level, bool write)
+{
+	int i = 0;
+	struct btree *b;
+
+	BUG_ON(level < 0);
+retry:
+	b = mca_find(c, k);
+
+	if (!b) {
+		if (current->bio_list)
+			return ERR_PTR(-EAGAIN);
+
+		mutex_lock(&c->bucket_lock);
+		b = mca_alloc(c, op, k, level);
+		mutex_unlock(&c->bucket_lock);
+
+		if (!b)
+			goto retry;
+		if (IS_ERR(b))
+			return b;
+
+		bch_btree_node_read(b);
+
+		if (!write)
+			downgrade_write(&b->lock);
+	} else {
+		rw_lock(write, b, level);
+		if (PTR_HASH(c, &b->key) != PTR_HASH(c, k)) {
+			rw_unlock(write, b);
+			goto retry;
+		}
+		BUG_ON(b->level != level);
+	}
+
+	b->accessed = 1;
+
+	for (; i <= b->keys.nsets && b->keys.set[i].size; i++) {
+		prefetch(b->keys.set[i].tree);
+		prefetch(b->keys.set[i].data);
+	}
+
+	for (; i <= b->keys.nsets; i++)
+		prefetch(b->keys.set[i].data);
+
+	if (btree_node_io_error(b)) {
+		rw_unlock(write, b);
+		return ERR_PTR(-EIO);
+	}
+
+	BUG_ON(!b->written);
+
+	return b;
+}
+
+static void btree_node_prefetch(struct cache_set *c, struct bkey *k, int level)
+{
+	struct btree *b;
+
+	mutex_lock(&c->bucket_lock);
+	b = mca_alloc(c, NULL, k, level);
+	mutex_unlock(&c->bucket_lock);
+
+	if (!IS_ERR_OR_NULL(b)) {
+		bch_btree_node_read(b);
+		rw_unlock(true, b);
+	}
+}
+
+/* Btree alloc */
+
+static void btree_node_free(struct btree *b)
+{
+	trace_bcache_btree_node_free(b);
+
+	BUG_ON(b == b->c->root);
+
+	mutex_lock(&b->write_lock);
+
+	if (btree_node_dirty(b))
+		btree_complete_write(b, btree_current_write(b));
+	clear_bit(BTREE_NODE_dirty, &b->flags);
+
+	mutex_unlock(&b->write_lock);
+
+	cancel_delayed_work(&b->work);
+
+	mutex_lock(&b->c->bucket_lock);
+	bch_bucket_free(b->c, &b->key);
+	mca_bucket_free(b);
+	mutex_unlock(&b->c->bucket_lock);
+}
+
+struct btree *bch_btree_node_alloc(struct cache_set *c, struct btree_op *op,
+				   int level)
+{
+	BKEY_PADDED(key) k;
+	struct btree *b = ERR_PTR(-EAGAIN);
+
+	mutex_lock(&c->bucket_lock);
+retry:
+	if (__bch_bucket_alloc_set(c, RESERVE_BTREE, &k.key, 1, op != NULL))
+		goto err;
+
+	bkey_put(c, &k.key);
+	SET_KEY_SIZE(&k.key, c->btree_pages * PAGE_SECTORS);
+
+	b = mca_alloc(c, op, &k.key, level);
+	if (IS_ERR(b))
+		goto err_free;
+
+	if (!b) {
+		cache_bug(c,
+			"Tried to allocate bucket that was in btree cache");
+		goto retry;
+	}
+
+	b->accessed = 1;
+	bch_bset_init_next(&b->keys, b->keys.set->data, bset_magic(&b->c->sb));
+
+	mutex_unlock(&c->bucket_lock);
+
+	trace_bcache_btree_node_alloc(b);
+	return b;
+err_free:
+	bch_bucket_free(c, &k.key);
+err:
+	mutex_unlock(&c->bucket_lock);
+
+	trace_bcache_btree_node_alloc_fail(b);
+	return b;
+}
+
+static struct btree *btree_node_alloc_replacement(struct btree *b,
+						  struct btree_op *op)
+{
+	struct btree *n = bch_btree_node_alloc(b->c, op, b->level);
+	if (!IS_ERR_OR_NULL(n)) {
+		mutex_lock(&n->write_lock);
+		bch_btree_sort_into(&b->keys, &n->keys, &b->c->sort);
+		bkey_copy_key(&n->key, &b->key);
+		mutex_unlock(&n->write_lock);
+	}
+
+	return n;
+}
+
+static void make_btree_freeing_key(struct btree *b, struct bkey *k)
+{
+	unsigned i;
+
+	mutex_lock(&b->c->bucket_lock);
+
+	atomic_inc(&b->c->prio_blocked);
+
+	bkey_copy(k, &b->key);
+	bkey_copy_key(k, &ZERO_KEY);
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		SET_PTR_GEN(k, i,
+			    bch_inc_gen(PTR_CACHE(b->c, &b->key, i),
+					PTR_BUCKET(b->c, &b->key, i)));
+
+	mutex_unlock(&b->c->bucket_lock);
+}
+
+static int btree_check_reserve(struct btree *b, struct btree_op *op)
+{
+	struct cache_set *c = b->c;
+	struct cache *ca;
+	unsigned i, reserve = (c->root->level - b->level) * 2 + 1;
+
+	mutex_lock(&c->bucket_lock);
+
+	for_each_cache(ca, c, i)
+		if (fifo_used(&ca->free[RESERVE_BTREE]) < reserve) {
+			if (op)
+				prepare_to_wait(&c->btree_cache_wait, &op->wait,
+						TASK_UNINTERRUPTIBLE);
+			mutex_unlock(&c->bucket_lock);
+			return -EINTR;
+		}
+
+	mutex_unlock(&c->bucket_lock);
+
+	return mca_cannibalize_lock(b->c, op);
+}
+
+/* Garbage collection */
+
+static uint8_t __bch_btree_mark_key(struct cache_set *c, int level,
+				    struct bkey *k)
+{
+	uint8_t stale = 0;
+	unsigned i;
+	struct bucket *g;
+
+	/*
+	 * ptr_invalid() can't return true for the keys that mark btree nodes as
+	 * freed, but since ptr_bad() returns true we'll never actually use them
+	 * for anything and thus we don't want mark their pointers here
+	 */
+	if (!bkey_cmp(k, &ZERO_KEY))
+		return stale;
+
+	for (i = 0; i < KEY_PTRS(k); i++) {
+		if (!ptr_available(c, k, i))
+			continue;
+
+		g = PTR_BUCKET(c, k, i);
+
+		if (gen_after(g->last_gc, PTR_GEN(k, i)))
+			g->last_gc = PTR_GEN(k, i);
+
+		if (ptr_stale(c, k, i)) {
+			stale = max(stale, ptr_stale(c, k, i));
+			continue;
+		}
+
+		cache_bug_on(GC_MARK(g) &&
+			     (GC_MARK(g) == GC_MARK_METADATA) != (level != 0),
+			     c, "inconsistent ptrs: mark = %llu, level = %i",
+			     GC_MARK(g), level);
+
+		if (level)
+			SET_GC_MARK(g, GC_MARK_METADATA);
+		else if (KEY_DIRTY(k))
+			SET_GC_MARK(g, GC_MARK_DIRTY);
+		else if (!GC_MARK(g))
+			SET_GC_MARK(g, GC_MARK_RECLAIMABLE);
+
+		/* guard against overflow */
+		SET_GC_SECTORS_USED(g, min_t(unsigned,
+					     GC_SECTORS_USED(g) + KEY_SIZE(k),
+					     MAX_GC_SECTORS_USED));
+
+		BUG_ON(!GC_SECTORS_USED(g));
+	}
+
+	return stale;
+}
+
+#define btree_mark_key(b, k)	__bch_btree_mark_key(b->c, b->level, k)
+
+void bch_initial_mark_key(struct cache_set *c, int level, struct bkey *k)
+{
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (ptr_available(c, k, i) &&
+		    !ptr_stale(c, k, i)) {
+			struct bucket *b = PTR_BUCKET(c, k, i);
+
+			b->gen = PTR_GEN(k, i);
+
+			if (level && bkey_cmp(k, &ZERO_KEY))
+				b->prio = BTREE_PRIO;
+			else if (!level && b->prio == BTREE_PRIO)
+				b->prio = INITIAL_PRIO;
+		}
+
+	__bch_btree_mark_key(c, level, k);
+}
+
+static bool btree_gc_mark_node(struct btree *b, struct gc_stat *gc)
+{
+	uint8_t stale = 0;
+	unsigned keys = 0, good_keys = 0;
+	struct bkey *k;
+	struct btree_iter iter;
+	struct bset_tree *t;
+
+	gc->nodes++;
+
+	for_each_key_filter(&b->keys, k, &iter, bch_ptr_invalid) {
+		stale = max(stale, btree_mark_key(b, k));
+		keys++;
+
+		if (bch_ptr_bad(&b->keys, k))
+			continue;
+
+		gc->key_bytes += bkey_u64s(k);
+		gc->nkeys++;
+		good_keys++;
+
+		gc->data += KEY_SIZE(k);
+	}
+
+	for (t = b->keys.set; t <= &b->keys.set[b->keys.nsets]; t++)
+		btree_bug_on(t->size &&
+			     bset_written(&b->keys, t) &&
+			     bkey_cmp(&b->key, &t->end) < 0,
+			     b, "found short btree key in gc");
+
+	if (b->c->gc_always_rewrite)
+		return true;
+
+	if (stale > 10)
+		return true;
+
+	if ((keys - good_keys) * 2 > keys)
+		return true;
+
+	return false;
+}
+
+#define GC_MERGE_NODES	4U
+
+struct gc_merge_info {
+	struct btree	*b;
+	unsigned	keys;
+};
+
+static int bch_btree_insert_node(struct btree *, struct btree_op *,
+				 struct keylist *, atomic_t *, struct bkey *);
+
+static int btree_gc_coalesce(struct btree *b, struct btree_op *op,
+			     struct gc_stat *gc, struct gc_merge_info *r)
+{
+	unsigned i, nodes = 0, keys = 0, blocks;
+	struct btree *new_nodes[GC_MERGE_NODES];
+	struct keylist keylist;
+	struct closure cl;
+	struct bkey *k;
+
+	bch_keylist_init(&keylist);
+
+	if (btree_check_reserve(b, NULL))
+		return 0;
+
+	memset(new_nodes, 0, sizeof(new_nodes));
+	closure_init_stack(&cl);
+
+	while (nodes < GC_MERGE_NODES && !IS_ERR_OR_NULL(r[nodes].b))
+		keys += r[nodes++].keys;
+
+	blocks = btree_default_blocks(b->c) * 2 / 3;
+
+	if (nodes < 2 ||
+	    __set_blocks(b->keys.set[0].data, keys,
+			 block_bytes(b->c)) > blocks * (nodes - 1))
+		return 0;
+
+	for (i = 0; i < nodes; i++) {
+		new_nodes[i] = btree_node_alloc_replacement(r[i].b, NULL);
+		if (IS_ERR_OR_NULL(new_nodes[i]))
+			goto out_nocoalesce;
+	}
+
+	/*
+	 * We have to check the reserve here, after we've allocated our new
+	 * nodes, to make sure the insert below will succeed - we also check
+	 * before as an optimization to potentially avoid a bunch of expensive
+	 * allocs/sorts
+	 */
+	if (btree_check_reserve(b, NULL))
+		goto out_nocoalesce;
+
+	for (i = 0; i < nodes; i++)
+		mutex_lock(&new_nodes[i]->write_lock);
+
+	for (i = nodes - 1; i > 0; --i) {
+		struct bset *n1 = btree_bset_first(new_nodes[i]);
+		struct bset *n2 = btree_bset_first(new_nodes[i - 1]);
+		struct bkey *k, *last = NULL;
+
+		keys = 0;
+
+		if (i > 1) {
+			for (k = n2->start;
+			     k < bset_bkey_last(n2);
+			     k = bkey_next(k)) {
+				if (__set_blocks(n1, n1->keys + keys +
+						 bkey_u64s(k),
+						 block_bytes(b->c)) > blocks)
+					break;
+
+				last = k;
+				keys += bkey_u64s(k);
+			}
+		} else {
+			/*
+			 * Last node we're not getting rid of - we're getting
+			 * rid of the node at r[0]. Have to try and fit all of
+			 * the remaining keys into this node; we can't ensure
+			 * they will always fit due to rounding and variable
+			 * length keys (shouldn't be possible in practice,
+			 * though)
+			 */
+			if (__set_blocks(n1, n1->keys + n2->keys,
+					 block_bytes(b->c)) >
+			    btree_blocks(new_nodes[i]))
+				goto out_nocoalesce;
+
+			keys = n2->keys;
+			/* Take the key of the node we're getting rid of */
+			last = &r->b->key;
+		}
+
+		BUG_ON(__set_blocks(n1, n1->keys + keys, block_bytes(b->c)) >
+		       btree_blocks(new_nodes[i]));
+
+		if (last)
+			bkey_copy_key(&new_nodes[i]->key, last);
+
+		memcpy(bset_bkey_last(n1),
+		       n2->start,
+		       (void *) bset_bkey_idx(n2, keys) - (void *) n2->start);
+
+		n1->keys += keys;
+		r[i].keys = n1->keys;
+
+		memmove(n2->start,
+			bset_bkey_idx(n2, keys),
+			(void *) bset_bkey_last(n2) -
+			(void *) bset_bkey_idx(n2, keys));
+
+		n2->keys -= keys;
+
+		if (__bch_keylist_realloc(&keylist,
+					  bkey_u64s(&new_nodes[i]->key)))
+			goto out_nocoalesce;
+
+		bch_btree_node_write(new_nodes[i], &cl);
+		bch_keylist_add(&keylist, &new_nodes[i]->key);
+	}
+
+	for (i = 0; i < nodes; i++)
+		mutex_unlock(&new_nodes[i]->write_lock);
+
+	closure_sync(&cl);
+
+	/* We emptied out this node */
+	BUG_ON(btree_bset_first(new_nodes[0])->keys);
+	btree_node_free(new_nodes[0]);
+	rw_unlock(true, new_nodes[0]);
+
+	for (i = 0; i < nodes; i++) {
+		if (__bch_keylist_realloc(&keylist, bkey_u64s(&r[i].b->key)))
+			goto out_nocoalesce;
+
+		make_btree_freeing_key(r[i].b, keylist.top);
+		bch_keylist_push(&keylist);
+	}
+
+	bch_btree_insert_node(b, op, &keylist, NULL, NULL);
+	BUG_ON(!bch_keylist_empty(&keylist));
+
+	for (i = 0; i < nodes; i++) {
+		btree_node_free(r[i].b);
+		rw_unlock(true, r[i].b);
+
+		r[i].b = new_nodes[i];
+	}
+
+	memmove(r, r + 1, sizeof(r[0]) * (nodes - 1));
+	r[nodes - 1].b = ERR_PTR(-EINTR);
+
+	trace_bcache_btree_gc_coalesce(nodes);
+	gc->nodes--;
+
+	bch_keylist_free(&keylist);
+
+	/* Invalidated our iterator */
+	return -EINTR;
+
+out_nocoalesce:
+	closure_sync(&cl);
+	bch_keylist_free(&keylist);
+
+	while ((k = bch_keylist_pop(&keylist)))
+		if (!bkey_cmp(k, &ZERO_KEY))
+			atomic_dec(&b->c->prio_blocked);
+
+	for (i = 0; i < nodes; i++)
+		if (!IS_ERR_OR_NULL(new_nodes[i])) {
+			btree_node_free(new_nodes[i]);
+			rw_unlock(true, new_nodes[i]);
+		}
+	return 0;
+}
+
+static int btree_gc_rewrite_node(struct btree *b, struct btree_op *op,
+				 struct btree *replace)
+{
+	struct keylist keys;
+	struct btree *n;
+
+	if (btree_check_reserve(b, NULL))
+		return 0;
+
+	n = btree_node_alloc_replacement(replace, NULL);
+
+	/* recheck reserve after allocating replacement node */
+	if (btree_check_reserve(b, NULL)) {
+		btree_node_free(n);
+		rw_unlock(true, n);
+		return 0;
+	}
+
+	bch_btree_node_write_sync(n);
+
+	bch_keylist_init(&keys);
+	bch_keylist_add(&keys, &n->key);
+
+	make_btree_freeing_key(replace, keys.top);
+	bch_keylist_push(&keys);
+
+	bch_btree_insert_node(b, op, &keys, NULL, NULL);
+	BUG_ON(!bch_keylist_empty(&keys));
+
+	btree_node_free(replace);
+	rw_unlock(true, n);
+
+	/* Invalidated our iterator */
+	return -EINTR;
+}
+
+static unsigned btree_gc_count_keys(struct btree *b)
+{
+	struct bkey *k;
+	struct btree_iter iter;
+	unsigned ret = 0;
+
+	for_each_key_filter(&b->keys, k, &iter, bch_ptr_bad)
+		ret += bkey_u64s(k);
+
+	return ret;
+}
+
+static int btree_gc_recurse(struct btree *b, struct btree_op *op,
+			    struct closure *writes, struct gc_stat *gc)
+{
+	int ret = 0;
+	bool should_rewrite;
+	struct bkey *k;
+	struct btree_iter iter;
+	struct gc_merge_info r[GC_MERGE_NODES];
+	struct gc_merge_info *i, *last = r + ARRAY_SIZE(r) - 1;
+
+	bch_btree_iter_init(&b->keys, &iter, &b->c->gc_done);
+
+	for (i = r; i < r + ARRAY_SIZE(r); i++)
+		i->b = ERR_PTR(-EINTR);
+
+	while (1) {
+		k = bch_btree_iter_next_filter(&iter, &b->keys, bch_ptr_bad);
+		if (k) {
+			r->b = bch_btree_node_get(b->c, op, k, b->level - 1,
+						  true);
+			if (IS_ERR(r->b)) {
+				ret = PTR_ERR(r->b);
+				break;
+			}
+
+			r->keys = btree_gc_count_keys(r->b);
+
+			ret = btree_gc_coalesce(b, op, gc, r);
+			if (ret)
+				break;
+		}
+
+		if (!last->b)
+			break;
+
+		if (!IS_ERR(last->b)) {
+			should_rewrite = btree_gc_mark_node(last->b, gc);
+			if (should_rewrite) {
+				ret = btree_gc_rewrite_node(b, op, last->b);
+				if (ret)
+					break;
+			}
+
+			if (last->b->level) {
+				ret = btree_gc_recurse(last->b, op, writes, gc);
+				if (ret)
+					break;
+			}
+
+			bkey_copy_key(&b->c->gc_done, &last->b->key);
+
+			/*
+			 * Must flush leaf nodes before gc ends, since replace
+			 * operations aren't journalled
+			 */
+			mutex_lock(&last->b->write_lock);
+			if (btree_node_dirty(last->b))
+				bch_btree_node_write(last->b, writes);
+			mutex_unlock(&last->b->write_lock);
+			rw_unlock(true, last->b);
+		}
+
+		memmove(r + 1, r, sizeof(r[0]) * (GC_MERGE_NODES - 1));
+		r->b = NULL;
+
+		if (need_resched()) {
+			ret = -EAGAIN;
+			break;
+		}
+	}
+
+	for (i = r; i < r + ARRAY_SIZE(r); i++)
+		if (!IS_ERR_OR_NULL(i->b)) {
+			mutex_lock(&i->b->write_lock);
+			if (btree_node_dirty(i->b))
+				bch_btree_node_write(i->b, writes);
+			mutex_unlock(&i->b->write_lock);
+			rw_unlock(true, i->b);
+		}
+
+	return ret;
+}
+
+static int bch_btree_gc_root(struct btree *b, struct btree_op *op,
+			     struct closure *writes, struct gc_stat *gc)
+{
+	struct btree *n = NULL;
+	int ret = 0;
+	bool should_rewrite;
+
+	should_rewrite = btree_gc_mark_node(b, gc);
+	if (should_rewrite) {
+		n = btree_node_alloc_replacement(b, NULL);
+
+		if (!IS_ERR_OR_NULL(n)) {
+			bch_btree_node_write_sync(n);
+
+			bch_btree_set_root(n);
+			btree_node_free(b);
+			rw_unlock(true, n);
+
+			return -EINTR;
+		}
+	}
+
+	__bch_btree_mark_key(b->c, b->level + 1, &b->key);
+
+	if (b->level) {
+		ret = btree_gc_recurse(b, op, writes, gc);
+		if (ret)
+			return ret;
+	}
+
+	bkey_copy_key(&b->c->gc_done, &b->key);
+
+	return ret;
+}
+
+static void btree_gc_start(struct cache_set *c)
+{
+	struct cache *ca;
+	struct bucket *b;
+	unsigned i;
+
+	if (!c->gc_mark_valid)
+		return;
+
+	mutex_lock(&c->bucket_lock);
+
+	c->gc_mark_valid = 0;
+	c->gc_done = ZERO_KEY;
+
+	for_each_cache(ca, c, i)
+		for_each_bucket(b, ca) {
+			b->last_gc = b->gen;
+			if (!atomic_read(&b->pin)) {
+				SET_GC_MARK(b, 0);
+				SET_GC_SECTORS_USED(b, 0);
+			}
+		}
+
+	mutex_unlock(&c->bucket_lock);
+}
+
+static size_t bch_btree_gc_finish(struct cache_set *c)
+{
+	size_t available = 0;
+	struct bucket *b;
+	struct cache *ca;
+	unsigned i;
+
+	mutex_lock(&c->bucket_lock);
+
+	set_gc_sectors(c);
+	c->gc_mark_valid = 1;
+	c->need_gc	= 0;
+
+	for (i = 0; i < KEY_PTRS(&c->uuid_bucket); i++)
+		SET_GC_MARK(PTR_BUCKET(c, &c->uuid_bucket, i),
+			    GC_MARK_METADATA);
+
+	/* don't reclaim buckets to which writeback keys point */
+	rcu_read_lock();
+	for (i = 0; i < c->nr_uuids; i++) {
+		struct bcache_device *d = c->devices[i];
+		struct cached_dev *dc;
+		struct keybuf_key *w, *n;
+		unsigned j;
+
+		if (!d || UUID_FLASH_ONLY(&c->uuids[i]))
+			continue;
+		dc = container_of(d, struct cached_dev, disk);
+
+		spin_lock(&dc->writeback_keys.lock);
+		rbtree_postorder_for_each_entry_safe(w, n,
+					&dc->writeback_keys.keys, node)
+			for (j = 0; j < KEY_PTRS(&w->key); j++)
+				SET_GC_MARK(PTR_BUCKET(c, &w->key, j),
+					    GC_MARK_DIRTY);
+		spin_unlock(&dc->writeback_keys.lock);
+	}
+	rcu_read_unlock();
+
+	for_each_cache(ca, c, i) {
+		uint64_t *i;
+
+		ca->invalidate_needs_gc = 0;
+
+		for (i = ca->sb.d; i < ca->sb.d + ca->sb.keys; i++)
+			SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
+
+		for (i = ca->prio_buckets;
+		     i < ca->prio_buckets + prio_buckets(ca) * 2; i++)
+			SET_GC_MARK(ca->buckets + *i, GC_MARK_METADATA);
+
+		for_each_bucket(b, ca) {
+			c->need_gc	= max(c->need_gc, bucket_gc_gen(b));
+
+			if (atomic_read(&b->pin))
+				continue;
+
+			BUG_ON(!GC_MARK(b) && GC_SECTORS_USED(b));
+
+			if (!GC_MARK(b) || GC_MARK(b) == GC_MARK_RECLAIMABLE)
+				available++;
+		}
+	}
+
+	mutex_unlock(&c->bucket_lock);
+	return available;
+}
+
+static void bch_btree_gc(struct cache_set *c)
+{
+	int ret;
+	unsigned long available;
+	struct gc_stat stats;
+	struct closure writes;
+	struct btree_op op;
+	uint64_t start_time = local_clock();
+
+	trace_bcache_gc_start(c);
+
+	memset(&stats, 0, sizeof(struct gc_stat));
+	closure_init_stack(&writes);
+	bch_btree_op_init(&op, SHRT_MAX);
+
+	btree_gc_start(c);
+
+	do {
+		ret = btree_root(gc_root, c, &op, &writes, &stats);
+		closure_sync(&writes);
+
+		if (ret && ret != -EAGAIN)
+			pr_warn("gc failed!");
+	} while (ret);
+
+	available = bch_btree_gc_finish(c);
+	wake_up_allocators(c);
+
+	bch_time_stats_update(&c->btree_gc_time, start_time);
+
+	stats.key_bytes *= sizeof(uint64_t);
+	stats.data	<<= 9;
+	stats.in_use	= (c->nbuckets - available) * 100 / c->nbuckets;
+	memcpy(&c->gc_stats, &stats, sizeof(struct gc_stat));
+
+	trace_bcache_gc_end(c);
+
+	bch_moving_gc(c);
+}
+
+static int bch_gc_thread(void *arg)
+{
+	struct cache_set *c = arg;
+	struct cache *ca;
+	unsigned i;
+
+	while (1) {
+again:
+		bch_btree_gc(c);
+
+		set_current_state(TASK_INTERRUPTIBLE);
+		if (kthread_should_stop())
+			break;
+
+		mutex_lock(&c->bucket_lock);
+
+		for_each_cache(ca, c, i)
+			if (ca->invalidate_needs_gc) {
+				mutex_unlock(&c->bucket_lock);
+				set_current_state(TASK_RUNNING);
+				goto again;
+			}
+
+		mutex_unlock(&c->bucket_lock);
+
+		try_to_freeze();
+		schedule();
+	}
+
+	return 0;
+}
+
+int bch_gc_thread_start(struct cache_set *c)
+{
+	c->gc_thread = kthread_create(bch_gc_thread, c, "bcache_gc");
+	if (IS_ERR(c->gc_thread))
+		return PTR_ERR(c->gc_thread);
+
+	set_task_state(c->gc_thread, TASK_INTERRUPTIBLE);
+	return 0;
+}
+
+/* Initial partial gc */
+
+static int bch_btree_check_recurse(struct btree *b, struct btree_op *op)
+{
+	int ret = 0;
+	struct bkey *k, *p = NULL;
+	struct btree_iter iter;
+
+	for_each_key_filter(&b->keys, k, &iter, bch_ptr_invalid)
+		bch_initial_mark_key(b->c, b->level, k);
+
+	bch_initial_mark_key(b->c, b->level + 1, &b->key);
+
+	if (b->level) {
+		bch_btree_iter_init(&b->keys, &iter, NULL);
+
+		do {
+			k = bch_btree_iter_next_filter(&iter, &b->keys,
+						       bch_ptr_bad);
+			if (k)
+				btree_node_prefetch(b->c, k, b->level - 1);
+
+			if (p)
+				ret = btree(check_recurse, p, b, op);
+
+			p = k;
+		} while (p && !ret);
+	}
+
+	return ret;
+}
+
+int bch_btree_check(struct cache_set *c)
+{
+	struct btree_op op;
+
+	bch_btree_op_init(&op, SHRT_MAX);
+
+	return btree_root(check_recurse, c, &op);
+}
+
+void bch_initial_gc_finish(struct cache_set *c)
+{
+	struct cache *ca;
+	struct bucket *b;
+	unsigned i;
+
+	bch_btree_gc_finish(c);
+
+	mutex_lock(&c->bucket_lock);
+
+	/*
+	 * We need to put some unused buckets directly on the prio freelist in
+	 * order to get the allocator thread started - it needs freed buckets in
+	 * order to rewrite the prios and gens, and it needs to rewrite prios
+	 * and gens in order to free buckets.
+	 *
+	 * This is only safe for buckets that have no live data in them, which
+	 * there should always be some of.
+	 */
+	for_each_cache(ca, c, i) {
+		for_each_bucket(b, ca) {
+			if (fifo_full(&ca->free[RESERVE_PRIO]))
+				break;
+
+			if (bch_can_invalidate_bucket(ca, b) &&
+			    !GC_MARK(b)) {
+				__bch_invalidate_one_bucket(ca, b);
+				fifo_push(&ca->free[RESERVE_PRIO],
+					  b - ca->buckets);
+			}
+		}
+	}
+
+	mutex_unlock(&c->bucket_lock);
+}
+
+/* Btree insertion */
+
+static bool btree_insert_key(struct btree *b, struct bkey *k,
+			     struct bkey *replace_key)
+{
+	unsigned status;
+
+	BUG_ON(bkey_cmp(k, &b->key) > 0);
+
+	status = bch_btree_insert_key(&b->keys, k, replace_key);
+	if (status != BTREE_INSERT_STATUS_NO_INSERT) {
+		bch_check_keys(&b->keys, "%u for %s", status,
+			       replace_key ? "replace" : "insert");
+
+		trace_bcache_btree_insert_key(b, k, replace_key != NULL,
+					      status);
+		return true;
+	} else
+		return false;
+}
+
+static size_t insert_u64s_remaining(struct btree *b)
+{
+	long ret = bch_btree_keys_u64s_remaining(&b->keys);
+
+	/*
+	 * Might land in the middle of an existing extent and have to split it
+	 */
+	if (b->keys.ops->is_extents)
+		ret -= KEY_MAX_U64S;
+
+	return max(ret, 0L);
+}
+
+static bool bch_btree_insert_keys(struct btree *b, struct btree_op *op,
+				  struct keylist *insert_keys,
+				  struct bkey *replace_key)
+{
+	bool ret = false;
+	int oldsize = bch_count_data(&b->keys);
+
+	while (!bch_keylist_empty(insert_keys)) {
+		struct bkey *k = insert_keys->keys;
+
+		if (bkey_u64s(k) > insert_u64s_remaining(b))
+			break;
+
+		if (bkey_cmp(k, &b->key) <= 0) {
+			if (!b->level)
+				bkey_put(b->c, k);
+
+			ret |= btree_insert_key(b, k, replace_key);
+			bch_keylist_pop_front(insert_keys);
+		} else if (bkey_cmp(&START_KEY(k), &b->key) < 0) {
+			BKEY_PADDED(key) temp;
+			bkey_copy(&temp.key, insert_keys->keys);
+
+			bch_cut_back(&b->key, &temp.key);
+			bch_cut_front(&b->key, insert_keys->keys);
+
+			ret |= btree_insert_key(b, &temp.key, replace_key);
+			break;
+		} else {
+			break;
+		}
+	}
+
+	if (!ret)
+		op->insert_collision = true;
+
+	BUG_ON(!bch_keylist_empty(insert_keys) && b->level);
+
+	BUG_ON(bch_count_data(&b->keys) < oldsize);
+	return ret;
+}
+
+static int btree_split(struct btree *b, struct btree_op *op,
+		       struct keylist *insert_keys,
+		       struct bkey *replace_key)
+{
+	bool split;
+	struct btree *n1, *n2 = NULL, *n3 = NULL;
+	uint64_t start_time = local_clock();
+	struct closure cl;
+	struct keylist parent_keys;
+
+	closure_init_stack(&cl);
+	bch_keylist_init(&parent_keys);
+
+	if (btree_check_reserve(b, op)) {
+		if (!b->level)
+			return -EINTR;
+		else
+			WARN(1, "insufficient reserve for split\n");
+	}
+
+	n1 = btree_node_alloc_replacement(b, op);
+	if (IS_ERR(n1))
+		goto err;
+
+	split = set_blocks(btree_bset_first(n1),
+			   block_bytes(n1->c)) > (btree_blocks(b) * 4) / 5;
+
+	if (split) {
+		unsigned keys = 0;
+
+		trace_bcache_btree_node_split(b, btree_bset_first(n1)->keys);
+
+		n2 = bch_btree_node_alloc(b->c, op, b->level);
+		if (IS_ERR(n2))
+			goto err_free1;
+
+		if (!b->parent) {
+			n3 = bch_btree_node_alloc(b->c, op, b->level + 1);
+			if (IS_ERR(n3))
+				goto err_free2;
+		}
+
+		mutex_lock(&n1->write_lock);
+		mutex_lock(&n2->write_lock);
+
+		bch_btree_insert_keys(n1, op, insert_keys, replace_key);
+
+		/*
+		 * Has to be a linear search because we don't have an auxiliary
+		 * search tree yet
+		 */
+
+		while (keys < (btree_bset_first(n1)->keys * 3) / 5)
+			keys += bkey_u64s(bset_bkey_idx(btree_bset_first(n1),
+							keys));
+
+		bkey_copy_key(&n1->key,
+			      bset_bkey_idx(btree_bset_first(n1), keys));
+		keys += bkey_u64s(bset_bkey_idx(btree_bset_first(n1), keys));
+
+		btree_bset_first(n2)->keys = btree_bset_first(n1)->keys - keys;
+		btree_bset_first(n1)->keys = keys;
+
+		memcpy(btree_bset_first(n2)->start,
+		       bset_bkey_last(btree_bset_first(n1)),
+		       btree_bset_first(n2)->keys * sizeof(uint64_t));
+
+		bkey_copy_key(&n2->key, &b->key);
+
+		bch_keylist_add(&parent_keys, &n2->key);
+		bch_btree_node_write(n2, &cl);
+		mutex_unlock(&n2->write_lock);
+		rw_unlock(true, n2);
+	} else {
+		trace_bcache_btree_node_compact(b, btree_bset_first(n1)->keys);
+
+		mutex_lock(&n1->write_lock);
+		bch_btree_insert_keys(n1, op, insert_keys, replace_key);
+	}
+
+	bch_keylist_add(&parent_keys, &n1->key);
+	bch_btree_node_write(n1, &cl);
+	mutex_unlock(&n1->write_lock);
+
+	if (n3) {
+		/* Depth increases, make a new root */
+		mutex_lock(&n3->write_lock);
+		bkey_copy_key(&n3->key, &MAX_KEY);
+		bch_btree_insert_keys(n3, op, &parent_keys, NULL);
+		bch_btree_node_write(n3, &cl);
+		mutex_unlock(&n3->write_lock);
+
+		closure_sync(&cl);
+		bch_btree_set_root(n3);
+		rw_unlock(true, n3);
+	} else if (!b->parent) {
+		/* Root filled up but didn't need to be split */
+		closure_sync(&cl);
+		bch_btree_set_root(n1);
+	} else {
+		/* Split a non root node */
+		closure_sync(&cl);
+		make_btree_freeing_key(b, parent_keys.top);
+		bch_keylist_push(&parent_keys);
+
+		bch_btree_insert_node(b->parent, op, &parent_keys, NULL, NULL);
+		BUG_ON(!bch_keylist_empty(&parent_keys));
+	}
+
+	btree_node_free(b);
+	rw_unlock(true, n1);
+
+	bch_time_stats_update(&b->c->btree_split_time, start_time);
+
+	return 0;
+err_free2:
+	bkey_put(b->c, &n2->key);
+	btree_node_free(n2);
+	rw_unlock(true, n2);
+err_free1:
+	bkey_put(b->c, &n1->key);
+	btree_node_free(n1);
+	rw_unlock(true, n1);
+err:
+	WARN(1, "bcache: btree split failed (level %u)", b->level);
+
+	if (n3 == ERR_PTR(-EAGAIN) ||
+	    n2 == ERR_PTR(-EAGAIN) ||
+	    n1 == ERR_PTR(-EAGAIN))
+		return -EAGAIN;
+
+	return -ENOMEM;
+}
+
+static int bch_btree_insert_node(struct btree *b, struct btree_op *op,
+				 struct keylist *insert_keys,
+				 atomic_t *journal_ref,
+				 struct bkey *replace_key)
+{
+	struct closure cl;
+
+	BUG_ON(b->level && replace_key);
+
+	closure_init_stack(&cl);
+
+	mutex_lock(&b->write_lock);
+
+	if (write_block(b) != btree_bset_last(b) &&
+	    b->keys.last_set_unwritten)
+		bch_btree_init_next(b); /* just wrote a set */
+
+	if (bch_keylist_nkeys(insert_keys) > insert_u64s_remaining(b)) {
+		mutex_unlock(&b->write_lock);
+		goto split;
+	}
+
+	BUG_ON(write_block(b) != btree_bset_last(b));
+
+	if (bch_btree_insert_keys(b, op, insert_keys, replace_key)) {
+		if (!b->level)
+			bch_btree_leaf_dirty(b, journal_ref);
+		else
+			bch_btree_node_write(b, &cl);
+	}
+
+	mutex_unlock(&b->write_lock);
+
+	/* wait for btree node write if necessary, after unlock */
+	closure_sync(&cl);
+
+	return 0;
+split:
+	if (current->bio_list) {
+		op->lock = b->c->root->level + 1;
+		return -EAGAIN;
+	} else if (op->lock <= b->c->root->level) {
+		op->lock = b->c->root->level + 1;
+		return -EINTR;
+	} else {
+		/* Invalidated all iterators */
+		int ret = btree_split(b, op, insert_keys, replace_key);
+
+		if (bch_keylist_empty(insert_keys))
+			return 0;
+		else if (!ret)
+			return -EINTR;
+		return ret;
+	}
+}
+
+int bch_btree_insert_check_key(struct btree *b, struct btree_op *op,
+			       struct bkey *check_key)
+{
+	int ret = -EINTR;
+	uint64_t btree_ptr = b->key.ptr[0];
+	unsigned long seq = b->seq;
+	struct keylist insert;
+	bool upgrade = op->lock == -1;
+
+	bch_keylist_init(&insert);
+
+	if (upgrade) {
+		rw_unlock(false, b);
+		rw_lock(true, b, b->level);
+
+		if (b->key.ptr[0] != btree_ptr ||
+		    b->seq != seq + 1)
+			goto out;
+	}
+
+	SET_KEY_PTRS(check_key, 1);
+	get_random_bytes(&check_key->ptr[0], sizeof(uint64_t));
+
+	SET_PTR_DEV(check_key, 0, PTR_CHECK_DEV);
+
+	bch_keylist_add(&insert, check_key);
+
+	ret = bch_btree_insert_node(b, op, &insert, NULL, NULL);
+
+	BUG_ON(!ret && !bch_keylist_empty(&insert));
+out:
+	if (upgrade)
+		downgrade_write(&b->lock);
+	return ret;
+}
+
+struct btree_insert_op {
+	struct btree_op	op;
+	struct keylist	*keys;
+	atomic_t	*journal_ref;
+	struct bkey	*replace_key;
+};
+
+static int btree_insert_fn(struct btree_op *b_op, struct btree *b)
+{
+	struct btree_insert_op *op = container_of(b_op,
+					struct btree_insert_op, op);
+
+	int ret = bch_btree_insert_node(b, &op->op, op->keys,
+					op->journal_ref, op->replace_key);
+	if (ret && !bch_keylist_empty(op->keys))
+		return ret;
+	else
+		return MAP_DONE;
+}
+
+int bch_btree_insert(struct cache_set *c, struct keylist *keys,
+		     atomic_t *journal_ref, struct bkey *replace_key)
+{
+	struct btree_insert_op op;
+	int ret = 0;
+
+	BUG_ON(current->bio_list);
+	BUG_ON(bch_keylist_empty(keys));
+
+	bch_btree_op_init(&op.op, 0);
+	op.keys		= keys;
+	op.journal_ref	= journal_ref;
+	op.replace_key	= replace_key;
+
+	while (!ret && !bch_keylist_empty(keys)) {
+		op.op.lock = 0;
+		ret = bch_btree_map_leaf_nodes(&op.op, c,
+					       &START_KEY(keys->keys),
+					       btree_insert_fn);
+	}
+
+	if (ret) {
+		struct bkey *k;
+
+		pr_err("error %i", ret);
+
+		while ((k = bch_keylist_pop(keys)))
+			bkey_put(c, k);
+	} else if (op.op.insert_collision)
+		ret = -ESRCH;
+
+	return ret;
+}
+
+void bch_btree_set_root(struct btree *b)
+{
+	unsigned i;
+	struct closure cl;
+
+	closure_init_stack(&cl);
+
+	trace_bcache_btree_set_root(b);
+
+	BUG_ON(!b->written);
+
+	for (i = 0; i < KEY_PTRS(&b->key); i++)
+		BUG_ON(PTR_BUCKET(b->c, &b->key, i)->prio != BTREE_PRIO);
+
+	mutex_lock(&b->c->bucket_lock);
+	list_del_init(&b->list);
+	mutex_unlock(&b->c->bucket_lock);
+
+	b->c->root = b;
+
+	bch_journal_meta(b->c, &cl);
+	closure_sync(&cl);
+}
+
+/* Map across nodes or keys */
+
+static int bch_btree_map_nodes_recurse(struct btree *b, struct btree_op *op,
+				       struct bkey *from,
+				       btree_map_nodes_fn *fn, int flags)
+{
+	int ret = MAP_CONTINUE;
+
+	if (b->level) {
+		struct bkey *k;
+		struct btree_iter iter;
+
+		bch_btree_iter_init(&b->keys, &iter, from);
+
+		while ((k = bch_btree_iter_next_filter(&iter, &b->keys,
+						       bch_ptr_bad))) {
+			ret = btree(map_nodes_recurse, k, b,
+				    op, from, fn, flags);
+			from = NULL;
+
+			if (ret != MAP_CONTINUE)
+				return ret;
+		}
+	}
+
+	if (!b->level || flags == MAP_ALL_NODES)
+		ret = fn(op, b);
+
+	return ret;
+}
+
+int __bch_btree_map_nodes(struct btree_op *op, struct cache_set *c,
+			  struct bkey *from, btree_map_nodes_fn *fn, int flags)
+{
+	return btree_root(map_nodes_recurse, c, op, from, fn, flags);
+}
+
+static int bch_btree_map_keys_recurse(struct btree *b, struct btree_op *op,
+				      struct bkey *from, btree_map_keys_fn *fn,
+				      int flags)
+{
+	int ret = MAP_CONTINUE;
+	struct bkey *k;
+	struct btree_iter iter;
+
+	bch_btree_iter_init(&b->keys, &iter, from);
+
+	while ((k = bch_btree_iter_next_filter(&iter, &b->keys, bch_ptr_bad))) {
+		ret = !b->level
+			? fn(op, b, k)
+			: btree(map_keys_recurse, k, b, op, from, fn, flags);
+		from = NULL;
+
+		if (ret != MAP_CONTINUE)
+			return ret;
+	}
+
+	if (!b->level && (flags & MAP_END_KEY))
+		ret = fn(op, b, &KEY(KEY_INODE(&b->key),
+				     KEY_OFFSET(&b->key), 0));
+
+	return ret;
+}
+
+int bch_btree_map_keys(struct btree_op *op, struct cache_set *c,
+		       struct bkey *from, btree_map_keys_fn *fn, int flags)
+{
+	return btree_root(map_keys_recurse, c, op, from, fn, flags);
+}
+
+/* Keybuf code */
+
+static inline int keybuf_cmp(struct keybuf_key *l, struct keybuf_key *r)
+{
+	/* Overlapping keys compare equal */
+	if (bkey_cmp(&l->key, &START_KEY(&r->key)) <= 0)
+		return -1;
+	if (bkey_cmp(&START_KEY(&l->key), &r->key) >= 0)
+		return 1;
+	return 0;
+}
+
+static inline int keybuf_nonoverlapping_cmp(struct keybuf_key *l,
+					    struct keybuf_key *r)
+{
+	return clamp_t(int64_t, bkey_cmp(&l->key, &r->key), -1, 1);
+}
+
+struct refill {
+	struct btree_op	op;
+	unsigned	nr_found;
+	struct keybuf	*buf;
+	struct bkey	*end;
+	keybuf_pred_fn	*pred;
+};
+
+static int refill_keybuf_fn(struct btree_op *op, struct btree *b,
+			    struct bkey *k)
+{
+	struct refill *refill = container_of(op, struct refill, op);
+	struct keybuf *buf = refill->buf;
+	int ret = MAP_CONTINUE;
+
+	if (bkey_cmp(k, refill->end) >= 0) {
+		ret = MAP_DONE;
+		goto out;
+	}
+
+	if (!KEY_SIZE(k)) /* end key */
+		goto out;
+
+	if (refill->pred(buf, k)) {
+		struct keybuf_key *w;
+
+		spin_lock(&buf->lock);
+
+		w = array_alloc(&buf->freelist);
+		if (!w) {
+			spin_unlock(&buf->lock);
+			return MAP_DONE;
+		}
+
+		w->private = NULL;
+		bkey_copy(&w->key, k);
+
+		if (RB_INSERT(&buf->keys, w, node, keybuf_cmp))
+			array_free(&buf->freelist, w);
+		else
+			refill->nr_found++;
+
+		if (array_freelist_empty(&buf->freelist))
+			ret = MAP_DONE;
+
+		spin_unlock(&buf->lock);
+	}
+out:
+	buf->last_scanned = *k;
+	return ret;
+}
+
+void bch_refill_keybuf(struct cache_set *c, struct keybuf *buf,
+		       struct bkey *end, keybuf_pred_fn *pred)
+{
+	struct bkey start = buf->last_scanned;
+	struct refill refill;
+
+	cond_resched();
+
+	bch_btree_op_init(&refill.op, -1);
+	refill.nr_found	= 0;
+	refill.buf	= buf;
+	refill.end	= end;
+	refill.pred	= pred;
+
+	bch_btree_map_keys(&refill.op, c, &buf->last_scanned,
+			   refill_keybuf_fn, MAP_END_KEY);
+
+	trace_bcache_keyscan(refill.nr_found,
+			     KEY_INODE(&start), KEY_OFFSET(&start),
+			     KEY_INODE(&buf->last_scanned),
+			     KEY_OFFSET(&buf->last_scanned));
+
+	spin_lock(&buf->lock);
+
+	if (!RB_EMPTY_ROOT(&buf->keys)) {
+		struct keybuf_key *w;
+		w = RB_FIRST(&buf->keys, struct keybuf_key, node);
+		buf->start	= START_KEY(&w->key);
+
+		w = RB_LAST(&buf->keys, struct keybuf_key, node);
+		buf->end	= w->key;
+	} else {
+		buf->start	= MAX_KEY;
+		buf->end	= MAX_KEY;
+	}
+
+	spin_unlock(&buf->lock);
+}
+
+static void __bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
+{
+	rb_erase(&w->node, &buf->keys);
+	array_free(&buf->freelist, w);
+}
+
+void bch_keybuf_del(struct keybuf *buf, struct keybuf_key *w)
+{
+	spin_lock(&buf->lock);
+	__bch_keybuf_del(buf, w);
+	spin_unlock(&buf->lock);
+}
+
+bool bch_keybuf_check_overlapping(struct keybuf *buf, struct bkey *start,
+				  struct bkey *end)
+{
+	bool ret = false;
+	struct keybuf_key *p, *w, s;
+	s.key = *start;
+
+	if (bkey_cmp(end, &buf->start) <= 0 ||
+	    bkey_cmp(start, &buf->end) >= 0)
+		return false;
+
+	spin_lock(&buf->lock);
+	w = RB_GREATER(&buf->keys, s, node, keybuf_nonoverlapping_cmp);
+
+	while (w && bkey_cmp(&START_KEY(&w->key), end) < 0) {
+		p = w;
+		w = RB_NEXT(w, node);
+
+		if (p->private)
+			ret = true;
+		else
+			__bch_keybuf_del(buf, p);
+	}
+
+	spin_unlock(&buf->lock);
+	return ret;
+}
+
+struct keybuf_key *bch_keybuf_next(struct keybuf *buf)
+{
+	struct keybuf_key *w;
+	spin_lock(&buf->lock);
+
+	w = RB_FIRST(&buf->keys, struct keybuf_key, node);
+
+	while (w && w->private)
+		w = RB_NEXT(w, node);
+
+	if (w)
+		w->private = ERR_PTR(-EINTR);
+
+	spin_unlock(&buf->lock);
+	return w;
+}
+
+struct keybuf_key *bch_keybuf_next_rescan(struct cache_set *c,
+					  struct keybuf *buf,
+					  struct bkey *end,
+					  keybuf_pred_fn *pred)
+{
+	struct keybuf_key *ret;
+
+	while (1) {
+		ret = bch_keybuf_next(buf);
+		if (ret)
+			break;
+
+		if (bkey_cmp(&buf->last_scanned, end) >= 0) {
+			pr_debug("scan finished");
+			break;
+		}
+
+		bch_refill_keybuf(c, buf, end, pred);
+	}
+
+	return ret;
+}
+
+void bch_keybuf_init(struct keybuf *buf)
+{
+	buf->last_scanned	= MAX_KEY;
+	buf->keys		= RB_ROOT;
+
+	spin_lock_init(&buf->lock);
+	array_allocator_init(&buf->freelist);
+}
diff --git a/drivers/md/bcache/btree.h b/drivers/md/bcache/btree.h
new file mode 100644
index 00000000000..91dfa5e6968
--- /dev/null
+++ b/drivers/md/bcache/btree.h
@@ -0,0 +1,309 @@
+#ifndef _BCACHE_BTREE_H
+#define _BCACHE_BTREE_H
+
+/*
+ * THE BTREE:
+ *
+ * At a high level, bcache's btree is relatively standard b+ tree. All keys and
+ * pointers are in the leaves; interior nodes only have pointers to the child
+ * nodes.
+ *
+ * In the interior nodes, a struct bkey always points to a child btree node, and
+ * the key is the highest key in the child node - except that the highest key in
+ * an interior node is always MAX_KEY. The size field refers to the size on disk
+ * of the child node - this would allow us to have variable sized btree nodes
+ * (handy for keeping the depth of the btree 1 by expanding just the root).
+ *
+ * Btree nodes are themselves log structured, but this is hidden fairly
+ * thoroughly. Btree nodes on disk will in practice have extents that overlap
+ * (because they were written at different times), but in memory we never have
+ * overlapping extents - when we read in a btree node from disk, the first thing
+ * we do is resort all the sets of keys with a mergesort, and in the same pass
+ * we check for overlapping extents and adjust them appropriately.
+ *
+ * struct btree_op is a central interface to the btree code. It's used for
+ * specifying read vs. write locking, and the embedded closure is used for
+ * waiting on IO or reserve memory.
+ *
+ * BTREE CACHE:
+ *
+ * Btree nodes are cached in memory; traversing the btree might require reading
+ * in btree nodes which is handled mostly transparently.
+ *
+ * bch_btree_node_get() looks up a btree node in the cache and reads it in from
+ * disk if necessary. This function is almost never called directly though - the
+ * btree() macro is used to get a btree node, call some function on it, and
+ * unlock the node after the function returns.
+ *
+ * The root is special cased - it's taken out of the cache's lru (thus pinning
+ * it in memory), so we can find the root of the btree by just dereferencing a
+ * pointer instead of looking it up in the cache. This makes locking a bit
+ * tricky, since the root pointer is protected by the lock in the btree node it
+ * points to - the btree_root() macro handles this.
+ *
+ * In various places we must be able to allocate memory for multiple btree nodes
+ * in order to make forward progress. To do this we use the btree cache itself
+ * as a reserve; if __get_free_pages() fails, we'll find a node in the btree
+ * cache we can reuse. We can't allow more than one thread to be doing this at a
+ * time, so there's a lock, implemented by a pointer to the btree_op closure -
+ * this allows the btree_root() macro to implicitly release this lock.
+ *
+ * BTREE IO:
+ *
+ * Btree nodes never have to be explicitly read in; bch_btree_node_get() handles
+ * this.
+ *
+ * For writing, we have two btree_write structs embeddded in struct btree - one
+ * write in flight, and one being set up, and we toggle between them.
+ *
+ * Writing is done with a single function -  bch_btree_write() really serves two
+ * different purposes and should be broken up into two different functions. When
+ * passing now = false, it merely indicates that the node is now dirty - calling
+ * it ensures that the dirty keys will be written at some point in the future.
+ *
+ * When passing now = true, bch_btree_write() causes a write to happen
+ * "immediately" (if there was already a write in flight, it'll cause the write
+ * to happen as soon as the previous write completes). It returns immediately
+ * though - but it takes a refcount on the closure in struct btree_op you passed
+ * to it, so a closure_sync() later can be used to wait for the write to
+ * complete.
+ *
+ * This is handy because btree_split() and garbage collection can issue writes
+ * in parallel, reducing the amount of time they have to hold write locks.
+ *
+ * LOCKING:
+ *
+ * When traversing the btree, we may need write locks starting at some level -
+ * inserting a key into the btree will typically only require a write lock on
+ * the leaf node.
+ *
+ * This is specified with the lock field in struct btree_op; lock = 0 means we
+ * take write locks at level <= 0, i.e. only leaf nodes. bch_btree_node_get()
+ * checks this field and returns the node with the appropriate lock held.
+ *
+ * If, after traversing the btree, the insertion code discovers it has to split
+ * then it must restart from the root and take new locks - to do this it changes
+ * the lock field and returns -EINTR, which causes the btree_root() macro to
+ * loop.
+ *
+ * Handling cache misses require a different mechanism for upgrading to a write
+ * lock. We do cache lookups with only a read lock held, but if we get a cache
+ * miss and we wish to insert this data into the cache, we have to insert a
+ * placeholder key to detect races - otherwise, we could race with a write and
+ * overwrite the data that was just written to the cache with stale data from
+ * the backing device.
+ *
+ * For this we use a sequence number that write locks and unlocks increment - to
+ * insert the check key it unlocks the btree node and then takes a write lock,
+ * and fails if the sequence number doesn't match.
+ */
+
+#include "bset.h"
+#include "debug.h"
+
+struct btree_write {
+	atomic_t		*journal;
+
+	/* If btree_split() frees a btree node, it writes a new pointer to that
+	 * btree node indicating it was freed; it takes a refcount on
+	 * c->prio_blocked because we can't write the gens until the new
+	 * pointer is on disk. This allows btree_write_endio() to release the
+	 * refcount that btree_split() took.
+	 */
+	int			prio_blocked;
+};
+
+struct btree {
+	/* Hottest entries first */
+	struct hlist_node	hash;
+
+	/* Key/pointer for this btree node */
+	BKEY_PADDED(key);
+
+	/* Single bit - set when accessed, cleared by shrinker */
+	unsigned long		accessed;
+	unsigned long		seq;
+	struct rw_semaphore	lock;
+	struct cache_set	*c;
+	struct btree		*parent;
+
+	struct mutex		write_lock;
+
+	unsigned long		flags;
+	uint16_t		written;	/* would be nice to kill */
+	uint8_t			level;
+
+	struct btree_keys	keys;
+
+	/* For outstanding btree writes, used as a lock - protects write_idx */
+	struct closure		io;
+	struct semaphore	io_mutex;
+
+	struct list_head	list;
+	struct delayed_work	work;
+
+	struct btree_write	writes[2];
+	struct bio		*bio;
+};
+
+#define BTREE_FLAG(flag)						\
+static inline bool btree_node_ ## flag(struct btree *b)			\
+{	return test_bit(BTREE_NODE_ ## flag, &b->flags); }		\
+									\
+static inline void set_btree_node_ ## flag(struct btree *b)		\
+{	set_bit(BTREE_NODE_ ## flag, &b->flags); }			\
+
+enum btree_flags {
+	BTREE_NODE_io_error,
+	BTREE_NODE_dirty,
+	BTREE_NODE_write_idx,
+};
+
+BTREE_FLAG(io_error);
+BTREE_FLAG(dirty);
+BTREE_FLAG(write_idx);
+
+static inline struct btree_write *btree_current_write(struct btree *b)
+{
+	return b->writes + btree_node_write_idx(b);
+}
+
+static inline struct btree_write *btree_prev_write(struct btree *b)
+{
+	return b->writes + (btree_node_write_idx(b) ^ 1);
+}
+
+static inline struct bset *btree_bset_first(struct btree *b)
+{
+	return b->keys.set->data;
+}
+
+static inline struct bset *btree_bset_last(struct btree *b)
+{
+	return bset_tree_last(&b->keys)->data;
+}
+
+static inline unsigned bset_block_offset(struct btree *b, struct bset *i)
+{
+	return bset_sector_offset(&b->keys, i) >> b->c->block_bits;
+}
+
+static inline void set_gc_sectors(struct cache_set *c)
+{
+	atomic_set(&c->sectors_to_gc, c->sb.bucket_size * c->nbuckets / 16);
+}
+
+void bkey_put(struct cache_set *c, struct bkey *k);
+
+/* Looping macros */
+
+#define for_each_cached_btree(b, c, iter)				\
+	for (iter = 0;							\
+	     iter < ARRAY_SIZE((c)->bucket_hash);			\
+	     iter++)							\
+		hlist_for_each_entry_rcu((b), (c)->bucket_hash + iter, hash)
+
+/* Recursing down the btree */
+
+struct btree_op {
+	/* for waiting on btree reserve in btree_split() */
+	wait_queue_t		wait;
+
+	/* Btree level at which we start taking write locks */
+	short			lock;
+
+	unsigned		insert_collision:1;
+};
+
+static inline void bch_btree_op_init(struct btree_op *op, int write_lock_level)
+{
+	memset(op, 0, sizeof(struct btree_op));
+	init_wait(&op->wait);
+	op->lock = write_lock_level;
+}
+
+static inline void rw_lock(bool w, struct btree *b, int level)
+{
+	w ? down_write_nested(&b->lock, level + 1)
+	  : down_read_nested(&b->lock, level + 1);
+	if (w)
+		b->seq++;
+}
+
+static inline void rw_unlock(bool w, struct btree *b)
+{
+	if (w)
+		b->seq++;
+	(w ? up_write : up_read)(&b->lock);
+}
+
+void bch_btree_node_read_done(struct btree *);
+void __bch_btree_node_write(struct btree *, struct closure *);
+void bch_btree_node_write(struct btree *, struct closure *);
+
+void bch_btree_set_root(struct btree *);
+struct btree *bch_btree_node_alloc(struct cache_set *, struct btree_op *, int);
+struct btree *bch_btree_node_get(struct cache_set *, struct btree_op *,
+				 struct bkey *, int, bool);
+
+int bch_btree_insert_check_key(struct btree *, struct btree_op *,
+			       struct bkey *);
+int bch_btree_insert(struct cache_set *, struct keylist *,
+		     atomic_t *, struct bkey *);
+
+int bch_gc_thread_start(struct cache_set *);
+void bch_initial_gc_finish(struct cache_set *);
+void bch_moving_gc(struct cache_set *);
+int bch_btree_check(struct cache_set *);
+void bch_initial_mark_key(struct cache_set *, int, struct bkey *);
+
+static inline void wake_up_gc(struct cache_set *c)
+{
+	if (c->gc_thread)
+		wake_up_process(c->gc_thread);
+}
+
+#define MAP_DONE	0
+#define MAP_CONTINUE	1
+
+#define MAP_ALL_NODES	0
+#define MAP_LEAF_NODES	1
+
+#define MAP_END_KEY	1
+
+typedef int (btree_map_nodes_fn)(struct btree_op *, struct btree *);
+int __bch_btree_map_nodes(struct btree_op *, struct cache_set *,
+			  struct bkey *, btree_map_nodes_fn *, int);
+
+static inline int bch_btree_map_nodes(struct btree_op *op, struct cache_set *c,
+				      struct bkey *from, btree_map_nodes_fn *fn)
+{
+	return __bch_btree_map_nodes(op, c, from, fn, MAP_ALL_NODES);
+}
+
+static inline int bch_btree_map_leaf_nodes(struct btree_op *op,
+					   struct cache_set *c,
+					   struct bkey *from,
+					   btree_map_nodes_fn *fn)
+{
+	return __bch_btree_map_nodes(op, c, from, fn, MAP_LEAF_NODES);
+}
+
+typedef int (btree_map_keys_fn)(struct btree_op *, struct btree *,
+				struct bkey *);
+int bch_btree_map_keys(struct btree_op *, struct cache_set *,
+		       struct bkey *, btree_map_keys_fn *, int);
+
+typedef bool (keybuf_pred_fn)(struct keybuf *, struct bkey *);
+
+void bch_keybuf_init(struct keybuf *);
+void bch_refill_keybuf(struct cache_set *, struct keybuf *,
+		       struct bkey *, keybuf_pred_fn *);
+bool bch_keybuf_check_overlapping(struct keybuf *, struct bkey *,
+				  struct bkey *);
+void bch_keybuf_del(struct keybuf *, struct keybuf_key *);
+struct keybuf_key *bch_keybuf_next(struct keybuf *);
+struct keybuf_key *bch_keybuf_next_rescan(struct cache_set *, struct keybuf *,
+					  struct bkey *, keybuf_pred_fn *);
+
+#endif
diff --git a/drivers/md/bcache/closure.c b/drivers/md/bcache/closure.c
new file mode 100644
index 00000000000..7a228de95fd
--- /dev/null
+++ b/drivers/md/bcache/closure.c
@@ -0,0 +1,222 @@
+/*
+ * Asynchronous refcounty things
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include <linux/debugfs.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+
+#include "closure.h"
+
+static inline void closure_put_after_sub(struct closure *cl, int flags)
+{
+	int r = flags & CLOSURE_REMAINING_MASK;
+
+	BUG_ON(flags & CLOSURE_GUARD_MASK);
+	BUG_ON(!r && (flags & ~CLOSURE_DESTRUCTOR));
+
+	/* Must deliver precisely one wakeup */
+	if (r == 1 && (flags & CLOSURE_SLEEPING))
+		wake_up_process(cl->task);
+
+	if (!r) {
+		if (cl->fn && !(flags & CLOSURE_DESTRUCTOR)) {
+			atomic_set(&cl->remaining,
+				   CLOSURE_REMAINING_INITIALIZER);
+			closure_queue(cl);
+		} else {
+			struct closure *parent = cl->parent;
+			closure_fn *destructor = cl->fn;
+
+			closure_debug_destroy(cl);
+
+			if (destructor)
+				destructor(cl);
+
+			if (parent)
+				closure_put(parent);
+		}
+	}
+}
+
+/* For clearing flags with the same atomic op as a put */
+void closure_sub(struct closure *cl, int v)
+{
+	closure_put_after_sub(cl, atomic_sub_return(v, &cl->remaining));
+}
+EXPORT_SYMBOL(closure_sub);
+
+/**
+ * closure_put - decrement a closure's refcount
+ */
+void closure_put(struct closure *cl)
+{
+	closure_put_after_sub(cl, atomic_dec_return(&cl->remaining));
+}
+EXPORT_SYMBOL(closure_put);
+
+/**
+ * closure_wake_up - wake up all closures on a wait list, without memory barrier
+ */
+void __closure_wake_up(struct closure_waitlist *wait_list)
+{
+	struct llist_node *list;
+	struct closure *cl;
+	struct llist_node *reverse = NULL;
+
+	list = llist_del_all(&wait_list->list);
+
+	/* We first reverse the list to preserve FIFO ordering and fairness */
+
+	while (list) {
+		struct llist_node *t = list;
+		list = llist_next(list);
+
+		t->next = reverse;
+		reverse = t;
+	}
+
+	/* Then do the wakeups */
+
+	while (reverse) {
+		cl = container_of(reverse, struct closure, list);
+		reverse = llist_next(reverse);
+
+		closure_set_waiting(cl, 0);
+		closure_sub(cl, CLOSURE_WAITING + 1);
+	}
+}
+EXPORT_SYMBOL(__closure_wake_up);
+
+/**
+ * closure_wait - add a closure to a waitlist
+ *
+ * @waitlist will own a ref on @cl, which will be released when
+ * closure_wake_up() is called on @waitlist.
+ *
+ */
+bool closure_wait(struct closure_waitlist *waitlist, struct closure *cl)
+{
+	if (atomic_read(&cl->remaining) & CLOSURE_WAITING)
+		return false;
+
+	closure_set_waiting(cl, _RET_IP_);
+	atomic_add(CLOSURE_WAITING + 1, &cl->remaining);
+	llist_add(&cl->list, &waitlist->list);
+
+	return true;
+}
+EXPORT_SYMBOL(closure_wait);
+
+/**
+ * closure_sync - sleep until a closure a closure has nothing left to wait on
+ *
+ * Sleeps until the refcount hits 1 - the thread that's running the closure owns
+ * the last refcount.
+ */
+void closure_sync(struct closure *cl)
+{
+	while (1) {
+		__closure_start_sleep(cl);
+		closure_set_ret_ip(cl);
+
+		if ((atomic_read(&cl->remaining) &
+		     CLOSURE_REMAINING_MASK) == 1)
+			break;
+
+		schedule();
+	}
+
+	__closure_end_sleep(cl);
+}
+EXPORT_SYMBOL(closure_sync);
+
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+
+static LIST_HEAD(closure_list);
+static DEFINE_SPINLOCK(closure_list_lock);
+
+void closure_debug_create(struct closure *cl)
+{
+	unsigned long flags;
+
+	BUG_ON(cl->magic == CLOSURE_MAGIC_ALIVE);
+	cl->magic = CLOSURE_MAGIC_ALIVE;
+
+	spin_lock_irqsave(&closure_list_lock, flags);
+	list_add(&cl->all, &closure_list);
+	spin_unlock_irqrestore(&closure_list_lock, flags);
+}
+EXPORT_SYMBOL(closure_debug_create);
+
+void closure_debug_destroy(struct closure *cl)
+{
+	unsigned long flags;
+
+	BUG_ON(cl->magic != CLOSURE_MAGIC_ALIVE);
+	cl->magic = CLOSURE_MAGIC_DEAD;
+
+	spin_lock_irqsave(&closure_list_lock, flags);
+	list_del(&cl->all);
+	spin_unlock_irqrestore(&closure_list_lock, flags);
+}
+EXPORT_SYMBOL(closure_debug_destroy);
+
+static struct dentry *debug;
+
+#define work_data_bits(work) ((unsigned long *)(&(work)->data))
+
+static int debug_seq_show(struct seq_file *f, void *data)
+{
+	struct closure *cl;
+	spin_lock_irq(&closure_list_lock);
+
+	list_for_each_entry(cl, &closure_list, all) {
+		int r = atomic_read(&cl->remaining);
+
+		seq_printf(f, "%p: %pF -> %pf p %p r %i ",
+			   cl, (void *) cl->ip, cl->fn, cl->parent,
+			   r & CLOSURE_REMAINING_MASK);
+
+		seq_printf(f, "%s%s%s%s\n",
+			   test_bit(WORK_STRUCT_PENDING,
+				    work_data_bits(&cl->work)) ? "Q" : "",
+			   r & CLOSURE_RUNNING	? "R" : "",
+			   r & CLOSURE_STACK	? "S" : "",
+			   r & CLOSURE_SLEEPING	? "Sl" : "");
+
+		if (r & CLOSURE_WAITING)
+			seq_printf(f, " W %pF\n",
+				   (void *) cl->waiting_on);
+
+		seq_printf(f, "\n");
+	}
+
+	spin_unlock_irq(&closure_list_lock);
+	return 0;
+}
+
+static int debug_seq_open(struct inode *inode, struct file *file)
+{
+	return single_open(file, debug_seq_show, NULL);
+}
+
+static const struct file_operations debug_ops = {
+	.owner		= THIS_MODULE,
+	.open		= debug_seq_open,
+	.read		= seq_read,
+	.release	= single_release
+};
+
+void __init closure_debug_init(void)
+{
+	debug = debugfs_create_file("closures", 0400, NULL, NULL, &debug_ops);
+}
+
+#endif
+
+MODULE_AUTHOR("Kent Overstreet <koverstreet@google.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/bcache/closure.h b/drivers/md/bcache/closure.h
new file mode 100644
index 00000000000..a08e3eeac3c
--- /dev/null
+++ b/drivers/md/bcache/closure.h
@@ -0,0 +1,386 @@
+#ifndef _LINUX_CLOSURE_H
+#define _LINUX_CLOSURE_H
+
+#include <linux/llist.h>
+#include <linux/sched.h>
+#include <linux/workqueue.h>
+
+/*
+ * Closure is perhaps the most overused and abused term in computer science, but
+ * since I've been unable to come up with anything better you're stuck with it
+ * again.
+ *
+ * What are closures?
+ *
+ * They embed a refcount. The basic idea is they count "things that are in
+ * progress" - in flight bios, some other thread that's doing something else -
+ * anything you might want to wait on.
+ *
+ * The refcount may be manipulated with closure_get() and closure_put().
+ * closure_put() is where many of the interesting things happen, when it causes
+ * the refcount to go to 0.
+ *
+ * Closures can be used to wait on things both synchronously and asynchronously,
+ * and synchronous and asynchronous use can be mixed without restriction. To
+ * wait synchronously, use closure_sync() - you will sleep until your closure's
+ * refcount hits 1.
+ *
+ * To wait asynchronously, use
+ *   continue_at(cl, next_function, workqueue);
+ *
+ * passing it, as you might expect, the function to run when nothing is pending
+ * and the workqueue to run that function out of.
+ *
+ * continue_at() also, critically, is a macro that returns the calling function.
+ * There's good reason for this.
+ *
+ * To use safely closures asynchronously, they must always have a refcount while
+ * they are running owned by the thread that is running them. Otherwise, suppose
+ * you submit some bios and wish to have a function run when they all complete:
+ *
+ * foo_endio(struct bio *bio, int error)
+ * {
+ *	closure_put(cl);
+ * }
+ *
+ * closure_init(cl);
+ *
+ * do_stuff();
+ * closure_get(cl);
+ * bio1->bi_endio = foo_endio;
+ * bio_submit(bio1);
+ *
+ * do_more_stuff();
+ * closure_get(cl);
+ * bio2->bi_endio = foo_endio;
+ * bio_submit(bio2);
+ *
+ * continue_at(cl, complete_some_read, system_wq);
+ *
+ * If closure's refcount started at 0, complete_some_read() could run before the
+ * second bio was submitted - which is almost always not what you want! More
+ * importantly, it wouldn't be possible to say whether the original thread or
+ * complete_some_read()'s thread owned the closure - and whatever state it was
+ * associated with!
+ *
+ * So, closure_init() initializes a closure's refcount to 1 - and when a
+ * closure_fn is run, the refcount will be reset to 1 first.
+ *
+ * Then, the rule is - if you got the refcount with closure_get(), release it
+ * with closure_put() (i.e, in a bio->bi_endio function). If you have a refcount
+ * on a closure because you called closure_init() or you were run out of a
+ * closure - _always_ use continue_at(). Doing so consistently will help
+ * eliminate an entire class of particularly pernicious races.
+ *
+ * Lastly, you might have a wait list dedicated to a specific event, and have no
+ * need for specifying the condition - you just want to wait until someone runs
+ * closure_wake_up() on the appropriate wait list. In that case, just use
+ * closure_wait(). It will return either true or false, depending on whether the
+ * closure was already on a wait list or not - a closure can only be on one wait
+ * list at a time.
+ *
+ * Parents:
+ *
+ * closure_init() takes two arguments - it takes the closure to initialize, and
+ * a (possibly null) parent.
+ *
+ * If parent is non null, the new closure will have a refcount for its lifetime;
+ * a closure is considered to be "finished" when its refcount hits 0 and the
+ * function to run is null. Hence
+ *
+ * continue_at(cl, NULL, NULL);
+ *
+ * returns up the (spaghetti) stack of closures, precisely like normal return
+ * returns up the C stack. continue_at() with non null fn is better thought of
+ * as doing a tail call.
+ *
+ * All this implies that a closure should typically be embedded in a particular
+ * struct (which its refcount will normally control the lifetime of), and that
+ * struct can very much be thought of as a stack frame.
+ */
+
+struct closure;
+typedef void (closure_fn) (struct closure *);
+
+struct closure_waitlist {
+	struct llist_head	list;
+};
+
+enum closure_state {
+	/*
+	 * CLOSURE_WAITING: Set iff the closure is on a waitlist. Must be set by
+	 * the thread that owns the closure, and cleared by the thread that's
+	 * waking up the closure.
+	 *
+	 * CLOSURE_SLEEPING: Must be set before a thread uses a closure to sleep
+	 * - indicates that cl->task is valid and closure_put() may wake it up.
+	 * Only set or cleared by the thread that owns the closure.
+	 *
+	 * The rest are for debugging and don't affect behaviour:
+	 *
+	 * CLOSURE_RUNNING: Set when a closure is running (i.e. by
+	 * closure_init() and when closure_put() runs then next function), and
+	 * must be cleared before remaining hits 0. Primarily to help guard
+	 * against incorrect usage and accidentally transferring references.
+	 * continue_at() and closure_return() clear it for you, if you're doing
+	 * something unusual you can use closure_set_dead() which also helps
+	 * annotate where references are being transferred.
+	 *
+	 * CLOSURE_STACK: Sanity check - remaining should never hit 0 on a
+	 * closure with this flag set
+	 */
+
+	CLOSURE_BITS_START	= (1 << 23),
+	CLOSURE_DESTRUCTOR	= (1 << 23),
+	CLOSURE_WAITING		= (1 << 25),
+	CLOSURE_SLEEPING	= (1 << 27),
+	CLOSURE_RUNNING		= (1 << 29),
+	CLOSURE_STACK		= (1 << 31),
+};
+
+#define CLOSURE_GUARD_MASK					\
+	((CLOSURE_DESTRUCTOR|CLOSURE_WAITING|CLOSURE_SLEEPING|	\
+	  CLOSURE_RUNNING|CLOSURE_STACK) << 1)
+
+#define CLOSURE_REMAINING_MASK		(CLOSURE_BITS_START - 1)
+#define CLOSURE_REMAINING_INITIALIZER	(1|CLOSURE_RUNNING)
+
+struct closure {
+	union {
+		struct {
+			struct workqueue_struct *wq;
+			struct task_struct	*task;
+			struct llist_node	list;
+			closure_fn		*fn;
+		};
+		struct work_struct	work;
+	};
+
+	struct closure		*parent;
+
+	atomic_t		remaining;
+
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+#define CLOSURE_MAGIC_DEAD	0xc054dead
+#define CLOSURE_MAGIC_ALIVE	0xc054a11e
+
+	unsigned		magic;
+	struct list_head	all;
+	unsigned long		ip;
+	unsigned long		waiting_on;
+#endif
+};
+
+void closure_sub(struct closure *cl, int v);
+void closure_put(struct closure *cl);
+void __closure_wake_up(struct closure_waitlist *list);
+bool closure_wait(struct closure_waitlist *list, struct closure *cl);
+void closure_sync(struct closure *cl);
+
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+
+void closure_debug_init(void);
+void closure_debug_create(struct closure *cl);
+void closure_debug_destroy(struct closure *cl);
+
+#else
+
+static inline void closure_debug_init(void) {}
+static inline void closure_debug_create(struct closure *cl) {}
+static inline void closure_debug_destroy(struct closure *cl) {}
+
+#endif
+
+static inline void closure_set_ip(struct closure *cl)
+{
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+	cl->ip = _THIS_IP_;
+#endif
+}
+
+static inline void closure_set_ret_ip(struct closure *cl)
+{
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+	cl->ip = _RET_IP_;
+#endif
+}
+
+static inline void closure_set_waiting(struct closure *cl, unsigned long f)
+{
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+	cl->waiting_on = f;
+#endif
+}
+
+static inline void __closure_end_sleep(struct closure *cl)
+{
+	__set_current_state(TASK_RUNNING);
+
+	if (atomic_read(&cl->remaining) & CLOSURE_SLEEPING)
+		atomic_sub(CLOSURE_SLEEPING, &cl->remaining);
+}
+
+static inline void __closure_start_sleep(struct closure *cl)
+{
+	closure_set_ip(cl);
+	cl->task = current;
+	set_current_state(TASK_UNINTERRUPTIBLE);
+
+	if (!(atomic_read(&cl->remaining) & CLOSURE_SLEEPING))
+		atomic_add(CLOSURE_SLEEPING, &cl->remaining);
+}
+
+static inline void closure_set_stopped(struct closure *cl)
+{
+	atomic_sub(CLOSURE_RUNNING, &cl->remaining);
+}
+
+static inline void set_closure_fn(struct closure *cl, closure_fn *fn,
+				  struct workqueue_struct *wq)
+{
+	BUG_ON(object_is_on_stack(cl));
+	closure_set_ip(cl);
+	cl->fn = fn;
+	cl->wq = wq;
+	/* between atomic_dec() in closure_put() */
+	smp_mb__before_atomic();
+}
+
+static inline void closure_queue(struct closure *cl)
+{
+	struct workqueue_struct *wq = cl->wq;
+	if (wq) {
+		INIT_WORK(&cl->work, cl->work.func);
+		BUG_ON(!queue_work(wq, &cl->work));
+	} else
+		cl->fn(cl);
+}
+
+/**
+ * closure_get - increment a closure's refcount
+ */
+static inline void closure_get(struct closure *cl)
+{
+#ifdef CONFIG_BCACHE_CLOSURES_DEBUG
+	BUG_ON((atomic_inc_return(&cl->remaining) &
+		CLOSURE_REMAINING_MASK) <= 1);
+#else
+	atomic_inc(&cl->remaining);
+#endif
+}
+
+/**
+ * closure_init - Initialize a closure, setting the refcount to 1
+ * @cl:		closure to initialize
+ * @parent:	parent of the new closure. cl will take a refcount on it for its
+ *		lifetime; may be NULL.
+ */
+static inline void closure_init(struct closure *cl, struct closure *parent)
+{
+	memset(cl, 0, sizeof(struct closure));
+	cl->parent = parent;
+	if (parent)
+		closure_get(parent);
+
+	atomic_set(&cl->remaining, CLOSURE_REMAINING_INITIALIZER);
+
+	closure_debug_create(cl);
+	closure_set_ip(cl);
+}
+
+static inline void closure_init_stack(struct closure *cl)
+{
+	memset(cl, 0, sizeof(struct closure));
+	atomic_set(&cl->remaining, CLOSURE_REMAINING_INITIALIZER|CLOSURE_STACK);
+}
+
+/**
+ * closure_wake_up - wake up all closures on a wait list.
+ */
+static inline void closure_wake_up(struct closure_waitlist *list)
+{
+	smp_mb();
+	__closure_wake_up(list);
+}
+
+/**
+ * continue_at - jump to another function with barrier
+ *
+ * After @cl is no longer waiting on anything (i.e. all outstanding refs have
+ * been dropped with closure_put()), it will resume execution at @fn running out
+ * of @wq (or, if @wq is NULL, @fn will be called by closure_put() directly).
+ *
+ * NOTE: This macro expands to a return in the calling function!
+ *
+ * This is because after calling continue_at() you no longer have a ref on @cl,
+ * and whatever @cl owns may be freed out from under you - a running closure fn
+ * has a ref on its own closure which continue_at() drops.
+ */
+#define continue_at(_cl, _fn, _wq)					\
+do {									\
+	set_closure_fn(_cl, _fn, _wq);					\
+	closure_sub(_cl, CLOSURE_RUNNING + 1);				\
+	return;								\
+} while (0)
+
+/**
+ * closure_return - finish execution of a closure
+ *
+ * This is used to indicate that @cl is finished: when all outstanding refs on
+ * @cl have been dropped @cl's ref on its parent closure (as passed to
+ * closure_init()) will be dropped, if one was specified - thus this can be
+ * thought of as returning to the parent closure.
+ */
+#define closure_return(_cl)	continue_at((_cl), NULL, NULL)
+
+/**
+ * continue_at_nobarrier - jump to another function without barrier
+ *
+ * Causes @fn to be executed out of @cl, in @wq context (or called directly if
+ * @wq is NULL).
+ *
+ * NOTE: like continue_at(), this macro expands to a return in the caller!
+ *
+ * The ref the caller of continue_at_nobarrier() had on @cl is now owned by @fn,
+ * thus it's not safe to touch anything protected by @cl after a
+ * continue_at_nobarrier().
+ */
+#define continue_at_nobarrier(_cl, _fn, _wq)				\
+do {									\
+	set_closure_fn(_cl, _fn, _wq);					\
+	closure_queue(_cl);						\
+	return;								\
+} while (0)
+
+/**
+ * closure_return - finish execution of a closure, with destructor
+ *
+ * Works like closure_return(), except @destructor will be called when all
+ * outstanding refs on @cl have been dropped; @destructor may be used to safely
+ * free the memory occupied by @cl, and it is called with the ref on the parent
+ * closure still held - so @destructor could safely return an item to a
+ * freelist protected by @cl's parent.
+ */
+#define closure_return_with_destructor(_cl, _destructor)		\
+do {									\
+	set_closure_fn(_cl, _destructor, NULL);				\
+	closure_sub(_cl, CLOSURE_RUNNING - CLOSURE_DESTRUCTOR + 1);	\
+	return;								\
+} while (0)
+
+/**
+ * closure_call - execute @fn out of a new, uninitialized closure
+ *
+ * Typically used when running out of one closure, and we want to run @fn
+ * asynchronously out of a new closure - @parent will then wait for @cl to
+ * finish.
+ */
+static inline void closure_call(struct closure *cl, closure_fn fn,
+				struct workqueue_struct *wq,
+				struct closure *parent)
+{
+	closure_init(cl, parent);
+	continue_at_nobarrier(cl, fn, wq);
+}
+
+#endif /* _LINUX_CLOSURE_H */
diff --git a/drivers/md/bcache/debug.c b/drivers/md/bcache/debug.c
new file mode 100644
index 00000000000..8b1f1d5c181
--- /dev/null
+++ b/drivers/md/bcache/debug.c
@@ -0,0 +1,252 @@
+/*
+ * Assorted bcache debug code
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "extents.h"
+
+#include <linux/console.h>
+#include <linux/debugfs.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/seq_file.h>
+
+static struct dentry *debug;
+
+#ifdef CONFIG_BCACHE_DEBUG
+
+#define for_each_written_bset(b, start, i)				\
+	for (i = (start);						\
+	     (void *) i < (void *) (start) + (KEY_SIZE(&b->key) << 9) &&\
+	     i->seq == (start)->seq;					\
+	     i = (void *) i + set_blocks(i, block_bytes(b->c)) *	\
+		 block_bytes(b->c))
+
+void bch_btree_verify(struct btree *b)
+{
+	struct btree *v = b->c->verify_data;
+	struct bset *ondisk, *sorted, *inmemory;
+	struct bio *bio;
+
+	if (!b->c->verify || !b->c->verify_ondisk)
+		return;
+
+	down(&b->io_mutex);
+	mutex_lock(&b->c->verify_lock);
+
+	ondisk = b->c->verify_ondisk;
+	sorted = b->c->verify_data->keys.set->data;
+	inmemory = b->keys.set->data;
+
+	bkey_copy(&v->key, &b->key);
+	v->written = 0;
+	v->level = b->level;
+	v->keys.ops = b->keys.ops;
+
+	bio = bch_bbio_alloc(b->c);
+	bio->bi_bdev		= PTR_CACHE(b->c, &b->key, 0)->bdev;
+	bio->bi_iter.bi_sector	= PTR_OFFSET(&b->key, 0);
+	bio->bi_iter.bi_size	= KEY_SIZE(&v->key) << 9;
+	bch_bio_map(bio, sorted);
+
+	submit_bio_wait(REQ_META|READ_SYNC, bio);
+	bch_bbio_free(bio, b->c);
+
+	memcpy(ondisk, sorted, KEY_SIZE(&v->key) << 9);
+
+	bch_btree_node_read_done(v);
+	sorted = v->keys.set->data;
+
+	if (inmemory->keys != sorted->keys ||
+	    memcmp(inmemory->start,
+		   sorted->start,
+		   (void *) bset_bkey_last(inmemory) - (void *) inmemory->start)) {
+		struct bset *i;
+		unsigned j;
+
+		console_lock();
+
+		printk(KERN_ERR "*** in memory:\n");
+		bch_dump_bset(&b->keys, inmemory, 0);
+
+		printk(KERN_ERR "*** read back in:\n");
+		bch_dump_bset(&v->keys, sorted, 0);
+
+		for_each_written_bset(b, ondisk, i) {
+			unsigned block = ((void *) i - (void *) ondisk) /
+				block_bytes(b->c);
+
+			printk(KERN_ERR "*** on disk block %u:\n", block);
+			bch_dump_bset(&b->keys, i, block);
+		}
+
+		printk(KERN_ERR "*** block %zu not written\n",
+		       ((void *) i - (void *) ondisk) / block_bytes(b->c));
+
+		for (j = 0; j < inmemory->keys; j++)
+			if (inmemory->d[j] != sorted->d[j])
+				break;
+
+		printk(KERN_ERR "b->written %u\n", b->written);
+
+		console_unlock();
+		panic("verify failed at %u\n", j);
+	}
+
+	mutex_unlock(&b->c->verify_lock);
+	up(&b->io_mutex);
+}
+
+void bch_data_verify(struct cached_dev *dc, struct bio *bio)
+{
+	char name[BDEVNAME_SIZE];
+	struct bio *check;
+	struct bio_vec bv, *bv2;
+	struct bvec_iter iter;
+	int i;
+
+	check = bio_clone(bio, GFP_NOIO);
+	if (!check)
+		return;
+
+	if (bio_alloc_pages(check, GFP_NOIO))
+		goto out_put;
+
+	submit_bio_wait(READ_SYNC, check);
+
+	bio_for_each_segment(bv, bio, iter) {
+		void *p1 = kmap_atomic(bv.bv_page);
+		void *p2 = page_address(check->bi_io_vec[iter.bi_idx].bv_page);
+
+		cache_set_err_on(memcmp(p1 + bv.bv_offset,
+					p2 + bv.bv_offset,
+					bv.bv_len),
+				 dc->disk.c,
+				 "verify failed at dev %s sector %llu",
+				 bdevname(dc->bdev, name),
+				 (uint64_t) bio->bi_iter.bi_sector);
+
+		kunmap_atomic(p1);
+	}
+
+	bio_for_each_segment_all(bv2, check, i)
+		__free_page(bv2->bv_page);
+out_put:
+	bio_put(check);
+}
+
+#endif
+
+#ifdef CONFIG_DEBUG_FS
+
+/* XXX: cache set refcounting */
+
+struct dump_iterator {
+	char			buf[PAGE_SIZE];
+	size_t			bytes;
+	struct cache_set	*c;
+	struct keybuf		keys;
+};
+
+static bool dump_pred(struct keybuf *buf, struct bkey *k)
+{
+	return true;
+}
+
+static ssize_t bch_dump_read(struct file *file, char __user *buf,
+			     size_t size, loff_t *ppos)
+{
+	struct dump_iterator *i = file->private_data;
+	ssize_t ret = 0;
+	char kbuf[80];
+
+	while (size) {
+		struct keybuf_key *w;
+		unsigned bytes = min(i->bytes, size);
+
+		int err = copy_to_user(buf, i->buf, bytes);
+		if (err)
+			return err;
+
+		ret	 += bytes;
+		buf	 += bytes;
+		size	 -= bytes;
+		i->bytes -= bytes;
+		memmove(i->buf, i->buf + bytes, i->bytes);
+
+		if (i->bytes)
+			break;
+
+		w = bch_keybuf_next_rescan(i->c, &i->keys, &MAX_KEY, dump_pred);
+		if (!w)
+			break;
+
+		bch_extent_to_text(kbuf, sizeof(kbuf), &w->key);
+		i->bytes = snprintf(i->buf, PAGE_SIZE, "%s\n", kbuf);
+		bch_keybuf_del(&i->keys, w);
+	}
+
+	return ret;
+}
+
+static int bch_dump_open(struct inode *inode, struct file *file)
+{
+	struct cache_set *c = inode->i_private;
+	struct dump_iterator *i;
+
+	i = kzalloc(sizeof(struct dump_iterator), GFP_KERNEL);
+	if (!i)
+		return -ENOMEM;
+
+	file->private_data = i;
+	i->c = c;
+	bch_keybuf_init(&i->keys);
+	i->keys.last_scanned = KEY(0, 0, 0);
+
+	return 0;
+}
+
+static int bch_dump_release(struct inode *inode, struct file *file)
+{
+	kfree(file->private_data);
+	return 0;
+}
+
+static const struct file_operations cache_set_debug_ops = {
+	.owner		= THIS_MODULE,
+	.open		= bch_dump_open,
+	.read		= bch_dump_read,
+	.release	= bch_dump_release
+};
+
+void bch_debug_init_cache_set(struct cache_set *c)
+{
+	if (!IS_ERR_OR_NULL(debug)) {
+		char name[50];
+		snprintf(name, 50, "bcache-%pU", c->sb.set_uuid);
+
+		c->debug = debugfs_create_file(name, 0400, debug, c,
+					       &cache_set_debug_ops);
+	}
+}
+
+#endif
+
+void bch_debug_exit(void)
+{
+	if (!IS_ERR_OR_NULL(debug))
+		debugfs_remove_recursive(debug);
+}
+
+int __init bch_debug_init(struct kobject *kobj)
+{
+	int ret = 0;
+
+	debug = debugfs_create_dir("bcache", NULL);
+	return ret;
+}
diff --git a/drivers/md/bcache/debug.h b/drivers/md/bcache/debug.h
new file mode 100644
index 00000000000..1f63c195d24
--- /dev/null
+++ b/drivers/md/bcache/debug.h
@@ -0,0 +1,34 @@
+#ifndef _BCACHE_DEBUG_H
+#define _BCACHE_DEBUG_H
+
+struct bio;
+struct cached_dev;
+struct cache_set;
+
+#ifdef CONFIG_BCACHE_DEBUG
+
+void bch_btree_verify(struct btree *);
+void bch_data_verify(struct cached_dev *, struct bio *);
+
+#define expensive_debug_checks(c)	((c)->expensive_debug_checks)
+#define key_merging_disabled(c)		((c)->key_merging_disabled)
+#define bypass_torture_test(d)		((d)->bypass_torture_test)
+
+#else /* DEBUG */
+
+static inline void bch_btree_verify(struct btree *b) {}
+static inline void bch_data_verify(struct cached_dev *dc, struct bio *bio) {}
+
+#define expensive_debug_checks(c)	0
+#define key_merging_disabled(c)		0
+#define bypass_torture_test(d)		0
+
+#endif
+
+#ifdef CONFIG_DEBUG_FS
+void bch_debug_init_cache_set(struct cache_set *);
+#else
+static inline void bch_debug_init_cache_set(struct cache_set *c) {}
+#endif
+
+#endif
diff --git a/drivers/md/bcache/extents.c b/drivers/md/bcache/extents.c
new file mode 100644
index 00000000000..3a0de4cf977
--- /dev/null
+++ b/drivers/md/bcache/extents.c
@@ -0,0 +1,620 @@
+/*
+ * Copyright (C) 2010 Kent Overstreet <kent.overstreet@gmail.com>
+ *
+ * Uses a block device as cache for other block devices; optimized for SSDs.
+ * All allocation is done in buckets, which should match the erase block size
+ * of the device.
+ *
+ * Buckets containing cached data are kept on a heap sorted by priority;
+ * bucket priority is increased on cache hit, and periodically all the buckets
+ * on the heap have their priority scaled down. This currently is just used as
+ * an LRU but in the future should allow for more intelligent heuristics.
+ *
+ * Buckets have an 8 bit counter; freeing is accomplished by incrementing the
+ * counter. Garbage collection is used to remove stale pointers.
+ *
+ * Indexing is done via a btree; nodes are not necessarily fully sorted, rather
+ * as keys are inserted we only sort the pages that have not yet been written.
+ * When garbage collection is run, we resort the entire node.
+ *
+ * All configuration is done via sysfs; see Documentation/bcache.txt.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "extents.h"
+#include "writeback.h"
+
+static void sort_key_next(struct btree_iter *iter,
+			  struct btree_iter_set *i)
+{
+	i->k = bkey_next(i->k);
+
+	if (i->k == i->end)
+		*i = iter->data[--iter->used];
+}
+
+static bool bch_key_sort_cmp(struct btree_iter_set l,
+			     struct btree_iter_set r)
+{
+	int64_t c = bkey_cmp(l.k, r.k);
+
+	return c ? c > 0 : l.k < r.k;
+}
+
+static bool __ptr_invalid(struct cache_set *c, const struct bkey *k)
+{
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (ptr_available(c, k, i)) {
+			struct cache *ca = PTR_CACHE(c, k, i);
+			size_t bucket = PTR_BUCKET_NR(c, k, i);
+			size_t r = bucket_remainder(c, PTR_OFFSET(k, i));
+
+			if (KEY_SIZE(k) + r > c->sb.bucket_size ||
+			    bucket <  ca->sb.first_bucket ||
+			    bucket >= ca->sb.nbuckets)
+				return true;
+		}
+
+	return false;
+}
+
+/* Common among btree and extent ptrs */
+
+static const char *bch_ptr_status(struct cache_set *c, const struct bkey *k)
+{
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (ptr_available(c, k, i)) {
+			struct cache *ca = PTR_CACHE(c, k, i);
+			size_t bucket = PTR_BUCKET_NR(c, k, i);
+			size_t r = bucket_remainder(c, PTR_OFFSET(k, i));
+
+			if (KEY_SIZE(k) + r > c->sb.bucket_size)
+				return "bad, length too big";
+			if (bucket <  ca->sb.first_bucket)
+				return "bad, short offset";
+			if (bucket >= ca->sb.nbuckets)
+				return "bad, offset past end of device";
+			if (ptr_stale(c, k, i))
+				return "stale";
+		}
+
+	if (!bkey_cmp(k, &ZERO_KEY))
+		return "bad, null key";
+	if (!KEY_PTRS(k))
+		return "bad, no pointers";
+	if (!KEY_SIZE(k))
+		return "zeroed key";
+	return "";
+}
+
+void bch_extent_to_text(char *buf, size_t size, const struct bkey *k)
+{
+	unsigned i = 0;
+	char *out = buf, *end = buf + size;
+
+#define p(...)	(out += scnprintf(out, end - out, __VA_ARGS__))
+
+	p("%llu:%llu len %llu -> [", KEY_INODE(k), KEY_START(k), KEY_SIZE(k));
+
+	for (i = 0; i < KEY_PTRS(k); i++) {
+		if (i)
+			p(", ");
+
+		if (PTR_DEV(k, i) == PTR_CHECK_DEV)
+			p("check dev");
+		else
+			p("%llu:%llu gen %llu", PTR_DEV(k, i),
+			  PTR_OFFSET(k, i), PTR_GEN(k, i));
+	}
+
+	p("]");
+
+	if (KEY_DIRTY(k))
+		p(" dirty");
+	if (KEY_CSUM(k))
+		p(" cs%llu %llx", KEY_CSUM(k), k->ptr[1]);
+#undef p
+}
+
+static void bch_bkey_dump(struct btree_keys *keys, const struct bkey *k)
+{
+	struct btree *b = container_of(keys, struct btree, keys);
+	unsigned j;
+	char buf[80];
+
+	bch_extent_to_text(buf, sizeof(buf), k);
+	printk(" %s", buf);
+
+	for (j = 0; j < KEY_PTRS(k); j++) {
+		size_t n = PTR_BUCKET_NR(b->c, k, j);
+		printk(" bucket %zu", n);
+
+		if (n >= b->c->sb.first_bucket && n < b->c->sb.nbuckets)
+			printk(" prio %i",
+			       PTR_BUCKET(b->c, k, j)->prio);
+	}
+
+	printk(" %s\n", bch_ptr_status(b->c, k));
+}
+
+/* Btree ptrs */
+
+bool __bch_btree_ptr_invalid(struct cache_set *c, const struct bkey *k)
+{
+	char buf[80];
+
+	if (!KEY_PTRS(k) || !KEY_SIZE(k) || KEY_DIRTY(k))
+		goto bad;
+
+	if (__ptr_invalid(c, k))
+		goto bad;
+
+	return false;
+bad:
+	bch_extent_to_text(buf, sizeof(buf), k);
+	cache_bug(c, "spotted btree ptr %s: %s", buf, bch_ptr_status(c, k));
+	return true;
+}
+
+static bool bch_btree_ptr_invalid(struct btree_keys *bk, const struct bkey *k)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+	return __bch_btree_ptr_invalid(b->c, k);
+}
+
+static bool btree_ptr_bad_expensive(struct btree *b, const struct bkey *k)
+{
+	unsigned i;
+	char buf[80];
+	struct bucket *g;
+
+	if (mutex_trylock(&b->c->bucket_lock)) {
+		for (i = 0; i < KEY_PTRS(k); i++)
+			if (ptr_available(b->c, k, i)) {
+				g = PTR_BUCKET(b->c, k, i);
+
+				if (KEY_DIRTY(k) ||
+				    g->prio != BTREE_PRIO ||
+				    (b->c->gc_mark_valid &&
+				     GC_MARK(g) != GC_MARK_METADATA))
+					goto err;
+			}
+
+		mutex_unlock(&b->c->bucket_lock);
+	}
+
+	return false;
+err:
+	mutex_unlock(&b->c->bucket_lock);
+	bch_extent_to_text(buf, sizeof(buf), k);
+	btree_bug(b,
+"inconsistent btree pointer %s: bucket %zi pin %i prio %i gen %i last_gc %i mark %llu",
+		  buf, PTR_BUCKET_NR(b->c, k, i), atomic_read(&g->pin),
+		  g->prio, g->gen, g->last_gc, GC_MARK(g));
+	return true;
+}
+
+static bool bch_btree_ptr_bad(struct btree_keys *bk, const struct bkey *k)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+	unsigned i;
+
+	if (!bkey_cmp(k, &ZERO_KEY) ||
+	    !KEY_PTRS(k) ||
+	    bch_ptr_invalid(bk, k))
+		return true;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (!ptr_available(b->c, k, i) ||
+		    ptr_stale(b->c, k, i))
+			return true;
+
+	if (expensive_debug_checks(b->c) &&
+	    btree_ptr_bad_expensive(b, k))
+		return true;
+
+	return false;
+}
+
+static bool bch_btree_ptr_insert_fixup(struct btree_keys *bk,
+				       struct bkey *insert,
+				       struct btree_iter *iter,
+				       struct bkey *replace_key)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+
+	if (!KEY_OFFSET(insert))
+		btree_current_write(b)->prio_blocked++;
+
+	return false;
+}
+
+const struct btree_keys_ops bch_btree_keys_ops = {
+	.sort_cmp	= bch_key_sort_cmp,
+	.insert_fixup	= bch_btree_ptr_insert_fixup,
+	.key_invalid	= bch_btree_ptr_invalid,
+	.key_bad	= bch_btree_ptr_bad,
+	.key_to_text	= bch_extent_to_text,
+	.key_dump	= bch_bkey_dump,
+};
+
+/* Extents */
+
+/*
+ * Returns true if l > r - unless l == r, in which case returns true if l is
+ * older than r.
+ *
+ * Necessary for btree_sort_fixup() - if there are multiple keys that compare
+ * equal in different sets, we have to process them newest to oldest.
+ */
+static bool bch_extent_sort_cmp(struct btree_iter_set l,
+				struct btree_iter_set r)
+{
+	int64_t c = bkey_cmp(&START_KEY(l.k), &START_KEY(r.k));
+
+	return c ? c > 0 : l.k < r.k;
+}
+
+static struct bkey *bch_extent_sort_fixup(struct btree_iter *iter,
+					  struct bkey *tmp)
+{
+	while (iter->used > 1) {
+		struct btree_iter_set *top = iter->data, *i = top + 1;
+
+		if (iter->used > 2 &&
+		    bch_extent_sort_cmp(i[0], i[1]))
+			i++;
+
+		if (bkey_cmp(top->k, &START_KEY(i->k)) <= 0)
+			break;
+
+		if (!KEY_SIZE(i->k)) {
+			sort_key_next(iter, i);
+			heap_sift(iter, i - top, bch_extent_sort_cmp);
+			continue;
+		}
+
+		if (top->k > i->k) {
+			if (bkey_cmp(top->k, i->k) >= 0)
+				sort_key_next(iter, i);
+			else
+				bch_cut_front(top->k, i->k);
+
+			heap_sift(iter, i - top, bch_extent_sort_cmp);
+		} else {
+			/* can't happen because of comparison func */
+			BUG_ON(!bkey_cmp(&START_KEY(top->k), &START_KEY(i->k)));
+
+			if (bkey_cmp(i->k, top->k) < 0) {
+				bkey_copy(tmp, top->k);
+
+				bch_cut_back(&START_KEY(i->k), tmp);
+				bch_cut_front(i->k, top->k);
+				heap_sift(iter, 0, bch_extent_sort_cmp);
+
+				return tmp;
+			} else {
+				bch_cut_back(&START_KEY(i->k), top->k);
+			}
+		}
+	}
+
+	return NULL;
+}
+
+static void bch_subtract_dirty(struct bkey *k,
+			   struct cache_set *c,
+			   uint64_t offset,
+			   int sectors)
+{
+	if (KEY_DIRTY(k))
+		bcache_dev_sectors_dirty_add(c, KEY_INODE(k),
+					     offset, -sectors);
+}
+
+static bool bch_extent_insert_fixup(struct btree_keys *b,
+				    struct bkey *insert,
+				    struct btree_iter *iter,
+				    struct bkey *replace_key)
+{
+	struct cache_set *c = container_of(b, struct btree, keys)->c;
+
+	uint64_t old_offset;
+	unsigned old_size, sectors_found = 0;
+
+	BUG_ON(!KEY_OFFSET(insert));
+	BUG_ON(!KEY_SIZE(insert));
+
+	while (1) {
+		struct bkey *k = bch_btree_iter_next(iter);
+		if (!k)
+			break;
+
+		if (bkey_cmp(&START_KEY(k), insert) >= 0) {
+			if (KEY_SIZE(k))
+				break;
+			else
+				continue;
+		}
+
+		if (bkey_cmp(k, &START_KEY(insert)) <= 0)
+			continue;
+
+		old_offset = KEY_START(k);
+		old_size = KEY_SIZE(k);
+
+		/*
+		 * We might overlap with 0 size extents; we can't skip these
+		 * because if they're in the set we're inserting to we have to
+		 * adjust them so they don't overlap with the key we're
+		 * inserting. But we don't want to check them for replace
+		 * operations.
+		 */
+
+		if (replace_key && KEY_SIZE(k)) {
+			/*
+			 * k might have been split since we inserted/found the
+			 * key we're replacing
+			 */
+			unsigned i;
+			uint64_t offset = KEY_START(k) -
+				KEY_START(replace_key);
+
+			/* But it must be a subset of the replace key */
+			if (KEY_START(k) < KEY_START(replace_key) ||
+			    KEY_OFFSET(k) > KEY_OFFSET(replace_key))
+				goto check_failed;
+
+			/* We didn't find a key that we were supposed to */
+			if (KEY_START(k) > KEY_START(insert) + sectors_found)
+				goto check_failed;
+
+			if (!bch_bkey_equal_header(k, replace_key))
+				goto check_failed;
+
+			/* skip past gen */
+			offset <<= 8;
+
+			BUG_ON(!KEY_PTRS(replace_key));
+
+			for (i = 0; i < KEY_PTRS(replace_key); i++)
+				if (k->ptr[i] != replace_key->ptr[i] + offset)
+					goto check_failed;
+
+			sectors_found = KEY_OFFSET(k) - KEY_START(insert);
+		}
+
+		if (bkey_cmp(insert, k) < 0 &&
+		    bkey_cmp(&START_KEY(insert), &START_KEY(k)) > 0) {
+			/*
+			 * We overlapped in the middle of an existing key: that
+			 * means we have to split the old key. But we have to do
+			 * slightly different things depending on whether the
+			 * old key has been written out yet.
+			 */
+
+			struct bkey *top;
+
+			bch_subtract_dirty(k, c, KEY_START(insert),
+				       KEY_SIZE(insert));
+
+			if (bkey_written(b, k)) {
+				/*
+				 * We insert a new key to cover the top of the
+				 * old key, and the old key is modified in place
+				 * to represent the bottom split.
+				 *
+				 * It's completely arbitrary whether the new key
+				 * is the top or the bottom, but it has to match
+				 * up with what btree_sort_fixup() does - it
+				 * doesn't check for this kind of overlap, it
+				 * depends on us inserting a new key for the top
+				 * here.
+				 */
+				top = bch_bset_search(b, bset_tree_last(b),
+						      insert);
+				bch_bset_insert(b, top, k);
+			} else {
+				BKEY_PADDED(key) temp;
+				bkey_copy(&temp.key, k);
+				bch_bset_insert(b, k, &temp.key);
+				top = bkey_next(k);
+			}
+
+			bch_cut_front(insert, top);
+			bch_cut_back(&START_KEY(insert), k);
+			bch_bset_fix_invalidated_key(b, k);
+			goto out;
+		}
+
+		if (bkey_cmp(insert, k) < 0) {
+			bch_cut_front(insert, k);
+		} else {
+			if (bkey_cmp(&START_KEY(insert), &START_KEY(k)) > 0)
+				old_offset = KEY_START(insert);
+
+			if (bkey_written(b, k) &&
+			    bkey_cmp(&START_KEY(insert), &START_KEY(k)) <= 0) {
+				/*
+				 * Completely overwrote, so we don't have to
+				 * invalidate the binary search tree
+				 */
+				bch_cut_front(k, k);
+			} else {
+				__bch_cut_back(&START_KEY(insert), k);
+				bch_bset_fix_invalidated_key(b, k);
+			}
+		}
+
+		bch_subtract_dirty(k, c, old_offset, old_size - KEY_SIZE(k));
+	}
+
+check_failed:
+	if (replace_key) {
+		if (!sectors_found) {
+			return true;
+		} else if (sectors_found < KEY_SIZE(insert)) {
+			SET_KEY_OFFSET(insert, KEY_OFFSET(insert) -
+				       (KEY_SIZE(insert) - sectors_found));
+			SET_KEY_SIZE(insert, sectors_found);
+		}
+	}
+out:
+	if (KEY_DIRTY(insert))
+		bcache_dev_sectors_dirty_add(c, KEY_INODE(insert),
+					     KEY_START(insert),
+					     KEY_SIZE(insert));
+
+	return false;
+}
+
+static bool bch_extent_invalid(struct btree_keys *bk, const struct bkey *k)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+	char buf[80];
+
+	if (!KEY_SIZE(k))
+		return true;
+
+	if (KEY_SIZE(k) > KEY_OFFSET(k))
+		goto bad;
+
+	if (__ptr_invalid(b->c, k))
+		goto bad;
+
+	return false;
+bad:
+	bch_extent_to_text(buf, sizeof(buf), k);
+	cache_bug(b->c, "spotted extent %s: %s", buf, bch_ptr_status(b->c, k));
+	return true;
+}
+
+static bool bch_extent_bad_expensive(struct btree *b, const struct bkey *k,
+				     unsigned ptr)
+{
+	struct bucket *g = PTR_BUCKET(b->c, k, ptr);
+	char buf[80];
+
+	if (mutex_trylock(&b->c->bucket_lock)) {
+		if (b->c->gc_mark_valid &&
+		    (!GC_MARK(g) ||
+		     GC_MARK(g) == GC_MARK_METADATA ||
+		     (GC_MARK(g) != GC_MARK_DIRTY && KEY_DIRTY(k))))
+			goto err;
+
+		if (g->prio == BTREE_PRIO)
+			goto err;
+
+		mutex_unlock(&b->c->bucket_lock);
+	}
+
+	return false;
+err:
+	mutex_unlock(&b->c->bucket_lock);
+	bch_extent_to_text(buf, sizeof(buf), k);
+	btree_bug(b,
+"inconsistent extent pointer %s:\nbucket %zu pin %i prio %i gen %i last_gc %i mark %llu",
+		  buf, PTR_BUCKET_NR(b->c, k, ptr), atomic_read(&g->pin),
+		  g->prio, g->gen, g->last_gc, GC_MARK(g));
+	return true;
+}
+
+static bool bch_extent_bad(struct btree_keys *bk, const struct bkey *k)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+	struct bucket *g;
+	unsigned i, stale;
+
+	if (!KEY_PTRS(k) ||
+	    bch_extent_invalid(bk, k))
+		return true;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (!ptr_available(b->c, k, i))
+			return true;
+
+	if (!expensive_debug_checks(b->c) && KEY_DIRTY(k))
+		return false;
+
+	for (i = 0; i < KEY_PTRS(k); i++) {
+		g = PTR_BUCKET(b->c, k, i);
+		stale = ptr_stale(b->c, k, i);
+
+		btree_bug_on(stale > 96, b,
+			     "key too stale: %i, need_gc %u",
+			     stale, b->c->need_gc);
+
+		btree_bug_on(stale && KEY_DIRTY(k) && KEY_SIZE(k),
+			     b, "stale dirty pointer");
+
+		if (stale)
+			return true;
+
+		if (expensive_debug_checks(b->c) &&
+		    bch_extent_bad_expensive(b, k, i))
+			return true;
+	}
+
+	return false;
+}
+
+static uint64_t merge_chksums(struct bkey *l, struct bkey *r)
+{
+	return (l->ptr[KEY_PTRS(l)] + r->ptr[KEY_PTRS(r)]) &
+		~((uint64_t)1 << 63);
+}
+
+static bool bch_extent_merge(struct btree_keys *bk, struct bkey *l, struct bkey *r)
+{
+	struct btree *b = container_of(bk, struct btree, keys);
+	unsigned i;
+
+	if (key_merging_disabled(b->c))
+		return false;
+
+	for (i = 0; i < KEY_PTRS(l); i++)
+		if (l->ptr[i] + PTR(0, KEY_SIZE(l), 0) != r->ptr[i] ||
+		    PTR_BUCKET_NR(b->c, l, i) != PTR_BUCKET_NR(b->c, r, i))
+			return false;
+
+	/* Keys with no pointers aren't restricted to one bucket and could
+	 * overflow KEY_SIZE
+	 */
+	if (KEY_SIZE(l) + KEY_SIZE(r) > USHRT_MAX) {
+		SET_KEY_OFFSET(l, KEY_OFFSET(l) + USHRT_MAX - KEY_SIZE(l));
+		SET_KEY_SIZE(l, USHRT_MAX);
+
+		bch_cut_front(l, r);
+		return false;
+	}
+
+	if (KEY_CSUM(l)) {
+		if (KEY_CSUM(r))
+			l->ptr[KEY_PTRS(l)] = merge_chksums(l, r);
+		else
+			SET_KEY_CSUM(l, 0);
+	}
+
+	SET_KEY_OFFSET(l, KEY_OFFSET(l) + KEY_SIZE(r));
+	SET_KEY_SIZE(l, KEY_SIZE(l) + KEY_SIZE(r));
+
+	return true;
+}
+
+const struct btree_keys_ops bch_extent_keys_ops = {
+	.sort_cmp	= bch_extent_sort_cmp,
+	.sort_fixup	= bch_extent_sort_fixup,
+	.insert_fixup	= bch_extent_insert_fixup,
+	.key_invalid	= bch_extent_invalid,
+	.key_bad	= bch_extent_bad,
+	.key_merge	= bch_extent_merge,
+	.key_to_text	= bch_extent_to_text,
+	.key_dump	= bch_bkey_dump,
+	.is_extents	= true,
+};
diff --git a/drivers/md/bcache/extents.h b/drivers/md/bcache/extents.h
new file mode 100644
index 00000000000..e4e23409782
--- /dev/null
+++ b/drivers/md/bcache/extents.h
@@ -0,0 +1,13 @@
+#ifndef _BCACHE_EXTENTS_H
+#define _BCACHE_EXTENTS_H
+
+extern const struct btree_keys_ops bch_btree_keys_ops;
+extern const struct btree_keys_ops bch_extent_keys_ops;
+
+struct bkey;
+struct cache_set;
+
+void bch_extent_to_text(char *, size_t, const struct bkey *);
+bool __bch_btree_ptr_invalid(struct cache_set *, const struct bkey *);
+
+#endif /* _BCACHE_EXTENTS_H */
diff --git a/drivers/md/bcache/io.c b/drivers/md/bcache/io.c
new file mode 100644
index 00000000000..fa028fa82df
--- /dev/null
+++ b/drivers/md/bcache/io.c
@@ -0,0 +1,243 @@
+/*
+ * Some low level IO code, and hacks for various block layer limitations
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "bset.h"
+#include "debug.h"
+
+#include <linux/blkdev.h>
+
+static unsigned bch_bio_max_sectors(struct bio *bio)
+{
+	struct request_queue *q = bdev_get_queue(bio->bi_bdev);
+	struct bio_vec bv;
+	struct bvec_iter iter;
+	unsigned ret = 0, seg = 0;
+
+	if (bio->bi_rw & REQ_DISCARD)
+		return min(bio_sectors(bio), q->limits.max_discard_sectors);
+
+	bio_for_each_segment(bv, bio, iter) {
+		struct bvec_merge_data bvm = {
+			.bi_bdev	= bio->bi_bdev,
+			.bi_sector	= bio->bi_iter.bi_sector,
+			.bi_size	= ret << 9,
+			.bi_rw		= bio->bi_rw,
+		};
+
+		if (seg == min_t(unsigned, BIO_MAX_PAGES,
+				 queue_max_segments(q)))
+			break;
+
+		if (q->merge_bvec_fn &&
+		    q->merge_bvec_fn(q, &bvm, &bv) < (int) bv.bv_len)
+			break;
+
+		seg++;
+		ret += bv.bv_len >> 9;
+	}
+
+	ret = min(ret, queue_max_sectors(q));
+
+	WARN_ON(!ret);
+	ret = max_t(int, ret, bio_iovec(bio).bv_len >> 9);
+
+	return ret;
+}
+
+static void bch_bio_submit_split_done(struct closure *cl)
+{
+	struct bio_split_hook *s = container_of(cl, struct bio_split_hook, cl);
+
+	s->bio->bi_end_io = s->bi_end_io;
+	s->bio->bi_private = s->bi_private;
+	bio_endio_nodec(s->bio, 0);
+
+	closure_debug_destroy(&s->cl);
+	mempool_free(s, s->p->bio_split_hook);
+}
+
+static void bch_bio_submit_split_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	struct bio_split_hook *s = container_of(cl, struct bio_split_hook, cl);
+
+	if (error)
+		clear_bit(BIO_UPTODATE, &s->bio->bi_flags);
+
+	bio_put(bio);
+	closure_put(cl);
+}
+
+void bch_generic_make_request(struct bio *bio, struct bio_split_pool *p)
+{
+	struct bio_split_hook *s;
+	struct bio *n;
+
+	if (!bio_has_data(bio) && !(bio->bi_rw & REQ_DISCARD))
+		goto submit;
+
+	if (bio_sectors(bio) <= bch_bio_max_sectors(bio))
+		goto submit;
+
+	s = mempool_alloc(p->bio_split_hook, GFP_NOIO);
+	closure_init(&s->cl, NULL);
+
+	s->bio		= bio;
+	s->p		= p;
+	s->bi_end_io	= bio->bi_end_io;
+	s->bi_private	= bio->bi_private;
+	bio_get(bio);
+
+	do {
+		n = bio_next_split(bio, bch_bio_max_sectors(bio),
+				   GFP_NOIO, s->p->bio_split);
+
+		n->bi_end_io	= bch_bio_submit_split_endio;
+		n->bi_private	= &s->cl;
+
+		closure_get(&s->cl);
+		generic_make_request(n);
+	} while (n != bio);
+
+	continue_at(&s->cl, bch_bio_submit_split_done, NULL);
+submit:
+	generic_make_request(bio);
+}
+
+/* Bios with headers */
+
+void bch_bbio_free(struct bio *bio, struct cache_set *c)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+	mempool_free(b, c->bio_meta);
+}
+
+struct bio *bch_bbio_alloc(struct cache_set *c)
+{
+	struct bbio *b = mempool_alloc(c->bio_meta, GFP_NOIO);
+	struct bio *bio = &b->bio;
+
+	bio_init(bio);
+	bio->bi_flags		|= BIO_POOL_NONE << BIO_POOL_OFFSET;
+	bio->bi_max_vecs	 = bucket_pages(c);
+	bio->bi_io_vec		 = bio->bi_inline_vecs;
+
+	return bio;
+}
+
+void __bch_submit_bbio(struct bio *bio, struct cache_set *c)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+
+	bio->bi_iter.bi_sector	= PTR_OFFSET(&b->key, 0);
+	bio->bi_bdev		= PTR_CACHE(c, &b->key, 0)->bdev;
+
+	b->submit_time_us = local_clock_us();
+	closure_bio_submit(bio, bio->bi_private, PTR_CACHE(c, &b->key, 0));
+}
+
+void bch_submit_bbio(struct bio *bio, struct cache_set *c,
+		     struct bkey *k, unsigned ptr)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+	bch_bkey_copy_single_ptr(&b->key, k, ptr);
+	__bch_submit_bbio(bio, c);
+}
+
+/* IO errors */
+
+void bch_count_io_errors(struct cache *ca, int error, const char *m)
+{
+	/*
+	 * The halflife of an error is:
+	 * log2(1/2)/log2(127/128) * refresh ~= 88 * refresh
+	 */
+
+	if (ca->set->error_decay) {
+		unsigned count = atomic_inc_return(&ca->io_count);
+
+		while (count > ca->set->error_decay) {
+			unsigned errors;
+			unsigned old = count;
+			unsigned new = count - ca->set->error_decay;
+
+			/*
+			 * First we subtract refresh from count; each time we
+			 * succesfully do so, we rescale the errors once:
+			 */
+
+			count = atomic_cmpxchg(&ca->io_count, old, new);
+
+			if (count == old) {
+				count = new;
+
+				errors = atomic_read(&ca->io_errors);
+				do {
+					old = errors;
+					new = ((uint64_t) errors * 127) / 128;
+					errors = atomic_cmpxchg(&ca->io_errors,
+								old, new);
+				} while (old != errors);
+			}
+		}
+	}
+
+	if (error) {
+		char buf[BDEVNAME_SIZE];
+		unsigned errors = atomic_add_return(1 << IO_ERROR_SHIFT,
+						    &ca->io_errors);
+		errors >>= IO_ERROR_SHIFT;
+
+		if (errors < ca->set->error_limit)
+			pr_err("%s: IO error on %s, recovering",
+			       bdevname(ca->bdev, buf), m);
+		else
+			bch_cache_set_error(ca->set,
+					    "%s: too many IO errors %s",
+					    bdevname(ca->bdev, buf), m);
+	}
+}
+
+void bch_bbio_count_io_errors(struct cache_set *c, struct bio *bio,
+			      int error, const char *m)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+	struct cache *ca = PTR_CACHE(c, &b->key, 0);
+
+	unsigned threshold = bio->bi_rw & REQ_WRITE
+		? c->congested_write_threshold_us
+		: c->congested_read_threshold_us;
+
+	if (threshold) {
+		unsigned t = local_clock_us();
+
+		int us = t - b->submit_time_us;
+		int congested = atomic_read(&c->congested);
+
+		if (us > (int) threshold) {
+			int ms = us / 1024;
+			c->congested_last_us = t;
+
+			ms = min(ms, CONGESTED_MAX + congested);
+			atomic_sub(ms, &c->congested);
+		} else if (congested < 0)
+			atomic_inc(&c->congested);
+	}
+
+	bch_count_io_errors(ca, error, m);
+}
+
+void bch_bbio_endio(struct cache_set *c, struct bio *bio,
+		    int error, const char *m)
+{
+	struct closure *cl = bio->bi_private;
+
+	bch_bbio_count_io_errors(c, bio, error, m);
+	bio_put(bio);
+	closure_put(cl);
+}
diff --git a/drivers/md/bcache/journal.c b/drivers/md/bcache/journal.c
new file mode 100644
index 00000000000..59e82021b5b
--- /dev/null
+++ b/drivers/md/bcache/journal.c
@@ -0,0 +1,815 @@
+/*
+ * bcache journalling code, for btree insertions
+ *
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+
+#include <trace/events/bcache.h>
+
+/*
+ * Journal replay/recovery:
+ *
+ * This code is all driven from run_cache_set(); we first read the journal
+ * entries, do some other stuff, then we mark all the keys in the journal
+ * entries (same as garbage collection would), then we replay them - reinserting
+ * them into the cache in precisely the same order as they appear in the
+ * journal.
+ *
+ * We only journal keys that go in leaf nodes, which simplifies things quite a
+ * bit.
+ */
+
+static void journal_read_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	closure_put(cl);
+}
+
+static int journal_read_bucket(struct cache *ca, struct list_head *list,
+			       unsigned bucket_index)
+{
+	struct journal_device *ja = &ca->journal;
+	struct bio *bio = &ja->bio;
+
+	struct journal_replay *i;
+	struct jset *j, *data = ca->set->journal.w[0].data;
+	struct closure cl;
+	unsigned len, left, offset = 0;
+	int ret = 0;
+	sector_t bucket = bucket_to_sector(ca->set, ca->sb.d[bucket_index]);
+
+	closure_init_stack(&cl);
+
+	pr_debug("reading %u", bucket_index);
+
+	while (offset < ca->sb.bucket_size) {
+reread:		left = ca->sb.bucket_size - offset;
+		len = min_t(unsigned, left, PAGE_SECTORS << JSET_BITS);
+
+		bio_reset(bio);
+		bio->bi_iter.bi_sector	= bucket + offset;
+		bio->bi_bdev	= ca->bdev;
+		bio->bi_rw	= READ;
+		bio->bi_iter.bi_size	= len << 9;
+
+		bio->bi_end_io	= journal_read_endio;
+		bio->bi_private = &cl;
+		bch_bio_map(bio, data);
+
+		closure_bio_submit(bio, &cl, ca);
+		closure_sync(&cl);
+
+		/* This function could be simpler now since we no longer write
+		 * journal entries that overlap bucket boundaries; this means
+		 * the start of a bucket will always have a valid journal entry
+		 * if it has any journal entries at all.
+		 */
+
+		j = data;
+		while (len) {
+			struct list_head *where;
+			size_t blocks, bytes = set_bytes(j);
+
+			if (j->magic != jset_magic(&ca->sb)) {
+				pr_debug("%u: bad magic", bucket_index);
+				return ret;
+			}
+
+			if (bytes > left << 9 ||
+			    bytes > PAGE_SIZE << JSET_BITS) {
+				pr_info("%u: too big, %zu bytes, offset %u",
+					bucket_index, bytes, offset);
+				return ret;
+			}
+
+			if (bytes > len << 9)
+				goto reread;
+
+			if (j->csum != csum_set(j)) {
+				pr_info("%u: bad csum, %zu bytes, offset %u",
+					bucket_index, bytes, offset);
+				return ret;
+			}
+
+			blocks = set_blocks(j, block_bytes(ca->set));
+
+			while (!list_empty(list)) {
+				i = list_first_entry(list,
+					struct journal_replay, list);
+				if (i->j.seq >= j->last_seq)
+					break;
+				list_del(&i->list);
+				kfree(i);
+			}
+
+			list_for_each_entry_reverse(i, list, list) {
+				if (j->seq == i->j.seq)
+					goto next_set;
+
+				if (j->seq < i->j.last_seq)
+					goto next_set;
+
+				if (j->seq > i->j.seq) {
+					where = &i->list;
+					goto add;
+				}
+			}
+
+			where = list;
+add:
+			i = kmalloc(offsetof(struct journal_replay, j) +
+				    bytes, GFP_KERNEL);
+			if (!i)
+				return -ENOMEM;
+			memcpy(&i->j, j, bytes);
+			list_add(&i->list, where);
+			ret = 1;
+
+			ja->seq[bucket_index] = j->seq;
+next_set:
+			offset	+= blocks * ca->sb.block_size;
+			len	-= blocks * ca->sb.block_size;
+			j = ((void *) j) + blocks * block_bytes(ca);
+		}
+	}
+
+	return ret;
+}
+
+int bch_journal_read(struct cache_set *c, struct list_head *list)
+{
+#define read_bucket(b)							\
+	({								\
+		int ret = journal_read_bucket(ca, list, b);		\
+		__set_bit(b, bitmap);					\
+		if (ret < 0)						\
+			return ret;					\
+		ret;							\
+	})
+
+	struct cache *ca;
+	unsigned iter;
+
+	for_each_cache(ca, c, iter) {
+		struct journal_device *ja = &ca->journal;
+		unsigned long bitmap[SB_JOURNAL_BUCKETS / BITS_PER_LONG];
+		unsigned i, l, r, m;
+		uint64_t seq;
+
+		bitmap_zero(bitmap, SB_JOURNAL_BUCKETS);
+		pr_debug("%u journal buckets", ca->sb.njournal_buckets);
+
+		/*
+		 * Read journal buckets ordered by golden ratio hash to quickly
+		 * find a sequence of buckets with valid journal entries
+		 */
+		for (i = 0; i < ca->sb.njournal_buckets; i++) {
+			l = (i * 2654435769U) % ca->sb.njournal_buckets;
+
+			if (test_bit(l, bitmap))
+				break;
+
+			if (read_bucket(l))
+				goto bsearch;
+		}
+
+		/*
+		 * If that fails, check all the buckets we haven't checked
+		 * already
+		 */
+		pr_debug("falling back to linear search");
+
+		for (l = find_first_zero_bit(bitmap, ca->sb.njournal_buckets);
+		     l < ca->sb.njournal_buckets;
+		     l = find_next_zero_bit(bitmap, ca->sb.njournal_buckets, l + 1))
+			if (read_bucket(l))
+				goto bsearch;
+
+		if (list_empty(list))
+			continue;
+bsearch:
+		/* Binary search */
+		m = r = find_next_bit(bitmap, ca->sb.njournal_buckets, l + 1);
+		pr_debug("starting binary search, l %u r %u", l, r);
+
+		while (l + 1 < r) {
+			seq = list_entry(list->prev, struct journal_replay,
+					 list)->j.seq;
+
+			m = (l + r) >> 1;
+			read_bucket(m);
+
+			if (seq != list_entry(list->prev, struct journal_replay,
+					      list)->j.seq)
+				l = m;
+			else
+				r = m;
+		}
+
+		/*
+		 * Read buckets in reverse order until we stop finding more
+		 * journal entries
+		 */
+		pr_debug("finishing up: m %u njournal_buckets %u",
+			 m, ca->sb.njournal_buckets);
+		l = m;
+
+		while (1) {
+			if (!l--)
+				l = ca->sb.njournal_buckets - 1;
+
+			if (l == m)
+				break;
+
+			if (test_bit(l, bitmap))
+				continue;
+
+			if (!read_bucket(l))
+				break;
+		}
+
+		seq = 0;
+
+		for (i = 0; i < ca->sb.njournal_buckets; i++)
+			if (ja->seq[i] > seq) {
+				seq = ja->seq[i];
+				/*
+				 * When journal_reclaim() goes to allocate for
+				 * the first time, it'll use the bucket after
+				 * ja->cur_idx
+				 */
+				ja->cur_idx = i;
+				ja->last_idx = ja->discard_idx = (i + 1) %
+					ca->sb.njournal_buckets;
+
+			}
+	}
+
+	if (!list_empty(list))
+		c->journal.seq = list_entry(list->prev,
+					    struct journal_replay,
+					    list)->j.seq;
+
+	return 0;
+#undef read_bucket
+}
+
+void bch_journal_mark(struct cache_set *c, struct list_head *list)
+{
+	atomic_t p = { 0 };
+	struct bkey *k;
+	struct journal_replay *i;
+	struct journal *j = &c->journal;
+	uint64_t last = j->seq;
+
+	/*
+	 * journal.pin should never fill up - we never write a journal
+	 * entry when it would fill up. But if for some reason it does, we
+	 * iterate over the list in reverse order so that we can just skip that
+	 * refcount instead of bugging.
+	 */
+
+	list_for_each_entry_reverse(i, list, list) {
+		BUG_ON(last < i->j.seq);
+		i->pin = NULL;
+
+		while (last-- != i->j.seq)
+			if (fifo_free(&j->pin) > 1) {
+				fifo_push_front(&j->pin, p);
+				atomic_set(&fifo_front(&j->pin), 0);
+			}
+
+		if (fifo_free(&j->pin) > 1) {
+			fifo_push_front(&j->pin, p);
+			i->pin = &fifo_front(&j->pin);
+			atomic_set(i->pin, 1);
+		}
+
+		for (k = i->j.start;
+		     k < bset_bkey_last(&i->j);
+		     k = bkey_next(k)) {
+			unsigned j;
+
+			for (j = 0; j < KEY_PTRS(k); j++)
+				if (ptr_available(c, k, j))
+					atomic_inc(&PTR_BUCKET(c, k, j)->pin);
+
+			bch_initial_mark_key(c, 0, k);
+		}
+	}
+}
+
+int bch_journal_replay(struct cache_set *s, struct list_head *list)
+{
+	int ret = 0, keys = 0, entries = 0;
+	struct bkey *k;
+	struct journal_replay *i =
+		list_entry(list->prev, struct journal_replay, list);
+
+	uint64_t start = i->j.last_seq, end = i->j.seq, n = start;
+	struct keylist keylist;
+
+	list_for_each_entry(i, list, list) {
+		BUG_ON(i->pin && atomic_read(i->pin) != 1);
+
+		cache_set_err_on(n != i->j.seq, s,
+"bcache: journal entries %llu-%llu missing! (replaying %llu-%llu)",
+				 n, i->j.seq - 1, start, end);
+
+		for (k = i->j.start;
+		     k < bset_bkey_last(&i->j);
+		     k = bkey_next(k)) {
+			trace_bcache_journal_replay_key(k);
+
+			bch_keylist_init_single(&keylist, k);
+
+			ret = bch_btree_insert(s, &keylist, i->pin, NULL);
+			if (ret)
+				goto err;
+
+			BUG_ON(!bch_keylist_empty(&keylist));
+			keys++;
+
+			cond_resched();
+		}
+
+		if (i->pin)
+			atomic_dec(i->pin);
+		n = i->j.seq + 1;
+		entries++;
+	}
+
+	pr_info("journal replay done, %i keys in %i entries, seq %llu",
+		keys, entries, end);
+err:
+	while (!list_empty(list)) {
+		i = list_first_entry(list, struct journal_replay, list);
+		list_del(&i->list);
+		kfree(i);
+	}
+
+	return ret;
+}
+
+/* Journalling */
+
+static void btree_flush_write(struct cache_set *c)
+{
+	/*
+	 * Try to find the btree node with that references the oldest journal
+	 * entry, best is our current candidate and is locked if non NULL:
+	 */
+	struct btree *b, *best;
+	unsigned i;
+retry:
+	best = NULL;
+
+	for_each_cached_btree(b, c, i)
+		if (btree_current_write(b)->journal) {
+			if (!best)
+				best = b;
+			else if (journal_pin_cmp(c,
+					btree_current_write(best)->journal,
+					btree_current_write(b)->journal)) {
+				best = b;
+			}
+		}
+
+	b = best;
+	if (b) {
+		mutex_lock(&b->write_lock);
+		if (!btree_current_write(b)->journal) {
+			mutex_unlock(&b->write_lock);
+			/* We raced */
+			goto retry;
+		}
+
+		__bch_btree_node_write(b, NULL);
+		mutex_unlock(&b->write_lock);
+	}
+}
+
+#define last_seq(j)	((j)->seq - fifo_used(&(j)->pin) + 1)
+
+static void journal_discard_endio(struct bio *bio, int error)
+{
+	struct journal_device *ja =
+		container_of(bio, struct journal_device, discard_bio);
+	struct cache *ca = container_of(ja, struct cache, journal);
+
+	atomic_set(&ja->discard_in_flight, DISCARD_DONE);
+
+	closure_wake_up(&ca->set->journal.wait);
+	closure_put(&ca->set->cl);
+}
+
+static void journal_discard_work(struct work_struct *work)
+{
+	struct journal_device *ja =
+		container_of(work, struct journal_device, discard_work);
+
+	submit_bio(0, &ja->discard_bio);
+}
+
+static void do_journal_discard(struct cache *ca)
+{
+	struct journal_device *ja = &ca->journal;
+	struct bio *bio = &ja->discard_bio;
+
+	if (!ca->discard) {
+		ja->discard_idx = ja->last_idx;
+		return;
+	}
+
+	switch (atomic_read(&ja->discard_in_flight)) {
+	case DISCARD_IN_FLIGHT:
+		return;
+
+	case DISCARD_DONE:
+		ja->discard_idx = (ja->discard_idx + 1) %
+			ca->sb.njournal_buckets;
+
+		atomic_set(&ja->discard_in_flight, DISCARD_READY);
+		/* fallthrough */
+
+	case DISCARD_READY:
+		if (ja->discard_idx == ja->last_idx)
+			return;
+
+		atomic_set(&ja->discard_in_flight, DISCARD_IN_FLIGHT);
+
+		bio_init(bio);
+		bio->bi_iter.bi_sector	= bucket_to_sector(ca->set,
+						ca->sb.d[ja->discard_idx]);
+		bio->bi_bdev		= ca->bdev;
+		bio->bi_rw		= REQ_WRITE|REQ_DISCARD;
+		bio->bi_max_vecs	= 1;
+		bio->bi_io_vec		= bio->bi_inline_vecs;
+		bio->bi_iter.bi_size	= bucket_bytes(ca);
+		bio->bi_end_io		= journal_discard_endio;
+
+		closure_get(&ca->set->cl);
+		INIT_WORK(&ja->discard_work, journal_discard_work);
+		schedule_work(&ja->discard_work);
+	}
+}
+
+static void journal_reclaim(struct cache_set *c)
+{
+	struct bkey *k = &c->journal.key;
+	struct cache *ca;
+	uint64_t last_seq;
+	unsigned iter, n = 0;
+	atomic_t p;
+
+	while (!atomic_read(&fifo_front(&c->journal.pin)))
+		fifo_pop(&c->journal.pin, p);
+
+	last_seq = last_seq(&c->journal);
+
+	/* Update last_idx */
+
+	for_each_cache(ca, c, iter) {
+		struct journal_device *ja = &ca->journal;
+
+		while (ja->last_idx != ja->cur_idx &&
+		       ja->seq[ja->last_idx] < last_seq)
+			ja->last_idx = (ja->last_idx + 1) %
+				ca->sb.njournal_buckets;
+	}
+
+	for_each_cache(ca, c, iter)
+		do_journal_discard(ca);
+
+	if (c->journal.blocks_free)
+		goto out;
+
+	/*
+	 * Allocate:
+	 * XXX: Sort by free journal space
+	 */
+
+	for_each_cache(ca, c, iter) {
+		struct journal_device *ja = &ca->journal;
+		unsigned next = (ja->cur_idx + 1) % ca->sb.njournal_buckets;
+
+		/* No space available on this device */
+		if (next == ja->discard_idx)
+			continue;
+
+		ja->cur_idx = next;
+		k->ptr[n++] = PTR(0,
+				  bucket_to_sector(c, ca->sb.d[ja->cur_idx]),
+				  ca->sb.nr_this_dev);
+	}
+
+	bkey_init(k);
+	SET_KEY_PTRS(k, n);
+
+	if (n)
+		c->journal.blocks_free = c->sb.bucket_size >> c->block_bits;
+out:
+	if (!journal_full(&c->journal))
+		__closure_wake_up(&c->journal.wait);
+}
+
+void bch_journal_next(struct journal *j)
+{
+	atomic_t p = { 1 };
+
+	j->cur = (j->cur == j->w)
+		? &j->w[1]
+		: &j->w[0];
+
+	/*
+	 * The fifo_push() needs to happen at the same time as j->seq is
+	 * incremented for last_seq() to be calculated correctly
+	 */
+	BUG_ON(!fifo_push(&j->pin, p));
+	atomic_set(&fifo_back(&j->pin), 1);
+
+	j->cur->data->seq	= ++j->seq;
+	j->cur->dirty		= false;
+	j->cur->need_write	= false;
+	j->cur->data->keys	= 0;
+
+	if (fifo_full(&j->pin))
+		pr_debug("journal_pin full (%zu)", fifo_used(&j->pin));
+}
+
+static void journal_write_endio(struct bio *bio, int error)
+{
+	struct journal_write *w = bio->bi_private;
+
+	cache_set_err_on(error, w->c, "journal io error");
+	closure_put(&w->c->journal.io);
+}
+
+static void journal_write(struct closure *);
+
+static void journal_write_done(struct closure *cl)
+{
+	struct journal *j = container_of(cl, struct journal, io);
+	struct journal_write *w = (j->cur == j->w)
+		? &j->w[1]
+		: &j->w[0];
+
+	__closure_wake_up(&w->wait);
+	continue_at_nobarrier(cl, journal_write, system_wq);
+}
+
+static void journal_write_unlock(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
+
+	c->journal.io_in_flight = 0;
+	spin_unlock(&c->journal.lock);
+}
+
+static void journal_write_unlocked(struct closure *cl)
+	__releases(c->journal.lock)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
+	struct cache *ca;
+	struct journal_write *w = c->journal.cur;
+	struct bkey *k = &c->journal.key;
+	unsigned i, sectors = set_blocks(w->data, block_bytes(c)) *
+		c->sb.block_size;
+
+	struct bio *bio;
+	struct bio_list list;
+	bio_list_init(&list);
+
+	if (!w->need_write) {
+		closure_return_with_destructor(cl, journal_write_unlock);
+	} else if (journal_full(&c->journal)) {
+		journal_reclaim(c);
+		spin_unlock(&c->journal.lock);
+
+		btree_flush_write(c);
+		continue_at(cl, journal_write, system_wq);
+	}
+
+	c->journal.blocks_free -= set_blocks(w->data, block_bytes(c));
+
+	w->data->btree_level = c->root->level;
+
+	bkey_copy(&w->data->btree_root, &c->root->key);
+	bkey_copy(&w->data->uuid_bucket, &c->uuid_bucket);
+
+	for_each_cache(ca, c, i)
+		w->data->prio_bucket[ca->sb.nr_this_dev] = ca->prio_buckets[0];
+
+	w->data->magic		= jset_magic(&c->sb);
+	w->data->version	= BCACHE_JSET_VERSION;
+	w->data->last_seq	= last_seq(&c->journal);
+	w->data->csum		= csum_set(w->data);
+
+	for (i = 0; i < KEY_PTRS(k); i++) {
+		ca = PTR_CACHE(c, k, i);
+		bio = &ca->journal.bio;
+
+		atomic_long_add(sectors, &ca->meta_sectors_written);
+
+		bio_reset(bio);
+		bio->bi_iter.bi_sector	= PTR_OFFSET(k, i);
+		bio->bi_bdev	= ca->bdev;
+		bio->bi_rw	= REQ_WRITE|REQ_SYNC|REQ_META|REQ_FLUSH|REQ_FUA;
+		bio->bi_iter.bi_size = sectors << 9;
+
+		bio->bi_end_io	= journal_write_endio;
+		bio->bi_private = w;
+		bch_bio_map(bio, w->data);
+
+		trace_bcache_journal_write(bio);
+		bio_list_add(&list, bio);
+
+		SET_PTR_OFFSET(k, i, PTR_OFFSET(k, i) + sectors);
+
+		ca->journal.seq[ca->journal.cur_idx] = w->data->seq;
+	}
+
+	atomic_dec_bug(&fifo_back(&c->journal.pin));
+	bch_journal_next(&c->journal);
+	journal_reclaim(c);
+
+	spin_unlock(&c->journal.lock);
+
+	while ((bio = bio_list_pop(&list)))
+		closure_bio_submit(bio, cl, c->cache[0]);
+
+	continue_at(cl, journal_write_done, NULL);
+}
+
+static void journal_write(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, journal.io);
+
+	spin_lock(&c->journal.lock);
+	journal_write_unlocked(cl);
+}
+
+static void journal_try_write(struct cache_set *c)
+	__releases(c->journal.lock)
+{
+	struct closure *cl = &c->journal.io;
+	struct journal_write *w = c->journal.cur;
+
+	w->need_write = true;
+
+	if (!c->journal.io_in_flight) {
+		c->journal.io_in_flight = 1;
+		closure_call(cl, journal_write_unlocked, NULL, &c->cl);
+	} else {
+		spin_unlock(&c->journal.lock);
+	}
+}
+
+static struct journal_write *journal_wait_for_write(struct cache_set *c,
+						    unsigned nkeys)
+{
+	size_t sectors;
+	struct closure cl;
+	bool wait = false;
+
+	closure_init_stack(&cl);
+
+	spin_lock(&c->journal.lock);
+
+	while (1) {
+		struct journal_write *w = c->journal.cur;
+
+		sectors = __set_blocks(w->data, w->data->keys + nkeys,
+				       block_bytes(c)) * c->sb.block_size;
+
+		if (sectors <= min_t(size_t,
+				     c->journal.blocks_free * c->sb.block_size,
+				     PAGE_SECTORS << JSET_BITS))
+			return w;
+
+		if (wait)
+			closure_wait(&c->journal.wait, &cl);
+
+		if (!journal_full(&c->journal)) {
+			if (wait)
+				trace_bcache_journal_entry_full(c);
+
+			/*
+			 * XXX: If we were inserting so many keys that they
+			 * won't fit in an _empty_ journal write, we'll
+			 * deadlock. For now, handle this in
+			 * bch_keylist_realloc() - but something to think about.
+			 */
+			BUG_ON(!w->data->keys);
+
+			journal_try_write(c); /* unlocks */
+		} else {
+			if (wait)
+				trace_bcache_journal_full(c);
+
+			journal_reclaim(c);
+			spin_unlock(&c->journal.lock);
+
+			btree_flush_write(c);
+		}
+
+		closure_sync(&cl);
+		spin_lock(&c->journal.lock);
+		wait = true;
+	}
+}
+
+static void journal_write_work(struct work_struct *work)
+{
+	struct cache_set *c = container_of(to_delayed_work(work),
+					   struct cache_set,
+					   journal.work);
+	spin_lock(&c->journal.lock);
+	if (c->journal.cur->dirty)
+		journal_try_write(c);
+	else
+		spin_unlock(&c->journal.lock);
+}
+
+/*
+ * Entry point to the journalling code - bio_insert() and btree_invalidate()
+ * pass bch_journal() a list of keys to be journalled, and then
+ * bch_journal() hands those same keys off to btree_insert_async()
+ */
+
+atomic_t *bch_journal(struct cache_set *c,
+		      struct keylist *keys,
+		      struct closure *parent)
+{
+	struct journal_write *w;
+	atomic_t *ret;
+
+	if (!CACHE_SYNC(&c->sb))
+		return NULL;
+
+	w = journal_wait_for_write(c, bch_keylist_nkeys(keys));
+
+	memcpy(bset_bkey_last(w->data), keys->keys, bch_keylist_bytes(keys));
+	w->data->keys += bch_keylist_nkeys(keys);
+
+	ret = &fifo_back(&c->journal.pin);
+	atomic_inc(ret);
+
+	if (parent) {
+		closure_wait(&w->wait, parent);
+		journal_try_write(c);
+	} else if (!w->dirty) {
+		w->dirty = true;
+		schedule_delayed_work(&c->journal.work,
+				      msecs_to_jiffies(c->journal_delay_ms));
+		spin_unlock(&c->journal.lock);
+	} else {
+		spin_unlock(&c->journal.lock);
+	}
+
+
+	return ret;
+}
+
+void bch_journal_meta(struct cache_set *c, struct closure *cl)
+{
+	struct keylist keys;
+	atomic_t *ref;
+
+	bch_keylist_init(&keys);
+
+	ref = bch_journal(c, &keys, cl);
+	if (ref)
+		atomic_dec_bug(ref);
+}
+
+void bch_journal_free(struct cache_set *c)
+{
+	free_pages((unsigned long) c->journal.w[1].data, JSET_BITS);
+	free_pages((unsigned long) c->journal.w[0].data, JSET_BITS);
+	free_fifo(&c->journal.pin);
+}
+
+int bch_journal_alloc(struct cache_set *c)
+{
+	struct journal *j = &c->journal;
+
+	spin_lock_init(&j->lock);
+	INIT_DELAYED_WORK(&j->work, journal_write_work);
+
+	c->journal_delay_ms = 100;
+
+	j->w[0].c = c;
+	j->w[1].c = c;
+
+	if (!(init_fifo(&j->pin, JOURNAL_PIN, GFP_KERNEL)) ||
+	    !(j->w[0].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)) ||
+	    !(j->w[1].data = (void *) __get_free_pages(GFP_KERNEL, JSET_BITS)))
+		return -ENOMEM;
+
+	return 0;
+}
diff --git a/drivers/md/bcache/journal.h b/drivers/md/bcache/journal.h
new file mode 100644
index 00000000000..e3c39457afb
--- /dev/null
+++ b/drivers/md/bcache/journal.h
@@ -0,0 +1,179 @@
+#ifndef _BCACHE_JOURNAL_H
+#define _BCACHE_JOURNAL_H
+
+/*
+ * THE JOURNAL:
+ *
+ * The journal is treated as a circular buffer of buckets - a journal entry
+ * never spans two buckets. This means (not implemented yet) we can resize the
+ * journal at runtime, and will be needed for bcache on raw flash support.
+ *
+ * Journal entries contain a list of keys, ordered by the time they were
+ * inserted; thus journal replay just has to reinsert the keys.
+ *
+ * We also keep some things in the journal header that are logically part of the
+ * superblock - all the things that are frequently updated. This is for future
+ * bcache on raw flash support; the superblock (which will become another
+ * journal) can't be moved or wear leveled, so it contains just enough
+ * information to find the main journal, and the superblock only has to be
+ * rewritten when we want to move/wear level the main journal.
+ *
+ * Currently, we don't journal BTREE_REPLACE operations - this will hopefully be
+ * fixed eventually. This isn't a bug - BTREE_REPLACE is used for insertions
+ * from cache misses, which don't have to be journaled, and for writeback and
+ * moving gc we work around it by flushing the btree to disk before updating the
+ * gc information. But it is a potential issue with incremental garbage
+ * collection, and it's fragile.
+ *
+ * OPEN JOURNAL ENTRIES:
+ *
+ * Each journal entry contains, in the header, the sequence number of the last
+ * journal entry still open - i.e. that has keys that haven't been flushed to
+ * disk in the btree.
+ *
+ * We track this by maintaining a refcount for every open journal entry, in a
+ * fifo; each entry in the fifo corresponds to a particular journal
+ * entry/sequence number. When the refcount at the tail of the fifo goes to
+ * zero, we pop it off - thus, the size of the fifo tells us the number of open
+ * journal entries
+ *
+ * We take a refcount on a journal entry when we add some keys to a journal
+ * entry that we're going to insert (held by struct btree_op), and then when we
+ * insert those keys into the btree the btree write we're setting up takes a
+ * copy of that refcount (held by struct btree_write). That refcount is dropped
+ * when the btree write completes.
+ *
+ * A struct btree_write can only hold a refcount on a single journal entry, but
+ * might contain keys for many journal entries - we handle this by making sure
+ * it always has a refcount on the _oldest_ journal entry of all the journal
+ * entries it has keys for.
+ *
+ * JOURNAL RECLAIM:
+ *
+ * As mentioned previously, our fifo of refcounts tells us the number of open
+ * journal entries; from that and the current journal sequence number we compute
+ * last_seq - the oldest journal entry we still need. We write last_seq in each
+ * journal entry, and we also have to keep track of where it exists on disk so
+ * we don't overwrite it when we loop around the journal.
+ *
+ * To do that we track, for each journal bucket, the sequence number of the
+ * newest journal entry it contains - if we don't need that journal entry we
+ * don't need anything in that bucket anymore. From that we track the last
+ * journal bucket we still need; all this is tracked in struct journal_device
+ * and updated by journal_reclaim().
+ *
+ * JOURNAL FILLING UP:
+ *
+ * There are two ways the journal could fill up; either we could run out of
+ * space to write to, or we could have too many open journal entries and run out
+ * of room in the fifo of refcounts. Since those refcounts are decremented
+ * without any locking we can't safely resize that fifo, so we handle it the
+ * same way.
+ *
+ * If the journal fills up, we start flushing dirty btree nodes until we can
+ * allocate space for a journal write again - preferentially flushing btree
+ * nodes that are pinning the oldest journal entries first.
+ */
+
+/*
+ * Only used for holding the journal entries we read in btree_journal_read()
+ * during cache_registration
+ */
+struct journal_replay {
+	struct list_head	list;
+	atomic_t		*pin;
+	struct jset		j;
+};
+
+/*
+ * We put two of these in struct journal; we used them for writes to the
+ * journal that are being staged or in flight.
+ */
+struct journal_write {
+	struct jset		*data;
+#define JSET_BITS		3
+
+	struct cache_set	*c;
+	struct closure_waitlist	wait;
+	bool			dirty;
+	bool			need_write;
+};
+
+/* Embedded in struct cache_set */
+struct journal {
+	spinlock_t		lock;
+	/* used when waiting because the journal was full */
+	struct closure_waitlist	wait;
+	struct closure		io;
+	int			io_in_flight;
+	struct delayed_work	work;
+
+	/* Number of blocks free in the bucket(s) we're currently writing to */
+	unsigned		blocks_free;
+	uint64_t		seq;
+	DECLARE_FIFO(atomic_t, pin);
+
+	BKEY_PADDED(key);
+
+	struct journal_write	w[2], *cur;
+};
+
+/*
+ * Embedded in struct cache. First three fields refer to the array of journal
+ * buckets, in cache_sb.
+ */
+struct journal_device {
+	/*
+	 * For each journal bucket, contains the max sequence number of the
+	 * journal writes it contains - so we know when a bucket can be reused.
+	 */
+	uint64_t		seq[SB_JOURNAL_BUCKETS];
+
+	/* Journal bucket we're currently writing to */
+	unsigned		cur_idx;
+
+	/* Last journal bucket that still contains an open journal entry */
+	unsigned		last_idx;
+
+	/* Next journal bucket to be discarded */
+	unsigned		discard_idx;
+
+#define DISCARD_READY		0
+#define DISCARD_IN_FLIGHT	1
+#define DISCARD_DONE		2
+	/* 1 - discard in flight, -1 - discard completed */
+	atomic_t		discard_in_flight;
+
+	struct work_struct	discard_work;
+	struct bio		discard_bio;
+	struct bio_vec		discard_bv;
+
+	/* Bio for journal reads/writes to this device */
+	struct bio		bio;
+	struct bio_vec		bv[8];
+};
+
+#define journal_pin_cmp(c, l, r)				\
+	(fifo_idx(&(c)->journal.pin, (l)) > fifo_idx(&(c)->journal.pin, (r)))
+
+#define JOURNAL_PIN	20000
+
+#define journal_full(j)						\
+	(!(j)->blocks_free || fifo_free(&(j)->pin) <= 1)
+
+struct closure;
+struct cache_set;
+struct btree_op;
+struct keylist;
+
+atomic_t *bch_journal(struct cache_set *, struct keylist *, struct closure *);
+void bch_journal_next(struct journal *);
+void bch_journal_mark(struct cache_set *, struct list_head *);
+void bch_journal_meta(struct cache_set *, struct closure *);
+int bch_journal_read(struct cache_set *, struct list_head *);
+int bch_journal_replay(struct cache_set *, struct list_head *);
+
+void bch_journal_free(struct cache_set *);
+int bch_journal_alloc(struct cache_set *);
+
+#endif /* _BCACHE_JOURNAL_H */
diff --git a/drivers/md/bcache/movinggc.c b/drivers/md/bcache/movinggc.c
new file mode 100644
index 00000000000..cd7490311e5
--- /dev/null
+++ b/drivers/md/bcache/movinggc.c
@@ -0,0 +1,256 @@
+/*
+ * Moving/copying garbage collector
+ *
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "request.h"
+
+#include <trace/events/bcache.h>
+
+struct moving_io {
+	struct closure		cl;
+	struct keybuf_key	*w;
+	struct data_insert_op	op;
+	struct bbio		bio;
+};
+
+static bool moving_pred(struct keybuf *buf, struct bkey *k)
+{
+	struct cache_set *c = container_of(buf, struct cache_set,
+					   moving_gc_keys);
+	unsigned i;
+
+	for (i = 0; i < KEY_PTRS(k); i++)
+		if (ptr_available(c, k, i) &&
+		    GC_MOVE(PTR_BUCKET(c, k, i)))
+			return true;
+
+	return false;
+}
+
+/* Moving GC - IO loop */
+
+static void moving_io_destructor(struct closure *cl)
+{
+	struct moving_io *io = container_of(cl, struct moving_io, cl);
+	kfree(io);
+}
+
+static void write_moving_finish(struct closure *cl)
+{
+	struct moving_io *io = container_of(cl, struct moving_io, cl);
+	struct bio *bio = &io->bio.bio;
+	struct bio_vec *bv;
+	int i;
+
+	bio_for_each_segment_all(bv, bio, i)
+		__free_page(bv->bv_page);
+
+	if (io->op.replace_collision)
+		trace_bcache_gc_copy_collision(&io->w->key);
+
+	bch_keybuf_del(&io->op.c->moving_gc_keys, io->w);
+
+	up(&io->op.c->moving_in_flight);
+
+	closure_return_with_destructor(cl, moving_io_destructor);
+}
+
+static void read_moving_endio(struct bio *bio, int error)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+	struct moving_io *io = container_of(bio->bi_private,
+					    struct moving_io, cl);
+
+	if (error)
+		io->op.error = error;
+	else if (!KEY_DIRTY(&b->key) &&
+		 ptr_stale(io->op.c, &b->key, 0)) {
+		io->op.error = -EINTR;
+	}
+
+	bch_bbio_endio(io->op.c, bio, error, "reading data to move");
+}
+
+static void moving_init(struct moving_io *io)
+{
+	struct bio *bio = &io->bio.bio;
+
+	bio_init(bio);
+	bio_get(bio);
+	bio_set_prio(bio, IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0));
+
+	bio->bi_iter.bi_size	= KEY_SIZE(&io->w->key) << 9;
+	bio->bi_max_vecs	= DIV_ROUND_UP(KEY_SIZE(&io->w->key),
+					       PAGE_SECTORS);
+	bio->bi_private		= &io->cl;
+	bio->bi_io_vec		= bio->bi_inline_vecs;
+	bch_bio_map(bio, NULL);
+}
+
+static void write_moving(struct closure *cl)
+{
+	struct moving_io *io = container_of(cl, struct moving_io, cl);
+	struct data_insert_op *op = &io->op;
+
+	if (!op->error) {
+		moving_init(io);
+
+		io->bio.bio.bi_iter.bi_sector = KEY_START(&io->w->key);
+		op->write_prio		= 1;
+		op->bio			= &io->bio.bio;
+
+		op->writeback		= KEY_DIRTY(&io->w->key);
+		op->csum		= KEY_CSUM(&io->w->key);
+
+		bkey_copy(&op->replace_key, &io->w->key);
+		op->replace		= true;
+
+		closure_call(&op->cl, bch_data_insert, NULL, cl);
+	}
+
+	continue_at(cl, write_moving_finish, op->wq);
+}
+
+static void read_moving_submit(struct closure *cl)
+{
+	struct moving_io *io = container_of(cl, struct moving_io, cl);
+	struct bio *bio = &io->bio.bio;
+
+	bch_submit_bbio(bio, io->op.c, &io->w->key, 0);
+
+	continue_at(cl, write_moving, io->op.wq);
+}
+
+static void read_moving(struct cache_set *c)
+{
+	struct keybuf_key *w;
+	struct moving_io *io;
+	struct bio *bio;
+	struct closure cl;
+
+	closure_init_stack(&cl);
+
+	/* XXX: if we error, background writeback could stall indefinitely */
+
+	while (!test_bit(CACHE_SET_STOPPING, &c->flags)) {
+		w = bch_keybuf_next_rescan(c, &c->moving_gc_keys,
+					   &MAX_KEY, moving_pred);
+		if (!w)
+			break;
+
+		if (ptr_stale(c, &w->key, 0)) {
+			bch_keybuf_del(&c->moving_gc_keys, w);
+			continue;
+		}
+
+		io = kzalloc(sizeof(struct moving_io) + sizeof(struct bio_vec)
+			     * DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS),
+			     GFP_KERNEL);
+		if (!io)
+			goto err;
+
+		w->private	= io;
+		io->w		= w;
+		io->op.inode	= KEY_INODE(&w->key);
+		io->op.c	= c;
+		io->op.wq	= c->moving_gc_wq;
+
+		moving_init(io);
+		bio = &io->bio.bio;
+
+		bio->bi_rw	= READ;
+		bio->bi_end_io	= read_moving_endio;
+
+		if (bio_alloc_pages(bio, GFP_KERNEL))
+			goto err;
+
+		trace_bcache_gc_copy(&w->key);
+
+		down(&c->moving_in_flight);
+		closure_call(&io->cl, read_moving_submit, NULL, &cl);
+	}
+
+	if (0) {
+err:		if (!IS_ERR_OR_NULL(w->private))
+			kfree(w->private);
+
+		bch_keybuf_del(&c->moving_gc_keys, w);
+	}
+
+	closure_sync(&cl);
+}
+
+static bool bucket_cmp(struct bucket *l, struct bucket *r)
+{
+	return GC_SECTORS_USED(l) < GC_SECTORS_USED(r);
+}
+
+static unsigned bucket_heap_top(struct cache *ca)
+{
+	struct bucket *b;
+	return (b = heap_peek(&ca->heap)) ? GC_SECTORS_USED(b) : 0;
+}
+
+void bch_moving_gc(struct cache_set *c)
+{
+	struct cache *ca;
+	struct bucket *b;
+	unsigned i;
+
+	if (!c->copy_gc_enabled)
+		return;
+
+	mutex_lock(&c->bucket_lock);
+
+	for_each_cache(ca, c, i) {
+		unsigned sectors_to_move = 0;
+		unsigned reserve_sectors = ca->sb.bucket_size *
+			fifo_used(&ca->free[RESERVE_MOVINGGC]);
+
+		ca->heap.used = 0;
+
+		for_each_bucket(b, ca) {
+			if (GC_MARK(b) == GC_MARK_METADATA ||
+			    !GC_SECTORS_USED(b) ||
+			    GC_SECTORS_USED(b) == ca->sb.bucket_size ||
+			    atomic_read(&b->pin))
+				continue;
+
+			if (!heap_full(&ca->heap)) {
+				sectors_to_move += GC_SECTORS_USED(b);
+				heap_add(&ca->heap, b, bucket_cmp);
+			} else if (bucket_cmp(b, heap_peek(&ca->heap))) {
+				sectors_to_move -= bucket_heap_top(ca);
+				sectors_to_move += GC_SECTORS_USED(b);
+
+				ca->heap.data[0] = b;
+				heap_sift(&ca->heap, 0, bucket_cmp);
+			}
+		}
+
+		while (sectors_to_move > reserve_sectors) {
+			heap_pop(&ca->heap, b, bucket_cmp);
+			sectors_to_move -= GC_SECTORS_USED(b);
+		}
+
+		while (heap_pop(&ca->heap, b, bucket_cmp))
+			SET_GC_MOVE(b, 1);
+	}
+
+	mutex_unlock(&c->bucket_lock);
+
+	c->moving_gc_keys.last_scanned = ZERO_KEY;
+
+	read_moving(c);
+}
+
+void bch_moving_init_cache_set(struct cache_set *c)
+{
+	bch_keybuf_init(&c->moving_gc_keys);
+	sema_init(&c->moving_in_flight, 64);
+}
diff --git a/drivers/md/bcache/request.c b/drivers/md/bcache/request.c
new file mode 100644
index 00000000000..15fff4f68a7
--- /dev/null
+++ b/drivers/md/bcache/request.c
@@ -0,0 +1,1159 @@
+/*
+ * Main bcache entry point - handle a read or a write request and decide what to
+ * do with it; the make_request functions are called by the block layer.
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "request.h"
+#include "writeback.h"
+
+#include <linux/module.h>
+#include <linux/hash.h>
+#include <linux/random.h>
+
+#include <trace/events/bcache.h>
+
+#define CUTOFF_CACHE_ADD	95
+#define CUTOFF_CACHE_READA	90
+
+struct kmem_cache *bch_search_cache;
+
+static void bch_data_insert_start(struct closure *);
+
+static unsigned cache_mode(struct cached_dev *dc, struct bio *bio)
+{
+	return BDEV_CACHE_MODE(&dc->sb);
+}
+
+static bool verify(struct cached_dev *dc, struct bio *bio)
+{
+	return dc->verify;
+}
+
+static void bio_csum(struct bio *bio, struct bkey *k)
+{
+	struct bio_vec bv;
+	struct bvec_iter iter;
+	uint64_t csum = 0;
+
+	bio_for_each_segment(bv, bio, iter) {
+		void *d = kmap(bv.bv_page) + bv.bv_offset;
+		csum = bch_crc64_update(csum, d, bv.bv_len);
+		kunmap(bv.bv_page);
+	}
+
+	k->ptr[KEY_PTRS(k)] = csum & (~0ULL >> 1);
+}
+
+/* Insert data into cache */
+
+static void bch_data_insert_keys(struct closure *cl)
+{
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+	atomic_t *journal_ref = NULL;
+	struct bkey *replace_key = op->replace ? &op->replace_key : NULL;
+	int ret;
+
+	/*
+	 * If we're looping, might already be waiting on
+	 * another journal write - can't wait on more than one journal write at
+	 * a time
+	 *
+	 * XXX: this looks wrong
+	 */
+#if 0
+	while (atomic_read(&s->cl.remaining) & CLOSURE_WAITING)
+		closure_sync(&s->cl);
+#endif
+
+	if (!op->replace)
+		journal_ref = bch_journal(op->c, &op->insert_keys,
+					  op->flush_journal ? cl : NULL);
+
+	ret = bch_btree_insert(op->c, &op->insert_keys,
+			       journal_ref, replace_key);
+	if (ret == -ESRCH) {
+		op->replace_collision = true;
+	} else if (ret) {
+		op->error		= -ENOMEM;
+		op->insert_data_done	= true;
+	}
+
+	if (journal_ref)
+		atomic_dec_bug(journal_ref);
+
+	if (!op->insert_data_done)
+		continue_at(cl, bch_data_insert_start, op->wq);
+
+	bch_keylist_free(&op->insert_keys);
+	closure_return(cl);
+}
+
+static int bch_keylist_realloc(struct keylist *l, unsigned u64s,
+			       struct cache_set *c)
+{
+	size_t oldsize = bch_keylist_nkeys(l);
+	size_t newsize = oldsize + u64s;
+
+	/*
+	 * The journalling code doesn't handle the case where the keys to insert
+	 * is bigger than an empty write: If we just return -ENOMEM here,
+	 * bio_insert() and bio_invalidate() will insert the keys created so far
+	 * and finish the rest when the keylist is empty.
+	 */
+	if (newsize * sizeof(uint64_t) > block_bytes(c) - sizeof(struct jset))
+		return -ENOMEM;
+
+	return __bch_keylist_realloc(l, u64s);
+}
+
+static void bch_data_invalidate(struct closure *cl)
+{
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+	struct bio *bio = op->bio;
+
+	pr_debug("invalidating %i sectors from %llu",
+		 bio_sectors(bio), (uint64_t) bio->bi_iter.bi_sector);
+
+	while (bio_sectors(bio)) {
+		unsigned sectors = min(bio_sectors(bio),
+				       1U << (KEY_SIZE_BITS - 1));
+
+		if (bch_keylist_realloc(&op->insert_keys, 2, op->c))
+			goto out;
+
+		bio->bi_iter.bi_sector	+= sectors;
+		bio->bi_iter.bi_size	-= sectors << 9;
+
+		bch_keylist_add(&op->insert_keys,
+				&KEY(op->inode, bio->bi_iter.bi_sector, sectors));
+	}
+
+	op->insert_data_done = true;
+	bio_put(bio);
+out:
+	continue_at(cl, bch_data_insert_keys, op->wq);
+}
+
+static void bch_data_insert_error(struct closure *cl)
+{
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+
+	/*
+	 * Our data write just errored, which means we've got a bunch of keys to
+	 * insert that point to data that wasn't succesfully written.
+	 *
+	 * We don't have to insert those keys but we still have to invalidate
+	 * that region of the cache - so, if we just strip off all the pointers
+	 * from the keys we'll accomplish just that.
+	 */
+
+	struct bkey *src = op->insert_keys.keys, *dst = op->insert_keys.keys;
+
+	while (src != op->insert_keys.top) {
+		struct bkey *n = bkey_next(src);
+
+		SET_KEY_PTRS(src, 0);
+		memmove(dst, src, bkey_bytes(src));
+
+		dst = bkey_next(dst);
+		src = n;
+	}
+
+	op->insert_keys.top = dst;
+
+	bch_data_insert_keys(cl);
+}
+
+static void bch_data_insert_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+
+	if (error) {
+		/* TODO: We could try to recover from this. */
+		if (op->writeback)
+			op->error = error;
+		else if (!op->replace)
+			set_closure_fn(cl, bch_data_insert_error, op->wq);
+		else
+			set_closure_fn(cl, NULL, NULL);
+	}
+
+	bch_bbio_endio(op->c, bio, error, "writing data to cache");
+}
+
+static void bch_data_insert_start(struct closure *cl)
+{
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+	struct bio *bio = op->bio, *n;
+
+	if (atomic_sub_return(bio_sectors(bio), &op->c->sectors_to_gc) < 0) {
+		set_gc_sectors(op->c);
+		wake_up_gc(op->c);
+	}
+
+	if (op->bypass)
+		return bch_data_invalidate(cl);
+
+	/*
+	 * Journal writes are marked REQ_FLUSH; if the original write was a
+	 * flush, it'll wait on the journal write.
+	 */
+	bio->bi_rw &= ~(REQ_FLUSH|REQ_FUA);
+
+	do {
+		unsigned i;
+		struct bkey *k;
+		struct bio_set *split = op->c->bio_split;
+
+		/* 1 for the device pointer and 1 for the chksum */
+		if (bch_keylist_realloc(&op->insert_keys,
+					3 + (op->csum ? 1 : 0),
+					op->c))
+			continue_at(cl, bch_data_insert_keys, op->wq);
+
+		k = op->insert_keys.top;
+		bkey_init(k);
+		SET_KEY_INODE(k, op->inode);
+		SET_KEY_OFFSET(k, bio->bi_iter.bi_sector);
+
+		if (!bch_alloc_sectors(op->c, k, bio_sectors(bio),
+				       op->write_point, op->write_prio,
+				       op->writeback))
+			goto err;
+
+		n = bio_next_split(bio, KEY_SIZE(k), GFP_NOIO, split);
+
+		n->bi_end_io	= bch_data_insert_endio;
+		n->bi_private	= cl;
+
+		if (op->writeback) {
+			SET_KEY_DIRTY(k, true);
+
+			for (i = 0; i < KEY_PTRS(k); i++)
+				SET_GC_MARK(PTR_BUCKET(op->c, k, i),
+					    GC_MARK_DIRTY);
+		}
+
+		SET_KEY_CSUM(k, op->csum);
+		if (KEY_CSUM(k))
+			bio_csum(n, k);
+
+		trace_bcache_cache_insert(k);
+		bch_keylist_push(&op->insert_keys);
+
+		n->bi_rw |= REQ_WRITE;
+		bch_submit_bbio(n, op->c, k, 0);
+	} while (n != bio);
+
+	op->insert_data_done = true;
+	continue_at(cl, bch_data_insert_keys, op->wq);
+err:
+	/* bch_alloc_sectors() blocks if s->writeback = true */
+	BUG_ON(op->writeback);
+
+	/*
+	 * But if it's not a writeback write we'd rather just bail out if
+	 * there aren't any buckets ready to write to - it might take awhile and
+	 * we might be starving btree writes for gc or something.
+	 */
+
+	if (!op->replace) {
+		/*
+		 * Writethrough write: We can't complete the write until we've
+		 * updated the index. But we don't want to delay the write while
+		 * we wait for buckets to be freed up, so just invalidate the
+		 * rest of the write.
+		 */
+		op->bypass = true;
+		return bch_data_invalidate(cl);
+	} else {
+		/*
+		 * From a cache miss, we can just insert the keys for the data
+		 * we have written or bail out if we didn't do anything.
+		 */
+		op->insert_data_done = true;
+		bio_put(bio);
+
+		if (!bch_keylist_empty(&op->insert_keys))
+			continue_at(cl, bch_data_insert_keys, op->wq);
+		else
+			closure_return(cl);
+	}
+}
+
+/**
+ * bch_data_insert - stick some data in the cache
+ *
+ * This is the starting point for any data to end up in a cache device; it could
+ * be from a normal write, or a writeback write, or a write to a flash only
+ * volume - it's also used by the moving garbage collector to compact data in
+ * mostly empty buckets.
+ *
+ * It first writes the data to the cache, creating a list of keys to be inserted
+ * (if the data had to be fragmented there will be multiple keys); after the
+ * data is written it calls bch_journal, and after the keys have been added to
+ * the next journal write they're inserted into the btree.
+ *
+ * It inserts the data in s->cache_bio; bi_sector is used for the key offset,
+ * and op->inode is used for the key inode.
+ *
+ * If s->bypass is true, instead of inserting the data it invalidates the
+ * region of the cache represented by s->cache_bio and op->inode.
+ */
+void bch_data_insert(struct closure *cl)
+{
+	struct data_insert_op *op = container_of(cl, struct data_insert_op, cl);
+
+	trace_bcache_write(op->bio, op->writeback, op->bypass);
+
+	bch_keylist_init(&op->insert_keys);
+	bio_get(op->bio);
+	bch_data_insert_start(cl);
+}
+
+/* Congested? */
+
+unsigned bch_get_congested(struct cache_set *c)
+{
+	int i;
+	long rand;
+
+	if (!c->congested_read_threshold_us &&
+	    !c->congested_write_threshold_us)
+		return 0;
+
+	i = (local_clock_us() - c->congested_last_us) / 1024;
+	if (i < 0)
+		return 0;
+
+	i += atomic_read(&c->congested);
+	if (i >= 0)
+		return 0;
+
+	i += CONGESTED_MAX;
+
+	if (i > 0)
+		i = fract_exp_two(i, 6);
+
+	rand = get_random_int();
+	i -= bitmap_weight(&rand, BITS_PER_LONG);
+
+	return i > 0 ? i : 1;
+}
+
+static void add_sequential(struct task_struct *t)
+{
+	ewma_add(t->sequential_io_avg,
+		 t->sequential_io, 8, 0);
+
+	t->sequential_io = 0;
+}
+
+static struct hlist_head *iohash(struct cached_dev *dc, uint64_t k)
+{
+	return &dc->io_hash[hash_64(k, RECENT_IO_BITS)];
+}
+
+static bool check_should_bypass(struct cached_dev *dc, struct bio *bio)
+{
+	struct cache_set *c = dc->disk.c;
+	unsigned mode = cache_mode(dc, bio);
+	unsigned sectors, congested = bch_get_congested(c);
+	struct task_struct *task = current;
+	struct io *i;
+
+	if (test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) ||
+	    c->gc_stats.in_use > CUTOFF_CACHE_ADD ||
+	    (bio->bi_rw & REQ_DISCARD))
+		goto skip;
+
+	if (mode == CACHE_MODE_NONE ||
+	    (mode == CACHE_MODE_WRITEAROUND &&
+	     (bio->bi_rw & REQ_WRITE)))
+		goto skip;
+
+	if (bio->bi_iter.bi_sector & (c->sb.block_size - 1) ||
+	    bio_sectors(bio) & (c->sb.block_size - 1)) {
+		pr_debug("skipping unaligned io");
+		goto skip;
+	}
+
+	if (bypass_torture_test(dc)) {
+		if ((get_random_int() & 3) == 3)
+			goto skip;
+		else
+			goto rescale;
+	}
+
+	if (!congested && !dc->sequential_cutoff)
+		goto rescale;
+
+	if (!congested &&
+	    mode == CACHE_MODE_WRITEBACK &&
+	    (bio->bi_rw & REQ_WRITE) &&
+	    (bio->bi_rw & REQ_SYNC))
+		goto rescale;
+
+	spin_lock(&dc->io_lock);
+
+	hlist_for_each_entry(i, iohash(dc, bio->bi_iter.bi_sector), hash)
+		if (i->last == bio->bi_iter.bi_sector &&
+		    time_before(jiffies, i->jiffies))
+			goto found;
+
+	i = list_first_entry(&dc->io_lru, struct io, lru);
+
+	add_sequential(task);
+	i->sequential = 0;
+found:
+	if (i->sequential + bio->bi_iter.bi_size > i->sequential)
+		i->sequential	+= bio->bi_iter.bi_size;
+
+	i->last			 = bio_end_sector(bio);
+	i->jiffies		 = jiffies + msecs_to_jiffies(5000);
+	task->sequential_io	 = i->sequential;
+
+	hlist_del(&i->hash);
+	hlist_add_head(&i->hash, iohash(dc, i->last));
+	list_move_tail(&i->lru, &dc->io_lru);
+
+	spin_unlock(&dc->io_lock);
+
+	sectors = max(task->sequential_io,
+		      task->sequential_io_avg) >> 9;
+
+	if (dc->sequential_cutoff &&
+	    sectors >= dc->sequential_cutoff >> 9) {
+		trace_bcache_bypass_sequential(bio);
+		goto skip;
+	}
+
+	if (congested && sectors >= congested) {
+		trace_bcache_bypass_congested(bio);
+		goto skip;
+	}
+
+rescale:
+	bch_rescale_priorities(c, bio_sectors(bio));
+	return false;
+skip:
+	bch_mark_sectors_bypassed(c, dc, bio_sectors(bio));
+	return true;
+}
+
+/* Cache lookup */
+
+struct search {
+	/* Stack frame for bio_complete */
+	struct closure		cl;
+
+	struct bbio		bio;
+	struct bio		*orig_bio;
+	struct bio		*cache_miss;
+	struct bcache_device	*d;
+
+	unsigned		insert_bio_sectors;
+	unsigned		recoverable:1;
+	unsigned		write:1;
+	unsigned		read_dirty_data:1;
+
+	unsigned long		start_time;
+
+	struct btree_op		op;
+	struct data_insert_op	iop;
+};
+
+static void bch_cache_read_endio(struct bio *bio, int error)
+{
+	struct bbio *b = container_of(bio, struct bbio, bio);
+	struct closure *cl = bio->bi_private;
+	struct search *s = container_of(cl, struct search, cl);
+
+	/*
+	 * If the bucket was reused while our bio was in flight, we might have
+	 * read the wrong data. Set s->error but not error so it doesn't get
+	 * counted against the cache device, but we'll still reread the data
+	 * from the backing device.
+	 */
+
+	if (error)
+		s->iop.error = error;
+	else if (!KEY_DIRTY(&b->key) &&
+		 ptr_stale(s->iop.c, &b->key, 0)) {
+		atomic_long_inc(&s->iop.c->cache_read_races);
+		s->iop.error = -EINTR;
+	}
+
+	bch_bbio_endio(s->iop.c, bio, error, "reading from cache");
+}
+
+/*
+ * Read from a single key, handling the initial cache miss if the key starts in
+ * the middle of the bio
+ */
+static int cache_lookup_fn(struct btree_op *op, struct btree *b, struct bkey *k)
+{
+	struct search *s = container_of(op, struct search, op);
+	struct bio *n, *bio = &s->bio.bio;
+	struct bkey *bio_key;
+	unsigned ptr;
+
+	if (bkey_cmp(k, &KEY(s->iop.inode, bio->bi_iter.bi_sector, 0)) <= 0)
+		return MAP_CONTINUE;
+
+	if (KEY_INODE(k) != s->iop.inode ||
+	    KEY_START(k) > bio->bi_iter.bi_sector) {
+		unsigned bio_sectors = bio_sectors(bio);
+		unsigned sectors = KEY_INODE(k) == s->iop.inode
+			? min_t(uint64_t, INT_MAX,
+				KEY_START(k) - bio->bi_iter.bi_sector)
+			: INT_MAX;
+
+		int ret = s->d->cache_miss(b, s, bio, sectors);
+		if (ret != MAP_CONTINUE)
+			return ret;
+
+		/* if this was a complete miss we shouldn't get here */
+		BUG_ON(bio_sectors <= sectors);
+	}
+
+	if (!KEY_SIZE(k))
+		return MAP_CONTINUE;
+
+	/* XXX: figure out best pointer - for multiple cache devices */
+	ptr = 0;
+
+	PTR_BUCKET(b->c, k, ptr)->prio = INITIAL_PRIO;
+
+	if (KEY_DIRTY(k))
+		s->read_dirty_data = true;
+
+	n = bio_next_split(bio, min_t(uint64_t, INT_MAX,
+				      KEY_OFFSET(k) - bio->bi_iter.bi_sector),
+			   GFP_NOIO, s->d->bio_split);
+
+	bio_key = &container_of(n, struct bbio, bio)->key;
+	bch_bkey_copy_single_ptr(bio_key, k, ptr);
+
+	bch_cut_front(&KEY(s->iop.inode, n->bi_iter.bi_sector, 0), bio_key);
+	bch_cut_back(&KEY(s->iop.inode, bio_end_sector(n), 0), bio_key);
+
+	n->bi_end_io	= bch_cache_read_endio;
+	n->bi_private	= &s->cl;
+
+	/*
+	 * The bucket we're reading from might be reused while our bio
+	 * is in flight, and we could then end up reading the wrong
+	 * data.
+	 *
+	 * We guard against this by checking (in cache_read_endio()) if
+	 * the pointer is stale again; if so, we treat it as an error
+	 * and reread from the backing device (but we don't pass that
+	 * error up anywhere).
+	 */
+
+	__bch_submit_bbio(n, b->c);
+	return n == bio ? MAP_DONE : MAP_CONTINUE;
+}
+
+static void cache_lookup(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, iop.cl);
+	struct bio *bio = &s->bio.bio;
+	int ret;
+
+	bch_btree_op_init(&s->op, -1);
+
+	ret = bch_btree_map_keys(&s->op, s->iop.c,
+				 &KEY(s->iop.inode, bio->bi_iter.bi_sector, 0),
+				 cache_lookup_fn, MAP_END_KEY);
+	if (ret == -EAGAIN)
+		continue_at(cl, cache_lookup, bcache_wq);
+
+	closure_return(cl);
+}
+
+/* Common code for the make_request functions */
+
+static void request_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+
+	if (error) {
+		struct search *s = container_of(cl, struct search, cl);
+		s->iop.error = error;
+		/* Only cache read errors are recoverable */
+		s->recoverable = false;
+	}
+
+	bio_put(bio);
+	closure_put(cl);
+}
+
+static void bio_complete(struct search *s)
+{
+	if (s->orig_bio) {
+		int cpu, rw = bio_data_dir(s->orig_bio);
+		unsigned long duration = jiffies - s->start_time;
+
+		cpu = part_stat_lock();
+		part_round_stats(cpu, &s->d->disk->part0);
+		part_stat_add(cpu, &s->d->disk->part0, ticks[rw], duration);
+		part_stat_unlock();
+
+		trace_bcache_request_end(s->d, s->orig_bio);
+		bio_endio(s->orig_bio, s->iop.error);
+		s->orig_bio = NULL;
+	}
+}
+
+static void do_bio_hook(struct search *s, struct bio *orig_bio)
+{
+	struct bio *bio = &s->bio.bio;
+
+	bio_init(bio);
+	__bio_clone_fast(bio, orig_bio);
+	bio->bi_end_io		= request_endio;
+	bio->bi_private		= &s->cl;
+
+	atomic_set(&bio->bi_cnt, 3);
+}
+
+static void search_free(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	bio_complete(s);
+
+	if (s->iop.bio)
+		bio_put(s->iop.bio);
+
+	closure_debug_destroy(cl);
+	mempool_free(s, s->d->c->search);
+}
+
+static inline struct search *search_alloc(struct bio *bio,
+					  struct bcache_device *d)
+{
+	struct search *s;
+
+	s = mempool_alloc(d->c->search, GFP_NOIO);
+
+	closure_init(&s->cl, NULL);
+	do_bio_hook(s, bio);
+
+	s->orig_bio		= bio;
+	s->cache_miss		= NULL;
+	s->d			= d;
+	s->recoverable		= 1;
+	s->write		= (bio->bi_rw & REQ_WRITE) != 0;
+	s->read_dirty_data	= 0;
+	s->start_time		= jiffies;
+
+	s->iop.c		= d->c;
+	s->iop.bio		= NULL;
+	s->iop.inode		= d->id;
+	s->iop.write_point	= hash_long((unsigned long) current, 16);
+	s->iop.write_prio	= 0;
+	s->iop.error		= 0;
+	s->iop.flags		= 0;
+	s->iop.flush_journal	= (bio->bi_rw & (REQ_FLUSH|REQ_FUA)) != 0;
+	s->iop.wq		= bcache_wq;
+
+	return s;
+}
+
+/* Cached devices */
+
+static void cached_dev_bio_complete(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
+
+	search_free(cl);
+	cached_dev_put(dc);
+}
+
+/* Process reads */
+
+static void cached_dev_cache_miss_done(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+
+	if (s->iop.replace_collision)
+		bch_mark_cache_miss_collision(s->iop.c, s->d);
+
+	if (s->iop.bio) {
+		int i;
+		struct bio_vec *bv;
+
+		bio_for_each_segment_all(bv, s->iop.bio, i)
+			__free_page(bv->bv_page);
+	}
+
+	cached_dev_bio_complete(cl);
+}
+
+static void cached_dev_read_error(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct bio *bio = &s->bio.bio;
+
+	if (s->recoverable) {
+		/* Retry from the backing device: */
+		trace_bcache_read_retry(s->orig_bio);
+
+		s->iop.error = 0;
+		do_bio_hook(s, s->orig_bio);
+
+		/* XXX: invalidate cache */
+
+		closure_bio_submit(bio, cl, s->d);
+	}
+
+	continue_at(cl, cached_dev_cache_miss_done, NULL);
+}
+
+static void cached_dev_read_done(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
+
+	/*
+	 * We had a cache miss; cache_bio now contains data ready to be inserted
+	 * into the cache.
+	 *
+	 * First, we copy the data we just read from cache_bio's bounce buffers
+	 * to the buffers the original bio pointed to:
+	 */
+
+	if (s->iop.bio) {
+		bio_reset(s->iop.bio);
+		s->iop.bio->bi_iter.bi_sector = s->cache_miss->bi_iter.bi_sector;
+		s->iop.bio->bi_bdev = s->cache_miss->bi_bdev;
+		s->iop.bio->bi_iter.bi_size = s->insert_bio_sectors << 9;
+		bch_bio_map(s->iop.bio, NULL);
+
+		bio_copy_data(s->cache_miss, s->iop.bio);
+
+		bio_put(s->cache_miss);
+		s->cache_miss = NULL;
+	}
+
+	if (verify(dc, &s->bio.bio) && s->recoverable && !s->read_dirty_data)
+		bch_data_verify(dc, s->orig_bio);
+
+	bio_complete(s);
+
+	if (s->iop.bio &&
+	    !test_bit(CACHE_SET_STOPPING, &s->iop.c->flags)) {
+		BUG_ON(!s->iop.replace);
+		closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
+	}
+
+	continue_at(cl, cached_dev_cache_miss_done, NULL);
+}
+
+static void cached_dev_read_done_bh(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
+
+	bch_mark_cache_accounting(s->iop.c, s->d,
+				  !s->cache_miss, s->iop.bypass);
+	trace_bcache_read(s->orig_bio, !s->cache_miss, s->iop.bypass);
+
+	if (s->iop.error)
+		continue_at_nobarrier(cl, cached_dev_read_error, bcache_wq);
+	else if (s->iop.bio || verify(dc, &s->bio.bio))
+		continue_at_nobarrier(cl, cached_dev_read_done, bcache_wq);
+	else
+		continue_at_nobarrier(cl, cached_dev_bio_complete, NULL);
+}
+
+static int cached_dev_cache_miss(struct btree *b, struct search *s,
+				 struct bio *bio, unsigned sectors)
+{
+	int ret = MAP_CONTINUE;
+	unsigned reada = 0;
+	struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
+	struct bio *miss, *cache_bio;
+
+	if (s->cache_miss || s->iop.bypass) {
+		miss = bio_next_split(bio, sectors, GFP_NOIO, s->d->bio_split);
+		ret = miss == bio ? MAP_DONE : MAP_CONTINUE;
+		goto out_submit;
+	}
+
+	if (!(bio->bi_rw & REQ_RAHEAD) &&
+	    !(bio->bi_rw & REQ_META) &&
+	    s->iop.c->gc_stats.in_use < CUTOFF_CACHE_READA)
+		reada = min_t(sector_t, dc->readahead >> 9,
+			      bdev_sectors(bio->bi_bdev) - bio_end_sector(bio));
+
+	s->insert_bio_sectors = min(sectors, bio_sectors(bio) + reada);
+
+	s->iop.replace_key = KEY(s->iop.inode,
+				 bio->bi_iter.bi_sector + s->insert_bio_sectors,
+				 s->insert_bio_sectors);
+
+	ret = bch_btree_insert_check_key(b, &s->op, &s->iop.replace_key);
+	if (ret)
+		return ret;
+
+	s->iop.replace = true;
+
+	miss = bio_next_split(bio, sectors, GFP_NOIO, s->d->bio_split);
+
+	/* btree_search_recurse()'s btree iterator is no good anymore */
+	ret = miss == bio ? MAP_DONE : -EINTR;
+
+	cache_bio = bio_alloc_bioset(GFP_NOWAIT,
+			DIV_ROUND_UP(s->insert_bio_sectors, PAGE_SECTORS),
+			dc->disk.bio_split);
+	if (!cache_bio)
+		goto out_submit;
+
+	cache_bio->bi_iter.bi_sector	= miss->bi_iter.bi_sector;
+	cache_bio->bi_bdev		= miss->bi_bdev;
+	cache_bio->bi_iter.bi_size	= s->insert_bio_sectors << 9;
+
+	cache_bio->bi_end_io	= request_endio;
+	cache_bio->bi_private	= &s->cl;
+
+	bch_bio_map(cache_bio, NULL);
+	if (bio_alloc_pages(cache_bio, __GFP_NOWARN|GFP_NOIO))
+		goto out_put;
+
+	if (reada)
+		bch_mark_cache_readahead(s->iop.c, s->d);
+
+	s->cache_miss	= miss;
+	s->iop.bio	= cache_bio;
+	bio_get(cache_bio);
+	closure_bio_submit(cache_bio, &s->cl, s->d);
+
+	return ret;
+out_put:
+	bio_put(cache_bio);
+out_submit:
+	miss->bi_end_io		= request_endio;
+	miss->bi_private	= &s->cl;
+	closure_bio_submit(miss, &s->cl, s->d);
+	return ret;
+}
+
+static void cached_dev_read(struct cached_dev *dc, struct search *s)
+{
+	struct closure *cl = &s->cl;
+
+	closure_call(&s->iop.cl, cache_lookup, NULL, cl);
+	continue_at(cl, cached_dev_read_done_bh, NULL);
+}
+
+/* Process writes */
+
+static void cached_dev_write_complete(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct cached_dev *dc = container_of(s->d, struct cached_dev, disk);
+
+	up_read_non_owner(&dc->writeback_lock);
+	cached_dev_bio_complete(cl);
+}
+
+static void cached_dev_write(struct cached_dev *dc, struct search *s)
+{
+	struct closure *cl = &s->cl;
+	struct bio *bio = &s->bio.bio;
+	struct bkey start = KEY(dc->disk.id, bio->bi_iter.bi_sector, 0);
+	struct bkey end = KEY(dc->disk.id, bio_end_sector(bio), 0);
+
+	bch_keybuf_check_overlapping(&s->iop.c->moving_gc_keys, &start, &end);
+
+	down_read_non_owner(&dc->writeback_lock);
+	if (bch_keybuf_check_overlapping(&dc->writeback_keys, &start, &end)) {
+		/*
+		 * We overlap with some dirty data undergoing background
+		 * writeback, force this write to writeback
+		 */
+		s->iop.bypass = false;
+		s->iop.writeback = true;
+	}
+
+	/*
+	 * Discards aren't _required_ to do anything, so skipping if
+	 * check_overlapping returned true is ok
+	 *
+	 * But check_overlapping drops dirty keys for which io hasn't started,
+	 * so we still want to call it.
+	 */
+	if (bio->bi_rw & REQ_DISCARD)
+		s->iop.bypass = true;
+
+	if (should_writeback(dc, s->orig_bio,
+			     cache_mode(dc, bio),
+			     s->iop.bypass)) {
+		s->iop.bypass = false;
+		s->iop.writeback = true;
+	}
+
+	if (s->iop.bypass) {
+		s->iop.bio = s->orig_bio;
+		bio_get(s->iop.bio);
+
+		if (!(bio->bi_rw & REQ_DISCARD) ||
+		    blk_queue_discard(bdev_get_queue(dc->bdev)))
+			closure_bio_submit(bio, cl, s->d);
+	} else if (s->iop.writeback) {
+		bch_writeback_add(dc);
+		s->iop.bio = bio;
+
+		if (bio->bi_rw & REQ_FLUSH) {
+			/* Also need to send a flush to the backing device */
+			struct bio *flush = bio_alloc_bioset(GFP_NOIO, 0,
+							     dc->disk.bio_split);
+
+			flush->bi_rw	= WRITE_FLUSH;
+			flush->bi_bdev	= bio->bi_bdev;
+			flush->bi_end_io = request_endio;
+			flush->bi_private = cl;
+
+			closure_bio_submit(flush, cl, s->d);
+		}
+	} else {
+		s->iop.bio = bio_clone_fast(bio, GFP_NOIO, dc->disk.bio_split);
+
+		closure_bio_submit(bio, cl, s->d);
+	}
+
+	closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
+	continue_at(cl, cached_dev_write_complete, NULL);
+}
+
+static void cached_dev_nodata(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+	struct bio *bio = &s->bio.bio;
+
+	if (s->iop.flush_journal)
+		bch_journal_meta(s->iop.c, cl);
+
+	/* If it's a flush, we send the flush to the backing device too */
+	closure_bio_submit(bio, cl, s->d);
+
+	continue_at(cl, cached_dev_bio_complete, NULL);
+}
+
+/* Cached devices - read & write stuff */
+
+static void cached_dev_make_request(struct request_queue *q, struct bio *bio)
+{
+	struct search *s;
+	struct bcache_device *d = bio->bi_bdev->bd_disk->private_data;
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	int cpu, rw = bio_data_dir(bio);
+
+	cpu = part_stat_lock();
+	part_stat_inc(cpu, &d->disk->part0, ios[rw]);
+	part_stat_add(cpu, &d->disk->part0, sectors[rw], bio_sectors(bio));
+	part_stat_unlock();
+
+	bio->bi_bdev = dc->bdev;
+	bio->bi_iter.bi_sector += dc->sb.data_offset;
+
+	if (cached_dev_get(dc)) {
+		s = search_alloc(bio, d);
+		trace_bcache_request_start(s->d, bio);
+
+		if (!bio->bi_iter.bi_size) {
+			/*
+			 * can't call bch_journal_meta from under
+			 * generic_make_request
+			 */
+			continue_at_nobarrier(&s->cl,
+					      cached_dev_nodata,
+					      bcache_wq);
+		} else {
+			s->iop.bypass = check_should_bypass(dc, bio);
+
+			if (rw)
+				cached_dev_write(dc, s);
+			else
+				cached_dev_read(dc, s);
+		}
+	} else {
+		if ((bio->bi_rw & REQ_DISCARD) &&
+		    !blk_queue_discard(bdev_get_queue(dc->bdev)))
+			bio_endio(bio, 0);
+		else
+			bch_generic_make_request(bio, &d->bio_split_hook);
+	}
+}
+
+static int cached_dev_ioctl(struct bcache_device *d, fmode_t mode,
+			    unsigned int cmd, unsigned long arg)
+{
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	return __blkdev_driver_ioctl(dc->bdev, mode, cmd, arg);
+}
+
+static int cached_dev_congested(void *data, int bits)
+{
+	struct bcache_device *d = data;
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	struct request_queue *q = bdev_get_queue(dc->bdev);
+	int ret = 0;
+
+	if (bdi_congested(&q->backing_dev_info, bits))
+		return 1;
+
+	if (cached_dev_get(dc)) {
+		unsigned i;
+		struct cache *ca;
+
+		for_each_cache(ca, d->c, i) {
+			q = bdev_get_queue(ca->bdev);
+			ret |= bdi_congested(&q->backing_dev_info, bits);
+		}
+
+		cached_dev_put(dc);
+	}
+
+	return ret;
+}
+
+void bch_cached_dev_request_init(struct cached_dev *dc)
+{
+	struct gendisk *g = dc->disk.disk;
+
+	g->queue->make_request_fn		= cached_dev_make_request;
+	g->queue->backing_dev_info.congested_fn = cached_dev_congested;
+	dc->disk.cache_miss			= cached_dev_cache_miss;
+	dc->disk.ioctl				= cached_dev_ioctl;
+}
+
+/* Flash backed devices */
+
+static int flash_dev_cache_miss(struct btree *b, struct search *s,
+				struct bio *bio, unsigned sectors)
+{
+	unsigned bytes = min(sectors, bio_sectors(bio)) << 9;
+
+	swap(bio->bi_iter.bi_size, bytes);
+	zero_fill_bio(bio);
+	swap(bio->bi_iter.bi_size, bytes);
+
+	bio_advance(bio, bytes);
+
+	if (!bio->bi_iter.bi_size)
+		return MAP_DONE;
+
+	return MAP_CONTINUE;
+}
+
+static void flash_dev_nodata(struct closure *cl)
+{
+	struct search *s = container_of(cl, struct search, cl);
+
+	if (s->iop.flush_journal)
+		bch_journal_meta(s->iop.c, cl);
+
+	continue_at(cl, search_free, NULL);
+}
+
+static void flash_dev_make_request(struct request_queue *q, struct bio *bio)
+{
+	struct search *s;
+	struct closure *cl;
+	struct bcache_device *d = bio->bi_bdev->bd_disk->private_data;
+	int cpu, rw = bio_data_dir(bio);
+
+	cpu = part_stat_lock();
+	part_stat_inc(cpu, &d->disk->part0, ios[rw]);
+	part_stat_add(cpu, &d->disk->part0, sectors[rw], bio_sectors(bio));
+	part_stat_unlock();
+
+	s = search_alloc(bio, d);
+	cl = &s->cl;
+	bio = &s->bio.bio;
+
+	trace_bcache_request_start(s->d, bio);
+
+	if (!bio->bi_iter.bi_size) {
+		/*
+		 * can't call bch_journal_meta from under
+		 * generic_make_request
+		 */
+		continue_at_nobarrier(&s->cl,
+				      flash_dev_nodata,
+				      bcache_wq);
+	} else if (rw) {
+		bch_keybuf_check_overlapping(&s->iop.c->moving_gc_keys,
+					&KEY(d->id, bio->bi_iter.bi_sector, 0),
+					&KEY(d->id, bio_end_sector(bio), 0));
+
+		s->iop.bypass		= (bio->bi_rw & REQ_DISCARD) != 0;
+		s->iop.writeback	= true;
+		s->iop.bio		= bio;
+
+		closure_call(&s->iop.cl, bch_data_insert, NULL, cl);
+	} else {
+		closure_call(&s->iop.cl, cache_lookup, NULL, cl);
+	}
+
+	continue_at(cl, search_free, NULL);
+}
+
+static int flash_dev_ioctl(struct bcache_device *d, fmode_t mode,
+			   unsigned int cmd, unsigned long arg)
+{
+	return -ENOTTY;
+}
+
+static int flash_dev_congested(void *data, int bits)
+{
+	struct bcache_device *d = data;
+	struct request_queue *q;
+	struct cache *ca;
+	unsigned i;
+	int ret = 0;
+
+	for_each_cache(ca, d->c, i) {
+		q = bdev_get_queue(ca->bdev);
+		ret |= bdi_congested(&q->backing_dev_info, bits);
+	}
+
+	return ret;
+}
+
+void bch_flash_dev_request_init(struct bcache_device *d)
+{
+	struct gendisk *g = d->disk;
+
+	g->queue->make_request_fn		= flash_dev_make_request;
+	g->queue->backing_dev_info.congested_fn = flash_dev_congested;
+	d->cache_miss				= flash_dev_cache_miss;
+	d->ioctl				= flash_dev_ioctl;
+}
+
+void bch_request_exit(void)
+{
+	if (bch_search_cache)
+		kmem_cache_destroy(bch_search_cache);
+}
+
+int __init bch_request_init(void)
+{
+	bch_search_cache = KMEM_CACHE(search, 0);
+	if (!bch_search_cache)
+		return -ENOMEM;
+
+	return 0;
+}
diff --git a/drivers/md/bcache/request.h b/drivers/md/bcache/request.h
new file mode 100644
index 00000000000..1ff36875c2b
--- /dev/null
+++ b/drivers/md/bcache/request.h
@@ -0,0 +1,43 @@
+#ifndef _BCACHE_REQUEST_H_
+#define _BCACHE_REQUEST_H_
+
+struct data_insert_op {
+	struct closure		cl;
+	struct cache_set	*c;
+	struct bio		*bio;
+	struct workqueue_struct *wq;
+
+	unsigned		inode;
+	uint16_t		write_point;
+	uint16_t		write_prio;
+	short			error;
+
+	union {
+		uint16_t	flags;
+
+	struct {
+		unsigned	bypass:1;
+		unsigned	writeback:1;
+		unsigned	flush_journal:1;
+		unsigned	csum:1;
+
+		unsigned	replace:1;
+		unsigned	replace_collision:1;
+
+		unsigned	insert_data_done:1;
+	};
+	};
+
+	struct keylist		insert_keys;
+	BKEY_PADDED(replace_key);
+};
+
+unsigned bch_get_congested(struct cache_set *);
+void bch_data_insert(struct closure *cl);
+
+void bch_cached_dev_request_init(struct cached_dev *dc);
+void bch_flash_dev_request_init(struct bcache_device *d);
+
+extern struct kmem_cache *bch_search_cache, *bch_passthrough_cache;
+
+#endif /* _BCACHE_REQUEST_H_ */
diff --git a/drivers/md/bcache/stats.c b/drivers/md/bcache/stats.c
new file mode 100644
index 00000000000..0ca072c20d0
--- /dev/null
+++ b/drivers/md/bcache/stats.c
@@ -0,0 +1,241 @@
+/*
+ * bcache stats code
+ *
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "stats.h"
+#include "btree.h"
+#include "sysfs.h"
+
+/*
+ * We keep absolute totals of various statistics, and addionally a set of three
+ * rolling averages.
+ *
+ * Every so often, a timer goes off and rescales the rolling averages.
+ * accounting_rescale[] is how many times the timer has to go off before we
+ * rescale each set of numbers; that gets us half lives of 5 minutes, one hour,
+ * and one day.
+ *
+ * accounting_delay is how often the timer goes off - 22 times in 5 minutes,
+ * and accounting_weight is what we use to rescale:
+ *
+ * pow(31 / 32, 22) ~= 1/2
+ *
+ * So that we don't have to increment each set of numbers every time we (say)
+ * get a cache hit, we increment a single atomic_t in acc->collector, and when
+ * the rescale function runs it resets the atomic counter to 0 and adds its
+ * old value to each of the exported numbers.
+ *
+ * To reduce rounding error, the numbers in struct cache_stats are all
+ * stored left shifted by 16, and scaled back in the sysfs show() function.
+ */
+
+static const unsigned DAY_RESCALE		= 288;
+static const unsigned HOUR_RESCALE		= 12;
+static const unsigned FIVE_MINUTE_RESCALE	= 1;
+static const unsigned accounting_delay		= (HZ * 300) / 22;
+static const unsigned accounting_weight		= 32;
+
+/* sysfs reading/writing */
+
+read_attribute(cache_hits);
+read_attribute(cache_misses);
+read_attribute(cache_bypass_hits);
+read_attribute(cache_bypass_misses);
+read_attribute(cache_hit_ratio);
+read_attribute(cache_readaheads);
+read_attribute(cache_miss_collisions);
+read_attribute(bypassed);
+
+SHOW(bch_stats)
+{
+	struct cache_stats *s =
+		container_of(kobj, struct cache_stats, kobj);
+#define var(stat)		(s->stat >> 16)
+	var_print(cache_hits);
+	var_print(cache_misses);
+	var_print(cache_bypass_hits);
+	var_print(cache_bypass_misses);
+
+	sysfs_print(cache_hit_ratio,
+		    DIV_SAFE(var(cache_hits) * 100,
+			     var(cache_hits) + var(cache_misses)));
+
+	var_print(cache_readaheads);
+	var_print(cache_miss_collisions);
+	sysfs_hprint(bypassed,	var(sectors_bypassed) << 9);
+#undef var
+	return 0;
+}
+
+STORE(bch_stats)
+{
+	return size;
+}
+
+static void bch_stats_release(struct kobject *k)
+{
+}
+
+static struct attribute *bch_stats_files[] = {
+	&sysfs_cache_hits,
+	&sysfs_cache_misses,
+	&sysfs_cache_bypass_hits,
+	&sysfs_cache_bypass_misses,
+	&sysfs_cache_hit_ratio,
+	&sysfs_cache_readaheads,
+	&sysfs_cache_miss_collisions,
+	&sysfs_bypassed,
+	NULL
+};
+static KTYPE(bch_stats);
+
+int bch_cache_accounting_add_kobjs(struct cache_accounting *acc,
+				   struct kobject *parent)
+{
+	int ret = kobject_add(&acc->total.kobj, parent,
+			      "stats_total");
+	ret = ret ?: kobject_add(&acc->five_minute.kobj, parent,
+				 "stats_five_minute");
+	ret = ret ?: kobject_add(&acc->hour.kobj, parent,
+				 "stats_hour");
+	ret = ret ?: kobject_add(&acc->day.kobj, parent,
+				 "stats_day");
+	return ret;
+}
+
+void bch_cache_accounting_clear(struct cache_accounting *acc)
+{
+	memset(&acc->total.cache_hits,
+	       0,
+	       sizeof(unsigned long) * 7);
+}
+
+void bch_cache_accounting_destroy(struct cache_accounting *acc)
+{
+	kobject_put(&acc->total.kobj);
+	kobject_put(&acc->five_minute.kobj);
+	kobject_put(&acc->hour.kobj);
+	kobject_put(&acc->day.kobj);
+
+	atomic_set(&acc->closing, 1);
+	if (del_timer_sync(&acc->timer))
+		closure_return(&acc->cl);
+}
+
+/* EWMA scaling */
+
+static void scale_stat(unsigned long *stat)
+{
+	*stat =  ewma_add(*stat, 0, accounting_weight, 0);
+}
+
+static void scale_stats(struct cache_stats *stats, unsigned long rescale_at)
+{
+	if (++stats->rescale == rescale_at) {
+		stats->rescale = 0;
+		scale_stat(&stats->cache_hits);
+		scale_stat(&stats->cache_misses);
+		scale_stat(&stats->cache_bypass_hits);
+		scale_stat(&stats->cache_bypass_misses);
+		scale_stat(&stats->cache_readaheads);
+		scale_stat(&stats->cache_miss_collisions);
+		scale_stat(&stats->sectors_bypassed);
+	}
+}
+
+static void scale_accounting(unsigned long data)
+{
+	struct cache_accounting *acc = (struct cache_accounting *) data;
+
+#define move_stat(name) do {						\
+	unsigned t = atomic_xchg(&acc->collector.name, 0);		\
+	t <<= 16;							\
+	acc->five_minute.name += t;					\
+	acc->hour.name += t;						\
+	acc->day.name += t;						\
+	acc->total.name += t;						\
+} while (0)
+
+	move_stat(cache_hits);
+	move_stat(cache_misses);
+	move_stat(cache_bypass_hits);
+	move_stat(cache_bypass_misses);
+	move_stat(cache_readaheads);
+	move_stat(cache_miss_collisions);
+	move_stat(sectors_bypassed);
+
+	scale_stats(&acc->total, 0);
+	scale_stats(&acc->day, DAY_RESCALE);
+	scale_stats(&acc->hour, HOUR_RESCALE);
+	scale_stats(&acc->five_minute, FIVE_MINUTE_RESCALE);
+
+	acc->timer.expires += accounting_delay;
+
+	if (!atomic_read(&acc->closing))
+		add_timer(&acc->timer);
+	else
+		closure_return(&acc->cl);
+}
+
+static void mark_cache_stats(struct cache_stat_collector *stats,
+			     bool hit, bool bypass)
+{
+	if (!bypass)
+		if (hit)
+			atomic_inc(&stats->cache_hits);
+		else
+			atomic_inc(&stats->cache_misses);
+	else
+		if (hit)
+			atomic_inc(&stats->cache_bypass_hits);
+		else
+			atomic_inc(&stats->cache_bypass_misses);
+}
+
+void bch_mark_cache_accounting(struct cache_set *c, struct bcache_device *d,
+			       bool hit, bool bypass)
+{
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	mark_cache_stats(&dc->accounting.collector, hit, bypass);
+	mark_cache_stats(&c->accounting.collector, hit, bypass);
+}
+
+void bch_mark_cache_readahead(struct cache_set *c, struct bcache_device *d)
+{
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	atomic_inc(&dc->accounting.collector.cache_readaheads);
+	atomic_inc(&c->accounting.collector.cache_readaheads);
+}
+
+void bch_mark_cache_miss_collision(struct cache_set *c, struct bcache_device *d)
+{
+	struct cached_dev *dc = container_of(d, struct cached_dev, disk);
+	atomic_inc(&dc->accounting.collector.cache_miss_collisions);
+	atomic_inc(&c->accounting.collector.cache_miss_collisions);
+}
+
+void bch_mark_sectors_bypassed(struct cache_set *c, struct cached_dev *dc,
+			       int sectors)
+{
+	atomic_add(sectors, &dc->accounting.collector.sectors_bypassed);
+	atomic_add(sectors, &c->accounting.collector.sectors_bypassed);
+}
+
+void bch_cache_accounting_init(struct cache_accounting *acc,
+			       struct closure *parent)
+{
+	kobject_init(&acc->total.kobj,		&bch_stats_ktype);
+	kobject_init(&acc->five_minute.kobj,	&bch_stats_ktype);
+	kobject_init(&acc->hour.kobj,		&bch_stats_ktype);
+	kobject_init(&acc->day.kobj,		&bch_stats_ktype);
+
+	closure_init(&acc->cl, parent);
+	init_timer(&acc->timer);
+	acc->timer.expires	= jiffies + accounting_delay;
+	acc->timer.data		= (unsigned long) acc;
+	acc->timer.function	= scale_accounting;
+	add_timer(&acc->timer);
+}
diff --git a/drivers/md/bcache/stats.h b/drivers/md/bcache/stats.h
new file mode 100644
index 00000000000..adbff141c88
--- /dev/null
+++ b/drivers/md/bcache/stats.h
@@ -0,0 +1,61 @@
+#ifndef _BCACHE_STATS_H_
+#define _BCACHE_STATS_H_
+
+struct cache_stat_collector {
+	atomic_t cache_hits;
+	atomic_t cache_misses;
+	atomic_t cache_bypass_hits;
+	atomic_t cache_bypass_misses;
+	atomic_t cache_readaheads;
+	atomic_t cache_miss_collisions;
+	atomic_t sectors_bypassed;
+};
+
+struct cache_stats {
+	struct kobject		kobj;
+
+	unsigned long cache_hits;
+	unsigned long cache_misses;
+	unsigned long cache_bypass_hits;
+	unsigned long cache_bypass_misses;
+	unsigned long cache_readaheads;
+	unsigned long cache_miss_collisions;
+	unsigned long sectors_bypassed;
+
+	unsigned		rescale;
+};
+
+struct cache_accounting {
+	struct closure		cl;
+	struct timer_list	timer;
+	atomic_t		closing;
+
+	struct cache_stat_collector collector;
+
+	struct cache_stats total;
+	struct cache_stats five_minute;
+	struct cache_stats hour;
+	struct cache_stats day;
+};
+
+struct cache_set;
+struct cached_dev;
+struct bcache_device;
+
+void bch_cache_accounting_init(struct cache_accounting *acc,
+			       struct closure *parent);
+
+int bch_cache_accounting_add_kobjs(struct cache_accounting *acc,
+				   struct kobject *parent);
+
+void bch_cache_accounting_clear(struct cache_accounting *acc);
+
+void bch_cache_accounting_destroy(struct cache_accounting *acc);
+
+void bch_mark_cache_accounting(struct cache_set *, struct bcache_device *,
+			       bool, bool);
+void bch_mark_cache_readahead(struct cache_set *, struct bcache_device *);
+void bch_mark_cache_miss_collision(struct cache_set *, struct bcache_device *);
+void bch_mark_sectors_bypassed(struct cache_set *, struct cached_dev *, int);
+
+#endif /* _BCACHE_STATS_H_ */
diff --git a/drivers/md/bcache/super.c b/drivers/md/bcache/super.c
new file mode 100644
index 00000000000..926ded8ccbf
--- /dev/null
+++ b/drivers/md/bcache/super.c
@@ -0,0 +1,2102 @@
+/*
+ * bcache setup/teardown code, and some metadata io - read a superblock and
+ * figure out what to do with it.
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "extents.h"
+#include "request.h"
+#include "writeback.h"
+
+#include <linux/blkdev.h>
+#include <linux/buffer_head.h>
+#include <linux/debugfs.h>
+#include <linux/genhd.h>
+#include <linux/idr.h>
+#include <linux/kthread.h>
+#include <linux/module.h>
+#include <linux/random.h>
+#include <linux/reboot.h>
+#include <linux/sysfs.h>
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Kent Overstreet <kent.overstreet@gmail.com>");
+
+static const char bcache_magic[] = {
+	0xc6, 0x85, 0x73, 0xf6, 0x4e, 0x1a, 0x45, 0xca,
+	0x82, 0x65, 0xf5, 0x7f, 0x48, 0xba, 0x6d, 0x81
+};
+
+static const char invalid_uuid[] = {
+	0xa0, 0x3e, 0xf8, 0xed, 0x3e, 0xe1, 0xb8, 0x78,
+	0xc8, 0x50, 0xfc, 0x5e, 0xcb, 0x16, 0xcd, 0x99
+};
+
+/* Default is -1; we skip past it for struct cached_dev's cache mode */
+const char * const bch_cache_modes[] = {
+	"default",
+	"writethrough",
+	"writeback",
+	"writearound",
+	"none",
+	NULL
+};
+
+static struct kobject *bcache_kobj;
+struct mutex bch_register_lock;
+LIST_HEAD(bch_cache_sets);
+static LIST_HEAD(uncached_devices);
+
+static int bcache_major;
+static DEFINE_IDA(bcache_minor);
+static wait_queue_head_t unregister_wait;
+struct workqueue_struct *bcache_wq;
+
+#define BTREE_MAX_PAGES		(256 * 1024 / PAGE_SIZE)
+
+static void bio_split_pool_free(struct bio_split_pool *p)
+{
+	if (p->bio_split_hook)
+		mempool_destroy(p->bio_split_hook);
+
+	if (p->bio_split)
+		bioset_free(p->bio_split);
+}
+
+static int bio_split_pool_init(struct bio_split_pool *p)
+{
+	p->bio_split = bioset_create(4, 0);
+	if (!p->bio_split)
+		return -ENOMEM;
+
+	p->bio_split_hook = mempool_create_kmalloc_pool(4,
+				sizeof(struct bio_split_hook));
+	if (!p->bio_split_hook)
+		return -ENOMEM;
+
+	return 0;
+}
+
+/* Superblock */
+
+static const char *read_super(struct cache_sb *sb, struct block_device *bdev,
+			      struct page **res)
+{
+	const char *err;
+	struct cache_sb *s;
+	struct buffer_head *bh = __bread(bdev, 1, SB_SIZE);
+	unsigned i;
+
+	if (!bh)
+		return "IO error";
+
+	s = (struct cache_sb *) bh->b_data;
+
+	sb->offset		= le64_to_cpu(s->offset);
+	sb->version		= le64_to_cpu(s->version);
+
+	memcpy(sb->magic,	s->magic, 16);
+	memcpy(sb->uuid,	s->uuid, 16);
+	memcpy(sb->set_uuid,	s->set_uuid, 16);
+	memcpy(sb->label,	s->label, SB_LABEL_SIZE);
+
+	sb->flags		= le64_to_cpu(s->flags);
+	sb->seq			= le64_to_cpu(s->seq);
+	sb->last_mount		= le32_to_cpu(s->last_mount);
+	sb->first_bucket	= le16_to_cpu(s->first_bucket);
+	sb->keys		= le16_to_cpu(s->keys);
+
+	for (i = 0; i < SB_JOURNAL_BUCKETS; i++)
+		sb->d[i] = le64_to_cpu(s->d[i]);
+
+	pr_debug("read sb version %llu, flags %llu, seq %llu, journal size %u",
+		 sb->version, sb->flags, sb->seq, sb->keys);
+
+	err = "Not a bcache superblock";
+	if (sb->offset != SB_SECTOR)
+		goto err;
+
+	if (memcmp(sb->magic, bcache_magic, 16))
+		goto err;
+
+	err = "Too many journal buckets";
+	if (sb->keys > SB_JOURNAL_BUCKETS)
+		goto err;
+
+	err = "Bad checksum";
+	if (s->csum != csum_set(s))
+		goto err;
+
+	err = "Bad UUID";
+	if (bch_is_zero(sb->uuid, 16))
+		goto err;
+
+	sb->block_size	= le16_to_cpu(s->block_size);
+
+	err = "Superblock block size smaller than device block size";
+	if (sb->block_size << 9 < bdev_logical_block_size(bdev))
+		goto err;
+
+	switch (sb->version) {
+	case BCACHE_SB_VERSION_BDEV:
+		sb->data_offset	= BDEV_DATA_START_DEFAULT;
+		break;
+	case BCACHE_SB_VERSION_BDEV_WITH_OFFSET:
+		sb->data_offset	= le64_to_cpu(s->data_offset);
+
+		err = "Bad data offset";
+		if (sb->data_offset < BDEV_DATA_START_DEFAULT)
+			goto err;
+
+		break;
+	case BCACHE_SB_VERSION_CDEV:
+	case BCACHE_SB_VERSION_CDEV_WITH_UUID:
+		sb->nbuckets	= le64_to_cpu(s->nbuckets);
+		sb->block_size	= le16_to_cpu(s->block_size);
+		sb->bucket_size	= le16_to_cpu(s->bucket_size);
+
+		sb->nr_in_set	= le16_to_cpu(s->nr_in_set);
+		sb->nr_this_dev	= le16_to_cpu(s->nr_this_dev);
+
+		err = "Too many buckets";
+		if (sb->nbuckets > LONG_MAX)
+			goto err;
+
+		err = "Not enough buckets";
+		if (sb->nbuckets < 1 << 7)
+			goto err;
+
+		err = "Bad block/bucket size";
+		if (!is_power_of_2(sb->block_size) ||
+		    sb->block_size > PAGE_SECTORS ||
+		    !is_power_of_2(sb->bucket_size) ||
+		    sb->bucket_size < PAGE_SECTORS)
+			goto err;
+
+		err = "Invalid superblock: device too small";
+		if (get_capacity(bdev->bd_disk) < sb->bucket_size * sb->nbuckets)
+			goto err;
+
+		err = "Bad UUID";
+		if (bch_is_zero(sb->set_uuid, 16))
+			goto err;
+
+		err = "Bad cache device number in set";
+		if (!sb->nr_in_set ||
+		    sb->nr_in_set <= sb->nr_this_dev ||
+		    sb->nr_in_set > MAX_CACHES_PER_SET)
+			goto err;
+
+		err = "Journal buckets not sequential";
+		for (i = 0; i < sb->keys; i++)
+			if (sb->d[i] != sb->first_bucket + i)
+				goto err;
+
+		err = "Too many journal buckets";
+		if (sb->first_bucket + sb->keys > sb->nbuckets)
+			goto err;
+
+		err = "Invalid superblock: first bucket comes before end of super";
+		if (sb->first_bucket * sb->bucket_size < 16)
+			goto err;
+
+		break;
+	default:
+		err = "Unsupported superblock version";
+		goto err;
+	}
+
+	sb->last_mount = get_seconds();
+	err = NULL;
+
+	get_page(bh->b_page);
+	*res = bh->b_page;
+err:
+	put_bh(bh);
+	return err;
+}
+
+static void write_bdev_super_endio(struct bio *bio, int error)
+{
+	struct cached_dev *dc = bio->bi_private;
+	/* XXX: error checking */
+
+	closure_put(&dc->sb_write);
+}
+
+static void __write_super(struct cache_sb *sb, struct bio *bio)
+{
+	struct cache_sb *out = page_address(bio->bi_io_vec[0].bv_page);
+	unsigned i;
+
+	bio->bi_iter.bi_sector	= SB_SECTOR;
+	bio->bi_rw		= REQ_SYNC|REQ_META;
+	bio->bi_iter.bi_size	= SB_SIZE;
+	bch_bio_map(bio, NULL);
+
+	out->offset		= cpu_to_le64(sb->offset);
+	out->version		= cpu_to_le64(sb->version);
+
+	memcpy(out->uuid,	sb->uuid, 16);
+	memcpy(out->set_uuid,	sb->set_uuid, 16);
+	memcpy(out->label,	sb->label, SB_LABEL_SIZE);
+
+	out->flags		= cpu_to_le64(sb->flags);
+	out->seq		= cpu_to_le64(sb->seq);
+
+	out->last_mount		= cpu_to_le32(sb->last_mount);
+	out->first_bucket	= cpu_to_le16(sb->first_bucket);
+	out->keys		= cpu_to_le16(sb->keys);
+
+	for (i = 0; i < sb->keys; i++)
+		out->d[i] = cpu_to_le64(sb->d[i]);
+
+	out->csum = csum_set(out);
+
+	pr_debug("ver %llu, flags %llu, seq %llu",
+		 sb->version, sb->flags, sb->seq);
+
+	submit_bio(REQ_WRITE, bio);
+}
+
+static void bch_write_bdev_super_unlock(struct closure *cl)
+{
+	struct cached_dev *dc = container_of(cl, struct cached_dev, sb_write);
+
+	up(&dc->sb_write_mutex);
+}
+
+void bch_write_bdev_super(struct cached_dev *dc, struct closure *parent)
+{
+	struct closure *cl = &dc->sb_write;
+	struct bio *bio = &dc->sb_bio;
+
+	down(&dc->sb_write_mutex);
+	closure_init(cl, parent);
+
+	bio_reset(bio);
+	bio->bi_bdev	= dc->bdev;
+	bio->bi_end_io	= write_bdev_super_endio;
+	bio->bi_private = dc;
+
+	closure_get(cl);
+	__write_super(&dc->sb, bio);
+
+	closure_return_with_destructor(cl, bch_write_bdev_super_unlock);
+}
+
+static void write_super_endio(struct bio *bio, int error)
+{
+	struct cache *ca = bio->bi_private;
+
+	bch_count_io_errors(ca, error, "writing superblock");
+	closure_put(&ca->set->sb_write);
+}
+
+static void bcache_write_super_unlock(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, sb_write);
+
+	up(&c->sb_write_mutex);
+}
+
+void bcache_write_super(struct cache_set *c)
+{
+	struct closure *cl = &c->sb_write;
+	struct cache *ca;
+	unsigned i;
+
+	down(&c->sb_write_mutex);
+	closure_init(cl, &c->cl);
+
+	c->sb.seq++;
+
+	for_each_cache(ca, c, i) {
+		struct bio *bio = &ca->sb_bio;
+
+		ca->sb.version		= BCACHE_SB_VERSION_CDEV_WITH_UUID;
+		ca->sb.seq		= c->sb.seq;
+		ca->sb.last_mount	= c->sb.last_mount;
+
+		SET_CACHE_SYNC(&ca->sb, CACHE_SYNC(&c->sb));
+
+		bio_reset(bio);
+		bio->bi_bdev	= ca->bdev;
+		bio->bi_end_io	= write_super_endio;
+		bio->bi_private = ca;
+
+		closure_get(cl);
+		__write_super(&ca->sb, bio);
+	}
+
+	closure_return_with_destructor(cl, bcache_write_super_unlock);
+}
+
+/* UUID io */
+
+static void uuid_endio(struct bio *bio, int error)
+{
+	struct closure *cl = bio->bi_private;
+	struct cache_set *c = container_of(cl, struct cache_set, uuid_write);
+
+	cache_set_err_on(error, c, "accessing uuids");
+	bch_bbio_free(bio, c);
+	closure_put(cl);
+}
+
+static void uuid_io_unlock(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, uuid_write);
+
+	up(&c->uuid_write_mutex);
+}
+
+static void uuid_io(struct cache_set *c, unsigned long rw,
+		    struct bkey *k, struct closure *parent)
+{
+	struct closure *cl = &c->uuid_write;
+	struct uuid_entry *u;
+	unsigned i;
+	char buf[80];
+
+	BUG_ON(!parent);
+	down(&c->uuid_write_mutex);
+	closure_init(cl, parent);
+
+	for (i = 0; i < KEY_PTRS(k); i++) {
+		struct bio *bio = bch_bbio_alloc(c);
+
+		bio->bi_rw	= REQ_SYNC|REQ_META|rw;
+		bio->bi_iter.bi_size = KEY_SIZE(k) << 9;
+
+		bio->bi_end_io	= uuid_endio;
+		bio->bi_private = cl;
+		bch_bio_map(bio, c->uuids);
+
+		bch_submit_bbio(bio, c, k, i);
+
+		if (!(rw & WRITE))
+			break;
+	}
+
+	bch_extent_to_text(buf, sizeof(buf), k);
+	pr_debug("%s UUIDs at %s", rw & REQ_WRITE ? "wrote" : "read", buf);
+
+	for (u = c->uuids; u < c->uuids + c->nr_uuids; u++)
+		if (!bch_is_zero(u->uuid, 16))
+			pr_debug("Slot %zi: %pU: %s: 1st: %u last: %u inv: %u",
+				 u - c->uuids, u->uuid, u->label,
+				 u->first_reg, u->last_reg, u->invalidated);
+
+	closure_return_with_destructor(cl, uuid_io_unlock);
+}
+
+static char *uuid_read(struct cache_set *c, struct jset *j, struct closure *cl)
+{
+	struct bkey *k = &j->uuid_bucket;
+
+	if (__bch_btree_ptr_invalid(c, k))
+		return "bad uuid pointer";
+
+	bkey_copy(&c->uuid_bucket, k);
+	uuid_io(c, READ_SYNC, k, cl);
+
+	if (j->version < BCACHE_JSET_VERSION_UUIDv1) {
+		struct uuid_entry_v0	*u0 = (void *) c->uuids;
+		struct uuid_entry	*u1 = (void *) c->uuids;
+		int i;
+
+		closure_sync(cl);
+
+		/*
+		 * Since the new uuid entry is bigger than the old, we have to
+		 * convert starting at the highest memory address and work down
+		 * in order to do it in place
+		 */
+
+		for (i = c->nr_uuids - 1;
+		     i >= 0;
+		     --i) {
+			memcpy(u1[i].uuid,	u0[i].uuid, 16);
+			memcpy(u1[i].label,	u0[i].label, 32);
+
+			u1[i].first_reg		= u0[i].first_reg;
+			u1[i].last_reg		= u0[i].last_reg;
+			u1[i].invalidated	= u0[i].invalidated;
+
+			u1[i].flags	= 0;
+			u1[i].sectors	= 0;
+		}
+	}
+
+	return NULL;
+}
+
+static int __uuid_write(struct cache_set *c)
+{
+	BKEY_PADDED(key) k;
+	struct closure cl;
+	closure_init_stack(&cl);
+
+	lockdep_assert_held(&bch_register_lock);
+
+	if (bch_bucket_alloc_set(c, RESERVE_BTREE, &k.key, 1, true))
+		return 1;
+
+	SET_KEY_SIZE(&k.key, c->sb.bucket_size);
+	uuid_io(c, REQ_WRITE, &k.key, &cl);
+	closure_sync(&cl);
+
+	bkey_copy(&c->uuid_bucket, &k.key);
+	bkey_put(c, &k.key);
+	return 0;
+}
+
+int bch_uuid_write(struct cache_set *c)
+{
+	int ret = __uuid_write(c);
+
+	if (!ret)
+		bch_journal_meta(c, NULL);
+
+	return ret;
+}
+
+static struct uuid_entry *uuid_find(struct cache_set *c, const char *uuid)
+{
+	struct uuid_entry *u;
+
+	for (u = c->uuids;
+	     u < c->uuids + c->nr_uuids; u++)
+		if (!memcmp(u->uuid, uuid, 16))
+			return u;
+
+	return NULL;
+}
+
+static struct uuid_entry *uuid_find_empty(struct cache_set *c)
+{
+	static const char zero_uuid[16] = "\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0\0";
+	return uuid_find(c, zero_uuid);
+}
+
+/*
+ * Bucket priorities/gens:
+ *
+ * For each bucket, we store on disk its
+   * 8 bit gen
+   * 16 bit priority
+ *
+ * See alloc.c for an explanation of the gen. The priority is used to implement
+ * lru (and in the future other) cache replacement policies; for most purposes
+ * it's just an opaque integer.
+ *
+ * The gens and the priorities don't have a whole lot to do with each other, and
+ * it's actually the gens that must be written out at specific times - it's no
+ * big deal if the priorities don't get written, if we lose them we just reuse
+ * buckets in suboptimal order.
+ *
+ * On disk they're stored in a packed array, and in as many buckets are required
+ * to fit them all. The buckets we use to store them form a list; the journal
+ * header points to the first bucket, the first bucket points to the second
+ * bucket, et cetera.
+ *
+ * This code is used by the allocation code; periodically (whenever it runs out
+ * of buckets to allocate from) the allocation code will invalidate some
+ * buckets, but it can't use those buckets until their new gens are safely on
+ * disk.
+ */
+
+static void prio_endio(struct bio *bio, int error)
+{
+	struct cache *ca = bio->bi_private;
+
+	cache_set_err_on(error, ca->set, "accessing priorities");
+	bch_bbio_free(bio, ca->set);
+	closure_put(&ca->prio);
+}
+
+static void prio_io(struct cache *ca, uint64_t bucket, unsigned long rw)
+{
+	struct closure *cl = &ca->prio;
+	struct bio *bio = bch_bbio_alloc(ca->set);
+
+	closure_init_stack(cl);
+
+	bio->bi_iter.bi_sector	= bucket * ca->sb.bucket_size;
+	bio->bi_bdev		= ca->bdev;
+	bio->bi_rw		= REQ_SYNC|REQ_META|rw;
+	bio->bi_iter.bi_size	= bucket_bytes(ca);
+
+	bio->bi_end_io	= prio_endio;
+	bio->bi_private = ca;
+	bch_bio_map(bio, ca->disk_buckets);
+
+	closure_bio_submit(bio, &ca->prio, ca);
+	closure_sync(cl);
+}
+
+void bch_prio_write(struct cache *ca)
+{
+	int i;
+	struct bucket *b;
+	struct closure cl;
+
+	closure_init_stack(&cl);
+
+	lockdep_assert_held(&ca->set->bucket_lock);
+
+	ca->disk_buckets->seq++;
+
+	atomic_long_add(ca->sb.bucket_size * prio_buckets(ca),
+			&ca->meta_sectors_written);
+
+	//pr_debug("free %zu, free_inc %zu, unused %zu", fifo_used(&ca->free),
+	//	 fifo_used(&ca->free_inc), fifo_used(&ca->unused));
+
+	for (i = prio_buckets(ca) - 1; i >= 0; --i) {
+		long bucket;
+		struct prio_set *p = ca->disk_buckets;
+		struct bucket_disk *d = p->data;
+		struct bucket_disk *end = d + prios_per_bucket(ca);
+
+		for (b = ca->buckets + i * prios_per_bucket(ca);
+		     b < ca->buckets + ca->sb.nbuckets && d < end;
+		     b++, d++) {
+			d->prio = cpu_to_le16(b->prio);
+			d->gen = b->gen;
+		}
+
+		p->next_bucket	= ca->prio_buckets[i + 1];
+		p->magic	= pset_magic(&ca->sb);
+		p->csum		= bch_crc64(&p->magic, bucket_bytes(ca) - 8);
+
+		bucket = bch_bucket_alloc(ca, RESERVE_PRIO, true);
+		BUG_ON(bucket == -1);
+
+		mutex_unlock(&ca->set->bucket_lock);
+		prio_io(ca, bucket, REQ_WRITE);
+		mutex_lock(&ca->set->bucket_lock);
+
+		ca->prio_buckets[i] = bucket;
+		atomic_dec_bug(&ca->buckets[bucket].pin);
+	}
+
+	mutex_unlock(&ca->set->bucket_lock);
+
+	bch_journal_meta(ca->set, &cl);
+	closure_sync(&cl);
+
+	mutex_lock(&ca->set->bucket_lock);
+
+	/*
+	 * Don't want the old priorities to get garbage collected until after we
+	 * finish writing the new ones, and they're journalled
+	 */
+	for (i = 0; i < prio_buckets(ca); i++) {
+		if (ca->prio_last_buckets[i])
+			__bch_bucket_free(ca,
+				&ca->buckets[ca->prio_last_buckets[i]]);
+
+		ca->prio_last_buckets[i] = ca->prio_buckets[i];
+	}
+}
+
+static void prio_read(struct cache *ca, uint64_t bucket)
+{
+	struct prio_set *p = ca->disk_buckets;
+	struct bucket_disk *d = p->data + prios_per_bucket(ca), *end = d;
+	struct bucket *b;
+	unsigned bucket_nr = 0;
+
+	for (b = ca->buckets;
+	     b < ca->buckets + ca->sb.nbuckets;
+	     b++, d++) {
+		if (d == end) {
+			ca->prio_buckets[bucket_nr] = bucket;
+			ca->prio_last_buckets[bucket_nr] = bucket;
+			bucket_nr++;
+
+			prio_io(ca, bucket, READ_SYNC);
+
+			if (p->csum != bch_crc64(&p->magic, bucket_bytes(ca) - 8))
+				pr_warn("bad csum reading priorities");
+
+			if (p->magic != pset_magic(&ca->sb))
+				pr_warn("bad magic reading priorities");
+
+			bucket = p->next_bucket;
+			d = p->data;
+		}
+
+		b->prio = le16_to_cpu(d->prio);
+		b->gen = b->last_gc = d->gen;
+	}
+}
+
+/* Bcache device */
+
+static int open_dev(struct block_device *b, fmode_t mode)
+{
+	struct bcache_device *d = b->bd_disk->private_data;
+	if (test_bit(BCACHE_DEV_CLOSING, &d->flags))
+		return -ENXIO;
+
+	closure_get(&d->cl);
+	return 0;
+}
+
+static void release_dev(struct gendisk *b, fmode_t mode)
+{
+	struct bcache_device *d = b->private_data;
+	closure_put(&d->cl);
+}
+
+static int ioctl_dev(struct block_device *b, fmode_t mode,
+		     unsigned int cmd, unsigned long arg)
+{
+	struct bcache_device *d = b->bd_disk->private_data;
+	return d->ioctl(d, mode, cmd, arg);
+}
+
+static const struct block_device_operations bcache_ops = {
+	.open		= open_dev,
+	.release	= release_dev,
+	.ioctl		= ioctl_dev,
+	.owner		= THIS_MODULE,
+};
+
+void bcache_device_stop(struct bcache_device *d)
+{
+	if (!test_and_set_bit(BCACHE_DEV_CLOSING, &d->flags))
+		closure_queue(&d->cl);
+}
+
+static void bcache_device_unlink(struct bcache_device *d)
+{
+	lockdep_assert_held(&bch_register_lock);
+
+	if (d->c && !test_and_set_bit(BCACHE_DEV_UNLINK_DONE, &d->flags)) {
+		unsigned i;
+		struct cache *ca;
+
+		sysfs_remove_link(&d->c->kobj, d->name);
+		sysfs_remove_link(&d->kobj, "cache");
+
+		for_each_cache(ca, d->c, i)
+			bd_unlink_disk_holder(ca->bdev, d->disk);
+	}
+}
+
+static void bcache_device_link(struct bcache_device *d, struct cache_set *c,
+			       const char *name)
+{
+	unsigned i;
+	struct cache *ca;
+
+	for_each_cache(ca, d->c, i)
+		bd_link_disk_holder(ca->bdev, d->disk);
+
+	snprintf(d->name, BCACHEDEVNAME_SIZE,
+		 "%s%u", name, d->id);
+
+	WARN(sysfs_create_link(&d->kobj, &c->kobj, "cache") ||
+	     sysfs_create_link(&c->kobj, &d->kobj, d->name),
+	     "Couldn't create device <-> cache set symlinks");
+}
+
+static void bcache_device_detach(struct bcache_device *d)
+{
+	lockdep_assert_held(&bch_register_lock);
+
+	if (test_bit(BCACHE_DEV_DETACHING, &d->flags)) {
+		struct uuid_entry *u = d->c->uuids + d->id;
+
+		SET_UUID_FLASH_ONLY(u, 0);
+		memcpy(u->uuid, invalid_uuid, 16);
+		u->invalidated = cpu_to_le32(get_seconds());
+		bch_uuid_write(d->c);
+	}
+
+	bcache_device_unlink(d);
+
+	d->c->devices[d->id] = NULL;
+	closure_put(&d->c->caching);
+	d->c = NULL;
+}
+
+static void bcache_device_attach(struct bcache_device *d, struct cache_set *c,
+				 unsigned id)
+{
+	BUG_ON(test_bit(CACHE_SET_STOPPING, &c->flags));
+
+	d->id = id;
+	d->c = c;
+	c->devices[id] = d;
+
+	closure_get(&c->caching);
+}
+
+static void bcache_device_free(struct bcache_device *d)
+{
+	lockdep_assert_held(&bch_register_lock);
+
+	pr_info("%s stopped", d->disk->disk_name);
+
+	if (d->c)
+		bcache_device_detach(d);
+	if (d->disk && d->disk->flags & GENHD_FL_UP)
+		del_gendisk(d->disk);
+	if (d->disk && d->disk->queue)
+		blk_cleanup_queue(d->disk->queue);
+	if (d->disk) {
+		ida_simple_remove(&bcache_minor, d->disk->first_minor);
+		put_disk(d->disk);
+	}
+
+	bio_split_pool_free(&d->bio_split_hook);
+	if (d->bio_split)
+		bioset_free(d->bio_split);
+	if (is_vmalloc_addr(d->full_dirty_stripes))
+		vfree(d->full_dirty_stripes);
+	else
+		kfree(d->full_dirty_stripes);
+	if (is_vmalloc_addr(d->stripe_sectors_dirty))
+		vfree(d->stripe_sectors_dirty);
+	else
+		kfree(d->stripe_sectors_dirty);
+
+	closure_debug_destroy(&d->cl);
+}
+
+static int bcache_device_init(struct bcache_device *d, unsigned block_size,
+			      sector_t sectors)
+{
+	struct request_queue *q;
+	size_t n;
+	int minor;
+
+	if (!d->stripe_size)
+		d->stripe_size = 1 << 31;
+
+	d->nr_stripes = DIV_ROUND_UP_ULL(sectors, d->stripe_size);
+
+	if (!d->nr_stripes ||
+	    d->nr_stripes > INT_MAX ||
+	    d->nr_stripes > SIZE_MAX / sizeof(atomic_t)) {
+		pr_err("nr_stripes too large");
+		return -ENOMEM;
+	}
+
+	n = d->nr_stripes * sizeof(atomic_t);
+	d->stripe_sectors_dirty = n < PAGE_SIZE << 6
+		? kzalloc(n, GFP_KERNEL)
+		: vzalloc(n);
+	if (!d->stripe_sectors_dirty)
+		return -ENOMEM;
+
+	n = BITS_TO_LONGS(d->nr_stripes) * sizeof(unsigned long);
+	d->full_dirty_stripes = n < PAGE_SIZE << 6
+		? kzalloc(n, GFP_KERNEL)
+		: vzalloc(n);
+	if (!d->full_dirty_stripes)
+		return -ENOMEM;
+
+	minor = ida_simple_get(&bcache_minor, 0, MINORMASK + 1, GFP_KERNEL);
+	if (minor < 0)
+		return minor;
+
+	if (!(d->bio_split = bioset_create(4, offsetof(struct bbio, bio))) ||
+	    bio_split_pool_init(&d->bio_split_hook) ||
+	    !(d->disk = alloc_disk(1))) {
+		ida_simple_remove(&bcache_minor, minor);
+		return -ENOMEM;
+	}
+
+	set_capacity(d->disk, sectors);
+	snprintf(d->disk->disk_name, DISK_NAME_LEN, "bcache%i", minor);
+
+	d->disk->major		= bcache_major;
+	d->disk->first_minor	= minor;
+	d->disk->fops		= &bcache_ops;
+	d->disk->private_data	= d;
+
+	q = blk_alloc_queue(GFP_KERNEL);
+	if (!q)
+		return -ENOMEM;
+
+	blk_queue_make_request(q, NULL);
+	d->disk->queue			= q;
+	q->queuedata			= d;
+	q->backing_dev_info.congested_data = d;
+	q->limits.max_hw_sectors	= UINT_MAX;
+	q->limits.max_sectors		= UINT_MAX;
+	q->limits.max_segment_size	= UINT_MAX;
+	q->limits.max_segments		= BIO_MAX_PAGES;
+	q->limits.max_discard_sectors	= UINT_MAX;
+	q->limits.discard_granularity	= 512;
+	q->limits.io_min		= block_size;
+	q->limits.logical_block_size	= block_size;
+	q->limits.physical_block_size	= block_size;
+	set_bit(QUEUE_FLAG_NONROT,	&d->disk->queue->queue_flags);
+	set_bit(QUEUE_FLAG_DISCARD,	&d->disk->queue->queue_flags);
+
+	blk_queue_flush(q, REQ_FLUSH|REQ_FUA);
+
+	return 0;
+}
+
+/* Cached device */
+
+static void calc_cached_dev_sectors(struct cache_set *c)
+{
+	uint64_t sectors = 0;
+	struct cached_dev *dc;
+
+	list_for_each_entry(dc, &c->cached_devs, list)
+		sectors += bdev_sectors(dc->bdev);
+
+	c->cached_dev_sectors = sectors;
+}
+
+void bch_cached_dev_run(struct cached_dev *dc)
+{
+	struct bcache_device *d = &dc->disk;
+	char buf[SB_LABEL_SIZE + 1];
+	char *env[] = {
+		"DRIVER=bcache",
+		kasprintf(GFP_KERNEL, "CACHED_UUID=%pU", dc->sb.uuid),
+		NULL,
+		NULL,
+	};
+
+	memcpy(buf, dc->sb.label, SB_LABEL_SIZE);
+	buf[SB_LABEL_SIZE] = '\0';
+	env[2] = kasprintf(GFP_KERNEL, "CACHED_LABEL=%s", buf);
+
+	if (atomic_xchg(&dc->running, 1))
+		return;
+
+	if (!d->c &&
+	    BDEV_STATE(&dc->sb) != BDEV_STATE_NONE) {
+		struct closure cl;
+		closure_init_stack(&cl);
+
+		SET_BDEV_STATE(&dc->sb, BDEV_STATE_STALE);
+		bch_write_bdev_super(dc, &cl);
+		closure_sync(&cl);
+	}
+
+	add_disk(d->disk);
+	bd_link_disk_holder(dc->bdev, dc->disk.disk);
+	/* won't show up in the uevent file, use udevadm monitor -e instead
+	 * only class / kset properties are persistent */
+	kobject_uevent_env(&disk_to_dev(d->disk)->kobj, KOBJ_CHANGE, env);
+	kfree(env[1]);
+	kfree(env[2]);
+
+	if (sysfs_create_link(&d->kobj, &disk_to_dev(d->disk)->kobj, "dev") ||
+	    sysfs_create_link(&disk_to_dev(d->disk)->kobj, &d->kobj, "bcache"))
+		pr_debug("error creating sysfs link");
+}
+
+static void cached_dev_detach_finish(struct work_struct *w)
+{
+	struct cached_dev *dc = container_of(w, struct cached_dev, detach);
+	char buf[BDEVNAME_SIZE];
+	struct closure cl;
+	closure_init_stack(&cl);
+
+	BUG_ON(!test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags));
+	BUG_ON(atomic_read(&dc->count));
+
+	mutex_lock(&bch_register_lock);
+
+	memset(&dc->sb.set_uuid, 0, 16);
+	SET_BDEV_STATE(&dc->sb, BDEV_STATE_NONE);
+
+	bch_write_bdev_super(dc, &cl);
+	closure_sync(&cl);
+
+	bcache_device_detach(&dc->disk);
+	list_move(&dc->list, &uncached_devices);
+
+	clear_bit(BCACHE_DEV_DETACHING, &dc->disk.flags);
+
+	mutex_unlock(&bch_register_lock);
+
+	pr_info("Caching disabled for %s", bdevname(dc->bdev, buf));
+
+	/* Drop ref we took in cached_dev_detach() */
+	closure_put(&dc->disk.cl);
+}
+
+void bch_cached_dev_detach(struct cached_dev *dc)
+{
+	lockdep_assert_held(&bch_register_lock);
+
+	if (test_bit(BCACHE_DEV_CLOSING, &dc->disk.flags))
+		return;
+
+	if (test_and_set_bit(BCACHE_DEV_DETACHING, &dc->disk.flags))
+		return;
+
+	/*
+	 * Block the device from being closed and freed until we're finished
+	 * detaching
+	 */
+	closure_get(&dc->disk.cl);
+
+	bch_writeback_queue(dc);
+	cached_dev_put(dc);
+}
+
+int bch_cached_dev_attach(struct cached_dev *dc, struct cache_set *c)
+{
+	uint32_t rtime = cpu_to_le32(get_seconds());
+	struct uuid_entry *u;
+	char buf[BDEVNAME_SIZE];
+
+	bdevname(dc->bdev, buf);
+
+	if (memcmp(dc->sb.set_uuid, c->sb.set_uuid, 16))
+		return -ENOENT;
+
+	if (dc->disk.c) {
+		pr_err("Can't attach %s: already attached", buf);
+		return -EINVAL;
+	}
+
+	if (test_bit(CACHE_SET_STOPPING, &c->flags)) {
+		pr_err("Can't attach %s: shutting down", buf);
+		return -EINVAL;
+	}
+
+	if (dc->sb.block_size < c->sb.block_size) {
+		/* Will die */
+		pr_err("Couldn't attach %s: block size less than set's block size",
+		       buf);
+		return -EINVAL;
+	}
+
+	u = uuid_find(c, dc->sb.uuid);
+
+	if (u &&
+	    (BDEV_STATE(&dc->sb) == BDEV_STATE_STALE ||
+	     BDEV_STATE(&dc->sb) == BDEV_STATE_NONE)) {
+		memcpy(u->uuid, invalid_uuid, 16);
+		u->invalidated = cpu_to_le32(get_seconds());
+		u = NULL;
+	}
+
+	if (!u) {
+		if (BDEV_STATE(&dc->sb) == BDEV_STATE_DIRTY) {
+			pr_err("Couldn't find uuid for %s in set", buf);
+			return -ENOENT;
+		}
+
+		u = uuid_find_empty(c);
+		if (!u) {
+			pr_err("Not caching %s, no room for UUID", buf);
+			return -EINVAL;
+		}
+	}
+
+	/* Deadlocks since we're called via sysfs...
+	sysfs_remove_file(&dc->kobj, &sysfs_attach);
+	 */
+
+	if (bch_is_zero(u->uuid, 16)) {
+		struct closure cl;
+		closure_init_stack(&cl);
+
+		memcpy(u->uuid, dc->sb.uuid, 16);
+		memcpy(u->label, dc->sb.label, SB_LABEL_SIZE);
+		u->first_reg = u->last_reg = rtime;
+		bch_uuid_write(c);
+
+		memcpy(dc->sb.set_uuid, c->sb.set_uuid, 16);
+		SET_BDEV_STATE(&dc->sb, BDEV_STATE_CLEAN);
+
+		bch_write_bdev_super(dc, &cl);
+		closure_sync(&cl);
+	} else {
+		u->last_reg = rtime;
+		bch_uuid_write(c);
+	}
+
+	bcache_device_attach(&dc->disk, c, u - c->uuids);
+	list_move(&dc->list, &c->cached_devs);
+	calc_cached_dev_sectors(c);
+
+	smp_wmb();
+	/*
+	 * dc->c must be set before dc->count != 0 - paired with the mb in
+	 * cached_dev_get()
+	 */
+	atomic_set(&dc->count, 1);
+
+	if (BDEV_STATE(&dc->sb) == BDEV_STATE_DIRTY) {
+		bch_sectors_dirty_init(dc);
+		atomic_set(&dc->has_dirty, 1);
+		atomic_inc(&dc->count);
+		bch_writeback_queue(dc);
+	}
+
+	bch_cached_dev_run(dc);
+	bcache_device_link(&dc->disk, c, "bdev");
+
+	pr_info("Caching %s as %s on set %pU",
+		bdevname(dc->bdev, buf), dc->disk.disk->disk_name,
+		dc->disk.c->sb.set_uuid);
+	return 0;
+}
+
+void bch_cached_dev_release(struct kobject *kobj)
+{
+	struct cached_dev *dc = container_of(kobj, struct cached_dev,
+					     disk.kobj);
+	kfree(dc);
+	module_put(THIS_MODULE);
+}
+
+static void cached_dev_free(struct closure *cl)
+{
+	struct cached_dev *dc = container_of(cl, struct cached_dev, disk.cl);
+
+	cancel_delayed_work_sync(&dc->writeback_rate_update);
+	kthread_stop(dc->writeback_thread);
+
+	mutex_lock(&bch_register_lock);
+
+	if (atomic_read(&dc->running))
+		bd_unlink_disk_holder(dc->bdev, dc->disk.disk);
+	bcache_device_free(&dc->disk);
+	list_del(&dc->list);
+
+	mutex_unlock(&bch_register_lock);
+
+	if (!IS_ERR_OR_NULL(dc->bdev)) {
+		if (dc->bdev->bd_disk)
+			blk_sync_queue(bdev_get_queue(dc->bdev));
+
+		blkdev_put(dc->bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
+	}
+
+	wake_up(&unregister_wait);
+
+	kobject_put(&dc->disk.kobj);
+}
+
+static void cached_dev_flush(struct closure *cl)
+{
+	struct cached_dev *dc = container_of(cl, struct cached_dev, disk.cl);
+	struct bcache_device *d = &dc->disk;
+
+	mutex_lock(&bch_register_lock);
+	bcache_device_unlink(d);
+	mutex_unlock(&bch_register_lock);
+
+	bch_cache_accounting_destroy(&dc->accounting);
+	kobject_del(&d->kobj);
+
+	continue_at(cl, cached_dev_free, system_wq);
+}
+
+static int cached_dev_init(struct cached_dev *dc, unsigned block_size)
+{
+	int ret;
+	struct io *io;
+	struct request_queue *q = bdev_get_queue(dc->bdev);
+
+	__module_get(THIS_MODULE);
+	INIT_LIST_HEAD(&dc->list);
+	closure_init(&dc->disk.cl, NULL);
+	set_closure_fn(&dc->disk.cl, cached_dev_flush, system_wq);
+	kobject_init(&dc->disk.kobj, &bch_cached_dev_ktype);
+	INIT_WORK(&dc->detach, cached_dev_detach_finish);
+	sema_init(&dc->sb_write_mutex, 1);
+	INIT_LIST_HEAD(&dc->io_lru);
+	spin_lock_init(&dc->io_lock);
+	bch_cache_accounting_init(&dc->accounting, &dc->disk.cl);
+
+	dc->sequential_cutoff		= 4 << 20;
+
+	for (io = dc->io; io < dc->io + RECENT_IO; io++) {
+		list_add(&io->lru, &dc->io_lru);
+		hlist_add_head(&io->hash, dc->io_hash + RECENT_IO);
+	}
+
+	dc->disk.stripe_size = q->limits.io_opt >> 9;
+
+	if (dc->disk.stripe_size)
+		dc->partial_stripes_expensive =
+			q->limits.raid_partial_stripes_expensive;
+
+	ret = bcache_device_init(&dc->disk, block_size,
+			 dc->bdev->bd_part->nr_sects - dc->sb.data_offset);
+	if (ret)
+		return ret;
+
+	set_capacity(dc->disk.disk,
+		     dc->bdev->bd_part->nr_sects - dc->sb.data_offset);
+
+	dc->disk.disk->queue->backing_dev_info.ra_pages =
+		max(dc->disk.disk->queue->backing_dev_info.ra_pages,
+		    q->backing_dev_info.ra_pages);
+
+	bch_cached_dev_request_init(dc);
+	bch_cached_dev_writeback_init(dc);
+	return 0;
+}
+
+/* Cached device - bcache superblock */
+
+static void register_bdev(struct cache_sb *sb, struct page *sb_page,
+				 struct block_device *bdev,
+				 struct cached_dev *dc)
+{
+	char name[BDEVNAME_SIZE];
+	const char *err = "cannot allocate memory";
+	struct cache_set *c;
+
+	memcpy(&dc->sb, sb, sizeof(struct cache_sb));
+	dc->bdev = bdev;
+	dc->bdev->bd_holder = dc;
+
+	bio_init(&dc->sb_bio);
+	dc->sb_bio.bi_max_vecs	= 1;
+	dc->sb_bio.bi_io_vec	= dc->sb_bio.bi_inline_vecs;
+	dc->sb_bio.bi_io_vec[0].bv_page = sb_page;
+	get_page(sb_page);
+
+	if (cached_dev_init(dc, sb->block_size << 9))
+		goto err;
+
+	err = "error creating kobject";
+	if (kobject_add(&dc->disk.kobj, &part_to_dev(bdev->bd_part)->kobj,
+			"bcache"))
+		goto err;
+	if (bch_cache_accounting_add_kobjs(&dc->accounting, &dc->disk.kobj))
+		goto err;
+
+	pr_info("registered backing device %s", bdevname(bdev, name));
+
+	list_add(&dc->list, &uncached_devices);
+	list_for_each_entry(c, &bch_cache_sets, list)
+		bch_cached_dev_attach(dc, c);
+
+	if (BDEV_STATE(&dc->sb) == BDEV_STATE_NONE ||
+	    BDEV_STATE(&dc->sb) == BDEV_STATE_STALE)
+		bch_cached_dev_run(dc);
+
+	return;
+err:
+	pr_notice("error opening %s: %s", bdevname(bdev, name), err);
+	bcache_device_stop(&dc->disk);
+}
+
+/* Flash only volumes */
+
+void bch_flash_dev_release(struct kobject *kobj)
+{
+	struct bcache_device *d = container_of(kobj, struct bcache_device,
+					       kobj);
+	kfree(d);
+}
+
+static void flash_dev_free(struct closure *cl)
+{
+	struct bcache_device *d = container_of(cl, struct bcache_device, cl);
+	bcache_device_free(d);
+	kobject_put(&d->kobj);
+}
+
+static void flash_dev_flush(struct closure *cl)
+{
+	struct bcache_device *d = container_of(cl, struct bcache_device, cl);
+
+	bcache_device_unlink(d);
+	kobject_del(&d->kobj);
+	continue_at(cl, flash_dev_free, system_wq);
+}
+
+static int flash_dev_run(struct cache_set *c, struct uuid_entry *u)
+{
+	struct bcache_device *d = kzalloc(sizeof(struct bcache_device),
+					  GFP_KERNEL);
+	if (!d)
+		return -ENOMEM;
+
+	closure_init(&d->cl, NULL);
+	set_closure_fn(&d->cl, flash_dev_flush, system_wq);
+
+	kobject_init(&d->kobj, &bch_flash_dev_ktype);
+
+	if (bcache_device_init(d, block_bytes(c), u->sectors))
+		goto err;
+
+	bcache_device_attach(d, c, u - c->uuids);
+	bch_flash_dev_request_init(d);
+	add_disk(d->disk);
+
+	if (kobject_add(&d->kobj, &disk_to_dev(d->disk)->kobj, "bcache"))
+		goto err;
+
+	bcache_device_link(d, c, "volume");
+
+	return 0;
+err:
+	kobject_put(&d->kobj);
+	return -ENOMEM;
+}
+
+static int flash_devs_run(struct cache_set *c)
+{
+	int ret = 0;
+	struct uuid_entry *u;
+
+	for (u = c->uuids;
+	     u < c->uuids + c->nr_uuids && !ret;
+	     u++)
+		if (UUID_FLASH_ONLY(u))
+			ret = flash_dev_run(c, u);
+
+	return ret;
+}
+
+int bch_flash_dev_create(struct cache_set *c, uint64_t size)
+{
+	struct uuid_entry *u;
+
+	if (test_bit(CACHE_SET_STOPPING, &c->flags))
+		return -EINTR;
+
+	u = uuid_find_empty(c);
+	if (!u) {
+		pr_err("Can't create volume, no room for UUID");
+		return -EINVAL;
+	}
+
+	get_random_bytes(u->uuid, 16);
+	memset(u->label, 0, 32);
+	u->first_reg = u->last_reg = cpu_to_le32(get_seconds());
+
+	SET_UUID_FLASH_ONLY(u, 1);
+	u->sectors = size >> 9;
+
+	bch_uuid_write(c);
+
+	return flash_dev_run(c, u);
+}
+
+/* Cache set */
+
+__printf(2, 3)
+bool bch_cache_set_error(struct cache_set *c, const char *fmt, ...)
+{
+	va_list args;
+
+	if (c->on_error != ON_ERROR_PANIC &&
+	    test_bit(CACHE_SET_STOPPING, &c->flags))
+		return false;
+
+	/* XXX: we can be called from atomic context
+	acquire_console_sem();
+	*/
+
+	printk(KERN_ERR "bcache: error on %pU: ", c->sb.set_uuid);
+
+	va_start(args, fmt);
+	vprintk(fmt, args);
+	va_end(args);
+
+	printk(", disabling caching\n");
+
+	if (c->on_error == ON_ERROR_PANIC)
+		panic("panic forced after error\n");
+
+	bch_cache_set_unregister(c);
+	return true;
+}
+
+void bch_cache_set_release(struct kobject *kobj)
+{
+	struct cache_set *c = container_of(kobj, struct cache_set, kobj);
+	kfree(c);
+	module_put(THIS_MODULE);
+}
+
+static void cache_set_free(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, cl);
+	struct cache *ca;
+	unsigned i;
+
+	if (!IS_ERR_OR_NULL(c->debug))
+		debugfs_remove(c->debug);
+
+	bch_open_buckets_free(c);
+	bch_btree_cache_free(c);
+	bch_journal_free(c);
+
+	for_each_cache(ca, c, i)
+		if (ca)
+			kobject_put(&ca->kobj);
+
+	bch_bset_sort_state_free(&c->sort);
+	free_pages((unsigned long) c->uuids, ilog2(bucket_pages(c)));
+
+	if (c->moving_gc_wq)
+		destroy_workqueue(c->moving_gc_wq);
+	if (c->bio_split)
+		bioset_free(c->bio_split);
+	if (c->fill_iter)
+		mempool_destroy(c->fill_iter);
+	if (c->bio_meta)
+		mempool_destroy(c->bio_meta);
+	if (c->search)
+		mempool_destroy(c->search);
+	kfree(c->devices);
+
+	mutex_lock(&bch_register_lock);
+	list_del(&c->list);
+	mutex_unlock(&bch_register_lock);
+
+	pr_info("Cache set %pU unregistered", c->sb.set_uuid);
+	wake_up(&unregister_wait);
+
+	closure_debug_destroy(&c->cl);
+	kobject_put(&c->kobj);
+}
+
+static void cache_set_flush(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, caching);
+	struct cache *ca;
+	struct btree *b;
+	unsigned i;
+
+	bch_cache_accounting_destroy(&c->accounting);
+
+	kobject_put(&c->internal);
+	kobject_del(&c->kobj);
+
+	if (c->gc_thread)
+		kthread_stop(c->gc_thread);
+
+	if (!IS_ERR_OR_NULL(c->root))
+		list_add(&c->root->list, &c->btree_cache);
+
+	/* Should skip this if we're unregistering because of an error */
+	list_for_each_entry(b, &c->btree_cache, list) {
+		mutex_lock(&b->write_lock);
+		if (btree_node_dirty(b))
+			__bch_btree_node_write(b, NULL);
+		mutex_unlock(&b->write_lock);
+	}
+
+	for_each_cache(ca, c, i)
+		if (ca->alloc_thread)
+			kthread_stop(ca->alloc_thread);
+
+	cancel_delayed_work_sync(&c->journal.work);
+	/* flush last journal entry if needed */
+	c->journal.work.work.func(&c->journal.work.work);
+
+	closure_return(cl);
+}
+
+static void __cache_set_unregister(struct closure *cl)
+{
+	struct cache_set *c = container_of(cl, struct cache_set, caching);
+	struct cached_dev *dc;
+	size_t i;
+
+	mutex_lock(&bch_register_lock);
+
+	for (i = 0; i < c->nr_uuids; i++)
+		if (c->devices[i]) {
+			if (!UUID_FLASH_ONLY(&c->uuids[i]) &&
+			    test_bit(CACHE_SET_UNREGISTERING, &c->flags)) {
+				dc = container_of(c->devices[i],
+						  struct cached_dev, disk);
+				bch_cached_dev_detach(dc);
+			} else {
+				bcache_device_stop(c->devices[i]);
+			}
+		}
+
+	mutex_unlock(&bch_register_lock);
+
+	continue_at(cl, cache_set_flush, system_wq);
+}
+
+void bch_cache_set_stop(struct cache_set *c)
+{
+	if (!test_and_set_bit(CACHE_SET_STOPPING, &c->flags))
+		closure_queue(&c->caching);
+}
+
+void bch_cache_set_unregister(struct cache_set *c)
+{
+	set_bit(CACHE_SET_UNREGISTERING, &c->flags);
+	bch_cache_set_stop(c);
+}
+
+#define alloc_bucket_pages(gfp, c)			\
+	((void *) __get_free_pages(__GFP_ZERO|gfp, ilog2(bucket_pages(c))))
+
+struct cache_set *bch_cache_set_alloc(struct cache_sb *sb)
+{
+	int iter_size;
+	struct cache_set *c = kzalloc(sizeof(struct cache_set), GFP_KERNEL);
+	if (!c)
+		return NULL;
+
+	__module_get(THIS_MODULE);
+	closure_init(&c->cl, NULL);
+	set_closure_fn(&c->cl, cache_set_free, system_wq);
+
+	closure_init(&c->caching, &c->cl);
+	set_closure_fn(&c->caching, __cache_set_unregister, system_wq);
+
+	/* Maybe create continue_at_noreturn() and use it here? */
+	closure_set_stopped(&c->cl);
+	closure_put(&c->cl);
+
+	kobject_init(&c->kobj, &bch_cache_set_ktype);
+	kobject_init(&c->internal, &bch_cache_set_internal_ktype);
+
+	bch_cache_accounting_init(&c->accounting, &c->cl);
+
+	memcpy(c->sb.set_uuid, sb->set_uuid, 16);
+	c->sb.block_size	= sb->block_size;
+	c->sb.bucket_size	= sb->bucket_size;
+	c->sb.nr_in_set		= sb->nr_in_set;
+	c->sb.last_mount	= sb->last_mount;
+	c->bucket_bits		= ilog2(sb->bucket_size);
+	c->block_bits		= ilog2(sb->block_size);
+	c->nr_uuids		= bucket_bytes(c) / sizeof(struct uuid_entry);
+
+	c->btree_pages		= bucket_pages(c);
+	if (c->btree_pages > BTREE_MAX_PAGES)
+		c->btree_pages = max_t(int, c->btree_pages / 4,
+				       BTREE_MAX_PAGES);
+
+	sema_init(&c->sb_write_mutex, 1);
+	mutex_init(&c->bucket_lock);
+	init_waitqueue_head(&c->btree_cache_wait);
+	init_waitqueue_head(&c->bucket_wait);
+	sema_init(&c->uuid_write_mutex, 1);
+
+	spin_lock_init(&c->btree_gc_time.lock);
+	spin_lock_init(&c->btree_split_time.lock);
+	spin_lock_init(&c->btree_read_time.lock);
+
+	bch_moving_init_cache_set(c);
+
+	INIT_LIST_HEAD(&c->list);
+	INIT_LIST_HEAD(&c->cached_devs);
+	INIT_LIST_HEAD(&c->btree_cache);
+	INIT_LIST_HEAD(&c->btree_cache_freeable);
+	INIT_LIST_HEAD(&c->btree_cache_freed);
+	INIT_LIST_HEAD(&c->data_buckets);
+
+	c->search = mempool_create_slab_pool(32, bch_search_cache);
+	if (!c->search)
+		goto err;
+
+	iter_size = (sb->bucket_size / sb->block_size + 1) *
+		sizeof(struct btree_iter_set);
+
+	if (!(c->devices = kzalloc(c->nr_uuids * sizeof(void *), GFP_KERNEL)) ||
+	    !(c->bio_meta = mempool_create_kmalloc_pool(2,
+				sizeof(struct bbio) + sizeof(struct bio_vec) *
+				bucket_pages(c))) ||
+	    !(c->fill_iter = mempool_create_kmalloc_pool(1, iter_size)) ||
+	    !(c->bio_split = bioset_create(4, offsetof(struct bbio, bio))) ||
+	    !(c->uuids = alloc_bucket_pages(GFP_KERNEL, c)) ||
+	    !(c->moving_gc_wq = create_workqueue("bcache_gc")) ||
+	    bch_journal_alloc(c) ||
+	    bch_btree_cache_alloc(c) ||
+	    bch_open_buckets_alloc(c) ||
+	    bch_bset_sort_state_init(&c->sort, ilog2(c->btree_pages)))
+		goto err;
+
+	c->congested_read_threshold_us	= 2000;
+	c->congested_write_threshold_us	= 20000;
+	c->error_limit	= 8 << IO_ERROR_SHIFT;
+
+	return c;
+err:
+	bch_cache_set_unregister(c);
+	return NULL;
+}
+
+static void run_cache_set(struct cache_set *c)
+{
+	const char *err = "cannot allocate memory";
+	struct cached_dev *dc, *t;
+	struct cache *ca;
+	struct closure cl;
+	unsigned i;
+
+	closure_init_stack(&cl);
+
+	for_each_cache(ca, c, i)
+		c->nbuckets += ca->sb.nbuckets;
+
+	if (CACHE_SYNC(&c->sb)) {
+		LIST_HEAD(journal);
+		struct bkey *k;
+		struct jset *j;
+
+		err = "cannot allocate memory for journal";
+		if (bch_journal_read(c, &journal))
+			goto err;
+
+		pr_debug("btree_journal_read() done");
+
+		err = "no journal entries found";
+		if (list_empty(&journal))
+			goto err;
+
+		j = &list_entry(journal.prev, struct journal_replay, list)->j;
+
+		err = "IO error reading priorities";
+		for_each_cache(ca, c, i)
+			prio_read(ca, j->prio_bucket[ca->sb.nr_this_dev]);
+
+		/*
+		 * If prio_read() fails it'll call cache_set_error and we'll
+		 * tear everything down right away, but if we perhaps checked
+		 * sooner we could avoid journal replay.
+		 */
+
+		k = &j->btree_root;
+
+		err = "bad btree root";
+		if (__bch_btree_ptr_invalid(c, k))
+			goto err;
+
+		err = "error reading btree root";
+		c->root = bch_btree_node_get(c, NULL, k, j->btree_level, true);
+		if (IS_ERR_OR_NULL(c->root))
+			goto err;
+
+		list_del_init(&c->root->list);
+		rw_unlock(true, c->root);
+
+		err = uuid_read(c, j, &cl);
+		if (err)
+			goto err;
+
+		err = "error in recovery";
+		if (bch_btree_check(c))
+			goto err;
+
+		bch_journal_mark(c, &journal);
+		bch_initial_gc_finish(c);
+		pr_debug("btree_check() done");
+
+		/*
+		 * bcache_journal_next() can't happen sooner, or
+		 * btree_gc_finish() will give spurious errors about last_gc >
+		 * gc_gen - this is a hack but oh well.
+		 */
+		bch_journal_next(&c->journal);
+
+		err = "error starting allocator thread";
+		for_each_cache(ca, c, i)
+			if (bch_cache_allocator_start(ca))
+				goto err;
+
+		/*
+		 * First place it's safe to allocate: btree_check() and
+		 * btree_gc_finish() have to run before we have buckets to
+		 * allocate, and bch_bucket_alloc_set() might cause a journal
+		 * entry to be written so bcache_journal_next() has to be called
+		 * first.
+		 *
+		 * If the uuids were in the old format we have to rewrite them
+		 * before the next journal entry is written:
+		 */
+		if (j->version < BCACHE_JSET_VERSION_UUID)
+			__uuid_write(c);
+
+		bch_journal_replay(c, &journal);
+	} else {
+		pr_notice("invalidating existing data");
+
+		for_each_cache(ca, c, i) {
+			unsigned j;
+
+			ca->sb.keys = clamp_t(int, ca->sb.nbuckets >> 7,
+					      2, SB_JOURNAL_BUCKETS);
+
+			for (j = 0; j < ca->sb.keys; j++)
+				ca->sb.d[j] = ca->sb.first_bucket + j;
+		}
+
+		bch_initial_gc_finish(c);
+
+		err = "error starting allocator thread";
+		for_each_cache(ca, c, i)
+			if (bch_cache_allocator_start(ca))
+				goto err;
+
+		mutex_lock(&c->bucket_lock);
+		for_each_cache(ca, c, i)
+			bch_prio_write(ca);
+		mutex_unlock(&c->bucket_lock);
+
+		err = "cannot allocate new UUID bucket";
+		if (__uuid_write(c))
+			goto err;
+
+		err = "cannot allocate new btree root";
+		c->root = bch_btree_node_alloc(c, NULL, 0);
+		if (IS_ERR_OR_NULL(c->root))
+			goto err;
+
+		mutex_lock(&c->root->write_lock);
+		bkey_copy_key(&c->root->key, &MAX_KEY);
+		bch_btree_node_write(c->root, &cl);
+		mutex_unlock(&c->root->write_lock);
+
+		bch_btree_set_root(c->root);
+		rw_unlock(true, c->root);
+
+		/*
+		 * We don't want to write the first journal entry until
+		 * everything is set up - fortunately journal entries won't be
+		 * written until the SET_CACHE_SYNC() here:
+		 */
+		SET_CACHE_SYNC(&c->sb, true);
+
+		bch_journal_next(&c->journal);
+		bch_journal_meta(c, &cl);
+	}
+
+	err = "error starting gc thread";
+	if (bch_gc_thread_start(c))
+		goto err;
+
+	closure_sync(&cl);
+	c->sb.last_mount = get_seconds();
+	bcache_write_super(c);
+
+	list_for_each_entry_safe(dc, t, &uncached_devices, list)
+		bch_cached_dev_attach(dc, c);
+
+	flash_devs_run(c);
+
+	return;
+err:
+	closure_sync(&cl);
+	/* XXX: test this, it's broken */
+	bch_cache_set_error(c, "%s", err);
+}
+
+static bool can_attach_cache(struct cache *ca, struct cache_set *c)
+{
+	return ca->sb.block_size	== c->sb.block_size &&
+		ca->sb.bucket_size	== c->sb.bucket_size &&
+		ca->sb.nr_in_set	== c->sb.nr_in_set;
+}
+
+static const char *register_cache_set(struct cache *ca)
+{
+	char buf[12];
+	const char *err = "cannot allocate memory";
+	struct cache_set *c;
+
+	list_for_each_entry(c, &bch_cache_sets, list)
+		if (!memcmp(c->sb.set_uuid, ca->sb.set_uuid, 16)) {
+			if (c->cache[ca->sb.nr_this_dev])
+				return "duplicate cache set member";
+
+			if (!can_attach_cache(ca, c))
+				return "cache sb does not match set";
+
+			if (!CACHE_SYNC(&ca->sb))
+				SET_CACHE_SYNC(&c->sb, false);
+
+			goto found;
+		}
+
+	c = bch_cache_set_alloc(&ca->sb);
+	if (!c)
+		return err;
+
+	err = "error creating kobject";
+	if (kobject_add(&c->kobj, bcache_kobj, "%pU", c->sb.set_uuid) ||
+	    kobject_add(&c->internal, &c->kobj, "internal"))
+		goto err;
+
+	if (bch_cache_accounting_add_kobjs(&c->accounting, &c->kobj))
+		goto err;
+
+	bch_debug_init_cache_set(c);
+
+	list_add(&c->list, &bch_cache_sets);
+found:
+	sprintf(buf, "cache%i", ca->sb.nr_this_dev);
+	if (sysfs_create_link(&ca->kobj, &c->kobj, "set") ||
+	    sysfs_create_link(&c->kobj, &ca->kobj, buf))
+		goto err;
+
+	if (ca->sb.seq > c->sb.seq) {
+		c->sb.version		= ca->sb.version;
+		memcpy(c->sb.set_uuid, ca->sb.set_uuid, 16);
+		c->sb.flags             = ca->sb.flags;
+		c->sb.seq		= ca->sb.seq;
+		pr_debug("set version = %llu", c->sb.version);
+	}
+
+	ca->set = c;
+	ca->set->cache[ca->sb.nr_this_dev] = ca;
+	c->cache_by_alloc[c->caches_loaded++] = ca;
+
+	if (c->caches_loaded == c->sb.nr_in_set)
+		run_cache_set(c);
+
+	return NULL;
+err:
+	bch_cache_set_unregister(c);
+	return err;
+}
+
+/* Cache device */
+
+void bch_cache_release(struct kobject *kobj)
+{
+	struct cache *ca = container_of(kobj, struct cache, kobj);
+	unsigned i;
+
+	if (ca->set)
+		ca->set->cache[ca->sb.nr_this_dev] = NULL;
+
+	bio_split_pool_free(&ca->bio_split_hook);
+
+	free_pages((unsigned long) ca->disk_buckets, ilog2(bucket_pages(ca)));
+	kfree(ca->prio_buckets);
+	vfree(ca->buckets);
+
+	free_heap(&ca->heap);
+	free_fifo(&ca->free_inc);
+
+	for (i = 0; i < RESERVE_NR; i++)
+		free_fifo(&ca->free[i]);
+
+	if (ca->sb_bio.bi_inline_vecs[0].bv_page)
+		put_page(ca->sb_bio.bi_io_vec[0].bv_page);
+
+	if (!IS_ERR_OR_NULL(ca->bdev)) {
+		blk_sync_queue(bdev_get_queue(ca->bdev));
+		blkdev_put(ca->bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
+	}
+
+	kfree(ca);
+	module_put(THIS_MODULE);
+}
+
+static int cache_alloc(struct cache_sb *sb, struct cache *ca)
+{
+	size_t free;
+	struct bucket *b;
+
+	__module_get(THIS_MODULE);
+	kobject_init(&ca->kobj, &bch_cache_ktype);
+
+	bio_init(&ca->journal.bio);
+	ca->journal.bio.bi_max_vecs = 8;
+	ca->journal.bio.bi_io_vec = ca->journal.bio.bi_inline_vecs;
+
+	free = roundup_pow_of_two(ca->sb.nbuckets) >> 10;
+
+	if (!init_fifo(&ca->free[RESERVE_BTREE], 8, GFP_KERNEL) ||
+	    !init_fifo(&ca->free[RESERVE_PRIO], prio_buckets(ca), GFP_KERNEL) ||
+	    !init_fifo(&ca->free[RESERVE_MOVINGGC], free, GFP_KERNEL) ||
+	    !init_fifo(&ca->free[RESERVE_NONE], free, GFP_KERNEL) ||
+	    !init_fifo(&ca->free_inc,	free << 2, GFP_KERNEL) ||
+	    !init_heap(&ca->heap,	free << 3, GFP_KERNEL) ||
+	    !(ca->buckets	= vzalloc(sizeof(struct bucket) *
+					  ca->sb.nbuckets)) ||
+	    !(ca->prio_buckets	= kzalloc(sizeof(uint64_t) * prio_buckets(ca) *
+					  2, GFP_KERNEL)) ||
+	    !(ca->disk_buckets	= alloc_bucket_pages(GFP_KERNEL, ca)) ||
+	    bio_split_pool_init(&ca->bio_split_hook))
+		return -ENOMEM;
+
+	ca->prio_last_buckets = ca->prio_buckets + prio_buckets(ca);
+
+	for_each_bucket(b, ca)
+		atomic_set(&b->pin, 0);
+
+	return 0;
+}
+
+static void register_cache(struct cache_sb *sb, struct page *sb_page,
+				  struct block_device *bdev, struct cache *ca)
+{
+	char name[BDEVNAME_SIZE];
+	const char *err = "cannot allocate memory";
+
+	memcpy(&ca->sb, sb, sizeof(struct cache_sb));
+	ca->bdev = bdev;
+	ca->bdev->bd_holder = ca;
+
+	bio_init(&ca->sb_bio);
+	ca->sb_bio.bi_max_vecs	= 1;
+	ca->sb_bio.bi_io_vec	= ca->sb_bio.bi_inline_vecs;
+	ca->sb_bio.bi_io_vec[0].bv_page = sb_page;
+	get_page(sb_page);
+
+	if (blk_queue_discard(bdev_get_queue(ca->bdev)))
+		ca->discard = CACHE_DISCARD(&ca->sb);
+
+	if (cache_alloc(sb, ca) != 0)
+		goto err;
+
+	err = "error creating kobject";
+	if (kobject_add(&ca->kobj, &part_to_dev(bdev->bd_part)->kobj, "bcache"))
+		goto err;
+
+	mutex_lock(&bch_register_lock);
+	err = register_cache_set(ca);
+	mutex_unlock(&bch_register_lock);
+
+	if (err)
+		goto err;
+
+	pr_info("registered cache device %s", bdevname(bdev, name));
+	return;
+err:
+	pr_notice("error opening %s: %s", bdevname(bdev, name), err);
+	kobject_put(&ca->kobj);
+}
+
+/* Global interfaces/init */
+
+static ssize_t register_bcache(struct kobject *, struct kobj_attribute *,
+			       const char *, size_t);
+
+kobj_attribute_write(register,		register_bcache);
+kobj_attribute_write(register_quiet,	register_bcache);
+
+static bool bch_is_open_backing(struct block_device *bdev) {
+	struct cache_set *c, *tc;
+	struct cached_dev *dc, *t;
+
+	list_for_each_entry_safe(c, tc, &bch_cache_sets, list)
+		list_for_each_entry_safe(dc, t, &c->cached_devs, list)
+			if (dc->bdev == bdev)
+				return true;
+	list_for_each_entry_safe(dc, t, &uncached_devices, list)
+		if (dc->bdev == bdev)
+			return true;
+	return false;
+}
+
+static bool bch_is_open_cache(struct block_device *bdev) {
+	struct cache_set *c, *tc;
+	struct cache *ca;
+	unsigned i;
+
+	list_for_each_entry_safe(c, tc, &bch_cache_sets, list)
+		for_each_cache(ca, c, i)
+			if (ca->bdev == bdev)
+				return true;
+	return false;
+}
+
+static bool bch_is_open(struct block_device *bdev) {
+	return bch_is_open_cache(bdev) || bch_is_open_backing(bdev);
+}
+
+static ssize_t register_bcache(struct kobject *k, struct kobj_attribute *attr,
+			       const char *buffer, size_t size)
+{
+	ssize_t ret = size;
+	const char *err = "cannot allocate memory";
+	char *path = NULL;
+	struct cache_sb *sb = NULL;
+	struct block_device *bdev = NULL;
+	struct page *sb_page = NULL;
+
+	if (!try_module_get(THIS_MODULE))
+		return -EBUSY;
+
+	if (!(path = kstrndup(buffer, size, GFP_KERNEL)) ||
+	    !(sb = kmalloc(sizeof(struct cache_sb), GFP_KERNEL)))
+		goto err;
+
+	err = "failed to open device";
+	bdev = blkdev_get_by_path(strim(path),
+				  FMODE_READ|FMODE_WRITE|FMODE_EXCL,
+				  sb);
+	if (IS_ERR(bdev)) {
+		if (bdev == ERR_PTR(-EBUSY)) {
+			bdev = lookup_bdev(strim(path));
+			if (!IS_ERR(bdev) && bch_is_open(bdev))
+				err = "device already registered";
+			else
+				err = "device busy";
+		}
+		goto err;
+	}
+
+	err = "failed to set blocksize";
+	if (set_blocksize(bdev, 4096))
+		goto err_close;
+
+	err = read_super(sb, bdev, &sb_page);
+	if (err)
+		goto err_close;
+
+	if (SB_IS_BDEV(sb)) {
+		struct cached_dev *dc = kzalloc(sizeof(*dc), GFP_KERNEL);
+		if (!dc)
+			goto err_close;
+
+		mutex_lock(&bch_register_lock);
+		register_bdev(sb, sb_page, bdev, dc);
+		mutex_unlock(&bch_register_lock);
+	} else {
+		struct cache *ca = kzalloc(sizeof(*ca), GFP_KERNEL);
+		if (!ca)
+			goto err_close;
+
+		register_cache(sb, sb_page, bdev, ca);
+	}
+out:
+	if (sb_page)
+		put_page(sb_page);
+	kfree(sb);
+	kfree(path);
+	module_put(THIS_MODULE);
+	return ret;
+
+err_close:
+	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
+err:
+	if (attr != &ksysfs_register_quiet)
+		pr_info("error opening %s: %s", path, err);
+	ret = -EINVAL;
+	goto out;
+}
+
+static int bcache_reboot(struct notifier_block *n, unsigned long code, void *x)
+{
+	if (code == SYS_DOWN ||
+	    code == SYS_HALT ||
+	    code == SYS_POWER_OFF) {
+		DEFINE_WAIT(wait);
+		unsigned long start = jiffies;
+		bool stopped = false;
+
+		struct cache_set *c, *tc;
+		struct cached_dev *dc, *tdc;
+
+		mutex_lock(&bch_register_lock);
+
+		if (list_empty(&bch_cache_sets) &&
+		    list_empty(&uncached_devices))
+			goto out;
+
+		pr_info("Stopping all devices:");
+
+		list_for_each_entry_safe(c, tc, &bch_cache_sets, list)
+			bch_cache_set_stop(c);
+
+		list_for_each_entry_safe(dc, tdc, &uncached_devices, list)
+			bcache_device_stop(&dc->disk);
+
+		/* What's a condition variable? */
+		while (1) {
+			long timeout = start + 2 * HZ - jiffies;
+
+			stopped = list_empty(&bch_cache_sets) &&
+				list_empty(&uncached_devices);
+
+			if (timeout < 0 || stopped)
+				break;
+
+			prepare_to_wait(&unregister_wait, &wait,
+					TASK_UNINTERRUPTIBLE);
+
+			mutex_unlock(&bch_register_lock);
+			schedule_timeout(timeout);
+			mutex_lock(&bch_register_lock);
+		}
+
+		finish_wait(&unregister_wait, &wait);
+
+		if (stopped)
+			pr_info("All devices stopped");
+		else
+			pr_notice("Timeout waiting for devices to be closed");
+out:
+		mutex_unlock(&bch_register_lock);
+	}
+
+	return NOTIFY_DONE;
+}
+
+static struct notifier_block reboot = {
+	.notifier_call	= bcache_reboot,
+	.priority	= INT_MAX, /* before any real devices */
+};
+
+static void bcache_exit(void)
+{
+	bch_debug_exit();
+	bch_request_exit();
+	if (bcache_kobj)
+		kobject_put(bcache_kobj);
+	if (bcache_wq)
+		destroy_workqueue(bcache_wq);
+	if (bcache_major)
+		unregister_blkdev(bcache_major, "bcache");
+	unregister_reboot_notifier(&reboot);
+}
+
+static int __init bcache_init(void)
+{
+	static const struct attribute *files[] = {
+		&ksysfs_register.attr,
+		&ksysfs_register_quiet.attr,
+		NULL
+	};
+
+	mutex_init(&bch_register_lock);
+	init_waitqueue_head(&unregister_wait);
+	register_reboot_notifier(&reboot);
+	closure_debug_init();
+
+	bcache_major = register_blkdev(0, "bcache");
+	if (bcache_major < 0)
+		return bcache_major;
+
+	if (!(bcache_wq = create_workqueue("bcache")) ||
+	    !(bcache_kobj = kobject_create_and_add("bcache", fs_kobj)) ||
+	    sysfs_create_files(bcache_kobj, files) ||
+	    bch_request_init() ||
+	    bch_debug_init(bcache_kobj))
+		goto err;
+
+	return 0;
+err:
+	bcache_exit();
+	return -ENOMEM;
+}
+
+module_exit(bcache_exit);
+module_init(bcache_init);
diff --git a/drivers/md/bcache/sysfs.c b/drivers/md/bcache/sysfs.c
new file mode 100644
index 00000000000..b3ff57d61dd
--- /dev/null
+++ b/drivers/md/bcache/sysfs.c
@@ -0,0 +1,898 @@
+/*
+ * bcache sysfs interfaces
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "sysfs.h"
+#include "btree.h"
+#include "request.h"
+#include "writeback.h"
+
+#include <linux/blkdev.h>
+#include <linux/sort.h>
+
+static const char * const cache_replacement_policies[] = {
+	"lru",
+	"fifo",
+	"random",
+	NULL
+};
+
+static const char * const error_actions[] = {
+	"unregister",
+	"panic",
+	NULL
+};
+
+write_attribute(attach);
+write_attribute(detach);
+write_attribute(unregister);
+write_attribute(stop);
+write_attribute(clear_stats);
+write_attribute(trigger_gc);
+write_attribute(prune_cache);
+write_attribute(flash_vol_create);
+
+read_attribute(bucket_size);
+read_attribute(block_size);
+read_attribute(nbuckets);
+read_attribute(tree_depth);
+read_attribute(root_usage_percent);
+read_attribute(priority_stats);
+read_attribute(btree_cache_size);
+read_attribute(btree_cache_max_chain);
+read_attribute(cache_available_percent);
+read_attribute(written);
+read_attribute(btree_written);
+read_attribute(metadata_written);
+read_attribute(active_journal_entries);
+
+sysfs_time_stats_attribute(btree_gc,	sec, ms);
+sysfs_time_stats_attribute(btree_split, sec, us);
+sysfs_time_stats_attribute(btree_sort,	ms,  us);
+sysfs_time_stats_attribute(btree_read,	ms,  us);
+
+read_attribute(btree_nodes);
+read_attribute(btree_used_percent);
+read_attribute(average_key_size);
+read_attribute(dirty_data);
+read_attribute(bset_tree_stats);
+
+read_attribute(state);
+read_attribute(cache_read_races);
+read_attribute(writeback_keys_done);
+read_attribute(writeback_keys_failed);
+read_attribute(io_errors);
+read_attribute(congested);
+rw_attribute(congested_read_threshold_us);
+rw_attribute(congested_write_threshold_us);
+
+rw_attribute(sequential_cutoff);
+rw_attribute(data_csum);
+rw_attribute(cache_mode);
+rw_attribute(writeback_metadata);
+rw_attribute(writeback_running);
+rw_attribute(writeback_percent);
+rw_attribute(writeback_delay);
+rw_attribute(writeback_rate);
+
+rw_attribute(writeback_rate_update_seconds);
+rw_attribute(writeback_rate_d_term);
+rw_attribute(writeback_rate_p_term_inverse);
+read_attribute(writeback_rate_debug);
+
+read_attribute(stripe_size);
+read_attribute(partial_stripes_expensive);
+
+rw_attribute(synchronous);
+rw_attribute(journal_delay_ms);
+rw_attribute(discard);
+rw_attribute(running);
+rw_attribute(label);
+rw_attribute(readahead);
+rw_attribute(errors);
+rw_attribute(io_error_limit);
+rw_attribute(io_error_halflife);
+rw_attribute(verify);
+rw_attribute(bypass_torture_test);
+rw_attribute(key_merging_disabled);
+rw_attribute(gc_always_rewrite);
+rw_attribute(expensive_debug_checks);
+rw_attribute(cache_replacement_policy);
+rw_attribute(btree_shrinker_disabled);
+rw_attribute(copy_gc_enabled);
+rw_attribute(size);
+
+SHOW(__bch_cached_dev)
+{
+	struct cached_dev *dc = container_of(kobj, struct cached_dev,
+					     disk.kobj);
+	const char *states[] = { "no cache", "clean", "dirty", "inconsistent" };
+
+#define var(stat)		(dc->stat)
+
+	if (attr == &sysfs_cache_mode)
+		return bch_snprint_string_list(buf, PAGE_SIZE,
+					       bch_cache_modes + 1,
+					       BDEV_CACHE_MODE(&dc->sb));
+
+	sysfs_printf(data_csum,		"%i", dc->disk.data_csum);
+	var_printf(verify,		"%i");
+	var_printf(bypass_torture_test,	"%i");
+	var_printf(writeback_metadata,	"%i");
+	var_printf(writeback_running,	"%i");
+	var_print(writeback_delay);
+	var_print(writeback_percent);
+	sysfs_hprint(writeback_rate,	dc->writeback_rate.rate << 9);
+
+	var_print(writeback_rate_update_seconds);
+	var_print(writeback_rate_d_term);
+	var_print(writeback_rate_p_term_inverse);
+
+	if (attr == &sysfs_writeback_rate_debug) {
+		char rate[20];
+		char dirty[20];
+		char target[20];
+		char proportional[20];
+		char derivative[20];
+		char change[20];
+		s64 next_io;
+
+		bch_hprint(rate,	dc->writeback_rate.rate << 9);
+		bch_hprint(dirty,	bcache_dev_sectors_dirty(&dc->disk) << 9);
+		bch_hprint(target,	dc->writeback_rate_target << 9);
+		bch_hprint(proportional,dc->writeback_rate_proportional << 9);
+		bch_hprint(derivative,	dc->writeback_rate_derivative << 9);
+		bch_hprint(change,	dc->writeback_rate_change << 9);
+
+		next_io = div64_s64(dc->writeback_rate.next - local_clock(),
+				    NSEC_PER_MSEC);
+
+		return sprintf(buf,
+			       "rate:\t\t%s/sec\n"
+			       "dirty:\t\t%s\n"
+			       "target:\t\t%s\n"
+			       "proportional:\t%s\n"
+			       "derivative:\t%s\n"
+			       "change:\t\t%s/sec\n"
+			       "next io:\t%llims\n",
+			       rate, dirty, target, proportional,
+			       derivative, change, next_io);
+	}
+
+	sysfs_hprint(dirty_data,
+		     bcache_dev_sectors_dirty(&dc->disk) << 9);
+
+	sysfs_hprint(stripe_size,	dc->disk.stripe_size << 9);
+	var_printf(partial_stripes_expensive,	"%u");
+
+	var_hprint(sequential_cutoff);
+	var_hprint(readahead);
+
+	sysfs_print(running,		atomic_read(&dc->running));
+	sysfs_print(state,		states[BDEV_STATE(&dc->sb)]);
+
+	if (attr == &sysfs_label) {
+		memcpy(buf, dc->sb.label, SB_LABEL_SIZE);
+		buf[SB_LABEL_SIZE + 1] = '\0';
+		strcat(buf, "\n");
+		return strlen(buf);
+	}
+
+#undef var
+	return 0;
+}
+SHOW_LOCKED(bch_cached_dev)
+
+STORE(__cached_dev)
+{
+	struct cached_dev *dc = container_of(kobj, struct cached_dev,
+					     disk.kobj);
+	unsigned v = size;
+	struct cache_set *c;
+	struct kobj_uevent_env *env;
+
+#define d_strtoul(var)		sysfs_strtoul(var, dc->var)
+#define d_strtoul_nonzero(var)	sysfs_strtoul_clamp(var, dc->var, 1, INT_MAX)
+#define d_strtoi_h(var)		sysfs_hatoi(var, dc->var)
+
+	sysfs_strtoul(data_csum,	dc->disk.data_csum);
+	d_strtoul(verify);
+	d_strtoul(bypass_torture_test);
+	d_strtoul(writeback_metadata);
+	d_strtoul(writeback_running);
+	d_strtoul(writeback_delay);
+
+	sysfs_strtoul_clamp(writeback_percent, dc->writeback_percent, 0, 40);
+
+	sysfs_strtoul_clamp(writeback_rate,
+			    dc->writeback_rate.rate, 1, INT_MAX);
+
+	d_strtoul_nonzero(writeback_rate_update_seconds);
+	d_strtoul(writeback_rate_d_term);
+	d_strtoul_nonzero(writeback_rate_p_term_inverse);
+
+	d_strtoi_h(sequential_cutoff);
+	d_strtoi_h(readahead);
+
+	if (attr == &sysfs_clear_stats)
+		bch_cache_accounting_clear(&dc->accounting);
+
+	if (attr == &sysfs_running &&
+	    strtoul_or_return(buf))
+		bch_cached_dev_run(dc);
+
+	if (attr == &sysfs_cache_mode) {
+		ssize_t v = bch_read_string_list(buf, bch_cache_modes + 1);
+
+		if (v < 0)
+			return v;
+
+		if ((unsigned) v != BDEV_CACHE_MODE(&dc->sb)) {
+			SET_BDEV_CACHE_MODE(&dc->sb, v);
+			bch_write_bdev_super(dc, NULL);
+		}
+	}
+
+	if (attr == &sysfs_label) {
+		if (size > SB_LABEL_SIZE)
+			return -EINVAL;
+		memcpy(dc->sb.label, buf, size);
+		if (size < SB_LABEL_SIZE)
+			dc->sb.label[size] = '\0';
+		if (size && dc->sb.label[size - 1] == '\n')
+			dc->sb.label[size - 1] = '\0';
+		bch_write_bdev_super(dc, NULL);
+		if (dc->disk.c) {
+			memcpy(dc->disk.c->uuids[dc->disk.id].label,
+			       buf, SB_LABEL_SIZE);
+			bch_uuid_write(dc->disk.c);
+		}
+		env = kzalloc(sizeof(struct kobj_uevent_env), GFP_KERNEL);
+		if (!env)
+			return -ENOMEM;
+		add_uevent_var(env, "DRIVER=bcache");
+		add_uevent_var(env, "CACHED_UUID=%pU", dc->sb.uuid),
+		add_uevent_var(env, "CACHED_LABEL=%s", buf);
+		kobject_uevent_env(
+			&disk_to_dev(dc->disk.disk)->kobj, KOBJ_CHANGE, env->envp);
+		kfree(env);
+	}
+
+	if (attr == &sysfs_attach) {
+		if (bch_parse_uuid(buf, dc->sb.set_uuid) < 16)
+			return -EINVAL;
+
+		list_for_each_entry(c, &bch_cache_sets, list) {
+			v = bch_cached_dev_attach(dc, c);
+			if (!v)
+				return size;
+		}
+
+		pr_err("Can't attach %s: cache set not found", buf);
+		size = v;
+	}
+
+	if (attr == &sysfs_detach && dc->disk.c)
+		bch_cached_dev_detach(dc);
+
+	if (attr == &sysfs_stop)
+		bcache_device_stop(&dc->disk);
+
+	return size;
+}
+
+STORE(bch_cached_dev)
+{
+	struct cached_dev *dc = container_of(kobj, struct cached_dev,
+					     disk.kobj);
+
+	mutex_lock(&bch_register_lock);
+	size = __cached_dev_store(kobj, attr, buf, size);
+
+	if (attr == &sysfs_writeback_running)
+		bch_writeback_queue(dc);
+
+	if (attr == &sysfs_writeback_percent)
+		schedule_delayed_work(&dc->writeback_rate_update,
+				      dc->writeback_rate_update_seconds * HZ);
+
+	mutex_unlock(&bch_register_lock);
+	return size;
+}
+
+static struct attribute *bch_cached_dev_files[] = {
+	&sysfs_attach,
+	&sysfs_detach,
+	&sysfs_stop,
+#if 0
+	&sysfs_data_csum,
+#endif
+	&sysfs_cache_mode,
+	&sysfs_writeback_metadata,
+	&sysfs_writeback_running,
+	&sysfs_writeback_delay,
+	&sysfs_writeback_percent,
+	&sysfs_writeback_rate,
+	&sysfs_writeback_rate_update_seconds,
+	&sysfs_writeback_rate_d_term,
+	&sysfs_writeback_rate_p_term_inverse,
+	&sysfs_writeback_rate_debug,
+	&sysfs_dirty_data,
+	&sysfs_stripe_size,
+	&sysfs_partial_stripes_expensive,
+	&sysfs_sequential_cutoff,
+	&sysfs_clear_stats,
+	&sysfs_running,
+	&sysfs_state,
+	&sysfs_label,
+	&sysfs_readahead,
+#ifdef CONFIG_BCACHE_DEBUG
+	&sysfs_verify,
+	&sysfs_bypass_torture_test,
+#endif
+	NULL
+};
+KTYPE(bch_cached_dev);
+
+SHOW(bch_flash_dev)
+{
+	struct bcache_device *d = container_of(kobj, struct bcache_device,
+					       kobj);
+	struct uuid_entry *u = &d->c->uuids[d->id];
+
+	sysfs_printf(data_csum,	"%i", d->data_csum);
+	sysfs_hprint(size,	u->sectors << 9);
+
+	if (attr == &sysfs_label) {
+		memcpy(buf, u->label, SB_LABEL_SIZE);
+		buf[SB_LABEL_SIZE + 1] = '\0';
+		strcat(buf, "\n");
+		return strlen(buf);
+	}
+
+	return 0;
+}
+
+STORE(__bch_flash_dev)
+{
+	struct bcache_device *d = container_of(kobj, struct bcache_device,
+					       kobj);
+	struct uuid_entry *u = &d->c->uuids[d->id];
+
+	sysfs_strtoul(data_csum,	d->data_csum);
+
+	if (attr == &sysfs_size) {
+		uint64_t v;
+		strtoi_h_or_return(buf, v);
+
+		u->sectors = v >> 9;
+		bch_uuid_write(d->c);
+		set_capacity(d->disk, u->sectors);
+	}
+
+	if (attr == &sysfs_label) {
+		memcpy(u->label, buf, SB_LABEL_SIZE);
+		bch_uuid_write(d->c);
+	}
+
+	if (attr == &sysfs_unregister) {
+		set_bit(BCACHE_DEV_DETACHING, &d->flags);
+		bcache_device_stop(d);
+	}
+
+	return size;
+}
+STORE_LOCKED(bch_flash_dev)
+
+static struct attribute *bch_flash_dev_files[] = {
+	&sysfs_unregister,
+#if 0
+	&sysfs_data_csum,
+#endif
+	&sysfs_label,
+	&sysfs_size,
+	NULL
+};
+KTYPE(bch_flash_dev);
+
+struct bset_stats_op {
+	struct btree_op op;
+	size_t nodes;
+	struct bset_stats stats;
+};
+
+static int bch_btree_bset_stats(struct btree_op *b_op, struct btree *b)
+{
+	struct bset_stats_op *op = container_of(b_op, struct bset_stats_op, op);
+
+	op->nodes++;
+	bch_btree_keys_stats(&b->keys, &op->stats);
+
+	return MAP_CONTINUE;
+}
+
+static int bch_bset_print_stats(struct cache_set *c, char *buf)
+{
+	struct bset_stats_op op;
+	int ret;
+
+	memset(&op, 0, sizeof(op));
+	bch_btree_op_init(&op.op, -1);
+
+	ret = bch_btree_map_nodes(&op.op, c, &ZERO_KEY, bch_btree_bset_stats);
+	if (ret < 0)
+		return ret;
+
+	return snprintf(buf, PAGE_SIZE,
+			"btree nodes:		%zu\n"
+			"written sets:		%zu\n"
+			"unwritten sets:		%zu\n"
+			"written key bytes:	%zu\n"
+			"unwritten key bytes:	%zu\n"
+			"floats:			%zu\n"
+			"failed:			%zu\n",
+			op.nodes,
+			op.stats.sets_written, op.stats.sets_unwritten,
+			op.stats.bytes_written, op.stats.bytes_unwritten,
+			op.stats.floats, op.stats.failed);
+}
+
+static unsigned bch_root_usage(struct cache_set *c)
+{
+	unsigned bytes = 0;
+	struct bkey *k;
+	struct btree *b;
+	struct btree_iter iter;
+
+	goto lock_root;
+
+	do {
+		rw_unlock(false, b);
+lock_root:
+		b = c->root;
+		rw_lock(false, b, b->level);
+	} while (b != c->root);
+
+	for_each_key_filter(&b->keys, k, &iter, bch_ptr_bad)
+		bytes += bkey_bytes(k);
+
+	rw_unlock(false, b);
+
+	return (bytes * 100) / btree_bytes(c);
+}
+
+static size_t bch_cache_size(struct cache_set *c)
+{
+	size_t ret = 0;
+	struct btree *b;
+
+	mutex_lock(&c->bucket_lock);
+	list_for_each_entry(b, &c->btree_cache, list)
+		ret += 1 << (b->keys.page_order + PAGE_SHIFT);
+
+	mutex_unlock(&c->bucket_lock);
+	return ret;
+}
+
+static unsigned bch_cache_max_chain(struct cache_set *c)
+{
+	unsigned ret = 0;
+	struct hlist_head *h;
+
+	mutex_lock(&c->bucket_lock);
+
+	for (h = c->bucket_hash;
+	     h < c->bucket_hash + (1 << BUCKET_HASH_BITS);
+	     h++) {
+		unsigned i = 0;
+		struct hlist_node *p;
+
+		hlist_for_each(p, h)
+			i++;
+
+		ret = max(ret, i);
+	}
+
+	mutex_unlock(&c->bucket_lock);
+	return ret;
+}
+
+static unsigned bch_btree_used(struct cache_set *c)
+{
+	return div64_u64(c->gc_stats.key_bytes * 100,
+			 (c->gc_stats.nodes ?: 1) * btree_bytes(c));
+}
+
+static unsigned bch_average_key_size(struct cache_set *c)
+{
+	return c->gc_stats.nkeys
+		? div64_u64(c->gc_stats.data, c->gc_stats.nkeys)
+		: 0;
+}
+
+SHOW(__bch_cache_set)
+{
+	struct cache_set *c = container_of(kobj, struct cache_set, kobj);
+
+	sysfs_print(synchronous,		CACHE_SYNC(&c->sb));
+	sysfs_print(journal_delay_ms,		c->journal_delay_ms);
+	sysfs_hprint(bucket_size,		bucket_bytes(c));
+	sysfs_hprint(block_size,		block_bytes(c));
+	sysfs_print(tree_depth,			c->root->level);
+	sysfs_print(root_usage_percent,		bch_root_usage(c));
+
+	sysfs_hprint(btree_cache_size,		bch_cache_size(c));
+	sysfs_print(btree_cache_max_chain,	bch_cache_max_chain(c));
+	sysfs_print(cache_available_percent,	100 - c->gc_stats.in_use);
+
+	sysfs_print_time_stats(&c->btree_gc_time,	btree_gc, sec, ms);
+	sysfs_print_time_stats(&c->btree_split_time,	btree_split, sec, us);
+	sysfs_print_time_stats(&c->sort.time,		btree_sort, ms, us);
+	sysfs_print_time_stats(&c->btree_read_time,	btree_read, ms, us);
+
+	sysfs_print(btree_used_percent,	bch_btree_used(c));
+	sysfs_print(btree_nodes,	c->gc_stats.nodes);
+	sysfs_hprint(average_key_size,	bch_average_key_size(c));
+
+	sysfs_print(cache_read_races,
+		    atomic_long_read(&c->cache_read_races));
+
+	sysfs_print(writeback_keys_done,
+		    atomic_long_read(&c->writeback_keys_done));
+	sysfs_print(writeback_keys_failed,
+		    atomic_long_read(&c->writeback_keys_failed));
+
+	if (attr == &sysfs_errors)
+		return bch_snprint_string_list(buf, PAGE_SIZE, error_actions,
+					       c->on_error);
+
+	/* See count_io_errors for why 88 */
+	sysfs_print(io_error_halflife,	c->error_decay * 88);
+	sysfs_print(io_error_limit,	c->error_limit >> IO_ERROR_SHIFT);
+
+	sysfs_hprint(congested,
+		     ((uint64_t) bch_get_congested(c)) << 9);
+	sysfs_print(congested_read_threshold_us,
+		    c->congested_read_threshold_us);
+	sysfs_print(congested_write_threshold_us,
+		    c->congested_write_threshold_us);
+
+	sysfs_print(active_journal_entries,	fifo_used(&c->journal.pin));
+	sysfs_printf(verify,			"%i", c->verify);
+	sysfs_printf(key_merging_disabled,	"%i", c->key_merging_disabled);
+	sysfs_printf(expensive_debug_checks,
+		     "%i", c->expensive_debug_checks);
+	sysfs_printf(gc_always_rewrite,		"%i", c->gc_always_rewrite);
+	sysfs_printf(btree_shrinker_disabled,	"%i", c->shrinker_disabled);
+	sysfs_printf(copy_gc_enabled,		"%i", c->copy_gc_enabled);
+
+	if (attr == &sysfs_bset_tree_stats)
+		return bch_bset_print_stats(c, buf);
+
+	return 0;
+}
+SHOW_LOCKED(bch_cache_set)
+
+STORE(__bch_cache_set)
+{
+	struct cache_set *c = container_of(kobj, struct cache_set, kobj);
+
+	if (attr == &sysfs_unregister)
+		bch_cache_set_unregister(c);
+
+	if (attr == &sysfs_stop)
+		bch_cache_set_stop(c);
+
+	if (attr == &sysfs_synchronous) {
+		bool sync = strtoul_or_return(buf);
+
+		if (sync != CACHE_SYNC(&c->sb)) {
+			SET_CACHE_SYNC(&c->sb, sync);
+			bcache_write_super(c);
+		}
+	}
+
+	if (attr == &sysfs_flash_vol_create) {
+		int r;
+		uint64_t v;
+		strtoi_h_or_return(buf, v);
+
+		r = bch_flash_dev_create(c, v);
+		if (r)
+			return r;
+	}
+
+	if (attr == &sysfs_clear_stats) {
+		atomic_long_set(&c->writeback_keys_done,	0);
+		atomic_long_set(&c->writeback_keys_failed,	0);
+
+		memset(&c->gc_stats, 0, sizeof(struct gc_stat));
+		bch_cache_accounting_clear(&c->accounting);
+	}
+
+	if (attr == &sysfs_trigger_gc)
+		wake_up_gc(c);
+
+	if (attr == &sysfs_prune_cache) {
+		struct shrink_control sc;
+		sc.gfp_mask = GFP_KERNEL;
+		sc.nr_to_scan = strtoul_or_return(buf);
+		c->shrink.scan_objects(&c->shrink, &sc);
+	}
+
+	sysfs_strtoul(congested_read_threshold_us,
+		      c->congested_read_threshold_us);
+	sysfs_strtoul(congested_write_threshold_us,
+		      c->congested_write_threshold_us);
+
+	if (attr == &sysfs_errors) {
+		ssize_t v = bch_read_string_list(buf, error_actions);
+
+		if (v < 0)
+			return v;
+
+		c->on_error = v;
+	}
+
+	if (attr == &sysfs_io_error_limit)
+		c->error_limit = strtoul_or_return(buf) << IO_ERROR_SHIFT;
+
+	/* See count_io_errors() for why 88 */
+	if (attr == &sysfs_io_error_halflife)
+		c->error_decay = strtoul_or_return(buf) / 88;
+
+	sysfs_strtoul(journal_delay_ms,		c->journal_delay_ms);
+	sysfs_strtoul(verify,			c->verify);
+	sysfs_strtoul(key_merging_disabled,	c->key_merging_disabled);
+	sysfs_strtoul(expensive_debug_checks,	c->expensive_debug_checks);
+	sysfs_strtoul(gc_always_rewrite,	c->gc_always_rewrite);
+	sysfs_strtoul(btree_shrinker_disabled,	c->shrinker_disabled);
+	sysfs_strtoul(copy_gc_enabled,		c->copy_gc_enabled);
+
+	return size;
+}
+STORE_LOCKED(bch_cache_set)
+
+SHOW(bch_cache_set_internal)
+{
+	struct cache_set *c = container_of(kobj, struct cache_set, internal);
+	return bch_cache_set_show(&c->kobj, attr, buf);
+}
+
+STORE(bch_cache_set_internal)
+{
+	struct cache_set *c = container_of(kobj, struct cache_set, internal);
+	return bch_cache_set_store(&c->kobj, attr, buf, size);
+}
+
+static void bch_cache_set_internal_release(struct kobject *k)
+{
+}
+
+static struct attribute *bch_cache_set_files[] = {
+	&sysfs_unregister,
+	&sysfs_stop,
+	&sysfs_synchronous,
+	&sysfs_journal_delay_ms,
+	&sysfs_flash_vol_create,
+
+	&sysfs_bucket_size,
+	&sysfs_block_size,
+	&sysfs_tree_depth,
+	&sysfs_root_usage_percent,
+	&sysfs_btree_cache_size,
+	&sysfs_cache_available_percent,
+
+	&sysfs_average_key_size,
+
+	&sysfs_errors,
+	&sysfs_io_error_limit,
+	&sysfs_io_error_halflife,
+	&sysfs_congested,
+	&sysfs_congested_read_threshold_us,
+	&sysfs_congested_write_threshold_us,
+	&sysfs_clear_stats,
+	NULL
+};
+KTYPE(bch_cache_set);
+
+static struct attribute *bch_cache_set_internal_files[] = {
+	&sysfs_active_journal_entries,
+
+	sysfs_time_stats_attribute_list(btree_gc, sec, ms)
+	sysfs_time_stats_attribute_list(btree_split, sec, us)
+	sysfs_time_stats_attribute_list(btree_sort, ms, us)
+	sysfs_time_stats_attribute_list(btree_read, ms, us)
+
+	&sysfs_btree_nodes,
+	&sysfs_btree_used_percent,
+	&sysfs_btree_cache_max_chain,
+
+	&sysfs_bset_tree_stats,
+	&sysfs_cache_read_races,
+	&sysfs_writeback_keys_done,
+	&sysfs_writeback_keys_failed,
+
+	&sysfs_trigger_gc,
+	&sysfs_prune_cache,
+#ifdef CONFIG_BCACHE_DEBUG
+	&sysfs_verify,
+	&sysfs_key_merging_disabled,
+	&sysfs_expensive_debug_checks,
+#endif
+	&sysfs_gc_always_rewrite,
+	&sysfs_btree_shrinker_disabled,
+	&sysfs_copy_gc_enabled,
+	NULL
+};
+KTYPE(bch_cache_set_internal);
+
+SHOW(__bch_cache)
+{
+	struct cache *ca = container_of(kobj, struct cache, kobj);
+
+	sysfs_hprint(bucket_size,	bucket_bytes(ca));
+	sysfs_hprint(block_size,	block_bytes(ca));
+	sysfs_print(nbuckets,		ca->sb.nbuckets);
+	sysfs_print(discard,		ca->discard);
+	sysfs_hprint(written, atomic_long_read(&ca->sectors_written) << 9);
+	sysfs_hprint(btree_written,
+		     atomic_long_read(&ca->btree_sectors_written) << 9);
+	sysfs_hprint(metadata_written,
+		     (atomic_long_read(&ca->meta_sectors_written) +
+		      atomic_long_read(&ca->btree_sectors_written)) << 9);
+
+	sysfs_print(io_errors,
+		    atomic_read(&ca->io_errors) >> IO_ERROR_SHIFT);
+
+	if (attr == &sysfs_cache_replacement_policy)
+		return bch_snprint_string_list(buf, PAGE_SIZE,
+					       cache_replacement_policies,
+					       CACHE_REPLACEMENT(&ca->sb));
+
+	if (attr == &sysfs_priority_stats) {
+		int cmp(const void *l, const void *r)
+		{	return *((uint16_t *) r) - *((uint16_t *) l); }
+
+		struct bucket *b;
+		size_t n = ca->sb.nbuckets, i;
+		size_t unused = 0, available = 0, dirty = 0, meta = 0;
+		uint64_t sum = 0;
+		/* Compute 31 quantiles */
+		uint16_t q[31], *p, *cached;
+		ssize_t ret;
+
+		cached = p = vmalloc(ca->sb.nbuckets * sizeof(uint16_t));
+		if (!p)
+			return -ENOMEM;
+
+		mutex_lock(&ca->set->bucket_lock);
+		for_each_bucket(b, ca) {
+			if (!GC_SECTORS_USED(b))
+				unused++;
+			if (GC_MARK(b) == GC_MARK_RECLAIMABLE)
+				available++;
+			if (GC_MARK(b) == GC_MARK_DIRTY)
+				dirty++;
+			if (GC_MARK(b) == GC_MARK_METADATA)
+				meta++;
+		}
+
+		for (i = ca->sb.first_bucket; i < n; i++)
+			p[i] = ca->buckets[i].prio;
+		mutex_unlock(&ca->set->bucket_lock);
+
+		sort(p, n, sizeof(uint16_t), cmp, NULL);
+
+		while (n &&
+		       !cached[n - 1])
+			--n;
+
+		unused = ca->sb.nbuckets - n;
+
+		while (cached < p + n &&
+		       *cached == BTREE_PRIO)
+			cached++, n--;
+
+		for (i = 0; i < n; i++)
+			sum += INITIAL_PRIO - cached[i];
+
+		if (n)
+			do_div(sum, n);
+
+		for (i = 0; i < ARRAY_SIZE(q); i++)
+			q[i] = INITIAL_PRIO - cached[n * (i + 1) /
+				(ARRAY_SIZE(q) + 1)];
+
+		vfree(p);
+
+		ret = scnprintf(buf, PAGE_SIZE,
+				"Unused:		%zu%%\n"
+				"Clean:		%zu%%\n"
+				"Dirty:		%zu%%\n"
+				"Metadata:	%zu%%\n"
+				"Average:	%llu\n"
+				"Sectors per Q:	%zu\n"
+				"Quantiles:	[",
+				unused * 100 / (size_t) ca->sb.nbuckets,
+				available * 100 / (size_t) ca->sb.nbuckets,
+				dirty * 100 / (size_t) ca->sb.nbuckets,
+				meta * 100 / (size_t) ca->sb.nbuckets, sum,
+				n * ca->sb.bucket_size / (ARRAY_SIZE(q) + 1));
+
+		for (i = 0; i < ARRAY_SIZE(q); i++)
+			ret += scnprintf(buf + ret, PAGE_SIZE - ret,
+					 "%u ", q[i]);
+		ret--;
+
+		ret += scnprintf(buf + ret, PAGE_SIZE - ret, "]\n");
+
+		return ret;
+	}
+
+	return 0;
+}
+SHOW_LOCKED(bch_cache)
+
+STORE(__bch_cache)
+{
+	struct cache *ca = container_of(kobj, struct cache, kobj);
+
+	if (attr == &sysfs_discard) {
+		bool v = strtoul_or_return(buf);
+
+		if (blk_queue_discard(bdev_get_queue(ca->bdev)))
+			ca->discard = v;
+
+		if (v != CACHE_DISCARD(&ca->sb)) {
+			SET_CACHE_DISCARD(&ca->sb, v);
+			bcache_write_super(ca->set);
+		}
+	}
+
+	if (attr == &sysfs_cache_replacement_policy) {
+		ssize_t v = bch_read_string_list(buf, cache_replacement_policies);
+
+		if (v < 0)
+			return v;
+
+		if ((unsigned) v != CACHE_REPLACEMENT(&ca->sb)) {
+			mutex_lock(&ca->set->bucket_lock);
+			SET_CACHE_REPLACEMENT(&ca->sb, v);
+			mutex_unlock(&ca->set->bucket_lock);
+
+			bcache_write_super(ca->set);
+		}
+	}
+
+	if (attr == &sysfs_clear_stats) {
+		atomic_long_set(&ca->sectors_written, 0);
+		atomic_long_set(&ca->btree_sectors_written, 0);
+		atomic_long_set(&ca->meta_sectors_written, 0);
+		atomic_set(&ca->io_count, 0);
+		atomic_set(&ca->io_errors, 0);
+	}
+
+	return size;
+}
+STORE_LOCKED(bch_cache)
+
+static struct attribute *bch_cache_files[] = {
+	&sysfs_bucket_size,
+	&sysfs_block_size,
+	&sysfs_nbuckets,
+	&sysfs_priority_stats,
+	&sysfs_discard,
+	&sysfs_written,
+	&sysfs_btree_written,
+	&sysfs_metadata_written,
+	&sysfs_io_errors,
+	&sysfs_clear_stats,
+	&sysfs_cache_replacement_policy,
+	NULL
+};
+KTYPE(bch_cache);
diff --git a/drivers/md/bcache/sysfs.h b/drivers/md/bcache/sysfs.h
new file mode 100644
index 00000000000..0526fe92a68
--- /dev/null
+++ b/drivers/md/bcache/sysfs.h
@@ -0,0 +1,110 @@
+#ifndef _BCACHE_SYSFS_H_
+#define _BCACHE_SYSFS_H_
+
+#define KTYPE(type)							\
+struct kobj_type type ## _ktype = {					\
+	.release	= type ## _release,				\
+	.sysfs_ops	= &((const struct sysfs_ops) {			\
+		.show	= type ## _show,				\
+		.store	= type ## _store				\
+	}),								\
+	.default_attrs	= type ## _files				\
+}
+
+#define SHOW(fn)							\
+static ssize_t fn ## _show(struct kobject *kobj, struct attribute *attr,\
+			   char *buf)					\
+
+#define STORE(fn)							\
+static ssize_t fn ## _store(struct kobject *kobj, struct attribute *attr,\
+			    const char *buf, size_t size)		\
+
+#define SHOW_LOCKED(fn)							\
+SHOW(fn)								\
+{									\
+	ssize_t ret;							\
+	mutex_lock(&bch_register_lock);					\
+	ret = __ ## fn ## _show(kobj, attr, buf);			\
+	mutex_unlock(&bch_register_lock);				\
+	return ret;							\
+}
+
+#define STORE_LOCKED(fn)						\
+STORE(fn)								\
+{									\
+	ssize_t ret;							\
+	mutex_lock(&bch_register_lock);					\
+	ret = __ ## fn ## _store(kobj, attr, buf, size);		\
+	mutex_unlock(&bch_register_lock);				\
+	return ret;							\
+}
+
+#define __sysfs_attribute(_name, _mode)					\
+	static struct attribute sysfs_##_name =				\
+		{ .name = #_name, .mode = _mode }
+
+#define write_attribute(n)	__sysfs_attribute(n, S_IWUSR)
+#define read_attribute(n)	__sysfs_attribute(n, S_IRUGO)
+#define rw_attribute(n)		__sysfs_attribute(n, S_IRUGO|S_IWUSR)
+
+#define sysfs_printf(file, fmt, ...)					\
+do {									\
+	if (attr == &sysfs_ ## file)					\
+		return snprintf(buf, PAGE_SIZE, fmt "\n", __VA_ARGS__);	\
+} while (0)
+
+#define sysfs_print(file, var)						\
+do {									\
+	if (attr == &sysfs_ ## file)					\
+		return snprint(buf, PAGE_SIZE, var);			\
+} while (0)
+
+#define sysfs_hprint(file, val)						\
+do {									\
+	if (attr == &sysfs_ ## file) {					\
+		ssize_t ret = bch_hprint(buf, val);			\
+		strcat(buf, "\n");					\
+		return ret + 1;						\
+	}								\
+} while (0)
+
+#define var_printf(_var, fmt)	sysfs_printf(_var, fmt, var(_var))
+#define var_print(_var)		sysfs_print(_var, var(_var))
+#define var_hprint(_var)	sysfs_hprint(_var, var(_var))
+
+#define sysfs_strtoul(file, var)					\
+do {									\
+	if (attr == &sysfs_ ## file)					\
+		return strtoul_safe(buf, var) ?: (ssize_t) size;	\
+} while (0)
+
+#define sysfs_strtoul_clamp(file, var, min, max)			\
+do {									\
+	if (attr == &sysfs_ ## file)					\
+		return strtoul_safe_clamp(buf, var, min, max)		\
+			?: (ssize_t) size;				\
+} while (0)
+
+#define strtoul_or_return(cp)						\
+({									\
+	unsigned long _v;						\
+	int _r = kstrtoul(cp, 10, &_v);					\
+	if (_r)								\
+		return _r;						\
+	_v;								\
+})
+
+#define strtoi_h_or_return(cp, v)					\
+do {									\
+	int _r = strtoi_h(cp, &v);					\
+	if (_r)								\
+		return _r;						\
+} while (0)
+
+#define sysfs_hatoi(file, var)						\
+do {									\
+	if (attr == &sysfs_ ## file)					\
+		return strtoi_h(buf, &var) ?: (ssize_t) size;		\
+} while (0)
+
+#endif  /* _BCACHE_SYSFS_H_ */
diff --git a/drivers/md/bcache/trace.c b/drivers/md/bcache/trace.c
new file mode 100644
index 00000000000..b7820b0d262
--- /dev/null
+++ b/drivers/md/bcache/trace.c
@@ -0,0 +1,52 @@
+#include "bcache.h"
+#include "btree.h"
+
+#include <linux/blktrace_api.h>
+#include <linux/module.h>
+
+#define CREATE_TRACE_POINTS
+#include <trace/events/bcache.h>
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_request_start);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_request_end);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_bypass_sequential);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_bypass_congested);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_read);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_write);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_read_retry);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_cache_insert);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_journal_replay_key);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_journal_write);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_journal_full);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_journal_entry_full);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_cache_cannibalize);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_read);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_write);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_node_alloc);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_node_alloc_fail);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_node_free);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_gc_coalesce);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_gc_start);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_gc_end);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_gc_copy);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_gc_copy_collision);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_insert_key);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_node_split);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_node_compact);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_btree_set_root);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_invalidate);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_alloc_fail);
+
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_writeback);
+EXPORT_TRACEPOINT_SYMBOL_GPL(bcache_writeback_collision);
diff --git a/drivers/md/bcache/util.c b/drivers/md/bcache/util.c
new file mode 100644
index 00000000000..db3ae4c2b22
--- /dev/null
+++ b/drivers/md/bcache/util.c
@@ -0,0 +1,381 @@
+/*
+ * random utiility code, for bcache but in theory not specific to bcache
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include <linux/bio.h>
+#include <linux/blkdev.h>
+#include <linux/ctype.h>
+#include <linux/debugfs.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include <linux/types.h>
+
+#include "util.h"
+
+#define simple_strtoint(c, end, base)	simple_strtol(c, end, base)
+#define simple_strtouint(c, end, base)	simple_strtoul(c, end, base)
+
+#define STRTO_H(name, type)					\
+int bch_ ## name ## _h(const char *cp, type *res)		\
+{								\
+	int u = 0;						\
+	char *e;						\
+	type i = simple_ ## name(cp, &e, 10);			\
+								\
+	switch (tolower(*e)) {					\
+	default:						\
+		return -EINVAL;					\
+	case 'y':						\
+	case 'z':						\
+		u++;						\
+	case 'e':						\
+		u++;						\
+	case 'p':						\
+		u++;						\
+	case 't':						\
+		u++;						\
+	case 'g':						\
+		u++;						\
+	case 'm':						\
+		u++;						\
+	case 'k':						\
+		u++;						\
+		if (e++ == cp)					\
+			return -EINVAL;				\
+	case '\n':						\
+	case '\0':						\
+		if (*e == '\n')					\
+			e++;					\
+	}							\
+								\
+	if (*e)							\
+		return -EINVAL;					\
+								\
+	while (u--) {						\
+		if ((type) ~0 > 0 &&				\
+		    (type) ~0 / 1024 <= i)			\
+			return -EINVAL;				\
+		if ((i > 0 && ANYSINT_MAX(type) / 1024 < i) ||	\
+		    (i < 0 && -ANYSINT_MAX(type) / 1024 > i))	\
+			return -EINVAL;				\
+		i *= 1024;					\
+	}							\
+								\
+	*res = i;						\
+	return 0;						\
+}								\
+
+STRTO_H(strtoint, int)
+STRTO_H(strtouint, unsigned int)
+STRTO_H(strtoll, long long)
+STRTO_H(strtoull, unsigned long long)
+
+ssize_t bch_hprint(char *buf, int64_t v)
+{
+	static const char units[] = "?kMGTPEZY";
+	char dec[4] = "";
+	int u, t = 0;
+
+	for (u = 0; v >= 1024 || v <= -1024; u++) {
+		t = v & ~(~0 << 10);
+		v >>= 10;
+	}
+
+	if (!u)
+		return sprintf(buf, "%llu", v);
+
+	if (v < 100 && v > -100)
+		snprintf(dec, sizeof(dec), ".%i", t / 100);
+
+	return sprintf(buf, "%lli%s%c", v, dec, units[u]);
+}
+
+ssize_t bch_snprint_string_list(char *buf, size_t size, const char * const list[],
+			    size_t selected)
+{
+	char *out = buf;
+	size_t i;
+
+	for (i = 0; list[i]; i++)
+		out += snprintf(out, buf + size - out,
+				i == selected ? "[%s] " : "%s ", list[i]);
+
+	out[-1] = '\n';
+	return out - buf;
+}
+
+ssize_t bch_read_string_list(const char *buf, const char * const list[])
+{
+	size_t i;
+	char *s, *d = kstrndup(buf, PAGE_SIZE - 1, GFP_KERNEL);
+	if (!d)
+		return -ENOMEM;
+
+	s = strim(d);
+
+	for (i = 0; list[i]; i++)
+		if (!strcmp(list[i], s))
+			break;
+
+	kfree(d);
+
+	if (!list[i])
+		return -EINVAL;
+
+	return i;
+}
+
+bool bch_is_zero(const char *p, size_t n)
+{
+	size_t i;
+
+	for (i = 0; i < n; i++)
+		if (p[i])
+			return false;
+	return true;
+}
+
+int bch_parse_uuid(const char *s, char *uuid)
+{
+	size_t i, j, x;
+	memset(uuid, 0, 16);
+
+	for (i = 0, j = 0;
+	     i < strspn(s, "-0123456789:ABCDEFabcdef") && j < 32;
+	     i++) {
+		x = s[i] | 32;
+
+		switch (x) {
+		case '0'...'9':
+			x -= '0';
+			break;
+		case 'a'...'f':
+			x -= 'a' - 10;
+			break;
+		default:
+			continue;
+		}
+
+		if (!(j & 1))
+			x <<= 4;
+		uuid[j++ >> 1] |= x;
+	}
+	return i;
+}
+
+void bch_time_stats_update(struct time_stats *stats, uint64_t start_time)
+{
+	uint64_t now, duration, last;
+
+	spin_lock(&stats->lock);
+
+	now		= local_clock();
+	duration	= time_after64(now, start_time)
+		? now - start_time : 0;
+	last		= time_after64(now, stats->last)
+		? now - stats->last : 0;
+
+	stats->max_duration = max(stats->max_duration, duration);
+
+	if (stats->last) {
+		ewma_add(stats->average_duration, duration, 8, 8);
+
+		if (stats->average_frequency)
+			ewma_add(stats->average_frequency, last, 8, 8);
+		else
+			stats->average_frequency  = last << 8;
+	} else {
+		stats->average_duration  = duration << 8;
+	}
+
+	stats->last = now ?: 1;
+
+	spin_unlock(&stats->lock);
+}
+
+/**
+ * bch_next_delay() - increment @d by the amount of work done, and return how
+ * long to delay until the next time to do some work.
+ *
+ * @d - the struct bch_ratelimit to update
+ * @done - the amount of work done, in arbitrary units
+ *
+ * Returns the amount of time to delay by, in jiffies
+ */
+uint64_t bch_next_delay(struct bch_ratelimit *d, uint64_t done)
+{
+	uint64_t now = local_clock();
+
+	d->next += div_u64(done * NSEC_PER_SEC, d->rate);
+
+	if (time_before64(now + NSEC_PER_SEC, d->next))
+		d->next = now + NSEC_PER_SEC;
+
+	if (time_after64(now - NSEC_PER_SEC * 2, d->next))
+		d->next = now - NSEC_PER_SEC * 2;
+
+	return time_after64(d->next, now)
+		? div_u64(d->next - now, NSEC_PER_SEC / HZ)
+		: 0;
+}
+
+void bch_bio_map(struct bio *bio, void *base)
+{
+	size_t size = bio->bi_iter.bi_size;
+	struct bio_vec *bv = bio->bi_io_vec;
+
+	BUG_ON(!bio->bi_iter.bi_size);
+	BUG_ON(bio->bi_vcnt);
+
+	bv->bv_offset = base ? ((unsigned long) base) % PAGE_SIZE : 0;
+	goto start;
+
+	for (; size; bio->bi_vcnt++, bv++) {
+		bv->bv_offset	= 0;
+start:		bv->bv_len	= min_t(size_t, PAGE_SIZE - bv->bv_offset,
+					size);
+		if (base) {
+			bv->bv_page = is_vmalloc_addr(base)
+				? vmalloc_to_page(base)
+				: virt_to_page(base);
+
+			base += bv->bv_len;
+		}
+
+		size -= bv->bv_len;
+	}
+}
+
+/*
+ * Portions Copyright (c) 1996-2001, PostgreSQL Global Development Group (Any
+ * use permitted, subject to terms of PostgreSQL license; see.)
+
+ * If we have a 64-bit integer type, then a 64-bit CRC looks just like the
+ * usual sort of implementation. (See Ross Williams' excellent introduction
+ * A PAINLESS GUIDE TO CRC ERROR DETECTION ALGORITHMS, available from
+ * ftp://ftp.rocksoft.com/papers/crc_v3.txt or several other net sites.)
+ * If we have no working 64-bit type, then fake it with two 32-bit registers.
+ *
+ * The present implementation is a normal (not "reflected", in Williams'
+ * terms) 64-bit CRC, using initial all-ones register contents and a final
+ * bit inversion. The chosen polynomial is borrowed from the DLT1 spec
+ * (ECMA-182, available from http://www.ecma.ch/ecma1/STAND/ECMA-182.HTM):
+ *
+ * x^64 + x^62 + x^57 + x^55 + x^54 + x^53 + x^52 + x^47 + x^46 + x^45 +
+ * x^40 + x^39 + x^38 + x^37 + x^35 + x^33 + x^32 + x^31 + x^29 + x^27 +
+ * x^24 + x^23 + x^22 + x^21 + x^19 + x^17 + x^13 + x^12 + x^10 + x^9 +
+ * x^7 + x^4 + x + 1
+*/
+
+static const uint64_t crc_table[256] = {
+	0x0000000000000000ULL, 0x42F0E1EBA9EA3693ULL, 0x85E1C3D753D46D26ULL,
+	0xC711223CFA3E5BB5ULL, 0x493366450E42ECDFULL, 0x0BC387AEA7A8DA4CULL,
+	0xCCD2A5925D9681F9ULL, 0x8E224479F47CB76AULL, 0x9266CC8A1C85D9BEULL,
+	0xD0962D61B56FEF2DULL, 0x17870F5D4F51B498ULL, 0x5577EEB6E6BB820BULL,
+	0xDB55AACF12C73561ULL, 0x99A54B24BB2D03F2ULL, 0x5EB4691841135847ULL,
+	0x1C4488F3E8F96ED4ULL, 0x663D78FF90E185EFULL, 0x24CD9914390BB37CULL,
+	0xE3DCBB28C335E8C9ULL, 0xA12C5AC36ADFDE5AULL, 0x2F0E1EBA9EA36930ULL,
+	0x6DFEFF5137495FA3ULL, 0xAAEFDD6DCD770416ULL, 0xE81F3C86649D3285ULL,
+	0xF45BB4758C645C51ULL, 0xB6AB559E258E6AC2ULL, 0x71BA77A2DFB03177ULL,
+	0x334A9649765A07E4ULL, 0xBD68D2308226B08EULL, 0xFF9833DB2BCC861DULL,
+	0x388911E7D1F2DDA8ULL, 0x7A79F00C7818EB3BULL, 0xCC7AF1FF21C30BDEULL,
+	0x8E8A101488293D4DULL, 0x499B3228721766F8ULL, 0x0B6BD3C3DBFD506BULL,
+	0x854997BA2F81E701ULL, 0xC7B97651866BD192ULL, 0x00A8546D7C558A27ULL,
+	0x4258B586D5BFBCB4ULL, 0x5E1C3D753D46D260ULL, 0x1CECDC9E94ACE4F3ULL,
+	0xDBFDFEA26E92BF46ULL, 0x990D1F49C77889D5ULL, 0x172F5B3033043EBFULL,
+	0x55DFBADB9AEE082CULL, 0x92CE98E760D05399ULL, 0xD03E790CC93A650AULL,
+	0xAA478900B1228E31ULL, 0xE8B768EB18C8B8A2ULL, 0x2FA64AD7E2F6E317ULL,
+	0x6D56AB3C4B1CD584ULL, 0xE374EF45BF6062EEULL, 0xA1840EAE168A547DULL,
+	0x66952C92ECB40FC8ULL, 0x2465CD79455E395BULL, 0x3821458AADA7578FULL,
+	0x7AD1A461044D611CULL, 0xBDC0865DFE733AA9ULL, 0xFF3067B657990C3AULL,
+	0x711223CFA3E5BB50ULL, 0x33E2C2240A0F8DC3ULL, 0xF4F3E018F031D676ULL,
+	0xB60301F359DBE0E5ULL, 0xDA050215EA6C212FULL, 0x98F5E3FE438617BCULL,
+	0x5FE4C1C2B9B84C09ULL, 0x1D14202910527A9AULL, 0x93366450E42ECDF0ULL,
+	0xD1C685BB4DC4FB63ULL, 0x16D7A787B7FAA0D6ULL, 0x5427466C1E109645ULL,
+	0x4863CE9FF6E9F891ULL, 0x0A932F745F03CE02ULL, 0xCD820D48A53D95B7ULL,
+	0x8F72ECA30CD7A324ULL, 0x0150A8DAF8AB144EULL, 0x43A04931514122DDULL,
+	0x84B16B0DAB7F7968ULL, 0xC6418AE602954FFBULL, 0xBC387AEA7A8DA4C0ULL,
+	0xFEC89B01D3679253ULL, 0x39D9B93D2959C9E6ULL, 0x7B2958D680B3FF75ULL,
+	0xF50B1CAF74CF481FULL, 0xB7FBFD44DD257E8CULL, 0x70EADF78271B2539ULL,
+	0x321A3E938EF113AAULL, 0x2E5EB66066087D7EULL, 0x6CAE578BCFE24BEDULL,
+	0xABBF75B735DC1058ULL, 0xE94F945C9C3626CBULL, 0x676DD025684A91A1ULL,
+	0x259D31CEC1A0A732ULL, 0xE28C13F23B9EFC87ULL, 0xA07CF2199274CA14ULL,
+	0x167FF3EACBAF2AF1ULL, 0x548F120162451C62ULL, 0x939E303D987B47D7ULL,
+	0xD16ED1D631917144ULL, 0x5F4C95AFC5EDC62EULL, 0x1DBC74446C07F0BDULL,
+	0xDAAD56789639AB08ULL, 0x985DB7933FD39D9BULL, 0x84193F60D72AF34FULL,
+	0xC6E9DE8B7EC0C5DCULL, 0x01F8FCB784FE9E69ULL, 0x43081D5C2D14A8FAULL,
+	0xCD2A5925D9681F90ULL, 0x8FDAB8CE70822903ULL, 0x48CB9AF28ABC72B6ULL,
+	0x0A3B7B1923564425ULL, 0x70428B155B4EAF1EULL, 0x32B26AFEF2A4998DULL,
+	0xF5A348C2089AC238ULL, 0xB753A929A170F4ABULL, 0x3971ED50550C43C1ULL,
+	0x7B810CBBFCE67552ULL, 0xBC902E8706D82EE7ULL, 0xFE60CF6CAF321874ULL,
+	0xE224479F47CB76A0ULL, 0xA0D4A674EE214033ULL, 0x67C58448141F1B86ULL,
+	0x253565A3BDF52D15ULL, 0xAB1721DA49899A7FULL, 0xE9E7C031E063ACECULL,
+	0x2EF6E20D1A5DF759ULL, 0x6C0603E6B3B7C1CAULL, 0xF6FAE5C07D3274CDULL,
+	0xB40A042BD4D8425EULL, 0x731B26172EE619EBULL, 0x31EBC7FC870C2F78ULL,
+	0xBFC9838573709812ULL, 0xFD39626EDA9AAE81ULL, 0x3A28405220A4F534ULL,
+	0x78D8A1B9894EC3A7ULL, 0x649C294A61B7AD73ULL, 0x266CC8A1C85D9BE0ULL,
+	0xE17DEA9D3263C055ULL, 0xA38D0B769B89F6C6ULL, 0x2DAF4F0F6FF541ACULL,
+	0x6F5FAEE4C61F773FULL, 0xA84E8CD83C212C8AULL, 0xEABE6D3395CB1A19ULL,
+	0x90C79D3FEDD3F122ULL, 0xD2377CD44439C7B1ULL, 0x15265EE8BE079C04ULL,
+	0x57D6BF0317EDAA97ULL, 0xD9F4FB7AE3911DFDULL, 0x9B041A914A7B2B6EULL,
+	0x5C1538ADB04570DBULL, 0x1EE5D94619AF4648ULL, 0x02A151B5F156289CULL,
+	0x4051B05E58BC1E0FULL, 0x87409262A28245BAULL, 0xC5B073890B687329ULL,
+	0x4B9237F0FF14C443ULL, 0x0962D61B56FEF2D0ULL, 0xCE73F427ACC0A965ULL,
+	0x8C8315CC052A9FF6ULL, 0x3A80143F5CF17F13ULL, 0x7870F5D4F51B4980ULL,
+	0xBF61D7E80F251235ULL, 0xFD913603A6CF24A6ULL, 0x73B3727A52B393CCULL,
+	0x31439391FB59A55FULL, 0xF652B1AD0167FEEAULL, 0xB4A25046A88DC879ULL,
+	0xA8E6D8B54074A6ADULL, 0xEA16395EE99E903EULL, 0x2D071B6213A0CB8BULL,
+	0x6FF7FA89BA4AFD18ULL, 0xE1D5BEF04E364A72ULL, 0xA3255F1BE7DC7CE1ULL,
+	0x64347D271DE22754ULL, 0x26C49CCCB40811C7ULL, 0x5CBD6CC0CC10FAFCULL,
+	0x1E4D8D2B65FACC6FULL, 0xD95CAF179FC497DAULL, 0x9BAC4EFC362EA149ULL,
+	0x158E0A85C2521623ULL, 0x577EEB6E6BB820B0ULL, 0x906FC95291867B05ULL,
+	0xD29F28B9386C4D96ULL, 0xCEDBA04AD0952342ULL, 0x8C2B41A1797F15D1ULL,
+	0x4B3A639D83414E64ULL, 0x09CA82762AAB78F7ULL, 0x87E8C60FDED7CF9DULL,
+	0xC51827E4773DF90EULL, 0x020905D88D03A2BBULL, 0x40F9E43324E99428ULL,
+	0x2CFFE7D5975E55E2ULL, 0x6E0F063E3EB46371ULL, 0xA91E2402C48A38C4ULL,
+	0xEBEEC5E96D600E57ULL, 0x65CC8190991CB93DULL, 0x273C607B30F68FAEULL,
+	0xE02D4247CAC8D41BULL, 0xA2DDA3AC6322E288ULL, 0xBE992B5F8BDB8C5CULL,
+	0xFC69CAB42231BACFULL, 0x3B78E888D80FE17AULL, 0x7988096371E5D7E9ULL,
+	0xF7AA4D1A85996083ULL, 0xB55AACF12C735610ULL, 0x724B8ECDD64D0DA5ULL,
+	0x30BB6F267FA73B36ULL, 0x4AC29F2A07BFD00DULL, 0x08327EC1AE55E69EULL,
+	0xCF235CFD546BBD2BULL, 0x8DD3BD16FD818BB8ULL, 0x03F1F96F09FD3CD2ULL,
+	0x41011884A0170A41ULL, 0x86103AB85A2951F4ULL, 0xC4E0DB53F3C36767ULL,
+	0xD8A453A01B3A09B3ULL, 0x9A54B24BB2D03F20ULL, 0x5D45907748EE6495ULL,
+	0x1FB5719CE1045206ULL, 0x919735E51578E56CULL, 0xD367D40EBC92D3FFULL,
+	0x1476F63246AC884AULL, 0x568617D9EF46BED9ULL, 0xE085162AB69D5E3CULL,
+	0xA275F7C11F7768AFULL, 0x6564D5FDE549331AULL, 0x279434164CA30589ULL,
+	0xA9B6706FB8DFB2E3ULL, 0xEB46918411358470ULL, 0x2C57B3B8EB0BDFC5ULL,
+	0x6EA7525342E1E956ULL, 0x72E3DAA0AA188782ULL, 0x30133B4B03F2B111ULL,
+	0xF7021977F9CCEAA4ULL, 0xB5F2F89C5026DC37ULL, 0x3BD0BCE5A45A6B5DULL,
+	0x79205D0E0DB05DCEULL, 0xBE317F32F78E067BULL, 0xFCC19ED95E6430E8ULL,
+	0x86B86ED5267CDBD3ULL, 0xC4488F3E8F96ED40ULL, 0x0359AD0275A8B6F5ULL,
+	0x41A94CE9DC428066ULL, 0xCF8B0890283E370CULL, 0x8D7BE97B81D4019FULL,
+	0x4A6ACB477BEA5A2AULL, 0x089A2AACD2006CB9ULL, 0x14DEA25F3AF9026DULL,
+	0x562E43B4931334FEULL, 0x913F6188692D6F4BULL, 0xD3CF8063C0C759D8ULL,
+	0x5DEDC41A34BBEEB2ULL, 0x1F1D25F19D51D821ULL, 0xD80C07CD676F8394ULL,
+	0x9AFCE626CE85B507ULL,
+};
+
+uint64_t bch_crc64_update(uint64_t crc, const void *_data, size_t len)
+{
+	const unsigned char *data = _data;
+
+	while (len--) {
+		int i = ((int) (crc >> 56) ^ *data++) & 0xFF;
+		crc = crc_table[i] ^ (crc << 8);
+	}
+
+	return crc;
+}
+
+uint64_t bch_crc64(const void *data, size_t len)
+{
+	uint64_t crc = 0xffffffffffffffffULL;
+
+	crc = bch_crc64_update(crc, data, len);
+
+	return crc ^ 0xffffffffffffffffULL;
+}
diff --git a/drivers/md/bcache/util.h b/drivers/md/bcache/util.h
new file mode 100644
index 00000000000..ac7d0d1f70d
--- /dev/null
+++ b/drivers/md/bcache/util.h
@@ -0,0 +1,588 @@
+
+#ifndef _BCACHE_UTIL_H
+#define _BCACHE_UTIL_H
+
+#include <linux/blkdev.h>
+#include <linux/errno.h>
+#include <linux/kernel.h>
+#include <linux/llist.h>
+#include <linux/ratelimit.h>
+#include <linux/vmalloc.h>
+#include <linux/workqueue.h>
+
+#include "closure.h"
+
+#define PAGE_SECTORS		(PAGE_SIZE / 512)
+
+struct closure;
+
+#ifdef CONFIG_BCACHE_DEBUG
+
+#define EBUG_ON(cond)			BUG_ON(cond)
+#define atomic_dec_bug(v)	BUG_ON(atomic_dec_return(v) < 0)
+#define atomic_inc_bug(v, i)	BUG_ON(atomic_inc_return(v) <= i)
+
+#else /* DEBUG */
+
+#define EBUG_ON(cond)			do { if (cond); } while (0)
+#define atomic_dec_bug(v)	atomic_dec(v)
+#define atomic_inc_bug(v, i)	atomic_inc(v)
+
+#endif
+
+#define DECLARE_HEAP(type, name)					\
+	struct {							\
+		size_t size, used;					\
+		type *data;						\
+	} name
+
+#define init_heap(heap, _size, gfp)					\
+({									\
+	size_t _bytes;							\
+	(heap)->used = 0;						\
+	(heap)->size = (_size);						\
+	_bytes = (heap)->size * sizeof(*(heap)->data);			\
+	(heap)->data = NULL;						\
+	if (_bytes < KMALLOC_MAX_SIZE)					\
+		(heap)->data = kmalloc(_bytes, (gfp));			\
+	if ((!(heap)->data) && ((gfp) & GFP_KERNEL))			\
+		(heap)->data = vmalloc(_bytes);				\
+	(heap)->data;							\
+})
+
+#define free_heap(heap)							\
+do {									\
+	if (is_vmalloc_addr((heap)->data))				\
+		vfree((heap)->data);					\
+	else								\
+		kfree((heap)->data);					\
+	(heap)->data = NULL;						\
+} while (0)
+
+#define heap_swap(h, i, j)	swap((h)->data[i], (h)->data[j])
+
+#define heap_sift(h, i, cmp)						\
+do {									\
+	size_t _r, _j = i;						\
+									\
+	for (; _j * 2 + 1 < (h)->used; _j = _r) {			\
+		_r = _j * 2 + 1;					\
+		if (_r + 1 < (h)->used &&				\
+		    cmp((h)->data[_r], (h)->data[_r + 1]))		\
+			_r++;						\
+									\
+		if (cmp((h)->data[_r], (h)->data[_j]))			\
+			break;						\
+		heap_swap(h, _r, _j);					\
+	}								\
+} while (0)
+
+#define heap_sift_down(h, i, cmp)					\
+do {									\
+	while (i) {							\
+		size_t p = (i - 1) / 2;					\
+		if (cmp((h)->data[i], (h)->data[p]))			\
+			break;						\
+		heap_swap(h, i, p);					\
+		i = p;							\
+	}								\
+} while (0)
+
+#define heap_add(h, d, cmp)						\
+({									\
+	bool _r = !heap_full(h);					\
+	if (_r) {							\
+		size_t _i = (h)->used++;				\
+		(h)->data[_i] = d;					\
+									\
+		heap_sift_down(h, _i, cmp);				\
+		heap_sift(h, _i, cmp);					\
+	}								\
+	_r;								\
+})
+
+#define heap_pop(h, d, cmp)						\
+({									\
+	bool _r = (h)->used;						\
+	if (_r) {							\
+		(d) = (h)->data[0];					\
+		(h)->used--;						\
+		heap_swap(h, 0, (h)->used);				\
+		heap_sift(h, 0, cmp);					\
+	}								\
+	_r;								\
+})
+
+#define heap_peek(h)	((h)->used ? (h)->data[0] : NULL)
+
+#define heap_full(h)	((h)->used == (h)->size)
+
+#define DECLARE_FIFO(type, name)					\
+	struct {							\
+		size_t front, back, size, mask;				\
+		type *data;						\
+	} name
+
+#define fifo_for_each(c, fifo, iter)					\
+	for (iter = (fifo)->front;					\
+	     c = (fifo)->data[iter], iter != (fifo)->back;		\
+	     iter = (iter + 1) & (fifo)->mask)
+
+#define __init_fifo(fifo, gfp)						\
+({									\
+	size_t _allocated_size, _bytes;					\
+	BUG_ON(!(fifo)->size);						\
+									\
+	_allocated_size = roundup_pow_of_two((fifo)->size + 1);		\
+	_bytes = _allocated_size * sizeof(*(fifo)->data);		\
+									\
+	(fifo)->mask = _allocated_size - 1;				\
+	(fifo)->front = (fifo)->back = 0;				\
+	(fifo)->data = NULL;						\
+									\
+	if (_bytes < KMALLOC_MAX_SIZE)					\
+		(fifo)->data = kmalloc(_bytes, (gfp));			\
+	if ((!(fifo)->data) && ((gfp) & GFP_KERNEL))			\
+		(fifo)->data = vmalloc(_bytes);				\
+	(fifo)->data;							\
+})
+
+#define init_fifo_exact(fifo, _size, gfp)				\
+({									\
+	(fifo)->size = (_size);						\
+	__init_fifo(fifo, gfp);						\
+})
+
+#define init_fifo(fifo, _size, gfp)					\
+({									\
+	(fifo)->size = (_size);						\
+	if ((fifo)->size > 4)						\
+		(fifo)->size = roundup_pow_of_two((fifo)->size) - 1;	\
+	__init_fifo(fifo, gfp);						\
+})
+
+#define free_fifo(fifo)							\
+do {									\
+	if (is_vmalloc_addr((fifo)->data))				\
+		vfree((fifo)->data);					\
+	else								\
+		kfree((fifo)->data);					\
+	(fifo)->data = NULL;						\
+} while (0)
+
+#define fifo_used(fifo)		(((fifo)->back - (fifo)->front) & (fifo)->mask)
+#define fifo_free(fifo)		((fifo)->size - fifo_used(fifo))
+
+#define fifo_empty(fifo)	(!fifo_used(fifo))
+#define fifo_full(fifo)		(!fifo_free(fifo))
+
+#define fifo_front(fifo)	((fifo)->data[(fifo)->front])
+#define fifo_back(fifo)							\
+	((fifo)->data[((fifo)->back - 1) & (fifo)->mask])
+
+#define fifo_idx(fifo, p)	(((p) - &fifo_front(fifo)) & (fifo)->mask)
+
+#define fifo_push_back(fifo, i)						\
+({									\
+	bool _r = !fifo_full((fifo));					\
+	if (_r) {							\
+		(fifo)->data[(fifo)->back++] = (i);			\
+		(fifo)->back &= (fifo)->mask;				\
+	}								\
+	_r;								\
+})
+
+#define fifo_pop_front(fifo, i)						\
+({									\
+	bool _r = !fifo_empty((fifo));					\
+	if (_r) {							\
+		(i) = (fifo)->data[(fifo)->front++];			\
+		(fifo)->front &= (fifo)->mask;				\
+	}								\
+	_r;								\
+})
+
+#define fifo_push_front(fifo, i)					\
+({									\
+	bool _r = !fifo_full((fifo));					\
+	if (_r) {							\
+		--(fifo)->front;					\
+		(fifo)->front &= (fifo)->mask;				\
+		(fifo)->data[(fifo)->front] = (i);			\
+	}								\
+	_r;								\
+})
+
+#define fifo_pop_back(fifo, i)						\
+({									\
+	bool _r = !fifo_empty((fifo));					\
+	if (_r) {							\
+		--(fifo)->back;						\
+		(fifo)->back &= (fifo)->mask;				\
+		(i) = (fifo)->data[(fifo)->back]			\
+	}								\
+	_r;								\
+})
+
+#define fifo_push(fifo, i)	fifo_push_back(fifo, (i))
+#define fifo_pop(fifo, i)	fifo_pop_front(fifo, (i))
+
+#define fifo_swap(l, r)							\
+do {									\
+	swap((l)->front, (r)->front);					\
+	swap((l)->back, (r)->back);					\
+	swap((l)->size, (r)->size);					\
+	swap((l)->mask, (r)->mask);					\
+	swap((l)->data, (r)->data);					\
+} while (0)
+
+#define fifo_move(dest, src)						\
+do {									\
+	typeof(*((dest)->data)) _t;					\
+	while (!fifo_full(dest) &&					\
+	       fifo_pop(src, _t))					\
+		fifo_push(dest, _t);					\
+} while (0)
+
+/*
+ * Simple array based allocator - preallocates a number of elements and you can
+ * never allocate more than that, also has no locking.
+ *
+ * Handy because if you know you only need a fixed number of elements you don't
+ * have to worry about memory allocation failure, and sometimes a mempool isn't
+ * what you want.
+ *
+ * We treat the free elements as entries in a singly linked list, and the
+ * freelist as a stack - allocating and freeing push and pop off the freelist.
+ */
+
+#define DECLARE_ARRAY_ALLOCATOR(type, name, size)			\
+	struct {							\
+		type	*freelist;					\
+		type	data[size];					\
+	} name
+
+#define array_alloc(array)						\
+({									\
+	typeof((array)->freelist) _ret = (array)->freelist;		\
+									\
+	if (_ret)							\
+		(array)->freelist = *((typeof((array)->freelist) *) _ret);\
+									\
+	_ret;								\
+})
+
+#define array_free(array, ptr)						\
+do {									\
+	typeof((array)->freelist) _ptr = ptr;				\
+									\
+	*((typeof((array)->freelist) *) _ptr) = (array)->freelist;	\
+	(array)->freelist = _ptr;					\
+} while (0)
+
+#define array_allocator_init(array)					\
+do {									\
+	typeof((array)->freelist) _i;					\
+									\
+	BUILD_BUG_ON(sizeof((array)->data[0]) < sizeof(void *));	\
+	(array)->freelist = NULL;					\
+									\
+	for (_i = (array)->data;					\
+	     _i < (array)->data + ARRAY_SIZE((array)->data);		\
+	     _i++)							\
+		array_free(array, _i);					\
+} while (0)
+
+#define array_freelist_empty(array)	((array)->freelist == NULL)
+
+#define ANYSINT_MAX(t)							\
+	((((t) 1 << (sizeof(t) * 8 - 2)) - (t) 1) * (t) 2 + (t) 1)
+
+int bch_strtoint_h(const char *, int *);
+int bch_strtouint_h(const char *, unsigned int *);
+int bch_strtoll_h(const char *, long long *);
+int bch_strtoull_h(const char *, unsigned long long *);
+
+static inline int bch_strtol_h(const char *cp, long *res)
+{
+#if BITS_PER_LONG == 32
+	return bch_strtoint_h(cp, (int *) res);
+#else
+	return bch_strtoll_h(cp, (long long *) res);
+#endif
+}
+
+static inline int bch_strtoul_h(const char *cp, long *res)
+{
+#if BITS_PER_LONG == 32
+	return bch_strtouint_h(cp, (unsigned int *) res);
+#else
+	return bch_strtoull_h(cp, (unsigned long long *) res);
+#endif
+}
+
+#define strtoi_h(cp, res)						\
+	(__builtin_types_compatible_p(typeof(*res), int)		\
+	? bch_strtoint_h(cp, (void *) res)				\
+	: __builtin_types_compatible_p(typeof(*res), long)		\
+	? bch_strtol_h(cp, (void *) res)				\
+	: __builtin_types_compatible_p(typeof(*res), long long)		\
+	? bch_strtoll_h(cp, (void *) res)				\
+	: __builtin_types_compatible_p(typeof(*res), unsigned int)	\
+	? bch_strtouint_h(cp, (void *) res)				\
+	: __builtin_types_compatible_p(typeof(*res), unsigned long)	\
+	? bch_strtoul_h(cp, (void *) res)				\
+	: __builtin_types_compatible_p(typeof(*res), unsigned long long)\
+	? bch_strtoull_h(cp, (void *) res) : -EINVAL)
+
+#define strtoul_safe(cp, var)						\
+({									\
+	unsigned long _v;						\
+	int _r = kstrtoul(cp, 10, &_v);					\
+	if (!_r)							\
+		var = _v;						\
+	_r;								\
+})
+
+#define strtoul_safe_clamp(cp, var, min, max)				\
+({									\
+	unsigned long _v;						\
+	int _r = kstrtoul(cp, 10, &_v);					\
+	if (!_r)							\
+		var = clamp_t(typeof(var), _v, min, max);		\
+	_r;								\
+})
+
+#define snprint(buf, size, var)						\
+	snprintf(buf, size,						\
+		__builtin_types_compatible_p(typeof(var), int)		\
+		     ? "%i\n" :						\
+		__builtin_types_compatible_p(typeof(var), unsigned)	\
+		     ? "%u\n" :						\
+		__builtin_types_compatible_p(typeof(var), long)		\
+		     ? "%li\n" :					\
+		__builtin_types_compatible_p(typeof(var), unsigned long)\
+		     ? "%lu\n" :					\
+		__builtin_types_compatible_p(typeof(var), int64_t)	\
+		     ? "%lli\n" :					\
+		__builtin_types_compatible_p(typeof(var), uint64_t)	\
+		     ? "%llu\n" :					\
+		__builtin_types_compatible_p(typeof(var), const char *)	\
+		     ? "%s\n" : "%i\n", var)
+
+ssize_t bch_hprint(char *buf, int64_t v);
+
+bool bch_is_zero(const char *p, size_t n);
+int bch_parse_uuid(const char *s, char *uuid);
+
+ssize_t bch_snprint_string_list(char *buf, size_t size, const char * const list[],
+			    size_t selected);
+
+ssize_t bch_read_string_list(const char *buf, const char * const list[]);
+
+struct time_stats {
+	spinlock_t	lock;
+	/*
+	 * all fields are in nanoseconds, averages are ewmas stored left shifted
+	 * by 8
+	 */
+	uint64_t	max_duration;
+	uint64_t	average_duration;
+	uint64_t	average_frequency;
+	uint64_t	last;
+};
+
+void bch_time_stats_update(struct time_stats *stats, uint64_t time);
+
+static inline unsigned local_clock_us(void)
+{
+	return local_clock() >> 10;
+}
+
+#define NSEC_PER_ns			1L
+#define NSEC_PER_us			NSEC_PER_USEC
+#define NSEC_PER_ms			NSEC_PER_MSEC
+#define NSEC_PER_sec			NSEC_PER_SEC
+
+#define __print_time_stat(stats, name, stat, units)			\
+	sysfs_print(name ## _ ## stat ## _ ## units,			\
+		    div_u64((stats)->stat >> 8, NSEC_PER_ ## units))
+
+#define sysfs_print_time_stats(stats, name,				\
+			       frequency_units,				\
+			       duration_units)				\
+do {									\
+	__print_time_stat(stats, name,					\
+			  average_frequency,	frequency_units);	\
+	__print_time_stat(stats, name,					\
+			  average_duration,	duration_units);	\
+	__print_time_stat(stats, name,					\
+			  max_duration,		duration_units);	\
+									\
+	sysfs_print(name ## _last_ ## frequency_units, (stats)->last	\
+		    ? div_s64(local_clock() - (stats)->last,		\
+			      NSEC_PER_ ## frequency_units)		\
+		    : -1LL);						\
+} while (0)
+
+#define sysfs_time_stats_attribute(name,				\
+				   frequency_units,			\
+				   duration_units)			\
+read_attribute(name ## _average_frequency_ ## frequency_units);		\
+read_attribute(name ## _average_duration_ ## duration_units);		\
+read_attribute(name ## _max_duration_ ## duration_units);		\
+read_attribute(name ## _last_ ## frequency_units)
+
+#define sysfs_time_stats_attribute_list(name,				\
+					frequency_units,		\
+					duration_units)			\
+&sysfs_ ## name ## _average_frequency_ ## frequency_units,		\
+&sysfs_ ## name ## _average_duration_ ## duration_units,		\
+&sysfs_ ## name ## _max_duration_ ## duration_units,			\
+&sysfs_ ## name ## _last_ ## frequency_units,
+
+#define ewma_add(ewma, val, weight, factor)				\
+({									\
+	(ewma) *= (weight) - 1;						\
+	(ewma) += (val) << factor;					\
+	(ewma) /= (weight);						\
+	(ewma) >> factor;						\
+})
+
+struct bch_ratelimit {
+	/* Next time we want to do some work, in nanoseconds */
+	uint64_t		next;
+
+	/*
+	 * Rate at which we want to do work, in units per nanosecond
+	 * The units here correspond to the units passed to bch_next_delay()
+	 */
+	unsigned		rate;
+};
+
+static inline void bch_ratelimit_reset(struct bch_ratelimit *d)
+{
+	d->next = local_clock();
+}
+
+uint64_t bch_next_delay(struct bch_ratelimit *d, uint64_t done);
+
+#define __DIV_SAFE(n, d, zero)						\
+({									\
+	typeof(n) _n = (n);						\
+	typeof(d) _d = (d);						\
+	_d ? _n / _d : zero;						\
+})
+
+#define DIV_SAFE(n, d)	__DIV_SAFE(n, d, 0)
+
+#define container_of_or_null(ptr, type, member)				\
+({									\
+	typeof(ptr) _ptr = ptr;						\
+	_ptr ? container_of(_ptr, type, member) : NULL;			\
+})
+
+#define RB_INSERT(root, new, member, cmp)				\
+({									\
+	__label__ dup;							\
+	struct rb_node **n = &(root)->rb_node, *parent = NULL;		\
+	typeof(new) this;						\
+	int res, ret = -1;						\
+									\
+	while (*n) {							\
+		parent = *n;						\
+		this = container_of(*n, typeof(*(new)), member);	\
+		res = cmp(new, this);					\
+		if (!res)						\
+			goto dup;					\
+		n = res < 0						\
+			? &(*n)->rb_left				\
+			: &(*n)->rb_right;				\
+	}								\
+									\
+	rb_link_node(&(new)->member, parent, n);			\
+	rb_insert_color(&(new)->member, root);				\
+	ret = 0;							\
+dup:									\
+	ret;								\
+})
+
+#define RB_SEARCH(root, search, member, cmp)				\
+({									\
+	struct rb_node *n = (root)->rb_node;				\
+	typeof(&(search)) this, ret = NULL;				\
+	int res;							\
+									\
+	while (n) {							\
+		this = container_of(n, typeof(search), member);		\
+		res = cmp(&(search), this);				\
+		if (!res) {						\
+			ret = this;					\
+			break;						\
+		}							\
+		n = res < 0						\
+			? n->rb_left					\
+			: n->rb_right;					\
+	}								\
+	ret;								\
+})
+
+#define RB_GREATER(root, search, member, cmp)				\
+({									\
+	struct rb_node *n = (root)->rb_node;				\
+	typeof(&(search)) this, ret = NULL;				\
+	int res;							\
+									\
+	while (n) {							\
+		this = container_of(n, typeof(search), member);		\
+		res = cmp(&(search), this);				\
+		if (res < 0) {						\
+			ret = this;					\
+			n = n->rb_left;					\
+		} else							\
+			n = n->rb_right;				\
+	}								\
+	ret;								\
+})
+
+#define RB_FIRST(root, type, member)					\
+	container_of_or_null(rb_first(root), type, member)
+
+#define RB_LAST(root, type, member)					\
+	container_of_or_null(rb_last(root), type, member)
+
+#define RB_NEXT(ptr, member)						\
+	container_of_or_null(rb_next(&(ptr)->member), typeof(*ptr), member)
+
+#define RB_PREV(ptr, member)						\
+	container_of_or_null(rb_prev(&(ptr)->member), typeof(*ptr), member)
+
+/* Does linear interpolation between powers of two */
+static inline unsigned fract_exp_two(unsigned x, unsigned fract_bits)
+{
+	unsigned fract = x & ~(~0 << fract_bits);
+
+	x >>= fract_bits;
+	x   = 1 << x;
+	x  += (x * fract) >> fract_bits;
+
+	return x;
+}
+
+void bch_bio_map(struct bio *bio, void *base);
+
+static inline sector_t bdev_sectors(struct block_device *bdev)
+{
+	return bdev->bd_inode->i_size >> 9;
+}
+
+#define closure_bio_submit(bio, cl, dev)				\
+do {									\
+	closure_get(cl);						\
+	bch_generic_make_request(bio, &(dev)->bio_split_hook);		\
+} while (0)
+
+uint64_t bch_crc64_update(uint64_t, const void *, size_t);
+uint64_t bch_crc64(const void *, size_t);
+
+#endif /* _BCACHE_UTIL_H */
diff --git a/drivers/md/bcache/writeback.c b/drivers/md/bcache/writeback.c
new file mode 100644
index 00000000000..f4300e4c011
--- /dev/null
+++ b/drivers/md/bcache/writeback.c
@@ -0,0 +1,507 @@
+/*
+ * background writeback - scan btree for dirty data and write it to the backing
+ * device
+ *
+ * Copyright 2010, 2011 Kent Overstreet <kent.overstreet@gmail.com>
+ * Copyright 2012 Google, Inc.
+ */
+
+#include "bcache.h"
+#include "btree.h"
+#include "debug.h"
+#include "writeback.h"
+
+#include <linux/delay.h>
+#include <linux/freezer.h>
+#include <linux/kthread.h>
+#include <trace/events/bcache.h>
+
+/* Rate limiting */
+
+static void __update_writeback_rate(struct cached_dev *dc)
+{
+	struct cache_set *c = dc->disk.c;
+	uint64_t cache_sectors = c->nbuckets * c->sb.bucket_size;
+	uint64_t cache_dirty_target =
+		div_u64(cache_sectors * dc->writeback_percent, 100);
+
+	int64_t target = div64_u64(cache_dirty_target * bdev_sectors(dc->bdev),
+				   c->cached_dev_sectors);
+
+	/* PD controller */
+
+	int64_t dirty = bcache_dev_sectors_dirty(&dc->disk);
+	int64_t derivative = dirty - dc->disk.sectors_dirty_last;
+	int64_t proportional = dirty - target;
+	int64_t change;
+
+	dc->disk.sectors_dirty_last = dirty;
+
+	/* Scale to sectors per second */
+
+	proportional *= dc->writeback_rate_update_seconds;
+	proportional = div_s64(proportional, dc->writeback_rate_p_term_inverse);
+
+	derivative = div_s64(derivative, dc->writeback_rate_update_seconds);
+
+	derivative = ewma_add(dc->disk.sectors_dirty_derivative, derivative,
+			      (dc->writeback_rate_d_term /
+			       dc->writeback_rate_update_seconds) ?: 1, 0);
+
+	derivative *= dc->writeback_rate_d_term;
+	derivative = div_s64(derivative, dc->writeback_rate_p_term_inverse);
+
+	change = proportional + derivative;
+
+	/* Don't increase writeback rate if the device isn't keeping up */
+	if (change > 0 &&
+	    time_after64(local_clock(),
+			 dc->writeback_rate.next + NSEC_PER_MSEC))
+		change = 0;
+
+	dc->writeback_rate.rate =
+		clamp_t(int64_t, (int64_t) dc->writeback_rate.rate + change,
+			1, NSEC_PER_MSEC);
+
+	dc->writeback_rate_proportional = proportional;
+	dc->writeback_rate_derivative = derivative;
+	dc->writeback_rate_change = change;
+	dc->writeback_rate_target = target;
+}
+
+static void update_writeback_rate(struct work_struct *work)
+{
+	struct cached_dev *dc = container_of(to_delayed_work(work),
+					     struct cached_dev,
+					     writeback_rate_update);
+
+	down_read(&dc->writeback_lock);
+
+	if (atomic_read(&dc->has_dirty) &&
+	    dc->writeback_percent)
+		__update_writeback_rate(dc);
+
+	up_read(&dc->writeback_lock);
+
+	schedule_delayed_work(&dc->writeback_rate_update,
+			      dc->writeback_rate_update_seconds * HZ);
+}
+
+static unsigned writeback_delay(struct cached_dev *dc, unsigned sectors)
+{
+	if (test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) ||
+	    !dc->writeback_percent)
+		return 0;
+
+	return bch_next_delay(&dc->writeback_rate, sectors);
+}
+
+struct dirty_io {
+	struct closure		cl;
+	struct cached_dev	*dc;
+	struct bio		bio;
+};
+
+static void dirty_init(struct keybuf_key *w)
+{
+	struct dirty_io *io = w->private;
+	struct bio *bio = &io->bio;
+
+	bio_init(bio);
+	if (!io->dc->writeback_percent)
+		bio_set_prio(bio, IOPRIO_PRIO_VALUE(IOPRIO_CLASS_IDLE, 0));
+
+	bio->bi_iter.bi_size	= KEY_SIZE(&w->key) << 9;
+	bio->bi_max_vecs	= DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS);
+	bio->bi_private		= w;
+	bio->bi_io_vec		= bio->bi_inline_vecs;
+	bch_bio_map(bio, NULL);
+}
+
+static void dirty_io_destructor(struct closure *cl)
+{
+	struct dirty_io *io = container_of(cl, struct dirty_io, cl);
+	kfree(io);
+}
+
+static void write_dirty_finish(struct closure *cl)
+{
+	struct dirty_io *io = container_of(cl, struct dirty_io, cl);
+	struct keybuf_key *w = io->bio.bi_private;
+	struct cached_dev *dc = io->dc;
+	struct bio_vec *bv;
+	int i;
+
+	bio_for_each_segment_all(bv, &io->bio, i)
+		__free_page(bv->bv_page);
+
+	/* This is kind of a dumb way of signalling errors. */
+	if (KEY_DIRTY(&w->key)) {
+		int ret;
+		unsigned i;
+		struct keylist keys;
+
+		bch_keylist_init(&keys);
+
+		bkey_copy(keys.top, &w->key);
+		SET_KEY_DIRTY(keys.top, false);
+		bch_keylist_push(&keys);
+
+		for (i = 0; i < KEY_PTRS(&w->key); i++)
+			atomic_inc(&PTR_BUCKET(dc->disk.c, &w->key, i)->pin);
+
+		ret = bch_btree_insert(dc->disk.c, &keys, NULL, &w->key);
+
+		if (ret)
+			trace_bcache_writeback_collision(&w->key);
+
+		atomic_long_inc(ret
+				? &dc->disk.c->writeback_keys_failed
+				: &dc->disk.c->writeback_keys_done);
+	}
+
+	bch_keybuf_del(&dc->writeback_keys, w);
+	up(&dc->in_flight);
+
+	closure_return_with_destructor(cl, dirty_io_destructor);
+}
+
+static void dirty_endio(struct bio *bio, int error)
+{
+	struct keybuf_key *w = bio->bi_private;
+	struct dirty_io *io = w->private;
+
+	if (error)
+		SET_KEY_DIRTY(&w->key, false);
+
+	closure_put(&io->cl);
+}
+
+static void write_dirty(struct closure *cl)
+{
+	struct dirty_io *io = container_of(cl, struct dirty_io, cl);
+	struct keybuf_key *w = io->bio.bi_private;
+
+	dirty_init(w);
+	io->bio.bi_rw		= WRITE;
+	io->bio.bi_iter.bi_sector = KEY_START(&w->key);
+	io->bio.bi_bdev		= io->dc->bdev;
+	io->bio.bi_end_io	= dirty_endio;
+
+	closure_bio_submit(&io->bio, cl, &io->dc->disk);
+
+	continue_at(cl, write_dirty_finish, system_wq);
+}
+
+static void read_dirty_endio(struct bio *bio, int error)
+{
+	struct keybuf_key *w = bio->bi_private;
+	struct dirty_io *io = w->private;
+
+	bch_count_io_errors(PTR_CACHE(io->dc->disk.c, &w->key, 0),
+			    error, "reading dirty data from cache");
+
+	dirty_endio(bio, error);
+}
+
+static void read_dirty_submit(struct closure *cl)
+{
+	struct dirty_io *io = container_of(cl, struct dirty_io, cl);
+
+	closure_bio_submit(&io->bio, cl, &io->dc->disk);
+
+	continue_at(cl, write_dirty, system_wq);
+}
+
+static void read_dirty(struct cached_dev *dc)
+{
+	unsigned delay = 0;
+	struct keybuf_key *w;
+	struct dirty_io *io;
+	struct closure cl;
+
+	closure_init_stack(&cl);
+
+	/*
+	 * XXX: if we error, background writeback just spins. Should use some
+	 * mempools.
+	 */
+
+	while (!kthread_should_stop()) {
+		try_to_freeze();
+
+		w = bch_keybuf_next(&dc->writeback_keys);
+		if (!w)
+			break;
+
+		BUG_ON(ptr_stale(dc->disk.c, &w->key, 0));
+
+		if (KEY_START(&w->key) != dc->last_read ||
+		    jiffies_to_msecs(delay) > 50)
+			while (!kthread_should_stop() && delay)
+				delay = schedule_timeout_uninterruptible(delay);
+
+		dc->last_read	= KEY_OFFSET(&w->key);
+
+		io = kzalloc(sizeof(struct dirty_io) + sizeof(struct bio_vec)
+			     * DIV_ROUND_UP(KEY_SIZE(&w->key), PAGE_SECTORS),
+			     GFP_KERNEL);
+		if (!io)
+			goto err;
+
+		w->private	= io;
+		io->dc		= dc;
+
+		dirty_init(w);
+		io->bio.bi_iter.bi_sector = PTR_OFFSET(&w->key, 0);
+		io->bio.bi_bdev		= PTR_CACHE(dc->disk.c,
+						    &w->key, 0)->bdev;
+		io->bio.bi_rw		= READ;
+		io->bio.bi_end_io	= read_dirty_endio;
+
+		if (bio_alloc_pages(&io->bio, GFP_KERNEL))
+			goto err_free;
+
+		trace_bcache_writeback(&w->key);
+
+		down(&dc->in_flight);
+		closure_call(&io->cl, read_dirty_submit, NULL, &cl);
+
+		delay = writeback_delay(dc, KEY_SIZE(&w->key));
+	}
+
+	if (0) {
+err_free:
+		kfree(w->private);
+err:
+		bch_keybuf_del(&dc->writeback_keys, w);
+	}
+
+	/*
+	 * Wait for outstanding writeback IOs to finish (and keybuf slots to be
+	 * freed) before refilling again
+	 */
+	closure_sync(&cl);
+}
+
+/* Scan for dirty data */
+
+void bcache_dev_sectors_dirty_add(struct cache_set *c, unsigned inode,
+				  uint64_t offset, int nr_sectors)
+{
+	struct bcache_device *d = c->devices[inode];
+	unsigned stripe_offset, stripe, sectors_dirty;
+
+	if (!d)
+		return;
+
+	stripe = offset_to_stripe(d, offset);
+	stripe_offset = offset & (d->stripe_size - 1);
+
+	while (nr_sectors) {
+		int s = min_t(unsigned, abs(nr_sectors),
+			      d->stripe_size - stripe_offset);
+
+		if (nr_sectors < 0)
+			s = -s;
+
+		if (stripe >= d->nr_stripes)
+			return;
+
+		sectors_dirty = atomic_add_return(s,
+					d->stripe_sectors_dirty + stripe);
+		if (sectors_dirty == d->stripe_size)
+			set_bit(stripe, d->full_dirty_stripes);
+		else
+			clear_bit(stripe, d->full_dirty_stripes);
+
+		nr_sectors -= s;
+		stripe_offset = 0;
+		stripe++;
+	}
+}
+
+static bool dirty_pred(struct keybuf *buf, struct bkey *k)
+{
+	return KEY_DIRTY(k);
+}
+
+static void refill_full_stripes(struct cached_dev *dc)
+{
+	struct keybuf *buf = &dc->writeback_keys;
+	unsigned start_stripe, stripe, next_stripe;
+	bool wrapped = false;
+
+	stripe = offset_to_stripe(&dc->disk, KEY_OFFSET(&buf->last_scanned));
+
+	if (stripe >= dc->disk.nr_stripes)
+		stripe = 0;
+
+	start_stripe = stripe;
+
+	while (1) {
+		stripe = find_next_bit(dc->disk.full_dirty_stripes,
+				       dc->disk.nr_stripes, stripe);
+
+		if (stripe == dc->disk.nr_stripes)
+			goto next;
+
+		next_stripe = find_next_zero_bit(dc->disk.full_dirty_stripes,
+						 dc->disk.nr_stripes, stripe);
+
+		buf->last_scanned = KEY(dc->disk.id,
+					stripe * dc->disk.stripe_size, 0);
+
+		bch_refill_keybuf(dc->disk.c, buf,
+				  &KEY(dc->disk.id,
+				       next_stripe * dc->disk.stripe_size, 0),
+				  dirty_pred);
+
+		if (array_freelist_empty(&buf->freelist))
+			return;
+
+		stripe = next_stripe;
+next:
+		if (wrapped && stripe > start_stripe)
+			return;
+
+		if (stripe == dc->disk.nr_stripes) {
+			stripe = 0;
+			wrapped = true;
+		}
+	}
+}
+
+static bool refill_dirty(struct cached_dev *dc)
+{
+	struct keybuf *buf = &dc->writeback_keys;
+	struct bkey end = KEY(dc->disk.id, MAX_KEY_OFFSET, 0);
+	bool searched_from_start = false;
+
+	if (dc->partial_stripes_expensive) {
+		refill_full_stripes(dc);
+		if (array_freelist_empty(&buf->freelist))
+			return false;
+	}
+
+	if (bkey_cmp(&buf->last_scanned, &end) >= 0) {
+		buf->last_scanned = KEY(dc->disk.id, 0, 0);
+		searched_from_start = true;
+	}
+
+	bch_refill_keybuf(dc->disk.c, buf, &end, dirty_pred);
+
+	return bkey_cmp(&buf->last_scanned, &end) >= 0 && searched_from_start;
+}
+
+static int bch_writeback_thread(void *arg)
+{
+	struct cached_dev *dc = arg;
+	bool searched_full_index;
+
+	while (!kthread_should_stop()) {
+		down_write(&dc->writeback_lock);
+		if (!atomic_read(&dc->has_dirty) ||
+		    (!test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) &&
+		     !dc->writeback_running)) {
+			up_write(&dc->writeback_lock);
+			set_current_state(TASK_INTERRUPTIBLE);
+
+			if (kthread_should_stop())
+				return 0;
+
+			try_to_freeze();
+			schedule();
+			continue;
+		}
+
+		searched_full_index = refill_dirty(dc);
+
+		if (searched_full_index &&
+		    RB_EMPTY_ROOT(&dc->writeback_keys.keys)) {
+			atomic_set(&dc->has_dirty, 0);
+			cached_dev_put(dc);
+			SET_BDEV_STATE(&dc->sb, BDEV_STATE_CLEAN);
+			bch_write_bdev_super(dc, NULL);
+		}
+
+		up_write(&dc->writeback_lock);
+
+		bch_ratelimit_reset(&dc->writeback_rate);
+		read_dirty(dc);
+
+		if (searched_full_index) {
+			unsigned delay = dc->writeback_delay * HZ;
+
+			while (delay &&
+			       !kthread_should_stop() &&
+			       !test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags))
+				delay = schedule_timeout_uninterruptible(delay);
+		}
+	}
+
+	return 0;
+}
+
+/* Init */
+
+struct sectors_dirty_init {
+	struct btree_op	op;
+	unsigned	inode;
+};
+
+static int sectors_dirty_init_fn(struct btree_op *_op, struct btree *b,
+				 struct bkey *k)
+{
+	struct sectors_dirty_init *op = container_of(_op,
+						struct sectors_dirty_init, op);
+	if (KEY_INODE(k) > op->inode)
+		return MAP_DONE;
+
+	if (KEY_DIRTY(k))
+		bcache_dev_sectors_dirty_add(b->c, KEY_INODE(k),
+					     KEY_START(k), KEY_SIZE(k));
+
+	return MAP_CONTINUE;
+}
+
+void bch_sectors_dirty_init(struct cached_dev *dc)
+{
+	struct sectors_dirty_init op;
+
+	bch_btree_op_init(&op.op, -1);
+	op.inode = dc->disk.id;
+
+	bch_btree_map_keys(&op.op, dc->disk.c, &KEY(op.inode, 0, 0),
+			   sectors_dirty_init_fn, 0);
+
+	dc->disk.sectors_dirty_last = bcache_dev_sectors_dirty(&dc->disk);
+}
+
+int bch_cached_dev_writeback_init(struct cached_dev *dc)
+{
+	sema_init(&dc->in_flight, 64);
+	init_rwsem(&dc->writeback_lock);
+	bch_keybuf_init(&dc->writeback_keys);
+
+	dc->writeback_metadata		= true;
+	dc->writeback_running		= true;
+	dc->writeback_percent		= 10;
+	dc->writeback_delay		= 30;
+	dc->writeback_rate.rate		= 1024;
+
+	dc->writeback_rate_update_seconds = 5;
+	dc->writeback_rate_d_term	= 30;
+	dc->writeback_rate_p_term_inverse = 6000;
+
+	dc->writeback_thread = kthread_create(bch_writeback_thread, dc,
+					      "bcache_writeback");
+	if (IS_ERR(dc->writeback_thread))
+		return PTR_ERR(dc->writeback_thread);
+
+	INIT_DELAYED_WORK(&dc->writeback_rate_update, update_writeback_rate);
+	schedule_delayed_work(&dc->writeback_rate_update,
+			      dc->writeback_rate_update_seconds * HZ);
+
+	return 0;
+}
diff --git a/drivers/md/bcache/writeback.h b/drivers/md/bcache/writeback.h
new file mode 100644
index 00000000000..e2f8598937a
--- /dev/null
+++ b/drivers/md/bcache/writeback.h
@@ -0,0 +1,90 @@
+#ifndef _BCACHE_WRITEBACK_H
+#define _BCACHE_WRITEBACK_H
+
+#define CUTOFF_WRITEBACK	40
+#define CUTOFF_WRITEBACK_SYNC	70
+
+static inline uint64_t bcache_dev_sectors_dirty(struct bcache_device *d)
+{
+	uint64_t i, ret = 0;
+
+	for (i = 0; i < d->nr_stripes; i++)
+		ret += atomic_read(d->stripe_sectors_dirty + i);
+
+	return ret;
+}
+
+static inline unsigned offset_to_stripe(struct bcache_device *d,
+					uint64_t offset)
+{
+	do_div(offset, d->stripe_size);
+	return offset;
+}
+
+static inline bool bcache_dev_stripe_dirty(struct cached_dev *dc,
+					   uint64_t offset,
+					   unsigned nr_sectors)
+{
+	unsigned stripe = offset_to_stripe(&dc->disk, offset);
+
+	while (1) {
+		if (atomic_read(dc->disk.stripe_sectors_dirty + stripe))
+			return true;
+
+		if (nr_sectors <= dc->disk.stripe_size)
+			return false;
+
+		nr_sectors -= dc->disk.stripe_size;
+		stripe++;
+	}
+}
+
+static inline bool should_writeback(struct cached_dev *dc, struct bio *bio,
+				    unsigned cache_mode, bool would_skip)
+{
+	unsigned in_use = dc->disk.c->gc_stats.in_use;
+
+	if (cache_mode != CACHE_MODE_WRITEBACK ||
+	    test_bit(BCACHE_DEV_DETACHING, &dc->disk.flags) ||
+	    in_use > CUTOFF_WRITEBACK_SYNC)
+		return false;
+
+	if (dc->partial_stripes_expensive &&
+	    bcache_dev_stripe_dirty(dc, bio->bi_iter.bi_sector,
+				    bio_sectors(bio)))
+		return true;
+
+	if (would_skip)
+		return false;
+
+	return bio->bi_rw & REQ_SYNC ||
+		in_use <= CUTOFF_WRITEBACK;
+}
+
+static inline void bch_writeback_queue(struct cached_dev *dc)
+{
+	wake_up_process(dc->writeback_thread);
+}
+
+static inline void bch_writeback_add(struct cached_dev *dc)
+{
+	if (!atomic_read(&dc->has_dirty) &&
+	    !atomic_xchg(&dc->has_dirty, 1)) {
+		atomic_inc(&dc->count);
+
+		if (BDEV_STATE(&dc->sb) != BDEV_STATE_DIRTY) {
+			SET_BDEV_STATE(&dc->sb, BDEV_STATE_DIRTY);
+			/* XXX: should do this synchronously */
+			bch_write_bdev_super(dc, NULL);
+		}
+
+		bch_writeback_queue(dc);
+	}
+}
+
+void bcache_dev_sectors_dirty_add(struct cache_set *, unsigned, uint64_t, int);
+
+void bch_sectors_dirty_init(struct cached_dev *dc);
+int bch_cached_dev_writeback_init(struct cached_dev *);
+
+#endif
diff --git a/drivers/md/bitmap.c b/drivers/md/bitmap.c
index ac89a5deaca..67f8b31e205 100644
--- a/drivers/md/bitmap.c
+++ b/drivers/md/bitmap.c
@@ -13,9 +13,9 @@
  * Still to do:
  *
  * flush after percent set rather than just time based. (maybe both).
- * wait if count gets too high, wake when it drops to half.
  */
 
+#include <linux/blkdev.h>
 #include <linux/module.h>
 #include <linux/errno.h>
 #include <linux/slab.h>
@@ -26,76 +26,15 @@
 #include <linux/file.h>
 #include <linux/mount.h>
 #include <linux/buffer_head.h>
-#include <linux/raid/md.h>
-#include <linux/raid/bitmap.h>
+#include <linux/seq_file.h>
+#include "md.h"
+#include "bitmap.h"
 
-/* debug macros */
-
-#define DEBUG 0
-
-#if DEBUG
-/* these are for debugging purposes only! */
-
-/* define one and only one of these */
-#define INJECT_FAULTS_1 0 /* cause bitmap_alloc_page to fail always */
-#define INJECT_FAULTS_2 0 /* cause bitmap file to be kicked when first bit set*/
-#define INJECT_FAULTS_3 0 /* treat bitmap file as kicked at init time */
-#define INJECT_FAULTS_4 0 /* undef */
-#define INJECT_FAULTS_5 0 /* undef */
-#define INJECT_FAULTS_6 0
-
-/* if these are defined, the driver will fail! debug only */
-#define INJECT_FATAL_FAULT_1 0 /* fail kmalloc, causing bitmap_create to fail */
-#define INJECT_FATAL_FAULT_2 0 /* undef */
-#define INJECT_FATAL_FAULT_3 0 /* undef */
-#endif
-
-//#define DPRINTK PRINTK /* set this NULL to avoid verbose debug output */
-#define DPRINTK(x...) do { } while(0)
-
-#ifndef PRINTK
-#  if DEBUG > 0
-#    define PRINTK(x...) printk(KERN_DEBUG x)
-#  else
-#    define PRINTK(x...)
-#  endif
-#endif
-
-static inline char * bmname(struct bitmap *bitmap)
+static inline char *bmname(struct bitmap *bitmap)
 {
 	return bitmap->mddev ? mdname(bitmap->mddev) : "mdX";
 }
 
-
-/*
- * just a placeholder - calls kmalloc for bitmap pages
- */
-static unsigned char *bitmap_alloc_page(struct bitmap *bitmap)
-{
-	unsigned char *page;
-
-#ifdef INJECT_FAULTS_1
-	page = NULL;
-#else
-	page = kmalloc(PAGE_SIZE, GFP_NOIO);
-#endif
-	if (!page)
-		printk("%s: bitmap_alloc_page FAILED\n", bmname(bitmap));
-	else
-		PRINTK("%s: bitmap_alloc_page: allocated page at %p\n",
-			bmname(bitmap), page);
-	return page;
-}
-
-/*
- * for now just a placeholder -- just calls kfree for bitmap pages
- */
-static void bitmap_free_page(struct bitmap *bitmap, unsigned char *page)
-{
-	PRINTK("%s: bitmap_free_page: free page %p\n", bmname(bitmap), page);
-	kfree(page);
-}
-
 /*
  * check a page and, if necessary, allocate it (or hijack it if the alloc fails)
  *
@@ -106,18 +45,21 @@ static void bitmap_free_page(struct bitmap *bitmap, unsigned char *page)
  * if we find our page, we increment the page's refcount so that it stays
  * allocated while we're using it
  */
-static int bitmap_checkpage(struct bitmap *bitmap, unsigned long page, int create)
+static int bitmap_checkpage(struct bitmap_counts *bitmap,
+			    unsigned long page, int create)
+__releases(bitmap->lock)
+__acquires(bitmap->lock)
 {
 	unsigned char *mappage;
 
 	if (page >= bitmap->pages) {
-		printk(KERN_ALERT
-			"%s: invalid bitmap page request: %lu (> %lu)\n",
-			bmname(bitmap), page, bitmap->pages-1);
+		/* This can happen if bitmap_start_sync goes beyond
+		 * End-of-device while looking for a whole page.
+		 * It is harmless.
+		 */
 		return -EINVAL;
 	}
 
-
 	if (bitmap->bp[page].hijacked) /* it's hijacked, don't try to alloc */
 		return 0;
 
@@ -127,47 +69,37 @@ static int bitmap_checkpage(struct bitmap *bitmap, unsigned long page, int creat
 	if (!create)
 		return -ENOENT;
 
-	spin_unlock_irq(&bitmap->lock);
-
 	/* this page has not been allocated yet */
 
-	if ((mappage = bitmap_alloc_page(bitmap)) == NULL) {
-		PRINTK("%s: bitmap map page allocation failed, hijacking\n",
-			bmname(bitmap));
+	spin_unlock_irq(&bitmap->lock);
+	mappage = kzalloc(PAGE_SIZE, GFP_NOIO);
+	spin_lock_irq(&bitmap->lock);
+
+	if (mappage == NULL) {
+		pr_debug("md/bitmap: map page allocation failed, hijacking\n");
 		/* failed - set the hijacked flag so that we can use the
 		 * pointer as a counter */
-		spin_lock_irq(&bitmap->lock);
 		if (!bitmap->bp[page].map)
 			bitmap->bp[page].hijacked = 1;
-		goto out;
-	}
-
-	/* got a page */
-
-	spin_lock_irq(&bitmap->lock);
-
-	/* recheck the page */
-
-	if (bitmap->bp[page].map || bitmap->bp[page].hijacked) {
+	} else if (bitmap->bp[page].map ||
+		   bitmap->bp[page].hijacked) {
 		/* somebody beat us to getting the page */
-		bitmap_free_page(bitmap, mappage);
+		kfree(mappage);
 		return 0;
-	}
+	} else {
 
-	/* no page was in place and we have one, so install it */
+		/* no page was in place and we have one, so install it */
 
-	memset(mappage, 0, PAGE_SIZE);
-	bitmap->bp[page].map = mappage;
-	bitmap->missing_pages--;
-out:
+		bitmap->bp[page].map = mappage;
+		bitmap->missing_pages--;
+	}
 	return 0;
 }
 
-
 /* if page is completely empty, put it back on the free list, or dealloc it */
 /* if page was hijacked, unmark the flag so it might get alloced next time */
 /* Note: lock should be held when calling this */
-static void bitmap_checkfree(struct bitmap *bitmap, unsigned long page)
+static void bitmap_checkfree(struct bitmap_counts *bitmap, unsigned long page)
 {
 	char *ptr;
 
@@ -179,26 +111,15 @@ static void bitmap_checkfree(struct bitmap *bitmap, unsigned long page)
 	if (bitmap->bp[page].hijacked) { /* page was hijacked, undo this now */
 		bitmap->bp[page].hijacked = 0;
 		bitmap->bp[page].map = NULL;
-		return;
+	} else {
+		/* normal case, free the page */
+		ptr = bitmap->bp[page].map;
+		bitmap->bp[page].map = NULL;
+		bitmap->missing_pages++;
+		kfree(ptr);
 	}
-
-	/* normal case, free the page */
-
-#if 0
-/* actually ... let's not.  We will probably need the page again exactly when
- * memory is tight and we are flusing to disk
- */
-	return;
-#else
-	ptr = bitmap->bp[page].map;
-	bitmap->bp[page].map = NULL;
-	bitmap->missing_pages++;
-	bitmap_free_page(bitmap, ptr);
-	return;
-#endif
 }
 
-
 /*
  * bitmap file handling - read and write the bitmap file and its superblock
  */
@@ -208,37 +129,33 @@ static void bitmap_checkfree(struct bitmap *bitmap, unsigned long page)
  */
 
 /* IO operations when bitmap is stored near all superblocks */
-static struct page *read_sb_page(mddev_t *mddev, long offset, unsigned long index)
+static int read_sb_page(struct mddev *mddev, loff_t offset,
+			struct page *page,
+			unsigned long index, int size)
 {
 	/* choose a good rdev and read the page from there */
 
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
-	struct page *page = alloc_page(GFP_KERNEL);
+	struct md_rdev *rdev;
 	sector_t target;
 
-	if (!page)
-		return ERR_PTR(-ENOMEM);
-
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev) {
 		if (! test_bit(In_sync, &rdev->flags)
 		    || test_bit(Faulty, &rdev->flags))
 			continue;
 
-		target = rdev->sb_start + offset + index * (PAGE_SIZE/512);
+		target = offset + index * (PAGE_SIZE/512);
 
-		if (sync_page_io(rdev->bdev, target, PAGE_SIZE, page, READ)) {
+		if (sync_page_io(rdev, target,
+				 roundup(size, bdev_logical_block_size(rdev->bdev)),
+				 page, READ, true)) {
 			page->index = index;
-			attach_page_buffers(page, NULL); /* so that free_buffer will
-							  * quietly no-op */
-			return page;
+			return 0;
 		}
 	}
-	return ERR_PTR(-EIO);
-
+	return -EIO;
 }
 
-static mdk_rdev_t *next_active_rdev(mdk_rdev_t *rdev, mddev_t *mddev)
+static struct md_rdev *next_active_rdev(struct md_rdev *rdev, struct mddev *mddev)
 {
 	/* Iterate the disks of an mddev, using rcu to protect access to the
 	 * linked list, and raising the refcount of devices we return to ensure
@@ -246,22 +163,18 @@ static mdk_rdev_t *next_active_rdev(mdk_rdev_t *rdev, mddev_t *mddev)
 	 * As devices are only added or removed when raid_disk is < 0 and
 	 * nr_pending is 0 and In_sync is clear, the entries we return will
 	 * still be in the same position on the list when we re-enter
-	 * list_for_each_continue_rcu.
+	 * list_for_each_entry_continue_rcu.
 	 */
-	struct list_head *pos;
 	rcu_read_lock();
 	if (rdev == NULL)
 		/* start at the beginning */
-		pos = &mddev->disks;
+		rdev = list_entry_rcu(&mddev->disks, struct md_rdev, same_set);
 	else {
 		/* release the previous rdev and start from there. */
 		rdev_dec_pending(rdev, mddev);
-		pos = &rdev->same_set;
 	}
-	list_for_each_continue_rcu(pos, &mddev->disks) {
-		rdev = list_entry(pos, mdk_rdev_t, same_set);
+	list_for_each_entry_continue_rcu(rdev, &mddev->disks, same_set) {
 		if (rdev->raid_disk >= 0 &&
-		    test_bit(In_sync, &rdev->flags) &&
 		    !test_bit(Faulty, &rdev->flags)) {
 			/* this is a usable devices */
 			atomic_inc(&rdev->nr_pending);
@@ -275,44 +188,64 @@ static mdk_rdev_t *next_active_rdev(mdk_rdev_t *rdev, mddev_t *mddev)
 
 static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
 {
-	mdk_rdev_t *rdev = NULL;
-	mddev_t *mddev = bitmap->mddev;
+	struct md_rdev *rdev = NULL;
+	struct block_device *bdev;
+	struct mddev *mddev = bitmap->mddev;
+	struct bitmap_storage *store = &bitmap->storage;
 
 	while ((rdev = next_active_rdev(rdev, mddev)) != NULL) {
-			int size = PAGE_SIZE;
-			if (page->index == bitmap->file_pages-1)
-				size = roundup(bitmap->last_page_size,
-					       bdev_hardsect_size(rdev->bdev));
-			/* Just make sure we aren't corrupting data or
-			 * metadata
-			 */
-			if (bitmap->offset < 0) {
-				/* DATA  BITMAP METADATA  */
-				if (bitmap->offset
-				    + (long)(page->index * (PAGE_SIZE/512))
-				    + size/512 > 0)
-					/* bitmap runs in to metadata */
-					goto bad_alignment;
-				if (rdev->data_offset + mddev->size*2
-				    > rdev->sb_start + bitmap->offset)
-					/* data runs in to bitmap */
-					goto bad_alignment;
-			} else if (rdev->sb_start < rdev->data_offset) {
-				/* METADATA BITMAP DATA */
-				if (rdev->sb_start
-				    + bitmap->offset
-				    + page->index*(PAGE_SIZE/512) + size/512
-				    > rdev->data_offset)
-					/* bitmap runs in to data */
-					goto bad_alignment;
-			} else {
-				/* DATA METADATA BITMAP - no problems */
-			}
-			md_super_write(mddev, rdev,
-				       rdev->sb_start + bitmap->offset
-				       + page->index * (PAGE_SIZE/512),
-				       size,
-				       page);
+		int size = PAGE_SIZE;
+		loff_t offset = mddev->bitmap_info.offset;
+
+		bdev = (rdev->meta_bdev) ? rdev->meta_bdev : rdev->bdev;
+
+		if (page->index == store->file_pages-1) {
+			int last_page_size = store->bytes & (PAGE_SIZE-1);
+			if (last_page_size == 0)
+				last_page_size = PAGE_SIZE;
+			size = roundup(last_page_size,
+				       bdev_logical_block_size(bdev));
+		}
+		/* Just make sure we aren't corrupting data or
+		 * metadata
+		 */
+		if (mddev->external) {
+			/* Bitmap could be anywhere. */
+			if (rdev->sb_start + offset + (page->index
+						       * (PAGE_SIZE/512))
+			    > rdev->data_offset
+			    &&
+			    rdev->sb_start + offset
+			    < (rdev->data_offset + mddev->dev_sectors
+			     + (PAGE_SIZE/512)))
+				goto bad_alignment;
+		} else if (offset < 0) {
+			/* DATA  BITMAP METADATA  */
+			if (offset
+			    + (long)(page->index * (PAGE_SIZE/512))
+			    + size/512 > 0)
+				/* bitmap runs in to metadata */
+				goto bad_alignment;
+			if (rdev->data_offset + mddev->dev_sectors
+			    > rdev->sb_start + offset)
+				/* data runs in to bitmap */
+				goto bad_alignment;
+		} else if (rdev->sb_start < rdev->data_offset) {
+			/* METADATA BITMAP DATA */
+			if (rdev->sb_start
+			    + offset
+			    + page->index*(PAGE_SIZE/512) + size/512
+			    > rdev->data_offset)
+				/* bitmap runs in to data */
+				goto bad_alignment;
+		} else {
+			/* DATA METADATA BITMAP - no problems */
+		}
+		md_super_write(mddev, rdev,
+			       rdev->sb_start + offset
+			       + page->index * (PAGE_SIZE/512),
+			       size,
+			       page);
 	}
 
 	if (wait)
@@ -320,7 +253,6 @@ static int write_sb_page(struct bitmap *bitmap, struct page *page, int wait)
 	return 0;
 
  bad_alignment:
-	rcu_read_unlock();
 	return -EINVAL;
 }
 
@@ -332,10 +264,10 @@ static void write_page(struct bitmap *bitmap, struct page *page, int wait)
 {
 	struct buffer_head *bh;
 
-	if (bitmap->file == NULL) {
+	if (bitmap->storage.file == NULL) {
 		switch (write_sb_page(bitmap, page, wait)) {
 		case -EINVAL:
-			bitmap->flags |= BITMAP_WRITE_ERROR;
+			set_bit(BITMAP_WRITE_ERROR, &bitmap->flags);
 		}
 	} else {
 
@@ -345,29 +277,24 @@ static void write_page(struct bitmap *bitmap, struct page *page, int wait)
 			atomic_inc(&bitmap->pending_writes);
 			set_buffer_locked(bh);
 			set_buffer_mapped(bh);
-			submit_bh(WRITE, bh);
+			submit_bh(WRITE | REQ_SYNC, bh);
 			bh = bh->b_this_page;
 		}
 
-		if (wait) {
+		if (wait)
 			wait_event(bitmap->write_wait,
 				   atomic_read(&bitmap->pending_writes)==0);
-		}
 	}
-	if (bitmap->flags & BITMAP_WRITE_ERROR)
+	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
 		bitmap_file_kick(bitmap);
 }
 
 static void end_bitmap_write(struct buffer_head *bh, int uptodate)
 {
 	struct bitmap *bitmap = bh->b_private;
-	unsigned long flags;
 
-	if (!uptodate) {
-		spin_lock_irqsave(&bitmap->lock, flags);
-		bitmap->flags |= BITMAP_WRITE_ERROR;
-		spin_unlock_irqrestore(&bitmap->lock, flags);
-	}
+	if (!uptodate)
+		set_bit(BITMAP_WRITE_ERROR, &bitmap->flags);
 	if (atomic_dec_and_test(&bitmap->pending_writes))
 		wake_up(&bitmap->write_wait);
 }
@@ -382,8 +309,12 @@ __clear_page_buffers(struct page *page)
 }
 static void free_buffers(struct page *page)
 {
-	struct buffer_head *bh = page_buffers(page);
+	struct buffer_head *bh;
+
+	if (!PagePrivate(page))
+		return;
 
+	bh = page_buffers(page);
 	while (bh) {
 		struct buffer_head *next = bh->b_this_page;
 		free_buffer_head(bh);
@@ -400,28 +331,22 @@ static void free_buffers(struct page *page)
  * This usage is similar to how swap files are handled, and allows us
  * to write to a file with no concerns of memory allocation failing.
  */
-static struct page *read_page(struct file *file, unsigned long index,
-			      struct bitmap *bitmap,
-			      unsigned long count)
+static int read_page(struct file *file, unsigned long index,
+		     struct bitmap *bitmap,
+		     unsigned long count,
+		     struct page *page)
 {
-	struct page *page = NULL;
-	struct inode *inode = file->f_path.dentry->d_inode;
+	int ret = 0;
+	struct inode *inode = file_inode(file);
 	struct buffer_head *bh;
 	sector_t block;
 
-	PRINTK("read bitmap file (%dB @ %Lu)\n", (int)PAGE_SIZE,
-			(unsigned long long)index << PAGE_SHIFT);
-
-	page = alloc_page(GFP_KERNEL);
-	if (!page)
-		page = ERR_PTR(-ENOMEM);
-	if (IS_ERR(page))
-		goto out;
+	pr_debug("read bitmap file (%dB @ %llu)\n", (int)PAGE_SIZE,
+		 (unsigned long long)index << PAGE_SHIFT);
 
 	bh = alloc_page_buffers(page, 1<<inode->i_blkbits, 0);
 	if (!bh) {
-		put_page(page);
-		page = ERR_PTR(-ENOMEM);
+		ret = -ENOMEM;
 		goto out;
 	}
 	attach_page_buffers(page, bh);
@@ -433,8 +358,7 @@ static struct page *read_page(struct file *file, unsigned long index,
 			bh->b_blocknr = bmap(inode, block);
 			if (bh->b_blocknr == 0) {
 				/* Cannot use this file! */
-				free_buffers(page);
-				page = ERR_PTR(-EINVAL);
+				ret = -EINVAL;
 				goto out;
 			}
 			bh->b_bdev = inode->i_sb->s_bdev;
@@ -457,17 +381,15 @@ static struct page *read_page(struct file *file, unsigned long index,
 
 	wait_event(bitmap->write_wait,
 		   atomic_read(&bitmap->pending_writes)==0);
-	if (bitmap->flags & BITMAP_WRITE_ERROR) {
-		free_buffers(page);
-		page = ERR_PTR(-EIO);
-	}
+	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
+		ret = -EIO;
 out:
-	if (IS_ERR(page))
-		printk(KERN_ALERT "md: bitmap read error: (%dB @ %Lu): %ld\n",
+	if (ret)
+		printk(KERN_ALERT "md: bitmap read error: (%dB @ %llu): %d\n",
 			(int)PAGE_SIZE,
 			(unsigned long long)index << PAGE_SHIFT,
-			PTR_ERR(page));
-	return page;
+			ret);
+	return ret;
 }
 
 /*
@@ -478,25 +400,30 @@ out:
 void bitmap_update_sb(struct bitmap *bitmap)
 {
 	bitmap_super_t *sb;
-	unsigned long flags;
 
 	if (!bitmap || !bitmap->mddev) /* no bitmap for this array */
 		return;
-	spin_lock_irqsave(&bitmap->lock, flags);
-	if (!bitmap->sb_page) { /* no superblock */
-		spin_unlock_irqrestore(&bitmap->lock, flags);
+	if (bitmap->mddev->bitmap_info.external)
 		return;
-	}
-	spin_unlock_irqrestore(&bitmap->lock, flags);
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	if (!bitmap->storage.sb_page) /* no superblock */
+		return;
+	sb = kmap_atomic(bitmap->storage.sb_page);
 	sb->events = cpu_to_le64(bitmap->mddev->events);
-	if (bitmap->mddev->events < bitmap->events_cleared) {
+	if (bitmap->mddev->events < bitmap->events_cleared)
 		/* rocking back to read-only */
 		bitmap->events_cleared = bitmap->mddev->events;
-		sb->events_cleared = cpu_to_le64(bitmap->events_cleared);
-	}
-	kunmap_atomic(sb, KM_USER0);
-	write_page(bitmap, bitmap->sb_page, 1);
+	sb->events_cleared = cpu_to_le64(bitmap->events_cleared);
+	sb->state = cpu_to_le32(bitmap->flags);
+	/* Just in case these have been changed via sysfs: */
+	sb->daemon_sleep = cpu_to_le32(bitmap->mddev->bitmap_info.daemon_sleep/HZ);
+	sb->write_behind = cpu_to_le32(bitmap->mddev->bitmap_info.max_write_behind);
+	/* This might have been changed by a reshape */
+	sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
+	sb->chunksize = cpu_to_le32(bitmap->mddev->bitmap_info.chunksize);
+	sb->sectors_reserved = cpu_to_le32(bitmap->mddev->
+					   bitmap_info.space);
+	kunmap_atomic(sb);
+	write_page(bitmap, bitmap->storage.sb_page, 1);
 }
 
 /* print out the bitmap file superblock */
@@ -504,9 +431,9 @@ void bitmap_print_sb(struct bitmap *bitmap)
 {
 	bitmap_super_t *sb;
 
-	if (!bitmap || !bitmap->sb_page)
+	if (!bitmap || !bitmap->storage.sb_page)
 		return;
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(bitmap->storage.sb_page);
 	printk(KERN_DEBUG "%s: bitmap file superblock:\n", bmname(bitmap));
 	printk(KERN_DEBUG "         magic: %08x\n", le32_to_cpu(sb->magic));
 	printk(KERN_DEBUG "       version: %d\n", le32_to_cpu(sb->version));
@@ -525,7 +452,76 @@ void bitmap_print_sb(struct bitmap *bitmap)
 	printk(KERN_DEBUG "     sync size: %llu KB\n",
 			(unsigned long long)le64_to_cpu(sb->sync_size)/2);
 	printk(KERN_DEBUG "max write behind: %d\n", le32_to_cpu(sb->write_behind));
-	kunmap_atomic(sb, KM_USER0);
+	kunmap_atomic(sb);
+}
+
+/*
+ * bitmap_new_disk_sb
+ * @bitmap
+ *
+ * This function is somewhat the reverse of bitmap_read_sb.  bitmap_read_sb
+ * reads and verifies the on-disk bitmap superblock and populates bitmap_info.
+ * This function verifies 'bitmap_info' and populates the on-disk bitmap
+ * structure, which is to be written to disk.
+ *
+ * Returns: 0 on success, -Exxx on error
+ */
+static int bitmap_new_disk_sb(struct bitmap *bitmap)
+{
+	bitmap_super_t *sb;
+	unsigned long chunksize, daemon_sleep, write_behind;
+
+	bitmap->storage.sb_page = alloc_page(GFP_KERNEL);
+	if (bitmap->storage.sb_page == NULL)
+		return -ENOMEM;
+	bitmap->storage.sb_page->index = 0;
+
+	sb = kmap_atomic(bitmap->storage.sb_page);
+
+	sb->magic = cpu_to_le32(BITMAP_MAGIC);
+	sb->version = cpu_to_le32(BITMAP_MAJOR_HI);
+
+	chunksize = bitmap->mddev->bitmap_info.chunksize;
+	BUG_ON(!chunksize);
+	if (!is_power_of_2(chunksize)) {
+		kunmap_atomic(sb);
+		printk(KERN_ERR "bitmap chunksize not a power of 2\n");
+		return -EINVAL;
+	}
+	sb->chunksize = cpu_to_le32(chunksize);
+
+	daemon_sleep = bitmap->mddev->bitmap_info.daemon_sleep;
+	if (!daemon_sleep ||
+	    (daemon_sleep < 1) || (daemon_sleep > MAX_SCHEDULE_TIMEOUT)) {
+		printk(KERN_INFO "Choosing daemon_sleep default (5 sec)\n");
+		daemon_sleep = 5 * HZ;
+	}
+	sb->daemon_sleep = cpu_to_le32(daemon_sleep);
+	bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
+
+	/*
+	 * FIXME: write_behind for RAID1.  If not specified, what
+	 * is a good choice?  We choose COUNTER_MAX / 2 arbitrarily.
+	 */
+	write_behind = bitmap->mddev->bitmap_info.max_write_behind;
+	if (write_behind > COUNTER_MAX)
+		write_behind = COUNTER_MAX / 2;
+	sb->write_behind = cpu_to_le32(write_behind);
+	bitmap->mddev->bitmap_info.max_write_behind = write_behind;
+
+	/* keep the array size field of the bitmap superblock up to date */
+	sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
+
+	memcpy(sb->uuid, bitmap->mddev->uuid, 16);
+
+	set_bit(BITMAP_STALE, &bitmap->flags);
+	sb->state = cpu_to_le32(bitmap->flags);
+	bitmap->events_cleared = bitmap->mddev->events;
+	sb->events_cleared = cpu_to_le64(bitmap->mddev->events);
+
+	kunmap_atomic(sb);
+
+	return 0;
 }
 
 /* read the superblock from the bitmap file and initialize some bitmap fields */
@@ -535,28 +531,45 @@ static int bitmap_read_sb(struct bitmap *bitmap)
 	bitmap_super_t *sb;
 	unsigned long chunksize, daemon_sleep, write_behind;
 	unsigned long long events;
+	unsigned long sectors_reserved = 0;
 	int err = -EINVAL;
-
+	struct page *sb_page;
+
+	if (!bitmap->storage.file && !bitmap->mddev->bitmap_info.offset) {
+		chunksize = 128 * 1024 * 1024;
+		daemon_sleep = 5 * HZ;
+		write_behind = 0;
+		set_bit(BITMAP_STALE, &bitmap->flags);
+		err = 0;
+		goto out_no_sb;
+	}
 	/* page 0 is the superblock, read it... */
-	if (bitmap->file) {
-		loff_t isize = i_size_read(bitmap->file->f_mapping->host);
+	sb_page = alloc_page(GFP_KERNEL);
+	if (!sb_page)
+		return -ENOMEM;
+	bitmap->storage.sb_page = sb_page;
+
+	if (bitmap->storage.file) {
+		loff_t isize = i_size_read(bitmap->storage.file->f_mapping->host);
 		int bytes = isize > PAGE_SIZE ? PAGE_SIZE : isize;
 
-		bitmap->sb_page = read_page(bitmap->file, 0, bitmap, bytes);
+		err = read_page(bitmap->storage.file, 0,
+				bitmap, bytes, sb_page);
 	} else {
-		bitmap->sb_page = read_sb_page(bitmap->mddev, bitmap->offset, 0);
+		err = read_sb_page(bitmap->mddev,
+				   bitmap->mddev->bitmap_info.offset,
+				   sb_page,
+				   0, sizeof(bitmap_super_t));
 	}
-	if (IS_ERR(bitmap->sb_page)) {
-		err = PTR_ERR(bitmap->sb_page);
-		bitmap->sb_page = NULL;
+	if (err)
 		return err;
-	}
 
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
+	sb = kmap_atomic(sb_page);
 
 	chunksize = le32_to_cpu(sb->chunksize);
-	daemon_sleep = le32_to_cpu(sb->daemon_sleep);
+	daemon_sleep = le32_to_cpu(sb->daemon_sleep) * HZ;
 	write_behind = le32_to_cpu(sb->write_behind);
+	sectors_reserved = le32_to_cpu(sb->sectors_reserved);
 
 	/* verify that the bitmap-specific fields are valid */
 	if (sb->magic != cpu_to_le32(BITMAP_MAGIC))
@@ -564,11 +577,11 @@ static int bitmap_read_sb(struct bitmap *bitmap)
 	else if (le32_to_cpu(sb->version) < BITMAP_MAJOR_LO ||
 		 le32_to_cpu(sb->version) > BITMAP_MAJOR_HI)
 		reason = "unrecognized superblock version";
-	else if (chunksize < PAGE_SIZE)
+	else if (chunksize < 512)
 		reason = "bitmap chunksize too small";
-	else if ((1 << ffz(~chunksize)) != chunksize)
+	else if (!is_power_of_2(chunksize))
 		reason = "bitmap chunksize not a power of 2";
-	else if (daemon_sleep < 1 || daemon_sleep > MAX_SCHEDULE_TIMEOUT / HZ)
+	else if (daemon_sleep < 1 || daemon_sleep > MAX_SCHEDULE_TIMEOUT)
 		reason = "daemon sleep period out of range";
 	else if (write_behind > COUNTER_MAX)
 		reason = "write-behind limit out of range (0 - 16383)";
@@ -581,159 +594,169 @@ static int bitmap_read_sb(struct bitmap *bitmap)
 	/* keep the array size field of the bitmap superblock up to date */
 	sb->sync_size = cpu_to_le64(bitmap->mddev->resync_max_sectors);
 
-	if (!bitmap->mddev->persistent)
-		goto success;
-
-	/*
-	 * if we have a persistent array superblock, compare the
-	 * bitmap's UUID and event counter to the mddev's
-	 */
-	if (memcmp(sb->uuid, bitmap->mddev->uuid, 16)) {
-		printk(KERN_INFO "%s: bitmap superblock UUID mismatch\n",
-			bmname(bitmap));
-		goto out;
-	}
-	events = le64_to_cpu(sb->events);
-	if (events < bitmap->mddev->events) {
-		printk(KERN_INFO "%s: bitmap file is out of date (%llu < %llu) "
-			"-- forcing full recovery\n", bmname(bitmap), events,
-			(unsigned long long) bitmap->mddev->events);
-		sb->state |= cpu_to_le32(BITMAP_STALE);
+	if (bitmap->mddev->persistent) {
+		/*
+		 * We have a persistent array superblock, so compare the
+		 * bitmap's UUID and event counter to the mddev's
+		 */
+		if (memcmp(sb->uuid, bitmap->mddev->uuid, 16)) {
+			printk(KERN_INFO
+			       "%s: bitmap superblock UUID mismatch\n",
+			       bmname(bitmap));
+			goto out;
+		}
+		events = le64_to_cpu(sb->events);
+		if (events < bitmap->mddev->events) {
+			printk(KERN_INFO
+			       "%s: bitmap file is out of date (%llu < %llu) "
+			       "-- forcing full recovery\n",
+			       bmname(bitmap), events,
+			       (unsigned long long) bitmap->mddev->events);
+			set_bit(BITMAP_STALE, &bitmap->flags);
+		}
 	}
-success:
+
 	/* assign fields using values from superblock */
-	bitmap->chunksize = chunksize;
-	bitmap->daemon_sleep = daemon_sleep;
-	bitmap->daemon_lastrun = jiffies;
-	bitmap->max_write_behind = write_behind;
 	bitmap->flags |= le32_to_cpu(sb->state);
 	if (le32_to_cpu(sb->version) == BITMAP_MAJOR_HOSTENDIAN)
-		bitmap->flags |= BITMAP_HOSTENDIAN;
+		set_bit(BITMAP_HOSTENDIAN, &bitmap->flags);
 	bitmap->events_cleared = le64_to_cpu(sb->events_cleared);
-	if (sb->state & cpu_to_le32(BITMAP_STALE))
-		bitmap->events_cleared = bitmap->mddev->events;
 	err = 0;
 out:
-	kunmap_atomic(sb, KM_USER0);
+	kunmap_atomic(sb);
+out_no_sb:
+	if (test_bit(BITMAP_STALE, &bitmap->flags))
+		bitmap->events_cleared = bitmap->mddev->events;
+	bitmap->mddev->bitmap_info.chunksize = chunksize;
+	bitmap->mddev->bitmap_info.daemon_sleep = daemon_sleep;
+	bitmap->mddev->bitmap_info.max_write_behind = write_behind;
+	if (bitmap->mddev->bitmap_info.space == 0 ||
+	    bitmap->mddev->bitmap_info.space > sectors_reserved)
+		bitmap->mddev->bitmap_info.space = sectors_reserved;
 	if (err)
 		bitmap_print_sb(bitmap);
 	return err;
 }
 
-enum bitmap_mask_op {
-	MASK_SET,
-	MASK_UNSET
-};
-
-/* record the state of the bitmap in the superblock.  Return the old value */
-static int bitmap_mask_state(struct bitmap *bitmap, enum bitmap_state bits,
-			     enum bitmap_mask_op op)
-{
-	bitmap_super_t *sb;
-	unsigned long flags;
-	int old;
-
-	spin_lock_irqsave(&bitmap->lock, flags);
-	if (!bitmap->sb_page) { /* can't set the state */
-		spin_unlock_irqrestore(&bitmap->lock, flags);
-		return 0;
-	}
-	spin_unlock_irqrestore(&bitmap->lock, flags);
-	sb = (bitmap_super_t *)kmap_atomic(bitmap->sb_page, KM_USER0);
-	old = le32_to_cpu(sb->state) & bits;
-	switch (op) {
-		case MASK_SET: sb->state |= cpu_to_le32(bits);
-				break;
-		case MASK_UNSET: sb->state &= cpu_to_le32(~bits);
-				break;
-		default: BUG();
-	}
-	kunmap_atomic(sb, KM_USER0);
-	return old;
-}
-
 /*
  * general bitmap file operations
  */
 
+/*
+ * on-disk bitmap:
+ *
+ * Use one bit per "chunk" (block set). We do the disk I/O on the bitmap
+ * file a page at a time. There's a superblock at the start of the file.
+ */
 /* calculate the index of the page that contains this bit */
-static inline unsigned long file_page_index(unsigned long chunk)
+static inline unsigned long file_page_index(struct bitmap_storage *store,
+					    unsigned long chunk)
 {
-	return CHUNK_BIT_OFFSET(chunk) >> PAGE_BIT_SHIFT;
+	if (store->sb_page)
+		chunk += sizeof(bitmap_super_t) << 3;
+	return chunk >> PAGE_BIT_SHIFT;
 }
 
 /* calculate the (bit) offset of this bit within a page */
-static inline unsigned long file_page_offset(unsigned long chunk)
+static inline unsigned long file_page_offset(struct bitmap_storage *store,
+					     unsigned long chunk)
 {
-	return CHUNK_BIT_OFFSET(chunk) & (PAGE_BITS - 1);
+	if (store->sb_page)
+		chunk += sizeof(bitmap_super_t) << 3;
+	return chunk & (PAGE_BITS - 1);
 }
 
 /*
  * return a pointer to the page in the filemap that contains the given bit
  *
- * this lookup is complicated by the fact that the bitmap sb might be exactly
- * 1 page (e.g., x86) or less than 1 page -- so the bitmap might start on page
- * 0 or page 1
  */
-static inline struct page *filemap_get_page(struct bitmap *bitmap,
-					unsigned long chunk)
+static inline struct page *filemap_get_page(struct bitmap_storage *store,
+					    unsigned long chunk)
 {
-	if (file_page_index(chunk) >= bitmap->file_pages) return NULL;
-	return bitmap->filemap[file_page_index(chunk) - file_page_index(0)];
+	if (file_page_index(store, chunk) >= store->file_pages)
+		return NULL;
+	return store->filemap[file_page_index(store, chunk)];
 }
 
+static int bitmap_storage_alloc(struct bitmap_storage *store,
+				unsigned long chunks, int with_super)
+{
+	int pnum;
+	unsigned long num_pages;
+	unsigned long bytes;
+
+	bytes = DIV_ROUND_UP(chunks, 8);
+	if (with_super)
+		bytes += sizeof(bitmap_super_t);
+
+	num_pages = DIV_ROUND_UP(bytes, PAGE_SIZE);
+
+	store->filemap = kmalloc(sizeof(struct page *)
+				 * num_pages, GFP_KERNEL);
+	if (!store->filemap)
+		return -ENOMEM;
+
+	if (with_super && !store->sb_page) {
+		store->sb_page = alloc_page(GFP_KERNEL|__GFP_ZERO);
+		if (store->sb_page == NULL)
+			return -ENOMEM;
+		store->sb_page->index = 0;
+	}
+	pnum = 0;
+	if (store->sb_page) {
+		store->filemap[0] = store->sb_page;
+		pnum = 1;
+	}
+	for ( ; pnum < num_pages; pnum++) {
+		store->filemap[pnum] = alloc_page(GFP_KERNEL|__GFP_ZERO);
+		if (!store->filemap[pnum]) {
+			store->file_pages = pnum;
+			return -ENOMEM;
+		}
+		store->filemap[pnum]->index = pnum;
+	}
+	store->file_pages = pnum;
+
+	/* We need 4 bits per page, rounded up to a multiple
+	 * of sizeof(unsigned long) */
+	store->filemap_attr = kzalloc(
+		roundup(DIV_ROUND_UP(num_pages*4, 8), sizeof(unsigned long)),
+		GFP_KERNEL);
+	if (!store->filemap_attr)
+		return -ENOMEM;
+
+	store->bytes = bytes;
+
+	return 0;
+}
 
-static void bitmap_file_unmap(struct bitmap *bitmap)
+static void bitmap_file_unmap(struct bitmap_storage *store)
 {
 	struct page **map, *sb_page;
-	unsigned long *attr;
 	int pages;
-	unsigned long flags;
+	struct file *file;
 
-	spin_lock_irqsave(&bitmap->lock, flags);
-	map = bitmap->filemap;
-	bitmap->filemap = NULL;
-	attr = bitmap->filemap_attr;
-	bitmap->filemap_attr = NULL;
-	pages = bitmap->file_pages;
-	bitmap->file_pages = 0;
-	sb_page = bitmap->sb_page;
-	bitmap->sb_page = NULL;
-	spin_unlock_irqrestore(&bitmap->lock, flags);
+	file = store->file;
+	map = store->filemap;
+	pages = store->file_pages;
+	sb_page = store->sb_page;
 
 	while (pages--)
-		if (map[pages]->index != 0) /* 0 is sb_page, release it below */
+		if (map[pages] != sb_page) /* 0 is sb_page, release it below */
 			free_buffers(map[pages]);
 	kfree(map);
-	kfree(attr);
+	kfree(store->filemap_attr);
 
 	if (sb_page)
 		free_buffers(sb_page);
-}
-
-static void bitmap_file_put(struct bitmap *bitmap)
-{
-	struct file *file;
-	unsigned long flags;
-
-	spin_lock_irqsave(&bitmap->lock, flags);
-	file = bitmap->file;
-	bitmap->file = NULL;
-	spin_unlock_irqrestore(&bitmap->lock, flags);
-
-	if (file)
-		wait_event(bitmap->write_wait,
-			   atomic_read(&bitmap->pending_writes)==0);
-	bitmap_file_unmap(bitmap);
 
 	if (file) {
-		struct inode *inode = file->f_path.dentry->d_inode;
+		struct inode *inode = file_inode(file);
 		invalidate_mapping_pages(inode->i_mapping, 0, -1);
 		fput(file);
 	}
 }
 
-
 /*
  * bitmap_file_kick - if an error occurs while manipulating the bitmap file
  * then it is no longer reliable, so we stop using it and we mark the file
@@ -743,15 +766,14 @@ static void bitmap_file_kick(struct bitmap *bitmap)
 {
 	char *path, *ptr = NULL;
 
-	if (bitmap_mask_state(bitmap, BITMAP_STALE, MASK_SET) == 0) {
+	if (!test_and_set_bit(BITMAP_STALE, &bitmap->flags)) {
 		bitmap_update_sb(bitmap);
 
-		if (bitmap->file) {
+		if (bitmap->storage.file) {
 			path = kmalloc(PAGE_SIZE, GFP_KERNEL);
 			if (path)
-				ptr = d_path(&bitmap->file->f_path, path,
-					     PAGE_SIZE);
-
+				ptr = d_path(&bitmap->storage.file->f_path,
+					     path, PAGE_SIZE);
 
 			printk(KERN_ALERT
 			      "%s: kicking failed bitmap file %s from array!\n",
@@ -763,36 +785,39 @@ static void bitmap_file_kick(struct bitmap *bitmap)
 			       "%s: disabling internal bitmap due to errors\n",
 			       bmname(bitmap));
 	}
-
-	bitmap_file_put(bitmap);
-
-	return;
 }
 
 enum bitmap_page_attr {
-	BITMAP_PAGE_DIRTY = 0, // there are set bits that need to be synced
-	BITMAP_PAGE_CLEAN = 1, // there are bits that might need to be cleared
-	BITMAP_PAGE_NEEDWRITE=2, // there are cleared bits that need to be synced
+	BITMAP_PAGE_DIRTY = 0,     /* there are set bits that need to be synced */
+	BITMAP_PAGE_PENDING = 1,   /* there are bits that are being cleaned.
+				    * i.e. counter is 1 or 2. */
+	BITMAP_PAGE_NEEDWRITE = 2, /* there are cleared bits that need to be synced */
 };
 
-static inline void set_page_attr(struct bitmap *bitmap, struct page *page,
-				enum bitmap_page_attr attr)
+static inline void set_page_attr(struct bitmap *bitmap, int pnum,
+				 enum bitmap_page_attr attr)
 {
-	__set_bit((page->index<<2) + attr, bitmap->filemap_attr);
+	set_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
 }
 
-static inline void clear_page_attr(struct bitmap *bitmap, struct page *page,
-				enum bitmap_page_attr attr)
+static inline void clear_page_attr(struct bitmap *bitmap, int pnum,
+				   enum bitmap_page_attr attr)
 {
-	__clear_bit((page->index<<2) + attr, bitmap->filemap_attr);
+	clear_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
 }
 
-static inline unsigned long test_page_attr(struct bitmap *bitmap, struct page *page,
-					   enum bitmap_page_attr attr)
+static inline int test_page_attr(struct bitmap *bitmap, int pnum,
+				 enum bitmap_page_attr attr)
 {
-	return test_bit((page->index<<2) + attr, bitmap->filemap_attr);
+	return test_bit((pnum<<2) + attr, bitmap->storage.filemap_attr);
 }
 
+static inline int test_and_clear_page_attr(struct bitmap *bitmap, int pnum,
+					   enum bitmap_page_attr attr)
+{
+	return test_and_clear_bit((pnum<<2) + attr,
+				  bitmap->storage.filemap_attr);
+}
 /*
  * bitmap_file_set_bit -- called before performing a write to the md device
  * to set (and eventually sync) a particular bit in the bitmap file
@@ -805,28 +830,46 @@ static void bitmap_file_set_bit(struct bitmap *bitmap, sector_t block)
 	unsigned long bit;
 	struct page *page;
 	void *kaddr;
-	unsigned long chunk = block >> CHUNK_BLOCK_SHIFT(bitmap);
+	unsigned long chunk = block >> bitmap->counts.chunkshift;
 
-	if (!bitmap->filemap) {
+	page = filemap_get_page(&bitmap->storage, chunk);
+	if (!page)
 		return;
-	}
+	bit = file_page_offset(&bitmap->storage, chunk);
 
-	page = filemap_get_page(bitmap, chunk);
-	if (!page) return;
-	bit = file_page_offset(chunk);
-
- 	/* set the bit */
-	kaddr = kmap_atomic(page, KM_USER0);
-	if (bitmap->flags & BITMAP_HOSTENDIAN)
+	/* set the bit */
+	kaddr = kmap_atomic(page);
+	if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
 		set_bit(bit, kaddr);
 	else
-		ext2_set_bit(bit, kaddr);
-	kunmap_atomic(kaddr, KM_USER0);
-	PRINTK("set file bit %lu page %lu\n", bit, page->index);
-
+		set_bit_le(bit, kaddr);
+	kunmap_atomic(kaddr);
+	pr_debug("set file bit %lu page %lu\n", bit, page->index);
 	/* record page number so it gets flushed to disk when unplug occurs */
-	set_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
+	set_page_attr(bitmap, page->index, BITMAP_PAGE_DIRTY);
+}
 
+static void bitmap_file_clear_bit(struct bitmap *bitmap, sector_t block)
+{
+	unsigned long bit;
+	struct page *page;
+	void *paddr;
+	unsigned long chunk = block >> bitmap->counts.chunkshift;
+
+	page = filemap_get_page(&bitmap->storage, chunk);
+	if (!page)
+		return;
+	bit = file_page_offset(&bitmap->storage, chunk);
+	paddr = kmap_atomic(page);
+	if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
+		clear_bit(bit, paddr);
+	else
+		clear_bit_le(bit, paddr);
+	kunmap_atomic(paddr);
+	if (!test_page_attr(bitmap, page->index, BITMAP_PAGE_NEEDWRITE)) {
+		set_page_attr(bitmap, page->index, BITMAP_PAGE_PENDING);
+		bitmap->allclean = 0;
+	}
 }
 
 /* this gets called when the md device is ready to unplug its underlying
@@ -834,44 +877,40 @@ static void bitmap_file_set_bit(struct bitmap *bitmap, sector_t block)
  * sync the dirty pages of the bitmap file to disk */
 void bitmap_unplug(struct bitmap *bitmap)
 {
-	unsigned long i, flags;
+	unsigned long i;
 	int dirty, need_write;
-	struct page *page;
 	int wait = 0;
 
-	if (!bitmap)
+	if (!bitmap || !bitmap->storage.filemap ||
+	    test_bit(BITMAP_STALE, &bitmap->flags))
 		return;
 
 	/* look at each page to see if there are any set bits that need to be
 	 * flushed out to disk */
-	for (i = 0; i < bitmap->file_pages; i++) {
-		spin_lock_irqsave(&bitmap->lock, flags);
-		if (!bitmap->filemap) {
-			spin_unlock_irqrestore(&bitmap->lock, flags);
+	for (i = 0; i < bitmap->storage.file_pages; i++) {
+		if (!bitmap->storage.filemap)
 			return;
+		dirty = test_and_clear_page_attr(bitmap, i, BITMAP_PAGE_DIRTY);
+		need_write = test_and_clear_page_attr(bitmap, i,
+						      BITMAP_PAGE_NEEDWRITE);
+		if (dirty || need_write) {
+			clear_page_attr(bitmap, i, BITMAP_PAGE_PENDING);
+			write_page(bitmap, bitmap->storage.filemap[i], 0);
 		}
-		page = bitmap->filemap[i];
-		dirty = test_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
-		need_write = test_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
-		clear_page_attr(bitmap, page, BITMAP_PAGE_DIRTY);
-		clear_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
 		if (dirty)
 			wait = 1;
-		spin_unlock_irqrestore(&bitmap->lock, flags);
-
-		if (dirty | need_write)
-			write_page(bitmap, page, 0);
 	}
 	if (wait) { /* if any writes were performed, we need to wait on them */
-		if (bitmap->file)
+		if (bitmap->storage.file)
 			wait_event(bitmap->write_wait,
 				   atomic_read(&bitmap->pending_writes)==0);
 		else
 			md_super_wait(bitmap->mddev);
 	}
-	if (bitmap->flags & BITMAP_WRITE_ERROR)
+	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
 		bitmap_file_kick(bitmap);
 }
+EXPORT_SYMBOL(bitmap_unplug);
 
 static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed);
 /* * bitmap_init_from_disk -- called at bitmap_create time to initialize
@@ -888,140 +927,117 @@ static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int n
 static int bitmap_init_from_disk(struct bitmap *bitmap, sector_t start)
 {
 	unsigned long i, chunks, index, oldindex, bit;
-	struct page *page = NULL, *oldpage = NULL;
-	unsigned long num_pages, bit_cnt = 0;
+	struct page *page = NULL;
+	unsigned long bit_cnt = 0;
 	struct file *file;
-	unsigned long bytes, offset;
+	unsigned long offset;
 	int outofdate;
 	int ret = -ENOSPC;
 	void *paddr;
+	struct bitmap_storage *store = &bitmap->storage;
 
-	chunks = bitmap->chunks;
-	file = bitmap->file;
+	chunks = bitmap->counts.chunks;
+	file = store->file;
 
-	BUG_ON(!file && !bitmap->offset);
+	if (!file && !bitmap->mddev->bitmap_info.offset) {
+		/* No permanent bitmap - fill with '1s'. */
+		store->filemap = NULL;
+		store->file_pages = 0;
+		for (i = 0; i < chunks ; i++) {
+			/* if the disk bit is set, set the memory bit */
+			int needed = ((sector_t)(i+1) << (bitmap->counts.chunkshift)
+				      >= start);
+			bitmap_set_memory_bits(bitmap,
+					       (sector_t)i << bitmap->counts.chunkshift,
+					       needed);
+		}
+		return 0;
+	}
 
-#ifdef INJECT_FAULTS_3
-	outofdate = 1;
-#else
-	outofdate = bitmap->flags & BITMAP_STALE;
-#endif
+	outofdate = test_bit(BITMAP_STALE, &bitmap->flags);
 	if (outofdate)
 		printk(KERN_INFO "%s: bitmap file is out of date, doing full "
 			"recovery\n", bmname(bitmap));
 
-	bytes = (chunks + 7) / 8;
-
-	num_pages = (bytes + sizeof(bitmap_super_t) + PAGE_SIZE - 1) / PAGE_SIZE;
-
-	if (file && i_size_read(file->f_mapping->host) < bytes + sizeof(bitmap_super_t)) {
+	if (file && i_size_read(file->f_mapping->host) < store->bytes) {
 		printk(KERN_INFO "%s: bitmap file too short %lu < %lu\n",
-			bmname(bitmap),
-			(unsigned long) i_size_read(file->f_mapping->host),
-			bytes + sizeof(bitmap_super_t));
+		       bmname(bitmap),
+		       (unsigned long) i_size_read(file->f_mapping->host),
+		       store->bytes);
 		goto err;
 	}
 
-	ret = -ENOMEM;
-
-	bitmap->filemap = kmalloc(sizeof(struct page *) * num_pages, GFP_KERNEL);
-	if (!bitmap->filemap)
-		goto err;
-
-	/* We need 4 bits per page, rounded up to a multiple of sizeof(unsigned long) */
-	bitmap->filemap_attr = kzalloc(
-		roundup( DIV_ROUND_UP(num_pages*4, 8), sizeof(unsigned long)),
-		GFP_KERNEL);
-	if (!bitmap->filemap_attr)
-		goto err;
-
 	oldindex = ~0L;
+	offset = 0;
+	if (!bitmap->mddev->bitmap_info.external)
+		offset = sizeof(bitmap_super_t);
 
 	for (i = 0; i < chunks; i++) {
 		int b;
-		index = file_page_index(i);
-		bit = file_page_offset(i);
+		index = file_page_index(&bitmap->storage, i);
+		bit = file_page_offset(&bitmap->storage, i);
 		if (index != oldindex) { /* this is a new page, read it in */
 			int count;
 			/* unmap the old page, we're done with it */
-			if (index == num_pages-1)
-				count = bytes + sizeof(bitmap_super_t)
-					- index * PAGE_SIZE;
+			if (index == store->file_pages-1)
+				count = store->bytes - index * PAGE_SIZE;
 			else
 				count = PAGE_SIZE;
-			if (index == 0) {
-				/*
-				 * if we're here then the superblock page
-				 * contains some bits (PAGE_SIZE != sizeof sb)
-				 * we've already read it in, so just use it
-				 */
-				page = bitmap->sb_page;
-				offset = sizeof(bitmap_super_t);
-			} else if (file) {
-				page = read_page(file, index, bitmap, count);
-				offset = 0;
-			} else {
-				page = read_sb_page(bitmap->mddev, bitmap->offset, index);
-				offset = 0;
-			}
-			if (IS_ERR(page)) { /* read error */
-				ret = PTR_ERR(page);
+			page = store->filemap[index];
+			if (file)
+				ret = read_page(file, index, bitmap,
+						count, page);
+			else
+				ret = read_sb_page(
+					bitmap->mddev,
+					bitmap->mddev->bitmap_info.offset,
+					page,
+					index, count);
+
+			if (ret)
 				goto err;
-			}
 
 			oldindex = index;
-			oldpage = page;
 
 			if (outofdate) {
 				/*
 				 * if bitmap is out of date, dirty the
-			 	 * whole page and write it out
+				 * whole page and write it out
 				 */
-				paddr = kmap_atomic(page, KM_USER0);
+				paddr = kmap_atomic(page);
 				memset(paddr + offset, 0xff,
 				       PAGE_SIZE - offset);
-				kunmap_atomic(paddr, KM_USER0);
+				kunmap_atomic(paddr);
 				write_page(bitmap, page, 1);
 
 				ret = -EIO;
-				if (bitmap->flags & BITMAP_WRITE_ERROR) {
-					/* release, page not in filemap yet */
-					put_page(page);
+				if (test_bit(BITMAP_WRITE_ERROR,
+					     &bitmap->flags))
 					goto err;
-				}
 			}
-
-			bitmap->filemap[bitmap->file_pages++] = page;
-			bitmap->last_page_size = count;
 		}
-		paddr = kmap_atomic(page, KM_USER0);
-		if (bitmap->flags & BITMAP_HOSTENDIAN)
+		paddr = kmap_atomic(page);
+		if (test_bit(BITMAP_HOSTENDIAN, &bitmap->flags))
 			b = test_bit(bit, paddr);
 		else
-			b = ext2_test_bit(bit, paddr);
-		kunmap_atomic(paddr, KM_USER0);
+			b = test_bit_le(bit, paddr);
+		kunmap_atomic(paddr);
 		if (b) {
 			/* if the disk bit is set, set the memory bit */
-			bitmap_set_memory_bits(bitmap, i << CHUNK_BLOCK_SHIFT(bitmap),
-					       ((i+1) << (CHUNK_BLOCK_SHIFT(bitmap)) >= start)
-				);
+			int needed = ((sector_t)(i+1) << bitmap->counts.chunkshift
+				      >= start);
+			bitmap_set_memory_bits(bitmap,
+					       (sector_t)i << bitmap->counts.chunkshift,
+					       needed);
 			bit_cnt++;
-			set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
 		}
-	}
-
- 	/* everything went OK */
-	ret = 0;
-	bitmap_mask_state(bitmap, BITMAP_STALE, MASK_UNSET);
-
-	if (bit_cnt) { /* Kick recovery if any bits were set */
-		set_bit(MD_RECOVERY_NEEDED, &bitmap->mddev->recovery);
-		md_wakeup_thread(bitmap->mddev->thread);
+		offset = 0;
 	}
 
 	printk(KERN_INFO "%s: bitmap initialized from disk: "
-		"read %lu/%lu pages, set %lu bits\n",
-		bmname(bitmap), bitmap->file_pages, num_pages, bit_cnt);
+	       "read %lu pages, set %lu of %lu bits\n",
+	       bmname(bitmap), store->file_pages,
+	       bit_cnt, chunks);
 
 	return 0;
 
@@ -1038,25 +1054,39 @@ void bitmap_write_all(struct bitmap *bitmap)
 	 */
 	int i;
 
-	for (i=0; i < bitmap->file_pages; i++)
-		set_page_attr(bitmap, bitmap->filemap[i],
+	if (!bitmap || !bitmap->storage.filemap)
+		return;
+	if (bitmap->storage.file)
+		/* Only one copy, so nothing needed */
+		return;
+
+	for (i = 0; i < bitmap->storage.file_pages; i++)
+		set_page_attr(bitmap, i,
 			      BITMAP_PAGE_NEEDWRITE);
+	bitmap->allclean = 0;
 }
 
-
-static void bitmap_count_page(struct bitmap *bitmap, sector_t offset, int inc)
+static void bitmap_count_page(struct bitmap_counts *bitmap,
+			      sector_t offset, int inc)
 {
-	sector_t chunk = offset >> CHUNK_BLOCK_SHIFT(bitmap);
+	sector_t chunk = offset >> bitmap->chunkshift;
 	unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
 	bitmap->bp[page].count += inc;
-/*
-	if (page == 0) printk("count page 0, offset %llu: %d gives %d\n",
-			      (unsigned long long)offset, inc, bitmap->bp[page].count);
-*/
 	bitmap_checkfree(bitmap, page);
 }
-static bitmap_counter_t *bitmap_get_counter(struct bitmap *bitmap,
-					    sector_t offset, int *blocks,
+
+static void bitmap_set_pending(struct bitmap_counts *bitmap, sector_t offset)
+{
+	sector_t chunk = offset >> bitmap->chunkshift;
+	unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
+	struct bitmap_page *bp = &bitmap->bp[page];
+
+	if (!bp->pending)
+		bp->pending = 1;
+}
+
+static bitmap_counter_t *bitmap_get_counter(struct bitmap_counts *bitmap,
+					    sector_t offset, sector_t *blocks,
 					    int create);
 
 /*
@@ -1064,189 +1094,198 @@ static bitmap_counter_t *bitmap_get_counter(struct bitmap *bitmap,
  *			out to disk
  */
 
-void bitmap_daemon_work(struct bitmap *bitmap)
+void bitmap_daemon_work(struct mddev *mddev)
 {
+	struct bitmap *bitmap;
 	unsigned long j;
-	unsigned long flags;
-	struct page *page = NULL, *lastpage = NULL;
-	int blocks;
-	void *paddr;
+	unsigned long nextpage;
+	sector_t blocks;
+	struct bitmap_counts *counts;
 
-	if (bitmap == NULL)
+	/* Use a mutex to guard daemon_work against
+	 * bitmap_destroy.
+	 */
+	mutex_lock(&mddev->bitmap_info.mutex);
+	bitmap = mddev->bitmap;
+	if (bitmap == NULL) {
+		mutex_unlock(&mddev->bitmap_info.mutex);
 		return;
-	if (time_before(jiffies, bitmap->daemon_lastrun + bitmap->daemon_sleep*HZ))
+	}
+	if (time_before(jiffies, bitmap->daemon_lastrun
+			+ mddev->bitmap_info.daemon_sleep))
 		goto done;
 
 	bitmap->daemon_lastrun = jiffies;
 	if (bitmap->allclean) {
-		bitmap->mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
-		return;
+		mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
+		goto done;
 	}
 	bitmap->allclean = 1;
 
-	for (j = 0; j < bitmap->chunks; j++) {
-		bitmap_counter_t *bmc;
-		spin_lock_irqsave(&bitmap->lock, flags);
-		if (!bitmap->filemap) {
-			/* error or shutdown */
-			spin_unlock_irqrestore(&bitmap->lock, flags);
-			break;
+	/* Any file-page which is PENDING now needs to be written.
+	 * So set NEEDWRITE now, then after we make any last-minute changes
+	 * we will write it.
+	 */
+	for (j = 0; j < bitmap->storage.file_pages; j++)
+		if (test_and_clear_page_attr(bitmap, j,
+					     BITMAP_PAGE_PENDING))
+			set_page_attr(bitmap, j,
+				      BITMAP_PAGE_NEEDWRITE);
+
+	if (bitmap->need_sync &&
+	    mddev->bitmap_info.external == 0) {
+		/* Arrange for superblock update as well as
+		 * other changes */
+		bitmap_super_t *sb;
+		bitmap->need_sync = 0;
+		if (bitmap->storage.filemap) {
+			sb = kmap_atomic(bitmap->storage.sb_page);
+			sb->events_cleared =
+				cpu_to_le64(bitmap->events_cleared);
+			kunmap_atomic(sb);
+			set_page_attr(bitmap, 0,
+				      BITMAP_PAGE_NEEDWRITE);
 		}
+	}
+	/* Now look at the bitmap counters and if any are '2' or '1',
+	 * decrement and handle accordingly.
+	 */
+	counts = &bitmap->counts;
+	spin_lock_irq(&counts->lock);
+	nextpage = 0;
+	for (j = 0; j < counts->chunks; j++) {
+		bitmap_counter_t *bmc;
+		sector_t  block = (sector_t)j << counts->chunkshift;
 
-		page = filemap_get_page(bitmap, j);
-
-		if (page != lastpage) {
-			/* skip this page unless it's marked as needing cleaning */
-			if (!test_page_attr(bitmap, page, BITMAP_PAGE_CLEAN)) {
-				int need_write = test_page_attr(bitmap, page,
-								BITMAP_PAGE_NEEDWRITE);
-				if (need_write)
-					clear_page_attr(bitmap, page, BITMAP_PAGE_NEEDWRITE);
-
-				spin_unlock_irqrestore(&bitmap->lock, flags);
-				if (need_write) {
-					write_page(bitmap, page, 0);
-					bitmap->allclean = 0;
-				}
+		if (j == nextpage) {
+			nextpage += PAGE_COUNTER_RATIO;
+			if (!counts->bp[j >> PAGE_COUNTER_SHIFT].pending) {
+				j |= PAGE_COUNTER_MASK;
 				continue;
 			}
-
-			/* grab the new page, sync and release the old */
-			if (lastpage != NULL) {
-				if (test_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE)) {
-					clear_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
-					spin_unlock_irqrestore(&bitmap->lock, flags);
-					write_page(bitmap, lastpage, 0);
-				} else {
-					set_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
-					spin_unlock_irqrestore(&bitmap->lock, flags);
-				}
-			} else
-				spin_unlock_irqrestore(&bitmap->lock, flags);
-			lastpage = page;
-
-			/* We are possibly going to clear some bits, so make
-			 * sure that events_cleared is up-to-date.
-			 */
-			if (bitmap->need_sync) {
-				bitmap_super_t *sb;
-				bitmap->need_sync = 0;
-				sb = kmap_atomic(bitmap->sb_page, KM_USER0);
-				sb->events_cleared =
-					cpu_to_le64(bitmap->events_cleared);
-				kunmap_atomic(sb, KM_USER0);
-				write_page(bitmap, bitmap->sb_page, 1);
-			}
-			spin_lock_irqsave(&bitmap->lock, flags);
-			clear_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
+			counts->bp[j >> PAGE_COUNTER_SHIFT].pending = 0;
 		}
-		bmc = bitmap_get_counter(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
-					&blocks, 0);
-		if (bmc) {
-/*
-  if (j < 100) printk("bitmap: j=%lu, *bmc = 0x%x\n", j, *bmc);
-*/
-			if (*bmc)
-				bitmap->allclean = 0;
+		bmc = bitmap_get_counter(counts,
+					 block,
+					 &blocks, 0);
 
-			if (*bmc == 2) {
-				*bmc=1; /* maybe clear the bit next time */
-				set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
-			} else if (*bmc == 1) {
-				/* we can clear the bit */
-				*bmc = 0;
-				bitmap_count_page(bitmap, j << CHUNK_BLOCK_SHIFT(bitmap),
-						  -1);
-
-				/* clear the bit */
-				paddr = kmap_atomic(page, KM_USER0);
-				if (bitmap->flags & BITMAP_HOSTENDIAN)
-					clear_bit(file_page_offset(j), paddr);
-				else
-					ext2_clear_bit(file_page_offset(j), paddr);
-				kunmap_atomic(paddr, KM_USER0);
-			}
+		if (!bmc) {
+			j |= PAGE_COUNTER_MASK;
+			continue;
+		}
+		if (*bmc == 1 && !bitmap->need_sync) {
+			/* We can clear the bit */
+			*bmc = 0;
+			bitmap_count_page(counts, block, -1);
+			bitmap_file_clear_bit(bitmap, block);
+		} else if (*bmc && *bmc <= 2) {
+			*bmc = 1;
+			bitmap_set_pending(counts, block);
+			bitmap->allclean = 0;
 		}
-		spin_unlock_irqrestore(&bitmap->lock, flags);
 	}
-
-	/* now sync the final page */
-	if (lastpage != NULL) {
-		spin_lock_irqsave(&bitmap->lock, flags);
-		if (test_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE)) {
-			clear_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
-			spin_unlock_irqrestore(&bitmap->lock, flags);
-			write_page(bitmap, lastpage, 0);
-		} else {
-			set_page_attr(bitmap, lastpage, BITMAP_PAGE_NEEDWRITE);
-			spin_unlock_irqrestore(&bitmap->lock, flags);
+	spin_unlock_irq(&counts->lock);
+
+	/* Now start writeout on any page in NEEDWRITE that isn't DIRTY.
+	 * DIRTY pages need to be written by bitmap_unplug so it can wait
+	 * for them.
+	 * If we find any DIRTY page we stop there and let bitmap_unplug
+	 * handle all the rest.  This is important in the case where
+	 * the first blocking holds the superblock and it has been updated.
+	 * We mustn't write any other blocks before the superblock.
+	 */
+	for (j = 0;
+	     j < bitmap->storage.file_pages
+		     && !test_bit(BITMAP_STALE, &bitmap->flags);
+	     j++) {
+
+		if (test_page_attr(bitmap, j,
+				   BITMAP_PAGE_DIRTY))
+			/* bitmap_unplug will handle the rest */
+			break;
+		if (test_and_clear_page_attr(bitmap, j,
+					     BITMAP_PAGE_NEEDWRITE)) {
+			write_page(bitmap, bitmap->storage.filemap[j], 0);
 		}
 	}
 
  done:
 	if (bitmap->allclean == 0)
-		bitmap->mddev->thread->timeout = bitmap->daemon_sleep * HZ;
+		mddev->thread->timeout =
+			mddev->bitmap_info.daemon_sleep;
+	mutex_unlock(&mddev->bitmap_info.mutex);
 }
 
-static bitmap_counter_t *bitmap_get_counter(struct bitmap *bitmap,
-					    sector_t offset, int *blocks,
+static bitmap_counter_t *bitmap_get_counter(struct bitmap_counts *bitmap,
+					    sector_t offset, sector_t *blocks,
 					    int create)
+__releases(bitmap->lock)
+__acquires(bitmap->lock)
 {
 	/* If 'create', we might release the lock and reclaim it.
 	 * The lock must have been taken with interrupts enabled.
 	 * If !create, we don't release the lock.
 	 */
-	sector_t chunk = offset >> CHUNK_BLOCK_SHIFT(bitmap);
+	sector_t chunk = offset >> bitmap->chunkshift;
 	unsigned long page = chunk >> PAGE_COUNTER_SHIFT;
 	unsigned long pageoff = (chunk & PAGE_COUNTER_MASK) << COUNTER_BYTE_SHIFT;
 	sector_t csize;
+	int err;
 
-	if (bitmap_checkpage(bitmap, page, create) < 0) {
-		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap));
-		*blocks = csize - (offset & (csize- 1));
+	err = bitmap_checkpage(bitmap, page, create);
+
+	if (bitmap->bp[page].hijacked ||
+	    bitmap->bp[page].map == NULL)
+		csize = ((sector_t)1) << (bitmap->chunkshift +
+					  PAGE_COUNTER_SHIFT - 1);
+	else
+		csize = ((sector_t)1) << bitmap->chunkshift;
+	*blocks = csize - (offset & (csize - 1));
+
+	if (err < 0)
 		return NULL;
-	}
+
 	/* now locked ... */
 
 	if (bitmap->bp[page].hijacked) { /* hijacked pointer */
 		/* should we use the first or second counter field
 		 * of the hijacked pointer? */
 		int hi = (pageoff > PAGE_COUNTER_MASK);
-		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap) +
-					  PAGE_COUNTER_SHIFT - 1);
-		*blocks = csize - (offset & (csize- 1));
 		return  &((bitmap_counter_t *)
 			  &bitmap->bp[page].map)[hi];
-	} else { /* page is allocated */
-		csize = ((sector_t)1) << (CHUNK_BLOCK_SHIFT(bitmap));
-		*blocks = csize - (offset & (csize- 1));
+	} else /* page is allocated */
 		return (bitmap_counter_t *)
 			&(bitmap->bp[page].map[pageoff]);
-	}
 }
 
 int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors, int behind)
 {
-	if (!bitmap) return 0;
+	if (!bitmap)
+		return 0;
 
 	if (behind) {
+		int bw;
 		atomic_inc(&bitmap->behind_writes);
-		PRINTK(KERN_DEBUG "inc write-behind count %d/%d\n",
-		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
+		bw = atomic_read(&bitmap->behind_writes);
+		if (bw > bitmap->behind_writes_used)
+			bitmap->behind_writes_used = bw;
+
+		pr_debug("inc write-behind count %d/%lu\n",
+			 bw, bitmap->mddev->bitmap_info.max_write_behind);
 	}
 
 	while (sectors) {
-		int blocks;
+		sector_t blocks;
 		bitmap_counter_t *bmc;
 
-		spin_lock_irq(&bitmap->lock);
-		bmc = bitmap_get_counter(bitmap, offset, &blocks, 1);
+		spin_lock_irq(&bitmap->counts.lock);
+		bmc = bitmap_get_counter(&bitmap->counts, offset, &blocks, 1);
 		if (!bmc) {
-			spin_unlock_irq(&bitmap->lock);
+			spin_unlock_irq(&bitmap->counts.lock);
 			return 0;
 		}
 
-		if (unlikely((*bmc & COUNTER_MAX) == COUNTER_MAX)) {
+		if (unlikely(COUNTER(*bmc) == COUNTER_MAX)) {
 			DEFINE_WAIT(__wait);
 			/* note that it is safe to do the prepare_to_wait
 			 * after the test as long as we do it before dropping
@@ -1254,18 +1293,16 @@ 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->lock);
-			blk_unplug(bitmap->mddev->queue);
+			spin_unlock_irq(&bitmap->counts.lock);
 			schedule();
 			finish_wait(&bitmap->overflow_wait, &__wait);
 			continue;
 		}
 
-		switch(*bmc) {
+		switch (*bmc) {
 		case 0:
 			bitmap_file_set_bit(bitmap, offset);
-			bitmap_count_page(bitmap,offset, 1);
-			blk_plug_device_unlocked(bitmap->mddev->queue);
+			bitmap_count_page(&bitmap->counts, offset, 1);
 			/* fall through */
 		case 1:
 			*bmc = 2;
@@ -1273,67 +1310,73 @@ int bitmap_startwrite(struct bitmap *bitmap, sector_t offset, unsigned long sect
 
 		(*bmc)++;
 
-		spin_unlock_irq(&bitmap->lock);
+		spin_unlock_irq(&bitmap->counts.lock);
 
 		offset += blocks;
 		if (sectors > blocks)
 			sectors -= blocks;
-		else sectors = 0;
+		else
+			sectors = 0;
 	}
-	bitmap->allclean = 0;
 	return 0;
 }
+EXPORT_SYMBOL(bitmap_startwrite);
 
 void bitmap_endwrite(struct bitmap *bitmap, sector_t offset, unsigned long sectors,
 		     int success, int behind)
 {
-	if (!bitmap) return;
+	if (!bitmap)
+		return;
 	if (behind) {
-		atomic_dec(&bitmap->behind_writes);
-		PRINTK(KERN_DEBUG "dec write-behind count %d/%d\n",
-		  atomic_read(&bitmap->behind_writes), bitmap->max_write_behind);
+		if (atomic_dec_and_test(&bitmap->behind_writes))
+			wake_up(&bitmap->behind_wait);
+		pr_debug("dec write-behind count %d/%lu\n",
+			 atomic_read(&bitmap->behind_writes),
+			 bitmap->mddev->bitmap_info.max_write_behind);
 	}
 
 	while (sectors) {
-		int blocks;
+		sector_t blocks;
 		unsigned long flags;
 		bitmap_counter_t *bmc;
 
-		spin_lock_irqsave(&bitmap->lock, flags);
-		bmc = bitmap_get_counter(bitmap, offset, &blocks, 0);
+		spin_lock_irqsave(&bitmap->counts.lock, flags);
+		bmc = bitmap_get_counter(&bitmap->counts, offset, &blocks, 0);
 		if (!bmc) {
-			spin_unlock_irqrestore(&bitmap->lock, flags);
+			spin_unlock_irqrestore(&bitmap->counts.lock, flags);
 			return;
 		}
 
-		if (success &&
+		if (success && !bitmap->mddev->degraded &&
 		    bitmap->events_cleared < bitmap->mddev->events) {
 			bitmap->events_cleared = bitmap->mddev->events;
 			bitmap->need_sync = 1;
+			sysfs_notify_dirent_safe(bitmap->sysfs_can_clear);
 		}
 
-		if (!success && ! (*bmc & NEEDED_MASK))
+		if (!success && !NEEDED(*bmc))
 			*bmc |= NEEDED_MASK;
 
-		if ((*bmc & COUNTER_MAX) == COUNTER_MAX)
+		if (COUNTER(*bmc) == COUNTER_MAX)
 			wake_up(&bitmap->overflow_wait);
 
 		(*bmc)--;
 		if (*bmc <= 2) {
-			set_page_attr(bitmap,
-				      filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap)),
-				      BITMAP_PAGE_CLEAN);
+			bitmap_set_pending(&bitmap->counts, offset);
+			bitmap->allclean = 0;
 		}
-		spin_unlock_irqrestore(&bitmap->lock, flags);
+		spin_unlock_irqrestore(&bitmap->counts.lock, flags);
 		offset += blocks;
 		if (sectors > blocks)
 			sectors -= blocks;
-		else sectors = 0;
+		else
+			sectors = 0;
 	}
 }
+EXPORT_SYMBOL(bitmap_endwrite);
 
-int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, int *blocks,
-			int degraded)
+static int __bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks,
+			       int degraded)
 {
 	bitmap_counter_t *bmc;
 	int rv;
@@ -1341,8 +1384,8 @@ int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, int *blocks,
 		*blocks = 1024;
 		return 1; /* always resync if no bitmap */
 	}
-	spin_lock_irq(&bitmap->lock);
-	bmc = bitmap_get_counter(bitmap, offset, blocks, 0);
+	spin_lock_irq(&bitmap->counts.lock);
+	bmc = bitmap_get_counter(&bitmap->counts, offset, blocks, 0);
 	rv = 0;
 	if (bmc) {
 		/* locked */
@@ -1356,29 +1399,48 @@ int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, int *blocks,
 			}
 		}
 	}
-	spin_unlock_irq(&bitmap->lock);
-	bitmap->allclean = 0;
+	spin_unlock_irq(&bitmap->counts.lock);
+	return rv;
+}
+
+int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks,
+		      int degraded)
+{
+	/* bitmap_start_sync must always report on multiples of whole
+	 * pages, otherwise resync (which is very PAGE_SIZE based) will
+	 * get confused.
+	 * So call __bitmap_start_sync repeatedly (if needed) until
+	 * At least PAGE_SIZE>>9 blocks are covered.
+	 * Return the 'or' of the result.
+	 */
+	int rv = 0;
+	sector_t blocks1;
+
+	*blocks = 0;
+	while (*blocks < (PAGE_SIZE>>9)) {
+		rv |= __bitmap_start_sync(bitmap, offset,
+					  &blocks1, degraded);
+		offset += blocks1;
+		*blocks += blocks1;
+	}
 	return rv;
 }
+EXPORT_SYMBOL(bitmap_start_sync);
 
-void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, int *blocks, int aborted)
+void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int aborted)
 {
 	bitmap_counter_t *bmc;
 	unsigned long flags;
-/*
-	if (offset == 0) printk("bitmap_end_sync 0 (%d)\n", aborted);
-*/	if (bitmap == NULL) {
+
+	if (bitmap == NULL) {
 		*blocks = 1024;
 		return;
 	}
-	spin_lock_irqsave(&bitmap->lock, flags);
-	bmc = bitmap_get_counter(bitmap, offset, blocks, 0);
+	spin_lock_irqsave(&bitmap->counts.lock, flags);
+	bmc = bitmap_get_counter(&bitmap->counts, offset, blocks, 0);
 	if (bmc == NULL)
 		goto unlock;
 	/* locked */
-/*
-	if (offset == 0) printk("bitmap_end sync found 0x%x, blocks %d\n", *bmc, *blocks);
-*/
 	if (RESYNC(*bmc)) {
 		*bmc &= ~RESYNC_MASK;
 
@@ -1386,16 +1448,15 @@ void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, int *blocks, int ab
 			*bmc |= NEEDED_MASK;
 		else {
 			if (*bmc <= 2) {
-				set_page_attr(bitmap,
-					      filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap)),
-					      BITMAP_PAGE_CLEAN);
+				bitmap_set_pending(&bitmap->counts, offset);
+				bitmap->allclean = 0;
 			}
 		}
 	}
  unlock:
-	spin_unlock_irqrestore(&bitmap->lock, flags);
-	bitmap->allclean = 0;
+	spin_unlock_irqrestore(&bitmap->counts.lock, flags);
 }
+EXPORT_SYMBOL(bitmap_end_sync);
 
 void bitmap_close_sync(struct bitmap *bitmap)
 {
@@ -1404,7 +1465,7 @@ void bitmap_close_sync(struct bitmap *bitmap)
 	 * RESYNC bit wherever it is still on
 	 */
 	sector_t sector = 0;
-	int blocks;
+	sector_t blocks;
 	if (!bitmap)
 		return;
 	while (sector < bitmap->mddev->resync_max_sectors) {
@@ -1412,11 +1473,12 @@ void bitmap_close_sync(struct bitmap *bitmap)
 		sector += blocks;
 	}
 }
+EXPORT_SYMBOL(bitmap_close_sync);
 
 void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector)
 {
 	sector_t s = 0;
-	int blocks;
+	sector_t blocks;
 
 	if (!bitmap)
 		return;
@@ -1425,44 +1487,46 @@ void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector)
 		return;
 	}
 	if (time_before(jiffies, (bitmap->last_end_sync
-				  + bitmap->daemon_sleep * HZ)))
+				  + bitmap->mddev->bitmap_info.daemon_sleep)))
 		return;
 	wait_event(bitmap->mddev->recovery_wait,
 		   atomic_read(&bitmap->mddev->recovery_active) == 0);
 
-	sector &= ~((1ULL << CHUNK_BLOCK_SHIFT(bitmap)) - 1);
+	bitmap->mddev->curr_resync_completed = sector;
+	set_bit(MD_CHANGE_CLEAN, &bitmap->mddev->flags);
+	sector &= ~((1ULL << bitmap->counts.chunkshift) - 1);
 	s = 0;
 	while (s < sector && s < bitmap->mddev->resync_max_sectors) {
 		bitmap_end_sync(bitmap, s, &blocks, 0);
 		s += blocks;
 	}
 	bitmap->last_end_sync = jiffies;
+	sysfs_notify(&bitmap->mddev->kobj, NULL, "sync_completed");
 }
+EXPORT_SYMBOL(bitmap_cond_end_sync);
 
 static void bitmap_set_memory_bits(struct bitmap *bitmap, sector_t offset, int needed)
 {
 	/* For each chunk covered by any of these sectors, set the
-	 * counter to 1 and set resync_needed.  They should all
+	 * counter to 2 and possibly set resync_needed.  They should all
 	 * be 0 at this point
 	 */
 
-	int secs;
+	sector_t secs;
 	bitmap_counter_t *bmc;
-	spin_lock_irq(&bitmap->lock);
-	bmc = bitmap_get_counter(bitmap, offset, &secs, 1);
+	spin_lock_irq(&bitmap->counts.lock);
+	bmc = bitmap_get_counter(&bitmap->counts, offset, &secs, 1);
 	if (!bmc) {
-		spin_unlock_irq(&bitmap->lock);
+		spin_unlock_irq(&bitmap->counts.lock);
 		return;
 	}
-	if (! *bmc) {
-		struct page *page;
-		*bmc = 1 | (needed?NEEDED_MASK:0);
-		bitmap_count_page(bitmap, offset, 1);
-		page = filemap_get_page(bitmap, offset >> CHUNK_BLOCK_SHIFT(bitmap));
-		set_page_attr(bitmap, page, BITMAP_PAGE_CLEAN);
+	if (!*bmc) {
+		*bmc = 2 | (needed ? NEEDED_MASK : 0);
+		bitmap_count_page(&bitmap->counts, offset, 1);
+		bitmap_set_pending(&bitmap->counts, offset);
+		bitmap->allclean = 0;
 	}
-	spin_unlock_irq(&bitmap->lock);
-	bitmap->allclean = 0;
+	spin_unlock_irq(&bitmap->counts.lock);
 }
 
 /* dirty the memory and file bits for bitmap chunks "s" to "e" */
@@ -1471,19 +1535,25 @@ void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e)
 	unsigned long chunk;
 
 	for (chunk = s; chunk <= e; chunk++) {
-		sector_t sec = chunk << CHUNK_BLOCK_SHIFT(bitmap);
+		sector_t sec = (sector_t)chunk << bitmap->counts.chunkshift;
 		bitmap_set_memory_bits(bitmap, sec, 1);
 		bitmap_file_set_bit(bitmap, sec);
+		if (sec < bitmap->mddev->recovery_cp)
+			/* We are asserting that the array is dirty,
+			 * so move the recovery_cp address back so
+			 * that it is obvious that it is dirty
+			 */
+			bitmap->mddev->recovery_cp = sec;
 	}
 }
 
 /*
  * flush out any pending updates
  */
-void bitmap_flush(mddev_t *mddev)
+void bitmap_flush(struct mddev *mddev)
 {
 	struct bitmap *bitmap = mddev->bitmap;
-	int sleep;
+	long sleep;
 
 	if (!bitmap) /* there was no bitmap */
 		return;
@@ -1491,12 +1561,13 @@ void bitmap_flush(mddev_t *mddev)
 	/* run the daemon_work three time to ensure everything is flushed
 	 * that can be
 	 */
-	sleep = bitmap->daemon_sleep;
-	bitmap->daemon_sleep = 0;
-	bitmap_daemon_work(bitmap);
-	bitmap_daemon_work(bitmap);
-	bitmap_daemon_work(bitmap);
-	bitmap->daemon_sleep = sleep;
+	sleep = mddev->bitmap_info.daemon_sleep * 2;
+	bitmap->daemon_lastrun -= sleep;
+	bitmap_daemon_work(mddev);
+	bitmap->daemon_lastrun -= sleep;
+	bitmap_daemon_work(mddev);
+	bitmap->daemon_lastrun -= sleep;
+	bitmap_daemon_work(mddev);
 	bitmap_update_sb(bitmap);
 }
 
@@ -1511,11 +1582,15 @@ static void bitmap_free(struct bitmap *bitmap)
 	if (!bitmap) /* there was no bitmap */
 		return;
 
-	/* release the bitmap file and kill the daemon */
-	bitmap_file_put(bitmap);
+	/* Shouldn't be needed - but just in case.... */
+	wait_event(bitmap->write_wait,
+		   atomic_read(&bitmap->pending_writes) == 0);
+
+	/* release the bitmap file  */
+	bitmap_file_unmap(&bitmap->storage);
 
-	bp = bitmap->bp;
-	pages = bitmap->pages;
+	bp = bitmap->counts.bp;
+	pages = bitmap->counts.pages;
 
 	/* free all allocated memory */
 
@@ -1526,17 +1601,23 @@ static void bitmap_free(struct bitmap *bitmap)
 	kfree(bp);
 	kfree(bitmap);
 }
-void bitmap_destroy(mddev_t *mddev)
+
+void bitmap_destroy(struct mddev *mddev)
 {
 	struct bitmap *bitmap = mddev->bitmap;
 
 	if (!bitmap) /* there was no bitmap */
 		return;
 
+	mutex_lock(&mddev->bitmap_info.mutex);
 	mddev->bitmap = NULL; /* disconnect from the md device */
+	mutex_unlock(&mddev->bitmap_info.mutex);
 	if (mddev->thread)
 		mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
 
+	if (bitmap->sysfs_can_clear)
+		sysfs_put(bitmap->sysfs_can_clear);
+
 	bitmap_free(bitmap);
 }
 
@@ -1544,106 +1625,641 @@ void bitmap_destroy(mddev_t *mddev)
  * initialize the bitmap structure
  * if this returns an error, bitmap_destroy must be called to do clean up
  */
-int bitmap_create(mddev_t *mddev)
+int bitmap_create(struct mddev *mddev)
 {
 	struct bitmap *bitmap;
-	unsigned long blocks = mddev->resync_max_sectors;
-	unsigned long chunks;
-	unsigned long pages;
-	struct file *file = mddev->bitmap_file;
+	sector_t blocks = mddev->resync_max_sectors;
+	struct file *file = mddev->bitmap_info.file;
 	int err;
-	sector_t start;
+	struct kernfs_node *bm = NULL;
 
 	BUILD_BUG_ON(sizeof(bitmap_super_t) != 256);
 
-	if (!file && !mddev->bitmap_offset) /* bitmap disabled, nothing to do */
-		return 0;
-
-	BUG_ON(file && mddev->bitmap_offset);
+	BUG_ON(file && mddev->bitmap_info.offset);
 
 	bitmap = kzalloc(sizeof(*bitmap), GFP_KERNEL);
 	if (!bitmap)
 		return -ENOMEM;
 
-	spin_lock_init(&bitmap->lock);
+	spin_lock_init(&bitmap->counts.lock);
 	atomic_set(&bitmap->pending_writes, 0);
 	init_waitqueue_head(&bitmap->write_wait);
 	init_waitqueue_head(&bitmap->overflow_wait);
+	init_waitqueue_head(&bitmap->behind_wait);
 
 	bitmap->mddev = mddev;
 
-	bitmap->file = file;
-	bitmap->offset = mddev->bitmap_offset;
+	if (mddev->kobj.sd)
+		bm = sysfs_get_dirent(mddev->kobj.sd, "bitmap");
+	if (bm) {
+		bitmap->sysfs_can_clear = sysfs_get_dirent(bm, "can_clear");
+		sysfs_put(bm);
+	} else
+		bitmap->sysfs_can_clear = NULL;
+
+	bitmap->storage.file = file;
 	if (file) {
 		get_file(file);
-		do_sync_mapping_range(file->f_mapping, 0, LLONG_MAX,
-				      SYNC_FILE_RANGE_WAIT_BEFORE |
-				      SYNC_FILE_RANGE_WRITE |
-				      SYNC_FILE_RANGE_WAIT_AFTER);
+		/* As future accesses to this file will use bmap,
+		 * and bypass the page cache, we must sync the file
+		 * first.
+		 */
+		vfs_fsync(file, 1);
+	}
+	/* read superblock from bitmap file (this sets mddev->bitmap_info.chunksize) */
+	if (!mddev->bitmap_info.external) {
+		/*
+		 * If 'MD_ARRAY_FIRST_USE' is set, then device-mapper is
+		 * instructing us to create a new on-disk bitmap instance.
+		 */
+		if (test_and_clear_bit(MD_ARRAY_FIRST_USE, &mddev->flags))
+			err = bitmap_new_disk_sb(bitmap);
+		else
+			err = bitmap_read_sb(bitmap);
+	} else {
+		err = 0;
+		if (mddev->bitmap_info.chunksize == 0 ||
+		    mddev->bitmap_info.daemon_sleep == 0)
+			/* chunksize and time_base need to be
+			 * set first. */
+			err = -EINVAL;
 	}
-	/* read superblock from bitmap file (this sets bitmap->chunksize) */
-	err = bitmap_read_sb(bitmap);
 	if (err)
 		goto error;
 
-	bitmap->chunkshift = ffz(~bitmap->chunksize);
+	bitmap->daemon_lastrun = jiffies;
+	err = bitmap_resize(bitmap, blocks, mddev->bitmap_info.chunksize, 1);
+	if (err)
+		goto error;
 
-	/* now that chunksize and chunkshift are set, we can use these macros */
- 	chunks = (blocks + CHUNK_BLOCK_RATIO(bitmap) - 1) /
-			CHUNK_BLOCK_RATIO(bitmap);
- 	pages = (chunks + PAGE_COUNTER_RATIO - 1) / PAGE_COUNTER_RATIO;
+	printk(KERN_INFO "created bitmap (%lu pages) for device %s\n",
+	       bitmap->counts.pages, bmname(bitmap));
 
-	BUG_ON(!pages);
+	mddev->bitmap = bitmap;
+	return test_bit(BITMAP_WRITE_ERROR, &bitmap->flags) ? -EIO : 0;
 
-	bitmap->chunks = chunks;
-	bitmap->pages = pages;
-	bitmap->missing_pages = pages;
-	bitmap->counter_bits = COUNTER_BITS;
+ error:
+	bitmap_free(bitmap);
+	return err;
+}
 
-	bitmap->syncchunk = ~0UL;
+int bitmap_load(struct mddev *mddev)
+{
+	int err = 0;
+	sector_t start = 0;
+	sector_t sector = 0;
+	struct bitmap *bitmap = mddev->bitmap;
 
-#ifdef INJECT_FATAL_FAULT_1
-	bitmap->bp = NULL;
-#else
-	bitmap->bp = kzalloc(pages * sizeof(*bitmap->bp), GFP_KERNEL);
-#endif
-	err = -ENOMEM;
-	if (!bitmap->bp)
-		goto error;
+	if (!bitmap)
+		goto out;
+
+	/* Clear out old bitmap info first:  Either there is none, or we
+	 * are resuming after someone else has possibly changed things,
+	 * so we should forget old cached info.
+	 * All chunks should be clean, but some might need_sync.
+	 */
+	while (sector < mddev->resync_max_sectors) {
+		sector_t blocks;
+		bitmap_start_sync(bitmap, sector, &blocks, 0);
+		sector += blocks;
+	}
+	bitmap_close_sync(bitmap);
 
-	/* now that we have some pages available, initialize the in-memory
-	 * bitmap from the on-disk bitmap */
-	start = 0;
 	if (mddev->degraded == 0
 	    || bitmap->events_cleared == mddev->events)
-		/* no need to keep dirty bits to optimise a re-add of a missing device */
+		/* no need to keep dirty bits to optimise a
+		 * re-add of a missing device */
 		start = mddev->recovery_cp;
+
+	mutex_lock(&mddev->bitmap_info.mutex);
 	err = bitmap_init_from_disk(bitmap, start);
+	mutex_unlock(&mddev->bitmap_info.mutex);
 
 	if (err)
-		goto error;
-
-	printk(KERN_INFO "created bitmap (%lu pages) for device %s\n",
-		pages, bmname(bitmap));
+		goto out;
+	clear_bit(BITMAP_STALE, &bitmap->flags);
 
-	mddev->bitmap = bitmap;
+	/* Kick recovery in case any bits were set */
+	set_bit(MD_RECOVERY_NEEDED, &bitmap->mddev->recovery);
 
-	mddev->thread->timeout = bitmap->daemon_sleep * HZ;
+	mddev->thread->timeout = mddev->bitmap_info.daemon_sleep;
+	md_wakeup_thread(mddev->thread);
 
 	bitmap_update_sb(bitmap);
 
-	return (bitmap->flags & BITMAP_WRITE_ERROR) ? -EIO : 0;
-
- error:
-	bitmap_free(bitmap);
+	if (test_bit(BITMAP_WRITE_ERROR, &bitmap->flags))
+		err = -EIO;
+out:
 	return err;
 }
+EXPORT_SYMBOL_GPL(bitmap_load);
+
+void bitmap_status(struct seq_file *seq, struct bitmap *bitmap)
+{
+	unsigned long chunk_kb;
+	struct bitmap_counts *counts;
+
+	if (!bitmap)
+		return;
+
+	counts = &bitmap->counts;
+
+	chunk_kb = bitmap->mddev->bitmap_info.chunksize >> 10;
+	seq_printf(seq, "bitmap: %lu/%lu pages [%luKB], "
+		   "%lu%s chunk",
+		   counts->pages - counts->missing_pages,
+		   counts->pages,
+		   (counts->pages - counts->missing_pages)
+		   << (PAGE_SHIFT - 10),
+		   chunk_kb ? chunk_kb : bitmap->mddev->bitmap_info.chunksize,
+		   chunk_kb ? "KB" : "B");
+	if (bitmap->storage.file) {
+		seq_printf(seq, ", file: ");
+		seq_path(seq, &bitmap->storage.file->f_path, " \t\n");
+	}
+
+	seq_printf(seq, "\n");
+}
+
+int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
+		  int chunksize, int init)
+{
+	/* If chunk_size is 0, choose an appropriate chunk size.
+	 * Then possibly allocate new storage space.
+	 * Then quiesce, copy bits, replace bitmap, and re-start
+	 *
+	 * This function is called both to set up the initial bitmap
+	 * and to resize the bitmap while the array is active.
+	 * If this happens as a result of the array being resized,
+	 * chunksize will be zero, and we need to choose a suitable
+	 * chunksize, otherwise we use what we are given.
+	 */
+	struct bitmap_storage store;
+	struct bitmap_counts old_counts;
+	unsigned long chunks;
+	sector_t block;
+	sector_t old_blocks, new_blocks;
+	int chunkshift;
+	int ret = 0;
+	long pages;
+	struct bitmap_page *new_bp;
+
+	if (chunksize == 0) {
+		/* If there is enough space, leave the chunk size unchanged,
+		 * else increase by factor of two until there is enough space.
+		 */
+		long bytes;
+		long space = bitmap->mddev->bitmap_info.space;
+
+		if (space == 0) {
+			/* We don't know how much space there is, so limit
+			 * to current size - in sectors.
+			 */
+			bytes = DIV_ROUND_UP(bitmap->counts.chunks, 8);
+			if (!bitmap->mddev->bitmap_info.external)
+				bytes += sizeof(bitmap_super_t);
+			space = DIV_ROUND_UP(bytes, 512);
+			bitmap->mddev->bitmap_info.space = space;
+		}
+		chunkshift = bitmap->counts.chunkshift;
+		chunkshift--;
+		do {
+			/* 'chunkshift' is shift from block size to chunk size */
+			chunkshift++;
+			chunks = DIV_ROUND_UP_SECTOR_T(blocks, 1 << chunkshift);
+			bytes = DIV_ROUND_UP(chunks, 8);
+			if (!bitmap->mddev->bitmap_info.external)
+				bytes += sizeof(bitmap_super_t);
+		} while (bytes > (space << 9));
+	} else
+		chunkshift = ffz(~chunksize) - BITMAP_BLOCK_SHIFT;
+
+	chunks = DIV_ROUND_UP_SECTOR_T(blocks, 1 << chunkshift);
+	memset(&store, 0, sizeof(store));
+	if (bitmap->mddev->bitmap_info.offset || bitmap->mddev->bitmap_info.file)
+		ret = bitmap_storage_alloc(&store, chunks,
+					   !bitmap->mddev->bitmap_info.external);
+	if (ret)
+		goto err;
+
+	pages = DIV_ROUND_UP(chunks, PAGE_COUNTER_RATIO);
+
+	new_bp = kzalloc(pages * sizeof(*new_bp), GFP_KERNEL);
+	ret = -ENOMEM;
+	if (!new_bp) {
+		bitmap_file_unmap(&store);
+		goto err;
+	}
+
+	if (!init)
+		bitmap->mddev->pers->quiesce(bitmap->mddev, 1);
+
+	store.file = bitmap->storage.file;
+	bitmap->storage.file = NULL;
+
+	if (store.sb_page && bitmap->storage.sb_page)
+		memcpy(page_address(store.sb_page),
+		       page_address(bitmap->storage.sb_page),
+		       sizeof(bitmap_super_t));
+	bitmap_file_unmap(&bitmap->storage);
+	bitmap->storage = store;
+
+	old_counts = bitmap->counts;
+	bitmap->counts.bp = new_bp;
+	bitmap->counts.pages = pages;
+	bitmap->counts.missing_pages = pages;
+	bitmap->counts.chunkshift = chunkshift;
+	bitmap->counts.chunks = chunks;
+	bitmap->mddev->bitmap_info.chunksize = 1 << (chunkshift +
+						     BITMAP_BLOCK_SHIFT);
+
+	blocks = min(old_counts.chunks << old_counts.chunkshift,
+		     chunks << chunkshift);
+
+	spin_lock_irq(&bitmap->counts.lock);
+	for (block = 0; block < blocks; ) {
+		bitmap_counter_t *bmc_old, *bmc_new;
+		int set;
+
+		bmc_old = bitmap_get_counter(&old_counts, block,
+					     &old_blocks, 0);
+		set = bmc_old && NEEDED(*bmc_old);
+
+		if (set) {
+			bmc_new = bitmap_get_counter(&bitmap->counts, block,
+						     &new_blocks, 1);
+			if (*bmc_new == 0) {
+				/* need to set on-disk bits too. */
+				sector_t end = block + new_blocks;
+				sector_t start = block >> chunkshift;
+				start <<= chunkshift;
+				while (start < end) {
+					bitmap_file_set_bit(bitmap, block);
+					start += 1 << chunkshift;
+				}
+				*bmc_new = 2;
+				bitmap_count_page(&bitmap->counts,
+						  block, 1);
+				bitmap_set_pending(&bitmap->counts,
+						   block);
+			}
+			*bmc_new |= NEEDED_MASK;
+			if (new_blocks < old_blocks)
+				old_blocks = new_blocks;
+		}
+		block += old_blocks;
+	}
+
+	if (!init) {
+		int i;
+		while (block < (chunks << chunkshift)) {
+			bitmap_counter_t *bmc;
+			bmc = bitmap_get_counter(&bitmap->counts, block,
+						 &new_blocks, 1);
+			if (bmc) {
+				/* new space.  It needs to be resynced, so
+				 * we set NEEDED_MASK.
+				 */
+				if (*bmc == 0) {
+					*bmc = NEEDED_MASK | 2;
+					bitmap_count_page(&bitmap->counts,
+							  block, 1);
+					bitmap_set_pending(&bitmap->counts,
+							   block);
+				}
+			}
+			block += new_blocks;
+		}
+		for (i = 0; i < bitmap->storage.file_pages; i++)
+			set_page_attr(bitmap, i, BITMAP_PAGE_DIRTY);
+	}
+	spin_unlock_irq(&bitmap->counts.lock);
+
+	if (!init) {
+		bitmap_unplug(bitmap);
+		bitmap->mddev->pers->quiesce(bitmap->mddev, 0);
+	}
+	ret = 0;
+err:
+	return ret;
+}
+EXPORT_SYMBOL_GPL(bitmap_resize);
+
+static ssize_t
+location_show(struct mddev *mddev, char *page)
+{
+	ssize_t len;
+	if (mddev->bitmap_info.file)
+		len = sprintf(page, "file");
+	else if (mddev->bitmap_info.offset)
+		len = sprintf(page, "%+lld", (long long)mddev->bitmap_info.offset);
+	else
+		len = sprintf(page, "none");
+	len += sprintf(page+len, "\n");
+	return len;
+}
+
+static ssize_t
+location_store(struct mddev *mddev, const char *buf, size_t len)
+{
+
+	if (mddev->pers) {
+		if (!mddev->pers->quiesce)
+			return -EBUSY;
+		if (mddev->recovery || mddev->sync_thread)
+			return -EBUSY;
+	}
+
+	if (mddev->bitmap || mddev->bitmap_info.file ||
+	    mddev->bitmap_info.offset) {
+		/* bitmap already configured.  Only option is to clear it */
+		if (strncmp(buf, "none", 4) != 0)
+			return -EBUSY;
+		if (mddev->pers) {
+			mddev->pers->quiesce(mddev, 1);
+			bitmap_destroy(mddev);
+			mddev->pers->quiesce(mddev, 0);
+		}
+		mddev->bitmap_info.offset = 0;
+		if (mddev->bitmap_info.file) {
+			struct file *f = mddev->bitmap_info.file;
+			mddev->bitmap_info.file = NULL;
+			fput(f);
+		}
+	} else {
+		/* No bitmap, OK to set a location */
+		long long offset;
+		if (strncmp(buf, "none", 4) == 0)
+			/* nothing to be done */;
+		else if (strncmp(buf, "file:", 5) == 0) {
+			/* Not supported yet */
+			return -EINVAL;
+		} else {
+			int rv;
+			if (buf[0] == '+')
+				rv = kstrtoll(buf+1, 10, &offset);
+			else
+				rv = kstrtoll(buf, 10, &offset);
+			if (rv)
+				return rv;
+			if (offset == 0)
+				return -EINVAL;
+			if (mddev->bitmap_info.external == 0 &&
+			    mddev->major_version == 0 &&
+			    offset != mddev->bitmap_info.default_offset)
+				return -EINVAL;
+			mddev->bitmap_info.offset = offset;
+			if (mddev->pers) {
+				mddev->pers->quiesce(mddev, 1);
+				rv = bitmap_create(mddev);
+				if (!rv)
+					rv = bitmap_load(mddev);
+				if (rv) {
+					bitmap_destroy(mddev);
+					mddev->bitmap_info.offset = 0;
+				}
+				mddev->pers->quiesce(mddev, 0);
+				if (rv)
+					return rv;
+			}
+		}
+	}
+	if (!mddev->external) {
+		/* Ensure new bitmap info is stored in
+		 * metadata promptly.
+		 */
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		md_wakeup_thread(mddev->thread);
+	}
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_location =
+__ATTR(location, S_IRUGO|S_IWUSR, location_show, location_store);
+
+/* 'bitmap/space' is the space available at 'location' for the
+ * bitmap.  This allows the kernel to know when it is safe to
+ * resize the bitmap to match a resized array.
+ */
+static ssize_t
+space_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%lu\n", mddev->bitmap_info.space);
+}
+
+static ssize_t
+space_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	unsigned long sectors;
+	int rv;
+
+	rv = kstrtoul(buf, 10, &sectors);
+	if (rv)
+		return rv;
+
+	if (sectors == 0)
+		return -EINVAL;
+
+	if (mddev->bitmap &&
+	    sectors < (mddev->bitmap->storage.bytes + 511) >> 9)
+		return -EFBIG; /* Bitmap is too big for this small space */
+
+	/* could make sure it isn't too big, but that isn't really
+	 * needed - user-space should be careful.
+	 */
+	mddev->bitmap_info.space = sectors;
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_space =
+__ATTR(space, S_IRUGO|S_IWUSR, space_show, space_store);
+
+static ssize_t
+timeout_show(struct mddev *mddev, char *page)
+{
+	ssize_t len;
+	unsigned long secs = mddev->bitmap_info.daemon_sleep / HZ;
+	unsigned long jifs = mddev->bitmap_info.daemon_sleep % HZ;
+
+	len = sprintf(page, "%lu", secs);
+	if (jifs)
+		len += sprintf(page+len, ".%03u", jiffies_to_msecs(jifs));
+	len += sprintf(page+len, "\n");
+	return len;
+}
+
+static ssize_t
+timeout_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	/* timeout can be set at any time */
+	unsigned long timeout;
+	int rv = strict_strtoul_scaled(buf, &timeout, 4);
+	if (rv)
+		return rv;
+
+	/* just to make sure we don't overflow... */
+	if (timeout >= LONG_MAX / HZ)
+		return -EINVAL;
+
+	timeout = timeout * HZ / 10000;
+
+	if (timeout >= MAX_SCHEDULE_TIMEOUT)
+		timeout = MAX_SCHEDULE_TIMEOUT-1;
+	if (timeout < 1)
+		timeout = 1;
+	mddev->bitmap_info.daemon_sleep = timeout;
+	if (mddev->thread) {
+		/* if thread->timeout is MAX_SCHEDULE_TIMEOUT, then
+		 * the bitmap is all clean and we don't need to
+		 * adjust the timeout right now
+		 */
+		if (mddev->thread->timeout < MAX_SCHEDULE_TIMEOUT) {
+			mddev->thread->timeout = timeout;
+			md_wakeup_thread(mddev->thread);
+		}
+	}
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_timeout =
+__ATTR(time_base, S_IRUGO|S_IWUSR, timeout_show, timeout_store);
+
+static ssize_t
+backlog_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%lu\n", mddev->bitmap_info.max_write_behind);
+}
+
+static ssize_t
+backlog_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	unsigned long backlog;
+	int rv = kstrtoul(buf, 10, &backlog);
+	if (rv)
+		return rv;
+	if (backlog > COUNTER_MAX)
+		return -EINVAL;
+	mddev->bitmap_info.max_write_behind = backlog;
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_backlog =
+__ATTR(backlog, S_IRUGO|S_IWUSR, backlog_show, backlog_store);
+
+static ssize_t
+chunksize_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%lu\n", mddev->bitmap_info.chunksize);
+}
+
+static ssize_t
+chunksize_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	/* Can only be changed when no bitmap is active */
+	int rv;
+	unsigned long csize;
+	if (mddev->bitmap)
+		return -EBUSY;
+	rv = kstrtoul(buf, 10, &csize);
+	if (rv)
+		return rv;
+	if (csize < 512 ||
+	    !is_power_of_2(csize))
+		return -EINVAL;
+	mddev->bitmap_info.chunksize = csize;
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_chunksize =
+__ATTR(chunksize, S_IRUGO|S_IWUSR, chunksize_show, chunksize_store);
+
+static ssize_t metadata_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%s\n", (mddev->bitmap_info.external
+				      ? "external" : "internal"));
+}
+
+static ssize_t metadata_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	if (mddev->bitmap ||
+	    mddev->bitmap_info.file ||
+	    mddev->bitmap_info.offset)
+		return -EBUSY;
+	if (strncmp(buf, "external", 8) == 0)
+		mddev->bitmap_info.external = 1;
+	else if (strncmp(buf, "internal", 8) == 0)
+		mddev->bitmap_info.external = 0;
+	else
+		return -EINVAL;
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_metadata =
+__ATTR(metadata, S_IRUGO|S_IWUSR, metadata_show, metadata_store);
+
+static ssize_t can_clear_show(struct mddev *mddev, char *page)
+{
+	int len;
+	if (mddev->bitmap)
+		len = sprintf(page, "%s\n", (mddev->bitmap->need_sync ?
+					     "false" : "true"));
+	else
+		len = sprintf(page, "\n");
+	return len;
+}
+
+static ssize_t can_clear_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	if (mddev->bitmap == NULL)
+		return -ENOENT;
+	if (strncmp(buf, "false", 5) == 0)
+		mddev->bitmap->need_sync = 1;
+	else if (strncmp(buf, "true", 4) == 0) {
+		if (mddev->degraded)
+			return -EBUSY;
+		mddev->bitmap->need_sync = 0;
+	} else
+		return -EINVAL;
+	return len;
+}
+
+static struct md_sysfs_entry bitmap_can_clear =
+__ATTR(can_clear, S_IRUGO|S_IWUSR, can_clear_show, can_clear_store);
+
+static ssize_t
+behind_writes_used_show(struct mddev *mddev, char *page)
+{
+	if (mddev->bitmap == NULL)
+		return sprintf(page, "0\n");
+	return sprintf(page, "%lu\n",
+		       mddev->bitmap->behind_writes_used);
+}
+
+static ssize_t
+behind_writes_used_reset(struct mddev *mddev, const char *buf, size_t len)
+{
+	if (mddev->bitmap)
+		mddev->bitmap->behind_writes_used = 0;
+	return len;
+}
+
+static struct md_sysfs_entry max_backlog_used =
+__ATTR(max_backlog_used, S_IRUGO | S_IWUSR,
+       behind_writes_used_show, behind_writes_used_reset);
+
+static struct attribute *md_bitmap_attrs[] = {
+	&bitmap_location.attr,
+	&bitmap_space.attr,
+	&bitmap_timeout.attr,
+	&bitmap_backlog.attr,
+	&bitmap_chunksize.attr,
+	&bitmap_metadata.attr,
+	&bitmap_can_clear.attr,
+	&max_backlog_used.attr,
+	NULL
+};
+struct attribute_group md_bitmap_group = {
+	.name = "bitmap",
+	.attrs = md_bitmap_attrs,
+};
 
-/* the bitmap API -- for raid personalities */
-EXPORT_SYMBOL(bitmap_startwrite);
-EXPORT_SYMBOL(bitmap_endwrite);
-EXPORT_SYMBOL(bitmap_start_sync);
-EXPORT_SYMBOL(bitmap_end_sync);
-EXPORT_SYMBOL(bitmap_unplug);
-EXPORT_SYMBOL(bitmap_close_sync);
-EXPORT_SYMBOL(bitmap_cond_end_sync);
diff --git a/drivers/md/bitmap.h b/drivers/md/bitmap.h
new file mode 100644
index 00000000000..30210b9c4ef
--- /dev/null
+++ b/drivers/md/bitmap.h
@@ -0,0 +1,265 @@
+/*
+ * bitmap.h: Copyright (C) Peter T. Breuer (ptb@ot.uc3m.es) 2003
+ *
+ * additions: Copyright (C) 2003-2004, Paul Clements, SteelEye Technology, Inc.
+ */
+#ifndef BITMAP_H
+#define BITMAP_H 1
+
+#define BITMAP_MAJOR_LO 3
+/* version 4 insists the bitmap is in little-endian order
+ * with version 3, it is host-endian which is non-portable
+ */
+#define BITMAP_MAJOR_HI 4
+#define	BITMAP_MAJOR_HOSTENDIAN 3
+
+/*
+ * in-memory bitmap:
+ *
+ * Use 16 bit block counters to track pending writes to each "chunk".
+ * The 2 high order bits are special-purpose, the first is a flag indicating
+ * whether a resync is needed.  The second is a flag indicating whether a
+ * resync is active.
+ * This means that the counter is actually 14 bits:
+ *
+ * +--------+--------+------------------------------------------------+
+ * | resync | resync |               counter                          |
+ * | needed | active |                                                |
+ * |  (0-1) |  (0-1) |              (0-16383)                         |
+ * +--------+--------+------------------------------------------------+
+ *
+ * The "resync needed" bit is set when:
+ *    a '1' bit is read from storage at startup.
+ *    a write request fails on some drives
+ *    a resync is aborted on a chunk with 'resync active' set
+ * It is cleared (and resync-active set) when a resync starts across all drives
+ * of the chunk.
+ *
+ *
+ * The "resync active" bit is set when:
+ *    a resync is started on all drives, and resync_needed is set.
+ *       resync_needed will be cleared (as long as resync_active wasn't already set).
+ * It is cleared when a resync completes.
+ *
+ * The counter counts pending write requests, plus the on-disk bit.
+ * When the counter is '1' and the resync bits are clear, the on-disk
+ * bit can be cleared as well, thus setting the counter to 0.
+ * When we set a bit, or in the counter (to start a write), if the fields is
+ * 0, we first set the disk bit and set the counter to 1.
+ *
+ * If the counter is 0, the on-disk bit is clear and the stipe is clean
+ * Anything that dirties the stipe pushes the counter to 2 (at least)
+ * and sets the on-disk bit (lazily).
+ * If a periodic sweep find the counter at 2, it is decremented to 1.
+ * If the sweep find the counter at 1, the on-disk bit is cleared and the
+ * counter goes to zero.
+ *
+ * Also, we'll hijack the "map" pointer itself and use it as two 16 bit block
+ * counters as a fallback when "page" memory cannot be allocated:
+ *
+ * Normal case (page memory allocated):
+ *
+ *     page pointer (32-bit)
+ *
+ *     [ ] ------+
+ *               |
+ *               +-------> [   ][   ]..[   ] (4096 byte page == 2048 counters)
+ *                          c1   c2    c2048
+ *
+ * Hijacked case (page memory allocation failed):
+ *
+ *     hijacked page pointer (32-bit)
+ *
+ *     [		  ][		  ] (no page memory allocated)
+ *      counter #1 (16-bit) counter #2 (16-bit)
+ *
+ */
+
+#ifdef __KERNEL__
+
+#define PAGE_BITS (PAGE_SIZE << 3)
+#define PAGE_BIT_SHIFT (PAGE_SHIFT + 3)
+
+typedef __u16 bitmap_counter_t;
+#define COUNTER_BITS 16
+#define COUNTER_BIT_SHIFT 4
+#define COUNTER_BYTE_SHIFT (COUNTER_BIT_SHIFT - 3)
+
+#define NEEDED_MASK ((bitmap_counter_t) (1 << (COUNTER_BITS - 1)))
+#define RESYNC_MASK ((bitmap_counter_t) (1 << (COUNTER_BITS - 2)))
+#define COUNTER_MAX ((bitmap_counter_t) RESYNC_MASK - 1)
+#define NEEDED(x) (((bitmap_counter_t) x) & NEEDED_MASK)
+#define RESYNC(x) (((bitmap_counter_t) x) & RESYNC_MASK)
+#define COUNTER(x) (((bitmap_counter_t) x) & COUNTER_MAX)
+
+/* how many counters per page? */
+#define PAGE_COUNTER_RATIO (PAGE_BITS / COUNTER_BITS)
+/* same, except a shift value for more efficient bitops */
+#define PAGE_COUNTER_SHIFT (PAGE_BIT_SHIFT - COUNTER_BIT_SHIFT)
+/* same, except a mask value for more efficient bitops */
+#define PAGE_COUNTER_MASK  (PAGE_COUNTER_RATIO - 1)
+
+#define BITMAP_BLOCK_SHIFT 9
+
+#endif
+
+/*
+ * bitmap structures:
+ */
+
+#define BITMAP_MAGIC 0x6d746962
+
+/* use these for bitmap->flags and bitmap->sb->state bit-fields */
+enum bitmap_state {
+	BITMAP_STALE	   = 1,  /* the bitmap file is out of date or had -EIO */
+	BITMAP_WRITE_ERROR = 2, /* A write error has occurred */
+	BITMAP_HOSTENDIAN  =15,
+};
+
+/* the superblock at the front of the bitmap file -- little endian */
+typedef struct bitmap_super_s {
+	__le32 magic;        /*  0  BITMAP_MAGIC */
+	__le32 version;      /*  4  the bitmap major for now, could change... */
+	__u8  uuid[16];      /*  8  128 bit uuid - must match md device uuid */
+	__le64 events;       /* 24  event counter for the bitmap (1)*/
+	__le64 events_cleared;/*32  event counter when last bit cleared (2) */
+	__le64 sync_size;    /* 40  the size of the md device's sync range(3) */
+	__le32 state;        /* 48  bitmap state information */
+	__le32 chunksize;    /* 52  the bitmap chunk size in bytes */
+	__le32 daemon_sleep; /* 56  seconds between disk flushes */
+	__le32 write_behind; /* 60  number of outstanding write-behind writes */
+	__le32 sectors_reserved; /* 64 number of 512-byte sectors that are
+				  * reserved for the bitmap. */
+
+	__u8  pad[256 - 68]; /* set to zero */
+} bitmap_super_t;
+
+/* notes:
+ * (1) This event counter is updated before the eventcounter in the md superblock
+ *    When a bitmap is loaded, it is only accepted if this event counter is equal
+ *    to, or one greater than, the event counter in the superblock.
+ * (2) This event counter is updated when the other one is *if*and*only*if* the
+ *    array is not degraded.  As bits are not cleared when the array is degraded,
+ *    this represents the last time that any bits were cleared.
+ *    If a device is being added that has an event count with this value or
+ *    higher, it is accepted as conforming to the bitmap.
+ * (3)This is the number of sectors represented by the bitmap, and is the range that
+ *    resync happens across.  For raid1 and raid5/6 it is the size of individual
+ *    devices.  For raid10 it is the size of the array.
+ */
+
+#ifdef __KERNEL__
+
+/* the in-memory bitmap is represented by bitmap_pages */
+struct bitmap_page {
+	/*
+	 * map points to the actual memory page
+	 */
+	char *map;
+	/*
+	 * in emergencies (when map cannot be alloced), hijack the map
+	 * pointer and use it as two counters itself
+	 */
+	unsigned int hijacked:1;
+	/*
+	 * If any counter in this page is '1' or '2' - and so could be
+	 * cleared then that page is marked as 'pending'
+	 */
+	unsigned int pending:1;
+	/*
+	 * count of dirty bits on the page
+	 */
+	unsigned int  count:30;
+};
+
+/* the main bitmap structure - one per mddev */
+struct bitmap {
+
+	struct bitmap_counts {
+		spinlock_t lock;
+		struct bitmap_page *bp;
+		unsigned long pages;		/* total number of pages
+						 * in the bitmap */
+		unsigned long missing_pages;	/* number of pages
+						 * not yet allocated */
+		unsigned long chunkshift;	/* chunksize = 2^chunkshift
+						 * (for bitops) */
+		unsigned long chunks;		/* Total number of data
+						 * chunks for the array */
+	} counts;
+
+	struct mddev *mddev; /* the md device that the bitmap is for */
+
+	__u64	events_cleared;
+	int need_sync;
+
+	struct bitmap_storage {
+		struct file *file;		/* backing disk file */
+		struct page *sb_page;		/* cached copy of the bitmap
+						 * file superblock */
+		struct page **filemap;		/* list of cache pages for
+						 * the file */
+		unsigned long *filemap_attr;	/* attributes associated
+						 * w/ filemap pages */
+		unsigned long file_pages;	/* number of pages in the file*/
+		unsigned long bytes;		/* total bytes in the bitmap */
+	} storage;
+
+	unsigned long flags;
+
+	int allclean;
+
+	atomic_t behind_writes;
+	unsigned long behind_writes_used; /* highest actual value at runtime */
+
+	/*
+	 * the bitmap daemon - periodically wakes up and sweeps the bitmap
+	 * file, cleaning up bits and flushing out pages to disk as necessary
+	 */
+	unsigned long daemon_lastrun; /* jiffies of last run */
+	unsigned long last_end_sync; /* when we lasted called end_sync to
+				      * update bitmap with resync progress */
+
+	atomic_t pending_writes; /* pending writes to the bitmap file */
+	wait_queue_head_t write_wait;
+	wait_queue_head_t overflow_wait;
+	wait_queue_head_t behind_wait;
+
+	struct kernfs_node *sysfs_can_clear;
+};
+
+/* the bitmap API */
+
+/* these are used only by md/bitmap */
+int  bitmap_create(struct mddev *mddev);
+int bitmap_load(struct mddev *mddev);
+void bitmap_flush(struct mddev *mddev);
+void bitmap_destroy(struct mddev *mddev);
+
+void bitmap_print_sb(struct bitmap *bitmap);
+void bitmap_update_sb(struct bitmap *bitmap);
+void bitmap_status(struct seq_file *seq, struct bitmap *bitmap);
+
+int  bitmap_setallbits(struct bitmap *bitmap);
+void bitmap_write_all(struct bitmap *bitmap);
+
+void bitmap_dirty_bits(struct bitmap *bitmap, unsigned long s, unsigned long e);
+
+/* these are exported */
+int bitmap_startwrite(struct bitmap *bitmap, sector_t offset,
+			unsigned long sectors, int behind);
+void bitmap_endwrite(struct bitmap *bitmap, sector_t offset,
+			unsigned long sectors, int success, int behind);
+int bitmap_start_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int degraded);
+void bitmap_end_sync(struct bitmap *bitmap, sector_t offset, sector_t *blocks, int aborted);
+void bitmap_close_sync(struct bitmap *bitmap);
+void bitmap_cond_end_sync(struct bitmap *bitmap, sector_t sector);
+
+void bitmap_unplug(struct bitmap *bitmap);
+void bitmap_daemon_work(struct mddev *mddev);
+
+int bitmap_resize(struct bitmap *bitmap, sector_t blocks,
+		  int chunksize, int init);
+#endif
+
+#endif
diff --git a/drivers/md/dm-bio-list.h b/drivers/md/dm-bio-list.h
deleted file mode 100644
index d4509be0fe6..00000000000
--- a/drivers/md/dm-bio-list.h
+++ /dev/null
@@ -1,107 +0,0 @@
-/*
- * Copyright (C) 2004 Red Hat UK Ltd.
- *
- * This file is released under the GPL.
- */
-
-#ifndef DM_BIO_LIST_H
-#define DM_BIO_LIST_H
-
-#include <linux/bio.h>
-
-#ifdef CONFIG_BLOCK
-
-struct bio_list {
-	struct bio *head;
-	struct bio *tail;
-};
-
-static inline int bio_list_empty(const struct bio_list *bl)
-{
-	return bl->head == NULL;
-}
-
-static inline void bio_list_init(struct bio_list *bl)
-{
-	bl->head = bl->tail = NULL;
-}
-
-#define bio_list_for_each(bio, bl) \
-	for (bio = (bl)->head; bio; bio = bio->bi_next)
-
-static inline unsigned bio_list_size(const struct bio_list *bl)
-{
-	unsigned sz = 0;
-	struct bio *bio;
-
-	bio_list_for_each(bio, bl)
-		sz++;
-
-	return sz;
-}
-
-static inline void bio_list_add(struct bio_list *bl, struct bio *bio)
-{
-	bio->bi_next = NULL;
-
-	if (bl->tail)
-		bl->tail->bi_next = bio;
-	else
-		bl->head = bio;
-
-	bl->tail = bio;
-}
-
-static inline void bio_list_merge(struct bio_list *bl, struct bio_list *bl2)
-{
-	if (!bl2->head)
-		return;
-
-	if (bl->tail)
-		bl->tail->bi_next = bl2->head;
-	else
-		bl->head = bl2->head;
-
-	bl->tail = bl2->tail;
-}
-
-static inline void bio_list_merge_head(struct bio_list *bl,
-				       struct bio_list *bl2)
-{
-	if (!bl2->head)
-		return;
-
-	if (bl->head)
-		bl2->tail->bi_next = bl->head;
-	else
-		bl->tail = bl2->tail;
-
-	bl->head = bl2->head;
-}
-
-static inline struct bio *bio_list_pop(struct bio_list *bl)
-{
-	struct bio *bio = bl->head;
-
-	if (bio) {
-		bl->head = bl->head->bi_next;
-		if (!bl->head)
-			bl->tail = NULL;
-
-		bio->bi_next = NULL;
-	}
-
-	return bio;
-}
-
-static inline struct bio *bio_list_get(struct bio_list *bl)
-{
-	struct bio *bio = bl->head;
-
-	bl->head = bl->tail = NULL;
-
-	return bio;
-}
-
-#endif /* CONFIG_BLOCK */
-#endif
diff --git a/drivers/md/dm-bio-prison.c b/drivers/md/dm-bio-prison.c
new file mode 100644
index 00000000000..f752d12081f
--- /dev/null
+++ b/drivers/md/dm-bio-prison.c
@@ -0,0 +1,408 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm.h"
+#include "dm-bio-prison.h"
+
+#include <linux/spinlock.h>
+#include <linux/mempool.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+
+/*----------------------------------------------------------------*/
+
+struct bucket {
+	spinlock_t lock;
+	struct hlist_head cells;
+};
+
+struct dm_bio_prison {
+	mempool_t *cell_pool;
+
+	unsigned nr_buckets;
+	unsigned hash_mask;
+	struct bucket *buckets;
+};
+
+/*----------------------------------------------------------------*/
+
+static uint32_t calc_nr_buckets(unsigned nr_cells)
+{
+	uint32_t n = 128;
+
+	nr_cells /= 4;
+	nr_cells = min(nr_cells, 8192u);
+
+	while (n < nr_cells)
+		n <<= 1;
+
+	return n;
+}
+
+static struct kmem_cache *_cell_cache;
+
+static void init_bucket(struct bucket *b)
+{
+	spin_lock_init(&b->lock);
+	INIT_HLIST_HEAD(&b->cells);
+}
+
+/*
+ * @nr_cells should be the number of cells you want in use _concurrently_.
+ * Don't confuse it with the number of distinct keys.
+ */
+struct dm_bio_prison *dm_bio_prison_create(unsigned nr_cells)
+{
+	unsigned i;
+	uint32_t nr_buckets = calc_nr_buckets(nr_cells);
+	size_t len = sizeof(struct dm_bio_prison) +
+		(sizeof(struct bucket) * nr_buckets);
+	struct dm_bio_prison *prison = kmalloc(len, GFP_KERNEL);
+
+	if (!prison)
+		return NULL;
+
+	prison->cell_pool = mempool_create_slab_pool(nr_cells, _cell_cache);
+	if (!prison->cell_pool) {
+		kfree(prison);
+		return NULL;
+	}
+
+	prison->nr_buckets = nr_buckets;
+	prison->hash_mask = nr_buckets - 1;
+	prison->buckets = (struct bucket *) (prison + 1);
+	for (i = 0; i < nr_buckets; i++)
+		init_bucket(prison->buckets + i);
+
+	return prison;
+}
+EXPORT_SYMBOL_GPL(dm_bio_prison_create);
+
+void dm_bio_prison_destroy(struct dm_bio_prison *prison)
+{
+	mempool_destroy(prison->cell_pool);
+	kfree(prison);
+}
+EXPORT_SYMBOL_GPL(dm_bio_prison_destroy);
+
+struct dm_bio_prison_cell *dm_bio_prison_alloc_cell(struct dm_bio_prison *prison, gfp_t gfp)
+{
+	return mempool_alloc(prison->cell_pool, gfp);
+}
+EXPORT_SYMBOL_GPL(dm_bio_prison_alloc_cell);
+
+void dm_bio_prison_free_cell(struct dm_bio_prison *prison,
+			     struct dm_bio_prison_cell *cell)
+{
+	mempool_free(cell, prison->cell_pool);
+}
+EXPORT_SYMBOL_GPL(dm_bio_prison_free_cell);
+
+static uint32_t hash_key(struct dm_bio_prison *prison, struct dm_cell_key *key)
+{
+	const unsigned long BIG_PRIME = 4294967291UL;
+	uint64_t hash = key->block * BIG_PRIME;
+
+	return (uint32_t) (hash & prison->hash_mask);
+}
+
+static int keys_equal(struct dm_cell_key *lhs, struct dm_cell_key *rhs)
+{
+	       return (lhs->virtual == rhs->virtual) &&
+		       (lhs->dev == rhs->dev) &&
+		       (lhs->block == rhs->block);
+}
+
+static struct bucket *get_bucket(struct dm_bio_prison *prison,
+				 struct dm_cell_key *key)
+{
+	return prison->buckets + hash_key(prison, key);
+}
+
+static struct dm_bio_prison_cell *__search_bucket(struct bucket *b,
+						  struct dm_cell_key *key)
+{
+	struct dm_bio_prison_cell *cell;
+
+	hlist_for_each_entry(cell, &b->cells, list)
+		if (keys_equal(&cell->key, key))
+			return cell;
+
+	return NULL;
+}
+
+static void __setup_new_cell(struct bucket *b,
+			     struct dm_cell_key *key,
+			     struct bio *holder,
+			     struct dm_bio_prison_cell *cell)
+{
+	memcpy(&cell->key, key, sizeof(cell->key));
+	cell->holder = holder;
+	bio_list_init(&cell->bios);
+	hlist_add_head(&cell->list, &b->cells);
+}
+
+static int __bio_detain(struct bucket *b,
+			struct dm_cell_key *key,
+			struct bio *inmate,
+			struct dm_bio_prison_cell *cell_prealloc,
+			struct dm_bio_prison_cell **cell_result)
+{
+	struct dm_bio_prison_cell *cell;
+
+	cell = __search_bucket(b, key);
+	if (cell) {
+		if (inmate)
+			bio_list_add(&cell->bios, inmate);
+		*cell_result = cell;
+		return 1;
+	}
+
+	__setup_new_cell(b, key, inmate, cell_prealloc);
+	*cell_result = cell_prealloc;
+	return 0;
+}
+
+static int bio_detain(struct dm_bio_prison *prison,
+		      struct dm_cell_key *key,
+		      struct bio *inmate,
+		      struct dm_bio_prison_cell *cell_prealloc,
+		      struct dm_bio_prison_cell **cell_result)
+{
+	int r;
+	unsigned long flags;
+	struct bucket *b = get_bucket(prison, key);
+
+	spin_lock_irqsave(&b->lock, flags);
+	r = __bio_detain(b, key, inmate, cell_prealloc, cell_result);
+	spin_unlock_irqrestore(&b->lock, flags);
+
+	return r;
+}
+
+int dm_bio_detain(struct dm_bio_prison *prison,
+		  struct dm_cell_key *key,
+		  struct bio *inmate,
+		  struct dm_bio_prison_cell *cell_prealloc,
+		  struct dm_bio_prison_cell **cell_result)
+{
+	return bio_detain(prison, key, inmate, cell_prealloc, cell_result);
+}
+EXPORT_SYMBOL_GPL(dm_bio_detain);
+
+int dm_get_cell(struct dm_bio_prison *prison,
+		struct dm_cell_key *key,
+		struct dm_bio_prison_cell *cell_prealloc,
+		struct dm_bio_prison_cell **cell_result)
+{
+	return bio_detain(prison, key, NULL, cell_prealloc, cell_result);
+}
+EXPORT_SYMBOL_GPL(dm_get_cell);
+
+/*
+ * @inmates must have been initialised prior to this call
+ */
+static void __cell_release(struct dm_bio_prison_cell *cell,
+			   struct bio_list *inmates)
+{
+	hlist_del(&cell->list);
+
+	if (inmates) {
+		if (cell->holder)
+			bio_list_add(inmates, cell->holder);
+		bio_list_merge(inmates, &cell->bios);
+	}
+}
+
+void dm_cell_release(struct dm_bio_prison *prison,
+		     struct dm_bio_prison_cell *cell,
+		     struct bio_list *bios)
+{
+	unsigned long flags;
+	struct bucket *b = get_bucket(prison, &cell->key);
+
+	spin_lock_irqsave(&b->lock, flags);
+	__cell_release(cell, bios);
+	spin_unlock_irqrestore(&b->lock, flags);
+}
+EXPORT_SYMBOL_GPL(dm_cell_release);
+
+/*
+ * Sometimes we don't want the holder, just the additional bios.
+ */
+static void __cell_release_no_holder(struct dm_bio_prison_cell *cell,
+				     struct bio_list *inmates)
+{
+	hlist_del(&cell->list);
+	bio_list_merge(inmates, &cell->bios);
+}
+
+void dm_cell_release_no_holder(struct dm_bio_prison *prison,
+			       struct dm_bio_prison_cell *cell,
+			       struct bio_list *inmates)
+{
+	unsigned long flags;
+	struct bucket *b = get_bucket(prison, &cell->key);
+
+	spin_lock_irqsave(&b->lock, flags);
+	__cell_release_no_holder(cell, inmates);
+	spin_unlock_irqrestore(&b->lock, flags);
+}
+EXPORT_SYMBOL_GPL(dm_cell_release_no_holder);
+
+void dm_cell_error(struct dm_bio_prison *prison,
+		   struct dm_bio_prison_cell *cell, int error)
+{
+	struct bio_list bios;
+	struct bio *bio;
+
+	bio_list_init(&bios);
+	dm_cell_release(prison, cell, &bios);
+
+	while ((bio = bio_list_pop(&bios)))
+		bio_endio(bio, error);
+}
+EXPORT_SYMBOL_GPL(dm_cell_error);
+
+/*----------------------------------------------------------------*/
+
+#define DEFERRED_SET_SIZE 64
+
+struct dm_deferred_entry {
+	struct dm_deferred_set *ds;
+	unsigned count;
+	struct list_head work_items;
+};
+
+struct dm_deferred_set {
+	spinlock_t lock;
+	unsigned current_entry;
+	unsigned sweeper;
+	struct dm_deferred_entry entries[DEFERRED_SET_SIZE];
+};
+
+struct dm_deferred_set *dm_deferred_set_create(void)
+{
+	int i;
+	struct dm_deferred_set *ds;
+
+	ds = kmalloc(sizeof(*ds), GFP_KERNEL);
+	if (!ds)
+		return NULL;
+
+	spin_lock_init(&ds->lock);
+	ds->current_entry = 0;
+	ds->sweeper = 0;
+	for (i = 0; i < DEFERRED_SET_SIZE; i++) {
+		ds->entries[i].ds = ds;
+		ds->entries[i].count = 0;
+		INIT_LIST_HEAD(&ds->entries[i].work_items);
+	}
+
+	return ds;
+}
+EXPORT_SYMBOL_GPL(dm_deferred_set_create);
+
+void dm_deferred_set_destroy(struct dm_deferred_set *ds)
+{
+	kfree(ds);
+}
+EXPORT_SYMBOL_GPL(dm_deferred_set_destroy);
+
+struct dm_deferred_entry *dm_deferred_entry_inc(struct dm_deferred_set *ds)
+{
+	unsigned long flags;
+	struct dm_deferred_entry *entry;
+
+	spin_lock_irqsave(&ds->lock, flags);
+	entry = ds->entries + ds->current_entry;
+	entry->count++;
+	spin_unlock_irqrestore(&ds->lock, flags);
+
+	return entry;
+}
+EXPORT_SYMBOL_GPL(dm_deferred_entry_inc);
+
+static unsigned ds_next(unsigned index)
+{
+	return (index + 1) % DEFERRED_SET_SIZE;
+}
+
+static void __sweep(struct dm_deferred_set *ds, struct list_head *head)
+{
+	while ((ds->sweeper != ds->current_entry) &&
+	       !ds->entries[ds->sweeper].count) {
+		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
+		ds->sweeper = ds_next(ds->sweeper);
+	}
+
+	if ((ds->sweeper == ds->current_entry) && !ds->entries[ds->sweeper].count)
+		list_splice_init(&ds->entries[ds->sweeper].work_items, head);
+}
+
+void dm_deferred_entry_dec(struct dm_deferred_entry *entry, struct list_head *head)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&entry->ds->lock, flags);
+	BUG_ON(!entry->count);
+	--entry->count;
+	__sweep(entry->ds, head);
+	spin_unlock_irqrestore(&entry->ds->lock, flags);
+}
+EXPORT_SYMBOL_GPL(dm_deferred_entry_dec);
+
+/*
+ * Returns 1 if deferred or 0 if no pending items to delay job.
+ */
+int dm_deferred_set_add_work(struct dm_deferred_set *ds, struct list_head *work)
+{
+	int r = 1;
+	unsigned long flags;
+	unsigned next_entry;
+
+	spin_lock_irqsave(&ds->lock, flags);
+	if ((ds->sweeper == ds->current_entry) &&
+	    !ds->entries[ds->current_entry].count)
+		r = 0;
+	else {
+		list_add(work, &ds->entries[ds->current_entry].work_items);
+		next_entry = ds_next(ds->current_entry);
+		if (!ds->entries[next_entry].count)
+			ds->current_entry = next_entry;
+	}
+	spin_unlock_irqrestore(&ds->lock, flags);
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_deferred_set_add_work);
+
+/*----------------------------------------------------------------*/
+
+static int __init dm_bio_prison_init(void)
+{
+	_cell_cache = KMEM_CACHE(dm_bio_prison_cell, 0);
+	if (!_cell_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+static void __exit dm_bio_prison_exit(void)
+{
+	kmem_cache_destroy(_cell_cache);
+	_cell_cache = NULL;
+}
+
+/*
+ * module hooks
+ */
+module_init(dm_bio_prison_init);
+module_exit(dm_bio_prison_exit);
+
+MODULE_DESCRIPTION(DM_NAME " bio prison");
+MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-bio-prison.h b/drivers/md/dm-bio-prison.h
new file mode 100644
index 00000000000..6805a142b75
--- /dev/null
+++ b/drivers/md/dm-bio-prison.h
@@ -0,0 +1,111 @@
+/*
+ * Copyright (C) 2011-2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_BIO_PRISON_H
+#define DM_BIO_PRISON_H
+
+#include "persistent-data/dm-block-manager.h" /* FIXME: for dm_block_t */
+#include "dm-thin-metadata.h" /* FIXME: for dm_thin_id */
+
+#include <linux/list.h>
+#include <linux/bio.h>
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Sometimes we can't deal with a bio straight away.  We put them in prison
+ * where they can't cause any mischief.  Bios are put in a cell identified
+ * by a key, multiple bios can be in the same cell.  When the cell is
+ * subsequently unlocked the bios become available.
+ */
+struct dm_bio_prison;
+
+/* FIXME: this needs to be more abstract */
+struct dm_cell_key {
+	int virtual;
+	dm_thin_id dev;
+	dm_block_t block;
+};
+
+/*
+ * Treat this as opaque, only in header so callers can manage allocation
+ * themselves.
+ */
+struct dm_bio_prison_cell {
+	struct hlist_node list;
+	struct dm_cell_key key;
+	struct bio *holder;
+	struct bio_list bios;
+};
+
+struct dm_bio_prison *dm_bio_prison_create(unsigned nr_cells);
+void dm_bio_prison_destroy(struct dm_bio_prison *prison);
+
+/*
+ * These two functions just wrap a mempool.  This is a transitory step:
+ * Eventually all bio prison clients should manage their own cell memory.
+ *
+ * Like mempool_alloc(), dm_bio_prison_alloc_cell() can only fail if called
+ * in interrupt context or passed GFP_NOWAIT.
+ */
+struct dm_bio_prison_cell *dm_bio_prison_alloc_cell(struct dm_bio_prison *prison,
+						    gfp_t gfp);
+void dm_bio_prison_free_cell(struct dm_bio_prison *prison,
+			     struct dm_bio_prison_cell *cell);
+
+/*
+ * Creates, or retrieves a cell for the given key.
+ *
+ * Returns 1 if pre-existing cell returned, zero if new cell created using
+ * @cell_prealloc.
+ */
+int dm_get_cell(struct dm_bio_prison *prison,
+		struct dm_cell_key *key,
+		struct dm_bio_prison_cell *cell_prealloc,
+		struct dm_bio_prison_cell **cell_result);
+
+/*
+ * An atomic op that combines retrieving a cell, and adding a bio to it.
+ *
+ * Returns 1 if the cell was already held, 0 if @inmate is the new holder.
+ */
+int dm_bio_detain(struct dm_bio_prison *prison,
+		  struct dm_cell_key *key,
+		  struct bio *inmate,
+		  struct dm_bio_prison_cell *cell_prealloc,
+		  struct dm_bio_prison_cell **cell_result);
+
+void dm_cell_release(struct dm_bio_prison *prison,
+		     struct dm_bio_prison_cell *cell,
+		     struct bio_list *bios);
+void dm_cell_release_no_holder(struct dm_bio_prison *prison,
+			       struct dm_bio_prison_cell *cell,
+			       struct bio_list *inmates);
+void dm_cell_error(struct dm_bio_prison *prison,
+		   struct dm_bio_prison_cell *cell, int error);
+
+/*----------------------------------------------------------------*/
+
+/*
+ * We use the deferred set to keep track of pending reads to shared blocks.
+ * We do this to ensure the new mapping caused by a write isn't performed
+ * until these prior reads have completed.  Otherwise the insertion of the
+ * new mapping could free the old block that the read bios are mapped to.
+ */
+
+struct dm_deferred_set;
+struct dm_deferred_entry;
+
+struct dm_deferred_set *dm_deferred_set_create(void);
+void dm_deferred_set_destroy(struct dm_deferred_set *ds);
+
+struct dm_deferred_entry *dm_deferred_entry_inc(struct dm_deferred_set *ds);
+void dm_deferred_entry_dec(struct dm_deferred_entry *entry, struct list_head *head);
+int dm_deferred_set_add_work(struct dm_deferred_set *ds, struct list_head *work);
+
+/*----------------------------------------------------------------*/
+
+#endif
diff --git a/drivers/md/dm-bio-record.h b/drivers/md/dm-bio-record.h
index d3ec217847d..dd364611156 100644
--- a/drivers/md/dm-bio-record.h
+++ b/drivers/md/dm-bio-record.h
@@ -16,30 +16,25 @@
  * functions in this file help the target record and restore the
  * original bio state.
  */
+
 struct dm_bio_details {
-	sector_t bi_sector;
 	struct block_device *bi_bdev;
-	unsigned int bi_size;
-	unsigned short bi_idx;
 	unsigned long bi_flags;
+	struct bvec_iter bi_iter;
 };
 
 static inline void dm_bio_record(struct dm_bio_details *bd, struct bio *bio)
 {
-	bd->bi_sector = bio->bi_sector;
 	bd->bi_bdev = bio->bi_bdev;
-	bd->bi_size = bio->bi_size;
-	bd->bi_idx = bio->bi_idx;
 	bd->bi_flags = bio->bi_flags;
+	bd->bi_iter = bio->bi_iter;
 }
 
 static inline void dm_bio_restore(struct dm_bio_details *bd, struct bio *bio)
 {
-	bio->bi_sector = bd->bi_sector;
 	bio->bi_bdev = bd->bi_bdev;
-	bio->bi_size = bd->bi_size;
-	bio->bi_idx = bd->bi_idx;
 	bio->bi_flags = bd->bi_flags;
+	bio->bi_iter = bd->bi_iter;
 }
 
 #endif
diff --git a/drivers/md/dm-bufio.c b/drivers/md/dm-bufio.c
new file mode 100644
index 00000000000..d724459860d
--- /dev/null
+++ b/drivers/md/dm-bufio.c
@@ -0,0 +1,1867 @@
+/*
+ * Copyright (C) 2009-2011 Red Hat, Inc.
+ *
+ * Author: Mikulas Patocka <mpatocka@redhat.com>
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-bufio.h"
+
+#include <linux/device-mapper.h>
+#include <linux/dm-io.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+#include <linux/shrinker.h>
+#include <linux/module.h>
+
+#define DM_MSG_PREFIX "bufio"
+
+/*
+ * Memory management policy:
+ *	Limit the number of buffers to DM_BUFIO_MEMORY_PERCENT of main memory
+ *	or DM_BUFIO_VMALLOC_PERCENT of vmalloc memory (whichever is lower).
+ *	Always allocate at least DM_BUFIO_MIN_BUFFERS buffers.
+ *	Start background writeback when there are DM_BUFIO_WRITEBACK_PERCENT
+ *	dirty buffers.
+ */
+#define DM_BUFIO_MIN_BUFFERS		8
+
+#define DM_BUFIO_MEMORY_PERCENT		2
+#define DM_BUFIO_VMALLOC_PERCENT	25
+#define DM_BUFIO_WRITEBACK_PERCENT	75
+
+/*
+ * Check buffer ages in this interval (seconds)
+ */
+#define DM_BUFIO_WORK_TIMER_SECS	10
+
+/*
+ * Free buffers when they are older than this (seconds)
+ */
+#define DM_BUFIO_DEFAULT_AGE_SECS	60
+
+/*
+ * The number of bvec entries that are embedded directly in the buffer.
+ * If the chunk size is larger, dm-io is used to do the io.
+ */
+#define DM_BUFIO_INLINE_VECS		16
+
+/*
+ * Buffer hash
+ */
+#define DM_BUFIO_HASH_BITS	20
+#define DM_BUFIO_HASH(block) \
+	((((block) >> DM_BUFIO_HASH_BITS) ^ (block)) & \
+	 ((1 << DM_BUFIO_HASH_BITS) - 1))
+
+/*
+ * Don't try to use kmem_cache_alloc for blocks larger than this.
+ * For explanation, see alloc_buffer_data below.
+ */
+#define DM_BUFIO_BLOCK_SIZE_SLAB_LIMIT	(PAGE_SIZE >> 1)
+#define DM_BUFIO_BLOCK_SIZE_GFP_LIMIT	(PAGE_SIZE << (MAX_ORDER - 1))
+
+/*
+ * dm_buffer->list_mode
+ */
+#define LIST_CLEAN	0
+#define LIST_DIRTY	1
+#define LIST_SIZE	2
+
+/*
+ * Linking of buffers:
+ *	All buffers are linked to cache_hash with their hash_list field.
+ *
+ *	Clean buffers that are not being written (B_WRITING not set)
+ *	are linked to lru[LIST_CLEAN] with their lru_list field.
+ *
+ *	Dirty and clean buffers that are being written are linked to
+ *	lru[LIST_DIRTY] with their lru_list field. When the write
+ *	finishes, the buffer cannot be relinked immediately (because we
+ *	are in an interrupt context and relinking requires process
+ *	context), so some clean-not-writing buffers can be held on
+ *	dirty_lru too.  They are later added to lru in the process
+ *	context.
+ */
+struct dm_bufio_client {
+	struct mutex lock;
+
+	struct list_head lru[LIST_SIZE];
+	unsigned long n_buffers[LIST_SIZE];
+
+	struct block_device *bdev;
+	unsigned block_size;
+	unsigned char sectors_per_block_bits;
+	unsigned char pages_per_block_bits;
+	unsigned char blocks_per_page_bits;
+	unsigned aux_size;
+	void (*alloc_callback)(struct dm_buffer *);
+	void (*write_callback)(struct dm_buffer *);
+
+	struct dm_io_client *dm_io;
+
+	struct list_head reserved_buffers;
+	unsigned need_reserved_buffers;
+
+	unsigned minimum_buffers;
+
+	struct hlist_head *cache_hash;
+	wait_queue_head_t free_buffer_wait;
+
+	int async_write_error;
+
+	struct list_head client_list;
+	struct shrinker shrinker;
+};
+
+/*
+ * Buffer state bits.
+ */
+#define B_READING	0
+#define B_WRITING	1
+#define B_DIRTY		2
+
+/*
+ * Describes how the block was allocated:
+ * kmem_cache_alloc(), __get_free_pages() or vmalloc().
+ * See the comment at alloc_buffer_data.
+ */
+enum data_mode {
+	DATA_MODE_SLAB = 0,
+	DATA_MODE_GET_FREE_PAGES = 1,
+	DATA_MODE_VMALLOC = 2,
+	DATA_MODE_LIMIT = 3
+};
+
+struct dm_buffer {
+	struct hlist_node hash_list;
+	struct list_head lru_list;
+	sector_t block;
+	void *data;
+	enum data_mode data_mode;
+	unsigned char list_mode;		/* LIST_* */
+	unsigned hold_count;
+	int read_error;
+	int write_error;
+	unsigned long state;
+	unsigned long last_accessed;
+	struct dm_bufio_client *c;
+	struct list_head write_list;
+	struct bio bio;
+	struct bio_vec bio_vec[DM_BUFIO_INLINE_VECS];
+};
+
+/*----------------------------------------------------------------*/
+
+static struct kmem_cache *dm_bufio_caches[PAGE_SHIFT - SECTOR_SHIFT];
+static char *dm_bufio_cache_names[PAGE_SHIFT - SECTOR_SHIFT];
+
+static inline int dm_bufio_cache_index(struct dm_bufio_client *c)
+{
+	unsigned ret = c->blocks_per_page_bits - 1;
+
+	BUG_ON(ret >= ARRAY_SIZE(dm_bufio_caches));
+
+	return ret;
+}
+
+#define DM_BUFIO_CACHE(c)	(dm_bufio_caches[dm_bufio_cache_index(c)])
+#define DM_BUFIO_CACHE_NAME(c)	(dm_bufio_cache_names[dm_bufio_cache_index(c)])
+
+#define dm_bufio_in_request()	(!!current->bio_list)
+
+static void dm_bufio_lock(struct dm_bufio_client *c)
+{
+	mutex_lock_nested(&c->lock, dm_bufio_in_request());
+}
+
+static int dm_bufio_trylock(struct dm_bufio_client *c)
+{
+	return mutex_trylock(&c->lock);
+}
+
+static void dm_bufio_unlock(struct dm_bufio_client *c)
+{
+	mutex_unlock(&c->lock);
+}
+
+/*
+ * FIXME Move to sched.h?
+ */
+#ifdef CONFIG_PREEMPT_VOLUNTARY
+#  define dm_bufio_cond_resched()		\
+do {						\
+	if (unlikely(need_resched()))		\
+		_cond_resched();		\
+} while (0)
+#else
+#  define dm_bufio_cond_resched()                do { } while (0)
+#endif
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Default cache size: available memory divided by the ratio.
+ */
+static unsigned long dm_bufio_default_cache_size;
+
+/*
+ * Total cache size set by the user.
+ */
+static unsigned long dm_bufio_cache_size;
+
+/*
+ * A copy of dm_bufio_cache_size because dm_bufio_cache_size can change
+ * at any time.  If it disagrees, the user has changed cache size.
+ */
+static unsigned long dm_bufio_cache_size_latch;
+
+static DEFINE_SPINLOCK(param_spinlock);
+
+/*
+ * Buffers are freed after this timeout
+ */
+static unsigned dm_bufio_max_age = DM_BUFIO_DEFAULT_AGE_SECS;
+
+static unsigned long dm_bufio_peak_allocated;
+static unsigned long dm_bufio_allocated_kmem_cache;
+static unsigned long dm_bufio_allocated_get_free_pages;
+static unsigned long dm_bufio_allocated_vmalloc;
+static unsigned long dm_bufio_current_allocated;
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Per-client cache: dm_bufio_cache_size / dm_bufio_client_count
+ */
+static unsigned long dm_bufio_cache_size_per_client;
+
+/*
+ * The current number of clients.
+ */
+static int dm_bufio_client_count;
+
+/*
+ * The list of all clients.
+ */
+static LIST_HEAD(dm_bufio_all_clients);
+
+/*
+ * This mutex protects dm_bufio_cache_size_latch,
+ * dm_bufio_cache_size_per_client and dm_bufio_client_count
+ */
+static DEFINE_MUTEX(dm_bufio_clients_lock);
+
+/*----------------------------------------------------------------*/
+
+static void adjust_total_allocated(enum data_mode data_mode, long diff)
+{
+	static unsigned long * const class_ptr[DATA_MODE_LIMIT] = {
+		&dm_bufio_allocated_kmem_cache,
+		&dm_bufio_allocated_get_free_pages,
+		&dm_bufio_allocated_vmalloc,
+	};
+
+	spin_lock(&param_spinlock);
+
+	*class_ptr[data_mode] += diff;
+
+	dm_bufio_current_allocated += diff;
+
+	if (dm_bufio_current_allocated > dm_bufio_peak_allocated)
+		dm_bufio_peak_allocated = dm_bufio_current_allocated;
+
+	spin_unlock(&param_spinlock);
+}
+
+/*
+ * Change the number of clients and recalculate per-client limit.
+ */
+static void __cache_size_refresh(void)
+{
+	BUG_ON(!mutex_is_locked(&dm_bufio_clients_lock));
+	BUG_ON(dm_bufio_client_count < 0);
+
+	dm_bufio_cache_size_latch = ACCESS_ONCE(dm_bufio_cache_size);
+
+	/*
+	 * Use default if set to 0 and report the actual cache size used.
+	 */
+	if (!dm_bufio_cache_size_latch) {
+		(void)cmpxchg(&dm_bufio_cache_size, 0,
+			      dm_bufio_default_cache_size);
+		dm_bufio_cache_size_latch = dm_bufio_default_cache_size;
+	}
+
+	dm_bufio_cache_size_per_client = dm_bufio_cache_size_latch /
+					 (dm_bufio_client_count ? : 1);
+}
+
+/*
+ * Allocating buffer data.
+ *
+ * Small buffers are allocated with kmem_cache, to use space optimally.
+ *
+ * For large buffers, we choose between get_free_pages and vmalloc.
+ * Each has advantages and disadvantages.
+ *
+ * __get_free_pages can randomly fail if the memory is fragmented.
+ * __vmalloc won't randomly fail, but vmalloc space is limited (it may be
+ * as low as 128M) so using it for caching is not appropriate.
+ *
+ * If the allocation may fail we use __get_free_pages. Memory fragmentation
+ * won't have a fatal effect here, but it just causes flushes of some other
+ * buffers and more I/O will be performed. Don't use __get_free_pages if it
+ * always fails (i.e. order >= MAX_ORDER).
+ *
+ * If the allocation shouldn't fail we use __vmalloc. This is only for the
+ * initial reserve allocation, so there's no risk of wasting all vmalloc
+ * space.
+ */
+static void *alloc_buffer_data(struct dm_bufio_client *c, gfp_t gfp_mask,
+			       enum data_mode *data_mode)
+{
+	unsigned noio_flag;
+	void *ptr;
+
+	if (c->block_size <= DM_BUFIO_BLOCK_SIZE_SLAB_LIMIT) {
+		*data_mode = DATA_MODE_SLAB;
+		return kmem_cache_alloc(DM_BUFIO_CACHE(c), gfp_mask);
+	}
+
+	if (c->block_size <= DM_BUFIO_BLOCK_SIZE_GFP_LIMIT &&
+	    gfp_mask & __GFP_NORETRY) {
+		*data_mode = DATA_MODE_GET_FREE_PAGES;
+		return (void *)__get_free_pages(gfp_mask,
+						c->pages_per_block_bits);
+	}
+
+	*data_mode = DATA_MODE_VMALLOC;
+
+	/*
+	 * __vmalloc allocates the data pages and auxiliary structures with
+	 * gfp_flags that were specified, but pagetables are always allocated
+	 * with GFP_KERNEL, no matter what was specified as gfp_mask.
+	 *
+	 * Consequently, we must set per-process flag PF_MEMALLOC_NOIO so that
+	 * all allocations done by this process (including pagetables) are done
+	 * as if GFP_NOIO was specified.
+	 */
+
+	if (gfp_mask & __GFP_NORETRY)
+		noio_flag = memalloc_noio_save();
+
+	ptr = __vmalloc(c->block_size, gfp_mask | __GFP_HIGHMEM, PAGE_KERNEL);
+
+	if (gfp_mask & __GFP_NORETRY)
+		memalloc_noio_restore(noio_flag);
+
+	return ptr;
+}
+
+/*
+ * Free buffer's data.
+ */
+static void free_buffer_data(struct dm_bufio_client *c,
+			     void *data, enum data_mode data_mode)
+{
+	switch (data_mode) {
+	case DATA_MODE_SLAB:
+		kmem_cache_free(DM_BUFIO_CACHE(c), data);
+		break;
+
+	case DATA_MODE_GET_FREE_PAGES:
+		free_pages((unsigned long)data, c->pages_per_block_bits);
+		break;
+
+	case DATA_MODE_VMALLOC:
+		vfree(data);
+		break;
+
+	default:
+		DMCRIT("dm_bufio_free_buffer_data: bad data mode: %d",
+		       data_mode);
+		BUG();
+	}
+}
+
+/*
+ * Allocate buffer and its data.
+ */
+static struct dm_buffer *alloc_buffer(struct dm_bufio_client *c, gfp_t gfp_mask)
+{
+	struct dm_buffer *b = kmalloc(sizeof(struct dm_buffer) + c->aux_size,
+				      gfp_mask);
+
+	if (!b)
+		return NULL;
+
+	b->c = c;
+
+	b->data = alloc_buffer_data(c, gfp_mask, &b->data_mode);
+	if (!b->data) {
+		kfree(b);
+		return NULL;
+	}
+
+	adjust_total_allocated(b->data_mode, (long)c->block_size);
+
+	return b;
+}
+
+/*
+ * Free buffer and its data.
+ */
+static void free_buffer(struct dm_buffer *b)
+{
+	struct dm_bufio_client *c = b->c;
+
+	adjust_total_allocated(b->data_mode, -(long)c->block_size);
+
+	free_buffer_data(c, b->data, b->data_mode);
+	kfree(b);
+}
+
+/*
+ * Link buffer to the hash list and clean or dirty queue.
+ */
+static void __link_buffer(struct dm_buffer *b, sector_t block, int dirty)
+{
+	struct dm_bufio_client *c = b->c;
+
+	c->n_buffers[dirty]++;
+	b->block = block;
+	b->list_mode = dirty;
+	list_add(&b->lru_list, &c->lru[dirty]);
+	hlist_add_head(&b->hash_list, &c->cache_hash[DM_BUFIO_HASH(block)]);
+	b->last_accessed = jiffies;
+}
+
+/*
+ * Unlink buffer from the hash list and dirty or clean queue.
+ */
+static void __unlink_buffer(struct dm_buffer *b)
+{
+	struct dm_bufio_client *c = b->c;
+
+	BUG_ON(!c->n_buffers[b->list_mode]);
+
+	c->n_buffers[b->list_mode]--;
+	hlist_del(&b->hash_list);
+	list_del(&b->lru_list);
+}
+
+/*
+ * Place the buffer to the head of dirty or clean LRU queue.
+ */
+static void __relink_lru(struct dm_buffer *b, int dirty)
+{
+	struct dm_bufio_client *c = b->c;
+
+	BUG_ON(!c->n_buffers[b->list_mode]);
+
+	c->n_buffers[b->list_mode]--;
+	c->n_buffers[dirty]++;
+	b->list_mode = dirty;
+	list_move(&b->lru_list, &c->lru[dirty]);
+}
+
+/*----------------------------------------------------------------
+ * Submit I/O on the buffer.
+ *
+ * Bio interface is faster but it has some problems:
+ *	the vector list is limited (increasing this limit increases
+ *	memory-consumption per buffer, so it is not viable);
+ *
+ *	the memory must be direct-mapped, not vmalloced;
+ *
+ *	the I/O driver can reject requests spuriously if it thinks that
+ *	the requests are too big for the device or if they cross a
+ *	controller-defined memory boundary.
+ *
+ * If the buffer is small enough (up to DM_BUFIO_INLINE_VECS pages) and
+ * it is not vmalloced, try using the bio interface.
+ *
+ * If the buffer is big, if it is vmalloced or if the underlying device
+ * rejects the bio because it is too large, use dm-io layer to do the I/O.
+ * The dm-io layer splits the I/O into multiple requests, avoiding the above
+ * shortcomings.
+ *--------------------------------------------------------------*/
+
+/*
+ * dm-io completion routine. It just calls b->bio.bi_end_io, pretending
+ * that the request was handled directly with bio interface.
+ */
+static void dmio_complete(unsigned long error, void *context)
+{
+	struct dm_buffer *b = context;
+
+	b->bio.bi_end_io(&b->bio, error ? -EIO : 0);
+}
+
+static void use_dmio(struct dm_buffer *b, int rw, sector_t block,
+		     bio_end_io_t *end_io)
+{
+	int r;
+	struct dm_io_request io_req = {
+		.bi_rw = rw,
+		.notify.fn = dmio_complete,
+		.notify.context = b,
+		.client = b->c->dm_io,
+	};
+	struct dm_io_region region = {
+		.bdev = b->c->bdev,
+		.sector = block << b->c->sectors_per_block_bits,
+		.count = b->c->block_size >> SECTOR_SHIFT,
+	};
+
+	if (b->data_mode != DATA_MODE_VMALLOC) {
+		io_req.mem.type = DM_IO_KMEM;
+		io_req.mem.ptr.addr = b->data;
+	} else {
+		io_req.mem.type = DM_IO_VMA;
+		io_req.mem.ptr.vma = b->data;
+	}
+
+	b->bio.bi_end_io = end_io;
+
+	r = dm_io(&io_req, 1, &region, NULL);
+	if (r)
+		end_io(&b->bio, r);
+}
+
+static void use_inline_bio(struct dm_buffer *b, int rw, sector_t block,
+			   bio_end_io_t *end_io)
+{
+	char *ptr;
+	int len;
+
+	bio_init(&b->bio);
+	b->bio.bi_io_vec = b->bio_vec;
+	b->bio.bi_max_vecs = DM_BUFIO_INLINE_VECS;
+	b->bio.bi_iter.bi_sector = block << b->c->sectors_per_block_bits;
+	b->bio.bi_bdev = b->c->bdev;
+	b->bio.bi_end_io = end_io;
+
+	/*
+	 * We assume that if len >= PAGE_SIZE ptr is page-aligned.
+	 * If len < PAGE_SIZE the buffer doesn't cross page boundary.
+	 */
+	ptr = b->data;
+	len = b->c->block_size;
+
+	if (len >= PAGE_SIZE)
+		BUG_ON((unsigned long)ptr & (PAGE_SIZE - 1));
+	else
+		BUG_ON((unsigned long)ptr & (len - 1));
+
+	do {
+		if (!bio_add_page(&b->bio, virt_to_page(ptr),
+				  len < PAGE_SIZE ? len : PAGE_SIZE,
+				  virt_to_phys(ptr) & (PAGE_SIZE - 1))) {
+			BUG_ON(b->c->block_size <= PAGE_SIZE);
+			use_dmio(b, rw, block, end_io);
+			return;
+		}
+
+		len -= PAGE_SIZE;
+		ptr += PAGE_SIZE;
+	} while (len > 0);
+
+	submit_bio(rw, &b->bio);
+}
+
+static void submit_io(struct dm_buffer *b, int rw, sector_t block,
+		      bio_end_io_t *end_io)
+{
+	if (rw == WRITE && b->c->write_callback)
+		b->c->write_callback(b);
+
+	if (b->c->block_size <= DM_BUFIO_INLINE_VECS * PAGE_SIZE &&
+	    b->data_mode != DATA_MODE_VMALLOC)
+		use_inline_bio(b, rw, block, end_io);
+	else
+		use_dmio(b, rw, block, end_io);
+}
+
+/*----------------------------------------------------------------
+ * Writing dirty buffers
+ *--------------------------------------------------------------*/
+
+/*
+ * The endio routine for write.
+ *
+ * Set the error, clear B_WRITING bit and wake anyone who was waiting on
+ * it.
+ */
+static void write_endio(struct bio *bio, int error)
+{
+	struct dm_buffer *b = container_of(bio, struct dm_buffer, bio);
+
+	b->write_error = error;
+	if (unlikely(error)) {
+		struct dm_bufio_client *c = b->c;
+		(void)cmpxchg(&c->async_write_error, 0, error);
+	}
+
+	BUG_ON(!test_bit(B_WRITING, &b->state));
+
+	smp_mb__before_atomic();
+	clear_bit(B_WRITING, &b->state);
+	smp_mb__after_atomic();
+
+	wake_up_bit(&b->state, B_WRITING);
+}
+
+/*
+ * This function is called when wait_on_bit is actually waiting.
+ */
+static int do_io_schedule(void *word)
+{
+	io_schedule();
+
+	return 0;
+}
+
+/*
+ * Initiate a write on a dirty buffer, but don't wait for it.
+ *
+ * - If the buffer is not dirty, exit.
+ * - If there some previous write going on, wait for it to finish (we can't
+ *   have two writes on the same buffer simultaneously).
+ * - Submit our write and don't wait on it. We set B_WRITING indicating
+ *   that there is a write in progress.
+ */
+static void __write_dirty_buffer(struct dm_buffer *b,
+				 struct list_head *write_list)
+{
+	if (!test_bit(B_DIRTY, &b->state))
+		return;
+
+	clear_bit(B_DIRTY, &b->state);
+	wait_on_bit_lock(&b->state, B_WRITING,
+			 do_io_schedule, TASK_UNINTERRUPTIBLE);
+
+	if (!write_list)
+		submit_io(b, WRITE, b->block, write_endio);
+	else
+		list_add_tail(&b->write_list, write_list);
+}
+
+static void __flush_write_list(struct list_head *write_list)
+{
+	struct blk_plug plug;
+	blk_start_plug(&plug);
+	while (!list_empty(write_list)) {
+		struct dm_buffer *b =
+			list_entry(write_list->next, struct dm_buffer, write_list);
+		list_del(&b->write_list);
+		submit_io(b, WRITE, b->block, write_endio);
+		dm_bufio_cond_resched();
+	}
+	blk_finish_plug(&plug);
+}
+
+/*
+ * Wait until any activity on the buffer finishes.  Possibly write the
+ * buffer if it is dirty.  When this function finishes, there is no I/O
+ * running on the buffer and the buffer is not dirty.
+ */
+static void __make_buffer_clean(struct dm_buffer *b)
+{
+	BUG_ON(b->hold_count);
+
+	if (!b->state)	/* fast case */
+		return;
+
+	wait_on_bit(&b->state, B_READING, do_io_schedule, TASK_UNINTERRUPTIBLE);
+	__write_dirty_buffer(b, NULL);
+	wait_on_bit(&b->state, B_WRITING, do_io_schedule, TASK_UNINTERRUPTIBLE);
+}
+
+/*
+ * Find some buffer that is not held by anybody, clean it, unlink it and
+ * return it.
+ */
+static struct dm_buffer *__get_unclaimed_buffer(struct dm_bufio_client *c)
+{
+	struct dm_buffer *b;
+
+	list_for_each_entry_reverse(b, &c->lru[LIST_CLEAN], lru_list) {
+		BUG_ON(test_bit(B_WRITING, &b->state));
+		BUG_ON(test_bit(B_DIRTY, &b->state));
+
+		if (!b->hold_count) {
+			__make_buffer_clean(b);
+			__unlink_buffer(b);
+			return b;
+		}
+		dm_bufio_cond_resched();
+	}
+
+	list_for_each_entry_reverse(b, &c->lru[LIST_DIRTY], lru_list) {
+		BUG_ON(test_bit(B_READING, &b->state));
+
+		if (!b->hold_count) {
+			__make_buffer_clean(b);
+			__unlink_buffer(b);
+			return b;
+		}
+		dm_bufio_cond_resched();
+	}
+
+	return NULL;
+}
+
+/*
+ * Wait until some other threads free some buffer or release hold count on
+ * some buffer.
+ *
+ * This function is entered with c->lock held, drops it and regains it
+ * before exiting.
+ */
+static void __wait_for_free_buffer(struct dm_bufio_client *c)
+{
+	DECLARE_WAITQUEUE(wait, current);
+
+	add_wait_queue(&c->free_buffer_wait, &wait);
+	set_task_state(current, TASK_UNINTERRUPTIBLE);
+	dm_bufio_unlock(c);
+
+	io_schedule();
+
+	set_task_state(current, TASK_RUNNING);
+	remove_wait_queue(&c->free_buffer_wait, &wait);
+
+	dm_bufio_lock(c);
+}
+
+enum new_flag {
+	NF_FRESH = 0,
+	NF_READ = 1,
+	NF_GET = 2,
+	NF_PREFETCH = 3
+};
+
+/*
+ * Allocate a new buffer. If the allocation is not possible, wait until
+ * some other thread frees a buffer.
+ *
+ * May drop the lock and regain it.
+ */
+static struct dm_buffer *__alloc_buffer_wait_no_callback(struct dm_bufio_client *c, enum new_flag nf)
+{
+	struct dm_buffer *b;
+
+	/*
+	 * dm-bufio is resistant to allocation failures (it just keeps
+	 * one buffer reserved in cases all the allocations fail).
+	 * So set flags to not try too hard:
+	 *	GFP_NOIO: don't recurse into the I/O layer
+	 *	__GFP_NORETRY: don't retry and rather return failure
+	 *	__GFP_NOMEMALLOC: don't use emergency reserves
+	 *	__GFP_NOWARN: don't print a warning in case of failure
+	 *
+	 * For debugging, if we set the cache size to 1, no new buffers will
+	 * be allocated.
+	 */
+	while (1) {
+		if (dm_bufio_cache_size_latch != 1) {
+			b = alloc_buffer(c, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
+			if (b)
+				return b;
+		}
+
+		if (nf == NF_PREFETCH)
+			return NULL;
+
+		if (!list_empty(&c->reserved_buffers)) {
+			b = list_entry(c->reserved_buffers.next,
+				       struct dm_buffer, lru_list);
+			list_del(&b->lru_list);
+			c->need_reserved_buffers++;
+
+			return b;
+		}
+
+		b = __get_unclaimed_buffer(c);
+		if (b)
+			return b;
+
+		__wait_for_free_buffer(c);
+	}
+}
+
+static struct dm_buffer *__alloc_buffer_wait(struct dm_bufio_client *c, enum new_flag nf)
+{
+	struct dm_buffer *b = __alloc_buffer_wait_no_callback(c, nf);
+
+	if (!b)
+		return NULL;
+
+	if (c->alloc_callback)
+		c->alloc_callback(b);
+
+	return b;
+}
+
+/*
+ * Free a buffer and wake other threads waiting for free buffers.
+ */
+static void __free_buffer_wake(struct dm_buffer *b)
+{
+	struct dm_bufio_client *c = b->c;
+
+	if (!c->need_reserved_buffers)
+		free_buffer(b);
+	else {
+		list_add(&b->lru_list, &c->reserved_buffers);
+		c->need_reserved_buffers--;
+	}
+
+	wake_up(&c->free_buffer_wait);
+}
+
+static void __write_dirty_buffers_async(struct dm_bufio_client *c, int no_wait,
+					struct list_head *write_list)
+{
+	struct dm_buffer *b, *tmp;
+
+	list_for_each_entry_safe_reverse(b, tmp, &c->lru[LIST_DIRTY], lru_list) {
+		BUG_ON(test_bit(B_READING, &b->state));
+
+		if (!test_bit(B_DIRTY, &b->state) &&
+		    !test_bit(B_WRITING, &b->state)) {
+			__relink_lru(b, LIST_CLEAN);
+			continue;
+		}
+
+		if (no_wait && test_bit(B_WRITING, &b->state))
+			return;
+
+		__write_dirty_buffer(b, write_list);
+		dm_bufio_cond_resched();
+	}
+}
+
+/*
+ * Get writeback threshold and buffer limit for a given client.
+ */
+static void __get_memory_limit(struct dm_bufio_client *c,
+			       unsigned long *threshold_buffers,
+			       unsigned long *limit_buffers)
+{
+	unsigned long buffers;
+
+	if (ACCESS_ONCE(dm_bufio_cache_size) != dm_bufio_cache_size_latch) {
+		mutex_lock(&dm_bufio_clients_lock);
+		__cache_size_refresh();
+		mutex_unlock(&dm_bufio_clients_lock);
+	}
+
+	buffers = dm_bufio_cache_size_per_client >>
+		  (c->sectors_per_block_bits + SECTOR_SHIFT);
+
+	if (buffers < c->minimum_buffers)
+		buffers = c->minimum_buffers;
+
+	*limit_buffers = buffers;
+	*threshold_buffers = buffers * DM_BUFIO_WRITEBACK_PERCENT / 100;
+}
+
+/*
+ * Check if we're over watermark.
+ * If we are over threshold_buffers, start freeing buffers.
+ * If we're over "limit_buffers", block until we get under the limit.
+ */
+static void __check_watermark(struct dm_bufio_client *c,
+			      struct list_head *write_list)
+{
+	unsigned long threshold_buffers, limit_buffers;
+
+	__get_memory_limit(c, &threshold_buffers, &limit_buffers);
+
+	while (c->n_buffers[LIST_CLEAN] + c->n_buffers[LIST_DIRTY] >
+	       limit_buffers) {
+
+		struct dm_buffer *b = __get_unclaimed_buffer(c);
+
+		if (!b)
+			return;
+
+		__free_buffer_wake(b);
+		dm_bufio_cond_resched();
+	}
+
+	if (c->n_buffers[LIST_DIRTY] > threshold_buffers)
+		__write_dirty_buffers_async(c, 1, write_list);
+}
+
+/*
+ * Find a buffer in the hash.
+ */
+static struct dm_buffer *__find(struct dm_bufio_client *c, sector_t block)
+{
+	struct dm_buffer *b;
+
+	hlist_for_each_entry(b, &c->cache_hash[DM_BUFIO_HASH(block)],
+			     hash_list) {
+		dm_bufio_cond_resched();
+		if (b->block == block)
+			return b;
+	}
+
+	return NULL;
+}
+
+/*----------------------------------------------------------------
+ * Getting a buffer
+ *--------------------------------------------------------------*/
+
+static struct dm_buffer *__bufio_new(struct dm_bufio_client *c, sector_t block,
+				     enum new_flag nf, int *need_submit,
+				     struct list_head *write_list)
+{
+	struct dm_buffer *b, *new_b = NULL;
+
+	*need_submit = 0;
+
+	b = __find(c, block);
+	if (b)
+		goto found_buffer;
+
+	if (nf == NF_GET)
+		return NULL;
+
+	new_b = __alloc_buffer_wait(c, nf);
+	if (!new_b)
+		return NULL;
+
+	/*
+	 * We've had a period where the mutex was unlocked, so need to
+	 * recheck the hash table.
+	 */
+	b = __find(c, block);
+	if (b) {
+		__free_buffer_wake(new_b);
+		goto found_buffer;
+	}
+
+	__check_watermark(c, write_list);
+
+	b = new_b;
+	b->hold_count = 1;
+	b->read_error = 0;
+	b->write_error = 0;
+	__link_buffer(b, block, LIST_CLEAN);
+
+	if (nf == NF_FRESH) {
+		b->state = 0;
+		return b;
+	}
+
+	b->state = 1 << B_READING;
+	*need_submit = 1;
+
+	return b;
+
+found_buffer:
+	if (nf == NF_PREFETCH)
+		return NULL;
+	/*
+	 * Note: it is essential that we don't wait for the buffer to be
+	 * read if dm_bufio_get function is used. Both dm_bufio_get and
+	 * dm_bufio_prefetch can be used in the driver request routine.
+	 * If the user called both dm_bufio_prefetch and dm_bufio_get on
+	 * the same buffer, it would deadlock if we waited.
+	 */
+	if (nf == NF_GET && unlikely(test_bit(B_READING, &b->state)))
+		return NULL;
+
+	b->hold_count++;
+	__relink_lru(b, test_bit(B_DIRTY, &b->state) ||
+		     test_bit(B_WRITING, &b->state));
+	return b;
+}
+
+/*
+ * The endio routine for reading: set the error, clear the bit and wake up
+ * anyone waiting on the buffer.
+ */
+static void read_endio(struct bio *bio, int error)
+{
+	struct dm_buffer *b = container_of(bio, struct dm_buffer, bio);
+
+	b->read_error = error;
+
+	BUG_ON(!test_bit(B_READING, &b->state));
+
+	smp_mb__before_atomic();
+	clear_bit(B_READING, &b->state);
+	smp_mb__after_atomic();
+
+	wake_up_bit(&b->state, B_READING);
+}
+
+/*
+ * A common routine for dm_bufio_new and dm_bufio_read.  Operation of these
+ * functions is similar except that dm_bufio_new doesn't read the
+ * buffer from the disk (assuming that the caller overwrites all the data
+ * and uses dm_bufio_mark_buffer_dirty to write new data back).
+ */
+static void *new_read(struct dm_bufio_client *c, sector_t block,
+		      enum new_flag nf, struct dm_buffer **bp)
+{
+	int need_submit;
+	struct dm_buffer *b;
+
+	LIST_HEAD(write_list);
+
+	dm_bufio_lock(c);
+	b = __bufio_new(c, block, nf, &need_submit, &write_list);
+	dm_bufio_unlock(c);
+
+	__flush_write_list(&write_list);
+
+	if (!b)
+		return b;
+
+	if (need_submit)
+		submit_io(b, READ, b->block, read_endio);
+
+	wait_on_bit(&b->state, B_READING, do_io_schedule, TASK_UNINTERRUPTIBLE);
+
+	if (b->read_error) {
+		int error = b->read_error;
+
+		dm_bufio_release(b);
+
+		return ERR_PTR(error);
+	}
+
+	*bp = b;
+
+	return b->data;
+}
+
+void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
+		   struct dm_buffer **bp)
+{
+	return new_read(c, block, NF_GET, bp);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get);
+
+void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
+		    struct dm_buffer **bp)
+{
+	BUG_ON(dm_bufio_in_request());
+
+	return new_read(c, block, NF_READ, bp);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_read);
+
+void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
+		   struct dm_buffer **bp)
+{
+	BUG_ON(dm_bufio_in_request());
+
+	return new_read(c, block, NF_FRESH, bp);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_new);
+
+void dm_bufio_prefetch(struct dm_bufio_client *c,
+		       sector_t block, unsigned n_blocks)
+{
+	struct blk_plug plug;
+
+	LIST_HEAD(write_list);
+
+	BUG_ON(dm_bufio_in_request());
+
+	blk_start_plug(&plug);
+	dm_bufio_lock(c);
+
+	for (; n_blocks--; block++) {
+		int need_submit;
+		struct dm_buffer *b;
+		b = __bufio_new(c, block, NF_PREFETCH, &need_submit,
+				&write_list);
+		if (unlikely(!list_empty(&write_list))) {
+			dm_bufio_unlock(c);
+			blk_finish_plug(&plug);
+			__flush_write_list(&write_list);
+			blk_start_plug(&plug);
+			dm_bufio_lock(c);
+		}
+		if (unlikely(b != NULL)) {
+			dm_bufio_unlock(c);
+
+			if (need_submit)
+				submit_io(b, READ, b->block, read_endio);
+			dm_bufio_release(b);
+
+			dm_bufio_cond_resched();
+
+			if (!n_blocks)
+				goto flush_plug;
+			dm_bufio_lock(c);
+		}
+	}
+
+	dm_bufio_unlock(c);
+
+flush_plug:
+	blk_finish_plug(&plug);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_prefetch);
+
+void dm_bufio_release(struct dm_buffer *b)
+{
+	struct dm_bufio_client *c = b->c;
+
+	dm_bufio_lock(c);
+
+	BUG_ON(!b->hold_count);
+
+	b->hold_count--;
+	if (!b->hold_count) {
+		wake_up(&c->free_buffer_wait);
+
+		/*
+		 * If there were errors on the buffer, and the buffer is not
+		 * to be written, free the buffer. There is no point in caching
+		 * invalid buffer.
+		 */
+		if ((b->read_error || b->write_error) &&
+		    !test_bit(B_READING, &b->state) &&
+		    !test_bit(B_WRITING, &b->state) &&
+		    !test_bit(B_DIRTY, &b->state)) {
+			__unlink_buffer(b);
+			__free_buffer_wake(b);
+		}
+	}
+
+	dm_bufio_unlock(c);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_release);
+
+void dm_bufio_mark_buffer_dirty(struct dm_buffer *b)
+{
+	struct dm_bufio_client *c = b->c;
+
+	dm_bufio_lock(c);
+
+	BUG_ON(test_bit(B_READING, &b->state));
+
+	if (!test_and_set_bit(B_DIRTY, &b->state))
+		__relink_lru(b, LIST_DIRTY);
+
+	dm_bufio_unlock(c);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_mark_buffer_dirty);
+
+void dm_bufio_write_dirty_buffers_async(struct dm_bufio_client *c)
+{
+	LIST_HEAD(write_list);
+
+	BUG_ON(dm_bufio_in_request());
+
+	dm_bufio_lock(c);
+	__write_dirty_buffers_async(c, 0, &write_list);
+	dm_bufio_unlock(c);
+	__flush_write_list(&write_list);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_write_dirty_buffers_async);
+
+/*
+ * For performance, it is essential that the buffers are written asynchronously
+ * and simultaneously (so that the block layer can merge the writes) and then
+ * waited upon.
+ *
+ * Finally, we flush hardware disk cache.
+ */
+int dm_bufio_write_dirty_buffers(struct dm_bufio_client *c)
+{
+	int a, f;
+	unsigned long buffers_processed = 0;
+	struct dm_buffer *b, *tmp;
+
+	LIST_HEAD(write_list);
+
+	dm_bufio_lock(c);
+	__write_dirty_buffers_async(c, 0, &write_list);
+	dm_bufio_unlock(c);
+	__flush_write_list(&write_list);
+	dm_bufio_lock(c);
+
+again:
+	list_for_each_entry_safe_reverse(b, tmp, &c->lru[LIST_DIRTY], lru_list) {
+		int dropped_lock = 0;
+
+		if (buffers_processed < c->n_buffers[LIST_DIRTY])
+			buffers_processed++;
+
+		BUG_ON(test_bit(B_READING, &b->state));
+
+		if (test_bit(B_WRITING, &b->state)) {
+			if (buffers_processed < c->n_buffers[LIST_DIRTY]) {
+				dropped_lock = 1;
+				b->hold_count++;
+				dm_bufio_unlock(c);
+				wait_on_bit(&b->state, B_WRITING,
+					    do_io_schedule,
+					    TASK_UNINTERRUPTIBLE);
+				dm_bufio_lock(c);
+				b->hold_count--;
+			} else
+				wait_on_bit(&b->state, B_WRITING,
+					    do_io_schedule,
+					    TASK_UNINTERRUPTIBLE);
+		}
+
+		if (!test_bit(B_DIRTY, &b->state) &&
+		    !test_bit(B_WRITING, &b->state))
+			__relink_lru(b, LIST_CLEAN);
+
+		dm_bufio_cond_resched();
+
+		/*
+		 * If we dropped the lock, the list is no longer consistent,
+		 * so we must restart the search.
+		 *
+		 * In the most common case, the buffer just processed is
+		 * relinked to the clean list, so we won't loop scanning the
+		 * same buffer again and again.
+		 *
+		 * This may livelock if there is another thread simultaneously
+		 * dirtying buffers, so we count the number of buffers walked
+		 * and if it exceeds the total number of buffers, it means that
+		 * someone is doing some writes simultaneously with us.  In
+		 * this case, stop, dropping the lock.
+		 */
+		if (dropped_lock)
+			goto again;
+	}
+	wake_up(&c->free_buffer_wait);
+	dm_bufio_unlock(c);
+
+	a = xchg(&c->async_write_error, 0);
+	f = dm_bufio_issue_flush(c);
+	if (a)
+		return a;
+
+	return f;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_write_dirty_buffers);
+
+/*
+ * Use dm-io to send and empty barrier flush the device.
+ */
+int dm_bufio_issue_flush(struct dm_bufio_client *c)
+{
+	struct dm_io_request io_req = {
+		.bi_rw = WRITE_FLUSH,
+		.mem.type = DM_IO_KMEM,
+		.mem.ptr.addr = NULL,
+		.client = c->dm_io,
+	};
+	struct dm_io_region io_reg = {
+		.bdev = c->bdev,
+		.sector = 0,
+		.count = 0,
+	};
+
+	BUG_ON(dm_bufio_in_request());
+
+	return dm_io(&io_req, 1, &io_reg, NULL);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_issue_flush);
+
+/*
+ * We first delete any other buffer that may be at that new location.
+ *
+ * Then, we write the buffer to the original location if it was dirty.
+ *
+ * Then, if we are the only one who is holding the buffer, relink the buffer
+ * in the hash queue for the new location.
+ *
+ * If there was someone else holding the buffer, we write it to the new
+ * location but not relink it, because that other user needs to have the buffer
+ * at the same place.
+ */
+void dm_bufio_release_move(struct dm_buffer *b, sector_t new_block)
+{
+	struct dm_bufio_client *c = b->c;
+	struct dm_buffer *new;
+
+	BUG_ON(dm_bufio_in_request());
+
+	dm_bufio_lock(c);
+
+retry:
+	new = __find(c, new_block);
+	if (new) {
+		if (new->hold_count) {
+			__wait_for_free_buffer(c);
+			goto retry;
+		}
+
+		/*
+		 * FIXME: Is there any point waiting for a write that's going
+		 * to be overwritten in a bit?
+		 */
+		__make_buffer_clean(new);
+		__unlink_buffer(new);
+		__free_buffer_wake(new);
+	}
+
+	BUG_ON(!b->hold_count);
+	BUG_ON(test_bit(B_READING, &b->state));
+
+	__write_dirty_buffer(b, NULL);
+	if (b->hold_count == 1) {
+		wait_on_bit(&b->state, B_WRITING,
+			    do_io_schedule, TASK_UNINTERRUPTIBLE);
+		set_bit(B_DIRTY, &b->state);
+		__unlink_buffer(b);
+		__link_buffer(b, new_block, LIST_DIRTY);
+	} else {
+		sector_t old_block;
+		wait_on_bit_lock(&b->state, B_WRITING,
+				 do_io_schedule, TASK_UNINTERRUPTIBLE);
+		/*
+		 * Relink buffer to "new_block" so that write_callback
+		 * sees "new_block" as a block number.
+		 * After the write, link the buffer back to old_block.
+		 * All this must be done in bufio lock, so that block number
+		 * change isn't visible to other threads.
+		 */
+		old_block = b->block;
+		__unlink_buffer(b);
+		__link_buffer(b, new_block, b->list_mode);
+		submit_io(b, WRITE, new_block, write_endio);
+		wait_on_bit(&b->state, B_WRITING,
+			    do_io_schedule, TASK_UNINTERRUPTIBLE);
+		__unlink_buffer(b);
+		__link_buffer(b, old_block, b->list_mode);
+	}
+
+	dm_bufio_unlock(c);
+	dm_bufio_release(b);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_release_move);
+
+/*
+ * Free the given buffer.
+ *
+ * This is just a hint, if the buffer is in use or dirty, this function
+ * does nothing.
+ */
+void dm_bufio_forget(struct dm_bufio_client *c, sector_t block)
+{
+	struct dm_buffer *b;
+
+	dm_bufio_lock(c);
+
+	b = __find(c, block);
+	if (b && likely(!b->hold_count) && likely(!b->state)) {
+		__unlink_buffer(b);
+		__free_buffer_wake(b);
+	}
+
+	dm_bufio_unlock(c);
+}
+EXPORT_SYMBOL(dm_bufio_forget);
+
+void dm_bufio_set_minimum_buffers(struct dm_bufio_client *c, unsigned n)
+{
+	c->minimum_buffers = n;
+}
+EXPORT_SYMBOL(dm_bufio_set_minimum_buffers);
+
+unsigned dm_bufio_get_block_size(struct dm_bufio_client *c)
+{
+	return c->block_size;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_block_size);
+
+sector_t dm_bufio_get_device_size(struct dm_bufio_client *c)
+{
+	return i_size_read(c->bdev->bd_inode) >>
+			   (SECTOR_SHIFT + c->sectors_per_block_bits);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_device_size);
+
+sector_t dm_bufio_get_block_number(struct dm_buffer *b)
+{
+	return b->block;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_block_number);
+
+void *dm_bufio_get_block_data(struct dm_buffer *b)
+{
+	return b->data;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_block_data);
+
+void *dm_bufio_get_aux_data(struct dm_buffer *b)
+{
+	return b + 1;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_aux_data);
+
+struct dm_bufio_client *dm_bufio_get_client(struct dm_buffer *b)
+{
+	return b->c;
+}
+EXPORT_SYMBOL_GPL(dm_bufio_get_client);
+
+static void drop_buffers(struct dm_bufio_client *c)
+{
+	struct dm_buffer *b;
+	int i;
+
+	BUG_ON(dm_bufio_in_request());
+
+	/*
+	 * An optimization so that the buffers are not written one-by-one.
+	 */
+	dm_bufio_write_dirty_buffers_async(c);
+
+	dm_bufio_lock(c);
+
+	while ((b = __get_unclaimed_buffer(c)))
+		__free_buffer_wake(b);
+
+	for (i = 0; i < LIST_SIZE; i++)
+		list_for_each_entry(b, &c->lru[i], lru_list)
+			DMERR("leaked buffer %llx, hold count %u, list %d",
+			      (unsigned long long)b->block, b->hold_count, i);
+
+	for (i = 0; i < LIST_SIZE; i++)
+		BUG_ON(!list_empty(&c->lru[i]));
+
+	dm_bufio_unlock(c);
+}
+
+/*
+ * Test if the buffer is unused and too old, and commit it.
+ * At if noio is set, we must not do any I/O because we hold
+ * dm_bufio_clients_lock and we would risk deadlock if the I/O gets rerouted to
+ * different bufio client.
+ */
+static int __cleanup_old_buffer(struct dm_buffer *b, gfp_t gfp,
+				unsigned long max_jiffies)
+{
+	if (jiffies - b->last_accessed < max_jiffies)
+		return 0;
+
+	if (!(gfp & __GFP_IO)) {
+		if (test_bit(B_READING, &b->state) ||
+		    test_bit(B_WRITING, &b->state) ||
+		    test_bit(B_DIRTY, &b->state))
+			return 0;
+	}
+
+	if (b->hold_count)
+		return 0;
+
+	__make_buffer_clean(b);
+	__unlink_buffer(b);
+	__free_buffer_wake(b);
+
+	return 1;
+}
+
+static long __scan(struct dm_bufio_client *c, unsigned long nr_to_scan,
+		   gfp_t gfp_mask)
+{
+	int l;
+	struct dm_buffer *b, *tmp;
+	long freed = 0;
+
+	for (l = 0; l < LIST_SIZE; l++) {
+		list_for_each_entry_safe_reverse(b, tmp, &c->lru[l], lru_list) {
+			freed += __cleanup_old_buffer(b, gfp_mask, 0);
+			if (!--nr_to_scan)
+				break;
+		}
+		dm_bufio_cond_resched();
+	}
+	return freed;
+}
+
+static unsigned long
+dm_bufio_shrink_scan(struct shrinker *shrink, struct shrink_control *sc)
+{
+	struct dm_bufio_client *c;
+	unsigned long freed;
+
+	c = container_of(shrink, struct dm_bufio_client, shrinker);
+	if (sc->gfp_mask & __GFP_IO)
+		dm_bufio_lock(c);
+	else if (!dm_bufio_trylock(c))
+		return SHRINK_STOP;
+
+	freed  = __scan(c, sc->nr_to_scan, sc->gfp_mask);
+	dm_bufio_unlock(c);
+	return freed;
+}
+
+static unsigned long
+dm_bufio_shrink_count(struct shrinker *shrink, struct shrink_control *sc)
+{
+	struct dm_bufio_client *c;
+	unsigned long count;
+
+	c = container_of(shrink, struct dm_bufio_client, shrinker);
+	if (sc->gfp_mask & __GFP_IO)
+		dm_bufio_lock(c);
+	else if (!dm_bufio_trylock(c))
+		return 0;
+
+	count = c->n_buffers[LIST_CLEAN] + c->n_buffers[LIST_DIRTY];
+	dm_bufio_unlock(c);
+	return count;
+}
+
+/*
+ * Create the buffering interface
+ */
+struct dm_bufio_client *dm_bufio_client_create(struct block_device *bdev, unsigned block_size,
+					       unsigned reserved_buffers, unsigned aux_size,
+					       void (*alloc_callback)(struct dm_buffer *),
+					       void (*write_callback)(struct dm_buffer *))
+{
+	int r;
+	struct dm_bufio_client *c;
+	unsigned i;
+
+	BUG_ON(block_size < 1 << SECTOR_SHIFT ||
+	       (block_size & (block_size - 1)));
+
+	c = kzalloc(sizeof(*c), GFP_KERNEL);
+	if (!c) {
+		r = -ENOMEM;
+		goto bad_client;
+	}
+	c->cache_hash = vmalloc(sizeof(struct hlist_head) << DM_BUFIO_HASH_BITS);
+	if (!c->cache_hash) {
+		r = -ENOMEM;
+		goto bad_hash;
+	}
+
+	c->bdev = bdev;
+	c->block_size = block_size;
+	c->sectors_per_block_bits = ffs(block_size) - 1 - SECTOR_SHIFT;
+	c->pages_per_block_bits = (ffs(block_size) - 1 >= PAGE_SHIFT) ?
+				  ffs(block_size) - 1 - PAGE_SHIFT : 0;
+	c->blocks_per_page_bits = (ffs(block_size) - 1 < PAGE_SHIFT ?
+				  PAGE_SHIFT - (ffs(block_size) - 1) : 0);
+
+	c->aux_size = aux_size;
+	c->alloc_callback = alloc_callback;
+	c->write_callback = write_callback;
+
+	for (i = 0; i < LIST_SIZE; i++) {
+		INIT_LIST_HEAD(&c->lru[i]);
+		c->n_buffers[i] = 0;
+	}
+
+	for (i = 0; i < 1 << DM_BUFIO_HASH_BITS; i++)
+		INIT_HLIST_HEAD(&c->cache_hash[i]);
+
+	mutex_init(&c->lock);
+	INIT_LIST_HEAD(&c->reserved_buffers);
+	c->need_reserved_buffers = reserved_buffers;
+
+	c->minimum_buffers = DM_BUFIO_MIN_BUFFERS;
+
+	init_waitqueue_head(&c->free_buffer_wait);
+	c->async_write_error = 0;
+
+	c->dm_io = dm_io_client_create();
+	if (IS_ERR(c->dm_io)) {
+		r = PTR_ERR(c->dm_io);
+		goto bad_dm_io;
+	}
+
+	mutex_lock(&dm_bufio_clients_lock);
+	if (c->blocks_per_page_bits) {
+		if (!DM_BUFIO_CACHE_NAME(c)) {
+			DM_BUFIO_CACHE_NAME(c) = kasprintf(GFP_KERNEL, "dm_bufio_cache-%u", c->block_size);
+			if (!DM_BUFIO_CACHE_NAME(c)) {
+				r = -ENOMEM;
+				mutex_unlock(&dm_bufio_clients_lock);
+				goto bad_cache;
+			}
+		}
+
+		if (!DM_BUFIO_CACHE(c)) {
+			DM_BUFIO_CACHE(c) = kmem_cache_create(DM_BUFIO_CACHE_NAME(c),
+							      c->block_size,
+							      c->block_size, 0, NULL);
+			if (!DM_BUFIO_CACHE(c)) {
+				r = -ENOMEM;
+				mutex_unlock(&dm_bufio_clients_lock);
+				goto bad_cache;
+			}
+		}
+	}
+	mutex_unlock(&dm_bufio_clients_lock);
+
+	while (c->need_reserved_buffers) {
+		struct dm_buffer *b = alloc_buffer(c, GFP_KERNEL);
+
+		if (!b) {
+			r = -ENOMEM;
+			goto bad_buffer;
+		}
+		__free_buffer_wake(b);
+	}
+
+	mutex_lock(&dm_bufio_clients_lock);
+	dm_bufio_client_count++;
+	list_add(&c->client_list, &dm_bufio_all_clients);
+	__cache_size_refresh();
+	mutex_unlock(&dm_bufio_clients_lock);
+
+	c->shrinker.count_objects = dm_bufio_shrink_count;
+	c->shrinker.scan_objects = dm_bufio_shrink_scan;
+	c->shrinker.seeks = 1;
+	c->shrinker.batch = 0;
+	register_shrinker(&c->shrinker);
+
+	return c;
+
+bad_buffer:
+bad_cache:
+	while (!list_empty(&c->reserved_buffers)) {
+		struct dm_buffer *b = list_entry(c->reserved_buffers.next,
+						 struct dm_buffer, lru_list);
+		list_del(&b->lru_list);
+		free_buffer(b);
+	}
+	dm_io_client_destroy(c->dm_io);
+bad_dm_io:
+	vfree(c->cache_hash);
+bad_hash:
+	kfree(c);
+bad_client:
+	return ERR_PTR(r);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_client_create);
+
+/*
+ * Free the buffering interface.
+ * It is required that there are no references on any buffers.
+ */
+void dm_bufio_client_destroy(struct dm_bufio_client *c)
+{
+	unsigned i;
+
+	drop_buffers(c);
+
+	unregister_shrinker(&c->shrinker);
+
+	mutex_lock(&dm_bufio_clients_lock);
+
+	list_del(&c->client_list);
+	dm_bufio_client_count--;
+	__cache_size_refresh();
+
+	mutex_unlock(&dm_bufio_clients_lock);
+
+	for (i = 0; i < 1 << DM_BUFIO_HASH_BITS; i++)
+		BUG_ON(!hlist_empty(&c->cache_hash[i]));
+
+	BUG_ON(c->need_reserved_buffers);
+
+	while (!list_empty(&c->reserved_buffers)) {
+		struct dm_buffer *b = list_entry(c->reserved_buffers.next,
+						 struct dm_buffer, lru_list);
+		list_del(&b->lru_list);
+		free_buffer(b);
+	}
+
+	for (i = 0; i < LIST_SIZE; i++)
+		if (c->n_buffers[i])
+			DMERR("leaked buffer count %d: %ld", i, c->n_buffers[i]);
+
+	for (i = 0; i < LIST_SIZE; i++)
+		BUG_ON(c->n_buffers[i]);
+
+	dm_io_client_destroy(c->dm_io);
+	vfree(c->cache_hash);
+	kfree(c);
+}
+EXPORT_SYMBOL_GPL(dm_bufio_client_destroy);
+
+static void cleanup_old_buffers(void)
+{
+	unsigned long max_age = ACCESS_ONCE(dm_bufio_max_age);
+	struct dm_bufio_client *c;
+
+	if (max_age > ULONG_MAX / HZ)
+		max_age = ULONG_MAX / HZ;
+
+	mutex_lock(&dm_bufio_clients_lock);
+	list_for_each_entry(c, &dm_bufio_all_clients, client_list) {
+		if (!dm_bufio_trylock(c))
+			continue;
+
+		while (!list_empty(&c->lru[LIST_CLEAN])) {
+			struct dm_buffer *b;
+			b = list_entry(c->lru[LIST_CLEAN].prev,
+				       struct dm_buffer, lru_list);
+			if (!__cleanup_old_buffer(b, 0, max_age * HZ))
+				break;
+			dm_bufio_cond_resched();
+		}
+
+		dm_bufio_unlock(c);
+		dm_bufio_cond_resched();
+	}
+	mutex_unlock(&dm_bufio_clients_lock);
+}
+
+static struct workqueue_struct *dm_bufio_wq;
+static struct delayed_work dm_bufio_work;
+
+static void work_fn(struct work_struct *w)
+{
+	cleanup_old_buffers();
+
+	queue_delayed_work(dm_bufio_wq, &dm_bufio_work,
+			   DM_BUFIO_WORK_TIMER_SECS * HZ);
+}
+
+/*----------------------------------------------------------------
+ * Module setup
+ *--------------------------------------------------------------*/
+
+/*
+ * This is called only once for the whole dm_bufio module.
+ * It initializes memory limit.
+ */
+static int __init dm_bufio_init(void)
+{
+	__u64 mem;
+
+	dm_bufio_allocated_kmem_cache = 0;
+	dm_bufio_allocated_get_free_pages = 0;
+	dm_bufio_allocated_vmalloc = 0;
+	dm_bufio_current_allocated = 0;
+
+	memset(&dm_bufio_caches, 0, sizeof dm_bufio_caches);
+	memset(&dm_bufio_cache_names, 0, sizeof dm_bufio_cache_names);
+
+	mem = (__u64)((totalram_pages - totalhigh_pages) *
+		      DM_BUFIO_MEMORY_PERCENT / 100) << PAGE_SHIFT;
+
+	if (mem > ULONG_MAX)
+		mem = ULONG_MAX;
+
+#ifdef CONFIG_MMU
+	/*
+	 * Get the size of vmalloc space the same way as VMALLOC_TOTAL
+	 * in fs/proc/internal.h
+	 */
+	if (mem > (VMALLOC_END - VMALLOC_START) * DM_BUFIO_VMALLOC_PERCENT / 100)
+		mem = (VMALLOC_END - VMALLOC_START) * DM_BUFIO_VMALLOC_PERCENT / 100;
+#endif
+
+	dm_bufio_default_cache_size = mem;
+
+	mutex_lock(&dm_bufio_clients_lock);
+	__cache_size_refresh();
+	mutex_unlock(&dm_bufio_clients_lock);
+
+	dm_bufio_wq = create_singlethread_workqueue("dm_bufio_cache");
+	if (!dm_bufio_wq)
+		return -ENOMEM;
+
+	INIT_DELAYED_WORK(&dm_bufio_work, work_fn);
+	queue_delayed_work(dm_bufio_wq, &dm_bufio_work,
+			   DM_BUFIO_WORK_TIMER_SECS * HZ);
+
+	return 0;
+}
+
+/*
+ * This is called once when unloading the dm_bufio module.
+ */
+static void __exit dm_bufio_exit(void)
+{
+	int bug = 0;
+	int i;
+
+	cancel_delayed_work_sync(&dm_bufio_work);
+	destroy_workqueue(dm_bufio_wq);
+
+	for (i = 0; i < ARRAY_SIZE(dm_bufio_caches); i++) {
+		struct kmem_cache *kc = dm_bufio_caches[i];
+
+		if (kc)
+			kmem_cache_destroy(kc);
+	}
+
+	for (i = 0; i < ARRAY_SIZE(dm_bufio_cache_names); i++)
+		kfree(dm_bufio_cache_names[i]);
+
+	if (dm_bufio_client_count) {
+		DMCRIT("%s: dm_bufio_client_count leaked: %d",
+			__func__, dm_bufio_client_count);
+		bug = 1;
+	}
+
+	if (dm_bufio_current_allocated) {
+		DMCRIT("%s: dm_bufio_current_allocated leaked: %lu",
+			__func__, dm_bufio_current_allocated);
+		bug = 1;
+	}
+
+	if (dm_bufio_allocated_get_free_pages) {
+		DMCRIT("%s: dm_bufio_allocated_get_free_pages leaked: %lu",
+		       __func__, dm_bufio_allocated_get_free_pages);
+		bug = 1;
+	}
+
+	if (dm_bufio_allocated_vmalloc) {
+		DMCRIT("%s: dm_bufio_vmalloc leaked: %lu",
+		       __func__, dm_bufio_allocated_vmalloc);
+		bug = 1;
+	}
+
+	if (bug)
+		BUG();
+}
+
+module_init(dm_bufio_init)
+module_exit(dm_bufio_exit)
+
+module_param_named(max_cache_size_bytes, dm_bufio_cache_size, ulong, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(max_cache_size_bytes, "Size of metadata cache");
+
+module_param_named(max_age_seconds, dm_bufio_max_age, uint, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(max_age_seconds, "Max age of a buffer in seconds");
+
+module_param_named(peak_allocated_bytes, dm_bufio_peak_allocated, ulong, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(peak_allocated_bytes, "Tracks the maximum allocated memory");
+
+module_param_named(allocated_kmem_cache_bytes, dm_bufio_allocated_kmem_cache, ulong, S_IRUGO);
+MODULE_PARM_DESC(allocated_kmem_cache_bytes, "Memory allocated with kmem_cache_alloc");
+
+module_param_named(allocated_get_free_pages_bytes, dm_bufio_allocated_get_free_pages, ulong, S_IRUGO);
+MODULE_PARM_DESC(allocated_get_free_pages_bytes, "Memory allocated with get_free_pages");
+
+module_param_named(allocated_vmalloc_bytes, dm_bufio_allocated_vmalloc, ulong, S_IRUGO);
+MODULE_PARM_DESC(allocated_vmalloc_bytes, "Memory allocated with vmalloc");
+
+module_param_named(current_allocated_bytes, dm_bufio_current_allocated, ulong, S_IRUGO);
+MODULE_PARM_DESC(current_allocated_bytes, "Memory currently used by the cache");
+
+MODULE_AUTHOR("Mikulas Patocka <dm-devel@redhat.com>");
+MODULE_DESCRIPTION(DM_NAME " buffered I/O library");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-bufio.h b/drivers/md/dm-bufio.h
new file mode 100644
index 00000000000..c096779a729
--- /dev/null
+++ b/drivers/md/dm-bufio.h
@@ -0,0 +1,132 @@
+/*
+ * Copyright (C) 2009-2011 Red Hat, Inc.
+ *
+ * Author: Mikulas Patocka <mpatocka@redhat.com>
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_BUFIO_H
+#define DM_BUFIO_H
+
+#include <linux/blkdev.h>
+#include <linux/types.h>
+
+/*----------------------------------------------------------------*/
+
+struct dm_bufio_client;
+struct dm_buffer;
+
+/*
+ * Create a buffered IO cache on a given device
+ */
+struct dm_bufio_client *
+dm_bufio_client_create(struct block_device *bdev, unsigned block_size,
+		       unsigned reserved_buffers, unsigned aux_size,
+		       void (*alloc_callback)(struct dm_buffer *),
+		       void (*write_callback)(struct dm_buffer *));
+
+/*
+ * Release a buffered IO cache.
+ */
+void dm_bufio_client_destroy(struct dm_bufio_client *c);
+
+/*
+ * WARNING: to avoid deadlocks, these conditions are observed:
+ *
+ * - At most one thread can hold at most "reserved_buffers" simultaneously.
+ * - Each other threads can hold at most one buffer.
+ * - Threads which call only dm_bufio_get can hold unlimited number of
+ *   buffers.
+ */
+
+/*
+ * Read a given block from disk. Returns pointer to data.  Returns a
+ * pointer to dm_buffer that can be used to release the buffer or to make
+ * it dirty.
+ */
+void *dm_bufio_read(struct dm_bufio_client *c, sector_t block,
+		    struct dm_buffer **bp);
+
+/*
+ * Like dm_bufio_read, but return buffer from cache, don't read
+ * it. If the buffer is not in the cache, return NULL.
+ */
+void *dm_bufio_get(struct dm_bufio_client *c, sector_t block,
+		   struct dm_buffer **bp);
+
+/*
+ * Like dm_bufio_read, but don't read anything from the disk.  It is
+ * expected that the caller initializes the buffer and marks it dirty.
+ */
+void *dm_bufio_new(struct dm_bufio_client *c, sector_t block,
+		   struct dm_buffer **bp);
+
+/*
+ * Prefetch the specified blocks to the cache.
+ * The function starts to read the blocks and returns without waiting for
+ * I/O to finish.
+ */
+void dm_bufio_prefetch(struct dm_bufio_client *c,
+		       sector_t block, unsigned n_blocks);
+
+/*
+ * Release a reference obtained with dm_bufio_{read,get,new}. The data
+ * pointer and dm_buffer pointer is no longer valid after this call.
+ */
+void dm_bufio_release(struct dm_buffer *b);
+
+/*
+ * Mark a buffer dirty. It should be called after the buffer is modified.
+ *
+ * In case of memory pressure, the buffer may be written after
+ * dm_bufio_mark_buffer_dirty, but before dm_bufio_write_dirty_buffers.  So
+ * dm_bufio_write_dirty_buffers guarantees that the buffer is on-disk but
+ * the actual writing may occur earlier.
+ */
+void dm_bufio_mark_buffer_dirty(struct dm_buffer *b);
+
+/*
+ * Initiate writing of dirty buffers, without waiting for completion.
+ */
+void dm_bufio_write_dirty_buffers_async(struct dm_bufio_client *c);
+
+/*
+ * Write all dirty buffers. Guarantees that all dirty buffers created prior
+ * to this call are on disk when this call exits.
+ */
+int dm_bufio_write_dirty_buffers(struct dm_bufio_client *c);
+
+/*
+ * Send an empty write barrier to the device to flush hardware disk cache.
+ */
+int dm_bufio_issue_flush(struct dm_bufio_client *c);
+
+/*
+ * Like dm_bufio_release but also move the buffer to the new
+ * block. dm_bufio_write_dirty_buffers is needed to commit the new block.
+ */
+void dm_bufio_release_move(struct dm_buffer *b, sector_t new_block);
+
+/*
+ * Free the given buffer.
+ * This is just a hint, if the buffer is in use or dirty, this function
+ * does nothing.
+ */
+void dm_bufio_forget(struct dm_bufio_client *c, sector_t block);
+
+/*
+ * Set the minimum number of buffers before cleanup happens.
+ */
+void dm_bufio_set_minimum_buffers(struct dm_bufio_client *c, unsigned n);
+
+unsigned dm_bufio_get_block_size(struct dm_bufio_client *c);
+sector_t dm_bufio_get_device_size(struct dm_bufio_client *c);
+sector_t dm_bufio_get_block_number(struct dm_buffer *b);
+void *dm_bufio_get_block_data(struct dm_buffer *b);
+void *dm_bufio_get_aux_data(struct dm_buffer *b);
+struct dm_bufio_client *dm_bufio_get_client(struct dm_buffer *b);
+
+/*----------------------------------------------------------------*/
+
+#endif
diff --git a/drivers/md/dm-builtin.c b/drivers/md/dm-builtin.c
new file mode 100644
index 00000000000..6c9049c51b2
--- /dev/null
+++ b/drivers/md/dm-builtin.c
@@ -0,0 +1,48 @@
+#include "dm.h"
+
+/*
+ * The kobject release method must not be placed in the module itself,
+ * otherwise we are subject to module unload races.
+ *
+ * The release method is called when the last reference to the kobject is
+ * dropped. It may be called by any other kernel code that drops the last
+ * reference.
+ *
+ * The release method suffers from module unload race. We may prevent the
+ * module from being unloaded at the start of the release method (using
+ * increased module reference count or synchronizing against the release
+ * method), however there is no way to prevent the module from being
+ * unloaded at the end of the release method.
+ *
+ * If this code were placed in the dm module, the following race may
+ * happen:
+ *  1. Some other process takes a reference to dm kobject
+ *  2. The user issues ioctl function to unload the dm device
+ *  3. dm_sysfs_exit calls kobject_put, however the object is not released
+ *     because of the other reference taken at step 1
+ *  4. dm_sysfs_exit waits on the completion
+ *  5. The other process that took the reference in step 1 drops it,
+ *     dm_kobject_release is called from this process
+ *  6. dm_kobject_release calls complete()
+ *  7. a reschedule happens before dm_kobject_release returns
+ *  8. dm_sysfs_exit continues, the dm device is unloaded, module reference
+ *     count is decremented
+ *  9. The user unloads the dm module
+ * 10. The other process that was rescheduled in step 7 continues to run,
+ *     it is now executing code in unloaded module, so it crashes
+ *
+ * Note that if the process that takes the foreign reference to dm kobject
+ * has a low priority and the system is sufficiently loaded with
+ * higher-priority processes that prevent the low-priority process from
+ * being scheduled long enough, this bug may really happen.
+ *
+ * In order to fix this module unload race, we place the release method
+ * into a helper code that is compiled directly into the kernel.
+ */
+
+void dm_kobject_release(struct kobject *kobj)
+{
+	complete(dm_get_completion_from_kobject(kobj));
+}
+
+EXPORT_SYMBOL(dm_kobject_release);
diff --git a/drivers/md/dm-cache-block-types.h b/drivers/md/dm-cache-block-types.h
new file mode 100644
index 00000000000..aac0e2df06b
--- /dev/null
+++ b/drivers/md/dm-cache-block-types.h
@@ -0,0 +1,43 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_CACHE_BLOCK_TYPES_H
+#define DM_CACHE_BLOCK_TYPES_H
+
+#include "persistent-data/dm-block-manager.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * It's helpful to get sparse to differentiate between indexes into the
+ * origin device, indexes into the cache device, and indexes into the
+ * discard bitset.
+ */
+
+typedef dm_block_t __bitwise__ dm_oblock_t;
+typedef uint32_t __bitwise__ dm_cblock_t;
+
+static inline dm_oblock_t to_oblock(dm_block_t b)
+{
+	return (__force dm_oblock_t) b;
+}
+
+static inline dm_block_t from_oblock(dm_oblock_t b)
+{
+	return (__force dm_block_t) b;
+}
+
+static inline dm_cblock_t to_cblock(uint32_t b)
+{
+	return (__force dm_cblock_t) b;
+}
+
+static inline uint32_t from_cblock(dm_cblock_t b)
+{
+	return (__force uint32_t) b;
+}
+
+#endif /* DM_CACHE_BLOCK_TYPES_H */
diff --git a/drivers/md/dm-cache-metadata.c b/drivers/md/dm-cache-metadata.c
new file mode 100644
index 00000000000..d2899e7eb3a
--- /dev/null
+++ b/drivers/md/dm-cache-metadata.c
@@ -0,0 +1,1299 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-cache-metadata.h"
+
+#include "persistent-data/dm-array.h"
+#include "persistent-data/dm-bitset.h"
+#include "persistent-data/dm-space-map.h"
+#include "persistent-data/dm-space-map-disk.h"
+#include "persistent-data/dm-transaction-manager.h"
+
+#include <linux/device-mapper.h>
+
+/*----------------------------------------------------------------*/
+
+#define DM_MSG_PREFIX   "cache metadata"
+
+#define CACHE_SUPERBLOCK_MAGIC 06142003
+#define CACHE_SUPERBLOCK_LOCATION 0
+
+/*
+ * defines a range of metadata versions that this module can handle.
+ */
+#define MIN_CACHE_VERSION 1
+#define MAX_CACHE_VERSION 1
+
+#define CACHE_METADATA_CACHE_SIZE 64
+
+/*
+ *  3 for btree insert +
+ *  2 for btree lookup used within space map
+ */
+#define CACHE_MAX_CONCURRENT_LOCKS 5
+#define SPACE_MAP_ROOT_SIZE 128
+
+enum superblock_flag_bits {
+	/* for spotting crashes that would invalidate the dirty bitset */
+	CLEAN_SHUTDOWN,
+};
+
+/*
+ * Each mapping from cache block -> origin block carries a set of flags.
+ */
+enum mapping_bits {
+	/*
+	 * A valid mapping.  Because we're using an array we clear this
+	 * flag for an non existant mapping.
+	 */
+	M_VALID = 1,
+
+	/*
+	 * The data on the cache is different from that on the origin.
+	 */
+	M_DIRTY = 2
+};
+
+struct cache_disk_superblock {
+	__le32 csum;
+	__le32 flags;
+	__le64 blocknr;
+
+	__u8 uuid[16];
+	__le64 magic;
+	__le32 version;
+
+	__u8 policy_name[CACHE_POLICY_NAME_SIZE];
+	__le32 policy_hint_size;
+
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+	__le64 mapping_root;
+	__le64 hint_root;
+
+	__le64 discard_root;
+	__le64 discard_block_size;
+	__le64 discard_nr_blocks;
+
+	__le32 data_block_size;
+	__le32 metadata_block_size;
+	__le32 cache_blocks;
+
+	__le32 compat_flags;
+	__le32 compat_ro_flags;
+	__le32 incompat_flags;
+
+	__le32 read_hits;
+	__le32 read_misses;
+	__le32 write_hits;
+	__le32 write_misses;
+
+	__le32 policy_version[CACHE_POLICY_VERSION_SIZE];
+} __packed;
+
+struct dm_cache_metadata {
+	struct block_device *bdev;
+	struct dm_block_manager *bm;
+	struct dm_space_map *metadata_sm;
+	struct dm_transaction_manager *tm;
+
+	struct dm_array_info info;
+	struct dm_array_info hint_info;
+	struct dm_disk_bitset discard_info;
+
+	struct rw_semaphore root_lock;
+	dm_block_t root;
+	dm_block_t hint_root;
+	dm_block_t discard_root;
+
+	sector_t discard_block_size;
+	dm_oblock_t discard_nr_blocks;
+
+	sector_t data_block_size;
+	dm_cblock_t cache_blocks;
+	bool changed:1;
+	bool clean_when_opened:1;
+
+	char policy_name[CACHE_POLICY_NAME_SIZE];
+	unsigned policy_version[CACHE_POLICY_VERSION_SIZE];
+	size_t policy_hint_size;
+	struct dm_cache_statistics stats;
+
+	/*
+	 * Reading the space map root can fail, so we read it into this
+	 * buffer before the superblock is locked and updated.
+	 */
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+};
+
+/*-------------------------------------------------------------------
+ * superblock validator
+ *-----------------------------------------------------------------*/
+
+#define SUPERBLOCK_CSUM_XOR 9031977
+
+static void sb_prepare_for_write(struct dm_block_validator *v,
+				 struct dm_block *b,
+				 size_t sb_block_size)
+{
+	struct cache_disk_superblock *disk_super = dm_block_data(b);
+
+	disk_super->blocknr = cpu_to_le64(dm_block_location(b));
+	disk_super->csum = cpu_to_le32(dm_bm_checksum(&disk_super->flags,
+						      sb_block_size - sizeof(__le32),
+						      SUPERBLOCK_CSUM_XOR));
+}
+
+static int check_metadata_version(struct cache_disk_superblock *disk_super)
+{
+	uint32_t metadata_version = le32_to_cpu(disk_super->version);
+	if (metadata_version < MIN_CACHE_VERSION || metadata_version > MAX_CACHE_VERSION) {
+		DMERR("Cache metadata version %u found, but only versions between %u and %u supported.",
+		      metadata_version, MIN_CACHE_VERSION, MAX_CACHE_VERSION);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int sb_check(struct dm_block_validator *v,
+		    struct dm_block *b,
+		    size_t sb_block_size)
+{
+	struct cache_disk_superblock *disk_super = dm_block_data(b);
+	__le32 csum_le;
+
+	if (dm_block_location(b) != le64_to_cpu(disk_super->blocknr)) {
+		DMERR("sb_check failed: blocknr %llu: wanted %llu",
+		      le64_to_cpu(disk_super->blocknr),
+		      (unsigned long long)dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	if (le64_to_cpu(disk_super->magic) != CACHE_SUPERBLOCK_MAGIC) {
+		DMERR("sb_check failed: magic %llu: wanted %llu",
+		      le64_to_cpu(disk_super->magic),
+		      (unsigned long long)CACHE_SUPERBLOCK_MAGIC);
+		return -EILSEQ;
+	}
+
+	csum_le = cpu_to_le32(dm_bm_checksum(&disk_super->flags,
+					     sb_block_size - sizeof(__le32),
+					     SUPERBLOCK_CSUM_XOR));
+	if (csum_le != disk_super->csum) {
+		DMERR("sb_check failed: csum %u: wanted %u",
+		      le32_to_cpu(csum_le), le32_to_cpu(disk_super->csum));
+		return -EILSEQ;
+	}
+
+	return check_metadata_version(disk_super);
+}
+
+static struct dm_block_validator sb_validator = {
+	.name = "superblock",
+	.prepare_for_write = sb_prepare_for_write,
+	.check = sb_check
+};
+
+/*----------------------------------------------------------------*/
+
+static int superblock_read_lock(struct dm_cache_metadata *cmd,
+				struct dm_block **sblock)
+{
+	return dm_bm_read_lock(cmd->bm, CACHE_SUPERBLOCK_LOCATION,
+			       &sb_validator, sblock);
+}
+
+static int superblock_lock_zero(struct dm_cache_metadata *cmd,
+				struct dm_block **sblock)
+{
+	return dm_bm_write_lock_zero(cmd->bm, CACHE_SUPERBLOCK_LOCATION,
+				     &sb_validator, sblock);
+}
+
+static int superblock_lock(struct dm_cache_metadata *cmd,
+			   struct dm_block **sblock)
+{
+	return dm_bm_write_lock(cmd->bm, CACHE_SUPERBLOCK_LOCATION,
+				&sb_validator, sblock);
+}
+
+/*----------------------------------------------------------------*/
+
+static int __superblock_all_zeroes(struct dm_block_manager *bm, bool *result)
+{
+	int r;
+	unsigned i;
+	struct dm_block *b;
+	__le64 *data_le, zero = cpu_to_le64(0);
+	unsigned sb_block_size = dm_bm_block_size(bm) / sizeof(__le64);
+
+	/*
+	 * We can't use a validator here - it may be all zeroes.
+	 */
+	r = dm_bm_read_lock(bm, CACHE_SUPERBLOCK_LOCATION, NULL, &b);
+	if (r)
+		return r;
+
+	data_le = dm_block_data(b);
+	*result = true;
+	for (i = 0; i < sb_block_size; i++) {
+		if (data_le[i] != zero) {
+			*result = false;
+			break;
+		}
+	}
+
+	return dm_bm_unlock(b);
+}
+
+static void __setup_mapping_info(struct dm_cache_metadata *cmd)
+{
+	struct dm_btree_value_type vt;
+
+	vt.context = NULL;
+	vt.size = sizeof(__le64);
+	vt.inc = NULL;
+	vt.dec = NULL;
+	vt.equal = NULL;
+	dm_array_info_init(&cmd->info, cmd->tm, &vt);
+
+	if (cmd->policy_hint_size) {
+		vt.size = sizeof(__le32);
+		dm_array_info_init(&cmd->hint_info, cmd->tm, &vt);
+	}
+}
+
+static int __save_sm_root(struct dm_cache_metadata *cmd)
+{
+	int r;
+	size_t metadata_len;
+
+	r = dm_sm_root_size(cmd->metadata_sm, &metadata_len);
+	if (r < 0)
+		return r;
+
+	return dm_sm_copy_root(cmd->metadata_sm, &cmd->metadata_space_map_root,
+			       metadata_len);
+}
+
+static void __copy_sm_root(struct dm_cache_metadata *cmd,
+			   struct cache_disk_superblock *disk_super)
+{
+	memcpy(&disk_super->metadata_space_map_root,
+	       &cmd->metadata_space_map_root,
+	       sizeof(cmd->metadata_space_map_root));
+}
+
+static int __write_initial_superblock(struct dm_cache_metadata *cmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct cache_disk_superblock *disk_super;
+	sector_t bdev_size = i_size_read(cmd->bdev->bd_inode) >> SECTOR_SHIFT;
+
+	/* FIXME: see if we can lose the max sectors limit */
+	if (bdev_size > DM_CACHE_METADATA_MAX_SECTORS)
+		bdev_size = DM_CACHE_METADATA_MAX_SECTORS;
+
+	r = dm_tm_pre_commit(cmd->tm);
+	if (r < 0)
+		return r;
+
+	/*
+	 * dm_sm_copy_root() can fail.  So we need to do it before we start
+	 * updating the superblock.
+	 */
+	r = __save_sm_root(cmd);
+	if (r)
+		return r;
+
+	r = superblock_lock_zero(cmd, &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(CACHE_SUPERBLOCK_MAGIC);
+	disk_super->version = cpu_to_le32(MAX_CACHE_VERSION);
+	memset(disk_super->policy_name, 0, sizeof(disk_super->policy_name));
+	memset(disk_super->policy_version, 0, sizeof(disk_super->policy_version));
+	disk_super->policy_hint_size = 0;
+
+	__copy_sm_root(cmd, disk_super);
+
+	disk_super->mapping_root = cpu_to_le64(cmd->root);
+	disk_super->hint_root = cpu_to_le64(cmd->hint_root);
+	disk_super->discard_root = cpu_to_le64(cmd->discard_root);
+	disk_super->discard_block_size = cpu_to_le64(cmd->discard_block_size);
+	disk_super->discard_nr_blocks = cpu_to_le64(from_oblock(cmd->discard_nr_blocks));
+	disk_super->metadata_block_size = cpu_to_le32(DM_CACHE_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
+	disk_super->data_block_size = cpu_to_le32(cmd->data_block_size);
+	disk_super->cache_blocks = cpu_to_le32(0);
+
+	disk_super->read_hits = cpu_to_le32(0);
+	disk_super->read_misses = cpu_to_le32(0);
+	disk_super->write_hits = cpu_to_le32(0);
+	disk_super->write_misses = cpu_to_le32(0);
+
+	return dm_tm_commit(cmd->tm, sblock);
+}
+
+static int __format_metadata(struct dm_cache_metadata *cmd)
+{
+	int r;
+
+	r = dm_tm_create_with_sm(cmd->bm, CACHE_SUPERBLOCK_LOCATION,
+				 &cmd->tm, &cmd->metadata_sm);
+	if (r < 0) {
+		DMERR("tm_create_with_sm failed");
+		return r;
+	}
+
+	__setup_mapping_info(cmd);
+
+	r = dm_array_empty(&cmd->info, &cmd->root);
+	if (r < 0)
+		goto bad;
+
+	dm_disk_bitset_init(cmd->tm, &cmd->discard_info);
+
+	r = dm_bitset_empty(&cmd->discard_info, &cmd->discard_root);
+	if (r < 0)
+		goto bad;
+
+	cmd->discard_block_size = 0;
+	cmd->discard_nr_blocks = 0;
+
+	r = __write_initial_superblock(cmd);
+	if (r)
+		goto bad;
+
+	cmd->clean_when_opened = true;
+	return 0;
+
+bad:
+	dm_tm_destroy(cmd->tm);
+	dm_sm_destroy(cmd->metadata_sm);
+
+	return r;
+}
+
+static int __check_incompat_features(struct cache_disk_superblock *disk_super,
+				     struct dm_cache_metadata *cmd)
+{
+	uint32_t features;
+
+	features = le32_to_cpu(disk_super->incompat_flags) & ~DM_CACHE_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(cmd->bdev->bd_disk))
+		return 0;
+
+	features = le32_to_cpu(disk_super->compat_ro_flags) & ~DM_CACHE_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_cache_metadata *cmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct cache_disk_superblock *disk_super;
+	unsigned long sb_flags;
+
+	r = superblock_read_lock(cmd, &sblock);
+	if (r < 0) {
+		DMERR("couldn't read lock superblock");
+		return r;
+	}
+
+	disk_super = dm_block_data(sblock);
+
+	/* Verify the data block size hasn't changed */
+	if (le32_to_cpu(disk_super->data_block_size) != cmd->data_block_size) {
+		DMERR("changing the data block size (from %u to %llu) is not supported",
+		      le32_to_cpu(disk_super->data_block_size),
+		      (unsigned long long)cmd->data_block_size);
+		r = -EINVAL;
+		goto bad;
+	}
+
+	r = __check_incompat_features(disk_super, cmd);
+	if (r < 0)
+		goto bad;
+
+	r = dm_tm_open_with_sm(cmd->bm, CACHE_SUPERBLOCK_LOCATION,
+			       disk_super->metadata_space_map_root,
+			       sizeof(disk_super->metadata_space_map_root),
+			       &cmd->tm, &cmd->metadata_sm);
+	if (r < 0) {
+		DMERR("tm_open_with_sm failed");
+		goto bad;
+	}
+
+	__setup_mapping_info(cmd);
+	dm_disk_bitset_init(cmd->tm, &cmd->discard_info);
+	sb_flags = le32_to_cpu(disk_super->flags);
+	cmd->clean_when_opened = test_bit(CLEAN_SHUTDOWN, &sb_flags);
+	return dm_bm_unlock(sblock);
+
+bad:
+	dm_bm_unlock(sblock);
+	return r;
+}
+
+static int __open_or_format_metadata(struct dm_cache_metadata *cmd,
+				     bool format_device)
+{
+	int r;
+	bool unformatted = false;
+
+	r = __superblock_all_zeroes(cmd->bm, &unformatted);
+	if (r)
+		return r;
+
+	if (unformatted)
+		return format_device ? __format_metadata(cmd) : -EPERM;
+
+	return __open_metadata(cmd);
+}
+
+static int __create_persistent_data_objects(struct dm_cache_metadata *cmd,
+					    bool may_format_device)
+{
+	int r;
+	cmd->bm = dm_block_manager_create(cmd->bdev, DM_CACHE_METADATA_BLOCK_SIZE,
+					  CACHE_METADATA_CACHE_SIZE,
+					  CACHE_MAX_CONCURRENT_LOCKS);
+	if (IS_ERR(cmd->bm)) {
+		DMERR("could not create block manager");
+		return PTR_ERR(cmd->bm);
+	}
+
+	r = __open_or_format_metadata(cmd, may_format_device);
+	if (r)
+		dm_block_manager_destroy(cmd->bm);
+
+	return r;
+}
+
+static void __destroy_persistent_data_objects(struct dm_cache_metadata *cmd)
+{
+	dm_sm_destroy(cmd->metadata_sm);
+	dm_tm_destroy(cmd->tm);
+	dm_block_manager_destroy(cmd->bm);
+}
+
+typedef unsigned long (*flags_mutator)(unsigned long);
+
+static void update_flags(struct cache_disk_superblock *disk_super,
+			 flags_mutator mutator)
+{
+	uint32_t sb_flags = mutator(le32_to_cpu(disk_super->flags));
+	disk_super->flags = cpu_to_le32(sb_flags);
+}
+
+static unsigned long set_clean_shutdown(unsigned long flags)
+{
+	set_bit(CLEAN_SHUTDOWN, &flags);
+	return flags;
+}
+
+static unsigned long clear_clean_shutdown(unsigned long flags)
+{
+	clear_bit(CLEAN_SHUTDOWN, &flags);
+	return flags;
+}
+
+static void read_superblock_fields(struct dm_cache_metadata *cmd,
+				   struct cache_disk_superblock *disk_super)
+{
+	cmd->root = le64_to_cpu(disk_super->mapping_root);
+	cmd->hint_root = le64_to_cpu(disk_super->hint_root);
+	cmd->discard_root = le64_to_cpu(disk_super->discard_root);
+	cmd->discard_block_size = le64_to_cpu(disk_super->discard_block_size);
+	cmd->discard_nr_blocks = to_oblock(le64_to_cpu(disk_super->discard_nr_blocks));
+	cmd->data_block_size = le32_to_cpu(disk_super->data_block_size);
+	cmd->cache_blocks = to_cblock(le32_to_cpu(disk_super->cache_blocks));
+	strncpy(cmd->policy_name, disk_super->policy_name, sizeof(cmd->policy_name));
+	cmd->policy_version[0] = le32_to_cpu(disk_super->policy_version[0]);
+	cmd->policy_version[1] = le32_to_cpu(disk_super->policy_version[1]);
+	cmd->policy_version[2] = le32_to_cpu(disk_super->policy_version[2]);
+	cmd->policy_hint_size = le32_to_cpu(disk_super->policy_hint_size);
+
+	cmd->stats.read_hits = le32_to_cpu(disk_super->read_hits);
+	cmd->stats.read_misses = le32_to_cpu(disk_super->read_misses);
+	cmd->stats.write_hits = le32_to_cpu(disk_super->write_hits);
+	cmd->stats.write_misses = le32_to_cpu(disk_super->write_misses);
+
+	cmd->changed = false;
+}
+
+/*
+ * The mutator updates the superblock flags.
+ */
+static int __begin_transaction_flags(struct dm_cache_metadata *cmd,
+				     flags_mutator mutator)
+{
+	int r;
+	struct cache_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	r = superblock_lock(cmd, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	update_flags(disk_super, mutator);
+	read_superblock_fields(cmd, disk_super);
+	dm_bm_unlock(sblock);
+
+	return dm_bm_flush(cmd->bm);
+}
+
+static int __begin_transaction(struct dm_cache_metadata *cmd)
+{
+	int r;
+	struct cache_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	/*
+	 * We re-read the superblock every time.  Shouldn't need to do this
+	 * really.
+	 */
+	r = superblock_read_lock(cmd, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	read_superblock_fields(cmd, disk_super);
+	dm_bm_unlock(sblock);
+
+	return 0;
+}
+
+static int __commit_transaction(struct dm_cache_metadata *cmd,
+				flags_mutator mutator)
+{
+	int r;
+	struct cache_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	/*
+	 * We need to know if the cache_disk_superblock exceeds a 512-byte sector.
+	 */
+	BUILD_BUG_ON(sizeof(struct cache_disk_superblock) > 512);
+
+	r = dm_bitset_flush(&cmd->discard_info, cmd->discard_root,
+			    &cmd->discard_root);
+	if (r)
+		return r;
+
+	r = dm_tm_pre_commit(cmd->tm);
+	if (r < 0)
+		return r;
+
+	r = __save_sm_root(cmd);
+	if (r)
+		return r;
+
+	r = superblock_lock(cmd, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+
+	if (mutator)
+		update_flags(disk_super, mutator);
+
+	disk_super->mapping_root = cpu_to_le64(cmd->root);
+	disk_super->hint_root = cpu_to_le64(cmd->hint_root);
+	disk_super->discard_root = cpu_to_le64(cmd->discard_root);
+	disk_super->discard_block_size = cpu_to_le64(cmd->discard_block_size);
+	disk_super->discard_nr_blocks = cpu_to_le64(from_oblock(cmd->discard_nr_blocks));
+	disk_super->cache_blocks = cpu_to_le32(from_cblock(cmd->cache_blocks));
+	strncpy(disk_super->policy_name, cmd->policy_name, sizeof(disk_super->policy_name));
+	disk_super->policy_version[0] = cpu_to_le32(cmd->policy_version[0]);
+	disk_super->policy_version[1] = cpu_to_le32(cmd->policy_version[1]);
+	disk_super->policy_version[2] = cpu_to_le32(cmd->policy_version[2]);
+
+	disk_super->read_hits = cpu_to_le32(cmd->stats.read_hits);
+	disk_super->read_misses = cpu_to_le32(cmd->stats.read_misses);
+	disk_super->write_hits = cpu_to_le32(cmd->stats.write_hits);
+	disk_super->write_misses = cpu_to_le32(cmd->stats.write_misses);
+	__copy_sm_root(cmd, disk_super);
+
+	return dm_tm_commit(cmd->tm, sblock);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * The mappings are held in a dm-array that has 64-bit values stored in
+ * little-endian format.  The index is the cblock, the high 48bits of the
+ * value are the oblock and the low 16 bit the flags.
+ */
+#define FLAGS_MASK ((1 << 16) - 1)
+
+static __le64 pack_value(dm_oblock_t block, unsigned flags)
+{
+	uint64_t value = from_oblock(block);
+	value <<= 16;
+	value = value | (flags & FLAGS_MASK);
+	return cpu_to_le64(value);
+}
+
+static void unpack_value(__le64 value_le, dm_oblock_t *block, unsigned *flags)
+{
+	uint64_t value = le64_to_cpu(value_le);
+	uint64_t b = value >> 16;
+	*block = to_oblock(b);
+	*flags = value & FLAGS_MASK;
+}
+
+/*----------------------------------------------------------------*/
+
+struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
+						 sector_t data_block_size,
+						 bool may_format_device,
+						 size_t policy_hint_size)
+{
+	int r;
+	struct dm_cache_metadata *cmd;
+
+	cmd = kzalloc(sizeof(*cmd), GFP_KERNEL);
+	if (!cmd) {
+		DMERR("could not allocate metadata struct");
+		return NULL;
+	}
+
+	init_rwsem(&cmd->root_lock);
+	cmd->bdev = bdev;
+	cmd->data_block_size = data_block_size;
+	cmd->cache_blocks = 0;
+	cmd->policy_hint_size = policy_hint_size;
+	cmd->changed = true;
+
+	r = __create_persistent_data_objects(cmd, may_format_device);
+	if (r) {
+		kfree(cmd);
+		return ERR_PTR(r);
+	}
+
+	r = __begin_transaction_flags(cmd, clear_clean_shutdown);
+	if (r < 0) {
+		dm_cache_metadata_close(cmd);
+		return ERR_PTR(r);
+	}
+
+	return cmd;
+}
+
+void dm_cache_metadata_close(struct dm_cache_metadata *cmd)
+{
+	__destroy_persistent_data_objects(cmd);
+	kfree(cmd);
+}
+
+/*
+ * Checks that the given cache block is either unmapped or clean.
+ */
+static int block_unmapped_or_clean(struct dm_cache_metadata *cmd, dm_cblock_t b,
+				   bool *result)
+{
+	int r;
+	__le64 value;
+	dm_oblock_t ob;
+	unsigned flags;
+
+	r = dm_array_get_value(&cmd->info, cmd->root, from_cblock(b), &value);
+	if (r) {
+		DMERR("block_unmapped_or_clean failed");
+		return r;
+	}
+
+	unpack_value(value, &ob, &flags);
+	*result = !((flags & M_VALID) && (flags & M_DIRTY));
+
+	return 0;
+}
+
+static int blocks_are_unmapped_or_clean(struct dm_cache_metadata *cmd,
+					dm_cblock_t begin, dm_cblock_t end,
+					bool *result)
+{
+	int r;
+	*result = true;
+
+	while (begin != end) {
+		r = block_unmapped_or_clean(cmd, begin, result);
+		if (r)
+			return r;
+
+		if (!*result) {
+			DMERR("cache block %llu is dirty",
+			      (unsigned long long) from_cblock(begin));
+			return 0;
+		}
+
+		begin = to_cblock(from_cblock(begin) + 1);
+	}
+
+	return 0;
+}
+
+int dm_cache_resize(struct dm_cache_metadata *cmd, dm_cblock_t new_cache_size)
+{
+	int r;
+	bool clean;
+	__le64 null_mapping = pack_value(0, 0);
+
+	down_write(&cmd->root_lock);
+	__dm_bless_for_disk(&null_mapping);
+
+	if (from_cblock(new_cache_size) < from_cblock(cmd->cache_blocks)) {
+		r = blocks_are_unmapped_or_clean(cmd, new_cache_size, cmd->cache_blocks, &clean);
+		if (r) {
+			__dm_unbless_for_disk(&null_mapping);
+			goto out;
+		}
+
+		if (!clean) {
+			DMERR("unable to shrink cache due to dirty blocks");
+			r = -EINVAL;
+			__dm_unbless_for_disk(&null_mapping);
+			goto out;
+		}
+	}
+
+	r = dm_array_resize(&cmd->info, cmd->root, from_cblock(cmd->cache_blocks),
+			    from_cblock(new_cache_size),
+			    &null_mapping, &cmd->root);
+	if (!r)
+		cmd->cache_blocks = new_cache_size;
+	cmd->changed = true;
+
+out:
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+int dm_cache_discard_bitset_resize(struct dm_cache_metadata *cmd,
+				   sector_t discard_block_size,
+				   dm_oblock_t new_nr_entries)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = dm_bitset_resize(&cmd->discard_info,
+			     cmd->discard_root,
+			     from_oblock(cmd->discard_nr_blocks),
+			     from_oblock(new_nr_entries),
+			     false, &cmd->discard_root);
+	if (!r) {
+		cmd->discard_block_size = discard_block_size;
+		cmd->discard_nr_blocks = new_nr_entries;
+	}
+
+	cmd->changed = true;
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+static int __set_discard(struct dm_cache_metadata *cmd, dm_oblock_t b)
+{
+	return dm_bitset_set_bit(&cmd->discard_info, cmd->discard_root,
+				 from_oblock(b), &cmd->discard_root);
+}
+
+static int __clear_discard(struct dm_cache_metadata *cmd, dm_oblock_t b)
+{
+	return dm_bitset_clear_bit(&cmd->discard_info, cmd->discard_root,
+				   from_oblock(b), &cmd->discard_root);
+}
+
+static int __is_discarded(struct dm_cache_metadata *cmd, dm_oblock_t b,
+			  bool *is_discarded)
+{
+	return dm_bitset_test_bit(&cmd->discard_info, cmd->discard_root,
+				  from_oblock(b), &cmd->discard_root,
+				  is_discarded);
+}
+
+static int __discard(struct dm_cache_metadata *cmd,
+		     dm_oblock_t dblock, bool discard)
+{
+	int r;
+
+	r = (discard ? __set_discard : __clear_discard)(cmd, dblock);
+	if (r)
+		return r;
+
+	cmd->changed = true;
+	return 0;
+}
+
+int dm_cache_set_discard(struct dm_cache_metadata *cmd,
+			 dm_oblock_t dblock, bool discard)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = __discard(cmd, dblock, discard);
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+static int __load_discards(struct dm_cache_metadata *cmd,
+			   load_discard_fn fn, void *context)
+{
+	int r = 0;
+	dm_block_t b;
+	bool discard;
+
+	for (b = 0; b < from_oblock(cmd->discard_nr_blocks); b++) {
+		dm_oblock_t dblock = to_oblock(b);
+
+		if (cmd->clean_when_opened) {
+			r = __is_discarded(cmd, dblock, &discard);
+			if (r)
+				return r;
+		} else
+			discard = false;
+
+		r = fn(context, cmd->discard_block_size, dblock, discard);
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+int dm_cache_load_discards(struct dm_cache_metadata *cmd,
+			   load_discard_fn fn, void *context)
+{
+	int r;
+
+	down_read(&cmd->root_lock);
+	r = __load_discards(cmd, fn, context);
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+dm_cblock_t dm_cache_size(struct dm_cache_metadata *cmd)
+{
+	dm_cblock_t r;
+
+	down_read(&cmd->root_lock);
+	r = cmd->cache_blocks;
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+static int __remove(struct dm_cache_metadata *cmd, dm_cblock_t cblock)
+{
+	int r;
+	__le64 value = pack_value(0, 0);
+
+	__dm_bless_for_disk(&value);
+	r = dm_array_set_value(&cmd->info, cmd->root, from_cblock(cblock),
+			       &value, &cmd->root);
+	if (r)
+		return r;
+
+	cmd->changed = true;
+	return 0;
+}
+
+int dm_cache_remove_mapping(struct dm_cache_metadata *cmd, dm_cblock_t cblock)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = __remove(cmd, cblock);
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+static int __insert(struct dm_cache_metadata *cmd,
+		    dm_cblock_t cblock, dm_oblock_t oblock)
+{
+	int r;
+	__le64 value = pack_value(oblock, M_VALID);
+	__dm_bless_for_disk(&value);
+
+	r = dm_array_set_value(&cmd->info, cmd->root, from_cblock(cblock),
+			       &value, &cmd->root);
+	if (r)
+		return r;
+
+	cmd->changed = true;
+	return 0;
+}
+
+int dm_cache_insert_mapping(struct dm_cache_metadata *cmd,
+			    dm_cblock_t cblock, dm_oblock_t oblock)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = __insert(cmd, cblock, oblock);
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+struct thunk {
+	load_mapping_fn fn;
+	void *context;
+
+	struct dm_cache_metadata *cmd;
+	bool respect_dirty_flags;
+	bool hints_valid;
+};
+
+static bool policy_unchanged(struct dm_cache_metadata *cmd,
+			     struct dm_cache_policy *policy)
+{
+	const char *policy_name = dm_cache_policy_get_name(policy);
+	const unsigned *policy_version = dm_cache_policy_get_version(policy);
+	size_t policy_hint_size = dm_cache_policy_get_hint_size(policy);
+
+	/*
+	 * Ensure policy names match.
+	 */
+	if (strncmp(cmd->policy_name, policy_name, sizeof(cmd->policy_name)))
+		return false;
+
+	/*
+	 * Ensure policy major versions match.
+	 */
+	if (cmd->policy_version[0] != policy_version[0])
+		return false;
+
+	/*
+	 * Ensure policy hint sizes match.
+	 */
+	if (cmd->policy_hint_size != policy_hint_size)
+		return false;
+
+	return true;
+}
+
+static bool hints_array_initialized(struct dm_cache_metadata *cmd)
+{
+	return cmd->hint_root && cmd->policy_hint_size;
+}
+
+static bool hints_array_available(struct dm_cache_metadata *cmd,
+				  struct dm_cache_policy *policy)
+{
+	return cmd->clean_when_opened && policy_unchanged(cmd, policy) &&
+		hints_array_initialized(cmd);
+}
+
+static int __load_mapping(void *context, uint64_t cblock, void *leaf)
+{
+	int r = 0;
+	bool dirty;
+	__le64 value;
+	__le32 hint_value = 0;
+	dm_oblock_t oblock;
+	unsigned flags;
+	struct thunk *thunk = context;
+	struct dm_cache_metadata *cmd = thunk->cmd;
+
+	memcpy(&value, leaf, sizeof(value));
+	unpack_value(value, &oblock, &flags);
+
+	if (flags & M_VALID) {
+		if (thunk->hints_valid) {
+			r = dm_array_get_value(&cmd->hint_info, cmd->hint_root,
+					       cblock, &hint_value);
+			if (r && r != -ENODATA)
+				return r;
+		}
+
+		dirty = thunk->respect_dirty_flags ? (flags & M_DIRTY) : true;
+		r = thunk->fn(thunk->context, oblock, to_cblock(cblock),
+			      dirty, le32_to_cpu(hint_value), thunk->hints_valid);
+	}
+
+	return r;
+}
+
+static int __load_mappings(struct dm_cache_metadata *cmd,
+			   struct dm_cache_policy *policy,
+			   load_mapping_fn fn, void *context)
+{
+	struct thunk thunk;
+
+	thunk.fn = fn;
+	thunk.context = context;
+
+	thunk.cmd = cmd;
+	thunk.respect_dirty_flags = cmd->clean_when_opened;
+	thunk.hints_valid = hints_array_available(cmd, policy);
+
+	return dm_array_walk(&cmd->info, cmd->root, __load_mapping, &thunk);
+}
+
+int dm_cache_load_mappings(struct dm_cache_metadata *cmd,
+			   struct dm_cache_policy *policy,
+			   load_mapping_fn fn, void *context)
+{
+	int r;
+
+	down_read(&cmd->root_lock);
+	r = __load_mappings(cmd, policy, fn, context);
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+static int __dump_mapping(void *context, uint64_t cblock, void *leaf)
+{
+	int r = 0;
+	__le64 value;
+	dm_oblock_t oblock;
+	unsigned flags;
+
+	memcpy(&value, leaf, sizeof(value));
+	unpack_value(value, &oblock, &flags);
+
+	return r;
+}
+
+static int __dump_mappings(struct dm_cache_metadata *cmd)
+{
+	return dm_array_walk(&cmd->info, cmd->root, __dump_mapping, NULL);
+}
+
+void dm_cache_dump(struct dm_cache_metadata *cmd)
+{
+	down_read(&cmd->root_lock);
+	__dump_mappings(cmd);
+	up_read(&cmd->root_lock);
+}
+
+int dm_cache_changed_this_transaction(struct dm_cache_metadata *cmd)
+{
+	int r;
+
+	down_read(&cmd->root_lock);
+	r = cmd->changed;
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+static int __dirty(struct dm_cache_metadata *cmd, dm_cblock_t cblock, bool dirty)
+{
+	int r;
+	unsigned flags;
+	dm_oblock_t oblock;
+	__le64 value;
+
+	r = dm_array_get_value(&cmd->info, cmd->root, from_cblock(cblock), &value);
+	if (r)
+		return r;
+
+	unpack_value(value, &oblock, &flags);
+
+	if (((flags & M_DIRTY) && dirty) || (!(flags & M_DIRTY) && !dirty))
+		/* nothing to be done */
+		return 0;
+
+	value = pack_value(oblock, (flags & ~M_DIRTY) | (dirty ? M_DIRTY : 0));
+	__dm_bless_for_disk(&value);
+
+	r = dm_array_set_value(&cmd->info, cmd->root, from_cblock(cblock),
+			       &value, &cmd->root);
+	if (r)
+		return r;
+
+	cmd->changed = true;
+	return 0;
+
+}
+
+int dm_cache_set_dirty(struct dm_cache_metadata *cmd,
+		       dm_cblock_t cblock, bool dirty)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = __dirty(cmd, cblock, dirty);
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+void dm_cache_metadata_get_stats(struct dm_cache_metadata *cmd,
+				 struct dm_cache_statistics *stats)
+{
+	down_read(&cmd->root_lock);
+	*stats = cmd->stats;
+	up_read(&cmd->root_lock);
+}
+
+void dm_cache_metadata_set_stats(struct dm_cache_metadata *cmd,
+				 struct dm_cache_statistics *stats)
+{
+	down_write(&cmd->root_lock);
+	cmd->stats = *stats;
+	up_write(&cmd->root_lock);
+}
+
+int dm_cache_commit(struct dm_cache_metadata *cmd, bool clean_shutdown)
+{
+	int r;
+	flags_mutator mutator = (clean_shutdown ? set_clean_shutdown :
+				 clear_clean_shutdown);
+
+	down_write(&cmd->root_lock);
+	r = __commit_transaction(cmd, mutator);
+	if (r)
+		goto out;
+
+	r = __begin_transaction(cmd);
+
+out:
+	up_write(&cmd->root_lock);
+	return r;
+}
+
+int dm_cache_get_free_metadata_block_count(struct dm_cache_metadata *cmd,
+					   dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&cmd->root_lock);
+	r = dm_sm_get_nr_free(cmd->metadata_sm, result);
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+int dm_cache_get_metadata_dev_size(struct dm_cache_metadata *cmd,
+				   dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&cmd->root_lock);
+	r = dm_sm_get_nr_blocks(cmd->metadata_sm, result);
+	up_read(&cmd->root_lock);
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+static int begin_hints(struct dm_cache_metadata *cmd, struct dm_cache_policy *policy)
+{
+	int r;
+	__le32 value;
+	size_t hint_size;
+	const char *policy_name = dm_cache_policy_get_name(policy);
+	const unsigned *policy_version = dm_cache_policy_get_version(policy);
+
+	if (!policy_name[0] ||
+	    (strlen(policy_name) > sizeof(cmd->policy_name) - 1))
+		return -EINVAL;
+
+	if (!policy_unchanged(cmd, policy)) {
+		strncpy(cmd->policy_name, policy_name, sizeof(cmd->policy_name));
+		memcpy(cmd->policy_version, policy_version, sizeof(cmd->policy_version));
+
+		hint_size = dm_cache_policy_get_hint_size(policy);
+		if (!hint_size)
+			return 0; /* short-circuit hints initialization */
+		cmd->policy_hint_size = hint_size;
+
+		if (cmd->hint_root) {
+			r = dm_array_del(&cmd->hint_info, cmd->hint_root);
+			if (r)
+				return r;
+		}
+
+		r = dm_array_empty(&cmd->hint_info, &cmd->hint_root);
+		if (r)
+			return r;
+
+		value = cpu_to_le32(0);
+		__dm_bless_for_disk(&value);
+		r = dm_array_resize(&cmd->hint_info, cmd->hint_root, 0,
+				    from_cblock(cmd->cache_blocks),
+				    &value, &cmd->hint_root);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int save_hint(void *context, dm_cblock_t cblock, dm_oblock_t oblock, uint32_t hint)
+{
+	struct dm_cache_metadata *cmd = context;
+	__le32 value = cpu_to_le32(hint);
+	int r;
+
+	__dm_bless_for_disk(&value);
+
+	r = dm_array_set_value(&cmd->hint_info, cmd->hint_root,
+			       from_cblock(cblock), &value, &cmd->hint_root);
+	cmd->changed = true;
+
+	return r;
+}
+
+static int write_hints(struct dm_cache_metadata *cmd, struct dm_cache_policy *policy)
+{
+	int r;
+
+	r = begin_hints(cmd, policy);
+	if (r) {
+		DMERR("begin_hints failed");
+		return r;
+	}
+
+	return policy_walk_mappings(policy, save_hint, cmd);
+}
+
+int dm_cache_write_hints(struct dm_cache_metadata *cmd, struct dm_cache_policy *policy)
+{
+	int r;
+
+	down_write(&cmd->root_lock);
+	r = write_hints(cmd, policy);
+	up_write(&cmd->root_lock);
+
+	return r;
+}
+
+int dm_cache_metadata_all_clean(struct dm_cache_metadata *cmd, bool *result)
+{
+	return blocks_are_unmapped_or_clean(cmd, 0, cmd->cache_blocks, result);
+}
diff --git a/drivers/md/dm-cache-metadata.h b/drivers/md/dm-cache-metadata.h
new file mode 100644
index 00000000000..cd70a78623a
--- /dev/null
+++ b/drivers/md/dm-cache-metadata.h
@@ -0,0 +1,140 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_CACHE_METADATA_H
+#define DM_CACHE_METADATA_H
+
+#include "dm-cache-block-types.h"
+#include "dm-cache-policy-internal.h"
+
+/*----------------------------------------------------------------*/
+
+#define DM_CACHE_METADATA_BLOCK_SIZE 4096
+
+/* FIXME: remove this restriction */
+/*
+ * The metadata device is currently limited in size.
+ *
+ * We have one block of index, which can hold 255 index entries.  Each
+ * index entry contains allocation info about 16k metadata blocks.
+ */
+#define DM_CACHE_METADATA_MAX_SECTORS (255 * (1 << 14) * (DM_CACHE_METADATA_BLOCK_SIZE / (1 << SECTOR_SHIFT)))
+
+/*
+ * A metadata device larger than 16GB triggers a warning.
+ */
+#define DM_CACHE_METADATA_MAX_SECTORS_WARNING (16 * (1024 * 1024 * 1024 >> SECTOR_SHIFT))
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Ext[234]-style compat feature flags.
+ *
+ * A new feature which old metadata will still be compatible with should
+ * define a DM_CACHE_FEATURE_COMPAT_* flag (rarely useful).
+ *
+ * A new feature that is not compatible with old code should define a
+ * DM_CACHE_FEATURE_INCOMPAT_* flag and guard the relevant code with
+ * that flag.
+ *
+ * A new feature that is not compatible with old code accessing the
+ * metadata RDWR should define a DM_CACHE_FEATURE_RO_COMPAT_* flag and
+ * guard the relevant code with that flag.
+ *
+ * As these various flags are defined they should be added to the
+ * following masks.
+ */
+#define DM_CACHE_FEATURE_COMPAT_SUPP	  0UL
+#define DM_CACHE_FEATURE_COMPAT_RO_SUPP	  0UL
+#define DM_CACHE_FEATURE_INCOMPAT_SUPP	  0UL
+
+/*
+ * Reopens or creates a new, empty metadata volume.
+ * Returns an ERR_PTR on failure.
+ */
+struct dm_cache_metadata *dm_cache_metadata_open(struct block_device *bdev,
+						 sector_t data_block_size,
+						 bool may_format_device,
+						 size_t policy_hint_size);
+
+void dm_cache_metadata_close(struct dm_cache_metadata *cmd);
+
+/*
+ * The metadata needs to know how many cache blocks there are.  We don't
+ * care about the origin, assuming the core target is giving us valid
+ * origin blocks to map to.
+ */
+int dm_cache_resize(struct dm_cache_metadata *cmd, dm_cblock_t new_cache_size);
+dm_cblock_t dm_cache_size(struct dm_cache_metadata *cmd);
+
+int dm_cache_discard_bitset_resize(struct dm_cache_metadata *cmd,
+				   sector_t discard_block_size,
+				   dm_oblock_t new_nr_entries);
+
+typedef int (*load_discard_fn)(void *context, sector_t discard_block_size,
+			       dm_oblock_t dblock, bool discarded);
+int dm_cache_load_discards(struct dm_cache_metadata *cmd,
+			   load_discard_fn fn, void *context);
+
+int dm_cache_set_discard(struct dm_cache_metadata *cmd, dm_oblock_t dblock, bool discard);
+
+int dm_cache_remove_mapping(struct dm_cache_metadata *cmd, dm_cblock_t cblock);
+int dm_cache_insert_mapping(struct dm_cache_metadata *cmd, dm_cblock_t cblock, dm_oblock_t oblock);
+int dm_cache_changed_this_transaction(struct dm_cache_metadata *cmd);
+
+typedef int (*load_mapping_fn)(void *context, dm_oblock_t oblock,
+			       dm_cblock_t cblock, bool dirty,
+			       uint32_t hint, bool hint_valid);
+int dm_cache_load_mappings(struct dm_cache_metadata *cmd,
+			   struct dm_cache_policy *policy,
+			   load_mapping_fn fn,
+			   void *context);
+
+int dm_cache_set_dirty(struct dm_cache_metadata *cmd, dm_cblock_t cblock, bool dirty);
+
+struct dm_cache_statistics {
+	uint32_t read_hits;
+	uint32_t read_misses;
+	uint32_t write_hits;
+	uint32_t write_misses;
+};
+
+void dm_cache_metadata_get_stats(struct dm_cache_metadata *cmd,
+				 struct dm_cache_statistics *stats);
+void dm_cache_metadata_set_stats(struct dm_cache_metadata *cmd,
+				 struct dm_cache_statistics *stats);
+
+int dm_cache_commit(struct dm_cache_metadata *cmd, bool clean_shutdown);
+
+int dm_cache_get_free_metadata_block_count(struct dm_cache_metadata *cmd,
+					   dm_block_t *result);
+
+int dm_cache_get_metadata_dev_size(struct dm_cache_metadata *cmd,
+				   dm_block_t *result);
+
+void dm_cache_dump(struct dm_cache_metadata *cmd);
+
+/*
+ * The policy is invited to save a 32bit hint value for every cblock (eg,
+ * for a hit count).  These are stored against the policy name.  If
+ * policies are changed, then hints will be lost.  If the machine crashes,
+ * hints will be lost.
+ *
+ * The hints are indexed by the cblock, but many policies will not
+ * neccessarily have a fast way of accessing efficiently via cblock.  So
+ * rather than querying the policy for each cblock, we let it walk its data
+ * structures and fill in the hints in whatever order it wishes.
+ */
+int dm_cache_write_hints(struct dm_cache_metadata *cmd, struct dm_cache_policy *p);
+
+/*
+ * Query method.  Are all the blocks in the cache clean?
+ */
+int dm_cache_metadata_all_clean(struct dm_cache_metadata *cmd, bool *result);
+
+/*----------------------------------------------------------------*/
+
+#endif /* DM_CACHE_METADATA_H */
diff --git a/drivers/md/dm-cache-policy-cleaner.c b/drivers/md/dm-cache-policy-cleaner.c
new file mode 100644
index 00000000000..b04d1f904d0
--- /dev/null
+++ b/drivers/md/dm-cache-policy-cleaner.c
@@ -0,0 +1,467 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * writeback cache policy supporting flushing out dirty cache blocks.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-cache-policy.h"
+#include "dm.h"
+
+#include <linux/hash.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+/*----------------------------------------------------------------*/
+
+#define DM_MSG_PREFIX "cache cleaner"
+
+/* Cache entry struct. */
+struct wb_cache_entry {
+	struct list_head list;
+	struct hlist_node hlist;
+
+	dm_oblock_t oblock;
+	dm_cblock_t cblock;
+	bool dirty:1;
+	bool pending:1;
+};
+
+struct hash {
+	struct hlist_head *table;
+	dm_block_t hash_bits;
+	unsigned nr_buckets;
+};
+
+struct policy {
+	struct dm_cache_policy policy;
+	spinlock_t lock;
+
+	struct list_head free;
+	struct list_head clean;
+	struct list_head clean_pending;
+	struct list_head dirty;
+
+	/*
+	 * We know exactly how many cblocks will be needed,
+	 * so we can allocate them up front.
+	 */
+	dm_cblock_t cache_size, nr_cblocks_allocated;
+	struct wb_cache_entry *cblocks;
+	struct hash chash;
+};
+
+/*----------------------------------------------------------------------------*/
+
+/*
+ * Low-level functions.
+ */
+static unsigned next_power(unsigned n, unsigned min)
+{
+	return roundup_pow_of_two(max(n, min));
+}
+
+static struct policy *to_policy(struct dm_cache_policy *p)
+{
+	return container_of(p, struct policy, policy);
+}
+
+static struct list_head *list_pop(struct list_head *q)
+{
+	struct list_head *r = q->next;
+
+	list_del(r);
+
+	return r;
+}
+
+/*----------------------------------------------------------------------------*/
+
+/* Allocate/free various resources. */
+static int alloc_hash(struct hash *hash, unsigned elts)
+{
+	hash->nr_buckets = next_power(elts >> 4, 16);
+	hash->hash_bits = ffs(hash->nr_buckets) - 1;
+	hash->table = vzalloc(sizeof(*hash->table) * hash->nr_buckets);
+
+	return hash->table ? 0 : -ENOMEM;
+}
+
+static void free_hash(struct hash *hash)
+{
+	vfree(hash->table);
+}
+
+static int alloc_cache_blocks_with_hash(struct policy *p, dm_cblock_t cache_size)
+{
+	int r = -ENOMEM;
+
+	p->cblocks = vzalloc(sizeof(*p->cblocks) * from_cblock(cache_size));
+	if (p->cblocks) {
+		unsigned u = from_cblock(cache_size);
+
+		while (u--)
+			list_add(&p->cblocks[u].list, &p->free);
+
+		p->nr_cblocks_allocated = 0;
+
+		/* Cache entries hash. */
+		r = alloc_hash(&p->chash, from_cblock(cache_size));
+		if (r)
+			vfree(p->cblocks);
+	}
+
+	return r;
+}
+
+static void free_cache_blocks_and_hash(struct policy *p)
+{
+	free_hash(&p->chash);
+	vfree(p->cblocks);
+}
+
+static struct wb_cache_entry *alloc_cache_entry(struct policy *p)
+{
+	struct wb_cache_entry *e;
+
+	BUG_ON(from_cblock(p->nr_cblocks_allocated) >= from_cblock(p->cache_size));
+
+	e = list_entry(list_pop(&p->free), struct wb_cache_entry, list);
+	p->nr_cblocks_allocated = to_cblock(from_cblock(p->nr_cblocks_allocated) + 1);
+
+	return e;
+}
+
+/*----------------------------------------------------------------------------*/
+
+/* Hash functions (lookup, insert, remove). */
+static struct wb_cache_entry *lookup_cache_entry(struct policy *p, dm_oblock_t oblock)
+{
+	struct hash *hash = &p->chash;
+	unsigned h = hash_64(from_oblock(oblock), hash->hash_bits);
+	struct wb_cache_entry *cur;
+	struct hlist_head *bucket = &hash->table[h];
+
+	hlist_for_each_entry(cur, bucket, hlist) {
+		if (cur->oblock == oblock) {
+			/* Move upfront bucket for faster access. */
+			hlist_del(&cur->hlist);
+			hlist_add_head(&cur->hlist, bucket);
+			return cur;
+		}
+	}
+
+	return NULL;
+}
+
+static void insert_cache_hash_entry(struct policy *p, struct wb_cache_entry *e)
+{
+	unsigned h = hash_64(from_oblock(e->oblock), p->chash.hash_bits);
+
+	hlist_add_head(&e->hlist, &p->chash.table[h]);
+}
+
+static void remove_cache_hash_entry(struct wb_cache_entry *e)
+{
+	hlist_del(&e->hlist);
+}
+
+/* Public interface (see dm-cache-policy.h */
+static int wb_map(struct dm_cache_policy *pe, dm_oblock_t oblock,
+		  bool can_block, bool can_migrate, bool discarded_oblock,
+		  struct bio *bio, struct policy_result *result)
+{
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+	unsigned long flags;
+
+	result->op = POLICY_MISS;
+
+	if (can_block)
+		spin_lock_irqsave(&p->lock, flags);
+
+	else if (!spin_trylock_irqsave(&p->lock, flags))
+		return -EWOULDBLOCK;
+
+	e = lookup_cache_entry(p, oblock);
+	if (e) {
+		result->op = POLICY_HIT;
+		result->cblock = e->cblock;
+
+	}
+
+	spin_unlock_irqrestore(&p->lock, flags);
+
+	return 0;
+}
+
+static int wb_lookup(struct dm_cache_policy *pe, dm_oblock_t oblock, dm_cblock_t *cblock)
+{
+	int r;
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+	unsigned long flags;
+
+	if (!spin_trylock_irqsave(&p->lock, flags))
+		return -EWOULDBLOCK;
+
+	e = lookup_cache_entry(p, oblock);
+	if (e) {
+		*cblock = e->cblock;
+		r = 0;
+
+	} else
+		r = -ENOENT;
+
+	spin_unlock_irqrestore(&p->lock, flags);
+
+	return r;
+}
+
+static void __set_clear_dirty(struct dm_cache_policy *pe, dm_oblock_t oblock, bool set)
+{
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+
+	e = lookup_cache_entry(p, oblock);
+	BUG_ON(!e);
+
+	if (set) {
+		if (!e->dirty) {
+			e->dirty = true;
+			list_move(&e->list, &p->dirty);
+		}
+
+	} else {
+		if (e->dirty) {
+			e->pending = false;
+			e->dirty = false;
+			list_move(&e->list, &p->clean);
+		}
+	}
+}
+
+static void wb_set_dirty(struct dm_cache_policy *pe, dm_oblock_t oblock)
+{
+	struct policy *p = to_policy(pe);
+	unsigned long flags;
+
+	spin_lock_irqsave(&p->lock, flags);
+	__set_clear_dirty(pe, oblock, true);
+	spin_unlock_irqrestore(&p->lock, flags);
+}
+
+static void wb_clear_dirty(struct dm_cache_policy *pe, dm_oblock_t oblock)
+{
+	struct policy *p = to_policy(pe);
+	unsigned long flags;
+
+	spin_lock_irqsave(&p->lock, flags);
+	__set_clear_dirty(pe, oblock, false);
+	spin_unlock_irqrestore(&p->lock, flags);
+}
+
+static void add_cache_entry(struct policy *p, struct wb_cache_entry *e)
+{
+	insert_cache_hash_entry(p, e);
+	if (e->dirty)
+		list_add(&e->list, &p->dirty);
+	else
+		list_add(&e->list, &p->clean);
+}
+
+static int wb_load_mapping(struct dm_cache_policy *pe,
+			   dm_oblock_t oblock, dm_cblock_t cblock,
+			   uint32_t hint, bool hint_valid)
+{
+	int r;
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e = alloc_cache_entry(p);
+
+	if (e) {
+		e->cblock = cblock;
+		e->oblock = oblock;
+		e->dirty = false; /* blocks default to clean */
+		add_cache_entry(p, e);
+		r = 0;
+
+	} else
+		r = -ENOMEM;
+
+	return r;
+}
+
+static void wb_destroy(struct dm_cache_policy *pe)
+{
+	struct policy *p = to_policy(pe);
+
+	free_cache_blocks_and_hash(p);
+	kfree(p);
+}
+
+static struct wb_cache_entry *__wb_force_remove_mapping(struct policy *p, dm_oblock_t oblock)
+{
+	struct wb_cache_entry *r = lookup_cache_entry(p, oblock);
+
+	BUG_ON(!r);
+
+	remove_cache_hash_entry(r);
+	list_del(&r->list);
+
+	return r;
+}
+
+static void wb_remove_mapping(struct dm_cache_policy *pe, dm_oblock_t oblock)
+{
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+	unsigned long flags;
+
+	spin_lock_irqsave(&p->lock, flags);
+	e = __wb_force_remove_mapping(p, oblock);
+	list_add_tail(&e->list, &p->free);
+	BUG_ON(!from_cblock(p->nr_cblocks_allocated));
+	p->nr_cblocks_allocated = to_cblock(from_cblock(p->nr_cblocks_allocated) - 1);
+	spin_unlock_irqrestore(&p->lock, flags);
+}
+
+static void wb_force_mapping(struct dm_cache_policy *pe,
+				dm_oblock_t current_oblock, dm_oblock_t oblock)
+{
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+	unsigned long flags;
+
+	spin_lock_irqsave(&p->lock, flags);
+	e = __wb_force_remove_mapping(p, current_oblock);
+	e->oblock = oblock;
+	add_cache_entry(p, e);
+	spin_unlock_irqrestore(&p->lock, flags);
+}
+
+static struct wb_cache_entry *get_next_dirty_entry(struct policy *p)
+{
+	struct list_head *l;
+	struct wb_cache_entry *r;
+
+	if (list_empty(&p->dirty))
+		return NULL;
+
+	l = list_pop(&p->dirty);
+	r = container_of(l, struct wb_cache_entry, list);
+	list_add(l, &p->clean_pending);
+
+	return r;
+}
+
+static int wb_writeback_work(struct dm_cache_policy *pe,
+			     dm_oblock_t *oblock,
+			     dm_cblock_t *cblock)
+{
+	int r = -ENOENT;
+	struct policy *p = to_policy(pe);
+	struct wb_cache_entry *e;
+	unsigned long flags;
+
+	spin_lock_irqsave(&p->lock, flags);
+
+	e = get_next_dirty_entry(p);
+	if (e) {
+		*oblock = e->oblock;
+		*cblock = e->cblock;
+		r = 0;
+	}
+
+	spin_unlock_irqrestore(&p->lock, flags);
+
+	return r;
+}
+
+static dm_cblock_t wb_residency(struct dm_cache_policy *pe)
+{
+	return to_policy(pe)->nr_cblocks_allocated;
+}
+
+/* Init the policy plugin interface function pointers. */
+static void init_policy_functions(struct policy *p)
+{
+	p->policy.destroy = wb_destroy;
+	p->policy.map = wb_map;
+	p->policy.lookup = wb_lookup;
+	p->policy.set_dirty = wb_set_dirty;
+	p->policy.clear_dirty = wb_clear_dirty;
+	p->policy.load_mapping = wb_load_mapping;
+	p->policy.walk_mappings = NULL;
+	p->policy.remove_mapping = wb_remove_mapping;
+	p->policy.writeback_work = wb_writeback_work;
+	p->policy.force_mapping = wb_force_mapping;
+	p->policy.residency = wb_residency;
+	p->policy.tick = NULL;
+}
+
+static struct dm_cache_policy *wb_create(dm_cblock_t cache_size,
+					 sector_t origin_size,
+					 sector_t cache_block_size)
+{
+	int r;
+	struct policy *p = kzalloc(sizeof(*p), GFP_KERNEL);
+
+	if (!p)
+		return NULL;
+
+	init_policy_functions(p);
+	INIT_LIST_HEAD(&p->free);
+	INIT_LIST_HEAD(&p->clean);
+	INIT_LIST_HEAD(&p->clean_pending);
+	INIT_LIST_HEAD(&p->dirty);
+
+	p->cache_size = cache_size;
+	spin_lock_init(&p->lock);
+
+	/* Allocate cache entry structs and add them to free list. */
+	r = alloc_cache_blocks_with_hash(p, cache_size);
+	if (!r)
+		return &p->policy;
+
+	kfree(p);
+
+	return NULL;
+}
+/*----------------------------------------------------------------------------*/
+
+static struct dm_cache_policy_type wb_policy_type = {
+	.name = "cleaner",
+	.version = {1, 0, 0},
+	.hint_size = 0,
+	.owner = THIS_MODULE,
+	.create = wb_create
+};
+
+static int __init wb_init(void)
+{
+	int r = dm_cache_policy_register(&wb_policy_type);
+
+	if (r < 0)
+		DMERR("register failed %d", r);
+	else
+		DMINFO("version %u.%u.%u loaded",
+		       wb_policy_type.version[0],
+		       wb_policy_type.version[1],
+		       wb_policy_type.version[2]);
+
+	return r;
+}
+
+static void __exit wb_exit(void)
+{
+	dm_cache_policy_unregister(&wb_policy_type);
+}
+
+module_init(wb_init);
+module_exit(wb_exit);
+
+MODULE_AUTHOR("Heinz Mauelshagen <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("cleaner cache policy");
diff --git a/drivers/md/dm-cache-policy-internal.h b/drivers/md/dm-cache-policy-internal.h
new file mode 100644
index 00000000000..2256a1f24f7
--- /dev/null
+++ b/drivers/md/dm-cache-policy-internal.h
@@ -0,0 +1,131 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_CACHE_POLICY_INTERNAL_H
+#define DM_CACHE_POLICY_INTERNAL_H
+
+#include "dm-cache-policy.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Little inline functions that simplify calling the policy methods.
+ */
+static inline int policy_map(struct dm_cache_policy *p, dm_oblock_t oblock,
+			     bool can_block, bool can_migrate, bool discarded_oblock,
+			     struct bio *bio, struct policy_result *result)
+{
+	return p->map(p, oblock, can_block, can_migrate, discarded_oblock, bio, result);
+}
+
+static inline int policy_lookup(struct dm_cache_policy *p, dm_oblock_t oblock, dm_cblock_t *cblock)
+{
+	BUG_ON(!p->lookup);
+	return p->lookup(p, oblock, cblock);
+}
+
+static inline void policy_set_dirty(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	if (p->set_dirty)
+		p->set_dirty(p, oblock);
+}
+
+static inline void policy_clear_dirty(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	if (p->clear_dirty)
+		p->clear_dirty(p, oblock);
+}
+
+static inline int policy_load_mapping(struct dm_cache_policy *p,
+				      dm_oblock_t oblock, dm_cblock_t cblock,
+				      uint32_t hint, bool hint_valid)
+{
+	return p->load_mapping(p, oblock, cblock, hint, hint_valid);
+}
+
+static inline int policy_walk_mappings(struct dm_cache_policy *p,
+				      policy_walk_fn fn, void *context)
+{
+	return p->walk_mappings ? p->walk_mappings(p, fn, context) : 0;
+}
+
+static inline int policy_writeback_work(struct dm_cache_policy *p,
+					dm_oblock_t *oblock,
+					dm_cblock_t *cblock)
+{
+	return p->writeback_work ? p->writeback_work(p, oblock, cblock) : -ENOENT;
+}
+
+static inline void policy_remove_mapping(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	p->remove_mapping(p, oblock);
+}
+
+static inline int policy_remove_cblock(struct dm_cache_policy *p, dm_cblock_t cblock)
+{
+	return p->remove_cblock(p, cblock);
+}
+
+static inline void policy_force_mapping(struct dm_cache_policy *p,
+					dm_oblock_t current_oblock, dm_oblock_t new_oblock)
+{
+	return p->force_mapping(p, current_oblock, new_oblock);
+}
+
+static inline dm_cblock_t policy_residency(struct dm_cache_policy *p)
+{
+	return p->residency(p);
+}
+
+static inline void policy_tick(struct dm_cache_policy *p)
+{
+	if (p->tick)
+		return p->tick(p);
+}
+
+static inline int policy_emit_config_values(struct dm_cache_policy *p, char *result, unsigned maxlen)
+{
+	ssize_t sz = 0;
+	if (p->emit_config_values)
+		return p->emit_config_values(p, result, maxlen);
+
+	DMEMIT("0");
+	return 0;
+}
+
+static inline int policy_set_config_value(struct dm_cache_policy *p,
+					  const char *key, const char *value)
+{
+	return p->set_config_value ? p->set_config_value(p, key, value) : -EINVAL;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Creates a new cache policy given a policy name, a cache size, an origin size and the block size.
+ */
+struct dm_cache_policy *dm_cache_policy_create(const char *name, dm_cblock_t cache_size,
+					       sector_t origin_size, sector_t block_size);
+
+/*
+ * Destroys the policy.  This drops references to the policy module as well
+ * as calling it's destroy method.  So always use this rather than calling
+ * the policy->destroy method directly.
+ */
+void dm_cache_policy_destroy(struct dm_cache_policy *p);
+
+/*
+ * In case we've forgotten.
+ */
+const char *dm_cache_policy_get_name(struct dm_cache_policy *p);
+
+const unsigned *dm_cache_policy_get_version(struct dm_cache_policy *p);
+
+size_t dm_cache_policy_get_hint_size(struct dm_cache_policy *p);
+
+/*----------------------------------------------------------------*/
+
+#endif /* DM_CACHE_POLICY_INTERNAL_H */
diff --git a/drivers/md/dm-cache-policy-mq.c b/drivers/md/dm-cache-policy-mq.c
new file mode 100644
index 00000000000..0e385e40909
--- /dev/null
+++ b/drivers/md/dm-cache-policy-mq.c
@@ -0,0 +1,1333 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-cache-policy.h"
+#include "dm.h"
+
+#include <linux/hash.h>
+#include <linux/module.h>
+#include <linux/mutex.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#define DM_MSG_PREFIX "cache-policy-mq"
+
+static struct kmem_cache *mq_entry_cache;
+
+/*----------------------------------------------------------------*/
+
+static unsigned next_power(unsigned n, unsigned min)
+{
+	return roundup_pow_of_two(max(n, min));
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Large, sequential ios are probably better left on the origin device since
+ * spindles tend to have good bandwidth.
+ *
+ * The io_tracker tries to spot when the io is in one of these sequential
+ * modes.
+ *
+ * Two thresholds to switch between random and sequential io mode are defaulting
+ * as follows and can be adjusted via the constructor and message interfaces.
+ */
+#define RANDOM_THRESHOLD_DEFAULT 4
+#define SEQUENTIAL_THRESHOLD_DEFAULT 512
+
+enum io_pattern {
+	PATTERN_SEQUENTIAL,
+	PATTERN_RANDOM
+};
+
+struct io_tracker {
+	enum io_pattern pattern;
+
+	unsigned nr_seq_samples;
+	unsigned nr_rand_samples;
+	unsigned thresholds[2];
+
+	dm_oblock_t last_end_oblock;
+};
+
+static void iot_init(struct io_tracker *t,
+		     int sequential_threshold, int random_threshold)
+{
+	t->pattern = PATTERN_RANDOM;
+	t->nr_seq_samples = 0;
+	t->nr_rand_samples = 0;
+	t->last_end_oblock = 0;
+	t->thresholds[PATTERN_RANDOM] = random_threshold;
+	t->thresholds[PATTERN_SEQUENTIAL] = sequential_threshold;
+}
+
+static enum io_pattern iot_pattern(struct io_tracker *t)
+{
+	return t->pattern;
+}
+
+static void iot_update_stats(struct io_tracker *t, struct bio *bio)
+{
+	if (bio->bi_iter.bi_sector == from_oblock(t->last_end_oblock) + 1)
+		t->nr_seq_samples++;
+	else {
+		/*
+		 * Just one non-sequential IO is enough to reset the
+		 * counters.
+		 */
+		if (t->nr_seq_samples) {
+			t->nr_seq_samples = 0;
+			t->nr_rand_samples = 0;
+		}
+
+		t->nr_rand_samples++;
+	}
+
+	t->last_end_oblock = to_oblock(bio_end_sector(bio) - 1);
+}
+
+static void iot_check_for_pattern_switch(struct io_tracker *t)
+{
+	switch (t->pattern) {
+	case PATTERN_SEQUENTIAL:
+		if (t->nr_rand_samples >= t->thresholds[PATTERN_RANDOM]) {
+			t->pattern = PATTERN_RANDOM;
+			t->nr_seq_samples = t->nr_rand_samples = 0;
+		}
+		break;
+
+	case PATTERN_RANDOM:
+		if (t->nr_seq_samples >= t->thresholds[PATTERN_SEQUENTIAL]) {
+			t->pattern = PATTERN_SEQUENTIAL;
+			t->nr_seq_samples = t->nr_rand_samples = 0;
+		}
+		break;
+	}
+}
+
+static void iot_examine_bio(struct io_tracker *t, struct bio *bio)
+{
+	iot_update_stats(t, bio);
+	iot_check_for_pattern_switch(t);
+}
+
+/*----------------------------------------------------------------*/
+
+
+/*
+ * This queue is divided up into different levels.  Allowing us to push
+ * entries to the back of any of the levels.  Think of it as a partially
+ * sorted queue.
+ */
+#define NR_QUEUE_LEVELS 16u
+
+struct queue {
+	struct list_head qs[NR_QUEUE_LEVELS];
+};
+
+static void queue_init(struct queue *q)
+{
+	unsigned i;
+
+	for (i = 0; i < NR_QUEUE_LEVELS; i++)
+		INIT_LIST_HEAD(q->qs + i);
+}
+
+/*
+ * Checks to see if the queue is empty.
+ * FIXME: reduce cpu usage.
+ */
+static bool queue_empty(struct queue *q)
+{
+	unsigned i;
+
+	for (i = 0; i < NR_QUEUE_LEVELS; i++)
+		if (!list_empty(q->qs + i))
+			return false;
+
+	return true;
+}
+
+/*
+ * Insert an entry to the back of the given level.
+ */
+static void queue_push(struct queue *q, unsigned level, struct list_head *elt)
+{
+	list_add_tail(elt, q->qs + level);
+}
+
+static void queue_remove(struct list_head *elt)
+{
+	list_del(elt);
+}
+
+/*
+ * Shifts all regions down one level.  This has no effect on the order of
+ * the queue.
+ */
+static void queue_shift_down(struct queue *q)
+{
+	unsigned level;
+
+	for (level = 1; level < NR_QUEUE_LEVELS; level++)
+		list_splice_init(q->qs + level, q->qs + level - 1);
+}
+
+/*
+ * Gives us the oldest entry of the lowest popoulated level.  If the first
+ * level is emptied then we shift down one level.
+ */
+static struct list_head *queue_pop(struct queue *q)
+{
+	unsigned level;
+	struct list_head *r;
+
+	for (level = 0; level < NR_QUEUE_LEVELS; level++)
+		if (!list_empty(q->qs + level)) {
+			r = q->qs[level].next;
+			list_del(r);
+
+			/* have we just emptied the bottom level? */
+			if (level == 0 && list_empty(q->qs))
+				queue_shift_down(q);
+
+			return r;
+		}
+
+	return NULL;
+}
+
+static struct list_head *list_pop(struct list_head *lh)
+{
+	struct list_head *r = lh->next;
+
+	BUG_ON(!r);
+	list_del_init(r);
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Describes a cache entry.  Used in both the cache and the pre_cache.
+ */
+struct entry {
+	struct hlist_node hlist;
+	struct list_head list;
+	dm_oblock_t oblock;
+
+	/*
+	 * FIXME: pack these better
+	 */
+	bool dirty:1;
+	unsigned hit_count;
+	unsigned generation;
+	unsigned tick;
+};
+
+/*
+ * Rather than storing the cblock in an entry, we allocate all entries in
+ * an array, and infer the cblock from the entry position.
+ *
+ * Free entries are linked together into a list.
+ */
+struct entry_pool {
+	struct entry *entries, *entries_end;
+	struct list_head free;
+	unsigned nr_allocated;
+};
+
+static int epool_init(struct entry_pool *ep, unsigned nr_entries)
+{
+	unsigned i;
+
+	ep->entries = vzalloc(sizeof(struct entry) * nr_entries);
+	if (!ep->entries)
+		return -ENOMEM;
+
+	ep->entries_end = ep->entries + nr_entries;
+
+	INIT_LIST_HEAD(&ep->free);
+	for (i = 0; i < nr_entries; i++)
+		list_add(&ep->entries[i].list, &ep->free);
+
+	ep->nr_allocated = 0;
+
+	return 0;
+}
+
+static void epool_exit(struct entry_pool *ep)
+{
+	vfree(ep->entries);
+}
+
+static struct entry *alloc_entry(struct entry_pool *ep)
+{
+	struct entry *e;
+
+	if (list_empty(&ep->free))
+		return NULL;
+
+	e = list_entry(list_pop(&ep->free), struct entry, list);
+	INIT_LIST_HEAD(&e->list);
+	INIT_HLIST_NODE(&e->hlist);
+	ep->nr_allocated++;
+
+	return e;
+}
+
+/*
+ * This assumes the cblock hasn't already been allocated.
+ */
+static struct entry *alloc_particular_entry(struct entry_pool *ep, dm_cblock_t cblock)
+{
+	struct entry *e = ep->entries + from_cblock(cblock);
+
+	list_del_init(&e->list);
+	INIT_HLIST_NODE(&e->hlist);
+	ep->nr_allocated++;
+
+	return e;
+}
+
+static void free_entry(struct entry_pool *ep, struct entry *e)
+{
+	BUG_ON(!ep->nr_allocated);
+	ep->nr_allocated--;
+	INIT_HLIST_NODE(&e->hlist);
+	list_add(&e->list, &ep->free);
+}
+
+/*
+ * Returns NULL if the entry is free.
+ */
+static struct entry *epool_find(struct entry_pool *ep, dm_cblock_t cblock)
+{
+	struct entry *e = ep->entries + from_cblock(cblock);
+	return !hlist_unhashed(&e->hlist) ? e : NULL;
+}
+
+static bool epool_empty(struct entry_pool *ep)
+{
+	return list_empty(&ep->free);
+}
+
+static bool in_pool(struct entry_pool *ep, struct entry *e)
+{
+	return e >= ep->entries && e < ep->entries_end;
+}
+
+static dm_cblock_t infer_cblock(struct entry_pool *ep, struct entry *e)
+{
+	return to_cblock(e - ep->entries);
+}
+
+/*----------------------------------------------------------------*/
+
+struct mq_policy {
+	struct dm_cache_policy policy;
+
+	/* protects everything */
+	struct mutex lock;
+	dm_cblock_t cache_size;
+	struct io_tracker tracker;
+
+	/*
+	 * Entries come from two pools, one of pre-cache entries, and one
+	 * for the cache proper.
+	 */
+	struct entry_pool pre_cache_pool;
+	struct entry_pool cache_pool;
+
+	/*
+	 * We maintain three queues of entries.  The cache proper,
+	 * consisting of a clean and dirty queue, contains the currently
+	 * active mappings.  Whereas the pre_cache tracks blocks that
+	 * are being hit frequently and potential candidates for promotion
+	 * to the cache.
+	 */
+	struct queue pre_cache;
+	struct queue cache_clean;
+	struct queue cache_dirty;
+
+	/*
+	 * Keeps track of time, incremented by the core.  We use this to
+	 * avoid attributing multiple hits within the same tick.
+	 *
+	 * Access to tick_protected should be done with the spin lock held.
+	 * It's copied to tick at the start of the map function (within the
+	 * mutex).
+	 */
+	spinlock_t tick_lock;
+	unsigned tick_protected;
+	unsigned tick;
+
+	/*
+	 * A count of the number of times the map function has been called
+	 * and found an entry in the pre_cache or cache.  Currently used to
+	 * calculate the generation.
+	 */
+	unsigned hit_count;
+
+	/*
+	 * A generation is a longish period that is used to trigger some
+	 * book keeping effects.  eg, decrementing hit counts on entries.
+	 * This is needed to allow the cache to evolve as io patterns
+	 * change.
+	 */
+	unsigned generation;
+	unsigned generation_period; /* in lookups (will probably change) */
+
+	/*
+	 * Entries in the pre_cache whose hit count passes the promotion
+	 * threshold move to the cache proper.  Working out the correct
+	 * value for the promotion_threshold is crucial to this policy.
+	 */
+	unsigned promote_threshold;
+
+	unsigned discard_promote_adjustment;
+	unsigned read_promote_adjustment;
+	unsigned write_promote_adjustment;
+
+	/*
+	 * The hash table allows us to quickly find an entry by origin
+	 * block.  Both pre_cache and cache entries are in here.
+	 */
+	unsigned nr_buckets;
+	dm_block_t hash_bits;
+	struct hlist_head *table;
+};
+
+#define DEFAULT_DISCARD_PROMOTE_ADJUSTMENT 1
+#define DEFAULT_READ_PROMOTE_ADJUSTMENT 4
+#define DEFAULT_WRITE_PROMOTE_ADJUSTMENT 8
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Simple hash table implementation.  Should replace with the standard hash
+ * table that's making its way upstream.
+ */
+static void hash_insert(struct mq_policy *mq, struct entry *e)
+{
+	unsigned h = hash_64(from_oblock(e->oblock), mq->hash_bits);
+
+	hlist_add_head(&e->hlist, mq->table + h);
+}
+
+static struct entry *hash_lookup(struct mq_policy *mq, dm_oblock_t oblock)
+{
+	unsigned h = hash_64(from_oblock(oblock), mq->hash_bits);
+	struct hlist_head *bucket = mq->table + h;
+	struct entry *e;
+
+	hlist_for_each_entry(e, bucket, hlist)
+		if (e->oblock == oblock) {
+			hlist_del(&e->hlist);
+			hlist_add_head(&e->hlist, bucket);
+			return e;
+		}
+
+	return NULL;
+}
+
+static void hash_remove(struct entry *e)
+{
+	hlist_del(&e->hlist);
+}
+
+/*----------------------------------------------------------------*/
+
+static bool any_free_cblocks(struct mq_policy *mq)
+{
+	return !epool_empty(&mq->cache_pool);
+}
+
+static bool any_clean_cblocks(struct mq_policy *mq)
+{
+	return !queue_empty(&mq->cache_clean);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Now we get to the meat of the policy.  This section deals with deciding
+ * when to to add entries to the pre_cache and cache, and move between
+ * them.
+ */
+
+/*
+ * The queue level is based on the log2 of the hit count.
+ */
+static unsigned queue_level(struct entry *e)
+{
+	return min((unsigned) ilog2(e->hit_count), NR_QUEUE_LEVELS - 1u);
+}
+
+static bool in_cache(struct mq_policy *mq, struct entry *e)
+{
+	return in_pool(&mq->cache_pool, e);
+}
+
+/*
+ * Inserts the entry into the pre_cache or the cache.  Ensures the cache
+ * block is marked as allocated if necc.  Inserts into the hash table.
+ * Sets the tick which records when the entry was last moved about.
+ */
+static void push(struct mq_policy *mq, struct entry *e)
+{
+	e->tick = mq->tick;
+	hash_insert(mq, e);
+
+	if (in_cache(mq, e))
+		queue_push(e->dirty ? &mq->cache_dirty : &mq->cache_clean,
+			   queue_level(e), &e->list);
+	else
+		queue_push(&mq->pre_cache, queue_level(e), &e->list);
+}
+
+/*
+ * Removes an entry from pre_cache or cache.  Removes from the hash table.
+ */
+static void del(struct mq_policy *mq, struct entry *e)
+{
+	queue_remove(&e->list);
+	hash_remove(e);
+}
+
+/*
+ * Like del, except it removes the first entry in the queue (ie. the least
+ * recently used).
+ */
+static struct entry *pop(struct mq_policy *mq, struct queue *q)
+{
+	struct entry *e;
+	struct list_head *h = queue_pop(q);
+
+	if (!h)
+		return NULL;
+
+	e = container_of(h, struct entry, list);
+	hash_remove(e);
+
+	return e;
+}
+
+/*
+ * Has this entry already been updated?
+ */
+static bool updated_this_tick(struct mq_policy *mq, struct entry *e)
+{
+	return mq->tick == e->tick;
+}
+
+/*
+ * The promotion threshold is adjusted every generation.  As are the counts
+ * of the entries.
+ *
+ * At the moment the threshold is taken by averaging the hit counts of some
+ * of the entries in the cache (the first 20 entries across all levels in
+ * ascending order, giving preference to the clean entries at each level).
+ *
+ * We can be much cleverer than this though.  For example, each promotion
+ * could bump up the threshold helping to prevent churn.  Much more to do
+ * here.
+ */
+
+#define MAX_TO_AVERAGE 20
+
+static void check_generation(struct mq_policy *mq)
+{
+	unsigned total = 0, nr = 0, count = 0, level;
+	struct list_head *head;
+	struct entry *e;
+
+	if ((mq->hit_count >= mq->generation_period) && (epool_empty(&mq->cache_pool))) {
+		mq->hit_count = 0;
+		mq->generation++;
+
+		for (level = 0; level < NR_QUEUE_LEVELS && count < MAX_TO_AVERAGE; level++) {
+			head = mq->cache_clean.qs + level;
+			list_for_each_entry(e, head, list) {
+				nr++;
+				total += e->hit_count;
+
+				if (++count >= MAX_TO_AVERAGE)
+					break;
+			}
+
+			head = mq->cache_dirty.qs + level;
+			list_for_each_entry(e, head, list) {
+				nr++;
+				total += e->hit_count;
+
+				if (++count >= MAX_TO_AVERAGE)
+					break;
+			}
+		}
+
+		mq->promote_threshold = nr ? total / nr : 1;
+		if (mq->promote_threshold * nr < total)
+			mq->promote_threshold++;
+	}
+}
+
+/*
+ * Whenever we use an entry we bump up it's hit counter, and push it to the
+ * back to it's current level.
+ */
+static void requeue_and_update_tick(struct mq_policy *mq, struct entry *e)
+{
+	if (updated_this_tick(mq, e))
+		return;
+
+	e->hit_count++;
+	mq->hit_count++;
+	check_generation(mq);
+
+	/* generation adjustment, to stop the counts increasing forever. */
+	/* FIXME: divide? */
+	/* e->hit_count -= min(e->hit_count - 1, mq->generation - e->generation); */
+	e->generation = mq->generation;
+
+	del(mq, e);
+	push(mq, e);
+}
+
+/*
+ * Demote the least recently used entry from the cache to the pre_cache.
+ * Returns the new cache entry to use, and the old origin block it was
+ * mapped to.
+ *
+ * We drop the hit count on the demoted entry back to 1 to stop it bouncing
+ * straight back into the cache if it's subsequently hit.  There are
+ * various options here, and more experimentation would be good:
+ *
+ * - just forget about the demoted entry completely (ie. don't insert it
+     into the pre_cache).
+ * - divide the hit count rather that setting to some hard coded value.
+ * - set the hit count to a hard coded value other than 1, eg, is it better
+ *   if it goes in at level 2?
+ */
+static int demote_cblock(struct mq_policy *mq, dm_oblock_t *oblock)
+{
+	struct entry *demoted = pop(mq, &mq->cache_clean);
+
+	if (!demoted)
+		/*
+		 * We could get a block from mq->cache_dirty, but that
+		 * would add extra latency to the triggering bio as it
+		 * waits for the writeback.  Better to not promote this
+		 * time and hope there's a clean block next time this block
+		 * is hit.
+		 */
+		return -ENOSPC;
+
+	*oblock = demoted->oblock;
+	free_entry(&mq->cache_pool, demoted);
+
+	/*
+	 * We used to put the demoted block into the pre-cache, but I think
+	 * it's simpler to just let it work it's way up from zero again.
+	 * Stops blocks flickering in and out of the cache.
+	 */
+
+	return 0;
+}
+
+/*
+ * We modify the basic promotion_threshold depending on the specific io.
+ *
+ * If the origin block has been discarded then there's no cost to copy it
+ * to the cache.
+ *
+ * We bias towards reads, since they can be demoted at no cost if they
+ * haven't been dirtied.
+ */
+static unsigned adjusted_promote_threshold(struct mq_policy *mq,
+					   bool discarded_oblock, int data_dir)
+{
+	if (data_dir == READ)
+		return mq->promote_threshold + mq->read_promote_adjustment;
+
+	if (discarded_oblock && (any_free_cblocks(mq) || any_clean_cblocks(mq))) {
+		/*
+		 * We don't need to do any copying at all, so give this a
+		 * very low threshold.
+		 */
+		return mq->discard_promote_adjustment;
+	}
+
+	return mq->promote_threshold + mq->write_promote_adjustment;
+}
+
+static bool should_promote(struct mq_policy *mq, struct entry *e,
+			   bool discarded_oblock, int data_dir)
+{
+	return e->hit_count >=
+		adjusted_promote_threshold(mq, discarded_oblock, data_dir);
+}
+
+static int cache_entry_found(struct mq_policy *mq,
+			     struct entry *e,
+			     struct policy_result *result)
+{
+	requeue_and_update_tick(mq, e);
+
+	if (in_cache(mq, e)) {
+		result->op = POLICY_HIT;
+		result->cblock = infer_cblock(&mq->cache_pool, e);
+	}
+
+	return 0;
+}
+
+/*
+ * Moves an entry from the pre_cache to the cache.  The main work is
+ * finding which cache block to use.
+ */
+static int pre_cache_to_cache(struct mq_policy *mq, struct entry *e,
+			      struct policy_result *result)
+{
+	int r;
+	struct entry *new_e;
+
+	/* Ensure there's a free cblock in the cache */
+	if (epool_empty(&mq->cache_pool)) {
+		result->op = POLICY_REPLACE;
+		r = demote_cblock(mq, &result->old_oblock);
+		if (r) {
+			result->op = POLICY_MISS;
+			return 0;
+		}
+	} else
+		result->op = POLICY_NEW;
+
+	new_e = alloc_entry(&mq->cache_pool);
+	BUG_ON(!new_e);
+
+	new_e->oblock = e->oblock;
+	new_e->dirty = false;
+	new_e->hit_count = e->hit_count;
+	new_e->generation = e->generation;
+	new_e->tick = e->tick;
+
+	del(mq, e);
+	free_entry(&mq->pre_cache_pool, e);
+	push(mq, new_e);
+
+	result->cblock = infer_cblock(&mq->cache_pool, new_e);
+
+	return 0;
+}
+
+static int pre_cache_entry_found(struct mq_policy *mq, struct entry *e,
+				 bool can_migrate, bool discarded_oblock,
+				 int data_dir, struct policy_result *result)
+{
+	int r = 0;
+	bool updated = updated_this_tick(mq, e);
+
+	if ((!discarded_oblock && updated) ||
+	    !should_promote(mq, e, discarded_oblock, data_dir)) {
+		requeue_and_update_tick(mq, e);
+		result->op = POLICY_MISS;
+
+	} else if (!can_migrate)
+		r = -EWOULDBLOCK;
+
+	else {
+		requeue_and_update_tick(mq, e);
+		r = pre_cache_to_cache(mq, e, result);
+	}
+
+	return r;
+}
+
+static void insert_in_pre_cache(struct mq_policy *mq,
+				dm_oblock_t oblock)
+{
+	struct entry *e = alloc_entry(&mq->pre_cache_pool);
+
+	if (!e)
+		/*
+		 * There's no spare entry structure, so we grab the least
+		 * used one from the pre_cache.
+		 */
+		e = pop(mq, &mq->pre_cache);
+
+	if (unlikely(!e)) {
+		DMWARN("couldn't pop from pre cache");
+		return;
+	}
+
+	e->dirty = false;
+	e->oblock = oblock;
+	e->hit_count = 1;
+	e->generation = mq->generation;
+	push(mq, e);
+}
+
+static void insert_in_cache(struct mq_policy *mq, dm_oblock_t oblock,
+			    struct policy_result *result)
+{
+	int r;
+	struct entry *e;
+
+	if (epool_empty(&mq->cache_pool)) {
+		result->op = POLICY_REPLACE;
+		r = demote_cblock(mq, &result->old_oblock);
+		if (unlikely(r)) {
+			result->op = POLICY_MISS;
+			insert_in_pre_cache(mq, oblock);
+			return;
+		}
+
+		/*
+		 * This will always succeed, since we've just demoted.
+		 */
+		e = alloc_entry(&mq->cache_pool);
+		BUG_ON(!e);
+
+	} else {
+		e = alloc_entry(&mq->cache_pool);
+		result->op = POLICY_NEW;
+	}
+
+	e->oblock = oblock;
+	e->dirty = false;
+	e->hit_count = 1;
+	e->generation = mq->generation;
+	push(mq, e);
+
+	result->cblock = infer_cblock(&mq->cache_pool, e);
+}
+
+static int no_entry_found(struct mq_policy *mq, dm_oblock_t oblock,
+			  bool can_migrate, bool discarded_oblock,
+			  int data_dir, struct policy_result *result)
+{
+	if (adjusted_promote_threshold(mq, discarded_oblock, data_dir) <= 1) {
+		if (can_migrate)
+			insert_in_cache(mq, oblock, result);
+		else
+			return -EWOULDBLOCK;
+	} else {
+		insert_in_pre_cache(mq, oblock);
+		result->op = POLICY_MISS;
+	}
+
+	return 0;
+}
+
+/*
+ * Looks the oblock up in the hash table, then decides whether to put in
+ * pre_cache, or cache etc.
+ */
+static int map(struct mq_policy *mq, dm_oblock_t oblock,
+	       bool can_migrate, bool discarded_oblock,
+	       int data_dir, struct policy_result *result)
+{
+	int r = 0;
+	struct entry *e = hash_lookup(mq, oblock);
+
+	if (e && in_cache(mq, e))
+		r = cache_entry_found(mq, e, result);
+
+	else if (iot_pattern(&mq->tracker) == PATTERN_SEQUENTIAL)
+		result->op = POLICY_MISS;
+
+	else if (e)
+		r = pre_cache_entry_found(mq, e, can_migrate, discarded_oblock,
+					  data_dir, result);
+
+	else
+		r = no_entry_found(mq, oblock, can_migrate, discarded_oblock,
+				   data_dir, result);
+
+	if (r == -EWOULDBLOCK)
+		result->op = POLICY_MISS;
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Public interface, via the policy struct.  See dm-cache-policy.h for a
+ * description of these.
+ */
+
+static struct mq_policy *to_mq_policy(struct dm_cache_policy *p)
+{
+	return container_of(p, struct mq_policy, policy);
+}
+
+static void mq_destroy(struct dm_cache_policy *p)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+
+	vfree(mq->table);
+	epool_exit(&mq->cache_pool);
+	epool_exit(&mq->pre_cache_pool);
+	kfree(mq);
+}
+
+static void copy_tick(struct mq_policy *mq)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&mq->tick_lock, flags);
+	mq->tick = mq->tick_protected;
+	spin_unlock_irqrestore(&mq->tick_lock, flags);
+}
+
+static int mq_map(struct dm_cache_policy *p, dm_oblock_t oblock,
+		  bool can_block, bool can_migrate, bool discarded_oblock,
+		  struct bio *bio, struct policy_result *result)
+{
+	int r;
+	struct mq_policy *mq = to_mq_policy(p);
+
+	result->op = POLICY_MISS;
+
+	if (can_block)
+		mutex_lock(&mq->lock);
+	else if (!mutex_trylock(&mq->lock))
+		return -EWOULDBLOCK;
+
+	copy_tick(mq);
+
+	iot_examine_bio(&mq->tracker, bio);
+	r = map(mq, oblock, can_migrate, discarded_oblock,
+		bio_data_dir(bio), result);
+
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static int mq_lookup(struct dm_cache_policy *p, dm_oblock_t oblock, dm_cblock_t *cblock)
+{
+	int r;
+	struct mq_policy *mq = to_mq_policy(p);
+	struct entry *e;
+
+	if (!mutex_trylock(&mq->lock))
+		return -EWOULDBLOCK;
+
+	e = hash_lookup(mq, oblock);
+	if (e && in_cache(mq, e)) {
+		*cblock = infer_cblock(&mq->cache_pool, e);
+		r = 0;
+	} else
+		r = -ENOENT;
+
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static void __mq_set_clear_dirty(struct mq_policy *mq, dm_oblock_t oblock, bool set)
+{
+	struct entry *e;
+
+	e = hash_lookup(mq, oblock);
+	BUG_ON(!e || !in_cache(mq, e));
+
+	del(mq, e);
+	e->dirty = set;
+	push(mq, e);
+}
+
+static void mq_set_dirty(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	__mq_set_clear_dirty(mq, oblock, true);
+	mutex_unlock(&mq->lock);
+}
+
+static void mq_clear_dirty(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	__mq_set_clear_dirty(mq, oblock, false);
+	mutex_unlock(&mq->lock);
+}
+
+static int mq_load_mapping(struct dm_cache_policy *p,
+			   dm_oblock_t oblock, dm_cblock_t cblock,
+			   uint32_t hint, bool hint_valid)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+	struct entry *e;
+
+	e = alloc_particular_entry(&mq->cache_pool, cblock);
+	e->oblock = oblock;
+	e->dirty = false;	/* this gets corrected in a minute */
+	e->hit_count = hint_valid ? hint : 1;
+	e->generation = mq->generation;
+	push(mq, e);
+
+	return 0;
+}
+
+static int mq_save_hints(struct mq_policy *mq, struct queue *q,
+			 policy_walk_fn fn, void *context)
+{
+	int r;
+	unsigned level;
+	struct entry *e;
+
+	for (level = 0; level < NR_QUEUE_LEVELS; level++)
+		list_for_each_entry(e, q->qs + level, list) {
+			r = fn(context, infer_cblock(&mq->cache_pool, e),
+			       e->oblock, e->hit_count);
+			if (r)
+				return r;
+		}
+
+	return 0;
+}
+
+static int mq_walk_mappings(struct dm_cache_policy *p, policy_walk_fn fn,
+			    void *context)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+	int r = 0;
+
+	mutex_lock(&mq->lock);
+
+	r = mq_save_hints(mq, &mq->cache_clean, fn, context);
+	if (!r)
+		r = mq_save_hints(mq, &mq->cache_dirty, fn, context);
+
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static void __remove_mapping(struct mq_policy *mq, dm_oblock_t oblock)
+{
+	struct entry *e;
+
+	e = hash_lookup(mq, oblock);
+	BUG_ON(!e || !in_cache(mq, e));
+
+	del(mq, e);
+	free_entry(&mq->cache_pool, e);
+}
+
+static void mq_remove_mapping(struct dm_cache_policy *p, dm_oblock_t oblock)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	__remove_mapping(mq, oblock);
+	mutex_unlock(&mq->lock);
+}
+
+static int __remove_cblock(struct mq_policy *mq, dm_cblock_t cblock)
+{
+	struct entry *e = epool_find(&mq->cache_pool, cblock);
+
+	if (!e)
+		return -ENODATA;
+
+	del(mq, e);
+	free_entry(&mq->cache_pool, e);
+
+	return 0;
+}
+
+static int mq_remove_cblock(struct dm_cache_policy *p, dm_cblock_t cblock)
+{
+	int r;
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	r = __remove_cblock(mq, cblock);
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static int __mq_writeback_work(struct mq_policy *mq, dm_oblock_t *oblock,
+			      dm_cblock_t *cblock)
+{
+	struct entry *e = pop(mq, &mq->cache_dirty);
+
+	if (!e)
+		return -ENODATA;
+
+	*oblock = e->oblock;
+	*cblock = infer_cblock(&mq->cache_pool, e);
+	e->dirty = false;
+	push(mq, e);
+
+	return 0;
+}
+
+static int mq_writeback_work(struct dm_cache_policy *p, dm_oblock_t *oblock,
+			     dm_cblock_t *cblock)
+{
+	int r;
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	r = __mq_writeback_work(mq, oblock, cblock);
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static void __force_mapping(struct mq_policy *mq,
+			    dm_oblock_t current_oblock, dm_oblock_t new_oblock)
+{
+	struct entry *e = hash_lookup(mq, current_oblock);
+
+	if (e && in_cache(mq, e)) {
+		del(mq, e);
+		e->oblock = new_oblock;
+		e->dirty = true;
+		push(mq, e);
+	}
+}
+
+static void mq_force_mapping(struct dm_cache_policy *p,
+			     dm_oblock_t current_oblock, dm_oblock_t new_oblock)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	__force_mapping(mq, current_oblock, new_oblock);
+	mutex_unlock(&mq->lock);
+}
+
+static dm_cblock_t mq_residency(struct dm_cache_policy *p)
+{
+	dm_cblock_t r;
+	struct mq_policy *mq = to_mq_policy(p);
+
+	mutex_lock(&mq->lock);
+	r = to_cblock(mq->cache_pool.nr_allocated);
+	mutex_unlock(&mq->lock);
+
+	return r;
+}
+
+static void mq_tick(struct dm_cache_policy *p)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+	unsigned long flags;
+
+	spin_lock_irqsave(&mq->tick_lock, flags);
+	mq->tick_protected++;
+	spin_unlock_irqrestore(&mq->tick_lock, flags);
+}
+
+static int mq_set_config_value(struct dm_cache_policy *p,
+			       const char *key, const char *value)
+{
+	struct mq_policy *mq = to_mq_policy(p);
+	unsigned long tmp;
+
+	if (kstrtoul(value, 10, &tmp))
+		return -EINVAL;
+
+	if (!strcasecmp(key, "random_threshold")) {
+		mq->tracker.thresholds[PATTERN_RANDOM] = tmp;
+
+	} else if (!strcasecmp(key, "sequential_threshold")) {
+		mq->tracker.thresholds[PATTERN_SEQUENTIAL] = tmp;
+
+	} else if (!strcasecmp(key, "discard_promote_adjustment"))
+		mq->discard_promote_adjustment = tmp;
+
+	else if (!strcasecmp(key, "read_promote_adjustment"))
+		mq->read_promote_adjustment = tmp;
+
+	else if (!strcasecmp(key, "write_promote_adjustment"))
+		mq->write_promote_adjustment = tmp;
+
+	else
+		return -EINVAL;
+
+	return 0;
+}
+
+static int mq_emit_config_values(struct dm_cache_policy *p, char *result, unsigned maxlen)
+{
+	ssize_t sz = 0;
+	struct mq_policy *mq = to_mq_policy(p);
+
+	DMEMIT("10 random_threshold %u "
+	       "sequential_threshold %u "
+	       "discard_promote_adjustment %u "
+	       "read_promote_adjustment %u "
+	       "write_promote_adjustment %u",
+	       mq->tracker.thresholds[PATTERN_RANDOM],
+	       mq->tracker.thresholds[PATTERN_SEQUENTIAL],
+	       mq->discard_promote_adjustment,
+	       mq->read_promote_adjustment,
+	       mq->write_promote_adjustment);
+
+	return 0;
+}
+
+/* Init the policy plugin interface function pointers. */
+static void init_policy_functions(struct mq_policy *mq)
+{
+	mq->policy.destroy = mq_destroy;
+	mq->policy.map = mq_map;
+	mq->policy.lookup = mq_lookup;
+	mq->policy.set_dirty = mq_set_dirty;
+	mq->policy.clear_dirty = mq_clear_dirty;
+	mq->policy.load_mapping = mq_load_mapping;
+	mq->policy.walk_mappings = mq_walk_mappings;
+	mq->policy.remove_mapping = mq_remove_mapping;
+	mq->policy.remove_cblock = mq_remove_cblock;
+	mq->policy.writeback_work = mq_writeback_work;
+	mq->policy.force_mapping = mq_force_mapping;
+	mq->policy.residency = mq_residency;
+	mq->policy.tick = mq_tick;
+	mq->policy.emit_config_values = mq_emit_config_values;
+	mq->policy.set_config_value = mq_set_config_value;
+}
+
+static struct dm_cache_policy *mq_create(dm_cblock_t cache_size,
+					 sector_t origin_size,
+					 sector_t cache_block_size)
+{
+	struct mq_policy *mq = kzalloc(sizeof(*mq), GFP_KERNEL);
+
+	if (!mq)
+		return NULL;
+
+	init_policy_functions(mq);
+	iot_init(&mq->tracker, SEQUENTIAL_THRESHOLD_DEFAULT, RANDOM_THRESHOLD_DEFAULT);
+	mq->cache_size = cache_size;
+
+	if (epool_init(&mq->pre_cache_pool, from_cblock(cache_size))) {
+		DMERR("couldn't initialize pool of pre-cache entries");
+		goto bad_pre_cache_init;
+	}
+
+	if (epool_init(&mq->cache_pool, from_cblock(cache_size))) {
+		DMERR("couldn't initialize pool of cache entries");
+		goto bad_cache_init;
+	}
+
+	mq->tick_protected = 0;
+	mq->tick = 0;
+	mq->hit_count = 0;
+	mq->generation = 0;
+	mq->promote_threshold = 0;
+	mq->discard_promote_adjustment = DEFAULT_DISCARD_PROMOTE_ADJUSTMENT;
+	mq->read_promote_adjustment = DEFAULT_READ_PROMOTE_ADJUSTMENT;
+	mq->write_promote_adjustment = DEFAULT_WRITE_PROMOTE_ADJUSTMENT;
+	mutex_init(&mq->lock);
+	spin_lock_init(&mq->tick_lock);
+
+	queue_init(&mq->pre_cache);
+	queue_init(&mq->cache_clean);
+	queue_init(&mq->cache_dirty);
+
+	mq->generation_period = max((unsigned) from_cblock(cache_size), 1024U);
+
+	mq->nr_buckets = next_power(from_cblock(cache_size) / 2, 16);
+	mq->hash_bits = ffs(mq->nr_buckets) - 1;
+	mq->table = vzalloc(sizeof(*mq->table) * mq->nr_buckets);
+	if (!mq->table)
+		goto bad_alloc_table;
+
+	return &mq->policy;
+
+bad_alloc_table:
+	epool_exit(&mq->cache_pool);
+bad_cache_init:
+	epool_exit(&mq->pre_cache_pool);
+bad_pre_cache_init:
+	kfree(mq);
+
+	return NULL;
+}
+
+/*----------------------------------------------------------------*/
+
+static struct dm_cache_policy_type mq_policy_type = {
+	.name = "mq",
+	.version = {1, 2, 0},
+	.hint_size = 4,
+	.owner = THIS_MODULE,
+	.create = mq_create
+};
+
+static struct dm_cache_policy_type default_policy_type = {
+	.name = "default",
+	.version = {1, 2, 0},
+	.hint_size = 4,
+	.owner = THIS_MODULE,
+	.create = mq_create,
+	.real = &mq_policy_type
+};
+
+static int __init mq_init(void)
+{
+	int r;
+
+	mq_entry_cache = kmem_cache_create("dm_mq_policy_cache_entry",
+					   sizeof(struct entry),
+					   __alignof__(struct entry),
+					   0, NULL);
+	if (!mq_entry_cache)
+		goto bad;
+
+	r = dm_cache_policy_register(&mq_policy_type);
+	if (r) {
+		DMERR("register failed %d", r);
+		goto bad_register_mq;
+	}
+
+	r = dm_cache_policy_register(&default_policy_type);
+	if (!r) {
+		DMINFO("version %u.%u.%u loaded",
+		       mq_policy_type.version[0],
+		       mq_policy_type.version[1],
+		       mq_policy_type.version[2]);
+		return 0;
+	}
+
+	DMERR("register failed (as default) %d", r);
+
+	dm_cache_policy_unregister(&mq_policy_type);
+bad_register_mq:
+	kmem_cache_destroy(mq_entry_cache);
+bad:
+	return -ENOMEM;
+}
+
+static void __exit mq_exit(void)
+{
+	dm_cache_policy_unregister(&mq_policy_type);
+	dm_cache_policy_unregister(&default_policy_type);
+
+	kmem_cache_destroy(mq_entry_cache);
+}
+
+module_init(mq_init);
+module_exit(mq_exit);
+
+MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("mq cache policy");
+
+MODULE_ALIAS("dm-cache-default");
diff --git a/drivers/md/dm-cache-policy.c b/drivers/md/dm-cache-policy.c
new file mode 100644
index 00000000000..c1a3cee99b4
--- /dev/null
+++ b/drivers/md/dm-cache-policy.c
@@ -0,0 +1,173 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-cache-policy-internal.h"
+#include "dm.h"
+
+#include <linux/module.h>
+#include <linux/slab.h>
+
+/*----------------------------------------------------------------*/
+
+#define DM_MSG_PREFIX "cache-policy"
+
+static DEFINE_SPINLOCK(register_lock);
+static LIST_HEAD(register_list);
+
+static struct dm_cache_policy_type *__find_policy(const char *name)
+{
+	struct dm_cache_policy_type *t;
+
+	list_for_each_entry(t, &register_list, list)
+		if (!strcmp(t->name, name))
+			return t;
+
+	return NULL;
+}
+
+static struct dm_cache_policy_type *__get_policy_once(const char *name)
+{
+	struct dm_cache_policy_type *t = __find_policy(name);
+
+	if (t && !try_module_get(t->owner)) {
+		DMWARN("couldn't get module %s", name);
+		t = ERR_PTR(-EINVAL);
+	}
+
+	return t;
+}
+
+static struct dm_cache_policy_type *get_policy_once(const char *name)
+{
+	struct dm_cache_policy_type *t;
+
+	spin_lock(&register_lock);
+	t = __get_policy_once(name);
+	spin_unlock(&register_lock);
+
+	return t;
+}
+
+static struct dm_cache_policy_type *get_policy(const char *name)
+{
+	struct dm_cache_policy_type *t;
+
+	t = get_policy_once(name);
+	if (IS_ERR(t))
+		return NULL;
+
+	if (t)
+		return t;
+
+	request_module("dm-cache-%s", name);
+
+	t = get_policy_once(name);
+	if (IS_ERR(t))
+		return NULL;
+
+	return t;
+}
+
+static void put_policy(struct dm_cache_policy_type *t)
+{
+	module_put(t->owner);
+}
+
+int dm_cache_policy_register(struct dm_cache_policy_type *type)
+{
+	int r;
+
+	/* One size fits all for now */
+	if (type->hint_size != 0 && type->hint_size != 4) {
+		DMWARN("hint size must be 0 or 4 but %llu supplied.", (unsigned long long) type->hint_size);
+		return -EINVAL;
+	}
+
+	spin_lock(&register_lock);
+	if (__find_policy(type->name)) {
+		DMWARN("attempt to register policy under duplicate name %s", type->name);
+		r = -EINVAL;
+	} else {
+		list_add(&type->list, &register_list);
+		r = 0;
+	}
+	spin_unlock(&register_lock);
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_register);
+
+void dm_cache_policy_unregister(struct dm_cache_policy_type *type)
+{
+	spin_lock(&register_lock);
+	list_del_init(&type->list);
+	spin_unlock(&register_lock);
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_unregister);
+
+struct dm_cache_policy *dm_cache_policy_create(const char *name,
+					       dm_cblock_t cache_size,
+					       sector_t origin_size,
+					       sector_t cache_block_size)
+{
+	struct dm_cache_policy *p = NULL;
+	struct dm_cache_policy_type *type;
+
+	type = get_policy(name);
+	if (!type) {
+		DMWARN("unknown policy type");
+		return ERR_PTR(-EINVAL);
+	}
+
+	p = type->create(cache_size, origin_size, cache_block_size);
+	if (!p) {
+		put_policy(type);
+		return ERR_PTR(-ENOMEM);
+	}
+	p->private = type;
+
+	return p;
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_create);
+
+void dm_cache_policy_destroy(struct dm_cache_policy *p)
+{
+	struct dm_cache_policy_type *t = p->private;
+
+	p->destroy(p);
+	put_policy(t);
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_destroy);
+
+const char *dm_cache_policy_get_name(struct dm_cache_policy *p)
+{
+	struct dm_cache_policy_type *t = p->private;
+
+	/* if t->real is set then an alias was used (e.g. "default") */
+	if (t->real)
+		return t->real->name;
+
+	return t->name;
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_get_name);
+
+const unsigned *dm_cache_policy_get_version(struct dm_cache_policy *p)
+{
+	struct dm_cache_policy_type *t = p->private;
+
+	return t->version;
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_get_version);
+
+size_t dm_cache_policy_get_hint_size(struct dm_cache_policy *p)
+{
+	struct dm_cache_policy_type *t = p->private;
+
+	return t->hint_size;
+}
+EXPORT_SYMBOL_GPL(dm_cache_policy_get_hint_size);
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/dm-cache-policy.h b/drivers/md/dm-cache-policy.h
new file mode 100644
index 00000000000..f50fe360c54
--- /dev/null
+++ b/drivers/md/dm-cache-policy.h
@@ -0,0 +1,249 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_CACHE_POLICY_H
+#define DM_CACHE_POLICY_H
+
+#include "dm-cache-block-types.h"
+
+#include <linux/device-mapper.h>
+
+/*----------------------------------------------------------------*/
+
+/* FIXME: make it clear which methods are optional.  Get debug policy to
+ * double check this at start.
+ */
+
+/*
+ * The cache policy makes the important decisions about which blocks get to
+ * live on the faster cache device.
+ *
+ * When the core target has to remap a bio it calls the 'map' method of the
+ * policy.  This returns an instruction telling the core target what to do.
+ *
+ * POLICY_HIT:
+ *   That block is in the cache.  Remap to the cache and carry on.
+ *
+ * POLICY_MISS:
+ *   This block is on the origin device.  Remap and carry on.
+ *
+ * POLICY_NEW:
+ *   This block is currently on the origin device, but the policy wants to
+ *   move it.  The core should:
+ *
+ *   - hold any further io to this origin block
+ *   - copy the origin to the given cache block
+ *   - release all the held blocks
+ *   - remap the original block to the cache
+ *
+ * POLICY_REPLACE:
+ *   This block is currently on the origin device.  The policy wants to
+ *   move it to the cache, with the added complication that the destination
+ *   cache block needs a writeback first.  The core should:
+ *
+ *   - hold any further io to this origin block
+ *   - hold any further io to the origin block that's being written back
+ *   - writeback
+ *   - copy new block to cache
+ *   - release held blocks
+ *   - remap bio to cache and reissue.
+ *
+ * Should the core run into trouble while processing a POLICY_NEW or
+ * POLICY_REPLACE instruction it will roll back the policies mapping using
+ * remove_mapping() or force_mapping().  These methods must not fail.  This
+ * approach avoids having transactional semantics in the policy (ie, the
+ * core informing the policy when a migration is complete), and hence makes
+ * it easier to write new policies.
+ *
+ * In general policy methods should never block, except in the case of the
+ * map function when can_migrate is set.  So be careful to implement using
+ * bounded, preallocated memory.
+ */
+enum policy_operation {
+	POLICY_HIT,
+	POLICY_MISS,
+	POLICY_NEW,
+	POLICY_REPLACE
+};
+
+/*
+ * This is the instruction passed back to the core target.
+ */
+struct policy_result {
+	enum policy_operation op;
+	dm_oblock_t old_oblock;	/* POLICY_REPLACE */
+	dm_cblock_t cblock;	/* POLICY_HIT, POLICY_NEW, POLICY_REPLACE */
+};
+
+typedef int (*policy_walk_fn)(void *context, dm_cblock_t cblock,
+			      dm_oblock_t oblock, uint32_t hint);
+
+/*
+ * The cache policy object.  Just a bunch of methods.  It is envisaged that
+ * this structure will be embedded in a bigger, policy specific structure
+ * (ie. use container_of()).
+ */
+struct dm_cache_policy {
+
+	/*
+	 * FIXME: make it clear which methods are optional, and which may
+	 * block.
+	 */
+
+	/*
+	 * Destroys this object.
+	 */
+	void (*destroy)(struct dm_cache_policy *p);
+
+	/*
+	 * See large comment above.
+	 *
+	 * oblock      - the origin block we're interested in.
+	 *
+	 * can_block - indicates whether the current thread is allowed to
+	 *             block.  -EWOULDBLOCK returned if it can't and would.
+	 *
+	 * can_migrate - gives permission for POLICY_NEW or POLICY_REPLACE
+	 *               instructions.  If denied and the policy would have
+	 *               returned one of these instructions it should
+	 *               return -EWOULDBLOCK.
+	 *
+	 * discarded_oblock - indicates whether the whole origin block is
+	 *               in a discarded state (FIXME: better to tell the
+	 *               policy about this sooner, so it can recycle that
+	 *               cache block if it wants.)
+	 * bio         - the bio that triggered this call.
+	 * result      - gets filled in with the instruction.
+	 *
+	 * May only return 0, or -EWOULDBLOCK (if !can_migrate)
+	 */
+	int (*map)(struct dm_cache_policy *p, dm_oblock_t oblock,
+		   bool can_block, bool can_migrate, bool discarded_oblock,
+		   struct bio *bio, struct policy_result *result);
+
+	/*
+	 * Sometimes we want to see if a block is in the cache, without
+	 * triggering any update of stats.  (ie. it's not a real hit).
+	 *
+	 * Must not block.
+	 *
+	 * Returns 0 if in cache, -ENOENT if not, < 0 for other errors
+	 * (-EWOULDBLOCK would be typical).
+	 */
+	int (*lookup)(struct dm_cache_policy *p, dm_oblock_t oblock, dm_cblock_t *cblock);
+
+	void (*set_dirty)(struct dm_cache_policy *p, dm_oblock_t oblock);
+	void (*clear_dirty)(struct dm_cache_policy *p, dm_oblock_t oblock);
+
+	/*
+	 * Called when a cache target is first created.  Used to load a
+	 * mapping from the metadata device into the policy.
+	 */
+	int (*load_mapping)(struct dm_cache_policy *p, dm_oblock_t oblock,
+			    dm_cblock_t cblock, uint32_t hint, bool hint_valid);
+
+	int (*walk_mappings)(struct dm_cache_policy *p, policy_walk_fn fn,
+			     void *context);
+
+	/*
+	 * Override functions used on the error paths of the core target.
+	 * They must succeed.
+	 */
+	void (*remove_mapping)(struct dm_cache_policy *p, dm_oblock_t oblock);
+	void (*force_mapping)(struct dm_cache_policy *p, dm_oblock_t current_oblock,
+			      dm_oblock_t new_oblock);
+
+	/*
+	 * This is called via the invalidate_cblocks message.  It is
+	 * possible the particular cblock has already been removed due to a
+	 * write io in passthrough mode.  In which case this should return
+	 * -ENODATA.
+	 */
+	int (*remove_cblock)(struct dm_cache_policy *p, dm_cblock_t cblock);
+
+	/*
+	 * Provide a dirty block to be written back by the core target.
+	 *
+	 * Returns:
+	 *
+	 * 0 and @cblock,@oblock: block to write back provided
+	 *
+	 * -ENODATA: no dirty blocks available
+	 */
+	int (*writeback_work)(struct dm_cache_policy *p, dm_oblock_t *oblock, dm_cblock_t *cblock);
+
+	/*
+	 * How full is the cache?
+	 */
+	dm_cblock_t (*residency)(struct dm_cache_policy *p);
+
+	/*
+	 * Because of where we sit in the block layer, we can be asked to
+	 * map a lot of little bios that are all in the same block (no
+	 * queue merging has occurred).  To stop the policy being fooled by
+	 * these the core target sends regular tick() calls to the policy.
+	 * The policy should only count an entry as hit once per tick.
+	 */
+	void (*tick)(struct dm_cache_policy *p);
+
+	/*
+	 * Configuration.
+	 */
+	int (*emit_config_values)(struct dm_cache_policy *p,
+				  char *result, unsigned maxlen);
+	int (*set_config_value)(struct dm_cache_policy *p,
+				const char *key, const char *value);
+
+	/*
+	 * Book keeping ptr for the policy register, not for general use.
+	 */
+	void *private;
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * We maintain a little register of the different policy types.
+ */
+#define CACHE_POLICY_NAME_SIZE 16
+#define CACHE_POLICY_VERSION_SIZE 3
+
+struct dm_cache_policy_type {
+	/* For use by the register code only. */
+	struct list_head list;
+
+	/*
+	 * Policy writers should fill in these fields.  The name field is
+	 * what gets passed on the target line to select your policy.
+	 */
+	char name[CACHE_POLICY_NAME_SIZE];
+	unsigned version[CACHE_POLICY_VERSION_SIZE];
+
+	/*
+	 * For use by an alias dm_cache_policy_type to point to the
+	 * real dm_cache_policy_type.
+	 */
+	struct dm_cache_policy_type *real;
+
+	/*
+	 * Policies may store a hint for each each cache block.
+	 * Currently the size of this hint must be 0 or 4 bytes but we
+	 * expect to relax this in future.
+	 */
+	size_t hint_size;
+
+	struct module *owner;
+	struct dm_cache_policy *(*create)(dm_cblock_t cache_size,
+					  sector_t origin_size,
+					  sector_t block_size);
+};
+
+int dm_cache_policy_register(struct dm_cache_policy_type *type);
+void dm_cache_policy_unregister(struct dm_cache_policy_type *type);
+
+/*----------------------------------------------------------------*/
+
+#endif	/* DM_CACHE_POLICY_H */
diff --git a/drivers/md/dm-cache-target.c b/drivers/md/dm-cache-target.c
new file mode 100644
index 00000000000..2c63326638b
--- /dev/null
+++ b/drivers/md/dm-cache-target.c
@@ -0,0 +1,3121 @@
+/*
+ * Copyright (C) 2012 Red Hat. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm.h"
+#include "dm-bio-prison.h"
+#include "dm-bio-record.h"
+#include "dm-cache-metadata.h"
+
+#include <linux/dm-io.h>
+#include <linux/dm-kcopyd.h>
+#include <linux/init.h>
+#include <linux/mempool.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#define DM_MSG_PREFIX "cache"
+
+DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(cache_copy_throttle,
+	"A percentage of time allocated for copying to and/or from cache");
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Glossary:
+ *
+ * oblock: index of an origin block
+ * cblock: index of a cache block
+ * promotion: movement of a block from origin to cache
+ * demotion: movement of a block from cache to origin
+ * migration: movement of a block between the origin and cache device,
+ *	      either direction
+ */
+
+/*----------------------------------------------------------------*/
+
+static size_t bitset_size_in_bytes(unsigned nr_entries)
+{
+	return sizeof(unsigned long) * dm_div_up(nr_entries, BITS_PER_LONG);
+}
+
+static unsigned long *alloc_bitset(unsigned nr_entries)
+{
+	size_t s = bitset_size_in_bytes(nr_entries);
+	return vzalloc(s);
+}
+
+static void clear_bitset(void *bitset, unsigned nr_entries)
+{
+	size_t s = bitset_size_in_bytes(nr_entries);
+	memset(bitset, 0, s);
+}
+
+static void free_bitset(unsigned long *bits)
+{
+	vfree(bits);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * There are a couple of places where we let a bio run, but want to do some
+ * work before calling its endio function.  We do this by temporarily
+ * changing the endio fn.
+ */
+struct dm_hook_info {
+	bio_end_io_t *bi_end_io;
+	void *bi_private;
+};
+
+static void dm_hook_bio(struct dm_hook_info *h, struct bio *bio,
+			bio_end_io_t *bi_end_io, void *bi_private)
+{
+	h->bi_end_io = bio->bi_end_io;
+	h->bi_private = bio->bi_private;
+
+	bio->bi_end_io = bi_end_io;
+	bio->bi_private = bi_private;
+}
+
+static void dm_unhook_bio(struct dm_hook_info *h, struct bio *bio)
+{
+	bio->bi_end_io = h->bi_end_io;
+	bio->bi_private = h->bi_private;
+
+	/*
+	 * Must bump bi_remaining to allow bio to complete with
+	 * restored bi_end_io.
+	 */
+	atomic_inc(&bio->bi_remaining);
+}
+
+/*----------------------------------------------------------------*/
+
+#define PRISON_CELLS 1024
+#define MIGRATION_POOL_SIZE 128
+#define COMMIT_PERIOD HZ
+#define MIGRATION_COUNT_WINDOW 10
+
+/*
+ * The block size of the device holding cache data must be
+ * between 32KB and 1GB.
+ */
+#define DATA_DEV_BLOCK_SIZE_MIN_SECTORS (32 * 1024 >> SECTOR_SHIFT)
+#define DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT)
+
+/*
+ * FIXME: the cache is read/write for the time being.
+ */
+enum cache_metadata_mode {
+	CM_WRITE,		/* metadata may be changed */
+	CM_READ_ONLY,		/* metadata may not be changed */
+};
+
+enum cache_io_mode {
+	/*
+	 * Data is written to cached blocks only.  These blocks are marked
+	 * dirty.  If you lose the cache device you will lose data.
+	 * Potential performance increase for both reads and writes.
+	 */
+	CM_IO_WRITEBACK,
+
+	/*
+	 * Data is written to both cache and origin.  Blocks are never
+	 * dirty.  Potential performance benfit for reads only.
+	 */
+	CM_IO_WRITETHROUGH,
+
+	/*
+	 * A degraded mode useful for various cache coherency situations
+	 * (eg, rolling back snapshots).  Reads and writes always go to the
+	 * origin.  If a write goes to a cached oblock, then the cache
+	 * block is invalidated.
+	 */
+	CM_IO_PASSTHROUGH
+};
+
+struct cache_features {
+	enum cache_metadata_mode mode;
+	enum cache_io_mode io_mode;
+};
+
+struct cache_stats {
+	atomic_t read_hit;
+	atomic_t read_miss;
+	atomic_t write_hit;
+	atomic_t write_miss;
+	atomic_t demotion;
+	atomic_t promotion;
+	atomic_t copies_avoided;
+	atomic_t cache_cell_clash;
+	atomic_t commit_count;
+	atomic_t discard_count;
+};
+
+/*
+ * Defines a range of cblocks, begin to (end - 1) are in the range.  end is
+ * the one-past-the-end value.
+ */
+struct cblock_range {
+	dm_cblock_t begin;
+	dm_cblock_t end;
+};
+
+struct invalidation_request {
+	struct list_head list;
+	struct cblock_range *cblocks;
+
+	atomic_t complete;
+	int err;
+
+	wait_queue_head_t result_wait;
+};
+
+struct cache {
+	struct dm_target *ti;
+	struct dm_target_callbacks callbacks;
+
+	struct dm_cache_metadata *cmd;
+
+	/*
+	 * Metadata is written to this device.
+	 */
+	struct dm_dev *metadata_dev;
+
+	/*
+	 * The slower of the two data devices.  Typically a spindle.
+	 */
+	struct dm_dev *origin_dev;
+
+	/*
+	 * The faster of the two data devices.  Typically an SSD.
+	 */
+	struct dm_dev *cache_dev;
+
+	/*
+	 * Size of the origin device in _complete_ blocks and native sectors.
+	 */
+	dm_oblock_t origin_blocks;
+	sector_t origin_sectors;
+
+	/*
+	 * Size of the cache device in blocks.
+	 */
+	dm_cblock_t cache_size;
+
+	/*
+	 * Fields for converting from sectors to blocks.
+	 */
+	uint32_t sectors_per_block;
+	int sectors_per_block_shift;
+
+	spinlock_t lock;
+	struct bio_list deferred_bios;
+	struct bio_list deferred_flush_bios;
+	struct bio_list deferred_writethrough_bios;
+	struct list_head quiesced_migrations;
+	struct list_head completed_migrations;
+	struct list_head need_commit_migrations;
+	sector_t migration_threshold;
+	wait_queue_head_t migration_wait;
+	atomic_t nr_migrations;
+
+	wait_queue_head_t quiescing_wait;
+	atomic_t quiescing;
+	atomic_t quiescing_ack;
+
+	/*
+	 * cache_size entries, dirty if set
+	 */
+	atomic_t nr_dirty;
+	unsigned long *dirty_bitset;
+
+	/*
+	 * origin_blocks entries, discarded if set.
+	 */
+	dm_oblock_t discard_nr_blocks;
+	unsigned long *discard_bitset;
+
+	/*
+	 * Rather than reconstructing the table line for the status we just
+	 * save it and regurgitate.
+	 */
+	unsigned nr_ctr_args;
+	const char **ctr_args;
+
+	struct dm_kcopyd_client *copier;
+	struct workqueue_struct *wq;
+	struct work_struct worker;
+
+	struct delayed_work waker;
+	unsigned long last_commit_jiffies;
+
+	struct dm_bio_prison *prison;
+	struct dm_deferred_set *all_io_ds;
+
+	mempool_t *migration_pool;
+	struct dm_cache_migration *next_migration;
+
+	struct dm_cache_policy *policy;
+	unsigned policy_nr_args;
+
+	bool need_tick_bio:1;
+	bool sized:1;
+	bool invalidate:1;
+	bool commit_requested:1;
+	bool loaded_mappings:1;
+	bool loaded_discards:1;
+
+	/*
+	 * Cache features such as write-through.
+	 */
+	struct cache_features features;
+
+	struct cache_stats stats;
+
+	/*
+	 * Invalidation fields.
+	 */
+	spinlock_t invalidation_lock;
+	struct list_head invalidation_requests;
+};
+
+struct per_bio_data {
+	bool tick:1;
+	unsigned req_nr:2;
+	struct dm_deferred_entry *all_io_entry;
+	struct dm_hook_info hook_info;
+
+	/*
+	 * writethrough fields.  These MUST remain at the end of this
+	 * structure and the 'cache' member must be the first as it
+	 * is used to determine the offset of the writethrough fields.
+	 */
+	struct cache *cache;
+	dm_cblock_t cblock;
+	struct dm_bio_details bio_details;
+};
+
+struct dm_cache_migration {
+	struct list_head list;
+	struct cache *cache;
+
+	unsigned long start_jiffies;
+	dm_oblock_t old_oblock;
+	dm_oblock_t new_oblock;
+	dm_cblock_t cblock;
+
+	bool err:1;
+	bool writeback:1;
+	bool demote:1;
+	bool promote:1;
+	bool requeue_holder:1;
+	bool invalidate:1;
+
+	struct dm_bio_prison_cell *old_ocell;
+	struct dm_bio_prison_cell *new_ocell;
+};
+
+/*
+ * Processing a bio in the worker thread may require these memory
+ * allocations.  We prealloc to avoid deadlocks (the same worker thread
+ * frees them back to the mempool).
+ */
+struct prealloc {
+	struct dm_cache_migration *mg;
+	struct dm_bio_prison_cell *cell1;
+	struct dm_bio_prison_cell *cell2;
+};
+
+static void wake_worker(struct cache *cache)
+{
+	queue_work(cache->wq, &cache->worker);
+}
+
+/*----------------------------------------------------------------*/
+
+static struct dm_bio_prison_cell *alloc_prison_cell(struct cache *cache)
+{
+	/* FIXME: change to use a local slab. */
+	return dm_bio_prison_alloc_cell(cache->prison, GFP_NOWAIT);
+}
+
+static void free_prison_cell(struct cache *cache, struct dm_bio_prison_cell *cell)
+{
+	dm_bio_prison_free_cell(cache->prison, cell);
+}
+
+static int prealloc_data_structs(struct cache *cache, struct prealloc *p)
+{
+	if (!p->mg) {
+		p->mg = mempool_alloc(cache->migration_pool, GFP_NOWAIT);
+		if (!p->mg)
+			return -ENOMEM;
+	}
+
+	if (!p->cell1) {
+		p->cell1 = alloc_prison_cell(cache);
+		if (!p->cell1)
+			return -ENOMEM;
+	}
+
+	if (!p->cell2) {
+		p->cell2 = alloc_prison_cell(cache);
+		if (!p->cell2)
+			return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void prealloc_free_structs(struct cache *cache, struct prealloc *p)
+{
+	if (p->cell2)
+		free_prison_cell(cache, p->cell2);
+
+	if (p->cell1)
+		free_prison_cell(cache, p->cell1);
+
+	if (p->mg)
+		mempool_free(p->mg, cache->migration_pool);
+}
+
+static struct dm_cache_migration *prealloc_get_migration(struct prealloc *p)
+{
+	struct dm_cache_migration *mg = p->mg;
+
+	BUG_ON(!mg);
+	p->mg = NULL;
+
+	return mg;
+}
+
+/*
+ * You must have a cell within the prealloc struct to return.  If not this
+ * function will BUG() rather than returning NULL.
+ */
+static struct dm_bio_prison_cell *prealloc_get_cell(struct prealloc *p)
+{
+	struct dm_bio_prison_cell *r = NULL;
+
+	if (p->cell1) {
+		r = p->cell1;
+		p->cell1 = NULL;
+
+	} else if (p->cell2) {
+		r = p->cell2;
+		p->cell2 = NULL;
+	} else
+		BUG();
+
+	return r;
+}
+
+/*
+ * You can't have more than two cells in a prealloc struct.  BUG() will be
+ * called if you try and overfill.
+ */
+static void prealloc_put_cell(struct prealloc *p, struct dm_bio_prison_cell *cell)
+{
+	if (!p->cell2)
+		p->cell2 = cell;
+
+	else if (!p->cell1)
+		p->cell1 = cell;
+
+	else
+		BUG();
+}
+
+/*----------------------------------------------------------------*/
+
+static void build_key(dm_oblock_t oblock, struct dm_cell_key *key)
+{
+	key->virtual = 0;
+	key->dev = 0;
+	key->block = from_oblock(oblock);
+}
+
+/*
+ * The caller hands in a preallocated cell, and a free function for it.
+ * The cell will be freed if there's an error, or if it wasn't used because
+ * a cell with that key already exists.
+ */
+typedef void (*cell_free_fn)(void *context, struct dm_bio_prison_cell *cell);
+
+static int bio_detain(struct cache *cache, dm_oblock_t oblock,
+		      struct bio *bio, struct dm_bio_prison_cell *cell_prealloc,
+		      cell_free_fn free_fn, void *free_context,
+		      struct dm_bio_prison_cell **cell_result)
+{
+	int r;
+	struct dm_cell_key key;
+
+	build_key(oblock, &key);
+	r = dm_bio_detain(cache->prison, &key, bio, cell_prealloc, cell_result);
+	if (r)
+		free_fn(free_context, cell_prealloc);
+
+	return r;
+}
+
+static int get_cell(struct cache *cache,
+		    dm_oblock_t oblock,
+		    struct prealloc *structs,
+		    struct dm_bio_prison_cell **cell_result)
+{
+	int r;
+	struct dm_cell_key key;
+	struct dm_bio_prison_cell *cell_prealloc;
+
+	cell_prealloc = prealloc_get_cell(structs);
+
+	build_key(oblock, &key);
+	r = dm_get_cell(cache->prison, &key, cell_prealloc, cell_result);
+	if (r)
+		prealloc_put_cell(structs, cell_prealloc);
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+static bool is_dirty(struct cache *cache, dm_cblock_t b)
+{
+	return test_bit(from_cblock(b), cache->dirty_bitset);
+}
+
+static void set_dirty(struct cache *cache, dm_oblock_t oblock, dm_cblock_t cblock)
+{
+	if (!test_and_set_bit(from_cblock(cblock), cache->dirty_bitset)) {
+		atomic_inc(&cache->nr_dirty);
+		policy_set_dirty(cache->policy, oblock);
+	}
+}
+
+static void clear_dirty(struct cache *cache, dm_oblock_t oblock, dm_cblock_t cblock)
+{
+	if (test_and_clear_bit(from_cblock(cblock), cache->dirty_bitset)) {
+		policy_clear_dirty(cache->policy, oblock);
+		if (atomic_dec_return(&cache->nr_dirty) == 0)
+			dm_table_event(cache->ti->table);
+	}
+}
+
+/*----------------------------------------------------------------*/
+
+static bool block_size_is_power_of_two(struct cache *cache)
+{
+	return cache->sectors_per_block_shift >= 0;
+}
+
+/* gcc on ARM generates spurious references to __udivdi3 and __umoddi3 */
+#if defined(CONFIG_ARM) && __GNUC__ == 4 && __GNUC_MINOR__ <= 6
+__always_inline
+#endif
+static dm_block_t block_div(dm_block_t b, uint32_t n)
+{
+	do_div(b, n);
+
+	return b;
+}
+
+static void set_discard(struct cache *cache, dm_oblock_t b)
+{
+	unsigned long flags;
+
+	atomic_inc(&cache->stats.discard_count);
+
+	spin_lock_irqsave(&cache->lock, flags);
+	set_bit(from_oblock(b), cache->discard_bitset);
+	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static void clear_discard(struct cache *cache, dm_oblock_t b)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	clear_bit(from_oblock(b), cache->discard_bitset);
+	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static bool is_discarded(struct cache *cache, dm_oblock_t b)
+{
+	int r;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	r = test_bit(from_oblock(b), cache->discard_bitset);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	return r;
+}
+
+static bool is_discarded_oblock(struct cache *cache, dm_oblock_t b)
+{
+	int r;
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	r = test_bit(from_oblock(b), cache->discard_bitset);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+static void load_stats(struct cache *cache)
+{
+	struct dm_cache_statistics stats;
+
+	dm_cache_metadata_get_stats(cache->cmd, &stats);
+	atomic_set(&cache->stats.read_hit, stats.read_hits);
+	atomic_set(&cache->stats.read_miss, stats.read_misses);
+	atomic_set(&cache->stats.write_hit, stats.write_hits);
+	atomic_set(&cache->stats.write_miss, stats.write_misses);
+}
+
+static void save_stats(struct cache *cache)
+{
+	struct dm_cache_statistics stats;
+
+	stats.read_hits = atomic_read(&cache->stats.read_hit);
+	stats.read_misses = atomic_read(&cache->stats.read_miss);
+	stats.write_hits = atomic_read(&cache->stats.write_hit);
+	stats.write_misses = atomic_read(&cache->stats.write_miss);
+
+	dm_cache_metadata_set_stats(cache->cmd, &stats);
+}
+
+/*----------------------------------------------------------------
+ * Per bio data
+ *--------------------------------------------------------------*/
+
+/*
+ * If using writeback, leave out struct per_bio_data's writethrough fields.
+ */
+#define PB_DATA_SIZE_WB (offsetof(struct per_bio_data, cache))
+#define PB_DATA_SIZE_WT (sizeof(struct per_bio_data))
+
+static bool writethrough_mode(struct cache_features *f)
+{
+	return f->io_mode == CM_IO_WRITETHROUGH;
+}
+
+static bool writeback_mode(struct cache_features *f)
+{
+	return f->io_mode == CM_IO_WRITEBACK;
+}
+
+static bool passthrough_mode(struct cache_features *f)
+{
+	return f->io_mode == CM_IO_PASSTHROUGH;
+}
+
+static size_t get_per_bio_data_size(struct cache *cache)
+{
+	return writethrough_mode(&cache->features) ? PB_DATA_SIZE_WT : PB_DATA_SIZE_WB;
+}
+
+static struct per_bio_data *get_per_bio_data(struct bio *bio, size_t data_size)
+{
+	struct per_bio_data *pb = dm_per_bio_data(bio, data_size);
+	BUG_ON(!pb);
+	return pb;
+}
+
+static struct per_bio_data *init_per_bio_data(struct bio *bio, size_t data_size)
+{
+	struct per_bio_data *pb = get_per_bio_data(bio, data_size);
+
+	pb->tick = false;
+	pb->req_nr = dm_bio_get_target_bio_nr(bio);
+	pb->all_io_entry = NULL;
+
+	return pb;
+}
+
+/*----------------------------------------------------------------
+ * Remapping
+ *--------------------------------------------------------------*/
+static void remap_to_origin(struct cache *cache, struct bio *bio)
+{
+	bio->bi_bdev = cache->origin_dev->bdev;
+}
+
+static void remap_to_cache(struct cache *cache, struct bio *bio,
+			   dm_cblock_t cblock)
+{
+	sector_t bi_sector = bio->bi_iter.bi_sector;
+	sector_t block = from_cblock(cblock);
+
+	bio->bi_bdev = cache->cache_dev->bdev;
+	if (!block_size_is_power_of_two(cache))
+		bio->bi_iter.bi_sector =
+			(block * cache->sectors_per_block) +
+			sector_div(bi_sector, cache->sectors_per_block);
+	else
+		bio->bi_iter.bi_sector =
+			(block << cache->sectors_per_block_shift) |
+			(bi_sector & (cache->sectors_per_block - 1));
+}
+
+static void check_if_tick_bio_needed(struct cache *cache, struct bio *bio)
+{
+	unsigned long flags;
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+
+	spin_lock_irqsave(&cache->lock, flags);
+	if (cache->need_tick_bio &&
+	    !(bio->bi_rw & (REQ_FUA | REQ_FLUSH | REQ_DISCARD))) {
+		pb->tick = true;
+		cache->need_tick_bio = false;
+	}
+	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static void remap_to_origin_clear_discard(struct cache *cache, struct bio *bio,
+				  dm_oblock_t oblock)
+{
+	check_if_tick_bio_needed(cache, bio);
+	remap_to_origin(cache, bio);
+	if (bio_data_dir(bio) == WRITE)
+		clear_discard(cache, oblock);
+}
+
+static void remap_to_cache_dirty(struct cache *cache, struct bio *bio,
+				 dm_oblock_t oblock, dm_cblock_t cblock)
+{
+	check_if_tick_bio_needed(cache, bio);
+	remap_to_cache(cache, bio, cblock);
+	if (bio_data_dir(bio) == WRITE) {
+		set_dirty(cache, oblock, cblock);
+		clear_discard(cache, oblock);
+	}
+}
+
+static dm_oblock_t get_bio_block(struct cache *cache, struct bio *bio)
+{
+	sector_t block_nr = bio->bi_iter.bi_sector;
+
+	if (!block_size_is_power_of_two(cache))
+		(void) sector_div(block_nr, cache->sectors_per_block);
+	else
+		block_nr >>= cache->sectors_per_block_shift;
+
+	return to_oblock(block_nr);
+}
+
+static int bio_triggers_commit(struct cache *cache, struct bio *bio)
+{
+	return bio->bi_rw & (REQ_FLUSH | REQ_FUA);
+}
+
+static void issue(struct cache *cache, struct bio *bio)
+{
+	unsigned long flags;
+
+	if (!bio_triggers_commit(cache, bio)) {
+		generic_make_request(bio);
+		return;
+	}
+
+	/*
+	 * Batch together any bios that trigger commits and then issue a
+	 * single commit for them in do_worker().
+	 */
+	spin_lock_irqsave(&cache->lock, flags);
+	cache->commit_requested = true;
+	bio_list_add(&cache->deferred_flush_bios, bio);
+	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static void defer_writethrough_bio(struct cache *cache, struct bio *bio)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	bio_list_add(&cache->deferred_writethrough_bios, bio);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void writethrough_endio(struct bio *bio, int err)
+{
+	struct per_bio_data *pb = get_per_bio_data(bio, PB_DATA_SIZE_WT);
+
+	dm_unhook_bio(&pb->hook_info, bio);
+
+	if (err) {
+		bio_endio(bio, err);
+		return;
+	}
+
+	dm_bio_restore(&pb->bio_details, bio);
+	remap_to_cache(pb->cache, bio, pb->cblock);
+
+	/*
+	 * We can't issue this bio directly, since we're in interrupt
+	 * context.  So it gets put on a bio list for processing by the
+	 * worker thread.
+	 */
+	defer_writethrough_bio(pb->cache, bio);
+}
+
+/*
+ * When running in writethrough mode we need to send writes to clean blocks
+ * to both the cache and origin devices.  In future we'd like to clone the
+ * bio and send them in parallel, but for now we're doing them in
+ * series as this is easier.
+ */
+static void remap_to_origin_then_cache(struct cache *cache, struct bio *bio,
+				       dm_oblock_t oblock, dm_cblock_t cblock)
+{
+	struct per_bio_data *pb = get_per_bio_data(bio, PB_DATA_SIZE_WT);
+
+	pb->cache = cache;
+	pb->cblock = cblock;
+	dm_hook_bio(&pb->hook_info, bio, writethrough_endio, NULL);
+	dm_bio_record(&pb->bio_details, bio);
+
+	remap_to_origin_clear_discard(pb->cache, bio, oblock);
+}
+
+/*----------------------------------------------------------------
+ * Migration processing
+ *
+ * Migration covers moving data from the origin device to the cache, or
+ * vice versa.
+ *--------------------------------------------------------------*/
+static void free_migration(struct dm_cache_migration *mg)
+{
+	mempool_free(mg, mg->cache->migration_pool);
+}
+
+static void inc_nr_migrations(struct cache *cache)
+{
+	atomic_inc(&cache->nr_migrations);
+}
+
+static void dec_nr_migrations(struct cache *cache)
+{
+	atomic_dec(&cache->nr_migrations);
+
+	/*
+	 * Wake the worker in case we're suspending the target.
+	 */
+	wake_up(&cache->migration_wait);
+}
+
+static void __cell_defer(struct cache *cache, struct dm_bio_prison_cell *cell,
+			 bool holder)
+{
+	(holder ? dm_cell_release : dm_cell_release_no_holder)
+		(cache->prison, cell, &cache->deferred_bios);
+	free_prison_cell(cache, cell);
+}
+
+static void cell_defer(struct cache *cache, struct dm_bio_prison_cell *cell,
+		       bool holder)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	__cell_defer(cache, cell, holder);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void cleanup_migration(struct dm_cache_migration *mg)
+{
+	struct cache *cache = mg->cache;
+	free_migration(mg);
+	dec_nr_migrations(cache);
+}
+
+static void migration_failure(struct dm_cache_migration *mg)
+{
+	struct cache *cache = mg->cache;
+
+	if (mg->writeback) {
+		DMWARN_LIMIT("writeback failed; couldn't copy block");
+		set_dirty(cache, mg->old_oblock, mg->cblock);
+		cell_defer(cache, mg->old_ocell, false);
+
+	} else if (mg->demote) {
+		DMWARN_LIMIT("demotion failed; couldn't copy block");
+		policy_force_mapping(cache->policy, mg->new_oblock, mg->old_oblock);
+
+		cell_defer(cache, mg->old_ocell, mg->promote ? false : true);
+		if (mg->promote)
+			cell_defer(cache, mg->new_ocell, true);
+	} else {
+		DMWARN_LIMIT("promotion failed; couldn't copy block");
+		policy_remove_mapping(cache->policy, mg->new_oblock);
+		cell_defer(cache, mg->new_ocell, true);
+	}
+
+	cleanup_migration(mg);
+}
+
+static void migration_success_pre_commit(struct dm_cache_migration *mg)
+{
+	unsigned long flags;
+	struct cache *cache = mg->cache;
+
+	if (mg->writeback) {
+		cell_defer(cache, mg->old_ocell, false);
+		clear_dirty(cache, mg->old_oblock, mg->cblock);
+		cleanup_migration(mg);
+		return;
+
+	} else if (mg->demote) {
+		if (dm_cache_remove_mapping(cache->cmd, mg->cblock)) {
+			DMWARN_LIMIT("demotion failed; couldn't update on disk metadata");
+			policy_force_mapping(cache->policy, mg->new_oblock,
+					     mg->old_oblock);
+			if (mg->promote)
+				cell_defer(cache, mg->new_ocell, true);
+			cleanup_migration(mg);
+			return;
+		}
+	} else {
+		if (dm_cache_insert_mapping(cache->cmd, mg->cblock, mg->new_oblock)) {
+			DMWARN_LIMIT("promotion failed; couldn't update on disk metadata");
+			policy_remove_mapping(cache->policy, mg->new_oblock);
+			cleanup_migration(mg);
+			return;
+		}
+	}
+
+	spin_lock_irqsave(&cache->lock, flags);
+	list_add_tail(&mg->list, &cache->need_commit_migrations);
+	cache->commit_requested = true;
+	spin_unlock_irqrestore(&cache->lock, flags);
+}
+
+static void migration_success_post_commit(struct dm_cache_migration *mg)
+{
+	unsigned long flags;
+	struct cache *cache = mg->cache;
+
+	if (mg->writeback) {
+		DMWARN("writeback unexpectedly triggered commit");
+		return;
+
+	} else if (mg->demote) {
+		cell_defer(cache, mg->old_ocell, mg->promote ? false : true);
+
+		if (mg->promote) {
+			mg->demote = false;
+
+			spin_lock_irqsave(&cache->lock, flags);
+			list_add_tail(&mg->list, &cache->quiesced_migrations);
+			spin_unlock_irqrestore(&cache->lock, flags);
+
+		} else {
+			if (mg->invalidate)
+				policy_remove_mapping(cache->policy, mg->old_oblock);
+			cleanup_migration(mg);
+		}
+
+	} else {
+		if (mg->requeue_holder)
+			cell_defer(cache, mg->new_ocell, true);
+		else {
+			bio_endio(mg->new_ocell->holder, 0);
+			cell_defer(cache, mg->new_ocell, false);
+		}
+		clear_dirty(cache, mg->new_oblock, mg->cblock);
+		cleanup_migration(mg);
+	}
+}
+
+static void copy_complete(int read_err, unsigned long write_err, void *context)
+{
+	unsigned long flags;
+	struct dm_cache_migration *mg = (struct dm_cache_migration *) context;
+	struct cache *cache = mg->cache;
+
+	if (read_err || write_err)
+		mg->err = true;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	list_add_tail(&mg->list, &cache->completed_migrations);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void issue_copy_real(struct dm_cache_migration *mg)
+{
+	int r;
+	struct dm_io_region o_region, c_region;
+	struct cache *cache = mg->cache;
+	sector_t cblock = from_cblock(mg->cblock);
+
+	o_region.bdev = cache->origin_dev->bdev;
+	o_region.count = cache->sectors_per_block;
+
+	c_region.bdev = cache->cache_dev->bdev;
+	c_region.sector = cblock * cache->sectors_per_block;
+	c_region.count = cache->sectors_per_block;
+
+	if (mg->writeback || mg->demote) {
+		/* demote */
+		o_region.sector = from_oblock(mg->old_oblock) * cache->sectors_per_block;
+		r = dm_kcopyd_copy(cache->copier, &c_region, 1, &o_region, 0, copy_complete, mg);
+	} else {
+		/* promote */
+		o_region.sector = from_oblock(mg->new_oblock) * cache->sectors_per_block;
+		r = dm_kcopyd_copy(cache->copier, &o_region, 1, &c_region, 0, copy_complete, mg);
+	}
+
+	if (r < 0) {
+		DMERR_LIMIT("issuing migration failed");
+		migration_failure(mg);
+	}
+}
+
+static void overwrite_endio(struct bio *bio, int err)
+{
+	struct dm_cache_migration *mg = bio->bi_private;
+	struct cache *cache = mg->cache;
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+	unsigned long flags;
+
+	dm_unhook_bio(&pb->hook_info, bio);
+
+	if (err)
+		mg->err = true;
+
+	mg->requeue_holder = false;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	list_add_tail(&mg->list, &cache->completed_migrations);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void issue_overwrite(struct dm_cache_migration *mg, struct bio *bio)
+{
+	size_t pb_data_size = get_per_bio_data_size(mg->cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+
+	dm_hook_bio(&pb->hook_info, bio, overwrite_endio, mg);
+	remap_to_cache_dirty(mg->cache, bio, mg->new_oblock, mg->cblock);
+	generic_make_request(bio);
+}
+
+static bool bio_writes_complete_block(struct cache *cache, struct bio *bio)
+{
+	return (bio_data_dir(bio) == WRITE) &&
+		(bio->bi_iter.bi_size == (cache->sectors_per_block << SECTOR_SHIFT));
+}
+
+static void avoid_copy(struct dm_cache_migration *mg)
+{
+	atomic_inc(&mg->cache->stats.copies_avoided);
+	migration_success_pre_commit(mg);
+}
+
+static void issue_copy(struct dm_cache_migration *mg)
+{
+	bool avoid;
+	struct cache *cache = mg->cache;
+
+	if (mg->writeback || mg->demote)
+		avoid = !is_dirty(cache, mg->cblock) ||
+			is_discarded_oblock(cache, mg->old_oblock);
+	else {
+		struct bio *bio = mg->new_ocell->holder;
+
+		avoid = is_discarded_oblock(cache, mg->new_oblock);
+
+		if (!avoid && bio_writes_complete_block(cache, bio)) {
+			issue_overwrite(mg, bio);
+			return;
+		}
+	}
+
+	avoid ? avoid_copy(mg) : issue_copy_real(mg);
+}
+
+static void complete_migration(struct dm_cache_migration *mg)
+{
+	if (mg->err)
+		migration_failure(mg);
+	else
+		migration_success_pre_commit(mg);
+}
+
+static void process_migrations(struct cache *cache, struct list_head *head,
+			       void (*fn)(struct dm_cache_migration *))
+{
+	unsigned long flags;
+	struct list_head list;
+	struct dm_cache_migration *mg, *tmp;
+
+	INIT_LIST_HEAD(&list);
+	spin_lock_irqsave(&cache->lock, flags);
+	list_splice_init(head, &list);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	list_for_each_entry_safe(mg, tmp, &list, list)
+		fn(mg);
+}
+
+static void __queue_quiesced_migration(struct dm_cache_migration *mg)
+{
+	list_add_tail(&mg->list, &mg->cache->quiesced_migrations);
+}
+
+static void queue_quiesced_migration(struct dm_cache_migration *mg)
+{
+	unsigned long flags;
+	struct cache *cache = mg->cache;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	__queue_quiesced_migration(mg);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void queue_quiesced_migrations(struct cache *cache, struct list_head *work)
+{
+	unsigned long flags;
+	struct dm_cache_migration *mg, *tmp;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	list_for_each_entry_safe(mg, tmp, work, list)
+		__queue_quiesced_migration(mg);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void check_for_quiesced_migrations(struct cache *cache,
+					  struct per_bio_data *pb)
+{
+	struct list_head work;
+
+	if (!pb->all_io_entry)
+		return;
+
+	INIT_LIST_HEAD(&work);
+	if (pb->all_io_entry)
+		dm_deferred_entry_dec(pb->all_io_entry, &work);
+
+	if (!list_empty(&work))
+		queue_quiesced_migrations(cache, &work);
+}
+
+static void quiesce_migration(struct dm_cache_migration *mg)
+{
+	if (!dm_deferred_set_add_work(mg->cache->all_io_ds, &mg->list))
+		queue_quiesced_migration(mg);
+}
+
+static void promote(struct cache *cache, struct prealloc *structs,
+		    dm_oblock_t oblock, dm_cblock_t cblock,
+		    struct dm_bio_prison_cell *cell)
+{
+	struct dm_cache_migration *mg = prealloc_get_migration(structs);
+
+	mg->err = false;
+	mg->writeback = false;
+	mg->demote = false;
+	mg->promote = true;
+	mg->requeue_holder = true;
+	mg->invalidate = false;
+	mg->cache = cache;
+	mg->new_oblock = oblock;
+	mg->cblock = cblock;
+	mg->old_ocell = NULL;
+	mg->new_ocell = cell;
+	mg->start_jiffies = jiffies;
+
+	inc_nr_migrations(cache);
+	quiesce_migration(mg);
+}
+
+static void writeback(struct cache *cache, struct prealloc *structs,
+		      dm_oblock_t oblock, dm_cblock_t cblock,
+		      struct dm_bio_prison_cell *cell)
+{
+	struct dm_cache_migration *mg = prealloc_get_migration(structs);
+
+	mg->err = false;
+	mg->writeback = true;
+	mg->demote = false;
+	mg->promote = false;
+	mg->requeue_holder = true;
+	mg->invalidate = false;
+	mg->cache = cache;
+	mg->old_oblock = oblock;
+	mg->cblock = cblock;
+	mg->old_ocell = cell;
+	mg->new_ocell = NULL;
+	mg->start_jiffies = jiffies;
+
+	inc_nr_migrations(cache);
+	quiesce_migration(mg);
+}
+
+static void demote_then_promote(struct cache *cache, struct prealloc *structs,
+				dm_oblock_t old_oblock, dm_oblock_t new_oblock,
+				dm_cblock_t cblock,
+				struct dm_bio_prison_cell *old_ocell,
+				struct dm_bio_prison_cell *new_ocell)
+{
+	struct dm_cache_migration *mg = prealloc_get_migration(structs);
+
+	mg->err = false;
+	mg->writeback = false;
+	mg->demote = true;
+	mg->promote = true;
+	mg->requeue_holder = true;
+	mg->invalidate = false;
+	mg->cache = cache;
+	mg->old_oblock = old_oblock;
+	mg->new_oblock = new_oblock;
+	mg->cblock = cblock;
+	mg->old_ocell = old_ocell;
+	mg->new_ocell = new_ocell;
+	mg->start_jiffies = jiffies;
+
+	inc_nr_migrations(cache);
+	quiesce_migration(mg);
+}
+
+/*
+ * Invalidate a cache entry.  No writeback occurs; any changes in the cache
+ * block are thrown away.
+ */
+static void invalidate(struct cache *cache, struct prealloc *structs,
+		       dm_oblock_t oblock, dm_cblock_t cblock,
+		       struct dm_bio_prison_cell *cell)
+{
+	struct dm_cache_migration *mg = prealloc_get_migration(structs);
+
+	mg->err = false;
+	mg->writeback = false;
+	mg->demote = true;
+	mg->promote = false;
+	mg->requeue_holder = true;
+	mg->invalidate = true;
+	mg->cache = cache;
+	mg->old_oblock = oblock;
+	mg->cblock = cblock;
+	mg->old_ocell = cell;
+	mg->new_ocell = NULL;
+	mg->start_jiffies = jiffies;
+
+	inc_nr_migrations(cache);
+	quiesce_migration(mg);
+}
+
+/*----------------------------------------------------------------
+ * bio processing
+ *--------------------------------------------------------------*/
+static void defer_bio(struct cache *cache, struct bio *bio)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&cache->lock, flags);
+	bio_list_add(&cache->deferred_bios, bio);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	wake_worker(cache);
+}
+
+static void process_flush_bio(struct cache *cache, struct bio *bio)
+{
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+
+	BUG_ON(bio->bi_iter.bi_size);
+	if (!pb->req_nr)
+		remap_to_origin(cache, bio);
+	else
+		remap_to_cache(cache, bio, 0);
+
+	issue(cache, bio);
+}
+
+/*
+ * People generally discard large parts of a device, eg, the whole device
+ * when formatting.  Splitting these large discards up into cache block
+ * sized ios and then quiescing (always neccessary for discard) takes too
+ * long.
+ *
+ * We keep it simple, and allow any size of discard to come in, and just
+ * mark off blocks on the discard bitset.  No passdown occurs!
+ *
+ * To implement passdown we need to change the bio_prison such that a cell
+ * can have a key that spans many blocks.
+ */
+static void process_discard_bio(struct cache *cache, struct bio *bio)
+{
+	dm_block_t start_block = dm_sector_div_up(bio->bi_iter.bi_sector,
+						  cache->sectors_per_block);
+	dm_block_t end_block = bio_end_sector(bio);
+	dm_block_t b;
+
+	end_block = block_div(end_block, cache->sectors_per_block);
+
+	for (b = start_block; b < end_block; b++)
+		set_discard(cache, to_oblock(b));
+
+	bio_endio(bio, 0);
+}
+
+static bool spare_migration_bandwidth(struct cache *cache)
+{
+	sector_t current_volume = (atomic_read(&cache->nr_migrations) + 1) *
+		cache->sectors_per_block;
+	return current_volume < cache->migration_threshold;
+}
+
+static void inc_hit_counter(struct cache *cache, struct bio *bio)
+{
+	atomic_inc(bio_data_dir(bio) == READ ?
+		   &cache->stats.read_hit : &cache->stats.write_hit);
+}
+
+static void inc_miss_counter(struct cache *cache, struct bio *bio)
+{
+	atomic_inc(bio_data_dir(bio) == READ ?
+		   &cache->stats.read_miss : &cache->stats.write_miss);
+}
+
+static void issue_cache_bio(struct cache *cache, struct bio *bio,
+			    struct per_bio_data *pb,
+			    dm_oblock_t oblock, dm_cblock_t cblock)
+{
+	pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+	remap_to_cache_dirty(cache, bio, oblock, cblock);
+	issue(cache, bio);
+}
+
+static void process_bio(struct cache *cache, struct prealloc *structs,
+			struct bio *bio)
+{
+	int r;
+	bool release_cell = true;
+	dm_oblock_t block = get_bio_block(cache, bio);
+	struct dm_bio_prison_cell *cell_prealloc, *old_ocell, *new_ocell;
+	struct policy_result lookup_result;
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+	bool discarded_block = is_discarded_oblock(cache, block);
+	bool passthrough = passthrough_mode(&cache->features);
+	bool can_migrate = !passthrough && (discarded_block || spare_migration_bandwidth(cache));
+
+	/*
+	 * Check to see if that block is currently migrating.
+	 */
+	cell_prealloc = prealloc_get_cell(structs);
+	r = bio_detain(cache, block, bio, cell_prealloc,
+		       (cell_free_fn) prealloc_put_cell,
+		       structs, &new_ocell);
+	if (r > 0)
+		return;
+
+	r = policy_map(cache->policy, block, true, can_migrate, discarded_block,
+		       bio, &lookup_result);
+
+	if (r == -EWOULDBLOCK)
+		/* migration has been denied */
+		lookup_result.op = POLICY_MISS;
+
+	switch (lookup_result.op) {
+	case POLICY_HIT:
+		if (passthrough) {
+			inc_miss_counter(cache, bio);
+
+			/*
+			 * Passthrough always maps to the origin,
+			 * invalidating any cache blocks that are written
+			 * to.
+			 */
+
+			if (bio_data_dir(bio) == WRITE) {
+				atomic_inc(&cache->stats.demotion);
+				invalidate(cache, structs, block, lookup_result.cblock, new_ocell);
+				release_cell = false;
+
+			} else {
+				/* FIXME: factor out issue_origin() */
+				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+				remap_to_origin_clear_discard(cache, bio, block);
+				issue(cache, bio);
+			}
+		} else {
+			inc_hit_counter(cache, bio);
+
+			if (bio_data_dir(bio) == WRITE &&
+			    writethrough_mode(&cache->features) &&
+			    !is_dirty(cache, lookup_result.cblock)) {
+				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+				remap_to_origin_then_cache(cache, bio, block, lookup_result.cblock);
+				issue(cache, bio);
+			} else
+				issue_cache_bio(cache, bio, pb, block, lookup_result.cblock);
+		}
+
+		break;
+
+	case POLICY_MISS:
+		inc_miss_counter(cache, bio);
+		pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+		remap_to_origin_clear_discard(cache, bio, block);
+		issue(cache, bio);
+		break;
+
+	case POLICY_NEW:
+		atomic_inc(&cache->stats.promotion);
+		promote(cache, structs, block, lookup_result.cblock, new_ocell);
+		release_cell = false;
+		break;
+
+	case POLICY_REPLACE:
+		cell_prealloc = prealloc_get_cell(structs);
+		r = bio_detain(cache, lookup_result.old_oblock, bio, cell_prealloc,
+			       (cell_free_fn) prealloc_put_cell,
+			       structs, &old_ocell);
+		if (r > 0) {
+			/*
+			 * We have to be careful to avoid lock inversion of
+			 * the cells.  So we back off, and wait for the
+			 * old_ocell to become free.
+			 */
+			policy_force_mapping(cache->policy, block,
+					     lookup_result.old_oblock);
+			atomic_inc(&cache->stats.cache_cell_clash);
+			break;
+		}
+		atomic_inc(&cache->stats.demotion);
+		atomic_inc(&cache->stats.promotion);
+
+		demote_then_promote(cache, structs, lookup_result.old_oblock,
+				    block, lookup_result.cblock,
+				    old_ocell, new_ocell);
+		release_cell = false;
+		break;
+
+	default:
+		DMERR_LIMIT("%s: erroring bio, unknown policy op: %u", __func__,
+			    (unsigned) lookup_result.op);
+		bio_io_error(bio);
+	}
+
+	if (release_cell)
+		cell_defer(cache, new_ocell, false);
+}
+
+static int need_commit_due_to_time(struct cache *cache)
+{
+	return jiffies < cache->last_commit_jiffies ||
+	       jiffies > cache->last_commit_jiffies + COMMIT_PERIOD;
+}
+
+static int commit_if_needed(struct cache *cache)
+{
+	int r = 0;
+
+	if ((cache->commit_requested || need_commit_due_to_time(cache)) &&
+	    dm_cache_changed_this_transaction(cache->cmd)) {
+		atomic_inc(&cache->stats.commit_count);
+		cache->commit_requested = false;
+		r = dm_cache_commit(cache->cmd, false);
+		cache->last_commit_jiffies = jiffies;
+	}
+
+	return r;
+}
+
+static void process_deferred_bios(struct cache *cache)
+{
+	unsigned long flags;
+	struct bio_list bios;
+	struct bio *bio;
+	struct prealloc structs;
+
+	memset(&structs, 0, sizeof(structs));
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&cache->lock, flags);
+	bio_list_merge(&bios, &cache->deferred_bios);
+	bio_list_init(&cache->deferred_bios);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	while (!bio_list_empty(&bios)) {
+		/*
+		 * If we've got no free migration structs, and processing
+		 * this bio might require one, we pause until there are some
+		 * prepared mappings to process.
+		 */
+		if (prealloc_data_structs(cache, &structs)) {
+			spin_lock_irqsave(&cache->lock, flags);
+			bio_list_merge(&cache->deferred_bios, &bios);
+			spin_unlock_irqrestore(&cache->lock, flags);
+			break;
+		}
+
+		bio = bio_list_pop(&bios);
+
+		if (bio->bi_rw & REQ_FLUSH)
+			process_flush_bio(cache, bio);
+		else if (bio->bi_rw & REQ_DISCARD)
+			process_discard_bio(cache, bio);
+		else
+			process_bio(cache, &structs, bio);
+	}
+
+	prealloc_free_structs(cache, &structs);
+}
+
+static void process_deferred_flush_bios(struct cache *cache, bool submit_bios)
+{
+	unsigned long flags;
+	struct bio_list bios;
+	struct bio *bio;
+
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&cache->lock, flags);
+	bio_list_merge(&bios, &cache->deferred_flush_bios);
+	bio_list_init(&cache->deferred_flush_bios);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	while ((bio = bio_list_pop(&bios)))
+		submit_bios ? generic_make_request(bio) : bio_io_error(bio);
+}
+
+static void process_deferred_writethrough_bios(struct cache *cache)
+{
+	unsigned long flags;
+	struct bio_list bios;
+	struct bio *bio;
+
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&cache->lock, flags);
+	bio_list_merge(&bios, &cache->deferred_writethrough_bios);
+	bio_list_init(&cache->deferred_writethrough_bios);
+	spin_unlock_irqrestore(&cache->lock, flags);
+
+	while ((bio = bio_list_pop(&bios)))
+		generic_make_request(bio);
+}
+
+static void writeback_some_dirty_blocks(struct cache *cache)
+{
+	int r = 0;
+	dm_oblock_t oblock;
+	dm_cblock_t cblock;
+	struct prealloc structs;
+	struct dm_bio_prison_cell *old_ocell;
+
+	memset(&structs, 0, sizeof(structs));
+
+	while (spare_migration_bandwidth(cache)) {
+		if (prealloc_data_structs(cache, &structs))
+			break;
+
+		r = policy_writeback_work(cache->policy, &oblock, &cblock);
+		if (r)
+			break;
+
+		r = get_cell(cache, oblock, &structs, &old_ocell);
+		if (r) {
+			policy_set_dirty(cache->policy, oblock);
+			break;
+		}
+
+		writeback(cache, &structs, oblock, cblock, old_ocell);
+	}
+
+	prealloc_free_structs(cache, &structs);
+}
+
+/*----------------------------------------------------------------
+ * Invalidations.
+ * Dropping something from the cache *without* writing back.
+ *--------------------------------------------------------------*/
+
+static void process_invalidation_request(struct cache *cache, struct invalidation_request *req)
+{
+	int r = 0;
+	uint64_t begin = from_cblock(req->cblocks->begin);
+	uint64_t end = from_cblock(req->cblocks->end);
+
+	while (begin != end) {
+		r = policy_remove_cblock(cache->policy, to_cblock(begin));
+		if (!r) {
+			r = dm_cache_remove_mapping(cache->cmd, to_cblock(begin));
+			if (r)
+				break;
+
+		} else if (r == -ENODATA) {
+			/* harmless, already unmapped */
+			r = 0;
+
+		} else {
+			DMERR("policy_remove_cblock failed");
+			break;
+		}
+
+		begin++;
+        }
+
+	cache->commit_requested = true;
+
+	req->err = r;
+	atomic_set(&req->complete, 1);
+
+	wake_up(&req->result_wait);
+}
+
+static void process_invalidation_requests(struct cache *cache)
+{
+	struct list_head list;
+	struct invalidation_request *req, *tmp;
+
+	INIT_LIST_HEAD(&list);
+	spin_lock(&cache->invalidation_lock);
+	list_splice_init(&cache->invalidation_requests, &list);
+	spin_unlock(&cache->invalidation_lock);
+
+	list_for_each_entry_safe (req, tmp, &list, list)
+		process_invalidation_request(cache, req);
+}
+
+/*----------------------------------------------------------------
+ * Main worker loop
+ *--------------------------------------------------------------*/
+static bool is_quiescing(struct cache *cache)
+{
+	return atomic_read(&cache->quiescing);
+}
+
+static void ack_quiescing(struct cache *cache)
+{
+	if (is_quiescing(cache)) {
+		atomic_inc(&cache->quiescing_ack);
+		wake_up(&cache->quiescing_wait);
+	}
+}
+
+static void wait_for_quiescing_ack(struct cache *cache)
+{
+	wait_event(cache->quiescing_wait, atomic_read(&cache->quiescing_ack));
+}
+
+static void start_quiescing(struct cache *cache)
+{
+	atomic_inc(&cache->quiescing);
+	wait_for_quiescing_ack(cache);
+}
+
+static void stop_quiescing(struct cache *cache)
+{
+	atomic_set(&cache->quiescing, 0);
+	atomic_set(&cache->quiescing_ack, 0);
+}
+
+static void wait_for_migrations(struct cache *cache)
+{
+	wait_event(cache->migration_wait, !atomic_read(&cache->nr_migrations));
+}
+
+static void stop_worker(struct cache *cache)
+{
+	cancel_delayed_work(&cache->waker);
+	flush_workqueue(cache->wq);
+}
+
+static void requeue_deferred_io(struct cache *cache)
+{
+	struct bio *bio;
+	struct bio_list bios;
+
+	bio_list_init(&bios);
+	bio_list_merge(&bios, &cache->deferred_bios);
+	bio_list_init(&cache->deferred_bios);
+
+	while ((bio = bio_list_pop(&bios)))
+		bio_endio(bio, DM_ENDIO_REQUEUE);
+}
+
+static int more_work(struct cache *cache)
+{
+	if (is_quiescing(cache))
+		return !list_empty(&cache->quiesced_migrations) ||
+			!list_empty(&cache->completed_migrations) ||
+			!list_empty(&cache->need_commit_migrations);
+	else
+		return !bio_list_empty(&cache->deferred_bios) ||
+			!bio_list_empty(&cache->deferred_flush_bios) ||
+			!bio_list_empty(&cache->deferred_writethrough_bios) ||
+			!list_empty(&cache->quiesced_migrations) ||
+			!list_empty(&cache->completed_migrations) ||
+			!list_empty(&cache->need_commit_migrations) ||
+			cache->invalidate;
+}
+
+static void do_worker(struct work_struct *ws)
+{
+	struct cache *cache = container_of(ws, struct cache, worker);
+
+	do {
+		if (!is_quiescing(cache)) {
+			writeback_some_dirty_blocks(cache);
+			process_deferred_writethrough_bios(cache);
+			process_deferred_bios(cache);
+			process_invalidation_requests(cache);
+		}
+
+		process_migrations(cache, &cache->quiesced_migrations, issue_copy);
+		process_migrations(cache, &cache->completed_migrations, complete_migration);
+
+		if (commit_if_needed(cache)) {
+			process_deferred_flush_bios(cache, false);
+
+			/*
+			 * FIXME: rollback metadata or just go into a
+			 * failure mode and error everything
+			 */
+		} else {
+			process_deferred_flush_bios(cache, true);
+			process_migrations(cache, &cache->need_commit_migrations,
+					   migration_success_post_commit);
+		}
+
+		ack_quiescing(cache);
+
+	} while (more_work(cache));
+}
+
+/*
+ * We want to commit periodically so that not too much
+ * unwritten metadata builds up.
+ */
+static void do_waker(struct work_struct *ws)
+{
+	struct cache *cache = container_of(to_delayed_work(ws), struct cache, waker);
+	policy_tick(cache->policy);
+	wake_worker(cache);
+	queue_delayed_work(cache->wq, &cache->waker, COMMIT_PERIOD);
+}
+
+/*----------------------------------------------------------------*/
+
+static int is_congested(struct dm_dev *dev, int bdi_bits)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+	return bdi_congested(&q->backing_dev_info, bdi_bits);
+}
+
+static int cache_is_congested(struct dm_target_callbacks *cb, int bdi_bits)
+{
+	struct cache *cache = container_of(cb, struct cache, callbacks);
+
+	return is_congested(cache->origin_dev, bdi_bits) ||
+		is_congested(cache->cache_dev, bdi_bits);
+}
+
+/*----------------------------------------------------------------
+ * Target methods
+ *--------------------------------------------------------------*/
+
+/*
+ * This function gets called on the error paths of the constructor, so we
+ * have to cope with a partially initialised struct.
+ */
+static void destroy(struct cache *cache)
+{
+	unsigned i;
+
+	if (cache->next_migration)
+		mempool_free(cache->next_migration, cache->migration_pool);
+
+	if (cache->migration_pool)
+		mempool_destroy(cache->migration_pool);
+
+	if (cache->all_io_ds)
+		dm_deferred_set_destroy(cache->all_io_ds);
+
+	if (cache->prison)
+		dm_bio_prison_destroy(cache->prison);
+
+	if (cache->wq)
+		destroy_workqueue(cache->wq);
+
+	if (cache->dirty_bitset)
+		free_bitset(cache->dirty_bitset);
+
+	if (cache->discard_bitset)
+		free_bitset(cache->discard_bitset);
+
+	if (cache->copier)
+		dm_kcopyd_client_destroy(cache->copier);
+
+	if (cache->cmd)
+		dm_cache_metadata_close(cache->cmd);
+
+	if (cache->metadata_dev)
+		dm_put_device(cache->ti, cache->metadata_dev);
+
+	if (cache->origin_dev)
+		dm_put_device(cache->ti, cache->origin_dev);
+
+	if (cache->cache_dev)
+		dm_put_device(cache->ti, cache->cache_dev);
+
+	if (cache->policy)
+		dm_cache_policy_destroy(cache->policy);
+
+	for (i = 0; i < cache->nr_ctr_args ; i++)
+		kfree(cache->ctr_args[i]);
+	kfree(cache->ctr_args);
+
+	kfree(cache);
+}
+
+static void cache_dtr(struct dm_target *ti)
+{
+	struct cache *cache = ti->private;
+
+	destroy(cache);
+}
+
+static sector_t get_dev_size(struct dm_dev *dev)
+{
+	return i_size_read(dev->bdev->bd_inode) >> SECTOR_SHIFT;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Construct a cache device mapping.
+ *
+ * cache <metadata dev> <cache dev> <origin dev> <block size>
+ *       <#feature args> [<feature arg>]*
+ *       <policy> <#policy args> [<policy arg>]*
+ *
+ * metadata dev    : fast device holding the persistent metadata
+ * cache dev	   : fast device holding cached data blocks
+ * origin dev	   : slow device holding original data blocks
+ * block size	   : cache unit size in sectors
+ *
+ * #feature args   : number of feature arguments passed
+ * feature args    : writethrough.  (The default is writeback.)
+ *
+ * policy	   : the replacement policy to use
+ * #policy args    : an even number of policy arguments corresponding
+ *		     to key/value pairs passed to the policy
+ * policy args	   : key/value pairs passed to the policy
+ *		     E.g. 'sequential_threshold 1024'
+ *		     See cache-policies.txt for details.
+ *
+ * Optional feature arguments are:
+ *   writethrough  : write through caching that prohibits cache block
+ *		     content from being different from origin block content.
+ *		     Without this argument, the default behaviour is to write
+ *		     back cache block contents later for performance reasons,
+ *		     so they may differ from the corresponding origin blocks.
+ */
+struct cache_args {
+	struct dm_target *ti;
+
+	struct dm_dev *metadata_dev;
+
+	struct dm_dev *cache_dev;
+	sector_t cache_sectors;
+
+	struct dm_dev *origin_dev;
+	sector_t origin_sectors;
+
+	uint32_t block_size;
+
+	const char *policy_name;
+	int policy_argc;
+	const char **policy_argv;
+
+	struct cache_features features;
+};
+
+static void destroy_cache_args(struct cache_args *ca)
+{
+	if (ca->metadata_dev)
+		dm_put_device(ca->ti, ca->metadata_dev);
+
+	if (ca->cache_dev)
+		dm_put_device(ca->ti, ca->cache_dev);
+
+	if (ca->origin_dev)
+		dm_put_device(ca->ti, ca->origin_dev);
+
+	kfree(ca);
+}
+
+static bool at_least_one_arg(struct dm_arg_set *as, char **error)
+{
+	if (!as->argc) {
+		*error = "Insufficient args";
+		return false;
+	}
+
+	return true;
+}
+
+static int parse_metadata_dev(struct cache_args *ca, struct dm_arg_set *as,
+			      char **error)
+{
+	int r;
+	sector_t metadata_dev_size;
+	char b[BDEVNAME_SIZE];
+
+	if (!at_least_one_arg(as, error))
+		return -EINVAL;
+
+	r = dm_get_device(ca->ti, dm_shift_arg(as), FMODE_READ | FMODE_WRITE,
+			  &ca->metadata_dev);
+	if (r) {
+		*error = "Error opening metadata device";
+		return r;
+	}
+
+	metadata_dev_size = get_dev_size(ca->metadata_dev);
+	if (metadata_dev_size > DM_CACHE_METADATA_MAX_SECTORS_WARNING)
+		DMWARN("Metadata device %s is larger than %u sectors: excess space will not be used.",
+		       bdevname(ca->metadata_dev->bdev, b), THIN_METADATA_MAX_SECTORS);
+
+	return 0;
+}
+
+static int parse_cache_dev(struct cache_args *ca, struct dm_arg_set *as,
+			   char **error)
+{
+	int r;
+
+	if (!at_least_one_arg(as, error))
+		return -EINVAL;
+
+	r = dm_get_device(ca->ti, dm_shift_arg(as), FMODE_READ | FMODE_WRITE,
+			  &ca->cache_dev);
+	if (r) {
+		*error = "Error opening cache device";
+		return r;
+	}
+	ca->cache_sectors = get_dev_size(ca->cache_dev);
+
+	return 0;
+}
+
+static int parse_origin_dev(struct cache_args *ca, struct dm_arg_set *as,
+			    char **error)
+{
+	int r;
+
+	if (!at_least_one_arg(as, error))
+		return -EINVAL;
+
+	r = dm_get_device(ca->ti, dm_shift_arg(as), FMODE_READ | FMODE_WRITE,
+			  &ca->origin_dev);
+	if (r) {
+		*error = "Error opening origin device";
+		return r;
+	}
+
+	ca->origin_sectors = get_dev_size(ca->origin_dev);
+	if (ca->ti->len > ca->origin_sectors) {
+		*error = "Device size larger than cached device";
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int parse_block_size(struct cache_args *ca, struct dm_arg_set *as,
+			    char **error)
+{
+	unsigned long block_size;
+
+	if (!at_least_one_arg(as, error))
+		return -EINVAL;
+
+	if (kstrtoul(dm_shift_arg(as), 10, &block_size) || !block_size ||
+	    block_size < DATA_DEV_BLOCK_SIZE_MIN_SECTORS ||
+	    block_size > DATA_DEV_BLOCK_SIZE_MAX_SECTORS ||
+	    block_size & (DATA_DEV_BLOCK_SIZE_MIN_SECTORS - 1)) {
+		*error = "Invalid data block size";
+		return -EINVAL;
+	}
+
+	if (block_size > ca->cache_sectors) {
+		*error = "Data block size is larger than the cache device";
+		return -EINVAL;
+	}
+
+	ca->block_size = block_size;
+
+	return 0;
+}
+
+static void init_features(struct cache_features *cf)
+{
+	cf->mode = CM_WRITE;
+	cf->io_mode = CM_IO_WRITEBACK;
+}
+
+static int parse_features(struct cache_args *ca, struct dm_arg_set *as,
+			  char **error)
+{
+	static struct dm_arg _args[] = {
+		{0, 1, "Invalid number of cache feature arguments"},
+	};
+
+	int r;
+	unsigned argc;
+	const char *arg;
+	struct cache_features *cf = &ca->features;
+
+	init_features(cf);
+
+	r = dm_read_arg_group(_args, as, &argc, error);
+	if (r)
+		return -EINVAL;
+
+	while (argc--) {
+		arg = dm_shift_arg(as);
+
+		if (!strcasecmp(arg, "writeback"))
+			cf->io_mode = CM_IO_WRITEBACK;
+
+		else if (!strcasecmp(arg, "writethrough"))
+			cf->io_mode = CM_IO_WRITETHROUGH;
+
+		else if (!strcasecmp(arg, "passthrough"))
+			cf->io_mode = CM_IO_PASSTHROUGH;
+
+		else {
+			*error = "Unrecognised cache feature requested";
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static int parse_policy(struct cache_args *ca, struct dm_arg_set *as,
+			char **error)
+{
+	static struct dm_arg _args[] = {
+		{0, 1024, "Invalid number of policy arguments"},
+	};
+
+	int r;
+
+	if (!at_least_one_arg(as, error))
+		return -EINVAL;
+
+	ca->policy_name = dm_shift_arg(as);
+
+	r = dm_read_arg_group(_args, as, &ca->policy_argc, error);
+	if (r)
+		return -EINVAL;
+
+	ca->policy_argv = (const char **)as->argv;
+	dm_consume_args(as, ca->policy_argc);
+
+	return 0;
+}
+
+static int parse_cache_args(struct cache_args *ca, int argc, char **argv,
+			    char **error)
+{
+	int r;
+	struct dm_arg_set as;
+
+	as.argc = argc;
+	as.argv = argv;
+
+	r = parse_metadata_dev(ca, &as, error);
+	if (r)
+		return r;
+
+	r = parse_cache_dev(ca, &as, error);
+	if (r)
+		return r;
+
+	r = parse_origin_dev(ca, &as, error);
+	if (r)
+		return r;
+
+	r = parse_block_size(ca, &as, error);
+	if (r)
+		return r;
+
+	r = parse_features(ca, &as, error);
+	if (r)
+		return r;
+
+	r = parse_policy(ca, &as, error);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
+static struct kmem_cache *migration_cache;
+
+#define NOT_CORE_OPTION 1
+
+static int process_config_option(struct cache *cache, const char *key, const char *value)
+{
+	unsigned long tmp;
+
+	if (!strcasecmp(key, "migration_threshold")) {
+		if (kstrtoul(value, 10, &tmp))
+			return -EINVAL;
+
+		cache->migration_threshold = tmp;
+		return 0;
+	}
+
+	return NOT_CORE_OPTION;
+}
+
+static int set_config_value(struct cache *cache, const char *key, const char *value)
+{
+	int r = process_config_option(cache, key, value);
+
+	if (r == NOT_CORE_OPTION)
+		r = policy_set_config_value(cache->policy, key, value);
+
+	if (r)
+		DMWARN("bad config value for %s: %s", key, value);
+
+	return r;
+}
+
+static int set_config_values(struct cache *cache, int argc, const char **argv)
+{
+	int r = 0;
+
+	if (argc & 1) {
+		DMWARN("Odd number of policy arguments given but they should be <key> <value> pairs.");
+		return -EINVAL;
+	}
+
+	while (argc) {
+		r = set_config_value(cache, argv[0], argv[1]);
+		if (r)
+			break;
+
+		argc -= 2;
+		argv += 2;
+	}
+
+	return r;
+}
+
+static int create_cache_policy(struct cache *cache, struct cache_args *ca,
+			       char **error)
+{
+	struct dm_cache_policy *p = dm_cache_policy_create(ca->policy_name,
+							   cache->cache_size,
+							   cache->origin_sectors,
+							   cache->sectors_per_block);
+	if (IS_ERR(p)) {
+		*error = "Error creating cache's policy";
+		return PTR_ERR(p);
+	}
+	cache->policy = p;
+
+	return 0;
+}
+
+#define DEFAULT_MIGRATION_THRESHOLD 2048
+
+static int cache_create(struct cache_args *ca, struct cache **result)
+{
+	int r = 0;
+	char **error = &ca->ti->error;
+	struct cache *cache;
+	struct dm_target *ti = ca->ti;
+	dm_block_t origin_blocks;
+	struct dm_cache_metadata *cmd;
+	bool may_format = ca->features.mode == CM_WRITE;
+
+	cache = kzalloc(sizeof(*cache), GFP_KERNEL);
+	if (!cache)
+		return -ENOMEM;
+
+	cache->ti = ca->ti;
+	ti->private = cache;
+	ti->num_flush_bios = 2;
+	ti->flush_supported = true;
+
+	ti->num_discard_bios = 1;
+	ti->discards_supported = true;
+	ti->discard_zeroes_data_unsupported = true;
+	/* Discard bios must be split on a block boundary */
+	ti->split_discard_bios = true;
+
+	cache->features = ca->features;
+	ti->per_bio_data_size = get_per_bio_data_size(cache);
+
+	cache->callbacks.congested_fn = cache_is_congested;
+	dm_table_add_target_callbacks(ti->table, &cache->callbacks);
+
+	cache->metadata_dev = ca->metadata_dev;
+	cache->origin_dev = ca->origin_dev;
+	cache->cache_dev = ca->cache_dev;
+
+	ca->metadata_dev = ca->origin_dev = ca->cache_dev = NULL;
+
+	/* FIXME: factor out this whole section */
+	origin_blocks = cache->origin_sectors = ca->origin_sectors;
+	origin_blocks = block_div(origin_blocks, ca->block_size);
+	cache->origin_blocks = to_oblock(origin_blocks);
+
+	cache->sectors_per_block = ca->block_size;
+	if (dm_set_target_max_io_len(ti, cache->sectors_per_block)) {
+		r = -EINVAL;
+		goto bad;
+	}
+
+	if (ca->block_size & (ca->block_size - 1)) {
+		dm_block_t cache_size = ca->cache_sectors;
+
+		cache->sectors_per_block_shift = -1;
+		cache_size = block_div(cache_size, ca->block_size);
+		cache->cache_size = to_cblock(cache_size);
+	} else {
+		cache->sectors_per_block_shift = __ffs(ca->block_size);
+		cache->cache_size = to_cblock(ca->cache_sectors >> cache->sectors_per_block_shift);
+	}
+
+	r = create_cache_policy(cache, ca, error);
+	if (r)
+		goto bad;
+
+	cache->policy_nr_args = ca->policy_argc;
+	cache->migration_threshold = DEFAULT_MIGRATION_THRESHOLD;
+
+	r = set_config_values(cache, ca->policy_argc, ca->policy_argv);
+	if (r) {
+		*error = "Error setting cache policy's config values";
+		goto bad;
+	}
+
+	cmd = dm_cache_metadata_open(cache->metadata_dev->bdev,
+				     ca->block_size, may_format,
+				     dm_cache_policy_get_hint_size(cache->policy));
+	if (IS_ERR(cmd)) {
+		*error = "Error creating metadata object";
+		r = PTR_ERR(cmd);
+		goto bad;
+	}
+	cache->cmd = cmd;
+
+	if (passthrough_mode(&cache->features)) {
+		bool all_clean;
+
+		r = dm_cache_metadata_all_clean(cache->cmd, &all_clean);
+		if (r) {
+			*error = "dm_cache_metadata_all_clean() failed";
+			goto bad;
+		}
+
+		if (!all_clean) {
+			*error = "Cannot enter passthrough mode unless all blocks are clean";
+			r = -EINVAL;
+			goto bad;
+		}
+	}
+
+	spin_lock_init(&cache->lock);
+	bio_list_init(&cache->deferred_bios);
+	bio_list_init(&cache->deferred_flush_bios);
+	bio_list_init(&cache->deferred_writethrough_bios);
+	INIT_LIST_HEAD(&cache->quiesced_migrations);
+	INIT_LIST_HEAD(&cache->completed_migrations);
+	INIT_LIST_HEAD(&cache->need_commit_migrations);
+	atomic_set(&cache->nr_migrations, 0);
+	init_waitqueue_head(&cache->migration_wait);
+
+	init_waitqueue_head(&cache->quiescing_wait);
+	atomic_set(&cache->quiescing, 0);
+	atomic_set(&cache->quiescing_ack, 0);
+
+	r = -ENOMEM;
+	atomic_set(&cache->nr_dirty, 0);
+	cache->dirty_bitset = alloc_bitset(from_cblock(cache->cache_size));
+	if (!cache->dirty_bitset) {
+		*error = "could not allocate dirty bitset";
+		goto bad;
+	}
+	clear_bitset(cache->dirty_bitset, from_cblock(cache->cache_size));
+
+	cache->discard_nr_blocks = cache->origin_blocks;
+	cache->discard_bitset = alloc_bitset(from_oblock(cache->discard_nr_blocks));
+	if (!cache->discard_bitset) {
+		*error = "could not allocate discard bitset";
+		goto bad;
+	}
+	clear_bitset(cache->discard_bitset, from_oblock(cache->discard_nr_blocks));
+
+	cache->copier = dm_kcopyd_client_create(&dm_kcopyd_throttle);
+	if (IS_ERR(cache->copier)) {
+		*error = "could not create kcopyd client";
+		r = PTR_ERR(cache->copier);
+		goto bad;
+	}
+
+	cache->wq = alloc_ordered_workqueue("dm-" DM_MSG_PREFIX, WQ_MEM_RECLAIM);
+	if (!cache->wq) {
+		*error = "could not create workqueue for metadata object";
+		goto bad;
+	}
+	INIT_WORK(&cache->worker, do_worker);
+	INIT_DELAYED_WORK(&cache->waker, do_waker);
+	cache->last_commit_jiffies = jiffies;
+
+	cache->prison = dm_bio_prison_create(PRISON_CELLS);
+	if (!cache->prison) {
+		*error = "could not create bio prison";
+		goto bad;
+	}
+
+	cache->all_io_ds = dm_deferred_set_create();
+	if (!cache->all_io_ds) {
+		*error = "could not create all_io deferred set";
+		goto bad;
+	}
+
+	cache->migration_pool = mempool_create_slab_pool(MIGRATION_POOL_SIZE,
+							 migration_cache);
+	if (!cache->migration_pool) {
+		*error = "Error creating cache's migration mempool";
+		goto bad;
+	}
+
+	cache->next_migration = NULL;
+
+	cache->need_tick_bio = true;
+	cache->sized = false;
+	cache->invalidate = false;
+	cache->commit_requested = false;
+	cache->loaded_mappings = false;
+	cache->loaded_discards = false;
+
+	load_stats(cache);
+
+	atomic_set(&cache->stats.demotion, 0);
+	atomic_set(&cache->stats.promotion, 0);
+	atomic_set(&cache->stats.copies_avoided, 0);
+	atomic_set(&cache->stats.cache_cell_clash, 0);
+	atomic_set(&cache->stats.commit_count, 0);
+	atomic_set(&cache->stats.discard_count, 0);
+
+	spin_lock_init(&cache->invalidation_lock);
+	INIT_LIST_HEAD(&cache->invalidation_requests);
+
+	*result = cache;
+	return 0;
+
+bad:
+	destroy(cache);
+	return r;
+}
+
+static int copy_ctr_args(struct cache *cache, int argc, const char **argv)
+{
+	unsigned i;
+	const char **copy;
+
+	copy = kcalloc(argc, sizeof(*copy), GFP_KERNEL);
+	if (!copy)
+		return -ENOMEM;
+	for (i = 0; i < argc; i++) {
+		copy[i] = kstrdup(argv[i], GFP_KERNEL);
+		if (!copy[i]) {
+			while (i--)
+				kfree(copy[i]);
+			kfree(copy);
+			return -ENOMEM;
+		}
+	}
+
+	cache->nr_ctr_args = argc;
+	cache->ctr_args = copy;
+
+	return 0;
+}
+
+static int cache_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int r = -EINVAL;
+	struct cache_args *ca;
+	struct cache *cache = NULL;
+
+	ca = kzalloc(sizeof(*ca), GFP_KERNEL);
+	if (!ca) {
+		ti->error = "Error allocating memory for cache";
+		return -ENOMEM;
+	}
+	ca->ti = ti;
+
+	r = parse_cache_args(ca, argc, argv, &ti->error);
+	if (r)
+		goto out;
+
+	r = cache_create(ca, &cache);
+	if (r)
+		goto out;
+
+	r = copy_ctr_args(cache, argc - 3, (const char **)argv + 3);
+	if (r) {
+		destroy(cache);
+		goto out;
+	}
+
+	ti->private = cache;
+
+out:
+	destroy_cache_args(ca);
+	return r;
+}
+
+static int cache_map(struct dm_target *ti, struct bio *bio)
+{
+	struct cache *cache = ti->private;
+
+	int r;
+	dm_oblock_t block = get_bio_block(cache, bio);
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	bool can_migrate = false;
+	bool discarded_block;
+	struct dm_bio_prison_cell *cell;
+	struct policy_result lookup_result;
+	struct per_bio_data *pb = init_per_bio_data(bio, pb_data_size);
+
+	if (unlikely(from_oblock(block) >= from_oblock(cache->origin_blocks))) {
+		/*
+		 * This can only occur if the io goes to a partial block at
+		 * the end of the origin device.  We don't cache these.
+		 * Just remap to the origin and carry on.
+		 */
+		remap_to_origin(cache, bio);
+		return DM_MAPIO_REMAPPED;
+	}
+
+	if (bio->bi_rw & (REQ_FLUSH | REQ_FUA | REQ_DISCARD)) {
+		defer_bio(cache, bio);
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	/*
+	 * Check to see if that block is currently migrating.
+	 */
+	cell = alloc_prison_cell(cache);
+	if (!cell) {
+		defer_bio(cache, bio);
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	r = bio_detain(cache, block, bio, cell,
+		       (cell_free_fn) free_prison_cell,
+		       cache, &cell);
+	if (r) {
+		if (r < 0)
+			defer_bio(cache, bio);
+
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	discarded_block = is_discarded_oblock(cache, block);
+
+	r = policy_map(cache->policy, block, false, can_migrate, discarded_block,
+		       bio, &lookup_result);
+	if (r == -EWOULDBLOCK) {
+		cell_defer(cache, cell, true);
+		return DM_MAPIO_SUBMITTED;
+
+	} else if (r) {
+		DMERR_LIMIT("Unexpected return from cache replacement policy: %d", r);
+		bio_io_error(bio);
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	r = DM_MAPIO_REMAPPED;
+	switch (lookup_result.op) {
+	case POLICY_HIT:
+		if (passthrough_mode(&cache->features)) {
+			if (bio_data_dir(bio) == WRITE) {
+				/*
+				 * We need to invalidate this block, so
+				 * defer for the worker thread.
+				 */
+				cell_defer(cache, cell, true);
+				r = DM_MAPIO_SUBMITTED;
+
+			} else {
+				pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+				inc_miss_counter(cache, bio);
+				remap_to_origin_clear_discard(cache, bio, block);
+
+				cell_defer(cache, cell, false);
+			}
+
+		} else {
+			inc_hit_counter(cache, bio);
+			pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+
+			if (bio_data_dir(bio) == WRITE && writethrough_mode(&cache->features) &&
+			    !is_dirty(cache, lookup_result.cblock))
+				remap_to_origin_then_cache(cache, bio, block, lookup_result.cblock);
+			else
+				remap_to_cache_dirty(cache, bio, block, lookup_result.cblock);
+
+			cell_defer(cache, cell, false);
+		}
+		break;
+
+	case POLICY_MISS:
+		inc_miss_counter(cache, bio);
+		pb->all_io_entry = dm_deferred_entry_inc(cache->all_io_ds);
+
+		if (pb->req_nr != 0) {
+			/*
+			 * This is a duplicate writethrough io that is no
+			 * longer needed because the block has been demoted.
+			 */
+			bio_endio(bio, 0);
+			cell_defer(cache, cell, false);
+			return DM_MAPIO_SUBMITTED;
+		} else {
+			remap_to_origin_clear_discard(cache, bio, block);
+			cell_defer(cache, cell, false);
+		}
+		break;
+
+	default:
+		DMERR_LIMIT("%s: erroring bio: unknown policy op: %u", __func__,
+			    (unsigned) lookup_result.op);
+		bio_io_error(bio);
+		r = DM_MAPIO_SUBMITTED;
+	}
+
+	return r;
+}
+
+static int cache_end_io(struct dm_target *ti, struct bio *bio, int error)
+{
+	struct cache *cache = ti->private;
+	unsigned long flags;
+	size_t pb_data_size = get_per_bio_data_size(cache);
+	struct per_bio_data *pb = get_per_bio_data(bio, pb_data_size);
+
+	if (pb->tick) {
+		policy_tick(cache->policy);
+
+		spin_lock_irqsave(&cache->lock, flags);
+		cache->need_tick_bio = true;
+		spin_unlock_irqrestore(&cache->lock, flags);
+	}
+
+	check_for_quiesced_migrations(cache, pb);
+
+	return 0;
+}
+
+static int write_dirty_bitset(struct cache *cache)
+{
+	unsigned i, r;
+
+	for (i = 0; i < from_cblock(cache->cache_size); i++) {
+		r = dm_cache_set_dirty(cache->cmd, to_cblock(i),
+				       is_dirty(cache, to_cblock(i)));
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int write_discard_bitset(struct cache *cache)
+{
+	unsigned i, r;
+
+	r = dm_cache_discard_bitset_resize(cache->cmd, cache->sectors_per_block,
+					   cache->origin_blocks);
+	if (r) {
+		DMERR("could not resize on-disk discard bitset");
+		return r;
+	}
+
+	for (i = 0; i < from_oblock(cache->discard_nr_blocks); i++) {
+		r = dm_cache_set_discard(cache->cmd, to_oblock(i),
+					 is_discarded(cache, to_oblock(i)));
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/*
+ * returns true on success
+ */
+static bool sync_metadata(struct cache *cache)
+{
+	int r1, r2, r3, r4;
+
+	r1 = write_dirty_bitset(cache);
+	if (r1)
+		DMERR("could not write dirty bitset");
+
+	r2 = write_discard_bitset(cache);
+	if (r2)
+		DMERR("could not write discard bitset");
+
+	save_stats(cache);
+
+	r3 = dm_cache_write_hints(cache->cmd, cache->policy);
+	if (r3)
+		DMERR("could not write hints");
+
+	/*
+	 * If writing the above metadata failed, we still commit, but don't
+	 * set the clean shutdown flag.  This will effectively force every
+	 * dirty bit to be set on reload.
+	 */
+	r4 = dm_cache_commit(cache->cmd, !r1 && !r2 && !r3);
+	if (r4)
+		DMERR("could not write cache metadata.  Data loss may occur.");
+
+	return !r1 && !r2 && !r3 && !r4;
+}
+
+static void cache_postsuspend(struct dm_target *ti)
+{
+	struct cache *cache = ti->private;
+
+	start_quiescing(cache);
+	wait_for_migrations(cache);
+	stop_worker(cache);
+	requeue_deferred_io(cache);
+	stop_quiescing(cache);
+
+	(void) sync_metadata(cache);
+}
+
+static int load_mapping(void *context, dm_oblock_t oblock, dm_cblock_t cblock,
+			bool dirty, uint32_t hint, bool hint_valid)
+{
+	int r;
+	struct cache *cache = context;
+
+	r = policy_load_mapping(cache->policy, oblock, cblock, hint, hint_valid);
+	if (r)
+		return r;
+
+	if (dirty)
+		set_dirty(cache, oblock, cblock);
+	else
+		clear_dirty(cache, oblock, cblock);
+
+	return 0;
+}
+
+static int load_discard(void *context, sector_t discard_block_size,
+			dm_oblock_t oblock, bool discard)
+{
+	struct cache *cache = context;
+
+	if (discard)
+		set_discard(cache, oblock);
+	else
+		clear_discard(cache, oblock);
+
+	return 0;
+}
+
+static dm_cblock_t get_cache_dev_size(struct cache *cache)
+{
+	sector_t size = get_dev_size(cache->cache_dev);
+	(void) sector_div(size, cache->sectors_per_block);
+	return to_cblock(size);
+}
+
+static bool can_resize(struct cache *cache, dm_cblock_t new_size)
+{
+	if (from_cblock(new_size) > from_cblock(cache->cache_size))
+		return true;
+
+	/*
+	 * We can't drop a dirty block when shrinking the cache.
+	 */
+	while (from_cblock(new_size) < from_cblock(cache->cache_size)) {
+		new_size = to_cblock(from_cblock(new_size) + 1);
+		if (is_dirty(cache, new_size)) {
+			DMERR("unable to shrink cache; cache block %llu is dirty",
+			      (unsigned long long) from_cblock(new_size));
+			return false;
+		}
+	}
+
+	return true;
+}
+
+static int resize_cache_dev(struct cache *cache, dm_cblock_t new_size)
+{
+	int r;
+
+	r = dm_cache_resize(cache->cmd, new_size);
+	if (r) {
+		DMERR("could not resize cache metadata");
+		return r;
+	}
+
+	cache->cache_size = new_size;
+
+	return 0;
+}
+
+static int cache_preresume(struct dm_target *ti)
+{
+	int r = 0;
+	struct cache *cache = ti->private;
+	dm_cblock_t csize = get_cache_dev_size(cache);
+
+	/*
+	 * Check to see if the cache has resized.
+	 */
+	if (!cache->sized) {
+		r = resize_cache_dev(cache, csize);
+		if (r)
+			return r;
+
+		cache->sized = true;
+
+	} else if (csize != cache->cache_size) {
+		if (!can_resize(cache, csize))
+			return -EINVAL;
+
+		r = resize_cache_dev(cache, csize);
+		if (r)
+			return r;
+	}
+
+	if (!cache->loaded_mappings) {
+		r = dm_cache_load_mappings(cache->cmd, cache->policy,
+					   load_mapping, cache);
+		if (r) {
+			DMERR("could not load cache mappings");
+			return r;
+		}
+
+		cache->loaded_mappings = true;
+	}
+
+	if (!cache->loaded_discards) {
+		r = dm_cache_load_discards(cache->cmd, load_discard, cache);
+		if (r) {
+			DMERR("could not load origin discards");
+			return r;
+		}
+
+		cache->loaded_discards = true;
+	}
+
+	return r;
+}
+
+static void cache_resume(struct dm_target *ti)
+{
+	struct cache *cache = ti->private;
+
+	cache->need_tick_bio = true;
+	do_waker(&cache->waker.work);
+}
+
+/*
+ * Status format:
+ *
+ * <metadata block size> <#used metadata blocks>/<#total metadata blocks>
+ * <cache block size> <#used cache blocks>/<#total cache blocks>
+ * <#read hits> <#read misses> <#write hits> <#write misses>
+ * <#demotions> <#promotions> <#dirty>
+ * <#features> <features>*
+ * <#core args> <core args>
+ * <policy name> <#policy args> <policy args>*
+ */
+static void cache_status(struct dm_target *ti, status_type_t type,
+			 unsigned status_flags, char *result, unsigned maxlen)
+{
+	int r = 0;
+	unsigned i;
+	ssize_t sz = 0;
+	dm_block_t nr_free_blocks_metadata = 0;
+	dm_block_t nr_blocks_metadata = 0;
+	char buf[BDEVNAME_SIZE];
+	struct cache *cache = ti->private;
+	dm_cblock_t residency;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		/* Commit to ensure statistics aren't out-of-date */
+		if (!(status_flags & DM_STATUS_NOFLUSH_FLAG) && !dm_suspended(ti)) {
+			r = dm_cache_commit(cache->cmd, false);
+			if (r)
+				DMERR("could not commit metadata for accurate status");
+		}
+
+		r = dm_cache_get_free_metadata_block_count(cache->cmd,
+							   &nr_free_blocks_metadata);
+		if (r) {
+			DMERR("could not get metadata free block count");
+			goto err;
+		}
+
+		r = dm_cache_get_metadata_dev_size(cache->cmd, &nr_blocks_metadata);
+		if (r) {
+			DMERR("could not get metadata device size");
+			goto err;
+		}
+
+		residency = policy_residency(cache->policy);
+
+		DMEMIT("%u %llu/%llu %u %llu/%llu %u %u %u %u %u %u %lu ",
+		       (unsigned)(DM_CACHE_METADATA_BLOCK_SIZE >> SECTOR_SHIFT),
+		       (unsigned long long)(nr_blocks_metadata - nr_free_blocks_metadata),
+		       (unsigned long long)nr_blocks_metadata,
+		       cache->sectors_per_block,
+		       (unsigned long long) from_cblock(residency),
+		       (unsigned long long) from_cblock(cache->cache_size),
+		       (unsigned) atomic_read(&cache->stats.read_hit),
+		       (unsigned) atomic_read(&cache->stats.read_miss),
+		       (unsigned) atomic_read(&cache->stats.write_hit),
+		       (unsigned) atomic_read(&cache->stats.write_miss),
+		       (unsigned) atomic_read(&cache->stats.demotion),
+		       (unsigned) atomic_read(&cache->stats.promotion),
+		       (unsigned long) atomic_read(&cache->nr_dirty));
+
+		if (writethrough_mode(&cache->features))
+			DMEMIT("1 writethrough ");
+
+		else if (passthrough_mode(&cache->features))
+			DMEMIT("1 passthrough ");
+
+		else if (writeback_mode(&cache->features))
+			DMEMIT("1 writeback ");
+
+		else {
+			DMERR("internal error: unknown io mode: %d", (int) cache->features.io_mode);
+			goto err;
+		}
+
+		DMEMIT("2 migration_threshold %llu ", (unsigned long long) cache->migration_threshold);
+
+		DMEMIT("%s ", dm_cache_policy_get_name(cache->policy));
+		if (sz < maxlen) {
+			r = policy_emit_config_values(cache->policy, result + sz, maxlen - sz);
+			if (r)
+				DMERR("policy_emit_config_values returned %d", r);
+		}
+
+		break;
+
+	case STATUSTYPE_TABLE:
+		format_dev_t(buf, cache->metadata_dev->bdev->bd_dev);
+		DMEMIT("%s ", buf);
+		format_dev_t(buf, cache->cache_dev->bdev->bd_dev);
+		DMEMIT("%s ", buf);
+		format_dev_t(buf, cache->origin_dev->bdev->bd_dev);
+		DMEMIT("%s", buf);
+
+		for (i = 0; i < cache->nr_ctr_args - 1; i++)
+			DMEMIT(" %s", cache->ctr_args[i]);
+		if (cache->nr_ctr_args)
+			DMEMIT(" %s", cache->ctr_args[cache->nr_ctr_args - 1]);
+	}
+
+	return;
+
+err:
+	DMEMIT("Error");
+}
+
+/*
+ * A cache block range can take two forms:
+ *
+ * i) A single cblock, eg. '3456'
+ * ii) A begin and end cblock with dots between, eg. 123-234
+ */
+static int parse_cblock_range(struct cache *cache, const char *str,
+			      struct cblock_range *result)
+{
+	char dummy;
+	uint64_t b, e;
+	int r;
+
+	/*
+	 * Try and parse form (ii) first.
+	 */
+	r = sscanf(str, "%llu-%llu%c", &b, &e, &dummy);
+	if (r < 0)
+		return r;
+
+	if (r == 2) {
+		result->begin = to_cblock(b);
+		result->end = to_cblock(e);
+		return 0;
+	}
+
+	/*
+	 * That didn't work, try form (i).
+	 */
+	r = sscanf(str, "%llu%c", &b, &dummy);
+	if (r < 0)
+		return r;
+
+	if (r == 1) {
+		result->begin = to_cblock(b);
+		result->end = to_cblock(from_cblock(result->begin) + 1u);
+		return 0;
+	}
+
+	DMERR("invalid cblock range '%s'", str);
+	return -EINVAL;
+}
+
+static int validate_cblock_range(struct cache *cache, struct cblock_range *range)
+{
+	uint64_t b = from_cblock(range->begin);
+	uint64_t e = from_cblock(range->end);
+	uint64_t n = from_cblock(cache->cache_size);
+
+	if (b >= n) {
+		DMERR("begin cblock out of range: %llu >= %llu", b, n);
+		return -EINVAL;
+	}
+
+	if (e > n) {
+		DMERR("end cblock out of range: %llu > %llu", e, n);
+		return -EINVAL;
+	}
+
+	if (b >= e) {
+		DMERR("invalid cblock range: %llu >= %llu", b, e);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int request_invalidation(struct cache *cache, struct cblock_range *range)
+{
+	struct invalidation_request req;
+
+	INIT_LIST_HEAD(&req.list);
+	req.cblocks = range;
+	atomic_set(&req.complete, 0);
+	req.err = 0;
+	init_waitqueue_head(&req.result_wait);
+
+	spin_lock(&cache->invalidation_lock);
+	list_add(&req.list, &cache->invalidation_requests);
+	spin_unlock(&cache->invalidation_lock);
+	wake_worker(cache);
+
+	wait_event(req.result_wait, atomic_read(&req.complete));
+	return req.err;
+}
+
+static int process_invalidate_cblocks_message(struct cache *cache, unsigned count,
+					      const char **cblock_ranges)
+{
+	int r = 0;
+	unsigned i;
+	struct cblock_range range;
+
+	if (!passthrough_mode(&cache->features)) {
+		DMERR("cache has to be in passthrough mode for invalidation");
+		return -EPERM;
+	}
+
+	for (i = 0; i < count; i++) {
+		r = parse_cblock_range(cache, cblock_ranges[i], &range);
+		if (r)
+			break;
+
+		r = validate_cblock_range(cache, &range);
+		if (r)
+			break;
+
+		/*
+		 * Pass begin and end origin blocks to the worker and wake it.
+		 */
+		r = request_invalidation(cache, &range);
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+/*
+ * Supports
+ *	"<key> <value>"
+ * and
+ *     "invalidate_cblocks [(<begin>)|(<begin>-<end>)]*
+ *
+ * The key migration_threshold is supported by the cache target core.
+ */
+static int cache_message(struct dm_target *ti, unsigned argc, char **argv)
+{
+	struct cache *cache = ti->private;
+
+	if (!argc)
+		return -EINVAL;
+
+	if (!strcasecmp(argv[0], "invalidate_cblocks"))
+		return process_invalidate_cblocks_message(cache, argc - 1, (const char **) argv + 1);
+
+	if (argc != 2)
+		return -EINVAL;
+
+	return set_config_value(cache, argv[0], argv[1]);
+}
+
+static int cache_iterate_devices(struct dm_target *ti,
+				 iterate_devices_callout_fn fn, void *data)
+{
+	int r = 0;
+	struct cache *cache = ti->private;
+
+	r = fn(ti, cache->cache_dev, 0, get_dev_size(cache->cache_dev), data);
+	if (!r)
+		r = fn(ti, cache->origin_dev, 0, ti->len, data);
+
+	return r;
+}
+
+/*
+ * We assume I/O is going to the origin (which is the volume
+ * more likely to have restrictions e.g. by being striped).
+ * (Looking up the exact location of the data would be expensive
+ * and could always be out of date by the time the bio is submitted.)
+ */
+static int cache_bvec_merge(struct dm_target *ti,
+			    struct bvec_merge_data *bvm,
+			    struct bio_vec *biovec, int max_size)
+{
+	struct cache *cache = ti->private;
+	struct request_queue *q = bdev_get_queue(cache->origin_dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = cache->origin_dev->bdev;
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static void set_discard_limits(struct cache *cache, struct queue_limits *limits)
+{
+	/*
+	 * FIXME: these limits may be incompatible with the cache device
+	 */
+	limits->max_discard_sectors = cache->sectors_per_block;
+	limits->discard_granularity = cache->sectors_per_block << SECTOR_SHIFT;
+}
+
+static void cache_io_hints(struct dm_target *ti, struct queue_limits *limits)
+{
+	struct cache *cache = ti->private;
+	uint64_t io_opt_sectors = limits->io_opt >> SECTOR_SHIFT;
+
+	/*
+	 * If the system-determined stacked limits are compatible with the
+	 * cache's blocksize (io_opt is a factor) do not override them.
+	 */
+	if (io_opt_sectors < cache->sectors_per_block ||
+	    do_div(io_opt_sectors, cache->sectors_per_block)) {
+		blk_limits_io_min(limits, 0);
+		blk_limits_io_opt(limits, cache->sectors_per_block << SECTOR_SHIFT);
+	}
+	set_discard_limits(cache, limits);
+}
+
+/*----------------------------------------------------------------*/
+
+static struct target_type cache_target = {
+	.name = "cache",
+	.version = {1, 4, 0},
+	.module = THIS_MODULE,
+	.ctr = cache_ctr,
+	.dtr = cache_dtr,
+	.map = cache_map,
+	.end_io = cache_end_io,
+	.postsuspend = cache_postsuspend,
+	.preresume = cache_preresume,
+	.resume = cache_resume,
+	.status = cache_status,
+	.message = cache_message,
+	.iterate_devices = cache_iterate_devices,
+	.merge = cache_bvec_merge,
+	.io_hints = cache_io_hints,
+};
+
+static int __init dm_cache_init(void)
+{
+	int r;
+
+	r = dm_register_target(&cache_target);
+	if (r) {
+		DMERR("cache target registration failed: %d", r);
+		return r;
+	}
+
+	migration_cache = KMEM_CACHE(dm_cache_migration, 0);
+	if (!migration_cache) {
+		dm_unregister_target(&cache_target);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void __exit dm_cache_exit(void)
+{
+	dm_unregister_target(&cache_target);
+	kmem_cache_destroy(migration_cache);
+}
+
+module_init(dm_cache_init);
+module_exit(dm_cache_exit);
+
+MODULE_DESCRIPTION(DM_NAME " cache target");
+MODULE_AUTHOR("Joe Thornber <ejt@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-crypt.c b/drivers/md/dm-crypt.c
index 13956437bc8..4cba2d808af 100644
--- a/drivers/md/dm-crypt.c
+++ b/drivers/md/dm-crypt.c
@@ -1,7 +1,8 @@
 /*
- * Copyright (C) 2003 Christophe Saout <christophe@saout.de>
+ * Copyright (C) 2003 Jana Saout <jana@saout.de>
  * Copyright (C) 2004 Clemens Fruhwirth <clemens@endorphin.org>
- * Copyright (C) 2006-2008 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2006-2009 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2013 Milan Broz <gmazyland@gmail.com>
  *
  * This file is released under the GPL.
  */
@@ -18,15 +19,17 @@
 #include <linux/crypto.h>
 #include <linux/workqueue.h>
 #include <linux/backing-dev.h>
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 #include <linux/scatterlist.h>
 #include <asm/page.h>
 #include <asm/unaligned.h>
+#include <crypto/hash.h>
+#include <crypto/md5.h>
+#include <crypto/algapi.h>
 
-#include "dm.h"
+#include <linux/device-mapper.h>
 
 #define DM_MSG_PREFIX "crypt"
-#define MESG_STR(x) x, sizeof(x)
 
 /*
  * context holding the current state of a multi-part conversion
@@ -35,32 +38,34 @@ struct convert_context {
 	struct completion restart;
 	struct bio *bio_in;
 	struct bio *bio_out;
-	unsigned int offset_in;
-	unsigned int offset_out;
-	unsigned int idx_in;
-	unsigned int idx_out;
-	sector_t sector;
-	atomic_t pending;
+	struct bvec_iter iter_in;
+	struct bvec_iter iter_out;
+	sector_t cc_sector;
+	atomic_t cc_pending;
+	struct ablkcipher_request *req;
 };
 
 /*
  * 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;
 };
 
 struct dm_crypt_request {
+	struct convert_context *ctx;
 	struct scatterlist sg_in;
 	struct scatterlist sg_out;
+	sector_t iv_sector;
 };
 
 struct crypt_config;
@@ -69,8 +74,34 @@ struct crypt_iv_operations {
 	int (*ctr)(struct crypt_config *cc, struct dm_target *ti,
 		   const char *opts);
 	void (*dtr)(struct crypt_config *cc);
-	const char *(*status)(struct crypt_config *cc);
-	int (*generator)(struct crypt_config *cc, u8 *iv, sector_t sector);
+	int (*init)(struct crypt_config *cc);
+	int (*wipe)(struct crypt_config *cc);
+	int (*generator)(struct crypt_config *cc, u8 *iv,
+			 struct dm_crypt_request *dmreq);
+	int (*post)(struct crypt_config *cc, u8 *iv,
+		    struct dm_crypt_request *dmreq);
+};
+
+struct iv_essiv_private {
+	struct crypto_hash *hash_tfm;
+	u8 *salt;
+};
+
+struct iv_benbi_private {
+	int shift;
+};
+
+#define LMK_SEED_SIZE 64 /* hash + 0 */
+struct iv_lmk_private {
+	struct crypto_shash *hash_tfm;
+	u8 *seed;
+};
+
+#define TCW_WHITENING_SIZE 16
+struct iv_tcw_private {
+	struct crypto_shash *crc32_tfm;
+	u8 *iv_seed;
+	u8 *whitening;
 };
 
 /*
@@ -78,6 +109,10 @@ struct crypt_iv_operations {
  * and encrypts / decrypts at the same time.
  */
 enum flags { DM_CRYPT_SUSPENDED, DM_CRYPT_KEY_VALID };
+
+/*
+ * The fields in here must be read only after initialization.
+ */
 struct crypt_config {
 	struct dm_dev *dev;
 	sector_t start;
@@ -93,20 +128,25 @@ struct crypt_config {
 
 	struct workqueue_struct *io_queue;
 	struct workqueue_struct *crypt_queue;
-	wait_queue_head_t writeq;
 
-	/*
-	 * crypto related data
-	 */
+	char *cipher;
+	char *cipher_string;
+
 	struct crypt_iv_operations *iv_gen_ops;
-	char *iv_mode;
 	union {
-		struct crypto_cipher *essiv_tfm;
-		int benbi_shift;
+		struct iv_essiv_private essiv;
+		struct iv_benbi_private benbi;
+		struct iv_lmk_private lmk;
+		struct iv_tcw_private tcw;
 	} iv_gen_private;
 	sector_t iv_offset;
 	unsigned int iv_size;
 
+	/* ESSIV: struct crypto_cipher *essiv_tfm */
+	void *iv_private;
+	struct crypto_ablkcipher **tfms;
+	unsigned tfms_count;
+
 	/*
 	 * Layout of each crypto request:
 	 *
@@ -121,24 +161,30 @@ struct crypt_config {
 	 * correctly aligned.
 	 */
 	unsigned int dmreq_start;
-	struct ablkcipher_request *req;
 
-	char cipher[CRYPTO_MAX_ALG_NAME];
-	char chainmode[CRYPTO_MAX_ALG_NAME];
-	struct crypto_ablkcipher *tfm;
 	unsigned long flags;
 	unsigned int key_size;
+	unsigned int key_parts;      /* independent parts in key buffer */
+	unsigned int key_extra_size; /* additional keys length */
 	u8 key[0];
 };
 
 #define MIN_IOS        16
 #define MIN_POOL_PAGES 32
-#define MIN_BIO_PAGES  8
 
 static struct kmem_cache *_crypt_io_pool;
 
 static void clone_init(struct dm_crypt_io *, struct bio *);
 static void kcryptd_queue_crypt(struct dm_crypt_io *io);
+static u8 *iv_of_dmreq(struct crypt_config *cc, struct dm_crypt_request *dmreq);
+
+/*
+ * Use this to access cipher attributes that are the same for each CPU.
+ */
+static struct crypto_ablkcipher *any_tfm(struct crypt_config *cc)
+{
+	return cc->tfms[0];
+}
 
 /*
  * Different IV generation algorithms:
@@ -146,6 +192,9 @@ static void kcryptd_queue_crypt(struct dm_crypt_io *io);
  * plain: the initial vector is the 32-bit little-endian version of the sector
  *        number, padded with zeros if necessary.
  *
+ * plain64: the initial vector is the 64-bit little-endian version of the sector
+ *        number, padded with zeros if necessary.
+ *
  * essiv: "encrypted sector|salt initial vector", the sector number is
  *        encrypted with the bulk cipher using a salt as key. The salt
  *        should be derived from the bulk cipher's key via hashing.
@@ -156,107 +205,213 @@ static void kcryptd_queue_crypt(struct dm_crypt_io *io);
  * null: the initial vector is always zero.  Provides compatibility with
  *       obsolete loop_fish2 devices.  Do not use for new devices.
  *
+ * lmk:  Compatible implementation of the block chaining mode used
+ *       by the Loop-AES block device encryption system
+ *       designed by Jari Ruusu. See http://loop-aes.sourceforge.net/
+ *       It operates on full 512 byte sectors and uses CBC
+ *       with an IV derived from the sector number, the data and
+ *       optionally extra IV seed.
+ *       This means that after decryption the first block
+ *       of sector must be tweaked according to decrypted data.
+ *       Loop-AES can use three encryption schemes:
+ *         version 1: is plain aes-cbc mode
+ *         version 2: uses 64 multikey scheme with lmk IV generator
+ *         version 3: the same as version 2 with additional IV seed
+ *                   (it uses 65 keys, last key is used as IV seed)
+ *
+ * tcw:  Compatible implementation of the block chaining mode used
+ *       by the TrueCrypt device encryption system (prior to version 4.1).
+ *       For more info see: http://www.truecrypt.org
+ *       It operates on full 512 byte sectors and uses CBC
+ *       with an IV derived from initial key and the sector number.
+ *       In addition, whitening value is applied on every sector, whitening
+ *       is calculated from initial key, sector number and mixed using CRC32.
+ *       Note that this encryption scheme is vulnerable to watermarking attacks
+ *       and should be used for old compatible containers access only.
+ *
  * plumb: unimplemented, see:
  * http://article.gmane.org/gmane.linux.kernel.device-mapper.dm-crypt/454
  */
 
-static int crypt_iv_plain_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
+static int crypt_iv_plain_gen(struct crypt_config *cc, u8 *iv,
+			      struct dm_crypt_request *dmreq)
 {
 	memset(iv, 0, cc->iv_size);
-	*(u32 *)iv = cpu_to_le32(sector & 0xffffffff);
+	*(__le32 *)iv = cpu_to_le32(dmreq->iv_sector & 0xffffffff);
 
 	return 0;
 }
 
-static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
-			      const char *opts)
+static int crypt_iv_plain64_gen(struct crypt_config *cc, u8 *iv,
+				struct dm_crypt_request *dmreq)
 {
-	struct crypto_cipher *essiv_tfm;
-	struct crypto_hash *hash_tfm;
+	memset(iv, 0, cc->iv_size);
+	*(__le64 *)iv = cpu_to_le64(dmreq->iv_sector);
+
+	return 0;
+}
+
+/* Initialise ESSIV - compute salt but no local memory allocations */
+static int crypt_iv_essiv_init(struct crypt_config *cc)
+{
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
 	struct hash_desc desc;
 	struct scatterlist sg;
-	unsigned int saltsize;
-	u8 *salt;
+	struct crypto_cipher *essiv_tfm;
 	int err;
 
-	if (opts == NULL) {
-		ti->error = "Digest algorithm missing for ESSIV mode";
-		return -EINVAL;
-	}
-
-	/* Hash the cipher key with the given hash algorithm */
-	hash_tfm = crypto_alloc_hash(opts, 0, CRYPTO_ALG_ASYNC);
-	if (IS_ERR(hash_tfm)) {
-		ti->error = "Error initializing ESSIV hash";
-		return PTR_ERR(hash_tfm);
-	}
-
-	saltsize = crypto_hash_digestsize(hash_tfm);
-	salt = kmalloc(saltsize, GFP_KERNEL);
-	if (salt == NULL) {
-		ti->error = "Error kmallocing salt storage in ESSIV";
-		crypto_free_hash(hash_tfm);
-		return -ENOMEM;
-	}
-
 	sg_init_one(&sg, cc->key, cc->key_size);
-	desc.tfm = hash_tfm;
+	desc.tfm = essiv->hash_tfm;
 	desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
-	err = crypto_hash_digest(&desc, &sg, cc->key_size, salt);
-	crypto_free_hash(hash_tfm);
 
-	if (err) {
-		ti->error = "Error calculating hash in ESSIV";
-		kfree(salt);
+	err = crypto_hash_digest(&desc, &sg, cc->key_size, essiv->salt);
+	if (err)
 		return err;
-	}
+
+	essiv_tfm = cc->iv_private;
+
+	err = crypto_cipher_setkey(essiv_tfm, essiv->salt,
+			    crypto_hash_digestsize(essiv->hash_tfm));
+	if (err)
+		return err;
+
+	return 0;
+}
+
+/* Wipe salt and reset key derived from volume key */
+static int crypt_iv_essiv_wipe(struct crypt_config *cc)
+{
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
+	unsigned salt_size = crypto_hash_digestsize(essiv->hash_tfm);
+	struct crypto_cipher *essiv_tfm;
+	int r, err = 0;
+
+	memset(essiv->salt, 0, salt_size);
+
+	essiv_tfm = cc->iv_private;
+	r = crypto_cipher_setkey(essiv_tfm, essiv->salt, salt_size);
+	if (r)
+		err = r;
+
+	return err;
+}
+
+/* Set up per cpu cipher state */
+static struct crypto_cipher *setup_essiv_cpu(struct crypt_config *cc,
+					     struct dm_target *ti,
+					     u8 *salt, unsigned saltsize)
+{
+	struct crypto_cipher *essiv_tfm;
+	int err;
 
 	/* Setup the essiv_tfm with the given salt */
 	essiv_tfm = crypto_alloc_cipher(cc->cipher, 0, CRYPTO_ALG_ASYNC);
 	if (IS_ERR(essiv_tfm)) {
 		ti->error = "Error allocating crypto tfm for ESSIV";
-		kfree(salt);
-		return PTR_ERR(essiv_tfm);
+		return essiv_tfm;
 	}
+
 	if (crypto_cipher_blocksize(essiv_tfm) !=
-	    crypto_ablkcipher_ivsize(cc->tfm)) {
+	    crypto_ablkcipher_ivsize(any_tfm(cc))) {
 		ti->error = "Block size of ESSIV cipher does "
 			    "not match IV size of block cipher";
 		crypto_free_cipher(essiv_tfm);
-		kfree(salt);
-		return -EINVAL;
+		return ERR_PTR(-EINVAL);
 	}
+
 	err = crypto_cipher_setkey(essiv_tfm, salt, saltsize);
 	if (err) {
 		ti->error = "Failed to set key for ESSIV cipher";
 		crypto_free_cipher(essiv_tfm);
-		kfree(salt);
-		return err;
+		return ERR_PTR(err);
 	}
-	kfree(salt);
 
-	cc->iv_gen_private.essiv_tfm = essiv_tfm;
-	return 0;
+	return essiv_tfm;
 }
 
 static void crypt_iv_essiv_dtr(struct crypt_config *cc)
 {
-	crypto_free_cipher(cc->iv_gen_private.essiv_tfm);
-	cc->iv_gen_private.essiv_tfm = NULL;
+	struct crypto_cipher *essiv_tfm;
+	struct iv_essiv_private *essiv = &cc->iv_gen_private.essiv;
+
+	crypto_free_hash(essiv->hash_tfm);
+	essiv->hash_tfm = NULL;
+
+	kzfree(essiv->salt);
+	essiv->salt = NULL;
+
+	essiv_tfm = cc->iv_private;
+
+	if (essiv_tfm)
+		crypto_free_cipher(essiv_tfm);
+
+	cc->iv_private = NULL;
 }
 
-static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
+static int crypt_iv_essiv_ctr(struct crypt_config *cc, struct dm_target *ti,
+			      const char *opts)
 {
+	struct crypto_cipher *essiv_tfm = NULL;
+	struct crypto_hash *hash_tfm = NULL;
+	u8 *salt = NULL;
+	int err;
+
+	if (!opts) {
+		ti->error = "Digest algorithm missing for ESSIV mode";
+		return -EINVAL;
+	}
+
+	/* Allocate hash algorithm */
+	hash_tfm = crypto_alloc_hash(opts, 0, CRYPTO_ALG_ASYNC);
+	if (IS_ERR(hash_tfm)) {
+		ti->error = "Error initializing ESSIV hash";
+		err = PTR_ERR(hash_tfm);
+		goto bad;
+	}
+
+	salt = kzalloc(crypto_hash_digestsize(hash_tfm), GFP_KERNEL);
+	if (!salt) {
+		ti->error = "Error kmallocing salt storage in ESSIV";
+		err = -ENOMEM;
+		goto bad;
+	}
+
+	cc->iv_gen_private.essiv.salt = salt;
+	cc->iv_gen_private.essiv.hash_tfm = hash_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;
+
+bad:
+	if (hash_tfm && !IS_ERR(hash_tfm))
+		crypto_free_hash(hash_tfm);
+	kfree(salt);
+	return err;
+}
+
+static int crypt_iv_essiv_gen(struct crypt_config *cc, u8 *iv,
+			      struct dm_crypt_request *dmreq)
+{
+	struct crypto_cipher *essiv_tfm = cc->iv_private;
+
 	memset(iv, 0, cc->iv_size);
-	*(u64 *)iv = cpu_to_le64(sector);
-	crypto_cipher_encrypt_one(cc->iv_gen_private.essiv_tfm, iv, iv);
+	*(__le64 *)iv = cpu_to_le64(dmreq->iv_sector);
+	crypto_cipher_encrypt_one(essiv_tfm, iv, iv);
+
 	return 0;
 }
 
 static int crypt_iv_benbi_ctr(struct crypt_config *cc, struct dm_target *ti,
 			      const char *opts)
 {
-	unsigned bs = crypto_ablkcipher_blocksize(cc->tfm);
+	unsigned bs = crypto_ablkcipher_blocksize(any_tfm(cc));
 	int log = ilog2(bs);
 
 	/* we need to calculate how far we must shift the sector count
@@ -272,7 +427,7 @@ static int crypt_iv_benbi_ctr(struct crypt_config *cc, struct dm_target *ti,
 		return -EINVAL;
 	}
 
-	cc->iv_gen_private.benbi_shift = 9 - log;
+	cc->iv_gen_private.benbi.shift = 9 - log;
 
 	return 0;
 }
@@ -281,32 +436,337 @@ static void crypt_iv_benbi_dtr(struct crypt_config *cc)
 {
 }
 
-static int crypt_iv_benbi_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
+static int crypt_iv_benbi_gen(struct crypt_config *cc, u8 *iv,
+			      struct dm_crypt_request *dmreq)
 {
 	__be64 val;
 
 	memset(iv, 0, cc->iv_size - sizeof(u64)); /* rest is cleared below */
 
-	val = cpu_to_be64(((u64)sector << cc->iv_gen_private.benbi_shift) + 1);
+	val = cpu_to_be64(((u64)dmreq->iv_sector << cc->iv_gen_private.benbi.shift) + 1);
 	put_unaligned(val, (__be64 *)(iv + cc->iv_size - sizeof(u64)));
 
 	return 0;
 }
 
-static int crypt_iv_null_gen(struct crypt_config *cc, u8 *iv, sector_t sector)
+static int crypt_iv_null_gen(struct crypt_config *cc, u8 *iv,
+			     struct dm_crypt_request *dmreq)
 {
 	memset(iv, 0, cc->iv_size);
 
 	return 0;
 }
 
+static void crypt_iv_lmk_dtr(struct crypt_config *cc)
+{
+	struct iv_lmk_private *lmk = &cc->iv_gen_private.lmk;
+
+	if (lmk->hash_tfm && !IS_ERR(lmk->hash_tfm))
+		crypto_free_shash(lmk->hash_tfm);
+	lmk->hash_tfm = NULL;
+
+	kzfree(lmk->seed);
+	lmk->seed = NULL;
+}
+
+static int crypt_iv_lmk_ctr(struct crypt_config *cc, struct dm_target *ti,
+			    const char *opts)
+{
+	struct iv_lmk_private *lmk = &cc->iv_gen_private.lmk;
+
+	lmk->hash_tfm = crypto_alloc_shash("md5", 0, 0);
+	if (IS_ERR(lmk->hash_tfm)) {
+		ti->error = "Error initializing LMK hash";
+		return PTR_ERR(lmk->hash_tfm);
+	}
+
+	/* No seed in LMK version 2 */
+	if (cc->key_parts == cc->tfms_count) {
+		lmk->seed = NULL;
+		return 0;
+	}
+
+	lmk->seed = kzalloc(LMK_SEED_SIZE, GFP_KERNEL);
+	if (!lmk->seed) {
+		crypt_iv_lmk_dtr(cc);
+		ti->error = "Error kmallocing seed storage in LMK";
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int crypt_iv_lmk_init(struct crypt_config *cc)
+{
+	struct iv_lmk_private *lmk = &cc->iv_gen_private.lmk;
+	int subkey_size = cc->key_size / cc->key_parts;
+
+	/* LMK seed is on the position of LMK_KEYS + 1 key */
+	if (lmk->seed)
+		memcpy(lmk->seed, cc->key + (cc->tfms_count * subkey_size),
+		       crypto_shash_digestsize(lmk->hash_tfm));
+
+	return 0;
+}
+
+static int crypt_iv_lmk_wipe(struct crypt_config *cc)
+{
+	struct iv_lmk_private *lmk = &cc->iv_gen_private.lmk;
+
+	if (lmk->seed)
+		memset(lmk->seed, 0, LMK_SEED_SIZE);
+
+	return 0;
+}
+
+static int crypt_iv_lmk_one(struct crypt_config *cc, u8 *iv,
+			    struct dm_crypt_request *dmreq,
+			    u8 *data)
+{
+	struct iv_lmk_private *lmk = &cc->iv_gen_private.lmk;
+	struct {
+		struct shash_desc desc;
+		char ctx[crypto_shash_descsize(lmk->hash_tfm)];
+	} sdesc;
+	struct md5_state md5state;
+	__le32 buf[4];
+	int i, r;
+
+	sdesc.desc.tfm = lmk->hash_tfm;
+	sdesc.desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+
+	r = crypto_shash_init(&sdesc.desc);
+	if (r)
+		return r;
+
+	if (lmk->seed) {
+		r = crypto_shash_update(&sdesc.desc, lmk->seed, LMK_SEED_SIZE);
+		if (r)
+			return r;
+	}
+
+	/* Sector is always 512B, block size 16, add data of blocks 1-31 */
+	r = crypto_shash_update(&sdesc.desc, data + 16, 16 * 31);
+	if (r)
+		return r;
+
+	/* Sector is cropped to 56 bits here */
+	buf[0] = cpu_to_le32(dmreq->iv_sector & 0xFFFFFFFF);
+	buf[1] = cpu_to_le32((((u64)dmreq->iv_sector >> 32) & 0x00FFFFFF) | 0x80000000);
+	buf[2] = cpu_to_le32(4024);
+	buf[3] = 0;
+	r = crypto_shash_update(&sdesc.desc, (u8 *)buf, sizeof(buf));
+	if (r)
+		return r;
+
+	/* No MD5 padding here */
+	r = crypto_shash_export(&sdesc.desc, &md5state);
+	if (r)
+		return r;
+
+	for (i = 0; i < MD5_HASH_WORDS; i++)
+		__cpu_to_le32s(&md5state.hash[i]);
+	memcpy(iv, &md5state.hash, cc->iv_size);
+
+	return 0;
+}
+
+static int crypt_iv_lmk_gen(struct crypt_config *cc, u8 *iv,
+			    struct dm_crypt_request *dmreq)
+{
+	u8 *src;
+	int r = 0;
+
+	if (bio_data_dir(dmreq->ctx->bio_in) == WRITE) {
+		src = kmap_atomic(sg_page(&dmreq->sg_in));
+		r = crypt_iv_lmk_one(cc, iv, dmreq, src + dmreq->sg_in.offset);
+		kunmap_atomic(src);
+	} else
+		memset(iv, 0, cc->iv_size);
+
+	return r;
+}
+
+static int crypt_iv_lmk_post(struct crypt_config *cc, u8 *iv,
+			     struct dm_crypt_request *dmreq)
+{
+	u8 *dst;
+	int r;
+
+	if (bio_data_dir(dmreq->ctx->bio_in) == WRITE)
+		return 0;
+
+	dst = kmap_atomic(sg_page(&dmreq->sg_out));
+	r = crypt_iv_lmk_one(cc, iv, dmreq, dst + dmreq->sg_out.offset);
+
+	/* Tweak the first block of plaintext sector */
+	if (!r)
+		crypto_xor(dst + dmreq->sg_out.offset, iv, cc->iv_size);
+
+	kunmap_atomic(dst);
+	return r;
+}
+
+static void crypt_iv_tcw_dtr(struct crypt_config *cc)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+
+	kzfree(tcw->iv_seed);
+	tcw->iv_seed = NULL;
+	kzfree(tcw->whitening);
+	tcw->whitening = NULL;
+
+	if (tcw->crc32_tfm && !IS_ERR(tcw->crc32_tfm))
+		crypto_free_shash(tcw->crc32_tfm);
+	tcw->crc32_tfm = NULL;
+}
+
+static int crypt_iv_tcw_ctr(struct crypt_config *cc, struct dm_target *ti,
+			    const char *opts)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+
+	if (cc->key_size <= (cc->iv_size + TCW_WHITENING_SIZE)) {
+		ti->error = "Wrong key size for TCW";
+		return -EINVAL;
+	}
+
+	tcw->crc32_tfm = crypto_alloc_shash("crc32", 0, 0);
+	if (IS_ERR(tcw->crc32_tfm)) {
+		ti->error = "Error initializing CRC32 in TCW";
+		return PTR_ERR(tcw->crc32_tfm);
+	}
+
+	tcw->iv_seed = kzalloc(cc->iv_size, GFP_KERNEL);
+	tcw->whitening = kzalloc(TCW_WHITENING_SIZE, GFP_KERNEL);
+	if (!tcw->iv_seed || !tcw->whitening) {
+		crypt_iv_tcw_dtr(cc);
+		ti->error = "Error allocating seed storage in TCW";
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int crypt_iv_tcw_init(struct crypt_config *cc)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+	int key_offset = cc->key_size - cc->iv_size - TCW_WHITENING_SIZE;
+
+	memcpy(tcw->iv_seed, &cc->key[key_offset], cc->iv_size);
+	memcpy(tcw->whitening, &cc->key[key_offset + cc->iv_size],
+	       TCW_WHITENING_SIZE);
+
+	return 0;
+}
+
+static int crypt_iv_tcw_wipe(struct crypt_config *cc)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+
+	memset(tcw->iv_seed, 0, cc->iv_size);
+	memset(tcw->whitening, 0, TCW_WHITENING_SIZE);
+
+	return 0;
+}
+
+static int crypt_iv_tcw_whitening(struct crypt_config *cc,
+				  struct dm_crypt_request *dmreq,
+				  u8 *data)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+	u64 sector = cpu_to_le64((u64)dmreq->iv_sector);
+	u8 buf[TCW_WHITENING_SIZE];
+	struct {
+		struct shash_desc desc;
+		char ctx[crypto_shash_descsize(tcw->crc32_tfm)];
+	} sdesc;
+	int i, r;
+
+	/* xor whitening with sector number */
+	memcpy(buf, tcw->whitening, TCW_WHITENING_SIZE);
+	crypto_xor(buf, (u8 *)&sector, 8);
+	crypto_xor(&buf[8], (u8 *)&sector, 8);
+
+	/* calculate crc32 for every 32bit part and xor it */
+	sdesc.desc.tfm = tcw->crc32_tfm;
+	sdesc.desc.flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+	for (i = 0; i < 4; i++) {
+		r = crypto_shash_init(&sdesc.desc);
+		if (r)
+			goto out;
+		r = crypto_shash_update(&sdesc.desc, &buf[i * 4], 4);
+		if (r)
+			goto out;
+		r = crypto_shash_final(&sdesc.desc, &buf[i * 4]);
+		if (r)
+			goto out;
+	}
+	crypto_xor(&buf[0], &buf[12], 4);
+	crypto_xor(&buf[4], &buf[8], 4);
+
+	/* apply whitening (8 bytes) to whole sector */
+	for (i = 0; i < ((1 << SECTOR_SHIFT) / 8); i++)
+		crypto_xor(data + i * 8, buf, 8);
+out:
+	memset(buf, 0, sizeof(buf));
+	return r;
+}
+
+static int crypt_iv_tcw_gen(struct crypt_config *cc, u8 *iv,
+			    struct dm_crypt_request *dmreq)
+{
+	struct iv_tcw_private *tcw = &cc->iv_gen_private.tcw;
+	u64 sector = cpu_to_le64((u64)dmreq->iv_sector);
+	u8 *src;
+	int r = 0;
+
+	/* Remove whitening from ciphertext */
+	if (bio_data_dir(dmreq->ctx->bio_in) != WRITE) {
+		src = kmap_atomic(sg_page(&dmreq->sg_in));
+		r = crypt_iv_tcw_whitening(cc, dmreq, src + dmreq->sg_in.offset);
+		kunmap_atomic(src);
+	}
+
+	/* Calculate IV */
+	memcpy(iv, tcw->iv_seed, cc->iv_size);
+	crypto_xor(iv, (u8 *)&sector, 8);
+	if (cc->iv_size > 8)
+		crypto_xor(&iv[8], (u8 *)&sector, cc->iv_size - 8);
+
+	return r;
+}
+
+static int crypt_iv_tcw_post(struct crypt_config *cc, u8 *iv,
+			     struct dm_crypt_request *dmreq)
+{
+	u8 *dst;
+	int r;
+
+	if (bio_data_dir(dmreq->ctx->bio_in) != WRITE)
+		return 0;
+
+	/* Apply whitening on ciphertext */
+	dst = kmap_atomic(sg_page(&dmreq->sg_out));
+	r = crypt_iv_tcw_whitening(cc, dmreq, dst + dmreq->sg_out.offset);
+	kunmap_atomic(dst);
+
+	return r;
+}
+
 static struct crypt_iv_operations crypt_iv_plain_ops = {
 	.generator = crypt_iv_plain_gen
 };
 
+static struct crypt_iv_operations crypt_iv_plain64_ops = {
+	.generator = crypt_iv_plain64_gen
+};
+
 static struct crypt_iv_operations crypt_iv_essiv_ops = {
 	.ctr       = crypt_iv_essiv_ctr,
 	.dtr       = crypt_iv_essiv_dtr,
+	.init      = crypt_iv_essiv_init,
+	.wipe      = crypt_iv_essiv_wipe,
 	.generator = crypt_iv_essiv_gen
 };
 
@@ -320,6 +780,24 @@ static struct crypt_iv_operations crypt_iv_null_ops = {
 	.generator = crypt_iv_null_gen
 };
 
+static struct crypt_iv_operations crypt_iv_lmk_ops = {
+	.ctr	   = crypt_iv_lmk_ctr,
+	.dtr	   = crypt_iv_lmk_dtr,
+	.init	   = crypt_iv_lmk_init,
+	.wipe	   = crypt_iv_lmk_wipe,
+	.generator = crypt_iv_lmk_gen,
+	.post	   = crypt_iv_lmk_post
+};
+
+static struct crypt_iv_operations crypt_iv_tcw_ops = {
+	.ctr	   = crypt_iv_tcw_ctr,
+	.dtr	   = crypt_iv_tcw_dtr,
+	.init	   = crypt_iv_tcw_init,
+	.wipe	   = crypt_iv_tcw_wipe,
+	.generator = crypt_iv_tcw_gen,
+	.post	   = crypt_iv_tcw_post
+};
+
 static void crypt_convert_init(struct crypt_config *cc,
 			       struct convert_context *ctx,
 			       struct bio *bio_out, struct bio *bio_in,
@@ -327,51 +805,61 @@ static void crypt_convert_init(struct crypt_config *cc,
 {
 	ctx->bio_in = bio_in;
 	ctx->bio_out = bio_out;
-	ctx->offset_in = 0;
-	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;
+	if (bio_in)
+		ctx->iter_in = bio_in->bi_iter;
+	if (bio_out)
+		ctx->iter_out = bio_out->bi_iter;
+	ctx->cc_sector = sector + cc->iv_offset;
 	init_completion(&ctx->restart);
-	atomic_set(&ctx->pending, 1);
+}
+
+static struct dm_crypt_request *dmreq_of_req(struct crypt_config *cc,
+					     struct ablkcipher_request *req)
+{
+	return (struct dm_crypt_request *)((char *)req + cc->dmreq_start);
+}
+
+static struct ablkcipher_request *req_of_dmreq(struct crypt_config *cc,
+					       struct dm_crypt_request *dmreq)
+{
+	return (struct ablkcipher_request *)((char *)dmreq - cc->dmreq_start);
+}
+
+static u8 *iv_of_dmreq(struct crypt_config *cc,
+		       struct dm_crypt_request *dmreq)
+{
+	return (u8 *)ALIGN((unsigned long)(dmreq + 1),
+		crypto_ablkcipher_alignmask(any_tfm(cc)) + 1);
 }
 
 static int crypt_convert_block(struct crypt_config *cc,
 			       struct convert_context *ctx,
 			       struct ablkcipher_request *req)
 {
-	struct bio_vec *bv_in = bio_iovec_idx(ctx->bio_in, ctx->idx_in);
-	struct bio_vec *bv_out = bio_iovec_idx(ctx->bio_out, ctx->idx_out);
+	struct bio_vec bv_in = bio_iter_iovec(ctx->bio_in, ctx->iter_in);
+	struct bio_vec bv_out = bio_iter_iovec(ctx->bio_out, ctx->iter_out);
 	struct dm_crypt_request *dmreq;
 	u8 *iv;
-	int r = 0;
+	int r;
 
-	dmreq = (struct dm_crypt_request *)((char *)req + cc->dmreq_start);
-	iv = (u8 *)ALIGN((unsigned long)(dmreq + 1),
-			 crypto_ablkcipher_alignmask(cc->tfm) + 1);
+	dmreq = dmreq_of_req(cc, req);
+	iv = iv_of_dmreq(cc, dmreq);
 
+	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,
-		    bv_in->bv_offset + ctx->offset_in);
+	sg_set_page(&dmreq->sg_in, bv_in.bv_page, 1 << SECTOR_SHIFT,
+		    bv_in.bv_offset);
 
 	sg_init_table(&dmreq->sg_out, 1);
-	sg_set_page(&dmreq->sg_out, bv_out->bv_page, 1 << SECTOR_SHIFT,
-		    bv_out->bv_offset + ctx->offset_out);
-
-	ctx->offset_in += 1 << SECTOR_SHIFT;
-	if (ctx->offset_in >= bv_in->bv_len) {
-		ctx->offset_in = 0;
-		ctx->idx_in++;
-	}
+	sg_set_page(&dmreq->sg_out, bv_out.bv_page, 1 << SECTOR_SHIFT,
+		    bv_out.bv_offset);
 
-	ctx->offset_out += 1 << SECTOR_SHIFT;
-	if (ctx->offset_out >= bv_out->bv_len) {
-		ctx->offset_out = 0;
-		ctx->idx_out++;
-	}
+	bio_advance_iter(ctx->bio_in, &ctx->iter_in, 1 << SECTOR_SHIFT);
+	bio_advance_iter(ctx->bio_out, &ctx->iter_out, 1 << SECTOR_SHIFT);
 
 	if (cc->iv_gen_ops) {
-		r = cc->iv_gen_ops->generator(cc, iv, ctx->sector);
+		r = cc->iv_gen_ops->generator(cc, iv, dmreq);
 		if (r < 0)
 			return r;
 	}
@@ -384,20 +872,27 @@ static int crypt_convert_block(struct crypt_config *cc,
 	else
 		r = crypto_ablkcipher_decrypt(req);
 
+	if (!r && cc->iv_gen_ops && cc->iv_gen_ops->post)
+		r = cc->iv_gen_ops->post(cc, iv, dmreq);
+
 	return r;
 }
 
 static void kcryptd_async_done(struct crypto_async_request *async_req,
 			       int error);
+
 static void crypt_alloc_req(struct crypt_config *cc,
 			    struct convert_context *ctx)
 {
-	if (!cc->req)
-		cc->req = mempool_alloc(cc->req_pool, GFP_NOIO);
-	ablkcipher_request_set_tfm(cc->req, cc->tfm);
-	ablkcipher_request_set_callback(cc->req, CRYPTO_TFM_REQ_MAY_BACKLOG |
-					     CRYPTO_TFM_REQ_MAY_SLEEP,
-					     kcryptd_async_done, ctx);
+	unsigned key_index = ctx->cc_sector & (cc->tfms_count - 1);
+
+	if (!ctx->req)
+		ctx->req = mempool_alloc(cc->req_pool, GFP_NOIO);
+
+	ablkcipher_request_set_tfm(ctx->req, cc->tfms[key_index]);
+	ablkcipher_request_set_callback(ctx->req,
+	    CRYPTO_TFM_REQ_MAY_BACKLOG | CRYPTO_TFM_REQ_MAY_SLEEP,
+	    kcryptd_async_done, dmreq_of_req(cc, ctx->req));
 }
 
 /*
@@ -408,36 +903,37 @@ static int crypt_convert(struct crypt_config *cc,
 {
 	int r;
 
-	while(ctx->idx_in < ctx->bio_in->bi_vcnt &&
-	      ctx->idx_out < ctx->bio_out->bi_vcnt) {
+	atomic_set(&ctx->cc_pending, 1);
+
+	while (ctx->iter_in.bi_size && ctx->iter_out.bi_size) {
 
 		crypt_alloc_req(cc, ctx);
 
-		atomic_inc(&ctx->pending);
+		atomic_inc(&ctx->cc_pending);
 
-		r = crypt_convert_block(cc, ctx, cc->req);
+		r = crypt_convert_block(cc, ctx, ctx->req);
 
 		switch (r) {
 		/* async */
 		case -EBUSY:
 			wait_for_completion(&ctx->restart);
-			INIT_COMPLETION(ctx->restart);
+			reinit_completion(&ctx->restart);
 			/* fall through*/
 		case -EINPROGRESS:
-			cc->req = NULL;
-			ctx->sector++;
+			ctx->req = NULL;
+			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;
 		}
 	}
@@ -445,22 +941,16 @@ static int crypt_convert(struct crypt_config *cc,
 	return 0;
 }
 
-static void dm_crypt_bio_destructor(struct bio *bio)
-{
-	struct dm_crypt_io *io = bio->bi_private;
-	struct crypt_config *cc = io->target->private;
-
-	bio_free(bio, cc->bs);
-}
-
 /*
  * Generate a new unfragmented bio with the given size
  * This should never violate the device limitations
- * May return a smaller bio when running out of pages
+ * May return a smaller bio when running out of pages, indicated by
+ * *out_of_pages set to 1.
  */
-static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size)
+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;
@@ -472,19 +962,21 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size)
 		return NULL;
 
 	clone_init(io, clone);
+	*out_of_pages = 0;
 
 	for (i = 0; i < nr_iovecs; i++) {
 		page = mempool_alloc(cc->page_pool, gfp_mask);
-		if (!page)
+		if (!page) {
+			*out_of_pages = 1;
 			break;
+		}
 
 		/*
-		 * if additional pages cannot be allocated without waiting,
-		 * return a partially allocated bio, the caller will then try
-		 * to allocate additional bios while submitting this partial bio
+		 * If additional pages cannot be allocated without waiting,
+		 * return a partially-allocated bio.  The caller will then try
+		 * to allocate more bios while submitting this partial bio.
 		 */
-		if (i == (MIN_BIO_PAGES - 1))
-			gfp_mask = (gfp_mask | __GFP_NOWARN) & ~__GFP_WAIT;
+		gfp_mask = (gfp_mask | __GFP_NOWARN) & ~__GFP_WAIT;
 
 		len = (size > PAGE_SIZE) ? PAGE_SIZE : size;
 
@@ -496,7 +988,7 @@ static struct bio *crypt_alloc_buffer(struct dm_crypt_io *io, unsigned size)
 		size -= len;
 	}
 
-	if (!clone->bi_size) {
+	if (!clone->bi_iter.bi_size) {
 		bio_put(clone);
 		return NULL;
 	}
@@ -509,27 +1001,61 @@ static void crypt_free_buffer_pages(struct crypt_config *cc, struct bio *clone)
 	unsigned int i;
 	struct bio_vec *bv;
 
-	for (i = 0; i < clone->bi_vcnt; i++) {
-		bv = bio_iovec_idx(clone, i);
+	bio_for_each_segment_all(bv, clone, i) {
 		BUG_ON(!bv->bv_page);
 		mempool_free(bv->bv_page, cc->page_pool);
 		bv->bv_page = NULL;
 	}
 }
 
+static struct dm_crypt_io *crypt_io_alloc(struct crypt_config *cc,
+					  struct bio *bio, sector_t sector)
+{
+	struct dm_crypt_io *io;
+
+	io = mempool_alloc(cc->io_pool, GFP_NOIO);
+	io->cc = cc;
+	io->base_bio = bio;
+	io->sector = sector;
+	io->error = 0;
+	io->base_io = NULL;
+	io->ctx.req = NULL;
+	atomic_set(&io->io_pending, 0);
+
+	return io;
+}
+
+static void crypt_inc_pending(struct dm_crypt_io *io)
+{
+	atomic_inc(&io->io_pending);
+}
+
 /*
  * One of the bios was finished. Check for completion of
  * the whole request and correctly clean up the buffer.
+ * If base_io is set, wait for the last fragment to complete.
  */
 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;
 
-	bio_endio(io->base_bio, io->error);
+	if (io->ctx.req)
+		mempool_free(io->ctx.req, cc->req_pool);
 	mempool_free(io, cc->io_pool);
+
+	if (likely(!base_io))
+		bio_endio(base_bio, error);
+	else {
+		if (error && !base_io->error)
+			base_io->error = error;
+		crypt_dec_pending(base_io);
+	}
 }
 
 /*
@@ -545,11 +1071,14 @@ static void crypt_dec_pending(struct dm_crypt_io *io)
  * They must be separated as otherwise the final stages could be
  * starved by new requests which can block in the first stages due
  * to memory allocation.
+ *
+ * The work is done per CPU global for all dm-crypt instances.
+ * They should not depend on each other and do not block.
  */
 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))
@@ -576,180 +1105,203 @@ 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;
 	clone->bi_bdev    = cc->dev->bdev;
 	clone->bi_rw      = io->base_bio->bi_rw;
-	clone->bi_destructor = dm_crypt_bio_destructor;
 }
 
-static void kcryptd_io_read(struct dm_crypt_io *io)
+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;
 
-	atomic_inc(&io->pending);
-
 	/*
 	 * The block layer might modify the bvec array, so always
 	 * copy the required bvecs because we need the original
 	 * one in order to decrypt the whole bio data *afterwards*.
 	 */
-	clone = bio_alloc_bioset(GFP_NOIO, bio_segments(base_bio), cc->bs);
-	if (unlikely(!clone)) {
-		io->error = -ENOMEM;
-		crypt_dec_pending(io);
-		return;
-	}
+	clone = bio_clone_bioset(base_bio, gfp, cc->bs);
+	if (!clone)
+		return 1;
+
+	crypt_inc_pending(io);
 
 	clone_init(io, clone);
-	clone->bi_idx = 0;
-	clone->bi_vcnt = bio_segments(base_bio);
-	clone->bi_size = base_bio->bi_size;
-	clone->bi_sector = cc->start + io->sector;
-	memcpy(clone->bi_io_vec, bio_iovec(base_bio),
-	       sizeof(struct bio_vec) * clone->bi_vcnt);
+	clone->bi_iter.bi_sector = cc->start + io->sector;
 
 	generic_make_request(clone);
+	return 0;
 }
 
 static void kcryptd_io_write(struct dm_crypt_io *io)
 {
 	struct bio *clone = io->ctx.bio_out;
-	struct crypt_config *cc = io->target->private;
-
 	generic_make_request(clone);
-	wake_up(&cc->writeq);
 }
 
 static void kcryptd_io(struct work_struct *work)
 {
 	struct dm_crypt_io *io = container_of(work, struct dm_crypt_io, work);
 
-	if (bio_data_dir(io->base_bio) == READ)
-		kcryptd_io_read(io);
-	else
+	if (bio_data_dir(io->base_bio) == READ) {
+		crypt_inc_pending(io);
+		if (kcryptd_io_read(io, GFP_NOIO))
+			io->error = -ENOMEM;
+		crypt_dec_pending(io);
+	} else
 		kcryptd_io_write(io);
 }
 
 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);
 }
 
-static void kcryptd_crypt_write_io_submit(struct dm_crypt_io *io,
-					  int error, int async)
+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(error < 0)) {
+	if (unlikely(io->error < 0)) {
 		crypt_free_buffer_pages(cc, clone);
 		bio_put(clone);
-		io->error = -EIO;
+		crypt_dec_pending(io);
 		return;
 	}
 
 	/* crypt_convert should have filled the clone bio */
-	BUG_ON(io->ctx.idx_out < clone->bi_vcnt);
+	BUG_ON(io->ctx.iter_out.bi_size);
 
-	clone->bi_sector = cc->start + io->sector;
-	io->sector += bio_sectors(clone);
+	clone->bi_iter.bi_sector = cc->start + io->sector;
 
 	if (async)
 		kcryptd_queue_io(io);
-	else {
-		atomic_inc(&io->pending);
+	else
 		generic_make_request(clone);
-	}
 }
 
-static void kcryptd_crypt_write_convert_loop(struct dm_crypt_io *io)
+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;
-	unsigned remaining = io->base_bio->bi_size;
+	struct dm_crypt_io *new_io;
+	int crypt_finished;
+	unsigned out_of_pages = 0;
+	unsigned remaining = io->base_bio->bi_iter.bi_size;
+	sector_t sector = io->sector;
 	int r;
 
 	/*
+	 * Prevent io from disappearing until this function completes.
+	 */
+	crypt_inc_pending(io);
+	crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, sector);
+
+	/*
 	 * The allocated buffers can be smaller than the whole bio,
 	 * so repeat the whole process until all the data can be handled.
 	 */
 	while (remaining) {
-		clone = crypt_alloc_buffer(io, remaining);
+		clone = crypt_alloc_buffer(io, remaining, &out_of_pages);
 		if (unlikely(!clone)) {
 			io->error = -ENOMEM;
-			return;
+			break;
 		}
 
 		io->ctx.bio_out = clone;
-		io->ctx.idx_out = 0;
+		io->ctx.iter_out = clone->bi_iter;
 
-		remaining -= clone->bi_size;
+		remaining -= clone->bi_iter.bi_size;
+		sector += bio_sectors(clone);
+
+		crypt_inc_pending(io);
 
 		r = crypt_convert(cc, &io->ctx);
+		if (r < 0)
+			io->error = -EIO;
+
+		crypt_finished = atomic_dec_and_test(&io->ctx.cc_pending);
+
+		/* Encryption was already finished, submit io now */
+		if (crypt_finished) {
+			kcryptd_crypt_write_io_submit(io, 0);
 
-		if (atomic_dec_and_test(&io->ctx.pending)) {
-			/* processed, no running async crypto  */
-			kcryptd_crypt_write_io_submit(io, r, 0);
+			/*
+			 * If there was an error, do not try next fragments.
+			 * For async, error is processed in async handler.
+			 */
 			if (unlikely(r < 0))
-				return;
-		} else
-			atomic_inc(&io->pending);
-
-		/* out of memory -> run queues */
-		if (unlikely(remaining)) {
-			/* wait for async crypto then reinitialize pending */
-			wait_event(cc->writeq, !atomic_read(&io->ctx.pending));
-			atomic_set(&io->ctx.pending, 1);
-			congestion_wait(WRITE, HZ/100);
-		}
-	}
-}
+				break;
 
-static void kcryptd_crypt_write_convert(struct dm_crypt_io *io)
-{
-	struct crypt_config *cc = io->target->private;
+			io->sector = sector;
+		}
 
-	/*
-	 * Prevent io from disappearing until this function completes.
-	 */
-	atomic_inc(&io->pending);
+		/*
+		 * Out of memory -> run queues
+		 * But don't wait if split was due to the io size restriction
+		 */
+		if (unlikely(out_of_pages))
+			congestion_wait(BLK_RW_ASYNC, HZ/100);
 
-	crypt_convert_init(cc, &io->ctx, NULL, io->base_bio, io->sector);
-	kcryptd_crypt_write_convert_loop(io);
+		/*
+		 * With async crypto it is unsafe to share the crypto context
+		 * between fragments, so switch to a new dm_crypt_io structure.
+		 */
+		if (unlikely(!crypt_finished && remaining)) {
+			new_io = crypt_io_alloc(io->cc, io->base_bio,
+						sector);
+			crypt_inc_pending(new_io);
+			crypt_convert_init(cc, &new_io->ctx, NULL,
+					   io->base_bio, sector);
+			new_io->ctx.iter_in = io->ctx.iter_in;
+
+			/*
+			 * Fragments after the first use the base_io
+			 * pending count.
+			 */
+			if (!io->base_io)
+				new_io->base_io = io;
+			else {
+				new_io->base_io = io->base_io;
+				crypt_inc_pending(io->base_io);
+				crypt_dec_pending(io);
+			}
+
+			io = new_io;
+		}
+	}
 
 	crypt_dec_pending(io);
 }
 
-static void kcryptd_crypt_read_done(struct dm_crypt_io *io, int error)
+static void kcryptd_crypt_read_done(struct dm_crypt_io *io)
 {
-	if (unlikely(error < 0))
-		io->error = -EIO;
-
 	crypt_dec_pending(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;
 
-	atomic_inc(&io->pending);
+	crypt_inc_pending(io);
 
 	crypt_convert_init(cc, &io->ctx, io->base_bio, io->base_bio,
 			   io->sector);
 
 	r = crypt_convert(cc, &io->ctx);
+	if (r < 0)
+		io->error = -EIO;
 
-	if (atomic_dec_and_test(&io->ctx.pending))
-		kcryptd_crypt_read_done(io, r);
+	if (atomic_dec_and_test(&io->ctx.cc_pending))
+		kcryptd_crypt_read_done(io);
 
 	crypt_dec_pending(io);
 }
@@ -757,24 +1309,31 @@ static void kcryptd_crypt_read_convert(struct dm_crypt_io *io)
 static void kcryptd_async_done(struct crypto_async_request *async_req,
 			       int error)
 {
-	struct convert_context *ctx = async_req->data;
+	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);
 		return;
 	}
 
-	mempool_free(ablkcipher_request_cast(async_req), cc->req_pool);
+	if (!error && cc->iv_gen_ops && cc->iv_gen_ops->post)
+		error = cc->iv_gen_ops->post(cc, iv_of_dmreq(cc, dmreq), dmreq);
+
+	if (error < 0)
+		io->error = -EIO;
+
+	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)
-		kcryptd_crypt_read_done(io, error);
+		kcryptd_crypt_read_done(io);
 	else
-		kcryptd_crypt_write_io_submit(io, error, 1);
+		kcryptd_crypt_write_io_submit(io, 1);
 }
 
 static void kcryptd_crypt(struct work_struct *work)
@@ -789,7 +1348,7 @@ 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);
@@ -801,7 +1360,6 @@ 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';
@@ -810,9 +1368,7 @@ 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;
 	}
 
@@ -822,315 +1378,468 @@ static int crypt_decode_key(u8 *key, char *hex, unsigned int size)
 	return 0;
 }
 
-/*
- * Encode key into its hex representation
- */
-static void crypt_encode_key(char *hex, u8 *key, unsigned int size)
+static void crypt_free_tfms(struct crypt_config *cc)
 {
-	unsigned int i;
+	unsigned i;
 
-	for (i = 0; i < size; i++) {
-		sprintf(hex, "%02x", *key);
-		hex += 2;
-		key++;
+	if (!cc->tfms)
+		return;
+
+	for (i = 0; i < cc->tfms_count; i++)
+		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, char *ciphermode)
+{
+	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++) {
+		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;
+		}
+	}
+
+	return 0;
+}
+
+static int crypt_setkey_allcpus(struct crypt_config *cc)
+{
+	unsigned subkey_size;
+	int err = 0, i, r;
+
+	/* Ignore extra keys (which are used for IV etc) */
+	subkey_size = (cc->key_size - cc->key_extra_size) >> ilog2(cc->tfms_count);
+
+	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;
 }
 
 static int crypt_set_key(struct crypt_config *cc, char *key)
 {
-	unsigned key_size = strlen(key) >> 1;
+	int r = -EINVAL;
+	int key_string_len = strlen(key);
 
-	if (cc->key_size && cc->key_size != key_size)
-		return -EINVAL;
+	/* The key size may not be changed. */
+	if (cc->key_size != (key_string_len >> 1))
+		goto out;
 
-	cc->key_size = key_size; /* initial settings */
+	/* Hyphen (which gives a key_size of zero) means there is no key. */
+	if (!cc->key_size && strcmp(key, "-"))
+		goto out;
 
-	if ((!key_size && strcmp(key, "-")) ||
-	   (key_size && crypt_decode_key(cc->key, key, key_size) < 0))
-		return -EINVAL;
+	if (cc->key_size && crypt_decode_key(cc->key, key, cc->key_size) < 0)
+		goto out;
 
 	set_bit(DM_CRYPT_KEY_VALID, &cc->flags);
 
-	return 0;
+	r = crypt_setkey_allcpus(cc);
+
+out:
+	/* Hex key string not needed after here, so wipe it. */
+	memset(key, '0', key_string_len);
+
+	return r;
 }
 
 static int crypt_wipe_key(struct crypt_config *cc)
 {
 	clear_bit(DM_CRYPT_KEY_VALID, &cc->flags);
 	memset(&cc->key, 0, cc->key_size * sizeof(u8));
-	return 0;
+
+	return crypt_setkey_allcpus(cc);
 }
 
-/*
- * Construct an encryption mapping:
- * <cipher> <key> <iv_offset> <dev_path> <start>
- */
-static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
+static void crypt_dtr(struct dm_target *ti)
 {
-	struct crypt_config *cc;
-	struct crypto_ablkcipher *tfm;
-	char *tmp;
-	char *cipher;
-	char *chainmode;
-	char *ivmode;
-	char *ivopts;
-	unsigned int key_size;
-	unsigned long long tmpll;
+	struct crypt_config *cc = ti->private;
 
-	if (argc != 5) {
-		ti->error = "Not enough arguments";
+	ti->private = NULL;
+
+	if (!cc)
+		return;
+
+	if (cc->io_queue)
+		destroy_workqueue(cc->io_queue);
+	if (cc->crypt_queue)
+		destroy_workqueue(cc->crypt_queue);
+
+	crypt_free_tfms(cc);
+
+	if (cc->bs)
+		bioset_free(cc->bs);
+
+	if (cc->page_pool)
+		mempool_destroy(cc->page_pool);
+	if (cc->req_pool)
+		mempool_destroy(cc->req_pool);
+	if (cc->io_pool)
+		mempool_destroy(cc->io_pool);
+
+	if (cc->iv_gen_ops && cc->iv_gen_ops->dtr)
+		cc->iv_gen_ops->dtr(cc);
+
+	if (cc->dev)
+		dm_put_device(ti, cc->dev);
+
+	kzfree(cc->cipher);
+	kzfree(cc->cipher_string);
+
+	/* Must zero key material before freeing */
+	kzfree(cc);
+}
+
+static int crypt_ctr_cipher(struct dm_target *ti,
+			    char *cipher_in, char *key)
+{
+	struct crypt_config *cc = ti->private;
+	char *tmp, *cipher, *chainmode, *ivmode, *ivopts, *keycount;
+	char *cipher_api = NULL;
+	int ret = -EINVAL;
+	char dummy;
+
+	/* Convert to crypto api definition? */
+	if (strchr(cipher_in, '(')) {
+		ti->error = "Bad cipher specification";
+		return -EINVAL;
+	}
+
+	cc->cipher_string = kstrdup(cipher_in, GFP_KERNEL);
+	if (!cc->cipher_string)
+		goto bad_mem;
+
+	/*
+	 * Legacy dm-crypt cipher specification
+	 * cipher[:keycount]-mode-iv:ivopts
+	 */
+	tmp = cipher_in;
+	keycount = strsep(&tmp, "-");
+	cipher = strsep(&keycount, ":");
+
+	if (!keycount)
+		cc->tfms_count = 1;
+	else if (sscanf(keycount, "%u%c", &cc->tfms_count, &dummy) != 1 ||
+		 !is_power_of_2(cc->tfms_count)) {
+		ti->error = "Bad cipher key count specification";
 		return -EINVAL;
 	}
+	cc->key_parts = cc->tfms_count;
+	cc->key_extra_size = 0;
+
+	cc->cipher = kstrdup(cipher, GFP_KERNEL);
+	if (!cc->cipher)
+		goto bad_mem;
 
-	tmp = argv[0];
-	cipher = strsep(&tmp, "-");
 	chainmode = strsep(&tmp, "-");
 	ivopts = strsep(&tmp, "-");
 	ivmode = strsep(&ivopts, ":");
 
 	if (tmp)
-		DMWARN("Unexpected additional cipher options");
-
-	key_size = strlen(argv[1]) >> 1;
-
- 	cc = kzalloc(sizeof(*cc) + key_size * sizeof(u8), GFP_KERNEL);
-	if (cc == NULL) {
-		ti->error =
-			"Cannot allocate transparent encryption context";
-		return -ENOMEM;
-	}
+		DMWARN("Ignoring unexpected additional cipher options");
 
- 	if (crypt_set_key(cc, argv[1])) {
-		ti->error = "Error decoding key";
-		goto bad_cipher;
-	}
-
-	/* Compatiblity mode for old dm-crypt cipher strings */
-	if (!chainmode || (strcmp(chainmode, "plain") == 0 && !ivmode)) {
+	/*
+	 * For compatibility with the original dm-crypt mapping format, if
+	 * only the cipher name is supplied, use cbc-plain.
+	 */
+	if (!chainmode || (!strcmp(chainmode, "plain") && !ivmode)) {
 		chainmode = "cbc";
 		ivmode = "plain";
 	}
 
 	if (strcmp(chainmode, "ecb") && !ivmode) {
-		ti->error = "This chaining mode requires an IV mechanism";
-		goto bad_cipher;
+		ti->error = "IV mechanism required";
+		return -EINVAL;
 	}
 
-	if (snprintf(cc->cipher, CRYPTO_MAX_ALG_NAME, "%s(%s)",
-		     chainmode, cipher) >= CRYPTO_MAX_ALG_NAME) {
-		ti->error = "Chain mode + cipher name is too long";
-		goto bad_cipher;
+	cipher_api = kmalloc(CRYPTO_MAX_ALG_NAME, GFP_KERNEL);
+	if (!cipher_api)
+		goto bad_mem;
+
+	ret = snprintf(cipher_api, CRYPTO_MAX_ALG_NAME,
+		       "%s(%s)", chainmode, cipher);
+	if (ret < 0) {
+		kfree(cipher_api);
+		goto bad_mem;
 	}
 
-	tfm = crypto_alloc_ablkcipher(cc->cipher, 0, 0);
-	if (IS_ERR(tfm)) {
+	/* Allocate cipher */
+	ret = crypt_alloc_tfms(cc, cipher_api);
+	if (ret < 0) {
 		ti->error = "Error allocating crypto tfm";
-		goto bad_cipher;
+		goto bad;
 	}
 
-	strcpy(cc->cipher, cipher);
-	strcpy(cc->chainmode, chainmode);
-	cc->tfm = tfm;
-
-	/*
-	 * Choose ivmode. Valid modes: "plain", "essiv:<esshash>", "benbi".
-	 * See comments at iv code
-	 */
+	/* Initialize IV */
+	cc->iv_size = crypto_ablkcipher_ivsize(any_tfm(cc));
+	if (cc->iv_size)
+		/* at least a 64 bit sector number should fit in our buffer */
+		cc->iv_size = max(cc->iv_size,
+				  (unsigned int)(sizeof(u64) / sizeof(u8)));
+	else if (ivmode) {
+		DMWARN("Selected cipher does not support IVs");
+		ivmode = NULL;
+	}
 
+	/* Choose ivmode, see comments at iv code. */
 	if (ivmode == NULL)
 		cc->iv_gen_ops = NULL;
 	else if (strcmp(ivmode, "plain") == 0)
 		cc->iv_gen_ops = &crypt_iv_plain_ops;
+	else if (strcmp(ivmode, "plain64") == 0)
+		cc->iv_gen_ops = &crypt_iv_plain64_ops;
 	else if (strcmp(ivmode, "essiv") == 0)
 		cc->iv_gen_ops = &crypt_iv_essiv_ops;
 	else if (strcmp(ivmode, "benbi") == 0)
 		cc->iv_gen_ops = &crypt_iv_benbi_ops;
 	else if (strcmp(ivmode, "null") == 0)
 		cc->iv_gen_ops = &crypt_iv_null_ops;
-	else {
+	else if (strcmp(ivmode, "lmk") == 0) {
+		cc->iv_gen_ops = &crypt_iv_lmk_ops;
+		/*
+		 * Version 2 and 3 is recognised according
+		 * to length of provided multi-key string.
+		 * If present (version 3), last key is used as IV seed.
+		 * All keys (including IV seed) are always the same size.
+		 */
+		if (cc->key_size % cc->key_parts) {
+			cc->key_parts++;
+			cc->key_extra_size = cc->key_size / cc->key_parts;
+		}
+	} else if (strcmp(ivmode, "tcw") == 0) {
+		cc->iv_gen_ops = &crypt_iv_tcw_ops;
+		cc->key_parts += 2; /* IV + whitening */
+		cc->key_extra_size = cc->iv_size + TCW_WHITENING_SIZE;
+	} else {
+		ret = -EINVAL;
 		ti->error = "Invalid IV mode";
-		goto bad_ivmode;
+		goto bad;
 	}
 
-	if (cc->iv_gen_ops && cc->iv_gen_ops->ctr &&
-	    cc->iv_gen_ops->ctr(cc, ti, ivopts) < 0)
-		goto bad_ivmode;
+	/* Initialize and set key */
+	ret = crypt_set_key(cc, key);
+	if (ret < 0) {
+		ti->error = "Error decoding and setting key";
+		goto bad;
+	}
 
-	cc->iv_size = crypto_ablkcipher_ivsize(tfm);
-	if (cc->iv_size)
-		/* at least a 64 bit sector number should fit in our buffer */
-		cc->iv_size = max(cc->iv_size,
-				  (unsigned int)(sizeof(u64) / sizeof(u8)));
-	else {
-		if (cc->iv_gen_ops) {
-			DMWARN("Selected cipher does not support IVs");
-			if (cc->iv_gen_ops->dtr)
-				cc->iv_gen_ops->dtr(cc);
-			cc->iv_gen_ops = NULL;
+	/* Allocate IV */
+	if (cc->iv_gen_ops && cc->iv_gen_ops->ctr) {
+		ret = cc->iv_gen_ops->ctr(cc, ti, ivopts);
+		if (ret < 0) {
+			ti->error = "Error creating IV";
+			goto bad;
 		}
 	}
 
+	/* Initialize IV (set keys for ESSIV etc) */
+	if (cc->iv_gen_ops && cc->iv_gen_ops->init) {
+		ret = cc->iv_gen_ops->init(cc);
+		if (ret < 0) {
+			ti->error = "Error initialising IV";
+			goto bad;
+		}
+	}
+
+	ret = 0;
+bad:
+	kfree(cipher_api);
+	return ret;
+
+bad_mem:
+	ti->error = "Cannot allocate cipher strings";
+	return -ENOMEM;
+}
+
+/*
+ * Construct an encryption mapping:
+ * <cipher> <key> <iv_offset> <dev_path> <start>
+ */
+static int crypt_ctr(struct dm_target *ti, unsigned int argc, char **argv)
+{
+	struct crypt_config *cc;
+	unsigned int key_size, opt_params;
+	unsigned long long tmpll;
+	int ret;
+	struct dm_arg_set as;
+	const char *opt_string;
+	char dummy;
+
+	static struct dm_arg _args[] = {
+		{0, 1, "Invalid number of feature args"},
+	};
+
+	if (argc < 5) {
+		ti->error = "Not enough arguments";
+		return -EINVAL;
+	}
+
+	key_size = strlen(argv[1]) >> 1;
+
+	cc = kzalloc(sizeof(*cc) + key_size * sizeof(u8), GFP_KERNEL);
+	if (!cc) {
+		ti->error = "Cannot allocate encryption context";
+		return -ENOMEM;
+	}
+	cc->key_size = key_size;
+
+	ti->private = cc;
+	ret = crypt_ctr_cipher(ti, argv[0], argv[1]);
+	if (ret < 0)
+		goto bad;
+
+	ret = -ENOMEM;
 	cc->io_pool = mempool_create_slab_pool(MIN_IOS, _crypt_io_pool);
 	if (!cc->io_pool) {
 		ti->error = "Cannot allocate crypt io mempool";
-		goto bad_slab_pool;
+		goto bad;
 	}
 
 	cc->dmreq_start = sizeof(struct ablkcipher_request);
-	cc->dmreq_start += crypto_ablkcipher_reqsize(tfm);
+	cc->dmreq_start += crypto_ablkcipher_reqsize(any_tfm(cc));
 	cc->dmreq_start = ALIGN(cc->dmreq_start, crypto_tfm_ctx_alignment());
-	cc->dmreq_start += crypto_ablkcipher_alignmask(tfm) &
+	cc->dmreq_start += crypto_ablkcipher_alignmask(any_tfm(cc)) &
 			   ~(crypto_tfm_ctx_alignment() - 1);
 
 	cc->req_pool = mempool_create_kmalloc_pool(MIN_IOS, cc->dmreq_start +
 			sizeof(struct dm_crypt_request) + cc->iv_size);
 	if (!cc->req_pool) {
 		ti->error = "Cannot allocate crypt request mempool";
-		goto bad_req_pool;
+		goto bad;
 	}
-	cc->req = NULL;
 
 	cc->page_pool = mempool_create_page_pool(MIN_POOL_PAGES, 0);
 	if (!cc->page_pool) {
 		ti->error = "Cannot allocate page mempool";
-		goto bad_page_pool;
+		goto bad;
 	}
 
-	cc->bs = bioset_create(MIN_IOS, MIN_IOS);
+	cc->bs = bioset_create(MIN_IOS, 0);
 	if (!cc->bs) {
 		ti->error = "Cannot allocate crypt bioset";
-		goto bad_bs;
+		goto bad;
 	}
 
-	if (crypto_ablkcipher_setkey(tfm, cc->key, key_size) < 0) {
-		ti->error = "Error setting key";
-		goto bad_device;
-	}
-
-	if (sscanf(argv[2], "%llu", &tmpll) != 1) {
+	ret = -EINVAL;
+	if (sscanf(argv[2], "%llu%c", &tmpll, &dummy) != 1) {
 		ti->error = "Invalid iv_offset sector";
-		goto bad_device;
+		goto bad;
 	}
 	cc->iv_offset = tmpll;
 
-	if (sscanf(argv[4], "%llu", &tmpll) != 1) {
+	if (dm_get_device(ti, argv[3], dm_table_get_mode(ti->table), &cc->dev)) {
+		ti->error = "Device lookup failed";
+		goto bad;
+	}
+
+	if (sscanf(argv[4], "%llu%c", &tmpll, &dummy) != 1) {
 		ti->error = "Invalid device sector";
-		goto bad_device;
+		goto bad;
 	}
 	cc->start = tmpll;
 
-	if (dm_get_device(ti, argv[3], cc->start, ti->len,
-			  dm_table_get_mode(ti->table), &cc->dev)) {
-		ti->error = "Device lookup failed";
-		goto bad_device;
-	}
+	argv += 5;
+	argc -= 5;
+
+	/* Optional parameters */
+	if (argc) {
+		as.argc = argc;
+		as.argv = argv;
 
-	if (ivmode && cc->iv_gen_ops) {
-		if (ivopts)
-			*(ivopts - 1) = ':';
-		cc->iv_mode = kmalloc(strlen(ivmode) + 1, GFP_KERNEL);
-		if (!cc->iv_mode) {
-			ti->error = "Error kmallocing iv_mode string";
-			goto bad_ivmode_string;
+		ret = dm_read_arg_group(_args, &as, &opt_params, &ti->error);
+		if (ret)
+			goto bad;
+
+		opt_string = dm_shift_arg(&as);
+
+		if (opt_params == 1 && opt_string &&
+		    !strcasecmp(opt_string, "allow_discards"))
+			ti->num_discard_bios = 1;
+		else if (opt_params) {
+			ret = -EINVAL;
+			ti->error = "Invalid feature arguments";
+			goto bad;
 		}
-		strcpy(cc->iv_mode, ivmode);
-	} else
-		cc->iv_mode = NULL;
+	}
 
-	cc->io_queue = create_singlethread_workqueue("kcryptd_io");
+	ret = -ENOMEM;
+	cc->io_queue = alloc_workqueue("kcryptd_io", WQ_MEM_RECLAIM, 1);
 	if (!cc->io_queue) {
 		ti->error = "Couldn't create kcryptd io queue";
-		goto bad_io_queue;
+		goto bad;
 	}
 
-	cc->crypt_queue = create_singlethread_workqueue("kcryptd");
+	cc->crypt_queue = alloc_workqueue("kcryptd",
+					  WQ_CPU_INTENSIVE | WQ_MEM_RECLAIM, 1);
 	if (!cc->crypt_queue) {
 		ti->error = "Couldn't create kcryptd queue";
-		goto bad_crypt_queue;
+		goto bad;
 	}
 
-	init_waitqueue_head(&cc->writeq);
-	ti->private = cc;
-	return 0;
-
-bad_crypt_queue:
-	destroy_workqueue(cc->io_queue);
-bad_io_queue:
-	kfree(cc->iv_mode);
-bad_ivmode_string:
-	dm_put_device(ti, cc->dev);
-bad_device:
-	bioset_free(cc->bs);
-bad_bs:
-	mempool_destroy(cc->page_pool);
-bad_page_pool:
-	mempool_destroy(cc->req_pool);
-bad_req_pool:
-	mempool_destroy(cc->io_pool);
-bad_slab_pool:
-	if (cc->iv_gen_ops && cc->iv_gen_ops->dtr)
-		cc->iv_gen_ops->dtr(cc);
-bad_ivmode:
-	crypto_free_ablkcipher(tfm);
-bad_cipher:
-	/* Must zero key material before freeing */
-	memset(cc, 0, sizeof(*cc) + cc->key_size * sizeof(u8));
-	kfree(cc);
-	return -EINVAL;
-}
+	ti->num_flush_bios = 1;
+	ti->discard_zeroes_data_unsupported = true;
 
-static void crypt_dtr(struct dm_target *ti)
-{
-	struct crypt_config *cc = (struct crypt_config *) ti->private;
-
-	destroy_workqueue(cc->io_queue);
-	destroy_workqueue(cc->crypt_queue);
-
-	if (cc->req)
-		mempool_free(cc->req, cc->req_pool);
-
-	bioset_free(cc->bs);
-	mempool_destroy(cc->page_pool);
-	mempool_destroy(cc->req_pool);
-	mempool_destroy(cc->io_pool);
-
-	kfree(cc->iv_mode);
-	if (cc->iv_gen_ops && cc->iv_gen_ops->dtr)
-		cc->iv_gen_ops->dtr(cc);
-	crypto_free_ablkcipher(cc->tfm);
-	dm_put_device(ti, cc->dev);
+	return 0;
 
-	/* Must zero key material before freeing */
-	memset(cc, 0, sizeof(*cc) + cc->key_size * sizeof(u8));
-	kfree(cc);
+bad:
+	crypt_dtr(ti);
+	return ret;
 }
 
-static int crypt_map(struct dm_target *ti, struct bio *bio,
-		     union map_info *map_context)
+static int crypt_map(struct dm_target *ti, struct bio *bio)
 {
-	struct crypt_config *cc = ti->private;
 	struct dm_crypt_io *io;
+	struct crypt_config *cc = ti->private;
 
-	io = mempool_alloc(cc->io_pool, GFP_NOIO);
-	io->target = ti;
-	io->base_bio = bio;
-	io->sector = bio->bi_sector - ti->begin;
-	io->error = 0;
-	atomic_set(&io->pending, 0);
+	/*
+	 * If bio is REQ_FLUSH or REQ_DISCARD, just bypass crypt queues.
+	 * - for REQ_FLUSH device-mapper core ensures that no IO is in-flight
+	 * - for REQ_DISCARD caller must use flush if IO ordering matters
+	 */
+	if (unlikely(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD))) {
+		bio->bi_bdev = cc->dev->bdev;
+		if (bio_sectors(bio))
+			bio->bi_iter.bi_sector = cc->start +
+				dm_target_offset(ti, bio->bi_iter.bi_sector);
+		return DM_MAPIO_REMAPPED;
+	}
 
-	if (bio_data_dir(io->base_bio) == READ)
-		kcryptd_queue_io(io);
-	else
+	io = crypt_io_alloc(cc, bio, dm_target_offset(ti, bio->bi_iter.bi_sector));
+
+	if (bio_data_dir(io->base_bio) == READ) {
+		if (kcryptd_io_read(io, GFP_NOWAIT))
+			kcryptd_queue_io(io);
+	} else
 		kcryptd_queue_crypt(io);
 
 	return DM_MAPIO_SUBMITTED;
 }
 
-static int crypt_status(struct dm_target *ti, status_type_t type,
-			char *result, unsigned int maxlen)
+static void crypt_status(struct dm_target *ti, status_type_t type,
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
-	struct crypt_config *cc = (struct crypt_config *) ti->private;
-	unsigned int sz = 0;
+	struct crypt_config *cc = ti->private;
+	unsigned i, sz = 0;
 
 	switch (type) {
 	case STATUSTYPE_INFO:
@@ -1138,29 +1847,22 @@ static int crypt_status(struct dm_target *ti, status_type_t type,
 		break;
 
 	case STATUSTYPE_TABLE:
-		if (cc->iv_mode)
-			DMEMIT("%s-%s-%s ", cc->cipher, cc->chainmode,
-			       cc->iv_mode);
+		DMEMIT("%s ", cc->cipher_string);
+
+		if (cc->key_size > 0)
+			for (i = 0; i < cc->key_size; i++)
+				DMEMIT("%02x", cc->key[i]);
 		else
-			DMEMIT("%s-%s ", cc->cipher, cc->chainmode);
-
-		if (cc->key_size > 0) {
-			if ((maxlen - sz) < ((cc->key_size << 1) + 1))
-				return -ENOMEM;
-
-			crypt_encode_key(result + sz, cc->key, cc->key_size);
-			sz += cc->key_size << 1;
-		} else {
-			if (sz >= maxlen)
-				return -ENOMEM;
-			result[sz++] = '-';
-		}
+			DMEMIT("-");
 
 		DMEMIT(" %llu %s %llu", (unsigned long long)cc->iv_offset,
 				cc->dev->name, (unsigned long long)cc->start);
+
+		if (ti->num_discard_bios)
+			DMEMIT(" 1 allow_discards");
+
 		break;
 	}
-	return 0;
 }
 
 static void crypt_postsuspend(struct dm_target *ti)
@@ -1196,19 +1898,32 @@ static void crypt_resume(struct dm_target *ti)
 static int crypt_message(struct dm_target *ti, unsigned argc, char **argv)
 {
 	struct crypt_config *cc = ti->private;
+	int ret = -EINVAL;
 
 	if (argc < 2)
 		goto error;
 
-	if (!strnicmp(argv[0], MESG_STR("key"))) {
+	if (!strcasecmp(argv[0], "key")) {
 		if (!test_bit(DM_CRYPT_SUSPENDED, &cc->flags)) {
 			DMWARN("not suspended during key manipulation.");
 			return -EINVAL;
 		}
-		if (argc == 3 && !strnicmp(argv[1], MESG_STR("set")))
-			return crypt_set_key(cc, argv[2]);
-		if (argc == 2 && !strnicmp(argv[1], MESG_STR("wipe")))
+		if (argc == 3 && !strcasecmp(argv[1], "set")) {
+			ret = crypt_set_key(cc, argv[2]);
+			if (ret)
+				return ret;
+			if (cc->iv_gen_ops && cc->iv_gen_ops->init)
+				ret = cc->iv_gen_ops->init(cc);
+			return ret;
+		}
+		if (argc == 2 && !strcasecmp(argv[1], "wipe")) {
+			if (cc->iv_gen_ops && cc->iv_gen_ops->wipe) {
+				ret = cc->iv_gen_ops->wipe(cc);
+				if (ret)
+					return ret;
+			}
 			return crypt_wipe_key(cc);
+		}
 	}
 
 error:
@@ -1226,14 +1941,22 @@ static int crypt_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
 		return max_size;
 
 	bvm->bi_bdev = cc->dev->bdev;
-	bvm->bi_sector = cc->start + bvm->bi_sector - ti->begin;
+	bvm->bi_sector = cc->start + dm_target_offset(ti, bvm->bi_sector);
 
 	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
 }
 
+static int crypt_iterate_devices(struct dm_target *ti,
+				 iterate_devices_callout_fn fn, void *data)
+{
+	struct crypt_config *cc = ti->private;
+
+	return fn(ti, cc->dev, cc->start, ti->len, data);
+}
+
 static struct target_type crypt_target = {
 	.name   = "crypt",
-	.version= {1, 6, 0},
+	.version = {1, 13, 0},
 	.module = THIS_MODULE,
 	.ctr    = crypt_ctr,
 	.dtr    = crypt_dtr,
@@ -1244,6 +1967,7 @@ static struct target_type crypt_target = {
 	.resume = crypt_resume,
 	.message = crypt_message,
 	.merge  = crypt_merge,
+	.iterate_devices = crypt_iterate_devices,
 };
 
 static int __init dm_crypt_init(void)
@@ -1265,17 +1989,13 @@ static int __init dm_crypt_init(void)
 
 static void __exit dm_crypt_exit(void)
 {
-	int r = dm_unregister_target(&crypt_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
-
+	dm_unregister_target(&crypt_target);
 	kmem_cache_destroy(_crypt_io_pool);
 }
 
 module_init(dm_crypt_init);
 module_exit(dm_crypt_exit);
 
-MODULE_AUTHOR("Christophe Saout <christophe@saout.de>");
+MODULE_AUTHOR("Jana Saout <jana@saout.de>");
 MODULE_DESCRIPTION(DM_NAME " target for transparent encryption / decryption");
 MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-delay.c b/drivers/md/dm-delay.c
index bdd37f881c4..42c3a27a14c 100644
--- a/drivers/md/dm-delay.c
+++ b/drivers/md/dm-delay.c
@@ -13,18 +13,17 @@
 #include <linux/bio.h>
 #include <linux/slab.h>
 
-#include "dm.h"
-#include "dm-bio-list.h"
+#include <linux/device-mapper.h>
 
 #define DM_MSG_PREFIX "delay"
 
 struct delay_c {
 	struct timer_list delay_timer;
 	struct mutex timer_lock;
+	struct workqueue_struct *kdelayd_wq;
 	struct work_struct flush_expired_bios;
 	struct list_head delayed_bios;
 	atomic_t may_delay;
-	mempool_t *delayed_pool;
 
 	struct dm_dev *dev_read;
 	sector_t start_read;
@@ -40,20 +39,16 @@ struct delay_c {
 struct dm_delay_info {
 	struct delay_c *context;
 	struct list_head list;
-	struct bio *bio;
 	unsigned long expires;
 };
 
 static DEFINE_MUTEX(delayed_bios_lock);
 
-static struct workqueue_struct *kdelayd_wq;
-static struct kmem_cache *delayed_cache;
-
 static void handle_delayed_timer(unsigned long data)
 {
 	struct delay_c *dc = (struct delay_c *)data;
 
-	queue_work(kdelayd_wq, &dc->flush_expired_bios);
+	queue_work(dc->kdelayd_wq, &dc->flush_expired_bios);
 }
 
 static void queue_timeout(struct delay_c *dc, unsigned long expires)
@@ -88,13 +83,14 @@ static struct bio *flush_delayed_bios(struct delay_c *dc, int flush_all)
 	mutex_lock(&delayed_bios_lock);
 	list_for_each_entry_safe(delayed, next, &dc->delayed_bios, list) {
 		if (flush_all || time_after_eq(jiffies, delayed->expires)) {
+			struct bio *bio = dm_bio_from_per_bio_data(delayed,
+						sizeof(struct dm_delay_info));
 			list_del(&delayed->list);
-			bio_list_add(&flush_bios, delayed->bio);
-			if ((bio_data_dir(delayed->bio) == WRITE))
+			bio_list_add(&flush_bios, bio);
+			if ((bio_data_dir(bio) == WRITE))
 				delayed->context->writes--;
 			else
 				delayed->context->reads--;
-			mempool_free(delayed, dc->delayed_pool);
 			continue;
 		}
 
@@ -132,6 +128,7 @@ static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
 	struct delay_c *dc;
 	unsigned long long tmpll;
+	char dummy;
 
 	if (argc != 3 && argc != 6) {
 		ti->error = "requires exactly 3 or 6 arguments";
@@ -146,19 +143,19 @@ static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	dc->reads = dc->writes = 0;
 
-	if (sscanf(argv[1], "%llu", &tmpll) != 1) {
+	if (sscanf(argv[1], "%llu%c", &tmpll, &dummy) != 1) {
 		ti->error = "Invalid device sector";
 		goto bad;
 	}
 	dc->start_read = tmpll;
 
-	if (sscanf(argv[2], "%u", &dc->read_delay) != 1) {
+	if (sscanf(argv[2], "%u%c", &dc->read_delay, &dummy) != 1) {
 		ti->error = "Invalid delay";
 		goto bad;
 	}
 
-	if (dm_get_device(ti, argv[0], dc->start_read, ti->len,
-			  dm_table_get_mode(ti->table), &dc->dev_read)) {
+	if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table),
+			  &dc->dev_read)) {
 		ti->error = "Device lookup failed";
 		goto bad;
 	}
@@ -167,28 +164,28 @@ static int delay_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	if (argc == 3)
 		goto out;
 
-	if (sscanf(argv[4], "%llu", &tmpll) != 1) {
+	if (sscanf(argv[4], "%llu%c", &tmpll, &dummy) != 1) {
 		ti->error = "Invalid write device sector";
 		goto bad_dev_read;
 	}
 	dc->start_write = tmpll;
 
-	if (sscanf(argv[5], "%u", &dc->write_delay) != 1) {
+	if (sscanf(argv[5], "%u%c", &dc->write_delay, &dummy) != 1) {
 		ti->error = "Invalid write delay";
 		goto bad_dev_read;
 	}
 
-	if (dm_get_device(ti, argv[3], dc->start_write, ti->len,
-			  dm_table_get_mode(ti->table), &dc->dev_write)) {
+	if (dm_get_device(ti, argv[3], dm_table_get_mode(ti->table),
+			  &dc->dev_write)) {
 		ti->error = "Write device lookup failed";
 		goto bad_dev_read;
 	}
 
 out:
-	dc->delayed_pool = mempool_create_slab_pool(128, delayed_cache);
-	if (!dc->delayed_pool) {
-		DMERR("Couldn't create delayed bio pool.");
-		goto bad_dev_write;
+	dc->kdelayd_wq = alloc_workqueue("kdelayd", WQ_MEM_RECLAIM, 0);
+	if (!dc->kdelayd_wq) {
+		DMERR("Couldn't start kdelayd");
+		goto bad_queue;
 	}
 
 	setup_timer(&dc->delay_timer, handle_delayed_timer, (unsigned long)dc);
@@ -198,10 +195,13 @@ out:
 	mutex_init(&dc->timer_lock);
 	atomic_set(&dc->may_delay, 1);
 
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->per_bio_data_size = sizeof(struct dm_delay_info);
 	ti->private = dc;
 	return 0;
 
-bad_dev_write:
+bad_queue:
 	if (dc->dev_write)
 		dm_put_device(ti, dc->dev_write);
 bad_dev_read:
@@ -215,14 +215,13 @@ static void delay_dtr(struct dm_target *ti)
 {
 	struct delay_c *dc = ti->private;
 
-	flush_workqueue(kdelayd_wq);
+	destroy_workqueue(dc->kdelayd_wq);
 
 	dm_put_device(ti, dc->dev_read);
 
 	if (dc->dev_write)
 		dm_put_device(ti, dc->dev_write);
 
-	mempool_destroy(dc->delayed_pool);
 	kfree(dc);
 }
 
@@ -234,10 +233,9 @@ static int delay_bio(struct delay_c *dc, int delay, struct bio *bio)
 	if (!delay || !atomic_read(&dc->may_delay))
 		return 1;
 
-	delayed = mempool_alloc(dc->delayed_pool, GFP_NOIO);
+	delayed = dm_per_bio_data(bio, sizeof(struct dm_delay_info));
 
 	delayed->context = dc;
-	delayed->bio = bio;
 	delayed->expires = expires = jiffies + (delay * HZ / 1000);
 
 	mutex_lock(&delayed_bios_lock);
@@ -272,28 +270,28 @@ static void delay_resume(struct dm_target *ti)
 	atomic_set(&dc->may_delay, 1);
 }
 
-static int delay_map(struct dm_target *ti, struct bio *bio,
-		     union map_info *map_context)
+static int delay_map(struct dm_target *ti, struct bio *bio)
 {
 	struct delay_c *dc = ti->private;
 
 	if ((bio_data_dir(bio) == WRITE) && (dc->dev_write)) {
 		bio->bi_bdev = dc->dev_write->bdev;
-		bio->bi_sector = dc->start_write +
-				 (bio->bi_sector - ti->begin);
+		if (bio_sectors(bio))
+			bio->bi_iter.bi_sector = dc->start_write +
+				dm_target_offset(ti, bio->bi_iter.bi_sector);
 
 		return delay_bio(dc, dc->write_delay, bio);
 	}
 
 	bio->bi_bdev = dc->dev_read->bdev;
-	bio->bi_sector = dc->start_read +
-			 (bio->bi_sector - ti->begin);
+	bio->bi_iter.bi_sector = dc->start_read +
+		dm_target_offset(ti, bio->bi_iter.bi_sector);
 
 	return delay_bio(dc, dc->read_delay, bio);
 }
 
-static int delay_status(struct dm_target *ti, status_type_t type,
-			char *result, unsigned maxlen)
+static void delay_status(struct dm_target *ti, status_type_t type,
+			 unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct delay_c *dc = ti->private;
 	int sz = 0;
@@ -313,13 +311,28 @@ static int delay_status(struct dm_target *ti, status_type_t type,
 			       dc->write_delay);
 		break;
 	}
+}
 
-	return 0;
+static int delay_iterate_devices(struct dm_target *ti,
+				 iterate_devices_callout_fn fn, void *data)
+{
+	struct delay_c *dc = ti->private;
+	int ret = 0;
+
+	ret = fn(ti, dc->dev_read, dc->start_read, ti->len, data);
+	if (ret)
+		goto out;
+
+	if (dc->dev_write)
+		ret = fn(ti, dc->dev_write, dc->start_write, ti->len, data);
+
+out:
+	return ret;
 }
 
 static struct target_type delay_target = {
 	.name	     = "delay",
-	.version     = {1, 0, 2},
+	.version     = {1, 2, 1},
 	.module      = THIS_MODULE,
 	.ctr	     = delay_ctr,
 	.dtr	     = delay_dtr,
@@ -327,23 +340,12 @@ static struct target_type delay_target = {
 	.presuspend  = delay_presuspend,
 	.resume	     = delay_resume,
 	.status	     = delay_status,
+	.iterate_devices = delay_iterate_devices,
 };
 
 static int __init dm_delay_init(void)
 {
-	int r = -ENOMEM;
-
-	kdelayd_wq = create_workqueue("kdelayd");
-	if (!kdelayd_wq) {
-		DMERR("Couldn't start kdelayd");
-		goto bad_queue;
-	}
-
-	delayed_cache = KMEM_CACHE(dm_delay_info, 0);
-	if (!delayed_cache) {
-		DMERR("Couldn't create delayed bio cache.");
-		goto bad_memcache;
-	}
+	int r;
 
 	r = dm_register_target(&delay_target);
 	if (r < 0) {
@@ -354,22 +356,12 @@ static int __init dm_delay_init(void)
 	return 0;
 
 bad_register:
-	kmem_cache_destroy(delayed_cache);
-bad_memcache:
-	destroy_workqueue(kdelayd_wq);
-bad_queue:
 	return r;
 }
 
 static void __exit dm_delay_exit(void)
 {
-	int r = dm_unregister_target(&delay_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
-
-	kmem_cache_destroy(delayed_cache);
-	destroy_workqueue(kdelayd_wq);
+	dm_unregister_target(&delay_target);
 }
 
 /* Module hooks */
diff --git a/drivers/md/dm-era-target.c b/drivers/md/dm-era-target.c
new file mode 100644
index 00000000000..ad913cd4ade
--- /dev/null
+++ b/drivers/md/dm-era-target.c
@@ -0,0 +1,1747 @@
+#include "dm.h"
+#include "persistent-data/dm-transaction-manager.h"
+#include "persistent-data/dm-bitset.h"
+#include "persistent-data/dm-space-map.h"
+
+#include <linux/dm-io.h>
+#include <linux/dm-kcopyd.h>
+#include <linux/init.h>
+#include <linux/mempool.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#define DM_MSG_PREFIX "era"
+
+#define SUPERBLOCK_LOCATION 0
+#define SUPERBLOCK_MAGIC 2126579579
+#define SUPERBLOCK_CSUM_XOR 146538381
+#define MIN_ERA_VERSION 1
+#define MAX_ERA_VERSION 1
+#define INVALID_WRITESET_ROOT SUPERBLOCK_LOCATION
+#define MIN_BLOCK_SIZE 8
+
+/*----------------------------------------------------------------
+ * Writeset
+ *--------------------------------------------------------------*/
+struct writeset_metadata {
+	uint32_t nr_bits;
+	dm_block_t root;
+};
+
+struct writeset {
+	struct writeset_metadata md;
+
+	/*
+	 * An in core copy of the bits to save constantly doing look ups on
+	 * disk.
+	 */
+	unsigned long *bits;
+};
+
+/*
+ * This does not free off the on disk bitset as this will normally be done
+ * after digesting into the era array.
+ */
+static void writeset_free(struct writeset *ws)
+{
+	vfree(ws->bits);
+}
+
+static int setup_on_disk_bitset(struct dm_disk_bitset *info,
+				unsigned nr_bits, dm_block_t *root)
+{
+	int r;
+
+	r = dm_bitset_empty(info, root);
+	if (r)
+		return r;
+
+	return dm_bitset_resize(info, *root, 0, nr_bits, false, root);
+}
+
+static size_t bitset_size(unsigned nr_bits)
+{
+	return sizeof(unsigned long) * dm_div_up(nr_bits, BITS_PER_LONG);
+}
+
+/*
+ * Allocates memory for the in core bitset.
+ */
+static int writeset_alloc(struct writeset *ws, dm_block_t nr_blocks)
+{
+	ws->md.nr_bits = nr_blocks;
+	ws->md.root = INVALID_WRITESET_ROOT;
+	ws->bits = vzalloc(bitset_size(nr_blocks));
+	if (!ws->bits) {
+		DMERR("%s: couldn't allocate in memory bitset", __func__);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+/*
+ * Wipes the in-core bitset, and creates a new on disk bitset.
+ */
+static int writeset_init(struct dm_disk_bitset *info, struct writeset *ws)
+{
+	int r;
+
+	memset(ws->bits, 0, bitset_size(ws->md.nr_bits));
+
+	r = setup_on_disk_bitset(info, ws->md.nr_bits, &ws->md.root);
+	if (r) {
+		DMERR("%s: setup_on_disk_bitset failed", __func__);
+		return r;
+	}
+
+	return 0;
+}
+
+static bool writeset_marked(struct writeset *ws, dm_block_t block)
+{
+	return test_bit(block, ws->bits);
+}
+
+static int writeset_marked_on_disk(struct dm_disk_bitset *info,
+				   struct writeset_metadata *m, dm_block_t block,
+				   bool *result)
+{
+	dm_block_t old = m->root;
+
+	/*
+	 * The bitset was flushed when it was archived, so we know there'll
+	 * be no change to the root.
+	 */
+	int r = dm_bitset_test_bit(info, m->root, block, &m->root, result);
+	if (r) {
+		DMERR("%s: dm_bitset_test_bit failed", __func__);
+		return r;
+	}
+
+	BUG_ON(m->root != old);
+
+	return r;
+}
+
+/*
+ * Returns < 0 on error, 0 if the bit wasn't previously set, 1 if it was.
+ */
+static int writeset_test_and_set(struct dm_disk_bitset *info,
+				 struct writeset *ws, uint32_t block)
+{
+	int r;
+
+	if (!test_and_set_bit(block, ws->bits)) {
+		r = dm_bitset_set_bit(info, ws->md.root, block, &ws->md.root);
+		if (r) {
+			/* FIXME: fail mode */
+			return r;
+		}
+
+		return 0;
+	}
+
+	return 1;
+}
+
+/*----------------------------------------------------------------
+ * On disk metadata layout
+ *--------------------------------------------------------------*/
+#define SPACE_MAP_ROOT_SIZE 128
+#define UUID_LEN 16
+
+struct writeset_disk {
+	__le32 nr_bits;
+	__le64 root;
+} __packed;
+
+struct superblock_disk {
+	__le32 csum;
+	__le32 flags;
+	__le64 blocknr;
+
+	__u8 uuid[UUID_LEN];
+	__le64 magic;
+	__le32 version;
+
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+
+	__le32 data_block_size;
+	__le32 metadata_block_size;
+	__le32 nr_blocks;
+
+	__le32 current_era;
+	struct writeset_disk current_writeset;
+
+	/*
+	 * Only these two fields are valid within the metadata snapshot.
+	 */
+	__le64 writeset_tree_root;
+	__le64 era_array_root;
+
+	__le64 metadata_snap;
+} __packed;
+
+/*----------------------------------------------------------------
+ * Superblock validation
+ *--------------------------------------------------------------*/
+static void sb_prepare_for_write(struct dm_block_validator *v,
+				 struct dm_block *b,
+				 size_t sb_block_size)
+{
+	struct superblock_disk *disk = dm_block_data(b);
+
+	disk->blocknr = cpu_to_le64(dm_block_location(b));
+	disk->csum = cpu_to_le32(dm_bm_checksum(&disk->flags,
+						sb_block_size - sizeof(__le32),
+						SUPERBLOCK_CSUM_XOR));
+}
+
+static int check_metadata_version(struct superblock_disk *disk)
+{
+	uint32_t metadata_version = le32_to_cpu(disk->version);
+	if (metadata_version < MIN_ERA_VERSION || metadata_version > MAX_ERA_VERSION) {
+		DMERR("Era metadata version %u found, but only versions between %u and %u supported.",
+		      metadata_version, MIN_ERA_VERSION, MAX_ERA_VERSION);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int sb_check(struct dm_block_validator *v,
+		    struct dm_block *b,
+		    size_t sb_block_size)
+{
+	struct superblock_disk *disk = dm_block_data(b);
+	__le32 csum_le;
+
+	if (dm_block_location(b) != le64_to_cpu(disk->blocknr)) {
+		DMERR("sb_check failed: blocknr %llu: wanted %llu",
+		      le64_to_cpu(disk->blocknr),
+		      (unsigned long long)dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	if (le64_to_cpu(disk->magic) != SUPERBLOCK_MAGIC) {
+		DMERR("sb_check failed: magic %llu: wanted %llu",
+		      le64_to_cpu(disk->magic),
+		      (unsigned long long) SUPERBLOCK_MAGIC);
+		return -EILSEQ;
+	}
+
+	csum_le = cpu_to_le32(dm_bm_checksum(&disk->flags,
+					     sb_block_size - sizeof(__le32),
+					     SUPERBLOCK_CSUM_XOR));
+	if (csum_le != disk->csum) {
+		DMERR("sb_check failed: csum %u: wanted %u",
+		      le32_to_cpu(csum_le), le32_to_cpu(disk->csum));
+		return -EILSEQ;
+	}
+
+	return check_metadata_version(disk);
+}
+
+static struct dm_block_validator sb_validator = {
+	.name = "superblock",
+	.prepare_for_write = sb_prepare_for_write,
+	.check = sb_check
+};
+
+/*----------------------------------------------------------------
+ * Low level metadata handling
+ *--------------------------------------------------------------*/
+#define DM_ERA_METADATA_BLOCK_SIZE 4096
+#define DM_ERA_METADATA_CACHE_SIZE 64
+#define ERA_MAX_CONCURRENT_LOCKS 5
+
+struct era_metadata {
+	struct block_device *bdev;
+	struct dm_block_manager *bm;
+	struct dm_space_map *sm;
+	struct dm_transaction_manager *tm;
+
+	dm_block_t block_size;
+	uint32_t nr_blocks;
+
+	uint32_t current_era;
+
+	/*
+	 * We preallocate 2 writesets.  When an era rolls over we
+	 * switch between them. This means the allocation is done at
+	 * preresume time, rather than on the io path.
+	 */
+	struct writeset writesets[2];
+	struct writeset *current_writeset;
+
+	dm_block_t writeset_tree_root;
+	dm_block_t era_array_root;
+
+	struct dm_disk_bitset bitset_info;
+	struct dm_btree_info writeset_tree_info;
+	struct dm_array_info era_array_info;
+
+	dm_block_t metadata_snap;
+
+	/*
+	 * A flag that is set whenever a writeset has been archived.
+	 */
+	bool archived_writesets;
+
+	/*
+	 * Reading the space map root can fail, so we read it into this
+	 * buffer before the superblock is locked and updated.
+	 */
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+};
+
+static int superblock_read_lock(struct era_metadata *md,
+				struct dm_block **sblock)
+{
+	return dm_bm_read_lock(md->bm, SUPERBLOCK_LOCATION,
+			       &sb_validator, sblock);
+}
+
+static int superblock_lock_zero(struct era_metadata *md,
+				struct dm_block **sblock)
+{
+	return dm_bm_write_lock_zero(md->bm, SUPERBLOCK_LOCATION,
+				     &sb_validator, sblock);
+}
+
+static int superblock_lock(struct era_metadata *md,
+			   struct dm_block **sblock)
+{
+	return dm_bm_write_lock(md->bm, SUPERBLOCK_LOCATION,
+				&sb_validator, sblock);
+}
+
+/* FIXME: duplication with cache and thin */
+static int superblock_all_zeroes(struct dm_block_manager *bm, bool *result)
+{
+	int r;
+	unsigned i;
+	struct dm_block *b;
+	__le64 *data_le, zero = cpu_to_le64(0);
+	unsigned sb_block_size = dm_bm_block_size(bm) / sizeof(__le64);
+
+	/*
+	 * We can't use a validator here - it may be all zeroes.
+	 */
+	r = dm_bm_read_lock(bm, SUPERBLOCK_LOCATION, NULL, &b);
+	if (r)
+		return r;
+
+	data_le = dm_block_data(b);
+	*result = true;
+	for (i = 0; i < sb_block_size; i++) {
+		if (data_le[i] != zero) {
+			*result = false;
+			break;
+		}
+	}
+
+	return dm_bm_unlock(b);
+}
+
+/*----------------------------------------------------------------*/
+
+static void ws_pack(const struct writeset_metadata *core, struct writeset_disk *disk)
+{
+	disk->nr_bits = cpu_to_le32(core->nr_bits);
+	disk->root = cpu_to_le64(core->root);
+}
+
+static void ws_unpack(const struct writeset_disk *disk, struct writeset_metadata *core)
+{
+	core->nr_bits = le32_to_cpu(disk->nr_bits);
+	core->root = le64_to_cpu(disk->root);
+}
+
+static void ws_inc(void *context, const void *value)
+{
+	struct era_metadata *md = context;
+	struct writeset_disk ws_d;
+	dm_block_t b;
+
+	memcpy(&ws_d, value, sizeof(ws_d));
+	b = le64_to_cpu(ws_d.root);
+
+	dm_tm_inc(md->tm, b);
+}
+
+static void ws_dec(void *context, const void *value)
+{
+	struct era_metadata *md = context;
+	struct writeset_disk ws_d;
+	dm_block_t b;
+
+	memcpy(&ws_d, value, sizeof(ws_d));
+	b = le64_to_cpu(ws_d.root);
+
+	dm_bitset_del(&md->bitset_info, b);
+}
+
+static int ws_eq(void *context, const void *value1, const void *value2)
+{
+	return !memcmp(value1, value2, sizeof(struct writeset_metadata));
+}
+
+/*----------------------------------------------------------------*/
+
+static void setup_writeset_tree_info(struct era_metadata *md)
+{
+	struct dm_btree_value_type *vt = &md->writeset_tree_info.value_type;
+	md->writeset_tree_info.tm = md->tm;
+	md->writeset_tree_info.levels = 1;
+	vt->context = md;
+	vt->size = sizeof(struct writeset_disk);
+	vt->inc = ws_inc;
+	vt->dec = ws_dec;
+	vt->equal = ws_eq;
+}
+
+static void setup_era_array_info(struct era_metadata *md)
+
+{
+	struct dm_btree_value_type vt;
+	vt.context = NULL;
+	vt.size = sizeof(__le32);
+	vt.inc = NULL;
+	vt.dec = NULL;
+	vt.equal = NULL;
+
+	dm_array_info_init(&md->era_array_info, md->tm, &vt);
+}
+
+static void setup_infos(struct era_metadata *md)
+{
+	dm_disk_bitset_init(md->tm, &md->bitset_info);
+	setup_writeset_tree_info(md);
+	setup_era_array_info(md);
+}
+
+/*----------------------------------------------------------------*/
+
+static int create_fresh_metadata(struct era_metadata *md)
+{
+	int r;
+
+	r = dm_tm_create_with_sm(md->bm, SUPERBLOCK_LOCATION,
+				 &md->tm, &md->sm);
+	if (r < 0) {
+		DMERR("dm_tm_create_with_sm failed");
+		return r;
+	}
+
+	setup_infos(md);
+
+	r = dm_btree_empty(&md->writeset_tree_info, &md->writeset_tree_root);
+	if (r) {
+		DMERR("couldn't create new writeset tree");
+		goto bad;
+	}
+
+	r = dm_array_empty(&md->era_array_info, &md->era_array_root);
+	if (r) {
+		DMERR("couldn't create era array");
+		goto bad;
+	}
+
+	return 0;
+
+bad:
+	dm_sm_destroy(md->sm);
+	dm_tm_destroy(md->tm);
+
+	return r;
+}
+
+static int save_sm_root(struct era_metadata *md)
+{
+	int r;
+	size_t metadata_len;
+
+	r = dm_sm_root_size(md->sm, &metadata_len);
+	if (r < 0)
+		return r;
+
+	return dm_sm_copy_root(md->sm, &md->metadata_space_map_root,
+			       metadata_len);
+}
+
+static void copy_sm_root(struct era_metadata *md, struct superblock_disk *disk)
+{
+	memcpy(&disk->metadata_space_map_root,
+	       &md->metadata_space_map_root,
+	       sizeof(md->metadata_space_map_root));
+}
+
+/*
+ * Writes a superblock, including the static fields that don't get updated
+ * with every commit (possible optimisation here).  'md' should be fully
+ * constructed when this is called.
+ */
+static void prepare_superblock(struct era_metadata *md, struct superblock_disk *disk)
+{
+	disk->magic = cpu_to_le64(SUPERBLOCK_MAGIC);
+	disk->flags = cpu_to_le32(0ul);
+
+	/* FIXME: can't keep blanking the uuid (uuid is currently unused though) */
+	memset(disk->uuid, 0, sizeof(disk->uuid));
+	disk->version = cpu_to_le32(MAX_ERA_VERSION);
+
+	copy_sm_root(md, disk);
+
+	disk->data_block_size = cpu_to_le32(md->block_size);
+	disk->metadata_block_size = cpu_to_le32(DM_ERA_METADATA_BLOCK_SIZE >> SECTOR_SHIFT);
+	disk->nr_blocks = cpu_to_le32(md->nr_blocks);
+	disk->current_era = cpu_to_le32(md->current_era);
+
+	ws_pack(&md->current_writeset->md, &disk->current_writeset);
+	disk->writeset_tree_root = cpu_to_le64(md->writeset_tree_root);
+	disk->era_array_root = cpu_to_le64(md->era_array_root);
+	disk->metadata_snap = cpu_to_le64(md->metadata_snap);
+}
+
+static int write_superblock(struct era_metadata *md)
+{
+	int r;
+	struct dm_block *sblock;
+	struct superblock_disk *disk;
+
+	r = save_sm_root(md);
+	if (r) {
+		DMERR("%s: save_sm_root failed", __func__);
+		return r;
+	}
+
+	r = superblock_lock_zero(md, &sblock);
+	if (r)
+		return r;
+
+	disk = dm_block_data(sblock);
+	prepare_superblock(md, disk);
+
+	return dm_tm_commit(md->tm, sblock);
+}
+
+/*
+ * Assumes block_size and the infos are set.
+ */
+static int format_metadata(struct era_metadata *md)
+{
+	int r;
+
+	r = create_fresh_metadata(md);
+	if (r)
+		return r;
+
+	r = write_superblock(md);
+	if (r) {
+		dm_sm_destroy(md->sm);
+		dm_tm_destroy(md->tm);
+		return r;
+	}
+
+	return 0;
+}
+
+static int open_metadata(struct era_metadata *md)
+{
+	int r;
+	struct dm_block *sblock;
+	struct superblock_disk *disk;
+
+	r = superblock_read_lock(md, &sblock);
+	if (r) {
+		DMERR("couldn't read_lock superblock");
+		return r;
+	}
+
+	disk = dm_block_data(sblock);
+	r = dm_tm_open_with_sm(md->bm, SUPERBLOCK_LOCATION,
+			       disk->metadata_space_map_root,
+			       sizeof(disk->metadata_space_map_root),
+			       &md->tm, &md->sm);
+	if (r) {
+		DMERR("dm_tm_open_with_sm failed");
+		goto bad;
+	}
+
+	setup_infos(md);
+
+	md->block_size = le32_to_cpu(disk->data_block_size);
+	md->nr_blocks = le32_to_cpu(disk->nr_blocks);
+	md->current_era = le32_to_cpu(disk->current_era);
+
+	md->writeset_tree_root = le64_to_cpu(disk->writeset_tree_root);
+	md->era_array_root = le64_to_cpu(disk->era_array_root);
+	md->metadata_snap = le64_to_cpu(disk->metadata_snap);
+	md->archived_writesets = true;
+
+	return dm_bm_unlock(sblock);
+
+bad:
+	dm_bm_unlock(sblock);
+	return r;
+}
+
+static int open_or_format_metadata(struct era_metadata *md,
+				   bool may_format)
+{
+	int r;
+	bool unformatted = false;
+
+	r = superblock_all_zeroes(md->bm, &unformatted);
+	if (r)
+		return r;
+
+	if (unformatted)
+		return may_format ? format_metadata(md) : -EPERM;
+
+	return open_metadata(md);
+}
+
+static int create_persistent_data_objects(struct era_metadata *md,
+					  bool may_format)
+{
+	int r;
+
+	md->bm = dm_block_manager_create(md->bdev, DM_ERA_METADATA_BLOCK_SIZE,
+					 DM_ERA_METADATA_CACHE_SIZE,
+					 ERA_MAX_CONCURRENT_LOCKS);
+	if (IS_ERR(md->bm)) {
+		DMERR("could not create block manager");
+		return PTR_ERR(md->bm);
+	}
+
+	r = open_or_format_metadata(md, may_format);
+	if (r)
+		dm_block_manager_destroy(md->bm);
+
+	return r;
+}
+
+static void destroy_persistent_data_objects(struct era_metadata *md)
+{
+	dm_sm_destroy(md->sm);
+	dm_tm_destroy(md->tm);
+	dm_block_manager_destroy(md->bm);
+}
+
+/*
+ * This waits until all era_map threads have picked up the new filter.
+ */
+static void swap_writeset(struct era_metadata *md, struct writeset *new_writeset)
+{
+	rcu_assign_pointer(md->current_writeset, new_writeset);
+	synchronize_rcu();
+}
+
+/*----------------------------------------------------------------
+ * Writesets get 'digested' into the main era array.
+ *
+ * We're using a coroutine here so the worker thread can do the digestion,
+ * thus avoiding synchronisation of the metadata.  Digesting a whole
+ * writeset in one go would cause too much latency.
+ *--------------------------------------------------------------*/
+struct digest {
+	uint32_t era;
+	unsigned nr_bits, current_bit;
+	struct writeset_metadata writeset;
+	__le32 value;
+	struct dm_disk_bitset info;
+
+	int (*step)(struct era_metadata *, struct digest *);
+};
+
+static int metadata_digest_lookup_writeset(struct era_metadata *md,
+					   struct digest *d);
+
+static int metadata_digest_remove_writeset(struct era_metadata *md,
+					   struct digest *d)
+{
+	int r;
+	uint64_t key = d->era;
+
+	r = dm_btree_remove(&md->writeset_tree_info, md->writeset_tree_root,
+			    &key, &md->writeset_tree_root);
+	if (r) {
+		DMERR("%s: dm_btree_remove failed", __func__);
+		return r;
+	}
+
+	d->step = metadata_digest_lookup_writeset;
+	return 0;
+}
+
+#define INSERTS_PER_STEP 100
+
+static int metadata_digest_transcribe_writeset(struct era_metadata *md,
+					       struct digest *d)
+{
+	int r;
+	bool marked;
+	unsigned b, e = min(d->current_bit + INSERTS_PER_STEP, d->nr_bits);
+
+	for (b = d->current_bit; b < e; b++) {
+		r = writeset_marked_on_disk(&d->info, &d->writeset, b, &marked);
+		if (r) {
+			DMERR("%s: writeset_marked_on_disk failed", __func__);
+			return r;
+		}
+
+		if (!marked)
+			continue;
+
+		__dm_bless_for_disk(&d->value);
+		r = dm_array_set_value(&md->era_array_info, md->era_array_root,
+				       b, &d->value, &md->era_array_root);
+		if (r) {
+			DMERR("%s: dm_array_set_value failed", __func__);
+			return r;
+		}
+	}
+
+	if (b == d->nr_bits)
+		d->step = metadata_digest_remove_writeset;
+	else
+		d->current_bit = b;
+
+	return 0;
+}
+
+static int metadata_digest_lookup_writeset(struct era_metadata *md,
+					   struct digest *d)
+{
+	int r;
+	uint64_t key;
+	struct writeset_disk disk;
+
+	r = dm_btree_find_lowest_key(&md->writeset_tree_info,
+				     md->writeset_tree_root, &key);
+	if (r < 0)
+		return r;
+
+	d->era = key;
+
+	r = dm_btree_lookup(&md->writeset_tree_info,
+			    md->writeset_tree_root, &key, &disk);
+	if (r) {
+		if (r == -ENODATA) {
+			d->step = NULL;
+			return 0;
+		}
+
+		DMERR("%s: dm_btree_lookup failed", __func__);
+		return r;
+	}
+
+	ws_unpack(&disk, &d->writeset);
+	d->value = cpu_to_le32(key);
+
+	d->nr_bits = min(d->writeset.nr_bits, md->nr_blocks);
+	d->current_bit = 0;
+	d->step = metadata_digest_transcribe_writeset;
+
+	return 0;
+}
+
+static int metadata_digest_start(struct era_metadata *md, struct digest *d)
+{
+	if (d->step)
+		return 0;
+
+	memset(d, 0, sizeof(*d));
+
+	/*
+	 * We initialise another bitset info to avoid any caching side
+	 * effects with the previous one.
+	 */
+	dm_disk_bitset_init(md->tm, &d->info);
+	d->step = metadata_digest_lookup_writeset;
+
+	return 0;
+}
+
+/*----------------------------------------------------------------
+ * High level metadata interface.  Target methods should use these, and not
+ * the lower level ones.
+ *--------------------------------------------------------------*/
+static struct era_metadata *metadata_open(struct block_device *bdev,
+					  sector_t block_size,
+					  bool may_format)
+{
+	int r;
+	struct era_metadata *md = kzalloc(sizeof(*md), GFP_KERNEL);
+
+	if (!md)
+		return NULL;
+
+	md->bdev = bdev;
+	md->block_size = block_size;
+
+	md->writesets[0].md.root = INVALID_WRITESET_ROOT;
+	md->writesets[1].md.root = INVALID_WRITESET_ROOT;
+	md->current_writeset = &md->writesets[0];
+
+	r = create_persistent_data_objects(md, may_format);
+	if (r) {
+		kfree(md);
+		return ERR_PTR(r);
+	}
+
+	return md;
+}
+
+static void metadata_close(struct era_metadata *md)
+{
+	destroy_persistent_data_objects(md);
+	kfree(md);
+}
+
+static bool valid_nr_blocks(dm_block_t n)
+{
+	/*
+	 * dm_bitset restricts us to 2^32.  test_bit & co. restrict us
+	 * further to 2^31 - 1
+	 */
+	return n < (1ull << 31);
+}
+
+static int metadata_resize(struct era_metadata *md, void *arg)
+{
+	int r;
+	dm_block_t *new_size = arg;
+	__le32 value;
+
+	if (!valid_nr_blocks(*new_size)) {
+		DMERR("Invalid number of origin blocks %llu",
+		      (unsigned long long) *new_size);
+		return -EINVAL;
+	}
+
+	writeset_free(&md->writesets[0]);
+	writeset_free(&md->writesets[1]);
+
+	r = writeset_alloc(&md->writesets[0], *new_size);
+	if (r) {
+		DMERR("%s: writeset_alloc failed for writeset 0", __func__);
+		return r;
+	}
+
+	r = writeset_alloc(&md->writesets[1], *new_size);
+	if (r) {
+		DMERR("%s: writeset_alloc failed for writeset 1", __func__);
+		return r;
+	}
+
+	value = cpu_to_le32(0u);
+	__dm_bless_for_disk(&value);
+	r = dm_array_resize(&md->era_array_info, md->era_array_root,
+			    md->nr_blocks, *new_size,
+			    &value, &md->era_array_root);
+	if (r) {
+		DMERR("%s: dm_array_resize failed", __func__);
+		return r;
+	}
+
+	md->nr_blocks = *new_size;
+	return 0;
+}
+
+static int metadata_era_archive(struct era_metadata *md)
+{
+	int r;
+	uint64_t keys[1];
+	struct writeset_disk value;
+
+	r = dm_bitset_flush(&md->bitset_info, md->current_writeset->md.root,
+			    &md->current_writeset->md.root);
+	if (r) {
+		DMERR("%s: dm_bitset_flush failed", __func__);
+		return r;
+	}
+
+	ws_pack(&md->current_writeset->md, &value);
+	md->current_writeset->md.root = INVALID_WRITESET_ROOT;
+
+	keys[0] = md->current_era;
+	__dm_bless_for_disk(&value);
+	r = dm_btree_insert(&md->writeset_tree_info, md->writeset_tree_root,
+			    keys, &value, &md->writeset_tree_root);
+	if (r) {
+		DMERR("%s: couldn't insert writeset into btree", __func__);
+		/* FIXME: fail mode */
+		return r;
+	}
+
+	md->archived_writesets = true;
+
+	return 0;
+}
+
+static struct writeset *next_writeset(struct era_metadata *md)
+{
+	return (md->current_writeset == &md->writesets[0]) ?
+		&md->writesets[1] : &md->writesets[0];
+}
+
+static int metadata_new_era(struct era_metadata *md)
+{
+	int r;
+	struct writeset *new_writeset = next_writeset(md);
+
+	r = writeset_init(&md->bitset_info, new_writeset);
+	if (r) {
+		DMERR("%s: writeset_init failed", __func__);
+		return r;
+	}
+
+	swap_writeset(md, new_writeset);
+	md->current_era++;
+
+	return 0;
+}
+
+static int metadata_era_rollover(struct era_metadata *md)
+{
+	int r;
+
+	if (md->current_writeset->md.root != INVALID_WRITESET_ROOT) {
+		r = metadata_era_archive(md);
+		if (r) {
+			DMERR("%s: metadata_archive_era failed", __func__);
+			/* FIXME: fail mode? */
+			return r;
+		}
+	}
+
+	r = metadata_new_era(md);
+	if (r) {
+		DMERR("%s: new era failed", __func__);
+		/* FIXME: fail mode */
+		return r;
+	}
+
+	return 0;
+}
+
+static bool metadata_current_marked(struct era_metadata *md, dm_block_t block)
+{
+	bool r;
+	struct writeset *ws;
+
+	rcu_read_lock();
+	ws = rcu_dereference(md->current_writeset);
+	r = writeset_marked(ws, block);
+	rcu_read_unlock();
+
+	return r;
+}
+
+static int metadata_commit(struct era_metadata *md)
+{
+	int r;
+	struct dm_block *sblock;
+
+	if (md->current_writeset->md.root != SUPERBLOCK_LOCATION) {
+		r = dm_bitset_flush(&md->bitset_info, md->current_writeset->md.root,
+				    &md->current_writeset->md.root);
+		if (r) {
+			DMERR("%s: bitset flush failed", __func__);
+			return r;
+		}
+	}
+
+	r = save_sm_root(md);
+	if (r) {
+		DMERR("%s: save_sm_root failed", __func__);
+		return r;
+	}
+
+	r = dm_tm_pre_commit(md->tm);
+	if (r) {
+		DMERR("%s: pre commit failed", __func__);
+		return r;
+	}
+
+	r = superblock_lock(md, &sblock);
+	if (r) {
+		DMERR("%s: superblock lock failed", __func__);
+		return r;
+	}
+
+	prepare_superblock(md, dm_block_data(sblock));
+
+	return dm_tm_commit(md->tm, sblock);
+}
+
+static int metadata_checkpoint(struct era_metadata *md)
+{
+	/*
+	 * For now we just rollover, but later I want to put a check in to
+	 * avoid this if the filter is still pretty fresh.
+	 */
+	return metadata_era_rollover(md);
+}
+
+/*
+ * Metadata snapshots allow userland to access era data.
+ */
+static int metadata_take_snap(struct era_metadata *md)
+{
+	int r, inc;
+	struct dm_block *clone;
+
+	if (md->metadata_snap != SUPERBLOCK_LOCATION) {
+		DMERR("%s: metadata snapshot already exists", __func__);
+		return -EINVAL;
+	}
+
+	r = metadata_era_rollover(md);
+	if (r) {
+		DMERR("%s: era rollover failed", __func__);
+		return r;
+	}
+
+	r = metadata_commit(md);
+	if (r) {
+		DMERR("%s: pre commit failed", __func__);
+		return r;
+	}
+
+	r = dm_sm_inc_block(md->sm, SUPERBLOCK_LOCATION);
+	if (r) {
+		DMERR("%s: couldn't increment superblock", __func__);
+		return r;
+	}
+
+	r = dm_tm_shadow_block(md->tm, SUPERBLOCK_LOCATION,
+			       &sb_validator, &clone, &inc);
+	if (r) {
+		DMERR("%s: couldn't shadow superblock", __func__);
+		dm_sm_dec_block(md->sm, SUPERBLOCK_LOCATION);
+		return r;
+	}
+	BUG_ON(!inc);
+
+	r = dm_sm_inc_block(md->sm, md->writeset_tree_root);
+	if (r) {
+		DMERR("%s: couldn't inc writeset tree root", __func__);
+		dm_tm_unlock(md->tm, clone);
+		return r;
+	}
+
+	r = dm_sm_inc_block(md->sm, md->era_array_root);
+	if (r) {
+		DMERR("%s: couldn't inc era tree root", __func__);
+		dm_sm_dec_block(md->sm, md->writeset_tree_root);
+		dm_tm_unlock(md->tm, clone);
+		return r;
+	}
+
+	md->metadata_snap = dm_block_location(clone);
+
+	r = dm_tm_unlock(md->tm, clone);
+	if (r) {
+		DMERR("%s: couldn't unlock clone", __func__);
+		md->metadata_snap = SUPERBLOCK_LOCATION;
+		return r;
+	}
+
+	return 0;
+}
+
+static int metadata_drop_snap(struct era_metadata *md)
+{
+	int r;
+	dm_block_t location;
+	struct dm_block *clone;
+	struct superblock_disk *disk;
+
+	if (md->metadata_snap == SUPERBLOCK_LOCATION) {
+		DMERR("%s: no snap to drop", __func__);
+		return -EINVAL;
+	}
+
+	r = dm_tm_read_lock(md->tm, md->metadata_snap, &sb_validator, &clone);
+	if (r) {
+		DMERR("%s: couldn't read lock superblock clone", __func__);
+		return r;
+	}
+
+	/*
+	 * Whatever happens now we'll commit with no record of the metadata
+	 * snap.
+	 */
+	md->metadata_snap = SUPERBLOCK_LOCATION;
+
+	disk = dm_block_data(clone);
+	r = dm_btree_del(&md->writeset_tree_info,
+			 le64_to_cpu(disk->writeset_tree_root));
+	if (r) {
+		DMERR("%s: error deleting writeset tree clone", __func__);
+		dm_tm_unlock(md->tm, clone);
+		return r;
+	}
+
+	r = dm_array_del(&md->era_array_info, le64_to_cpu(disk->era_array_root));
+	if (r) {
+		DMERR("%s: error deleting era array clone", __func__);
+		dm_tm_unlock(md->tm, clone);
+		return r;
+	}
+
+	location = dm_block_location(clone);
+	dm_tm_unlock(md->tm, clone);
+
+	return dm_sm_dec_block(md->sm, location);
+}
+
+struct metadata_stats {
+	dm_block_t used;
+	dm_block_t total;
+	dm_block_t snap;
+	uint32_t era;
+};
+
+static int metadata_get_stats(struct era_metadata *md, void *ptr)
+{
+	int r;
+	struct metadata_stats *s = ptr;
+	dm_block_t nr_free, nr_total;
+
+	r = dm_sm_get_nr_free(md->sm, &nr_free);
+	if (r) {
+		DMERR("dm_sm_get_nr_free returned %d", r);
+		return r;
+	}
+
+	r = dm_sm_get_nr_blocks(md->sm, &nr_total);
+	if (r) {
+		DMERR("dm_pool_get_metadata_dev_size returned %d", r);
+		return r;
+	}
+
+	s->used = nr_total - nr_free;
+	s->total = nr_total;
+	s->snap = md->metadata_snap;
+	s->era = md->current_era;
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
+struct era {
+	struct dm_target *ti;
+	struct dm_target_callbacks callbacks;
+
+	struct dm_dev *metadata_dev;
+	struct dm_dev *origin_dev;
+
+	dm_block_t nr_blocks;
+	uint32_t sectors_per_block;
+	int sectors_per_block_shift;
+	struct era_metadata *md;
+
+	struct workqueue_struct *wq;
+	struct work_struct worker;
+
+	spinlock_t deferred_lock;
+	struct bio_list deferred_bios;
+
+	spinlock_t rpc_lock;
+	struct list_head rpc_calls;
+
+	struct digest digest;
+	atomic_t suspended;
+};
+
+struct rpc {
+	struct list_head list;
+
+	int (*fn0)(struct era_metadata *);
+	int (*fn1)(struct era_metadata *, void *);
+	void *arg;
+	int result;
+
+	struct completion complete;
+};
+
+/*----------------------------------------------------------------
+ * Remapping.
+ *---------------------------------------------------------------*/
+static bool block_size_is_power_of_two(struct era *era)
+{
+	return era->sectors_per_block_shift >= 0;
+}
+
+static dm_block_t get_block(struct era *era, struct bio *bio)
+{
+	sector_t block_nr = bio->bi_iter.bi_sector;
+
+	if (!block_size_is_power_of_two(era))
+		(void) sector_div(block_nr, era->sectors_per_block);
+	else
+		block_nr >>= era->sectors_per_block_shift;
+
+	return block_nr;
+}
+
+static void remap_to_origin(struct era *era, struct bio *bio)
+{
+	bio->bi_bdev = era->origin_dev->bdev;
+}
+
+/*----------------------------------------------------------------
+ * Worker thread
+ *--------------------------------------------------------------*/
+static void wake_worker(struct era *era)
+{
+	if (!atomic_read(&era->suspended))
+		queue_work(era->wq, &era->worker);
+}
+
+static void process_old_eras(struct era *era)
+{
+	int r;
+
+	if (!era->digest.step)
+		return;
+
+	r = era->digest.step(era->md, &era->digest);
+	if (r < 0) {
+		DMERR("%s: digest step failed, stopping digestion", __func__);
+		era->digest.step = NULL;
+
+	} else if (era->digest.step)
+		wake_worker(era);
+}
+
+static void process_deferred_bios(struct era *era)
+{
+	int r;
+	struct bio_list deferred_bios, marked_bios;
+	struct bio *bio;
+	bool commit_needed = false;
+	bool failed = false;
+
+	bio_list_init(&deferred_bios);
+	bio_list_init(&marked_bios);
+
+	spin_lock(&era->deferred_lock);
+	bio_list_merge(&deferred_bios, &era->deferred_bios);
+	bio_list_init(&era->deferred_bios);
+	spin_unlock(&era->deferred_lock);
+
+	while ((bio = bio_list_pop(&deferred_bios))) {
+		r = writeset_test_and_set(&era->md->bitset_info,
+					  era->md->current_writeset,
+					  get_block(era, bio));
+		if (r < 0) {
+			/*
+			 * This is bad news, we need to rollback.
+			 * FIXME: finish.
+			 */
+			failed = true;
+
+		} else if (r == 0)
+			commit_needed = true;
+
+		bio_list_add(&marked_bios, bio);
+	}
+
+	if (commit_needed) {
+		r = metadata_commit(era->md);
+		if (r)
+			failed = true;
+	}
+
+	if (failed)
+		while ((bio = bio_list_pop(&marked_bios)))
+			bio_io_error(bio);
+	else
+		while ((bio = bio_list_pop(&marked_bios)))
+			generic_make_request(bio);
+}
+
+static void process_rpc_calls(struct era *era)
+{
+	int r;
+	bool need_commit = false;
+	struct list_head calls;
+	struct rpc *rpc, *tmp;
+
+	INIT_LIST_HEAD(&calls);
+	spin_lock(&era->rpc_lock);
+	list_splice_init(&era->rpc_calls, &calls);
+	spin_unlock(&era->rpc_lock);
+
+	list_for_each_entry_safe(rpc, tmp, &calls, list) {
+		rpc->result = rpc->fn0 ? rpc->fn0(era->md) : rpc->fn1(era->md, rpc->arg);
+		need_commit = true;
+	}
+
+	if (need_commit) {
+		r = metadata_commit(era->md);
+		if (r)
+			list_for_each_entry_safe(rpc, tmp, &calls, list)
+				rpc->result = r;
+	}
+
+	list_for_each_entry_safe(rpc, tmp, &calls, list)
+		complete(&rpc->complete);
+}
+
+static void kick_off_digest(struct era *era)
+{
+	if (era->md->archived_writesets) {
+		era->md->archived_writesets = false;
+		metadata_digest_start(era->md, &era->digest);
+	}
+}
+
+static void do_work(struct work_struct *ws)
+{
+	struct era *era = container_of(ws, struct era, worker);
+
+	kick_off_digest(era);
+	process_old_eras(era);
+	process_deferred_bios(era);
+	process_rpc_calls(era);
+}
+
+static void defer_bio(struct era *era, struct bio *bio)
+{
+	spin_lock(&era->deferred_lock);
+	bio_list_add(&era->deferred_bios, bio);
+	spin_unlock(&era->deferred_lock);
+
+	wake_worker(era);
+}
+
+/*
+ * Make an rpc call to the worker to change the metadata.
+ */
+static int perform_rpc(struct era *era, struct rpc *rpc)
+{
+	rpc->result = 0;
+	init_completion(&rpc->complete);
+
+	spin_lock(&era->rpc_lock);
+	list_add(&rpc->list, &era->rpc_calls);
+	spin_unlock(&era->rpc_lock);
+
+	wake_worker(era);
+	wait_for_completion(&rpc->complete);
+
+	return rpc->result;
+}
+
+static int in_worker0(struct era *era, int (*fn)(struct era_metadata *))
+{
+	struct rpc rpc;
+	rpc.fn0 = fn;
+	rpc.fn1 = NULL;
+
+	return perform_rpc(era, &rpc);
+}
+
+static int in_worker1(struct era *era,
+		      int (*fn)(struct era_metadata *, void *), void *arg)
+{
+	struct rpc rpc;
+	rpc.fn0 = NULL;
+	rpc.fn1 = fn;
+	rpc.arg = arg;
+
+	return perform_rpc(era, &rpc);
+}
+
+static void start_worker(struct era *era)
+{
+	atomic_set(&era->suspended, 0);
+}
+
+static void stop_worker(struct era *era)
+{
+	atomic_set(&era->suspended, 1);
+	flush_workqueue(era->wq);
+}
+
+/*----------------------------------------------------------------
+ * Target methods
+ *--------------------------------------------------------------*/
+static int dev_is_congested(struct dm_dev *dev, int bdi_bits)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+	return bdi_congested(&q->backing_dev_info, bdi_bits);
+}
+
+static int era_is_congested(struct dm_target_callbacks *cb, int bdi_bits)
+{
+	struct era *era = container_of(cb, struct era, callbacks);
+	return dev_is_congested(era->origin_dev, bdi_bits);
+}
+
+static void era_destroy(struct era *era)
+{
+	if (era->md)
+		metadata_close(era->md);
+
+	if (era->wq)
+		destroy_workqueue(era->wq);
+
+	if (era->origin_dev)
+		dm_put_device(era->ti, era->origin_dev);
+
+	if (era->metadata_dev)
+		dm_put_device(era->ti, era->metadata_dev);
+
+	kfree(era);
+}
+
+static dm_block_t calc_nr_blocks(struct era *era)
+{
+	return dm_sector_div_up(era->ti->len, era->sectors_per_block);
+}
+
+static bool valid_block_size(dm_block_t block_size)
+{
+	bool greater_than_zero = block_size > 0;
+	bool multiple_of_min_block_size = (block_size & (MIN_BLOCK_SIZE - 1)) == 0;
+
+	return greater_than_zero && multiple_of_min_block_size;
+}
+
+/*
+ * <metadata dev> <data dev> <data block size (sectors)>
+ */
+static int era_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int r;
+	char dummy;
+	struct era *era;
+	struct era_metadata *md;
+
+	if (argc != 3) {
+		ti->error = "Invalid argument count";
+		return -EINVAL;
+	}
+
+	era = kzalloc(sizeof(*era), GFP_KERNEL);
+	if (!era) {
+		ti->error = "Error allocating era structure";
+		return -ENOMEM;
+	}
+
+	era->ti = ti;
+
+	r = dm_get_device(ti, argv[0], FMODE_READ | FMODE_WRITE, &era->metadata_dev);
+	if (r) {
+		ti->error = "Error opening metadata device";
+		era_destroy(era);
+		return -EINVAL;
+	}
+
+	r = dm_get_device(ti, argv[1], FMODE_READ | FMODE_WRITE, &era->origin_dev);
+	if (r) {
+		ti->error = "Error opening data device";
+		era_destroy(era);
+		return -EINVAL;
+	}
+
+	r = sscanf(argv[2], "%u%c", &era->sectors_per_block, &dummy);
+	if (r != 1) {
+		ti->error = "Error parsing block size";
+		era_destroy(era);
+		return -EINVAL;
+	}
+
+	r = dm_set_target_max_io_len(ti, era->sectors_per_block);
+	if (r) {
+		ti->error = "could not set max io len";
+		era_destroy(era);
+		return -EINVAL;
+	}
+
+	if (!valid_block_size(era->sectors_per_block)) {
+		ti->error = "Invalid block size";
+		era_destroy(era);
+		return -EINVAL;
+	}
+	if (era->sectors_per_block & (era->sectors_per_block - 1))
+		era->sectors_per_block_shift = -1;
+	else
+		era->sectors_per_block_shift = __ffs(era->sectors_per_block);
+
+	md = metadata_open(era->metadata_dev->bdev, era->sectors_per_block, true);
+	if (IS_ERR(md)) {
+		ti->error = "Error reading metadata";
+		era_destroy(era);
+		return PTR_ERR(md);
+	}
+	era->md = md;
+
+	era->nr_blocks = calc_nr_blocks(era);
+
+	r = metadata_resize(era->md, &era->nr_blocks);
+	if (r) {
+		ti->error = "couldn't resize metadata";
+		era_destroy(era);
+		return -ENOMEM;
+	}
+
+	era->wq = alloc_ordered_workqueue("dm-" DM_MSG_PREFIX, WQ_MEM_RECLAIM);
+	if (!era->wq) {
+		ti->error = "could not create workqueue for metadata object";
+		era_destroy(era);
+		return -ENOMEM;
+	}
+	INIT_WORK(&era->worker, do_work);
+
+	spin_lock_init(&era->deferred_lock);
+	bio_list_init(&era->deferred_bios);
+
+	spin_lock_init(&era->rpc_lock);
+	INIT_LIST_HEAD(&era->rpc_calls);
+
+	ti->private = era;
+	ti->num_flush_bios = 1;
+	ti->flush_supported = true;
+
+	ti->num_discard_bios = 1;
+	ti->discards_supported = true;
+	era->callbacks.congested_fn = era_is_congested;
+	dm_table_add_target_callbacks(ti->table, &era->callbacks);
+
+	return 0;
+}
+
+static void era_dtr(struct dm_target *ti)
+{
+	era_destroy(ti->private);
+}
+
+static int era_map(struct dm_target *ti, struct bio *bio)
+{
+	struct era *era = ti->private;
+	dm_block_t block = get_block(era, bio);
+
+	/*
+	 * All bios get remapped to the origin device.  We do this now, but
+	 * it may not get issued until later.  Depending on whether the
+	 * block is marked in this era.
+	 */
+	remap_to_origin(era, bio);
+
+	/*
+	 * REQ_FLUSH bios carry no data, so we're not interested in them.
+	 */
+	if (!(bio->bi_rw & REQ_FLUSH) &&
+	    (bio_data_dir(bio) == WRITE) &&
+	    !metadata_current_marked(era->md, block)) {
+		defer_bio(era, bio);
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	return DM_MAPIO_REMAPPED;
+}
+
+static void era_postsuspend(struct dm_target *ti)
+{
+	int r;
+	struct era *era = ti->private;
+
+	r = in_worker0(era, metadata_era_archive);
+	if (r) {
+		DMERR("%s: couldn't archive current era", __func__);
+		/* FIXME: fail mode */
+	}
+
+	stop_worker(era);
+}
+
+static int era_preresume(struct dm_target *ti)
+{
+	int r;
+	struct era *era = ti->private;
+	dm_block_t new_size = calc_nr_blocks(era);
+
+	if (era->nr_blocks != new_size) {
+		r = in_worker1(era, metadata_resize, &new_size);
+		if (r)
+			return r;
+
+		era->nr_blocks = new_size;
+	}
+
+	start_worker(era);
+
+	r = in_worker0(era, metadata_new_era);
+	if (r) {
+		DMERR("%s: metadata_era_rollover failed", __func__);
+		return r;
+	}
+
+	return 0;
+}
+
+/*
+ * Status format:
+ *
+ * <metadata block size> <#used metadata blocks>/<#total metadata blocks>
+ * <current era> <held metadata root | '-'>
+ */
+static void era_status(struct dm_target *ti, status_type_t type,
+		       unsigned status_flags, char *result, unsigned maxlen)
+{
+	int r;
+	struct era *era = ti->private;
+	ssize_t sz = 0;
+	struct metadata_stats stats;
+	char buf[BDEVNAME_SIZE];
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		r = in_worker1(era, metadata_get_stats, &stats);
+		if (r)
+			goto err;
+
+		DMEMIT("%u %llu/%llu %u",
+		       (unsigned) (DM_ERA_METADATA_BLOCK_SIZE >> SECTOR_SHIFT),
+		       (unsigned long long) stats.used,
+		       (unsigned long long) stats.total,
+		       (unsigned) stats.era);
+
+		if (stats.snap != SUPERBLOCK_LOCATION)
+			DMEMIT(" %llu", stats.snap);
+		else
+			DMEMIT(" -");
+		break;
+
+	case STATUSTYPE_TABLE:
+		format_dev_t(buf, era->metadata_dev->bdev->bd_dev);
+		DMEMIT("%s ", buf);
+		format_dev_t(buf, era->origin_dev->bdev->bd_dev);
+		DMEMIT("%s %u", buf, era->sectors_per_block);
+		break;
+	}
+
+	return;
+
+err:
+	DMEMIT("Error");
+}
+
+static int era_message(struct dm_target *ti, unsigned argc, char **argv)
+{
+	struct era *era = ti->private;
+
+	if (argc != 1) {
+		DMERR("incorrect number of message arguments");
+		return -EINVAL;
+	}
+
+	if (!strcasecmp(argv[0], "checkpoint"))
+		return in_worker0(era, metadata_checkpoint);
+
+	if (!strcasecmp(argv[0], "take_metadata_snap"))
+		return in_worker0(era, metadata_take_snap);
+
+	if (!strcasecmp(argv[0], "drop_metadata_snap"))
+		return in_worker0(era, metadata_drop_snap);
+
+	DMERR("unsupported message '%s'", argv[0]);
+	return -EINVAL;
+}
+
+static sector_t get_dev_size(struct dm_dev *dev)
+{
+	return i_size_read(dev->bdev->bd_inode) >> SECTOR_SHIFT;
+}
+
+static int era_iterate_devices(struct dm_target *ti,
+			       iterate_devices_callout_fn fn, void *data)
+{
+	struct era *era = ti->private;
+	return fn(ti, era->origin_dev, 0, get_dev_size(era->origin_dev), data);
+}
+
+static int era_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+		     struct bio_vec *biovec, int max_size)
+{
+	struct era *era = ti->private;
+	struct request_queue *q = bdev_get_queue(era->origin_dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = era->origin_dev->bdev;
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static void era_io_hints(struct dm_target *ti, struct queue_limits *limits)
+{
+	struct era *era = ti->private;
+	uint64_t io_opt_sectors = limits->io_opt >> SECTOR_SHIFT;
+
+	/*
+	 * If the system-determined stacked limits are compatible with the
+	 * era device's blocksize (io_opt is a factor) do not override them.
+	 */
+	if (io_opt_sectors < era->sectors_per_block ||
+	    do_div(io_opt_sectors, era->sectors_per_block)) {
+		blk_limits_io_min(limits, 0);
+		blk_limits_io_opt(limits, era->sectors_per_block << SECTOR_SHIFT);
+	}
+}
+
+/*----------------------------------------------------------------*/
+
+static struct target_type era_target = {
+	.name = "era",
+	.version = {1, 0, 0},
+	.module = THIS_MODULE,
+	.ctr = era_ctr,
+	.dtr = era_dtr,
+	.map = era_map,
+	.postsuspend = era_postsuspend,
+	.preresume = era_preresume,
+	.status = era_status,
+	.message = era_message,
+	.iterate_devices = era_iterate_devices,
+	.merge = era_merge,
+	.io_hints = era_io_hints
+};
+
+static int __init dm_era_init(void)
+{
+	int r;
+
+	r = dm_register_target(&era_target);
+	if (r) {
+		DMERR("era target registration failed: %d", r);
+		return r;
+	}
+
+	return 0;
+}
+
+static void __exit dm_era_exit(void)
+{
+	dm_unregister_target(&era_target);
+}
+
+module_init(dm_era_init);
+module_exit(dm_era_exit);
+
+MODULE_DESCRIPTION(DM_NAME " era target");
+MODULE_AUTHOR("Joe Thornber <ejt@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-exception-store.c b/drivers/md/dm-exception-store.c
index 41f408068a7..ebaa4f803ee 100644
--- a/drivers/md/dm-exception-store.c
+++ b/drivers/md/dm-exception-store.c
@@ -1,709 +1,290 @@
 /*
- * dm-exception-store.c
- *
  * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
- * Copyright (C) 2006 Red Hat GmbH
+ * Copyright (C) 2006-2008 Red Hat GmbH
  *
  * This file is released under the GPL.
  */
 
-#include "dm.h"
-#include "dm-snap.h"
+#include "dm-exception-store.h"
 
+#include <linux/ctype.h>
 #include <linux/mm.h>
 #include <linux/pagemap.h>
 #include <linux/vmalloc.h>
+#include <linux/module.h>
 #include <linux/slab.h>
-#include <linux/dm-io.h>
-#include <linux/dm-kcopyd.h>
-
-#define DM_MSG_PREFIX "snapshots"
-#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32	/* 16KB */
-
-/*-----------------------------------------------------------------
- * Persistent snapshots, by persistent we mean that the snapshot
- * will survive a reboot.
- *---------------------------------------------------------------*/
-
-/*
- * We need to store a record of which parts of the origin have
- * been copied to the snapshot device.  The snapshot code
- * requires that we copy exception chunks to chunk aligned areas
- * of the COW store.  It makes sense therefore, to store the
- * metadata in chunk size blocks.
- *
- * There is no backward or forward compatibility implemented,
- * snapshots with different disk versions than the kernel will
- * not be usable.  It is expected that "lvcreate" will blank out
- * the start of a fresh COW device before calling the snapshot
- * constructor.
- *
- * The first chunk of the COW device just contains the header.
- * After this there is a chunk filled with exception metadata,
- * followed by as many exception chunks as can fit in the
- * metadata areas.
- *
- * All on disk structures are in little-endian format.  The end
- * of the exceptions info is indicated by an exception with a
- * new_chunk of 0, which is invalid since it would point to the
- * header chunk.
- */
-
-/*
- * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
- */
-#define SNAP_MAGIC 0x70416e53
-
-/*
- * The on-disk version of the metadata.
- */
-#define SNAPSHOT_DISK_VERSION 1
-
-struct disk_header {
-	uint32_t magic;
-
-	/*
-	 * Is this snapshot valid.  There is no way of recovering
-	 * an invalid snapshot.
-	 */
-	uint32_t valid;
-
-	/*
-	 * Simple, incrementing version. no backward
-	 * compatibility.
-	 */
-	uint32_t version;
-
-	/* In sectors */
-	uint32_t chunk_size;
-};
 
-struct disk_exception {
-	uint64_t old_chunk;
-	uint64_t new_chunk;
-};
+#define DM_MSG_PREFIX "snapshot exception stores"
 
-struct commit_callback {
-	void (*callback)(void *, int success);
-	void *context;
-};
+static LIST_HEAD(_exception_store_types);
+static DEFINE_SPINLOCK(_lock);
 
-/*
- * The top level structure for a persistent exception store.
- */
-struct pstore {
-	struct dm_snapshot *snap;	/* up pointer to my snapshot */
-	int version;
-	int valid;
-	uint32_t exceptions_per_area;
-
-	/*
-	 * Now that we have an asynchronous kcopyd there is no
-	 * need for large chunk sizes, so it wont hurt to have a
-	 * whole chunks worth of metadata in memory at once.
-	 */
-	void *area;
-
-	/*
-	 * Used to keep track of which metadata area the data in
-	 * 'chunk' refers to.
-	 */
-	uint32_t current_area;
-
-	/*
-	 * The next free chunk for an exception.
-	 */
-	uint32_t next_free;
-
-	/*
-	 * The index of next free exception in the current
-	 * metadata area.
-	 */
-	uint32_t current_committed;
-
-	atomic_t pending_count;
-	uint32_t callback_count;
-	struct commit_callback *callbacks;
-	struct dm_io_client *io_client;
-
-	struct workqueue_struct *metadata_wq;
-};
-
-static unsigned sectors_to_pages(unsigned sectors)
-{
-	return DIV_ROUND_UP(sectors, PAGE_SIZE >> 9);
-}
-
-static int alloc_area(struct pstore *ps)
+static struct dm_exception_store_type *__find_exception_store_type(const char *name)
 {
-	int r = -ENOMEM;
-	size_t len;
+	struct dm_exception_store_type *type;
 
-	len = ps->snap->chunk_size << SECTOR_SHIFT;
+	list_for_each_entry(type, &_exception_store_types, list)
+		if (!strcmp(name, type->name))
+			return type;
 
-	/*
-	 * Allocate the chunk_size block of memory that will hold
-	 * a single metadata area.
-	 */
-	ps->area = vmalloc(len);
-	if (!ps->area)
-		return r;
-
-	return 0;
+	return NULL;
 }
 
-static void free_area(struct pstore *ps)
+static struct dm_exception_store_type *_get_exception_store_type(const char *name)
 {
-	vfree(ps->area);
-	ps->area = NULL;
-}
+	struct dm_exception_store_type *type;
 
-struct mdata_req {
-	struct dm_io_region *where;
-	struct dm_io_request *io_req;
-	struct work_struct work;
-	int result;
-};
+	spin_lock(&_lock);
 
-static void do_metadata(struct work_struct *work)
-{
-	struct mdata_req *req = container_of(work, struct mdata_req, work);
+	type = __find_exception_store_type(name);
 
-	req->result = dm_io(req->io_req, 1, req->where, NULL);
-}
+	if (type && !try_module_get(type->module))
+		type = NULL;
 
-/*
- * Read or write a chunk aligned and sized block of data from a device.
- */
-static int chunk_io(struct pstore *ps, uint32_t chunk, int rw, int metadata)
-{
-	struct dm_io_region where = {
-		.bdev = ps->snap->cow->bdev,
-		.sector = ps->snap->chunk_size * chunk,
-		.count = ps->snap->chunk_size,
-	};
-	struct dm_io_request io_req = {
-		.bi_rw = rw,
-		.mem.type = DM_IO_VMA,
-		.mem.ptr.vma = ps->area,
-		.client = ps->io_client,
-		.notify.fn = NULL,
-	};
-	struct mdata_req req;
-
-	if (!metadata)
-		return dm_io(&io_req, 1, &where, NULL);
-
-	req.where = &where;
-	req.io_req = &io_req;
-
-	/*
-	 * Issue the synchronous I/O from a different thread
-	 * to avoid generic_make_request recursion.
-	 */
-	INIT_WORK(&req.work, do_metadata);
-	queue_work(ps->metadata_wq, &req.work);
-	flush_workqueue(ps->metadata_wq);
-
-	return req.result;
+	spin_unlock(&_lock);
+
+	return type;
 }
 
 /*
- * Read or write a metadata area.  Remembering to skip the first
- * chunk which holds the header.
+ * get_type
+ * @type_name
+ *
+ * Attempt to retrieve the dm_exception_store_type by name.  If not already
+ * available, attempt to load the appropriate module.
+ *
+ * Exstore modules are named "dm-exstore-" followed by the 'type_name'.
+ * Modules may contain multiple types.
+ * This function will first try the module "dm-exstore-<type_name>",
+ * then truncate 'type_name' on the last '-' and try again.
+ *
+ * For example, if type_name was "clustered-shared", it would search
+ * 'dm-exstore-clustered-shared' then 'dm-exstore-clustered'.
+ *
+ * 'dm-exception-store-<type_name>' is too long of a name in my
+ * opinion, which is why I've chosen to have the files
+ * containing exception store implementations be 'dm-exstore-<type_name>'.
+ * If you want your module to be autoloaded, you will follow this
+ * naming convention.
+ *
+ * Returns: dm_exception_store_type* on success, NULL on failure
  */
-static int area_io(struct pstore *ps, uint32_t area, int rw)
-{
-	int r;
-	uint32_t chunk;
-
-	/* convert a metadata area index to a chunk index */
-	chunk = 1 + ((ps->exceptions_per_area + 1) * area);
-
-	r = chunk_io(ps, chunk, rw, 0);
-	if (r)
-		return r;
-
-	ps->current_area = area;
-	return 0;
-}
-
-static int zero_area(struct pstore *ps, uint32_t area)
+static struct dm_exception_store_type *get_type(const char *type_name)
 {
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
-	return area_io(ps, area, WRITE);
-}
-
-static int read_header(struct pstore *ps, int *new_snapshot)
-{
-	int r;
-	struct disk_header *dh;
-	chunk_t chunk_size;
-	int chunk_size_supplied = 1;
-
-	/*
-	 * Use default chunk size (or hardsect_size, if larger) if none supplied
-	 */
-	if (!ps->snap->chunk_size) {
-        	ps->snap->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
-		    bdev_hardsect_size(ps->snap->cow->bdev) >> 9);
-		ps->snap->chunk_mask = ps->snap->chunk_size - 1;
-		ps->snap->chunk_shift = ffs(ps->snap->chunk_size) - 1;
-		chunk_size_supplied = 0;
-	}
-
-	ps->io_client = dm_io_client_create(sectors_to_pages(ps->snap->
-							     chunk_size));
-	if (IS_ERR(ps->io_client))
-		return PTR_ERR(ps->io_client);
-
-	r = alloc_area(ps);
-	if (r)
-		return r;
+	char *p, *type_name_dup;
+	struct dm_exception_store_type *type;
 
-	r = chunk_io(ps, 0, READ, 1);
-	if (r)
-		goto bad;
-
-	dh = (struct disk_header *) ps->area;
+	type = _get_exception_store_type(type_name);
+	if (type)
+		return type;
 
-	if (le32_to_cpu(dh->magic) == 0) {
-		*new_snapshot = 1;
-		return 0;
+	type_name_dup = kstrdup(type_name, GFP_KERNEL);
+	if (!type_name_dup) {
+		DMERR("No memory left to attempt load for \"%s\"", type_name);
+		return NULL;
 	}
 
-	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
-		DMWARN("Invalid or corrupt snapshot");
-		r = -ENXIO;
-		goto bad;
+	while (request_module("dm-exstore-%s", type_name_dup) ||
+	       !(type = _get_exception_store_type(type_name))) {
+		p = strrchr(type_name_dup, '-');
+		if (!p)
+			break;
+		p[0] = '\0';
 	}
 
-	*new_snapshot = 0;
-	ps->valid = le32_to_cpu(dh->valid);
-	ps->version = le32_to_cpu(dh->version);
-	chunk_size = le32_to_cpu(dh->chunk_size);
-
-	if (!chunk_size_supplied || ps->snap->chunk_size == chunk_size)
-		return 0;
-
-	DMWARN("chunk size %llu in device metadata overrides "
-	       "table chunk size of %llu.",
-	       (unsigned long long)chunk_size,
-	       (unsigned long long)ps->snap->chunk_size);
+	if (!type)
+		DMWARN("Module for exstore type \"%s\" not found.", type_name);
 
-	/* We had a bogus chunk_size. Fix stuff up. */
-	free_area(ps);
+	kfree(type_name_dup);
 
-	ps->snap->chunk_size = chunk_size;
-	ps->snap->chunk_mask = chunk_size - 1;
-	ps->snap->chunk_shift = ffs(chunk_size) - 1;
-
-	r = dm_io_client_resize(sectors_to_pages(ps->snap->chunk_size),
-				ps->io_client);
-	if (r)
-		return r;
-
-	r = alloc_area(ps);
-	return r;
-
-bad:
-	free_area(ps);
-	return r;
+	return type;
 }
 
-static int write_header(struct pstore *ps)
+static void put_type(struct dm_exception_store_type *type)
 {
-	struct disk_header *dh;
-
-	memset(ps->area, 0, ps->snap->chunk_size << SECTOR_SHIFT);
-
-	dh = (struct disk_header *) ps->area;
-	dh->magic = cpu_to_le32(SNAP_MAGIC);
-	dh->valid = cpu_to_le32(ps->valid);
-	dh->version = cpu_to_le32(ps->version);
-	dh->chunk_size = cpu_to_le32(ps->snap->chunk_size);
-
-	return chunk_io(ps, 0, WRITE, 1);
+	spin_lock(&_lock);
+	module_put(type->module);
+	spin_unlock(&_lock);
 }
 
-/*
- * Access functions for the disk exceptions, these do the endian conversions.
- */
-static struct disk_exception *get_exception(struct pstore *ps, uint32_t index)
+int dm_exception_store_type_register(struct dm_exception_store_type *type)
 {
-	BUG_ON(index >= ps->exceptions_per_area);
-
-	return ((struct disk_exception *) ps->area) + index;
-}
+	int r = 0;
 
-static void read_exception(struct pstore *ps,
-			   uint32_t index, struct disk_exception *result)
-{
-	struct disk_exception *e = get_exception(ps, index);
+	spin_lock(&_lock);
+	if (!__find_exception_store_type(type->name))
+		list_add(&type->list, &_exception_store_types);
+	else
+		r = -EEXIST;
+	spin_unlock(&_lock);
 
-	/* copy it */
-	result->old_chunk = le64_to_cpu(e->old_chunk);
-	result->new_chunk = le64_to_cpu(e->new_chunk);
+	return r;
 }
+EXPORT_SYMBOL(dm_exception_store_type_register);
 
-static void write_exception(struct pstore *ps,
-			    uint32_t index, struct disk_exception *de)
+int dm_exception_store_type_unregister(struct dm_exception_store_type *type)
 {
-	struct disk_exception *e = get_exception(ps, index);
+	spin_lock(&_lock);
 
-	/* copy it */
-	e->old_chunk = cpu_to_le64(de->old_chunk);
-	e->new_chunk = cpu_to_le64(de->new_chunk);
-}
-
-/*
- * Registers the exceptions that are present in the current area.
- * 'full' is filled in to indicate if the area has been
- * filled.
- */
-static int insert_exceptions(struct pstore *ps, int *full)
-{
-	int r;
-	unsigned int i;
-	struct disk_exception de;
-
-	/* presume the area is full */
-	*full = 1;
-
-	for (i = 0; i < ps->exceptions_per_area; i++) {
-		read_exception(ps, i, &de);
-
-		/*
-		 * If the new_chunk is pointing at the start of
-		 * the COW device, where the first metadata area
-		 * is we know that we've hit the end of the
-		 * exceptions.  Therefore the area is not full.
-		 */
-		if (de.new_chunk == 0LL) {
-			ps->current_committed = i;
-			*full = 0;
-			break;
-		}
-
-		/*
-		 * Keep track of the start of the free chunks.
-		 */
-		if (ps->next_free <= de.new_chunk)
-			ps->next_free = de.new_chunk + 1;
-
-		/*
-		 * Otherwise we add the exception to the snapshot.
-		 */
-		r = dm_add_exception(ps->snap, de.old_chunk, de.new_chunk);
-		if (r)
-			return r;
+	if (!__find_exception_store_type(type->name)) {
+		spin_unlock(&_lock);
+		return -EINVAL;
 	}
 
-	return 0;
-}
+	list_del(&type->list);
 
-static int read_exceptions(struct pstore *ps)
-{
-	uint32_t area;
-	int r, full = 1;
-
-	/*
-	 * Keeping reading chunks and inserting exceptions until
-	 * we find a partially full area.
-	 */
-	for (area = 0; full; area++) {
-		r = area_io(ps, area, READ);
-		if (r)
-			return r;
-
-		r = insert_exceptions(ps, &full);
-		if (r)
-			return r;
-	}
+	spin_unlock(&_lock);
 
 	return 0;
 }
+EXPORT_SYMBOL(dm_exception_store_type_unregister);
 
-static struct pstore *get_info(struct exception_store *store)
+static int set_chunk_size(struct dm_exception_store *store,
+			  const char *chunk_size_arg, char **error)
 {
-	return (struct pstore *) store->context;
-}
+	unsigned chunk_size;
 
-static void persistent_fraction_full(struct exception_store *store,
-				     sector_t *numerator, sector_t *denominator)
-{
-	*numerator = get_info(store)->next_free * store->snap->chunk_size;
-	*denominator = get_dev_size(store->snap->cow->bdev);
-}
-
-static void persistent_destroy(struct exception_store *store)
-{
-	struct pstore *ps = get_info(store);
-
-	destroy_workqueue(ps->metadata_wq);
-	dm_io_client_destroy(ps->io_client);
-	vfree(ps->callbacks);
-	free_area(ps);
-	kfree(ps);
-}
-
-static int persistent_read_metadata(struct exception_store *store)
-{
-	int r, uninitialized_var(new_snapshot);
-	struct pstore *ps = get_info(store);
-
-	/*
-	 * Read the snapshot header.
-	 */
-	r = read_header(ps, &new_snapshot);
-	if (r)
-		return r;
-
-	/*
-	 * Now we know correct chunk_size, complete the initialisation.
-	 */
-	ps->exceptions_per_area = (ps->snap->chunk_size << SECTOR_SHIFT) /
-				  sizeof(struct disk_exception);
-	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
-			sizeof(*ps->callbacks));
-	if (!ps->callbacks)
-		return -ENOMEM;
+	if (kstrtouint(chunk_size_arg, 10, &chunk_size)) {
+		*error = "Invalid chunk size";
+		return -EINVAL;
+	}
 
-	/*
-	 * Do we need to setup a new snapshot ?
-	 */
-	if (new_snapshot) {
-		r = write_header(ps);
-		if (r) {
-			DMWARN("write_header failed");
-			return r;
-		}
-
-		r = zero_area(ps, 0);
-		if (r) {
-			DMWARN("zero_area(0) failed");
-			return r;
-		}
-
-	} else {
-		/*
-		 * Sanity checks.
-		 */
-		if (ps->version != SNAPSHOT_DISK_VERSION) {
-			DMWARN("unable to handle snapshot disk version %d",
-			       ps->version);
-			return -EINVAL;
-		}
-
-		/*
-		 * Metadata are valid, but snapshot is invalidated
-		 */
-		if (!ps->valid)
-			return 1;
-
-		/*
-		 * Read the metadata.
-		 */
-		r = read_exceptions(ps);
-		if (r)
-			return r;
+	if (!chunk_size) {
+		store->chunk_size = store->chunk_mask = store->chunk_shift = 0;
+		return 0;
 	}
 
-	return 0;
+	return dm_exception_store_set_chunk_size(store, chunk_size, error);
 }
 
-static int persistent_prepare(struct exception_store *store,
-			      struct dm_snap_exception *e)
+int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
+				      unsigned chunk_size,
+				      char **error)
 {
-	struct pstore *ps = get_info(store);
-	uint32_t stride;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
+	/* Check chunk_size is a power of 2 */
+	if (!is_power_of_2(chunk_size)) {
+		*error = "Chunk size is not a power of 2";
+		return -EINVAL;
+	}
 
-	/* Is there enough room ? */
-	if (size < ((ps->next_free + 1) * store->snap->chunk_size))
-		return -ENOSPC;
+	/* Validate the chunk size against the device block size */
+	if (chunk_size %
+	    (bdev_logical_block_size(dm_snap_cow(store->snap)->bdev) >> 9) ||
+	    chunk_size %
+	    (bdev_logical_block_size(dm_snap_origin(store->snap)->bdev) >> 9)) {
+		*error = "Chunk size is not a multiple of device blocksize";
+		return -EINVAL;
+	}
 
-	e->new_chunk = ps->next_free;
+	if (chunk_size > INT_MAX >> SECTOR_SHIFT) {
+		*error = "Chunk size is too high";
+		return -EINVAL;
+	}
 
-	/*
-	 * Move onto the next free pending, making sure to take
-	 * into account the location of the metadata chunks.
-	 */
-	stride = (ps->exceptions_per_area + 1);
-	if ((++ps->next_free % stride) == 1)
-		ps->next_free++;
+	store->chunk_size = chunk_size;
+	store->chunk_mask = chunk_size - 1;
+	store->chunk_shift = ffs(chunk_size) - 1;
 
-	atomic_inc(&ps->pending_count);
 	return 0;
 }
 
-static void persistent_commit(struct exception_store *store,
-			      struct dm_snap_exception *e,
-			      void (*callback) (void *, int success),
-			      void *callback_context)
+int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
+			      unsigned *args_used,
+			      struct dm_exception_store **store)
 {
-	int r;
-	unsigned int i;
-	struct pstore *ps = get_info(store);
-	struct disk_exception de;
-	struct commit_callback *cb;
-
-	de.old_chunk = e->old_chunk;
-	de.new_chunk = e->new_chunk;
-	write_exception(ps, ps->current_committed++, &de);
-
-	/*
-	 * Add the callback to the back of the array.  This code
-	 * is the only place where the callback array is
-	 * manipulated, and we know that it will never be called
-	 * multiple times concurrently.
-	 */
-	cb = ps->callbacks + ps->callback_count++;
-	cb->callback = callback;
-	cb->context = callback_context;
-
-	/*
-	 * If there are no more exceptions in flight, or we have
-	 * filled this metadata area we commit the exceptions to
-	 * disk.
-	 */
-	if (atomic_dec_and_test(&ps->pending_count) ||
-	    (ps->current_committed == ps->exceptions_per_area)) {
-		r = area_io(ps, ps->current_area, WRITE);
-		if (r)
-			ps->valid = 0;
-
-		/*
-		 * Have we completely filled the current area ?
-		 */
-		if (ps->current_committed == ps->exceptions_per_area) {
-			ps->current_committed = 0;
-			r = zero_area(ps, ps->current_area + 1);
-			if (r)
-				ps->valid = 0;
-		}
-
-		for (i = 0; i < ps->callback_count; i++) {
-			cb = ps->callbacks + i;
-			cb->callback(cb->context, r == 0 ? 1 : 0);
-		}
-
-		ps->callback_count = 0;
+	int r = 0;
+	struct dm_exception_store_type *type = NULL;
+	struct dm_exception_store *tmp_store;
+	char persistent;
+
+	if (argc < 2) {
+		ti->error = "Insufficient exception store arguments";
+		return -EINVAL;
 	}
-}
 
-static void persistent_drop(struct exception_store *store)
-{
-	struct pstore *ps = get_info(store);
+	tmp_store = kmalloc(sizeof(*tmp_store), GFP_KERNEL);
+	if (!tmp_store) {
+		ti->error = "Exception store allocation failed";
+		return -ENOMEM;
+	}
 
-	ps->valid = 0;
-	if (write_header(ps))
-		DMWARN("write header failed");
-}
+	persistent = toupper(*argv[0]);
+	if (persistent == 'P')
+		type = get_type("P");
+	else if (persistent == 'N')
+		type = get_type("N");
+	else {
+		ti->error = "Persistent flag is not P or N";
+		r = -EINVAL;
+		goto bad_type;
+	}
 
-int dm_create_persistent(struct exception_store *store)
-{
-	struct pstore *ps;
+	if (!type) {
+		ti->error = "Exception store type not recognised";
+		r = -EINVAL;
+		goto bad_type;
+	}
 
-	/* allocate the pstore */
-	ps = kmalloc(sizeof(*ps), GFP_KERNEL);
-	if (!ps)
-		return -ENOMEM;
+	tmp_store->type = type;
+	tmp_store->snap = snap;
 
-	ps->snap = store->snap;
-	ps->valid = 1;
-	ps->version = SNAPSHOT_DISK_VERSION;
-	ps->area = NULL;
-	ps->next_free = 2;	/* skipping the header and first area */
-	ps->current_committed = 0;
-
-	ps->callback_count = 0;
-	atomic_set(&ps->pending_count, 0);
-	ps->callbacks = NULL;
-
-	ps->metadata_wq = create_singlethread_workqueue("ksnaphd");
-	if (!ps->metadata_wq) {
-		kfree(ps);
-		DMERR("couldn't start header metadata update thread");
-		return -ENOMEM;
-	}
+	r = set_chunk_size(tmp_store, argv[1], &ti->error);
+	if (r)
+		goto bad;
 
-	store->destroy = persistent_destroy;
-	store->read_metadata = persistent_read_metadata;
-	store->prepare_exception = persistent_prepare;
-	store->commit_exception = persistent_commit;
-	store->drop_snapshot = persistent_drop;
-	store->fraction_full = persistent_fraction_full;
-	store->context = ps;
+	r = type->ctr(tmp_store, 0, NULL);
+	if (r) {
+		ti->error = "Exception store type constructor failed";
+		goto bad;
+	}
 
+	*args_used = 2;
+	*store = tmp_store;
 	return 0;
-}
 
-/*-----------------------------------------------------------------
- * Implementation of the store for non-persistent snapshots.
- *---------------------------------------------------------------*/
-struct transient_c {
-	sector_t next_free;
-};
-
-static void transient_destroy(struct exception_store *store)
-{
-	kfree(store->context);
+bad:
+	put_type(type);
+bad_type:
+	kfree(tmp_store);
+	return r;
 }
+EXPORT_SYMBOL(dm_exception_store_create);
 
-static int transient_read_metadata(struct exception_store *store)
+void dm_exception_store_destroy(struct dm_exception_store *store)
 {
-	return 0;
+	store->type->dtr(store);
+	put_type(store->type);
+	kfree(store);
 }
+EXPORT_SYMBOL(dm_exception_store_destroy);
 
-static int transient_prepare(struct exception_store *store,
-			     struct dm_snap_exception *e)
+int dm_exception_store_init(void)
 {
-	struct transient_c *tc = (struct transient_c *) store->context;
-	sector_t size = get_dev_size(store->snap->cow->bdev);
+	int r;
 
-	if (size < (tc->next_free + store->snap->chunk_size))
-		return -1;
+	r = dm_transient_snapshot_init();
+	if (r) {
+		DMERR("Unable to register transient exception store type.");
+		goto transient_fail;
+	}
 
-	e->new_chunk = sector_to_chunk(store->snap, tc->next_free);
-	tc->next_free += store->snap->chunk_size;
+	r = dm_persistent_snapshot_init();
+	if (r) {
+		DMERR("Unable to register persistent exception store type");
+		goto persistent_fail;
+	}
 
 	return 0;
-}
-
-static void transient_commit(struct exception_store *store,
-			     struct dm_snap_exception *e,
-			     void (*callback) (void *, int success),
-			     void *callback_context)
-{
-	/* Just succeed */
-	callback(callback_context, 1);
-}
 
-static void transient_fraction_full(struct exception_store *store,
-				    sector_t *numerator, sector_t *denominator)
-{
-	*numerator = ((struct transient_c *) store->context)->next_free;
-	*denominator = get_dev_size(store->snap->cow->bdev);
+persistent_fail:
+	dm_transient_snapshot_exit();
+transient_fail:
+	return r;
 }
 
-int dm_create_transient(struct exception_store *store)
+void dm_exception_store_exit(void)
 {
-	struct transient_c *tc;
-
-	store->destroy = transient_destroy;
-	store->read_metadata = transient_read_metadata;
-	store->prepare_exception = transient_prepare;
-	store->commit_exception = transient_commit;
-	store->drop_snapshot = NULL;
-	store->fraction_full = transient_fraction_full;
-
-	tc = kmalloc(sizeof(struct transient_c), GFP_KERNEL);
-	if (!tc)
-		return -ENOMEM;
-
-	tc->next_free = 0;
-	store->context = tc;
-
-	return 0;
+	dm_persistent_snapshot_exit();
+	dm_transient_snapshot_exit();
 }
diff --git a/drivers/md/dm-exception-store.h b/drivers/md/dm-exception-store.h
new file mode 100644
index 00000000000..0b2536247cf
--- /dev/null
+++ b/drivers/md/dm-exception-store.h
@@ -0,0 +1,227 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2008 Red Hat, Inc. All rights reserved.
+ *
+ * Device-mapper snapshot exception store.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_EXCEPTION_STORE
+#define _LINUX_DM_EXCEPTION_STORE
+
+#include <linux/blkdev.h>
+#include <linux/device-mapper.h>
+
+/*
+ * The snapshot code deals with largish chunks of the disk at a
+ * time. Typically 32k - 512k.
+ */
+typedef sector_t chunk_t;
+
+/*
+ * An exception is used where an old chunk of data has been
+ * replaced by a new one.
+ * If chunk_t is 64 bits in size, the top 8 bits of new_chunk hold the number
+ * of chunks that follow contiguously.  Remaining bits hold the number of the
+ * chunk within the device.
+ */
+struct dm_exception {
+	struct list_head hash_list;
+
+	chunk_t old_chunk;
+	chunk_t new_chunk;
+};
+
+/*
+ * Abstraction to handle the meta/layout of exception stores (the
+ * COW device).
+ */
+struct dm_exception_store;
+struct dm_exception_store_type {
+	const char *name;
+	struct module *module;
+
+	int (*ctr) (struct dm_exception_store *store,
+		    unsigned argc, char **argv);
+
+	/*
+	 * Destroys this object when you've finished with it.
+	 */
+	void (*dtr) (struct dm_exception_store *store);
+
+	/*
+	 * The target shouldn't read the COW device until this is
+	 * called.  As exceptions are read from the COW, they are
+	 * reported back via the callback.
+	 */
+	int (*read_metadata) (struct dm_exception_store *store,
+			      int (*callback)(void *callback_context,
+					      chunk_t old, chunk_t new),
+			      void *callback_context);
+
+	/*
+	 * Find somewhere to store the next exception.
+	 */
+	int (*prepare_exception) (struct dm_exception_store *store,
+				  struct dm_exception *e);
+
+	/*
+	 * Update the metadata with this exception.
+	 */
+	void (*commit_exception) (struct dm_exception_store *store,
+				  struct dm_exception *e,
+				  void (*callback) (void *, int success),
+				  void *callback_context);
+
+	/*
+	 * Returns 0 if the exception store is empty.
+	 *
+	 * If there are exceptions still to be merged, sets
+	 * *last_old_chunk and *last_new_chunk to the most recent
+	 * still-to-be-merged chunk and returns the number of
+	 * consecutive previous ones.
+	 */
+	int (*prepare_merge) (struct dm_exception_store *store,
+			      chunk_t *last_old_chunk, chunk_t *last_new_chunk);
+
+	/*
+	 * Clear the last n exceptions.
+	 * nr_merged must be <= the value returned by prepare_merge.
+	 */
+	int (*commit_merge) (struct dm_exception_store *store, int nr_merged);
+
+	/*
+	 * The snapshot is invalid, note this in the metadata.
+	 */
+	void (*drop_snapshot) (struct dm_exception_store *store);
+
+	unsigned (*status) (struct dm_exception_store *store,
+			    status_type_t status, char *result,
+			    unsigned maxlen);
+
+	/*
+	 * Return how full the snapshot is.
+	 */
+	void (*usage) (struct dm_exception_store *store,
+		       sector_t *total_sectors, sector_t *sectors_allocated,
+		       sector_t *metadata_sectors);
+
+	/* For internal device-mapper use only. */
+	struct list_head list;
+};
+
+struct dm_snapshot;
+
+struct dm_exception_store {
+	struct dm_exception_store_type *type;
+	struct dm_snapshot *snap;
+
+	/* Size of data blocks saved - must be a power of 2 */
+	unsigned chunk_size;
+	unsigned chunk_mask;
+	unsigned chunk_shift;
+
+	void *context;
+};
+
+/*
+ * Obtain the origin or cow device used by a given snapshot.
+ */
+struct dm_dev *dm_snap_origin(struct dm_snapshot *snap);
+struct dm_dev *dm_snap_cow(struct dm_snapshot *snap);
+
+/*
+ * Funtions to manipulate consecutive chunks
+ */
+#  if defined(CONFIG_LBDAF) || (BITS_PER_LONG == 64)
+#    define DM_CHUNK_CONSECUTIVE_BITS 8
+#    define DM_CHUNK_NUMBER_BITS 56
+
+static inline chunk_t dm_chunk_number(chunk_t chunk)
+{
+	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
+}
+
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
+{
+	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
+}
+
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
+{
+	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
+
+	BUG_ON(!dm_consecutive_chunk_count(e));
+}
+
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
+{
+	BUG_ON(!dm_consecutive_chunk_count(e));
+
+	e->new_chunk -= (1ULL << DM_CHUNK_NUMBER_BITS);
+}
+
+#  else
+#    define DM_CHUNK_CONSECUTIVE_BITS 0
+
+static inline chunk_t dm_chunk_number(chunk_t chunk)
+{
+	return chunk;
+}
+
+static inline unsigned dm_consecutive_chunk_count(struct dm_exception *e)
+{
+	return 0;
+}
+
+static inline void dm_consecutive_chunk_count_inc(struct dm_exception *e)
+{
+}
+
+static inline void dm_consecutive_chunk_count_dec(struct dm_exception *e)
+{
+}
+
+#  endif
+
+/*
+ * Return the number of sectors in the device.
+ */
+static inline sector_t get_dev_size(struct block_device *bdev)
+{
+	return i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
+}
+
+static inline chunk_t sector_to_chunk(struct dm_exception_store *store,
+				      sector_t sector)
+{
+	return sector >> store->chunk_shift;
+}
+
+int dm_exception_store_type_register(struct dm_exception_store_type *type);
+int dm_exception_store_type_unregister(struct dm_exception_store_type *type);
+
+int dm_exception_store_set_chunk_size(struct dm_exception_store *store,
+				      unsigned chunk_size,
+				      char **error);
+
+int dm_exception_store_create(struct dm_target *ti, int argc, char **argv,
+			      struct dm_snapshot *snap,
+			      unsigned *args_used,
+			      struct dm_exception_store **store);
+void dm_exception_store_destroy(struct dm_exception_store *store);
+
+int dm_exception_store_init(void);
+void dm_exception_store_exit(void);
+
+/*
+ * Two exception store implementations.
+ */
+int dm_persistent_snapshot_init(void);
+void dm_persistent_snapshot_exit(void);
+
+int dm_transient_snapshot_init(void);
+void dm_transient_snapshot_exit(void);
+
+#endif /* _LINUX_DM_EXCEPTION_STORE */
diff --git a/drivers/md/dm-flakey.c b/drivers/md/dm-flakey.c
new file mode 100644
index 00000000000..b257e46876d
--- /dev/null
+++ b/drivers/md/dm-flakey.c
@@ -0,0 +1,447 @@
+/*
+ * Copyright (C) 2003 Sistina Software (UK) Limited.
+ * Copyright (C) 2004, 2010-2011 Red Hat, Inc. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/device-mapper.h>
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/blkdev.h>
+#include <linux/bio.h>
+#include <linux/slab.h>
+
+#define DM_MSG_PREFIX "flakey"
+
+#define all_corrupt_bio_flags_match(bio, fc)	\
+	(((bio)->bi_rw & (fc)->corrupt_bio_flags) == (fc)->corrupt_bio_flags)
+
+/*
+ * Flakey: Used for testing only, simulates intermittent,
+ * catastrophic device failure.
+ */
+struct flakey_c {
+	struct dm_dev *dev;
+	unsigned long start_time;
+	sector_t start;
+	unsigned up_interval;
+	unsigned down_interval;
+	unsigned long flags;
+	unsigned corrupt_bio_byte;
+	unsigned corrupt_bio_rw;
+	unsigned corrupt_bio_value;
+	unsigned corrupt_bio_flags;
+};
+
+enum feature_flag_bits {
+	DROP_WRITES
+};
+
+struct per_bio_data {
+	bool bio_submitted;
+};
+
+static int parse_features(struct dm_arg_set *as, struct flakey_c *fc,
+			  struct dm_target *ti)
+{
+	int r;
+	unsigned argc;
+	const char *arg_name;
+
+	static struct dm_arg _args[] = {
+		{0, 6, "Invalid number of feature args"},
+		{1, UINT_MAX, "Invalid corrupt bio byte"},
+		{0, 255, "Invalid corrupt value to write into bio byte (0-255)"},
+		{0, UINT_MAX, "Invalid corrupt bio flags mask"},
+	};
+
+	/* No feature arguments supplied. */
+	if (!as->argc)
+		return 0;
+
+	r = dm_read_arg_group(_args, as, &argc, &ti->error);
+	if (r)
+		return r;
+
+	while (argc) {
+		arg_name = dm_shift_arg(as);
+		argc--;
+
+		/*
+		 * drop_writes
+		 */
+		if (!strcasecmp(arg_name, "drop_writes")) {
+			if (test_and_set_bit(DROP_WRITES, &fc->flags)) {
+				ti->error = "Feature drop_writes duplicated";
+				return -EINVAL;
+			}
+
+			continue;
+		}
+
+		/*
+		 * corrupt_bio_byte <Nth_byte> <direction> <value> <bio_flags>
+		 */
+		if (!strcasecmp(arg_name, "corrupt_bio_byte")) {
+			if (!argc) {
+				ti->error = "Feature corrupt_bio_byte requires parameters";
+				return -EINVAL;
+			}
+
+			r = dm_read_arg(_args + 1, as, &fc->corrupt_bio_byte, &ti->error);
+			if (r)
+				return r;
+			argc--;
+
+			/*
+			 * Direction r or w?
+			 */
+			arg_name = dm_shift_arg(as);
+			if (!strcasecmp(arg_name, "w"))
+				fc->corrupt_bio_rw = WRITE;
+			else if (!strcasecmp(arg_name, "r"))
+				fc->corrupt_bio_rw = READ;
+			else {
+				ti->error = "Invalid corrupt bio direction (r or w)";
+				return -EINVAL;
+			}
+			argc--;
+
+			/*
+			 * Value of byte (0-255) to write in place of correct one.
+			 */
+			r = dm_read_arg(_args + 2, as, &fc->corrupt_bio_value, &ti->error);
+			if (r)
+				return r;
+			argc--;
+
+			/*
+			 * Only corrupt bios with these flags set.
+			 */
+			r = dm_read_arg(_args + 3, as, &fc->corrupt_bio_flags, &ti->error);
+			if (r)
+				return r;
+			argc--;
+
+			continue;
+		}
+
+		ti->error = "Unrecognised flakey feature requested";
+		return -EINVAL;
+	}
+
+	if (test_bit(DROP_WRITES, &fc->flags) && (fc->corrupt_bio_rw == WRITE)) {
+		ti->error = "drop_writes is incompatible with corrupt_bio_byte with the WRITE flag set";
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/*
+ * Construct a flakey mapping:
+ * <dev_path> <offset> <up interval> <down interval> [<#feature args> [<arg>]*]
+ *
+ *   Feature args:
+ *     [drop_writes]
+ *     [corrupt_bio_byte <Nth_byte> <direction> <value> <bio_flags>]
+ *
+ *   Nth_byte starts from 1 for the first byte.
+ *   Direction is r for READ or w for WRITE.
+ *   bio_flags is ignored if 0.
+ */
+static int flakey_ctr(struct dm_target *ti, unsigned int argc, char **argv)
+{
+	static struct dm_arg _args[] = {
+		{0, UINT_MAX, "Invalid up interval"},
+		{0, UINT_MAX, "Invalid down interval"},
+	};
+
+	int r;
+	struct flakey_c *fc;
+	unsigned long long tmpll;
+	struct dm_arg_set as;
+	const char *devname;
+	char dummy;
+
+	as.argc = argc;
+	as.argv = argv;
+
+	if (argc < 4) {
+		ti->error = "Invalid argument count";
+		return -EINVAL;
+	}
+
+	fc = kzalloc(sizeof(*fc), GFP_KERNEL);
+	if (!fc) {
+		ti->error = "Cannot allocate context";
+		return -ENOMEM;
+	}
+	fc->start_time = jiffies;
+
+	devname = dm_shift_arg(&as);
+
+	if (sscanf(dm_shift_arg(&as), "%llu%c", &tmpll, &dummy) != 1) {
+		ti->error = "Invalid device sector";
+		goto bad;
+	}
+	fc->start = tmpll;
+
+	r = dm_read_arg(_args, &as, &fc->up_interval, &ti->error);
+	if (r)
+		goto bad;
+
+	r = dm_read_arg(_args, &as, &fc->down_interval, &ti->error);
+	if (r)
+		goto bad;
+
+	if (!(fc->up_interval + fc->down_interval)) {
+		ti->error = "Total (up + down) interval is zero";
+		goto bad;
+	}
+
+	if (fc->up_interval + fc->down_interval < fc->up_interval) {
+		ti->error = "Interval overflow";
+		goto bad;
+	}
+
+	r = parse_features(&as, fc, ti);
+	if (r)
+		goto bad;
+
+	if (dm_get_device(ti, devname, dm_table_get_mode(ti->table), &fc->dev)) {
+		ti->error = "Device lookup failed";
+		goto bad;
+	}
+
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->per_bio_data_size = sizeof(struct per_bio_data);
+	ti->private = fc;
+	return 0;
+
+bad:
+	kfree(fc);
+	return -EINVAL;
+}
+
+static void flakey_dtr(struct dm_target *ti)
+{
+	struct flakey_c *fc = ti->private;
+
+	dm_put_device(ti, fc->dev);
+	kfree(fc);
+}
+
+static sector_t flakey_map_sector(struct dm_target *ti, sector_t bi_sector)
+{
+	struct flakey_c *fc = ti->private;
+
+	return fc->start + dm_target_offset(ti, bi_sector);
+}
+
+static void flakey_map_bio(struct dm_target *ti, struct bio *bio)
+{
+	struct flakey_c *fc = ti->private;
+
+	bio->bi_bdev = fc->dev->bdev;
+	if (bio_sectors(bio))
+		bio->bi_iter.bi_sector =
+			flakey_map_sector(ti, bio->bi_iter.bi_sector);
+}
+
+static void corrupt_bio_data(struct bio *bio, struct flakey_c *fc)
+{
+	unsigned bio_bytes = bio_cur_bytes(bio);
+	char *data = bio_data(bio);
+
+	/*
+	 * Overwrite the Nth byte of the data returned.
+	 */
+	if (data && bio_bytes >= fc->corrupt_bio_byte) {
+		data[fc->corrupt_bio_byte - 1] = fc->corrupt_bio_value;
+
+		DMDEBUG("Corrupting data bio=%p by writing %u to byte %u "
+			"(rw=%c bi_rw=%lu bi_sector=%llu cur_bytes=%u)\n",
+			bio, fc->corrupt_bio_value, fc->corrupt_bio_byte,
+			(bio_data_dir(bio) == WRITE) ? 'w' : 'r', bio->bi_rw,
+			(unsigned long long)bio->bi_iter.bi_sector, bio_bytes);
+	}
+}
+
+static int flakey_map(struct dm_target *ti, struct bio *bio)
+{
+	struct flakey_c *fc = ti->private;
+	unsigned elapsed;
+	struct per_bio_data *pb = dm_per_bio_data(bio, sizeof(struct per_bio_data));
+	pb->bio_submitted = false;
+
+	/* Are we alive ? */
+	elapsed = (jiffies - fc->start_time) / HZ;
+	if (elapsed % (fc->up_interval + fc->down_interval) >= fc->up_interval) {
+		/*
+		 * Flag this bio as submitted while down.
+		 */
+		pb->bio_submitted = true;
+
+		/*
+		 * Map reads as normal.
+		 */
+		if (bio_data_dir(bio) == READ)
+			goto map_bio;
+
+		/*
+		 * Drop writes?
+		 */
+		if (test_bit(DROP_WRITES, &fc->flags)) {
+			bio_endio(bio, 0);
+			return DM_MAPIO_SUBMITTED;
+		}
+
+		/*
+		 * Corrupt matching writes.
+		 */
+		if (fc->corrupt_bio_byte && (fc->corrupt_bio_rw == WRITE)) {
+			if (all_corrupt_bio_flags_match(bio, fc))
+				corrupt_bio_data(bio, fc);
+			goto map_bio;
+		}
+
+		/*
+		 * By default, error all I/O.
+		 */
+		return -EIO;
+	}
+
+map_bio:
+	flakey_map_bio(ti, bio);
+
+	return DM_MAPIO_REMAPPED;
+}
+
+static int flakey_end_io(struct dm_target *ti, struct bio *bio, int error)
+{
+	struct flakey_c *fc = ti->private;
+	struct per_bio_data *pb = dm_per_bio_data(bio, sizeof(struct per_bio_data));
+
+	/*
+	 * Corrupt successful READs while in down state.
+	 * If flags were specified, only corrupt those that match.
+	 */
+	if (fc->corrupt_bio_byte && !error && pb->bio_submitted &&
+	    (bio_data_dir(bio) == READ) && (fc->corrupt_bio_rw == READ) &&
+	    all_corrupt_bio_flags_match(bio, fc))
+		corrupt_bio_data(bio, fc);
+
+	return error;
+}
+
+static void flakey_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
+{
+	unsigned sz = 0;
+	struct flakey_c *fc = ti->private;
+	unsigned drop_writes;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		result[0] = '\0';
+		break;
+
+	case STATUSTYPE_TABLE:
+		DMEMIT("%s %llu %u %u ", fc->dev->name,
+		       (unsigned long long)fc->start, fc->up_interval,
+		       fc->down_interval);
+
+		drop_writes = test_bit(DROP_WRITES, &fc->flags);
+		DMEMIT("%u ", drop_writes + (fc->corrupt_bio_byte > 0) * 5);
+
+		if (drop_writes)
+			DMEMIT("drop_writes ");
+
+		if (fc->corrupt_bio_byte)
+			DMEMIT("corrupt_bio_byte %u %c %u %u ",
+			       fc->corrupt_bio_byte,
+			       (fc->corrupt_bio_rw == WRITE) ? 'w' : 'r',
+			       fc->corrupt_bio_value, fc->corrupt_bio_flags);
+
+		break;
+	}
+}
+
+static int flakey_ioctl(struct dm_target *ti, unsigned int cmd, unsigned long arg)
+{
+	struct flakey_c *fc = ti->private;
+	struct dm_dev *dev = fc->dev;
+	int r = 0;
+
+	/*
+	 * Only pass ioctls through if the device sizes match exactly.
+	 */
+	if (fc->start ||
+	    ti->len != i_size_read(dev->bdev->bd_inode) >> SECTOR_SHIFT)
+		r = scsi_verify_blk_ioctl(NULL, cmd);
+
+	return r ? : __blkdev_driver_ioctl(dev->bdev, dev->mode, cmd, arg);
+}
+
+static int flakey_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+			struct bio_vec *biovec, int max_size)
+{
+	struct flakey_c *fc = ti->private;
+	struct request_queue *q = bdev_get_queue(fc->dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = fc->dev->bdev;
+	bvm->bi_sector = flakey_map_sector(ti, bvm->bi_sector);
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static int flakey_iterate_devices(struct dm_target *ti, iterate_devices_callout_fn fn, void *data)
+{
+	struct flakey_c *fc = ti->private;
+
+	return fn(ti, fc->dev, fc->start, ti->len, data);
+}
+
+static struct target_type flakey_target = {
+	.name   = "flakey",
+	.version = {1, 3, 1},
+	.module = THIS_MODULE,
+	.ctr    = flakey_ctr,
+	.dtr    = flakey_dtr,
+	.map    = flakey_map,
+	.end_io = flakey_end_io,
+	.status = flakey_status,
+	.ioctl	= flakey_ioctl,
+	.merge	= flakey_merge,
+	.iterate_devices = flakey_iterate_devices,
+};
+
+static int __init dm_flakey_init(void)
+{
+	int r = dm_register_target(&flakey_target);
+
+	if (r < 0)
+		DMERR("register failed %d", r);
+
+	return r;
+}
+
+static void __exit dm_flakey_exit(void)
+{
+	dm_unregister_target(&flakey_target);
+}
+
+/* Module hooks */
+module_init(dm_flakey_init);
+module_exit(dm_flakey_exit);
+
+MODULE_DESCRIPTION(DM_NAME " flakey target");
+MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-io.c b/drivers/md/dm-io.c
index 4789c42d9a3..db404a0f7e2 100644
--- a/drivers/md/dm-io.c
+++ b/drivers/md/dm-io.c
@@ -7,56 +7,59 @@
 
 #include "dm.h"
 
+#include <linux/device-mapper.h>
+
 #include <linux/bio.h>
+#include <linux/completion.h>
 #include <linux/mempool.h>
 #include <linux/module.h>
 #include <linux/sched.h>
 #include <linux/slab.h>
 #include <linux/dm-io.h>
 
+#define DM_MSG_PREFIX "io"
+
+#define DM_IO_MAX_REGIONS	BITS_PER_LONG
+
 struct dm_io_client {
 	mempool_t *pool;
 	struct bio_set *bios;
 };
 
-/* FIXME: can we shrink this ? */
+/*
+ * Aligning 'struct io' reduces the number of bits required to store
+ * its address.  Refer to store_io_and_region_in_bio() below.
+ */
 struct io {
 	unsigned long error_bits;
 	atomic_t count;
-	struct task_struct *sleeper;
+	struct completion *wait;
 	struct dm_io_client *client;
 	io_notify_fn callback;
 	void *context;
-};
-
-/*
- * io contexts are only dynamically allocated for asynchronous
- * io.  Since async io is likely to be the majority of io we'll
- * have the same number of io contexts as bios! (FIXME: must reduce this).
- */
+	void *vma_invalidate_address;
+	unsigned long vma_invalidate_size;
+} __attribute__((aligned(DM_IO_MAX_REGIONS)));
 
-static unsigned int pages_to_ios(unsigned int pages)
-{
-	return 4 * pages;	/* too many ? */
-}
+static struct kmem_cache *_dm_io_cache;
 
 /*
  * Create a client with mempool and bioset.
  */
-struct dm_io_client *dm_io_client_create(unsigned num_pages)
+struct dm_io_client *dm_io_client_create(void)
 {
-	unsigned ios = pages_to_ios(num_pages);
 	struct dm_io_client *client;
+	unsigned min_ios = dm_get_reserved_bio_based_ios();
 
 	client = kmalloc(sizeof(*client), GFP_KERNEL);
 	if (!client)
 		return ERR_PTR(-ENOMEM);
 
-	client->pool = mempool_create_kmalloc_pool(ios, sizeof(struct io));
+	client->pool = mempool_create_slab_pool(min_ios, _dm_io_cache);
 	if (!client->pool)
 		goto bad;
 
-	client->bios = bioset_create(16, 16);
+	client->bios = bioset_create(min_ios, 0);
 	if (!client->bios)
 		goto bad;
 
@@ -70,13 +73,6 @@ struct dm_io_client *dm_io_client_create(unsigned num_pages)
 }
 EXPORT_SYMBOL(dm_io_client_create);
 
-int dm_io_client_resize(unsigned num_pages, struct dm_io_client *client)
-{
-	return mempool_resize(client->pool, pages_to_ios(num_pages),
-			      GFP_KERNEL);
-}
-EXPORT_SYMBOL(dm_io_client_resize);
-
 void dm_io_client_destroy(struct dm_io_client *client)
 {
 	mempool_destroy(client->pool);
@@ -87,18 +83,29 @@ EXPORT_SYMBOL(dm_io_client_destroy);
 
 /*-----------------------------------------------------------------
  * We need to keep track of which region a bio is doing io for.
- * In order to save a memory allocation we store this the last
- * bvec which we know is unused (blech).
- * XXX This is ugly and can OOPS with some configs... find another way.
+ * To avoid a memory allocation to store just 5 or 6 bits, we
+ * ensure the 'struct io' pointer is aligned so enough low bits are
+ * always zero and then combine it with the region number directly in
+ * bi_private.
  *---------------------------------------------------------------*/
-static inline void bio_set_region(struct bio *bio, unsigned region)
+static void store_io_and_region_in_bio(struct bio *bio, struct io *io,
+				       unsigned region)
 {
-	bio->bi_io_vec[bio->bi_max_vecs].bv_len = region;
+	if (unlikely(!IS_ALIGNED((unsigned long)io, DM_IO_MAX_REGIONS))) {
+		DMCRIT("Unaligned struct io pointer %p", io);
+		BUG();
+	}
+
+	bio->bi_private = (void *)((unsigned long)io | region);
 }
 
-static inline unsigned bio_get_region(struct bio *bio)
+static void retrieve_io_and_region_from_bio(struct bio *bio, struct io **io,
+				       unsigned *region)
 {
-	return bio->bi_io_vec[bio->bi_max_vecs].bv_len;
+	unsigned long val = (unsigned long)bio->bi_private;
+
+	*io = (void *)(val & -(unsigned long)DM_IO_MAX_REGIONS);
+	*region = val & (DM_IO_MAX_REGIONS - 1);
 }
 
 /*-----------------------------------------------------------------
@@ -111,8 +118,12 @@ static void dec_count(struct io *io, unsigned int region, int error)
 		set_bit(region, &io->error_bits);
 
 	if (atomic_dec_and_test(&io->count)) {
-		if (io->sleeper)
-			wake_up_process(io->sleeper);
+		if (io->vma_invalidate_size)
+			invalidate_kernel_vmap_range(io->vma_invalidate_address,
+						     io->vma_invalidate_size);
+
+		if (io->wait)
+			complete(io->wait);
 
 		else {
 			unsigned long r = io->error_bits;
@@ -136,10 +147,8 @@ static void endio(struct bio *bio, int error)
 	/*
 	 * The bio destructor in bio_put() may use the io object.
 	 */
-	io = bio->bi_private;
-	region = bio_get_region(bio);
+	retrieve_io_and_region_from_bio(bio, &io, &region);
 
-	bio->bi_max_vecs++;
 	bio_put(bio);
 
 	dec_count(io, region, error);
@@ -156,6 +165,9 @@ struct dpages {
 
 	unsigned context_u;
 	void *context_ptr;
+
+	void *vma_invalidate_address;
+	unsigned long vma_invalidate_size;
 };
 
 /*
@@ -190,26 +202,28 @@ static void list_dp_init(struct dpages *dp, struct page_list *pl, unsigned offse
 /*
  * Functions for getting the pages from a bvec.
  */
-static void bvec_get_page(struct dpages *dp,
-		  struct page **p, unsigned long *len, unsigned *offset)
+static void bio_get_page(struct dpages *dp, struct page **p,
+			 unsigned long *len, unsigned *offset)
 {
-	struct bio_vec *bvec = (struct bio_vec *) dp->context_ptr;
+	struct bio_vec *bvec = dp->context_ptr;
 	*p = bvec->bv_page;
-	*len = bvec->bv_len;
-	*offset = bvec->bv_offset;
+	*len = bvec->bv_len - dp->context_u;
+	*offset = bvec->bv_offset + dp->context_u;
 }
 
-static void bvec_next_page(struct dpages *dp)
+static void bio_next_page(struct dpages *dp)
 {
-	struct bio_vec *bvec = (struct bio_vec *) dp->context_ptr;
+	struct bio_vec *bvec = dp->context_ptr;
 	dp->context_ptr = bvec + 1;
+	dp->context_u = 0;
 }
 
-static void bvec_dp_init(struct dpages *dp, struct bio_vec *bvec)
+static void bio_dp_init(struct dpages *dp, struct bio *bio)
 {
-	dp->get_page = bvec_get_page;
-	dp->next_page = bvec_next_page;
-	dp->context_ptr = bvec;
+	dp->get_page = bio_get_page;
+	dp->next_page = bio_next_page;
+	dp->context_ptr = __bvec_iter_bvec(bio->bi_io_vec, bio->bi_iter);
+	dp->context_u = bio->bi_iter.bi_bvec_done;
 }
 
 /*
@@ -237,13 +251,6 @@ static void vm_dp_init(struct dpages *dp, void *data)
 	dp->context_ptr = data;
 }
 
-static void dm_bio_destructor(struct bio *bio)
-{
-	struct io *io = bio->bi_private;
-
-	bio_free(bio, io->client->bios);
-}
-
 /*
  * Functions for getting the pages from kernel memory.
  */
@@ -281,30 +288,50 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 	unsigned offset;
 	unsigned num_bvecs;
 	sector_t remaining = where->count;
+	struct request_queue *q = bdev_get_queue(where->bdev);
+	unsigned short logical_block_size = queue_logical_block_size(q);
+	sector_t num_sectors;
 
-	while (remaining) {
+	/*
+	 * where->count may be zero if rw holds a flush and we need to
+	 * send a zero-sized flush.
+	 */
+	do {
 		/*
-		 * Allocate a suitably sized-bio: we add an extra
-		 * bvec for bio_get/set_region() and decrement bi_max_vecs
-		 * to hide it from bio_add_page().
+		 * Allocate a suitably sized-bio.
 		 */
-		num_bvecs = dm_sector_div_up(remaining,
-					     (PAGE_SIZE >> SECTOR_SHIFT));
-		num_bvecs = 1 + min_t(int, bio_get_nr_vecs(where->bdev),
-				      num_bvecs);
+		if ((rw & REQ_DISCARD) || (rw & REQ_WRITE_SAME))
+			num_bvecs = 1;
+		else
+			num_bvecs = min_t(int, bio_get_nr_vecs(where->bdev),
+					  dm_sector_div_up(remaining, (PAGE_SIZE >> SECTOR_SHIFT)));
+
 		bio = bio_alloc_bioset(GFP_NOIO, num_bvecs, io->client->bios);
-		bio->bi_sector = where->sector + (where->count - remaining);
+		bio->bi_iter.bi_sector = where->sector + (where->count - remaining);
 		bio->bi_bdev = where->bdev;
 		bio->bi_end_io = endio;
-		bio->bi_private = io;
-		bio->bi_destructor = dm_bio_destructor;
-		bio->bi_max_vecs--;
-		bio_set_region(bio, region);
+		store_io_and_region_in_bio(bio, io, region);
+
+		if (rw & REQ_DISCARD) {
+			num_sectors = min_t(sector_t, q->limits.max_discard_sectors, remaining);
+			bio->bi_iter.bi_size = num_sectors << SECTOR_SHIFT;
+			remaining -= num_sectors;
+		} else if (rw & REQ_WRITE_SAME) {
+			/*
+			 * WRITE SAME only uses a single page.
+			 */
+			dp->get_page(dp, &page, &len, &offset);
+			bio_add_page(bio, page, logical_block_size, offset);
+			num_sectors = min_t(sector_t, q->limits.max_write_same_sectors, remaining);
+			bio->bi_iter.bi_size = num_sectors << SECTOR_SHIFT;
 
-		/*
-		 * Try and add as many pages as possible.
-		 */
-		while (remaining) {
+			offset = 0;
+			remaining -= num_sectors;
+			dp->next_page(dp);
+		} else while (remaining) {
+			/*
+			 * Try and add as many pages as possible.
+			 */
 			dp->get_page(dp, &page, &len, &offset);
 			len = min(len, to_bytes(remaining));
 			if (!bio_add_page(bio, page, len, offset))
@@ -317,7 +344,7 @@ static void do_region(int rw, unsigned region, struct dm_io_region *where,
 
 		atomic_inc(&io->count);
 		submit_bio(rw, bio);
-	}
+	} while (remaining);
 }
 
 static void dispatch_io(int rw, unsigned int num_regions,
@@ -327,8 +354,10 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	int i;
 	struct dpages old_pages = *dp;
 
+	BUG_ON(num_regions > DM_IO_MAX_REGIONS);
+
 	if (sync)
-		rw |= (1 << BIO_RW_SYNC);
+		rw |= REQ_SYNC;
 
 	/*
 	 * For multiple regions we need to be careful to rewind
@@ -336,7 +365,7 @@ static void dispatch_io(int rw, unsigned int num_regions,
 	 */
 	for (i = 0; i < num_regions; i++) {
 		*dp = old_pages;
-		if (where[i].count)
+		if (where[i].count || (rw & REQ_FLUSH))
 			do_region(rw, i, where + i, dp, io);
 	}
 
@@ -351,37 +380,37 @@ static int sync_io(struct dm_io_client *client, unsigned int num_regions,
 		   struct dm_io_region *where, int rw, struct dpages *dp,
 		   unsigned long *error_bits)
 {
-	struct io io;
+	/*
+	 * gcc <= 4.3 can't do the alignment for stack variables, so we must
+	 * align it on our own.
+	 * volatile prevents the optimizer from removing or reusing
+	 * "io_" field from the stack frame (allowed in ANSI C).
+	 */
+	volatile char io_[sizeof(struct io) + __alignof__(struct io) - 1];
+	struct io *io = (struct io *)PTR_ALIGN(&io_, __alignof__(struct io));
+	DECLARE_COMPLETION_ONSTACK(wait);
 
 	if (num_regions > 1 && (rw & RW_MASK) != WRITE) {
 		WARN_ON(1);
 		return -EIO;
 	}
 
-	io.error_bits = 0;
-	atomic_set(&io.count, 1); /* see dispatch_io() */
-	io.sleeper = current;
-	io.client = client;
-
-	dispatch_io(rw, num_regions, where, dp, &io, 1);
+	io->error_bits = 0;
+	atomic_set(&io->count, 1); /* see dispatch_io() */
+	io->wait = &wait;
+	io->client = client;
 
-	while (1) {
-		set_current_state(TASK_UNINTERRUPTIBLE);
+	io->vma_invalidate_address = dp->vma_invalidate_address;
+	io->vma_invalidate_size = dp->vma_invalidate_size;
 
-		if (!atomic_read(&io.count) || signal_pending(current))
-			break;
+	dispatch_io(rw, num_regions, where, dp, io, 1);
 
-		io_schedule();
-	}
-	set_current_state(TASK_RUNNING);
-
-	if (atomic_read(&io.count))
-		return -EINTR;
+	wait_for_completion_io(&wait);
 
 	if (error_bits)
-		*error_bits = io.error_bits;
+		*error_bits = io->error_bits;
 
-	return io.error_bits ? -EIO : 0;
+	return io->error_bits ? -EIO : 0;
 }
 
 static int async_io(struct dm_io_client *client, unsigned int num_regions,
@@ -399,28 +428,41 @@ static int async_io(struct dm_io_client *client, unsigned int num_regions,
 	io = mempool_alloc(client->pool, GFP_NOIO);
 	io->error_bits = 0;
 	atomic_set(&io->count, 1); /* see dispatch_io() */
-	io->sleeper = NULL;
+	io->wait = NULL;
 	io->client = client;
 	io->callback = fn;
 	io->context = context;
 
+	io->vma_invalidate_address = dp->vma_invalidate_address;
+	io->vma_invalidate_size = dp->vma_invalidate_size;
+
 	dispatch_io(rw, num_regions, where, dp, io, 0);
 	return 0;
 }
 
-static int dp_init(struct dm_io_request *io_req, struct dpages *dp)
+static int dp_init(struct dm_io_request *io_req, struct dpages *dp,
+		   unsigned long size)
 {
 	/* Set up dpages based on memory type */
+
+	dp->vma_invalidate_address = NULL;
+	dp->vma_invalidate_size = 0;
+
 	switch (io_req->mem.type) {
 	case DM_IO_PAGE_LIST:
 		list_dp_init(dp, io_req->mem.ptr.pl, io_req->mem.offset);
 		break;
 
-	case DM_IO_BVEC:
-		bvec_dp_init(dp, io_req->mem.ptr.bvec);
+	case DM_IO_BIO:
+		bio_dp_init(dp, io_req->mem.ptr.bio);
 		break;
 
 	case DM_IO_VMA:
+		flush_kernel_vmap_range(io_req->mem.ptr.vma, size);
+		if ((io_req->bi_rw & RW_MASK) == READ) {
+			dp->vma_invalidate_address = io_req->mem.ptr.vma;
+			dp->vma_invalidate_size = size;
+		}
 		vm_dp_init(dp, io_req->mem.ptr.vma);
 		break;
 
@@ -439,8 +481,8 @@ static int dp_init(struct dm_io_request *io_req, struct dpages *dp)
  * New collapsed (a)synchronous interface.
  *
  * If the IO is asynchronous (i.e. it has notify.fn), you must either unplug
- * the queue with blk_unplug() some time later or set the BIO_RW_SYNC bit in
- * io_req->bi_rw. If you fail to do one of these, the IO will be submitted to
+ * the queue with blk_unplug() some time later or set REQ_SYNC in
+io_req->bi_rw. If you fail to do one of these, the IO will be submitted to
  * the disk after q->unplug_delay, which defaults to 3ms in blk-settings.c.
  */
 int dm_io(struct dm_io_request *io_req, unsigned num_regions,
@@ -449,7 +491,7 @@ int dm_io(struct dm_io_request *io_req, unsigned num_regions,
 	int r;
 	struct dpages dp;
 
-	r = dp_init(io_req, &dp);
+	r = dp_init(io_req, &dp, (unsigned long)where->count << SECTOR_SHIFT);
 	if (r)
 		return r;
 
@@ -461,3 +503,18 @@ int dm_io(struct dm_io_request *io_req, unsigned num_regions,
 			&dp, io_req->notify.fn, io_req->notify.context);
 }
 EXPORT_SYMBOL(dm_io);
+
+int __init dm_io_init(void)
+{
+	_dm_io_cache = KMEM_CACHE(io, 0);
+	if (!_dm_io_cache)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void dm_io_exit(void)
+{
+	kmem_cache_destroy(_dm_io_cache);
+	_dm_io_cache = NULL;
+}
diff --git a/drivers/md/dm-ioctl.c b/drivers/md/dm-ioctl.c
index b262c0042de..51521429fb5 100644
--- a/drivers/md/dm-ioctl.c
+++ b/drivers/md/dm-ioctl.c
@@ -36,6 +36,14 @@ struct hash_cell {
 	struct dm_table *new_map;
 };
 
+/*
+ * A dummy definition to make RCU happy.
+ * struct dm_table should never be dereferenced in this file.
+ */
+struct dm_table {
+	int undefined__;
+};
+
 struct vers_iter {
     size_t param_size;
     struct dm_target_versions *vers, *old_vers;
@@ -49,13 +57,18 @@ struct vers_iter {
 static struct list_head _name_buckets[NUM_BUCKETS];
 static struct list_head _uuid_buckets[NUM_BUCKETS];
 
-static void dm_hash_remove_all(int keep_open_devices);
+static void dm_hash_remove_all(bool keep_open_devices, bool mark_deferred, bool only_deferred);
 
 /*
  * Guards access to both hash tables.
  */
 static DECLARE_RWSEM(_hash_lock);
 
+/*
+ * Protects use of mdptr to obtain hash cell name and uuid from mapped device.
+ */
+static DEFINE_MUTEX(dm_hash_cells_mutex);
+
 static void init_buckets(struct list_head *buckets)
 {
 	unsigned int i;
@@ -73,7 +86,7 @@ static int dm_hash_init(void)
 
 static void dm_hash_exit(void)
 {
-	dm_hash_remove_all(0);
+	dm_hash_remove_all(false, false, false);
 }
 
 /*-----------------------------------------------------------------
@@ -123,6 +136,24 @@ static struct hash_cell *__get_uuid_cell(const char *str)
 	return NULL;
 }
 
+static struct hash_cell *__get_dev_cell(uint64_t dev)
+{
+	struct mapped_device *md;
+	struct hash_cell *hc;
+
+	md = dm_get_md(huge_decode_dev(dev));
+	if (!md)
+		return NULL;
+
+	hc = dm_get_mdptr(md);
+	if (!hc) {
+		dm_put(md);
+		return NULL;
+	}
+
+	return hc;
+}
+
 /*-----------------------------------------------------------------
  * Inserting, removing and renaming a device.
  *---------------------------------------------------------------*/
@@ -206,7 +237,9 @@ static int dm_hash_insert(const char *name, const char *uuid, struct mapped_devi
 		list_add(&cell->uuid_list, _uuid_buckets + hash_str(uuid));
 	}
 	dm_get(md);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(md, cell);
+	mutex_unlock(&dm_hash_cells_mutex);
 	up_write(&_hash_lock);
 
 	return 0;
@@ -217,128 +250,210 @@ static int dm_hash_insert(const char *name, const char *uuid, struct mapped_devi
 	return -EBUSY;
 }
 
-static void __hash_remove(struct hash_cell *hc)
+static struct dm_table *__hash_remove(struct hash_cell *hc)
 {
 	struct dm_table *table;
+	int srcu_idx;
 
 	/* remove from the dev hash */
 	list_del(&hc->uuid_list);
 	list_del(&hc->name_list);
+	mutex_lock(&dm_hash_cells_mutex);
 	dm_set_mdptr(hc->md, NULL);
+	mutex_unlock(&dm_hash_cells_mutex);
 
-	table = dm_get_table(hc->md);
-	if (table) {
+	table = dm_get_live_table(hc->md, &srcu_idx);
+	if (table)
 		dm_table_event(table);
-		dm_table_put(table);
-	}
+	dm_put_live_table(hc->md, srcu_idx);
 
+	table = NULL;
 	if (hc->new_map)
-		dm_table_put(hc->new_map);
+		table = hc->new_map;
 	dm_put(hc->md);
 	free_cell(hc);
+
+	return table;
 }
 
-static void dm_hash_remove_all(int keep_open_devices)
+static void dm_hash_remove_all(bool keep_open_devices, bool mark_deferred, bool only_deferred)
 {
-	int i, dev_skipped, dev_removed;
+	int i, dev_skipped;
 	struct hash_cell *hc;
-	struct list_head *tmp, *n;
+	struct mapped_device *md;
+	struct dm_table *t;
+
+retry:
+	dev_skipped = 0;
 
 	down_write(&_hash_lock);
 
-retry:
-	dev_skipped = dev_removed = 0;
 	for (i = 0; i < NUM_BUCKETS; i++) {
-		list_for_each_safe (tmp, n, _name_buckets + i) {
-			hc = list_entry(tmp, struct hash_cell, name_list);
+		list_for_each_entry(hc, _name_buckets + i, name_list) {
+			md = hc->md;
+			dm_get(md);
 
 			if (keep_open_devices &&
-			    dm_lock_for_deletion(hc->md)) {
+			    dm_lock_for_deletion(md, mark_deferred, only_deferred)) {
+				dm_put(md);
 				dev_skipped++;
 				continue;
 			}
-			__hash_remove(hc);
-			dev_removed = 1;
+
+			t = __hash_remove(hc);
+
+			up_write(&_hash_lock);
+
+			if (t) {
+				dm_sync_table(md);
+				dm_table_destroy(t);
+			}
+			dm_put(md);
+			if (likely(keep_open_devices))
+				dm_destroy(md);
+			else
+				dm_destroy_immediate(md);
+
+			/*
+			 * Some mapped devices may be using other mapped
+			 * devices, so repeat until we make no further
+			 * progress.  If a new mapped device is created
+			 * here it will also get removed.
+			 */
+			goto retry;
 		}
 	}
 
+	up_write(&_hash_lock);
+
+	if (dev_skipped)
+		DMWARN("remove_all left %d open device(s)", dev_skipped);
+}
+
+/*
+ * Set the uuid of a hash_cell that isn't already set.
+ */
+static void __set_cell_uuid(struct hash_cell *hc, char *new_uuid)
+{
+	mutex_lock(&dm_hash_cells_mutex);
+	hc->uuid = new_uuid;
+	mutex_unlock(&dm_hash_cells_mutex);
+
+	list_add(&hc->uuid_list, _uuid_buckets + hash_str(new_uuid));
+}
+
+/*
+ * Changes the name of a hash_cell and returns the old name for
+ * the caller to free.
+ */
+static char *__change_cell_name(struct hash_cell *hc, char *new_name)
+{
+	char *old_name;
+
 	/*
-	 * Some mapped devices may be using other mapped devices, so if any
-	 * still exist, repeat until we make no further progress.
+	 * Rename and move the name cell.
 	 */
-	if (dev_skipped) {
-		if (dev_removed)
-			goto retry;
+	list_del(&hc->name_list);
+	old_name = hc->name;
 
-		DMWARN("remove_all left %d open device(s)", dev_skipped);
-	}
+	mutex_lock(&dm_hash_cells_mutex);
+	hc->name = new_name;
+	mutex_unlock(&dm_hash_cells_mutex);
 
-	up_write(&_hash_lock);
+	list_add(&hc->name_list, _name_buckets + hash_str(new_name));
+
+	return old_name;
 }
 
-static int dm_hash_rename(const char *old, const char *new)
+static struct mapped_device *dm_hash_rename(struct dm_ioctl *param,
+					    const char *new)
 {
-	char *new_name, *old_name;
+	char *new_data, *old_name = NULL;
 	struct hash_cell *hc;
 	struct dm_table *table;
+	struct mapped_device *md;
+	unsigned change_uuid = (param->flags & DM_UUID_FLAG) ? 1 : 0;
+	int srcu_idx;
 
 	/*
 	 * duplicate new.
 	 */
-	new_name = kstrdup(new, GFP_KERNEL);
-	if (!new_name)
-		return -ENOMEM;
+	new_data = kstrdup(new, GFP_KERNEL);
+	if (!new_data)
+		return ERR_PTR(-ENOMEM);
 
 	down_write(&_hash_lock);
 
 	/*
 	 * Is new free ?
 	 */
-	hc = __get_name_cell(new);
+	if (change_uuid)
+		hc = __get_uuid_cell(new);
+	else
+		hc = __get_name_cell(new);
+
 	if (hc) {
-		DMWARN("asked to rename to an already existing name %s -> %s",
-		       old, new);
+		DMWARN("Unable to change %s on mapped device %s to one that "
+		       "already exists: %s",
+		       change_uuid ? "uuid" : "name",
+		       param->name, new);
 		dm_put(hc->md);
 		up_write(&_hash_lock);
-		kfree(new_name);
-		return -EBUSY;
+		kfree(new_data);
+		return ERR_PTR(-EBUSY);
 	}
 
 	/*
 	 * Is there such a device as 'old' ?
 	 */
-	hc = __get_name_cell(old);
+	hc = __get_name_cell(param->name);
 	if (!hc) {
-		DMWARN("asked to rename a non existent device %s -> %s",
-		       old, new);
+		DMWARN("Unable to rename non-existent device, %s to %s%s",
+		       param->name, change_uuid ? "uuid " : "", new);
 		up_write(&_hash_lock);
-		kfree(new_name);
-		return -ENXIO;
+		kfree(new_data);
+		return ERR_PTR(-ENXIO);
 	}
 
 	/*
-	 * rename and move the name cell.
+	 * Does this device already have a uuid?
 	 */
-	list_del(&hc->name_list);
-	old_name = hc->name;
-	hc->name = new_name;
-	list_add(&hc->name_list, _name_buckets + hash_str(new_name));
+	if (change_uuid && hc->uuid) {
+		DMWARN("Unable to change uuid of mapped device %s to %s "
+		       "because uuid is already set to %s",
+		       param->name, new, hc->uuid);
+		dm_put(hc->md);
+		up_write(&_hash_lock);
+		kfree(new_data);
+		return ERR_PTR(-EINVAL);
+	}
+
+	if (change_uuid)
+		__set_cell_uuid(hc, new_data);
+	else
+		old_name = __change_cell_name(hc, new_data);
 
 	/*
 	 * Wake up any dm event waiters.
 	 */
-	table = dm_get_table(hc->md);
-	if (table) {
+	table = dm_get_live_table(hc->md, &srcu_idx);
+	if (table)
 		dm_table_event(table);
-		dm_table_put(table);
-	}
+	dm_put_live_table(hc->md, srcu_idx);
 
-	dm_kobject_uevent(hc->md);
+	if (!dm_kobject_uevent(hc->md, KOBJ_CHANGE, param->event_nr))
+		param->flags |= DM_UEVENT_GENERATED_FLAG;
 
-	dm_put(hc->md);
+	md = hc->md;
 	up_write(&_hash_lock);
 	kfree(old_name);
-	return 0;
+
+	return md;
+}
+
+void dm_deferred_remove(void)
+{
+	dm_hash_remove_all(true, false, true);
 }
 
 /*-----------------------------------------------------------------
@@ -352,7 +467,7 @@ typedef int (*ioctl_fn)(struct dm_ioctl *param, size_t param_size);
 
 static int remove_all(struct dm_ioctl *param, size_t param_size)
 {
-	dm_hash_remove_all(1);
+	dm_hash_remove_all(true, !!(param->flags & DM_DEFERRED_REMOVE), false);
 	param->data_size = 0;
 	return 0;
 }
@@ -426,7 +541,7 @@ static int list_devices(struct dm_ioctl *param, size_t param_size)
 				old_nl->next = (uint32_t) ((void *) nl -
 							   (void *) old_nl);
 			disk = dm_disk(hc->md);
-			nl->dev = huge_encode_dev(MKDEV(disk->major, disk->first_minor));
+			nl->dev = huge_encode_dev(disk_devt(disk));
 			nl->next = 0;
 			strcpy(nl->name, hc->name);
 
@@ -512,8 +627,6 @@ static int list_versions(struct dm_ioctl *param, size_t param_size)
 	return 0;
 }
 
-
-
 static int check_name(const char *name)
 {
 	if (strchr(name, '/')) {
@@ -525,21 +638,61 @@ static int check_name(const char *name)
 }
 
 /*
+ * On successful return, the caller must not attempt to acquire
+ * _hash_lock without first calling dm_table_put, because dm_table_destroy
+ * waits for this dm_table_put and could be called under this lock.
+ */
+static struct dm_table *dm_get_inactive_table(struct mapped_device *md, int *srcu_idx)
+{
+	struct hash_cell *hc;
+	struct dm_table *table = NULL;
+
+	/* increment rcu count, we don't care about the table pointer */
+	dm_get_live_table(md, srcu_idx);
+
+	down_read(&_hash_lock);
+	hc = dm_get_mdptr(md);
+	if (!hc || hc->md != md) {
+		DMWARN("device has been removed from the dev hash table.");
+		goto out;
+	}
+
+	table = hc->new_map;
+
+out:
+	up_read(&_hash_lock);
+
+	return table;
+}
+
+static struct dm_table *dm_get_live_or_inactive_table(struct mapped_device *md,
+						      struct dm_ioctl *param,
+						      int *srcu_idx)
+{
+	return (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) ?
+		dm_get_inactive_table(md, srcu_idx) : dm_get_live_table(md, srcu_idx);
+}
+
+/*
  * Fills in a dm_ioctl structure, ready for sending back to
  * userland.
  */
-static int __dev_status(struct mapped_device *md, struct dm_ioctl *param)
+static void __dev_status(struct mapped_device *md, struct dm_ioctl *param)
 {
 	struct gendisk *disk = dm_disk(md);
 	struct dm_table *table;
+	int srcu_idx;
 
 	param->flags &= ~(DM_SUSPEND_FLAG | DM_READONLY_FLAG |
 			  DM_ACTIVE_PRESENT_FLAG);
 
-	if (dm_suspended(md))
+	if (dm_suspended_md(md))
 		param->flags |= DM_SUSPEND_FLAG;
 
-	param->dev = huge_encode_dev(MKDEV(disk->major, disk->first_minor));
+	if (dm_test_deferred_remove_flag(md))
+		param->flags |= DM_DEFERRED_REMOVE;
+
+	param->dev = huge_encode_dev(disk_devt(disk));
 
 	/*
 	 * Yes, this will be out of date by the time it gets back
@@ -548,20 +701,31 @@ static int __dev_status(struct mapped_device *md, struct dm_ioctl *param)
 	 */
 	param->open_count = dm_open_count(md);
 
-	if (disk->policy)
-		param->flags |= DM_READONLY_FLAG;
-
 	param->event_nr = dm_get_event_nr(md);
+	param->target_count = 0;
 
-	table = dm_get_table(md);
+	table = dm_get_live_table(md, &srcu_idx);
 	if (table) {
-		param->flags |= DM_ACTIVE_PRESENT_FLAG;
-		param->target_count = dm_table_get_num_targets(table);
-		dm_table_put(table);
-	} else
-		param->target_count = 0;
+		if (!(param->flags & DM_QUERY_INACTIVE_TABLE_FLAG)) {
+			if (get_disk_ro(disk))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+		}
 
-	return 0;
+		param->flags |= DM_ACTIVE_PRESENT_FLAG;
+	}
+	dm_put_live_table(md, srcu_idx);
+
+	if (param->flags & DM_QUERY_INACTIVE_TABLE_FLAG) {
+		int srcu_idx;
+		table = dm_get_inactive_table(md, &srcu_idx);
+		if (table) {
+			if (!(dm_table_get_mode(table) & FMODE_WRITE))
+				param->flags |= DM_READONLY_FLAG;
+			param->target_count = dm_table_get_num_targets(table);
+		}
+		dm_put_live_table(md, srcu_idx);
+	}
 }
 
 static int dev_create(struct dm_ioctl *param, size_t param_size)
@@ -583,15 +747,17 @@ static int dev_create(struct dm_ioctl *param, size_t param_size)
 	r = dm_hash_insert(param->name, *param->uuid ? param->uuid : NULL, md);
 	if (r) {
 		dm_put(md);
+		dm_destroy(md);
 		return r;
 	}
 
 	param->flags &= ~DM_INACTIVE_PRESENT_FLAG;
 
-	r = __dev_status(md, param);
+	__dev_status(md, param);
+
 	dm_put(md);
 
-	return r;
+	return 0;
 }
 
 /*
@@ -599,25 +765,45 @@ static int dev_create(struct dm_ioctl *param, size_t param_size)
  */
 static struct hash_cell *__find_device_hash_cell(struct dm_ioctl *param)
 {
-	struct mapped_device *md;
-	void *mdptr = NULL;
+	struct hash_cell *hc = NULL;
 
-	if (*param->uuid)
-		return __get_uuid_cell(param->uuid);
+	if (*param->uuid) {
+		if (*param->name || param->dev)
+			return NULL;
 
-	if (*param->name)
-		return __get_name_cell(param->name);
+		hc = __get_uuid_cell(param->uuid);
+		if (!hc)
+			return NULL;
+	} else if (*param->name) {
+		if (param->dev)
+			return NULL;
 
-	md = dm_get_md(huge_decode_dev(param->dev));
-	if (!md)
-		goto out;
+		hc = __get_name_cell(param->name);
+		if (!hc)
+			return NULL;
+	} else if (param->dev) {
+		hc = __get_dev_cell(param->dev);
+		if (!hc)
+			return NULL;
+	} else
+		return NULL;
 
-	mdptr = dm_get_mdptr(md);
-	if (!mdptr)
-		dm_put(md);
+	/*
+	 * Sneakily write in both the name and the uuid
+	 * while we have the cell.
+	 */
+	strlcpy(param->name, hc->name, sizeof(param->name));
+	if (hc->uuid)
+		strlcpy(param->uuid, hc->uuid, sizeof(param->uuid));
+	else
+		param->uuid[0] = '\0';
 
-out:
-	return mdptr;
+	if (hc->new_map)
+		param->flags |= DM_INACTIVE_PRESENT_FLAG;
+	else
+		param->flags &= ~DM_INACTIVE_PRESENT_FLAG;
+
+	return hc;
 }
 
 static struct mapped_device *find_device(struct dm_ioctl *param)
@@ -627,24 +813,8 @@ static struct mapped_device *find_device(struct dm_ioctl *param)
 
 	down_read(&_hash_lock);
 	hc = __find_device_hash_cell(param);
-	if (hc) {
+	if (hc)
 		md = hc->md;
-
-		/*
-		 * Sneakily write in both the name and the uuid
-		 * while we have the cell.
-		 */
-		strncpy(param->name, hc->name, sizeof(param->name));
-		if (hc->uuid)
-			strncpy(param->uuid, hc->uuid, sizeof(param->uuid)-1);
-		else
-			param->uuid[0] = '\0';
-
-		if (hc->new_map)
-			param->flags |= DM_INACTIVE_PRESENT_FLAG;
-		else
-			param->flags &= ~DM_INACTIVE_PRESENT_FLAG;
-	}
 	up_read(&_hash_lock);
 
 	return md;
@@ -655,12 +825,13 @@ static int dev_remove(struct dm_ioctl *param, size_t param_size)
 	struct hash_cell *hc;
 	struct mapped_device *md;
 	int r;
+	struct dm_table *t;
 
 	down_write(&_hash_lock);
 	hc = __find_device_hash_cell(param);
 
 	if (!hc) {
-		DMWARN("device doesn't appear to be in the dev hash table.");
+		DMDEBUG_LIMIT("device doesn't appear to be in the dev hash table.");
 		up_write(&_hash_lock);
 		return -ENXIO;
 	}
@@ -670,18 +841,34 @@ static int dev_remove(struct dm_ioctl *param, size_t param_size)
 	/*
 	 * Ensure the device is not open and nothing further can open it.
 	 */
-	r = dm_lock_for_deletion(md);
+	r = dm_lock_for_deletion(md, !!(param->flags & DM_DEFERRED_REMOVE), false);
 	if (r) {
-		DMWARN("unable to remove open device %s", hc->name);
+		if (r == -EBUSY && param->flags & DM_DEFERRED_REMOVE) {
+			up_write(&_hash_lock);
+			dm_put(md);
+			return 0;
+		}
+		DMDEBUG_LIMIT("unable to remove open device %s", hc->name);
 		up_write(&_hash_lock);
 		dm_put(md);
 		return r;
 	}
 
-	__hash_remove(hc);
+	t = __hash_remove(hc);
 	up_write(&_hash_lock);
+
+	if (t) {
+		dm_sync_table(md);
+		dm_table_destroy(t);
+	}
+
+	param->flags &= ~DM_DEFERRED_REMOVE;
+
+	if (!dm_kobject_uevent(md, KOBJ_REMOVE, param->event_nr))
+		param->flags |= DM_UEVENT_GENERATED_FLAG;
+
 	dm_put(md);
-	param->data_size = 0;
+	dm_destroy(md);
 	return 0;
 }
 
@@ -701,20 +888,31 @@ static int invalid_str(char *str, void *end)
 static int dev_rename(struct dm_ioctl *param, size_t param_size)
 {
 	int r;
-	char *new_name = (char *) param + param->data_start;
+	char *new_data = (char *) param + param->data_start;
+	struct mapped_device *md;
+	unsigned change_uuid = (param->flags & DM_UUID_FLAG) ? 1 : 0;
 
-	if (new_name < param->data ||
-	    invalid_str(new_name, (void *) param + param_size)) {
-		DMWARN("Invalid new logical volume name supplied.");
+	if (new_data < param->data ||
+	    invalid_str(new_data, (void *) param + param_size) || !*new_data ||
+	    strlen(new_data) > (change_uuid ? DM_UUID_LEN - 1 : DM_NAME_LEN - 1)) {
+		DMWARN("Invalid new mapped device name or uuid string supplied.");
 		return -EINVAL;
 	}
 
-	r = check_name(new_name);
-	if (r)
-		return r;
+	if (!change_uuid) {
+		r = check_name(new_data);
+		if (r)
+			return r;
+	}
 
-	param->data_size = 0;
-	return dm_hash_rename(param->name, new_name);
+	md = dm_hash_rename(param, new_data);
+	if (IS_ERR(md))
+		return PTR_ERR(md);
+
+	__dev_status(md, param);
+	dm_put(md);
+
+	return 0;
 }
 
 static int dev_set_geometry(struct dm_ioctl *param, size_t param_size)
@@ -724,6 +922,7 @@ static int dev_set_geometry(struct dm_ioctl *param, size_t param_size)
 	struct hd_geometry geometry;
 	unsigned long indata[4];
 	char *geostr = (char *) param + param->data_start;
+	char dummy;
 
 	md = find_device(param);
 	if (!md)
@@ -735,8 +934,8 @@ static int dev_set_geometry(struct dm_ioctl *param, size_t param_size)
 		goto out;
 	}
 
-	x = sscanf(geostr, "%lu %lu %lu %lu", indata,
-		   indata + 1, indata + 2, indata + 3);
+	x = sscanf(geostr, "%lu %lu %lu %lu%c", indata,
+		   indata + 1, indata + 2, indata + 3, &dummy);
 
 	if (x != 4) {
 		DMWARN("Unable to interpret geometry settings.");
@@ -755,8 +954,6 @@ static int dev_set_geometry(struct dm_ioctl *param, size_t param_size)
 	geometry.start = indata[3];
 
 	r = dm_set_geometry(md, &geometry);
-	if (!r)
-		r = __dev_status(md, param);
 
 	param->data_size = 0;
 
@@ -780,13 +977,17 @@ static int do_suspend(struct dm_ioctl *param)
 	if (param->flags & DM_NOFLUSH_FLAG)
 		suspend_flags |= DM_SUSPEND_NOFLUSH_FLAG;
 
-	if (!dm_suspended(md))
+	if (!dm_suspended_md(md)) {
 		r = dm_suspend(md, suspend_flags);
+		if (r)
+			goto out;
+	}
 
-	if (!r)
-		r = __dev_status(md, param);
+	__dev_status(md, param);
 
+out:
 	dm_put(md);
+
 	return r;
 }
 
@@ -796,13 +997,13 @@ static int do_resume(struct dm_ioctl *param)
 	unsigned suspend_flags = DM_SUSPEND_LOCKFS_FLAG;
 	struct hash_cell *hc;
 	struct mapped_device *md;
-	struct dm_table *new_map;
+	struct dm_table *new_map, *old_map = NULL;
 
 	down_write(&_hash_lock);
 
 	hc = __find_device_hash_cell(param);
 	if (!hc) {
-		DMWARN("device doesn't appear to be in the dev hash table.");
+		DMDEBUG_LIMIT("device doesn't appear to be in the dev hash table.");
 		up_write(&_hash_lock);
 		return -ENXIO;
 	}
@@ -822,29 +1023,38 @@ static int do_resume(struct dm_ioctl *param)
 			suspend_flags &= ~DM_SUSPEND_LOCKFS_FLAG;
 		if (param->flags & DM_NOFLUSH_FLAG)
 			suspend_flags |= DM_SUSPEND_NOFLUSH_FLAG;
-		if (!dm_suspended(md))
+		if (!dm_suspended_md(md))
 			dm_suspend(md, suspend_flags);
 
-		r = dm_swap_table(md, new_map);
-		if (r) {
+		old_map = dm_swap_table(md, new_map);
+		if (IS_ERR(old_map)) {
+			dm_sync_table(md);
+			dm_table_destroy(new_map);
 			dm_put(md);
-			dm_table_put(new_map);
-			return r;
+			return PTR_ERR(old_map);
 		}
 
 		if (dm_table_get_mode(new_map) & FMODE_WRITE)
 			set_disk_ro(dm_disk(md), 0);
 		else
 			set_disk_ro(dm_disk(md), 1);
-
-		dm_table_put(new_map);
 	}
 
-	if (dm_suspended(md))
+	if (dm_suspended_md(md)) {
 		r = dm_resume(md);
+		if (!r && !dm_kobject_uevent(md, KOBJ_CHANGE, param->event_nr))
+			param->flags |= DM_UEVENT_GENERATED_FLAG;
+	}
+
+	/*
+	 * Since dm_swap_table synchronizes RCU, nobody should be in
+	 * read-side critical section already.
+	 */
+	if (old_map)
+		dm_table_destroy(old_map);
 
 	if (!r)
-		r = __dev_status(md, param);
+		__dev_status(md, param);
 
 	dm_put(md);
 	return r;
@@ -868,16 +1078,16 @@ static int dev_suspend(struct dm_ioctl *param, size_t param_size)
  */
 static int dev_status(struct dm_ioctl *param, size_t param_size)
 {
-	int r;
 	struct mapped_device *md;
 
 	md = find_device(param);
 	if (!md)
 		return -ENXIO;
 
-	r = __dev_status(md, param);
+	__dev_status(md, param);
 	dm_put(md);
-	return r;
+
+	return 0;
 }
 
 /*
@@ -891,6 +1101,7 @@ 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);
 
@@ -903,6 +1114,7 @@ static void retrieve_status(struct dm_table *table,
 	num_targets = dm_table_get_num_targets(table);
 	for (i = 0; i < num_targets; i++) {
 		struct dm_target *ti = dm_table_get_target(table, i);
+		size_t l;
 
 		remaining = len - (outptr - outbuf);
 		if (remaining <= sizeof(struct dm_target_spec)) {
@@ -927,14 +1139,19 @@ 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)) {
-				param->flags |= DM_BUFFER_FULL_FLAG;
-				break;
-			}
+			if (param->flags & DM_NOFLUSH_FLAG)
+				status_flags |= DM_STATUS_NOFLUSH_FLAG;
+			ti->type->status(ti, type, status_flags, outptr, remaining);
 		} else
 			outptr[0] = '\0';
 
-		outptr += strlen(outptr) + 1;
+		l = strlen(outptr) + 1;
+		if (l == remaining) {
+			param->flags |= DM_BUFFER_FULL_FLAG;
+			break;
+		}
+
+		outptr += l;
 		used = param->data_start + (outptr - outbuf);
 
 		outptr = align_ptr(outptr);
@@ -952,9 +1169,10 @@ static void retrieve_status(struct dm_table *table,
  */
 static int dev_wait(struct dm_ioctl *param, size_t param_size)
 {
-	int r;
+	int r = 0;
 	struct mapped_device *md;
 	struct dm_table *table;
+	int srcu_idx;
 
 	md = find_device(param);
 	if (!md)
@@ -973,24 +1191,22 @@ static int dev_wait(struct dm_ioctl *param, size_t param_size)
 	 * changed to trigger the event, so we may as well tell
 	 * him and save an ioctl.
 	 */
-	r = __dev_status(md, param);
-	if (r)
-		goto out;
+	__dev_status(md, param);
 
-	table = dm_get_table(md);
-	if (table) {
+	table = dm_get_live_or_inactive_table(md, param, &srcu_idx);
+	if (table)
 		retrieve_status(table, param, param_size);
-		dm_table_put(table);
-	}
+	dm_put_live_table(md, srcu_idx);
 
- out:
+out:
 	dm_put(md);
+
 	return r;
 }
 
-static inline int get_mode(struct dm_ioctl *param)
+static inline fmode_t get_mode(struct dm_ioctl *param)
 {
-	int mode = FMODE_READ | FMODE_WRITE;
+	fmode_t mode = FMODE_READ | FMODE_WRITE;
 
 	if (param->flags & DM_READONLY_FLAG)
 		mode = FMODE_READ;
@@ -1052,8 +1268,9 @@ static int table_load(struct dm_ioctl *param, size_t param_size)
 {
 	int r;
 	struct hash_cell *hc;
-	struct dm_table *t;
+	struct dm_table *t, *old_map = NULL;
 	struct mapped_device *md;
+	struct target_type *immutable_target_type;
 
 	md = find_device(param);
 	if (!md)
@@ -1061,33 +1278,72 @@ static int table_load(struct dm_ioctl *param, size_t param_size)
 
 	r = dm_table_create(&t, get_mode(param), param->target_count, md);
 	if (r)
-		goto out;
+		goto err;
 
+	/* Protect md->type and md->queue against concurrent table loads. */
+	dm_lock_md_type(md);
 	r = populate_table(t, param, param_size);
+	if (r)
+		goto err_unlock_md_type;
+
+	immutable_target_type = dm_get_immutable_target_type(md);
+	if (immutable_target_type &&
+	    (immutable_target_type != dm_table_get_immutable_target_type(t))) {
+		DMWARN("can't replace immutable target type %s",
+		       immutable_target_type->name);
+		r = -EINVAL;
+		goto err_unlock_md_type;
+	}
+
+	if (dm_get_md_type(md) == DM_TYPE_NONE)
+		/* Initial table load: acquire type of table. */
+		dm_set_md_type(md, dm_table_get_type(t));
+	else if (dm_get_md_type(md) != dm_table_get_type(t)) {
+		DMWARN("can't change device type after initial table load.");
+		r = -EINVAL;
+		goto err_unlock_md_type;
+	}
+
+	/* setup md->queue to reflect md's type (may block) */
+	r = dm_setup_md_queue(md);
 	if (r) {
-		dm_table_put(t);
-		goto out;
+		DMWARN("unable to set up device queue for new table.");
+		goto err_unlock_md_type;
 	}
+	dm_unlock_md_type(md);
 
+	/* stage inactive table */
 	down_write(&_hash_lock);
 	hc = dm_get_mdptr(md);
 	if (!hc || hc->md != md) {
 		DMWARN("device has been removed from the dev hash table.");
-		dm_table_put(t);
 		up_write(&_hash_lock);
 		r = -ENXIO;
-		goto out;
+		goto err_destroy_table;
 	}
 
 	if (hc->new_map)
-		dm_table_put(hc->new_map);
+		old_map = hc->new_map;
 	hc->new_map = t;
 	up_write(&_hash_lock);
 
 	param->flags |= DM_INACTIVE_PRESENT_FLAG;
-	r = __dev_status(md, param);
+	__dev_status(md, param);
+
+	if (old_map) {
+		dm_sync_table(md);
+		dm_table_destroy(old_map);
+	}
 
-out:
+	dm_put(md);
+
+	return 0;
+
+err_unlock_md_type:
+	dm_unlock_md_type(md);
+err_destroy_table:
+	dm_table_destroy(t);
+err:
 	dm_put(md);
 
 	return r;
@@ -1095,31 +1351,36 @@ out:
 
 static int table_clear(struct dm_ioctl *param, size_t param_size)
 {
-	int r;
 	struct hash_cell *hc;
 	struct mapped_device *md;
+	struct dm_table *old_map = NULL;
 
 	down_write(&_hash_lock);
 
 	hc = __find_device_hash_cell(param);
 	if (!hc) {
-		DMWARN("device doesn't appear to be in the dev hash table.");
+		DMDEBUG_LIMIT("device doesn't appear to be in the dev hash table.");
 		up_write(&_hash_lock);
 		return -ENXIO;
 	}
 
 	if (hc->new_map) {
-		dm_table_put(hc->new_map);
+		old_map = hc->new_map;
 		hc->new_map = NULL;
 	}
 
 	param->flags &= ~DM_INACTIVE_PRESENT_FLAG;
 
-	r = __dev_status(hc->md, param);
+	__dev_status(hc->md, param);
 	md = hc->md;
 	up_write(&_hash_lock);
+	if (old_map) {
+		dm_sync_table(md);
+		dm_table_destroy(old_map);
+	}
 	dm_put(md);
-	return r;
+
+	return 0;
 }
 
 /*
@@ -1131,7 +1392,7 @@ static void retrieve_deps(struct dm_table *table,
 	unsigned int count = 0;
 	struct list_head *tmp;
 	size_t len, needed;
-	struct dm_dev *dd;
+	struct dm_dev_internal *dd;
 	struct dm_target_deps *deps;
 
 	deps = get_result_buffer(param, param_size, &len);
@@ -1157,34 +1418,31 @@ static void retrieve_deps(struct dm_table *table,
 	deps->count = count;
 	count = 0;
 	list_for_each_entry (dd, dm_table_get_devices(table), list)
-		deps->dev[count++] = huge_encode_dev(dd->bdev->bd_dev);
+		deps->dev[count++] = huge_encode_dev(dd->dm_dev.bdev->bd_dev);
 
 	param->data_size = param->data_start + needed;
 }
 
 static int table_deps(struct dm_ioctl *param, size_t param_size)
 {
-	int r = 0;
 	struct mapped_device *md;
 	struct dm_table *table;
+	int srcu_idx;
 
 	md = find_device(param);
 	if (!md)
 		return -ENXIO;
 
-	r = __dev_status(md, param);
-	if (r)
-		goto out;
+	__dev_status(md, param);
 
-	table = dm_get_table(md);
-	if (table) {
+	table = dm_get_live_or_inactive_table(md, param, &srcu_idx);
+	if (table)
 		retrieve_deps(table, param, param_size);
-		dm_table_put(table);
-	}
+	dm_put_live_table(md, srcu_idx);
 
- out:
 	dm_put(md);
-	return r;
+
+	return 0;
 }
 
 /*
@@ -1193,27 +1451,54 @@ static int table_deps(struct dm_ioctl *param, size_t param_size)
  */
 static int table_status(struct dm_ioctl *param, size_t param_size)
 {
-	int r;
 	struct mapped_device *md;
 	struct dm_table *table;
+	int srcu_idx;
 
 	md = find_device(param);
 	if (!md)
 		return -ENXIO;
 
-	r = __dev_status(md, param);
-	if (r)
-		goto out;
+	__dev_status(md, param);
 
-	table = dm_get_table(md);
-	if (table) {
+	table = dm_get_live_or_inactive_table(md, param, &srcu_idx);
+	if (table)
 		retrieve_status(table, param, param_size);
-		dm_table_put(table);
-	}
+	dm_put_live_table(md, srcu_idx);
 
- out:
 	dm_put(md);
-	return r;
+
+	return 0;
+}
+
+/*
+ * Process device-mapper dependent messages.  Messages prefixed with '@'
+ * are processed by the DM core.  All others are delivered to the target.
+ * Returns a number <= 1 if message was processed by device mapper.
+ * Returns 2 if message should be delivered to the target.
+ */
+static int message_for_md(struct mapped_device *md, unsigned argc, char **argv,
+			  char *result, unsigned maxlen)
+{
+	int r;
+
+	if (**argv != '@')
+		return 2; /* no '@' prefix, deliver to target */
+
+	if (!strcasecmp(argv[0], "@cancel_deferred_remove")) {
+		if (argc != 1) {
+			DMERR("Invalid arguments for @cancel_deferred_remove");
+			return -EINVAL;
+		}
+		return dm_cancel_deferred_remove(md);
+	}
+
+	r = dm_stats_message(md, argc, argv, result, maxlen);
+	if (r < 2)
+		return r;
+
+	DMERR("Unsupported message sent to DM core: %s", argv[0]);
+	return -EINVAL;
 }
 
 /*
@@ -1227,15 +1512,14 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 	struct dm_table *table;
 	struct dm_target *ti;
 	struct dm_target_msg *tmsg = (void *) param + param->data_start;
+	size_t maxlen;
+	char *result = get_result_buffer(param, param_size, &maxlen);
+	int srcu_idx;
 
 	md = find_device(param);
 	if (!md)
 		return -ENXIO;
 
-	r = __dev_status(md, param);
-	if (r)
-		goto out;
-
 	if (tmsg < (struct dm_target_msg *) param->data ||
 	    invalid_str(tmsg->message, (void *) param + param_size)) {
 		DMWARN("Invalid target message parameters.");
@@ -1249,10 +1533,24 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		goto out;
 	}
 
-	table = dm_get_table(md);
-	if (!table)
+	if (!argc) {
+		DMWARN("Empty message received.");
+		goto out_argv;
+	}
+
+	r = message_for_md(md, argc, argv, result, maxlen);
+	if (r <= 1)
 		goto out_argv;
 
+	table = dm_get_live_table(md, &srcu_idx);
+	if (!table)
+		goto out_table;
+
+	if (dm_deleting_md(md)) {
+		r = -ENXIO;
+		goto out_table;
+	}
+
 	ti = dm_table_find_target(table, tmsg->sector);
 	if (!dm_target_is_valid(ti)) {
 		DMWARN("Target message sector outside device.");
@@ -1264,48 +1562,73 @@ static int target_message(struct dm_ioctl *param, size_t param_size)
 		r = -EINVAL;
 	}
 
-	dm_table_put(table);
+ out_table:
+	dm_put_live_table(md, srcu_idx);
  out_argv:
 	kfree(argv);
  out:
-	param->data_size = 0;
+	if (r >= 0)
+		__dev_status(md, param);
+
+	if (r == 1) {
+		param->flags |= DM_DATA_OUT_FLAG;
+		if (dm_message_test_buffer_overflow(result, maxlen))
+			param->flags |= DM_BUFFER_FULL_FLAG;
+		else
+			param->data_size = param->data_start + strlen(result) + 1;
+		r = 0;
+	}
+
 	dm_put(md);
 	return r;
 }
 
+/*
+ * The ioctl parameter block consists of two parts, a dm_ioctl struct
+ * followed by a data buffer.  This flag is set if the second part,
+ * which has a variable size, is not used by the function processing
+ * the ioctl.
+ */
+#define IOCTL_FLAGS_NO_PARAMS	1
+
 /*-----------------------------------------------------------------
  * Implementation of open/close/ioctl on the special char
  * device.
  *---------------------------------------------------------------*/
-static ioctl_fn lookup_ioctl(unsigned int cmd)
+static ioctl_fn lookup_ioctl(unsigned int cmd, int *ioctl_flags)
 {
 	static struct {
 		int cmd;
+		int flags;
 		ioctl_fn fn;
 	} _ioctls[] = {
-		{DM_VERSION_CMD, NULL},	/* version is dealt with elsewhere */
-		{DM_REMOVE_ALL_CMD, remove_all},
-		{DM_LIST_DEVICES_CMD, list_devices},
-
-		{DM_DEV_CREATE_CMD, dev_create},
-		{DM_DEV_REMOVE_CMD, dev_remove},
-		{DM_DEV_RENAME_CMD, dev_rename},
-		{DM_DEV_SUSPEND_CMD, dev_suspend},
-		{DM_DEV_STATUS_CMD, dev_status},
-		{DM_DEV_WAIT_CMD, dev_wait},
-
-		{DM_TABLE_LOAD_CMD, table_load},
-		{DM_TABLE_CLEAR_CMD, table_clear},
-		{DM_TABLE_DEPS_CMD, table_deps},
-		{DM_TABLE_STATUS_CMD, table_status},
-
-		{DM_LIST_VERSIONS_CMD, list_versions},
-
-		{DM_TARGET_MSG_CMD, target_message},
-		{DM_DEV_SET_GEOMETRY_CMD, dev_set_geometry}
+		{DM_VERSION_CMD, 0, NULL}, /* version is dealt with elsewhere */
+		{DM_REMOVE_ALL_CMD, IOCTL_FLAGS_NO_PARAMS, remove_all},
+		{DM_LIST_DEVICES_CMD, 0, list_devices},
+
+		{DM_DEV_CREATE_CMD, IOCTL_FLAGS_NO_PARAMS, dev_create},
+		{DM_DEV_REMOVE_CMD, IOCTL_FLAGS_NO_PARAMS, dev_remove},
+		{DM_DEV_RENAME_CMD, 0, dev_rename},
+		{DM_DEV_SUSPEND_CMD, IOCTL_FLAGS_NO_PARAMS, dev_suspend},
+		{DM_DEV_STATUS_CMD, IOCTL_FLAGS_NO_PARAMS, dev_status},
+		{DM_DEV_WAIT_CMD, 0, dev_wait},
+
+		{DM_TABLE_LOAD_CMD, 0, table_load},
+		{DM_TABLE_CLEAR_CMD, IOCTL_FLAGS_NO_PARAMS, table_clear},
+		{DM_TABLE_DEPS_CMD, 0, table_deps},
+		{DM_TABLE_STATUS_CMD, 0, table_status},
+
+		{DM_LIST_VERSIONS_CMD, 0, list_versions},
+
+		{DM_TARGET_MSG_CMD, 0, target_message},
+		{DM_DEV_SET_GEOMETRY_CMD, 0, dev_set_geometry}
 	};
 
-	return (cmd >= ARRAY_SIZE(_ioctls)) ? NULL : _ioctls[cmd].fn;
+	if (unlikely(cmd >= ARRAY_SIZE(_ioctls)))
+		return NULL;
+
+	*ioctl_flags = _ioctls[cmd].flags;
+	return _ioctls[cmd].fn;
 }
 
 /*
@@ -1342,38 +1665,103 @@ static int check_version(unsigned int cmd, struct dm_ioctl __user *user)
 	return r;
 }
 
-static void free_params(struct dm_ioctl *param)
+#define DM_PARAMS_KMALLOC	0x0001	/* Params alloced with kmalloc */
+#define DM_PARAMS_VMALLOC	0x0002	/* Params alloced with vmalloc */
+#define DM_WIPE_BUFFER		0x0010	/* Wipe input buffer before returning from ioctl */
+
+static void free_params(struct dm_ioctl *param, size_t param_size, int param_flags)
 {
-	vfree(param);
+	if (param_flags & DM_WIPE_BUFFER)
+		memset(param, 0, param_size);
+
+	if (param_flags & DM_PARAMS_KMALLOC)
+		kfree(param);
+	if (param_flags & DM_PARAMS_VMALLOC)
+		vfree(param);
 }
 
-static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl **param)
+static int copy_params(struct dm_ioctl __user *user, struct dm_ioctl *param_kernel,
+		       int ioctl_flags,
+		       struct dm_ioctl **param, int *param_flags)
 {
-	struct dm_ioctl tmp, *dmi;
+	struct dm_ioctl *dmi;
+	int secure_data;
+	const size_t minimum_data_size = sizeof(*param_kernel) - sizeof(param_kernel->data);
 
-	if (copy_from_user(&tmp, user, sizeof(tmp) - sizeof(tmp.data)))
+	if (copy_from_user(param_kernel, user, minimum_data_size))
 		return -EFAULT;
 
-	if (tmp.data_size < (sizeof(tmp) - sizeof(tmp.data)))
+	if (param_kernel->data_size < minimum_data_size)
 		return -EINVAL;
 
-	dmi = vmalloc(tmp.data_size);
-	if (!dmi)
+	secure_data = param_kernel->flags & DM_SECURE_DATA_FLAG;
+
+	*param_flags = secure_data ? DM_WIPE_BUFFER : 0;
+
+	if (ioctl_flags & IOCTL_FLAGS_NO_PARAMS) {
+		dmi = param_kernel;
+		dmi->data_size = minimum_data_size;
+		goto data_copied;
+	}
+
+	/*
+	 * Try to avoid low memory issues when a device is suspended.
+	 * Use kmalloc() rather than vmalloc() when we can.
+	 */
+	dmi = NULL;
+	if (param_kernel->data_size <= KMALLOC_MAX_SIZE) {
+		dmi = kmalloc(param_kernel->data_size, GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
+		if (dmi)
+			*param_flags |= DM_PARAMS_KMALLOC;
+	}
+
+	if (!dmi) {
+		unsigned noio_flag;
+		noio_flag = memalloc_noio_save();
+		dmi = __vmalloc(param_kernel->data_size, GFP_NOIO | __GFP_REPEAT | __GFP_HIGH | __GFP_HIGHMEM, PAGE_KERNEL);
+		memalloc_noio_restore(noio_flag);
+		if (dmi)
+			*param_flags |= DM_PARAMS_VMALLOC;
+	}
+
+	if (!dmi) {
+		if (secure_data && clear_user(user, param_kernel->data_size))
+			return -EFAULT;
 		return -ENOMEM;
+	}
 
-	if (copy_from_user(dmi, user, tmp.data_size)) {
-		vfree(dmi);
-		return -EFAULT;
+	if (copy_from_user(dmi, user, param_kernel->data_size))
+		goto bad;
+
+data_copied:
+	/*
+	 * Abort if something changed the ioctl data while it was being copied.
+	 */
+	if (dmi->data_size != param_kernel->data_size) {
+		DMERR("rejecting ioctl: data size modified while processing parameters");
+		goto bad;
 	}
 
+	/* Wipe the user buffer so we do not return it to userspace */
+	if (secure_data && clear_user(user, param_kernel->data_size))
+		goto bad;
+
 	*param = dmi;
 	return 0;
+
+bad:
+	free_params(dmi, param_kernel->data_size, *param_flags);
+
+	return -EFAULT;
 }
 
 static int validate_params(uint cmd, struct dm_ioctl *param)
 {
 	/* Always clear this flag */
 	param->flags &= ~DM_BUFFER_FULL_FLAG;
+	param->flags &= ~DM_UEVENT_GENERATED_FLAG;
+	param->flags &= ~DM_SECURE_DATA_FLAG;
+	param->flags &= ~DM_DATA_OUT_FLAG;
 
 	/* Ignores parameters */
 	if (cmd == DM_REMOVE_ALL_CMD ||
@@ -1401,10 +1789,13 @@ static int validate_params(uint cmd, struct dm_ioctl *param)
 static int ctl_ioctl(uint command, struct dm_ioctl __user *user)
 {
 	int r = 0;
+	int ioctl_flags;
+	int param_flags;
 	unsigned int cmd;
 	struct dm_ioctl *uninitialized_var(param);
 	ioctl_fn fn = NULL;
-	size_t param_size;
+	size_t input_param_size;
+	struct dm_ioctl param_kernel;
 
 	/* only root can play with this */
 	if (!capable(CAP_SYS_ADMIN))
@@ -1429,35 +1820,31 @@ static int ctl_ioctl(uint command, struct dm_ioctl __user *user)
 	if (cmd == DM_VERSION_CMD)
 		return 0;
 
-	fn = lookup_ioctl(cmd);
+	fn = lookup_ioctl(cmd, &ioctl_flags);
 	if (!fn) {
 		DMWARN("dm_ctl_ioctl: unknown command 0x%x", command);
 		return -ENOTTY;
 	}
 
 	/*
-	 * Trying to avoid low memory issues when a device is
-	 * suspended.
-	 */
-	current->flags |= PF_MEMALLOC;
-
-	/*
 	 * Copy the parameters into kernel space.
 	 */
-	r = copy_params(user, &param);
-
-	current->flags &= ~PF_MEMALLOC;
+	r = copy_params(user, &param_kernel, ioctl_flags, &param, &param_flags);
 
 	if (r)
 		return r;
 
+	input_param_size = param->data_size;
 	r = validate_params(cmd, param);
 	if (r)
 		goto out;
 
-	param_size = param->data_size;
 	param->data_size = sizeof(*param);
-	r = fn(param, param_size);
+	r = fn(param, input_param_size);
+
+	if (unlikely(param->flags & DM_BUFFER_FULL_FLAG) &&
+	    unlikely(ioctl_flags & IOCTL_FLAGS_NO_PARAMS))
+		DMERR("ioctl %d tried to output some data but has IOCTL_FLAGS_NO_PARAMS set", cmd);
 
 	/*
 	 * Copy the results back to userland.
@@ -1465,8 +1852,8 @@ static int ctl_ioctl(uint command, struct dm_ioctl __user *user)
 	if (!r && copy_to_user(user, param, param->data_size))
 		r = -EFAULT;
 
- out:
-	free_params(param);
+out:
+	free_params(param, input_param_size, param_flags);
 	return r;
 }
 
@@ -1485,17 +1872,23 @@ static long dm_compat_ctl_ioctl(struct file *file, uint command, ulong u)
 #endif
 
 static const struct file_operations _ctl_fops = {
+	.open = nonseekable_open,
 	.unlocked_ioctl	 = dm_ctl_ioctl,
 	.compat_ioctl = dm_compat_ctl_ioctl,
 	.owner	 = THIS_MODULE,
+	.llseek  = noop_llseek,
 };
 
 static struct miscdevice _dm_misc = {
-	.minor 		= MISC_DYNAMIC_MINOR,
+	.minor		= MAPPER_CTRL_MINOR,
 	.name  		= DM_NAME,
+	.nodename	= DM_DIR "/" DM_CONTROL_NODE,
 	.fops  		= &_ctl_fops
 };
 
+MODULE_ALIAS_MISCDEV(MAPPER_CTRL_MINOR);
+MODULE_ALIAS("devname:" DM_DIR "/" DM_CONTROL_NODE);
+
 /*
  * Create misc character device and link to DM_DIR/control.
  */
@@ -1542,20 +1935,20 @@ int dm_copy_name_and_uuid(struct mapped_device *md, char *name, char *uuid)
 	if (!md)
 		return -ENXIO;
 
-	dm_get(md);
-	down_read(&_hash_lock);
+	mutex_lock(&dm_hash_cells_mutex);
 	hc = dm_get_mdptr(md);
 	if (!hc || hc->md != md) {
 		r = -ENXIO;
 		goto out;
 	}
 
-	strcpy(name, hc->name);
-	strcpy(uuid, hc->uuid ? : "");
+	if (name)
+		strcpy(name, hc->name);
+	if (uuid)
+		strcpy(uuid, hc->uuid ? : "");
 
 out:
-	up_read(&_hash_lock);
-	dm_put(md);
+	mutex_unlock(&dm_hash_cells_mutex);
 
 	return r;
 }
diff --git a/drivers/md/dm-kcopyd.c b/drivers/md/dm-kcopyd.c
index 996802b8a45..3a7cade5e27 100644
--- a/drivers/md/dm-kcopyd.c
+++ b/drivers/md/dm-kcopyd.c
@@ -10,7 +10,7 @@
  */
 
 #include <linux/types.h>
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 #include <linux/blkdev.h>
 #include <linux/fs.h>
 #include <linux/init.h>
@@ -22,19 +22,25 @@
 #include <linux/vmalloc.h>
 #include <linux/workqueue.h>
 #include <linux/mutex.h>
+#include <linux/delay.h>
+#include <linux/device-mapper.h>
 #include <linux/dm-kcopyd.h>
 
 #include "dm.h"
 
+#define SUB_JOB_SIZE	128
+#define SPLIT_COUNT	8
+#define MIN_JOBS	8
+#define RESERVE_PAGES	(DIV_ROUND_UP(SUB_JOB_SIZE << SECTOR_SHIFT, PAGE_SIZE))
+
 /*-----------------------------------------------------------------
  * Each kcopyd client has its own little pool of preallocated
  * pages for kcopyd io.
  *---------------------------------------------------------------*/
 struct dm_kcopyd_client {
-	spinlock_t lock;
 	struct page_list *pages;
-	unsigned int nr_pages;
-	unsigned int nr_free_pages;
+	unsigned nr_reserved_pages;
+	unsigned nr_free_pages;
 
 	struct dm_io_client *io_client;
 
@@ -46,6 +52,8 @@ struct dm_kcopyd_client {
 	struct workqueue_struct *kcopyd_wq;
 	struct work_struct kcopyd_work;
 
+	struct dm_kcopyd_throttle *throttle;
+
 /*
  * We maintain three lists of jobs:
  *
@@ -61,20 +69,136 @@ struct dm_kcopyd_client {
 	struct list_head pages_jobs;
 };
 
+static struct page_list zero_page_list;
+
+static DEFINE_SPINLOCK(throttle_spinlock);
+
+/*
+ * IO/IDLE accounting slowly decays after (1 << ACCOUNT_INTERVAL_SHIFT) period.
+ * When total_period >= (1 << ACCOUNT_INTERVAL_SHIFT) the counters are divided
+ * by 2.
+ */
+#define ACCOUNT_INTERVAL_SHIFT		SHIFT_HZ
+
+/*
+ * Sleep this number of milliseconds.
+ *
+ * The value was decided experimentally.
+ * Smaller values seem to cause an increased copy rate above the limit.
+ * The reason for this is unknown but possibly due to jiffies rounding errors
+ * or read/write cache inside the disk.
+ */
+#define SLEEP_MSEC			100
+
+/*
+ * Maximum number of sleep events. There is a theoretical livelock if more
+ * kcopyd clients do work simultaneously which this limit avoids.
+ */
+#define MAX_SLEEPS			10
+
+static void io_job_start(struct dm_kcopyd_throttle *t)
+{
+	unsigned throttle, now, difference;
+	int slept = 0, skew;
+
+	if (unlikely(!t))
+		return;
+
+try_again:
+	spin_lock_irq(&throttle_spinlock);
+
+	throttle = ACCESS_ONCE(t->throttle);
+
+	if (likely(throttle >= 100))
+		goto skip_limit;
+
+	now = jiffies;
+	difference = now - t->last_jiffies;
+	t->last_jiffies = now;
+	if (t->num_io_jobs)
+		t->io_period += difference;
+	t->total_period += difference;
+
+	/*
+	 * Maintain sane values if we got a temporary overflow.
+	 */
+	if (unlikely(t->io_period > t->total_period))
+		t->io_period = t->total_period;
+
+	if (unlikely(t->total_period >= (1 << ACCOUNT_INTERVAL_SHIFT))) {
+		int shift = fls(t->total_period >> ACCOUNT_INTERVAL_SHIFT);
+		t->total_period >>= shift;
+		t->io_period >>= shift;
+	}
+
+	skew = t->io_period - throttle * t->total_period / 100;
+
+	if (unlikely(skew > 0) && slept < MAX_SLEEPS) {
+		slept++;
+		spin_unlock_irq(&throttle_spinlock);
+		msleep(SLEEP_MSEC);
+		goto try_again;
+	}
+
+skip_limit:
+	t->num_io_jobs++;
+
+	spin_unlock_irq(&throttle_spinlock);
+}
+
+static void io_job_finish(struct dm_kcopyd_throttle *t)
+{
+	unsigned long flags;
+
+	if (unlikely(!t))
+		return;
+
+	spin_lock_irqsave(&throttle_spinlock, flags);
+
+	t->num_io_jobs--;
+
+	if (likely(ACCESS_ONCE(t->throttle) >= 100))
+		goto skip_limit;
+
+	if (!t->num_io_jobs) {
+		unsigned now, difference;
+
+		now = jiffies;
+		difference = now - t->last_jiffies;
+		t->last_jiffies = now;
+
+		t->io_period += difference;
+		t->total_period += difference;
+
+		/*
+		 * Maintain sane values if we got a temporary overflow.
+		 */
+		if (unlikely(t->io_period > t->total_period))
+			t->io_period = t->total_period;
+	}
+
+skip_limit:
+	spin_unlock_irqrestore(&throttle_spinlock, flags);
+}
+
+
 static void wake(struct dm_kcopyd_client *kc)
 {
 	queue_work(kc->kcopyd_wq, &kc->kcopyd_work);
 }
 
-static struct page_list *alloc_pl(void)
+/*
+ * Obtain one page for the use of kcopyd.
+ */
+static struct page_list *alloc_pl(gfp_t gfp)
 {
 	struct page_list *pl;
 
-	pl = kmalloc(sizeof(*pl), GFP_KERNEL);
+	pl = kmalloc(sizeof(*pl), gfp);
 	if (!pl)
 		return NULL;
 
-	pl->page = alloc_page(GFP_KERNEL);
+	pl->page = alloc_page(gfp);
 	if (!pl->page) {
 		kfree(pl);
 		return NULL;
@@ -89,41 +213,56 @@ static void free_pl(struct page_list *pl)
 	kfree(pl);
 }
 
-static int kcopyd_get_pages(struct dm_kcopyd_client *kc,
-			    unsigned int nr, struct page_list **pages)
+/*
+ * Add the provided pages to a client's free page list, releasing
+ * back to the system any beyond the reserved_pages limit.
+ */
+static void kcopyd_put_pages(struct dm_kcopyd_client *kc, struct page_list *pl)
 {
-	struct page_list *pl;
-
-	spin_lock(&kc->lock);
-	if (kc->nr_free_pages < nr) {
-		spin_unlock(&kc->lock);
-		return -ENOMEM;
-	}
-
-	kc->nr_free_pages -= nr;
-	for (*pages = pl = kc->pages; --nr; pl = pl->next)
-		;
+	struct page_list *next;
 
-	kc->pages = pl->next;
-	pl->next = NULL;
+	do {
+		next = pl->next;
 
-	spin_unlock(&kc->lock);
+		if (kc->nr_free_pages >= kc->nr_reserved_pages)
+			free_pl(pl);
+		else {
+			pl->next = kc->pages;
+			kc->pages = pl;
+			kc->nr_free_pages++;
+		}
 
-	return 0;
+		pl = next;
+	} while (pl);
 }
 
-static void kcopyd_put_pages(struct dm_kcopyd_client *kc, struct page_list *pl)
+static int kcopyd_get_pages(struct dm_kcopyd_client *kc,
+			    unsigned int nr, struct page_list **pages)
 {
-	struct page_list *cursor;
+	struct page_list *pl;
 
-	spin_lock(&kc->lock);
-	for (cursor = pl; cursor->next; cursor = cursor->next)
-		kc->nr_free_pages++;
+	*pages = NULL;
+
+	do {
+		pl = alloc_pl(__GFP_NOWARN | __GFP_NORETRY);
+		if (unlikely(!pl)) {
+			/* Use reserved pages */
+			pl = kc->pages;
+			if (unlikely(!pl))
+				goto out_of_memory;
+			kc->pages = pl->next;
+			kc->nr_free_pages--;
+		}
+		pl->next = *pages;
+		*pages = pl;
+	} while (--nr);
 
-	kc->nr_free_pages++;
-	cursor->next = kc->pages;
-	kc->pages = pl;
-	spin_unlock(&kc->lock);
+	return 0;
+
+out_of_memory:
+	if (*pages)
+		kcopyd_put_pages(kc, *pages);
+	return -ENOMEM;
 }
 
 /*
@@ -140,13 +279,16 @@ static void drop_pages(struct page_list *pl)
 	}
 }
 
-static int client_alloc_pages(struct dm_kcopyd_client *kc, unsigned int nr)
+/*
+ * Allocate and reserve nr_pages for the use of a specific client.
+ */
+static int client_reserve_pages(struct dm_kcopyd_client *kc, unsigned nr_pages)
 {
-	unsigned int i;
+	unsigned i;
 	struct page_list *pl = NULL, *next;
 
-	for (i = 0; i < nr; i++) {
-		next = alloc_pl();
+	for (i = 0; i < nr_pages; i++) {
+		next = alloc_pl(GFP_KERNEL);
 		if (!next) {
 			if (pl)
 				drop_pages(pl);
@@ -156,17 +298,18 @@ static int client_alloc_pages(struct dm_kcopyd_client *kc, unsigned int nr)
 		pl = next;
 	}
 
+	kc->nr_reserved_pages += nr_pages;
 	kcopyd_put_pages(kc, pl);
-	kc->nr_pages += nr;
+
 	return 0;
 }
 
 static void client_free_pages(struct dm_kcopyd_client *kc)
 {
-	BUG_ON(kc->nr_free_pages != kc->nr_pages);
+	BUG_ON(kc->nr_free_pages != kc->nr_reserved_pages);
 	drop_pages(kc->pages);
 	kc->pages = NULL;
-	kc->nr_free_pages = kc->nr_pages = 0;
+	kc->nr_free_pages = kc->nr_reserved_pages = 0;
 }
 
 /*-----------------------------------------------------------------
@@ -197,8 +340,6 @@ struct kcopyd_job {
 	unsigned int num_dests;
 	struct dm_io_region dests[DM_KCOPYD_MAX_REGIONS];
 
-	sector_t offset;
-	unsigned int nr_pages;
 	struct page_list *pages;
 
 	/*
@@ -215,19 +356,23 @@ struct kcopyd_job {
 	struct mutex lock;
 	atomic_t sub_jobs;
 	sector_t progress;
-};
 
-/* FIXME: this should scale with the number of pages */
-#define MIN_JOBS 512
+	struct kcopyd_job *master_job;
+};
 
 static struct kmem_cache *_job_cache;
 
 int __init dm_kcopyd_init(void)
 {
-	_job_cache = KMEM_CACHE(kcopyd_job, 0);
+	_job_cache = kmem_cache_create("kcopyd_job",
+				sizeof(struct kcopyd_job) * (SPLIT_COUNT + 1),
+				__alignof__(struct kcopyd_job), 0, NULL);
 	if (!_job_cache)
 		return -ENOMEM;
 
+	zero_page_list.next = &zero_page_list;
+	zero_page_list.page = ZERO_PAGE(0);
+
 	return 0;
 }
 
@@ -268,6 +413,17 @@ static void push(struct list_head *jobs, struct kcopyd_job *job)
 	spin_unlock_irqrestore(&kc->job_lock, flags);
 }
 
+
+static void push_head(struct list_head *jobs, struct kcopyd_job *job)
+{
+	unsigned long flags;
+	struct dm_kcopyd_client *kc = job->kc;
+
+	spin_lock_irqsave(&kc->job_lock, flags);
+	list_add(&job->list, jobs);
+	spin_unlock_irqrestore(&kc->job_lock, flags);
+}
+
 /*
  * These three functions process 1 item from the corresponding
  * job list.
@@ -285,8 +441,14 @@ static int run_complete_job(struct kcopyd_job *job)
 	dm_kcopyd_notify_fn fn = job->fn;
 	struct dm_kcopyd_client *kc = job->kc;
 
-	kcopyd_put_pages(kc, job->pages);
-	mempool_free(job, kc->job_pool);
+	if (job->pages && job->pages != &zero_page_list)
+		kcopyd_put_pages(kc, job->pages);
+	/*
+	 * If this is the master job, the sub jobs have already
+	 * completed so we can free everything.
+	 */
+	if (job->master_job == job)
+		mempool_free(job, kc->job_pool);
 	fn(read_err, write_err, context);
 
 	if (atomic_dec_and_test(&kc->nr_jobs))
@@ -300,8 +462,10 @@ static void complete_io(unsigned long error, void *context)
 	struct kcopyd_job *job = (struct kcopyd_job *) context;
 	struct dm_kcopyd_client *kc = job->kc;
 
+	io_job_finish(kc->throttle);
+
 	if (error) {
-		if (job->rw == WRITE)
+		if (job->rw & WRITE)
 			job->write_err |= error;
 		else
 			job->read_err = 1;
@@ -313,7 +477,7 @@ static void complete_io(unsigned long error, void *context)
 		}
 	}
 
-	if (job->rw == WRITE)
+	if (job->rw & WRITE)
 		push(&kc->complete_jobs, job);
 
 	else {
@@ -332,15 +496,17 @@ static int run_io_job(struct kcopyd_job *job)
 {
 	int r;
 	struct dm_io_request io_req = {
-		.bi_rw = job->rw | (1 << BIO_RW_SYNC),
+		.bi_rw = job->rw,
 		.mem.type = DM_IO_PAGE_LIST,
 		.mem.ptr.pl = job->pages,
-		.mem.offset = job->offset,
+		.mem.offset = 0,
 		.notify.fn = complete_io,
 		.notify.context = job,
 		.client = job->kc->io_client,
 	};
 
+	io_job_start(job->kc->throttle);
+
 	if (job->rw == READ)
 		r = dm_io(&io_req, 1, &job->source, NULL);
 	else
@@ -352,10 +518,9 @@ static int run_io_job(struct kcopyd_job *job)
 static int run_pages_job(struct kcopyd_job *job)
 {
 	int r;
+	unsigned nr_pages = dm_div_up(job->dests[0].count, PAGE_SIZE >> 9);
 
-	job->nr_pages = dm_div_up(job->dests[0].count + job->offset,
-				  PAGE_SIZE >> 9);
-	r = kcopyd_get_pages(job->kc, job->nr_pages, &job->pages);
+	r = kcopyd_get_pages(job->kc, nr_pages, &job->pages);
 	if (!r) {
 		/* this job is ready for io */
 		push(&job->kc->io_jobs, job);
@@ -385,7 +550,7 @@ static int process_jobs(struct list_head *jobs, struct dm_kcopyd_client *kc,
 
 		if (r < 0) {
 			/* error this rogue job */
-			if (job->rw == WRITE)
+			if (job->rw & WRITE)
 				job->write_err = (unsigned long) -1L;
 			else
 				job->read_err = 1;
@@ -398,7 +563,7 @@ static int process_jobs(struct list_head *jobs, struct dm_kcopyd_client *kc,
 			 * We couldn't service this job ATM, so
 			 * push this job back onto the list.
 			 */
-			push(jobs, job);
+			push_head(jobs, job);
 			break;
 		}
 
@@ -415,6 +580,7 @@ static void do_work(struct work_struct *work)
 {
 	struct dm_kcopyd_client *kc = container_of(work,
 					struct dm_kcopyd_client, kcopyd_work);
+	struct blk_plug plug;
 
 	/*
 	 * The order that these are called is *very* important.
@@ -423,9 +589,11 @@ static void do_work(struct work_struct *work)
 	 * list.  io jobs call wake when they complete and it all
 	 * starts again.
 	 */
+	blk_start_plug(&plug);
 	process_jobs(&kc->complete_jobs, kc, run_complete_job);
 	process_jobs(&kc->pages_jobs, kc, run_pages_job);
 	process_jobs(&kc->io_jobs, kc, run_io_job);
+	blk_finish_plug(&plug);
 }
 
 /*
@@ -437,18 +605,24 @@ static void dispatch_job(struct kcopyd_job *job)
 {
 	struct dm_kcopyd_client *kc = job->kc;
 	atomic_inc(&kc->nr_jobs);
-	push(&kc->pages_jobs, job);
+	if (unlikely(!job->source.count))
+		push(&kc->complete_jobs, job);
+	else if (job->pages == &zero_page_list)
+		push(&kc->io_jobs, job);
+	else
+		push(&kc->pages_jobs, job);
 	wake(kc);
 }
 
-#define SUB_JOB_SIZE 128
 static void segment_complete(int read_err, unsigned long write_err,
 			     void *context)
 {
 	/* FIXME: tidy this function */
 	sector_t progress = 0;
 	sector_t count = 0;
-	struct kcopyd_job *job = (struct kcopyd_job *) context;
+	struct kcopyd_job *sub_job = (struct kcopyd_job *) context;
+	struct kcopyd_job *job = sub_job->master_job;
+	struct dm_kcopyd_client *kc = job->kc;
 
 	mutex_lock(&job->lock);
 
@@ -478,8 +652,6 @@ static void segment_complete(int read_err, unsigned long write_err,
 
 	if (count) {
 		int i;
-		struct kcopyd_job *sub_job = mempool_alloc(job->kc->job_pool,
-							   GFP_NOIO);
 
 		*sub_job = *job;
 		sub_job->source.sector += progress;
@@ -491,34 +663,39 @@ static void segment_complete(int read_err, unsigned long write_err,
 		}
 
 		sub_job->fn = segment_complete;
-		sub_job->context = job;
+		sub_job->context = sub_job;
 		dispatch_job(sub_job);
 
 	} else if (atomic_dec_and_test(&job->sub_jobs)) {
 
 		/*
-		 * To avoid a race we must keep the job around
-		 * until after the notify function has completed.
-		 * Otherwise the client may try and stop the job
-		 * after we've completed.
+		 * Queue the completion callback to the kcopyd thread.
+		 *
+		 * Some callers assume that all the completions are called
+		 * from a single thread and don't race with each other.
+		 *
+		 * We must not call the callback directly here because this
+		 * code may not be executing in the thread.
 		 */
-		job->fn(read_err, write_err, job->context);
-		mempool_free(job, job->kc->job_pool);
+		push(&kc->complete_jobs, job);
+		wake(kc);
 	}
 }
 
 /*
- * Create some little jobs that will do the move between
- * them.
+ * Create some sub jobs to share the work between them.
  */
-#define SPLIT_COUNT 8
-static void split_job(struct kcopyd_job *job)
+static void split_job(struct kcopyd_job *master_job)
 {
 	int i;
 
-	atomic_set(&job->sub_jobs, SPLIT_COUNT);
-	for (i = 0; i < SPLIT_COUNT; i++)
-		segment_complete(0, 0u, job);
+	atomic_inc(&master_job->kc->nr_jobs);
+
+	atomic_set(&master_job->sub_jobs, SPLIT_COUNT);
+	for (i = 0; i < SPLIT_COUNT; i++) {
+		master_job[i + 1].master_job = master_job;
+		segment_complete(0, 0u, &master_job[i + 1]);
+	}
 }
 
 int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from,
@@ -526,9 +703,11 @@ int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from,
 		   unsigned int flags, dm_kcopyd_notify_fn fn, void *context)
 {
 	struct kcopyd_job *job;
+	int i;
 
 	/*
-	 * Allocate a new job.
+	 * Allocate an array of jobs consisting of one master job
+	 * followed by SPLIT_COUNT sub jobs.
 	 */
 	job = mempool_alloc(kc->job_pool, GFP_NOIO);
 
@@ -539,23 +718,36 @@ int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from,
 	job->flags = flags;
 	job->read_err = 0;
 	job->write_err = 0;
-	job->rw = READ;
-
-	job->source = *from;
 
 	job->num_dests = num_dests;
 	memcpy(&job->dests, dests, sizeof(*dests) * num_dests);
 
-	job->offset = 0;
-	job->nr_pages = 0;
-	job->pages = NULL;
+	if (from) {
+		job->source = *from;
+		job->pages = NULL;
+		job->rw = READ;
+	} else {
+		memset(&job->source, 0, sizeof job->source);
+		job->source.count = job->dests[0].count;
+		job->pages = &zero_page_list;
+
+		/*
+		 * Use WRITE SAME to optimize zeroing if all dests support it.
+		 */
+		job->rw = WRITE | REQ_WRITE_SAME;
+		for (i = 0; i < job->num_dests; i++)
+			if (!bdev_write_same(job->dests[i].bdev)) {
+				job->rw = WRITE;
+				break;
+			}
+	}
 
 	job->fn = fn;
 	job->context = context;
+	job->master_job = job;
 
-	if (job->source.count < SUB_JOB_SIZE)
+	if (job->source.count <= SUB_JOB_SIZE)
 		dispatch_job(job);
-
 	else {
 		mutex_init(&job->lock);
 		job->progress = 0;
@@ -566,6 +758,46 @@ int dm_kcopyd_copy(struct dm_kcopyd_client *kc, struct dm_io_region *from,
 }
 EXPORT_SYMBOL(dm_kcopyd_copy);
 
+int dm_kcopyd_zero(struct dm_kcopyd_client *kc,
+		   unsigned num_dests, struct dm_io_region *dests,
+		   unsigned flags, dm_kcopyd_notify_fn fn, void *context)
+{
+	return dm_kcopyd_copy(kc, NULL, num_dests, dests, flags, fn, context);
+}
+EXPORT_SYMBOL(dm_kcopyd_zero);
+
+void *dm_kcopyd_prepare_callback(struct dm_kcopyd_client *kc,
+				 dm_kcopyd_notify_fn fn, void *context)
+{
+	struct kcopyd_job *job;
+
+	job = mempool_alloc(kc->job_pool, GFP_NOIO);
+
+	memset(job, 0, sizeof(struct kcopyd_job));
+	job->kc = kc;
+	job->fn = fn;
+	job->context = context;
+	job->master_job = job;
+
+	atomic_inc(&kc->nr_jobs);
+
+	return job;
+}
+EXPORT_SYMBOL(dm_kcopyd_prepare_callback);
+
+void dm_kcopyd_do_callback(void *j, int read_err, unsigned long write_err)
+{
+	struct kcopyd_job *job = j;
+	struct dm_kcopyd_client *kc = job->kc;
+
+	job->read_err = read_err;
+	job->write_err = write_err;
+
+	push(&kc->complete_jobs, job);
+	wake(kc);
+}
+EXPORT_SYMBOL(dm_kcopyd_do_callback);
+
 /*
  * Cancels a kcopyd job, eg. someone might be deactivating a
  * mirror.
@@ -581,38 +813,37 @@ int kcopyd_cancel(struct kcopyd_job *job, int block)
 /*-----------------------------------------------------------------
  * Client setup
  *---------------------------------------------------------------*/
-int dm_kcopyd_client_create(unsigned int nr_pages,
-			    struct dm_kcopyd_client **result)
+struct dm_kcopyd_client *dm_kcopyd_client_create(struct dm_kcopyd_throttle *throttle)
 {
 	int r = -ENOMEM;
 	struct dm_kcopyd_client *kc;
 
 	kc = kmalloc(sizeof(*kc), GFP_KERNEL);
 	if (!kc)
-		return -ENOMEM;
+		return ERR_PTR(-ENOMEM);
 
-	spin_lock_init(&kc->lock);
 	spin_lock_init(&kc->job_lock);
 	INIT_LIST_HEAD(&kc->complete_jobs);
 	INIT_LIST_HEAD(&kc->io_jobs);
 	INIT_LIST_HEAD(&kc->pages_jobs);
+	kc->throttle = throttle;
 
 	kc->job_pool = mempool_create_slab_pool(MIN_JOBS, _job_cache);
 	if (!kc->job_pool)
 		goto bad_slab;
 
 	INIT_WORK(&kc->kcopyd_work, do_work);
-	kc->kcopyd_wq = create_singlethread_workqueue("kcopyd");
+	kc->kcopyd_wq = alloc_workqueue("kcopyd", WQ_MEM_RECLAIM, 0);
 	if (!kc->kcopyd_wq)
 		goto bad_workqueue;
 
 	kc->pages = NULL;
-	kc->nr_pages = kc->nr_free_pages = 0;
-	r = client_alloc_pages(kc, nr_pages);
+	kc->nr_reserved_pages = kc->nr_free_pages = 0;
+	r = client_reserve_pages(kc, RESERVE_PAGES);
 	if (r)
 		goto bad_client_pages;
 
-	kc->io_client = dm_io_client_create(nr_pages);
+	kc->io_client = dm_io_client_create();
 	if (IS_ERR(kc->io_client)) {
 		r = PTR_ERR(kc->io_client);
 		goto bad_io_client;
@@ -621,8 +852,7 @@ int dm_kcopyd_client_create(unsigned int nr_pages,
 	init_waitqueue_head(&kc->destroyq);
 	atomic_set(&kc->nr_jobs, 0);
 
-	*result = kc;
-	return 0;
+	return kc;
 
 bad_io_client:
 	client_free_pages(kc);
@@ -633,7 +863,7 @@ bad_workqueue:
 bad_slab:
 	kfree(kc);
 
-	return r;
+	return ERR_PTR(r);
 }
 EXPORT_SYMBOL(dm_kcopyd_client_create);
 
diff --git a/drivers/md/dm-linear.c b/drivers/md/dm-linear.c
index 6449bcdf84c..53e848c1093 100644
--- a/drivers/md/dm-linear.c
+++ b/drivers/md/dm-linear.c
@@ -5,12 +5,12 @@
  */
 
 #include "dm.h"
-
 #include <linux/module.h>
 #include <linux/init.h>
 #include <linux/blkdev.h>
 #include <linux/bio.h>
 #include <linux/slab.h>
+#include <linux/device-mapper.h>
 
 #define DM_MSG_PREFIX "linear"
 
@@ -29,6 +29,7 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
 	struct linear_c *lc;
 	unsigned long long tmp;
+	char dummy;
 
 	if (argc != 2) {
 		ti->error = "Invalid argument count";
@@ -41,18 +42,20 @@ static int linear_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -ENOMEM;
 	}
 
-	if (sscanf(argv[1], "%llu", &tmp) != 1) {
+	if (sscanf(argv[1], "%llu%c", &tmp, &dummy) != 1) {
 		ti->error = "dm-linear: Invalid device sector";
 		goto bad;
 	}
 	lc->start = tmp;
 
-	if (dm_get_device(ti, argv[0], lc->start, ti->len,
-			  dm_table_get_mode(ti->table), &lc->dev)) {
+	if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &lc->dev)) {
 		ti->error = "dm-linear: Device lookup failed";
 		goto bad;
 	}
 
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->num_write_same_bios = 1;
 	ti->private = lc;
 	return 0;
 
@@ -73,7 +76,7 @@ static sector_t linear_map_sector(struct dm_target *ti, sector_t bi_sector)
 {
 	struct linear_c *lc = ti->private;
 
-	return lc->start + (bi_sector - ti->begin);
+	return lc->start + dm_target_offset(ti, bi_sector);
 }
 
 static void linear_map_bio(struct dm_target *ti, struct bio *bio)
@@ -81,19 +84,20 @@ static void linear_map_bio(struct dm_target *ti, struct bio *bio)
 	struct linear_c *lc = ti->private;
 
 	bio->bi_bdev = lc->dev->bdev;
-	bio->bi_sector = linear_map_sector(ti, bio->bi_sector);
+	if (bio_sectors(bio))
+		bio->bi_iter.bi_sector =
+			linear_map_sector(ti, bio->bi_iter.bi_sector);
 }
 
-static int linear_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static int linear_map(struct dm_target *ti, struct bio *bio)
 {
 	linear_map_bio(ti, bio);
 
 	return DM_MAPIO_REMAPPED;
 }
 
-static int linear_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned int maxlen)
+static void linear_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
 {
 	struct linear_c *lc = (struct linear_c *) ti->private;
 
@@ -107,23 +111,23 @@ static int linear_status(struct dm_target *ti, status_type_t type,
 				(unsigned long long)lc->start);
 		break;
 	}
-	return 0;
 }
 
-static int linear_ioctl(struct dm_target *ti, struct inode *inode,
-			struct file *filp, unsigned int cmd,
+static int linear_ioctl(struct dm_target *ti, unsigned int cmd,
 			unsigned long arg)
 {
 	struct linear_c *lc = (struct linear_c *) ti->private;
-	struct block_device *bdev = lc->dev->bdev;
-	struct file fake_file = {};
-	struct dentry fake_dentry = {};
+	struct dm_dev *dev = lc->dev;
+	int r = 0;
 
-	fake_file.f_mode = lc->dev->mode;
-	fake_file.f_path.dentry = &fake_dentry;
-	fake_dentry.d_inode = bdev->bd_inode;
+	/*
+	 * Only pass ioctls through if the device sizes match exactly.
+	 */
+	if (lc->start ||
+	    ti->len != i_size_read(dev->bdev->bd_inode) >> SECTOR_SHIFT)
+		r = scsi_verify_blk_ioctl(NULL, cmd);
 
-	return blkdev_driver_ioctl(bdev->bd_inode, &fake_file, bdev->bd_disk, cmd, arg);
+	return r ? : __blkdev_driver_ioctl(dev->bdev, dev->mode, cmd, arg);
 }
 
 static int linear_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
@@ -141,9 +145,17 @@ static int linear_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
 	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
 }
 
+static int linear_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct linear_c *lc = ti->private;
+
+	return fn(ti, lc->dev, lc->start, ti->len, data);
+}
+
 static struct target_type linear_target = {
 	.name   = "linear",
-	.version= {1, 0, 3},
+	.version = {1, 2, 1},
 	.module = THIS_MODULE,
 	.ctr    = linear_ctr,
 	.dtr    = linear_dtr,
@@ -151,6 +163,7 @@ static struct target_type linear_target = {
 	.status = linear_status,
 	.ioctl  = linear_ioctl,
 	.merge  = linear_merge,
+	.iterate_devices = linear_iterate_devices,
 };
 
 int __init dm_linear_init(void)
@@ -165,8 +178,5 @@ int __init dm_linear_init(void)
 
 void dm_linear_exit(void)
 {
-	int r = dm_unregister_target(&linear_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&linear_target);
 }
diff --git a/drivers/md/dm-log-userspace-base.c b/drivers/md/dm-log-userspace-base.c
new file mode 100644
index 00000000000..b953db6cc22
--- /dev/null
+++ b/drivers/md/dm-log-userspace-base.c
@@ -0,0 +1,930 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#include <linux/bio.h>
+#include <linux/slab.h>
+#include <linux/dm-dirty-log.h>
+#include <linux/device-mapper.h>
+#include <linux/dm-log-userspace.h>
+#include <linux/module.h>
+#include <linux/workqueue.h>
+
+#include "dm-log-userspace-transfer.h"
+
+#define DM_LOG_USERSPACE_VSN "1.3.0"
+
+struct flush_entry {
+	int type;
+	region_t region;
+	struct list_head list;
+};
+
+/*
+ * This limit on the number of mark and clear request is, to a degree,
+ * arbitrary.  However, there is some basis for the choice in the limits
+ * imposed on the size of data payload by dm-log-userspace-transfer.c:
+ * dm_consult_userspace().
+ */
+#define MAX_FLUSH_GROUP_COUNT 32
+
+struct log_c {
+	struct dm_target *ti;
+	struct dm_dev *log_dev;
+	uint32_t region_size;
+	region_t region_count;
+	uint64_t luid;
+	char uuid[DM_UUID_LEN];
+
+	char *usr_argv_str;
+	uint32_t usr_argc;
+
+	/*
+	 * in_sync_hint gets set when doing is_remote_recovering.  It
+	 * represents the first region that needs recovery.  IOW, the
+	 * first zero bit of sync_bits.  This can be useful for to limit
+	 * traffic for calls like is_remote_recovering and get_resync_work,
+	 * but be take care in its use for anything else.
+	 */
+	uint64_t in_sync_hint;
+
+	/*
+	 * Mark and clear requests are held until a flush is issued
+	 * so that we can group, and thereby limit, the amount of
+	 * network traffic between kernel and userspace.  The 'flush_lock'
+	 * is used to protect these lists.
+	 */
+	spinlock_t flush_lock;
+	struct list_head mark_list;
+	struct list_head clear_list;
+
+	/*
+	 * Workqueue for flush of clear region requests.
+	 */
+	struct workqueue_struct *dmlog_wq;
+	struct delayed_work flush_log_work;
+	atomic_t sched_flush;
+
+	/*
+	 * Combine userspace flush and mark requests for efficiency.
+	 */
+	uint32_t integrated_flush;
+};
+
+static mempool_t *flush_entry_pool;
+
+static void *flush_entry_alloc(gfp_t gfp_mask, void *pool_data)
+{
+	return kmalloc(sizeof(struct flush_entry), gfp_mask);
+}
+
+static void flush_entry_free(void *element, void *pool_data)
+{
+	kfree(element);
+}
+
+static int userspace_do_request(struct log_c *lc, const char *uuid,
+				int request_type, char *data, size_t data_size,
+				char *rdata, size_t *rdata_size)
+{
+	int r;
+
+	/*
+	 * If the server isn't there, -ESRCH is returned,
+	 * and we must keep trying until the server is
+	 * restored.
+	 */
+retry:
+	r = dm_consult_userspace(uuid, lc->luid, request_type, data,
+				 data_size, rdata, rdata_size);
+
+	if (r != -ESRCH)
+		return r;
+
+	DMERR(" Userspace log server not found.");
+	while (1) {
+		set_current_state(TASK_INTERRUPTIBLE);
+		schedule_timeout(2*HZ);
+		DMWARN("Attempting to contact userspace log server...");
+		r = dm_consult_userspace(uuid, lc->luid, DM_ULOG_CTR,
+					 lc->usr_argv_str,
+					 strlen(lc->usr_argv_str) + 1,
+					 NULL, NULL);
+		if (!r)
+			break;
+	}
+	DMINFO("Reconnected to userspace log server... DM_ULOG_CTR complete");
+	r = dm_consult_userspace(uuid, lc->luid, DM_ULOG_RESUME, NULL,
+				 0, NULL, NULL);
+	if (!r)
+		goto retry;
+
+	DMERR("Error trying to resume userspace log: %d", r);
+
+	return -ESRCH;
+}
+
+static int build_constructor_string(struct dm_target *ti,
+				    unsigned argc, char **argv,
+				    char **ctr_str)
+{
+	int i, str_size;
+	char *str = NULL;
+
+	*ctr_str = NULL;
+
+	/*
+	 * Determine overall size of the string.
+	 */
+	for (i = 0, str_size = 0; i < argc; i++)
+		str_size += strlen(argv[i]) + 1; /* +1 for space between args */
+
+	str_size += 20; /* Max number of chars in a printed u64 number */
+
+	str = kzalloc(str_size, GFP_KERNEL);
+	if (!str) {
+		DMWARN("Unable to allocate memory for constructor string");
+		return -ENOMEM;
+	}
+
+	str_size = sprintf(str, "%llu", (unsigned long long)ti->len);
+	for (i = 0; i < argc; i++)
+		str_size += sprintf(str + str_size, " %s", argv[i]);
+
+	*ctr_str = str;
+	return str_size;
+}
+
+static void do_flush(struct work_struct *work)
+{
+	int r;
+	struct log_c *lc = container_of(work, struct log_c, flush_log_work.work);
+
+	atomic_set(&lc->sched_flush, 0);
+
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH, NULL, 0, NULL, NULL);
+
+	if (r)
+		dm_table_event(lc->ti->table);
+}
+
+/*
+ * userspace_ctr
+ *
+ * argv contains:
+ *	<UUID> [integrated_flush] <other args>
+ * Where 'other args' are the userspace implementation-specific log
+ * arguments.
+ *
+ * Example:
+ *	<UUID> [integrated_flush] clustered-disk <arg count> <log dev>
+ *	<region_size> [[no]sync]
+ *
+ * This module strips off the <UUID> and uses it for identification
+ * purposes when communicating with userspace about a log.
+ *
+ * If integrated_flush is defined, the kernel combines flush
+ * and mark requests.
+ *
+ * The rest of the line, beginning with 'clustered-disk', is passed
+ * to the userspace ctr function.
+ */
+static int userspace_ctr(struct dm_dirty_log *log, struct dm_target *ti,
+			 unsigned argc, char **argv)
+{
+	int r = 0;
+	int str_size;
+	char *ctr_str = NULL;
+	struct log_c *lc = NULL;
+	uint64_t rdata;
+	size_t rdata_size = sizeof(rdata);
+	char *devices_rdata = NULL;
+	size_t devices_rdata_size = DM_NAME_LEN;
+
+	if (argc < 3) {
+		DMWARN("Too few arguments to userspace dirty log");
+		return -EINVAL;
+	}
+
+	lc = kzalloc(sizeof(*lc), GFP_KERNEL);
+	if (!lc) {
+		DMWARN("Unable to allocate userspace log context.");
+		return -ENOMEM;
+	}
+
+	/* The ptr value is sufficient for local unique id */
+	lc->luid = (unsigned long)lc;
+
+	lc->ti = ti;
+
+	if (strlen(argv[0]) > (DM_UUID_LEN - 1)) {
+		DMWARN("UUID argument too long.");
+		kfree(lc);
+		return -EINVAL;
+	}
+
+	lc->usr_argc = argc;
+
+	strncpy(lc->uuid, argv[0], DM_UUID_LEN);
+	argc--;
+	argv++;
+	spin_lock_init(&lc->flush_lock);
+	INIT_LIST_HEAD(&lc->mark_list);
+	INIT_LIST_HEAD(&lc->clear_list);
+
+	if (!strcasecmp(argv[0], "integrated_flush")) {
+		lc->integrated_flush = 1;
+		argc--;
+		argv++;
+	}
+
+	str_size = build_constructor_string(ti, argc, argv, &ctr_str);
+	if (str_size < 0) {
+		kfree(lc);
+		return str_size;
+	}
+
+	devices_rdata = kzalloc(devices_rdata_size, GFP_KERNEL);
+	if (!devices_rdata) {
+		DMERR("Failed to allocate memory for device information");
+		r = -ENOMEM;
+		goto out;
+	}
+
+	/*
+	 * Send table string and get back any opened device.
+	 */
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_CTR,
+				 ctr_str, str_size,
+				 devices_rdata, &devices_rdata_size);
+
+	if (r < 0) {
+		if (r == -ESRCH)
+			DMERR("Userspace log server not found");
+		else
+			DMERR("Userspace log server failed to create log");
+		goto out;
+	}
+
+	/* Since the region size does not change, get it now */
+	rdata_size = sizeof(rdata);
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_GET_REGION_SIZE,
+				 NULL, 0, (char *)&rdata, &rdata_size);
+
+	if (r) {
+		DMERR("Failed to get region size of dirty log");
+		goto out;
+	}
+
+	lc->region_size = (uint32_t)rdata;
+	lc->region_count = dm_sector_div_up(ti->len, lc->region_size);
+
+	if (devices_rdata_size) {
+		if (devices_rdata[devices_rdata_size - 1] != '\0') {
+			DMERR("DM_ULOG_CTR device return string not properly terminated");
+			r = -EINVAL;
+			goto out;
+		}
+		r = dm_get_device(ti, devices_rdata,
+				  dm_table_get_mode(ti->table), &lc->log_dev);
+		if (r)
+			DMERR("Failed to register %s with device-mapper",
+			      devices_rdata);
+	}
+
+	if (lc->integrated_flush) {
+		lc->dmlog_wq = alloc_workqueue("dmlogd", WQ_MEM_RECLAIM, 0);
+		if (!lc->dmlog_wq) {
+			DMERR("couldn't start dmlogd");
+			r = -ENOMEM;
+			goto out;
+		}
+
+		INIT_DELAYED_WORK(&lc->flush_log_work, do_flush);
+		atomic_set(&lc->sched_flush, 0);
+	}
+
+out:
+	kfree(devices_rdata);
+	if (r) {
+		kfree(lc);
+		kfree(ctr_str);
+	} else {
+		lc->usr_argv_str = ctr_str;
+		log->context = lc;
+	}
+
+	return r;
+}
+
+static void userspace_dtr(struct dm_dirty_log *log)
+{
+	struct log_c *lc = log->context;
+
+	if (lc->integrated_flush) {
+		/* flush workqueue */
+		if (atomic_read(&lc->sched_flush))
+			flush_delayed_work(&lc->flush_log_work);
+
+		destroy_workqueue(lc->dmlog_wq);
+	}
+
+	(void) dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_DTR,
+				    NULL, 0, NULL, NULL);
+
+	if (lc->log_dev)
+		dm_put_device(lc->ti, lc->log_dev);
+
+	kfree(lc->usr_argv_str);
+	kfree(lc);
+
+	return;
+}
+
+static int userspace_presuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_PRESUSPEND,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static int userspace_postsuspend(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	/*
+	 * Run planned flush earlier.
+	 */
+	if (lc->integrated_flush && atomic_read(&lc->sched_flush))
+		flush_delayed_work(&lc->flush_log_work);
+
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_POSTSUSPEND,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static int userspace_resume(struct dm_dirty_log *log)
+{
+	int r;
+	struct log_c *lc = log->context;
+
+	lc->in_sync_hint = 0;
+	r = dm_consult_userspace(lc->uuid, lc->luid, DM_ULOG_RESUME,
+				 NULL, 0, NULL, NULL);
+
+	return r;
+}
+
+static uint32_t userspace_get_region_size(struct dm_dirty_log *log)
+{
+	struct log_c *lc = log->context;
+
+	return lc->region_size;
+}
+
+/*
+ * userspace_is_clean
+ *
+ * Check whether a region is clean.  If there is any sort of
+ * failure when consulting the server, we return not clean.
+ *
+ * Returns: 1 if clean, 0 otherwise
+ */
+static int userspace_is_clean(struct dm_dirty_log *log, region_t region)
+{
+	int r;
+	uint64_t region64 = (uint64_t)region;
+	int64_t is_clean;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+
+	rdata_size = sizeof(is_clean);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_IS_CLEAN,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&is_clean, &rdata_size);
+
+	return (r) ? 0 : (int)is_clean;
+}
+
+/*
+ * userspace_in_sync
+ *
+ * Check if the region is in-sync.  If there is any sort
+ * of failure when consulting the server, we assume that
+ * the region is not in sync.
+ *
+ * If 'can_block' is set, return immediately
+ *
+ * Returns: 1 if in-sync, 0 if not-in-sync, -EWOULDBLOCK
+ */
+static int userspace_in_sync(struct dm_dirty_log *log, region_t region,
+			     int can_block)
+{
+	int r;
+	uint64_t region64 = region;
+	int64_t in_sync;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+
+	/*
+	 * We can never respond directly - even if in_sync_hint is
+	 * set.  This is because another machine could see a device
+	 * failure and mark the region out-of-sync.  If we don't go
+	 * to userspace to ask, we might think the region is in-sync
+	 * and allow a read to pick up data that is stale.  (This is
+	 * very unlikely if a device actually fails; but it is very
+	 * likely if a connection to one device from one machine fails.)
+	 *
+	 * There still might be a problem if the mirror caches the region
+	 * state as in-sync... but then this call would not be made.  So,
+	 * that is a mirror problem.
+	 */
+	if (!can_block)
+		return -EWOULDBLOCK;
+
+	rdata_size = sizeof(in_sync);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_IN_SYNC,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&in_sync, &rdata_size);
+	return (r) ? 0 : (int)in_sync;
+}
+
+static int flush_one_by_one(struct log_c *lc, struct list_head *flush_list)
+{
+	int r = 0;
+	struct flush_entry *fe;
+
+	list_for_each_entry(fe, flush_list, list) {
+		r = userspace_do_request(lc, lc->uuid, fe->type,
+					 (char *)&fe->region,
+					 sizeof(fe->region),
+					 NULL, NULL);
+		if (r)
+			break;
+	}
+
+	return r;
+}
+
+static int flush_by_group(struct log_c *lc, struct list_head *flush_list,
+			  int flush_with_payload)
+{
+	int r = 0;
+	int count;
+	uint32_t type = 0;
+	struct flush_entry *fe, *tmp_fe;
+	LIST_HEAD(tmp_list);
+	uint64_t group[MAX_FLUSH_GROUP_COUNT];
+
+	/*
+	 * Group process the requests
+	 */
+	while (!list_empty(flush_list)) {
+		count = 0;
+
+		list_for_each_entry_safe(fe, tmp_fe, flush_list, list) {
+			group[count] = fe->region;
+			count++;
+
+			list_move(&fe->list, &tmp_list);
+
+			type = fe->type;
+			if (count >= MAX_FLUSH_GROUP_COUNT)
+				break;
+		}
+
+		if (flush_with_payload) {
+			r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH,
+						 (char *)(group),
+						 count * sizeof(uint64_t),
+						 NULL, NULL);
+			/*
+			 * Integrated flush failed.
+			 */
+			if (r)
+				break;
+		} else {
+			r = userspace_do_request(lc, lc->uuid, type,
+						 (char *)(group),
+						 count * sizeof(uint64_t),
+						 NULL, NULL);
+			if (r) {
+				/*
+				 * Group send failed.  Attempt one-by-one.
+				 */
+				list_splice_init(&tmp_list, flush_list);
+				r = flush_one_by_one(lc, flush_list);
+				break;
+			}
+		}
+	}
+
+	/*
+	 * Must collect flush_entrys that were successfully processed
+	 * as a group so that they will be free'd by the caller.
+	 */
+	list_splice_init(&tmp_list, flush_list);
+
+	return r;
+}
+
+/*
+ * userspace_flush
+ *
+ * This function is ok to block.
+ * The flush happens in two stages.  First, it sends all
+ * clear/mark requests that are on the list.  Then it
+ * tells the server to commit them.  This gives the
+ * server a chance to optimise the commit, instead of
+ * doing it for every request.
+ *
+ * Additionally, we could implement another thread that
+ * sends the requests up to the server - reducing the
+ * load on flush.  Then the flush would have less in
+ * the list and be responsible for the finishing commit.
+ *
+ * Returns: 0 on success, < 0 on failure
+ */
+static int userspace_flush(struct dm_dirty_log *log)
+{
+	int r = 0;
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	LIST_HEAD(mark_list);
+	LIST_HEAD(clear_list);
+	int mark_list_is_empty;
+	int clear_list_is_empty;
+	struct flush_entry *fe, *tmp_fe;
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	list_splice_init(&lc->mark_list, &mark_list);
+	list_splice_init(&lc->clear_list, &clear_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	mark_list_is_empty = list_empty(&mark_list);
+	clear_list_is_empty = list_empty(&clear_list);
+
+	if (mark_list_is_empty && clear_list_is_empty)
+		return 0;
+
+	r = flush_by_group(lc, &clear_list, 0);
+	if (r)
+		goto out;
+
+	if (!lc->integrated_flush) {
+		r = flush_by_group(lc, &mark_list, 0);
+		if (r)
+			goto out;
+		r = userspace_do_request(lc, lc->uuid, DM_ULOG_FLUSH,
+					 NULL, 0, NULL, NULL);
+		goto out;
+	}
+
+	/*
+	 * Send integrated flush request with mark_list as payload.
+	 */
+	r = flush_by_group(lc, &mark_list, 1);
+	if (r)
+		goto out;
+
+	if (mark_list_is_empty && !atomic_read(&lc->sched_flush)) {
+		/*
+		 * When there are only clear region requests,
+		 * we schedule a flush in the future.
+		 */
+		queue_delayed_work(lc->dmlog_wq, &lc->flush_log_work, 3 * HZ);
+		atomic_set(&lc->sched_flush, 1);
+	} else {
+		/*
+		 * Cancel pending flush because we
+		 * have already flushed in mark_region.
+		 */
+		cancel_delayed_work(&lc->flush_log_work);
+		atomic_set(&lc->sched_flush, 0);
+	}
+
+out:
+	/*
+	 * We can safely remove these entries, even after failure.
+	 * Calling code will receive an error and will know that
+	 * the log facility has failed.
+	 */
+	list_for_each_entry_safe(fe, tmp_fe, &mark_list, list) {
+		list_del(&fe->list);
+		mempool_free(fe, flush_entry_pool);
+	}
+	list_for_each_entry_safe(fe, tmp_fe, &clear_list, list) {
+		list_del(&fe->list);
+		mempool_free(fe, flush_entry_pool);
+	}
+
+	if (r)
+		dm_table_event(lc->ti->table);
+
+	return r;
+}
+
+/*
+ * userspace_mark_region
+ *
+ * This function should avoid blocking unless absolutely required.
+ * (Memory allocation is valid for blocking.)
+ */
+static void userspace_mark_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct flush_entry *fe;
+
+	/* Wait for an allocation, but _never_ fail */
+	fe = mempool_alloc(flush_entry_pool, GFP_NOIO);
+	BUG_ON(!fe);
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	fe->type = DM_ULOG_MARK_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->mark_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	return;
+}
+
+/*
+ * userspace_clear_region
+ *
+ * This function must not block.
+ * So, the alloc can't block.  In the worst case, it is ok to
+ * fail.  It would simply mean we can't clear the region.
+ * Does nothing to current sync context, but does mean
+ * the region will be re-sync'ed on a reload of the mirror
+ * even though it is in-sync.
+ */
+static void userspace_clear_region(struct dm_dirty_log *log, region_t region)
+{
+	unsigned long flags;
+	struct log_c *lc = log->context;
+	struct flush_entry *fe;
+
+	/*
+	 * If we fail to allocate, we skip the clearing of
+	 * the region.  This doesn't hurt us in any way, except
+	 * to cause the region to be resync'ed when the
+	 * device is activated next time.
+	 */
+	fe = mempool_alloc(flush_entry_pool, GFP_ATOMIC);
+	if (!fe) {
+		DMERR("Failed to allocate memory to clear region.");
+		return;
+	}
+
+	spin_lock_irqsave(&lc->flush_lock, flags);
+	fe->type = DM_ULOG_CLEAR_REGION;
+	fe->region = region;
+	list_add(&fe->list, &lc->clear_list);
+	spin_unlock_irqrestore(&lc->flush_lock, flags);
+
+	return;
+}
+
+/*
+ * userspace_get_resync_work
+ *
+ * Get a region that needs recovery.  It is valid to return
+ * an error for this function.
+ *
+ * Returns: 1 if region filled, 0 if no work, <0 on error
+ */
+static int userspace_get_resync_work(struct dm_dirty_log *log, region_t *region)
+{
+	int r;
+	size_t rdata_size;
+	struct log_c *lc = log->context;
+	struct {
+		int64_t i; /* 64-bit for mix arch compatibility */
+		region_t r;
+	} pkg;
+
+	if (lc->in_sync_hint >= lc->region_count)
+		return 0;
+
+	rdata_size = sizeof(pkg);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_GET_RESYNC_WORK,
+				 NULL, 0, (char *)&pkg, &rdata_size);
+
+	*region = pkg.r;
+	return (r) ? r : (int)pkg.i;
+}
+
+/*
+ * userspace_set_region_sync
+ *
+ * Set the sync status of a given region.  This function
+ * must not fail.
+ */
+static void userspace_set_region_sync(struct dm_dirty_log *log,
+				      region_t region, int in_sync)
+{
+	int r;
+	struct log_c *lc = log->context;
+	struct {
+		region_t r;
+		int64_t i;
+	} pkg;
+
+	pkg.r = region;
+	pkg.i = (int64_t)in_sync;
+
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_SET_REGION_SYNC,
+				 (char *)&pkg, sizeof(pkg), NULL, NULL);
+
+	/*
+	 * It would be nice to be able to report failures.
+	 * However, it is easy emough to detect and resolve.
+	 */
+	return;
+}
+
+/*
+ * userspace_get_sync_count
+ *
+ * If there is any sort of failure when consulting the server,
+ * we assume that the sync count is zero.
+ *
+ * Returns: sync count on success, 0 on failure
+ */
+static region_t userspace_get_sync_count(struct dm_dirty_log *log)
+{
+	int r;
+	size_t rdata_size;
+	uint64_t sync_count;
+	struct log_c *lc = log->context;
+
+	rdata_size = sizeof(sync_count);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_GET_SYNC_COUNT,
+				 NULL, 0, (char *)&sync_count, &rdata_size);
+
+	if (r)
+		return 0;
+
+	if (sync_count >= lc->region_count)
+		lc->in_sync_hint = lc->region_count;
+
+	return (region_t)sync_count;
+}
+
+/*
+ * userspace_status
+ *
+ * Returns: amount of space consumed
+ */
+static int userspace_status(struct dm_dirty_log *log, status_type_t status_type,
+			    char *result, unsigned maxlen)
+{
+	int r = 0;
+	char *table_args;
+	size_t sz = (size_t)maxlen;
+	struct log_c *lc = log->context;
+
+	switch (status_type) {
+	case STATUSTYPE_INFO:
+		r = userspace_do_request(lc, lc->uuid, DM_ULOG_STATUS_INFO,
+					 NULL, 0, result, &sz);
+
+		if (r) {
+			sz = 0;
+			DMEMIT("%s 1 COM_FAILURE", log->type->name);
+		}
+		break;
+	case STATUSTYPE_TABLE:
+		sz = 0;
+		table_args = strchr(lc->usr_argv_str, ' ');
+		BUG_ON(!table_args); /* There will always be a ' ' */
+		table_args++;
+
+		DMEMIT("%s %u %s ", log->type->name, lc->usr_argc, lc->uuid);
+		if (lc->integrated_flush)
+			DMEMIT("integrated_flush ");
+		DMEMIT("%s ", table_args);
+		break;
+	}
+	return (r) ? 0 : (int)sz;
+}
+
+/*
+ * userspace_is_remote_recovering
+ *
+ * Returns: 1 if region recovering, 0 otherwise
+ */
+static int userspace_is_remote_recovering(struct dm_dirty_log *log,
+					  region_t region)
+{
+	int r;
+	uint64_t region64 = region;
+	struct log_c *lc = log->context;
+	static unsigned long long limit;
+	struct {
+		int64_t is_recovering;
+		uint64_t in_sync_hint;
+	} pkg;
+	size_t rdata_size = sizeof(pkg);
+
+	/*
+	 * Once the mirror has been reported to be in-sync,
+	 * it will never again ask for recovery work.  So,
+	 * we can safely say there is not a remote machine
+	 * recovering if the device is in-sync.  (in_sync_hint
+	 * must be reset at resume time.)
+	 */
+	if (region < lc->in_sync_hint)
+		return 0;
+	else if (jiffies < limit)
+		return 1;
+
+	limit = jiffies + (HZ / 4);
+	r = userspace_do_request(lc, lc->uuid, DM_ULOG_IS_REMOTE_RECOVERING,
+				 (char *)&region64, sizeof(region64),
+				 (char *)&pkg, &rdata_size);
+	if (r)
+		return 1;
+
+	lc->in_sync_hint = pkg.in_sync_hint;
+
+	return (int)pkg.is_recovering;
+}
+
+static struct dm_dirty_log_type _userspace_type = {
+	.name = "userspace",
+	.module = THIS_MODULE,
+	.ctr = userspace_ctr,
+	.dtr = userspace_dtr,
+	.presuspend = userspace_presuspend,
+	.postsuspend = userspace_postsuspend,
+	.resume = userspace_resume,
+	.get_region_size = userspace_get_region_size,
+	.is_clean = userspace_is_clean,
+	.in_sync = userspace_in_sync,
+	.flush = userspace_flush,
+	.mark_region = userspace_mark_region,
+	.clear_region = userspace_clear_region,
+	.get_resync_work = userspace_get_resync_work,
+	.set_region_sync = userspace_set_region_sync,
+	.get_sync_count = userspace_get_sync_count,
+	.status = userspace_status,
+	.is_remote_recovering = userspace_is_remote_recovering,
+};
+
+static int __init userspace_dirty_log_init(void)
+{
+	int r = 0;
+
+	flush_entry_pool = mempool_create(100, flush_entry_alloc,
+					  flush_entry_free, NULL);
+
+	if (!flush_entry_pool) {
+		DMWARN("Unable to create flush_entry_pool:  No memory.");
+		return -ENOMEM;
+	}
+
+	r = dm_ulog_tfr_init();
+	if (r) {
+		DMWARN("Unable to initialize userspace log communications");
+		mempool_destroy(flush_entry_pool);
+		return r;
+	}
+
+	r = dm_dirty_log_type_register(&_userspace_type);
+	if (r) {
+		DMWARN("Couldn't register userspace dirty log type");
+		dm_ulog_tfr_exit();
+		mempool_destroy(flush_entry_pool);
+		return r;
+	}
+
+	DMINFO("version " DM_LOG_USERSPACE_VSN " loaded");
+	return 0;
+}
+
+static void __exit userspace_dirty_log_exit(void)
+{
+	dm_dirty_log_type_unregister(&_userspace_type);
+	dm_ulog_tfr_exit();
+	mempool_destroy(flush_entry_pool);
+
+	DMINFO("version " DM_LOG_USERSPACE_VSN " unloaded");
+	return;
+}
+
+module_init(userspace_dirty_log_init);
+module_exit(userspace_dirty_log_exit);
+
+MODULE_DESCRIPTION(DM_NAME " userspace dirty log link");
+MODULE_AUTHOR("Jonathan Brassow <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-log-userspace-transfer.c b/drivers/md/dm-log-userspace-transfer.c
new file mode 100644
index 00000000000..b428c0ae63d
--- /dev/null
+++ b/drivers/md/dm-log-userspace-transfer.c
@@ -0,0 +1,286 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#include <linux/kernel.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <net/sock.h>
+#include <linux/workqueue.h>
+#include <linux/connector.h>
+#include <linux/device-mapper.h>
+#include <linux/dm-log-userspace.h>
+
+#include "dm-log-userspace-transfer.h"
+
+static uint32_t dm_ulog_seq;
+
+/*
+ * Netlink/Connector is an unreliable protocol.  How long should
+ * we wait for a response before assuming it was lost and retrying?
+ * (If we do receive a response after this time, it will be discarded
+ * and the response to the resent request will be waited for.
+ */
+#define DM_ULOG_RETRY_TIMEOUT (15 * HZ)
+
+/*
+ * Pre-allocated space for speed
+ */
+#define DM_ULOG_PREALLOCED_SIZE 512
+static struct cn_msg *prealloced_cn_msg;
+static struct dm_ulog_request *prealloced_ulog_tfr;
+
+static struct cb_id ulog_cn_id = {
+	.idx = CN_IDX_DM,
+	.val = CN_VAL_DM_USERSPACE_LOG
+};
+
+static DEFINE_MUTEX(dm_ulog_lock);
+
+struct receiving_pkg {
+	struct list_head list;
+	struct completion complete;
+
+	uint32_t seq;
+
+	int error;
+	size_t *data_size;
+	char *data;
+};
+
+static DEFINE_SPINLOCK(receiving_list_lock);
+static struct list_head receiving_list;
+
+static int dm_ulog_sendto_server(struct dm_ulog_request *tfr)
+{
+	int r;
+	struct cn_msg *msg = prealloced_cn_msg;
+
+	memset(msg, 0, sizeof(struct cn_msg));
+
+	msg->id.idx = ulog_cn_id.idx;
+	msg->id.val = ulog_cn_id.val;
+	msg->ack = 0;
+	msg->seq = tfr->seq;
+	msg->len = sizeof(struct dm_ulog_request) + tfr->data_size;
+
+	r = cn_netlink_send(msg, 0, 0, gfp_any());
+
+	return r;
+}
+
+/*
+ * Parameters for this function can be either msg or tfr, but not
+ * both.  This function fills in the reply for a waiting request.
+ * If just msg is given, then the reply is simply an ACK from userspace
+ * that the request was received.
+ *
+ * Returns: 0 on success, -ENOENT on failure
+ */
+static int fill_pkg(struct cn_msg *msg, struct dm_ulog_request *tfr)
+{
+	uint32_t rtn_seq = (msg) ? msg->seq : (tfr) ? tfr->seq : 0;
+	struct receiving_pkg *pkg;
+
+	/*
+	 * The 'receiving_pkg' entries in this list are statically
+	 * allocated on the stack in 'dm_consult_userspace'.
+	 * Each process that is waiting for a reply from the user
+	 * space server will have an entry in this list.
+	 *
+	 * We are safe to do it this way because the stack space
+	 * is unique to each process, but still addressable by
+	 * other processes.
+	 */
+	list_for_each_entry(pkg, &receiving_list, list) {
+		if (rtn_seq != pkg->seq)
+			continue;
+
+		if (msg) {
+			pkg->error = -msg->ack;
+			/*
+			 * If we are trying again, we will need to know our
+			 * storage capacity.  Otherwise, along with the
+			 * error code, we make explicit that we have no data.
+			 */
+			if (pkg->error != -EAGAIN)
+				*(pkg->data_size) = 0;
+		} else if (tfr->data_size > *(pkg->data_size)) {
+			DMERR("Insufficient space to receive package [%u] "
+			      "(%u vs %zu)", tfr->request_type,
+			      tfr->data_size, *(pkg->data_size));
+
+			*(pkg->data_size) = 0;
+			pkg->error = -ENOSPC;
+		} else {
+			pkg->error = tfr->error;
+			memcpy(pkg->data, tfr->data, tfr->data_size);
+			*(pkg->data_size) = tfr->data_size;
+		}
+		complete(&pkg->complete);
+		return 0;
+	}
+
+	return -ENOENT;
+}
+
+/*
+ * This is the connector callback that delivers data
+ * that was sent from userspace.
+ */
+static void cn_ulog_callback(struct cn_msg *msg, struct netlink_skb_parms *nsp)
+{
+	struct dm_ulog_request *tfr = (struct dm_ulog_request *)(msg + 1);
+
+	if (!capable(CAP_SYS_ADMIN))
+		return;
+
+	spin_lock(&receiving_list_lock);
+	if (msg->len == 0)
+		fill_pkg(msg, NULL);
+	else if (msg->len < sizeof(*tfr))
+		DMERR("Incomplete message received (expected %u, got %u): [%u]",
+		      (unsigned)sizeof(*tfr), msg->len, msg->seq);
+	else
+		fill_pkg(NULL, tfr);
+	spin_unlock(&receiving_list_lock);
+}
+
+/**
+ * dm_consult_userspace
+ * @uuid: log's universal unique identifier (must be DM_UUID_LEN in size)
+ * @luid: log's local unique identifier
+ * @request_type:  found in include/linux/dm-log-userspace.h
+ * @data: data to tx to the server
+ * @data_size: size of data in bytes
+ * @rdata: place to put return data from server
+ * @rdata_size: value-result (amount of space given/amount of space used)
+ *
+ * rdata_size is undefined on failure.
+ *
+ * Memory used to communicate with userspace is zero'ed
+ * before populating to ensure that no unwanted bits leak
+ * from kernel space to user-space.  All userspace log communications
+ * between kernel and user space go through this function.
+ *
+ * Returns: 0 on success, -EXXX on failure
+ **/
+int dm_consult_userspace(const char *uuid, uint64_t luid, int request_type,
+			 char *data, size_t data_size,
+			 char *rdata, size_t *rdata_size)
+{
+	int r = 0;
+	size_t dummy = 0;
+	int overhead_size = sizeof(struct dm_ulog_request) + sizeof(struct cn_msg);
+	struct dm_ulog_request *tfr = prealloced_ulog_tfr;
+	struct receiving_pkg pkg;
+
+	/*
+	 * Given the space needed to hold the 'struct cn_msg' and
+	 * 'struct dm_ulog_request' - do we have enough payload
+	 * space remaining?
+	 */
+	if (data_size > (DM_ULOG_PREALLOCED_SIZE - overhead_size)) {
+		DMINFO("Size of tfr exceeds preallocated size");
+		return -EINVAL;
+	}
+
+	if (!rdata_size)
+		rdata_size = &dummy;
+resend:
+	/*
+	 * We serialize the sending of requests so we can
+	 * use the preallocated space.
+	 */
+	mutex_lock(&dm_ulog_lock);
+
+	memset(tfr, 0, DM_ULOG_PREALLOCED_SIZE - sizeof(struct cn_msg));
+	memcpy(tfr->uuid, uuid, DM_UUID_LEN);
+	tfr->version = DM_ULOG_REQUEST_VERSION;
+	tfr->luid = luid;
+	tfr->seq = dm_ulog_seq++;
+
+	/*
+	 * Must be valid request type (all other bits set to
+	 * zero).  This reserves other bits for possible future
+	 * use.
+	 */
+	tfr->request_type = request_type & DM_ULOG_REQUEST_MASK;
+
+	tfr->data_size = data_size;
+	if (data && data_size)
+		memcpy(tfr->data, data, data_size);
+
+	memset(&pkg, 0, sizeof(pkg));
+	init_completion(&pkg.complete);
+	pkg.seq = tfr->seq;
+	pkg.data_size = rdata_size;
+	pkg.data = rdata;
+	spin_lock(&receiving_list_lock);
+	list_add(&(pkg.list), &receiving_list);
+	spin_unlock(&receiving_list_lock);
+
+	r = dm_ulog_sendto_server(tfr);
+
+	mutex_unlock(&dm_ulog_lock);
+
+	if (r) {
+		DMERR("Unable to send log request [%u] to userspace: %d",
+		      request_type, r);
+		spin_lock(&receiving_list_lock);
+		list_del_init(&(pkg.list));
+		spin_unlock(&receiving_list_lock);
+
+		goto out;
+	}
+
+	r = wait_for_completion_timeout(&(pkg.complete), DM_ULOG_RETRY_TIMEOUT);
+	spin_lock(&receiving_list_lock);
+	list_del_init(&(pkg.list));
+	spin_unlock(&receiving_list_lock);
+	if (!r) {
+		DMWARN("[%s] Request timed out: [%u/%u] - retrying",
+		       (strlen(uuid) > 8) ?
+		       (uuid + (strlen(uuid) - 8)) : (uuid),
+		       request_type, pkg.seq);
+		goto resend;
+	}
+
+	r = pkg.error;
+	if (r == -EAGAIN)
+		goto resend;
+
+out:
+	return r;
+}
+
+int dm_ulog_tfr_init(void)
+{
+	int r;
+	void *prealloced;
+
+	INIT_LIST_HEAD(&receiving_list);
+
+	prealloced = kmalloc(DM_ULOG_PREALLOCED_SIZE, GFP_KERNEL);
+	if (!prealloced)
+		return -ENOMEM;
+
+	prealloced_cn_msg = prealloced;
+	prealloced_ulog_tfr = prealloced + sizeof(struct cn_msg);
+
+	r = cn_add_callback(&ulog_cn_id, "dmlogusr", cn_ulog_callback);
+	if (r) {
+		cn_del_callback(&ulog_cn_id);
+		return r;
+	}
+
+	return 0;
+}
+
+void dm_ulog_tfr_exit(void)
+{
+	cn_del_callback(&ulog_cn_id);
+	kfree(prealloced_cn_msg);
+}
diff --git a/drivers/md/dm-log-userspace-transfer.h b/drivers/md/dm-log-userspace-transfer.h
new file mode 100644
index 00000000000..04ee874f915
--- /dev/null
+++ b/drivers/md/dm-log-userspace-transfer.h
@@ -0,0 +1,18 @@
+/*
+ * Copyright (C) 2006-2009 Red Hat, Inc.
+ *
+ * This file is released under the LGPL.
+ */
+
+#ifndef __DM_LOG_USERSPACE_TRANSFER_H__
+#define __DM_LOG_USERSPACE_TRANSFER_H__
+
+#define DM_MSG_PREFIX "dm-log-userspace"
+
+int dm_ulog_tfr_init(void);
+void dm_ulog_tfr_exit(void);
+int dm_consult_userspace(const char *uuid, uint64_t luid, int request_type,
+			 char *data, size_t data_size,
+			 char *rdata, size_t *rdata_size);
+
+#endif /* __DM_LOG_USERSPACE_TRANSFER_H__ */
diff --git a/drivers/md/dm-log.c b/drivers/md/dm-log.c
index 5b48478c79f..627d19186d5 100644
--- a/drivers/md/dm-log.c
+++ b/drivers/md/dm-log.c
@@ -12,44 +12,33 @@
 #include <linux/dm-io.h>
 #include <linux/dm-dirty-log.h>
 
-#include "dm.h"
+#include <linux/device-mapper.h>
 
 #define DM_MSG_PREFIX "dirty region log"
 
-struct dm_dirty_log_internal {
-	struct dm_dirty_log_type *type;
-
-	struct list_head list;
-	long use;
-};
-
 static LIST_HEAD(_log_types);
 static DEFINE_SPINLOCK(_lock);
 
-static struct dm_dirty_log_internal *__find_dirty_log_type(const char *name)
+static struct dm_dirty_log_type *__find_dirty_log_type(const char *name)
 {
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	list_for_each_entry(log_type, &_log_types, list)
-		if (!strcmp(name, log_type->type->name))
+		if (!strcmp(name, log_type->name))
 			return log_type;
 
 	return NULL;
 }
 
-static struct dm_dirty_log_internal *_get_dirty_log_type(const char *name)
+static struct dm_dirty_log_type *_get_dirty_log_type(const char *name)
 {
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	spin_lock(&_lock);
 
 	log_type = __find_dirty_log_type(name);
-	if (log_type) {
-		if (!log_type->use && !try_module_get(log_type->type->module))
-			log_type = NULL;
-		else
-			log_type->use++;
-	}
+	if (log_type && !try_module_get(log_type->module))
+		log_type = NULL;
 
 	spin_unlock(&_lock);
 
@@ -76,14 +65,14 @@ static struct dm_dirty_log_internal *_get_dirty_log_type(const char *name)
 static struct dm_dirty_log_type *get_type(const char *type_name)
 {
 	char *p, *type_name_dup;
-	struct dm_dirty_log_internal *log_type;
+	struct dm_dirty_log_type *log_type;
 
 	if (!type_name)
 		return NULL;
 
 	log_type = _get_dirty_log_type(type_name);
 	if (log_type)
-		return log_type->type;
+		return log_type;
 
 	type_name_dup = kstrdup(type_name, GFP_KERNEL);
 	if (!type_name_dup) {
@@ -105,56 +94,33 @@ static struct dm_dirty_log_type *get_type(const char *type_name)
 
 	kfree(type_name_dup);
 
-	return log_type ? log_type->type : NULL;
+	return log_type;
 }
 
 static void put_type(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type;
-
 	if (!type)
 		return;
 
 	spin_lock(&_lock);
-	log_type = __find_dirty_log_type(type->name);
-	if (!log_type)
+	if (!__find_dirty_log_type(type->name))
 		goto out;
 
-	if (!--log_type->use)
-		module_put(type->module);
-
-	BUG_ON(log_type->use < 0);
+	module_put(type->module);
 
 out:
 	spin_unlock(&_lock);
 }
 
-static struct dm_dirty_log_internal *_alloc_dirty_log_type(struct dm_dirty_log_type *type)
-{
-	struct dm_dirty_log_internal *log_type = kzalloc(sizeof(*log_type),
-							 GFP_KERNEL);
-
-	if (log_type)
-		log_type->type = type;
-
-	return log_type;
-}
-
 int dm_dirty_log_type_register(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type = _alloc_dirty_log_type(type);
 	int r = 0;
 
-	if (!log_type)
-		return -ENOMEM;
-
 	spin_lock(&_lock);
 	if (!__find_dirty_log_type(type->name))
-		list_add(&log_type->list, &_log_types);
-	else {
-		kfree(log_type);
+		list_add(&type->list, &_log_types);
+	else
 		r = -EEXIST;
-	}
 	spin_unlock(&_lock);
 
 	return r;
@@ -163,33 +129,25 @@ EXPORT_SYMBOL(dm_dirty_log_type_register);
 
 int dm_dirty_log_type_unregister(struct dm_dirty_log_type *type)
 {
-	struct dm_dirty_log_internal *log_type;
-
 	spin_lock(&_lock);
 
-	log_type = __find_dirty_log_type(type->name);
-	if (!log_type) {
+	if (!__find_dirty_log_type(type->name)) {
 		spin_unlock(&_lock);
 		return -EINVAL;
 	}
 
-	if (log_type->use) {
-		spin_unlock(&_lock);
-		return -ETXTBSY;
-	}
-
-	list_del(&log_type->list);
+	list_del(&type->list);
 
 	spin_unlock(&_lock);
-	kfree(log_type);
 
 	return 0;
 }
 EXPORT_SYMBOL(dm_dirty_log_type_unregister);
 
 struct dm_dirty_log *dm_dirty_log_create(const char *type_name,
-					 struct dm_target *ti,
-					 unsigned int argc, char **argv)
+			struct dm_target *ti,
+			int (*flush_callback_fn)(struct dm_target *ti),
+			unsigned int argc, char **argv)
 {
 	struct dm_dirty_log_type *type;
 	struct dm_dirty_log *log;
@@ -204,6 +162,7 @@ struct dm_dirty_log *dm_dirty_log_create(const char *type_name,
 		return NULL;
 	}
 
+	log->flush_callback_fn = flush_callback_fn;
 	log->type = type;
 	if (type->ctr(log, ti, argc, argv)) {
 		kfree(log);
@@ -238,20 +197,28 @@ EXPORT_SYMBOL(dm_dirty_log_destroy);
 #define MIRROR_DISK_VERSION 2
 #define LOG_OFFSET 2
 
-struct log_header {
-	uint32_t magic;
+struct log_header_disk {
+	__le32 magic;
 
 	/*
 	 * Simple, incrementing version. no backward
 	 * compatibility.
 	 */
+	__le32 version;
+	__le64 nr_regions;
+} __packed;
+
+struct log_header_core {
+	uint32_t magic;
 	uint32_t version;
-	sector_t nr_regions;
+	uint64_t nr_regions;
 };
 
 struct log_c {
 	struct dm_target *ti;
-	int touched;
+	int touched_dirtied;
+	int touched_cleaned;
+	int flush_failed;
 	uint32_t region_size;
 	unsigned int region_count;
 	region_t sync_count;
@@ -276,11 +243,12 @@ struct log_c {
 	 * Disk log fields
 	 */
 	int log_dev_failed;
+	int log_dev_flush_failed;
 	struct dm_dev *log_dev;
-	struct log_header header;
+	struct log_header_core header;
 
 	struct dm_io_region header_location;
-	struct log_header *disk_header;
+	struct log_header_disk *disk_header;
 };
 
 /*
@@ -289,34 +257,34 @@ struct log_c {
  */
 static inline int log_test_bit(uint32_t *bs, unsigned bit)
 {
-	return ext2_test_bit(bit, (unsigned long *) bs) ? 1 : 0;
+	return test_bit_le(bit, bs) ? 1 : 0;
 }
 
 static inline void log_set_bit(struct log_c *l,
 			       uint32_t *bs, unsigned bit)
 {
-	ext2_set_bit(bit, (unsigned long *) bs);
-	l->touched = 1;
+	__set_bit_le(bit, bs);
+	l->touched_cleaned = 1;
 }
 
 static inline void log_clear_bit(struct log_c *l,
 				 uint32_t *bs, unsigned bit)
 {
-	ext2_clear_bit(bit, (unsigned long *) bs);
-	l->touched = 1;
+	__clear_bit_le(bit, bs);
+	l->touched_dirtied = 1;
 }
 
 /*----------------------------------------------------------------
  * Header IO
  *--------------------------------------------------------------*/
-static void header_to_disk(struct log_header *core, struct log_header *disk)
+static void header_to_disk(struct log_header_core *core, struct log_header_disk *disk)
 {
 	disk->magic = cpu_to_le32(core->magic);
 	disk->version = cpu_to_le32(core->version);
 	disk->nr_regions = cpu_to_le64(core->nr_regions);
 }
 
-static void header_from_disk(struct log_header *core, struct log_header *disk)
+static void header_from_disk(struct log_header_core *core, struct log_header_disk *disk)
 {
 	core->magic = le32_to_cpu(disk->magic);
 	core->version = le32_to_cpu(disk->version);
@@ -326,12 +294,23 @@ static void header_from_disk(struct log_header *core, struct log_header *disk)
 static int rw_header(struct log_c *lc, int rw)
 {
 	lc->io_req.bi_rw = rw;
-	lc->io_req.mem.ptr.vma = lc->disk_header;
-	lc->io_req.notify.fn = NULL;
 
 	return dm_io(&lc->io_req, 1, &lc->header_location, NULL);
 }
 
+static int flush_header(struct log_c *lc)
+{
+	struct dm_io_region null_location = {
+		.bdev = lc->header_location.bdev,
+		.sector = 0,
+		.count = 0,
+	};
+
+	lc->io_req.bi_rw = WRITE_FLUSH;
+
+	return dm_io(&lc->io_req, 1, &null_location, NULL);
+}
+
 static int read_header(struct log_c *log)
 {
 	int r;
@@ -362,10 +341,15 @@ static int read_header(struct log_c *log)
 	return 0;
 }
 
-static inline int write_header(struct log_c *log)
+static int _check_region_size(struct dm_target *ti, uint32_t region_size)
 {
-	header_to_disk(&log->header, log->disk_header);
-	return rw_header(log, WRITE);
+	if (region_size < 2 || region_size > ti->len)
+		return 0;
+
+	if (!is_power_of_2(region_size))
+		return 0;
+
+	return 1;
 }
 
 /*----------------------------------------------------------------
@@ -385,6 +369,7 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	unsigned int region_count;
 	size_t bitset_size, buf_size;
 	int r;
+	char dummy;
 
 	if (argc < 1 || argc > 2) {
 		DMWARN("wrong number of arguments to dirty region log");
@@ -403,8 +388,9 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		}
 	}
 
-	if (sscanf(argv[0], "%u", &region_size) != 1) {
-		DMWARN("invalid region size string");
+	if (sscanf(argv[0], "%u%c", &region_size, &dummy) != 1 ||
+	    !_check_region_size(ti, region_size)) {
+		DMWARN("invalid region size %s", argv[0]);
 		return -EINVAL;
 	}
 
@@ -417,7 +403,9 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	}
 
 	lc->ti = ti;
-	lc->touched = 0;
+	lc->touched_dirtied = 0;
+	lc->touched_cleaned = 0;
+	lc->flush_failed = 0;
 	lc->region_size = region_size;
 	lc->region_count = region_count;
 	lc->sync = sync;
@@ -445,32 +433,46 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	} else {
 		lc->log_dev = dev;
 		lc->log_dev_failed = 0;
+		lc->log_dev_flush_failed = 0;
 		lc->header_location.bdev = lc->log_dev->bdev;
 		lc->header_location.sector = 0;
 
 		/*
 		 * Buffer holds both header and bitset.
 		 */
-		buf_size = dm_round_up((LOG_OFFSET << SECTOR_SHIFT) +
-				       bitset_size, ti->limits.hardsect_size);
+		buf_size =
+		    dm_round_up((LOG_OFFSET << SECTOR_SHIFT) + bitset_size,
+				bdev_logical_block_size(lc->header_location.
+							    bdev));
+
+		if (buf_size > i_size_read(dev->bdev->bd_inode)) {
+			DMWARN("log device %s too small: need %llu bytes",
+				dev->name, (unsigned long long)buf_size);
+			kfree(lc);
+			return -EINVAL;
+		}
+
 		lc->header_location.count = buf_size >> SECTOR_SHIFT;
+
 		lc->io_req.mem.type = DM_IO_VMA;
-		lc->io_req.client = dm_io_client_create(dm_div_up(buf_size,
-								   PAGE_SIZE));
+		lc->io_req.notify.fn = NULL;
+		lc->io_req.client = dm_io_client_create();
 		if (IS_ERR(lc->io_req.client)) {
 			r = PTR_ERR(lc->io_req.client);
 			DMWARN("couldn't allocate disk io client");
 			kfree(lc);
-			return -ENOMEM;
+			return r;
 		}
 
 		lc->disk_header = vmalloc(buf_size);
 		if (!lc->disk_header) {
 			DMWARN("couldn't allocate disk log buffer");
+			dm_io_client_destroy(lc->io_req.client);
 			kfree(lc);
 			return -ENOMEM;
 		}
 
+		lc->io_req.mem.ptr.vma = lc->disk_header;
 		lc->clean_bits = (void *)lc->disk_header +
 				 (LOG_OFFSET << SECTOR_SHIFT);
 	}
@@ -482,6 +484,8 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 		DMWARN("couldn't allocate sync bitset");
 		if (!dev)
 			vfree(lc->clean_bits);
+		else
+			dm_io_client_destroy(lc->io_req.client);
 		vfree(lc->disk_header);
 		kfree(lc);
 		return -ENOMEM;
@@ -489,17 +493,18 @@ static int create_log_context(struct dm_dirty_log *log, struct dm_target *ti,
 	memset(lc->sync_bits, (sync == NOSYNC) ? -1 : 0, bitset_size);
 	lc->sync_count = (sync == NOSYNC) ? region_count : 0;
 
-	lc->recovering_bits = vmalloc(bitset_size);
+	lc->recovering_bits = vzalloc(bitset_size);
 	if (!lc->recovering_bits) {
 		DMWARN("couldn't allocate sync bitset");
 		vfree(lc->sync_bits);
 		if (!dev)
 			vfree(lc->clean_bits);
+		else
+			dm_io_client_destroy(lc->io_req.client);
 		vfree(lc->disk_header);
 		kfree(lc);
 		return -ENOMEM;
 	}
-	memset(lc->recovering_bits, 0, bitset_size);
 	lc->sync_search = 0;
 	log->context = lc;
 
@@ -543,8 +548,7 @@ static int disk_ctr(struct dm_dirty_log *log, struct dm_target *ti,
 		return -EINVAL;
 	}
 
-	r = dm_get_device(ti, argv[0], 0, 0 /* FIXME */,
-			  FMODE_READ | FMODE_WRITE, &dev);
+	r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &dev);
 	if (r)
 		return r;
 
@@ -567,16 +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)
@@ -625,14 +619,22 @@ 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 */
 	lc->header.nr_regions = lc->region_count;
 
+	header_to_disk(&lc->header, lc->disk_header);
+
 	/* write the new header */
-	r = write_header(lc);
+	r = rw_header(lc, WRITE);
+	if (!r) {
+		r = flush_header(lc);
+		if (r)
+			lc->log_dev_flush_failed = 1;
+	}
 	if (r) {
 		DMWARN("%s: Failed to write header on dirty region log device",
 		       lc->log_dev->name);
@@ -675,18 +677,40 @@ static int core_flush(struct dm_dirty_log *log)
 
 static int disk_flush(struct dm_dirty_log *log)
 {
-	int r;
-	struct log_c *lc = (struct log_c *) log->context;
+	int r, i;
+	struct log_c *lc = log->context;
 
 	/* only write if the log has changed */
-	if (!lc->touched)
+	if (!lc->touched_cleaned && !lc->touched_dirtied)
 		return 0;
 
-	r = write_header(lc);
+	if (lc->touched_cleaned && log->flush_callback_fn &&
+	    log->flush_callback_fn(lc->ti)) {
+		/*
+		 * At this point it is impossible to determine which
+		 * regions are clean and which are dirty (without
+		 * re-reading the log off disk). So mark all of them
+		 * dirty.
+		 */
+		lc->flush_failed = 1;
+		for (i = 0; i < lc->region_count; i++)
+			log_clear_bit(lc, lc->clean_bits, i);
+	}
+
+	r = rw_header(lc, WRITE);
 	if (r)
 		fail_log_device(lc);
-	else
-		lc->touched = 0;
+	else {
+		if (lc->touched_dirtied) {
+			r = flush_header(lc);
+			if (r) {
+				lc->log_dev_flush_failed = 1;
+				fail_log_device(lc);
+			} else
+				lc->touched_dirtied = 0;
+		}
+		lc->touched_cleaned = 0;
+	}
 
 	return r;
 }
@@ -700,7 +724,8 @@ static void core_mark_region(struct dm_dirty_log *log, region_t region)
 static void core_clear_region(struct dm_dirty_log *log, region_t region)
 {
 	struct log_c *lc = (struct log_c *) log->context;
-	log_set_bit(lc, lc->clean_bits, region);
+	if (likely(!lc->flush_failed))
+		log_set_bit(lc, lc->clean_bits, region);
 }
 
 static int core_get_resync_work(struct dm_dirty_log *log, region_t *region)
@@ -711,8 +736,7 @@ static int core_get_resync_work(struct dm_dirty_log *log, region_t *region)
 		return 0;
 
 	do {
-		*region = ext2_find_next_zero_bit(
-					     (unsigned long *) lc->sync_bits,
+		*region = find_next_zero_bit_le(lc->sync_bits,
 					     lc->region_count,
 					     lc->sync_search);
 		lc->sync_search = *region + 1;
@@ -781,7 +805,9 @@ static int disk_status(struct dm_dirty_log *log, status_type_t status,
 	switch(status) {
 	case STATUSTYPE_INFO:
 		DMEMIT("3 %s %s %c", log->type->name, lc->log_dev->name,
-		       lc->log_dev_failed ? 'D' : 'A');
+		       lc->log_dev_flush_failed ? 'F' :
+		       lc->log_dev_failed ? 'D' :
+		       'A');
 		break;
 
 	case STATUSTYPE_TABLE:
diff --git a/drivers/md/dm-mpath.c b/drivers/md/dm-mpath.c
index 71dd65aa31b..f4167b013d9 100644
--- a/drivers/md/dm-mpath.c
+++ b/drivers/md/dm-mpath.c
@@ -5,10 +5,10 @@
  * This file is released under the GPL.
  */
 
+#include <linux/device-mapper.h>
+
 #include "dm.h"
 #include "dm-path-selector.h"
-#include "dm-bio-list.h"
-#include "dm-bio-record.h"
 #include "dm-uevent.h"
 
 #include <linux/ctype.h>
@@ -19,20 +19,24 @@
 #include <linux/slab.h>
 #include <linux/time.h>
 #include <linux/workqueue.h>
+#include <linux/delay.h>
 #include <scsi/scsi_dh.h>
-#include <asm/atomic.h>
+#include <linux/atomic.h>
 
 #define DM_MSG_PREFIX "multipath"
-#define MESG_STR(x) x, sizeof(x)
+#define DM_PG_INIT_DELAY_MSECS 2000
+#define DM_PG_INIT_DELAY_DEFAULT ((unsigned) -1)
 
 /* Path properties */
 struct pgpath {
 	struct list_head list;
 
 	struct priority_group *pg;	/* Owning PG */
+	unsigned is_active;		/* Path status */
 	unsigned fail_count;		/* Cumulative failure count */
 
 	struct dm_path path;
+	struct delayed_work activate_path;
 };
 
 #define path_to_pgpath(__pgp) container_of((__pgp), struct pgpath, path)
@@ -59,14 +63,19 @@ struct multipath {
 	struct list_head list;
 	struct dm_target *ti;
 
+	const char *hw_handler_name;
+	char *hw_handler_params;
+
 	spinlock_t lock;
 
-	const char *hw_handler_name;
-	struct work_struct activate_path;
 	unsigned nr_priority_groups;
 	struct list_head priority_groups;
+
+	wait_queue_head_t pg_init_wait;	/* Wait for pg_init completion */
+
 	unsigned pg_init_required;	/* pg_init needs calling? */
 	unsigned pg_init_in_progress;	/* Only one pg_init allowed at once */
+	unsigned pg_init_delay_retry;	/* Delay pg_init retry? */
 
 	unsigned nr_valid_paths;	/* Total number of usable paths */
 	struct pgpath *current_pgpath;
@@ -74,15 +83,15 @@ struct multipath {
 	struct priority_group *next_pg;	/* Switch to this PG if set */
 	unsigned repeat_count;		/* I/Os left before calling PS again */
 
-	unsigned queue_io;		/* Must we queue all I/O? */
-	unsigned queue_if_no_path;	/* Queue I/O if last path fails? */
-	unsigned saved_queue_if_no_path;/* Saved state during suspension */
+	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_disabled:1;	/* pg_init is not currently allowed */
+
 	unsigned pg_init_retries;	/* Number of times to retry pg_init */
 	unsigned pg_init_count;		/* Number of times pg_init called */
-
-	struct work_struct process_queued_ios;
-	struct bio_list queued_ios;
-	unsigned queue_size;
+	unsigned pg_init_delay_msecs;	/* Number of msecs before pg_init retry */
 
 	struct work_struct trigger_event;
 
@@ -91,6 +100,8 @@ struct multipath {
 	 * can resubmit bios on error.
 	 */
 	mempool_t *mpio_pool;
+
+	struct mutex work_mutex;
 };
 
 /*
@@ -98,19 +109,17 @@ struct multipath {
  */
 struct dm_mpath_io {
 	struct pgpath *pgpath;
-	struct dm_bio_details details;
+	size_t nr_bytes;
 };
 
 typedef int (*action_fn) (struct pgpath *pgpath);
 
-#define MIN_IOS 256	/* Mempool size */
-
 static struct kmem_cache *_mpio_cache;
 
 static struct workqueue_struct *kmultipathd, *kmpath_handlerd;
-static void process_queued_ios(struct work_struct *work);
 static void trigger_event(struct work_struct *work);
 static void activate_path(struct work_struct *work);
+static int __pgpath_busy(struct pgpath *pgpath);
 
 
 /*-----------------------------------------------
@@ -121,8 +130,10 @@ static struct pgpath *alloc_pgpath(void)
 {
 	struct pgpath *pgpath = kzalloc(sizeof(*pgpath), GFP_KERNEL);
 
-	if (pgpath)
-		pgpath->path.is_active = 1;
+	if (pgpath) {
+		pgpath->is_active = 1;
+		INIT_DELAYED_WORK(&pgpath->activate_path, activate_path);
+	}
 
 	return pgpath;
 }
@@ -175,16 +186,18 @@ static void free_priority_group(struct priority_group *pg,
 static struct multipath *alloc_multipath(struct dm_target *ti)
 {
 	struct multipath *m;
+	unsigned min_ios = dm_get_reserved_rq_based_ios();
 
 	m = kzalloc(sizeof(*m), GFP_KERNEL);
 	if (m) {
 		INIT_LIST_HEAD(&m->priority_groups);
 		spin_lock_init(&m->lock);
 		m->queue_io = 1;
-		INIT_WORK(&m->process_queued_ios, process_queued_ios);
+		m->pg_init_delay_msecs = DM_PG_INIT_DELAY_DEFAULT;
 		INIT_WORK(&m->trigger_event, trigger_event);
-		INIT_WORK(&m->activate_path, activate_path);
-		m->mpio_pool = mempool_create_slab_pool(MIN_IOS, _mpio_cache);
+		init_waitqueue_head(&m->pg_init_wait);
+		mutex_init(&m->work_mutex);
+		m->mpio_pool = mempool_create_slab_pool(min_ios, _mpio_cache);
 		if (!m->mpio_pool) {
 			kfree(m);
 			return NULL;
@@ -206,15 +219,66 @@ static void free_multipath(struct multipath *m)
 	}
 
 	kfree(m->hw_handler_name);
+	kfree(m->hw_handler_params);
 	mempool_destroy(m->mpio_pool);
 	kfree(m);
 }
 
+static int set_mapinfo(struct multipath *m, union map_info *info)
+{
+	struct dm_mpath_io *mpio;
+
+	mpio = mempool_alloc(m->mpio_pool, GFP_ATOMIC);
+	if (!mpio)
+		return -ENOMEM;
+
+	memset(mpio, 0, sizeof(*mpio));
+	info->ptr = mpio;
+
+	return 0;
+}
+
+static void clear_mapinfo(struct multipath *m, union map_info *info)
+{
+	struct dm_mpath_io *mpio = info->ptr;
+
+	info->ptr = NULL;
+	mempool_free(mpio, m->mpio_pool);
+}
 
 /*-----------------------------------------------
  * Path selection
  *-----------------------------------------------*/
 
+static int __pg_init_all_paths(struct multipath *m)
+{
+	struct pgpath *pgpath;
+	unsigned long pg_init_delay = 0;
+
+	if (m->pg_init_in_progress || m->pg_init_disabled)
+		return 0;
+
+	m->pg_init_count++;
+	m->pg_init_required = 0;
+
+	/* Check here to reset pg_init_required */
+	if (!m->current_pg)
+		return 0;
+
+	if (m->pg_init_delay_retry)
+		pg_init_delay = msecs_to_jiffies(m->pg_init_delay_msecs != DM_PG_INIT_DELAY_DEFAULT ?
+						 m->pg_init_delay_msecs : DM_PG_INIT_DELAY_MSECS);
+	list_for_each_entry(pgpath, &m->current_pg->pgpaths, list) {
+		/* Skip failed paths */
+		if (!pgpath->is_active)
+			continue;
+		if (queue_delayed_work(kmpath_handlerd, &pgpath->activate_path,
+				       pg_init_delay))
+			m->pg_init_in_progress++;
+	}
+	return m->pg_init_in_progress;
+}
+
 static void __switch_pg(struct multipath *m, struct pgpath *pgpath)
 {
 	m->current_pg = pgpath->pg;
@@ -231,11 +295,12 @@ static void __switch_pg(struct multipath *m, struct pgpath *pgpath)
 	m->pg_init_count = 0;
 }
 
-static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg)
+static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg,
+			       size_t nr_bytes)
 {
 	struct dm_path *path;
 
-	path = pg->ps.type->select_path(&pg->ps, &m->repeat_count);
+	path = pg->ps.type->select_path(&pg->ps, &m->repeat_count, nr_bytes);
 	if (!path)
 		return -ENXIO;
 
@@ -247,7 +312,7 @@ static int __choose_path_in_pg(struct multipath *m, struct priority_group *pg)
 	return 0;
 }
 
-static void __choose_pgpath(struct multipath *m)
+static void __choose_pgpath(struct multipath *m, size_t nr_bytes)
 {
 	struct priority_group *pg;
 	unsigned bypassed = 1;
@@ -259,25 +324,29 @@ static void __choose_pgpath(struct multipath *m)
 	if (m->next_pg) {
 		pg = m->next_pg;
 		m->next_pg = NULL;
-		if (!__choose_path_in_pg(m, pg))
+		if (!__choose_path_in_pg(m, pg, nr_bytes))
 			return;
 	}
 
 	/* Don't change PG until it has no remaining paths */
-	if (m->current_pg && !__choose_path_in_pg(m, m->current_pg))
+	if (m->current_pg && !__choose_path_in_pg(m, m->current_pg, nr_bytes))
 		return;
 
 	/*
 	 * Loop through priority groups until we find a valid path.
 	 * First time we skip PGs marked 'bypassed'.
-	 * Second time we only try the ones we skipped.
+	 * Second time we only try the ones we skipped, but set
+	 * pg_init_delay_retry so we do not hammer controllers.
 	 */
 	do {
 		list_for_each_entry(pg, &m->priority_groups, list) {
 			if (pg->bypassed == bypassed)
 				continue;
-			if (!__choose_path_in_pg(m, pg))
+			if (!__choose_path_in_pg(m, pg, nr_bytes)) {
+				if (!bypassed)
+					m->pg_init_delay_retry = 1;
 				return;
+			}
 		}
 	} while (bypassed--);
 
@@ -299,48 +368,63 @@ failed:
  */
 static int __must_push_back(struct multipath *m)
 {
-	return (m->queue_if_no_path != m->saved_queue_if_no_path &&
-		dm_noflush_suspending(m->ti));
+	return (m->queue_if_no_path ||
+		(m->queue_if_no_path != m->saved_queue_if_no_path &&
+		 dm_noflush_suspending(m->ti)));
 }
 
-static int map_io(struct multipath *m, struct bio *bio,
-		  struct dm_mpath_io *mpio, unsigned was_queued)
+#define pg_ready(m) (!(m)->queue_io && !(m)->pg_init_required)
+
+/*
+ * Map cloned requests
+ */
+static int multipath_map(struct dm_target *ti, struct request *clone,
+			 union map_info *map_context)
 {
-	int r = DM_MAPIO_REMAPPED;
+	struct multipath *m = (struct multipath *) ti->private;
+	int r = DM_MAPIO_REQUEUE;
+	size_t nr_bytes = blk_rq_bytes(clone);
 	unsigned long flags;
 	struct pgpath *pgpath;
+	struct block_device *bdev;
+	struct dm_mpath_io *mpio;
 
 	spin_lock_irqsave(&m->lock, flags);
 
 	/* Do we need to select a new pgpath? */
 	if (!m->current_pgpath ||
 	    (!m->queue_io && (m->repeat_count && --m->repeat_count == 0)))
-		__choose_pgpath(m);
+		__choose_pgpath(m, nr_bytes);
 
 	pgpath = m->current_pgpath;
 
-	if (was_queued)
-		m->queue_size--;
-
-	if ((pgpath && m->queue_io) ||
-	    (!pgpath && m->queue_if_no_path)) {
-		/* Queue for the daemon to resubmit */
-		bio_list_add(&m->queued_ios, bio);
-		m->queue_size++;
-		if ((m->pg_init_required && !m->pg_init_in_progress) ||
-		    !m->queue_io)
-			queue_work(kmultipathd, &m->process_queued_ios);
-		pgpath = NULL;
-		r = DM_MAPIO_SUBMITTED;
-	} else if (pgpath)
-		bio->bi_bdev = pgpath->path.dev->bdev;
-	else if (__must_push_back(m))
-		r = DM_MAPIO_REQUEUE;
-	else
-		r = -EIO;	/* Failed */
-
+	if (!pgpath) {
+		if (!__must_push_back(m))
+			r = -EIO;	/* Failed */
+		goto out_unlock;
+	}
+	if (!pg_ready(m)) {
+		__pg_init_all_paths(m);
+		goto out_unlock;
+	}
+	if (set_mapinfo(m, map_context) < 0)
+		/* ENOMEM, requeue */
+		goto out_unlock;
+
+	bdev = pgpath->path.dev->bdev;
+	clone->q = bdev_get_queue(bdev);
+	clone->rq_disk = bdev->bd_disk;
+	clone->cmd_flags |= REQ_FAILFAST_TRANSPORT;
+	mpio = map_context->ptr;
 	mpio->pgpath = pgpath;
-
+	mpio->nr_bytes = nr_bytes;
+	if (pgpath->pg->ps.type->start_io)
+		pgpath->pg->ps.type->start_io(&pgpath->pg->ps,
+					      &pgpath->path,
+					      nr_bytes);
+	r = DM_MAPIO_REMAPPED;
+
+out_unlock:
 	spin_unlock_irqrestore(&m->lock, flags);
 
 	return r;
@@ -361,86 +445,12 @@ static int queue_if_no_path(struct multipath *m, unsigned queue_if_no_path,
 	else
 		m->saved_queue_if_no_path = queue_if_no_path;
 	m->queue_if_no_path = queue_if_no_path;
-	if (!m->queue_if_no_path && m->queue_size)
-		queue_work(kmultipathd, &m->process_queued_ios);
-
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	return 0;
-}
-
-/*-----------------------------------------------------------------
- * The multipath daemon is responsible for resubmitting queued ios.
- *---------------------------------------------------------------*/
-
-static void dispatch_queued_ios(struct multipath *m)
-{
-	int r;
-	unsigned long flags;
-	struct bio *bio = NULL, *next;
-	struct dm_mpath_io *mpio;
-	union map_info *info;
-
-	spin_lock_irqsave(&m->lock, flags);
-	bio = bio_list_get(&m->queued_ios);
-	spin_unlock_irqrestore(&m->lock, flags);
-
-	while (bio) {
-		next = bio->bi_next;
-		bio->bi_next = NULL;
-
-		info = dm_get_mapinfo(bio);
-		mpio = info->ptr;
-
-		r = map_io(m, bio, mpio, 1);
-		if (r < 0)
-			bio_endio(bio, r);
-		else if (r == DM_MAPIO_REMAPPED)
-			generic_make_request(bio);
-		else if (r == DM_MAPIO_REQUEUE)
-			bio_endio(bio, -EIO);
-
-		bio = next;
-	}
-}
-
-static void process_queued_ios(struct work_struct *work)
-{
-	struct multipath *m =
-		container_of(work, struct multipath, process_queued_ios);
-	struct pgpath *pgpath = NULL;
-	unsigned init_required = 0, must_queue = 1;
-	unsigned long flags;
-
-	spin_lock_irqsave(&m->lock, flags);
-
-	if (!m->queue_size)
-		goto out;
-
-	if (!m->current_pgpath)
-		__choose_pgpath(m);
+	if (!queue_if_no_path)
+		dm_table_run_md_queue_async(m->ti->table);
 
-	pgpath = m->current_pgpath;
-
-	if ((pgpath && !m->queue_io) ||
-	    (!pgpath && !m->queue_if_no_path))
-		must_queue = 0;
-
-	if (m->pg_init_required && !m->pg_init_in_progress) {
-		m->pg_init_count++;
-		m->pg_init_required = 0;
-		m->pg_init_in_progress = 1;
-		init_required = 1;
-	}
-
-out:
-	spin_unlock_irqrestore(&m->lock, flags);
-
-	if (init_required)
-		queue_work(kmpath_handlerd, &m->activate_path);
-
-	if (!must_queue)
-		dispatch_queued_ios(m);
+	return 0;
 }
 
 /*
@@ -465,69 +475,24 @@ static void trigger_event(struct work_struct *work)
  *      <#paths> <#per-path selector args>
  *         [<path> [<arg>]* ]+ ]+
  *---------------------------------------------------------------*/
-struct param {
-	unsigned min;
-	unsigned max;
-	char *error;
-};
-
-static int read_param(struct param *param, char *str, unsigned *v, char **error)
-{
-	if (!str ||
-	    (sscanf(str, "%u", v) != 1) ||
-	    (*v < param->min) ||
-	    (*v > param->max)) {
-		*error = param->error;
-		return -EINVAL;
-	}
-
-	return 0;
-}
-
-struct arg_set {
-	unsigned argc;
-	char **argv;
-};
-
-static char *shift(struct arg_set *as)
-{
-	char *r;
-
-	if (as->argc) {
-		as->argc--;
-		r = *as->argv;
-		as->argv++;
-		return r;
-	}
-
-	return NULL;
-}
-
-static void consume(struct arg_set *as, unsigned n)
-{
-	BUG_ON (as->argc < n);
-	as->argc -= n;
-	as->argv += n;
-}
-
-static int parse_path_selector(struct arg_set *as, struct priority_group *pg,
+static int parse_path_selector(struct dm_arg_set *as, struct priority_group *pg,
 			       struct dm_target *ti)
 {
 	int r;
 	struct path_selector_type *pst;
 	unsigned ps_argc;
 
-	static struct param _params[] = {
+	static struct dm_arg _args[] = {
 		{0, 1024, "invalid number of path selector args"},
 	};
 
-	pst = dm_get_path_selector(shift(as));
+	pst = dm_get_path_selector(dm_shift_arg(as));
 	if (!pst) {
 		ti->error = "unknown path selector type";
 		return -EINVAL;
 	}
 
-	r = read_param(_params, shift(as), &ps_argc, &ti->error);
+	r = dm_read_arg_group(_args, as, &ps_argc, &ti->error);
 	if (r) {
 		dm_put_path_selector(pst);
 		return -EINVAL;
@@ -541,42 +506,90 @@ static int parse_path_selector(struct arg_set *as, struct priority_group *pg,
 	}
 
 	pg->ps.type = pst;
-	consume(as, ps_argc);
+	dm_consume_args(as, ps_argc);
 
 	return 0;
 }
 
-static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
+static struct pgpath *parse_path(struct dm_arg_set *as, struct path_selector *ps,
 			       struct dm_target *ti)
 {
 	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) {
 		ti->error = "no device given";
-		return NULL;
+		return ERR_PTR(-EINVAL);
 	}
 
 	p = alloc_pgpath();
 	if (!p)
-		return NULL;
+		return ERR_PTR(-ENOMEM);
 
-	r = dm_get_device(ti, shift(as), ti->begin, ti->len,
-			  dm_table_get_mode(ti->table), &p->path.dev);
+	r = dm_get_device(ti, dm_shift_arg(as), dm_table_get_mode(ti->table),
+			  &p->path.dev);
 	if (r) {
 		ti->error = "error getting device";
 		goto bad;
 	}
 
+	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) {
-		r = scsi_dh_attach(bdev_get_queue(p->path.dev->bdev),
-				   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:
+			 * try to reattach with correct one.
+			 */
+			scsi_dh_detach(q);
+			r = scsi_dh_attach(q, m->hw_handler_name);
+		}
+
 		if (r < 0) {
+			ti->error = "error attaching hardware handler";
 			dm_put_device(ti, p->path.dev);
 			goto bad;
 		}
+
+		if (m->hw_handler_params) {
+			r = scsi_dh_set_params(q, m->hw_handler_params);
+			if (r < 0) {
+				ti->error = "unable to set hardware "
+							"handler parameters";
+				scsi_dh_detach(q);
+				dm_put_device(ti, p->path.dev);
+				goto bad;
+			}
+		}
 	}
 
 	r = ps->type->add_path(ps, &p->path, as->argc, as->argv, &ti->error);
@@ -589,32 +602,32 @@ static struct pgpath *parse_path(struct arg_set *as, struct path_selector *ps,
 
  bad:
 	free_pgpath(p);
-	return NULL;
+	return ERR_PTR(r);
 }
 
-static struct priority_group *parse_priority_group(struct arg_set *as,
+static struct priority_group *parse_priority_group(struct dm_arg_set *as,
 						   struct multipath *m)
 {
-	static struct param _params[] = {
+	static struct dm_arg _args[] = {
 		{1, 1024, "invalid number of paths"},
 		{0, 1024, "invalid number of selector args"}
 	};
 
 	int r;
-	unsigned i, nr_selector_args, nr_params;
+	unsigned i, nr_selector_args, nr_args;
 	struct priority_group *pg;
 	struct dm_target *ti = m->ti;
 
 	if (as->argc < 2) {
 		as->argc = 0;
-		ti->error = "not enough priority group aruments";
-		return NULL;
+		ti->error = "not enough priority group arguments";
+		return ERR_PTR(-EINVAL);
 	}
 
 	pg = alloc_priority_group();
 	if (!pg) {
 		ti->error = "couldn't allocate priority group";
-		return NULL;
+		return ERR_PTR(-ENOMEM);
 	}
 	pg->m = m;
 
@@ -625,84 +638,109 @@ static struct priority_group *parse_priority_group(struct arg_set *as,
 	/*
 	 * read the paths
 	 */
-	r = read_param(_params, shift(as), &pg->nr_pgpaths, &ti->error);
+	r = dm_read_arg(_args, as, &pg->nr_pgpaths, &ti->error);
 	if (r)
 		goto bad;
 
-	r = read_param(_params + 1, shift(as), &nr_selector_args, &ti->error);
+	r = dm_read_arg(_args + 1, as, &nr_selector_args, &ti->error);
 	if (r)
 		goto bad;
 
-	nr_params = 1 + nr_selector_args;
+	nr_args = 1 + nr_selector_args;
 	for (i = 0; i < pg->nr_pgpaths; i++) {
 		struct pgpath *pgpath;
-		struct arg_set path_args;
+		struct dm_arg_set path_args;
 
-		if (as->argc < nr_params) {
+		if (as->argc < nr_args) {
 			ti->error = "not enough path parameters";
+			r = -EINVAL;
 			goto bad;
 		}
 
-		path_args.argc = nr_params;
+		path_args.argc = nr_args;
 		path_args.argv = as->argv;
 
 		pgpath = parse_path(&path_args, &pg->ps, ti);
-		if (!pgpath)
+		if (IS_ERR(pgpath)) {
+			r = PTR_ERR(pgpath);
 			goto bad;
+		}
 
 		pgpath->pg = pg;
 		list_add_tail(&pgpath->list, &pg->pgpaths);
-		consume(as, nr_params);
+		dm_consume_args(as, nr_args);
 	}
 
 	return pg;
 
  bad:
 	free_priority_group(pg, ti);
-	return NULL;
+	return ERR_PTR(r);
 }
 
-static int parse_hw_handler(struct arg_set *as, struct multipath *m)
+static int parse_hw_handler(struct dm_arg_set *as, struct multipath *m)
 {
 	unsigned hw_argc;
+	int ret;
 	struct dm_target *ti = m->ti;
 
-	static struct param _params[] = {
+	static struct dm_arg _args[] = {
 		{0, 1024, "invalid number of hardware handler args"},
 	};
 
-	if (read_param(_params, shift(as), &hw_argc, &ti->error))
+	if (dm_read_arg_group(_args, as, &hw_argc, &ti->error))
 		return -EINVAL;
 
 	if (!hw_argc)
 		return 0;
 
-	m->hw_handler_name = kstrdup(shift(as), GFP_KERNEL);
-	request_module("scsi_dh_%s", m->hw_handler_name);
-	if (scsi_dh_handler_exist(m->hw_handler_name) == 0) {
+	m->hw_handler_name = kstrdup(dm_shift_arg(as), GFP_KERNEL);
+	if (!try_then_request_module(scsi_dh_handler_exist(m->hw_handler_name),
+				     "scsi_dh_%s", m->hw_handler_name)) {
 		ti->error = "unknown hardware handler type";
-		kfree(m->hw_handler_name);
-		m->hw_handler_name = NULL;
-		return -EINVAL;
+		ret = -EINVAL;
+		goto fail;
 	}
-	consume(as, hw_argc - 1);
+
+	if (hw_argc > 1) {
+		char *p;
+		int i, j, len = 4;
+
+		for (i = 0; i <= hw_argc - 2; i++)
+			len += strlen(as->argv[i]) + 1;
+		p = m->hw_handler_params = kzalloc(len, GFP_KERNEL);
+		if (!p) {
+			ti->error = "memory allocation failed";
+			ret = -ENOMEM;
+			goto fail;
+		}
+		j = sprintf(p, "%d", hw_argc - 1);
+		for (i = 0, p+=j+1; i <= hw_argc - 2; i++, p+=j+1)
+			j = sprintf(p, "%s", as->argv[i]);
+	}
+	dm_consume_args(as, hw_argc - 1);
 
 	return 0;
+fail:
+	kfree(m->hw_handler_name);
+	m->hw_handler_name = NULL;
+	return ret;
 }
 
-static int parse_features(struct arg_set *as, struct multipath *m)
+static int parse_features(struct dm_arg_set *as, struct multipath *m)
 {
 	int r;
 	unsigned argc;
 	struct dm_target *ti = m->ti;
-	const char *param_name;
+	const char *arg_name;
 
-	static struct param _params[] = {
-		{0, 3, "invalid number of feature args"},
+	static struct dm_arg _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"},
 	};
 
-	r = read_param(_params, shift(as), &argc, &ti->error);
+	r = dm_read_arg_group(_args, as, &argc, &ti->error);
 	if (r)
 		return -EINVAL;
 
@@ -710,18 +748,29 @@ static int parse_features(struct arg_set *as, struct multipath *m)
 		return 0;
 
 	do {
-		param_name = shift(as);
+		arg_name = dm_shift_arg(as);
 		argc--;
 
-		if (!strnicmp(param_name, MESG_STR("queue_if_no_path"))) {
+		if (!strcasecmp(arg_name, "queue_if_no_path")) {
 			r = queue_if_no_path(m, 1, 0);
 			continue;
 		}
 
-		if (!strnicmp(param_name, MESG_STR("pg_init_retries")) &&
+		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);
+			argc--;
+			continue;
+		}
+
+		if (!strcasecmp(arg_name, "pg_init_delay_msecs") &&
 		    (argc >= 1)) {
-			r = read_param(_params + 1, shift(as),
-				       &m->pg_init_retries, &ti->error);
+			r = dm_read_arg(_args + 2, as, &m->pg_init_delay_msecs, &ti->error);
 			argc--;
 			continue;
 		}
@@ -736,15 +785,15 @@ static int parse_features(struct arg_set *as, struct multipath *m)
 static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 			 char **argv)
 {
-	/* target parameters */
-	static struct param _params[] = {
-		{1, 1024, "invalid number of priority groups"},
-		{1, 1024, "invalid initial priority group number"},
+	/* target arguments */
+	static struct dm_arg _args[] = {
+		{0, 1024, "invalid number of priority groups"},
+		{0, 1024, "invalid initial priority group number"},
 	};
 
 	int r;
 	struct multipath *m;
-	struct arg_set as;
+	struct dm_arg_set as;
 	unsigned pg_count = 0;
 	unsigned next_pg_num;
 
@@ -765,21 +814,28 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	if (r)
 		goto bad;
 
-	r = read_param(_params, shift(&as), &m->nr_priority_groups, &ti->error);
+	r = dm_read_arg(_args, &as, &m->nr_priority_groups, &ti->error);
 	if (r)
 		goto bad;
 
-	r = read_param(_params + 1, shift(&as), &next_pg_num, &ti->error);
+	r = dm_read_arg(_args + 1, &as, &next_pg_num, &ti->error);
 	if (r)
 		goto bad;
 
+	if ((!m->nr_priority_groups && next_pg_num) ||
+	    (m->nr_priority_groups && !next_pg_num)) {
+		ti->error = "invalid initial priority group";
+		r = -EINVAL;
+		goto bad;
+	}
+
 	/* parse the priority groups */
 	while (as.argc) {
 		struct priority_group *pg;
 
 		pg = parse_priority_group(&as, m);
-		if (!pg) {
-			r = -EINVAL;
+		if (IS_ERR(pg)) {
+			r = PTR_ERR(pg);
 			goto bad;
 		}
 
@@ -797,6 +853,10 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 		goto bad;
 	}
 
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->num_write_same_bios = 1;
+
 	return 0;
 
  bad:
@@ -804,35 +864,54 @@ static int multipath_ctr(struct dm_target *ti, unsigned int argc,
 	return r;
 }
 
-static void multipath_dtr(struct dm_target *ti)
+static void multipath_wait_for_pg_init_completion(struct multipath *m)
 {
-	struct multipath *m = (struct multipath *) ti->private;
+	DECLARE_WAITQUEUE(wait, current);
+	unsigned long flags;
 
-	flush_workqueue(kmpath_handlerd);
-	flush_workqueue(kmultipathd);
-	free_multipath(m);
+	add_wait_queue(&m->pg_init_wait, &wait);
+
+	while (1) {
+		set_current_state(TASK_UNINTERRUPTIBLE);
+
+		spin_lock_irqsave(&m->lock, flags);
+		if (!m->pg_init_in_progress) {
+			spin_unlock_irqrestore(&m->lock, flags);
+			break;
+		}
+		spin_unlock_irqrestore(&m->lock, flags);
+
+		io_schedule();
+	}
+	set_current_state(TASK_RUNNING);
+
+	remove_wait_queue(&m->pg_init_wait, &wait);
 }
 
-/*
- * Map bios, recording original fields for later in case we have to resubmit
- */
-static int multipath_map(struct dm_target *ti, struct bio *bio,
-			 union map_info *map_context)
+static void flush_multipath_work(struct multipath *m)
 {
-	int r;
-	struct dm_mpath_io *mpio;
-	struct multipath *m = (struct multipath *) ti->private;
+	unsigned long flags;
 
-	mpio = mempool_alloc(m->mpio_pool, GFP_NOIO);
-	dm_bio_record(&mpio->details, bio);
+	spin_lock_irqsave(&m->lock, flags);
+	m->pg_init_disabled = 1;
+	spin_unlock_irqrestore(&m->lock, flags);
 
-	map_context->ptr = mpio;
-	bio->bi_rw |= (1 << BIO_RW_FAILFAST);
-	r = map_io(m, bio, mpio, 0);
-	if (r < 0 || r == DM_MAPIO_REQUEUE)
-		mempool_free(mpio, m->mpio_pool);
+	flush_workqueue(kmpath_handlerd);
+	multipath_wait_for_pg_init_completion(m);
+	flush_workqueue(kmultipathd);
+	flush_work(&m->trigger_event);
 
-	return r;
+	spin_lock_irqsave(&m->lock, flags);
+	m->pg_init_disabled = 0;
+	spin_unlock_irqrestore(&m->lock, flags);
+}
+
+static void multipath_dtr(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	flush_multipath_work(m);
+	free_multipath(m);
 }
 
 /*
@@ -845,13 +924,13 @@ static int fail_path(struct pgpath *pgpath)
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (!pgpath->path.is_active)
+	if (!pgpath->is_active)
 		goto out;
 
 	DMWARN("Failing path %s.", pgpath->path.dev->name);
 
 	pgpath->pg->ps.type->fail_path(&pgpath->pg->ps, &pgpath->path);
-	pgpath->path.is_active = 0;
+	pgpath->is_active = 0;
 	pgpath->fail_count++;
 
 	m->nr_valid_paths--;
@@ -862,7 +941,7 @@ static int fail_path(struct pgpath *pgpath)
 	dm_path_uevent(DM_UEVENT_PATH_FAILED, m->ti,
 		      pgpath->path.dev->name, m->nr_valid_paths);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
@@ -875,13 +954,13 @@ out:
  */
 static int reinstate_path(struct pgpath *pgpath)
 {
-	int r = 0;
+	int r = 0, run_queue = 0;
 	unsigned long flags;
 	struct multipath *m = pgpath->pg->m;
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (pgpath->path.is_active)
+	if (pgpath->is_active)
 		goto out;
 
 	if (!pgpath->pg->ps.type->reinstate_path) {
@@ -895,19 +974,25 @@ static int reinstate_path(struct pgpath *pgpath)
 	if (r)
 		goto out;
 
-	pgpath->path.is_active = 1;
+	pgpath->is_active = 1;
 
-	m->current_pgpath = NULL;
-	if (!m->nr_valid_paths++ && m->queue_size)
-		queue_work(kmultipathd, &m->process_queued_ios);
+	if (!m->nr_valid_paths++) {
+		m->current_pgpath = NULL;
+		run_queue = 1;
+	} else if (m->hw_handler_name && (m->current_pg == pgpath->pg)) {
+		if (queue_work(kmpath_handlerd, &pgpath->activate_path.work))
+			m->pg_init_in_progress++;
+	}
 
 	dm_path_uevent(DM_UEVENT_PATH_REINSTATED, m->ti,
 		      pgpath->path.dev->name, m->nr_valid_paths);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 
 out:
 	spin_unlock_irqrestore(&m->lock, flags);
+	if (run_queue)
+		dm_table_run_md_queue_async(m->ti->table);
 
 	return r;
 }
@@ -918,7 +1003,7 @@ out:
 static int action_dev(struct multipath *m, struct dm_dev *dev,
 		      action_fn action)
 {
-	int r = 0;
+	int r = -EINVAL;
 	struct pgpath *pgpath;
 	struct priority_group *pg;
 
@@ -948,7 +1033,7 @@ static void bypass_pg(struct multipath *m, struct priority_group *pg,
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 }
 
 /*
@@ -959,8 +1044,9 @@ static int switch_pg_num(struct multipath *m, const char *pgstr)
 	struct priority_group *pg;
 	unsigned pgnum;
 	unsigned long flags;
+	char dummy;
 
-	if (!pgstr || (sscanf(pgstr, "%u", &pgnum) != 1) || !pgnum ||
+	if (!pgstr || (sscanf(pgstr, "%u%c", &pgnum, &dummy) != 1) || !pgnum ||
 	    (pgnum > m->nr_priority_groups)) {
 		DMWARN("invalid PG number supplied to switch_pg_num");
 		return -EINVAL;
@@ -978,7 +1064,7 @@ static int switch_pg_num(struct multipath *m, const char *pgstr)
 	}
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	queue_work(kmultipathd, &m->trigger_event);
+	schedule_work(&m->trigger_event);
 	return 0;
 }
 
@@ -990,8 +1076,9 @@ static int bypass_pg_num(struct multipath *m, const char *pgstr, int bypassed)
 {
 	struct priority_group *pg;
 	unsigned pgnum;
+	char dummy;
 
-	if (!pgstr || (sscanf(pgstr, "%u", &pgnum) != 1) || !pgnum ||
+	if (!pgstr || (sscanf(pgstr, "%u%c", &pgnum, &dummy) != 1) || !pgnum ||
 	    (pgnum > m->nr_priority_groups)) {
 		DMWARN("invalid PG number supplied to bypass_pg");
 		return -EINVAL;
@@ -1016,7 +1103,7 @@ static int pg_init_limit_reached(struct multipath *m, struct pgpath *pgpath)
 
 	spin_lock_irqsave(&m->lock, flags);
 
-	if (m->pg_init_count <= m->pg_init_retries)
+	if (m->pg_init_count <= m->pg_init_retries && !m->pg_init_disabled)
 		m->pg_init_required = 1;
 	else
 		limit_reached = 1;
@@ -1026,12 +1113,13 @@ static int pg_init_limit_reached(struct multipath *m, struct pgpath *pgpath)
 	return limit_reached;
 }
 
-static void pg_init_done(struct dm_path *path, int errors)
+static void pg_init_done(void *data, int errors)
 {
-	struct pgpath *pgpath = path_to_pgpath(path);
+	struct pgpath *pgpath = data;
 	struct priority_group *pg = pgpath->pg;
 	struct multipath *m = pg->m;
 	unsigned long flags;
+	unsigned delay_retry = 0;
 
 	/* device or driver problems */
 	switch (errors) {
@@ -1042,8 +1130,8 @@ static void pg_init_done(struct dm_path *path, int errors)
 			errors = 0;
 			break;
 		}
-		DMERR("Cannot failover device because scsi_dh_%s was not "
-		      "loaded.", m->hw_handler_name);
+		DMERR("Could not failover the device: Handler scsi_dh_%s "
+		      "Error %d.", m->hw_handler_name, errors);
 		/*
 		 * Fail path for now, so we do not ping pong
 		 */
@@ -1056,8 +1144,9 @@ static void pg_init_done(struct dm_path *path, int errors)
 		 */
 		bypass_pg(m, pg, 1);
 		break;
-	/* TODO: For SCSI_DH_RETRY we should wait a couple seconds */
 	case SCSI_DH_RETRY:
+		/* Wait before retrying. */
+		delay_retry = 1;
 	case SCSI_DH_IMM_RETRY:
 	case SCSI_DH_RES_TEMP_UNAVAIL:
 		if (pg_init_limit_reached(m, pgpath))
@@ -1075,96 +1164,124 @@ static void pg_init_done(struct dm_path *path, int errors)
 
 	spin_lock_irqsave(&m->lock, flags);
 	if (errors) {
-		DMERR("Could not failover device. Error %d.", errors);
-		m->current_pgpath = NULL;
-		m->current_pg = NULL;
-	} else if (!m->pg_init_required) {
-		m->queue_io = 0;
+		if (pgpath == m->current_pgpath) {
+			DMERR("Could not failover device. Error %d.", errors);
+			m->current_pgpath = NULL;
+			m->current_pg = NULL;
+		}
+	} else if (!m->pg_init_required)
 		pg->bypassed = 0;
+
+	if (--m->pg_init_in_progress)
+		/* Activations of other paths are still on going */
+		goto out;
+
+	if (m->pg_init_required) {
+		m->pg_init_delay_retry = delay_retry;
+		if (__pg_init_all_paths(m))
+			goto out;
 	}
+	m->queue_io = 0;
 
-	m->pg_init_in_progress = 0;
-	queue_work(kmultipathd, &m->process_queued_ios);
+	/*
+	 * Wake up any thread waiting to suspend.
+	 */
+	wake_up(&m->pg_init_wait);
+
+out:
 	spin_unlock_irqrestore(&m->lock, flags);
 }
 
 static void activate_path(struct work_struct *work)
 {
-	int ret;
-	struct multipath *m =
-		container_of(work, struct multipath, activate_path);
-	struct dm_path *path = &m->current_pgpath->path;
+	struct pgpath *pgpath =
+		container_of(work, struct pgpath, activate_path.work);
 
-	ret = scsi_dh_activate(bdev_get_queue(path->dev->bdev));
-	pg_init_done(path, ret);
+	if (pgpath->is_active)
+		scsi_dh_activate(bdev_get_queue(pgpath->path.dev->bdev),
+				 pg_init_done, pgpath);
+	else
+		pg_init_done(pgpath, SCSI_DH_DEV_OFFLINED);
+}
+
+static int noretry_error(int error)
+{
+	switch (error) {
+	case -EOPNOTSUPP:
+	case -EREMOTEIO:
+	case -EILSEQ:
+	case -ENODATA:
+	case -ENOSPC:
+		return 1;
+	}
+
+	/* Anything else could be a path failure, so should be retried */
+	return 0;
 }
 
 /*
  * end_io handling
  */
-static int do_end_io(struct multipath *m, struct bio *bio,
+static int do_end_io(struct multipath *m, struct request *clone,
 		     int error, struct dm_mpath_io *mpio)
 {
+	/*
+	 * We don't queue any clone request inside the multipath target
+	 * during end I/O handling, since those clone requests don't have
+	 * bio clones.  If we queue them inside the multipath target,
+	 * we need to make bio clones, that requires memory allocation.
+	 * (See drivers/md/dm.c:end_clone_bio() about why the clone requests
+	 *  don't have bio clones.)
+	 * Instead of queueing the clone request here, we queue the original
+	 * request into dm core, which will remake a clone request and
+	 * clone bios for it and resubmit it later.
+	 */
+	int r = DM_ENDIO_REQUEUE;
 	unsigned long flags;
 
-	if (!error)
+	if (!error && !clone->errors)
 		return 0;	/* I/O complete */
 
-	if ((error == -EWOULDBLOCK) && bio_rw_ahead(bio))
+	if (noretry_error(error))
 		return error;
 
-	if (error == -EOPNOTSUPP)
-		return error;
+	if (mpio->pgpath)
+		fail_path(mpio->pgpath);
 
 	spin_lock_irqsave(&m->lock, flags);
 	if (!m->nr_valid_paths) {
-		if (__must_push_back(m)) {
-			spin_unlock_irqrestore(&m->lock, flags);
-			return DM_ENDIO_REQUEUE;
-		} else if (!m->queue_if_no_path) {
-			spin_unlock_irqrestore(&m->lock, flags);
-			return -EIO;
+		if (!m->queue_if_no_path) {
+			if (!__must_push_back(m))
+				r = -EIO;
 		} else {
-			spin_unlock_irqrestore(&m->lock, flags);
-			goto requeue;
+			if (error == -EBADE)
+				r = error;
 		}
 	}
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	if (mpio->pgpath)
-		fail_path(mpio->pgpath);
-
-      requeue:
-	dm_bio_restore(&mpio->details, bio);
-
-	/* queue for the daemon to resubmit or fail */
-	spin_lock_irqsave(&m->lock, flags);
-	bio_list_add(&m->queued_ios, bio);
-	m->queue_size++;
-	if (!m->queue_io)
-		queue_work(kmultipathd, &m->process_queued_ios);
-	spin_unlock_irqrestore(&m->lock, flags);
-
-	return DM_ENDIO_INCOMPLETE;	/* io not complete */
+	return r;
 }
 
-static int multipath_end_io(struct dm_target *ti, struct bio *bio,
+static int multipath_end_io(struct dm_target *ti, struct request *clone,
 			    int error, union map_info *map_context)
 {
 	struct multipath *m = ti->private;
 	struct dm_mpath_io *mpio = map_context->ptr;
-	struct pgpath *pgpath = mpio->pgpath;
+	struct pgpath *pgpath;
 	struct path_selector *ps;
 	int r;
 
-	r  = do_end_io(m, bio, error, mpio);
+	BUG_ON(!mpio);
+
+	r  = do_end_io(m, clone, error, mpio);
+	pgpath = mpio->pgpath;
 	if (pgpath) {
 		ps = &pgpath->pg->ps;
 		if (ps->type->end_io)
-			ps->type->end_io(ps, &pgpath->path);
+			ps->type->end_io(ps, &pgpath->path, mpio->nr_bytes);
 	}
-	if (r != DM_ENDIO_INCOMPLETE)
-		mempool_free(mpio, m->mpio_pool);
+	clear_mapinfo(m, map_context);
 
 	return r;
 }
@@ -1182,6 +1299,15 @@ static void multipath_presuspend(struct dm_target *ti)
 	queue_if_no_path(m, 0, 1);
 }
 
+static void multipath_postsuspend(struct dm_target *ti)
+{
+	struct multipath *m = ti->private;
+
+	mutex_lock(&m->work_mutex);
+	flush_multipath_work(m);
+	mutex_unlock(&m->work_mutex);
+}
+
 /*
  * Restore the queue_if_no_path setting.
  */
@@ -1211,8 +1337,8 @@ static void multipath_resume(struct dm_target *ti)
  *     [priority selector-name num_ps_args [ps_args]*
  *      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)
+static void multipath_status(struct dm_target *ti, status_type_t type,
+			     unsigned status_flags, char *result, unsigned maxlen)
 {
 	int sz = 0;
 	unsigned long flags;
@@ -1226,14 +1352,20 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 	/* Features */
 	if (type == STATUSTYPE_INFO)
-		DMEMIT("2 %u %u ", m->queue_size, m->pg_init_count);
+		DMEMIT("2 %u %u ", m->queue_io, m->pg_init_count);
 	else {
 		DMEMIT("%u ", m->queue_if_no_path +
-			      (m->pg_init_retries > 0) * 2);
+			      (m->pg_init_retries > 0) * 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)
@@ -1248,7 +1380,7 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 	else if (m->current_pg)
 		pg_num = m->current_pg->pg_num;
 	else
-			pg_num = 1;
+		pg_num = (m->nr_priority_groups ? 1 : 0);
 
 	DMEMIT("%u ", pg_num);
 
@@ -1276,7 +1408,7 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 
 			list_for_each_entry(p, &pg->pgpaths, list) {
 				DMEMIT("%s %s %u ", p->path.dev->name,
-				       p->path.is_active ? "A" : "F",
+				       p->is_active ? "A" : "F",
 				       p->fail_count);
 				if (pg->ps.type->status)
 					sz += pg->ps.type->status(&pg->ps,
@@ -1312,92 +1444,221 @@ static int multipath_status(struct dm_target *ti, status_type_t type,
 	}
 
 	spin_unlock_irqrestore(&m->lock, flags);
-
-	return 0;
 }
 
 static int multipath_message(struct dm_target *ti, unsigned argc, char **argv)
 {
-	int r;
+	int r = -EINVAL;
 	struct dm_dev *dev;
 	struct multipath *m = (struct multipath *) ti->private;
 	action_fn action;
 
+	mutex_lock(&m->work_mutex);
+
+	if (dm_suspended(ti)) {
+		r = -EBUSY;
+		goto out;
+	}
+
 	if (argc == 1) {
-		if (!strnicmp(argv[0], MESG_STR("queue_if_no_path")))
-			return queue_if_no_path(m, 1, 0);
-		else if (!strnicmp(argv[0], MESG_STR("fail_if_no_path")))
-			return queue_if_no_path(m, 0, 0);
+		if (!strcasecmp(argv[0], "queue_if_no_path")) {
+			r = queue_if_no_path(m, 1, 0);
+			goto out;
+		} else if (!strcasecmp(argv[0], "fail_if_no_path")) {
+			r = queue_if_no_path(m, 0, 0);
+			goto out;
+		}
 	}
 
-	if (argc != 2)
-		goto error;
+	if (argc != 2) {
+		DMWARN("Invalid multipath message arguments. Expected 2 arguments, got %d.", argc);
+		goto out;
+	}
 
-	if (!strnicmp(argv[0], MESG_STR("disable_group")))
-		return bypass_pg_num(m, argv[1], 1);
-	else if (!strnicmp(argv[0], MESG_STR("enable_group")))
-		return bypass_pg_num(m, argv[1], 0);
-	else if (!strnicmp(argv[0], MESG_STR("switch_group")))
-		return switch_pg_num(m, argv[1]);
-	else if (!strnicmp(argv[0], MESG_STR("reinstate_path")))
+	if (!strcasecmp(argv[0], "disable_group")) {
+		r = bypass_pg_num(m, argv[1], 1);
+		goto out;
+	} else if (!strcasecmp(argv[0], "enable_group")) {
+		r = bypass_pg_num(m, argv[1], 0);
+		goto out;
+	} else if (!strcasecmp(argv[0], "switch_group")) {
+		r = switch_pg_num(m, argv[1]);
+		goto out;
+	} else if (!strcasecmp(argv[0], "reinstate_path"))
 		action = reinstate_path;
-	else if (!strnicmp(argv[0], MESG_STR("fail_path")))
+	else if (!strcasecmp(argv[0], "fail_path"))
 		action = fail_path;
-	else
-		goto error;
+	else {
+		DMWARN("Unrecognised multipath message received: %s", argv[0]);
+		goto out;
+	}
 
-	r = dm_get_device(ti, argv[1], ti->begin, ti->len,
-			  dm_table_get_mode(ti->table), &dev);
+	r = dm_get_device(ti, argv[1], dm_table_get_mode(ti->table), &dev);
 	if (r) {
 		DMWARN("message: error getting device %s",
 		       argv[1]);
-		return -EINVAL;
+		goto out;
 	}
 
 	r = action_dev(m, dev, action);
 
 	dm_put_device(ti, dev);
 
+out:
+	mutex_unlock(&m->work_mutex);
 	return r;
-
-error:
-	DMWARN("Unrecognised multipath message received.");
-	return -EINVAL;
 }
 
-static int multipath_ioctl(struct dm_target *ti, struct inode *inode,
-			   struct file *filp, unsigned int cmd,
+static int multipath_ioctl(struct dm_target *ti, unsigned int cmd,
 			   unsigned long arg)
 {
-	struct multipath *m = (struct multipath *) ti->private;
-	struct block_device *bdev = NULL;
+	struct multipath *m = ti->private;
+	struct pgpath *pgpath;
+	struct block_device *bdev;
+	fmode_t mode;
 	unsigned long flags;
-	struct file fake_file = {};
-	struct dentry fake_dentry = {};
-	int r = 0;
+	int r;
 
-	fake_file.f_path.dentry = &fake_dentry;
+	bdev = NULL;
+	mode = 0;
+	r = 0;
 
 	spin_lock_irqsave(&m->lock, flags);
 
 	if (!m->current_pgpath)
-		__choose_pgpath(m);
+		__choose_pgpath(m, 0);
+
+	pgpath = m->current_pgpath;
 
-	if (m->current_pgpath) {
-		bdev = m->current_pgpath->path.dev->bdev;
-		fake_dentry.d_inode = bdev->bd_inode;
-		fake_file.f_mode = m->current_pgpath->path.dev->mode;
+	if (pgpath) {
+		bdev = pgpath->path.dev->bdev;
+		mode = pgpath->path.dev->mode;
 	}
 
-	if (m->queue_io)
-		r = -EAGAIN;
+	if ((pgpath && m->queue_io) || (!pgpath && m->queue_if_no_path))
+		r = -ENOTCONN;
 	else if (!bdev)
 		r = -EIO;
 
 	spin_unlock_irqrestore(&m->lock, flags);
 
-	return r ? : blkdev_driver_ioctl(bdev->bd_inode, &fake_file,
-					 bdev->bd_disk, cmd, arg);
+	/*
+	 * Only pass ioctls through if the device sizes match exactly.
+	 */
+	if (!bdev || ti->len != i_size_read(bdev->bd_inode) >> SECTOR_SHIFT) {
+		int err = scsi_verify_blk_ioctl(NULL, cmd);
+		if (err)
+			r = err;
+	}
+
+	if (r == -ENOTCONN && !fatal_signal_pending(current)) {
+		spin_lock_irqsave(&m->lock, flags);
+		if (!m->current_pg) {
+			/* Path status changed, redo selection */
+			__choose_pgpath(m, 0);
+		}
+		if (m->pg_init_required)
+			__pg_init_all_paths(m);
+		spin_unlock_irqrestore(&m->lock, flags);
+		dm_table_run_md_queue_async(m->ti->table);
+	}
+
+	return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);
+}
+
+static int multipath_iterate_devices(struct dm_target *ti,
+				     iterate_devices_callout_fn fn, void *data)
+{
+	struct multipath *m = ti->private;
+	struct priority_group *pg;
+	struct pgpath *p;
+	int ret = 0;
+
+	list_for_each_entry(pg, &m->priority_groups, list) {
+		list_for_each_entry(p, &pg->pgpaths, list) {
+			ret = fn(ti, p->path.dev, ti->begin, ti->len, data);
+			if (ret)
+				goto out;
+		}
+	}
+
+out:
+	return ret;
+}
+
+static int __pgpath_busy(struct pgpath *pgpath)
+{
+	struct request_queue *q = bdev_get_queue(pgpath->path.dev->bdev);
+
+	return dm_underlying_device_busy(q);
+}
+
+/*
+ * We return "busy", only when we can map I/Os but underlying devices
+ * are busy (so even if we map I/Os now, the I/Os will wait on
+ * the underlying queue).
+ * In other words, if we want to kill I/Os or queue them inside us
+ * due to map unavailability, we don't return "busy".  Otherwise,
+ * dm core won't give us the I/Os and we can't do what we want.
+ */
+static int multipath_busy(struct dm_target *ti)
+{
+	int busy = 0, has_active = 0;
+	struct multipath *m = ti->private;
+	struct priority_group *pg;
+	struct pgpath *pgpath;
+	unsigned long flags;
+
+	spin_lock_irqsave(&m->lock, flags);
+
+	/* pg_init in progress or no paths available */
+	if (m->pg_init_in_progress ||
+	    (!m->nr_valid_paths && m->queue_if_no_path)) {
+		busy = 1;
+		goto out;
+	}
+	/* Guess which priority_group will be used at next mapping time */
+	if (unlikely(!m->current_pgpath && m->next_pg))
+		pg = m->next_pg;
+	else if (likely(m->current_pg))
+		pg = m->current_pg;
+	else
+		/*
+		 * We don't know which pg will be used at next mapping time.
+		 * We don't call __choose_pgpath() here to avoid to trigger
+		 * pg_init just by busy checking.
+		 * So we don't know whether underlying devices we will be using
+		 * at next mapping time are busy or not. Just try mapping.
+		 */
+		goto out;
+
+	/*
+	 * If there is one non-busy active path at least, the path selector
+	 * will be able to select it. So we consider such a pg as not busy.
+	 */
+	busy = 1;
+	list_for_each_entry(pgpath, &pg->pgpaths, list)
+		if (pgpath->is_active) {
+			has_active = 1;
+
+			if (!__pgpath_busy(pgpath)) {
+				busy = 0;
+				break;
+			}
+		}
+
+	if (!has_active)
+		/*
+		 * No active path in this pg, so this pg won't be used and
+		 * the current_pg will be changed at next mapping time.
+		 * We need to try mapping to determine it.
+		 */
+		busy = 0;
+
+out:
+	spin_unlock_irqrestore(&m->lock, flags);
+
+	return busy;
 }
 
 /*-----------------------------------------------------------------
@@ -1405,17 +1666,20 @@ static int multipath_ioctl(struct dm_target *ti, struct inode *inode,
  *---------------------------------------------------------------*/
 static struct target_type multipath_target = {
 	.name = "multipath",
-	.version = {1, 0, 5},
+	.version = {1, 7, 0},
 	.module = THIS_MODULE,
 	.ctr = multipath_ctr,
 	.dtr = multipath_dtr,
-	.map = multipath_map,
-	.end_io = multipath_end_io,
+	.map_rq = multipath_map,
+	.rq_end_io = multipath_end_io,
 	.presuspend = multipath_presuspend,
+	.postsuspend = multipath_postsuspend,
 	.resume = multipath_resume,
 	.status = multipath_status,
 	.message = multipath_message,
 	.ioctl  = multipath_ioctl,
+	.iterate_devices = multipath_iterate_devices,
+	.busy = multipath_busy,
 };
 
 static int __init dm_multipath_init(void)
@@ -1434,7 +1698,7 @@ static int __init dm_multipath_init(void)
 		return -EINVAL;
 	}
 
-	kmultipathd = create_workqueue("kmpathd");
+	kmultipathd = alloc_workqueue("kmpathd", WQ_MEM_RECLAIM, 0);
 	if (!kmultipathd) {
 		DMERR("failed to create workqueue kmpathd");
 		dm_unregister_target(&multipath_target);
@@ -1448,7 +1712,8 @@ static int __init dm_multipath_init(void)
 	 * old workqueue would also create a bottleneck in the
 	 * path of the storage hardware device activation.
 	 */
-	kmpath_handlerd = create_singlethread_workqueue("kmpath_handlerd");
+	kmpath_handlerd = alloc_ordered_workqueue("kmpath_handlerd",
+						  WQ_MEM_RECLAIM);
 	if (!kmpath_handlerd) {
 		DMERR("failed to create workqueue kmpath_handlerd");
 		destroy_workqueue(kmultipathd);
@@ -1466,14 +1731,10 @@ static int __init dm_multipath_init(void)
 
 static void __exit dm_multipath_exit(void)
 {
-	int r;
-
 	destroy_workqueue(kmpath_handlerd);
 	destroy_workqueue(kmultipathd);
 
-	r = dm_unregister_target(&multipath_target);
-	if (r < 0)
-		DMERR("target unregister failed %d", r);
+	dm_unregister_target(&multipath_target);
 	kmem_cache_destroy(_mpio_cache);
 }
 
diff --git a/drivers/md/dm-mpath.h b/drivers/md/dm-mpath.h
index c198b856a45..e230f719625 100644
--- a/drivers/md/dm-mpath.h
+++ b/drivers/md/dm-mpath.h
@@ -13,8 +13,6 @@ struct dm_dev;
 
 struct dm_path {
 	struct dm_dev *dev;	/* Read-only */
-	unsigned is_active;	/* Read-only */
-
 	void *pscontext;	/* For path-selector use */
 };
 
diff --git a/drivers/md/dm-path-selector.c b/drivers/md/dm-path-selector.c
index ca1bb636a3e..fa0ccc585cb 100644
--- a/drivers/md/dm-path-selector.c
+++ b/drivers/md/dm-path-selector.c
@@ -9,16 +9,16 @@
  * Path selector registration.
  */
 
-#include "dm.h"
+#include <linux/device-mapper.h>
+#include <linux/module.h>
+
 #include "dm-path-selector.h"
 
 #include <linux/slab.h>
 
 struct ps_internal {
 	struct path_selector_type pst;
-
 	struct list_head list;
-	long use;
 };
 
 #define pst_to_psi(__pst) container_of((__pst), struct ps_internal, pst)
@@ -44,12 +44,8 @@ static struct ps_internal *get_path_selector(const char *name)
 
 	down_read(&_ps_lock);
 	psi = __find_path_selector_type(name);
-	if (psi) {
-		if ((psi->use == 0) && !try_module_get(psi->pst.module))
-			psi = NULL;
-		else
-			psi->use++;
-	}
+	if (psi && !try_module_get(psi->pst.module))
+		psi = NULL;
 	up_read(&_ps_lock);
 
 	return psi;
@@ -83,11 +79,7 @@ void dm_put_path_selector(struct path_selector_type *pst)
 	if (!psi)
 		goto out;
 
-	if (--psi->use == 0)
-		module_put(psi->pst.module);
-
-	BUG_ON(psi->use < 0);
-
+	module_put(psi->pst.module);
 out:
 	up_read(&_ps_lock);
 }
@@ -135,11 +127,6 @@ int dm_unregister_path_selector(struct path_selector_type *pst)
 		return -EINVAL;
 	}
 
-	if (psi->use) {
-		up_write(&_ps_lock);
-		return -ETXTBSY;
-	}
-
 	list_del(&psi->list);
 
 	up_write(&_ps_lock);
diff --git a/drivers/md/dm-path-selector.h b/drivers/md/dm-path-selector.h
index 27357b85d73..e7d1fa8b045 100644
--- a/drivers/md/dm-path-selector.h
+++ b/drivers/md/dm-path-selector.h
@@ -56,7 +56,8 @@ struct path_selector_type {
 	 * the path fails.
 	 */
 	struct dm_path *(*select_path) (struct path_selector *ps,
-				     unsigned *repeat_count);
+					unsigned *repeat_count,
+					size_t nr_bytes);
 
 	/*
 	 * Notify the selector that a path has failed.
@@ -75,7 +76,10 @@ struct path_selector_type {
 	int (*status) (struct path_selector *ps, struct dm_path *path,
 		       status_type_t type, char *result, unsigned int maxlen);
 
-	int (*end_io) (struct path_selector *ps, struct dm_path *path);
+	int (*start_io) (struct path_selector *ps, struct dm_path *path,
+			 size_t nr_bytes);
+	int (*end_io) (struct path_selector *ps, struct dm_path *path,
+		       size_t nr_bytes);
 };
 
 /* Register a path selector */
diff --git a/drivers/md/dm-queue-length.c b/drivers/md/dm-queue-length.c
new file mode 100644
index 00000000000..3941fae0de9
--- /dev/null
+++ b/drivers/md/dm-queue-length.c
@@ -0,0 +1,264 @@
+/*
+ * Copyright (C) 2004-2005 IBM Corp.  All Rights Reserved.
+ * Copyright (C) 2006-2009 NEC Corporation.
+ *
+ * dm-queue-length.c
+ *
+ * Module Author: Stefan Bader, IBM
+ * Modified by: Kiyoshi Ueda, NEC
+ *
+ * This file is released under the GPL.
+ *
+ * queue-length path selector - choose a path with the least number of
+ * in-flight I/Os.
+ */
+
+#include "dm.h"
+#include "dm-path-selector.h"
+
+#include <linux/slab.h>
+#include <linux/ctype.h>
+#include <linux/errno.h>
+#include <linux/module.h>
+#include <linux/atomic.h>
+
+#define DM_MSG_PREFIX	"multipath queue-length"
+#define QL_MIN_IO	128
+#define QL_VERSION	"0.1.0"
+
+struct selector {
+	struct list_head	valid_paths;
+	struct list_head	failed_paths;
+};
+
+struct path_info {
+	struct list_head	list;
+	struct dm_path		*path;
+	unsigned		repeat_count;
+	atomic_t		qlen;	/* the number of in-flight I/Os */
+};
+
+static struct selector *alloc_selector(void)
+{
+	struct selector *s = kmalloc(sizeof(*s), GFP_KERNEL);
+
+	if (s) {
+		INIT_LIST_HEAD(&s->valid_paths);
+		INIT_LIST_HEAD(&s->failed_paths);
+	}
+
+	return s;
+}
+
+static int ql_create(struct path_selector *ps, unsigned argc, char **argv)
+{
+	struct selector *s = alloc_selector();
+
+	if (!s)
+		return -ENOMEM;
+
+	ps->context = s;
+	return 0;
+}
+
+static void ql_free_paths(struct list_head *paths)
+{
+	struct path_info *pi, *next;
+
+	list_for_each_entry_safe(pi, next, paths, list) {
+		list_del(&pi->list);
+		kfree(pi);
+	}
+}
+
+static void ql_destroy(struct path_selector *ps)
+{
+	struct selector *s = ps->context;
+
+	ql_free_paths(&s->valid_paths);
+	ql_free_paths(&s->failed_paths);
+	kfree(s);
+	ps->context = NULL;
+}
+
+static int ql_status(struct path_selector *ps, struct dm_path *path,
+		     status_type_t type, char *result, unsigned maxlen)
+{
+	unsigned sz = 0;
+	struct path_info *pi;
+
+	/* When called with NULL path, return selector status/args. */
+	if (!path)
+		DMEMIT("0 ");
+	else {
+		pi = path->pscontext;
+
+		switch (type) {
+		case STATUSTYPE_INFO:
+			DMEMIT("%d ", atomic_read(&pi->qlen));
+			break;
+		case STATUSTYPE_TABLE:
+			DMEMIT("%u ", pi->repeat_count);
+			break;
+		}
+	}
+
+	return sz;
+}
+
+static int ql_add_path(struct path_selector *ps, struct dm_path *path,
+		       int argc, char **argv, char **error)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi;
+	unsigned repeat_count = QL_MIN_IO;
+	char dummy;
+
+	/*
+	 * Arguments: [<repeat_count>]
+	 * 	<repeat_count>: The number of I/Os before switching path.
+	 * 			If not given, default (QL_MIN_IO) is used.
+	 */
+	if (argc > 1) {
+		*error = "queue-length ps: incorrect number of arguments";
+		return -EINVAL;
+	}
+
+	if ((argc == 1) && (sscanf(argv[0], "%u%c", &repeat_count, &dummy) != 1)) {
+		*error = "queue-length ps: invalid repeat count";
+		return -EINVAL;
+	}
+
+	/* Allocate the path information structure */
+	pi = kmalloc(sizeof(*pi), GFP_KERNEL);
+	if (!pi) {
+		*error = "queue-length ps: Error allocating path information";
+		return -ENOMEM;
+	}
+
+	pi->path = path;
+	pi->repeat_count = repeat_count;
+	atomic_set(&pi->qlen, 0);
+
+	path->pscontext = pi;
+
+	list_add_tail(&pi->list, &s->valid_paths);
+
+	return 0;
+}
+
+static void ql_fail_path(struct path_selector *ps, struct dm_path *path)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = path->pscontext;
+
+	list_move(&pi->list, &s->failed_paths);
+}
+
+static int ql_reinstate_path(struct path_selector *ps, struct dm_path *path)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = path->pscontext;
+
+	list_move_tail(&pi->list, &s->valid_paths);
+
+	return 0;
+}
+
+/*
+ * Select a path having the minimum number of in-flight I/Os
+ */
+static struct dm_path *ql_select_path(struct path_selector *ps,
+				      unsigned *repeat_count, size_t nr_bytes)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = NULL, *best = NULL;
+
+	if (list_empty(&s->valid_paths))
+		return NULL;
+
+	/* Change preferred (first in list) path to evenly balance. */
+	list_move_tail(s->valid_paths.next, &s->valid_paths);
+
+	list_for_each_entry(pi, &s->valid_paths, list) {
+		if (!best ||
+		    (atomic_read(&pi->qlen) < atomic_read(&best->qlen)))
+			best = pi;
+
+		if (!atomic_read(&best->qlen))
+			break;
+	}
+
+	if (!best)
+		return NULL;
+
+	*repeat_count = best->repeat_count;
+
+	return best->path;
+}
+
+static int ql_start_io(struct path_selector *ps, struct dm_path *path,
+		       size_t nr_bytes)
+{
+	struct path_info *pi = path->pscontext;
+
+	atomic_inc(&pi->qlen);
+
+	return 0;
+}
+
+static int ql_end_io(struct path_selector *ps, struct dm_path *path,
+		     size_t nr_bytes)
+{
+	struct path_info *pi = path->pscontext;
+
+	atomic_dec(&pi->qlen);
+
+	return 0;
+}
+
+static struct path_selector_type ql_ps = {
+	.name		= "queue-length",
+	.module		= THIS_MODULE,
+	.table_args	= 1,
+	.info_args	= 1,
+	.create		= ql_create,
+	.destroy	= ql_destroy,
+	.status		= ql_status,
+	.add_path	= ql_add_path,
+	.fail_path	= ql_fail_path,
+	.reinstate_path	= ql_reinstate_path,
+	.select_path	= ql_select_path,
+	.start_io	= ql_start_io,
+	.end_io		= ql_end_io,
+};
+
+static int __init dm_ql_init(void)
+{
+	int r = dm_register_path_selector(&ql_ps);
+
+	if (r < 0)
+		DMERR("register failed %d", r);
+
+	DMINFO("version " QL_VERSION " loaded");
+
+	return r;
+}
+
+static void __exit dm_ql_exit(void)
+{
+	int r = dm_unregister_path_selector(&ql_ps);
+
+	if (r < 0)
+		DMERR("unregister failed %d", r);
+}
+
+module_init(dm_ql_init);
+module_exit(dm_ql_exit);
+
+MODULE_AUTHOR("Stefan Bader <Stefan.Bader at de.ibm.com>");
+MODULE_DESCRIPTION(
+	"(C) Copyright IBM Corp. 2004,2005   All Rights Reserved.\n"
+	DM_NAME " path selector to balance the number of in-flight I/Os"
+);
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-raid.c b/drivers/md/dm-raid.c
new file mode 100644
index 00000000000..4880b69e2e9
--- /dev/null
+++ b/drivers/md/dm-raid.c
@@ -0,0 +1,1693 @@
+/*
+ * Copyright (C) 2010-2011 Neil Brown
+ * Copyright (C) 2010-2011 Red Hat, Inc. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#include "md.h"
+#include "raid1.h"
+#include "raid5.h"
+#include "raid10.h"
+#include "bitmap.h"
+
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "raid"
+
+/*
+ * The following flags are used by dm-raid.c to set up the array state.
+ * They must be cleared before md_run is called.
+ */
+#define FirstUse 10             /* rdev flag */
+
+struct raid_dev {
+	/*
+	 * Two DM devices, one to hold metadata and one to hold the
+	 * actual data/parity.  The reason for this is to not confuse
+	 * ti->len and give more flexibility in altering size and
+	 * characteristics.
+	 *
+	 * While it is possible for this device to be associated
+	 * with a different physical device than the data_dev, it
+	 * is intended for it to be the same.
+	 *    |--------- Physical Device ---------|
+	 *    |- meta_dev -|------ data_dev ------|
+	 */
+	struct dm_dev *meta_dev;
+	struct dm_dev *data_dev;
+	struct md_rdev rdev;
+};
+
+/*
+ * Flags for rs->print_flags field.
+ */
+#define DMPF_SYNC              0x1
+#define DMPF_NOSYNC            0x2
+#define DMPF_REBUILD           0x4
+#define DMPF_DAEMON_SLEEP      0x8
+#define DMPF_MIN_RECOVERY_RATE 0x10
+#define DMPF_MAX_RECOVERY_RATE 0x20
+#define DMPF_MAX_WRITE_BEHIND  0x40
+#define DMPF_STRIPE_CACHE      0x80
+#define DMPF_REGION_SIZE       0x100
+#define DMPF_RAID10_COPIES     0x200
+#define DMPF_RAID10_FORMAT     0x400
+
+struct raid_set {
+	struct dm_target *ti;
+
+	uint32_t bitmap_loaded;
+	uint32_t print_flags;
+
+	struct mddev md;
+	struct raid_type *raid_type;
+	struct dm_target_callbacks callbacks;
+
+	struct raid_dev dev[0];
+};
+
+/* Supported raid types and properties. */
+static struct raid_type {
+	const char *name;		/* RAID algorithm. */
+	const char *descr;		/* Descriptor text for logging. */
+	const unsigned parity_devs;	/* # of parity devices. */
+	const unsigned minimal_devs;	/* minimal # of devices in set. */
+	const unsigned level;		/* RAID level. */
+	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},
+	{"raid5_ls", "RAID5 (left symmetric)",		1, 2, 5, ALGORITHM_LEFT_SYMMETRIC},
+	{"raid5_rs", "RAID5 (right symmetric)",		1, 2, 5, ALGORITHM_RIGHT_SYMMETRIC},
+	{"raid6_zr", "RAID6 (zero restart)",		2, 4, 6, ALGORITHM_ROTATING_ZERO_RESTART},
+	{"raid6_nr", "RAID6 (N restart)",		2, 4, 6, ALGORITHM_ROTATING_N_RESTART},
+	{"raid6_nc", "RAID6 (N continue)",		2, 4, 6, ALGORITHM_ROTATING_N_CONTINUE}
+};
+
+static char *raid10_md_layout_to_format(int layout)
+{
+	/*
+	 * Bit 16 and 17 stand for "offset" and "use_far_sets"
+	 * Refer to MD's raid10.c for details
+	 */
+	if ((layout & 0x10000) && (layout & 0x20000))
+		return "offset";
+
+	if ((layout & 0xFF) > 1)
+		return "near";
+
+	return "far";
+}
+
+static unsigned raid10_md_layout_to_copies(int layout)
+{
+	if ((layout & 0xFF) > 1)
+		return layout & 0xFF;
+	return (layout >> 8) & 0xFF;
+}
+
+static int raid10_format_to_md_layout(char *format, unsigned copies)
+{
+	unsigned n = 1, f = 1;
+
+	if (!strcmp("near", format))
+		n = copies;
+	else
+		f = copies;
+
+	if (!strcmp("offset", format))
+		return 0x30000 | (f << 8) | n;
+
+	if (!strcmp("far", format))
+		return 0x20000 | (f << 8) | n;
+
+	return (f << 8) | n;
+}
+
+static struct raid_type *get_raid_type(char *name)
+{
+	int i;
+
+	for (i = 0; i < ARRAY_SIZE(raid_types); i++)
+		if (!strcmp(raid_types[i].name, name))
+			return &raid_types[i];
+
+	return NULL;
+}
+
+static struct raid_set *context_alloc(struct dm_target *ti, struct raid_type *raid_type, unsigned raid_devs)
+{
+	unsigned i;
+	struct raid_set *rs;
+
+	if (raid_devs <= raid_type->parity_devs) {
+		ti->error = "Insufficient number of 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";
+		return ERR_PTR(-ENOMEM);
+	}
+
+	mddev_init(&rs->md);
+
+	rs->ti = ti;
+	rs->raid_type = raid_type;
+	rs->md.raid_disks = raid_devs;
+	rs->md.level = raid_type->level;
+	rs->md.new_level = rs->md.level;
+	rs->md.layout = raid_type->algorithm;
+	rs->md.new_layout = rs->md.layout;
+	rs->md.delta_disks = 0;
+	rs->md.recovery_cp = 0;
+
+	for (i = 0; i < raid_devs; i++)
+		md_rdev_init(&rs->dev[i].rdev);
+
+	/*
+	 * Remaining items to be initialized by further RAID params:
+	 *  rs->md.persistent
+	 *  rs->md.external
+	 *  rs->md.chunk_sectors
+	 *  rs->md.new_chunk_sectors
+	 *  rs->md.dev_sectors
+	 */
+
+	return rs;
+}
+
+static void context_free(struct raid_set *rs)
+{
+	int i;
+
+	for (i = 0; i < rs->md.raid_disks; i++) {
+		if (rs->dev[i].meta_dev)
+			dm_put_device(rs->ti, rs->dev[i].meta_dev);
+		md_rdev_clear(&rs->dev[i].rdev);
+		if (rs->dev[i].data_dev)
+			dm_put_device(rs->ti, rs->dev[i].data_dev);
+	}
+
+	kfree(rs);
+}
+
+/*
+ * For every device we have two words
+ *  <meta_dev>: meta device name or '-' if missing
+ *  <data_dev>: data device name or '-' if missing
+ *
+ * The following are permitted:
+ *    - -
+ *    - <data_dev>
+ *    <meta_dev> <data_dev>
+ *
+ * The following is not allowed:
+ *    <meta_dev> -
+ *
+ * This code parses those words.  If there is a failure,
+ * the caller must use context_free to unwind the operations.
+ */
+static int dev_parms(struct raid_set *rs, char **argv)
+{
+	int i;
+	int rebuild = 0;
+	int metadata_available = 0;
+	int ret = 0;
+
+	for (i = 0; i < rs->md.raid_disks; i++, argv += 2) {
+		rs->dev[i].rdev.raid_disk = i;
+
+		rs->dev[i].meta_dev = NULL;
+		rs->dev[i].data_dev = NULL;
+
+		/*
+		 * There are no offsets, since there is a separate device
+		 * for data and metadata.
+		 */
+		rs->dev[i].rdev.data_offset = 0;
+		rs->dev[i].rdev.mddev = &rs->md;
+
+		if (strcmp(argv[0], "-")) {
+			ret = dm_get_device(rs->ti, argv[0],
+					    dm_table_get_mode(rs->ti->table),
+					    &rs->dev[i].meta_dev);
+			rs->ti->error = "RAID metadata device lookup failure";
+			if (ret)
+				return ret;
+
+			rs->dev[i].rdev.sb_page = alloc_page(GFP_KERNEL);
+			if (!rs->dev[i].rdev.sb_page)
+				return -ENOMEM;
+		}
+
+		if (!strcmp(argv[1], "-")) {
+			if (!test_bit(In_sync, &rs->dev[i].rdev.flags) &&
+			    (!rs->dev[i].rdev.recovery_offset)) {
+				rs->ti->error = "Drive designated for rebuild not specified";
+				return -EINVAL;
+			}
+
+			rs->ti->error = "No data device supplied with metadata device";
+			if (rs->dev[i].meta_dev)
+				return -EINVAL;
+
+			continue;
+		}
+
+		ret = dm_get_device(rs->ti, argv[1],
+				    dm_table_get_mode(rs->ti->table),
+				    &rs->dev[i].data_dev);
+		if (ret) {
+			rs->ti->error = "RAID device lookup failure";
+			return ret;
+		}
+
+		if (rs->dev[i].meta_dev) {
+			metadata_available = 1;
+			rs->dev[i].rdev.meta_bdev = rs->dev[i].meta_dev->bdev;
+		}
+		rs->dev[i].rdev.bdev = rs->dev[i].data_dev->bdev;
+		list_add(&rs->dev[i].rdev.same_set, &rs->md.disks);
+		if (!test_bit(In_sync, &rs->dev[i].rdev.flags))
+			rebuild++;
+	}
+
+	if (metadata_available) {
+		rs->md.external = 0;
+		rs->md.persistent = 1;
+		rs->md.major_version = 2;
+	} else if (rebuild && !rs->md.recovery_cp) {
+		/*
+		 * Without metadata, we will not be able to tell if the array
+		 * is in-sync or not - we must assume it is not.  Therefore,
+		 * it is impossible to rebuild a drive.
+		 *
+		 * Even if there is metadata, the on-disk information may
+		 * indicate that the array is not in-sync and it will then
+		 * fail at that time.
+		 *
+		 * User could specify 'nosync' option if desperate.
+		 */
+		DMERR("Unable to rebuild drive while array is not in-sync");
+		rs->ti->error = "RAID device lookup failure";
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+/*
+ * validate_region_size
+ * @rs
+ * @region_size:  region size in sectors.  If 0, pick a size (4MiB default).
+ *
+ * Set rs->md.bitmap_info.chunksize (which really refers to 'region size').
+ * Ensure that (ti->len/region_size < 2^21) - required by MD bitmap.
+ *
+ * Returns: 0 on success, -EINVAL on failure.
+ */
+static int validate_region_size(struct raid_set *rs, unsigned long region_size)
+{
+	unsigned long min_region_size = rs->ti->len / (1 << 21);
+
+	if (!region_size) {
+		/*
+		 * Choose a reasonable default.  All figures in sectors.
+		 */
+		if (min_region_size > (1 << 13)) {
+			/* If not a power of 2, make it the next power of 2 */
+			if (min_region_size & (min_region_size - 1))
+				region_size = 1 << fls(region_size);
+			DMINFO("Choosing default region size of %lu sectors",
+			       region_size);
+		} else {
+			DMINFO("Choosing default region size of 4MiB");
+			region_size = 1 << 13; /* sectors */
+		}
+	} else {
+		/*
+		 * Validate user-supplied value.
+		 */
+		if (region_size > rs->ti->len) {
+			rs->ti->error = "Supplied region size is too large";
+			return -EINVAL;
+		}
+
+		if (region_size < min_region_size) {
+			DMERR("Supplied region_size (%lu sectors) below minimum (%lu)",
+			      region_size, min_region_size);
+			rs->ti->error = "Supplied region size is too small";
+			return -EINVAL;
+		}
+
+		if (!is_power_of_2(region_size)) {
+			rs->ti->error = "Region size is not a power of 2";
+			return -EINVAL;
+		}
+
+		if (region_size < rs->md.chunk_sectors) {
+			rs->ti->error = "Region size is smaller than the chunk size";
+			return -EINVAL;
+		}
+	}
+
+	/*
+	 * Convert sectors to bytes.
+	 */
+	rs->md.bitmap_info.chunksize = (region_size << 9);
+
+	return 0;
+}
+
+/*
+ * validate_raid_redundancy
+ * @rs
+ *
+ * Determine if there are enough devices in the array that haven't
+ * failed (or are being rebuilt) to form a usable array.
+ *
+ * Returns: 0 on success, -EINVAL on failure.
+ */
+static int validate_raid_redundancy(struct raid_set *rs)
+{
+	unsigned i, rebuild_cnt = 0;
+	unsigned rebuilds_per_group = 0, copies, d;
+	unsigned group_size, last_group_start;
+
+	for (i = 0; i < rs->md.raid_disks; i++)
+		if (!test_bit(In_sync, &rs->dev[i].rdev.flags) ||
+		    !rs->dev[i].rdev.sb_page)
+			rebuild_cnt++;
+
+	switch (rs->raid_type->level) {
+	case 1:
+		if (rebuild_cnt >= rs->md.raid_disks)
+			goto too_many;
+		break;
+	case 4:
+	case 5:
+	case 6:
+		if (rebuild_cnt > rs->raid_type->parity_devs)
+			goto too_many;
+		break;
+	case 10:
+		copies = raid10_md_layout_to_copies(rs->md.layout);
+		if (rebuild_cnt < copies)
+			break;
+
+		/*
+		 * It is possible to have a higher rebuild count for RAID10,
+		 * as long as the failed devices occur in different mirror
+		 * groups (i.e. different stripes).
+		 *
+		 * When checking "near" format, make sure no adjacent devices
+		 * have failed beyond what can be handled.  In addition to the
+		 * simple case where the number of devices is a multiple of the
+		 * number of copies, we must also handle cases where the number
+		 * of devices is not a multiple of the number of copies.
+		 * E.g.    dev1 dev2 dev3 dev4 dev5
+		 *          A    A    B    B    C
+		 *          C    D    D    E    E
+		 */
+		if (!strcmp("near", raid10_md_layout_to_format(rs->md.layout))) {
+			for (i = 0; i < rs->md.raid_disks * copies; i++) {
+				if (!(i % copies))
+					rebuilds_per_group = 0;
+				d = i % rs->md.raid_disks;
+				if ((!rs->dev[d].rdev.sb_page ||
+				     !test_bit(In_sync, &rs->dev[d].rdev.flags)) &&
+				    (++rebuilds_per_group >= copies))
+					goto too_many;
+			}
+			break;
+		}
+
+		/*
+		 * When checking "far" and "offset" formats, we need to ensure
+		 * that the device that holds its copy is not also dead or
+		 * being rebuilt.  (Note that "far" and "offset" formats only
+		 * support two copies right now.  These formats also only ever
+		 * use the 'use_far_sets' variant.)
+		 *
+		 * This check is somewhat complicated by the need to account
+		 * for arrays that are not a multiple of (far) copies.  This
+		 * results in the need to treat the last (potentially larger)
+		 * set differently.
+		 */
+		group_size = (rs->md.raid_disks / copies);
+		last_group_start = (rs->md.raid_disks / group_size) - 1;
+		last_group_start *= group_size;
+		for (i = 0; i < rs->md.raid_disks; i++) {
+			if (!(i % copies) && !(i > last_group_start))
+				rebuilds_per_group = 0;
+			if ((!rs->dev[i].rdev.sb_page ||
+			     !test_bit(In_sync, &rs->dev[i].rdev.flags)) &&
+			    (++rebuilds_per_group >= copies))
+					goto too_many;
+		}
+		break;
+	default:
+		if (rebuild_cnt)
+			return -EINVAL;
+	}
+
+	return 0;
+
+too_many:
+	return -EINVAL;
+}
+
+/*
+ * Possible arguments are...
+ *	<chunk_size> [optional_args]
+ *
+ * Argument definitions
+ *    <chunk_size>			The number of sectors per disk that
+ *                                      will form the "stripe"
+ *    [[no]sync]			Force or prevent recovery of the
+ *                                      entire array
+ *    [rebuild <idx>]			Rebuild the drive indicated by the index
+ *    [daemon_sleep <ms>]		Time between bitmap daemon work to
+ *                                      clear bits
+ *    [min_recovery_rate <kB/sec/disk>]	Throttle RAID initialization
+ *    [max_recovery_rate <kB/sec/disk>]	Throttle RAID initialization
+ *    [write_mostly <idx>]		Indicate a write mostly drive via index
+ *    [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|far|offset>] 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;
+	unsigned long value, region_size = 0;
+	sector_t sectors_per_dev = rs->ti->len;
+	sector_t max_io_len;
+	char *key;
+
+	/*
+	 * First, parse the in-order required arguments
+	 * "chunk_size" is the only argument of this type.
+	 */
+	if ((kstrtoul(argv[0], 10, &value) < 0)) {
+		rs->ti->error = "Bad chunk size";
+		return -EINVAL;
+	} else if (rs->raid_type->level == 1) {
+		if (value)
+			DMERR("Ignoring chunk size parameter for RAID 1");
+		value = 0;
+	} else if (!is_power_of_2(value)) {
+		rs->ti->error = "Chunk size must be a power of 2";
+		return -EINVAL;
+	} else if (value < 8) {
+		rs->ti->error = "Chunk size value is too small";
+		return -EINVAL;
+	}
+
+	rs->md.new_chunk_sectors = rs->md.chunk_sectors = value;
+	argv++;
+	num_raid_params--;
+
+	/*
+	 * We set each individual device as In_sync with a completed
+	 * 'recovery_offset'.  If there has been a device failure or
+	 * replacement then one of the following cases applies:
+	 *
+	 *   1) User specifies 'rebuild'.
+	 *      - Device is reset when param is read.
+	 *   2) A new device is supplied.
+	 *      - No matching superblock found, resets device.
+	 *   3) Device failure was transient and returns on reload.
+	 *      - Failure noticed, resets device for bitmap replay.
+	 *   4) Device hadn't completed recovery after previous failure.
+	 *      - Superblock is read and overrides recovery_offset.
+	 *
+	 * What is found in the superblocks of the devices is always
+	 * authoritative, unless 'rebuild' or '[no]sync' was specified.
+	 */
+	for (i = 0; i < rs->md.raid_disks; i++) {
+		set_bit(In_sync, &rs->dev[i].rdev.flags);
+		rs->dev[i].rdev.recovery_offset = MaxSector;
+	}
+
+	/*
+	 * Second, parse the unordered optional arguments
+	 */
+	for (i = 0; i < num_raid_params; i++) {
+		if (!strcasecmp(argv[i], "nosync")) {
+			rs->md.recovery_cp = MaxSector;
+			rs->print_flags |= DMPF_NOSYNC;
+			continue;
+		}
+		if (!strcasecmp(argv[i], "sync")) {
+			rs->md.recovery_cp = 0;
+			rs->print_flags |= DMPF_SYNC;
+			continue;
+		}
+
+		/* The rest of the optional arguments come in key/value pairs */
+		if ((i + 1) >= num_raid_params) {
+			rs->ti->error = "Wrong number of raid parameters given";
+			return -EINVAL;
+		}
+
+		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]) &&
+			    strcmp("far", argv[i]) &&
+			    strcmp("offset", argv[i])) {
+				rs->ti->error = "Invalid 'raid10_format' value given";
+				return -EINVAL;
+			}
+			raid10_format = argv[i];
+			rs->print_flags |= DMPF_RAID10_FORMAT;
+			continue;
+		}
+
+		if (kstrtoul(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")) {
+			if (value >= rs->md.raid_disks) {
+				rs->ti->error = "Invalid rebuild index given";
+				return -EINVAL;
+			}
+			clear_bit(In_sync, &rs->dev[value].rdev.flags);
+			rs->dev[value].rdev.recovery_offset = 0;
+			rs->print_flags |= DMPF_REBUILD;
+		} else if (!strcasecmp(key, "write_mostly")) {
+			if (rs->raid_type->level != 1) {
+				rs->ti->error = "write_mostly option is only valid for RAID1";
+				return -EINVAL;
+			}
+			if (value >= rs->md.raid_disks) {
+				rs->ti->error = "Invalid write_mostly drive index given";
+				return -EINVAL;
+			}
+			set_bit(WriteMostly, &rs->dev[value].rdev.flags);
+		} else if (!strcasecmp(key, "max_write_behind")) {
+			if (rs->raid_type->level != 1) {
+				rs->ti->error = "max_write_behind option is only valid for RAID1";
+				return -EINVAL;
+			}
+			rs->print_flags |= DMPF_MAX_WRITE_BEHIND;
+
+			/*
+			 * In device-mapper, we specify things in sectors, but
+			 * MD records this value in kB
+			 */
+			value /= 2;
+			if (value > COUNTER_MAX) {
+				rs->ti->error = "Max write-behind limit out of range";
+				return -EINVAL;
+			}
+			rs->md.bitmap_info.max_write_behind = value;
+		} else if (!strcasecmp(key, "daemon_sleep")) {
+			rs->print_flags |= DMPF_DAEMON_SLEEP;
+			if (!value || (value > MAX_SCHEDULE_TIMEOUT)) {
+				rs->ti->error = "daemon sleep period out of range";
+				return -EINVAL;
+			}
+			rs->md.bitmap_info.daemon_sleep = value;
+		} else if (!strcasecmp(key, "stripe_cache")) {
+			rs->print_flags |= DMPF_STRIPE_CACHE;
+
+			/*
+			 * In device-mapper, we specify things in sectors, but
+			 * MD records this value in kB
+			 */
+			value /= 2;
+
+			if ((rs->raid_type->level != 5) &&
+			    (rs->raid_type->level != 6)) {
+				rs->ti->error = "Inappropriate argument: stripe_cache";
+				return -EINVAL;
+			}
+			if (raid5_set_cache_size(&rs->md, (int)value)) {
+				rs->ti->error = "Bad stripe_cache size";
+				return -EINVAL;
+			}
+		} else if (!strcasecmp(key, "min_recovery_rate")) {
+			rs->print_flags |= DMPF_MIN_RECOVERY_RATE;
+			if (value > INT_MAX) {
+				rs->ti->error = "min_recovery_rate out of range";
+				return -EINVAL;
+			}
+			rs->md.sync_speed_min = (int)value;
+		} else if (!strcasecmp(key, "max_recovery_rate")) {
+			rs->print_flags |= DMPF_MAX_RECOVERY_RATE;
+			if (value > INT_MAX) {
+				rs->ti->error = "max_recovery_rate out of range";
+				return -EINVAL;
+			}
+			rs->md.sync_speed_max = (int)value;
+		} 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";
+			return -EINVAL;
+		}
+	}
+
+	if (validate_region_size(rs, region_size))
+		return -EINVAL;
+
+	if (rs->md.chunk_sectors)
+		max_io_len = rs->md.chunk_sectors;
+	else
+		max_io_len = 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;
+		}
+
+		/*
+		 * If the format is not "near", we only support
+		 * two copies at the moment.
+		 */
+		if (strcmp("near", raid10_format) && (raid10_copies > 2)) {
+			rs->ti->error = "Too many copies for given RAID10 format.";
+			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;
+	rs->md.external = 1;
+
+	return 0;
+}
+
+static void do_table_event(struct work_struct *ws)
+{
+	struct raid_set *rs = container_of(ws, struct raid_set, md.event_work);
+
+	dm_table_event(rs->ti->table);
+}
+
+static int raid_is_congested(struct dm_target_callbacks *cb, int bits)
+{
+	struct raid_set *rs = container_of(cb, struct raid_set, callbacks);
+
+	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);
+}
+
+/*
+ * This structure is never routinely used by userspace, unlike md superblocks.
+ * Devices with this superblock should only ever be accessed via device-mapper.
+ */
+#define DM_RAID_MAGIC 0x64526D44
+struct dm_raid_superblock {
+	__le32 magic;		/* "DmRd" */
+	__le32 features;	/* Used to indicate possible future changes */
+
+	__le32 num_devices;	/* Number of devices in this array. (Max 64) */
+	__le32 array_position;	/* The position of this drive in the array */
+
+	__le64 events;		/* Incremented by md when superblock updated */
+	__le64 failed_devices;	/* Bit field of devices to indicate failures */
+
+	/*
+	 * This offset tracks the progress of the repair or replacement of
+	 * an individual drive.
+	 */
+	__le64 disk_recovery_offset;
+
+	/*
+	 * This offset tracks the progress of the initial array
+	 * synchronisation/parity calculation.
+	 */
+	__le64 array_resync_offset;
+
+	/*
+	 * RAID characteristics
+	 */
+	__le32 level;
+	__le32 layout;
+	__le32 stripe_sectors;
+
+	__u8 pad[452];		/* Round struct to 512 bytes. */
+				/* Always set to 0 when writing. */
+} __packed;
+
+static int read_disk_sb(struct md_rdev *rdev, int size)
+{
+	BUG_ON(!rdev->sb_page);
+
+	if (rdev->sb_loaded)
+		return 0;
+
+	if (!sync_page_io(rdev, 0, size, rdev->sb_page, READ, 1)) {
+		DMERR("Failed to read superblock of device at position %d",
+		      rdev->raid_disk);
+		md_error(rdev->mddev, rdev);
+		return -EINVAL;
+	}
+
+	rdev->sb_loaded = 1;
+
+	return 0;
+}
+
+static void super_sync(struct mddev *mddev, struct md_rdev *rdev)
+{
+	int i;
+	uint64_t failed_devices;
+	struct dm_raid_superblock *sb;
+	struct raid_set *rs = container_of(mddev, struct raid_set, md);
+
+	sb = page_address(rdev->sb_page);
+	failed_devices = le64_to_cpu(sb->failed_devices);
+
+	for (i = 0; i < mddev->raid_disks; i++)
+		if (!rs->dev[i].data_dev ||
+		    test_bit(Faulty, &(rs->dev[i].rdev.flags)))
+			failed_devices |= (1ULL << i);
+
+	memset(sb, 0, sizeof(*sb));
+
+	sb->magic = cpu_to_le32(DM_RAID_MAGIC);
+	sb->features = cpu_to_le32(0);	/* No features yet */
+
+	sb->num_devices = cpu_to_le32(mddev->raid_disks);
+	sb->array_position = cpu_to_le32(rdev->raid_disk);
+
+	sb->events = cpu_to_le64(mddev->events);
+	sb->failed_devices = cpu_to_le64(failed_devices);
+
+	sb->disk_recovery_offset = cpu_to_le64(rdev->recovery_offset);
+	sb->array_resync_offset = cpu_to_le64(mddev->recovery_cp);
+
+	sb->level = cpu_to_le32(mddev->level);
+	sb->layout = cpu_to_le32(mddev->layout);
+	sb->stripe_sectors = cpu_to_le32(mddev->chunk_sectors);
+}
+
+/*
+ * super_load
+ *
+ * This function creates a superblock if one is not found on the device
+ * and will decide which superblock to use if there's a choice.
+ *
+ * Return: 1 if use rdev, 0 if use refdev, -Exxx otherwise
+ */
+static int super_load(struct md_rdev *rdev, struct md_rdev *refdev)
+{
+	int ret;
+	struct dm_raid_superblock *sb;
+	struct dm_raid_superblock *refsb;
+	uint64_t events_sb, events_refsb;
+
+	rdev->sb_start = 0;
+	rdev->sb_size = sizeof(*sb);
+
+	ret = read_disk_sb(rdev, rdev->sb_size);
+	if (ret)
+		return ret;
+
+	sb = page_address(rdev->sb_page);
+
+	/*
+	 * Two cases that we want to write new superblocks and rebuild:
+	 * 1) New device (no matching magic number)
+	 * 2) Device specified for rebuild (!In_sync w/ offset == 0)
+	 */
+	if ((sb->magic != cpu_to_le32(DM_RAID_MAGIC)) ||
+	    (!test_bit(In_sync, &rdev->flags) && !rdev->recovery_offset)) {
+		super_sync(rdev->mddev, rdev);
+
+		set_bit(FirstUse, &rdev->flags);
+
+		/* Force writing of superblocks to disk */
+		set_bit(MD_CHANGE_DEVS, &rdev->mddev->flags);
+
+		/* Any superblock is better than none, choose that if given */
+		return refdev ? 0 : 1;
+	}
+
+	if (!refdev)
+		return 1;
+
+	events_sb = le64_to_cpu(sb->events);
+
+	refsb = page_address(refdev->sb_page);
+	events_refsb = le64_to_cpu(refsb->events);
+
+	return (events_sb > events_refsb) ? 1 : 0;
+}
+
+static int super_init_validation(struct mddev *mddev, struct md_rdev *rdev)
+{
+	int role;
+	struct raid_set *rs = container_of(mddev, struct raid_set, md);
+	uint64_t events_sb;
+	uint64_t failed_devices;
+	struct dm_raid_superblock *sb;
+	uint32_t new_devs = 0;
+	uint32_t rebuilds = 0;
+	struct md_rdev *r;
+	struct dm_raid_superblock *sb2;
+
+	sb = page_address(rdev->sb_page);
+	events_sb = le64_to_cpu(sb->events);
+	failed_devices = le64_to_cpu(sb->failed_devices);
+
+	/*
+	 * Initialise to 1 if this is a new superblock.
+	 */
+	mddev->events = events_sb ? : 1;
+
+	/*
+	 * Reshaping is not currently allowed
+	 */
+	if (le32_to_cpu(sb->level) != mddev->level) {
+		DMERR("Reshaping arrays not yet supported. (RAID level change)");
+		return -EINVAL;
+	}
+	if (le32_to_cpu(sb->layout) != mddev->layout) {
+		DMERR("Reshaping arrays not yet supported. (RAID layout change)");
+		DMERR("  0x%X vs 0x%X", le32_to_cpu(sb->layout), mddev->layout);
+		DMERR("  Old layout: %s w/ %d copies",
+		      raid10_md_layout_to_format(le32_to_cpu(sb->layout)),
+		      raid10_md_layout_to_copies(le32_to_cpu(sb->layout)));
+		DMERR("  New layout: %s w/ %d copies",
+		      raid10_md_layout_to_format(mddev->layout),
+		      raid10_md_layout_to_copies(mddev->layout));
+		return -EINVAL;
+	}
+	if (le32_to_cpu(sb->stripe_sectors) != mddev->chunk_sectors) {
+		DMERR("Reshaping arrays not yet supported. (stripe sectors change)");
+		return -EINVAL;
+	}
+
+	/* We can only change the number of devices in RAID1 right now */
+	if ((rs->raid_type->level != 1) &&
+	    (le32_to_cpu(sb->num_devices) != mddev->raid_disks)) {
+		DMERR("Reshaping arrays not yet supported. (device count change)");
+		return -EINVAL;
+	}
+
+	if (!(rs->print_flags & (DMPF_SYNC | DMPF_NOSYNC)))
+		mddev->recovery_cp = le64_to_cpu(sb->array_resync_offset);
+
+	/*
+	 * During load, we set FirstUse if a new superblock was written.
+	 * There are two reasons we might not have a superblock:
+	 * 1) The array is brand new - in which case, all of the
+	 *    devices must have their In_sync bit set.  Also,
+	 *    recovery_cp must be 0, unless forced.
+	 * 2) This is a new device being added to an old array
+	 *    and the new device needs to be rebuilt - in which
+	 *    case the In_sync bit will /not/ be set and
+	 *    recovery_cp must be MaxSector.
+	 */
+	rdev_for_each(r, mddev) {
+		if (!test_bit(In_sync, &r->flags)) {
+			DMINFO("Device %d specified for rebuild: "
+			       "Clearing superblock", r->raid_disk);
+			rebuilds++;
+		} else if (test_bit(FirstUse, &r->flags))
+			new_devs++;
+	}
+
+	if (!rebuilds) {
+		if (new_devs == mddev->raid_disks) {
+			DMINFO("Superblocks created for new array");
+			set_bit(MD_ARRAY_FIRST_USE, &mddev->flags);
+		} else if (new_devs) {
+			DMERR("New device injected "
+			      "into existing array without 'rebuild' "
+			      "parameter specified");
+			return -EINVAL;
+		}
+	} else if (new_devs) {
+		DMERR("'rebuild' devices cannot be "
+		      "injected into an array with other first-time devices");
+		return -EINVAL;
+	} else if (mddev->recovery_cp != MaxSector) {
+		DMERR("'rebuild' specified while array is not in-sync");
+		return -EINVAL;
+	}
+
+	/*
+	 * Now we set the Faulty bit for those devices that are
+	 * recorded in the superblock as failed.
+	 */
+	rdev_for_each(r, mddev) {
+		if (!r->sb_page)
+			continue;
+		sb2 = page_address(r->sb_page);
+		sb2->failed_devices = 0;
+
+		/*
+		 * Check for any device re-ordering.
+		 */
+		if (!test_bit(FirstUse, &r->flags) && (r->raid_disk >= 0)) {
+			role = le32_to_cpu(sb2->array_position);
+			if (role != r->raid_disk) {
+				if (rs->raid_type->level != 1) {
+					rs->ti->error = "Cannot change device "
+						"positions in RAID array";
+					return -EINVAL;
+				}
+				DMINFO("RAID1 device #%d now at position #%d",
+				       role, r->raid_disk);
+			}
+
+			/*
+			 * Partial recovery is performed on
+			 * returning failed devices.
+			 */
+			if (failed_devices & (1 << role))
+				set_bit(Faulty, &r->flags);
+		}
+	}
+
+	return 0;
+}
+
+static int super_validate(struct mddev *mddev, struct md_rdev *rdev)
+{
+	struct dm_raid_superblock *sb = page_address(rdev->sb_page);
+
+	/*
+	 * If mddev->events is not set, we know we have not yet initialized
+	 * the array.
+	 */
+	if (!mddev->events && super_init_validation(mddev, rdev))
+		return -EINVAL;
+
+	mddev->bitmap_info.offset = 4096 >> 9; /* Enable bitmap creation */
+	rdev->mddev->bitmap_info.default_offset = 4096 >> 9;
+	if (!test_bit(FirstUse, &rdev->flags)) {
+		rdev->recovery_offset = le64_to_cpu(sb->disk_recovery_offset);
+		if (rdev->recovery_offset != MaxSector)
+			clear_bit(In_sync, &rdev->flags);
+	}
+
+	/*
+	 * If a device comes back, set it as not In_sync and no longer faulty.
+	 */
+	if (test_bit(Faulty, &rdev->flags)) {
+		clear_bit(Faulty, &rdev->flags);
+		clear_bit(In_sync, &rdev->flags);
+		rdev->saved_raid_disk = rdev->raid_disk;
+		rdev->recovery_offset = 0;
+	}
+
+	clear_bit(FirstUse, &rdev->flags);
+
+	return 0;
+}
+
+/*
+ * Analyse superblocks and select the freshest.
+ */
+static int analyse_superblocks(struct dm_target *ti, struct raid_set *rs)
+{
+	int ret;
+	struct raid_dev *dev;
+	struct md_rdev *rdev, *tmp, *freshest;
+	struct mddev *mddev = &rs->md;
+
+	freshest = NULL;
+	rdev_for_each_safe(rdev, tmp, mddev) {
+		/*
+		 * Skipping super_load due to DMPF_SYNC will cause
+		 * the array to undergo initialization again as
+		 * though it were new.  This is the intended effect
+		 * of the "sync" directive.
+		 *
+		 * When reshaping capability is added, we must ensure
+		 * that the "sync" directive is disallowed during the
+		 * reshape.
+		 */
+		if (rs->print_flags & DMPF_SYNC)
+			continue;
+
+		if (!rdev->meta_bdev)
+			continue;
+
+		ret = super_load(rdev, freshest);
+
+		switch (ret) {
+		case 1:
+			freshest = rdev;
+			break;
+		case 0:
+			break;
+		default:
+			dev = container_of(rdev, struct raid_dev, rdev);
+			if (dev->meta_dev)
+				dm_put_device(ti, dev->meta_dev);
+
+			dev->meta_dev = NULL;
+			rdev->meta_bdev = NULL;
+
+			if (rdev->sb_page)
+				put_page(rdev->sb_page);
+
+			rdev->sb_page = NULL;
+
+			rdev->sb_loaded = 0;
+
+			/*
+			 * We might be able to salvage the data device
+			 * even though the meta device has failed.  For
+			 * now, we behave as though '- -' had been
+			 * set for this device in the table.
+			 */
+			if (dev->data_dev)
+				dm_put_device(ti, dev->data_dev);
+
+			dev->data_dev = NULL;
+			rdev->bdev = NULL;
+
+			list_del(&rdev->same_set);
+		}
+	}
+
+	if (!freshest)
+		return 0;
+
+	if (validate_raid_redundancy(rs)) {
+		rs->ti->error = "Insufficient redundancy to activate array";
+		return -EINVAL;
+	}
+
+	/*
+	 * Validation of the freshest device provides the source of
+	 * validation for the remaining devices.
+	 */
+	ti->error = "Unable to assemble array: Invalid superblocks";
+	if (super_validate(mddev, freshest))
+		return -EINVAL;
+
+	rdev_for_each(rdev, mddev)
+		if ((rdev != freshest) && super_validate(mddev, rdev))
+			return -EINVAL;
+
+	return 0;
+}
+
+/*
+ * Construct a RAID4/5/6 mapping:
+ * Args:
+ *	<raid_type> <#raid_params> <raid_params>		\
+ *	<#raid_devs> { <meta_dev1> <dev1> .. <meta_devN> <devN> }
+ *
+ * <raid_params> varies by <raid_type>.  See 'parse_raid_params' for
+ * details on possible <raid_params>.
+ */
+static int raid_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int ret;
+	struct raid_type *rt;
+	unsigned long num_raid_params, num_raid_devs;
+	struct raid_set *rs = NULL;
+
+	/* Must have at least <raid_type> <#raid_params> */
+	if (argc < 2) {
+		ti->error = "Too few arguments";
+		return -EINVAL;
+	}
+
+	/* raid type */
+	rt = get_raid_type(argv[0]);
+	if (!rt) {
+		ti->error = "Unrecognised raid_type";
+		return -EINVAL;
+	}
+	argc--;
+	argv++;
+
+	/* number of RAID parameters */
+	if (kstrtoul(argv[0], 10, &num_raid_params) < 0) {
+		ti->error = "Cannot understand number of RAID parameters";
+		return -EINVAL;
+	}
+	argc--;
+	argv++;
+
+	/* Skip over RAID params for now and find out # of devices */
+	if (num_raid_params + 1 > argc) {
+		ti->error = "Arguments do not agree with counts given";
+		return -EINVAL;
+	}
+
+	if ((kstrtoul(argv[num_raid_params], 10, &num_raid_devs) < 0) ||
+	    (num_raid_devs >= INT_MAX)) {
+		ti->error = "Cannot understand number of raid devices";
+		return -EINVAL;
+	}
+
+	rs = context_alloc(ti, rt, (unsigned)num_raid_devs);
+	if (IS_ERR(rs))
+		return PTR_ERR(rs);
+
+	ret = parse_raid_params(rs, argv, (unsigned)num_raid_params);
+	if (ret)
+		goto bad;
+
+	ret = -EINVAL;
+
+	argc -= num_raid_params + 1; /* +1: we already have num_raid_devs */
+	argv += num_raid_params + 1;
+
+	if (argc != (num_raid_devs * 2)) {
+		ti->error = "Supplied RAID devices does not match the count given";
+		goto bad;
+	}
+
+	ret = dev_parms(rs, argv);
+	if (ret)
+		goto bad;
+
+	rs->md.sync_super = super_sync;
+	ret = analyse_superblocks(ti, rs);
+	if (ret)
+		goto bad;
+
+	INIT_WORK(&rs->md.event_work, do_table_event);
+	ti->private = rs;
+	ti->num_flush_bios = 1;
+
+	mutex_lock(&rs->md.reconfig_mutex);
+	ret = md_run(&rs->md);
+	rs->md.in_sync = 0; /* Assume already marked dirty */
+	mutex_unlock(&rs->md.reconfig_mutex);
+
+	if (ret) {
+		ti->error = "Fail to run raid array";
+		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);
+
+	return ret;
+}
+
+static void raid_dtr(struct dm_target *ti)
+{
+	struct raid_set *rs = ti->private;
+
+	list_del_init(&rs->callbacks.list);
+	md_stop(&rs->md);
+	context_free(rs);
+}
+
+static int raid_map(struct dm_target *ti, struct bio *bio)
+{
+	struct raid_set *rs = ti->private;
+	struct mddev *mddev = &rs->md;
+
+	mddev->pers->make_request(mddev, bio);
+
+	return DM_MAPIO_SUBMITTED;
+}
+
+static const char *decipher_sync_action(struct mddev *mddev)
+{
+	if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
+		return "frozen";
+
+	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
+	    (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) {
+		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+			return "reshape";
+
+		if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
+			if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+				return "resync";
+			else if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
+				return "check";
+			return "repair";
+		}
+
+		if (test_bit(MD_RECOVERY_RECOVER, &mddev->recovery))
+			return "recover";
+	}
+
+	return "idle";
+}
+
+static void raid_status(struct dm_target *ti, status_type_t type,
+			unsigned status_flags, char *result, unsigned maxlen)
+{
+	struct raid_set *rs = ti->private;
+	unsigned raid_param_cnt = 1; /* at least 1 for chunksize */
+	unsigned sz = 0;
+	int i, array_in_sync = 0;
+	sector_t sync;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		DMEMIT("%s %d ", rs->raid_type->name, rs->md.raid_disks);
+
+		if (test_bit(MD_RECOVERY_RUNNING, &rs->md.recovery))
+			sync = rs->md.curr_resync_completed;
+		else
+			sync = rs->md.recovery_cp;
+
+		if (sync >= rs->md.resync_max_sectors) {
+			/*
+			 * Sync complete.
+			 */
+			array_in_sync = 1;
+			sync = rs->md.resync_max_sectors;
+		} else if (test_bit(MD_RECOVERY_REQUESTED, &rs->md.recovery)) {
+			/*
+			 * If "check" or "repair" is occurring, the array has
+			 * undergone and initial sync and the health characters
+			 * should not be 'a' anymore.
+			 */
+			array_in_sync = 1;
+		} else {
+			/*
+			 * The array may be doing an initial sync, or it may
+			 * be rebuilding individual components.  If all the
+			 * devices are In_sync, then it is the array that is
+			 * being initialized.
+			 */
+			for (i = 0; i < rs->md.raid_disks; i++)
+				if (!test_bit(In_sync, &rs->dev[i].rdev.flags))
+					array_in_sync = 1;
+		}
+
+		/*
+		 * Status characters:
+		 *  'D' = Dead/Failed device
+		 *  'a' = Alive but not in-sync
+		 *  'A' = Alive and in-sync
+		 */
+		for (i = 0; i < rs->md.raid_disks; i++) {
+			if (test_bit(Faulty, &rs->dev[i].rdev.flags))
+				DMEMIT("D");
+			else if (!array_in_sync ||
+				 !test_bit(In_sync, &rs->dev[i].rdev.flags))
+				DMEMIT("a");
+			else
+				DMEMIT("A");
+		}
+
+		/*
+		 * In-sync ratio:
+		 *  The in-sync ratio shows the progress of:
+		 *   - Initializing the array
+		 *   - Rebuilding a subset of devices of the array
+		 *  The user can distinguish between the two by referring
+		 *  to the status characters.
+		 */
+		DMEMIT(" %llu/%llu",
+		       (unsigned long long) sync,
+		       (unsigned long long) rs->md.resync_max_sectors);
+
+		/*
+		 * Sync action:
+		 *   See Documentation/device-mapper/dm-raid.c for
+		 *   information on each of these states.
+		 */
+		DMEMIT(" %s", decipher_sync_action(&rs->md));
+
+		/*
+		 * resync_mismatches/mismatch_cnt
+		 *   This field shows the number of discrepancies found when
+		 *   performing a "check" of the array.
+		 */
+		DMEMIT(" %llu",
+		       (strcmp(rs->md.last_sync_action, "check")) ? 0 :
+		       (unsigned long long)
+		       atomic64_read(&rs->md.resync_mismatches));
+		break;
+	case STATUSTYPE_TABLE:
+		/* The string you would use to construct this array */
+		for (i = 0; i < rs->md.raid_disks; i++) {
+			if ((rs->print_flags & DMPF_REBUILD) &&
+			    rs->dev[i].data_dev &&
+			    !test_bit(In_sync, &rs->dev[i].rdev.flags))
+				raid_param_cnt += 2; /* for rebuilds */
+			if (rs->dev[i].data_dev &&
+			    test_bit(WriteMostly, &rs->dev[i].rdev.flags))
+				raid_param_cnt += 2;
+		}
+
+		raid_param_cnt += (hweight32(rs->print_flags & ~DMPF_REBUILD) * 2);
+		if (rs->print_flags & (DMPF_SYNC | DMPF_NOSYNC))
+			raid_param_cnt--;
+
+		DMEMIT("%s %u %u", rs->raid_type->name,
+		       raid_param_cnt, rs->md.chunk_sectors);
+
+		if ((rs->print_flags & DMPF_SYNC) &&
+		    (rs->md.recovery_cp == MaxSector))
+			DMEMIT(" sync");
+		if (rs->print_flags & DMPF_NOSYNC)
+			DMEMIT(" nosync");
+
+		for (i = 0; i < rs->md.raid_disks; i++)
+			if ((rs->print_flags & DMPF_REBUILD) &&
+			    rs->dev[i].data_dev &&
+			    !test_bit(In_sync, &rs->dev[i].rdev.flags))
+				DMEMIT(" rebuild %u", i);
+
+		if (rs->print_flags & DMPF_DAEMON_SLEEP)
+			DMEMIT(" daemon_sleep %lu",
+			       rs->md.bitmap_info.daemon_sleep);
+
+		if (rs->print_flags & DMPF_MIN_RECOVERY_RATE)
+			DMEMIT(" min_recovery_rate %d", rs->md.sync_speed_min);
+
+		if (rs->print_flags & DMPF_MAX_RECOVERY_RATE)
+			DMEMIT(" max_recovery_rate %d", rs->md.sync_speed_max);
+
+		for (i = 0; i < rs->md.raid_disks; i++)
+			if (rs->dev[i].data_dev &&
+			    test_bit(WriteMostly, &rs->dev[i].rdev.flags))
+				DMEMIT(" write_mostly %u", i);
+
+		if (rs->print_flags & DMPF_MAX_WRITE_BEHIND)
+			DMEMIT(" max_write_behind %lu",
+			       rs->md.bitmap_info.max_write_behind);
+
+		if (rs->print_flags & DMPF_STRIPE_CACHE) {
+			struct r5conf *conf = rs->md.private;
+
+			/* convert from kiB to sectors */
+			DMEMIT(" stripe_cache %d",
+			       conf ? conf->max_nr_stripes * 2 : 0);
+		}
+
+		if (rs->print_flags & DMPF_REGION_SIZE)
+			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 %s",
+			       raid10_md_layout_to_format(rs->md.layout));
+
+		DMEMIT(" %d", rs->md.raid_disks);
+		for (i = 0; i < rs->md.raid_disks; i++) {
+			if (rs->dev[i].meta_dev)
+				DMEMIT(" %s", rs->dev[i].meta_dev->name);
+			else
+				DMEMIT(" -");
+
+			if (rs->dev[i].data_dev)
+				DMEMIT(" %s", rs->dev[i].data_dev->name);
+			else
+				DMEMIT(" -");
+		}
+	}
+}
+
+static int raid_message(struct dm_target *ti, unsigned argc, char **argv)
+{
+	struct raid_set *rs = ti->private;
+	struct mddev *mddev = &rs->md;
+
+	if (!strcasecmp(argv[0], "reshape")) {
+		DMERR("Reshape not supported.");
+		return -EINVAL;
+	}
+
+	if (!mddev->pers || !mddev->pers->sync_request)
+		return -EINVAL;
+
+	if (!strcasecmp(argv[0], "frozen"))
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+	else
+		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+
+	if (!strcasecmp(argv[0], "idle") || !strcasecmp(argv[0], "frozen")) {
+		if (mddev->sync_thread) {
+			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+			md_reap_sync_thread(mddev);
+		}
+	} else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
+		   test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
+		return -EBUSY;
+	else if (!strcasecmp(argv[0], "resync"))
+		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	else if (!strcasecmp(argv[0], "recover")) {
+		set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
+		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	} else {
+		if (!strcasecmp(argv[0], "check"))
+			set_bit(MD_RECOVERY_CHECK, &mddev->recovery);
+		else if (!!strcasecmp(argv[0], "repair"))
+			return -EINVAL;
+		set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
+		set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+	}
+	if (mddev->ro == 2) {
+		/* A write to sync_action is enough to justify
+		 * canceling read-auto mode
+		 */
+		mddev->ro = 0;
+		if (!mddev->suspended)
+			md_wakeup_thread(mddev->sync_thread);
+	}
+	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	if (!mddev->suspended)
+		md_wakeup_thread(mddev->thread);
+
+	return 0;
+}
+
+static int raid_iterate_devices(struct dm_target *ti,
+				iterate_devices_callout_fn fn, void *data)
+{
+	struct raid_set *rs = ti->private;
+	unsigned i;
+	int ret = 0;
+
+	for (i = 0; !ret && i < rs->md.raid_disks; i++)
+		if (rs->dev[i].data_dev)
+			ret = fn(ti,
+				 rs->dev[i].data_dev,
+				 0, /* No offset on data devs */
+				 rs->md.dev_sectors,
+				 data);
+
+	return ret;
+}
+
+static void raid_io_hints(struct dm_target *ti, struct queue_limits *limits)
+{
+	struct raid_set *rs = ti->private;
+	unsigned chunk_size = rs->md.chunk_sectors << 9;
+	struct r5conf *conf = rs->md.private;
+
+	blk_limits_io_min(limits, chunk_size);
+	blk_limits_io_opt(limits, chunk_size * (conf->raid_disks - conf->max_degraded));
+}
+
+static void raid_presuspend(struct dm_target *ti)
+{
+	struct raid_set *rs = ti->private;
+
+	md_stop_writes(&rs->md);
+}
+
+static void raid_postsuspend(struct dm_target *ti)
+{
+	struct raid_set *rs = ti->private;
+
+	mddev_suspend(&rs->md);
+}
+
+static void attempt_restore_of_faulty_devices(struct raid_set *rs)
+{
+	int i;
+	uint64_t failed_devices, cleared_failed_devices = 0;
+	unsigned long flags;
+	struct dm_raid_superblock *sb;
+	struct md_rdev *r;
+
+	for (i = 0; i < rs->md.raid_disks; i++) {
+		r = &rs->dev[i].rdev;
+		if (test_bit(Faulty, &r->flags) && r->sb_page &&
+		    sync_page_io(r, 0, r->sb_size, r->sb_page, READ, 1)) {
+			DMINFO("Faulty %s device #%d has readable super block."
+			       "  Attempting to revive it.",
+			       rs->raid_type->name, i);
+
+			/*
+			 * Faulty bit may be set, but sometimes the array can
+			 * be suspended before the personalities can respond
+			 * by removing the device from the array (i.e. calling
+			 * 'hot_remove_disk').  If they haven't yet removed
+			 * the failed device, its 'raid_disk' number will be
+			 * '>= 0' - meaning we must call this function
+			 * ourselves.
+			 */
+			if ((r->raid_disk >= 0) &&
+			    (r->mddev->pers->hot_remove_disk(r->mddev, r) != 0))
+				/* Failed to revive this device, try next */
+				continue;
+
+			r->raid_disk = i;
+			r->saved_raid_disk = i;
+			flags = r->flags;
+			clear_bit(Faulty, &r->flags);
+			clear_bit(WriteErrorSeen, &r->flags);
+			clear_bit(In_sync, &r->flags);
+			if (r->mddev->pers->hot_add_disk(r->mddev, r)) {
+				r->raid_disk = -1;
+				r->saved_raid_disk = -1;
+				r->flags = flags;
+			} else {
+				r->recovery_offset = 0;
+				cleared_failed_devices |= 1 << i;
+			}
+		}
+	}
+	if (cleared_failed_devices) {
+		rdev_for_each(r, &rs->md) {
+			sb = page_address(r->sb_page);
+			failed_devices = le64_to_cpu(sb->failed_devices);
+			failed_devices &= ~cleared_failed_devices;
+			sb->failed_devices = cpu_to_le64(failed_devices);
+		}
+	}
+}
+
+static void raid_resume(struct dm_target *ti)
+{
+	struct raid_set *rs = ti->private;
+
+	set_bit(MD_CHANGE_DEVS, &rs->md.flags);
+	if (!rs->bitmap_loaded) {
+		bitmap_load(&rs->md);
+		rs->bitmap_loaded = 1;
+	} else {
+		/*
+		 * A secondary resume while the device is active.
+		 * Take this opportunity to check whether any failed
+		 * devices are reachable again.
+		 */
+		attempt_restore_of_faulty_devices(rs);
+	}
+
+	clear_bit(MD_RECOVERY_FROZEN, &rs->md.recovery);
+	mddev_resume(&rs->md);
+}
+
+static struct target_type raid_target = {
+	.name = "raid",
+	.version = {1, 5, 2},
+	.module = THIS_MODULE,
+	.ctr = raid_ctr,
+	.dtr = raid_dtr,
+	.map = raid_map,
+	.status = raid_status,
+	.message = raid_message,
+	.iterate_devices = raid_iterate_devices,
+	.io_hints = raid_io_hints,
+	.presuspend = raid_presuspend,
+	.postsuspend = raid_postsuspend,
+	.resume = raid_resume,
+};
+
+static int __init dm_raid_init(void)
+{
+	DMINFO("Loading target version %u.%u.%u",
+	       raid_target.version[0],
+	       raid_target.version[1],
+	       raid_target.version[2]);
+	return dm_register_target(&raid_target);
+}
+
+static void __exit dm_raid_exit(void)
+{
+	dm_unregister_target(&raid_target);
+}
+
+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");
+MODULE_AUTHOR("Neil Brown <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-raid1.c b/drivers/md/dm-raid1.c
index ff05fe89308..7dfdb5c746d 100644
--- a/drivers/md/dm-raid1.c
+++ b/drivers/md/dm-raid1.c
@@ -1,30 +1,27 @@
 /*
  * Copyright (C) 2003 Sistina Software Limited.
+ * Copyright (C) 2005-2008 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
 
-#include "dm.h"
-#include "dm-bio-list.h"
 #include "dm-bio-record.h"
 
-#include <linux/ctype.h>
 #include <linux/init.h>
 #include <linux/mempool.h>
 #include <linux/module.h>
 #include <linux/pagemap.h>
 #include <linux/slab.h>
-#include <linux/time.h>
-#include <linux/vmalloc.h>
 #include <linux/workqueue.h>
-#include <linux/log2.h>
-#include <linux/hardirq.h>
+#include <linux/device-mapper.h>
 #include <linux/dm-io.h>
 #include <linux/dm-dirty-log.h>
 #include <linux/dm-kcopyd.h>
+#include <linux/dm-region-hash.h>
 
 #define DM_MSG_PREFIX "raid1"
-#define DM_IO_PAGES 64
+
+#define MAX_RECOVERY 1	/* Maximum number of regions recovered in parallel. */
 
 #define DM_RAID1_HANDLE_ERRORS 0x01
 #define errors_handled(p)	((p)->features & DM_RAID1_HANDLE_ERRORS)
@@ -32,91 +29,11 @@
 static DECLARE_WAIT_QUEUE_HEAD(_kmirrord_recovery_stopped);
 
 /*-----------------------------------------------------------------
- * Region hash
- *
- * The mirror splits itself up into discrete regions.  Each
- * region can be in one of three states: clean, dirty,
- * nosync.  There is no need to put clean regions in the hash.
- *
- * In addition to being present in the hash table a region _may_
- * be present on one of three lists.
- *
- *   clean_regions: Regions on this list have no io pending to
- *   them, they are in sync, we are no longer interested in them,
- *   they are dull.  rh_update_states() will remove them from the
- *   hash table.
- *
- *   quiesced_regions: These regions have been spun down, ready
- *   for recovery.  rh_recovery_start() will remove regions from
- *   this list and hand them to kmirrord, which will schedule the
- *   recovery io with kcopyd.
- *
- *   recovered_regions: Regions that kcopyd has successfully
- *   recovered.  rh_update_states() will now schedule any delayed
- *   io, up the recovery_count, and remove the region from the
- *   hash.
- *
- * There are 2 locks:
- *   A rw spin lock 'hash_lock' protects just the hash table,
- *   this is never held in write mode from interrupt context,
- *   which I believe means that we only have to disable irqs when
- *   doing a write lock.
- *
- *   An ordinary spin lock 'region_lock' that protects the three
- *   lists in the region_hash, with the 'state', 'list' and
- *   'bhs_delayed' fields of the regions.  This is used from irq
- *   context, so all other uses will have to suspend local irqs.
- *---------------------------------------------------------------*/
-struct mirror_set;
-struct region_hash {
-	struct mirror_set *ms;
-	uint32_t region_size;
-	unsigned region_shift;
-
-	/* holds persistent region state */
-	struct dm_dirty_log *log;
-
-	/* hash table */
-	rwlock_t hash_lock;
-	mempool_t *region_pool;
-	unsigned int mask;
-	unsigned int nr_buckets;
-	struct list_head *buckets;
-
-	spinlock_t region_lock;
-	atomic_t recovery_in_flight;
-	struct semaphore recovery_count;
-	struct list_head clean_regions;
-	struct list_head quiesced_regions;
-	struct list_head recovered_regions;
-	struct list_head failed_recovered_regions;
-};
-
-enum {
-	RH_CLEAN,
-	RH_DIRTY,
-	RH_NOSYNC,
-	RH_RECOVERING
-};
-
-struct region {
-	struct region_hash *rh;	/* FIXME: can we get rid of this ? */
-	region_t key;
-	int state;
-
-	struct list_head hash_list;
-	struct list_head list;
-
-	atomic_t pending;
-	struct bio_list delayed_bios;
-};
-
-
-/*-----------------------------------------------------------------
  * Mirror set structures.
  *---------------------------------------------------------------*/
 enum dm_raid1_error {
 	DM_RAID1_WRITE_ERROR,
+	DM_RAID1_FLUSH_ERROR,
 	DM_RAID1_SYNC_ERROR,
 	DM_RAID1_READ_ERROR
 };
@@ -132,22 +49,24 @@ struct mirror {
 struct mirror_set {
 	struct dm_target *ti;
 	struct list_head list;
-	struct region_hash rh;
-	struct dm_kcopyd_client *kcopyd_client;
+
 	uint64_t features;
 
 	spinlock_t lock;	/* protects the lists */
 	struct bio_list reads;
 	struct bio_list writes;
 	struct bio_list failures;
+	struct bio_list holds;	/* bios are waiting until suspend */
 
+	struct dm_region_hash *rh;
+	struct dm_kcopyd_client *kcopyd_client;
 	struct dm_io_client *io_client;
-	mempool_t *read_record_pool;
 
 	/* recovery */
 	region_t nr_regions;
 	int in_sync;
 	int log_failure;
+	int leg_failure;
 	atomic_t suspend;
 
 	atomic_t default_mirror;	/* Default mirror */
@@ -159,25 +78,17 @@ struct mirror_set {
 
 	struct work_struct trigger_event;
 
-	unsigned int nr_mirrors;
+	unsigned nr_mirrors;
 	struct mirror mirror[0];
 };
 
-/*
- * Conversion fns
- */
-static inline region_t bio_to_region(struct region_hash *rh, struct bio *bio)
-{
-	return (bio->bi_sector - rh->ms->ti->begin) >> rh->region_shift;
-}
+DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(raid1_resync_throttle,
+		"A percentage of time allocated for raid resynchronization");
 
-static inline sector_t region_to_sector(struct region_hash *rh, region_t region)
+static void wakeup_mirrord(void *context)
 {
-	return region << rh->region_shift;
-}
+	struct mirror_set *ms = context;
 
-static void wake(struct mirror_set *ms)
-{
 	queue_work(ms->kmirrord_wq, &ms->kmirrord_work);
 }
 
@@ -186,7 +97,7 @@ static void delayed_wake_fn(unsigned long data)
 	struct mirror_set *ms = (struct mirror_set *) data;
 
 	clear_bit(0, &ms->timer_pending);
-	wake(ms);
+	wakeup_mirrord(ms);
 }
 
 static void delayed_wake(struct mirror_set *ms)
@@ -200,479 +111,41 @@ static void delayed_wake(struct mirror_set *ms)
 	add_timer(&ms->timer);
 }
 
-/* FIXME move this */
-static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw);
-
-#define MIN_REGIONS 64
-#define MAX_RECOVERY 1
-static int rh_init(struct region_hash *rh, struct mirror_set *ms,
-		   struct dm_dirty_log *log, uint32_t region_size,
-		   region_t nr_regions)
-{
-	unsigned int nr_buckets, max_buckets;
-	size_t i;
-
-	/*
-	 * Calculate a suitable number of buckets for our hash
-	 * table.
-	 */
-	max_buckets = nr_regions >> 6;
-	for (nr_buckets = 128u; nr_buckets < max_buckets; nr_buckets <<= 1)
-		;
-	nr_buckets >>= 1;
-
-	rh->ms = ms;
-	rh->log = log;
-	rh->region_size = region_size;
-	rh->region_shift = ffs(region_size) - 1;
-	rwlock_init(&rh->hash_lock);
-	rh->mask = nr_buckets - 1;
-	rh->nr_buckets = nr_buckets;
-
-	rh->buckets = vmalloc(nr_buckets * sizeof(*rh->buckets));
-	if (!rh->buckets) {
-		DMERR("unable to allocate region hash memory");
-		return -ENOMEM;
-	}
-
-	for (i = 0; i < nr_buckets; i++)
-		INIT_LIST_HEAD(rh->buckets + i);
-
-	spin_lock_init(&rh->region_lock);
-	sema_init(&rh->recovery_count, 0);
-	atomic_set(&rh->recovery_in_flight, 0);
-	INIT_LIST_HEAD(&rh->clean_regions);
-	INIT_LIST_HEAD(&rh->quiesced_regions);
-	INIT_LIST_HEAD(&rh->recovered_regions);
-	INIT_LIST_HEAD(&rh->failed_recovered_regions);
-
-	rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
-						      sizeof(struct region));
-	if (!rh->region_pool) {
-		vfree(rh->buckets);
-		rh->buckets = NULL;
-		return -ENOMEM;
-	}
-
-	return 0;
-}
-
-static void rh_exit(struct region_hash *rh)
-{
-	unsigned int h;
-	struct region *reg, *nreg;
-
-	BUG_ON(!list_empty(&rh->quiesced_regions));
-	for (h = 0; h < rh->nr_buckets; h++) {
-		list_for_each_entry_safe(reg, nreg, rh->buckets + h, hash_list) {
-			BUG_ON(atomic_read(&reg->pending));
-			mempool_free(reg, rh->region_pool);
-		}
-	}
-
-	if (rh->log)
-		dm_dirty_log_destroy(rh->log);
-	if (rh->region_pool)
-		mempool_destroy(rh->region_pool);
-	vfree(rh->buckets);
-}
-
-#define RH_HASH_MULT 2654435387U
-
-static inline unsigned int rh_hash(struct region_hash *rh, region_t region)
-{
-	return (unsigned int) ((region * RH_HASH_MULT) >> 12) & rh->mask;
-}
-
-static struct region *__rh_lookup(struct region_hash *rh, region_t region)
-{
-	struct region *reg;
-
-	list_for_each_entry (reg, rh->buckets + rh_hash(rh, region), hash_list)
-		if (reg->key == region)
-			return reg;
-
-	return NULL;
-}
-
-static void __rh_insert(struct region_hash *rh, struct region *reg)
-{
-	unsigned int h = rh_hash(rh, reg->key);
-	list_add(&reg->hash_list, rh->buckets + h);
-}
-
-static struct region *__rh_alloc(struct region_hash *rh, region_t region)
+static void wakeup_all_recovery_waiters(void *context)
 {
-	struct region *reg, *nreg;
-
-	read_unlock(&rh->hash_lock);
-	nreg = mempool_alloc(rh->region_pool, GFP_ATOMIC);
-	if (unlikely(!nreg))
-		nreg = kmalloc(sizeof(struct region), GFP_NOIO);
-	nreg->state = rh->log->type->in_sync(rh->log, region, 1) ?
-		RH_CLEAN : RH_NOSYNC;
-	nreg->rh = rh;
-	nreg->key = region;
-
-	INIT_LIST_HEAD(&nreg->list);
-
-	atomic_set(&nreg->pending, 0);
-	bio_list_init(&nreg->delayed_bios);
-	write_lock_irq(&rh->hash_lock);
-
-	reg = __rh_lookup(rh, region);
-	if (reg)
-		/* we lost the race */
-		mempool_free(nreg, rh->region_pool);
-
-	else {
-		__rh_insert(rh, nreg);
-		if (nreg->state == RH_CLEAN) {
-			spin_lock(&rh->region_lock);
-			list_add(&nreg->list, &rh->clean_regions);
-			spin_unlock(&rh->region_lock);
-		}
-		reg = nreg;
-	}
-	write_unlock_irq(&rh->hash_lock);
-	read_lock(&rh->hash_lock);
-
-	return reg;
+	wake_up_all(&_kmirrord_recovery_stopped);
 }
 
-static inline struct region *__rh_find(struct region_hash *rh, region_t region)
-{
-	struct region *reg;
-
-	reg = __rh_lookup(rh, region);
-	if (!reg)
-		reg = __rh_alloc(rh, region);
-
-	return reg;
-}
-
-static int rh_state(struct region_hash *rh, region_t region, int may_block)
-{
-	int r;
-	struct region *reg;
-
-	read_lock(&rh->hash_lock);
-	reg = __rh_lookup(rh, region);
-	read_unlock(&rh->hash_lock);
-
-	if (reg)
-		return reg->state;
-
-	/*
-	 * The region wasn't in the hash, so we fall back to the
-	 * dirty log.
-	 */
-	r = rh->log->type->in_sync(rh->log, region, may_block);
-
-	/*
-	 * Any error from the dirty log (eg. -EWOULDBLOCK) gets
-	 * taken as a RH_NOSYNC
-	 */
-	return r == 1 ? RH_CLEAN : RH_NOSYNC;
-}
-
-static inline int rh_in_sync(struct region_hash *rh,
-			     region_t region, int may_block)
-{
-	int state = rh_state(rh, region, may_block);
-	return state == RH_CLEAN || state == RH_DIRTY;
-}
-
-static void dispatch_bios(struct mirror_set *ms, struct bio_list *bio_list)
-{
-	struct bio *bio;
-
-	while ((bio = bio_list_pop(bio_list))) {
-		queue_bio(ms, bio, WRITE);
-	}
-}
-
-static void complete_resync_work(struct region *reg, int success)
-{
-	struct region_hash *rh = reg->rh;
-
-	rh->log->type->set_region_sync(rh->log, reg->key, success);
-
-	/*
-	 * Dispatch the bios before we call 'wake_up_all'.
-	 * This is important because if we are suspending,
-	 * we want to know that recovery is complete and
-	 * the work queue is flushed.  If we wake_up_all
-	 * before we dispatch_bios (queue bios and call wake()),
-	 * then we risk suspending before the work queue
-	 * has been properly flushed.
-	 */
-	dispatch_bios(rh->ms, &reg->delayed_bios);
-	if (atomic_dec_and_test(&rh->recovery_in_flight))
-		wake_up_all(&_kmirrord_recovery_stopped);
-	up(&rh->recovery_count);
-}
-
-static void rh_update_states(struct region_hash *rh)
-{
-	struct region *reg, *next;
-
-	LIST_HEAD(clean);
-	LIST_HEAD(recovered);
-	LIST_HEAD(failed_recovered);
-
-	/*
-	 * Quickly grab the lists.
-	 */
-	write_lock_irq(&rh->hash_lock);
-	spin_lock(&rh->region_lock);
-	if (!list_empty(&rh->clean_regions)) {
-		list_splice_init(&rh->clean_regions, &clean);
-
-		list_for_each_entry(reg, &clean, list)
-			list_del(&reg->hash_list);
-	}
-
-	if (!list_empty(&rh->recovered_regions)) {
-		list_splice_init(&rh->recovered_regions, &recovered);
-
-		list_for_each_entry (reg, &recovered, list)
-			list_del(&reg->hash_list);
-	}
-
-	if (!list_empty(&rh->failed_recovered_regions)) {
-		list_splice_init(&rh->failed_recovered_regions,
-				 &failed_recovered);
-
-		list_for_each_entry(reg, &failed_recovered, list)
-			list_del(&reg->hash_list);
-	}
-
-	spin_unlock(&rh->region_lock);
-	write_unlock_irq(&rh->hash_lock);
-
-	/*
-	 * All the regions on the recovered and clean lists have
-	 * now been pulled out of the system, so no need to do
-	 * any more locking.
-	 */
-	list_for_each_entry_safe (reg, next, &recovered, list) {
-		rh->log->type->clear_region(rh->log, reg->key);
-		complete_resync_work(reg, 1);
-		mempool_free(reg, rh->region_pool);
-	}
-
-	list_for_each_entry_safe(reg, next, &failed_recovered, list) {
-		complete_resync_work(reg, errors_handled(rh->ms) ? 0 : 1);
-		mempool_free(reg, rh->region_pool);
-	}
-
-	list_for_each_entry_safe(reg, next, &clean, list) {
-		rh->log->type->clear_region(rh->log, reg->key);
-		mempool_free(reg, rh->region_pool);
-	}
-
-	rh->log->type->flush(rh->log);
-}
-
-static void rh_inc(struct region_hash *rh, region_t region)
-{
-	struct region *reg;
-
-	read_lock(&rh->hash_lock);
-	reg = __rh_find(rh, region);
-
-	spin_lock_irq(&rh->region_lock);
-	atomic_inc(&reg->pending);
-
-	if (reg->state == RH_CLEAN) {
-		reg->state = RH_DIRTY;
-		list_del_init(&reg->list);	/* take off the clean list */
-		spin_unlock_irq(&rh->region_lock);
-
-		rh->log->type->mark_region(rh->log, reg->key);
-	} else
-		spin_unlock_irq(&rh->region_lock);
-
-
-	read_unlock(&rh->hash_lock);
-}
-
-static void rh_inc_pending(struct region_hash *rh, struct bio_list *bios)
-{
-	struct bio *bio;
-
-	for (bio = bios->head; bio; bio = bio->bi_next)
-		rh_inc(rh, bio_to_region(rh, bio));
-}
-
-static void rh_dec(struct region_hash *rh, region_t region)
+static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw)
 {
 	unsigned long flags;
-	struct region *reg;
 	int should_wake = 0;
+	struct bio_list *bl;
 
-	read_lock(&rh->hash_lock);
-	reg = __rh_lookup(rh, region);
-	read_unlock(&rh->hash_lock);
-
-	spin_lock_irqsave(&rh->region_lock, flags);
-	if (atomic_dec_and_test(&reg->pending)) {
-		/*
-		 * There is no pending I/O for this region.
-		 * We can move the region to corresponding list for next action.
-		 * At this point, the region is not yet connected to any list.
-		 *
-		 * If the state is RH_NOSYNC, the region should be kept off
-		 * from clean list.
-		 * The hash entry for RH_NOSYNC will remain in memory
-		 * until the region is recovered or the map is reloaded.
-		 */
-
-		/* do nothing for RH_NOSYNC */
-		if (reg->state == RH_RECOVERING) {
-			list_add_tail(&reg->list, &rh->quiesced_regions);
-		} else if (reg->state == RH_DIRTY) {
-			reg->state = RH_CLEAN;
-			list_add(&reg->list, &rh->clean_regions);
-		}
-		should_wake = 1;
-	}
-	spin_unlock_irqrestore(&rh->region_lock, flags);
+	bl = (rw == WRITE) ? &ms->writes : &ms->reads;
+	spin_lock_irqsave(&ms->lock, flags);
+	should_wake = !(bl->head);
+	bio_list_add(bl, bio);
+	spin_unlock_irqrestore(&ms->lock, flags);
 
 	if (should_wake)
-		wake(rh->ms);
-}
-
-/*
- * Starts quiescing a region in preparation for recovery.
- */
-static int __rh_recovery_prepare(struct region_hash *rh)
-{
-	int r;
-	struct region *reg;
-	region_t region;
-
-	/*
-	 * Ask the dirty log what's next.
-	 */
-	r = rh->log->type->get_resync_work(rh->log, &region);
-	if (r <= 0)
-		return r;
-
-	/*
-	 * Get this region, and start it quiescing by setting the
-	 * recovering flag.
-	 */
-	read_lock(&rh->hash_lock);
-	reg = __rh_find(rh, region);
-	read_unlock(&rh->hash_lock);
-
-	spin_lock_irq(&rh->region_lock);
-	reg->state = RH_RECOVERING;
-
-	/* Already quiesced ? */
-	if (atomic_read(&reg->pending))
-		list_del_init(&reg->list);
-	else
-		list_move(&reg->list, &rh->quiesced_regions);
-
-	spin_unlock_irq(&rh->region_lock);
-
-	return 1;
-}
-
-static void rh_recovery_prepare(struct region_hash *rh)
-{
-	/* Extra reference to avoid race with rh_stop_recovery */
-	atomic_inc(&rh->recovery_in_flight);
-
-	while (!down_trylock(&rh->recovery_count)) {
-		atomic_inc(&rh->recovery_in_flight);
-		if (__rh_recovery_prepare(rh) <= 0) {
-			atomic_dec(&rh->recovery_in_flight);
-			up(&rh->recovery_count);
-			break;
-		}
-	}
-
-	/* Drop the extra reference */
-	if (atomic_dec_and_test(&rh->recovery_in_flight))
-		wake_up_all(&_kmirrord_recovery_stopped);
-}
-
-/*
- * Returns any quiesced regions.
- */
-static struct region *rh_recovery_start(struct region_hash *rh)
-{
-	struct region *reg = NULL;
-
-	spin_lock_irq(&rh->region_lock);
-	if (!list_empty(&rh->quiesced_regions)) {
-		reg = list_entry(rh->quiesced_regions.next,
-				 struct region, list);
-		list_del_init(&reg->list);	/* remove from the quiesced list */
-	}
-	spin_unlock_irq(&rh->region_lock);
-
-	return reg;
-}
-
-static void rh_recovery_end(struct region *reg, int success)
-{
-	struct region_hash *rh = reg->rh;
-
-	spin_lock_irq(&rh->region_lock);
-	if (success)
-		list_add(&reg->list, &reg->rh->recovered_regions);
-	else {
-		reg->state = RH_NOSYNC;
-		list_add(&reg->list, &reg->rh->failed_recovered_regions);
-	}
-	spin_unlock_irq(&rh->region_lock);
-
-	wake(rh->ms);
-}
-
-static int rh_flush(struct region_hash *rh)
-{
-	return rh->log->type->flush(rh->log);
+		wakeup_mirrord(ms);
 }
 
-static void rh_delay(struct region_hash *rh, struct bio *bio)
+static void dispatch_bios(void *context, struct bio_list *bio_list)
 {
-	struct region *reg;
-
-	read_lock(&rh->hash_lock);
-	reg = __rh_find(rh, bio_to_region(rh, bio));
-	bio_list_add(&reg->delayed_bios, bio);
-	read_unlock(&rh->hash_lock);
-}
-
-static void rh_stop_recovery(struct region_hash *rh)
-{
-	int i;
-
-	/* wait for any recovering regions */
-	for (i = 0; i < MAX_RECOVERY; i++)
-		down(&rh->recovery_count);
-}
-
-static void rh_start_recovery(struct region_hash *rh)
-{
-	int i;
-
-	for (i = 0; i < MAX_RECOVERY; i++)
-		up(&rh->recovery_count);
+	struct mirror_set *ms = context;
+	struct bio *bio;
 
-	wake(rh->ms);
+	while ((bio = bio_list_pop(bio_list)))
+		queue_bio(ms, bio, WRITE);
 }
 
-#define MIN_READ_RECORDS 20
-struct dm_raid1_read_record {
+struct dm_raid1_bio_record {
 	struct mirror *m;
+	/* if details->bi_bdev == NULL, details were not saved */
 	struct dm_bio_details details;
+	region_t write_region;
 };
 
 /*
@@ -708,6 +181,17 @@ static void set_default_mirror(struct mirror *m)
 	atomic_set(&ms->default_mirror, m - m0);
 }
 
+static struct mirror *get_valid_mirror(struct mirror_set *ms)
+{
+	struct mirror *m;
+
+	for (m = ms->mirror; m < ms->mirror + ms->nr_mirrors; m++)
+		if (!atomic_read(&m->error_count))
+			return m;
+
+	return NULL;
+}
+
 /* fail_mirror
  * @m: mirror device to fail
  * @error_type: one of the enum's, DM_RAID1_*_ERROR
@@ -727,8 +211,7 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 	struct mirror_set *ms = m->ms;
 	struct mirror *new;
 
-	if (!errors_handled(ms))
-		return;
+	ms->leg_failure = 1;
 
 	/*
 	 * error_count is used for nothing more than a
@@ -740,6 +223,9 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 	if (test_and_set_bit(error_type, &m->error_type))
 		return;
 
+	if (!errors_handled(ms))
+		return;
+
 	if (m != get_default_mirror(ms))
 		goto out;
 
@@ -753,19 +239,50 @@ static void fail_mirror(struct mirror *m, enum dm_raid1_error error_type)
 		goto out;
 	}
 
-	for (new = ms->mirror; new < ms->mirror + ms->nr_mirrors; new++)
-		if (!atomic_read(&new->error_count)) {
-			set_default_mirror(new);
-			break;
-		}
-
-	if (unlikely(new == ms->mirror + ms->nr_mirrors))
+	new = get_valid_mirror(ms);
+	if (new)
+		set_default_mirror(new);
+	else
 		DMWARN("All sides of mirror have failed.");
 
 out:
 	schedule_work(&ms->trigger_event);
 }
 
+static int mirror_flush(struct dm_target *ti)
+{
+	struct mirror_set *ms = ti->private;
+	unsigned long error_bits;
+
+	unsigned int i;
+	struct dm_io_region io[ms->nr_mirrors];
+	struct mirror *m;
+	struct dm_io_request io_req = {
+		.bi_rw = WRITE_FLUSH,
+		.mem.type = DM_IO_KMEM,
+		.mem.ptr.addr = NULL,
+		.client = ms->io_client,
+	};
+
+	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++) {
+		io[i].bdev = m->dev->bdev;
+		io[i].sector = 0;
+		io[i].count = 0;
+	}
+
+	error_bits = -1;
+	dm_io(&io_req, ms->nr_mirrors, io, &error_bits);
+	if (unlikely(error_bits != 0)) {
+		for (i = 0; i < ms->nr_mirrors; i++)
+			if (test_bit(i, &error_bits))
+				fail_mirror(ms->mirror + i,
+					    DM_RAID1_FLUSH_ERROR);
+		return -EIO;
+	}
+
+	return 0;
+}
+
 /*-----------------------------------------------------------------
  * Recovery.
  *
@@ -776,8 +293,8 @@ out:
 static void recovery_complete(int read_err, unsigned long write_err,
 			      void *context)
 {
-	struct region *reg = (struct region *)context;
-	struct mirror_set *ms = reg->rh->ms;
+	struct dm_region *reg = context;
+	struct mirror_set *ms = dm_rh_region_context(reg);
 	int m, bit = 0;
 
 	if (read_err) {
@@ -803,31 +320,33 @@ static void recovery_complete(int read_err, unsigned long write_err,
 		}
 	}
 
-	rh_recovery_end(reg, !(read_err || write_err));
+	dm_rh_recovery_end(reg, !(read_err || write_err));
 }
 
-static int recover(struct mirror_set *ms, struct region *reg)
+static int recover(struct mirror_set *ms, struct dm_region *reg)
 {
 	int r;
-	unsigned int i;
+	unsigned i;
 	struct dm_io_region from, to[DM_KCOPYD_MAX_REGIONS], *dest;
 	struct mirror *m;
 	unsigned long flags = 0;
+	region_t key = dm_rh_get_region_key(reg);
+	sector_t region_size = dm_rh_get_region_size(ms->rh);
 
 	/* fill in the source */
 	m = get_default_mirror(ms);
 	from.bdev = m->dev->bdev;
-	from.sector = m->offset + region_to_sector(reg->rh, reg->key);
-	if (reg->key == (ms->nr_regions - 1)) {
+	from.sector = m->offset + dm_rh_region_to_sector(ms->rh, key);
+	if (key == (ms->nr_regions - 1)) {
 		/*
 		 * The final region may be smaller than
 		 * region_size.
 		 */
-		from.count = ms->ti->len & (reg->rh->region_size - 1);
+		from.count = ms->ti->len & (region_size - 1);
 		if (!from.count)
-			from.count = reg->rh->region_size;
+			from.count = region_size;
 	} else
-		from.count = reg->rh->region_size;
+		from.count = region_size;
 
 	/* fill in the destinations */
 	for (i = 0, dest = to; i < ms->nr_mirrors; i++) {
@@ -836,13 +355,15 @@ static int recover(struct mirror_set *ms, struct region *reg)
 
 		m = ms->mirror + i;
 		dest->bdev = m->dev->bdev;
-		dest->sector = m->offset + region_to_sector(reg->rh, reg->key);
+		dest->sector = m->offset + dm_rh_region_to_sector(ms->rh, key);
 		dest->count = from.count;
 		dest++;
 	}
 
 	/* hand to kcopyd */
-	set_bit(DM_KCOPYD_IGNORE_ERROR, &flags);
+	if (!errors_handled(ms))
+		set_bit(DM_KCOPYD_IGNORE_ERROR, &flags);
+
 	r = dm_kcopyd_copy(ms->kcopyd_client, &from, ms->nr_mirrors - 1, to,
 			   flags, recovery_complete, reg);
 
@@ -851,22 +372,22 @@ static int recover(struct mirror_set *ms, struct region *reg)
 
 static void do_recovery(struct mirror_set *ms)
 {
+	struct dm_region *reg;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
 	int r;
-	struct region *reg;
-	struct dm_dirty_log *log = ms->rh.log;
 
 	/*
 	 * Start quiescing some regions.
 	 */
-	rh_recovery_prepare(&ms->rh);
+	dm_rh_recovery_prepare(ms->rh);
 
 	/*
 	 * Copy any already quiesced regions.
 	 */
-	while ((reg = rh_recovery_start(&ms->rh))) {
+	while ((reg = dm_rh_recovery_start(ms->rh))) {
 		r = recover(ms, reg);
 		if (r)
-			rh_recovery_end(reg, 0);
+			dm_rh_recovery_end(reg, 0);
 	}
 
 	/*
@@ -907,10 +428,11 @@ static int default_ok(struct mirror *m)
 
 static int mirror_available(struct mirror_set *ms, struct bio *bio)
 {
-	region_t region = bio_to_region(&ms->rh, bio);
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
+	region_t region = dm_rh_bio_to_region(ms->rh, bio);
 
-	if (ms->rh.log->type->in_sync(ms->rh.log, region, 0))
-		return choose_mirror(ms,  bio->bi_sector) ? 1 : 0;
+	if (log->type->in_sync(log, region, 0))
+		return choose_mirror(ms,  bio->bi_iter.bi_sector) ? 1 : 0;
 
 	return 0;
 }
@@ -920,13 +442,15 @@ static int mirror_available(struct mirror_set *ms, struct bio *bio)
  */
 static sector_t map_sector(struct mirror *m, struct bio *bio)
 {
-	return m->offset + (bio->bi_sector - m->ms->ti->begin);
+	if (unlikely(!bio->bi_iter.bi_size))
+		return 0;
+	return m->offset + dm_target_offset(m->ms->ti, bio->bi_iter.bi_sector);
 }
 
 static void map_bio(struct mirror *m, struct bio *bio)
 {
 	bio->bi_bdev = m->dev->bdev;
-	bio->bi_sector = map_sector(m, bio);
+	bio->bi_iter.bi_sector = map_sector(m, bio);
 }
 
 static void map_region(struct dm_io_region *io, struct mirror *m,
@@ -934,7 +458,35 @@ static void map_region(struct dm_io_region *io, struct mirror *m,
 {
 	io->bdev = m->dev->bdev;
 	io->sector = map_sector(m, bio);
-	io->count = bio->bi_size >> 9;
+	io->count = bio_sectors(bio);
+}
+
+static void hold_bio(struct mirror_set *ms, struct bio *bio)
+{
+	/*
+	 * Lock is required to avoid race condition during suspend
+	 * process.
+	 */
+	spin_lock_irq(&ms->lock);
+
+	if (atomic_read(&ms->suspend)) {
+		spin_unlock_irq(&ms->lock);
+
+		/*
+		 * If device is suspended, complete the bio.
+		 */
+		if (dm_noflush_suspending(ms->ti))
+			bio_endio(bio, DM_ENDIO_REQUEUE);
+		else
+			bio_endio(bio, -EIO);
+		return;
+	}
+
+	/*
+	 * Hold bio until the suspend is complete.
+	 */
+	bio_list_add(&ms->holds, bio);
+	spin_unlock_irq(&ms->lock);
 }
 
 /*-----------------------------------------------------------------
@@ -974,8 +526,8 @@ static void read_async_bio(struct mirror *m, struct bio *bio)
 	struct dm_io_region io;
 	struct dm_io_request io_req = {
 		.bi_rw = READ,
-		.mem.type = DM_IO_BVEC,
-		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
+		.mem.type = DM_IO_BIO,
+		.mem.ptr.bio = bio,
 		.notify.fn = read_callback,
 		.notify.context = bio,
 		.client = m->ms->io_client,
@@ -983,7 +535,14 @@ static void read_async_bio(struct mirror *m, struct bio *bio)
 
 	map_region(&io, m, bio);
 	bio_set_m(bio, m);
-	(void) dm_io(&io_req, 1, &io, NULL);
+	BUG_ON(dm_io(&io_req, 1, &io, NULL));
+}
+
+static inline int region_in_sync(struct mirror_set *ms, region_t region,
+				 int may_block)
+{
+	int state = dm_rh_get_state(ms->rh, region, may_block);
+	return state == DM_RH_CLEAN || state == DM_RH_DIRTY;
 }
 
 static void do_reads(struct mirror_set *ms, struct bio_list *reads)
@@ -993,14 +552,14 @@ static void do_reads(struct mirror_set *ms, struct bio_list *reads)
 	struct mirror *m;
 
 	while ((bio = bio_list_pop(reads))) {
-		region = bio_to_region(&ms->rh, bio);
+		region = dm_rh_bio_to_region(ms->rh, bio);
 		m = get_default_mirror(ms);
 
 		/*
 		 * We can only read balance if the region is in sync.
 		 */
-		if (likely(rh_in_sync(&ms->rh, region, 1)))
-			m = choose_mirror(ms, bio->bi_sector);
+		if (likely(region_in_sync(ms, region, 1)))
+			m = choose_mirror(ms, bio->bi_iter.bi_sector);
 		else if (m && atomic_read(&m->error_count))
 			m = NULL;
 
@@ -1022,64 +581,12 @@ static void do_reads(struct mirror_set *ms, struct bio_list *reads)
  * NOSYNC:	increment pending, just write to the default mirror
  *---------------------------------------------------------------*/
 
-/* __bio_mark_nosync
- * @ms
- * @bio
- * @done
- * @error
- *
- * The bio was written on some mirror(s) but failed on other mirror(s).
- * We can successfully endio the bio but should avoid the region being
- * marked clean by setting the state RH_NOSYNC.
- *
- * This function is _not_ safe in interrupt context!
- */
-static void __bio_mark_nosync(struct mirror_set *ms,
-			      struct bio *bio, unsigned done, int error)
-{
-	unsigned long flags;
-	struct region_hash *rh = &ms->rh;
-	struct dm_dirty_log *log = ms->rh.log;
-	struct region *reg;
-	region_t region = bio_to_region(rh, bio);
-	int recovering = 0;
-
-	/* We must inform the log that the sync count has changed. */
-	log->type->set_region_sync(log, region, 0);
-	ms->in_sync = 0;
-
-	read_lock(&rh->hash_lock);
-	reg = __rh_find(rh, region);
-	read_unlock(&rh->hash_lock);
-
-	/* region hash entry should exist because write was in-flight */
-	BUG_ON(!reg);
-	BUG_ON(!list_empty(&reg->list));
-
-	spin_lock_irqsave(&rh->region_lock, flags);
-	/*
-	 * Possible cases:
-	 *   1) RH_DIRTY
-	 *   2) RH_NOSYNC: was dirty, other preceeding writes failed
-	 *   3) RH_RECOVERING: flushing pending writes
-	 * Either case, the region should have not been connected to list.
-	 */
-	recovering = (reg->state == RH_RECOVERING);
-	reg->state = RH_NOSYNC;
-	BUG_ON(!list_empty(&reg->list));
-	spin_unlock_irqrestore(&rh->region_lock, flags);
-
-	bio_endio(bio, error);
-	if (recovering)
-		complete_resync_work(reg, 0);
-}
 
 static void write_callback(unsigned long error, void *context)
 {
 	unsigned i, ret = 0;
 	struct bio *bio = (struct bio *) context;
 	struct mirror_set *ms;
-	int uptodate = 0;
 	int should_wake = 0;
 	unsigned long flags;
 
@@ -1092,36 +599,27 @@ static void write_callback(unsigned long error, void *context)
 	 * This way we handle both writes to SYNC and NOSYNC
 	 * regions with the same code.
 	 */
-	if (likely(!error))
-		goto out;
+	if (likely(!error)) {
+		bio_endio(bio, ret);
+		return;
+	}
 
 	for (i = 0; i < ms->nr_mirrors; i++)
 		if (test_bit(i, &error))
 			fail_mirror(ms->mirror + i, DM_RAID1_WRITE_ERROR);
-		else
-			uptodate = 1;
 
-	if (unlikely(!uptodate)) {
-		DMERR("All replicated volumes dead, failing I/O");
-		/* None of the writes succeeded, fail the I/O. */
-		ret = -EIO;
-	} else if (errors_handled(ms)) {
-		/*
-		 * Need to raise event.  Since raising
-		 * events can block, we need to do it in
-		 * the main thread.
-		 */
-		spin_lock_irqsave(&ms->lock, flags);
-		if (!ms->failures.head)
-			should_wake = 1;
-		bio_list_add(&ms->failures, bio);
-		spin_unlock_irqrestore(&ms->lock, flags);
-		if (should_wake)
-			wake(ms);
-		return;
-	}
-out:
-	bio_endio(bio, ret);
+	/*
+	 * Need to raise event.  Since raising
+	 * events can block, we need to do it in
+	 * the main thread.
+	 */
+	spin_lock_irqsave(&ms->lock, flags);
+	if (!ms->failures.head)
+		should_wake = 1;
+	bio_list_add(&ms->failures, bio);
+	spin_unlock_irqrestore(&ms->lock, flags);
+	if (should_wake)
+		wakeup_mirrord(ms);
 }
 
 static void do_write(struct mirror_set *ms, struct bio *bio)
@@ -1130,14 +628,20 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	struct dm_io_region io[ms->nr_mirrors], *dest = io;
 	struct mirror *m;
 	struct dm_io_request io_req = {
-		.bi_rw = WRITE,
-		.mem.type = DM_IO_BVEC,
-		.mem.ptr.bvec = bio->bi_io_vec + bio->bi_idx,
+		.bi_rw = WRITE | (bio->bi_rw & WRITE_FLUSH_FUA),
+		.mem.type = DM_IO_BIO,
+		.mem.ptr.bio = bio,
 		.notify.fn = write_callback,
 		.notify.context = bio,
 		.client = ms->io_client,
 	};
 
+	if (bio->bi_rw & REQ_DISCARD) {
+		io_req.bi_rw |= REQ_DISCARD;
+		io_req.mem.type = DM_IO_KMEM;
+		io_req.mem.ptr.addr = NULL;
+	}
+
 	for (i = 0, m = ms->mirror; i < ms->nr_mirrors; i++, m++)
 		map_region(dest++, m, bio);
 
@@ -1147,7 +651,7 @@ static void do_write(struct mirror_set *ms, struct bio *bio)
 	 */
 	bio_set_m(bio, get_default_mirror(ms));
 
-	(void) dm_io(&io_req, ms->nr_mirrors, io, NULL);
+	BUG_ON(dm_io(&io_req, ms->nr_mirrors, io, NULL));
 }
 
 static void do_writes(struct mirror_set *ms, struct bio_list *writes)
@@ -1155,6 +659,9 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	int state;
 	struct bio *bio;
 	struct bio_list sync, nosync, recover, *this_list = NULL;
+	struct bio_list requeue;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
+	region_t region;
 
 	if (!writes->head)
 		return;
@@ -1165,20 +672,35 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	bio_list_init(&sync);
 	bio_list_init(&nosync);
 	bio_list_init(&recover);
+	bio_list_init(&requeue);
 
 	while ((bio = bio_list_pop(writes))) {
-		state = rh_state(&ms->rh, bio_to_region(&ms->rh, bio), 1);
+		if ((bio->bi_rw & REQ_FLUSH) ||
+		    (bio->bi_rw & REQ_DISCARD)) {
+			bio_list_add(&sync, bio);
+			continue;
+		}
+
+		region = dm_rh_bio_to_region(ms->rh, bio);
+
+		if (log->type->is_remote_recovering &&
+		    log->type->is_remote_recovering(log, region)) {
+			bio_list_add(&requeue, bio);
+			continue;
+		}
+
+		state = dm_rh_get_state(ms->rh, region, 1);
 		switch (state) {
-		case RH_CLEAN:
-		case RH_DIRTY:
+		case DM_RH_CLEAN:
+		case DM_RH_DIRTY:
 			this_list = &sync;
 			break;
 
-		case RH_NOSYNC:
+		case DM_RH_NOSYNC:
 			this_list = &nosync;
 			break;
 
-		case RH_RECOVERING:
+		case DM_RH_RECOVERING:
 			this_list = &recover;
 			break;
 		}
@@ -1187,32 +709,56 @@ static void do_writes(struct mirror_set *ms, struct bio_list *writes)
 	}
 
 	/*
+	 * Add bios that are delayed due to remote recovery
+	 * back on to the write queue
+	 */
+	if (unlikely(requeue.head)) {
+		spin_lock_irq(&ms->lock);
+		bio_list_merge(&ms->writes, &requeue);
+		spin_unlock_irq(&ms->lock);
+		delayed_wake(ms);
+	}
+
+	/*
 	 * Increment the pending counts for any regions that will
 	 * be written to (writes to recover regions are going to
 	 * be delayed).
 	 */
-	rh_inc_pending(&ms->rh, &sync);
-	rh_inc_pending(&ms->rh, &nosync);
-	ms->log_failure = rh_flush(&ms->rh) ? 1 : 0;
+	dm_rh_inc_pending(ms->rh, &sync);
+	dm_rh_inc_pending(ms->rh, &nosync);
+
+	/*
+	 * If the flush fails on a previous call and succeeds here,
+	 * we must not reset the log_failure variable.  We need
+	 * userspace interaction to do that.
+	 */
+	ms->log_failure = dm_rh_flush(ms->rh) ? 1 : ms->log_failure;
 
 	/*
 	 * Dispatch io.
 	 */
-	if (unlikely(ms->log_failure)) {
+	if (unlikely(ms->log_failure) && errors_handled(ms)) {
 		spin_lock_irq(&ms->lock);
 		bio_list_merge(&ms->failures, &sync);
 		spin_unlock_irq(&ms->lock);
-		wake(ms);
+		wakeup_mirrord(ms);
 	} else
 		while ((bio = bio_list_pop(&sync)))
 			do_write(ms, bio);
 
 	while ((bio = bio_list_pop(&recover)))
-		rh_delay(&ms->rh, bio);
+		dm_rh_delay(ms->rh, bio);
 
 	while ((bio = bio_list_pop(&nosync))) {
-		map_bio(get_default_mirror(ms), bio);
-		generic_make_request(bio);
+		if (unlikely(ms->leg_failure) && errors_handled(ms)) {
+			spin_lock_irq(&ms->lock);
+			bio_list_add(&ms->failures, bio);
+			spin_unlock_irq(&ms->lock);
+			wakeup_mirrord(ms);
+		} else {
+			map_bio(get_default_mirror(ms), bio);
+			generic_make_request(bio);
+		}
 	}
 }
 
@@ -1220,18 +766,12 @@ static void do_failures(struct mirror_set *ms, struct bio_list *failures)
 {
 	struct bio *bio;
 
-	if (!failures->head)
-		return;
-
-	if (!ms->log_failure) {
-		while ((bio = bio_list_pop(failures)))
-			__bio_mark_nosync(ms, bio, bio->bi_size, 0);
+	if (likely(!failures->head))
 		return;
-	}
 
 	/*
 	 * If the log has failed, unattempted writes are being
-	 * put on the failures list.  We can't issue those writes
+	 * put on the holds list.  We can't issue those writes
 	 * until a log has been marked, so we must store them.
 	 *
 	 * If a 'noflush' suspend is in progress, we can requeue
@@ -1246,23 +786,27 @@ static void do_failures(struct mirror_set *ms, struct bio_list *failures)
 	 * for us to treat them the same and requeue them
 	 * as well.
 	 */
-	if (dm_noflush_suspending(ms->ti)) {
-		while ((bio = bio_list_pop(failures)))
-			bio_endio(bio, DM_ENDIO_REQUEUE);
-		return;
-	}
+	while ((bio = bio_list_pop(failures))) {
+		if (!ms->log_failure) {
+			ms->in_sync = 0;
+			dm_rh_mark_nosync(ms->rh, bio);
+		}
 
-	if (atomic_read(&ms->suspend)) {
-		while ((bio = bio_list_pop(failures)))
+		/*
+		 * If all the legs are dead, fail the I/O.
+		 * If we have been told to handle errors, hold the bio
+		 * and wait for userspace to deal with the problem.
+		 * Otherwise pretend that the I/O succeeded. (This would
+		 * be wrong if the failed leg returned after reboot and
+		 * got replicated back to the good legs.)
+		 */
+		if (!get_valid_mirror(ms))
 			bio_endio(bio, -EIO);
-		return;
+		else if (errors_handled(ms))
+			hold_bio(ms, bio);
+		else
+			bio_endio(bio, 0);
 	}
-
-	spin_lock_irq(&ms->lock);
-	bio_list_merge(&ms->failures, failures);
-	spin_unlock_irq(&ms->lock);
-
-	delayed_wake(ms);
 }
 
 static void trigger_event(struct work_struct *work)
@@ -1278,8 +822,8 @@ static void trigger_event(struct work_struct *work)
  *---------------------------------------------------------------*/
 static void do_mirror(struct work_struct *work)
 {
-	struct mirror_set *ms =container_of(work, struct mirror_set,
-					    kmirrord_work);
+	struct mirror_set *ms = container_of(work, struct mirror_set,
+					     kmirrord_work);
 	struct bio_list reads, writes, failures;
 	unsigned long flags;
 
@@ -1292,16 +836,13 @@ static void do_mirror(struct work_struct *work)
 	bio_list_init(&ms->failures);
 	spin_unlock_irqrestore(&ms->lock, flags);
 
-	rh_update_states(&ms->rh);
+	dm_rh_update_states(ms->rh, errors_handled(ms));
 	do_recovery(ms);
 	do_reads(ms, &reads);
 	do_writes(ms, &writes);
 	do_failures(ms, &failures);
-
-	dm_table_unplug_all(ms->ti->table);
 }
 
-
 /*-----------------------------------------------------------------
  * Target functions
  *---------------------------------------------------------------*/
@@ -1313,9 +854,6 @@ static struct mirror_set *alloc_context(unsigned int nr_mirrors,
 	size_t len;
 	struct mirror_set *ms = NULL;
 
-	if (array_too_big(sizeof(*ms), sizeof(ms->mirror[0]), nr_mirrors))
-		return NULL;
-
 	len = sizeof(*ms) + (sizeof(ms->mirror[0]) * nr_mirrors);
 
 	ms = kzalloc(len, GFP_KERNEL);
@@ -1325,36 +863,34 @@ static struct mirror_set *alloc_context(unsigned int nr_mirrors,
 	}
 
 	spin_lock_init(&ms->lock);
+	bio_list_init(&ms->reads);
+	bio_list_init(&ms->writes);
+	bio_list_init(&ms->failures);
+	bio_list_init(&ms->holds);
 
 	ms->ti = ti;
 	ms->nr_mirrors = nr_mirrors;
 	ms->nr_regions = dm_sector_div_up(ti->len, region_size);
 	ms->in_sync = 0;
 	ms->log_failure = 0;
+	ms->leg_failure = 0;
 	atomic_set(&ms->suspend, 0);
 	atomic_set(&ms->default_mirror, DEFAULT_MIRROR);
 
-	len = sizeof(struct dm_raid1_read_record);
-	ms->read_record_pool = mempool_create_kmalloc_pool(MIN_READ_RECORDS,
-							   len);
-	if (!ms->read_record_pool) {
-		ti->error = "Error creating mirror read_record_pool";
-		kfree(ms);
-		return NULL;
-	}
-
-	ms->io_client = dm_io_client_create(DM_IO_PAGES);
+	ms->io_client = dm_io_client_create();
 	if (IS_ERR(ms->io_client)) {
 		ti->error = "Error creating dm_io client";
-		mempool_destroy(ms->read_record_pool);
 		kfree(ms);
  		return NULL;
 	}
 
-	if (rh_init(&ms->rh, ms, dl, region_size, ms->nr_regions)) {
+	ms->rh = dm_region_hash_create(ms, dispatch_bios, wakeup_mirrord,
+				       wakeup_all_recovery_waiters,
+				       ms->ti->begin, MAX_RECOVERY,
+				       dl, region_size, ms->nr_regions);
+	if (IS_ERR(ms->rh)) {
 		ti->error = "Error creating dirty region hash";
 		dm_io_client_destroy(ms->io_client);
-		mempool_destroy(ms->read_record_pool);
 		kfree(ms);
 		return NULL;
 	}
@@ -1369,29 +905,22 @@ static void free_context(struct mirror_set *ms, struct dm_target *ti,
 		dm_put_device(ti, ms->mirror[m].dev);
 
 	dm_io_client_destroy(ms->io_client);
-	rh_exit(&ms->rh);
-	mempool_destroy(ms->read_record_pool);
+	dm_region_hash_destroy(ms->rh);
 	kfree(ms);
 }
 
-static inline int _check_region_size(struct dm_target *ti, uint32_t size)
-{
-	return !(size % (PAGE_SIZE >> 9) || !is_power_of_2(size) ||
-		 size > ti->len);
-}
-
 static int get_mirror(struct mirror_set *ms, struct dm_target *ti,
 		      unsigned int mirror, char **argv)
 {
 	unsigned long long offset;
+	char dummy;
 
-	if (sscanf(argv[1], "%llu", &offset) != 1) {
+	if (sscanf(argv[1], "%llu%c", &offset, &dummy) != 1) {
 		ti->error = "Invalid offset";
 		return -EINVAL;
 	}
 
-	if (dm_get_device(ti, argv[0], offset, ti->len,
-			  dm_table_get_mode(ti->table),
+	if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table),
 			  &ms->mirror[mirror].dev)) {
 		ti->error = "Device lookup failure";
 		return -ENXIO;
@@ -1409,18 +938,19 @@ static int get_mirror(struct mirror_set *ms, struct dm_target *ti,
  * Create dirty log: log_type #log_params <log_params>
  */
 static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
-					  unsigned int argc, char **argv,
-					  unsigned int *args_used)
+					     unsigned argc, char **argv,
+					     unsigned *args_used)
 {
-	unsigned int param_count;
+	unsigned param_count;
 	struct dm_dirty_log *dl;
+	char dummy;
 
 	if (argc < 2) {
 		ti->error = "Insufficient mirror log arguments";
 		return NULL;
 	}
 
-	if (sscanf(argv[1], "%u", &param_count) != 1) {
+	if (sscanf(argv[1], "%u%c", &param_count, &dummy) != 1) {
 		ti->error = "Invalid mirror log argument count";
 		return NULL;
 	}
@@ -1432,18 +962,13 @@ static struct dm_dirty_log *create_dirty_log(struct dm_target *ti,
 		return NULL;
 	}
 
-	dl = dm_dirty_log_create(argv[0], ti, param_count, argv + 2);
+	dl = dm_dirty_log_create(argv[0], ti, mirror_flush, param_count,
+				 argv + 2);
 	if (!dl) {
 		ti->error = "Error creating mirror dirty log";
 		return NULL;
 	}
 
-	if (!_check_region_size(ti, dl->type->get_region_size(dl))) {
-		ti->error = "Invalid region size";
-		dm_dirty_log_destroy(dl);
-		return NULL;
-	}
-
 	return dl;
 }
 
@@ -1452,13 +977,14 @@ static int parse_features(struct mirror_set *ms, unsigned argc, char **argv,
 {
 	unsigned num_features;
 	struct dm_target *ti = ms->ti;
+	char dummy;
 
 	*args_used = 0;
 
 	if (!argc)
 		return 0;
 
-	if (sscanf(argv[0], "%u", &num_features) != 1) {
+	if (sscanf(argv[0], "%u%c", &num_features, &dummy) != 1) {
 		ti->error = "Invalid number of features";
 		return -EINVAL;
 	}
@@ -1502,6 +1028,7 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	unsigned int nr_mirrors, m, args_used;
 	struct mirror_set *ms;
 	struct dm_dirty_log *dl;
+	char dummy;
 
 	dl = create_dirty_log(ti, argc, argv, &args_used);
 	if (!dl)
@@ -1510,7 +1037,7 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	argv += args_used;
 	argc -= args_used;
 
-	if (!argc || sscanf(argv[0], "%u", &nr_mirrors) != 1 ||
+	if (!argc || sscanf(argv[0], "%u%c", &nr_mirrors, &dummy) != 1 ||
 	    nr_mirrors < 2 || nr_mirrors > DM_KCOPYD_MAX_REGIONS + 1) {
 		ti->error = "Invalid number of mirrors";
 		dm_dirty_log_destroy(dl);
@@ -1543,9 +1070,17 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	}
 
 	ti->private = ms;
- 	ti->split_io = ms->rh.region_size;
 
-	ms->kmirrord_wq = create_singlethread_workqueue("kmirrord");
+	r = dm_set_target_max_io_len(ti, dm_rh_get_region_size(ms->rh));
+	if (r)
+		goto err_free_context;
+
+	ti->num_flush_bios = 1;
+	ti->num_discard_bios = 1;
+	ti->per_bio_data_size = sizeof(struct dm_raid1_bio_record);
+	ti->discard_zeroes_data_unsupported = true;
+
+	ms->kmirrord_wq = alloc_workqueue("kmirrord", WQ_MEM_RECLAIM, 0);
 	if (!ms->kmirrord_wq) {
 		DMERR("couldn't start kmirrord");
 		r = -ENOMEM;
@@ -1578,11 +1113,13 @@ static int mirror_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		goto err_destroy_wq;
 	}
 
-	r = dm_kcopyd_client_create(DM_IO_PAGES, &ms->kcopyd_client);
-	if (r)
+	ms->kcopyd_client = dm_kcopyd_client_create(&dm_kcopyd_throttle);
+	if (IS_ERR(ms->kcopyd_client)) {
+		r = PTR_ERR(ms->kcopyd_client);
 		goto err_destroy_wq;
+	}
 
-	wake(ms);
+	wakeup_mirrord(ms);
 	return 0;
 
 err_destroy_wq:
@@ -1598,47 +1135,34 @@ static void mirror_dtr(struct dm_target *ti)
 
 	del_timer_sync(&ms->timer);
 	flush_workqueue(ms->kmirrord_wq);
+	flush_work(&ms->trigger_event);
 	dm_kcopyd_client_destroy(ms->kcopyd_client);
 	destroy_workqueue(ms->kmirrord_wq);
 	free_context(ms, ti, ms->nr_mirrors);
 }
 
-static void queue_bio(struct mirror_set *ms, struct bio *bio, int rw)
-{
-	unsigned long flags;
-	int should_wake = 0;
-	struct bio_list *bl;
-
-	bl = (rw == WRITE) ? &ms->writes : &ms->reads;
-	spin_lock_irqsave(&ms->lock, flags);
-	should_wake = !(bl->head);
-	bio_list_add(bl, bio);
-	spin_unlock_irqrestore(&ms->lock, flags);
-
-	if (should_wake)
-		wake(ms);
-}
-
 /*
  * Mirror mapping function
  */
-static int mirror_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static int mirror_map(struct dm_target *ti, struct bio *bio)
 {
 	int r, rw = bio_rw(bio);
 	struct mirror *m;
 	struct mirror_set *ms = ti->private;
-	struct dm_raid1_read_record *read_record = NULL;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
+	struct dm_raid1_bio_record *bio_record =
+	  dm_per_bio_data(bio, sizeof(struct dm_raid1_bio_record));
+
+	bio_record->details.bi_bdev = NULL;
 
 	if (rw == WRITE) {
 		/* Save region for mirror_end_io() handler */
-		map_context->ll = bio_to_region(&ms->rh, bio);
+		bio_record->write_region = dm_rh_bio_to_region(ms->rh, bio);
 		queue_bio(ms, bio, rw);
 		return DM_MAPIO_SUBMITTED;
 	}
 
-	r = ms->rh.log->type->in_sync(ms->rh.log,
-				      bio_to_region(&ms->rh, bio), 0);
+	r = log->type->in_sync(log, dm_rh_bio_to_region(ms->rh, bio), 0);
 	if (r < 0 && r != -EWOULDBLOCK)
 		return r;
 
@@ -1657,47 +1181,44 @@ static int mirror_map(struct dm_target *ti, struct bio *bio,
 	 * The region is in-sync and we can perform reads directly.
 	 * Store enough information so we can retry if it fails.
 	 */
-	m = choose_mirror(ms, bio->bi_sector);
+	m = choose_mirror(ms, bio->bi_iter.bi_sector);
 	if (unlikely(!m))
 		return -EIO;
 
-	read_record = mempool_alloc(ms->read_record_pool, GFP_NOIO);
-	if (likely(read_record)) {
-		dm_bio_record(&read_record->details, bio);
-		map_context->ptr = read_record;
-		read_record->m = m;
-	}
+	dm_bio_record(&bio_record->details, bio);
+	bio_record->m = m;
 
 	map_bio(m, bio);
 
 	return DM_MAPIO_REMAPPED;
 }
 
-static int mirror_end_io(struct dm_target *ti, struct bio *bio,
-			 int error, union map_info *map_context)
+static int mirror_end_io(struct dm_target *ti, struct bio *bio, int error)
 {
 	int rw = bio_rw(bio);
 	struct mirror_set *ms = (struct mirror_set *) ti->private;
 	struct mirror *m = NULL;
 	struct dm_bio_details *bd = NULL;
-	struct dm_raid1_read_record *read_record = map_context->ptr;
+	struct dm_raid1_bio_record *bio_record =
+	  dm_per_bio_data(bio, sizeof(struct dm_raid1_bio_record));
 
 	/*
 	 * We need to dec pending if this was a write.
 	 */
 	if (rw == WRITE) {
-		rh_dec(&ms->rh, map_context->ll);
+		if (!(bio->bi_rw & (REQ_FLUSH | REQ_DISCARD)))
+			dm_rh_dec(ms->rh, bio_record->write_region);
 		return error;
 	}
 
 	if (error == -EOPNOTSUPP)
 		goto out;
 
-	if ((error == -EWOULDBLOCK) && bio_rw_ahead(bio))
+	if ((error == -EWOULDBLOCK) && (bio->bi_rw & REQ_RAHEAD))
 		goto out;
 
 	if (unlikely(error)) {
-		if (!read_record) {
+		if (!bio_record->details.bi_bdev) {
 			/*
 			 * There wasn't enough memory to record necessary
 			 * information for a retry or there was no other
@@ -1707,7 +1228,7 @@ static int mirror_end_io(struct dm_target *ti, struct bio *bio,
 			return -EIO;
 		}
 
-		m = read_record->m;
+		m = bio_record->m;
 
 		DMERR("Mirror read failed from %s. Trying alternative device.",
 		      m->dev->name);
@@ -1719,22 +1240,21 @@ static int mirror_end_io(struct dm_target *ti, struct bio *bio,
 		 * mirror.
 		 */
 		if (default_ok(m) || mirror_available(ms, bio)) {
-			bd = &read_record->details;
+			bd = &bio_record->details;
 
 			dm_bio_restore(bd, bio);
-			mempool_free(read_record, ms->read_record_pool);
-			map_context->ptr = NULL;
+			bio_record->details.bi_bdev = NULL;
+
+			atomic_inc(&bio->bi_remaining);
+
 			queue_bio(ms, bio, rw);
-			return 1;
+			return DM_ENDIO_INCOMPLETE;
 		}
 		DMERR("All replicated volumes dead, failing I/O");
 	}
 
 out:
-	if (read_record) {
-		mempool_free(read_record, ms->read_record_pool);
-		map_context->ptr = NULL;
-	}
+	bio_record->details.bi_bdev = NULL;
 
 	return error;
 }
@@ -1742,18 +1262,35 @@ out:
 static void mirror_presuspend(struct dm_target *ti)
 {
 	struct mirror_set *ms = (struct mirror_set *) ti->private;
-	struct dm_dirty_log *log = ms->rh.log;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
+
+	struct bio_list holds;
+	struct bio *bio;
 
 	atomic_set(&ms->suspend, 1);
 
 	/*
+	 * Process bios in the hold list to start recovery waiting
+	 * for bios in the hold list. After the process, no bio has
+	 * a chance to be added in the hold list because ms->suspend
+	 * is set.
+	 */
+	spin_lock_irq(&ms->lock);
+	holds = ms->holds;
+	bio_list_init(&ms->holds);
+	spin_unlock_irq(&ms->lock);
+
+	while ((bio = bio_list_pop(&holds)))
+		hold_bio(ms, bio);
+
+	/*
 	 * We must finish up all the work that we've
 	 * generated (i.e. recovery work).
 	 */
-	rh_stop_recovery(&ms->rh);
+	dm_rh_stop_recovery(ms->rh);
 
 	wait_event(_kmirrord_recovery_stopped,
-		   !atomic_read(&ms->rh.recovery_in_flight));
+		   !dm_rh_recovery_in_flight(ms->rh));
 
 	if (log->type->presuspend && log->type->presuspend(log))
 		/* FIXME: need better error handling */
@@ -1771,7 +1308,7 @@ static void mirror_presuspend(struct dm_target *ti)
 static void mirror_postsuspend(struct dm_target *ti)
 {
 	struct mirror_set *ms = ti->private;
-	struct dm_dirty_log *log = ms->rh.log;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
 
 	if (log->type->postsuspend && log->type->postsuspend(log))
 		/* FIXME: need better error handling */
@@ -1781,13 +1318,13 @@ static void mirror_postsuspend(struct dm_target *ti)
 static void mirror_resume(struct dm_target *ti)
 {
 	struct mirror_set *ms = ti->private;
-	struct dm_dirty_log *log = ms->rh.log;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
 
 	atomic_set(&ms->suspend, 0);
 	if (log->type->resume && log->type->resume(log))
 		/* FIXME: need better error handling */
 		DMWARN("log resume failed");
-	rh_start_recovery(&ms->rh);
+	dm_rh_start_recovery(ms->rh);
 }
 
 /*
@@ -1808,18 +1345,19 @@ static char device_status_char(struct mirror *m)
 	if (!atomic_read(&(m->error_count)))
 		return 'A';
 
-	return (test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
+	return (test_bit(DM_RAID1_FLUSH_ERROR, &(m->error_type))) ? 'F' :
+		(test_bit(DM_RAID1_WRITE_ERROR, &(m->error_type))) ? 'D' :
 		(test_bit(DM_RAID1_SYNC_ERROR, &(m->error_type))) ? 'S' :
 		(test_bit(DM_RAID1_READ_ERROR, &(m->error_type))) ? 'R' : 'U';
 }
 
 
-static int mirror_status(struct dm_target *ti, status_type_t type,
-			 char *result, unsigned int maxlen)
+static void mirror_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
 {
 	unsigned int m, sz = 0;
 	struct mirror_set *ms = (struct mirror_set *) ti->private;
-	struct dm_dirty_log *log = ms->rh.log;
+	struct dm_dirty_log *log = dm_rh_dirty_log(ms->rh);
 	char buffer[ms->nr_mirrors + 1];
 
 	switch (type) {
@@ -1832,15 +1370,15 @@ static int mirror_status(struct dm_target *ti, status_type_t type,
 		buffer[m] = '\0';
 
 		DMEMIT("%llu/%llu 1 %s ",
-		      (unsigned long long)log->type->get_sync_count(ms->rh.log),
+		      (unsigned long long)log->type->get_sync_count(log),
 		      (unsigned long long)ms->nr_regions, buffer);
 
-		sz += log->type->status(ms->rh.log, type, result+sz, maxlen-sz);
+		sz += log->type->status(log, type, result+sz, maxlen-sz);
 
 		break;
 
 	case STATUSTYPE_TABLE:
-		sz = log->type->status(ms->rh.log, type, result, maxlen);
+		sz = log->type->status(log, type, result, maxlen);
 
 		DMEMIT("%d", ms->nr_mirrors);
 		for (m = 0; m < ms->nr_mirrors; m++)
@@ -1850,13 +1388,25 @@ static int mirror_status(struct dm_target *ti, status_type_t type,
 		if (ms->features & DM_RAID1_HANDLE_ERRORS)
 			DMEMIT(" 1 handle_errors");
 	}
+}
 
-	return 0;
+static int mirror_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct mirror_set *ms = ti->private;
+	int ret = 0;
+	unsigned i;
+
+	for (i = 0; !ret && i < ms->nr_mirrors; i++)
+		ret = fn(ti, ms->mirror[i].dev,
+			 ms->mirror[i].offset, ti->len, data);
+
+	return ret;
 }
 
 static struct target_type mirror_target = {
 	.name	 = "mirror",
-	.version = {1, 0, 20},
+	.version = {1, 13, 2},
 	.module	 = THIS_MODULE,
 	.ctr	 = mirror_ctr,
 	.dtr	 = mirror_dtr,
@@ -1866,6 +1416,7 @@ static struct target_type mirror_target = {
 	.postsuspend = mirror_postsuspend,
 	.resume	 = mirror_resume,
 	.status	 = mirror_status,
+	.iterate_devices = mirror_iterate_devices,
 };
 
 static int __init dm_mirror_init(void)
@@ -1873,19 +1424,20 @@ static int __init dm_mirror_init(void)
 	int r;
 
 	r = dm_register_target(&mirror_target);
-	if (r < 0)
+	if (r < 0) {
 		DMERR("Failed to register mirror target");
+		goto bad_target;
+	}
+
+	return 0;
 
+bad_target:
 	return r;
 }
 
 static void __exit dm_mirror_exit(void)
 {
-	int r;
-
-	r = dm_unregister_target(&mirror_target);
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&mirror_target);
 }
 
 /* Module hooks */
diff --git a/drivers/md/dm-region-hash.c b/drivers/md/dm-region-hash.c
new file mode 100644
index 00000000000..b929fd5f498
--- /dev/null
+++ b/drivers/md/dm-region-hash.c
@@ -0,0 +1,724 @@
+/*
+ * Copyright (C) 2003 Sistina Software Limited.
+ * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/dm-dirty-log.h>
+#include <linux/dm-region-hash.h>
+
+#include <linux/ctype.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/vmalloc.h>
+
+#include "dm.h"
+
+#define	DM_MSG_PREFIX	"region hash"
+
+/*-----------------------------------------------------------------
+ * Region hash
+ *
+ * The mirror splits itself up into discrete regions.  Each
+ * region can be in one of three states: clean, dirty,
+ * nosync.  There is no need to put clean regions in the hash.
+ *
+ * In addition to being present in the hash table a region _may_
+ * be present on one of three lists.
+ *
+ *   clean_regions: Regions on this list have no io pending to
+ *   them, they are in sync, we are no longer interested in them,
+ *   they are dull.  dm_rh_update_states() will remove them from the
+ *   hash table.
+ *
+ *   quiesced_regions: These regions have been spun down, ready
+ *   for recovery.  rh_recovery_start() will remove regions from
+ *   this list and hand them to kmirrord, which will schedule the
+ *   recovery io with kcopyd.
+ *
+ *   recovered_regions: Regions that kcopyd has successfully
+ *   recovered.  dm_rh_update_states() will now schedule any delayed
+ *   io, up the recovery_count, and remove the region from the
+ *   hash.
+ *
+ * There are 2 locks:
+ *   A rw spin lock 'hash_lock' protects just the hash table,
+ *   this is never held in write mode from interrupt context,
+ *   which I believe means that we only have to disable irqs when
+ *   doing a write lock.
+ *
+ *   An ordinary spin lock 'region_lock' that protects the three
+ *   lists in the region_hash, with the 'state', 'list' and
+ *   'delayed_bios' fields of the regions.  This is used from irq
+ *   context, so all other uses will have to suspend local irqs.
+ *---------------------------------------------------------------*/
+struct dm_region_hash {
+	uint32_t region_size;
+	unsigned region_shift;
+
+	/* holds persistent region state */
+	struct dm_dirty_log *log;
+
+	/* hash table */
+	rwlock_t hash_lock;
+	mempool_t *region_pool;
+	unsigned mask;
+	unsigned nr_buckets;
+	unsigned prime;
+	unsigned shift;
+	struct list_head *buckets;
+
+	unsigned max_recovery; /* Max # of regions to recover in parallel */
+
+	spinlock_t region_lock;
+	atomic_t recovery_in_flight;
+	struct semaphore recovery_count;
+	struct list_head clean_regions;
+	struct list_head quiesced_regions;
+	struct list_head recovered_regions;
+	struct list_head failed_recovered_regions;
+
+	/*
+	 * If there was a flush failure no regions can be marked clean.
+	 */
+	int flush_failure;
+
+	void *context;
+	sector_t target_begin;
+
+	/* Callback function to schedule bios writes */
+	void (*dispatch_bios)(void *context, struct bio_list *bios);
+
+	/* Callback function to wakeup callers worker thread. */
+	void (*wakeup_workers)(void *context);
+
+	/* Callback function to wakeup callers recovery waiters. */
+	void (*wakeup_all_recovery_waiters)(void *context);
+};
+
+struct dm_region {
+	struct dm_region_hash *rh;	/* FIXME: can we get rid of this ? */
+	region_t key;
+	int state;
+
+	struct list_head hash_list;
+	struct list_head list;
+
+	atomic_t pending;
+	struct bio_list delayed_bios;
+};
+
+/*
+ * Conversion fns
+ */
+static region_t dm_rh_sector_to_region(struct dm_region_hash *rh, sector_t sector)
+{
+	return sector >> rh->region_shift;
+}
+
+sector_t dm_rh_region_to_sector(struct dm_region_hash *rh, region_t region)
+{
+	return region << rh->region_shift;
+}
+EXPORT_SYMBOL_GPL(dm_rh_region_to_sector);
+
+region_t dm_rh_bio_to_region(struct dm_region_hash *rh, struct bio *bio)
+{
+	return dm_rh_sector_to_region(rh, bio->bi_iter.bi_sector -
+				      rh->target_begin);
+}
+EXPORT_SYMBOL_GPL(dm_rh_bio_to_region);
+
+void *dm_rh_region_context(struct dm_region *reg)
+{
+	return reg->rh->context;
+}
+EXPORT_SYMBOL_GPL(dm_rh_region_context);
+
+region_t dm_rh_get_region_key(struct dm_region *reg)
+{
+	return reg->key;
+}
+EXPORT_SYMBOL_GPL(dm_rh_get_region_key);
+
+sector_t dm_rh_get_region_size(struct dm_region_hash *rh)
+{
+	return rh->region_size;
+}
+EXPORT_SYMBOL_GPL(dm_rh_get_region_size);
+
+/*
+ * FIXME: shall we pass in a structure instead of all these args to
+ * dm_region_hash_create()????
+ */
+#define RH_HASH_MULT 2654435387U
+#define RH_HASH_SHIFT 12
+
+#define MIN_REGIONS 64
+struct dm_region_hash *dm_region_hash_create(
+		void *context, void (*dispatch_bios)(void *context,
+						     struct bio_list *bios),
+		void (*wakeup_workers)(void *context),
+		void (*wakeup_all_recovery_waiters)(void *context),
+		sector_t target_begin, unsigned max_recovery,
+		struct dm_dirty_log *log, uint32_t region_size,
+		region_t nr_regions)
+{
+	struct dm_region_hash *rh;
+	unsigned nr_buckets, max_buckets;
+	size_t i;
+
+	/*
+	 * Calculate a suitable number of buckets for our hash
+	 * table.
+	 */
+	max_buckets = nr_regions >> 6;
+	for (nr_buckets = 128u; nr_buckets < max_buckets; nr_buckets <<= 1)
+		;
+	nr_buckets >>= 1;
+
+	rh = kmalloc(sizeof(*rh), GFP_KERNEL);
+	if (!rh) {
+		DMERR("unable to allocate region hash memory");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	rh->context = context;
+	rh->dispatch_bios = dispatch_bios;
+	rh->wakeup_workers = wakeup_workers;
+	rh->wakeup_all_recovery_waiters = wakeup_all_recovery_waiters;
+	rh->target_begin = target_begin;
+	rh->max_recovery = max_recovery;
+	rh->log = log;
+	rh->region_size = region_size;
+	rh->region_shift = ffs(region_size) - 1;
+	rwlock_init(&rh->hash_lock);
+	rh->mask = nr_buckets - 1;
+	rh->nr_buckets = nr_buckets;
+
+	rh->shift = RH_HASH_SHIFT;
+	rh->prime = RH_HASH_MULT;
+
+	rh->buckets = vmalloc(nr_buckets * sizeof(*rh->buckets));
+	if (!rh->buckets) {
+		DMERR("unable to allocate region hash bucket memory");
+		kfree(rh);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	for (i = 0; i < nr_buckets; i++)
+		INIT_LIST_HEAD(rh->buckets + i);
+
+	spin_lock_init(&rh->region_lock);
+	sema_init(&rh->recovery_count, 0);
+	atomic_set(&rh->recovery_in_flight, 0);
+	INIT_LIST_HEAD(&rh->clean_regions);
+	INIT_LIST_HEAD(&rh->quiesced_regions);
+	INIT_LIST_HEAD(&rh->recovered_regions);
+	INIT_LIST_HEAD(&rh->failed_recovered_regions);
+	rh->flush_failure = 0;
+
+	rh->region_pool = mempool_create_kmalloc_pool(MIN_REGIONS,
+						      sizeof(struct dm_region));
+	if (!rh->region_pool) {
+		vfree(rh->buckets);
+		kfree(rh);
+		rh = ERR_PTR(-ENOMEM);
+	}
+
+	return rh;
+}
+EXPORT_SYMBOL_GPL(dm_region_hash_create);
+
+void dm_region_hash_destroy(struct dm_region_hash *rh)
+{
+	unsigned h;
+	struct dm_region *reg, *nreg;
+
+	BUG_ON(!list_empty(&rh->quiesced_regions));
+	for (h = 0; h < rh->nr_buckets; h++) {
+		list_for_each_entry_safe(reg, nreg, rh->buckets + h,
+					 hash_list) {
+			BUG_ON(atomic_read(&reg->pending));
+			mempool_free(reg, rh->region_pool);
+		}
+	}
+
+	if (rh->log)
+		dm_dirty_log_destroy(rh->log);
+
+	if (rh->region_pool)
+		mempool_destroy(rh->region_pool);
+
+	vfree(rh->buckets);
+	kfree(rh);
+}
+EXPORT_SYMBOL_GPL(dm_region_hash_destroy);
+
+struct dm_dirty_log *dm_rh_dirty_log(struct dm_region_hash *rh)
+{
+	return rh->log;
+}
+EXPORT_SYMBOL_GPL(dm_rh_dirty_log);
+
+static unsigned rh_hash(struct dm_region_hash *rh, region_t region)
+{
+	return (unsigned) ((region * rh->prime) >> rh->shift) & rh->mask;
+}
+
+static struct dm_region *__rh_lookup(struct dm_region_hash *rh, region_t region)
+{
+	struct dm_region *reg;
+	struct list_head *bucket = rh->buckets + rh_hash(rh, region);
+
+	list_for_each_entry(reg, bucket, hash_list)
+		if (reg->key == region)
+			return reg;
+
+	return NULL;
+}
+
+static void __rh_insert(struct dm_region_hash *rh, struct dm_region *reg)
+{
+	list_add(&reg->hash_list, rh->buckets + rh_hash(rh, reg->key));
+}
+
+static struct dm_region *__rh_alloc(struct dm_region_hash *rh, region_t region)
+{
+	struct dm_region *reg, *nreg;
+
+	nreg = mempool_alloc(rh->region_pool, GFP_ATOMIC);
+	if (unlikely(!nreg))
+		nreg = kmalloc(sizeof(*nreg), GFP_NOIO | __GFP_NOFAIL);
+
+	nreg->state = rh->log->type->in_sync(rh->log, region, 1) ?
+		      DM_RH_CLEAN : DM_RH_NOSYNC;
+	nreg->rh = rh;
+	nreg->key = region;
+	INIT_LIST_HEAD(&nreg->list);
+	atomic_set(&nreg->pending, 0);
+	bio_list_init(&nreg->delayed_bios);
+
+	write_lock_irq(&rh->hash_lock);
+	reg = __rh_lookup(rh, region);
+	if (reg)
+		/* We lost the race. */
+		mempool_free(nreg, rh->region_pool);
+	else {
+		__rh_insert(rh, nreg);
+		if (nreg->state == DM_RH_CLEAN) {
+			spin_lock(&rh->region_lock);
+			list_add(&nreg->list, &rh->clean_regions);
+			spin_unlock(&rh->region_lock);
+		}
+
+		reg = nreg;
+	}
+	write_unlock_irq(&rh->hash_lock);
+
+	return reg;
+}
+
+static struct dm_region *__rh_find(struct dm_region_hash *rh, region_t region)
+{
+	struct dm_region *reg;
+
+	reg = __rh_lookup(rh, region);
+	if (!reg) {
+		read_unlock(&rh->hash_lock);
+		reg = __rh_alloc(rh, region);
+		read_lock(&rh->hash_lock);
+	}
+
+	return reg;
+}
+
+int dm_rh_get_state(struct dm_region_hash *rh, region_t region, int may_block)
+{
+	int r;
+	struct dm_region *reg;
+
+	read_lock(&rh->hash_lock);
+	reg = __rh_lookup(rh, region);
+	read_unlock(&rh->hash_lock);
+
+	if (reg)
+		return reg->state;
+
+	/*
+	 * The region wasn't in the hash, so we fall back to the
+	 * dirty log.
+	 */
+	r = rh->log->type->in_sync(rh->log, region, may_block);
+
+	/*
+	 * Any error from the dirty log (eg. -EWOULDBLOCK) gets
+	 * taken as a DM_RH_NOSYNC
+	 */
+	return r == 1 ? DM_RH_CLEAN : DM_RH_NOSYNC;
+}
+EXPORT_SYMBOL_GPL(dm_rh_get_state);
+
+static void complete_resync_work(struct dm_region *reg, int success)
+{
+	struct dm_region_hash *rh = reg->rh;
+
+	rh->log->type->set_region_sync(rh->log, reg->key, success);
+
+	/*
+	 * Dispatch the bios before we call 'wake_up_all'.
+	 * This is important because if we are suspending,
+	 * we want to know that recovery is complete and
+	 * the work queue is flushed.  If we wake_up_all
+	 * before we dispatch_bios (queue bios and call wake()),
+	 * then we risk suspending before the work queue
+	 * has been properly flushed.
+	 */
+	rh->dispatch_bios(rh->context, &reg->delayed_bios);
+	if (atomic_dec_and_test(&rh->recovery_in_flight))
+		rh->wakeup_all_recovery_waiters(rh->context);
+	up(&rh->recovery_count);
+}
+
+/* dm_rh_mark_nosync
+ * @ms
+ * @bio
+ *
+ * The bio was written on some mirror(s) but failed on other mirror(s).
+ * We can successfully endio the bio but should avoid the region being
+ * marked clean by setting the state DM_RH_NOSYNC.
+ *
+ * This function is _not_ safe in interrupt context!
+ */
+void dm_rh_mark_nosync(struct dm_region_hash *rh, struct bio *bio)
+{
+	unsigned long flags;
+	struct dm_dirty_log *log = rh->log;
+	struct dm_region *reg;
+	region_t region = dm_rh_bio_to_region(rh, bio);
+	int recovering = 0;
+
+	if (bio->bi_rw & REQ_FLUSH) {
+		rh->flush_failure = 1;
+		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);
+
+	read_lock(&rh->hash_lock);
+	reg = __rh_find(rh, region);
+	read_unlock(&rh->hash_lock);
+
+	/* region hash entry should exist because write was in-flight */
+	BUG_ON(!reg);
+	BUG_ON(!list_empty(&reg->list));
+
+	spin_lock_irqsave(&rh->region_lock, flags);
+	/*
+	 * Possible cases:
+	 *   1) DM_RH_DIRTY
+	 *   2) DM_RH_NOSYNC: was dirty, other preceding writes failed
+	 *   3) DM_RH_RECOVERING: flushing pending writes
+	 * Either case, the region should have not been connected to list.
+	 */
+	recovering = (reg->state == DM_RH_RECOVERING);
+	reg->state = DM_RH_NOSYNC;
+	BUG_ON(!list_empty(&reg->list));
+	spin_unlock_irqrestore(&rh->region_lock, flags);
+
+	if (recovering)
+		complete_resync_work(reg, 0);
+}
+EXPORT_SYMBOL_GPL(dm_rh_mark_nosync);
+
+void dm_rh_update_states(struct dm_region_hash *rh, int errors_handled)
+{
+	struct dm_region *reg, *next;
+
+	LIST_HEAD(clean);
+	LIST_HEAD(recovered);
+	LIST_HEAD(failed_recovered);
+
+	/*
+	 * Quickly grab the lists.
+	 */
+	write_lock_irq(&rh->hash_lock);
+	spin_lock(&rh->region_lock);
+	if (!list_empty(&rh->clean_regions)) {
+		list_splice_init(&rh->clean_regions, &clean);
+
+		list_for_each_entry(reg, &clean, list)
+			list_del(&reg->hash_list);
+	}
+
+	if (!list_empty(&rh->recovered_regions)) {
+		list_splice_init(&rh->recovered_regions, &recovered);
+
+		list_for_each_entry(reg, &recovered, list)
+			list_del(&reg->hash_list);
+	}
+
+	if (!list_empty(&rh->failed_recovered_regions)) {
+		list_splice_init(&rh->failed_recovered_regions,
+				 &failed_recovered);
+
+		list_for_each_entry(reg, &failed_recovered, list)
+			list_del(&reg->hash_list);
+	}
+
+	spin_unlock(&rh->region_lock);
+	write_unlock_irq(&rh->hash_lock);
+
+	/*
+	 * All the regions on the recovered and clean lists have
+	 * now been pulled out of the system, so no need to do
+	 * any more locking.
+	 */
+	list_for_each_entry_safe(reg, next, &recovered, list) {
+		rh->log->type->clear_region(rh->log, reg->key);
+		complete_resync_work(reg, 1);
+		mempool_free(reg, rh->region_pool);
+	}
+
+	list_for_each_entry_safe(reg, next, &failed_recovered, list) {
+		complete_resync_work(reg, errors_handled ? 0 : 1);
+		mempool_free(reg, rh->region_pool);
+	}
+
+	list_for_each_entry_safe(reg, next, &clean, list) {
+		rh->log->type->clear_region(rh->log, reg->key);
+		mempool_free(reg, rh->region_pool);
+	}
+
+	rh->log->type->flush(rh->log);
+}
+EXPORT_SYMBOL_GPL(dm_rh_update_states);
+
+static void rh_inc(struct dm_region_hash *rh, region_t region)
+{
+	struct dm_region *reg;
+
+	read_lock(&rh->hash_lock);
+	reg = __rh_find(rh, region);
+
+	spin_lock_irq(&rh->region_lock);
+	atomic_inc(&reg->pending);
+
+	if (reg->state == DM_RH_CLEAN) {
+		reg->state = DM_RH_DIRTY;
+		list_del_init(&reg->list);	/* take off the clean list */
+		spin_unlock_irq(&rh->region_lock);
+
+		rh->log->type->mark_region(rh->log, reg->key);
+	} else
+		spin_unlock_irq(&rh->region_lock);
+
+
+	read_unlock(&rh->hash_lock);
+}
+
+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 | REQ_DISCARD))
+			continue;
+		rh_inc(rh, dm_rh_bio_to_region(rh, bio));
+	}
+}
+EXPORT_SYMBOL_GPL(dm_rh_inc_pending);
+
+void dm_rh_dec(struct dm_region_hash *rh, region_t region)
+{
+	unsigned long flags;
+	struct dm_region *reg;
+	int should_wake = 0;
+
+	read_lock(&rh->hash_lock);
+	reg = __rh_lookup(rh, region);
+	read_unlock(&rh->hash_lock);
+
+	spin_lock_irqsave(&rh->region_lock, flags);
+	if (atomic_dec_and_test(&reg->pending)) {
+		/*
+		 * There is no pending I/O for this region.
+		 * We can move the region to corresponding list for next action.
+		 * At this point, the region is not yet connected to any list.
+		 *
+		 * If the state is DM_RH_NOSYNC, the region should be kept off
+		 * from clean list.
+		 * The hash entry for DM_RH_NOSYNC will remain in memory
+		 * until the region is recovered or the map is reloaded.
+		 */
+
+		/* do nothing for DM_RH_NOSYNC */
+		if (unlikely(rh->flush_failure)) {
+			/*
+			 * If a write flush failed some time ago, we
+			 * don't know whether or not this write made it
+			 * to the disk, so we must resync the device.
+			 */
+			reg->state = DM_RH_NOSYNC;
+		} else if (reg->state == DM_RH_RECOVERING) {
+			list_add_tail(&reg->list, &rh->quiesced_regions);
+		} else if (reg->state == DM_RH_DIRTY) {
+			reg->state = DM_RH_CLEAN;
+			list_add(&reg->list, &rh->clean_regions);
+		}
+		should_wake = 1;
+	}
+	spin_unlock_irqrestore(&rh->region_lock, flags);
+
+	if (should_wake)
+		rh->wakeup_workers(rh->context);
+}
+EXPORT_SYMBOL_GPL(dm_rh_dec);
+
+/*
+ * Starts quiescing a region in preparation for recovery.
+ */
+static int __rh_recovery_prepare(struct dm_region_hash *rh)
+{
+	int r;
+	region_t region;
+	struct dm_region *reg;
+
+	/*
+	 * Ask the dirty log what's next.
+	 */
+	r = rh->log->type->get_resync_work(rh->log, &region);
+	if (r <= 0)
+		return r;
+
+	/*
+	 * Get this region, and start it quiescing by setting the
+	 * recovering flag.
+	 */
+	read_lock(&rh->hash_lock);
+	reg = __rh_find(rh, region);
+	read_unlock(&rh->hash_lock);
+
+	spin_lock_irq(&rh->region_lock);
+	reg->state = DM_RH_RECOVERING;
+
+	/* Already quiesced ? */
+	if (atomic_read(&reg->pending))
+		list_del_init(&reg->list);
+	else
+		list_move(&reg->list, &rh->quiesced_regions);
+
+	spin_unlock_irq(&rh->region_lock);
+
+	return 1;
+}
+
+void dm_rh_recovery_prepare(struct dm_region_hash *rh)
+{
+	/* Extra reference to avoid race with dm_rh_stop_recovery */
+	atomic_inc(&rh->recovery_in_flight);
+
+	while (!down_trylock(&rh->recovery_count)) {
+		atomic_inc(&rh->recovery_in_flight);
+		if (__rh_recovery_prepare(rh) <= 0) {
+			atomic_dec(&rh->recovery_in_flight);
+			up(&rh->recovery_count);
+			break;
+		}
+	}
+
+	/* Drop the extra reference */
+	if (atomic_dec_and_test(&rh->recovery_in_flight))
+		rh->wakeup_all_recovery_waiters(rh->context);
+}
+EXPORT_SYMBOL_GPL(dm_rh_recovery_prepare);
+
+/*
+ * Returns any quiesced regions.
+ */
+struct dm_region *dm_rh_recovery_start(struct dm_region_hash *rh)
+{
+	struct dm_region *reg = NULL;
+
+	spin_lock_irq(&rh->region_lock);
+	if (!list_empty(&rh->quiesced_regions)) {
+		reg = list_entry(rh->quiesced_regions.next,
+				 struct dm_region, list);
+		list_del_init(&reg->list);  /* remove from the quiesced list */
+	}
+	spin_unlock_irq(&rh->region_lock);
+
+	return reg;
+}
+EXPORT_SYMBOL_GPL(dm_rh_recovery_start);
+
+void dm_rh_recovery_end(struct dm_region *reg, int success)
+{
+	struct dm_region_hash *rh = reg->rh;
+
+	spin_lock_irq(&rh->region_lock);
+	if (success)
+		list_add(&reg->list, &reg->rh->recovered_regions);
+	else
+		list_add(&reg->list, &reg->rh->failed_recovered_regions);
+
+	spin_unlock_irq(&rh->region_lock);
+
+	rh->wakeup_workers(rh->context);
+}
+EXPORT_SYMBOL_GPL(dm_rh_recovery_end);
+
+/* Return recovery in flight count. */
+int dm_rh_recovery_in_flight(struct dm_region_hash *rh)
+{
+	return atomic_read(&rh->recovery_in_flight);
+}
+EXPORT_SYMBOL_GPL(dm_rh_recovery_in_flight);
+
+int dm_rh_flush(struct dm_region_hash *rh)
+{
+	return rh->log->type->flush(rh->log);
+}
+EXPORT_SYMBOL_GPL(dm_rh_flush);
+
+void dm_rh_delay(struct dm_region_hash *rh, struct bio *bio)
+{
+	struct dm_region *reg;
+
+	read_lock(&rh->hash_lock);
+	reg = __rh_find(rh, dm_rh_bio_to_region(rh, bio));
+	bio_list_add(&reg->delayed_bios, bio);
+	read_unlock(&rh->hash_lock);
+}
+EXPORT_SYMBOL_GPL(dm_rh_delay);
+
+void dm_rh_stop_recovery(struct dm_region_hash *rh)
+{
+	int i;
+
+	/* wait for any recovering regions */
+	for (i = 0; i < rh->max_recovery; i++)
+		down(&rh->recovery_count);
+}
+EXPORT_SYMBOL_GPL(dm_rh_stop_recovery);
+
+void dm_rh_start_recovery(struct dm_region_hash *rh)
+{
+	int i;
+
+	for (i = 0; i < rh->max_recovery; i++)
+		up(&rh->recovery_count);
+
+	rh->wakeup_workers(rh->context);
+}
+EXPORT_SYMBOL_GPL(dm_rh_start_recovery);
+
+MODULE_DESCRIPTION(DM_NAME " region hash");
+MODULE_AUTHOR("Joe Thornber/Heinz Mauelshagen <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-round-robin.c b/drivers/md/dm-round-robin.c
index 391dfa2ad43..6ab1192cdd5 100644
--- a/drivers/md/dm-round-robin.c
+++ b/drivers/md/dm-round-robin.c
@@ -9,10 +9,12 @@
  * Round-robin path selector.
  */
 
-#include "dm.h"
+#include <linux/device-mapper.h>
+
 #include "dm-path-selector.h"
 
 #include <linux/slab.h>
+#include <linux/module.h>
 
 #define DM_MSG_PREFIX "multipath round-robin"
 
@@ -112,6 +114,7 @@ static int rr_add_path(struct path_selector *ps, struct dm_path *path,
 	struct selector *s = (struct selector *) ps->context;
 	struct path_info *pi;
 	unsigned repeat_count = RR_MIN_IO;
+	char dummy;
 
 	if (argc > 1) {
 		*error = "round-robin ps: incorrect number of arguments";
@@ -119,7 +122,7 @@ static int rr_add_path(struct path_selector *ps, struct dm_path *path,
 	}
 
 	/* First path argument is number of I/Os before switching path */
-	if ((argc == 1) && (sscanf(argv[0], "%u", &repeat_count) != 1)) {
+	if ((argc == 1) && (sscanf(argv[0], "%u%c", &repeat_count, &dummy) != 1)) {
 		*error = "round-robin ps: invalid repeat count";
 		return -EINVAL;
 	}
@@ -160,7 +163,7 @@ static int rr_reinstate_path(struct path_selector *ps, struct dm_path *p)
 }
 
 static struct dm_path *rr_select_path(struct path_selector *ps,
-				   unsigned *repeat_count)
+				      unsigned *repeat_count, size_t nr_bytes)
 {
 	struct selector *s = (struct selector *) ps->context;
 	struct path_info *pi = NULL;
diff --git a/drivers/md/dm-service-time.c b/drivers/md/dm-service-time.c
new file mode 100644
index 00000000000..9df8f6bd641
--- /dev/null
+++ b/drivers/md/dm-service-time.c
@@ -0,0 +1,343 @@
+/*
+ * Copyright (C) 2007-2009 NEC Corporation.  All Rights Reserved.
+ *
+ * Module Author: Kiyoshi Ueda
+ *
+ * This file is released under the GPL.
+ *
+ * Throughput oriented path selector.
+ */
+
+#include "dm.h"
+#include "dm-path-selector.h"
+
+#include <linux/slab.h>
+#include <linux/module.h>
+
+#define DM_MSG_PREFIX	"multipath service-time"
+#define ST_MIN_IO	1
+#define ST_MAX_RELATIVE_THROUGHPUT	100
+#define ST_MAX_RELATIVE_THROUGHPUT_SHIFT	7
+#define ST_MAX_INFLIGHT_SIZE	((size_t)-1 >> ST_MAX_RELATIVE_THROUGHPUT_SHIFT)
+#define ST_VERSION	"0.2.0"
+
+struct selector {
+	struct list_head valid_paths;
+	struct list_head failed_paths;
+};
+
+struct path_info {
+	struct list_head list;
+	struct dm_path *path;
+	unsigned repeat_count;
+	unsigned relative_throughput;
+	atomic_t in_flight_size;	/* Total size of in-flight I/Os */
+};
+
+static struct selector *alloc_selector(void)
+{
+	struct selector *s = kmalloc(sizeof(*s), GFP_KERNEL);
+
+	if (s) {
+		INIT_LIST_HEAD(&s->valid_paths);
+		INIT_LIST_HEAD(&s->failed_paths);
+	}
+
+	return s;
+}
+
+static int st_create(struct path_selector *ps, unsigned argc, char **argv)
+{
+	struct selector *s = alloc_selector();
+
+	if (!s)
+		return -ENOMEM;
+
+	ps->context = s;
+	return 0;
+}
+
+static void free_paths(struct list_head *paths)
+{
+	struct path_info *pi, *next;
+
+	list_for_each_entry_safe(pi, next, paths, list) {
+		list_del(&pi->list);
+		kfree(pi);
+	}
+}
+
+static void st_destroy(struct path_selector *ps)
+{
+	struct selector *s = ps->context;
+
+	free_paths(&s->valid_paths);
+	free_paths(&s->failed_paths);
+	kfree(s);
+	ps->context = NULL;
+}
+
+static int st_status(struct path_selector *ps, struct dm_path *path,
+		     status_type_t type, char *result, unsigned maxlen)
+{
+	unsigned sz = 0;
+	struct path_info *pi;
+
+	if (!path)
+		DMEMIT("0 ");
+	else {
+		pi = path->pscontext;
+
+		switch (type) {
+		case STATUSTYPE_INFO:
+			DMEMIT("%d %u ", atomic_read(&pi->in_flight_size),
+			       pi->relative_throughput);
+			break;
+		case STATUSTYPE_TABLE:
+			DMEMIT("%u %u ", pi->repeat_count,
+			       pi->relative_throughput);
+			break;
+		}
+	}
+
+	return sz;
+}
+
+static int st_add_path(struct path_selector *ps, struct dm_path *path,
+		       int argc, char **argv, char **error)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi;
+	unsigned repeat_count = ST_MIN_IO;
+	unsigned relative_throughput = 1;
+	char dummy;
+
+	/*
+	 * Arguments: [<repeat_count> [<relative_throughput>]]
+	 * 	<repeat_count>: The number of I/Os before switching path.
+	 * 			If not given, default (ST_MIN_IO) is used.
+	 * 	<relative_throughput>: The relative throughput value of
+	 *			the path among all paths in the path-group.
+	 * 			The valid range: 0-<ST_MAX_RELATIVE_THROUGHPUT>
+	 *			If not given, minimum value '1' is used.
+	 *			If '0' is given, the path isn't selected while
+	 * 			other paths having a positive value are
+	 * 			available.
+	 */
+	if (argc > 2) {
+		*error = "service-time ps: incorrect number of arguments";
+		return -EINVAL;
+	}
+
+	if (argc && (sscanf(argv[0], "%u%c", &repeat_count, &dummy) != 1)) {
+		*error = "service-time ps: invalid repeat count";
+		return -EINVAL;
+	}
+
+	if ((argc == 2) &&
+	    (sscanf(argv[1], "%u%c", &relative_throughput, &dummy) != 1 ||
+	     relative_throughput > ST_MAX_RELATIVE_THROUGHPUT)) {
+		*error = "service-time ps: invalid relative_throughput value";
+		return -EINVAL;
+	}
+
+	/* allocate the path */
+	pi = kmalloc(sizeof(*pi), GFP_KERNEL);
+	if (!pi) {
+		*error = "service-time ps: Error allocating path context";
+		return -ENOMEM;
+	}
+
+	pi->path = path;
+	pi->repeat_count = repeat_count;
+	pi->relative_throughput = relative_throughput;
+	atomic_set(&pi->in_flight_size, 0);
+
+	path->pscontext = pi;
+
+	list_add_tail(&pi->list, &s->valid_paths);
+
+	return 0;
+}
+
+static void st_fail_path(struct path_selector *ps, struct dm_path *path)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = path->pscontext;
+
+	list_move(&pi->list, &s->failed_paths);
+}
+
+static int st_reinstate_path(struct path_selector *ps, struct dm_path *path)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = path->pscontext;
+
+	list_move_tail(&pi->list, &s->valid_paths);
+
+	return 0;
+}
+
+/*
+ * Compare the estimated service time of 2 paths, pi1 and pi2,
+ * for the incoming I/O.
+ *
+ * Returns:
+ * < 0 : pi1 is better
+ * 0   : no difference between pi1 and pi2
+ * > 0 : pi2 is better
+ *
+ * Description:
+ * Basically, the service time is estimated by:
+ *     ('pi->in-flight-size' + 'incoming') / 'pi->relative_throughput'
+ * To reduce the calculation, some optimizations are made.
+ * (See comments inline)
+ */
+static int st_compare_load(struct path_info *pi1, struct path_info *pi2,
+			   size_t incoming)
+{
+	size_t sz1, sz2, st1, st2;
+
+	sz1 = atomic_read(&pi1->in_flight_size);
+	sz2 = atomic_read(&pi2->in_flight_size);
+
+	/*
+	 * Case 1: Both have same throughput value. Choose less loaded path.
+	 */
+	if (pi1->relative_throughput == pi2->relative_throughput)
+		return sz1 - sz2;
+
+	/*
+	 * Case 2a: Both have same load. Choose higher throughput path.
+	 * Case 2b: One path has no throughput value. Choose the other one.
+	 */
+	if (sz1 == sz2 ||
+	    !pi1->relative_throughput || !pi2->relative_throughput)
+		return pi2->relative_throughput - pi1->relative_throughput;
+
+	/*
+	 * Case 3: Calculate service time. Choose faster path.
+	 *         Service time using pi1:
+	 *             st1 = (sz1 + incoming) / pi1->relative_throughput
+	 *         Service time using pi2:
+	 *             st2 = (sz2 + incoming) / pi2->relative_throughput
+	 *
+	 *         To avoid the division, transform the expression to use
+	 *         multiplication.
+	 *         Because ->relative_throughput > 0 here, if st1 < st2,
+	 *         the expressions below are the same meaning:
+	 *             (sz1 + incoming) / pi1->relative_throughput <
+	 *                 (sz2 + incoming) / pi2->relative_throughput
+	 *             (sz1 + incoming) * pi2->relative_throughput <
+	 *                 (sz2 + incoming) * pi1->relative_throughput
+	 *         So use the later one.
+	 */
+	sz1 += incoming;
+	sz2 += incoming;
+	if (unlikely(sz1 >= ST_MAX_INFLIGHT_SIZE ||
+		     sz2 >= ST_MAX_INFLIGHT_SIZE)) {
+		/*
+		 * Size may be too big for multiplying pi->relative_throughput
+		 * and overflow.
+		 * To avoid the overflow and mis-selection, shift down both.
+		 */
+		sz1 >>= ST_MAX_RELATIVE_THROUGHPUT_SHIFT;
+		sz2 >>= ST_MAX_RELATIVE_THROUGHPUT_SHIFT;
+	}
+	st1 = sz1 * pi2->relative_throughput;
+	st2 = sz2 * pi1->relative_throughput;
+	if (st1 != st2)
+		return st1 - st2;
+
+	/*
+	 * Case 4: Service time is equal. Choose higher throughput path.
+	 */
+	return pi2->relative_throughput - pi1->relative_throughput;
+}
+
+static struct dm_path *st_select_path(struct path_selector *ps,
+				      unsigned *repeat_count, size_t nr_bytes)
+{
+	struct selector *s = ps->context;
+	struct path_info *pi = NULL, *best = NULL;
+
+	if (list_empty(&s->valid_paths))
+		return NULL;
+
+	/* Change preferred (first in list) path to evenly balance. */
+	list_move_tail(s->valid_paths.next, &s->valid_paths);
+
+	list_for_each_entry(pi, &s->valid_paths, list)
+		if (!best || (st_compare_load(pi, best, nr_bytes) < 0))
+			best = pi;
+
+	if (!best)
+		return NULL;
+
+	*repeat_count = best->repeat_count;
+
+	return best->path;
+}
+
+static int st_start_io(struct path_selector *ps, struct dm_path *path,
+		       size_t nr_bytes)
+{
+	struct path_info *pi = path->pscontext;
+
+	atomic_add(nr_bytes, &pi->in_flight_size);
+
+	return 0;
+}
+
+static int st_end_io(struct path_selector *ps, struct dm_path *path,
+		     size_t nr_bytes)
+{
+	struct path_info *pi = path->pscontext;
+
+	atomic_sub(nr_bytes, &pi->in_flight_size);
+
+	return 0;
+}
+
+static struct path_selector_type st_ps = {
+	.name		= "service-time",
+	.module		= THIS_MODULE,
+	.table_args	= 2,
+	.info_args	= 2,
+	.create		= st_create,
+	.destroy	= st_destroy,
+	.status		= st_status,
+	.add_path	= st_add_path,
+	.fail_path	= st_fail_path,
+	.reinstate_path	= st_reinstate_path,
+	.select_path	= st_select_path,
+	.start_io	= st_start_io,
+	.end_io		= st_end_io,
+};
+
+static int __init dm_st_init(void)
+{
+	int r = dm_register_path_selector(&st_ps);
+
+	if (r < 0)
+		DMERR("register failed %d", r);
+
+	DMINFO("version " ST_VERSION " loaded");
+
+	return r;
+}
+
+static void __exit dm_st_exit(void)
+{
+	int r = dm_unregister_path_selector(&st_ps);
+
+	if (r < 0)
+		DMERR("unregister failed %d", r);
+}
+
+module_init(dm_st_init);
+module_exit(dm_st_exit);
+
+MODULE_DESCRIPTION(DM_NAME " throughput oriented path selector");
+MODULE_AUTHOR("Kiyoshi Ueda <k-ueda@ct.jp.nec.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-snap-persistent.c b/drivers/md/dm-snap-persistent.c
new file mode 100644
index 00000000000..d6e88178d22
--- /dev/null
+++ b/drivers/md/dm-snap-persistent.c
@@ -0,0 +1,958 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2006-2008 Red Hat GmbH
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-exception-store.h"
+
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/vmalloc.h>
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/dm-io.h>
+#include "dm-bufio.h"
+
+#define DM_MSG_PREFIX "persistent snapshot"
+#define DM_CHUNK_SIZE_DEFAULT_SECTORS 32	/* 16KB */
+
+#define DM_PREFETCH_CHUNKS		12
+
+/*-----------------------------------------------------------------
+ * Persistent snapshots, by persistent we mean that the snapshot
+ * will survive a reboot.
+ *---------------------------------------------------------------*/
+
+/*
+ * We need to store a record of which parts of the origin have
+ * been copied to the snapshot device.  The snapshot code
+ * requires that we copy exception chunks to chunk aligned areas
+ * of the COW store.  It makes sense therefore, to store the
+ * metadata in chunk size blocks.
+ *
+ * There is no backward or forward compatibility implemented,
+ * snapshots with different disk versions than the kernel will
+ * not be usable.  It is expected that "lvcreate" will blank out
+ * the start of a fresh COW device before calling the snapshot
+ * constructor.
+ *
+ * The first chunk of the COW device just contains the header.
+ * After this there is a chunk filled with exception metadata,
+ * followed by as many exception chunks as can fit in the
+ * metadata areas.
+ *
+ * All on disk structures are in little-endian format.  The end
+ * of the exceptions info is indicated by an exception with a
+ * new_chunk of 0, which is invalid since it would point to the
+ * header chunk.
+ */
+
+/*
+ * Magic for persistent snapshots: "SnAp" - Feeble isn't it.
+ */
+#define SNAP_MAGIC 0x70416e53
+
+/*
+ * The on-disk version of the metadata.
+ */
+#define SNAPSHOT_DISK_VERSION 1
+
+#define NUM_SNAPSHOT_HDR_CHUNKS 1
+
+struct disk_header {
+	__le32 magic;
+
+	/*
+	 * Is this snapshot valid.  There is no way of recovering
+	 * an invalid snapshot.
+	 */
+	__le32 valid;
+
+	/*
+	 * Simple, incrementing version. no backward
+	 * compatibility.
+	 */
+	__le32 version;
+
+	/* In sectors */
+	__le32 chunk_size;
+} __packed;
+
+struct disk_exception {
+	__le64 old_chunk;
+	__le64 new_chunk;
+} __packed;
+
+struct core_exception {
+	uint64_t old_chunk;
+	uint64_t new_chunk;
+};
+
+struct commit_callback {
+	void (*callback)(void *, int success);
+	void *context;
+};
+
+/*
+ * The top level structure for a persistent exception store.
+ */
+struct pstore {
+	struct dm_exception_store *store;
+	int version;
+	int valid;
+	uint32_t exceptions_per_area;
+
+	/*
+	 * Now that we have an asynchronous kcopyd there is no
+	 * need for large chunk sizes, so it wont hurt to have a
+	 * whole chunks worth of metadata in memory at once.
+	 */
+	void *area;
+
+	/*
+	 * An area of zeros used to clear the next area.
+	 */
+	void *zero_area;
+
+	/*
+	 * An area used for header. The header can be written
+	 * concurrently with metadata (when invalidating the snapshot),
+	 * so it needs a separate buffer.
+	 */
+	void *header_area;
+
+	/*
+	 * Used to keep track of which metadata area the data in
+	 * 'chunk' refers to.
+	 */
+	chunk_t current_area;
+
+	/*
+	 * The next free chunk for an exception.
+	 *
+	 * When creating exceptions, all the chunks here and above are
+	 * free.  It holds the next chunk to be allocated.  On rare
+	 * occasions (e.g. after a system crash) holes can be left in
+	 * the exception store because chunks can be committed out of
+	 * order.
+	 *
+	 * When merging exceptions, it does not necessarily mean all the
+	 * chunks here and above are free.  It holds the value it would
+	 * have held if all chunks had been committed in order of
+	 * allocation.  Consequently the value may occasionally be
+	 * slightly too low, but since it's only used for 'status' and
+	 * it can never reach its minimum value too early this doesn't
+	 * matter.
+	 */
+
+	chunk_t next_free;
+
+	/*
+	 * The index of next free exception in the current
+	 * metadata area.
+	 */
+	uint32_t current_committed;
+
+	atomic_t pending_count;
+	uint32_t callback_count;
+	struct commit_callback *callbacks;
+	struct dm_io_client *io_client;
+
+	struct workqueue_struct *metadata_wq;
+};
+
+static int alloc_area(struct pstore *ps)
+{
+	int r = -ENOMEM;
+	size_t len;
+
+	len = ps->store->chunk_size << SECTOR_SHIFT;
+
+	/*
+	 * Allocate the chunk_size block of memory that will hold
+	 * a single metadata area.
+	 */
+	ps->area = vmalloc(len);
+	if (!ps->area)
+		goto err_area;
+
+	ps->zero_area = vzalloc(len);
+	if (!ps->zero_area)
+		goto err_zero_area;
+
+	ps->header_area = vmalloc(len);
+	if (!ps->header_area)
+		goto err_header_area;
+
+	return 0;
+
+err_header_area:
+	vfree(ps->zero_area);
+
+err_zero_area:
+	vfree(ps->area);
+
+err_area:
+	return r;
+}
+
+static void free_area(struct pstore *ps)
+{
+	if (ps->area)
+		vfree(ps->area);
+	ps->area = NULL;
+
+	if (ps->zero_area)
+		vfree(ps->zero_area);
+	ps->zero_area = NULL;
+
+	if (ps->header_area)
+		vfree(ps->header_area);
+	ps->header_area = NULL;
+}
+
+struct mdata_req {
+	struct dm_io_region *where;
+	struct dm_io_request *io_req;
+	struct work_struct work;
+	int result;
+};
+
+static void do_metadata(struct work_struct *work)
+{
+	struct mdata_req *req = container_of(work, struct mdata_req, work);
+
+	req->result = dm_io(req->io_req, 1, req->where, NULL);
+}
+
+/*
+ * Read or write a chunk aligned and sized block of data from a device.
+ */
+static int chunk_io(struct pstore *ps, void *area, chunk_t chunk, int rw,
+		    int metadata)
+{
+	struct dm_io_region where = {
+		.bdev = dm_snap_cow(ps->store->snap)->bdev,
+		.sector = ps->store->chunk_size * chunk,
+		.count = ps->store->chunk_size,
+	};
+	struct dm_io_request io_req = {
+		.bi_rw = rw,
+		.mem.type = DM_IO_VMA,
+		.mem.ptr.vma = area,
+		.client = ps->io_client,
+		.notify.fn = NULL,
+	};
+	struct mdata_req req;
+
+	if (!metadata)
+		return dm_io(&io_req, 1, &where, NULL);
+
+	req.where = &where;
+	req.io_req = &io_req;
+
+	/*
+	 * Issue the synchronous I/O from a different thread
+	 * to avoid generic_make_request recursion.
+	 */
+	INIT_WORK_ONSTACK(&req.work, do_metadata);
+	queue_work(ps->metadata_wq, &req.work);
+	flush_workqueue(ps->metadata_wq);
+	destroy_work_on_stack(&req.work);
+
+	return req.result;
+}
+
+/*
+ * Convert a metadata area index to a chunk index.
+ */
+static chunk_t area_location(struct pstore *ps, chunk_t area)
+{
+	return NUM_SNAPSHOT_HDR_CHUNKS + ((ps->exceptions_per_area + 1) * area);
+}
+
+static void skip_metadata(struct pstore *ps)
+{
+	uint32_t stride = ps->exceptions_per_area + 1;
+	chunk_t next_free = ps->next_free;
+	if (sector_div(next_free, stride) == NUM_SNAPSHOT_HDR_CHUNKS)
+		ps->next_free++;
+}
+
+/*
+ * Read or write a metadata area.  Remembering to skip the first
+ * chunk which holds the header.
+ */
+static int area_io(struct pstore *ps, int rw)
+{
+	int r;
+	chunk_t chunk;
+
+	chunk = area_location(ps, ps->current_area);
+
+	r = chunk_io(ps, ps->area, chunk, rw, 0);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static void zero_memory_area(struct pstore *ps)
+{
+	memset(ps->area, 0, ps->store->chunk_size << SECTOR_SHIFT);
+}
+
+static int zero_disk_area(struct pstore *ps, chunk_t area)
+{
+	return chunk_io(ps, ps->zero_area, area_location(ps, area), WRITE, 0);
+}
+
+static int read_header(struct pstore *ps, int *new_snapshot)
+{
+	int r;
+	struct disk_header *dh;
+	unsigned chunk_size;
+	int chunk_size_supplied = 1;
+	char *chunk_err;
+
+	/*
+	 * Use default chunk size (or logical_block_size, if larger)
+	 * if none supplied
+	 */
+	if (!ps->store->chunk_size) {
+		ps->store->chunk_size = max(DM_CHUNK_SIZE_DEFAULT_SECTORS,
+		    bdev_logical_block_size(dm_snap_cow(ps->store->snap)->
+					    bdev) >> 9);
+		ps->store->chunk_mask = ps->store->chunk_size - 1;
+		ps->store->chunk_shift = ffs(ps->store->chunk_size) - 1;
+		chunk_size_supplied = 0;
+	}
+
+	ps->io_client = dm_io_client_create();
+	if (IS_ERR(ps->io_client))
+		return PTR_ERR(ps->io_client);
+
+	r = alloc_area(ps);
+	if (r)
+		return r;
+
+	r = chunk_io(ps, ps->header_area, 0, READ, 1);
+	if (r)
+		goto bad;
+
+	dh = ps->header_area;
+
+	if (le32_to_cpu(dh->magic) == 0) {
+		*new_snapshot = 1;
+		return 0;
+	}
+
+	if (le32_to_cpu(dh->magic) != SNAP_MAGIC) {
+		DMWARN("Invalid or corrupt snapshot");
+		r = -ENXIO;
+		goto bad;
+	}
+
+	*new_snapshot = 0;
+	ps->valid = le32_to_cpu(dh->valid);
+	ps->version = le32_to_cpu(dh->version);
+	chunk_size = le32_to_cpu(dh->chunk_size);
+
+	if (ps->store->chunk_size == chunk_size)
+		return 0;
+
+	if (chunk_size_supplied)
+		DMWARN("chunk size %u in device metadata overrides "
+		       "table chunk size of %u.",
+		       chunk_size, ps->store->chunk_size);
+
+	/* We had a bogus chunk_size. Fix stuff up. */
+	free_area(ps);
+
+	r = dm_exception_store_set_chunk_size(ps->store, chunk_size,
+					      &chunk_err);
+	if (r) {
+		DMERR("invalid on-disk chunk size %u: %s.",
+		      chunk_size, chunk_err);
+		return r;
+	}
+
+	r = alloc_area(ps);
+	return r;
+
+bad:
+	free_area(ps);
+	return r;
+}
+
+static int write_header(struct pstore *ps)
+{
+	struct disk_header *dh;
+
+	memset(ps->header_area, 0, ps->store->chunk_size << SECTOR_SHIFT);
+
+	dh = ps->header_area;
+	dh->magic = cpu_to_le32(SNAP_MAGIC);
+	dh->valid = cpu_to_le32(ps->valid);
+	dh->version = cpu_to_le32(ps->version);
+	dh->chunk_size = cpu_to_le32(ps->store->chunk_size);
+
+	return chunk_io(ps, ps->header_area, 0, WRITE, 1);
+}
+
+/*
+ * Access functions for the disk exceptions, these do the endian conversions.
+ */
+static struct disk_exception *get_exception(struct pstore *ps, void *ps_area,
+					    uint32_t index)
+{
+	BUG_ON(index >= ps->exceptions_per_area);
+
+	return ((struct disk_exception *) ps_area) + index;
+}
+
+static void read_exception(struct pstore *ps, void *ps_area,
+			   uint32_t index, struct core_exception *result)
+{
+	struct disk_exception *de = get_exception(ps, ps_area, index);
+
+	/* copy it */
+	result->old_chunk = le64_to_cpu(de->old_chunk);
+	result->new_chunk = le64_to_cpu(de->new_chunk);
+}
+
+static void write_exception(struct pstore *ps,
+			    uint32_t index, struct core_exception *e)
+{
+	struct disk_exception *de = get_exception(ps, ps->area, index);
+
+	/* copy it */
+	de->old_chunk = cpu_to_le64(e->old_chunk);
+	de->new_chunk = cpu_to_le64(e->new_chunk);
+}
+
+static void clear_exception(struct pstore *ps, uint32_t index)
+{
+	struct disk_exception *de = get_exception(ps, ps->area, index);
+
+	/* clear it */
+	de->old_chunk = 0;
+	de->new_chunk = 0;
+}
+
+/*
+ * Registers the exceptions that are present in the current area.
+ * 'full' is filled in to indicate if the area has been
+ * filled.
+ */
+static int insert_exceptions(struct pstore *ps, void *ps_area,
+			     int (*callback)(void *callback_context,
+					     chunk_t old, chunk_t new),
+			     void *callback_context,
+			     int *full)
+{
+	int r;
+	unsigned int i;
+	struct core_exception e;
+
+	/* presume the area is full */
+	*full = 1;
+
+	for (i = 0; i < ps->exceptions_per_area; i++) {
+		read_exception(ps, ps_area, i, &e);
+
+		/*
+		 * If the new_chunk is pointing at the start of
+		 * the COW device, where the first metadata area
+		 * is we know that we've hit the end of the
+		 * exceptions.  Therefore the area is not full.
+		 */
+		if (e.new_chunk == 0LL) {
+			ps->current_committed = i;
+			*full = 0;
+			break;
+		}
+
+		/*
+		 * Keep track of the start of the free chunks.
+		 */
+		if (ps->next_free <= e.new_chunk)
+			ps->next_free = e.new_chunk + 1;
+
+		/*
+		 * Otherwise we add the exception to the snapshot.
+		 */
+		r = callback(callback_context, e.old_chunk, e.new_chunk);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static int read_exceptions(struct pstore *ps,
+			   int (*callback)(void *callback_context, chunk_t old,
+					   chunk_t new),
+			   void *callback_context)
+{
+	int r, full = 1;
+	struct dm_bufio_client *client;
+	chunk_t prefetch_area = 0;
+
+	client = dm_bufio_client_create(dm_snap_cow(ps->store->snap)->bdev,
+					ps->store->chunk_size << SECTOR_SHIFT,
+					1, 0, NULL, NULL);
+
+	if (IS_ERR(client))
+		return PTR_ERR(client);
+
+	/*
+	 * Setup for one current buffer + desired readahead buffers.
+	 */
+	dm_bufio_set_minimum_buffers(client, 1 + DM_PREFETCH_CHUNKS);
+
+	/*
+	 * Keeping reading chunks and inserting exceptions until
+	 * we find a partially full area.
+	 */
+	for (ps->current_area = 0; full; ps->current_area++) {
+		struct dm_buffer *bp;
+		void *area;
+		chunk_t chunk;
+
+		if (unlikely(prefetch_area < ps->current_area))
+			prefetch_area = ps->current_area;
+
+		if (DM_PREFETCH_CHUNKS) do {
+			chunk_t pf_chunk = area_location(ps, prefetch_area);
+			if (unlikely(pf_chunk >= dm_bufio_get_device_size(client)))
+				break;
+			dm_bufio_prefetch(client, pf_chunk, 1);
+			prefetch_area++;
+			if (unlikely(!prefetch_area))
+				break;
+		} while (prefetch_area <= ps->current_area + DM_PREFETCH_CHUNKS);
+
+		chunk = area_location(ps, ps->current_area);
+
+		area = dm_bufio_read(client, chunk, &bp);
+		if (unlikely(IS_ERR(area))) {
+			r = PTR_ERR(area);
+			goto ret_destroy_bufio;
+		}
+
+		r = insert_exceptions(ps, area, callback, callback_context,
+				      &full);
+
+		if (!full)
+			memcpy(ps->area, area, ps->store->chunk_size << SECTOR_SHIFT);
+
+		dm_bufio_release(bp);
+
+		dm_bufio_forget(client, chunk);
+
+		if (unlikely(r))
+			goto ret_destroy_bufio;
+	}
+
+	ps->current_area--;
+
+	skip_metadata(ps);
+
+	r = 0;
+
+ret_destroy_bufio:
+	dm_bufio_client_destroy(client);
+
+	return r;
+}
+
+static struct pstore *get_info(struct dm_exception_store *store)
+{
+	return (struct pstore *) store->context;
+}
+
+static void persistent_usage(struct dm_exception_store *store,
+			     sector_t *total_sectors,
+			     sector_t *sectors_allocated,
+			     sector_t *metadata_sectors)
+{
+	struct pstore *ps = get_info(store);
+
+	*sectors_allocated = ps->next_free * store->chunk_size;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+
+	/*
+	 * First chunk is the fixed header.
+	 * Then there are (ps->current_area + 1) metadata chunks, each one
+	 * separated from the next by ps->exceptions_per_area data chunks.
+	 */
+	*metadata_sectors = (ps->current_area + 1 + NUM_SNAPSHOT_HDR_CHUNKS) *
+			    store->chunk_size;
+}
+
+static void persistent_dtr(struct dm_exception_store *store)
+{
+	struct pstore *ps = get_info(store);
+
+	destroy_workqueue(ps->metadata_wq);
+
+	/* Created in read_header */
+	if (ps->io_client)
+		dm_io_client_destroy(ps->io_client);
+	free_area(ps);
+
+	/* Allocated in persistent_read_metadata */
+	if (ps->callbacks)
+		vfree(ps->callbacks);
+
+	kfree(ps);
+}
+
+static int persistent_read_metadata(struct dm_exception_store *store,
+				    int (*callback)(void *callback_context,
+						    chunk_t old, chunk_t new),
+				    void *callback_context)
+{
+	int r, uninitialized_var(new_snapshot);
+	struct pstore *ps = get_info(store);
+
+	/*
+	 * Read the snapshot header.
+	 */
+	r = read_header(ps, &new_snapshot);
+	if (r)
+		return r;
+
+	/*
+	 * Now we know correct chunk_size, complete the initialisation.
+	 */
+	ps->exceptions_per_area = (ps->store->chunk_size << SECTOR_SHIFT) /
+				  sizeof(struct disk_exception);
+	ps->callbacks = dm_vcalloc(ps->exceptions_per_area,
+				   sizeof(*ps->callbacks));
+	if (!ps->callbacks)
+		return -ENOMEM;
+
+	/*
+	 * Do we need to setup a new snapshot ?
+	 */
+	if (new_snapshot) {
+		r = write_header(ps);
+		if (r) {
+			DMWARN("write_header failed");
+			return r;
+		}
+
+		ps->current_area = 0;
+		zero_memory_area(ps);
+		r = zero_disk_area(ps, 0);
+		if (r)
+			DMWARN("zero_disk_area(0) failed");
+		return r;
+	}
+	/*
+	 * Sanity checks.
+	 */
+	if (ps->version != SNAPSHOT_DISK_VERSION) {
+		DMWARN("unable to handle snapshot disk version %d",
+		       ps->version);
+		return -EINVAL;
+	}
+
+	/*
+	 * Metadata are valid, but snapshot is invalidated
+	 */
+	if (!ps->valid)
+		return 1;
+
+	/*
+	 * Read the metadata.
+	 */
+	r = read_exceptions(ps, callback, callback_context);
+
+	return r;
+}
+
+static int persistent_prepare_exception(struct dm_exception_store *store,
+					struct dm_exception *e)
+{
+	struct pstore *ps = get_info(store);
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
+
+	/* Is there enough room ? */
+	if (size < ((ps->next_free + 1) * store->chunk_size))
+		return -ENOSPC;
+
+	e->new_chunk = ps->next_free;
+
+	/*
+	 * Move onto the next free pending, making sure to take
+	 * into account the location of the metadata chunks.
+	 */
+	ps->next_free++;
+	skip_metadata(ps);
+
+	atomic_inc(&ps->pending_count);
+	return 0;
+}
+
+static void persistent_commit_exception(struct dm_exception_store *store,
+					struct dm_exception *e,
+					void (*callback) (void *, int success),
+					void *callback_context)
+{
+	unsigned int i;
+	struct pstore *ps = get_info(store);
+	struct core_exception ce;
+	struct commit_callback *cb;
+
+	ce.old_chunk = e->old_chunk;
+	ce.new_chunk = e->new_chunk;
+	write_exception(ps, ps->current_committed++, &ce);
+
+	/*
+	 * Add the callback to the back of the array.  This code
+	 * is the only place where the callback array is
+	 * manipulated, and we know that it will never be called
+	 * multiple times concurrently.
+	 */
+	cb = ps->callbacks + ps->callback_count++;
+	cb->callback = callback;
+	cb->context = callback_context;
+
+	/*
+	 * If there are exceptions in flight and we have not yet
+	 * filled this metadata area there's nothing more to do.
+	 */
+	if (!atomic_dec_and_test(&ps->pending_count) &&
+	    (ps->current_committed != ps->exceptions_per_area))
+		return;
+
+	/*
+	 * If we completely filled the current area, then wipe the next one.
+	 */
+	if ((ps->current_committed == ps->exceptions_per_area) &&
+	    zero_disk_area(ps, ps->current_area + 1))
+		ps->valid = 0;
+
+	/*
+	 * Commit exceptions to disk.
+	 */
+	if (ps->valid && area_io(ps, WRITE_FLUSH_FUA))
+		ps->valid = 0;
+
+	/*
+	 * Advance to the next area if this one is full.
+	 */
+	if (ps->current_committed == ps->exceptions_per_area) {
+		ps->current_committed = 0;
+		ps->current_area++;
+		zero_memory_area(ps);
+	}
+
+	for (i = 0; i < ps->callback_count; i++) {
+		cb = ps->callbacks + i;
+		cb->callback(cb->context, ps->valid);
+	}
+
+	ps->callback_count = 0;
+}
+
+static int persistent_prepare_merge(struct dm_exception_store *store,
+				    chunk_t *last_old_chunk,
+				    chunk_t *last_new_chunk)
+{
+	struct pstore *ps = get_info(store);
+	struct core_exception ce;
+	int nr_consecutive;
+	int r;
+
+	/*
+	 * When current area is empty, move back to preceding area.
+	 */
+	if (!ps->current_committed) {
+		/*
+		 * Have we finished?
+		 */
+		if (!ps->current_area)
+			return 0;
+
+		ps->current_area--;
+		r = area_io(ps, READ);
+		if (r < 0)
+			return r;
+		ps->current_committed = ps->exceptions_per_area;
+	}
+
+	read_exception(ps, ps->area, ps->current_committed - 1, &ce);
+	*last_old_chunk = ce.old_chunk;
+	*last_new_chunk = ce.new_chunk;
+
+	/*
+	 * Find number of consecutive chunks within the current area,
+	 * working backwards.
+	 */
+	for (nr_consecutive = 1; nr_consecutive < ps->current_committed;
+	     nr_consecutive++) {
+		read_exception(ps, ps->area,
+			       ps->current_committed - 1 - nr_consecutive, &ce);
+		if (ce.old_chunk != *last_old_chunk - nr_consecutive ||
+		    ce.new_chunk != *last_new_chunk - nr_consecutive)
+			break;
+	}
+
+	return nr_consecutive;
+}
+
+static int persistent_commit_merge(struct dm_exception_store *store,
+				   int nr_merged)
+{
+	int r, i;
+	struct pstore *ps = get_info(store);
+
+	BUG_ON(nr_merged > ps->current_committed);
+
+	for (i = 0; i < nr_merged; i++)
+		clear_exception(ps, ps->current_committed - 1 - i);
+
+	r = area_io(ps, WRITE_FLUSH_FUA);
+	if (r < 0)
+		return r;
+
+	ps->current_committed -= nr_merged;
+
+	/*
+	 * At this stage, only persistent_usage() uses ps->next_free, so
+	 * we make no attempt to keep ps->next_free strictly accurate
+	 * as exceptions may have been committed out-of-order originally.
+	 * Once a snapshot has become merging, we set it to the value it
+	 * would have held had all the exceptions been committed in order.
+	 *
+	 * ps->current_area does not get reduced by prepare_merge() until
+	 * after commit_merge() has removed the nr_merged previous exceptions.
+	 */
+	ps->next_free = area_location(ps, ps->current_area) +
+			ps->current_committed + 1;
+
+	return 0;
+}
+
+static void persistent_drop_snapshot(struct dm_exception_store *store)
+{
+	struct pstore *ps = get_info(store);
+
+	ps->valid = 0;
+	if (write_header(ps))
+		DMWARN("write header failed");
+}
+
+static int persistent_ctr(struct dm_exception_store *store,
+			  unsigned argc, char **argv)
+{
+	struct pstore *ps;
+
+	/* allocate the pstore */
+	ps = kzalloc(sizeof(*ps), GFP_KERNEL);
+	if (!ps)
+		return -ENOMEM;
+
+	ps->store = store;
+	ps->valid = 1;
+	ps->version = SNAPSHOT_DISK_VERSION;
+	ps->area = NULL;
+	ps->zero_area = NULL;
+	ps->header_area = NULL;
+	ps->next_free = NUM_SNAPSHOT_HDR_CHUNKS + 1; /* header and 1st area */
+	ps->current_committed = 0;
+
+	ps->callback_count = 0;
+	atomic_set(&ps->pending_count, 0);
+	ps->callbacks = NULL;
+
+	ps->metadata_wq = alloc_workqueue("ksnaphd", WQ_MEM_RECLAIM, 0);
+	if (!ps->metadata_wq) {
+		kfree(ps);
+		DMERR("couldn't start header metadata update thread");
+		return -ENOMEM;
+	}
+
+	store->context = ps;
+
+	return 0;
+}
+
+static unsigned persistent_status(struct dm_exception_store *store,
+				  status_type_t status, char *result,
+				  unsigned maxlen)
+{
+	unsigned sz = 0;
+
+	switch (status) {
+	case STATUSTYPE_INFO:
+		break;
+	case STATUSTYPE_TABLE:
+		DMEMIT(" P %llu", (unsigned long long)store->chunk_size);
+	}
+
+	return sz;
+}
+
+static struct dm_exception_store_type _persistent_type = {
+	.name = "persistent",
+	.module = THIS_MODULE,
+	.ctr = persistent_ctr,
+	.dtr = persistent_dtr,
+	.read_metadata = persistent_read_metadata,
+	.prepare_exception = persistent_prepare_exception,
+	.commit_exception = persistent_commit_exception,
+	.prepare_merge = persistent_prepare_merge,
+	.commit_merge = persistent_commit_merge,
+	.drop_snapshot = persistent_drop_snapshot,
+	.usage = persistent_usage,
+	.status = persistent_status,
+};
+
+static struct dm_exception_store_type _persistent_compat_type = {
+	.name = "P",
+	.module = THIS_MODULE,
+	.ctr = persistent_ctr,
+	.dtr = persistent_dtr,
+	.read_metadata = persistent_read_metadata,
+	.prepare_exception = persistent_prepare_exception,
+	.commit_exception = persistent_commit_exception,
+	.prepare_merge = persistent_prepare_merge,
+	.commit_merge = persistent_commit_merge,
+	.drop_snapshot = persistent_drop_snapshot,
+	.usage = persistent_usage,
+	.status = persistent_status,
+};
+
+int dm_persistent_snapshot_init(void)
+{
+	int r;
+
+	r = dm_exception_store_type_register(&_persistent_type);
+	if (r) {
+		DMERR("Unable to register persistent exception store type");
+		return r;
+	}
+
+	r = dm_exception_store_type_register(&_persistent_compat_type);
+	if (r) {
+		DMERR("Unable to register old-style persistent exception "
+		      "store type");
+		dm_exception_store_type_unregister(&_persistent_type);
+		return r;
+	}
+
+	return r;
+}
+
+void dm_persistent_snapshot_exit(void)
+{
+	dm_exception_store_type_unregister(&_persistent_type);
+	dm_exception_store_type_unregister(&_persistent_compat_type);
+}
diff --git a/drivers/md/dm-snap-transient.c b/drivers/md/dm-snap-transient.c
new file mode 100644
index 00000000000..1ce9a2586e4
--- /dev/null
+++ b/drivers/md/dm-snap-transient.c
@@ -0,0 +1,153 @@
+/*
+ * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
+ * Copyright (C) 2006-2008 Red Hat GmbH
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-exception-store.h"
+
+#include <linux/mm.h>
+#include <linux/pagemap.h>
+#include <linux/vmalloc.h>
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/dm-io.h>
+
+#define DM_MSG_PREFIX "transient snapshot"
+
+/*-----------------------------------------------------------------
+ * Implementation of the store for non-persistent snapshots.
+ *---------------------------------------------------------------*/
+struct transient_c {
+	sector_t next_free;
+};
+
+static void transient_dtr(struct dm_exception_store *store)
+{
+	kfree(store->context);
+}
+
+static int transient_read_metadata(struct dm_exception_store *store,
+				   int (*callback)(void *callback_context,
+						   chunk_t old, chunk_t new),
+				   void *callback_context)
+{
+	return 0;
+}
+
+static int transient_prepare_exception(struct dm_exception_store *store,
+				       struct dm_exception *e)
+{
+	struct transient_c *tc = store->context;
+	sector_t size = get_dev_size(dm_snap_cow(store->snap)->bdev);
+
+	if (size < (tc->next_free + store->chunk_size))
+		return -1;
+
+	e->new_chunk = sector_to_chunk(store, tc->next_free);
+	tc->next_free += store->chunk_size;
+
+	return 0;
+}
+
+static void transient_commit_exception(struct dm_exception_store *store,
+				       struct dm_exception *e,
+				       void (*callback) (void *, int success),
+				       void *callback_context)
+{
+	/* Just succeed */
+	callback(callback_context, 1);
+}
+
+static void transient_usage(struct dm_exception_store *store,
+			    sector_t *total_sectors,
+			    sector_t *sectors_allocated,
+			    sector_t *metadata_sectors)
+{
+	*sectors_allocated = ((struct transient_c *) store->context)->next_free;
+	*total_sectors = get_dev_size(dm_snap_cow(store->snap)->bdev);
+	*metadata_sectors = 0;
+}
+
+static int transient_ctr(struct dm_exception_store *store,
+			 unsigned argc, char **argv)
+{
+	struct transient_c *tc;
+
+	tc = kmalloc(sizeof(struct transient_c), GFP_KERNEL);
+	if (!tc)
+		return -ENOMEM;
+
+	tc->next_free = 0;
+	store->context = tc;
+
+	return 0;
+}
+
+static unsigned transient_status(struct dm_exception_store *store,
+				 status_type_t status, char *result,
+				 unsigned maxlen)
+{
+	unsigned sz = 0;
+
+	switch (status) {
+	case STATUSTYPE_INFO:
+		break;
+	case STATUSTYPE_TABLE:
+		DMEMIT(" N %llu", (unsigned long long)store->chunk_size);
+	}
+
+	return sz;
+}
+
+static struct dm_exception_store_type _transient_type = {
+	.name = "transient",
+	.module = THIS_MODULE,
+	.ctr = transient_ctr,
+	.dtr = transient_dtr,
+	.read_metadata = transient_read_metadata,
+	.prepare_exception = transient_prepare_exception,
+	.commit_exception = transient_commit_exception,
+	.usage = transient_usage,
+	.status = transient_status,
+};
+
+static struct dm_exception_store_type _transient_compat_type = {
+	.name = "N",
+	.module = THIS_MODULE,
+	.ctr = transient_ctr,
+	.dtr = transient_dtr,
+	.read_metadata = transient_read_metadata,
+	.prepare_exception = transient_prepare_exception,
+	.commit_exception = transient_commit_exception,
+	.usage = transient_usage,
+	.status = transient_status,
+};
+
+int dm_transient_snapshot_init(void)
+{
+	int r;
+
+	r = dm_exception_store_type_register(&_transient_type);
+	if (r) {
+		DMWARN("Unable to register transient exception store type");
+		return r;
+	}
+
+	r = dm_exception_store_type_register(&_transient_compat_type);
+	if (r) {
+		DMWARN("Unable to register old-style transient "
+		       "exception store type");
+		dm_exception_store_type_unregister(&_transient_type);
+		return r;
+	}
+
+	return r;
+}
+
+void dm_transient_snapshot_exit(void)
+{
+	dm_exception_store_type_unregister(&_transient_type);
+	dm_exception_store_type_unregister(&_transient_compat_type);
+}
diff --git a/drivers/md/dm-snap.c b/drivers/md/dm-snap.c
index 6e5528aecc9..5bd2290cfb1 100644
--- a/drivers/md/dm-snap.c
+++ b/drivers/md/dm-snap.c
@@ -7,8 +7,8 @@
  */
 
 #include <linux/blkdev.h>
-#include <linux/ctype.h>
 #include <linux/device-mapper.h>
+#include <linux/delay.h>
 #include <linux/fs.h>
 #include <linux/init.h>
 #include <linux/kdev_t.h>
@@ -20,65 +20,161 @@
 #include <linux/log2.h>
 #include <linux/dm-kcopyd.h>
 
-#include "dm-snap.h"
-#include "dm-bio-list.h"
+#include "dm-exception-store.h"
 
 #define DM_MSG_PREFIX "snapshots"
 
-/*
- * The percentage increment we will wake up users at
- */
-#define WAKE_UP_PERCENT 5
+static const char dm_snapshot_merge_target_name[] = "snapshot-merge";
 
-/*
- * kcopyd priority of snapshot operations
- */
-#define SNAPSHOT_COPY_PRIORITY 2
-
-/*
- * Reserve 1MB for each snapshot initially (with minimum of 1 page).
- */
-#define SNAPSHOT_PAGES (((1UL << 20) >> PAGE_SHIFT) ? : 1)
+#define dm_target_is_snapshot_merge(ti) \
+	((ti)->type->name == dm_snapshot_merge_target_name)
 
 /*
  * The size of the mempool used to track chunks in use.
  */
 #define MIN_IOS 256
 
-static struct workqueue_struct *ksnapd;
-static void flush_queued_bios(struct work_struct *work);
+#define DM_TRACKED_CHUNK_HASH_SIZE	16
+#define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
+					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
 
-struct dm_snap_pending_exception {
-	struct dm_snap_exception e;
+struct dm_exception_table {
+	uint32_t hash_mask;
+	unsigned hash_shift;
+	struct list_head *table;
+};
+
+struct dm_snapshot {
+	struct rw_semaphore lock;
+
+	struct dm_dev *origin;
+	struct dm_dev *cow;
+
+	struct dm_target *ti;
+
+	/* List of snapshots per Origin */
+	struct list_head list;
 
 	/*
-	 * Origin buffers waiting for this to complete are held
-	 * in a bio list
+	 * You can't use a snapshot if this is 0 (e.g. if full).
+	 * A snapshot-merge target never clears this.
 	 */
-	struct bio_list origin_bios;
-	struct bio_list snapshot_bios;
+	int valid;
+
+	/* Origin writes don't trigger exceptions until this is set */
+	int active;
+
+	atomic_t pending_exceptions_count;
+
+	/* Protected by "lock" */
+	sector_t exception_start_sequence;
+
+	/* Protected by kcopyd single-threaded callback */
+	sector_t exception_complete_sequence;
 
 	/*
-	 * Short-term queue of pending exceptions prior to submission.
+	 * A list of pending exceptions that completed out of order.
+	 * Protected by kcopyd single-threaded callback.
 	 */
-	struct list_head list;
+	struct list_head out_of_order_list;
+
+	mempool_t *pending_pool;
+
+	struct dm_exception_table pending;
+	struct dm_exception_table complete;
+
+	/*
+	 * pe_lock protects all pending_exception operations and access
+	 * as well as the snapshot_bios list.
+	 */
+	spinlock_t pe_lock;
+
+	/* Chunks with outstanding reads */
+	spinlock_t tracked_chunk_lock;
+	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
+
+	/* The on disk metadata handler */
+	struct dm_exception_store *store;
+
+	struct dm_kcopyd_client *kcopyd_client;
+
+	/* Wait for events based on state_bits */
+	unsigned long state_bits;
+
+	/* Range of chunks currently being merged. */
+	chunk_t first_merging_chunk;
+	int num_merging_chunks;
+
+	/*
+	 * The merge operation failed if this flag is set.
+	 * Failure modes are handled as follows:
+	 * - I/O error reading the header
+	 *   	=> don't load the target; abort.
+	 * - Header does not have "valid" flag set
+	 *   	=> use the origin; forget about the snapshot.
+	 * - I/O error when reading exceptions
+	 *   	=> don't load the target; abort.
+	 *         (We can't use the intermediate origin state.)
+	 * - I/O error while merging
+	 *	=> stop merging; set merge_failed; process I/O normally.
+	 */
+	int merge_failed;
+
+	/*
+	 * Incoming bios that overlap with chunks being merged must wait
+	 * for them to be committed.
+	 */
+	struct bio_list bios_queued_during_merge;
+};
+
+/*
+ * state_bits:
+ *   RUNNING_MERGE  - Merge operation is in progress.
+ *   SHUTDOWN_MERGE - Set to signal that merge needs to be stopped;
+ *                    cleared afterwards.
+ */
+#define RUNNING_MERGE          0
+#define SHUTDOWN_MERGE         1
+
+DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(snapshot_copy_throttle,
+		"A percentage of time allocated for copy on write");
+
+struct dm_dev *dm_snap_origin(struct dm_snapshot *s)
+{
+	return s->origin;
+}
+EXPORT_SYMBOL(dm_snap_origin);
+
+struct dm_dev *dm_snap_cow(struct dm_snapshot *s)
+{
+	return s->cow;
+}
+EXPORT_SYMBOL(dm_snap_cow);
+
+static sector_t chunk_to_sector(struct dm_exception_store *store,
+				chunk_t chunk)
+{
+	return chunk << store->chunk_shift;
+}
 
+static int bdev_equal(struct block_device *lhs, struct block_device *rhs)
+{
 	/*
-	 * The primary pending_exception is the one that holds
-	 * the ref_count and the list of origin_bios for a
-	 * group of pending_exceptions.  It is always last to get freed.
-	 * These fields get set up when writing to the origin.
+	 * There is only ever one instance of a particular block
+	 * device so we can compare pointers safely.
 	 */
-	struct dm_snap_pending_exception *primary_pe;
+	return lhs == rhs;
+}
+
+struct dm_snap_pending_exception {
+	struct dm_exception e;
 
 	/*
-	 * Number of pending_exceptions processing this chunk.
-	 * When this drops to zero we must complete the origin bios.
-	 * If incrementing or decrementing this, hold pe->snap->lock for
-	 * the sibling concerned and not pe->primary_pe->snap->lock unless
-	 * they are the same.
+	 * Origin buffers waiting for this to complete are held
+	 * in a bio list
 	 */
-	atomic_t ref_count;
+	struct bio_list origin_bios;
+	struct bio_list snapshot_bios;
 
 	/* Pointer back to snapshot context */
 	struct dm_snapshot *snap;
@@ -88,6 +184,21 @@ struct dm_snap_pending_exception {
 	 * kcopyd.
 	 */
 	int started;
+
+	/* There was copying error. */
+	int copy_error;
+
+	/* A sequence number, it is used for in-order completion. */
+	sector_t exception_sequence;
+
+	struct list_head out_of_order_entry;
+
+	/*
+	 * For writing a complete chunk, bypassing the copy.
+	 */
+	struct bio *full_bio;
+	bio_end_io_t *full_bio_end_io;
+	void *full_bio_private;
 };
 
 /*
@@ -102,46 +213,48 @@ struct dm_snap_tracked_chunk {
 	chunk_t chunk;
 };
 
-static struct kmem_cache *tracked_chunk_cache;
+static void init_tracked_chunk(struct bio *bio)
+{
+	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
+	INIT_HLIST_NODE(&c->node);
+}
 
-static struct dm_snap_tracked_chunk *track_chunk(struct dm_snapshot *s,
-						 chunk_t chunk)
+static bool is_bio_tracked(struct bio *bio)
 {
-	struct dm_snap_tracked_chunk *c = mempool_alloc(s->tracked_chunk_pool,
-							GFP_NOIO);
-	unsigned long flags;
+	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
+	return !hlist_unhashed(&c->node);
+}
+
+static void track_chunk(struct dm_snapshot *s, struct bio *bio, chunk_t chunk)
+{
+	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
 
 	c->chunk = chunk;
 
-	spin_lock_irqsave(&s->tracked_chunk_lock, flags);
+	spin_lock_irq(&s->tracked_chunk_lock);
 	hlist_add_head(&c->node,
 		       &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)]);
-	spin_unlock_irqrestore(&s->tracked_chunk_lock, flags);
-
-	return c;
+	spin_unlock_irq(&s->tracked_chunk_lock);
 }
 
-static void stop_tracking_chunk(struct dm_snapshot *s,
-				struct dm_snap_tracked_chunk *c)
+static void stop_tracking_chunk(struct dm_snapshot *s, struct bio *bio)
 {
+	struct dm_snap_tracked_chunk *c = dm_per_bio_data(bio, sizeof(struct dm_snap_tracked_chunk));
 	unsigned long flags;
 
 	spin_lock_irqsave(&s->tracked_chunk_lock, flags);
 	hlist_del(&c->node);
 	spin_unlock_irqrestore(&s->tracked_chunk_lock, flags);
-
-	mempool_free(c, s->tracked_chunk_pool);
 }
 
 static int __chunk_is_tracked(struct dm_snapshot *s, chunk_t chunk)
 {
 	struct dm_snap_tracked_chunk *c;
-	struct hlist_node *hn;
 	int found = 0;
 
 	spin_lock_irq(&s->tracked_chunk_lock);
 
-	hlist_for_each_entry(c, hn,
+	hlist_for_each_entry(c,
 	    &s->tracked_chunk_hash[DM_TRACKED_CHUNK_HASH(chunk)], node) {
 		if (c->chunk == chunk) {
 			found = 1;
@@ -155,6 +268,16 @@ static int __chunk_is_tracked(struct dm_snapshot *s, chunk_t chunk)
 }
 
 /*
+ * This conflicting I/O is extremely improbable in the caller,
+ * so msleep(1) is sufficient and there is no need for a wait queue.
+ */
+static void __check_for_conflicting_io(struct dm_snapshot *s, chunk_t chunk)
+{
+	while (__chunk_is_tracked(s, chunk))
+		msleep(1);
+}
+
+/*
  * One of these per registered origin, held in the snapshot_origins hash
  */
 struct origin {
@@ -176,6 +299,10 @@ struct origin {
 static struct list_head *_origins;
 static struct rw_semaphore _origins_lock;
 
+static DECLARE_WAIT_QUEUE_HEAD(_pending_exceptions_done);
+static DEFINE_SPINLOCK(_pending_exceptions_done_spinlock);
+static uint64_t _pending_exceptions_done_count;
+
 static int init_origin_hash(void)
 {
 	int i;
@@ -224,24 +351,149 @@ static void __insert_origin(struct origin *o)
 }
 
 /*
+ * _origins_lock must be held when calling this function.
+ * Returns number of snapshots registered using the supplied cow device, plus:
+ * snap_src - a snapshot suitable for use as a source of exception handover
+ * snap_dest - a snapshot capable of receiving exception handover.
+ * snap_merge - an existing snapshot-merge target linked to the same origin.
+ *   There can be at most one snapshot-merge target. The parameter is optional.
+ *
+ * Possible return values and states of snap_src and snap_dest.
+ *   0: NULL, NULL  - first new snapshot
+ *   1: snap_src, NULL - normal snapshot
+ *   2: snap_src, snap_dest  - waiting for handover
+ *   2: snap_src, NULL - handed over, waiting for old to be deleted
+ *   1: NULL, snap_dest - source got destroyed without handover
+ */
+static int __find_snapshots_sharing_cow(struct dm_snapshot *snap,
+					struct dm_snapshot **snap_src,
+					struct dm_snapshot **snap_dest,
+					struct dm_snapshot **snap_merge)
+{
+	struct dm_snapshot *s;
+	struct origin *o;
+	int count = 0;
+	int active;
+
+	o = __lookup_origin(snap->origin->bdev);
+	if (!o)
+		goto out;
+
+	list_for_each_entry(s, &o->snapshots, list) {
+		if (dm_target_is_snapshot_merge(s->ti) && snap_merge)
+			*snap_merge = s;
+		if (!bdev_equal(s->cow->bdev, snap->cow->bdev))
+			continue;
+
+		down_read(&s->lock);
+		active = s->active;
+		up_read(&s->lock);
+
+		if (active) {
+			if (snap_src)
+				*snap_src = s;
+		} else if (snap_dest)
+			*snap_dest = s;
+
+		count++;
+	}
+
+out:
+	return count;
+}
+
+/*
+ * On success, returns 1 if this snapshot is a handover destination,
+ * otherwise returns 0.
+ */
+static int __validate_exception_handover(struct dm_snapshot *snap)
+{
+	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;
+	struct dm_snapshot *snap_merge = NULL;
+
+	/* Does snapshot need exceptions handed over to it? */
+	if ((__find_snapshots_sharing_cow(snap, &snap_src, &snap_dest,
+					  &snap_merge) == 2) ||
+	    snap_dest) {
+		snap->ti->error = "Snapshot cow pairing for exception "
+				  "table handover failed";
+		return -EINVAL;
+	}
+
+	/*
+	 * If no snap_src was found, snap cannot become a handover
+	 * destination.
+	 */
+	if (!snap_src)
+		return 0;
+
+	/*
+	 * Non-snapshot-merge handover?
+	 */
+	if (!dm_target_is_snapshot_merge(snap->ti))
+		return 1;
+
+	/*
+	 * Do not allow more than one merging snapshot.
+	 */
+	if (snap_merge) {
+		snap->ti->error = "A snapshot is already merging.";
+		return -EINVAL;
+	}
+
+	if (!snap_src->store->type->prepare_merge ||
+	    !snap_src->store->type->commit_merge) {
+		snap->ti->error = "Snapshot exception store does not "
+				  "support snapshot-merge.";
+		return -EINVAL;
+	}
+
+	return 1;
+}
+
+static void __insert_snapshot(struct origin *o, struct dm_snapshot *s)
+{
+	struct dm_snapshot *l;
+
+	/* Sort the list according to chunk size, largest-first smallest-last */
+	list_for_each_entry(l, &o->snapshots, list)
+		if (l->store->chunk_size < s->store->chunk_size)
+			break;
+	list_add_tail(&s->list, &l->list);
+}
+
+/*
  * Make a note of the snapshot and its origin so we can look it
  * up when the origin has a write on it.
+ *
+ * Also validate snapshot exception store handovers.
+ * On success, returns 1 if this registration is a handover destination,
+ * otherwise returns 0.
  */
 static int register_snapshot(struct dm_snapshot *snap)
 {
-	struct origin *o;
+	struct origin *o, *new_o = NULL;
 	struct block_device *bdev = snap->origin->bdev;
+	int r = 0;
+
+	new_o = kmalloc(sizeof(*new_o), GFP_KERNEL);
+	if (!new_o)
+		return -ENOMEM;
 
 	down_write(&_origins_lock);
-	o = __lookup_origin(bdev);
 
-	if (!o) {
+	r = __validate_exception_handover(snap);
+	if (r < 0) {
+		kfree(new_o);
+		goto out;
+	}
+
+	o = __lookup_origin(bdev);
+	if (o)
+		kfree(new_o);
+	else {
 		/* New origin */
-		o = kmalloc(sizeof(*o), GFP_KERNEL);
-		if (!o) {
-			up_write(&_origins_lock);
-			return -ENOMEM;
-		}
+		o = new_o;
 
 		/* Initialise the struct */
 		INIT_LIST_HEAD(&o->snapshots);
@@ -250,10 +502,27 @@ static int register_snapshot(struct dm_snapshot *snap)
 		__insert_origin(o);
 	}
 
-	list_add_tail(&snap->list, &o->snapshots);
+	__insert_snapshot(o, snap);
+
+out:
+	up_write(&_origins_lock);
+
+	return r;
+}
+
+/*
+ * Move snapshot to correct place in list according to chunk size.
+ */
+static void reregister_snapshot(struct dm_snapshot *s)
+{
+	struct block_device *bdev = s->origin->bdev;
+
+	down_write(&_origins_lock);
+
+	list_del(&s->list);
+	__insert_snapshot(__lookup_origin(bdev), s);
 
 	up_write(&_origins_lock);
-	return 0;
 }
 
 static void unregister_snapshot(struct dm_snapshot *s)
@@ -264,7 +533,7 @@ static void unregister_snapshot(struct dm_snapshot *s)
 	o = __lookup_origin(s->origin->bdev);
 
 	list_del(&s->list);
-	if (list_empty(&o->snapshots)) {
+	if (o && list_empty(&o->snapshots)) {
 		list_del(&o->hash_list);
 		kfree(o);
 	}
@@ -277,8 +546,8 @@ static void unregister_snapshot(struct dm_snapshot *s)
  * The lowest hash_shift bits of the chunk number are ignored, allowing
  * some consecutive chunks to be grouped together.
  */
-static int init_exception_table(struct exception_table *et, uint32_t size,
-				unsigned hash_shift)
+static int dm_exception_table_init(struct dm_exception_table *et,
+				   uint32_t size, unsigned hash_shift)
 {
 	unsigned int i;
 
@@ -294,10 +563,11 @@ static int init_exception_table(struct exception_table *et, uint32_t size,
 	return 0;
 }
 
-static void exit_exception_table(struct exception_table *et, struct kmem_cache *mem)
+static void dm_exception_table_exit(struct dm_exception_table *et,
+				    struct kmem_cache *mem)
 {
 	struct list_head *slot;
-	struct dm_snap_exception *ex, *next;
+	struct dm_exception *ex, *next;
 	int i, size;
 
 	size = et->hash_mask + 1;
@@ -311,19 +581,12 @@ static void exit_exception_table(struct exception_table *et, struct kmem_cache *
 	vfree(et->table);
 }
 
-static uint32_t exception_hash(struct exception_table *et, chunk_t chunk)
+static uint32_t exception_hash(struct dm_exception_table *et, chunk_t chunk)
 {
 	return (chunk >> et->hash_shift) & et->hash_mask;
 }
 
-static void insert_exception(struct exception_table *eh,
-			     struct dm_snap_exception *e)
-{
-	struct list_head *l = &eh->table[exception_hash(eh, e->old_chunk)];
-	list_add(&e->hash_list, l);
-}
-
-static void remove_exception(struct dm_snap_exception *e)
+static void dm_remove_exception(struct dm_exception *e)
 {
 	list_del(&e->hash_list);
 }
@@ -332,11 +595,11 @@ static void remove_exception(struct dm_snap_exception *e)
  * Return the exception data for a sector, or NULL if not
  * remapped.
  */
-static struct dm_snap_exception *lookup_exception(struct exception_table *et,
-						  chunk_t chunk)
+static struct dm_exception *dm_lookup_exception(struct dm_exception_table *et,
+						chunk_t chunk)
 {
 	struct list_head *slot;
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 
 	slot = &et->table[exception_hash(et, chunk)];
 	list_for_each_entry (e, slot, hash_list)
@@ -347,18 +610,18 @@ static struct dm_snap_exception *lookup_exception(struct exception_table *et,
 	return NULL;
 }
 
-static struct dm_snap_exception *alloc_exception(void)
+static struct dm_exception *alloc_completed_exception(gfp_t gfp)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 
-	e = kmem_cache_alloc(exception_cache, GFP_NOIO);
-	if (!e)
+	e = kmem_cache_alloc(exception_cache, gfp);
+	if (!e && gfp == GFP_NOIO)
 		e = kmem_cache_alloc(exception_cache, GFP_ATOMIC);
 
 	return e;
 }
 
-static void free_exception(struct dm_snap_exception *e)
+static void free_completed_exception(struct dm_exception *e)
 {
 	kmem_cache_free(exception_cache, e);
 }
@@ -368,6 +631,7 @@ static struct dm_snap_pending_exception *alloc_pending_exception(struct dm_snaps
 	struct dm_snap_pending_exception *pe = mempool_alloc(s->pending_pool,
 							     GFP_NOIO);
 
+	atomic_inc(&s->pending_exceptions_count);
 	pe->snap = s;
 
 	return pe;
@@ -375,15 +639,18 @@ static struct dm_snap_pending_exception *alloc_pending_exception(struct dm_snaps
 
 static void free_pending_exception(struct dm_snap_pending_exception *pe)
 {
-	mempool_free(pe, pe->snap->pending_pool);
+	struct dm_snapshot *s = pe->snap;
+
+	mempool_free(pe, s->pending_pool);
+	smp_mb__before_atomic();
+	atomic_dec(&s->pending_exceptions_count);
 }
 
-static void insert_completed_exception(struct dm_snapshot *s,
-				       struct dm_snap_exception *new_e)
+static void dm_insert_exception(struct dm_exception_table *eh,
+				struct dm_exception *new_e)
 {
-	struct exception_table *eh = &s->complete;
 	struct list_head *l;
-	struct dm_snap_exception *e = NULL;
+	struct dm_exception *e = NULL;
 
 	l = &eh->table[exception_hash(eh, new_e->old_chunk)];
 
@@ -399,7 +666,7 @@ static void insert_completed_exception(struct dm_snapshot *s,
 		    new_e->new_chunk == (dm_chunk_number(e->new_chunk) +
 					 dm_consecutive_chunk_count(e) + 1)) {
 			dm_consecutive_chunk_count_inc(e);
-			free_exception(new_e);
+			free_completed_exception(new_e);
 			return;
 		}
 
@@ -409,7 +676,7 @@ static void insert_completed_exception(struct dm_snapshot *s,
 			dm_consecutive_chunk_count_inc(e);
 			e->old_chunk--;
 			e->new_chunk--;
-			free_exception(new_e);
+			free_completed_exception(new_e);
 			return;
 		}
 
@@ -421,11 +688,16 @@ out:
 	list_add(&new_e->hash_list, e ? &e->hash_list : l);
 }
 
-int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
+/*
+ * Callback used by the exception stores to load exceptions when
+ * initialising.
+ */
+static int dm_add_exception(void *context, chunk_t old, chunk_t new)
 {
-	struct dm_snap_exception *e;
+	struct dm_snapshot *s = context;
+	struct dm_exception *e;
 
-	e = alloc_exception();
+	e = alloc_completed_exception(GFP_KERNEL);
 	if (!e)
 		return -ENOMEM;
 
@@ -434,12 +706,29 @@ int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new)
 	/* Consecutive_count is implicitly initialised to zero */
 	e->new_chunk = new;
 
-	insert_completed_exception(s, e);
+	dm_insert_exception(&s->complete, e);
 
 	return 0;
 }
 
 /*
+ * Return a minimum chunk size of all snapshots that have the specified origin.
+ * Return zero if the origin has no snapshots.
+ */
+static uint32_t __minimum_chunk_size(struct origin *o)
+{
+	struct dm_snapshot *snap;
+	unsigned chunk_size = 0;
+
+	if (o)
+		list_for_each_entry(snap, &o->snapshots, list)
+			chunk_size = min_not_zero(chunk_size,
+						  snap->store->chunk_size);
+
+	return (uint32_t) chunk_size;
+}
+
+/*
  * Hard coded magic.
  */
 static int calc_max_buckets(void)
@@ -456,22 +745,23 @@ static int calc_max_buckets(void)
  */
 static int init_hash_tables(struct dm_snapshot *s)
 {
-	sector_t hash_size, cow_dev_size, origin_dev_size, max_buckets;
+	sector_t hash_size, cow_dev_size, max_buckets;
 
 	/*
 	 * Calculate based on the size of the original volume or
 	 * the COW volume...
 	 */
 	cow_dev_size = get_dev_size(s->cow->bdev);
-	origin_dev_size = get_dev_size(s->origin->bdev);
 	max_buckets = calc_max_buckets();
 
-	hash_size = min(origin_dev_size, cow_dev_size) >> s->chunk_shift;
+	hash_size = cow_dev_size >> s->store->chunk_shift;
 	hash_size = min(hash_size, max_buckets);
 
+	if (hash_size < 64)
+		hash_size = 64;
 	hash_size = rounddown_pow_of_two(hash_size);
-	if (init_exception_table(&s->complete, hash_size,
-				 DM_CHUNK_CONSECUTIVE_BITS))
+	if (dm_exception_table_init(&s->complete, hash_size,
+				    DM_CHUNK_CONSECUTIVE_BITS))
 		return -ENOMEM;
 
 	/*
@@ -482,67 +772,285 @@ static int init_hash_tables(struct dm_snapshot *s)
 	if (hash_size < 64)
 		hash_size = 64;
 
-	if (init_exception_table(&s->pending, hash_size, 0)) {
-		exit_exception_table(&s->complete, exception_cache);
+	if (dm_exception_table_init(&s->pending, hash_size, 0)) {
+		dm_exception_table_exit(&s->complete, exception_cache);
 		return -ENOMEM;
 	}
 
 	return 0;
 }
 
-/*
- * Round a number up to the nearest 'size' boundary.  size must
- * be a power of 2.
- */
-static ulong round_up(ulong n, ulong size)
+static void merge_shutdown(struct dm_snapshot *s)
+{
+	clear_bit_unlock(RUNNING_MERGE, &s->state_bits);
+	smp_mb__after_atomic();
+	wake_up_bit(&s->state_bits, RUNNING_MERGE);
+}
+
+static struct bio *__release_queued_bios_after_merge(struct dm_snapshot *s)
 {
-	size--;
-	return (n + size) & ~size;
+	s->first_merging_chunk = 0;
+	s->num_merging_chunks = 0;
+
+	return bio_list_get(&s->bios_queued_during_merge);
 }
 
-static int set_chunk_size(struct dm_snapshot *s, const char *chunk_size_arg,
-			  char **error)
+/*
+ * Remove one chunk from the index of completed exceptions.
+ */
+static int __remove_single_exception_chunk(struct dm_snapshot *s,
+					   chunk_t old_chunk)
 {
-	unsigned long chunk_size;
-	char *value;
+	struct dm_exception *e;
 
-	chunk_size = simple_strtoul(chunk_size_arg, &value, 10);
-	if (*chunk_size_arg == '\0' || *value != '\0') {
-		*error = "Invalid chunk size";
+	e = dm_lookup_exception(&s->complete, old_chunk);
+	if (!e) {
+		DMERR("Corruption detected: exception for block %llu is "
+		      "on disk but not in memory",
+		      (unsigned long long)old_chunk);
 		return -EINVAL;
 	}
 
-	if (!chunk_size) {
-		s->chunk_size = s->chunk_mask = s->chunk_shift = 0;
+	/*
+	 * If this is the only chunk using this exception, remove exception.
+	 */
+	if (!dm_consecutive_chunk_count(e)) {
+		dm_remove_exception(e);
+		free_completed_exception(e);
 		return 0;
 	}
 
 	/*
-	 * Chunk size must be multiple of page size.  Silently
-	 * round up if it's not.
+	 * The chunk may be either at the beginning or the end of a
+	 * group of consecutive chunks - never in the middle.  We are
+	 * removing chunks in the opposite order to that in which they
+	 * were added, so this should always be true.
+	 * Decrement the consecutive chunk counter and adjust the
+	 * starting point if necessary.
 	 */
-	chunk_size = round_up(chunk_size, PAGE_SIZE >> 9);
-
-	/* Check chunk_size is a power of 2 */
-	if (!is_power_of_2(chunk_size)) {
-		*error = "Chunk size is not a power of 2";
+	if (old_chunk == e->old_chunk) {
+		e->old_chunk++;
+		e->new_chunk++;
+	} else if (old_chunk != e->old_chunk +
+		   dm_consecutive_chunk_count(e)) {
+		DMERR("Attempt to merge block %llu from the "
+		      "middle of a chunk range [%llu - %llu]",
+		      (unsigned long long)old_chunk,
+		      (unsigned long long)e->old_chunk,
+		      (unsigned long long)
+		      e->old_chunk + dm_consecutive_chunk_count(e));
 		return -EINVAL;
 	}
 
-	/* Validate the chunk size against the device block size */
-	if (chunk_size % (bdev_hardsect_size(s->cow->bdev) >> 9)) {
-		*error = "Chunk size is not a multiple of device blocksize";
-		return -EINVAL;
+	dm_consecutive_chunk_count_dec(e);
+
+	return 0;
+}
+
+static void flush_bios(struct bio *bio);
+
+static int remove_single_exception_chunk(struct dm_snapshot *s)
+{
+	struct bio *b = NULL;
+	int r;
+	chunk_t old_chunk = s->first_merging_chunk + s->num_merging_chunks - 1;
+
+	down_write(&s->lock);
+
+	/*
+	 * Process chunks (and associated exceptions) in reverse order
+	 * so that dm_consecutive_chunk_count_dec() accounting works.
+	 */
+	do {
+		r = __remove_single_exception_chunk(s, old_chunk);
+		if (r)
+			goto out;
+	} while (old_chunk-- > s->first_merging_chunk);
+
+	b = __release_queued_bios_after_merge(s);
+
+out:
+	up_write(&s->lock);
+	if (b)
+		flush_bios(b);
+
+	return r;
+}
+
+static int origin_write_extent(struct dm_snapshot *merging_snap,
+			       sector_t sector, unsigned chunk_size);
+
+static void merge_callback(int read_err, unsigned long write_err,
+			   void *context);
+
+static uint64_t read_pending_exceptions_done_count(void)
+{
+	uint64_t pending_exceptions_done;
+
+	spin_lock(&_pending_exceptions_done_spinlock);
+	pending_exceptions_done = _pending_exceptions_done_count;
+	spin_unlock(&_pending_exceptions_done_spinlock);
+
+	return pending_exceptions_done;
+}
+
+static void increment_pending_exceptions_done_count(void)
+{
+	spin_lock(&_pending_exceptions_done_spinlock);
+	_pending_exceptions_done_count++;
+	spin_unlock(&_pending_exceptions_done_spinlock);
+
+	wake_up_all(&_pending_exceptions_done);
+}
+
+static void snapshot_merge_next_chunks(struct dm_snapshot *s)
+{
+	int i, linear_chunks;
+	chunk_t old_chunk, new_chunk;
+	struct dm_io_region src, dest;
+	sector_t io_size;
+	uint64_t previous_count;
+
+	BUG_ON(!test_bit(RUNNING_MERGE, &s->state_bits));
+	if (unlikely(test_bit(SHUTDOWN_MERGE, &s->state_bits)))
+		goto shut;
+
+	/*
+	 * valid flag never changes during merge, so no lock required.
+	 */
+	if (!s->valid) {
+		DMERR("Snapshot is invalid: can't merge");
+		goto shut;
+	}
+
+	linear_chunks = s->store->type->prepare_merge(s->store, &old_chunk,
+						      &new_chunk);
+	if (linear_chunks <= 0) {
+		if (linear_chunks < 0) {
+			DMERR("Read error in exception store: "
+			      "shutting down merge");
+			down_write(&s->lock);
+			s->merge_failed = 1;
+			up_write(&s->lock);
+		}
+		goto shut;
+	}
+
+	/* Adjust old_chunk and new_chunk to reflect start of linear region */
+	old_chunk = old_chunk + 1 - linear_chunks;
+	new_chunk = new_chunk + 1 - linear_chunks;
+
+	/*
+	 * Use one (potentially large) I/O to copy all 'linear_chunks'
+	 * from the exception store to the origin
+	 */
+	io_size = linear_chunks * s->store->chunk_size;
+
+	dest.bdev = s->origin->bdev;
+	dest.sector = chunk_to_sector(s->store, old_chunk);
+	dest.count = min(io_size, get_dev_size(dest.bdev) - dest.sector);
+
+	src.bdev = s->cow->bdev;
+	src.sector = chunk_to_sector(s->store, new_chunk);
+	src.count = dest.count;
+
+	/*
+	 * Reallocate any exceptions needed in other snapshots then
+	 * wait for the pending exceptions to complete.
+	 * Each time any pending exception (globally on the system)
+	 * completes we are woken and repeat the process to find out
+	 * if we can proceed.  While this may not seem a particularly
+	 * efficient algorithm, it is not expected to have any
+	 * significant impact on performance.
+	 */
+	previous_count = read_pending_exceptions_done_count();
+	while (origin_write_extent(s, dest.sector, io_size)) {
+		wait_event(_pending_exceptions_done,
+			   (read_pending_exceptions_done_count() !=
+			    previous_count));
+		/* Retry after the wait, until all exceptions are done. */
+		previous_count = read_pending_exceptions_done_count();
+	}
+
+	down_write(&s->lock);
+	s->first_merging_chunk = old_chunk;
+	s->num_merging_chunks = linear_chunks;
+	up_write(&s->lock);
+
+	/* Wait until writes to all 'linear_chunks' drain */
+	for (i = 0; i < linear_chunks; i++)
+		__check_for_conflicting_io(s, old_chunk + i);
+
+	dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, merge_callback, s);
+	return;
+
+shut:
+	merge_shutdown(s);
+}
+
+static void error_bios(struct bio *bio);
+
+static void merge_callback(int read_err, unsigned long write_err, void *context)
+{
+	struct dm_snapshot *s = context;
+	struct bio *b = NULL;
+
+	if (read_err || write_err) {
+		if (read_err)
+			DMERR("Read error: shutting down merge.");
+		else
+			DMERR("Write error: shutting down merge.");
+		goto shut;
 	}
 
-	s->chunk_size = chunk_size;
-	s->chunk_mask = chunk_size - 1;
-	s->chunk_shift = ffs(chunk_size) - 1;
+	if (s->store->type->commit_merge(s->store,
+					 s->num_merging_chunks) < 0) {
+		DMERR("Write error in exception store: shutting down merge");
+		goto shut;
+	}
+
+	if (remove_single_exception_chunk(s) < 0)
+		goto shut;
+
+	snapshot_merge_next_chunks(s);
+
+	return;
+
+shut:
+	down_write(&s->lock);
+	s->merge_failed = 1;
+	b = __release_queued_bios_after_merge(s);
+	up_write(&s->lock);
+	error_bios(b);
+
+	merge_shutdown(s);
+}
+
+static void start_merge(struct dm_snapshot *s)
+{
+	if (!test_and_set_bit(RUNNING_MERGE, &s->state_bits))
+		snapshot_merge_next_chunks(s);
+}
+
+static int wait_schedule(void *ptr)
+{
+	schedule();
 
 	return 0;
 }
 
 /*
+ * Stop the merging process and wait until it finishes.
+ */
+static void stop_merge(struct dm_snapshot *s)
+{
+	set_bit(SHUTDOWN_MERGE, &s->state_bits);
+	wait_on_bit(&s->state_bits, RUNNING_MERGE, wait_schedule,
+		    TASK_UNINTERRUPTIBLE);
+	clear_bit(SHUTDOWN_MERGE, &s->state_bits);
+}
+
+/*
  * Construct a snapshot mapping: <origin_dev> <COW-dev> <p/n> <chunk-size>
  */
 static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
@@ -550,99 +1058,93 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 	struct dm_snapshot *s;
 	int i;
 	int r = -EINVAL;
-	char persistent;
-	char *origin_path;
-	char *cow_path;
+	char *origin_path, *cow_path;
+	unsigned args_used, num_flush_bios = 1;
+	fmode_t origin_mode = FMODE_READ;
 
 	if (argc != 4) {
 		ti->error = "requires exactly 4 arguments";
 		r = -EINVAL;
-		goto bad1;
+		goto bad;
 	}
 
-	origin_path = argv[0];
-	cow_path = argv[1];
-	persistent = toupper(*argv[2]);
-
-	if (persistent != 'P' && persistent != 'N') {
-		ti->error = "Persistent flag is not P or N";
-		r = -EINVAL;
-		goto bad1;
+	if (dm_target_is_snapshot_merge(ti)) {
+		num_flush_bios = 2;
+		origin_mode = FMODE_WRITE;
 	}
 
 	s = kmalloc(sizeof(*s), GFP_KERNEL);
-	if (s == NULL) {
-		ti->error = "Cannot allocate snapshot context private "
-		    "structure";
+	if (!s) {
+		ti->error = "Cannot allocate private snapshot structure";
 		r = -ENOMEM;
-		goto bad1;
+		goto bad;
 	}
 
-	r = dm_get_device(ti, origin_path, 0, ti->len, FMODE_READ, &s->origin);
+	origin_path = argv[0];
+	argv++;
+	argc--;
+
+	r = dm_get_device(ti, origin_path, origin_mode, &s->origin);
 	if (r) {
 		ti->error = "Cannot get origin device";
-		goto bad2;
+		goto bad_origin;
 	}
 
-	r = dm_get_device(ti, cow_path, 0, 0,
-			  FMODE_READ | FMODE_WRITE, &s->cow);
+	cow_path = argv[0];
+	argv++;
+	argc--;
+
+	r = dm_get_device(ti, cow_path, dm_table_get_mode(ti->table), &s->cow);
 	if (r) {
-		dm_put_device(ti, s->origin);
 		ti->error = "Cannot get COW device";
-		goto bad2;
+		goto bad_cow;
 	}
 
-	r = set_chunk_size(s, argv[3], &ti->error);
-	if (r)
-		goto bad3;
+	r = dm_exception_store_create(ti, argc, argv, s, &args_used, &s->store);
+	if (r) {
+		ti->error = "Couldn't create exception store";
+		r = -EINVAL;
+		goto bad_store;
+	}
 
-	s->type = persistent;
+	argv += args_used;
+	argc -= args_used;
 
+	s->ti = ti;
 	s->valid = 1;
 	s->active = 0;
-	s->last_percent = 0;
+	atomic_set(&s->pending_exceptions_count, 0);
+	s->exception_start_sequence = 0;
+	s->exception_complete_sequence = 0;
+	INIT_LIST_HEAD(&s->out_of_order_list);
 	init_rwsem(&s->lock);
+	INIT_LIST_HEAD(&s->list);
 	spin_lock_init(&s->pe_lock);
-	s->ti = ti;
+	s->state_bits = 0;
+	s->merge_failed = 0;
+	s->first_merging_chunk = 0;
+	s->num_merging_chunks = 0;
+	bio_list_init(&s->bios_queued_during_merge);
 
 	/* Allocate hash table for COW data */
 	if (init_hash_tables(s)) {
 		ti->error = "Unable to allocate hash table space";
 		r = -ENOMEM;
-		goto bad3;
+		goto bad_hash_tables;
 	}
 
-	s->store.snap = s;
-
-	if (persistent == 'P')
-		r = dm_create_persistent(&s->store);
-	else
-		r = dm_create_transient(&s->store);
-
-	if (r) {
-		ti->error = "Couldn't create exception store";
-		r = -EINVAL;
-		goto bad4;
-	}
-
-	r = dm_kcopyd_client_create(SNAPSHOT_PAGES, &s->kcopyd_client);
-	if (r) {
+	s->kcopyd_client = dm_kcopyd_client_create(&dm_kcopyd_throttle);
+	if (IS_ERR(s->kcopyd_client)) {
+		r = PTR_ERR(s->kcopyd_client);
 		ti->error = "Could not create kcopyd client";
-		goto bad5;
+		goto bad_kcopyd;
 	}
 
 	s->pending_pool = mempool_create_slab_pool(MIN_IOS, pending_cache);
 	if (!s->pending_pool) {
 		ti->error = "Could not allocate mempool for pending exceptions";
-		goto bad6;
-	}
-
-	s->tracked_chunk_pool = mempool_create_slab_pool(MIN_IOS,
-							 tracked_chunk_cache);
-	if (!s->tracked_chunk_pool) {
-		ti->error = "Could not allocate tracked_chunk mempool for "
-			    "tracking reads";
-		goto bad_tracked_chunk_pool;
+		r = -ENOMEM;
+		goto bad_pending_pool;
 	}
 
 	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
@@ -650,56 +1152,79 @@ static int snapshot_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 
 	spin_lock_init(&s->tracked_chunk_lock);
 
-	/* Metadata must only be loaded into one table at once */
-	r = s->store.read_metadata(&s->store);
+	ti->private = s;
+	ti->num_flush_bios = num_flush_bios;
+	ti->per_bio_data_size = sizeof(struct dm_snap_tracked_chunk);
+
+	/* Add snapshot to the list of snapshots for this origin */
+	/* Exceptions aren't triggered till snapshot_resume() is called */
+	r = register_snapshot(s);
+	if (r == -ENOMEM) {
+		ti->error = "Snapshot origin struct allocation failed";
+		goto bad_load_and_register;
+	} else if (r < 0) {
+		/* invalid handover, register_snapshot has set ti->error */
+		goto bad_load_and_register;
+	}
+
+	/*
+	 * Metadata must only be loaded into one table at once, so skip this
+	 * if metadata will be handed over during resume.
+	 * Chunk size will be set during the handover - set it to zero to
+	 * ensure it's ignored.
+	 */
+	if (r > 0) {
+		s->store->chunk_size = 0;
+		return 0;
+	}
+
+	r = s->store->type->read_metadata(s->store, dm_add_exception,
+					  (void *)s);
 	if (r < 0) {
 		ti->error = "Failed to read snapshot metadata";
-		goto bad_load_and_register;
+		goto bad_read_metadata;
 	} else if (r > 0) {
 		s->valid = 0;
 		DMWARN("Snapshot is marked invalid.");
 	}
 
-	bio_list_init(&s->queued_bios);
-	INIT_WORK(&s->queued_bios_work, flush_queued_bios);
-
-	/* Add snapshot to the list of snapshots for this origin */
-	/* Exceptions aren't triggered till snapshot_resume() is called */
-	if (register_snapshot(s)) {
-		r = -EINVAL;
-		ti->error = "Cannot register snapshot origin";
-		goto bad_load_and_register;
+	if (!s->store->chunk_size) {
+		ti->error = "Chunk size not set";
+		goto bad_read_metadata;
 	}
 
-	ti->private = s;
-	ti->split_io = s->chunk_size;
+	r = dm_set_target_max_io_len(ti, s->store->chunk_size);
+	if (r)
+		goto bad_read_metadata;
 
 	return 0;
 
- bad_load_and_register:
-	mempool_destroy(s->tracked_chunk_pool);
+bad_read_metadata:
+	unregister_snapshot(s);
 
- bad_tracked_chunk_pool:
+bad_load_and_register:
 	mempool_destroy(s->pending_pool);
 
- bad6:
+bad_pending_pool:
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 
- bad5:
-	s->store.destroy(&s->store);
+bad_kcopyd:
+	dm_exception_table_exit(&s->pending, pending_cache);
+	dm_exception_table_exit(&s->complete, exception_cache);
 
- bad4:
-	exit_exception_table(&s->pending, pending_cache);
-	exit_exception_table(&s->complete, exception_cache);
+bad_hash_tables:
+	dm_exception_store_destroy(s->store);
 
- bad3:
+bad_store:
 	dm_put_device(ti, s->cow);
+
+bad_cow:
 	dm_put_device(ti, s->origin);
 
- bad2:
+bad_origin:
 	kfree(s);
 
- bad1:
+bad:
 	return r;
 }
 
@@ -708,10 +1233,39 @@ static void __free_exceptions(struct dm_snapshot *s)
 	dm_kcopyd_client_destroy(s->kcopyd_client);
 	s->kcopyd_client = NULL;
 
-	exit_exception_table(&s->pending, pending_cache);
-	exit_exception_table(&s->complete, exception_cache);
+	dm_exception_table_exit(&s->pending, pending_cache);
+	dm_exception_table_exit(&s->complete, exception_cache);
+}
+
+static void __handover_exceptions(struct dm_snapshot *snap_src,
+				  struct dm_snapshot *snap_dest)
+{
+	union {
+		struct dm_exception_table table_swap;
+		struct dm_exception_store *store_swap;
+	} u;
+
+	/*
+	 * Swap all snapshot context information between the two instances.
+	 */
+	u.table_swap = snap_dest->complete;
+	snap_dest->complete = snap_src->complete;
+	snap_src->complete = u.table_swap;
+
+	u.store_swap = snap_dest->store;
+	snap_dest->store = snap_src->store;
+	snap_src->store = u.store_swap;
+
+	snap_dest->store->snap = snap_dest;
+	snap_src->store->snap = snap_src;
 
-	s->store.destroy(&s->store);
+	snap_dest->ti->max_io_len = snap_dest->store->chunk_size;
+	snap_dest->valid = snap_src->valid;
+
+	/*
+	 * Set source invalid to ensure it receives no further I/O.
+	 */
+	snap_src->valid = 0;
 }
 
 static void snapshot_dtr(struct dm_target *ti)
@@ -720,27 +1274,49 @@ static void snapshot_dtr(struct dm_target *ti)
 	int i;
 #endif
 	struct dm_snapshot *s = ti->private;
+	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;
+
+	down_read(&_origins_lock);
+	/* Check whether exception handover must be cancelled */
+	(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL);
+	if (snap_src && snap_dest && (s == snap_src)) {
+		down_write(&snap_dest->lock);
+		snap_dest->valid = 0;
+		up_write(&snap_dest->lock);
+		DMERR("Cancelling snapshot handover.");
+	}
+	up_read(&_origins_lock);
 
-	flush_workqueue(ksnapd);
+	if (dm_target_is_snapshot_merge(ti))
+		stop_merge(s);
 
 	/* Prevent further origin writes from using this snapshot. */
 	/* After this returns there can be no new kcopyd jobs. */
 	unregister_snapshot(s);
 
+	while (atomic_read(&s->pending_exceptions_count))
+		msleep(1);
+	/*
+	 * Ensure instructions in mempool_destroy aren't reordered
+	 * before atomic_read.
+	 */
+	smp_mb();
+
 #ifdef CONFIG_DM_DEBUG
 	for (i = 0; i < DM_TRACKED_CHUNK_HASH_SIZE; i++)
 		BUG_ON(!hlist_empty(&s->tracked_chunk_hash[i]));
 #endif
 
-	mempool_destroy(s->tracked_chunk_pool);
-
 	__free_exceptions(s);
 
 	mempool_destroy(s->pending_pool);
 
-	dm_put_device(ti, s->origin);
+	dm_exception_store_destroy(s->store);
+
 	dm_put_device(ti, s->cow);
 
+	dm_put_device(ti, s->origin);
+
 	kfree(s);
 }
 
@@ -759,18 +1335,24 @@ static void flush_bios(struct bio *bio)
 	}
 }
 
-static void flush_queued_bios(struct work_struct *work)
-{
-	struct dm_snapshot *s =
-		container_of(work, struct dm_snapshot, queued_bios_work);
-	struct bio *queued_bios;
-	unsigned long flags;
+static int do_origin(struct dm_dev *origin, struct bio *bio);
 
-	spin_lock_irqsave(&s->pe_lock, flags);
-	queued_bios = bio_list_get(&s->queued_bios);
-	spin_unlock_irqrestore(&s->pe_lock, flags);
+/*
+ * Flush a list of buffers.
+ */
+static void retry_origin_bios(struct dm_snapshot *s, struct bio *bio)
+{
+	struct bio *n;
+	int r;
 
-	flush_bios(queued_bios);
+	while (bio) {
+		n = bio->bi_next;
+		bio->bi_next = NULL;
+		r = do_origin(s->origin, bio);
+		if (r == DM_MAPIO_REMAPPED)
+			generic_make_request(bio);
+		bio = n;
+	}
 }
 
 /*
@@ -798,57 +1380,21 @@ static void __invalidate_snapshot(struct dm_snapshot *s, int err)
 	else if (err == -ENOMEM)
 		DMERR("Invalidating snapshot: Unable to allocate exception.");
 
-	if (s->store.drop_snapshot)
-		s->store.drop_snapshot(&s->store);
+	if (s->store->type->drop_snapshot)
+		s->store->type->drop_snapshot(s->store);
 
 	s->valid = 0;
 
 	dm_table_event(s->ti->table);
 }
 
-static void get_pending_exception(struct dm_snap_pending_exception *pe)
-{
-	atomic_inc(&pe->ref_count);
-}
-
-static struct bio *put_pending_exception(struct dm_snap_pending_exception *pe)
-{
-	struct dm_snap_pending_exception *primary_pe;
-	struct bio *origin_bios = NULL;
-
-	primary_pe = pe->primary_pe;
-
-	/*
-	 * If this pe is involved in a write to the origin and
-	 * it is the last sibling to complete then release
-	 * the bios for the original write to the origin.
-	 */
-	if (primary_pe &&
-	    atomic_dec_and_test(&primary_pe->ref_count))
-		origin_bios = bio_list_get(&primary_pe->origin_bios);
-
-	/*
-	 * Free the pe if it's not linked to an origin write or if
-	 * it's not itself a primary pe.
-	 */
-	if (!primary_pe || primary_pe != pe)
-		free_pending_exception(pe);
-
-	/*
-	 * Free the primary pe if nothing references it.
-	 */
-	if (primary_pe && !atomic_read(&primary_pe->ref_count))
-		free_pending_exception(primary_pe);
-
-	return origin_bios;
-}
-
 static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 	struct dm_snapshot *s = pe->snap;
 	struct bio *origin_bios = NULL;
 	struct bio *snapshot_bios = NULL;
+	struct bio *full_bio = NULL;
 	int error = 0;
 
 	if (!success) {
@@ -859,7 +1405,7 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 		goto out;
 	}
 
-	e = alloc_exception();
+	e = alloc_completed_exception(GFP_NOIO);
 	if (!e) {
 		down_write(&s->lock);
 		__invalidate_snapshot(s, -ENOMEM);
@@ -870,38 +1416,48 @@ static void pending_complete(struct dm_snap_pending_exception *pe, int success)
 
 	down_write(&s->lock);
 	if (!s->valid) {
-		free_exception(e);
+		free_completed_exception(e);
 		error = 1;
 		goto out;
 	}
 
-	/*
-	 * Check for conflicting reads. This is extremely improbable,
-	 * so yield() is sufficient and there is no need for a wait queue.
-	 */
-	while (__chunk_is_tracked(s, pe->e.old_chunk))
-		yield();
+	/* Check for conflicting reads */
+	__check_for_conflicting_io(s, pe->e.old_chunk);
 
 	/*
 	 * Add a proper exception, and remove the
 	 * in-flight exception from the list.
 	 */
-	insert_completed_exception(s, e);
+	dm_insert_exception(&s->complete, e);
 
- out:
-	remove_exception(&pe->e);
+out:
+	dm_remove_exception(&pe->e);
 	snapshot_bios = bio_list_get(&pe->snapshot_bios);
-	origin_bios = put_pending_exception(pe);
+	origin_bios = bio_list_get(&pe->origin_bios);
+	full_bio = pe->full_bio;
+	if (full_bio) {
+		full_bio->bi_end_io = pe->full_bio_end_io;
+		full_bio->bi_private = pe->full_bio_private;
+		atomic_inc(&full_bio->bi_remaining);
+	}
+	free_pending_exception(pe);
+
+	increment_pending_exceptions_done_count();
 
 	up_write(&s->lock);
 
 	/* Submit any pending write bios */
-	if (error)
+	if (error) {
+		if (full_bio)
+			bio_io_error(full_bio);
 		error_bios(snapshot_bios);
-	else
+	} else {
+		if (full_bio)
+			bio_endio(full_bio, 0);
 		flush_bios(snapshot_bios);
+	}
 
-	flush_bios(origin_bios);
+	retry_origin_bios(s, origin_bios);
 }
 
 static void commit_callback(void *context, int success)
@@ -911,6 +1467,19 @@ static void commit_callback(void *context, int success)
 	pending_complete(pe, success);
 }
 
+static void complete_exception(struct dm_snap_pending_exception *pe)
+{
+	struct dm_snapshot *s = pe->snap;
+
+	if (unlikely(pe->copy_error))
+		pending_complete(pe, 0);
+
+	else
+		/* Update the metadata if we are persistent */
+		s->store->type->commit_exception(s->store, &pe->e,
+						 commit_callback, pe);
+}
+
 /*
  * Called when the copy I/O has finished.  kcopyd actually runs
  * this code so don't block.
@@ -920,13 +1489,32 @@ static void copy_callback(int read_err, unsigned long write_err, void *context)
 	struct dm_snap_pending_exception *pe = context;
 	struct dm_snapshot *s = pe->snap;
 
-	if (read_err || write_err)
-		pending_complete(pe, 0);
+	pe->copy_error = read_err || write_err;
 
-	else
-		/* Update the metadata if we are persistent */
-		s->store.commit_exception(&s->store, &pe->e, commit_callback,
-					  pe);
+	if (pe->exception_sequence == s->exception_complete_sequence) {
+		s->exception_complete_sequence++;
+		complete_exception(pe);
+
+		while (!list_empty(&s->out_of_order_list)) {
+			pe = list_entry(s->out_of_order_list.next,
+					struct dm_snap_pending_exception, out_of_order_entry);
+			if (pe->exception_sequence != s->exception_complete_sequence)
+				break;
+			s->exception_complete_sequence++;
+			list_del(&pe->out_of_order_entry);
+			complete_exception(pe);
+		}
+	} else {
+		struct list_head *lh;
+		struct dm_snap_pending_exception *pe2;
+
+		list_for_each_prev(lh, &s->out_of_order_list) {
+			pe2 = list_entry(lh, struct dm_snap_pending_exception, out_of_order_entry);
+			if (pe2->exception_sequence < pe->exception_sequence)
+				break;
+		}
+		list_add(&pe->out_of_order_entry, lh);
+	}
 }
 
 /*
@@ -942,16 +1530,52 @@ static void start_copy(struct dm_snap_pending_exception *pe)
 	dev_size = get_dev_size(bdev);
 
 	src.bdev = bdev;
-	src.sector = chunk_to_sector(s, pe->e.old_chunk);
-	src.count = min(s->chunk_size, dev_size - src.sector);
+	src.sector = chunk_to_sector(s->store, pe->e.old_chunk);
+	src.count = min((sector_t)s->store->chunk_size, dev_size - src.sector);
 
 	dest.bdev = s->cow->bdev;
-	dest.sector = chunk_to_sector(s, pe->e.new_chunk);
+	dest.sector = chunk_to_sector(s->store, pe->e.new_chunk);
 	dest.count = src.count;
 
 	/* Hand over to kcopyd */
-	dm_kcopyd_copy(s->kcopyd_client,
-		    &src, 1, &dest, 0, copy_callback, pe);
+	dm_kcopyd_copy(s->kcopyd_client, &src, 1, &dest, 0, copy_callback, pe);
+}
+
+static void full_bio_end_io(struct bio *bio, int error)
+{
+	void *callback_data = bio->bi_private;
+
+	dm_kcopyd_do_callback(callback_data, 0, error ? 1 : 0);
+}
+
+static void start_full_bio(struct dm_snap_pending_exception *pe,
+			   struct bio *bio)
+{
+	struct dm_snapshot *s = pe->snap;
+	void *callback_data;
+
+	pe->full_bio = bio;
+	pe->full_bio_end_io = bio->bi_end_io;
+	pe->full_bio_private = bio->bi_private;
+
+	callback_data = dm_kcopyd_prepare_callback(s->kcopyd_client,
+						   copy_callback, pe);
+
+	bio->bi_end_io = full_bio_end_io;
+	bio->bi_private = callback_data;
+
+	generic_make_request(bio);
+}
+
+static struct dm_snap_pending_exception *
+__lookup_pending_exception(struct dm_snapshot *s, chunk_t chunk)
+{
+	struct dm_exception *e = dm_lookup_exception(&s->pending, chunk);
+
+	if (!e)
+		return NULL;
+
+	return container_of(e, struct dm_snap_pending_exception, e);
 }
 
 /*
@@ -963,80 +1587,61 @@ static void start_copy(struct dm_snap_pending_exception *pe)
  * this.
  */
 static struct dm_snap_pending_exception *
-__find_pending_exception(struct dm_snapshot *s, struct bio *bio)
+__find_pending_exception(struct dm_snapshot *s,
+			 struct dm_snap_pending_exception *pe, chunk_t chunk)
 {
-	struct dm_snap_exception *e;
-	struct dm_snap_pending_exception *pe;
-	chunk_t chunk = sector_to_chunk(s, bio->bi_sector);
-
-	/*
-	 * Is there a pending exception for this already ?
-	 */
-	e = lookup_exception(&s->pending, chunk);
-	if (e) {
-		/* cast the exception to a pending exception */
-		pe = container_of(e, struct dm_snap_pending_exception, e);
-		goto out;
-	}
-
-	/*
-	 * Create a new pending exception, we don't want
-	 * to hold the lock while we do this.
-	 */
-	up_write(&s->lock);
-	pe = alloc_pending_exception(s);
-	down_write(&s->lock);
-
-	if (!s->valid) {
-		free_pending_exception(pe);
-		return NULL;
-	}
+	struct dm_snap_pending_exception *pe2;
 
-	e = lookup_exception(&s->pending, chunk);
-	if (e) {
+	pe2 = __lookup_pending_exception(s, chunk);
+	if (pe2) {
 		free_pending_exception(pe);
-		pe = container_of(e, struct dm_snap_pending_exception, e);
-		goto out;
+		return pe2;
 	}
 
 	pe->e.old_chunk = chunk;
 	bio_list_init(&pe->origin_bios);
 	bio_list_init(&pe->snapshot_bios);
-	pe->primary_pe = NULL;
-	atomic_set(&pe->ref_count, 0);
 	pe->started = 0;
+	pe->full_bio = NULL;
 
-	if (s->store.prepare_exception(&s->store, &pe->e)) {
+	if (s->store->type->prepare_exception(s->store, &pe->e)) {
 		free_pending_exception(pe);
 		return NULL;
 	}
 
-	get_pending_exception(pe);
-	insert_exception(&s->pending, &pe->e);
+	pe->exception_sequence = s->exception_start_sequence++;
+
+	dm_insert_exception(&s->pending, &pe->e);
 
- out:
 	return pe;
 }
 
-static void remap_exception(struct dm_snapshot *s, struct dm_snap_exception *e,
+static void remap_exception(struct dm_snapshot *s, struct dm_exception *e,
 			    struct bio *bio, chunk_t chunk)
 {
 	bio->bi_bdev = s->cow->bdev;
-	bio->bi_sector = chunk_to_sector(s, dm_chunk_number(e->new_chunk) +
-			 (chunk - e->old_chunk)) +
-			 (bio->bi_sector & s->chunk_mask);
+	bio->bi_iter.bi_sector =
+		chunk_to_sector(s->store, dm_chunk_number(e->new_chunk) +
+				(chunk - e->old_chunk)) +
+		(bio->bi_iter.bi_sector & s->store->chunk_mask);
 }
 
-static int snapshot_map(struct dm_target *ti, struct bio *bio,
-			union map_info *map_context)
+static int snapshot_map(struct dm_target *ti, struct bio *bio)
 {
-	struct dm_snap_exception *e;
+	struct dm_exception *e;
 	struct dm_snapshot *s = ti->private;
 	int r = DM_MAPIO_REMAPPED;
 	chunk_t chunk;
 	struct dm_snap_pending_exception *pe = NULL;
 
-	chunk = sector_to_chunk(s, bio->bi_sector);
+	init_tracked_chunk(bio);
+
+	if (bio->bi_rw & REQ_FLUSH) {
+		bio->bi_bdev = s->cow->bdev;
+		return DM_MAPIO_REMAPPED;
+	}
+
+	chunk = sector_to_chunk(s->store, bio->bi_iter.bi_sector);
 
 	/* Full snapshots are not usable */
 	/* To get here the table must be live so s->active is always set. */
@@ -1053,7 +1658,7 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	}
 
 	/* If the block is already remapped - use that, else remap it */
-	e = lookup_exception(&s->complete, chunk);
+	e = dm_lookup_exception(&s->complete, chunk);
 	if (e) {
 		remap_exception(s, e, bio, chunk);
 		goto out_unlock;
@@ -1065,18 +1670,48 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 	 * writeable.
 	 */
 	if (bio_rw(bio) == WRITE) {
-		pe = __find_pending_exception(s, bio);
+		pe = __lookup_pending_exception(s, chunk);
 		if (!pe) {
-			__invalidate_snapshot(s, -ENOMEM);
-			r = -EIO;
-			goto out_unlock;
+			up_write(&s->lock);
+			pe = alloc_pending_exception(s);
+			down_write(&s->lock);
+
+			if (!s->valid) {
+				free_pending_exception(pe);
+				r = -EIO;
+				goto out_unlock;
+			}
+
+			e = dm_lookup_exception(&s->complete, chunk);
+			if (e) {
+				free_pending_exception(pe);
+				remap_exception(s, e, bio, chunk);
+				goto out_unlock;
+			}
+
+			pe = __find_pending_exception(s, pe, chunk);
+			if (!pe) {
+				__invalidate_snapshot(s, -ENOMEM);
+				r = -EIO;
+				goto out_unlock;
+			}
 		}
 
 		remap_exception(s, &pe->e, bio, chunk);
-		bio_list_add(&pe->snapshot_bios, bio);
 
 		r = DM_MAPIO_SUBMITTED;
 
+		if (!pe->started &&
+		    bio->bi_iter.bi_size ==
+		    (s->store->chunk_size << SECTOR_SHIFT)) {
+			pe->started = 1;
+			up_write(&s->lock);
+			start_full_bio(pe, bio);
+			goto out;
+		}
+
+		bio_list_add(&pe->snapshot_bios, bio);
+
 		if (!pe->started) {
 			/* this is protected by snap->lock */
 			pe->started = 1;
@@ -1086,58 +1721,216 @@ static int snapshot_map(struct dm_target *ti, struct bio *bio,
 		}
 	} else {
 		bio->bi_bdev = s->origin->bdev;
-		map_context->ptr = track_chunk(s, chunk);
+		track_chunk(s, bio, chunk);
 	}
 
- out_unlock:
+out_unlock:
 	up_write(&s->lock);
- out:
+out:
+	return r;
+}
+
+/*
+ * A snapshot-merge target behaves like a combination of a snapshot
+ * target and a snapshot-origin target.  It only generates new
+ * exceptions in other snapshots and not in the one that is being
+ * merged.
+ *
+ * For each chunk, if there is an existing exception, it is used to
+ * redirect I/O to the cow device.  Otherwise I/O is sent to the origin,
+ * which in turn might generate exceptions in other snapshots.
+ * If merging is currently taking place on the chunk in question, the
+ * I/O is deferred by adding it to s->bios_queued_during_merge.
+ */
+static int snapshot_merge_map(struct dm_target *ti, struct bio *bio)
+{
+	struct dm_exception *e;
+	struct dm_snapshot *s = ti->private;
+	int r = DM_MAPIO_REMAPPED;
+	chunk_t chunk;
+
+	init_tracked_chunk(bio);
+
+	if (bio->bi_rw & REQ_FLUSH) {
+		if (!dm_bio_get_target_bio_nr(bio))
+			bio->bi_bdev = s->origin->bdev;
+		else
+			bio->bi_bdev = s->cow->bdev;
+		return DM_MAPIO_REMAPPED;
+	}
+
+	chunk = sector_to_chunk(s->store, bio->bi_iter.bi_sector);
+
+	down_write(&s->lock);
+
+	/* Full merging snapshots are redirected to the origin */
+	if (!s->valid)
+		goto redirect_to_origin;
+
+	/* If the block is already remapped - use that */
+	e = dm_lookup_exception(&s->complete, chunk);
+	if (e) {
+		/* Queue writes overlapping with chunks being merged */
+		if (bio_rw(bio) == WRITE &&
+		    chunk >= s->first_merging_chunk &&
+		    chunk < (s->first_merging_chunk +
+			     s->num_merging_chunks)) {
+			bio->bi_bdev = s->origin->bdev;
+			bio_list_add(&s->bios_queued_during_merge, bio);
+			r = DM_MAPIO_SUBMITTED;
+			goto out_unlock;
+		}
+
+		remap_exception(s, e, bio, chunk);
+
+		if (bio_rw(bio) == WRITE)
+			track_chunk(s, bio, chunk);
+		goto out_unlock;
+	}
+
+redirect_to_origin:
+	bio->bi_bdev = s->origin->bdev;
+
+	if (bio_rw(bio) == WRITE) {
+		up_write(&s->lock);
+		return do_origin(s->origin, bio);
+	}
+
+out_unlock:
+	up_write(&s->lock);
+
 	return r;
 }
 
-static int snapshot_end_io(struct dm_target *ti, struct bio *bio,
-			   int error, union map_info *map_context)
+static int snapshot_end_io(struct dm_target *ti, struct bio *bio, int error)
 {
 	struct dm_snapshot *s = ti->private;
-	struct dm_snap_tracked_chunk *c = map_context->ptr;
 
-	if (c)
-		stop_tracking_chunk(s, c);
+	if (is_bio_tracked(bio))
+		stop_tracking_chunk(s, bio);
 
 	return 0;
 }
 
+static void snapshot_merge_presuspend(struct dm_target *ti)
+{
+	struct dm_snapshot *s = ti->private;
+
+	stop_merge(s);
+}
+
+static int snapshot_preresume(struct dm_target *ti)
+{
+	int r = 0;
+	struct dm_snapshot *s = ti->private;
+	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;
+
+	down_read(&_origins_lock);
+	(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL);
+	if (snap_src && snap_dest) {
+		down_read(&snap_src->lock);
+		if (s == snap_src) {
+			DMERR("Unable to resume snapshot source until "
+			      "handover completes.");
+			r = -EINVAL;
+		} else if (!dm_suspended(snap_src->ti)) {
+			DMERR("Unable to perform snapshot handover until "
+			      "source is suspended.");
+			r = -EINVAL;
+		}
+		up_read(&snap_src->lock);
+	}
+	up_read(&_origins_lock);
+
+	return r;
+}
+
 static void snapshot_resume(struct dm_target *ti)
 {
 	struct dm_snapshot *s = ti->private;
+	struct dm_snapshot *snap_src = NULL, *snap_dest = NULL;
+
+	down_read(&_origins_lock);
+	(void) __find_snapshots_sharing_cow(s, &snap_src, &snap_dest, NULL);
+	if (snap_src && snap_dest) {
+		down_write(&snap_src->lock);
+		down_write_nested(&snap_dest->lock, SINGLE_DEPTH_NESTING);
+		__handover_exceptions(snap_src, snap_dest);
+		up_write(&snap_dest->lock);
+		up_write(&snap_src->lock);
+	}
+	up_read(&_origins_lock);
+
+	/* Now we have correct chunk size, reregister */
+	reregister_snapshot(s);
 
 	down_write(&s->lock);
 	s->active = 1;
 	up_write(&s->lock);
 }
 
-static int snapshot_status(struct dm_target *ti, status_type_t type,
-			   char *result, unsigned int maxlen)
+static uint32_t get_origin_minimum_chunksize(struct block_device *bdev)
 {
+	uint32_t min_chunksize;
+
+	down_read(&_origins_lock);
+	min_chunksize = __minimum_chunk_size(__lookup_origin(bdev));
+	up_read(&_origins_lock);
+
+	return min_chunksize;
+}
+
+static void snapshot_merge_resume(struct dm_target *ti)
+{
+	struct dm_snapshot *s = ti->private;
+
+	/*
+	 * Handover exceptions from existing snapshot.
+	 */
+	snapshot_resume(ti);
+
+	/*
+	 * snapshot-merge acts as an origin, so set ti->max_io_len
+	 */
+	ti->max_io_len = get_origin_minimum_chunksize(s->origin->bdev);
+
+	start_merge(s);
+}
+
+static void snapshot_status(struct dm_target *ti, status_type_t type,
+			    unsigned status_flags, char *result, unsigned maxlen)
+{
+	unsigned sz = 0;
 	struct dm_snapshot *snap = ti->private;
 
 	switch (type) {
 	case STATUSTYPE_INFO:
+
+		down_write(&snap->lock);
+
 		if (!snap->valid)
-			snprintf(result, maxlen, "Invalid");
+			DMEMIT("Invalid");
+		else if (snap->merge_failed)
+			DMEMIT("Merge failed");
 		else {
-			if (snap->store.fraction_full) {
-				sector_t numerator, denominator;
-				snap->store.fraction_full(&snap->store,
-							  &numerator,
-							  &denominator);
-				snprintf(result, maxlen, "%llu/%llu",
-					(unsigned long long)numerator,
-					(unsigned long long)denominator);
+			if (snap->store->type->usage) {
+				sector_t total_sectors, sectors_allocated,
+					 metadata_sectors;
+				snap->store->type->usage(snap->store,
+							 &total_sectors,
+							 &sectors_allocated,
+							 &metadata_sectors);
+				DMEMIT("%llu/%llu %llu",
+				       (unsigned long long)sectors_allocated,
+				       (unsigned long long)total_sectors,
+				       (unsigned long long)metadata_sectors);
 			}
 			else
-				snprintf(result, maxlen, "Unknown");
+				DMEMIT("Unknown");
 		}
+
+		up_write(&snap->lock);
+
 		break;
 
 	case STATUSTYPE_TABLE:
@@ -1146,30 +1939,61 @@ static int snapshot_status(struct dm_target *ti, status_type_t type,
 		 * to make private copies if the output is to
 		 * make sense.
 		 */
-		snprintf(result, maxlen, "%s %s %c %llu",
-			 snap->origin->name, snap->cow->name,
-			 snap->type,
-			 (unsigned long long)snap->chunk_size);
+		DMEMIT("%s %s", snap->origin->name, snap->cow->name);
+		snap->store->type->status(snap->store, type, result + sz,
+					  maxlen - sz);
 		break;
 	}
+}
 
-	return 0;
+static int snapshot_iterate_devices(struct dm_target *ti,
+				    iterate_devices_callout_fn fn, void *data)
+{
+	struct dm_snapshot *snap = ti->private;
+	int r;
+
+	r = fn(ti, snap->origin, 0, ti->len, data);
+
+	if (!r)
+		r = fn(ti, snap->cow, 0, get_dev_size(snap->cow->bdev), data);
+
+	return r;
 }
 
+
 /*-----------------------------------------------------------------
  * Origin methods
  *---------------------------------------------------------------*/
-static int __origin_write(struct list_head *snapshots, struct bio *bio)
+
+/*
+ * If no exceptions need creating, DM_MAPIO_REMAPPED is returned and any
+ * supplied bio was ignored.  The caller may submit it immediately.
+ * (No remapping actually occurs as the origin is always a direct linear
+ * map.)
+ *
+ * If further exceptions are required, DM_MAPIO_SUBMITTED is returned
+ * and any supplied bio is added to a list to be submitted once all
+ * the necessary exceptions exist.
+ */
+static int __origin_write(struct list_head *snapshots, sector_t sector,
+			  struct bio *bio)
 {
-	int r = DM_MAPIO_REMAPPED, first = 0;
+	int r = DM_MAPIO_REMAPPED;
 	struct dm_snapshot *snap;
-	struct dm_snap_exception *e;
-	struct dm_snap_pending_exception *pe, *next_pe, *primary_pe = NULL;
+	struct dm_exception *e;
+	struct dm_snap_pending_exception *pe;
+	struct dm_snap_pending_exception *pe_to_start_now = NULL;
+	struct dm_snap_pending_exception *pe_to_start_last = NULL;
 	chunk_t chunk;
-	LIST_HEAD(pe_queue);
 
 	/* Do all the snapshots on this origin */
 	list_for_each_entry (snap, snapshots, list) {
+		/*
+		 * Don't make new exceptions in a merging snapshot
+		 * because it has effectively been deleted
+		 */
+		if (dm_target_is_snapshot_merge(snap->ti))
+			continue;
 
 		down_write(&snap->lock);
 
@@ -1178,86 +2002,85 @@ static int __origin_write(struct list_head *snapshots, struct bio *bio)
 			goto next_snapshot;
 
 		/* Nothing to do if writing beyond end of snapshot */
-		if (bio->bi_sector >= dm_table_get_size(snap->ti->table))
+		if (sector >= dm_table_get_size(snap->ti->table))
 			goto next_snapshot;
 
 		/*
 		 * Remember, different snapshots can have
 		 * different chunk sizes.
 		 */
-		chunk = sector_to_chunk(snap, bio->bi_sector);
+		chunk = sector_to_chunk(snap->store, sector);
 
 		/*
 		 * Check exception table to see if block
 		 * is already remapped in this snapshot
 		 * and trigger an exception if not.
-		 *
-		 * ref_count is initialised to 1 so pending_complete()
-		 * won't destroy the primary_pe while we're inside this loop.
 		 */
-		e = lookup_exception(&snap->complete, chunk);
+		e = dm_lookup_exception(&snap->complete, chunk);
 		if (e)
 			goto next_snapshot;
 
-		pe = __find_pending_exception(snap, bio);
+		pe = __lookup_pending_exception(snap, chunk);
 		if (!pe) {
-			__invalidate_snapshot(snap, -ENOMEM);
-			goto next_snapshot;
-		}
+			up_write(&snap->lock);
+			pe = alloc_pending_exception(snap);
+			down_write(&snap->lock);
 
-		if (!primary_pe) {
-			/*
-			 * Either every pe here has same
-			 * primary_pe or none has one yet.
-			 */
-			if (pe->primary_pe)
-				primary_pe = pe->primary_pe;
-			else {
-				primary_pe = pe;
-				first = 1;
+			if (!snap->valid) {
+				free_pending_exception(pe);
+				goto next_snapshot;
 			}
 
-			bio_list_add(&primary_pe->origin_bios, bio);
+			e = dm_lookup_exception(&snap->complete, chunk);
+			if (e) {
+				free_pending_exception(pe);
+				goto next_snapshot;
+			}
 
-			r = DM_MAPIO_SUBMITTED;
+			pe = __find_pending_exception(snap, pe, chunk);
+			if (!pe) {
+				__invalidate_snapshot(snap, -ENOMEM);
+				goto next_snapshot;
+			}
 		}
 
-		if (!pe->primary_pe) {
-			pe->primary_pe = primary_pe;
-			get_pending_exception(primary_pe);
+		r = DM_MAPIO_SUBMITTED;
+
+		/*
+		 * If an origin bio was supplied, queue it to wait for the
+		 * completion of this exception, and start this one last,
+		 * at the end of the function.
+		 */
+		if (bio) {
+			bio_list_add(&pe->origin_bios, bio);
+			bio = NULL;
+
+			if (!pe->started) {
+				pe->started = 1;
+				pe_to_start_last = pe;
+			}
 		}
 
 		if (!pe->started) {
 			pe->started = 1;
-			list_add_tail(&pe->list, &pe_queue);
+			pe_to_start_now = pe;
 		}
 
- next_snapshot:
+next_snapshot:
 		up_write(&snap->lock);
-	}
-
-	if (!primary_pe)
-		return r;
-
-	/*
-	 * If this is the first time we're processing this chunk and
-	 * ref_count is now 1 it means all the pending exceptions
-	 * got completed while we were in the loop above, so it falls to
-	 * us here to remove the primary_pe and submit any origin_bios.
-	 */
 
-	if (first && atomic_dec_and_test(&primary_pe->ref_count)) {
-		flush_bios(bio_list_get(&primary_pe->origin_bios));
-		free_pending_exception(primary_pe);
-		/* If we got here, pe_queue is necessarily empty. */
-		return r;
+		if (pe_to_start_now) {
+			start_copy(pe_to_start_now);
+			pe_to_start_now = NULL;
+		}
 	}
 
 	/*
-	 * Now that we have a complete pe list we can start the copying.
+	 * Submit the exception against which the bio is queued last,
+	 * to give the other exceptions a head start.
 	 */
-	list_for_each_entry_safe(pe, next_pe, &pe_queue, list)
-		start_copy(pe);
+	if (pe_to_start_last)
+		start_copy(pe_to_start_last);
 
 	return r;
 }
@@ -1273,16 +2096,56 @@ static int do_origin(struct dm_dev *origin, struct bio *bio)
 	down_read(&_origins_lock);
 	o = __lookup_origin(origin->bdev);
 	if (o)
-		r = __origin_write(&o->snapshots, bio);
+		r = __origin_write(&o->snapshots, bio->bi_iter.bi_sector, bio);
 	up_read(&_origins_lock);
 
 	return r;
 }
 
 /*
+ * Trigger exceptions in all non-merging snapshots.
+ *
+ * The chunk size of the merging snapshot may be larger than the chunk
+ * size of some other snapshot so we may need to reallocate multiple
+ * chunks in other snapshots.
+ *
+ * We scan all the overlapping exceptions in the other snapshots.
+ * Returns 1 if anything was reallocated and must be waited for,
+ * otherwise returns 0.
+ *
+ * size must be a multiple of merging_snap's chunk_size.
+ */
+static int origin_write_extent(struct dm_snapshot *merging_snap,
+			       sector_t sector, unsigned size)
+{
+	int must_wait = 0;
+	sector_t n;
+	struct origin *o;
+
+	/*
+	 * 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->max_io_len)
+		if (__origin_write(&o->snapshots, sector + n, NULL) ==
+		    DM_MAPIO_SUBMITTED)
+			must_wait = 1;
+	up_read(&_origins_lock);
+
+	return must_wait;
+}
+
+/*
  * Origin: maps a linear range of a device, with hooks for snapshotting.
  */
 
+struct dm_origin {
+	struct dm_dev *dev;
+	unsigned split_boundary;
+};
+
 /*
  * Construct an origin mapping: <dev_path>
  * The context for an origin is merely a 'struct dm_dev *'
@@ -1291,67 +2154,82 @@ static int do_origin(struct dm_dev *origin, struct bio *bio)
 static int origin_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 {
 	int r;
-	struct dm_dev *dev;
+	struct dm_origin *o;
 
 	if (argc != 1) {
 		ti->error = "origin: incorrect number of arguments";
 		return -EINVAL;
 	}
 
-	r = dm_get_device(ti, argv[0], 0, ti->len,
-			  dm_table_get_mode(ti->table), &dev);
+	o = kmalloc(sizeof(struct dm_origin), GFP_KERNEL);
+	if (!o) {
+		ti->error = "Cannot allocate private origin structure";
+		r = -ENOMEM;
+		goto bad_alloc;
+	}
+
+	r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &o->dev);
 	if (r) {
 		ti->error = "Cannot get target device";
-		return r;
+		goto bad_open;
 	}
 
-	ti->private = dev;
+	ti->private = o;
+	ti->num_flush_bios = 1;
+
 	return 0;
+
+bad_open:
+	kfree(o);
+bad_alloc:
+	return r;
 }
 
 static void origin_dtr(struct dm_target *ti)
 {
-	struct dm_dev *dev = ti->private;
-	dm_put_device(ti, dev);
+	struct dm_origin *o = ti->private;
+	dm_put_device(ti, o->dev);
+	kfree(o);
 }
 
-static int origin_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static int origin_map(struct dm_target *ti, struct bio *bio)
 {
-	struct dm_dev *dev = ti->private;
-	bio->bi_bdev = dev->bdev;
+	struct dm_origin *o = ti->private;
+	unsigned available_sectors;
+
+	bio->bi_bdev = o->dev->bdev;
+
+	if (unlikely(bio->bi_rw & REQ_FLUSH))
+		return DM_MAPIO_REMAPPED;
+
+	if (bio_rw(bio) != WRITE)
+		return DM_MAPIO_REMAPPED;
+
+	available_sectors = o->split_boundary -
+		((unsigned)bio->bi_iter.bi_sector & (o->split_boundary - 1));
+
+	if (bio_sectors(bio) > available_sectors)
+		dm_accept_partial_bio(bio, available_sectors);
 
 	/* Only tell snapshots if this is a write */
-	return (bio_rw(bio) == WRITE) ? do_origin(dev, bio) : DM_MAPIO_REMAPPED;
+	return do_origin(o->dev, bio);
 }
 
-#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))
-
 /*
- * Set the target "split_io" field to the minimum of all the snapshots'
+ * 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;
-	struct dm_snapshot *snap;
-	struct origin *o;
-	chunk_t chunk_size = 0;
-
-	down_read(&_origins_lock);
-	o = __lookup_origin(dev->bdev);
-	if (o)
-		list_for_each_entry (snap, &o->snapshots, list)
-			chunk_size = min_not_zero(chunk_size, snap->chunk_size);
-	up_read(&_origins_lock);
+	struct dm_origin *o = ti->private;
 
-	ti->split_io = chunk_size;
+	o->split_boundary = get_origin_minimum_chunksize(o->dev->bdev);
 }
 
-static int origin_status(struct dm_target *ti, status_type_t type, char *result,
-			 unsigned int maxlen)
+static void origin_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
 {
-	struct dm_dev *dev = ti->private;
+	struct dm_origin *o = ti->private;
 
 	switch (type) {
 	case STATUSTYPE_INFO:
@@ -1359,121 +2237,152 @@ static int origin_status(struct dm_target *ti, status_type_t type, char *result,
 		break;
 
 	case STATUSTYPE_TABLE:
-		snprintf(result, maxlen, "%s", dev->name);
+		snprintf(result, maxlen, "%s", o->dev->name);
 		break;
 	}
+}
 
-	return 0;
+static int origin_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+			struct bio_vec *biovec, int max_size)
+{
+	struct dm_origin *o = ti->private;
+	struct request_queue *q = bdev_get_queue(o->dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = o->dev->bdev;
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static int origin_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct dm_origin *o = ti->private;
+
+	return fn(ti, o->dev, 0, ti->len, data);
 }
 
 static struct target_type origin_target = {
 	.name    = "snapshot-origin",
-	.version = {1, 6, 0},
+	.version = {1, 8, 1},
 	.module  = THIS_MODULE,
 	.ctr     = origin_ctr,
 	.dtr     = origin_dtr,
 	.map     = origin_map,
 	.resume  = origin_resume,
 	.status  = origin_status,
+	.merge	 = origin_merge,
+	.iterate_devices = origin_iterate_devices,
 };
 
 static struct target_type snapshot_target = {
 	.name    = "snapshot",
-	.version = {1, 6, 0},
+	.version = {1, 12, 0},
 	.module  = THIS_MODULE,
 	.ctr     = snapshot_ctr,
 	.dtr     = snapshot_dtr,
 	.map     = snapshot_map,
 	.end_io  = snapshot_end_io,
+	.preresume  = snapshot_preresume,
 	.resume  = snapshot_resume,
 	.status  = snapshot_status,
+	.iterate_devices = snapshot_iterate_devices,
+};
+
+static struct target_type merge_target = {
+	.name    = dm_snapshot_merge_target_name,
+	.version = {1, 2, 0},
+	.module  = THIS_MODULE,
+	.ctr     = snapshot_ctr,
+	.dtr     = snapshot_dtr,
+	.map     = snapshot_merge_map,
+	.end_io  = snapshot_end_io,
+	.presuspend = snapshot_merge_presuspend,
+	.preresume  = snapshot_preresume,
+	.resume  = snapshot_merge_resume,
+	.status  = snapshot_status,
+	.iterate_devices = snapshot_iterate_devices,
 };
 
 static int __init dm_snapshot_init(void)
 {
 	int r;
 
-	r = dm_register_target(&snapshot_target);
+	r = dm_exception_store_init();
 	if (r) {
-		DMERR("snapshot target register failed %d", r);
+		DMERR("Failed to initialize exception stores");
 		return r;
 	}
 
+	r = dm_register_target(&snapshot_target);
+	if (r < 0) {
+		DMERR("snapshot target register failed %d", r);
+		goto bad_register_snapshot_target;
+	}
+
 	r = dm_register_target(&origin_target);
 	if (r < 0) {
 		DMERR("Origin target register failed %d", r);
-		goto bad1;
+		goto bad_register_origin_target;
+	}
+
+	r = dm_register_target(&merge_target);
+	if (r < 0) {
+		DMERR("Merge target register failed %d", r);
+		goto bad_register_merge_target;
 	}
 
 	r = init_origin_hash();
 	if (r) {
 		DMERR("init_origin_hash failed.");
-		goto bad2;
+		goto bad_origin_hash;
 	}
 
-	exception_cache = KMEM_CACHE(dm_snap_exception, 0);
+	exception_cache = KMEM_CACHE(dm_exception, 0);
 	if (!exception_cache) {
 		DMERR("Couldn't create exception cache.");
 		r = -ENOMEM;
-		goto bad3;
+		goto bad_exception_cache;
 	}
 
 	pending_cache = KMEM_CACHE(dm_snap_pending_exception, 0);
 	if (!pending_cache) {
 		DMERR("Couldn't create pending cache.");
 		r = -ENOMEM;
-		goto bad4;
-	}
-
-	tracked_chunk_cache = KMEM_CACHE(dm_snap_tracked_chunk, 0);
-	if (!tracked_chunk_cache) {
-		DMERR("Couldn't create cache to track chunks in use.");
-		r = -ENOMEM;
-		goto bad5;
-	}
-
-	ksnapd = create_singlethread_workqueue("ksnapd");
-	if (!ksnapd) {
-		DMERR("Failed to create ksnapd workqueue.");
-		r = -ENOMEM;
-		goto bad_pending_pool;
+		goto bad_pending_cache;
 	}
 
 	return 0;
 
-      bad_pending_pool:
-	kmem_cache_destroy(tracked_chunk_cache);
-      bad5:
-	kmem_cache_destroy(pending_cache);
-      bad4:
+bad_pending_cache:
 	kmem_cache_destroy(exception_cache);
-      bad3:
+bad_exception_cache:
 	exit_origin_hash();
-      bad2:
+bad_origin_hash:
+	dm_unregister_target(&merge_target);
+bad_register_merge_target:
 	dm_unregister_target(&origin_target);
-      bad1:
+bad_register_origin_target:
 	dm_unregister_target(&snapshot_target);
+bad_register_snapshot_target:
+	dm_exception_store_exit();
+
 	return r;
 }
 
 static void __exit dm_snapshot_exit(void)
 {
-	int r;
-
-	destroy_workqueue(ksnapd);
-
-	r = dm_unregister_target(&snapshot_target);
-	if (r)
-		DMERR("snapshot unregister failed %d", r);
-
-	r = dm_unregister_target(&origin_target);
-	if (r)
-		DMERR("origin unregister failed %d", r);
+	dm_unregister_target(&snapshot_target);
+	dm_unregister_target(&origin_target);
+	dm_unregister_target(&merge_target);
 
 	exit_origin_hash();
 	kmem_cache_destroy(pending_cache);
 	kmem_cache_destroy(exception_cache);
-	kmem_cache_destroy(tracked_chunk_cache);
+
+	dm_exception_store_exit();
 }
 
 /* Module hooks */
@@ -1483,3 +2392,5 @@ module_exit(dm_snapshot_exit);
 MODULE_DESCRIPTION(DM_NAME " snapshot target");
 MODULE_AUTHOR("Joe Thornber");
 MODULE_LICENSE("GPL");
+MODULE_ALIAS("dm-snapshot-origin");
+MODULE_ALIAS("dm-snapshot-merge");
diff --git a/drivers/md/dm-snap.h b/drivers/md/dm-snap.h
deleted file mode 100644
index 292c15609ae..00000000000
--- a/drivers/md/dm-snap.h
+++ /dev/null
@@ -1,231 +0,0 @@
-/*
- * dm-snapshot.c
- *
- * Copyright (C) 2001-2002 Sistina Software (UK) Limited.
- *
- * This file is released under the GPL.
- */
-
-#ifndef DM_SNAPSHOT_H
-#define DM_SNAPSHOT_H
-
-#include "dm.h"
-#include "dm-bio-list.h"
-#include <linux/blkdev.h>
-#include <linux/workqueue.h>
-
-struct exception_table {
-	uint32_t hash_mask;
-	unsigned hash_shift;
-	struct list_head *table;
-};
-
-/*
- * The snapshot code deals with largish chunks of the disk at a
- * time. Typically 32k - 512k.
- */
-typedef sector_t chunk_t;
-
-/*
- * An exception is used where an old chunk of data has been
- * replaced by a new one.
- * If chunk_t is 64 bits in size, the top 8 bits of new_chunk hold the number
- * of chunks that follow contiguously.  Remaining bits hold the number of the
- * chunk within the device.
- */
-struct dm_snap_exception {
-	struct list_head hash_list;
-
-	chunk_t old_chunk;
-	chunk_t new_chunk;
-};
-
-/*
- * Funtions to manipulate consecutive chunks
- */
-#  if defined(CONFIG_LBD) || (BITS_PER_LONG == 64)
-#    define DM_CHUNK_CONSECUTIVE_BITS 8
-#    define DM_CHUNK_NUMBER_BITS 56
-
-static inline chunk_t dm_chunk_number(chunk_t chunk)
-{
-	return chunk & (chunk_t)((1ULL << DM_CHUNK_NUMBER_BITS) - 1ULL);
-}
-
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
-{
-	return e->new_chunk >> DM_CHUNK_NUMBER_BITS;
-}
-
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
-{
-	e->new_chunk += (1ULL << DM_CHUNK_NUMBER_BITS);
-
-	BUG_ON(!dm_consecutive_chunk_count(e));
-}
-
-#  else
-#    define DM_CHUNK_CONSECUTIVE_BITS 0
-
-static inline chunk_t dm_chunk_number(chunk_t chunk)
-{
-	return chunk;
-}
-
-static inline unsigned dm_consecutive_chunk_count(struct dm_snap_exception *e)
-{
-	return 0;
-}
-
-static inline void dm_consecutive_chunk_count_inc(struct dm_snap_exception *e)
-{
-}
-
-#  endif
-
-/*
- * Abstraction to handle the meta/layout of exception stores (the
- * COW device).
- */
-struct exception_store {
-
-	/*
-	 * Destroys this object when you've finished with it.
-	 */
-	void (*destroy) (struct exception_store *store);
-
-	/*
-	 * The target shouldn't read the COW device until this is
-	 * called.
-	 */
-	int (*read_metadata) (struct exception_store *store);
-
-	/*
-	 * Find somewhere to store the next exception.
-	 */
-	int (*prepare_exception) (struct exception_store *store,
-				  struct dm_snap_exception *e);
-
-	/*
-	 * Update the metadata with this exception.
-	 */
-	void (*commit_exception) (struct exception_store *store,
-				  struct dm_snap_exception *e,
-				  void (*callback) (void *, int success),
-				  void *callback_context);
-
-	/*
-	 * The snapshot is invalid, note this in the metadata.
-	 */
-	void (*drop_snapshot) (struct exception_store *store);
-
-	/*
-	 * Return how full the snapshot is.
-	 */
-	void (*fraction_full) (struct exception_store *store,
-			       sector_t *numerator,
-			       sector_t *denominator);
-
-	struct dm_snapshot *snap;
-	void *context;
-};
-
-#define DM_TRACKED_CHUNK_HASH_SIZE	16
-#define DM_TRACKED_CHUNK_HASH(x)	((unsigned long)(x) & \
-					 (DM_TRACKED_CHUNK_HASH_SIZE - 1))
-
-struct dm_snapshot {
-	struct rw_semaphore lock;
-	struct dm_target *ti;
-
-	struct dm_dev *origin;
-	struct dm_dev *cow;
-
-	/* List of snapshots per Origin */
-	struct list_head list;
-
-	/* Size of data blocks saved - must be a power of 2 */
-	chunk_t chunk_size;
-	chunk_t chunk_mask;
-	chunk_t chunk_shift;
-
-	/* You can't use a snapshot if this is 0 (e.g. if full) */
-	int valid;
-
-	/* Origin writes don't trigger exceptions until this is set */
-	int active;
-
-	/* Used for display of table */
-	char type;
-
-	/* The last percentage we notified */
-	int last_percent;
-
-	mempool_t *pending_pool;
-
-	struct exception_table pending;
-	struct exception_table complete;
-
-	/*
-	 * pe_lock protects all pending_exception operations and access
-	 * as well as the snapshot_bios list.
-	 */
-	spinlock_t pe_lock;
-
-	/* The on disk metadata handler */
-	struct exception_store store;
-
-	struct dm_kcopyd_client *kcopyd_client;
-
-	/* Queue of snapshot writes for ksnapd to flush */
-	struct bio_list queued_bios;
-	struct work_struct queued_bios_work;
-
-	/* Chunks with outstanding reads */
-	mempool_t *tracked_chunk_pool;
-	spinlock_t tracked_chunk_lock;
-	struct hlist_head tracked_chunk_hash[DM_TRACKED_CHUNK_HASH_SIZE];
-};
-
-/*
- * Used by the exception stores to load exceptions hen
- * initialising.
- */
-int dm_add_exception(struct dm_snapshot *s, chunk_t old, chunk_t new);
-
-/*
- * Constructor and destructor for the default persistent
- * store.
- */
-int dm_create_persistent(struct exception_store *store);
-
-int dm_create_transient(struct exception_store *store);
-
-/*
- * Return the number of sectors in the device.
- */
-static inline sector_t get_dev_size(struct block_device *bdev)
-{
-	return bdev->bd_inode->i_size >> SECTOR_SHIFT;
-}
-
-static inline chunk_t sector_to_chunk(struct dm_snapshot *s, sector_t sector)
-{
-	return (sector & ~s->chunk_mask) >> s->chunk_shift;
-}
-
-static inline sector_t chunk_to_sector(struct dm_snapshot *s, chunk_t chunk)
-{
-	return chunk << s->chunk_shift;
-}
-
-static inline int bdev_equal(struct block_device *lhs, struct block_device *rhs)
-{
-	/*
-	 * There is only ever one instance of a particular block
-	 * device so we can compare pointers safely.
-	 */
-	return lhs == rhs;
-}
-
-#endif
diff --git a/drivers/md/dm-stats.c b/drivers/md/dm-stats.c
new file mode 100644
index 00000000000..28a90122a5a
--- /dev/null
+++ b/drivers/md/dm-stats.c
@@ -0,0 +1,981 @@
+#include <linux/errno.h>
+#include <linux/numa.h>
+#include <linux/slab.h>
+#include <linux/rculist.h>
+#include <linux/threads.h>
+#include <linux/preempt.h>
+#include <linux/irqflags.h>
+#include <linux/vmalloc.h>
+#include <linux/mm.h>
+#include <linux/module.h>
+#include <linux/device-mapper.h>
+
+#include "dm.h"
+#include "dm-stats.h"
+
+#define DM_MSG_PREFIX "stats"
+
+static int dm_stat_need_rcu_barrier;
+
+/*
+ * Using 64-bit values to avoid overflow (which is a
+ * problem that block/genhd.c's IO accounting has).
+ */
+struct dm_stat_percpu {
+	unsigned long long sectors[2];
+	unsigned long long ios[2];
+	unsigned long long merges[2];
+	unsigned long long ticks[2];
+	unsigned long long io_ticks[2];
+	unsigned long long io_ticks_total;
+	unsigned long long time_in_queue;
+};
+
+struct dm_stat_shared {
+	atomic_t in_flight[2];
+	unsigned long stamp;
+	struct dm_stat_percpu tmp;
+};
+
+struct dm_stat {
+	struct list_head list_entry;
+	int id;
+	size_t n_entries;
+	sector_t start;
+	sector_t end;
+	sector_t step;
+	const char *program_id;
+	const char *aux_data;
+	struct rcu_head rcu_head;
+	size_t shared_alloc_size;
+	size_t percpu_alloc_size;
+	struct dm_stat_percpu *stat_percpu[NR_CPUS];
+	struct dm_stat_shared stat_shared[0];
+};
+
+struct dm_stats_last_position {
+	sector_t last_sector;
+	unsigned last_rw;
+};
+
+/*
+ * A typo on the command line could possibly make the kernel run out of memory
+ * and crash. To prevent the crash we account all used memory. We fail if we
+ * exhaust 1/4 of all memory or 1/2 of vmalloc space.
+ */
+#define DM_STATS_MEMORY_FACTOR		4
+#define DM_STATS_VMALLOC_FACTOR		2
+
+static DEFINE_SPINLOCK(shared_memory_lock);
+
+static unsigned long shared_memory_amount;
+
+static bool __check_shared_memory(size_t alloc_size)
+{
+	size_t a;
+
+	a = shared_memory_amount + alloc_size;
+	if (a < shared_memory_amount)
+		return false;
+	if (a >> PAGE_SHIFT > totalram_pages / DM_STATS_MEMORY_FACTOR)
+		return false;
+#ifdef CONFIG_MMU
+	if (a > (VMALLOC_END - VMALLOC_START) / DM_STATS_VMALLOC_FACTOR)
+		return false;
+#endif
+	return true;
+}
+
+static bool check_shared_memory(size_t alloc_size)
+{
+	bool ret;
+
+	spin_lock_irq(&shared_memory_lock);
+
+	ret = __check_shared_memory(alloc_size);
+
+	spin_unlock_irq(&shared_memory_lock);
+
+	return ret;
+}
+
+static bool claim_shared_memory(size_t alloc_size)
+{
+	spin_lock_irq(&shared_memory_lock);
+
+	if (!__check_shared_memory(alloc_size)) {
+		spin_unlock_irq(&shared_memory_lock);
+		return false;
+	}
+
+	shared_memory_amount += alloc_size;
+
+	spin_unlock_irq(&shared_memory_lock);
+
+	return true;
+}
+
+static void free_shared_memory(size_t alloc_size)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(&shared_memory_lock, flags);
+
+	if (WARN_ON_ONCE(shared_memory_amount < alloc_size)) {
+		spin_unlock_irqrestore(&shared_memory_lock, flags);
+		DMCRIT("Memory usage accounting bug.");
+		return;
+	}
+
+	shared_memory_amount -= alloc_size;
+
+	spin_unlock_irqrestore(&shared_memory_lock, flags);
+}
+
+static void *dm_kvzalloc(size_t alloc_size, int node)
+{
+	void *p;
+
+	if (!claim_shared_memory(alloc_size))
+		return NULL;
+
+	if (alloc_size <= KMALLOC_MAX_SIZE) {
+		p = kzalloc_node(alloc_size, GFP_KERNEL | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN, node);
+		if (p)
+			return p;
+	}
+	p = vzalloc_node(alloc_size, node);
+	if (p)
+		return p;
+
+	free_shared_memory(alloc_size);
+
+	return NULL;
+}
+
+static void dm_kvfree(void *ptr, size_t alloc_size)
+{
+	if (!ptr)
+		return;
+
+	free_shared_memory(alloc_size);
+
+	if (is_vmalloc_addr(ptr))
+		vfree(ptr);
+	else
+		kfree(ptr);
+}
+
+static void dm_stat_free(struct rcu_head *head)
+{
+	int cpu;
+	struct dm_stat *s = container_of(head, struct dm_stat, rcu_head);
+
+	kfree(s->program_id);
+	kfree(s->aux_data);
+	for_each_possible_cpu(cpu)
+		dm_kvfree(s->stat_percpu[cpu], s->percpu_alloc_size);
+	dm_kvfree(s, s->shared_alloc_size);
+}
+
+static int dm_stat_in_flight(struct dm_stat_shared *shared)
+{
+	return atomic_read(&shared->in_flight[READ]) +
+	       atomic_read(&shared->in_flight[WRITE]);
+}
+
+void dm_stats_init(struct dm_stats *stats)
+{
+	int cpu;
+	struct dm_stats_last_position *last;
+
+	mutex_init(&stats->mutex);
+	INIT_LIST_HEAD(&stats->list);
+	stats->last = alloc_percpu(struct dm_stats_last_position);
+	for_each_possible_cpu(cpu) {
+		last = per_cpu_ptr(stats->last, cpu);
+		last->last_sector = (sector_t)ULLONG_MAX;
+		last->last_rw = UINT_MAX;
+	}
+}
+
+void dm_stats_cleanup(struct dm_stats *stats)
+{
+	size_t ni;
+	struct dm_stat *s;
+	struct dm_stat_shared *shared;
+
+	while (!list_empty(&stats->list)) {
+		s = container_of(stats->list.next, struct dm_stat, list_entry);
+		list_del(&s->list_entry);
+		for (ni = 0; ni < s->n_entries; ni++) {
+			shared = &s->stat_shared[ni];
+			if (WARN_ON(dm_stat_in_flight(shared))) {
+				DMCRIT("leaked in-flight counter at index %lu "
+				       "(start %llu, end %llu, step %llu): reads %d, writes %d",
+				       (unsigned long)ni,
+				       (unsigned long long)s->start,
+				       (unsigned long long)s->end,
+				       (unsigned long long)s->step,
+				       atomic_read(&shared->in_flight[READ]),
+				       atomic_read(&shared->in_flight[WRITE]));
+			}
+		}
+		dm_stat_free(&s->rcu_head);
+	}
+	free_percpu(stats->last);
+}
+
+static int dm_stats_create(struct dm_stats *stats, sector_t start, sector_t end,
+			   sector_t step, const char *program_id, const char *aux_data,
+			   void (*suspend_callback)(struct mapped_device *),
+			   void (*resume_callback)(struct mapped_device *),
+			   struct mapped_device *md)
+{
+	struct list_head *l;
+	struct dm_stat *s, *tmp_s;
+	sector_t n_entries;
+	size_t ni;
+	size_t shared_alloc_size;
+	size_t percpu_alloc_size;
+	struct dm_stat_percpu *p;
+	int cpu;
+	int ret_id;
+	int r;
+
+	if (end < start || !step)
+		return -EINVAL;
+
+	n_entries = end - start;
+	if (dm_sector_div64(n_entries, step))
+		n_entries++;
+
+	if (n_entries != (size_t)n_entries || !(size_t)(n_entries + 1))
+		return -EOVERFLOW;
+
+	shared_alloc_size = sizeof(struct dm_stat) + (size_t)n_entries * sizeof(struct dm_stat_shared);
+	if ((shared_alloc_size - sizeof(struct dm_stat)) / sizeof(struct dm_stat_shared) != n_entries)
+		return -EOVERFLOW;
+
+	percpu_alloc_size = (size_t)n_entries * sizeof(struct dm_stat_percpu);
+	if (percpu_alloc_size / sizeof(struct dm_stat_percpu) != n_entries)
+		return -EOVERFLOW;
+
+	if (!check_shared_memory(shared_alloc_size + num_possible_cpus() * percpu_alloc_size))
+		return -ENOMEM;
+
+	s = dm_kvzalloc(shared_alloc_size, NUMA_NO_NODE);
+	if (!s)
+		return -ENOMEM;
+
+	s->n_entries = n_entries;
+	s->start = start;
+	s->end = end;
+	s->step = step;
+	s->shared_alloc_size = shared_alloc_size;
+	s->percpu_alloc_size = percpu_alloc_size;
+
+	s->program_id = kstrdup(program_id, GFP_KERNEL);
+	if (!s->program_id) {
+		r = -ENOMEM;
+		goto out;
+	}
+	s->aux_data = kstrdup(aux_data, GFP_KERNEL);
+	if (!s->aux_data) {
+		r = -ENOMEM;
+		goto out;
+	}
+
+	for (ni = 0; ni < n_entries; ni++) {
+		atomic_set(&s->stat_shared[ni].in_flight[READ], 0);
+		atomic_set(&s->stat_shared[ni].in_flight[WRITE], 0);
+	}
+
+	for_each_possible_cpu(cpu) {
+		p = dm_kvzalloc(percpu_alloc_size, cpu_to_node(cpu));
+		if (!p) {
+			r = -ENOMEM;
+			goto out;
+		}
+		s->stat_percpu[cpu] = p;
+	}
+
+	/*
+	 * Suspend/resume to make sure there is no i/o in flight,
+	 * so that newly created statistics will be exact.
+	 *
+	 * (note: we couldn't suspend earlier because we must not
+	 * allocate memory while suspended)
+	 */
+	suspend_callback(md);
+
+	mutex_lock(&stats->mutex);
+	s->id = 0;
+	list_for_each(l, &stats->list) {
+		tmp_s = container_of(l, struct dm_stat, list_entry);
+		if (WARN_ON(tmp_s->id < s->id)) {
+			r = -EINVAL;
+			goto out_unlock_resume;
+		}
+		if (tmp_s->id > s->id)
+			break;
+		if (unlikely(s->id == INT_MAX)) {
+			r = -ENFILE;
+			goto out_unlock_resume;
+		}
+		s->id++;
+	}
+	ret_id = s->id;
+	list_add_tail_rcu(&s->list_entry, l);
+	mutex_unlock(&stats->mutex);
+
+	resume_callback(md);
+
+	return ret_id;
+
+out_unlock_resume:
+	mutex_unlock(&stats->mutex);
+	resume_callback(md);
+out:
+	dm_stat_free(&s->rcu_head);
+	return r;
+}
+
+static struct dm_stat *__dm_stats_find(struct dm_stats *stats, int id)
+{
+	struct dm_stat *s;
+
+	list_for_each_entry(s, &stats->list, list_entry) {
+		if (s->id > id)
+			break;
+		if (s->id == id)
+			return s;
+	}
+
+	return NULL;
+}
+
+static int dm_stats_delete(struct dm_stats *stats, int id)
+{
+	struct dm_stat *s;
+	int cpu;
+
+	mutex_lock(&stats->mutex);
+
+	s = __dm_stats_find(stats, id);
+	if (!s) {
+		mutex_unlock(&stats->mutex);
+		return -ENOENT;
+	}
+
+	list_del_rcu(&s->list_entry);
+	mutex_unlock(&stats->mutex);
+
+	/*
+	 * vfree can't be called from RCU callback
+	 */
+	for_each_possible_cpu(cpu)
+		if (is_vmalloc_addr(s->stat_percpu))
+			goto do_sync_free;
+	if (is_vmalloc_addr(s)) {
+do_sync_free:
+		synchronize_rcu_expedited();
+		dm_stat_free(&s->rcu_head);
+	} else {
+		ACCESS_ONCE(dm_stat_need_rcu_barrier) = 1;
+		call_rcu(&s->rcu_head, dm_stat_free);
+	}
+	return 0;
+}
+
+static int dm_stats_list(struct dm_stats *stats, const char *program,
+			 char *result, unsigned maxlen)
+{
+	struct dm_stat *s;
+	sector_t len;
+	unsigned sz = 0;
+
+	/*
+	 * Output format:
+	 *   <region_id>: <start_sector>+<length> <step> <program_id> <aux_data>
+	 */
+
+	mutex_lock(&stats->mutex);
+	list_for_each_entry(s, &stats->list, list_entry) {
+		if (!program || !strcmp(program, s->program_id)) {
+			len = s->end - s->start;
+			DMEMIT("%d: %llu+%llu %llu %s %s\n", s->id,
+				(unsigned long long)s->start,
+				(unsigned long long)len,
+				(unsigned long long)s->step,
+				s->program_id,
+				s->aux_data);
+		}
+	}
+	mutex_unlock(&stats->mutex);
+
+	return 1;
+}
+
+static void dm_stat_round(struct dm_stat_shared *shared, struct dm_stat_percpu *p)
+{
+	/*
+	 * This is racy, but so is part_round_stats_single.
+	 */
+	unsigned long now = jiffies;
+	unsigned in_flight_read;
+	unsigned in_flight_write;
+	unsigned long difference = now - shared->stamp;
+
+	if (!difference)
+		return;
+	in_flight_read = (unsigned)atomic_read(&shared->in_flight[READ]);
+	in_flight_write = (unsigned)atomic_read(&shared->in_flight[WRITE]);
+	if (in_flight_read)
+		p->io_ticks[READ] += difference;
+	if (in_flight_write)
+		p->io_ticks[WRITE] += difference;
+	if (in_flight_read + in_flight_write) {
+		p->io_ticks_total += difference;
+		p->time_in_queue += (in_flight_read + in_flight_write) * difference;
+	}
+	shared->stamp = now;
+}
+
+static void dm_stat_for_entry(struct dm_stat *s, size_t entry,
+			      unsigned long bi_rw, sector_t len, bool merged,
+			      bool end, unsigned long duration)
+{
+	unsigned long idx = bi_rw & REQ_WRITE;
+	struct dm_stat_shared *shared = &s->stat_shared[entry];
+	struct dm_stat_percpu *p;
+
+	/*
+	 * For strict correctness we should use local_irq_save/restore
+	 * instead of preempt_disable/enable.
+	 *
+	 * preempt_disable/enable is racy if the driver finishes bios
+	 * from non-interrupt context as well as from interrupt context
+	 * or from more different interrupts.
+	 *
+	 * On 64-bit architectures the race only results in not counting some
+	 * events, so it is acceptable.  On 32-bit architectures the race could
+	 * cause the counter going off by 2^32, so we need to do proper locking
+	 * there.
+	 *
+	 * part_stat_lock()/part_stat_unlock() have this race too.
+	 */
+#if BITS_PER_LONG == 32
+	unsigned long flags;
+	local_irq_save(flags);
+#else
+	preempt_disable();
+#endif
+	p = &s->stat_percpu[smp_processor_id()][entry];
+
+	if (!end) {
+		dm_stat_round(shared, p);
+		atomic_inc(&shared->in_flight[idx]);
+	} else {
+		dm_stat_round(shared, p);
+		atomic_dec(&shared->in_flight[idx]);
+		p->sectors[idx] += len;
+		p->ios[idx] += 1;
+		p->merges[idx] += merged;
+		p->ticks[idx] += duration;
+	}
+
+#if BITS_PER_LONG == 32
+	local_irq_restore(flags);
+#else
+	preempt_enable();
+#endif
+}
+
+static void __dm_stat_bio(struct dm_stat *s, unsigned long bi_rw,
+			  sector_t bi_sector, sector_t end_sector,
+			  bool end, unsigned long duration,
+			  struct dm_stats_aux *stats_aux)
+{
+	sector_t rel_sector, offset, todo, fragment_len;
+	size_t entry;
+
+	if (end_sector <= s->start || bi_sector >= s->end)
+		return;
+	if (unlikely(bi_sector < s->start)) {
+		rel_sector = 0;
+		todo = end_sector - s->start;
+	} else {
+		rel_sector = bi_sector - s->start;
+		todo = end_sector - bi_sector;
+	}
+	if (unlikely(end_sector > s->end))
+		todo -= (end_sector - s->end);
+
+	offset = dm_sector_div64(rel_sector, s->step);
+	entry = rel_sector;
+	do {
+		if (WARN_ON_ONCE(entry >= s->n_entries)) {
+			DMCRIT("Invalid area access in region id %d", s->id);
+			return;
+		}
+		fragment_len = todo;
+		if (fragment_len > s->step - offset)
+			fragment_len = s->step - offset;
+		dm_stat_for_entry(s, entry, bi_rw, fragment_len,
+				  stats_aux->merged, end, duration);
+		todo -= fragment_len;
+		entry++;
+		offset = 0;
+	} while (unlikely(todo != 0));
+}
+
+void dm_stats_account_io(struct dm_stats *stats, unsigned long bi_rw,
+			 sector_t bi_sector, unsigned bi_sectors, bool end,
+			 unsigned long duration, struct dm_stats_aux *stats_aux)
+{
+	struct dm_stat *s;
+	sector_t end_sector;
+	struct dm_stats_last_position *last;
+
+	if (unlikely(!bi_sectors))
+		return;
+
+	end_sector = bi_sector + bi_sectors;
+
+	if (!end) {
+		/*
+		 * A race condition can at worst result in the merged flag being
+		 * misrepresented, so we don't have to disable preemption here.
+		 */
+		last = __this_cpu_ptr(stats->last);
+		stats_aux->merged =
+			(bi_sector == (ACCESS_ONCE(last->last_sector) &&
+				       ((bi_rw & (REQ_WRITE | REQ_DISCARD)) ==
+					(ACCESS_ONCE(last->last_rw) & (REQ_WRITE | REQ_DISCARD)))
+				       ));
+		ACCESS_ONCE(last->last_sector) = end_sector;
+		ACCESS_ONCE(last->last_rw) = bi_rw;
+	}
+
+	rcu_read_lock();
+
+	list_for_each_entry_rcu(s, &stats->list, list_entry)
+		__dm_stat_bio(s, bi_rw, bi_sector, end_sector, end, duration, stats_aux);
+
+	rcu_read_unlock();
+}
+
+static void __dm_stat_init_temporary_percpu_totals(struct dm_stat_shared *shared,
+						   struct dm_stat *s, size_t x)
+{
+	int cpu;
+	struct dm_stat_percpu *p;
+
+	local_irq_disable();
+	p = &s->stat_percpu[smp_processor_id()][x];
+	dm_stat_round(shared, p);
+	local_irq_enable();
+
+	memset(&shared->tmp, 0, sizeof(shared->tmp));
+	for_each_possible_cpu(cpu) {
+		p = &s->stat_percpu[cpu][x];
+		shared->tmp.sectors[READ] += ACCESS_ONCE(p->sectors[READ]);
+		shared->tmp.sectors[WRITE] += ACCESS_ONCE(p->sectors[WRITE]);
+		shared->tmp.ios[READ] += ACCESS_ONCE(p->ios[READ]);
+		shared->tmp.ios[WRITE] += ACCESS_ONCE(p->ios[WRITE]);
+		shared->tmp.merges[READ] += ACCESS_ONCE(p->merges[READ]);
+		shared->tmp.merges[WRITE] += ACCESS_ONCE(p->merges[WRITE]);
+		shared->tmp.ticks[READ] += ACCESS_ONCE(p->ticks[READ]);
+		shared->tmp.ticks[WRITE] += ACCESS_ONCE(p->ticks[WRITE]);
+		shared->tmp.io_ticks[READ] += ACCESS_ONCE(p->io_ticks[READ]);
+		shared->tmp.io_ticks[WRITE] += ACCESS_ONCE(p->io_ticks[WRITE]);
+		shared->tmp.io_ticks_total += ACCESS_ONCE(p->io_ticks_total);
+		shared->tmp.time_in_queue += ACCESS_ONCE(p->time_in_queue);
+	}
+}
+
+static void __dm_stat_clear(struct dm_stat *s, size_t idx_start, size_t idx_end,
+			    bool init_tmp_percpu_totals)
+{
+	size_t x;
+	struct dm_stat_shared *shared;
+	struct dm_stat_percpu *p;
+
+	for (x = idx_start; x < idx_end; x++) {
+		shared = &s->stat_shared[x];
+		if (init_tmp_percpu_totals)
+			__dm_stat_init_temporary_percpu_totals(shared, s, x);
+		local_irq_disable();
+		p = &s->stat_percpu[smp_processor_id()][x];
+		p->sectors[READ] -= shared->tmp.sectors[READ];
+		p->sectors[WRITE] -= shared->tmp.sectors[WRITE];
+		p->ios[READ] -= shared->tmp.ios[READ];
+		p->ios[WRITE] -= shared->tmp.ios[WRITE];
+		p->merges[READ] -= shared->tmp.merges[READ];
+		p->merges[WRITE] -= shared->tmp.merges[WRITE];
+		p->ticks[READ] -= shared->tmp.ticks[READ];
+		p->ticks[WRITE] -= shared->tmp.ticks[WRITE];
+		p->io_ticks[READ] -= shared->tmp.io_ticks[READ];
+		p->io_ticks[WRITE] -= shared->tmp.io_ticks[WRITE];
+		p->io_ticks_total -= shared->tmp.io_ticks_total;
+		p->time_in_queue -= shared->tmp.time_in_queue;
+		local_irq_enable();
+	}
+}
+
+static int dm_stats_clear(struct dm_stats *stats, int id)
+{
+	struct dm_stat *s;
+
+	mutex_lock(&stats->mutex);
+
+	s = __dm_stats_find(stats, id);
+	if (!s) {
+		mutex_unlock(&stats->mutex);
+		return -ENOENT;
+	}
+
+	__dm_stat_clear(s, 0, s->n_entries, true);
+
+	mutex_unlock(&stats->mutex);
+
+	return 1;
+}
+
+/*
+ * This is like jiffies_to_msec, but works for 64-bit values.
+ */
+static unsigned long long dm_jiffies_to_msec64(unsigned long long j)
+{
+	unsigned long long result = 0;
+	unsigned mult;
+
+	if (j)
+		result = jiffies_to_msecs(j & 0x3fffff);
+	if (j >= 1 << 22) {
+		mult = jiffies_to_msecs(1 << 22);
+		result += (unsigned long long)mult * (unsigned long long)jiffies_to_msecs((j >> 22) & 0x3fffff);
+	}
+	if (j >= 1ULL << 44)
+		result += (unsigned long long)mult * (unsigned long long)mult * (unsigned long long)jiffies_to_msecs(j >> 44);
+
+	return result;
+}
+
+static int dm_stats_print(struct dm_stats *stats, int id,
+			  size_t idx_start, size_t idx_len,
+			  bool clear, char *result, unsigned maxlen)
+{
+	unsigned sz = 0;
+	struct dm_stat *s;
+	size_t x;
+	sector_t start, end, step;
+	size_t idx_end;
+	struct dm_stat_shared *shared;
+
+	/*
+	 * Output format:
+	 *   <start_sector>+<length> counters
+	 */
+
+	mutex_lock(&stats->mutex);
+
+	s = __dm_stats_find(stats, id);
+	if (!s) {
+		mutex_unlock(&stats->mutex);
+		return -ENOENT;
+	}
+
+	idx_end = idx_start + idx_len;
+	if (idx_end < idx_start ||
+	    idx_end > s->n_entries)
+		idx_end = s->n_entries;
+
+	if (idx_start > idx_end)
+		idx_start = idx_end;
+
+	step = s->step;
+	start = s->start + (step * idx_start);
+
+	for (x = idx_start; x < idx_end; x++, start = end) {
+		shared = &s->stat_shared[x];
+		end = start + step;
+		if (unlikely(end > s->end))
+			end = s->end;
+
+		__dm_stat_init_temporary_percpu_totals(shared, s, x);
+
+		DMEMIT("%llu+%llu %llu %llu %llu %llu %llu %llu %llu %llu %d %llu %llu %llu %llu\n",
+		       (unsigned long long)start,
+		       (unsigned long long)step,
+		       shared->tmp.ios[READ],
+		       shared->tmp.merges[READ],
+		       shared->tmp.sectors[READ],
+		       dm_jiffies_to_msec64(shared->tmp.ticks[READ]),
+		       shared->tmp.ios[WRITE],
+		       shared->tmp.merges[WRITE],
+		       shared->tmp.sectors[WRITE],
+		       dm_jiffies_to_msec64(shared->tmp.ticks[WRITE]),
+		       dm_stat_in_flight(shared),
+		       dm_jiffies_to_msec64(shared->tmp.io_ticks_total),
+		       dm_jiffies_to_msec64(shared->tmp.time_in_queue),
+		       dm_jiffies_to_msec64(shared->tmp.io_ticks[READ]),
+		       dm_jiffies_to_msec64(shared->tmp.io_ticks[WRITE]));
+
+		if (unlikely(sz + 1 >= maxlen))
+			goto buffer_overflow;
+	}
+
+	if (clear)
+		__dm_stat_clear(s, idx_start, idx_end, false);
+
+buffer_overflow:
+	mutex_unlock(&stats->mutex);
+
+	return 1;
+}
+
+static int dm_stats_set_aux(struct dm_stats *stats, int id, const char *aux_data)
+{
+	struct dm_stat *s;
+	const char *new_aux_data;
+
+	mutex_lock(&stats->mutex);
+
+	s = __dm_stats_find(stats, id);
+	if (!s) {
+		mutex_unlock(&stats->mutex);
+		return -ENOENT;
+	}
+
+	new_aux_data = kstrdup(aux_data, GFP_KERNEL);
+	if (!new_aux_data) {
+		mutex_unlock(&stats->mutex);
+		return -ENOMEM;
+	}
+
+	kfree(s->aux_data);
+	s->aux_data = new_aux_data;
+
+	mutex_unlock(&stats->mutex);
+
+	return 0;
+}
+
+static int message_stats_create(struct mapped_device *md,
+				unsigned argc, char **argv,
+				char *result, unsigned maxlen)
+{
+	int id;
+	char dummy;
+	unsigned long long start, end, len, step;
+	unsigned divisor;
+	const char *program_id, *aux_data;
+
+	/*
+	 * Input format:
+	 *   <range> <step> [<program_id> [<aux_data>]]
+	 */
+
+	if (argc < 3 || argc > 5)
+		return -EINVAL;
+
+	if (!strcmp(argv[1], "-")) {
+		start = 0;
+		len = dm_get_size(md);
+		if (!len)
+			len = 1;
+	} else if (sscanf(argv[1], "%llu+%llu%c", &start, &len, &dummy) != 2 ||
+		   start != (sector_t)start || len != (sector_t)len)
+		return -EINVAL;
+
+	end = start + len;
+	if (start >= end)
+		return -EINVAL;
+
+	if (sscanf(argv[2], "/%u%c", &divisor, &dummy) == 1) {
+		step = end - start;
+		if (do_div(step, divisor))
+			step++;
+		if (!step)
+			step = 1;
+	} else if (sscanf(argv[2], "%llu%c", &step, &dummy) != 1 ||
+		   step != (sector_t)step || !step)
+		return -EINVAL;
+
+	program_id = "-";
+	aux_data = "-";
+
+	if (argc > 3)
+		program_id = argv[3];
+
+	if (argc > 4)
+		aux_data = argv[4];
+
+	/*
+	 * If a buffer overflow happens after we created the region,
+	 * it's too late (the userspace would retry with a larger
+	 * buffer, but the region id that caused the overflow is already
+	 * leaked).  So we must detect buffer overflow in advance.
+	 */
+	snprintf(result, maxlen, "%d", INT_MAX);
+	if (dm_message_test_buffer_overflow(result, maxlen))
+		return 1;
+
+	id = dm_stats_create(dm_get_stats(md), start, end, step, program_id, aux_data,
+			     dm_internal_suspend, dm_internal_resume, md);
+	if (id < 0)
+		return id;
+
+	snprintf(result, maxlen, "%d", id);
+
+	return 1;
+}
+
+static int message_stats_delete(struct mapped_device *md,
+				unsigned argc, char **argv)
+{
+	int id;
+	char dummy;
+
+	if (argc != 2)
+		return -EINVAL;
+
+	if (sscanf(argv[1], "%d%c", &id, &dummy) != 1 || id < 0)
+		return -EINVAL;
+
+	return dm_stats_delete(dm_get_stats(md), id);
+}
+
+static int message_stats_clear(struct mapped_device *md,
+			       unsigned argc, char **argv)
+{
+	int id;
+	char dummy;
+
+	if (argc != 2)
+		return -EINVAL;
+
+	if (sscanf(argv[1], "%d%c", &id, &dummy) != 1 || id < 0)
+		return -EINVAL;
+
+	return dm_stats_clear(dm_get_stats(md), id);
+}
+
+static int message_stats_list(struct mapped_device *md,
+			      unsigned argc, char **argv,
+			      char *result, unsigned maxlen)
+{
+	int r;
+	const char *program = NULL;
+
+	if (argc < 1 || argc > 2)
+		return -EINVAL;
+
+	if (argc > 1) {
+		program = kstrdup(argv[1], GFP_KERNEL);
+		if (!program)
+			return -ENOMEM;
+	}
+
+	r = dm_stats_list(dm_get_stats(md), program, result, maxlen);
+
+	kfree(program);
+
+	return r;
+}
+
+static int message_stats_print(struct mapped_device *md,
+			       unsigned argc, char **argv, bool clear,
+			       char *result, unsigned maxlen)
+{
+	int id;
+	char dummy;
+	unsigned long idx_start = 0, idx_len = ULONG_MAX;
+
+	if (argc != 2 && argc != 4)
+		return -EINVAL;
+
+	if (sscanf(argv[1], "%d%c", &id, &dummy) != 1 || id < 0)
+		return -EINVAL;
+
+	if (argc > 3) {
+		if (strcmp(argv[2], "-") &&
+		    sscanf(argv[2], "%lu%c", &idx_start, &dummy) != 1)
+			return -EINVAL;
+		if (strcmp(argv[3], "-") &&
+		    sscanf(argv[3], "%lu%c", &idx_len, &dummy) != 1)
+			return -EINVAL;
+	}
+
+	return dm_stats_print(dm_get_stats(md), id, idx_start, idx_len, clear,
+			      result, maxlen);
+}
+
+static int message_stats_set_aux(struct mapped_device *md,
+				 unsigned argc, char **argv)
+{
+	int id;
+	char dummy;
+
+	if (argc != 3)
+		return -EINVAL;
+
+	if (sscanf(argv[1], "%d%c", &id, &dummy) != 1 || id < 0)
+		return -EINVAL;
+
+	return dm_stats_set_aux(dm_get_stats(md), id, argv[2]);
+}
+
+int dm_stats_message(struct mapped_device *md, unsigned argc, char **argv,
+		     char *result, unsigned maxlen)
+{
+	int r;
+
+	if (dm_request_based(md)) {
+		DMWARN("Statistics are only supported for bio-based devices");
+		return -EOPNOTSUPP;
+	}
+
+	/* All messages here must start with '@' */
+	if (!strcasecmp(argv[0], "@stats_create"))
+		r = message_stats_create(md, argc, argv, result, maxlen);
+	else if (!strcasecmp(argv[0], "@stats_delete"))
+		r = message_stats_delete(md, argc, argv);
+	else if (!strcasecmp(argv[0], "@stats_clear"))
+		r = message_stats_clear(md, argc, argv);
+	else if (!strcasecmp(argv[0], "@stats_list"))
+		r = message_stats_list(md, argc, argv, result, maxlen);
+	else if (!strcasecmp(argv[0], "@stats_print"))
+		r = message_stats_print(md, argc, argv, false, result, maxlen);
+	else if (!strcasecmp(argv[0], "@stats_print_clear"))
+		r = message_stats_print(md, argc, argv, true, result, maxlen);
+	else if (!strcasecmp(argv[0], "@stats_set_aux"))
+		r = message_stats_set_aux(md, argc, argv);
+	else
+		return 2; /* this wasn't a stats message */
+
+	if (r == -EINVAL)
+		DMWARN("Invalid parameters for message %s", argv[0]);
+
+	return r;
+}
+
+int __init dm_statistics_init(void)
+{
+	shared_memory_amount = 0;
+	dm_stat_need_rcu_barrier = 0;
+	return 0;
+}
+
+void dm_statistics_exit(void)
+{
+	if (dm_stat_need_rcu_barrier)
+		rcu_barrier();
+	if (WARN_ON(shared_memory_amount))
+		DMCRIT("shared_memory_amount leaked: %lu", shared_memory_amount);
+}
+
+module_param_named(stats_current_allocated_bytes, shared_memory_amount, ulong, S_IRUGO);
+MODULE_PARM_DESC(stats_current_allocated_bytes, "Memory currently used by statistics");
diff --git a/drivers/md/dm-stats.h b/drivers/md/dm-stats.h
new file mode 100644
index 00000000000..e7c4984bf23
--- /dev/null
+++ b/drivers/md/dm-stats.h
@@ -0,0 +1,40 @@
+#ifndef DM_STATS_H
+#define DM_STATS_H
+
+#include <linux/types.h>
+#include <linux/mutex.h>
+#include <linux/list.h>
+
+int dm_statistics_init(void);
+void dm_statistics_exit(void);
+
+struct dm_stats {
+	struct mutex mutex;
+	struct list_head list;	/* list of struct dm_stat */
+	struct dm_stats_last_position __percpu *last;
+	sector_t last_sector;
+	unsigned last_rw;
+};
+
+struct dm_stats_aux {
+	bool merged;
+};
+
+void dm_stats_init(struct dm_stats *st);
+void dm_stats_cleanup(struct dm_stats *st);
+
+struct mapped_device;
+
+int dm_stats_message(struct mapped_device *md, unsigned argc, char **argv,
+		     char *result, unsigned maxlen);
+
+void dm_stats_account_io(struct dm_stats *stats, unsigned long bi_rw,
+			 sector_t bi_sector, unsigned bi_sectors, bool end,
+			 unsigned long duration, struct dm_stats_aux *aux);
+
+static inline bool dm_stats_used(struct dm_stats *st)
+{
+	return !list_empty(&st->list);
+}
+
+#endif
diff --git a/drivers/md/dm-stripe.c b/drivers/md/dm-stripe.c
index 4de90ab3968..d1600d2aa2e 100644
--- a/drivers/md/dm-stripe.c
+++ b/drivers/md/dm-stripe.c
@@ -5,6 +5,7 @@
  */
 
 #include "dm.h"
+#include <linux/device-mapper.h>
 
 #include <linux/module.h>
 #include <linux/init.h>
@@ -25,43 +26,40 @@ struct stripe {
 
 struct stripe_c {
 	uint32_t stripes;
+	int stripes_shift;
 
 	/* 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;
 
 	/* Work struct used for triggering events*/
-	struct work_struct kstriped_ws;
+	struct work_struct trigger_event;
 
 	struct stripe stripe[0];
 };
 
-static struct workqueue_struct *kstriped;
-
 /*
  * An event is triggered whenever a drive
  * drops out of a stripe volume.
  */
 static void trigger_event(struct work_struct *work)
 {
-	struct stripe_c *sc = container_of(work, struct stripe_c, kstriped_ws);
-
+	struct stripe_c *sc = container_of(work, struct stripe_c,
+					   trigger_event);
 	dm_table_event(sc->ti->table);
-
 }
 
 static inline struct stripe_c *alloc_context(unsigned int stripes)
 {
 	size_t len;
 
-	if (array_too_big(sizeof(struct stripe_c), sizeof(struct stripe),
-			  stripes))
+	if (dm_array_too_big(sizeof(struct stripe_c), sizeof(struct stripe),
+			     stripes))
 		return NULL;
 
 	len = sizeof(struct stripe_c) + (sizeof(struct stripe) * stripes);
@@ -76,12 +74,12 @@ static int get_stripe(struct dm_target *ti, struct stripe_c *sc,
 		      unsigned int stripe, char **argv)
 {
 	unsigned long long start;
+	char dummy;
 
-	if (sscanf(argv[1], "%llu", &start) != 1)
+	if (sscanf(argv[1], "%llu%c", &start, &dummy) != 1)
 		return -EINVAL;
 
-	if (dm_get_device(ti, argv[0], start, sc->stripe_width,
-			  dm_table_get_mode(ti->table),
+	if (dm_get_device(ti, argv[0], dm_table_get_mode(ti->table),
 			  &sc->stripe[stripe].dev))
 		return -ENXIO;
 
@@ -92,15 +90,14 @@ 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)
 {
 	struct stripe_c *sc;
-	sector_t width;
+	sector_t width, tmp_len;
 	uint32_t stripes;
 	uint32_t chunk_size;
-	char *end;
 	int r;
 	unsigned int i;
 
@@ -109,37 +106,27 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -EINVAL;
 	}
 
-	stripes = simple_strtoul(argv[0], &end, 10);
-	if (*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, stripes)) {
 		ti->error = "Target length not divisible by "
-		    "chunk size";
+		    "number of stripes";
 		return -EINVAL;
 	}
 
-	width = ti->len;
-	if (sector_div(width, stripes)) {
+	tmp_len = width;
+	if (sector_div(tmp_len, chunk_size)) {
 		ti->error = "Target length not divisible by "
-		    "number of stripes";
+		    "chunk size";
 		return -EINVAL;
 	}
 
@@ -159,19 +146,31 @@ static int stripe_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -ENOMEM;
 	}
 
-	INIT_WORK(&sc->kstriped_ws, trigger_event);
+	INIT_WORK(&sc->trigger_event, trigger_event);
 
 	/* Set pointer to dm target; used in trigger_event */
 	sc->ti = ti;
-
 	sc->stripes = stripes;
 	sc->stripe_width = width;
-	ti->split_io = chunk_size;
 
-	sc->chunk_mask = ((sector_t) chunk_size) - 1;
-	for (sc->chunk_shift = 0; chunk_size; sc->chunk_shift++)
-		chunk_size >>= 1;
-	sc->chunk_shift--;
+	if (stripes & (stripes - 1))
+		sc->stripes_shift = -1;
+	else
+		sc->stripes_shift = __ffs(stripes);
+
+	r = dm_set_target_max_io_len(ti, chunk_size);
+	if (r)
+		return r;
+
+	ti->num_flush_bios = stripes;
+	ti->num_discard_bios = stripes;
+	ti->num_write_same_bios = stripes;
+
+	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.
@@ -203,22 +202,105 @@ static void stripe_dtr(struct dm_target *ti)
 	for (i = 0; i < sc->stripes; i++)
 		dm_put_device(ti, sc->stripe[i].dev);
 
-	flush_workqueue(kstriped);
+	flush_work(&sc->trigger_event);
 	kfree(sc);
 }
 
-static int stripe_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static void stripe_map_sector(struct stripe_c *sc, sector_t sector,
+			      uint32_t *stripe, sector_t *result)
 {
-	struct stripe_c *sc = (struct stripe_c *) ti->private;
+	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 - 1);
+		chunk >>= sc->stripes_shift;
+	}
+
+	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,
+				    uint32_t target_stripe, sector_t *result)
+{
+	uint32_t stripe;
+
+	stripe_map_sector(sc, sector, &stripe, result);
+	if (stripe == target_stripe)
+		return;
+
+	/* 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_size;		/* next chunk */
+}
 
-	sector_t offset = bio->bi_sector - ti->begin;
-	sector_t chunk = offset >> sc->chunk_shift;
-	uint32_t stripe = sector_div(chunk, sc->stripes);
+static int stripe_map_range(struct stripe_c *sc, struct bio *bio,
+			    uint32_t target_stripe)
+{
+	sector_t begin, end;
+
+	stripe_map_range_sector(sc, bio->bi_iter.bi_sector,
+				target_stripe, &begin);
+	stripe_map_range_sector(sc, bio_end_sector(bio),
+				target_stripe, &end);
+	if (begin < end) {
+		bio->bi_bdev = sc->stripe[target_stripe].dev->bdev;
+		bio->bi_iter.bi_sector = begin +
+			sc->stripe[target_stripe].physical_start;
+		bio->bi_iter.bi_size = to_bytes(end - begin);
+		return DM_MAPIO_REMAPPED;
+	} else {
+		/* The range doesn't map to the target stripe */
+		bio_endio(bio, 0);
+		return DM_MAPIO_SUBMITTED;
+	}
+}
+
+static int stripe_map(struct dm_target *ti, struct bio *bio)
+{
+	struct stripe_c *sc = ti->private;
+	uint32_t stripe;
+	unsigned target_bio_nr;
+
+	if (bio->bi_rw & REQ_FLUSH) {
+		target_bio_nr = dm_bio_get_target_bio_nr(bio);
+		BUG_ON(target_bio_nr >= sc->stripes);
+		bio->bi_bdev = sc->stripe[target_bio_nr].dev->bdev;
+		return DM_MAPIO_REMAPPED;
+	}
+	if (unlikely(bio->bi_rw & REQ_DISCARD) ||
+	    unlikely(bio->bi_rw & REQ_WRITE_SAME)) {
+		target_bio_nr = dm_bio_get_target_bio_nr(bio);
+		BUG_ON(target_bio_nr >= sc->stripes);
+		return stripe_map_range(sc, bio, target_bio_nr);
+	}
+
+	stripe_map_sector(sc, bio->bi_iter.bi_sector,
+			  &stripe, &bio->bi_iter.bi_sector);
 
+	bio->bi_iter.bi_sector += sc->stripe[stripe].physical_start;
 	bio->bi_bdev = sc->stripe[stripe].dev->bdev;
-	bio->bi_sector = sc->stripe[stripe].physical_start +
-	    (chunk << sc->chunk_shift) + (offset & sc->chunk_mask);
+
 	return DM_MAPIO_REMAPPED;
 }
 
@@ -235,8 +317,8 @@ 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 void 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];
@@ -257,17 +339,15 @@ 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);
 		break;
 	}
-	return 0;
 }
 
-static int stripe_end_io(struct dm_target *ti, struct bio *bio,
-			 int error, union map_info *map_context)
+static int stripe_end_io(struct dm_target *ti, struct bio *bio, int error)
 {
 	unsigned i;
 	char major_minor[16];
@@ -276,7 +356,7 @@ static int stripe_end_io(struct dm_target *ti, struct bio *bio,
 	if (!error)
 		return 0; /* I/O complete */
 
-	if ((error == -EWOULDBLOCK) && bio_rw_ahead(bio))
+	if ((error == -EWOULDBLOCK) && (bio->bi_rw & REQ_RAHEAD))
 		return error;
 
 	if (error == -EOPNOTSUPP)
@@ -284,8 +364,8 @@ static int stripe_end_io(struct dm_target *ti, struct bio *bio,
 
 	memset(major_minor, 0, sizeof(major_minor));
 	sprintf(major_minor, "%d:%d",
-		bio->bi_bdev->bd_disk->major,
-		bio->bi_bdev->bd_disk->first_minor);
+		MAJOR(disk_devt(bio->bi_bdev->bd_disk)),
+		MINOR(disk_devt(bio->bi_bdev->bd_disk)));
 
 	/*
 	 * Test to see which stripe drive triggered the event
@@ -298,21 +378,70 @@ static int stripe_end_io(struct dm_target *ti, struct bio *bio,
 			atomic_inc(&(sc->stripe[i].error_count));
 			if (atomic_read(&(sc->stripe[i].error_count)) <
 			    DM_IO_ERROR_THRESHOLD)
-				queue_work(kstriped, &sc->kstriped_ws);
+				schedule_work(&sc->trigger_event);
 		}
 
 	return error;
 }
 
+static int stripe_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct stripe_c *sc = ti->private;
+	int ret = 0;
+	unsigned i = 0;
+
+	do {
+		ret = fn(ti, sc->stripe[i].dev,
+			 sc->stripe[i].physical_start,
+			 sc->stripe_width, data);
+	} while (!ret && ++i < sc->stripes);
+
+	return ret;
+}
+
+static void stripe_io_hints(struct dm_target *ti,
+			    struct queue_limits *limits)
+{
+	struct stripe_c *sc = ti->private;
+	unsigned chunk_size = sc->chunk_size << SECTOR_SHIFT;
+
+	blk_limits_io_min(limits, chunk_size);
+	blk_limits_io_opt(limits, chunk_size * sc->stripes);
+}
+
+static int stripe_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+			struct bio_vec *biovec, int max_size)
+{
+	struct stripe_c *sc = ti->private;
+	sector_t bvm_sector = bvm->bi_sector;
+	uint32_t stripe;
+	struct request_queue *q;
+
+	stripe_map_sector(sc, bvm_sector, &stripe, &bvm_sector);
+
+	q = bdev_get_queue(sc->stripe[stripe].dev->bdev);
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = sc->stripe[stripe].dev->bdev;
+	bvm->bi_sector = sc->stripe[stripe].physical_start + bvm_sector;
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
 static struct target_type stripe_target = {
 	.name   = "striped",
-	.version = {1, 1, 0},
+	.version = {1, 5, 1},
 	.module = THIS_MODULE,
 	.ctr    = stripe_ctr,
 	.dtr    = stripe_dtr,
 	.map    = stripe_map,
 	.end_io = stripe_end_io,
 	.status = stripe_status,
+	.iterate_devices = stripe_iterate_devices,
+	.io_hints = stripe_io_hints,
+	.merge  = stripe_merge,
 };
 
 int __init dm_stripe_init(void)
@@ -320,14 +449,9 @@ int __init dm_stripe_init(void)
 	int r;
 
 	r = dm_register_target(&stripe_target);
-	if (r < 0)
+	if (r < 0) {
 		DMWARN("target registration failed");
-
-	kstriped = create_singlethread_workqueue("kstriped");
-	if (!kstriped) {
-		DMERR("failed to create workqueue kstriped");
-		dm_unregister_target(&stripe_target);
-		return -ENOMEM;
+		return r;
 	}
 
 	return r;
@@ -335,10 +459,5 @@ int __init dm_stripe_init(void)
 
 void dm_stripe_exit(void)
 {
-	if (dm_unregister_target(&stripe_target))
-		DMWARN("target unregistration failed");
-
-	destroy_workqueue(kstriped);
-
-	return;
+	dm_unregister_target(&stripe_target);
 }
diff --git a/drivers/md/dm-switch.c b/drivers/md/dm-switch.c
new file mode 100644
index 00000000000..09a688b3d48
--- /dev/null
+++ b/drivers/md/dm-switch.c
@@ -0,0 +1,538 @@
+/*
+ * Copyright (C) 2010-2012 by Dell Inc.  All rights reserved.
+ * Copyright (C) 2011-2013 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ *
+ * dm-switch is a device-mapper target that maps IO to underlying block
+ * devices efficiently when there are a large number of fixed-sized
+ * address regions but there is no simple pattern to allow for a compact
+ * mapping representation such as dm-stripe.
+ */
+
+#include <linux/device-mapper.h>
+
+#include <linux/module.h>
+#include <linux/init.h>
+#include <linux/vmalloc.h>
+
+#define DM_MSG_PREFIX "switch"
+
+/*
+ * One region_table_slot_t holds <region_entries_per_slot> region table
+ * entries each of which is <region_table_entry_bits> in size.
+ */
+typedef unsigned long region_table_slot_t;
+
+/*
+ * A device with the offset to its start sector.
+ */
+struct switch_path {
+	struct dm_dev *dmdev;
+	sector_t start;
+};
+
+/*
+ * Context block for a dm switch device.
+ */
+struct switch_ctx {
+	struct dm_target *ti;
+
+	unsigned nr_paths;		/* Number of paths in path_list. */
+
+	unsigned region_size;		/* Region size in 512-byte sectors */
+	unsigned long nr_regions;	/* Number of regions making up the device */
+	signed char region_size_bits;	/* log2 of region_size or -1 */
+
+	unsigned char region_table_entry_bits;	/* Number of bits in one region table entry */
+	unsigned char region_entries_per_slot;	/* Number of entries in one region table slot */
+	signed char region_entries_per_slot_bits;	/* log2 of region_entries_per_slot or -1 */
+
+	region_table_slot_t *region_table;	/* Region table */
+
+	/*
+	 * Array of dm devices to switch between.
+	 */
+	struct switch_path path_list[0];
+};
+
+static struct switch_ctx *alloc_switch_ctx(struct dm_target *ti, unsigned nr_paths,
+					   unsigned region_size)
+{
+	struct switch_ctx *sctx;
+
+	sctx = kzalloc(sizeof(struct switch_ctx) + nr_paths * sizeof(struct switch_path),
+		       GFP_KERNEL);
+	if (!sctx)
+		return NULL;
+
+	sctx->ti = ti;
+	sctx->region_size = region_size;
+
+	ti->private = sctx;
+
+	return sctx;
+}
+
+static int alloc_region_table(struct dm_target *ti, unsigned nr_paths)
+{
+	struct switch_ctx *sctx = ti->private;
+	sector_t nr_regions = ti->len;
+	sector_t nr_slots;
+
+	if (!(sctx->region_size & (sctx->region_size - 1)))
+		sctx->region_size_bits = __ffs(sctx->region_size);
+	else
+		sctx->region_size_bits = -1;
+
+	sctx->region_table_entry_bits = 1;
+	while (sctx->region_table_entry_bits < sizeof(region_table_slot_t) * 8 &&
+	       (region_table_slot_t)1 << sctx->region_table_entry_bits < nr_paths)
+		sctx->region_table_entry_bits++;
+
+	sctx->region_entries_per_slot = (sizeof(region_table_slot_t) * 8) / sctx->region_table_entry_bits;
+	if (!(sctx->region_entries_per_slot & (sctx->region_entries_per_slot - 1)))
+		sctx->region_entries_per_slot_bits = __ffs(sctx->region_entries_per_slot);
+	else
+		sctx->region_entries_per_slot_bits = -1;
+
+	if (sector_div(nr_regions, sctx->region_size))
+		nr_regions++;
+
+	sctx->nr_regions = nr_regions;
+	if (sctx->nr_regions != nr_regions || sctx->nr_regions >= ULONG_MAX) {
+		ti->error = "Region table too large";
+		return -EINVAL;
+	}
+
+	nr_slots = nr_regions;
+	if (sector_div(nr_slots, sctx->region_entries_per_slot))
+		nr_slots++;
+
+	if (nr_slots > ULONG_MAX / sizeof(region_table_slot_t)) {
+		ti->error = "Region table too large";
+		return -EINVAL;
+	}
+
+	sctx->region_table = vmalloc(nr_slots * sizeof(region_table_slot_t));
+	if (!sctx->region_table) {
+		ti->error = "Cannot allocate region table";
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void switch_get_position(struct switch_ctx *sctx, unsigned long region_nr,
+				unsigned long *region_index, unsigned *bit)
+{
+	if (sctx->region_entries_per_slot_bits >= 0) {
+		*region_index = region_nr >> sctx->region_entries_per_slot_bits;
+		*bit = region_nr & (sctx->region_entries_per_slot - 1);
+	} else {
+		*region_index = region_nr / sctx->region_entries_per_slot;
+		*bit = region_nr % sctx->region_entries_per_slot;
+	}
+
+	*bit *= sctx->region_table_entry_bits;
+}
+
+/*
+ * Find which path to use at given offset.
+ */
+static unsigned switch_get_path_nr(struct switch_ctx *sctx, sector_t offset)
+{
+	unsigned long region_index;
+	unsigned bit, path_nr;
+	sector_t p;
+
+	p = offset;
+	if (sctx->region_size_bits >= 0)
+		p >>= sctx->region_size_bits;
+	else
+		sector_div(p, sctx->region_size);
+
+	switch_get_position(sctx, p, &region_index, &bit);
+	path_nr = (ACCESS_ONCE(sctx->region_table[region_index]) >> bit) &
+	       ((1 << sctx->region_table_entry_bits) - 1);
+
+	/* This can only happen if the processor uses non-atomic stores. */
+	if (unlikely(path_nr >= sctx->nr_paths))
+		path_nr = 0;
+
+	return path_nr;
+}
+
+static void switch_region_table_write(struct switch_ctx *sctx, unsigned long region_nr,
+				      unsigned value)
+{
+	unsigned long region_index;
+	unsigned bit;
+	region_table_slot_t pte;
+
+	switch_get_position(sctx, region_nr, &region_index, &bit);
+
+	pte = sctx->region_table[region_index];
+	pte &= ~((((region_table_slot_t)1 << sctx->region_table_entry_bits) - 1) << bit);
+	pte |= (region_table_slot_t)value << bit;
+	sctx->region_table[region_index] = pte;
+}
+
+/*
+ * Fill the region table with an initial round robin pattern.
+ */
+static void initialise_region_table(struct switch_ctx *sctx)
+{
+	unsigned path_nr = 0;
+	unsigned long region_nr;
+
+	for (region_nr = 0; region_nr < sctx->nr_regions; region_nr++) {
+		switch_region_table_write(sctx, region_nr, path_nr);
+		if (++path_nr >= sctx->nr_paths)
+			path_nr = 0;
+	}
+}
+
+static int parse_path(struct dm_arg_set *as, struct dm_target *ti)
+{
+	struct switch_ctx *sctx = ti->private;
+	unsigned long long start;
+	int r;
+
+	r = dm_get_device(ti, dm_shift_arg(as), dm_table_get_mode(ti->table),
+			  &sctx->path_list[sctx->nr_paths].dmdev);
+	if (r) {
+		ti->error = "Device lookup failed";
+		return r;
+	}
+
+	if (kstrtoull(dm_shift_arg(as), 10, &start) || start != (sector_t)start) {
+		ti->error = "Invalid device starting offset";
+		dm_put_device(ti, sctx->path_list[sctx->nr_paths].dmdev);
+		return -EINVAL;
+	}
+
+	sctx->path_list[sctx->nr_paths].start = start;
+
+	sctx->nr_paths++;
+
+	return 0;
+}
+
+/*
+ * Destructor: Don't free the dm_target, just the ti->private data (if any).
+ */
+static void switch_dtr(struct dm_target *ti)
+{
+	struct switch_ctx *sctx = ti->private;
+
+	while (sctx->nr_paths--)
+		dm_put_device(ti, sctx->path_list[sctx->nr_paths].dmdev);
+
+	vfree(sctx->region_table);
+	kfree(sctx);
+}
+
+/*
+ * Constructor arguments:
+ *   <num_paths> <region_size> <num_optional_args> [<optional_args>...]
+ *   [<dev_path> <offset>]+
+ *
+ * Optional args are to allow for future extension: currently this
+ * parameter must be 0.
+ */
+static int switch_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	static struct dm_arg _args[] = {
+		{1, (KMALLOC_MAX_SIZE - sizeof(struct switch_ctx)) / sizeof(struct switch_path), "Invalid number of paths"},
+		{1, UINT_MAX, "Invalid region size"},
+		{0, 0, "Invalid number of optional args"},
+	};
+
+	struct switch_ctx *sctx;
+	struct dm_arg_set as;
+	unsigned nr_paths, region_size, nr_optional_args;
+	int r;
+
+	as.argc = argc;
+	as.argv = argv;
+
+	r = dm_read_arg(_args, &as, &nr_paths, &ti->error);
+	if (r)
+		return -EINVAL;
+
+	r = dm_read_arg(_args + 1, &as, &region_size, &ti->error);
+	if (r)
+		return r;
+
+	r = dm_read_arg_group(_args + 2, &as, &nr_optional_args, &ti->error);
+	if (r)
+		return r;
+	/* parse optional arguments here, if we add any */
+
+	if (as.argc != nr_paths * 2) {
+		ti->error = "Incorrect number of path arguments";
+		return -EINVAL;
+	}
+
+	sctx = alloc_switch_ctx(ti, nr_paths, region_size);
+	if (!sctx) {
+		ti->error = "Cannot allocate redirection context";
+		return -ENOMEM;
+	}
+
+	r = dm_set_target_max_io_len(ti, region_size);
+	if (r)
+		goto error;
+
+	while (as.argc) {
+		r = parse_path(&as, ti);
+		if (r)
+			goto error;
+	}
+
+	r = alloc_region_table(ti, nr_paths);
+	if (r)
+		goto error;
+
+	initialise_region_table(sctx);
+
+	/* For UNMAP, sending the request down any path is sufficient */
+	ti->num_discard_bios = 1;
+
+	return 0;
+
+error:
+	switch_dtr(ti);
+
+	return r;
+}
+
+static int switch_map(struct dm_target *ti, struct bio *bio)
+{
+	struct switch_ctx *sctx = ti->private;
+	sector_t offset = dm_target_offset(ti, bio->bi_iter.bi_sector);
+	unsigned path_nr = switch_get_path_nr(sctx, offset);
+
+	bio->bi_bdev = sctx->path_list[path_nr].dmdev->bdev;
+	bio->bi_iter.bi_sector = sctx->path_list[path_nr].start + offset;
+
+	return DM_MAPIO_REMAPPED;
+}
+
+/*
+ * We need to parse hex numbers in the message as quickly as possible.
+ *
+ * This table-based hex parser improves performance.
+ * It improves a time to load 1000000 entries compared to the condition-based
+ * parser.
+ *		table-based parser	condition-based parser
+ * PA-RISC	0.29s			0.31s
+ * Opteron	0.0495s			0.0498s
+ */
+static const unsigned char hex_table[256] = {
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 255, 255, 255, 255, 255, 255,
+255, 10, 11, 12, 13, 14, 15, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 10, 11, 12, 13, 14, 15, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255,
+255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255, 255
+};
+
+static __always_inline unsigned long parse_hex(const char **string)
+{
+	unsigned char d;
+	unsigned long r = 0;
+
+	while ((d = hex_table[(unsigned char)**string]) < 16) {
+		r = (r << 4) | d;
+		(*string)++;
+	}
+
+	return r;
+}
+
+static int process_set_region_mappings(struct switch_ctx *sctx,
+			     unsigned argc, char **argv)
+{
+	unsigned i;
+	unsigned long region_index = 0;
+
+	for (i = 1; i < argc; i++) {
+		unsigned long path_nr;
+		const char *string = argv[i];
+
+		if (*string == ':')
+			region_index++;
+		else {
+			region_index = parse_hex(&string);
+			if (unlikely(*string != ':')) {
+				DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+				return -EINVAL;
+			}
+		}
+
+		string++;
+		if (unlikely(!*string)) {
+			DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+			return -EINVAL;
+		}
+
+		path_nr = parse_hex(&string);
+		if (unlikely(*string)) {
+			DMWARN("invalid set_region_mappings argument: '%s'", argv[i]);
+			return -EINVAL;
+		}
+		if (unlikely(region_index >= sctx->nr_regions)) {
+			DMWARN("invalid set_region_mappings region number: %lu >= %lu", region_index, sctx->nr_regions);
+			return -EINVAL;
+		}
+		if (unlikely(path_nr >= sctx->nr_paths)) {
+			DMWARN("invalid set_region_mappings device: %lu >= %u", path_nr, sctx->nr_paths);
+			return -EINVAL;
+		}
+
+		switch_region_table_write(sctx, region_index, path_nr);
+	}
+
+	return 0;
+}
+
+/*
+ * Messages are processed one-at-a-time.
+ *
+ * Only set_region_mappings is supported.
+ */
+static int switch_message(struct dm_target *ti, unsigned argc, char **argv)
+{
+	static DEFINE_MUTEX(message_mutex);
+
+	struct switch_ctx *sctx = ti->private;
+	int r = -EINVAL;
+
+	mutex_lock(&message_mutex);
+
+	if (!strcasecmp(argv[0], "set_region_mappings"))
+		r = process_set_region_mappings(sctx, argc, argv);
+	else
+		DMWARN("Unrecognised message received.");
+
+	mutex_unlock(&message_mutex);
+
+	return r;
+}
+
+static void switch_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
+{
+	struct switch_ctx *sctx = ti->private;
+	unsigned sz = 0;
+	int path_nr;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		result[0] = '\0';
+		break;
+
+	case STATUSTYPE_TABLE:
+		DMEMIT("%u %u 0", sctx->nr_paths, sctx->region_size);
+		for (path_nr = 0; path_nr < sctx->nr_paths; path_nr++)
+			DMEMIT(" %s %llu", sctx->path_list[path_nr].dmdev->name,
+			       (unsigned long long)sctx->path_list[path_nr].start);
+		break;
+	}
+}
+
+/*
+ * Switch ioctl:
+ *
+ * Passthrough all ioctls to the path for sector 0
+ */
+static int switch_ioctl(struct dm_target *ti, unsigned cmd,
+			unsigned long arg)
+{
+	struct switch_ctx *sctx = ti->private;
+	struct block_device *bdev;
+	fmode_t mode;
+	unsigned path_nr;
+	int r = 0;
+
+	path_nr = switch_get_path_nr(sctx, 0);
+
+	bdev = sctx->path_list[path_nr].dmdev->bdev;
+	mode = sctx->path_list[path_nr].dmdev->mode;
+
+	/*
+	 * Only pass ioctls through if the device sizes match exactly.
+	 */
+	if (ti->len + sctx->path_list[path_nr].start != i_size_read(bdev->bd_inode) >> SECTOR_SHIFT)
+		r = scsi_verify_blk_ioctl(NULL, cmd);
+
+	return r ? : __blkdev_driver_ioctl(bdev, mode, cmd, arg);
+}
+
+static int switch_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct switch_ctx *sctx = ti->private;
+	int path_nr;
+	int r;
+
+	for (path_nr = 0; path_nr < sctx->nr_paths; path_nr++) {
+		r = fn(ti, sctx->path_list[path_nr].dmdev,
+			 sctx->path_list[path_nr].start, ti->len, data);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+static struct target_type switch_target = {
+	.name = "switch",
+	.version = {1, 0, 0},
+	.module = THIS_MODULE,
+	.ctr = switch_ctr,
+	.dtr = switch_dtr,
+	.map = switch_map,
+	.message = switch_message,
+	.status = switch_status,
+	.ioctl = switch_ioctl,
+	.iterate_devices = switch_iterate_devices,
+};
+
+static int __init dm_switch_init(void)
+{
+	int r;
+
+	r = dm_register_target(&switch_target);
+	if (r < 0)
+		DMERR("dm_register_target() failed %d", r);
+
+	return r;
+}
+
+static void __exit dm_switch_exit(void)
+{
+	dm_unregister_target(&switch_target);
+}
+
+module_init(dm_switch_init);
+module_exit(dm_switch_exit);
+
+MODULE_DESCRIPTION(DM_NAME " dynamic path switching target");
+MODULE_AUTHOR("Kevin D. O'Kelley <Kevin_OKelley@dell.com>");
+MODULE_AUTHOR("Narendran Ganapathy <Narendran_Ganapathy@dell.com>");
+MODULE_AUTHOR("Jim Ramsay <Jim_Ramsay@dell.com>");
+MODULE_AUTHOR("Mikulas Patocka <mpatocka@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-sysfs.c b/drivers/md/dm-sysfs.c
new file mode 100644
index 00000000000..c62c5ab6aed
--- /dev/null
+++ b/drivers/md/dm-sysfs.c
@@ -0,0 +1,111 @@
+/*
+ * Copyright (C) 2008 Red Hat, Inc. All rights reserved.
+ *
+ * This file is released under the GPL.
+ */
+
+#include <linux/sysfs.h>
+#include <linux/dm-ioctl.h>
+#include "dm.h"
+
+struct dm_sysfs_attr {
+	struct attribute attr;
+	ssize_t (*show)(struct mapped_device *, char *);
+	ssize_t (*store)(struct mapped_device *, char *);
+};
+
+#define DM_ATTR_RO(_name) \
+struct dm_sysfs_attr dm_attr_##_name = \
+	__ATTR(_name, S_IRUGO, dm_attr_##_name##_show, NULL)
+
+static ssize_t dm_attr_show(struct kobject *kobj, struct attribute *attr,
+			    char *page)
+{
+	struct dm_sysfs_attr *dm_attr;
+	struct mapped_device *md;
+	ssize_t ret;
+
+	dm_attr = container_of(attr, struct dm_sysfs_attr, attr);
+	if (!dm_attr->show)
+		return -EIO;
+
+	md = dm_get_from_kobject(kobj);
+	if (!md)
+		return -EINVAL;
+
+	ret = dm_attr->show(md, page);
+	dm_put(md);
+
+	return ret;
+}
+
+static ssize_t dm_attr_name_show(struct mapped_device *md, char *buf)
+{
+	if (dm_copy_name_and_uuid(md, buf, NULL))
+		return -EIO;
+
+	strcat(buf, "\n");
+	return strlen(buf);
+}
+
+static ssize_t dm_attr_uuid_show(struct mapped_device *md, char *buf)
+{
+	if (dm_copy_name_and_uuid(md, NULL, buf))
+		return -EIO;
+
+	strcat(buf, "\n");
+	return strlen(buf);
+}
+
+static ssize_t dm_attr_suspended_show(struct mapped_device *md, char *buf)
+{
+	sprintf(buf, "%d\n", dm_suspended_md(md));
+
+	return strlen(buf);
+}
+
+static DM_ATTR_RO(name);
+static DM_ATTR_RO(uuid);
+static DM_ATTR_RO(suspended);
+
+static struct attribute *dm_attrs[] = {
+	&dm_attr_name.attr,
+	&dm_attr_uuid.attr,
+	&dm_attr_suspended.attr,
+	NULL,
+};
+
+static const struct sysfs_ops dm_sysfs_ops = {
+	.show	= dm_attr_show,
+};
+
+/*
+ * dm kobject is embedded in mapped_device structure
+ * no need to define release function here
+ */
+static struct kobj_type dm_ktype = {
+	.sysfs_ops	= &dm_sysfs_ops,
+	.default_attrs	= dm_attrs,
+	.release	= dm_kobject_release,
+};
+
+/*
+ * Initialize kobj
+ * because nobody using md yet, no need to call explicit dm_get/put
+ */
+int dm_sysfs_init(struct mapped_device *md)
+{
+	return kobject_init_and_add(dm_kobject(md), &dm_ktype,
+				    &disk_to_dev(dm_disk(md))->kobj,
+				    "%s", "dm");
+}
+
+/*
+ * Remove kobj, called after all references removed
+ */
+void dm_sysfs_exit(struct mapped_device *md)
+{
+	struct kobject *kobj = dm_kobject(md);
+	kobject_put(kobj);
+	wait_for_completion(dm_get_completion_from_kobject(kobj));
+}
diff --git a/drivers/md/dm-table.c b/drivers/md/dm-table.c
index 61f44140923..5f59f1e3e5b 100644
--- a/drivers/md/dm-table.c
+++ b/drivers/md/dm-table.c
@@ -1,6 +1,6 @@
 /*
  * Copyright (C) 2001 Sistina Software (UK) Limited.
- * Copyright (C) 2004 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
@@ -12,10 +12,12 @@
 #include <linux/blkdev.h>
 #include <linux/namei.h>
 #include <linux/ctype.h>
+#include <linux/string.h>
 #include <linux/slab.h>
 #include <linux/interrupt.h>
 #include <linux/mutex.h>
-#include <asm/atomic.h>
+#include <linux/delay.h>
+#include <linux/atomic.h>
 
 #define DM_MSG_PREFIX "table"
 
@@ -26,7 +28,7 @@
 
 struct dm_table {
 	struct mapped_device *md;
-	atomic_t holders;
+	unsigned type;
 
 	/* btree table */
 	unsigned int depth;
@@ -38,25 +40,27 @@ struct dm_table {
 	sector_t *highs;
 	struct dm_target *targets;
 
+	struct target_type *immutable_target_type;
+	unsigned integrity_supported:1;
+	unsigned singleton:1;
+
 	/*
 	 * Indicates the rw permissions for the new logical
 	 * device.  This should be a combination of FMODE_READ
 	 * and FMODE_WRITE.
 	 */
-	int mode;
+	fmode_t mode;
 
 	/* a list of devices used by this table */
 	struct list_head devices;
 
-	/*
-	 * These are optimistic limits taken from all the
-	 * targets, some targets will need smaller limits.
-	 */
-	struct io_restrictions limits;
-
 	/* events get handed up using this callback */
 	void (*event_fn)(void *);
 	void *event_context;
+
+	struct dm_md_mempools *mempools;
+
+	struct list_head target_callbacks;
 };
 
 /*
@@ -75,42 +79,6 @@ static unsigned int int_log(unsigned int n, unsigned int base)
 }
 
 /*
- * Returns the minimum that is _not_ zero, unless both are zero.
- */
-#define min_not_zero(l, r) (l == 0) ? r : ((r == 0) ? l : min(l, r))
-
-/*
- * Combine two io_restrictions, always taking the lower value.
- */
-static void combine_restrictions_low(struct io_restrictions *lhs,
-				     struct io_restrictions *rhs)
-{
-	lhs->max_sectors =
-		min_not_zero(lhs->max_sectors, rhs->max_sectors);
-
-	lhs->max_phys_segments =
-		min_not_zero(lhs->max_phys_segments, rhs->max_phys_segments);
-
-	lhs->max_hw_segments =
-		min_not_zero(lhs->max_hw_segments, rhs->max_hw_segments);
-
-	lhs->hardsect_size = max(lhs->hardsect_size, rhs->hardsect_size);
-
-	lhs->max_segment_size =
-		min_not_zero(lhs->max_segment_size, rhs->max_segment_size);
-
-	lhs->max_hw_sectors =
-		min_not_zero(lhs->max_hw_sectors, rhs->max_hw_sectors);
-
-	lhs->seg_boundary_mask =
-		min_not_zero(lhs->seg_boundary_mask, rhs->seg_boundary_mask);
-
-	lhs->bounce_pfn = min_not_zero(lhs->bounce_pfn, rhs->bounce_pfn);
-
-	lhs->no_cluster |= rhs->no_cluster;
-}
-
-/*
  * Calculate the index of the child node of the n'th node k'th key.
  */
 static inline unsigned int get_child(unsigned int n, unsigned int k)
@@ -173,12 +141,11 @@ void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size)
 		return NULL;
 
 	size = nmemb * elem_size;
-	addr = vmalloc(size);
-	if (addr)
-		memset(addr, 0, size);
+	addr = vzalloc(size);
 
 	return addr;
 }
+EXPORT_SYMBOL(dm_vcalloc);
 
 /*
  * highs, and targets are managed as dynamic arrays during a
@@ -188,7 +155,6 @@ static int alloc_targets(struct dm_table *t, unsigned int num)
 {
 	sector_t *n_highs;
 	struct dm_target *n_targets;
-	int n = t->num_targets;
 
 	/*
 	 * Allocate both the target array and offset array at once.
@@ -202,12 +168,7 @@ static int alloc_targets(struct dm_table *t, unsigned int num)
 
 	n_targets = (struct dm_target *) (n_highs + num);
 
-	if (n) {
-		memcpy(n_highs, t->highs, sizeof(*n_highs) * n);
-		memcpy(n_targets, t->targets, sizeof(*n_targets) * n);
-	}
-
-	memset(n_highs + n, -1, sizeof(*n_highs) * (num - n));
+	memset(n_highs, -1, sizeof(*n_highs) * num);
 	vfree(t->highs);
 
 	t->num_allocated = num;
@@ -217,7 +178,7 @@ static int alloc_targets(struct dm_table *t, unsigned int num)
 	return 0;
 }
 
-int dm_table_create(struct dm_table **result, int mode,
+int dm_table_create(struct dm_table **result, fmode_t mode,
 		    unsigned num_targets, struct mapped_device *md)
 {
 	struct dm_table *t = kzalloc(sizeof(*t), GFP_KERNEL);
@@ -226,16 +187,20 @@ int dm_table_create(struct dm_table **result, int mode,
 		return -ENOMEM;
 
 	INIT_LIST_HEAD(&t->devices);
-	atomic_set(&t->holders, 1);
+	INIT_LIST_HEAD(&t->target_callbacks);
 
 	if (!num_targets)
 		num_targets = KEYS_PER_NODE;
 
 	num_targets = dm_round_up(num_targets, KEYS_PER_NODE);
 
+	if (!num_targets) {
+		kfree(t);
+		return -ENOMEM;
+	}
+
 	if (alloc_targets(t, num_targets)) {
 		kfree(t);
-		t = NULL;
 		return -ENOMEM;
 	}
 
@@ -250,16 +215,22 @@ static void free_devices(struct list_head *devices)
 	struct list_head *tmp, *next;
 
 	list_for_each_safe(tmp, next, devices) {
-		struct dm_dev *dd = list_entry(tmp, struct dm_dev, list);
+		struct dm_dev_internal *dd =
+		    list_entry(tmp, struct dm_dev_internal, list);
+		DMWARN("dm_table_destroy: dm_put_device call missing for %s",
+		       dd->dm_dev.name);
 		kfree(dd);
 	}
 }
 
-static void table_destroy(struct dm_table *t)
+void dm_table_destroy(struct dm_table *t)
 {
 	unsigned int i;
 
-	/* free the indexes (see dm_table_complete) */
+	if (!t)
+		return;
+
+	/* free the indexes */
 	if (t->depth >= 2)
 		vfree(t->index[t->depth - 2]);
 
@@ -276,63 +247,22 @@ static void table_destroy(struct dm_table *t)
 	vfree(t->highs);
 
 	/* free the device list */
-	if (t->devices.next != &t->devices) {
-		DMWARN("devices still present during destroy: "
-		       "dm_table_remove_device calls missing");
+	free_devices(&t->devices);
 
-		free_devices(&t->devices);
-	}
+	dm_free_md_mempools(t->mempools);
 
 	kfree(t);
 }
 
-void dm_table_get(struct dm_table *t)
-{
-	atomic_inc(&t->holders);
-}
-
-void dm_table_put(struct dm_table *t)
-{
-	if (!t)
-		return;
-
-	if (atomic_dec_and_test(&t->holders))
-		table_destroy(t);
-}
-
-/*
- * Checks to see if we need to extend highs or targets.
- */
-static inline int check_space(struct dm_table *t)
-{
-	if (t->num_targets >= t->num_allocated)
-		return alloc_targets(t, t->num_allocated * 2);
-
-	return 0;
-}
-
-/*
- * Convert a device path to a dev_t.
- */
-static int lookup_device(const char *path, dev_t *dev)
-{
-	struct block_device *bdev = lookup_bdev(path);
-	if (IS_ERR(bdev))
-		return PTR_ERR(bdev);
-	*dev = bdev->bd_dev;
-	bdput(bdev);
-	return 0;
-}
-
 /*
  * See if we've already got a device in the list.
  */
-static struct dm_dev *find_device(struct list_head *l, dev_t dev)
+static struct dm_dev_internal *find_device(struct list_head *l, dev_t dev)
 {
-	struct dm_dev *dd;
+	struct dm_dev_internal *dd;
 
 	list_for_each_entry (dd, l, list)
-		if (dd->bdev->bd_dev == dev)
+		if (dd->dm_dev.bdev->bd_dev == dev)
 			return dd;
 
 	return NULL;
@@ -341,100 +271,167 @@ static struct dm_dev *find_device(struct list_head *l, dev_t dev)
 /*
  * Open a device so we can use it as a map destination.
  */
-static int open_dev(struct dm_dev *d, dev_t dev, struct mapped_device *md)
+static int open_dev(struct dm_dev_internal *d, dev_t dev,
+		    struct mapped_device *md)
 {
 	static char *_claim_ptr = "I belong to device-mapper";
 	struct block_device *bdev;
 
 	int r;
 
-	BUG_ON(d->bdev);
+	BUG_ON(d->dm_dev.bdev);
 
-	bdev = open_by_devnum(dev, d->mode);
+	bdev = blkdev_get_by_dev(dev, d->dm_dev.mode | FMODE_EXCL, _claim_ptr);
 	if (IS_ERR(bdev))
 		return PTR_ERR(bdev);
-	r = bd_claim_by_disk(bdev, _claim_ptr, dm_disk(md));
-	if (r)
-		blkdev_put(bdev);
-	else
-		d->bdev = bdev;
-	return r;
+
+	r = bd_link_disk_holder(bdev, dm_disk(md));
+	if (r) {
+		blkdev_put(bdev, d->dm_dev.mode | FMODE_EXCL);
+		return r;
+	}
+
+	d->dm_dev.bdev = bdev;
+	return 0;
 }
 
 /*
  * Close a device that we've been using.
  */
-static void close_dev(struct dm_dev *d, struct mapped_device *md)
+static void close_dev(struct dm_dev_internal *d, struct mapped_device *md)
 {
-	if (!d->bdev)
+	if (!d->dm_dev.bdev)
 		return;
 
-	bd_release_from_disk(d->bdev, dm_disk(md));
-	blkdev_put(d->bdev);
-	d->bdev = NULL;
+	bd_unlink_disk_holder(d->dm_dev.bdev, dm_disk(md));
+	blkdev_put(d->dm_dev.bdev, d->dm_dev.mode | FMODE_EXCL);
+	d->dm_dev.bdev = NULL;
 }
 
 /*
- * If possible, this checks an area of a destination device is valid.
+ * If possible, this checks an area of a destination device is invalid.
  */
-static int check_device_area(struct dm_dev *dd, sector_t start, sector_t len)
+static int device_area_is_invalid(struct dm_target *ti, struct dm_dev *dev,
+				  sector_t start, sector_t len, void *data)
 {
-	sector_t dev_size = dd->bdev->bd_inode->i_size >> SECTOR_SHIFT;
+	struct request_queue *q;
+	struct queue_limits *limits = data;
+	struct block_device *bdev = dev->bdev;
+	sector_t dev_size =
+		i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
+	unsigned short logical_block_size_sectors =
+		limits->logical_block_size >> SECTOR_SHIFT;
+	char b[BDEVNAME_SIZE];
+
+	/*
+	 * Some devices exist without request functions,
+	 * such as loop devices not yet bound to backing files.
+	 * Forbid the use of such devices.
+	 */
+	q = bdev_get_queue(bdev);
+	if (!q || !q->make_request_fn) {
+		DMWARN("%s: %s is not yet initialised: "
+		       "start=%llu, len=%llu, dev_size=%llu",
+		       dm_device_name(ti->table->md), bdevname(bdev, b),
+		       (unsigned long long)start,
+		       (unsigned long long)len,
+		       (unsigned long long)dev_size);
+		return 1;
+	}
 
 	if (!dev_size)
+		return 0;
+
+	if ((start >= dev_size) || (start + len > dev_size)) {
+		DMWARN("%s: %s too small for target: "
+		       "start=%llu, len=%llu, dev_size=%llu",
+		       dm_device_name(ti->table->md), bdevname(bdev, b),
+		       (unsigned long long)start,
+		       (unsigned long long)len,
+		       (unsigned long long)dev_size);
+		return 1;
+	}
+
+	if (logical_block_size_sectors <= 1)
+		return 0;
+
+	if (start & (logical_block_size_sectors - 1)) {
+		DMWARN("%s: start=%llu not aligned to h/w "
+		       "logical block size %u of %s",
+		       dm_device_name(ti->table->md),
+		       (unsigned long long)start,
+		       limits->logical_block_size, bdevname(bdev, b));
 		return 1;
+	}
 
-	return ((start < dev_size) && (len <= (dev_size - start)));
+	if (len & (logical_block_size_sectors - 1)) {
+		DMWARN("%s: len=%llu not aligned to h/w "
+		       "logical block size %u of %s",
+		       dm_device_name(ti->table->md),
+		       (unsigned long long)len,
+		       limits->logical_block_size, bdevname(bdev, b));
+		return 1;
+	}
+
+	return 0;
 }
 
 /*
- * This upgrades the mode on an already open dm_dev.  Being
+ * This upgrades the mode on an already open dm_dev, being
  * careful to leave things as they were if we fail to reopen the
- * device.
+ * device and not to touch the existing bdev field in case
+ * it is accessed concurrently inside dm_table_any_congested().
  */
-static int upgrade_mode(struct dm_dev *dd, int new_mode, struct mapped_device *md)
+static int upgrade_mode(struct dm_dev_internal *dd, fmode_t new_mode,
+			struct mapped_device *md)
 {
 	int r;
-	struct dm_dev dd_copy;
-	dev_t dev = dd->bdev->bd_dev;
+	struct dm_dev_internal dd_new, dd_old;
 
-	dd_copy = *dd;
+	dd_new = dd_old = *dd;
 
-	dd->mode |= new_mode;
-	dd->bdev = NULL;
-	r = open_dev(dd, dev, md);
-	if (!r)
-		close_dev(&dd_copy, md);
-	else
-		*dd = dd_copy;
+	dd_new.dm_dev.mode |= new_mode;
+	dd_new.dm_dev.bdev = NULL;
 
-	return r;
+	r = open_dev(&dd_new, dd->dm_dev.bdev->bd_dev, md);
+	if (r)
+		return r;
+
+	dd->dm_dev.mode |= new_mode;
+	close_dev(&dd_old, md);
+
+	return 0;
 }
 
 /*
  * Add a device to the list, or just increment the usage count if
  * it's already present.
  */
-static int __table_get_device(struct dm_table *t, struct dm_target *ti,
-			      const char *path, sector_t start, sector_t len,
-			      int mode, struct dm_dev **result)
+int dm_get_device(struct dm_target *ti, const char *path, fmode_t mode,
+		  struct dm_dev **result)
 {
 	int r;
 	dev_t uninitialized_var(dev);
-	struct dm_dev *dd;
+	struct dm_dev_internal *dd;
 	unsigned int major, minor;
+	struct dm_table *t = ti->table;
+	char dummy;
 
 	BUG_ON(!t);
 
-	if (sscanf(path, "%u:%u", &major, &minor) == 2) {
+	if (sscanf(path, "%u:%u%c", &major, &minor, &dummy) == 2) {
 		/* Extract the major/minor numbers */
 		dev = MKDEV(major, minor);
 		if (MAJOR(dev) != major || MINOR(dev) != minor)
 			return -EOVERFLOW;
 	} else {
 		/* convert the path to a device */
-		if ((r = lookup_device(path, &dev)))
-			return r;
+		struct block_device *bdev = lookup_bdev(path);
+
+		if (IS_ERR(bdev))
+			return PTR_ERR(bdev);
+		dev = bdev->bd_dev;
+		bdput(bdev);
 	}
 
 	dd = find_device(&t->devices, dev);
@@ -443,111 +440,81 @@ static int __table_get_device(struct dm_table *t, struct dm_target *ti,
 		if (!dd)
 			return -ENOMEM;
 
-		dd->mode = mode;
-		dd->bdev = NULL;
+		dd->dm_dev.mode = mode;
+		dd->dm_dev.bdev = NULL;
 
 		if ((r = open_dev(dd, dev, t->md))) {
 			kfree(dd);
 			return r;
 		}
 
-		format_dev_t(dd->name, dev);
+		format_dev_t(dd->dm_dev.name, dev);
 
 		atomic_set(&dd->count, 0);
 		list_add(&dd->list, &t->devices);
 
-	} else if (dd->mode != (mode | dd->mode)) {
+	} else if (dd->dm_dev.mode != (mode | dd->dm_dev.mode)) {
 		r = upgrade_mode(dd, mode, t->md);
 		if (r)
 			return r;
 	}
 	atomic_inc(&dd->count);
 
-	if (!check_device_area(dd, start, len)) {
-		DMWARN("device %s too small for target", path);
-		dm_put_device(ti, dd);
-		return -EINVAL;
-	}
-
-	*result = dd;
-
+	*result = &dd->dm_dev;
 	return 0;
 }
+EXPORT_SYMBOL(dm_get_device);
 
-void dm_set_device_limits(struct dm_target *ti, struct block_device *bdev)
+static int dm_set_device_limits(struct dm_target *ti, struct dm_dev *dev,
+				sector_t start, sector_t len, void *data)
 {
+	struct queue_limits *limits = data;
+	struct block_device *bdev = dev->bdev;
 	struct request_queue *q = bdev_get_queue(bdev);
-	struct io_restrictions *rs = &ti->limits;
+	char b[BDEVNAME_SIZE];
 
-	/*
-	 * Combine the device limits low.
-	 *
-	 * FIXME: if we move an io_restriction struct
-	 *        into q this would just be a call to
-	 *        combine_restrictions_low()
-	 */
-	rs->max_sectors =
-		min_not_zero(rs->max_sectors, q->max_sectors);
+	if (unlikely(!q)) {
+		DMWARN("%s: Cannot set limits for nonexistent device %s",
+		       dm_device_name(ti->table->md), bdevname(bdev, b));
+		return 0;
+	}
+
+	if (bdev_stack_limits(limits, bdev, start) < 0)
+		DMWARN("%s: adding target device %s caused an alignment inconsistency: "
+		       "physical_block_size=%u, logical_block_size=%u, "
+		       "alignment_offset=%u, start=%llu",
+		       dm_device_name(ti->table->md), bdevname(bdev, b),
+		       q->limits.physical_block_size,
+		       q->limits.logical_block_size,
+		       q->limits.alignment_offset,
+		       (unsigned long long) start << SECTOR_SHIFT);
 
 	/*
 	 * Check if merge fn is supported.
 	 * If not we'll force DM to use PAGE_SIZE or
 	 * smaller I/O, just to be safe.
 	 */
-
-	if (q->merge_bvec_fn && !ti->type->merge)
-		rs->max_sectors =
-			min_not_zero(rs->max_sectors,
-				     (unsigned int) (PAGE_SIZE >> 9));
-
-	rs->max_phys_segments =
-		min_not_zero(rs->max_phys_segments,
-			     q->max_phys_segments);
-
-	rs->max_hw_segments =
-		min_not_zero(rs->max_hw_segments, q->max_hw_segments);
-
-	rs->hardsect_size = max(rs->hardsect_size, q->hardsect_size);
-
-	rs->max_segment_size =
-		min_not_zero(rs->max_segment_size, q->max_segment_size);
-
-	rs->max_hw_sectors =
-		min_not_zero(rs->max_hw_sectors, q->max_hw_sectors);
-
-	rs->seg_boundary_mask =
-		min_not_zero(rs->seg_boundary_mask,
-			     q->seg_boundary_mask);
-
-	rs->bounce_pfn = min_not_zero(rs->bounce_pfn, q->bounce_pfn);
-
-	rs->no_cluster |= !test_bit(QUEUE_FLAG_CLUSTER, &q->queue_flags);
-}
-EXPORT_SYMBOL_GPL(dm_set_device_limits);
-
-int dm_get_device(struct dm_target *ti, const char *path, sector_t start,
-		  sector_t len, int mode, struct dm_dev **result)
-{
-	int r = __table_get_device(ti->table, ti, path,
-				   start, len, mode, result);
-
-	if (!r)
-		dm_set_device_limits(ti, (*result)->bdev);
-
-	return r;
+	if (dm_queue_merge_is_compulsory(q) && !ti->type->merge)
+		blk_limits_max_hw_sectors(limits,
+					  (unsigned int) (PAGE_SIZE >> 9));
+	return 0;
 }
 
 /*
- * Decrement a devices use count and remove it if necessary.
+ * Decrement a device's use count and remove it if necessary.
  */
-void dm_put_device(struct dm_target *ti, struct dm_dev *dd)
+void dm_put_device(struct dm_target *ti, struct dm_dev *d)
 {
+	struct dm_dev_internal *dd = container_of(d, struct dm_dev_internal,
+						  dm_dev);
+
 	if (atomic_dec_and_test(&dd->count)) {
 		close_dev(dd, ti->table->md);
 		list_del(&dd->list);
 		kfree(dd);
 	}
 }
+EXPORT_SYMBOL(dm_put_device);
 
 /*
  * Checks to see if the target joins onto the end of the table.
@@ -565,14 +532,28 @@ static int adjoin(struct dm_table *table, struct dm_target *ti)
 
 /*
  * Used to dynamically allocate the arg array.
+ *
+ * We do first allocation with GFP_NOIO because dm-mpath and dm-thin must
+ * process messages even if some device is suspended. These messages have a
+ * small fixed number of arguments.
+ *
+ * On the other hand, dm-switch needs to process bulk data using messages and
+ * excessive use of GFP_NOIO could cause trouble.
  */
 static char **realloc_argv(unsigned *array_size, char **old_argv)
 {
 	char **argv;
 	unsigned new_size;
+	gfp_t gfp;
 
-	new_size = *array_size ? *array_size * 2 : 64;
-	argv = kmalloc(new_size * sizeof(*argv), GFP_KERNEL);
+	if (*array_size) {
+		new_size = *array_size * 2;
+		gfp = GFP_KERNEL;
+	} else {
+		new_size = 8;
+		gfp = GFP_NOIO;
+	}
+	argv = kmalloc(new_size * sizeof(*argv), gfp);
 	if (argv) {
 		memcpy(argv, old_argv, *array_size * sizeof(*argv));
 		*array_size = new_size;
@@ -602,11 +583,8 @@ int dm_split_args(int *argc, char ***argvp, char *input)
 		return -ENOMEM;
 
 	while (1) {
-		start = end;
-
 		/* Skip whitespace */
-		while (*start && isspace(*start))
-			start++;
+		start = skip_spaces(end);
 
 		if (!*start)
 			break;	/* success, we hit the end */
@@ -648,24 +626,78 @@ int dm_split_args(int *argc, char ***argvp, char *input)
 	return 0;
 }
 
-static void check_for_valid_limits(struct io_restrictions *rs)
+/*
+ * Impose necessary and sufficient conditions on a devices's table such
+ * that any incoming bio which respects its logical_block_size can be
+ * processed successfully.  If it falls across the boundary between
+ * two or more targets, the size of each piece it gets split into must
+ * be compatible with the logical_block_size of the target processing it.
+ */
+static int validate_hardware_logical_block_alignment(struct dm_table *table,
+						 struct queue_limits *limits)
 {
-	if (!rs->max_sectors)
-		rs->max_sectors = SAFE_MAX_SECTORS;
-	if (!rs->max_hw_sectors)
-		rs->max_hw_sectors = SAFE_MAX_SECTORS;
-	if (!rs->max_phys_segments)
-		rs->max_phys_segments = MAX_PHYS_SEGMENTS;
-	if (!rs->max_hw_segments)
-		rs->max_hw_segments = MAX_HW_SEGMENTS;
-	if (!rs->hardsect_size)
-		rs->hardsect_size = 1 << SECTOR_SHIFT;
-	if (!rs->max_segment_size)
-		rs->max_segment_size = MAX_SEGMENT_SIZE;
-	if (!rs->seg_boundary_mask)
-		rs->seg_boundary_mask = -1;
-	if (!rs->bounce_pfn)
-		rs->bounce_pfn = -1;
+	/*
+	 * This function uses arithmetic modulo the logical_block_size
+	 * (in units of 512-byte sectors).
+	 */
+	unsigned short device_logical_block_size_sects =
+		limits->logical_block_size >> SECTOR_SHIFT;
+
+	/*
+	 * Offset of the start of the next table entry, mod logical_block_size.
+	 */
+	unsigned short next_target_start = 0;
+
+	/*
+	 * Given an aligned bio that extends beyond the end of a
+	 * target, how many sectors must the next target handle?
+	 */
+	unsigned short remaining = 0;
+
+	struct dm_target *uninitialized_var(ti);
+	struct queue_limits ti_limits;
+	unsigned i = 0;
+
+	/*
+	 * Check each entry in the table in turn.
+	 */
+	while (i < dm_table_get_num_targets(table)) {
+		ti = dm_table_get_target(table, i++);
+
+		blk_set_stacking_limits(&ti_limits);
+
+		/* combine all target devices' limits */
+		if (ti->type->iterate_devices)
+			ti->type->iterate_devices(ti, dm_set_device_limits,
+						  &ti_limits);
+
+		/*
+		 * If the remaining sectors fall entirely within this
+		 * table entry are they compatible with its logical_block_size?
+		 */
+		if (remaining < ti->len &&
+		    remaining & ((ti_limits.logical_block_size >>
+				  SECTOR_SHIFT) - 1))
+			break;	/* Error */
+
+		next_target_start =
+		    (unsigned short) ((next_target_start + ti->len) &
+				      (device_logical_block_size_sects - 1));
+		remaining = next_target_start ?
+		    device_logical_block_size_sects - next_target_start : 0;
+	}
+
+	if (remaining) {
+		DMWARN("%s: table line %u (start sect %llu len %llu) "
+		       "not aligned to h/w logical block size %u",
+		       dm_device_name(table->md), i,
+		       (unsigned long long) ti->begin,
+		       (unsigned long long) ti->len,
+		       limits->logical_block_size);
+		return -EINVAL;
+	}
+
+	return 0;
 }
 
 int dm_table_add_target(struct dm_table *t, const char *type,
@@ -675,8 +707,13 @@ int dm_table_add_target(struct dm_table *t, const char *type,
 	char **argv;
 	struct dm_target *tgt;
 
-	if ((r = check_space(t)))
-		return r;
+	if (t->singleton) {
+		DMERR("%s: target type %s must appear alone in table",
+		      dm_device_name(t->md), t->targets->type->name);
+		return -EINVAL;
+	}
+
+	BUG_ON(t->num_targets >= t->num_allocated);
 
 	tgt = t->targets + t->num_targets;
 	memset(tgt, 0, sizeof(*tgt));
@@ -693,6 +730,36 @@ int dm_table_add_target(struct dm_table *t, const char *type,
 		return -EINVAL;
 	}
 
+	if (dm_target_needs_singleton(tgt->type)) {
+		if (t->num_targets) {
+			DMERR("%s: target type %s must appear alone in table",
+			      dm_device_name(t->md), type);
+			return -EINVAL;
+		}
+		t->singleton = 1;
+	}
+
+	if (dm_target_always_writeable(tgt->type) && !(t->mode & FMODE_WRITE)) {
+		DMERR("%s: target type %s may not be included in read-only tables",
+		      dm_device_name(t->md), type);
+		return -EINVAL;
+	}
+
+	if (t->immutable_target_type) {
+		if (t->immutable_target_type != tgt->type) {
+			DMERR("%s: immutable target type %s cannot be mixed with other target types",
+			      dm_device_name(t->md), t->immutable_target_type->name);
+			return -EINVAL;
+		}
+	} else if (dm_target_is_immutable(tgt->type)) {
+		if (t->num_targets) {
+			DMERR("%s: immutable target type %s cannot be mixed with other target types",
+			      dm_device_name(t->md), tgt->type->name);
+			return -EINVAL;
+		}
+		t->immutable_target_type = tgt->type;
+	}
+
 	tgt->table = t;
 	tgt->begin = start;
 	tgt->len = len;
@@ -720,9 +787,10 @@ int dm_table_add_target(struct dm_table *t, const char *type,
 
 	t->highs[t->num_targets++] = tgt->begin + tgt->len - 1;
 
-	/* FIXME: the plan is to combine high here and then have
-	 * the merge fn apply the target level restrictions. */
-	combine_restrictions_low(&t->limits, &tgt->limits);
+	if (!tgt->num_discard_bios && tgt->discards_supported)
+		DMWARN("%s: %s: ignoring discards_supported because num_discard_bios is zero.",
+		       dm_device_name(t->md), type);
+
 	return 0;
 
  bad:
@@ -731,6 +799,187 @@ int dm_table_add_target(struct dm_table *t, const char *type,
 	return r;
 }
 
+/*
+ * Target argument parsing helpers.
+ */
+static int validate_next_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
+			     unsigned *value, char **error, unsigned grouped)
+{
+	const char *arg_str = dm_shift_arg(arg_set);
+	char dummy;
+
+	if (!arg_str ||
+	    (sscanf(arg_str, "%u%c", value, &dummy) != 1) ||
+	    (*value < arg->min) ||
+	    (*value > arg->max) ||
+	    (grouped && arg_set->argc < *value)) {
+		*error = arg->error;
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+int dm_read_arg(struct dm_arg *arg, struct dm_arg_set *arg_set,
+		unsigned *value, char **error)
+{
+	return validate_next_arg(arg, arg_set, value, error, 0);
+}
+EXPORT_SYMBOL(dm_read_arg);
+
+int dm_read_arg_group(struct dm_arg *arg, struct dm_arg_set *arg_set,
+		      unsigned *value, char **error)
+{
+	return validate_next_arg(arg, arg_set, value, error, 1);
+}
+EXPORT_SYMBOL(dm_read_arg_group);
+
+const char *dm_shift_arg(struct dm_arg_set *as)
+{
+	char *r;
+
+	if (as->argc) {
+		as->argc--;
+		r = *as->argv;
+		as->argv++;
+		return r;
+	}
+
+	return NULL;
+}
+EXPORT_SYMBOL(dm_shift_arg);
+
+void dm_consume_args(struct dm_arg_set *as, unsigned num_args)
+{
+	BUG_ON(as->argc < num_args);
+	as->argc -= num_args;
+	as->argv += num_args;
+}
+EXPORT_SYMBOL(dm_consume_args);
+
+static int dm_table_set_type(struct dm_table *t)
+{
+	unsigned i;
+	unsigned bio_based = 0, request_based = 0, hybrid = 0;
+	struct dm_target *tgt;
+	struct dm_dev_internal *dd;
+	struct list_head *devices;
+	unsigned live_md_type;
+
+	for (i = 0; i < t->num_targets; i++) {
+		tgt = t->targets + i;
+		if (dm_target_hybrid(tgt))
+			hybrid = 1;
+		else if (dm_target_request_based(tgt))
+			request_based = 1;
+		else
+			bio_based = 1;
+
+		if (bio_based && request_based) {
+			DMWARN("Inconsistent table: different target types"
+			       " can't be mixed up");
+			return -EINVAL;
+		}
+	}
+
+	if (hybrid && !bio_based && !request_based) {
+		/*
+		 * The targets can work either way.
+		 * Determine the type from the live device.
+		 * Default to bio-based if device is new.
+		 */
+		live_md_type = dm_get_md_type(t->md);
+		if (live_md_type == DM_TYPE_REQUEST_BASED)
+			request_based = 1;
+		else
+			bio_based = 1;
+	}
+
+	if (bio_based) {
+		/* We must use this table as bio-based */
+		t->type = DM_TYPE_BIO_BASED;
+		return 0;
+	}
+
+	BUG_ON(!request_based); /* No targets in this table */
+
+	/* Non-request-stackable devices can't be used for request-based dm */
+	devices = dm_table_get_devices(t);
+	list_for_each_entry(dd, devices, list) {
+		if (!blk_queue_stackable(bdev_get_queue(dd->dm_dev.bdev))) {
+			DMWARN("table load rejected: including"
+			       " non-request-stackable devices");
+			return -EINVAL;
+		}
+	}
+
+	/*
+	 * Request-based dm supports only tables that have a single target now.
+	 * To support multiple targets, request splitting support is needed,
+	 * and that needs lots of changes in the block-layer.
+	 * (e.g. request completion process for partial completion.)
+	 */
+	if (t->num_targets > 1) {
+		DMWARN("Request-based dm doesn't support multiple targets yet");
+		return -EINVAL;
+	}
+
+	t->type = DM_TYPE_REQUEST_BASED;
+
+	return 0;
+}
+
+unsigned dm_table_get_type(struct dm_table *t)
+{
+	return t->type;
+}
+
+struct target_type *dm_table_get_immutable_target_type(struct dm_table *t)
+{
+	return t->immutable_target_type;
+}
+
+bool dm_table_request_based(struct dm_table *t)
+{
+	return dm_table_get_type(t) == DM_TYPE_REQUEST_BASED;
+}
+
+static int dm_table_alloc_md_mempools(struct dm_table *t)
+{
+	unsigned type = dm_table_get_type(t);
+	unsigned per_bio_data_size = 0;
+	struct dm_target *tgt;
+	unsigned i;
+
+	if (unlikely(type == DM_TYPE_NONE)) {
+		DMWARN("no table type is set, can't allocate mempools");
+		return -EINVAL;
+	}
+
+	if (type == DM_TYPE_BIO_BASED)
+		for (i = 0; i < t->num_targets; i++) {
+			tgt = t->targets + i;
+			per_bio_data_size = max(per_bio_data_size, tgt->per_bio_data_size);
+		}
+
+	t->mempools = dm_alloc_md_mempools(type, t->integrity_supported, per_bio_data_size);
+	if (!t->mempools)
+		return -ENOMEM;
+
+	return 0;
+}
+
+void dm_table_free_md_mempools(struct dm_table *t)
+{
+	dm_free_md_mempools(t->mempools);
+	t->mempools = NULL;
+}
+
+struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t)
+{
+	return t->mempools;
+}
+
 static int setup_indexes(struct dm_table *t)
 {
 	int i;
@@ -760,13 +1009,11 @@ static int setup_indexes(struct dm_table *t)
 /*
  * Builds the btree to index the map.
  */
-int dm_table_complete(struct dm_table *t)
+static int dm_table_build_index(struct dm_table *t)
 {
 	int r = 0;
 	unsigned int leaf_nodes;
 
-	check_for_valid_limits(&t->limits);
-
 	/* how many indexes will the btree have ? */
 	leaf_nodes = dm_div_up(t->num_targets, KEYS_PER_NODE);
 	t->depth = 1 + int_log(leaf_nodes, CHILDREN_PER_NODE);
@@ -781,6 +1028,117 @@ int dm_table_complete(struct dm_table *t)
 	return r;
 }
 
+/*
+ * Get a disk whose integrity profile reflects the table's profile.
+ * If %match_all is true, all devices' profiles must match.
+ * If %match_all is false, all devices must at least have an
+ * allocated integrity profile; but uninitialized is ok.
+ * Returns NULL if integrity support was inconsistent or unavailable.
+ */
+static struct gendisk * dm_table_get_integrity_disk(struct dm_table *t,
+						    bool match_all)
+{
+	struct list_head *devices = dm_table_get_devices(t);
+	struct dm_dev_internal *dd = NULL;
+	struct gendisk *prev_disk = NULL, *template_disk = NULL;
+
+	list_for_each_entry(dd, devices, list) {
+		template_disk = dd->dm_dev.bdev->bd_disk;
+		if (!blk_get_integrity(template_disk))
+			goto no_integrity;
+		if (!match_all && !blk_integrity_is_initialized(template_disk))
+			continue; /* skip uninitialized profiles */
+		else if (prev_disk &&
+			 blk_integrity_compare(prev_disk, template_disk) < 0)
+			goto no_integrity;
+		prev_disk = template_disk;
+	}
+
+	return template_disk;
+
+no_integrity:
+	if (prev_disk)
+		DMWARN("%s: integrity not set: %s and %s profile mismatch",
+		       dm_device_name(t->md),
+		       prev_disk->disk_name,
+		       template_disk->disk_name);
+	return NULL;
+}
+
+/*
+ * Register the mapped device for blk_integrity support if
+ * the underlying devices have an integrity profile.  But all devices
+ * may not have matching profiles (checking all devices isn't reliable
+ * during table load because this table may use other DM device(s) which
+ * must be resumed before they will have an initialized integity profile).
+ * Stacked DM devices force a 2 stage integrity profile validation:
+ * 1 - during load, validate all initialized integrity profiles match
+ * 2 - during resume, validate all integrity profiles match
+ */
+static int dm_table_prealloc_integrity(struct dm_table *t, struct mapped_device *md)
+{
+	struct gendisk *template_disk = NULL;
+
+	template_disk = dm_table_get_integrity_disk(t, false);
+	if (!template_disk)
+		return 0;
+
+	if (!blk_integrity_is_initialized(dm_disk(md))) {
+		t->integrity_supported = 1;
+		return blk_integrity_register(dm_disk(md), NULL);
+	}
+
+	/*
+	 * If DM device already has an initalized integrity
+	 * profile the new profile should not conflict.
+	 */
+	if (blk_integrity_is_initialized(template_disk) &&
+	    blk_integrity_compare(dm_disk(md), template_disk) < 0) {
+		DMWARN("%s: conflict with existing integrity profile: "
+		       "%s profile mismatch",
+		       dm_device_name(t->md),
+		       template_disk->disk_name);
+		return 1;
+	}
+
+	/* Preserve existing initialized integrity profile */
+	t->integrity_supported = 1;
+	return 0;
+}
+
+/*
+ * Prepares the table for use by building the indices,
+ * setting the type, and allocating mempools.
+ */
+int dm_table_complete(struct dm_table *t)
+{
+	int r;
+
+	r = dm_table_set_type(t);
+	if (r) {
+		DMERR("unable to set table type");
+		return r;
+	}
+
+	r = dm_table_build_index(t);
+	if (r) {
+		DMERR("unable to build btrees");
+		return r;
+	}
+
+	r = dm_table_prealloc_integrity(t, t->md);
+	if (r) {
+		DMERR("could not register integrity profile.");
+		return r;
+	}
+
+	r = dm_table_alloc_md_mempools(t);
+	if (r)
+		DMERR("unable to allocate mempools");
+
+	return r;
+}
+
 static DEFINE_MUTEX(_event_lock);
 void dm_table_event_callback(struct dm_table *t,
 			     void (*fn)(void *), void *context)
@@ -804,11 +1162,13 @@ void dm_table_event(struct dm_table *t)
 		t->event_fn(t->event_context);
 	mutex_unlock(&_event_lock);
 }
+EXPORT_SYMBOL(dm_table_event);
 
 sector_t dm_table_get_size(struct dm_table *t)
 {
 	return t->num_targets ? (t->highs[t->num_targets - 1] + 1) : 0;
 }
+EXPORT_SYMBOL(dm_table_get_size);
 
 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index)
 {
@@ -841,26 +1201,292 @@ struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector)
 	return &t->targets[(KEYS_PER_NODE * n) + k];
 }
 
-void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q)
+static int count_device(struct dm_target *ti, struct dm_dev *dev,
+			sector_t start, sector_t len, void *data)
 {
+	unsigned *num_devices = data;
+
+	(*num_devices)++;
+
+	return 0;
+}
+
+/*
+ * Check whether a table has no data devices attached using each
+ * target's iterate_devices method.
+ * Returns false if the result is unknown because a target doesn't
+ * support iterate_devices.
+ */
+bool dm_table_has_no_data_devices(struct dm_table *table)
+{
+	struct dm_target *uninitialized_var(ti);
+	unsigned i = 0, num_devices = 0;
+
+	while (i < dm_table_get_num_targets(table)) {
+		ti = dm_table_get_target(table, i++);
+
+		if (!ti->type->iterate_devices)
+			return false;
+
+		ti->type->iterate_devices(ti, count_device, &num_devices);
+		if (num_devices)
+			return false;
+	}
+
+	return true;
+}
+
+/*
+ * Establish the new table's queue_limits and validate them.
+ */
+int dm_calculate_queue_limits(struct dm_table *table,
+			      struct queue_limits *limits)
+{
+	struct dm_target *uninitialized_var(ti);
+	struct queue_limits ti_limits;
+	unsigned i = 0;
+
+	blk_set_stacking_limits(limits);
+
+	while (i < dm_table_get_num_targets(table)) {
+		blk_set_stacking_limits(&ti_limits);
+
+		ti = dm_table_get_target(table, i++);
+
+		if (!ti->type->iterate_devices)
+			goto combine_limits;
+
+		/*
+		 * Combine queue limits of all the devices this target uses.
+		 */
+		ti->type->iterate_devices(ti, dm_set_device_limits,
+					  &ti_limits);
+
+		/* Set I/O hints portion of queue limits */
+		if (ti->type->io_hints)
+			ti->type->io_hints(ti, &ti_limits);
+
+		/*
+		 * Check each device area is consistent with the target's
+		 * overall queue limits.
+		 */
+		if (ti->type->iterate_devices(ti, device_area_is_invalid,
+					      &ti_limits))
+			return -EINVAL;
+
+combine_limits:
+		/*
+		 * Merge this target's queue limits into the overall limits
+		 * for the table.
+		 */
+		if (blk_stack_limits(limits, &ti_limits, 0) < 0)
+			DMWARN("%s: adding target device "
+			       "(start sect %llu len %llu) "
+			       "caused an alignment inconsistency",
+			       dm_device_name(table->md),
+			       (unsigned long long) ti->begin,
+			       (unsigned long long) ti->len);
+	}
+
+	return validate_hardware_logical_block_alignment(table, limits);
+}
+
+/*
+ * Set the integrity profile for this device if all devices used have
+ * matching profiles.  We're quite deep in the resume path but still
+ * don't know if all devices (particularly DM devices this device
+ * may be stacked on) have matching profiles.  Even if the profiles
+ * don't match we have no way to fail (to resume) at this point.
+ */
+static void dm_table_set_integrity(struct dm_table *t)
+{
+	struct gendisk *template_disk = NULL;
+
+	if (!blk_get_integrity(dm_disk(t->md)))
+		return;
+
+	template_disk = dm_table_get_integrity_disk(t, true);
+	if (template_disk)
+		blk_integrity_register(dm_disk(t->md),
+				       blk_get_integrity(template_disk));
+	else if (blk_integrity_is_initialized(dm_disk(t->md)))
+		DMWARN("%s: device no longer has a valid integrity profile",
+		       dm_device_name(t->md));
+	else
+		DMWARN("%s: unable to establish an integrity profile",
+		       dm_device_name(t->md));
+}
+
+static int device_flush_capable(struct dm_target *ti, struct dm_dev *dev,
+				sector_t start, sector_t len, void *data)
+{
+	unsigned flush = (*(unsigned *)data);
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && (q->flush_flags & flush);
+}
+
+static bool dm_table_supports_flush(struct dm_table *t, unsigned flush)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
 	/*
-	 * Make sure we obey the optimistic sub devices
-	 * restrictions.
+	 * Require at least one underlying device to support flushes.
+	 * t->devices includes internal dm devices such as mirror logs
+	 * so we need to use iterate_devices here, which targets
+	 * supporting flushes must provide.
 	 */
-	blk_queue_max_sectors(q, t->limits.max_sectors);
-	q->max_phys_segments = t->limits.max_phys_segments;
-	q->max_hw_segments = t->limits.max_hw_segments;
-	q->hardsect_size = t->limits.hardsect_size;
-	q->max_segment_size = t->limits.max_segment_size;
-	q->max_hw_sectors = t->limits.max_hw_sectors;
-	q->seg_boundary_mask = t->limits.seg_boundary_mask;
-	q->bounce_pfn = t->limits.bounce_pfn;
-
-	if (t->limits.no_cluster)
-		queue_flag_clear_unlocked(QUEUE_FLAG_CLUSTER, q);
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (!ti->num_flush_bios)
+			continue;
+
+		if (ti->flush_supported)
+			return 1;
+
+		if (ti->type->iterate_devices &&
+		    ti->type->iterate_devices(ti, device_flush_capable, &flush))
+			return 1;
+	}
+
+	return 0;
+}
+
+static bool dm_table_discard_zeroes_data(struct dm_table *t)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
+	/* Ensure that all targets supports discard_zeroes_data. */
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (ti->discard_zeroes_data_unsupported)
+			return 0;
+	}
+
+	return 1;
+}
+
+static int device_is_nonrot(struct dm_target *ti, struct dm_dev *dev,
+			    sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && blk_queue_nonrot(q);
+}
+
+static int device_is_not_random(struct dm_target *ti, struct dm_dev *dev,
+			     sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && !blk_queue_add_random(q);
+}
+
+static bool dm_table_all_devices_attribute(struct dm_table *t,
+					   iterate_devices_callout_fn func)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (!ti->type->iterate_devices ||
+		    !ti->type->iterate_devices(ti, func, NULL))
+			return 0;
+	}
+
+	return 1;
+}
+
+static int device_not_write_same_capable(struct dm_target *ti, struct dm_dev *dev,
+					 sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && !q->limits.max_write_same_sectors;
+}
+
+static bool dm_table_supports_write_same(struct dm_table *t)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (!ti->num_write_same_bios)
+			return false;
+
+		if (!ti->type->iterate_devices ||
+		    ti->type->iterate_devices(ti, device_not_write_same_capable, NULL))
+			return false;
+	}
+
+	return true;
+}
+
+void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
+			       struct queue_limits *limits)
+{
+	unsigned flush = 0;
+
+	/*
+	 * Copy table's limits to the DM device's request_queue
+	 */
+	q->limits = *limits;
+
+	if (!dm_table_supports_discards(t))
+		queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, q);
 	else
-		queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, q);
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, q);
+
+	if (dm_table_supports_flush(t, REQ_FLUSH)) {
+		flush |= REQ_FLUSH;
+		if (dm_table_supports_flush(t, REQ_FUA))
+			flush |= REQ_FUA;
+	}
+	blk_queue_flush(q, flush);
+
+	if (!dm_table_discard_zeroes_data(t))
+		q->limits.discard_zeroes_data = 0;
 
+	/* Ensure that all underlying devices are non-rotational. */
+	if (dm_table_all_devices_attribute(t, device_is_nonrot))
+		queue_flag_set_unlocked(QUEUE_FLAG_NONROT, q);
+	else
+		queue_flag_clear_unlocked(QUEUE_FLAG_NONROT, q);
+
+	if (!dm_table_supports_write_same(t))
+		q->limits.max_write_same_sectors = 0;
+
+	dm_table_set_integrity(t);
+
+	/*
+	 * Determine whether or not this queue's I/O timings contribute
+	 * to the entropy pool, Only request-based targets use this.
+	 * Clear QUEUE_FLAG_ADD_RANDOM if any underlying device does not
+	 * have it set.
+	 */
+	if (blk_queue_add_random(q) && dm_table_all_devices_attribute(t, device_is_not_random))
+		queue_flag_clear_unlocked(QUEUE_FLAG_ADD_RANDOM, q);
+
+	/*
+	 * QUEUE_FLAG_STACKABLE must be set after all queue settings are
+	 * visible to other CPUs because, once the flag is set, incoming bios
+	 * are processed by request-based dm, which refers to the queue
+	 * settings.
+	 * Until the flag set, bios are passed to bio-based dm and queued to
+	 * md->deferred where queue settings are not needed yet.
+	 * Those bios are passed to request-based dm at the resume time.
+	 */
+	smp_mb();
+	if (dm_table_request_based(t))
+		queue_flag_set_unlocked(QUEUE_FLAG_STACKABLE, q);
 }
 
 unsigned int dm_table_get_num_targets(struct dm_table *t)
@@ -873,10 +1499,11 @@ struct list_head *dm_table_get_devices(struct dm_table *t)
 	return &t->devices;
 }
 
-int dm_table_get_mode(struct dm_table *t)
+fmode_t dm_table_get_mode(struct dm_table *t)
 {
 	return t->mode;
 }
+EXPORT_SYMBOL(dm_table_get_mode);
 
 static void suspend_targets(struct dm_table *t, unsigned postsuspend)
 {
@@ -921,8 +1548,11 @@ int dm_table_resume_targets(struct dm_table *t)
 			continue;
 
 		r = ti->type->preresume(ti);
-		if (r)
+		if (r) {
+			DMERR("%s: %s: preresume failed, error = %d",
+			      dm_device_name(t->md), ti->type->name, r);
 			return r;
+		}
 	}
 
 	for (i = 0; i < t->num_targets; i++) {
@@ -935,46 +1565,110 @@ int dm_table_resume_targets(struct dm_table *t)
 	return 0;
 }
 
+void dm_table_add_target_callbacks(struct dm_table *t, struct dm_target_callbacks *cb)
+{
+	list_add(&cb->list, &t->target_callbacks);
+}
+EXPORT_SYMBOL_GPL(dm_table_add_target_callbacks);
+
 int dm_table_any_congested(struct dm_table *t, int bdi_bits)
 {
-	struct dm_dev *dd;
+	struct dm_dev_internal *dd;
 	struct list_head *devices = dm_table_get_devices(t);
+	struct dm_target_callbacks *cb;
 	int r = 0;
 
 	list_for_each_entry(dd, devices, list) {
-		struct request_queue *q = bdev_get_queue(dd->bdev);
-		r |= bdi_congested(&q->backing_dev_info, bdi_bits);
+		struct request_queue *q = bdev_get_queue(dd->dm_dev.bdev);
+		char b[BDEVNAME_SIZE];
+
+		if (likely(q))
+			r |= bdi_congested(&q->backing_dev_info, bdi_bits);
+		else
+			DMWARN_LIMIT("%s: any_congested: nonexistent device %s",
+				     dm_device_name(t->md),
+				     bdevname(dd->dm_dev.bdev, b));
 	}
 
+	list_for_each_entry(cb, &t->target_callbacks, list)
+		if (cb->congested_fn)
+			r |= cb->congested_fn(cb, bdi_bits);
+
 	return r;
 }
 
-void dm_table_unplug_all(struct dm_table *t)
+int dm_table_any_busy_target(struct dm_table *t)
 {
-	struct dm_dev *dd;
-	struct list_head *devices = dm_table_get_devices(t);
-
-	list_for_each_entry(dd, devices, list) {
-		struct request_queue *q = bdev_get_queue(dd->bdev);
+	unsigned i;
+	struct dm_target *ti;
 
-		blk_unplug(q);
+	for (i = 0; i < t->num_targets; i++) {
+		ti = t->targets + i;
+		if (ti->type->busy && ti->type->busy(ti))
+			return 1;
 	}
+
+	return 0;
 }
 
 struct mapped_device *dm_table_get_md(struct dm_table *t)
 {
-	dm_get(t->md);
-
 	return t->md;
 }
-
-EXPORT_SYMBOL(dm_vcalloc);
-EXPORT_SYMBOL(dm_get_device);
-EXPORT_SYMBOL(dm_put_device);
-EXPORT_SYMBOL(dm_table_event);
-EXPORT_SYMBOL(dm_table_get_size);
-EXPORT_SYMBOL(dm_table_get_mode);
 EXPORT_SYMBOL(dm_table_get_md);
-EXPORT_SYMBOL(dm_table_put);
-EXPORT_SYMBOL(dm_table_get);
-EXPORT_SYMBOL(dm_table_unplug_all);
+
+void dm_table_run_md_queue_async(struct dm_table *t)
+{
+	struct mapped_device *md;
+	struct request_queue *queue;
+	unsigned long flags;
+
+	if (!dm_table_request_based(t))
+		return;
+
+	md = dm_table_get_md(t);
+	queue = dm_get_md_queue(md);
+	if (queue) {
+		spin_lock_irqsave(queue->queue_lock, flags);
+		blk_run_queue_async(queue);
+		spin_unlock_irqrestore(queue->queue_lock, flags);
+	}
+}
+EXPORT_SYMBOL(dm_table_run_md_queue_async);
+
+static int device_discard_capable(struct dm_target *ti, struct dm_dev *dev,
+				  sector_t start, sector_t len, void *data)
+{
+	struct request_queue *q = bdev_get_queue(dev->bdev);
+
+	return q && blk_queue_discard(q);
+}
+
+bool dm_table_supports_discards(struct dm_table *t)
+{
+	struct dm_target *ti;
+	unsigned i = 0;
+
+	/*
+	 * Unless any target used by the table set discards_supported,
+	 * require at least one underlying device to support discards.
+	 * t->devices includes internal dm devices such as mirror logs
+	 * so we need to use iterate_devices here, which targets
+	 * supporting discard selectively must provide.
+	 */
+	while (i < dm_table_get_num_targets(t)) {
+		ti = dm_table_get_target(t, i++);
+
+		if (!ti->num_discard_bios)
+			continue;
+
+		if (ti->discards_supported)
+			return 1;
+
+		if (ti->type->iterate_devices &&
+		    ti->type->iterate_devices(ti, device_discard_capable, NULL))
+			return 1;
+	}
+
+	return 0;
+}
diff --git a/drivers/md/dm-target.c b/drivers/md/dm-target.c
index 835cf95b857..242e3cec397 100644
--- a/drivers/md/dm-target.c
+++ b/drivers/md/dm-target.c
@@ -10,49 +10,37 @@
 #include <linux/init.h>
 #include <linux/kmod.h>
 #include <linux/bio.h>
-#include <linux/slab.h>
 
 #define DM_MSG_PREFIX "target"
 
-struct tt_internal {
-	struct target_type tt;
-
-	struct list_head list;
-	long use;
-};
-
 static LIST_HEAD(_targets);
 static DECLARE_RWSEM(_lock);
 
 #define DM_MOD_NAME_SIZE 32
 
-static inline struct tt_internal *__find_target_type(const char *name)
+static inline struct target_type *__find_target_type(const char *name)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
-	list_for_each_entry (ti, &_targets, list)
-		if (!strcmp(name, ti->tt.name))
-			return ti;
+	list_for_each_entry(tt, &_targets, list)
+		if (!strcmp(name, tt->name))
+			return tt;
 
 	return NULL;
 }
 
-static struct tt_internal *get_target_type(const char *name)
+static struct target_type *get_target_type(const char *name)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
 	down_read(&_lock);
 
-	ti = __find_target_type(name);
-	if (ti) {
-		if ((ti->use == 0) && !try_module_get(ti->tt.module))
-			ti = NULL;
-		else
-			ti->use++;
-	}
+	tt = __find_target_type(name);
+	if (tt && !try_module_get(tt->module))
+		tt = NULL;
 
 	up_read(&_lock);
-	return ti;
+	return tt;
 }
 
 static void load_module(const char *name)
@@ -62,122 +50,100 @@ static void load_module(const char *name)
 
 struct target_type *dm_get_target_type(const char *name)
 {
-	struct tt_internal *ti = get_target_type(name);
+	struct target_type *tt = get_target_type(name);
 
-	if (!ti) {
+	if (!tt) {
 		load_module(name);
-		ti = get_target_type(name);
+		tt = get_target_type(name);
 	}
 
-	return ti ? &ti->tt : NULL;
+	return tt;
 }
 
-void dm_put_target_type(struct target_type *t)
+void dm_put_target_type(struct target_type *tt)
 {
-	struct tt_internal *ti = (struct tt_internal *) t;
-
 	down_read(&_lock);
-	if (--ti->use == 0)
-		module_put(ti->tt.module);
-
-	BUG_ON(ti->use < 0);
+	module_put(tt->module);
 	up_read(&_lock);
-
-	return;
-}
-
-static struct tt_internal *alloc_target(struct target_type *t)
-{
-	struct tt_internal *ti = kzalloc(sizeof(*ti), GFP_KERNEL);
-
-	if (ti)
-		ti->tt = *t;
-
-	return ti;
 }
 
-
 int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 					void *param), void *param)
 {
-	struct tt_internal *ti;
+	struct target_type *tt;
 
 	down_read(&_lock);
-	list_for_each_entry (ti, &_targets, list)
-		iter_func(&ti->tt, param);
+	list_for_each_entry(tt, &_targets, list)
+		iter_func(tt, param);
 	up_read(&_lock);
 
 	return 0;
 }
 
-int dm_register_target(struct target_type *t)
+int dm_register_target(struct target_type *tt)
 {
 	int rv = 0;
-	struct tt_internal *ti = alloc_target(t);
-
-	if (!ti)
-		return -ENOMEM;
 
 	down_write(&_lock);
-	if (__find_target_type(t->name))
+	if (__find_target_type(tt->name))
 		rv = -EEXIST;
 	else
-		list_add(&ti->list, &_targets);
+		list_add(&tt->list, &_targets);
 
 	up_write(&_lock);
-	if (rv)
-		kfree(ti);
 	return rv;
 }
 
-int dm_unregister_target(struct target_type *t)
+void dm_unregister_target(struct target_type *tt)
 {
-	struct tt_internal *ti;
-
 	down_write(&_lock);
-	if (!(ti = __find_target_type(t->name))) {
-		up_write(&_lock);
-		return -EINVAL;
+	if (!__find_target_type(tt->name)) {
+		DMCRIT("Unregistering unrecognised target: %s", tt->name);
+		BUG();
 	}
 
-	if (ti->use) {
-		up_write(&_lock);
-		return -ETXTBSY;
-	}
-
-	list_del(&ti->list);
-	kfree(ti);
+	list_del(&tt->list);
 
 	up_write(&_lock);
-	return 0;
 }
 
 /*
  * io-err: always fails an io, useful for bringing
  * up LVs that have holes in them.
  */
-static int io_err_ctr(struct dm_target *ti, unsigned int argc, char **args)
+static int io_err_ctr(struct dm_target *tt, unsigned int argc, char **args)
 {
+	/*
+	 * Return error for discards instead of -EOPNOTSUPP
+	 */
+	tt->num_discard_bios = 1;
+
 	return 0;
 }
 
-static void io_err_dtr(struct dm_target *ti)
+static void io_err_dtr(struct dm_target *tt)
 {
 	/* empty */
 }
 
-static int io_err_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static int io_err_map(struct dm_target *tt, struct bio *bio)
+{
+	return -EIO;
+}
+
+static int io_err_map_rq(struct dm_target *ti, struct request *clone,
+			 union map_info *map_context)
 {
 	return -EIO;
 }
 
 static struct target_type error_target = {
 	.name = "error",
-	.version = {1, 0, 1},
+	.version = {1, 2, 0},
 	.ctr  = io_err_ctr,
 	.dtr  = io_err_dtr,
 	.map  = io_err_map,
+	.map_rq = io_err_map_rq,
 };
 
 int __init dm_target_init(void)
@@ -187,8 +153,7 @@ int __init dm_target_init(void)
 
 void dm_target_exit(void)
 {
-	if (dm_unregister_target(&error_target))
-		DMWARN("error target unregistration failed");
+	dm_unregister_target(&error_target);
 }
 
 EXPORT_SYMBOL(dm_register_target);
diff --git a/drivers/md/dm-thin-metadata.c b/drivers/md/dm-thin-metadata.c
new file mode 100644
index 00000000000..e9d33ad59df
--- /dev/null
+++ b/drivers/md/dm-thin-metadata.c
@@ -0,0 +1,1815 @@
+/*
+ * Copyright (C) 2011-2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-thin-metadata.h"
+#include "persistent-data/dm-btree.h"
+#include "persistent-data/dm-space-map.h"
+#include "persistent-data/dm-space-map-disk.h"
+#include "persistent-data/dm-transaction-manager.h"
+
+#include <linux/list.h>
+#include <linux/device-mapper.h>
+#include <linux/workqueue.h>
+
+/*--------------------------------------------------------------------------
+ * As far as the metadata goes, there is:
+ *
+ * - A superblock in block zero, taking up fewer than 512 bytes for
+ *   atomic writes.
+ *
+ * - A space map managing the metadata blocks.
+ *
+ * - A space map managing the data blocks.
+ *
+ * - A btree mapping our internal thin dev ids onto struct disk_device_details.
+ *
+ * - A hierarchical btree, with 2 levels which effectively maps (thin
+ *   dev id, virtual block) -> block_time.  Block time is a 64-bit
+ *   field holding the time in the low 24 bits, and block in the top 48
+ *   bits.
+ *
+ * BTrees consist solely of btree_nodes, that fill a block.  Some are
+ * internal nodes, as such their values are a __le64 pointing to other
+ * nodes.  Leaf nodes can store data of any reasonable size (ie. much
+ * smaller than the block size).  The nodes consist of the header,
+ * followed by an array of keys, followed by an array of values.  We have
+ * to binary search on the keys so they're all held together to help the
+ * cpu cache.
+ *
+ * Space maps have 2 btrees:
+ *
+ * - One maps a uint64_t onto a struct index_entry.  Which points to a
+ *   bitmap block, and has some details about how many free entries there
+ *   are etc.
+ *
+ * - The bitmap blocks have a header (for the checksum).  Then the rest
+ *   of the block is pairs of bits.  With the meaning being:
+ *
+ *   0 - ref count is 0
+ *   1 - ref count is 1
+ *   2 - ref count is 2
+ *   3 - ref count is higher than 2
+ *
+ * - If the count is higher than 2 then the ref count is entered in a
+ *   second btree that directly maps the block_address to a uint32_t ref
+ *   count.
+ *
+ * The space map metadata variant doesn't have a bitmaps btree.  Instead
+ * it has one single blocks worth of index_entries.  This avoids
+ * recursive issues with the bitmap btree needing to allocate space in
+ * order to insert.  With a small data block size such as 64k the
+ * metadata support data devices that are hundreds of terrabytes.
+ *
+ * The space maps allocate space linearly from front to back.  Space that
+ * is freed in a transaction is never recycled within that transaction.
+ * To try and avoid fragmenting _free_ space the allocator always goes
+ * back and fills in gaps.
+ *
+ * All metadata io is in THIN_METADATA_BLOCK_SIZE sized/aligned chunks
+ * from the block manager.
+ *--------------------------------------------------------------------------*/
+
+#define DM_MSG_PREFIX   "thin metadata"
+
+#define THIN_SUPERBLOCK_MAGIC 27022010
+#define THIN_SUPERBLOCK_LOCATION 0
+#define THIN_VERSION 2
+#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
+
+/*
+ * Little endian on-disk superblock and device details.
+ */
+struct thin_disk_superblock {
+	__le32 csum;	/* Checksum of superblock except for this field. */
+	__le32 flags;
+	__le64 blocknr;	/* This block number, dm_block_t. */
+
+	__u8 uuid[16];
+	__le64 magic;
+	__le32 version;
+	__le32 time;
+
+	__le64 trans_id;
+
+	/*
+	 * Root held by userspace transactions.
+	 */
+	__le64 held_root;
+
+	__u8 data_space_map_root[SPACE_MAP_ROOT_SIZE];
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+
+	/*
+	 * 2-level btree mapping (dev_id, (dev block, time)) -> data block
+	 */
+	__le64 data_mapping_root;
+
+	/*
+	 * Device detail root mapping dev_id -> device_details
+	 */
+	__le64 device_details_root;
+
+	__le32 data_block_size;		/* In 512-byte sectors. */
+
+	__le32 metadata_block_size;	/* In 512-byte sectors. */
+	__le64 metadata_nr_blocks;
+
+	__le32 compat_flags;
+	__le32 compat_ro_flags;
+	__le32 incompat_flags;
+} __packed;
+
+struct disk_device_details {
+	__le64 mapped_blocks;
+	__le64 transaction_id;		/* When created. */
+	__le32 creation_time;
+	__le32 snapshotted_time;
+} __packed;
+
+struct dm_pool_metadata {
+	struct hlist_node hash;
+
+	struct block_device *bdev;
+	struct dm_block_manager *bm;
+	struct dm_space_map *metadata_sm;
+	struct dm_space_map *data_sm;
+	struct dm_transaction_manager *tm;
+	struct dm_transaction_manager *nb_tm;
+
+	/*
+	 * Two-level btree.
+	 * First level holds thin_dev_t.
+	 * Second level holds mappings.
+	 */
+	struct dm_btree_info info;
+
+	/*
+	 * Non-blocking version of the above.
+	 */
+	struct dm_btree_info nb_info;
+
+	/*
+	 * Just the top level for deleting whole devices.
+	 */
+	struct dm_btree_info tl_info;
+
+	/*
+	 * Just the bottom level for creating new devices.
+	 */
+	struct dm_btree_info bl_info;
+
+	/*
+	 * Describes the device details btree.
+	 */
+	struct dm_btree_info details_info;
+
+	struct rw_semaphore root_lock;
+	uint32_t time;
+	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;
+
+	/*
+	 * Reading the space map roots can fail, so we read it into these
+	 * buffers before the superblock is locked and updated.
+	 */
+	__u8 data_space_map_root[SPACE_MAP_ROOT_SIZE];
+	__u8 metadata_space_map_root[SPACE_MAP_ROOT_SIZE];
+};
+
+struct dm_thin_device {
+	struct list_head list;
+	struct dm_pool_metadata *pmd;
+	dm_thin_id id;
+
+	int open_count;
+	bool changed:1;
+	bool aborted_with_changes:1;
+	uint64_t mapped_blocks;
+	uint64_t transaction_id;
+	uint32_t creation_time;
+	uint32_t snapshotted_time;
+};
+
+/*----------------------------------------------------------------
+ * superblock validator
+ *--------------------------------------------------------------*/
+
+#define SUPERBLOCK_CSUM_XOR 160774
+
+static void sb_prepare_for_write(struct dm_block_validator *v,
+				 struct dm_block *b,
+				 size_t block_size)
+{
+	struct thin_disk_superblock *disk_super = dm_block_data(b);
+
+	disk_super->blocknr = cpu_to_le64(dm_block_location(b));
+	disk_super->csum = cpu_to_le32(dm_bm_checksum(&disk_super->flags,
+						      block_size - sizeof(__le32),
+						      SUPERBLOCK_CSUM_XOR));
+}
+
+static int sb_check(struct dm_block_validator *v,
+		    struct dm_block *b,
+		    size_t block_size)
+{
+	struct thin_disk_superblock *disk_super = dm_block_data(b);
+	__le32 csum_le;
+
+	if (dm_block_location(b) != le64_to_cpu(disk_super->blocknr)) {
+		DMERR("sb_check failed: blocknr %llu: "
+		      "wanted %llu", le64_to_cpu(disk_super->blocknr),
+		      (unsigned long long)dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	if (le64_to_cpu(disk_super->magic) != THIN_SUPERBLOCK_MAGIC) {
+		DMERR("sb_check failed: magic %llu: "
+		      "wanted %llu", le64_to_cpu(disk_super->magic),
+		      (unsigned long long)THIN_SUPERBLOCK_MAGIC);
+		return -EILSEQ;
+	}
+
+	csum_le = cpu_to_le32(dm_bm_checksum(&disk_super->flags,
+					     block_size - sizeof(__le32),
+					     SUPERBLOCK_CSUM_XOR));
+	if (csum_le != disk_super->csum) {
+		DMERR("sb_check failed: csum %u: wanted %u",
+		      le32_to_cpu(csum_le), le32_to_cpu(disk_super->csum));
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+static struct dm_block_validator sb_validator = {
+	.name = "superblock",
+	.prepare_for_write = sb_prepare_for_write,
+	.check = sb_check
+};
+
+/*----------------------------------------------------------------
+ * Methods for the btree value types
+ *--------------------------------------------------------------*/
+
+static uint64_t pack_block_time(dm_block_t b, uint32_t t)
+{
+	return (b << 24) | t;
+}
+
+static void unpack_block_time(uint64_t v, dm_block_t *b, uint32_t *t)
+{
+	*b = v >> 24;
+	*t = v & ((1 << 24) - 1);
+}
+
+static void data_block_inc(void *context, const void *value_le)
+{
+	struct dm_space_map *sm = context;
+	__le64 v_le;
+	uint64_t b;
+	uint32_t t;
+
+	memcpy(&v_le, value_le, sizeof(v_le));
+	unpack_block_time(le64_to_cpu(v_le), &b, &t);
+	dm_sm_inc_block(sm, b);
+}
+
+static void data_block_dec(void *context, const void *value_le)
+{
+	struct dm_space_map *sm = context;
+	__le64 v_le;
+	uint64_t b;
+	uint32_t t;
+
+	memcpy(&v_le, value_le, sizeof(v_le));
+	unpack_block_time(le64_to_cpu(v_le), &b, &t);
+	dm_sm_dec_block(sm, b);
+}
+
+static int data_block_equal(void *context, const void *value1_le, const void *value2_le)
+{
+	__le64 v1_le, v2_le;
+	uint64_t b1, b2;
+	uint32_t t;
+
+	memcpy(&v1_le, value1_le, sizeof(v1_le));
+	memcpy(&v2_le, value2_le, sizeof(v2_le));
+	unpack_block_time(le64_to_cpu(v1_le), &b1, &t);
+	unpack_block_time(le64_to_cpu(v2_le), &b2, &t);
+
+	return b1 == b2;
+}
+
+static void subtree_inc(void *context, const void *value)
+{
+	struct dm_btree_info *info = context;
+	__le64 root_le;
+	uint64_t root;
+
+	memcpy(&root_le, value, sizeof(root_le));
+	root = le64_to_cpu(root_le);
+	dm_tm_inc(info->tm, root);
+}
+
+static void subtree_dec(void *context, const void *value)
+{
+	struct dm_btree_info *info = context;
+	__le64 root_le;
+	uint64_t root;
+
+	memcpy(&root_le, value, sizeof(root_le));
+	root = le64_to_cpu(root_le);
+	if (dm_btree_del(info, root))
+		DMERR("btree delete failed\n");
+}
+
+static int subtree_equal(void *context, const void *value1_le, const void *value2_le)
+{
+	__le64 v1_le, v2_le;
+	memcpy(&v1_le, value1_le, sizeof(v1_le));
+	memcpy(&v2_le, value2_le, sizeof(v2_le));
+
+	return v1_le == v2_le;
+}
+
+/*----------------------------------------------------------------*/
+
+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;
+	struct dm_block *b;
+	__le64 *data_le, zero = cpu_to_le64(0);
+	unsigned block_size = dm_bm_block_size(bm) / sizeof(__le64);
+
+	/*
+	 * We can't use a validator here - it may be all zeroes.
+	 */
+	r = dm_bm_read_lock(bm, THIN_SUPERBLOCK_LOCATION, NULL, &b);
+	if (r)
+		return r;
+
+	data_le = dm_block_data(b);
+	*result = 1;
+	for (i = 0; i < block_size; i++) {
+		if (data_le[i] != zero) {
+			*result = 0;
+			break;
+		}
+	}
+
+	return dm_bm_unlock(b);
+}
+
+static void __setup_btree_details(struct dm_pool_metadata *pmd)
+{
+	pmd->info.tm = pmd->tm;
+	pmd->info.levels = 2;
+	pmd->info.value_type.context = pmd->data_sm;
+	pmd->info.value_type.size = sizeof(__le64);
+	pmd->info.value_type.inc = data_block_inc;
+	pmd->info.value_type.dec = data_block_dec;
+	pmd->info.value_type.equal = data_block_equal;
+
+	memcpy(&pmd->nb_info, &pmd->info, sizeof(pmd->nb_info));
+	pmd->nb_info.tm = pmd->nb_tm;
+
+	pmd->tl_info.tm = pmd->tm;
+	pmd->tl_info.levels = 1;
+	pmd->tl_info.value_type.context = &pmd->bl_info;
+	pmd->tl_info.value_type.size = sizeof(__le64);
+	pmd->tl_info.value_type.inc = subtree_inc;
+	pmd->tl_info.value_type.dec = subtree_dec;
+	pmd->tl_info.value_type.equal = subtree_equal;
+
+	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);
+	pmd->bl_info.value_type.inc = data_block_inc;
+	pmd->bl_info.value_type.dec = data_block_dec;
+	pmd->bl_info.value_type.equal = data_block_equal;
+
+	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;
+}
+
+static int save_sm_roots(struct dm_pool_metadata *pmd)
+{
+	int r;
+	size_t len;
+
+	r = dm_sm_root_size(pmd->metadata_sm, &len);
+	if (r < 0)
+		return r;
+
+	r = dm_sm_copy_root(pmd->metadata_sm, &pmd->metadata_space_map_root, len);
+	if (r < 0)
+		return r;
+
+	r = dm_sm_root_size(pmd->data_sm, &len);
+	if (r < 0)
+		return r;
+
+	return dm_sm_copy_root(pmd->data_sm, &pmd->data_space_map_root, len);
+}
+
+static void copy_sm_roots(struct dm_pool_metadata *pmd,
+			  struct thin_disk_superblock *disk)
+{
+	memcpy(&disk->metadata_space_map_root,
+	       &pmd->metadata_space_map_root,
+	       sizeof(pmd->metadata_space_map_root));
+
+	memcpy(&disk->data_space_map_root,
+	       &pmd->data_space_map_root,
+	       sizeof(pmd->data_space_map_root));
+}
+
+static int __write_initial_superblock(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct thin_disk_superblock *disk_super;
+	sector_t bdev_size = i_size_read(pmd->bdev->bd_inode) >> SECTOR_SHIFT;
+
+	if (bdev_size > THIN_METADATA_MAX_SECTORS)
+		bdev_size = THIN_METADATA_MAX_SECTORS;
+
+	r = dm_sm_commit(pmd->data_sm);
+	if (r < 0)
+		return r;
+
+	r = save_sm_roots(pmd);
+	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;
+
+	copy_sm_roots(pmd, disk_super);
+
+	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);
+	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);
+}
+
+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_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);
+
+	/* Verify the data block size hasn't changed */
+	if (le32_to_cpu(disk_super->data_block_size) != pmd->data_block_size) {
+		DMERR("changing the data block size (from %u to %llu) is not supported",
+		      le32_to_cpu(disk_super->data_block_size),
+		      (unsigned long long)pmd->data_block_size);
+		r = -EINVAL;
+		goto bad_unlock_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 << SECTOR_SHIFT,
+					  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;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	/*
+	 * We re-read the superblock every time.  Shouldn't need to do this
+	 * really.
+	 */
+	r = dm_bm_read_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			    &sb_validator, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	pmd->time = le32_to_cpu(disk_super->time);
+	pmd->root = le64_to_cpu(disk_super->data_mapping_root);
+	pmd->details_root = le64_to_cpu(disk_super->device_details_root);
+	pmd->trans_id = le64_to_cpu(disk_super->trans_id);
+	pmd->flags = le32_to_cpu(disk_super->flags);
+	pmd->data_block_size = le32_to_cpu(disk_super->data_block_size);
+
+	dm_bm_unlock(sblock);
+	return 0;
+}
+
+static int __write_changed_details(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_thin_device *td, *tmp;
+	struct disk_device_details details;
+	uint64_t key;
+
+	list_for_each_entry_safe(td, tmp, &pmd->thin_devices, list) {
+		if (!td->changed)
+			continue;
+
+		key = td->id;
+
+		details.mapped_blocks = cpu_to_le64(td->mapped_blocks);
+		details.transaction_id = cpu_to_le64(td->transaction_id);
+		details.creation_time = cpu_to_le32(td->creation_time);
+		details.snapshotted_time = cpu_to_le32(td->snapshotted_time);
+		__dm_bless_for_disk(&details);
+
+		r = dm_btree_insert(&pmd->details_info, pmd->details_root,
+				    &key, &details, &pmd->details_root);
+		if (r)
+			return r;
+
+		if (td->open_count)
+			td->changed = 0;
+		else {
+			list_del(&td->list);
+			kfree(td);
+		}
+	}
+
+	return 0;
+}
+
+static int __commit_transaction(struct dm_pool_metadata *pmd)
+{
+	int r;
+	size_t metadata_len, data_len;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	/*
+	 * We need to know if the thin_disk_superblock exceeds a 512-byte sector.
+	 */
+	BUILD_BUG_ON(sizeof(struct thin_disk_superblock) > 512);
+
+	r = __write_changed_details(pmd);
+	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 = 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 = save_sm_roots(pmd);
+	if (r < 0)
+		return r;
+
+	r = superblock_lock(pmd, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	disk_super->time = cpu_to_le32(pmd->time);
+	disk_super->data_mapping_root = cpu_to_le64(pmd->root);
+	disk_super->device_details_root = cpu_to_le64(pmd->details_root);
+	disk_super->trans_id = cpu_to_le64(pmd->trans_id);
+	disk_super->flags = cpu_to_le32(pmd->flags);
+
+	copy_sm_roots(pmd, disk_super);
+
+	return dm_tm_commit(pmd->tm, sblock);
+}
+
+struct dm_pool_metadata *dm_pool_metadata_open(struct block_device *bdev,
+					       sector_t data_block_size,
+					       bool format_device)
+{
+	int r;
+	struct dm_pool_metadata *pmd;
+
+	pmd = kmalloc(sizeof(*pmd), GFP_KERNEL);
+	if (!pmd) {
+		DMERR("could not allocate metadata struct");
+		return ERR_PTR(-ENOMEM);
+	}
+
+	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 = __create_persistent_data_objects(pmd, format_device);
+	if (r) {
+		kfree(pmd);
+		return ERR_PTR(r);
+	}
+
+	r = __begin_transaction(pmd);
+	if (r < 0) {
+		if (dm_pool_metadata_close(pmd) < 0)
+			DMWARN("%s: dm_pool_metadata_close() failed.", __func__);
+		return ERR_PTR(r);
+	}
+
+	return pmd;
+}
+
+int dm_pool_metadata_close(struct dm_pool_metadata *pmd)
+{
+	int r;
+	unsigned open_devices = 0;
+	struct dm_thin_device *td, *tmp;
+
+	down_read(&pmd->root_lock);
+	list_for_each_entry_safe(td, tmp, &pmd->thin_devices, list) {
+		if (td->open_count)
+			open_devices++;
+		else {
+			list_del(&td->list);
+			kfree(td);
+		}
+	}
+	up_read(&pmd->root_lock);
+
+	if (open_devices) {
+		DMERR("attempt to close pmd when %u device(s) are still open",
+		       open_devices);
+		return -EBUSY;
+	}
+
+	if (!pmd->read_only && !pmd->fail_io) {
+		r = __commit_transaction(pmd);
+		if (r < 0)
+			DMWARN("%s: __commit_transaction() failed, error = %d",
+			       __func__, r);
+	}
+
+	if (!pmd->fail_io)
+		__destroy_persistent_data_objects(pmd);
+
+	kfree(pmd);
+	return 0;
+}
+
+/*
+ * __open_device: Returns @td corresponding to device with id @dev,
+ * creating it if @create is set and incrementing @td->open_count.
+ * On failure, @td is undefined.
+ */
+static int __open_device(struct dm_pool_metadata *pmd,
+			 dm_thin_id dev, int create,
+			 struct dm_thin_device **td)
+{
+	int r, changed = 0;
+	struct dm_thin_device *td2;
+	uint64_t key = dev;
+	struct disk_device_details details_le;
+
+	/*
+	 * If the device is already open, return it.
+	 */
+	list_for_each_entry(td2, &pmd->thin_devices, list)
+		if (td2->id == dev) {
+			/*
+			 * May not create an already-open device.
+			 */
+			if (create)
+				return -EEXIST;
+
+			td2->open_count++;
+			*td = td2;
+			return 0;
+		}
+
+	/*
+	 * Check the device exists.
+	 */
+	r = dm_btree_lookup(&pmd->details_info, pmd->details_root,
+			    &key, &details_le);
+	if (r) {
+		if (r != -ENODATA || !create)
+			return r;
+
+		/*
+		 * Create new device.
+		 */
+		changed = 1;
+		details_le.mapped_blocks = 0;
+		details_le.transaction_id = cpu_to_le64(pmd->trans_id);
+		details_le.creation_time = cpu_to_le32(pmd->time);
+		details_le.snapshotted_time = cpu_to_le32(pmd->time);
+	}
+
+	*td = kmalloc(sizeof(**td), GFP_NOIO);
+	if (!*td)
+		return -ENOMEM;
+
+	(*td)->pmd = 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);
+	(*td)->snapshotted_time = le32_to_cpu(details_le.snapshotted_time);
+
+	list_add(&(*td)->list, &pmd->thin_devices);
+
+	return 0;
+}
+
+static void __close_device(struct dm_thin_device *td)
+{
+	--td->open_count;
+}
+
+static int __create_thin(struct dm_pool_metadata *pmd,
+			 dm_thin_id dev)
+{
+	int r;
+	dm_block_t dev_root;
+	uint64_t key = dev;
+	struct disk_device_details details_le;
+	struct dm_thin_device *td;
+	__le64 value;
+
+	r = dm_btree_lookup(&pmd->details_info, pmd->details_root,
+			    &key, &details_le);
+	if (!r)
+		return -EEXIST;
+
+	/*
+	 * Create an empty btree for the mappings.
+	 */
+	r = dm_btree_empty(&pmd->bl_info, &dev_root);
+	if (r)
+		return r;
+
+	/*
+	 * Insert it into the main mapping tree.
+	 */
+	value = cpu_to_le64(dev_root);
+	__dm_bless_for_disk(&value);
+	r = dm_btree_insert(&pmd->tl_info, pmd->root, &key, &value, &pmd->root);
+	if (r) {
+		dm_btree_del(&pmd->bl_info, dev_root);
+		return r;
+	}
+
+	r = __open_device(pmd, dev, 1, &td);
+	if (r) {
+		dm_btree_remove(&pmd->tl_info, pmd->root, &key, &pmd->root);
+		dm_btree_del(&pmd->bl_info, dev_root);
+		return r;
+	}
+	__close_device(td);
+
+	return r;
+}
+
+int dm_pool_create_thin(struct dm_pool_metadata *pmd, dm_thin_id dev)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __create_thin(pmd, dev);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __set_snapshot_details(struct dm_pool_metadata *pmd,
+				  struct dm_thin_device *snap,
+				  dm_thin_id origin, uint32_t time)
+{
+	int r;
+	struct dm_thin_device *td;
+
+	r = __open_device(pmd, origin, 0, &td);
+	if (r)
+		return r;
+
+	td->changed = 1;
+	td->snapshotted_time = time;
+
+	snap->mapped_blocks = td->mapped_blocks;
+	snap->snapshotted_time = time;
+	__close_device(td);
+
+	return 0;
+}
+
+static int __create_snap(struct dm_pool_metadata *pmd,
+			 dm_thin_id dev, dm_thin_id origin)
+{
+	int r;
+	dm_block_t origin_root;
+	uint64_t key = origin, dev_key = dev;
+	struct dm_thin_device *td;
+	struct disk_device_details details_le;
+	__le64 value;
+
+	/* check this device is unused */
+	r = dm_btree_lookup(&pmd->details_info, pmd->details_root,
+			    &dev_key, &details_le);
+	if (!r)
+		return -EEXIST;
+
+	/* find the mapping tree for the origin */
+	r = dm_btree_lookup(&pmd->tl_info, pmd->root, &key, &value);
+	if (r)
+		return r;
+	origin_root = le64_to_cpu(value);
+
+	/* clone the origin, an inc will do */
+	dm_tm_inc(pmd->tm, origin_root);
+
+	/* insert into the main mapping tree */
+	value = cpu_to_le64(origin_root);
+	__dm_bless_for_disk(&value);
+	key = dev;
+	r = dm_btree_insert(&pmd->tl_info, pmd->root, &key, &value, &pmd->root);
+	if (r) {
+		dm_tm_dec(pmd->tm, origin_root);
+		return r;
+	}
+
+	pmd->time++;
+
+	r = __open_device(pmd, dev, 1, &td);
+	if (r)
+		goto bad;
+
+	r = __set_snapshot_details(pmd, td, origin, pmd->time);
+	__close_device(td);
+
+	if (r)
+		goto bad;
+
+	return 0;
+
+bad:
+	dm_btree_remove(&pmd->tl_info, pmd->root, &key, &pmd->root);
+	dm_btree_remove(&pmd->details_info, pmd->details_root,
+			&key, &pmd->details_root);
+	return r;
+}
+
+int dm_pool_create_snap(struct dm_pool_metadata *pmd,
+				 dm_thin_id dev,
+				 dm_thin_id origin)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __create_snap(pmd, dev, origin);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __delete_device(struct dm_pool_metadata *pmd, dm_thin_id dev)
+{
+	int r;
+	uint64_t key = dev;
+	struct dm_thin_device *td;
+
+	/* TODO: failure should mark the transaction invalid */
+	r = __open_device(pmd, dev, 0, &td);
+	if (r)
+		return r;
+
+	if (td->open_count > 1) {
+		__close_device(td);
+		return -EBUSY;
+	}
+
+	list_del(&td->list);
+	kfree(td);
+	r = dm_btree_remove(&pmd->details_info, pmd->details_root,
+			    &key, &pmd->details_root);
+	if (r)
+		return r;
+
+	r = dm_btree_remove(&pmd->tl_info, pmd->root, &key, &pmd->root);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+int dm_pool_delete_thin_device(struct dm_pool_metadata *pmd,
+			       dm_thin_id dev)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __delete_device(pmd, dev);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+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) {
+		DMERR("mismatched transaction id");
+		goto out;
+	}
+
+	pmd->trans_id = new_id;
+	r = 0;
+
+out:
+	up_write(&pmd->root_lock);
+
+	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);
+	if (!pmd->fail_io) {
+		*result = pmd->trans_id;
+		r = 0;
+	}
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+static int __reserve_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r, inc;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *copy, *sblock;
+	dm_block_t held_root;
+
+	/*
+	 * Copy the superblock.
+	 */
+	dm_sm_inc_block(pmd->metadata_sm, THIN_SUPERBLOCK_LOCATION);
+	r = dm_tm_shadow_block(pmd->tm, THIN_SUPERBLOCK_LOCATION,
+			       &sb_validator, &copy, &inc);
+	if (r)
+		return r;
+
+	BUG_ON(!inc);
+
+	held_root = dm_block_location(copy);
+	disk_super = dm_block_data(copy);
+
+	if (le64_to_cpu(disk_super->held_root)) {
+		DMWARN("Pool metadata snapshot already exists: release this before taking another.");
+
+		dm_tm_dec(pmd->tm, held_root);
+		dm_tm_unlock(pmd->tm, copy);
+		return -EBUSY;
+	}
+
+	/*
+	 * Wipe the spacemap since we're not publishing this.
+	 */
+	memset(&disk_super->data_space_map_root, 0,
+	       sizeof(disk_super->data_space_map_root));
+	memset(&disk_super->metadata_space_map_root, 0,
+	       sizeof(disk_super->metadata_space_map_root));
+
+	/*
+	 * Increment the data structures that need to be preserved.
+	 */
+	dm_tm_inc(pmd->tm, le64_to_cpu(disk_super->data_mapping_root));
+	dm_tm_inc(pmd->tm, le64_to_cpu(disk_super->device_details_root));
+	dm_tm_unlock(pmd->tm, copy);
+
+	/*
+	 * Write the held root into the superblock.
+	 */
+	r = superblock_lock(pmd, &sblock);
+	if (r) {
+		dm_tm_dec(pmd->tm, held_root);
+		return r;
+	}
+
+	disk_super = dm_block_data(sblock);
+	disk_super->held_root = cpu_to_le64(held_root);
+	dm_bm_unlock(sblock);
+	return 0;
+}
+
+int dm_pool_reserve_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __reserve_metadata_snap(pmd);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __release_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *sblock, *copy;
+	dm_block_t held_root;
+
+	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);
+
+	dm_bm_unlock(sblock);
+
+	if (!held_root) {
+		DMWARN("No pool metadata snapshot found: nothing to release.");
+		return -EINVAL;
+	}
+
+	r = dm_tm_read_lock(pmd->tm, held_root, &sb_validator, &copy);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(copy);
+	dm_sm_dec_block(pmd->metadata_sm, le64_to_cpu(disk_super->data_mapping_root));
+	dm_sm_dec_block(pmd->metadata_sm, le64_to_cpu(disk_super->device_details_root));
+	dm_sm_dec_block(pmd->metadata_sm, held_root);
+
+	return dm_tm_unlock(pmd->tm, copy);
+}
+
+int dm_pool_release_metadata_snap(struct dm_pool_metadata *pmd)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __release_metadata_snap(pmd);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+static int __get_metadata_snap(struct dm_pool_metadata *pmd,
+			       dm_block_t *result)
+{
+	int r;
+	struct thin_disk_superblock *disk_super;
+	struct dm_block *sblock;
+
+	r = dm_bm_read_lock(pmd->bm, THIN_SUPERBLOCK_LOCATION,
+			    &sb_validator, &sblock);
+	if (r)
+		return r;
+
+	disk_super = dm_block_data(sblock);
+	*result = le64_to_cpu(disk_super->held_root);
+
+	return dm_bm_unlock(sblock);
+}
+
+int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
+			      dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __get_metadata_snap(pmd, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_open_thin_device(struct dm_pool_metadata *pmd, dm_thin_id dev,
+			     struct dm_thin_device **td)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __open_device(pmd, dev, 0, td);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_close_thin_device(struct dm_thin_device *td)
+{
+	down_write(&td->pmd->root_lock);
+	__close_device(td);
+	up_write(&td->pmd->root_lock);
+
+	return 0;
+}
+
+dm_thin_id dm_thin_dev_id(struct dm_thin_device *td)
+{
+	return td->id;
+}
+
+/*
+ * Check whether @time (of block creation) is older than @td's last snapshot.
+ * If so then the associated block is shared with the last snapshot device.
+ * Any block on a device created *after* the device last got snapshotted is
+ * necessarily not shared.
+ */
+static bool __snapshotted_since(struct dm_thin_device *td, uint32_t time)
+{
+	return td->snapshotted_time > 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 = -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);
+		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;
+		unpack_block_time(block_time, &exception_block,
+				  &exception_time);
+		result->block = exception_block;
+		result->shared = __snapshotted_since(td, exception_time);
+	}
+
+	return r;
+}
+
+static int __insert(struct dm_thin_device *td, dm_block_t block,
+		    dm_block_t data_block)
+{
+	int r, inserted;
+	__le64 value;
+	struct dm_pool_metadata *pmd = td->pmd;
+	dm_block_t keys[2] = { td->id, block };
+
+	value = cpu_to_le64(pack_block_time(data_block, pmd->time));
+	__dm_bless_for_disk(&value);
+
+	r = dm_btree_insert_notify(&pmd->info, pmd->root, keys, &value,
+				   &pmd->root, &inserted);
+	if (r)
+		return r;
+
+	td->changed = 1;
+	if (inserted)
+		td->mapped_blocks++;
+
+	return 0;
+}
+
+int dm_thin_insert_block(struct dm_thin_device *td, dm_block_t block,
+			 dm_block_t data_block)
+{
+	int r = -EINVAL;
+
+	down_write(&td->pmd->root_lock);
+	if (!td->pmd->fail_io)
+		r = __insert(td, block, data_block);
+	up_write(&td->pmd->root_lock);
+
+	return r;
+}
+
+static int __remove(struct dm_thin_device *td, dm_block_t block)
+{
+	int r;
+	struct dm_pool_metadata *pmd = td->pmd;
+	dm_block_t keys[2] = { td->id, block };
+
+	r = dm_btree_remove(&pmd->info, pmd->root, keys, &pmd->root);
+	if (r)
+		return r;
+
+	td->mapped_blocks--;
+	td->changed = 1;
+
+	return 0;
+}
+
+int dm_thin_remove_block(struct dm_thin_device *td, dm_block_t block)
+{
+	int r = -EINVAL;
+
+	down_write(&td->pmd->root_lock);
+	if (!td->pmd->fail_io)
+		r = __remove(td, block);
+	up_write(&td->pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_block_is_used(struct dm_pool_metadata *pmd, dm_block_t b, bool *result)
+{
+	int r;
+	uint32_t ref_count;
+
+	down_read(&pmd->root_lock);
+	r = dm_sm_get_count(pmd->data_sm, b, &ref_count);
+	if (!r)
+		*result = (ref_count != 0);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+bool dm_thin_changed_this_transaction(struct dm_thin_device *td)
+{
+	int r;
+
+	down_read(&td->pmd->root_lock);
+	r = td->changed;
+	up_read(&td->pmd->root_lock);
+
+	return r;
+}
+
+bool dm_pool_changed_this_transaction(struct dm_pool_metadata *pmd)
+{
+	bool r = false;
+	struct dm_thin_device *td, *tmp;
+
+	down_read(&pmd->root_lock);
+	list_for_each_entry_safe(td, tmp, &pmd->thin_devices, list) {
+		if (td->changed) {
+			r = td->changed;
+			break;
+		}
+	}
+	up_read(&pmd->root_lock);
+
+	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;
+}
+
+int dm_pool_commit_metadata(struct dm_pool_metadata *pmd)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (pmd->fail_io)
+		goto out;
+
+	r = __commit_transaction(pmd);
+	if (r <= 0)
+		goto out;
+
+	/*
+	 * Open the next transaction.
+	 */
+	r = __begin_transaction(pmd);
+out:
+	up_write(&pmd->root_lock);
+	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 = -EINVAL;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_free(pmd->data_sm, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_get_free_metadata_block_count(struct dm_pool_metadata *pmd,
+					  dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_free(pmd->metadata_sm, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_get_metadata_dev_size(struct dm_pool_metadata *pmd,
+				  dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_blocks(pmd->metadata_sm, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_get_data_block_size(struct dm_pool_metadata *pmd, sector_t *result)
+{
+	down_read(&pmd->root_lock);
+	*result = pmd->data_block_size;
+	up_read(&pmd->root_lock);
+
+	return 0;
+}
+
+int dm_pool_get_data_dev_size(struct dm_pool_metadata *pmd, dm_block_t *result)
+{
+	int r = -EINVAL;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = dm_sm_get_nr_blocks(pmd->data_sm, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+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);
+	if (!pmd->fail_io) {
+		*result = td->mapped_blocks;
+		r = 0;
+	}
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+static int __highest_block(struct dm_thin_device *td, dm_block_t *result)
+{
+	int r;
+	__le64 value_le;
+	dm_block_t thin_root;
+	struct dm_pool_metadata *pmd = td->pmd;
+
+	r = dm_btree_lookup(&pmd->tl_info, pmd->root, &td->id, &value_le);
+	if (r)
+		return r;
+
+	thin_root = le64_to_cpu(value_le);
+
+	return dm_btree_find_highest_key(&pmd->bl_info, thin_root, result);
+}
+
+int dm_thin_get_highest_mapped_block(struct dm_thin_device *td,
+				     dm_block_t *result)
+{
+	int r = -EINVAL;
+	struct dm_pool_metadata *pmd = td->pmd;
+
+	down_read(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __highest_block(td, result);
+	up_read(&pmd->root_lock);
+
+	return r;
+}
+
+static int __resize_space_map(struct dm_space_map *sm, dm_block_t new_count)
+{
+	int r;
+	dm_block_t old_count;
+
+	r = dm_sm_get_nr_blocks(sm, &old_count);
+	if (r)
+		return r;
+
+	if (new_count == old_count)
+		return 0;
+
+	if (new_count < old_count) {
+		DMERR("cannot reduce size of space map");
+		return -EINVAL;
+	}
+
+	return dm_sm_extend(sm, new_count - old_count);
+}
+
+int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __resize_space_map(pmd->data_sm, new_count);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_resize_metadata_dev(struct dm_pool_metadata *pmd, dm_block_t new_count)
+{
+	int r = -EINVAL;
+
+	down_write(&pmd->root_lock);
+	if (!pmd->fail_io)
+		r = __resize_space_map(pmd->metadata_sm, 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);
+}
+
+void dm_pool_metadata_read_write(struct dm_pool_metadata *pmd)
+{
+	down_write(&pmd->root_lock);
+	pmd->read_only = false;
+	dm_bm_set_read_write(pmd->bm);
+	up_write(&pmd->root_lock);
+}
+
+int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
+					dm_block_t threshold,
+					dm_sm_threshold_fn fn,
+					void *context)
+{
+	int r;
+
+	down_write(&pmd->root_lock);
+	r = dm_sm_register_threshold_callback(pmd->metadata_sm, threshold, fn, context);
+	up_write(&pmd->root_lock);
+
+	return r;
+}
+
+int dm_pool_metadata_set_needs_check(struct dm_pool_metadata *pmd)
+{
+	int r;
+	struct dm_block *sblock;
+	struct thin_disk_superblock *disk_super;
+
+	down_write(&pmd->root_lock);
+	pmd->flags |= THIN_METADATA_NEEDS_CHECK_FLAG;
+
+	r = superblock_lock(pmd, &sblock);
+	if (r) {
+		DMERR("couldn't read superblock");
+		goto out;
+	}
+
+	disk_super = dm_block_data(sblock);
+	disk_super->flags = cpu_to_le32(pmd->flags);
+
+	dm_bm_unlock(sblock);
+out:
+	up_write(&pmd->root_lock);
+	return r;
+}
+
+bool dm_pool_metadata_needs_check(struct dm_pool_metadata *pmd)
+{
+	bool needs_check;
+
+	down_read(&pmd->root_lock);
+	needs_check = pmd->flags & THIN_METADATA_NEEDS_CHECK_FLAG;
+	up_read(&pmd->root_lock);
+
+	return needs_check;
+}
diff --git a/drivers/md/dm-thin-metadata.h b/drivers/md/dm-thin-metadata.h
new file mode 100644
index 00000000000..e3c857db195
--- /dev/null
+++ b/drivers/md/dm-thin-metadata.h
@@ -0,0 +1,218 @@
+/*
+ * Copyright (C) 2010-2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_THIN_METADATA_H
+#define DM_THIN_METADATA_H
+
+#include "persistent-data/dm-block-manager.h"
+#include "persistent-data/dm-space-map.h"
+#include "persistent-data/dm-space-map-metadata.h"
+
+#define THIN_METADATA_BLOCK_SIZE DM_SM_METADATA_BLOCK_SIZE
+
+/*
+ * The metadata device is currently limited in size.
+ */
+#define THIN_METADATA_MAX_SECTORS DM_SM_METADATA_MAX_SECTORS
+
+/*
+ * A metadata device larger than 16GB triggers a warning.
+ */
+#define THIN_METADATA_MAX_SECTORS_WARNING (16 * (1024 * 1024 * 1024 >> SECTOR_SHIFT))
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Thin metadata superblock flags.
+ */
+#define THIN_METADATA_NEEDS_CHECK_FLAG (1 << 0)
+
+struct dm_pool_metadata;
+struct dm_thin_device;
+
+/*
+ * Device identifier
+ */
+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,
+					       bool format_device);
+
+int dm_pool_metadata_close(struct dm_pool_metadata *pmd);
+
+/*
+ * Compat feature flags.  Any incompat flags beyond the ones
+ * specified below will prevent use of the thin metadata.
+ */
+#define THIN_FEATURE_COMPAT_SUPP	  0UL
+#define THIN_FEATURE_COMPAT_RO_SUPP	  0UL
+#define THIN_FEATURE_INCOMPAT_SUPP	  0UL
+
+/*
+ * Device creation/deletion.
+ */
+int dm_pool_create_thin(struct dm_pool_metadata *pmd, dm_thin_id dev);
+
+/*
+ * An internal snapshot.
+ *
+ * You can only snapshot a quiesced origin i.e. one that is either
+ * suspended or not instanced at all.
+ */
+int dm_pool_create_snap(struct dm_pool_metadata *pmd, dm_thin_id dev,
+			dm_thin_id origin);
+
+/*
+ * Deletes a virtual device from the metadata.  It _is_ safe to call this
+ * when that device is open.  Operations on that device will just start
+ * failing.  You still need to call close() on the device.
+ */
+int dm_pool_delete_thin_device(struct dm_pool_metadata *pmd,
+			       dm_thin_id dev);
+
+/*
+ * Commits _all_ metadata changes: device creation, deletion, mapping
+ * updates.
+ */
+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,
+					uint64_t current_id,
+					uint64_t new_id);
+
+int dm_pool_get_metadata_transaction_id(struct dm_pool_metadata *pmd,
+					uint64_t *result);
+
+/*
+ * Hold/get root for userspace transaction.
+ *
+ * The metadata snapshot is a copy of the current superblock (minus the
+ * space maps).  Userland can access the data structures for READ
+ * operations only.  A small performance hit is incurred by providing this
+ * copy of the metadata to userland due to extra copy-on-write operations
+ * on the metadata nodes.  Release this as soon as you finish with it.
+ */
+int dm_pool_reserve_metadata_snap(struct dm_pool_metadata *pmd);
+int dm_pool_release_metadata_snap(struct dm_pool_metadata *pmd);
+
+int dm_pool_get_metadata_snap(struct dm_pool_metadata *pmd,
+			      dm_block_t *result);
+
+/*
+ * Actions on a single virtual device.
+ */
+
+/*
+ * Opening the same device more than once will fail with -EBUSY.
+ */
+int dm_pool_open_thin_device(struct dm_pool_metadata *pmd, dm_thin_id dev,
+			     struct dm_thin_device **td);
+
+int dm_pool_close_thin_device(struct dm_thin_device *td);
+
+dm_thin_id dm_thin_dev_id(struct dm_thin_device *td);
+
+struct dm_thin_lookup_result {
+	dm_block_t block;
+	bool shared:1;
+};
+
+/*
+ * Returns:
+ *   -EWOULDBLOCK iff @can_block is set and would block.
+ *   -ENODATA iff that mapping is not present.
+ *   0 success
+ */
+int dm_thin_find_block(struct dm_thin_device *td, dm_block_t block,
+		       int can_block, struct dm_thin_lookup_result *result);
+
+/*
+ * Obtain an unused block.
+ */
+int dm_pool_alloc_data_block(struct dm_pool_metadata *pmd, dm_block_t *result);
+
+/*
+ * Insert or remove block.
+ */
+int dm_thin_insert_block(struct dm_thin_device *td, dm_block_t block,
+			 dm_block_t data_block);
+
+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_pool_changed_this_transaction(struct dm_pool_metadata *pmd);
+
+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);
+
+int dm_thin_get_mapped_count(struct dm_thin_device *td, dm_block_t *result);
+
+int dm_pool_get_free_block_count(struct dm_pool_metadata *pmd,
+				 dm_block_t *result);
+
+int dm_pool_get_free_metadata_block_count(struct dm_pool_metadata *pmd,
+					  dm_block_t *result);
+
+int dm_pool_get_metadata_dev_size(struct dm_pool_metadata *pmd,
+				  dm_block_t *result);
+
+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 dm_pool_block_is_used(struct dm_pool_metadata *pmd, dm_block_t b, bool *result);
+
+/*
+ * Returns -ENOSPC if the new size is too small and already allocated
+ * blocks would be lost.
+ */
+int dm_pool_resize_data_dev(struct dm_pool_metadata *pmd, dm_block_t new_size);
+int dm_pool_resize_metadata_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);
+void dm_pool_metadata_read_write(struct dm_pool_metadata *pmd);
+
+int dm_pool_register_metadata_threshold(struct dm_pool_metadata *pmd,
+					dm_block_t threshold,
+					dm_sm_threshold_fn fn,
+					void *context);
+
+/*
+ * Updates the superblock immediately.
+ */
+int dm_pool_metadata_set_needs_check(struct dm_pool_metadata *pmd);
+bool dm_pool_metadata_needs_check(struct dm_pool_metadata *pmd);
+
+/*----------------------------------------------------------------*/
+
+#endif
diff --git a/drivers/md/dm-thin.c b/drivers/md/dm-thin.c
new file mode 100644
index 00000000000..fc9c848a60c
--- /dev/null
+++ b/drivers/md/dm-thin.c
@@ -0,0 +1,3546 @@
+/*
+ * Copyright (C) 2011-2012 Red Hat UK.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-thin-metadata.h"
+#include "dm-bio-prison.h"
+#include "dm.h"
+
+#include <linux/device-mapper.h>
+#include <linux/dm-io.h>
+#include <linux/dm-kcopyd.h>
+#include <linux/list.h>
+#include <linux/rculist.h>
+#include <linux/init.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/rbtree.h>
+
+#define	DM_MSG_PREFIX	"thin"
+
+/*
+ * Tunable constants
+ */
+#define ENDIO_HOOK_POOL_SIZE 1024
+#define MAPPING_POOL_SIZE 1024
+#define PRISON_CELLS 1024
+#define COMMIT_PERIOD HZ
+#define NO_SPACE_TIMEOUT_SECS 60
+
+static unsigned no_space_timeout_secs = NO_SPACE_TIMEOUT_SECS;
+
+DECLARE_DM_KCOPYD_THROTTLE_WITH_MODULE_PARM(snapshot_copy_throttle,
+		"A percentage of time allocated for copy on write");
+
+/*
+ * The block size of the device holding pool data must be
+ * between 64KB and 1GB.
+ */
+#define DATA_DEV_BLOCK_SIZE_MIN_SECTORS (64 * 1024 >> SECTOR_SHIFT)
+#define DATA_DEV_BLOCK_SIZE_MAX_SECTORS (1024 * 1024 * 1024 >> SECTOR_SHIFT)
+
+/*
+ * Device id is restricted to 24 bits.
+ */
+#define MAX_DEV_ID ((1 << 24) - 1)
+
+/*
+ * How do we handle breaking sharing of data blocks?
+ * =================================================
+ *
+ * We use a standard copy-on-write btree to store the mappings for the
+ * devices (note I'm talking about copy-on-write of the metadata here, not
+ * the data).  When you take an internal snapshot you clone the root node
+ * of the origin btree.  After this there is no concept of an origin or a
+ * snapshot.  They are just two device trees that happen to point to the
+ * same data blocks.
+ *
+ * When we get a write in we decide if it's to a shared data block using
+ * some timestamp magic.  If it is, we have to break sharing.
+ *
+ * Let's say we write to a shared block in what was the origin.  The
+ * steps are:
+ *
+ * i) plug io further to this physical block. (see bio_prison code).
+ *
+ * ii) quiesce any read io to that shared data block.  Obviously
+ * including all devices that share this block.  (see dm_deferred_set code)
+ *
+ * iii) copy the data block to a newly allocate block.  This step can be
+ * missed out if the io covers the block. (schedule_copy).
+ *
+ * iv) insert the new mapping into the origin's btree
+ * (process_prepared_mapping).  This act of inserting breaks some
+ * sharing of btree nodes between the two devices.  Breaking sharing only
+ * effects the btree of that specific device.  Btrees for the other
+ * devices that share the block never change.  The btree for the origin
+ * device as it was after the last commit is untouched, ie. we're using
+ * persistent data structures in the functional programming sense.
+ *
+ * v) unplug io to this physical block, including the io that triggered
+ * the breaking of sharing.
+ *
+ * Steps (ii) and (iii) occur in parallel.
+ *
+ * The metadata _doesn't_ need to be committed before the io continues.  We
+ * get away with this because the io is always written to a _new_ block.
+ * If there's a crash, then:
+ *
+ * - The origin mapping will point to the old origin block (the shared
+ * one).  This will contain the data as it was before the io that triggered
+ * the breaking of sharing came in.
+ *
+ * - The snap mapping still points to the old block.  As it would after
+ * the commit.
+ *
+ * The downside of this scheme is the timestamp magic isn't perfect, and
+ * will continue to think that data block in the snapshot device is shared
+ * even after the write to the origin has broken sharing.  I suspect data
+ * blocks will typically be shared by many different devices, so we're
+ * breaking sharing n + 1 times, rather than n, where n is the number of
+ * devices that reference this data block.  At the moment I think the
+ * benefits far, far outweigh the disadvantages.
+ */
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Key building.
+ */
+static void build_data_key(struct dm_thin_device *td,
+			   dm_block_t b, struct dm_cell_key *key)
+{
+	key->virtual = 0;
+	key->dev = dm_thin_dev_id(td);
+	key->block = b;
+}
+
+static void build_virtual_key(struct dm_thin_device *td, dm_block_t b,
+			      struct dm_cell_key *key)
+{
+	key->virtual = 1;
+	key->dev = dm_thin_dev_id(td);
+	key->block = b;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * A pool device ties together a metadata device and a data device.  It
+ * also provides the interface for creating and destroying internal
+ * devices.
+ */
+struct dm_thin_new_mapping;
+
+/*
+ * The pool runs in 4 modes.  Ordered in degraded order for comparisons.
+ */
+enum pool_mode {
+	PM_WRITE,		/* metadata may be changed */
+	PM_OUT_OF_DATA_SPACE,	/* metadata may be changed, though data may not be allocated */
+	PM_READ_ONLY,		/* metadata may not be changed */
+	PM_FAIL,		/* all I/O fails */
+};
+
+struct pool_features {
+	enum pool_mode mode;
+
+	bool zero_new_blocks:1;
+	bool discard_enabled:1;
+	bool discard_passdown:1;
+	bool error_if_no_space: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 */
+
+	struct mapped_device *pool_md;
+	struct block_device *md_dev;
+	struct dm_pool_metadata *pmd;
+
+	dm_block_t low_water_blocks;
+	uint32_t sectors_per_block;
+	int sectors_per_block_shift;
+
+	struct pool_features pf;
+	bool low_water_triggered:1;	/* A dm event has been sent */
+
+	struct dm_bio_prison *prison;
+	struct dm_kcopyd_client *copier;
+
+	struct workqueue_struct *wq;
+	struct work_struct worker;
+	struct delayed_work waker;
+	struct delayed_work no_space_timeout;
+
+	unsigned long last_commit_jiffies;
+	unsigned ref_count;
+
+	spinlock_t lock;
+	struct bio_list deferred_flush_bios;
+	struct list_head prepared_mappings;
+	struct list_head prepared_discards;
+	struct list_head active_thins;
+
+	struct dm_deferred_set *shared_read_ds;
+	struct dm_deferred_set *all_io_ds;
+
+	struct dm_thin_new_mapping *next_mapping;
+	mempool_t *mapping_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 metadata_operation_failed(struct pool *pool, const char *op, int r);
+
+/*
+ * Target context for a pool.
+ */
+struct pool_c {
+	struct dm_target *ti;
+	struct pool *pool;
+	struct dm_dev *data_dev;
+	struct dm_dev *metadata_dev;
+	struct dm_target_callbacks callbacks;
+
+	dm_block_t low_water_blocks;
+	struct pool_features requested_pf; /* Features requested during table load */
+	struct pool_features adjusted_pf;  /* Features used after adjusting for constituent devices */
+};
+
+/*
+ * Target context for a thin.
+ */
+struct thin_c {
+	struct list_head list;
+	struct dm_dev *pool_dev;
+	struct dm_dev *origin_dev;
+	dm_thin_id dev_id;
+
+	struct pool *pool;
+	struct dm_thin_device *td;
+	bool requeue_mode:1;
+	spinlock_t lock;
+	struct bio_list deferred_bio_list;
+	struct bio_list retry_on_resume_list;
+	struct rb_root sort_bio_list; /* sorted list of deferred bios */
+
+	/*
+	 * Ensures the thin is not destroyed until the worker has finished
+	 * iterating the active_thins list.
+	 */
+	atomic_t refcount;
+	struct completion can_destroy;
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * wake_worker() is used when new work is queued and when pool_resume is
+ * ready to continue deferred IO processing.
+ */
+static void wake_worker(struct pool *pool)
+{
+	queue_work(pool->wq, &pool->worker);
+}
+
+/*----------------------------------------------------------------*/
+
+static int bio_detain(struct pool *pool, struct dm_cell_key *key, struct bio *bio,
+		      struct dm_bio_prison_cell **cell_result)
+{
+	int r;
+	struct dm_bio_prison_cell *cell_prealloc;
+
+	/*
+	 * Allocate a cell from the prison's mempool.
+	 * This might block but it can't fail.
+	 */
+	cell_prealloc = dm_bio_prison_alloc_cell(pool->prison, GFP_NOIO);
+
+	r = dm_bio_detain(pool->prison, key, bio, cell_prealloc, cell_result);
+	if (r)
+		/*
+		 * We reused an old cell; we can get rid of
+		 * the new one.
+		 */
+		dm_bio_prison_free_cell(pool->prison, cell_prealloc);
+
+	return r;
+}
+
+static void cell_release(struct pool *pool,
+			 struct dm_bio_prison_cell *cell,
+			 struct bio_list *bios)
+{
+	dm_cell_release(pool->prison, cell, bios);
+	dm_bio_prison_free_cell(pool->prison, cell);
+}
+
+static void cell_release_no_holder(struct pool *pool,
+				   struct dm_bio_prison_cell *cell,
+				   struct bio_list *bios)
+{
+	dm_cell_release_no_holder(pool->prison, cell, bios);
+	dm_bio_prison_free_cell(pool->prison, cell);
+}
+
+static void cell_defer_no_holder_no_free(struct thin_c *tc,
+					 struct dm_bio_prison_cell *cell)
+{
+	struct pool *pool = tc->pool;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tc->lock, flags);
+	dm_cell_release_no_holder(pool->prison, cell, &tc->deferred_bio_list);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	wake_worker(pool);
+}
+
+static void cell_error_with_code(struct pool *pool,
+				 struct dm_bio_prison_cell *cell, int error_code)
+{
+	dm_cell_error(pool->prison, cell, error_code);
+	dm_bio_prison_free_cell(pool->prison, cell);
+}
+
+static void cell_error(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+	cell_error_with_code(pool, cell, -EIO);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * A global list of pools that uses a struct mapped_device as a key.
+ */
+static struct dm_thin_pool_table {
+	struct mutex mutex;
+	struct list_head pools;
+} dm_thin_pool_table;
+
+static void pool_table_init(void)
+{
+	mutex_init(&dm_thin_pool_table.mutex);
+	INIT_LIST_HEAD(&dm_thin_pool_table.pools);
+}
+
+static void __pool_table_insert(struct pool *pool)
+{
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+	list_add(&pool->list, &dm_thin_pool_table.pools);
+}
+
+static void __pool_table_remove(struct pool *pool)
+{
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+	list_del(&pool->list);
+}
+
+static struct pool *__pool_table_lookup(struct mapped_device *md)
+{
+	struct pool *pool = NULL, *tmp;
+
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+
+	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
+		if (tmp->pool_md == md) {
+			pool = tmp;
+			break;
+		}
+	}
+
+	return pool;
+}
+
+static struct pool *__pool_table_lookup_metadata_dev(struct block_device *md_dev)
+{
+	struct pool *pool = NULL, *tmp;
+
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+
+	list_for_each_entry(tmp, &dm_thin_pool_table.pools, list) {
+		if (tmp->md_dev == md_dev) {
+			pool = tmp;
+			break;
+		}
+	}
+
+	return pool;
+}
+
+/*----------------------------------------------------------------*/
+
+struct dm_thin_endio_hook {
+	struct thin_c *tc;
+	struct dm_deferred_entry *shared_read_entry;
+	struct dm_deferred_entry *all_io_entry;
+	struct dm_thin_new_mapping *overwrite_mapping;
+	struct rb_node rb_node;
+};
+
+static void requeue_bio_list(struct thin_c *tc, struct bio_list *master)
+{
+	struct bio *bio;
+	struct bio_list bios;
+	unsigned long flags;
+
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&tc->lock, flags);
+	bio_list_merge(&bios, master);
+	bio_list_init(master);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	while ((bio = bio_list_pop(&bios)))
+		bio_endio(bio, DM_ENDIO_REQUEUE);
+}
+
+static void requeue_io(struct thin_c *tc)
+{
+	requeue_bio_list(tc, &tc->deferred_bio_list);
+	requeue_bio_list(tc, &tc->retry_on_resume_list);
+}
+
+static void error_thin_retry_list(struct thin_c *tc)
+{
+	struct bio *bio;
+	unsigned long flags;
+	struct bio_list bios;
+
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&tc->lock, flags);
+	bio_list_merge(&bios, &tc->retry_on_resume_list);
+	bio_list_init(&tc->retry_on_resume_list);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	while ((bio = bio_list_pop(&bios)))
+		bio_io_error(bio);
+}
+
+static void error_retry_list(struct pool *pool)
+{
+	struct thin_c *tc;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(tc, &pool->active_thins, list)
+		error_thin_retry_list(tc);
+	rcu_read_unlock();
+}
+
+/*
+ * This section of code contains the logic for processing a thin device's IO.
+ * Much of the code depends on pool object resources (lists, workqueues, etc)
+ * but most is exclusively called from the thin target rather than the thin-pool
+ * target.
+ */
+
+static bool block_size_is_power_of_two(struct pool *pool)
+{
+	return pool->sectors_per_block_shift >= 0;
+}
+
+static dm_block_t get_bio_block(struct thin_c *tc, struct bio *bio)
+{
+	struct pool *pool = tc->pool;
+	sector_t block_nr = bio->bi_iter.bi_sector;
+
+	if (block_size_is_power_of_two(pool))
+		block_nr >>= pool->sectors_per_block_shift;
+	else
+		(void) sector_div(block_nr, pool->sectors_per_block);
+
+	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_iter.bi_sector;
+
+	bio->bi_bdev = tc->pool_dev->bdev;
+	if (block_size_is_power_of_two(pool))
+		bio->bi_iter.bi_sector =
+			(block << pool->sectors_per_block_shift) |
+			(bi_sector & (pool->sectors_per_block - 1));
+	else
+		bio->bi_iter.bi_sector = (block * pool->sectors_per_block) +
+				 sector_div(bi_sector, pool->sectors_per_block);
+}
+
+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 inc_all_io_entry(struct pool *pool, struct bio *bio)
+{
+	struct dm_thin_endio_hook *h;
+
+	if (bio->bi_rw & REQ_DISCARD)
+		return;
+
+	h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+	h->all_io_entry = dm_deferred_entry_inc(pool->all_io_ds);
+}
+
+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;
+	}
+
+	/*
+	 * 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 (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)
+{
+	remap_to_origin(tc, bio);
+	issue(tc, bio);
+}
+
+static void remap_and_issue(struct thin_c *tc, struct bio *bio,
+			    dm_block_t block)
+{
+	remap(tc, bio, block);
+	issue(tc, bio);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Bio endio functions.
+ */
+struct dm_thin_new_mapping {
+	struct list_head list;
+
+	bool quiesced:1;
+	bool prepared:1;
+	bool pass_discard:1;
+	bool definitely_not_shared:1;
+
+	int err;
+	struct thin_c *tc;
+	dm_block_t virt_block;
+	dm_block_t data_block;
+	struct dm_bio_prison_cell *cell, *cell2;
+
+	/*
+	 * If the bio covers the whole area of a block then we can avoid
+	 * zeroing or copying.  Instead this bio is hooked.  The bio will
+	 * still be in the cell, so care has to be taken to avoid issuing
+	 * the bio twice.
+	 */
+	struct bio *bio;
+	bio_end_io_t *saved_bi_end_io;
+};
+
+static void __maybe_add_mapping(struct dm_thin_new_mapping *m)
+{
+	struct pool *pool = m->tc->pool;
+
+	if (m->quiesced && m->prepared) {
+		list_add_tail(&m->list, &pool->prepared_mappings);
+		wake_worker(pool);
+	}
+}
+
+static void copy_complete(int read_err, unsigned long write_err, void *context)
+{
+	unsigned long flags;
+	struct dm_thin_new_mapping *m = context;
+	struct pool *pool = m->tc->pool;
+
+	m->err = read_err || write_err ? -EIO : 0;
+
+	spin_lock_irqsave(&pool->lock, flags);
+	m->prepared = true;
+	__maybe_add_mapping(m);
+	spin_unlock_irqrestore(&pool->lock, flags);
+}
+
+static void overwrite_endio(struct bio *bio, int err)
+{
+	unsigned long flags;
+	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+	struct dm_thin_new_mapping *m = h->overwrite_mapping;
+	struct pool *pool = m->tc->pool;
+
+	m->err = err;
+
+	spin_lock_irqsave(&pool->lock, flags);
+	m->prepared = true;
+	__maybe_add_mapping(m);
+	spin_unlock_irqrestore(&pool->lock, flags);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Workqueue.
+ */
+
+/*
+ * Prepared mapping jobs.
+ */
+
+/*
+ * This sends the bios in the cell back to the deferred_bios list.
+ */
+static void cell_defer(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+	struct pool *pool = tc->pool;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tc->lock, flags);
+	cell_release(pool, cell, &tc->deferred_bio_list);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	wake_worker(pool);
+}
+
+/*
+ * Same as cell_defer above, except it omits the original holder of the cell.
+ */
+static void cell_defer_no_holder(struct thin_c *tc, struct dm_bio_prison_cell *cell)
+{
+	struct pool *pool = tc->pool;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tc->lock, flags);
+	cell_release_no_holder(pool, cell, &tc->deferred_bio_list);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	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;
+		atomic_inc(&m->bio->bi_remaining);
+	}
+	cell_error(m->tc->pool, 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;
+	struct pool *pool = tc->pool;
+	struct bio *bio;
+	int r;
+
+	bio = m->bio;
+	if (bio) {
+		bio->bi_end_io = m->saved_bi_end_io;
+		atomic_inc(&bio->bi_remaining);
+	}
+
+	if (m->err) {
+		cell_error(pool, m->cell);
+		goto out;
+	}
+
+	/*
+	 * Commit the prepared block into the mapping btree.
+	 * Any I/O for this block arriving after this point will get
+	 * remapped to it directly.
+	 */
+	r = dm_thin_insert_block(tc->td, m->virt_block, m->data_block);
+	if (r) {
+		metadata_operation_failed(pool, "dm_thin_insert_block", r);
+		cell_error(pool, m->cell);
+		goto out;
+	}
+
+	/*
+	 * Release any bios held while the block was being provisioned.
+	 * If we are processing a write bio that completely covers the block,
+	 * we already processed it so can ignore it now when processing
+	 * the bios in the cell.
+	 */
+	if (bio) {
+		cell_defer_no_holder(tc, m->cell);
+		bio_endio(bio, 0);
+	} else
+		cell_defer(tc, m->cell);
+
+out:
+	list_del(&m->list);
+	mempool_free(m, pool->mapping_pool);
+}
+
+static void process_prepared_discard_fail(struct dm_thin_new_mapping *m)
+{
+	struct thin_c *tc = m->tc;
+
+	bio_io_error(m->bio);
+	cell_defer_no_holder(tc, m->cell);
+	cell_defer_no_holder(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;
+
+	inc_all_io_entry(tc->pool, m->bio);
+	cell_defer_no_holder(tc, m->cell);
+	cell_defer_no_holder(tc, m->cell2);
+
+	if (m->pass_discard)
+		if (m->definitely_not_shared)
+			remap_and_issue(tc, m->bio, m->data_block);
+		else {
+			bool used = false;
+			if (dm_pool_block_is_used(tc->pool->pmd, m->data_block, &used) || used)
+				bio_endio(m->bio, 0);
+			else
+				remap_and_issue(tc, m->bio, m->data_block);
+		}
+	else
+		bio_endio(m->bio, 0);
+
+	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_LIMIT("dm_thin_remove_block() failed");
+
+	process_prepared_discard_passdown(m);
+}
+
+static void process_prepared(struct pool *pool, struct list_head *head,
+			     process_mapping_fn *fn)
+{
+	unsigned long flags;
+	struct list_head maps;
+	struct dm_thin_new_mapping *m, *tmp;
+
+	INIT_LIST_HEAD(&maps);
+	spin_lock_irqsave(&pool->lock, flags);
+	list_splice_init(head, &maps);
+	spin_unlock_irqrestore(&pool->lock, flags);
+
+	list_for_each_entry_safe(m, tmp, &maps, list)
+		(*fn)(m);
+}
+
+/*
+ * Deferred bio jobs.
+ */
+static int io_overlaps_block(struct pool *pool, struct bio *bio)
+{
+	return bio->bi_iter.bi_size ==
+		(pool->sectors_per_block << SECTOR_SHIFT);
+}
+
+static int io_overwrites_block(struct pool *pool, struct bio *bio)
+{
+	return (bio_data_dir(bio) == WRITE) &&
+		io_overlaps_block(pool, bio);
+}
+
+static void save_and_set_endio(struct bio *bio, bio_end_io_t **save,
+			       bio_end_io_t *fn)
+{
+	*save = bio->bi_end_io;
+	bio->bi_end_io = fn;
+}
+
+static int ensure_next_mapping(struct pool *pool)
+{
+	if (pool->next_mapping)
+		return 0;
+
+	pool->next_mapping = mempool_alloc(pool->mapping_pool, GFP_ATOMIC);
+
+	return pool->next_mapping ? 0 : -ENOMEM;
+}
+
+static struct dm_thin_new_mapping *get_next_mapping(struct pool *pool)
+{
+	struct dm_thin_new_mapping *m = pool->next_mapping;
+
+	BUG_ON(!pool->next_mapping);
+
+	memset(m, 0, sizeof(struct dm_thin_new_mapping));
+	INIT_LIST_HEAD(&m->list);
+	m->bio = NULL;
+
+	pool->next_mapping = NULL;
+
+	return m;
+}
+
+static void schedule_copy(struct thin_c *tc, dm_block_t virt_block,
+			  struct dm_dev *origin, dm_block_t data_origin,
+			  dm_block_t data_dest,
+			  struct dm_bio_prison_cell *cell, struct bio *bio)
+{
+	int r;
+	struct pool *pool = tc->pool;
+	struct dm_thin_new_mapping *m = get_next_mapping(pool);
+
+	m->tc = tc;
+	m->virt_block = virt_block;
+	m->data_block = data_dest;
+	m->cell = cell;
+
+	if (!dm_deferred_set_add_work(pool->shared_read_ds, &m->list))
+		m->quiesced = true;
+
+	/*
+	 * IO to pool_dev remaps to the pool target's data_dev.
+	 *
+	 * If the whole block of data is being overwritten, we can issue the
+	 * bio immediately. Otherwise we use kcopyd to clone the data first.
+	 */
+	if (io_overwrites_block(pool, bio)) {
+		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+
+		h->overwrite_mapping = m;
+		m->bio = bio;
+		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
+		inc_all_io_entry(pool, bio);
+		remap_and_issue(tc, bio, data_dest);
+	} else {
+		struct dm_io_region from, to;
+
+		from.bdev = origin->bdev;
+		from.sector = data_origin * pool->sectors_per_block;
+		from.count = pool->sectors_per_block;
+
+		to.bdev = tc->pool_dev->bdev;
+		to.sector = data_dest * pool->sectors_per_block;
+		to.count = pool->sectors_per_block;
+
+		r = dm_kcopyd_copy(pool->copier, &from, 1, &to,
+				   0, copy_complete, m);
+		if (r < 0) {
+			mempool_free(m, pool->mapping_pool);
+			DMERR_LIMIT("dm_kcopyd_copy() failed");
+			cell_error(pool, cell);
+		}
+	}
+}
+
+static void schedule_internal_copy(struct thin_c *tc, dm_block_t virt_block,
+				   dm_block_t data_origin, dm_block_t data_dest,
+				   struct dm_bio_prison_cell *cell, struct bio *bio)
+{
+	schedule_copy(tc, virt_block, tc->pool_dev,
+		      data_origin, data_dest, cell, bio);
+}
+
+static void schedule_external_copy(struct thin_c *tc, dm_block_t virt_block,
+				   dm_block_t data_dest,
+				   struct dm_bio_prison_cell *cell, struct bio *bio)
+{
+	schedule_copy(tc, virt_block, tc->origin_dev,
+		      virt_block, data_dest, cell, bio);
+}
+
+static void schedule_zero(struct thin_c *tc, dm_block_t virt_block,
+			  dm_block_t data_block, struct dm_bio_prison_cell *cell,
+			  struct bio *bio)
+{
+	struct pool *pool = tc->pool;
+	struct dm_thin_new_mapping *m = get_next_mapping(pool);
+
+	m->quiesced = true;
+	m->prepared = false;
+	m->tc = tc;
+	m->virt_block = virt_block;
+	m->data_block = data_block;
+	m->cell = cell;
+
+	/*
+	 * If the whole block of data is being overwritten or we are not
+	 * zeroing pre-existing data, we can issue the bio immediately.
+	 * Otherwise we use kcopyd to zero the data first.
+	 */
+	if (!pool->pf.zero_new_blocks)
+		process_prepared_mapping(m);
+
+	else if (io_overwrites_block(pool, bio)) {
+		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+
+		h->overwrite_mapping = m;
+		m->bio = bio;
+		save_and_set_endio(bio, &m->saved_bi_end_io, overwrite_endio);
+		inc_all_io_entry(pool, bio);
+		remap_and_issue(tc, bio, data_block);
+	} else {
+		int r;
+		struct dm_io_region to;
+
+		to.bdev = tc->pool_dev->bdev;
+		to.sector = data_block * pool->sectors_per_block;
+		to.count = pool->sectors_per_block;
+
+		r = dm_kcopyd_zero(pool->copier, 1, &to, 0, copy_complete, m);
+		if (r < 0) {
+			mempool_free(m, pool->mapping_pool);
+			DMERR_LIMIT("dm_kcopyd_zero() failed");
+			cell_error(pool, cell);
+		}
+	}
+}
+
+/*
+ * A non-zero return indicates read_only or fail_io mode.
+ * Many callers don't care about the return value.
+ */
+static int commit(struct pool *pool)
+{
+	int r;
+
+	if (get_pool_mode(pool) >= PM_READ_ONLY)
+		return -EINVAL;
+
+	r = dm_pool_commit_metadata(pool->pmd);
+	if (r)
+		metadata_operation_failed(pool, "dm_pool_commit_metadata", r);
+
+	return r;
+}
+
+static void check_low_water_mark(struct pool *pool, dm_block_t free_blocks)
+{
+	unsigned long flags;
+
+	if (free_blocks <= pool->low_water_blocks && !pool->low_water_triggered) {
+		DMWARN("%s: reached low water mark for data device: sending event.",
+		       dm_device_name(pool->pool_md));
+		spin_lock_irqsave(&pool->lock, flags);
+		pool->low_water_triggered = true;
+		spin_unlock_irqrestore(&pool->lock, flags);
+		dm_table_event(pool->ti->table);
+	}
+}
+
+static void set_pool_mode(struct pool *pool, enum pool_mode new_mode);
+
+static int alloc_data_block(struct thin_c *tc, dm_block_t *result)
+{
+	int r;
+	dm_block_t free_blocks;
+	struct pool *pool = tc->pool;
+
+	if (WARN_ON(get_pool_mode(pool) != PM_WRITE))
+		return -EINVAL;
+
+	r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
+	if (r) {
+		metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);
+		return r;
+	}
+
+	check_low_water_mark(pool, free_blocks);
+
+	if (!free_blocks) {
+		/*
+		 * Try to commit to see if that will free up some
+		 * more space.
+		 */
+		r = commit(pool);
+		if (r)
+			return r;
+
+		r = dm_pool_get_free_block_count(pool->pmd, &free_blocks);
+		if (r) {
+			metadata_operation_failed(pool, "dm_pool_get_free_block_count", r);
+			return r;
+		}
+
+		if (!free_blocks) {
+			set_pool_mode(pool, PM_OUT_OF_DATA_SPACE);
+			return -ENOSPC;
+		}
+	}
+
+	r = dm_pool_alloc_data_block(pool->pmd, result);
+	if (r) {
+		metadata_operation_failed(pool, "dm_pool_alloc_data_block", r);
+		return r;
+	}
+
+	return 0;
+}
+
+/*
+ * If we have run out of space, queue bios until the device is
+ * resumed, presumably after having been reloaded with more space.
+ */
+static void retry_on_resume(struct bio *bio)
+{
+	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+	struct thin_c *tc = h->tc;
+	unsigned long flags;
+
+	spin_lock_irqsave(&tc->lock, flags);
+	bio_list_add(&tc->retry_on_resume_list, bio);
+	spin_unlock_irqrestore(&tc->lock, flags);
+}
+
+static int should_error_unserviceable_bio(struct pool *pool)
+{
+	enum pool_mode m = get_pool_mode(pool);
+
+	switch (m) {
+	case PM_WRITE:
+		/* Shouldn't get here */
+		DMERR_LIMIT("bio unserviceable, yet pool is in PM_WRITE mode");
+		return -EIO;
+
+	case PM_OUT_OF_DATA_SPACE:
+		return pool->pf.error_if_no_space ? -ENOSPC : 0;
+
+	case PM_READ_ONLY:
+	case PM_FAIL:
+		return -EIO;
+	default:
+		/* Shouldn't get here */
+		DMERR_LIMIT("bio unserviceable, yet pool has an unknown mode");
+		return -EIO;
+	}
+}
+
+static void handle_unserviceable_bio(struct pool *pool, struct bio *bio)
+{
+	int error = should_error_unserviceable_bio(pool);
+
+	if (error)
+		bio_endio(bio, error);
+	else
+		retry_on_resume(bio);
+}
+
+static void retry_bios_on_resume(struct pool *pool, struct dm_bio_prison_cell *cell)
+{
+	struct bio *bio;
+	struct bio_list bios;
+	int error;
+
+	error = should_error_unserviceable_bio(pool);
+	if (error) {
+		cell_error_with_code(pool, cell, error);
+		return;
+	}
+
+	bio_list_init(&bios);
+	cell_release(pool, cell, &bios);
+
+	error = should_error_unserviceable_bio(pool);
+	if (error)
+		while ((bio = bio_list_pop(&bios)))
+			bio_endio(bio, error);
+	else
+		while ((bio = bio_list_pop(&bios)))
+			retry_on_resume(bio);
+}
+
+static void process_discard(struct thin_c *tc, struct bio *bio)
+{
+	int r;
+	unsigned long flags;
+	struct pool *pool = tc->pool;
+	struct dm_bio_prison_cell *cell, *cell2;
+	struct dm_cell_key key, key2;
+	dm_block_t block = get_bio_block(tc, bio);
+	struct dm_thin_lookup_result lookup_result;
+	struct dm_thin_new_mapping *m;
+
+	build_virtual_key(tc->td, block, &key);
+	if (bio_detain(tc->pool, &key, bio, &cell))
+		return;
+
+	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
+	switch (r) {
+	case 0:
+		/*
+		 * Check nobody is fiddling with this pool block.  This can
+		 * happen if someone's in the process of breaking sharing
+		 * on this block.
+		 */
+		build_data_key(tc->td, lookup_result.block, &key2);
+		if (bio_detain(tc->pool, &key2, bio, &cell2)) {
+			cell_defer_no_holder(tc, cell);
+			break;
+		}
+
+		if (io_overlaps_block(pool, bio)) {
+			/*
+			 * IO may still be going to the destination block.  We must
+			 * quiesce before we can do the removal.
+			 */
+			m = get_next_mapping(pool);
+			m->tc = tc;
+			m->pass_discard = pool->pf.discard_passdown;
+			m->definitely_not_shared = !lookup_result.shared;
+			m->virt_block = block;
+			m->data_block = lookup_result.block;
+			m->cell = cell;
+			m->cell2 = cell2;
+			m->bio = bio;
+
+			if (!dm_deferred_set_add_work(pool->all_io_ds, &m->list)) {
+				spin_lock_irqsave(&pool->lock, flags);
+				list_add_tail(&m->list, &pool->prepared_discards);
+				spin_unlock_irqrestore(&pool->lock, flags);
+				wake_worker(pool);
+			}
+		} else {
+			inc_all_io_entry(pool, bio);
+			cell_defer_no_holder(tc, cell);
+			cell_defer_no_holder(tc, cell2);
+
+			/*
+			 * The DM core makes sure that the discard doesn't span
+			 * a block boundary.  So we submit the discard of a
+			 * partial block appropriately.
+			 */
+			if ((!lookup_result.shared) && pool->pf.discard_passdown)
+				remap_and_issue(tc, bio, lookup_result.block);
+			else
+				bio_endio(bio, 0);
+		}
+		break;
+
+	case -ENODATA:
+		/*
+		 * It isn't provisioned, just forget it.
+		 */
+		cell_defer_no_holder(tc, cell);
+		bio_endio(bio, 0);
+		break;
+
+	default:
+		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
+			    __func__, r);
+		cell_defer_no_holder(tc, cell);
+		bio_io_error(bio);
+		break;
+	}
+}
+
+static void break_sharing(struct thin_c *tc, struct bio *bio, dm_block_t block,
+			  struct dm_cell_key *key,
+			  struct dm_thin_lookup_result *lookup_result,
+			  struct dm_bio_prison_cell *cell)
+{
+	int r;
+	dm_block_t data_block;
+	struct pool *pool = tc->pool;
+
+	r = alloc_data_block(tc, &data_block);
+	switch (r) {
+	case 0:
+		schedule_internal_copy(tc, block, lookup_result->block,
+				       data_block, cell, bio);
+		break;
+
+	case -ENOSPC:
+		retry_bios_on_resume(pool, cell);
+		break;
+
+	default:
+		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
+			    __func__, r);
+		cell_error(pool, cell);
+		break;
+	}
+}
+
+static void process_shared_bio(struct thin_c *tc, struct bio *bio,
+			       dm_block_t block,
+			       struct dm_thin_lookup_result *lookup_result)
+{
+	struct dm_bio_prison_cell *cell;
+	struct pool *pool = tc->pool;
+	struct dm_cell_key key;
+
+	/*
+	 * If cell is already occupied, then sharing is already in the process
+	 * of being broken so we have nothing further to do here.
+	 */
+	build_data_key(tc->td, lookup_result->block, &key);
+	if (bio_detain(pool, &key, bio, &cell))
+		return;
+
+	if (bio_data_dir(bio) == WRITE && bio->bi_iter.bi_size)
+		break_sharing(tc, bio, block, &key, lookup_result, cell);
+	else {
+		struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+
+		h->shared_read_entry = dm_deferred_entry_inc(pool->shared_read_ds);
+		inc_all_io_entry(pool, bio);
+		cell_defer_no_holder(tc, cell);
+
+		remap_and_issue(tc, bio, lookup_result->block);
+	}
+}
+
+static void provision_block(struct thin_c *tc, struct bio *bio, dm_block_t block,
+			    struct dm_bio_prison_cell *cell)
+{
+	int r;
+	dm_block_t data_block;
+	struct pool *pool = tc->pool;
+
+	/*
+	 * Remap empty bios (flushes) immediately, without provisioning.
+	 */
+	if (!bio->bi_iter.bi_size) {
+		inc_all_io_entry(pool, bio);
+		cell_defer_no_holder(tc, cell);
+
+		remap_and_issue(tc, bio, 0);
+		return;
+	}
+
+	/*
+	 * Fill read bios with zeroes and complete them immediately.
+	 */
+	if (bio_data_dir(bio) == READ) {
+		zero_fill_bio(bio);
+		cell_defer_no_holder(tc, cell);
+		bio_endio(bio, 0);
+		return;
+	}
+
+	r = alloc_data_block(tc, &data_block);
+	switch (r) {
+	case 0:
+		if (tc->origin_dev)
+			schedule_external_copy(tc, block, data_block, cell, bio);
+		else
+			schedule_zero(tc, block, data_block, cell, bio);
+		break;
+
+	case -ENOSPC:
+		retry_bios_on_resume(pool, cell);
+		break;
+
+	default:
+		DMERR_LIMIT("%s: alloc_data_block() failed: error = %d",
+			    __func__, r);
+		cell_error(pool, cell);
+		break;
+	}
+}
+
+static void process_bio(struct thin_c *tc, struct bio *bio)
+{
+	int r;
+	struct pool *pool = tc->pool;
+	dm_block_t block = get_bio_block(tc, bio);
+	struct dm_bio_prison_cell *cell;
+	struct dm_cell_key key;
+	struct dm_thin_lookup_result lookup_result;
+
+	/*
+	 * If cell is already occupied, then the block is already
+	 * being provisioned so we have nothing further to do here.
+	 */
+	build_virtual_key(tc->td, block, &key);
+	if (bio_detain(pool, &key, bio, &cell))
+		return;
+
+	r = dm_thin_find_block(tc->td, block, 1, &lookup_result);
+	switch (r) {
+	case 0:
+		if (lookup_result.shared) {
+			process_shared_bio(tc, bio, block, &lookup_result);
+			cell_defer_no_holder(tc, cell); /* FIXME: pass this cell into process_shared? */
+		} else {
+			inc_all_io_entry(pool, bio);
+			cell_defer_no_holder(tc, cell);
+
+			remap_and_issue(tc, bio, lookup_result.block);
+		}
+		break;
+
+	case -ENODATA:
+		if (bio_data_dir(bio) == READ && tc->origin_dev) {
+			inc_all_io_entry(pool, bio);
+			cell_defer_no_holder(tc, cell);
+
+			remap_to_origin_and_issue(tc, bio);
+		} else
+			provision_block(tc, bio, block, cell);
+		break;
+
+	default:
+		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
+			    __func__, r);
+		cell_defer_no_holder(tc, cell);
+		bio_io_error(bio);
+		break;
+	}
+}
+
+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_iter.bi_size)
+			handle_unserviceable_bio(tc->pool, bio);
+		else {
+			inc_all_io_entry(tc->pool, bio);
+			remap_and_issue(tc, bio, lookup_result.block);
+		}
+		break;
+
+	case -ENODATA:
+		if (rw != READ) {
+			handle_unserviceable_bio(tc->pool, bio);
+			break;
+		}
+
+		if (tc->origin_dev) {
+			inc_all_io_entry(tc->pool, bio);
+			remap_to_origin_and_issue(tc, bio);
+			break;
+		}
+
+		zero_fill_bio(bio);
+		bio_endio(bio, 0);
+		break;
+
+	default:
+		DMERR_LIMIT("%s: dm_thin_find_block() failed: error = %d",
+			    __func__, r);
+		bio_io_error(bio);
+		break;
+	}
+}
+
+static void process_bio_success(struct thin_c *tc, struct bio *bio)
+{
+	bio_endio(bio, 0);
+}
+
+static void process_bio_fail(struct thin_c *tc, struct bio *bio)
+{
+	bio_io_error(bio);
+}
+
+/*
+ * FIXME: should we also commit due to size of transaction, measured in
+ * metadata blocks?
+ */
+static int need_commit_due_to_time(struct pool *pool)
+{
+	return jiffies < pool->last_commit_jiffies ||
+	       jiffies > pool->last_commit_jiffies + COMMIT_PERIOD;
+}
+
+#define thin_pbd(node) rb_entry((node), struct dm_thin_endio_hook, rb_node)
+#define thin_bio(pbd) dm_bio_from_per_bio_data((pbd), sizeof(struct dm_thin_endio_hook))
+
+static void __thin_bio_rb_add(struct thin_c *tc, struct bio *bio)
+{
+	struct rb_node **rbp, *parent;
+	struct dm_thin_endio_hook *pbd;
+	sector_t bi_sector = bio->bi_iter.bi_sector;
+
+	rbp = &tc->sort_bio_list.rb_node;
+	parent = NULL;
+	while (*rbp) {
+		parent = *rbp;
+		pbd = thin_pbd(parent);
+
+		if (bi_sector < thin_bio(pbd)->bi_iter.bi_sector)
+			rbp = &(*rbp)->rb_left;
+		else
+			rbp = &(*rbp)->rb_right;
+	}
+
+	pbd = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+	rb_link_node(&pbd->rb_node, parent, rbp);
+	rb_insert_color(&pbd->rb_node, &tc->sort_bio_list);
+}
+
+static void __extract_sorted_bios(struct thin_c *tc)
+{
+	struct rb_node *node;
+	struct dm_thin_endio_hook *pbd;
+	struct bio *bio;
+
+	for (node = rb_first(&tc->sort_bio_list); node; node = rb_next(node)) {
+		pbd = thin_pbd(node);
+		bio = thin_bio(pbd);
+
+		bio_list_add(&tc->deferred_bio_list, bio);
+		rb_erase(&pbd->rb_node, &tc->sort_bio_list);
+	}
+
+	WARN_ON(!RB_EMPTY_ROOT(&tc->sort_bio_list));
+}
+
+static void __sort_thin_deferred_bios(struct thin_c *tc)
+{
+	struct bio *bio;
+	struct bio_list bios;
+
+	bio_list_init(&bios);
+	bio_list_merge(&bios, &tc->deferred_bio_list);
+	bio_list_init(&tc->deferred_bio_list);
+
+	/* Sort deferred_bio_list using rb-tree */
+	while ((bio = bio_list_pop(&bios)))
+		__thin_bio_rb_add(tc, bio);
+
+	/*
+	 * Transfer the sorted bios in sort_bio_list back to
+	 * deferred_bio_list to allow lockless submission of
+	 * all bios.
+	 */
+	__extract_sorted_bios(tc);
+}
+
+static void process_thin_deferred_bios(struct thin_c *tc)
+{
+	struct pool *pool = tc->pool;
+	unsigned long flags;
+	struct bio *bio;
+	struct bio_list bios;
+	struct blk_plug plug;
+
+	if (tc->requeue_mode) {
+		requeue_bio_list(tc, &tc->deferred_bio_list);
+		return;
+	}
+
+	bio_list_init(&bios);
+
+	spin_lock_irqsave(&tc->lock, flags);
+
+	if (bio_list_empty(&tc->deferred_bio_list)) {
+		spin_unlock_irqrestore(&tc->lock, flags);
+		return;
+	}
+
+	__sort_thin_deferred_bios(tc);
+
+	bio_list_merge(&bios, &tc->deferred_bio_list);
+	bio_list_init(&tc->deferred_bio_list);
+
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	blk_start_plug(&plug);
+	while ((bio = bio_list_pop(&bios))) {
+		/*
+		 * If we've got no free new_mapping structs, and processing
+		 * this bio might require one, we pause until there are some
+		 * prepared mappings to process.
+		 */
+		if (ensure_next_mapping(pool)) {
+			spin_lock_irqsave(&tc->lock, flags);
+			bio_list_add(&tc->deferred_bio_list, bio);
+			bio_list_merge(&tc->deferred_bio_list, &bios);
+			spin_unlock_irqrestore(&tc->lock, flags);
+			break;
+		}
+
+		if (bio->bi_rw & REQ_DISCARD)
+			pool->process_discard(tc, bio);
+		else
+			pool->process_bio(tc, bio);
+	}
+	blk_finish_plug(&plug);
+}
+
+static void thin_get(struct thin_c *tc);
+static void thin_put(struct thin_c *tc);
+
+/*
+ * We can't hold rcu_read_lock() around code that can block.  So we
+ * find a thin with the rcu lock held; bump a refcount; then drop
+ * the lock.
+ */
+static struct thin_c *get_first_thin(struct pool *pool)
+{
+	struct thin_c *tc = NULL;
+
+	rcu_read_lock();
+	if (!list_empty(&pool->active_thins)) {
+		tc = list_entry_rcu(pool->active_thins.next, struct thin_c, list);
+		thin_get(tc);
+	}
+	rcu_read_unlock();
+
+	return tc;
+}
+
+static struct thin_c *get_next_thin(struct pool *pool, struct thin_c *tc)
+{
+	struct thin_c *old_tc = tc;
+
+	rcu_read_lock();
+	list_for_each_entry_continue_rcu(tc, &pool->active_thins, list) {
+		thin_get(tc);
+		thin_put(old_tc);
+		rcu_read_unlock();
+		return tc;
+	}
+	thin_put(old_tc);
+	rcu_read_unlock();
+
+	return NULL;
+}
+
+static void process_deferred_bios(struct pool *pool)
+{
+	unsigned long flags;
+	struct bio *bio;
+	struct bio_list bios;
+	struct thin_c *tc;
+
+	tc = get_first_thin(pool);
+	while (tc) {
+		process_thin_deferred_bios(tc);
+		tc = get_next_thin(pool, tc);
+	}
+
+	/*
+	 * If there are any deferred flush bios, we must commit
+	 * the metadata before issuing them.
+	 */
+	bio_list_init(&bios);
+	spin_lock_irqsave(&pool->lock, flags);
+	bio_list_merge(&bios, &pool->deferred_flush_bios);
+	bio_list_init(&pool->deferred_flush_bios);
+	spin_unlock_irqrestore(&pool->lock, flags);
+
+	if (bio_list_empty(&bios) &&
+	    !(dm_pool_changed_this_transaction(pool->pmd) && need_commit_due_to_time(pool)))
+		return;
+
+	if (commit(pool)) {
+		while ((bio = bio_list_pop(&bios)))
+			bio_io_error(bio);
+		return;
+	}
+	pool->last_commit_jiffies = jiffies;
+
+	while ((bio = bio_list_pop(&bios)))
+		generic_make_request(bio);
+}
+
+static void do_worker(struct work_struct *ws)
+{
+	struct pool *pool = container_of(ws, struct pool, worker);
+
+	process_prepared(pool, &pool->prepared_mappings, &pool->process_prepared_mapping);
+	process_prepared(pool, &pool->prepared_discards, &pool->process_prepared_discard);
+	process_deferred_bios(pool);
+}
+
+/*
+ * We want to commit periodically so that not too much
+ * unwritten data builds up.
+ */
+static void do_waker(struct work_struct *ws)
+{
+	struct pool *pool = container_of(to_delayed_work(ws), struct pool, waker);
+	wake_worker(pool);
+	queue_delayed_work(pool->wq, &pool->waker, COMMIT_PERIOD);
+}
+
+/*
+ * We're holding onto IO to allow userland time to react.  After the
+ * timeout either the pool will have been resized (and thus back in
+ * PM_WRITE mode), or we degrade to PM_READ_ONLY and start erroring IO.
+ */
+static void do_no_space_timeout(struct work_struct *ws)
+{
+	struct pool *pool = container_of(to_delayed_work(ws), struct pool,
+					 no_space_timeout);
+
+	if (get_pool_mode(pool) == PM_OUT_OF_DATA_SPACE && !pool->pf.error_if_no_space)
+		set_pool_mode(pool, PM_READ_ONLY);
+}
+
+/*----------------------------------------------------------------*/
+
+struct pool_work {
+	struct work_struct worker;
+	struct completion complete;
+};
+
+static struct pool_work *to_pool_work(struct work_struct *ws)
+{
+	return container_of(ws, struct pool_work, worker);
+}
+
+static void pool_work_complete(struct pool_work *pw)
+{
+	complete(&pw->complete);
+}
+
+static void pool_work_wait(struct pool_work *pw, struct pool *pool,
+			   void (*fn)(struct work_struct *))
+{
+	INIT_WORK_ONSTACK(&pw->worker, fn);
+	init_completion(&pw->complete);
+	queue_work(pool->wq, &pw->worker);
+	wait_for_completion(&pw->complete);
+}
+
+/*----------------------------------------------------------------*/
+
+struct noflush_work {
+	struct pool_work pw;
+	struct thin_c *tc;
+};
+
+static struct noflush_work *to_noflush(struct work_struct *ws)
+{
+	return container_of(to_pool_work(ws), struct noflush_work, pw);
+}
+
+static void do_noflush_start(struct work_struct *ws)
+{
+	struct noflush_work *w = to_noflush(ws);
+	w->tc->requeue_mode = true;
+	requeue_io(w->tc);
+	pool_work_complete(&w->pw);
+}
+
+static void do_noflush_stop(struct work_struct *ws)
+{
+	struct noflush_work *w = to_noflush(ws);
+	w->tc->requeue_mode = false;
+	pool_work_complete(&w->pw);
+}
+
+static void noflush_work(struct thin_c *tc, void (*fn)(struct work_struct *))
+{
+	struct noflush_work w;
+
+	w.tc = tc;
+	pool_work_wait(&w.pw, tc->pool, fn);
+}
+
+/*----------------------------------------------------------------*/
+
+static enum pool_mode get_pool_mode(struct pool *pool)
+{
+	return pool->pf.mode;
+}
+
+static void notify_of_pool_mode_change(struct pool *pool, const char *new_mode)
+{
+	dm_table_event(pool->ti->table);
+	DMINFO("%s: switching pool to %s mode",
+	       dm_device_name(pool->pool_md), new_mode);
+}
+
+static void set_pool_mode(struct pool *pool, enum pool_mode new_mode)
+{
+	struct pool_c *pt = pool->ti->private;
+	bool needs_check = dm_pool_metadata_needs_check(pool->pmd);
+	enum pool_mode old_mode = get_pool_mode(pool);
+	unsigned long no_space_timeout = ACCESS_ONCE(no_space_timeout_secs) * HZ;
+
+	/*
+	 * Never allow the pool to transition to PM_WRITE mode if user
+	 * intervention is required to verify metadata and data consistency.
+	 */
+	if (new_mode == PM_WRITE && needs_check) {
+		DMERR("%s: unable to switch pool to write mode until repaired.",
+		      dm_device_name(pool->pool_md));
+		if (old_mode != new_mode)
+			new_mode = old_mode;
+		else
+			new_mode = PM_READ_ONLY;
+	}
+	/*
+	 * If we were in PM_FAIL mode, rollback of metadata failed.  We're
+	 * not going to recover without a thin_repair.	So we never let the
+	 * pool move out of the old mode.
+	 */
+	if (old_mode == PM_FAIL)
+		new_mode = old_mode;
+
+	switch (new_mode) {
+	case PM_FAIL:
+		if (old_mode != new_mode)
+			notify_of_pool_mode_change(pool, "failure");
+		dm_pool_metadata_read_only(pool->pmd);
+		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;
+
+		error_retry_list(pool);
+		break;
+
+	case PM_READ_ONLY:
+		if (old_mode != new_mode)
+			notify_of_pool_mode_change(pool, "read-only");
+		dm_pool_metadata_read_only(pool->pmd);
+		pool->process_bio = process_bio_read_only;
+		pool->process_discard = process_bio_success;
+		pool->process_prepared_mapping = process_prepared_mapping_fail;
+		pool->process_prepared_discard = process_prepared_discard_passdown;
+
+		error_retry_list(pool);
+		break;
+
+	case PM_OUT_OF_DATA_SPACE:
+		/*
+		 * Ideally we'd never hit this state; the low water mark
+		 * would trigger userland to extend the pool before we
+		 * completely run out of data space.  However, many small
+		 * IOs to unprovisioned space can consume data space at an
+		 * alarming rate.  Adjust your low water mark if you're
+		 * frequently seeing this mode.
+		 */
+		if (old_mode != new_mode)
+			notify_of_pool_mode_change(pool, "out-of-data-space");
+		pool->process_bio = process_bio_read_only;
+		pool->process_discard = process_discard;
+		pool->process_prepared_mapping = process_prepared_mapping;
+		pool->process_prepared_discard = process_prepared_discard_passdown;
+
+		if (!pool->pf.error_if_no_space && no_space_timeout)
+			queue_delayed_work(pool->wq, &pool->no_space_timeout, no_space_timeout);
+		break;
+
+	case PM_WRITE:
+		if (old_mode != new_mode)
+			notify_of_pool_mode_change(pool, "write");
+		dm_pool_metadata_read_write(pool->pmd);
+		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;
+	}
+
+	pool->pf.mode = new_mode;
+	/*
+	 * The pool mode may have changed, sync it so bind_control_target()
+	 * doesn't cause an unexpected mode transition on resume.
+	 */
+	pt->adjusted_pf.mode = new_mode;
+}
+
+static void abort_transaction(struct pool *pool)
+{
+	const char *dev_name = dm_device_name(pool->pool_md);
+
+	DMERR_LIMIT("%s: aborting current metadata transaction", dev_name);
+	if (dm_pool_abort_metadata(pool->pmd)) {
+		DMERR("%s: failed to abort metadata transaction", dev_name);
+		set_pool_mode(pool, PM_FAIL);
+	}
+
+	if (dm_pool_metadata_set_needs_check(pool->pmd)) {
+		DMERR("%s: failed to set 'needs_check' flag in metadata", dev_name);
+		set_pool_mode(pool, PM_FAIL);
+	}
+}
+
+static void metadata_operation_failed(struct pool *pool, const char *op, int r)
+{
+	DMERR_LIMIT("%s: metadata operation '%s' failed: error = %d",
+		    dm_device_name(pool->pool_md), op, r);
+
+	abort_transaction(pool);
+	set_pool_mode(pool, PM_READ_ONLY);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Mapping functions.
+ */
+
+/*
+ * Called only while mapping a thin bio to hand it over to the workqueue.
+ */
+static void thin_defer_bio(struct thin_c *tc, struct bio *bio)
+{
+	unsigned long flags;
+	struct pool *pool = tc->pool;
+
+	spin_lock_irqsave(&tc->lock, flags);
+	bio_list_add(&tc->deferred_bio_list, bio);
+	spin_unlock_irqrestore(&tc->lock, flags);
+
+	wake_worker(pool);
+}
+
+static void thin_hook_bio(struct thin_c *tc, struct bio *bio)
+{
+	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+
+	h->tc = tc;
+	h->shared_read_entry = NULL;
+	h->all_io_entry = NULL;
+	h->overwrite_mapping = NULL;
+}
+
+/*
+ * Non-blocking function called from the thin target's map function.
+ */
+static int thin_bio_map(struct dm_target *ti, struct bio *bio)
+{
+	int r;
+	struct thin_c *tc = ti->private;
+	dm_block_t block = get_bio_block(tc, bio);
+	struct dm_thin_device *td = tc->td;
+	struct dm_thin_lookup_result result;
+	struct dm_bio_prison_cell cell1, cell2;
+	struct dm_bio_prison_cell *cell_result;
+	struct dm_cell_key key;
+
+	thin_hook_bio(tc, bio);
+
+	if (tc->requeue_mode) {
+		bio_endio(bio, DM_ENDIO_REQUEUE);
+		return DM_MAPIO_SUBMITTED;
+	}
+
+	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;
+	}
+
+	r = dm_thin_find_block(td, block, 0, &result);
+
+	/*
+	 * Note that we defer readahead too.
+	 */
+	switch (r) {
+	case 0:
+		if (unlikely(result.shared)) {
+			/*
+			 * We have a race condition here between the
+			 * result.shared value returned by the lookup and
+			 * snapshot creation, which may cause new
+			 * sharing.
+			 *
+			 * To avoid this always quiesce the origin before
+			 * taking the snap.  You want to do this anyway to
+			 * ensure a consistent application view
+			 * (i.e. lockfs).
+			 *
+			 * More distant ancestors are irrelevant. The
+			 * shared flag will be set in their case.
+			 */
+			thin_defer_bio(tc, bio);
+			return DM_MAPIO_SUBMITTED;
+		}
+
+		build_virtual_key(tc->td, block, &key);
+		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell1, &cell_result))
+			return DM_MAPIO_SUBMITTED;
+
+		build_data_key(tc->td, result.block, &key);
+		if (dm_bio_detain(tc->pool->prison, &key, bio, &cell2, &cell_result)) {
+			cell_defer_no_holder_no_free(tc, &cell1);
+			return DM_MAPIO_SUBMITTED;
+		}
+
+		inc_all_io_entry(tc->pool, bio);
+		cell_defer_no_holder_no_free(tc, &cell2);
+		cell_defer_no_holder_no_free(tc, &cell1);
+
+		remap(tc, bio, result.block);
+		return DM_MAPIO_REMAPPED;
+
+	case -ENODATA:
+		if (get_pool_mode(tc->pool) == PM_READ_ONLY) {
+			/*
+			 * This block isn't provisioned, and we have no way
+			 * of doing so.
+			 */
+			handle_unserviceable_bio(tc->pool, bio);
+			return DM_MAPIO_SUBMITTED;
+		}
+		/* fall through */
+
+	case -EWOULDBLOCK:
+		/*
+		 * In future, the failed dm_thin_find_block above could
+		 * provide the hint to load the metadata into cache.
+		 */
+		thin_defer_bio(tc, bio);
+		return DM_MAPIO_SUBMITTED;
+
+	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);
+		return DM_MAPIO_SUBMITTED;
+	}
+}
+
+static int pool_is_congested(struct dm_target_callbacks *cb, int bdi_bits)
+{
+	struct pool_c *pt = container_of(cb, struct pool_c, callbacks);
+	struct request_queue *q;
+
+	if (get_pool_mode(pt->pool) == PM_OUT_OF_DATA_SPACE)
+		return 1;
+
+	q = bdev_get_queue(pt->data_dev->bdev);
+	return bdi_congested(&q->backing_dev_info, bdi_bits);
+}
+
+static void requeue_bios(struct pool *pool)
+{
+	unsigned long flags;
+	struct thin_c *tc;
+
+	rcu_read_lock();
+	list_for_each_entry_rcu(tc, &pool->active_thins, list) {
+		spin_lock_irqsave(&tc->lock, flags);
+		bio_list_merge(&tc->deferred_bio_list, &tc->retry_on_resume_list);
+		bio_list_init(&tc->retry_on_resume_list);
+		spin_unlock_irqrestore(&tc->lock, flags);
+	}
+	rcu_read_unlock();
+}
+
+/*----------------------------------------------------------------
+ * Binding of control targets to a pool object
+ *--------------------------------------------------------------*/
+static bool data_dev_supports_discard(struct pool_c *pt)
+{
+	struct request_queue *q = bdev_get_queue(pt->data_dev->bdev);
+
+	return q && blk_queue_discard(q);
+}
+
+static bool is_factor(sector_t block_size, uint32_t n)
+{
+	return !sector_div(block_size, n);
+}
+
+/*
+ * If discard_passdown was enabled verify that the data device
+ * supports discards.  Disable discard_passdown if not.
+ */
+static void disable_passdown_if_not_supported(struct pool_c *pt)
+{
+	struct pool *pool = pt->pool;
+	struct block_device *data_bdev = pt->data_dev->bdev;
+	struct queue_limits *data_limits = &bdev_get_queue(data_bdev)->limits;
+	sector_t block_size = pool->sectors_per_block << SECTOR_SHIFT;
+	const char *reason = NULL;
+	char buf[BDEVNAME_SIZE];
+
+	if (!pt->adjusted_pf.discard_passdown)
+		return;
+
+	if (!data_dev_supports_discard(pt))
+		reason = "discard unsupported";
+
+	else if (data_limits->max_discard_sectors < pool->sectors_per_block)
+		reason = "max discard sectors smaller than a block";
+
+	else if (data_limits->discard_granularity > block_size)
+		reason = "discard granularity larger than a block";
+
+	else if (!is_factor(block_size, data_limits->discard_granularity))
+		reason = "discard granularity not a factor of block size";
+
+	if (reason) {
+		DMWARN("Data device (%s) %s: Disabling discard passdown.", bdevname(data_bdev, buf), reason);
+		pt->adjusted_pf.discard_passdown = false;
+	}
+}
+
+static int bind_control_target(struct pool *pool, struct dm_target *ti)
+{
+	struct pool_c *pt = ti->private;
+
+	/*
+	 * We want to make sure that a pool in PM_FAIL mode is never upgraded.
+	 */
+	enum pool_mode old_mode = get_pool_mode(pool);
+	enum pool_mode new_mode = pt->adjusted_pf.mode;
+
+	/*
+	 * Don't change the pool's mode until set_pool_mode() below.
+	 * Otherwise the pool's process_* function pointers may
+	 * not match the desired pool mode.
+	 */
+	pt->adjusted_pf.mode = old_mode;
+
+	pool->ti = ti;
+	pool->pf = pt->adjusted_pf;
+	pool->low_water_blocks = pt->low_water_blocks;
+
+	set_pool_mode(pool, new_mode);
+
+	return 0;
+}
+
+static void unbind_control_target(struct pool *pool, struct dm_target *ti)
+{
+	if (pool->ti == ti)
+		pool->ti = NULL;
+}
+
+/*----------------------------------------------------------------
+ * Pool creation
+ *--------------------------------------------------------------*/
+/* Initialize pool features. */
+static void pool_features_init(struct pool_features *pf)
+{
+	pf->mode = PM_WRITE;
+	pf->zero_new_blocks = true;
+	pf->discard_enabled = true;
+	pf->discard_passdown = true;
+	pf->error_if_no_space = false;
+}
+
+static void __pool_destroy(struct pool *pool)
+{
+	__pool_table_remove(pool);
+
+	if (dm_pool_metadata_close(pool->pmd) < 0)
+		DMWARN("%s: dm_pool_metadata_close() failed.", __func__);
+
+	dm_bio_prison_destroy(pool->prison);
+	dm_kcopyd_client_destroy(pool->copier);
+
+	if (pool->wq)
+		destroy_workqueue(pool->wq);
+
+	if (pool->next_mapping)
+		mempool_free(pool->next_mapping, pool->mapping_pool);
+	mempool_destroy(pool->mapping_pool);
+	dm_deferred_set_destroy(pool->shared_read_ds);
+	dm_deferred_set_destroy(pool->all_io_ds);
+	kfree(pool);
+}
+
+static struct kmem_cache *_new_mapping_cache;
+
+static struct pool *pool_create(struct mapped_device *pool_md,
+				struct block_device *metadata_dev,
+				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, format_device);
+	if (IS_ERR(pmd)) {
+		*error = "Error creating metadata object";
+		return (struct pool *)pmd;
+	}
+
+	pool = kmalloc(sizeof(*pool), GFP_KERNEL);
+	if (!pool) {
+		*error = "Error allocating memory for pool";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_pool;
+	}
+
+	pool->pmd = pmd;
+	pool->sectors_per_block = block_size;
+	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 = dm_bio_prison_create(PRISON_CELLS);
+	if (!pool->prison) {
+		*error = "Error creating pool's bio prison";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_prison;
+	}
+
+	pool->copier = dm_kcopyd_client_create(&dm_kcopyd_throttle);
+	if (IS_ERR(pool->copier)) {
+		r = PTR_ERR(pool->copier);
+		*error = "Error creating pool's kcopyd client";
+		err_p = ERR_PTR(r);
+		goto bad_kcopyd_client;
+	}
+
+	/*
+	 * Create singlethreaded workqueue that will service all devices
+	 * that use this metadata.
+	 */
+	pool->wq = alloc_ordered_workqueue("dm-" DM_MSG_PREFIX, WQ_MEM_RECLAIM);
+	if (!pool->wq) {
+		*error = "Error creating pool's workqueue";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_wq;
+	}
+
+	INIT_WORK(&pool->worker, do_worker);
+	INIT_DELAYED_WORK(&pool->waker, do_waker);
+	INIT_DELAYED_WORK(&pool->no_space_timeout, do_no_space_timeout);
+	spin_lock_init(&pool->lock);
+	bio_list_init(&pool->deferred_flush_bios);
+	INIT_LIST_HEAD(&pool->prepared_mappings);
+	INIT_LIST_HEAD(&pool->prepared_discards);
+	INIT_LIST_HEAD(&pool->active_thins);
+	pool->low_water_triggered = false;
+
+	pool->shared_read_ds = dm_deferred_set_create();
+	if (!pool->shared_read_ds) {
+		*error = "Error creating pool's shared read deferred set";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_shared_read_ds;
+	}
+
+	pool->all_io_ds = dm_deferred_set_create();
+	if (!pool->all_io_ds) {
+		*error = "Error creating pool's all io deferred set";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_all_io_ds;
+	}
+
+	pool->next_mapping = NULL;
+	pool->mapping_pool = mempool_create_slab_pool(MAPPING_POOL_SIZE,
+						      _new_mapping_cache);
+	if (!pool->mapping_pool) {
+		*error = "Error creating pool's mapping mempool";
+		err_p = ERR_PTR(-ENOMEM);
+		goto bad_mapping_pool;
+	}
+
+	pool->ref_count = 1;
+	pool->last_commit_jiffies = jiffies;
+	pool->pool_md = pool_md;
+	pool->md_dev = metadata_dev;
+	__pool_table_insert(pool);
+
+	return pool;
+
+bad_mapping_pool:
+	dm_deferred_set_destroy(pool->all_io_ds);
+bad_all_io_ds:
+	dm_deferred_set_destroy(pool->shared_read_ds);
+bad_shared_read_ds:
+	destroy_workqueue(pool->wq);
+bad_wq:
+	dm_kcopyd_client_destroy(pool->copier);
+bad_kcopyd_client:
+	dm_bio_prison_destroy(pool->prison);
+bad_prison:
+	kfree(pool);
+bad_pool:
+	if (dm_pool_metadata_close(pmd))
+		DMWARN("%s: dm_pool_metadata_close() failed.", __func__);
+
+	return err_p;
+}
+
+static void __pool_inc(struct pool *pool)
+{
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+	pool->ref_count++;
+}
+
+static void __pool_dec(struct pool *pool)
+{
+	BUG_ON(!mutex_is_locked(&dm_thin_pool_table.mutex));
+	BUG_ON(!pool->ref_count);
+	if (!--pool->ref_count)
+		__pool_destroy(pool);
+}
+
+static struct pool *__pool_find(struct mapped_device *pool_md,
+				struct block_device *metadata_dev,
+				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) {
+			*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) {
+				*error = "different pool cannot replace a pool";
+				return ERR_PTR(-EINVAL);
+			}
+			__pool_inc(pool);
+
+		} else {
+			pool = pool_create(pool_md, metadata_dev, block_size, read_only, error);
+			*created = 1;
+		}
+	}
+
+	return pool;
+}
+
+/*----------------------------------------------------------------
+ * Pool target methods
+ *--------------------------------------------------------------*/
+static void pool_dtr(struct dm_target *ti)
+{
+	struct pool_c *pt = ti->private;
+
+	mutex_lock(&dm_thin_pool_table.mutex);
+
+	unbind_control_target(pt->pool, ti);
+	__pool_dec(pt->pool);
+	dm_put_device(ti, pt->metadata_dev);
+	dm_put_device(ti, pt->data_dev);
+	kfree(pt);
+
+	mutex_unlock(&dm_thin_pool_table.mutex);
+}
+
+static int parse_pool_features(struct dm_arg_set *as, struct pool_features *pf,
+			       struct dm_target *ti)
+{
+	int r;
+	unsigned argc;
+	const char *arg_name;
+
+	static struct dm_arg _args[] = {
+		{0, 4, "Invalid number of pool feature arguments"},
+	};
+
+	/*
+	 * No feature arguments supplied.
+	 */
+	if (!as->argc)
+		return 0;
+
+	r = dm_read_arg_group(_args, as, &argc, &ti->error);
+	if (r)
+		return -EINVAL;
+
+	while (argc && !r) {
+		arg_name = dm_shift_arg(as);
+		argc--;
+
+		if (!strcasecmp(arg_name, "skip_block_zeroing"))
+			pf->zero_new_blocks = false;
+
+		else if (!strcasecmp(arg_name, "ignore_discard"))
+			pf->discard_enabled = false;
+
+		else if (!strcasecmp(arg_name, "no_discard_passdown"))
+			pf->discard_passdown = false;
+
+		else if (!strcasecmp(arg_name, "read_only"))
+			pf->mode = PM_READ_ONLY;
+
+		else if (!strcasecmp(arg_name, "error_if_no_space"))
+			pf->error_if_no_space = true;
+
+		else {
+			ti->error = "Unrecognised pool feature requested";
+			r = -EINVAL;
+			break;
+		}
+	}
+
+	return r;
+}
+
+static void metadata_low_callback(void *context)
+{
+	struct pool *pool = context;
+
+	DMWARN("%s: reached low water mark for metadata device: sending event.",
+	       dm_device_name(pool->pool_md));
+
+	dm_table_event(pool->ti->table);
+}
+
+static sector_t get_dev_size(struct block_device *bdev)
+{
+	return i_size_read(bdev->bd_inode) >> SECTOR_SHIFT;
+}
+
+static void warn_if_metadata_device_too_big(struct block_device *bdev)
+{
+	sector_t metadata_dev_size = get_dev_size(bdev);
+	char buffer[BDEVNAME_SIZE];
+
+	if (metadata_dev_size > THIN_METADATA_MAX_SECTORS_WARNING)
+		DMWARN("Metadata device %s is larger than %u sectors: excess space will not be used.",
+		       bdevname(bdev, buffer), THIN_METADATA_MAX_SECTORS);
+}
+
+static sector_t get_metadata_dev_size(struct block_device *bdev)
+{
+	sector_t metadata_dev_size = get_dev_size(bdev);
+
+	if (metadata_dev_size > THIN_METADATA_MAX_SECTORS)
+		metadata_dev_size = THIN_METADATA_MAX_SECTORS;
+
+	return metadata_dev_size;
+}
+
+static dm_block_t get_metadata_dev_size_in_blocks(struct block_device *bdev)
+{
+	sector_t metadata_dev_size = get_metadata_dev_size(bdev);
+
+	sector_div(metadata_dev_size, THIN_METADATA_BLOCK_SIZE);
+
+	return metadata_dev_size;
+}
+
+/*
+ * When a metadata threshold is crossed a dm event is triggered, and
+ * userland should respond by growing the metadata device.  We could let
+ * userland set the threshold, like we do with the data threshold, but I'm
+ * not sure they know enough to do this well.
+ */
+static dm_block_t calc_metadata_threshold(struct pool_c *pt)
+{
+	/*
+	 * 4M is ample for all ops with the possible exception of thin
+	 * device deletion which is harmless if it fails (just retry the
+	 * delete after you've grown the device).
+	 */
+	dm_block_t quarter = get_metadata_dev_size_in_blocks(pt->metadata_dev->bdev) / 4;
+	return min((dm_block_t)1024ULL /* 4M */, quarter);
+}
+
+/*
+ * thin-pool <metadata dev> <data dev>
+ *	     <data block size (sectors)>
+ *	     <low water mark (blocks)>
+ *	     [<#feature args> [<arg>]*]
+ *
+ * Optional feature arguments are:
+ *	     skip_block_zeroing: skips the zeroing of newly-provisioned blocks.
+ *	     ignore_discard: disable discard
+ *	     no_discard_passdown: don't pass discards down to the data device
+ *	     read_only: Don't allow any changes to be made to the pool metadata.
+ *	     error_if_no_space: error IOs, instead of queueing, if no space.
+ */
+static int pool_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int r, pool_created = 0;
+	struct pool_c *pt;
+	struct pool *pool;
+	struct pool_features pf;
+	struct dm_arg_set as;
+	struct dm_dev *data_dev;
+	unsigned long block_size;
+	dm_block_t low_water_blocks;
+	struct dm_dev *metadata_dev;
+	fmode_t metadata_mode;
+
+	/*
+	 * FIXME Remove validation from scope of lock.
+	 */
+	mutex_lock(&dm_thin_pool_table.mutex);
+
+	if (argc < 4) {
+		ti->error = "Invalid argument count";
+		r = -EINVAL;
+		goto out_unlock;
+	}
+
+	as.argc = argc;
+	as.argv = argv;
+
+	/*
+	 * Set default pool features.
+	 */
+	pool_features_init(&pf);
+
+	dm_consume_args(&as, 4);
+	r = parse_pool_features(&as, &pf, ti);
+	if (r)
+		goto out_unlock;
+
+	metadata_mode = FMODE_READ | ((pf.mode == PM_READ_ONLY) ? 0 : FMODE_WRITE);
+	r = dm_get_device(ti, argv[0], metadata_mode, &metadata_dev);
+	if (r) {
+		ti->error = "Error opening metadata block device";
+		goto out_unlock;
+	}
+	warn_if_metadata_device_too_big(metadata_dev->bdev);
+
+	r = dm_get_device(ti, argv[1], FMODE_READ | FMODE_WRITE, &data_dev);
+	if (r) {
+		ti->error = "Error getting data device";
+		goto out_metadata;
+	}
+
+	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 ||
+	    block_size & (DATA_DEV_BLOCK_SIZE_MIN_SECTORS - 1)) {
+		ti->error = "Invalid block size";
+		r = -EINVAL;
+		goto out;
+	}
+
+	if (kstrtoull(argv[3], 10, (unsigned long long *)&low_water_blocks)) {
+		ti->error = "Invalid low water mark";
+		r = -EINVAL;
+		goto out;
+	}
+
+	pt = kzalloc(sizeof(*pt), GFP_KERNEL);
+	if (!pt) {
+		r = -ENOMEM;
+		goto out;
+	}
+
+	pool = __pool_find(dm_table_get_md(ti->table), metadata_dev->bdev,
+			   block_size, pf.mode == PM_READ_ONLY, &ti->error, &pool_created);
+	if (IS_ERR(pool)) {
+		r = PTR_ERR(pool);
+		goto out_free_pt;
+	}
+
+	/*
+	 * 'pool_created' reflects whether this is the first table load.
+	 * Top level discard support is not allowed to be changed after
+	 * initial load.  This would require a pool reload to trigger thin
+	 * device changes.
+	 */
+	if (!pool_created && pf.discard_enabled != pool->pf.discard_enabled) {
+		ti->error = "Discard support cannot be disabled once enabled";
+		r = -EINVAL;
+		goto out_flags_changed;
+	}
+
+	pt->pool = pool;
+	pt->ti = ti;
+	pt->metadata_dev = metadata_dev;
+	pt->data_dev = data_dev;
+	pt->low_water_blocks = low_water_blocks;
+	pt->adjusted_pf = pt->requested_pf = pf;
+	ti->num_flush_bios = 1;
+
+	/*
+	 * Only need to enable discards if the pool should pass
+	 * them down to the data device.  The thin device's discard
+	 * processing will cause mappings to be removed from the btree.
+	 */
+	ti->discard_zeroes_data_unsupported = true;
+	if (pf.discard_enabled && pf.discard_passdown) {
+		ti->num_discard_bios = 1;
+
+		/*
+		 * Setting 'discards_supported' circumvents the normal
+		 * stacking of discard limits (this keeps the pool and
+		 * thin devices' discard limits consistent).
+		 */
+		ti->discards_supported = true;
+	}
+	ti->private = pt;
+
+	r = dm_pool_register_metadata_threshold(pt->pool->pmd,
+						calc_metadata_threshold(pt),
+						metadata_low_callback,
+						pool);
+	if (r)
+		goto out_free_pt;
+
+	pt->callbacks.congested_fn = pool_is_congested;
+	dm_table_add_target_callbacks(ti->table, &pt->callbacks);
+
+	mutex_unlock(&dm_thin_pool_table.mutex);
+
+	return 0;
+
+out_flags_changed:
+	__pool_dec(pool);
+out_free_pt:
+	kfree(pt);
+out:
+	dm_put_device(ti, data_dev);
+out_metadata:
+	dm_put_device(ti, metadata_dev);
+out_unlock:
+	mutex_unlock(&dm_thin_pool_table.mutex);
+
+	return r;
+}
+
+static int pool_map(struct dm_target *ti, struct bio *bio)
+{
+	int r;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	unsigned long flags;
+
+	/*
+	 * As this is a singleton target, ti->begin is always zero.
+	 */
+	spin_lock_irqsave(&pool->lock, flags);
+	bio->bi_bdev = pt->data_dev->bdev;
+	r = DM_MAPIO_REMAPPED;
+	spin_unlock_irqrestore(&pool->lock, flags);
+
+	return r;
+}
+
+static int maybe_resize_data_dev(struct dm_target *ti, bool *need_commit)
+{
+	int r;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	sector_t data_size = ti->len;
+	dm_block_t sb_data_size;
+
+	*need_commit = false;
+
+	(void) sector_div(data_size, pool->sectors_per_block);
+
+	r = dm_pool_get_data_dev_size(pool->pmd, &sb_data_size);
+	if (r) {
+		DMERR("%s: failed to retrieve data device size",
+		      dm_device_name(pool->pool_md));
+		return r;
+	}
+
+	if (data_size < sb_data_size) {
+		DMERR("%s: pool target (%llu blocks) too small: expected %llu",
+		      dm_device_name(pool->pool_md),
+		      (unsigned long long)data_size, sb_data_size);
+		return -EINVAL;
+
+	} else if (data_size > sb_data_size) {
+		if (dm_pool_metadata_needs_check(pool->pmd)) {
+			DMERR("%s: unable to grow the data device until repaired.",
+			      dm_device_name(pool->pool_md));
+			return 0;
+		}
+
+		if (sb_data_size)
+			DMINFO("%s: growing the data device from %llu to %llu blocks",
+			       dm_device_name(pool->pool_md),
+			       sb_data_size, (unsigned long long)data_size);
+		r = dm_pool_resize_data_dev(pool->pmd, data_size);
+		if (r) {
+			metadata_operation_failed(pool, "dm_pool_resize_data_dev", r);
+			return r;
+		}
+
+		*need_commit = true;
+	}
+
+	return 0;
+}
+
+static int maybe_resize_metadata_dev(struct dm_target *ti, bool *need_commit)
+{
+	int r;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	dm_block_t metadata_dev_size, sb_metadata_dev_size;
+
+	*need_commit = false;
+
+	metadata_dev_size = get_metadata_dev_size_in_blocks(pool->md_dev);
+
+	r = dm_pool_get_metadata_dev_size(pool->pmd, &sb_metadata_dev_size);
+	if (r) {
+		DMERR("%s: failed to retrieve metadata device size",
+		      dm_device_name(pool->pool_md));
+		return r;
+	}
+
+	if (metadata_dev_size < sb_metadata_dev_size) {
+		DMERR("%s: metadata device (%llu blocks) too small: expected %llu",
+		      dm_device_name(pool->pool_md),
+		      metadata_dev_size, sb_metadata_dev_size);
+		return -EINVAL;
+
+	} else if (metadata_dev_size > sb_metadata_dev_size) {
+		if (dm_pool_metadata_needs_check(pool->pmd)) {
+			DMERR("%s: unable to grow the metadata device until repaired.",
+			      dm_device_name(pool->pool_md));
+			return 0;
+		}
+
+		warn_if_metadata_device_too_big(pool->md_dev);
+		DMINFO("%s: growing the metadata device from %llu to %llu blocks",
+		       dm_device_name(pool->pool_md),
+		       sb_metadata_dev_size, metadata_dev_size);
+		r = dm_pool_resize_metadata_dev(pool->pmd, metadata_dev_size);
+		if (r) {
+			metadata_operation_failed(pool, "dm_pool_resize_metadata_dev", r);
+			return r;
+		}
+
+		*need_commit = true;
+	}
+
+	return 0;
+}
+
+/*
+ * Retrieves the number of blocks of the data device from
+ * the superblock and compares it to the actual device size,
+ * thus resizing the data device in case it has grown.
+ *
+ * This both copes with opening preallocated data devices in the ctr
+ * being followed by a resume
+ * -and-
+ * calling the resume method individually after userspace has
+ * grown the data device in reaction to a table event.
+ */
+static int pool_preresume(struct dm_target *ti)
+{
+	int r;
+	bool need_commit1, need_commit2;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+
+	/*
+	 * Take control of the pool object.
+	 */
+	r = bind_control_target(pool, ti);
+	if (r)
+		return r;
+
+	r = maybe_resize_data_dev(ti, &need_commit1);
+	if (r)
+		return r;
+
+	r = maybe_resize_metadata_dev(ti, &need_commit2);
+	if (r)
+		return r;
+
+	if (need_commit1 || need_commit2)
+		(void) commit(pool);
+
+	return 0;
+}
+
+static void pool_resume(struct dm_target *ti)
+{
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	unsigned long flags;
+
+	spin_lock_irqsave(&pool->lock, flags);
+	pool->low_water_triggered = false;
+	spin_unlock_irqrestore(&pool->lock, flags);
+	requeue_bios(pool);
+
+	do_waker(&pool->waker.work);
+}
+
+static void pool_postsuspend(struct dm_target *ti)
+{
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+
+	cancel_delayed_work(&pool->waker);
+	cancel_delayed_work(&pool->no_space_timeout);
+	flush_workqueue(pool->wq);
+	(void) commit(pool);
+}
+
+static int check_arg_count(unsigned argc, unsigned args_required)
+{
+	if (argc != args_required) {
+		DMWARN("Message received with %u arguments instead of %u.",
+		       argc, args_required);
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
+static int read_dev_id(char *arg, dm_thin_id *dev_id, int warning)
+{
+	if (!kstrtoull(arg, 10, (unsigned long long *)dev_id) &&
+	    *dev_id <= MAX_DEV_ID)
+		return 0;
+
+	if (warning)
+		DMWARN("Message received with invalid device id: %s", arg);
+
+	return -EINVAL;
+}
+
+static int process_create_thin_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	dm_thin_id dev_id;
+	int r;
+
+	r = check_arg_count(argc, 2);
+	if (r)
+		return r;
+
+	r = read_dev_id(argv[1], &dev_id, 1);
+	if (r)
+		return r;
+
+	r = dm_pool_create_thin(pool->pmd, dev_id);
+	if (r) {
+		DMWARN("Creation of new thinly-provisioned device with id %s failed.",
+		       argv[1]);
+		return r;
+	}
+
+	return 0;
+}
+
+static int process_create_snap_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	dm_thin_id dev_id;
+	dm_thin_id origin_dev_id;
+	int r;
+
+	r = check_arg_count(argc, 3);
+	if (r)
+		return r;
+
+	r = read_dev_id(argv[1], &dev_id, 1);
+	if (r)
+		return r;
+
+	r = read_dev_id(argv[2], &origin_dev_id, 1);
+	if (r)
+		return r;
+
+	r = dm_pool_create_snap(pool->pmd, dev_id, origin_dev_id);
+	if (r) {
+		DMWARN("Creation of new snapshot %s of device %s failed.",
+		       argv[1], argv[2]);
+		return r;
+	}
+
+	return 0;
+}
+
+static int process_delete_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	dm_thin_id dev_id;
+	int r;
+
+	r = check_arg_count(argc, 2);
+	if (r)
+		return r;
+
+	r = read_dev_id(argv[1], &dev_id, 1);
+	if (r)
+		return r;
+
+	r = dm_pool_delete_thin_device(pool->pmd, dev_id);
+	if (r)
+		DMWARN("Deletion of thin device %s failed.", argv[1]);
+
+	return r;
+}
+
+static int process_set_transaction_id_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	dm_thin_id old_id, new_id;
+	int r;
+
+	r = check_arg_count(argc, 3);
+	if (r)
+		return r;
+
+	if (kstrtoull(argv[1], 10, (unsigned long long *)&old_id)) {
+		DMWARN("set_transaction_id message: Unrecognised id %s.", argv[1]);
+		return -EINVAL;
+	}
+
+	if (kstrtoull(argv[2], 10, (unsigned long long *)&new_id)) {
+		DMWARN("set_transaction_id message: Unrecognised new id %s.", argv[2]);
+		return -EINVAL;
+	}
+
+	r = dm_pool_set_metadata_transaction_id(pool->pmd, old_id, new_id);
+	if (r) {
+		DMWARN("Failed to change transaction id from %s to %s.",
+		       argv[1], argv[2]);
+		return r;
+	}
+
+	return 0;
+}
+
+static int process_reserve_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	int r;
+
+	r = check_arg_count(argc, 1);
+	if (r)
+		return r;
+
+	(void) commit(pool);
+
+	r = dm_pool_reserve_metadata_snap(pool->pmd);
+	if (r)
+		DMWARN("reserve_metadata_snap message failed.");
+
+	return r;
+}
+
+static int process_release_metadata_snap_mesg(unsigned argc, char **argv, struct pool *pool)
+{
+	int r;
+
+	r = check_arg_count(argc, 1);
+	if (r)
+		return r;
+
+	r = dm_pool_release_metadata_snap(pool->pmd);
+	if (r)
+		DMWARN("release_metadata_snap message failed.");
+
+	return r;
+}
+
+/*
+ * Messages supported:
+ *   create_thin	<dev_id>
+ *   create_snap	<dev_id> <origin_id>
+ *   delete		<dev_id>
+ *   trim		<dev_id> <new_size_in_sectors>
+ *   set_transaction_id <current_trans_id> <new_trans_id>
+ *   reserve_metadata_snap
+ *   release_metadata_snap
+ */
+static int pool_message(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int r = -EINVAL;
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+
+	if (!strcasecmp(argv[0], "create_thin"))
+		r = process_create_thin_mesg(argc, argv, pool);
+
+	else if (!strcasecmp(argv[0], "create_snap"))
+		r = process_create_snap_mesg(argc, argv, pool);
+
+	else if (!strcasecmp(argv[0], "delete"))
+		r = process_delete_mesg(argc, argv, pool);
+
+	else if (!strcasecmp(argv[0], "set_transaction_id"))
+		r = process_set_transaction_id_mesg(argc, argv, pool);
+
+	else if (!strcasecmp(argv[0], "reserve_metadata_snap"))
+		r = process_reserve_metadata_snap_mesg(argc, argv, pool);
+
+	else if (!strcasecmp(argv[0], "release_metadata_snap"))
+		r = process_release_metadata_snap_mesg(argc, argv, pool);
+
+	else
+		DMWARN("Unrecognised thin pool target message received: %s", argv[0]);
+
+	if (!r)
+		(void) commit(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) +
+		pf->error_if_no_space;
+	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 ");
+
+	if (pf->error_if_no_space)
+		DMEMIT("error_if_no_space ");
+}
+
+/*
+ * Status line is:
+ *    <transaction id> <used metadata sectors>/<total metadata sectors>
+ *    <used data sectors>/<total data sectors> <held metadata root>
+ */
+static void pool_status(struct dm_target *ti, status_type_t type,
+			unsigned status_flags, char *result, unsigned maxlen)
+{
+	int r;
+	unsigned sz = 0;
+	uint64_t transaction_id;
+	dm_block_t nr_free_blocks_data;
+	dm_block_t nr_free_blocks_metadata;
+	dm_block_t nr_blocks_data;
+	dm_block_t nr_blocks_metadata;
+	dm_block_t held_root;
+	char buf[BDEVNAME_SIZE];
+	char buf2[BDEVNAME_SIZE];
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+
+	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(pool);
+
+		r = dm_pool_get_metadata_transaction_id(pool->pmd, &transaction_id);
+		if (r) {
+			DMERR("%s: dm_pool_get_metadata_transaction_id returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		r = dm_pool_get_free_metadata_block_count(pool->pmd, &nr_free_blocks_metadata);
+		if (r) {
+			DMERR("%s: dm_pool_get_free_metadata_block_count returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		r = dm_pool_get_metadata_dev_size(pool->pmd, &nr_blocks_metadata);
+		if (r) {
+			DMERR("%s: dm_pool_get_metadata_dev_size returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		r = dm_pool_get_free_block_count(pool->pmd, &nr_free_blocks_data);
+		if (r) {
+			DMERR("%s: dm_pool_get_free_block_count returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		r = dm_pool_get_data_dev_size(pool->pmd, &nr_blocks_data);
+		if (r) {
+			DMERR("%s: dm_pool_get_data_dev_size returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		r = dm_pool_get_metadata_snap(pool->pmd, &held_root);
+		if (r) {
+			DMERR("%s: dm_pool_get_metadata_snap returned %d",
+			      dm_device_name(pool->pool_md), r);
+			goto err;
+		}
+
+		DMEMIT("%llu %llu/%llu %llu/%llu ",
+		       (unsigned long long)transaction_id,
+		       (unsigned long long)(nr_blocks_metadata - nr_free_blocks_metadata),
+		       (unsigned long long)nr_blocks_metadata,
+		       (unsigned long long)(nr_blocks_data - nr_free_blocks_data),
+		       (unsigned long long)nr_blocks_data);
+
+		if (held_root)
+			DMEMIT("%llu ", held_root);
+		else
+			DMEMIT("- ");
+
+		if (pool->pf.mode == PM_OUT_OF_DATA_SPACE)
+			DMEMIT("out_of_data_space ");
+		else if (pool->pf.mode == PM_READ_ONLY)
+			DMEMIT("ro ");
+		else
+			DMEMIT("rw ");
+
+		if (!pool->pf.discard_enabled)
+			DMEMIT("ignore_discard ");
+		else if (pool->pf.discard_passdown)
+			DMEMIT("discard_passdown ");
+		else
+			DMEMIT("no_discard_passdown ");
+
+		if (pool->pf.error_if_no_space)
+			DMEMIT("error_if_no_space ");
+		else
+			DMEMIT("queue_if_no_space ");
+
+		break;
+
+	case STATUSTYPE_TABLE:
+		DMEMIT("%s %s %lu %llu ",
+		       format_dev_t(buf, pt->metadata_dev->bdev->bd_dev),
+		       format_dev_t(buf2, pt->data_dev->bdev->bd_dev),
+		       (unsigned long)pool->sectors_per_block,
+		       (unsigned long long)pt->low_water_blocks);
+		emit_flags(&pt->requested_pf, result, sz, maxlen);
+		break;
+	}
+	return;
+
+err:
+	DMEMIT("Error");
+}
+
+static int pool_iterate_devices(struct dm_target *ti,
+				iterate_devices_callout_fn fn, void *data)
+{
+	struct pool_c *pt = ti->private;
+
+	return fn(ti, pt->data_dev, 0, ti->len, data);
+}
+
+static int pool_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+		      struct bio_vec *biovec, int max_size)
+{
+	struct pool_c *pt = ti->private;
+	struct request_queue *q = bdev_get_queue(pt->data_dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = pt->data_dev->bdev;
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static void set_discard_limits(struct pool_c *pt, struct queue_limits *limits)
+{
+	struct pool *pool = pt->pool;
+	struct queue_limits *data_limits;
+
+	limits->max_discard_sectors = pool->sectors_per_block;
+
+	/*
+	 * discard_granularity is just a hint, and not enforced.
+	 */
+	if (pt->adjusted_pf.discard_passdown) {
+		data_limits = &bdev_get_queue(pt->data_dev->bdev)->limits;
+		limits->discard_granularity = max(data_limits->discard_granularity,
+						  pool->sectors_per_block << SECTOR_SHIFT);
+	} else
+		limits->discard_granularity = pool->sectors_per_block << SECTOR_SHIFT;
+}
+
+static void pool_io_hints(struct dm_target *ti, struct queue_limits *limits)
+{
+	struct pool_c *pt = ti->private;
+	struct pool *pool = pt->pool;
+	uint64_t io_opt_sectors = limits->io_opt >> SECTOR_SHIFT;
+
+	/*
+	 * If the system-determined stacked limits are compatible with the
+	 * pool's blocksize (io_opt is a factor) do not override them.
+	 */
+	if (io_opt_sectors < pool->sectors_per_block ||
+	    do_div(io_opt_sectors, pool->sectors_per_block)) {
+		blk_limits_io_min(limits, 0);
+		blk_limits_io_opt(limits, pool->sectors_per_block << SECTOR_SHIFT);
+	}
+
+	/*
+	 * pt->adjusted_pf is a staging area for the actual features to use.
+	 * They get transferred to the live pool in bind_control_target()
+	 * called from pool_preresume().
+	 */
+	if (!pt->adjusted_pf.discard_enabled) {
+		/*
+		 * Must explicitly disallow stacking discard limits otherwise the
+		 * block layer will stack them if pool's data device has support.
+		 * QUEUE_FLAG_DISCARD wouldn't be set but there is no way for the
+		 * user to see that, so make sure to set all discard limits to 0.
+		 */
+		limits->discard_granularity = 0;
+		return;
+	}
+
+	disable_passdown_if_not_supported(pt);
+
+	set_discard_limits(pt, limits);
+}
+
+static struct target_type pool_target = {
+	.name = "thin-pool",
+	.features = DM_TARGET_SINGLETON | DM_TARGET_ALWAYS_WRITEABLE |
+		    DM_TARGET_IMMUTABLE,
+	.version = {1, 12, 0},
+	.module = THIS_MODULE,
+	.ctr = pool_ctr,
+	.dtr = pool_dtr,
+	.map = pool_map,
+	.postsuspend = pool_postsuspend,
+	.preresume = pool_preresume,
+	.resume = pool_resume,
+	.message = pool_message,
+	.status = pool_status,
+	.merge = pool_merge,
+	.iterate_devices = pool_iterate_devices,
+	.io_hints = pool_io_hints,
+};
+
+/*----------------------------------------------------------------
+ * Thin target methods
+ *--------------------------------------------------------------*/
+static void thin_get(struct thin_c *tc)
+{
+	atomic_inc(&tc->refcount);
+}
+
+static void thin_put(struct thin_c *tc)
+{
+	if (atomic_dec_and_test(&tc->refcount))
+		complete(&tc->can_destroy);
+}
+
+static void thin_dtr(struct dm_target *ti)
+{
+	struct thin_c *tc = ti->private;
+	unsigned long flags;
+
+	thin_put(tc);
+	wait_for_completion(&tc->can_destroy);
+
+	spin_lock_irqsave(&tc->pool->lock, flags);
+	list_del_rcu(&tc->list);
+	spin_unlock_irqrestore(&tc->pool->lock, flags);
+	synchronize_rcu();
+
+	mutex_lock(&dm_thin_pool_table.mutex);
+
+	__pool_dec(tc->pool);
+	dm_pool_close_thin_device(tc->td);
+	dm_put_device(ti, tc->pool_dev);
+	if (tc->origin_dev)
+		dm_put_device(ti, tc->origin_dev);
+	kfree(tc);
+
+	mutex_unlock(&dm_thin_pool_table.mutex);
+}
+
+/*
+ * Thin target parameters:
+ *
+ * <pool_dev> <dev_id> [origin_dev]
+ *
+ * pool_dev: the path to the pool (eg, /dev/mapper/my_pool)
+ * dev_id: the internal device identifier
+ * origin_dev: a device external to the pool that should act as the origin
+ *
+ * If the pool device has discards disabled, they get disabled for the thin
+ * device as well.
+ */
+static int thin_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	int r;
+	struct thin_c *tc;
+	struct dm_dev *pool_dev, *origin_dev;
+	struct mapped_device *pool_md;
+	unsigned long flags;
+
+	mutex_lock(&dm_thin_pool_table.mutex);
+
+	if (argc != 2 && argc != 3) {
+		ti->error = "Invalid argument count";
+		r = -EINVAL;
+		goto out_unlock;
+	}
+
+	tc = ti->private = kzalloc(sizeof(*tc), GFP_KERNEL);
+	if (!tc) {
+		ti->error = "Out of memory";
+		r = -ENOMEM;
+		goto out_unlock;
+	}
+	spin_lock_init(&tc->lock);
+	bio_list_init(&tc->deferred_bio_list);
+	bio_list_init(&tc->retry_on_resume_list);
+	tc->sort_bio_list = RB_ROOT;
+
+	if (argc == 3) {
+		r = dm_get_device(ti, argv[2], FMODE_READ, &origin_dev);
+		if (r) {
+			ti->error = "Error opening origin device";
+			goto bad_origin_dev;
+		}
+		tc->origin_dev = origin_dev;
+	}
+
+	r = dm_get_device(ti, argv[0], dm_table_get_mode(ti->table), &pool_dev);
+	if (r) {
+		ti->error = "Error opening pool device";
+		goto bad_pool_dev;
+	}
+	tc->pool_dev = pool_dev;
+
+	if (read_dev_id(argv[1], (unsigned long long *)&tc->dev_id, 0)) {
+		ti->error = "Invalid device id";
+		r = -EINVAL;
+		goto bad_common;
+	}
+
+	pool_md = dm_get_md(tc->pool_dev->bdev->bd_dev);
+	if (!pool_md) {
+		ti->error = "Couldn't get pool mapped device";
+		r = -EINVAL;
+		goto bad_common;
+	}
+
+	tc->pool = __pool_table_lookup(pool_md);
+	if (!tc->pool) {
+		ti->error = "Couldn't find pool object";
+		r = -EINVAL;
+		goto bad_pool_lookup;
+	}
+	__pool_inc(tc->pool);
+
+	if (get_pool_mode(tc->pool) == PM_FAIL) {
+		ti->error = "Couldn't open thin device, Pool is in fail mode";
+		r = -EINVAL;
+		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;
+	}
+
+	r = dm_set_target_max_io_len(ti, tc->pool->sectors_per_block);
+	if (r)
+		goto bad_target_max_io_len;
+
+	ti->num_flush_bios = 1;
+	ti->flush_supported = true;
+	ti->per_bio_data_size = sizeof(struct dm_thin_endio_hook);
+
+	/* In case the pool supports discards, pass them on. */
+	ti->discard_zeroes_data_unsupported = true;
+	if (tc->pool->pf.discard_enabled) {
+		ti->discards_supported = true;
+		ti->num_discard_bios = 1;
+		/* Discard bios must be split on a block boundary */
+		ti->split_discard_bios = true;
+	}
+
+	dm_put(pool_md);
+
+	mutex_unlock(&dm_thin_pool_table.mutex);
+
+	atomic_set(&tc->refcount, 1);
+	init_completion(&tc->can_destroy);
+
+	spin_lock_irqsave(&tc->pool->lock, flags);
+	list_add_tail_rcu(&tc->list, &tc->pool->active_thins);
+	spin_unlock_irqrestore(&tc->pool->lock, flags);
+	/*
+	 * This synchronize_rcu() call is needed here otherwise we risk a
+	 * wake_worker() call finding no bios to process (because the newly
+	 * added tc isn't yet visible).  So this reduces latency since we
+	 * aren't then dependent on the periodic commit to wake_worker().
+	 */
+	synchronize_rcu();
+
+	return 0;
+
+bad_target_max_io_len:
+	dm_pool_close_thin_device(tc->td);
+bad_thin_open:
+	__pool_dec(tc->pool);
+bad_pool_lookup:
+	dm_put(pool_md);
+bad_common:
+	dm_put_device(ti, tc->pool_dev);
+bad_pool_dev:
+	if (tc->origin_dev)
+		dm_put_device(ti, tc->origin_dev);
+bad_origin_dev:
+	kfree(tc);
+out_unlock:
+	mutex_unlock(&dm_thin_pool_table.mutex);
+
+	return r;
+}
+
+static int thin_map(struct dm_target *ti, struct bio *bio)
+{
+	bio->bi_iter.bi_sector = dm_target_offset(ti, bio->bi_iter.bi_sector);
+
+	return thin_bio_map(ti, bio);
+}
+
+static int thin_endio(struct dm_target *ti, struct bio *bio, int err)
+{
+	unsigned long flags;
+	struct dm_thin_endio_hook *h = dm_per_bio_data(bio, sizeof(struct dm_thin_endio_hook));
+	struct list_head work;
+	struct dm_thin_new_mapping *m, *tmp;
+	struct pool *pool = h->tc->pool;
+
+	if (h->shared_read_entry) {
+		INIT_LIST_HEAD(&work);
+		dm_deferred_entry_dec(h->shared_read_entry, &work);
+
+		spin_lock_irqsave(&pool->lock, flags);
+		list_for_each_entry_safe(m, tmp, &work, list) {
+			list_del(&m->list);
+			m->quiesced = true;
+			__maybe_add_mapping(m);
+		}
+		spin_unlock_irqrestore(&pool->lock, flags);
+	}
+
+	if (h->all_io_entry) {
+		INIT_LIST_HEAD(&work);
+		dm_deferred_entry_dec(h->all_io_entry, &work);
+		if (!list_empty(&work)) {
+			spin_lock_irqsave(&pool->lock, flags);
+			list_for_each_entry_safe(m, tmp, &work, list)
+				list_add_tail(&m->list, &pool->prepared_discards);
+			spin_unlock_irqrestore(&pool->lock, flags);
+			wake_worker(pool);
+		}
+	}
+
+	return 0;
+}
+
+static void thin_presuspend(struct dm_target *ti)
+{
+	struct thin_c *tc = ti->private;
+
+	if (dm_noflush_suspending(ti))
+		noflush_work(tc, do_noflush_start);
+}
+
+static void thin_postsuspend(struct dm_target *ti)
+{
+	struct thin_c *tc = ti->private;
+
+	/*
+	 * The dm_noflush_suspending flag has been cleared by now, so
+	 * unfortunately we must always run this.
+	 */
+	noflush_work(tc, do_noflush_stop);
+}
+
+/*
+ * <nr mapped sectors> <highest mapped sector>
+ */
+static void thin_status(struct dm_target *ti, status_type_t type,
+			unsigned status_flags, char *result, unsigned maxlen)
+{
+	int r;
+	ssize_t sz = 0;
+	dm_block_t mapped, highest;
+	char buf[BDEVNAME_SIZE];
+	struct thin_c *tc = ti->private;
+
+	if (get_pool_mode(tc->pool) == PM_FAIL) {
+		DMEMIT("Fail");
+		return;
+	}
+
+	if (!tc->td)
+		DMEMIT("-");
+	else {
+		switch (type) {
+		case STATUSTYPE_INFO:
+			r = dm_thin_get_mapped_count(tc->td, &mapped);
+			if (r) {
+				DMERR("dm_thin_get_mapped_count returned %d", r);
+				goto err;
+			}
+
+			r = dm_thin_get_highest_mapped_block(tc->td, &highest);
+			if (r < 0) {
+				DMERR("dm_thin_get_highest_mapped_block returned %d", r);
+				goto err;
+			}
+
+			DMEMIT("%llu ", mapped * tc->pool->sectors_per_block);
+			if (r)
+				DMEMIT("%llu", ((highest + 1) *
+						tc->pool->sectors_per_block) - 1);
+			else
+				DMEMIT("-");
+			break;
+
+		case STATUSTYPE_TABLE:
+			DMEMIT("%s %lu",
+			       format_dev_t(buf, tc->pool_dev->bdev->bd_dev),
+			       (unsigned long) tc->dev_id);
+			if (tc->origin_dev)
+				DMEMIT(" %s", format_dev_t(buf, tc->origin_dev->bdev->bd_dev));
+			break;
+		}
+	}
+
+	return;
+
+err:
+	DMEMIT("Error");
+}
+
+static int thin_iterate_devices(struct dm_target *ti,
+				iterate_devices_callout_fn fn, void *data)
+{
+	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 (!pool->ti)
+		return 0;	/* nothing is bound */
+
+	blocks = pool->ti->len;
+	(void) sector_div(blocks, pool->sectors_per_block);
+	if (blocks)
+		return fn(ti, tc->pool_dev, 0, pool->sectors_per_block * blocks, data);
+
+	return 0;
+}
+
+static struct target_type thin_target = {
+	.name = "thin",
+	.version = {1, 12, 0},
+	.module	= THIS_MODULE,
+	.ctr = thin_ctr,
+	.dtr = thin_dtr,
+	.map = thin_map,
+	.end_io = thin_endio,
+	.presuspend = thin_presuspend,
+	.postsuspend = thin_postsuspend,
+	.status = thin_status,
+	.iterate_devices = thin_iterate_devices,
+};
+
+/*----------------------------------------------------------------*/
+
+static int __init dm_thin_init(void)
+{
+	int r;
+
+	pool_table_init();
+
+	r = dm_register_target(&thin_target);
+	if (r)
+		return r;
+
+	r = dm_register_target(&pool_target);
+	if (r)
+		goto bad_pool_target;
+
+	r = -ENOMEM;
+
+	_new_mapping_cache = KMEM_CACHE(dm_thin_new_mapping, 0);
+	if (!_new_mapping_cache)
+		goto bad_new_mapping_cache;
+
+	return 0;
+
+bad_new_mapping_cache:
+	dm_unregister_target(&pool_target);
+bad_pool_target:
+	dm_unregister_target(&thin_target);
+
+	return r;
+}
+
+static void dm_thin_exit(void)
+{
+	dm_unregister_target(&thin_target);
+	dm_unregister_target(&pool_target);
+
+	kmem_cache_destroy(_new_mapping_cache);
+}
+
+module_init(dm_thin_init);
+module_exit(dm_thin_exit);
+
+module_param_named(no_space_timeout, no_space_timeout_secs, uint, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(no_space_timeout, "Out of data space queue IO timeout in seconds");
+
+MODULE_DESCRIPTION(DM_NAME " thin provisioning target");
+MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-uevent.c b/drivers/md/dm-uevent.c
index 6f65883aef1..8efe033bab5 100644
--- a/drivers/md/dm-uevent.c
+++ b/drivers/md/dm-uevent.c
@@ -22,6 +22,7 @@
 #include <linux/slab.h>
 #include <linux/kobject.h>
 #include <linux/dm-ioctl.h>
+#include <linux/export.h>
 
 #include "dm.h"
 #include "dm-uevent.h"
@@ -139,14 +140,13 @@ void dm_send_uevents(struct list_head *events, struct kobject *kobj)
 		list_del_init(&event->elist);
 
 		/*
-		 * Need to call dm_copy_name_and_uuid from here for now.
-		 * Context of previous var adds and locking used for
-		 * hash_cell not compatable.
+		 * When a device is being removed this copy fails and we
+		 * discard these unsent events.
 		 */
 		if (dm_copy_name_and_uuid(event->md, event->name,
 					  event->uuid)) {
-			DMERR("%s: dm_copy_name_and_uuid() failed",
-			      __func__);
+			DMINFO("%s: skipping sending uevent for lost device",
+			       __func__);
 			goto uevent_free;
 		}
 
@@ -188,7 +188,7 @@ void dm_path_uevent(enum dm_uevent_type event_type, struct dm_target *ti,
 
 	if (event_type >= ARRAY_SIZE(_dm_uevent_type_names)) {
 		DMERR("%s: Invalid event_type %d", __func__, event_type);
-		goto out;
+		return;
 	}
 
 	event = dm_build_path_uevent(md, ti,
@@ -196,12 +196,9 @@ void dm_path_uevent(enum dm_uevent_type event_type, struct dm_target *ti,
 				     _dm_uevent_type_names[event_type].name,
 				     path, nr_valid_paths);
 	if (IS_ERR(event))
-		goto out;
+		return;
 
 	dm_uevent_add(md, &event->elist);
-
-out:
-	dm_put(md);
 }
 EXPORT_SYMBOL_GPL(dm_path_uevent);
 
diff --git a/drivers/md/dm-verity.c b/drivers/md/dm-verity.c
new file mode 100644
index 00000000000..7a7bab8947a
--- /dev/null
+++ b/drivers/md/dm-verity.c
@@ -0,0 +1,903 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * Author: Mikulas Patocka <mpatocka@redhat.com>
+ *
+ * Based on Chromium dm-verity driver (C) 2011 The Chromium OS Authors
+ *
+ * This file is released under the GPLv2.
+ *
+ * In the file "/sys/module/dm_verity/parameters/prefetch_cluster" you can set
+ * default prefetch value. Data are read in "prefetch_cluster" chunks from the
+ * hash device. Setting this greatly improves performance when data and hash
+ * are on the same disk on different partitions on devices with poor random
+ * access behavior.
+ */
+
+#include "dm-bufio.h"
+
+#include <linux/module.h>
+#include <linux/device-mapper.h>
+#include <crypto/hash.h>
+
+#define DM_MSG_PREFIX			"verity"
+
+#define DM_VERITY_IO_VEC_INLINE		16
+#define DM_VERITY_MEMPOOL_SIZE		4
+#define DM_VERITY_DEFAULT_PREFETCH_SIZE	262144
+
+#define DM_VERITY_MAX_LEVELS		63
+
+static unsigned dm_verity_prefetch_cluster = DM_VERITY_DEFAULT_PREFETCH_SIZE;
+
+module_param_named(prefetch_cluster, dm_verity_prefetch_cluster, uint, S_IRUGO | S_IWUSR);
+
+struct dm_verity {
+	struct dm_dev *data_dev;
+	struct dm_dev *hash_dev;
+	struct dm_target *ti;
+	struct dm_bufio_client *bufio;
+	char *alg_name;
+	struct crypto_shash *tfm;
+	u8 *root_digest;	/* digest of the root block */
+	u8 *salt;		/* salt: its size is salt_size */
+	unsigned salt_size;
+	sector_t data_start;	/* data offset in 512-byte sectors */
+	sector_t hash_start;	/* hash start in blocks */
+	sector_t data_blocks;	/* the number of data blocks */
+	sector_t hash_blocks;	/* the number of hash blocks */
+	unsigned char data_dev_block_bits;	/* log2(data blocksize) */
+	unsigned char hash_dev_block_bits;	/* log2(hash blocksize) */
+	unsigned char hash_per_block_bits;	/* log2(hashes in hash block) */
+	unsigned char levels;	/* the number of tree levels */
+	unsigned char version;
+	unsigned digest_size;	/* digest size for the current hash algorithm */
+	unsigned shash_descsize;/* the size of temporary space for crypto */
+	int hash_failed;	/* set to 1 if hash of any block failed */
+
+	mempool_t *vec_mempool;	/* mempool of bio vector */
+
+	struct workqueue_struct *verify_wq;
+
+	/* starting blocks for each tree level. 0 is the lowest level. */
+	sector_t hash_level_block[DM_VERITY_MAX_LEVELS];
+};
+
+struct dm_verity_io {
+	struct dm_verity *v;
+
+	/* original values of bio->bi_end_io and bio->bi_private */
+	bio_end_io_t *orig_bi_end_io;
+	void *orig_bi_private;
+
+	sector_t block;
+	unsigned n_blocks;
+
+	struct bvec_iter iter;
+
+	struct work_struct work;
+
+	/*
+	 * Three variably-size fields follow this struct:
+	 *
+	 * u8 hash_desc[v->shash_descsize];
+	 * u8 real_digest[v->digest_size];
+	 * u8 want_digest[v->digest_size];
+	 *
+	 * To access them use: io_hash_desc(), io_real_digest() and io_want_digest().
+	 */
+};
+
+struct dm_verity_prefetch_work {
+	struct work_struct work;
+	struct dm_verity *v;
+	sector_t block;
+	unsigned n_blocks;
+};
+
+static struct shash_desc *io_hash_desc(struct dm_verity *v, struct dm_verity_io *io)
+{
+	return (struct shash_desc *)(io + 1);
+}
+
+static u8 *io_real_digest(struct dm_verity *v, struct dm_verity_io *io)
+{
+	return (u8 *)(io + 1) + v->shash_descsize;
+}
+
+static u8 *io_want_digest(struct dm_verity *v, struct dm_verity_io *io)
+{
+	return (u8 *)(io + 1) + v->shash_descsize + v->digest_size;
+}
+
+/*
+ * Auxiliary structure appended to each dm-bufio buffer. If the value
+ * hash_verified is nonzero, hash of the block has been verified.
+ *
+ * The variable hash_verified is set to 0 when allocating the buffer, then
+ * it can be changed to 1 and it is never reset to 0 again.
+ *
+ * There is no lock around this value, a race condition can at worst cause
+ * that multiple processes verify the hash of the same buffer simultaneously
+ * and write 1 to hash_verified simultaneously.
+ * This condition is harmless, so we don't need locking.
+ */
+struct buffer_aux {
+	int hash_verified;
+};
+
+/*
+ * Initialize struct buffer_aux for a freshly created buffer.
+ */
+static void dm_bufio_alloc_callback(struct dm_buffer *buf)
+{
+	struct buffer_aux *aux = dm_bufio_get_aux_data(buf);
+
+	aux->hash_verified = 0;
+}
+
+/*
+ * Translate input sector number to the sector number on the target device.
+ */
+static sector_t verity_map_sector(struct dm_verity *v, sector_t bi_sector)
+{
+	return v->data_start + dm_target_offset(v->ti, bi_sector);
+}
+
+/*
+ * Return hash position of a specified block at a specified tree level
+ * (0 is the lowest level).
+ * The lowest "hash_per_block_bits"-bits of the result denote hash position
+ * inside a hash block. The remaining bits denote location of the hash block.
+ */
+static sector_t verity_position_at_level(struct dm_verity *v, sector_t block,
+					 int level)
+{
+	return block >> (level * v->hash_per_block_bits);
+}
+
+static void verity_hash_at_level(struct dm_verity *v, sector_t block, int level,
+				 sector_t *hash_block, unsigned *offset)
+{
+	sector_t position = verity_position_at_level(v, block, level);
+	unsigned idx;
+
+	*hash_block = v->hash_level_block[level] + (position >> v->hash_per_block_bits);
+
+	if (!offset)
+		return;
+
+	idx = position & ((1 << v->hash_per_block_bits) - 1);
+	if (!v->version)
+		*offset = idx * v->digest_size;
+	else
+		*offset = idx << (v->hash_dev_block_bits - v->hash_per_block_bits);
+}
+
+/*
+ * Verify hash of a metadata block pertaining to the specified data block
+ * ("block" argument) at a specified level ("level" argument).
+ *
+ * On successful return, io_want_digest(v, io) contains the hash value for
+ * a lower tree level or for the data block (if we're at the lowest leve).
+ *
+ * If "skip_unverified" is true, unverified buffer is skipped and 1 is returned.
+ * If "skip_unverified" is false, unverified buffer is hashed and verified
+ * against current value of io_want_digest(v, io).
+ */
+static int verity_verify_level(struct dm_verity_io *io, sector_t block,
+			       int level, bool skip_unverified)
+{
+	struct dm_verity *v = io->v;
+	struct dm_buffer *buf;
+	struct buffer_aux *aux;
+	u8 *data;
+	int r;
+	sector_t hash_block;
+	unsigned offset;
+
+	verity_hash_at_level(v, block, level, &hash_block, &offset);
+
+	data = dm_bufio_read(v->bufio, hash_block, &buf);
+	if (unlikely(IS_ERR(data)))
+		return PTR_ERR(data);
+
+	aux = dm_bufio_get_aux_data(buf);
+
+	if (!aux->hash_verified) {
+		struct shash_desc *desc;
+		u8 *result;
+
+		if (skip_unverified) {
+			r = 1;
+			goto release_ret_r;
+		}
+
+		desc = io_hash_desc(v, io);
+		desc->tfm = v->tfm;
+		desc->flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+		r = crypto_shash_init(desc);
+		if (r < 0) {
+			DMERR("crypto_shash_init failed: %d", r);
+			goto release_ret_r;
+		}
+
+		if (likely(v->version >= 1)) {
+			r = crypto_shash_update(desc, v->salt, v->salt_size);
+			if (r < 0) {
+				DMERR("crypto_shash_update failed: %d", r);
+				goto release_ret_r;
+			}
+		}
+
+		r = crypto_shash_update(desc, data, 1 << v->hash_dev_block_bits);
+		if (r < 0) {
+			DMERR("crypto_shash_update failed: %d", r);
+			goto release_ret_r;
+		}
+
+		if (!v->version) {
+			r = crypto_shash_update(desc, v->salt, v->salt_size);
+			if (r < 0) {
+				DMERR("crypto_shash_update failed: %d", r);
+				goto release_ret_r;
+			}
+		}
+
+		result = io_real_digest(v, io);
+		r = crypto_shash_final(desc, result);
+		if (r < 0) {
+			DMERR("crypto_shash_final failed: %d", r);
+			goto release_ret_r;
+		}
+		if (unlikely(memcmp(result, io_want_digest(v, io), v->digest_size))) {
+			DMERR_LIMIT("metadata block %llu is corrupted",
+				(unsigned long long)hash_block);
+			v->hash_failed = 1;
+			r = -EIO;
+			goto release_ret_r;
+		} else
+			aux->hash_verified = 1;
+	}
+
+	data += offset;
+
+	memcpy(io_want_digest(v, io), data, v->digest_size);
+
+	dm_bufio_release(buf);
+	return 0;
+
+release_ret_r:
+	dm_bufio_release(buf);
+
+	return r;
+}
+
+/*
+ * Verify one "dm_verity_io" structure.
+ */
+static int verity_verify_io(struct dm_verity_io *io)
+{
+	struct dm_verity *v = io->v;
+	struct bio *bio = dm_bio_from_per_bio_data(io,
+						   v->ti->per_bio_data_size);
+	unsigned b;
+	int i;
+
+	for (b = 0; b < io->n_blocks; b++) {
+		struct shash_desc *desc;
+		u8 *result;
+		int r;
+		unsigned todo;
+
+		if (likely(v->levels)) {
+			/*
+			 * First, we try to get the requested hash for
+			 * the current block. If the hash block itself is
+			 * verified, zero is returned. If it isn't, this
+			 * function returns 0 and we fall back to whole
+			 * chain verification.
+			 */
+			int r = verity_verify_level(io, io->block + b, 0, true);
+			if (likely(!r))
+				goto test_block_hash;
+			if (r < 0)
+				return r;
+		}
+
+		memcpy(io_want_digest(v, io), v->root_digest, v->digest_size);
+
+		for (i = v->levels - 1; i >= 0; i--) {
+			int r = verity_verify_level(io, io->block + b, i, false);
+			if (unlikely(r))
+				return r;
+		}
+
+test_block_hash:
+		desc = io_hash_desc(v, io);
+		desc->tfm = v->tfm;
+		desc->flags = CRYPTO_TFM_REQ_MAY_SLEEP;
+		r = crypto_shash_init(desc);
+		if (r < 0) {
+			DMERR("crypto_shash_init failed: %d", r);
+			return r;
+		}
+
+		if (likely(v->version >= 1)) {
+			r = crypto_shash_update(desc, v->salt, v->salt_size);
+			if (r < 0) {
+				DMERR("crypto_shash_update failed: %d", r);
+				return r;
+			}
+		}
+		todo = 1 << v->data_dev_block_bits;
+		do {
+			u8 *page;
+			unsigned len;
+			struct bio_vec bv = bio_iter_iovec(bio, io->iter);
+
+			page = kmap_atomic(bv.bv_page);
+			len = bv.bv_len;
+			if (likely(len >= todo))
+				len = todo;
+			r = crypto_shash_update(desc, page + bv.bv_offset, len);
+			kunmap_atomic(page);
+
+			if (r < 0) {
+				DMERR("crypto_shash_update failed: %d", r);
+				return r;
+			}
+
+			bio_advance_iter(bio, &io->iter, len);
+			todo -= len;
+		} while (todo);
+
+		if (!v->version) {
+			r = crypto_shash_update(desc, v->salt, v->salt_size);
+			if (r < 0) {
+				DMERR("crypto_shash_update failed: %d", r);
+				return r;
+			}
+		}
+
+		result = io_real_digest(v, io);
+		r = crypto_shash_final(desc, result);
+		if (r < 0) {
+			DMERR("crypto_shash_final failed: %d", r);
+			return r;
+		}
+		if (unlikely(memcmp(result, io_want_digest(v, io), v->digest_size))) {
+			DMERR_LIMIT("data block %llu is corrupted",
+				(unsigned long long)(io->block + b));
+			v->hash_failed = 1;
+			return -EIO;
+		}
+	}
+
+	return 0;
+}
+
+/*
+ * End one "io" structure with a given error.
+ */
+static void verity_finish_io(struct dm_verity_io *io, int error)
+{
+	struct dm_verity *v = io->v;
+	struct bio *bio = dm_bio_from_per_bio_data(io, v->ti->per_bio_data_size);
+
+	bio->bi_end_io = io->orig_bi_end_io;
+	bio->bi_private = io->orig_bi_private;
+
+	bio_endio_nodec(bio, error);
+}
+
+static void verity_work(struct work_struct *w)
+{
+	struct dm_verity_io *io = container_of(w, struct dm_verity_io, work);
+
+	verity_finish_io(io, verity_verify_io(io));
+}
+
+static void verity_end_io(struct bio *bio, int error)
+{
+	struct dm_verity_io *io = bio->bi_private;
+
+	if (error) {
+		verity_finish_io(io, error);
+		return;
+	}
+
+	INIT_WORK(&io->work, verity_work);
+	queue_work(io->v->verify_wq, &io->work);
+}
+
+/*
+ * Prefetch buffers for the specified io.
+ * The root buffer is not prefetched, it is assumed that it will be cached
+ * all the time.
+ */
+static void verity_prefetch_io(struct work_struct *work)
+{
+	struct dm_verity_prefetch_work *pw =
+		container_of(work, struct dm_verity_prefetch_work, work);
+	struct dm_verity *v = pw->v;
+	int i;
+
+	for (i = v->levels - 2; i >= 0; i--) {
+		sector_t hash_block_start;
+		sector_t hash_block_end;
+		verity_hash_at_level(v, pw->block, i, &hash_block_start, NULL);
+		verity_hash_at_level(v, pw->block + pw->n_blocks - 1, i, &hash_block_end, NULL);
+		if (!i) {
+			unsigned cluster = ACCESS_ONCE(dm_verity_prefetch_cluster);
+
+			cluster >>= v->data_dev_block_bits;
+			if (unlikely(!cluster))
+				goto no_prefetch_cluster;
+
+			if (unlikely(cluster & (cluster - 1)))
+				cluster = 1 << __fls(cluster);
+
+			hash_block_start &= ~(sector_t)(cluster - 1);
+			hash_block_end |= cluster - 1;
+			if (unlikely(hash_block_end >= v->hash_blocks))
+				hash_block_end = v->hash_blocks - 1;
+		}
+no_prefetch_cluster:
+		dm_bufio_prefetch(v->bufio, hash_block_start,
+				  hash_block_end - hash_block_start + 1);
+	}
+
+	kfree(pw);
+}
+
+static void verity_submit_prefetch(struct dm_verity *v, struct dm_verity_io *io)
+{
+	struct dm_verity_prefetch_work *pw;
+
+	pw = kmalloc(sizeof(struct dm_verity_prefetch_work),
+		GFP_NOIO | __GFP_NORETRY | __GFP_NOMEMALLOC | __GFP_NOWARN);
+
+	if (!pw)
+		return;
+
+	INIT_WORK(&pw->work, verity_prefetch_io);
+	pw->v = v;
+	pw->block = io->block;
+	pw->n_blocks = io->n_blocks;
+	queue_work(v->verify_wq, &pw->work);
+}
+
+/*
+ * Bio map function. It allocates dm_verity_io structure and bio vector and
+ * fills them. Then it issues prefetches and the I/O.
+ */
+static int verity_map(struct dm_target *ti, struct bio *bio)
+{
+	struct dm_verity *v = ti->private;
+	struct dm_verity_io *io;
+
+	bio->bi_bdev = v->data_dev->bdev;
+	bio->bi_iter.bi_sector = verity_map_sector(v, bio->bi_iter.bi_sector);
+
+	if (((unsigned)bio->bi_iter.bi_sector | bio_sectors(bio)) &
+	    ((1 << (v->data_dev_block_bits - SECTOR_SHIFT)) - 1)) {
+		DMERR_LIMIT("unaligned io");
+		return -EIO;
+	}
+
+	if (bio_end_sector(bio) >>
+	    (v->data_dev_block_bits - SECTOR_SHIFT) > v->data_blocks) {
+		DMERR_LIMIT("io out of range");
+		return -EIO;
+	}
+
+	if (bio_data_dir(bio) == WRITE)
+		return -EIO;
+
+	io = dm_per_bio_data(bio, ti->per_bio_data_size);
+	io->v = v;
+	io->orig_bi_end_io = bio->bi_end_io;
+	io->orig_bi_private = bio->bi_private;
+	io->block = bio->bi_iter.bi_sector >> (v->data_dev_block_bits - SECTOR_SHIFT);
+	io->n_blocks = bio->bi_iter.bi_size >> v->data_dev_block_bits;
+
+	bio->bi_end_io = verity_end_io;
+	bio->bi_private = io;
+	io->iter = bio->bi_iter;
+
+	verity_submit_prefetch(v, io);
+
+	generic_make_request(bio);
+
+	return DM_MAPIO_SUBMITTED;
+}
+
+/*
+ * Status: V (valid) or C (corruption found)
+ */
+static void verity_status(struct dm_target *ti, status_type_t type,
+			  unsigned status_flags, char *result, unsigned maxlen)
+{
+	struct dm_verity *v = ti->private;
+	unsigned sz = 0;
+	unsigned x;
+
+	switch (type) {
+	case STATUSTYPE_INFO:
+		DMEMIT("%c", v->hash_failed ? 'C' : 'V');
+		break;
+	case STATUSTYPE_TABLE:
+		DMEMIT("%u %s %s %u %u %llu %llu %s ",
+			v->version,
+			v->data_dev->name,
+			v->hash_dev->name,
+			1 << v->data_dev_block_bits,
+			1 << v->hash_dev_block_bits,
+			(unsigned long long)v->data_blocks,
+			(unsigned long long)v->hash_start,
+			v->alg_name
+			);
+		for (x = 0; x < v->digest_size; x++)
+			DMEMIT("%02x", v->root_digest[x]);
+		DMEMIT(" ");
+		if (!v->salt_size)
+			DMEMIT("-");
+		else
+			for (x = 0; x < v->salt_size; x++)
+				DMEMIT("%02x", v->salt[x]);
+		break;
+	}
+}
+
+static int verity_ioctl(struct dm_target *ti, unsigned cmd,
+			unsigned long arg)
+{
+	struct dm_verity *v = ti->private;
+	int r = 0;
+
+	if (v->data_start ||
+	    ti->len != i_size_read(v->data_dev->bdev->bd_inode) >> SECTOR_SHIFT)
+		r = scsi_verify_blk_ioctl(NULL, cmd);
+
+	return r ? : __blkdev_driver_ioctl(v->data_dev->bdev, v->data_dev->mode,
+				     cmd, arg);
+}
+
+static int verity_merge(struct dm_target *ti, struct bvec_merge_data *bvm,
+			struct bio_vec *biovec, int max_size)
+{
+	struct dm_verity *v = ti->private;
+	struct request_queue *q = bdev_get_queue(v->data_dev->bdev);
+
+	if (!q->merge_bvec_fn)
+		return max_size;
+
+	bvm->bi_bdev = v->data_dev->bdev;
+	bvm->bi_sector = verity_map_sector(v, bvm->bi_sector);
+
+	return min(max_size, q->merge_bvec_fn(q, bvm, biovec));
+}
+
+static int verity_iterate_devices(struct dm_target *ti,
+				  iterate_devices_callout_fn fn, void *data)
+{
+	struct dm_verity *v = ti->private;
+
+	return fn(ti, v->data_dev, v->data_start, ti->len, data);
+}
+
+static void verity_io_hints(struct dm_target *ti, struct queue_limits *limits)
+{
+	struct dm_verity *v = ti->private;
+
+	if (limits->logical_block_size < 1 << v->data_dev_block_bits)
+		limits->logical_block_size = 1 << v->data_dev_block_bits;
+
+	if (limits->physical_block_size < 1 << v->data_dev_block_bits)
+		limits->physical_block_size = 1 << v->data_dev_block_bits;
+
+	blk_limits_io_min(limits, limits->logical_block_size);
+}
+
+static void verity_dtr(struct dm_target *ti)
+{
+	struct dm_verity *v = ti->private;
+
+	if (v->verify_wq)
+		destroy_workqueue(v->verify_wq);
+
+	if (v->vec_mempool)
+		mempool_destroy(v->vec_mempool);
+
+	if (v->bufio)
+		dm_bufio_client_destroy(v->bufio);
+
+	kfree(v->salt);
+	kfree(v->root_digest);
+
+	if (v->tfm)
+		crypto_free_shash(v->tfm);
+
+	kfree(v->alg_name);
+
+	if (v->hash_dev)
+		dm_put_device(ti, v->hash_dev);
+
+	if (v->data_dev)
+		dm_put_device(ti, v->data_dev);
+
+	kfree(v);
+}
+
+/*
+ * Target parameters:
+ *	<version>	The current format is version 1.
+ *			Vsn 0 is compatible with original Chromium OS releases.
+ *	<data device>
+ *	<hash device>
+ *	<data block size>
+ *	<hash block size>
+ *	<the number of data blocks>
+ *	<hash start block>
+ *	<algorithm>
+ *	<digest>
+ *	<salt>		Hex string or "-" if no salt.
+ */
+static int verity_ctr(struct dm_target *ti, unsigned argc, char **argv)
+{
+	struct dm_verity *v;
+	unsigned num;
+	unsigned long long num_ll;
+	int r;
+	int i;
+	sector_t hash_position;
+	char dummy;
+
+	v = kzalloc(sizeof(struct dm_verity), GFP_KERNEL);
+	if (!v) {
+		ti->error = "Cannot allocate verity structure";
+		return -ENOMEM;
+	}
+	ti->private = v;
+	v->ti = ti;
+
+	if ((dm_table_get_mode(ti->table) & ~FMODE_READ)) {
+		ti->error = "Device must be readonly";
+		r = -EINVAL;
+		goto bad;
+	}
+
+	if (argc != 10) {
+		ti->error = "Invalid argument count: exactly 10 arguments required";
+		r = -EINVAL;
+		goto bad;
+	}
+
+	if (sscanf(argv[0], "%u%c", &num, &dummy) != 1 ||
+	    num > 1) {
+		ti->error = "Invalid version";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->version = num;
+
+	r = dm_get_device(ti, argv[1], FMODE_READ, &v->data_dev);
+	if (r) {
+		ti->error = "Data device lookup failed";
+		goto bad;
+	}
+
+	r = dm_get_device(ti, argv[2], FMODE_READ, &v->hash_dev);
+	if (r) {
+		ti->error = "Data device lookup failed";
+		goto bad;
+	}
+
+	if (sscanf(argv[3], "%u%c", &num, &dummy) != 1 ||
+	    !num || (num & (num - 1)) ||
+	    num < bdev_logical_block_size(v->data_dev->bdev) ||
+	    num > PAGE_SIZE) {
+		ti->error = "Invalid data device block size";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->data_dev_block_bits = __ffs(num);
+
+	if (sscanf(argv[4], "%u%c", &num, &dummy) != 1 ||
+	    !num || (num & (num - 1)) ||
+	    num < bdev_logical_block_size(v->hash_dev->bdev) ||
+	    num > INT_MAX) {
+		ti->error = "Invalid hash device block size";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->hash_dev_block_bits = __ffs(num);
+
+	if (sscanf(argv[5], "%llu%c", &num_ll, &dummy) != 1 ||
+	    (sector_t)(num_ll << (v->data_dev_block_bits - SECTOR_SHIFT))
+	    >> (v->data_dev_block_bits - SECTOR_SHIFT) != num_ll) {
+		ti->error = "Invalid data blocks";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->data_blocks = num_ll;
+
+	if (ti->len > (v->data_blocks << (v->data_dev_block_bits - SECTOR_SHIFT))) {
+		ti->error = "Data device is too small";
+		r = -EINVAL;
+		goto bad;
+	}
+
+	if (sscanf(argv[6], "%llu%c", &num_ll, &dummy) != 1 ||
+	    (sector_t)(num_ll << (v->hash_dev_block_bits - SECTOR_SHIFT))
+	    >> (v->hash_dev_block_bits - SECTOR_SHIFT) != num_ll) {
+		ti->error = "Invalid hash start";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->hash_start = num_ll;
+
+	v->alg_name = kstrdup(argv[7], GFP_KERNEL);
+	if (!v->alg_name) {
+		ti->error = "Cannot allocate algorithm name";
+		r = -ENOMEM;
+		goto bad;
+	}
+
+	v->tfm = crypto_alloc_shash(v->alg_name, 0, 0);
+	if (IS_ERR(v->tfm)) {
+		ti->error = "Cannot initialize hash function";
+		r = PTR_ERR(v->tfm);
+		v->tfm = NULL;
+		goto bad;
+	}
+	v->digest_size = crypto_shash_digestsize(v->tfm);
+	if ((1 << v->hash_dev_block_bits) < v->digest_size * 2) {
+		ti->error = "Digest size too big";
+		r = -EINVAL;
+		goto bad;
+	}
+	v->shash_descsize =
+		sizeof(struct shash_desc) + crypto_shash_descsize(v->tfm);
+
+	v->root_digest = kmalloc(v->digest_size, GFP_KERNEL);
+	if (!v->root_digest) {
+		ti->error = "Cannot allocate root digest";
+		r = -ENOMEM;
+		goto bad;
+	}
+	if (strlen(argv[8]) != v->digest_size * 2 ||
+	    hex2bin(v->root_digest, argv[8], v->digest_size)) {
+		ti->error = "Invalid root digest";
+		r = -EINVAL;
+		goto bad;
+	}
+
+	if (strcmp(argv[9], "-")) {
+		v->salt_size = strlen(argv[9]) / 2;
+		v->salt = kmalloc(v->salt_size, GFP_KERNEL);
+		if (!v->salt) {
+			ti->error = "Cannot allocate salt";
+			r = -ENOMEM;
+			goto bad;
+		}
+		if (strlen(argv[9]) != v->salt_size * 2 ||
+		    hex2bin(v->salt, argv[9], v->salt_size)) {
+			ti->error = "Invalid salt";
+			r = -EINVAL;
+			goto bad;
+		}
+	}
+
+	v->hash_per_block_bits =
+		__fls((1 << v->hash_dev_block_bits) / v->digest_size);
+
+	v->levels = 0;
+	if (v->data_blocks)
+		while (v->hash_per_block_bits * v->levels < 64 &&
+		       (unsigned long long)(v->data_blocks - 1) >>
+		       (v->hash_per_block_bits * v->levels))
+			v->levels++;
+
+	if (v->levels > DM_VERITY_MAX_LEVELS) {
+		ti->error = "Too many tree levels";
+		r = -E2BIG;
+		goto bad;
+	}
+
+	hash_position = v->hash_start;
+	for (i = v->levels - 1; i >= 0; i--) {
+		sector_t s;
+		v->hash_level_block[i] = hash_position;
+		s = (v->data_blocks + ((sector_t)1 << ((i + 1) * v->hash_per_block_bits)) - 1)
+					>> ((i + 1) * v->hash_per_block_bits);
+		if (hash_position + s < hash_position) {
+			ti->error = "Hash device offset overflow";
+			r = -E2BIG;
+			goto bad;
+		}
+		hash_position += s;
+	}
+	v->hash_blocks = hash_position;
+
+	v->bufio = dm_bufio_client_create(v->hash_dev->bdev,
+		1 << v->hash_dev_block_bits, 1, sizeof(struct buffer_aux),
+		dm_bufio_alloc_callback, NULL);
+	if (IS_ERR(v->bufio)) {
+		ti->error = "Cannot initialize dm-bufio";
+		r = PTR_ERR(v->bufio);
+		v->bufio = NULL;
+		goto bad;
+	}
+
+	if (dm_bufio_get_device_size(v->bufio) < v->hash_blocks) {
+		ti->error = "Hash device is too small";
+		r = -E2BIG;
+		goto bad;
+	}
+
+	ti->per_bio_data_size = roundup(sizeof(struct dm_verity_io) + v->shash_descsize + v->digest_size * 2, __alignof__(struct dm_verity_io));
+
+	v->vec_mempool = mempool_create_kmalloc_pool(DM_VERITY_MEMPOOL_SIZE,
+					BIO_MAX_PAGES * sizeof(struct bio_vec));
+	if (!v->vec_mempool) {
+		ti->error = "Cannot allocate vector mempool";
+		r = -ENOMEM;
+		goto bad;
+	}
+
+	/* WQ_UNBOUND greatly improves performance when running on ramdisk */
+	v->verify_wq = alloc_workqueue("kverityd", WQ_CPU_INTENSIVE | WQ_MEM_RECLAIM | WQ_UNBOUND, num_online_cpus());
+	if (!v->verify_wq) {
+		ti->error = "Cannot allocate workqueue";
+		r = -ENOMEM;
+		goto bad;
+	}
+
+	return 0;
+
+bad:
+	verity_dtr(ti);
+
+	return r;
+}
+
+static struct target_type verity_target = {
+	.name		= "verity",
+	.version	= {1, 2, 0},
+	.module		= THIS_MODULE,
+	.ctr		= verity_ctr,
+	.dtr		= verity_dtr,
+	.map		= verity_map,
+	.status		= verity_status,
+	.ioctl		= verity_ioctl,
+	.merge		= verity_merge,
+	.iterate_devices = verity_iterate_devices,
+	.io_hints	= verity_io_hints,
+};
+
+static int __init dm_verity_init(void)
+{
+	int r;
+
+	r = dm_register_target(&verity_target);
+	if (r < 0)
+		DMERR("register failed %d", r);
+
+	return r;
+}
+
+static void __exit dm_verity_exit(void)
+{
+	dm_unregister_target(&verity_target);
+}
+
+module_init(dm_verity_init);
+module_exit(dm_verity_exit);
+
+MODULE_AUTHOR("Mikulas Patocka <mpatocka@redhat.com>");
+MODULE_AUTHOR("Mandeep Baines <msb@chromium.org>");
+MODULE_AUTHOR("Will Drewry <wad@chromium.org>");
+MODULE_DESCRIPTION(DM_NAME " target for transparent disk integrity checking");
+MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm-zero.c b/drivers/md/dm-zero.c
index bdec206c404..b9a64bbce30 100644
--- a/drivers/md/dm-zero.c
+++ b/drivers/md/dm-zero.c
@@ -1,10 +1,10 @@
 /*
- * Copyright (C) 2003 Christophe Saout <christophe@saout.de>
+ * Copyright (C) 2003 Jana Saout <jana@saout.de>
  *
  * This file is released under the GPL.
  */
 
-#include "dm.h"
+#include <linux/device-mapper.h>
 
 #include <linux/module.h>
 #include <linux/init.h>
@@ -22,14 +22,18 @@ static int zero_ctr(struct dm_target *ti, unsigned int argc, char **argv)
 		return -EINVAL;
 	}
 
+	/*
+	 * Silently drop discards, avoiding -EOPNOTSUPP.
+	 */
+	ti->num_discard_bios = 1;
+
 	return 0;
 }
 
 /*
  * Return zeros only on reads
  */
-static int zero_map(struct dm_target *ti, struct bio *bio,
-		      union map_info *map_context)
+static int zero_map(struct dm_target *ti, struct bio *bio)
 {
 	switch(bio_rw(bio)) {
 	case READ:
@@ -51,7 +55,7 @@ static int zero_map(struct dm_target *ti, struct bio *bio,
 
 static struct target_type zero_target = {
 	.name   = "zero",
-	.version = {1, 0, 0},
+	.version = {1, 1, 0},
 	.module = THIS_MODULE,
 	.ctr    = zero_ctr,
 	.map    = zero_map,
@@ -69,15 +73,12 @@ static int __init dm_zero_init(void)
 
 static void __exit dm_zero_exit(void)
 {
-	int r = dm_unregister_target(&zero_target);
-
-	if (r < 0)
-		DMERR("unregister failed %d", r);
+	dm_unregister_target(&zero_target);
 }
 
 module_init(dm_zero_init)
 module_exit(dm_zero_exit)
 
-MODULE_AUTHOR("Christophe Saout <christophe@saout.de>");
+MODULE_AUTHOR("Jana Saout <jana@saout.de>");
 MODULE_DESCRIPTION(DM_NAME " dummy target returning zeros");
 MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm.c b/drivers/md/dm.c
index bca448e1187..32b958dbc49 100644
--- a/drivers/md/dm.c
+++ b/drivers/md/dm.c
@@ -1,12 +1,11 @@
 /*
  * Copyright (C) 2001, 2002 Sistina Software (UK) Limited.
- * Copyright (C) 2004-2006 Red Hat, Inc. All rights reserved.
+ * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
  *
  * This file is released under the GPL.
  */
 
 #include "dm.h"
-#include "dm-bio-list.h"
 #include "dm-uevent.h"
 
 #include <linux/init.h>
@@ -15,23 +14,50 @@
 #include <linux/moduleparam.h>
 #include <linux/blkpg.h>
 #include <linux/bio.h>
-#include <linux/buffer_head.h>
 #include <linux/mempool.h>
 #include <linux/slab.h>
 #include <linux/idr.h>
 #include <linux/hdreg.h>
-#include <linux/blktrace_api.h>
-#include <linux/smp_lock.h>
+#include <linux/delay.h>
+
+#include <trace/events/block.h>
 
 #define DM_MSG_PREFIX "core"
 
+#ifdef CONFIG_PRINTK
+/*
+ * ratelimit state to be used in DMXXX_LIMIT().
+ */
+DEFINE_RATELIMIT_STATE(dm_ratelimit_state,
+		       DEFAULT_RATELIMIT_INTERVAL,
+		       DEFAULT_RATELIMIT_BURST);
+EXPORT_SYMBOL(dm_ratelimit_state);
+#endif
+
+/*
+ * Cookies are numeric values sent with CHANGE and REMOVE
+ * uevents while resuming, removing or renaming the device.
+ */
+#define DM_COOKIE_ENV_VAR_NAME "DM_COOKIE"
+#define DM_COOKIE_LENGTH 24
+
 static const char *_name = DM_NAME;
 
 static unsigned int major = 0;
 static unsigned int _major = 0;
 
+static DEFINE_IDR(_minor_idr);
+
 static DEFINE_SPINLOCK(_minor_lock);
+
+static void do_deferred_remove(struct work_struct *w);
+
+static DECLARE_WORK(deferred_remove_work, do_deferred_remove);
+
+static struct workqueue_struct *deferred_remove_workqueue;
+
 /*
+ * For bio-based dm.
  * One of these is allocated per bio.
  */
 struct dm_io {
@@ -40,61 +66,91 @@ struct dm_io {
 	atomic_t io_count;
 	struct bio *bio;
 	unsigned long start_time;
+	spinlock_t endio_lock;
+	struct dm_stats_aux stats_aux;
 };
 
 /*
- * One of these is allocated per target within a bio.  Hopefully
- * this will be simplified out one day.
+ * For request-based dm.
+ * One of these is allocated per request.
  */
-struct dm_target_io {
-	struct dm_io *io;
+struct dm_rq_target_io {
+	struct mapped_device *md;
 	struct dm_target *ti;
+	struct request *orig, clone;
+	int error;
 	union map_info info;
 };
 
-union map_info *dm_get_mapinfo(struct bio *bio)
+/*
+ * For request-based dm - the bio clones we allocate are embedded in these
+ * structs.
+ *
+ * We allocate these with bio_alloc_bioset, using the front_pad parameter when
+ * the bioset is created - this means the bio has to come at the end of the
+ * struct.
+ */
+struct dm_rq_clone_bio_info {
+	struct bio *orig;
+	struct dm_rq_target_io *tio;
+	struct bio clone;
+};
+
+union map_info *dm_get_rq_mapinfo(struct request *rq)
 {
-	if (bio && bio->bi_private)
-		return &((struct dm_target_io *)bio->bi_private)->info;
+	if (rq && rq->end_io_data)
+		return &((struct dm_rq_target_io *)rq->end_io_data)->info;
 	return NULL;
 }
+EXPORT_SYMBOL_GPL(dm_get_rq_mapinfo);
 
 #define MINOR_ALLOCED ((void *)-1)
 
 /*
  * Bits for the md->flags field.
  */
-#define DMF_BLOCK_IO 0
+#define DMF_BLOCK_IO_FOR_SUSPEND 0
 #define DMF_SUSPENDED 1
 #define DMF_FROZEN 2
 #define DMF_FREEING 3
 #define DMF_DELETING 4
 #define DMF_NOFLUSH_SUSPENDING 5
+#define DMF_MERGE_IS_OPTIONAL 6
+#define DMF_DEFERRED_REMOVE 7
 
 /*
- * Work processed by per-device workqueue.
+ * A dummy definition to make RCU happy.
+ * struct dm_table should never be dereferenced in this file.
  */
-struct dm_wq_req {
-	enum {
-		DM_WQ_FLUSH_ALL,
-		DM_WQ_FLUSH_DEFERRED,
-	} type;
-	struct work_struct work;
-	struct mapped_device *md;
-	void *context;
+struct dm_table {
+	int undefined__;
 };
 
+/*
+ * Work processed by per-device workqueue.
+ */
 struct mapped_device {
-	struct rw_semaphore io_lock;
+	struct srcu_struct io_barrier;
 	struct mutex suspend_lock;
-	spinlock_t pushback_lock;
-	rwlock_t map_lock;
 	atomic_t holders;
 	atomic_t open_count;
 
+	/*
+	 * The current mapping.
+	 * Use dm_get_live_table{_fast} or take suspend_lock for
+	 * dereference.
+	 */
+	struct dm_table *map;
+
 	unsigned long flags;
 
 	struct request_queue *queue;
+	unsigned type;
+	/* Protect queue and type against concurrent access. */
+	struct mutex type_lock;
+
+	struct target_type *immutable_target_type;
+
 	struct gendisk *disk;
 	char name[16];
 
@@ -103,26 +159,21 @@ struct mapped_device {
 	/*
 	 * A list of ios that arrived while we were suspended.
 	 */
-	atomic_t pending;
+	atomic_t pending[2];
 	wait_queue_head_t wait;
+	struct work_struct work;
 	struct bio_list deferred;
-	struct bio_list pushback;
+	spinlock_t deferred_lock;
 
 	/*
-	 * Processing queue (flush/barriers)
+	 * Processing queue (flush)
 	 */
 	struct workqueue_struct *wq;
 
 	/*
-	 * The current mapping.
-	 */
-	struct dm_table *map;
-
-	/*
 	 * io objects are allocated from here.
 	 */
 	mempool_t *io_pool;
-	mempool_t *tio_pool;
 
 	struct bio_set *bs;
 
@@ -139,57 +190,128 @@ struct mapped_device {
 	 * freeze/thaw support require holding onto a super block
 	 */
 	struct super_block *frozen_sb;
-	struct block_device *suspended_bdev;
+	struct block_device *bdev;
 
 	/* forced geometry settings */
 	struct hd_geometry geometry;
+
+	/* kobject and completion */
+	struct dm_kobject_holder kobj_holder;
+
+	/* zero-length flush that will be cloned and submitted to targets */
+	struct bio flush_bio;
+
+	struct dm_stats stats;
 };
 
-#define MIN_IOS 256
+/*
+ * For mempools pre-allocation at the table loading time.
+ */
+struct dm_md_mempools {
+	mempool_t *io_pool;
+	struct bio_set *bs;
+};
+
+#define RESERVED_BIO_BASED_IOS		16
+#define RESERVED_REQUEST_BASED_IOS	256
+#define RESERVED_MAX_IOS		1024
 static struct kmem_cache *_io_cache;
-static struct kmem_cache *_tio_cache;
+static struct kmem_cache *_rq_tio_cache;
+
+/*
+ * Bio-based DM's mempools' reserved IOs set by the user.
+ */
+static unsigned reserved_bio_based_ios = RESERVED_BIO_BASED_IOS;
+
+/*
+ * Request-based DM's mempools' reserved IOs set by the user.
+ */
+static unsigned reserved_rq_based_ios = RESERVED_REQUEST_BASED_IOS;
+
+static unsigned __dm_get_reserved_ios(unsigned *reserved_ios,
+				      unsigned def, unsigned max)
+{
+	unsigned ios = ACCESS_ONCE(*reserved_ios);
+	unsigned modified_ios = 0;
+
+	if (!ios)
+		modified_ios = def;
+	else if (ios > max)
+		modified_ios = max;
+
+	if (modified_ios) {
+		(void)cmpxchg(reserved_ios, ios, modified_ios);
+		ios = modified_ios;
+	}
+
+	return ios;
+}
+
+unsigned dm_get_reserved_bio_based_ios(void)
+{
+	return __dm_get_reserved_ios(&reserved_bio_based_ios,
+				     RESERVED_BIO_BASED_IOS, RESERVED_MAX_IOS);
+}
+EXPORT_SYMBOL_GPL(dm_get_reserved_bio_based_ios);
+
+unsigned dm_get_reserved_rq_based_ios(void)
+{
+	return __dm_get_reserved_ios(&reserved_rq_based_ios,
+				     RESERVED_REQUEST_BASED_IOS, RESERVED_MAX_IOS);
+}
+EXPORT_SYMBOL_GPL(dm_get_reserved_rq_based_ios);
 
 static int __init local_init(void)
 {
-	int r;
+	int r = -ENOMEM;
 
 	/* allocate a slab for the dm_ios */
 	_io_cache = KMEM_CACHE(dm_io, 0);
 	if (!_io_cache)
-		return -ENOMEM;
+		return r;
 
-	/* allocate a slab for the target ios */
-	_tio_cache = KMEM_CACHE(dm_target_io, 0);
-	if (!_tio_cache) {
-		kmem_cache_destroy(_io_cache);
-		return -ENOMEM;
-	}
+	_rq_tio_cache = KMEM_CACHE(dm_rq_target_io, 0);
+	if (!_rq_tio_cache)
+		goto out_free_io_cache;
 
 	r = dm_uevent_init();
-	if (r) {
-		kmem_cache_destroy(_tio_cache);
-		kmem_cache_destroy(_io_cache);
-		return r;
+	if (r)
+		goto out_free_rq_tio_cache;
+
+	deferred_remove_workqueue = alloc_workqueue("kdmremove", WQ_UNBOUND, 1);
+	if (!deferred_remove_workqueue) {
+		r = -ENOMEM;
+		goto out_uevent_exit;
 	}
 
 	_major = major;
 	r = register_blkdev(_major, _name);
-	if (r < 0) {
-		kmem_cache_destroy(_tio_cache);
-		kmem_cache_destroy(_io_cache);
-		dm_uevent_exit();
-		return r;
-	}
+	if (r < 0)
+		goto out_free_workqueue;
 
 	if (!_major)
 		_major = r;
 
 	return 0;
+
+out_free_workqueue:
+	destroy_workqueue(deferred_remove_workqueue);
+out_uevent_exit:
+	dm_uevent_exit();
+out_free_rq_tio_cache:
+	kmem_cache_destroy(_rq_tio_cache);
+out_free_io_cache:
+	kmem_cache_destroy(_io_cache);
+
+	return r;
 }
 
 static void local_exit(void)
 {
-	kmem_cache_destroy(_tio_cache);
+	flush_scheduled_work();
+	destroy_workqueue(deferred_remove_workqueue);
+
+	kmem_cache_destroy(_rq_tio_cache);
 	kmem_cache_destroy(_io_cache);
 	unregister_blkdev(_major, _name);
 	dm_uevent_exit();
@@ -204,8 +326,10 @@ static int (*_inits[])(void) __initdata = {
 	dm_target_init,
 	dm_linear_init,
 	dm_stripe_init,
+	dm_io_init,
 	dm_kcopyd_init,
 	dm_interface_init,
+	dm_statistics_init,
 };
 
 static void (*_exits[])(void) = {
@@ -213,8 +337,10 @@ static void (*_exits[])(void) = {
 	dm_target_exit,
 	dm_linear_exit,
 	dm_stripe_exit,
+	dm_io_exit,
 	dm_kcopyd_exit,
 	dm_interface_exit,
+	dm_statistics_exit,
 };
 
 static int __init dm_init(void)
@@ -244,23 +370,33 @@ static void __exit dm_exit(void)
 
 	while (i--)
 		_exits[i]();
+
+	/*
+	 * Should be empty by this point.
+	 */
+	idr_destroy(&_minor_idr);
 }
 
 /*
  * Block device functions
  */
-static int dm_blk_open(struct inode *inode, struct file *file)
+int dm_deleting_md(struct mapped_device *md)
+{
+	return test_bit(DMF_DELETING, &md->flags);
+}
+
+static int dm_blk_open(struct block_device *bdev, fmode_t mode)
 {
 	struct mapped_device *md;
 
 	spin_lock(&_minor_lock);
 
-	md = inode->i_bdev->bd_disk->private_data;
+	md = bdev->bd_disk->private_data;
 	if (!md)
 		goto out;
 
 	if (test_bit(DMF_FREEING, &md->flags) ||
-	    test_bit(DMF_DELETING, &md->flags)) {
+	    dm_deleting_md(md)) {
 		md = NULL;
 		goto out;
 	}
@@ -274,14 +410,19 @@ out:
 	return md ? 0 : -ENXIO;
 }
 
-static int dm_blk_close(struct inode *inode, struct file *file)
+static void dm_blk_close(struct gendisk *disk, fmode_t mode)
 {
-	struct mapped_device *md;
+	struct mapped_device *md = disk->private_data;
+
+	spin_lock(&_minor_lock);
+
+	if (atomic_dec_and_test(&md->open_count) &&
+	    (test_bit(DMF_DEFERRED_REMOVE, &md->flags)))
+		queue_work(deferred_remove_workqueue, &deferred_remove_work);
 
-	md = inode->i_bdev->bd_disk->private_data;
-	atomic_dec(&md->open_count);
 	dm_put(md);
-	return 0;
+
+	spin_unlock(&_minor_lock);
 }
 
 int dm_open_count(struct mapped_device *md)
@@ -292,14 +433,18 @@ int dm_open_count(struct mapped_device *md)
 /*
  * Guarantees nothing is using the device before it's deleted.
  */
-int dm_lock_for_deletion(struct mapped_device *md)
+int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred)
 {
 	int r = 0;
 
 	spin_lock(&_minor_lock);
 
-	if (dm_open_count(md))
+	if (dm_open_count(md)) {
 		r = -EBUSY;
+		if (mark_deferred)
+			set_bit(DMF_DEFERRED_REMOVE, &md->flags);
+	} else if (only_deferred && !test_bit(DMF_DEFERRED_REMOVE, &md->flags))
+		r = -EEXIST;
 	else
 		set_bit(DMF_DELETING, &md->flags);
 
@@ -308,6 +453,42 @@ int dm_lock_for_deletion(struct mapped_device *md)
 	return r;
 }
 
+int dm_cancel_deferred_remove(struct mapped_device *md)
+{
+	int r = 0;
+
+	spin_lock(&_minor_lock);
+
+	if (test_bit(DMF_DELETING, &md->flags))
+		r = -EBUSY;
+	else
+		clear_bit(DMF_DEFERRED_REMOVE, &md->flags);
+
+	spin_unlock(&_minor_lock);
+
+	return r;
+}
+
+static void do_deferred_remove(struct work_struct *w)
+{
+	dm_deferred_remove();
+}
+
+sector_t dm_get_size(struct mapped_device *md)
+{
+	return get_capacity(md->disk);
+}
+
+struct request_queue *dm_get_md_queue(struct mapped_device *md)
+{
+	return md->queue;
+}
+
+struct dm_stats *dm_get_stats(struct mapped_device *md)
+{
+	return &md->stats;
+}
+
 static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 {
 	struct mapped_device *md = bdev->bd_disk->private_data;
@@ -315,20 +496,17 @@ static int dm_blk_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 	return dm_get_geometry(md, geo);
 }
 
-static int dm_blk_ioctl(struct inode *inode, struct file *file,
+static int dm_blk_ioctl(struct block_device *bdev, fmode_t mode,
 			unsigned int cmd, unsigned long arg)
 {
-	struct mapped_device *md;
+	struct mapped_device *md = bdev->bd_disk->private_data;
+	int srcu_idx;
 	struct dm_table *map;
 	struct dm_target *tgt;
 	int r = -ENOTTY;
 
-	/* We don't really need this lock, but we do need 'inode'. */
-	unlock_kernel();
-
-	md = inode->i_bdev->bd_disk->private_data;
-
-	map = dm_get_table(md);
+retry:
+	map = dm_get_live_table(md, &srcu_idx);
 
 	if (!map || !dm_table_get_size(map))
 		goto out;
@@ -339,18 +517,22 @@ static int dm_blk_ioctl(struct inode *inode, struct file *file,
 
 	tgt = dm_table_get_target(map, 0);
 
-	if (dm_suspended(md)) {
+	if (dm_suspended_md(md)) {
 		r = -EAGAIN;
 		goto out;
 	}
 
 	if (tgt->type->ioctl)
-		r = tgt->type->ioctl(tgt, inode, file, cmd, arg);
+		r = tgt->type->ioctl(tgt, cmd, arg);
 
 out:
-	dm_table_put(map);
+	dm_put_live_table(md, srcu_idx);
+
+	if (r == -ENOTCONN) {
+		msleep(10);
+		goto retry;
+	}
 
-	lock_kernel();
 	return r;
 }
 
@@ -364,80 +546,127 @@ static void free_io(struct mapped_device *md, struct dm_io *io)
 	mempool_free(io, md->io_pool);
 }
 
-static struct dm_target_io *alloc_tio(struct mapped_device *md)
+static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
 {
-	return mempool_alloc(md->tio_pool, GFP_NOIO);
+	bio_put(&tio->clone);
 }
 
-static void free_tio(struct mapped_device *md, struct dm_target_io *tio)
+static struct dm_rq_target_io *alloc_rq_tio(struct mapped_device *md,
+					    gfp_t gfp_mask)
+{
+	return mempool_alloc(md->io_pool, gfp_mask);
+}
+
+static void free_rq_tio(struct dm_rq_target_io *tio)
+{
+	mempool_free(tio, tio->md->io_pool);
+}
+
+static int md_in_flight(struct mapped_device *md)
 {
-	mempool_free(tio, md->tio_pool);
+	return atomic_read(&md->pending[READ]) +
+	       atomic_read(&md->pending[WRITE]);
 }
 
 static void start_io_acct(struct dm_io *io)
 {
 	struct mapped_device *md = io->md;
+	struct bio *bio = io->bio;
+	int cpu;
+	int rw = bio_data_dir(bio);
 
 	io->start_time = jiffies;
 
-	preempt_disable();
-	disk_round_stats(dm_disk(md));
-	preempt_enable();
-	dm_disk(md)->in_flight = atomic_inc_return(&md->pending);
+	cpu = part_stat_lock();
+	part_round_stats(cpu, &dm_disk(md)->part0);
+	part_stat_unlock();
+	atomic_set(&dm_disk(md)->part0.in_flight[rw],
+		atomic_inc_return(&md->pending[rw]));
+
+	if (unlikely(dm_stats_used(&md->stats)))
+		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
+				    bio_sectors(bio), false, 0, &io->stats_aux);
 }
 
-static int end_io_acct(struct dm_io *io)
+static void end_io_acct(struct dm_io *io)
 {
 	struct mapped_device *md = io->md;
 	struct bio *bio = io->bio;
 	unsigned long duration = jiffies - io->start_time;
-	int pending;
+	int pending, cpu;
 	int rw = bio_data_dir(bio);
 
-	preempt_disable();
-	disk_round_stats(dm_disk(md));
-	preempt_enable();
-	dm_disk(md)->in_flight = pending = atomic_dec_return(&md->pending);
+	cpu = part_stat_lock();
+	part_round_stats(cpu, &dm_disk(md)->part0);
+	part_stat_add(cpu, &dm_disk(md)->part0, ticks[rw], duration);
+	part_stat_unlock();
 
-	disk_stat_add(dm_disk(md), ticks[rw], duration);
+	if (unlikely(dm_stats_used(&md->stats)))
+		dm_stats_account_io(&md->stats, bio->bi_rw, bio->bi_iter.bi_sector,
+				    bio_sectors(bio), true, duration, &io->stats_aux);
+
+	/*
+	 * After this is decremented the bio must not be touched if it is
+	 * a flush.
+	 */
+	pending = atomic_dec_return(&md->pending[rw]);
+	atomic_set(&dm_disk(md)->part0.in_flight[rw], pending);
+	pending += atomic_read(&md->pending[rw^0x1]);
 
-	return !pending;
+	/* nudge anyone waiting on suspend queue */
+	if (!pending)
+		wake_up(&md->wait);
 }
 
 /*
  * Add the bio to the list of deferred io.
  */
-static int queue_io(struct mapped_device *md, struct bio *bio)
+static void queue_io(struct mapped_device *md, struct bio *bio)
 {
-	down_write(&md->io_lock);
-
-	if (!test_bit(DMF_BLOCK_IO, &md->flags)) {
-		up_write(&md->io_lock);
-		return 1;
-	}
+	unsigned long flags;
 
+	spin_lock_irqsave(&md->deferred_lock, flags);
 	bio_list_add(&md->deferred, bio);
-
-	up_write(&md->io_lock);
-	return 0;		/* deferred successfully */
+	spin_unlock_irqrestore(&md->deferred_lock, flags);
+	queue_work(md->wq, &md->work);
 }
 
 /*
  * Everyone (including functions in this file), should use this
  * function to access the md->map field, and make sure they call
- * dm_table_put() when finished.
+ * dm_put_live_table() when finished.
  */
-struct dm_table *dm_get_table(struct mapped_device *md)
+struct dm_table *dm_get_live_table(struct mapped_device *md, int *srcu_idx) __acquires(md->io_barrier)
 {
-	struct dm_table *t;
+	*srcu_idx = srcu_read_lock(&md->io_barrier);
 
-	read_lock(&md->map_lock);
-	t = md->map;
-	if (t)
-		dm_table_get(t);
-	read_unlock(&md->map_lock);
+	return srcu_dereference(md->map, &md->io_barrier);
+}
 
-	return t;
+void dm_put_live_table(struct mapped_device *md, int srcu_idx) __releases(md->io_barrier)
+{
+	srcu_read_unlock(&md->io_barrier, srcu_idx);
+}
+
+void dm_sync_table(struct mapped_device *md)
+{
+	synchronize_srcu(&md->io_barrier);
+	synchronize_rcu_expedited();
+}
+
+/*
+ * A fast alternative to dm_get_live_table/dm_put_live_table.
+ * The caller must not block between these two functions.
+ */
+static struct dm_table *dm_get_live_table_fast(struct mapped_device *md) __acquires(RCU)
+{
+	rcu_read_lock();
+	return rcu_dereference(md->map);
+}
+
+static void dm_put_live_table_fast(struct mapped_device *md) __releases(RCU)
+{
+	rcu_read_unlock();
 }
 
 /*
@@ -488,46 +717,68 @@ static int __noflush_suspending(struct mapped_device *md)
 static void dec_pending(struct dm_io *io, int error)
 {
 	unsigned long flags;
+	int io_error;
+	struct bio *bio;
+	struct mapped_device *md = io->md;
 
 	/* Push-back supersedes any I/O errors */
-	if (error && !(io->error > 0 && __noflush_suspending(io->md)))
-		io->error = error;
+	if (unlikely(error)) {
+		spin_lock_irqsave(&io->endio_lock, flags);
+		if (!(io->error > 0 && __noflush_suspending(md)))
+			io->error = error;
+		spin_unlock_irqrestore(&io->endio_lock, flags);
+	}
 
 	if (atomic_dec_and_test(&io->io_count)) {
 		if (io->error == DM_ENDIO_REQUEUE) {
 			/*
 			 * Target requested pushing back the I/O.
-			 * This must be handled before the sleeper on
-			 * suspend queue merges the pushback list.
 			 */
-			spin_lock_irqsave(&io->md->pushback_lock, flags);
-			if (__noflush_suspending(io->md))
-				bio_list_add(&io->md->pushback, io->bio);
+			spin_lock_irqsave(&md->deferred_lock, flags);
+			if (__noflush_suspending(md))
+				bio_list_add_head(&md->deferred, io->bio);
 			else
 				/* noflush suspend was interrupted. */
 				io->error = -EIO;
-			spin_unlock_irqrestore(&io->md->pushback_lock, flags);
+			spin_unlock_irqrestore(&md->deferred_lock, flags);
 		}
 
-		if (end_io_acct(io))
-			/* nudge anyone waiting on suspend queue */
-			wake_up(&io->md->wait);
+		io_error = io->error;
+		bio = io->bio;
+		end_io_acct(io);
+		free_io(md, io);
 
-		if (io->error != DM_ENDIO_REQUEUE) {
-			blk_add_trace_bio(io->md->queue, io->bio,
-					  BLK_TA_COMPLETE);
+		if (io_error == DM_ENDIO_REQUEUE)
+			return;
 
-			bio_endio(io->bio, io->error);
+		if ((bio->bi_rw & REQ_FLUSH) && bio->bi_iter.bi_size) {
+			/*
+			 * Preflush done for flush with data, reissue
+			 * without REQ_FLUSH.
+			 */
+			bio->bi_rw &= ~REQ_FLUSH;
+			queue_io(md, bio);
+		} else {
+			/* done with normal IO or empty flush */
+			trace_block_bio_complete(md->queue, bio, io_error);
+			bio_endio(bio, io_error);
 		}
-
-		free_io(io->md, io);
 	}
 }
 
+static void disable_write_same(struct mapped_device *md)
+{
+	struct queue_limits *limits = dm_get_queue_limits(md);
+
+	/* device doesn't really support WRITE SAME, disable it */
+	limits->max_write_same_sectors = 0;
+}
+
 static void clone_endio(struct bio *bio, int error)
 {
 	int r = 0;
-	struct dm_target_io *tio = bio->bi_private;
+	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
+	struct dm_io *io = tio->io;
 	struct mapped_device *md = tio->io->md;
 	dm_endio_fn endio = tio->ti->type->end_io;
 
@@ -535,7 +786,7 @@ static void clone_endio(struct bio *bio, int error)
 		error = -EIO;
 
 	if (endio) {
-		r = endio(tio->ti, bio, error, &tio->info);
+		r = endio(tio->ti, bio, error);
 		if (r < 0 || r == DM_ENDIO_REQUEUE)
 			/*
 			 * error and requeue request are handled
@@ -551,51 +802,390 @@ static void clone_endio(struct bio *bio, int error)
 		}
 	}
 
-	dec_pending(tio->io, error);
+	if (unlikely(r == -EREMOTEIO && (bio->bi_rw & REQ_WRITE_SAME) &&
+		     !bdev_get_queue(bio->bi_bdev)->limits.max_write_same_sectors))
+		disable_write_same(md);
+
+	free_tio(md, tio);
+	dec_pending(io, error);
+}
+
+/*
+ * Partial completion handling for request-based dm
+ */
+static void end_clone_bio(struct bio *clone, int error)
+{
+	struct dm_rq_clone_bio_info *info =
+		container_of(clone, struct dm_rq_clone_bio_info, clone);
+	struct dm_rq_target_io *tio = info->tio;
+	struct bio *bio = info->orig;
+	unsigned int nr_bytes = info->orig->bi_iter.bi_size;
+
+	bio_put(clone);
+
+	if (tio->error)
+		/*
+		 * An error has already been detected on the request.
+		 * Once error occurred, just let clone->end_io() handle
+		 * the remainder.
+		 */
+		return;
+	else if (error) {
+		/*
+		 * Don't notice the error to the upper layer yet.
+		 * The error handling decision is made by the target driver,
+		 * when the request is completed.
+		 */
+		tio->error = error;
+		return;
+	}
 
 	/*
-	 * Store md for cleanup instead of tio which is about to get freed.
+	 * I/O for the bio successfully completed.
+	 * Notice the data completion to the upper layer.
 	 */
-	bio->bi_private = md->bs;
 
-	bio_put(bio);
-	free_tio(md, tio);
+	/*
+	 * bios are processed from the head of the list.
+	 * So the completing bio should always be rq->bio.
+	 * If it's not, something wrong is happening.
+	 */
+	if (tio->orig->bio != bio)
+		DMERR("bio completion is going in the middle of the request");
+
+	/*
+	 * Update the original request.
+	 * Do not use blk_end_request() here, because it may complete
+	 * the original request before the clone, and break the ordering.
+	 */
+	blk_update_request(tio->orig, 0, nr_bytes);
+}
+
+/*
+ * Don't touch any member of the md after calling this function because
+ * the md may be freed in dm_put() at the end of this function.
+ * Or do dm_get() before calling this function and dm_put() later.
+ */
+static void rq_completed(struct mapped_device *md, int rw, int run_queue)
+{
+	atomic_dec(&md->pending[rw]);
+
+	/* nudge anyone waiting on suspend queue */
+	if (!md_in_flight(md))
+		wake_up(&md->wait);
+
+	/*
+	 * Run this off this callpath, as drivers could invoke end_io while
+	 * inside their request_fn (and holding the queue lock). Calling
+	 * back into ->request_fn() could deadlock attempting to grab the
+	 * queue lock again.
+	 */
+	if (run_queue)
+		blk_run_queue_async(md->queue);
+
+	/*
+	 * dm_put() must be at the end of this function. See the comment above
+	 */
+	dm_put(md);
+}
+
+static void free_rq_clone(struct request *clone)
+{
+	struct dm_rq_target_io *tio = clone->end_io_data;
+
+	blk_rq_unprep_clone(clone);
+	free_rq_tio(tio);
+}
+
+/*
+ * Complete the clone and the original request.
+ * Must be called without queue lock.
+ */
+static void dm_end_request(struct request *clone, int error)
+{
+	int rw = rq_data_dir(clone);
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	struct mapped_device *md = tio->md;
+	struct request *rq = tio->orig;
+
+	if (rq->cmd_type == REQ_TYPE_BLOCK_PC) {
+		rq->errors = clone->errors;
+		rq->resid_len = clone->resid_len;
+
+		if (rq->sense)
+			/*
+			 * We are using the sense buffer of the original
+			 * request.
+			 * So setting the length of the sense data is enough.
+			 */
+			rq->sense_len = clone->sense_len;
+	}
+
+	free_rq_clone(clone);
+	blk_end_request_all(rq, error);
+	rq_completed(md, rw, true);
+}
+
+static void dm_unprep_request(struct request *rq)
+{
+	struct request *clone = rq->special;
+
+	rq->special = NULL;
+	rq->cmd_flags &= ~REQ_DONTPREP;
+
+	free_rq_clone(clone);
+}
+
+/*
+ * Requeue the original request of a clone.
+ */
+void dm_requeue_unmapped_request(struct request *clone)
+{
+	int rw = rq_data_dir(clone);
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	struct mapped_device *md = tio->md;
+	struct request *rq = tio->orig;
+	struct request_queue *q = rq->q;
+	unsigned long flags;
+
+	dm_unprep_request(rq);
+
+	spin_lock_irqsave(q->queue_lock, flags);
+	blk_requeue_request(q, rq);
+	spin_unlock_irqrestore(q->queue_lock, flags);
+
+	rq_completed(md, rw, 0);
+}
+EXPORT_SYMBOL_GPL(dm_requeue_unmapped_request);
+
+static void __stop_queue(struct request_queue *q)
+{
+	blk_stop_queue(q);
+}
+
+static void stop_queue(struct request_queue *q)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+	__stop_queue(q);
+	spin_unlock_irqrestore(q->queue_lock, flags);
+}
+
+static void __start_queue(struct request_queue *q)
+{
+	if (blk_queue_stopped(q))
+		blk_start_queue(q);
+}
+
+static void start_queue(struct request_queue *q)
+{
+	unsigned long flags;
+
+	spin_lock_irqsave(q->queue_lock, flags);
+	__start_queue(q);
+	spin_unlock_irqrestore(q->queue_lock, flags);
+}
+
+static void dm_done(struct request *clone, int error, bool mapped)
+{
+	int r = error;
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	dm_request_endio_fn rq_end_io = NULL;
+
+	if (tio->ti) {
+		rq_end_io = tio->ti->type->rq_end_io;
+
+		if (mapped && rq_end_io)
+			r = rq_end_io(tio->ti, clone, error, &tio->info);
+	}
+
+	if (unlikely(r == -EREMOTEIO && (clone->cmd_flags & REQ_WRITE_SAME) &&
+		     !clone->q->limits.max_write_same_sectors))
+		disable_write_same(tio->md);
+
+	if (r <= 0)
+		/* The target wants to complete the I/O */
+		dm_end_request(clone, r);
+	else if (r == DM_ENDIO_INCOMPLETE)
+		/* The target will handle the I/O */
+		return;
+	else if (r == DM_ENDIO_REQUEUE)
+		/* The target wants to requeue the I/O */
+		dm_requeue_unmapped_request(clone);
+	else {
+		DMWARN("unimplemented target endio return value: %d", r);
+		BUG();
+	}
+}
+
+/*
+ * Request completion handler for request-based dm
+ */
+static void dm_softirq_done(struct request *rq)
+{
+	bool mapped = true;
+	struct request *clone = rq->completion_data;
+	struct dm_rq_target_io *tio = clone->end_io_data;
+
+	if (rq->cmd_flags & REQ_FAILED)
+		mapped = false;
+
+	dm_done(clone, tio->error, mapped);
+}
+
+/*
+ * Complete the clone and the original request with the error status
+ * through softirq context.
+ */
+static void dm_complete_request(struct request *clone, int error)
+{
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	struct request *rq = tio->orig;
+
+	tio->error = error;
+	rq->completion_data = clone;
+	blk_complete_request(rq);
+}
+
+/*
+ * Complete the not-mapped clone and the original request with the error status
+ * through softirq context.
+ * Target's rq_end_io() function isn't called.
+ * This may be used when the target's map_rq() function fails.
+ */
+void dm_kill_unmapped_request(struct request *clone, int error)
+{
+	struct dm_rq_target_io *tio = clone->end_io_data;
+	struct request *rq = tio->orig;
+
+	rq->cmd_flags |= REQ_FAILED;
+	dm_complete_request(clone, error);
+}
+EXPORT_SYMBOL_GPL(dm_kill_unmapped_request);
+
+/*
+ * Called with the queue lock held
+ */
+static void end_clone_request(struct request *clone, int error)
+{
+	/*
+	 * For just cleaning up the information of the queue in which
+	 * the clone was dispatched.
+	 * The clone is *NOT* freed actually here because it is alloced from
+	 * dm own mempool and REQ_ALLOCED isn't set in clone->cmd_flags.
+	 */
+	__blk_put_request(clone->q, clone);
+
+	/*
+	 * Actual request completion is done in a softirq context which doesn't
+	 * hold the queue lock.  Otherwise, deadlock could occur because:
+	 *     - another request may be submitted by the upper level driver
+	 *       of the stacking during the completion
+	 *     - the submission which requires queue lock may be done
+	 *       against this queue
+	 */
+	dm_complete_request(clone, error);
+}
+
+/*
+ * Return maximum size of I/O possible at the supplied sector up to the current
+ * target boundary.
+ */
+static sector_t max_io_len_target_boundary(sector_t sector, struct dm_target *ti)
+{
+	sector_t target_offset = dm_target_offset(ti, sector);
+
+	return ti->len - target_offset;
 }
 
-static sector_t max_io_len(struct mapped_device *md,
-			   sector_t sector, struct dm_target *ti)
+static sector_t max_io_len(sector_t sector, struct dm_target *ti)
 {
-	sector_t offset = sector - ti->begin;
-	sector_t len = ti->len - offset;
+	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;
-		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;
 }
 
-static void __map_bio(struct dm_target *ti, struct bio *clone,
-		      struct dm_target_io *tio)
+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);
+
+/*
+ * A target may call dm_accept_partial_bio only from the map routine.  It is
+ * allowed for all bio types except REQ_FLUSH.
+ *
+ * dm_accept_partial_bio informs the dm that the target only wants to process
+ * additional n_sectors sectors of the bio and the rest of the data should be
+ * sent in a next bio.
+ *
+ * A diagram that explains the arithmetics:
+ * +--------------------+---------------+-------+
+ * |         1          |       2       |   3   |
+ * +--------------------+---------------+-------+
+ *
+ * <-------------- *tio->len_ptr --------------->
+ *                      <------- bi_size ------->
+ *                      <-- n_sectors -->
+ *
+ * Region 1 was already iterated over with bio_advance or similar function.
+ *	(it may be empty if the target doesn't use bio_advance)
+ * Region 2 is the remaining bio size that the target wants to process.
+ *	(it may be empty if region 1 is non-empty, although there is no reason
+ *	 to make it empty)
+ * The target requires that region 3 is to be sent in the next bio.
+ *
+ * If the target wants to receive multiple copies of the bio (via num_*bios, etc),
+ * the partially processed part (the sum of regions 1+2) must be the same for all
+ * copies of the bio.
+ */
+void dm_accept_partial_bio(struct bio *bio, unsigned n_sectors)
+{
+	struct dm_target_io *tio = container_of(bio, struct dm_target_io, clone);
+	unsigned bi_size = bio->bi_iter.bi_size >> SECTOR_SHIFT;
+	BUG_ON(bio->bi_rw & REQ_FLUSH);
+	BUG_ON(bi_size > *tio->len_ptr);
+	BUG_ON(n_sectors > bi_size);
+	*tio->len_ptr -= bi_size - n_sectors;
+	bio->bi_iter.bi_size = n_sectors << SECTOR_SHIFT;
+}
+EXPORT_SYMBOL_GPL(dm_accept_partial_bio);
+
+static void __map_bio(struct dm_target_io *tio)
 {
 	int r;
 	sector_t sector;
 	struct mapped_device *md;
-
-	/*
-	 * Sanity checks.
-	 */
-	BUG_ON(!clone->bi_size);
+	struct bio *clone = &tio->clone;
+	struct dm_target *ti = tio->ti;
 
 	clone->bi_end_io = clone_endio;
-	clone->bi_private = tio;
 
 	/*
 	 * Map the clone.  If r == 0 we don't need to do
@@ -603,25 +1193,19 @@ static void __map_bio(struct dm_target *ti, struct bio *clone,
 	 * this io.
 	 */
 	atomic_inc(&tio->io->io_count);
-	sector = clone->bi_sector;
-	r = ti->type->map(ti, clone, &tio->info);
+	sector = clone->bi_iter.bi_sector;
+	r = ti->type->map(ti, clone);
 	if (r == DM_MAPIO_REMAPPED) {
 		/* the bio has been remapped so dispatch it */
 
-		blk_add_trace_remap(bdev_get_queue(clone->bi_bdev), clone,
-				    tio->io->bio->bi_bdev->bd_dev,
-				    clone->bi_sector, sector);
+		trace_block_bio_remap(bdev_get_queue(clone->bi_bdev), clone,
+				      tio->io->bio->bi_bdev->bd_dev, sector);
 
 		generic_make_request(clone);
 	} else if (r < 0 || r == DM_MAPIO_REQUEUE) {
 		/* error the io and bail out, or requeue it if needed */
 		md = tio->io->md;
 		dec_pending(tio->io, r);
-		/*
-		 * Store bio_set for cleanup.
-		 */
-		clone->bi_private = md->bs;
-		bio_put(clone);
 		free_tio(md, tio);
 	} else if (r) {
 		DMWARN("unimplemented target map return value: %d", r);
@@ -635,195 +1219,250 @@ struct clone_info {
 	struct bio *bio;
 	struct dm_io *io;
 	sector_t sector;
-	sector_t sector_count;
-	unsigned short idx;
+	unsigned sector_count;
 };
 
-static void dm_bio_destructor(struct bio *bio)
+static void bio_setup_sector(struct bio *bio, sector_t sector, unsigned len)
 {
-	struct bio_set *bs = bio->bi_private;
-
-	bio_free(bio, bs);
+	bio->bi_iter.bi_sector = sector;
+	bio->bi_iter.bi_size = to_bytes(len);
 }
 
 /*
- * Creates a little bio that is just does part of a bvec.
+ * Creates a bio that consists of range of complete bvecs.
  */
-static struct bio *split_bvec(struct bio *bio, sector_t sector,
-			      unsigned short idx, unsigned int offset,
-			      unsigned int len, struct bio_set *bs)
+static void clone_bio(struct dm_target_io *tio, struct bio *bio,
+		      sector_t sector, unsigned len)
 {
-	struct bio *clone;
-	struct bio_vec *bv = bio->bi_io_vec + idx;
+	struct bio *clone = &tio->clone;
 
-	clone = bio_alloc_bioset(GFP_NOIO, 1, bs);
-	clone->bi_destructor = dm_bio_destructor;
-	*clone->bi_io_vec = *bv;
+	__bio_clone_fast(clone, bio);
 
-	clone->bi_sector = sector;
-	clone->bi_bdev = bio->bi_bdev;
-	clone->bi_rw = bio->bi_rw;
-	clone->bi_vcnt = 1;
-	clone->bi_size = to_bytes(len);
-	clone->bi_io_vec->bv_offset = offset;
-	clone->bi_io_vec->bv_len = clone->bi_size;
+	if (bio_integrity(bio))
+		bio_integrity_clone(clone, bio, GFP_NOIO);
 
-	return clone;
+	bio_advance(clone, to_bytes(sector - clone->bi_iter.bi_sector));
+	clone->bi_iter.bi_size = to_bytes(len);
+
+	if (bio_integrity(bio))
+		bio_integrity_trim(clone, 0, len);
 }
 
-/*
- * Creates a bio that consists of range of complete bvecs.
- */
-static struct bio *clone_bio(struct bio *bio, sector_t sector,
-			     unsigned short idx, unsigned short bv_count,
-			     unsigned int len, struct bio_set *bs)
+static struct dm_target_io *alloc_tio(struct clone_info *ci,
+				      struct dm_target *ti, int nr_iovecs,
+				      unsigned target_bio_nr)
 {
+	struct dm_target_io *tio;
 	struct bio *clone;
 
-	clone = bio_alloc_bioset(GFP_NOIO, bio->bi_max_vecs, bs);
-	__bio_clone(clone, bio);
-	clone->bi_destructor = dm_bio_destructor;
-	clone->bi_sector = sector;
-	clone->bi_idx = idx;
-	clone->bi_vcnt = idx + bv_count;
-	clone->bi_size = to_bytes(len);
-	clone->bi_flags &= ~(1 << BIO_SEG_VALID);
+	clone = bio_alloc_bioset(GFP_NOIO, nr_iovecs, ci->md->bs);
+	tio = container_of(clone, struct dm_target_io, clone);
 
-	return clone;
+	tio->io = ci->io;
+	tio->ti = ti;
+	tio->target_bio_nr = target_bio_nr;
+
+	return tio;
 }
 
-static int __clone_and_map(struct clone_info *ci)
+static void __clone_and_map_simple_bio(struct clone_info *ci,
+				       struct dm_target *ti,
+				       unsigned target_bio_nr, unsigned *len)
 {
-	struct bio *clone, *bio = ci->bio;
+	struct dm_target_io *tio = alloc_tio(ci, ti, ci->bio->bi_max_vecs, target_bio_nr);
+	struct bio *clone = &tio->clone;
+
+	tio->len_ptr = len;
+
+	/*
+	 * Discard requests require the bio's inline iovecs be initialized.
+	 * ci->bio->bi_max_vecs is BIO_INLINE_VECS anyway, for both flush
+	 * and discard, so no need for concern about wasted bvec allocations.
+	 */
+	 __bio_clone_fast(clone, ci->bio);
+	if (len)
+		bio_setup_sector(clone, ci->sector, *len);
+
+	__map_bio(tio);
+}
+
+static void __send_duplicate_bios(struct clone_info *ci, struct dm_target *ti,
+				  unsigned num_bios, unsigned *len)
+{
+	unsigned target_bio_nr;
+
+	for (target_bio_nr = 0; target_bio_nr < num_bios; target_bio_nr++)
+		__clone_and_map_simple_bio(ci, ti, target_bio_nr, len);
+}
+
+static int __send_empty_flush(struct clone_info *ci)
+{
+	unsigned target_nr = 0;
 	struct dm_target *ti;
-	sector_t len = 0, max;
-	struct dm_target_io *tio;
 
-	ti = dm_table_find_target(ci->map, ci->sector);
-	if (!dm_target_is_valid(ti))
-		return -EIO;
+	BUG_ON(bio_has_data(ci->bio));
+	while ((ti = dm_table_get_target(ci->map, target_nr++)))
+		__send_duplicate_bios(ci, ti, ti->num_flush_bios, NULL);
+
+	return 0;
+}
 
-	max = max_io_len(ci->md, ci->sector, ti);
+static void __clone_and_map_data_bio(struct clone_info *ci, struct dm_target *ti,
+				     sector_t sector, unsigned *len)
+{
+	struct bio *bio = ci->bio;
+	struct dm_target_io *tio;
+	unsigned target_bio_nr;
+	unsigned num_target_bios = 1;
 
 	/*
-	 * Allocate a target io object.
+	 * Does the target want to receive duplicate copies of the bio?
 	 */
-	tio = alloc_tio(ci->md);
-	tio->io = ci->io;
-	tio->ti = ti;
-	memset(&tio->info, 0, sizeof(tio->info));
+	if (bio_data_dir(bio) == WRITE && ti->num_write_bios)
+		num_target_bios = ti->num_write_bios(ti, bio);
+
+	for (target_bio_nr = 0; target_bio_nr < num_target_bios; target_bio_nr++) {
+		tio = alloc_tio(ci, ti, 0, target_bio_nr);
+		tio->len_ptr = len;
+		clone_bio(tio, bio, sector, *len);
+		__map_bio(tio);
+	}
+}
 
-	if (ci->sector_count <= max) {
-		/*
-		 * Optimise for the simple case where we can do all of
-		 * the remaining io with a single clone.
-		 */
-		clone = clone_bio(bio, ci->sector, ci->idx,
-				  bio->bi_vcnt - ci->idx, ci->sector_count,
-				  ci->md->bs);
-		__map_bio(ti, clone, tio);
-		ci->sector_count = 0;
+typedef unsigned (*get_num_bios_fn)(struct dm_target *ti);
 
-	} else if (to_sector(bio->bi_io_vec[ci->idx].bv_len) <= max) {
-		/*
-		 * There are some bvecs that don't span targets.
-		 * Do as many of these as possible.
-		 */
-		int i;
-		sector_t remaining = max;
-		sector_t bv_len;
+static unsigned get_num_discard_bios(struct dm_target *ti)
+{
+	return ti->num_discard_bios;
+}
 
-		for (i = ci->idx; remaining && (i < bio->bi_vcnt); i++) {
-			bv_len = to_sector(bio->bi_io_vec[i].bv_len);
+static unsigned get_num_write_same_bios(struct dm_target *ti)
+{
+	return ti->num_write_same_bios;
+}
 
-			if (bv_len > remaining)
-				break;
+typedef bool (*is_split_required_fn)(struct dm_target *ti);
 
-			remaining -= bv_len;
-			len += bv_len;
-		}
+static bool is_split_required_for_discard(struct dm_target *ti)
+{
+	return ti->split_discard_bios;
+}
 
-		clone = clone_bio(bio, ci->sector, ci->idx, i - ci->idx, len,
-				  ci->md->bs);
-		__map_bio(ti, clone, tio);
+static int __send_changing_extent_only(struct clone_info *ci,
+				       get_num_bios_fn get_num_bios,
+				       is_split_required_fn is_split_required)
+{
+	struct dm_target *ti;
+	unsigned len;
+	unsigned num_bios;
 
-		ci->sector += len;
-		ci->sector_count -= len;
-		ci->idx = i;
+	do {
+		ti = dm_table_find_target(ci->map, ci->sector);
+		if (!dm_target_is_valid(ti))
+			return -EIO;
 
-	} else {
 		/*
-		 * Handle a bvec that must be split between two or more targets.
+		 * Even though the device advertised support for this type of
+		 * request, that does not mean every target supports it, and
+		 * reconfiguration might also have changed that since the
+		 * check was performed.
 		 */
-		struct bio_vec *bv = bio->bi_io_vec + ci->idx;
-		sector_t remaining = to_sector(bv->bv_len);
-		unsigned int offset = 0;
+		num_bios = get_num_bios ? get_num_bios(ti) : 0;
+		if (!num_bios)
+			return -EOPNOTSUPP;
+
+		if (is_split_required && !is_split_required(ti))
+			len = min((sector_t)ci->sector_count, max_io_len_target_boundary(ci->sector, ti));
+		else
+			len = min((sector_t)ci->sector_count, max_io_len(ci->sector, ti));
+
+		__send_duplicate_bios(ci, ti, num_bios, &len);
+
+		ci->sector += len;
+	} while (ci->sector_count -= len);
 
-		do {
-			if (offset) {
-				ti = dm_table_find_target(ci->map, ci->sector);
-				if (!dm_target_is_valid(ti))
-					return -EIO;
+	return 0;
+}
 
-				max = max_io_len(ci->md, ci->sector, ti);
+static int __send_discard(struct clone_info *ci)
+{
+	return __send_changing_extent_only(ci, get_num_discard_bios,
+					   is_split_required_for_discard);
+}
 
-				tio = alloc_tio(ci->md);
-				tio->io = ci->io;
-				tio->ti = ti;
-				memset(&tio->info, 0, sizeof(tio->info));
-			}
+static int __send_write_same(struct clone_info *ci)
+{
+	return __send_changing_extent_only(ci, get_num_write_same_bios, NULL);
+}
 
-			len = min(remaining, max);
+/*
+ * Select the correct strategy for processing a non-flush bio.
+ */
+static int __split_and_process_non_flush(struct clone_info *ci)
+{
+	struct bio *bio = ci->bio;
+	struct dm_target *ti;
+	unsigned len;
 
-			clone = split_bvec(bio, ci->sector, ci->idx,
-					   bv->bv_offset + offset, len,
-					   ci->md->bs);
+	if (unlikely(bio->bi_rw & REQ_DISCARD))
+		return __send_discard(ci);
+	else if (unlikely(bio->bi_rw & REQ_WRITE_SAME))
+		return __send_write_same(ci);
 
-			__map_bio(ti, clone, tio);
+	ti = dm_table_find_target(ci->map, ci->sector);
+	if (!dm_target_is_valid(ti))
+		return -EIO;
 
-			ci->sector += len;
-			ci->sector_count -= len;
-			offset += to_bytes(len);
-		} while (remaining -= len);
+	len = min_t(sector_t, max_io_len(ci->sector, ti), ci->sector_count);
 
-		ci->idx++;
-	}
+	__clone_and_map_data_bio(ci, ti, ci->sector, &len);
+
+	ci->sector += len;
+	ci->sector_count -= len;
 
 	return 0;
 }
 
 /*
- * Split the bio into several clones.
+ * Entry point to split a bio into clones and submit them to the targets.
  */
-static int __split_bio(struct mapped_device *md, struct bio *bio)
+static void __split_and_process_bio(struct mapped_device *md,
+				    struct dm_table *map, struct bio *bio)
 {
 	struct clone_info ci;
 	int error = 0;
 
-	ci.map = dm_get_table(md);
-	if (unlikely(!ci.map))
-		return -EIO;
+	if (unlikely(!map)) {
+		bio_io_error(bio);
+		return;
+	}
 
+	ci.map = map;
 	ci.md = md;
-	ci.bio = bio;
 	ci.io = alloc_io(md);
 	ci.io->error = 0;
 	atomic_set(&ci.io->io_count, 1);
 	ci.io->bio = bio;
 	ci.io->md = md;
-	ci.sector = bio->bi_sector;
-	ci.sector_count = bio_sectors(bio);
-	ci.idx = bio->bi_idx;
+	spin_lock_init(&ci.io->endio_lock);
+	ci.sector = bio->bi_iter.bi_sector;
 
 	start_io_acct(ci.io);
-	while (ci.sector_count && !error)
-		error = __clone_and_map(&ci);
+
+	if (bio->bi_rw & REQ_FLUSH) {
+		ci.bio = &ci.md->flush_bio;
+		ci.sector_count = 0;
+		error = __send_empty_flush(&ci);
+		/* dec_pending submits any data associated with flush */
+	} else {
+		ci.bio = bio;
+		ci.sector_count = bio_sectors(bio);
+		while (ci.sector_count && !error)
+			error = __split_and_process_non_flush(&ci);
+	}
 
 	/* drop the extra reference count */
 	dec_pending(ci.io, error);
-	dm_table_put(ci.map);
-
-	return 0;
 }
 /*-----------------------------------------------------------------
  * CRUD END
@@ -834,20 +1473,22 @@ static int dm_merge_bvec(struct request_queue *q,
 			 struct bio_vec *biovec)
 {
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table_fast(md);
 	struct dm_target *ti;
 	sector_t max_sectors;
-	int max_size;
+	int max_size = 0;
 
 	if (unlikely(!map))
-		return 0;
+		goto out;
 
 	ti = dm_table_find_target(map, bvm->bi_sector);
+	if (!dm_target_is_valid(ti))
+		goto out;
 
 	/*
 	 * Find maximum amount of I/O that won't need splitting
 	 */
-	max_sectors = min(max_io_len(md, bvm->bi_sector, ti),
+	max_sectors = min(max_io_len(bvm->bi_sector, ti),
 			  (sector_t) BIO_MAX_SECTORS);
 	max_size = (max_sectors << SECTOR_SHIFT) - bvm->bi_size;
 	if (max_size < 0)
@@ -860,15 +1501,24 @@ static int dm_merge_bvec(struct request_queue *q,
 	 */
 	if (max_size && ti->type->merge)
 		max_size = ti->type->merge(ti, bvm, biovec, max_size);
+	/*
+	 * If the target doesn't support merge method and some of the devices
+	 * provided their merge_bvec method (we know this by looking at
+	 * queue_max_hw_sectors), then we can't allow bios with multiple vector
+	 * entries.  So always set max_size to 0, and the code below allows
+	 * just one page.
+	 */
+	else if (queue_max_hw_sectors(q) <= PAGE_SIZE >> 9)
+		max_size = 0;
 
+out:
+	dm_put_live_table_fast(md);
 	/*
 	 * Always allow an entire first page
 	 */
 	if (max_size <= biovec->bv_len && !(bvm->bi_size >> SECTOR_SHIFT))
 		max_size = biovec->bv_len;
 
-	dm_table_put(map);
-
 	return max_size;
 }
 
@@ -876,87 +1526,326 @@ static int dm_merge_bvec(struct request_queue *q,
  * The request function that just remaps the bio built up by
  * dm_merge_bvec.
  */
-static int dm_request(struct request_queue *q, struct bio *bio)
+static void _dm_request(struct request_queue *q, struct bio *bio)
 {
-	int r = -EIO;
 	int rw = bio_data_dir(bio);
 	struct mapped_device *md = q->queuedata;
+	int cpu;
+	int srcu_idx;
+	struct dm_table *map;
+
+	map = dm_get_live_table(md, &srcu_idx);
+
+	cpu = part_stat_lock();
+	part_stat_inc(cpu, &dm_disk(md)->part0, ios[rw]);
+	part_stat_add(cpu, &dm_disk(md)->part0, sectors[rw], bio_sectors(bio));
+	part_stat_unlock();
+
+	/* if we're suspended, we have to queue this io for later */
+	if (unlikely(test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))) {
+		dm_put_live_table(md, srcu_idx);
+
+		if (bio_rw(bio) != READA)
+			queue_io(md, bio);
+		else
+			bio_io_error(bio);
+		return;
+	}
+
+	__split_and_process_bio(md, map, bio);
+	dm_put_live_table(md, srcu_idx);
+	return;
+}
+
+int dm_request_based(struct mapped_device *md)
+{
+	return blk_queue_stackable(md->queue);
+}
+
+static void dm_request(struct request_queue *q, struct bio *bio)
+{
+	struct mapped_device *md = q->queuedata;
+
+	if (dm_request_based(md))
+		blk_queue_bio(q, bio);
+	else
+		_dm_request(q, bio);
+}
+
+void dm_dispatch_request(struct request *rq)
+{
+	int r;
+
+	if (blk_queue_io_stat(rq->q))
+		rq->cmd_flags |= REQ_IO_STAT;
+
+	rq->start_time = jiffies;
+	r = blk_insert_cloned_request(rq->q, rq);
+	if (r)
+		dm_complete_request(rq, r);
+}
+EXPORT_SYMBOL_GPL(dm_dispatch_request);
+
+static int dm_rq_bio_constructor(struct bio *bio, struct bio *bio_orig,
+				 void *data)
+{
+	struct dm_rq_target_io *tio = data;
+	struct dm_rq_clone_bio_info *info =
+		container_of(bio, struct dm_rq_clone_bio_info, clone);
+
+	info->orig = bio_orig;
+	info->tio = tio;
+	bio->bi_end_io = end_clone_bio;
+
+	return 0;
+}
+
+static int setup_clone(struct request *clone, struct request *rq,
+		       struct dm_rq_target_io *tio)
+{
+	int r;
+
+	r = blk_rq_prep_clone(clone, rq, tio->md->bs, GFP_ATOMIC,
+			      dm_rq_bio_constructor, tio);
+	if (r)
+		return r;
+
+	clone->cmd = rq->cmd;
+	clone->cmd_len = rq->cmd_len;
+	clone->sense = rq->sense;
+	clone->end_io = end_clone_request;
+	clone->end_io_data = tio;
+
+	return 0;
+}
+
+static struct request *clone_rq(struct request *rq, struct mapped_device *md,
+				gfp_t gfp_mask)
+{
+	struct request *clone;
+	struct dm_rq_target_io *tio;
+
+	tio = alloc_rq_tio(md, gfp_mask);
+	if (!tio)
+		return NULL;
+
+	tio->md = md;
+	tio->ti = NULL;
+	tio->orig = rq;
+	tio->error = 0;
+	memset(&tio->info, 0, sizeof(tio->info));
+
+	clone = &tio->clone;
+	if (setup_clone(clone, rq, tio)) {
+		/* -ENOMEM */
+		free_rq_tio(tio);
+		return NULL;
+	}
+
+	return clone;
+}
+
+/*
+ * Called with the queue lock held.
+ */
+static int dm_prep_fn(struct request_queue *q, struct request *rq)
+{
+	struct mapped_device *md = q->queuedata;
+	struct request *clone;
+
+	if (unlikely(rq->special)) {
+		DMWARN("Already has something in rq->special.");
+		return BLKPREP_KILL;
+	}
+
+	clone = clone_rq(rq, md, GFP_ATOMIC);
+	if (!clone)
+		return BLKPREP_DEFER;
+
+	rq->special = clone;
+	rq->cmd_flags |= REQ_DONTPREP;
+
+	return BLKPREP_OK;
+}
+
+/*
+ * Returns:
+ * 0  : the request has been processed (not requeued)
+ * !0 : the request has been requeued
+ */
+static int map_request(struct dm_target *ti, struct request *clone,
+		       struct mapped_device *md)
+{
+	int r, requeued = 0;
+	struct dm_rq_target_io *tio = clone->end_io_data;
+
+	tio->ti = ti;
+	r = ti->type->map_rq(ti, clone, &tio->info);
+	switch (r) {
+	case DM_MAPIO_SUBMITTED:
+		/* The target has taken the I/O to submit by itself later */
+		break;
+	case DM_MAPIO_REMAPPED:
+		/* The target has remapped the I/O so dispatch it */
+		trace_block_rq_remap(clone->q, clone, disk_devt(dm_disk(md)),
+				     blk_rq_pos(tio->orig));
+		dm_dispatch_request(clone);
+		break;
+	case DM_MAPIO_REQUEUE:
+		/* The target wants to requeue the I/O */
+		dm_requeue_unmapped_request(clone);
+		requeued = 1;
+		break;
+	default:
+		if (r > 0) {
+			DMWARN("unimplemented target map return value: %d", r);
+			BUG();
+		}
+
+		/* The target wants to complete the I/O */
+		dm_kill_unmapped_request(clone, r);
+		break;
+	}
+
+	return requeued;
+}
+
+static struct request *dm_start_request(struct mapped_device *md, struct request *orig)
+{
+	struct request *clone;
+
+	blk_start_request(orig);
+	clone = orig->special;
+	atomic_inc(&md->pending[rq_data_dir(clone)]);
 
 	/*
-	 * There is no use in forwarding any barrier request since we can't
-	 * guarantee it is (or can be) handled by the targets correctly.
+	 * Hold the md reference here for the in-flight I/O.
+	 * We can't rely on the reference count by device opener,
+	 * because the device may be closed during the request completion
+	 * when all bios are completed.
+	 * See the comment in rq_completed() too.
 	 */
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
-	}
+	dm_get(md);
 
-	down_read(&md->io_lock);
+	return clone;
+}
 
-	disk_stat_inc(dm_disk(md), ios[rw]);
-	disk_stat_add(dm_disk(md), sectors[rw], bio_sectors(bio));
+/*
+ * q->request_fn for request-based dm.
+ * Called with the queue lock held.
+ */
+static void dm_request_fn(struct request_queue *q)
+{
+	struct mapped_device *md = q->queuedata;
+	int srcu_idx;
+	struct dm_table *map = dm_get_live_table(md, &srcu_idx);
+	struct dm_target *ti;
+	struct request *rq, *clone;
+	sector_t pos;
 
 	/*
-	 * If we're suspended we have to queue
-	 * this io for later.
+	 * For suspend, check blk_queue_stopped() and increment
+	 * ->pending within a single queue_lock not to increment the
+	 * number of in-flight I/Os after the queue is stopped in
+	 * dm_suspend().
 	 */
-	while (test_bit(DMF_BLOCK_IO, &md->flags)) {
-		up_read(&md->io_lock);
+	while (!blk_queue_stopped(q)) {
+		rq = blk_peek_request(q);
+		if (!rq)
+			goto delay_and_out;
+
+		/* always use block 0 to find the target for flushes for now */
+		pos = 0;
+		if (!(rq->cmd_flags & REQ_FLUSH))
+			pos = blk_rq_pos(rq);
+
+		ti = dm_table_find_target(map, pos);
+		if (!dm_target_is_valid(ti)) {
+			/*
+			 * Must perform setup, that dm_done() requires,
+			 * before calling dm_kill_unmapped_request
+			 */
+			DMERR_LIMIT("request attempted access beyond the end of device");
+			clone = dm_start_request(md, rq);
+			dm_kill_unmapped_request(clone, -EIO);
+			continue;
+		}
 
-		if (bio_rw(bio) != READA)
-			r = queue_io(md, bio);
+		if (ti->type->busy && ti->type->busy(ti))
+			goto delay_and_out;
 
-		if (r <= 0)
-			goto out_req;
+		clone = dm_start_request(md, rq);
 
-		/*
-		 * We're in a while loop, because someone could suspend
-		 * before we get to the following read lock.
-		 */
-		down_read(&md->io_lock);
+		spin_unlock(q->queue_lock);
+		if (map_request(ti, clone, md))
+			goto requeued;
+
+		BUG_ON(!irqs_disabled());
+		spin_lock(q->queue_lock);
 	}
 
-	r = __split_bio(md, bio);
-	up_read(&md->io_lock);
+	goto out;
 
-out_req:
-	if (r < 0)
-		bio_io_error(bio);
+requeued:
+	BUG_ON(!irqs_disabled());
+	spin_lock(q->queue_lock);
 
-	return 0;
+delay_and_out:
+	blk_delay_queue(q, HZ / 10);
+out:
+	dm_put_live_table(md, srcu_idx);
+}
+
+int dm_underlying_device_busy(struct request_queue *q)
+{
+	return blk_lld_busy(q);
 }
+EXPORT_SYMBOL_GPL(dm_underlying_device_busy);
 
-static void dm_unplug_all(struct request_queue *q)
+static int dm_lld_busy(struct request_queue *q)
 {
+	int r;
 	struct mapped_device *md = q->queuedata;
-	struct dm_table *map = dm_get_table(md);
+	struct dm_table *map = dm_get_live_table_fast(md);
 
-	if (map) {
-		dm_table_unplug_all(map);
-		dm_table_put(map);
-	}
+	if (!map || test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags))
+		r = 1;
+	else
+		r = dm_table_any_busy_target(map);
+
+	dm_put_live_table_fast(md);
+
+	return r;
 }
 
 static int dm_any_congested(void *congested_data, int bdi_bits)
 {
-	int r;
-	struct mapped_device *md = (struct mapped_device *) congested_data;
-	struct dm_table *map = dm_get_table(md);
+	int r = bdi_bits;
+	struct mapped_device *md = congested_data;
+	struct dm_table *map;
 
-	if (!map || test_bit(DMF_BLOCK_IO, &md->flags))
-		r = bdi_bits;
-	else
-		r = dm_table_any_congested(map, bdi_bits);
+	if (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
+		map = dm_get_live_table_fast(md);
+		if (map) {
+			/*
+			 * Request-based dm cares about only own queue for
+			 * the query about congestion status of request_queue
+			 */
+			if (dm_request_based(md))
+				r = md->queue->backing_dev_info.state &
+				    bdi_bits;
+			else
+				r = dm_table_any_congested(map, bdi_bits);
+		}
+		dm_put_live_table_fast(md);
+	}
 
-	dm_table_put(map);
 	return r;
 }
 
 /*-----------------------------------------------------------------
  * An IDR is used to keep track of allocated minor numbers.
  *---------------------------------------------------------------*/
-static DEFINE_IDR(_minor_idr);
-
 static void free_minor(int minor)
 {
 	spin_lock(&_minor_lock);
@@ -969,65 +1858,64 @@ static void free_minor(int minor)
  */
 static int specific_minor(int minor)
 {
-	int r, m;
+	int r;
 
 	if (minor >= (1 << MINORBITS))
 		return -EINVAL;
 
-	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
-	if (!r)
-		return -ENOMEM;
-
+	idr_preload(GFP_KERNEL);
 	spin_lock(&_minor_lock);
 
-	if (idr_find(&_minor_idr, minor)) {
-		r = -EBUSY;
-		goto out;
-	}
-
-	r = idr_get_new_above(&_minor_idr, MINOR_ALLOCED, minor, &m);
-	if (r)
-		goto out;
-
-	if (m != minor) {
-		idr_remove(&_minor_idr, m);
-		r = -EBUSY;
-		goto out;
-	}
+	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, minor, minor + 1, GFP_NOWAIT);
 
-out:
 	spin_unlock(&_minor_lock);
-	return r;
+	idr_preload_end();
+	if (r < 0)
+		return r == -ENOSPC ? -EBUSY : r;
+	return 0;
 }
 
 static int next_free_minor(int *minor)
 {
-	int r, m;
-
-	r = idr_pre_get(&_minor_idr, GFP_KERNEL);
-	if (!r)
-		return -ENOMEM;
+	int r;
 
+	idr_preload(GFP_KERNEL);
 	spin_lock(&_minor_lock);
 
-	r = idr_get_new(&_minor_idr, MINOR_ALLOCED, &m);
-	if (r)
-		goto out;
+	r = idr_alloc(&_minor_idr, MINOR_ALLOCED, 0, 1 << MINORBITS, GFP_NOWAIT);
 
-	if (m >= (1 << MINORBITS)) {
-		idr_remove(&_minor_idr, m);
-		r = -ENOSPC;
-		goto out;
-	}
-
-	*minor = m;
-
-out:
 	spin_unlock(&_minor_lock);
-	return r;
+	idr_preload_end();
+	if (r < 0)
+		return r;
+	*minor = r;
+	return 0;
 }
 
-static struct block_device_operations dm_blk_dops;
+static const struct block_device_operations dm_blk_dops;
+
+static void dm_wq_work(struct work_struct *work);
+
+static void dm_init_md_queue(struct mapped_device *md)
+{
+	/*
+	 * Request-based dm devices cannot be stacked on top of bio-based dm
+	 * devices.  The type of this dm device has not been decided yet.
+	 * The type is decided at the first table loading time.
+	 * To prevent problematic device stacking, clear the queue flag
+	 * for request stacking support until then.
+	 *
+	 * This queue is new, so no concurrency on the queue_flags.
+	 */
+	queue_flag_clear_unlocked(QUEUE_FLAG_STACKABLE, md->queue);
+
+	md->queue->queuedata = md;
+	md->queue->backing_dev_info.congested_fn = dm_any_congested;
+	md->queue->backing_dev_info.congested_data = md;
+	blk_queue_make_request(md->queue, dm_request);
+	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
+	blk_queue_merge_bvec(md->queue, dm_merge_bvec);
+}
 
 /*
  * Allocate and initialise a blank device with a given minor.
@@ -1054,10 +1942,14 @@ static struct mapped_device *alloc_dev(int minor)
 	if (r < 0)
 		goto bad_minor;
 
-	init_rwsem(&md->io_lock);
+	r = init_srcu_struct(&md->io_barrier);
+	if (r < 0)
+		goto bad_io_barrier;
+
+	md->type = DM_TYPE_NONE;
 	mutex_init(&md->suspend_lock);
-	spin_lock_init(&md->pushback_lock);
-	rwlock_init(&md->map_lock);
+	mutex_init(&md->type_lock);
+	spin_lock_init(&md->deferred_lock);
 	atomic_set(&md->holders, 1);
 	atomic_set(&md->open_count, 0);
 	atomic_set(&md->event_nr, 0);
@@ -1069,33 +1961,18 @@ static struct mapped_device *alloc_dev(int minor)
 	if (!md->queue)
 		goto bad_queue;
 
-	md->queue->queuedata = md;
-	md->queue->backing_dev_info.congested_fn = dm_any_congested;
-	md->queue->backing_dev_info.congested_data = md;
-	blk_queue_make_request(md->queue, dm_request);
-	blk_queue_bounce_limit(md->queue, BLK_BOUNCE_ANY);
-	md->queue->unplug_fn = dm_unplug_all;
-	blk_queue_merge_bvec(md->queue, dm_merge_bvec);
-
-	md->io_pool = mempool_create_slab_pool(MIN_IOS, _io_cache);
-	if (!md->io_pool)
-		goto bad_io_pool;
-
-	md->tio_pool = mempool_create_slab_pool(MIN_IOS, _tio_cache);
-	if (!md->tio_pool)
-		goto bad_tio_pool;
-
-	md->bs = bioset_create(16, 16);
-	if (!md->bs)
-		goto bad_no_bioset;
+	dm_init_md_queue(md);
 
 	md->disk = alloc_disk(1);
 	if (!md->disk)
 		goto bad_disk;
 
-	atomic_set(&md->pending, 0);
+	atomic_set(&md->pending[0], 0);
+	atomic_set(&md->pending[1], 0);
 	init_waitqueue_head(&md->wait);
+	INIT_WORK(&md->work, dm_wq_work);
 	init_waitqueue_head(&md->eventq);
+	init_completion(&md->kobj_holder.completion);
 
 	md->disk->major = _major;
 	md->disk->first_minor = minor;
@@ -1106,10 +1983,20 @@ static struct mapped_device *alloc_dev(int minor)
 	add_disk(md->disk);
 	format_dev_t(md->name, MKDEV(_major, minor));
 
-	md->wq = create_singlethread_workqueue("kdmflush");
+	md->wq = alloc_workqueue("kdmflush", WQ_MEM_RECLAIM, 0);
 	if (!md->wq)
 		goto bad_thread;
 
+	md->bdev = bdget_disk(md->disk, 0);
+	if (!md->bdev)
+		goto bad_bdev;
+
+	bio_init(&md->flush_bio);
+	md->flush_bio.bi_bdev = md->bdev;
+	md->flush_bio.bi_rw = WRITE_FLUSH;
+
+	dm_stats_init(&md->stats);
+
 	/* Populate the mapping, nobody knows we exist yet */
 	spin_lock(&_minor_lock);
 	old_md = idr_replace(&_minor_idr, md, minor);
@@ -1119,17 +2006,16 @@ static struct mapped_device *alloc_dev(int minor)
 
 	return md;
 
+bad_bdev:
+	destroy_workqueue(md->wq);
 bad_thread:
+	del_gendisk(md->disk);
 	put_disk(md->disk);
 bad_disk:
-	bioset_free(md->bs);
-bad_no_bioset:
-	mempool_destroy(md->tio_pool);
-bad_tio_pool:
-	mempool_destroy(md->io_pool);
-bad_io_pool:
 	blk_cleanup_queue(md->queue);
 bad_queue:
+	cleanup_srcu_struct(&md->io_barrier);
+bad_io_barrier:
 	free_minor(minor);
 bad_minor:
 	module_put(THIS_MODULE);
@@ -1142,17 +2028,18 @@ static void unlock_fs(struct mapped_device *md);
 
 static void free_dev(struct mapped_device *md)
 {
-	int minor = md->disk->first_minor;
+	int minor = MINOR(disk_devt(md->disk));
 
-	if (md->suspended_bdev) {
-		unlock_fs(md);
-		bdput(md->suspended_bdev);
-	}
+	unlock_fs(md);
+	bdput(md->bdev);
 	destroy_workqueue(md->wq);
-	mempool_destroy(md->tio_pool);
-	mempool_destroy(md->io_pool);
-	bioset_free(md->bs);
+	if (md->io_pool)
+		mempool_destroy(md->io_pool);
+	if (md->bs)
+		bioset_free(md->bs);
+	blk_integrity_unregister(md->disk);
 	del_gendisk(md->disk);
+	cleanup_srcu_struct(&md->io_barrier);
 	free_minor(minor);
 
 	spin_lock(&_minor_lock);
@@ -1161,10 +2048,50 @@ static void free_dev(struct mapped_device *md)
 
 	put_disk(md->disk);
 	blk_cleanup_queue(md->queue);
+	dm_stats_cleanup(&md->stats);
 	module_put(THIS_MODULE);
 	kfree(md);
 }
 
+static void __bind_mempools(struct mapped_device *md, struct dm_table *t)
+{
+	struct dm_md_mempools *p = dm_table_get_md_mempools(t);
+
+	if (md->io_pool && md->bs) {
+		/* The md already has necessary mempools. */
+		if (dm_table_get_type(t) == DM_TYPE_BIO_BASED) {
+			/*
+			 * Reload bioset because front_pad may have changed
+			 * because a different table was loaded.
+			 */
+			bioset_free(md->bs);
+			md->bs = p->bs;
+			p->bs = NULL;
+		} else if (dm_table_get_type(t) == DM_TYPE_REQUEST_BASED) {
+			/*
+			 * There's no need to reload with request-based dm
+			 * because the size of front_pad doesn't change.
+			 * Note for future: If you are to reload bioset,
+			 * prep-ed requests in the queue may refer
+			 * to bio from the old bioset, so you must walk
+			 * through the queue to unprep.
+			 */
+		}
+		goto out;
+	}
+
+	BUG_ON(!p || md->io_pool || md->bs);
+
+	md->io_pool = p->io_pool;
+	p->io_pool = NULL;
+	md->bs = p->bs;
+	p->bs = NULL;
+
+out:
+	/* mempool bind completed, now no need any mempools in the table */
+	dm_table_free_md_mempools(t);
+}
+
 /*
  * Bind a table to the device.
  */
@@ -1178,62 +2105,141 @@ static void event_callback(void *context)
 	list_splice_init(&md->uevent_list, &uevents);
 	spin_unlock_irqrestore(&md->uevent_lock, flags);
 
-	dm_send_uevents(&uevents, &md->disk->dev.kobj);
+	dm_send_uevents(&uevents, &disk_to_dev(md->disk)->kobj);
 
 	atomic_inc(&md->event_nr);
 	wake_up(&md->eventq);
 }
 
+/*
+ * Protected by md->suspend_lock obtained by dm_swap_table().
+ */
 static void __set_size(struct mapped_device *md, sector_t size)
 {
 	set_capacity(md->disk, size);
 
-	mutex_lock(&md->suspended_bdev->bd_inode->i_mutex);
-	i_size_write(md->suspended_bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
-	mutex_unlock(&md->suspended_bdev->bd_inode->i_mutex);
+	i_size_write(md->bdev->bd_inode, (loff_t)size << SECTOR_SHIFT);
+}
+
+/*
+ * Return 1 if the queue has a compulsory merge_bvec_fn function.
+ *
+ * If this function returns 0, then the device is either a non-dm
+ * device without a merge_bvec_fn, or it is a dm device that is
+ * able to split any bios it receives that are too big.
+ */
+int dm_queue_merge_is_compulsory(struct request_queue *q)
+{
+	struct mapped_device *dev_md;
+
+	if (!q->merge_bvec_fn)
+		return 0;
+
+	if (q->make_request_fn == dm_request) {
+		dev_md = q->queuedata;
+		if (test_bit(DMF_MERGE_IS_OPTIONAL, &dev_md->flags))
+			return 0;
+	}
+
+	return 1;
+}
+
+static int dm_device_merge_is_compulsory(struct dm_target *ti,
+					 struct dm_dev *dev, sector_t start,
+					 sector_t len, void *data)
+{
+	struct block_device *bdev = dev->bdev;
+	struct request_queue *q = bdev_get_queue(bdev);
+
+	return dm_queue_merge_is_compulsory(q);
 }
 
-static int __bind(struct mapped_device *md, struct dm_table *t)
+/*
+ * Return 1 if it is acceptable to ignore merge_bvec_fn based
+ * on the properties of the underlying devices.
+ */
+static int dm_table_merge_is_optional(struct dm_table *table)
 {
+	unsigned i = 0;
+	struct dm_target *ti;
+
+	while (i < dm_table_get_num_targets(table)) {
+		ti = dm_table_get_target(table, i++);
+
+		if (ti->type->iterate_devices &&
+		    ti->type->iterate_devices(ti, dm_device_merge_is_compulsory, NULL))
+			return 0;
+	}
+
+	return 1;
+}
+
+/*
+ * Returns old map, which caller must destroy.
+ */
+static struct dm_table *__bind(struct mapped_device *md, struct dm_table *t,
+			       struct queue_limits *limits)
+{
+	struct dm_table *old_map;
 	struct request_queue *q = md->queue;
 	sector_t size;
+	int merge_is_optional;
 
 	size = dm_table_get_size(t);
 
 	/*
 	 * Wipe any geometry if the size of the table changed.
 	 */
-	if (size != get_capacity(md->disk))
+	if (size != dm_get_size(md))
 		memset(&md->geometry, 0, sizeof(md->geometry));
 
-	if (md->suspended_bdev)
-		__set_size(md, size);
-	if (size == 0)
-		return 0;
+	__set_size(md, size);
 
-	dm_table_get(t);
 	dm_table_event_callback(t, event_callback, md);
 
-	write_lock(&md->map_lock);
-	md->map = t;
-	dm_table_set_restrictions(t, q);
-	write_unlock(&md->map_lock);
+	/*
+	 * The queue hasn't been stopped yet, if the old table type wasn't
+	 * for request-based during suspension.  So stop it to prevent
+	 * I/O mapping before resume.
+	 * This must be done before setting the queue restrictions,
+	 * because request-based dm may be run just after the setting.
+	 */
+	if (dm_table_request_based(t) && !blk_queue_stopped(q))
+		stop_queue(q);
 
-	return 0;
+	__bind_mempools(md, t);
+
+	merge_is_optional = dm_table_merge_is_optional(t);
+
+	old_map = md->map;
+	rcu_assign_pointer(md->map, t);
+	md->immutable_target_type = dm_table_get_immutable_target_type(t);
+
+	dm_table_set_restrictions(t, q, limits);
+	if (merge_is_optional)
+		set_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
+	else
+		clear_bit(DMF_MERGE_IS_OPTIONAL, &md->flags);
+	dm_sync_table(md);
+
+	return old_map;
 }
 
-static void __unbind(struct mapped_device *md)
+/*
+ * Returns unbound table for the caller to free.
+ */
+static struct dm_table *__unbind(struct mapped_device *md)
 {
 	struct dm_table *map = md->map;
 
 	if (!map)
-		return;
+		return NULL;
 
 	dm_table_event_callback(map, NULL, NULL);
-	write_lock(&md->map_lock);
-	md->map = NULL;
-	write_unlock(&md->map_lock);
-	dm_table_put(map);
+	RCU_INIT_POINTER(md->map, NULL);
+	dm_sync_table(md);
+
+	return map;
 }
 
 /*
@@ -1247,10 +2253,94 @@ int dm_create(int minor, struct mapped_device **result)
 	if (!md)
 		return -ENXIO;
 
+	dm_sysfs_init(md);
+
 	*result = md;
 	return 0;
 }
 
+/*
+ * Functions to manage md->type.
+ * All are required to hold md->type_lock.
+ */
+void dm_lock_md_type(struct mapped_device *md)
+{
+	mutex_lock(&md->type_lock);
+}
+
+void dm_unlock_md_type(struct mapped_device *md)
+{
+	mutex_unlock(&md->type_lock);
+}
+
+void dm_set_md_type(struct mapped_device *md, unsigned type)
+{
+	BUG_ON(!mutex_is_locked(&md->type_lock));
+	md->type = type;
+}
+
+unsigned dm_get_md_type(struct mapped_device *md)
+{
+	BUG_ON(!mutex_is_locked(&md->type_lock));
+	return md->type;
+}
+
+struct target_type *dm_get_immutable_target_type(struct mapped_device *md)
+{
+	return md->immutable_target_type;
+}
+
+/*
+ * The queue_limits are only valid as long as you have a reference
+ * count on 'md'.
+ */
+struct queue_limits *dm_get_queue_limits(struct mapped_device *md)
+{
+	BUG_ON(!atomic_read(&md->holders));
+	return &md->queue->limits;
+}
+EXPORT_SYMBOL_GPL(dm_get_queue_limits);
+
+/*
+ * Fully initialize a request-based queue (->elevator, ->request_fn, etc).
+ */
+static int dm_init_request_based_queue(struct mapped_device *md)
+{
+	struct request_queue *q = NULL;
+
+	if (md->queue->elevator)
+		return 1;
+
+	/* Fully initialize the queue */
+	q = blk_init_allocated_queue(md->queue, dm_request_fn, NULL);
+	if (!q)
+		return 0;
+
+	md->queue = q;
+	dm_init_md_queue(md);
+	blk_queue_softirq_done(md->queue, dm_softirq_done);
+	blk_queue_prep_rq(md->queue, dm_prep_fn);
+	blk_queue_lld_busy(md->queue, dm_lld_busy);
+
+	elv_register_queue(md->queue);
+
+	return 1;
+}
+
+/*
+ * Setup the DM device's queue based on md's type
+ */
+int dm_setup_md_queue(struct mapped_device *md)
+{
+	if ((dm_get_md_type(md) == DM_TYPE_REQUEST_BASED) &&
+	    !dm_init_request_based_queue(md)) {
+		DMWARN("Cannot initialize queue for request-based mapped device");
+		return -EINVAL;
+	}
+
+	return 0;
+}
+
 static struct mapped_device *dm_find_md(dev_t dev)
 {
 	struct mapped_device *md;
@@ -1263,7 +2353,8 @@ static struct mapped_device *dm_find_md(dev_t dev)
 
 	md = idr_find(&_minor_idr, minor);
 	if (md && (md == MINOR_ALLOCED ||
-		   (dm_disk(md)->first_minor != minor) ||
+		   (MINOR(disk_devt(dm_disk(md))) != minor) ||
+		   dm_deleting_md(md) ||
 		   test_bit(DMF_FREEING, &md->flags))) {
 		md = NULL;
 		goto out;
@@ -1284,6 +2375,7 @@ struct mapped_device *dm_get_md(dev_t dev)
 
 	return md;
 }
+EXPORT_SYMBOL_GPL(dm_get_md);
 
 void *dm_get_mdptr(struct mapped_device *md)
 {
@@ -1298,6 +2390,7 @@ void dm_set_mdptr(struct mapped_device *md, void *ptr)
 void dm_get(struct mapped_device *md)
 {
 	atomic_inc(&md->holders);
+	BUG_ON(test_bit(DMF_FREEING, &md->flags));
 }
 
 const char *dm_device_name(struct mapped_device *md)
@@ -1306,40 +2399,76 @@ const char *dm_device_name(struct mapped_device *md)
 }
 EXPORT_SYMBOL_GPL(dm_device_name);
 
-void dm_put(struct mapped_device *md)
+static void __dm_destroy(struct mapped_device *md, bool wait)
 {
 	struct dm_table *map;
+	int srcu_idx;
 
-	BUG_ON(test_bit(DMF_FREEING, &md->flags));
+	might_sleep();
 
-	if (atomic_dec_and_lock(&md->holders, &_minor_lock)) {
-		map = dm_get_table(md);
-		idr_replace(&_minor_idr, MINOR_ALLOCED, dm_disk(md)->first_minor);
-		set_bit(DMF_FREEING, &md->flags);
-		spin_unlock(&_minor_lock);
-		if (!dm_suspended(md)) {
-			dm_table_presuspend_targets(map);
-			dm_table_postsuspend_targets(map);
-		}
-		__unbind(md);
-		dm_table_put(map);
-		free_dev(md);
+	spin_lock(&_minor_lock);
+	map = dm_get_live_table(md, &srcu_idx);
+	idr_replace(&_minor_idr, MINOR_ALLOCED, MINOR(disk_devt(dm_disk(md))));
+	set_bit(DMF_FREEING, &md->flags);
+	spin_unlock(&_minor_lock);
+
+	if (!dm_suspended_md(md)) {
+		dm_table_presuspend_targets(map);
+		dm_table_postsuspend_targets(map);
 	}
+
+	/* dm_put_live_table must be before msleep, otherwise deadlock is possible */
+	dm_put_live_table(md, srcu_idx);
+
+	/*
+	 * Rare, but there may be I/O requests still going to complete,
+	 * for example.  Wait for all references to disappear.
+	 * No one should increment the reference count of the mapped_device,
+	 * after the mapped_device state becomes DMF_FREEING.
+	 */
+	if (wait)
+		while (atomic_read(&md->holders))
+			msleep(1);
+	else if (atomic_read(&md->holders))
+		DMWARN("%s: Forcibly removing mapped_device still in use! (%d users)",
+		       dm_device_name(md), atomic_read(&md->holders));
+
+	dm_sysfs_exit(md);
+	dm_table_destroy(__unbind(md));
+	free_dev(md);
+}
+
+void dm_destroy(struct mapped_device *md)
+{
+	__dm_destroy(md, true);
+}
+
+void dm_destroy_immediate(struct mapped_device *md)
+{
+	__dm_destroy(md, false);
+}
+
+void dm_put(struct mapped_device *md)
+{
+	atomic_dec(&md->holders);
 }
 EXPORT_SYMBOL_GPL(dm_put);
 
-static int dm_wait_for_completion(struct mapped_device *md)
+static int dm_wait_for_completion(struct mapped_device *md, int interruptible)
 {
 	int r = 0;
+	DECLARE_WAITQUEUE(wait, current);
+
+	add_wait_queue(&md->wait, &wait);
 
 	while (1) {
-		set_current_state(TASK_INTERRUPTIBLE);
+		set_current_state(interruptible);
 
-		smp_mb();
-		if (!atomic_read(&md->pending))
+		if (!md_in_flight(md))
 			break;
 
-		if (signal_pending(current)) {
+		if (interruptible == TASK_INTERRUPTIBLE &&
+		    signal_pending(current)) {
 			r = -EINTR;
 			break;
 		}
@@ -1348,97 +2477,89 @@ static int dm_wait_for_completion(struct mapped_device *md)
 	}
 	set_current_state(TASK_RUNNING);
 
+	remove_wait_queue(&md->wait, &wait);
+
 	return r;
 }
 
 /*
  * Process the deferred bios
  */
-static void __flush_deferred_io(struct mapped_device *md)
+static void dm_wq_work(struct work_struct *work)
 {
+	struct mapped_device *md = container_of(work, struct mapped_device,
+						work);
 	struct bio *c;
+	int srcu_idx;
+	struct dm_table *map;
 
-	while ((c = bio_list_pop(&md->deferred))) {
-		if (__split_bio(md, c))
-			bio_io_error(c);
-	}
-
-	clear_bit(DMF_BLOCK_IO, &md->flags);
-}
+	map = dm_get_live_table(md, &srcu_idx);
 
-static void __merge_pushback_list(struct mapped_device *md)
-{
-	unsigned long flags;
-
-	spin_lock_irqsave(&md->pushback_lock, flags);
-	clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
-	bio_list_merge_head(&md->deferred, &md->pushback);
-	bio_list_init(&md->pushback);
-	spin_unlock_irqrestore(&md->pushback_lock, flags);
-}
+	while (!test_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags)) {
+		spin_lock_irq(&md->deferred_lock);
+		c = bio_list_pop(&md->deferred);
+		spin_unlock_irq(&md->deferred_lock);
 
-static void dm_wq_work(struct work_struct *work)
-{
-	struct dm_wq_req *req = container_of(work, struct dm_wq_req, work);
-	struct mapped_device *md = req->md;
+		if (!c)
+			break;
 
-	down_write(&md->io_lock);
-	switch (req->type) {
-	case DM_WQ_FLUSH_ALL:
-		__merge_pushback_list(md);
-		/* pass through */
-	case DM_WQ_FLUSH_DEFERRED:
-		__flush_deferred_io(md);
-		break;
-	default:
-		DMERR("dm_wq_work: unrecognised work type %d", req->type);
-		BUG();
+		if (dm_request_based(md))
+			generic_make_request(c);
+		else
+			__split_and_process_bio(md, map, c);
 	}
-	up_write(&md->io_lock);
-}
 
-static void dm_wq_queue(struct mapped_device *md, int type, void *context,
-			struct dm_wq_req *req)
-{
-	req->type = type;
-	req->md = md;
-	req->context = context;
-	INIT_WORK(&req->work, dm_wq_work);
-	queue_work(md->wq, &req->work);
+	dm_put_live_table(md, srcu_idx);
 }
 
-static void dm_queue_flush(struct mapped_device *md, int type, void *context)
+static void dm_queue_flush(struct mapped_device *md)
 {
-	struct dm_wq_req req;
-
-	dm_wq_queue(md, type, context, &req);
-	flush_workqueue(md->wq);
+	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
+	smp_mb__after_atomic();
+	queue_work(md->wq, &md->work);
 }
 
 /*
- * Swap in a new table (destroying old one).
+ * Swap in a new table, returning the old one for the caller to destroy.
  */
-int dm_swap_table(struct mapped_device *md, struct dm_table *table)
+struct dm_table *dm_swap_table(struct mapped_device *md, struct dm_table *table)
 {
-	int r = -EINVAL;
+	struct dm_table *live_map = NULL, *map = ERR_PTR(-EINVAL);
+	struct queue_limits limits;
+	int r;
 
 	mutex_lock(&md->suspend_lock);
 
 	/* device must be suspended */
-	if (!dm_suspended(md))
+	if (!dm_suspended_md(md))
 		goto out;
 
-	/* without bdev, the device size cannot be changed */
-	if (!md->suspended_bdev)
-		if (get_capacity(md->disk) != dm_table_get_size(table))
+	/*
+	 * If the new table has no data devices, retain the existing limits.
+	 * This helps multipath with queue_if_no_path if all paths disappear,
+	 * then new I/O is queued based on these limits, and then some paths
+	 * reappear.
+	 */
+	if (dm_table_has_no_data_devices(table)) {
+		live_map = dm_get_live_table_fast(md);
+		if (live_map)
+			limits = md->queue->limits;
+		dm_put_live_table_fast(md);
+	}
+
+	if (!live_map) {
+		r = dm_calculate_queue_limits(table, &limits);
+		if (r) {
+			map = ERR_PTR(r);
 			goto out;
+		}
+	}
 
-	__unbind(md);
-	r = __bind(md, table);
+	map = __bind(md, table, &limits);
 
 out:
 	mutex_unlock(&md->suspend_lock);
-	return r;
+	return map;
 }
 
 /*
@@ -1451,7 +2572,7 @@ static int lock_fs(struct mapped_device *md)
 
 	WARN_ON(md->frozen_sb);
 
-	md->frozen_sb = freeze_bdev(md->suspended_bdev);
+	md->frozen_sb = freeze_bdev(md->bdev);
 	if (IS_ERR(md->frozen_sb)) {
 		r = PTR_ERR(md->frozen_sb);
 		md->frozen_sb = NULL;
@@ -1460,9 +2581,6 @@ static int lock_fs(struct mapped_device *md)
 
 	set_bit(DMF_FROZEN, &md->flags);
 
-	/* don't bdput right now, we don't want the bdev
-	 * to go away while it is locked.
-	 */
 	return 0;
 }
 
@@ -1471,7 +2589,7 @@ static void unlock_fs(struct mapped_device *md)
 	if (!test_bit(DMF_FROZEN, &md->flags))
 		return;
 
-	thaw_bdev(md->suspended_bdev, md->frozen_sb);
+	thaw_bdev(md->bdev, md->frozen_sb);
 	md->frozen_sb = NULL;
 	clear_bit(DMF_FROZEN, &md->flags);
 }
@@ -1483,22 +2601,30 @@ static void unlock_fs(struct mapped_device *md)
  * dm_bind_table, dm_suspend must be called to flush any in
  * flight bios and ensure that any further io gets deferred.
  */
+/*
+ * Suspend mechanism in request-based dm.
+ *
+ * 1. Flush all I/Os by lock_fs() if needed.
+ * 2. Stop dispatching any I/O by stopping the request_queue.
+ * 3. Wait for all in-flight I/Os to be completed or requeued.
+ *
+ * To abort suspend, start the request_queue.
+ */
 int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 {
 	struct dm_table *map = NULL;
-	DECLARE_WAITQUEUE(wait, current);
 	int r = 0;
 	int do_lockfs = suspend_flags & DM_SUSPEND_LOCKFS_FLAG ? 1 : 0;
 	int noflush = suspend_flags & DM_SUSPEND_NOFLUSH_FLAG ? 1 : 0;
 
 	mutex_lock(&md->suspend_lock);
 
-	if (dm_suspended(md)) {
+	if (dm_suspended_md(md)) {
 		r = -EINVAL;
 		goto out_unlock;
 	}
 
-	map = dm_get_table(md);
+	map = md->map;
 
 	/*
 	 * DMF_NOFLUSH_SUSPENDING must be set before presuspend.
@@ -1510,78 +2636,73 @@ int dm_suspend(struct mapped_device *md, unsigned suspend_flags)
 	/* This does not get reverted if there's an error later. */
 	dm_table_presuspend_targets(map);
 
-	/* bdget() can stall if the pending I/Os are not flushed */
-	if (!noflush) {
-		md->suspended_bdev = bdget_disk(md->disk, 0);
-		if (!md->suspended_bdev) {
-			DMWARN("bdget failed in dm_suspend");
-			r = -ENOMEM;
-			goto flush_and_out;
-		}
-
-		/*
-		 * Flush I/O to the device. noflush supersedes do_lockfs,
-		 * because lock_fs() needs to flush I/Os.
-		 */
-		if (do_lockfs) {
-			r = lock_fs(md);
-			if (r)
-				goto out;
-		}
+	/*
+	 * Flush I/O to the device.
+	 * Any I/O submitted after lock_fs() may not be flushed.
+	 * noflush takes precedence over do_lockfs.
+	 * (lock_fs() flushes I/Os and waits for them to complete.)
+	 */
+	if (!noflush && do_lockfs) {
+		r = lock_fs(md);
+		if (r)
+			goto out_unlock;
 	}
 
 	/*
-	 * First we set the BLOCK_IO flag so no more ios will be mapped.
+	 * Here we must make sure that no processes are submitting requests
+	 * to target drivers i.e. no one may be executing
+	 * __split_and_process_bio. This is called from dm_request and
+	 * dm_wq_work.
+	 *
+	 * To get all processes out of __split_and_process_bio in dm_request,
+	 * we take the write lock. To prevent any process from reentering
+	 * __split_and_process_bio from dm_request and quiesce the thread
+	 * (dm_wq_work), we set BMF_BLOCK_IO_FOR_SUSPEND and call
+	 * flush_workqueue(md->wq).
 	 */
-	down_write(&md->io_lock);
-	set_bit(DMF_BLOCK_IO, &md->flags);
+	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
+	synchronize_srcu(&md->io_barrier);
 
-	add_wait_queue(&md->wait, &wait);
-	up_write(&md->io_lock);
+	/*
+	 * Stop md->queue before flushing md->wq in case request-based
+	 * dm defers requests to md->wq from md->queue.
+	 */
+	if (dm_request_based(md))
+		stop_queue(md->queue);
 
-	/* unplug */
-	if (map)
-		dm_table_unplug_all(map);
+	flush_workqueue(md->wq);
 
 	/*
-	 * Wait for the already-mapped ios to complete.
+	 * At this point no more requests are entering target request routines.
+	 * We call dm_wait_for_completion to wait for all existing requests
+	 * to finish.
 	 */
-	r = dm_wait_for_completion(md);
-
-	down_write(&md->io_lock);
-	remove_wait_queue(&md->wait, &wait);
+	r = dm_wait_for_completion(md, TASK_INTERRUPTIBLE);
 
 	if (noflush)
-		__merge_pushback_list(md);
-	up_write(&md->io_lock);
+		clear_bit(DMF_NOFLUSH_SUSPENDING, &md->flags);
+	synchronize_srcu(&md->io_barrier);
 
 	/* were we interrupted ? */
 	if (r < 0) {
-		dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL);
+		dm_queue_flush(md);
+
+		if (dm_request_based(md))
+			start_queue(md->queue);
 
 		unlock_fs(md);
-		goto out; /* pushback list is already flushed, so skip flush */
+		goto out_unlock; /* pushback list is already flushed, so skip flush */
 	}
 
-	dm_table_postsuspend_targets(map);
+	/*
+	 * If dm_wait_for_completion returned 0, the device is completely
+	 * quiescent now. There is no request-processing activity. All new
+	 * requests are being added to md->deferred list.
+	 */
 
 	set_bit(DMF_SUSPENDED, &md->flags);
 
-flush_and_out:
-	if (r && noflush)
-		/*
-		 * Because there may be already I/Os in the pushback list,
-		 * flush them before return.
-		 */
-		dm_queue_flush(md, DM_WQ_FLUSH_ALL, NULL);
-
-out:
-	if (r && md->suspended_bdev) {
-		bdput(md->suspended_bdev);
-		md->suspended_bdev = NULL;
-	}
-
-	dm_table_put(map);
+	dm_table_postsuspend_targets(map);
 
 out_unlock:
 	mutex_unlock(&md->suspend_lock);
@@ -1594,10 +2715,10 @@ int dm_resume(struct mapped_device *md)
 	struct dm_table *map = NULL;
 
 	mutex_lock(&md->suspend_lock);
-	if (!dm_suspended(md))
+	if (!dm_suspended_md(md))
 		goto out;
 
-	map = dm_get_table(md);
+	map = md->map;
 	if (!map || !dm_table_get_size(map))
 		goto out;
 
@@ -1605,36 +2726,76 @@ int dm_resume(struct mapped_device *md)
 	if (r)
 		goto out;
 
-	dm_queue_flush(md, DM_WQ_FLUSH_DEFERRED, NULL);
+	dm_queue_flush(md);
 
-	unlock_fs(md);
+	/*
+	 * Flushing deferred I/Os must be done after targets are resumed
+	 * so that mapping of targets can work correctly.
+	 * Request-based dm is queueing the deferred I/Os in its request_queue.
+	 */
+	if (dm_request_based(md))
+		start_queue(md->queue);
 
-	if (md->suspended_bdev) {
-		bdput(md->suspended_bdev);
-		md->suspended_bdev = NULL;
-	}
+	unlock_fs(md);
 
 	clear_bit(DMF_SUSPENDED, &md->flags);
 
-	dm_table_unplug_all(map);
-
-	dm_kobject_uevent(md);
-
 	r = 0;
-
 out:
-	dm_table_put(map);
 	mutex_unlock(&md->suspend_lock);
 
 	return r;
 }
 
+/*
+ * Internal suspend/resume works like userspace-driven suspend. It waits
+ * until all bios finish and prevents issuing new bios to the target drivers.
+ * It may be used only from the kernel.
+ *
+ * Internal suspend holds md->suspend_lock, which prevents interaction with
+ * userspace-driven suspend.
+ */
+
+void dm_internal_suspend(struct mapped_device *md)
+{
+	mutex_lock(&md->suspend_lock);
+	if (dm_suspended_md(md))
+		return;
+
+	set_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
+	synchronize_srcu(&md->io_barrier);
+	flush_workqueue(md->wq);
+	dm_wait_for_completion(md, TASK_UNINTERRUPTIBLE);
+}
+
+void dm_internal_resume(struct mapped_device *md)
+{
+	if (dm_suspended_md(md))
+		goto done;
+
+	dm_queue_flush(md);
+
+done:
+	mutex_unlock(&md->suspend_lock);
+}
+
 /*-----------------------------------------------------------------
  * Event notification.
  *---------------------------------------------------------------*/
-void dm_kobject_uevent(struct mapped_device *md)
-{
-	kobject_uevent(&md->disk->dev.kobj, KOBJ_CHANGE);
+int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
+		       unsigned cookie)
+{
+	char udev_cookie[DM_COOKIE_LENGTH];
+	char *envp[] = { udev_cookie, NULL };
+
+	if (!cookie)
+		return kobject_uevent(&disk_to_dev(md->disk)->kobj, action);
+	else {
+		snprintf(udev_cookie, DM_COOKIE_LENGTH, "%s=%u",
+			 DM_COOKIE_ENV_VAR_NAME, cookie);
+		return kobject_uevent_env(&disk_to_dev(md->disk)->kobj,
+					  action, envp);
+	}
 }
 
 uint32_t dm_next_uevent_seq(struct mapped_device *md)
@@ -1671,23 +2832,104 @@ struct gendisk *dm_disk(struct mapped_device *md)
 	return md->disk;
 }
 
-int dm_suspended(struct mapped_device *md)
+struct kobject *dm_kobject(struct mapped_device *md)
+{
+	return &md->kobj_holder.kobj;
+}
+
+struct mapped_device *dm_get_from_kobject(struct kobject *kobj)
+{
+	struct mapped_device *md;
+
+	md = container_of(kobj, struct mapped_device, kobj_holder.kobj);
+
+	if (test_bit(DMF_FREEING, &md->flags) ||
+	    dm_deleting_md(md))
+		return NULL;
+
+	dm_get(md);
+	return md;
+}
+
+int dm_suspended_md(struct mapped_device *md)
 {
 	return test_bit(DMF_SUSPENDED, &md->flags);
 }
 
-int dm_noflush_suspending(struct dm_target *ti)
+int dm_test_deferred_remove_flag(struct mapped_device *md)
 {
-	struct mapped_device *md = dm_table_get_md(ti->table);
-	int r = __noflush_suspending(md);
+	return test_bit(DMF_DEFERRED_REMOVE, &md->flags);
+}
 
-	dm_put(md);
+int dm_suspended(struct dm_target *ti)
+{
+	return dm_suspended_md(dm_table_get_md(ti->table));
+}
+EXPORT_SYMBOL_GPL(dm_suspended);
 
-	return r;
+int dm_noflush_suspending(struct dm_target *ti)
+{
+	return __noflush_suspending(dm_table_get_md(ti->table));
 }
 EXPORT_SYMBOL_GPL(dm_noflush_suspending);
 
-static struct block_device_operations dm_blk_dops = {
+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size)
+{
+	struct dm_md_mempools *pools = kzalloc(sizeof(*pools), GFP_KERNEL);
+	struct kmem_cache *cachep;
+	unsigned int pool_size;
+	unsigned int front_pad;
+
+	if (!pools)
+		return NULL;
+
+	if (type == DM_TYPE_BIO_BASED) {
+		cachep = _io_cache;
+		pool_size = dm_get_reserved_bio_based_ios();
+		front_pad = roundup(per_bio_data_size, __alignof__(struct dm_target_io)) + offsetof(struct dm_target_io, clone);
+	} else if (type == DM_TYPE_REQUEST_BASED) {
+		cachep = _rq_tio_cache;
+		pool_size = dm_get_reserved_rq_based_ios();
+		front_pad = offsetof(struct dm_rq_clone_bio_info, clone);
+		/* per_bio_data_size is not used. See __bind_mempools(). */
+		WARN_ON(per_bio_data_size != 0);
+	} else
+		goto out;
+
+	pools->io_pool = mempool_create_slab_pool(pool_size, cachep);
+	if (!pools->io_pool)
+		goto out;
+
+	pools->bs = bioset_create(pool_size, front_pad);
+	if (!pools->bs)
+		goto out;
+
+	if (integrity && bioset_integrity_create(pools->bs, pool_size))
+		goto out;
+
+	return pools;
+
+out:
+	dm_free_md_mempools(pools);
+
+	return NULL;
+}
+
+void dm_free_md_mempools(struct dm_md_mempools *pools)
+{
+	if (!pools)
+		return;
+
+	if (pools->io_pool)
+		mempool_destroy(pools->io_pool);
+
+	if (pools->bs)
+		bioset_free(pools->bs);
+
+	kfree(pools);
+}
+
+static const struct block_device_operations dm_blk_dops = {
 	.open = dm_blk_open,
 	.release = dm_blk_close,
 	.ioctl = dm_blk_ioctl,
@@ -1695,8 +2937,6 @@ static struct block_device_operations dm_blk_dops = {
 	.owner = THIS_MODULE
 };
 
-EXPORT_SYMBOL(dm_get_mapinfo);
-
 /*
  * module hooks
  */
@@ -1705,6 +2945,13 @@ module_exit(dm_exit);
 
 module_param(major, uint, 0);
 MODULE_PARM_DESC(major, "The major number of the device mapper");
+
+module_param(reserved_bio_based_ios, uint, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(reserved_bio_based_ios, "Reserved IOs in bio-based mempools");
+
+module_param(reserved_rq_based_ios, uint, S_IRUGO | S_IWUSR);
+MODULE_PARM_DESC(reserved_rq_based_ios, "Reserved IOs in request-based mempools");
+
 MODULE_DESCRIPTION(DM_NAME " driver");
 MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
 MODULE_LICENSE("GPL");
diff --git a/drivers/md/dm.h b/drivers/md/dm.h
index 1e59a0b0a78..ed76126aac5 100644
--- a/drivers/md/dm.h
+++ b/drivers/md/dm.h
@@ -15,6 +15,10 @@
 #include <linux/list.h>
 #include <linux/blkdev.h>
 #include <linux/hdreg.h>
+#include <linux/completion.h>
+#include <linux/kobject.h>
+
+#include "dm-stats.h"
 
 /*
  * Suspend feature flags
@@ -23,61 +27,119 @@
 #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
+#define DM_TYPE_BIO_BASED	1
+#define DM_TYPE_REQUEST_BASED	2
+
+/*
  * List of devices that a metadevice uses and should open/close.
  */
-struct dm_dev {
+struct dm_dev_internal {
 	struct list_head list;
-
 	atomic_t count;
-	int mode;
-	struct block_device *bdev;
-	char name[16];
+	struct dm_dev dm_dev;
 };
 
 struct dm_table;
+struct dm_md_mempools;
 
 /*-----------------------------------------------------------------
  * Internal table functions.
  *---------------------------------------------------------------*/
+void dm_table_destroy(struct dm_table *t);
 void dm_table_event_callback(struct dm_table *t,
 			     void (*fn)(void *), void *context);
 struct dm_target *dm_table_get_target(struct dm_table *t, unsigned int index);
 struct dm_target *dm_table_find_target(struct dm_table *t, sector_t sector);
-void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q);
+bool dm_table_has_no_data_devices(struct dm_table *table);
+int dm_calculate_queue_limits(struct dm_table *table,
+			      struct queue_limits *limits);
+void dm_table_set_restrictions(struct dm_table *t, struct request_queue *q,
+			       struct queue_limits *limits);
 struct list_head *dm_table_get_devices(struct dm_table *t);
 void dm_table_presuspend_targets(struct dm_table *t);
 void dm_table_postsuspend_targets(struct dm_table *t);
 int dm_table_resume_targets(struct dm_table *t);
 int dm_table_any_congested(struct dm_table *t, int bdi_bits);
-void dm_table_unplug_all(struct dm_table *t);
+int dm_table_any_busy_target(struct dm_table *t);
+unsigned dm_table_get_type(struct dm_table *t);
+struct target_type *dm_table_get_immutable_target_type(struct dm_table *t);
+bool dm_table_request_based(struct dm_table *t);
+bool dm_table_supports_discards(struct dm_table *t);
+void dm_table_free_md_mempools(struct dm_table *t);
+struct dm_md_mempools *dm_table_get_md_mempools(struct dm_table *t);
+
+int dm_queue_merge_is_compulsory(struct request_queue *q);
+
+void dm_lock_md_type(struct mapped_device *md);
+void dm_unlock_md_type(struct mapped_device *md);
+void dm_set_md_type(struct mapped_device *md, unsigned type);
+unsigned dm_get_md_type(struct mapped_device *md);
+struct target_type *dm_get_immutable_target_type(struct mapped_device *md);
+
+int dm_setup_md_queue(struct mapped_device *md);
 
 /*
  * To check the return value from dm_table_find_target().
  */
 #define dm_target_is_valid(t) ((t)->table)
 
+/*
+ * To check whether the target type is bio-based or not (request-based).
+ */
+#define dm_target_bio_based(t) ((t)->type->map != NULL)
+
+/*
+ * To check whether the target type is request-based or not (bio-based).
+ */
+#define dm_target_request_based(t) ((t)->type->map_rq != NULL)
+
+/*
+ * To check whether the target type is a hybrid (capable of being
+ * either request-based or bio-based).
+ */
+#define dm_target_hybrid(t) (dm_target_bio_based(t) && dm_target_request_based(t))
+
 /*-----------------------------------------------------------------
  * A registry of target types.
  *---------------------------------------------------------------*/
 int dm_target_init(void);
 void dm_target_exit(void);
 struct target_type *dm_get_target_type(const char *name);
-void dm_put_target_type(struct target_type *t);
+void dm_put_target_type(struct target_type *tt);
 int dm_target_iterate(void (*iter_func)(struct target_type *tt,
 					void *param), void *param);
 
-/*-----------------------------------------------------------------
- * Useful inlines.
- *---------------------------------------------------------------*/
-static inline int array_too_big(unsigned long fixed, unsigned long obj,
-				unsigned long num)
-{
-	return (num > (ULONG_MAX - fixed) / obj);
-}
-
 int dm_split_args(int *argc, char ***argvp, char *input);
 
 /*
+ * Is this mapped_device being deleted?
+ */
+int dm_deleting_md(struct mapped_device *md);
+
+/*
+ * Is this mapped_device suspended?
+ */
+int dm_suspended_md(struct mapped_device *md);
+
+/*
+ * Test if the device is scheduled for deferred remove.
+ */
+int dm_test_deferred_remove_flag(struct mapped_device *md);
+
+/*
+ * Try to remove devices marked for deferred removal.
+ */
+void dm_deferred_remove(void);
+
+/*
  * The device-mapper can be driven through one of two interfaces;
  * ioctl or filesystem, depending which patch you have applied.
  */
@@ -85,6 +147,29 @@ int dm_interface_init(void);
 void dm_interface_exit(void);
 
 /*
+ * sysfs interface
+ */
+struct dm_kobject_holder {
+	struct kobject kobj;
+	struct completion completion;
+};
+
+static inline struct completion *dm_get_completion_from_kobject(struct kobject *kobj)
+{
+	return &container_of(kobj, struct dm_kobject_holder, kobj)->completion;
+}
+
+int dm_sysfs_init(struct mapped_device *md);
+void dm_sysfs_exit(struct mapped_device *md);
+struct kobject *dm_kobject(struct mapped_device *md);
+struct mapped_device *dm_get_from_kobject(struct kobject *kobj);
+
+/*
+ * The kobject helper
+ */
+void dm_kobject_release(struct kobject *kobj);
+
+/*
  * Targets for linear and striped mappings
  */
 int dm_linear_init(void);
@@ -93,14 +178,46 @@ void dm_linear_exit(void);
 int dm_stripe_init(void);
 void dm_stripe_exit(void);
 
-void *dm_vcalloc(unsigned long nmemb, unsigned long elem_size);
-union map_info *dm_get_mapinfo(struct bio *bio);
+/*
+ * mapped_device operations
+ */
+void dm_destroy(struct mapped_device *md);
+void dm_destroy_immediate(struct mapped_device *md);
 int dm_open_count(struct mapped_device *md);
-int dm_lock_for_deletion(struct mapped_device *md);
+int dm_lock_for_deletion(struct mapped_device *md, bool mark_deferred, bool only_deferred);
+int dm_cancel_deferred_remove(struct mapped_device *md);
+int dm_request_based(struct mapped_device *md);
+sector_t dm_get_size(struct mapped_device *md);
+struct request_queue *dm_get_md_queue(struct mapped_device *md);
+struct dm_stats *dm_get_stats(struct mapped_device *md);
 
-void dm_kobject_uevent(struct mapped_device *md);
+int dm_kobject_uevent(struct mapped_device *md, enum kobject_action action,
+		      unsigned cookie);
+
+void dm_internal_suspend(struct mapped_device *md);
+void dm_internal_resume(struct mapped_device *md);
+
+int dm_io_init(void);
+void dm_io_exit(void);
 
 int dm_kcopyd_init(void);
 void dm_kcopyd_exit(void);
 
+/*
+ * Mempool operations
+ */
+struct dm_md_mempools *dm_alloc_md_mempools(unsigned type, unsigned integrity, unsigned per_bio_data_size);
+void dm_free_md_mempools(struct dm_md_mempools *pools);
+
+/*
+ * Helpers that are used by DM core
+ */
+unsigned dm_get_reserved_bio_based_ios(void);
+unsigned dm_get_reserved_rq_based_ios(void);
+
+static inline bool dm_message_test_buffer_overflow(char *result, unsigned maxlen)
+{
+	return !maxlen || strlen(result) + 1 >= maxlen;
+}
+
 #endif
diff --git a/drivers/md/faulty.c b/drivers/md/faulty.c
index 268547dbfbd..e8b4574956c 100644
--- a/drivers/md/faulty.c
+++ b/drivers/md/faulty.c
@@ -30,7 +30,7 @@
  *
  * Different modes can be active at a time, but only
  * one can be set at array creation.  Others can be added later.
- * A mode can be one-shot or recurrent with the recurrance being
+ * A mode can be one-shot or recurrent with the recurrence being
  * once in every N requests.
  * The bottom 5 bits of the "layout" indicate the mode.  The
  * remainder indicate a period, or 0 for one-shot.
@@ -62,31 +62,36 @@
 #define	ModeShift	5
 
 #define MaxFault	50
-#include <linux/raid/md.h>
+#include <linux/blkdev.h>
+#include <linux/module.h>
+#include <linux/raid/md_u.h>
+#include <linux/slab.h>
+#include "md.h"
+#include <linux/seq_file.h>
 
 
 static void faulty_fail(struct bio *bio, int error)
 {
 	struct bio *b = bio->bi_private;
 
-	b->bi_size = bio->bi_size;
-	b->bi_sector = bio->bi_sector;
+	b->bi_iter.bi_size = bio->bi_iter.bi_size;
+	b->bi_iter.bi_sector = bio->bi_iter.bi_sector;
 
 	bio_put(bio);
 
 	bio_io_error(b);
 }
 
-typedef struct faulty_conf {
+struct faulty_conf {
 	int period[Modes];
 	atomic_t counters[Modes];
 	sector_t faults[MaxFault];
 	int	modes[MaxFault];
 	int nfaults;
-	mdk_rdev_t *rdev;
-} conf_t;
+	struct md_rdev *rdev;
+};
 
-static int check_mode(conf_t *conf, int mode)
+static int check_mode(struct faulty_conf *conf, int mode)
 {
 	if (conf->period[mode] == 0 &&
 	    atomic_read(&conf->counters[mode]) <= 0)
@@ -101,7 +106,7 @@ static int check_mode(conf_t *conf, int mode)
 	return 0;
 }
 
-static int check_sector(conf_t *conf, sector_t start, sector_t end, int dir)
+static int check_sector(struct faulty_conf *conf, sector_t start, sector_t end, int dir)
 {
 	/* If we find a ReadFixable sector, we fix it ... */
 	int i;
@@ -125,7 +130,7 @@ static int check_sector(conf_t *conf, sector_t start, sector_t end, int dir)
 	return 0;
 }
 
-static void add_sector(conf_t *conf, sector_t start, int mode)
+static void add_sector(struct faulty_conf *conf, sector_t start, int mode)
 {
 	int i;
 	int n = conf->nfaults;
@@ -165,10 +170,9 @@ static void add_sector(conf_t *conf, sector_t start, int mode)
 		conf->nfaults = n+1;
 }
 
-static int make_request(struct request_queue *q, struct bio *bio)
+static void make_request(struct mddev *mddev, struct bio *bio)
 {
-	mddev_t *mddev = q->queuedata;
-	conf_t *conf = (conf_t*)mddev->private;
+	struct faulty_conf *conf = mddev->private;
 	int failit = 0;
 
 	if (bio_data_dir(bio) == WRITE) {
@@ -178,50 +182,53 @@ static int make_request(struct request_queue *q, struct bio *bio)
 			 * just fail immediately
 			 */
 			bio_endio(bio, -EIO);
-			return 0;
+			return;
 		}
 
-		if (check_sector(conf, bio->bi_sector, bio->bi_sector+(bio->bi_size>>9),
-				 WRITE))
+		if (check_sector(conf, bio->bi_iter.bi_sector,
+				 bio_end_sector(bio), WRITE))
 			failit = 1;
 		if (check_mode(conf, WritePersistent)) {
-			add_sector(conf, bio->bi_sector, WritePersistent);
+			add_sector(conf, bio->bi_iter.bi_sector,
+				   WritePersistent);
 			failit = 1;
 		}
 		if (check_mode(conf, WriteTransient))
 			failit = 1;
 	} else {
 		/* read request */
-		if (check_sector(conf, bio->bi_sector, bio->bi_sector + (bio->bi_size>>9),
-				 READ))
+		if (check_sector(conf, bio->bi_iter.bi_sector,
+				 bio_end_sector(bio), READ))
 			failit = 1;
 		if (check_mode(conf, ReadTransient))
 			failit = 1;
 		if (check_mode(conf, ReadPersistent)) {
-			add_sector(conf, bio->bi_sector, ReadPersistent);
+			add_sector(conf, bio->bi_iter.bi_sector,
+				   ReadPersistent);
 			failit = 1;
 		}
 		if (check_mode(conf, ReadFixable)) {
-			add_sector(conf, bio->bi_sector, ReadFixable);
+			add_sector(conf, bio->bi_iter.bi_sector,
+				   ReadFixable);
 			failit = 1;
 		}
 	}
 	if (failit) {
-		struct bio *b = bio_clone(bio, GFP_NOIO);
+		struct bio *b = bio_clone_mddev(bio, GFP_NOIO, mddev);
+
 		b->bi_bdev = conf->rdev->bdev;
 		b->bi_private = bio;
 		b->bi_end_io = faulty_fail;
-		generic_make_request(b);
-		return 0;
-	} else {
+		bio = b;
+	} else
 		bio->bi_bdev = conf->rdev->bdev;
-		return 1;
-	}
+
+	generic_make_request(bio);
 }
 
-static void status(struct seq_file *seq, mddev_t *mddev)
+static void status(struct seq_file *seq, struct mddev *mddev)
 {
-	conf_t *conf = (conf_t*)mddev->private;
+	struct faulty_conf *conf = mddev->private;
 	int n;
 
 	if ((n=atomic_read(&conf->counters[WriteTransient])) != 0)
@@ -252,14 +259,14 @@ static void status(struct seq_file *seq, mddev_t *mddev)
 }
 
 
-static int reconfig(mddev_t *mddev, int layout, int chunk_size)
+static int reshape(struct mddev *mddev)
 {
-	int mode = layout & ModeMask;
-	int count = layout >> ModeShift;
-	conf_t *conf = mddev->private;
+	int mode = mddev->new_layout & ModeMask;
+	int count = mddev->new_layout >> ModeShift;
+	struct faulty_conf *conf = mddev->private;
 
-	if (chunk_size != -1)
-		return -EINVAL;
+	if (mddev->new_layout < 0)
+		return 0;
 
 	/* new layout */
 	if (mode == ClearFaults)
@@ -276,17 +283,34 @@ static int reconfig(mddev_t *mddev, int layout, int chunk_size)
 		atomic_set(&conf->counters[mode], count);
 	} else
 		return -EINVAL;
+	mddev->new_layout = -1;
 	mddev->layout = -1; /* makes sure further changes come through */
 	return 0;
 }
 
-static int run(mddev_t *mddev)
+static sector_t faulty_size(struct mddev *mddev, sector_t sectors, int raid_disks)
 {
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	WARN_ONCE(raid_disks,
+		  "%s does not support generic reshape\n", __func__);
+
+	if (sectors == 0)
+		return mddev->dev_sectors;
+
+	return sectors;
+}
+
+static int run(struct mddev *mddev)
+{
+	struct md_rdev *rdev;
 	int i;
+	struct faulty_conf *conf;
+
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
 
-	conf_t *conf = kmalloc(sizeof(*conf), GFP_KERNEL);
+	conf = kmalloc(sizeof(*conf), GFP_KERNEL);
+	if (!conf)
+		return -ENOMEM;
 
 	for (i=0; i<Modes; i++) {
 		atomic_set(&conf->counters[i], 0);
@@ -294,27 +318,30 @@ static int run(mddev_t *mddev)
 	}
 	conf->nfaults = 0;
 
-	rdev_for_each(rdev, tmp, mddev)
+	rdev_for_each(rdev, mddev) {
 		conf->rdev = rdev;
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
+	}
 
-	mddev->array_sectors = mddev->size * 2;
+	md_set_array_sectors(mddev, faulty_size(mddev, 0, 0));
 	mddev->private = conf;
 
-	reconfig(mddev, mddev->layout, -1);
+	reshape(mddev);
 
 	return 0;
 }
 
-static int stop(mddev_t *mddev)
+static int stop(struct mddev *mddev)
 {
-	conf_t *conf = (conf_t *)mddev->private;
+	struct faulty_conf *conf = mddev->private;
 
 	kfree(conf);
 	mddev->private = NULL;
 	return 0;
 }
 
-static struct mdk_personality faulty_personality =
+static struct md_personality faulty_personality =
 {
 	.name		= "faulty",
 	.level		= LEVEL_FAULTY,
@@ -323,7 +350,8 @@ static struct mdk_personality faulty_personality =
 	.run		= run,
 	.stop		= stop,
 	.status		= status,
-	.reconfig	= reconfig,
+	.check_reshape	= reshape,
+	.size		= faulty_size,
 };
 
 static int __init raid_init(void)
@@ -339,6 +367,7 @@ static void raid_exit(void)
 module_init(raid_init);
 module_exit(raid_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Fault injection personality for MD");
 MODULE_ALIAS("md-personality-10"); /* faulty */
 MODULE_ALIAS("md-faulty");
 MODULE_ALIAS("md-level--5");
diff --git a/drivers/md/linear.c b/drivers/md/linear.c
index b1eebf88c20..56f534b4a2d 100644
--- a/drivers/md/linear.c
+++ b/drivers/md/linear.c
@@ -16,35 +16,40 @@
    Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  
 */
 
+#include <linux/blkdev.h>
+#include <linux/raid/md_u.h>
+#include <linux/seq_file.h>
 #include <linux/module.h>
-
-#include <linux/raid/md.h>
 #include <linux/slab.h>
-#include <linux/raid/linear.h>
-
-#define MAJOR_NR MD_MAJOR
-#define MD_DRIVER
-#define MD_PERSONALITY
+#include "md.h"
+#include "linear.h"
 
 /*
  * find which device holds a particular offset 
  */
-static inline dev_info_t *which_dev(mddev_t *mddev, sector_t sector)
+static inline struct dev_info *which_dev(struct mddev *mddev, sector_t sector)
 {
-	dev_info_t *hash;
-	linear_conf_t *conf = mddev_to_conf(mddev);
-	sector_t block = sector >> 1;
+	int lo, mid, hi;
+	struct linear_conf *conf;
+
+	lo = 0;
+	hi = mddev->raid_disks - 1;
+	conf = rcu_dereference(mddev->private);
 
 	/*
-	 * sector_div(a,b) returns the remainer and sets a to a/b
+	 * Binary Search
 	 */
-	block >>= conf->preshift;
-	(void)sector_div(block, conf->hash_spacing);
-	hash = conf->hash_table[block];
 
-	while ((sector>>1) >= (hash->size + hash->offset))
-		hash++;
-	return hash;
+	while (hi > lo) {
+
+		mid = (hi + lo) / 2;
+		if (sector < conf->disks[mid].end_sector)
+			hi = mid;
+		else
+			lo = mid + 1;
+	}
+
+	return conf->disks + lo;
 }
 
 /**
@@ -59,13 +64,24 @@ static int linear_mergeable_bvec(struct request_queue *q,
 				 struct bvec_merge_data *bvm,
 				 struct bio_vec *biovec)
 {
-	mddev_t *mddev = q->queuedata;
-	dev_info_t *dev0;
+	struct mddev *mddev = q->queuedata;
+	struct dev_info *dev0;
 	unsigned long maxsectors, bio_sectors = bvm->bi_size >> 9;
 	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
+	int maxbytes = biovec->bv_len;
+	struct request_queue *subq;
 
+	rcu_read_lock();
 	dev0 = which_dev(mddev, sector);
-	maxsectors = (dev0->size << 1) - (sector - (dev0->offset<<1));
+	maxsectors = dev0->end_sector - sector;
+	subq = bdev_get_queue(dev0->rdev->bdev);
+	if (subq->merge_bvec_fn) {
+		bvm->bi_bdev = dev0->rdev->bdev;
+		bvm->bi_sector -= dev0->end_sector - dev0->rdev->sectors;
+		maxbytes = min(maxbytes, subq->merge_bvec_fn(subq, bvm,
+							     biovec));
+	}
+	rcu_read_unlock();
 
 	if (maxsectors < bio_sectors)
 		maxsectors = 0;
@@ -73,50 +89,58 @@ static int linear_mergeable_bvec(struct request_queue *q,
 		maxsectors -= bio_sectors;
 
 	if (maxsectors <= (PAGE_SIZE >> 9 ) && bio_sectors == 0)
-		return biovec->bv_len;
-	/* The bytes available at this offset could be really big,
-	 * so we cap at 2^31 to avoid overflow */
-	if (maxsectors > (1 << (31-9)))
-		return 1<<31;
-	return maxsectors << 9;
-}
-
-static void linear_unplug(struct request_queue *q)
-{
-	mddev_t *mddev = q->queuedata;
-	linear_conf_t *conf = mddev_to_conf(mddev);
-	int i;
+		return maxbytes;
 
-	for (i=0; i < mddev->raid_disks; i++) {
-		struct request_queue *r_queue = bdev_get_queue(conf->disks[i].rdev->bdev);
-		blk_unplug(r_queue);
-	}
+	if (maxsectors > (maxbytes >> 9))
+		return maxbytes;
+	else
+		return maxsectors << 9;
 }
 
 static int linear_congested(void *data, int bits)
 {
-	mddev_t *mddev = data;
-	linear_conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = data;
+	struct linear_conf *conf;
 	int i, ret = 0;
 
+	if (mddev_congested(mddev, bits))
+		return 1;
+
+	rcu_read_lock();
+	conf = rcu_dereference(mddev->private);
+
 	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
 		struct request_queue *q = bdev_get_queue(conf->disks[i].rdev->bdev);
 		ret |= bdi_congested(&q->backing_dev_info, bits);
 	}
+
+	rcu_read_unlock();
 	return ret;
 }
 
-static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
+static sector_t linear_size(struct mddev *mddev, sector_t sectors, int raid_disks)
+{
+	struct linear_conf *conf;
+	sector_t array_sectors;
+
+	rcu_read_lock();
+	conf = rcu_dereference(mddev->private);
+	WARN_ONCE(sectors || raid_disks,
+		  "%s does not support generic reshape\n", __func__);
+	array_sectors = conf->array_sectors;
+	rcu_read_unlock();
+
+	return array_sectors;
+}
+
+static struct linear_conf *linear_conf(struct mddev *mddev, int raid_disks)
 {
-	linear_conf_t *conf;
-	dev_info_t **table;
-	mdk_rdev_t *rdev;
-	int i, nb_zone, cnt;
-	sector_t min_spacing;
-	sector_t curr_offset;
-	struct list_head *tmp;
-
-	conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(dev_info_t),
+	struct linear_conf *conf;
+	struct md_rdev *rdev;
+	int i, cnt;
+	bool discard_supported = false;
+
+	conf = kzalloc (sizeof (*conf) + raid_disks*sizeof(struct dev_info),
 			GFP_KERNEL);
 	if (!conf)
 		return NULL;
@@ -124,122 +148,53 @@ static linear_conf_t *linear_conf(mddev_t *mddev, int raid_disks)
 	cnt = 0;
 	conf->array_sectors = 0;
 
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev) {
 		int j = rdev->raid_disk;
-		dev_info_t *disk = conf->disks + j;
+		struct dev_info *disk = conf->disks + j;
+		sector_t sectors;
 
 		if (j < 0 || j >= raid_disks || disk->rdev) {
-			printk("linear: disk numbering problem. Aborting!\n");
+			printk(KERN_ERR "md/linear:%s: disk numbering problem. Aborting!\n",
+			       mdname(mddev));
 			goto out;
 		}
 
 		disk->rdev = rdev;
+		if (mddev->chunk_sectors) {
+			sectors = rdev->sectors;
+			sector_div(sectors, mddev->chunk_sectors);
+			rdev->sectors = sectors * mddev->chunk_sectors;
+		}
 
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
-		/* as we don't honour merge_bvec_fn, we must never risk
-		 * violating it, so limit ->max_sector to one PAGE, as
-		 * a one page request is never in violation.
-		 */
-		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
-
-		disk->size = rdev->size;
-		conf->array_sectors += rdev->size * 2;
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
 
+		conf->array_sectors += rdev->sectors;
 		cnt++;
+
+		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
+			discard_supported = true;
 	}
 	if (cnt != raid_disks) {
-		printk("linear: not enough drives present. Aborting!\n");
+		printk(KERN_ERR "md/linear:%s: not enough drives present. Aborting!\n",
+		       mdname(mddev));
 		goto out;
 	}
 
-	min_spacing = conf->array_sectors / 2;
-	sector_div(min_spacing, PAGE_SIZE/sizeof(struct dev_info *));
-
-	/* min_spacing is the minimum spacing that will fit the hash
-	 * table in one PAGE.  This may be much smaller than needed.
-	 * We find the smallest non-terminal set of consecutive devices
-	 * that is larger than min_spacing as use the size of that as
-	 * the actual spacing
-	 */
-	conf->hash_spacing = conf->array_sectors / 2;
-	for (i=0; i < cnt-1 ; i++) {
-		sector_t sz = 0;
-		int j;
-		for (j = i; j < cnt - 1 && sz < min_spacing; j++)
-			sz += conf->disks[j].size;
-		if (sz >= min_spacing && sz < conf->hash_spacing)
-			conf->hash_spacing = sz;
-	}
+	if (!discard_supported)
+		queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
+	else
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
 
-	/* hash_spacing may be too large for sector_div to work with,
-	 * so we might need to pre-shift
-	 */
-	conf->preshift = 0;
-	if (sizeof(sector_t) > sizeof(u32)) {
-		sector_t space = conf->hash_spacing;
-		while (space > (sector_t)(~(u32)0)) {
-			space >>= 1;
-			conf->preshift++;
-		}
-	}
 	/*
-	 * This code was restructured to work around a gcc-2.95.3 internal
-	 * compiler error.  Alter it with care.
+	 * Here we calculate the device offsets.
 	 */
-	{
-		sector_t sz;
-		unsigned round;
-		unsigned long base;
-
-		sz = conf->array_sectors >> (conf->preshift + 1);
-		sz += 1; /* force round-up */
-		base = conf->hash_spacing >> conf->preshift;
-		round = sector_div(sz, base);
-		nb_zone = sz + (round ? 1 : 0);
-	}
-	BUG_ON(nb_zone > PAGE_SIZE / sizeof(struct dev_info *));
+	conf->disks[0].end_sector = conf->disks[0].rdev->sectors;
 
-	conf->hash_table = kmalloc (sizeof (struct dev_info *) * nb_zone,
-					GFP_KERNEL);
-	if (!conf->hash_table)
-		goto out;
-
-	/*
-	 * Here we generate the linear hash table
-	 * First calculate the device offsets.
-	 */
-	conf->disks[0].offset = 0;
 	for (i = 1; i < raid_disks; i++)
-		conf->disks[i].offset =
-			conf->disks[i-1].offset +
-			conf->disks[i-1].size;
-
-	table = conf->hash_table;
-	curr_offset = 0;
-	i = 0;
-	for (curr_offset = 0;
-	     curr_offset < conf->array_sectors / 2;
-	     curr_offset += conf->hash_spacing) {
-
-		while (i < raid_disks-1 &&
-		       curr_offset >= conf->disks[i+1].offset)
-			i++;
-
-		*table ++ = conf->disks + i;
-	}
-
-	if (conf->preshift) {
-		conf->hash_spacing >>= conf->preshift;
-		/* round hash_spacing up so that when we divide by it,
-		 * we err on the side of "too-low", which is safest.
-		 */
-		conf->hash_spacing++;
-	}
-
-	BUG_ON(table - conf->hash_table > nb_zone);
+		conf->disks[i].end_sector =
+			conf->disks[i-1].end_sector +
+			conf->disks[i].rdev->sectors;
 
 	return conf;
 
@@ -248,26 +203,33 @@ out:
 	return NULL;
 }
 
-static int linear_run (mddev_t *mddev)
+static int linear_run (struct mddev *mddev)
 {
-	linear_conf_t *conf;
+	struct linear_conf *conf;
+	int ret;
 
-	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
 	conf = linear_conf(mddev, mddev->raid_disks);
 
 	if (!conf)
 		return 1;
 	mddev->private = conf;
-	mddev->array_sectors = conf->array_sectors;
+	md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
 
 	blk_queue_merge_bvec(mddev->queue, linear_mergeable_bvec);
-	mddev->queue->unplug_fn = linear_unplug;
 	mddev->queue->backing_dev_info.congested_fn = linear_congested;
 	mddev->queue->backing_dev_info.congested_data = mddev;
-	return 0;
+
+	ret =  md_integrity_register(mddev);
+	if (ret) {
+		kfree(conf);
+		mddev->private = NULL;
+	}
+	return ret;
 }
 
-static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
+static int linear_add(struct mddev *mddev, struct md_rdev *rdev)
 {
 	/* Adding a drive to a linear array allows the array to grow.
 	 * It is permitted if the new drive has a matching superblock
@@ -277,125 +239,124 @@ static int linear_add(mddev_t *mddev, mdk_rdev_t *rdev)
 	 * The current one is never freed until the array is stopped.
 	 * This avoids races.
 	 */
-	linear_conf_t *newconf;
+	struct linear_conf *newconf, *oldconf;
 
 	if (rdev->saved_raid_disk != mddev->raid_disks)
 		return -EINVAL;
 
 	rdev->raid_disk = rdev->saved_raid_disk;
+	rdev->saved_raid_disk = -1;
 
 	newconf = linear_conf(mddev,mddev->raid_disks+1);
 
 	if (!newconf)
 		return -ENOMEM;
 
-	newconf->prev = mddev_to_conf(mddev);
-	mddev->private = newconf;
+	oldconf = rcu_dereference_protected(mddev->private,
+					    lockdep_is_held(
+						    &mddev->reconfig_mutex));
 	mddev->raid_disks++;
-	mddev->array_sectors = newconf->array_sectors;
+	rcu_assign_pointer(mddev->private, newconf);
+	md_set_array_sectors(mddev, linear_size(mddev, 0, 0));
 	set_capacity(mddev->gendisk, mddev->array_sectors);
+	revalidate_disk(mddev->gendisk);
+	kfree_rcu(oldconf, rcu);
 	return 0;
 }
 
-static int linear_stop (mddev_t *mddev)
+static int linear_stop (struct mddev *mddev)
 {
-	linear_conf_t *conf = mddev_to_conf(mddev);
-  
+	struct linear_conf *conf =
+		rcu_dereference_protected(mddev->private,
+					  lockdep_is_held(
+						  &mddev->reconfig_mutex));
+
+	/*
+	 * We do not require rcu protection here since
+	 * we hold reconfig_mutex for both linear_add and
+	 * linear_stop, so they cannot race.
+	 * We should make sure any old 'conf's are properly
+	 * freed though.
+	 */
+	rcu_barrier();
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
-	do {
-		linear_conf_t *t = conf->prev;
-		kfree(conf->hash_table);
-		kfree(conf);
-		conf = t;
-	} while (conf);
+	kfree(conf);
+	mddev->private = NULL;
 
 	return 0;
 }
 
-static int linear_make_request (struct request_queue *q, struct bio *bio)
+static void linear_make_request(struct mddev *mddev, struct bio *bio)
 {
-	const int rw = bio_data_dir(bio);
-	mddev_t *mddev = q->queuedata;
-	dev_info_t *tmp_dev;
-	sector_t block;
-
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+	char b[BDEVNAME_SIZE];
+	struct dev_info *tmp_dev;
+	struct bio *split;
+	sector_t start_sector, end_sector, data_offset;
+
+	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
+		md_flush_request(mddev, bio);
+		return;
 	}
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
-
-	tmp_dev = which_dev(mddev, bio->bi_sector);
-	block = bio->bi_sector >> 1;
-    
-	if (unlikely(block >= (tmp_dev->size + tmp_dev->offset)
-		     || block < tmp_dev->offset)) {
-		char b[BDEVNAME_SIZE];
-
-		printk("linear_make_request: Block %llu out of bounds on "
-			"dev %s size %llu offset %llu\n",
-			(unsigned long long)block,
-			bdevname(tmp_dev->rdev->bdev, b),
-			(unsigned long long)tmp_dev->size,
-		        (unsigned long long)tmp_dev->offset);
-		bio_io_error(bio);
-		return 0;
-	}
-	if (unlikely(bio->bi_sector + (bio->bi_size >> 9) >
-		     (tmp_dev->offset + tmp_dev->size)<<1)) {
-		/* This bio crosses a device boundary, so we have to
-		 * split it.
-		 */
-		struct bio_pair *bp;
-		bp = bio_split(bio, bio_split_pool,
-			       ((tmp_dev->offset + tmp_dev->size)<<1) - bio->bi_sector);
-		if (linear_make_request(q, &bp->bio1))
-			generic_make_request(&bp->bio1);
-		if (linear_make_request(q, &bp->bio2))
-			generic_make_request(&bp->bio2);
-		bio_pair_release(bp);
-		return 0;
-	}
-		    
-	bio->bi_bdev = tmp_dev->rdev->bdev;
-	bio->bi_sector = bio->bi_sector - (tmp_dev->offset << 1) + tmp_dev->rdev->data_offset;
+	do {
+		rcu_read_lock();
+
+		tmp_dev = which_dev(mddev, bio->bi_iter.bi_sector);
+		start_sector = tmp_dev->end_sector - tmp_dev->rdev->sectors;
+		end_sector = tmp_dev->end_sector;
+		data_offset = tmp_dev->rdev->data_offset;
+		bio->bi_bdev = tmp_dev->rdev->bdev;
+
+		rcu_read_unlock();
+
+		if (unlikely(bio->bi_iter.bi_sector >= end_sector ||
+			     bio->bi_iter.bi_sector < start_sector))
+			goto out_of_bounds;
+
+		if (unlikely(bio_end_sector(bio) > end_sector)) {
+			/* This bio crosses a device boundary, so we have to
+			 * split it.
+			 */
+			split = bio_split(bio, end_sector -
+					  bio->bi_iter.bi_sector,
+					  GFP_NOIO, fs_bio_set);
+			bio_chain(split, bio);
+		} else {
+			split = bio;
+		}
 
-	return 1;
+		split->bi_iter.bi_sector = split->bi_iter.bi_sector -
+			start_sector + data_offset;
+
+		if (unlikely((split->bi_rw & REQ_DISCARD) &&
+			 !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
+			/* Just ignore it */
+			bio_endio(split, 0);
+		} else
+			generic_make_request(split);
+	} while (split != bio);
+	return;
+
+out_of_bounds:
+	printk(KERN_ERR
+	       "md/linear:%s: make_request: Sector %llu out of bounds on "
+	       "dev %s: %llu sectors, offset %llu\n",
+	       mdname(mddev),
+	       (unsigned long long)bio->bi_iter.bi_sector,
+	       bdevname(tmp_dev->rdev->bdev, b),
+	       (unsigned long long)tmp_dev->rdev->sectors,
+	       (unsigned long long)start_sector);
+	bio_io_error(bio);
 }
 
-static void linear_status (struct seq_file *seq, mddev_t *mddev)
+static void linear_status (struct seq_file *seq, struct mddev *mddev)
 {
 
-#undef MD_DEBUG
-#ifdef MD_DEBUG
-	int j;
-	linear_conf_t *conf = mddev_to_conf(mddev);
-	sector_t s = 0;
-  
-	seq_printf(seq, "      ");
-	for (j = 0; j < mddev->raid_disks; j++)
-	{
-		char b[BDEVNAME_SIZE];
-		s += conf->smallest_size;
-		seq_printf(seq, "[%s",
-			   bdevname(conf->hash_table[j][0].rdev->bdev,b));
-
-		while (s > conf->hash_table[j][0].offset +
-		           conf->hash_table[j][0].size)
-			seq_printf(seq, "/%s] ",
-				   bdevname(conf->hash_table[j][1].rdev->bdev,b));
-		else
-			seq_printf(seq, "] ");
-	}
-	seq_printf(seq, "\n");
-#endif
-	seq_printf(seq, " %dk rounding", mddev->chunk_size/1024);
+	seq_printf(seq, " %dk rounding", mddev->chunk_sectors / 2);
 }
 
 
-static struct mdk_personality linear_personality =
+static struct md_personality linear_personality =
 {
 	.name		= "linear",
 	.level		= LEVEL_LINEAR,
@@ -405,6 +366,7 @@ static struct mdk_personality linear_personality =
 	.stop		= linear_stop,
 	.status		= linear_status,
 	.hot_add_disk	= linear_add,
+	.size		= linear_size,
 };
 
 static int __init linear_init (void)
@@ -421,6 +383,7 @@ static void linear_exit (void)
 module_init(linear_init);
 module_exit(linear_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("Linear device concatenation personality for MD");
 MODULE_ALIAS("md-personality-1"); /* LINEAR - deprecated*/
 MODULE_ALIAS("md-linear");
 MODULE_ALIAS("md-level--1");
diff --git a/drivers/md/linear.h b/drivers/md/linear.h
new file mode 100644
index 00000000000..b685ddd7d7f
--- /dev/null
+++ b/drivers/md/linear.h
@@ -0,0 +1,15 @@
+#ifndef _LINEAR_H
+#define _LINEAR_H
+
+struct dev_info {
+	struct md_rdev	*rdev;
+	sector_t	end_sector;
+};
+
+struct linear_conf
+{
+	struct rcu_head		rcu;
+	sector_t		array_sectors;
+	struct dev_info		disks[0];
+};
+#endif
diff --git a/drivers/md/md.c b/drivers/md/md.c
index 4790c83d78d..32fc19c540d 100644
--- a/drivers/md/md.c
+++ b/drivers/md/md.c
@@ -32,53 +32,58 @@
    Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
 */
 
-#include <linux/module.h>
-#include <linux/kernel.h>
 #include <linux/kthread.h>
-#include <linux/linkage.h>
-#include <linux/raid/md.h>
-#include <linux/raid/bitmap.h>
+#include <linux/blkdev.h>
 #include <linux/sysctl.h>
-#include <linux/buffer_head.h> /* for invalidate_bdev */
+#include <linux/seq_file.h>
+#include <linux/fs.h>
 #include <linux/poll.h>
-#include <linux/mutex.h>
 #include <linux/ctype.h>
-#include <linux/freezer.h>
-
-#include <linux/init.h>
-
+#include <linux/string.h>
+#include <linux/hdreg.h>
+#include <linux/proc_fs.h>
+#include <linux/random.h>
+#include <linux/module.h>
+#include <linux/reboot.h>
 #include <linux/file.h>
-
-#ifdef CONFIG_KMOD
-#include <linux/kmod.h>
-#endif
-
-#include <asm/unaligned.h>
-
-#define MAJOR_NR MD_MAJOR
-#define MD_DRIVER
-
-/* 63 partitions with the alternate major number (mdp) */
-#define MdpMinorShift 6
-
-#define DEBUG 0
-#define dprintk(x...) ((void)(DEBUG && printk(x)))
-
+#include <linux/compat.h>
+#include <linux/delay.h>
+#include <linux/raid/md_p.h>
+#include <linux/raid/md_u.h>
+#include <linux/slab.h>
+#include "md.h"
+#include "bitmap.h"
 
 #ifndef MODULE
-static void autostart_arrays (int part);
+static void autostart_arrays(int part);
 #endif
 
+/* pers_list is a list of registered personalities protected
+ * by pers_lock.
+ * pers_lock does extra service to protect accesses to
+ * mddev->thread when the mutex cannot be held.
+ */
 static LIST_HEAD(pers_list);
 static DEFINE_SPINLOCK(pers_lock);
 
 static void md_print_devices(void);
 
 static DECLARE_WAIT_QUEUE_HEAD(resync_wait);
+static struct workqueue_struct *md_wq;
+static struct workqueue_struct *md_misc_wq;
+
+static int remove_and_add_spares(struct mddev *mddev,
+				 struct md_rdev *this);
 
 #define MD_BUG(x...) { printk("md: bug in file %s, line %d\n", __FILE__, __LINE__); md_print_devices(); }
 
 /*
+ * Default number of read corrections we'll attempt on an rdev
+ * before ejecting it from the array. We divide the read error
+ * count by 2 for every hour elapsed between read errors.
+ */
+#define MD_DEFAULT_MAX_CORRECTED_READ_ERRORS 20
+/*
  * Current RAID-1,4,5 parallel reconstruction 'guaranteed speed limit'
  * is 1000 KB/sec, so the extra system load does not show up that much.
  * Increase it if you want to have more _guaranteed_ speed. Note that
@@ -93,13 +98,13 @@ static DECLARE_WAIT_QUEUE_HEAD(resync_wait);
 
 static int sysctl_speed_limit_min = 1000;
 static int sysctl_speed_limit_max = 200000;
-static inline int speed_min(mddev_t *mddev)
+static inline int speed_min(struct mddev *mddev)
 {
 	return mddev->sync_speed_min ?
 		mddev->sync_speed_min : sysctl_speed_limit_min;
 }
 
-static inline int speed_max(mddev_t *mddev)
+static inline int speed_max(struct mddev *mddev)
 {
 	return mddev->sync_speed_max ?
 		mddev->sync_speed_max : sysctl_speed_limit_max;
@@ -107,52 +112,77 @@ static inline int speed_max(mddev_t *mddev)
 
 static struct ctl_table_header *raid_table_header;
 
-static ctl_table raid_table[] = {
+static struct ctl_table raid_table[] = {
 	{
-		.ctl_name	= DEV_RAID_SPEED_LIMIT_MIN,
 		.procname	= "speed_limit_min",
 		.data		= &sysctl_speed_limit_min,
 		.maxlen		= sizeof(int),
 		.mode		= S_IRUGO|S_IWUSR,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
 	{
-		.ctl_name	= DEV_RAID_SPEED_LIMIT_MAX,
 		.procname	= "speed_limit_max",
 		.data		= &sysctl_speed_limit_max,
 		.maxlen		= sizeof(int),
 		.mode		= S_IRUGO|S_IWUSR,
-		.proc_handler	= &proc_dointvec,
+		.proc_handler	= proc_dointvec,
 	},
-	{ .ctl_name = 0 }
+	{ }
 };
 
-static ctl_table raid_dir_table[] = {
+static struct ctl_table raid_dir_table[] = {
 	{
-		.ctl_name	= DEV_RAID,
 		.procname	= "raid",
 		.maxlen		= 0,
 		.mode		= S_IRUGO|S_IXUGO,
 		.child		= raid_table,
 	},
-	{ .ctl_name = 0 }
+	{ }
 };
 
-static ctl_table raid_root_table[] = {
+static struct ctl_table raid_root_table[] = {
 	{
-		.ctl_name	= CTL_DEV,
 		.procname	= "dev",
 		.maxlen		= 0,
 		.mode		= 0555,
 		.child		= raid_dir_table,
 	},
-	{ .ctl_name = 0 }
+	{  }
 };
 
-static struct block_device_operations md_fops;
+static const struct block_device_operations md_fops;
 
 static int start_readonly;
 
+/* bio_clone_mddev
+ * like bio_clone, but with a local bio set
+ */
+
+struct bio *bio_alloc_mddev(gfp_t gfp_mask, int nr_iovecs,
+			    struct mddev *mddev)
+{
+	struct bio *b;
+
+	if (!mddev || !mddev->bio_set)
+		return bio_alloc(gfp_mask, nr_iovecs);
+
+	b = bio_alloc_bioset(gfp_mask, nr_iovecs, mddev->bio_set);
+	if (!b)
+		return NULL;
+	return b;
+}
+EXPORT_SYMBOL_GPL(bio_alloc_mddev);
+
+struct bio *bio_clone_mddev(struct bio *bio, gfp_t gfp_mask,
+			    struct mddev *mddev)
+{
+	if (!mddev || !mddev->bio_set)
+		return bio_clone(bio, gfp_mask);
+
+	return bio_clone_bioset(bio, gfp_mask, mddev->bio_set);
+}
+EXPORT_SYMBOL_GPL(bio_clone_mddev);
+
 /*
  * We have a system wide 'event count' that is incremented
  * on any 'interesting' event, and readers of /proc/mdstat
@@ -165,7 +195,7 @@ static int start_readonly;
  */
 static DECLARE_WAIT_QUEUE_HEAD(md_event_waiters);
 static atomic_t md_event_count;
-void md_new_event(mddev_t *mddev)
+void md_new_event(struct mddev *mddev)
 {
 	atomic_inc(&md_event_count);
 	wake_up(&md_event_waiters);
@@ -175,7 +205,7 @@ EXPORT_SYMBOL_GPL(md_new_event);
 /* Alternate version that can be called from interrupts
  * when calling sysfs_notify isn't needed.
  */
-static void md_new_event_inintr(mddev_t *mddev)
+static void md_new_event_inintr(struct mddev *mddev)
 {
 	atomic_inc(&md_event_count);
 	wake_up(&md_event_waiters);
@@ -196,62 +226,321 @@ static DEFINE_SPINLOCK(all_mddevs_lock);
  * Any code which breaks out of this loop while own
  * a reference to the current mddev and must mddev_put it.
  */
-#define for_each_mddev(mddev,tmp)					\
+#define for_each_mddev(_mddev,_tmp)					\
 									\
 	for (({ spin_lock(&all_mddevs_lock); 				\
-		tmp = all_mddevs.next;					\
-		mddev = NULL;});					\
-	     ({ if (tmp != &all_mddevs)					\
-			mddev_get(list_entry(tmp, mddev_t, all_mddevs));\
+		_tmp = all_mddevs.next;					\
+		_mddev = NULL;});					\
+	     ({ if (_tmp != &all_mddevs)				\
+			mddev_get(list_entry(_tmp, struct mddev, all_mddevs));\
 		spin_unlock(&all_mddevs_lock);				\
-		if (mddev) mddev_put(mddev);				\
-		mddev = list_entry(tmp, mddev_t, all_mddevs);		\
-		tmp != &all_mddevs;});					\
+		if (_mddev) mddev_put(_mddev);				\
+		_mddev = list_entry(_tmp, struct mddev, all_mddevs);	\
+		_tmp != &all_mddevs;});					\
 	     ({ spin_lock(&all_mddevs_lock);				\
-		tmp = tmp->next;})					\
+		_tmp = _tmp->next;})					\
 		)
 
 
-static int md_fail_request (struct request_queue *q, struct bio *bio)
+/* Rather than calling directly into the personality make_request function,
+ * IO requests come here first so that we can check if the device is
+ * being suspended pending a reconfiguration.
+ * We hold a refcount over the call to ->make_request.  By the time that
+ * call has finished, the bio has been linked into some internal structure
+ * and so is visible to ->quiesce(), so we don't need the refcount any more.
+ */
+static void md_make_request(struct request_queue *q, struct bio *bio)
 {
-	bio_io_error(bio);
-	return 0;
+	const int rw = bio_data_dir(bio);
+	struct mddev *mddev = q->queuedata;
+	int cpu;
+	unsigned int sectors;
+
+	if (mddev == NULL || mddev->pers == NULL
+	    || !mddev->ready) {
+		bio_io_error(bio);
+		return;
+	}
+	if (mddev->ro == 1 && unlikely(rw == WRITE)) {
+		bio_endio(bio, bio_sectors(bio) == 0 ? 0 : -EROFS);
+		return;
+	}
+	smp_rmb(); /* Ensure implications of  'active' are visible */
+	rcu_read_lock();
+	if (mddev->suspended) {
+		DEFINE_WAIT(__wait);
+		for (;;) {
+			prepare_to_wait(&mddev->sb_wait, &__wait,
+					TASK_UNINTERRUPTIBLE);
+			if (!mddev->suspended)
+				break;
+			rcu_read_unlock();
+			schedule();
+			rcu_read_lock();
+		}
+		finish_wait(&mddev->sb_wait, &__wait);
+	}
+	atomic_inc(&mddev->active_io);
+	rcu_read_unlock();
+
+	/*
+	 * save the sectors now since our bio can
+	 * go away inside make_request
+	 */
+	sectors = bio_sectors(bio);
+	mddev->pers->make_request(mddev, bio);
+
+	cpu = part_stat_lock();
+	part_stat_inc(cpu, &mddev->gendisk->part0, ios[rw]);
+	part_stat_add(cpu, &mddev->gendisk->part0, sectors[rw], sectors);
+	part_stat_unlock();
+
+	if (atomic_dec_and_test(&mddev->active_io) && mddev->suspended)
+		wake_up(&mddev->sb_wait);
+}
+
+/* mddev_suspend makes sure no new requests are submitted
+ * to the device, and that any requests that have been submitted
+ * are completely handled.
+ * Once ->stop is called and completes, the module will be completely
+ * unused.
+ */
+void mddev_suspend(struct mddev *mddev)
+{
+	BUG_ON(mddev->suspended);
+	mddev->suspended = 1;
+	synchronize_rcu();
+	wait_event(mddev->sb_wait, atomic_read(&mddev->active_io) == 0);
+	mddev->pers->quiesce(mddev, 1);
+
+	del_timer_sync(&mddev->safemode_timer);
+}
+EXPORT_SYMBOL_GPL(mddev_suspend);
+
+void mddev_resume(struct mddev *mddev)
+{
+	mddev->suspended = 0;
+	wake_up(&mddev->sb_wait);
+	mddev->pers->quiesce(mddev, 0);
+
+	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	md_wakeup_thread(mddev->thread);
+	md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
+}
+EXPORT_SYMBOL_GPL(mddev_resume);
+
+int mddev_congested(struct mddev *mddev, int bits)
+{
+	return mddev->suspended;
 }
+EXPORT_SYMBOL(mddev_congested);
+
+/*
+ * Generic flush handling for md
+ */
+
+static void md_end_flush(struct bio *bio, int err)
+{
+	struct md_rdev *rdev = bio->bi_private;
+	struct mddev *mddev = rdev->mddev;
+
+	rdev_dec_pending(rdev, mddev);
+
+	if (atomic_dec_and_test(&mddev->flush_pending)) {
+		/* The pre-request flush has finished */
+		queue_work(md_wq, &mddev->flush_work);
+	}
+	bio_put(bio);
+}
+
+static void md_submit_flush_data(struct work_struct *ws);
 
-static inline mddev_t *mddev_get(mddev_t *mddev)
+static void submit_flushes(struct work_struct *ws)
+{
+	struct mddev *mddev = container_of(ws, struct mddev, flush_work);
+	struct md_rdev *rdev;
+
+	INIT_WORK(&mddev->flush_work, md_submit_flush_data);
+	atomic_set(&mddev->flush_pending, 1);
+	rcu_read_lock();
+	rdev_for_each_rcu(rdev, mddev)
+		if (rdev->raid_disk >= 0 &&
+		    !test_bit(Faulty, &rdev->flags)) {
+			/* Take two references, one is dropped
+			 * when request finishes, one after
+			 * we reclaim rcu_read_lock
+			 */
+			struct bio *bi;
+			atomic_inc(&rdev->nr_pending);
+			atomic_inc(&rdev->nr_pending);
+			rcu_read_unlock();
+			bi = bio_alloc_mddev(GFP_NOIO, 0, mddev);
+			bi->bi_end_io = md_end_flush;
+			bi->bi_private = rdev;
+			bi->bi_bdev = rdev->bdev;
+			atomic_inc(&mddev->flush_pending);
+			submit_bio(WRITE_FLUSH, bi);
+			rcu_read_lock();
+			rdev_dec_pending(rdev, mddev);
+		}
+	rcu_read_unlock();
+	if (atomic_dec_and_test(&mddev->flush_pending))
+		queue_work(md_wq, &mddev->flush_work);
+}
+
+static void md_submit_flush_data(struct work_struct *ws)
+{
+	struct mddev *mddev = container_of(ws, struct mddev, flush_work);
+	struct bio *bio = mddev->flush_bio;
+
+	if (bio->bi_iter.bi_size == 0)
+		/* an empty barrier - all done */
+		bio_endio(bio, 0);
+	else {
+		bio->bi_rw &= ~REQ_FLUSH;
+		mddev->pers->make_request(mddev, bio);
+	}
+
+	mddev->flush_bio = NULL;
+	wake_up(&mddev->sb_wait);
+}
+
+void md_flush_request(struct mddev *mddev, struct bio *bio)
+{
+	spin_lock_irq(&mddev->write_lock);
+	wait_event_lock_irq(mddev->sb_wait,
+			    !mddev->flush_bio,
+			    mddev->write_lock);
+	mddev->flush_bio = bio;
+	spin_unlock_irq(&mddev->write_lock);
+
+	INIT_WORK(&mddev->flush_work, submit_flushes);
+	queue_work(md_wq, &mddev->flush_work);
+}
+EXPORT_SYMBOL(md_flush_request);
+
+void md_unplug(struct blk_plug_cb *cb, bool from_schedule)
+{
+	struct mddev *mddev = cb->data;
+	md_wakeup_thread(mddev->thread);
+	kfree(cb);
+}
+EXPORT_SYMBOL(md_unplug);
+
+static inline struct mddev *mddev_get(struct mddev *mddev)
 {
 	atomic_inc(&mddev->active);
 	return mddev;
 }
 
-static void mddev_put(mddev_t *mddev)
+static void mddev_delayed_delete(struct work_struct *ws);
+
+static void mddev_put(struct mddev *mddev)
 {
+	struct bio_set *bs = NULL;
+
 	if (!atomic_dec_and_lock(&mddev->active, &all_mddevs_lock))
 		return;
-	if (!mddev->raid_disks && list_empty(&mddev->disks)) {
-		list_del(&mddev->all_mddevs);
-		spin_unlock(&all_mddevs_lock);
-		blk_cleanup_queue(mddev->queue);
-		kobject_put(&mddev->kobj);
-	} else
-		spin_unlock(&all_mddevs_lock);
+	if (!mddev->raid_disks && list_empty(&mddev->disks) &&
+	    mddev->ctime == 0 && !mddev->hold_active) {
+		/* Array is not configured at all, and not held active,
+		 * so destroy it */
+		list_del_init(&mddev->all_mddevs);
+		bs = mddev->bio_set;
+		mddev->bio_set = NULL;
+		if (mddev->gendisk) {
+			/* We did a probe so need to clean up.  Call
+			 * queue_work inside the spinlock so that
+			 * flush_workqueue() after mddev_find will
+			 * succeed in waiting for the work to be done.
+			 */
+			INIT_WORK(&mddev->del_work, mddev_delayed_delete);
+			queue_work(md_misc_wq, &mddev->del_work);
+		} else
+			kfree(mddev);
+	}
+	spin_unlock(&all_mddevs_lock);
+	if (bs)
+		bioset_free(bs);
+}
+
+void mddev_init(struct mddev *mddev)
+{
+	mutex_init(&mddev->open_mutex);
+	mutex_init(&mddev->reconfig_mutex);
+	mutex_init(&mddev->bitmap_info.mutex);
+	INIT_LIST_HEAD(&mddev->disks);
+	INIT_LIST_HEAD(&mddev->all_mddevs);
+	init_timer(&mddev->safemode_timer);
+	atomic_set(&mddev->active, 1);
+	atomic_set(&mddev->openers, 0);
+	atomic_set(&mddev->active_io, 0);
+	spin_lock_init(&mddev->write_lock);
+	atomic_set(&mddev->flush_pending, 0);
+	init_waitqueue_head(&mddev->sb_wait);
+	init_waitqueue_head(&mddev->recovery_wait);
+	mddev->reshape_position = MaxSector;
+	mddev->reshape_backwards = 0;
+	mddev->last_sync_action = "none";
+	mddev->resync_min = 0;
+	mddev->resync_max = MaxSector;
+	mddev->level = LEVEL_NONE;
 }
+EXPORT_SYMBOL_GPL(mddev_init);
 
-static mddev_t * mddev_find(dev_t unit)
+static struct mddev * mddev_find(dev_t unit)
 {
-	mddev_t *mddev, *new = NULL;
+	struct mddev *mddev, *new = NULL;
+
+	if (unit && MAJOR(unit) != MD_MAJOR)
+		unit &= ~((1<<MdpMinorShift)-1);
 
  retry:
 	spin_lock(&all_mddevs_lock);
-	list_for_each_entry(mddev, &all_mddevs, all_mddevs)
-		if (mddev->unit == unit) {
-			mddev_get(mddev);
+
+	if (unit) {
+		list_for_each_entry(mddev, &all_mddevs, all_mddevs)
+			if (mddev->unit == unit) {
+				mddev_get(mddev);
+				spin_unlock(&all_mddevs_lock);
+				kfree(new);
+				return mddev;
+			}
+
+		if (new) {
+			list_add(&new->all_mddevs, &all_mddevs);
 			spin_unlock(&all_mddevs_lock);
-			kfree(new);
-			return mddev;
+			new->hold_active = UNTIL_IOCTL;
+			return new;
 		}
-
-	if (new) {
+	} else if (new) {
+		/* find an unused unit number */
+		static int next_minor = 512;
+		int start = next_minor;
+		int is_free = 0;
+		int dev = 0;
+		while (!is_free) {
+			dev = MKDEV(MD_MAJOR, next_minor);
+			next_minor++;
+			if (next_minor > MINORMASK)
+				next_minor = 0;
+			if (next_minor == start) {
+				/* Oh dear, all in use. */
+				spin_unlock(&all_mddevs_lock);
+				kfree(new);
+				return NULL;
+			}
+				
+			is_free = 1;
+			list_for_each_entry(mddev, &all_mddevs, all_mddevs)
+				if (mddev->unit == dev) {
+					is_free = 0;
+					break;
+				}
+		}
+		new->unit = dev;
+		new->md_minor = MINOR(dev);
+		new->hold_active = UNTIL_STOP;
 		list_add(&new->all_mddevs, &all_mddevs);
 		spin_unlock(&all_mddevs_lock);
 		return new;
@@ -268,77 +557,126 @@ static mddev_t * mddev_find(dev_t unit)
 	else
 		new->md_minor = MINOR(unit) >> MdpMinorShift;
 
-	mutex_init(&new->reconfig_mutex);
-	INIT_LIST_HEAD(&new->disks);
-	INIT_LIST_HEAD(&new->all_mddevs);
-	init_timer(&new->safemode_timer);
-	atomic_set(&new->active, 1);
-	atomic_set(&new->openers, 0);
-	spin_lock_init(&new->write_lock);
-	init_waitqueue_head(&new->sb_wait);
-	init_waitqueue_head(&new->recovery_wait);
-	new->reshape_position = MaxSector;
-	new->resync_min = 0;
-	new->resync_max = MaxSector;
-	new->level = LEVEL_NONE;
-
-	new->queue = blk_alloc_queue(GFP_KERNEL);
-	if (!new->queue) {
-		kfree(new);
-		return NULL;
-	}
-	/* Can be unlocked because the queue is new: no concurrency */
-	queue_flag_set_unlocked(QUEUE_FLAG_CLUSTER, new->queue);
-
-	blk_queue_make_request(new->queue, md_fail_request);
+	mddev_init(new);
 
 	goto retry;
 }
 
-static inline int mddev_lock(mddev_t * mddev)
+static inline int __must_check mddev_lock(struct mddev * mddev)
 {
 	return mutex_lock_interruptible(&mddev->reconfig_mutex);
 }
 
-static inline int mddev_trylock(mddev_t * mddev)
+/* Sometimes we need to take the lock in a situation where
+ * failure due to interrupts is not acceptable.
+ */
+static inline void mddev_lock_nointr(struct mddev * mddev)
+{
+	mutex_lock(&mddev->reconfig_mutex);
+}
+
+static inline int mddev_is_locked(struct mddev *mddev)
+{
+	return mutex_is_locked(&mddev->reconfig_mutex);
+}
+
+static inline int mddev_trylock(struct mddev * mddev)
 {
 	return mutex_trylock(&mddev->reconfig_mutex);
 }
 
-static inline void mddev_unlock(mddev_t * mddev)
+static struct attribute_group md_redundancy_group;
+
+static void mddev_unlock(struct mddev * mddev)
 {
-	mutex_unlock(&mddev->reconfig_mutex);
+	if (mddev->to_remove) {
+		/* These cannot be removed under reconfig_mutex as
+		 * an access to the files will try to take reconfig_mutex
+		 * while holding the file unremovable, which leads to
+		 * a deadlock.
+		 * So hold set sysfs_active while the remove in happeing,
+		 * and anything else which might set ->to_remove or my
+		 * otherwise change the sysfs namespace will fail with
+		 * -EBUSY if sysfs_active is still set.
+		 * We set sysfs_active under reconfig_mutex and elsewhere
+		 * test it under the same mutex to ensure its correct value
+		 * is seen.
+		 */
+		struct attribute_group *to_remove = mddev->to_remove;
+		mddev->to_remove = NULL;
+		mddev->sysfs_active = 1;
+		mutex_unlock(&mddev->reconfig_mutex);
+
+		if (mddev->kobj.sd) {
+			if (to_remove != &md_redundancy_group)
+				sysfs_remove_group(&mddev->kobj, to_remove);
+			if (mddev->pers == NULL ||
+			    mddev->pers->sync_request == NULL) {
+				sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
+				if (mddev->sysfs_action)
+					sysfs_put(mddev->sysfs_action);
+				mddev->sysfs_action = NULL;
+			}
+		}
+		mddev->sysfs_active = 0;
+	} else
+		mutex_unlock(&mddev->reconfig_mutex);
 
+	/* As we've dropped the mutex we need a spinlock to
+	 * make sure the thread doesn't disappear
+	 */
+	spin_lock(&pers_lock);
 	md_wakeup_thread(mddev->thread);
+	spin_unlock(&pers_lock);
 }
 
-static mdk_rdev_t * find_rdev_nr(mddev_t *mddev, int nr)
+static struct md_rdev * find_rdev_nr(struct mddev *mddev, int nr)
 {
-	mdk_rdev_t * rdev;
-	struct list_head *tmp;
+	struct md_rdev *rdev;
 
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev)
 		if (rdev->desc_nr == nr)
 			return rdev;
-	}
+
 	return NULL;
 }
 
-static mdk_rdev_t * find_rdev(mddev_t * mddev, dev_t dev)
+static struct md_rdev *find_rdev_nr_rcu(struct mddev *mddev, int nr)
 {
-	struct list_head *tmp;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
+
+	rdev_for_each_rcu(rdev, mddev)
+		if (rdev->desc_nr == nr)
+			return rdev;
+
+	return NULL;
+}
+
+static struct md_rdev *find_rdev(struct mddev *mddev, dev_t dev)
+{
+	struct md_rdev *rdev;
 
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev)
 		if (rdev->bdev->bd_dev == dev)
 			return rdev;
-	}
+
 	return NULL;
 }
 
-static struct mdk_personality *find_pers(int level, char *clevel)
+static struct md_rdev *find_rdev_rcu(struct mddev *mddev, dev_t dev)
 {
-	struct mdk_personality *pers;
+	struct md_rdev *rdev;
+
+	rdev_for_each_rcu(rdev, mddev)
+		if (rdev->bdev->bd_dev == dev)
+			return rdev;
+
+	return NULL;
+}
+
+static struct md_personality *find_pers(int level, char *clevel)
+{
+	struct md_personality *pers;
 	list_for_each_entry(pers, &pers_list, list) {
 		if (level != LEVEL_NONE && pers->level == level)
 			return pers;
@@ -349,22 +687,13 @@ static struct mdk_personality *find_pers(int level, char *clevel)
 }
 
 /* return the offset of the super block in 512byte sectors */
-static inline sector_t calc_dev_sboffset(struct block_device *bdev)
+static inline sector_t calc_dev_sboffset(struct md_rdev *rdev)
 {
-	sector_t num_sectors = bdev->bd_inode->i_size / 512;
+	sector_t num_sectors = i_size_read(rdev->bdev->bd_inode) / 512;
 	return MD_NEW_SIZE_SECTORS(num_sectors);
 }
 
-static sector_t calc_num_sectors(mdk_rdev_t *rdev, unsigned chunk_size)
-{
-	sector_t num_sectors = rdev->sb_start;
-
-	if (chunk_size)
-		num_sectors &= ~((sector_t)chunk_size/512 - 1);
-	return num_sectors;
-}
-
-static int alloc_disk_sb(mdk_rdev_t * rdev)
+static int alloc_disk_sb(struct md_rdev * rdev)
 {
 	if (rdev->sb_page)
 		MD_BUG();
@@ -378,22 +707,28 @@ static int alloc_disk_sb(mdk_rdev_t * rdev)
 	return 0;
 }
 
-static void free_disk_sb(mdk_rdev_t * rdev)
+void md_rdev_clear(struct md_rdev *rdev)
 {
 	if (rdev->sb_page) {
 		put_page(rdev->sb_page);
 		rdev->sb_loaded = 0;
 		rdev->sb_page = NULL;
 		rdev->sb_start = 0;
-		rdev->size = 0;
+		rdev->sectors = 0;
+	}
+	if (rdev->bb_page) {
+		put_page(rdev->bb_page);
+		rdev->bb_page = NULL;
 	}
+	kfree(rdev->badblocks.page);
+	rdev->badblocks.page = NULL;
 }
-
+EXPORT_SYMBOL_GPL(md_rdev_clear);
 
 static void super_written(struct bio *bio, int error)
 {
-	mdk_rdev_t *rdev = bio->bi_private;
-	mddev_t *mddev = rdev->mddev;
+	struct md_rdev *rdev = bio->bi_private;
+	struct mddev *mddev = rdev->mddev;
 
 	if (error || !test_bit(BIO_UPTODATE, &bio->bi_flags)) {
 		printk("md: super_written gets error=%d, uptodate=%d\n",
@@ -407,32 +742,7 @@ static void super_written(struct bio *bio, int error)
 	bio_put(bio);
 }
 
-static void super_written_barrier(struct bio *bio, int error)
-{
-	struct bio *bio2 = bio->bi_private;
-	mdk_rdev_t *rdev = bio2->bi_private;
-	mddev_t *mddev = rdev->mddev;
-
-	if (!test_bit(BIO_UPTODATE, &bio->bi_flags) &&
-	    error == -EOPNOTSUPP) {
-		unsigned long flags;
-		/* barriers don't appear to be supported :-( */
-		set_bit(BarriersNotsupp, &rdev->flags);
-		mddev->barriers_work = 0;
-		spin_lock_irqsave(&mddev->write_lock, flags);
-		bio2->bi_next = mddev->biolist;
-		mddev->biolist = bio2;
-		spin_unlock_irqrestore(&mddev->write_lock, flags);
-		wake_up(&mddev->sb_wait);
-		bio_put(bio);
-	} else {
-		bio_put(bio2);
-		bio->bi_private = rdev;
-		super_written(bio, error);
-	}
-}
-
-void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
+void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
 		   sector_t sector, int size, struct page *page)
 {
 	/* write first size bytes of page to sector of rdev
@@ -440,78 +750,50 @@ void md_super_write(mddev_t *mddev, mdk_rdev_t *rdev,
 	 * and decrement it on completion, waking up sb_wait
 	 * if zero is reached.
 	 * If an error occurred, call md_error
-	 *
-	 * As we might need to resubmit the request if BIO_RW_BARRIER
-	 * causes ENOTSUPP, we allocate a spare bio...
 	 */
-	struct bio *bio = bio_alloc(GFP_NOIO, 1);
-	int rw = (1<<BIO_RW) | (1<<BIO_RW_SYNC);
+	struct bio *bio = bio_alloc_mddev(GFP_NOIO, 1, mddev);
 
-	bio->bi_bdev = rdev->bdev;
-	bio->bi_sector = sector;
+	bio->bi_bdev = rdev->meta_bdev ? rdev->meta_bdev : rdev->bdev;
+	bio->bi_iter.bi_sector = sector;
 	bio_add_page(bio, page, size, 0);
 	bio->bi_private = rdev;
 	bio->bi_end_io = super_written;
-	bio->bi_rw = rw;
 
 	atomic_inc(&mddev->pending_writes);
-	if (!test_bit(BarriersNotsupp, &rdev->flags)) {
-		struct bio *rbio;
-		rw |= (1<<BIO_RW_BARRIER);
-		rbio = bio_clone(bio, GFP_NOIO);
-		rbio->bi_private = bio;
-		rbio->bi_end_io = super_written_barrier;
-		submit_bio(rw, rbio);
-	} else
-		submit_bio(rw, bio);
+	submit_bio(WRITE_FLUSH_FUA, bio);
 }
 
-void md_super_wait(mddev_t *mddev)
+void md_super_wait(struct mddev *mddev)
 {
-	/* wait for all superblock writes that were scheduled to complete.
-	 * if any had to be retried (due to BARRIER problems), retry them
-	 */
+	/* wait for all superblock writes that were scheduled to complete */
 	DEFINE_WAIT(wq);
 	for(;;) {
 		prepare_to_wait(&mddev->sb_wait, &wq, TASK_UNINTERRUPTIBLE);
 		if (atomic_read(&mddev->pending_writes)==0)
 			break;
-		while (mddev->biolist) {
-			struct bio *bio;
-			spin_lock_irq(&mddev->write_lock);
-			bio = mddev->biolist;
-			mddev->biolist = bio->bi_next ;
-			bio->bi_next = NULL;
-			spin_unlock_irq(&mddev->write_lock);
-			submit_bio(bio->bi_rw, bio);
-		}
 		schedule();
 	}
 	finish_wait(&mddev->sb_wait, &wq);
 }
 
-static void bi_complete(struct bio *bio, int error)
-{
-	complete((struct completion*)bio->bi_private);
-}
-
-int sync_page_io(struct block_device *bdev, sector_t sector, int size,
-		   struct page *page, int rw)
+int sync_page_io(struct md_rdev *rdev, sector_t sector, int size,
+		 struct page *page, int rw, bool metadata_op)
 {
-	struct bio *bio = bio_alloc(GFP_NOIO, 1);
-	struct completion event;
+	struct bio *bio = bio_alloc_mddev(GFP_NOIO, 1, rdev->mddev);
 	int ret;
 
-	rw |= (1 << BIO_RW_SYNC);
-
-	bio->bi_bdev = bdev;
-	bio->bi_sector = sector;
+	bio->bi_bdev = (metadata_op && rdev->meta_bdev) ?
+		rdev->meta_bdev : rdev->bdev;
+	if (metadata_op)
+		bio->bi_iter.bi_sector = sector + rdev->sb_start;
+	else if (rdev->mddev->reshape_position != MaxSector &&
+		 (rdev->mddev->reshape_backwards ==
+		  (sector >= rdev->mddev->reshape_position)))
+		bio->bi_iter.bi_sector = sector + rdev->new_data_offset;
+	else
+		bio->bi_iter.bi_sector = sector + rdev->data_offset;
 	bio_add_page(bio, page, size, 0);
-	init_completion(&event);
-	bio->bi_private = &event;
-	bio->bi_end_io = bi_complete;
-	submit_bio(rw, bio);
-	wait_for_completion(&event);
+	submit_bio_wait(rw, bio);
 
 	ret = test_bit(BIO_UPTODATE, &bio->bi_flags);
 	bio_put(bio);
@@ -519,7 +801,7 @@ int sync_page_io(struct block_device *bdev, sector_t sector, int size,
 }
 EXPORT_SYMBOL_GPL(sync_page_io);
 
-static int read_disk_sb(mdk_rdev_t * rdev, int size)
+static int read_disk_sb(struct md_rdev * rdev, int size)
 {
 	char b[BDEVNAME_SIZE];
 	if (!rdev->sb_page) {
@@ -530,7 +812,7 @@ static int read_disk_sb(mdk_rdev_t * rdev, int size)
 		return 0;
 
 
-	if (!sync_page_io(rdev->bdev, rdev->sb_start, size, rdev->sb_page, READ))
+	if (!sync_page_io(rdev, 0, size, rdev->sb_page, READ, true))
 		goto fail;
 	rdev->sb_loaded = 1;
 	return 0;
@@ -626,7 +908,7 @@ static unsigned int calc_sb_csum(mdp_super_t * sb)
  * We rely on user-space to write the initial superblock, and support
  * reading and updating of superblocks.
  * Interface methods are:
- *   int load_super(mdk_rdev_t *dev, mdk_rdev_t *refdev, int minor_version)
+ *   int load_super(struct md_rdev *dev, struct md_rdev *refdev, int minor_version)
  *      loads and validates a superblock on dev.
  *      if refdev != NULL, compare superblocks on both devices
  *    Return:
@@ -636,13 +918,13 @@ static unsigned int calc_sb_csum(mdp_super_t * sb)
  *     -EINVAL superblock incompatible or invalid
  *     -othererror e.g. -EIO
  *
- *   int validate_super(mddev_t *mddev, mdk_rdev_t *dev)
+ *   int validate_super(struct mddev *mddev, struct md_rdev *dev)
  *      Verify that dev is acceptable into mddev.
  *       The first time, mddev->raid_disks will be 0, and data from
  *       dev should be merged in.  Subsequent calls check that dev
  *       is new enough.  Return 0 or -EINVAL
  *
- *   void sync_super(mddev_t *mddev, mdk_rdev_t *dev)
+ *   void sync_super(struct mddev *mddev, struct md_rdev *dev)
  *     Update the superblock for rdev with data in mddev
  *     This does not write to disc.
  *
@@ -651,18 +933,41 @@ static unsigned int calc_sb_csum(mdp_super_t * sb)
 struct super_type  {
 	char		    *name;
 	struct module	    *owner;
-	int		    (*load_super)(mdk_rdev_t *rdev, mdk_rdev_t *refdev,
+	int		    (*load_super)(struct md_rdev *rdev,
+					  struct md_rdev *refdev,
 					  int minor_version);
-	int		    (*validate_super)(mddev_t *mddev, mdk_rdev_t *rdev);
-	void		    (*sync_super)(mddev_t *mddev, mdk_rdev_t *rdev);
-	unsigned long long  (*rdev_size_change)(mdk_rdev_t *rdev,
+	int		    (*validate_super)(struct mddev *mddev,
+					      struct md_rdev *rdev);
+	void		    (*sync_super)(struct mddev *mddev,
+					  struct md_rdev *rdev);
+	unsigned long long  (*rdev_size_change)(struct md_rdev *rdev,
 						sector_t num_sectors);
+	int		    (*allow_new_offset)(struct md_rdev *rdev,
+						unsigned long long new_offset);
 };
 
 /*
+ * Check that the given mddev has no bitmap.
+ *
+ * This function is called from the run method of all personalities that do not
+ * support bitmaps. It prints an error message and returns non-zero if mddev
+ * has a bitmap. Otherwise, it returns 0.
+ *
+ */
+int md_check_no_bitmap(struct mddev *mddev)
+{
+	if (!mddev->bitmap_info.file && !mddev->bitmap_info.offset)
+		return 0;
+	printk(KERN_ERR "%s: bitmaps are not supported for %s\n",
+		mdname(mddev), mddev->pers->name);
+	return 1;
+}
+EXPORT_SYMBOL(md_check_no_bitmap);
+
+/*
  * load_super for 0.90.0 
  */
-static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
+static int super_90_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor_version)
 {
 	char b[BDEVNAME_SIZE], b2[BDEVNAME_SIZE];
 	mdp_super_t *sb;
@@ -674,7 +979,7 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 	 *
 	 * It also happens to be a multiple of 4Kb.
 	 */
-	rdev->sb_start = calc_dev_sboffset(rdev->bdev);
+	rdev->sb_start = calc_dev_sboffset(rdev);
 
 	ret = read_disk_sb(rdev, MD_SB_BYTES);
 	if (ret) return ret;
@@ -682,7 +987,7 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 	ret = -EINVAL;
 
 	bdevname(rdev->bdev, b);
-	sb = (mdp_super_t*)page_address(rdev->sb_page);
+	sb = page_address(rdev->sb_page);
 
 	if (sb->md_magic != MD_SB_MAGIC) {
 		printk(KERN_ERR "md: invalid raid superblock magic on %s\n",
@@ -710,18 +1015,9 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 
 	rdev->preferred_minor = sb->md_minor;
 	rdev->data_offset = 0;
+	rdev->new_data_offset = 0;
 	rdev->sb_size = MD_SB_BYTES;
-
-	if (sb->state & (1<<MD_SB_BITMAP_PRESENT)) {
-		if (sb->level != 1 && sb->level != 4
-		    && sb->level != 5 && sb->level != 6
-		    && sb->level != 10) {
-			/* FIXME use a better test */
-			printk(KERN_WARNING
-			       "md: bitmaps not supported for this level.\n");
-			goto abort;
-		}
-	}
+	rdev->badblocks.shift = -1;
 
 	if (sb->level == LEVEL_MULTIPATH)
 		rdev->desc_nr = -1;
@@ -732,7 +1028,7 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 		ret = 1;
 	} else {
 		__u64 ev1, ev2;
-		mdp_super_t *refsb = (mdp_super_t*)page_address(refdev->sb_page);
+		mdp_super_t *refsb = page_address(refdev->sb_page);
 		if (!uuid_equal(refsb, sb)) {
 			printk(KERN_WARNING "md: %s has different UUID to %s\n",
 				b, bdevname(refdev->bdev,b2));
@@ -751,9 +1047,15 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 		else 
 			ret = 0;
 	}
-	rdev->size = calc_num_sectors(rdev, sb->chunk_size) / 2;
+	rdev->sectors = rdev->sb_start;
+	/* Limit to 4TB as metadata cannot record more than that.
+	 * (not needed for Linear and RAID0 as metadata doesn't
+	 * record this size)
+	 */
+	if (rdev->sectors >= (2ULL << 32) && sb->level >= 1)
+		rdev->sectors = (2ULL << 32) - 2;
 
-	if (rdev->size < sb->size && sb->level > 1)
+	if (rdev->sectors < ((sector_t)sb->size) * 2 && sb->level >= 1)
 		/* "this cannot possibly happen" ... */
 		ret = -EINVAL;
 
@@ -764,47 +1066,53 @@ static int super_90_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version
 /*
  * validate_super for 0.90.0
  */
-static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
+static int super_90_validate(struct mddev *mddev, struct md_rdev *rdev)
 {
 	mdp_disk_t *desc;
-	mdp_super_t *sb = (mdp_super_t *)page_address(rdev->sb_page);
+	mdp_super_t *sb = page_address(rdev->sb_page);
 	__u64 ev1 = md_event(sb);
 
 	rdev->raid_disk = -1;
 	clear_bit(Faulty, &rdev->flags);
 	clear_bit(In_sync, &rdev->flags);
+	clear_bit(Bitmap_sync, &rdev->flags);
 	clear_bit(WriteMostly, &rdev->flags);
-	clear_bit(BarriersNotsupp, &rdev->flags);
 
 	if (mddev->raid_disks == 0) {
 		mddev->major_version = 0;
 		mddev->minor_version = sb->minor_version;
 		mddev->patch_version = sb->patch_version;
 		mddev->external = 0;
-		mddev->chunk_size = sb->chunk_size;
+		mddev->chunk_sectors = sb->chunk_size >> 9;
 		mddev->ctime = sb->ctime;
 		mddev->utime = sb->utime;
 		mddev->level = sb->level;
 		mddev->clevel[0] = 0;
 		mddev->layout = sb->layout;
 		mddev->raid_disks = sb->raid_disks;
-		mddev->size = sb->size;
+		mddev->dev_sectors = ((sector_t)sb->size) * 2;
 		mddev->events = ev1;
-		mddev->bitmap_offset = 0;
-		mddev->default_bitmap_offset = MD_SB_BYTES >> 9;
+		mddev->bitmap_info.offset = 0;
+		mddev->bitmap_info.space = 0;
+		/* bitmap can use 60 K after the 4K superblocks */
+		mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
+		mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
+		mddev->reshape_backwards = 0;
 
 		if (mddev->minor_version >= 91) {
 			mddev->reshape_position = sb->reshape_position;
 			mddev->delta_disks = sb->delta_disks;
 			mddev->new_level = sb->new_level;
 			mddev->new_layout = sb->new_layout;
-			mddev->new_chunk = sb->new_chunk;
+			mddev->new_chunk_sectors = sb->new_chunk >> 9;
+			if (mddev->delta_disks < 0)
+				mddev->reshape_backwards = 1;
 		} else {
 			mddev->reshape_position = MaxSector;
 			mddev->delta_disks = 0;
 			mddev->new_level = mddev->level;
 			mddev->new_layout = mddev->layout;
-			mddev->new_chunk = mddev->chunk_size;
+			mddev->new_chunk_sectors = mddev->chunk_sectors;
 		}
 
 		if (sb->state & (1<<MD_SB_CLEAN))
@@ -825,20 +1133,29 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 		mddev->max_disks = MD_SB_DISKS;
 
 		if (sb->state & (1<<MD_SB_BITMAP_PRESENT) &&
-		    mddev->bitmap_file == NULL)
-			mddev->bitmap_offset = mddev->default_bitmap_offset;
+		    mddev->bitmap_info.file == NULL) {
+			mddev->bitmap_info.offset =
+				mddev->bitmap_info.default_offset;
+			mddev->bitmap_info.space =
+				mddev->bitmap_info.default_space;
+		}
 
 	} else if (mddev->pers == NULL) {
-		/* Insist on good event counter while assembling */
+		/* Insist on good event counter while assembling, except
+		 * for spares (which don't need an event count) */
 		++ev1;
-		if (ev1 < mddev->events) 
-			return -EINVAL;
+		if (sb->disks[rdev->desc_nr].state & (
+			    (1<<MD_DISK_SYNC) | (1 << MD_DISK_ACTIVE)))
+			if (ev1 < mddev->events) 
+				return -EINVAL;
 	} else if (mddev->bitmap) {
 		/* if adding to array with a bitmap, then we can accept an
 		 * older device ... but not too old.
 		 */
 		if (ev1 < mddev->bitmap->events_cleared)
 			return 0;
+		if (ev1 < mddev->events)
+			set_bit(Bitmap_sync, &rdev->flags);
 	} else {
 		if (ev1 < mddev->events)
 			/* just a hot-add of a new device, leave raid_disk at -1 */
@@ -854,6 +1171,15 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 			    desc->raid_disk < mddev->raid_disks */) {
 			set_bit(In_sync, &rdev->flags);
 			rdev->raid_disk = desc->raid_disk;
+			rdev->saved_raid_disk = desc->raid_disk;
+		} else if (desc->state & (1<<MD_DISK_ACTIVE)) {
+			/* active but not in sync implies recovery up to
+			 * reshape position.  We don't know exactly where
+			 * that is, so set to zero for now */
+			if (mddev->minor_version >= 91) {
+				rdev->recovery_offset = 0;
+				rdev->raid_disk = desc->raid_disk;
+			}
 		}
 		if (desc->state & (1<<MD_DISK_WRITEMOSTLY))
 			set_bit(WriteMostly, &rdev->flags);
@@ -865,11 +1191,10 @@ static int super_90_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 /*
  * sync_super for 0.90.0
  */
-static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
+static void super_90_sync(struct mddev *mddev, struct md_rdev *rdev)
 {
 	mdp_super_t *sb;
-	struct list_head *tmp;
-	mdk_rdev_t *rdev2;
+	struct md_rdev *rdev2;
 	int next_spare = mddev->raid_disks;
 
 
@@ -888,7 +1213,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 
 	rdev->sb_size = MD_SB_BYTES;
 
-	sb = (mdp_super_t*)page_address(rdev->sb_page);
+	sb = page_address(rdev->sb_page);
 
 	memset(sb, 0, sizeof(*sb));
 
@@ -903,7 +1228,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 
 	sb->ctime = mddev->ctime;
 	sb->level = mddev->level;
-	sb->size  = mddev->size;
+	sb->size = mddev->dev_sectors / 2;
 	sb->raid_disks = mddev->raid_disks;
 	sb->md_minor = mddev->md_minor;
 	sb->not_persistent = 0;
@@ -920,7 +1245,7 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 		sb->new_level = mddev->new_level;
 		sb->delta_disks = mddev->delta_disks;
 		sb->new_layout = mddev->new_layout;
-		sb->new_chunk = mddev->new_chunk;
+		sb->new_chunk = mddev->new_chunk_sectors << 9;
 	}
 	mddev->minor_version = sb->minor_version;
 	if (mddev->in_sync)
@@ -934,17 +1259,28 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 		sb->recovery_cp = 0;
 
 	sb->layout = mddev->layout;
-	sb->chunk_size = mddev->chunk_size;
+	sb->chunk_size = mddev->chunk_sectors << 9;
 
-	if (mddev->bitmap && mddev->bitmap_file == NULL)
+	if (mddev->bitmap && mddev->bitmap_info.file == NULL)
 		sb->state |= (1<<MD_SB_BITMAP_PRESENT);
 
 	sb->disks[0].state = (1<<MD_DISK_REMOVED);
-	rdev_for_each(rdev2, tmp, mddev) {
+	rdev_for_each(rdev2, mddev) {
 		mdp_disk_t *d;
 		int desc_nr;
-		if (rdev2->raid_disk >= 0 && test_bit(In_sync, &rdev2->flags)
-		    && !test_bit(Faulty, &rdev2->flags))
+		int is_active = test_bit(In_sync, &rdev2->flags);
+
+		if (rdev2->raid_disk >= 0 &&
+		    sb->minor_version >= 91)
+			/* we have nowhere to store the recovery_offset,
+			 * but if it is not below the reshape_position,
+			 * we can piggy-back on that.
+			 */
+			is_active = 1;
+		if (rdev2->raid_disk < 0 ||
+		    test_bit(Faulty, &rdev2->flags))
+			is_active = 0;
+		if (is_active)
 			desc_nr = rdev2->raid_disk;
 		else
 			desc_nr = next_spare++;
@@ -954,16 +1290,16 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 		d->number = rdev2->desc_nr;
 		d->major = MAJOR(rdev2->bdev->bd_dev);
 		d->minor = MINOR(rdev2->bdev->bd_dev);
-		if (rdev2->raid_disk >= 0 && test_bit(In_sync, &rdev2->flags)
-		    && !test_bit(Faulty, &rdev2->flags))
+		if (is_active)
 			d->raid_disk = rdev2->raid_disk;
 		else
 			d->raid_disk = rdev2->desc_nr; /* compatibility */
 		if (test_bit(Faulty, &rdev2->flags))
 			d->state = (1<<MD_DISK_FAULTY);
-		else if (test_bit(In_sync, &rdev2->flags)) {
+		else if (is_active) {
 			d->state = (1<<MD_DISK_ACTIVE);
-			d->state |= (1<<MD_DISK_SYNC);
+			if (test_bit(In_sync, &rdev2->flags))
+				d->state |= (1<<MD_DISK_SYNC);
 			active++;
 			working++;
 		} else {
@@ -999,21 +1335,32 @@ static void super_90_sync(mddev_t *mddev, mdk_rdev_t *rdev)
  * rdev_size_change for 0.90.0
  */
 static unsigned long long
-super_90_rdev_size_change(mdk_rdev_t *rdev, sector_t num_sectors)
+super_90_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
 {
-	if (num_sectors && num_sectors < rdev->mddev->size * 2)
+	if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
 		return 0; /* component must fit device */
-	if (rdev->mddev->bitmap_offset)
+	if (rdev->mddev->bitmap_info.offset)
 		return 0; /* can't move bitmap */
-	rdev->sb_start = calc_dev_sboffset(rdev->bdev);
+	rdev->sb_start = calc_dev_sboffset(rdev);
 	if (!num_sectors || num_sectors > rdev->sb_start)
 		num_sectors = rdev->sb_start;
+	/* Limit to 4TB as metadata cannot record more than that.
+	 * 4TB == 2^32 KB, or 2*2^32 sectors.
+	 */
+	if (num_sectors >= (2ULL << 32) && rdev->mddev->level >= 1)
+		num_sectors = (2ULL << 32) - 2;
 	md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
 		       rdev->sb_page);
 	md_super_wait(rdev->mddev);
-	return num_sectors / 2; /* kB for sysfs */
+	return num_sectors;
 }
 
+static int
+super_90_allow_new_offset(struct md_rdev *rdev, unsigned long long new_offset)
+{
+	/* non-zero offset changes not possible with v0.90 */
+	return new_offset == 0;
+}
 
 /*
  * version 1 superblock
@@ -1026,12 +1373,11 @@ static __le32 calc_sb_1_csum(struct mdp_superblock_1 * sb)
 	unsigned long long newcsum;
 	int size = 256 + le32_to_cpu(sb->max_dev)*2;
 	__le32 *isuper = (__le32*)sb;
-	int i;
 
 	disk_csum = sb->sb_csum;
 	sb->sb_csum = 0;
 	newcsum = 0;
-	for (i=0; size>=4; size -= 4 )
+	for (; size >= 4; size -= 4)
 		newcsum += le32_to_cpu(*isuper++);
 
 	if (size == 2)
@@ -1042,11 +1388,14 @@ static __le32 calc_sb_1_csum(struct mdp_superblock_1 * sb)
 	return cpu_to_le32(csum);
 }
 
-static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
+static int md_set_badblocks(struct badblocks *bb, sector_t s, int sectors,
+			    int acknowledged);
+static int super_1_load(struct md_rdev *rdev, struct md_rdev *refdev, int minor_version)
 {
 	struct mdp_superblock_1 *sb;
 	int ret;
 	sector_t sb_start;
+	sector_t sectors;
 	char b[BDEVNAME_SIZE], b2[BDEVNAME_SIZE];
 	int bmask;
 
@@ -1060,7 +1409,7 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
 	 */
 	switch(minor_version) {
 	case 0:
-		sb_start = rdev->bdev->bd_inode->i_size >> 9;
+		sb_start = i_size_read(rdev->bdev->bd_inode) >> 9;
 		sb_start -= 8*2;
 		sb_start &= ~(sector_t)(4*2-1);
 		break;
@@ -1082,7 +1431,7 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
 	if (ret) return ret;
 
 
-	sb = (struct mdp_superblock_1*)page_address(rdev->sb_page);
+	sb = page_address(rdev->sb_page);
 
 	if (sb->magic != cpu_to_le32(MD_SB_MAGIC) ||
 	    sb->major_version != cpu_to_le32(1) ||
@@ -1101,42 +1450,83 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
 		       bdevname(rdev->bdev,b));
 		return -EINVAL;
 	}
-	if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET)) {
-		if (sb->level != cpu_to_le32(1) &&
-		    sb->level != cpu_to_le32(4) &&
-		    sb->level != cpu_to_le32(5) &&
-		    sb->level != cpu_to_le32(6) &&
-		    sb->level != cpu_to_le32(10)) {
-			printk(KERN_WARNING
-			       "md: bitmaps not supported for this level.\n");
-			return -EINVAL;
-		}
-	}
+	if (sb->pad0 ||
+	    sb->pad3[0] ||
+	    memcmp(sb->pad3, sb->pad3+1, sizeof(sb->pad3) - sizeof(sb->pad3[1])))
+		/* Some padding is non-zero, might be a new feature */
+		return -EINVAL;
 
 	rdev->preferred_minor = 0xffff;
 	rdev->data_offset = le64_to_cpu(sb->data_offset);
+	rdev->new_data_offset = rdev->data_offset;
+	if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE) &&
+	    (le32_to_cpu(sb->feature_map) & MD_FEATURE_NEW_OFFSET))
+		rdev->new_data_offset += (s32)le32_to_cpu(sb->new_offset);
 	atomic_set(&rdev->corrected_errors, le32_to_cpu(sb->cnt_corrected_read));
 
 	rdev->sb_size = le32_to_cpu(sb->max_dev) * 2 + 256;
-	bmask = queue_hardsect_size(rdev->bdev->bd_disk->queue)-1;
+	bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
 	if (rdev->sb_size & bmask)
 		rdev->sb_size = (rdev->sb_size | bmask) + 1;
 
 	if (minor_version
 	    && rdev->data_offset < sb_start + (rdev->sb_size/512))
 		return -EINVAL;
+	if (minor_version
+	    && rdev->new_data_offset < sb_start + (rdev->sb_size/512))
+		return -EINVAL;
 
 	if (sb->level == cpu_to_le32(LEVEL_MULTIPATH))
 		rdev->desc_nr = -1;
 	else
 		rdev->desc_nr = le32_to_cpu(sb->dev_number);
 
+	if (!rdev->bb_page) {
+		rdev->bb_page = alloc_page(GFP_KERNEL);
+		if (!rdev->bb_page)
+			return -ENOMEM;
+	}
+	if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BAD_BLOCKS) &&
+	    rdev->badblocks.count == 0) {
+		/* need to load the bad block list.
+		 * Currently we limit it to one page.
+		 */
+		s32 offset;
+		sector_t bb_sector;
+		u64 *bbp;
+		int i;
+		int sectors = le16_to_cpu(sb->bblog_size);
+		if (sectors > (PAGE_SIZE / 512))
+			return -EINVAL;
+		offset = le32_to_cpu(sb->bblog_offset);
+		if (offset == 0)
+			return -EINVAL;
+		bb_sector = (long long)offset;
+		if (!sync_page_io(rdev, bb_sector, sectors << 9,
+				  rdev->bb_page, READ, true))
+			return -EIO;
+		bbp = (u64 *)page_address(rdev->bb_page);
+		rdev->badblocks.shift = sb->bblog_shift;
+		for (i = 0 ; i < (sectors << (9-3)) ; i++, bbp++) {
+			u64 bb = le64_to_cpu(*bbp);
+			int count = bb & (0x3ff);
+			u64 sector = bb >> 10;
+			sector <<= sb->bblog_shift;
+			count <<= sb->bblog_shift;
+			if (bb + 1 == 0)
+				break;
+			if (md_set_badblocks(&rdev->badblocks,
+					     sector, count, 1) == 0)
+				return -EINVAL;
+		}
+	} else if (sb->bblog_offset != 0)
+		rdev->badblocks.shift = 0;
+
 	if (!refdev) {
 		ret = 1;
 	} else {
 		__u64 ev1, ev2;
-		struct mdp_superblock_1 *refsb = 
-			(struct mdp_superblock_1*)page_address(refdev->sb_page);
+		struct mdp_superblock_1 *refsb = page_address(refdev->sb_page);
 
 		if (memcmp(sb->set_uuid, refsb->set_uuid, 16) != 0 ||
 		    sb->level != refsb->level ||
@@ -1156,82 +1546,110 @@ static int super_1_load(mdk_rdev_t *rdev, mdk_rdev_t *refdev, int minor_version)
 		else
 			ret = 0;
 	}
-	if (minor_version)
-		rdev->size = ((rdev->bdev->bd_inode->i_size>>9) - le64_to_cpu(sb->data_offset)) / 2;
-	else
-		rdev->size = rdev->sb_start / 2;
-	if (rdev->size < le64_to_cpu(sb->data_size)/2)
-		return -EINVAL;
-	rdev->size = le64_to_cpu(sb->data_size)/2;
-	if (le32_to_cpu(sb->chunksize))
-		rdev->size &= ~((sector_t)le32_to_cpu(sb->chunksize)/2 - 1);
-
-	if (le64_to_cpu(sb->size) > rdev->size*2)
+	if (minor_version) {
+		sectors = (i_size_read(rdev->bdev->bd_inode) >> 9);
+		sectors -= rdev->data_offset;
+	} else
+		sectors = rdev->sb_start;
+	if (sectors < le64_to_cpu(sb->data_size))
 		return -EINVAL;
+	rdev->sectors = le64_to_cpu(sb->data_size);
 	return ret;
 }
 
-static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
+static int super_1_validate(struct mddev *mddev, struct md_rdev *rdev)
 {
-	struct mdp_superblock_1 *sb = (struct mdp_superblock_1*)page_address(rdev->sb_page);
+	struct mdp_superblock_1 *sb = page_address(rdev->sb_page);
 	__u64 ev1 = le64_to_cpu(sb->events);
 
 	rdev->raid_disk = -1;
 	clear_bit(Faulty, &rdev->flags);
 	clear_bit(In_sync, &rdev->flags);
+	clear_bit(Bitmap_sync, &rdev->flags);
 	clear_bit(WriteMostly, &rdev->flags);
-	clear_bit(BarriersNotsupp, &rdev->flags);
 
 	if (mddev->raid_disks == 0) {
 		mddev->major_version = 1;
 		mddev->patch_version = 0;
 		mddev->external = 0;
-		mddev->chunk_size = le32_to_cpu(sb->chunksize) << 9;
+		mddev->chunk_sectors = le32_to_cpu(sb->chunksize);
 		mddev->ctime = le64_to_cpu(sb->ctime) & ((1ULL << 32)-1);
 		mddev->utime = le64_to_cpu(sb->utime) & ((1ULL << 32)-1);
 		mddev->level = le32_to_cpu(sb->level);
 		mddev->clevel[0] = 0;
 		mddev->layout = le32_to_cpu(sb->layout);
 		mddev->raid_disks = le32_to_cpu(sb->raid_disks);
-		mddev->size = le64_to_cpu(sb->size)/2;
+		mddev->dev_sectors = le64_to_cpu(sb->size);
 		mddev->events = ev1;
-		mddev->bitmap_offset = 0;
-		mddev->default_bitmap_offset = 1024 >> 9;
-		
+		mddev->bitmap_info.offset = 0;
+		mddev->bitmap_info.space = 0;
+		/* Default location for bitmap is 1K after superblock
+		 * using 3K - total of 4K
+		 */
+		mddev->bitmap_info.default_offset = 1024 >> 9;
+		mddev->bitmap_info.default_space = (4096-1024) >> 9;
+		mddev->reshape_backwards = 0;
+
 		mddev->recovery_cp = le64_to_cpu(sb->resync_offset);
 		memcpy(mddev->uuid, sb->set_uuid, 16);
 
 		mddev->max_disks =  (4096-256)/2;
 
 		if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_BITMAP_OFFSET) &&
-		    mddev->bitmap_file == NULL )
-			mddev->bitmap_offset = (__s32)le32_to_cpu(sb->bitmap_offset);
+		    mddev->bitmap_info.file == NULL) {
+			mddev->bitmap_info.offset =
+				(__s32)le32_to_cpu(sb->bitmap_offset);
+			/* Metadata doesn't record how much space is available.
+			 * For 1.0, we assume we can use up to the superblock
+			 * if before, else to 4K beyond superblock.
+			 * For others, assume no change is possible.
+			 */
+			if (mddev->minor_version > 0)
+				mddev->bitmap_info.space = 0;
+			else if (mddev->bitmap_info.offset > 0)
+				mddev->bitmap_info.space =
+					8 - mddev->bitmap_info.offset;
+			else
+				mddev->bitmap_info.space =
+					-mddev->bitmap_info.offset;
+		}
 
 		if ((le32_to_cpu(sb->feature_map) & MD_FEATURE_RESHAPE_ACTIVE)) {
 			mddev->reshape_position = le64_to_cpu(sb->reshape_position);
 			mddev->delta_disks = le32_to_cpu(sb->delta_disks);
 			mddev->new_level = le32_to_cpu(sb->new_level);
 			mddev->new_layout = le32_to_cpu(sb->new_layout);
-			mddev->new_chunk = le32_to_cpu(sb->new_chunk)<<9;
+			mddev->new_chunk_sectors = le32_to_cpu(sb->new_chunk);
+			if (mddev->delta_disks < 0 ||
+			    (mddev->delta_disks == 0 &&
+			     (le32_to_cpu(sb->feature_map)
+			      & MD_FEATURE_RESHAPE_BACKWARDS)))
+				mddev->reshape_backwards = 1;
 		} else {
 			mddev->reshape_position = MaxSector;
 			mddev->delta_disks = 0;
 			mddev->new_level = mddev->level;
 			mddev->new_layout = mddev->layout;
-			mddev->new_chunk = mddev->chunk_size;
+			mddev->new_chunk_sectors = mddev->chunk_sectors;
 		}
 
 	} else if (mddev->pers == NULL) {
-		/* Insist of good event counter while assembling */
+		/* Insist of good event counter while assembling, except for
+		 * spares (which don't need an event count) */
 		++ev1;
-		if (ev1 < mddev->events)
-			return -EINVAL;
+		if (rdev->desc_nr >= 0 &&
+		    rdev->desc_nr < le32_to_cpu(sb->max_dev) &&
+		    le16_to_cpu(sb->dev_roles[rdev->desc_nr]) < 0xfffe)
+			if (ev1 < mddev->events)
+				return -EINVAL;
 	} else if (mddev->bitmap) {
 		/* If adding to array with a bitmap, then we can accept an
 		 * older device, but not too old.
 		 */
 		if (ev1 < mddev->bitmap->events_cleared)
 			return 0;
+		if (ev1 < mddev->events)
+			set_bit(Bitmap_sync, &rdev->flags);
 	} else {
 		if (ev1 < mddev->events)
 			/* just a hot-add of a new device, leave raid_disk at -1 */
@@ -1239,7 +1657,12 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 	}
 	if (mddev->level != LEVEL_MULTIPATH) {
 		int role;
-		role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
+		if (rdev->desc_nr < 0 ||
+		    rdev->desc_nr >= le32_to_cpu(sb->max_dev)) {
+			role = 0xffff;
+			rdev->desc_nr = -1;
+		} else
+			role = le16_to_cpu(sb->dev_roles[rdev->desc_nr]);
 		switch(role) {
 		case 0xffff: /* spare */
 			break;
@@ -1247,37 +1670,40 @@ static int super_1_validate(mddev_t *mddev, mdk_rdev_t *rdev)
 			set_bit(Faulty, &rdev->flags);
 			break;
 		default:
+			rdev->saved_raid_disk = role;
 			if ((le32_to_cpu(sb->feature_map) &
-			     MD_FEATURE_RECOVERY_OFFSET))
+			     MD_FEATURE_RECOVERY_OFFSET)) {
 				rdev->recovery_offset = le64_to_cpu(sb->recovery_offset);
-			else
+				if (!(le32_to_cpu(sb->feature_map) &
+				      MD_FEATURE_RECOVERY_BITMAP))
+					rdev->saved_raid_disk = -1;
+			} else
 				set_bit(In_sync, &rdev->flags);
 			rdev->raid_disk = role;
 			break;
 		}
 		if (sb->devflags & WriteMostly1)
 			set_bit(WriteMostly, &rdev->flags);
+		if (le32_to_cpu(sb->feature_map) & MD_FEATURE_REPLACEMENT)
+			set_bit(Replacement, &rdev->flags);
 	} else /* MULTIPATH are always insync */
 		set_bit(In_sync, &rdev->flags);
 
 	return 0;
 }
 
-static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
+static void super_1_sync(struct mddev *mddev, struct md_rdev *rdev)
 {
 	struct mdp_superblock_1 *sb;
-	struct list_head *tmp;
-	mdk_rdev_t *rdev2;
+	struct md_rdev *rdev2;
 	int max_dev, i;
 	/* make rdev->sb match mddev and rdev data. */
 
-	sb = (struct mdp_superblock_1*)page_address(rdev->sb_page);
+	sb = page_address(rdev->sb_page);
 
 	sb->feature_map = 0;
 	sb->pad0 = 0;
 	sb->recovery_offset = cpu_to_le64(0);
-	memset(sb->pad1, 0, sizeof(sb->pad1));
-	memset(sb->pad2, 0, sizeof(sb->pad2));
 	memset(sb->pad3, 0, sizeof(sb->pad3));
 
 	sb->utime = cpu_to_le64((__u64)mddev->utime);
@@ -1290,19 +1716,36 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 	sb->cnt_corrected_read = cpu_to_le32(atomic_read(&rdev->corrected_errors));
 
 	sb->raid_disks = cpu_to_le32(mddev->raid_disks);
-	sb->size = cpu_to_le64(mddev->size<<1);
+	sb->size = cpu_to_le64(mddev->dev_sectors);
+	sb->chunksize = cpu_to_le32(mddev->chunk_sectors);
+	sb->level = cpu_to_le32(mddev->level);
+	sb->layout = cpu_to_le32(mddev->layout);
+
+	if (test_bit(WriteMostly, &rdev->flags))
+		sb->devflags |= WriteMostly1;
+	else
+		sb->devflags &= ~WriteMostly1;
+	sb->data_offset = cpu_to_le64(rdev->data_offset);
+	sb->data_size = cpu_to_le64(rdev->sectors);
 
-	if (mddev->bitmap && mddev->bitmap_file == NULL) {
-		sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_offset);
+	if (mddev->bitmap && mddev->bitmap_info.file == NULL) {
+		sb->bitmap_offset = cpu_to_le32((__u32)mddev->bitmap_info.offset);
 		sb->feature_map = cpu_to_le32(MD_FEATURE_BITMAP_OFFSET);
 	}
 
 	if (rdev->raid_disk >= 0 &&
-	    !test_bit(In_sync, &rdev->flags) &&
-	    rdev->recovery_offset > 0) {
-		sb->feature_map |= cpu_to_le32(MD_FEATURE_RECOVERY_OFFSET);
-		sb->recovery_offset = cpu_to_le64(rdev->recovery_offset);
-	}
+	    !test_bit(In_sync, &rdev->flags)) {
+		sb->feature_map |=
+			cpu_to_le32(MD_FEATURE_RECOVERY_OFFSET);
+		sb->recovery_offset =
+			cpu_to_le64(rdev->recovery_offset);
+		if (rdev->saved_raid_disk >= 0 && mddev->bitmap)
+			sb->feature_map |=
+				cpu_to_le32(MD_FEATURE_RECOVERY_BITMAP);
+	}
+	if (test_bit(Replacement, &rdev->flags))
+		sb->feature_map |=
+			cpu_to_le32(MD_FEATURE_REPLACEMENT);
 
 	if (mddev->reshape_position != MaxSector) {
 		sb->feature_map |= cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE);
@@ -1310,26 +1753,78 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 		sb->new_layout = cpu_to_le32(mddev->new_layout);
 		sb->delta_disks = cpu_to_le32(mddev->delta_disks);
 		sb->new_level = cpu_to_le32(mddev->new_level);
-		sb->new_chunk = cpu_to_le32(mddev->new_chunk>>9);
+		sb->new_chunk = cpu_to_le32(mddev->new_chunk_sectors);
+		if (mddev->delta_disks == 0 &&
+		    mddev->reshape_backwards)
+			sb->feature_map
+				|= cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS);
+		if (rdev->new_data_offset != rdev->data_offset) {
+			sb->feature_map
+				|= cpu_to_le32(MD_FEATURE_NEW_OFFSET);
+			sb->new_offset = cpu_to_le32((__u32)(rdev->new_data_offset
+							     - rdev->data_offset));
+		}
+	}
+
+	if (rdev->badblocks.count == 0)
+		/* Nothing to do for bad blocks*/ ;
+	else if (sb->bblog_offset == 0)
+		/* Cannot record bad blocks on this device */
+		md_error(mddev, rdev);
+	else {
+		struct badblocks *bb = &rdev->badblocks;
+		u64 *bbp = (u64 *)page_address(rdev->bb_page);
+		u64 *p = bb->page;
+		sb->feature_map |= cpu_to_le32(MD_FEATURE_BAD_BLOCKS);
+		if (bb->changed) {
+			unsigned seq;
+
+retry:
+			seq = read_seqbegin(&bb->lock);
+
+			memset(bbp, 0xff, PAGE_SIZE);
+
+			for (i = 0 ; i < bb->count ; i++) {
+				u64 internal_bb = p[i];
+				u64 store_bb = ((BB_OFFSET(internal_bb) << 10)
+						| BB_LEN(internal_bb));
+				bbp[i] = cpu_to_le64(store_bb);
+			}
+			bb->changed = 0;
+			if (read_seqretry(&bb->lock, seq))
+				goto retry;
+
+			bb->sector = (rdev->sb_start +
+				      (int)le32_to_cpu(sb->bblog_offset));
+			bb->size = le16_to_cpu(sb->bblog_size);
+		}
 	}
 
 	max_dev = 0;
-	rdev_for_each(rdev2, tmp, mddev)
+	rdev_for_each(rdev2, mddev)
 		if (rdev2->desc_nr+1 > max_dev)
 			max_dev = rdev2->desc_nr+1;
 
-	if (max_dev > le32_to_cpu(sb->max_dev))
+	if (max_dev > le32_to_cpu(sb->max_dev)) {
+		int bmask;
 		sb->max_dev = cpu_to_le32(max_dev);
+		rdev->sb_size = max_dev * 2 + 256;
+		bmask = queue_logical_block_size(rdev->bdev->bd_disk->queue)-1;
+		if (rdev->sb_size & bmask)
+			rdev->sb_size = (rdev->sb_size | bmask) + 1;
+	} else
+		max_dev = le32_to_cpu(sb->max_dev);
+
 	for (i=0; i<max_dev;i++)
 		sb->dev_roles[i] = cpu_to_le16(0xfffe);
 	
-	rdev_for_each(rdev2, tmp, mddev) {
+	rdev_for_each(rdev2, mddev) {
 		i = rdev2->desc_nr;
 		if (test_bit(Faulty, &rdev2->flags))
 			sb->dev_roles[i] = cpu_to_le16(0xfffe);
 		else if (test_bit(In_sync, &rdev2->flags))
 			sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
-		else if (rdev2->raid_disk >= 0 && rdev2->recovery_offset > 0)
+		else if (rdev2->raid_disk >= 0)
 			sb->dev_roles[i] = cpu_to_le16(rdev2->raid_disk);
 		else
 			sb->dev_roles[i] = cpu_to_le16(0xffff);
@@ -1339,39 +1834,75 @@ static void super_1_sync(mddev_t *mddev, mdk_rdev_t *rdev)
 }
 
 static unsigned long long
-super_1_rdev_size_change(mdk_rdev_t *rdev, sector_t num_sectors)
+super_1_rdev_size_change(struct md_rdev *rdev, sector_t num_sectors)
 {
 	struct mdp_superblock_1 *sb;
 	sector_t max_sectors;
-	if (num_sectors && num_sectors < rdev->mddev->size * 2)
+	if (num_sectors && num_sectors < rdev->mddev->dev_sectors)
 		return 0; /* component must fit device */
+	if (rdev->data_offset != rdev->new_data_offset)
+		return 0; /* too confusing */
 	if (rdev->sb_start < rdev->data_offset) {
 		/* minor versions 1 and 2; superblock before data */
-		max_sectors = rdev->bdev->bd_inode->i_size >> 9;
+		max_sectors = i_size_read(rdev->bdev->bd_inode) >> 9;
 		max_sectors -= rdev->data_offset;
 		if (!num_sectors || num_sectors > max_sectors)
 			num_sectors = max_sectors;
-	} else if (rdev->mddev->bitmap_offset) {
+	} else if (rdev->mddev->bitmap_info.offset) {
 		/* minor version 0 with bitmap we can't move */
 		return 0;
 	} else {
 		/* minor version 0; superblock after data */
 		sector_t sb_start;
-		sb_start = (rdev->bdev->bd_inode->i_size >> 9) - 8*2;
+		sb_start = (i_size_read(rdev->bdev->bd_inode) >> 9) - 8*2;
 		sb_start &= ~(sector_t)(4*2 - 1);
-		max_sectors = rdev->size * 2 + sb_start - rdev->sb_start;
+		max_sectors = rdev->sectors + sb_start - rdev->sb_start;
 		if (!num_sectors || num_sectors > max_sectors)
 			num_sectors = max_sectors;
 		rdev->sb_start = sb_start;
 	}
-	sb = (struct mdp_superblock_1 *) page_address(rdev->sb_page);
+	sb = page_address(rdev->sb_page);
 	sb->data_size = cpu_to_le64(num_sectors);
 	sb->super_offset = rdev->sb_start;
 	sb->sb_csum = calc_sb_1_csum(sb);
 	md_super_write(rdev->mddev, rdev, rdev->sb_start, rdev->sb_size,
 		       rdev->sb_page);
 	md_super_wait(rdev->mddev);
-	return num_sectors / 2; /* kB for sysfs */
+	return num_sectors;
+
+}
+
+static int
+super_1_allow_new_offset(struct md_rdev *rdev,
+			 unsigned long long new_offset)
+{
+	/* All necessary checks on new >= old have been done */
+	struct bitmap *bitmap;
+	if (new_offset >= rdev->data_offset)
+		return 1;
+
+	/* with 1.0 metadata, there is no metadata to tread on
+	 * so we can always move back */
+	if (rdev->mddev->minor_version == 0)
+		return 1;
+
+	/* otherwise we must be sure not to step on
+	 * any metadata, so stay:
+	 * 36K beyond start of superblock
+	 * beyond end of badblocks
+	 * beyond write-intent bitmap
+	 */
+	if (rdev->sb_start + (32+4)*2 > new_offset)
+		return 0;
+	bitmap = rdev->mddev->bitmap;
+	if (bitmap && !rdev->mddev->bitmap_info.file &&
+	    rdev->sb_start + rdev->mddev->bitmap_info.offset +
+	    bitmap->storage.file_pages * (PAGE_SIZE>>9) > new_offset)
+		return 0;
+	if (rdev->badblocks.sector + rdev->badblocks.size > new_offset)
+		return 0;
+
+	return 1;
 }
 
 static struct super_type super_types[] = {
@@ -1382,6 +1913,7 @@ static struct super_type super_types[] = {
 		.validate_super	    = super_90_validate,
 		.sync_super	    = super_90_sync,
 		.rdev_size_change   = super_90_rdev_size_change,
+		.allow_new_offset   = super_90_allow_new_offset,
 	},
 	[1] = {
 		.name	= "md-1",
@@ -1390,12 +1922,25 @@ static struct super_type super_types[] = {
 		.validate_super	    = super_1_validate,
 		.sync_super	    = super_1_sync,
 		.rdev_size_change   = super_1_rdev_size_change,
+		.allow_new_offset   = super_1_allow_new_offset,
 	},
 };
 
-static int match_mddev_units(mddev_t *mddev1, mddev_t *mddev2)
+static void sync_super(struct mddev *mddev, struct md_rdev *rdev)
 {
-	mdk_rdev_t *rdev, *rdev2;
+	if (mddev->sync_super) {
+		mddev->sync_super(mddev, rdev);
+		return;
+	}
+
+	BUG_ON(mddev->major_version >= ARRAY_SIZE(super_types));
+
+	super_types[mddev->major_version].sync_super(mddev, rdev);
+}
+
+static int match_mddev_units(struct mddev *mddev1, struct mddev *mddev2)
+{
+	struct md_rdev *rdev, *rdev2;
 
 	rcu_read_lock();
 	rdev_for_each_rcu(rdev, mddev1)
@@ -1411,7 +1956,84 @@ static int match_mddev_units(mddev_t *mddev1, mddev_t *mddev2)
 
 static LIST_HEAD(pending_raid_disks);
 
-static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
+/*
+ * Try to register data integrity profile for an mddev
+ *
+ * This is called when an array is started and after a disk has been kicked
+ * from the array. It only succeeds if all working and active component devices
+ * are integrity capable with matching profiles.
+ */
+int md_integrity_register(struct mddev *mddev)
+{
+	struct md_rdev *rdev, *reference = NULL;
+
+	if (list_empty(&mddev->disks))
+		return 0; /* nothing to do */
+	if (!mddev->gendisk || blk_get_integrity(mddev->gendisk))
+		return 0; /* shouldn't register, or already is */
+	rdev_for_each(rdev, mddev) {
+		/* skip spares and non-functional disks */
+		if (test_bit(Faulty, &rdev->flags))
+			continue;
+		if (rdev->raid_disk < 0)
+			continue;
+		if (!reference) {
+			/* Use the first rdev as the reference */
+			reference = rdev;
+			continue;
+		}
+		/* does this rdev's profile match the reference profile? */
+		if (blk_integrity_compare(reference->bdev->bd_disk,
+				rdev->bdev->bd_disk) < 0)
+			return -EINVAL;
+	}
+	if (!reference || !bdev_get_integrity(reference->bdev))
+		return 0;
+	/*
+	 * All component devices are integrity capable and have matching
+	 * profiles, register the common profile for the md device.
+	 */
+	if (blk_integrity_register(mddev->gendisk,
+			bdev_get_integrity(reference->bdev)) != 0) {
+		printk(KERN_ERR "md: failed to register integrity for %s\n",
+			mdname(mddev));
+		return -EINVAL;
+	}
+	printk(KERN_NOTICE "md: data integrity enabled on %s\n", mdname(mddev));
+	if (bioset_integrity_create(mddev->bio_set, BIO_POOL_SIZE)) {
+		printk(KERN_ERR "md: failed to create integrity pool for %s\n",
+		       mdname(mddev));
+		return -EINVAL;
+	}
+	return 0;
+}
+EXPORT_SYMBOL(md_integrity_register);
+
+/* Disable data integrity if non-capable/non-matching disk is being added */
+void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev)
+{
+	struct blk_integrity *bi_rdev;
+	struct blk_integrity *bi_mddev;
+
+	if (!mddev->gendisk)
+		return;
+
+	bi_rdev = bdev_get_integrity(rdev->bdev);
+	bi_mddev = blk_get_integrity(mddev->gendisk);
+
+	if (!bi_mddev) /* nothing to do */
+		return;
+	if (rdev->raid_disk < 0) /* skip spares */
+		return;
+	if (bi_rdev && blk_integrity_compare(mddev->gendisk,
+					     rdev->bdev->bd_disk) >= 0)
+		return;
+	printk(KERN_NOTICE "disabling data integrity on %s\n", mdname(mddev));
+	blk_integrity_unregister(mddev->gendisk);
+}
+EXPORT_SYMBOL(md_integrity_add_rdev);
+
+static int bind_rdev_to_array(struct md_rdev * rdev, struct mddev * mddev)
 {
 	char b[BDEVNAME_SIZE];
 	struct kobject *ko;
@@ -1427,8 +2049,9 @@ static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
 	if (find_rdev(mddev, rdev->bdev->bd_dev))
 		return -EEXIST;
 
-	/* make sure rdev->size exceeds mddev->size */
-	if (rdev->size && (mddev->size == 0 || rdev->size < mddev->size)) {
+	/* make sure rdev->sectors exceeds mddev->dev_sectors */
+	if (rdev->sectors && (mddev->dev_sectors == 0 ||
+			rdev->sectors < mddev->dev_sectors)) {
 		if (mddev->pers) {
 			/* Cannot change size, so fail
 			 * If mddev->level <= 0, then we don't care
@@ -1437,7 +2060,7 @@ static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
 			if (mddev->level > 0)
 				return -ENOSPC;
 		} else
-			mddev->size = rdev->size;
+			mddev->dev_sectors = rdev->sectors;
 	}
 
 	/* Verify rdev->desc_nr is unique.
@@ -1454,6 +2077,11 @@ static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
 		if (find_rdev_nr(mddev, rdev->desc_nr))
 			return -EBUSY;
 	}
+	if (mddev->max_disks && rdev->desc_nr >= mddev->max_disks) {
+		printk(KERN_WARNING "md: %s: array is limited to %d devices\n",
+		       mdname(mddev), mddev->max_disks);
+		return -EBUSY;
+	}
 	bdevname(rdev->bdev,b);
 	while ( (s=strchr(b, '/')) != NULL)
 		*s = '!';
@@ -1464,16 +2092,17 @@ static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
 	if ((err = kobject_add(&rdev->kobj, &mddev->kobj, "dev-%s", b)))
 		goto fail;
 
-	if (rdev->bdev->bd_part)
-		ko = &rdev->bdev->bd_part->dev.kobj;
-	else
-		ko = &rdev->bdev->bd_disk->dev.kobj;
-	if ((err = sysfs_create_link(&rdev->kobj, ko, "block"))) {
-		kobject_del(&rdev->kobj);
-		goto fail;
-	}
+	ko = &part_to_dev(rdev->bdev->bd_part)->kobj;
+	if (sysfs_create_link(&rdev->kobj, ko, "block"))
+		/* failure here is OK */;
+	rdev->sysfs_state = sysfs_get_dirent_safe(rdev->kobj.sd, "state");
+
 	list_add_rcu(&rdev->same_set, &mddev->disks);
-	bd_claim_by_disk(rdev->bdev, rdev->bdev->bd_holder, mddev->gendisk);
+	bd_link_disk_holder(rdev->bdev, mddev->gendisk);
+
+	/* May as well allow recovery to be retried once */
+	mddev->recovery_disabled++;
+
 	return 0;
 
  fail:
@@ -1484,24 +2113,26 @@ static int bind_rdev_to_array(mdk_rdev_t * rdev, mddev_t * mddev)
 
 static void md_delayed_delete(struct work_struct *ws)
 {
-	mdk_rdev_t *rdev = container_of(ws, mdk_rdev_t, del_work);
+	struct md_rdev *rdev = container_of(ws, struct md_rdev, del_work);
 	kobject_del(&rdev->kobj);
 	kobject_put(&rdev->kobj);
 }
 
-static void unbind_rdev_from_array(mdk_rdev_t * rdev)
+static void unbind_rdev_from_array(struct md_rdev * rdev)
 {
 	char b[BDEVNAME_SIZE];
 	if (!rdev->mddev) {
 		MD_BUG();
 		return;
 	}
-	bd_release_from_disk(rdev->bdev, rdev->mddev->gendisk);
+	bd_unlink_disk_holder(rdev->bdev, rdev->mddev->gendisk);
 	list_del_rcu(&rdev->same_set);
 	printk(KERN_INFO "md: unbind<%s>\n", bdevname(rdev->bdev,b));
 	rdev->mddev = NULL;
 	sysfs_remove_link(&rdev->kobj, "block");
-
+	sysfs_put(rdev->sysfs_state);
+	rdev->sysfs_state = NULL;
+	rdev->badblocks.count = 0;
 	/* We need to delay this, otherwise we can deadlock when
 	 * writing to 'remove' to "dev/state".  We also need
 	 * to delay it due to rcu usage.
@@ -1509,7 +2140,7 @@ static void unbind_rdev_from_array(mdk_rdev_t * rdev)
 	synchronize_rcu();
 	INIT_WORK(&rdev->del_work, md_delayed_delete);
 	kobject_get(&rdev->kobj);
-	schedule_work(&rdev->del_work);
+	queue_work(md_misc_wq, &rdev->del_work);
 }
 
 /*
@@ -1517,51 +2148,42 @@ static void unbind_rdev_from_array(mdk_rdev_t * rdev)
  * otherwise reused by a RAID array (or any other kernel
  * subsystem), by bd_claiming the device.
  */
-static int lock_rdev(mdk_rdev_t *rdev, dev_t dev, int shared)
+static int lock_rdev(struct md_rdev *rdev, dev_t dev, int shared)
 {
 	int err = 0;
 	struct block_device *bdev;
 	char b[BDEVNAME_SIZE];
 
-	bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
+	bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL,
+				 shared ? (struct md_rdev *)lock_rdev : rdev);
 	if (IS_ERR(bdev)) {
 		printk(KERN_ERR "md: could not open %s.\n",
 			__bdevname(dev, b));
 		return PTR_ERR(bdev);
 	}
-	err = bd_claim(bdev, shared ? (mdk_rdev_t *)lock_rdev : rdev);
-	if (err) {
-		printk(KERN_ERR "md: could not bd_claim %s.\n",
-			bdevname(bdev, b));
-		blkdev_put(bdev);
-		return err;
-	}
-	if (!shared)
-		set_bit(AllReserved, &rdev->flags);
 	rdev->bdev = bdev;
 	return err;
 }
 
-static void unlock_rdev(mdk_rdev_t *rdev)
+static void unlock_rdev(struct md_rdev *rdev)
 {
 	struct block_device *bdev = rdev->bdev;
 	rdev->bdev = NULL;
 	if (!bdev)
 		MD_BUG();
-	bd_release(bdev);
-	blkdev_put(bdev);
+	blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
 }
 
 void md_autodetect_dev(dev_t dev);
 
-static void export_rdev(mdk_rdev_t * rdev)
+static void export_rdev(struct md_rdev * rdev)
 {
 	char b[BDEVNAME_SIZE];
 	printk(KERN_INFO "md: export_rdev(%s)\n",
 		bdevname(rdev->bdev,b));
 	if (rdev->mddev)
 		MD_BUG();
-	free_disk_sb(rdev);
+	md_rdev_clear(rdev);
 #ifndef MODULE
 	if (test_bit(AutoDetected, &rdev->flags))
 		md_autodetect_dev(rdev->bdev->bd_dev);
@@ -1570,18 +2192,17 @@ static void export_rdev(mdk_rdev_t * rdev)
 	kobject_put(&rdev->kobj);
 }
 
-static void kick_rdev_from_array(mdk_rdev_t * rdev)
+static void kick_rdev_from_array(struct md_rdev * rdev)
 {
 	unbind_rdev_from_array(rdev);
 	export_rdev(rdev);
 }
 
-static void export_array(mddev_t *mddev)
+static void export_array(struct mddev *mddev)
 {
-	struct list_head *tmp;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev, *tmp;
 
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each_safe(rdev, tmp, mddev) {
 		if (!rdev->mddev) {
 			MD_BUG();
 			continue;
@@ -1600,7 +2221,7 @@ static void print_desc(mdp_disk_t *desc)
 		desc->major,desc->minor,desc->raid_disk,desc->state);
 }
 
-static void print_sb(mdp_super_t *sb)
+static void print_sb_90(mdp_super_t *sb)
 {
 	int i;
 
@@ -1631,28 +2252,76 @@ static void print_sb(mdp_super_t *sb)
 	}
 	printk(KERN_INFO "md:     THIS: ");
 	print_desc(&sb->this_disk);
-
 }
 
-static void print_rdev(mdk_rdev_t *rdev)
+static void print_sb_1(struct mdp_superblock_1 *sb)
+{
+	__u8 *uuid;
+
+	uuid = sb->set_uuid;
+	printk(KERN_INFO
+	       "md:  SB: (V:%u) (F:0x%08x) Array-ID:<%pU>\n"
+	       "md:    Name: \"%s\" CT:%llu\n",
+		le32_to_cpu(sb->major_version),
+		le32_to_cpu(sb->feature_map),
+		uuid,
+		sb->set_name,
+		(unsigned long long)le64_to_cpu(sb->ctime)
+		       & MD_SUPERBLOCK_1_TIME_SEC_MASK);
+
+	uuid = sb->device_uuid;
+	printk(KERN_INFO
+	       "md:       L%u SZ%llu RD:%u LO:%u CS:%u DO:%llu DS:%llu SO:%llu"
+			" RO:%llu\n"
+	       "md:     Dev:%08x UUID: %pU\n"
+	       "md:       (F:0x%08x) UT:%llu Events:%llu ResyncOffset:%llu CSUM:0x%08x\n"
+	       "md:         (MaxDev:%u) \n",
+		le32_to_cpu(sb->level),
+		(unsigned long long)le64_to_cpu(sb->size),
+		le32_to_cpu(sb->raid_disks),
+		le32_to_cpu(sb->layout),
+		le32_to_cpu(sb->chunksize),
+		(unsigned long long)le64_to_cpu(sb->data_offset),
+		(unsigned long long)le64_to_cpu(sb->data_size),
+		(unsigned long long)le64_to_cpu(sb->super_offset),
+		(unsigned long long)le64_to_cpu(sb->recovery_offset),
+		le32_to_cpu(sb->dev_number),
+		uuid,
+		sb->devflags,
+		(unsigned long long)le64_to_cpu(sb->utime) & MD_SUPERBLOCK_1_TIME_SEC_MASK,
+		(unsigned long long)le64_to_cpu(sb->events),
+		(unsigned long long)le64_to_cpu(sb->resync_offset),
+		le32_to_cpu(sb->sb_csum),
+		le32_to_cpu(sb->max_dev)
+		);
+}
+
+static void print_rdev(struct md_rdev *rdev, int major_version)
 {
 	char b[BDEVNAME_SIZE];
-	printk(KERN_INFO "md: rdev %s, SZ:%08llu F:%d S:%d DN:%u\n",
-		bdevname(rdev->bdev,b), (unsigned long long)rdev->size,
+	printk(KERN_INFO "md: rdev %s, Sect:%08llu F:%d S:%d DN:%u\n",
+		bdevname(rdev->bdev, b), (unsigned long long)rdev->sectors,
 	        test_bit(Faulty, &rdev->flags), test_bit(In_sync, &rdev->flags),
 	        rdev->desc_nr);
 	if (rdev->sb_loaded) {
-		printk(KERN_INFO "md: rdev superblock:\n");
-		print_sb((mdp_super_t*)page_address(rdev->sb_page));
+		printk(KERN_INFO "md: rdev superblock (MJ:%d):\n", major_version);
+		switch (major_version) {
+		case 0:
+			print_sb_90(page_address(rdev->sb_page));
+			break;
+		case 1:
+			print_sb_1(page_address(rdev->sb_page));
+			break;
+		}
 	} else
 		printk(KERN_INFO "md: no rdev superblock!\n");
 }
 
 static void md_print_devices(void)
 {
-	struct list_head *tmp, *tmp2;
-	mdk_rdev_t *rdev;
-	mddev_t *mddev;
+	struct list_head *tmp;
+	struct md_rdev *rdev;
+	struct mddev *mddev;
 	char b[BDEVNAME_SIZE];
 
 	printk("\n");
@@ -1665,19 +2334,19 @@ static void md_print_devices(void)
 			bitmap_print_sb(mddev->bitmap);
 		else
 			printk("%s: ", mdname(mddev));
-		rdev_for_each(rdev, tmp2, mddev)
+		rdev_for_each(rdev, mddev)
 			printk("<%s>", bdevname(rdev->bdev,b));
 		printk("\n");
 
-		rdev_for_each(rdev, tmp2, mddev)
-			print_rdev(rdev);
+		rdev_for_each(rdev, mddev)
+			print_rdev(rdev, mddev->major_version);
 	}
 	printk("md:	**********************************\n");
 	printk("\n");
 }
 
 
-static void sync_sbs(mddev_t * mddev, int nospares)
+static void sync_sbs(struct mddev * mddev, int nospares)
 {
 	/* Update each superblock (in-memory image), but
 	 * if we are allowed to, skip spares which already
@@ -1685,38 +2354,67 @@ static void sync_sbs(mddev_t * mddev, int nospares)
 	 * (which would mean they aren't being marked as dirty
 	 * with the rest of the array)
 	 */
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
-
-	rdev_for_each(rdev, tmp, mddev) {
+	struct md_rdev *rdev;
+	rdev_for_each(rdev, mddev) {
 		if (rdev->sb_events == mddev->events ||
 		    (nospares &&
 		     rdev->raid_disk < 0 &&
-		     (rdev->sb_events&1)==0 &&
 		     rdev->sb_events+1 == mddev->events)) {
 			/* Don't update this superblock */
 			rdev->sb_loaded = 2;
 		} else {
-			super_types[mddev->major_version].
-				sync_super(mddev, rdev);
+			sync_super(mddev, rdev);
 			rdev->sb_loaded = 1;
 		}
 	}
 }
 
-static void md_update_sb(mddev_t * mddev, int force_change)
+static void md_update_sb(struct mddev * mddev, int force_change)
 {
-	struct list_head *tmp;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 	int sync_req;
 	int nospares = 0;
+	int any_badblocks_changed = 0;
 
-	if (mddev->external)
+	if (mddev->ro) {
+		if (force_change)
+			set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		return;
+	}
 repeat:
+	/* First make sure individual recovery_offsets are correct */
+	rdev_for_each(rdev, mddev) {
+		if (rdev->raid_disk >= 0 &&
+		    mddev->delta_disks >= 0 &&
+		    !test_bit(In_sync, &rdev->flags) &&
+		    mddev->curr_resync_completed > rdev->recovery_offset)
+				rdev->recovery_offset = mddev->curr_resync_completed;
+
+	}	
+	if (!mddev->persistent) {
+		clear_bit(MD_CHANGE_CLEAN, &mddev->flags);
+		clear_bit(MD_CHANGE_DEVS, &mddev->flags);
+		if (!mddev->external) {
+			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
+			rdev_for_each(rdev, mddev) {
+				if (rdev->badblocks.changed) {
+					rdev->badblocks.changed = 0;
+					md_ack_all_badblocks(&rdev->badblocks);
+					md_error(mddev, rdev);
+				}
+				clear_bit(Blocked, &rdev->flags);
+				clear_bit(BlockedBadBlocks, &rdev->flags);
+				wake_up(&rdev->blocked_wait);
+			}
+		}
+		wake_up(&mddev->sb_wait);
+		return;
+	}
+
 	spin_lock_irq(&mddev->write_lock);
 
-	set_bit(MD_CHANGE_PENDING, &mddev->flags);
+	mddev->utime = get_seconds();
+
 	if (test_and_clear_bit(MD_CHANGE_DEVS, &mddev->flags))
 		force_change = 1;
 	if (test_and_clear_bit(MD_CHANGE_CLEAN, &mddev->flags))
@@ -1740,31 +2438,19 @@ repeat:
 		nospares = 0;
 
 	sync_req = mddev->in_sync;
-	mddev->utime = get_seconds();
 
 	/* If this is just a dirty<->clean transition, and the array is clean
 	 * and 'events' is odd, we can roll back to the previous clean state */
 	if (nospares
 	    && (mddev->in_sync && mddev->recovery_cp == MaxSector)
-	    && (mddev->events & 1)
-	    && mddev->events != 1)
+	    && mddev->can_decrease_events
+	    && mddev->events != 1) {
 		mddev->events--;
-	else {
+		mddev->can_decrease_events = 0;
+	} else {
 		/* otherwise we have to go forward and ... */
 		mddev->events ++;
-		if (!mddev->in_sync || mddev->recovery_cp != MaxSector) { /* not clean */
-			/* .. if the array isn't clean, insist on an odd 'events' */
-			if ((mddev->events&1)==0) {
-				mddev->events++;
-				nospares = 0;
-			}
-		} else {
-			/* otherwise insist on an even 'events' (for clean states) */
-			if ((mddev->events&1)) {
-				mddev->events++;
-				nospares = 0;
-			}
-		}
+		mddev->can_decrease_events = nospares;
 	}
 
 	if (!mddev->events) {
@@ -1777,46 +2463,46 @@ repeat:
 		mddev->events --;
 	}
 
-	/*
-	 * do not write anything to disk if using
-	 * nonpersistent superblocks
-	 */
-	if (!mddev->persistent) {
-		if (!mddev->external)
-			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
-
-		spin_unlock_irq(&mddev->write_lock);
-		wake_up(&mddev->sb_wait);
-		return;
+	rdev_for_each(rdev, mddev) {
+		if (rdev->badblocks.changed)
+			any_badblocks_changed++;
+		if (test_bit(Faulty, &rdev->flags))
+			set_bit(FaultRecorded, &rdev->flags);
 	}
+
 	sync_sbs(mddev, nospares);
 	spin_unlock_irq(&mddev->write_lock);
 
-	dprintk(KERN_INFO 
-		"md: updating %s RAID superblock on device (in sync %d)\n",
-		mdname(mddev),mddev->in_sync);
+	pr_debug("md: updating %s RAID superblock on device (in sync %d)\n",
+		 mdname(mddev), mddev->in_sync);
 
 	bitmap_update_sb(mddev->bitmap);
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev) {
 		char b[BDEVNAME_SIZE];
-		dprintk(KERN_INFO "md: ");
+
 		if (rdev->sb_loaded != 1)
 			continue; /* no noise on spare devices */
-		if (test_bit(Faulty, &rdev->flags))
-			dprintk("(skipping faulty ");
 
-		dprintk("%s ", bdevname(rdev->bdev,b));
 		if (!test_bit(Faulty, &rdev->flags)) {
 			md_super_write(mddev,rdev,
 				       rdev->sb_start, rdev->sb_size,
 				       rdev->sb_page);
-			dprintk(KERN_INFO "(write) %s's sb offset: %llu\n",
-				bdevname(rdev->bdev,b),
-				(unsigned long long)rdev->sb_start);
+			pr_debug("md: (write) %s's sb offset: %llu\n",
+				 bdevname(rdev->bdev, b),
+				 (unsigned long long)rdev->sb_start);
 			rdev->sb_events = mddev->events;
+			if (rdev->badblocks.size) {
+				md_super_write(mddev, rdev,
+					       rdev->badblocks.sector,
+					       rdev->badblocks.size << 9,
+					       rdev->bb_page);
+				rdev->badblocks.size = 0;
+			}
 
 		} else
-			dprintk(")\n");
+			pr_debug("md: %s (skipping faulty)\n",
+				 bdevname(rdev->bdev, b));
+
 		if (mddev->level == LEVEL_MULTIPATH)
 			/* only need to write one superblock... */
 			break;
@@ -1834,7 +2520,18 @@ repeat:
 	clear_bit(MD_CHANGE_PENDING, &mddev->flags);
 	spin_unlock_irq(&mddev->write_lock);
 	wake_up(&mddev->sb_wait);
+	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
+		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+
+	rdev_for_each(rdev, mddev) {
+		if (test_and_clear_bit(FaultRecorded, &rdev->flags))
+			clear_bit(Blocked, &rdev->flags);
 
+		if (any_badblocks_changed)
+			md_ack_all_badblocks(&rdev->badblocks);
+		clear_bit(BlockedBadBlocks, &rdev->flags);
+		wake_up(&rdev->blocked_wait);
+	}
 }
 
 /* words written to sysfs files may, or may not, be \n terminated.
@@ -1859,17 +2556,18 @@ static int cmd_match(const char *cmd, const char *str)
 
 struct rdev_sysfs_entry {
 	struct attribute attr;
-	ssize_t (*show)(mdk_rdev_t *, char *);
-	ssize_t (*store)(mdk_rdev_t *, const char *, size_t);
+	ssize_t (*show)(struct md_rdev *, char *);
+	ssize_t (*store)(struct md_rdev *, const char *, size_t);
 };
 
 static ssize_t
-state_show(mdk_rdev_t *rdev, char *page)
+state_show(struct md_rdev *rdev, char *page)
 {
 	char *sep = "";
 	size_t len = 0;
 
-	if (test_bit(Faulty, &rdev->flags)) {
+	if (test_bit(Faulty, &rdev->flags) ||
+	    rdev->badblocks.unacked_exist) {
 		len+= sprintf(page+len, "%sfaulty",sep);
 		sep = ",";
 	}
@@ -1881,7 +2579,9 @@ state_show(mdk_rdev_t *rdev, char *page)
 		len += sprintf(page+len, "%swrite_mostly",sep);
 		sep = ",";
 	}
-	if (test_bit(Blocked, &rdev->flags)) {
+	if (test_bit(Blocked, &rdev->flags) ||
+	    (rdev->badblocks.unacked_exist
+	     && !test_bit(Faulty, &rdev->flags))) {
 		len += sprintf(page+len, "%sblocked", sep);
 		sep = ",";
 	}
@@ -1890,29 +2590,50 @@ state_show(mdk_rdev_t *rdev, char *page)
 		len += sprintf(page+len, "%sspare", sep);
 		sep = ",";
 	}
+	if (test_bit(WriteErrorSeen, &rdev->flags)) {
+		len += sprintf(page+len, "%swrite_error", sep);
+		sep = ",";
+	}
+	if (test_bit(WantReplacement, &rdev->flags)) {
+		len += sprintf(page+len, "%swant_replacement", sep);
+		sep = ",";
+	}
+	if (test_bit(Replacement, &rdev->flags)) {
+		len += sprintf(page+len, "%sreplacement", sep);
+		sep = ",";
+	}
+
 	return len+sprintf(page+len, "\n");
 }
 
 static ssize_t
-state_store(mdk_rdev_t *rdev, const char *buf, size_t len)
+state_store(struct md_rdev *rdev, const char *buf, size_t len)
 {
 	/* can write
-	 *  faulty  - simulates and error
+	 *  faulty  - simulates an error
 	 *  remove  - disconnects the device
 	 *  writemostly - sets write_mostly
 	 *  -writemostly - clears write_mostly
-	 *  blocked - sets the Blocked flag
-	 *  -blocked - clears the Blocked flag
+	 *  blocked - sets the Blocked flags
+	 *  -blocked - clears the Blocked and possibly simulates an error
+	 *  insync - sets Insync providing device isn't active
+	 *  -insync - clear Insync for a device with a slot assigned,
+	 *            so that it gets rebuilt based on bitmap
+	 *  write_error - sets WriteErrorSeen
+	 *  -write_error - clears WriteErrorSeen
 	 */
 	int err = -EINVAL;
 	if (cmd_match(buf, "faulty") && rdev->mddev->pers) {
 		md_error(rdev->mddev, rdev);
-		err = 0;
+		if (test_bit(Faulty, &rdev->flags))
+			err = 0;
+		else
+			err = -EBUSY;
 	} else if (cmd_match(buf, "remove")) {
 		if (rdev->raid_disk >= 0)
 			err = -EBUSY;
 		else {
-			mddev_t *mddev = rdev->mddev;
+			struct mddev *mddev = rdev->mddev;
 			kick_rdev_from_array(rdev);
 			if (mddev->pers)
 				md_update_sb(mddev, 1);
@@ -1929,28 +2650,86 @@ state_store(mdk_rdev_t *rdev, const char *buf, size_t len)
 		set_bit(Blocked, &rdev->flags);
 		err = 0;
 	} else if (cmd_match(buf, "-blocked")) {
+		if (!test_bit(Faulty, &rdev->flags) &&
+		    rdev->badblocks.unacked_exist) {
+			/* metadata handler doesn't understand badblocks,
+			 * so we need to fail the device
+			 */
+			md_error(rdev->mddev, rdev);
+		}
 		clear_bit(Blocked, &rdev->flags);
+		clear_bit(BlockedBadBlocks, &rdev->flags);
 		wake_up(&rdev->blocked_wait);
 		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
 		md_wakeup_thread(rdev->mddev->thread);
 
 		err = 0;
+	} else if (cmd_match(buf, "insync") && rdev->raid_disk == -1) {
+		set_bit(In_sync, &rdev->flags);
+		err = 0;
+	} else if (cmd_match(buf, "-insync") && rdev->raid_disk >= 0) {
+		clear_bit(In_sync, &rdev->flags);
+		rdev->saved_raid_disk = rdev->raid_disk;
+		rdev->raid_disk = -1;
+		err = 0;
+	} else if (cmd_match(buf, "write_error")) {
+		set_bit(WriteErrorSeen, &rdev->flags);
+		err = 0;
+	} else if (cmd_match(buf, "-write_error")) {
+		clear_bit(WriteErrorSeen, &rdev->flags);
+		err = 0;
+	} else if (cmd_match(buf, "want_replacement")) {
+		/* Any non-spare device that is not a replacement can
+		 * become want_replacement at any time, but we then need to
+		 * check if recovery is needed.
+		 */
+		if (rdev->raid_disk >= 0 &&
+		    !test_bit(Replacement, &rdev->flags))
+			set_bit(WantReplacement, &rdev->flags);
+		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
+		md_wakeup_thread(rdev->mddev->thread);
+		err = 0;
+	} else if (cmd_match(buf, "-want_replacement")) {
+		/* Clearing 'want_replacement' is always allowed.
+		 * Once replacements starts it is too late though.
+		 */
+		err = 0;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else if (cmd_match(buf, "replacement")) {
+		/* Can only set a device as a replacement when array has not
+		 * yet been started.  Once running, replacement is automatic
+		 * from spares, or by assigning 'slot'.
+		 */
+		if (rdev->mddev->pers)
+			err = -EBUSY;
+		else {
+			set_bit(Replacement, &rdev->flags);
+			err = 0;
+		}
+	} else if (cmd_match(buf, "-replacement")) {
+		/* Similarly, can only clear Replacement before start */
+		if (rdev->mddev->pers)
+			err = -EBUSY;
+		else {
+			clear_bit(Replacement, &rdev->flags);
+			err = 0;
+		}
 	}
 	if (!err)
-		sysfs_notify(&rdev->kobj, NULL, "state");
+		sysfs_notify_dirent_safe(rdev->sysfs_state);
 	return err ? err : len;
 }
 static struct rdev_sysfs_entry rdev_state =
 __ATTR(state, S_IRUGO|S_IWUSR, state_show, state_store);
 
 static ssize_t
-errors_show(mdk_rdev_t *rdev, char *page)
+errors_show(struct md_rdev *rdev, char *page)
 {
 	return sprintf(page, "%d\n", atomic_read(&rdev->corrected_errors));
 }
 
 static ssize_t
-errors_store(mdk_rdev_t *rdev, const char *buf, size_t len)
+errors_store(struct md_rdev *rdev, const char *buf, size_t len)
 {
 	char *e;
 	unsigned long n = simple_strtoul(buf, &e, 10);
@@ -1964,7 +2743,7 @@ static struct rdev_sysfs_entry rdev_errors =
 __ATTR(errors, S_IRUGO|S_IWUSR, errors_show, errors_store);
 
 static ssize_t
-slot_show(mdk_rdev_t *rdev, char *page)
+slot_show(struct md_rdev *rdev, char *page)
 {
 	if (rdev->raid_disk < 0)
 		return sprintf(page, "none\n");
@@ -1973,11 +2752,10 @@ slot_show(mdk_rdev_t *rdev, char *page)
 }
 
 static ssize_t
-slot_store(mdk_rdev_t *rdev, const char *buf, size_t len)
+slot_store(struct md_rdev *rdev, const char *buf, size_t len)
 {
 	char *e;
 	int err;
-	char nm[20];
 	int slot = simple_strtoul(buf, &e, 10);
 	if (strncmp(buf, "none", 4)==0)
 		slot = -1;
@@ -1994,62 +2772,59 @@ slot_store(mdk_rdev_t *rdev, const char *buf, size_t len)
 		if (rdev->raid_disk == -1)
 			return -EEXIST;
 		/* personality does all needed checks */
-		if (rdev->mddev->pers->hot_add_disk == NULL)
+		if (rdev->mddev->pers->hot_remove_disk == NULL)
 			return -EINVAL;
-		err = rdev->mddev->pers->
-			hot_remove_disk(rdev->mddev, rdev->raid_disk);
-		if (err)
-			return err;
-		sprintf(nm, "rd%d", rdev->raid_disk);
-		sysfs_remove_link(&rdev->mddev->kobj, nm);
+		clear_bit(Blocked, &rdev->flags);
+		remove_and_add_spares(rdev->mddev, rdev);
+		if (rdev->raid_disk >= 0)
+			return -EBUSY;
 		set_bit(MD_RECOVERY_NEEDED, &rdev->mddev->recovery);
 		md_wakeup_thread(rdev->mddev->thread);
 	} else if (rdev->mddev->pers) {
-		mdk_rdev_t *rdev2;
-		struct list_head *tmp;
 		/* Activating a spare .. or possibly reactivating
-		 * if we every get bitmaps working here.
+		 * if we ever get bitmaps working here.
 		 */
 
 		if (rdev->raid_disk != -1)
 			return -EBUSY;
 
+		if (test_bit(MD_RECOVERY_RUNNING, &rdev->mddev->recovery))
+			return -EBUSY;
+
 		if (rdev->mddev->pers->hot_add_disk == NULL)
 			return -EINVAL;
 
-		rdev_for_each(rdev2, tmp, rdev->mddev)
-			if (rdev2->raid_disk == slot)
-				return -EEXIST;
+		if (slot >= rdev->mddev->raid_disks &&
+		    slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
+			return -ENOSPC;
 
 		rdev->raid_disk = slot;
 		if (test_bit(In_sync, &rdev->flags))
 			rdev->saved_raid_disk = slot;
 		else
 			rdev->saved_raid_disk = -1;
+		clear_bit(In_sync, &rdev->flags);
+		clear_bit(Bitmap_sync, &rdev->flags);
 		err = rdev->mddev->pers->
 			hot_add_disk(rdev->mddev, rdev);
 		if (err) {
 			rdev->raid_disk = -1;
 			return err;
 		} else
-			sysfs_notify(&rdev->kobj, NULL, "state");
-		sprintf(nm, "rd%d", rdev->raid_disk);
-		if (sysfs_create_link(&rdev->mddev->kobj, &rdev->kobj, nm))
-			printk(KERN_WARNING
-			       "md: cannot register "
-			       "%s for %s\n",
-			       nm, mdname(rdev->mddev));
-
+			sysfs_notify_dirent_safe(rdev->sysfs_state);
+		if (sysfs_link_rdev(rdev->mddev, rdev))
+			/* failure here is OK */;
 		/* don't wakeup anyone, leave that to userspace. */
 	} else {
-		if (slot >= rdev->mddev->raid_disks)
+		if (slot >= rdev->mddev->raid_disks &&
+		    slot >= rdev->mddev->raid_disks + rdev->mddev->delta_disks)
 			return -ENOSPC;
 		rdev->raid_disk = slot;
 		/* assume it is working */
 		clear_bit(Faulty, &rdev->flags);
 		clear_bit(WriteMostly, &rdev->flags);
 		set_bit(In_sync, &rdev->flags);
-		sysfs_notify(&rdev->kobj, NULL, "state");
+		sysfs_notify_dirent_safe(rdev->sysfs_state);
 	}
 	return len;
 }
@@ -2059,35 +2834,92 @@ static struct rdev_sysfs_entry rdev_slot =
 __ATTR(slot, S_IRUGO|S_IWUSR, slot_show, slot_store);
 
 static ssize_t
-offset_show(mdk_rdev_t *rdev, char *page)
+offset_show(struct md_rdev *rdev, char *page)
 {
 	return sprintf(page, "%llu\n", (unsigned long long)rdev->data_offset);
 }
 
 static ssize_t
-offset_store(mdk_rdev_t *rdev, const char *buf, size_t len)
+offset_store(struct md_rdev *rdev, const char *buf, size_t len)
 {
-	char *e;
-	unsigned long long offset = simple_strtoull(buf, &e, 10);
-	if (e==buf || (*e && *e != '\n'))
+	unsigned long long offset;
+	if (kstrtoull(buf, 10, &offset) < 0)
 		return -EINVAL;
 	if (rdev->mddev->pers && rdev->raid_disk >= 0)
 		return -EBUSY;
-	if (rdev->size && rdev->mddev->external)
+	if (rdev->sectors && rdev->mddev->external)
 		/* Must set offset before size, so overlap checks
 		 * can be sane */
 		return -EBUSY;
 	rdev->data_offset = offset;
+	rdev->new_data_offset = offset;
 	return len;
 }
 
 static struct rdev_sysfs_entry rdev_offset =
 __ATTR(offset, S_IRUGO|S_IWUSR, offset_show, offset_store);
 
+static ssize_t new_offset_show(struct md_rdev *rdev, char *page)
+{
+	return sprintf(page, "%llu\n",
+		       (unsigned long long)rdev->new_data_offset);
+}
+
+static ssize_t new_offset_store(struct md_rdev *rdev,
+				const char *buf, size_t len)
+{
+	unsigned long long new_offset;
+	struct mddev *mddev = rdev->mddev;
+
+	if (kstrtoull(buf, 10, &new_offset) < 0)
+		return -EINVAL;
+
+	if (mddev->sync_thread)
+		return -EBUSY;
+	if (new_offset == rdev->data_offset)
+		/* reset is always permitted */
+		;
+	else if (new_offset > rdev->data_offset) {
+		/* must not push array size beyond rdev_sectors */
+		if (new_offset - rdev->data_offset
+		    + mddev->dev_sectors > rdev->sectors)
+				return -E2BIG;
+	}
+	/* Metadata worries about other space details. */
+
+	/* decreasing the offset is inconsistent with a backwards
+	 * reshape.
+	 */
+	if (new_offset < rdev->data_offset &&
+	    mddev->reshape_backwards)
+		return -EINVAL;
+	/* Increasing offset is inconsistent with forwards
+	 * reshape.  reshape_direction should be set to
+	 * 'backwards' first.
+	 */
+	if (new_offset > rdev->data_offset &&
+	    !mddev->reshape_backwards)
+		return -EINVAL;
+
+	if (mddev->pers && mddev->persistent &&
+	    !super_types[mddev->major_version]
+	    .allow_new_offset(rdev, new_offset))
+		return -E2BIG;
+	rdev->new_data_offset = new_offset;
+	if (new_offset > rdev->data_offset)
+		mddev->reshape_backwards = 1;
+	else if (new_offset < rdev->data_offset)
+		mddev->reshape_backwards = 0;
+
+	return len;
+}
+static struct rdev_sysfs_entry rdev_new_offset =
+__ATTR(new_offset, S_IRUGO|S_IWUSR, new_offset_show, new_offset_store);
+
 static ssize_t
-rdev_size_show(mdk_rdev_t *rdev, char *page)
+rdev_size_show(struct md_rdev *rdev, char *page)
 {
-	return sprintf(page, "%llu\n", (unsigned long long)rdev->size);
+	return sprintf(page, "%llu\n", (unsigned long long)rdev->sectors / 2);
 }
 
 static int overlaps(sector_t s1, sector_t l1, sector_t s2, sector_t l2)
@@ -2100,54 +2932,74 @@ static int overlaps(sector_t s1, sector_t l1, sector_t s2, sector_t l2)
 	return 1;
 }
 
-static ssize_t
-rdev_size_store(mdk_rdev_t *rdev, const char *buf, size_t len)
+static int strict_blocks_to_sectors(const char *buf, sector_t *sectors)
 {
-	unsigned long long size;
-	unsigned long long oldsize = rdev->size;
-	mddev_t *my_mddev = rdev->mddev;
+	unsigned long long blocks;
+	sector_t new;
 
-	if (strict_strtoull(buf, 10, &size) < 0)
+	if (kstrtoull(buf, 10, &blocks) < 0)
 		return -EINVAL;
-	if (size < my_mddev->size)
+
+	if (blocks & 1ULL << (8 * sizeof(blocks) - 1))
+		return -EINVAL; /* sector conversion overflow */
+
+	new = blocks * 2;
+	if (new != blocks * 2)
+		return -EINVAL; /* unsigned long long to sector_t overflow */
+
+	*sectors = new;
+	return 0;
+}
+
+static ssize_t
+rdev_size_store(struct md_rdev *rdev, const char *buf, size_t len)
+{
+	struct mddev *my_mddev = rdev->mddev;
+	sector_t oldsectors = rdev->sectors;
+	sector_t sectors;
+
+	if (strict_blocks_to_sectors(buf, &sectors) < 0)
 		return -EINVAL;
+	if (rdev->data_offset != rdev->new_data_offset)
+		return -EINVAL; /* too confusing */
 	if (my_mddev->pers && rdev->raid_disk >= 0) {
 		if (my_mddev->persistent) {
-			size = super_types[my_mddev->major_version].
-				rdev_size_change(rdev, size * 2);
-			if (!size)
+			sectors = super_types[my_mddev->major_version].
+				rdev_size_change(rdev, sectors);
+			if (!sectors)
 				return -EBUSY;
-		} else if (!size) {
-			size = (rdev->bdev->bd_inode->i_size >> 10);
-			size -= rdev->data_offset/2;
-		}
-		if (size < my_mddev->size)
-			return -EINVAL; /* component must fit device */
+		} else if (!sectors)
+			sectors = (i_size_read(rdev->bdev->bd_inode) >> 9) -
+				rdev->data_offset;
+		if (!my_mddev->pers->resize)
+			/* Cannot change size for RAID0 or Linear etc */
+			return -EINVAL;
 	}
+	if (sectors < my_mddev->dev_sectors)
+		return -EINVAL; /* component must fit device */
 
-	rdev->size = size;
-	if (size > oldsize && my_mddev->external) {
+	rdev->sectors = sectors;
+	if (sectors > oldsectors && my_mddev->external) {
 		/* need to check that all other rdevs with the same ->bdev
 		 * do not overlap.  We need to unlock the mddev to avoid
-		 * a deadlock.  We have already changed rdev->size, and if
+		 * a deadlock.  We have already changed rdev->sectors, and if
 		 * we have to change it back, we will have the lock again.
 		 */
-		mddev_t *mddev;
+		struct mddev *mddev;
 		int overlap = 0;
-		struct list_head *tmp, *tmp2;
+		struct list_head *tmp;
 
 		mddev_unlock(my_mddev);
 		for_each_mddev(mddev, tmp) {
-			mdk_rdev_t *rdev2;
-
-			mddev_lock(mddev);
-			rdev_for_each(rdev2, tmp2, mddev)
-				if (test_bit(AllReserved, &rdev2->flags) ||
-				    (rdev->bdev == rdev2->bdev &&
-				     rdev != rdev2 &&
-				     overlaps(rdev->data_offset, rdev->size * 2,
-					      rdev2->data_offset,
-					      rdev2->size * 2))) {
+			struct md_rdev *rdev2;
+
+			mddev_lock_nointr(mddev);
+			rdev_for_each(rdev2, mddev)
+				if (rdev->bdev == rdev2->bdev &&
+				    rdev != rdev2 &&
+				    overlaps(rdev->data_offset, rdev->sectors,
+					     rdev2->data_offset,
+					     rdev2->sectors)) {
 					overlap = 1;
 					break;
 				}
@@ -2157,15 +3009,15 @@ rdev_size_store(mdk_rdev_t *rdev, const char *buf, size_t len)
 				break;
 			}
 		}
-		mddev_lock(my_mddev);
+		mddev_lock_nointr(my_mddev);
 		if (overlap) {
 			/* Someone else could have slipped in a size
 			 * change here, but doing so is just silly.
-			 * We put oldsize back because we *know* it is
+			 * We put oldsectors back because we *know* it is
 			 * safe, and trust userspace not to race with
 			 * itself
 			 */
-			rdev->size = oldsize;
+			rdev->sectors = oldsectors;
 			return -EBUSY;
 		}
 	}
@@ -2175,20 +3027,93 @@ rdev_size_store(mdk_rdev_t *rdev, const char *buf, size_t len)
 static struct rdev_sysfs_entry rdev_size =
 __ATTR(size, S_IRUGO|S_IWUSR, rdev_size_show, rdev_size_store);
 
+
+static ssize_t recovery_start_show(struct md_rdev *rdev, char *page)
+{
+	unsigned long long recovery_start = rdev->recovery_offset;
+
+	if (test_bit(In_sync, &rdev->flags) ||
+	    recovery_start == MaxSector)
+		return sprintf(page, "none\n");
+
+	return sprintf(page, "%llu\n", recovery_start);
+}
+
+static ssize_t recovery_start_store(struct md_rdev *rdev, const char *buf, size_t len)
+{
+	unsigned long long recovery_start;
+
+	if (cmd_match(buf, "none"))
+		recovery_start = MaxSector;
+	else if (kstrtoull(buf, 10, &recovery_start))
+		return -EINVAL;
+
+	if (rdev->mddev->pers &&
+	    rdev->raid_disk >= 0)
+		return -EBUSY;
+
+	rdev->recovery_offset = recovery_start;
+	if (recovery_start == MaxSector)
+		set_bit(In_sync, &rdev->flags);
+	else
+		clear_bit(In_sync, &rdev->flags);
+	return len;
+}
+
+static struct rdev_sysfs_entry rdev_recovery_start =
+__ATTR(recovery_start, S_IRUGO|S_IWUSR, recovery_start_show, recovery_start_store);
+
+
+static ssize_t
+badblocks_show(struct badblocks *bb, char *page, int unack);
+static ssize_t
+badblocks_store(struct badblocks *bb, const char *page, size_t len, int unack);
+
+static ssize_t bb_show(struct md_rdev *rdev, char *page)
+{
+	return badblocks_show(&rdev->badblocks, page, 0);
+}
+static ssize_t bb_store(struct md_rdev *rdev, const char *page, size_t len)
+{
+	int rv = badblocks_store(&rdev->badblocks, page, len, 0);
+	/* Maybe that ack was all we needed */
+	if (test_and_clear_bit(BlockedBadBlocks, &rdev->flags))
+		wake_up(&rdev->blocked_wait);
+	return rv;
+}
+static struct rdev_sysfs_entry rdev_bad_blocks =
+__ATTR(bad_blocks, S_IRUGO|S_IWUSR, bb_show, bb_store);
+
+
+static ssize_t ubb_show(struct md_rdev *rdev, char *page)
+{
+	return badblocks_show(&rdev->badblocks, page, 1);
+}
+static ssize_t ubb_store(struct md_rdev *rdev, const char *page, size_t len)
+{
+	return badblocks_store(&rdev->badblocks, page, len, 1);
+}
+static struct rdev_sysfs_entry rdev_unack_bad_blocks =
+__ATTR(unacknowledged_bad_blocks, S_IRUGO|S_IWUSR, ubb_show, ubb_store);
+
 static struct attribute *rdev_default_attrs[] = {
 	&rdev_state.attr,
 	&rdev_errors.attr,
 	&rdev_slot.attr,
 	&rdev_offset.attr,
+	&rdev_new_offset.attr,
 	&rdev_size.attr,
+	&rdev_recovery_start.attr,
+	&rdev_bad_blocks.attr,
+	&rdev_unack_bad_blocks.attr,
 	NULL,
 };
 static ssize_t
 rdev_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 {
 	struct rdev_sysfs_entry *entry = container_of(attr, struct rdev_sysfs_entry, attr);
-	mdk_rdev_t *rdev = container_of(kobj, mdk_rdev_t, kobj);
-	mddev_t *mddev = rdev->mddev;
+	struct md_rdev *rdev = container_of(kobj, struct md_rdev, kobj);
+	struct mddev *mddev = rdev->mddev;
 	ssize_t rv;
 
 	if (!entry->show)
@@ -2210,9 +3135,9 @@ rdev_attr_store(struct kobject *kobj, struct attribute *attr,
 	      const char *page, size_t length)
 {
 	struct rdev_sysfs_entry *entry = container_of(attr, struct rdev_sysfs_entry, attr);
-	mdk_rdev_t *rdev = container_of(kobj, mdk_rdev_t, kobj);
+	struct md_rdev *rdev = container_of(kobj, struct md_rdev, kobj);
 	ssize_t rv;
-	mddev_t *mddev = rdev->mddev;
+	struct mddev *mddev = rdev->mddev;
 
 	if (!entry->store)
 		return -EIO;
@@ -2231,10 +3156,10 @@ rdev_attr_store(struct kobject *kobj, struct attribute *attr,
 
 static void rdev_free(struct kobject *ko)
 {
-	mdk_rdev_t *rdev = container_of(ko, mdk_rdev_t, kobj);
+	struct md_rdev *rdev = container_of(ko, struct md_rdev, kobj);
 	kfree(rdev);
 }
-static struct sysfs_ops rdev_sysfs_ops = {
+static const struct sysfs_ops rdev_sysfs_ops = {
 	.show		= rdev_attr_show,
 	.store		= rdev_attr_store,
 };
@@ -2244,6 +3169,40 @@ static struct kobj_type rdev_ktype = {
 	.default_attrs	= rdev_default_attrs,
 };
 
+int md_rdev_init(struct md_rdev *rdev)
+{
+	rdev->desc_nr = -1;
+	rdev->saved_raid_disk = -1;
+	rdev->raid_disk = -1;
+	rdev->flags = 0;
+	rdev->data_offset = 0;
+	rdev->new_data_offset = 0;
+	rdev->sb_events = 0;
+	rdev->last_read_error.tv_sec  = 0;
+	rdev->last_read_error.tv_nsec = 0;
+	rdev->sb_loaded = 0;
+	rdev->bb_page = NULL;
+	atomic_set(&rdev->nr_pending, 0);
+	atomic_set(&rdev->read_errors, 0);
+	atomic_set(&rdev->corrected_errors, 0);
+
+	INIT_LIST_HEAD(&rdev->same_set);
+	init_waitqueue_head(&rdev->blocked_wait);
+
+	/* Add space to store bad block list.
+	 * This reserves the space even on arrays where it cannot
+	 * be used - I wonder if that matters
+	 */
+	rdev->badblocks.count = 0;
+	rdev->badblocks.shift = -1; /* disabled until explicitly enabled */
+	rdev->badblocks.page = kmalloc(PAGE_SIZE, GFP_KERNEL);
+	seqlock_init(&rdev->badblocks.lock);
+	if (rdev->badblocks.page == NULL)
+		return -ENOMEM;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(md_rdev_init);
 /*
  * Import a device. If 'super_format' >= 0, then sanity check the superblock
  *
@@ -2254,11 +3213,11 @@ static struct kobj_type rdev_ktype = {
  *
  * a faulty rdev _never_ has rdev->sb set.
  */
-static mdk_rdev_t *md_import_device(dev_t newdev, int super_format, int super_minor)
+static struct md_rdev *md_import_device(dev_t newdev, int super_format, int super_minor)
 {
 	char b[BDEVNAME_SIZE];
 	int err;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 	sector_t size;
 
 	rdev = kzalloc(sizeof(*rdev), GFP_KERNEL);
@@ -2267,7 +3226,11 @@ static mdk_rdev_t *md_import_device(dev_t newdev, int super_format, int super_mi
 		return ERR_PTR(-ENOMEM);
 	}
 
-	if ((err = alloc_disk_sb(rdev)))
+	err = md_rdev_init(rdev);
+	if (err)
+		goto abort_free;
+	err = alloc_disk_sb(rdev);
+	if (err)
 		goto abort_free;
 
 	err = lock_rdev(rdev, newdev, super_format == -2);
@@ -2276,17 +3239,7 @@ static mdk_rdev_t *md_import_device(dev_t newdev, int super_format, int super_mi
 
 	kobject_init(&rdev->kobj, &rdev_ktype);
 
-	rdev->desc_nr = -1;
-	rdev->saved_raid_disk = -1;
-	rdev->raid_disk = -1;
-	rdev->flags = 0;
-	rdev->data_offset = 0;
-	rdev->sb_events = 0;
-	atomic_set(&rdev->nr_pending, 0);
-	atomic_set(&rdev->read_errors, 0);
-	atomic_set(&rdev->corrected_errors, 0);
-
-	size = rdev->bdev->bd_inode->i_size >> BLOCK_SIZE_BITS;
+	size = i_size_read(rdev->bdev->bd_inode) >> BLOCK_SIZE_BITS;
 	if (!size) {
 		printk(KERN_WARNING 
 			"md: %s has zero or unknown size, marking faulty!\n",
@@ -2314,17 +3267,12 @@ static mdk_rdev_t *md_import_device(dev_t newdev, int super_format, int super_mi
 		}
 	}
 
-	INIT_LIST_HEAD(&rdev->same_set);
-	init_waitqueue_head(&rdev->blocked_wait);
-
 	return rdev;
 
 abort_free:
-	if (rdev->sb_page) {
-		if (rdev->bdev)
-			unlock_rdev(rdev);
-		free_disk_sb(rdev);
-	}
+	if (rdev->bdev)
+		unlock_rdev(rdev);
+	md_rdev_clear(rdev);
 	kfree(rdev);
 	return ERR_PTR(err);
 }
@@ -2334,15 +3282,14 @@ abort_free:
  */
 
 
-static void analyze_sbs(mddev_t * mddev)
+static void analyze_sbs(struct mddev * mddev)
 {
 	int i;
-	struct list_head *tmp;
-	mdk_rdev_t *rdev, *freshest;
+	struct md_rdev *rdev, *freshest, *tmp;
 	char b[BDEVNAME_SIZE];
 
 	freshest = NULL;
-	rdev_for_each(rdev, tmp, mddev)
+	rdev_for_each_safe(rdev, tmp, mddev)
 		switch (super_types[mddev->major_version].
 			load_super(rdev, freshest, mddev->minor_version)) {
 		case 1:
@@ -2363,7 +3310,17 @@ static void analyze_sbs(mddev_t * mddev)
 		validate_super(mddev, freshest);
 
 	i = 0;
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each_safe(rdev, tmp, mddev) {
+		if (mddev->max_disks &&
+		    (rdev->desc_nr >= mddev->max_disks ||
+		     i > mddev->max_disks)) {
+			printk(KERN_WARNING
+			       "md: %s: %s: only %d devices permitted\n",
+			       mdname(mddev), bdevname(rdev->bdev, b),
+			       mddev->max_disks);
+			kick_rdev_from_array(rdev);
+			continue;
+		}
 		if (rdev != freshest)
 			if (super_types[mddev->major_version].
 			    validate_super(mddev, rdev)) {
@@ -2377,60 +3334,69 @@ static void analyze_sbs(mddev_t * mddev)
 			rdev->desc_nr = i++;
 			rdev->raid_disk = rdev->desc_nr;
 			set_bit(In_sync, &rdev->flags);
-		} else if (rdev->raid_disk >= mddev->raid_disks) {
+		} else if (rdev->raid_disk >= (mddev->raid_disks - min(0, mddev->delta_disks))) {
 			rdev->raid_disk = -1;
 			clear_bit(In_sync, &rdev->flags);
 		}
 	}
+}
 
-
-
-	if (mddev->recovery_cp != MaxSector &&
-	    mddev->level >= 1)
-		printk(KERN_ERR "md: %s: raid array is not clean"
-		       " -- starting background reconstruction\n",
-		       mdname(mddev));
-
+/* Read a fixed-point number.
+ * Numbers in sysfs attributes should be in "standard" units where
+ * possible, so time should be in seconds.
+ * However we internally use a a much smaller unit such as 
+ * milliseconds or jiffies.
+ * This function takes a decimal number with a possible fractional
+ * component, and produces an integer which is the result of
+ * multiplying that number by 10^'scale'.
+ * all without any floating-point arithmetic.
+ */
+int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale)
+{
+	unsigned long result = 0;
+	long decimals = -1;
+	while (isdigit(*cp) || (*cp == '.' && decimals < 0)) {
+		if (*cp == '.')
+			decimals = 0;
+		else if (decimals < scale) {
+			unsigned int value;
+			value = *cp - '0';
+			result = result * 10 + value;
+			if (decimals >= 0)
+				decimals++;
+		}
+		cp++;
+	}
+	if (*cp == '\n')
+		cp++;
+	if (*cp)
+		return -EINVAL;
+	if (decimals < 0)
+		decimals = 0;
+	while (decimals < scale) {
+		result *= 10;
+		decimals ++;
+	}
+	*res = result;
+	return 0;
 }
 
+
 static void md_safemode_timeout(unsigned long data);
 
 static ssize_t
-safe_delay_show(mddev_t *mddev, char *page)
+safe_delay_show(struct mddev *mddev, char *page)
 {
 	int msec = (mddev->safemode_delay*1000)/HZ;
 	return sprintf(page, "%d.%03d\n", msec/1000, msec%1000);
 }
 static ssize_t
-safe_delay_store(mddev_t *mddev, const char *cbuf, size_t len)
+safe_delay_store(struct mddev *mddev, const char *cbuf, size_t len)
 {
-	int scale=1;
-	int dot=0;
-	int i;
 	unsigned long msec;
-	char buf[30];
-	char *e;
-	/* remove a period, and count digits after it */
-	if (len >= sizeof(buf))
-		return -EINVAL;
-	strlcpy(buf, cbuf, len);
-	buf[len] = 0;
-	for (i=0; i<len; i++) {
-		if (dot) {
-			if (isdigit(buf[i])) {
-				buf[i-1] = buf[i];
-				scale *= 10;
-			}
-			buf[i] = 0;
-		} else if (buf[i] == '.') {
-			dot=1;
-			buf[i] = 0;
-		}
-	}
-	msec = simple_strtoul(buf, &e, 10);
-	if (e == buf || (*e && *e != '\n'))
+
+	if (strict_strtoul_scaled(cbuf, &msec, 3) < 0)
 		return -EINVAL;
-	msec = (msec * 1000) / scale;
 	if (msec == 0)
 		mddev->safemode_delay = 0;
 	else {
@@ -2438,7 +3404,7 @@ safe_delay_store(mddev_t *mddev, const char *cbuf, size_t len)
 		mddev->safemode_delay = (msec*HZ)/1000;
 		if (mddev->safemode_delay == 0)
 			mddev->safemode_delay = 1;
-		if (mddev->safemode_delay < old_delay)
+		if (mddev->safemode_delay < old_delay || old_delay == 0)
 			md_safemode_timeout((unsigned long)mddev);
 	}
 	return len;
@@ -2447,9 +3413,9 @@ static struct md_sysfs_entry md_safe_delay =
 __ATTR(safe_mode_delay, S_IRUGO|S_IWUSR,safe_delay_show, safe_delay_store);
 
 static ssize_t
-level_show(mddev_t *mddev, char *page)
+level_show(struct mddev *mddev, char *page)
 {
-	struct mdk_personality *p = mddev->pers;
+	struct md_personality *p = mddev->pers;
 	if (p)
 		return sprintf(page, "%s\n", p->name);
 	else if (mddev->clevel[0])
@@ -2461,20 +3427,184 @@ level_show(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-level_store(mddev_t *mddev, const char *buf, size_t len)
+level_store(struct mddev *mddev, const char *buf, size_t len)
 {
+	char clevel[16];
 	ssize_t rv = len;
-	if (mddev->pers)
+	struct md_personality *pers;
+	long level;
+	void *priv;
+	struct md_rdev *rdev;
+
+	if (mddev->pers == NULL) {
+		if (len == 0)
+			return 0;
+		if (len >= sizeof(mddev->clevel))
+			return -ENOSPC;
+		strncpy(mddev->clevel, buf, len);
+		if (mddev->clevel[len-1] == '\n')
+			len--;
+		mddev->clevel[len] = 0;
+		mddev->level = LEVEL_NONE;
+		return rv;
+	}
+	if (mddev->ro)
+		return  -EROFS;
+
+	/* request to change the personality.  Need to ensure:
+	 *  - array is not engaged in resync/recovery/reshape
+	 *  - old personality can be suspended
+	 *  - new personality will access other array.
+	 */
+
+	if (mddev->sync_thread ||
+	    mddev->reshape_position != MaxSector ||
+	    mddev->sysfs_active)
 		return -EBUSY;
-	if (len == 0)
-		return 0;
-	if (len >= sizeof(mddev->clevel))
-		return -ENOSPC;
-	strncpy(mddev->clevel, buf, len);
-	if (mddev->clevel[len-1] == '\n')
+
+	if (!mddev->pers->quiesce) {
+		printk(KERN_WARNING "md: %s: %s does not support online personality change\n",
+		       mdname(mddev), mddev->pers->name);
+		return -EINVAL;
+	}
+
+	/* Now find the new personality */
+	if (len == 0 || len >= sizeof(clevel))
+		return -EINVAL;
+	strncpy(clevel, buf, len);
+	if (clevel[len-1] == '\n')
 		len--;
-	mddev->clevel[len] = 0;
-	mddev->level = LEVEL_NONE;
+	clevel[len] = 0;
+	if (kstrtol(clevel, 10, &level))
+		level = LEVEL_NONE;
+
+	if (request_module("md-%s", clevel) != 0)
+		request_module("md-level-%s", clevel);
+	spin_lock(&pers_lock);
+	pers = find_pers(level, clevel);
+	if (!pers || !try_module_get(pers->owner)) {
+		spin_unlock(&pers_lock);
+		printk(KERN_WARNING "md: personality %s not loaded\n", clevel);
+		return -EINVAL;
+	}
+	spin_unlock(&pers_lock);
+
+	if (pers == mddev->pers) {
+		/* Nothing to do! */
+		module_put(pers->owner);
+		return rv;
+	}
+	if (!pers->takeover) {
+		module_put(pers->owner);
+		printk(KERN_WARNING "md: %s: %s does not support personality takeover\n",
+		       mdname(mddev), clevel);
+		return -EINVAL;
+	}
+
+	rdev_for_each(rdev, mddev)
+		rdev->new_raid_disk = rdev->raid_disk;
+
+	/* ->takeover must set new_* and/or delta_disks
+	 * if it succeeds, and may set them when it fails.
+	 */
+	priv = pers->takeover(mddev);
+	if (IS_ERR(priv)) {
+		mddev->new_level = mddev->level;
+		mddev->new_layout = mddev->layout;
+		mddev->new_chunk_sectors = mddev->chunk_sectors;
+		mddev->raid_disks -= mddev->delta_disks;
+		mddev->delta_disks = 0;
+		mddev->reshape_backwards = 0;
+		module_put(pers->owner);
+		printk(KERN_WARNING "md: %s: %s would not accept array\n",
+		       mdname(mddev), clevel);
+		return PTR_ERR(priv);
+	}
+
+	/* Looks like we have a winner */
+	mddev_suspend(mddev);
+	mddev->pers->stop(mddev);
+	
+	if (mddev->pers->sync_request == NULL &&
+	    pers->sync_request != NULL) {
+		/* need to add the md_redundancy_group */
+		if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
+			printk(KERN_WARNING
+			       "md: cannot register extra attributes for %s\n",
+			       mdname(mddev));
+		mddev->sysfs_action = sysfs_get_dirent(mddev->kobj.sd, "sync_action");
+	}		
+	if (mddev->pers->sync_request != NULL &&
+	    pers->sync_request == NULL) {
+		/* need to remove the md_redundancy_group */
+		if (mddev->to_remove == NULL)
+			mddev->to_remove = &md_redundancy_group;
+	}
+
+	if (mddev->pers->sync_request == NULL &&
+	    mddev->external) {
+		/* We are converting from a no-redundancy array
+		 * to a redundancy array and metadata is managed
+		 * externally so we need to be sure that writes
+		 * won't block due to a need to transition
+		 *      clean->dirty
+		 * until external management is started.
+		 */
+		mddev->in_sync = 0;
+		mddev->safemode_delay = 0;
+		mddev->safemode = 0;
+	}
+
+	rdev_for_each(rdev, mddev) {
+		if (rdev->raid_disk < 0)
+			continue;
+		if (rdev->new_raid_disk >= mddev->raid_disks)
+			rdev->new_raid_disk = -1;
+		if (rdev->new_raid_disk == rdev->raid_disk)
+			continue;
+		sysfs_unlink_rdev(mddev, rdev);
+	}
+	rdev_for_each(rdev, mddev) {
+		if (rdev->raid_disk < 0)
+			continue;
+		if (rdev->new_raid_disk == rdev->raid_disk)
+			continue;
+		rdev->raid_disk = rdev->new_raid_disk;
+		if (rdev->raid_disk < 0)
+			clear_bit(In_sync, &rdev->flags);
+		else {
+			if (sysfs_link_rdev(mddev, rdev))
+				printk(KERN_WARNING "md: cannot register rd%d"
+				       " for %s after level change\n",
+				       rdev->raid_disk, mdname(mddev));
+		}
+	}
+
+	module_put(mddev->pers->owner);
+	mddev->pers = pers;
+	mddev->private = priv;
+	strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
+	mddev->level = mddev->new_level;
+	mddev->layout = mddev->new_layout;
+	mddev->chunk_sectors = mddev->new_chunk_sectors;
+	mddev->delta_disks = 0;
+	mddev->reshape_backwards = 0;
+	mddev->degraded = 0;
+	if (mddev->pers->sync_request == NULL) {
+		/* this is now an array without redundancy, so
+		 * it must always be in_sync
+		 */
+		mddev->in_sync = 1;
+		del_timer_sync(&mddev->safemode_timer);
+	}
+	blk_set_stacking_limits(&mddev->queue->limits);
+	pers->run(mddev);
+	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	mddev_resume(mddev);
+	if (!mddev->thread)
+		md_update_sb(mddev, 1);
+	sysfs_notify(&mddev->kobj, NULL, "level");
+	md_new_event(mddev);
 	return rv;
 }
 
@@ -2483,7 +3613,7 @@ __ATTR(level, S_IRUGO|S_IWUSR, level_show, level_store);
 
 
 static ssize_t
-layout_show(mddev_t *mddev, char *page)
+layout_show(struct mddev *mddev, char *page)
 {
 	/* just a number, not meaningful for all levels */
 	if (mddev->reshape_position != MaxSector &&
@@ -2494,7 +3624,7 @@ layout_show(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-layout_store(mddev_t *mddev, const char *buf, size_t len)
+layout_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *e;
 	unsigned long n = simple_strtoul(buf, &e, 10);
@@ -2502,12 +3632,23 @@ layout_store(mddev_t *mddev, const char *buf, size_t len)
 	if (!*buf || (*e && *e != '\n'))
 		return -EINVAL;
 
-	if (mddev->pers)
-		return -EBUSY;
-	if (mddev->reshape_position != MaxSector)
+	if (mddev->pers) {
+		int err;
+		if (mddev->pers->check_reshape == NULL)
+			return -EBUSY;
+		if (mddev->ro)
+			return -EROFS;
 		mddev->new_layout = n;
-	else
-		mddev->layout = n;
+		err = mddev->pers->check_reshape(mddev);
+		if (err) {
+			mddev->new_layout = mddev->layout;
+			return err;
+		}
+	} else {
+		mddev->new_layout = n;
+		if (mddev->reshape_position == MaxSector)
+			mddev->layout = n;
+	}
 	return len;
 }
 static struct md_sysfs_entry md_layout =
@@ -2515,7 +3656,7 @@ __ATTR(layout, S_IRUGO|S_IWUSR, layout_show, layout_store);
 
 
 static ssize_t
-raid_disks_show(mddev_t *mddev, char *page)
+raid_disks_show(struct mddev *mddev, char *page)
 {
 	if (mddev->raid_disks == 0)
 		return 0;
@@ -2526,10 +3667,10 @@ raid_disks_show(mddev_t *mddev, char *page)
 	return sprintf(page, "%d\n", mddev->raid_disks);
 }
 
-static int update_raid_disks(mddev_t *mddev, int raid_disks);
+static int update_raid_disks(struct mddev *mddev, int raid_disks);
 
 static ssize_t
-raid_disks_store(mddev_t *mddev, const char *buf, size_t len)
+raid_disks_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *e;
 	int rv = 0;
@@ -2541,9 +3682,20 @@ raid_disks_store(mddev_t *mddev, const char *buf, size_t len)
 	if (mddev->pers)
 		rv = update_raid_disks(mddev, n);
 	else if (mddev->reshape_position != MaxSector) {
+		struct md_rdev *rdev;
 		int olddisks = mddev->raid_disks - mddev->delta_disks;
+
+		rdev_for_each(rdev, mddev) {
+			if (olddisks < n &&
+			    rdev->data_offset < rdev->new_data_offset)
+				return -EINVAL;
+			if (olddisks > n &&
+			    rdev->data_offset > rdev->new_data_offset)
+				return -EINVAL;
+		}
 		mddev->delta_disks = n - olddisks;
 		mddev->raid_disks = n;
+		mddev->reshape_backwards = (mddev->delta_disks < 0);
 	} else
 		mddev->raid_disks = n;
 	return rv ? rv : len;
@@ -2552,54 +3704,71 @@ static struct md_sysfs_entry md_raid_disks =
 __ATTR(raid_disks, S_IRUGO|S_IWUSR, raid_disks_show, raid_disks_store);
 
 static ssize_t
-chunk_size_show(mddev_t *mddev, char *page)
+chunk_size_show(struct mddev *mddev, char *page)
 {
 	if (mddev->reshape_position != MaxSector &&
-	    mddev->chunk_size != mddev->new_chunk)
-		return sprintf(page, "%d (%d)\n", mddev->new_chunk,
-			       mddev->chunk_size);
-	return sprintf(page, "%d\n", mddev->chunk_size);
+	    mddev->chunk_sectors != mddev->new_chunk_sectors)
+		return sprintf(page, "%d (%d)\n",
+			       mddev->new_chunk_sectors << 9,
+			       mddev->chunk_sectors << 9);
+	return sprintf(page, "%d\n", mddev->chunk_sectors << 9);
 }
 
 static ssize_t
-chunk_size_store(mddev_t *mddev, const char *buf, size_t len)
+chunk_size_store(struct mddev *mddev, const char *buf, size_t len)
 {
-	/* can only set chunk_size if array is not yet active */
 	char *e;
 	unsigned long n = simple_strtoul(buf, &e, 10);
 
 	if (!*buf || (*e && *e != '\n'))
 		return -EINVAL;
 
-	if (mddev->pers)
-		return -EBUSY;
-	else if (mddev->reshape_position != MaxSector)
-		mddev->new_chunk = n;
-	else
-		mddev->chunk_size = n;
+	if (mddev->pers) {
+		int err;
+		if (mddev->pers->check_reshape == NULL)
+			return -EBUSY;
+		if (mddev->ro)
+			return -EROFS;
+		mddev->new_chunk_sectors = n >> 9;
+		err = mddev->pers->check_reshape(mddev);
+		if (err) {
+			mddev->new_chunk_sectors = mddev->chunk_sectors;
+			return err;
+		}
+	} else {
+		mddev->new_chunk_sectors = n >> 9;
+		if (mddev->reshape_position == MaxSector)
+			mddev->chunk_sectors = n >> 9;
+	}
 	return len;
 }
 static struct md_sysfs_entry md_chunk_size =
 __ATTR(chunk_size, S_IRUGO|S_IWUSR, chunk_size_show, chunk_size_store);
 
 static ssize_t
-resync_start_show(mddev_t *mddev, char *page)
+resync_start_show(struct mddev *mddev, char *page)
 {
+	if (mddev->recovery_cp == MaxSector)
+		return sprintf(page, "none\n");
 	return sprintf(page, "%llu\n", (unsigned long long)mddev->recovery_cp);
 }
 
 static ssize_t
-resync_start_store(mddev_t *mddev, const char *buf, size_t len)
+resync_start_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *e;
 	unsigned long long n = simple_strtoull(buf, &e, 10);
 
-	if (mddev->pers)
+	if (mddev->pers && !test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
 		return -EBUSY;
-	if (!*buf || (*e && *e != '\n'))
+	if (cmd_match(buf, "none"))
+		n = MaxSector;
+	else if (!*buf || (*e && *e != '\n'))
 		return -EINVAL;
 
 	mddev->recovery_cp = n;
+	if (mddev->pers)
+		set_bit(MD_CHANGE_CLEAN, &mddev->flags);
 	return len;
 }
 static struct md_sysfs_entry md_resync_start =
@@ -2657,7 +3826,7 @@ static int match_word(const char *word, char **list)
 }
 
 static ssize_t
-array_state_show(mddev_t *mddev, char *page)
+array_state_show(struct mddev *mddev, char *page)
 {
 	enum array_state st = inactive;
 
@@ -2672,7 +3841,7 @@ array_state_show(mddev_t *mddev, char *page)
 		case 0:
 			if (mddev->in_sync)
 				st = clean;
-			else if (test_bit(MD_CHANGE_CLEAN, &mddev->flags))
+			else if (test_bit(MD_CHANGE_PENDING, &mddev->flags))
 				st = write_pending;
 			else if (mddev->safemode)
 				st = active_idle;
@@ -2682,7 +3851,7 @@ array_state_show(mddev_t *mddev, char *page)
 	else {
 		if (list_empty(&mddev->disks) &&
 		    mddev->raid_disks == 0 &&
-		    mddev->size == 0)
+		    mddev->dev_sectors == 0)
 			st = clear;
 		else
 			st = inactive;
@@ -2690,12 +3859,13 @@ array_state_show(mddev_t *mddev, char *page)
 	return sprintf(page, "%s\n", array_states[st]);
 }
 
-static int do_md_stop(mddev_t * mddev, int ro, int is_open);
-static int do_md_run(mddev_t * mddev);
-static int restart_array(mddev_t *mddev);
+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);
 
 static ssize_t
-array_state_store(mddev_t *mddev, const char *buf, size_t len)
+array_state_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	int err = -EINVAL;
 	enum array_state st = match_word(buf, array_states);
@@ -2704,24 +3874,20 @@ array_state_store(mddev_t *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 = do_md_stop(mddev, 1, 0);
+			err = md_set_readonly(mddev, NULL);
 		else {
 			mddev->ro = 1;
 			set_disk_ro(mddev->gendisk, 1);
@@ -2730,9 +3896,9 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
 		break;
 	case read_auto:
 		if (mddev->pers) {
-			if (mddev->ro != 1)
-				err = do_md_stop(mddev, 1, 0);
-			else
+			if (mddev->ro == 0)
+				err = md_set_readonly(mddev, NULL);
+			else if (mddev->ro == 1)
 				err = restart_array(mddev);
 			if (err == 0) {
 				mddev->ro = 2;
@@ -2752,25 +3918,19 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
 					mddev->in_sync = 1;
 					if (mddev->safemode == 1)
 						mddev->safemode = 0;
-					if (mddev->persistent)
-						set_bit(MD_CHANGE_CLEAN,
-							&mddev->flags);
+					set_bit(MD_CHANGE_CLEAN, &mddev->flags);
 				}
 				err = 0;
 			} else
 				err = -EBUSY;
 			spin_unlock_irq(&mddev->write_lock);
-		} else {
-			mddev->ro = 0;
-			mddev->recovery_cp = MaxSector;
-			err = do_md_run(mddev);
-		}
+		} else
+			err = -EINVAL;
 		break;
 	case active:
 		if (mddev->pers) {
 			restart_array(mddev);
-			if (mddev->external)
-				clear_bit(MD_CHANGE_CLEAN, &mddev->flags);
+			clear_bit(MD_CHANGE_PENDING, &mddev->flags);
 			wake_up(&mddev->sb_wait);
 			err = 0;
 		} else {
@@ -2787,7 +3947,9 @@ array_state_store(mddev_t *mddev, const char *buf, size_t len)
 	if (err)
 		return err;
 	else {
-		sysfs_notify(&mddev->kobj, NULL, "array_state");
+		if (mddev->hold_active == UNTIL_IOCTL)
+			mddev->hold_active = 0;
+		sysfs_notify_dirent_safe(mddev->sysfs_state);
 		return len;
 	}
 }
@@ -2795,13 +3957,36 @@ static struct md_sysfs_entry md_array_state =
 __ATTR(array_state, S_IRUGO|S_IWUSR, array_state_show, array_state_store);
 
 static ssize_t
-null_show(mddev_t *mddev, char *page)
+max_corrected_read_errors_show(struct mddev *mddev, char *page) {
+	return sprintf(page, "%d\n",
+		       atomic_read(&mddev->max_corr_read_errors));
+}
+
+static ssize_t
+max_corrected_read_errors_store(struct mddev *mddev, const char *buf, size_t len)
 {
+	char *e;
+	unsigned long n = simple_strtoul(buf, &e, 10);
+
+	if (*buf && (*e == 0 || *e == '\n')) {
+		atomic_set(&mddev->max_corr_read_errors, n);
+		return len;
+	}
 	return -EINVAL;
 }
 
+static struct md_sysfs_entry max_corr_read_errors =
+__ATTR(max_read_errors, S_IRUGO|S_IWUSR, max_corrected_read_errors_show,
+	max_corrected_read_errors_store);
+
 static ssize_t
-new_dev_store(mddev_t *mddev, const char *buf, size_t len)
+null_show(struct mddev *mddev, char *page)
+{
+	return -EINVAL;
+}
+
+static ssize_t
+new_dev_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	/* buf must be %d:%d\n? giving major and minor numbers */
 	/* The new device is added to the array.
@@ -2814,7 +3999,7 @@ new_dev_store(mddev_t *mddev, const char *buf, size_t len)
 	int major = simple_strtoul(buf, &e, 10);
 	int minor;
 	dev_t dev;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 	int err;
 
 	if (!*buf || *e != ':' || !e[1] || e[1] == '\n')
@@ -2832,8 +4017,9 @@ new_dev_store(mddev_t *mddev, const char *buf, size_t len)
 		rdev = md_import_device(dev, mddev->major_version,
 					mddev->minor_version);
 		if (!IS_ERR(rdev) && !list_empty(&mddev->disks)) {
-			mdk_rdev_t *rdev0 = list_entry(mddev->disks.next,
-						       mdk_rdev_t, same_set);
+			struct md_rdev *rdev0
+				= list_entry(mddev->disks.next,
+					     struct md_rdev, same_set);
 			err = super_types[mddev->major_version]
 				.load_super(rdev, rdev0, mddev->minor_version);
 			if (err < 0)
@@ -2857,7 +4043,7 @@ static struct md_sysfs_entry md_new_device =
 __ATTR(new_dev, S_IWUSR, null_show, new_dev_store);
 
 static ssize_t
-bitmap_store(mddev_t *mddev, const char *buf, size_t len)
+bitmap_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *end;
 	unsigned long chunk, end_chunk;
@@ -2875,8 +4061,7 @@ bitmap_store(mddev_t *mddev, const char *buf, size_t len)
 		}
 		if (*end && !isspace(*end)) break;
 		bitmap_dirty_bits(mddev->bitmap, chunk, end_chunk);
-		buf = end;
-		while (isspace(*buf)) buf++;
+		buf = skip_spaces(end);
 	}
 	bitmap_unplug(mddev->bitmap); /* flush the bits to disk */
 out:
@@ -2887,34 +4072,33 @@ static struct md_sysfs_entry md_bitmap =
 __ATTR(bitmap_set_bits, S_IWUSR, null_show, bitmap_store);
 
 static ssize_t
-size_show(mddev_t *mddev, char *page)
+size_show(struct mddev *mddev, char *page)
 {
-	return sprintf(page, "%llu\n", (unsigned long long)mddev->size);
+	return sprintf(page, "%llu\n",
+		(unsigned long long)mddev->dev_sectors / 2);
 }
 
-static int update_size(mddev_t *mddev, sector_t num_sectors);
+static int update_size(struct mddev *mddev, sector_t num_sectors);
 
 static ssize_t
-size_store(mddev_t *mddev, const char *buf, size_t len)
+size_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	/* If array is inactive, we can reduce the component size, but
 	 * not increase it (except from 0).
 	 * If array is active, we can try an on-line resize
 	 */
-	char *e;
-	int err = 0;
-	unsigned long long size = simple_strtoull(buf, &e, 10);
-	if (!*buf || *buf == '\n' ||
-	    (*e && *e != '\n'))
-		return -EINVAL;
+	sector_t sectors;
+	int err = strict_blocks_to_sectors(buf, &sectors);
 
+	if (err < 0)
+		return err;
 	if (mddev->pers) {
-		err = update_size(mddev, size * 2);
+		err = update_size(mddev, sectors);
 		md_update_sb(mddev, 1);
 	} else {
-		if (mddev->size == 0 ||
-		    mddev->size > size)
-			mddev->size = size;
+		if (mddev->dev_sectors == 0 ||
+		    mddev->dev_sectors > sectors)
+			mddev->dev_sectors = sectors;
 		else
 			err = -ENOSPC;
 	}
@@ -2925,14 +4109,14 @@ static struct md_sysfs_entry md_size =
 __ATTR(component_size, S_IRUGO|S_IWUSR, size_show, size_store);
 
 
-/* Metdata version.
+/* Metadata version.
  * This is one of
  *   'none' for arrays with no metadata (good luck...)
  *   'external' for arrays with externally managed metadata,
  * or N.M for internally known formats
  */
 static ssize_t
-metadata_show(mddev_t *mddev, char *page)
+metadata_show(struct mddev *mddev, char *page)
 {
 	if (mddev->persistent)
 		return sprintf(page, "%d.%d\n",
@@ -2944,11 +4128,17 @@ metadata_show(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-metadata_store(mddev_t *mddev, const char *buf, size_t len)
+metadata_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	int major, minor;
 	char *e;
-	if (!list_empty(&mddev->disks))
+	/* Changing the details of 'external' metadata is
+	 * always permitted.  Otherwise there must be
+	 * no devices attached to the array.
+	 */
+	if (mddev->external && strncmp(buf, "external:", 9) == 0)
+		;
+	else if (!list_empty(&mddev->disks))
 		return -EBUSY;
 
 	if (cmd_match(buf, "none")) {
@@ -2992,10 +4182,12 @@ static struct md_sysfs_entry md_metadata =
 __ATTR(metadata_version, S_IRUGO|S_IWUSR, metadata_show, metadata_store);
 
 static ssize_t
-action_show(mddev_t *mddev, char *page)
+action_show(struct mddev *mddev, char *page)
 {
 	char *type = "idle";
-	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
+	if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
+		type = "frozen";
+	else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
 	    (!mddev->ro && test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))) {
 		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 			type = "reshape";
@@ -3013,17 +4205,20 @@ action_show(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-action_store(mddev_t *mddev, const char *page, size_t len)
+action_store(struct mddev *mddev, const char *page, size_t len)
 {
 	if (!mddev->pers || !mddev->pers->sync_request)
 		return -EINVAL;
 
-	if (cmd_match(page, "idle")) {
+	if (cmd_match(page, "frozen"))
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+	else
+		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+
+	if (cmd_match(page, "idle") || cmd_match(page, "frozen")) {
 		if (mddev->sync_thread) {
 			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-			md_unregister_thread(mddev->sync_thread);
-			mddev->sync_thread = NULL;
-			mddev->recovery = 0;
+			md_reap_sync_thread(mddev);
 		}
 	} else if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) ||
 		   test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
@@ -3049,34 +4244,49 @@ action_store(mddev_t *mddev, const char *page, size_t len)
 		set_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
 		set_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 	}
+	if (mddev->ro == 2) {
+		/* A write to sync_action is enough to justify
+		 * canceling read-auto mode
+		 */
+		mddev->ro = 0;
+		md_wakeup_thread(mddev->sync_thread);
+	}
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
-	sysfs_notify(&mddev->kobj, NULL, "sync_action");
+	sysfs_notify_dirent_safe(mddev->sysfs_action);
 	return len;
 }
 
+static struct md_sysfs_entry md_scan_mode =
+__ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store);
+
 static ssize_t
-mismatch_cnt_show(mddev_t *mddev, char *page)
+last_sync_action_show(struct mddev *mddev, char *page)
 {
-	return sprintf(page, "%llu\n",
-		       (unsigned long long) mddev->resync_mismatches);
+	return sprintf(page, "%s\n", mddev->last_sync_action);
 }
 
-static struct md_sysfs_entry md_scan_mode =
-__ATTR(sync_action, S_IRUGO|S_IWUSR, action_show, action_store);
+static struct md_sysfs_entry md_last_scan_mode = __ATTR_RO(last_sync_action);
 
+static ssize_t
+mismatch_cnt_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%llu\n",
+		       (unsigned long long)
+		       atomic64_read(&mddev->resync_mismatches));
+}
 
 static struct md_sysfs_entry md_mismatches = __ATTR_RO(mismatch_cnt);
 
 static ssize_t
-sync_min_show(mddev_t *mddev, char *page)
+sync_min_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%d (%s)\n", speed_min(mddev),
 		       mddev->sync_speed_min ? "local": "system");
 }
 
 static ssize_t
-sync_min_store(mddev_t *mddev, const char *buf, size_t len)
+sync_min_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	int min;
 	char *e;
@@ -3095,14 +4305,14 @@ static struct md_sysfs_entry md_sync_min =
 __ATTR(sync_speed_min, S_IRUGO|S_IWUSR, sync_min_show, sync_min_store);
 
 static ssize_t
-sync_max_show(mddev_t *mddev, char *page)
+sync_max_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%d (%s)\n", speed_max(mddev),
 		       mddev->sync_speed_max ? "local": "system");
 }
 
 static ssize_t
-sync_max_store(mddev_t *mddev, const char *buf, size_t len)
+sync_max_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	int max;
 	char *e;
@@ -3121,24 +4331,24 @@ static struct md_sysfs_entry md_sync_max =
 __ATTR(sync_speed_max, S_IRUGO|S_IWUSR, sync_max_show, sync_max_store);
 
 static ssize_t
-degraded_show(mddev_t *mddev, char *page)
+degraded_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%d\n", mddev->degraded);
 }
 static struct md_sysfs_entry md_degraded = __ATTR_RO(degraded);
 
 static ssize_t
-sync_force_parallel_show(mddev_t *mddev, char *page)
+sync_force_parallel_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%d\n", mddev->parallel_resync);
 }
 
 static ssize_t
-sync_force_parallel_store(mddev_t *mddev, const char *buf, size_t len)
+sync_force_parallel_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	long n;
 
-	if (strict_strtol(buf, 10, &n))
+	if (kstrtol(buf, 10, &n))
 		return -EINVAL;
 
 	if (n != 0 && n != 1)
@@ -3158,9 +4368,11 @@ __ATTR(sync_force_parallel, S_IRUGO|S_IWUSR,
        sync_force_parallel_show, sync_force_parallel_store);
 
 static ssize_t
-sync_speed_show(mddev_t *mddev, char *page)
+sync_speed_show(struct mddev *mddev, char *page)
 {
 	unsigned long resync, dt, db;
+	if (mddev->curr_resync == 0)
+		return sprintf(page, "none\n");
 	resync = mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active);
 	dt = (jiffies - mddev->resync_mark) / HZ;
 	if (!dt) dt++;
@@ -3171,32 +4383,40 @@ sync_speed_show(mddev_t *mddev, char *page)
 static struct md_sysfs_entry md_sync_speed = __ATTR_RO(sync_speed);
 
 static ssize_t
-sync_completed_show(mddev_t *mddev, char *page)
+sync_completed_show(struct mddev *mddev, char *page)
 {
-	unsigned long max_blocks, resync;
+	unsigned long long max_sectors, resync;
+
+	if (!test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
+		return sprintf(page, "none\n");
 
-	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
-		max_blocks = mddev->resync_max_sectors;
+	if (mddev->curr_resync == 1 ||
+	    mddev->curr_resync == 2)
+		return sprintf(page, "delayed\n");
+
+	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+		max_sectors = mddev->resync_max_sectors;
 	else
-		max_blocks = mddev->size << 1;
+		max_sectors = mddev->dev_sectors;
 
-	resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active));
-	return sprintf(page, "%lu / %lu\n", resync, max_blocks);
+	resync = mddev->curr_resync_completed;
+	return sprintf(page, "%llu / %llu\n", resync, max_sectors);
 }
 
 static struct md_sysfs_entry md_sync_completed = __ATTR_RO(sync_completed);
 
 static ssize_t
-min_sync_show(mddev_t *mddev, char *page)
+min_sync_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%llu\n",
 		       (unsigned long long)mddev->resync_min);
 }
 static ssize_t
-min_sync_store(mddev_t *mddev, const char *buf, size_t len)
+min_sync_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	unsigned long long min;
-	if (strict_strtoull(buf, 10, &min))
+	if (kstrtoull(buf, 10, &min))
 		return -EINVAL;
 	if (min > mddev->resync_max)
 		return -EINVAL;
@@ -3204,8 +4424,9 @@ min_sync_store(mddev_t *mddev, const char *buf, size_t len)
 		return -EBUSY;
 
 	/* Must be a multiple of chunk_size */
-	if (mddev->chunk_size) {
-		if (min & (sector_t)((mddev->chunk_size>>9)-1))
+	if (mddev->chunk_sectors) {
+		sector_t temp = min;
+		if (sector_div(temp, mddev->chunk_sectors))
 			return -EINVAL;
 	}
 	mddev->resync_min = min;
@@ -3217,7 +4438,7 @@ static struct md_sysfs_entry md_min_sync =
 __ATTR(sync_min, S_IRUGO|S_IWUSR, min_sync_show, min_sync_store);
 
 static ssize_t
-max_sync_show(mddev_t *mddev, char *page)
+max_sync_show(struct mddev *mddev, char *page)
 {
 	if (mddev->resync_max == MaxSector)
 		return sprintf(page, "max\n");
@@ -3226,23 +4447,25 @@ max_sync_show(mddev_t *mddev, char *page)
 			       (unsigned long long)mddev->resync_max);
 }
 static ssize_t
-max_sync_store(mddev_t *mddev, const char *buf, size_t len)
+max_sync_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	if (strncmp(buf, "max", 3) == 0)
 		mddev->resync_max = MaxSector;
 	else {
 		unsigned long long max;
-		if (strict_strtoull(buf, 10, &max))
+		if (kstrtoull(buf, 10, &max))
 			return -EINVAL;
 		if (max < mddev->resync_min)
 			return -EINVAL;
 		if (max < mddev->resync_max &&
+		    mddev->ro == 0 &&
 		    test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
 			return -EBUSY;
 
 		/* Must be a multiple of chunk_size */
-		if (mddev->chunk_size) {
-			if (max & (sector_t)((mddev->chunk_size>>9)-1))
+		if (mddev->chunk_sectors) {
+			sector_t temp = max;
+			if (sector_div(temp, mddev->chunk_sectors))
 				return -EINVAL;
 		}
 		mddev->resync_max = max;
@@ -3255,63 +4478,74 @@ static struct md_sysfs_entry md_max_sync =
 __ATTR(sync_max, S_IRUGO|S_IWUSR, max_sync_show, max_sync_store);
 
 static ssize_t
-suspend_lo_show(mddev_t *mddev, char *page)
+suspend_lo_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_lo);
 }
 
 static ssize_t
-suspend_lo_store(mddev_t *mddev, const char *buf, size_t len)
+suspend_lo_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *e;
 	unsigned long long new = simple_strtoull(buf, &e, 10);
+	unsigned long long old = mddev->suspend_lo;
 
-	if (mddev->pers->quiesce == NULL)
+	if (mddev->pers == NULL || 
+	    mddev->pers->quiesce == NULL)
 		return -EINVAL;
 	if (buf == e || (*e && *e != '\n'))
 		return -EINVAL;
-	if (new >= mddev->suspend_hi ||
-	    (new > mddev->suspend_lo && new < mddev->suspend_hi)) {
-		mddev->suspend_lo = new;
+
+	mddev->suspend_lo = new;
+	if (new >= old)
+		/* Shrinking suspended region */
 		mddev->pers->quiesce(mddev, 2);
-		return len;
-	} else
-		return -EINVAL;
+	else {
+		/* Expanding suspended region - need to wait */
+		mddev->pers->quiesce(mddev, 1);
+		mddev->pers->quiesce(mddev, 0);
+	}
+	return len;
 }
 static struct md_sysfs_entry md_suspend_lo =
 __ATTR(suspend_lo, S_IRUGO|S_IWUSR, suspend_lo_show, suspend_lo_store);
 
 
 static ssize_t
-suspend_hi_show(mddev_t *mddev, char *page)
+suspend_hi_show(struct mddev *mddev, char *page)
 {
 	return sprintf(page, "%llu\n", (unsigned long long)mddev->suspend_hi);
 }
 
 static ssize_t
-suspend_hi_store(mddev_t *mddev, const char *buf, size_t len)
+suspend_hi_store(struct mddev *mddev, const char *buf, size_t len)
 {
 	char *e;
 	unsigned long long new = simple_strtoull(buf, &e, 10);
+	unsigned long long old = mddev->suspend_hi;
 
-	if (mddev->pers->quiesce == NULL)
+	if (mddev->pers == NULL ||
+	    mddev->pers->quiesce == NULL)
 		return -EINVAL;
 	if (buf == e || (*e && *e != '\n'))
 		return -EINVAL;
-	if ((new <= mddev->suspend_lo && mddev->suspend_lo >= mddev->suspend_hi) ||
-	    (new > mddev->suspend_lo && new > mddev->suspend_hi)) {
-		mddev->suspend_hi = new;
+
+	mddev->suspend_hi = new;
+	if (new <= old)
+		/* Shrinking suspended region */
+		mddev->pers->quiesce(mddev, 2);
+	else {
+		/* Expanding suspended region - need to wait */
 		mddev->pers->quiesce(mddev, 1);
 		mddev->pers->quiesce(mddev, 0);
-		return len;
-	} else
-		return -EINVAL;
+	}
+	return len;
 }
 static struct md_sysfs_entry md_suspend_hi =
 __ATTR(suspend_hi, S_IRUGO|S_IWUSR, suspend_hi_show, suspend_hi_store);
 
 static ssize_t
-reshape_position_show(mddev_t *mddev, char *page)
+reshape_position_show(struct mddev *mddev, char *page)
 {
 	if (mddev->reshape_position != MaxSector)
 		return sprintf(page, "%llu\n",
@@ -3321,8 +4555,9 @@ reshape_position_show(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-reshape_position_store(mddev_t *mddev, const char *buf, size_t len)
+reshape_position_store(struct mddev *mddev, const char *buf, size_t len)
 {
+	struct md_rdev *rdev;
 	char *e;
 	unsigned long long new = simple_strtoull(buf, &e, 10);
 	if (mddev->pers)
@@ -3331,9 +4566,12 @@ reshape_position_store(mddev_t *mddev, const char *buf, size_t len)
 		return -EINVAL;
 	mddev->reshape_position = new;
 	mddev->delta_disks = 0;
+	mddev->reshape_backwards = 0;
 	mddev->new_level = mddev->level;
 	mddev->new_layout = mddev->layout;
-	mddev->new_chunk = mddev->chunk_size;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	rdev_for_each(rdev, mddev)
+		rdev->new_data_offset = rdev->data_offset;
 	return len;
 }
 
@@ -3341,6 +4579,84 @@ static struct md_sysfs_entry md_reshape_position =
 __ATTR(reshape_position, S_IRUGO|S_IWUSR, reshape_position_show,
        reshape_position_store);
 
+static ssize_t
+reshape_direction_show(struct mddev *mddev, char *page)
+{
+	return sprintf(page, "%s\n",
+		       mddev->reshape_backwards ? "backwards" : "forwards");
+}
+
+static ssize_t
+reshape_direction_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	int backwards = 0;
+	if (cmd_match(buf, "forwards"))
+		backwards = 0;
+	else if (cmd_match(buf, "backwards"))
+		backwards = 1;
+	else
+		return -EINVAL;
+	if (mddev->reshape_backwards == backwards)
+		return len;
+
+	/* check if we are allowed to change */
+	if (mddev->delta_disks)
+		return -EBUSY;
+
+	if (mddev->persistent &&
+	    mddev->major_version == 0)
+		return -EINVAL;
+
+	mddev->reshape_backwards = backwards;
+	return len;
+}
+
+static struct md_sysfs_entry md_reshape_direction =
+__ATTR(reshape_direction, S_IRUGO|S_IWUSR, reshape_direction_show,
+       reshape_direction_store);
+
+static ssize_t
+array_size_show(struct mddev *mddev, char *page)
+{
+	if (mddev->external_size)
+		return sprintf(page, "%llu\n",
+			       (unsigned long long)mddev->array_sectors/2);
+	else
+		return sprintf(page, "default\n");
+}
+
+static ssize_t
+array_size_store(struct mddev *mddev, const char *buf, size_t len)
+{
+	sector_t sectors;
+
+	if (strncmp(buf, "default", 7) == 0) {
+		if (mddev->pers)
+			sectors = mddev->pers->size(mddev, 0, 0);
+		else
+			sectors = mddev->array_sectors;
+
+		mddev->external_size = 0;
+	} else {
+		if (strict_blocks_to_sectors(buf, &sectors) < 0)
+			return -EINVAL;
+		if (mddev->pers && mddev->pers->size(mddev, 0, 0) < sectors)
+			return -E2BIG;
+
+		mddev->external_size = 1;
+	}
+
+	mddev->array_sectors = sectors;
+	if (mddev->pers) {
+		set_capacity(mddev->gendisk, mddev->array_sectors);
+		revalidate_disk(mddev->gendisk);
+	}
+	return len;
+}
+
+static struct md_sysfs_entry md_array_size =
+__ATTR(array_size, S_IRUGO|S_IWUSR, array_size_show,
+       array_size_store);
 
 static struct attribute *md_default_attrs[] = {
 	&md_level.attr,
@@ -3354,11 +4670,15 @@ static struct attribute *md_default_attrs[] = {
 	&md_safe_delay.attr,
 	&md_array_state.attr,
 	&md_reshape_position.attr,
+	&md_reshape_direction.attr,
+	&md_array_size.attr,
+	&max_corr_read_errors.attr,
 	NULL,
 };
 
 static struct attribute *md_redundancy_attrs[] = {
 	&md_scan_mode.attr,
+	&md_last_scan_mode.attr,
 	&md_mismatches.attr,
 	&md_sync_min.attr,
 	&md_sync_max.attr,
@@ -3383,16 +4703,25 @@ static ssize_t
 md_attr_show(struct kobject *kobj, struct attribute *attr, char *page)
 {
 	struct md_sysfs_entry *entry = container_of(attr, struct md_sysfs_entry, attr);
-	mddev_t *mddev = container_of(kobj, struct mddev_s, kobj);
+	struct mddev *mddev = container_of(kobj, struct mddev, kobj);
 	ssize_t rv;
 
 	if (!entry->show)
 		return -EIO;
+	spin_lock(&all_mddevs_lock);
+	if (list_empty(&mddev->all_mddevs)) {
+		spin_unlock(&all_mddevs_lock);
+		return -EBUSY;
+	}
+	mddev_get(mddev);
+	spin_unlock(&all_mddevs_lock);
+
 	rv = mddev_lock(mddev);
 	if (!rv) {
 		rv = entry->show(mddev, page);
 		mddev_unlock(mddev);
 	}
+	mddev_put(mddev);
 	return rv;
 }
 
@@ -3401,28 +4730,49 @@ md_attr_store(struct kobject *kobj, struct attribute *attr,
 	      const char *page, size_t length)
 {
 	struct md_sysfs_entry *entry = container_of(attr, struct md_sysfs_entry, attr);
-	mddev_t *mddev = container_of(kobj, struct mddev_s, kobj);
+	struct mddev *mddev = container_of(kobj, struct mddev, kobj);
 	ssize_t rv;
 
 	if (!entry->store)
 		return -EIO;
 	if (!capable(CAP_SYS_ADMIN))
 		return -EACCES;
+	spin_lock(&all_mddevs_lock);
+	if (list_empty(&mddev->all_mddevs)) {
+		spin_unlock(&all_mddevs_lock);
+		return -EBUSY;
+	}
+	mddev_get(mddev);
+	spin_unlock(&all_mddevs_lock);
+	if (entry->store == new_dev_store)
+		flush_workqueue(md_misc_wq);
 	rv = mddev_lock(mddev);
 	if (!rv) {
 		rv = entry->store(mddev, page, length);
 		mddev_unlock(mddev);
 	}
+	mddev_put(mddev);
 	return rv;
 }
 
 static void md_free(struct kobject *ko)
 {
-	mddev_t *mddev = container_of(ko, mddev_t, kobj);
+	struct mddev *mddev = container_of(ko, struct mddev, kobj);
+
+	if (mddev->sysfs_state)
+		sysfs_put(mddev->sysfs_state);
+
+	if (mddev->gendisk) {
+		del_gendisk(mddev->gendisk);
+		put_disk(mddev->gendisk);
+	}
+	if (mddev->queue)
+		blk_cleanup_queue(mddev->queue);
+
 	kfree(mddev);
 }
 
-static struct sysfs_ops md_sysfs_ops = {
+static const struct sysfs_ops md_sysfs_ops = {
 	.show	= md_attr_show,
 	.store	= md_attr_store,
 };
@@ -3434,76 +4784,164 @@ static struct kobj_type md_ktype = {
 
 int mdp_major = 0;
 
-static struct kobject *md_probe(dev_t dev, int *part, void *data)
+static void mddev_delayed_delete(struct work_struct *ws)
+{
+	struct mddev *mddev = container_of(ws, struct mddev, del_work);
+
+	sysfs_remove_group(&mddev->kobj, &md_bitmap_group);
+	kobject_del(&mddev->kobj);
+	kobject_put(&mddev->kobj);
+}
+
+static int md_alloc(dev_t dev, char *name)
 {
 	static DEFINE_MUTEX(disks_mutex);
-	mddev_t *mddev = mddev_find(dev);
+	struct mddev *mddev = mddev_find(dev);
 	struct gendisk *disk;
-	int partitioned = (MAJOR(dev) != MD_MAJOR);
-	int shift = partitioned ? MdpMinorShift : 0;
-	int unit = MINOR(dev) >> shift;
+	int partitioned;
+	int shift;
+	int unit;
 	int error;
 
 	if (!mddev)
-		return NULL;
+		return -ENODEV;
+
+	partitioned = (MAJOR(mddev->unit) != MD_MAJOR);
+	shift = partitioned ? MdpMinorShift : 0;
+	unit = MINOR(mddev->unit) >> shift;
+
+	/* wait for any previous instance of this device to be
+	 * completely removed (mddev_delayed_delete).
+	 */
+	flush_workqueue(md_misc_wq);
 
 	mutex_lock(&disks_mutex);
-	if (mddev->gendisk) {
-		mutex_unlock(&disks_mutex);
-		mddev_put(mddev);
-		return NULL;
+	error = -EEXIST;
+	if (mddev->gendisk)
+		goto abort;
+
+	if (name) {
+		/* Need to ensure that 'name' is not a duplicate.
+		 */
+		struct mddev *mddev2;
+		spin_lock(&all_mddevs_lock);
+
+		list_for_each_entry(mddev2, &all_mddevs, all_mddevs)
+			if (mddev2->gendisk &&
+			    strcmp(mddev2->gendisk->disk_name, name) == 0) {
+				spin_unlock(&all_mddevs_lock);
+				goto abort;
+			}
+		spin_unlock(&all_mddevs_lock);
 	}
+
+	error = -ENOMEM;
+	mddev->queue = blk_alloc_queue(GFP_KERNEL);
+	if (!mddev->queue)
+		goto abort;
+	mddev->queue->queuedata = mddev;
+
+	blk_queue_make_request(mddev->queue, md_make_request);
+	blk_set_stacking_limits(&mddev->queue->limits);
+
 	disk = alloc_disk(1 << shift);
 	if (!disk) {
-		mutex_unlock(&disks_mutex);
-		mddev_put(mddev);
-		return NULL;
+		blk_cleanup_queue(mddev->queue);
+		mddev->queue = NULL;
+		goto abort;
 	}
-	disk->major = MAJOR(dev);
+	disk->major = MAJOR(mddev->unit);
 	disk->first_minor = unit << shift;
-	if (partitioned)
+	if (name)
+		strcpy(disk->disk_name, name);
+	else if (partitioned)
 		sprintf(disk->disk_name, "md_d%d", unit);
 	else
 		sprintf(disk->disk_name, "md%d", unit);
 	disk->fops = &md_fops;
 	disk->private_data = mddev;
 	disk->queue = mddev->queue;
-	add_disk(disk);
+	blk_queue_flush(mddev->queue, REQ_FLUSH | REQ_FUA);
+	/* Allow extended partitions.  This makes the
+	 * 'mdp' device redundant, but we can't really
+	 * remove it now.
+	 */
+	disk->flags |= GENHD_FL_EXT_DEVT;
 	mddev->gendisk = disk;
-	error = kobject_init_and_add(&mddev->kobj, &md_ktype, &disk->dev.kobj,
-				     "%s", "md");
-	mutex_unlock(&disks_mutex);
-	if (error)
+	/* As soon as we call add_disk(), another thread could get
+	 * through to md_open, so make sure it doesn't get too far
+	 */
+	mutex_lock(&mddev->open_mutex);
+	add_disk(disk);
+
+	error = kobject_init_and_add(&mddev->kobj, &md_ktype,
+				     &disk_to_dev(disk)->kobj, "%s", "md");
+	if (error) {
+		/* This isn't possible, but as kobject_init_and_add is marked
+		 * __must_check, we must do something with the result
+		 */
 		printk(KERN_WARNING "md: cannot register %s/md - name in use\n",
 		       disk->disk_name);
-	else
+		error = 0;
+	}
+	if (mddev->kobj.sd &&
+	    sysfs_create_group(&mddev->kobj, &md_bitmap_group))
+		printk(KERN_DEBUG "pointless warning\n");
+	mutex_unlock(&mddev->open_mutex);
+ abort:
+	mutex_unlock(&disks_mutex);
+	if (!error && mddev->kobj.sd) {
 		kobject_uevent(&mddev->kobj, KOBJ_ADD);
+		mddev->sysfs_state = sysfs_get_dirent_safe(mddev->kobj.sd, "array_state");
+	}
+	mddev_put(mddev);
+	return error;
+}
+
+static struct kobject *md_probe(dev_t dev, int *part, void *data)
+{
+	md_alloc(dev, NULL);
 	return NULL;
 }
 
+static int add_named_array(const char *val, struct kernel_param *kp)
+{
+	/* val must be "md_*" where * is not all digits.
+	 * We allocate an array with a large free minor number, and
+	 * set the name to val.  val must not already be an active name.
+	 */
+	int len = strlen(val);
+	char buf[DISK_NAME_LEN];
+
+	while (len && val[len-1] == '\n')
+		len--;
+	if (len >= DISK_NAME_LEN)
+		return -E2BIG;
+	strlcpy(buf, val, len+1);
+	if (strncmp(buf, "md_", 3) != 0)
+		return -EINVAL;
+	return md_alloc(0, buf);
+}
+
 static void md_safemode_timeout(unsigned long data)
 {
-	mddev_t *mddev = (mddev_t *) data;
+	struct mddev *mddev = (struct mddev *) data;
 
 	if (!atomic_read(&mddev->writes_pending)) {
 		mddev->safemode = 1;
 		if (mddev->external)
-			set_bit(MD_NOTIFY_ARRAY_STATE, &mddev->flags);
+			sysfs_notify_dirent_safe(mddev->sysfs_state);
 	}
 	md_wakeup_thread(mddev->thread);
 }
 
 static int start_dirty_degraded;
 
-static int do_md_run(mddev_t * mddev)
+int md_run(struct mddev *mddev)
 {
 	int err;
-	int chunk_size;
-	struct list_head *tmp;
-	mdk_rdev_t *rdev;
-	struct gendisk *disk;
-	struct mdk_personality *pers;
-	char b[BDEVNAME_SIZE];
+	struct md_rdev *rdev;
+	struct md_personality *pers;
 
 	if (list_empty(&mddev->disks))
 		/* cannot run an array with no devices.. */
@@ -3511,6 +4949,9 @@ static int do_md_run(mddev_t * mddev)
 
 	if (mddev->pers)
 		return -EBUSY;
+	/* Cannot run until previous stop completes properly */
+	if (mddev->sysfs_active)
+		return -EBUSY;
 
 	/*
 	 * Analyze all RAID superblock(s)
@@ -3521,56 +4962,17 @@ static int do_md_run(mddev_t * mddev)
 		analyze_sbs(mddev);
 	}
 
-	chunk_size = mddev->chunk_size;
-
-	if (chunk_size) {
-		if (chunk_size > MAX_CHUNK_SIZE) {
-			printk(KERN_ERR "too big chunk_size: %d > %d\n",
-				chunk_size, MAX_CHUNK_SIZE);
-			return -EINVAL;
-		}
-		/*
-		 * chunk-size has to be a power of 2 and multiples of PAGE_SIZE
-		 */
-		if ( (1 << ffz(~chunk_size)) != chunk_size) {
-			printk(KERN_ERR "chunk_size of %d not valid\n", chunk_size);
-			return -EINVAL;
-		}
-		if (chunk_size < PAGE_SIZE) {
-			printk(KERN_ERR "too small chunk_size: %d < %ld\n",
-				chunk_size, PAGE_SIZE);
-			return -EINVAL;
-		}
-
-		/* devices must have minimum size of one chunk */
-		rdev_for_each(rdev, tmp, mddev) {
-			if (test_bit(Faulty, &rdev->flags))
-				continue;
-			if (rdev->size < chunk_size / 1024) {
-				printk(KERN_WARNING
-					"md: Dev %s smaller than chunk_size:"
-					" %lluk < %dk\n",
-					bdevname(rdev->bdev,b),
-					(unsigned long long)rdev->size,
-					chunk_size / 1024);
-				return -EINVAL;
-			}
-		}
-	}
-
-#ifdef CONFIG_KMOD
 	if (mddev->level != LEVEL_NONE)
 		request_module("md-level-%d", mddev->level);
 	else if (mddev->clevel[0])
 		request_module("md-%s", mddev->clevel);
-#endif
 
 	/*
 	 * Drop all container device buffers, from now on
 	 * the only valid external interface is through the md
 	 * device.
 	 */
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev) {
 		if (test_bit(Faulty, &rdev->flags))
 			continue;
 		sync_blockdev(rdev->bdev);
@@ -3578,11 +4980,13 @@ static int do_md_run(mddev_t * mddev)
 
 		/* perform some consistency tests on the device.
 		 * We don't want the data to overlap the metadata,
-		 * Internal Bitmap issues has handled elsewhere.
+		 * Internal Bitmap issues have been handled elsewhere.
 		 */
-		if (rdev->data_offset < rdev->sb_start) {
-			if (mddev->size &&
-			    rdev->data_offset + mddev->size*2
+		if (rdev->meta_bdev) {
+			/* Nothing to check */;
+		} else if (rdev->data_offset < rdev->sb_start) {
+			if (mddev->dev_sectors &&
+			    rdev->data_offset + mddev->dev_sectors
 			    > rdev->sb_start) {
 				printk("md: %s: data overlaps metadata\n",
 				       mdname(mddev));
@@ -3596,13 +5000,11 @@ static int do_md_run(mddev_t * mddev)
 				return -EINVAL;
 			}
 		}
-		sysfs_notify(&rdev->kobj, NULL, "state");
+		sysfs_notify_dirent_safe(rdev->sysfs_state);
 	}
 
-	md_probe(mddev->unit, NULL, NULL);
-	disk = mddev->gendisk;
-	if (!disk)
-		return -ENOMEM;
+	if (mddev->bio_set == NULL)
+		mddev->bio_set = bioset_create(BIO_POOL_SIZE, 0);
 
 	spin_lock(&pers_lock);
 	pers = find_pers(mddev->level, mddev->clevel);
@@ -3618,7 +5020,10 @@ static int do_md_run(mddev_t * mddev)
 	}
 	mddev->pers = pers;
 	spin_unlock(&pers_lock);
-	mddev->level = pers->level;
+	if (mddev->level != pers->level) {
+		mddev->level = pers->level;
+		mddev->new_level = pers->level;
+	}
 	strlcpy(mddev->clevel, pers->name, sizeof(mddev->clevel));
 
 	if (mddev->reshape_position != MaxSector &&
@@ -3634,11 +5039,11 @@ static int do_md_run(mddev_t * mddev)
 		 * configuration.
 		 */
 		char b[BDEVNAME_SIZE], b2[BDEVNAME_SIZE];
-		mdk_rdev_t *rdev2;
-		struct list_head *tmp2;
+		struct md_rdev *rdev2;
 		int warned = 0;
-		rdev_for_each(rdev, tmp, mddev) {
-			rdev_for_each(rdev2, tmp2, mddev) {
+
+		rdev_for_each(rdev, mddev)
+			rdev_for_each(rdev2, mddev) {
 				if (rdev < rdev2 &&
 				    rdev->bdev->bd_contains ==
 				    rdev2->bdev->bd_contains) {
@@ -3652,7 +5057,7 @@ static int do_md_run(mddev_t * mddev)
 					warned = 1;
 				}
 			}
-		}
+
 		if (warned)
 			printk(KERN_WARNING
 			       "True protection against single-disk"
@@ -3660,17 +5065,29 @@ static int do_md_run(mddev_t * mddev)
 	}
 
 	mddev->recovery = 0;
-	mddev->resync_max_sectors = mddev->size << 1; /* may be over-ridden by personality */
-	mddev->barriers_work = 1;
+	/* may be over-ridden by personality */
+	mddev->resync_max_sectors = mddev->dev_sectors;
+
 	mddev->ok_start_degraded = start_dirty_degraded;
 
-	if (start_readonly)
+	if (start_readonly && mddev->ro == 0)
 		mddev->ro = 2; /* read-only, but switch on first write */
 
 	err = mddev->pers->run(mddev);
 	if (err)
 		printk(KERN_ERR "md: pers->run() failed ...\n");
-	else if (mddev->pers->sync_request) {
+	else if (mddev->pers->size(mddev, 0, 0) < mddev->array_sectors) {
+		WARN_ONCE(!mddev->external_size, "%s: default size too small,"
+			  " but 'external_size' not in effect?\n", __func__);
+		printk(KERN_ERR
+		       "md: invalid array_size %llu > default size %llu\n",
+		       (unsigned long long)mddev->array_sectors / 2,
+		       (unsigned long long)mddev->pers->size(mddev, 0, 0) / 2);
+		err = -EINVAL;
+		mddev->pers->stop(mddev);
+	}
+	if (err == 0 && mddev->pers->sync_request &&
+	    (mddev->bitmap_info.file || mddev->bitmap_info.offset)) {
 		err = bitmap_create(mddev);
 		if (err) {
 			printk(KERN_ERR "%s: failed to create bitmap (%d)\n",
@@ -3685,87 +5102,68 @@ static int do_md_run(mddev_t * mddev)
 		return err;
 	}
 	if (mddev->pers->sync_request) {
-		if (sysfs_create_group(&mddev->kobj, &md_redundancy_group))
+		if (mddev->kobj.sd &&
+		    sysfs_create_group(&mddev->kobj, &md_redundancy_group))
 			printk(KERN_WARNING
 			       "md: cannot register extra attributes for %s\n",
 			       mdname(mddev));
+		mddev->sysfs_action = sysfs_get_dirent_safe(mddev->kobj.sd, "sync_action");
 	} else if (mddev->ro == 2) /* auto-readonly not meaningful */
 		mddev->ro = 0;
 
  	atomic_set(&mddev->writes_pending,0);
+	atomic_set(&mddev->max_corr_read_errors,
+		   MD_DEFAULT_MAX_CORRECTED_READ_ERRORS);
 	mddev->safemode = 0;
 	mddev->safemode_timer.function = md_safemode_timeout;
 	mddev->safemode_timer.data = (unsigned long) mddev;
 	mddev->safemode_delay = (200 * HZ)/1000 +1; /* 200 msec delay */
 	mddev->in_sync = 1;
-
-	rdev_for_each(rdev, tmp, mddev)
-		if (rdev->raid_disk >= 0) {
-			char nm[20];
-			sprintf(nm, "rd%d", rdev->raid_disk);
-			if (sysfs_create_link(&mddev->kobj, &rdev->kobj, nm))
-				printk("md: cannot register %s for %s\n",
-				       nm, mdname(mddev));
-		}
+	smp_wmb();
+	mddev->ready = 1;
+	rdev_for_each(rdev, mddev)
+		if (rdev->raid_disk >= 0)
+			if (sysfs_link_rdev(mddev, rdev))
+				/* failure here is OK */;
 	
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	
-	if (mddev->flags)
+	if (mddev->flags & MD_UPDATE_SB_FLAGS)
 		md_update_sb(mddev, 0);
 
-	set_capacity(disk, mddev->array_sectors);
+	md_new_event(mddev);
+	sysfs_notify_dirent_safe(mddev->sysfs_state);
+	sysfs_notify_dirent_safe(mddev->sysfs_action);
+	sysfs_notify(&mddev->kobj, NULL, "degraded");
+	return 0;
+}
+EXPORT_SYMBOL_GPL(md_run);
 
-	/* If we call blk_queue_make_request here, it will
-	 * re-initialise max_sectors etc which may have been
-	 * refined inside -> run.  So just set the bits we need to set.
-	 * Most initialisation happended when we called
-	 * blk_queue_make_request(..., md_fail_request)
-	 * earlier.
-	 */
-	mddev->queue->queuedata = mddev;
-	mddev->queue->make_request_fn = mddev->pers->make_request;
+static int do_md_run(struct mddev *mddev)
+{
+	int err;
 
-	/* If there is a partially-recovered drive we need to
-	 * start recovery here.  If we leave it to md_check_recovery,
-	 * it will remove the drives and not do the right thing
-	 */
-	if (mddev->degraded && !mddev->sync_thread) {
-		struct list_head *rtmp;
-		int spares = 0;
-		rdev_for_each(rdev, rtmp, mddev)
-			if (rdev->raid_disk >= 0 &&
-			    !test_bit(In_sync, &rdev->flags) &&
-			    !test_bit(Faulty, &rdev->flags))
-				/* complete an interrupted recovery */
-				spares++;
-		if (spares && mddev->pers->sync_request) {
-			mddev->recovery = 0;
-			set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
-			mddev->sync_thread = md_register_thread(md_do_sync,
-								mddev,
-								"%s_resync");
-			if (!mddev->sync_thread) {
-				printk(KERN_ERR "%s: could not start resync"
-				       " thread...\n",
-				       mdname(mddev));
-				/* leave the spares where they are, it shouldn't hurt */
-				mddev->recovery = 0;
-			}
-		}
+	err = md_run(mddev);
+	if (err)
+		goto out;
+	err = bitmap_load(mddev);
+	if (err) {
+		bitmap_destroy(mddev);
+		goto out;
 	}
+
 	md_wakeup_thread(mddev->thread);
 	md_wakeup_thread(mddev->sync_thread); /* possibly kick off a reshape */
 
+	set_capacity(mddev->gendisk, mddev->array_sectors);
+	revalidate_disk(mddev->gendisk);
 	mddev->changed = 1;
-	md_new_event(mddev);
-	sysfs_notify(&mddev->kobj, NULL, "array_state");
-	sysfs_notify(&mddev->kobj, NULL, "sync_action");
-	sysfs_notify(&mddev->kobj, NULL, "degraded");
-	kobject_uevent(&mddev->gendisk->dev.kobj, KOBJ_CHANGE);
-	return 0;
+	kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
+out:
+	return err;
 }
 
-static int restart_array(mddev_t *mddev)
+static int restart_array(struct mddev *mddev)
 {
 	struct gendisk *disk = mddev->gendisk;
 
@@ -3785,182 +5183,252 @@ static int restart_array(mddev_t *mddev)
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
 	md_wakeup_thread(mddev->sync_thread);
-	sysfs_notify(&mddev->kobj, NULL, "array_state");
+	sysfs_notify_dirent_safe(mddev->sysfs_state);
 	return 0;
 }
 
-/* similar to deny_write_access, but accounts for our holding a reference
- * to the file ourselves */
-static int deny_bitmap_write_access(struct file * file)
+static void md_clean(struct mddev *mddev)
 {
-	struct inode *inode = file->f_mapping->host;
+	mddev->array_sectors = 0;
+	mddev->external_size = 0;
+	mddev->dev_sectors = 0;
+	mddev->raid_disks = 0;
+	mddev->recovery_cp = 0;
+	mddev->resync_min = 0;
+	mddev->resync_max = MaxSector;
+	mddev->reshape_position = MaxSector;
+	mddev->external = 0;
+	mddev->persistent = 0;
+	mddev->level = LEVEL_NONE;
+	mddev->clevel[0] = 0;
+	mddev->flags = 0;
+	mddev->ro = 0;
+	mddev->metadata_type[0] = 0;
+	mddev->chunk_sectors = 0;
+	mddev->ctime = mddev->utime = 0;
+	mddev->layout = 0;
+	mddev->max_disks = 0;
+	mddev->events = 0;
+	mddev->can_decrease_events = 0;
+	mddev->delta_disks = 0;
+	mddev->reshape_backwards = 0;
+	mddev->new_level = LEVEL_NONE;
+	mddev->new_layout = 0;
+	mddev->new_chunk_sectors = 0;
+	mddev->curr_resync = 0;
+	atomic64_set(&mddev->resync_mismatches, 0);
+	mddev->suspend_lo = mddev->suspend_hi = 0;
+	mddev->sync_speed_min = mddev->sync_speed_max = 0;
+	mddev->recovery = 0;
+	mddev->in_sync = 0;
+	mddev->changed = 0;
+	mddev->degraded = 0;
+	mddev->safemode = 0;
+	mddev->merge_check_needed = 0;
+	mddev->bitmap_info.offset = 0;
+	mddev->bitmap_info.default_offset = 0;
+	mddev->bitmap_info.default_space = 0;
+	mddev->bitmap_info.chunksize = 0;
+	mddev->bitmap_info.daemon_sleep = 0;
+	mddev->bitmap_info.max_write_behind = 0;
+}
 
-	spin_lock(&inode->i_lock);
-	if (atomic_read(&inode->i_writecount) > 1) {
-		spin_unlock(&inode->i_lock);
-		return -ETXTBSY;
+static void __md_stop_writes(struct mddev *mddev)
+{
+	set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		md_reap_sync_thread(mddev);
 	}
-	atomic_set(&inode->i_writecount, -1);
-	spin_unlock(&inode->i_lock);
 
-	return 0;
+	del_timer_sync(&mddev->safemode_timer);
+
+	bitmap_flush(mddev);
+	md_super_wait(mddev);
+
+	if (mddev->ro == 0 &&
+	    (!mddev->in_sync || (mddev->flags & MD_UPDATE_SB_FLAGS))) {
+		/* mark array as shutdown cleanly */
+		mddev->in_sync = 1;
+		md_update_sb(mddev, 1);
+	}
+}
+
+void md_stop_writes(struct mddev *mddev)
+{
+	mddev_lock_nointr(mddev);
+	__md_stop_writes(mddev);
+	mddev_unlock(mddev);
 }
+EXPORT_SYMBOL_GPL(md_stop_writes);
 
-static void restore_bitmap_write_access(struct file *file)
+static void __md_stop(struct mddev *mddev)
 {
-	struct inode *inode = file->f_mapping->host;
+	mddev->ready = 0;
+	mddev->pers->stop(mddev);
+	if (mddev->pers->sync_request && mddev->to_remove == NULL)
+		mddev->to_remove = &md_redundancy_group;
+	module_put(mddev->pers->owner);
+	mddev->pers = NULL;
+	clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+}
 
-	spin_lock(&inode->i_lock);
-	atomic_set(&inode->i_writecount, 1);
-	spin_unlock(&inode->i_lock);
+void md_stop(struct mddev *mddev)
+{
+	/* stop the array and free an attached data structures.
+	 * This is called from dm-raid
+	 */
+	__md_stop(mddev);
+	bitmap_destroy(mddev);
+	if (mddev->bio_set)
+		bioset_free(mddev->bio_set);
+}
+
+EXPORT_SYMBOL_GPL(md_stop);
+
+static int md_set_readonly(struct mddev *mddev, struct block_device *bdev)
+{
+	int err = 0;
+	int did_freeze = 0;
+
+	if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+		did_freeze = 1;
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		md_wakeup_thread(mddev->thread);
+	}
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		/* Thread might be blocked waiting for metadata update
+		 * which will now never happen */
+		wake_up_process(mddev->sync_thread->tsk);
+	}
+	mddev_unlock(mddev);
+	wait_event(resync_wait, mddev->sync_thread == NULL);
+	mddev_lock_nointr(mddev);
+
+	mutex_lock(&mddev->open_mutex);
+	if (atomic_read(&mddev->openers) > !!bdev ||
+	    mddev->sync_thread ||
+	    (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
+		printk("md: %s still in use.\n",mdname(mddev));
+		if (did_freeze) {
+			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+			md_wakeup_thread(mddev->thread);
+		}
+		err = -EBUSY;
+		goto out;
+	}
+	if (mddev->pers) {
+		__md_stop_writes(mddev);
+
+		err  = -ENXIO;
+		if (mddev->ro==1)
+			goto out;
+		mddev->ro = 1;
+		set_disk_ro(mddev->gendisk, 1);
+		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		sysfs_notify_dirent_safe(mddev->sysfs_state);
+		err = 0;
+	}
+out:
+	mutex_unlock(&mddev->open_mutex);
+	return err;
 }
 
 /* mode:
  *   0 - completely stop and dis-assemble array
- *   1 - switch to readonly
  *   2 - stop but do not disassemble array
  */
-static int do_md_stop(mddev_t * mddev, int mode, int is_open)
+static int do_md_stop(struct mddev * mddev, int mode,
+		      struct block_device *bdev)
 {
-	int err = 0;
 	struct gendisk *disk = mddev->gendisk;
+	struct md_rdev *rdev;
+	int did_freeze = 0;
 
-	if (atomic_read(&mddev->openers) > is_open) {
+	if (!test_bit(MD_RECOVERY_FROZEN, &mddev->recovery)) {
+		did_freeze = 1;
+		set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+		md_wakeup_thread(mddev->thread);
+	}
+	if (mddev->sync_thread) {
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		/* Thread might be blocked waiting for metadata update
+		 * which will now never happen */
+		wake_up_process(mddev->sync_thread->tsk);
+	}
+	mddev_unlock(mddev);
+	wait_event(resync_wait, mddev->sync_thread == NULL);
+	mddev_lock_nointr(mddev);
+
+	mutex_lock(&mddev->open_mutex);
+	if (atomic_read(&mddev->openers) > !!bdev ||
+	    mddev->sysfs_active ||
+	    mddev->sync_thread ||
+	    (bdev && !test_bit(MD_STILL_CLOSED, &mddev->flags))) {
 		printk("md: %s still in use.\n",mdname(mddev));
+		mutex_unlock(&mddev->open_mutex);
+		if (did_freeze) {
+			clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
+			md_wakeup_thread(mddev->thread);
+		}
 		return -EBUSY;
 	}
-
 	if (mddev->pers) {
+		if (mddev->ro)
+			set_disk_ro(disk, 0);
 
-		if (mddev->sync_thread) {
-			set_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
-			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-			md_unregister_thread(mddev->sync_thread);
-			mddev->sync_thread = NULL;
-		}
-
-		del_timer_sync(&mddev->safemode_timer);
-
-		switch(mode) {
-		case 1: /* readonly */
-			err  = -ENXIO;
-			if (mddev->ro==1)
-				goto out;
-			mddev->ro = 1;
-			break;
-		case 0: /* disassemble */
-		case 2: /* stop */
-			bitmap_flush(mddev);
-			md_super_wait(mddev);
-			if (mddev->ro)
-				set_disk_ro(disk, 0);
-			blk_queue_make_request(mddev->queue, md_fail_request);
-			mddev->pers->stop(mddev);
-			mddev->queue->merge_bvec_fn = NULL;
-			mddev->queue->unplug_fn = NULL;
-			mddev->queue->backing_dev_info.congested_fn = NULL;
-			if (mddev->pers->sync_request)
-				sysfs_remove_group(&mddev->kobj, &md_redundancy_group);
+		__md_stop_writes(mddev);
+		__md_stop(mddev);
+		mddev->queue->merge_bvec_fn = NULL;
+		mddev->queue->backing_dev_info.congested_fn = NULL;
 
-			module_put(mddev->pers->owner);
-			mddev->pers = NULL;
-			/* tell userspace to handle 'inactive' */
-			sysfs_notify(&mddev->kobj, NULL, "array_state");
+		/* tell userspace to handle 'inactive' */
+		sysfs_notify_dirent_safe(mddev->sysfs_state);
 
-			set_capacity(disk, 0);
-			mddev->changed = 1;
+		rdev_for_each(rdev, mddev)
+			if (rdev->raid_disk >= 0)
+				sysfs_unlink_rdev(mddev, rdev);
 
-			if (mddev->ro)
-				mddev->ro = 0;
-		}
-		if (!mddev->in_sync || mddev->flags) {
-			/* mark array as shutdown cleanly */
-			mddev->in_sync = 1;
-			md_update_sb(mddev, 1);
-		}
-		if (mode == 1)
-			set_disk_ro(disk, 1);
-		clear_bit(MD_RECOVERY_FROZEN, &mddev->recovery);
-	}
+		set_capacity(disk, 0);
+		mutex_unlock(&mddev->open_mutex);
+		mddev->changed = 1;
+		revalidate_disk(disk);
 
+		if (mddev->ro)
+			mddev->ro = 0;
+	} else
+		mutex_unlock(&mddev->open_mutex);
 	/*
 	 * Free resources if final stop
 	 */
 	if (mode == 0) {
-		mdk_rdev_t *rdev;
-		struct list_head *tmp;
-
 		printk(KERN_INFO "md: %s stopped.\n", mdname(mddev));
 
 		bitmap_destroy(mddev);
-		if (mddev->bitmap_file) {
-			restore_bitmap_write_access(mddev->bitmap_file);
-			fput(mddev->bitmap_file);
-			mddev->bitmap_file = NULL;
+		if (mddev->bitmap_info.file) {
+			fput(mddev->bitmap_info.file);
+			mddev->bitmap_info.file = NULL;
 		}
-		mddev->bitmap_offset = 0;
-
-		rdev_for_each(rdev, tmp, mddev)
-			if (rdev->raid_disk >= 0) {
-				char nm[20];
-				sprintf(nm, "rd%d", rdev->raid_disk);
-				sysfs_remove_link(&mddev->kobj, nm);
-			}
-
-		/* make sure all md_delayed_delete calls have finished */
-		flush_scheduled_work();
+		mddev->bitmap_info.offset = 0;
 
 		export_array(mddev);
 
-		mddev->array_sectors = 0;
-		mddev->size = 0;
-		mddev->raid_disks = 0;
-		mddev->recovery_cp = 0;
-		mddev->resync_min = 0;
-		mddev->resync_max = MaxSector;
-		mddev->reshape_position = MaxSector;
-		mddev->external = 0;
-		mddev->persistent = 0;
-		mddev->level = LEVEL_NONE;
-		mddev->clevel[0] = 0;
-		mddev->flags = 0;
-		mddev->ro = 0;
-		mddev->metadata_type[0] = 0;
-		mddev->chunk_size = 0;
-		mddev->ctime = mddev->utime = 0;
-		mddev->layout = 0;
-		mddev->max_disks = 0;
-		mddev->events = 0;
-		mddev->delta_disks = 0;
-		mddev->new_level = LEVEL_NONE;
-		mddev->new_layout = 0;
-		mddev->new_chunk = 0;
-		mddev->curr_resync = 0;
-		mddev->resync_mismatches = 0;
-		mddev->suspend_lo = mddev->suspend_hi = 0;
-		mddev->sync_speed_min = mddev->sync_speed_max = 0;
-		mddev->recovery = 0;
-		mddev->in_sync = 0;
-		mddev->changed = 0;
-		mddev->degraded = 0;
-		mddev->barriers_work = 0;
-		mddev->safemode = 0;
-
-	} else if (mddev->pers)
-		printk(KERN_INFO "md: %s switched to read-only mode.\n",
-			mdname(mddev));
-	err = 0;
+		md_clean(mddev);
+		kobject_uevent(&disk_to_dev(mddev->gendisk)->kobj, KOBJ_CHANGE);
+		if (mddev->hold_active == UNTIL_STOP)
+			mddev->hold_active = 0;
+	}
+	blk_integrity_unregister(disk);
 	md_new_event(mddev);
-	sysfs_notify(&mddev->kobj, NULL, "array_state");
-out:
-	return err;
+	sysfs_notify_dirent_safe(mddev->sysfs_state);
+	return 0;
 }
 
 #ifndef MODULE
-static void autorun_array(mddev_t *mddev)
+static void autorun_array(struct mddev *mddev)
 {
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	struct md_rdev *rdev;
 	int err;
 
 	if (list_empty(&mddev->disks))
@@ -3968,16 +5436,16 @@ static void autorun_array(mddev_t *mddev)
 
 	printk(KERN_INFO "md: running: ");
 
-	rdev_for_each(rdev, tmp, mddev) {
+	rdev_for_each(rdev, mddev) {
 		char b[BDEVNAME_SIZE];
 		printk("<%s>", bdevname(rdev->bdev,b));
 	}
 	printk("\n");
 
-	err = do_md_run (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);
 	}
 }
 
@@ -3995,9 +5463,8 @@ static void autorun_array(mddev_t *mddev)
  */
 static void autorun_devices(int part)
 {
-	struct list_head *tmp;
-	mdk_rdev_t *rdev0, *rdev;
-	mddev_t *mddev;
+	struct md_rdev *rdev0, *rdev, *tmp;
+	struct mddev *mddev;
 	char b[BDEVNAME_SIZE];
 
 	printk(KERN_INFO "md: autorun ...\n");
@@ -4006,12 +5473,12 @@ static void autorun_devices(int part)
 		dev_t dev;
 		LIST_HEAD(candidates);
 		rdev0 = list_entry(pending_raid_disks.next,
-					 mdk_rdev_t, same_set);
+					 struct md_rdev, same_set);
 
 		printk(KERN_INFO "md: considering %s ...\n",
 			bdevname(rdev0->bdev,b));
 		INIT_LIST_HEAD(&candidates);
-		rdev_for_each_list(rdev, tmp, pending_raid_disks)
+		rdev_for_each_list(rdev, tmp, &pending_raid_disks)
 			if (super_90_load(rdev, rdev0, 0) >= 0) {
 				printk(KERN_INFO "md:  adding %s ...\n",
 					bdevname(rdev->bdev,b));
@@ -4057,7 +5524,7 @@ static void autorun_devices(int part)
 		} else {
 			printk(KERN_INFO "md: created %s\n", mdname(mddev));
 			mddev->persistent = 1;
-			rdev_for_each_list(rdev, tmp, candidates) {
+			rdev_for_each_list(rdev, tmp, &candidates) {
 				list_del_init(&rdev->same_set);
 				if (bind_rdev_to_array(rdev, mddev))
 					export_rdev(rdev);
@@ -4068,7 +5535,7 @@ static void autorun_devices(int part)
 		/* on success, candidates will be empty, on error
 		 * it won't...
 		 */
-		rdev_for_each_list(rdev, tmp, candidates) {
+		rdev_for_each_list(rdev, tmp, &candidates) {
 			list_del_init(&rdev->same_set);
 			export_rdev(rdev);
 		}
@@ -4092,34 +5559,35 @@ static int get_version(void __user * arg)
 	return 0;
 }
 
-static int get_array_info(mddev_t * mddev, void __user * arg)
+static int get_array_info(struct mddev * mddev, void __user * arg)
 {
 	mdu_array_info_t info;
-	int nr,working,active,failed,spare;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	int nr,working,insync,failed,spare;
+	struct md_rdev *rdev;
 
-	nr=working=active=failed=spare=0;
-	rdev_for_each(rdev, tmp, mddev) {
+	nr = working = insync = failed = spare = 0;
+	rcu_read_lock();
+	rdev_for_each_rcu(rdev, mddev) {
 		nr++;
 		if (test_bit(Faulty, &rdev->flags))
 			failed++;
 		else {
 			working++;
 			if (test_bit(In_sync, &rdev->flags))
-				active++;	
+				insync++;	
 			else
 				spare++;
 		}
 	}
+	rcu_read_unlock();
 
 	info.major_version = mddev->major_version;
 	info.minor_version = mddev->minor_version;
 	info.patch_version = MD_PATCHLEVEL_VERSION;
 	info.ctime         = mddev->ctime;
 	info.level         = mddev->level;
-	info.size          = mddev->size;
-	if (info.size != mddev->size) /* overflow */
+	info.size          = mddev->dev_sectors / 2;
+	if (info.size != mddev->dev_sectors / 2) /* overflow */
 		info.size = -1;
 	info.nr_disks      = nr;
 	info.raid_disks    = mddev->raid_disks;
@@ -4130,15 +5598,15 @@ static int get_array_info(mddev_t * mddev, void __user * arg)
 	info.state         = 0;
 	if (mddev->in_sync)
 		info.state = (1<<MD_SB_CLEAN);
-	if (mddev->bitmap && mddev->bitmap_offset)
-		info.state = (1<<MD_SB_BITMAP_PRESENT);
-	info.active_disks  = active;
+	if (mddev->bitmap && mddev->bitmap_info.offset)
+		info.state |= (1<<MD_SB_BITMAP_PRESENT);
+	info.active_disks  = insync;
 	info.working_disks = working;
 	info.failed_disks  = failed;
 	info.spare_disks   = spare;
 
 	info.layout        = mddev->layout;
-	info.chunk_size    = mddev->chunk_size;
+	info.chunk_size    = mddev->chunk_sectors << 9;
 
 	if (copy_to_user(arg, &info, sizeof(info)))
 		return -EFAULT;
@@ -4146,22 +5614,19 @@ static int get_array_info(mddev_t * mddev, void __user * arg)
 	return 0;
 }
 
-static int get_bitmap_file(mddev_t * mddev, void __user * arg)
+static int get_bitmap_file(struct mddev * mddev, void __user * arg)
 {
 	mdu_bitmap_file_t *file = NULL; /* too big for stack allocation */
 	char *ptr, *buf = NULL;
 	int err = -ENOMEM;
 
-	if (md_allow_write(mddev))
-		file = kmalloc(sizeof(*file), GFP_NOIO);
-	else
-		file = kmalloc(sizeof(*file), GFP_KERNEL);
+	file = kmalloc(sizeof(*file), GFP_NOIO);
 
 	if (!file)
 		goto out;
 
 	/* bitmap disabled, zero the first byte and copy out */
-	if (!mddev->bitmap || !mddev->bitmap->file) {
+	if (!mddev->bitmap || !mddev->bitmap->storage.file) {
 		file->pathname[0] = '\0';
 		goto copy_out;
 	}
@@ -4170,7 +5635,8 @@ static int get_bitmap_file(mddev_t * mddev, void __user * arg)
 	if (!buf)
 		goto out;
 
-	ptr = d_path(&mddev->bitmap->file->f_path, buf, sizeof(file->pathname));
+	ptr = d_path(&mddev->bitmap->storage.file->f_path,
+		     buf, sizeof(file->pathname));
 	if (IS_ERR(ptr))
 		goto out;
 
@@ -4186,15 +5652,16 @@ out:
 	return err;
 }
 
-static int get_disk_info(mddev_t * mddev, void __user * arg)
+static int get_disk_info(struct mddev * mddev, void __user * arg)
 {
 	mdu_disk_info_t info;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 
 	if (copy_from_user(&info, arg, sizeof(info)))
 		return -EFAULT;
 
-	rdev = find_rdev_nr(mddev, info.number);
+	rcu_read_lock();
+	rdev = find_rdev_nr_rcu(mddev, info.number);
 	if (rdev) {
 		info.major = MAJOR(rdev->bdev->bd_dev);
 		info.minor = MINOR(rdev->bdev->bd_dev);
@@ -4213,6 +5680,7 @@ static int get_disk_info(mddev_t * mddev, void __user * arg)
 		info.raid_disk = -1;
 		info.state = (1<<MD_DISK_REMOVED);
 	}
+	rcu_read_unlock();
 
 	if (copy_to_user(arg, &info, sizeof(info)))
 		return -EFAULT;
@@ -4220,10 +5688,10 @@ static int get_disk_info(mddev_t * mddev, void __user * arg)
 	return 0;
 }
 
-static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
+static int add_new_disk(struct mddev * mddev, mdu_disk_info_t *info)
 {
 	char b[BDEVNAME_SIZE], b2[BDEVNAME_SIZE];
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 	dev_t dev = MKDEV(info->major,info->minor);
 
 	if (info->major != MAJOR(dev) || info->minor != MINOR(dev))
@@ -4240,9 +5708,10 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 			return PTR_ERR(rdev);
 		}
 		if (!list_empty(&mddev->disks)) {
-			mdk_rdev_t *rdev0 = list_entry(mddev->disks.next,
-							mdk_rdev_t, same_set);
-			int err = super_types[mddev->major_version]
+			struct md_rdev *rdev0
+				= list_entry(mddev->disks.next,
+					     struct md_rdev, same_set);
+			err = super_types[mddev->major_version]
 				.load_super(rdev, rdev0, mddev->minor_version);
 			if (err < 0) {
 				printk(KERN_WARNING 
@@ -4283,21 +5752,33 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 				PTR_ERR(rdev));
 			return PTR_ERR(rdev);
 		}
-		/* set save_raid_disk if appropriate */
+		/* set saved_raid_disk if appropriate */
 		if (!mddev->persistent) {
 			if (info->state & (1<<MD_DISK_SYNC)  &&
-			    info->raid_disk < mddev->raid_disks)
+			    info->raid_disk < mddev->raid_disks) {
 				rdev->raid_disk = info->raid_disk;
-			else
+				set_bit(In_sync, &rdev->flags);
+				clear_bit(Bitmap_sync, &rdev->flags);
+			} else
 				rdev->raid_disk = -1;
+			rdev->saved_raid_disk = rdev->raid_disk;
 		} else
 			super_types[mddev->major_version].
 				validate_super(mddev, rdev);
-		rdev->saved_raid_disk = rdev->raid_disk;
+		if ((info->state & (1<<MD_DISK_SYNC)) &&
+		     rdev->raid_disk != info->raid_disk) {
+			/* This was a hot-add request, but events doesn't
+			 * match, so reject it.
+			 */
+			export_rdev(rdev);
+			return -EINVAL;
+		}
 
 		clear_bit(In_sync, &rdev->flags); /* just to be sure */
 		if (info->state & (1<<MD_DISK_WRITEMOSTLY))
 			set_bit(WriteMostly, &rdev->flags);
+		else
+			clear_bit(WriteMostly, &rdev->flags);
 
 		rdev->raid_disk = -1;
 		err = bind_rdev_to_array(rdev, mddev);
@@ -4315,12 +5796,14 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 		if (err)
 			export_rdev(rdev);
 		else
-			sysfs_notify(&rdev->kobj, NULL, "state");
+			sysfs_notify_dirent_safe(rdev->sysfs_state);
 
-		md_update_sb(mddev, 1);
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		if (mddev->degraded)
 			set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+		if (!err)
+			md_new_event(mddev);
 		md_wakeup_thread(mddev->thread);
 		return err;
 	}
@@ -4336,7 +5819,7 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 
 	if (!(info->state & (1<<MD_DISK_FAULTY))) {
 		int err;
-		rdev = md_import_device (dev, -1, 0);
+		rdev = md_import_device(dev, -1, 0);
 		if (IS_ERR(rdev)) {
 			printk(KERN_WARNING 
 				"md: error, md_import_device() returned %ld\n",
@@ -4358,10 +5841,10 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 
 		if (!mddev->persistent) {
 			printk(KERN_INFO "md: nonpersistent superblock ...\n");
-			rdev->sb_start = rdev->bdev->bd_inode->i_size / 512;
-		} else 
-			rdev->sb_start = calc_dev_sboffset(rdev->bdev);
-		rdev->size = calc_num_sectors(rdev, mddev->chunk_size) / 2;
+			rdev->sb_start = i_size_read(rdev->bdev->bd_inode) / 512;
+		} else
+			rdev->sb_start = calc_dev_sboffset(rdev);
+		rdev->sectors = rdev->sb_start;
 
 		err = bind_rdev_to_array(rdev, mddev);
 		if (err) {
@@ -4373,15 +5856,18 @@ static int add_new_disk(mddev_t * mddev, mdu_disk_info_t *info)
 	return 0;
 }
 
-static int hot_remove_disk(mddev_t * mddev, dev_t dev)
+static int hot_remove_disk(struct mddev * mddev, dev_t dev)
 {
 	char b[BDEVNAME_SIZE];
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 
 	rdev = find_rdev(mddev, dev);
 	if (!rdev)
 		return -ENXIO;
 
+	clear_bit(Blocked, &rdev->flags);
+	remove_and_add_spares(mddev, rdev);
+
 	if (rdev->raid_disk >= 0)
 		goto busy;
 
@@ -4396,11 +5882,11 @@ busy:
 	return -EBUSY;
 }
 
-static int hot_add_disk(mddev_t * mddev, dev_t dev)
+static int hot_add_disk(struct mddev * mddev, dev_t dev)
 {
 	char b[BDEVNAME_SIZE];
 	int err;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 
 	if (!mddev->pers)
 		return -ENODEV;
@@ -4418,7 +5904,7 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
 		return -EINVAL;
 	}
 
-	rdev = md_import_device (dev, -1, 0);
+	rdev = md_import_device(dev, -1, 0);
 	if (IS_ERR(rdev)) {
 		printk(KERN_WARNING 
 			"md: error, md_import_device() returned %ld\n",
@@ -4427,11 +5913,11 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
 	}
 
 	if (mddev->persistent)
-		rdev->sb_start = calc_dev_sboffset(rdev->bdev);
+		rdev->sb_start = calc_dev_sboffset(rdev);
 	else
-		rdev->sb_start = rdev->bdev->bd_inode->i_size / 512;
+		rdev->sb_start = i_size_read(rdev->bdev->bd_inode) / 512;
 
-	rdev->size = calc_num_sectors(rdev, mddev->chunk_size) / 2;
+	rdev->sectors = rdev->sb_start;
 
 	if (test_bit(Faulty, &rdev->flags)) {
 		printk(KERN_WARNING 
@@ -4452,13 +5938,6 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
 	 * noticed in interrupt contexts ...
 	 */
 
-	if (rdev->desc_nr == mddev->max_disks) {
-		printk(KERN_WARNING "%s: can not hot-add to full array!\n",
-			mdname(mddev));
-		err = -EBUSY;
-		goto abort_unbind_export;
-	}
-
 	rdev->raid_disk = -1;
 
 	md_update_sb(mddev, 1);
@@ -4472,17 +5951,14 @@ static int hot_add_disk(mddev_t * mddev, dev_t dev)
 	md_new_event(mddev);
 	return 0;
 
-abort_unbind_export:
-	unbind_rdev_from_array(rdev);
-
 abort_export:
 	export_rdev(rdev);
 	return err;
 }
 
-static int set_bitmap_file(mddev_t *mddev, int fd)
+static int set_bitmap_file(struct mddev *mddev, int fd)
 {
-	int err;
+	int err = 0;
 
 	if (mddev->pers) {
 		if (!mddev->pers->quiesce)
@@ -4494,32 +5970,47 @@ static int set_bitmap_file(mddev_t *mddev, int fd)
 
 
 	if (fd >= 0) {
+		struct inode *inode;
 		if (mddev->bitmap)
 			return -EEXIST; /* cannot add when bitmap is present */
-		mddev->bitmap_file = fget(fd);
+		mddev->bitmap_info.file = fget(fd);
 
-		if (mddev->bitmap_file == NULL) {
+		if (mddev->bitmap_info.file == NULL) {
 			printk(KERN_ERR "%s: error: failed to get bitmap file\n",
 			       mdname(mddev));
 			return -EBADF;
 		}
 
-		err = deny_bitmap_write_access(mddev->bitmap_file);
-		if (err) {
+		inode = mddev->bitmap_info.file->f_mapping->host;
+		if (!S_ISREG(inode->i_mode)) {
+			printk(KERN_ERR "%s: error: bitmap file must be a regular file\n",
+			       mdname(mddev));
+			err = -EBADF;
+		} else if (!(mddev->bitmap_info.file->f_mode & FMODE_WRITE)) {
+			printk(KERN_ERR "%s: error: bitmap file must open for write\n",
+			       mdname(mddev));
+			err = -EBADF;
+		} else if (atomic_read(&inode->i_writecount) != 1) {
 			printk(KERN_ERR "%s: error: bitmap file is already in use\n",
 			       mdname(mddev));
-			fput(mddev->bitmap_file);
-			mddev->bitmap_file = NULL;
+			err = -EBUSY;
+		}
+		if (err) {
+			fput(mddev->bitmap_info.file);
+			mddev->bitmap_info.file = NULL;
 			return err;
 		}
-		mddev->bitmap_offset = 0; /* file overrides offset */
+		mddev->bitmap_info.offset = 0; /* file overrides offset */
 	} else if (mddev->bitmap == NULL)
 		return -ENOENT; /* cannot remove what isn't there */
 	err = 0;
 	if (mddev->pers) {
 		mddev->pers->quiesce(mddev, 1);
-		if (fd >= 0)
+		if (fd >= 0) {
 			err = bitmap_create(mddev);
+			if (!err)
+				err = bitmap_load(mddev);
+		}
 		if (fd < 0 || err) {
 			bitmap_destroy(mddev);
 			fd = -1; /* make sure to put the file */
@@ -4527,11 +6018,9 @@ static int set_bitmap_file(mddev_t *mddev, int fd)
 		mddev->pers->quiesce(mddev, 0);
 	}
 	if (fd < 0) {
-		if (mddev->bitmap_file) {
-			restore_bitmap_write_access(mddev->bitmap_file);
-			fput(mddev->bitmap_file);
-		}
-		mddev->bitmap_file = NULL;
+		if (mddev->bitmap_info.file)
+			fput(mddev->bitmap_info.file);
+		mddev->bitmap_info.file = NULL;
 	}
 
 	return err;
@@ -4550,7 +6039,7 @@ static int set_bitmap_file(mddev_t *mddev, int fd)
  *  The minor and patch _version numbers are also kept incase the
  *  super_block handler wishes to interpret them.
  */
-static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
+static int set_array_info(struct mddev * mddev, mdu_array_info_t *info)
 {
 
 	if (info->raid_disks == 0) {
@@ -4568,6 +6057,10 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 		mddev->minor_version = info->minor_version;
 		mddev->patch_version = info->patch_version;
 		mddev->persistent = !info->not_persistent;
+		/* ensure mddev_put doesn't delete this now that there
+		 * is some minimal configuration.
+		 */
+		mddev->ctime         = get_seconds();
 		return 0;
 	}
 	mddev->major_version = MD_MAJOR_VERSION;
@@ -4577,7 +6070,7 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 
 	mddev->level         = info->level;
 	mddev->clevel[0]     = 0;
-	mddev->size          = info->size;
+	mddev->dev_sectors   = 2 * (sector_t)info->size;
 	mddev->raid_disks    = info->raid_disks;
 	/* don't set md_minor, it is determined by which /dev/md* was
 	 * openned
@@ -4590,7 +6083,7 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 	mddev->external	     = 0;
 
 	mddev->layout        = info->layout;
-	mddev->chunk_size    = info->chunk_size;
+	mddev->chunk_sectors = info->chunk_size >> 9;
 
 	mddev->max_disks     = MD_SB_DISKS;
 
@@ -4598,8 +6091,9 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 		mddev->flags         = 0;
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
 
-	mddev->default_bitmap_offset = MD_SB_BYTES >> 9;
-	mddev->bitmap_offset = 0;
+	mddev->bitmap_info.default_offset = MD_SB_BYTES >> 9;
+	mddev->bitmap_info.default_space = 64*2 - (MD_SB_BYTES >> 9);
+	mddev->bitmap_info.offset = 0;
 
 	mddev->reshape_position = MaxSector;
 
@@ -4609,18 +6103,29 @@ static int set_array_info(mddev_t * mddev, mdu_array_info_t *info)
 	get_random_bytes(mddev->uuid, 16);
 
 	mddev->new_level = mddev->level;
-	mddev->new_chunk = mddev->chunk_size;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
 	mddev->new_layout = mddev->layout;
 	mddev->delta_disks = 0;
+	mddev->reshape_backwards = 0;
 
 	return 0;
 }
 
-static int update_size(mddev_t *mddev, sector_t num_sectors)
+void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors)
 {
-	mdk_rdev_t * rdev;
+	WARN(!mddev_is_locked(mddev), "%s: unlocked mddev!\n", __func__);
+
+	if (mddev->external_size)
+		return;
+
+	mddev->array_sectors = array_sectors;
+}
+EXPORT_SYMBOL(md_set_array_sectors);
+
+static int update_size(struct mddev *mddev, sector_t num_sectors)
+{
+	struct md_rdev *rdev;
 	int rv;
-	struct list_head *tmp;
 	int fit = (num_sectors == 0);
 
 	if (mddev->pers->resize == NULL)
@@ -4633,18 +6138,14 @@ static int update_size(mddev_t *mddev, sector_t num_sectors)
 	 * sb_start or, if that is <data_offset, it must fit before the size
 	 * of each device.  If num_sectors is zero, we find the largest size
 	 * that fits.
-
 	 */
 	if (mddev->sync_thread)
 		return -EBUSY;
-	if (mddev->bitmap)
-		/* Sorry, cannot grow a bitmap yet, just remove it,
-		 * grow, and re-add.
-		 */
-		return -EBUSY;
-	rdev_for_each(rdev, tmp, mddev) {
-		sector_t avail;
-		avail = rdev->size * 2;
+	if (mddev->ro)
+		return -EROFS;
+
+	rdev_for_each(rdev, mddev) {
+		sector_t avail = rdev->sectors;
 
 		if (fit && (num_sectors == 0 || num_sectors > avail))
 			num_sectors = avail;
@@ -4652,35 +6153,46 @@ static int update_size(mddev_t *mddev, sector_t num_sectors)
 			return -ENOSPC;
 	}
 	rv = mddev->pers->resize(mddev, num_sectors);
-	if (!rv) {
-		struct block_device *bdev;
-
-		bdev = bdget_disk(mddev->gendisk, 0);
-		if (bdev) {
-			mutex_lock(&bdev->bd_inode->i_mutex);
-			i_size_write(bdev->bd_inode,
-				     (loff_t)mddev->array_sectors << 9);
-			mutex_unlock(&bdev->bd_inode->i_mutex);
-			bdput(bdev);
-		}
-	}
+	if (!rv)
+		revalidate_disk(mddev->gendisk);
 	return rv;
 }
 
-static int update_raid_disks(mddev_t *mddev, int raid_disks)
+static int update_raid_disks(struct mddev *mddev, int raid_disks)
 {
 	int rv;
+	struct md_rdev *rdev;
 	/* change the number of raid disks */
 	if (mddev->pers->check_reshape == NULL)
 		return -EINVAL;
+	if (mddev->ro)
+		return -EROFS;
 	if (raid_disks <= 0 ||
-	    raid_disks >= mddev->max_disks)
+	    (mddev->max_disks && raid_disks >= mddev->max_disks))
 		return -EINVAL;
 	if (mddev->sync_thread || mddev->reshape_position != MaxSector)
 		return -EBUSY;
+
+	rdev_for_each(rdev, mddev) {
+		if (mddev->raid_disks < raid_disks &&
+		    rdev->data_offset < rdev->new_data_offset)
+			return -EINVAL;
+		if (mddev->raid_disks > raid_disks &&
+		    rdev->data_offset > rdev->new_data_offset)
+			return -EINVAL;
+	}
+
 	mddev->delta_disks = raid_disks - mddev->raid_disks;
+	if (mddev->delta_disks < 0)
+		mddev->reshape_backwards = 1;
+	else if (mddev->delta_disks > 0)
+		mddev->reshape_backwards = 0;
 
 	rv = mddev->pers->check_reshape(mddev);
+	if (rv < 0) {
+		mddev->delta_disks = 0;
+		mddev->reshape_backwards = 0;
+	}
 	return rv;
 }
 
@@ -4693,14 +6205,14 @@ static int update_raid_disks(mddev_t *mddev, int raid_disks)
  * Any differences that cannot be handled will cause an error.
  * Normally, only one change can be managed at a time.
  */
-static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
+static int update_array_info(struct mddev *mddev, mdu_array_info_t *info)
 {
 	int rv = 0;
 	int cnt = 0;
 	int state = 0;
 
 	/* calculate expected state,ignoring low bits */
-	if (mddev->bitmap && mddev->bitmap_offset)
+	if (mddev->bitmap && mddev->bitmap_info.offset)
 		state |= (1 << MD_SB_BITMAP_PRESENT);
 
 	if (mddev->major_version != info->major_version ||
@@ -4710,30 +6222,41 @@ static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
 	    mddev->level         != info->level         ||
 /*	    mddev->layout        != info->layout        || */
 	    !mddev->persistent	 != info->not_persistent||
-	    mddev->chunk_size    != info->chunk_size    ||
+	    mddev->chunk_sectors != info->chunk_size >> 9 ||
 	    /* ignore bottom 8 bits of state, and allow SB_BITMAP_PRESENT to change */
 	    ((state^info->state) & 0xfffffe00)
 		)
 		return -EINVAL;
 	/* Check there is only one change */
-	if (info->size >= 0 && mddev->size != info->size) cnt++;
-	if (mddev->raid_disks != info->raid_disks) cnt++;
-	if (mddev->layout != info->layout) cnt++;
-	if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT)) cnt++;
-	if (cnt == 0) return 0;
-	if (cnt > 1) return -EINVAL;
+	if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
+		cnt++;
+	if (mddev->raid_disks != info->raid_disks)
+		cnt++;
+	if (mddev->layout != info->layout)
+		cnt++;
+	if ((state ^ info->state) & (1<<MD_SB_BITMAP_PRESENT))
+		cnt++;
+	if (cnt == 0)
+		return 0;
+	if (cnt > 1)
+		return -EINVAL;
 
 	if (mddev->layout != info->layout) {
 		/* Change layout
 		 * we don't need to do anything at the md level, the
 		 * personality will take care of it all.
 		 */
-		if (mddev->pers->reconfig == NULL)
+		if (mddev->pers->check_reshape == NULL)
 			return -EINVAL;
-		else
-			return mddev->pers->reconfig(mddev, info->layout, -1);
+		else {
+			mddev->new_layout = info->layout;
+			rv = mddev->pers->check_reshape(mddev);
+			if (rv)
+				mddev->new_layout = mddev->layout;
+			return rv;
+		}
 	}
-	if (info->size >= 0 && mddev->size != info->size)
+	if (info->size >= 0 && mddev->dev_sectors / 2 != info->size)
 		rv = update_size(mddev, (sector_t)info->size * 2);
 
 	if (mddev->raid_disks    != info->raid_disks)
@@ -4748,11 +6271,16 @@ static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
 			/* add the bitmap */
 			if (mddev->bitmap)
 				return -EEXIST;
-			if (mddev->default_bitmap_offset == 0)
+			if (mddev->bitmap_info.default_offset == 0)
 				return -EINVAL;
-			mddev->bitmap_offset = mddev->default_bitmap_offset;
+			mddev->bitmap_info.offset =
+				mddev->bitmap_info.default_offset;
+			mddev->bitmap_info.space =
+				mddev->bitmap_info.default_space;
 			mddev->pers->quiesce(mddev, 1);
 			rv = bitmap_create(mddev);
+			if (!rv)
+				rv = bitmap_load(mddev);
 			if (rv)
 				bitmap_destroy(mddev);
 			mddev->pers->quiesce(mddev, 0);
@@ -4760,31 +6288,37 @@ static int update_array_info(mddev_t *mddev, mdu_array_info_t *info)
 			/* remove the bitmap */
 			if (!mddev->bitmap)
 				return -ENOENT;
-			if (mddev->bitmap->file)
+			if (mddev->bitmap->storage.file)
 				return -EINVAL;
 			mddev->pers->quiesce(mddev, 1);
 			bitmap_destroy(mddev);
 			mddev->pers->quiesce(mddev, 0);
-			mddev->bitmap_offset = 0;
+			mddev->bitmap_info.offset = 0;
 		}
 	}
 	md_update_sb(mddev, 1);
 	return rv;
 }
 
-static int set_disk_faulty(mddev_t *mddev, dev_t dev)
+static int set_disk_faulty(struct mddev *mddev, dev_t dev)
 {
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
+	int err = 0;
 
 	if (mddev->pers == NULL)
 		return -ENODEV;
 
-	rdev = find_rdev(mddev, dev);
+	rcu_read_lock();
+	rdev = find_rdev_rcu(mddev, dev);
 	if (!rdev)
-		return -ENODEV;
-
-	md_error(mddev, rdev);
-	return 0;
+		err =  -ENODEV;
+	else {
+		md_error(mddev, rdev);
+		if (!test_bit(Faulty, &rdev->flags))
+			err = -EBUSY;
+	}
+	rcu_read_unlock();
+	return err;
 }
 
 /*
@@ -4795,59 +6329,140 @@ static int set_disk_faulty(mddev_t *mddev, dev_t dev)
  */
 static int md_getgeo(struct block_device *bdev, struct hd_geometry *geo)
 {
-	mddev_t *mddev = bdev->bd_disk->private_data;
+	struct mddev *mddev = bdev->bd_disk->private_data;
 
 	geo->heads = 2;
 	geo->sectors = 4;
-	geo->cylinders = get_capacity(mddev->gendisk) / 8;
+	geo->cylinders = mddev->array_sectors / 8;
 	return 0;
 }
 
-static int md_ioctl(struct inode *inode, struct file *file,
+static inline bool md_ioctl_valid(unsigned int cmd)
+{
+	switch (cmd) {
+	case ADD_NEW_DISK:
+	case BLKROSET:
+	case GET_ARRAY_INFO:
+	case GET_BITMAP_FILE:
+	case GET_DISK_INFO:
+	case HOT_ADD_DISK:
+	case HOT_REMOVE_DISK:
+	case PRINT_RAID_DEBUG:
+	case RAID_AUTORUN:
+	case RAID_VERSION:
+	case RESTART_ARRAY_RW:
+	case RUN_ARRAY:
+	case SET_ARRAY_INFO:
+	case SET_BITMAP_FILE:
+	case SET_DISK_FAULTY:
+	case STOP_ARRAY:
+	case STOP_ARRAY_RO:
+		return true;
+	default:
+		return false;
+	}
+}
+
+static int md_ioctl(struct block_device *bdev, fmode_t mode,
 			unsigned int cmd, unsigned long arg)
 {
 	int err = 0;
 	void __user *argp = (void __user *)arg;
-	mddev_t *mddev = NULL;
+	struct mddev *mddev = NULL;
+	int ro;
 
-	if (!capable(CAP_SYS_ADMIN))
-		return -EACCES;
+	if (!md_ioctl_valid(cmd))
+		return -ENOTTY;
+
+	switch (cmd) {
+	case RAID_VERSION:
+	case GET_ARRAY_INFO:
+	case GET_DISK_INFO:
+		break;
+	default:
+		if (!capable(CAP_SYS_ADMIN))
+			return -EACCES;
+	}
 
 	/*
 	 * Commands dealing with the RAID driver but not any
 	 * particular array:
 	 */
-	switch (cmd)
-	{
-		case RAID_VERSION:
-			err = get_version(argp);
-			goto done;
+	switch (cmd) {
+	case RAID_VERSION:
+		err = get_version(argp);
+		goto done;
 
-		case PRINT_RAID_DEBUG:
-			err = 0;
-			md_print_devices();
-			goto done;
+	case PRINT_RAID_DEBUG:
+		err = 0;
+		md_print_devices();
+		goto done;
 
 #ifndef MODULE
-		case RAID_AUTORUN:
-			err = 0;
-			autostart_arrays(arg);
-			goto done;
+	case RAID_AUTORUN:
+		err = 0;
+		autostart_arrays(arg);
+		goto done;
 #endif
-		default:;
+	default:;
 	}
 
 	/*
 	 * Commands creating/starting a new array:
 	 */
 
-	mddev = inode->i_bdev->bd_disk->private_data;
+	mddev = bdev->bd_disk->private_data;
 
 	if (!mddev) {
 		BUG();
 		goto abort;
 	}
 
+	/* Some actions do not requires the mutex */
+	switch (cmd) {
+	case GET_ARRAY_INFO:
+		if (!mddev->raid_disks && !mddev->external)
+			err = -ENODEV;
+		else
+			err = get_array_info(mddev, argp);
+		goto abort;
+
+	case GET_DISK_INFO:
+		if (!mddev->raid_disks && !mddev->external)
+			err = -ENODEV;
+		else
+			err = get_disk_info(mddev, argp);
+		goto abort;
+
+	case SET_DISK_FAULTY:
+		err = set_disk_faulty(mddev, new_decode_dev(arg));
+		goto abort;
+	}
+
+	if (cmd == ADD_NEW_DISK)
+		/* need to ensure md_delayed_delete() has completed */
+		flush_workqueue(md_misc_wq);
+
+	if (cmd == HOT_REMOVE_DISK)
+		/* need to ensure recovery thread has run */
+		wait_event_interruptible_timeout(mddev->sb_wait,
+						 !test_bit(MD_RECOVERY_NEEDED,
+							   &mddev->flags),
+						 msecs_to_jiffies(5000));
+	if (cmd == STOP_ARRAY || cmd == STOP_ARRAY_RO) {
+		/* Need to flush page cache, and ensure no-one else opens
+		 * and writes
+		 */
+		mutex_lock(&mddev->open_mutex);
+		if (atomic_read(&mddev->openers) > 1) {
+			mutex_unlock(&mddev->open_mutex);
+			err = -EBUSY;
+			goto abort;
+		}
+		set_bit(MD_STILL_CLOSED, &mddev->flags);
+		mutex_unlock(&mddev->open_mutex);
+		sync_blockdev(bdev);
+	}
 	err = mddev_lock(mddev);
 	if (err) {
 		printk(KERN_INFO 
@@ -4856,50 +6471,44 @@ static int md_ioctl(struct inode *inode, struct file *file,
 		goto abort;
 	}
 
-	switch (cmd)
-	{
-		case SET_ARRAY_INFO:
-			{
-				mdu_array_info_t info;
-				if (!arg)
-					memset(&info, 0, sizeof(info));
-				else if (copy_from_user(&info, argp, sizeof(info))) {
-					err = -EFAULT;
-					goto abort_unlock;
-				}
-				if (mddev->pers) {
-					err = update_array_info(mddev, &info);
-					if (err) {
-						printk(KERN_WARNING "md: couldn't update"
-						       " array info. %d\n", err);
-						goto abort_unlock;
-					}
-					goto done_unlock;
-				}
-				if (!list_empty(&mddev->disks)) {
-					printk(KERN_WARNING
-					       "md: array %s already has disks!\n",
-					       mdname(mddev));
-					err = -EBUSY;
-					goto abort_unlock;
-				}
-				if (mddev->raid_disks) {
-					printk(KERN_WARNING
-					       "md: array %s already initialised!\n",
-					       mdname(mddev));
-					err = -EBUSY;
-					goto abort_unlock;
-				}
-				err = set_array_info(mddev, &info);
-				if (err) {
-					printk(KERN_WARNING "md: couldn't set"
-					       " array info. %d\n", err);
-					goto abort_unlock;
-				}
+	if (cmd == SET_ARRAY_INFO) {
+		mdu_array_info_t info;
+		if (!arg)
+			memset(&info, 0, sizeof(info));
+		else if (copy_from_user(&info, argp, sizeof(info))) {
+			err = -EFAULT;
+			goto abort_unlock;
+		}
+		if (mddev->pers) {
+			err = update_array_info(mddev, &info);
+			if (err) {
+				printk(KERN_WARNING "md: couldn't update"
+				       " array info. %d\n", err);
+				goto abort_unlock;
 			}
 			goto done_unlock;
-
-		default:;
+		}
+		if (!list_empty(&mddev->disks)) {
+			printk(KERN_WARNING
+			       "md: array %s already has disks!\n",
+			       mdname(mddev));
+			err = -EBUSY;
+			goto abort_unlock;
+		}
+		if (mddev->raid_disks) {
+			printk(KERN_WARNING
+			       "md: array %s already initialised!\n",
+			       mdname(mddev));
+			err = -EBUSY;
+			goto abort_unlock;
+		}
+		err = set_array_info(mddev, &info);
+		if (err) {
+			printk(KERN_WARNING "md: couldn't set"
+			       " array info. %d\n", err);
+			goto abort_unlock;
+		}
+		goto done_unlock;
 	}
 
 	/*
@@ -4918,32 +6527,73 @@ static int md_ioctl(struct inode *inode, struct file *file,
 	/*
 	 * Commands even a read-only array can execute:
 	 */
-	switch (cmd)
-	{
-		case GET_ARRAY_INFO:
-			err = get_array_info(mddev, argp);
-			goto done_unlock;
-
-		case GET_BITMAP_FILE:
-			err = get_bitmap_file(mddev, argp);
-			goto done_unlock;
-
-		case GET_DISK_INFO:
-			err = get_disk_info(mddev, argp);
+	switch (cmd) {
+	case GET_BITMAP_FILE:
+		err = get_bitmap_file(mddev, argp);
+		goto done_unlock;
+
+	case RESTART_ARRAY_RW:
+		err = restart_array(mddev);
+		goto done_unlock;
+
+	case STOP_ARRAY:
+		err = do_md_stop(mddev, 0, bdev);
+		goto done_unlock;
+
+	case STOP_ARRAY_RO:
+		err = md_set_readonly(mddev, bdev);
+		goto done_unlock;
+
+	case HOT_REMOVE_DISK:
+		err = hot_remove_disk(mddev, new_decode_dev(arg));
+		goto done_unlock;
+
+	case ADD_NEW_DISK:
+		/* We can support ADD_NEW_DISK on read-only arrays
+		 * on if we are re-adding a preexisting device.
+		 * So require mddev->pers and MD_DISK_SYNC.
+		 */
+		if (mddev->pers) {
+			mdu_disk_info_t info;
+			if (copy_from_user(&info, argp, sizeof(info)))
+				err = -EFAULT;
+			else if (!(info.state & (1<<MD_DISK_SYNC)))
+				/* Need to clear read-only for this */
+				break;
+			else
+				err = add_new_disk(mddev, &info);
 			goto done_unlock;
+		}
+		break;
 
-		case RESTART_ARRAY_RW:
-			err = restart_array(mddev);
+	case BLKROSET:
+		if (get_user(ro, (int __user *)(arg))) {
+			err = -EFAULT;
 			goto done_unlock;
+		}
+		err = -EINVAL;
 
-		case STOP_ARRAY:
-			err = do_md_stop (mddev, 0, 1);
+		/* if the bdev is going readonly the value of mddev->ro
+		 * does not matter, no writes are coming
+		 */
+		if (ro)
 			goto done_unlock;
 
-		case STOP_ARRAY_RO:
-			err = do_md_stop (mddev, 1, 1);
+		/* are we are already prepared for writes? */
+		if (mddev->ro != 1)
 			goto done_unlock;
 
+		/* transitioning to readauto need only happen for
+		 * arrays that call md_write_start
+		 */
+		if (mddev->pers) {
+			err = restart_array(mddev);
+			if (err == 0) {
+				mddev->ro = 2;
+				set_disk_ro(mddev->gendisk, 0);
+			}
+		}
+		goto done_unlock;
 	}
 
 	/*
@@ -4956,54 +6606,58 @@ static int md_ioctl(struct inode *inode, struct file *file,
 	if (_IOC_TYPE(cmd) == MD_MAJOR && mddev->ro && mddev->pers) {
 		if (mddev->ro == 2) {
 			mddev->ro = 0;
-			sysfs_notify(&mddev->kobj, NULL, "array_state");
+			sysfs_notify_dirent_safe(mddev->sysfs_state);
 			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
-			md_wakeup_thread(mddev->thread);
+			/* mddev_unlock will wake thread */
+			/* If a device failed while we were read-only, we
+			 * need to make sure the metadata is updated now.
+			 */
+			if (test_bit(MD_CHANGE_DEVS, &mddev->flags)) {
+				mddev_unlock(mddev);
+				wait_event(mddev->sb_wait,
+					   !test_bit(MD_CHANGE_DEVS, &mddev->flags) &&
+					   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+				mddev_lock_nointr(mddev);
+			}
 		} else {
 			err = -EROFS;
 			goto abort_unlock;
 		}
 	}
 
-	switch (cmd)
+	switch (cmd) {
+	case ADD_NEW_DISK:
 	{
-		case ADD_NEW_DISK:
-		{
-			mdu_disk_info_t info;
-			if (copy_from_user(&info, argp, sizeof(info)))
-				err = -EFAULT;
-			else
-				err = add_new_disk(mddev, &info);
-			goto done_unlock;
-		}
-
-		case HOT_REMOVE_DISK:
-			err = hot_remove_disk(mddev, new_decode_dev(arg));
-			goto done_unlock;
-
-		case HOT_ADD_DISK:
-			err = hot_add_disk(mddev, new_decode_dev(arg));
-			goto done_unlock;
+		mdu_disk_info_t info;
+		if (copy_from_user(&info, argp, sizeof(info)))
+			err = -EFAULT;
+		else
+			err = add_new_disk(mddev, &info);
+		goto done_unlock;
+	}
 
-		case SET_DISK_FAULTY:
-			err = set_disk_faulty(mddev, new_decode_dev(arg));
-			goto done_unlock;
+	case HOT_ADD_DISK:
+		err = hot_add_disk(mddev, new_decode_dev(arg));
+		goto done_unlock;
 
-		case RUN_ARRAY:
-			err = do_md_run (mddev);
-			goto done_unlock;
+	case RUN_ARRAY:
+		err = do_md_run(mddev);
+		goto done_unlock;
 
-		case SET_BITMAP_FILE:
-			err = set_bitmap_file(mddev, (int)arg);
-			goto done_unlock;
+	case SET_BITMAP_FILE:
+		err = set_bitmap_file(mddev, (int)arg);
+		goto done_unlock;
 
-		default:
-			err = -EINVAL;
-			goto abort_unlock;
+	default:
+		err = -EINVAL;
+		goto abort_unlock;
 	}
 
 done_unlock:
 abort_unlock:
+	if (mddev->hold_active == UNTIL_IOCTL &&
+	    err != -EINVAL)
+		mddev->hold_active = 0;
 	mddev_unlock(mddev);
 
 	return err;
@@ -5013,68 +6667,103 @@ done:
 abort:
 	return err;
 }
+#ifdef CONFIG_COMPAT
+static int md_compat_ioctl(struct block_device *bdev, fmode_t mode,
+		    unsigned int cmd, unsigned long arg)
+{
+	switch (cmd) {
+	case HOT_REMOVE_DISK:
+	case HOT_ADD_DISK:
+	case SET_DISK_FAULTY:
+	case SET_BITMAP_FILE:
+		/* These take in integer arg, do not convert */
+		break;
+	default:
+		arg = (unsigned long)compat_ptr(arg);
+		break;
+	}
+
+	return md_ioctl(bdev, mode, cmd, arg);
+}
+#endif /* CONFIG_COMPAT */
 
-static int md_open(struct inode *inode, struct file *file)
+static int md_open(struct block_device *bdev, fmode_t mode)
 {
 	/*
 	 * Succeed if we can lock the mddev, which confirms that
 	 * it isn't being stopped right now.
 	 */
-	mddev_t *mddev = inode->i_bdev->bd_disk->private_data;
+	struct mddev *mddev = mddev_find(bdev->bd_dev);
 	int err;
 
-	if ((err = mutex_lock_interruptible_nested(&mddev->reconfig_mutex, 1)))
+	if (!mddev)
+		return -ENODEV;
+
+	if (mddev->gendisk != bdev->bd_disk) {
+		/* we are racing with mddev_put which is discarding this
+		 * bd_disk.
+		 */
+		mddev_put(mddev);
+		/* Wait until bdev->bd_disk is definitely gone */
+		flush_workqueue(md_misc_wq);
+		/* Then retry the open from the top */
+		return -ERESTARTSYS;
+	}
+	BUG_ON(mddev != bdev->bd_disk->private_data);
+
+	if ((err = mutex_lock_interruptible(&mddev->open_mutex)))
 		goto out;
 
 	err = 0;
-	mddev_get(mddev);
 	atomic_inc(&mddev->openers);
-	mddev_unlock(mddev);
+	clear_bit(MD_STILL_CLOSED, &mddev->flags);
+	mutex_unlock(&mddev->open_mutex);
 
-	check_disk_change(inode->i_bdev);
+	check_disk_change(bdev);
  out:
 	return err;
 }
 
-static int md_release(struct inode *inode, struct file * file)
+static void md_release(struct gendisk *disk, fmode_t mode)
 {
- 	mddev_t *mddev = inode->i_bdev->bd_disk->private_data;
+ 	struct mddev *mddev = disk->private_data;
 
 	BUG_ON(!mddev);
 	atomic_dec(&mddev->openers);
 	mddev_put(mddev);
-
-	return 0;
 }
 
 static int md_media_changed(struct gendisk *disk)
 {
-	mddev_t *mddev = disk->private_data;
+	struct mddev *mddev = disk->private_data;
 
 	return mddev->changed;
 }
 
 static int md_revalidate(struct gendisk *disk)
 {
-	mddev_t *mddev = disk->private_data;
+	struct mddev *mddev = disk->private_data;
 
 	mddev->changed = 0;
 	return 0;
 }
-static struct block_device_operations md_fops =
+static const struct block_device_operations md_fops =
 {
 	.owner		= THIS_MODULE,
 	.open		= md_open,
 	.release	= md_release,
 	.ioctl		= md_ioctl,
+#ifdef CONFIG_COMPAT
+	.compat_ioctl	= md_compat_ioctl,
+#endif
 	.getgeo		= md_getgeo,
-	.media_changed	= md_media_changed,
+	.media_changed  = md_media_changed,
 	.revalidate_disk= md_revalidate,
 };
 
 static int md_thread(void * arg)
 {
-	mdk_thread_t *thread = arg;
+	struct md_thread *thread = arg;
 
 	/*
 	 * md_thread is a 'system-thread', it's priority should be very
@@ -5106,28 +6795,28 @@ static int md_thread(void * arg)
 			 thread->timeout);
 
 		clear_bit(THREAD_WAKEUP, &thread->flags);
-
-		thread->run(thread->mddev);
+		if (!kthread_should_stop())
+			thread->run(thread);
 	}
 
 	return 0;
 }
 
-void md_wakeup_thread(mdk_thread_t *thread)
+void md_wakeup_thread(struct md_thread *thread)
 {
 	if (thread) {
-		dprintk("md: waking up MD thread %s.\n", thread->tsk->comm);
+		pr_debug("md: waking up MD thread %s.\n", thread->tsk->comm);
 		set_bit(THREAD_WAKEUP, &thread->flags);
 		wake_up(&thread->wqueue);
 	}
 }
 
-mdk_thread_t *md_register_thread(void (*run) (mddev_t *), mddev_t *mddev,
-				 const char *name)
+struct md_thread *md_register_thread(void (*run) (struct md_thread *),
+		struct mddev *mddev, const char *name)
 {
-	mdk_thread_t *thread;
+	struct md_thread *thread;
 
-	thread = kzalloc(sizeof(mdk_thread_t), GFP_KERNEL);
+	thread = kzalloc(sizeof(struct md_thread), GFP_KERNEL);
 	if (!thread)
 		return NULL;
 
@@ -5136,7 +6825,10 @@ mdk_thread_t *md_register_thread(void (*run) (mddev_t *), mddev_t *mddev,
 	thread->run = run;
 	thread->mddev = mddev;
 	thread->timeout = MAX_SCHEDULE_TIMEOUT;
-	thread->tsk = kthread_run(md_thread, thread, name, mdname(thread->mddev));
+	thread->tsk = kthread_run(md_thread, thread,
+				  "%s_%s",
+				  mdname(thread->mddev),
+				  name);
 	if (IS_ERR(thread->tsk)) {
 		kfree(thread);
 		return NULL;
@@ -5144,15 +6836,24 @@ mdk_thread_t *md_register_thread(void (*run) (mddev_t *), mddev_t *mddev,
 	return thread;
 }
 
-void md_unregister_thread(mdk_thread_t *thread)
+void md_unregister_thread(struct md_thread **threadp)
 {
-	dprintk("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
+	struct md_thread *thread = *threadp;
+	if (!thread)
+		return;
+	pr_debug("interrupting MD-thread pid %d\n", task_pid_nr(thread->tsk));
+	/* Locking ensures that mddev_unlock does not wake_up a
+	 * non-existent thread
+	 */
+	spin_lock(&pers_lock);
+	*threadp = NULL;
+	spin_unlock(&pers_lock);
 
 	kthread_stop(thread->tsk);
 	kfree(thread);
 }
 
-void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
+void md_error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	if (!mddev) {
 		MD_BUG();
@@ -5162,26 +6863,17 @@ void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
 	if (!rdev || test_bit(Faulty, &rdev->flags))
 		return;
 
-	if (mddev->external)
-		set_bit(Blocked, &rdev->flags);
-/*
-	dprintk("md_error dev:%s, rdev:(%d:%d), (caller: %p,%p,%p,%p).\n",
-		mdname(mddev),
-		MAJOR(rdev->bdev->bd_dev), MINOR(rdev->bdev->bd_dev),
-		__builtin_return_address(0),__builtin_return_address(1),
-		__builtin_return_address(2),__builtin_return_address(3));
-*/
-	if (!mddev->pers)
-		return;
-	if (!mddev->pers->error_handler)
+	if (!mddev->pers || !mddev->pers->error_handler)
 		return;
 	mddev->pers->error_handler(mddev,rdev);
 	if (mddev->degraded)
 		set_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
-	set_bit(StateChanged, &rdev->flags);
+	sysfs_notify_dirent_safe(rdev->sysfs_state);
 	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
+	if (mddev->event_work.func)
+		queue_work(md_misc_wq, &mddev->event_work);
 	md_new_event_inintr(mddev);
 }
 
@@ -5190,12 +6882,11 @@ void md_error(mddev_t *mddev, mdk_rdev_t *rdev)
 static void status_unused(struct seq_file *seq)
 {
 	int i = 0;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	struct md_rdev *rdev;
 
 	seq_printf(seq, "unused devices: ");
 
-	rdev_for_each_list(rdev, tmp, pending_raid_disks) {
+	list_for_each_entry(rdev, &pending_raid_disks, same_set) {
 		char b[BDEVNAME_SIZE];
 		i++;
 		seq_printf(seq, "%s ",
@@ -5208,39 +6899,45 @@ static void status_unused(struct seq_file *seq)
 }
 
 
-static void status_resync(struct seq_file *seq, mddev_t * mddev)
+static void status_resync(struct seq_file *seq, struct mddev * mddev)
 {
-	sector_t max_blocks, resync, res;
-	unsigned long dt, db, rt;
+	sector_t max_sectors, resync, res;
+	unsigned long dt, db;
+	sector_t rt;
 	int scale;
 	unsigned int per_milli;
 
-	resync = (mddev->curr_resync - atomic_read(&mddev->recovery_active))/2;
+	if (mddev->curr_resync <= 3)
+		resync = 0;
+	else
+		resync = mddev->curr_resync
+			- atomic_read(&mddev->recovery_active);
 
-	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
-		max_blocks = mddev->resync_max_sectors >> 1;
+	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+		max_sectors = mddev->resync_max_sectors;
 	else
-		max_blocks = mddev->size;
+		max_sectors = mddev->dev_sectors;
 
 	/*
 	 * Should not happen.
 	 */
-	if (!max_blocks) {
+	if (!max_sectors) {
 		MD_BUG();
 		return;
 	}
 	/* Pick 'scale' such that (resync>>scale)*1000 will fit
-	 * in a sector_t, and (max_blocks>>scale) will fit in a
+	 * in a sector_t, and (max_sectors>>scale) will fit in a
 	 * u32, as those are the requirements for sector_div.
 	 * Thus 'scale' must be at least 10
 	 */
 	scale = 10;
 	if (sizeof(sector_t) > sizeof(unsigned long)) {
-		while ( max_blocks/2 > (1ULL<<(scale+32)))
+		while ( max_sectors/2 > (1ULL<<(scale+32)))
 			scale++;
 	}
 	res = (resync>>scale)*1000;
-	sector_div(res, (u32)((max_blocks>>scale)+1));
+	sector_div(res, (u32)((max_sectors>>scale)+1));
 
 	per_milli = res;
 	{
@@ -5261,25 +6958,35 @@ static void status_resync(struct seq_file *seq, mddev_t * mddev)
 		     (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ?
 		      "resync" : "recovery"))),
 		   per_milli/10, per_milli % 10,
-		   (unsigned long long) resync,
-		   (unsigned long long) max_blocks);
+		   (unsigned long long) resync/2,
+		   (unsigned long long) max_sectors/2);
 
 	/*
-	 * We do not want to overflow, so the order of operands and
-	 * the * 100 / 100 trick are important. We do a +1 to be
-	 * safe against division by zero. We only estimate anyway.
-	 *
 	 * dt: time from mark until now
 	 * db: blocks written from mark until now
 	 * rt: remaining time
+	 *
+	 * rt is a sector_t, so could be 32bit or 64bit.
+	 * So we divide before multiply in case it is 32bit and close
+	 * to the limit.
+	 * We scale the divisor (db) by 32 to avoid losing precision
+	 * near the end of resync when the number of remaining sectors
+	 * is close to 'db'.
+	 * We then divide rt by 32 after multiplying by db to compensate.
+	 * The '+1' avoids division by zero if db is very small.
 	 */
 	dt = ((jiffies - mddev->resync_mark) / HZ);
 	if (!dt) dt++;
 	db = (mddev->curr_mark_cnt - atomic_read(&mddev->recovery_active))
 		- mddev->resync_mark_cnt;
-	rt = (dt * ((unsigned long)(max_blocks-resync) / (db/2/100+1)))/100;
 
-	seq_printf(seq, " finish=%lu.%lumin", rt / 60, (rt % 60)/6);
+	rt = max_sectors - resync;    /* number of remaining sectors */
+	sector_div(rt, db/32+1);
+	rt *= dt;
+	rt >>= 5;
+
+	seq_printf(seq, " finish=%lu.%lumin", (unsigned long)rt / 60,
+		   ((unsigned long)rt % 60)/6);
 
 	seq_printf(seq, " speed=%ldK/sec", db/2/dt);
 }
@@ -5288,7 +6995,7 @@ static void *md_seq_start(struct seq_file *seq, loff_t *pos)
 {
 	struct list_head *tmp;
 	loff_t l = *pos;
-	mddev_t *mddev;
+	struct mddev *mddev;
 
 	if (l >= 0x10000)
 		return NULL;
@@ -5299,7 +7006,7 @@ static void *md_seq_start(struct seq_file *seq, loff_t *pos)
 	spin_lock(&all_mddevs_lock);
 	list_for_each(tmp,&all_mddevs)
 		if (!l--) {
-			mddev = list_entry(tmp, mddev_t, all_mddevs);
+			mddev = list_entry(tmp, struct mddev, all_mddevs);
 			mddev_get(mddev);
 			spin_unlock(&all_mddevs_lock);
 			return mddev;
@@ -5313,7 +7020,7 @@ static void *md_seq_start(struct seq_file *seq, loff_t *pos)
 static void *md_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 {
 	struct list_head *tmp;
-	mddev_t *next_mddev, *mddev = v;
+	struct mddev *next_mddev, *mddev = v;
 	
 	++*pos;
 	if (v == (void*)2)
@@ -5325,7 +7032,7 @@ static void *md_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 	else
 		tmp = mddev->all_mddevs.next;
 	if (tmp != &all_mddevs)
-		next_mddev = mddev_get(list_entry(tmp,mddev_t,all_mddevs));
+		next_mddev = mddev_get(list_entry(tmp,struct mddev,all_mddevs));
 	else {
 		next_mddev = (void*)2;
 		*pos = 0x10000;
@@ -5340,27 +7047,20 @@ static void *md_seq_next(struct seq_file *seq, void *v, loff_t *pos)
 
 static void md_seq_stop(struct seq_file *seq, void *v)
 {
-	mddev_t *mddev = v;
+	struct mddev *mddev = v;
 
 	if (mddev && v != (void*)1 && v != (void*)2)
 		mddev_put(mddev);
 }
 
-struct mdstat_info {
-	int event;
-};
-
 static int md_seq_show(struct seq_file *seq, void *v)
 {
-	mddev_t *mddev = v;
-	sector_t size;
-	struct list_head *tmp2;
-	mdk_rdev_t *rdev;
-	struct mdstat_info *mi = seq->private;
-	struct bitmap *bitmap;
+	struct mddev *mddev = v;
+	sector_t sectors;
+	struct md_rdev *rdev;
 
 	if (v == (void*)1) {
-		struct mdk_personality *pers;
+		struct md_personality *pers;
 		seq_printf(seq, "Personalities : ");
 		spin_lock(&pers_lock);
 		list_for_each_entry(pers, &pers_list, list)
@@ -5368,7 +7068,7 @@ static int md_seq_show(struct seq_file *seq, void *v)
 
 		spin_unlock(&pers_lock);
 		seq_printf(seq, "\n");
-		mi->event = atomic_read(&md_event_count);
+		seq->poll_event = atomic_read(&md_event_count);
 		return 0;
 	}
 	if (v == (void*)2) {
@@ -5390,8 +7090,8 @@ static int md_seq_show(struct seq_file *seq, void *v)
 			seq_printf(seq, " %s", mddev->pers->name);
 		}
 
-		size = 0;
-		rdev_for_each(rdev, tmp2, mddev) {
+		sectors = 0;
+		rdev_for_each(rdev, mddev) {
 			char b[BDEVNAME_SIZE];
 			seq_printf(seq, " %s[%d]",
 				bdevname(rdev->bdev,b), rdev->desc_nr);
@@ -5400,9 +7100,12 @@ static int md_seq_show(struct seq_file *seq, void *v)
 			if (test_bit(Faulty, &rdev->flags)) {
 				seq_printf(seq, "(F)");
 				continue;
-			} else if (rdev->raid_disk < 0)
+			}
+			if (rdev->raid_disk < 0)
 				seq_printf(seq, "(S)"); /* spare */
-			size += rdev->size;
+			if (test_bit(Replacement, &rdev->flags))
+				seq_printf(seq, "(R)");
+			sectors += rdev->sectors;
 		}
 
 		if (!list_empty(&mddev->disks)) {
@@ -5412,7 +7115,7 @@ static int md_seq_show(struct seq_file *seq, void *v)
 					   mddev->array_sectors / 2);
 			else
 				seq_printf(seq, "\n      %llu blocks",
-					   (unsigned long long)size);
+					   (unsigned long long)sectors / 2);
 		}
 		if (mddev->persistent) {
 			if (mddev->major_version != 0 ||
@@ -5428,13 +7131,13 @@ static int md_seq_show(struct seq_file *seq, void *v)
 			seq_printf(seq, " super non-persistent");
 
 		if (mddev->pers) {
-			mddev->pers->status (seq, mddev);
+			mddev->pers->status(seq, mddev);
 	 		seq_printf(seq, "\n      ");
 			if (mddev->pers->sync_request) {
 				if (mddev->curr_resync > 2) {
-					status_resync (seq, mddev);
+					status_resync(seq, mddev);
 					seq_printf(seq, "\n      ");
-				} else if (mddev->curr_resync == 1 || mddev->curr_resync == 2)
+				} else if (mddev->curr_resync >= 1)
 					seq_printf(seq, "\tresync=DELAYED\n      ");
 				else if (mddev->recovery_cp < MaxSector)
 					seq_printf(seq, "\tresync=PENDING\n      ");
@@ -5442,27 +7145,7 @@ static int md_seq_show(struct seq_file *seq, void *v)
 		} else
 			seq_printf(seq, "\n       ");
 
-		if ((bitmap = mddev->bitmap)) {
-			unsigned long chunk_kb;
-			unsigned long flags;
-			spin_lock_irqsave(&bitmap->lock, flags);
-			chunk_kb = bitmap->chunksize >> 10;
-			seq_printf(seq, "bitmap: %lu/%lu pages [%luKB], "
-				"%lu%s chunk",
-				bitmap->pages - bitmap->missing_pages,
-				bitmap->pages,
-				(bitmap->pages - bitmap->missing_pages)
-					<< (PAGE_SHIFT - 10),
-				chunk_kb ? chunk_kb : bitmap->chunksize,
-				chunk_kb ? "KB" : "B");
-			if (bitmap->file) {
-				seq_printf(seq, ", file: ");
-				seq_path(seq, &bitmap->file->f_path, " \t\n");
-			}
-
-			seq_printf(seq, "\n");
-			spin_unlock_irqrestore(&bitmap->lock, flags);
-		}
+		bitmap_status(seq, mddev->bitmap);
 
 		seq_printf(seq, "\n");
 	}
@@ -5471,7 +7154,7 @@ static int md_seq_show(struct seq_file *seq, void *v)
 	return 0;
 }
 
-static struct seq_operations md_seq_ops = {
+static const struct seq_operations md_seq_ops = {
 	.start  = md_seq_start,
 	.next   = md_seq_next,
 	.stop   = md_seq_stop,
@@ -5480,34 +7163,32 @@ static struct seq_operations md_seq_ops = {
 
 static int md_seq_open(struct inode *inode, struct file *file)
 {
+	struct seq_file *seq;
 	int error;
-	struct mdstat_info *mi = kmalloc(sizeof(*mi), GFP_KERNEL);
-	if (mi == NULL)
-		return -ENOMEM;
 
 	error = seq_open(file, &md_seq_ops);
 	if (error)
-		kfree(mi);
-	else {
-		struct seq_file *p = file->private_data;
-		p->private = mi;
-		mi->event = atomic_read(&md_event_count);
-	}
+		return error;
+
+	seq = file->private_data;
+	seq->poll_event = atomic_read(&md_event_count);
 	return error;
 }
 
+static int md_unloading;
 static unsigned int mdstat_poll(struct file *filp, poll_table *wait)
 {
-	struct seq_file *m = filp->private_data;
-	struct mdstat_info *mi = m->private;
+	struct seq_file *seq = filp->private_data;
 	int mask;
 
+	if (md_unloading)
+		return POLLIN|POLLRDNORM|POLLERR|POLLPRI;;
 	poll_wait(filp, &md_event_waiters, wait);
 
 	/* always allow read */
 	mask = POLLIN | POLLRDNORM;
 
-	if (mi->event != atomic_read(&md_event_count))
+	if (seq->poll_event != atomic_read(&md_event_count))
 		mask |= POLLERR | POLLPRI;
 	return mask;
 }
@@ -5521,7 +7202,7 @@ static const struct file_operations md_seq_fops = {
 	.poll		= mdstat_poll,
 };
 
-int register_md_personality(struct mdk_personality *p)
+int register_md_personality(struct md_personality *p)
 {
 	spin_lock(&pers_lock);
 	list_add_tail(&p->list, &pers_list);
@@ -5530,7 +7211,7 @@ int register_md_personality(struct mdk_personality *p)
 	return 0;
 }
 
-int unregister_md_personality(struct mdk_personality *p)
+int unregister_md_personality(struct md_personality *p)
 {
 	printk(KERN_INFO "md: %s personality unregistered\n", p->name);
 	spin_lock(&pers_lock);
@@ -5539,19 +7220,19 @@ int unregister_md_personality(struct mdk_personality *p)
 	return 0;
 }
 
-static int is_mddev_idle(mddev_t *mddev)
+static int is_mddev_idle(struct mddev *mddev, int init)
 {
-	mdk_rdev_t * rdev;
+	struct md_rdev * rdev;
 	int idle;
-	long curr_events;
+	int curr_events;
 
 	idle = 1;
 	rcu_read_lock();
 	rdev_for_each_rcu(rdev, mddev) {
 		struct gendisk *disk = rdev->bdev->bd_contains->bd_disk;
-		curr_events = disk_stat_read(disk, sectors[0]) + 
-				disk_stat_read(disk, sectors[1]) - 
-				atomic_read(&disk->sync_io);
+		curr_events = (int)part_stat_read(&disk->part0, sectors[0]) +
+			      (int)part_stat_read(&disk->part0, sectors[1]) -
+			      atomic_read(&disk->sync_io);
 		/* sync IO will cause sync_io to increase before the disk_stats
 		 * as sync_io is counted when a request starts, and
 		 * disk_stats is counted when it completes.
@@ -5574,7 +7255,7 @@ static int is_mddev_idle(mddev_t *mddev)
 		 * always make curr_events less than last_events.
 		 *
 		 */
-		if (curr_events - rdev->last_events > 4096) {
+		if (init || curr_events - rdev->last_events > 64) {
 			rdev->last_events = curr_events;
 			idle = 0;
 		}
@@ -5583,13 +7264,14 @@ static int is_mddev_idle(mddev_t *mddev)
 	return idle;
 }
 
-void md_done_sync(mddev_t *mddev, int blocks, int ok)
+void md_done_sync(struct mddev *mddev, int blocks, int ok)
 {
 	/* another "blocks" (512byte) blocks have been synced */
 	atomic_sub(blocks, &mddev->recovery_active);
 	wake_up(&mddev->recovery_wait);
 	if (!ok) {
 		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		set_bit(MD_RECOVERY_ERROR, &mddev->recovery);
 		md_wakeup_thread(mddev->thread);
 		// stop recovery, signal do_sync ....
 	}
@@ -5601,7 +7283,7 @@ void md_done_sync(mddev_t *mddev, int blocks, int ok)
  * in superblock) before writing, schedule a superblock update
  * and wait for it to complete.
  */
-void md_write_start(mddev_t *mddev, struct bio *bi)
+void md_write_start(struct mddev *mddev, struct bio *bi)
 {
 	int did_change = 0;
 	if (bio_data_dir(bi) != WRITE)
@@ -5624,19 +7306,19 @@ void md_write_start(mddev_t *mddev, struct bio *bi)
 		if (mddev->in_sync) {
 			mddev->in_sync = 0;
 			set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+			set_bit(MD_CHANGE_PENDING, &mddev->flags);
 			md_wakeup_thread(mddev->thread);
 			did_change = 1;
 		}
 		spin_unlock_irq(&mddev->write_lock);
 	}
 	if (did_change)
-		sysfs_notify(&mddev->kobj, NULL, "array_state");
+		sysfs_notify_dirent_safe(mddev->sysfs_state);
 	wait_event(mddev->sb_wait,
-		   !test_bit(MD_CHANGE_CLEAN, &mddev->flags) &&
 		   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
 }
 
-void md_write_end(mddev_t *mddev)
+void md_write_end(struct mddev *mddev)
 {
 	if (atomic_dec_and_test(&mddev->writes_pending)) {
 		if (mddev->safemode == 2)
@@ -5655,7 +7337,7 @@ void md_write_end(mddev_t *mddev)
  * In the ->external case MD_CHANGE_CLEAN can not be cleared until mddev->lock
  * is dropped, so return -EAGAIN after notifying userspace.
  */
-int md_allow_write(mddev_t *mddev)
+int md_allow_write(struct mddev *mddev)
 {
 	if (!mddev->pers)
 		return 0;
@@ -5668,16 +7350,17 @@ int md_allow_write(mddev_t *mddev)
 	if (mddev->in_sync) {
 		mddev->in_sync = 0;
 		set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+		set_bit(MD_CHANGE_PENDING, &mddev->flags);
 		if (mddev->safemode_delay &&
 		    mddev->safemode == 0)
 			mddev->safemode = 1;
 		spin_unlock_irq(&mddev->write_lock);
 		md_update_sb(mddev, 0);
-		sysfs_notify(&mddev->kobj, NULL, "array_state");
+		sysfs_notify_dirent_safe(mddev->sysfs_state);
 	} else
 		spin_unlock_irq(&mddev->write_lock);
 
-	if (test_bit(MD_CHANGE_CLEAN, &mddev->flags))
+	if (test_bit(MD_CHANGE_PENDING, &mddev->flags))
 		return -EAGAIN;
 	else
 		return 0;
@@ -5686,40 +7369,49 @@ EXPORT_SYMBOL_GPL(md_allow_write);
 
 #define SYNC_MARKS	10
 #define	SYNC_MARK_STEP	(3*HZ)
-void md_do_sync(mddev_t *mddev)
+#define UPDATE_FREQUENCY (5*60*HZ)
+void md_do_sync(struct md_thread *thread)
 {
-	mddev_t *mddev2;
+	struct mddev *mddev = thread->mddev;
+	struct mddev *mddev2;
 	unsigned int currspeed = 0,
 		 window;
 	sector_t max_sectors,j, io_sectors;
 	unsigned long mark[SYNC_MARKS];
+	unsigned long update_time;
 	sector_t mark_cnt[SYNC_MARKS];
 	int last_mark,m;
 	struct list_head *tmp;
 	sector_t last_check;
 	int skipped = 0;
-	struct list_head *rtmp;
-	mdk_rdev_t *rdev;
-	char *desc;
+	struct md_rdev *rdev;
+	char *desc, *action = NULL;
+	struct blk_plug plug;
 
 	/* just incase thread restarts... */
 	if (test_bit(MD_RECOVERY_DONE, &mddev->recovery))
 		return;
-	if (mddev->ro) /* never try to sync a read-only array */
+	if (mddev->ro) {/* never try to sync a read-only array */
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 		return;
+	}
 
 	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
-		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
+		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)) {
 			desc = "data-check";
-		else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+			action = "check";
+		} else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 			desc = "requested-resync";
-		else
+			action = "repair";
+		} else
 			desc = "resync";
 	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 		desc = "reshape";
 	else
 		desc = "recovery";
 
+	mddev->last_sync_action = action ?: desc;
+
 	/* we overload curr_resync somewhat here.
 	 * 0 == not engaged in resync at all
 	 * 2 == checking that there is no conflict with another sync
@@ -5740,10 +7432,8 @@ void md_do_sync(mddev_t *mddev)
 		mddev->curr_resync = 2;
 
 	try_again:
-		if (kthread_should_stop()) {
-			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
 			goto skip;
-		}
 		for_each_mddev(mddev2, tmp) {
 			if (mddev2 == mddev)
 				continue;
@@ -5761,14 +7451,20 @@ void md_do_sync(mddev_t *mddev)
 					 * time 'round when curr_resync == 2
 					 */
 					continue;
-				prepare_to_wait(&resync_wait, &wq, TASK_UNINTERRUPTIBLE);
-				if (!kthread_should_stop() &&
+				/* We need to wait 'interruptible' so as not to
+				 * contribute to the load average, and not to
+				 * be caught by 'softlockup'
+				 */
+				prepare_to_wait(&resync_wait, &wq, TASK_INTERRUPTIBLE);
+				if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
 				    mddev2->curr_resync >= mddev->curr_resync) {
 					printk(KERN_INFO "md: delaying %s of %s"
 					       " until %s has finished (they"
 					       " share one or more physical units)\n",
 					       desc, mdname(mddev), mdname(mddev2));
 					mddev_put(mddev2);
+					if (signal_pending(current))
+						flush_signals(current);
 					schedule();
 					finish_wait(&resync_wait, &wq);
 					goto try_again;
@@ -5784,7 +7480,7 @@ void md_do_sync(mddev_t *mddev)
 		 * which defaults to physical size, but can be virtual size
 		 */
 		max_sectors = mddev->resync_max_sectors;
-		mddev->resync_mismatches = 0;
+		atomic64_set(&mddev->resync_mismatches, 0);
 		/* we don't use the checkpoint if there's a bitmap */
 		if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
 			j = mddev->resync_min;
@@ -5792,17 +7488,32 @@ void md_do_sync(mddev_t *mddev)
 			j = mddev->recovery_cp;
 
 	} else if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
-		max_sectors = mddev->size << 1;
+		max_sectors = mddev->resync_max_sectors;
 	else {
 		/* recovery follows the physical size of devices */
-		max_sectors = mddev->size << 1;
+		max_sectors = mddev->dev_sectors;
 		j = MaxSector;
-		rdev_for_each(rdev, rtmp, mddev)
+		rcu_read_lock();
+		rdev_for_each_rcu(rdev, mddev)
 			if (rdev->raid_disk >= 0 &&
 			    !test_bit(Faulty, &rdev->flags) &&
 			    !test_bit(In_sync, &rdev->flags) &&
 			    rdev->recovery_offset < j)
 				j = rdev->recovery_offset;
+		rcu_read_unlock();
+
+		/* If there is a bitmap, we need to make sure all
+		 * writes that started before we added a spare
+		 * complete before we start doing a recovery.
+		 * Otherwise the write might complete and (via
+		 * bitmap_endwrite) set a bit in the bitmap after the
+		 * recovery has checked that bit and skipped that
+		 * region.
+		 */
+		if (mddev->bitmap) {
+			mddev->pers->quiesce(mddev, 1);
+			mddev->pers->quiesce(mddev, 0);
+		}
 	}
 
 	printk(KERN_INFO "md: %s of RAID array %s\n", desc, mdname(mddev));
@@ -5812,7 +7523,7 @@ void md_do_sync(mddev_t *mddev)
 	       "(but not more than %d KB/sec) for %s.\n",
 	       speed_max(mddev), desc);
 
-	is_mddev_idle(mddev); /* this also initializes IO event counters */
+	is_mddev_idle(mddev, 1); /* this initializes IO event counters */
 
 	io_sectors = 0;
 	for (m = 0; m < SYNC_MARKS; m++) {
@@ -5827,36 +7538,71 @@ void md_do_sync(mddev_t *mddev)
 	 * Tune reconstruction:
 	 */
 	window = 32*(PAGE_SIZE/512);
-	printk(KERN_INFO "md: using %dk window, over a total of %llu blocks.\n",
-		window/2,(unsigned long long) max_sectors/2);
+	printk(KERN_INFO "md: using %dk window, over a total of %lluk.\n",
+		window/2, (unsigned long long)max_sectors/2);
 
 	atomic_set(&mddev->recovery_active, 0);
 	last_check = 0;
 
 	if (j>2) {
-		printk(KERN_INFO 
+		printk(KERN_INFO
 		       "md: resuming %s of %s from checkpoint.\n",
 		       desc, mdname(mddev));
 		mddev->curr_resync = j;
-	}
+	} else
+		mddev->curr_resync = 3; /* no longer delayed */
+	mddev->curr_resync_completed = j;
+	sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+	md_new_event(mddev);
+	update_time = jiffies;
 
+	blk_start_plug(&plug);
 	while (j < max_sectors) {
 		sector_t sectors;
 
 		skipped = 0;
-		if (j >= mddev->resync_max) {
-			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+
+		if (!test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+		    ((mddev->curr_resync > mddev->curr_resync_completed &&
+		      (mddev->curr_resync - mddev->curr_resync_completed)
+		      > (max_sectors >> 4)) ||
+		     time_after_eq(jiffies, update_time + UPDATE_FREQUENCY) ||
+		     (j - mddev->curr_resync_completed)*2
+		     >= mddev->resync_max - mddev->curr_resync_completed
+			    )) {
+			/* time to update curr_resync_completed */
 			wait_event(mddev->recovery_wait,
-				   mddev->resync_max > j
-				   || kthread_should_stop());
+				   atomic_read(&mddev->recovery_active) == 0);
+			mddev->curr_resync_completed = j;
+			if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
+			    j > mddev->recovery_cp)
+				mddev->recovery_cp = j;
+			update_time = jiffies;
+			set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+		}
+
+		while (j >= mddev->resync_max &&
+		       !test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
+			/* As this condition is controlled by user-space,
+			 * we can block indefinitely, so use '_interruptible'
+			 * to avoid triggering warnings.
+			 */
+			flush_signals(current); /* just in case */
+			wait_event_interruptible(mddev->recovery_wait,
+						 mddev->resync_max > j
+						 || test_bit(MD_RECOVERY_INTR,
+							     &mddev->recovery));
 		}
-		if (kthread_should_stop())
-			goto interrupted;
+
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			break;
+
 		sectors = mddev->pers->sync_request(mddev, j, &skipped,
 						  currspeed < speed_min(mddev));
 		if (sectors == 0) {
 			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-			goto out;
+			break;
 		}
 
 		if (!skipped) { /* actual IO requested */
@@ -5864,11 +7610,15 @@ void md_do_sync(mddev_t *mddev)
 			atomic_add(sectors, &mddev->recovery_active);
 		}
 
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			break;
+
 		j += sectors;
-		if (j>1) mddev->curr_resync = j;
+		if (j > 2)
+			mddev->curr_resync = j;
 		mddev->curr_mark_cnt = io_sectors;
 		if (last_check == 0)
-			/* this is the earliers that rebuilt will be
+			/* this is the earliest that rebuild will be
 			 * visible in /proc/mdstat
 			 */
 			md_new_event(mddev);
@@ -5877,10 +7627,6 @@ void md_do_sync(mddev_t *mddev)
 			continue;
 
 		last_check = io_sectors;
-
-		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
-			break;
-
 	repeat:
 		if (time_after_eq(jiffies, mark[last_mark] + SYNC_MARK_STEP )) {
 			/* step marks */
@@ -5893,10 +7639,8 @@ void md_do_sync(mddev_t *mddev)
 			last_mark = next;
 		}
 
-
-		if (kthread_should_stop())
-			goto interrupted;
-
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			break;
 
 		/*
 		 * this loop exits only if either when we are slower than
@@ -5906,7 +7650,6 @@ void md_do_sync(mddev_t *mddev)
 		 * about not overloading the IO subsystem. (things like an
 		 * e2fsck being done on the RAID array should execute fast)
 		 */
-		blk_unplug(mddev->queue);
 		cond_resched();
 
 		currspeed = ((unsigned long)(io_sectors-mddev->resync_mark_cnt))/2
@@ -5914,19 +7657,19 @@ void md_do_sync(mddev_t *mddev)
 
 		if (currspeed > speed_min(mddev)) {
 			if ((currspeed > speed_max(mddev)) ||
-					!is_mddev_idle(mddev)) {
+					!is_mddev_idle(mddev, 0)) {
 				msleep(500);
 				goto repeat;
 			}
 		}
 	}
-	printk(KERN_INFO "md: %s: %s done.\n",mdname(mddev), desc);
+	printk(KERN_INFO "md: %s: %s %s.\n",mdname(mddev), desc,
+	       test_bit(MD_RECOVERY_INTR, &mddev->recovery)
+	       ? "interrupted" : "done");
 	/*
 	 * this also signals 'finished resyncing' to md_stop
 	 */
- out:
-	blk_unplug(mddev->queue);
-
+	blk_finish_plug(&plug);
 	wait_event(mddev->recovery_wait, !atomic_read(&mddev->recovery_active));
 
 	/* tell personality that we are finished */
@@ -5940,95 +7683,106 @@ void md_do_sync(mddev_t *mddev)
 					printk(KERN_INFO
 					       "md: checkpointing %s of %s.\n",
 					       desc, mdname(mddev));
-					mddev->recovery_cp = mddev->curr_resync;
+					if (test_bit(MD_RECOVERY_ERROR,
+						&mddev->recovery))
+						mddev->recovery_cp =
+							mddev->curr_resync_completed;
+					else
+						mddev->recovery_cp =
+							mddev->curr_resync;
 				}
 			} else
 				mddev->recovery_cp = MaxSector;
 		} else {
 			if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery))
 				mddev->curr_resync = MaxSector;
-			rdev_for_each(rdev, rtmp, mddev)
+			rcu_read_lock();
+			rdev_for_each_rcu(rdev, mddev)
 				if (rdev->raid_disk >= 0 &&
+				    mddev->delta_disks >= 0 &&
 				    !test_bit(Faulty, &rdev->flags) &&
 				    !test_bit(In_sync, &rdev->flags) &&
 				    rdev->recovery_offset < mddev->curr_resync)
 					rdev->recovery_offset = mddev->curr_resync;
+			rcu_read_unlock();
 		}
 	}
+ skip:
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
 
- skip:
+	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
+		/* We completed so min/max setting can be forgotten if used. */
+		if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+			mddev->resync_min = 0;
+		mddev->resync_max = MaxSector;
+	} else if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+		mddev->resync_min = mddev->curr_resync_completed;
 	mddev->curr_resync = 0;
-	mddev->resync_min = 0;
-	mddev->resync_max = MaxSector;
-	sysfs_notify(&mddev->kobj, NULL, "sync_completed");
 	wake_up(&resync_wait);
 	set_bit(MD_RECOVERY_DONE, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
 	return;
-
- interrupted:
-	/*
-	 * got a signal, exit.
-	 */
-	printk(KERN_INFO
-	       "md: md_do_sync() got signal ... exiting\n");
-	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-	goto out;
-
 }
 EXPORT_SYMBOL_GPL(md_do_sync);
 
-
-static int remove_and_add_spares(mddev_t *mddev)
+static int remove_and_add_spares(struct mddev *mddev,
+				 struct md_rdev *this)
 {
-	mdk_rdev_t *rdev;
-	struct list_head *rtmp;
+	struct md_rdev *rdev;
 	int spares = 0;
+	int removed = 0;
 
-	rdev_for_each(rdev, rtmp, mddev)
-		if (rdev->raid_disk >= 0 &&
+	rdev_for_each(rdev, mddev)
+		if ((this == NULL || rdev == this) &&
+		    rdev->raid_disk >= 0 &&
 		    !test_bit(Blocked, &rdev->flags) &&
 		    (test_bit(Faulty, &rdev->flags) ||
 		     ! test_bit(In_sync, &rdev->flags)) &&
 		    atomic_read(&rdev->nr_pending)==0) {
 			if (mddev->pers->hot_remove_disk(
-				    mddev, rdev->raid_disk)==0) {
-				char nm[20];
-				sprintf(nm,"rd%d", rdev->raid_disk);
-				sysfs_remove_link(&mddev->kobj, nm);
+				    mddev, rdev) == 0) {
+				sysfs_unlink_rdev(mddev, rdev);
 				rdev->raid_disk = -1;
+				removed++;
 			}
 		}
+	if (removed && mddev->kobj.sd)
+		sysfs_notify(&mddev->kobj, NULL, "degraded");
 
-	if (mddev->degraded && ! mddev->ro) {
-		rdev_for_each(rdev, rtmp, mddev) {
-			if (rdev->raid_disk >= 0 &&
-			    !test_bit(In_sync, &rdev->flags) &&
-			    !test_bit(Blocked, &rdev->flags))
-				spares++;
-			if (rdev->raid_disk < 0
-			    && !test_bit(Faulty, &rdev->flags)) {
-				rdev->recovery_offset = 0;
-				if (mddev->pers->
-				    hot_add_disk(mddev, rdev) == 0) {
-					char nm[20];
-					sprintf(nm, "rd%d", rdev->raid_disk);
-					if (sysfs_create_link(&mddev->kobj,
-							      &rdev->kobj, nm))
-						printk(KERN_WARNING
-						       "md: cannot register "
-						       "%s for %s\n",
-						       nm, mdname(mddev));
-					spares++;
-					md_new_event(mddev);
-				} else
-					break;
-			}
+	if (this)
+		goto no_add;
+
+	rdev_for_each(rdev, mddev) {
+		if (rdev->raid_disk >= 0 &&
+		    !test_bit(In_sync, &rdev->flags) &&
+		    !test_bit(Faulty, &rdev->flags))
+			spares++;
+		if (rdev->raid_disk >= 0)
+			continue;
+		if (test_bit(Faulty, &rdev->flags))
+			continue;
+		if (mddev->ro &&
+		    ! (rdev->saved_raid_disk >= 0 &&
+		       !test_bit(Bitmap_sync, &rdev->flags)))
+			continue;
+
+		if (rdev->saved_raid_disk < 0)
+			rdev->recovery_offset = 0;
+		if (mddev->pers->
+		    hot_add_disk(mddev, rdev) == 0) {
+			if (sysfs_link_rdev(mddev, rdev))
+				/* failure here is OK */;
+			spares++;
+			md_new_event(mddev);
+			set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		}
 	}
+no_add:
+	if (removed)
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 	return spares;
 }
+
 /*
  * This routine is regularly called by all per-raid-array threads to
  * deal with generic issues like resync and super-block update.
@@ -6051,20 +7805,13 @@ static int remove_and_add_spares(mddev_t *mddev)
  *  5/ If array is degraded, try to add spares devices
  *  6/ If array has spares or is not in-sync, start a resync thread.
  */
-void md_check_recovery(mddev_t *mddev)
+void md_check_recovery(struct mddev *mddev)
 {
-	mdk_rdev_t *rdev;
-	struct list_head *rtmp;
-
+	if (mddev->suspended)
+		return;
 
 	if (mddev->bitmap)
-		bitmap_daemon_work(mddev->bitmap);
-
-	if (test_and_clear_bit(MD_NOTIFY_ARRAY_STATE, &mddev->flags))
-		sysfs_notify(&mddev->kobj, NULL, "array_state");
-
-	if (mddev->ro)
-		return;
+		bitmap_daemon_work(mddev);
 
 	if (signal_pending(current)) {
 		if (mddev->pers->sync_request && !mddev->external) {
@@ -6078,7 +7825,7 @@ void md_check_recovery(mddev_t *mddev)
 	if (mddev->ro && !test_bit(MD_RECOVERY_NEEDED, &mddev->recovery))
 		return;
 	if ( ! (
-		(mddev->flags && !mddev->external) ||
+		(mddev->flags & MD_UPDATE_SB_FLAGS & ~ (1<<MD_CHANGE_PENDING)) ||
 		test_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
 		test_bit(MD_RECOVERY_DONE, &mddev->recovery) ||
 		(mddev->external == 0 && mddev->safemode == 1) ||
@@ -6091,10 +7838,19 @@ void md_check_recovery(mddev_t *mddev)
 		int spares = 0;
 
 		if (mddev->ro) {
-			/* Only thing we do on a ro array is remove
-			 * failed devices.
+			/* On a read-only array we can:
+			 * - remove failed devices
+			 * - add already-in_sync devices if the array itself
+			 *   is in-sync.
+			 * As we only add devices that are already in-sync,
+			 * we can activate the spares immediately.
+			 */
+			remove_and_add_spares(mddev, NULL);
+			/* There is no thread, but we need to call
+			 * ->spare_active and clear saved_raid_disk
 			 */
-			remove_and_add_spares(mddev);
+			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+			md_reap_sync_thread(mddev);
 			clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 			goto unlock;
 		}
@@ -6108,24 +7864,18 @@ void md_check_recovery(mddev_t *mddev)
 			    mddev->recovery_cp == MaxSector) {
 				mddev->in_sync = 1;
 				did_change = 1;
-				if (mddev->persistent)
-					set_bit(MD_CHANGE_CLEAN, &mddev->flags);
+				set_bit(MD_CHANGE_CLEAN, &mddev->flags);
 			}
 			if (mddev->safemode == 1)
 				mddev->safemode = 0;
 			spin_unlock_irq(&mddev->write_lock);
 			if (did_change)
-				sysfs_notify(&mddev->kobj, NULL, "array_state");
+				sysfs_notify_dirent_safe(mddev->sysfs_state);
 		}
 
-		if (mddev->flags)
+		if (mddev->flags & MD_UPDATE_SB_FLAGS)
 			md_update_sb(mddev, 0);
 
-		rdev_for_each(rdev, rtmp, mddev)
-			if (test_and_clear_bit(StateChanged, &rdev->flags))
-				sysfs_notify(&rdev->kobj, NULL, "state");
-
-
 		if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery) &&
 		    !test_bit(MD_RECOVERY_DONE, &mddev->recovery)) {
 			/* resync/recovery still happening */
@@ -6133,60 +7883,38 @@ void md_check_recovery(mddev_t *mddev)
 			goto unlock;
 		}
 		if (mddev->sync_thread) {
-			/* resync has finished, collect result */
-			md_unregister_thread(mddev->sync_thread);
-			mddev->sync_thread = NULL;
-			if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
-			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
-				/* success...*/
-				/* activate any spares */
-				if (mddev->pers->spare_active(mddev))
-					sysfs_notify(&mddev->kobj, NULL,
-						     "degraded");
-			}
-			md_update_sb(mddev, 1);
-
-			/* if array is no-longer degraded, then any saved_raid_disk
-			 * information must be scrapped
-			 */
-			if (!mddev->degraded)
-				rdev_for_each(rdev, rtmp, mddev)
-					rdev->saved_raid_disk = -1;
-
-			mddev->recovery = 0;
-			/* flag recovery needed just to double check */
-			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
-			sysfs_notify(&mddev->kobj, NULL, "sync_action");
-			md_new_event(mddev);
+			md_reap_sync_thread(mddev);
 			goto unlock;
 		}
 		/* Set RUNNING before clearing NEEDED to avoid
 		 * any transients in the value of "sync_action".
 		 */
+		mddev->curr_resync_completed = 0;
 		set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
-		clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 		/* Clear some bits that don't mean anything, but
 		 * might be left set
 		 */
 		clear_bit(MD_RECOVERY_INTR, &mddev->recovery);
 		clear_bit(MD_RECOVERY_DONE, &mddev->recovery);
 
-		if (test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
+		if (!test_and_clear_bit(MD_RECOVERY_NEEDED, &mddev->recovery) ||
+		    test_bit(MD_RECOVERY_FROZEN, &mddev->recovery))
 			goto unlock;
 		/* no recovery is running.
 		 * remove any failed drives, then
 		 * add spares if possible.
-		 * Spare are also removed and re-added, to allow
+		 * Spares are also removed and re-added, to allow
 		 * the personality to fail the re-add.
 		 */
 
 		if (mddev->reshape_position != MaxSector) {
-			if (mddev->pers->check_reshape(mddev) != 0)
+			if (mddev->pers->check_reshape == NULL ||
+			    mddev->pers->check_reshape(mddev) != 0)
 				/* Cannot proceed */
 				goto unlock;
 			set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
 			clear_bit(MD_RECOVERY_RECOVER, &mddev->recovery);
-		} else if ((spares = remove_and_add_spares(mddev))) {
+		} else if ((spares = remove_and_add_spares(mddev, NULL))) {
 			clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 			clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
 			clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
@@ -6199,7 +7927,7 @@ void md_check_recovery(mddev_t *mddev)
 			goto unlock;
 
 		if (mddev->pers->sync_request) {
-			if (spares && mddev->bitmap && ! mddev->bitmap->file) {
+			if (spares) {
 				/* We are adding a device or devices to an array
 				 * which has the bitmap stored on all devices.
 				 * So make sure all bitmap pages get written
@@ -6208,66 +7936,627 @@ void md_check_recovery(mddev_t *mddev)
 			}
 			mddev->sync_thread = md_register_thread(md_do_sync,
 								mddev,
-								"%s_resync");
+								"resync");
 			if (!mddev->sync_thread) {
 				printk(KERN_ERR "%s: could not start resync"
 					" thread...\n", 
 					mdname(mddev));
 				/* leave the spares where they are, it shouldn't hurt */
-				mddev->recovery = 0;
+				clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+				clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+				clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
+				clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
+				clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
 			} else
 				md_wakeup_thread(mddev->sync_thread);
-			sysfs_notify(&mddev->kobj, NULL, "sync_action");
+			sysfs_notify_dirent_safe(mddev->sysfs_action);
 			md_new_event(mddev);
 		}
 	unlock:
+		wake_up(&mddev->sb_wait);
+
 		if (!mddev->sync_thread) {
 			clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
 			if (test_and_clear_bit(MD_RECOVERY_RECOVER,
 					       &mddev->recovery))
-				sysfs_notify(&mddev->kobj, NULL, "sync_action");
+				if (mddev->sysfs_action)
+					sysfs_notify_dirent_safe(mddev->sysfs_action);
 		}
 		mddev_unlock(mddev);
 	}
 }
 
-void md_wait_for_blocked_rdev(mdk_rdev_t *rdev, mddev_t *mddev)
+void md_reap_sync_thread(struct mddev *mddev)
 {
-	sysfs_notify(&rdev->kobj, NULL, "state");
+	struct md_rdev *rdev;
+
+	/* resync has finished, collect result */
+	md_unregister_thread(&mddev->sync_thread);
+	wake_up(&resync_wait);
+	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery) &&
+	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
+		/* success...*/
+		/* activate any spares */
+		if (mddev->pers->spare_active(mddev)) {
+			sysfs_notify(&mddev->kobj, NULL,
+				     "degraded");
+			set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		}
+	}
+	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    mddev->pers->finish_reshape)
+		mddev->pers->finish_reshape(mddev);
+
+	/* If array is no-longer degraded, then any saved_raid_disk
+	 * information must be scrapped.
+	 */
+	if (!mddev->degraded)
+		rdev_for_each(rdev, mddev)
+			rdev->saved_raid_disk = -1;
+
+	md_update_sb(mddev, 1);
+	clear_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+	clear_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
+	clear_bit(MD_RECOVERY_REQUESTED, &mddev->recovery);
+	clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
+	/* flag recovery needed just to double check */
+	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	sysfs_notify_dirent_safe(mddev->sysfs_action);
+	md_new_event(mddev);
+	if (mddev->event_work.func)
+		queue_work(md_misc_wq, &mddev->event_work);
+}
+
+void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev)
+{
+	sysfs_notify_dirent_safe(rdev->sysfs_state);
 	wait_event_timeout(rdev->blocked_wait,
-			   !test_bit(Blocked, &rdev->flags),
+			   !test_bit(Blocked, &rdev->flags) &&
+			   !test_bit(BlockedBadBlocks, &rdev->flags),
 			   msecs_to_jiffies(5000));
 	rdev_dec_pending(rdev, mddev);
 }
 EXPORT_SYMBOL(md_wait_for_blocked_rdev);
 
-static int md_notify_reboot(struct notifier_block *this,
-			    unsigned long code, void *x)
+void md_finish_reshape(struct mddev *mddev)
 {
-	struct list_head *tmp;
-	mddev_t *mddev;
+	/* called be personality module when reshape completes. */
+	struct md_rdev *rdev;
 
-	if ((code == SYS_DOWN) || (code == SYS_HALT) || (code == SYS_POWER_OFF)) {
+	rdev_for_each(rdev, mddev) {
+		if (rdev->data_offset > rdev->new_data_offset)
+			rdev->sectors += rdev->data_offset - rdev->new_data_offset;
+		else
+			rdev->sectors -= rdev->new_data_offset - rdev->data_offset;
+		rdev->data_offset = rdev->new_data_offset;
+	}
+}
+EXPORT_SYMBOL(md_finish_reshape);
 
-		printk(KERN_INFO "md: stopping all md devices.\n");
+/* Bad block management.
+ * We can record which blocks on each device are 'bad' and so just
+ * fail those blocks, or that stripe, rather than the whole device.
+ * Entries in the bad-block table are 64bits wide.  This comprises:
+ * Length of bad-range, in sectors: 0-511 for lengths 1-512
+ * Start of bad-range, sector offset, 54 bits (allows 8 exbibytes)
+ *  A 'shift' can be set so that larger blocks are tracked and
+ *  consequently larger devices can be covered.
+ * 'Acknowledged' flag - 1 bit. - the most significant bit.
+ *
+ * Locking of the bad-block table uses a seqlock so md_is_badblock
+ * might need to retry if it is very unlucky.
+ * We will sometimes want to check for bad blocks in a bi_end_io function,
+ * so we use the write_seqlock_irq variant.
+ *
+ * When looking for a bad block we specify a range and want to
+ * know if any block in the range is bad.  So we binary-search
+ * to the last range that starts at-or-before the given endpoint,
+ * (or "before the sector after the target range")
+ * then see if it ends after the given start.
+ * We return
+ *  0 if there are no known bad blocks in the range
+ *  1 if there are known bad block which are all acknowledged
+ * -1 if there are bad blocks which have not yet been acknowledged in metadata.
+ * plus the start/length of the first bad section we overlap.
+ */
+int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
+		   sector_t *first_bad, int *bad_sectors)
+{
+	int hi;
+	int lo;
+	u64 *p = bb->page;
+	int rv;
+	sector_t target = s + sectors;
+	unsigned seq;
+
+	if (bb->shift > 0) {
+		/* round the start down, and the end up */
+		s >>= bb->shift;
+		target += (1<<bb->shift) - 1;
+		target >>= bb->shift;
+		sectors = target - s;
+	}
+	/* 'target' is now the first block after the bad range */
+
+retry:
+	seq = read_seqbegin(&bb->lock);
+	lo = 0;
+	rv = 0;
+	hi = bb->count;
+
+	/* Binary search between lo and hi for 'target'
+	 * i.e. for the last range that starts before 'target'
+	 */
+	/* INVARIANT: ranges before 'lo' and at-or-after 'hi'
+	 * are known not to be the last range before target.
+	 * VARIANT: hi-lo is the number of possible
+	 * ranges, and decreases until it reaches 1
+	 */
+	while (hi - lo > 1) {
+		int mid = (lo + hi) / 2;
+		sector_t a = BB_OFFSET(p[mid]);
+		if (a < target)
+			/* This could still be the one, earlier ranges
+			 * could not. */
+			lo = mid;
+		else
+			/* This and later ranges are definitely out. */
+			hi = mid;
+	}
+	/* 'lo' might be the last that started before target, but 'hi' isn't */
+	if (hi > lo) {
+		/* need to check all range that end after 's' to see if
+		 * any are unacknowledged.
+		 */
+		while (lo >= 0 &&
+		       BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > s) {
+			if (BB_OFFSET(p[lo]) < target) {
+				/* starts before the end, and finishes after
+				 * the start, so they must overlap
+				 */
+				if (rv != -1 && BB_ACK(p[lo]))
+					rv = 1;
+				else
+					rv = -1;
+				*first_bad = BB_OFFSET(p[lo]);
+				*bad_sectors = BB_LEN(p[lo]);
+			}
+			lo--;
+		}
+	}
 
-		for_each_mddev(mddev, tmp)
-			if (mddev_trylock(mddev)) {
-				/* Force a switch to readonly even array
-				 * appears to still be in use.  Hence
-				 * the '100'.
+	if (read_seqretry(&bb->lock, seq))
+		goto retry;
+
+	return rv;
+}
+EXPORT_SYMBOL_GPL(md_is_badblock);
+
+/*
+ * Add a range of bad blocks to the table.
+ * This might extend the table, or might contract it
+ * if two adjacent ranges can be merged.
+ * We binary-search to find the 'insertion' point, then
+ * decide how best to handle it.
+ */
+static int md_set_badblocks(struct badblocks *bb, sector_t s, int sectors,
+			    int acknowledged)
+{
+	u64 *p;
+	int lo, hi;
+	int rv = 1;
+	unsigned long flags;
+
+	if (bb->shift < 0)
+		/* badblocks are disabled */
+		return 0;
+
+	if (bb->shift) {
+		/* round the start down, and the end up */
+		sector_t next = s + sectors;
+		s >>= bb->shift;
+		next += (1<<bb->shift) - 1;
+		next >>= bb->shift;
+		sectors = next - s;
+	}
+
+	write_seqlock_irqsave(&bb->lock, flags);
+
+	p = bb->page;
+	lo = 0;
+	hi = bb->count;
+	/* Find the last range that starts at-or-before 's' */
+	while (hi - lo > 1) {
+		int mid = (lo + hi) / 2;
+		sector_t a = BB_OFFSET(p[mid]);
+		if (a <= s)
+			lo = mid;
+		else
+			hi = mid;
+	}
+	if (hi > lo && BB_OFFSET(p[lo]) > s)
+		hi = lo;
+
+	if (hi > lo) {
+		/* we found a range that might merge with the start
+		 * of our new range
+		 */
+		sector_t a = BB_OFFSET(p[lo]);
+		sector_t e = a + BB_LEN(p[lo]);
+		int ack = BB_ACK(p[lo]);
+		if (e >= s) {
+			/* Yes, we can merge with a previous range */
+			if (s == a && s + sectors >= e)
+				/* new range covers old */
+				ack = acknowledged;
+			else
+				ack = ack && acknowledged;
+
+			if (e < s + sectors)
+				e = s + sectors;
+			if (e - a <= BB_MAX_LEN) {
+				p[lo] = BB_MAKE(a, e-a, ack);
+				s = e;
+			} else {
+				/* does not all fit in one range,
+				 * make p[lo] maximal
 				 */
-				do_md_stop (mddev, 1, 100);
-				mddev_unlock(mddev);
+				if (BB_LEN(p[lo]) != BB_MAX_LEN)
+					p[lo] = BB_MAKE(a, BB_MAX_LEN, ack);
+				s = a + BB_MAX_LEN;
 			}
-		/*
-		 * certain more exotic SCSI devices are known to be
-		 * volatile wrt too early system reboots. While the
-		 * right place to handle this issue is the given
-		 * driver, we do want to have a safe RAID driver ...
+			sectors = e - s;
+		}
+	}
+	if (sectors && hi < bb->count) {
+		/* 'hi' points to the first range that starts after 's'.
+		 * Maybe we can merge with the start of that range */
+		sector_t a = BB_OFFSET(p[hi]);
+		sector_t e = a + BB_LEN(p[hi]);
+		int ack = BB_ACK(p[hi]);
+		if (a <= s + sectors) {
+			/* merging is possible */
+			if (e <= s + sectors) {
+				/* full overlap */
+				e = s + sectors;
+				ack = acknowledged;
+			} else
+				ack = ack && acknowledged;
+
+			a = s;
+			if (e - a <= BB_MAX_LEN) {
+				p[hi] = BB_MAKE(a, e-a, ack);
+				s = e;
+			} else {
+				p[hi] = BB_MAKE(a, BB_MAX_LEN, ack);
+				s = a + BB_MAX_LEN;
+			}
+			sectors = e - s;
+			lo = hi;
+			hi++;
+		}
+	}
+	if (sectors == 0 && hi < bb->count) {
+		/* we might be able to combine lo and hi */
+		/* Note: 's' is at the end of 'lo' */
+		sector_t a = BB_OFFSET(p[hi]);
+		int lolen = BB_LEN(p[lo]);
+		int hilen = BB_LEN(p[hi]);
+		int newlen = lolen + hilen - (s - a);
+		if (s >= a && newlen < BB_MAX_LEN) {
+			/* yes, we can combine them */
+			int ack = BB_ACK(p[lo]) && BB_ACK(p[hi]);
+			p[lo] = BB_MAKE(BB_OFFSET(p[lo]), newlen, ack);
+			memmove(p + hi, p + hi + 1,
+				(bb->count - hi - 1) * 8);
+			bb->count--;
+		}
+	}
+	while (sectors) {
+		/* didn't merge (it all).
+		 * Need to add a range just before 'hi' */
+		if (bb->count >= MD_MAX_BADBLOCKS) {
+			/* No room for more */
+			rv = 0;
+			break;
+		} else {
+			int this_sectors = sectors;
+			memmove(p + hi + 1, p + hi,
+				(bb->count - hi) * 8);
+			bb->count++;
+
+			if (this_sectors > BB_MAX_LEN)
+				this_sectors = BB_MAX_LEN;
+			p[hi] = BB_MAKE(s, this_sectors, acknowledged);
+			sectors -= this_sectors;
+			s += this_sectors;
+		}
+	}
+
+	bb->changed = 1;
+	if (!acknowledged)
+		bb->unacked_exist = 1;
+	write_sequnlock_irqrestore(&bb->lock, flags);
+
+	return rv;
+}
+
+int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
+		       int is_new)
+{
+	int rv;
+	if (is_new)
+		s += rdev->new_data_offset;
+	else
+		s += rdev->data_offset;
+	rv = md_set_badblocks(&rdev->badblocks,
+			      s, sectors, 0);
+	if (rv) {
+		/* Make sure they get written out promptly */
+		sysfs_notify_dirent_safe(rdev->sysfs_state);
+		set_bit(MD_CHANGE_CLEAN, &rdev->mddev->flags);
+		md_wakeup_thread(rdev->mddev->thread);
+	}
+	return rv;
+}
+EXPORT_SYMBOL_GPL(rdev_set_badblocks);
+
+/*
+ * Remove a range of bad blocks from the table.
+ * This may involve extending the table if we spilt a region,
+ * but it must not fail.  So if the table becomes full, we just
+ * drop the remove request.
+ */
+static int md_clear_badblocks(struct badblocks *bb, sector_t s, int sectors)
+{
+	u64 *p;
+	int lo, hi;
+	sector_t target = s + sectors;
+	int rv = 0;
+
+	if (bb->shift > 0) {
+		/* When clearing we round the start up and the end down.
+		 * This should not matter as the shift should align with
+		 * the block size and no rounding should ever be needed.
+		 * However it is better the think a block is bad when it
+		 * isn't than to think a block is not bad when it is.
 		 */
-		mdelay(1000*1);
+		s += (1<<bb->shift) - 1;
+		s >>= bb->shift;
+		target >>= bb->shift;
+		sectors = target - s;
+	}
+
+	write_seqlock_irq(&bb->lock);
+
+	p = bb->page;
+	lo = 0;
+	hi = bb->count;
+	/* Find the last range that starts before 'target' */
+	while (hi - lo > 1) {
+		int mid = (lo + hi) / 2;
+		sector_t a = BB_OFFSET(p[mid]);
+		if (a < target)
+			lo = mid;
+		else
+			hi = mid;
 	}
+	if (hi > lo) {
+		/* p[lo] is the last range that could overlap the
+		 * current range.  Earlier ranges could also overlap,
+		 * but only this one can overlap the end of the range.
+		 */
+		if (BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > target) {
+			/* Partial overlap, leave the tail of this range */
+			int ack = BB_ACK(p[lo]);
+			sector_t a = BB_OFFSET(p[lo]);
+			sector_t end = a + BB_LEN(p[lo]);
+
+			if (a < s) {
+				/* we need to split this range */
+				if (bb->count >= MD_MAX_BADBLOCKS) {
+					rv = -ENOSPC;
+					goto out;
+				}
+				memmove(p+lo+1, p+lo, (bb->count - lo) * 8);
+				bb->count++;
+				p[lo] = BB_MAKE(a, s-a, ack);
+				lo++;
+			}
+			p[lo] = BB_MAKE(target, end - target, ack);
+			/* there is no longer an overlap */
+			hi = lo;
+			lo--;
+		}
+		while (lo >= 0 &&
+		       BB_OFFSET(p[lo]) + BB_LEN(p[lo]) > s) {
+			/* This range does overlap */
+			if (BB_OFFSET(p[lo]) < s) {
+				/* Keep the early parts of this range. */
+				int ack = BB_ACK(p[lo]);
+				sector_t start = BB_OFFSET(p[lo]);
+				p[lo] = BB_MAKE(start, s - start, ack);
+				/* now low doesn't overlap, so.. */
+				break;
+			}
+			lo--;
+		}
+		/* 'lo' is strictly before, 'hi' is strictly after,
+		 * anything between needs to be discarded
+		 */
+		if (hi - lo > 1) {
+			memmove(p+lo+1, p+hi, (bb->count - hi) * 8);
+			bb->count -= (hi - lo - 1);
+		}
+	}
+
+	bb->changed = 1;
+out:
+	write_sequnlock_irq(&bb->lock);
+	return rv;
+}
+
+int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
+			 int is_new)
+{
+	if (is_new)
+		s += rdev->new_data_offset;
+	else
+		s += rdev->data_offset;
+	return md_clear_badblocks(&rdev->badblocks,
+				  s, sectors);
+}
+EXPORT_SYMBOL_GPL(rdev_clear_badblocks);
+
+/*
+ * Acknowledge all bad blocks in a list.
+ * This only succeeds if ->changed is clear.  It is used by
+ * in-kernel metadata updates
+ */
+void md_ack_all_badblocks(struct badblocks *bb)
+{
+	if (bb->page == NULL || bb->changed)
+		/* no point even trying */
+		return;
+	write_seqlock_irq(&bb->lock);
+
+	if (bb->changed == 0 && bb->unacked_exist) {
+		u64 *p = bb->page;
+		int i;
+		for (i = 0; i < bb->count ; i++) {
+			if (!BB_ACK(p[i])) {
+				sector_t start = BB_OFFSET(p[i]);
+				int len = BB_LEN(p[i]);
+				p[i] = BB_MAKE(start, len, 1);
+			}
+		}
+		bb->unacked_exist = 0;
+	}
+	write_sequnlock_irq(&bb->lock);
+}
+EXPORT_SYMBOL_GPL(md_ack_all_badblocks);
+
+/* sysfs access to bad-blocks list.
+ * We present two files.
+ * 'bad-blocks' lists sector numbers and lengths of ranges that
+ *    are recorded as bad.  The list is truncated to fit within
+ *    the one-page limit of sysfs.
+ *    Writing "sector length" to this file adds an acknowledged
+ *    bad block list.
+ * 'unacknowledged-bad-blocks' lists bad blocks that have not yet
+ *    been acknowledged.  Writing to this file adds bad blocks
+ *    without acknowledging them.  This is largely for testing.
+ */
+
+static ssize_t
+badblocks_show(struct badblocks *bb, char *page, int unack)
+{
+	size_t len;
+	int i;
+	u64 *p = bb->page;
+	unsigned seq;
+
+	if (bb->shift < 0)
+		return 0;
+
+retry:
+	seq = read_seqbegin(&bb->lock);
+
+	len = 0;
+	i = 0;
+
+	while (len < PAGE_SIZE && i < bb->count) {
+		sector_t s = BB_OFFSET(p[i]);
+		unsigned int length = BB_LEN(p[i]);
+		int ack = BB_ACK(p[i]);
+		i++;
+
+		if (unack && ack)
+			continue;
+
+		len += snprintf(page+len, PAGE_SIZE-len, "%llu %u\n",
+				(unsigned long long)s << bb->shift,
+				length << bb->shift);
+	}
+	if (unack && len == 0)
+		bb->unacked_exist = 0;
+
+	if (read_seqretry(&bb->lock, seq))
+		goto retry;
+
+	return len;
+}
+
+#define DO_DEBUG 1
+
+static ssize_t
+badblocks_store(struct badblocks *bb, const char *page, size_t len, int unack)
+{
+	unsigned long long sector;
+	int length;
+	char newline;
+#ifdef DO_DEBUG
+	/* Allow clearing via sysfs *only* for testing/debugging.
+	 * Normally only a successful write may clear a badblock
+	 */
+	int clear = 0;
+	if (page[0] == '-') {
+		clear = 1;
+		page++;
+	}
+#endif /* DO_DEBUG */
+
+	switch (sscanf(page, "%llu %d%c", &sector, &length, &newline)) {
+	case 3:
+		if (newline != '\n')
+			return -EINVAL;
+	case 2:
+		if (length <= 0)
+			return -EINVAL;
+		break;
+	default:
+		return -EINVAL;
+	}
+
+#ifdef DO_DEBUG
+	if (clear) {
+		md_clear_badblocks(bb, sector, length);
+		return len;
+	}
+#endif /* DO_DEBUG */
+	if (md_set_badblocks(bb, sector, length, !unack))
+		return len;
+	else
+		return -ENOSPC;
+}
+
+static int md_notify_reboot(struct notifier_block *this,
+			    unsigned long code, void *x)
+{
+	struct list_head *tmp;
+	struct mddev *mddev;
+	int need_delay = 0;
+
+	for_each_mddev(mddev, tmp) {
+		if (mddev_trylock(mddev)) {
+			if (mddev->pers)
+				__md_stop_writes(mddev);
+			if (mddev->persistent)
+				mddev->safemode = 2;
+			mddev_unlock(mddev);
+		}
+		need_delay = 1;
+	}
+	/*
+	 * certain more exotic SCSI devices are known to be
+	 * volatile wrt too early system reboots. While the
+	 * right place to handle this issue is the given
+	 * driver, we do want to have a safe RAID driver ...
+	 */
+	if (need_delay)
+		mdelay(1000*1);
+
 	return NOTIFY_DONE;
 }
 
@@ -6279,20 +8568,31 @@ static struct notifier_block md_notifier = {
 
 static void md_geninit(void)
 {
-	dprintk("md: sizeof(mdp_super_t) = %d\n", (int)sizeof(mdp_super_t));
+	pr_debug("md: sizeof(mdp_super_t) = %d\n", (int)sizeof(mdp_super_t));
 
 	proc_create("mdstat", S_IRUGO, NULL, &md_seq_fops);
 }
 
 static int __init md_init(void)
 {
-	if (register_blkdev(MAJOR_NR, "md"))
-		return -1;
-	if ((mdp_major=register_blkdev(0, "mdp"))<=0) {
-		unregister_blkdev(MAJOR_NR, "md");
-		return -1;
-	}
-	blk_register_region(MKDEV(MAJOR_NR, 0), 1UL<<MINORBITS, THIS_MODULE,
+	int ret = -ENOMEM;
+
+	md_wq = alloc_workqueue("md", WQ_MEM_RECLAIM, 0);
+	if (!md_wq)
+		goto err_wq;
+
+	md_misc_wq = alloc_workqueue("md_misc", 0, 0);
+	if (!md_misc_wq)
+		goto err_misc_wq;
+
+	if ((ret = register_blkdev(MD_MAJOR, "md")) < 0)
+		goto err_md;
+
+	if ((ret = register_blkdev(0, "mdp")) < 0)
+		goto err_mdp;
+	mdp_major = ret;
+
+	blk_register_region(MKDEV(MD_MAJOR, 0), 1UL<<MINORBITS, THIS_MODULE,
 			    md_probe, NULL, NULL);
 	blk_register_region(MKDEV(mdp_major, 0), 1UL<<MINORBITS, THIS_MODULE,
 			    md_probe, NULL, NULL);
@@ -6301,9 +8601,17 @@ static int __init md_init(void)
 	raid_table_header = register_sysctl_table(raid_root_table);
 
 	md_geninit();
-	return (0);
-}
+	return 0;
 
+err_mdp:
+	unregister_blkdev(MD_MAJOR, "md");
+err_md:
+	destroy_workqueue(md_misc_wq);
+err_misc_wq:
+	destroy_workqueue(md_wq);
+err_wq:
+	return ret;
+}
 
 #ifndef MODULE
 
@@ -6335,7 +8643,7 @@ void md_autodetect_dev(dev_t dev)
 
 static void autostart_arrays(int part)
 {
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 	struct detected_devices_node *node_detected_dev;
 	dev_t dev;
 	int i_scanned, i_passed;
@@ -6375,27 +8683,36 @@ static void autostart_arrays(int part)
 
 static __exit void md_exit(void)
 {
-	mddev_t *mddev;
+	struct mddev *mddev;
 	struct list_head *tmp;
+	int delay = 1;
 
-	blk_unregister_region(MKDEV(MAJOR_NR,0), 1U << MINORBITS);
+	blk_unregister_region(MKDEV(MD_MAJOR,0), 1U << MINORBITS);
 	blk_unregister_region(MKDEV(mdp_major,0), 1U << MINORBITS);
 
-	unregister_blkdev(MAJOR_NR,"md");
+	unregister_blkdev(MD_MAJOR,"md");
 	unregister_blkdev(mdp_major, "mdp");
 	unregister_reboot_notifier(&md_notifier);
 	unregister_sysctl_table(raid_table_header);
+
+	/* We cannot unload the modules while some process is
+	 * waiting for us in select() or poll() - wake them up
+	 */
+	md_unloading = 1;
+	while (waitqueue_active(&md_event_waiters)) {
+		/* not safe to leave yet */
+		wake_up(&md_event_waiters);
+		msleep(delay);
+		delay += delay;
+	}
 	remove_proc_entry("mdstat", NULL);
+
 	for_each_mddev(mddev, tmp) {
-		struct gendisk *disk = mddev->gendisk;
-		if (!disk)
-			continue;
 		export_array(mddev);
-		del_gendisk(disk);
-		put_disk(disk);
-		mddev->gendisk = NULL;
-		mddev_put(mddev);
+		mddev->hold_active = 0;
 	}
+	destroy_workqueue(md_misc_wq);
+	destroy_workqueue(md_wq);
 }
 
 subsys_initcall(md_init);
@@ -6419,6 +8736,7 @@ static int set_ro(const char *val, struct kernel_param *kp)
 module_param_call(start_ro, set_ro, get_ro, NULL, S_IRUSR|S_IWUSR);
 module_param(start_dirty_degraded, int, S_IRUGO|S_IWUSR);
 
+module_param_call(new_array, add_named_array, NULL, NULL, S_IWUSR);
 
 EXPORT_SYMBOL(register_md_personality);
 EXPORT_SYMBOL(unregister_md_personality);
@@ -6430,6 +8748,8 @@ EXPORT_SYMBOL(md_register_thread);
 EXPORT_SYMBOL(md_unregister_thread);
 EXPORT_SYMBOL(md_wakeup_thread);
 EXPORT_SYMBOL(md_check_recovery);
+EXPORT_SYMBOL(md_reap_sync_thread);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("MD RAID framework");
 MODULE_ALIAS("md");
 MODULE_ALIAS_BLOCKDEV_MAJOR(MD_MAJOR);
diff --git a/drivers/md/md.h b/drivers/md/md.h
new file mode 100644
index 00000000000..a49d991f3fe
--- /dev/null
+++ b/drivers/md/md.h
@@ -0,0 +1,629 @@
+/*
+   md.h : kernel internal structure of the Linux MD driver
+          Copyright (C) 1996-98 Ingo Molnar, Gadi Oxman
+	  
+   This program is free software; you can redistribute it and/or modify
+   it under the terms of the GNU General Public License as published by
+   the Free Software Foundation; either version 2, or (at your option)
+   any later version.
+   
+   You should have received a copy of the GNU General Public License
+   (for example /usr/src/linux/COPYING); if not, write to the Free
+   Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  
+*/
+
+#ifndef _MD_MD_H
+#define _MD_MD_H
+
+#include <linux/blkdev.h>
+#include <linux/kobject.h>
+#include <linux/list.h>
+#include <linux/mm.h>
+#include <linux/mutex.h>
+#include <linux/timer.h>
+#include <linux/wait.h>
+#include <linux/workqueue.h>
+
+#define MaxSector (~(sector_t)0)
+
+/* Bad block numbers are stored sorted in a single page.
+ * 64bits is used for each block or extent.
+ * 54 bits are sector number, 9 bits are extent size,
+ * 1 bit is an 'acknowledged' flag.
+ */
+#define MD_MAX_BADBLOCKS	(PAGE_SIZE/8)
+
+/*
+ * MD's 'extended' device
+ */
+struct md_rdev {
+	struct list_head same_set;	/* RAID devices within the same set */
+
+	sector_t sectors;		/* Device size (in 512bytes sectors) */
+	struct mddev *mddev;		/* RAID array if running */
+	int last_events;		/* IO event timestamp */
+
+	/*
+	 * If meta_bdev is non-NULL, it means that a separate device is
+	 * being used to store the metadata (superblock/bitmap) which
+	 * would otherwise be contained on the same device as the data (bdev).
+	 */
+	struct block_device *meta_bdev;
+	struct block_device *bdev;	/* block device handle */
+
+	struct page	*sb_page, *bb_page;
+	int		sb_loaded;
+	__u64		sb_events;
+	sector_t	data_offset;	/* start of data in array */
+	sector_t	new_data_offset;/* only relevant while reshaping */
+	sector_t 	sb_start;	/* offset of the super block (in 512byte sectors) */
+	int		sb_size;	/* bytes in the superblock */
+	int		preferred_minor;	/* autorun support */
+
+	struct kobject	kobj;
+
+	/* A device can be in one of three states based on two flags:
+	 * Not working:   faulty==1 in_sync==0
+	 * Fully working: faulty==0 in_sync==1
+	 * Working, but not
+	 * in sync with array
+	 *                faulty==0 in_sync==0
+	 *
+	 * It can never have faulty==1, in_sync==1
+	 * This reduces the burden of testing multiple flags in many cases
+	 */
+
+	unsigned long	flags;	/* bit set of 'enum flag_bits' bits. */
+	wait_queue_head_t blocked_wait;
+
+	int desc_nr;			/* descriptor index in the superblock */
+	int raid_disk;			/* role of device in array */
+	int new_raid_disk;		/* role that the device will have in
+					 * the array after a level-change completes.
+					 */
+	int saved_raid_disk;		/* role that device used to have in the
+					 * array and could again if we did a partial
+					 * resync from the bitmap
+					 */
+	sector_t	recovery_offset;/* If this device has been partially
+					 * recovered, this is where we were
+					 * up to.
+					 */
+
+	atomic_t	nr_pending;	/* number of pending requests.
+					 * only maintained for arrays that
+					 * support hot removal
+					 */
+	atomic_t	read_errors;	/* number of consecutive read errors that
+					 * we have tried to ignore.
+					 */
+	struct timespec last_read_error;	/* monotonic time since our
+						 * last read error
+						 */
+	atomic_t	corrected_errors; /* number of corrected read errors,
+					   * for reporting to userspace and storing
+					   * in superblock.
+					   */
+	struct work_struct del_work;	/* used for delayed sysfs removal */
+
+	struct kernfs_node *sysfs_state; /* handle for 'state'
+					   * sysfs entry */
+
+	struct badblocks {
+		int	count;		/* count of bad blocks */
+		int	unacked_exist;	/* there probably are unacknowledged
+					 * bad blocks.  This is only cleared
+					 * when a read discovers none
+					 */
+		int	shift;		/* shift from sectors to block size
+					 * a -ve shift means badblocks are
+					 * disabled.*/
+		u64	*page;		/* badblock list */
+		int	changed;
+		seqlock_t lock;
+
+		sector_t sector;
+		sector_t size;		/* in sectors */
+	} badblocks;
+};
+enum flag_bits {
+	Faulty,			/* device is known to have a fault */
+	In_sync,		/* device is in_sync with rest of array */
+	Bitmap_sync,		/* ..actually, not quite In_sync.  Need a
+				 * bitmap-based recovery to get fully in sync
+				 */
+	Unmerged,		/* device is being added to array and should
+				 * be considerred for bvec_merge_fn but not
+				 * yet for actual IO
+				 */
+	WriteMostly,		/* Avoid reading if at all possible */
+	AutoDetected,		/* added by auto-detect */
+	Blocked,		/* An error occurred but has not yet
+				 * been acknowledged by the metadata
+				 * handler, so don't allow writes
+				 * until it is cleared */
+	WriteErrorSeen,		/* A write error has been seen on this
+				 * device
+				 */
+	FaultRecorded,		/* Intermediate state for clearing
+				 * Blocked.  The Fault is/will-be
+				 * recorded in the metadata, but that
+				 * metadata hasn't been stored safely
+				 * on disk yet.
+				 */
+	BlockedBadBlocks,	/* A writer is blocked because they
+				 * found an unacknowledged bad-block.
+				 * This can safely be cleared at any
+				 * time, and the writer will re-check.
+				 * It may be set at any time, and at
+				 * worst the writer will timeout and
+				 * re-check.  So setting it as
+				 * accurately as possible is good, but
+				 * not absolutely critical.
+				 */
+	WantReplacement,	/* This device is a candidate to be
+				 * hot-replaced, either because it has
+				 * reported some faults, or because
+				 * of explicit request.
+				 */
+	Replacement,		/* This device is a replacement for
+				 * a want_replacement device with same
+				 * raid_disk number.
+				 */
+};
+
+#define BB_LEN_MASK	(0x00000000000001FFULL)
+#define BB_OFFSET_MASK	(0x7FFFFFFFFFFFFE00ULL)
+#define BB_ACK_MASK	(0x8000000000000000ULL)
+#define BB_MAX_LEN	512
+#define BB_OFFSET(x)	(((x) & BB_OFFSET_MASK) >> 9)
+#define BB_LEN(x)	(((x) & BB_LEN_MASK) + 1)
+#define BB_ACK(x)	(!!((x) & BB_ACK_MASK))
+#define BB_MAKE(a, l, ack) (((a)<<9) | ((l)-1) | ((u64)(!!(ack)) << 63))
+
+extern int md_is_badblock(struct badblocks *bb, sector_t s, int sectors,
+			  sector_t *first_bad, int *bad_sectors);
+static inline int is_badblock(struct md_rdev *rdev, sector_t s, int sectors,
+			      sector_t *first_bad, int *bad_sectors)
+{
+	if (unlikely(rdev->badblocks.count)) {
+		int rv = md_is_badblock(&rdev->badblocks, rdev->data_offset + s,
+					sectors,
+					first_bad, bad_sectors);
+		if (rv)
+			*first_bad -= rdev->data_offset;
+		return rv;
+	}
+	return 0;
+}
+extern int rdev_set_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
+			      int is_new);
+extern int rdev_clear_badblocks(struct md_rdev *rdev, sector_t s, int sectors,
+				int is_new);
+extern void md_ack_all_badblocks(struct badblocks *bb);
+
+struct mddev {
+	void				*private;
+	struct md_personality		*pers;
+	dev_t				unit;
+	int				md_minor;
+	struct list_head		disks;
+	unsigned long			flags;
+#define MD_CHANGE_DEVS	0	/* Some device status has changed */
+#define MD_CHANGE_CLEAN 1	/* transition to or from 'clean' */
+#define MD_CHANGE_PENDING 2	/* switch from 'clean' to 'active' in progress */
+#define MD_UPDATE_SB_FLAGS (1 | 2 | 4)	/* If these are set, md_update_sb needed */
+#define MD_ARRAY_FIRST_USE 3    /* First use of array, needs initialization */
+#define MD_STILL_CLOSED	4	/* If set, then array has not been opened since
+				 * md_ioctl checked on it.
+				 */
+
+	int				suspended;
+	atomic_t			active_io;
+	int				ro;
+	int				sysfs_active; /* set when sysfs deletes
+						       * are happening, so run/
+						       * takeover/stop are not safe
+						       */
+	int				ready; /* See when safe to pass
+						* IO requests down */
+	struct gendisk			*gendisk;
+
+	struct kobject			kobj;
+	int				hold_active;
+#define	UNTIL_IOCTL	1
+#define	UNTIL_STOP	2
+
+	/* Superblock information */
+	int				major_version,
+					minor_version,
+					patch_version;
+	int				persistent;
+	int 				external;	/* metadata is
+							 * managed externally */
+	char				metadata_type[17]; /* externally set*/
+	int				chunk_sectors;
+	time_t				ctime, utime;
+	int				level, layout;
+	char				clevel[16];
+	int				raid_disks;
+	int				max_disks;
+	sector_t			dev_sectors; 	/* used size of
+							 * component devices */
+	sector_t			array_sectors; /* exported array size */
+	int				external_size; /* size managed
+							* externally */
+	__u64				events;
+	/* If the last 'event' was simply a clean->dirty transition, and
+	 * we didn't write it to the spares, then it is safe and simple
+	 * to just decrement the event count on a dirty->clean transition.
+	 * So we record that possibility here.
+	 */
+	int				can_decrease_events;
+
+	char				uuid[16];
+
+	/* If the array is being reshaped, we need to record the
+	 * new shape and an indication of where we are up to.
+	 * This is written to the superblock.
+	 * If reshape_position is MaxSector, then no reshape is happening (yet).
+	 */
+	sector_t			reshape_position;
+	int				delta_disks, new_level, new_layout;
+	int				new_chunk_sectors;
+	int				reshape_backwards;
+
+	struct md_thread		*thread;	/* management thread */
+	struct md_thread		*sync_thread;	/* doing resync or reconstruct */
+
+	/* 'last_sync_action' is initialized to "none".  It is set when a
+	 * sync operation (i.e "data-check", "requested-resync", "resync",
+	 * "recovery", or "reshape") is started.  It holds this value even
+	 * when the sync thread is "frozen" (interrupted) or "idle" (stopped
+	 * or finished).  It is overwritten when a new sync operation is begun.
+	 */
+	char				*last_sync_action;
+	sector_t			curr_resync;	/* last block scheduled */
+	/* As resync requests can complete out of order, we cannot easily track
+	 * how much resync has been completed.  So we occasionally pause until
+	 * everything completes, then set curr_resync_completed to curr_resync.
+	 * As such it may be well behind the real resync mark, but it is a value
+	 * we are certain of.
+	 */
+	sector_t			curr_resync_completed;
+	unsigned long			resync_mark;	/* a recent timestamp */
+	sector_t			resync_mark_cnt;/* blocks written at resync_mark */
+	sector_t			curr_mark_cnt; /* blocks scheduled now */
+
+	sector_t			resync_max_sectors; /* may be set by personality */
+
+	atomic64_t			resync_mismatches; /* count of sectors where
+							    * parity/replica mismatch found
+							    */
+
+	/* allow user-space to request suspension of IO to regions of the array */
+	sector_t			suspend_lo;
+	sector_t			suspend_hi;
+	/* if zero, use the system-wide default */
+	int				sync_speed_min;
+	int				sync_speed_max;
+
+	/* resync even though the same disks are shared among md-devices */
+	int				parallel_resync;
+
+	int				ok_start_degraded;
+	/* recovery/resync flags 
+	 * NEEDED:   we might need to start a resync/recover
+	 * RUNNING:  a thread is running, or about to be started
+	 * SYNC:     actually doing a resync, not a recovery
+	 * RECOVER:  doing recovery, or need to try it.
+	 * INTR:     resync needs to be aborted for some reason
+	 * DONE:     thread is done and is waiting to be reaped
+	 * REQUEST:  user-space has requested a sync (used with SYNC)
+	 * CHECK:    user-space request for check-only, no repair
+	 * RESHAPE:  A reshape is happening
+	 * ERROR:    sync-action interrupted because io-error
+	 *
+	 * If neither SYNC or RESHAPE are set, then it is a recovery.
+	 */
+#define	MD_RECOVERY_RUNNING	0
+#define	MD_RECOVERY_SYNC	1
+#define	MD_RECOVERY_RECOVER	2
+#define	MD_RECOVERY_INTR	3
+#define	MD_RECOVERY_DONE	4
+#define	MD_RECOVERY_NEEDED	5
+#define	MD_RECOVERY_REQUESTED	6
+#define	MD_RECOVERY_CHECK	7
+#define MD_RECOVERY_RESHAPE	8
+#define	MD_RECOVERY_FROZEN	9
+#define	MD_RECOVERY_ERROR	10
+
+	unsigned long			recovery;
+	/* If a RAID personality determines that recovery (of a particular
+	 * device) will fail due to a read error on the source device, it
+	 * takes a copy of this number and does not attempt recovery again
+	 * until this number changes.
+	 */
+	int				recovery_disabled;
+
+	int				in_sync;	/* know to not need resync */
+	/* 'open_mutex' avoids races between 'md_open' and 'do_md_stop', so
+	 * that we are never stopping an array while it is open.
+	 * 'reconfig_mutex' protects all other reconfiguration.
+	 * These locks are separate due to conflicting interactions
+	 * with bdev->bd_mutex.
+	 * Lock ordering is:
+	 *  reconfig_mutex -> bd_mutex : e.g. do_md_run -> revalidate_disk
+	 *  bd_mutex -> open_mutex:  e.g. __blkdev_get -> md_open
+	 */
+	struct mutex			open_mutex;
+	struct mutex			reconfig_mutex;
+	atomic_t			active;		/* general refcount */
+	atomic_t			openers;	/* number of active opens */
+
+	int				changed;	/* True if we might need to
+							 * reread partition info */
+	int				degraded;	/* whether md should consider
+							 * adding a spare
+							 */
+	int				merge_check_needed; /* at least one
+							     * member device
+							     * has a
+							     * merge_bvec_fn */
+
+	atomic_t			recovery_active; /* blocks scheduled, but not written */
+	wait_queue_head_t		recovery_wait;
+	sector_t			recovery_cp;
+	sector_t			resync_min;	/* user requested sync
+							 * starts here */
+	sector_t			resync_max;	/* resync should pause
+							 * when it gets here */
+
+	struct kernfs_node		*sysfs_state;	/* handle for 'array_state'
+							 * file in sysfs.
+							 */
+	struct kernfs_node		*sysfs_action;  /* handle for 'sync_action' */
+
+	struct work_struct del_work;	/* used for delayed sysfs removal */
+
+	spinlock_t			write_lock;
+	wait_queue_head_t		sb_wait;	/* for waiting on superblock updates */
+	atomic_t			pending_writes;	/* number of active superblock writes */
+
+	unsigned int			safemode;	/* if set, update "clean" superblock
+							 * when no writes pending.
+							 */ 
+	unsigned int			safemode_delay;
+	struct timer_list		safemode_timer;
+	atomic_t			writes_pending; 
+	struct request_queue		*queue;	/* for plugging ... */
+
+	struct bitmap                   *bitmap; /* the bitmap for the device */
+	struct {
+		struct file		*file; /* the bitmap file */
+		loff_t			offset; /* offset from superblock of
+						 * start of bitmap. May be
+						 * negative, but not '0'
+						 * For external metadata, offset
+						 * from start of device. 
+						 */
+		unsigned long		space; /* space available at this offset */
+		loff_t			default_offset; /* this is the offset to use when
+							 * hot-adding a bitmap.  It should
+							 * eventually be settable by sysfs.
+							 */
+		unsigned long		default_space; /* space available at
+							* default offset */
+		struct mutex		mutex;
+		unsigned long		chunksize;
+		unsigned long		daemon_sleep; /* how many jiffies between updates? */
+		unsigned long		max_write_behind; /* write-behind mode */
+		int			external;
+	} bitmap_info;
+
+	atomic_t 			max_corr_read_errors; /* max read retries */
+	struct list_head		all_mddevs;
+
+	struct attribute_group		*to_remove;
+
+	struct bio_set			*bio_set;
+
+	/* Generic flush handling.
+	 * The last to finish preflush schedules a worker to submit
+	 * the rest of the request (without the REQ_FLUSH flag).
+	 */
+	struct bio *flush_bio;
+	atomic_t flush_pending;
+	struct work_struct flush_work;
+	struct work_struct event_work;	/* used by dm to report failure event */
+	void (*sync_super)(struct mddev *mddev, struct md_rdev *rdev);
+};
+
+
+static inline void rdev_dec_pending(struct md_rdev *rdev, struct mddev *mddev)
+{
+	int faulty = test_bit(Faulty, &rdev->flags);
+	if (atomic_dec_and_test(&rdev->nr_pending) && faulty)
+		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+}
+
+static inline void md_sync_acct(struct block_device *bdev, unsigned long nr_sectors)
+{
+        atomic_add(nr_sectors, &bdev->bd_contains->bd_disk->sync_io);
+}
+
+struct md_personality
+{
+	char *name;
+	int level;
+	struct list_head list;
+	struct module *owner;
+	void (*make_request)(struct mddev *mddev, struct bio *bio);
+	int (*run)(struct mddev *mddev);
+	int (*stop)(struct mddev *mddev);
+	void (*status)(struct seq_file *seq, struct mddev *mddev);
+	/* error_handler must set ->faulty and clear ->in_sync
+	 * if appropriate, and should abort recovery if needed 
+	 */
+	void (*error_handler)(struct mddev *mddev, struct md_rdev *rdev);
+	int (*hot_add_disk) (struct mddev *mddev, struct md_rdev *rdev);
+	int (*hot_remove_disk) (struct mddev *mddev, struct md_rdev *rdev);
+	int (*spare_active) (struct mddev *mddev);
+	sector_t (*sync_request)(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster);
+	int (*resize) (struct mddev *mddev, sector_t sectors);
+	sector_t (*size) (struct mddev *mddev, sector_t sectors, int raid_disks);
+	int (*check_reshape) (struct mddev *mddev);
+	int (*start_reshape) (struct mddev *mddev);
+	void (*finish_reshape) (struct mddev *mddev);
+	/* quiesce moves between quiescence states
+	 * 0 - fully active
+	 * 1 - no new requests allowed
+	 * others - reserved
+	 */
+	void (*quiesce) (struct mddev *mddev, int state);
+	/* takeover is used to transition an array from one
+	 * personality to another.  The new personality must be able
+	 * to handle the data in the current layout.
+	 * e.g. 2drive raid1 -> 2drive raid5
+	 *      ndrive raid5 -> degraded n+1drive raid6 with special layout
+	 * If the takeover succeeds, a new 'private' structure is returned.
+	 * This needs to be installed and then ->run used to activate the
+	 * array.
+	 */
+	void *(*takeover) (struct mddev *mddev);
+};
+
+
+struct md_sysfs_entry {
+	struct attribute attr;
+	ssize_t (*show)(struct mddev *, char *);
+	ssize_t (*store)(struct mddev *, const char *, size_t);
+};
+extern struct attribute_group md_bitmap_group;
+
+static inline struct kernfs_node *sysfs_get_dirent_safe(struct kernfs_node *sd, char *name)
+{
+	if (sd)
+		return sysfs_get_dirent(sd, name);
+	return sd;
+}
+static inline void sysfs_notify_dirent_safe(struct kernfs_node *sd)
+{
+	if (sd)
+		sysfs_notify_dirent(sd);
+}
+
+static inline char * mdname (struct mddev * mddev)
+{
+	return mddev->gendisk ? mddev->gendisk->disk_name : "mdX";
+}
+
+static inline int sysfs_link_rdev(struct mddev *mddev, struct md_rdev *rdev)
+{
+	char nm[20];
+	if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
+		sprintf(nm, "rd%d", rdev->raid_disk);
+		return sysfs_create_link(&mddev->kobj, &rdev->kobj, nm);
+	} else
+		return 0;
+}
+
+static inline void sysfs_unlink_rdev(struct mddev *mddev, struct md_rdev *rdev)
+{
+	char nm[20];
+	if (!test_bit(Replacement, &rdev->flags) && mddev->kobj.sd) {
+		sprintf(nm, "rd%d", rdev->raid_disk);
+		sysfs_remove_link(&mddev->kobj, nm);
+	}
+}
+
+/*
+ * iterates through some rdev ringlist. It's safe to remove the
+ * current 'rdev'. Dont touch 'tmp' though.
+ */
+#define rdev_for_each_list(rdev, tmp, head)				\
+	list_for_each_entry_safe(rdev, tmp, head, same_set)
+
+/*
+ * iterates through the 'same array disks' ringlist
+ */
+#define rdev_for_each(rdev, mddev)				\
+	list_for_each_entry(rdev, &((mddev)->disks), same_set)
+
+#define rdev_for_each_safe(rdev, tmp, mddev)				\
+	list_for_each_entry_safe(rdev, tmp, &((mddev)->disks), same_set)
+
+#define rdev_for_each_rcu(rdev, mddev)				\
+	list_for_each_entry_rcu(rdev, &((mddev)->disks), same_set)
+
+struct md_thread {
+	void			(*run) (struct md_thread *thread);
+	struct mddev		*mddev;
+	wait_queue_head_t	wqueue;
+	unsigned long           flags;
+	struct task_struct	*tsk;
+	unsigned long		timeout;
+	void			*private;
+};
+
+#define THREAD_WAKEUP  0
+
+static inline void safe_put_page(struct page *p)
+{
+	if (p) put_page(p);
+}
+
+extern int register_md_personality(struct md_personality *p);
+extern int unregister_md_personality(struct md_personality *p);
+extern struct md_thread *md_register_thread(
+	void (*run)(struct md_thread *thread),
+	struct mddev *mddev,
+	const char *name);
+extern void md_unregister_thread(struct md_thread **threadp);
+extern void md_wakeup_thread(struct md_thread *thread);
+extern void md_check_recovery(struct mddev *mddev);
+extern void md_reap_sync_thread(struct mddev *mddev);
+extern void md_write_start(struct mddev *mddev, struct bio *bi);
+extern void md_write_end(struct mddev *mddev);
+extern void md_done_sync(struct mddev *mddev, int blocks, int ok);
+extern void md_error(struct mddev *mddev, struct md_rdev *rdev);
+extern void md_finish_reshape(struct mddev *mddev);
+
+extern int mddev_congested(struct mddev *mddev, int bits);
+extern void md_flush_request(struct mddev *mddev, struct bio *bio);
+extern void md_super_write(struct mddev *mddev, struct md_rdev *rdev,
+			   sector_t sector, int size, struct page *page);
+extern void md_super_wait(struct mddev *mddev);
+extern int sync_page_io(struct md_rdev *rdev, sector_t sector, int size, 
+			struct page *page, int rw, bool metadata_op);
+extern void md_do_sync(struct md_thread *thread);
+extern void md_new_event(struct mddev *mddev);
+extern int md_allow_write(struct mddev *mddev);
+extern void md_wait_for_blocked_rdev(struct md_rdev *rdev, struct mddev *mddev);
+extern void md_set_array_sectors(struct mddev *mddev, sector_t array_sectors);
+extern int md_check_no_bitmap(struct mddev *mddev);
+extern int md_integrity_register(struct mddev *mddev);
+extern void md_integrity_add_rdev(struct md_rdev *rdev, struct mddev *mddev);
+extern int strict_strtoul_scaled(const char *cp, unsigned long *res, int scale);
+
+extern void mddev_init(struct mddev *mddev);
+extern int md_run(struct mddev *mddev);
+extern void md_stop(struct mddev *mddev);
+extern void md_stop_writes(struct mddev *mddev);
+extern int md_rdev_init(struct md_rdev *rdev);
+extern void md_rdev_clear(struct md_rdev *rdev);
+
+extern void mddev_suspend(struct mddev *mddev);
+extern void mddev_resume(struct mddev *mddev);
+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 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/mktables.c b/drivers/md/mktables.c
deleted file mode 100644
index b61d5767aae..00000000000
--- a/drivers/md/mktables.c
+++ /dev/null
@@ -1,120 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002-2007 H. Peter Anvin - All Rights Reserved
- *
- *   This file is part of the Linux kernel, and is made available under
- *   the terms of the GNU General Public License version 2 or (at your
- *   option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * mktables.c
- *
- * Make RAID-6 tables.  This is a host user space program to be run at
- * compile time.
- */
-
-#include <stdio.h>
-#include <string.h>
-#include <inttypes.h>
-#include <stdlib.h>
-#include <time.h>
-
-static uint8_t gfmul(uint8_t a, uint8_t b)
-{
-	uint8_t v = 0;
-
-	while (b) {
-		if (b & 1)
-			v ^= a;
-		a = (a << 1) ^ (a & 0x80 ? 0x1d : 0);
-		b >>= 1;
-	}
-
-	return v;
-}
-
-static uint8_t gfpow(uint8_t a, int b)
-{
-	uint8_t v = 1;
-
-	b %= 255;
-	if (b < 0)
-		b += 255;
-
-	while (b) {
-		if (b & 1)
-			v = gfmul(v, a);
-		a = gfmul(a, a);
-		b >>= 1;
-	}
-
-	return v;
-}
-
-int main(int argc, char *argv[])
-{
-	int i, j, k;
-	uint8_t v;
-	uint8_t exptbl[256], invtbl[256];
-
-	printf("#include \"raid6.h\"\n");
-
-	/* Compute multiplication table */
-	printf("\nconst u8  __attribute__((aligned(256)))\n"
-		"raid6_gfmul[256][256] =\n"
-		"{\n");
-	for (i = 0; i < 256; i++) {
-		printf("\t{\n");
-		for (j = 0; j < 256; j += 8) {
-			printf("\t\t");
-			for (k = 0; k < 8; k++)
-				printf("0x%02x,%c", gfmul(i, j + k),
-				       (k == 7) ? '\n' : ' ');
-		}
-		printf("\t},\n");
-	}
-	printf("};\n");
-
-	/* Compute power-of-2 table (exponent) */
-	v = 1;
-	printf("\nconst u8 __attribute__((aligned(256)))\n"
-	       "raid6_gfexp[256] =\n" "{\n");
-	for (i = 0; i < 256; i += 8) {
-		printf("\t");
-		for (j = 0; j < 8; j++) {
-			exptbl[i + j] = v;
-			printf("0x%02x,%c", v, (j == 7) ? '\n' : ' ');
-			v = gfmul(v, 2);
-			if (v == 1)
-				v = 0;	/* For entry 255, not a real entry */
-		}
-	}
-	printf("};\n");
-
-	/* Compute inverse table x^-1 == x^254 */
-	printf("\nconst u8 __attribute__((aligned(256)))\n"
-	       "raid6_gfinv[256] =\n" "{\n");
-	for (i = 0; i < 256; i += 8) {
-		printf("\t");
-		for (j = 0; j < 8; j++) {
-			invtbl[i + j] = v = gfpow(i + j, 254);
-			printf("0x%02x,%c", v, (j == 7) ? '\n' : ' ');
-		}
-	}
-	printf("};\n");
-
-	/* Compute inv(2^x + 1) (exponent-xor-inverse) table */
-	printf("\nconst u8 __attribute__((aligned(256)))\n"
-	       "raid6_gfexi[256] =\n" "{\n");
-	for (i = 0; i < 256; i += 8) {
-		printf("\t");
-		for (j = 0; j < 8; j++)
-			printf("0x%02x,%c", invtbl[exptbl[i + j] ^ 1],
-			       (j == 7) ? '\n' : ' ');
-	}
-	printf("};\n");
-
-	return 0;
-}
diff --git a/drivers/md/multipath.c b/drivers/md/multipath.c
index c4779ccba1c..849ad39f547 100644
--- a/drivers/md/multipath.c
+++ b/drivers/md/multipath.c
@@ -19,23 +19,20 @@
  * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
+#include <linux/blkdev.h>
 #include <linux/module.h>
+#include <linux/raid/md_u.h>
+#include <linux/seq_file.h>
 #include <linux/slab.h>
-#include <linux/spinlock.h>
-#include <linux/raid/multipath.h>
-#include <linux/buffer_head.h>
-#include <asm/atomic.h>
-
-#define MAJOR_NR MD_MAJOR
-#define MD_DRIVER
-#define MD_PERSONALITY
+#include "md.h"
+#include "multipath.h"
 
 #define MAX_WORK_PER_DISK 128
 
 #define	NR_RESERVED_BUFS	32
 
 
-static int multipath_map (multipath_conf_t *conf)
+static int multipath_map (struct mpconf *conf)
 {
 	int i, disks = conf->raid_disks;
 
@@ -46,7 +43,7 @@ static int multipath_map (multipath_conf_t *conf)
 
 	rcu_read_lock();
 	for (i = 0; i < disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
 		if (rdev && test_bit(In_sync, &rdev->flags)) {
 			atomic_inc(&rdev->nr_pending);
 			rcu_read_unlock();
@@ -62,8 +59,8 @@ static int multipath_map (multipath_conf_t *conf)
 static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
 {
 	unsigned long flags;
-	mddev_t *mddev = mp_bh->mddev;
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = mp_bh->mddev;
+	struct mpconf *conf = mddev->private;
 
 	spin_lock_irqsave(&conf->device_lock, flags);
 	list_add(&mp_bh->retry_list, &conf->retry_list);
@@ -80,7 +77,7 @@ static void multipath_reschedule_retry (struct multipath_bh *mp_bh)
 static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
 {
 	struct bio *bio = mp_bh->master_bio;
-	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);
+	struct mpconf *conf = mp_bh->mddev->private;
 
 	bio_endio(bio, err);
 	mempool_free(mp_bh, conf->pool);
@@ -89,13 +86,13 @@ static void multipath_end_bh_io (struct multipath_bh *mp_bh, int err)
 static void multipath_end_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	struct multipath_bh * mp_bh = (struct multipath_bh *)(bio->bi_private);
-	multipath_conf_t *conf = mddev_to_conf(mp_bh->mddev);
-	mdk_rdev_t *rdev = conf->multipaths[mp_bh->path].rdev;
+	struct multipath_bh *mp_bh = bio->bi_private;
+	struct mpconf *conf = mp_bh->mddev->private;
+	struct md_rdev *rdev = conf->multipaths[mp_bh->path].rdev;
 
 	if (uptodate)
 		multipath_end_bh_io(mp_bh, 0);
-	else if (!bio_rw_ahead(bio)) {
+	else if (!(bio->bi_rw & REQ_RAHEAD)) {
 		/*
 		 * oops, IO error:
 		 */
@@ -103,54 +100,22 @@ static void multipath_end_request(struct bio *bio, int error)
 		md_error (mp_bh->mddev, rdev);
 		printk(KERN_ERR "multipath: %s: rescheduling sector %llu\n", 
 		       bdevname(rdev->bdev,b), 
-		       (unsigned long long)bio->bi_sector);
+		       (unsigned long long)bio->bi_iter.bi_sector);
 		multipath_reschedule_retry(mp_bh);
 	} else
 		multipath_end_bh_io(mp_bh, error);
 	rdev_dec_pending(rdev, conf->mddev);
 }
 
-static void unplug_slaves(mddev_t *mddev)
+static void multipath_make_request(struct mddev *mddev, struct bio * bio)
 {
-	multipath_conf_t *conf = mddev_to_conf(mddev);
-	int i;
-
-	rcu_read_lock();
-	for (i=0; i<mddev->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
-		if (rdev && !test_bit(Faulty, &rdev->flags)
-		    && atomic_read(&rdev->nr_pending)) {
-			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
-
-			atomic_inc(&rdev->nr_pending);
-			rcu_read_unlock();
-
-			blk_unplug(r_queue);
-
-			rdev_dec_pending(rdev, mddev);
-			rcu_read_lock();
-		}
-	}
-	rcu_read_unlock();
-}
-
-static void multipath_unplug(struct request_queue *q)
-{
-	unplug_slaves(q->queuedata);
-}
-
-
-static int multipath_make_request (struct request_queue *q, struct bio * bio)
-{
-	mddev_t *mddev = q->queuedata;
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mpconf *conf = mddev->private;
 	struct multipath_bh * mp_bh;
 	struct multipath_info *multipath;
-	const int rw = bio_data_dir(bio);
 
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
+		md_flush_request(mddev, bio);
+		return;
 	}
 
 	mp_bh = mempool_alloc(conf->pool, GFP_NOIO);
@@ -158,34 +123,31 @@ static int multipath_make_request (struct request_queue *q, struct bio * bio)
 	mp_bh->master_bio = bio;
 	mp_bh->mddev = mddev;
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
-
 	mp_bh->path = multipath_map(conf);
 	if (mp_bh->path < 0) {
 		bio_endio(bio, -EIO);
 		mempool_free(mp_bh, conf->pool);
-		return 0;
+		return;
 	}
 	multipath = conf->multipaths + mp_bh->path;
 
 	mp_bh->bio = *bio;
-	mp_bh->bio.bi_sector += multipath->rdev->data_offset;
+	mp_bh->bio.bi_iter.bi_sector += multipath->rdev->data_offset;
 	mp_bh->bio.bi_bdev = multipath->rdev->bdev;
-	mp_bh->bio.bi_rw |= (1 << BIO_RW_FAILFAST);
+	mp_bh->bio.bi_rw |= REQ_FAILFAST_TRANSPORT;
 	mp_bh->bio.bi_end_io = multipath_end_request;
 	mp_bh->bio.bi_private = mp_bh;
 	generic_make_request(&mp_bh->bio);
-	return 0;
+	return;
 }
 
-static void multipath_status (struct seq_file *seq, mddev_t *mddev)
+static void multipath_status (struct seq_file *seq, struct mddev *mddev)
 {
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mpconf *conf = mddev->private;
 	int i;
 	
 	seq_printf (seq, " [%d/%d] [", conf->raid_disks,
-						 conf->working_disks);
+		    conf->raid_disks - mddev->degraded);
 	for (i = 0; i < conf->raid_disks; i++)
 		seq_printf (seq, "%s",
 			       conf->multipaths[i].rdev && 
@@ -195,13 +157,16 @@ static void multipath_status (struct seq_file *seq, mddev_t *mddev)
 
 static int multipath_congested(void *data, int bits)
 {
-	mddev_t *mddev = data;
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = data;
+	struct mpconf *conf = mddev->private;
 	int i, ret = 0;
 
+	if (mddev_congested(mddev, bits))
+		return 1;
+
 	rcu_read_lock();
 	for (i = 0; i < mddev->raid_disks ; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->multipaths[i].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->multipaths[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags)) {
 			struct request_queue *q = bdev_get_queue(rdev->bdev);
 
@@ -219,41 +184,42 @@ static int multipath_congested(void *data, int bits)
 /*
  * Careful, this can execute in IRQ contexts as well!
  */
-static void multipath_error (mddev_t *mddev, mdk_rdev_t *rdev)
+static void multipath_error (struct mddev *mddev, struct md_rdev *rdev)
 {
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mpconf *conf = mddev->private;
+	char b[BDEVNAME_SIZE];
 
-	if (conf->working_disks <= 1) {
+	if (conf->raid_disks - mddev->degraded <= 1) {
 		/*
 		 * Uh oh, we can do nothing if this is our last path, but
 		 * first check if this is a queued request for a device
 		 * which has just failed.
 		 */
 		printk(KERN_ALERT 
-			"multipath: only one IO path left and IO error.\n");
+		       "multipath: only one IO path left and IO error.\n");
 		/* leave it active... it's all we have */
-	} else {
-		/*
-		 * Mark disk as unusable
-		 */
-		if (!test_bit(Faulty, &rdev->flags)) {
-			char b[BDEVNAME_SIZE];
-			clear_bit(In_sync, &rdev->flags);
-			set_bit(Faulty, &rdev->flags);
-			set_bit(MD_CHANGE_DEVS, &mddev->flags);
-			conf->working_disks--;
-			mddev->degraded++;
-			printk(KERN_ALERT "multipath: IO failure on %s,"
-				" disabling IO path.\n"
-				"multipath: Operation continuing"
-				" on %d IO paths.\n",
-				bdevname (rdev->bdev,b),
-				conf->working_disks);
-		}
+		return;
 	}
+	/*
+	 * Mark disk as unusable
+	 */
+	if (test_and_clear_bit(In_sync, &rdev->flags)) {
+		unsigned long flags;
+		spin_lock_irqsave(&conf->device_lock, flags);
+		mddev->degraded++;
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+	}
+	set_bit(Faulty, &rdev->flags);
+	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	printk(KERN_ALERT "multipath: IO failure on %s,"
+	       " disabling IO path.\n"
+	       "multipath: Operation continuing"
+	       " on %d IO paths.\n",
+	       bdevname(rdev->bdev, b),
+	       conf->raid_disks - mddev->degraded);
 }
 
-static void print_multipath_conf (multipath_conf_t *conf)
+static void print_multipath_conf (struct mpconf *conf)
 {
 	int i;
 	struct multipath_info *tmp;
@@ -263,7 +229,7 @@ static void print_multipath_conf (multipath_conf_t *conf)
 		printk("(conf==NULL)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->working_disks,
+	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
 			 conf->raid_disks);
 
 	for (i = 0; i < conf->raid_disks; i++) {
@@ -277,9 +243,9 @@ static void print_multipath_conf (multipath_conf_t *conf)
 }
 
 
-static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
+static int multipath_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	multipath_conf_t *conf = mddev->private;
+	struct mpconf *conf = mddev->private;
 	struct request_queue *q;
 	int err = -EEXIST;
 	int path;
@@ -295,24 +261,29 @@ static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 	for (path = first; path <= last; path++)
 		if ((p=conf->multipaths+path)->rdev == NULL) {
 			q = rdev->bdev->bd_disk->queue;
-			blk_queue_stack_limits(mddev->queue, q);
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 
 		/* as we don't honour merge_bvec_fn, we must never risk
-		 * violating it, so limit ->max_sector to one PAGE, as
-		 * a one page request is never in violation.
+		 * violating it, so limit ->max_segments to one, lying
+		 * within a single page.
 		 * (Note: it is very unlikely that a device with
 		 * merge_bvec_fn will be involved in multipath.)
 		 */
-			if (q->merge_bvec_fn &&
-			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-				blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
+			if (q->merge_bvec_fn) {
+				blk_queue_max_segments(mddev->queue, 1);
+				blk_queue_segment_boundary(mddev->queue,
+							   PAGE_CACHE_SIZE - 1);
+			}
 
-			conf->working_disks++;
+			spin_lock_irq(&conf->device_lock);
 			mddev->degraded--;
 			rdev->raid_disk = path;
 			set_bit(In_sync, &rdev->flags);
+			spin_unlock_irq(&conf->device_lock);
 			rcu_assign_pointer(p->rdev, rdev);
 			err = 0;
+			md_integrity_add_rdev(rdev, mddev);
 			break;
 		}
 
@@ -321,17 +292,16 @@ static int multipath_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 	return err;
 }
 
-static int multipath_remove_disk(mddev_t *mddev, int number)
+static int multipath_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	multipath_conf_t *conf = mddev->private;
+	struct mpconf *conf = mddev->private;
 	int err = 0;
-	mdk_rdev_t *rdev;
+	int number = rdev->raid_disk;
 	struct multipath_info *p = conf->multipaths + number;
 
 	print_multipath_conf(conf);
 
-	rdev = p->rdev;
-	if (rdev) {
+	if (rdev == p->rdev) {
 		if (test_bit(In_sync, &rdev->flags) ||
 		    atomic_read(&rdev->nr_pending)) {
 			printk(KERN_ERR "hot-remove-disk, slot %d is identified"
@@ -345,7 +315,9 @@ static int multipath_remove_disk(mddev_t *mddev, int number)
 			/* lost the race, try later */
 			err = -EBUSY;
 			p->rdev = rdev;
+			goto abort;
 		}
+		err = md_integrity_register(mddev);
 	}
 abort:
 
@@ -363,12 +335,13 @@ abort:
  *	3.	Performs writes following reads for array syncronising.
  */
 
-static void multipathd (mddev_t *mddev)
+static void multipathd(struct md_thread *thread)
 {
+	struct mddev *mddev = thread->mddev;
 	struct multipath_bh *mp_bh;
 	struct bio *bio;
 	unsigned long flags;
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mpconf *conf = mddev->private;
 	struct list_head *head = &conf->retry_list;
 
 	md_check_recovery(mddev);
@@ -382,23 +355,24 @@ static void multipathd (mddev_t *mddev)
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 
 		bio = &mp_bh->bio;
-		bio->bi_sector = mp_bh->master_bio->bi_sector;
+		bio->bi_iter.bi_sector = mp_bh->master_bio->bi_iter.bi_sector;
 		
 		if ((mp_bh->path = multipath_map (conf))<0) {
 			printk(KERN_ALERT "multipath: %s: unrecoverable IO read"
 				" error for block %llu\n",
 				bdevname(bio->bi_bdev,b),
-				(unsigned long long)bio->bi_sector);
+				(unsigned long long)bio->bi_iter.bi_sector);
 			multipath_end_bh_io(mp_bh, -EIO);
 		} else {
 			printk(KERN_ERR "multipath: %s: redirecting sector %llu"
 				" to another IO path\n",
 				bdevname(bio->bi_bdev,b),
-				(unsigned long long)bio->bi_sector);
+				(unsigned long long)bio->bi_iter.bi_sector);
 			*bio = *(mp_bh->master_bio);
-			bio->bi_sector += conf->multipaths[mp_bh->path].rdev->data_offset;
+			bio->bi_iter.bi_sector +=
+				conf->multipaths[mp_bh->path].rdev->data_offset;
 			bio->bi_bdev = conf->multipaths[mp_bh->path].rdev->bdev;
-			bio->bi_rw |= (1 << BIO_RW_FAILFAST);
+			bio->bi_rw |= REQ_FAILFAST_TRANSPORT;
 			bio->bi_end_io = multipath_end_request;
 			bio->bi_private = mp_bh;
 			generic_make_request(bio);
@@ -407,13 +381,24 @@ static void multipathd (mddev_t *mddev)
 	spin_unlock_irqrestore(&conf->device_lock, flags);
 }
 
-static int multipath_run (mddev_t *mddev)
+static sector_t multipath_size(struct mddev *mddev, sector_t sectors, int raid_disks)
+{
+	WARN_ONCE(sectors || raid_disks,
+		  "%s does not support generic reshape\n", __func__);
+
+	return mddev->dev_sectors;
+}
+
+static int multipath_run (struct mddev *mddev)
 {
-	multipath_conf_t *conf;
+	struct mpconf *conf;
 	int disk_idx;
 	struct multipath_info *disk;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	struct md_rdev *rdev;
+	int working_disks;
+
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
 
 	if (mddev->level != LEVEL_MULTIPATH) {
 		printk("multipath: %s: raid level not set to multipath IO (%d)\n",
@@ -425,9 +410,8 @@ static int multipath_run (mddev_t *mddev)
 	 * bookkeeping area. [whatever we allocate in multipath_run(),
 	 * should be freed in multipath_stop()]
 	 */
-	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
 
-	conf = kzalloc(sizeof(multipath_conf_t), GFP_KERNEL);
+	conf = kzalloc(sizeof(struct mpconf), GFP_KERNEL);
 	mddev->private = conf;
 	if (!conf) {
 		printk(KERN_ERR 
@@ -445,8 +429,8 @@ static int multipath_run (mddev_t *mddev)
 		goto out_free_conf;
 	}
 
-	conf->working_disks = 0;
-	rdev_for_each(rdev, tmp, mddev) {
+	working_disks = 0;
+	rdev_for_each(rdev, mddev) {
 		disk_idx = rdev->raid_disk;
 		if (disk_idx < 0 ||
 		    disk_idx >= mddev->raid_disks)
@@ -454,18 +438,20 @@ static int multipath_run (mddev_t *mddev)
 
 		disk = conf->multipaths + disk_idx;
 		disk->rdev = rdev;
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
 
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
 		/* as we don't honour merge_bvec_fn, we must never risk
 		 * violating it, not that we ever expect a device with
 		 * a merge_bvec_fn to be involved in multipath */
-		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
+		if (rdev->bdev->bd_disk->queue->merge_bvec_fn) {
+			blk_queue_max_segments(mddev->queue, 1);
+			blk_queue_segment_boundary(mddev->queue,
+						   PAGE_CACHE_SIZE - 1);
+		}
 
 		if (!test_bit(Faulty, &rdev->flags))
-			conf->working_disks++;
+			working_disks++;
 	}
 
 	conf->raid_disks = mddev->raid_disks;
@@ -473,14 +459,14 @@ static int multipath_run (mddev_t *mddev)
 	spin_lock_init(&conf->device_lock);
 	INIT_LIST_HEAD(&conf->retry_list);
 
-	if (!conf->working_disks) {
+	if (!working_disks) {
 		printk(KERN_ERR "multipath: no operational IO paths for %s\n",
 			mdname(mddev));
 		goto out_free_conf;
 	}
-	mddev->degraded = conf->raid_disks - conf->working_disks;
+	mddev->degraded = conf->raid_disks - working_disks;
 
-	conf->pool = mempool_create_kzalloc_pool(NR_RESERVED_BUFS,
+	conf->pool = mempool_create_kmalloc_pool(NR_RESERVED_BUFS,
 						 sizeof(struct multipath_bh));
 	if (conf->pool == NULL) {
 		printk(KERN_ERR 
@@ -490,7 +476,8 @@ static int multipath_run (mddev_t *mddev)
 	}
 
 	{
-		mddev->thread = md_register_thread(multipathd, mddev, "%s_multipath");
+		mddev->thread = md_register_thread(multipathd, mddev,
+						   "multipath");
 		if (!mddev->thread) {
 			printk(KERN_ERR "multipath: couldn't allocate thread"
 				" for %s\n", mdname(mddev));
@@ -500,16 +487,19 @@ static int multipath_run (mddev_t *mddev)
 
 	printk(KERN_INFO 
 		"multipath: array %s active with %d out of %d IO paths\n",
-		mdname(mddev), conf->working_disks, mddev->raid_disks);
+		mdname(mddev), conf->raid_disks - mddev->degraded,
+	       mddev->raid_disks);
 	/*
 	 * Ok, everything is just fine now
 	 */
-	mddev->array_sectors = mddev->size * 2;
+	md_set_array_sectors(mddev, multipath_size(mddev, 0, 0));
 
-	mddev->queue->unplug_fn = multipath_unplug;
 	mddev->queue->backing_dev_info.congested_fn = multipath_congested;
 	mddev->queue->backing_dev_info.congested_data = mddev;
 
+	if (md_integrity_register(mddev))
+		goto out_free_conf;
+
 	return 0;
 
 out_free_conf:
@@ -523,12 +513,11 @@ out:
 }
 
 
-static int multipath_stop (mddev_t *mddev)
+static int multipath_stop (struct mddev *mddev)
 {
-	multipath_conf_t *conf = mddev_to_conf(mddev);
+	struct mpconf *conf = mddev->private;
 
-	md_unregister_thread(mddev->thread);
-	mddev->thread = NULL;
+	md_unregister_thread(&mddev->thread);
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
 	mempool_destroy(conf->pool);
 	kfree(conf->multipaths);
@@ -537,7 +526,7 @@ static int multipath_stop (mddev_t *mddev)
 	return 0;
 }
 
-static struct mdk_personality multipath_personality =
+static struct md_personality multipath_personality =
 {
 	.name		= "multipath",
 	.level		= LEVEL_MULTIPATH,
@@ -549,6 +538,7 @@ static struct mdk_personality multipath_personality =
 	.error_handler	= multipath_error,
 	.hot_add_disk	= multipath_add_disk,
 	.hot_remove_disk= multipath_remove_disk,
+	.size		= multipath_size,
 };
 
 static int __init multipath_init (void)
@@ -564,6 +554,7 @@ static void __exit multipath_exit (void)
 module_init(multipath_init);
 module_exit(multipath_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("simple multi-path personality for MD");
 MODULE_ALIAS("md-personality-7"); /* MULTIPATH */
 MODULE_ALIAS("md-multipath");
 MODULE_ALIAS("md-level--4");
diff --git a/drivers/md/multipath.h b/drivers/md/multipath.h
new file mode 100644
index 00000000000..717c60f6289
--- /dev/null
+++ b/drivers/md/multipath.h
@@ -0,0 +1,31 @@
+#ifndef _MULTIPATH_H
+#define _MULTIPATH_H
+
+struct multipath_info {
+	struct md_rdev	*rdev;
+};
+
+struct mpconf {
+	struct mddev			*mddev;
+	struct multipath_info	*multipaths;
+	int			raid_disks;
+	spinlock_t		device_lock;
+	struct list_head	retry_list;
+
+	mempool_t		*pool;
+};
+
+/*
+ * this is our 'private' 'collective' MULTIPATH buffer head.
+ * it contains information about what kind of IO operations were started
+ * for this MULTIPATH operation, and about their status:
+ */
+
+struct multipath_bh {
+	struct mddev			*mddev;
+	struct bio		*master_bio;
+	struct bio		bio;
+	int			path;
+	struct list_head	retry_list;
+};
+#endif
diff --git a/drivers/md/persistent-data/Kconfig b/drivers/md/persistent-data/Kconfig
new file mode 100644
index 00000000000..0c2dec7aec2
--- /dev/null
+++ b/drivers/md/persistent-data/Kconfig
@@ -0,0 +1,18 @@
+config DM_PERSISTENT_DATA
+       tristate
+       depends on BLK_DEV_DM
+       select LIBCRC32C
+       select DM_BUFIO
+       ---help---
+	 Library providing immutable on-disk data structure support for
+	 device-mapper targets such as the thin provisioning target.
+
+config DM_DEBUG_BLOCK_STACK_TRACING
+       boolean "Keep stack trace of persistent data block lock holders"
+       depends on STACKTRACE_SUPPORT && DM_PERSISTENT_DATA
+       select STACKTRACE
+       ---help---
+	 Enable this for messages that may help debug problems with the
+	 block manager locking used by thin provisioning and caching.
+
+	 If unsure, say N.
diff --git a/drivers/md/persistent-data/Makefile b/drivers/md/persistent-data/Makefile
new file mode 100644
index 00000000000..ff528792c35
--- /dev/null
+++ b/drivers/md/persistent-data/Makefile
@@ -0,0 +1,12 @@
+obj-$(CONFIG_DM_PERSISTENT_DATA) += dm-persistent-data.o
+dm-persistent-data-objs := \
+	dm-array.o \
+	dm-bitset.o \
+	dm-block-manager.o \
+	dm-space-map-common.o \
+	dm-space-map-disk.o \
+	dm-space-map-metadata.o \
+	dm-transaction-manager.o \
+	dm-btree.o \
+	dm-btree-remove.o \
+	dm-btree-spine.o
diff --git a/drivers/md/persistent-data/dm-array.c b/drivers/md/persistent-data/dm-array.c
new file mode 100644
index 00000000000..1d75b1dc1e2
--- /dev/null
+++ b/drivers/md/persistent-data/dm-array.c
@@ -0,0 +1,819 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-array.h"
+#include "dm-space-map.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/export.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "array"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * The array is implemented as a fully populated btree, which points to
+ * blocks that contain the packed values.  This is more space efficient
+ * than just using a btree since we don't store 1 key per value.
+ */
+struct array_block {
+	__le32 csum;
+	__le32 max_entries;
+	__le32 nr_entries;
+	__le32 value_size;
+	__le64 blocknr; /* Block this node is supposed to live in. */
+} __packed;
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Validator methods.  As usual we calculate a checksum, and also write the
+ * block location into the header (paranoia about ssds remapping areas by
+ * mistake).
+ */
+#define CSUM_XOR 595846735
+
+static void array_block_prepare_for_write(struct dm_block_validator *v,
+					  struct dm_block *b,
+					  size_t size_of_block)
+{
+	struct array_block *bh_le = dm_block_data(b);
+
+	bh_le->blocknr = cpu_to_le64(dm_block_location(b));
+	bh_le->csum = cpu_to_le32(dm_bm_checksum(&bh_le->max_entries,
+						 size_of_block - sizeof(__le32),
+						 CSUM_XOR));
+}
+
+static int array_block_check(struct dm_block_validator *v,
+			     struct dm_block *b,
+			     size_t size_of_block)
+{
+	struct array_block *bh_le = dm_block_data(b);
+	__le32 csum_disk;
+
+	if (dm_block_location(b) != le64_to_cpu(bh_le->blocknr)) {
+		DMERR_LIMIT("array_block_check failed: blocknr %llu != wanted %llu",
+			    (unsigned long long) le64_to_cpu(bh_le->blocknr),
+			    (unsigned long long) dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	csum_disk = cpu_to_le32(dm_bm_checksum(&bh_le->max_entries,
+					       size_of_block - sizeof(__le32),
+					       CSUM_XOR));
+	if (csum_disk != bh_le->csum) {
+		DMERR_LIMIT("array_block_check failed: csum %u != wanted %u",
+			    (unsigned) le32_to_cpu(csum_disk),
+			    (unsigned) le32_to_cpu(bh_le->csum));
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+static struct dm_block_validator array_validator = {
+	.name = "array",
+	.prepare_for_write = array_block_prepare_for_write,
+	.check = array_block_check
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Functions for manipulating the array blocks.
+ */
+
+/*
+ * Returns a pointer to a value within an array block.
+ *
+ * index - The index into _this_ specific block.
+ */
+static void *element_at(struct dm_array_info *info, struct array_block *ab,
+			unsigned index)
+{
+	unsigned char *entry = (unsigned char *) (ab + 1);
+
+	entry += index * info->value_type.size;
+
+	return entry;
+}
+
+/*
+ * Utility function that calls one of the value_type methods on every value
+ * in an array block.
+ */
+static void on_entries(struct dm_array_info *info, struct array_block *ab,
+		       void (*fn)(void *, const void *))
+{
+	unsigned i, nr_entries = le32_to_cpu(ab->nr_entries);
+
+	for (i = 0; i < nr_entries; i++)
+		fn(info->value_type.context, element_at(info, ab, i));
+}
+
+/*
+ * Increment every value in an array block.
+ */
+static void inc_ablock_entries(struct dm_array_info *info, struct array_block *ab)
+{
+	struct dm_btree_value_type *vt = &info->value_type;
+
+	if (vt->inc)
+		on_entries(info, ab, vt->inc);
+}
+
+/*
+ * Decrement every value in an array block.
+ */
+static void dec_ablock_entries(struct dm_array_info *info, struct array_block *ab)
+{
+	struct dm_btree_value_type *vt = &info->value_type;
+
+	if (vt->dec)
+		on_entries(info, ab, vt->dec);
+}
+
+/*
+ * Each array block can hold this many values.
+ */
+static uint32_t calc_max_entries(size_t value_size, size_t size_of_block)
+{
+	return (size_of_block - sizeof(struct array_block)) / value_size;
+}
+
+/*
+ * Allocate a new array block.  The caller will need to unlock block.
+ */
+static int alloc_ablock(struct dm_array_info *info, size_t size_of_block,
+			uint32_t max_entries,
+			struct dm_block **block, struct array_block **ab)
+{
+	int r;
+
+	r = dm_tm_new_block(info->btree_info.tm, &array_validator, block);
+	if (r)
+		return r;
+
+	(*ab) = dm_block_data(*block);
+	(*ab)->max_entries = cpu_to_le32(max_entries);
+	(*ab)->nr_entries = cpu_to_le32(0);
+	(*ab)->value_size = cpu_to_le32(info->value_type.size);
+
+	return 0;
+}
+
+/*
+ * Pad an array block out with a particular value.  Every instance will
+ * cause an increment of the value_type.  new_nr must always be more than
+ * the current number of entries.
+ */
+static void fill_ablock(struct dm_array_info *info, struct array_block *ab,
+			const void *value, unsigned new_nr)
+{
+	unsigned i;
+	uint32_t nr_entries;
+	struct dm_btree_value_type *vt = &info->value_type;
+
+	BUG_ON(new_nr > le32_to_cpu(ab->max_entries));
+	BUG_ON(new_nr < le32_to_cpu(ab->nr_entries));
+
+	nr_entries = le32_to_cpu(ab->nr_entries);
+	for (i = nr_entries; i < new_nr; i++) {
+		if (vt->inc)
+			vt->inc(vt->context, value);
+		memcpy(element_at(info, ab, i), value, vt->size);
+	}
+	ab->nr_entries = cpu_to_le32(new_nr);
+}
+
+/*
+ * Remove some entries from the back of an array block.  Every value
+ * removed will be decremented.  new_nr must be <= the current number of
+ * entries.
+ */
+static void trim_ablock(struct dm_array_info *info, struct array_block *ab,
+			unsigned new_nr)
+{
+	unsigned i;
+	uint32_t nr_entries;
+	struct dm_btree_value_type *vt = &info->value_type;
+
+	BUG_ON(new_nr > le32_to_cpu(ab->max_entries));
+	BUG_ON(new_nr > le32_to_cpu(ab->nr_entries));
+
+	nr_entries = le32_to_cpu(ab->nr_entries);
+	for (i = nr_entries; i > new_nr; i--)
+		if (vt->dec)
+			vt->dec(vt->context, element_at(info, ab, i - 1));
+	ab->nr_entries = cpu_to_le32(new_nr);
+}
+
+/*
+ * Read locks a block, and coerces it to an array block.  The caller must
+ * unlock 'block' when finished.
+ */
+static int get_ablock(struct dm_array_info *info, dm_block_t b,
+		      struct dm_block **block, struct array_block **ab)
+{
+	int r;
+
+	r = dm_tm_read_lock(info->btree_info.tm, b, &array_validator, block);
+	if (r)
+		return r;
+
+	*ab = dm_block_data(*block);
+	return 0;
+}
+
+/*
+ * Unlocks an array block.
+ */
+static int unlock_ablock(struct dm_array_info *info, struct dm_block *block)
+{
+	return dm_tm_unlock(info->btree_info.tm, block);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Btree manipulation.
+ */
+
+/*
+ * Looks up an array block in the btree, and then read locks it.
+ *
+ * index is the index of the index of the array_block, (ie. the array index
+ * / max_entries).
+ */
+static int lookup_ablock(struct dm_array_info *info, dm_block_t root,
+			 unsigned index, struct dm_block **block,
+			 struct array_block **ab)
+{
+	int r;
+	uint64_t key = index;
+	__le64 block_le;
+
+	r = dm_btree_lookup(&info->btree_info, root, &key, &block_le);
+	if (r)
+		return r;
+
+	return get_ablock(info, le64_to_cpu(block_le), block, ab);
+}
+
+/*
+ * Insert an array block into the btree.  The block is _not_ unlocked.
+ */
+static int insert_ablock(struct dm_array_info *info, uint64_t index,
+			 struct dm_block *block, dm_block_t *root)
+{
+	__le64 block_le = cpu_to_le64(dm_block_location(block));
+
+	__dm_bless_for_disk(block_le);
+	return dm_btree_insert(&info->btree_info, *root, &index, &block_le, root);
+}
+
+/*
+ * Looks up an array block in the btree.  Then shadows it, and updates the
+ * btree to point to this new shadow.  'root' is an input/output parameter
+ * for both the current root block, and the new one.
+ */
+static int shadow_ablock(struct dm_array_info *info, dm_block_t *root,
+			 unsigned index, struct dm_block **block,
+			 struct array_block **ab)
+{
+	int r, inc;
+	uint64_t key = index;
+	dm_block_t b;
+	__le64 block_le;
+
+	/*
+	 * lookup
+	 */
+	r = dm_btree_lookup(&info->btree_info, *root, &key, &block_le);
+	if (r)
+		return r;
+	b = le64_to_cpu(block_le);
+
+	/*
+	 * shadow
+	 */
+	r = dm_tm_shadow_block(info->btree_info.tm, b,
+			       &array_validator, block, &inc);
+	if (r)
+		return r;
+
+	*ab = dm_block_data(*block);
+	if (inc)
+		inc_ablock_entries(info, *ab);
+
+	/*
+	 * Reinsert.
+	 *
+	 * The shadow op will often be a noop.  Only insert if it really
+	 * copied data.
+	 */
+	if (dm_block_location(*block) != b) {
+		/*
+		 * dm_tm_shadow_block will have already decremented the old
+		 * block, but it is still referenced by the btree.  We
+		 * increment to stop the insert decrementing it below zero
+		 * when overwriting the old value.
+		 */
+		dm_tm_inc(info->btree_info.tm, b);
+		r = insert_ablock(info, index, *block, root);
+	}
+
+	return r;
+}
+
+/*
+ * Allocate an new array block, and fill it with some values.
+ */
+static int insert_new_ablock(struct dm_array_info *info, size_t size_of_block,
+			     uint32_t max_entries,
+			     unsigned block_index, uint32_t nr,
+			     const void *value, dm_block_t *root)
+{
+	int r;
+	struct dm_block *block;
+	struct array_block *ab;
+
+	r = alloc_ablock(info, size_of_block, max_entries, &block, &ab);
+	if (r)
+		return r;
+
+	fill_ablock(info, ab, value, nr);
+	r = insert_ablock(info, block_index, block, root);
+	unlock_ablock(info, block);
+
+	return r;
+}
+
+static int insert_full_ablocks(struct dm_array_info *info, size_t size_of_block,
+			       unsigned begin_block, unsigned end_block,
+			       unsigned max_entries, const void *value,
+			       dm_block_t *root)
+{
+	int r = 0;
+
+	for (; !r && begin_block != end_block; begin_block++)
+		r = insert_new_ablock(info, size_of_block, max_entries, begin_block, max_entries, value, root);
+
+	return r;
+}
+
+/*
+ * There are a bunch of functions involved with resizing an array.  This
+ * structure holds information that commonly needed by them.  Purely here
+ * to reduce parameter count.
+ */
+struct resize {
+	/*
+	 * Describes the array.
+	 */
+	struct dm_array_info *info;
+
+	/*
+	 * The current root of the array.  This gets updated.
+	 */
+	dm_block_t root;
+
+	/*
+	 * Metadata block size.  Used to calculate the nr entries in an
+	 * array block.
+	 */
+	size_t size_of_block;
+
+	/*
+	 * Maximum nr entries in an array block.
+	 */
+	unsigned max_entries;
+
+	/*
+	 * nr of completely full blocks in the array.
+	 *
+	 * 'old' refers to before the resize, 'new' after.
+	 */
+	unsigned old_nr_full_blocks, new_nr_full_blocks;
+
+	/*
+	 * Number of entries in the final block.  0 iff only full blocks in
+	 * the array.
+	 */
+	unsigned old_nr_entries_in_last_block, new_nr_entries_in_last_block;
+
+	/*
+	 * The default value used when growing the array.
+	 */
+	const void *value;
+};
+
+/*
+ * Removes a consecutive set of array blocks from the btree.  The values
+ * in block are decremented as a side effect of the btree remove.
+ *
+ * begin_index - the index of the first array block to remove.
+ * end_index - the one-past-the-end value.  ie. this block is not removed.
+ */
+static int drop_blocks(struct resize *resize, unsigned begin_index,
+		       unsigned end_index)
+{
+	int r;
+
+	while (begin_index != end_index) {
+		uint64_t key = begin_index++;
+		r = dm_btree_remove(&resize->info->btree_info, resize->root,
+				    &key, &resize->root);
+		if (r)
+			return r;
+	}
+
+	return 0;
+}
+
+/*
+ * Calculates how many blocks are needed for the array.
+ */
+static unsigned total_nr_blocks_needed(unsigned nr_full_blocks,
+				       unsigned nr_entries_in_last_block)
+{
+	return nr_full_blocks + (nr_entries_in_last_block ? 1 : 0);
+}
+
+/*
+ * Shrink an array.
+ */
+static int shrink(struct resize *resize)
+{
+	int r;
+	unsigned begin, end;
+	struct dm_block *block;
+	struct array_block *ab;
+
+	/*
+	 * Lose some blocks from the back?
+	 */
+	if (resize->new_nr_full_blocks < resize->old_nr_full_blocks) {
+		begin = total_nr_blocks_needed(resize->new_nr_full_blocks,
+					       resize->new_nr_entries_in_last_block);
+		end = total_nr_blocks_needed(resize->old_nr_full_blocks,
+					     resize->old_nr_entries_in_last_block);
+
+		r = drop_blocks(resize, begin, end);
+		if (r)
+			return r;
+	}
+
+	/*
+	 * Trim the new tail block
+	 */
+	if (resize->new_nr_entries_in_last_block) {
+		r = shadow_ablock(resize->info, &resize->root,
+				  resize->new_nr_full_blocks, &block, &ab);
+		if (r)
+			return r;
+
+		trim_ablock(resize->info, ab, resize->new_nr_entries_in_last_block);
+		unlock_ablock(resize->info, block);
+	}
+
+	return 0;
+}
+
+/*
+ * Grow an array.
+ */
+static int grow_extend_tail_block(struct resize *resize, uint32_t new_nr_entries)
+{
+	int r;
+	struct dm_block *block;
+	struct array_block *ab;
+
+	r = shadow_ablock(resize->info, &resize->root,
+			  resize->old_nr_full_blocks, &block, &ab);
+	if (r)
+		return r;
+
+	fill_ablock(resize->info, ab, resize->value, new_nr_entries);
+	unlock_ablock(resize->info, block);
+
+	return r;
+}
+
+static int grow_add_tail_block(struct resize *resize)
+{
+	return insert_new_ablock(resize->info, resize->size_of_block,
+				 resize->max_entries,
+				 resize->new_nr_full_blocks,
+				 resize->new_nr_entries_in_last_block,
+				 resize->value, &resize->root);
+}
+
+static int grow_needs_more_blocks(struct resize *resize)
+{
+	int r;
+	unsigned old_nr_blocks = resize->old_nr_full_blocks;
+
+	if (resize->old_nr_entries_in_last_block > 0) {
+		old_nr_blocks++;
+
+		r = grow_extend_tail_block(resize, resize->max_entries);
+		if (r)
+			return r;
+	}
+
+	r = insert_full_ablocks(resize->info, resize->size_of_block,
+				old_nr_blocks,
+				resize->new_nr_full_blocks,
+				resize->max_entries, resize->value,
+				&resize->root);
+	if (r)
+		return r;
+
+	if (resize->new_nr_entries_in_last_block)
+		r = grow_add_tail_block(resize);
+
+	return r;
+}
+
+static int grow(struct resize *resize)
+{
+	if (resize->new_nr_full_blocks > resize->old_nr_full_blocks)
+		return grow_needs_more_blocks(resize);
+
+	else if (resize->old_nr_entries_in_last_block)
+		return grow_extend_tail_block(resize, resize->new_nr_entries_in_last_block);
+
+	else
+		return grow_add_tail_block(resize);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * These are the value_type functions for the btree elements, which point
+ * to array blocks.
+ */
+static void block_inc(void *context, const void *value)
+{
+	__le64 block_le;
+	struct dm_array_info *info = context;
+
+	memcpy(&block_le, value, sizeof(block_le));
+	dm_tm_inc(info->btree_info.tm, le64_to_cpu(block_le));
+}
+
+static void block_dec(void *context, const void *value)
+{
+	int r;
+	uint64_t b;
+	__le64 block_le;
+	uint32_t ref_count;
+	struct dm_block *block;
+	struct array_block *ab;
+	struct dm_array_info *info = context;
+
+	memcpy(&block_le, value, sizeof(block_le));
+	b = le64_to_cpu(block_le);
+
+	r = dm_tm_ref(info->btree_info.tm, b, &ref_count);
+	if (r) {
+		DMERR_LIMIT("couldn't get reference count for block %llu",
+			    (unsigned long long) b);
+		return;
+	}
+
+	if (ref_count == 1) {
+		/*
+		 * We're about to drop the last reference to this ablock.
+		 * So we need to decrement the ref count of the contents.
+		 */
+		r = get_ablock(info, b, &block, &ab);
+		if (r) {
+			DMERR_LIMIT("couldn't get array block %llu",
+				    (unsigned long long) b);
+			return;
+		}
+
+		dec_ablock_entries(info, ab);
+		unlock_ablock(info, block);
+	}
+
+	dm_tm_dec(info->btree_info.tm, b);
+}
+
+static int block_equal(void *context, const void *value1, const void *value2)
+{
+	return !memcmp(value1, value2, sizeof(__le64));
+}
+
+/*----------------------------------------------------------------*/
+
+void dm_array_info_init(struct dm_array_info *info,
+			struct dm_transaction_manager *tm,
+			struct dm_btree_value_type *vt)
+{
+	struct dm_btree_value_type *bvt = &info->btree_info.value_type;
+
+	memcpy(&info->value_type, vt, sizeof(info->value_type));
+	info->btree_info.tm = tm;
+	info->btree_info.levels = 1;
+
+	bvt->context = info;
+	bvt->size = sizeof(__le64);
+	bvt->inc = block_inc;
+	bvt->dec = block_dec;
+	bvt->equal = block_equal;
+}
+EXPORT_SYMBOL_GPL(dm_array_info_init);
+
+int dm_array_empty(struct dm_array_info *info, dm_block_t *root)
+{
+	return dm_btree_empty(&info->btree_info, root);
+}
+EXPORT_SYMBOL_GPL(dm_array_empty);
+
+static int array_resize(struct dm_array_info *info, dm_block_t root,
+			uint32_t old_size, uint32_t new_size,
+			const void *value, dm_block_t *new_root)
+{
+	int r;
+	struct resize resize;
+
+	if (old_size == new_size)
+		return 0;
+
+	resize.info = info;
+	resize.root = root;
+	resize.size_of_block = dm_bm_block_size(dm_tm_get_bm(info->btree_info.tm));
+	resize.max_entries = calc_max_entries(info->value_type.size,
+					      resize.size_of_block);
+
+	resize.old_nr_full_blocks = old_size / resize.max_entries;
+	resize.old_nr_entries_in_last_block = old_size % resize.max_entries;
+	resize.new_nr_full_blocks = new_size / resize.max_entries;
+	resize.new_nr_entries_in_last_block = new_size % resize.max_entries;
+	resize.value = value;
+
+	r = ((new_size > old_size) ? grow : shrink)(&resize);
+	if (r)
+		return r;
+
+	*new_root = resize.root;
+	return 0;
+}
+
+int dm_array_resize(struct dm_array_info *info, dm_block_t root,
+		    uint32_t old_size, uint32_t new_size,
+		    const void *value, dm_block_t *new_root)
+		    __dm_written_to_disk(value)
+{
+	int r = array_resize(info, root, old_size, new_size, value, new_root);
+	__dm_unbless_for_disk(value);
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_array_resize);
+
+int dm_array_del(struct dm_array_info *info, dm_block_t root)
+{
+	return dm_btree_del(&info->btree_info, root);
+}
+EXPORT_SYMBOL_GPL(dm_array_del);
+
+int dm_array_get_value(struct dm_array_info *info, dm_block_t root,
+		       uint32_t index, void *value_le)
+{
+	int r;
+	struct dm_block *block;
+	struct array_block *ab;
+	size_t size_of_block;
+	unsigned entry, max_entries;
+
+	size_of_block = dm_bm_block_size(dm_tm_get_bm(info->btree_info.tm));
+	max_entries = calc_max_entries(info->value_type.size, size_of_block);
+
+	r = lookup_ablock(info, root, index / max_entries, &block, &ab);
+	if (r)
+		return r;
+
+	entry = index % max_entries;
+	if (entry >= le32_to_cpu(ab->nr_entries))
+		r = -ENODATA;
+	else
+		memcpy(value_le, element_at(info, ab, entry),
+		       info->value_type.size);
+
+	unlock_ablock(info, block);
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_array_get_value);
+
+static int array_set_value(struct dm_array_info *info, dm_block_t root,
+			   uint32_t index, const void *value, dm_block_t *new_root)
+{
+	int r;
+	struct dm_block *block;
+	struct array_block *ab;
+	size_t size_of_block;
+	unsigned max_entries;
+	unsigned entry;
+	void *old_value;
+	struct dm_btree_value_type *vt = &info->value_type;
+
+	size_of_block = dm_bm_block_size(dm_tm_get_bm(info->btree_info.tm));
+	max_entries = calc_max_entries(info->value_type.size, size_of_block);
+
+	r = shadow_ablock(info, &root, index / max_entries, &block, &ab);
+	if (r)
+		return r;
+	*new_root = root;
+
+	entry = index % max_entries;
+	if (entry >= le32_to_cpu(ab->nr_entries)) {
+		r = -ENODATA;
+		goto out;
+	}
+
+	old_value = element_at(info, ab, entry);
+	if (vt->dec &&
+	    (!vt->equal || !vt->equal(vt->context, old_value, value))) {
+		vt->dec(vt->context, old_value);
+		if (vt->inc)
+			vt->inc(vt->context, value);
+	}
+
+	memcpy(old_value, value, info->value_type.size);
+
+out:
+	unlock_ablock(info, block);
+	return r;
+}
+
+int dm_array_set_value(struct dm_array_info *info, dm_block_t root,
+		 uint32_t index, const void *value, dm_block_t *new_root)
+		 __dm_written_to_disk(value)
+{
+	int r;
+
+	r = array_set_value(info, root, index, value, new_root);
+	__dm_unbless_for_disk(value);
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_array_set_value);
+
+struct walk_info {
+	struct dm_array_info *info;
+	int (*fn)(void *context, uint64_t key, void *leaf);
+	void *context;
+};
+
+static int walk_ablock(void *context, uint64_t *keys, void *leaf)
+{
+	struct walk_info *wi = context;
+
+	int r;
+	unsigned i;
+	__le64 block_le;
+	unsigned nr_entries, max_entries;
+	struct dm_block *block;
+	struct array_block *ab;
+
+	memcpy(&block_le, leaf, sizeof(block_le));
+	r = get_ablock(wi->info, le64_to_cpu(block_le), &block, &ab);
+	if (r)
+		return r;
+
+	max_entries = le32_to_cpu(ab->max_entries);
+	nr_entries = le32_to_cpu(ab->nr_entries);
+	for (i = 0; i < nr_entries; i++) {
+		r = wi->fn(wi->context, keys[0] * max_entries + i,
+			   element_at(wi->info, ab, i));
+
+		if (r)
+			break;
+	}
+
+	unlock_ablock(wi->info, block);
+	return r;
+}
+
+int dm_array_walk(struct dm_array_info *info, dm_block_t root,
+		  int (*fn)(void *, uint64_t key, void *leaf),
+		  void *context)
+{
+	struct walk_info wi;
+
+	wi.info = info;
+	wi.fn = fn;
+	wi.context = context;
+
+	return dm_btree_walk(&info->btree_info, root, walk_ablock, &wi);
+}
+EXPORT_SYMBOL_GPL(dm_array_walk);
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-array.h b/drivers/md/persistent-data/dm-array.h
new file mode 100644
index 00000000000..ea177d6fa58
--- /dev/null
+++ b/drivers/md/persistent-data/dm-array.h
@@ -0,0 +1,166 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+#ifndef _LINUX_DM_ARRAY_H
+#define _LINUX_DM_ARRAY_H
+
+#include "dm-btree.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * The dm-array is a persistent version of an array.  It packs the data
+ * more efficiently than a btree which will result in less disk space use,
+ * and a performance boost.  The element get and set operations are still
+ * O(ln(n)), but with a much smaller constant.
+ *
+ * The value type structure is reused from the btree type to support proper
+ * reference counting of values.
+ *
+ * The arrays implicitly know their length, and bounds are checked for
+ * lookups and updated.  It doesn't store this in an accessible place
+ * because it would waste a whole metadata block.  Make sure you store the
+ * size along with the array root in your encompassing data.
+ *
+ * Array entries are indexed via an unsigned integer starting from zero.
+ * Arrays are not sparse; if you resize an array to have 'n' entries then
+ * 'n - 1' will be the last valid index.
+ *
+ * Typical use:
+ *
+ * a) initialise a dm_array_info structure.  This describes the array
+ *    values and ties it into a specific transaction manager.  It holds no
+ *    instance data; the same info can be used for many similar arrays if
+ *    you wish.
+ *
+ * b) Get yourself a root.  The root is the index of a block of data on the
+ *    disk that holds a particular instance of an array.  You may have a
+ *    pre existing root in your metadata that you wish to use, or you may
+ *    want to create a brand new, empty array with dm_array_empty().
+ *
+ * Like the other data structures in this library, dm_array objects are
+ * immutable between transactions.  Update functions will return you the
+ * root for a _new_ array.  If you've incremented the old root, via
+ * dm_tm_inc(), before calling the update function you may continue to use
+ * it in parallel with the new root.
+ *
+ * c) resize an array with dm_array_resize().
+ *
+ * d) Get a value from the array with dm_array_get_value().
+ *
+ * e) Set a value in the array with dm_array_set_value().
+ *
+ * f) Walk an array of values in index order with dm_array_walk().  More
+ *    efficient than making many calls to dm_array_get_value().
+ *
+ * g) Destroy the array with dm_array_del().  This tells the transaction
+ *    manager that you're no longer using this data structure so it can
+ *    recycle it's blocks.  (dm_array_dec() would be a better name for it,
+ *    but del is in keeping with dm_btree_del()).
+ */
+
+/*
+ * Describes an array.  Don't initialise this structure yourself, use the
+ * init function below.
+ */
+struct dm_array_info {
+	struct dm_transaction_manager *tm;
+	struct dm_btree_value_type value_type;
+	struct dm_btree_info btree_info;
+};
+
+/*
+ * Sets up a dm_array_info structure.  You don't need to do anything with
+ * this structure when you finish using it.
+ *
+ * info - the structure being filled in.
+ * tm   - the transaction manager that should supervise this structure.
+ * vt   - describes the leaf values.
+ */
+void dm_array_info_init(struct dm_array_info *info,
+			struct dm_transaction_manager *tm,
+			struct dm_btree_value_type *vt);
+
+/*
+ * Create an empty, zero length array.
+ *
+ * info - describes the array
+ * root - on success this will be filled out with the root block
+ */
+int dm_array_empty(struct dm_array_info *info, dm_block_t *root);
+
+/*
+ * Resizes the array.
+ *
+ * info - describes the array
+ * root - the root block of the array on disk
+ * old_size - the caller is responsible for remembering the size of
+ *            the array
+ * new_size - can be bigger or smaller than old_size
+ * value - if we're growing the array the new entries will have this value
+ * new_root - on success, points to the new root block
+ *
+ * If growing the inc function for 'value' will be called the appropriate
+ * number of times.  So if the caller is holding a reference they may want
+ * to drop it.
+ */
+int dm_array_resize(struct dm_array_info *info, dm_block_t root,
+		    uint32_t old_size, uint32_t new_size,
+		    const void *value, dm_block_t *new_root)
+	__dm_written_to_disk(value);
+
+/*
+ * Frees a whole array.  The value_type's decrement operation will be called
+ * for all values in the array
+ */
+int dm_array_del(struct dm_array_info *info, dm_block_t root);
+
+/*
+ * Lookup a value in the array
+ *
+ * info - describes the array
+ * root - root block of the array
+ * index - array index
+ * value - the value to be read.  Will be in on-disk format of course.
+ *
+ * -ENODATA will be returned if the index is out of bounds.
+ */
+int dm_array_get_value(struct dm_array_info *info, dm_block_t root,
+		       uint32_t index, void *value);
+
+/*
+ * Set an entry in the array.
+ *
+ * info - describes the array
+ * root - root block of the array
+ * index - array index
+ * value - value to be written to disk.  Make sure you confirm the value is
+ *         in on-disk format with__dm_bless_for_disk() before calling.
+ * new_root - the new root block
+ *
+ * The old value being overwritten will be decremented, the new value
+ * incremented.
+ *
+ * -ENODATA will be returned if the index is out of bounds.
+ */
+int dm_array_set_value(struct dm_array_info *info, dm_block_t root,
+		       uint32_t index, const void *value, dm_block_t *new_root)
+	__dm_written_to_disk(value);
+
+/*
+ * Walk through all the entries in an array.
+ *
+ * info - describes the array
+ * root - root block of the array
+ * fn - called back for every element
+ * context - passed to the callback
+ */
+int dm_array_walk(struct dm_array_info *info, dm_block_t root,
+		  int (*fn)(void *context, uint64_t key, void *leaf),
+		  void *context);
+
+/*----------------------------------------------------------------*/
+
+#endif	/* _LINUX_DM_ARRAY_H */
diff --git a/drivers/md/persistent-data/dm-bitset.c b/drivers/md/persistent-data/dm-bitset.c
new file mode 100644
index 00000000000..36f7cc2c710
--- /dev/null
+++ b/drivers/md/persistent-data/dm-bitset.c
@@ -0,0 +1,171 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-bitset.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/export.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "bitset"
+#define BITS_PER_ARRAY_ENTRY 64
+
+/*----------------------------------------------------------------*/
+
+static struct dm_btree_value_type bitset_bvt = {
+	.context = NULL,
+	.size = sizeof(__le64),
+	.inc = NULL,
+	.dec = NULL,
+	.equal = NULL,
+};
+
+/*----------------------------------------------------------------*/
+
+void dm_disk_bitset_init(struct dm_transaction_manager *tm,
+			 struct dm_disk_bitset *info)
+{
+	dm_array_info_init(&info->array_info, tm, &bitset_bvt);
+	info->current_index_set = false;
+}
+EXPORT_SYMBOL_GPL(dm_disk_bitset_init);
+
+int dm_bitset_empty(struct dm_disk_bitset *info, dm_block_t *root)
+{
+	return dm_array_empty(&info->array_info, root);
+}
+EXPORT_SYMBOL_GPL(dm_bitset_empty);
+
+int dm_bitset_resize(struct dm_disk_bitset *info, dm_block_t root,
+		     uint32_t old_nr_entries, uint32_t new_nr_entries,
+		     bool default_value, dm_block_t *new_root)
+{
+	uint32_t old_blocks = dm_div_up(old_nr_entries, BITS_PER_ARRAY_ENTRY);
+	uint32_t new_blocks = dm_div_up(new_nr_entries, BITS_PER_ARRAY_ENTRY);
+	__le64 value = default_value ? cpu_to_le64(~0) : cpu_to_le64(0);
+
+	__dm_bless_for_disk(&value);
+	return dm_array_resize(&info->array_info, root, old_blocks, new_blocks,
+			       &value, new_root);
+}
+EXPORT_SYMBOL_GPL(dm_bitset_resize);
+
+int dm_bitset_del(struct dm_disk_bitset *info, dm_block_t root)
+{
+	return dm_array_del(&info->array_info, root);
+}
+EXPORT_SYMBOL_GPL(dm_bitset_del);
+
+int dm_bitset_flush(struct dm_disk_bitset *info, dm_block_t root,
+		    dm_block_t *new_root)
+{
+	int r;
+	__le64 value;
+
+	if (!info->current_index_set || !info->dirty)
+		return 0;
+
+	value = cpu_to_le64(info->current_bits);
+
+	__dm_bless_for_disk(&value);
+	r = dm_array_set_value(&info->array_info, root, info->current_index,
+			       &value, new_root);
+	if (r)
+		return r;
+
+	info->current_index_set = false;
+	info->dirty = false;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bitset_flush);
+
+static int read_bits(struct dm_disk_bitset *info, dm_block_t root,
+		     uint32_t array_index)
+{
+	int r;
+	__le64 value;
+
+	r = dm_array_get_value(&info->array_info, root, array_index, &value);
+	if (r)
+		return r;
+
+	info->current_bits = le64_to_cpu(value);
+	info->current_index_set = true;
+	info->current_index = array_index;
+	info->dirty = false;
+
+	return 0;
+}
+
+static int get_array_entry(struct dm_disk_bitset *info, dm_block_t root,
+			   uint32_t index, dm_block_t *new_root)
+{
+	int r;
+	unsigned array_index = index / BITS_PER_ARRAY_ENTRY;
+
+	if (info->current_index_set) {
+		if (info->current_index == array_index)
+			return 0;
+
+		r = dm_bitset_flush(info, root, new_root);
+		if (r)
+			return r;
+	}
+
+	return read_bits(info, root, array_index);
+}
+
+int dm_bitset_set_bit(struct dm_disk_bitset *info, dm_block_t root,
+		      uint32_t index, dm_block_t *new_root)
+{
+	int r;
+	unsigned b = index % BITS_PER_ARRAY_ENTRY;
+
+	r = get_array_entry(info, root, index, new_root);
+	if (r)
+		return r;
+
+	set_bit(b, (unsigned long *) &info->current_bits);
+	info->dirty = true;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bitset_set_bit);
+
+int dm_bitset_clear_bit(struct dm_disk_bitset *info, dm_block_t root,
+			uint32_t index, dm_block_t *new_root)
+{
+	int r;
+	unsigned b = index % BITS_PER_ARRAY_ENTRY;
+
+	r = get_array_entry(info, root, index, new_root);
+	if (r)
+		return r;
+
+	clear_bit(b, (unsigned long *) &info->current_bits);
+	info->dirty = true;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bitset_clear_bit);
+
+int dm_bitset_test_bit(struct dm_disk_bitset *info, dm_block_t root,
+		       uint32_t index, dm_block_t *new_root, bool *result)
+{
+	int r;
+	unsigned b = index % BITS_PER_ARRAY_ENTRY;
+
+	r = get_array_entry(info, root, index, new_root);
+	if (r)
+		return r;
+
+	*result = test_bit(b, (unsigned long *) &info->current_bits);
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bitset_test_bit);
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-bitset.h b/drivers/md/persistent-data/dm-bitset.h
new file mode 100644
index 00000000000..c2287d672ef
--- /dev/null
+++ b/drivers/md/persistent-data/dm-bitset.h
@@ -0,0 +1,166 @@
+/*
+ * Copyright (C) 2012 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+#ifndef _LINUX_DM_BITSET_H
+#define _LINUX_DM_BITSET_H
+
+#include "dm-array.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * This bitset type is a thin wrapper round a dm_array of 64bit words.  It
+ * uses a tiny, one word cache to reduce the number of array lookups and so
+ * increase performance.
+ *
+ * Like the dm-array that it's based on, the caller needs to keep track of
+ * the size of the bitset separately.  The underlying dm-array implicitly
+ * knows how many words it's storing and will return -ENODATA if you try
+ * and access an out of bounds word.  However, an out of bounds bit in the
+ * final word will _not_ be detected, you have been warned.
+ *
+ * Bits are indexed from zero.
+
+ * Typical use:
+ *
+ * a) Initialise a dm_disk_bitset structure with dm_disk_bitset_init().
+ *    This describes the bitset and includes the cache.  It's not called it
+ *    dm_bitset_info in line with other data structures because it does
+ *    include instance data.
+ *
+ * b) Get yourself a root.  The root is the index of a block of data on the
+ *    disk that holds a particular instance of an bitset.  You may have a
+ *    pre existing root in your metadata that you wish to use, or you may
+ *    want to create a brand new, empty bitset with dm_bitset_empty().
+ *
+ * Like the other data structures in this library, dm_bitset objects are
+ * immutable between transactions.  Update functions will return you the
+ * root for a _new_ array.  If you've incremented the old root, via
+ * dm_tm_inc(), before calling the update function you may continue to use
+ * it in parallel with the new root.
+ *
+ * Even read operations may trigger the cache to be flushed and as such
+ * return a root for a new, updated bitset.
+ *
+ * c) resize a bitset with dm_bitset_resize().
+ *
+ * d) Set a bit with dm_bitset_set_bit().
+ *
+ * e) Clear a bit with dm_bitset_clear_bit().
+ *
+ * f) Test a bit with dm_bitset_test_bit().
+ *
+ * g) Flush all updates from the cache with dm_bitset_flush().
+ *
+ * h) Destroy the bitset with dm_bitset_del().  This tells the transaction
+ *    manager that you're no longer using this data structure so it can
+ *    recycle it's blocks.  (dm_bitset_dec() would be a better name for it,
+ *    but del is in keeping with dm_btree_del()).
+ */
+
+/*
+ * Opaque object.  Unlike dm_array_info, you should have one of these per
+ * bitset.  Initialise with dm_disk_bitset_init().
+ */
+struct dm_disk_bitset {
+	struct dm_array_info array_info;
+
+	uint32_t current_index;
+	uint64_t current_bits;
+
+	bool current_index_set:1;
+	bool dirty:1;
+};
+
+/*
+ * Sets up a dm_disk_bitset structure.  You don't need to do anything with
+ * this structure when you finish using it.
+ *
+ * tm - the transaction manager that should supervise this structure
+ * info - the structure being initialised
+ */
+void dm_disk_bitset_init(struct dm_transaction_manager *tm,
+			 struct dm_disk_bitset *info);
+
+/*
+ * Create an empty, zero length bitset.
+ *
+ * info - describes the bitset
+ * new_root - on success, points to the new root block
+ */
+int dm_bitset_empty(struct dm_disk_bitset *info, dm_block_t *new_root);
+
+/*
+ * Resize the bitset.
+ *
+ * info - describes the bitset
+ * old_root - the root block of the array on disk
+ * old_nr_entries - the number of bits in the old bitset
+ * new_nr_entries - the number of bits you want in the new bitset
+ * default_value - the value for any new bits
+ * new_root - on success, points to the new root block
+ */
+int dm_bitset_resize(struct dm_disk_bitset *info, dm_block_t old_root,
+		     uint32_t old_nr_entries, uint32_t new_nr_entries,
+		     bool default_value, dm_block_t *new_root);
+
+/*
+ * Frees the bitset.
+ */
+int dm_bitset_del(struct dm_disk_bitset *info, dm_block_t root);
+
+/*
+ * Set a bit.
+ *
+ * info - describes the bitset
+ * root - the root block of the bitset
+ * index - the bit index
+ * new_root - on success, points to the new root block
+ *
+ * -ENODATA will be returned if the index is out of bounds.
+ */
+int dm_bitset_set_bit(struct dm_disk_bitset *info, dm_block_t root,
+		      uint32_t index, dm_block_t *new_root);
+
+/*
+ * Clears a bit.
+ *
+ * info - describes the bitset
+ * root - the root block of the bitset
+ * index - the bit index
+ * new_root - on success, points to the new root block
+ *
+ * -ENODATA will be returned if the index is out of bounds.
+ */
+int dm_bitset_clear_bit(struct dm_disk_bitset *info, dm_block_t root,
+			uint32_t index, dm_block_t *new_root);
+
+/*
+ * Tests a bit.
+ *
+ * info - describes the bitset
+ * root - the root block of the bitset
+ * index - the bit index
+ * new_root - on success, points to the new root block (cached values may have been written)
+ * result - the bit value you're after
+ *
+ * -ENODATA will be returned if the index is out of bounds.
+ */
+int dm_bitset_test_bit(struct dm_disk_bitset *info, dm_block_t root,
+		       uint32_t index, dm_block_t *new_root, bool *result);
+
+/*
+ * Flush any cached changes to disk.
+ *
+ * info - describes the bitset
+ * root - the root block of the bitset
+ * new_root - on success, points to the new root block
+ */
+int dm_bitset_flush(struct dm_disk_bitset *info, dm_block_t root,
+		    dm_block_t *new_root);
+
+/*----------------------------------------------------------------*/
+
+#endif /* _LINUX_DM_BITSET_H */
diff --git a/drivers/md/persistent-data/dm-block-manager.c b/drivers/md/persistent-data/dm-block-manager.c
new file mode 100644
index 00000000000..087411c95ff
--- /dev/null
+++ b/drivers/md/persistent-data/dm-block-manager.c
@@ -0,0 +1,636 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+#include "dm-block-manager.h"
+#include "dm-persistent-data-internal.h"
+#include "../dm-bufio.h"
+
+#include <linux/crc32c.h>
+#include <linux/module.h>
+#include <linux/slab.h>
+#include <linux/rwsem.h>
+#include <linux/device-mapper.h>
+#include <linux/stacktrace.h>
+
+#define DM_MSG_PREFIX "block manager"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * This is a read/write semaphore with a couple of differences.
+ *
+ * i) There is a restriction on the number of concurrent read locks that
+ * may be held at once.  This is just an implementation detail.
+ *
+ * ii) Recursive locking attempts are detected and return EINVAL.  A stack
+ * trace is also emitted for the previous lock acquisition.
+ *
+ * iii) Priority is given to write locks.
+ */
+#define MAX_HOLDERS 4
+#define MAX_STACK 10
+
+typedef unsigned long stack_entries[MAX_STACK];
+
+struct block_lock {
+	spinlock_t lock;
+	__s32 count;
+	struct list_head waiters;
+	struct task_struct *holders[MAX_HOLDERS];
+
+#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
+	struct stack_trace traces[MAX_HOLDERS];
+	stack_entries entries[MAX_HOLDERS];
+#endif
+};
+
+struct waiter {
+	struct list_head list;
+	struct task_struct *task;
+	int wants_write;
+};
+
+static unsigned __find_holder(struct block_lock *lock,
+			      struct task_struct *task)
+{
+	unsigned i;
+
+	for (i = 0; i < MAX_HOLDERS; i++)
+		if (lock->holders[i] == task)
+			break;
+
+	BUG_ON(i == MAX_HOLDERS);
+	return i;
+}
+
+/* call this *after* you increment lock->count */
+static void __add_holder(struct block_lock *lock, struct task_struct *task)
+{
+	unsigned h = __find_holder(lock, NULL);
+#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
+	struct stack_trace *t;
+#endif
+
+	get_task_struct(task);
+	lock->holders[h] = task;
+
+#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
+	t = lock->traces + h;
+	t->nr_entries = 0;
+	t->max_entries = MAX_STACK;
+	t->entries = lock->entries[h];
+	t->skip = 2;
+	save_stack_trace(t);
+#endif
+}
+
+/* call this *before* you decrement lock->count */
+static void __del_holder(struct block_lock *lock, struct task_struct *task)
+{
+	unsigned h = __find_holder(lock, task);
+	lock->holders[h] = NULL;
+	put_task_struct(task);
+}
+
+static int __check_holder(struct block_lock *lock)
+{
+	unsigned i;
+#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
+	static struct stack_trace t;
+	static stack_entries entries;
+#endif
+
+	for (i = 0; i < MAX_HOLDERS; i++) {
+		if (lock->holders[i] == current) {
+			DMERR("recursive lock detected in metadata");
+#ifdef CONFIG_DM_DEBUG_BLOCK_STACK_TRACING
+			DMERR("previously held here:");
+			print_stack_trace(lock->traces + i, 4);
+
+			DMERR("subsequent acquisition attempted here:");
+			t.nr_entries = 0;
+			t.max_entries = MAX_STACK;
+			t.entries = entries;
+			t.skip = 3;
+			save_stack_trace(&t);
+			print_stack_trace(&t, 4);
+#endif
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+
+static void __wait(struct waiter *w)
+{
+	for (;;) {
+		set_task_state(current, TASK_UNINTERRUPTIBLE);
+
+		if (!w->task)
+			break;
+
+		schedule();
+	}
+
+	set_task_state(current, TASK_RUNNING);
+}
+
+static void __wake_waiter(struct waiter *w)
+{
+	struct task_struct *task;
+
+	list_del(&w->list);
+	task = w->task;
+	smp_mb();
+	w->task = NULL;
+	wake_up_process(task);
+}
+
+/*
+ * We either wake a few readers or a single writer.
+ */
+static void __wake_many(struct block_lock *lock)
+{
+	struct waiter *w, *tmp;
+
+	BUG_ON(lock->count < 0);
+	list_for_each_entry_safe(w, tmp, &lock->waiters, list) {
+		if (lock->count >= MAX_HOLDERS)
+			return;
+
+		if (w->wants_write) {
+			if (lock->count > 0)
+				return; /* still read locked */
+
+			lock->count = -1;
+			__add_holder(lock, w->task);
+			__wake_waiter(w);
+			return;
+		}
+
+		lock->count++;
+		__add_holder(lock, w->task);
+		__wake_waiter(w);
+	}
+}
+
+static void bl_init(struct block_lock *lock)
+{
+	int i;
+
+	spin_lock_init(&lock->lock);
+	lock->count = 0;
+	INIT_LIST_HEAD(&lock->waiters);
+	for (i = 0; i < MAX_HOLDERS; i++)
+		lock->holders[i] = NULL;
+}
+
+static int __available_for_read(struct block_lock *lock)
+{
+	return lock->count >= 0 &&
+		lock->count < MAX_HOLDERS &&
+		list_empty(&lock->waiters);
+}
+
+static int bl_down_read(struct block_lock *lock)
+{
+	int r;
+	struct waiter w;
+
+	spin_lock(&lock->lock);
+	r = __check_holder(lock);
+	if (r) {
+		spin_unlock(&lock->lock);
+		return r;
+	}
+
+	if (__available_for_read(lock)) {
+		lock->count++;
+		__add_holder(lock, current);
+		spin_unlock(&lock->lock);
+		return 0;
+	}
+
+	get_task_struct(current);
+
+	w.task = current;
+	w.wants_write = 0;
+	list_add_tail(&w.list, &lock->waiters);
+	spin_unlock(&lock->lock);
+
+	__wait(&w);
+	put_task_struct(current);
+	return 0;
+}
+
+static int bl_down_read_nonblock(struct block_lock *lock)
+{
+	int r;
+
+	spin_lock(&lock->lock);
+	r = __check_holder(lock);
+	if (r)
+		goto out;
+
+	if (__available_for_read(lock)) {
+		lock->count++;
+		__add_holder(lock, current);
+		r = 0;
+	} else
+		r = -EWOULDBLOCK;
+
+out:
+	spin_unlock(&lock->lock);
+	return r;
+}
+
+static void bl_up_read(struct block_lock *lock)
+{
+	spin_lock(&lock->lock);
+	BUG_ON(lock->count <= 0);
+	__del_holder(lock, current);
+	--lock->count;
+	if (!list_empty(&lock->waiters))
+		__wake_many(lock);
+	spin_unlock(&lock->lock);
+}
+
+static int bl_down_write(struct block_lock *lock)
+{
+	int r;
+	struct waiter w;
+
+	spin_lock(&lock->lock);
+	r = __check_holder(lock);
+	if (r) {
+		spin_unlock(&lock->lock);
+		return r;
+	}
+
+	if (lock->count == 0 && list_empty(&lock->waiters)) {
+		lock->count = -1;
+		__add_holder(lock, current);
+		spin_unlock(&lock->lock);
+		return 0;
+	}
+
+	get_task_struct(current);
+	w.task = current;
+	w.wants_write = 1;
+
+	/*
+	 * Writers given priority. We know there's only one mutator in the
+	 * system, so ignoring the ordering reversal.
+	 */
+	list_add(&w.list, &lock->waiters);
+	spin_unlock(&lock->lock);
+
+	__wait(&w);
+	put_task_struct(current);
+
+	return 0;
+}
+
+static void bl_up_write(struct block_lock *lock)
+{
+	spin_lock(&lock->lock);
+	__del_holder(lock, current);
+	lock->count = 0;
+	if (!list_empty(&lock->waiters))
+		__wake_many(lock);
+	spin_unlock(&lock->lock);
+}
+
+static void report_recursive_bug(dm_block_t b, int r)
+{
+	if (r == -EINVAL)
+		DMERR("recursive acquisition of block %llu requested.",
+		      (unsigned long long) b);
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Block manager is currently implemented using dm-bufio.  struct
+ * dm_block_manager and struct dm_block map directly onto a couple of
+ * structs in the bufio interface.  I want to retain the freedom to move
+ * away from bufio in the future.  So these structs are just cast within
+ * this .c file, rather than making it through to the public interface.
+ */
+static struct dm_buffer *to_buffer(struct dm_block *b)
+{
+	return (struct dm_buffer *) b;
+}
+
+dm_block_t dm_block_location(struct dm_block *b)
+{
+	return dm_bufio_get_block_number(to_buffer(b));
+}
+EXPORT_SYMBOL_GPL(dm_block_location);
+
+void *dm_block_data(struct dm_block *b)
+{
+	return dm_bufio_get_block_data(to_buffer(b));
+}
+EXPORT_SYMBOL_GPL(dm_block_data);
+
+struct buffer_aux {
+	struct dm_block_validator *validator;
+	struct block_lock lock;
+	int write_locked;
+};
+
+static void dm_block_manager_alloc_callback(struct dm_buffer *buf)
+{
+	struct buffer_aux *aux = dm_bufio_get_aux_data(buf);
+	aux->validator = NULL;
+	bl_init(&aux->lock);
+}
+
+static void dm_block_manager_write_callback(struct dm_buffer *buf)
+{
+	struct buffer_aux *aux = dm_bufio_get_aux_data(buf);
+	if (aux->validator) {
+		aux->validator->prepare_for_write(aux->validator, (struct dm_block *) buf,
+			 dm_bufio_get_block_size(dm_bufio_get_client(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)
+{
+	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)
+{
+	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(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(bm->bufio);
+}
+
+static int dm_bm_validate_buffer(struct dm_block_manager *bm,
+				 struct dm_buffer *buf,
+				 struct buffer_aux *aux,
+				 struct dm_block_validator *v)
+{
+	if (unlikely(!aux->validator)) {
+		int r;
+		if (!v)
+			return 0;
+		r = v->check(v, (struct dm_block *) buf, dm_bufio_get_block_size(bm->bufio));
+		if (unlikely(r)) {
+			DMERR_LIMIT("%s validator check failed for block %llu", v->name,
+				    (unsigned long long) dm_bufio_get_block_number(buf));
+			return r;
+		}
+		aux->validator = v;
+	} else {
+		if (unlikely(aux->validator != v)) {
+			DMERR_LIMIT("validator mismatch (old=%s vs new=%s) for block %llu",
+				    aux->validator->name, v ? v->name : "NULL",
+				    (unsigned long long) dm_bufio_get_block_number(buf));
+			return -EINVAL;
+		}
+	}
+
+	return 0;
+}
+int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
+		    struct dm_block_validator *v,
+		    struct dm_block **result)
+{
+	struct buffer_aux *aux;
+	void *p;
+	int r;
+
+	p = dm_bufio_read(bm->bufio, b, (struct dm_buffer **) result);
+	if (unlikely(IS_ERR(p)))
+		return PTR_ERR(p);
+
+	aux = dm_bufio_get_aux_data(to_buffer(*result));
+	r = bl_down_read(&aux->lock);
+	if (unlikely(r)) {
+		dm_bufio_release(to_buffer(*result));
+		report_recursive_bug(b, r);
+		return r;
+	}
+
+	aux->write_locked = 0;
+
+	r = dm_bm_validate_buffer(bm, to_buffer(*result), aux, v);
+	if (unlikely(r)) {
+		bl_up_read(&aux->lock);
+		dm_bufio_release(to_buffer(*result));
+		return r;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bm_read_lock);
+
+int dm_bm_write_lock(struct dm_block_manager *bm,
+		     dm_block_t b, struct dm_block_validator *v,
+		     struct dm_block **result)
+{
+	struct buffer_aux *aux;
+	void *p;
+	int r;
+
+	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);
+
+	aux = dm_bufio_get_aux_data(to_buffer(*result));
+	r = bl_down_write(&aux->lock);
+	if (r) {
+		dm_bufio_release(to_buffer(*result));
+		report_recursive_bug(b, r);
+		return r;
+	}
+
+	aux->write_locked = 1;
+
+	r = dm_bm_validate_buffer(bm, to_buffer(*result), aux, v);
+	if (unlikely(r)) {
+		bl_up_write(&aux->lock);
+		dm_bufio_release(to_buffer(*result));
+		return r;
+	}
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bm_write_lock);
+
+int dm_bm_read_try_lock(struct dm_block_manager *bm,
+			dm_block_t b, struct dm_block_validator *v,
+			struct dm_block **result)
+{
+	struct buffer_aux *aux;
+	void *p;
+	int r;
+
+	p = dm_bufio_get(bm->bufio, b, (struct dm_buffer **) result);
+	if (unlikely(IS_ERR(p)))
+		return PTR_ERR(p);
+	if (unlikely(!p))
+		return -EWOULDBLOCK;
+
+	aux = dm_bufio_get_aux_data(to_buffer(*result));
+	r = bl_down_read_nonblock(&aux->lock);
+	if (r < 0) {
+		dm_bufio_release(to_buffer(*result));
+		report_recursive_bug(b, r);
+		return r;
+	}
+	aux->write_locked = 0;
+
+	r = dm_bm_validate_buffer(bm, to_buffer(*result), aux, v);
+	if (unlikely(r)) {
+		bl_up_read(&aux->lock);
+		dm_bufio_release(to_buffer(*result));
+		return r;
+	}
+
+	return 0;
+}
+
+int dm_bm_write_lock_zero(struct dm_block_manager *bm,
+			  dm_block_t b, struct dm_block_validator *v,
+			  struct dm_block **result)
+{
+	int r;
+	struct buffer_aux *aux;
+	void *p;
+
+	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);
+
+	memset(p, 0, dm_bm_block_size(bm));
+
+	aux = dm_bufio_get_aux_data(to_buffer(*result));
+	r = bl_down_write(&aux->lock);
+	if (r) {
+		dm_bufio_release(to_buffer(*result));
+		return r;
+	}
+
+	aux->write_locked = 1;
+	aux->validator = v;
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bm_write_lock_zero);
+
+int dm_bm_unlock(struct dm_block *b)
+{
+	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(to_buffer(b));
+
+	return 0;
+}
+EXPORT_SYMBOL_GPL(dm_bm_unlock);
+
+int dm_bm_flush(struct dm_block_manager *bm)
+{
+	if (bm->read_only)
+		return -EPERM;
+
+	return dm_bufio_write_dirty_buffers(bm->bufio);
+}
+EXPORT_SYMBOL_GPL(dm_bm_flush);
+
+void dm_bm_prefetch(struct dm_block_manager *bm, dm_block_t b)
+{
+	dm_bufio_prefetch(bm->bufio, b, 1);
+}
+
+void dm_bm_set_read_only(struct dm_block_manager *bm)
+{
+	bm->read_only = true;
+}
+EXPORT_SYMBOL_GPL(dm_bm_set_read_only);
+
+void dm_bm_set_read_write(struct dm_block_manager *bm)
+{
+	bm->read_only = false;
+}
+EXPORT_SYMBOL_GPL(dm_bm_set_read_write);
+
+u32 dm_bm_checksum(const void *data, size_t len, u32 init_xor)
+{
+	return crc32c(~(u32) 0, data, len) ^ init_xor;
+}
+EXPORT_SYMBOL_GPL(dm_bm_checksum);
+
+/*----------------------------------------------------------------*/
+
+MODULE_LICENSE("GPL");
+MODULE_AUTHOR("Joe Thornber <dm-devel@redhat.com>");
+MODULE_DESCRIPTION("Immutable metadata library for dm");
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-block-manager.h b/drivers/md/persistent-data/dm-block-manager.h
new file mode 100644
index 00000000000..1b95dfc1778
--- /dev/null
+++ b/drivers/md/persistent-data/dm-block-manager.h
@@ -0,0 +1,133 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_BLOCK_MANAGER_H
+#define _LINUX_DM_BLOCK_MANAGER_H
+
+#include <linux/types.h>
+#include <linux/blkdev.h>
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Block number.
+ */
+typedef uint64_t dm_block_t;
+struct dm_block;
+
+dm_block_t dm_block_location(struct dm_block *b);
+void *dm_block_data(struct dm_block *b);
+
+/*----------------------------------------------------------------*/
+
+/*
+ * @name should be a unique identifier for the block manager, no longer
+ * than 32 chars.
+ *
+ * @max_held_per_thread should be the maximum number of locks, read or
+ * write, that an individual thread holds at any one time.
+ */
+struct dm_block_manager;
+struct dm_block_manager *dm_block_manager_create(
+	struct block_device *bdev, unsigned block_size,
+	unsigned cache_size, unsigned max_held_per_thread);
+void dm_block_manager_destroy(struct dm_block_manager *bm);
+
+unsigned dm_bm_block_size(struct dm_block_manager *bm);
+dm_block_t dm_bm_nr_blocks(struct dm_block_manager *bm);
+
+/*----------------------------------------------------------------*/
+
+/*
+ * The validator allows the caller to verify newly-read data and modify
+ * the data just before writing, e.g. to calculate checksums.  It's
+ * important to be consistent with your use of validators.  The only time
+ * you can change validators is if you call dm_bm_write_lock_zero.
+ */
+struct dm_block_validator {
+	const char *name;
+	void (*prepare_for_write)(struct dm_block_validator *v, struct dm_block *b, size_t block_size);
+
+	/*
+	 * Return 0 if the checksum is valid or < 0 on error.
+	 */
+	int (*check)(struct dm_block_validator *v, struct dm_block *b, size_t block_size);
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * You can have multiple concurrent readers or a single writer holding a
+ * block lock.
+ */
+
+/*
+ * dm_bm_lock() locks a block and returns through @result a pointer to
+ * memory that holds a copy of that block.  If you have write-locked the
+ * block then any changes you make to memory pointed to by @result will be
+ * written back to the disk sometime after dm_bm_unlock is called.
+ */
+int dm_bm_read_lock(struct dm_block_manager *bm, dm_block_t b,
+		    struct dm_block_validator *v,
+		    struct dm_block **result);
+
+int dm_bm_write_lock(struct dm_block_manager *bm, dm_block_t b,
+		     struct dm_block_validator *v,
+		     struct dm_block **result);
+
+/*
+ * The *_try_lock variants return -EWOULDBLOCK if the block isn't
+ * available immediately.
+ */
+int dm_bm_read_try_lock(struct dm_block_manager *bm, dm_block_t b,
+			struct dm_block_validator *v,
+			struct dm_block **result);
+
+/*
+ * Use dm_bm_write_lock_zero() when you know you're going to
+ * overwrite the block completely.  It saves a disk read.
+ */
+int dm_bm_write_lock_zero(struct dm_block_manager *bm, dm_block_t b,
+			  struct dm_block_validator *v,
+			  struct dm_block **result);
+
+int dm_bm_unlock(struct dm_block *b);
+
+/*
+ * 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
+ * this function.  All dirty blocks are guaranteed to be written and flushed
+ * before the superblock.
+ *
+ * This method always blocks.
+ */
+int dm_bm_flush(struct dm_block_manager *bm);
+
+/*
+ * Request data is prefetched into the cache.
+ */
+void dm_bm_prefetch(struct dm_block_manager *bm, dm_block_t b);
+
+/*
+ * 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);
+void dm_bm_set_read_write(struct dm_block_manager *bm);
+
+u32 dm_bm_checksum(const void *data, size_t len, u32 init_xor);
+
+/*----------------------------------------------------------------*/
+
+#endif	/* _LINUX_DM_BLOCK_MANAGER_H */
diff --git a/drivers/md/persistent-data/dm-btree-internal.h b/drivers/md/persistent-data/dm-btree-internal.h
new file mode 100644
index 00000000000..37d367bb9aa
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree-internal.h
@@ -0,0 +1,135 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_BTREE_INTERNAL_H
+#define DM_BTREE_INTERNAL_H
+
+#include "dm-btree.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * We'll need 2 accessor functions for n->csum and n->blocknr
+ * to support dm-btree-spine.c in that case.
+ */
+
+enum node_flags {
+	INTERNAL_NODE = 1,
+	LEAF_NODE = 1 << 1
+};
+
+/*
+ * Every btree node begins with this structure.  Make sure it's a multiple
+ * of 8-bytes in size, otherwise the 64bit keys will be mis-aligned.
+ */
+struct node_header {
+	__le32 csum;
+	__le32 flags;
+	__le64 blocknr; /* Block this node is supposed to live in. */
+
+	__le32 nr_entries;
+	__le32 max_entries;
+	__le32 value_size;
+	__le32 padding;
+} __packed;
+
+struct btree_node {
+	struct node_header header;
+	__le64 keys[0];
+} __packed;
+
+
+void inc_children(struct dm_transaction_manager *tm, struct btree_node *n,
+		  struct dm_btree_value_type *vt);
+
+int new_block(struct dm_btree_info *info, struct dm_block **result);
+int unlock_block(struct dm_btree_info *info, struct dm_block *b);
+
+/*
+ * Spines keep track of the rolling locks.  There are 2 variants, read-only
+ * and one that uses shadowing.  These are separate structs to allow the
+ * type checker to spot misuse, for example accidentally calling read_lock
+ * on a shadow spine.
+ */
+struct ro_spine {
+	struct dm_btree_info *info;
+
+	int count;
+	struct dm_block *nodes[2];
+};
+
+void init_ro_spine(struct ro_spine *s, struct dm_btree_info *info);
+int exit_ro_spine(struct ro_spine *s);
+int ro_step(struct ro_spine *s, dm_block_t new_child);
+void ro_pop(struct ro_spine *s);
+struct btree_node *ro_node(struct ro_spine *s);
+
+struct shadow_spine {
+	struct dm_btree_info *info;
+
+	int count;
+	struct dm_block *nodes[2];
+
+	dm_block_t root;
+};
+
+void init_shadow_spine(struct shadow_spine *s, struct dm_btree_info *info);
+int exit_shadow_spine(struct shadow_spine *s);
+
+int shadow_step(struct shadow_spine *s, dm_block_t b,
+		struct dm_btree_value_type *vt);
+
+/*
+ * The spine must have at least one entry before calling this.
+ */
+struct dm_block *shadow_current(struct shadow_spine *s);
+
+/*
+ * The spine must have at least two entries before calling this.
+ */
+struct dm_block *shadow_parent(struct shadow_spine *s);
+
+int shadow_has_parent(struct shadow_spine *s);
+
+int shadow_root(struct shadow_spine *s);
+
+/*
+ * Some inlines.
+ */
+static inline __le64 *key_ptr(struct btree_node *n, uint32_t index)
+{
+	return n->keys + index;
+}
+
+static inline void *value_base(struct btree_node *n)
+{
+	return &n->keys[le32_to_cpu(n->header.max_entries)];
+}
+
+static inline void *value_ptr(struct btree_node *n, uint32_t index)
+{
+	uint32_t value_size = le32_to_cpu(n->header.value_size);
+	return value_base(n) + (value_size * index);
+}
+
+/*
+ * Assumes the values are suitably-aligned and converts to core format.
+ */
+static inline uint64_t value64(struct btree_node *n, uint32_t index)
+{
+	__le64 *values_le = value_base(n);
+
+	return le64_to_cpu(values_le[index]);
+}
+
+/*
+ * Searching for a key within a single node.
+ */
+int lower_bound(struct btree_node *n, uint64_t key);
+
+extern struct dm_block_validator btree_node_validator;
+
+#endif	/* DM_BTREE_INTERNAL_H */
diff --git a/drivers/md/persistent-data/dm-btree-remove.c b/drivers/md/persistent-data/dm-btree-remove.c
new file mode 100644
index 00000000000..b88757cd0d1
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree-remove.c
@@ -0,0 +1,592 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-btree.h"
+#include "dm-btree-internal.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/export.h>
+
+/*
+ * Removing an entry from a btree
+ * ==============================
+ *
+ * A very important constraint for our btree is that no node, except the
+ * root, may have fewer than a certain number of entries.
+ * (MIN_ENTRIES <= nr_entries <= MAX_ENTRIES).
+ *
+ * Ensuring this is complicated by the way we want to only ever hold the
+ * locks on 2 nodes concurrently, and only change nodes in a top to bottom
+ * fashion.
+ *
+ * Each node may have a left or right sibling.  When decending the spine,
+ * if a node contains only MIN_ENTRIES then we try and increase this to at
+ * least MIN_ENTRIES + 1.  We do this in the following ways:
+ *
+ * [A] No siblings => this can only happen if the node is the root, in which
+ *     case we copy the childs contents over the root.
+ *
+ * [B] No left sibling
+ *     ==> rebalance(node, right sibling)
+ *
+ * [C] No right sibling
+ *     ==> rebalance(left sibling, node)
+ *
+ * [D] Both siblings, total_entries(left, node, right) <= DEL_THRESHOLD
+ *     ==> delete node adding it's contents to left and right
+ *
+ * [E] Both siblings, total_entries(left, node, right) > DEL_THRESHOLD
+ *     ==> rebalance(left, node, right)
+ *
+ * After these operations it's possible that the our original node no
+ * longer contains the desired sub tree.  For this reason this rebalancing
+ * is performed on the children of the current node.  This also avoids
+ * having a special case for the root.
+ *
+ * Once this rebalancing has occurred we can then step into the child node
+ * for internal nodes.  Or delete the entry for leaf nodes.
+ */
+
+/*
+ * Some little utilities for moving node data around.
+ */
+static void node_shift(struct btree_node *n, int shift)
+{
+	uint32_t nr_entries = le32_to_cpu(n->header.nr_entries);
+	uint32_t value_size = le32_to_cpu(n->header.value_size);
+
+	if (shift < 0) {
+		shift = -shift;
+		BUG_ON(shift > nr_entries);
+		BUG_ON((void *) key_ptr(n, shift) >= value_ptr(n, shift));
+		memmove(key_ptr(n, 0),
+			key_ptr(n, shift),
+			(nr_entries - shift) * sizeof(__le64));
+		memmove(value_ptr(n, 0),
+			value_ptr(n, shift),
+			(nr_entries - shift) * value_size);
+	} else {
+		BUG_ON(nr_entries + shift > le32_to_cpu(n->header.max_entries));
+		memmove(key_ptr(n, shift),
+			key_ptr(n, 0),
+			nr_entries * sizeof(__le64));
+		memmove(value_ptr(n, shift),
+			value_ptr(n, 0),
+			nr_entries * value_size);
+	}
+}
+
+static void node_copy(struct btree_node *left, struct btree_node *right, int shift)
+{
+	uint32_t nr_left = le32_to_cpu(left->header.nr_entries);
+	uint32_t value_size = le32_to_cpu(left->header.value_size);
+	BUG_ON(value_size != le32_to_cpu(right->header.value_size));
+
+	if (shift < 0) {
+		shift = -shift;
+		BUG_ON(nr_left + shift > le32_to_cpu(left->header.max_entries));
+		memcpy(key_ptr(left, nr_left),
+		       key_ptr(right, 0),
+		       shift * sizeof(__le64));
+		memcpy(value_ptr(left, nr_left),
+		       value_ptr(right, 0),
+		       shift * value_size);
+	} else {
+		BUG_ON(shift > le32_to_cpu(right->header.max_entries));
+		memcpy(key_ptr(right, 0),
+		       key_ptr(left, nr_left - shift),
+		       shift * sizeof(__le64));
+		memcpy(value_ptr(right, 0),
+		       value_ptr(left, nr_left - shift),
+		       shift * value_size);
+	}
+}
+
+/*
+ * Delete a specific entry from a leaf node.
+ */
+static void delete_at(struct btree_node *n, unsigned index)
+{
+	unsigned nr_entries = le32_to_cpu(n->header.nr_entries);
+	unsigned nr_to_copy = nr_entries - (index + 1);
+	uint32_t value_size = le32_to_cpu(n->header.value_size);
+	BUG_ON(index >= nr_entries);
+
+	if (nr_to_copy) {
+		memmove(key_ptr(n, index),
+			key_ptr(n, index + 1),
+			nr_to_copy * sizeof(__le64));
+
+		memmove(value_ptr(n, index),
+			value_ptr(n, index + 1),
+			nr_to_copy * value_size);
+	}
+
+	n->header.nr_entries = cpu_to_le32(nr_entries - 1);
+}
+
+static unsigned merge_threshold(struct btree_node *n)
+{
+	return le32_to_cpu(n->header.max_entries) / 3;
+}
+
+struct child {
+	unsigned index;
+	struct dm_block *block;
+	struct btree_node *n;
+};
+
+static int init_child(struct dm_btree_info *info, struct dm_btree_value_type *vt,
+		      struct btree_node *parent,
+		      unsigned index, struct child *result)
+{
+	int r, inc;
+	dm_block_t root;
+
+	result->index = index;
+	root = value64(parent, index);
+
+	r = dm_tm_shadow_block(info->tm, root, &btree_node_validator,
+			       &result->block, &inc);
+	if (r)
+		return r;
+
+	result->n = dm_block_data(result->block);
+
+	if (inc)
+		inc_children(info->tm, result->n, vt);
+
+	*((__le64 *) value_ptr(parent, index)) =
+		cpu_to_le64(dm_block_location(result->block));
+
+	return 0;
+}
+
+static int exit_child(struct dm_btree_info *info, struct child *c)
+{
+	return dm_tm_unlock(info->tm, c->block);
+}
+
+static void shift(struct btree_node *left, struct btree_node *right, int count)
+{
+	uint32_t nr_left = le32_to_cpu(left->header.nr_entries);
+	uint32_t nr_right = le32_to_cpu(right->header.nr_entries);
+	uint32_t max_entries = le32_to_cpu(left->header.max_entries);
+	uint32_t r_max_entries = le32_to_cpu(right->header.max_entries);
+
+	BUG_ON(max_entries != r_max_entries);
+	BUG_ON(nr_left - count > max_entries);
+	BUG_ON(nr_right + count > max_entries);
+
+	if (!count)
+		return;
+
+	if (count > 0) {
+		node_shift(right, count);
+		node_copy(left, right, count);
+	} else {
+		node_copy(left, right, count);
+		node_shift(right, count);
+	}
+
+	left->header.nr_entries = cpu_to_le32(nr_left - count);
+	right->header.nr_entries = cpu_to_le32(nr_right + count);
+}
+
+static void __rebalance2(struct dm_btree_info *info, struct btree_node *parent,
+			 struct child *l, struct child *r)
+{
+	struct btree_node *left = l->n;
+	struct btree_node *right = r->n;
+	uint32_t nr_left = le32_to_cpu(left->header.nr_entries);
+	uint32_t nr_right = le32_to_cpu(right->header.nr_entries);
+	unsigned threshold = 2 * merge_threshold(left) + 1;
+
+	if (nr_left + nr_right < threshold) {
+		/*
+		 * Merge
+		 */
+		node_copy(left, right, -nr_right);
+		left->header.nr_entries = cpu_to_le32(nr_left + nr_right);
+		delete_at(parent, r->index);
+
+		/*
+		 * We need to decrement the right block, but not it's
+		 * children, since they're still referenced by left.
+		 */
+		dm_tm_dec(info->tm, dm_block_location(r->block));
+	} else {
+		/*
+		 * Rebalance.
+		 */
+		unsigned target_left = (nr_left + nr_right) / 2;
+		shift(left, right, nr_left - target_left);
+		*key_ptr(parent, r->index) = right->keys[0];
+	}
+}
+
+static int rebalance2(struct shadow_spine *s, struct dm_btree_info *info,
+		      struct dm_btree_value_type *vt, unsigned left_index)
+{
+	int r;
+	struct btree_node *parent;
+	struct child left, right;
+
+	parent = dm_block_data(shadow_current(s));
+
+	r = init_child(info, vt, parent, left_index, &left);
+	if (r)
+		return r;
+
+	r = init_child(info, vt, parent, left_index + 1, &right);
+	if (r) {
+		exit_child(info, &left);
+		return r;
+	}
+
+	__rebalance2(info, parent, &left, &right);
+
+	r = exit_child(info, &left);
+	if (r) {
+		exit_child(info, &right);
+		return r;
+	}
+
+	return exit_child(info, &right);
+}
+
+/*
+ * We dump as many entries from center as possible into left, then the rest
+ * in right, then rebalance2.  This wastes some cpu, but I want something
+ * simple atm.
+ */
+static void delete_center_node(struct dm_btree_info *info, struct btree_node *parent,
+			       struct child *l, struct child *c, struct child *r,
+			       struct btree_node *left, struct btree_node *center, struct btree_node *right,
+			       uint32_t nr_left, uint32_t nr_center, uint32_t nr_right)
+{
+	uint32_t max_entries = le32_to_cpu(left->header.max_entries);
+	unsigned shift = min(max_entries - nr_left, nr_center);
+
+	BUG_ON(nr_left + shift > max_entries);
+	node_copy(left, center, -shift);
+	left->header.nr_entries = cpu_to_le32(nr_left + shift);
+
+	if (shift != nr_center) {
+		shift = nr_center - shift;
+		BUG_ON((nr_right + shift) > max_entries);
+		node_shift(right, shift);
+		node_copy(center, right, shift);
+		right->header.nr_entries = cpu_to_le32(nr_right + shift);
+	}
+	*key_ptr(parent, r->index) = right->keys[0];
+
+	delete_at(parent, c->index);
+	r->index--;
+
+	dm_tm_dec(info->tm, dm_block_location(c->block));
+	__rebalance2(info, parent, l, r);
+}
+
+/*
+ * Redistributes entries among 3 sibling nodes.
+ */
+static void redistribute3(struct dm_btree_info *info, struct btree_node *parent,
+			  struct child *l, struct child *c, struct child *r,
+			  struct btree_node *left, struct btree_node *center, struct btree_node *right,
+			  uint32_t nr_left, uint32_t nr_center, uint32_t nr_right)
+{
+	int s;
+	uint32_t max_entries = le32_to_cpu(left->header.max_entries);
+	unsigned target = (nr_left + nr_center + nr_right) / 3;
+	BUG_ON(target > max_entries);
+
+	if (nr_left < nr_right) {
+		s = nr_left - target;
+
+		if (s < 0 && nr_center < -s) {
+			/* not enough in central node */
+			shift(left, center, nr_center);
+			s = nr_center - target;
+			shift(left, right, s);
+			nr_right += s;
+		} else
+			shift(left, center, s);
+
+		shift(center, right, target - nr_right);
+
+	} else {
+		s = target - nr_right;
+		if (s > 0 && nr_center < s) {
+			/* not enough in central node */
+			shift(center, right, nr_center);
+			s = target - nr_center;
+			shift(left, right, s);
+			nr_left -= s;
+		} else
+			shift(center, right, s);
+
+		shift(left, center, nr_left - target);
+	}
+
+	*key_ptr(parent, c->index) = center->keys[0];
+	*key_ptr(parent, r->index) = right->keys[0];
+}
+
+static void __rebalance3(struct dm_btree_info *info, struct btree_node *parent,
+			 struct child *l, struct child *c, struct child *r)
+{
+	struct btree_node *left = l->n;
+	struct btree_node *center = c->n;
+	struct btree_node *right = r->n;
+
+	uint32_t nr_left = le32_to_cpu(left->header.nr_entries);
+	uint32_t nr_center = le32_to_cpu(center->header.nr_entries);
+	uint32_t nr_right = le32_to_cpu(right->header.nr_entries);
+
+	unsigned threshold = merge_threshold(left) * 4 + 1;
+
+	BUG_ON(left->header.max_entries != center->header.max_entries);
+	BUG_ON(center->header.max_entries != right->header.max_entries);
+
+	if ((nr_left + nr_center + nr_right) < threshold)
+		delete_center_node(info, parent, l, c, r, left, center, right,
+				   nr_left, nr_center, nr_right);
+	else
+		redistribute3(info, parent, l, c, r, left, center, right,
+			      nr_left, nr_center, nr_right);
+}
+
+static int rebalance3(struct shadow_spine *s, struct dm_btree_info *info,
+		      struct dm_btree_value_type *vt, unsigned left_index)
+{
+	int r;
+	struct btree_node *parent = dm_block_data(shadow_current(s));
+	struct child left, center, right;
+
+	/*
+	 * FIXME: fill out an array?
+	 */
+	r = init_child(info, vt, parent, left_index, &left);
+	if (r)
+		return r;
+
+	r = init_child(info, vt, parent, left_index + 1, &center);
+	if (r) {
+		exit_child(info, &left);
+		return r;
+	}
+
+	r = init_child(info, vt, parent, left_index + 2, &right);
+	if (r) {
+		exit_child(info, &left);
+		exit_child(info, &center);
+		return r;
+	}
+
+	__rebalance3(info, parent, &left, &center, &right);
+
+	r = exit_child(info, &left);
+	if (r) {
+		exit_child(info, &center);
+		exit_child(info, &right);
+		return r;
+	}
+
+	r = exit_child(info, &center);
+	if (r) {
+		exit_child(info, &right);
+		return r;
+	}
+
+	r = exit_child(info, &right);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int get_nr_entries(struct dm_transaction_manager *tm,
+			  dm_block_t b, uint32_t *result)
+{
+	int r;
+	struct dm_block *block;
+	struct btree_node *n;
+
+	r = dm_tm_read_lock(tm, b, &btree_node_validator, &block);
+	if (r)
+		return r;
+
+	n = dm_block_data(block);
+	*result = le32_to_cpu(n->header.nr_entries);
+
+	return dm_tm_unlock(tm, block);
+}
+
+static int rebalance_children(struct shadow_spine *s,
+			      struct dm_btree_info *info,
+			      struct dm_btree_value_type *vt, uint64_t key)
+{
+	int i, r, has_left_sibling, has_right_sibling;
+	uint32_t child_entries;
+	struct btree_node *n;
+
+	n = dm_block_data(shadow_current(s));
+
+	if (le32_to_cpu(n->header.nr_entries) == 1) {
+		struct dm_block *child;
+		dm_block_t b = value64(n, 0);
+
+		r = dm_tm_read_lock(info->tm, b, &btree_node_validator, &child);
+		if (r)
+			return r;
+
+		memcpy(n, dm_block_data(child),
+		       dm_bm_block_size(dm_tm_get_bm(info->tm)));
+		r = dm_tm_unlock(info->tm, child);
+		if (r)
+			return r;
+
+		dm_tm_dec(info->tm, dm_block_location(child));
+		return 0;
+	}
+
+	i = lower_bound(n, key);
+	if (i < 0)
+		return -ENODATA;
+
+	r = get_nr_entries(info->tm, value64(n, i), &child_entries);
+	if (r)
+		return r;
+
+	has_left_sibling = i > 0;
+	has_right_sibling = i < (le32_to_cpu(n->header.nr_entries) - 1);
+
+	if (!has_left_sibling)
+		r = rebalance2(s, info, vt, i);
+
+	else if (!has_right_sibling)
+		r = rebalance2(s, info, vt, i - 1);
+
+	else
+		r = rebalance3(s, info, vt, i - 1);
+
+	return r;
+}
+
+static int do_leaf(struct btree_node *n, uint64_t key, unsigned *index)
+{
+	int i = lower_bound(n, key);
+
+	if ((i < 0) ||
+	    (i >= le32_to_cpu(n->header.nr_entries)) ||
+	    (le64_to_cpu(n->keys[i]) != key))
+		return -ENODATA;
+
+	*index = i;
+
+	return 0;
+}
+
+/*
+ * Prepares for removal from one level of the hierarchy.  The caller must
+ * call delete_at() to remove the entry at index.
+ */
+static int remove_raw(struct shadow_spine *s, struct dm_btree_info *info,
+		      struct dm_btree_value_type *vt, dm_block_t root,
+		      uint64_t key, unsigned *index)
+{
+	int i = *index, r;
+	struct btree_node *n;
+
+	for (;;) {
+		r = shadow_step(s, root, vt);
+		if (r < 0)
+			break;
+
+		/*
+		 * We have to patch up the parent node, ugly, but I don't
+		 * see a way to do this automatically as part of the spine
+		 * op.
+		 */
+		if (shadow_has_parent(s)) {
+			__le64 location = cpu_to_le64(dm_block_location(shadow_current(s)));
+			memcpy(value_ptr(dm_block_data(shadow_parent(s)), i),
+			       &location, sizeof(__le64));
+		}
+
+		n = dm_block_data(shadow_current(s));
+
+		if (le32_to_cpu(n->header.flags) & LEAF_NODE)
+			return do_leaf(n, key, index);
+
+		r = rebalance_children(s, info, vt, key);
+		if (r)
+			break;
+
+		n = dm_block_data(shadow_current(s));
+		if (le32_to_cpu(n->header.flags) & LEAF_NODE)
+			return do_leaf(n, key, index);
+
+		i = lower_bound(n, key);
+
+		/*
+		 * We know the key is present, or else
+		 * rebalance_children would have returned
+		 * -ENODATA
+		 */
+		root = value64(n, i);
+	}
+
+	return r;
+}
+
+static struct dm_btree_value_type le64_type = {
+	.context = NULL,
+	.size = sizeof(__le64),
+	.inc = NULL,
+	.dec = NULL,
+	.equal = NULL
+};
+
+int dm_btree_remove(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, dm_block_t *new_root)
+{
+	unsigned level, last_level = info->levels - 1;
+	int index = 0, r = 0;
+	struct shadow_spine spine;
+	struct btree_node *n;
+
+	init_shadow_spine(&spine, info);
+	for (level = 0; level < info->levels; level++) {
+		r = remove_raw(&spine, info,
+			       (level == last_level ?
+				&info->value_type : &le64_type),
+			       root, keys[level], (unsigned *)&index);
+		if (r < 0)
+			break;
+
+		n = dm_block_data(shadow_current(&spine));
+		if (level != last_level) {
+			root = value64(n, index);
+			continue;
+		}
+
+		BUG_ON(index < 0 || index >= le32_to_cpu(n->header.nr_entries));
+
+		if (info->value_type.dec)
+			info->value_type.dec(info->value_type.context,
+					     value_ptr(n, index));
+
+		delete_at(n, index);
+	}
+
+	*new_root = shadow_root(&spine);
+	exit_shadow_spine(&spine);
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_btree_remove);
diff --git a/drivers/md/persistent-data/dm-btree-spine.c b/drivers/md/persistent-data/dm-btree-spine.c
new file mode 100644
index 00000000000..cf9fd676ae4
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree-spine.c
@@ -0,0 +1,251 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-btree-internal.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "btree spine"
+
+/*----------------------------------------------------------------*/
+
+#define BTREE_CSUM_XOR 121107
+
+static int node_check(struct dm_block_validator *v,
+		      struct dm_block *b,
+		      size_t block_size);
+
+static void node_prepare_for_write(struct dm_block_validator *v,
+				   struct dm_block *b,
+				   size_t block_size)
+{
+	struct btree_node *n = dm_block_data(b);
+	struct node_header *h = &n->header;
+
+	h->blocknr = cpu_to_le64(dm_block_location(b));
+	h->csum = cpu_to_le32(dm_bm_checksum(&h->flags,
+					     block_size - sizeof(__le32),
+					     BTREE_CSUM_XOR));
+
+	BUG_ON(node_check(v, b, 4096));
+}
+
+static int node_check(struct dm_block_validator *v,
+		      struct dm_block *b,
+		      size_t block_size)
+{
+	struct btree_node *n = dm_block_data(b);
+	struct node_header *h = &n->header;
+	size_t value_size;
+	__le32 csum_disk;
+	uint32_t flags;
+
+	if (dm_block_location(b) != le64_to_cpu(h->blocknr)) {
+		DMERR_LIMIT("node_check failed: blocknr %llu != wanted %llu",
+			    le64_to_cpu(h->blocknr), dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	csum_disk = cpu_to_le32(dm_bm_checksum(&h->flags,
+					       block_size - sizeof(__le32),
+					       BTREE_CSUM_XOR));
+	if (csum_disk != h->csum) {
+		DMERR_LIMIT("node_check failed: csum %u != wanted %u",
+			    le32_to_cpu(csum_disk), le32_to_cpu(h->csum));
+		return -EILSEQ;
+	}
+
+	value_size = le32_to_cpu(h->value_size);
+
+	if (sizeof(struct node_header) +
+	    (sizeof(__le64) + value_size) * le32_to_cpu(h->max_entries) > block_size) {
+		DMERR_LIMIT("node_check failed: max_entries too large");
+		return -EILSEQ;
+	}
+
+	if (le32_to_cpu(h->nr_entries) > le32_to_cpu(h->max_entries)) {
+		DMERR_LIMIT("node_check failed: too many entries");
+		return -EILSEQ;
+	}
+
+	/*
+	 * The node must be either INTERNAL or LEAF.
+	 */
+	flags = le32_to_cpu(h->flags);
+	if (!(flags & INTERNAL_NODE) && !(flags & LEAF_NODE)) {
+		DMERR_LIMIT("node_check failed: node is neither INTERNAL or LEAF");
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+struct dm_block_validator btree_node_validator = {
+	.name = "btree_node",
+	.prepare_for_write = node_prepare_for_write,
+	.check = node_check
+};
+
+/*----------------------------------------------------------------*/
+
+static int bn_read_lock(struct dm_btree_info *info, dm_block_t b,
+		 struct dm_block **result)
+{
+	return dm_tm_read_lock(info->tm, b, &btree_node_validator, result);
+}
+
+static int bn_shadow(struct dm_btree_info *info, dm_block_t orig,
+	      struct dm_btree_value_type *vt,
+	      struct dm_block **result)
+{
+	int r, inc;
+
+	r = dm_tm_shadow_block(info->tm, orig, &btree_node_validator,
+			       result, &inc);
+	if (!r && inc)
+		inc_children(info->tm, dm_block_data(*result), vt);
+
+	return r;
+}
+
+int new_block(struct dm_btree_info *info, struct dm_block **result)
+{
+	return dm_tm_new_block(info->tm, &btree_node_validator, result);
+}
+
+int unlock_block(struct dm_btree_info *info, struct dm_block *b)
+{
+	return dm_tm_unlock(info->tm, b);
+}
+
+/*----------------------------------------------------------------*/
+
+void init_ro_spine(struct ro_spine *s, struct dm_btree_info *info)
+{
+	s->info = info;
+	s->count = 0;
+	s->nodes[0] = NULL;
+	s->nodes[1] = NULL;
+}
+
+int exit_ro_spine(struct ro_spine *s)
+{
+	int r = 0, i;
+
+	for (i = 0; i < s->count; i++) {
+		int r2 = unlock_block(s->info, s->nodes[i]);
+		if (r2 < 0)
+			r = r2;
+	}
+
+	return r;
+}
+
+int ro_step(struct ro_spine *s, dm_block_t new_child)
+{
+	int r;
+
+	if (s->count == 2) {
+		r = unlock_block(s->info, s->nodes[0]);
+		if (r < 0)
+			return r;
+		s->nodes[0] = s->nodes[1];
+		s->count--;
+	}
+
+	r = bn_read_lock(s->info, new_child, s->nodes + s->count);
+	if (!r)
+		s->count++;
+
+	return r;
+}
+
+void ro_pop(struct ro_spine *s)
+{
+	BUG_ON(!s->count);
+	--s->count;
+	unlock_block(s->info, s->nodes[s->count]);
+}
+
+struct btree_node *ro_node(struct ro_spine *s)
+{
+	struct dm_block *block;
+
+	BUG_ON(!s->count);
+	block = s->nodes[s->count - 1];
+
+	return dm_block_data(block);
+}
+
+/*----------------------------------------------------------------*/
+
+void init_shadow_spine(struct shadow_spine *s, struct dm_btree_info *info)
+{
+	s->info = info;
+	s->count = 0;
+}
+
+int exit_shadow_spine(struct shadow_spine *s)
+{
+	int r = 0, i;
+
+	for (i = 0; i < s->count; i++) {
+		int r2 = unlock_block(s->info, s->nodes[i]);
+		if (r2 < 0)
+			r = r2;
+	}
+
+	return r;
+}
+
+int shadow_step(struct shadow_spine *s, dm_block_t b,
+		struct dm_btree_value_type *vt)
+{
+	int r;
+
+	if (s->count == 2) {
+		r = unlock_block(s->info, s->nodes[0]);
+		if (r < 0)
+			return r;
+		s->nodes[0] = s->nodes[1];
+		s->count--;
+	}
+
+	r = bn_shadow(s->info, b, vt, s->nodes + s->count);
+	if (!r) {
+		if (!s->count)
+			s->root = dm_block_location(s->nodes[0]);
+
+		s->count++;
+	}
+
+	return r;
+}
+
+struct dm_block *shadow_current(struct shadow_spine *s)
+{
+	BUG_ON(!s->count);
+
+	return s->nodes[s->count - 1];
+}
+
+struct dm_block *shadow_parent(struct shadow_spine *s)
+{
+	BUG_ON(s->count != 2);
+
+	return s->count == 2 ? s->nodes[0] : NULL;
+}
+
+int shadow_has_parent(struct shadow_spine *s)
+{
+	return s->count >= 2;
+}
+
+int shadow_root(struct shadow_spine *s)
+{
+	return s->root;
+}
diff --git a/drivers/md/persistent-data/dm-btree.c b/drivers/md/persistent-data/dm-btree.c
new file mode 100644
index 00000000000..416060c2570
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree.c
@@ -0,0 +1,896 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-btree-internal.h"
+#include "dm-space-map.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/export.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "btree"
+
+/*----------------------------------------------------------------
+ * Array manipulation
+ *--------------------------------------------------------------*/
+static void memcpy_disk(void *dest, const void *src, size_t len)
+	__dm_written_to_disk(src)
+{
+	memcpy(dest, src, len);
+	__dm_unbless_for_disk(src);
+}
+
+static void array_insert(void *base, size_t elt_size, unsigned nr_elts,
+			 unsigned index, void *elt)
+	__dm_written_to_disk(elt)
+{
+	if (index < nr_elts)
+		memmove(base + (elt_size * (index + 1)),
+			base + (elt_size * index),
+			(nr_elts - index) * elt_size);
+
+	memcpy_disk(base + (elt_size * index), elt, elt_size);
+}
+
+/*----------------------------------------------------------------*/
+
+/* makes the assumption that no two keys are the same. */
+static int bsearch(struct btree_node *n, uint64_t key, int want_hi)
+{
+	int lo = -1, hi = le32_to_cpu(n->header.nr_entries);
+
+	while (hi - lo > 1) {
+		int mid = lo + ((hi - lo) / 2);
+		uint64_t mid_key = le64_to_cpu(n->keys[mid]);
+
+		if (mid_key == key)
+			return mid;
+
+		if (mid_key < key)
+			lo = mid;
+		else
+			hi = mid;
+	}
+
+	return want_hi ? hi : lo;
+}
+
+int lower_bound(struct btree_node *n, uint64_t key)
+{
+	return bsearch(n, key, 0);
+}
+
+void inc_children(struct dm_transaction_manager *tm, struct btree_node *n,
+		  struct dm_btree_value_type *vt)
+{
+	unsigned i;
+	uint32_t nr_entries = le32_to_cpu(n->header.nr_entries);
+
+	if (le32_to_cpu(n->header.flags) & INTERNAL_NODE)
+		for (i = 0; i < nr_entries; i++)
+			dm_tm_inc(tm, value64(n, i));
+	else if (vt->inc)
+		for (i = 0; i < nr_entries; i++)
+			vt->inc(vt->context, value_ptr(n, i));
+}
+
+static int insert_at(size_t value_size, struct btree_node *node, unsigned index,
+		      uint64_t key, void *value)
+		      __dm_written_to_disk(value)
+{
+	uint32_t nr_entries = le32_to_cpu(node->header.nr_entries);
+	__le64 key_le = cpu_to_le64(key);
+
+	if (index > nr_entries ||
+	    index >= le32_to_cpu(node->header.max_entries)) {
+		DMERR("too many entries in btree node for insert");
+		__dm_unbless_for_disk(value);
+		return -ENOMEM;
+	}
+
+	__dm_bless_for_disk(&key_le);
+
+	array_insert(node->keys, sizeof(*node->keys), nr_entries, index, &key_le);
+	array_insert(value_base(node), value_size, nr_entries, index, value);
+	node->header.nr_entries = cpu_to_le32(nr_entries + 1);
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * We want 3n entries (for some n).  This works more nicely for repeated
+ * insert remove loops than (2n + 1).
+ */
+static uint32_t calc_max_entries(size_t value_size, size_t block_size)
+{
+	uint32_t total, n;
+	size_t elt_size = sizeof(uint64_t) + value_size; /* key + value */
+
+	block_size -= sizeof(struct node_header);
+	total = block_size / elt_size;
+	n = total / 3;		/* rounds down */
+
+	return 3 * n;
+}
+
+int dm_btree_empty(struct dm_btree_info *info, dm_block_t *root)
+{
+	int r;
+	struct dm_block *b;
+	struct btree_node *n;
+	size_t block_size;
+	uint32_t max_entries;
+
+	r = new_block(info, &b);
+	if (r < 0)
+		return r;
+
+	block_size = dm_bm_block_size(dm_tm_get_bm(info->tm));
+	max_entries = calc_max_entries(info->value_type.size, block_size);
+
+	n = dm_block_data(b);
+	memset(n, 0, block_size);
+	n->header.flags = cpu_to_le32(LEAF_NODE);
+	n->header.nr_entries = cpu_to_le32(0);
+	n->header.max_entries = cpu_to_le32(max_entries);
+	n->header.value_size = cpu_to_le32(info->value_type.size);
+
+	*root = dm_block_location(b);
+	return unlock_block(info, b);
+}
+EXPORT_SYMBOL_GPL(dm_btree_empty);
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Deletion uses a recursive algorithm, since we have limited stack space
+ * we explicitly manage our own stack on the heap.
+ */
+#define MAX_SPINE_DEPTH 64
+struct frame {
+	struct dm_block *b;
+	struct btree_node *n;
+	unsigned level;
+	unsigned nr_children;
+	unsigned current_child;
+};
+
+struct del_stack {
+	struct dm_btree_info *info;
+	struct dm_transaction_manager *tm;
+	int top;
+	struct frame spine[MAX_SPINE_DEPTH];
+};
+
+static int top_frame(struct del_stack *s, struct frame **f)
+{
+	if (s->top < 0) {
+		DMERR("btree deletion stack empty");
+		return -EINVAL;
+	}
+
+	*f = s->spine + s->top;
+
+	return 0;
+}
+
+static int unprocessed_frames(struct del_stack *s)
+{
+	return s->top >= 0;
+}
+
+static void prefetch_children(struct del_stack *s, struct frame *f)
+{
+	unsigned i;
+	struct dm_block_manager *bm = dm_tm_get_bm(s->tm);
+
+	for (i = 0; i < f->nr_children; i++)
+		dm_bm_prefetch(bm, value64(f->n, i));
+}
+
+static bool is_internal_level(struct dm_btree_info *info, struct frame *f)
+{
+	return f->level < (info->levels - 1);
+}
+
+static int push_frame(struct del_stack *s, dm_block_t b, unsigned level)
+{
+	int r;
+	uint32_t ref_count;
+
+	if (s->top >= MAX_SPINE_DEPTH - 1) {
+		DMERR("btree deletion stack out of memory");
+		return -ENOMEM;
+	}
+
+	r = dm_tm_ref(s->tm, b, &ref_count);
+	if (r)
+		return r;
+
+	if (ref_count > 1)
+		/*
+		 * This is a shared node, so we can just decrement it's
+		 * reference counter and leave the children.
+		 */
+		dm_tm_dec(s->tm, b);
+
+	else {
+		uint32_t flags;
+		struct frame *f = s->spine + ++s->top;
+
+		r = dm_tm_read_lock(s->tm, b, &btree_node_validator, &f->b);
+		if (r) {
+			s->top--;
+			return r;
+		}
+
+		f->n = dm_block_data(f->b);
+		f->level = level;
+		f->nr_children = le32_to_cpu(f->n->header.nr_entries);
+		f->current_child = 0;
+
+		flags = le32_to_cpu(f->n->header.flags);
+		if (flags & INTERNAL_NODE || is_internal_level(s->info, f))
+			prefetch_children(s, f);
+	}
+
+	return 0;
+}
+
+static void pop_frame(struct del_stack *s)
+{
+	struct frame *f = s->spine + s->top--;
+
+	dm_tm_dec(s->tm, dm_block_location(f->b));
+	dm_tm_unlock(s->tm, f->b);
+}
+
+int dm_btree_del(struct dm_btree_info *info, dm_block_t root)
+{
+	int r;
+	struct del_stack *s;
+
+	s = kmalloc(sizeof(*s), GFP_KERNEL);
+	if (!s)
+		return -ENOMEM;
+	s->info = info;
+	s->tm = info->tm;
+	s->top = -1;
+
+	r = push_frame(s, root, 0);
+	if (r)
+		goto out;
+
+	while (unprocessed_frames(s)) {
+		uint32_t flags;
+		struct frame *f;
+		dm_block_t b;
+
+		r = top_frame(s, &f);
+		if (r)
+			goto out;
+
+		if (f->current_child >= f->nr_children) {
+			pop_frame(s);
+			continue;
+		}
+
+		flags = le32_to_cpu(f->n->header.flags);
+		if (flags & INTERNAL_NODE) {
+			b = value64(f->n, f->current_child);
+			f->current_child++;
+			r = push_frame(s, b, f->level);
+			if (r)
+				goto out;
+
+		} else if (is_internal_level(info, f)) {
+			b = value64(f->n, f->current_child);
+			f->current_child++;
+			r = push_frame(s, b, f->level + 1);
+			if (r)
+				goto out;
+
+		} else {
+			if (info->value_type.dec) {
+				unsigned i;
+
+				for (i = 0; i < f->nr_children; i++)
+					info->value_type.dec(info->value_type.context,
+							     value_ptr(f->n, i));
+			}
+			pop_frame(s);
+		}
+	}
+
+out:
+	kfree(s);
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_btree_del);
+
+/*----------------------------------------------------------------*/
+
+static int btree_lookup_raw(struct ro_spine *s, dm_block_t block, uint64_t key,
+			    int (*search_fn)(struct btree_node *, uint64_t),
+			    uint64_t *result_key, void *v, size_t value_size)
+{
+	int i, r;
+	uint32_t flags, nr_entries;
+
+	do {
+		r = ro_step(s, block);
+		if (r < 0)
+			return r;
+
+		i = search_fn(ro_node(s), key);
+
+		flags = le32_to_cpu(ro_node(s)->header.flags);
+		nr_entries = le32_to_cpu(ro_node(s)->header.nr_entries);
+		if (i < 0 || i >= nr_entries)
+			return -ENODATA;
+
+		if (flags & INTERNAL_NODE)
+			block = value64(ro_node(s), i);
+
+	} while (!(flags & LEAF_NODE));
+
+	*result_key = le64_to_cpu(ro_node(s)->keys[i]);
+	memcpy(v, value_ptr(ro_node(s), i), value_size);
+
+	return 0;
+}
+
+int dm_btree_lookup(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, void *value_le)
+{
+	unsigned level, last_level = info->levels - 1;
+	int r = -ENODATA;
+	uint64_t rkey;
+	__le64 internal_value_le;
+	struct ro_spine spine;
+
+	init_ro_spine(&spine, info);
+	for (level = 0; level < info->levels; level++) {
+		size_t size;
+		void *value_p;
+
+		if (level == last_level) {
+			value_p = value_le;
+			size = info->value_type.size;
+
+		} else {
+			value_p = &internal_value_le;
+			size = sizeof(uint64_t);
+		}
+
+		r = btree_lookup_raw(&spine, root, keys[level],
+				     lower_bound, &rkey,
+				     value_p, size);
+
+		if (!r) {
+			if (rkey != keys[level]) {
+				exit_ro_spine(&spine);
+				return -ENODATA;
+			}
+		} else {
+			exit_ro_spine(&spine);
+			return r;
+		}
+
+		root = le64_to_cpu(internal_value_le);
+	}
+	exit_ro_spine(&spine);
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_btree_lookup);
+
+/*
+ * Splits a node by creating a sibling node and shifting half the nodes
+ * contents across.  Assumes there is a parent node, and it has room for
+ * another child.
+ *
+ * Before:
+ *	  +--------+
+ *	  | Parent |
+ *	  +--------+
+ *	     |
+ *	     v
+ *	+----------+
+ *	| A ++++++ |
+ *	+----------+
+ *
+ *
+ * After:
+ *		+--------+
+ *		| Parent |
+ *		+--------+
+ *		  |	|
+ *		  v	+------+
+ *	    +---------+	       |
+ *	    | A* +++  |	       v
+ *	    +---------+	  +-------+
+ *			  | B +++ |
+ *			  +-------+
+ *
+ * Where A* is a shadow of A.
+ */
+static int btree_split_sibling(struct shadow_spine *s, dm_block_t root,
+			       unsigned parent_index, uint64_t key)
+{
+	int r;
+	size_t size;
+	unsigned nr_left, nr_right;
+	struct dm_block *left, *right, *parent;
+	struct btree_node *ln, *rn, *pn;
+	__le64 location;
+
+	left = shadow_current(s);
+
+	r = new_block(s->info, &right);
+	if (r < 0)
+		return r;
+
+	ln = dm_block_data(left);
+	rn = dm_block_data(right);
+
+	nr_left = le32_to_cpu(ln->header.nr_entries) / 2;
+	nr_right = le32_to_cpu(ln->header.nr_entries) - nr_left;
+
+	ln->header.nr_entries = cpu_to_le32(nr_left);
+
+	rn->header.flags = ln->header.flags;
+	rn->header.nr_entries = cpu_to_le32(nr_right);
+	rn->header.max_entries = ln->header.max_entries;
+	rn->header.value_size = ln->header.value_size;
+	memcpy(rn->keys, ln->keys + nr_left, nr_right * sizeof(rn->keys[0]));
+
+	size = le32_to_cpu(ln->header.flags) & INTERNAL_NODE ?
+		sizeof(uint64_t) : s->info->value_type.size;
+	memcpy(value_ptr(rn, 0), value_ptr(ln, nr_left),
+	       size * nr_right);
+
+	/*
+	 * Patch up the parent
+	 */
+	parent = shadow_parent(s);
+
+	pn = dm_block_data(parent);
+	location = cpu_to_le64(dm_block_location(left));
+	__dm_bless_for_disk(&location);
+	memcpy_disk(value_ptr(pn, parent_index),
+		    &location, sizeof(__le64));
+
+	location = cpu_to_le64(dm_block_location(right));
+	__dm_bless_for_disk(&location);
+
+	r = insert_at(sizeof(__le64), pn, parent_index + 1,
+		      le64_to_cpu(rn->keys[0]), &location);
+	if (r)
+		return r;
+
+	if (key < le64_to_cpu(rn->keys[0])) {
+		unlock_block(s->info, right);
+		s->nodes[1] = left;
+	} else {
+		unlock_block(s->info, left);
+		s->nodes[1] = right;
+	}
+
+	return 0;
+}
+
+/*
+ * Splits a node by creating two new children beneath the given node.
+ *
+ * Before:
+ *	  +----------+
+ *	  | A ++++++ |
+ *	  +----------+
+ *
+ *
+ * After:
+ *	+------------+
+ *	| A (shadow) |
+ *	+------------+
+ *	    |	|
+ *   +------+	+----+
+ *   |		     |
+ *   v		     v
+ * +-------+	 +-------+
+ * | B +++ |	 | C +++ |
+ * +-------+	 +-------+
+ */
+static int btree_split_beneath(struct shadow_spine *s, uint64_t key)
+{
+	int r;
+	size_t size;
+	unsigned nr_left, nr_right;
+	struct dm_block *left, *right, *new_parent;
+	struct btree_node *pn, *ln, *rn;
+	__le64 val;
+
+	new_parent = shadow_current(s);
+
+	r = new_block(s->info, &left);
+	if (r < 0)
+		return r;
+
+	r = new_block(s->info, &right);
+	if (r < 0) {
+		/* FIXME: put left */
+		return r;
+	}
+
+	pn = dm_block_data(new_parent);
+	ln = dm_block_data(left);
+	rn = dm_block_data(right);
+
+	nr_left = le32_to_cpu(pn->header.nr_entries) / 2;
+	nr_right = le32_to_cpu(pn->header.nr_entries) - nr_left;
+
+	ln->header.flags = pn->header.flags;
+	ln->header.nr_entries = cpu_to_le32(nr_left);
+	ln->header.max_entries = pn->header.max_entries;
+	ln->header.value_size = pn->header.value_size;
+
+	rn->header.flags = pn->header.flags;
+	rn->header.nr_entries = cpu_to_le32(nr_right);
+	rn->header.max_entries = pn->header.max_entries;
+	rn->header.value_size = pn->header.value_size;
+
+	memcpy(ln->keys, pn->keys, nr_left * sizeof(pn->keys[0]));
+	memcpy(rn->keys, pn->keys + nr_left, nr_right * sizeof(pn->keys[0]));
+
+	size = le32_to_cpu(pn->header.flags) & INTERNAL_NODE ?
+		sizeof(__le64) : s->info->value_type.size;
+	memcpy(value_ptr(ln, 0), value_ptr(pn, 0), nr_left * size);
+	memcpy(value_ptr(rn, 0), value_ptr(pn, nr_left),
+	       nr_right * size);
+
+	/* new_parent should just point to l and r now */
+	pn->header.flags = cpu_to_le32(INTERNAL_NODE);
+	pn->header.nr_entries = cpu_to_le32(2);
+	pn->header.max_entries = cpu_to_le32(
+		calc_max_entries(sizeof(__le64),
+				 dm_bm_block_size(
+					 dm_tm_get_bm(s->info->tm))));
+	pn->header.value_size = cpu_to_le32(sizeof(__le64));
+
+	val = cpu_to_le64(dm_block_location(left));
+	__dm_bless_for_disk(&val);
+	pn->keys[0] = ln->keys[0];
+	memcpy_disk(value_ptr(pn, 0), &val, sizeof(__le64));
+
+	val = cpu_to_le64(dm_block_location(right));
+	__dm_bless_for_disk(&val);
+	pn->keys[1] = rn->keys[0];
+	memcpy_disk(value_ptr(pn, 1), &val, sizeof(__le64));
+
+	/*
+	 * rejig the spine.  This is ugly, since it knows too
+	 * much about the spine
+	 */
+	if (s->nodes[0] != new_parent) {
+		unlock_block(s->info, s->nodes[0]);
+		s->nodes[0] = new_parent;
+	}
+	if (key < le64_to_cpu(rn->keys[0])) {
+		unlock_block(s->info, right);
+		s->nodes[1] = left;
+	} else {
+		unlock_block(s->info, left);
+		s->nodes[1] = right;
+	}
+	s->count = 2;
+
+	return 0;
+}
+
+static int btree_insert_raw(struct shadow_spine *s, dm_block_t root,
+			    struct dm_btree_value_type *vt,
+			    uint64_t key, unsigned *index)
+{
+	int r, i = *index, top = 1;
+	struct btree_node *node;
+
+	for (;;) {
+		r = shadow_step(s, root, vt);
+		if (r < 0)
+			return r;
+
+		node = dm_block_data(shadow_current(s));
+
+		/*
+		 * We have to patch up the parent node, ugly, but I don't
+		 * see a way to do this automatically as part of the spine
+		 * op.
+		 */
+		if (shadow_has_parent(s) && i >= 0) { /* FIXME: second clause unness. */
+			__le64 location = cpu_to_le64(dm_block_location(shadow_current(s)));
+
+			__dm_bless_for_disk(&location);
+			memcpy_disk(value_ptr(dm_block_data(shadow_parent(s)), i),
+				    &location, sizeof(__le64));
+		}
+
+		node = dm_block_data(shadow_current(s));
+
+		if (node->header.nr_entries == node->header.max_entries) {
+			if (top)
+				r = btree_split_beneath(s, key);
+			else
+				r = btree_split_sibling(s, root, i, key);
+
+			if (r < 0)
+				return r;
+		}
+
+		node = dm_block_data(shadow_current(s));
+
+		i = lower_bound(node, key);
+
+		if (le32_to_cpu(node->header.flags) & LEAF_NODE)
+			break;
+
+		if (i < 0) {
+			/* change the bounds on the lowest key */
+			node->keys[0] = cpu_to_le64(key);
+			i = 0;
+		}
+
+		root = value64(node, i);
+		top = 0;
+	}
+
+	if (i < 0 || le64_to_cpu(node->keys[i]) != key)
+		i++;
+
+	*index = i;
+	return 0;
+}
+
+static int insert(struct dm_btree_info *info, dm_block_t root,
+		  uint64_t *keys, void *value, dm_block_t *new_root,
+		  int *inserted)
+		  __dm_written_to_disk(value)
+{
+	int r, need_insert;
+	unsigned level, index = -1, last_level = info->levels - 1;
+	dm_block_t block = root;
+	struct shadow_spine spine;
+	struct btree_node *n;
+	struct dm_btree_value_type le64_type;
+
+	le64_type.context = NULL;
+	le64_type.size = sizeof(__le64);
+	le64_type.inc = NULL;
+	le64_type.dec = NULL;
+	le64_type.equal = NULL;
+
+	init_shadow_spine(&spine, info);
+
+	for (level = 0; level < (info->levels - 1); level++) {
+		r = btree_insert_raw(&spine, block, &le64_type, keys[level], &index);
+		if (r < 0)
+			goto bad;
+
+		n = dm_block_data(shadow_current(&spine));
+		need_insert = ((index >= le32_to_cpu(n->header.nr_entries)) ||
+			       (le64_to_cpu(n->keys[index]) != keys[level]));
+
+		if (need_insert) {
+			dm_block_t new_tree;
+			__le64 new_le;
+
+			r = dm_btree_empty(info, &new_tree);
+			if (r < 0)
+				goto bad;
+
+			new_le = cpu_to_le64(new_tree);
+			__dm_bless_for_disk(&new_le);
+
+			r = insert_at(sizeof(uint64_t), n, index,
+				      keys[level], &new_le);
+			if (r)
+				goto bad;
+		}
+
+		if (level < last_level)
+			block = value64(n, index);
+	}
+
+	r = btree_insert_raw(&spine, block, &info->value_type,
+			     keys[level], &index);
+	if (r < 0)
+		goto bad;
+
+	n = dm_block_data(shadow_current(&spine));
+	need_insert = ((index >= le32_to_cpu(n->header.nr_entries)) ||
+		       (le64_to_cpu(n->keys[index]) != keys[level]));
+
+	if (need_insert) {
+		if (inserted)
+			*inserted = 1;
+
+		r = insert_at(info->value_type.size, n, index,
+			      keys[level], value);
+		if (r)
+			goto bad_unblessed;
+	} else {
+		if (inserted)
+			*inserted = 0;
+
+		if (info->value_type.dec &&
+		    (!info->value_type.equal ||
+		     !info->value_type.equal(
+			     info->value_type.context,
+			     value_ptr(n, index),
+			     value))) {
+			info->value_type.dec(info->value_type.context,
+					     value_ptr(n, index));
+		}
+		memcpy_disk(value_ptr(n, index),
+			    value, info->value_type.size);
+	}
+
+	*new_root = shadow_root(&spine);
+	exit_shadow_spine(&spine);
+
+	return 0;
+
+bad:
+	__dm_unbless_for_disk(value);
+bad_unblessed:
+	exit_shadow_spine(&spine);
+	return r;
+}
+
+int dm_btree_insert(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, void *value, dm_block_t *new_root)
+		    __dm_written_to_disk(value)
+{
+	return insert(info, root, keys, value, new_root, NULL);
+}
+EXPORT_SYMBOL_GPL(dm_btree_insert);
+
+int dm_btree_insert_notify(struct dm_btree_info *info, dm_block_t root,
+			   uint64_t *keys, void *value, dm_block_t *new_root,
+			   int *inserted)
+			   __dm_written_to_disk(value)
+{
+	return insert(info, root, keys, value, new_root, inserted);
+}
+EXPORT_SYMBOL_GPL(dm_btree_insert_notify);
+
+/*----------------------------------------------------------------*/
+
+static int find_key(struct ro_spine *s, dm_block_t block, bool find_highest,
+		    uint64_t *result_key, dm_block_t *next_block)
+{
+	int i, r;
+	uint32_t flags;
+
+	do {
+		r = ro_step(s, block);
+		if (r < 0)
+			return r;
+
+		flags = le32_to_cpu(ro_node(s)->header.flags);
+		i = le32_to_cpu(ro_node(s)->header.nr_entries);
+		if (!i)
+			return -ENODATA;
+		else
+			i--;
+
+		if (find_highest)
+			*result_key = le64_to_cpu(ro_node(s)->keys[i]);
+		else
+			*result_key = le64_to_cpu(ro_node(s)->keys[0]);
+
+		if (next_block || flags & INTERNAL_NODE)
+			block = value64(ro_node(s), i);
+
+	} while (flags & INTERNAL_NODE);
+
+	if (next_block)
+		*next_block = block;
+	return 0;
+}
+
+static int dm_btree_find_key(struct dm_btree_info *info, dm_block_t root,
+			     bool find_highest, uint64_t *result_keys)
+{
+	int r = 0, count = 0, level;
+	struct ro_spine spine;
+
+	init_ro_spine(&spine, info);
+	for (level = 0; level < info->levels; level++) {
+		r = find_key(&spine, root, find_highest, result_keys + level,
+			     level == info->levels - 1 ? NULL : &root);
+		if (r == -ENODATA) {
+			r = 0;
+			break;
+
+		} else if (r)
+			break;
+
+		count++;
+	}
+	exit_ro_spine(&spine);
+
+	return r ? r : count;
+}
+
+int dm_btree_find_highest_key(struct dm_btree_info *info, dm_block_t root,
+			      uint64_t *result_keys)
+{
+	return dm_btree_find_key(info, root, true, result_keys);
+}
+EXPORT_SYMBOL_GPL(dm_btree_find_highest_key);
+
+int dm_btree_find_lowest_key(struct dm_btree_info *info, dm_block_t root,
+			     uint64_t *result_keys)
+{
+	return dm_btree_find_key(info, root, false, result_keys);
+}
+EXPORT_SYMBOL_GPL(dm_btree_find_lowest_key);
+
+/*----------------------------------------------------------------*/
+
+/*
+ * FIXME: We shouldn't use a recursive algorithm when we have limited stack
+ * space.  Also this only works for single level trees.
+ */
+static int walk_node(struct ro_spine *s, dm_block_t block,
+		     int (*fn)(void *context, uint64_t *keys, void *leaf),
+		     void *context)
+{
+	int r;
+	unsigned i, nr;
+	struct btree_node *n;
+	uint64_t keys;
+
+	r = ro_step(s, block);
+	n = ro_node(s);
+
+	nr = le32_to_cpu(n->header.nr_entries);
+	for (i = 0; i < nr; i++) {
+		if (le32_to_cpu(n->header.flags) & INTERNAL_NODE) {
+			r = walk_node(s, value64(n, i), fn, context);
+			if (r)
+				goto out;
+		} else {
+			keys = le64_to_cpu(*key_ptr(n, i));
+			r = fn(context, &keys, value_ptr(n, i));
+			if (r)
+				goto out;
+		}
+	}
+
+out:
+	ro_pop(s);
+	return r;
+}
+
+int dm_btree_walk(struct dm_btree_info *info, dm_block_t root,
+		  int (*fn)(void *context, uint64_t *keys, void *leaf),
+		  void *context)
+{
+	int r;
+	struct ro_spine spine;
+
+	BUG_ON(info->levels > 1);
+
+	init_ro_spine(&spine, info);
+	r = walk_node(&spine, root, fn, context);
+	exit_ro_spine(&spine);
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_btree_walk);
diff --git a/drivers/md/persistent-data/dm-btree.h b/drivers/md/persistent-data/dm-btree.h
new file mode 100644
index 00000000000..dacfc34180b
--- /dev/null
+++ b/drivers/md/persistent-data/dm-btree.h
@@ -0,0 +1,162 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+#ifndef _LINUX_DM_BTREE_H
+#define _LINUX_DM_BTREE_H
+
+#include "dm-block-manager.h"
+
+struct dm_transaction_manager;
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Annotations used to check on-disk metadata is handled as little-endian.
+ */
+#ifdef __CHECKER__
+#  define __dm_written_to_disk(x) __releases(x)
+#  define __dm_reads_from_disk(x) __acquires(x)
+#  define __dm_bless_for_disk(x) __acquire(x)
+#  define __dm_unbless_for_disk(x) __release(x)
+#else
+#  define __dm_written_to_disk(x)
+#  define __dm_reads_from_disk(x)
+#  define __dm_bless_for_disk(x)
+#  define __dm_unbless_for_disk(x)
+#endif
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Manipulates hierarchical B+ trees with 64-bit keys and arbitrary-sized
+ * values.
+ */
+
+/*
+ * Information about the values stored within the btree.
+ */
+struct dm_btree_value_type {
+	void *context;
+
+	/*
+	 * The size in bytes of each value.
+	 */
+	uint32_t size;
+
+	/*
+	 * Any of these methods can be safely set to NULL if you do not
+	 * need the corresponding feature.
+	 */
+
+	/*
+	 * The btree is making a duplicate of the value, for instance
+	 * because previously-shared btree nodes have now diverged.
+	 * @value argument is the new copy that the copy function may modify.
+	 * (Probably it just wants to increment a reference count
+	 * somewhere.) This method is _not_ called for insertion of a new
+	 * value: It is assumed the ref count is already 1.
+	 */
+	void (*inc)(void *context, const void *value);
+
+	/*
+	 * This value is being deleted.  The btree takes care of freeing
+	 * the memory pointed to by @value.  Often the del function just
+	 * needs to decrement a reference count somewhere.
+	 */
+	void (*dec)(void *context, const void *value);
+
+	/*
+	 * A test for equality between two values.  When a value is
+	 * overwritten with a new one, the old one has the dec method
+	 * called _unless_ the new and old value are deemed equal.
+	 */
+	int (*equal)(void *context, const void *value1, const void *value2);
+};
+
+/*
+ * The shape and contents of a btree.
+ */
+struct dm_btree_info {
+	struct dm_transaction_manager *tm;
+
+	/*
+	 * Number of nested btrees. (Not the depth of a single tree.)
+	 */
+	unsigned levels;
+	struct dm_btree_value_type value_type;
+};
+
+/*
+ * Set up an empty tree.  O(1).
+ */
+int dm_btree_empty(struct dm_btree_info *info, dm_block_t *root);
+
+/*
+ * Delete a tree.  O(n) - this is the slow one!  It can also block, so
+ * please don't call it on an IO path.
+ */
+int dm_btree_del(struct dm_btree_info *info, dm_block_t root);
+
+/*
+ * All the lookup functions return -ENODATA if the key cannot be found.
+ */
+
+/*
+ * Tries to find a key that matches exactly.  O(ln(n))
+ */
+int dm_btree_lookup(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, void *value_le);
+
+/*
+ * Insertion (or overwrite an existing value).  O(ln(n))
+ */
+int dm_btree_insert(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, void *value, dm_block_t *new_root)
+		    __dm_written_to_disk(value);
+
+/*
+ * A variant of insert that indicates whether it actually inserted or just
+ * overwrote.  Useful if you're keeping track of the number of entries in a
+ * tree.
+ */
+int dm_btree_insert_notify(struct dm_btree_info *info, dm_block_t root,
+			   uint64_t *keys, void *value, dm_block_t *new_root,
+			   int *inserted)
+			   __dm_written_to_disk(value);
+
+/*
+ * Remove a key if present.  This doesn't remove empty sub trees.  Normally
+ * subtrees represent a separate entity, like a snapshot map, so this is
+ * correct behaviour.  O(ln(n)).
+ */
+int dm_btree_remove(struct dm_btree_info *info, dm_block_t root,
+		    uint64_t *keys, dm_block_t *new_root);
+
+/*
+ * Returns < 0 on failure.  Otherwise the number of key entries that have
+ * been filled out.  Remember trees can have zero entries, and as such have
+ * no lowest key.
+ */
+int dm_btree_find_lowest_key(struct dm_btree_info *info, dm_block_t root,
+			     uint64_t *result_keys);
+
+/*
+ * Returns < 0 on failure.  Otherwise the number of key entries that have
+ * been filled out.  Remember trees can have zero entries, and as such have
+ * no highest key.
+ */
+int dm_btree_find_highest_key(struct dm_btree_info *info, dm_block_t root,
+			      uint64_t *result_keys);
+
+/*
+ * Iterate through the a btree, calling fn() on each entry.
+ * It only works for single level trees and is internally recursive, so
+ * monitor stack usage carefully.
+ */
+int dm_btree_walk(struct dm_btree_info *info, dm_block_t root,
+		  int (*fn)(void *context, uint64_t *keys, void *leaf),
+		  void *context);
+
+#endif	/* _LINUX_DM_BTREE_H */
diff --git a/drivers/md/persistent-data/dm-persistent-data-internal.h b/drivers/md/persistent-data/dm-persistent-data-internal.h
new file mode 100644
index 00000000000..c49e26fff36
--- /dev/null
+++ b/drivers/md/persistent-data/dm-persistent-data-internal.h
@@ -0,0 +1,19 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _DM_PERSISTENT_DATA_INTERNAL_H
+#define _DM_PERSISTENT_DATA_INTERNAL_H
+
+#include "dm-block-manager.h"
+
+static inline unsigned dm_hash_block(dm_block_t b, unsigned hash_mask)
+{
+	const unsigned BIG_PRIME = 4294967291UL;
+
+	return (((unsigned) b) * BIG_PRIME) & hash_mask;
+}
+
+#endif	/* _PERSISTENT_DATA_INTERNAL_H */
diff --git a/drivers/md/persistent-data/dm-space-map-common.c b/drivers/md/persistent-data/dm-space-map-common.c
new file mode 100644
index 00000000000..aacbe70c2c2
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-common.c
@@ -0,0 +1,751 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-space-map-common.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/bitops.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "space map common"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Index validator.
+ */
+#define INDEX_CSUM_XOR 160478
+
+static void index_prepare_for_write(struct dm_block_validator *v,
+				    struct dm_block *b,
+				    size_t block_size)
+{
+	struct disk_metadata_index *mi_le = dm_block_data(b);
+
+	mi_le->blocknr = cpu_to_le64(dm_block_location(b));
+	mi_le->csum = cpu_to_le32(dm_bm_checksum(&mi_le->padding,
+						 block_size - sizeof(__le32),
+						 INDEX_CSUM_XOR));
+}
+
+static int index_check(struct dm_block_validator *v,
+		       struct dm_block *b,
+		       size_t block_size)
+{
+	struct disk_metadata_index *mi_le = dm_block_data(b);
+	__le32 csum_disk;
+
+	if (dm_block_location(b) != le64_to_cpu(mi_le->blocknr)) {
+		DMERR_LIMIT("index_check failed: blocknr %llu != wanted %llu",
+			    le64_to_cpu(mi_le->blocknr), dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	csum_disk = cpu_to_le32(dm_bm_checksum(&mi_le->padding,
+					       block_size - sizeof(__le32),
+					       INDEX_CSUM_XOR));
+	if (csum_disk != mi_le->csum) {
+		DMERR_LIMIT("index_check failed: csum %u != wanted %u",
+			    le32_to_cpu(csum_disk), le32_to_cpu(mi_le->csum));
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+static struct dm_block_validator index_validator = {
+	.name = "index",
+	.prepare_for_write = index_prepare_for_write,
+	.check = index_check
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Bitmap validator
+ */
+#define BITMAP_CSUM_XOR 240779
+
+static void bitmap_prepare_for_write(struct dm_block_validator *v,
+				     struct dm_block *b,
+				     size_t block_size)
+{
+	struct disk_bitmap_header *disk_header = dm_block_data(b);
+
+	disk_header->blocknr = cpu_to_le64(dm_block_location(b));
+	disk_header->csum = cpu_to_le32(dm_bm_checksum(&disk_header->not_used,
+						       block_size - sizeof(__le32),
+						       BITMAP_CSUM_XOR));
+}
+
+static int bitmap_check(struct dm_block_validator *v,
+			struct dm_block *b,
+			size_t block_size)
+{
+	struct disk_bitmap_header *disk_header = dm_block_data(b);
+	__le32 csum_disk;
+
+	if (dm_block_location(b) != le64_to_cpu(disk_header->blocknr)) {
+		DMERR_LIMIT("bitmap check failed: blocknr %llu != wanted %llu",
+			    le64_to_cpu(disk_header->blocknr), dm_block_location(b));
+		return -ENOTBLK;
+	}
+
+	csum_disk = cpu_to_le32(dm_bm_checksum(&disk_header->not_used,
+					       block_size - sizeof(__le32),
+					       BITMAP_CSUM_XOR));
+	if (csum_disk != disk_header->csum) {
+		DMERR_LIMIT("bitmap check failed: csum %u != wanted %u",
+			    le32_to_cpu(csum_disk), le32_to_cpu(disk_header->csum));
+		return -EILSEQ;
+	}
+
+	return 0;
+}
+
+static struct dm_block_validator dm_sm_bitmap_validator = {
+	.name = "sm_bitmap",
+	.prepare_for_write = bitmap_prepare_for_write,
+	.check = bitmap_check
+};
+
+/*----------------------------------------------------------------*/
+
+#define ENTRIES_PER_WORD 32
+#define ENTRIES_SHIFT	5
+
+static void *dm_bitmap_data(struct dm_block *b)
+{
+	return dm_block_data(b) + sizeof(struct disk_bitmap_header);
+}
+
+#define WORD_MASK_HIGH 0xAAAAAAAAAAAAAAAAULL
+
+static unsigned bitmap_word_used(void *addr, unsigned b)
+{
+	__le64 *words_le = addr;
+	__le64 *w_le = words_le + (b >> ENTRIES_SHIFT);
+
+	uint64_t bits = le64_to_cpu(*w_le);
+	uint64_t mask = (bits + WORD_MASK_HIGH + 1) & WORD_MASK_HIGH;
+
+	return !(~bits & mask);
+}
+
+static unsigned sm_lookup_bitmap(void *addr, unsigned b)
+{
+	__le64 *words_le = addr;
+	__le64 *w_le = words_le + (b >> ENTRIES_SHIFT);
+	unsigned hi, lo;
+
+	b = (b & (ENTRIES_PER_WORD - 1)) << 1;
+	hi = !!test_bit_le(b, (void *) w_le);
+	lo = !!test_bit_le(b + 1, (void *) w_le);
+	return (hi << 1) | lo;
+}
+
+static void sm_set_bitmap(void *addr, unsigned b, unsigned val)
+{
+	__le64 *words_le = addr;
+	__le64 *w_le = words_le + (b >> ENTRIES_SHIFT);
+
+	b = (b & (ENTRIES_PER_WORD - 1)) << 1;
+
+	if (val & 2)
+		__set_bit_le(b, (void *) w_le);
+	else
+		__clear_bit_le(b, (void *) w_le);
+
+	if (val & 1)
+		__set_bit_le(b + 1, (void *) w_le);
+	else
+		__clear_bit_le(b + 1, (void *) w_le);
+}
+
+static int sm_find_free(void *addr, unsigned begin, unsigned end,
+			unsigned *result)
+{
+	while (begin < end) {
+		if (!(begin & (ENTRIES_PER_WORD - 1)) &&
+		    bitmap_word_used(addr, begin)) {
+			begin += ENTRIES_PER_WORD;
+			continue;
+		}
+
+		if (!sm_lookup_bitmap(addr, begin)) {
+			*result = begin;
+			return 0;
+		}
+
+		begin++;
+	}
+
+	return -ENOSPC;
+}
+
+/*----------------------------------------------------------------*/
+
+static int sm_ll_init(struct ll_disk *ll, struct dm_transaction_manager *tm)
+{
+	ll->tm = tm;
+
+	ll->bitmap_info.tm = tm;
+	ll->bitmap_info.levels = 1;
+
+	/*
+	 * Because the new bitmap blocks are created via a shadow
+	 * operation, the old entry has already had its reference count
+	 * decremented and we don't need the btree to do any bookkeeping.
+	 */
+	ll->bitmap_info.value_type.size = sizeof(struct disk_index_entry);
+	ll->bitmap_info.value_type.inc = NULL;
+	ll->bitmap_info.value_type.dec = NULL;
+	ll->bitmap_info.value_type.equal = NULL;
+
+	ll->ref_count_info.tm = tm;
+	ll->ref_count_info.levels = 1;
+	ll->ref_count_info.value_type.size = sizeof(uint32_t);
+	ll->ref_count_info.value_type.inc = NULL;
+	ll->ref_count_info.value_type.dec = NULL;
+	ll->ref_count_info.value_type.equal = NULL;
+
+	ll->block_size = dm_bm_block_size(dm_tm_get_bm(tm));
+
+	if (ll->block_size > (1 << 30)) {
+		DMERR("block size too big to hold bitmaps");
+		return -EINVAL;
+	}
+
+	ll->entries_per_block = (ll->block_size - sizeof(struct disk_bitmap_header)) *
+		ENTRIES_PER_BYTE;
+	ll->nr_blocks = 0;
+	ll->bitmap_root = 0;
+	ll->ref_count_root = 0;
+	ll->bitmap_index_changed = false;
+
+	return 0;
+}
+
+int sm_ll_extend(struct ll_disk *ll, dm_block_t extra_blocks)
+{
+	int r;
+	dm_block_t i, nr_blocks, nr_indexes;
+	unsigned old_blocks, blocks;
+
+	nr_blocks = ll->nr_blocks + extra_blocks;
+	old_blocks = dm_sector_div_up(ll->nr_blocks, ll->entries_per_block);
+	blocks = dm_sector_div_up(nr_blocks, ll->entries_per_block);
+
+	nr_indexes = dm_sector_div_up(nr_blocks, ll->entries_per_block);
+	if (nr_indexes > ll->max_entries(ll)) {
+		DMERR("space map too large");
+		return -EINVAL;
+	}
+
+	/*
+	 * We need to set this before the dm_tm_new_block() call below.
+	 */
+	ll->nr_blocks = nr_blocks;
+	for (i = old_blocks; i < blocks; i++) {
+		struct dm_block *b;
+		struct disk_index_entry idx;
+
+		r = dm_tm_new_block(ll->tm, &dm_sm_bitmap_validator, &b);
+		if (r < 0)
+			return r;
+
+		idx.blocknr = cpu_to_le64(dm_block_location(b));
+
+		r = dm_tm_unlock(ll->tm, b);
+		if (r < 0)
+			return r;
+
+		idx.nr_free = cpu_to_le32(ll->entries_per_block);
+		idx.none_free_before = 0;
+
+		r = ll->save_ie(ll, i, &idx);
+		if (r < 0)
+			return r;
+	}
+
+	return 0;
+}
+
+int sm_ll_lookup_bitmap(struct ll_disk *ll, dm_block_t b, uint32_t *result)
+{
+	int r;
+	dm_block_t index = b;
+	struct disk_index_entry ie_disk;
+	struct dm_block *blk;
+
+	b = do_div(index, ll->entries_per_block);
+	r = ll->load_ie(ll, index, &ie_disk);
+	if (r < 0)
+		return r;
+
+	r = dm_tm_read_lock(ll->tm, le64_to_cpu(ie_disk.blocknr),
+			    &dm_sm_bitmap_validator, &blk);
+	if (r < 0)
+		return r;
+
+	*result = sm_lookup_bitmap(dm_bitmap_data(blk), b);
+
+	return dm_tm_unlock(ll->tm, blk);
+}
+
+static int sm_ll_lookup_big_ref_count(struct ll_disk *ll, dm_block_t b,
+				      uint32_t *result)
+{
+	__le32 le_rc;
+	int r;
+
+	r = dm_btree_lookup(&ll->ref_count_info, ll->ref_count_root, &b, &le_rc);
+	if (r < 0)
+		return r;
+
+	*result = le32_to_cpu(le_rc);
+
+	return r;
+}
+
+int sm_ll_lookup(struct ll_disk *ll, dm_block_t b, uint32_t *result)
+{
+	int r = sm_ll_lookup_bitmap(ll, b, result);
+
+	if (r)
+		return r;
+
+	if (*result != 3)
+		return r;
+
+	return sm_ll_lookup_big_ref_count(ll, b, result);
+}
+
+int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin,
+			  dm_block_t end, dm_block_t *result)
+{
+	int r;
+	struct disk_index_entry ie_disk;
+	dm_block_t i, index_begin = begin;
+	dm_block_t index_end = dm_sector_div_up(end, ll->entries_per_block);
+
+	/*
+	 * FIXME: Use shifts
+	 */
+	begin = do_div(index_begin, ll->entries_per_block);
+	end = do_div(end, ll->entries_per_block);
+
+	for (i = index_begin; i < index_end; i++, begin = 0) {
+		struct dm_block *blk;
+		unsigned position;
+		uint32_t bit_end;
+
+		r = ll->load_ie(ll, i, &ie_disk);
+		if (r < 0)
+			return r;
+
+		if (le32_to_cpu(ie_disk.nr_free) == 0)
+			continue;
+
+		r = dm_tm_read_lock(ll->tm, le64_to_cpu(ie_disk.blocknr),
+				    &dm_sm_bitmap_validator, &blk);
+		if (r < 0)
+			return r;
+
+		bit_end = (i == index_end - 1) ?  end : ll->entries_per_block;
+
+		r = sm_find_free(dm_bitmap_data(blk),
+				 max_t(unsigned, begin, le32_to_cpu(ie_disk.none_free_before)),
+				 bit_end, &position);
+		if (r == -ENOSPC) {
+			/*
+			 * This might happen because we started searching
+			 * part way through the bitmap.
+			 */
+			dm_tm_unlock(ll->tm, blk);
+			continue;
+
+		} else if (r < 0) {
+			dm_tm_unlock(ll->tm, blk);
+			return r;
+		}
+
+		r = dm_tm_unlock(ll->tm, blk);
+		if (r < 0)
+			return r;
+
+		*result = i * ll->entries_per_block + (dm_block_t) position;
+		return 0;
+	}
+
+	return -ENOSPC;
+}
+
+static int sm_ll_mutate(struct ll_disk *ll, dm_block_t b,
+			int (*mutator)(void *context, uint32_t old, uint32_t *new),
+			void *context, enum allocation_event *ev)
+{
+	int r;
+	uint32_t bit, old, ref_count;
+	struct dm_block *nb;
+	dm_block_t index = b;
+	struct disk_index_entry ie_disk;
+	void *bm_le;
+	int inc;
+
+	bit = do_div(index, ll->entries_per_block);
+	r = ll->load_ie(ll, index, &ie_disk);
+	if (r < 0)
+		return r;
+
+	r = dm_tm_shadow_block(ll->tm, le64_to_cpu(ie_disk.blocknr),
+			       &dm_sm_bitmap_validator, &nb, &inc);
+	if (r < 0) {
+		DMERR("dm_tm_shadow_block() failed");
+		return r;
+	}
+	ie_disk.blocknr = cpu_to_le64(dm_block_location(nb));
+
+	bm_le = dm_bitmap_data(nb);
+	old = sm_lookup_bitmap(bm_le, bit);
+
+	if (old > 2) {
+		r = sm_ll_lookup_big_ref_count(ll, b, &old);
+		if (r < 0) {
+			dm_tm_unlock(ll->tm, nb);
+			return r;
+		}
+	}
+
+	r = mutator(context, old, &ref_count);
+	if (r) {
+		dm_tm_unlock(ll->tm, nb);
+		return r;
+	}
+
+	if (ref_count <= 2) {
+		sm_set_bitmap(bm_le, bit, ref_count);
+
+		r = dm_tm_unlock(ll->tm, nb);
+		if (r < 0)
+			return r;
+
+		if (old > 2) {
+			r = dm_btree_remove(&ll->ref_count_info,
+					    ll->ref_count_root,
+					    &b, &ll->ref_count_root);
+			if (r)
+				return r;
+		}
+
+	} else {
+		__le32 le_rc = cpu_to_le32(ref_count);
+
+		sm_set_bitmap(bm_le, bit, 3);
+		r = dm_tm_unlock(ll->tm, nb);
+		if (r < 0)
+			return r;
+
+		__dm_bless_for_disk(&le_rc);
+		r = dm_btree_insert(&ll->ref_count_info, ll->ref_count_root,
+				    &b, &le_rc, &ll->ref_count_root);
+		if (r < 0) {
+			DMERR("ref count insert failed");
+			return r;
+		}
+	}
+
+	if (ref_count && !old) {
+		*ev = SM_ALLOC;
+		ll->nr_allocated++;
+		le32_add_cpu(&ie_disk.nr_free, -1);
+		if (le32_to_cpu(ie_disk.none_free_before) == bit)
+			ie_disk.none_free_before = cpu_to_le32(bit + 1);
+
+	} else if (old && !ref_count) {
+		*ev = SM_FREE;
+		ll->nr_allocated--;
+		le32_add_cpu(&ie_disk.nr_free, 1);
+		ie_disk.none_free_before = cpu_to_le32(min(le32_to_cpu(ie_disk.none_free_before), bit));
+	}
+
+	return ll->save_ie(ll, index, &ie_disk);
+}
+
+static int set_ref_count(void *context, uint32_t old, uint32_t *new)
+{
+	*new = *((uint32_t *) context);
+	return 0;
+}
+
+int sm_ll_insert(struct ll_disk *ll, dm_block_t b,
+		 uint32_t ref_count, enum allocation_event *ev)
+{
+	return sm_ll_mutate(ll, b, set_ref_count, &ref_count, ev);
+}
+
+static int inc_ref_count(void *context, uint32_t old, uint32_t *new)
+{
+	*new = old + 1;
+	return 0;
+}
+
+int sm_ll_inc(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev)
+{
+	return sm_ll_mutate(ll, b, inc_ref_count, NULL, ev);
+}
+
+static int dec_ref_count(void *context, uint32_t old, uint32_t *new)
+{
+	if (!old) {
+		DMERR_LIMIT("unable to decrement a reference count below 0");
+		return -EINVAL;
+	}
+
+	*new = old - 1;
+	return 0;
+}
+
+int sm_ll_dec(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev)
+{
+	return sm_ll_mutate(ll, b, dec_ref_count, NULL, ev);
+}
+
+int sm_ll_commit(struct ll_disk *ll)
+{
+	int r = 0;
+
+	if (ll->bitmap_index_changed) {
+		r = ll->commit(ll);
+		if (!r)
+			ll->bitmap_index_changed = false;
+	}
+
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+static int metadata_ll_load_ie(struct ll_disk *ll, dm_block_t index,
+			       struct disk_index_entry *ie)
+{
+	memcpy(ie, ll->mi_le.index + index, sizeof(*ie));
+	return 0;
+}
+
+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;
+}
+
+static int metadata_ll_init_index(struct ll_disk *ll)
+{
+	int r;
+	struct dm_block *b;
+
+	r = dm_tm_new_block(ll->tm, &index_validator, &b);
+	if (r < 0)
+		return r;
+
+	memcpy(dm_block_data(b), &ll->mi_le, sizeof(ll->mi_le));
+	ll->bitmap_root = dm_block_location(b);
+
+	return dm_tm_unlock(ll->tm, b);
+}
+
+static int metadata_ll_open(struct ll_disk *ll)
+{
+	int r;
+	struct dm_block *block;
+
+	r = dm_tm_read_lock(ll->tm, ll->bitmap_root,
+			    &index_validator, &block);
+	if (r)
+		return r;
+
+	memcpy(&ll->mi_le, dm_block_data(block), sizeof(ll->mi_le));
+	return dm_tm_unlock(ll->tm, block);
+}
+
+static dm_block_t metadata_ll_max_entries(struct ll_disk *ll)
+{
+	return MAX_METADATA_BITMAPS;
+}
+
+static int metadata_ll_commit(struct ll_disk *ll)
+{
+	int r, inc;
+	struct dm_block *b;
+
+	r = dm_tm_shadow_block(ll->tm, ll->bitmap_root, &index_validator, &b, &inc);
+	if (r)
+		return r;
+
+	memcpy(dm_block_data(b), &ll->mi_le, sizeof(ll->mi_le));
+	ll->bitmap_root = dm_block_location(b);
+
+	return dm_tm_unlock(ll->tm, b);
+}
+
+int sm_ll_new_metadata(struct ll_disk *ll, struct dm_transaction_manager *tm)
+{
+	int r;
+
+	r = sm_ll_init(ll, tm);
+	if (r < 0)
+		return r;
+
+	ll->load_ie = metadata_ll_load_ie;
+	ll->save_ie = metadata_ll_save_ie;
+	ll->init_index = metadata_ll_init_index;
+	ll->open_index = metadata_ll_open;
+	ll->max_entries = metadata_ll_max_entries;
+	ll->commit = metadata_ll_commit;
+
+	ll->nr_blocks = 0;
+	ll->nr_allocated = 0;
+
+	r = ll->init_index(ll);
+	if (r < 0)
+		return r;
+
+	r = dm_btree_empty(&ll->ref_count_info, &ll->ref_count_root);
+	if (r < 0)
+		return r;
+
+	return 0;
+}
+
+int sm_ll_open_metadata(struct ll_disk *ll, struct dm_transaction_manager *tm,
+			void *root_le, size_t len)
+{
+	int r;
+	struct disk_sm_root *smr = root_le;
+
+	if (len < sizeof(struct disk_sm_root)) {
+		DMERR("sm_metadata root too small");
+		return -ENOMEM;
+	}
+
+	r = sm_ll_init(ll, tm);
+	if (r < 0)
+		return r;
+
+	ll->load_ie = metadata_ll_load_ie;
+	ll->save_ie = metadata_ll_save_ie;
+	ll->init_index = metadata_ll_init_index;
+	ll->open_index = metadata_ll_open;
+	ll->max_entries = metadata_ll_max_entries;
+	ll->commit = metadata_ll_commit;
+
+	ll->nr_blocks = le64_to_cpu(smr->nr_blocks);
+	ll->nr_allocated = le64_to_cpu(smr->nr_allocated);
+	ll->bitmap_root = le64_to_cpu(smr->bitmap_root);
+	ll->ref_count_root = le64_to_cpu(smr->ref_count_root);
+
+	return ll->open_index(ll);
+}
+
+/*----------------------------------------------------------------*/
+
+static int disk_ll_load_ie(struct ll_disk *ll, dm_block_t index,
+			   struct disk_index_entry *ie)
+{
+	return dm_btree_lookup(&ll->bitmap_info, ll->bitmap_root, &index, ie);
+}
+
+static int disk_ll_save_ie(struct ll_disk *ll, dm_block_t index,
+			   struct disk_index_entry *ie)
+{
+	__dm_bless_for_disk(ie);
+	return dm_btree_insert(&ll->bitmap_info, ll->bitmap_root,
+			       &index, ie, &ll->bitmap_root);
+}
+
+static int disk_ll_init_index(struct ll_disk *ll)
+{
+	return dm_btree_empty(&ll->bitmap_info, &ll->bitmap_root);
+}
+
+static int disk_ll_open(struct ll_disk *ll)
+{
+	/* nothing to do */
+	return 0;
+}
+
+static dm_block_t disk_ll_max_entries(struct ll_disk *ll)
+{
+	return -1ULL;
+}
+
+static int disk_ll_commit(struct ll_disk *ll)
+{
+	return 0;
+}
+
+int sm_ll_new_disk(struct ll_disk *ll, struct dm_transaction_manager *tm)
+{
+	int r;
+
+	r = sm_ll_init(ll, tm);
+	if (r < 0)
+		return r;
+
+	ll->load_ie = disk_ll_load_ie;
+	ll->save_ie = disk_ll_save_ie;
+	ll->init_index = disk_ll_init_index;
+	ll->open_index = disk_ll_open;
+	ll->max_entries = disk_ll_max_entries;
+	ll->commit = disk_ll_commit;
+
+	ll->nr_blocks = 0;
+	ll->nr_allocated = 0;
+
+	r = ll->init_index(ll);
+	if (r < 0)
+		return r;
+
+	r = dm_btree_empty(&ll->ref_count_info, &ll->ref_count_root);
+	if (r < 0)
+		return r;
+
+	return 0;
+}
+
+int sm_ll_open_disk(struct ll_disk *ll, struct dm_transaction_manager *tm,
+		    void *root_le, size_t len)
+{
+	int r;
+	struct disk_sm_root *smr = root_le;
+
+	if (len < sizeof(struct disk_sm_root)) {
+		DMERR("sm_metadata root too small");
+		return -ENOMEM;
+	}
+
+	r = sm_ll_init(ll, tm);
+	if (r < 0)
+		return r;
+
+	ll->load_ie = disk_ll_load_ie;
+	ll->save_ie = disk_ll_save_ie;
+	ll->init_index = disk_ll_init_index;
+	ll->open_index = disk_ll_open;
+	ll->max_entries = disk_ll_max_entries;
+	ll->commit = disk_ll_commit;
+
+	ll->nr_blocks = le64_to_cpu(smr->nr_blocks);
+	ll->nr_allocated = le64_to_cpu(smr->nr_allocated);
+	ll->bitmap_root = le64_to_cpu(smr->bitmap_root);
+	ll->ref_count_root = le64_to_cpu(smr->ref_count_root);
+
+	return ll->open_index(ll);
+}
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-space-map-common.h b/drivers/md/persistent-data/dm-space-map-common.h
new file mode 100644
index 00000000000..b3078d5eda0
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-common.h
@@ -0,0 +1,127 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_SPACE_MAP_COMMON_H
+#define DM_SPACE_MAP_COMMON_H
+
+#include "dm-btree.h"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Low level disk format
+ *
+ * Bitmap btree
+ * ------------
+ *
+ * Each value stored in the btree is an index_entry.  This points to a
+ * block that is used as a bitmap.  Within the bitmap hold 2 bits per
+ * entry, which represent UNUSED = 0, REF_COUNT = 1, REF_COUNT = 2 and
+ * REF_COUNT = many.
+ *
+ * Refcount btree
+ * --------------
+ *
+ * Any entry that has a ref count higher than 2 gets entered in the ref
+ * count tree.  The leaf values for this tree is the 32-bit ref count.
+ */
+
+struct disk_index_entry {
+	__le64 blocknr;
+	__le32 nr_free;
+	__le32 none_free_before;
+} __packed;
+
+
+#define MAX_METADATA_BITMAPS 255
+struct disk_metadata_index {
+	__le32 csum;
+	__le32 padding;
+	__le64 blocknr;
+
+	struct disk_index_entry index[MAX_METADATA_BITMAPS];
+} __packed;
+
+struct ll_disk;
+
+typedef int (*load_ie_fn)(struct ll_disk *ll, dm_block_t index, struct disk_index_entry *result);
+typedef int (*save_ie_fn)(struct ll_disk *ll, dm_block_t index, struct disk_index_entry *ie);
+typedef int (*init_index_fn)(struct ll_disk *ll);
+typedef int (*open_index_fn)(struct ll_disk *ll);
+typedef dm_block_t (*max_index_entries_fn)(struct ll_disk *ll);
+typedef int (*commit_fn)(struct ll_disk *ll);
+
+struct ll_disk {
+	struct dm_transaction_manager *tm;
+	struct dm_btree_info bitmap_info;
+	struct dm_btree_info ref_count_info;
+
+	uint32_t block_size;
+	uint32_t entries_per_block;
+	dm_block_t nr_blocks;
+	dm_block_t nr_allocated;
+
+	/*
+	 * bitmap_root may be a btree root or a simple index.
+	 */
+	dm_block_t bitmap_root;
+
+	dm_block_t ref_count_root;
+
+	struct disk_metadata_index mi_le;
+	load_ie_fn load_ie;
+	save_ie_fn save_ie;
+	init_index_fn init_index;
+	open_index_fn open_index;
+	max_index_entries_fn max_entries;
+	commit_fn commit;
+	bool bitmap_index_changed:1;
+};
+
+struct disk_sm_root {
+	__le64 nr_blocks;
+	__le64 nr_allocated;
+	__le64 bitmap_root;
+	__le64 ref_count_root;
+} __packed;
+
+#define ENTRIES_PER_BYTE 4
+
+struct disk_bitmap_header {
+	__le32 csum;
+	__le32 not_used;
+	__le64 blocknr;
+} __packed;
+
+enum allocation_event {
+	SM_NONE,
+	SM_ALLOC,
+	SM_FREE,
+};
+
+/*----------------------------------------------------------------*/
+
+int sm_ll_extend(struct ll_disk *ll, dm_block_t extra_blocks);
+int sm_ll_lookup_bitmap(struct ll_disk *ll, dm_block_t b, uint32_t *result);
+int sm_ll_lookup(struct ll_disk *ll, dm_block_t b, uint32_t *result);
+int sm_ll_find_free_block(struct ll_disk *ll, dm_block_t begin,
+			  dm_block_t end, dm_block_t *result);
+int sm_ll_insert(struct ll_disk *ll, dm_block_t b, uint32_t ref_count, enum allocation_event *ev);
+int sm_ll_inc(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev);
+int sm_ll_dec(struct ll_disk *ll, dm_block_t b, enum allocation_event *ev);
+int sm_ll_commit(struct ll_disk *ll);
+
+int sm_ll_new_metadata(struct ll_disk *ll, struct dm_transaction_manager *tm);
+int sm_ll_open_metadata(struct ll_disk *ll, struct dm_transaction_manager *tm,
+			void *root_le, size_t len);
+
+int sm_ll_new_disk(struct ll_disk *ll, struct dm_transaction_manager *tm);
+int sm_ll_open_disk(struct ll_disk *ll, struct dm_transaction_manager *tm,
+		    void *root_le, size_t len);
+
+/*----------------------------------------------------------------*/
+
+#endif	/* DM_SPACE_MAP_COMMON_H */
diff --git a/drivers/md/persistent-data/dm-space-map-disk.c b/drivers/md/persistent-data/dm-space-map-disk.c
new file mode 100644
index 00000000000..cfbf9617e46
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-disk.c
@@ -0,0 +1,303 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-space-map-common.h"
+#include "dm-space-map-disk.h"
+#include "dm-space-map.h"
+#include "dm-transaction-manager.h"
+
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/export.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "space map disk"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Space map interface.
+ */
+struct sm_disk {
+	struct dm_space_map sm;
+
+	struct ll_disk ll;
+	struct ll_disk old_ll;
+
+	dm_block_t begin;
+	dm_block_t nr_allocated_this_transaction;
+};
+
+static void sm_disk_destroy(struct dm_space_map *sm)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	kfree(smd);
+}
+
+static int sm_disk_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	return sm_ll_extend(&smd->ll, extra_blocks);
+}
+
+static int sm_disk_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+	*count = smd->old_ll.nr_blocks;
+
+	return 0;
+}
+
+static int sm_disk_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+	*count = (smd->old_ll.nr_blocks - smd->old_ll.nr_allocated) - smd->nr_allocated_this_transaction;
+
+	return 0;
+}
+
+static int sm_disk_get_count(struct dm_space_map *sm, dm_block_t b,
+			     uint32_t *result)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+	return sm_ll_lookup(&smd->ll, b, result);
+}
+
+static int sm_disk_count_is_more_than_one(struct dm_space_map *sm, dm_block_t b,
+					  int *result)
+{
+	int r;
+	uint32_t count;
+
+	r = sm_disk_get_count(sm, b, &count);
+	if (r)
+		return r;
+
+	return count > 1;
+}
+
+static int sm_disk_set_count(struct dm_space_map *sm, dm_block_t b,
+			     uint32_t count)
+{
+	int r;
+	uint32_t old_count;
+	enum allocation_event ev;
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	r = sm_ll_insert(&smd->ll, b, count, &ev);
+	if (!r) {
+		switch (ev) {
+		case SM_NONE:
+			break;
+
+		case SM_ALLOC:
+			/*
+			 * This _must_ be free in the prior transaction
+			 * otherwise we've lost atomicity.
+			 */
+			smd->nr_allocated_this_transaction++;
+			break;
+
+		case SM_FREE:
+			/*
+			 * It's only free if it's also free in the last
+			 * transaction.
+			 */
+			r = sm_ll_lookup(&smd->old_ll, b, &old_count);
+			if (r)
+				return r;
+
+			if (!old_count)
+				smd->nr_allocated_this_transaction--;
+			break;
+		}
+	}
+
+	return r;
+}
+
+static int sm_disk_inc_block(struct dm_space_map *sm, dm_block_t b)
+{
+	int r;
+	enum allocation_event ev;
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	r = sm_ll_inc(&smd->ll, b, &ev);
+	if (!r && (ev == SM_ALLOC))
+		/*
+		 * This _must_ be free in the prior transaction
+		 * otherwise we've lost atomicity.
+		 */
+		smd->nr_allocated_this_transaction++;
+
+	return r;
+}
+
+static int sm_disk_dec_block(struct dm_space_map *sm, dm_block_t b)
+{
+	enum allocation_event ev;
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	return sm_ll_dec(&smd->ll, b, &ev);
+}
+
+static int sm_disk_new_block(struct dm_space_map *sm, dm_block_t *b)
+{
+	int r;
+	enum allocation_event ev;
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	/* FIXME: we should loop round a couple of times */
+	r = sm_ll_find_free_block(&smd->old_ll, smd->begin, smd->old_ll.nr_blocks, b);
+	if (r)
+		return r;
+
+	smd->begin = *b + 1;
+	r = sm_ll_inc(&smd->ll, *b, &ev);
+	if (!r) {
+		BUG_ON(ev != SM_ALLOC);
+		smd->nr_allocated_this_transaction++;
+	}
+
+	return r;
+}
+
+static int sm_disk_commit(struct dm_space_map *sm)
+{
+	int r;
+	dm_block_t nr_free;
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+
+	r = sm_disk_get_nr_free(sm, &nr_free);
+	if (r)
+		return r;
+
+	r = sm_ll_commit(&smd->ll);
+	if (r)
+		return r;
+
+	memcpy(&smd->old_ll, &smd->ll, sizeof(smd->old_ll));
+	smd->begin = 0;
+	smd->nr_allocated_this_transaction = 0;
+
+	r = sm_disk_get_nr_free(sm, &nr_free);
+	if (r)
+		return r;
+
+	return 0;
+}
+
+static int sm_disk_root_size(struct dm_space_map *sm, size_t *result)
+{
+	*result = sizeof(struct disk_sm_root);
+
+	return 0;
+}
+
+static int sm_disk_copy_root(struct dm_space_map *sm, void *where_le, size_t max)
+{
+	struct sm_disk *smd = container_of(sm, struct sm_disk, sm);
+	struct disk_sm_root root_le;
+
+	root_le.nr_blocks = cpu_to_le64(smd->ll.nr_blocks);
+	root_le.nr_allocated = cpu_to_le64(smd->ll.nr_allocated);
+	root_le.bitmap_root = cpu_to_le64(smd->ll.bitmap_root);
+	root_le.ref_count_root = cpu_to_le64(smd->ll.ref_count_root);
+
+	if (max < sizeof(root_le))
+		return -ENOSPC;
+
+	memcpy(where_le, &root_le, sizeof(root_le));
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
+static struct dm_space_map ops = {
+	.destroy = sm_disk_destroy,
+	.extend = sm_disk_extend,
+	.get_nr_blocks = sm_disk_get_nr_blocks,
+	.get_nr_free = sm_disk_get_nr_free,
+	.get_count = sm_disk_get_count,
+	.count_is_more_than_one = sm_disk_count_is_more_than_one,
+	.set_count = sm_disk_set_count,
+	.inc_block = sm_disk_inc_block,
+	.dec_block = sm_disk_dec_block,
+	.new_block = sm_disk_new_block,
+	.commit = sm_disk_commit,
+	.root_size = sm_disk_root_size,
+	.copy_root = sm_disk_copy_root,
+	.register_threshold_callback = NULL
+};
+
+struct dm_space_map *dm_sm_disk_create(struct dm_transaction_manager *tm,
+				       dm_block_t nr_blocks)
+{
+	int r;
+	struct sm_disk *smd;
+
+	smd = kmalloc(sizeof(*smd), GFP_KERNEL);
+	if (!smd)
+		return ERR_PTR(-ENOMEM);
+
+	smd->begin = 0;
+	smd->nr_allocated_this_transaction = 0;
+	memcpy(&smd->sm, &ops, sizeof(smd->sm));
+
+	r = sm_ll_new_disk(&smd->ll, tm);
+	if (r)
+		goto bad;
+
+	r = sm_ll_extend(&smd->ll, nr_blocks);
+	if (r)
+		goto bad;
+
+	r = sm_disk_commit(&smd->sm);
+	if (r)
+		goto bad;
+
+	return &smd->sm;
+
+bad:
+	kfree(smd);
+	return ERR_PTR(r);
+}
+EXPORT_SYMBOL_GPL(dm_sm_disk_create);
+
+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;
+
+	smd = kmalloc(sizeof(*smd), GFP_KERNEL);
+	if (!smd)
+		return ERR_PTR(-ENOMEM);
+
+	smd->begin = 0;
+	smd->nr_allocated_this_transaction = 0;
+	memcpy(&smd->sm, &ops, sizeof(smd->sm));
+
+	r = sm_ll_open_disk(&smd->ll, tm, root_le, len);
+	if (r)
+		goto bad;
+
+	r = sm_disk_commit(&smd->sm);
+	if (r)
+		goto bad;
+
+	return &smd->sm;
+
+bad:
+	kfree(smd);
+	return ERR_PTR(r);
+}
+EXPORT_SYMBOL_GPL(dm_sm_disk_open);
+
+/*----------------------------------------------------------------*/
diff --git a/drivers/md/persistent-data/dm-space-map-disk.h b/drivers/md/persistent-data/dm-space-map-disk.h
new file mode 100644
index 00000000000..447a0a9a2d9
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-disk.h
@@ -0,0 +1,25 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_SPACE_MAP_DISK_H
+#define _LINUX_DM_SPACE_MAP_DISK_H
+
+#include "dm-block-manager.h"
+
+struct dm_space_map;
+struct dm_transaction_manager;
+
+/*
+ * Unfortunately we have to use two-phase construction due to the cycle
+ * between the tm and sm.
+ */
+struct dm_space_map *dm_sm_disk_create(struct dm_transaction_manager *tm,
+				       dm_block_t nr_blocks);
+
+struct dm_space_map *dm_sm_disk_open(struct dm_transaction_manager *tm,
+				     void *root, size_t len);
+
+#endif /* _LINUX_DM_SPACE_MAP_DISK_H */
diff --git a/drivers/md/persistent-data/dm-space-map-metadata.c b/drivers/md/persistent-data/dm-space-map-metadata.c
new file mode 100644
index 00000000000..786b689bdfc
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-metadata.c
@@ -0,0 +1,794 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#include "dm-space-map.h"
+#include "dm-space-map-common.h"
+#include "dm-space-map-metadata.h"
+
+#include <linux/list.h>
+#include <linux/slab.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "space map metadata"
+
+/*----------------------------------------------------------------*/
+
+/*
+ * An edge triggered threshold.
+ */
+struct threshold {
+	bool threshold_set;
+	bool value_set;
+	dm_block_t threshold;
+	dm_block_t current_value;
+	dm_sm_threshold_fn fn;
+	void *context;
+};
+
+static void threshold_init(struct threshold *t)
+{
+	t->threshold_set = false;
+	t->value_set = false;
+}
+
+static void set_threshold(struct threshold *t, dm_block_t value,
+			  dm_sm_threshold_fn fn, void *context)
+{
+	t->threshold_set = true;
+	t->threshold = value;
+	t->fn = fn;
+	t->context = context;
+}
+
+static bool below_threshold(struct threshold *t, dm_block_t value)
+{
+	return t->threshold_set && value <= t->threshold;
+}
+
+static bool threshold_already_triggered(struct threshold *t)
+{
+	return t->value_set && below_threshold(t, t->current_value);
+}
+
+static void check_threshold(struct threshold *t, dm_block_t value)
+{
+	if (below_threshold(t, value) &&
+	    !threshold_already_triggered(t))
+		t->fn(t->context);
+
+	t->value_set = true;
+	t->current_value = value;
+}
+
+/*----------------------------------------------------------------*/
+
+/*
+ * Space map interface.
+ *
+ * The low level disk format is written using the standard btree and
+ * transaction manager.  This means that performing disk operations may
+ * cause us to recurse into the space map in order to allocate new blocks.
+ * For this reason we have a pool of pre-allocated blocks large enough to
+ * service any metadata_ll_disk operation.
+ */
+
+/*
+ * FIXME: we should calculate this based on the size of the device.
+ * Only the metadata space map needs this functionality.
+ */
+#define MAX_RECURSIVE_ALLOCATIONS 1024
+
+enum block_op_type {
+	BOP_INC,
+	BOP_DEC
+};
+
+struct block_op {
+	enum block_op_type type;
+	dm_block_t block;
+};
+
+struct bop_ring_buffer {
+	unsigned begin;
+	unsigned end;
+	struct block_op bops[MAX_RECURSIVE_ALLOCATIONS + 1];
+};
+
+static void brb_init(struct bop_ring_buffer *brb)
+{
+	brb->begin = 0;
+	brb->end = 0;
+}
+
+static bool brb_empty(struct bop_ring_buffer *brb)
+{
+	return brb->begin == brb->end;
+}
+
+static unsigned brb_next(struct bop_ring_buffer *brb, unsigned old)
+{
+	unsigned r = old + 1;
+	return (r >= (sizeof(brb->bops) / sizeof(*brb->bops))) ? 0 : r;
+}
+
+static int brb_push(struct bop_ring_buffer *brb,
+		    enum block_op_type type, dm_block_t b)
+{
+	struct block_op *bop;
+	unsigned next = brb_next(brb, brb->end);
+
+	/*
+	 * We don't allow the last bop to be filled, this way we can
+	 * differentiate between full and empty.
+	 */
+	if (next == brb->begin)
+		return -ENOMEM;
+
+	bop = brb->bops + brb->end;
+	bop->type = type;
+	bop->block = b;
+
+	brb->end = next;
+
+	return 0;
+}
+
+static int brb_pop(struct bop_ring_buffer *brb, struct block_op *result)
+{
+	struct block_op *bop;
+
+	if (brb_empty(brb))
+		return -ENODATA;
+
+	bop = brb->bops + brb->begin;
+	result->type = bop->type;
+	result->block = bop->block;
+
+	brb->begin = brb_next(brb, brb->begin);
+
+	return 0;
+}
+
+/*----------------------------------------------------------------*/
+
+struct sm_metadata {
+	struct dm_space_map sm;
+
+	struct ll_disk ll;
+	struct ll_disk old_ll;
+
+	dm_block_t begin;
+
+	unsigned recursion_count;
+	unsigned allocated_this_transaction;
+	struct bop_ring_buffer uncommitted;
+
+	struct threshold threshold;
+};
+
+static int add_bop(struct sm_metadata *smm, enum block_op_type type, dm_block_t b)
+{
+	int r = brb_push(&smm->uncommitted, type, b);
+
+	if (r) {
+		DMERR("too many recursive allocations");
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static int commit_bop(struct sm_metadata *smm, struct block_op *op)
+{
+	int r = 0;
+	enum allocation_event ev;
+
+	switch (op->type) {
+	case BOP_INC:
+		r = sm_ll_inc(&smm->ll, op->block, &ev);
+		break;
+
+	case BOP_DEC:
+		r = sm_ll_dec(&smm->ll, op->block, &ev);
+		break;
+	}
+
+	return r;
+}
+
+static void in(struct sm_metadata *smm)
+{
+	smm->recursion_count++;
+}
+
+static int out(struct sm_metadata *smm)
+{
+	int r = 0;
+
+	/*
+	 * If we're not recursing then very bad things are happening.
+	 */
+	if (!smm->recursion_count) {
+		DMERR("lost track of recursion depth");
+		return -ENOMEM;
+	}
+
+	if (smm->recursion_count == 1) {
+		while (!brb_empty(&smm->uncommitted)) {
+			struct block_op bop;
+
+			r = brb_pop(&smm->uncommitted, &bop);
+			if (r) {
+				DMERR("bug in bop ring buffer");
+				break;
+			}
+
+			r = commit_bop(smm, &bop);
+			if (r)
+				break;
+		}
+	}
+
+	smm->recursion_count--;
+
+	return r;
+}
+
+/*
+ * When using the out() function above, we often want to combine an error
+ * code for the operation run in the recursive context with that from
+ * out().
+ */
+static int combine_errors(int r1, int r2)
+{
+	return r1 ? r1 : r2;
+}
+
+static int recursing(struct sm_metadata *smm)
+{
+	return smm->recursion_count;
+}
+
+static void sm_metadata_destroy(struct dm_space_map *sm)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	kfree(smm);
+}
+
+static int sm_metadata_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	*count = smm->ll.nr_blocks;
+
+	return 0;
+}
+
+static int sm_metadata_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	*count = smm->old_ll.nr_blocks - smm->old_ll.nr_allocated -
+		 smm->allocated_this_transaction;
+
+	return 0;
+}
+
+static int sm_metadata_get_count(struct dm_space_map *sm, dm_block_t b,
+				 uint32_t *result)
+{
+	int r;
+	unsigned i;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+	unsigned adjustment = 0;
+
+	/*
+	 * We may have some uncommitted adjustments to add.  This list
+	 * should always be really short.
+	 */
+	for (i = smm->uncommitted.begin;
+	     i != smm->uncommitted.end;
+	     i = brb_next(&smm->uncommitted, i)) {
+		struct block_op *op = smm->uncommitted.bops + i;
+
+		if (op->block != b)
+			continue;
+
+		switch (op->type) {
+		case BOP_INC:
+			adjustment++;
+			break;
+
+		case BOP_DEC:
+			adjustment--;
+			break;
+		}
+	}
+
+	r = sm_ll_lookup(&smm->ll, b, result);
+	if (r)
+		return r;
+
+	*result += adjustment;
+
+	return 0;
+}
+
+static int sm_metadata_count_is_more_than_one(struct dm_space_map *sm,
+					      dm_block_t b, int *result)
+{
+	int r, adjustment = 0;
+	unsigned i;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+	uint32_t rc;
+
+	/*
+	 * We may have some uncommitted adjustments to add.  This list
+	 * should always be really short.
+	 */
+	for (i = smm->uncommitted.begin;
+	     i != smm->uncommitted.end;
+	     i = brb_next(&smm->uncommitted, i)) {
+
+		struct block_op *op = smm->uncommitted.bops + i;
+
+		if (op->block != b)
+			continue;
+
+		switch (op->type) {
+		case BOP_INC:
+			adjustment++;
+			break;
+
+		case BOP_DEC:
+			adjustment--;
+			break;
+		}
+	}
+
+	if (adjustment > 1) {
+		*result = 1;
+		return 0;
+	}
+
+	r = sm_ll_lookup_bitmap(&smm->ll, b, &rc);
+	if (r)
+		return r;
+
+	if (rc == 3)
+		/*
+		 * We err on the side of caution, and always return true.
+		 */
+		*result = 1;
+	else
+		*result = rc + adjustment > 1;
+
+	return 0;
+}
+
+static int sm_metadata_set_count(struct dm_space_map *sm, dm_block_t b,
+				 uint32_t count)
+{
+	int r, r2;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	if (smm->recursion_count) {
+		DMERR("cannot recurse set_count()");
+		return -EINVAL;
+	}
+
+	in(smm);
+	r = sm_ll_insert(&smm->ll, b, count, &ev);
+	r2 = out(smm);
+
+	return combine_errors(r, r2);
+}
+
+static int sm_metadata_inc_block(struct dm_space_map *sm, dm_block_t b)
+{
+	int r, r2 = 0;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	if (recursing(smm))
+		r = add_bop(smm, BOP_INC, b);
+	else {
+		in(smm);
+		r = sm_ll_inc(&smm->ll, b, &ev);
+		r2 = out(smm);
+	}
+
+	return combine_errors(r, r2);
+}
+
+static int sm_metadata_dec_block(struct dm_space_map *sm, dm_block_t b)
+{
+	int r, r2 = 0;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	if (recursing(smm))
+		r = add_bop(smm, BOP_DEC, b);
+	else {
+		in(smm);
+		r = sm_ll_dec(&smm->ll, b, &ev);
+		r2 = out(smm);
+	}
+
+	return combine_errors(r, r2);
+}
+
+static int sm_metadata_new_block_(struct dm_space_map *sm, dm_block_t *b)
+{
+	int r, r2 = 0;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	r = sm_ll_find_free_block(&smm->old_ll, smm->begin, smm->old_ll.nr_blocks, b);
+	if (r)
+		return r;
+
+	smm->begin = *b + 1;
+
+	if (recursing(smm))
+		r = add_bop(smm, BOP_INC, *b);
+	else {
+		in(smm);
+		r = sm_ll_inc(&smm->ll, *b, &ev);
+		r2 = out(smm);
+	}
+
+	if (!r)
+		smm->allocated_this_transaction++;
+
+	return combine_errors(r, r2);
+}
+
+static int sm_metadata_new_block(struct dm_space_map *sm, dm_block_t *b)
+{
+	dm_block_t count;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	int r = sm_metadata_new_block_(sm, b);
+	if (r) {
+		DMERR_LIMIT("unable to allocate new metadata block");
+		return r;
+	}
+
+	r = sm_metadata_get_nr_free(sm, &count);
+	if (r) {
+		DMERR_LIMIT("couldn't get free block count");
+		return r;
+	}
+
+	check_threshold(&smm->threshold, count);
+
+	return r;
+}
+
+static int sm_metadata_commit(struct dm_space_map *sm)
+{
+	int r;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	r = sm_ll_commit(&smm->ll);
+	if (r)
+		return r;
+
+	memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
+	smm->begin = 0;
+	smm->allocated_this_transaction = 0;
+
+	return 0;
+}
+
+static int sm_metadata_register_threshold_callback(struct dm_space_map *sm,
+						   dm_block_t threshold,
+						   dm_sm_threshold_fn fn,
+						   void *context)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	set_threshold(&smm->threshold, threshold, fn, context);
+
+	return 0;
+}
+
+static int sm_metadata_root_size(struct dm_space_map *sm, size_t *result)
+{
+	*result = sizeof(struct disk_sm_root);
+
+	return 0;
+}
+
+static int sm_metadata_copy_root(struct dm_space_map *sm, void *where_le, size_t max)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+	struct disk_sm_root root_le;
+
+	root_le.nr_blocks = cpu_to_le64(smm->ll.nr_blocks);
+	root_le.nr_allocated = cpu_to_le64(smm->ll.nr_allocated);
+	root_le.bitmap_root = cpu_to_le64(smm->ll.bitmap_root);
+	root_le.ref_count_root = cpu_to_le64(smm->ll.ref_count_root);
+
+	if (max < sizeof(root_le))
+		return -ENOSPC;
+
+	memcpy(where_le, &root_le, sizeof(root_le));
+
+	return 0;
+}
+
+static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks);
+
+static struct dm_space_map ops = {
+	.destroy = sm_metadata_destroy,
+	.extend = sm_metadata_extend,
+	.get_nr_blocks = sm_metadata_get_nr_blocks,
+	.get_nr_free = sm_metadata_get_nr_free,
+	.get_count = sm_metadata_get_count,
+	.count_is_more_than_one = sm_metadata_count_is_more_than_one,
+	.set_count = sm_metadata_set_count,
+	.inc_block = sm_metadata_inc_block,
+	.dec_block = sm_metadata_dec_block,
+	.new_block = sm_metadata_new_block,
+	.commit = sm_metadata_commit,
+	.root_size = sm_metadata_root_size,
+	.copy_root = sm_metadata_copy_root,
+	.register_threshold_callback = sm_metadata_register_threshold_callback
+};
+
+/*----------------------------------------------------------------*/
+
+/*
+ * When a new space map is created that manages its own space.  We use
+ * this tiny bootstrap allocator.
+ */
+static void sm_bootstrap_destroy(struct dm_space_map *sm)
+{
+}
+
+static int sm_bootstrap_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
+{
+	DMERR("bootstrap doesn't support extend");
+
+	return -EINVAL;
+}
+
+static int sm_bootstrap_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	return smm->ll.nr_blocks;
+}
+
+static int sm_bootstrap_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	*count = smm->ll.nr_blocks - smm->begin;
+
+	return 0;
+}
+
+static int sm_bootstrap_get_count(struct dm_space_map *sm, dm_block_t b,
+				  uint32_t *result)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	return b < smm->begin ? 1 : 0;
+}
+
+static int sm_bootstrap_count_is_more_than_one(struct dm_space_map *sm,
+					       dm_block_t b, int *result)
+{
+	*result = 0;
+
+	return 0;
+}
+
+static int sm_bootstrap_set_count(struct dm_space_map *sm, dm_block_t b,
+				  uint32_t count)
+{
+	DMERR("bootstrap doesn't support set_count");
+
+	return -EINVAL;
+}
+
+static int sm_bootstrap_new_block(struct dm_space_map *sm, dm_block_t *b)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	/*
+	 * We know the entire device is unused.
+	 */
+	if (smm->begin == smm->ll.nr_blocks)
+		return -ENOSPC;
+
+	*b = smm->begin++;
+
+	return 0;
+}
+
+static int sm_bootstrap_inc_block(struct dm_space_map *sm, dm_block_t b)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	return add_bop(smm, BOP_INC, b);
+}
+
+static int sm_bootstrap_dec_block(struct dm_space_map *sm, dm_block_t b)
+{
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	return add_bop(smm, BOP_DEC, b);
+}
+
+static int sm_bootstrap_commit(struct dm_space_map *sm)
+{
+	return 0;
+}
+
+static int sm_bootstrap_root_size(struct dm_space_map *sm, size_t *result)
+{
+	DMERR("bootstrap doesn't support root_size");
+
+	return -EINVAL;
+}
+
+static int sm_bootstrap_copy_root(struct dm_space_map *sm, void *where,
+				  size_t max)
+{
+	DMERR("bootstrap doesn't support copy_root");
+
+	return -EINVAL;
+}
+
+static struct dm_space_map bootstrap_ops = {
+	.destroy = sm_bootstrap_destroy,
+	.extend = sm_bootstrap_extend,
+	.get_nr_blocks = sm_bootstrap_get_nr_blocks,
+	.get_nr_free = sm_bootstrap_get_nr_free,
+	.get_count = sm_bootstrap_get_count,
+	.count_is_more_than_one = sm_bootstrap_count_is_more_than_one,
+	.set_count = sm_bootstrap_set_count,
+	.inc_block = sm_bootstrap_inc_block,
+	.dec_block = sm_bootstrap_dec_block,
+	.new_block = sm_bootstrap_new_block,
+	.commit = sm_bootstrap_commit,
+	.root_size = sm_bootstrap_root_size,
+	.copy_root = sm_bootstrap_copy_root,
+	.register_threshold_callback = NULL
+};
+
+/*----------------------------------------------------------------*/
+
+static int sm_metadata_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
+{
+	int r, i;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+	dm_block_t old_len = smm->ll.nr_blocks;
+
+	/*
+	 * Flick into a mode where all blocks get allocated in the new area.
+	 */
+	smm->begin = old_len;
+	memcpy(sm, &bootstrap_ops, sizeof(*sm));
+
+	/*
+	 * Extend.
+	 */
+	r = sm_ll_extend(&smm->ll, extra_blocks);
+	if (r)
+		goto out;
+
+	/*
+	 * We repeatedly increment then commit until the commit doesn't
+	 * allocate any new blocks.
+	 */
+	do {
+		for (i = old_len; !r && i < smm->begin; i++) {
+			r = sm_ll_inc(&smm->ll, i, &ev);
+			if (r)
+				goto out;
+		}
+		old_len = smm->begin;
+
+		r = sm_ll_commit(&smm->ll);
+		if (r)
+			goto out;
+
+	} while (old_len != smm->begin);
+
+out:
+	/*
+	 * Switch back to normal behaviour.
+	 */
+	memcpy(sm, &ops, sizeof(*sm));
+	return r;
+}
+
+/*----------------------------------------------------------------*/
+
+struct dm_space_map *dm_sm_metadata_init(void)
+{
+	struct sm_metadata *smm;
+
+	smm = kmalloc(sizeof(*smm), GFP_KERNEL);
+	if (!smm)
+		return ERR_PTR(-ENOMEM);
+
+	memcpy(&smm->sm, &ops, sizeof(smm->sm));
+
+	return &smm->sm;
+}
+
+int dm_sm_metadata_create(struct dm_space_map *sm,
+			  struct dm_transaction_manager *tm,
+			  dm_block_t nr_blocks,
+			  dm_block_t superblock)
+{
+	int r;
+	dm_block_t i;
+	enum allocation_event ev;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	smm->begin = superblock + 1;
+	smm->recursion_count = 0;
+	smm->allocated_this_transaction = 0;
+	brb_init(&smm->uncommitted);
+	threshold_init(&smm->threshold);
+
+	memcpy(&smm->sm, &bootstrap_ops, sizeof(smm->sm));
+
+	r = sm_ll_new_metadata(&smm->ll, tm);
+	if (r)
+		return r;
+
+	if (nr_blocks > DM_SM_METADATA_MAX_BLOCKS)
+		nr_blocks = DM_SM_METADATA_MAX_BLOCKS;
+	r = sm_ll_extend(&smm->ll, nr_blocks);
+	if (r)
+		return r;
+
+	memcpy(&smm->sm, &ops, sizeof(smm->sm));
+
+	/*
+	 * Now we need to update the newly created data structures with the
+	 * allocated blocks that they were built from.
+	 */
+	for (i = superblock; !r && i < smm->begin; i++)
+		r = sm_ll_inc(&smm->ll, i, &ev);
+
+	if (r)
+		return r;
+
+	return sm_metadata_commit(sm);
+}
+
+int dm_sm_metadata_open(struct dm_space_map *sm,
+			struct dm_transaction_manager *tm,
+			void *root_le, size_t len)
+{
+	int r;
+	struct sm_metadata *smm = container_of(sm, struct sm_metadata, sm);
+
+	r = sm_ll_open_metadata(&smm->ll, tm, root_le, len);
+	if (r)
+		return r;
+
+	smm->begin = 0;
+	smm->recursion_count = 0;
+	smm->allocated_this_transaction = 0;
+	brb_init(&smm->uncommitted);
+	threshold_init(&smm->threshold);
+
+	memcpy(&smm->old_ll, &smm->ll, sizeof(smm->old_ll));
+	return 0;
+}
diff --git a/drivers/md/persistent-data/dm-space-map-metadata.h b/drivers/md/persistent-data/dm-space-map-metadata.h
new file mode 100644
index 00000000000..64df923974d
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map-metadata.h
@@ -0,0 +1,44 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef DM_SPACE_MAP_METADATA_H
+#define DM_SPACE_MAP_METADATA_H
+
+#include "dm-transaction-manager.h"
+
+#define DM_SM_METADATA_BLOCK_SIZE (4096 >> SECTOR_SHIFT)
+
+/*
+ * The metadata device is currently limited in size.
+ *
+ * We have one block of index, which can hold 255 index entries.  Each
+ * index entry contains allocation info about ~16k metadata blocks.
+ */
+#define DM_SM_METADATA_MAX_BLOCKS (255 * ((1 << 14) - 64))
+#define DM_SM_METADATA_MAX_SECTORS (DM_SM_METADATA_MAX_BLOCKS * DM_SM_METADATA_BLOCK_SIZE)
+
+/*
+ * Unfortunately we have to use two-phase construction due to the cycle
+ * between the tm and sm.
+ */
+struct dm_space_map *dm_sm_metadata_init(void);
+
+/*
+ * Create a fresh space map.
+ */
+int dm_sm_metadata_create(struct dm_space_map *sm,
+			  struct dm_transaction_manager *tm,
+			  dm_block_t nr_blocks,
+			  dm_block_t superblock);
+
+/*
+ * Open from a previously-recorded root.
+ */
+int dm_sm_metadata_open(struct dm_space_map *sm,
+			struct dm_transaction_manager *tm,
+			void *root_le, size_t len);
+
+#endif	/* DM_SPACE_MAP_METADATA_H */
diff --git a/drivers/md/persistent-data/dm-space-map.h b/drivers/md/persistent-data/dm-space-map.h
new file mode 100644
index 00000000000..3e6d1153b7c
--- /dev/null
+++ b/drivers/md/persistent-data/dm-space-map.h
@@ -0,0 +1,157 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_SPACE_MAP_H
+#define _LINUX_DM_SPACE_MAP_H
+
+#include "dm-block-manager.h"
+
+typedef void (*dm_sm_threshold_fn)(void *context);
+
+/*
+ * struct dm_space_map keeps a record of how many times each block in a device
+ * is referenced.  It needs to be fixed on disk as part of the transaction.
+ */
+struct dm_space_map {
+	void (*destroy)(struct dm_space_map *sm);
+
+	/*
+	 * You must commit before allocating the newly added space.
+	 */
+	int (*extend)(struct dm_space_map *sm, dm_block_t extra_blocks);
+
+	/*
+	 * Extensions do not appear in this count until after commit has
+	 * been called.
+	 */
+	int (*get_nr_blocks)(struct dm_space_map *sm, dm_block_t *count);
+
+	/*
+	 * Space maps must never allocate a block from the previous
+	 * transaction, in case we need to rollback.  This complicates the
+	 * semantics of get_nr_free(), it should return the number of blocks
+	 * that are available for allocation _now_.  For instance you may
+	 * have blocks with a zero reference count that will not be
+	 * available for allocation until after the next commit.
+	 */
+	int (*get_nr_free)(struct dm_space_map *sm, dm_block_t *count);
+
+	int (*get_count)(struct dm_space_map *sm, dm_block_t b, uint32_t *result);
+	int (*count_is_more_than_one)(struct dm_space_map *sm, dm_block_t b,
+				      int *result);
+	int (*set_count)(struct dm_space_map *sm, dm_block_t b, uint32_t count);
+
+	int (*commit)(struct dm_space_map *sm);
+
+	int (*inc_block)(struct dm_space_map *sm, dm_block_t b);
+	int (*dec_block)(struct dm_space_map *sm, dm_block_t b);
+
+	/*
+	 * new_block will increment the returned block.
+	 */
+	int (*new_block)(struct dm_space_map *sm, dm_block_t *b);
+
+	/*
+	 * The root contains all the information needed to fix the space map.
+	 * Generally this info is small, so squirrel it away in a disk block
+	 * along with other info.
+	 */
+	int (*root_size)(struct dm_space_map *sm, size_t *result);
+	int (*copy_root)(struct dm_space_map *sm, void *copy_to_here_le, size_t len);
+
+	/*
+	 * You can register one threshold callback which is edge-triggered
+	 * when the free space in the space map drops below the threshold.
+	 */
+	int (*register_threshold_callback)(struct dm_space_map *sm,
+					   dm_block_t threshold,
+					   dm_sm_threshold_fn fn,
+					   void *context);
+};
+
+/*----------------------------------------------------------------*/
+
+static inline void dm_sm_destroy(struct dm_space_map *sm)
+{
+	sm->destroy(sm);
+}
+
+static inline int dm_sm_extend(struct dm_space_map *sm, dm_block_t extra_blocks)
+{
+	return sm->extend(sm, extra_blocks);
+}
+
+static inline int dm_sm_get_nr_blocks(struct dm_space_map *sm, dm_block_t *count)
+{
+	return sm->get_nr_blocks(sm, count);
+}
+
+static inline int dm_sm_get_nr_free(struct dm_space_map *sm, dm_block_t *count)
+{
+	return sm->get_nr_free(sm, count);
+}
+
+static inline int dm_sm_get_count(struct dm_space_map *sm, dm_block_t b,
+				  uint32_t *result)
+{
+	return sm->get_count(sm, b, result);
+}
+
+static inline int dm_sm_count_is_more_than_one(struct dm_space_map *sm,
+					       dm_block_t b, int *result)
+{
+	return sm->count_is_more_than_one(sm, b, result);
+}
+
+static inline int dm_sm_set_count(struct dm_space_map *sm, dm_block_t b,
+				  uint32_t count)
+{
+	return sm->set_count(sm, b, count);
+}
+
+static inline int dm_sm_commit(struct dm_space_map *sm)
+{
+	return sm->commit(sm);
+}
+
+static inline int dm_sm_inc_block(struct dm_space_map *sm, dm_block_t b)
+{
+	return sm->inc_block(sm, b);
+}
+
+static inline int dm_sm_dec_block(struct dm_space_map *sm, dm_block_t b)
+{
+	return sm->dec_block(sm, b);
+}
+
+static inline int dm_sm_new_block(struct dm_space_map *sm, dm_block_t *b)
+{
+	return sm->new_block(sm, b);
+}
+
+static inline int dm_sm_root_size(struct dm_space_map *sm, size_t *result)
+{
+	return sm->root_size(sm, result);
+}
+
+static inline int dm_sm_copy_root(struct dm_space_map *sm, void *copy_to_here_le, size_t len)
+{
+	return sm->copy_root(sm, copy_to_here_le, len);
+}
+
+static inline int dm_sm_register_threshold_callback(struct dm_space_map *sm,
+						    dm_block_t threshold,
+						    dm_sm_threshold_fn fn,
+						    void *context)
+{
+	if (sm->register_threshold_callback)
+		return sm->register_threshold_callback(sm, threshold, fn, context);
+
+	return -EINVAL;
+}
+
+
+#endif	/* _LINUX_DM_SPACE_MAP_H */
diff --git a/drivers/md/persistent-data/dm-transaction-manager.c b/drivers/md/persistent-data/dm-transaction-manager.c
new file mode 100644
index 00000000000..3bc30a0ae3d
--- /dev/null
+++ b/drivers/md/persistent-data/dm-transaction-manager.c
@@ -0,0 +1,382 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+#include "dm-transaction-manager.h"
+#include "dm-space-map.h"
+#include "dm-space-map-disk.h"
+#include "dm-space-map-metadata.h"
+#include "dm-persistent-data-internal.h"
+
+#include <linux/export.h>
+#include <linux/slab.h>
+#include <linux/device-mapper.h>
+
+#define DM_MSG_PREFIX "transaction manager"
+
+/*----------------------------------------------------------------*/
+
+struct shadow_info {
+	struct hlist_node hlist;
+	dm_block_t where;
+};
+
+/*
+ * It would be nice if we scaled with the size of transaction.
+ */
+#define DM_HASH_SIZE 256
+#define DM_HASH_MASK (DM_HASH_SIZE - 1)
+
+struct dm_transaction_manager {
+	int is_clone;
+	struct dm_transaction_manager *real;
+
+	struct dm_block_manager *bm;
+	struct dm_space_map *sm;
+
+	spinlock_t lock;
+	struct hlist_head buckets[DM_HASH_SIZE];
+};
+
+/*----------------------------------------------------------------*/
+
+static int is_shadow(struct dm_transaction_manager *tm, dm_block_t b)
+{
+	int r = 0;
+	unsigned bucket = dm_hash_block(b, DM_HASH_MASK);
+	struct shadow_info *si;
+
+	spin_lock(&tm->lock);
+	hlist_for_each_entry(si, tm->buckets + bucket, hlist)
+		if (si->where == b) {
+			r = 1;
+			break;
+		}
+	spin_unlock(&tm->lock);
+
+	return r;
+}
+
+/*
+ * This can silently fail if there's no memory.  We're ok with this since
+ * creating redundant shadows causes no harm.
+ */
+static void insert_shadow(struct dm_transaction_manager *tm, dm_block_t b)
+{
+	unsigned bucket;
+	struct shadow_info *si;
+
+	si = kmalloc(sizeof(*si), GFP_NOIO);
+	if (si) {
+		si->where = b;
+		bucket = dm_hash_block(b, DM_HASH_MASK);
+		spin_lock(&tm->lock);
+		hlist_add_head(&si->hlist, tm->buckets + bucket);
+		spin_unlock(&tm->lock);
+	}
+}
+
+static void wipe_shadow_table(struct dm_transaction_manager *tm)
+{
+	struct shadow_info *si;
+	struct hlist_node *tmp;
+	struct hlist_head *bucket;
+	int i;
+
+	spin_lock(&tm->lock);
+	for (i = 0; i < DM_HASH_SIZE; i++) {
+		bucket = tm->buckets + i;
+		hlist_for_each_entry_safe(si, tmp, bucket, hlist)
+			kfree(si);
+
+		INIT_HLIST_HEAD(bucket);
+	}
+
+	spin_unlock(&tm->lock);
+}
+
+/*----------------------------------------------------------------*/
+
+static struct dm_transaction_manager *dm_tm_create(struct dm_block_manager *bm,
+						   struct dm_space_map *sm)
+{
+	int i;
+	struct dm_transaction_manager *tm;
+
+	tm = kmalloc(sizeof(*tm), GFP_KERNEL);
+	if (!tm)
+		return ERR_PTR(-ENOMEM);
+
+	tm->is_clone = 0;
+	tm->real = NULL;
+	tm->bm = bm;
+	tm->sm = sm;
+
+	spin_lock_init(&tm->lock);
+	for (i = 0; i < DM_HASH_SIZE; i++)
+		INIT_HLIST_HEAD(tm->buckets + i);
+
+	return tm;
+}
+
+struct dm_transaction_manager *dm_tm_create_non_blocking_clone(struct dm_transaction_manager *real)
+{
+	struct dm_transaction_manager *tm;
+
+	tm = kmalloc(sizeof(*tm), GFP_KERNEL);
+	if (tm) {
+		tm->is_clone = 1;
+		tm->real = real;
+	}
+
+	return tm;
+}
+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);
+
+int dm_tm_pre_commit(struct dm_transaction_manager *tm)
+{
+	int r;
+
+	if (tm->is_clone)
+		return -EWOULDBLOCK;
+
+	r = dm_sm_commit(tm->sm);
+	if (r < 0)
+		return r;
+
+	return dm_bm_flush(tm->bm);
+}
+EXPORT_SYMBOL_GPL(dm_tm_pre_commit);
+
+int dm_tm_commit(struct dm_transaction_manager *tm, struct dm_block *root)
+{
+	if (tm->is_clone)
+		return -EWOULDBLOCK;
+
+	wipe_shadow_table(tm);
+	dm_bm_unlock(root);
+
+	return dm_bm_flush(tm->bm);
+}
+EXPORT_SYMBOL_GPL(dm_tm_commit);
+
+int dm_tm_new_block(struct dm_transaction_manager *tm,
+		    struct dm_block_validator *v,
+		    struct dm_block **result)
+{
+	int r;
+	dm_block_t new_block;
+
+	if (tm->is_clone)
+		return -EWOULDBLOCK;
+
+	r = dm_sm_new_block(tm->sm, &new_block);
+	if (r < 0)
+		return r;
+
+	r = dm_bm_write_lock_zero(tm->bm, new_block, v, result);
+	if (r < 0) {
+		dm_sm_dec_block(tm->sm, new_block);
+		return r;
+	}
+
+	/*
+	 * New blocks count as shadows in that they don't need to be
+	 * shadowed again.
+	 */
+	insert_shadow(tm, new_block);
+
+	return 0;
+}
+
+static int __shadow_block(struct dm_transaction_manager *tm, dm_block_t orig,
+			  struct dm_block_validator *v,
+			  struct dm_block **result)
+{
+	int r;
+	dm_block_t new;
+	struct dm_block *orig_block;
+
+	r = dm_sm_new_block(tm->sm, &new);
+	if (r < 0)
+		return r;
+
+	r = dm_sm_dec_block(tm->sm, orig);
+	if (r < 0)
+		return r;
+
+	r = dm_bm_read_lock(tm->bm, orig, v, &orig_block);
+	if (r < 0)
+		return r;
+
+	/*
+	 * 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;
+	}
+
+	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,
+		       struct dm_block_validator *v, struct dm_block **result,
+		       int *inc_children)
+{
+	int r;
+
+	if (tm->is_clone)
+		return -EWOULDBLOCK;
+
+	r = dm_sm_count_is_more_than_one(tm->sm, orig, inc_children);
+	if (r < 0)
+		return r;
+
+	if (is_shadow(tm, orig) && !*inc_children)
+		return dm_bm_write_lock(tm->bm, orig, v, result);
+
+	r = __shadow_block(tm, orig, v, result);
+	if (r < 0)
+		return r;
+	insert_shadow(tm, dm_block_location(*result));
+
+	return r;
+}
+EXPORT_SYMBOL_GPL(dm_tm_shadow_block);
+
+int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b,
+		    struct dm_block_validator *v,
+		    struct dm_block **blk)
+{
+	if (tm->is_clone)
+		return dm_bm_read_try_lock(tm->real->bm, b, v, blk);
+
+	return dm_bm_read_lock(tm->bm, b, v, blk);
+}
+EXPORT_SYMBOL_GPL(dm_tm_read_lock);
+
+int dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b)
+{
+	return dm_bm_unlock(b);
+}
+EXPORT_SYMBOL_GPL(dm_tm_unlock);
+
+void dm_tm_inc(struct dm_transaction_manager *tm, dm_block_t b)
+{
+	/*
+	 * The non-blocking clone doesn't support this.
+	 */
+	BUG_ON(tm->is_clone);
+
+	dm_sm_inc_block(tm->sm, b);
+}
+EXPORT_SYMBOL_GPL(dm_tm_inc);
+
+void dm_tm_dec(struct dm_transaction_manager *tm, dm_block_t b)
+{
+	/*
+	 * The non-blocking clone doesn't support this.
+	 */
+	BUG_ON(tm->is_clone);
+
+	dm_sm_dec_block(tm->sm, b);
+}
+EXPORT_SYMBOL_GPL(dm_tm_dec);
+
+int dm_tm_ref(struct dm_transaction_manager *tm, dm_block_t b,
+	      uint32_t *result)
+{
+	if (tm->is_clone)
+		return -EWOULDBLOCK;
+
+	return dm_sm_get_count(tm->sm, b, result);
+}
+
+struct dm_block_manager *dm_tm_get_bm(struct dm_transaction_manager *tm)
+{
+	return tm->bm;
+}
+
+/*----------------------------------------------------------------*/
+
+static int dm_tm_create_internal(struct dm_block_manager *bm,
+				 dm_block_t sb_location,
+				 struct dm_transaction_manager **tm,
+				 struct dm_space_map **sm,
+				 int create,
+				 void *sm_root, size_t sm_len)
+{
+	int r;
+
+	*sm = dm_sm_metadata_init();
+	if (IS_ERR(*sm))
+		return PTR_ERR(*sm);
+
+	*tm = dm_tm_create(bm, *sm);
+	if (IS_ERR(*tm)) {
+		dm_sm_destroy(*sm);
+		return PTR_ERR(*tm);
+	}
+
+	if (create) {
+		r = dm_sm_metadata_create(*sm, *tm, dm_bm_nr_blocks(bm),
+					  sb_location);
+		if (r) {
+			DMERR("couldn't create metadata space map");
+			goto bad;
+		}
+
+	} else {
+		r = dm_sm_metadata_open(*sm, *tm, sm_root, sm_len);
+		if (r) {
+			DMERR("couldn't open metadata space map");
+			goto bad;
+		}
+	}
+
+	return 0;
+
+bad:
+	dm_tm_destroy(*tm);
+	dm_sm_destroy(*sm);
+	return r;
+}
+
+int dm_tm_create_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
+			 struct dm_transaction_manager **tm,
+			 struct dm_space_map **sm)
+{
+	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,
+		       void *sm_root, size_t root_len,
+		       struct dm_transaction_manager **tm,
+		       struct dm_space_map **sm)
+{
+	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
new file mode 100644
index 00000000000..2772ed2a781
--- /dev/null
+++ b/drivers/md/persistent-data/dm-transaction-manager.h
@@ -0,0 +1,130 @@
+/*
+ * Copyright (C) 2011 Red Hat, Inc.
+ *
+ * This file is released under the GPL.
+ */
+
+#ifndef _LINUX_DM_TRANSACTION_MANAGER_H
+#define _LINUX_DM_TRANSACTION_MANAGER_H
+
+#include "dm-block-manager.h"
+
+struct dm_transaction_manager;
+struct dm_space_map;
+
+/*----------------------------------------------------------------*/
+
+/*
+ * This manages the scope of a transaction.  It also enforces immutability
+ * of the on-disk data structures by limiting access to writeable blocks.
+ *
+ * Clients should not fiddle with the block manager directly.
+ */
+
+void dm_tm_destroy(struct dm_transaction_manager *tm);
+
+/*
+ * The non-blocking version of a transaction manager is intended for use in
+ * fast path code that needs to do lookups e.g. a dm mapping function.
+ * You create the non-blocking variant from a normal tm.  The interface is
+ * the same, except that most functions will just return -EWOULDBLOCK.
+ * Methods that return void yet may block should not be called on a clone
+ * viz. dm_tm_inc, dm_tm_dec.  Call dm_tm_destroy() as you would with a normal
+ * tm when you've finished with it.  You may not destroy the original prior
+ * to clones.
+ */
+struct dm_transaction_manager *dm_tm_create_non_blocking_clone(struct dm_transaction_manager *real);
+
+/*
+ * We use a 2-phase commit here.
+ *
+ * i) Make all changes for the transaction *except* for the superblock.
+ * Then call dm_tm_pre_commit() to flush them to disk.
+ *
+ * ii) Lock your superblock.  Update.  Then call dm_tm_commit() which will
+ * unlock the superblock and flush it.  No other blocks should be updated
+ * during this period.  Care should be taken to never unlock a partially
+ * updated superblock; perform any operations that could fail *before* you
+ * take the superblock lock.
+ */
+int dm_tm_pre_commit(struct dm_transaction_manager *tm);
+int dm_tm_commit(struct dm_transaction_manager *tm, struct dm_block *superblock);
+
+/*
+ * These methods are the only way to get hold of a writeable block.
+ */
+
+/*
+ * dm_tm_new_block() is pretty self-explanatory.  Make sure you do actually
+ * write to the whole of @data before you unlock, otherwise you could get
+ * a data leak.  (The other option is for tm_new_block() to zero new blocks
+ * before handing them out, which will be redundant in most, if not all,
+ * cases).
+ * Zeroes the new block and returns with write lock held.
+ */
+int dm_tm_new_block(struct dm_transaction_manager *tm,
+		    struct dm_block_validator *v,
+		    struct dm_block **result);
+
+/*
+ * dm_tm_shadow_block() allocates a new block and copies the data from @orig
+ * to it.  It then decrements the reference count on original block.  Use
+ * this to update the contents of a block in a data structure, don't
+ * confuse this with a clone - you shouldn't access the orig block after
+ * this operation.  Because the tm knows the scope of the transaction it
+ * can optimise requests for a shadow of a shadow to a no-op.  Don't forget
+ * to unlock when you've finished with the shadow.
+ *
+ * The @inc_children flag is used to tell the caller whether it needs to
+ * adjust reference counts for children.  (Data in the block may refer to
+ * other blocks.)
+ *
+ * Shadowing implicitly drops a reference on @orig so you must not have
+ * it locked when you call this.
+ */
+int dm_tm_shadow_block(struct dm_transaction_manager *tm, dm_block_t orig,
+		       struct dm_block_validator *v,
+		       struct dm_block **result, int *inc_children);
+
+/*
+ * Read access.  You can lock any block you want.  If there's a write lock
+ * on it outstanding then it'll block.
+ */
+int dm_tm_read_lock(struct dm_transaction_manager *tm, dm_block_t b,
+		    struct dm_block_validator *v,
+		    struct dm_block **result);
+
+int dm_tm_unlock(struct dm_transaction_manager *tm, struct dm_block *b);
+
+/*
+ * Functions for altering the reference count of a block directly.
+ */
+void dm_tm_inc(struct dm_transaction_manager *tm, dm_block_t b);
+
+void dm_tm_dec(struct dm_transaction_manager *tm, dm_block_t b);
+
+int dm_tm_ref(struct dm_transaction_manager *tm, dm_block_t b,
+	      uint32_t *result);
+
+struct dm_block_manager *dm_tm_get_bm(struct dm_transaction_manager *tm);
+
+/*
+ * A little utility that ties the knot by producing a transaction manager
+ * that has a space map managed by the transaction manager...
+ *
+ * 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_transaction_manager **tm,
+			 struct dm_space_map **sm);
+
+int dm_tm_open_with_sm(struct dm_block_manager *bm, dm_block_t sb_location,
+		       void *sm_root, size_t root_len,
+		       struct dm_transaction_manager **tm,
+		       struct dm_space_map **sm);
+
+#endif	/* _LINUX_DM_TRANSACTION_MANAGER_H */
diff --git a/drivers/md/raid0.c b/drivers/md/raid0.c
index 18361063566..407a99e46f6 100644
--- a/drivers/md/raid0.c
+++ b/drivers/md/raid0.c
@@ -18,35 +18,26 @@
    Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  
 */
 
+#include <linux/blkdev.h>
+#include <linux/seq_file.h>
 #include <linux/module.h>
-#include <linux/raid/raid0.h>
-
-#define MAJOR_NR MD_MAJOR
-#define MD_DRIVER
-#define MD_PERSONALITY
-
-static void raid0_unplug(struct request_queue *q)
-{
-	mddev_t *mddev = q->queuedata;
-	raid0_conf_t *conf = mddev_to_conf(mddev);
-	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
-	int i;
-
-	for (i=0; i<mddev->raid_disks; i++) {
-		struct request_queue *r_queue = bdev_get_queue(devlist[i]->bdev);
-
-		blk_unplug(r_queue);
-	}
-}
+#include <linux/slab.h>
+#include "md.h"
+#include "raid0.h"
+#include "raid5.h"
 
 static int raid0_congested(void *data, int bits)
 {
-	mddev_t *mddev = data;
-	raid0_conf_t *conf = mddev_to_conf(mddev);
-	mdk_rdev_t **devlist = conf->strip_zone[0].dev;
+	struct mddev *mddev = data;
+	struct r0conf *conf = mddev->private;
+	struct md_rdev **devlist = conf->devlist;
+	int raid_disks = conf->strip_zone[0].nb_dev;
 	int i, ret = 0;
 
-	for (i = 0; i < mddev->raid_disks && !ret ; i++) {
+	if (mddev_congested(mddev, bits))
+		return 1;
+
+	for (i = 0; i < raid_disks && !ret ; i++) {
 		struct request_queue *q = bdev_get_queue(devlist[i]->bdev);
 
 		ret |= bdi_congested(&q->backing_dev_info, bits);
@@ -54,68 +45,110 @@ static int raid0_congested(void *data, int bits)
 	return ret;
 }
 
+/*
+ * inform the user of the raid configuration
+*/
+static void dump_zones(struct mddev *mddev)
+{
+	int j, k;
+	sector_t zone_size = 0;
+	sector_t zone_start = 0;
+	char b[BDEVNAME_SIZE];
+	struct r0conf *conf = mddev->private;
+	int raid_disks = conf->strip_zone[0].nb_dev;
+	printk(KERN_INFO "md: RAID0 configuration for %s - %d zone%s\n",
+	       mdname(mddev),
+	       conf->nr_strip_zones, conf->nr_strip_zones==1?"":"s");
+	for (j = 0; j < conf->nr_strip_zones; j++) {
+		printk(KERN_INFO "md: zone%d=[", j);
+		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
+			printk(KERN_CONT "%s%s", k?"/":"",
+			bdevname(conf->devlist[j*raid_disks
+						+ k]->bdev, b));
+		printk(KERN_CONT "]\n");
+
+		zone_size  = conf->strip_zone[j].zone_end - zone_start;
+		printk(KERN_INFO "      zone-offset=%10lluKB, "
+				"device-offset=%10lluKB, size=%10lluKB\n",
+			(unsigned long long)zone_start>>1,
+			(unsigned long long)conf->strip_zone[j].dev_start>>1,
+			(unsigned long long)zone_size>>1);
+		zone_start = conf->strip_zone[j].zone_end;
+	}
+	printk(KERN_INFO "\n");
+}
 
-static int create_strip_zones (mddev_t *mddev)
+static int create_strip_zones(struct mddev *mddev, struct r0conf **private_conf)
 {
-	int i, c, j;
-	sector_t current_offset, curr_zone_offset;
-	sector_t min_spacing;
-	raid0_conf_t *conf = mddev_to_conf(mddev);
-	mdk_rdev_t *smallest, *rdev1, *rdev2, *rdev;
-	struct list_head *tmp1, *tmp2;
+	int i, c, err;
+	sector_t curr_zone_end, sectors;
+	struct md_rdev *smallest, *rdev1, *rdev2, *rdev, **dev;
 	struct strip_zone *zone;
 	int cnt;
 	char b[BDEVNAME_SIZE];
- 
-	/*
-	 * The number of 'same size groups'
-	 */
-	conf->nr_strip_zones = 0;
- 
-	rdev_for_each(rdev1, tmp1, mddev) {
-		printk("raid0: looking at %s\n",
-			bdevname(rdev1->bdev,b));
+	char b2[BDEVNAME_SIZE];
+	struct r0conf *conf = kzalloc(sizeof(*conf), GFP_KERNEL);
+	bool discard_supported = false;
+
+	if (!conf)
+		return -ENOMEM;
+	rdev_for_each(rdev1, mddev) {
+		pr_debug("md/raid0:%s: looking at %s\n",
+			 mdname(mddev),
+			 bdevname(rdev1->bdev, b));
 		c = 0;
-		rdev_for_each(rdev2, tmp2, mddev) {
-			printk("raid0:   comparing %s(%llu)",
-			       bdevname(rdev1->bdev,b),
-			       (unsigned long long)rdev1->size);
-			printk(" with %s(%llu)\n",
-			       bdevname(rdev2->bdev,b),
-			       (unsigned long long)rdev2->size);
+
+		/* round size to chunk_size */
+		sectors = rdev1->sectors;
+		sector_div(sectors, mddev->chunk_sectors);
+		rdev1->sectors = sectors * mddev->chunk_sectors;
+
+		rdev_for_each(rdev2, mddev) {
+			pr_debug("md/raid0:%s:   comparing %s(%llu)"
+				 " with %s(%llu)\n",
+				 mdname(mddev),
+				 bdevname(rdev1->bdev,b),
+				 (unsigned long long)rdev1->sectors,
+				 bdevname(rdev2->bdev,b2),
+				 (unsigned long long)rdev2->sectors);
 			if (rdev2 == rdev1) {
-				printk("raid0:   END\n");
+				pr_debug("md/raid0:%s:   END\n",
+					 mdname(mddev));
 				break;
 			}
-			if (rdev2->size == rdev1->size)
-			{
+			if (rdev2->sectors == rdev1->sectors) {
 				/*
 				 * Not unique, don't count it as a new
 				 * group
 				 */
-				printk("raid0:   EQUAL\n");
+				pr_debug("md/raid0:%s:   EQUAL\n",
+					 mdname(mddev));
 				c = 1;
 				break;
 			}
-			printk("raid0:   NOT EQUAL\n");
+			pr_debug("md/raid0:%s:   NOT EQUAL\n",
+				 mdname(mddev));
 		}
 		if (!c) {
-			printk("raid0:   ==> UNIQUE\n");
+			pr_debug("md/raid0:%s:   ==> UNIQUE\n",
+				 mdname(mddev));
 			conf->nr_strip_zones++;
-			printk("raid0: %d zones\n", conf->nr_strip_zones);
+			pr_debug("md/raid0:%s: %d zones\n",
+				 mdname(mddev), conf->nr_strip_zones);
 		}
 	}
-	printk("raid0: FINAL %d zones\n", conf->nr_strip_zones);
-
+	pr_debug("md/raid0:%s: FINAL %d zones\n",
+		 mdname(mddev), conf->nr_strip_zones);
+	err = -ENOMEM;
 	conf->strip_zone = kzalloc(sizeof(struct strip_zone)*
 				conf->nr_strip_zones, GFP_KERNEL);
 	if (!conf->strip_zone)
-		return 1;
-	conf->devlist = kzalloc(sizeof(mdk_rdev_t*)*
+		goto abort;
+	conf->devlist = kzalloc(sizeof(struct md_rdev*)*
 				conf->nr_strip_zones*mddev->raid_disks,
 				GFP_KERNEL);
 	if (!conf->devlist)
-		return 1;
+		goto abort;
 
 	/* The first zone must contain all devices, so here we check that
 	 * there is a proper alignment of slots to devices and find them all
@@ -123,123 +156,207 @@ static int create_strip_zones (mddev_t *mddev)
 	zone = &conf->strip_zone[0];
 	cnt = 0;
 	smallest = NULL;
-	zone->dev = conf->devlist;
-	rdev_for_each(rdev1, tmp1, mddev) {
+	dev = conf->devlist;
+	err = -EINVAL;
+	rdev_for_each(rdev1, mddev) {
 		int j = rdev1->raid_disk;
 
-		if (j < 0 || j >= mddev->raid_disks) {
-			printk("raid0: bad disk number %d - aborting!\n", j);
+		if (mddev->level == 10) {
+			/* taking over a raid10-n2 array */
+			j /= 2;
+			rdev1->new_raid_disk = j;
+		}
+
+		if (mddev->level == 1) {
+			/* taiking over a raid1 array-
+			 * we have only one active disk
+			 */
+			j = 0;
+			rdev1->new_raid_disk = j;
+		}
+
+		if (j < 0) {
+			printk(KERN_ERR
+			       "md/raid0:%s: remove inactive devices before converting to RAID0\n",
+			       mdname(mddev));
 			goto abort;
 		}
-		if (zone->dev[j]) {
-			printk("raid0: multiple devices for %d - aborting!\n",
-				j);
+		if (j >= mddev->raid_disks) {
+			printk(KERN_ERR "md/raid0:%s: bad disk number %d - "
+			       "aborting!\n", mdname(mddev), j);
 			goto abort;
 		}
-		zone->dev[j] = rdev1;
+		if (dev[j]) {
+			printk(KERN_ERR "md/raid0:%s: multiple devices for %d - "
+			       "aborting!\n", mdname(mddev), j);
+			goto abort;
+		}
+		dev[j] = rdev1;
 
-		blk_queue_stack_limits(mddev->queue,
-				       rdev1->bdev->bd_disk->queue);
-		/* as we don't honour merge_bvec_fn, we must never risk
-		 * violating it, so limit ->max_sector to one PAGE, as
-		 * a one page request is never in violation.
-		 */
+		disk_stack_limits(mddev->gendisk, rdev1->bdev,
+				  rdev1->data_offset << 9);
 
-		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn &&
-		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
+		if (rdev1->bdev->bd_disk->queue->merge_bvec_fn)
+			conf->has_merge_bvec = 1;
 
-		if (!smallest || (rdev1->size <smallest->size))
+		if (!smallest || (rdev1->sectors < smallest->sectors))
 			smallest = rdev1;
 		cnt++;
+
+		if (blk_queue_discard(bdev_get_queue(rdev1->bdev)))
+			discard_supported = true;
 	}
 	if (cnt != mddev->raid_disks) {
-		printk("raid0: too few disks (%d of %d) - aborting!\n",
-			cnt, mddev->raid_disks);
+		printk(KERN_ERR "md/raid0:%s: too few disks (%d of %d) - "
+		       "aborting!\n", mdname(mddev), cnt, mddev->raid_disks);
 		goto abort;
 	}
 	zone->nb_dev = cnt;
-	zone->size = smallest->size * cnt;
-	zone->zone_offset = 0;
+	zone->zone_end = smallest->sectors * cnt;
 
-	current_offset = smallest->size;
-	curr_zone_offset = zone->size;
+	curr_zone_end = zone->zone_end;
 
 	/* now do the other zones */
 	for (i = 1; i < conf->nr_strip_zones; i++)
 	{
+		int j;
+
 		zone = conf->strip_zone + i;
-		zone->dev = conf->strip_zone[i-1].dev + mddev->raid_disks;
+		dev = conf->devlist + i * mddev->raid_disks;
 
-		printk("raid0: zone %d\n", i);
-		zone->dev_offset = current_offset;
+		pr_debug("md/raid0:%s: zone %d\n", mdname(mddev), i);
+		zone->dev_start = smallest->sectors;
 		smallest = NULL;
 		c = 0;
 
 		for (j=0; j<cnt; j++) {
-			char b[BDEVNAME_SIZE];
-			rdev = conf->strip_zone[0].dev[j];
-			printk("raid0: checking %s ...", bdevname(rdev->bdev,b));
-			if (rdev->size > current_offset)
-			{
-				printk(" contained as device %d\n", c);
-				zone->dev[c] = rdev;
-				c++;
-				if (!smallest || (rdev->size <smallest->size)) {
-					smallest = rdev;
-					printk("  (%llu) is smallest!.\n", 
-						(unsigned long long)rdev->size);
-				}
-			} else
-				printk(" nope.\n");
+			rdev = conf->devlist[j];
+			if (rdev->sectors <= zone->dev_start) {
+				pr_debug("md/raid0:%s: checking %s ... nope\n",
+					 mdname(mddev),
+					 bdevname(rdev->bdev, b));
+				continue;
+			}
+			pr_debug("md/raid0:%s: checking %s ..."
+				 " contained as device %d\n",
+				 mdname(mddev),
+				 bdevname(rdev->bdev, b), c);
+			dev[c] = rdev;
+			c++;
+			if (!smallest || rdev->sectors < smallest->sectors) {
+				smallest = rdev;
+				pr_debug("md/raid0:%s:  (%llu) is smallest!.\n",
+					 mdname(mddev),
+					 (unsigned long long)rdev->sectors);
+			}
 		}
 
 		zone->nb_dev = c;
-		zone->size = (smallest->size - current_offset) * c;
-		printk("raid0: zone->nb_dev: %d, size: %llu\n",
-			zone->nb_dev, (unsigned long long)zone->size);
+		sectors = (smallest->sectors - zone->dev_start) * c;
+		pr_debug("md/raid0:%s: zone->nb_dev: %d, sectors: %llu\n",
+			 mdname(mddev),
+			 zone->nb_dev, (unsigned long long)sectors);
 
-		zone->zone_offset = curr_zone_offset;
-		curr_zone_offset += zone->size;
+		curr_zone_end += sectors;
+		zone->zone_end = curr_zone_end;
 
-		current_offset = smallest->size;
-		printk("raid0: current zone offset: %llu\n",
-			(unsigned long long)current_offset);
+		pr_debug("md/raid0:%s: current zone start: %llu\n",
+			 mdname(mddev),
+			 (unsigned long long)smallest->sectors);
 	}
+	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
+	mddev->queue->backing_dev_info.congested_data = mddev;
 
-	/* Now find appropriate hash spacing.
-	 * We want a number which causes most hash entries to cover
-	 * at most two strips, but the hash table must be at most
-	 * 1 PAGE.  We choose the smallest strip, or contiguous collection
-	 * of strips, that has big enough size.  We never consider the last
-	 * strip though as it's size has no bearing on the efficacy of the hash
-	 * table.
+	/*
+	 * now since we have the hard sector sizes, we can make sure
+	 * chunk size is a multiple of that sector size
 	 */
-	conf->hash_spacing = curr_zone_offset;
-	min_spacing = curr_zone_offset;
-	sector_div(min_spacing, PAGE_SIZE/sizeof(struct strip_zone*));
-	for (i=0; i < conf->nr_strip_zones-1; i++) {
-		sector_t sz = 0;
-		for (j=i; j<conf->nr_strip_zones-1 &&
-			     sz < min_spacing ; j++)
-			sz += conf->strip_zone[j].size;
-		if (sz >= min_spacing && sz < conf->hash_spacing)
-			conf->hash_spacing = sz;
+	if ((mddev->chunk_sectors << 9) % queue_logical_block_size(mddev->queue)) {
+		printk(KERN_ERR "md/raid0:%s: chunk_size of %d not valid\n",
+		       mdname(mddev),
+		       mddev->chunk_sectors << 9);
+		goto abort;
 	}
 
-	mddev->queue->unplug_fn = raid0_unplug;
+	blk_queue_io_min(mddev->queue, mddev->chunk_sectors << 9);
+	blk_queue_io_opt(mddev->queue,
+			 (mddev->chunk_sectors << 9) * mddev->raid_disks);
 
-	mddev->queue->backing_dev_info.congested_fn = raid0_congested;
-	mddev->queue->backing_dev_info.congested_data = mddev;
+	if (!discard_supported)
+		queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
+	else
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
+
+	pr_debug("md/raid0:%s: done.\n", mdname(mddev));
+	*private_conf = conf;
 
-	printk("raid0: done.\n");
 	return 0;
- abort:
-	return 1;
+abort:
+	kfree(conf->strip_zone);
+	kfree(conf->devlist);
+	kfree(conf);
+	*private_conf = ERR_PTR(err);
+	return err;
+}
+
+/* Find the zone which holds a particular offset
+ * Update *sectorp to be an offset in that zone
+ */
+static struct strip_zone *find_zone(struct r0conf *conf,
+				    sector_t *sectorp)
+{
+	int i;
+	struct strip_zone *z = conf->strip_zone;
+	sector_t sector = *sectorp;
+
+	for (i = 0; i < conf->nr_strip_zones; i++)
+		if (sector < z[i].zone_end) {
+			if (i)
+				*sectorp = sector - z[i-1].zone_end;
+			return z + i;
+		}
+	BUG();
+}
+
+/*
+ * remaps the bio to the target device. we separate two flows.
+ * power 2 flow and a general flow for the sake of perfromance
+*/
+static struct md_rdev *map_sector(struct mddev *mddev, struct strip_zone *zone,
+				sector_t sector, sector_t *sector_offset)
+{
+	unsigned int sect_in_chunk;
+	sector_t chunk;
+	struct r0conf *conf = mddev->private;
+	int raid_disks = conf->strip_zone[0].nb_dev;
+	unsigned int chunk_sects = mddev->chunk_sectors;
+
+	if (is_power_of_2(chunk_sects)) {
+		int chunksect_bits = ffz(~chunk_sects);
+		/* find the sector offset inside the chunk */
+		sect_in_chunk  = sector & (chunk_sects - 1);
+		sector >>= chunksect_bits;
+		/* chunk in zone */
+		chunk = *sector_offset;
+		/* quotient is the chunk in real device*/
+		sector_div(chunk, zone->nb_dev << chunksect_bits);
+	} else{
+		sect_in_chunk = sector_div(sector, chunk_sects);
+		chunk = *sector_offset;
+		sector_div(chunk, chunk_sects * zone->nb_dev);
+	}
+	/*
+	*  position the bio over the real device
+	*  real sector = chunk in device + starting of zone
+	*	+ the position in the chunk
+	*/
+	*sector_offset = (chunk * chunk_sects) + sect_in_chunk;
+	return conf->devlist[(zone - conf->strip_zone)*raid_disks
+			     + sector_div(sector, zone->nb_dev)];
 }
 
 /**
- *	raid0_mergeable_bvec -- tell bio layer if a two requests can be merged
+ *	raid0_mergeable_bvec -- tell bio layer if two requests can be merged
  *	@q: request queue
  *	@bvm: properties of new bio
  *	@biovec: the request that could be merged to it.
@@ -250,102 +367,95 @@ static int raid0_mergeable_bvec(struct request_queue *q,
 				struct bvec_merge_data *bvm,
 				struct bio_vec *biovec)
 {
-	mddev_t *mddev = q->queuedata;
+	struct mddev *mddev = q->queuedata;
+	struct r0conf *conf = mddev->private;
 	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
+	sector_t sector_offset = sector;
 	int max;
-	unsigned int chunk_sectors = mddev->chunk_size >> 9;
+	unsigned int chunk_sectors = mddev->chunk_sectors;
 	unsigned int bio_sectors = bvm->bi_size >> 9;
-
-	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
-	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
+	struct strip_zone *zone;
+	struct md_rdev *rdev;
+	struct request_queue *subq;
+
+	if (is_power_of_2(chunk_sectors))
+		max =  (chunk_sectors - ((sector & (chunk_sectors-1))
+						+ bio_sectors)) << 9;
+	else
+		max =  (chunk_sectors - (sector_div(sector, chunk_sectors)
+						+ bio_sectors)) << 9;
+	if (max < 0)
+		max = 0; /* bio_add cannot handle a negative return */
 	if (max <= biovec->bv_len && bio_sectors == 0)
 		return biovec->bv_len;
-	else 
+	if (max < biovec->bv_len)
+		/* too small already, no need to check further */
+		return max;
+	if (!conf->has_merge_bvec)
+		return max;
+
+	/* May need to check subordinate device */
+	sector = sector_offset;
+	zone = find_zone(mddev->private, &sector_offset);
+	rdev = map_sector(mddev, zone, sector, &sector_offset);
+	subq = bdev_get_queue(rdev->bdev);
+	if (subq->merge_bvec_fn) {
+		bvm->bi_bdev = rdev->bdev;
+		bvm->bi_sector = sector_offset + zone->dev_start +
+			rdev->data_offset;
+		return min(max, subq->merge_bvec_fn(subq, bvm, biovec));
+	} else
 		return max;
 }
 
-static int raid0_run (mddev_t *mddev)
+static sector_t raid0_size(struct mddev *mddev, sector_t sectors, int raid_disks)
 {
-	unsigned  cur=0, i=0, nb_zone;
-	s64 size;
-	raid0_conf_t *conf;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
-
-	if (mddev->chunk_size == 0) {
-		printk(KERN_ERR "md/raid0: non-zero chunk size required.\n");
-		return -EINVAL;
-	}
-	printk(KERN_INFO "%s: setting max_sectors to %d, segment boundary to %d\n",
-	       mdname(mddev),
-	       mddev->chunk_size >> 9,
-	       (mddev->chunk_size>>1)-1);
-	blk_queue_max_sectors(mddev->queue, mddev->chunk_size >> 9);
-	blk_queue_segment_boundary(mddev->queue, (mddev->chunk_size>>1) - 1);
-	mddev->queue->queue_lock = &mddev->queue->__queue_lock;
+	sector_t array_sectors = 0;
+	struct md_rdev *rdev;
 
-	conf = kmalloc(sizeof (raid0_conf_t), GFP_KERNEL);
-	if (!conf)
-		goto out;
-	mddev->private = (void *)conf;
- 
-	conf->strip_zone = NULL;
-	conf->devlist = NULL;
-	if (create_strip_zones (mddev)) 
-		goto out_free_conf;
+	WARN_ONCE(sectors || raid_disks,
+		  "%s does not support generic reshape\n", __func__);
 
-	/* calculate array device size */
-	mddev->array_sectors = 0;
-	rdev_for_each(rdev, tmp, mddev)
-		mddev->array_sectors += rdev->size * 2;
-
-	printk("raid0 : md_size is %llu blocks.\n", 
-		(unsigned long long)mddev->array_sectors / 2);
-	printk("raid0 : conf->hash_spacing is %llu blocks.\n",
-		(unsigned long long)conf->hash_spacing);
-	{
-		sector_t s = mddev->array_sectors / 2;
-		sector_t space = conf->hash_spacing;
-		int round;
-		conf->preshift = 0;
-		if (sizeof(sector_t) > sizeof(u32)) {
-			/*shift down space and s so that sector_div will work */
-			while (space > (sector_t) (~(u32)0)) {
-				s >>= 1;
-				space >>= 1;
-				s += 1; /* force round-up */
-				conf->preshift++;
-			}
-		}
-		round = sector_div(s, (u32)space) ? 1 : 0;
-		nb_zone = s + round;
-	}
-	printk("raid0 : nb_zone is %d.\n", nb_zone);
-
-	printk("raid0 : Allocating %Zd bytes for hash.\n",
-				nb_zone*sizeof(struct strip_zone*));
-	conf->hash_table = kmalloc (sizeof (struct strip_zone *)*nb_zone, GFP_KERNEL);
-	if (!conf->hash_table)
-		goto out_free_conf;
-	size = conf->strip_zone[cur].size;
-
-	conf->hash_table[0] = conf->strip_zone + cur;
-	for (i=1; i< nb_zone; i++) {
-		while (size <= conf->hash_spacing) {
-			cur++;
-			size += conf->strip_zone[cur].size;
-		}
-		size -= conf->hash_spacing;
-		conf->hash_table[i] = conf->strip_zone + cur;
+	rdev_for_each(rdev, mddev)
+		array_sectors += (rdev->sectors &
+				  ~(sector_t)(mddev->chunk_sectors-1));
+
+	return array_sectors;
+}
+
+static int raid0_stop(struct mddev *mddev);
+
+static int raid0_run(struct mddev *mddev)
+{
+	struct r0conf *conf;
+	int ret;
+
+	if (mddev->chunk_sectors == 0) {
+		printk(KERN_ERR "md/raid0:%s: chunk size must be set.\n",
+		       mdname(mddev));
+		return -EINVAL;
 	}
-	if (conf->preshift) {
-		conf->hash_spacing >>= conf->preshift;
-		/* round hash_spacing up so when we divide by it, we
-		 * err on the side of too-low, which is safest
-		 */
-		conf->hash_spacing++;
+	if (md_check_no_bitmap(mddev))
+		return -EINVAL;
+	blk_queue_max_hw_sectors(mddev->queue, mddev->chunk_sectors);
+	blk_queue_max_write_same_sectors(mddev->queue, mddev->chunk_sectors);
+	blk_queue_max_discard_sectors(mddev->queue, mddev->chunk_sectors);
+
+	/* if private is not null, we are here after takeover */
+	if (mddev->private == NULL) {
+		ret = create_strip_zones(mddev, &conf);
+		if (ret < 0)
+			return ret;
+		mddev->private = conf;
 	}
+	conf = mddev->private;
+
+	/* calculate array device size */
+	md_set_array_sectors(mddev, raid0_size(mddev, 0, 0));
 
+	printk(KERN_INFO "md/raid0:%s: md_size is %llu sectors.\n",
+	       mdname(mddev),
+	       (unsigned long long)mddev->array_sectors);
 	/* calculate the max read-ahead size.
 	 * For read-ahead of large files to be effective, we need to
 	 * readahead at least twice a whole stripe. i.e. number of devices
@@ -356,154 +466,253 @@ static int raid0_run (mddev_t *mddev)
 	 * chunksize should be used in that case.
 	 */
 	{
-		int stripe = mddev->raid_disks * mddev->chunk_size / PAGE_SIZE;
+		int stripe = mddev->raid_disks *
+			(mddev->chunk_sectors << 9) / PAGE_SIZE;
 		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, raid0_mergeable_bvec);
-	return 0;
+	dump_zones(mddev);
 
-out_free_conf:
-	kfree(conf->strip_zone);
-	kfree(conf->devlist);
-	kfree(conf);
-	mddev->private = NULL;
-out:
-	return -ENOMEM;
+	ret = md_integrity_register(mddev);
+	if (ret)
+		raid0_stop(mddev);
+
+	return ret;
 }
 
-static int raid0_stop (mddev_t *mddev)
+static int raid0_stop(struct mddev *mddev)
 {
-	raid0_conf_t *conf = mddev_to_conf(mddev);
+	struct r0conf *conf = mddev->private;
 
 	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
-	kfree(conf->hash_table);
-	conf->hash_table = NULL;
 	kfree(conf->strip_zone);
-	conf->strip_zone = NULL;
+	kfree(conf->devlist);
 	kfree(conf);
 	mddev->private = NULL;
-
 	return 0;
 }
 
-static int raid0_make_request (struct request_queue *q, struct bio *bio)
+/*
+ * Is io distribute over 1 or more chunks ?
+*/
+static inline int is_io_in_chunk_boundary(struct mddev *mddev,
+			unsigned int chunk_sects, struct bio *bio)
+{
+	if (likely(is_power_of_2(chunk_sects))) {
+		return chunk_sects >=
+			((bio->bi_iter.bi_sector & (chunk_sects-1))
+					+ bio_sectors(bio));
+	} else{
+		sector_t sector = bio->bi_iter.bi_sector;
+		return chunk_sects >= (sector_div(sector, chunk_sects)
+						+ bio_sectors(bio));
+	}
+}
+
+static void raid0_make_request(struct mddev *mddev, struct bio *bio)
 {
-	mddev_t *mddev = q->queuedata;
-	unsigned int sect_in_chunk, chunksize_bits,  chunk_size, chunk_sects;
-	raid0_conf_t *conf = mddev_to_conf(mddev);
 	struct strip_zone *zone;
-	mdk_rdev_t *tmp_dev;
-	sector_t chunk;
-	sector_t block, rsect;
-	const int rw = bio_data_dir(bio);
+	struct md_rdev *tmp_dev;
+	struct bio *split;
 
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
+		md_flush_request(mddev, bio);
+		return;
 	}
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
-
-	chunk_size = mddev->chunk_size >> 10;
-	chunk_sects = mddev->chunk_size >> 9;
-	chunksize_bits = ffz(~chunk_size);
-	block = bio->bi_sector >> 1;
-	
-
-	if (unlikely(chunk_sects < (bio->bi_sector & (chunk_sects - 1)) + (bio->bi_size >> 9))) {
-		struct bio_pair *bp;
-		/* Sanity check -- queue functions should prevent this happening */
-		if (bio->bi_vcnt != 1 ||
-		    bio->bi_idx != 0)
-			goto bad_map;
-		/* This is a one page bio that upper layers
-		 * refuse to split for us, so we need to split it.
-		 */
-		bp = bio_split(bio, bio_split_pool, chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
-		if (raid0_make_request(q, &bp->bio1))
-			generic_make_request(&bp->bio1);
-		if (raid0_make_request(q, &bp->bio2))
-			generic_make_request(&bp->bio2);
-
-		bio_pair_release(bp);
-		return 0;
+	do {
+		sector_t sector = bio->bi_iter.bi_sector;
+		unsigned chunk_sects = mddev->chunk_sectors;
+
+		unsigned sectors = chunk_sects -
+			(likely(is_power_of_2(chunk_sects))
+			 ? (sector & (chunk_sects-1))
+			 : sector_div(sector, chunk_sects));
+
+		if (sectors < bio_sectors(bio)) {
+			split = bio_split(bio, sectors, GFP_NOIO, fs_bio_set);
+			bio_chain(split, bio);
+		} else {
+			split = bio;
+		}
+
+		zone = find_zone(mddev->private, &sector);
+		tmp_dev = map_sector(mddev, zone, sector, &sector);
+		split->bi_bdev = tmp_dev->bdev;
+		split->bi_iter.bi_sector = sector + zone->dev_start +
+			tmp_dev->data_offset;
+
+		if (unlikely((split->bi_rw & REQ_DISCARD) &&
+			 !blk_queue_discard(bdev_get_queue(split->bi_bdev)))) {
+			/* Just ignore it */
+			bio_endio(split, 0);
+		} else
+			generic_make_request(split);
+	} while (split != bio);
+}
+
+static void raid0_status(struct seq_file *seq, struct mddev *mddev)
+{
+	seq_printf(seq, " %dk chunks", mddev->chunk_sectors / 2);
+	return;
+}
+
+static void *raid0_takeover_raid45(struct mddev *mddev)
+{
+	struct md_rdev *rdev;
+	struct r0conf *priv_conf;
+
+	if (mddev->degraded != 1) {
+		printk(KERN_ERR "md/raid0:%s: raid5 must be degraded! Degraded disks: %d\n",
+		       mdname(mddev),
+		       mddev->degraded);
+		return ERR_PTR(-EINVAL);
 	}
- 
 
-	{
-		sector_t x = block >> conf->preshift;
-		sector_div(x, (u32)conf->hash_spacing);
-		zone = conf->hash_table[x];
+	rdev_for_each(rdev, mddev) {
+		/* check slot number for a disk */
+		if (rdev->raid_disk == mddev->raid_disks-1) {
+			printk(KERN_ERR "md/raid0:%s: raid5 must have missing parity disk!\n",
+			       mdname(mddev));
+			return ERR_PTR(-EINVAL);
+		}
+		rdev->sectors = mddev->dev_sectors;
 	}
- 
-	while (block >= (zone->zone_offset + zone->size)) 
-		zone++;
-    
-	sect_in_chunk = bio->bi_sector & ((chunk_size<<1) -1);
 
+	/* Set new parameters */
+	mddev->new_level = 0;
+	mddev->new_layout = 0;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->raid_disks--;
+	mddev->delta_disks = -1;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	create_strip_zones(mddev, &priv_conf);
+	return priv_conf;
+}
 
-	{
-		sector_t x =  (block - zone->zone_offset) >> chunksize_bits;
+static void *raid0_takeover_raid10(struct mddev *mddev)
+{
+	struct r0conf *priv_conf;
+
+	/* Check layout:
+	 *  - far_copies must be 1
+	 *  - near_copies must be 2
+	 *  - disks number must be even
+	 *  - all mirrors must be already degraded
+	 */
+	if (mddev->layout != ((1 << 8) + 2)) {
+		printk(KERN_ERR "md/raid0:%s:: Raid0 cannot takover layout: 0x%x\n",
+		       mdname(mddev),
+		       mddev->layout);
+		return ERR_PTR(-EINVAL);
+	}
+	if (mddev->raid_disks & 1) {
+		printk(KERN_ERR "md/raid0:%s: Raid0 cannot takover Raid10 with odd disk number.\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+	if (mddev->degraded != (mddev->raid_disks>>1)) {
+		printk(KERN_ERR "md/raid0:%s: All mirrors must be already degraded!\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	/* Set new parameters */
+	mddev->new_level = 0;
+	mddev->new_layout = 0;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->delta_disks = - mddev->raid_disks / 2;
+	mddev->raid_disks += mddev->delta_disks;
+	mddev->degraded = 0;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	create_strip_zones(mddev, &priv_conf);
+	return priv_conf;
+}
 
-		sector_div(x, zone->nb_dev);
-		chunk = x;
+static void *raid0_takeover_raid1(struct mddev *mddev)
+{
+	struct r0conf *priv_conf;
+	int chunksect;
 
-		x = block >> chunksize_bits;
-		tmp_dev = zone->dev[sector_div(x, zone->nb_dev)];
+	/* Check layout:
+	 *  - (N - 1) mirror drives must be already faulty
+	 */
+	if ((mddev->raid_disks - 1) != mddev->degraded) {
+		printk(KERN_ERR "md/raid0:%s: (N - 1) mirrors drives must be already faulty!\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
 	}
-	rsect = (((chunk << chunksize_bits) + zone->dev_offset)<<1)
-		+ sect_in_chunk;
- 
-	bio->bi_bdev = tmp_dev->bdev;
-	bio->bi_sector = rsect + tmp_dev->data_offset;
 
 	/*
-	 * Let the main block layer submit the IO and resolve recursion:
+	 * a raid1 doesn't have the notion of chunk size, so
+	 * figure out the largest suitable size we can use.
 	 */
-	return 1;
-
-bad_map:
-	printk("raid0_make_request bug: can't convert block across chunks"
-		" or bigger than %dk %llu %d\n", chunk_size, 
-		(unsigned long long)bio->bi_sector, bio->bi_size >> 10);
-
-	bio_io_error(bio);
-	return 0;
+	chunksect = 64 * 2; /* 64K by default */
+
+	/* The array must be an exact multiple of chunksize */
+	while (chunksect && (mddev->array_sectors & (chunksect - 1)))
+		chunksect >>= 1;
+
+	if ((chunksect << 9) < PAGE_SIZE)
+		/* array size does not allow a suitable chunk size */
+		return ERR_PTR(-EINVAL);
+
+	/* Set new parameters */
+	mddev->new_level = 0;
+	mddev->new_layout = 0;
+	mddev->new_chunk_sectors = chunksect;
+	mddev->chunk_sectors = chunksect;
+	mddev->delta_disks = 1 - mddev->raid_disks;
+	mddev->raid_disks = 1;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	create_strip_zones(mddev, &priv_conf);
+	return priv_conf;
 }
 
-static void raid0_status (struct seq_file *seq, mddev_t *mddev)
+static void *raid0_takeover(struct mddev *mddev)
 {
-#undef MD_DEBUG
-#ifdef MD_DEBUG
-	int j, k, h;
-	char b[BDEVNAME_SIZE];
-	raid0_conf_t *conf = mddev_to_conf(mddev);
+	/* raid0 can take over:
+	 *  raid4 - if all data disks are active.
+	 *  raid5 - providing it is Raid4 layout and one disk is faulty
+	 *  raid10 - assuming we have all necessary active disks
+	 *  raid1 - with (N -1) mirror drives faulty
+	 */
+	if (mddev->level == 4)
+		return raid0_takeover_raid45(mddev);
 
-	h = 0;
-	for (j = 0; j < conf->nr_strip_zones; j++) {
-		seq_printf(seq, "      z%d", j);
-		if (conf->hash_table[h] == conf->strip_zone+j)
-			seq_printf(seq, "(h%d)", h++);
-		seq_printf(seq, "=[");
-		for (k = 0; k < conf->strip_zone[j].nb_dev; k++)
-			seq_printf(seq, "%s/", bdevname(
-				conf->strip_zone[j].dev[k]->bdev,b));
+	if (mddev->level == 5) {
+		if (mddev->layout == ALGORITHM_PARITY_N)
+			return raid0_takeover_raid45(mddev);
 
-		seq_printf(seq, "] zo=%d do=%d s=%d\n",
-				conf->strip_zone[j].zone_offset,
-				conf->strip_zone[j].dev_offset,
-				conf->strip_zone[j].size);
+		printk(KERN_ERR "md/raid0:%s: Raid can only takeover Raid5 with layout: %d\n",
+		       mdname(mddev), ALGORITHM_PARITY_N);
 	}
-#endif
-	seq_printf(seq, " %dk chunks", mddev->chunk_size/1024);
-	return;
+
+	if (mddev->level == 10)
+		return raid0_takeover_raid10(mddev);
+
+	if (mddev->level == 1)
+		return raid0_takeover_raid1(mddev);
+
+	printk(KERN_ERR "Takeover from raid%i to raid0 not supported\n",
+		mddev->level);
+
+	return ERR_PTR(-EINVAL);
+}
+
+static void raid0_quiesce(struct mddev *mddev, int state)
+{
 }
 
-static struct mdk_personality raid0_personality=
+static struct md_personality raid0_personality=
 {
 	.name		= "raid0",
 	.level		= 0,
@@ -512,6 +721,9 @@ static struct mdk_personality raid0_personality=
 	.run		= raid0_run,
 	.stop		= raid0_stop,
 	.status		= raid0_status,
+	.size		= raid0_size,
+	.takeover	= raid0_takeover,
+	.quiesce	= raid0_quiesce,
 };
 
 static int __init raid0_init (void)
@@ -527,6 +739,7 @@ static void raid0_exit (void)
 module_init(raid0_init);
 module_exit(raid0_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("RAID0 (striping) personality for MD");
 MODULE_ALIAS("md-personality-2"); /* RAID0 */
 MODULE_ALIAS("md-raid0");
 MODULE_ALIAS("md-level-0");
diff --git a/drivers/md/raid0.h b/drivers/md/raid0.h
new file mode 100644
index 00000000000..05539d9c97f
--- /dev/null
+++ b/drivers/md/raid0.h
@@ -0,0 +1,19 @@
+#ifndef _RAID0_H
+#define _RAID0_H
+
+struct strip_zone {
+	sector_t zone_end;	/* Start of the next zone (in sectors) */
+	sector_t dev_start;	/* Zone offset in real dev (in sectors) */
+	int	 nb_dev;	/* # of devices attached to the zone */
+};
+
+struct r0conf {
+	struct strip_zone	*strip_zone;
+	struct md_rdev		**devlist; /* lists of rdevs, pointed to
+					    * by strip_zone->dev */
+	int			nr_strip_zones;
+	int			has_merge_bvec;	/* at least one member has
+						 * a merge_bvec_fn */
+};
+
+#endif
diff --git a/drivers/md/raid1.c b/drivers/md/raid1.c
index 03a5ab705c2..56e24c072b6 100644
--- a/drivers/md/raid1.c
+++ b/drivers/md/raid1.c
@@ -31,40 +31,52 @@
  * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
-#include "dm-bio-list.h"
-#include <linux/raid/raid1.h>
-#include <linux/raid/bitmap.h>
-
-#define DEBUG 0
-#if DEBUG
-#define PRINTK(x...) printk(x)
-#else
-#define PRINTK(x...)
-#endif
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/blkdev.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include <linux/ratelimit.h>
+#include "md.h"
+#include "raid1.h"
+#include "bitmap.h"
 
 /*
  * Number of guaranteed r1bios in case of extreme VM load:
  */
 #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)
 
-static void unplug_slaves(mddev_t *mddev);
+#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.
+ */
+static int max_queued_requests = 1024;
 
-static void allow_barrier(conf_t *conf);
-static void lower_barrier(conf_t *conf);
+static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
+			  sector_t bi_sector);
+static void lower_barrier(struct r1conf *conf);
 
 static void * r1bio_pool_alloc(gfp_t gfp_flags, void *data)
 {
 	struct pool_info *pi = data;
-	r1bio_t *r1_bio;
-	int size = offsetof(r1bio_t, bios[pi->raid_disks]);
+	int size = offsetof(struct r1bio, bios[pi->raid_disks]);
 
 	/* allocate a r1bio with room for raid_disks entries in the bios array */
-	r1_bio = kzalloc(size, gfp_flags);
-	if (!r1_bio)
-		unplug_slaves(pi->mddev);
-
-	return r1_bio;
+	return kzalloc(size, gfp_flags);
 }
 
 static void r1bio_pool_free(void *r1_bio, void *data)
@@ -73,30 +85,30 @@ static void r1bio_pool_free(void *r1_bio, void *data)
 }
 
 #define RESYNC_BLOCK_SIZE (64*1024)
-//#define RESYNC_BLOCK_SIZE PAGE_SIZE
+#define RESYNC_DEPTH 32
 #define RESYNC_SECTORS (RESYNC_BLOCK_SIZE >> 9)
 #define RESYNC_PAGES ((RESYNC_BLOCK_SIZE + PAGE_SIZE-1) / PAGE_SIZE)
-#define RESYNC_WINDOW (2048*1024)
+#define RESYNC_WINDOW (RESYNC_BLOCK_SIZE * RESYNC_DEPTH)
+#define RESYNC_WINDOW_SECTORS (RESYNC_WINDOW >> 9)
+#define NEXT_NORMALIO_DISTANCE (3 * RESYNC_WINDOW_SECTORS)
 
 static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
 {
 	struct pool_info *pi = data;
-	struct page *page;
-	r1bio_t *r1_bio;
+	struct r1bio *r1_bio;
 	struct bio *bio;
+	int need_pages;
 	int i, j;
 
 	r1_bio = r1bio_pool_alloc(gfp_flags, pi);
-	if (!r1_bio) {
-		unplug_slaves(pi->mddev);
+	if (!r1_bio)
 		return NULL;
-	}
 
 	/*
 	 * Allocate bios : 1 for reading, n-1 for writing
 	 */
 	for (j = pi->raid_disks ; j-- ; ) {
-		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
+		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
 		if (!bio)
 			goto out_free_bio;
 		r1_bio->bios[j] = bio;
@@ -108,18 +120,15 @@ static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
 	 * RESYNC_PAGES for each bio.
 	 */
 	if (test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery))
-		j = pi->raid_disks;
+		need_pages = pi->raid_disks;
 	else
-		j = 1;
-	while(j--) {
+		need_pages = 1;
+	for (j = 0; j < need_pages; j++) {
 		bio = r1_bio->bios[j];
-		for (i = 0; i < RESYNC_PAGES; i++) {
-			page = alloc_page(gfp_flags);
-			if (unlikely(!page))
-				goto out_free_pages;
+		bio->bi_vcnt = RESYNC_PAGES;
 
-			bio->bi_io_vec[i].bv_page = page;
-		}
+		if (bio_alloc_pages(bio, gfp_flags))
+			goto out_free_pages;
 	}
 	/* If not user-requests, copy the page pointers to all bios */
 	if (!test_bit(MD_RECOVERY_REQUESTED, &pi->mddev->recovery)) {
@@ -134,12 +143,15 @@ static void * r1buf_pool_alloc(gfp_t gfp_flags, void *data)
 	return r1_bio;
 
 out_free_pages:
-	for (i=0; i < RESYNC_PAGES ; i++)
-		for (j=0 ; j < pi->raid_disks; j++)
-			safe_put_page(r1_bio->bios[j]->bi_io_vec[i].bv_page);
-	j = -1;
+	while (--j >= 0) {
+		struct bio_vec *bv;
+
+		bio_for_each_segment_all(bv, r1_bio->bios[j], i)
+			__free_page(bv->bv_page);
+	}
+
 out_free_bio:
-	while ( ++j < pi->raid_disks )
+	while (++j < pi->raid_disks)
 		bio_put(r1_bio->bios[j]);
 	r1bio_pool_free(r1_bio, data);
 	return NULL;
@@ -149,7 +161,7 @@ static void r1buf_pool_free(void *__r1_bio, void *data)
 {
 	struct pool_info *pi = data;
 	int i,j;
-	r1bio_t *r1bio = __r1_bio;
+	struct r1bio *r1bio = __r1_bio;
 
 	for (i = 0; i < RESYNC_PAGES; i++)
 		for (j = pi->raid_disks; j-- ;) {
@@ -164,38 +176,32 @@ static void r1buf_pool_free(void *__r1_bio, void *data)
 	r1bio_pool_free(r1bio, data);
 }
 
-static void put_all_bios(conf_t *conf, r1bio_t *r1_bio)
+static void put_all_bios(struct r1conf *conf, struct r1bio *r1_bio)
 {
 	int i;
 
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		struct bio **bio = r1_bio->bios + i;
-		if (*bio && *bio != IO_BLOCKED)
+		if (!BIO_SPECIAL(*bio))
 			bio_put(*bio);
 		*bio = NULL;
 	}
 }
 
-static void free_r1bio(r1bio_t *r1_bio)
+static void free_r1bio(struct r1bio *r1_bio)
 {
-	conf_t *conf = mddev_to_conf(r1_bio->mddev);
-
-	/*
-	 * Wake up any possible resync thread that waits for the device
-	 * to go idle.
-	 */
-	allow_barrier(conf);
+	struct r1conf *conf = r1_bio->mddev->private;
 
 	put_all_bios(conf, r1_bio);
 	mempool_free(r1_bio, conf->r1bio_pool);
 }
 
-static void put_buf(r1bio_t *r1_bio)
+static void put_buf(struct r1bio *r1_bio)
 {
-	conf_t *conf = mddev_to_conf(r1_bio->mddev);
+	struct r1conf *conf = r1_bio->mddev->private;
 	int i;
 
-	for (i=0; i<conf->raid_disks; i++) {
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 		struct bio *bio = r1_bio->bios[i];
 		if (bio->bi_end_io)
 			rdev_dec_pending(conf->mirrors[i].rdev, r1_bio->mddev);
@@ -206,11 +212,11 @@ static void put_buf(r1bio_t *r1_bio)
 	lower_barrier(conf);
 }
 
-static void reschedule_retry(r1bio_t *r1_bio)
+static void reschedule_retry(struct r1bio *r1_bio)
 {
 	unsigned long flags;
-	mddev_t *mddev = r1_bio->mddev;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = r1_bio->mddev;
+	struct r1conf *conf = mddev->private;
 
 	spin_lock_irqsave(&conf->device_lock, flags);
 	list_add(&r1_bio->retry_list, &conf->retry_list);
@@ -226,20 +232,52 @@ static void reschedule_retry(r1bio_t *r1_bio)
  * operation and are ready to return a success/failure code to the buffer
  * cache layer.
  */
-static void raid_end_bio_io(r1bio_t *r1_bio)
+static void call_bio_endio(struct r1bio *r1_bio)
+{
+	struct bio *bio = r1_bio->master_bio;
+	int done;
+	struct r1conf *conf = r1_bio->mddev->private;
+	sector_t start_next_window = r1_bio->start_next_window;
+	sector_t bi_sector = bio->bi_iter.bi_sector;
+
+	if (bio->bi_phys_segments) {
+		unsigned long flags;
+		spin_lock_irqsave(&conf->device_lock, flags);
+		bio->bi_phys_segments--;
+		done = (bio->bi_phys_segments == 0);
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+		/*
+		 * make_request() might be waiting for
+		 * bi_phys_segments to decrease
+		 */
+		wake_up(&conf->wait_barrier);
+	} else
+		done = 1;
+
+	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
+		clear_bit(BIO_UPTODATE, &bio->bi_flags);
+	if (done) {
+		bio_endio(bio, 0);
+		/*
+		 * Wake up any possible resync thread that waits for the device
+		 * to go idle.
+		 */
+		allow_barrier(conf, start_next_window, bi_sector);
+	}
+}
+
+static void raid_end_bio_io(struct r1bio *r1_bio)
 {
 	struct bio *bio = r1_bio->master_bio;
 
 	/* if nobody has done the final endio yet, do it now */
 	if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
-		PRINTK(KERN_DEBUG "raid1: sync end %s on sectors %llu-%llu\n",
-			(bio_data_dir(bio) == WRITE) ? "write" : "read",
-			(unsigned long long) bio->bi_sector,
-			(unsigned long long) bio->bi_sector +
-				(bio->bi_size >> 9) - 1);
+		pr_debug("raid1: sync end %s on sectors %llu-%llu\n",
+			 (bio_data_dir(bio) == WRITE) ? "write" : "read",
+			 (unsigned long long) bio->bi_iter.bi_sector,
+			 (unsigned long long) bio_end_sector(bio) - 1);
 
-		bio_endio(bio,
-			test_bit(R1BIO_Uptodate, &r1_bio->state) ? 0 : -EIO);
+		call_bio_endio(r1_bio);
 	}
 	free_r1bio(r1_bio);
 }
@@ -247,20 +285,39 @@ static void raid_end_bio_io(r1bio_t *r1_bio)
 /*
  * Update disk head position estimator based on IRQ completion info.
  */
-static inline void update_head_pos(int disk, r1bio_t *r1_bio)
+static inline void update_head_pos(int disk, struct r1bio *r1_bio)
 {
-	conf_t *conf = mddev_to_conf(r1_bio->mddev);
+	struct r1conf *conf = r1_bio->mddev->private;
 
 	conf->mirrors[disk].head_position =
 		r1_bio->sector + (r1_bio->sectors);
 }
 
+/*
+ * Find the disk number which triggered given bio
+ */
+static int find_bio_disk(struct r1bio *r1_bio, struct bio *bio)
+{
+	int mirror;
+	struct r1conf *conf = r1_bio->mddev->private;
+	int raid_disks = conf->raid_disks;
+
+	for (mirror = 0; mirror < raid_disks * 2; mirror++)
+		if (r1_bio->bios[mirror] == bio)
+			break;
+
+	BUG_ON(mirror == raid_disks * 2);
+	update_head_pos(mirror, r1_bio);
+
+	return mirror;
+}
+
 static void raid1_end_read_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	struct r1bio *r1_bio = bio->bi_private;
 	int mirror;
-	conf_t *conf = mddev_to_conf(r1_bio->mddev);
+	struct r1conf *conf = r1_bio->mddev->private;
 
 	mirror = r1_bio->read_disk;
 	/*
@@ -284,113 +341,154 @@ static void raid1_end_read_request(struct bio *bio, int error)
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 	}
 
-	if (uptodate)
+	if (uptodate) {
 		raid_end_bio_io(r1_bio);
-	else {
+		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
+	} else {
 		/*
 		 * oops, read error:
 		 */
 		char b[BDEVNAME_SIZE];
-		if (printk_ratelimit())
-			printk(KERN_ERR "raid1: %s: rescheduling sector %llu\n",
-			       bdevname(conf->mirrors[mirror].rdev->bdev,b), (unsigned long long)r1_bio->sector);
+		printk_ratelimited(
+			KERN_ERR "md/raid1:%s: %s: "
+			"rescheduling sector %llu\n",
+			mdname(conf->mddev),
+			bdevname(conf->mirrors[mirror].rdev->bdev,
+				 b),
+			(unsigned long long)r1_bio->sector);
+		set_bit(R1BIO_ReadError, &r1_bio->state);
 		reschedule_retry(r1_bio);
+		/* don't drop the reference on read_disk yet */
 	}
+}
 
-	rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
+static void close_write(struct r1bio *r1_bio)
+{
+	/* it really is the end of this request */
+	if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
+		/* free extra copy of the data pages */
+		int i = r1_bio->behind_page_count;
+		while (i--)
+			safe_put_page(r1_bio->behind_bvecs[i].bv_page);
+		kfree(r1_bio->behind_bvecs);
+		r1_bio->behind_bvecs = NULL;
+	}
+	/* clear the bitmap if all writes complete successfully */
+	bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector,
+			r1_bio->sectors,
+			!test_bit(R1BIO_Degraded, &r1_bio->state),
+			test_bit(R1BIO_BehindIO, &r1_bio->state));
+	md_write_end(r1_bio->mddev);
+}
+
+static void r1_bio_write_done(struct r1bio *r1_bio)
+{
+	if (!atomic_dec_and_test(&r1_bio->remaining))
+		return;
+
+	if (test_bit(R1BIO_WriteError, &r1_bio->state))
+		reschedule_retry(r1_bio);
+	else {
+		close_write(r1_bio);
+		if (test_bit(R1BIO_MadeGood, &r1_bio->state))
+			reschedule_retry(r1_bio);
+		else
+			raid_end_bio_io(r1_bio);
+	}
 }
 
 static void raid1_end_write_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
+	struct r1bio *r1_bio = bio->bi_private;
 	int mirror, behind = test_bit(R1BIO_BehindIO, &r1_bio->state);
-	conf_t *conf = mddev_to_conf(r1_bio->mddev);
+	struct r1conf *conf = r1_bio->mddev->private;
 	struct bio *to_put = NULL;
 
+	mirror = find_bio_disk(r1_bio, bio);
 
-	for (mirror = 0; mirror < conf->raid_disks; mirror++)
-		if (r1_bio->bios[mirror] == bio)
-			break;
-
-	if (error == -EOPNOTSUPP && test_bit(R1BIO_Barrier, &r1_bio->state)) {
-		set_bit(BarriersNotsupp, &conf->mirrors[mirror].rdev->flags);
-		set_bit(R1BIO_BarrierRetry, &r1_bio->state);
-		r1_bio->mddev->barriers_work = 0;
-		/* Don't rdev_dec_pending in this branch - keep it for the retry */
+	/*
+	 * 'one mirror IO has finished' event handler:
+	 */
+	if (!uptodate) {
+		set_bit(WriteErrorSeen,
+			&conf->mirrors[mirror].rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &conf->mirrors[mirror].rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				conf->mddev->recovery);
+
+		set_bit(R1BIO_WriteError, &r1_bio->state);
 	} else {
 		/*
-		 * this branch is our 'one mirror IO has finished' event handler:
+		 * Set R1BIO_Uptodate in our master bio, so that we
+		 * will return a good error code for to the higher
+		 * levels even if IO on some other mirrored buffer
+		 * fails.
+		 *
+		 * The 'master' represents the composite IO operation
+		 * to user-side. So if something waits for IO, then it
+		 * will wait for the 'master' bio.
 		 */
+		sector_t first_bad;
+		int bad_sectors;
+
 		r1_bio->bios[mirror] = NULL;
 		to_put = bio;
-		if (!uptodate) {
-			md_error(r1_bio->mddev, conf->mirrors[mirror].rdev);
-			/* an I/O failed, we can't clear the bitmap */
-			set_bit(R1BIO_Degraded, &r1_bio->state);
-		} else
-			/*
-			 * Set R1BIO_Uptodate in our master bio, so that
-			 * we will return a good error code for to the higher
-			 * levels even if IO on some other mirrored buffer fails.
-			 *
-			 * The 'master' represents the composite IO operation to
-			 * user-side. So if something waits for IO, then it will
-			 * wait for the 'master' bio.
-			 */
+		/*
+		 * Do not set R1BIO_Uptodate if the current device is
+		 * rebuilding or Faulty. This is because we cannot use
+		 * such device for properly reading the data back (we could
+		 * potentially use it, if the current write would have felt
+		 * before rdev->recovery_offset, but for simplicity we don't
+		 * check this here.
+		 */
+		if (test_bit(In_sync, &conf->mirrors[mirror].rdev->flags) &&
+		    !test_bit(Faulty, &conf->mirrors[mirror].rdev->flags))
 			set_bit(R1BIO_Uptodate, &r1_bio->state);
 
-		update_head_pos(mirror, r1_bio);
-
-		if (behind) {
-			if (test_bit(WriteMostly, &conf->mirrors[mirror].rdev->flags))
-				atomic_dec(&r1_bio->behind_remaining);
-
-			/* In behind mode, we ACK the master bio once the I/O has safely
-			 * reached all non-writemostly disks. Setting the Returned bit
-			 * ensures that this gets done only once -- we don't ever want to
-			 * return -EIO here, instead we'll wait */
-
-			if (atomic_read(&r1_bio->behind_remaining) >= (atomic_read(&r1_bio->remaining)-1) &&
-			    test_bit(R1BIO_Uptodate, &r1_bio->state)) {
-				/* Maybe we can return now */
-				if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
-					struct bio *mbio = r1_bio->master_bio;
-					PRINTK(KERN_DEBUG "raid1: behind end write sectors %llu-%llu\n",
-					       (unsigned long long) mbio->bi_sector,
-					       (unsigned long long) mbio->bi_sector +
-					       (mbio->bi_size >> 9) - 1);
-					bio_endio(mbio, 0);
-				}
+		/* Maybe we can clear some bad blocks. */
+		if (is_badblock(conf->mirrors[mirror].rdev,
+				r1_bio->sector, r1_bio->sectors,
+				&first_bad, &bad_sectors)) {
+			r1_bio->bios[mirror] = IO_MADE_GOOD;
+			set_bit(R1BIO_MadeGood, &r1_bio->state);
+		}
+	}
+
+	if (behind) {
+		if (test_bit(WriteMostly, &conf->mirrors[mirror].rdev->flags))
+			atomic_dec(&r1_bio->behind_remaining);
+
+		/*
+		 * In behind mode, we ACK the master bio once the I/O
+		 * has safely reached all non-writemostly
+		 * disks. Setting the Returned bit ensures that this
+		 * gets done only once -- we don't ever want to return
+		 * -EIO here, instead we'll wait
+		 */
+		if (atomic_read(&r1_bio->behind_remaining) >= (atomic_read(&r1_bio->remaining)-1) &&
+		    test_bit(R1BIO_Uptodate, &r1_bio->state)) {
+			/* Maybe we can return now */
+			if (!test_and_set_bit(R1BIO_Returned, &r1_bio->state)) {
+				struct bio *mbio = r1_bio->master_bio;
+				pr_debug("raid1: behind end write sectors"
+					 " %llu-%llu\n",
+					 (unsigned long long) mbio->bi_iter.bi_sector,
+					 (unsigned long long) bio_end_sector(mbio) - 1);
+				call_bio_endio(r1_bio);
 			}
 		}
-		rdev_dec_pending(conf->mirrors[mirror].rdev, conf->mddev);
 	}
+	if (r1_bio->bios[mirror] == NULL)
+		rdev_dec_pending(conf->mirrors[mirror].rdev,
+				 conf->mddev);
+
 	/*
-	 *
 	 * Let's see if all mirrored write operations have finished
 	 * already.
 	 */
-	if (atomic_dec_and_test(&r1_bio->remaining)) {
-		if (test_bit(R1BIO_BarrierRetry, &r1_bio->state))
-			reschedule_retry(r1_bio);
-		else {
-			/* it really is the end of this request */
-			if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
-				/* free extra copy of the data pages */
-				int i = bio->bi_vcnt;
-				while (i--)
-					safe_put_page(bio->bi_io_vec[i].bv_page);
-			}
-			/* clear the bitmap if all writes complete successfully */
-			bitmap_endwrite(r1_bio->mddev->bitmap, r1_bio->sector,
-					r1_bio->sectors,
-					!test_bit(R1BIO_Degraded, &r1_bio->state),
-					behind);
-			md_write_end(r1_bio->mddev);
-			raid_end_bio_io(r1_bio);
-		}
-	}
+	r1_bio_write_done(r1_bio);
 
 	if (to_put)
 		bio_put(to_put);
@@ -411,14 +509,19 @@ static void raid1_end_write_request(struct bio *bio, int error)
  *
  * The rdev for the device selected will have nr_pending incremented.
  */
-static int read_balance(conf_t *conf, r1bio_t *r1_bio)
+static int read_balance(struct r1conf *conf, struct r1bio *r1_bio, int *max_sectors)
 {
-	const unsigned long this_sector = r1_bio->sector;
-	int new_disk = conf->last_used, disk = new_disk;
-	int wonly_disk = -1;
-	const int sectors = r1_bio->sectors;
-	sector_t new_distance, current_distance;
-	mdk_rdev_t *rdev;
+	const sector_t this_sector = r1_bio->sector;
+	int sectors;
+	int best_good_sectors;
+	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();
 	/*
@@ -427,161 +530,236 @@ static int read_balance(conf_t *conf, r1bio_t *r1_bio)
 	 * We take the first readable disk when above the resync window.
 	 */
  retry:
-	if (conf->mddev->recovery_cp < MaxSector &&
-	    (this_sector + sectors >= conf->next_resync)) {
-		/* Choose the first operation device, for consistancy */
-		new_disk = 0;
-
-		for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
-		     r1_bio->bios[new_disk] == IO_BLOCKED ||
-		     !rdev || !test_bit(In_sync, &rdev->flags)
-			     || test_bit(WriteMostly, &rdev->flags);
-		     rdev = rcu_dereference(conf->mirrors[++new_disk].rdev)) {
-
-			if (rdev && test_bit(In_sync, &rdev->flags) &&
-				r1_bio->bios[new_disk] != IO_BLOCKED)
-				wonly_disk = new_disk;
-
-			if (new_disk == conf->raid_disks - 1) {
-				new_disk = wonly_disk;
-				break;
-			}
-		}
-		goto rb_out;
-	}
+	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))
+		choose_first = 1;
+	else
+		choose_first = 0;
 
-	/* make sure the disk is operational */
-	for (rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
-	     r1_bio->bios[new_disk] == IO_BLOCKED ||
-	     !rdev || !test_bit(In_sync, &rdev->flags) ||
-		     test_bit(WriteMostly, &rdev->flags);
-	     rdev = rcu_dereference(conf->mirrors[new_disk].rdev)) {
+	for (disk = 0 ; disk < conf->raid_disks * 2 ; disk++) {
+		sector_t dist;
+		sector_t first_bad;
+		int bad_sectors;
+		unsigned int pending;
+		bool nonrot;
 
-		if (rdev && test_bit(In_sync, &rdev->flags) &&
-		    r1_bio->bios[new_disk] != IO_BLOCKED)
-			wonly_disk = new_disk;
+		rdev = rcu_dereference(conf->mirrors[disk].rdev);
+		if (r1_bio->bios[disk] == IO_BLOCKED
+		    || rdev == NULL
+		    || test_bit(Unmerged, &rdev->flags)
+		    || test_bit(Faulty, &rdev->flags))
+			continue;
+		if (!test_bit(In_sync, &rdev->flags) &&
+		    rdev->recovery_offset < this_sector + sectors)
+			continue;
+		if (test_bit(WriteMostly, &rdev->flags)) {
+			/* Don't balance among write-mostly, just
+			 * use the first as a last resort */
+			if (best_disk < 0) {
+				if (is_badblock(rdev, this_sector, sectors,
+						&first_bad, &bad_sectors)) {
+					if (first_bad < this_sector)
+						/* Cannot use this */
+						continue;
+					best_good_sectors = first_bad - this_sector;
+				} else
+					best_good_sectors = sectors;
+				best_disk = disk;
+			}
+			continue;
+		}
+		/* This is a reasonable device to use.  It might
+		 * even be best.
+		 */
+		if (is_badblock(rdev, this_sector, sectors,
+				&first_bad, &bad_sectors)) {
+			if (best_dist < MaxSector)
+				/* already have a better device */
+				continue;
+			if (first_bad <= this_sector) {
+				/* cannot read here. If this is the 'primary'
+				 * device, then we must not read beyond
+				 * bad_sectors from another device..
+				 */
+				bad_sectors -= (this_sector - first_bad);
+				if (choose_first && sectors > bad_sectors)
+					sectors = bad_sectors;
+				if (best_good_sectors > sectors)
+					best_good_sectors = sectors;
 
-		if (new_disk <= 0)
-			new_disk = conf->raid_disks;
-		new_disk--;
-		if (new_disk == disk) {
-			new_disk = wonly_disk;
+			} else {
+				sector_t good_sectors = first_bad - this_sector;
+				if (good_sectors > best_good_sectors) {
+					best_good_sectors = good_sectors;
+					best_disk = disk;
+				}
+				if (choose_first)
+					break;
+			}
+			continue;
+		} 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) {
+			best_disk = disk;
 			break;
 		}
-	}
-
-	if (new_disk < 0)
-		goto rb_out;
+		/* 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];
 
-	disk = new_disk;
-	/* now disk == new_disk == starting point for search */
-
-	/*
-	 * Don't change to another disk for sequential reads:
-	 */
-	if (conf->next_seq_sect == this_sector)
-		goto rb_out;
-	if (this_sector == conf->mirrors[new_disk].head_position)
-		goto rb_out;
-
-	current_distance = abs(this_sector - conf->mirrors[disk].head_position);
-
-	/* Find the disk whose head is closest */
-
-	do {
-		if (disk <= 0)
-			disk = conf->raid_disks;
-		disk--;
-
-		rdev = rcu_dereference(conf->mirrors[disk].rdev);
+			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 (!rdev || r1_bio->bios[disk] == IO_BLOCKED ||
-		    !test_bit(In_sync, &rdev->flags) ||
-		    test_bit(WriteMostly, &rdev->flags))
+		if (choose_next_idle)
 			continue;
 
-		if (!atomic_read(&rdev->nr_pending)) {
-			new_disk = disk;
-			break;
+		if (min_pending > pending) {
+			min_pending = pending;
+			best_pending_disk = disk;
 		}
-		new_distance = abs(this_sector - conf->mirrors[disk].head_position);
-		if (new_distance < current_distance) {
-			current_distance = new_distance;
-			new_disk = disk;
-		}
-	} while (disk != conf->last_used);
 
- rb_out:
+		if (dist < best_dist) {
+			best_dist = dist;
+			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 (new_disk >= 0) {
-		rdev = rcu_dereference(conf->mirrors[new_disk].rdev);
+	if (best_disk >= 0) {
+		rdev = rcu_dereference(conf->mirrors[best_disk].rdev);
 		if (!rdev)
 			goto retry;
 		atomic_inc(&rdev->nr_pending);
-		if (!test_bit(In_sync, &rdev->flags)) {
+		if (test_bit(Faulty, &rdev->flags)) {
 			/* cannot risk returning a device that failed
 			 * before we inc'ed nr_pending
 			 */
 			rdev_dec_pending(rdev, conf->mddev);
 			goto retry;
 		}
-		conf->next_seq_sect = this_sector + sectors;
-		conf->last_used = new_disk;
+		sectors = best_good_sectors;
+
+		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;
 
-	return new_disk;
+	return best_disk;
 }
 
-static void unplug_slaves(mddev_t *mddev)
+static int raid1_mergeable_bvec(struct request_queue *q,
+				struct bvec_merge_data *bvm,
+				struct bio_vec *biovec)
 {
-	conf_t *conf = mddev_to_conf(mddev);
-	int i;
-
-	rcu_read_lock();
-	for (i=0; i<mddev->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
-		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
-			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
-
-			atomic_inc(&rdev->nr_pending);
-			rcu_read_unlock();
-
-			blk_unplug(r_queue);
-
-			rdev_dec_pending(rdev, mddev);
-			rcu_read_lock();
+	struct mddev *mddev = q->queuedata;
+	struct r1conf *conf = mddev->private;
+	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
+	int max = biovec->bv_len;
+
+	if (mddev->merge_check_needed) {
+		int disk;
+		rcu_read_lock();
+		for (disk = 0; disk < conf->raid_disks * 2; disk++) {
+			struct md_rdev *rdev = rcu_dereference(
+				conf->mirrors[disk].rdev);
+			if (rdev && !test_bit(Faulty, &rdev->flags)) {
+				struct request_queue *q =
+					bdev_get_queue(rdev->bdev);
+				if (q->merge_bvec_fn) {
+					bvm->bi_sector = sector +
+						rdev->data_offset;
+					bvm->bi_bdev = rdev->bdev;
+					max = min(max, q->merge_bvec_fn(
+							  q, bvm, biovec));
+				}
+			}
 		}
+		rcu_read_unlock();
 	}
-	rcu_read_unlock();
-}
+	return max;
 
-static void raid1_unplug(struct request_queue *q)
-{
-	mddev_t *mddev = q->queuedata;
-
-	unplug_slaves(mddev);
-	md_wakeup_thread(mddev->thread);
 }
 
-static int raid1_congested(void *data, int bits)
+int md_raid1_congested(struct mddev *mddev, int bits)
 {
-	mddev_t *mddev = data;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 	int i, ret = 0;
 
+	if ((bits & (1 << BDI_async_congested)) &&
+	    conf->pending_count >= max_queued_requests)
+		return 1;
+
 	rcu_read_lock();
-	for (i = 0; i < mddev->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
+	for (i = 0; i < conf->raid_disks * 2; i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags)) {
 			struct request_queue *q = bdev_get_queue(rdev->bdev);
 
+			BUG_ON(!q);
+
 			/* Note the '|| 1' - when read_balance prefers
 			 * non-congested targets, it can be removed
 			 */
-			if ((bits & (1<<BDI_write_congested)) || 1)
+			if ((bits & (1<<BDI_async_congested)) || 1)
 				ret |= bdi_congested(&q->backing_dev_info, bits);
 			else
 				ret &= bdi_congested(&q->backing_dev_info, bits);
@@ -590,37 +768,46 @@ static int raid1_congested(void *data, int bits)
 	rcu_read_unlock();
 	return ret;
 }
+EXPORT_SYMBOL_GPL(md_raid1_congested);
+
+static int raid1_congested(void *data, int bits)
+{
+	struct mddev *mddev = data;
 
+	return mddev_congested(mddev, bits) ||
+		md_raid1_congested(mddev, bits);
+}
 
-static int flush_pending_writes(conf_t *conf)
+static void flush_pending_writes(struct r1conf *conf)
 {
 	/* Any writes that have been queued but are awaiting
 	 * bitmap updates get flushed here.
-	 * We return 1 if any requests were actually submitted.
 	 */
-	int rv = 0;
-
 	spin_lock_irq(&conf->device_lock);
 
 	if (conf->pending_bio_list.head) {
 		struct bio *bio;
 		bio = bio_list_get(&conf->pending_bio_list);
-		blk_remove_plug(conf->mddev->queue);
+		conf->pending_count = 0;
 		spin_unlock_irq(&conf->device_lock);
 		/* flush any pending bitmap writes to
 		 * disk before proceeding w/ I/O */
 		bitmap_unplug(conf->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);
+			if (unlikely((bio->bi_rw & REQ_DISCARD) &&
+			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+				/* Just ignore it */
+				bio_endio(bio, 0);
+			else
+				generic_make_request(bio);
 			bio = next;
 		}
-		rv = 1;
 	} else
 		spin_unlock_irq(&conf->device_lock);
-	return rv;
 }
 
 /* Barriers....
@@ -644,169 +831,312 @@ static int flush_pending_writes(conf_t *conf)
  *    there is no normal IO happeing.  It must arrange to call
  *    lower_barrier when the particular background IO completes.
  */
-#define RESYNC_DEPTH 32
-
-static void raise_barrier(conf_t *conf)
+static void raise_barrier(struct r1conf *conf)
 {
 	spin_lock_irq(&conf->resync_lock);
 
 	/* Wait until no block IO is waiting */
 	wait_event_lock_irq(conf->wait_barrier, !conf->nr_waiting,
-			    conf->resync_lock,
-			    raid1_unplug(conf->mddev->queue));
+			    conf->resync_lock);
 
 	/* block any new IO from starting */
 	conf->barrier++;
 
-	/* No wait for all pending IO to complete */
+	/* For these conditions we must wait:
+	 * A: while the array is in frozen state
+	 * B: while barrier >= RESYNC_DEPTH, meaning resync reach
+	 *    the max count which allowed.
+	 * C: next_resync + RESYNC_SECTORS > start_next_window, meaning
+	 *    next resync will reach to the window which normal bios are
+	 *    handling.
+	 */
 	wait_event_lock_irq(conf->wait_barrier,
-			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
-			    conf->resync_lock,
-			    raid1_unplug(conf->mddev->queue));
+			    !conf->array_frozen &&
+			    conf->barrier < RESYNC_DEPTH &&
+			    (conf->start_next_window >=
+			     conf->next_resync + RESYNC_SECTORS),
+			    conf->resync_lock);
 
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static void lower_barrier(conf_t *conf)
+static void lower_barrier(struct r1conf *conf)
 {
 	unsigned long flags;
+	BUG_ON(conf->barrier <= 0);
 	spin_lock_irqsave(&conf->resync_lock, flags);
 	conf->barrier--;
 	spin_unlock_irqrestore(&conf->resync_lock, flags);
 	wake_up(&conf->wait_barrier);
 }
 
-static void wait_barrier(conf_t *conf)
+static bool need_to_wait_for_sync(struct r1conf *conf, struct bio *bio)
+{
+	bool wait = false;
+
+	if (conf->array_frozen || !bio)
+		wait = true;
+	else if (conf->barrier && bio_data_dir(bio) == WRITE) {
+		if (conf->next_resync < RESYNC_WINDOW_SECTORS)
+			wait = true;
+		else if ((conf->next_resync - RESYNC_WINDOW_SECTORS
+				>= bio_end_sector(bio)) ||
+			 (conf->next_resync + NEXT_NORMALIO_DISTANCE
+				<= bio->bi_iter.bi_sector))
+			wait = false;
+		else
+			wait = true;
+	}
+
+	return wait;
+}
+
+static sector_t wait_barrier(struct r1conf *conf, struct bio *bio)
 {
+	sector_t sector = 0;
+
 	spin_lock_irq(&conf->resync_lock);
-	if (conf->barrier) {
+	if (need_to_wait_for_sync(conf, bio)) {
 		conf->nr_waiting++;
-		wait_event_lock_irq(conf->wait_barrier, !conf->barrier,
-				    conf->resync_lock,
-				    raid1_unplug(conf->mddev->queue));
+		/* Wait for the barrier to drop.
+		 * However if there are already pending
+		 * requests (preventing the barrier from
+		 * rising completely), and the
+		 * pre-process bio queue isn't empty,
+		 * then don't wait, as we need to empty
+		 * that queue to get the nr_pending
+		 * count down.
+		 */
+		wait_event_lock_irq(conf->wait_barrier,
+				    !conf->array_frozen &&
+				    (!conf->barrier ||
+				    ((conf->start_next_window <
+				      conf->next_resync + RESYNC_SECTORS) &&
+				     current->bio_list &&
+				     !bio_list_empty(current->bio_list))),
+				    conf->resync_lock);
 		conf->nr_waiting--;
 	}
+
+	if (bio && bio_data_dir(bio) == WRITE) {
+		if (conf->next_resync + NEXT_NORMALIO_DISTANCE
+		    <= bio->bi_iter.bi_sector) {
+			if (conf->start_next_window == MaxSector)
+				conf->start_next_window =
+					conf->next_resync +
+					NEXT_NORMALIO_DISTANCE;
+
+			if ((conf->start_next_window + NEXT_NORMALIO_DISTANCE)
+			    <= bio->bi_iter.bi_sector)
+				conf->next_window_requests++;
+			else
+				conf->current_window_requests++;
+			sector = conf->start_next_window;
+		}
+	}
+
 	conf->nr_pending++;
 	spin_unlock_irq(&conf->resync_lock);
+	return sector;
 }
 
-static void allow_barrier(conf_t *conf)
+static void allow_barrier(struct r1conf *conf, sector_t start_next_window,
+			  sector_t bi_sector)
 {
 	unsigned long flags;
+
 	spin_lock_irqsave(&conf->resync_lock, flags);
 	conf->nr_pending--;
+	if (start_next_window) {
+		if (start_next_window == conf->start_next_window) {
+			if (conf->start_next_window + NEXT_NORMALIO_DISTANCE
+			    <= bi_sector)
+				conf->next_window_requests--;
+			else
+				conf->current_window_requests--;
+		} else
+			conf->current_window_requests--;
+
+		if (!conf->current_window_requests) {
+			if (conf->next_window_requests) {
+				conf->current_window_requests =
+					conf->next_window_requests;
+				conf->next_window_requests = 0;
+				conf->start_next_window +=
+					NEXT_NORMALIO_DISTANCE;
+			} else
+				conf->start_next_window = MaxSector;
+		}
+	}
 	spin_unlock_irqrestore(&conf->resync_lock, flags);
 	wake_up(&conf->wait_barrier);
 }
 
-static void freeze_array(conf_t *conf)
+static void freeze_array(struct r1conf *conf, int extra)
 {
 	/* stop syncio and normal IO and wait for everything to
 	 * go quite.
-	 * We increment barrier and nr_waiting, and then
-	 * wait until nr_pending match nr_queued+1
+	 * We wait until nr_pending match nr_queued+extra
 	 * This is called in the context of one normal IO request
 	 * that has failed. Thus any sync request that might be pending
 	 * will be blocked by nr_pending, and we need to wait for
 	 * pending IO requests to complete or be queued for re-try.
-	 * Thus the number queued (nr_queued) plus this request (1)
+	 * Thus the number queued (nr_queued) plus this request (extra)
 	 * must match the number of pending IOs (nr_pending) before
 	 * we continue.
 	 */
 	spin_lock_irq(&conf->resync_lock);
-	conf->barrier++;
-	conf->nr_waiting++;
-	wait_event_lock_irq(conf->wait_barrier,
-			    conf->nr_pending == conf->nr_queued+1,
-			    conf->resync_lock,
-			    ({ flush_pending_writes(conf);
-			       raid1_unplug(conf->mddev->queue); }));
+	conf->array_frozen = 1;
+	wait_event_lock_irq_cmd(conf->wait_barrier,
+				conf->nr_pending == conf->nr_queued+extra,
+				conf->resync_lock,
+				flush_pending_writes(conf));
 	spin_unlock_irq(&conf->resync_lock);
 }
-static void unfreeze_array(conf_t *conf)
+static void unfreeze_array(struct r1conf *conf)
 {
 	/* reverse the effect of the freeze */
 	spin_lock_irq(&conf->resync_lock);
-	conf->barrier--;
-	conf->nr_waiting--;
+	conf->array_frozen = 0;
 	wake_up(&conf->wait_barrier);
 	spin_unlock_irq(&conf->resync_lock);
 }
 
 
-/* duplicate the data pages for behind I/O */
-static struct page **alloc_behind_pages(struct bio *bio)
+/* duplicate the data pages for behind I/O 
+ */
+static void alloc_behind_pages(struct bio *bio, struct r1bio *r1_bio)
 {
 	int i;
 	struct bio_vec *bvec;
-	struct page **pages = kzalloc(bio->bi_vcnt * sizeof(struct page *),
+	struct bio_vec *bvecs = kzalloc(bio->bi_vcnt * sizeof(struct bio_vec),
 					GFP_NOIO);
-	if (unlikely(!pages))
-		goto do_sync_io;
+	if (unlikely(!bvecs))
+		return;
 
-	bio_for_each_segment(bvec, bio, i) {
-		pages[i] = alloc_page(GFP_NOIO);
-		if (unlikely(!pages[i]))
+	bio_for_each_segment_all(bvec, bio, i) {
+		bvecs[i] = *bvec;
+		bvecs[i].bv_page = alloc_page(GFP_NOIO);
+		if (unlikely(!bvecs[i].bv_page))
 			goto do_sync_io;
-		memcpy(kmap(pages[i]) + bvec->bv_offset,
-			kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
-		kunmap(pages[i]);
+		memcpy(kmap(bvecs[i].bv_page) + bvec->bv_offset,
+		       kmap(bvec->bv_page) + bvec->bv_offset, bvec->bv_len);
+		kunmap(bvecs[i].bv_page);
 		kunmap(bvec->bv_page);
 	}
-
-	return pages;
+	r1_bio->behind_bvecs = bvecs;
+	r1_bio->behind_page_count = bio->bi_vcnt;
+	set_bit(R1BIO_BehindIO, &r1_bio->state);
+	return;
 
 do_sync_io:
-	if (pages)
-		for (i = 0; i < bio->bi_vcnt && pages[i]; i++)
-			put_page(pages[i]);
-	kfree(pages);
-	PRINTK("%dB behind alloc failed, doing sync I/O\n", bio->bi_size);
-	return NULL;
+	for (i = 0; i < bio->bi_vcnt; i++)
+		if (bvecs[i].bv_page)
+			put_page(bvecs[i].bv_page);
+	kfree(bvecs);
+	pr_debug("%dB behind alloc failed, doing sync I/O\n",
+		 bio->bi_iter.bi_size);
 }
 
-static int make_request(struct request_queue *q, struct bio * bio)
+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)
 {
-	mddev_t *mddev = q->queuedata;
-	conf_t *conf = mddev_to_conf(mddev);
-	mirror_info_t *mirror;
-	r1bio_t *r1_bio;
+	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 || current->bio_list) {
+		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);
+		wake_up(&conf->wait_barrier);
+		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;
+		if (unlikely((bio->bi_rw & REQ_DISCARD) &&
+		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+			/* Just ignore it */
+			bio_endio(bio, 0);
+		else
+			generic_make_request(bio);
+		bio = next;
+	}
+	kfree(plug);
+}
+
+static void make_request(struct mddev *mddev, struct bio * bio)
+{
+	struct r1conf *conf = mddev->private;
+	struct raid1_info *mirror;
+	struct r1bio *r1_bio;
 	struct bio *read_bio;
-	int i, targets = 0, disks;
+	int i, disks;
 	struct bitmap *bitmap;
 	unsigned long flags;
-	struct bio_list bl;
-	struct page **behind_pages = NULL;
 	const int rw = bio_data_dir(bio);
-	const int do_sync = bio_sync(bio);
-	int do_barriers;
-	mdk_rdev_t *blocked_rdev;
+	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
+	const unsigned long do_flush_fua = (bio->bi_rw & (REQ_FLUSH | REQ_FUA));
+	const unsigned long do_discard = (bio->bi_rw
+					  & (REQ_DISCARD | REQ_SECURE));
+	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
+	struct md_rdev *blocked_rdev;
+	struct blk_plug_cb *cb;
+	struct raid1_plug_cb *plug = NULL;
+	int first_clone;
+	int sectors_handled;
+	int max_sectors;
+	sector_t start_next_window;
 
 	/*
 	 * Register the new request and wait if the reconstruction
 	 * thread has put up a bar for new requests.
 	 * Continue immediately if no resync is active currently.
-	 * We test barriers_work *after* md_write_start as md_write_start
-	 * may cause the first superblock write, and that will check out
-	 * if barriers work.
 	 */
 
 	md_write_start(mddev, bio); /* wait on superblock update early */
 
-	if (unlikely(!mddev->barriers_work && bio_barrier(bio))) {
-		if (rw == WRITE)
-			md_write_end(mddev);
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+	if (bio_data_dir(bio) == WRITE &&
+	    bio_end_sector(bio) > mddev->suspend_lo &&
+	    bio->bi_iter.bi_sector < mddev->suspend_hi) {
+		/* As the suspend_* range is controlled by
+		 * userspace, we want an interruptible
+		 * wait.
+		 */
+		DEFINE_WAIT(w);
+		for (;;) {
+			flush_signals(current);
+			prepare_to_wait(&conf->wait_barrier,
+					&w, TASK_INTERRUPTIBLE);
+			if (bio_end_sector(bio) <= mddev->suspend_lo ||
+			    bio->bi_iter.bi_sector >= mddev->suspend_hi)
+				break;
+			schedule();
+		}
+		finish_wait(&conf->wait_barrier, &w);
 	}
 
-	wait_barrier(conf);
+	start_next_window = wait_barrier(conf, bio);
 
 	bitmap = mddev->bitmap;
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
-
 	/*
 	 * make_request() can abort the operation when READA is being
 	 * used and no empty request is available.
@@ -815,183 +1145,325 @@ static int make_request(struct request_queue *q, struct bio * bio)
 	r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);
 
 	r1_bio->master_bio = bio;
-	r1_bio->sectors = bio->bi_size >> 9;
+	r1_bio->sectors = bio_sectors(bio);
 	r1_bio->state = 0;
 	r1_bio->mddev = mddev;
-	r1_bio->sector = bio->bi_sector;
+	r1_bio->sector = bio->bi_iter.bi_sector;
+
+	/* We might need to issue multiple reads to different
+	 * devices if there are bad blocks around, so we keep
+	 * track of the number of reads in bio->bi_phys_segments.
+	 * If this is 0, there is only one r1_bio and no locking
+	 * will be needed when requests complete.  If it is
+	 * non-zero, then it is the number of not-completed requests.
+	 */
+	bio->bi_phys_segments = 0;
+	clear_bit(BIO_SEG_VALID, &bio->bi_flags);
 
 	if (rw == READ) {
 		/*
 		 * read balancing logic:
 		 */
-		int rdisk = read_balance(conf, r1_bio);
+		int rdisk;
+
+read_again:
+		rdisk = read_balance(conf, r1_bio, &max_sectors);
 
 		if (rdisk < 0) {
 			/* couldn't find anywhere to read from */
 			raid_end_bio_io(r1_bio);
-			return 0;
+			return;
 		}
 		mirror = conf->mirrors + rdisk;
 
+		if (test_bit(WriteMostly, &mirror->rdev->flags) &&
+		    bitmap) {
+			/* Reading from a write-mostly device must
+			 * take care not to over-take any writes
+			 * that are 'behind'
+			 */
+			wait_event(bitmap->behind_wait,
+				   atomic_read(&bitmap->behind_writes) == 0);
+		}
 		r1_bio->read_disk = rdisk;
 
-		read_bio = bio_clone(bio, GFP_NOIO);
+		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		bio_trim(read_bio, r1_bio->sector - bio->bi_iter.bi_sector,
+			 max_sectors);
 
 		r1_bio->bios[rdisk] = read_bio;
 
-		read_bio->bi_sector = r1_bio->sector + mirror->rdev->data_offset;
+		read_bio->bi_iter.bi_sector = r1_bio->sector +
+			mirror->rdev->data_offset;
 		read_bio->bi_bdev = mirror->rdev->bdev;
 		read_bio->bi_end_io = raid1_end_read_request;
 		read_bio->bi_rw = READ | do_sync;
 		read_bio->bi_private = r1_bio;
 
-		generic_make_request(read_bio);
-		return 0;
+		if (max_sectors < r1_bio->sectors) {
+			/* could not read all from this device, so we will
+			 * need another r1_bio.
+			 */
+
+			sectors_handled = (r1_bio->sector + max_sectors
+					   - bio->bi_iter.bi_sector);
+			r1_bio->sectors = max_sectors;
+			spin_lock_irq(&conf->device_lock);
+			if (bio->bi_phys_segments == 0)
+				bio->bi_phys_segments = 2;
+			else
+				bio->bi_phys_segments++;
+			spin_unlock_irq(&conf->device_lock);
+			/* Cannot call generic_make_request directly
+			 * as that will be queued in __make_request
+			 * and subsequent mempool_alloc might block waiting
+			 * for it.  So hand bio over to raid1d.
+			 */
+			reschedule_retry(r1_bio);
+
+			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);
+
+			r1_bio->master_bio = bio;
+			r1_bio->sectors = bio_sectors(bio) - sectors_handled;
+			r1_bio->state = 0;
+			r1_bio->mddev = mddev;
+			r1_bio->sector = bio->bi_iter.bi_sector +
+				sectors_handled;
+			goto read_again;
+		} else
+			generic_make_request(read_bio);
+		return;
 	}
 
 	/*
 	 * WRITE:
 	 */
-	/* first select target devices under spinlock and
+	if (conf->pending_count >= max_queued_requests) {
+		md_wakeup_thread(mddev->thread);
+		wait_event(conf->wait_barrier,
+			   conf->pending_count < max_queued_requests);
+	}
+	/* first select target devices under rcu_lock and
 	 * inc refcount on their rdev.  Record them by setting
 	 * bios[x] to bio
+	 * If there are known/acknowledged bad blocks on any device on
+	 * which we have seen a write error, we want to avoid writing those
+	 * blocks.
+	 * This potentially requires several writes to write around
+	 * the bad blocks.  Each set of writes gets it's own r1bio
+	 * with a set of bios attached.
 	 */
-	disks = conf->raid_disks;
-#if 0
-	{ static int first=1;
-	if (first) printk("First Write sector %llu disks %d\n",
-			  (unsigned long long)r1_bio->sector, disks);
-	first = 0;
-	}
-#endif
+
+	disks = conf->raid_disks * 2;
  retry_write:
+	r1_bio->start_next_window = start_next_window;
 	blocked_rdev = NULL;
 	rcu_read_lock();
+	max_sectors = r1_bio->sectors;
 	for (i = 0;  i < disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
 			atomic_inc(&rdev->nr_pending);
 			blocked_rdev = rdev;
 			break;
 		}
-		if (rdev && !test_bit(Faulty, &rdev->flags)) {
-			atomic_inc(&rdev->nr_pending);
-			if (test_bit(Faulty, &rdev->flags)) {
+		r1_bio->bios[i] = NULL;
+		if (!rdev || test_bit(Faulty, &rdev->flags)
+		    || test_bit(Unmerged, &rdev->flags)) {
+			if (i < conf->raid_disks)
+				set_bit(R1BIO_Degraded, &r1_bio->state);
+			continue;
+		}
+
+		atomic_inc(&rdev->nr_pending);
+		if (test_bit(WriteErrorSeen, &rdev->flags)) {
+			sector_t first_bad;
+			int bad_sectors;
+			int is_bad;
+
+			is_bad = is_badblock(rdev, r1_bio->sector,
+					     max_sectors,
+					     &first_bad, &bad_sectors);
+			if (is_bad < 0) {
+				/* mustn't write here until the bad block is
+				 * acknowledged*/
+				set_bit(BlockedBadBlocks, &rdev->flags);
+				blocked_rdev = rdev;
+				break;
+			}
+			if (is_bad && first_bad <= r1_bio->sector) {
+				/* Cannot write here at all */
+				bad_sectors -= (r1_bio->sector - first_bad);
+				if (bad_sectors < max_sectors)
+					/* mustn't write more than bad_sectors
+					 * to other devices yet
+					 */
+					max_sectors = bad_sectors;
 				rdev_dec_pending(rdev, mddev);
-				r1_bio->bios[i] = NULL;
-			} else
-				r1_bio->bios[i] = bio;
-			targets++;
-		} else
-			r1_bio->bios[i] = NULL;
+				/* We don't set R1BIO_Degraded as that
+				 * only applies if the disk is
+				 * missing, so it might be re-added,
+				 * and we want to know to recover this
+				 * chunk.
+				 * In this case the device is here,
+				 * and the fact that this chunk is not
+				 * in-sync is recorded in the bad
+				 * block log
+				 */
+				continue;
+			}
+			if (is_bad) {
+				int good_sectors = first_bad - r1_bio->sector;
+				if (good_sectors < max_sectors)
+					max_sectors = good_sectors;
+			}
+		}
+		r1_bio->bios[i] = bio;
 	}
 	rcu_read_unlock();
 
 	if (unlikely(blocked_rdev)) {
 		/* Wait for this device to become unblocked */
 		int j;
+		sector_t old = start_next_window;
 
 		for (j = 0; j < i; j++)
 			if (r1_bio->bios[j])
 				rdev_dec_pending(conf->mirrors[j].rdev, mddev);
-
-		allow_barrier(conf);
+		r1_bio->state = 0;
+		allow_barrier(conf, start_next_window, bio->bi_iter.bi_sector);
 		md_wait_for_blocked_rdev(blocked_rdev, mddev);
-		wait_barrier(conf);
+		start_next_window = wait_barrier(conf, bio);
+		/*
+		 * We must make sure the multi r1bios of bio have
+		 * the same value of bi_phys_segments
+		 */
+		if (bio->bi_phys_segments && old &&
+		    old != start_next_window)
+			/* Wait for the former r1bio(s) to complete */
+			wait_event(conf->wait_barrier,
+				   bio->bi_phys_segments == 1);
 		goto retry_write;
 	}
 
-	BUG_ON(targets == 0); /* we never fail the last device */
-
-	if (targets < conf->raid_disks) {
-		/* array is degraded, we will not clear the bitmap
-		 * on I/O completion (see raid1_end_write_request) */
-		set_bit(R1BIO_Degraded, &r1_bio->state);
+	if (max_sectors < r1_bio->sectors) {
+		/* We are splitting this write into multiple parts, so
+		 * we need to prepare for allocating another r1_bio.
+		 */
+		r1_bio->sectors = max_sectors;
+		spin_lock_irq(&conf->device_lock);
+		if (bio->bi_phys_segments == 0)
+			bio->bi_phys_segments = 2;
+		else
+			bio->bi_phys_segments++;
+		spin_unlock_irq(&conf->device_lock);
 	}
+	sectors_handled = r1_bio->sector + max_sectors - bio->bi_iter.bi_sector;
 
-	/* do behind I/O ? */
-	if (bitmap &&
-	    atomic_read(&bitmap->behind_writes) < bitmap->max_write_behind &&
-	    (behind_pages = alloc_behind_pages(bio)) != NULL)
-		set_bit(R1BIO_BehindIO, &r1_bio->state);
-
-	atomic_set(&r1_bio->remaining, 0);
+	atomic_set(&r1_bio->remaining, 1);
 	atomic_set(&r1_bio->behind_remaining, 0);
 
-	do_barriers = bio_barrier(bio);
-	if (do_barriers)
-		set_bit(R1BIO_Barrier, &r1_bio->state);
-
-	bio_list_init(&bl);
+	first_clone = 1;
 	for (i = 0; i < disks; i++) {
 		struct bio *mbio;
 		if (!r1_bio->bios[i])
 			continue;
 
-		mbio = bio_clone(bio, GFP_NOIO);
-		r1_bio->bios[i] = mbio;
-
-		mbio->bi_sector	= r1_bio->sector + conf->mirrors[i].rdev->data_offset;
-		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
-		mbio->bi_end_io	= raid1_end_write_request;
-		mbio->bi_rw = WRITE | do_barriers | do_sync;
-		mbio->bi_private = r1_bio;
-
-		if (behind_pages) {
+		mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		bio_trim(mbio, r1_bio->sector - bio->bi_iter.bi_sector, max_sectors);
+
+		if (first_clone) {
+			/* do behind I/O ?
+			 * Not if there are too many, or cannot
+			 * allocate memory, or a reader on WriteMostly
+			 * is waiting for behind writes to flush */
+			if (bitmap &&
+			    (atomic_read(&bitmap->behind_writes)
+			     < mddev->bitmap_info.max_write_behind) &&
+			    !waitqueue_active(&bitmap->behind_wait))
+				alloc_behind_pages(mbio, r1_bio);
+
+			bitmap_startwrite(bitmap, r1_bio->sector,
+					  r1_bio->sectors,
+					  test_bit(R1BIO_BehindIO,
+						   &r1_bio->state));
+			first_clone = 0;
+		}
+		if (r1_bio->behind_bvecs) {
 			struct bio_vec *bvec;
 			int j;
 
-			/* Yes, I really want the '__' version so that
-			 * we clear any unused pointer in the io_vec, rather
-			 * than leave them unchanged.  This is important
-			 * because when we come to free the pages, we won't
-			 * know the originial bi_idx, so we just free
-			 * them all
+			/*
+			 * We trimmed the bio, so _all is legit
 			 */
-			__bio_for_each_segment(bvec, mbio, j, 0)
-				bvec->bv_page = behind_pages[j];
+			bio_for_each_segment_all(bvec, mbio, j)
+				bvec->bv_page = r1_bio->behind_bvecs[j].bv_page;
 			if (test_bit(WriteMostly, &conf->mirrors[i].rdev->flags))
 				atomic_inc(&r1_bio->behind_remaining);
 		}
 
+		r1_bio->bios[i] = mbio;
+
+		mbio->bi_iter.bi_sector	= (r1_bio->sector +
+				   conf->mirrors[i].rdev->data_offset);
+		mbio->bi_bdev = conf->mirrors[i].rdev->bdev;
+		mbio->bi_end_io	= raid1_end_write_request;
+		mbio->bi_rw =
+			WRITE | do_flush_fua | do_sync | do_discard | do_same;
+		mbio->bi_private = r1_bio;
+
 		atomic_inc(&r1_bio->remaining);
 
-		bio_list_add(&bl, mbio);
+		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);
+		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.
+	 */
+	if (sectors_handled < bio_sectors(bio)) {
+		r1_bio_write_done(r1_bio);
+		/* We need another r1_bio.  It has already been counted
+		 * in bio->bi_phys_segments
+		 */
+		r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);
+		r1_bio->master_bio = bio;
+		r1_bio->sectors = bio_sectors(bio) - sectors_handled;
+		r1_bio->state = 0;
+		r1_bio->mddev = mddev;
+		r1_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
+		goto retry_write;
 	}
-	kfree(behind_pages); /* the behind pages are attached to the bios now */
 
-	bitmap_startwrite(bitmap, bio->bi_sector, r1_bio->sectors,
-				test_bit(R1BIO_BehindIO, &r1_bio->state));
-	spin_lock_irqsave(&conf->device_lock, flags);
-	bio_list_merge(&conf->pending_bio_list, &bl);
-	bio_list_init(&bl);
+	r1_bio_write_done(r1_bio);
 
-	blk_plug_device(mddev->queue);
-	spin_unlock_irqrestore(&conf->device_lock, flags);
-
-	/* In case raid1d snuck into freeze_array */
+	/* In case raid1d snuck in to freeze_array */
 	wake_up(&conf->wait_barrier);
-
-	if (do_sync)
-		md_wakeup_thread(mddev->thread);
-#if 0
-	while ((bio = bio_list_pop(&bl)) != NULL)
-		generic_make_request(bio);
-#endif
-
-	return 0;
 }
 
-static void status(struct seq_file *seq, mddev_t *mddev)
+static void status(struct seq_file *seq, struct mddev *mddev)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 	int i;
 
 	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
 		   conf->raid_disks - mddev->degraded);
 	rcu_read_lock();
 	for (i = 0; i < conf->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		seq_printf(seq, "%s",
 			   rdev && test_bit(In_sync, &rdev->flags) ? "U" : "_");
 	}
@@ -1000,10 +1472,10 @@ static void status(struct seq_file *seq, mddev_t *mddev)
 }
 
 
-static void error(mddev_t *mddev, mdk_rdev_t *rdev)
+static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 
 	/*
 	 * If it is not operational, then we have already marked it as dead
@@ -1012,12 +1484,17 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 	 * else mark the drive as failed
 	 */
 	if (test_bit(In_sync, &rdev->flags)
-	    && (conf->raid_disks - mddev->degraded) == 1)
+	    && (conf->raid_disks - mddev->degraded) == 1) {
 		/*
 		 * Don't fail the drive, act as though we were just a
-		 * normal single drive
+		 * normal single drive.
+		 * However don't try a recovery from this drive as
+		 * it is very likely to fail.
 		 */
+		conf->recovery_disabled = mddev->recovery_disabled;
 		return;
+	}
+	set_bit(Blocked, &rdev->flags);
 	if (test_and_clear_bit(In_sync, &rdev->flags)) {
 		unsigned long flags;
 		spin_lock_irqsave(&conf->device_lock, flags);
@@ -1031,29 +1508,31 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 	} else
 		set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
-	printk(KERN_ALERT "raid1: Disk failure on %s, disabling device.\n"
-		"raid1: Operation continuing on %d devices.\n",
-		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
+	printk(KERN_ALERT
+	       "md/raid1:%s: Disk failure on %s, disabling device.\n"
+	       "md/raid1:%s: Operation continuing on %d devices.\n",
+	       mdname(mddev), bdevname(rdev->bdev, b),
+	       mdname(mddev), conf->raid_disks - mddev->degraded);
 }
 
-static void print_conf(conf_t *conf)
+static void print_conf(struct r1conf *conf)
 {
 	int i;
 
-	printk("RAID1 conf printout:\n");
+	printk(KERN_DEBUG "RAID1 conf printout:\n");
 	if (!conf) {
-		printk("(!conf)\n");
+		printk(KERN_DEBUG "(!conf)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
+	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
 		conf->raid_disks);
 
 	rcu_read_lock();
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev)
-			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
+			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
 			       i, !test_bit(In_sync, &rdev->flags),
 			       !test_bit(Faulty, &rdev->flags),
 			       bdevname(rdev->bdev,b));
@@ -1061,19 +1540,24 @@ static void print_conf(conf_t *conf)
 	rcu_read_unlock();
 }
 
-static void close_sync(conf_t *conf)
+static void close_sync(struct r1conf *conf)
 {
-	wait_barrier(conf);
-	allow_barrier(conf);
+	wait_barrier(conf, NULL);
+	allow_barrier(conf, 0, 0);
 
 	mempool_destroy(conf->r1buf_pool);
 	conf->r1buf_pool = NULL;
+
+	conf->next_resync = 0;
+	conf->start_next_window = MaxSector;
 }
 
-static int raid1_spare_active(mddev_t *mddev)
+static int raid1_spare_active(struct mddev *mddev)
 {
 	int i;
-	conf_t *conf = mddev->private;
+	struct r1conf *conf = mddev->private;
+	int count = 0;
+	unsigned long flags;
 
 	/*
 	 * Find all failed disks within the RAID1 configuration 
@@ -1081,46 +1565,71 @@ static int raid1_spare_active(mddev_t *mddev)
 	 * Called under mddev lock, so rcu protection not needed.
 	 */
 	for (i = 0; i < conf->raid_disks; i++) {
-		mdk_rdev_t *rdev = conf->mirrors[i].rdev;
+		struct md_rdev *rdev = conf->mirrors[i].rdev;
+		struct md_rdev *repl = conf->mirrors[conf->raid_disks + i].rdev;
+		if (repl
+		    && repl->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &repl->flags)
+		    && !test_and_set_bit(In_sync, &repl->flags)) {
+			/* replacement has just become active */
+			if (!rdev ||
+			    !test_and_clear_bit(In_sync, &rdev->flags))
+				count++;
+			if (rdev) {
+				/* Replaced device not technically
+				 * faulty, but we need to be sure
+				 * it gets removed and never re-added
+				 */
+				set_bit(Faulty, &rdev->flags);
+				sysfs_notify_dirent_safe(
+					rdev->sysfs_state);
+			}
+		}
 		if (rdev
+		    && rdev->recovery_offset == MaxSector
 		    && !test_bit(Faulty, &rdev->flags)
 		    && !test_and_set_bit(In_sync, &rdev->flags)) {
-			unsigned long flags;
-			spin_lock_irqsave(&conf->device_lock, flags);
-			mddev->degraded--;
-			spin_unlock_irqrestore(&conf->device_lock, flags);
+			count++;
+			sysfs_notify_dirent_safe(rdev->sysfs_state);
 		}
 	}
+	spin_lock_irqsave(&conf->device_lock, flags);
+	mddev->degraded -= count;
+	spin_unlock_irqrestore(&conf->device_lock, flags);
 
 	print_conf(conf);
-	return 0;
+	return count;
 }
 
 
-static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
+static int raid1_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	conf_t *conf = mddev->private;
+	struct r1conf *conf = mddev->private;
 	int err = -EEXIST;
 	int mirror = 0;
-	mirror_info_t *p;
+	struct raid1_info *p;
 	int first = 0;
-	int last = mddev->raid_disks - 1;
+	int last = conf->raid_disks - 1;
+	struct request_queue *q = bdev_get_queue(rdev->bdev);
+
+	if (mddev->recovery_disabled == conf->recovery_disabled)
+		return -EBUSY;
 
 	if (rdev->raid_disk >= 0)
 		first = last = rdev->raid_disk;
 
-	for (mirror = first; mirror <= last; mirror++)
-		if ( !(p=conf->mirrors+mirror)->rdev) {
+	if (q->merge_bvec_fn) {
+		set_bit(Unmerged, &rdev->flags);
+		mddev->merge_check_needed = 1;
+	}
 
-			blk_queue_stack_limits(mddev->queue,
-					       rdev->bdev->bd_disk->queue);
-			/* as we don't honour merge_bvec_fn, we must never risk
-			 * violating it, so limit ->max_sector to one PAGE, as
-			 * a one page request is never in violation.
-			 */
-			if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-				blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
+	for (mirror = first; mirror <= last; mirror++) {
+		p = conf->mirrors+mirror;
+		if (!p->rdev) {
+
+			if (mddev->gendisk)
+				disk_stack_limits(mddev->gendisk, rdev->bdev,
+						  rdev->data_offset << 9);
 
 			p->head_position = 0;
 			rdev->raid_disk = mirror;
@@ -1133,30 +1642,60 @@ static int raid1_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
 			rcu_assign_pointer(p->rdev, rdev);
 			break;
 		}
-
+		if (test_bit(WantReplacement, &p->rdev->flags) &&
+		    p[conf->raid_disks].rdev == NULL) {
+			/* Add this device as a replacement */
+			clear_bit(In_sync, &rdev->flags);
+			set_bit(Replacement, &rdev->flags);
+			rdev->raid_disk = mirror;
+			err = 0;
+			conf->fullsync = 1;
+			rcu_assign_pointer(p[conf->raid_disks].rdev, rdev);
+			break;
+		}
+	}
+	if (err == 0 && test_bit(Unmerged, &rdev->flags)) {
+		/* Some requests might not have seen this new
+		 * merge_bvec_fn.  We must wait for them to complete
+		 * before merging the device fully.
+		 * First we make sure any code which has tested
+		 * our function has submitted the request, then
+		 * we wait for all outstanding requests to complete.
+		 */
+		synchronize_sched();
+		freeze_array(conf, 0);
+		unfreeze_array(conf);
+		clear_bit(Unmerged, &rdev->flags);
+	}
+	md_integrity_add_rdev(rdev, mddev);
+	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
 	print_conf(conf);
 	return err;
 }
 
-static int raid1_remove_disk(mddev_t *mddev, int number)
+static int raid1_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	conf_t *conf = mddev->private;
+	struct r1conf *conf = mddev->private;
 	int err = 0;
-	mdk_rdev_t *rdev;
-	mirror_info_t *p = conf->mirrors+ number;
+	int number = rdev->raid_disk;
+	struct raid1_info *p = conf->mirrors + number;
+
+	if (rdev != p->rdev)
+		p = conf->mirrors + conf->raid_disks + number;
 
 	print_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
+	if (rdev == p->rdev) {
 		if (test_bit(In_sync, &rdev->flags) ||
 		    atomic_read(&rdev->nr_pending)) {
 			err = -EBUSY;
 			goto abort;
 		}
-		/* Only remove non-faulty devices is recovery
+		/* Only remove non-faulty devices if recovery
 		 * is not possible.
 		 */
 		if (!test_bit(Faulty, &rdev->flags) &&
+		    mddev->recovery_disabled != conf->recovery_disabled &&
 		    mddev->degraded < conf->raid_disks) {
 			err = -EBUSY;
 			goto abort;
@@ -1167,7 +1706,23 @@ static int raid1_remove_disk(mddev_t *mddev, int number)
 			/* lost the race, try later */
 			err = -EBUSY;
 			p->rdev = rdev;
-		}
+			goto abort;
+		} else if (conf->mirrors[conf->raid_disks + number].rdev) {
+			/* We just removed a device that is being replaced.
+			 * Move down the replacement.  We drain all IO before
+			 * doing this to avoid confusion.
+			 */
+			struct md_rdev *repl =
+				conf->mirrors[conf->raid_disks + number].rdev;
+			freeze_array(conf, 0);
+			clear_bit(Replacement, &repl->flags);
+			p->rdev = repl;
+			conf->mirrors[conf->raid_disks + number].rdev = NULL;
+			unfreeze_array(conf);
+			clear_bit(WantReplacement, &rdev->flags);
+		} else
+			clear_bit(WantReplacement, &rdev->flags);
+		err = md_integrity_register(mddev);
 	}
 abort:
 
@@ -1178,14 +1733,10 @@ abort:
 
 static void end_sync_read(struct bio *bio, int error)
 {
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
-	int i;
+	struct r1bio *r1_bio = bio->bi_private;
+
+	update_head_pos(r1_bio->read_disk, r1_bio);
 
-	for (i=r1_bio->mddev->raid_disks; i--; )
-		if (r1_bio->bios[i] == bio)
-			break;
-	BUG_ON(i < 0);
-	update_head_pos(i, r1_bio);
 	/*
 	 * we have read a block, now it needs to be re-written,
 	 * or re-read if the read failed.
@@ -1201,19 +1752,17 @@ static void end_sync_read(struct bio *bio, int error)
 static void end_sync_write(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r1bio_t * r1_bio = (r1bio_t *)(bio->bi_private);
-	mddev_t *mddev = r1_bio->mddev;
-	conf_t *conf = mddev_to_conf(mddev);
-	int i;
+	struct r1bio *r1_bio = bio->bi_private;
+	struct mddev *mddev = r1_bio->mddev;
+	struct r1conf *conf = mddev->private;
 	int mirror=0;
+	sector_t first_bad;
+	int bad_sectors;
+
+	mirror = find_bio_disk(r1_bio, bio);
 
-	for (i = 0; i < conf->raid_disks; i++)
-		if (r1_bio->bios[i] == bio) {
-			mirror = i;
-			break;
-		}
 	if (!uptodate) {
-		int sync_blocks = 0;
+		sector_t sync_blocks = 0;
 		sector_t s = r1_bio->sector;
 		long sectors_to_go = r1_bio->sectors;
 		/* make sure these bits doesn't get cleared. */
@@ -1223,207 +1772,288 @@ static void end_sync_write(struct bio *bio, int error)
 			s += sync_blocks;
 			sectors_to_go -= sync_blocks;
 		} while (sectors_to_go > 0);
-		md_error(mddev, conf->mirrors[mirror].rdev);
-	}
-
-	update_head_pos(mirror, r1_bio);
+		set_bit(WriteErrorSeen,
+			&conf->mirrors[mirror].rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &conf->mirrors[mirror].rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				mddev->recovery);
+		set_bit(R1BIO_WriteError, &r1_bio->state);
+	} else if (is_badblock(conf->mirrors[mirror].rdev,
+			       r1_bio->sector,
+			       r1_bio->sectors,
+			       &first_bad, &bad_sectors) &&
+		   !is_badblock(conf->mirrors[r1_bio->read_disk].rdev,
+				r1_bio->sector,
+				r1_bio->sectors,
+				&first_bad, &bad_sectors)
+		)
+		set_bit(R1BIO_MadeGood, &r1_bio->state);
 
 	if (atomic_dec_and_test(&r1_bio->remaining)) {
-		md_done_sync(mddev, r1_bio->sectors, uptodate);
-		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, uptodate);
+		}
 	}
 }
 
-static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
+static int r1_sync_page_io(struct md_rdev *rdev, sector_t sector,
+			    int sectors, struct page *page, int rw)
 {
-	conf_t *conf = mddev_to_conf(mddev);
-	int i;
-	int disks = conf->raid_disks;
-	struct bio *bio, *wbio;
-
-	bio = r1_bio->bios[r1_bio->read_disk];
+	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
+		/* success */
+		return 1;
+	if (rw == WRITE) {
+		set_bit(WriteErrorSeen, &rdev->flags);
+		if (!test_and_set_bit(WantReplacement,
+				      &rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED, &
+				rdev->mddev->recovery);
+	}
+	/* need to record an error - either for the block or the device */
+	if (!rdev_set_badblocks(rdev, sector, sectors, 0))
+		md_error(rdev->mddev, rdev);
+	return 0;
+}
 
+static int fix_sync_read_error(struct r1bio *r1_bio)
+{
+	/* Try some synchronous reads of other devices to get
+	 * good data, much like with normal read errors.  Only
+	 * read into the pages we already have so we don't
+	 * need to re-issue the read request.
+	 * We don't need to freeze the array, because being in an
+	 * active sync request, there is no normal IO, and
+	 * no overlapping syncs.
+	 * We don't need to check is_badblock() again as we
+	 * made sure that anything with a bad block in range
+	 * will have bi_end_io clear.
+	 */
+	struct mddev *mddev = r1_bio->mddev;
+	struct r1conf *conf = mddev->private;
+	struct bio *bio = r1_bio->bios[r1_bio->read_disk];
+	sector_t sect = r1_bio->sector;
+	int sectors = r1_bio->sectors;
+	int idx = 0;
 
-	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
-		/* We have read all readable devices.  If we haven't
-		 * got the block, then there is no hope left.
-		 * If we have, then we want to do a comparison
-		 * and skip the write if everything is the same.
-		 * If any blocks failed to read, then we need to
-		 * attempt an over-write
-		 */
-		int primary;
-		if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
-			for (i=0; i<mddev->raid_disks; i++)
-				if (r1_bio->bios[i]->bi_end_io == end_sync_read)
-					md_error(mddev, conf->mirrors[i].rdev);
+	while(sectors) {
+		int s = sectors;
+		int d = r1_bio->read_disk;
+		int success = 0;
+		struct md_rdev *rdev;
+		int start;
 
-			md_done_sync(mddev, r1_bio->sectors, 1);
-			put_buf(r1_bio);
-			return;
-		}
-		for (primary=0; primary<mddev->raid_disks; primary++)
-			if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
-			    test_bit(BIO_UPTODATE, &r1_bio->bios[primary]->bi_flags)) {
-				r1_bio->bios[primary]->bi_end_io = NULL;
-				rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
-				break;
+		if (s > (PAGE_SIZE>>9))
+			s = PAGE_SIZE >> 9;
+		do {
+			if (r1_bio->bios[d]->bi_end_io == end_sync_read) {
+				/* No rcu protection needed here devices
+				 * can only be removed when no resync is
+				 * active, and resync is currently active
+				 */
+				rdev = conf->mirrors[d].rdev;
+				if (sync_page_io(rdev, sect, s<<9,
+						 bio->bi_io_vec[idx].bv_page,
+						 READ, false)) {
+					success = 1;
+					break;
+				}
 			}
-		r1_bio->read_disk = primary;
-		for (i=0; i<mddev->raid_disks; i++)
-			if (r1_bio->bios[i]->bi_end_io == end_sync_read) {
-				int j;
-				int vcnt = r1_bio->sectors >> (PAGE_SHIFT- 9);
-				struct bio *pbio = r1_bio->bios[primary];
-				struct bio *sbio = r1_bio->bios[i];
-
-				if (test_bit(BIO_UPTODATE, &sbio->bi_flags)) {
-					for (j = vcnt; j-- ; ) {
-						struct page *p, *s;
-						p = pbio->bi_io_vec[j].bv_page;
-						s = sbio->bi_io_vec[j].bv_page;
-						if (memcmp(page_address(p),
-							   page_address(s),
-							   PAGE_SIZE))
-							break;
-					}
-				} else
-					j = 0;
-				if (j >= 0)
-					mddev->resync_mismatches += r1_bio->sectors;
-				if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
-					      && test_bit(BIO_UPTODATE, &sbio->bi_flags))) {
-					sbio->bi_end_io = NULL;
-					rdev_dec_pending(conf->mirrors[i].rdev, mddev);
-				} else {
-					/* fixup the bio for reuse */
-					int size;
-					sbio->bi_vcnt = vcnt;
-					sbio->bi_size = r1_bio->sectors << 9;
-					sbio->bi_idx = 0;
-					sbio->bi_phys_segments = 0;
-					sbio->bi_hw_segments = 0;
-					sbio->bi_hw_front_size = 0;
-					sbio->bi_hw_back_size = 0;
-					sbio->bi_flags &= ~(BIO_POOL_MASK - 1);
-					sbio->bi_flags |= 1 << BIO_UPTODATE;
-					sbio->bi_next = NULL;
-					sbio->bi_sector = r1_bio->sector +
-						conf->mirrors[i].rdev->data_offset;
-					sbio->bi_bdev = conf->mirrors[i].rdev->bdev;
-					size = sbio->bi_size;
-					for (j = 0; j < vcnt ; j++) {
-						struct bio_vec *bi;
-						bi = &sbio->bi_io_vec[j];
-						bi->bv_offset = 0;
-						if (size > PAGE_SIZE)
-							bi->bv_len = PAGE_SIZE;
-						else
-							bi->bv_len = size;
-						size -= PAGE_SIZE;
-						memcpy(page_address(bi->bv_page),
-						       page_address(pbio->bi_io_vec[j].bv_page),
-						       PAGE_SIZE);
-					}
+			d++;
+			if (d == conf->raid_disks * 2)
+				d = 0;
+		} while (!success && d != r1_bio->read_disk);
 
-				}
+		if (!success) {
+			char b[BDEVNAME_SIZE];
+			int abort = 0;
+			/* Cannot read from anywhere, this block is lost.
+			 * Record a bad block on each device.  If that doesn't
+			 * work just disable and interrupt the recovery.
+			 * Don't fail devices as that won't really help.
+			 */
+			printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O read error"
+			       " for block %llu\n",
+			       mdname(mddev),
+			       bdevname(bio->bi_bdev, b),
+			       (unsigned long long)r1_bio->sector);
+			for (d = 0; d < conf->raid_disks * 2; d++) {
+				rdev = conf->mirrors[d].rdev;
+				if (!rdev || test_bit(Faulty, &rdev->flags))
+					continue;
+				if (!rdev_set_badblocks(rdev, sect, s, 0))
+					abort = 1;
 			}
-	}
-	if (!test_bit(R1BIO_Uptodate, &r1_bio->state)) {
-		/* ouch - failed to read all of that.
-		 * Try some synchronous reads of other devices to get
-		 * good data, much like with normal read errors.  Only
-		 * read into the pages we already have so we don't
-		 * need to re-issue the read request.
-		 * We don't need to freeze the array, because being in an
-		 * active sync request, there is no normal IO, and
-		 * no overlapping syncs.
-		 */
-		sector_t sect = r1_bio->sector;
-		int sectors = r1_bio->sectors;
-		int idx = 0;
-
-		while(sectors) {
-			int s = sectors;
-			int d = r1_bio->read_disk;
-			int success = 0;
-			mdk_rdev_t *rdev;
-
-			if (s > (PAGE_SIZE>>9))
-				s = PAGE_SIZE >> 9;
-			do {
-				if (r1_bio->bios[d]->bi_end_io == end_sync_read) {
-					/* No rcu protection needed here devices
-					 * can only be removed when no resync is
-					 * active, and resync is currently active
-					 */
-					rdev = conf->mirrors[d].rdev;
-					if (sync_page_io(rdev->bdev,
-							 sect + rdev->data_offset,
-							 s<<9,
-							 bio->bi_io_vec[idx].bv_page,
-							 READ)) {
-						success = 1;
-						break;
-					}
-				}
-				d++;
-				if (d == conf->raid_disks)
-					d = 0;
-			} while (!success && d != r1_bio->read_disk);
-
-			if (success) {
-				int start = d;
-				/* write it back and re-read */
-				set_bit(R1BIO_Uptodate, &r1_bio->state);
-				while (d != r1_bio->read_disk) {
-					if (d == 0)
-						d = conf->raid_disks;
-					d--;
-					if (r1_bio->bios[d]->bi_end_io != end_sync_read)
-						continue;
-					rdev = conf->mirrors[d].rdev;
-					atomic_add(s, &rdev->corrected_errors);
-					if (sync_page_io(rdev->bdev,
-							 sect + rdev->data_offset,
-							 s<<9,
-							 bio->bi_io_vec[idx].bv_page,
-							 WRITE) == 0)
-						md_error(mddev, rdev);
-				}
-				d = start;
-				while (d != r1_bio->read_disk) {
-					if (d == 0)
-						d = conf->raid_disks;
-					d--;
-					if (r1_bio->bios[d]->bi_end_io != end_sync_read)
-						continue;
-					rdev = conf->mirrors[d].rdev;
-					if (sync_page_io(rdev->bdev,
-							 sect + rdev->data_offset,
-							 s<<9,
-							 bio->bi_io_vec[idx].bv_page,
-							 READ) == 0)
-						md_error(mddev, rdev);
-				}
-			} else {
-				char b[BDEVNAME_SIZE];
-				/* Cannot read from anywhere, array is toast */
-				md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
-				printk(KERN_ALERT "raid1: %s: unrecoverable I/O read error"
-				       " for block %llu\n",
-				       bdevname(bio->bi_bdev,b),
-				       (unsigned long long)r1_bio->sector);
+			if (abort) {
+				conf->recovery_disabled =
+					mddev->recovery_disabled;
+				set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 				md_done_sync(mddev, r1_bio->sectors, 0);
 				put_buf(r1_bio);
-				return;
+				return 0;
 			}
+			/* Try next page */
 			sectors -= s;
 			sect += s;
-			idx ++;
+			idx++;
+			continue;
+		}
+
+		start = d;
+		/* write it back and re-read */
+		while (d != r1_bio->read_disk) {
+			if (d == 0)
+				d = conf->raid_disks * 2;
+			d--;
+			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
+				continue;
+			rdev = conf->mirrors[d].rdev;
+			if (r1_sync_page_io(rdev, sect, s,
+					    bio->bi_io_vec[idx].bv_page,
+					    WRITE) == 0) {
+				r1_bio->bios[d]->bi_end_io = NULL;
+				rdev_dec_pending(rdev, mddev);
+			}
+		}
+		d = start;
+		while (d != r1_bio->read_disk) {
+			if (d == 0)
+				d = conf->raid_disks * 2;
+			d--;
+			if (r1_bio->bios[d]->bi_end_io != end_sync_read)
+				continue;
+			rdev = conf->mirrors[d].rdev;
+			if (r1_sync_page_io(rdev, sect, s,
+					    bio->bi_io_vec[idx].bv_page,
+					    READ) != 0)
+				atomic_add(s, &rdev->corrected_errors);
+		}
+		sectors -= s;
+		sect += s;
+		idx ++;
+	}
+	set_bit(R1BIO_Uptodate, &r1_bio->state);
+	set_bit(BIO_UPTODATE, &bio->bi_flags);
+	return 1;
+}
+
+static int process_checks(struct r1bio *r1_bio)
+{
+	/* We have read all readable devices.  If we haven't
+	 * got the block, then there is no hope left.
+	 * If we have, then we want to do a comparison
+	 * and skip the write if everything is the same.
+	 * If any blocks failed to read, then we need to
+	 * attempt an over-write
+	 */
+	struct mddev *mddev = r1_bio->mddev;
+	struct r1conf *conf = mddev->private;
+	int primary;
+	int i;
+	int vcnt;
+
+	/* Fix variable parts of all bios */
+	vcnt = (r1_bio->sectors + PAGE_SIZE / 512 - 1) >> (PAGE_SHIFT - 9);
+	for (i = 0; i < conf->raid_disks * 2; i++) {
+		int j;
+		int size;
+		int uptodate;
+		struct bio *b = r1_bio->bios[i];
+		if (b->bi_end_io != end_sync_read)
+			continue;
+		/* fixup the bio for reuse, but preserve BIO_UPTODATE */
+		uptodate = test_bit(BIO_UPTODATE, &b->bi_flags);
+		bio_reset(b);
+		if (!uptodate)
+			clear_bit(BIO_UPTODATE, &b->bi_flags);
+		b->bi_vcnt = vcnt;
+		b->bi_iter.bi_size = r1_bio->sectors << 9;
+		b->bi_iter.bi_sector = r1_bio->sector +
+			conf->mirrors[i].rdev->data_offset;
+		b->bi_bdev = conf->mirrors[i].rdev->bdev;
+		b->bi_end_io = end_sync_read;
+		b->bi_private = r1_bio;
+
+		size = b->bi_iter.bi_size;
+		for (j = 0; j < vcnt ; j++) {
+			struct bio_vec *bi;
+			bi = &b->bi_io_vec[j];
+			bi->bv_offset = 0;
+			if (size > PAGE_SIZE)
+				bi->bv_len = PAGE_SIZE;
+			else
+				bi->bv_len = size;
+			size -= PAGE_SIZE;
 		}
 	}
+	for (primary = 0; primary < conf->raid_disks * 2; primary++)
+		if (r1_bio->bios[primary]->bi_end_io == end_sync_read &&
+		    test_bit(BIO_UPTODATE, &r1_bio->bios[primary]->bi_flags)) {
+			r1_bio->bios[primary]->bi_end_io = NULL;
+			rdev_dec_pending(conf->mirrors[primary].rdev, mddev);
+			break;
+		}
+	r1_bio->read_disk = primary;
+	for (i = 0; i < conf->raid_disks * 2; i++) {
+		int j;
+		struct bio *pbio = r1_bio->bios[primary];
+		struct bio *sbio = r1_bio->bios[i];
+		int uptodate = test_bit(BIO_UPTODATE, &sbio->bi_flags);
 
+		if (sbio->bi_end_io != end_sync_read)
+			continue;
+		/* Now we can 'fixup' the BIO_UPTODATE flag */
+		set_bit(BIO_UPTODATE, &sbio->bi_flags);
+
+		if (uptodate) {
+			for (j = vcnt; j-- ; ) {
+				struct page *p, *s;
+				p = pbio->bi_io_vec[j].bv_page;
+				s = sbio->bi_io_vec[j].bv_page;
+				if (memcmp(page_address(p),
+					   page_address(s),
+					   sbio->bi_io_vec[j].bv_len))
+					break;
+			}
+		} else
+			j = 0;
+		if (j >= 0)
+			atomic64_add(r1_bio->sectors, &mddev->resync_mismatches);
+		if (j < 0 || (test_bit(MD_RECOVERY_CHECK, &mddev->recovery)
+			      && uptodate)) {
+			/* No need to write to this device. */
+			sbio->bi_end_io = NULL;
+			rdev_dec_pending(conf->mirrors[i].rdev, mddev);
+			continue;
+		}
+
+		bio_copy_data(sbio, pbio);
+	}
+	return 0;
+}
+
+static void sync_request_write(struct mddev *mddev, struct r1bio *r1_bio)
+{
+	struct r1conf *conf = mddev->private;
+	int i;
+	int disks = conf->raid_disks * 2;
+	struct bio *bio, *wbio;
+
+	bio = r1_bio->bios[r1_bio->read_disk];
+
+	if (!test_bit(R1BIO_Uptodate, &r1_bio->state))
+		/* ouch - failed to read all of that. */
+		if (!fix_sync_read_error(r1_bio))
+			return;
+
+	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
+		if (process_checks(r1_bio) < 0)
+			return;
 	/*
 	 * schedule writes
 	 */
@@ -1439,15 +2069,21 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
 		wbio->bi_rw = WRITE;
 		wbio->bi_end_io = end_sync_write;
 		atomic_inc(&r1_bio->remaining);
-		md_sync_acct(conf->mirrors[i].rdev->bdev, wbio->bi_size >> 9);
+		md_sync_acct(conf->mirrors[i].rdev->bdev, bio_sectors(wbio));
 
 		generic_make_request(wbio);
 	}
 
 	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);
+		}
 	}
 }
 
@@ -1456,19 +2092,19 @@ static void sync_request_write(mddev_t *mddev, r1bio_t *r1_bio)
  *
  *	1.	Retries failed read operations on working mirrors.
  *	2.	Updates the raid superblock when problems encounter.
- *	3.	Performs writes following reads for array syncronising.
+ *	3.	Performs writes following reads for array synchronising.
  */
 
-static void fix_read_error(conf_t *conf, int read_disk,
+static void fix_read_error(struct r1conf *conf, int read_disk,
 			   sector_t sect, int sectors)
 {
-	mddev_t *mddev = conf->mddev;
+	struct mddev *mddev = conf->mddev;
 	while(sectors) {
 		int s = sectors;
 		int d = read_disk;
 		int success = 0;
 		int start;
-		mdk_rdev_t *rdev;
+		struct md_rdev *rdev;
 
 		if (s > (PAGE_SIZE>>9))
 			s = PAGE_SIZE >> 9;
@@ -1479,62 +2115,59 @@ static void fix_read_error(conf_t *conf, int read_disk,
 			 * which is the thread that might remove
 			 * a device.  If raid1d ever becomes multi-threaded....
 			 */
+			sector_t first_bad;
+			int bad_sectors;
+
 			rdev = conf->mirrors[d].rdev;
 			if (rdev &&
-			    test_bit(In_sync, &rdev->flags) &&
-			    sync_page_io(rdev->bdev,
-					 sect + rdev->data_offset,
-					 s<<9,
-					 conf->tmppage, READ))
+			    (test_bit(In_sync, &rdev->flags) ||
+			     (!test_bit(Faulty, &rdev->flags) &&
+			      rdev->recovery_offset >= sect + s)) &&
+			    is_badblock(rdev, sect, s,
+					&first_bad, &bad_sectors) == 0 &&
+			    sync_page_io(rdev, sect, s<<9,
+					 conf->tmppage, READ, false))
 				success = 1;
 			else {
 				d++;
-				if (d == conf->raid_disks)
+				if (d == conf->raid_disks * 2)
 					d = 0;
 			}
 		} while (!success && d != read_disk);
 
 		if (!success) {
-			/* Cannot read from anywhere -- bye bye array */
-			md_error(mddev, conf->mirrors[read_disk].rdev);
+			/* Cannot read from anywhere - mark it bad */
+			struct md_rdev *rdev = conf->mirrors[read_disk].rdev;
+			if (!rdev_set_badblocks(rdev, sect, s, 0))
+				md_error(mddev, rdev);
 			break;
 		}
 		/* write it back and re-read */
 		start = d;
 		while (d != read_disk) {
 			if (d==0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			rdev = conf->mirrors[d].rdev;
 			if (rdev &&
-			    test_bit(In_sync, &rdev->flags)) {
-				if (sync_page_io(rdev->bdev,
-						 sect + rdev->data_offset,
-						 s<<9, conf->tmppage, WRITE)
-				    == 0)
-					/* Well, this device is dead */
-					md_error(mddev, rdev);
-			}
+			    test_bit(In_sync, &rdev->flags))
+				r1_sync_page_io(rdev, sect, s,
+						conf->tmppage, WRITE);
 		}
 		d = start;
 		while (d != read_disk) {
 			char b[BDEVNAME_SIZE];
 			if (d==0)
-				d = conf->raid_disks;
+				d = conf->raid_disks * 2;
 			d--;
 			rdev = conf->mirrors[d].rdev;
 			if (rdev &&
 			    test_bit(In_sync, &rdev->flags)) {
-				if (sync_page_io(rdev->bdev,
-						 sect + rdev->data_offset,
-						 s<<9, conf->tmppage, READ)
-				    == 0)
-					/* Well, this device is dead */
-					md_error(mddev, rdev);
-				else {
+				if (r1_sync_page_io(rdev, sect, s,
+						    conf->tmppage, READ)) {
 					atomic_add(s, &rdev->corrected_errors);
 					printk(KERN_INFO
-					       "raid1:%s: read error corrected "
+					       "md/raid1:%s: read error corrected "
 					       "(%d sectors at %llu on %s)\n",
 					       mdname(mddev), s,
 					       (unsigned long long)(sect +
@@ -1548,129 +2181,280 @@ static void fix_read_error(conf_t *conf, int read_disk,
 	}
 }
 
-static void raid1d(mddev_t *mddev)
+static int narrow_write_error(struct r1bio *r1_bio, int i)
+{
+	struct mddev *mddev = r1_bio->mddev;
+	struct r1conf *conf = mddev->private;
+	struct md_rdev *rdev = conf->mirrors[i].rdev;
+
+	/* bio has the data to be written to device 'i' where
+	 * we just recently had a write error.
+	 * We repeatedly clone the bio and trim down to one block,
+	 * then try the write.  Where the write fails we record
+	 * a bad block.
+	 * It is conceivable that the bio doesn't exactly align with
+	 * blocks.  We must handle this somehow.
+	 *
+	 * We currently own a reference on the rdev.
+	 */
+
+	int block_sectors;
+	sector_t sector;
+	int sectors;
+	int sect_to_write = r1_bio->sectors;
+	int ok = 1;
+
+	if (rdev->badblocks.shift < 0)
+		return 0;
+
+	block_sectors = 1 << rdev->badblocks.shift;
+	sector = r1_bio->sector;
+	sectors = ((sector + block_sectors)
+		   & ~(sector_t)(block_sectors - 1))
+		- sector;
+
+	while (sect_to_write) {
+		struct bio *wbio;
+		if (sectors > sect_to_write)
+			sectors = sect_to_write;
+		/* Write at 'sector' for 'sectors'*/
+
+		if (test_bit(R1BIO_BehindIO, &r1_bio->state)) {
+			unsigned vcnt = r1_bio->behind_page_count;
+			struct bio_vec *vec = r1_bio->behind_bvecs;
+
+			while (!vec->bv_page) {
+				vec++;
+				vcnt--;
+			}
+
+			wbio = bio_alloc_mddev(GFP_NOIO, vcnt, mddev);
+			memcpy(wbio->bi_io_vec, vec, vcnt * sizeof(struct bio_vec));
+
+			wbio->bi_vcnt = vcnt;
+		} else {
+			wbio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev);
+		}
+
+		wbio->bi_rw = WRITE;
+		wbio->bi_iter.bi_sector = r1_bio->sector;
+		wbio->bi_iter.bi_size = r1_bio->sectors << 9;
+
+		bio_trim(wbio, sector - r1_bio->sector, sectors);
+		wbio->bi_iter.bi_sector += rdev->data_offset;
+		wbio->bi_bdev = rdev->bdev;
+		if (submit_bio_wait(WRITE, wbio) == 0)
+			/* failure! */
+			ok = rdev_set_badblocks(rdev, sector,
+						sectors, 0)
+				&& ok;
+
+		bio_put(wbio);
+		sect_to_write -= sectors;
+		sector += sectors;
+		sectors = block_sectors;
+	}
+	return ok;
+}
+
+static void handle_sync_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
+{
+	int m;
+	int s = r1_bio->sectors;
+	for (m = 0; m < conf->raid_disks * 2 ; m++) {
+		struct md_rdev *rdev = conf->mirrors[m].rdev;
+		struct bio *bio = r1_bio->bios[m];
+		if (bio->bi_end_io == NULL)
+			continue;
+		if (test_bit(BIO_UPTODATE, &bio->bi_flags) &&
+		    test_bit(R1BIO_MadeGood, &r1_bio->state)) {
+			rdev_clear_badblocks(rdev, r1_bio->sector, s, 0);
+		}
+		if (!test_bit(BIO_UPTODATE, &bio->bi_flags) &&
+		    test_bit(R1BIO_WriteError, &r1_bio->state)) {
+			if (!rdev_set_badblocks(rdev, r1_bio->sector, s, 0))
+				md_error(conf->mddev, rdev);
+		}
+	}
+	put_buf(r1_bio);
+	md_done_sync(conf->mddev, s, 1);
+}
+
+static void handle_write_finished(struct r1conf *conf, struct r1bio *r1_bio)
 {
-	r1bio_t *r1_bio;
+	int m;
+	for (m = 0; m < conf->raid_disks * 2 ; m++)
+		if (r1_bio->bios[m] == IO_MADE_GOOD) {
+			struct md_rdev *rdev = conf->mirrors[m].rdev;
+			rdev_clear_badblocks(rdev,
+					     r1_bio->sector,
+					     r1_bio->sectors, 0);
+			rdev_dec_pending(rdev, conf->mddev);
+		} else if (r1_bio->bios[m] != NULL) {
+			/* This drive got a write error.  We need to
+			 * narrow down and record precise write
+			 * errors.
+			 */
+			if (!narrow_write_error(r1_bio, m)) {
+				md_error(conf->mddev,
+					 conf->mirrors[m].rdev);
+				/* an I/O failed, we can't clear the bitmap */
+				set_bit(R1BIO_Degraded, &r1_bio->state);
+			}
+			rdev_dec_pending(conf->mirrors[m].rdev,
+					 conf->mddev);
+		}
+	if (test_bit(R1BIO_WriteError, &r1_bio->state))
+		close_write(r1_bio);
+	raid_end_bio_io(r1_bio);
+}
+
+static void handle_read_error(struct r1conf *conf, struct r1bio *r1_bio)
+{
+	int disk;
+	int max_sectors;
+	struct mddev *mddev = conf->mddev;
 	struct bio *bio;
+	char b[BDEVNAME_SIZE];
+	struct md_rdev *rdev;
+
+	clear_bit(R1BIO_ReadError, &r1_bio->state);
+	/* we got a read error. Maybe the drive is bad.  Maybe just
+	 * the block and we can fix it.
+	 * We freeze all other IO, and try reading the block from
+	 * other devices.  When we find one, we re-write
+	 * and check it that fixes the read error.
+	 * This is all done synchronously while the array is
+	 * frozen
+	 */
+	if (mddev->ro == 0) {
+		freeze_array(conf, 1);
+		fix_read_error(conf, r1_bio->read_disk,
+			       r1_bio->sector, r1_bio->sectors);
+		unfreeze_array(conf);
+	} else
+		md_error(mddev, conf->mirrors[r1_bio->read_disk].rdev);
+	rdev_dec_pending(conf->mirrors[r1_bio->read_disk].rdev, conf->mddev);
+
+	bio = r1_bio->bios[r1_bio->read_disk];
+	bdevname(bio->bi_bdev, b);
+read_more:
+	disk = read_balance(conf, r1_bio, &max_sectors);
+	if (disk == -1) {
+		printk(KERN_ALERT "md/raid1:%s: %s: unrecoverable I/O"
+		       " read error for block %llu\n",
+		       mdname(mddev), b, (unsigned long long)r1_bio->sector);
+		raid_end_bio_io(r1_bio);
+	} else {
+		const unsigned long do_sync
+			= r1_bio->master_bio->bi_rw & REQ_SYNC;
+		if (bio) {
+			r1_bio->bios[r1_bio->read_disk] =
+				mddev->ro ? IO_BLOCKED : NULL;
+			bio_put(bio);
+		}
+		r1_bio->read_disk = disk;
+		bio = bio_clone_mddev(r1_bio->master_bio, GFP_NOIO, mddev);
+		bio_trim(bio, r1_bio->sector - bio->bi_iter.bi_sector,
+			 max_sectors);
+		r1_bio->bios[r1_bio->read_disk] = bio;
+		rdev = conf->mirrors[disk].rdev;
+		printk_ratelimited(KERN_ERR
+				   "md/raid1:%s: redirecting sector %llu"
+				   " to other mirror: %s\n",
+				   mdname(mddev),
+				   (unsigned long long)r1_bio->sector,
+				   bdevname(rdev->bdev, b));
+		bio->bi_iter.bi_sector = r1_bio->sector + rdev->data_offset;
+		bio->bi_bdev = rdev->bdev;
+		bio->bi_end_io = raid1_end_read_request;
+		bio->bi_rw = READ | do_sync;
+		bio->bi_private = r1_bio;
+		if (max_sectors < r1_bio->sectors) {
+			/* Drat - have to split this up more */
+			struct bio *mbio = r1_bio->master_bio;
+			int sectors_handled = (r1_bio->sector + max_sectors
+					       - mbio->bi_iter.bi_sector);
+			r1_bio->sectors = max_sectors;
+			spin_lock_irq(&conf->device_lock);
+			if (mbio->bi_phys_segments == 0)
+				mbio->bi_phys_segments = 2;
+			else
+				mbio->bi_phys_segments++;
+			spin_unlock_irq(&conf->device_lock);
+			generic_make_request(bio);
+			bio = NULL;
+
+			r1_bio = mempool_alloc(conf->r1bio_pool, GFP_NOIO);
+
+			r1_bio->master_bio = mbio;
+			r1_bio->sectors = bio_sectors(mbio) - sectors_handled;
+			r1_bio->state = 0;
+			set_bit(R1BIO_ReadError, &r1_bio->state);
+			r1_bio->mddev = mddev;
+			r1_bio->sector = mbio->bi_iter.bi_sector +
+				sectors_handled;
+
+			goto read_more;
+		} else
+			generic_make_request(bio);
+	}
+}
+
+static void raid1d(struct md_thread *thread)
+{
+	struct mddev *mddev = thread->mddev;
+	struct r1bio *r1_bio;
 	unsigned long flags;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 	struct list_head *head = &conf->retry_list;
-	int unplug=0;
-	mdk_rdev_t *rdev;
+	struct blk_plug plug;
 
 	md_check_recovery(mddev);
-	
+
+	blk_start_plug(&plug);
 	for (;;) {
-		char b[BDEVNAME_SIZE];
 
-		unplug += flush_pending_writes(conf);
+		flush_pending_writes(conf);
 
 		spin_lock_irqsave(&conf->device_lock, flags);
 		if (list_empty(head)) {
 			spin_unlock_irqrestore(&conf->device_lock, flags);
 			break;
 		}
-		r1_bio = list_entry(head->prev, r1bio_t, retry_list);
+		r1_bio = list_entry(head->prev, struct r1bio, retry_list);
 		list_del(head->prev);
 		conf->nr_queued--;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 
 		mddev = r1_bio->mddev;
-		conf = mddev_to_conf(mddev);
+		conf = mddev->private;
 		if (test_bit(R1BIO_IsSync, &r1_bio->state)) {
-			sync_request_write(mddev, r1_bio);
-			unplug = 1;
-		} else if (test_bit(R1BIO_BarrierRetry, &r1_bio->state)) {
-			/* some requests in the r1bio were BIO_RW_BARRIER
-			 * requests which failed with -EOPNOTSUPP.  Hohumm..
-			 * Better resubmit without the barrier.
-			 * We know which devices to resubmit for, because
-			 * all others have had their bios[] entry cleared.
-			 * We already have a nr_pending reference on these rdevs.
-			 */
-			int i;
-			const int do_sync = bio_sync(r1_bio->master_bio);
-			clear_bit(R1BIO_BarrierRetry, &r1_bio->state);
-			clear_bit(R1BIO_Barrier, &r1_bio->state);
-			for (i=0; i < conf->raid_disks; i++)
-				if (r1_bio->bios[i])
-					atomic_inc(&r1_bio->remaining);
-			for (i=0; i < conf->raid_disks; i++)
-				if (r1_bio->bios[i]) {
-					struct bio_vec *bvec;
-					int j;
-
-					bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
-					/* copy pages from the failed bio, as
-					 * this might be a write-behind device */
-					__bio_for_each_segment(bvec, bio, j, 0)
-						bvec->bv_page = bio_iovec_idx(r1_bio->bios[i], j)->bv_page;
-					bio_put(r1_bio->bios[i]);
-					bio->bi_sector = r1_bio->sector +
-						conf->mirrors[i].rdev->data_offset;
-					bio->bi_bdev = conf->mirrors[i].rdev->bdev;
-					bio->bi_end_io = raid1_end_write_request;
-					bio->bi_rw = WRITE | do_sync;
-					bio->bi_private = r1_bio;
-					r1_bio->bios[i] = bio;
-					generic_make_request(bio);
-				}
-		} else {
-			int disk;
-
-			/* we got a read error. Maybe the drive is bad.  Maybe just
-			 * the block and we can fix it.
-			 * We freeze all other IO, and try reading the block from
-			 * other devices.  When we find one, we re-write
-			 * and check it that fixes the read error.
-			 * This is all done synchronously while the array is
-			 * frozen
+			if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
+			    test_bit(R1BIO_WriteError, &r1_bio->state))
+				handle_sync_write_finished(conf, r1_bio);
+			else
+				sync_request_write(mddev, r1_bio);
+		} else if (test_bit(R1BIO_MadeGood, &r1_bio->state) ||
+			   test_bit(R1BIO_WriteError, &r1_bio->state))
+			handle_write_finished(conf, r1_bio);
+		else if (test_bit(R1BIO_ReadError, &r1_bio->state))
+			handle_read_error(conf, r1_bio);
+		else
+			/* just a partial read to be scheduled from separate
+			 * context
 			 */
-			if (mddev->ro == 0) {
-				freeze_array(conf);
-				fix_read_error(conf, r1_bio->read_disk,
-					       r1_bio->sector,
-					       r1_bio->sectors);
-				unfreeze_array(conf);
-			}
+			generic_make_request(r1_bio->bios[r1_bio->read_disk]);
 
-			bio = r1_bio->bios[r1_bio->read_disk];
-			if ((disk=read_balance(conf, r1_bio)) == -1) {
-				printk(KERN_ALERT "raid1: %s: unrecoverable I/O"
-				       " read error for block %llu\n",
-				       bdevname(bio->bi_bdev,b),
-				       (unsigned long long)r1_bio->sector);
-				raid_end_bio_io(r1_bio);
-			} else {
-				const int do_sync = bio_sync(r1_bio->master_bio);
-				r1_bio->bios[r1_bio->read_disk] =
-					mddev->ro ? IO_BLOCKED : NULL;
-				r1_bio->read_disk = disk;
-				bio_put(bio);
-				bio = bio_clone(r1_bio->master_bio, GFP_NOIO);
-				r1_bio->bios[r1_bio->read_disk] = bio;
-				rdev = conf->mirrors[disk].rdev;
-				if (printk_ratelimit())
-					printk(KERN_ERR "raid1: %s: redirecting sector %llu to"
-					       " another mirror\n",
-					       bdevname(rdev->bdev,b),
-					       (unsigned long long)r1_bio->sector);
-				bio->bi_sector = r1_bio->sector + rdev->data_offset;
-				bio->bi_bdev = rdev->bdev;
-				bio->bi_end_io = raid1_end_read_request;
-				bio->bi_rw = READ | do_sync;
-				bio->bi_private = r1_bio;
-				unplug = 1;
-				generic_make_request(bio);
-			}
-		}
+		cond_resched();
+		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
+			md_check_recovery(mddev);
 	}
-	if (unplug)
-		unplug_slaves(mddev);
+	blk_finish_plug(&plug);
 }
 
 
-static int init_resync(conf_t *conf)
+static int init_resync(struct r1conf *conf)
 {
 	int buffs;
 
@@ -1694,29 +2478,26 @@ static int init_resync(conf_t *conf)
  * that can be installed to exclude normal IO requests.
  */
 
-static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
+static sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
 {
-	conf_t *conf = mddev_to_conf(mddev);
-	r1bio_t *r1_bio;
+	struct r1conf *conf = mddev->private;
+	struct r1bio *r1_bio;
 	struct bio *bio;
 	sector_t max_sector, nr_sectors;
 	int disk = -1;
 	int i;
 	int wonly = -1;
 	int write_targets = 0, read_targets = 0;
-	int sync_blocks;
+	sector_t sync_blocks;
 	int still_degraded = 0;
+	int good_sectors = RESYNC_SECTORS;
+	int min_bad = 0; /* number of sectors that are bad in all devices */
 
 	if (!conf->r1buf_pool)
-	{
-/*
-		printk("sync start - bitmap %p\n", mddev->bitmap);
-*/
 		if (init_resync(conf))
 			return 0;
-	}
 
-	max_sector = mddev->size << 1;
+	max_sector = mddev->dev_sectors;
 	if (sector_nr >= max_sector) {
 		/* If we aborted, we need to abort the
 		 * sync on the 'current' bitmap chunk (there will
@@ -1759,11 +2540,11 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		msleep_interruptible(1000);
 
 	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
+	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
 	raise_barrier(conf);
 
 	conf->next_resync = sector_nr;
 
-	r1_bio = mempool_alloc(conf->r1buf_pool, GFP_NOIO);
 	rcu_read_lock();
 	/*
 	 * If we get a correctably read error during resync or recovery,
@@ -1779,54 +2560,108 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	r1_bio->state = 0;
 	set_bit(R1BIO_IsSync, &r1_bio->state);
 
-	for (i=0; i < conf->raid_disks; i++) {
-		mdk_rdev_t *rdev;
+	for (i = 0; i < conf->raid_disks * 2; i++) {
+		struct md_rdev *rdev;
 		bio = r1_bio->bios[i];
-
-		/* take from bio_init */
-		bio->bi_next = NULL;
-		bio->bi_flags |= 1 << BIO_UPTODATE;
-		bio->bi_rw = READ;
-		bio->bi_vcnt = 0;
-		bio->bi_idx = 0;
-		bio->bi_phys_segments = 0;
-		bio->bi_hw_segments = 0;
-		bio->bi_size = 0;
-		bio->bi_end_io = NULL;
-		bio->bi_private = NULL;
+		bio_reset(bio);
 
 		rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev == NULL ||
-			   test_bit(Faulty, &rdev->flags)) {
-			still_degraded = 1;
-			continue;
+		    test_bit(Faulty, &rdev->flags)) {
+			if (i < conf->raid_disks)
+				still_degraded = 1;
 		} else if (!test_bit(In_sync, &rdev->flags)) {
 			bio->bi_rw = WRITE;
 			bio->bi_end_io = end_sync_write;
 			write_targets ++;
 		} else {
 			/* may need to read from here */
-			bio->bi_rw = READ;
-			bio->bi_end_io = end_sync_read;
-			if (test_bit(WriteMostly, &rdev->flags)) {
-				if (wonly < 0)
-					wonly = i;
-			} else {
-				if (disk < 0)
-					disk = i;
+			sector_t first_bad = MaxSector;
+			int bad_sectors;
+
+			if (is_badblock(rdev, sector_nr, good_sectors,
+					&first_bad, &bad_sectors)) {
+				if (first_bad > sector_nr)
+					good_sectors = first_bad - sector_nr;
+				else {
+					bad_sectors -= (sector_nr - first_bad);
+					if (min_bad == 0 ||
+					    min_bad > bad_sectors)
+						min_bad = bad_sectors;
+				}
+			}
+			if (sector_nr < first_bad) {
+				if (test_bit(WriteMostly, &rdev->flags)) {
+					if (wonly < 0)
+						wonly = i;
+				} else {
+					if (disk < 0)
+						disk = i;
+				}
+				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++;
 			}
-			read_targets++;
 		}
-		atomic_inc(&rdev->nr_pending);
-		bio->bi_sector = sector_nr + rdev->data_offset;
-		bio->bi_bdev = rdev->bdev;
-		bio->bi_private = r1_bio;
+		if (bio->bi_end_io) {
+			atomic_inc(&rdev->nr_pending);
+			bio->bi_iter.bi_sector = sector_nr + rdev->data_offset;
+			bio->bi_bdev = rdev->bdev;
+			bio->bi_private = r1_bio;
+		}
 	}
 	rcu_read_unlock();
 	if (disk < 0)
 		disk = wonly;
 	r1_bio->read_disk = disk;
 
+	if (read_targets == 0 && min_bad > 0) {
+		/* These sectors are bad on all InSync devices, so we
+		 * need to mark them bad on all write targets
+		 */
+		int ok = 1;
+		for (i = 0 ; i < conf->raid_disks * 2 ; i++)
+			if (r1_bio->bios[i]->bi_end_io == end_sync_write) {
+				struct md_rdev *rdev = conf->mirrors[i].rdev;
+				ok = rdev_set_badblocks(rdev, sector_nr,
+							min_bad, 0
+					) && ok;
+			}
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		*skipped = 1;
+		put_buf(r1_bio);
+
+		if (!ok) {
+			/* Cannot record the badblocks, so need to
+			 * abort the resync.
+			 * If there are multiple read targets, could just
+			 * fail the really bad ones ???
+			 */
+			conf->recovery_disabled = mddev->recovery_disabled;
+			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+			return 0;
+		} else
+			return min_bad;
+
+	}
+	if (min_bad > 0 && min_bad < good_sectors) {
+		/* only resync enough to reach the next bad->good
+		 * transition */
+		good_sectors = min_bad;
+	}
+
 	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) && read_targets > 0)
 		/* extra read targets are also write targets */
 		write_targets += read_targets-1;
@@ -1835,7 +2670,10 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		/* 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;
@@ -1843,6 +2681,8 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 
 	if (max_sector > mddev->resync_max)
 		max_sector = mddev->resync_max; /* Don't do IO beyond here */
+	if (max_sector > sector_nr + good_sectors)
+		max_sector = sector_nr + good_sectors;
 	nr_sectors = 0;
 	sync_blocks = 0;
 	do {
@@ -1859,11 +2699,11 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 			    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery))
 				break;
 			BUG_ON(sync_blocks < (PAGE_SIZE>>9));
-			if (len > (sync_blocks<<9))
+			if ((len >> 9) > sync_blocks)
 				len = sync_blocks<<9;
 		}
 
-		for (i=0 ; i < conf->raid_disks; i++) {
+		for (i = 0 ; i < conf->raid_disks * 2; i++) {
 			bio = r1_bio->bios[i];
 			if (bio->bi_end_io) {
 				page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
@@ -1877,7 +2717,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 							continue;
 						/* remove last page from this bio */
 						bio->bi_vcnt--;
-						bio->bi_size -= len;
+						bio->bi_iter.bi_size -= len;
 						bio->bi_flags &= ~(1<< BIO_SEG_VALID);
 					}
 					goto bio_full;
@@ -1896,9 +2736,10 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	 */
 	if (test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
 		atomic_set(&r1_bio->remaining, read_targets);
-		for (i=0; i<conf->raid_disks; 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);
 			}
@@ -1913,77 +2754,70 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	return nr_sectors;
 }
 
-static int run(mddev_t *mddev)
+static sector_t raid1_size(struct mddev *mddev, sector_t sectors, int raid_disks)
 {
-	conf_t *conf;
-	int i, j, disk_idx;
-	mirror_info_t *disk;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
+	if (sectors)
+		return sectors;
 
-	if (mddev->level != 1) {
-		printk("raid1: %s: raid level not set to mirroring (%d)\n",
-		       mdname(mddev), mddev->level);
-		goto out;
-	}
-	if (mddev->reshape_position != MaxSector) {
-		printk("raid1: %s: reshape_position set but not supported\n",
-		       mdname(mddev));
-		goto out;
-	}
-	/*
-	 * copy the already verified devices into our private RAID1
-	 * bookkeeping area. [whatever we allocate in run(),
-	 * should be freed in stop()]
-	 */
-	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
-	mddev->private = conf;
+	return mddev->dev_sectors;
+}
+
+static struct r1conf *setup_conf(struct mddev *mddev)
+{
+	struct r1conf *conf;
+	int i;
+	struct raid1_info *disk;
+	struct md_rdev *rdev;
+	int err = -ENOMEM;
+
+	conf = kzalloc(sizeof(struct r1conf), GFP_KERNEL);
 	if (!conf)
-		goto out_no_mem;
+		goto abort;
 
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
+	conf->mirrors = kzalloc(sizeof(struct raid1_info)
+				* mddev->raid_disks * 2,
 				 GFP_KERNEL);
 	if (!conf->mirrors)
-		goto out_no_mem;
+		goto abort;
 
 	conf->tmppage = alloc_page(GFP_KERNEL);
 	if (!conf->tmppage)
-		goto out_no_mem;
+		goto abort;
 
-	conf->poolinfo = kmalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
+	conf->poolinfo = kzalloc(sizeof(*conf->poolinfo), GFP_KERNEL);
 	if (!conf->poolinfo)
-		goto out_no_mem;
-	conf->poolinfo->mddev = mddev;
-	conf->poolinfo->raid_disks = mddev->raid_disks;
+		goto abort;
+	conf->poolinfo->raid_disks = mddev->raid_disks * 2;
 	conf->r1bio_pool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
 					  r1bio_pool_free,
 					  conf->poolinfo);
 	if (!conf->r1bio_pool)
-		goto out_no_mem;
+		goto abort;
 
-	spin_lock_init(&conf->device_lock);
-	mddev->queue->queue_lock = &conf->device_lock;
+	conf->poolinfo->mddev = mddev;
 
-	rdev_for_each(rdev, tmp, mddev) {
-		disk_idx = rdev->raid_disk;
+	err = -EINVAL;
+	spin_lock_init(&conf->device_lock);
+	rdev_for_each(rdev, mddev) {
+		struct request_queue *q;
+		int disk_idx = rdev->raid_disk;
 		if (disk_idx >= mddev->raid_disks
 		    || disk_idx < 0)
 			continue;
-		disk = conf->mirrors + disk_idx;
+		if (test_bit(Replacement, &rdev->flags))
+			disk = conf->mirrors + mddev->raid_disks + disk_idx;
+		else
+			disk = conf->mirrors + disk_idx;
 
+		if (disk->rdev)
+			goto abort;
 		disk->rdev = rdev;
-
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
-		/* as we don't honour merge_bvec_fn, we must never risk
-		 * violating it, so limit ->max_sector to one PAGE, as
-		 * a one page request is never in violation.
-		 */
-		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-			blk_queue_max_sectors(mddev->queue, PAGE_SIZE>>9);
+		q = bdev_get_queue(rdev->bdev);
+		if (q->merge_bvec_fn)
+			mddev->merge_check_needed = 1;
 
 		disk->head_position = 0;
+		disk->seq_start = MaxSector;
 	}
 	conf->raid_disks = mddev->raid_disks;
 	conf->mddev = mddev;
@@ -1993,110 +2827,184 @@ static int run(mddev_t *mddev)
 	init_waitqueue_head(&conf->wait_barrier);
 
 	bio_list_init(&conf->pending_bio_list);
-	bio_list_init(&conf->flushing_bio_list);
+	conf->pending_count = 0;
+	conf->recovery_disabled = mddev->recovery_disabled - 1;
 
+	conf->start_next_window = MaxSector;
+	conf->current_window_requests = conf->next_window_requests = 0;
 
-	mddev->degraded = 0;
-	for (i = 0; i < conf->raid_disks; i++) {
+	err = -EIO;
+	for (i = 0; i < conf->raid_disks * 2; i++) {
 
 		disk = conf->mirrors + i;
 
+		if (i < conf->raid_disks &&
+		    disk[conf->raid_disks].rdev) {
+			/* This slot has a replacement. */
+			if (!disk->rdev) {
+				/* No original, just make the replacement
+				 * a recovering spare
+				 */
+				disk->rdev =
+					disk[conf->raid_disks].rdev;
+				disk[conf->raid_disks].rdev = NULL;
+			} else if (!test_bit(In_sync, &disk->rdev->flags))
+				/* Original is not in_sync - bad */
+				goto abort;
+		}
+
 		if (!disk->rdev ||
 		    !test_bit(In_sync, &disk->rdev->flags)) {
 			disk->head_position = 0;
-			mddev->degraded++;
-			if (disk->rdev)
+			if (disk->rdev &&
+			    (disk->rdev->saved_raid_disk < 0))
 				conf->fullsync = 1;
 		}
 	}
-	if (mddev->degraded == conf->raid_disks) {
-		printk(KERN_ERR "raid1: no operational mirrors for %s\n",
-			mdname(mddev));
-		goto out_free_conf;
+
+	err = -ENOMEM;
+	conf->thread = md_register_thread(raid1d, mddev, "raid1");
+	if (!conf->thread) {
+		printk(KERN_ERR
+		       "md/raid1:%s: couldn't allocate thread\n",
+		       mdname(mddev));
+		goto abort;
 	}
-	if (conf->raid_disks - mddev->degraded == 1)
-		mddev->recovery_cp = MaxSector;
 
+	return conf;
+
+ abort:
+	if (conf) {
+		if (conf->r1bio_pool)
+			mempool_destroy(conf->r1bio_pool);
+		kfree(conf->mirrors);
+		safe_put_page(conf->tmppage);
+		kfree(conf->poolinfo);
+		kfree(conf);
+	}
+	return ERR_PTR(err);
+}
+
+static int stop(struct mddev *mddev);
+static int run(struct mddev *mddev)
+{
+	struct r1conf *conf;
+	int i;
+	struct md_rdev *rdev;
+	int ret;
+	bool discard_supported = false;
+
+	if (mddev->level != 1) {
+		printk(KERN_ERR "md/raid1:%s: raid level not set to mirroring (%d)\n",
+		       mdname(mddev), mddev->level);
+		return -EIO;
+	}
+	if (mddev->reshape_position != MaxSector) {
+		printk(KERN_ERR "md/raid1:%s: reshape_position set but not supported\n",
+		       mdname(mddev));
+		return -EIO;
+	}
 	/*
-	 * find the first working one and use it as a starting point
-	 * to read balancing.
+	 * copy the already verified devices into our private RAID1
+	 * bookkeeping area. [whatever we allocate in run(),
+	 * should be freed in stop()]
 	 */
-	for (j = 0; j < conf->raid_disks &&
-		     (!conf->mirrors[j].rdev ||
-		      !test_bit(In_sync, &conf->mirrors[j].rdev->flags)) ; j++)
-		/* nothing */;
-	conf->last_used = j;
+	if (mddev->private == NULL)
+		conf = setup_conf(mddev);
+	else
+		conf = mddev->private;
 
+	if (IS_ERR(conf))
+		return PTR_ERR(conf);
 
-	mddev->thread = md_register_thread(raid1d, mddev, "%s_raid1");
-	if (!mddev->thread) {
-		printk(KERN_ERR
-		       "raid1: couldn't allocate thread for %s\n",
-		       mdname(mddev));
-		goto out_free_conf;
+	if (mddev->queue)
+		blk_queue_max_write_same_sectors(mddev->queue, 0);
+
+	rdev_for_each(rdev, mddev) {
+		if (!mddev->gendisk)
+			continue;
+		disk_stack_limits(mddev->gendisk, rdev->bdev,
+				  rdev->data_offset << 9);
+		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
+			discard_supported = true;
 	}
 
+	mddev->degraded = 0;
+	for (i=0; i < conf->raid_disks; i++)
+		if (conf->mirrors[i].rdev == NULL ||
+		    !test_bit(In_sync, &conf->mirrors[i].rdev->flags) ||
+		    test_bit(Faulty, &conf->mirrors[i].rdev->flags))
+			mddev->degraded++;
+
+	if (conf->raid_disks - mddev->degraded == 1)
+		mddev->recovery_cp = MaxSector;
+
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "md/raid1:%s: not clean"
+		       " -- starting background reconstruction\n",
+		       mdname(mddev));
 	printk(KERN_INFO 
-		"raid1: raid set %s active with %d out of %d mirrors\n",
+		"md/raid1:%s: active with %d out of %d mirrors\n",
 		mdname(mddev), mddev->raid_disks - mddev->degraded, 
 		mddev->raid_disks);
+
 	/*
 	 * Ok, everything is just fine now
 	 */
-	mddev->array_sectors = mddev->size * 2;
-
-	mddev->queue->unplug_fn = raid1_unplug;
-	mddev->queue->backing_dev_info.congested_fn = raid1_congested;
-	mddev->queue->backing_dev_info.congested_data = mddev;
+	mddev->thread = conf->thread;
+	conf->thread = NULL;
+	mddev->private = conf;
 
-	return 0;
+	md_set_array_sectors(mddev, raid1_size(mddev, 0, 0));
 
-out_no_mem:
-	printk(KERN_ERR "raid1: couldn't allocate memory for %s\n",
-	       mdname(mddev));
+	if (mddev->queue) {
+		mddev->queue->backing_dev_info.congested_fn = raid1_congested;
+		mddev->queue->backing_dev_info.congested_data = mddev;
+		blk_queue_merge_bvec(mddev->queue, raid1_mergeable_bvec);
 
-out_free_conf:
-	if (conf) {
-		if (conf->r1bio_pool)
-			mempool_destroy(conf->r1bio_pool);
-		kfree(conf->mirrors);
-		safe_put_page(conf->tmppage);
-		kfree(conf->poolinfo);
-		kfree(conf);
-		mddev->private = NULL;
+		if (discard_supported)
+			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
+						mddev->queue);
+		else
+			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
+						  mddev->queue);
 	}
-out:
-	return -EIO;
+
+	ret =  md_integrity_register(mddev);
+	if (ret)
+		stop(mddev);
+	return ret;
 }
 
-static int stop(mddev_t *mddev)
+static int stop(struct mddev *mddev)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 	struct bitmap *bitmap = mddev->bitmap;
-	int behind_wait = 0;
 
 	/* wait for behind writes to complete */
-	while (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
-		behind_wait++;
-		printk(KERN_INFO "raid1: behind writes in progress on device %s, waiting to stop (%d)\n", mdname(mddev), behind_wait);
-		set_current_state(TASK_UNINTERRUPTIBLE);
-		schedule_timeout(HZ); /* wait a second */
+	if (bitmap && atomic_read(&bitmap->behind_writes) > 0) {
+		printk(KERN_INFO "md/raid1:%s: behind writes in progress - waiting to stop.\n",
+		       mdname(mddev));
 		/* need to kick something here to make sure I/O goes? */
+		wait_event(bitmap->behind_wait,
+			   atomic_read(&bitmap->behind_writes) == 0);
 	}
 
-	md_unregister_thread(mddev->thread);
-	mddev->thread = NULL;
-	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
+	freeze_array(conf, 0);
+	unfreeze_array(conf);
+
+	md_unregister_thread(&mddev->thread);
 	if (conf->r1bio_pool)
 		mempool_destroy(conf->r1bio_pool);
 	kfree(conf->mirrors);
+	safe_put_page(conf->tmppage);
 	kfree(conf->poolinfo);
 	kfree(conf);
 	mddev->private = NULL;
 	return 0;
 }
 
-static int raid1_resize(mddev_t *mddev, sector_t sectors)
+static int raid1_resize(struct mddev *mddev, sector_t sectors)
 {
 	/* no resync is happening, and there is enough space
 	 * on all devices, so we can resize.
@@ -2105,20 +3013,29 @@ static int raid1_resize(mddev_t *mddev, sector_t sectors)
 	 * any io in the removed space completes, but it hardly seems
 	 * worth it.
 	 */
-	mddev->array_sectors = sectors;
+	sector_t newsize = raid1_size(mddev, sectors, 0);
+	if (mddev->external_size &&
+	    mddev->array_sectors > newsize)
+		return -EINVAL;
+	if (mddev->bitmap) {
+		int ret = bitmap_resize(mddev->bitmap, newsize, 0, 0);
+		if (ret)
+			return ret;
+	}
+	md_set_array_sectors(mddev, newsize);
 	set_capacity(mddev->gendisk, mddev->array_sectors);
-	mddev->changed = 1;
-	if (mddev->array_sectors / 2 > mddev->size &&
-	    mddev->recovery_cp == MaxSector) {
-		mddev->recovery_cp = mddev->size << 1;
+	revalidate_disk(mddev->gendisk);
+	if (sectors > mddev->dev_sectors &&
+	    mddev->recovery_cp > mddev->dev_sectors) {
+		mddev->recovery_cp = mddev->dev_sectors;
 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	}
-	mddev->size = mddev->array_sectors / 2;
+	mddev->dev_sectors = sectors;
 	mddev->resync_max_sectors = sectors;
 	return 0;
 }
 
-static int raid1_reshape(mddev_t *mddev)
+static int raid1_reshape(struct mddev *mddev)
 {
 	/* We need to:
 	 * 1/ resize the r1bio_pool
@@ -2133,17 +3050,17 @@ static int raid1_reshape(mddev_t *mddev)
 	 */
 	mempool_t *newpool, *oldpool;
 	struct pool_info *newpoolinfo;
-	mirror_info_t *newmirrors;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct raid1_info *newmirrors;
+	struct r1conf *conf = mddev->private;
 	int cnt, raid_disks;
 	unsigned long flags;
 	int d, d2, err;
 
 	/* Cannot change chunk_size, layout, or level */
-	if (mddev->chunk_size != mddev->new_chunk ||
+	if (mddev->chunk_sectors != mddev->new_chunk_sectors ||
 	    mddev->layout != mddev->new_layout ||
 	    mddev->level != mddev->new_level) {
-		mddev->new_chunk = mddev->chunk_size;
+		mddev->new_chunk_sectors = mddev->chunk_sectors;
 		mddev->new_layout = mddev->layout;
 		mddev->new_level = mddev->level;
 		return -EINVAL;
@@ -2168,7 +3085,7 @@ static int raid1_reshape(mddev_t *mddev)
 	if (!newpoolinfo)
 		return -ENOMEM;
 	newpoolinfo->mddev = mddev;
-	newpoolinfo->raid_disks = raid_disks;
+	newpoolinfo->raid_disks = raid_disks * 2;
 
 	newpool = mempool_create(NR_RAID1_BIOS, r1bio_pool_alloc,
 				 r1bio_pool_free, newpoolinfo);
@@ -2176,34 +3093,30 @@ static int raid1_reshape(mddev_t *mddev)
 		kfree(newpoolinfo);
 		return -ENOMEM;
 	}
-	newmirrors = kzalloc(sizeof(struct mirror_info) * raid_disks, GFP_KERNEL);
+	newmirrors = kzalloc(sizeof(struct raid1_info) * raid_disks * 2,
+			     GFP_KERNEL);
 	if (!newmirrors) {
 		kfree(newpoolinfo);
 		mempool_destroy(newpool);
 		return -ENOMEM;
 	}
 
-	raise_barrier(conf);
+	freeze_array(conf, 0);
 
 	/* ok, everything is stopped */
 	oldpool = conf->r1bio_pool;
 	conf->r1bio_pool = newpool;
 
 	for (d = d2 = 0; d < conf->raid_disks; d++) {
-		mdk_rdev_t *rdev = conf->mirrors[d].rdev;
+		struct md_rdev *rdev = conf->mirrors[d].rdev;
 		if (rdev && rdev->raid_disk != d2) {
-			char nm[20];
-			sprintf(nm, "rd%d", rdev->raid_disk);
-			sysfs_remove_link(&mddev->kobj, nm);
+			sysfs_unlink_rdev(mddev, rdev);
 			rdev->raid_disk = d2;
-			sprintf(nm, "rd%d", rdev->raid_disk);
-			sysfs_remove_link(&mddev->kobj, nm);
-			if (sysfs_create_link(&mddev->kobj,
-					      &rdev->kobj, nm))
+			sysfs_unlink_rdev(mddev, rdev);
+			if (sysfs_link_rdev(mddev, rdev))
 				printk(KERN_WARNING
-				       "md/raid1: cannot register "
-				       "%s for %s\n",
-				       nm, mdname(mddev));
+				       "md/raid1:%s: cannot register rd%d\n",
+				       mdname(mddev), rdev->raid_disk);
 		}
 		if (rdev)
 			newmirrors[d2++].rdev = rdev;
@@ -2219,8 +3132,7 @@ static int raid1_reshape(mddev_t *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);
+	unfreeze_array(conf);
 
 	set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	md_wakeup_thread(mddev->thread);
@@ -2229,22 +3141,43 @@ static int raid1_reshape(mddev_t *mddev)
 	return 0;
 }
 
-static void raid1_quiesce(mddev_t *mddev, int state)
+static void raid1_quiesce(struct mddev *mddev, int state)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r1conf *conf = mddev->private;
 
 	switch(state) {
+	case 2: /* wake for suspend */
+		wake_up(&conf->wait_barrier);
+		break;
 	case 1:
-		raise_barrier(conf);
+		freeze_array(conf, 0);
 		break;
 	case 0:
-		lower_barrier(conf);
+		unfreeze_array(conf);
 		break;
 	}
 }
 
+static void *raid1_takeover(struct mddev *mddev)
+{
+	/* raid1 can take over:
+	 *  raid5 with 2 devices, any layout or chunk size
+	 */
+	if (mddev->level == 5 && mddev->raid_disks == 2) {
+		struct r1conf *conf;
+		mddev->new_level = 1;
+		mddev->new_layout = 0;
+		mddev->new_chunk_sectors = 0;
+		conf = setup_conf(mddev);
+		if (!IS_ERR(conf))
+			/* Array must appear to be quiesced */
+			conf->array_frozen = 1;
+		return conf;
+	}
+	return ERR_PTR(-EINVAL);
+}
 
-static struct mdk_personality raid1_personality =
+static struct md_personality raid1_personality =
 {
 	.name		= "raid1",
 	.level		= 1,
@@ -2259,8 +3192,10 @@ static struct mdk_personality raid1_personality =
 	.spare_active	= raid1_spare_active,
 	.sync_request	= sync_request,
 	.resize		= raid1_resize,
+	.size		= raid1_size,
 	.check_reshape	= raid1_reshape,
 	.quiesce	= raid1_quiesce,
+	.takeover	= raid1_takeover,
 };
 
 static int __init raid_init(void)
@@ -2276,6 +3211,9 @@ static void raid_exit(void)
 module_init(raid_init);
 module_exit(raid_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("RAID1 (mirroring) personality for MD");
 MODULE_ALIAS("md-personality-3"); /* RAID1 */
 MODULE_ALIAS("md-raid1");
 MODULE_ALIAS("md-level-1");
+
+module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);
diff --git a/drivers/md/raid1.h b/drivers/md/raid1.h
new file mode 100644
index 00000000000..9bebca7bff2
--- /dev/null
+++ b/drivers/md/raid1.h
@@ -0,0 +1,178 @@
+#ifndef _RAID1_H
+#define _RAID1_H
+
+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;
+};
+
+/*
+ * memory pools need a pointer to the mddev, so they can force an unplug
+ * when memory is tight, and a count of the number of drives that the
+ * pool was allocated for, so they know how much to allocate and free.
+ * mddev->raid_disks cannot be used, as it can change while a pool is active
+ * These two datums are stored in a kmalloced struct.
+ * The 'raid_disks' here is twice the raid_disks in r1conf.
+ * This allows space for each 'real' device can have a replacement in the
+ * second half of the array.
+ */
+
+struct pool_info {
+	struct mddev *mddev;
+	int	raid_disks;
+};
+
+struct r1conf {
+	struct mddev		*mddev;
+	struct raid1_info	*mirrors;	/* twice 'raid_disks' to
+						 * allow for replacements.
+						 */
+	int			raid_disks;
+
+	/* During resync, read_balancing is only allowed on the part
+	 * of the array that has been resynced.  'next_resync' tells us
+	 * where that is.
+	 */
+	sector_t		next_resync;
+
+	/* When raid1 starts resync, we divide array into four partitions
+	 * |---------|--------------|---------------------|-------------|
+	 *        next_resync   start_next_window       end_window
+	 * start_next_window = next_resync + NEXT_NORMALIO_DISTANCE
+	 * end_window = start_next_window + NEXT_NORMALIO_DISTANCE
+	 * current_window_requests means the count of normalIO between
+	 *   start_next_window and end_window.
+	 * next_window_requests means the count of normalIO after end_window.
+	 * */
+	sector_t		start_next_window;
+	int			current_window_requests;
+	int			next_window_requests;
+
+	spinlock_t		device_lock;
+
+	/* list of 'struct r1bio' that need to be processed by raid1d,
+	 * whether to retry a read, writeout a resync or recovery
+	 * block, or anything else.
+	 */
+	struct list_head	retry_list;
+
+	/* queue pending writes to be submitted on unplug */
+	struct bio_list		pending_bio_list;
+	int			pending_count;
+
+	/* for use when syncing mirrors:
+	 * We don't allow both normal IO and resync/recovery IO at
+	 * the same time - resync/recovery can only happen when there
+	 * is no other IO.  So when either is active, the other has to wait.
+	 * See more details description in raid1.c near raise_barrier().
+	 */
+	wait_queue_head_t	wait_barrier;
+	spinlock_t		resync_lock;
+	int			nr_pending;
+	int			nr_waiting;
+	int			nr_queued;
+	int			barrier;
+	int			array_frozen;
+
+	/* Set to 1 if a full sync is needed, (fresh device added).
+	 * Cleared when a sync completes.
+	 */
+	int			fullsync;
+
+	/* When the same as mddev->recovery_disabled we don't allow
+	 * recovery to be attempted as we expect a read error.
+	 */
+	int			recovery_disabled;
+
+
+	/* poolinfo contains information about the content of the
+	 * mempools - it changes when the array grows or shrinks
+	 */
+	struct pool_info	*poolinfo;
+	mempool_t		*r1bio_pool;
+	mempool_t		*r1buf_pool;
+
+	/* temporary buffer to synchronous IO when attempting to repair
+	 * a read error.
+	 */
+	struct page		*tmppage;
+
+
+	/* When taking over an array from a different personality, we store
+	 * the new thread here until we fully activate the array.
+	 */
+	struct md_thread	*thread;
+};
+
+/*
+ * this is our 'private' RAID1 bio.
+ *
+ * it contains information about what kind of IO operations were started
+ * for this RAID1 operation, and about their status:
+ */
+
+struct r1bio {
+	atomic_t		remaining; /* 'have we finished' count,
+					    * used from IRQ handlers
+					    */
+	atomic_t		behind_remaining; /* number of write-behind ios remaining
+						 * in this BehindIO request
+						 */
+	sector_t		sector;
+	sector_t		start_next_window;
+	int			sectors;
+	unsigned long		state;
+	struct mddev		*mddev;
+	/*
+	 * original bio going to /dev/mdx
+	 */
+	struct bio		*master_bio;
+	/*
+	 * if the IO is in READ direction, then this is where we read
+	 */
+	int			read_disk;
+
+	struct list_head	retry_list;
+	/* Next two are only valid when R1BIO_BehindIO is set */
+	struct bio_vec		*behind_bvecs;
+	int			behind_page_count;
+	/*
+	 * if the IO is in WRITE direction, then multiple bios are used.
+	 * We choose the number when they are allocated.
+	 */
+	struct bio		*bios[0];
+	/* DO NOT PUT ANY NEW FIELDS HERE - bios array is contiguously alloced*/
+};
+
+/* bits for r1bio.state */
+#define	R1BIO_Uptodate	0
+#define	R1BIO_IsSync	1
+#define	R1BIO_Degraded	2
+#define	R1BIO_BehindIO	3
+/* Set ReadError on bios that experience a readerror so that
+ * raid1d knows what to do with them.
+ */
+#define R1BIO_ReadError 4
+/* For write-behind requests, we call bi_end_io when
+ * the last non-write-behind device completes, providing
+ * any write was successful.  Otherwise we call when
+ * any write-behind write succeeds, otherwise we call
+ * with failure when last write completes (and all failed).
+ * Record that bi_end_io was called with this flag...
+ */
+#define	R1BIO_Returned 6
+/* If a write for this request means we can clear some
+ * known-bad-block records, we set this flag
+ */
+#define	R1BIO_MadeGood 7
+#define	R1BIO_WriteError 8
+
+extern int md_raid1_congested(struct mddev *mddev, int bits);
+
+#endif
diff --git a/drivers/md/raid10.c b/drivers/md/raid10.c
index e34cd0e6247..cb882aae9e2 100644
--- a/drivers/md/raid10.c
+++ b/drivers/md/raid10.c
@@ -5,7 +5,7 @@
  *
  * RAID-10 support for md.
  *
- * Base on code in raid1.c.  See raid1.c for futher copyright information.
+ * Base on code in raid1.c.  See raid1.c for further copyright information.
  *
  *
  * This program is free software; you can redistribute it and/or modify
@@ -18,9 +18,17 @@
  * Software Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.
  */
 
-#include "dm-bio-list.h"
-#include <linux/raid/raid10.h>
-#include <linux/raid/bitmap.h>
+#include <linux/slab.h>
+#include <linux/delay.h>
+#include <linux/blkdev.h>
+#include <linux/module.h>
+#include <linux/seq_file.h>
+#include <linux/ratelimit.h>
+#include <linux/kthread.h>
+#include "md.h"
+#include "raid10.h"
+#include "raid0.h"
+#include "bitmap.h"
 
 /*
  * RAID10 provides a combination of RAID0 and RAID1 functionality.
@@ -30,21 +38,36 @@
  *    near_copies (stored in low byte of layout)
  *    far_copies (stored in second byte of layout)
  *    far_offset (stored in bit 16 of layout )
+ *    use_far_sets (stored in bit 17 of layout )
  *
- * The data to be stored is divided into chunks using chunksize.
- * Each device is divided into far_copies sections.
- * In each section, chunks are laid out in a style similar to raid0, but
- * near_copies copies of each chunk is stored (each on a different drive).
- * The starting device for each section is offset near_copies from the starting
- * device of the previous section.
- * Thus they are (near_copies*far_copies) of each chunk, and each is on a different
- * drive.
- * near_copies and far_copies must be at least one, and their product is at most
- * raid_disks.
+ * The data to be stored is divided into chunks using chunksize.  Each device
+ * is divided into far_copies sections.   In each section, chunks are laid out
+ * in a style similar to raid0, but near_copies copies of each chunk is stored
+ * (each on a different drive).  The starting device for each section is offset
+ * near_copies from the starting device of the previous section.  Thus there
+ * are (near_copies * far_copies) of each chunk, and each is on a different
+ * drive.  near_copies and far_copies must be at least one, and their product
+ * is at most raid_disks.
  *
  * If far_offset is true, then the far_copies are handled a bit differently.
- * The copies are still in different stripes, but instead of be very far apart
- * on disk, there are adjacent stripes.
+ * The copies are still in different stripes, but instead of being very far
+ * apart on disk, there are adjacent stripes.
+ *
+ * The far and offset algorithms are handled slightly differently if
+ * 'use_far_sets' is true.  In this case, the array's devices are grouped into
+ * sets that are (near_copies * far_copies) in size.  The far copied stripes
+ * are still shifted by 'near_copies' devices, but this shifting stays confined
+ * to the set rather than the entire array.  This is done to improve the number
+ * of device combinations that can fail without causing the array to fail.
+ * Example 'far' algorithm w/o 'use_far_sets' (each letter represents a chunk
+ * on a device):
+ *    A B C D    A B C D E
+ *      ...         ...
+ *    D A B C    E A B C D
+ * Example 'far' algorithm w/ 'use_far_sets' enabled (sets illustrated w/ []'s):
+ *    [A B] [C D]    [A B] [C D E]
+ *    |...| |...|    |...| | ... |
+ *    [B A] [D C]    [B A] [E C D]
  */
 
 /*
@@ -52,23 +75,43 @@
  */
 #define	NR_RAID10_BIOS 256
 
-static void unplug_slaves(mddev_t *mddev);
+/* 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)
 
-static void allow_barrier(conf_t *conf);
-static void lower_barrier(conf_t *conf);
+/* When there are this many requests queued to be written by
+ * the raid10 thread, we become 'congested' to provide back-pressure
+ * for writeback.
+ */
+static int max_queued_requests = 1024;
+
+static void allow_barrier(struct r10conf *conf);
+static void lower_barrier(struct r10conf *conf);
+static int _enough(struct r10conf *conf, int previous, int ignore);
+static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
+				int *skipped);
+static void reshape_request_write(struct mddev *mddev, struct r10bio *r10_bio);
+static void end_reshape_write(struct bio *bio, int error);
+static void end_reshape(struct r10conf *conf);
 
 static void * r10bio_pool_alloc(gfp_t gfp_flags, void *data)
 {
-	conf_t *conf = data;
-	r10bio_t *r10_bio;
-	int size = offsetof(struct r10bio_s, devs[conf->copies]);
-
-	/* allocate a r10bio with room for raid_disks entries in the bios array */
-	r10_bio = kzalloc(size, gfp_flags);
-	if (!r10_bio)
-		unplug_slaves(conf->mddev);
+	struct r10conf *conf = data;
+	int size = offsetof(struct r10bio, devs[conf->copies]);
 
-	return r10_bio;
+	/* allocate a r10bio with room for raid_disks entries in the
+	 * bios array */
+	return kzalloc(size, gfp_flags);
 }
 
 static void r10bio_pool_free(void *r10_bio, void *data)
@@ -93,20 +136,19 @@ static void r10bio_pool_free(void *r10_bio, void *data)
  */
 static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 {
-	conf_t *conf = data;
+	struct r10conf *conf = data;
 	struct page *page;
-	r10bio_t *r10_bio;
+	struct r10bio *r10_bio;
 	struct bio *bio;
 	int i, j;
 	int nalloc;
 
 	r10_bio = r10bio_pool_alloc(gfp_flags, conf);
-	if (!r10_bio) {
-		unplug_slaves(conf->mddev);
+	if (!r10_bio)
 		return NULL;
-	}
 
-	if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
+	if (test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &conf->mddev->recovery))
 		nalloc = conf->copies; /* resync */
 	else
 		nalloc = 2; /* recovery */
@@ -115,23 +157,40 @@ static void * r10buf_pool_alloc(gfp_t gfp_flags, void *data)
 	 * Allocate bios.
 	 */
 	for (j = nalloc ; j-- ; ) {
-		bio = bio_alloc(gfp_flags, RESYNC_PAGES);
+		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
 		if (!bio)
 			goto out_free_bio;
 		r10_bio->devs[j].bio = bio;
+		if (!conf->have_replacement)
+			continue;
+		bio = bio_kmalloc(gfp_flags, RESYNC_PAGES);
+		if (!bio)
+			goto out_free_bio;
+		r10_bio->devs[j].repl_bio = bio;
 	}
 	/*
 	 * Allocate RESYNC_PAGES data pages and attach them
 	 * where needed.
 	 */
 	for (j = 0 ; j < nalloc; j++) {
+		struct bio *rbio = r10_bio->devs[j].repl_bio;
 		bio = r10_bio->devs[j].bio;
 		for (i = 0; i < RESYNC_PAGES; i++) {
-			page = alloc_page(gfp_flags);
+			if (j > 0 && !test_bit(MD_RECOVERY_SYNC,
+					       &conf->mddev->recovery)) {
+				/* we can share bv_page's during recovery
+				 * and reshape */
+				struct bio *rbio = r10_bio->devs[0].bio;
+				page = rbio->bi_io_vec[i].bv_page;
+				get_page(page);
+			} else
+				page = alloc_page(gfp_flags);
 			if (unlikely(!page))
 				goto out_free_pages;
 
 			bio->bi_io_vec[i].bv_page = page;
+			if (rbio)
+				rbio->bi_io_vec[i].bv_page = page;
 		}
 	}
 
@@ -143,10 +202,14 @@ out_free_pages:
 	while (j--)
 		for (i = 0; i < RESYNC_PAGES ; i++)
 			safe_put_page(r10_bio->devs[j].bio->bi_io_vec[i].bv_page);
-	j = -1;
+	j = 0;
 out_free_bio:
-	while ( ++j < nalloc )
-		bio_put(r10_bio->devs[j].bio);
+	for ( ; j < nalloc; j++) {
+		if (r10_bio->devs[j].bio)
+			bio_put(r10_bio->devs[j].bio);
+		if (r10_bio->devs[j].repl_bio)
+			bio_put(r10_bio->devs[j].repl_bio);
+	}
 	r10bio_pool_free(r10_bio, conf);
 	return NULL;
 }
@@ -154,8 +217,8 @@ out_free_bio:
 static void r10buf_pool_free(void *__r10_bio, void *data)
 {
 	int i;
-	conf_t *conf = data;
-	r10bio_t *r10bio = __r10_bio;
+	struct r10conf *conf = data;
+	struct r10bio *r10bio = __r10_bio;
 	int j;
 
 	for (j=0; j < conf->copies; j++) {
@@ -167,50 +230,51 @@ static void r10buf_pool_free(void *__r10_bio, void *data)
 			}
 			bio_put(bio);
 		}
+		bio = r10bio->devs[j].repl_bio;
+		if (bio)
+			bio_put(bio);
 	}
 	r10bio_pool_free(r10bio, conf);
 }
 
-static void put_all_bios(conf_t *conf, r10bio_t *r10_bio)
+static void put_all_bios(struct r10conf *conf, struct r10bio *r10_bio)
 {
 	int i;
 
 	for (i = 0; i < conf->copies; i++) {
 		struct bio **bio = & r10_bio->devs[i].bio;
-		if (*bio && *bio != IO_BLOCKED)
+		if (!BIO_SPECIAL(*bio))
+			bio_put(*bio);
+		*bio = NULL;
+		bio = &r10_bio->devs[i].repl_bio;
+		if (r10_bio->read_slot < 0 && !BIO_SPECIAL(*bio))
 			bio_put(*bio);
 		*bio = NULL;
 	}
 }
 
-static void free_r10bio(r10bio_t *r10_bio)
+static void free_r10bio(struct r10bio *r10_bio)
 {
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
-
-	/*
-	 * Wake up any possible resync thread that waits for the device
-	 * to go idle.
-	 */
-	allow_barrier(conf);
+	struct r10conf *conf = r10_bio->mddev->private;
 
 	put_all_bios(conf, r10_bio);
 	mempool_free(r10_bio, conf->r10bio_pool);
 }
 
-static void put_buf(r10bio_t *r10_bio)
+static void put_buf(struct r10bio *r10_bio)
 {
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
+	struct r10conf *conf = r10_bio->mddev->private;
 
 	mempool_free(r10_bio, conf->r10buf_pool);
 
 	lower_barrier(conf);
 }
 
-static void reschedule_retry(r10bio_t *r10_bio)
+static void reschedule_retry(struct r10bio *r10_bio)
 {
 	unsigned long flags;
-	mddev_t *mddev = r10_bio->mddev;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = r10_bio->mddev;
+	struct r10conf *conf = mddev->private;
 
 	spin_lock_irqsave(&conf->device_lock, flags);
 	list_add(&r10_bio->retry_list, &conf->retry_list);
@@ -228,36 +292,84 @@ static void reschedule_retry(r10bio_t *r10_bio)
  * operation and are ready to return a success/failure code to the buffer
  * cache layer.
  */
-static void raid_end_bio_io(r10bio_t *r10_bio)
+static void raid_end_bio_io(struct r10bio *r10_bio)
 {
 	struct bio *bio = r10_bio->master_bio;
+	int done;
+	struct r10conf *conf = r10_bio->mddev->private;
 
-	bio_endio(bio,
-		test_bit(R10BIO_Uptodate, &r10_bio->state) ? 0 : -EIO);
+	if (bio->bi_phys_segments) {
+		unsigned long flags;
+		spin_lock_irqsave(&conf->device_lock, flags);
+		bio->bi_phys_segments--;
+		done = (bio->bi_phys_segments == 0);
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+	} else
+		done = 1;
+	if (!test_bit(R10BIO_Uptodate, &r10_bio->state))
+		clear_bit(BIO_UPTODATE, &bio->bi_flags);
+	if (done) {
+		bio_endio(bio, 0);
+		/*
+		 * Wake up any possible resync thread that waits for the device
+		 * to go idle.
+		 */
+		allow_barrier(conf);
+	}
 	free_r10bio(r10_bio);
 }
 
 /*
  * Update disk head position estimator based on IRQ completion info.
  */
-static inline void update_head_pos(int slot, r10bio_t *r10_bio)
+static inline void update_head_pos(int slot, struct r10bio *r10_bio)
 {
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
+	struct r10conf *conf = r10_bio->mddev->private;
 
 	conf->mirrors[r10_bio->devs[slot].devnum].head_position =
 		r10_bio->devs[slot].addr + (r10_bio->sectors);
 }
 
+/*
+ * Find the disk number which triggered given bio
+ */
+static int find_bio_disk(struct r10conf *conf, struct r10bio *r10_bio,
+			 struct bio *bio, int *slotp, int *replp)
+{
+	int slot;
+	int repl = 0;
+
+	for (slot = 0; slot < conf->copies; slot++) {
+		if (r10_bio->devs[slot].bio == bio)
+			break;
+		if (r10_bio->devs[slot].repl_bio == bio) {
+			repl = 1;
+			break;
+		}
+	}
+
+	BUG_ON(slot == conf->copies);
+	update_head_pos(slot, r10_bio);
+
+	if (slotp)
+		*slotp = slot;
+	if (replp)
+		*replp = repl;
+	return r10_bio->devs[slot].devnum;
+}
+
 static void raid10_end_read_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
+	struct r10bio *r10_bio = bio->bi_private;
 	int slot, dev;
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
+	struct md_rdev *rdev;
+	struct r10conf *conf = r10_bio->mddev->private;
 
 
 	slot = r10_bio->read_slot;
 	dev = r10_bio->devs[slot].devnum;
+	rdev = r10_bio->devs[slot].rdev;
 	/*
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
@@ -274,41 +386,96 @@ static void raid10_end_read_request(struct bio *bio, int error)
 		 * wait for the 'master' bio.
 		 */
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
+	} else {
+		/* If all other devices that store this block have
+		 * failed, we want to return the error upwards rather
+		 * than fail the last device.  Here we redefine
+		 * "uptodate" to mean "Don't want to retry"
+		 */
+		if (!_enough(conf, test_bit(R10BIO_Previous, &r10_bio->state),
+			     rdev->raid_disk))
+			uptodate = 1;
+	}
+	if (uptodate) {
 		raid_end_bio_io(r10_bio);
+		rdev_dec_pending(rdev, conf->mddev);
 	} else {
 		/*
-		 * oops, read error:
+		 * oops, read error - keep the refcount on the rdev
 		 */
 		char b[BDEVNAME_SIZE];
-		if (printk_ratelimit())
-			printk(KERN_ERR "raid10: %s: rescheduling sector %llu\n",
-			       bdevname(conf->mirrors[dev].rdev->bdev,b), (unsigned long long)r10_bio->sector);
+		printk_ratelimited(KERN_ERR
+				   "md/raid10:%s: %s: rescheduling sector %llu\n",
+				   mdname(conf->mddev),
+				   bdevname(rdev->bdev, b),
+				   (unsigned long long)r10_bio->sector);
+		set_bit(R10BIO_ReadError, &r10_bio->state);
 		reschedule_retry(r10_bio);
 	}
+}
+
+static void close_write(struct r10bio *r10_bio)
+{
+	/* clear the bitmap if all writes complete successfully */
+	bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
+			r10_bio->sectors,
+			!test_bit(R10BIO_Degraded, &r10_bio->state),
+			0);
+	md_write_end(r10_bio->mddev);
+}
 
-	rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
+static void one_write_done(struct r10bio *r10_bio)
+{
+	if (atomic_dec_and_test(&r10_bio->remaining)) {
+		if (test_bit(R10BIO_WriteError, &r10_bio->state))
+			reschedule_retry(r10_bio);
+		else {
+			close_write(r10_bio);
+			if (test_bit(R10BIO_MadeGood, &r10_bio->state))
+				reschedule_retry(r10_bio);
+			else
+				raid_end_bio_io(r10_bio);
+		}
+	}
 }
 
 static void raid10_end_write_request(struct bio *bio, int error)
 {
 	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
-	int slot, dev;
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
-
-	for (slot = 0; slot < conf->copies; slot++)
-		if (r10_bio->devs[slot].bio == bio)
-			break;
-	dev = r10_bio->devs[slot].devnum;
-
+	struct r10bio *r10_bio = bio->bi_private;
+	int dev;
+	int dec_rdev = 1;
+	struct r10conf *conf = r10_bio->mddev->private;
+	int slot, repl;
+	struct md_rdev *rdev = NULL;
+
+	dev = find_bio_disk(conf, r10_bio, bio, &slot, &repl);
+
+	if (repl)
+		rdev = conf->mirrors[dev].replacement;
+	if (!rdev) {
+		smp_rmb();
+		repl = 0;
+		rdev = conf->mirrors[dev].rdev;
+	}
 	/*
 	 * this branch is our 'one mirror IO has finished' event handler:
 	 */
 	if (!uptodate) {
-		md_error(r10_bio->mddev, conf->mirrors[dev].rdev);
-		/* an I/O failed, we can't clear the bitmap */
-		set_bit(R10BIO_Degraded, &r10_bio->state);
-	} else
+		if (repl)
+			/* Never record new bad blocks to replacement,
+			 * just fail it.
+			 */
+			md_error(rdev->mddev, rdev);
+		else {
+			set_bit(WriteErrorSeen,	&rdev->flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+			set_bit(R10BIO_WriteError, &r10_bio->state);
+			dec_rdev = 0;
+		}
+	} else {
 		/*
 		 * Set R10BIO_Uptodate in our master bio, so that
 		 * we will return a good error code for to the higher
@@ -318,39 +485,56 @@ static void raid10_end_write_request(struct bio *bio, int error)
 		 * user-side. So if something waits for IO, then it will
 		 * wait for the 'master' bio.
 		 */
-		set_bit(R10BIO_Uptodate, &r10_bio->state);
+		sector_t first_bad;
+		int bad_sectors;
 
-	update_head_pos(slot, r10_bio);
+		/*
+		 * Do not set R10BIO_Uptodate if the current device is
+		 * rebuilding or Faulty. This is because we cannot use
+		 * such device for properly reading the data back (we could
+		 * potentially use it, if the current write would have felt
+		 * before rdev->recovery_offset, but for simplicity we don't
+		 * check this here.
+		 */
+		if (test_bit(In_sync, &rdev->flags) &&
+		    !test_bit(Faulty, &rdev->flags))
+			set_bit(R10BIO_Uptodate, &r10_bio->state);
+
+		/* Maybe we can clear some bad blocks. */
+		if (is_badblock(rdev,
+				r10_bio->devs[slot].addr,
+				r10_bio->sectors,
+				&first_bad, &bad_sectors)) {
+			bio_put(bio);
+			if (repl)
+				r10_bio->devs[slot].repl_bio = IO_MADE_GOOD;
+			else
+				r10_bio->devs[slot].bio = IO_MADE_GOOD;
+			dec_rdev = 0;
+			set_bit(R10BIO_MadeGood, &r10_bio->state);
+		}
+	}
 
 	/*
 	 *
 	 * Let's see if all mirrored write operations have finished
 	 * already.
 	 */
-	if (atomic_dec_and_test(&r10_bio->remaining)) {
-		/* clear the bitmap if all writes complete successfully */
-		bitmap_endwrite(r10_bio->mddev->bitmap, r10_bio->sector,
-				r10_bio->sectors,
-				!test_bit(R10BIO_Degraded, &r10_bio->state),
-				0);
-		md_write_end(r10_bio->mddev);
-		raid_end_bio_io(r10_bio);
-	}
-
-	rdev_dec_pending(conf->mirrors[dev].rdev, conf->mddev);
+	one_write_done(r10_bio);
+	if (dec_rdev)
+		rdev_dec_pending(rdev, conf->mddev);
 }
 
-
 /*
  * RAID10 layout manager
- * Aswell as the chunksize and raid_disks count, there are two
+ * As well as the chunksize and raid_disks count, there are two
  * parameters: near_copies and far_copies.
  * near_copies * far_copies must be <= raid_disks.
  * Normally one of these will be 1.
  * If both are 1, we get raid0.
  * If near_copies == raid_disks, we get raid1.
  *
- * Chunks are layed out in raid0 style with near_copies copies of the
+ * Chunks are laid out in raid0 style with near_copies copies of the
  * first chunk, followed by near_copies copies of the next chunk and
  * so on.
  * If far_copies > 1, then after 1/far_copies of the array has been assigned
@@ -366,79 +550,127 @@ static void raid10_end_write_request(struct bio *bio, int error)
  * sector offset to a virtual address
  */
 
-static void raid10_find_phys(conf_t *conf, r10bio_t *r10bio)
+static void __raid10_find_phys(struct geom *geo, struct r10bio *r10bio)
 {
 	int n,f;
 	sector_t sector;
 	sector_t chunk;
 	sector_t stripe;
 	int dev;
-
 	int slot = 0;
+	int last_far_set_start, last_far_set_size;
+
+	last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1;
+	last_far_set_start *= geo->far_set_size;
+
+	last_far_set_size = geo->far_set_size;
+	last_far_set_size += (geo->raid_disks % geo->far_set_size);
 
 	/* now calculate first sector/dev */
-	chunk = r10bio->sector >> conf->chunk_shift;
-	sector = r10bio->sector & conf->chunk_mask;
+	chunk = r10bio->sector >> geo->chunk_shift;
+	sector = r10bio->sector & geo->chunk_mask;
 
-	chunk *= conf->near_copies;
+	chunk *= geo->near_copies;
 	stripe = chunk;
-	dev = sector_div(stripe, conf->raid_disks);
-	if (conf->far_offset)
-		stripe *= conf->far_copies;
+	dev = sector_div(stripe, geo->raid_disks);
+	if (geo->far_offset)
+		stripe *= geo->far_copies;
 
-	sector += stripe << conf->chunk_shift;
+	sector += stripe << geo->chunk_shift;
 
 	/* and calculate all the others */
-	for (n=0; n < conf->near_copies; n++) {
+	for (n = 0; n < geo->near_copies; n++) {
 		int d = dev;
+		int set;
 		sector_t s = sector;
-		r10bio->devs[slot].addr = sector;
 		r10bio->devs[slot].devnum = d;
+		r10bio->devs[slot].addr = s;
 		slot++;
 
-		for (f = 1; f < conf->far_copies; f++) {
-			d += conf->near_copies;
-			if (d >= conf->raid_disks)
-				d -= conf->raid_disks;
-			s += conf->stride;
+		for (f = 1; f < geo->far_copies; f++) {
+			set = d / geo->far_set_size;
+			d += geo->near_copies;
+
+			if ((geo->raid_disks % geo->far_set_size) &&
+			    (d > last_far_set_start)) {
+				d -= last_far_set_start;
+				d %= last_far_set_size;
+				d += last_far_set_start;
+			} else {
+				d %= geo->far_set_size;
+				d += geo->far_set_size * set;
+			}
+			s += geo->stride;
 			r10bio->devs[slot].devnum = d;
 			r10bio->devs[slot].addr = s;
 			slot++;
 		}
 		dev++;
-		if (dev >= conf->raid_disks) {
+		if (dev >= geo->raid_disks) {
 			dev = 0;
-			sector += (conf->chunk_mask + 1);
+			sector += (geo->chunk_mask + 1);
 		}
 	}
-	BUG_ON(slot != conf->copies);
 }
 
-static sector_t raid10_find_virt(conf_t *conf, sector_t sector, int dev)
+static void raid10_find_phys(struct r10conf *conf, struct r10bio *r10bio)
+{
+	struct geom *geo = &conf->geo;
+
+	if (conf->reshape_progress != MaxSector &&
+	    ((r10bio->sector >= conf->reshape_progress) !=
+	     conf->mddev->reshape_backwards)) {
+		set_bit(R10BIO_Previous, &r10bio->state);
+		geo = &conf->prev;
+	} else
+		clear_bit(R10BIO_Previous, &r10bio->state);
+
+	__raid10_find_phys(geo, r10bio);
+}
+
+static sector_t raid10_find_virt(struct r10conf *conf, sector_t sector, int dev)
 {
 	sector_t offset, chunk, vchunk;
+	/* Never use conf->prev as this is only called during resync
+	 * or recovery, so reshape isn't happening
+	 */
+	struct geom *geo = &conf->geo;
+	int far_set_start = (dev / geo->far_set_size) * geo->far_set_size;
+	int far_set_size = geo->far_set_size;
+	int last_far_set_start;
+
+	if (geo->raid_disks % geo->far_set_size) {
+		last_far_set_start = (geo->raid_disks / geo->far_set_size) - 1;
+		last_far_set_start *= geo->far_set_size;
+
+		if (dev >= last_far_set_start) {
+			far_set_size = geo->far_set_size;
+			far_set_size += (geo->raid_disks % geo->far_set_size);
+			far_set_start = last_far_set_start;
+		}
+	}
 
-	offset = sector & conf->chunk_mask;
-	if (conf->far_offset) {
+	offset = sector & geo->chunk_mask;
+	if (geo->far_offset) {
 		int fc;
-		chunk = sector >> conf->chunk_shift;
-		fc = sector_div(chunk, conf->far_copies);
-		dev -= fc * conf->near_copies;
-		if (dev < 0)
-			dev += conf->raid_disks;
+		chunk = sector >> geo->chunk_shift;
+		fc = sector_div(chunk, geo->far_copies);
+		dev -= fc * geo->near_copies;
+		if (dev < far_set_start)
+			dev += far_set_size;
 	} else {
-		while (sector >= conf->stride) {
-			sector -= conf->stride;
-			if (dev < conf->near_copies)
-				dev += conf->raid_disks - conf->near_copies;
+		while (sector >= geo->stride) {
+			sector -= geo->stride;
+			if (dev < (geo->near_copies + far_set_start))
+				dev += far_set_size - geo->near_copies;
 			else
-				dev -= conf->near_copies;
+				dev -= geo->near_copies;
 		}
-		chunk = sector >> conf->chunk_shift;
+		chunk = sector >> geo->chunk_shift;
 	}
-	vchunk = chunk * conf->raid_disks + dev;
-	sector_div(vchunk, conf->near_copies);
-	return (vchunk << conf->chunk_shift) + offset;
+	vchunk = chunk * geo->raid_disks + dev;
+	sector_div(vchunk, geo->near_copies);
+	return (vchunk << geo->chunk_shift) + offset;
 }
 
 /**
@@ -448,25 +680,85 @@ static sector_t raid10_find_virt(conf_t *conf, sector_t sector, int dev)
  *	@biovec: the request that could be merged to it.
  *
  *	Return amount of bytes we can accept at this offset
- *      If near_copies == raid_disk, there are no striping issues,
- *      but in that case, the function isn't called at all.
+ *	This requires checking for end-of-chunk if near_copies != raid_disks,
+ *	and for subordinate merge_bvec_fns if merge_check_needed.
  */
 static int raid10_mergeable_bvec(struct request_queue *q,
 				 struct bvec_merge_data *bvm,
 				 struct bio_vec *biovec)
 {
-	mddev_t *mddev = q->queuedata;
+	struct mddev *mddev = q->queuedata;
+	struct r10conf *conf = mddev->private;
 	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
 	int max;
-	unsigned int chunk_sectors = mddev->chunk_size >> 9;
+	unsigned int chunk_sectors;
 	unsigned int bio_sectors = bvm->bi_size >> 9;
-
-	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
-	if (max < 0) max = 0; /* bio_add cannot handle a negative return */
-	if (max <= biovec->bv_len && bio_sectors == 0)
-		return biovec->bv_len;
-	else
-		return max;
+	struct geom *geo = &conf->geo;
+
+	chunk_sectors = (conf->geo.chunk_mask & conf->prev.chunk_mask) + 1;
+	if (conf->reshape_progress != MaxSector &&
+	    ((sector >= conf->reshape_progress) !=
+	     conf->mddev->reshape_backwards))
+		geo = &conf->prev;
+
+	if (geo->near_copies < geo->raid_disks) {
+		max = (chunk_sectors - ((sector & (chunk_sectors - 1))
+					+ bio_sectors)) << 9;
+		if (max < 0)
+			/* bio_add cannot handle a negative return */
+			max = 0;
+		if (max <= biovec->bv_len && bio_sectors == 0)
+			return biovec->bv_len;
+	} else
+		max = biovec->bv_len;
+
+	if (mddev->merge_check_needed) {
+		struct {
+			struct r10bio r10_bio;
+			struct r10dev devs[conf->copies];
+		} on_stack;
+		struct r10bio *r10_bio = &on_stack.r10_bio;
+		int s;
+		if (conf->reshape_progress != MaxSector) {
+			/* Cannot give any guidance during reshape */
+			if (max <= biovec->bv_len && bio_sectors == 0)
+				return biovec->bv_len;
+			return 0;
+		}
+		r10_bio->sector = sector;
+		raid10_find_phys(conf, r10_bio);
+		rcu_read_lock();
+		for (s = 0; s < conf->copies; s++) {
+			int disk = r10_bio->devs[s].devnum;
+			struct md_rdev *rdev = rcu_dereference(
+				conf->mirrors[disk].rdev);
+			if (rdev && !test_bit(Faulty, &rdev->flags)) {
+				struct request_queue *q =
+					bdev_get_queue(rdev->bdev);
+				if (q->merge_bvec_fn) {
+					bvm->bi_sector = r10_bio->devs[s].addr
+						+ rdev->data_offset;
+					bvm->bi_bdev = rdev->bdev;
+					max = min(max, q->merge_bvec_fn(
+							  q, bvm, biovec));
+				}
+			}
+			rdev = rcu_dereference(conf->mirrors[disk].replacement);
+			if (rdev && !test_bit(Faulty, &rdev->flags)) {
+				struct request_queue *q =
+					bdev_get_queue(rdev->bdev);
+				if (q->merge_bvec_fn) {
+					bvm->bi_sector = r10_bio->devs[s].addr
+						+ rdev->data_offset;
+					bvm->bi_bdev = rdev->bdev;
+					max = min(max, q->merge_bvec_fn(
+							  q, bvm, biovec));
+				}
+			}
+		}
+		rcu_read_unlock();
+	}
+	return max;
 }
 
 /*
@@ -488,16 +780,29 @@ static int raid10_mergeable_bvec(struct request_queue *q,
  * FIXME: possibly should rethink readbalancing and do it differently
  * depending on near_copies / far_copies geometry.
  */
-static int read_balance(conf_t *conf, r10bio_t *r10_bio)
+static struct md_rdev *read_balance(struct r10conf *conf,
+				    struct r10bio *r10_bio,
+				    int *max_sectors)
 {
-	const unsigned long this_sector = r10_bio->sector;
-	int disk, slot, nslot;
-	const int sectors = r10_bio->sectors;
-	sector_t new_distance, current_distance;
-	mdk_rdev_t *rdev;
+	const sector_t this_sector = r10_bio->sector;
+	int disk, slot;
+	int sectors = r10_bio->sectors;
+	int best_good_sectors;
+	sector_t new_distance, best_dist;
+	struct md_rdev *best_rdev, *rdev = NULL;
+	int do_balance;
+	int best_slot;
+	struct geom *geo = &conf->geo;
 
 	raid10_find_phys(conf, r10_bio);
 	rcu_read_lock();
+retry:
+	sectors = r10_bio->sectors;
+	best_slot = -1;
+	best_rdev = NULL;
+	best_dist = MaxSector;
+	best_good_sectors = 0;
+	do_balance = 1;
 	/*
 	 * Check if we can balance. We can balance on the whole
 	 * device if no resync is going on (recovery is ok), or below
@@ -505,132 +810,122 @@ static int read_balance(conf_t *conf, r10bio_t *r10_bio)
 	 * above the resync window.
 	 */
 	if (conf->mddev->recovery_cp < MaxSector
-	    && (this_sector + sectors >= conf->next_resync)) {
-		/* make sure that disk is operational */
-		slot = 0;
-		disk = r10_bio->devs[slot].devnum;
-
-		while ((rdev = rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
-		       r10_bio->devs[slot].bio == IO_BLOCKED ||
-		       !test_bit(In_sync, &rdev->flags)) {
-			slot++;
-			if (slot == conf->copies) {
-				slot = 0;
-				disk = -1;
-				break;
-			}
-			disk = r10_bio->devs[slot].devnum;
-		}
-		goto rb_out;
-	}
+	    && (this_sector + sectors >= conf->next_resync))
+		do_balance = 0;
 
+	for (slot = 0; slot < conf->copies ; slot++) {
+		sector_t first_bad;
+		int bad_sectors;
+		sector_t dev_sector;
 
-	/* make sure the disk is operational */
-	slot = 0;
-	disk = r10_bio->devs[slot].devnum;
-	while ((rdev=rcu_dereference(conf->mirrors[disk].rdev)) == NULL ||
-	       r10_bio->devs[slot].bio == IO_BLOCKED ||
-	       !test_bit(In_sync, &rdev->flags)) {
-		slot ++;
-		if (slot == conf->copies) {
-			disk = -1;
-			goto rb_out;
-		}
+		if (r10_bio->devs[slot].bio == IO_BLOCKED)
+			continue;
 		disk = r10_bio->devs[slot].devnum;
-	}
-
-
-	current_distance = abs(r10_bio->devs[slot].addr -
-			       conf->mirrors[disk].head_position);
-
-	/* Find the disk whose head is closest,
-	 * or - for far > 1 - find the closest to partition beginning */
-
-	for (nslot = slot; nslot < conf->copies; nslot++) {
-		int ndisk = r10_bio->devs[nslot].devnum;
-
+		rdev = rcu_dereference(conf->mirrors[disk].replacement);
+		if (rdev == NULL || test_bit(Faulty, &rdev->flags) ||
+		    test_bit(Unmerged, &rdev->flags) ||
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
+			rdev = rcu_dereference(conf->mirrors[disk].rdev);
+		if (rdev == NULL ||
+		    test_bit(Faulty, &rdev->flags) ||
+		    test_bit(Unmerged, &rdev->flags))
+			continue;
+		if (!test_bit(In_sync, &rdev->flags) &&
+		    r10_bio->devs[slot].addr + sectors > rdev->recovery_offset)
+			continue;
 
-		if ((rdev=rcu_dereference(conf->mirrors[ndisk].rdev)) == NULL ||
-		    r10_bio->devs[nslot].bio == IO_BLOCKED ||
-		    !test_bit(In_sync, &rdev->flags))
+		dev_sector = r10_bio->devs[slot].addr;
+		if (is_badblock(rdev, dev_sector, sectors,
+				&first_bad, &bad_sectors)) {
+			if (best_dist < MaxSector)
+				/* Already have a better slot */
+				continue;
+			if (first_bad <= dev_sector) {
+				/* Cannot read here.  If this is the
+				 * 'primary' device, then we must not read
+				 * beyond 'bad_sectors' from another device.
+				 */
+				bad_sectors -= (dev_sector - first_bad);
+				if (!do_balance && sectors > bad_sectors)
+					sectors = bad_sectors;
+				if (best_good_sectors > sectors)
+					best_good_sectors = sectors;
+			} else {
+				sector_t good_sectors =
+					first_bad - dev_sector;
+				if (good_sectors > best_good_sectors) {
+					best_good_sectors = good_sectors;
+					best_slot = slot;
+					best_rdev = rdev;
+				}
+				if (!do_balance)
+					/* Must read from here */
+					break;
+			}
 			continue;
+		} else
+			best_good_sectors = sectors;
+
+		if (!do_balance)
+			break;
 
 		/* This optimisation is debatable, and completely destroys
 		 * sequential read speed for 'far copies' arrays.  So only
 		 * keep it for 'near' arrays, and review those later.
 		 */
-		if (conf->near_copies > 1 && !atomic_read(&rdev->nr_pending)) {
-			disk = ndisk;
-			slot = nslot;
+		if (geo->near_copies > 1 && !atomic_read(&rdev->nr_pending))
 			break;
-		}
 
 		/* for far > 1 always use the lowest address */
-		if (conf->far_copies > 1)
-			new_distance = r10_bio->devs[nslot].addr;
+		if (geo->far_copies > 1)
+			new_distance = r10_bio->devs[slot].addr;
 		else
-			new_distance = abs(r10_bio->devs[nslot].addr -
-					   conf->mirrors[ndisk].head_position);
-		if (new_distance < current_distance) {
-			current_distance = new_distance;
-			disk = ndisk;
-			slot = nslot;
+			new_distance = abs(r10_bio->devs[slot].addr -
+					   conf->mirrors[disk].head_position);
+		if (new_distance < best_dist) {
+			best_dist = new_distance;
+			best_slot = slot;
+			best_rdev = rdev;
 		}
 	}
+	if (slot >= conf->copies) {
+		slot = best_slot;
+		rdev = best_rdev;
+	}
 
-rb_out:
-	r10_bio->read_slot = slot;
-/*	conf->next_seq_sect = this_sector + sectors;*/
-
-	if (disk >= 0 && (rdev=rcu_dereference(conf->mirrors[disk].rdev))!= NULL)
-		atomic_inc(&conf->mirrors[disk].rdev->nr_pending);
-	else
-		disk = -1;
-	rcu_read_unlock();
-
-	return disk;
-}
-
-static void unplug_slaves(mddev_t *mddev)
-{
-	conf_t *conf = mddev_to_conf(mddev);
-	int i;
-
-	rcu_read_lock();
-	for (i=0; i<mddev->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
-		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
-			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
-
-			atomic_inc(&rdev->nr_pending);
-			rcu_read_unlock();
-
-			blk_unplug(r_queue);
-
-			rdev_dec_pending(rdev, mddev);
-			rcu_read_lock();
+	if (slot >= 0) {
+		atomic_inc(&rdev->nr_pending);
+		if (test_bit(Faulty, &rdev->flags)) {
+			/* Cannot risk returning a device that failed
+			 * before we inc'ed nr_pending
+			 */
+			rdev_dec_pending(rdev, conf->mddev);
+			goto retry;
 		}
-	}
+		r10_bio->read_slot = slot;
+	} else
+		rdev = NULL;
 	rcu_read_unlock();
-}
-
-static void raid10_unplug(struct request_queue *q)
-{
-	mddev_t *mddev = q->queuedata;
+	*max_sectors = best_good_sectors;
 
-	unplug_slaves(q->queuedata);
-	md_wakeup_thread(mddev->thread);
+	return rdev;
 }
 
-static int raid10_congested(void *data, int bits)
+int md_raid10_congested(struct mddev *mddev, int bits)
 {
-	mddev_t *mddev = data;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
 	int i, ret = 0;
 
+	if ((bits & (1 << BDI_async_congested)) &&
+	    conf->pending_count >= max_queued_requests)
+		return 1;
+
 	rcu_read_lock();
-	for (i = 0; i < mddev->raid_disks && ret == 0; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[i].rdev);
+	for (i = 0;
+	     (i < conf->geo.raid_disks || i < conf->prev.raid_disks)
+		     && ret == 0;
+	     i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
 		if (rdev && !test_bit(Faulty, &rdev->flags)) {
 			struct request_queue *q = bdev_get_queue(rdev->bdev);
 
@@ -640,37 +935,48 @@ 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 int flush_pending_writes(conf_t *conf)
+static void flush_pending_writes(struct r10conf *conf)
 {
 	/* Any writes that have been queued but are awaiting
 	 * bitmap updates get flushed here.
-	 * We return 1 if any requests were actually submitted.
 	 */
-	int rv = 0;
-
 	spin_lock_irq(&conf->device_lock);
 
 	if (conf->pending_bio_list.head) {
 		struct bio *bio;
 		bio = bio_list_get(&conf->pending_bio_list);
-		blk_remove_plug(conf->mddev->queue);
+		conf->pending_count = 0;
 		spin_unlock_irq(&conf->device_lock);
 		/* flush any pending bitmap writes to disk
 		 * before proceeding w/ I/O */
 		bitmap_unplug(conf->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);
+			if (unlikely((bio->bi_rw & REQ_DISCARD) &&
+			    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+				/* Just ignore it */
+				bio_endio(bio, 0);
+			else
+				generic_make_request(bio);
 			bio = next;
 		}
-		rv = 1;
 	} else
 		spin_unlock_irq(&conf->device_lock);
-	return rv;
 }
+
 /* Barriers....
  * Sometimes we need to suspend IO while we do something else,
  * either some resync/recovery, or reconfigure the array.
@@ -693,29 +999,27 @@ static int flush_pending_writes(conf_t *conf)
  *    lower_barrier when the particular background IO completes.
  */
 
-static void raise_barrier(conf_t *conf, int force)
+static void raise_barrier(struct r10conf *conf, int force)
 {
 	BUG_ON(force && !conf->barrier);
 	spin_lock_irq(&conf->resync_lock);
 
 	/* Wait until no block IO is waiting (unless 'force') */
 	wait_event_lock_irq(conf->wait_barrier, force || !conf->nr_waiting,
-			    conf->resync_lock,
-			    raid10_unplug(conf->mddev->queue));
+			    conf->resync_lock);
 
 	/* block any new IO from starting */
 	conf->barrier++;
 
-	/* No wait for all pending IO to complete */
+	/* Now wait for all pending IO to complete */
 	wait_event_lock_irq(conf->wait_barrier,
 			    !conf->nr_pending && conf->barrier < RESYNC_DEPTH,
-			    conf->resync_lock,
-			    raid10_unplug(conf->mddev->queue));
+			    conf->resync_lock);
 
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static void lower_barrier(conf_t *conf)
+static void lower_barrier(struct r10conf *conf)
 {
 	unsigned long flags;
 	spin_lock_irqsave(&conf->resync_lock, flags);
@@ -724,21 +1028,33 @@ static void lower_barrier(conf_t *conf)
 	wake_up(&conf->wait_barrier);
 }
 
-static void wait_barrier(conf_t *conf)
+static void wait_barrier(struct r10conf *conf)
 {
 	spin_lock_irq(&conf->resync_lock);
 	if (conf->barrier) {
 		conf->nr_waiting++;
-		wait_event_lock_irq(conf->wait_barrier, !conf->barrier,
-				    conf->resync_lock,
-				    raid10_unplug(conf->mddev->queue));
+		/* Wait for the barrier to drop.
+		 * However if there are already pending
+		 * requests (preventing the barrier from
+		 * rising completely), and the
+		 * pre-process bio queue isn't empty,
+		 * then don't wait, as we need to empty
+		 * that queue to get the nr_pending
+		 * count down.
+		 */
+		wait_event_lock_irq(conf->wait_barrier,
+				    !conf->barrier ||
+				    (conf->nr_pending &&
+				     current->bio_list &&
+				     !bio_list_empty(current->bio_list)),
+				    conf->resync_lock);
 		conf->nr_waiting--;
 	}
 	conf->nr_pending++;
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static void allow_barrier(conf_t *conf)
+static void allow_barrier(struct r10conf *conf)
 {
 	unsigned long flags;
 	spin_lock_irqsave(&conf->resync_lock, flags);
@@ -747,32 +1063,32 @@ static void allow_barrier(conf_t *conf)
 	wake_up(&conf->wait_barrier);
 }
 
-static void freeze_array(conf_t *conf)
+static void freeze_array(struct r10conf *conf, int extra)
 {
 	/* stop syncio and normal IO and wait for everything to
 	 * go quiet.
 	 * We increment barrier and nr_waiting, and then
-	 * wait until nr_pending match nr_queued+1
+	 * wait until nr_pending match nr_queued+extra
 	 * This is called in the context of one normal IO request
 	 * that has failed. Thus any sync request that might be pending
 	 * will be blocked by nr_pending, and we need to wait for
 	 * pending IO requests to complete or be queued for re-try.
-	 * Thus the number queued (nr_queued) plus this request (1)
+	 * Thus the number queued (nr_queued) plus this request (extra)
 	 * must match the number of pending IOs (nr_pending) before
 	 * we continue.
 	 */
 	spin_lock_irq(&conf->resync_lock);
 	conf->barrier++;
 	conf->nr_waiting++;
-	wait_event_lock_irq(conf->wait_barrier,
-			    conf->nr_pending == conf->nr_queued+1,
-			    conf->resync_lock,
-			    ({ flush_pending_writes(conf);
-			       raid10_unplug(conf->mddev->queue); }));
+	wait_event_lock_irq_cmd(conf->wait_barrier,
+				conf->nr_pending == conf->nr_queued+extra,
+				conf->resync_lock,
+				flush_pending_writes(conf));
+
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static void unfreeze_array(conf_t *conf)
+static void unfreeze_array(struct r10conf *conf)
 {
 	/* reverse the effect of the freeze */
 	spin_lock_irq(&conf->resync_lock);
@@ -782,59 +1098,79 @@ static void unfreeze_array(conf_t *conf)
 	spin_unlock_irq(&conf->resync_lock);
 }
 
-static int make_request(struct request_queue *q, struct bio * bio)
+static sector_t choose_data_offset(struct r10bio *r10_bio,
+				   struct md_rdev *rdev)
 {
-	mddev_t *mddev = q->queuedata;
-	conf_t *conf = mddev_to_conf(mddev);
-	mirror_info_t *mirror;
-	r10bio_t *r10_bio;
-	struct bio *read_bio;
-	int i;
-	int chunk_sects = conf->chunk_mask + 1;
-	const int rw = bio_data_dir(bio);
-	const int do_sync = bio_sync(bio);
-	struct bio_list bl;
-	unsigned long flags;
-	mdk_rdev_t *blocked_rdev;
+	if (!test_bit(MD_RECOVERY_RESHAPE, &rdev->mddev->recovery) ||
+	    test_bit(R10BIO_Previous, &r10_bio->state))
+		return rdev->data_offset;
+	else
+		return rdev->new_data_offset;
+}
 
-	if (unlikely(bio_barrier(bio))) {
-		bio_endio(bio, -EOPNOTSUPP);
-		return 0;
+struct raid10_plug_cb {
+	struct blk_plug_cb	cb;
+	struct bio_list		pending;
+	int			pending_cnt;
+};
+
+static void raid10_unplug(struct blk_plug_cb *cb, bool from_schedule)
+{
+	struct raid10_plug_cb *plug = container_of(cb, struct raid10_plug_cb,
+						   cb);
+	struct mddev *mddev = plug->cb.data;
+	struct r10conf *conf = mddev->private;
+	struct bio *bio;
+
+	if (from_schedule || current->bio_list) {
+		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);
+		wake_up(&conf->wait_barrier);
+		md_wakeup_thread(mddev->thread);
+		kfree(plug);
+		return;
 	}
 
-	/* If this request crosses a chunk boundary, we need to
-	 * split it.  This will only happen for 1 PAGE (or less) requests.
-	 */
-	if (unlikely( (bio->bi_sector & conf->chunk_mask) + (bio->bi_size >> 9)
-		      > chunk_sects &&
-		    conf->near_copies < conf->raid_disks)) {
-		struct bio_pair *bp;
-		/* Sanity check -- queue functions should prevent this happening */
-		if (bio->bi_vcnt != 1 ||
-		    bio->bi_idx != 0)
-			goto bad_map;
-		/* This is a one page bio that upper layers
-		 * refuse to split for us, so we need to split it.
-		 */
-		bp = bio_split(bio, bio_split_pool,
-			       chunk_sects - (bio->bi_sector & (chunk_sects - 1)) );
-		if (make_request(q, &bp->bio1))
-			generic_make_request(&bp->bio1);
-		if (make_request(q, &bp->bio2))
-			generic_make_request(&bp->bio2);
-
-		bio_pair_release(bp);
-		return 0;
-	bad_map:
-		printk("raid10_make_request bug: can't convert block across chunks"
-		       " or bigger than %dk %llu %d\n", chunk_sects/2,
-		       (unsigned long long)bio->bi_sector, bio->bi_size >> 10);
+	/* 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);
 
-		bio_io_error(bio);
-		return 0;
+	while (bio) { /* submit pending writes */
+		struct bio *next = bio->bi_next;
+		bio->bi_next = NULL;
+		if (unlikely((bio->bi_rw & REQ_DISCARD) &&
+		    !blk_queue_discard(bdev_get_queue(bio->bi_bdev))))
+			/* Just ignore it */
+			bio_endio(bio, 0);
+		else
+			generic_make_request(bio);
+		bio = next;
 	}
+	kfree(plug);
+}
 
-	md_write_start(mddev, bio);
+static void __make_request(struct mddev *mddev, struct bio *bio)
+{
+	struct r10conf *conf = mddev->private;
+	struct r10bio *r10_bio;
+	struct bio *read_bio;
+	int i;
+	const int rw = bio_data_dir(bio);
+	const unsigned long do_sync = (bio->bi_rw & REQ_SYNC);
+	const unsigned long do_fua = (bio->bi_rw & REQ_FUA);
+	const unsigned long do_discard = (bio->bi_rw
+					  & (REQ_DISCARD | REQ_SECURE));
+	const unsigned long do_same = (bio->bi_rw & REQ_WRITE_SAME);
+	unsigned long flags;
+	struct md_rdev *blocked_rdev;
+	struct blk_plug_cb *cb;
+	struct raid10_plug_cb *plug = NULL;
+	int sectors_handled;
+	int max_sectors;
+	int sectors;
 
 	/*
 	 * Register the new request and wait if the reconstruction
@@ -843,70 +1179,227 @@ static int make_request(struct request_queue *q, struct bio * bio)
 	 */
 	wait_barrier(conf);
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bio));
+	sectors = bio_sectors(bio);
+	while (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    bio->bi_iter.bi_sector < conf->reshape_progress &&
+	    bio->bi_iter.bi_sector + sectors > conf->reshape_progress) {
+		/* IO spans the reshape position.  Need to wait for
+		 * reshape to pass
+		 */
+		allow_barrier(conf);
+		wait_event(conf->wait_barrier,
+			   conf->reshape_progress <= bio->bi_iter.bi_sector ||
+			   conf->reshape_progress >= bio->bi_iter.bi_sector +
+			   sectors);
+		wait_barrier(conf);
+	}
+	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    bio_data_dir(bio) == WRITE &&
+	    (mddev->reshape_backwards
+	     ? (bio->bi_iter.bi_sector < conf->reshape_safe &&
+		bio->bi_iter.bi_sector + sectors > conf->reshape_progress)
+	     : (bio->bi_iter.bi_sector + sectors > conf->reshape_safe &&
+		bio->bi_iter.bi_sector < conf->reshape_progress))) {
+		/* Need to update reshape_position in metadata */
+		mddev->reshape_position = conf->reshape_progress;
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		set_bit(MD_CHANGE_PENDING, &mddev->flags);
+		md_wakeup_thread(mddev->thread);
+		wait_event(mddev->sb_wait,
+			   !test_bit(MD_CHANGE_PENDING, &mddev->flags));
+
+		conf->reshape_safe = mddev->reshape_position;
+	}
 
 	r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
 
 	r10_bio->master_bio = bio;
-	r10_bio->sectors = bio->bi_size >> 9;
+	r10_bio->sectors = sectors;
 
 	r10_bio->mddev = mddev;
-	r10_bio->sector = bio->bi_sector;
+	r10_bio->sector = bio->bi_iter.bi_sector;
 	r10_bio->state = 0;
 
+	/* We might need to issue multiple reads to different
+	 * devices if there are bad blocks around, so we keep
+	 * track of the number of reads in bio->bi_phys_segments.
+	 * If this is 0, there is only one r10_bio and no locking
+	 * will be needed when the request completes.  If it is
+	 * non-zero, then it is the number of not-completed requests.
+	 */
+	bio->bi_phys_segments = 0;
+	clear_bit(BIO_SEG_VALID, &bio->bi_flags);
+
 	if (rw == READ) {
 		/*
 		 * read balancing logic:
 		 */
-		int disk = read_balance(conf, r10_bio);
-		int slot = r10_bio->read_slot;
-		if (disk < 0) {
+		struct md_rdev *rdev;
+		int slot;
+
+read_again:
+		rdev = read_balance(conf, r10_bio, &max_sectors);
+		if (!rdev) {
 			raid_end_bio_io(r10_bio);
-			return 0;
+			return;
 		}
-		mirror = conf->mirrors + disk;
+		slot = r10_bio->read_slot;
 
-		read_bio = bio_clone(bio, GFP_NOIO);
+		read_bio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		bio_trim(read_bio, r10_bio->sector - bio->bi_iter.bi_sector,
+			 max_sectors);
 
 		r10_bio->devs[slot].bio = read_bio;
+		r10_bio->devs[slot].rdev = rdev;
 
-		read_bio->bi_sector = r10_bio->devs[slot].addr +
-			mirror->rdev->data_offset;
-		read_bio->bi_bdev = mirror->rdev->bdev;
+		read_bio->bi_iter.bi_sector = r10_bio->devs[slot].addr +
+			choose_data_offset(r10_bio, rdev);
+		read_bio->bi_bdev = rdev->bdev;
 		read_bio->bi_end_io = raid10_end_read_request;
 		read_bio->bi_rw = READ | do_sync;
 		read_bio->bi_private = r10_bio;
 
-		generic_make_request(read_bio);
-		return 0;
+		if (max_sectors < r10_bio->sectors) {
+			/* Could not read all from this device, so we will
+			 * need another r10_bio.
+			 */
+			sectors_handled = (r10_bio->sector + max_sectors
+					   - bio->bi_iter.bi_sector);
+			r10_bio->sectors = max_sectors;
+			spin_lock_irq(&conf->device_lock);
+			if (bio->bi_phys_segments == 0)
+				bio->bi_phys_segments = 2;
+			else
+				bio->bi_phys_segments++;
+			spin_unlock_irq(&conf->device_lock);
+			/* Cannot call generic_make_request directly
+			 * as that will be queued in __generic_make_request
+			 * and subsequent mempool_alloc might block
+			 * waiting for it.  so hand bio over to raid10d.
+			 */
+			reschedule_retry(r10_bio);
+
+			r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
+
+			r10_bio->master_bio = bio;
+			r10_bio->sectors = bio_sectors(bio) - sectors_handled;
+			r10_bio->state = 0;
+			r10_bio->mddev = mddev;
+			r10_bio->sector = bio->bi_iter.bi_sector +
+				sectors_handled;
+			goto read_again;
+		} else
+			generic_make_request(read_bio);
+		return;
 	}
 
 	/*
 	 * WRITE:
 	 */
+	if (conf->pending_count >= max_queued_requests) {
+		md_wakeup_thread(mddev->thread);
+		wait_event(conf->wait_barrier,
+			   conf->pending_count < max_queued_requests);
+	}
 	/* first select target devices under rcu_lock and
 	 * inc refcount on their rdev.  Record them by setting
 	 * bios[x] to bio
+	 * If there are known/acknowledged bad blocks on any device
+	 * on which we have seen a write error, we want to avoid
+	 * writing to those blocks.  This potentially requires several
+	 * writes to write around the bad blocks.  Each set of writes
+	 * gets its own r10_bio with a set of bios attached.  The number
+	 * of r10_bios is recored in bio->bi_phys_segments just as with
+	 * the read case.
 	 */
+
+	r10_bio->read_slot = -1; /* make sure repl_bio gets freed */
 	raid10_find_phys(conf, r10_bio);
- retry_write:
+retry_write:
 	blocked_rdev = NULL;
 	rcu_read_lock();
+	max_sectors = r10_bio->sectors;
+
 	for (i = 0;  i < conf->copies; i++) {
 		int d = r10_bio->devs[i].devnum;
-		mdk_rdev_t *rdev = rcu_dereference(conf->mirrors[d].rdev);
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[d].rdev);
+		struct md_rdev *rrdev = rcu_dereference(
+			conf->mirrors[d].replacement);
+		if (rdev == rrdev)
+			rrdev = NULL;
 		if (rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
 			atomic_inc(&rdev->nr_pending);
 			blocked_rdev = rdev;
 			break;
 		}
-		if (rdev && !test_bit(Faulty, &rdev->flags)) {
-			atomic_inc(&rdev->nr_pending);
-			r10_bio->devs[i].bio = bio;
-		} else {
-			r10_bio->devs[i].bio = NULL;
+		if (rrdev && unlikely(test_bit(Blocked, &rrdev->flags))) {
+			atomic_inc(&rrdev->nr_pending);
+			blocked_rdev = rrdev;
+			break;
+		}
+		if (rdev && (test_bit(Faulty, &rdev->flags)
+			     || test_bit(Unmerged, &rdev->flags)))
+			rdev = NULL;
+		if (rrdev && (test_bit(Faulty, &rrdev->flags)
+			      || test_bit(Unmerged, &rrdev->flags)))
+			rrdev = NULL;
+
+		r10_bio->devs[i].bio = NULL;
+		r10_bio->devs[i].repl_bio = NULL;
+
+		if (!rdev && !rrdev) {
 			set_bit(R10BIO_Degraded, &r10_bio->state);
+			continue;
+		}
+		if (rdev && test_bit(WriteErrorSeen, &rdev->flags)) {
+			sector_t first_bad;
+			sector_t dev_sector = r10_bio->devs[i].addr;
+			int bad_sectors;
+			int is_bad;
+
+			is_bad = is_badblock(rdev, dev_sector,
+					     max_sectors,
+					     &first_bad, &bad_sectors);
+			if (is_bad < 0) {
+				/* Mustn't write here until the bad block
+				 * is acknowledged
+				 */
+				atomic_inc(&rdev->nr_pending);
+				set_bit(BlockedBadBlocks, &rdev->flags);
+				blocked_rdev = rdev;
+				break;
+			}
+			if (is_bad && first_bad <= dev_sector) {
+				/* Cannot write here at all */
+				bad_sectors -= (dev_sector - first_bad);
+				if (bad_sectors < max_sectors)
+					/* Mustn't write more than bad_sectors
+					 * to other devices yet
+					 */
+					max_sectors = bad_sectors;
+				/* We don't set R10BIO_Degraded as that
+				 * only applies if the disk is missing,
+				 * so it might be re-added, and we want to
+				 * know to recover this chunk.
+				 * In this case the device is here, and the
+				 * fact that this chunk is not in-sync is
+				 * recorded in the bad block log.
+				 */
+				continue;
+			}
+			if (is_bad) {
+				int good_sectors = first_bad - dev_sector;
+				if (good_sectors < max_sectors)
+					max_sectors = good_sectors;
+			}
+		}
+		if (rdev) {
+			r10_bio->devs[i].bio = bio;
+			atomic_inc(&rdev->nr_pending);
+		}
+		if (rrdev) {
+			r10_bio->devs[i].repl_bio = bio;
+			atomic_inc(&rrdev->nr_pending);
 		}
 	}
 	rcu_read_unlock();
@@ -916,90 +1409,265 @@ static int make_request(struct request_queue *q, struct bio * bio)
 		int j;
 		int d;
 
-		for (j = 0; j < i; j++)
+		for (j = 0; j < i; j++) {
 			if (r10_bio->devs[j].bio) {
 				d = r10_bio->devs[j].devnum;
 				rdev_dec_pending(conf->mirrors[d].rdev, mddev);
 			}
+			if (r10_bio->devs[j].repl_bio) {
+				struct md_rdev *rdev;
+				d = r10_bio->devs[j].devnum;
+				rdev = conf->mirrors[d].replacement;
+				if (!rdev) {
+					/* Race with remove_disk */
+					smp_mb();
+					rdev = conf->mirrors[d].rdev;
+				}
+				rdev_dec_pending(rdev, mddev);
+			}
+		}
 		allow_barrier(conf);
 		md_wait_for_blocked_rdev(blocked_rdev, mddev);
 		wait_barrier(conf);
 		goto retry_write;
 	}
 
-	atomic_set(&r10_bio->remaining, 0);
+	if (max_sectors < r10_bio->sectors) {
+		/* We are splitting this into multiple parts, so
+		 * we need to prepare for allocating another r10_bio.
+		 */
+		r10_bio->sectors = max_sectors;
+		spin_lock_irq(&conf->device_lock);
+		if (bio->bi_phys_segments == 0)
+			bio->bi_phys_segments = 2;
+		else
+			bio->bi_phys_segments++;
+		spin_unlock_irq(&conf->device_lock);
+	}
+	sectors_handled = r10_bio->sector + max_sectors -
+		bio->bi_iter.bi_sector;
+
+	atomic_set(&r10_bio->remaining, 1);
+	bitmap_startwrite(mddev->bitmap, r10_bio->sector, r10_bio->sectors, 0);
 
-	bio_list_init(&bl);
 	for (i = 0; i < conf->copies; i++) {
 		struct bio *mbio;
 		int d = r10_bio->devs[i].devnum;
-		if (!r10_bio->devs[i].bio)
-			continue;
+		if (r10_bio->devs[i].bio) {
+			struct md_rdev *rdev = conf->mirrors[d].rdev;
+			mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+			bio_trim(mbio, r10_bio->sector - bio->bi_iter.bi_sector,
+				 max_sectors);
+			r10_bio->devs[i].bio = mbio;
+
+			mbio->bi_iter.bi_sector	= (r10_bio->devs[i].addr+
+					   choose_data_offset(r10_bio,
+							      rdev));
+			mbio->bi_bdev = rdev->bdev;
+			mbio->bi_end_io	= raid10_end_write_request;
+			mbio->bi_rw =
+				WRITE | do_sync | do_fua | do_discard | do_same;
+			mbio->bi_private = r10_bio;
+
+			atomic_inc(&r10_bio->remaining);
 
-		mbio = bio_clone(bio, GFP_NOIO);
-		r10_bio->devs[i].bio = mbio;
+			cb = blk_check_plugged(raid10_unplug, mddev,
+					       sizeof(*plug));
+			if (cb)
+				plug = container_of(cb, struct raid10_plug_cb,
+						    cb);
+			else
+				plug = NULL;
+			spin_lock_irqsave(&conf->device_lock, flags);
+			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);
+		}
 
-		mbio->bi_sector	= r10_bio->devs[i].addr+
-			conf->mirrors[d].rdev->data_offset;
-		mbio->bi_bdev = conf->mirrors[d].rdev->bdev;
-		mbio->bi_end_io	= raid10_end_write_request;
-		mbio->bi_rw = WRITE | do_sync;
-		mbio->bi_private = r10_bio;
+		if (r10_bio->devs[i].repl_bio) {
+			struct md_rdev *rdev = conf->mirrors[d].replacement;
+			if (rdev == NULL) {
+				/* Replacement just got moved to main 'rdev' */
+				smp_mb();
+				rdev = conf->mirrors[d].rdev;
+			}
+			mbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+			bio_trim(mbio, r10_bio->sector - bio->bi_iter.bi_sector,
+				 max_sectors);
+			r10_bio->devs[i].repl_bio = mbio;
+
+			mbio->bi_iter.bi_sector	= (r10_bio->devs[i].addr +
+					   choose_data_offset(
+						   r10_bio, rdev));
+			mbio->bi_bdev = rdev->bdev;
+			mbio->bi_end_io	= raid10_end_write_request;
+			mbio->bi_rw =
+				WRITE | do_sync | do_fua | do_discard | do_same;
+			mbio->bi_private = r10_bio;
 
-		atomic_inc(&r10_bio->remaining);
-		bio_list_add(&bl, mbio);
+			atomic_inc(&r10_bio->remaining);
+			spin_lock_irqsave(&conf->device_lock, flags);
+			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 (unlikely(!atomic_read(&r10_bio->remaining))) {
-		/* the array is dead */
-		md_write_end(mddev);
-		raid_end_bio_io(r10_bio);
-		return 0;
+	/* Don't remove the bias on 'remaining' (one_write_done) until
+	 * after checking if we need to go around again.
+	 */
+
+	if (sectors_handled < bio_sectors(bio)) {
+		one_write_done(r10_bio);
+		/* We need another r10_bio.  It has already been counted
+		 * in bio->bi_phys_segments.
+		 */
+		r10_bio = mempool_alloc(conf->r10bio_pool, GFP_NOIO);
+
+		r10_bio->master_bio = bio;
+		r10_bio->sectors = bio_sectors(bio) - sectors_handled;
+
+		r10_bio->mddev = mddev;
+		r10_bio->sector = bio->bi_iter.bi_sector + sectors_handled;
+		r10_bio->state = 0;
+		goto retry_write;
 	}
+	one_write_done(r10_bio);
+}
 
-	bitmap_startwrite(mddev->bitmap, bio->bi_sector, r10_bio->sectors, 0);
-	spin_lock_irqsave(&conf->device_lock, flags);
-	bio_list_merge(&conf->pending_bio_list, &bl);
-	blk_plug_device(mddev->queue);
-	spin_unlock_irqrestore(&conf->device_lock, flags);
+static void make_request(struct mddev *mddev, struct bio *bio)
+{
+	struct r10conf *conf = mddev->private;
+	sector_t chunk_mask = (conf->geo.chunk_mask & conf->prev.chunk_mask);
+	int chunk_sects = chunk_mask + 1;
 
-	/* In case raid10d snuck in to freeze_array */
-	wake_up(&conf->wait_barrier);
+	struct bio *split;
 
-	if (do_sync)
-		md_wakeup_thread(mddev->thread);
+	if (unlikely(bio->bi_rw & REQ_FLUSH)) {
+		md_flush_request(mddev, bio);
+		return;
+	}
 
-	return 0;
+	md_write_start(mddev, bio);
+
+
+	do {
+
+		/*
+		 * If this request crosses a chunk boundary, we need to split
+		 * it.
+		 */
+		if (unlikely((bio->bi_iter.bi_sector & chunk_mask) +
+			     bio_sectors(bio) > chunk_sects
+			     && (conf->geo.near_copies < conf->geo.raid_disks
+				 || conf->prev.near_copies <
+				 conf->prev.raid_disks))) {
+			split = bio_split(bio, chunk_sects -
+					  (bio->bi_iter.bi_sector &
+					   (chunk_sects - 1)),
+					  GFP_NOIO, fs_bio_set);
+			bio_chain(split, bio);
+		} else {
+			split = bio;
+		}
+
+		__make_request(mddev, split);
+	} while (split != bio);
+
+	/* In case raid10d snuck in to freeze_array */
+	wake_up(&conf->wait_barrier);
 }
 
-static void status(struct seq_file *seq, mddev_t *mddev)
+static void status(struct seq_file *seq, struct mddev *mddev)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
 	int i;
 
-	if (conf->near_copies < conf->raid_disks)
-		seq_printf(seq, " %dK chunks", mddev->chunk_size/1024);
-	if (conf->near_copies > 1)
-		seq_printf(seq, " %d near-copies", conf->near_copies);
-	if (conf->far_copies > 1) {
-		if (conf->far_offset)
-			seq_printf(seq, " %d offset-copies", conf->far_copies);
+	if (conf->geo.near_copies < conf->geo.raid_disks)
+		seq_printf(seq, " %dK chunks", mddev->chunk_sectors / 2);
+	if (conf->geo.near_copies > 1)
+		seq_printf(seq, " %d near-copies", conf->geo.near_copies);
+	if (conf->geo.far_copies > 1) {
+		if (conf->geo.far_offset)
+			seq_printf(seq, " %d offset-copies", conf->geo.far_copies);
 		else
-			seq_printf(seq, " %d far-copies", conf->far_copies);
+			seq_printf(seq, " %d far-copies", conf->geo.far_copies);
 	}
-	seq_printf(seq, " [%d/%d] [", conf->raid_disks,
-					conf->raid_disks - mddev->degraded);
-	for (i = 0; i < conf->raid_disks; i++)
+	seq_printf(seq, " [%d/%d] [", conf->geo.raid_disks,
+					conf->geo.raid_disks - mddev->degraded);
+	for (i = 0; i < conf->geo.raid_disks; i++)
 		seq_printf(seq, "%s",
 			      conf->mirrors[i].rdev &&
 			      test_bit(In_sync, &conf->mirrors[i].rdev->flags) ? "U" : "_");
 	seq_printf(seq, "]");
 }
 
-static void error(mddev_t *mddev, mdk_rdev_t *rdev)
+/* check if there are enough drives for
+ * every block to appear on atleast one.
+ * Don't consider the device numbered 'ignore'
+ * as we might be about to remove it.
+ */
+static int _enough(struct r10conf *conf, int previous, int ignore)
+{
+	int first = 0;
+	int has_enough = 0;
+	int disks, ncopies;
+	if (previous) {
+		disks = conf->prev.raid_disks;
+		ncopies = conf->prev.near_copies;
+	} else {
+		disks = conf->geo.raid_disks;
+		ncopies = conf->geo.near_copies;
+	}
+
+	rcu_read_lock();
+	do {
+		int n = conf->copies;
+		int cnt = 0;
+		int this = first;
+		while (n--) {
+			struct md_rdev *rdev;
+			if (this != ignore &&
+			    (rdev = rcu_dereference(conf->mirrors[this].rdev)) &&
+			    test_bit(In_sync, &rdev->flags))
+				cnt++;
+			this = (this+1) % disks;
+		}
+		if (cnt == 0)
+			goto out;
+		first = (first + ncopies) % disks;
+	} while (first != 0);
+	has_enough = 1;
+out:
+	rcu_read_unlock();
+	return has_enough;
+}
+
+static int enough(struct r10conf *conf, int ignore)
+{
+	/* when calling 'enough', both 'prev' and 'geo' must
+	 * be stable.
+	 * This is ensured if ->reconfig_mutex or ->device_lock
+	 * is held.
+	 */
+	return _enough(conf, 0, ignore) &&
+		_enough(conf, 1, ignore);
+}
+
+static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
+	unsigned long flags;
 
 	/*
 	 * If it is not operational, then we have already marked it as dead
@@ -1007,58 +1675,58 @@ static void error(mddev_t *mddev, mdk_rdev_t *rdev)
 	 * next level up know.
 	 * else mark the drive as failed
 	 */
+	spin_lock_irqsave(&conf->device_lock, flags);
 	if (test_bit(In_sync, &rdev->flags)
-	    && conf->raid_disks-mddev->degraded == 1)
+	    && !enough(conf, rdev->raid_disk)) {
 		/*
 		 * Don't fail the drive, just return an IO error.
-		 * The test should really be more sophisticated than
-		 * "working_disks == 1", but it isn't critical, and
-		 * can wait until we do more sophisticated "is the drive
-		 * really dead" tests...
 		 */
+		spin_unlock_irqrestore(&conf->device_lock, flags);
 		return;
+	}
 	if (test_and_clear_bit(In_sync, &rdev->flags)) {
-		unsigned long flags;
-		spin_lock_irqsave(&conf->device_lock, flags);
 		mddev->degraded++;
-		spin_unlock_irqrestore(&conf->device_lock, flags);
-		/*
+			/*
 		 * if recovery is running, make sure it aborts.
 		 */
 		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
 	}
+	set_bit(Blocked, &rdev->flags);
 	set_bit(Faulty, &rdev->flags);
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
-	printk(KERN_ALERT "raid10: Disk failure on %s, disabling device.\n"
-		"raid10: Operation continuing on %d devices.\n",
-		bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
+	spin_unlock_irqrestore(&conf->device_lock, flags);
+	printk(KERN_ALERT
+	       "md/raid10:%s: Disk failure on %s, disabling device.\n"
+	       "md/raid10:%s: Operation continuing on %d devices.\n",
+	       mdname(mddev), bdevname(rdev->bdev, b),
+	       mdname(mddev), conf->geo.raid_disks - mddev->degraded);
 }
 
-static void print_conf(conf_t *conf)
+static void print_conf(struct r10conf *conf)
 {
 	int i;
-	mirror_info_t *tmp;
+	struct raid10_info *tmp;
 
-	printk("RAID10 conf printout:\n");
+	printk(KERN_DEBUG "RAID10 conf printout:\n");
 	if (!conf) {
-		printk("(!conf)\n");
+		printk(KERN_DEBUG "(!conf)\n");
 		return;
 	}
-	printk(" --- wd:%d rd:%d\n", conf->raid_disks - conf->mddev->degraded,
-		conf->raid_disks);
+	printk(KERN_DEBUG " --- wd:%d rd:%d\n", conf->geo.raid_disks - conf->mddev->degraded,
+		conf->geo.raid_disks);
 
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->geo.raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->mirrors + i;
 		if (tmp->rdev)
-			printk(" disk %d, wo:%d, o:%d, dev:%s\n",
+			printk(KERN_DEBUG " disk %d, wo:%d, o:%d, dev:%s\n",
 				i, !test_bit(In_sync, &tmp->rdev->flags),
 			        !test_bit(Faulty, &tmp->rdev->flags),
 				bdevname(tmp->rdev->bdev,b));
 	}
 }
 
-static void close_sync(conf_t *conf)
+static void close_sync(struct r10conf *conf)
 {
 	wait_barrier(conf);
 	allow_barrier(conf);
@@ -1067,137 +1735,195 @@ static void close_sync(conf_t *conf)
 	conf->r10buf_pool = NULL;
 }
 
-/* check if there are enough drives for
- * every block to appear on atleast one
- */
-static int enough(conf_t *conf)
-{
-	int first = 0;
-
-	do {
-		int n = conf->copies;
-		int cnt = 0;
-		while (n--) {
-			if (conf->mirrors[first].rdev)
-				cnt++;
-			first = (first+1) % conf->raid_disks;
-		}
-		if (cnt == 0)
-			return 0;
-	} while (first != 0);
-	return 1;
-}
-
-static int raid10_spare_active(mddev_t *mddev)
+static int raid10_spare_active(struct mddev *mddev)
 {
 	int i;
-	conf_t *conf = mddev->private;
-	mirror_info_t *tmp;
+	struct r10conf *conf = mddev->private;
+	struct raid10_info *tmp;
+	int count = 0;
+	unsigned long flags;
 
 	/*
 	 * Find all non-in_sync disks within the RAID10 configuration
 	 * and mark them in_sync
 	 */
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0; i < conf->geo.raid_disks; i++) {
 		tmp = conf->mirrors + i;
-		if (tmp->rdev
-		    && !test_bit(Faulty, &tmp->rdev->flags)
-		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
-			unsigned long flags;
-			spin_lock_irqsave(&conf->device_lock, flags);
-			mddev->degraded--;
-			spin_unlock_irqrestore(&conf->device_lock, flags);
+		if (tmp->replacement
+		    && tmp->replacement->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &tmp->replacement->flags)
+		    && !test_and_set_bit(In_sync, &tmp->replacement->flags)) {
+			/* Replacement has just become active */
+			if (!tmp->rdev
+			    || !test_and_clear_bit(In_sync, &tmp->rdev->flags))
+				count++;
+			if (tmp->rdev) {
+				/* Replaced device not technically faulty,
+				 * but we need to be sure it gets removed
+				 * and never re-added.
+				 */
+				set_bit(Faulty, &tmp->rdev->flags);
+				sysfs_notify_dirent_safe(
+					tmp->rdev->sysfs_state);
+			}
+			sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
+		} else if (tmp->rdev
+			   && tmp->rdev->recovery_offset == MaxSector
+			   && !test_bit(Faulty, &tmp->rdev->flags)
+			   && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
+			count++;
+			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
 		}
 	}
+	spin_lock_irqsave(&conf->device_lock, flags);
+	mddev->degraded -= count;
+	spin_unlock_irqrestore(&conf->device_lock, flags);
 
 	print_conf(conf);
-	return 0;
+	return count;
 }
 
 
-static int raid10_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
+static int raid10_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	conf_t *conf = mddev->private;
+	struct r10conf *conf = mddev->private;
 	int err = -EEXIST;
 	int mirror;
-	mirror_info_t *p;
 	int first = 0;
-	int last = mddev->raid_disks - 1;
+	int last = conf->geo.raid_disks - 1;
+	struct request_queue *q = bdev_get_queue(rdev->bdev);
 
 	if (mddev->recovery_cp < MaxSector)
 		/* only hot-add to in-sync arrays, as recovery is
 		 * very different from resync
 		 */
 		return -EBUSY;
-	if (!enough(conf))
+	if (rdev->saved_raid_disk < 0 && !_enough(conf, 1, -1))
 		return -EINVAL;
 
-	if (rdev->raid_disk)
+	if (rdev->raid_disk >= 0)
 		first = last = rdev->raid_disk;
 
-	if (rdev->saved_raid_disk >= 0 &&
-	    rdev->saved_raid_disk >= first &&
+	if (q->merge_bvec_fn) {
+		set_bit(Unmerged, &rdev->flags);
+		mddev->merge_check_needed = 1;
+	}
+
+	if (rdev->saved_raid_disk >= first &&
 	    conf->mirrors[rdev->saved_raid_disk].rdev == NULL)
 		mirror = rdev->saved_raid_disk;
 	else
 		mirror = first;
-	for ( ; mirror <= last ; mirror++)
-		if ( !(p=conf->mirrors+mirror)->rdev) {
-
-			blk_queue_stack_limits(mddev->queue,
-					       rdev->bdev->bd_disk->queue);
-			/* as we don't honour merge_bvec_fn, we must never risk
-			 * violating it, so limit ->max_sector to one PAGE, as
-			 * a one page request is never in violation.
-			 */
-			if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-			    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-				mddev->queue->max_sectors = (PAGE_SIZE>>9);
-
-			p->head_position = 0;
+	for ( ; mirror <= last ; mirror++) {
+		struct raid10_info *p = &conf->mirrors[mirror];
+		if (p->recovery_disabled == mddev->recovery_disabled)
+			continue;
+		if (p->rdev) {
+			if (!test_bit(WantReplacement, &p->rdev->flags) ||
+			    p->replacement != NULL)
+				continue;
+			clear_bit(In_sync, &rdev->flags);
+			set_bit(Replacement, &rdev->flags);
 			rdev->raid_disk = mirror;
 			err = 0;
-			if (rdev->saved_raid_disk != mirror)
-				conf->fullsync = 1;
-			rcu_assign_pointer(p->rdev, rdev);
+			if (mddev->gendisk)
+				disk_stack_limits(mddev->gendisk, rdev->bdev,
+						  rdev->data_offset << 9);
+			conf->fullsync = 1;
+			rcu_assign_pointer(p->replacement, rdev);
 			break;
 		}
 
+		if (mddev->gendisk)
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
+
+		p->head_position = 0;
+		p->recovery_disabled = mddev->recovery_disabled - 1;
+		rdev->raid_disk = mirror;
+		err = 0;
+		if (rdev->saved_raid_disk != mirror)
+			conf->fullsync = 1;
+		rcu_assign_pointer(p->rdev, rdev);
+		break;
+	}
+	if (err == 0 && test_bit(Unmerged, &rdev->flags)) {
+		/* Some requests might not have seen this new
+		 * merge_bvec_fn.  We must wait for them to complete
+		 * before merging the device fully.
+		 * First we make sure any code which has tested
+		 * our function has submitted the request, then
+		 * we wait for all outstanding requests to complete.
+		 */
+		synchronize_sched();
+		freeze_array(conf, 0);
+		unfreeze_array(conf);
+		clear_bit(Unmerged, &rdev->flags);
+	}
+	md_integrity_add_rdev(rdev, mddev);
+	if (mddev->queue && blk_queue_discard(bdev_get_queue(rdev->bdev)))
+		queue_flag_set_unlocked(QUEUE_FLAG_DISCARD, mddev->queue);
+
 	print_conf(conf);
 	return err;
 }
 
-static int raid10_remove_disk(mddev_t *mddev, int number)
+static int raid10_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	conf_t *conf = mddev->private;
+	struct r10conf *conf = mddev->private;
 	int err = 0;
-	mdk_rdev_t *rdev;
-	mirror_info_t *p = conf->mirrors+ number;
+	int number = rdev->raid_disk;
+	struct md_rdev **rdevp;
+	struct raid10_info *p = conf->mirrors + number;
 
 	print_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
-		if (test_bit(In_sync, &rdev->flags) ||
-		    atomic_read(&rdev->nr_pending)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		/* Only remove faulty devices in recovery
-		 * is not possible.
-		 */
-		if (!test_bit(Faulty, &rdev->flags) &&
-		    enough(conf)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		p->rdev = NULL;
-		synchronize_rcu();
-		if (atomic_read(&rdev->nr_pending)) {
-			/* lost the race, try later */
-			err = -EBUSY;
-			p->rdev = rdev;
-		}
+	if (rdev == p->rdev)
+		rdevp = &p->rdev;
+	else if (rdev == p->replacement)
+		rdevp = &p->replacement;
+	else
+		return 0;
+
+	if (test_bit(In_sync, &rdev->flags) ||
+	    atomic_read(&rdev->nr_pending)) {
+		err = -EBUSY;
+		goto abort;
 	}
+	/* Only remove faulty devices if recovery
+	 * is not possible.
+	 */
+	if (!test_bit(Faulty, &rdev->flags) &&
+	    mddev->recovery_disabled != p->recovery_disabled &&
+	    (!p->replacement || p->replacement == rdev) &&
+	    number < conf->geo.raid_disks &&
+	    enough(conf, -1)) {
+		err = -EBUSY;
+		goto abort;
+	}
+	*rdevp = NULL;
+	synchronize_rcu();
+	if (atomic_read(&rdev->nr_pending)) {
+		/* lost the race, try later */
+		err = -EBUSY;
+		*rdevp = rdev;
+		goto abort;
+	} else if (p->replacement) {
+		/* We must have just cleared 'rdev' */
+		p->rdev = p->replacement;
+		clear_bit(Replacement, &p->replacement->flags);
+		smp_mb(); /* Make sure other CPUs may see both as identical
+			   * but will never see neither -- if they are careful.
+			   */
+		p->replacement = NULL;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else
+		/* We might have just remove the Replacement as faulty
+		 * Clear the flag just in case
+		 */
+		clear_bit(WantReplacement, &rdev->flags);
+
+	err = md_integrity_register(mddev);
+
 abort:
 
 	print_conf(conf);
@@ -1207,30 +1933,29 @@ abort:
 
 static void end_sync_read(struct bio *bio, int error)
 {
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
-	conf_t *conf = mddev_to_conf(r10_bio->mddev);
-	int i,d;
+	struct r10bio *r10_bio = bio->bi_private;
+	struct r10conf *conf = r10_bio->mddev->private;
+	int d;
 
-	for (i=0; i<conf->copies; i++)
-		if (r10_bio->devs[i].bio == bio)
-			break;
-	BUG_ON(i == conf->copies);
-	update_head_pos(i, r10_bio);
-	d = r10_bio->devs[i].devnum;
+	if (bio == r10_bio->master_bio) {
+		/* this is a reshape read */
+		d = r10_bio->read_slot; /* really the read dev */
+	} else
+		d = find_bio_disk(conf, r10_bio, bio, NULL, NULL);
 
 	if (test_bit(BIO_UPTODATE, &bio->bi_flags))
 		set_bit(R10BIO_Uptodate, &r10_bio->state);
-	else {
+	else
+		/* The write handler will notice the lack of
+		 * R10BIO_Uptodate and record any errors etc
+		 */
 		atomic_add(r10_bio->sectors,
 			   &conf->mirrors[d].rdev->corrected_errors);
-		if (!test_bit(MD_RECOVERY_SYNC, &conf->mddev->recovery))
-			md_error(r10_bio->mddev,
-				 conf->mirrors[d].rdev);
-	}
 
 	/* for reconstruct, we always reschedule after a read.
 	 * for resync, only after all reads
 	 */
+	rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
 	if (test_bit(R10BIO_IsRecover, &r10_bio->state) ||
 	    atomic_dec_and_test(&r10_bio->remaining)) {
 		/* we have read all the blocks,
@@ -1238,40 +1963,73 @@ static void end_sync_read(struct bio *bio, int error)
 		 */
 		reschedule_retry(r10_bio);
 	}
-	rdev_dec_pending(conf->mirrors[d].rdev, conf->mddev);
 }
 
-static void end_sync_write(struct bio *bio, int error)
+static void end_sync_request(struct r10bio *r10_bio)
 {
-	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
-	r10bio_t * r10_bio = (r10bio_t *)(bio->bi_private);
-	mddev_t *mddev = r10_bio->mddev;
-	conf_t *conf = mddev_to_conf(mddev);
-	int i,d;
-
-	for (i = 0; i < conf->copies; i++)
-		if (r10_bio->devs[i].bio == bio)
-			break;
-	d = r10_bio->devs[i].devnum;
-
-	if (!uptodate)
-		md_error(mddev, conf->mirrors[d].rdev);
-
-	update_head_pos(i, r10_bio);
+	struct mddev *mddev = r10_bio->mddev;
 
 	while (atomic_dec_and_test(&r10_bio->remaining)) {
 		if (r10_bio->master_bio == NULL) {
 			/* the primary of several recovery bios */
-			md_done_sync(mddev, r10_bio->sectors, 1);
-			put_buf(r10_bio);
+			sector_t s = r10_bio->sectors;
+			if (test_bit(R10BIO_MadeGood, &r10_bio->state) ||
+			    test_bit(R10BIO_WriteError, &r10_bio->state))
+				reschedule_retry(r10_bio);
+			else
+				put_buf(r10_bio);
+			md_done_sync(mddev, s, 1);
 			break;
 		} else {
-			r10bio_t *r10_bio2 = (r10bio_t *)r10_bio->master_bio;
-			put_buf(r10_bio);
+			struct r10bio *r10_bio2 = (struct r10bio *)r10_bio->master_bio;
+			if (test_bit(R10BIO_MadeGood, &r10_bio->state) ||
+			    test_bit(R10BIO_WriteError, &r10_bio->state))
+				reschedule_retry(r10_bio);
+			else
+				put_buf(r10_bio);
 			r10_bio = r10_bio2;
 		}
 	}
-	rdev_dec_pending(conf->mirrors[d].rdev, mddev);
+}
+
+static void end_sync_write(struct bio *bio, int error)
+{
+	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct r10bio *r10_bio = bio->bi_private;
+	struct mddev *mddev = r10_bio->mddev;
+	struct r10conf *conf = mddev->private;
+	int d;
+	sector_t first_bad;
+	int bad_sectors;
+	int slot;
+	int repl;
+	struct md_rdev *rdev = NULL;
+
+	d = find_bio_disk(conf, r10_bio, bio, &slot, &repl);
+	if (repl)
+		rdev = conf->mirrors[d].replacement;
+	else
+		rdev = conf->mirrors[d].rdev;
+
+	if (!uptodate) {
+		if (repl)
+			md_error(mddev, rdev);
+		else {
+			set_bit(WriteErrorSeen, &rdev->flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+			set_bit(R10BIO_WriteError, &r10_bio->state);
+		}
+	} else if (is_badblock(rdev,
+			     r10_bio->devs[slot].addr,
+			     r10_bio->sectors,
+			     &first_bad, &bad_sectors))
+		set_bit(R10BIO_MadeGood, &r10_bio->state);
+
+	rdev_dec_pending(rdev, mddev);
+
+	end_sync_request(r10_bio);
 }
 
 /*
@@ -1290,11 +2048,12 @@ static void end_sync_write(struct bio *bio, int error)
  * We check if all blocks are in-sync and only write to blocks that
  * aren't in sync
  */
-static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
+static void sync_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
 	int i, first;
 	struct bio *tbio, *fbio;
+	int vcnt;
 
 	atomic_set(&r10_bio->remaining, 1);
 
@@ -1309,10 +2068,10 @@ static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
 	first = i;
 	fbio = r10_bio->devs[i].bio;
 
+	vcnt = (r10_bio->sectors + (PAGE_SIZE >> 9) - 1) >> (PAGE_SHIFT - 9);
 	/* now find blocks with errors */
 	for (i=0 ; i < conf->copies ; i++) {
 		int  j, d;
-		int vcnt = r10_bio->sectors >> (PAGE_SHIFT-9);
 
 		tbio = r10_bio->devs[i].bio;
 
@@ -1325,35 +2084,36 @@ static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
 			 * both 'first' and 'i', so we just compare them.
 			 * All vec entries are PAGE_SIZE;
 			 */
-			for (j = 0; j < vcnt; j++)
+			int sectors = r10_bio->sectors;
+			for (j = 0; j < vcnt; j++) {
+				int len = PAGE_SIZE;
+				if (sectors < (len / 512))
+					len = sectors * 512;
 				if (memcmp(page_address(fbio->bi_io_vec[j].bv_page),
 					   page_address(tbio->bi_io_vec[j].bv_page),
-					   PAGE_SIZE))
+					   len))
 					break;
+				sectors -= len/512;
+			}
 			if (j == vcnt)
 				continue;
-			mddev->resync_mismatches += r10_bio->sectors;
+			atomic64_add(r10_bio->sectors, &mddev->resync_mismatches);
+			if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
+				/* Don't fix anything. */
+				continue;
 		}
-		if (test_bit(MD_RECOVERY_CHECK, &mddev->recovery))
-			/* Don't fix anything. */
-			continue;
-		/* Ok, we need to write this bio
+		/* Ok, we need to write this bio, either to correct an
+		 * inconsistency or to correct an unreadable block.
 		 * First we need to fixup bv_offset, bv_len and
 		 * bi_vecs, as the read request might have corrupted these
 		 */
+		bio_reset(tbio);
+
 		tbio->bi_vcnt = vcnt;
-		tbio->bi_size = r10_bio->sectors << 9;
-		tbio->bi_idx = 0;
-		tbio->bi_phys_segments = 0;
-		tbio->bi_hw_segments = 0;
-		tbio->bi_hw_front_size = 0;
-		tbio->bi_hw_back_size = 0;
-		tbio->bi_flags &= ~(BIO_POOL_MASK - 1);
-		tbio->bi_flags |= 1 << BIO_UPTODATE;
-		tbio->bi_next = NULL;
+		tbio->bi_iter.bi_size = r10_bio->sectors << 9;
 		tbio->bi_rw = WRITE;
 		tbio->bi_private = r10_bio;
-		tbio->bi_sector = r10_bio->devs[i].addr;
+		tbio->bi_iter.bi_sector = r10_bio->devs[i].addr;
 
 		for (j=0; j < vcnt ; j++) {
 			tbio->bi_io_vec[j].bv_offset = 0;
@@ -1368,13 +2128,35 @@ static void sync_request_write(mddev_t *mddev, r10bio_t *r10_bio)
 		d = r10_bio->devs[i].devnum;
 		atomic_inc(&conf->mirrors[d].rdev->nr_pending);
 		atomic_inc(&r10_bio->remaining);
-		md_sync_acct(conf->mirrors[d].rdev->bdev, tbio->bi_size >> 9);
+		md_sync_acct(conf->mirrors[d].rdev->bdev, bio_sectors(tbio));
 
-		tbio->bi_sector += conf->mirrors[d].rdev->data_offset;
+		tbio->bi_iter.bi_sector += conf->mirrors[d].rdev->data_offset;
 		tbio->bi_bdev = conf->mirrors[d].rdev->bdev;
 		generic_make_request(tbio);
 	}
 
+	/* Now write out to any replacement devices
+	 * that are active
+	 */
+	for (i = 0; i < conf->copies; i++) {
+		int j, d;
+
+		tbio = r10_bio->devs[i].repl_bio;
+		if (!tbio || !tbio->bi_end_io)
+			continue;
+		if (r10_bio->devs[i].bio->bi_end_io != end_sync_write
+		    && r10_bio->devs[i].bio != fbio)
+			for (j = 0; j < vcnt; j++)
+				memcpy(page_address(tbio->bi_io_vec[j].bv_page),
+				       page_address(fbio->bi_io_vec[j].bv_page),
+				       PAGE_SIZE);
+		d = r10_bio->devs[i].devnum;
+		atomic_inc(&r10_bio->remaining);
+		md_sync_acct(conf->mirrors[d].replacement->bdev,
+			     bio_sectors(tbio));
+		generic_make_request(tbio);
+	}
+
 done:
 	if (atomic_dec_and_test(&r10_bio->remaining)) {
 		md_done_sync(mddev, r10_bio->sectors, 1);
@@ -1392,34 +2174,189 @@ done:
  * The second for writing.
  *
  */
+static void fix_recovery_read_error(struct r10bio *r10_bio)
+{
+	/* We got a read error during recovery.
+	 * We repeat the read in smaller page-sized sections.
+	 * If a read succeeds, write it to the new device or record
+	 * a bad block if we cannot.
+	 * If a read fails, record a bad block on both old and
+	 * new devices.
+	 */
+	struct mddev *mddev = r10_bio->mddev;
+	struct r10conf *conf = mddev->private;
+	struct bio *bio = r10_bio->devs[0].bio;
+	sector_t sect = 0;
+	int sectors = r10_bio->sectors;
+	int idx = 0;
+	int dr = r10_bio->devs[0].devnum;
+	int dw = r10_bio->devs[1].devnum;
+
+	while (sectors) {
+		int s = sectors;
+		struct md_rdev *rdev;
+		sector_t addr;
+		int ok;
+
+		if (s > (PAGE_SIZE>>9))
+			s = PAGE_SIZE >> 9;
+
+		rdev = conf->mirrors[dr].rdev;
+		addr = r10_bio->devs[0].addr + sect,
+		ok = sync_page_io(rdev,
+				  addr,
+				  s << 9,
+				  bio->bi_io_vec[idx].bv_page,
+				  READ, false);
+		if (ok) {
+			rdev = conf->mirrors[dw].rdev;
+			addr = r10_bio->devs[1].addr + sect;
+			ok = sync_page_io(rdev,
+					  addr,
+					  s << 9,
+					  bio->bi_io_vec[idx].bv_page,
+					  WRITE, false);
+			if (!ok) {
+				set_bit(WriteErrorSeen, &rdev->flags);
+				if (!test_and_set_bit(WantReplacement,
+						      &rdev->flags))
+					set_bit(MD_RECOVERY_NEEDED,
+						&rdev->mddev->recovery);
+			}
+		}
+		if (!ok) {
+			/* We don't worry if we cannot set a bad block -
+			 * it really is bad so there is no loss in not
+			 * recording it yet
+			 */
+			rdev_set_badblocks(rdev, addr, s, 0);
+
+			if (rdev != conf->mirrors[dw].rdev) {
+				/* need bad block on destination too */
+				struct md_rdev *rdev2 = conf->mirrors[dw].rdev;
+				addr = r10_bio->devs[1].addr + sect;
+				ok = rdev_set_badblocks(rdev2, addr, s, 0);
+				if (!ok) {
+					/* just abort the recovery */
+					printk(KERN_NOTICE
+					       "md/raid10:%s: recovery aborted"
+					       " due to read error\n",
+					       mdname(mddev));
+
+					conf->mirrors[dw].recovery_disabled
+						= mddev->recovery_disabled;
+					set_bit(MD_RECOVERY_INTR,
+						&mddev->recovery);
+					break;
+				}
+			}
+		}
 
-static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio)
+		sectors -= s;
+		sect += s;
+		idx++;
+	}
+}
+
+static void recovery_request_write(struct mddev *mddev, struct r10bio *r10_bio)
 {
-	conf_t *conf = mddev_to_conf(mddev);
-	int i, d;
-	struct bio *bio, *wbio;
+	struct r10conf *conf = mddev->private;
+	int d;
+	struct bio *wbio, *wbio2;
 
+	if (!test_bit(R10BIO_Uptodate, &r10_bio->state)) {
+		fix_recovery_read_error(r10_bio);
+		end_sync_request(r10_bio);
+		return;
+	}
 
-	/* move the pages across to the second bio
+	/*
+	 * share the pages with the first bio
 	 * and submit the write request
 	 */
-	bio = r10_bio->devs[0].bio;
+	d = r10_bio->devs[1].devnum;
 	wbio = r10_bio->devs[1].bio;
-	for (i=0; i < wbio->bi_vcnt; i++) {
-		struct page *p = bio->bi_io_vec[i].bv_page;
-		bio->bi_io_vec[i].bv_page = wbio->bi_io_vec[i].bv_page;
-		wbio->bi_io_vec[i].bv_page = p;
+	wbio2 = r10_bio->devs[1].repl_bio;
+	/* Need to test wbio2->bi_end_io before we call
+	 * generic_make_request as if the former is NULL,
+	 * the latter is free to free wbio2.
+	 */
+	if (wbio2 && !wbio2->bi_end_io)
+		wbio2 = NULL;
+	if (wbio->bi_end_io) {
+		atomic_inc(&conf->mirrors[d].rdev->nr_pending);
+		md_sync_acct(conf->mirrors[d].rdev->bdev, bio_sectors(wbio));
+		generic_make_request(wbio);
 	}
-	d = r10_bio->devs[1].devnum;
+	if (wbio2) {
+		atomic_inc(&conf->mirrors[d].replacement->nr_pending);
+		md_sync_acct(conf->mirrors[d].replacement->bdev,
+			     bio_sectors(wbio2));
+		generic_make_request(wbio2);
+	}
+}
 
-	atomic_inc(&conf->mirrors[d].rdev->nr_pending);
-	md_sync_acct(conf->mirrors[d].rdev->bdev, wbio->bi_size >> 9);
-	if (test_bit(R10BIO_Uptodate, &r10_bio->state))
-		generic_make_request(wbio);
+
+/*
+ * Used by fix_read_error() to decay the per rdev read_errors.
+ * We halve the read error count for every hour that has elapsed
+ * since the last recorded read error.
+ *
+ */
+static void check_decay_read_errors(struct mddev *mddev, struct md_rdev *rdev)
+{
+	struct timespec cur_time_mon;
+	unsigned long hours_since_last;
+	unsigned int read_errors = atomic_read(&rdev->read_errors);
+
+	ktime_get_ts(&cur_time_mon);
+
+	if (rdev->last_read_error.tv_sec == 0 &&
+	    rdev->last_read_error.tv_nsec == 0) {
+		/* first time we've seen a read error */
+		rdev->last_read_error = cur_time_mon;
+		return;
+	}
+
+	hours_since_last = (cur_time_mon.tv_sec -
+			    rdev->last_read_error.tv_sec) / 3600;
+
+	rdev->last_read_error = cur_time_mon;
+
+	/*
+	 * if hours_since_last is > the number of bits in read_errors
+	 * just set read errors to 0. We do this to avoid
+	 * overflowing the shift of read_errors by hours_since_last.
+	 */
+	if (hours_since_last >= 8 * sizeof(read_errors))
+		atomic_set(&rdev->read_errors, 0);
 	else
-		bio_endio(wbio, -EIO);
+		atomic_set(&rdev->read_errors, read_errors >> hours_since_last);
 }
 
+static int r10_sync_page_io(struct md_rdev *rdev, sector_t sector,
+			    int sectors, struct page *page, int rw)
+{
+	sector_t first_bad;
+	int bad_sectors;
+
+	if (is_badblock(rdev, sector, sectors, &first_bad, &bad_sectors)
+	    && (rw == READ || test_bit(WriteErrorSeen, &rdev->flags)))
+		return -1;
+	if (sync_page_io(rdev, sector, sectors << 9, page, rw, false))
+		/* success */
+		return 1;
+	if (rw == WRITE) {
+		set_bit(WriteErrorSeen, &rdev->flags);
+		if (!test_and_set_bit(WantReplacement, &rdev->flags))
+			set_bit(MD_RECOVERY_NEEDED,
+				&rdev->mddev->recovery);
+	}
+	/* need to record an error - either for the block or the device */
+	if (!rdev_set_badblocks(rdev, sector, sectors, 0))
+		md_error(rdev->mddev, rdev);
+	return 0;
+}
 
 /*
  * This is a kernel thread which:
@@ -1429,11 +2366,43 @@ static void recovery_request_write(mddev_t *mddev, r10bio_t *r10_bio)
  *	3.	Performs writes following reads for array synchronising.
  */
 
-static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
+static void fix_read_error(struct r10conf *conf, struct mddev *mddev, struct r10bio *r10_bio)
 {
 	int sect = 0; /* Offset from r10_bio->sector */
 	int sectors = r10_bio->sectors;
-	mdk_rdev_t*rdev;
+	struct md_rdev*rdev;
+	int max_read_errors = atomic_read(&mddev->max_corr_read_errors);
+	int d = r10_bio->devs[r10_bio->read_slot].devnum;
+
+	/* still own a reference to this rdev, so it cannot
+	 * have been cleared recently.
+	 */
+	rdev = conf->mirrors[d].rdev;
+
+	if (test_bit(Faulty, &rdev->flags))
+		/* drive has already been failed, just ignore any
+		   more fix_read_error() attempts */
+		return;
+
+	check_decay_read_errors(mddev, rdev);
+	atomic_inc(&rdev->read_errors);
+	if (atomic_read(&rdev->read_errors) > max_read_errors) {
+		char b[BDEVNAME_SIZE];
+		bdevname(rdev->bdev, b);
+
+		printk(KERN_NOTICE
+		       "md/raid10:%s: %s: Raid device exceeded "
+		       "read_error threshold [cur %d:max %d]\n",
+		       mdname(mddev), b,
+		       atomic_read(&rdev->read_errors), max_read_errors);
+		printk(KERN_NOTICE
+		       "md/raid10:%s: %s: Failing raid device\n",
+		       mdname(mddev), b);
+		md_error(mddev, conf->mirrors[d].rdev);
+		r10_bio->devs[r10_bio->read_slot].bio = IO_BLOCKED;
+		return;
+	}
+
 	while(sectors) {
 		int s = sectors;
 		int sl = r10_bio->read_slot;
@@ -1445,17 +2414,23 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 
 		rcu_read_lock();
 		do {
-			int d = r10_bio->devs[sl].devnum;
+			sector_t first_bad;
+			int bad_sectors;
+
+			d = r10_bio->devs[sl].devnum;
 			rdev = rcu_dereference(conf->mirrors[d].rdev);
 			if (rdev &&
-			    test_bit(In_sync, &rdev->flags)) {
+			    !test_bit(Unmerged, &rdev->flags) &&
+			    test_bit(In_sync, &rdev->flags) &&
+			    is_badblock(rdev, r10_bio->devs[sl].addr + sect, s,
+					&first_bad, &bad_sectors) == 0) {
 				atomic_inc(&rdev->nr_pending);
 				rcu_read_unlock();
-				success = sync_page_io(rdev->bdev,
+				success = sync_page_io(rdev,
 						       r10_bio->devs[sl].addr +
-						       sect + rdev->data_offset,
+						       sect,
 						       s<<9,
-						       conf->tmppage, READ);
+						       conf->tmppage, READ, false);
 				rdev_dec_pending(rdev, mddev);
 				rcu_read_lock();
 				if (success)
@@ -1468,9 +2443,22 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 		rcu_read_unlock();
 
 		if (!success) {
-			/* Cannot read from anywhere -- bye bye array */
+			/* Cannot read from anywhere, just mark the block
+			 * as bad on the first device to discourage future
+			 * reads.
+			 */
 			int dn = r10_bio->devs[r10_bio->read_slot].devnum;
-			md_error(mddev, conf->mirrors[dn].rdev);
+			rdev = conf->mirrors[dn].rdev;
+
+			if (!rdev_set_badblocks(
+				    rdev,
+				    r10_bio->devs[r10_bio->read_slot].addr
+				    + sect,
+				    s, 0)) {
+				md_error(mddev, rdev);
+				r10_bio->devs[r10_bio->read_slot].bio
+					= IO_BLOCKED;
+			}
 			break;
 		}
 
@@ -1478,59 +2466,94 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 		/* write it back and re-read */
 		rcu_read_lock();
 		while (sl != r10_bio->read_slot) {
-			int d;
+			char b[BDEVNAME_SIZE];
+
 			if (sl==0)
 				sl = conf->copies;
 			sl--;
 			d = r10_bio->devs[sl].devnum;
 			rdev = rcu_dereference(conf->mirrors[d].rdev);
-			if (rdev &&
-			    test_bit(In_sync, &rdev->flags)) {
-				atomic_inc(&rdev->nr_pending);
-				rcu_read_unlock();
-				atomic_add(s, &rdev->corrected_errors);
-				if (sync_page_io(rdev->bdev,
-						 r10_bio->devs[sl].addr +
-						 sect + rdev->data_offset,
-						 s<<9, conf->tmppage, WRITE)
-				    == 0)
-					/* Well, this device is dead */
-					md_error(mddev, rdev);
-				rdev_dec_pending(rdev, mddev);
-				rcu_read_lock();
+			if (!rdev ||
+			    test_bit(Unmerged, &rdev->flags) ||
+			    !test_bit(In_sync, &rdev->flags))
+				continue;
+
+			atomic_inc(&rdev->nr_pending);
+			rcu_read_unlock();
+			if (r10_sync_page_io(rdev,
+					     r10_bio->devs[sl].addr +
+					     sect,
+					     s, conf->tmppage, WRITE)
+			    == 0) {
+				/* Well, this device is dead */
+				printk(KERN_NOTICE
+				       "md/raid10:%s: read correction "
+				       "write failed"
+				       " (%d sectors at %llu on %s)\n",
+				       mdname(mddev), s,
+				       (unsigned long long)(
+					       sect +
+					       choose_data_offset(r10_bio,
+								  rdev)),
+				       bdevname(rdev->bdev, b));
+				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
+				       "drive\n",
+				       mdname(mddev),
+				       bdevname(rdev->bdev, b));
 			}
+			rdev_dec_pending(rdev, mddev);
+			rcu_read_lock();
 		}
 		sl = start;
 		while (sl != r10_bio->read_slot) {
-			int d;
+			char b[BDEVNAME_SIZE];
+
 			if (sl==0)
 				sl = conf->copies;
 			sl--;
 			d = r10_bio->devs[sl].devnum;
 			rdev = rcu_dereference(conf->mirrors[d].rdev);
-			if (rdev &&
-			    test_bit(In_sync, &rdev->flags)) {
-				char b[BDEVNAME_SIZE];
-				atomic_inc(&rdev->nr_pending);
-				rcu_read_unlock();
-				if (sync_page_io(rdev->bdev,
-						 r10_bio->devs[sl].addr +
-						 sect + rdev->data_offset,
-						 s<<9, conf->tmppage, READ) == 0)
-					/* Well, this device is dead */
-					md_error(mddev, rdev);
-				else
-					printk(KERN_INFO
-					       "raid10:%s: read error corrected"
-					       " (%d sectors at %llu on %s)\n",
-					       mdname(mddev), s,
-					       (unsigned long long)(sect+
-					            rdev->data_offset),
-					       bdevname(rdev->bdev, b));
+			if (!rdev ||
+			    !test_bit(In_sync, &rdev->flags))
+				continue;
 
-				rdev_dec_pending(rdev, mddev);
-				rcu_read_lock();
+			atomic_inc(&rdev->nr_pending);
+			rcu_read_unlock();
+			switch (r10_sync_page_io(rdev,
+					     r10_bio->devs[sl].addr +
+					     sect,
+					     s, conf->tmppage,
+						 READ)) {
+			case 0:
+				/* Well, this device is dead */
+				printk(KERN_NOTICE
+				       "md/raid10:%s: unable to read back "
+				       "corrected sectors"
+				       " (%d sectors at %llu on %s)\n",
+				       mdname(mddev), s,
+				       (unsigned long long)(
+					       sect +
+					       choose_data_offset(r10_bio, rdev)),
+				       bdevname(rdev->bdev, b));
+				printk(KERN_NOTICE "md/raid10:%s: %s: failing "
+				       "drive\n",
+				       mdname(mddev),
+				       bdevname(rdev->bdev, b));
+				break;
+			case 1:
+				printk(KERN_INFO
+				       "md/raid10:%s: read error corrected"
+				       " (%d sectors at %llu on %s)\n",
+				       mdname(mddev), s,
+				       (unsigned long long)(
+					       sect +
+					       choose_data_offset(r10_bio, rdev)),
+				       bdevname(rdev->bdev, b));
+				atomic_add(s, &rdev->corrected_errors);
 			}
+
+			rdev_dec_pending(rdev, mddev);
+			rcu_read_lock();
 		}
 		rcu_read_unlock();
 
@@ -1539,101 +2562,309 @@ static void fix_read_error(conf_t *conf, mddev_t *mddev, r10bio_t *r10_bio)
 	}
 }
 
-static void raid10d(mddev_t *mddev)
+static int narrow_write_error(struct r10bio *r10_bio, int i)
+{
+	struct bio *bio = r10_bio->master_bio;
+	struct mddev *mddev = r10_bio->mddev;
+	struct r10conf *conf = mddev->private;
+	struct md_rdev *rdev = conf->mirrors[r10_bio->devs[i].devnum].rdev;
+	/* bio has the data to be written to slot 'i' where
+	 * we just recently had a write error.
+	 * We repeatedly clone the bio and trim down to one block,
+	 * then try the write.  Where the write fails we record
+	 * a bad block.
+	 * It is conceivable that the bio doesn't exactly align with
+	 * blocks.  We must handle this.
+	 *
+	 * We currently own a reference to the rdev.
+	 */
+
+	int block_sectors;
+	sector_t sector;
+	int sectors;
+	int sect_to_write = r10_bio->sectors;
+	int ok = 1;
+
+	if (rdev->badblocks.shift < 0)
+		return 0;
+
+	block_sectors = 1 << rdev->badblocks.shift;
+	sector = r10_bio->sector;
+	sectors = ((r10_bio->sector + block_sectors)
+		   & ~(sector_t)(block_sectors - 1))
+		- sector;
+
+	while (sect_to_write) {
+		struct bio *wbio;
+		if (sectors > sect_to_write)
+			sectors = sect_to_write;
+		/* Write at 'sector' for 'sectors' */
+		wbio = bio_clone_mddev(bio, GFP_NOIO, mddev);
+		bio_trim(wbio, sector - bio->bi_iter.bi_sector, sectors);
+		wbio->bi_iter.bi_sector = (r10_bio->devs[i].addr+
+				   choose_data_offset(r10_bio, rdev) +
+				   (sector - r10_bio->sector));
+		wbio->bi_bdev = rdev->bdev;
+		if (submit_bio_wait(WRITE, wbio) == 0)
+			/* Failure! */
+			ok = rdev_set_badblocks(rdev, sector,
+						sectors, 0)
+				&& ok;
+
+		bio_put(wbio);
+		sect_to_write -= sectors;
+		sector += sectors;
+		sectors = block_sectors;
+	}
+	return ok;
+}
+
+static void handle_read_error(struct mddev *mddev, struct r10bio *r10_bio)
 {
-	r10bio_t *r10_bio;
+	int slot = r10_bio->read_slot;
 	struct bio *bio;
+	struct r10conf *conf = mddev->private;
+	struct md_rdev *rdev = r10_bio->devs[slot].rdev;
+	char b[BDEVNAME_SIZE];
+	unsigned long do_sync;
+	int max_sectors;
+
+	/* we got a read error. Maybe the drive is bad.  Maybe just
+	 * the block and we can fix it.
+	 * We freeze all other IO, and try reading the block from
+	 * other devices.  When we find one, we re-write
+	 * and check it that fixes the read error.
+	 * This is all done synchronously while the array is
+	 * frozen.
+	 */
+	bio = r10_bio->devs[slot].bio;
+	bdevname(bio->bi_bdev, b);
+	bio_put(bio);
+	r10_bio->devs[slot].bio = NULL;
+
+	if (mddev->ro == 0) {
+		freeze_array(conf, 1);
+		fix_read_error(conf, mddev, r10_bio);
+		unfreeze_array(conf);
+	} else
+		r10_bio->devs[slot].bio = IO_BLOCKED;
+
+	rdev_dec_pending(rdev, mddev);
+
+read_more:
+	rdev = read_balance(conf, r10_bio, &max_sectors);
+	if (rdev == NULL) {
+		printk(KERN_ALERT "md/raid10:%s: %s: unrecoverable I/O"
+		       " read error for block %llu\n",
+		       mdname(mddev), b,
+		       (unsigned long long)r10_bio->sector);
+		raid_end_bio_io(r10_bio);
+		return;
+	}
+
+	do_sync = (r10_bio->master_bio->bi_rw & REQ_SYNC);
+	slot = r10_bio->read_slot;
+	printk_ratelimited(
+		KERN_ERR
+		"md/raid10:%s: %s: redirecting "
+		"sector %llu to another mirror\n",
+		mdname(mddev),
+		bdevname(rdev->bdev, b),
+		(unsigned long long)r10_bio->sector);
+	bio = bio_clone_mddev(r10_bio->master_bio,
+			      GFP_NOIO, mddev);
+	bio_trim(bio, r10_bio->sector - bio->bi_iter.bi_sector, max_sectors);
+	r10_bio->devs[slot].bio = bio;
+	r10_bio->devs[slot].rdev = rdev;
+	bio->bi_iter.bi_sector = r10_bio->devs[slot].addr
+		+ choose_data_offset(r10_bio, rdev);
+	bio->bi_bdev = rdev->bdev;
+	bio->bi_rw = READ | do_sync;
+	bio->bi_private = r10_bio;
+	bio->bi_end_io = raid10_end_read_request;
+	if (max_sectors < r10_bio->sectors) {
+		/* Drat - have to split this up more */
+		struct bio *mbio = r10_bio->master_bio;
+		int sectors_handled =
+			r10_bio->sector + max_sectors
+			- mbio->bi_iter.bi_sector;
+		r10_bio->sectors = max_sectors;
+		spin_lock_irq(&conf->device_lock);
+		if (mbio->bi_phys_segments == 0)
+			mbio->bi_phys_segments = 2;
+		else
+			mbio->bi_phys_segments++;
+		spin_unlock_irq(&conf->device_lock);
+		generic_make_request(bio);
+
+		r10_bio = mempool_alloc(conf->r10bio_pool,
+					GFP_NOIO);
+		r10_bio->master_bio = mbio;
+		r10_bio->sectors = bio_sectors(mbio) - sectors_handled;
+		r10_bio->state = 0;
+		set_bit(R10BIO_ReadError,
+			&r10_bio->state);
+		r10_bio->mddev = mddev;
+		r10_bio->sector = mbio->bi_iter.bi_sector
+			+ sectors_handled;
+
+		goto read_more;
+	} else
+		generic_make_request(bio);
+}
+
+static void handle_write_completed(struct r10conf *conf, struct r10bio *r10_bio)
+{
+	/* Some sort of write request has finished and it
+	 * succeeded in writing where we thought there was a
+	 * bad block.  So forget the bad block.
+	 * Or possibly if failed and we need to record
+	 * a bad block.
+	 */
+	int m;
+	struct md_rdev *rdev;
+
+	if (test_bit(R10BIO_IsSync, &r10_bio->state) ||
+	    test_bit(R10BIO_IsRecover, &r10_bio->state)) {
+		for (m = 0; m < conf->copies; m++) {
+			int dev = r10_bio->devs[m].devnum;
+			rdev = conf->mirrors[dev].rdev;
+			if (r10_bio->devs[m].bio == NULL)
+				continue;
+			if (test_bit(BIO_UPTODATE,
+				     &r10_bio->devs[m].bio->bi_flags)) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors, 0);
+			} else {
+				if (!rdev_set_badblocks(
+					    rdev,
+					    r10_bio->devs[m].addr,
+					    r10_bio->sectors, 0))
+					md_error(conf->mddev, rdev);
+			}
+			rdev = conf->mirrors[dev].replacement;
+			if (r10_bio->devs[m].repl_bio == NULL)
+				continue;
+			if (test_bit(BIO_UPTODATE,
+				     &r10_bio->devs[m].repl_bio->bi_flags)) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors, 0);
+			} else {
+				if (!rdev_set_badblocks(
+					    rdev,
+					    r10_bio->devs[m].addr,
+					    r10_bio->sectors, 0))
+					md_error(conf->mddev, rdev);
+			}
+		}
+		put_buf(r10_bio);
+	} else {
+		for (m = 0; m < conf->copies; m++) {
+			int dev = r10_bio->devs[m].devnum;
+			struct bio *bio = r10_bio->devs[m].bio;
+			rdev = conf->mirrors[dev].rdev;
+			if (bio == IO_MADE_GOOD) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors, 0);
+				rdev_dec_pending(rdev, conf->mddev);
+			} else if (bio != NULL &&
+				   !test_bit(BIO_UPTODATE, &bio->bi_flags)) {
+				if (!narrow_write_error(r10_bio, m)) {
+					md_error(conf->mddev, rdev);
+					set_bit(R10BIO_Degraded,
+						&r10_bio->state);
+				}
+				rdev_dec_pending(rdev, conf->mddev);
+			}
+			bio = r10_bio->devs[m].repl_bio;
+			rdev = conf->mirrors[dev].replacement;
+			if (rdev && bio == IO_MADE_GOOD) {
+				rdev_clear_badblocks(
+					rdev,
+					r10_bio->devs[m].addr,
+					r10_bio->sectors, 0);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
+		}
+		if (test_bit(R10BIO_WriteError,
+			     &r10_bio->state))
+			close_write(r10_bio);
+		raid_end_bio_io(r10_bio);
+	}
+}
+
+static void raid10d(struct md_thread *thread)
+{
+	struct mddev *mddev = thread->mddev;
+	struct r10bio *r10_bio;
 	unsigned long flags;
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
 	struct list_head *head = &conf->retry_list;
-	int unplug=0;
-	mdk_rdev_t *rdev;
+	struct blk_plug plug;
 
 	md_check_recovery(mddev);
 
+	blk_start_plug(&plug);
 	for (;;) {
-		char b[BDEVNAME_SIZE];
 
-		unplug += flush_pending_writes(conf);
+		flush_pending_writes(conf);
 
 		spin_lock_irqsave(&conf->device_lock, flags);
 		if (list_empty(head)) {
 			spin_unlock_irqrestore(&conf->device_lock, flags);
 			break;
 		}
-		r10_bio = list_entry(head->prev, r10bio_t, retry_list);
+		r10_bio = list_entry(head->prev, struct r10bio, retry_list);
 		list_del(head->prev);
 		conf->nr_queued--;
 		spin_unlock_irqrestore(&conf->device_lock, flags);
 
 		mddev = r10_bio->mddev;
-		conf = mddev_to_conf(mddev);
-		if (test_bit(R10BIO_IsSync, &r10_bio->state)) {
+		conf = mddev->private;
+		if (test_bit(R10BIO_MadeGood, &r10_bio->state) ||
+		    test_bit(R10BIO_WriteError, &r10_bio->state))
+			handle_write_completed(conf, r10_bio);
+		else if (test_bit(R10BIO_IsReshape, &r10_bio->state))
+			reshape_request_write(mddev, r10_bio);
+		else if (test_bit(R10BIO_IsSync, &r10_bio->state))
 			sync_request_write(mddev, r10_bio);
-			unplug = 1;
-		} else 	if (test_bit(R10BIO_IsRecover, &r10_bio->state)) {
+		else if (test_bit(R10BIO_IsRecover, &r10_bio->state))
 			recovery_request_write(mddev, r10_bio);
-			unplug = 1;
-		} else {
-			int mirror;
-			/* we got a read error. Maybe the drive is bad.  Maybe just
-			 * the block and we can fix it.
-			 * We freeze all other IO, and try reading the block from
-			 * other devices.  When we find one, we re-write
-			 * and check it that fixes the read error.
-			 * This is all done synchronously while the array is
-			 * frozen.
+		else if (test_bit(R10BIO_ReadError, &r10_bio->state))
+			handle_read_error(mddev, r10_bio);
+		else {
+			/* just a partial read to be scheduled from a
+			 * separate context
 			 */
-			if (mddev->ro == 0) {
-				freeze_array(conf);
-				fix_read_error(conf, mddev, r10_bio);
-				unfreeze_array(conf);
-			}
-
-			bio = r10_bio->devs[r10_bio->read_slot].bio;
-			r10_bio->devs[r10_bio->read_slot].bio =
-				mddev->ro ? IO_BLOCKED : NULL;
-			mirror = read_balance(conf, r10_bio);
-			if (mirror == -1) {
-				printk(KERN_ALERT "raid10: %s: unrecoverable I/O"
-				       " read error for block %llu\n",
-				       bdevname(bio->bi_bdev,b),
-				       (unsigned long long)r10_bio->sector);
-				raid_end_bio_io(r10_bio);
-				bio_put(bio);
-			} else {
-				const int do_sync = bio_sync(r10_bio->master_bio);
-				bio_put(bio);
-				rdev = conf->mirrors[mirror].rdev;
-				if (printk_ratelimit())
-					printk(KERN_ERR "raid10: %s: redirecting sector %llu to"
-					       " another mirror\n",
-					       bdevname(rdev->bdev,b),
-					       (unsigned long long)r10_bio->sector);
-				bio = bio_clone(r10_bio->master_bio, GFP_NOIO);
-				r10_bio->devs[r10_bio->read_slot].bio = bio;
-				bio->bi_sector = r10_bio->devs[r10_bio->read_slot].addr
-					+ rdev->data_offset;
-				bio->bi_bdev = rdev->bdev;
-				bio->bi_rw = READ | do_sync;
-				bio->bi_private = r10_bio;
-				bio->bi_end_io = raid10_end_read_request;
-				unplug = 1;
-				generic_make_request(bio);
-			}
+			int slot = r10_bio->read_slot;
+			generic_make_request(r10_bio->devs[slot].bio);
 		}
+
+		cond_resched();
+		if (mddev->flags & ~(1<<MD_CHANGE_PENDING))
+			md_check_recovery(mddev);
 	}
-	if (unplug)
-		unplug_slaves(mddev);
+	blk_finish_plug(&plug);
 }
 
 
-static int init_resync(conf_t *conf)
+static int init_resync(struct r10conf *conf)
 {
 	int buffs;
+	int i;
 
 	buffs = RESYNC_WINDOW / RESYNC_BLOCK_SIZE;
 	BUG_ON(conf->r10buf_pool);
+	conf->have_replacement = 0;
+	for (i = 0; i < conf->geo.raid_disks; i++)
+		if (conf->mirrors[i].replacement)
+			conf->have_replacement = 1;
 	conf->r10buf_pool = mempool_create(buffs, r10buf_pool_alloc, r10buf_pool_free, conf);
 	if (!conf->r10buf_pool)
 		return -ENOMEM;
@@ -1673,27 +2904,43 @@ static int init_resync(conf_t *conf)
  *
  */
 
-static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
+static sector_t sync_request(struct mddev *mddev, sector_t sector_nr,
+			     int *skipped, int go_faster)
 {
-	conf_t *conf = mddev_to_conf(mddev);
-	r10bio_t *r10_bio;
+	struct r10conf *conf = mddev->private;
+	struct r10bio *r10_bio;
 	struct bio *biolist = NULL, *bio;
 	sector_t max_sector, nr_sectors;
-	int disk;
 	int i;
 	int max_sync;
-	int sync_blocks;
-
+	sector_t sync_blocks;
 	sector_t sectors_skipped = 0;
 	int chunks_skipped = 0;
+	sector_t chunk_mask = conf->geo.chunk_mask;
 
 	if (!conf->r10buf_pool)
 		if (init_resync(conf))
 			return 0;
 
+	/*
+	 * Allow skipping a full rebuild for incremental assembly
+	 * of a clean array, like RAID1 does.
+	 */
+	if (mddev->bitmap == NULL &&
+	    mddev->recovery_cp == MaxSector &&
+	    mddev->reshape_position == MaxSector &&
+	    !test_bit(MD_RECOVERY_SYNC, &mddev->recovery) &&
+	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
+	    !test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery) &&
+	    conf->fullsync == 0) {
+		*skipped = 1;
+		return mddev->dev_sectors - sector_nr;
+	}
+
  skipped:
-	max_sector = mddev->size << 1;
-	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
+	max_sector = mddev->dev_sectors;
+	if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery) ||
+	    test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 		max_sector = mddev->resync_max_sectors;
 	if (sector_nr >= max_sector) {
 		/* If we aborted, we need to abort the
@@ -1705,25 +2952,47 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		 * we need to convert that to several
 		 * virtual addresses.
 		 */
+		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
+			end_reshape(conf);
+			return 0;
+		}
+
 		if (mddev->curr_resync < max_sector) { /* aborted */
 			if (test_bit(MD_RECOVERY_SYNC, &mddev->recovery))
 				bitmap_end_sync(mddev->bitmap, mddev->curr_resync,
 						&sync_blocks, 1);
-			else for (i=0; i<conf->raid_disks; i++) {
+			else for (i = 0; i < conf->geo.raid_disks; i++) {
 				sector_t sect =
 					raid10_find_virt(conf, mddev->curr_resync, i);
 				bitmap_end_sync(mddev->bitmap, sect,
 						&sync_blocks, 1);
 			}
-		} else /* completed sync */
+		} else {
+			/* completed sync */
+			if ((!mddev->bitmap || conf->fullsync)
+			    && conf->have_replacement
+			    && test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
+				/* Completed a full sync so the replacements
+				 * are now fully recovered.
+				 */
+				for (i = 0; i < conf->geo.raid_disks; i++)
+					if (conf->mirrors[i].replacement)
+						conf->mirrors[i].replacement
+							->recovery_offset
+							= MaxSector;
+			}
 			conf->fullsync = 0;
-
+		}
 		bitmap_close_sync(mddev->bitmap);
 		close_sync(conf);
 		*skipped = 1;
 		return sectors_skipped;
 	}
-	if (chunks_skipped >= conf->raid_disks) {
+
+	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
+		return reshape_request(mddev, sector_nr, skipped);
+
+	if (chunks_skipped >= conf->geo.raid_disks) {
 		/* if there has been nothing to do on any drive,
 		 * then there is nothing to do at all..
 		 */
@@ -1737,9 +3006,9 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	/* make sure whole request will fit in a chunk - if chunks
 	 * are meaningful
 	 */
-	if (conf->near_copies < conf->raid_disks &&
-	    max_sector > (sector_nr | conf->chunk_mask))
-		max_sector = (sector_nr | conf->chunk_mask) + 1;
+	if (conf->geo.near_copies < conf->geo.raid_disks &&
+	    max_sector > (sector_nr | chunk_mask))
+		max_sector = (sector_nr | chunk_mask) + 1;
 	/*
 	 * If there is non-resync activity waiting for us then
 	 * put in a delay to throttle resync.
@@ -1747,8 +3016,6 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	if (!go_faster && conf->nr_waiting)
 		msleep_interruptible(1000);
 
-	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
-
 	/* Again, very different code for resync and recovery.
 	 * Both must result in an r10bio with a list of bios that
 	 * have bi_end_io, bi_sector, bi_bdev set,
@@ -1767,114 +3034,220 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	max_sync = RESYNC_PAGES << (PAGE_SHIFT-9);
 	if (!test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
 		/* recovery... the complicated one */
-		int i, j, k;
+		int j;
 		r10_bio = NULL;
 
-		for (i=0 ; i<conf->raid_disks; i++)
-			if (conf->mirrors[i].rdev &&
-			    !test_bit(In_sync, &conf->mirrors[i].rdev->flags)) {
-				int still_degraded = 0;
-				/* want to reconstruct this device */
-				r10bio_t *rb2 = r10_bio;
-				sector_t sect = raid10_find_virt(conf, sector_nr, i);
-				int must_sync;
-				/* Unless we are doing a full sync, we only need
-				 * to recover the block if it is set in the bitmap
+		for (i = 0 ; i < conf->geo.raid_disks; i++) {
+			int still_degraded;
+			struct r10bio *rb2;
+			sector_t sect;
+			int must_sync;
+			int any_working;
+			struct raid10_info *mirror = &conf->mirrors[i];
+
+			if ((mirror->rdev == NULL ||
+			     test_bit(In_sync, &mirror->rdev->flags))
+			    &&
+			    (mirror->replacement == NULL ||
+			     test_bit(Faulty,
+				      &mirror->replacement->flags)))
+				continue;
+
+			still_degraded = 0;
+			/* 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.
 				 */
-				must_sync = bitmap_start_sync(mddev->bitmap, sect,
-							      &sync_blocks, 1);
-				if (sync_blocks < max_sync)
-					max_sync = sync_blocks;
-				if (!must_sync &&
-				    !conf->fullsync) {
-					/* yep, skip the sync_blocks here, but don't assume
-					 * that there will never be anything to do here
-					 */
-					chunks_skipped = -1;
-					continue;
-				}
+				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
+			 */
+			must_sync = bitmap_start_sync(mddev->bitmap, sect,
+						      &sync_blocks, 1);
+			if (sync_blocks < max_sync)
+				max_sync = sync_blocks;
+			if (!must_sync &&
+			    mirror->replacement == NULL &&
+			    !conf->fullsync) {
+				/* yep, skip the sync_blocks here, but don't assume
+				 * that there will never be anything to do here
+				 */
+				chunks_skipped = -1;
+				continue;
+			}
 
-				r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
-				raise_barrier(conf, rb2 != NULL);
-				atomic_set(&r10_bio->remaining, 0);
+			r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
+			raise_barrier(conf, rb2 != NULL);
+			atomic_set(&r10_bio->remaining, 0);
 
-				r10_bio->master_bio = (struct bio*)rb2;
-				if (rb2)
-					atomic_inc(&rb2->remaining);
-				r10_bio->mddev = mddev;
-				set_bit(R10BIO_IsRecover, &r10_bio->state);
-				r10_bio->sector = sect;
-
-				raid10_find_phys(conf, r10_bio);
-				/* Need to check if this section will still be
-				 * degraded
-				 */
-				for (j=0; j<conf->copies;j++) {
-					int d = r10_bio->devs[j].devnum;
-					if (conf->mirrors[d].rdev == NULL ||
-					    test_bit(Faulty, &conf->mirrors[d].rdev->flags)) {
-						still_degraded = 1;
-						break;
+			r10_bio->master_bio = (struct bio*)rb2;
+			if (rb2)
+				atomic_inc(&rb2->remaining);
+			r10_bio->mddev = mddev;
+			set_bit(R10BIO_IsRecover, &r10_bio->state);
+			r10_bio->sector = sect;
+
+			raid10_find_phys(conf, r10_bio);
+
+			/* Need to check if the array will still be
+			 * degraded
+			 */
+			for (j = 0; j < conf->geo.raid_disks; j++)
+				if (conf->mirrors[j].rdev == NULL ||
+				    test_bit(Faulty, &conf->mirrors[j].rdev->flags)) {
+					still_degraded = 1;
+					break;
+				}
+
+			must_sync = bitmap_start_sync(mddev->bitmap, sect,
+						      &sync_blocks, still_degraded);
+
+			any_working = 0;
+			for (j=0; j<conf->copies;j++) {
+				int k;
+				int d = r10_bio->devs[j].devnum;
+				sector_t from_addr, to_addr;
+				struct md_rdev *rdev;
+				sector_t sector, first_bad;
+				int bad_sectors;
+				if (!conf->mirrors[d].rdev ||
+				    !test_bit(In_sync, &conf->mirrors[d].rdev->flags))
+					continue;
+				/* This is where we read from */
+				any_working = 1;
+				rdev = conf->mirrors[d].rdev;
+				sector = r10_bio->devs[j].addr;
+
+				if (is_badblock(rdev, sector, max_sync,
+						&first_bad, &bad_sectors)) {
+					if (first_bad > sector)
+						max_sync = first_bad - sector;
+					else {
+						bad_sectors -= (sector
+								- first_bad);
+						if (max_sync > bad_sectors)
+							max_sync = bad_sectors;
+						continue;
 					}
 				}
-				must_sync = bitmap_start_sync(mddev->bitmap, sect,
-							      &sync_blocks, still_degraded);
-
-				for (j=0; j<conf->copies;j++) {
-					int d = r10_bio->devs[j].devnum;
-					if (conf->mirrors[d].rdev &&
-					    test_bit(In_sync, &conf->mirrors[d].rdev->flags)) {
-						/* This is where we read from */
-						bio = r10_bio->devs[0].bio;
-						bio->bi_next = biolist;
-						biolist = bio;
-						bio->bi_private = r10_bio;
-						bio->bi_end_io = end_sync_read;
-						bio->bi_rw = READ;
-						bio->bi_sector = r10_bio->devs[j].addr +
-							conf->mirrors[d].rdev->data_offset;
-						bio->bi_bdev = conf->mirrors[d].rdev->bdev;
-						atomic_inc(&conf->mirrors[d].rdev->nr_pending);
-						atomic_inc(&r10_bio->remaining);
-						/* and we write to 'i' */
-
-						for (k=0; k<conf->copies; k++)
-							if (r10_bio->devs[k].devnum == i)
-								break;
-						BUG_ON(k == conf->copies);
-						bio = r10_bio->devs[1].bio;
-						bio->bi_next = biolist;
-						biolist = bio;
-						bio->bi_private = r10_bio;
-						bio->bi_end_io = end_sync_write;
-						bio->bi_rw = WRITE;
-						bio->bi_sector = r10_bio->devs[k].addr +
-							conf->mirrors[i].rdev->data_offset;
-						bio->bi_bdev = conf->mirrors[i].rdev->bdev;
-
-						r10_bio->devs[0].devnum = d;
-						r10_bio->devs[1].devnum = i;
+				bio = r10_bio->devs[0].bio;
+				bio_reset(bio);
+				bio->bi_next = biolist;
+				biolist = bio;
+				bio->bi_private = r10_bio;
+				bio->bi_end_io = end_sync_read;
+				bio->bi_rw = READ;
+				from_addr = r10_bio->devs[j].addr;
+				bio->bi_iter.bi_sector = from_addr +
+					rdev->data_offset;
+				bio->bi_bdev = rdev->bdev;
+				atomic_inc(&rdev->nr_pending);
+				/* and we write to 'i' (if not in_sync) */
 
+				for (k=0; k<conf->copies; k++)
+					if (r10_bio->devs[k].devnum == i)
 						break;
-					}
+				BUG_ON(k == conf->copies);
+				to_addr = r10_bio->devs[k].addr;
+				r10_bio->devs[0].devnum = d;
+				r10_bio->devs[0].addr = from_addr;
+				r10_bio->devs[1].devnum = i;
+				r10_bio->devs[1].addr = to_addr;
+
+				rdev = mirror->rdev;
+				if (!test_bit(In_sync, &rdev->flags)) {
+					bio = r10_bio->devs[1].bio;
+					bio_reset(bio);
+					bio->bi_next = biolist;
+					biolist = bio;
+					bio->bi_private = r10_bio;
+					bio->bi_end_io = end_sync_write;
+					bio->bi_rw = WRITE;
+					bio->bi_iter.bi_sector = to_addr
+						+ rdev->data_offset;
+					bio->bi_bdev = rdev->bdev;
+					atomic_inc(&r10_bio->remaining);
+				} else
+					r10_bio->devs[1].bio->bi_end_io = NULL;
+
+				/* and maybe write to replacement */
+				bio = r10_bio->devs[1].repl_bio;
+				if (bio)
+					bio->bi_end_io = NULL;
+				rdev = mirror->replacement;
+				/* Note: if rdev != NULL, then bio
+				 * cannot be NULL as r10buf_pool_alloc will
+				 * have allocated it.
+				 * So the second test here is pointless.
+				 * But it keeps semantic-checkers happy, and
+				 * this comment keeps human reviewers
+				 * happy.
+				 */
+				if (rdev == NULL || bio == NULL ||
+				    test_bit(Faulty, &rdev->flags))
+					break;
+				bio_reset(bio);
+				bio->bi_next = biolist;
+				biolist = bio;
+				bio->bi_private = r10_bio;
+				bio->bi_end_io = end_sync_write;
+				bio->bi_rw = WRITE;
+				bio->bi_iter.bi_sector = to_addr +
+					rdev->data_offset;
+				bio->bi_bdev = rdev->bdev;
+				atomic_inc(&r10_bio->remaining);
+				break;
+			}
+			if (j == conf->copies) {
+				/* Cannot recover, so abort the recovery or
+				 * record a bad block */
+				if (any_working) {
+					/* problem is that there are bad blocks
+					 * on other device(s)
+					 */
+					int k;
+					for (k = 0; k < conf->copies; k++)
+						if (r10_bio->devs[k].devnum == i)
+							break;
+					if (!test_bit(In_sync,
+						      &mirror->rdev->flags)
+					    && !rdev_set_badblocks(
+						    mirror->rdev,
+						    r10_bio->devs[k].addr,
+						    max_sync, 0))
+						any_working = 0;
+					if (mirror->replacement &&
+					    !rdev_set_badblocks(
+						    mirror->replacement,
+						    r10_bio->devs[k].addr,
+						    max_sync, 0))
+						any_working = 0;
 				}
-				if (j == conf->copies) {
-					/* Cannot recover, so abort the recovery */
-					put_buf(r10_bio);
-					if (rb2)
-						atomic_dec(&rb2->remaining);
-					r10_bio = rb2;
+				if (!any_working)  {
 					if (!test_and_set_bit(MD_RECOVERY_INTR,
 							      &mddev->recovery))
-						printk(KERN_INFO "raid10: %s: insufficient working devices for recovery.\n",
+						printk(KERN_INFO "md/raid10:%s: insufficient "
+						       "working devices for recovery.\n",
 						       mdname(mddev));
-					break;
+					mirror->recovery_disabled
+						= mddev->recovery_disabled;
 				}
+				put_buf(r10_bio);
+				if (rb2)
+					atomic_dec(&rb2->remaining);
+				r10_bio = rb2;
+				break;
 			}
+		}
 		if (biolist == NULL) {
 			while (r10_bio) {
-				r10bio_t *rb2 = r10_bio;
-				r10_bio = (r10bio_t*) rb2->master_bio;
+				struct r10bio *rb2 = r10_bio;
+				r10_bio = (struct r10bio*) rb2->master_bio;
 				rb2->master_bio = NULL;
 				put_buf(rb2);
 			}
@@ -1884,9 +3257,12 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		/* resync. Schedule a read for every block at this virt offset */
 		int count = 0;
 
+		bitmap_cond_end_sync(mddev->bitmap, sector_nr);
+
 		if (!bitmap_start_sync(mddev->bitmap, sector_nr,
 				       &sync_blocks, mddev->degraded) &&
-		    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery)) {
+		    !conf->fullsync && !test_bit(MD_RECOVERY_REQUESTED,
+						 &mddev->recovery)) {
 			/* We can skip this block */
 			*skipped = 1;
 			return sync_blocks + sectors_skipped;
@@ -1904,16 +3280,35 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		r10_bio->sector = sector_nr;
 		set_bit(R10BIO_IsSync, &r10_bio->state);
 		raid10_find_phys(conf, r10_bio);
-		r10_bio->sectors = (sector_nr | conf->chunk_mask) - sector_nr +1;
+		r10_bio->sectors = (sector_nr | chunk_mask) - sector_nr + 1;
 
-		for (i=0; i<conf->copies; i++) {
+		for (i = 0; i < conf->copies; i++) {
 			int d = r10_bio->devs[i].devnum;
+			sector_t first_bad, sector;
+			int bad_sectors;
+
+			if (r10_bio->devs[i].repl_bio)
+				r10_bio->devs[i].repl_bio->bi_end_io = NULL;
+
 			bio = r10_bio->devs[i].bio;
-			bio->bi_end_io = NULL;
+			bio_reset(bio);
 			clear_bit(BIO_UPTODATE, &bio->bi_flags);
 			if (conf->mirrors[d].rdev == NULL ||
 			    test_bit(Faulty, &conf->mirrors[d].rdev->flags))
 				continue;
+			sector = r10_bio->devs[i].addr;
+			if (is_badblock(conf->mirrors[d].rdev,
+					sector, max_sync,
+					&first_bad, &bad_sectors)) {
+				if (first_bad > sector)
+					max_sync = first_bad - sector;
+				else {
+					bad_sectors -= (sector - first_bad);
+					if (max_sync > bad_sectors)
+						max_sync = bad_sectors;
+					continue;
+				}
+			}
 			atomic_inc(&conf->mirrors[d].rdev->nr_pending);
 			atomic_inc(&r10_bio->remaining);
 			bio->bi_next = biolist;
@@ -1921,17 +3316,45 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 			bio->bi_private = r10_bio;
 			bio->bi_end_io = end_sync_read;
 			bio->bi_rw = READ;
-			bio->bi_sector = r10_bio->devs[i].addr +
+			bio->bi_iter.bi_sector = sector +
 				conf->mirrors[d].rdev->data_offset;
 			bio->bi_bdev = conf->mirrors[d].rdev->bdev;
 			count++;
+
+			if (conf->mirrors[d].replacement == NULL ||
+			    test_bit(Faulty,
+				     &conf->mirrors[d].replacement->flags))
+				continue;
+
+			/* Need to set up for writing to the replacement */
+			bio = r10_bio->devs[i].repl_bio;
+			bio_reset(bio);
+			clear_bit(BIO_UPTODATE, &bio->bi_flags);
+
+			sector = r10_bio->devs[i].addr;
+			atomic_inc(&conf->mirrors[d].rdev->nr_pending);
+			bio->bi_next = biolist;
+			biolist = bio;
+			bio->bi_private = r10_bio;
+			bio->bi_end_io = end_sync_write;
+			bio->bi_rw = WRITE;
+			bio->bi_iter.bi_sector = sector +
+				conf->mirrors[d].replacement->data_offset;
+			bio->bi_bdev = conf->mirrors[d].replacement->bdev;
+			count++;
 		}
 
 		if (count < 2) {
 			for (i=0; i<conf->copies; i++) {
 				int d = r10_bio->devs[i].devnum;
 				if (r10_bio->devs[i].bio->bi_end_io)
-					rdev_dec_pending(conf->mirrors[d].rdev, mddev);
+					rdev_dec_pending(conf->mirrors[d].rdev,
+							 mddev);
+				if (r10_bio->devs[i].repl_bio &&
+				    r10_bio->devs[i].repl_bio->bi_end_io)
+					rdev_dec_pending(
+						conf->mirrors[d].replacement,
+						mddev);
 			}
 			put_buf(r10_bio);
 			biolist = NULL;
@@ -1939,44 +3362,33 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 		}
 	}
 
-	for (bio = biolist; bio ; bio=bio->bi_next) {
-
-		bio->bi_flags &= ~(BIO_POOL_MASK - 1);
-		if (bio->bi_end_io)
-			bio->bi_flags |= 1 << BIO_UPTODATE;
-		bio->bi_vcnt = 0;
-		bio->bi_idx = 0;
-		bio->bi_phys_segments = 0;
-		bio->bi_hw_segments = 0;
-		bio->bi_size = 0;
-	}
-
 	nr_sectors = 0;
 	if (sector_nr + max_sync < max_sector)
 		max_sector = sector_nr + max_sync;
 	do {
 		struct page *page;
 		int len = PAGE_SIZE;
-		disk = 0;
 		if (sector_nr + (len>>9) > max_sector)
 			len = (max_sector - sector_nr) << 9;
 		if (len == 0)
 			break;
 		for (bio= biolist ; bio ; bio=bio->bi_next) {
+			struct bio *bio2;
 			page = bio->bi_io_vec[bio->bi_vcnt].bv_page;
-			if (bio_add_page(bio, page, len, 0) == 0) {
-				/* stop here */
-				struct bio *bio2;
-				bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
-				for (bio2 = biolist; bio2 && bio2 != bio; bio2 = bio2->bi_next) {
-					/* remove last page from this bio */
-					bio2->bi_vcnt--;
-					bio2->bi_size -= len;
-					bio2->bi_flags &= ~(1<< BIO_SEG_VALID);
-				}
-				goto bio_full;
+			if (bio_add_page(bio, page, len, 0))
+				continue;
+
+			/* stop here */
+			bio->bi_io_vec[bio->bi_vcnt].bv_page = page;
+			for (bio2 = biolist;
+			     bio2 && bio2 != bio;
+			     bio2 = bio2->bi_next) {
+				/* remove last page from this bio */
+				bio2->bi_vcnt--;
+				bio2->bi_iter.bi_size -= len;
+				bio2->bi_flags &= ~(1<< BIO_SEG_VALID);
 			}
-			disk = i;
+			goto bio_full;
 		}
 		nr_sectors += len>>9;
 		sector_nr += len>>9;
@@ -1994,6 +3406,7 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 
 		if (bio->bi_end_io == end_sync_read) {
 			md_sync_acct(bio->bi_bdev, nr_sectors);
+			set_bit(BIO_UPTODATE, &bio->bi_flags);
 			generic_make_request(bio);
 		}
 	}
@@ -2007,191 +3420,392 @@ static sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, i
 	return sectors_skipped + nr_sectors;
  giveup:
 	/* There is nowhere to write, so all non-sync
-	 * drives must be failed, so try the next chunk...
+	 * drives must be failed or in resync, all drives
+	 * have a bad block, so try the next chunk...
 	 */
-	{
-	sector_t sec = max_sector - sector_nr;
-	sectors_skipped += sec;
+	if (sector_nr + max_sync < max_sector)
+		max_sector = sector_nr + max_sync;
+
+	sectors_skipped += (max_sector - sector_nr);
 	chunks_skipped ++;
 	sector_nr = max_sector;
 	goto skipped;
+}
+
+static sector_t
+raid10_size(struct mddev *mddev, sector_t sectors, int raid_disks)
+{
+	sector_t size;
+	struct r10conf *conf = mddev->private;
+
+	if (!raid_disks)
+		raid_disks = min(conf->geo.raid_disks,
+				 conf->prev.raid_disks);
+	if (!sectors)
+		sectors = conf->dev_sectors;
+
+	size = sectors >> conf->geo.chunk_shift;
+	sector_div(size, conf->geo.far_copies);
+	size = size * raid_disks;
+	sector_div(size, conf->geo.near_copies);
+
+	return size << conf->geo.chunk_shift;
+}
+
+static void calc_sectors(struct r10conf *conf, sector_t size)
+{
+	/* Calculate the number of sectors-per-device that will
+	 * actually be used, and set conf->dev_sectors and
+	 * conf->stride
+	 */
+
+	size = size >> conf->geo.chunk_shift;
+	sector_div(size, conf->geo.far_copies);
+	size = size * conf->geo.raid_disks;
+	sector_div(size, conf->geo.near_copies);
+	/* 'size' is now the number of chunks in the array */
+	/* calculate "used chunks per device" */
+	size = size * conf->copies;
+
+	/* We need to round up when dividing by raid_disks to
+	 * get the stride size.
+	 */
+	size = DIV_ROUND_UP_SECTOR_T(size, conf->geo.raid_disks);
+
+	conf->dev_sectors = size << conf->geo.chunk_shift;
+
+	if (conf->geo.far_offset)
+		conf->geo.stride = 1 << conf->geo.chunk_shift;
+	else {
+		sector_div(size, conf->geo.far_copies);
+		conf->geo.stride = size << conf->geo.chunk_shift;
 	}
 }
 
-static int run(mddev_t *mddev)
+enum geo_type {geo_new, geo_old, geo_start};
+static int setup_geo(struct geom *geo, struct mddev *mddev, enum geo_type new)
 {
-	conf_t *conf;
-	int i, disk_idx;
-	mirror_info_t *disk;
-	mdk_rdev_t *rdev;
-	struct list_head *tmp;
 	int nc, fc, fo;
-	sector_t stride, size;
-
-	if (mddev->chunk_size == 0) {
-		printk(KERN_ERR "md/raid10: non-zero chunk size required.\n");
-		return -EINVAL;
+	int layout, chunk, disks;
+	switch (new) {
+	case geo_old:
+		layout = mddev->layout;
+		chunk = mddev->chunk_sectors;
+		disks = mddev->raid_disks - mddev->delta_disks;
+		break;
+	case geo_new:
+		layout = mddev->new_layout;
+		chunk = mddev->new_chunk_sectors;
+		disks = mddev->raid_disks;
+		break;
+	default: /* avoid 'may be unused' warnings */
+	case geo_start: /* new when starting reshape - raid_disks not
+			 * updated yet. */
+		layout = mddev->new_layout;
+		chunk = mddev->new_chunk_sectors;
+		disks = mddev->raid_disks + mddev->delta_disks;
+		break;
 	}
+	if (layout >> 18)
+		return -1;
+	if (chunk < (PAGE_SIZE >> 9) ||
+	    !is_power_of_2(chunk))
+		return -2;
+	nc = layout & 255;
+	fc = (layout >> 8) & 255;
+	fo = layout & (1<<16);
+	geo->raid_disks = disks;
+	geo->near_copies = nc;
+	geo->far_copies = fc;
+	geo->far_offset = fo;
+	geo->far_set_size = (layout & (1<<17)) ? disks / fc : disks;
+	geo->chunk_mask = chunk - 1;
+	geo->chunk_shift = ffz(~chunk);
+	return nc*fc;
+}
+
+static struct r10conf *setup_conf(struct mddev *mddev)
+{
+	struct r10conf *conf = NULL;
+	int err = -EINVAL;
+	struct geom geo;
+	int copies;
 
-	nc = mddev->layout & 255;
-	fc = (mddev->layout >> 8) & 255;
-	fo = mddev->layout & (1<<16);
-	if ((nc*fc) <2 || (nc*fc) > mddev->raid_disks ||
-	    (mddev->layout >> 17)) {
-		printk(KERN_ERR "raid10: %s: unsupported raid10 layout: 0x%8x\n",
-		       mdname(mddev), mddev->layout);
+	copies = setup_geo(&geo, mddev, geo_new);
+
+	if (copies == -2) {
+		printk(KERN_ERR "md/raid10:%s: chunk size must be "
+		       "at least PAGE_SIZE(%ld) and be a power of 2.\n",
+		       mdname(mddev), PAGE_SIZE);
 		goto out;
 	}
-	/*
-	 * copy the already verified devices into our private RAID10
-	 * bookkeeping area. [whatever we allocate in run(),
-	 * should be freed in stop()]
-	 */
-	conf = kzalloc(sizeof(conf_t), GFP_KERNEL);
-	mddev->private = conf;
-	if (!conf) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-			mdname(mddev));
+
+	if (copies < 2 || copies > mddev->raid_disks) {
+		printk(KERN_ERR "md/raid10:%s: unsupported raid10 layout: 0x%8x\n",
+		       mdname(mddev), mddev->new_layout);
 		goto out;
 	}
-	conf->mirrors = kzalloc(sizeof(struct mirror_info)*mddev->raid_disks,
-				 GFP_KERNEL);
-	if (!conf->mirrors) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-		       mdname(mddev));
-		goto out_free_conf;
-	}
+
+	err = -ENOMEM;
+	conf = kzalloc(sizeof(struct r10conf), GFP_KERNEL);
+	if (!conf)
+		goto out;
+
+	/* FIXME calc properly */
+	conf->mirrors = kzalloc(sizeof(struct raid10_info)*(mddev->raid_disks +
+							    max(0,-mddev->delta_disks)),
+				GFP_KERNEL);
+	if (!conf->mirrors)
+		goto out;
 
 	conf->tmppage = alloc_page(GFP_KERNEL);
 	if (!conf->tmppage)
-		goto out_free_conf;
+		goto out;
+
+	conf->geo = geo;
+	conf->copies = copies;
+	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
+					   r10bio_pool_free, conf);
+	if (!conf->r10bio_pool)
+		goto out;
+
+	calc_sectors(conf, mddev->dev_sectors);
+	if (mddev->reshape_position == MaxSector) {
+		conf->prev = conf->geo;
+		conf->reshape_progress = MaxSector;
+	} else {
+		if (setup_geo(&conf->prev, mddev, geo_old) != conf->copies) {
+			err = -EINVAL;
+			goto out;
+		}
+		conf->reshape_progress = mddev->reshape_position;
+		if (conf->prev.far_offset)
+			conf->prev.stride = 1 << conf->prev.chunk_shift;
+		else
+			/* far_copies must be 1 */
+			conf->prev.stride = conf->dev_sectors;
+	}
+	spin_lock_init(&conf->device_lock);
+	INIT_LIST_HEAD(&conf->retry_list);
+
+	spin_lock_init(&conf->resync_lock);
+	init_waitqueue_head(&conf->wait_barrier);
+
+	conf->thread = md_register_thread(raid10d, mddev, "raid10");
+	if (!conf->thread)
+		goto out;
 
 	conf->mddev = mddev;
-	conf->raid_disks = mddev->raid_disks;
-	conf->near_copies = nc;
-	conf->far_copies = fc;
-	conf->copies = nc*fc;
-	conf->far_offset = fo;
-	conf->chunk_mask = (sector_t)(mddev->chunk_size>>9)-1;
-	conf->chunk_shift = ffz(~mddev->chunk_size) - 9;
-	size = mddev->size >> (conf->chunk_shift-1);
-	sector_div(size, fc);
-	size = size * conf->raid_disks;
-	sector_div(size, nc);
-	/* 'size' is now the number of chunks in the array */
-	/* calculate "used chunks per device" in 'stride' */
-	stride = size * conf->copies;
+	return conf;
 
-	/* We need to round up when dividing by raid_disks to
-	 * get the stride size.
-	 */
-	stride += conf->raid_disks - 1;
-	sector_div(stride, conf->raid_disks);
-	mddev->size = stride  << (conf->chunk_shift-1);
+ out:
+	if (err == -ENOMEM)
+		printk(KERN_ERR "md/raid10:%s: couldn't allocate memory.\n",
+		       mdname(mddev));
+	if (conf) {
+		if (conf->r10bio_pool)
+			mempool_destroy(conf->r10bio_pool);
+		kfree(conf->mirrors);
+		safe_put_page(conf->tmppage);
+		kfree(conf);
+	}
+	return ERR_PTR(err);
+}
 
-	if (fo)
-		stride = 1;
-	else
-		sector_div(stride, fc);
-	conf->stride = stride << conf->chunk_shift;
+static int run(struct mddev *mddev)
+{
+	struct r10conf *conf;
+	int i, disk_idx, chunk_size;
+	struct raid10_info *disk;
+	struct md_rdev *rdev;
+	sector_t size;
+	sector_t min_offset_diff = 0;
+	int first = 1;
+	bool discard_supported = false;
+
+	if (mddev->private == NULL) {
+		conf = setup_conf(mddev);
+		if (IS_ERR(conf))
+			return PTR_ERR(conf);
+		mddev->private = conf;
+	}
+	conf = mddev->private;
+	if (!conf)
+		goto out;
 
-	conf->r10bio_pool = mempool_create(NR_RAID10_BIOS, r10bio_pool_alloc,
-						r10bio_pool_free, conf);
-	if (!conf->r10bio_pool) {
-		printk(KERN_ERR "raid10: couldn't allocate memory for %s\n",
-			mdname(mddev));
-		goto out_free_conf;
+	mddev->thread = conf->thread;
+	conf->thread = NULL;
+
+	chunk_size = mddev->chunk_sectors << 9;
+	if (mddev->queue) {
+		blk_queue_max_discard_sectors(mddev->queue,
+					      mddev->chunk_sectors);
+		blk_queue_max_write_same_sectors(mddev->queue, 0);
+		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));
 	}
 
-	spin_lock_init(&conf->device_lock);
-	mddev->queue->queue_lock = &conf->device_lock;
+	rdev_for_each(rdev, mddev) {
+		long long diff;
+		struct request_queue *q;
 
-	rdev_for_each(rdev, tmp, mddev) {
 		disk_idx = rdev->raid_disk;
-		if (disk_idx >= mddev->raid_disks
-		    || disk_idx < 0)
+		if (disk_idx < 0)
+			continue;
+		if (disk_idx >= conf->geo.raid_disks &&
+		    disk_idx >= conf->prev.raid_disks)
 			continue;
 		disk = conf->mirrors + disk_idx;
 
-		disk->rdev = rdev;
-
-		blk_queue_stack_limits(mddev->queue,
-				       rdev->bdev->bd_disk->queue);
-		/* as we don't honour merge_bvec_fn, we must never risk
-		 * violating it, so limit ->max_sector to one PAGE, as
-		 * a one page request is never in violation.
-		 */
-		if (rdev->bdev->bd_disk->queue->merge_bvec_fn &&
-		    mddev->queue->max_sectors > (PAGE_SIZE>>9))
-			mddev->queue->max_sectors = (PAGE_SIZE>>9);
+		if (test_bit(Replacement, &rdev->flags)) {
+			if (disk->replacement)
+				goto out_free_conf;
+			disk->replacement = rdev;
+		} else {
+			if (disk->rdev)
+				goto out_free_conf;
+			disk->rdev = rdev;
+		}
+		q = bdev_get_queue(rdev->bdev);
+		if (q->merge_bvec_fn)
+			mddev->merge_check_needed = 1;
+		diff = (rdev->new_data_offset - rdev->data_offset);
+		if (!mddev->reshape_backwards)
+			diff = -diff;
+		if (diff < 0)
+			diff = 0;
+		if (first || diff < min_offset_diff)
+			min_offset_diff = diff;
+
+		if (mddev->gendisk)
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
 
 		disk->head_position = 0;
-	}
-	INIT_LIST_HEAD(&conf->retry_list);
 
-	spin_lock_init(&conf->resync_lock);
-	init_waitqueue_head(&conf->wait_barrier);
+		if (blk_queue_discard(bdev_get_queue(rdev->bdev)))
+			discard_supported = true;
+	}
 
+	if (mddev->queue) {
+		if (discard_supported)
+			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
+						mddev->queue);
+		else
+			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
+						  mddev->queue);
+	}
 	/* need to check that every block has at least one working mirror */
-	if (!enough(conf)) {
-		printk(KERN_ERR "raid10: not enough operational mirrors for %s\n",
+	if (!enough(conf, -1)) {
+		printk(KERN_ERR "md/raid10:%s: not enough operational mirrors.\n",
 		       mdname(mddev));
 		goto out_free_conf;
 	}
 
+	if (conf->reshape_progress != MaxSector) {
+		/* must ensure that shape change is supported */
+		if (conf->geo.far_copies != 1 &&
+		    conf->geo.far_offset == 0)
+			goto out_free_conf;
+		if (conf->prev.far_copies != 1 &&
+		    conf->prev.far_offset == 0)
+			goto out_free_conf;
+	}
+
 	mddev->degraded = 0;
-	for (i = 0; i < conf->raid_disks; i++) {
+	for (i = 0;
+	     i < conf->geo.raid_disks
+		     || i < conf->prev.raid_disks;
+	     i++) {
 
 		disk = conf->mirrors + i;
 
+		if (!disk->rdev && disk->replacement) {
+			/* The replacement is all we have - use it */
+			disk->rdev = disk->replacement;
+			disk->replacement = NULL;
+			clear_bit(Replacement, &disk->rdev->flags);
+		}
+
 		if (!disk->rdev ||
 		    !test_bit(In_sync, &disk->rdev->flags)) {
 			disk->head_position = 0;
 			mddev->degraded++;
-			if (disk->rdev)
+			if (disk->rdev &&
+			    disk->rdev->saved_raid_disk < 0)
 				conf->fullsync = 1;
 		}
+		disk->recovery_disabled = mddev->recovery_disabled - 1;
 	}
 
-
-	mddev->thread = md_register_thread(raid10d, mddev, "%s_raid10");
-	if (!mddev->thread) {
-		printk(KERN_ERR
-		       "raid10: couldn't allocate thread for %s\n",
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "md/raid10:%s: not clean"
+		       " -- starting background reconstruction\n",
 		       mdname(mddev));
-		goto out_free_conf;
-	}
-
 	printk(KERN_INFO
-		"raid10: raid set %s active with %d out of %d devices\n",
-		mdname(mddev), mddev->raid_disks - mddev->degraded,
-		mddev->raid_disks);
+		"md/raid10:%s: active with %d out of %d devices\n",
+		mdname(mddev), conf->geo.raid_disks - mddev->degraded,
+		conf->geo.raid_disks);
 	/*
 	 * Ok, everything is just fine now
 	 */
-	mddev->array_sectors = size << conf->chunk_shift;
-	mddev->resync_max_sectors = size << conf->chunk_shift;
+	mddev->dev_sectors = conf->dev_sectors;
+	size = raid10_size(mddev, 0, 0);
+	md_set_array_sectors(mddev, size);
+	mddev->resync_max_sectors = size;
+
+	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);
+	}
 
-	mddev->queue->unplug_fn = raid10_unplug;
-	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...
-	 */
-	{
-		int stripe = conf->raid_disks * (mddev->chunk_size / PAGE_SIZE);
-		stripe /= conf->near_copies;
-		if (mddev->queue->backing_dev_info.ra_pages < 2* stripe)
-			mddev->queue->backing_dev_info.ra_pages = 2* stripe;
+	if (md_integrity_register(mddev))
+		goto out_free_conf;
+
+	if (conf->reshape_progress != MaxSector) {
+		unsigned long before_length, after_length;
+
+		before_length = ((1 << conf->prev.chunk_shift) *
+				 conf->prev.far_copies);
+		after_length = ((1 << conf->geo.chunk_shift) *
+				conf->geo.far_copies);
+
+		if (max(before_length, after_length) > min_offset_diff) {
+			/* This cannot work */
+			printk("md/raid10: offset difference not enough to continue reshape\n");
+			goto out_free_conf;
+		}
+		conf->offset_diff = min_offset_diff;
+
+		conf->reshape_safe = conf->reshape_progress;
+		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+		clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
+		set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
+		set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+		mddev->sync_thread = md_register_thread(md_do_sync, mddev,
+							"reshape");
 	}
 
-	if (conf->near_copies < mddev->raid_disks)
-		blk_queue_merge_bvec(mddev->queue, raid10_mergeable_bvec);
 	return 0;
 
 out_free_conf:
+	md_unregister_thread(&mddev->thread);
 	if (conf->r10bio_pool)
 		mempool_destroy(conf->r10bio_pool);
 	safe_put_page(conf->tmppage);
@@ -2202,24 +3816,30 @@ out:
 	return -EIO;
 }
 
-static int stop(mddev_t *mddev)
+static int stop(struct mddev *mddev)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
+
+	raise_barrier(conf, 0);
+	lower_barrier(conf);
+
+	md_unregister_thread(&mddev->thread);
+	if (mddev->queue)
+		/* the unplug fn references 'conf'*/
+		blk_sync_queue(mddev->queue);
 
-	md_unregister_thread(mddev->thread);
-	mddev->thread = NULL;
-	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
 	if (conf->r10bio_pool)
 		mempool_destroy(conf->r10bio_pool);
+	safe_put_page(conf->tmppage);
 	kfree(conf->mirrors);
 	kfree(conf);
 	mddev->private = NULL;
 	return 0;
 }
 
-static void raid10_quiesce(mddev_t *mddev, int state)
+static void raid10_quiesce(struct mddev *mddev, int state)
 {
-	conf_t *conf = mddev_to_conf(mddev);
+	struct r10conf *conf = mddev->private;
 
 	switch(state) {
 	case 1:
@@ -2229,16 +3849,866 @@ static void raid10_quiesce(mddev_t *mddev, int state)
 		lower_barrier(conf);
 		break;
 	}
-	if (mddev->thread) {
-		if (mddev->bitmap)
-			mddev->thread->timeout = mddev->bitmap->daemon_sleep * HZ;
+}
+
+static int raid10_resize(struct mddev *mddev, sector_t sectors)
+{
+	/* Resize of 'far' arrays is not supported.
+	 * For 'near' and 'offset' arrays we can set the
+	 * number of sectors used to be an appropriate multiple
+	 * of the chunk size.
+	 * For 'offset', this is far_copies*chunksize.
+	 * For 'near' the multiplier is the LCM of
+	 * near_copies and raid_disks.
+	 * So if far_copies > 1 && !far_offset, fail.
+	 * Else find LCM(raid_disks, near_copy)*far_copies and
+	 * multiply by chunk_size.  Then round to this number.
+	 * This is mostly done by raid10_size()
+	 */
+	struct r10conf *conf = mddev->private;
+	sector_t oldsize, size;
+
+	if (mddev->reshape_position != MaxSector)
+		return -EBUSY;
+
+	if (conf->geo.far_copies > 1 && !conf->geo.far_offset)
+		return -EINVAL;
+
+	oldsize = raid10_size(mddev, 0, 0);
+	size = raid10_size(mddev, sectors, 0);
+	if (mddev->external_size &&
+	    mddev->array_sectors > size)
+		return -EINVAL;
+	if (mddev->bitmap) {
+		int ret = bitmap_resize(mddev->bitmap, size, 0, 0);
+		if (ret)
+			return ret;
+	}
+	md_set_array_sectors(mddev, size);
+	set_capacity(mddev->gendisk, mddev->array_sectors);
+	revalidate_disk(mddev->gendisk);
+	if (sectors > mddev->dev_sectors &&
+	    mddev->recovery_cp > oldsize) {
+		mddev->recovery_cp = oldsize;
+		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+	}
+	calc_sectors(conf, sectors);
+	mddev->dev_sectors = conf->dev_sectors;
+	mddev->resync_max_sectors = size;
+	return 0;
+}
+
+static void *raid10_takeover_raid0(struct mddev *mddev)
+{
+	struct md_rdev *rdev;
+	struct r10conf *conf;
+
+	if (mddev->degraded > 0) {
+		printk(KERN_ERR "md/raid10:%s: Error: degraded raid0!\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	/* Set new parameters */
+	mddev->new_level = 10;
+	/* new layout: far_copies = 1, near_copies = 2 */
+	mddev->new_layout = (1<<8) + 2;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->delta_disks = mddev->raid_disks;
+	mddev->raid_disks *= 2;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	conf = setup_conf(mddev);
+	if (!IS_ERR(conf)) {
+		rdev_for_each(rdev, mddev)
+			if (rdev->raid_disk >= 0)
+				rdev->new_raid_disk = rdev->raid_disk * 2;
+		conf->barrier = 1;
+	}
+
+	return conf;
+}
+
+static void *raid10_takeover(struct mddev *mddev)
+{
+	struct r0conf *raid0_conf;
+
+	/* raid10 can take over:
+	 *  raid0 - providing it has only two drives
+	 */
+	if (mddev->level == 0) {
+		/* for raid0 takeover only one zone is supported */
+		raid0_conf = mddev->private;
+		if (raid0_conf->nr_strip_zones > 1) {
+			printk(KERN_ERR "md/raid10:%s: cannot takeover raid 0"
+			       " with more than one zone.\n",
+			       mdname(mddev));
+			return ERR_PTR(-EINVAL);
+		}
+		return raid10_takeover_raid0(mddev);
+	}
+	return ERR_PTR(-EINVAL);
+}
+
+static int raid10_check_reshape(struct mddev *mddev)
+{
+	/* Called when there is a request to change
+	 * - layout (to ->new_layout)
+	 * - chunk size (to ->new_chunk_sectors)
+	 * - raid_disks (by delta_disks)
+	 * or when trying to restart a reshape that was ongoing.
+	 *
+	 * We need to validate the request and possibly allocate
+	 * space if that might be an issue later.
+	 *
+	 * Currently we reject any reshape of a 'far' mode array,
+	 * allow chunk size to change if new is generally acceptable,
+	 * allow raid_disks to increase, and allow
+	 * a switch between 'near' mode and 'offset' mode.
+	 */
+	struct r10conf *conf = mddev->private;
+	struct geom geo;
+
+	if (conf->geo.far_copies != 1 && !conf->geo.far_offset)
+		return -EINVAL;
+
+	if (setup_geo(&geo, mddev, geo_start) != conf->copies)
+		/* mustn't change number of copies */
+		return -EINVAL;
+	if (geo.far_copies > 1 && !geo.far_offset)
+		/* Cannot switch to 'far' mode */
+		return -EINVAL;
+
+	if (mddev->array_sectors & geo.chunk_mask)
+			/* not factor of array size */
+			return -EINVAL;
+
+	if (!enough(conf, -1))
+		return -EINVAL;
+
+	kfree(conf->mirrors_new);
+	conf->mirrors_new = NULL;
+	if (mddev->delta_disks > 0) {
+		/* allocate new 'mirrors' list */
+		conf->mirrors_new = kzalloc(
+			sizeof(struct raid10_info)
+			*(mddev->raid_disks +
+			  mddev->delta_disks),
+			GFP_KERNEL);
+		if (!conf->mirrors_new)
+			return -ENOMEM;
+	}
+	return 0;
+}
+
+/*
+ * Need to check if array has failed when deciding whether to:
+ *  - start an array
+ *  - remove non-faulty devices
+ *  - add a spare
+ *  - allow a reshape
+ * This determination is simple when no reshape is happening.
+ * However if there is a reshape, we need to carefully check
+ * both the before and after sections.
+ * This is because some failed devices may only affect one
+ * of the two sections, and some non-in_sync devices may
+ * be insync in the section most affected by failed devices.
+ */
+static int calc_degraded(struct r10conf *conf)
+{
+	int degraded, degraded2;
+	int i;
+
+	rcu_read_lock();
+	degraded = 0;
+	/* 'prev' section first */
+	for (i = 0; i < conf->prev.raid_disks; i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
+		if (!rdev || test_bit(Faulty, &rdev->flags))
+			degraded++;
+		else if (!test_bit(In_sync, &rdev->flags))
+			/* When we can reduce the number of devices in
+			 * an array, this might not contribute to
+			 * 'degraded'.  It does now.
+			 */
+			degraded++;
+	}
+	rcu_read_unlock();
+	if (conf->geo.raid_disks == conf->prev.raid_disks)
+		return degraded;
+	rcu_read_lock();
+	degraded2 = 0;
+	for (i = 0; i < conf->geo.raid_disks; i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->mirrors[i].rdev);
+		if (!rdev || test_bit(Faulty, &rdev->flags))
+			degraded2++;
+		else if (!test_bit(In_sync, &rdev->flags)) {
+			/* If reshape is increasing the number of devices,
+			 * this section has already been recovered, so
+			 * it doesn't contribute to degraded.
+			 * else it does.
+			 */
+			if (conf->geo.raid_disks <= conf->prev.raid_disks)
+				degraded2++;
+		}
+	}
+	rcu_read_unlock();
+	if (degraded2 > degraded)
+		return degraded2;
+	return degraded;
+}
+
+static int raid10_start_reshape(struct mddev *mddev)
+{
+	/* A 'reshape' has been requested. This commits
+	 * the various 'new' fields and sets MD_RECOVER_RESHAPE
+	 * This also checks if there are enough spares and adds them
+	 * to the array.
+	 * We currently require enough spares to make the final
+	 * array non-degraded.  We also require that the difference
+	 * between old and new data_offset - on each device - is
+	 * enough that we never risk over-writing.
+	 */
+
+	unsigned long before_length, after_length;
+	sector_t min_offset_diff = 0;
+	int first = 1;
+	struct geom new;
+	struct r10conf *conf = mddev->private;
+	struct md_rdev *rdev;
+	int spares = 0;
+	int ret;
+
+	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
+		return -EBUSY;
+
+	if (setup_geo(&new, mddev, geo_start) != conf->copies)
+		return -EINVAL;
+
+	before_length = ((1 << conf->prev.chunk_shift) *
+			 conf->prev.far_copies);
+	after_length = ((1 << conf->geo.chunk_shift) *
+			conf->geo.far_copies);
+
+	rdev_for_each(rdev, mddev) {
+		if (!test_bit(In_sync, &rdev->flags)
+		    && !test_bit(Faulty, &rdev->flags))
+			spares++;
+		if (rdev->raid_disk >= 0) {
+			long long diff = (rdev->new_data_offset
+					  - rdev->data_offset);
+			if (!mddev->reshape_backwards)
+				diff = -diff;
+			if (diff < 0)
+				diff = 0;
+			if (first || diff < min_offset_diff)
+				min_offset_diff = diff;
+		}
+	}
+
+	if (max(before_length, after_length) > min_offset_diff)
+		return -EINVAL;
+
+	if (spares < mddev->delta_disks)
+		return -EINVAL;
+
+	conf->offset_diff = min_offset_diff;
+	spin_lock_irq(&conf->device_lock);
+	if (conf->mirrors_new) {
+		memcpy(conf->mirrors_new, conf->mirrors,
+		       sizeof(struct raid10_info)*conf->prev.raid_disks);
+		smp_mb();
+		kfree(conf->mirrors_old); /* FIXME and elsewhere */
+		conf->mirrors_old = conf->mirrors;
+		conf->mirrors = conf->mirrors_new;
+		conf->mirrors_new = NULL;
+	}
+	setup_geo(&conf->geo, mddev, geo_start);
+	smp_mb();
+	if (mddev->reshape_backwards) {
+		sector_t size = raid10_size(mddev, 0, 0);
+		if (size < mddev->array_sectors) {
+			spin_unlock_irq(&conf->device_lock);
+			printk(KERN_ERR "md/raid10:%s: array size must be reduce before number of disks\n",
+			       mdname(mddev));
+			return -EINVAL;
+		}
+		mddev->resync_max_sectors = size;
+		conf->reshape_progress = size;
+	} else
+		conf->reshape_progress = 0;
+	spin_unlock_irq(&conf->device_lock);
+
+	if (mddev->delta_disks && mddev->bitmap) {
+		ret = bitmap_resize(mddev->bitmap,
+				    raid10_size(mddev, 0,
+						conf->geo.raid_disks),
+				    0, 0);
+		if (ret)
+			goto abort;
+	}
+	if (mddev->delta_disks > 0) {
+		rdev_for_each(rdev, mddev)
+			if (rdev->raid_disk < 0 &&
+			    !test_bit(Faulty, &rdev->flags)) {
+				if (raid10_add_disk(mddev, rdev) == 0) {
+					if (rdev->raid_disk >=
+					    conf->prev.raid_disks)
+						set_bit(In_sync, &rdev->flags);
+					else
+						rdev->recovery_offset = 0;
+
+					if (sysfs_link_rdev(mddev, rdev))
+						/* Failure here  is OK */;
+				}
+			} else if (rdev->raid_disk >= conf->prev.raid_disks
+				   && !test_bit(Faulty, &rdev->flags)) {
+				/* This is a spare that was manually added */
+				set_bit(In_sync, &rdev->flags);
+			}
+	}
+	/* When a reshape changes the number of devices,
+	 * ->degraded is measured against the larger of the
+	 * pre and  post numbers.
+	 */
+	spin_lock_irq(&conf->device_lock);
+	mddev->degraded = calc_degraded(conf);
+	spin_unlock_irq(&conf->device_lock);
+	mddev->raid_disks = conf->geo.raid_disks;
+	mddev->reshape_position = conf->reshape_progress;
+	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+
+	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
+	clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
+	set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
+	set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
+
+	mddev->sync_thread = md_register_thread(md_do_sync, mddev,
+						"reshape");
+	if (!mddev->sync_thread) {
+		ret = -EAGAIN;
+		goto abort;
+	}
+	conf->reshape_checkpoint = jiffies;
+	md_wakeup_thread(mddev->sync_thread);
+	md_new_event(mddev);
+	return 0;
+
+abort:
+	mddev->recovery = 0;
+	spin_lock_irq(&conf->device_lock);
+	conf->geo = conf->prev;
+	mddev->raid_disks = conf->geo.raid_disks;
+	rdev_for_each(rdev, mddev)
+		rdev->new_data_offset = rdev->data_offset;
+	smp_wmb();
+	conf->reshape_progress = MaxSector;
+	mddev->reshape_position = MaxSector;
+	spin_unlock_irq(&conf->device_lock);
+	return ret;
+}
+
+/* Calculate the last device-address that could contain
+ * any block from the chunk that includes the array-address 's'
+ * and report the next address.
+ * i.e. the address returned will be chunk-aligned and after
+ * any data that is in the chunk containing 's'.
+ */
+static sector_t last_dev_address(sector_t s, struct geom *geo)
+{
+	s = (s | geo->chunk_mask) + 1;
+	s >>= geo->chunk_shift;
+	s *= geo->near_copies;
+	s = DIV_ROUND_UP_SECTOR_T(s, geo->raid_disks);
+	s *= geo->far_copies;
+	s <<= geo->chunk_shift;
+	return s;
+}
+
+/* Calculate the first device-address that could contain
+ * any block from the chunk that includes the array-address 's'.
+ * This too will be the start of a chunk
+ */
+static sector_t first_dev_address(sector_t s, struct geom *geo)
+{
+	s >>= geo->chunk_shift;
+	s *= geo->near_copies;
+	sector_div(s, geo->raid_disks);
+	s *= geo->far_copies;
+	s <<= geo->chunk_shift;
+	return s;
+}
+
+static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr,
+				int *skipped)
+{
+	/* We simply copy at most one chunk (smallest of old and new)
+	 * at a time, possibly less if that exceeds RESYNC_PAGES,
+	 * or we hit a bad block or something.
+	 * This might mean we pause for normal IO in the middle of
+	 * a chunk, but that is not a problem was mddev->reshape_position
+	 * can record any location.
+	 *
+	 * If we will want to write to a location that isn't
+	 * yet recorded as 'safe' (i.e. in metadata on disk) then
+	 * we need to flush all reshape requests and update the metadata.
+	 *
+	 * When reshaping forwards (e.g. to more devices), we interpret
+	 * 'safe' as the earliest block which might not have been copied
+	 * down yet.  We divide this by previous stripe size and multiply
+	 * by previous stripe length to get lowest device offset that we
+	 * cannot write to yet.
+	 * We interpret 'sector_nr' as an address that we want to write to.
+	 * From this we use last_device_address() to find where we might
+	 * write to, and first_device_address on the  'safe' position.
+	 * If this 'next' write position is after the 'safe' position,
+	 * we must update the metadata to increase the 'safe' position.
+	 *
+	 * When reshaping backwards, we round in the opposite direction
+	 * and perform the reverse test:  next write position must not be
+	 * less than current safe position.
+	 *
+	 * In all this the minimum difference in data offsets
+	 * (conf->offset_diff - always positive) allows a bit of slack,
+	 * so next can be after 'safe', but not by more than offset_disk
+	 *
+	 * We need to prepare all the bios here before we start any IO
+	 * to ensure the size we choose is acceptable to all devices.
+	 * The means one for each copy for write-out and an extra one for
+	 * read-in.
+	 * We store the read-in bio in ->master_bio and the others in
+	 * ->devs[x].bio and ->devs[x].repl_bio.
+	 */
+	struct r10conf *conf = mddev->private;
+	struct r10bio *r10_bio;
+	sector_t next, safe, last;
+	int max_sectors;
+	int nr_sectors;
+	int s;
+	struct md_rdev *rdev;
+	int need_flush = 0;
+	struct bio *blist;
+	struct bio *bio, *read_bio;
+	int sectors_done = 0;
+
+	if (sector_nr == 0) {
+		/* If restarting in the middle, skip the initial sectors */
+		if (mddev->reshape_backwards &&
+		    conf->reshape_progress < raid10_size(mddev, 0, 0)) {
+			sector_nr = (raid10_size(mddev, 0, 0)
+				     - conf->reshape_progress);
+		} else if (!mddev->reshape_backwards &&
+			   conf->reshape_progress > 0)
+			sector_nr = conf->reshape_progress;
+		if (sector_nr) {
+			mddev->curr_resync_completed = sector_nr;
+			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+			*skipped = 1;
+			return sector_nr;
+		}
+	}
+
+	/* We don't use sector_nr to track where we are up to
+	 * as that doesn't work well for ->reshape_backwards.
+	 * So just use ->reshape_progress.
+	 */
+	if (mddev->reshape_backwards) {
+		/* 'next' is the earliest device address that we might
+		 * write to for this chunk in the new layout
+		 */
+		next = first_dev_address(conf->reshape_progress - 1,
+					 &conf->geo);
+
+		/* 'safe' is the last device address that we might read from
+		 * in the old layout after a restart
+		 */
+		safe = last_dev_address(conf->reshape_safe - 1,
+					&conf->prev);
+
+		if (next + conf->offset_diff < safe)
+			need_flush = 1;
+
+		last = conf->reshape_progress - 1;
+		sector_nr = last & ~(sector_t)(conf->geo.chunk_mask
+					       & conf->prev.chunk_mask);
+		if (sector_nr + RESYNC_BLOCK_SIZE/512 < last)
+			sector_nr = last + 1 - RESYNC_BLOCK_SIZE/512;
+	} else {
+		/* 'next' is after the last device address that we
+		 * might write to for this chunk in the new layout
+		 */
+		next = last_dev_address(conf->reshape_progress, &conf->geo);
+
+		/* 'safe' is the earliest device address that we might
+		 * read from in the old layout after a restart
+		 */
+		safe = first_dev_address(conf->reshape_safe, &conf->prev);
+
+		/* Need to update metadata if 'next' might be beyond 'safe'
+		 * as that would possibly corrupt data
+		 */
+		if (next > safe + conf->offset_diff)
+			need_flush = 1;
+
+		sector_nr = conf->reshape_progress;
+		last  = sector_nr | (conf->geo.chunk_mask
+				     & conf->prev.chunk_mask);
+
+		if (sector_nr + RESYNC_BLOCK_SIZE/512 <= last)
+			last = sector_nr + RESYNC_BLOCK_SIZE/512 - 1;
+	}
+
+	if (need_flush ||
+	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
+		/* Need to update reshape_position in metadata */
+		wait_barrier(conf);
+		mddev->reshape_position = conf->reshape_progress;
+		if (mddev->reshape_backwards)
+			mddev->curr_resync_completed = raid10_size(mddev, 0, 0)
+				- conf->reshape_progress;
 		else
-			mddev->thread->timeout = MAX_SCHEDULE_TIMEOUT;
+			mddev->curr_resync_completed = conf->reshape_progress;
+		conf->reshape_checkpoint = jiffies;
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		md_wakeup_thread(mddev->thread);
+		wait_event(mddev->sb_wait, mddev->flags == 0 ||
+			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
+			allow_barrier(conf);
+			return sectors_done;
+		}
+		conf->reshape_safe = mddev->reshape_position;
+		allow_barrier(conf);
+	}
+
+read_more:
+	/* Now schedule reads for blocks from sector_nr to last */
+	r10_bio = mempool_alloc(conf->r10buf_pool, GFP_NOIO);
+	raise_barrier(conf, sectors_done != 0);
+	atomic_set(&r10_bio->remaining, 0);
+	r10_bio->mddev = mddev;
+	r10_bio->sector = sector_nr;
+	set_bit(R10BIO_IsReshape, &r10_bio->state);
+	r10_bio->sectors = last - sector_nr + 1;
+	rdev = read_balance(conf, r10_bio, &max_sectors);
+	BUG_ON(!test_bit(R10BIO_Previous, &r10_bio->state));
+
+	if (!rdev) {
+		/* Cannot read from here, so need to record bad blocks
+		 * on all the target devices.
+		 */
+		// FIXME
+		set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+		return sectors_done;
+	}
+
+	read_bio = bio_alloc_mddev(GFP_KERNEL, RESYNC_PAGES, mddev);
+
+	read_bio->bi_bdev = rdev->bdev;
+	read_bio->bi_iter.bi_sector = (r10_bio->devs[r10_bio->read_slot].addr
+			       + rdev->data_offset);
+	read_bio->bi_private = r10_bio;
+	read_bio->bi_end_io = end_sync_read;
+	read_bio->bi_rw = READ;
+	read_bio->bi_flags &= ~(BIO_POOL_MASK - 1);
+	read_bio->bi_flags |= 1 << BIO_UPTODATE;
+	read_bio->bi_vcnt = 0;
+	read_bio->bi_iter.bi_size = 0;
+	r10_bio->master_bio = read_bio;
+	r10_bio->read_slot = r10_bio->devs[r10_bio->read_slot].devnum;
+
+	/* Now find the locations in the new layout */
+	__raid10_find_phys(&conf->geo, r10_bio);
+
+	blist = read_bio;
+	read_bio->bi_next = NULL;
+
+	for (s = 0; s < conf->copies*2; s++) {
+		struct bio *b;
+		int d = r10_bio->devs[s/2].devnum;
+		struct md_rdev *rdev2;
+		if (s&1) {
+			rdev2 = conf->mirrors[d].replacement;
+			b = r10_bio->devs[s/2].repl_bio;
+		} else {
+			rdev2 = conf->mirrors[d].rdev;
+			b = r10_bio->devs[s/2].bio;
+		}
+		if (!rdev2 || test_bit(Faulty, &rdev2->flags))
+			continue;
+
+		bio_reset(b);
+		b->bi_bdev = rdev2->bdev;
+		b->bi_iter.bi_sector = r10_bio->devs[s/2].addr +
+			rdev2->new_data_offset;
+		b->bi_private = r10_bio;
+		b->bi_end_io = end_reshape_write;
+		b->bi_rw = WRITE;
+		b->bi_next = blist;
+		blist = b;
+	}
+
+	/* Now add as many pages as possible to all of these bios. */
+
+	nr_sectors = 0;
+	for (s = 0 ; s < max_sectors; s += PAGE_SIZE >> 9) {
+		struct page *page = r10_bio->devs[0].bio->bi_io_vec[s/(PAGE_SIZE>>9)].bv_page;
+		int len = (max_sectors - s) << 9;
+		if (len > PAGE_SIZE)
+			len = PAGE_SIZE;
+		for (bio = blist; bio ; bio = bio->bi_next) {
+			struct bio *bio2;
+			if (bio_add_page(bio, page, len, 0))
+				continue;
+
+			/* Didn't fit, must stop */
+			for (bio2 = blist;
+			     bio2 && bio2 != bio;
+			     bio2 = bio2->bi_next) {
+				/* Remove last page from this bio */
+				bio2->bi_vcnt--;
+				bio2->bi_iter.bi_size -= len;
+				bio2->bi_flags &= ~(1<<BIO_SEG_VALID);
+			}
+			goto bio_full;
+		}
+		sector_nr += len >> 9;
+		nr_sectors += len >> 9;
 	}
+bio_full:
+	r10_bio->sectors = nr_sectors;
+
+	/* Now submit the read */
+	md_sync_acct(read_bio->bi_bdev, r10_bio->sectors);
+	atomic_inc(&r10_bio->remaining);
+	read_bio->bi_next = NULL;
+	generic_make_request(read_bio);
+	sector_nr += nr_sectors;
+	sectors_done += nr_sectors;
+	if (sector_nr <= last)
+		goto read_more;
+
+	/* Now that we have done the whole section we can
+	 * update reshape_progress
+	 */
+	if (mddev->reshape_backwards)
+		conf->reshape_progress -= sectors_done;
+	else
+		conf->reshape_progress += sectors_done;
+
+	return sectors_done;
 }
 
-static struct mdk_personality raid10_personality =
+static void end_reshape_request(struct r10bio *r10_bio);
+static int handle_reshape_read_error(struct mddev *mddev,
+				     struct r10bio *r10_bio);
+static void reshape_request_write(struct mddev *mddev, struct r10bio *r10_bio)
+{
+	/* Reshape read completed.  Hopefully we have a block
+	 * to write out.
+	 * If we got a read error then we do sync 1-page reads from
+	 * elsewhere until we find the data - or give up.
+	 */
+	struct r10conf *conf = mddev->private;
+	int s;
+
+	if (!test_bit(R10BIO_Uptodate, &r10_bio->state))
+		if (handle_reshape_read_error(mddev, r10_bio) < 0) {
+			/* Reshape has been aborted */
+			md_done_sync(mddev, r10_bio->sectors, 0);
+			return;
+		}
+
+	/* We definitely have the data in the pages, schedule the
+	 * writes.
+	 */
+	atomic_set(&r10_bio->remaining, 1);
+	for (s = 0; s < conf->copies*2; s++) {
+		struct bio *b;
+		int d = r10_bio->devs[s/2].devnum;
+		struct md_rdev *rdev;
+		if (s&1) {
+			rdev = conf->mirrors[d].replacement;
+			b = r10_bio->devs[s/2].repl_bio;
+		} else {
+			rdev = conf->mirrors[d].rdev;
+			b = r10_bio->devs[s/2].bio;
+		}
+		if (!rdev || test_bit(Faulty, &rdev->flags))
+			continue;
+		atomic_inc(&rdev->nr_pending);
+		md_sync_acct(b->bi_bdev, r10_bio->sectors);
+		atomic_inc(&r10_bio->remaining);
+		b->bi_next = NULL;
+		generic_make_request(b);
+	}
+	end_reshape_request(r10_bio);
+}
+
+static void end_reshape(struct r10conf *conf)
+{
+	if (test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery))
+		return;
+
+	spin_lock_irq(&conf->device_lock);
+	conf->prev = conf->geo;
+	md_finish_reshape(conf->mddev);
+	smp_wmb();
+	conf->reshape_progress = MaxSector;
+	spin_unlock_irq(&conf->device_lock);
+
+	/* read-ahead size must cover two whole stripes, which is
+	 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
+	 */
+	if (conf->mddev->queue) {
+		int stripe = conf->geo.raid_disks *
+			((conf->mddev->chunk_sectors << 9) / PAGE_SIZE);
+		stripe /= conf->geo.near_copies;
+		if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
+			conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
+	}
+	conf->fullsync = 0;
+}
+
+
+static int handle_reshape_read_error(struct mddev *mddev,
+				     struct r10bio *r10_bio)
+{
+	/* Use sync reads to get the blocks from somewhere else */
+	int sectors = r10_bio->sectors;
+	struct r10conf *conf = mddev->private;
+	struct {
+		struct r10bio r10_bio;
+		struct r10dev devs[conf->copies];
+	} on_stack;
+	struct r10bio *r10b = &on_stack.r10_bio;
+	int slot = 0;
+	int idx = 0;
+	struct bio_vec *bvec = r10_bio->master_bio->bi_io_vec;
+
+	r10b->sector = r10_bio->sector;
+	__raid10_find_phys(&conf->prev, r10b);
+
+	while (sectors) {
+		int s = sectors;
+		int success = 0;
+		int first_slot = slot;
+
+		if (s > (PAGE_SIZE >> 9))
+			s = PAGE_SIZE >> 9;
+
+		while (!success) {
+			int d = r10b->devs[slot].devnum;
+			struct md_rdev *rdev = conf->mirrors[d].rdev;
+			sector_t addr;
+			if (rdev == NULL ||
+			    test_bit(Faulty, &rdev->flags) ||
+			    !test_bit(In_sync, &rdev->flags))
+				goto failed;
+
+			addr = r10b->devs[slot].addr + idx * PAGE_SIZE;
+			success = sync_page_io(rdev,
+					       addr,
+					       s << 9,
+					       bvec[idx].bv_page,
+					       READ, false);
+			if (success)
+				break;
+		failed:
+			slot++;
+			if (slot >= conf->copies)
+				slot = 0;
+			if (slot == first_slot)
+				break;
+		}
+		if (!success) {
+			/* couldn't read this block, must give up */
+			set_bit(MD_RECOVERY_INTR,
+				&mddev->recovery);
+			return -EIO;
+		}
+		sectors -= s;
+		idx++;
+	}
+	return 0;
+}
+
+static void end_reshape_write(struct bio *bio, int error)
+{
+	int uptodate = test_bit(BIO_UPTODATE, &bio->bi_flags);
+	struct r10bio *r10_bio = bio->bi_private;
+	struct mddev *mddev = r10_bio->mddev;
+	struct r10conf *conf = mddev->private;
+	int d;
+	int slot;
+	int repl;
+	struct md_rdev *rdev = NULL;
+
+	d = find_bio_disk(conf, r10_bio, bio, &slot, &repl);
+	if (repl)
+		rdev = conf->mirrors[d].replacement;
+	if (!rdev) {
+		smp_mb();
+		rdev = conf->mirrors[d].rdev;
+	}
+
+	if (!uptodate) {
+		/* FIXME should record badblock */
+		md_error(mddev, rdev);
+	}
+
+	rdev_dec_pending(rdev, mddev);
+	end_reshape_request(r10_bio);
+}
+
+static void end_reshape_request(struct r10bio *r10_bio)
+{
+	if (!atomic_dec_and_test(&r10_bio->remaining))
+		return;
+	md_done_sync(r10_bio->mddev, r10_bio->sectors, 1);
+	bio_put(r10_bio->master_bio);
+	put_buf(r10_bio);
+}
+
+static void raid10_finish_reshape(struct mddev *mddev)
+{
+	struct r10conf *conf = mddev->private;
+
+	if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+		return;
+
+	if (mddev->delta_disks > 0) {
+		sector_t size = raid10_size(mddev, 0, 0);
+		md_set_array_sectors(mddev, size);
+		if (mddev->recovery_cp > mddev->resync_max_sectors) {
+			mddev->recovery_cp = mddev->resync_max_sectors;
+			set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
+		}
+		mddev->resync_max_sectors = size;
+		set_capacity(mddev->gendisk, mddev->array_sectors);
+		revalidate_disk(mddev->gendisk);
+	} else {
+		int d;
+		for (d = conf->geo.raid_disks ;
+		     d < conf->geo.raid_disks - mddev->delta_disks;
+		     d++) {
+			struct md_rdev *rdev = conf->mirrors[d].rdev;
+			if (rdev)
+				clear_bit(In_sync, &rdev->flags);
+			rdev = conf->mirrors[d].replacement;
+			if (rdev)
+				clear_bit(In_sync, &rdev->flags);
+		}
+	}
+	mddev->layout = mddev->new_layout;
+	mddev->chunk_sectors = 1 << conf->geo.chunk_shift;
+	mddev->reshape_position = MaxSector;
+	mddev->delta_disks = 0;
+	mddev->reshape_backwards = 0;
+}
+
+static struct md_personality raid10_personality =
 {
 	.name		= "raid10",
 	.level		= 10,
@@ -2253,6 +4723,12 @@ static struct mdk_personality raid10_personality =
 	.spare_active	= raid10_spare_active,
 	.sync_request	= sync_request,
 	.quiesce	= raid10_quiesce,
+	.size		= raid10_size,
+	.resize		= raid10_resize,
+	.takeover	= raid10_takeover,
+	.check_reshape	= raid10_check_reshape,
+	.start_reshape	= raid10_start_reshape,
+	.finish_reshape	= raid10_finish_reshape,
 };
 
 static int __init raid_init(void)
@@ -2268,6 +4744,9 @@ static void raid_exit(void)
 module_init(raid_init);
 module_exit(raid_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("RAID10 (striped mirror) personality for MD");
 MODULE_ALIAS("md-personality-9"); /* RAID10 */
 MODULE_ALIAS("md-raid10");
 MODULE_ALIAS("md-level-10");
+
+module_param(max_queued_requests, int, S_IRUGO|S_IWUSR);
diff --git a/drivers/md/raid10.h b/drivers/md/raid10.h
new file mode 100644
index 00000000000..157d69e83ff
--- /dev/null
+++ b/drivers/md/raid10.h
@@ -0,0 +1,156 @@
+#ifndef _RAID10_H
+#define _RAID10_H
+
+struct raid10_info {
+	struct md_rdev	*rdev, *replacement;
+	sector_t	head_position;
+	int		recovery_disabled;	/* matches
+						 * mddev->recovery_disabled
+						 * when we shouldn't try
+						 * recovering this device.
+						 */
+};
+
+struct r10conf {
+	struct mddev		*mddev;
+	struct raid10_info	*mirrors;
+	struct raid10_info	*mirrors_new, *mirrors_old;
+	spinlock_t		device_lock;
+
+	/* geometry */
+	struct geom {
+		int		raid_disks;
+		int		near_copies;  /* number of copies laid out
+					       * raid0 style */
+		int		far_copies;   /* number of copies laid out
+					       * at large strides across drives
+					       */
+		int		far_offset;   /* far_copies are offset by 1
+					       * stripe instead of many
+					       */
+		sector_t	stride;	      /* distance between far copies.
+					       * This is size / far_copies unless
+					       * far_offset, in which case it is
+					       * 1 stripe.
+					       */
+		int             far_set_size; /* The number of devices in a set,
+					       * where a 'set' are devices that
+					       * contain far/offset copies of
+					       * each other.
+					       */
+		int		chunk_shift; /* shift from chunks to sectors */
+		sector_t	chunk_mask;
+	} prev, geo;
+	int			copies;	      /* near_copies * far_copies.
+					       * must be <= raid_disks
+					       */
+
+	sector_t		dev_sectors;  /* temp copy of
+					       * mddev->dev_sectors */
+	sector_t		reshape_progress;
+	sector_t		reshape_safe;
+	unsigned long		reshape_checkpoint;
+	sector_t		offset_diff;
+
+	struct list_head	retry_list;
+	/* queue pending writes and submit them on unplug */
+	struct bio_list		pending_bio_list;
+	int			pending_count;
+
+	spinlock_t		resync_lock;
+	int			nr_pending;
+	int			nr_waiting;
+	int			nr_queued;
+	int			barrier;
+	sector_t		next_resync;
+	int			fullsync;  /* set to 1 if a full sync is needed,
+					    * (fresh device added).
+					    * Cleared when a sync completes.
+					    */
+	int			have_replacement; /* There is at least one
+						   * replacement device.
+						   */
+	wait_queue_head_t	wait_barrier;
+
+	mempool_t		*r10bio_pool;
+	mempool_t		*r10buf_pool;
+	struct page		*tmppage;
+
+	/* When taking over an array from a different personality, we store
+	 * the new thread here until we fully activate the array.
+	 */
+	struct md_thread	*thread;
+};
+
+/*
+ * this is our 'private' RAID10 bio.
+ *
+ * it contains information about what kind of IO operations were started
+ * for this RAID10 operation, and about their status:
+ */
+
+struct r10bio {
+	atomic_t		remaining; /* 'have we finished' count,
+					    * used from IRQ handlers
+					    */
+	sector_t		sector;	/* virtual sector number */
+	int			sectors;
+	unsigned long		state;
+	struct mddev		*mddev;
+	/*
+	 * original bio going to /dev/mdx
+	 */
+	struct bio		*master_bio;
+	/*
+	 * if the IO is in READ direction, then this is where we read
+	 */
+	int			read_slot;
+
+	struct list_head	retry_list;
+	/*
+	 * if the IO is in WRITE direction, then multiple bios are used,
+	 * one for each copy.
+	 * When resyncing we also use one for each copy.
+	 * When reconstructing, we use 2 bios, one for read, one for write.
+	 * We choose the number when they are allocated.
+	 * We sometimes need an extra bio to write to the replacement.
+	 */
+	struct r10dev {
+		struct bio	*bio;
+		union {
+			struct bio	*repl_bio; /* used for resync and
+						    * writes */
+			struct md_rdev	*rdev;	   /* used for reads
+						    * (read_slot >= 0) */
+		};
+		sector_t	addr;
+		int		devnum;
+	} devs[0];
+};
+
+/* bits for r10bio.state */
+enum r10bio_state {
+	R10BIO_Uptodate,
+	R10BIO_IsSync,
+	R10BIO_IsRecover,
+	R10BIO_IsReshape,
+	R10BIO_Degraded,
+/* Set ReadError on bios that experience a read error
+ * so that raid10d knows what to do with them.
+ */
+	R10BIO_ReadError,
+/* If a write for this request means we can clear some
+ * known-bad-block records, we set this flag.
+ */
+	R10BIO_MadeGood,
+	R10BIO_WriteError,
+/* During a reshape we might be performing IO on the
+ * 'previous' part of the array, in which case this
+ * flag is set
+ */
+	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 224de022e7c..6234b2e8458 100644
--- a/drivers/md/raid5.c
+++ b/drivers/md/raid5.c
@@ -27,12 +27,12 @@
  *
  * We group bitmap updates into batches.  Each batch has a number.
  * We may write out several batches at once, but that isn't very important.
- * conf->bm_write is the number of the last batch successfully written.
- * conf->bm_flush is the number of the last batch that was closed to
+ * conf->seq_write is the number of the last batch successfully written.
+ * conf->seq_flush is the number of the last batch that was closed to
  *    new additions.
  * When we discover that we will need to write to any block in a stripe
  * (in add_stripe_bio) we update the in-memory bitmap and record in sh->bm_seq
- * the number of the batch it will be in. This is bm_flush+1.
+ * the number of the batch it will be in. This is seq_flush+1.
  * When we are ready to do a write, if that batch hasn't been written yet,
  *   we plug the array and queue the stripe for later.
  * When an unplug happens, we increment bm_flush, thus closing the current
@@ -43,17 +43,28 @@
  * miss any bits.
  */
 
+#include <linux/blkdev.h>
+#include <linux/kthread.h>
+#include <linux/raid/pq.h>
+#include <linux/async_tx.h>
 #include <linux/module.h>
+#include <linux/async.h>
+#include <linux/seq_file.h>
+#include <linux/cpu.h>
 #include <linux/slab.h>
-#include <linux/highmem.h>
-#include <linux/bitops.h>
-#include <linux/kthread.h>
-#include <asm/atomic.h>
-#include "raid6.h"
+#include <linux/ratelimit.h>
+#include <linux/nodemask.h>
+#include <trace/events/block.h>
 
-#include <linux/raid/bitmap.h>
-#include <linux/async_tx.h>
+#include "md.h"
+#include "raid5.h"
+#include "raid0.h"
+#include "bitmap.h"
 
+#define cpu_to_group(cpu) cpu_to_node(cpu)
+#define ANY_GROUP NUMA_NO_NODE
+
+static struct workqueue_struct *raid5_wq;
 /*
  * Stripe cache
  */
@@ -66,8 +77,49 @@
 #define BYPASS_THRESHOLD	1
 #define NR_HASH			(PAGE_SIZE / sizeof(struct hlist_head))
 #define HASH_MASK		(NR_HASH - 1)
+#define MAX_STRIPE_BATCH	8
+
+static inline struct hlist_head *stripe_hash(struct r5conf *conf, sector_t sect)
+{
+	int hash = (sect >> STRIPE_SHIFT) & HASH_MASK;
+	return &conf->stripe_hashtbl[hash];
+}
+
+static inline int stripe_hash_locks_hash(sector_t sect)
+{
+	return (sect >> STRIPE_SHIFT) & STRIPE_HASH_LOCKS_MASK;
+}
+
+static inline void lock_device_hash_lock(struct r5conf *conf, int hash)
+{
+	spin_lock_irq(conf->hash_locks + hash);
+	spin_lock(&conf->device_lock);
+}
+
+static inline void unlock_device_hash_lock(struct r5conf *conf, int hash)
+{
+	spin_unlock(&conf->device_lock);
+	spin_unlock_irq(conf->hash_locks + hash);
+}
 
-#define stripe_hash(conf, sect)	(&((conf)->stripe_hashtbl[((sect) >> STRIPE_SHIFT) & HASH_MASK]))
+static inline void lock_all_device_hash_locks_irq(struct r5conf *conf)
+{
+	int i;
+	local_irq_disable();
+	spin_lock(conf->hash_locks);
+	for (i = 1; i < NR_STRIPE_HASH_LOCKS; i++)
+		spin_lock_nest_lock(conf->hash_locks + i, conf->hash_locks);
+	spin_lock(&conf->device_lock);
+}
+
+static inline void unlock_all_device_hash_locks_irq(struct r5conf *conf)
+{
+	int i;
+	spin_unlock(&conf->device_lock);
+	for (i = NR_STRIPE_HASH_LOCKS; i; i--)
+		spin_unlock(conf->hash_locks + i - 1);
+	local_irq_enable();
+}
 
 /* bio's attached to a stripe+device for I/O are linked together in bi_sector
  * order without overlap.  There may be several bio's per stripe+device, and
@@ -75,38 +127,97 @@
  * When walking this list for a particular stripe+device, we must never proceed
  * beyond a bio that extends past this device, as the next bio might no longer
  * be valid.
- * This macro is used to determine the 'next' bio in the list, given the sector
+ * This function is used to determine the 'next' bio in the list, given the sector
  * of the current stripe+device
  */
-#define r5_next_bio(bio, sect) ( ( (bio)->bi_sector + ((bio)->bi_size>>9) < sect + STRIPE_SECTORS) ? (bio)->bi_next : NULL)
+static inline struct bio *r5_next_bio(struct bio *bio, sector_t sector)
+{
+	int sectors = bio_sectors(bio);
+	if (bio->bi_iter.bi_sector + sectors < sector + STRIPE_SECTORS)
+		return bio->bi_next;
+	else
+		return NULL;
+}
+
 /*
- * The following can be used to debug the driver
+ * 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
  */
-#define RAID5_PARANOIA	1
-#if RAID5_PARANOIA && defined(CONFIG_SMP)
-# define CHECK_DEVLOCK() assert_spin_locked(&conf->device_lock)
-#else
-# define CHECK_DEVLOCK()
-#endif
+static inline int raid5_bi_processed_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return (atomic_read(segments) >> 16) & 0xffff;
+}
 
-#ifdef DEBUG
-#define inline
-#define __inline__
-#endif
+static inline int raid5_dec_bi_active_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	return atomic_sub_return(1, segments) & 0xffff;
+}
 
-#define printk_rl(args...) ((void) (printk_ratelimit() && printk(args)))
+static inline void raid5_inc_bi_active_stripes(struct bio *bio)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_inc(segments);
+}
 
-#if !RAID6_USE_EMPTY_ZERO_PAGE
-/* In .bss so it's zeroed */
-const char raid6_empty_zero_page[PAGE_SIZE] __attribute__((aligned(256)));
-#endif
+static inline void raid5_set_bi_processed_stripes(struct bio *bio,
+	unsigned int cnt)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	int old, new;
+
+	do {
+		old = atomic_read(segments);
+		new = (old & 0xffff) | (cnt << 16);
+	} while (atomic_cmpxchg(segments, old, new) != old);
+}
+
+static inline void raid5_set_bi_stripes(struct bio *bio, unsigned int cnt)
+{
+	atomic_t *segments = (atomic_t *)&bio->bi_phys_segments;
+	atomic_set(segments, cnt);
+}
 
+/* Find first data disk in a raid6 stripe */
+static inline int raid6_d0(struct stripe_head *sh)
+{
+	if (sh->ddf_layout)
+		/* ddf always start from first device */
+		return 0;
+	/* md starts just after Q block */
+	if (sh->qd_idx == sh->disks - 1)
+		return 0;
+	else
+		return sh->qd_idx + 1;
+}
 static inline int raid6_next_disk(int disk, int raid_disks)
 {
 	disk++;
 	return (disk < raid_disks) ? disk : 0;
 }
 
+/* When walking through the disks in a raid5, starting at raid6_d0,
+ * We need to map each disk to a 'slot', where the data disks are slot
+ * 0 .. raid_disks-3, the parity disk is raid_disks-2 and the Q disk
+ * is raid_disks-1.  This help does that mapping.
+ */
+static int raid6_idx_to_slot(int idx, struct stripe_head *sh,
+			     int *count, int syndrome_disks)
+{
+	int slot = *count;
+
+	if (sh->ddf_layout)
+		(*count)++;
+	if (idx == sh->pd_idx)
+		return syndrome_disks;
+	if (idx == sh->qd_idx)
+		return syndrome_disks + 1;
+	if (!sh->ddf_layout)
+		(*count)++;
+	return slot;
+}
+
 static void return_io(struct bio *return_bi)
 {
 	struct bio *bi = return_bi;
@@ -114,13 +225,15 @@ static void return_io(struct bio *return_bi)
 
 		return_bi = bi->bi_next;
 		bi->bi_next = NULL;
-		bi->bi_size = 0;
+		bi->bi_iter.bi_size = 0;
+		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
+					 bi, 0);
 		bio_endio(bi, 0);
 		bi = return_bi;
 	}
 }
 
-static void print_raid5_conf (raid5_conf_t *conf);
+static void print_raid5_conf (struct r5conf *conf);
 
 static int stripe_operations_active(struct stripe_head *sh)
 {
@@ -129,49 +242,203 @@ static int stripe_operations_active(struct stripe_head *sh)
 	       test_bit(STRIPE_COMPUTE_RUN, &sh->state);
 }
 
-static void __release_stripe(raid5_conf_t *conf, struct stripe_head *sh)
+static void raid5_wakeup_stripe_thread(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);
-				blk_plug_device(conf->mddev->queue);
-			} else if (test_bit(STRIPE_BIT_DELAY, &sh->state) &&
-				   sh->bm_seq - conf->seq_write > 0) {
-				list_add_tail(&sh->lru, &conf->bitmap_list);
-				blk_plug_device(conf->mddev->queue);
-			} else {
-				clear_bit(STRIPE_BIT_DELAY, &sh->state);
+	struct r5conf *conf = sh->raid_conf;
+	struct r5worker_group *group;
+	int thread_cnt;
+	int i, cpu = sh->cpu;
+
+	if (!cpu_online(cpu)) {
+		cpu = cpumask_any(cpu_online_mask);
+		sh->cpu = cpu;
+	}
+
+	if (list_empty(&sh->lru)) {
+		struct r5worker_group *group;
+		group = conf->worker_groups + cpu_to_group(cpu);
+		list_add_tail(&sh->lru, &group->handle_list);
+		group->stripes_cnt++;
+		sh->group = group;
+	}
+
+	if (conf->worker_cnt_per_group == 0) {
+		md_wakeup_thread(conf->mddev->thread);
+		return;
+	}
+
+	group = conf->worker_groups + cpu_to_group(sh->cpu);
+
+	group->workers[0].working = true;
+	/* at least one worker should run to avoid race */
+	queue_work_on(sh->cpu, raid5_wq, &group->workers[0].work);
+
+	thread_cnt = group->stripes_cnt / MAX_STRIPE_BATCH - 1;
+	/* wakeup more workers */
+	for (i = 1; i < conf->worker_cnt_per_group && thread_cnt > 0; i++) {
+		if (group->workers[i].working == false) {
+			group->workers[i].working = true;
+			queue_work_on(sh->cpu, raid5_wq,
+				      &group->workers[i].work);
+			thread_cnt--;
+		}
+	}
+}
+
+static void do_release_stripe(struct r5conf *conf, struct stripe_head *sh,
+			      struct list_head *temp_inactive_list)
+{
+	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);
+			if (atomic_read(&conf->preread_active_stripes)
+			    < IO_THRESHOLD)
+				md_wakeup_thread(conf->mddev->thread);
+		} 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);
+			if (conf->worker_cnt_per_group == 0) {
 				list_add_tail(&sh->lru, &conf->handle_list);
+			} else {
+				raid5_wakeup_stripe_thread(sh);
+				return;
 			}
-			md_wakeup_thread(conf->mddev->thread);
-		} else {
-			BUG_ON(stripe_operations_active(sh));
-			if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-				atomic_dec(&conf->preread_active_stripes);
-				if (atomic_read(&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);
-			}
 		}
+		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, temp_inactive_list);
+	}
+}
+
+static void __release_stripe(struct r5conf *conf, struct stripe_head *sh,
+			     struct list_head *temp_inactive_list)
+{
+	if (atomic_dec_and_test(&sh->count))
+		do_release_stripe(conf, sh, temp_inactive_list);
+}
+
+/*
+ * @hash could be NR_STRIPE_HASH_LOCKS, then we have a list of inactive_list
+ *
+ * Be careful: Only one task can add/delete stripes from temp_inactive_list at
+ * given time. Adding stripes only takes device lock, while deleting stripes
+ * only takes hash lock.
+ */
+static void release_inactive_stripe_list(struct r5conf *conf,
+					 struct list_head *temp_inactive_list,
+					 int hash)
+{
+	int size;
+	bool do_wakeup = false;
+	unsigned long flags;
+
+	if (hash == NR_STRIPE_HASH_LOCKS) {
+		size = NR_STRIPE_HASH_LOCKS;
+		hash = NR_STRIPE_HASH_LOCKS - 1;
+	} else
+		size = 1;
+	while (size) {
+		struct list_head *list = &temp_inactive_list[size - 1];
+
+		/*
+		 * We don't hold any lock here yet, get_active_stripe() might
+		 * remove stripes from the list
+		 */
+		if (!list_empty_careful(list)) {
+			spin_lock_irqsave(conf->hash_locks + hash, flags);
+			if (list_empty(conf->inactive_list + hash) &&
+			    !list_empty(list))
+				atomic_dec(&conf->empty_inactive_list_nr);
+			list_splice_tail_init(list, conf->inactive_list + hash);
+			do_wakeup = true;
+			spin_unlock_irqrestore(conf->hash_locks + hash, flags);
+		}
+		size--;
+		hash--;
+	}
+
+	if (do_wakeup) {
+		wake_up(&conf->wait_for_stripe);
+		if (conf->retry_read_aligned)
+			md_wakeup_thread(conf->mddev->thread);
+	}
+}
+
+/* should hold conf->device_lock already */
+static int release_stripe_list(struct r5conf *conf,
+			       struct list_head *temp_inactive_list)
+{
+	struct stripe_head *sh;
+	int count = 0;
+	struct llist_node *head;
+
+	head = llist_del_all(&conf->released_stripes);
+	head = llist_reverse_order(head);
+	while (head) {
+		int hash;
+
+		sh = llist_entry(head, struct stripe_head, release_list);
+		head = llist_next(head);
+		/* sh could be readded after STRIPE_ON_RELEASE_LIST is cleard */
+		smp_mb();
+		clear_bit(STRIPE_ON_RELEASE_LIST, &sh->state);
+		/*
+		 * Don't worry the bit is set here, because if the bit is set
+		 * again, the count is always > 1. This is true for
+		 * STRIPE_ON_UNPLUG_LIST bit too.
+		 */
+		hash = sh->hash_lock_index;
+		__release_stripe(conf, sh, &temp_inactive_list[hash]);
+		count++;
 	}
+
+	return count;
 }
+
 static void release_stripe(struct stripe_head *sh)
 {
-	raid5_conf_t *conf = sh->raid_conf;
+	struct r5conf *conf = sh->raid_conf;
 	unsigned long flags;
+	struct list_head list;
+	int hash;
+	bool wakeup;
 
-	spin_lock_irqsave(&conf->device_lock, flags);
-	__release_stripe(conf, sh);
-	spin_unlock_irqrestore(&conf->device_lock, flags);
+	/* Avoid release_list until the last reference.
+	 */
+	if (atomic_add_unless(&sh->count, -1, 1))
+		return;
+
+	if (unlikely(!conf->mddev->thread) ||
+		test_and_set_bit(STRIPE_ON_RELEASE_LIST, &sh->state))
+		goto slow_path;
+	wakeup = llist_add(&sh->release_list, &conf->released_stripes);
+	if (wakeup)
+		md_wakeup_thread(conf->mddev->thread);
+	return;
+slow_path:
+	local_irq_save(flags);
+	/* we are ok here if STRIPE_ON_RELEASE_LIST is set or not */
+	if (atomic_dec_and_lock(&sh->count, &conf->device_lock)) {
+		INIT_LIST_HEAD(&list);
+		hash = sh->hash_lock_index;
+		do_release_stripe(conf, sh, &list);
+		spin_unlock(&conf->device_lock);
+		release_inactive_stripe_list(conf, &list, hash);
+	}
+	local_irq_restore(flags);
 }
 
 static inline void remove_hash(struct stripe_head *sh)
@@ -182,42 +449,45 @@ static inline void remove_hash(struct stripe_head *sh)
 	hlist_del_init(&sh->hash);
 }
 
-static inline void insert_hash(raid5_conf_t *conf, struct stripe_head *sh)
+static inline void insert_hash(struct r5conf *conf, struct stripe_head *sh)
 {
 	struct hlist_head *hp = stripe_hash(conf, sh->sector);
 
 	pr_debug("insert_hash(), stripe %llu\n",
 		(unsigned long long)sh->sector);
 
-	CHECK_DEVLOCK();
 	hlist_add_head(&sh->hash, hp);
 }
 
 
 /* find an idle stripe, make sure it is unhashed, and return it. */
-static struct stripe_head *get_free_stripe(raid5_conf_t *conf)
+static struct stripe_head *get_free_stripe(struct r5conf *conf, int hash)
 {
 	struct stripe_head *sh = NULL;
 	struct list_head *first;
 
-	CHECK_DEVLOCK();
-	if (list_empty(&conf->inactive_list))
+	if (list_empty(conf->inactive_list + hash))
 		goto out;
-	first = conf->inactive_list.next;
+	first = (conf->inactive_list + hash)->next;
 	sh = list_entry(first, struct stripe_head, lru);
 	list_del_init(first);
 	remove_hash(sh);
 	atomic_inc(&conf->active_stripes);
+	BUG_ON(hash != sh->hash_lock_index);
+	if (list_empty(conf->inactive_list + hash))
+		atomic_inc(&conf->empty_inactive_list_nr);
 out:
 	return sh;
 }
 
-static void shrink_buffers(struct stripe_head *sh, int num)
+static void shrink_buffers(struct stripe_head *sh)
 {
 	struct page *p;
 	int i;
+	int num = sh->raid_conf->pool_size;
 
-	for (i=0; i<num ; i++) {
+	for (i = 0; i < num ; i++) {
+		WARN_ON(sh->dev[i].page != sh->dev[i].orig_page);
 		p = sh->dev[i].page;
 		if (!p)
 			continue;
@@ -226,43 +496,48 @@ static void shrink_buffers(struct stripe_head *sh, int num)
 	}
 }
 
-static int grow_buffers(struct stripe_head *sh, int num)
+static int grow_buffers(struct stripe_head *sh)
 {
 	int i;
+	int num = sh->raid_conf->pool_size;
 
-	for (i=0; i<num; i++) {
+	for (i = 0; i < num; i++) {
 		struct page *page;
 
 		if (!(page = alloc_page(GFP_KERNEL))) {
 			return 1;
 		}
 		sh->dev[i].page = page;
+		sh->dev[i].orig_page = page;
 	}
 	return 0;
 }
 
-static void raid5_build_block (struct stripe_head *sh, int i);
+static void raid5_build_block(struct stripe_head *sh, int i, int previous);
+static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
+			    struct stripe_head *sh);
 
-static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int disks)
+static void init_stripe(struct stripe_head *sh, sector_t sector, int previous)
 {
-	raid5_conf_t *conf = sh->raid_conf;
-	int i;
+	struct r5conf *conf = sh->raid_conf;
+	int i, seq;
 
 	BUG_ON(atomic_read(&sh->count) != 0);
 	BUG_ON(test_bit(STRIPE_HANDLE, &sh->state));
 	BUG_ON(stripe_operations_active(sh));
 
-	CHECK_DEVLOCK();
 	pr_debug("init_stripe called, stripe %llu\n",
 		(unsigned long long)sh->sector);
 
 	remove_hash(sh);
-
+retry:
+	seq = read_seqcount_begin(&conf->gen_lock);
+	sh->generation = conf->generation - previous;
+	sh->disks = previous ? conf->previous_raid_disks : conf->raid_disks;
 	sh->sector = sector;
-	sh->pd_idx = pd_idx;
+	stripe_set_idx(sector, conf, previous, sh);
 	sh->state = 0;
 
-	sh->disks = disks;
 
 	for (i = sh->disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
@@ -273,84 +548,190 @@ static void init_stripe(struct stripe_head *sh, sector_t sector, int pd_idx, int
 			       (unsigned long long)sh->sector, i, dev->toread,
 			       dev->read, dev->towrite, dev->written,
 			       test_bit(R5_LOCKED, &dev->flags));
-			BUG();
+			WARN_ON(1);
 		}
 		dev->flags = 0;
-		raid5_build_block(sh, i);
+		raid5_build_block(sh, i, previous);
 	}
+	if (read_seqcount_retry(&conf->gen_lock, seq))
+		goto retry;
 	insert_hash(conf, sh);
+	sh->cpu = smp_processor_id();
 }
 
-static struct stripe_head *__find_stripe(raid5_conf_t *conf, sector_t sector, int disks)
+static struct stripe_head *__find_stripe(struct r5conf *conf, sector_t sector,
+					 short generation)
 {
 	struct stripe_head *sh;
-	struct hlist_node *hn;
 
-	CHECK_DEVLOCK();
 	pr_debug("__find_stripe, sector %llu\n", (unsigned long long)sector);
-	hlist_for_each_entry(sh, hn, stripe_hash(conf, sector), hash)
-		if (sh->sector == sector && sh->disks == disks)
+	hlist_for_each_entry(sh, stripe_hash(conf, sector), hash)
+		if (sh->sector == sector && sh->generation == generation)
 			return sh;
 	pr_debug("__stripe %llu not in cache\n", (unsigned long long)sector);
 	return NULL;
 }
 
-static void unplug_slaves(mddev_t *mddev);
-static void raid5_unplug_device(struct request_queue *q);
+/*
+ * Need to check if array has failed when deciding whether to:
+ *  - start an array
+ *  - remove non-faulty devices
+ *  - add a spare
+ *  - allow a reshape
+ * This determination is simple when no reshape is happening.
+ * However if there is a reshape, we need to carefully check
+ * both the before and after sections.
+ * This is because some failed devices may only affect one
+ * of the two sections, and some non-in_sync devices may
+ * be insync in the section most affected by failed devices.
+ */
+static int calc_degraded(struct r5conf *conf)
+{
+	int degraded, degraded2;
+	int i;
+
+	rcu_read_lock();
+	degraded = 0;
+	for (i = 0; i < conf->previous_raid_disks; i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
+		if (rdev && test_bit(Faulty, &rdev->flags))
+			rdev = rcu_dereference(conf->disks[i].replacement);
+		if (!rdev || test_bit(Faulty, &rdev->flags))
+			degraded++;
+		else if (test_bit(In_sync, &rdev->flags))
+			;
+		else
+			/* not in-sync or faulty.
+			 * If the reshape increases the number of devices,
+			 * this is being recovered by the reshape, so
+			 * this 'previous' section is not in_sync.
+			 * If the number of devices is being reduced however,
+			 * the device can only be part of the array if
+			 * we are reverting a reshape, so this section will
+			 * be in-sync.
+			 */
+			if (conf->raid_disks >= conf->previous_raid_disks)
+				degraded++;
+	}
+	rcu_read_unlock();
+	if (conf->raid_disks == conf->previous_raid_disks)
+		return degraded;
+	rcu_read_lock();
+	degraded2 = 0;
+	for (i = 0; i < conf->raid_disks; i++) {
+		struct md_rdev *rdev = rcu_dereference(conf->disks[i].rdev);
+		if (rdev && test_bit(Faulty, &rdev->flags))
+			rdev = rcu_dereference(conf->disks[i].replacement);
+		if (!rdev || test_bit(Faulty, &rdev->flags))
+			degraded2++;
+		else if (test_bit(In_sync, &rdev->flags))
+			;
+		else
+			/* not in-sync or faulty.
+			 * If reshape increases the number of devices, this
+			 * section has already been recovered, else it
+			 * almost certainly hasn't.
+			 */
+			if (conf->raid_disks <= conf->previous_raid_disks)
+				degraded2++;
+	}
+	rcu_read_unlock();
+	if (degraded2 > degraded)
+		return degraded2;
+	return degraded;
+}
+
+static int has_failed(struct r5conf *conf)
+{
+	int degraded;
+
+	if (conf->mddev->reshape_position == MaxSector)
+		return conf->mddev->degraded > conf->max_degraded;
+
+	degraded = calc_degraded(conf);
+	if (degraded > conf->max_degraded)
+		return 1;
+	return 0;
+}
 
-static struct stripe_head *get_active_stripe(raid5_conf_t *conf, sector_t sector, int disks,
-					     int pd_idx, int noblock)
+static struct stripe_head *
+get_active_stripe(struct r5conf *conf, sector_t sector,
+		  int previous, int noblock, int noquiesce)
 {
 	struct stripe_head *sh;
+	int hash = stripe_hash_locks_hash(sector);
 
 	pr_debug("get_stripe, sector %llu\n", (unsigned long long)sector);
 
-	spin_lock_irq(&conf->device_lock);
+	spin_lock_irq(conf->hash_locks + hash);
 
 	do {
 		wait_event_lock_irq(conf->wait_for_stripe,
-				    conf->quiesce == 0,
-				    conf->device_lock, /* nothing */);
-		sh = __find_stripe(conf, sector, disks);
+				    conf->quiesce == 0 || noquiesce,
+				    *(conf->hash_locks + hash));
+		sh = __find_stripe(conf, sector, conf->generation - previous);
 		if (!sh) {
 			if (!conf->inactive_blocked)
-				sh = get_free_stripe(conf);
+				sh = get_free_stripe(conf, hash);
 			if (noblock && sh == NULL)
 				break;
 			if (!sh) {
 				conf->inactive_blocked = 1;
-				wait_event_lock_irq(conf->wait_for_stripe,
-						    !list_empty(&conf->inactive_list) &&
-						    (atomic_read(&conf->active_stripes)
-						     < (conf->max_nr_stripes *3/4)
-						     || !conf->inactive_blocked),
-						    conf->device_lock,
-						    raid5_unplug_device(conf->mddev->queue)
-					);
+				wait_event_lock_irq(
+					conf->wait_for_stripe,
+					!list_empty(conf->inactive_list + hash) &&
+					(atomic_read(&conf->active_stripes)
+					 < (conf->max_nr_stripes * 3 / 4)
+					 || !conf->inactive_blocked),
+					*(conf->hash_locks + hash));
 				conf->inactive_blocked = 0;
-			} else
-				init_stripe(sh, sector, pd_idx, disks);
-		} else {
-			if (atomic_read(&sh->count)) {
-			  BUG_ON(!list_empty(&sh->lru));
 			} else {
+				init_stripe(sh, sector, previous);
+				atomic_inc(&sh->count);
+			}
+		} else if (!atomic_inc_not_zero(&sh->count)) {
+			spin_lock(&conf->device_lock);
+			if (!atomic_read(&sh->count)) {
 				if (!test_bit(STRIPE_HANDLE, &sh->state))
 					atomic_inc(&conf->active_stripes);
-				if (list_empty(&sh->lru) &&
-				    !test_bit(STRIPE_EXPANDING, &sh->state))
-					BUG();
+				BUG_ON(list_empty(&sh->lru) &&
+				       !test_bit(STRIPE_EXPANDING, &sh->state));
 				list_del_init(&sh->lru);
+				if (sh->group) {
+					sh->group->stripes_cnt--;
+					sh->group = NULL;
+				}
 			}
+			atomic_inc(&sh->count);
+			spin_unlock(&conf->device_lock);
 		}
 	} while (sh == NULL);
 
-	if (sh)
-		atomic_inc(&sh->count);
-
-	spin_unlock_irq(&conf->device_lock);
+	spin_unlock_irq(conf->hash_locks + hash);
 	return sh;
 }
 
+/* Determine if 'data_offset' or 'new_data_offset' should be used
+ * in this stripe_head.
+ */
+static int use_new_offset(struct r5conf *conf, struct stripe_head *sh)
+{
+	sector_t progress = conf->reshape_progress;
+	/* Need a memory barrier to make sure we see the value
+	 * of conf->generation, or ->data_offset that was set before
+	 * reshape_progress was updated.
+	 */
+	smp_rmb();
+	if (progress == MaxSector)
+		return 0;
+	if (sh->generation == conf->generation - 1)
+		return 0;
+	/* We are in a reshape, and this is a new-generation stripe,
+	 * so use new_data_offset.
+	 */
+	return 1;
+}
+
 static void
 raid5_end_read_request(struct bio *bi, int error);
 static void
@@ -358,66 +739,200 @@ raid5_end_write_request(struct bio *bi, int error);
 
 static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 {
-	raid5_conf_t *conf = sh->raid_conf;
+	struct r5conf *conf = sh->raid_conf;
 	int i, disks = sh->disks;
 
 	might_sleep();
 
 	for (i = disks; i--; ) {
 		int rw;
-		struct bio *bi;
-		mdk_rdev_t *rdev;
-		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags))
-			rw = WRITE;
-		else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
+		int replace_only = 0;
+		struct bio *bi, *rbi;
+		struct md_rdev *rdev, *rrdev = NULL;
+		if (test_and_clear_bit(R5_Wantwrite, &sh->dev[i].flags)) {
+			if (test_and_clear_bit(R5_WantFUA, &sh->dev[i].flags))
+				rw = WRITE_FUA;
+			else
+				rw = WRITE;
+			if (test_bit(R5_Discard, &sh->dev[i].flags))
+				rw |= REQ_DISCARD;
+		} else if (test_and_clear_bit(R5_Wantread, &sh->dev[i].flags))
 			rw = READ;
-		else
+		else if (test_and_clear_bit(R5_WantReplace,
+					    &sh->dev[i].flags)) {
+			rw = WRITE;
+			replace_only = 1;
+		} else
 			continue;
+		if (test_and_clear_bit(R5_SyncIO, &sh->dev[i].flags))
+			rw |= REQ_SYNC;
 
 		bi = &sh->dev[i].req;
-
-		bi->bi_rw = rw;
-		if (rw == WRITE)
-			bi->bi_end_io = raid5_end_write_request;
-		else
-			bi->bi_end_io = raid5_end_read_request;
+		rbi = &sh->dev[i].rreq; /* For writing to replacement */
 
 		rcu_read_lock();
+		rrdev = rcu_dereference(conf->disks[i].replacement);
+		smp_mb(); /* Ensure that if rrdev is NULL, rdev won't be */
 		rdev = rcu_dereference(conf->disks[i].rdev);
+		if (!rdev) {
+			rdev = rrdev;
+			rrdev = NULL;
+		}
+		if (rw & WRITE) {
+			if (replace_only)
+				rdev = NULL;
+			if (rdev == rrdev)
+				/* We raced and saw duplicates */
+				rrdev = NULL;
+		} else {
+			if (test_bit(R5_ReadRepl, &sh->dev[i].flags) && rrdev)
+				rdev = rrdev;
+			rrdev = NULL;
+		}
+
 		if (rdev && test_bit(Faulty, &rdev->flags))
 			rdev = NULL;
 		if (rdev)
 			atomic_inc(&rdev->nr_pending);
+		if (rrdev && test_bit(Faulty, &rrdev->flags))
+			rrdev = NULL;
+		if (rrdev)
+			atomic_inc(&rrdev->nr_pending);
 		rcu_read_unlock();
 
+		/* We have already checked bad blocks for reads.  Now
+		 * need to check for writes.  We never accept write errors
+		 * on the replacement, so we don't to check rrdev.
+		 */
+		while ((rw & WRITE) && rdev &&
+		       test_bit(WriteErrorSeen, &rdev->flags)) {
+			sector_t first_bad;
+			int bad_sectors;
+			int bad = is_badblock(rdev, sh->sector, STRIPE_SECTORS,
+					      &first_bad, &bad_sectors);
+			if (!bad)
+				break;
+
+			if (bad < 0) {
+				set_bit(BlockedBadBlocks, &rdev->flags);
+				if (!conf->mddev->external &&
+				    conf->mddev->flags) {
+					/* It is very unlikely, but we might
+					 * still need to write out the
+					 * bad block log - better give it
+					 * 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 */
+				rdev_dec_pending(rdev, conf->mddev);
+				rdev = NULL;
+			}
+		}
+
 		if (rdev) {
-			if (s->syncing || s->expanding || s->expanded)
+			if (s->syncing || s->expanding || s->expanded
+			    || s->replacing)
 				md_sync_acct(rdev->bdev, STRIPE_SECTORS);
 
 			set_bit(STRIPE_IO_STARTED, &sh->state);
 
+			bio_reset(bi);
 			bi->bi_bdev = rdev->bdev;
+			bi->bi_rw = rw;
+			bi->bi_end_io = (rw & WRITE)
+				? raid5_end_write_request
+				: raid5_end_read_request;
+			bi->bi_private = sh;
+
 			pr_debug("%s: for %llu schedule op %ld on disc %d\n",
 				__func__, (unsigned long long)sh->sector,
 				bi->bi_rw, i);
 			atomic_inc(&sh->count);
-			bi->bi_sector = sh->sector + rdev->data_offset;
-			bi->bi_flags = 1 << BIO_UPTODATE;
+			if (use_new_offset(conf, sh))
+				bi->bi_iter.bi_sector = (sh->sector
+						 + rdev->new_data_offset);
+			else
+				bi->bi_iter.bi_sector = (sh->sector
+						 + rdev->data_offset);
+			if (test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+				bi->bi_rw |= REQ_NOMERGE;
+
+			if (test_bit(R5_SkipCopy, &sh->dev[i].flags))
+				WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
+			sh->dev[i].vec.bv_page = sh->dev[i].page;
 			bi->bi_vcnt = 1;
-			bi->bi_max_vecs = 1;
-			bi->bi_idx = 0;
-			bi->bi_io_vec = &sh->dev[i].vec;
 			bi->bi_io_vec[0].bv_len = STRIPE_SIZE;
 			bi->bi_io_vec[0].bv_offset = 0;
-			bi->bi_size = STRIPE_SIZE;
-			bi->bi_next = NULL;
-			if (rw == WRITE &&
-			    test_bit(R5_ReWrite, &sh->dev[i].flags))
-				atomic_add(STRIPE_SECTORS,
-					&rdev->corrected_errors);
+			bi->bi_iter.bi_size = STRIPE_SIZE;
+			/*
+			 * If this is discard request, set bi_vcnt 0. We don't
+			 * want to confuse SCSI because SCSI will replace payload
+			 */
+			if (rw & REQ_DISCARD)
+				bi->bi_vcnt = 0;
+			if (rrdev)
+				set_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags);
+
+			if (conf->mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(bi->bi_bdev),
+						      bi, disk_devt(conf->mddev->gendisk),
+						      sh->dev[i].sector);
 			generic_make_request(bi);
-		} else {
-			if (rw == WRITE)
+		}
+		if (rrdev) {
+			if (s->syncing || s->expanding || s->expanded
+			    || s->replacing)
+				md_sync_acct(rrdev->bdev, STRIPE_SECTORS);
+
+			set_bit(STRIPE_IO_STARTED, &sh->state);
+
+			bio_reset(rbi);
+			rbi->bi_bdev = rrdev->bdev;
+			rbi->bi_rw = rw;
+			BUG_ON(!(rw & WRITE));
+			rbi->bi_end_io = raid5_end_write_request;
+			rbi->bi_private = sh;
+
+			pr_debug("%s: for %llu schedule op %ld on "
+				 "replacement disc %d\n",
+				__func__, (unsigned long long)sh->sector,
+				rbi->bi_rw, i);
+			atomic_inc(&sh->count);
+			if (use_new_offset(conf, sh))
+				rbi->bi_iter.bi_sector = (sh->sector
+						  + rrdev->new_data_offset);
+			else
+				rbi->bi_iter.bi_sector = (sh->sector
+						  + rrdev->data_offset);
+			if (test_bit(R5_SkipCopy, &sh->dev[i].flags))
+				WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
+			sh->dev[i].rvec.bv_page = sh->dev[i].page;
+			rbi->bi_vcnt = 1;
+			rbi->bi_io_vec[0].bv_len = STRIPE_SIZE;
+			rbi->bi_io_vec[0].bv_offset = 0;
+			rbi->bi_iter.bi_size = STRIPE_SIZE;
+			/*
+			 * If this is discard request, set bi_vcnt 0. We don't
+			 * want to confuse SCSI because SCSI will replace payload
+			 */
+			if (rw & REQ_DISCARD)
+				rbi->bi_vcnt = 0;
+			if (conf->mddev->gendisk)
+				trace_block_bio_remap(bdev_get_queue(rbi->bi_bdev),
+						      rbi, disk_devt(conf->mddev->gendisk),
+						      sh->dev[i].sector);
+			generic_make_request(rbi);
+		}
+		if (!rdev && !rrdev) {
+			if (rw & WRITE)
 				set_bit(STRIPE_DEGRADED, &sh->state);
 			pr_debug("skip op %ld on disc %d for sector %llu\n",
 				bi->bi_rw, i, (unsigned long long)sh->sector);
@@ -428,20 +943,28 @@ static void ops_run_io(struct stripe_head *sh, struct stripe_head_state *s)
 }
 
 static struct dma_async_tx_descriptor *
-async_copy_data(int frombio, struct bio *bio, struct page *page,
-	sector_t sector, struct dma_async_tx_descriptor *tx)
+async_copy_data(int frombio, struct bio *bio, struct page **page,
+	sector_t sector, struct dma_async_tx_descriptor *tx,
+	struct stripe_head *sh)
 {
-	struct bio_vec *bvl;
+	struct bio_vec bvl;
+	struct bvec_iter iter;
 	struct page *bio_page;
-	int i;
 	int page_offset;
+	struct async_submit_ctl submit;
+	enum async_tx_flags flags = 0;
 
-	if (bio->bi_sector >= sector)
-		page_offset = (signed)(bio->bi_sector - sector) * 512;
+	if (bio->bi_iter.bi_sector >= sector)
+		page_offset = (signed)(bio->bi_iter.bi_sector - sector) * 512;
 	else
-		page_offset = (signed)(sector - bio->bi_sector) * -512;
-	bio_for_each_segment(bvl, bio, i) {
-		int len = bio_iovec_idx(bio, i)->bv_len;
+		page_offset = (signed)(sector - bio->bi_iter.bi_sector) * -512;
+
+	if (frombio)
+		flags |= ASYNC_TX_FENCE;
+	init_async_submit(&submit, flags, tx, NULL, NULL, NULL);
+
+	bio_for_each_segment(bvl, bio, iter) {
+		int len = bvl.bv_len;
 		int clen;
 		int b_offset = 0;
 
@@ -457,19 +980,23 @@ async_copy_data(int frombio, struct bio *bio, struct page *page,
 			clen = len;
 
 		if (clen > 0) {
-			b_offset += bio_iovec_idx(bio, i)->bv_offset;
-			bio_page = bio_iovec_idx(bio, i)->bv_page;
-			if (frombio)
-				tx = async_memcpy(page, bio_page, page_offset,
-					b_offset, clen,
-					ASYNC_TX_DEP_ACK,
-					tx, NULL, NULL);
-			else
-				tx = async_memcpy(bio_page, page, b_offset,
-					page_offset, clen,
-					ASYNC_TX_DEP_ACK,
-					tx, NULL, NULL);
+			b_offset += bvl.bv_offset;
+			bio_page = bvl.bv_page;
+			if (frombio) {
+				if (sh->raid_conf->skip_copy &&
+				    b_offset == 0 && page_offset == 0 &&
+				    clen == STRIPE_SIZE)
+					*page = bio_page;
+				else
+					tx = async_memcpy(*page, bio_page, page_offset,
+						  b_offset, clen, &submit);
+			} else
+				tx = async_memcpy(bio_page, *page, b_offset,
+						  page_offset, clen, &submit);
 		}
+		/* chain the operations */
+		submit.depend_tx = tx;
+
 		if (clen < len) /* hit end of page */
 			break;
 		page_offset +=  len;
@@ -482,14 +1009,12 @@ static void ops_complete_biofill(void *stripe_head_ref)
 {
 	struct stripe_head *sh = stripe_head_ref;
 	struct bio *return_bi = NULL;
-	raid5_conf_t *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];
 
@@ -504,10 +1029,10 @@ static void ops_complete_biofill(void *stripe_head_ref)
 			BUG_ON(!dev->read);
 			rbi = dev->read;
 			dev->read = NULL;
-			while (rbi && rbi->bi_sector <
+			while (rbi && rbi->bi_iter.bi_sector <
 				dev->sector + STRIPE_SECTORS) {
 				rbi2 = r5_next_bio(rbi, dev->sector);
-				if (--rbi->bi_phys_segments == 0) {
+				if (!raid5_dec_bi_active_stripes(rbi)) {
 					rbi->bi_next = return_bi;
 					return_bi = rbi;
 				}
@@ -515,7 +1040,6 @@ 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);
@@ -527,7 +1051,7 @@ 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;
-	raid5_conf_t *conf = sh->raid_conf;
+	struct async_submit_ctl submit;
 	int i;
 
 	pr_debug("%s: stripe %llu\n", __func__,
@@ -537,36 +1061,48 @@ 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);
-			while (rbi && rbi->bi_sector <
+			spin_unlock_irq(&sh->stripe_lock);
+			while (rbi && rbi->bi_iter.bi_sector <
 				dev->sector + STRIPE_SECTORS) {
-				tx = async_copy_data(0, rbi, dev->page,
-					dev->sector, tx);
+				tx = async_copy_data(0, rbi, &dev->page,
+					dev->sector, tx, sh);
 				rbi = r5_next_bio(rbi, dev->sector);
 			}
 		}
 	}
 
 	atomic_inc(&sh->count);
-	async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx,
-		ops_complete_biofill, sh);
+	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_biofill, sh, NULL);
+	async_trigger_callback(&submit);
 }
 
-static void ops_complete_compute5(void *stripe_head_ref)
+static void mark_target_uptodate(struct stripe_head *sh, int target)
 {
-	struct stripe_head *sh = stripe_head_ref;
-	int target = sh->ops.target;
-	struct r5dev *tgt = &sh->dev[target];
+	struct r5dev *tgt;
 
-	pr_debug("%s: stripe %llu\n", __func__,
-		(unsigned long long)sh->sector);
+	if (target < 0)
+		return;
 
+	tgt = &sh->dev[target];
 	set_bit(R5_UPTODATE, &tgt->flags);
 	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
 	clear_bit(R5_Wantcompute, &tgt->flags);
+}
+
+static void ops_complete_compute(void *stripe_head_ref)
+{
+	struct stripe_head *sh = stripe_head_ref;
+
+	pr_debug("%s: stripe %llu\n", __func__,
+		(unsigned long long)sh->sector);
+
+	/* mark the computed target(s) as uptodate */
+	mark_target_uptodate(sh, sh->ops.target);
+	mark_target_uptodate(sh, sh->ops.target2);
+
 	clear_bit(STRIPE_COMPUTE_RUN, &sh->state);
 	if (sh->check_state == check_state_compute_run)
 		sh->check_state = check_state_compute_result;
@@ -574,16 +1110,24 @@ static void ops_complete_compute5(void *stripe_head_ref)
 	release_stripe(sh);
 }
 
-static struct dma_async_tx_descriptor *ops_run_compute5(struct stripe_head *sh)
+/* return a pointer to the address conversion region of the scribble buffer */
+static addr_conv_t *to_addr_conv(struct stripe_head *sh,
+				 struct raid5_percpu *percpu)
+{
+	return percpu->scribble + sizeof(struct page *) * (sh->disks + 2);
+}
+
+static struct dma_async_tx_descriptor *
+ops_run_compute5(struct stripe_head *sh, struct raid5_percpu *percpu)
 {
-	/* kernel stack size limits the total number of disks */
 	int disks = sh->disks;
-	struct page *xor_srcs[disks];
+	struct page **xor_srcs = percpu->scribble;
 	int target = sh->ops.target;
 	struct r5dev *tgt = &sh->dev[target];
 	struct page *xor_dest = tgt->page;
 	int count = 0;
 	struct dma_async_tx_descriptor *tx;
+	struct async_submit_ctl submit;
 	int i;
 
 	pr_debug("%s: stripe %llu block: %d\n",
@@ -596,17 +1140,213 @@ static struct dma_async_tx_descriptor *ops_run_compute5(struct stripe_head *sh)
 
 	atomic_inc(&sh->count);
 
+	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST, NULL,
+			  ops_complete_compute, sh, to_addr_conv(sh, percpu));
 	if (unlikely(count == 1))
-		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE,
-			0, NULL, ops_complete_compute5, sh);
+		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
 	else
-		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
-			ASYNC_TX_XOR_ZERO_DST, NULL,
-			ops_complete_compute5, sh);
+		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
 
 	return tx;
 }
 
+/* set_syndrome_sources - populate source buffers for gen_syndrome
+ * @srcs - (struct page *) array of size sh->disks
+ * @sh - stripe_head to parse
+ *
+ * Populates srcs in proper layout order for the stripe and returns the
+ * 'count' of sources to be used in a call to async_gen_syndrome.  The P
+ * destination buffer is recorded in srcs[count] and the Q destination
+ * is recorded in srcs[count+1]].
+ */
+static int set_syndrome_sources(struct page **srcs, struct stripe_head *sh)
+{
+	int disks = sh->disks;
+	int syndrome_disks = sh->ddf_layout ? disks : (disks - 2);
+	int d0_idx = raid6_d0(sh);
+	int count;
+	int i;
+
+	for (i = 0; i < disks; i++)
+		srcs[i] = NULL;
+
+	count = 0;
+	i = d0_idx;
+	do {
+		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);
+
+		srcs[slot] = sh->dev[i].page;
+		i = raid6_next_disk(i, disks);
+	} while (i != d0_idx);
+
+	return syndrome_disks;
+}
+
+static struct dma_async_tx_descriptor *
+ops_run_compute6_1(struct stripe_head *sh, struct raid5_percpu *percpu)
+{
+	int disks = sh->disks;
+	struct page **blocks = percpu->scribble;
+	int target;
+	int qd_idx = sh->qd_idx;
+	struct dma_async_tx_descriptor *tx;
+	struct async_submit_ctl submit;
+	struct r5dev *tgt;
+	struct page *dest;
+	int i;
+	int count;
+
+	if (sh->ops.target < 0)
+		target = sh->ops.target2;
+	else if (sh->ops.target2 < 0)
+		target = sh->ops.target;
+	else
+		/* we should only have one valid target */
+		BUG();
+	BUG_ON(target < 0);
+	pr_debug("%s: stripe %llu block: %d\n",
+		__func__, (unsigned long long)sh->sector, target);
+
+	tgt = &sh->dev[target];
+	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
+	dest = tgt->page;
+
+	atomic_inc(&sh->count);
+
+	if (target == qd_idx) {
+		count = set_syndrome_sources(blocks, sh);
+		blocks[count] = NULL; /* regenerating p is not necessary */
+		BUG_ON(blocks[count+1] != dest); /* q should already be set */
+		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
+				  ops_complete_compute, sh,
+				  to_addr_conv(sh, percpu));
+		tx = async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE, &submit);
+	} else {
+		/* Compute any data- or p-drive using XOR */
+		count = 0;
+		for (i = disks; i-- ; ) {
+			if (i == target || i == qd_idx)
+				continue;
+			blocks[count++] = sh->dev[i].page;
+		}
+
+		init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
+				  NULL, ops_complete_compute, sh,
+				  to_addr_conv(sh, percpu));
+		tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE, &submit);
+	}
+
+	return tx;
+}
+
+static struct dma_async_tx_descriptor *
+ops_run_compute6_2(struct stripe_head *sh, struct raid5_percpu *percpu)
+{
+	int i, count, disks = sh->disks;
+	int syndrome_disks = sh->ddf_layout ? disks : disks-2;
+	int d0_idx = raid6_d0(sh);
+	int faila = -1, failb = -1;
+	int target = sh->ops.target;
+	int target2 = sh->ops.target2;
+	struct r5dev *tgt = &sh->dev[target];
+	struct r5dev *tgt2 = &sh->dev[target2];
+	struct dma_async_tx_descriptor *tx;
+	struct page **blocks = percpu->scribble;
+	struct async_submit_ctl submit;
+
+	pr_debug("%s: stripe %llu block1: %d block2: %d\n",
+		 __func__, (unsigned long long)sh->sector, target, target2);
+	BUG_ON(target < 0 || target2 < 0);
+	BUG_ON(!test_bit(R5_Wantcompute, &tgt->flags));
+	BUG_ON(!test_bit(R5_Wantcompute, &tgt2->flags));
+
+	/* we need to open-code set_syndrome_sources to handle the
+	 * slot number conversion for 'faila' and 'failb'
+	 */
+	for (i = 0; i < disks ; i++)
+		blocks[i] = NULL;
+	count = 0;
+	i = d0_idx;
+	do {
+		int slot = raid6_idx_to_slot(i, sh, &count, syndrome_disks);
+
+		blocks[slot] = sh->dev[i].page;
+
+		if (i == target)
+			faila = slot;
+		if (i == target2)
+			failb = slot;
+		i = raid6_next_disk(i, disks);
+	} while (i != d0_idx);
+
+	BUG_ON(faila == failb);
+	if (failb < faila)
+		swap(faila, failb);
+	pr_debug("%s: stripe: %llu faila: %d failb: %d\n",
+		 __func__, (unsigned long long)sh->sector, faila, failb);
+
+	atomic_inc(&sh->count);
+
+	if (failb == syndrome_disks+1) {
+		/* Q disk is one of the missing disks */
+		if (faila == syndrome_disks) {
+			/* Missing P+Q, just recompute */
+			init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
+					  ops_complete_compute, sh,
+					  to_addr_conv(sh, percpu));
+			return async_gen_syndrome(blocks, 0, syndrome_disks+2,
+						  STRIPE_SIZE, &submit);
+		} else {
+			struct page *dest;
+			int data_target;
+			int qd_idx = sh->qd_idx;
+
+			/* Missing D+Q: recompute D from P, then recompute Q */
+			if (target == qd_idx)
+				data_target = target2;
+			else
+				data_target = target;
+
+			count = 0;
+			for (i = disks; i-- ; ) {
+				if (i == data_target || i == qd_idx)
+					continue;
+				blocks[count++] = sh->dev[i].page;
+			}
+			dest = sh->dev[data_target].page;
+			init_async_submit(&submit,
+					  ASYNC_TX_FENCE|ASYNC_TX_XOR_ZERO_DST,
+					  NULL, NULL, NULL,
+					  to_addr_conv(sh, percpu));
+			tx = async_xor(dest, blocks, 0, count, STRIPE_SIZE,
+				       &submit);
+
+			count = set_syndrome_sources(blocks, sh);
+			init_async_submit(&submit, ASYNC_TX_FENCE, tx,
+					  ops_complete_compute, sh,
+					  to_addr_conv(sh, percpu));
+			return async_gen_syndrome(blocks, 0, count+2,
+						  STRIPE_SIZE, &submit);
+		}
+	} else {
+		init_async_submit(&submit, ASYNC_TX_FENCE, NULL,
+				  ops_complete_compute, sh,
+				  to_addr_conv(sh, percpu));
+		if (failb == syndrome_disks) {
+			/* We're missing D+P. */
+			return async_raid6_datap_recov(syndrome_disks+2,
+						       STRIPE_SIZE, faila,
+						       blocks, &submit);
+		} else {
+			/* We're missing D+D. */
+			return async_raid6_2data_recov(syndrome_disks+2,
+						       STRIPE_SIZE, faila, failb,
+						       blocks, &submit);
+		}
+	}
+}
+
+
 static void ops_complete_prexor(void *stripe_head_ref)
 {
 	struct stripe_head *sh = stripe_head_ref;
@@ -616,12 +1356,13 @@ static void ops_complete_prexor(void *stripe_head_ref)
 }
 
 static struct dma_async_tx_descriptor *
-ops_run_prexor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
+ops_run_prexor(struct stripe_head *sh, struct raid5_percpu *percpu,
+	       struct dma_async_tx_descriptor *tx)
 {
-	/* kernel stack size limits the total number of disks */
 	int disks = sh->disks;
-	struct page *xor_srcs[disks];
+	struct page **xor_srcs = percpu->scribble;
 	int count = 0, pd_idx = sh->pd_idx, i;
+	struct async_submit_ctl submit;
 
 	/* existing parity data subtracted */
 	struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;
@@ -636,9 +1377,9 @@ ops_run_prexor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 			xor_srcs[count++] = dev->page;
 	}
 
-	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
-		ASYNC_TX_DEP_ACK | ASYNC_TX_XOR_DROP_DST, tx,
-		ops_complete_prexor, sh);
+	init_async_submit(&submit, ASYNC_TX_FENCE|ASYNC_TX_XOR_DROP_DST, tx,
+			  ops_complete_prexor, sh, to_addr_conv(sh, percpu));
+	tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
 
 	return tx;
 }
@@ -659,17 +1400,31 @@ 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(&sh->lock);
+			spin_lock_irq(&sh->stripe_lock);
 			chosen = dev->towrite;
 			dev->towrite = NULL;
 			BUG_ON(dev->written);
 			wbi = dev->written = chosen;
-			spin_unlock(&sh->lock);
+			spin_unlock_irq(&sh->stripe_lock);
+			WARN_ON(dev->page != dev->orig_page);
 
-			while (wbi && wbi->bi_sector <
+			while (wbi && wbi->bi_iter.bi_sector <
 				dev->sector + STRIPE_SECTORS) {
-				tx = async_copy_data(1, wbi, dev->page,
-					dev->sector, tx);
+				if (wbi->bi_rw & REQ_FUA)
+					set_bit(R5_WantFUA, &dev->flags);
+				if (wbi->bi_rw & REQ_SYNC)
+					set_bit(R5_SyncIO, &dev->flags);
+				if (wbi->bi_rw & REQ_DISCARD)
+					set_bit(R5_Discard, &dev->flags);
+				else {
+					tx = async_copy_data(1, wbi, &dev->page,
+						dev->sector, tx, sh);
+					if (dev->page != dev->orig_page) {
+						set_bit(R5_SkipCopy, &dev->flags);
+						clear_bit(R5_UPTODATE, &dev->flags);
+						clear_bit(R5_OVERWRITE, &dev->flags);
+					}
+				}
 				wbi = r5_next_bio(wbi, dev->sector);
 			}
 		}
@@ -678,18 +1433,35 @@ ops_run_biodrain(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 	return tx;
 }
 
-static void ops_complete_postxor(void *stripe_head_ref)
+static void ops_complete_reconstruct(void *stripe_head_ref)
 {
 	struct stripe_head *sh = stripe_head_ref;
-	int disks = sh->disks, i, pd_idx = sh->pd_idx;
+	int disks = sh->disks;
+	int pd_idx = sh->pd_idx;
+	int qd_idx = sh->qd_idx;
+	int i;
+	bool fua = false, sync = false, discard = false;
 
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
 
 	for (i = disks; i--; ) {
+		fua |= test_bit(R5_WantFUA, &sh->dev[i].flags);
+		sync |= test_bit(R5_SyncIO, &sh->dev[i].flags);
+		discard |= test_bit(R5_Discard, &sh->dev[i].flags);
+	}
+
+	for (i = disks; i--; ) {
 		struct r5dev *dev = &sh->dev[i];
-		if (dev->written || i == pd_idx)
-			set_bit(R5_UPTODATE, &dev->flags);
+
+		if (dev->written || i == pd_idx || i == qd_idx) {
+			if (!discard && !test_bit(R5_SkipCopy, &dev->flags))
+				set_bit(R5_UPTODATE, &dev->flags);
+			if (fua)
+				set_bit(R5_WantFUA, &dev->flags);
+			if (sync)
+				set_bit(R5_SyncIO, &dev->flags);
+		}
 	}
 
 	if (sh->reconstruct_state == reconstruct_state_drain_run)
@@ -706,12 +1478,12 @@ static void ops_complete_postxor(void *stripe_head_ref)
 }
 
 static void
-ops_run_postxor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
+ops_run_reconstruct5(struct stripe_head *sh, struct raid5_percpu *percpu,
+		     struct dma_async_tx_descriptor *tx)
 {
-	/* kernel stack size limits the total number of disks */
 	int disks = sh->disks;
-	struct page *xor_srcs[disks];
-
+	struct page **xor_srcs = percpu->scribble;
+	struct async_submit_ctl submit;
 	int count = 0, pd_idx = sh->pd_idx, i;
 	struct page *xor_dest;
 	int prexor = 0;
@@ -720,6 +1492,18 @@ ops_run_postxor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
 
+	for (i = 0; i < sh->disks; i++) {
+		if (pd_idx == i)
+			continue;
+		if (!test_bit(R5_Discard, &sh->dev[i].flags))
+			break;
+	}
+	if (i >= sh->disks) {
+		atomic_inc(&sh->count);
+		set_bit(R5_Discard, &sh->dev[pd_idx].flags);
+		ops_complete_reconstruct(sh);
+		return;
+	}
 	/* check if prexor is active which means only process blocks
 	 * that are part of a read-modify-write (written)
 	 */
@@ -745,18 +1529,50 @@ ops_run_postxor(struct stripe_head *sh, struct dma_async_tx_descriptor *tx)
 	 * set ASYNC_TX_XOR_DROP_DST and ASYNC_TX_XOR_ZERO_DST
 	 * for the synchronous xor case
 	 */
-	flags = ASYNC_TX_DEP_ACK | ASYNC_TX_ACK |
+	flags = ASYNC_TX_ACK |
 		(prexor ? ASYNC_TX_XOR_DROP_DST : ASYNC_TX_XOR_ZERO_DST);
 
 	atomic_inc(&sh->count);
 
-	if (unlikely(count == 1)) {
-		flags &= ~(ASYNC_TX_XOR_DROP_DST | ASYNC_TX_XOR_ZERO_DST);
-		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE,
-			flags, tx, ops_complete_postxor, sh);
-	} else
-		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
-			flags, tx, ops_complete_postxor, sh);
+	init_async_submit(&submit, flags, tx, ops_complete_reconstruct, sh,
+			  to_addr_conv(sh, percpu));
+	if (unlikely(count == 1))
+		tx = async_memcpy(xor_dest, xor_srcs[0], 0, 0, STRIPE_SIZE, &submit);
+	else
+		tx = async_xor(xor_dest, xor_srcs, 0, count, STRIPE_SIZE, &submit);
+}
+
+static void
+ops_run_reconstruct6(struct stripe_head *sh, struct raid5_percpu *percpu,
+		     struct dma_async_tx_descriptor *tx)
+{
+	struct async_submit_ctl submit;
+	struct page **blocks = percpu->scribble;
+	int count, i;
+
+	pr_debug("%s: stripe %llu\n", __func__, (unsigned long long)sh->sector);
+
+	for (i = 0; i < sh->disks; i++) {
+		if (sh->pd_idx == i || sh->qd_idx == i)
+			continue;
+		if (!test_bit(R5_Discard, &sh->dev[i].flags))
+			break;
+	}
+	if (i >= sh->disks) {
+		atomic_inc(&sh->count);
+		set_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
+		set_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
+		ops_complete_reconstruct(sh);
+		return;
+	}
+
+	count = set_syndrome_sources(blocks, sh);
+
+	atomic_inc(&sh->count);
+
+	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_reconstruct,
+			  sh, to_addr_conv(sh, percpu));
+	async_gen_syndrome(blocks, 0, count+2, STRIPE_SIZE,  &submit);
 }
 
 static void ops_complete_check(void *stripe_head_ref)
@@ -771,63 +1587,115 @@ static void ops_complete_check(void *stripe_head_ref)
 	release_stripe(sh);
 }
 
-static void ops_run_check(struct stripe_head *sh)
+static void ops_run_check_p(struct stripe_head *sh, struct raid5_percpu *percpu)
 {
-	/* kernel stack size limits the total number of disks */
 	int disks = sh->disks;
-	struct page *xor_srcs[disks];
+	int pd_idx = sh->pd_idx;
+	int qd_idx = sh->qd_idx;
+	struct page *xor_dest;
+	struct page **xor_srcs = percpu->scribble;
 	struct dma_async_tx_descriptor *tx;
-
-	int count = 0, pd_idx = sh->pd_idx, i;
-	struct page *xor_dest = xor_srcs[count++] = sh->dev[pd_idx].page;
+	struct async_submit_ctl submit;
+	int count;
+	int i;
 
 	pr_debug("%s: stripe %llu\n", __func__,
 		(unsigned long long)sh->sector);
 
+	count = 0;
+	xor_dest = sh->dev[pd_idx].page;
+	xor_srcs[count++] = xor_dest;
 	for (i = disks; i--; ) {
-		struct r5dev *dev = &sh->dev[i];
-		if (i != pd_idx)
-			xor_srcs[count++] = dev->page;
+		if (i == pd_idx || i == qd_idx)
+			continue;
+		xor_srcs[count++] = sh->dev[i].page;
 	}
 
-	tx = async_xor_zero_sum(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
-		&sh->ops.zero_sum_result, 0, NULL, NULL, NULL);
+	init_async_submit(&submit, 0, NULL, NULL, NULL,
+			  to_addr_conv(sh, percpu));
+	tx = async_xor_val(xor_dest, xor_srcs, 0, count, STRIPE_SIZE,
+			   &sh->ops.zero_sum_result, &submit);
 
 	atomic_inc(&sh->count);
-	tx = async_trigger_callback(ASYNC_TX_DEP_ACK | ASYNC_TX_ACK, tx,
-		ops_complete_check, sh);
+	init_async_submit(&submit, ASYNC_TX_ACK, tx, ops_complete_check, sh, NULL);
+	tx = async_trigger_callback(&submit);
 }
 
-static void raid5_run_ops(struct stripe_head *sh, unsigned long ops_request)
+static void ops_run_check_pq(struct stripe_head *sh, struct raid5_percpu *percpu, int checkp)
+{
+	struct page **srcs = percpu->scribble;
+	struct async_submit_ctl submit;
+	int count;
+
+	pr_debug("%s: stripe %llu checkp: %d\n", __func__,
+		(unsigned long long)sh->sector, checkp);
+
+	count = set_syndrome_sources(srcs, sh);
+	if (!checkp)
+		srcs[count] = NULL;
+
+	atomic_inc(&sh->count);
+	init_async_submit(&submit, ASYNC_TX_ACK, NULL, ops_complete_check,
+			  sh, to_addr_conv(sh, percpu));
+	async_syndrome_val(srcs, 0, count+2, STRIPE_SIZE,
+			   &sh->ops.zero_sum_result, percpu->spare_page, &submit);
+}
+
+static void raid_run_ops(struct stripe_head *sh, unsigned long ops_request)
 {
 	int overlap_clear = 0, i, disks = sh->disks;
 	struct dma_async_tx_descriptor *tx = NULL;
+	struct r5conf *conf = sh->raid_conf;
+	int level = conf->level;
+	struct raid5_percpu *percpu;
+	unsigned long cpu;
 
+	cpu = get_cpu();
+	percpu = per_cpu_ptr(conf->percpu, cpu);
 	if (test_bit(STRIPE_OP_BIOFILL, &ops_request)) {
 		ops_run_biofill(sh);
 		overlap_clear++;
 	}
 
 	if (test_bit(STRIPE_OP_COMPUTE_BLK, &ops_request)) {
-		tx = ops_run_compute5(sh);
-		/* terminate the chain if postxor is not set to be run */
-		if (tx && !test_bit(STRIPE_OP_POSTXOR, &ops_request))
+		if (level < 6)
+			tx = ops_run_compute5(sh, percpu);
+		else {
+			if (sh->ops.target2 < 0 || sh->ops.target < 0)
+				tx = ops_run_compute6_1(sh, percpu);
+			else
+				tx = ops_run_compute6_2(sh, percpu);
+		}
+		/* terminate the chain if reconstruct is not set to be run */
+		if (tx && !test_bit(STRIPE_OP_RECONSTRUCT, &ops_request))
 			async_tx_ack(tx);
 	}
 
 	if (test_bit(STRIPE_OP_PREXOR, &ops_request))
-		tx = ops_run_prexor(sh, tx);
+		tx = ops_run_prexor(sh, percpu, tx);
 
 	if (test_bit(STRIPE_OP_BIODRAIN, &ops_request)) {
 		tx = ops_run_biodrain(sh, tx);
 		overlap_clear++;
 	}
 
-	if (test_bit(STRIPE_OP_POSTXOR, &ops_request))
-		ops_run_postxor(sh, tx);
+	if (test_bit(STRIPE_OP_RECONSTRUCT, &ops_request)) {
+		if (level < 6)
+			ops_run_reconstruct5(sh, percpu, tx);
+		else
+			ops_run_reconstruct6(sh, percpu, tx);
+	}
 
-	if (test_bit(STRIPE_OP_CHECK, &ops_request))
-		ops_run_check(sh);
+	if (test_bit(STRIPE_OP_CHECK, &ops_request)) {
+		if (sh->check_state == check_state_run)
+			ops_run_check_p(sh, percpu);
+		else if (sh->check_state == check_state_run_q)
+			ops_run_check_pq(sh, percpu, 0);
+		else if (sh->check_state == check_state_run_pq)
+			ops_run_check_pq(sh, percpu, 1);
+		else
+			BUG();
+	}
 
 	if (overlap_clear)
 		for (i = disks; i--; ) {
@@ -835,24 +1703,26 @@ static void raid5_run_ops(struct stripe_head *sh, unsigned long ops_request)
 			if (test_and_clear_bit(R5_Overlap, &dev->flags))
 				wake_up(&sh->raid_conf->wait_for_overlap);
 		}
+	put_cpu();
 }
 
-static int grow_one_stripe(raid5_conf_t *conf)
+static int grow_one_stripe(struct r5conf *conf, int hash)
 {
 	struct stripe_head *sh;
-	sh = kmem_cache_alloc(conf->slab_cache, GFP_KERNEL);
+	sh = kmem_cache_zalloc(conf->slab_cache, GFP_KERNEL);
 	if (!sh)
 		return 0;
-	memset(sh, 0, sizeof(*sh) + (conf->raid_disks-1)*sizeof(struct r5dev));
+
 	sh->raid_conf = conf;
-	spin_lock_init(&sh->lock);
 
-	if (grow_buffers(sh, conf->raid_disks)) {
-		shrink_buffers(sh, conf->raid_disks);
+	spin_lock_init(&sh->stripe_lock);
+
+	if (grow_buffers(sh)) {
+		shrink_buffers(sh);
 		kmem_cache_free(conf->slab_cache, sh);
 		return 0;
 	}
-	sh->disks = conf->raid_disks;
+	sh->hash_lock_index = hash;
 	/* we just created an active stripe so... */
 	atomic_set(&sh->count, 1);
 	atomic_inc(&conf->active_stripes);
@@ -861,13 +1731,20 @@ static int grow_one_stripe(raid5_conf_t *conf)
 	return 1;
 }
 
-static int grow_stripes(raid5_conf_t *conf, int num)
+static int grow_stripes(struct r5conf *conf, int num)
 {
 	struct kmem_cache *sc;
-	int devs = conf->raid_disks;
+	int devs = max(conf->raid_disks, conf->previous_raid_disks);
+	int hash;
+
+	if (conf->mddev->gendisk)
+		sprintf(conf->cache_name[0],
+			"raid%d-%s", conf->level, mdname(conf->mddev));
+	else
+		sprintf(conf->cache_name[0],
+			"raid%d-%p", conf->level, conf->mddev);
+	sprintf(conf->cache_name[1], "%s-alt", conf->cache_name[0]);
 
-	sprintf(conf->cache_name[0], "raid5-%s", mdname(conf->mddev));
-	sprintf(conf->cache_name[1], "raid5-%s-alt", mdname(conf->mddev));
 	conf->active_name = 0;
 	sc = kmem_cache_create(conf->cache_name[conf->active_name],
 			       sizeof(struct stripe_head)+(devs-1)*sizeof(struct r5dev),
@@ -876,14 +1753,39 @@ static int grow_stripes(raid5_conf_t *conf, int num)
 		return 1;
 	conf->slab_cache = sc;
 	conf->pool_size = devs;
-	while (num--)
-		if (!grow_one_stripe(conf))
+	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
+	while (num--) {
+		if (!grow_one_stripe(conf, hash))
 			return 1;
+		conf->max_nr_stripes++;
+		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
+	}
 	return 0;
 }
 
-#ifdef CONFIG_MD_RAID5_RESHAPE
-static int resize_stripes(raid5_conf_t *conf, int newsize)
+/**
+ * scribble_len - return the required size of the scribble region
+ * @num - total number of disks in the array
+ *
+ * The size must be enough to contain:
+ * 1/ a struct page pointer for each device in the array +2
+ * 2/ room to convert each entry in (1) to its corresponding dma
+ *    (dma_map_page()) or page (page_address()) address.
+ *
+ * Note: the +2 is for the destination buffers of the ddf/raid6 case where we
+ * calculate over all devices (not just the data blocks), using zeros in place
+ * of the P and Q blocks.
+ */
+static size_t scribble_len(int num)
+{
+	size_t len;
+
+	len = sizeof(struct page *) * (num+2) + sizeof(addr_conv_t) * (num+2);
+
+	return len;
+}
+
+static int resize_stripes(struct r5conf *conf, int newsize)
 {
 	/* Make all the stripes able to hold 'newsize' devices.
 	 * New slots in each stripe get 'page' set to a new page.
@@ -891,7 +1793,7 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 	 * This happens in stages:
 	 * 1/ create a new kmem_cache and allocate the required number of
 	 *    stripe_heads.
-	 * 2/ gather all the old stripe_heads and tranfer the pages across
+	 * 2/ gather all the old stripe_heads and transfer the pages across
 	 *    to the new stripe_heads.  This will have the side effect of
 	 *    freezing the array as once all stripe_heads have been collected,
 	 *    no IO will be possible.  Old stripe heads are freed once their
@@ -911,9 +1813,11 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 	struct stripe_head *osh, *nsh;
 	LIST_HEAD(newstripes);
 	struct disk_info *ndisks;
+	unsigned long cpu;
 	int err;
 	struct kmem_cache *sc;
 	int i;
+	int hash, cnt;
 
 	if (newsize <= conf->pool_size)
 		return 0; /* never bother to shrink */
@@ -930,14 +1834,12 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 		return -ENOMEM;
 
 	for (i = conf->max_nr_stripes; i; i--) {
-		nsh = kmem_cache_alloc(sc, GFP_KERNEL);
+		nsh = kmem_cache_zalloc(sc, GFP_KERNEL);
 		if (!nsh)
 			break;
 
-		memset(nsh, 0, sizeof(*nsh) + (newsize-1)*sizeof(struct r5dev));
-
 		nsh->raid_conf = conf;
-		spin_lock_init(&nsh->lock);
+		spin_lock_init(&nsh->stripe_lock);
 
 		list_add(&nsh->lru, &newstripes);
 	}
@@ -955,28 +1857,38 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 	 * OK, we have enough stripes, start collecting inactive
 	 * stripes and copying them over
 	 */
+	hash = 0;
+	cnt = 0;
 	list_for_each_entry(nsh, &newstripes, lru) {
-		spin_lock_irq(&conf->device_lock);
-		wait_event_lock_irq(conf->wait_for_stripe,
-				    !list_empty(&conf->inactive_list),
-				    conf->device_lock,
-				    unplug_slaves(conf->mddev)
-			);
-		osh = get_free_stripe(conf);
-		spin_unlock_irq(&conf->device_lock);
+		lock_device_hash_lock(conf, hash);
+		wait_event_cmd(conf->wait_for_stripe,
+				    !list_empty(conf->inactive_list + hash),
+				    unlock_device_hash_lock(conf, hash),
+				    lock_device_hash_lock(conf, hash));
+		osh = get_free_stripe(conf, hash);
+		unlock_device_hash_lock(conf, hash);
 		atomic_set(&nsh->count, 1);
-		for(i=0; i<conf->pool_size; i++)
+		for(i=0; i<conf->pool_size; i++) {
 			nsh->dev[i].page = osh->dev[i].page;
+			nsh->dev[i].orig_page = osh->dev[i].page;
+		}
 		for( ; i<newsize; i++)
 			nsh->dev[i].page = NULL;
+		nsh->hash_lock_index = hash;
 		kmem_cache_free(conf->slab_cache, osh);
+		cnt++;
+		if (cnt >= conf->max_nr_stripes / NR_STRIPE_HASH_LOCKS +
+		    !!((conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS) > hash)) {
+			hash++;
+			cnt = 0;
+		}
 	}
 	kmem_cache_destroy(conf->slab_cache);
 
 	/* Step 3.
 	 * At this point, we are holding all the stripes so the array
 	 * is completely stalled, so now is a good time to resize
-	 * conf->disks.
+	 * conf->disks and the scribble region
 	 */
 	ndisks = kzalloc(newsize * sizeof(struct disk_info), GFP_NOIO);
 	if (ndisks) {
@@ -987,14 +1899,35 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 	} else
 		err = -ENOMEM;
 
+	get_online_cpus();
+	conf->scribble_len = scribble_len(newsize);
+	for_each_present_cpu(cpu) {
+		struct raid5_percpu *percpu;
+		void *scribble;
+
+		percpu = per_cpu_ptr(conf->percpu, cpu);
+		scribble = kmalloc(conf->scribble_len, GFP_NOIO);
+
+		if (scribble) {
+			kfree(percpu->scribble);
+			percpu->scribble = scribble;
+		} else {
+			err = -ENOMEM;
+			break;
+		}
+	}
+	put_online_cpus();
+
 	/* Step 4, return new stripes to service */
 	while(!list_empty(&newstripes)) {
 		nsh = list_entry(newstripes.next, struct stripe_head, lru);
 		list_del_init(&nsh->lru);
+
 		for (i=conf->raid_disks; i < newsize; i++)
 			if (nsh->dev[i].page == NULL) {
 				struct page *p = alloc_page(GFP_NOIO);
 				nsh->dev[i].page = p;
+				nsh->dev[i].orig_page = p;
 				if (!p)
 					err = -ENOMEM;
 			}
@@ -1007,28 +1940,29 @@ static int resize_stripes(raid5_conf_t *conf, int newsize)
 	conf->pool_size = newsize;
 	return err;
 }
-#endif
 
-static int drop_one_stripe(raid5_conf_t *conf)
+static int drop_one_stripe(struct r5conf *conf, int hash)
 {
 	struct stripe_head *sh;
 
-	spin_lock_irq(&conf->device_lock);
-	sh = get_free_stripe(conf);
-	spin_unlock_irq(&conf->device_lock);
+	spin_lock_irq(conf->hash_locks + hash);
+	sh = get_free_stripe(conf, hash);
+	spin_unlock_irq(conf->hash_locks + hash);
 	if (!sh)
 		return 0;
 	BUG_ON(atomic_read(&sh->count));
-	shrink_buffers(sh, conf->pool_size);
+	shrink_buffers(sh);
 	kmem_cache_free(conf->slab_cache, sh);
 	atomic_dec(&conf->active_stripes);
 	return 1;
 }
 
-static void shrink_stripes(raid5_conf_t *conf)
+static void shrink_stripes(struct r5conf *conf)
 {
-	while (drop_one_stripe(conf))
-		;
+	int hash;
+	for (hash = 0; hash < NR_STRIPE_HASH_LOCKS; hash++)
+		while (drop_one_stripe(conf, hash))
+			;
 
 	if (conf->slab_cache)
 		kmem_cache_destroy(conf->slab_cache);
@@ -1037,13 +1971,13 @@ static void shrink_stripes(raid5_conf_t *conf)
 
 static void raid5_end_read_request(struct bio * bi, int error)
 {
- 	struct stripe_head *sh = bi->bi_private;
-	raid5_conf_t *conf = sh->raid_conf;
+	struct stripe_head *sh = bi->bi_private;
+	struct r5conf *conf = sh->raid_conf;
 	int disks = sh->disks, i;
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
 	char b[BDEVNAME_SIZE];
-	mdk_rdev_t *rdev;
-
+	struct md_rdev *rdev = NULL;
+	sector_t s;
 
 	for (i=0 ; i<disks; i++)
 		if (bi == &sh->dev[i].req)
@@ -1056,78 +1990,137 @@ static void raid5_end_read_request(struct bio * bi, int error)
 		BUG();
 		return;
 	}
+	if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
+		/* If replacement finished while this request was outstanding,
+		 * 'replacement' might be NULL already.
+		 * In that case it moved down to 'rdev'.
+		 * rdev is not removed until all requests are finished.
+		 */
+		rdev = conf->disks[i].replacement;
+	if (!rdev)
+		rdev = conf->disks[i].rdev;
 
+	if (use_new_offset(conf, sh))
+		s = sh->sector + rdev->new_data_offset;
+	else
+		s = sh->sector + rdev->data_offset;
 	if (uptodate) {
 		set_bit(R5_UPTODATE, &sh->dev[i].flags);
 		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
-			rdev = conf->disks[i].rdev;
-			printk_rl(KERN_INFO "raid5:%s: read error corrected"
-				  " (%lu sectors at %llu on %s)\n",
-				  mdname(conf->mddev), STRIPE_SECTORS,
-				  (unsigned long long)(sh->sector
-						       + rdev->data_offset),
-				  bdevname(rdev->bdev, b));
+			/* Note that this cannot happen on a
+			 * replacement device.  We just fail those on
+			 * any error
+			 */
+			printk_ratelimited(
+				KERN_INFO
+				"md/raid:%s: read error corrected"
+				" (%lu sectors at %llu on %s)\n",
+				mdname(conf->mddev), STRIPE_SECTORS,
+				(unsigned long long)s,
+				bdevname(rdev->bdev, b));
+			atomic_add(STRIPE_SECTORS, &rdev->corrected_errors);
 			clear_bit(R5_ReadError, &sh->dev[i].flags);
 			clear_bit(R5_ReWrite, &sh->dev[i].flags);
-		}
-		if (atomic_read(&conf->disks[i].rdev->read_errors))
-			atomic_set(&conf->disks[i].rdev->read_errors, 0);
+		} 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(conf->disks[i].rdev->bdev, b);
+		const char *bdn = bdevname(rdev->bdev, b);
 		int retry = 0;
-		rdev = conf->disks[i].rdev;
+		int set_bad = 0;
 
 		clear_bit(R5_UPTODATE, &sh->dev[i].flags);
 		atomic_inc(&rdev->read_errors);
-		if (conf->mddev->degraded)
-			printk_rl(KERN_WARNING
-				  "raid5:%s: read error not correctable "
-				  "(sector %llu on %s).\n",
-				  mdname(conf->mddev),
-				  (unsigned long long)(sh->sector
-						       + rdev->data_offset),
-				  bdn);
-		else if (test_bit(R5_ReWrite, &sh->dev[i].flags))
+		if (test_bit(R5_ReadRepl, &sh->dev[i].flags))
+			printk_ratelimited(
+				KERN_WARNING
+				"md/raid:%s: read error on replacement device "
+				"(sector %llu on %s).\n",
+				mdname(conf->mddev),
+				(unsigned long long)s,
+				bdn);
+		else if (conf->mddev->degraded >= conf->max_degraded) {
+			set_bad = 1;
+			printk_ratelimited(
+				KERN_WARNING
+				"md/raid:%s: read error not correctable "
+				"(sector %llu on %s).\n",
+				mdname(conf->mddev),
+				(unsigned long long)s,
+				bdn);
+		} else if (test_bit(R5_ReWrite, &sh->dev[i].flags)) {
 			/* Oh, no!!! */
-			printk_rl(KERN_WARNING
-				  "raid5:%s: read error NOT corrected!! "
-				  "(sector %llu on %s).\n",
-				  mdname(conf->mddev),
-				  (unsigned long long)(sh->sector
-						       + rdev->data_offset),
-				  bdn);
-		else if (atomic_read(&rdev->read_errors)
+			set_bad = 1;
+			printk_ratelimited(
+				KERN_WARNING
+				"md/raid:%s: read error NOT corrected!! "
+				"(sector %llu on %s).\n",
+				mdname(conf->mddev),
+				(unsigned long long)s,
+				bdn);
+		} else if (atomic_read(&rdev->read_errors)
 			 > conf->max_nr_stripes)
 			printk(KERN_WARNING
-			       "raid5:%s: Too many read errors, failing device %s.\n",
+			       "md/raid:%s: Too many read errors, failing device %s.\n",
 			       mdname(conf->mddev), bdn);
 		else
 			retry = 1;
+		if (set_bad && test_bit(In_sync, &rdev->flags)
+		    && !test_bit(R5_ReadNoMerge, &sh->dev[i].flags))
+			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(conf->disks[i].rdev, conf->mddev);
+	rdev_dec_pending(rdev, conf->mddev);
 	clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	set_bit(STRIPE_HANDLE, &sh->state);
 	release_stripe(sh);
 }
 
-static void raid5_end_write_request (struct bio *bi, int error)
+static void raid5_end_write_request(struct bio *bi, int error)
 {
- 	struct stripe_head *sh = bi->bi_private;
-	raid5_conf_t *conf = sh->raid_conf;
+	struct stripe_head *sh = bi->bi_private;
+	struct r5conf *conf = sh->raid_conf;
 	int disks = sh->disks, i;
+	struct md_rdev *uninitialized_var(rdev);
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
+	sector_t first_bad;
+	int bad_sectors;
+	int replacement = 0;
 
-	for (i=0 ; i<disks; i++)
-		if (bi == &sh->dev[i].req)
+	for (i = 0 ; i < disks; i++) {
+		if (bi == &sh->dev[i].req) {
+			rdev = conf->disks[i].rdev;
 			break;
-
+		}
+		if (bi == &sh->dev[i].rreq) {
+			rdev = conf->disks[i].replacement;
+			if (rdev)
+				replacement = 1;
+			else
+				/* rdev was removed and 'replacement'
+				 * replaced it.  rdev is not removed
+				 * until all requests are finished.
+				 */
+				rdev = conf->disks[i].rdev;
+			break;
+		}
+	}
 	pr_debug("end_write_request %llu/%d, count %d, uptodate: %d.\n",
 		(unsigned long long)sh->sector, i, atomic_read(&sh->count),
 		uptodate);
@@ -1136,77 +2129,107 @@ static void raid5_end_write_request (struct bio *bi, int error)
 		return;
 	}
 
-	if (!uptodate)
-		md_error(conf->mddev, conf->disks[i].rdev);
+	if (replacement) {
+		if (!uptodate)
+			md_error(conf->mddev, rdev);
+		else if (is_badblock(rdev, sh->sector,
+				     STRIPE_SECTORS,
+				     &first_bad, &bad_sectors))
+			set_bit(R5_MadeGoodRepl, &sh->dev[i].flags);
+	} else {
+		if (!uptodate) {
+			set_bit(STRIPE_DEGRADED, &sh->state);
+			set_bit(WriteErrorSeen, &rdev->flags);
+			set_bit(R5_WriteError, &sh->dev[i].flags);
+			if (!test_and_set_bit(WantReplacement, &rdev->flags))
+				set_bit(MD_RECOVERY_NEEDED,
+					&rdev->mddev->recovery);
+		} else if (is_badblock(rdev, sh->sector,
+				       STRIPE_SECTORS,
+				       &first_bad, &bad_sectors)) {
+			set_bit(R5_MadeGood, &sh->dev[i].flags);
+			if (test_bit(R5_ReadError, &sh->dev[i].flags))
+				/* That was a successful write so make
+				 * sure it looks like we already did
+				 * a re-write.
+				 */
+				set_bit(R5_ReWrite, &sh->dev[i].flags);
+		}
+	}
+	rdev_dec_pending(rdev, conf->mddev);
 
-	rdev_dec_pending(conf->disks[i].rdev, conf->mddev);
-	
-	clear_bit(R5_LOCKED, &sh->dev[i].flags);
+	if (!test_and_clear_bit(R5_DOUBLE_LOCKED, &sh->dev[i].flags))
+		clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	set_bit(STRIPE_HANDLE, &sh->state);
 	release_stripe(sh);
 }
 
+static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous);
 
-static sector_t compute_blocknr(struct stripe_head *sh, int i);
-	
-static void raid5_build_block (struct stripe_head *sh, int i)
+static void raid5_build_block(struct stripe_head *sh, int i, int previous)
 {
 	struct r5dev *dev = &sh->dev[i];
 
 	bio_init(&dev->req);
 	dev->req.bi_io_vec = &dev->vec;
-	dev->req.bi_vcnt++;
-	dev->req.bi_max_vecs++;
-	dev->vec.bv_page = dev->page;
-	dev->vec.bv_len = STRIPE_SIZE;
-	dev->vec.bv_offset = 0;
-
-	dev->req.bi_sector = sh->sector;
+	dev->req.bi_max_vecs = 1;
 	dev->req.bi_private = sh;
 
+	bio_init(&dev->rreq);
+	dev->rreq.bi_io_vec = &dev->rvec;
+	dev->rreq.bi_max_vecs = 1;
+	dev->rreq.bi_private = sh;
+
 	dev->flags = 0;
-	dev->sector = compute_blocknr(sh, i);
+	dev->sector = compute_blocknr(sh, i, previous);
 }
 
-static void error(mddev_t *mddev, mdk_rdev_t *rdev)
+static void error(struct mddev *mddev, struct md_rdev *rdev)
 {
 	char b[BDEVNAME_SIZE];
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
-	pr_debug("raid5: error called\n");
+	struct r5conf *conf = mddev->private;
+	unsigned long flags;
+	pr_debug("raid456: error called\n");
 
-	if (!test_bit(Faulty, &rdev->flags)) {
-		set_bit(MD_CHANGE_DEVS, &mddev->flags);
-		if (test_and_clear_bit(In_sync, &rdev->flags)) {
-			unsigned long flags;
-			spin_lock_irqsave(&conf->device_lock, flags);
-			mddev->degraded++;
-			spin_unlock_irqrestore(&conf->device_lock, flags);
-			/*
-			 * if recovery was running, make sure it aborts.
-			 */
-			set_bit(MD_RECOVERY_INTR, &mddev->recovery);
-		}
-		set_bit(Faulty, &rdev->flags);
-		printk (KERN_ALERT
-			"raid5: Disk failure on %s, disabling device.\n"
-			"raid5: Operation continuing on %d devices.\n",
-			bdevname(rdev->bdev,b), conf->raid_disks - mddev->degraded);
-	}
+	spin_lock_irqsave(&conf->device_lock, flags);
+	clear_bit(In_sync, &rdev->flags);
+	mddev->degraded = calc_degraded(conf);
+	spin_unlock_irqrestore(&conf->device_lock, flags);
+	set_bit(MD_RECOVERY_INTR, &mddev->recovery);
+
+	set_bit(Blocked, &rdev->flags);
+	set_bit(Faulty, &rdev->flags);
+	set_bit(MD_CHANGE_DEVS, &mddev->flags);
+	printk(KERN_ALERT
+	       "md/raid:%s: Disk failure on %s, disabling device.\n"
+	       "md/raid:%s: Operation continuing on %d devices.\n",
+	       mdname(mddev),
+	       bdevname(rdev->bdev, b),
+	       mdname(mddev),
+	       conf->raid_disks - mddev->degraded);
 }
 
 /*
  * Input: a 'big' sector number,
  * Output: index of the data and parity disk, and the sector # in them.
  */
-static sector_t raid5_compute_sector(sector_t r_sector, unsigned int raid_disks,
-			unsigned int data_disks, unsigned int * dd_idx,
-			unsigned int * pd_idx, raid5_conf_t *conf)
+static sector_t raid5_compute_sector(struct r5conf *conf, sector_t r_sector,
+				     int previous, int *dd_idx,
+				     struct stripe_head *sh)
 {
-	long stripe;
-	unsigned long chunk_number;
+	sector_t stripe, stripe2;
+	sector_t chunk_number;
 	unsigned int chunk_offset;
+	int pd_idx, qd_idx;
+	int ddf_layout = 0;
 	sector_t new_sector;
-	int sectors_per_chunk = conf->chunk_size >> 9;
+	int algorithm = previous ? conf->prev_algo
+				 : conf->algorithm;
+	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
+					 : conf->chunk_sectors;
+	int raid_disks = previous ? conf->previous_raid_disks
+				  : conf->raid_disks;
+	int data_disks = raid_disks - conf->max_degraded;
 
 	/* First compute the information on this sector */
 
@@ -1215,83 +2238,176 @@ static sector_t raid5_compute_sector(sector_t r_sector, unsigned int raid_disks,
 	 */
 	chunk_offset = sector_div(r_sector, sectors_per_chunk);
 	chunk_number = r_sector;
-	BUG_ON(r_sector != chunk_number);
 
 	/*
 	 * Compute the stripe number
 	 */
-	stripe = chunk_number / data_disks;
-
-	/*
-	 * Compute the data disk and parity disk indexes inside the stripe
-	 */
-	*dd_idx = chunk_number % data_disks;
-
+	stripe = chunk_number;
+	*dd_idx = sector_div(stripe, data_disks);
+	stripe2 = stripe;
 	/*
 	 * Select the parity disk based on the user selected algorithm.
 	 */
+	pd_idx = qd_idx = -1;
 	switch(conf->level) {
 	case 4:
-		*pd_idx = data_disks;
+		pd_idx = data_disks;
 		break;
 	case 5:
-		switch (conf->algorithm) {
+		switch (algorithm) {
 		case ALGORITHM_LEFT_ASYMMETRIC:
-			*pd_idx = data_disks - stripe % raid_disks;
-			if (*dd_idx >= *pd_idx)
+			pd_idx = data_disks - sector_div(stripe2, raid_disks);
+			if (*dd_idx >= pd_idx)
 				(*dd_idx)++;
 			break;
 		case ALGORITHM_RIGHT_ASYMMETRIC:
-			*pd_idx = stripe % raid_disks;
-			if (*dd_idx >= *pd_idx)
+			pd_idx = sector_div(stripe2, raid_disks);
+			if (*dd_idx >= pd_idx)
 				(*dd_idx)++;
 			break;
 		case ALGORITHM_LEFT_SYMMETRIC:
-			*pd_idx = data_disks - stripe % raid_disks;
-			*dd_idx = (*pd_idx + 1 + *dd_idx) % raid_disks;
+			pd_idx = data_disks - sector_div(stripe2, raid_disks);
+			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
 			break;
 		case ALGORITHM_RIGHT_SYMMETRIC:
-			*pd_idx = stripe % raid_disks;
-			*dd_idx = (*pd_idx + 1 + *dd_idx) % raid_disks;
+			pd_idx = sector_div(stripe2, raid_disks);
+			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
+			break;
+		case ALGORITHM_PARITY_0:
+			pd_idx = 0;
+			(*dd_idx)++;
+			break;
+		case ALGORITHM_PARITY_N:
+			pd_idx = data_disks;
 			break;
 		default:
-			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
-				conf->algorithm);
+			BUG();
 		}
 		break;
 	case 6:
 
-		/**** FIX THIS ****/
-		switch (conf->algorithm) {
+		switch (algorithm) {
 		case ALGORITHM_LEFT_ASYMMETRIC:
-			*pd_idx = raid_disks - 1 - (stripe % raid_disks);
-			if (*pd_idx == raid_disks-1)
-				(*dd_idx)++; 	/* Q D D D P */
-			else if (*dd_idx >= *pd_idx)
+			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
+			qd_idx = pd_idx + 1;
+			if (pd_idx == raid_disks-1) {
+				(*dd_idx)++;	/* Q D D D P */
+				qd_idx = 0;
+			} else if (*dd_idx >= pd_idx)
 				(*dd_idx) += 2; /* D D P Q D */
 			break;
 		case ALGORITHM_RIGHT_ASYMMETRIC:
-			*pd_idx = stripe % raid_disks;
-			if (*pd_idx == raid_disks-1)
-				(*dd_idx)++; 	/* Q D D D P */
-			else if (*dd_idx >= *pd_idx)
+			pd_idx = sector_div(stripe2, raid_disks);
+			qd_idx = pd_idx + 1;
+			if (pd_idx == raid_disks-1) {
+				(*dd_idx)++;	/* Q D D D P */
+				qd_idx = 0;
+			} else if (*dd_idx >= pd_idx)
 				(*dd_idx) += 2; /* D D P Q D */
 			break;
 		case ALGORITHM_LEFT_SYMMETRIC:
-			*pd_idx = raid_disks - 1 - (stripe % raid_disks);
-			*dd_idx = (*pd_idx + 2 + *dd_idx) % raid_disks;
+			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
+			qd_idx = (pd_idx + 1) % raid_disks;
+			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
 			break;
 		case ALGORITHM_RIGHT_SYMMETRIC:
-			*pd_idx = stripe % raid_disks;
-			*dd_idx = (*pd_idx + 2 + *dd_idx) % raid_disks;
+			pd_idx = sector_div(stripe2, raid_disks);
+			qd_idx = (pd_idx + 1) % raid_disks;
+			*dd_idx = (pd_idx + 2 + *dd_idx) % raid_disks;
+			break;
+
+		case ALGORITHM_PARITY_0:
+			pd_idx = 0;
+			qd_idx = 1;
+			(*dd_idx) += 2;
+			break;
+		case ALGORITHM_PARITY_N:
+			pd_idx = data_disks;
+			qd_idx = data_disks + 1;
 			break;
+
+		case ALGORITHM_ROTATING_ZERO_RESTART:
+			/* Exactly the same as RIGHT_ASYMMETRIC, but or
+			 * of blocks for computing Q is different.
+			 */
+			pd_idx = sector_div(stripe2, raid_disks);
+			qd_idx = pd_idx + 1;
+			if (pd_idx == raid_disks-1) {
+				(*dd_idx)++;	/* Q D D D P */
+				qd_idx = 0;
+			} else if (*dd_idx >= pd_idx)
+				(*dd_idx) += 2; /* D D P Q D */
+			ddf_layout = 1;
+			break;
+
+		case ALGORITHM_ROTATING_N_RESTART:
+			/* Same a left_asymmetric, by first stripe is
+			 * D D D P Q  rather than
+			 * Q D D D P
+			 */
+			stripe2 += 1;
+			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
+			qd_idx = pd_idx + 1;
+			if (pd_idx == raid_disks-1) {
+				(*dd_idx)++;	/* Q D D D P */
+				qd_idx = 0;
+			} else if (*dd_idx >= pd_idx)
+				(*dd_idx) += 2; /* D D P Q D */
+			ddf_layout = 1;
+			break;
+
+		case ALGORITHM_ROTATING_N_CONTINUE:
+			/* Same as left_symmetric but Q is before P */
+			pd_idx = raid_disks - 1 - sector_div(stripe2, raid_disks);
+			qd_idx = (pd_idx + raid_disks - 1) % raid_disks;
+			*dd_idx = (pd_idx + 1 + *dd_idx) % raid_disks;
+			ddf_layout = 1;
+			break;
+
+		case ALGORITHM_LEFT_ASYMMETRIC_6:
+			/* RAID5 left_asymmetric, with Q on last device */
+			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
+			if (*dd_idx >= pd_idx)
+				(*dd_idx)++;
+			qd_idx = raid_disks - 1;
+			break;
+
+		case ALGORITHM_RIGHT_ASYMMETRIC_6:
+			pd_idx = sector_div(stripe2, raid_disks-1);
+			if (*dd_idx >= pd_idx)
+				(*dd_idx)++;
+			qd_idx = raid_disks - 1;
+			break;
+
+		case ALGORITHM_LEFT_SYMMETRIC_6:
+			pd_idx = data_disks - sector_div(stripe2, raid_disks-1);
+			*dd_idx = (pd_idx + 1 + *dd_idx) % (raid_disks-1);
+			qd_idx = raid_disks - 1;
+			break;
+
+		case ALGORITHM_RIGHT_SYMMETRIC_6:
+			pd_idx = sector_div(stripe2, raid_disks-1);
+			*dd_idx = (pd_idx + 1 + *dd_idx) % (raid_disks-1);
+			qd_idx = raid_disks - 1;
+			break;
+
+		case ALGORITHM_PARITY_0_6:
+			pd_idx = 0;
+			(*dd_idx)++;
+			qd_idx = raid_disks - 1;
+			break;
+
 		default:
-			printk (KERN_CRIT "raid6: unsupported algorithm %d\n",
-				conf->algorithm);
+			BUG();
 		}
 		break;
 	}
 
+	if (sh) {
+		sh->pd_idx = pd_idx;
+		sh->qd_idx = qd_idx;
+		sh->ddf_layout = ddf_layout;
+	}
 	/*
 	 * Finally, compute the new sector number
 	 */
@@ -1300,29 +2416,33 @@ static sector_t raid5_compute_sector(sector_t r_sector, unsigned int raid_disks,
 }
 
 
-static sector_t compute_blocknr(struct stripe_head *sh, int i)
+static sector_t compute_blocknr(struct stripe_head *sh, int i, int previous)
 {
-	raid5_conf_t *conf = sh->raid_conf;
+	struct r5conf *conf = sh->raid_conf;
 	int raid_disks = sh->disks;
 	int data_disks = raid_disks - conf->max_degraded;
 	sector_t new_sector = sh->sector, check;
-	int sectors_per_chunk = conf->chunk_size >> 9;
+	int sectors_per_chunk = previous ? conf->prev_chunk_sectors
+					 : conf->chunk_sectors;
+	int algorithm = previous ? conf->prev_algo
+				 : conf->algorithm;
 	sector_t stripe;
 	int chunk_offset;
-	int chunk_number, dummy1, dummy2, dd_idx = i;
+	sector_t chunk_number;
+	int dummy1, dd_idx = i;
 	sector_t r_sector;
+	struct stripe_head sh2;
 
 
 	chunk_offset = sector_div(new_sector, sectors_per_chunk);
 	stripe = new_sector;
-	BUG_ON(new_sector != stripe);
 
 	if (i == sh->pd_idx)
 		return 0;
 	switch(conf->level) {
 	case 4: break;
 	case 5:
-		switch (conf->algorithm) {
+		switch (algorithm) {
 		case ALGORITHM_LEFT_ASYMMETRIC:
 		case ALGORITHM_RIGHT_ASYMMETRIC:
 			if (i > sh->pd_idx)
@@ -1334,19 +2454,25 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i)
 				i += raid_disks;
 			i -= (sh->pd_idx + 1);
 			break;
+		case ALGORITHM_PARITY_0:
+			i -= 1;
+			break;
+		case ALGORITHM_PARITY_N:
+			break;
 		default:
-			printk(KERN_ERR "raid5: unsupported algorithm %d\n",
-			       conf->algorithm);
+			BUG();
 		}
 		break;
 	case 6:
-		if (i == raid6_next_disk(sh->pd_idx, raid_disks))
+		if (i == sh->qd_idx)
 			return 0; /* It is the Q disk */
-		switch (conf->algorithm) {
+		switch (algorithm) {
 		case ALGORITHM_LEFT_ASYMMETRIC:
 		case ALGORITHM_RIGHT_ASYMMETRIC:
-		  	if (sh->pd_idx == raid_disks-1)
-				i--; 	/* Q D D D P */
+		case ALGORITHM_ROTATING_ZERO_RESTART:
+		case ALGORITHM_ROTATING_N_RESTART:
+			if (sh->pd_idx == raid_disks-1)
+				i--;	/* Q D D D P */
 			else if (i > sh->pd_idx)
 				i -= 2; /* D D P Q D */
 			break;
@@ -1361,282 +2487,66 @@ static sector_t compute_blocknr(struct stripe_head *sh, int i)
 				i -= (sh->pd_idx + 2);
 			}
 			break;
+		case ALGORITHM_PARITY_0:
+			i -= 2;
+			break;
+		case ALGORITHM_PARITY_N:
+			break;
+		case ALGORITHM_ROTATING_N_CONTINUE:
+			/* Like left_symmetric, but P is before Q */
+			if (sh->pd_idx == 0)
+				i--;	/* P D D D Q */
+			else {
+				/* D D Q P D */
+				if (i < sh->pd_idx)
+					i += raid_disks;
+				i -= (sh->pd_idx + 1);
+			}
+			break;
+		case ALGORITHM_LEFT_ASYMMETRIC_6:
+		case ALGORITHM_RIGHT_ASYMMETRIC_6:
+			if (i > sh->pd_idx)
+				i--;
+			break;
+		case ALGORITHM_LEFT_SYMMETRIC_6:
+		case ALGORITHM_RIGHT_SYMMETRIC_6:
+			if (i < sh->pd_idx)
+				i += data_disks + 1;
+			i -= (sh->pd_idx + 1);
+			break;
+		case ALGORITHM_PARITY_0_6:
+			i -= 1;
+			break;
 		default:
-			printk (KERN_CRIT "raid6: unsupported algorithm %d\n",
-				conf->algorithm);
+			BUG();
 		}
 		break;
 	}
 
 	chunk_number = stripe * data_disks + i;
-	r_sector = (sector_t)chunk_number * sectors_per_chunk + chunk_offset;
-
-	check = raid5_compute_sector (r_sector, raid_disks, data_disks, &dummy1, &dummy2, conf);
-	if (check != sh->sector || dummy1 != dd_idx || dummy2 != sh->pd_idx) {
-		printk(KERN_ERR "compute_blocknr: map not correct\n");
+	r_sector = chunk_number * sectors_per_chunk + chunk_offset;
+
+	check = raid5_compute_sector(conf, r_sector,
+				     previous, &dummy1, &sh2);
+	if (check != sh->sector || dummy1 != dd_idx || sh2.pd_idx != sh->pd_idx
+		|| sh2.qd_idx != sh->qd_idx) {
+		printk(KERN_ERR "md/raid:%s: compute_blocknr: map not correct\n",
+		       mdname(conf->mddev));
 		return 0;
 	}
 	return r_sector;
 }
 
 
-
-/*
- * Copy data between a page in the stripe cache, and one or more bion
- * The page could align with the middle of the bio, or there could be
- * several bion, each with several bio_vecs, which cover part of the page
- * Multiple bion are linked together on bi_next.  There may be extras
- * at the end of this list.  We ignore them.
- */
-static void copy_data(int frombio, struct bio *bio,
-		     struct page *page,
-		     sector_t sector)
-{
-	char *pa = page_address(page);
-	struct bio_vec *bvl;
-	int i;
-	int page_offset;
-
-	if (bio->bi_sector >= sector)
-		page_offset = (signed)(bio->bi_sector - sector) * 512;
-	else
-		page_offset = (signed)(sector - bio->bi_sector) * -512;
-	bio_for_each_segment(bvl, bio, i) {
-		int len = bio_iovec_idx(bio,i)->bv_len;
-		int clen;
-		int b_offset = 0;
-
-		if (page_offset < 0) {
-			b_offset = -page_offset;
-			page_offset += b_offset;
-			len -= b_offset;
-		}
-
-		if (len > 0 && page_offset + len > STRIPE_SIZE)
-			clen = STRIPE_SIZE - page_offset;
-		else clen = len;
-
-		if (clen > 0) {
-			char *ba = __bio_kmap_atomic(bio, i, KM_USER0);
-			if (frombio)
-				memcpy(pa+page_offset, ba+b_offset, clen);
-			else
-				memcpy(ba+b_offset, pa+page_offset, clen);
-			__bio_kunmap_atomic(ba, KM_USER0);
-		}
-		if (clen < len) /* hit end of page */
-			break;
-		page_offset +=  len;
-	}
-}
-
-#define check_xor()	do {						  \
-				if (count == MAX_XOR_BLOCKS) {		  \
-				xor_blocks(count, STRIPE_SIZE, dest, ptr);\
-				count = 0;				  \
-			   }						  \
-			} while(0)
-
-static void compute_parity6(struct stripe_head *sh, int method)
-{
-	raid6_conf_t *conf = sh->raid_conf;
-	int i, pd_idx = sh->pd_idx, qd_idx, d0_idx, disks = sh->disks, count;
-	struct bio *chosen;
-	/**** FIX THIS: This could be very bad if disks is close to 256 ****/
-	void *ptrs[disks];
-
-	qd_idx = raid6_next_disk(pd_idx, disks);
-	d0_idx = raid6_next_disk(qd_idx, disks);
-
-	pr_debug("compute_parity, stripe %llu, method %d\n",
-		(unsigned long long)sh->sector, method);
-
-	switch(method) {
-	case READ_MODIFY_WRITE:
-		BUG();		/* READ_MODIFY_WRITE N/A for RAID-6 */
-	case RECONSTRUCT_WRITE:
-		for (i= disks; i-- ;)
-			if ( i != pd_idx && i != qd_idx && sh->dev[i].towrite ) {
-				chosen = sh->dev[i].towrite;
-				sh->dev[i].towrite = NULL;
-
-				if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
-					wake_up(&conf->wait_for_overlap);
-
-				BUG_ON(sh->dev[i].written);
-				sh->dev[i].written = chosen;
-			}
-		break;
-	case CHECK_PARITY:
-		BUG();		/* Not implemented yet */
-	}
-
-	for (i = disks; i--;)
-		if (sh->dev[i].written) {
-			sector_t sector = sh->dev[i].sector;
-			struct bio *wbi = sh->dev[i].written;
-			while (wbi && wbi->bi_sector < sector + STRIPE_SECTORS) {
-				copy_data(1, wbi, sh->dev[i].page, sector);
-				wbi = r5_next_bio(wbi, sector);
-			}
-
-			set_bit(R5_LOCKED, &sh->dev[i].flags);
-			set_bit(R5_UPTODATE, &sh->dev[i].flags);
-		}
-
-//	switch(method) {
-//	case RECONSTRUCT_WRITE:
-//	case CHECK_PARITY:
-//	case UPDATE_PARITY:
-		/* Note that unlike RAID-5, the ordering of the disks matters greatly. */
-		/* FIX: Is this ordering of drives even remotely optimal? */
-		count = 0;
-		i = d0_idx;
-		do {
-			ptrs[count++] = page_address(sh->dev[i].page);
-			if (count <= disks-2 && !test_bit(R5_UPTODATE, &sh->dev[i].flags))
-				printk("block %d/%d not uptodate on parity calc\n", i,count);
-			i = raid6_next_disk(i, disks);
-		} while ( i != d0_idx );
-//		break;
-//	}
-
-	raid6_call.gen_syndrome(disks, STRIPE_SIZE, ptrs);
-
-	switch(method) {
-	case RECONSTRUCT_WRITE:
-		set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
-		set_bit(R5_UPTODATE, &sh->dev[qd_idx].flags);
-		set_bit(R5_LOCKED,   &sh->dev[pd_idx].flags);
-		set_bit(R5_LOCKED,   &sh->dev[qd_idx].flags);
-		break;
-	case UPDATE_PARITY:
-		set_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
-		set_bit(R5_UPTODATE, &sh->dev[qd_idx].flags);
-		break;
-	}
-}
-
-
-/* Compute one missing block */
-static void compute_block_1(struct stripe_head *sh, int dd_idx, int nozero)
-{
-	int i, count, disks = sh->disks;
-	void *ptr[MAX_XOR_BLOCKS], *dest, *p;
-	int pd_idx = sh->pd_idx;
-	int qd_idx = raid6_next_disk(pd_idx, disks);
-
-	pr_debug("compute_block_1, stripe %llu, idx %d\n",
-		(unsigned long long)sh->sector, dd_idx);
-
-	if ( dd_idx == qd_idx ) {
-		/* We're actually computing the Q drive */
-		compute_parity6(sh, UPDATE_PARITY);
-	} else {
-		dest = page_address(sh->dev[dd_idx].page);
-		if (!nozero) memset(dest, 0, STRIPE_SIZE);
-		count = 0;
-		for (i = disks ; i--; ) {
-			if (i == dd_idx || i == qd_idx)
-				continue;
-			p = page_address(sh->dev[i].page);
-			if (test_bit(R5_UPTODATE, &sh->dev[i].flags))
-				ptr[count++] = p;
-			else
-				printk("compute_block() %d, stripe %llu, %d"
-				       " not present\n", dd_idx,
-				       (unsigned long long)sh->sector, i);
-
-			check_xor();
-		}
-		if (count)
-			xor_blocks(count, STRIPE_SIZE, dest, ptr);
-		if (!nozero) set_bit(R5_UPTODATE, &sh->dev[dd_idx].flags);
-		else clear_bit(R5_UPTODATE, &sh->dev[dd_idx].flags);
-	}
-}
-
-/* Compute two missing blocks */
-static void compute_block_2(struct stripe_head *sh, int dd_idx1, int dd_idx2)
-{
-	int i, count, disks = sh->disks;
-	int pd_idx = sh->pd_idx;
-	int qd_idx = raid6_next_disk(pd_idx, disks);
-	int d0_idx = raid6_next_disk(qd_idx, disks);
-	int faila, failb;
-
-	/* faila and failb are disk numbers relative to d0_idx */
-	/* pd_idx become disks-2 and qd_idx become disks-1 */
-	faila = (dd_idx1 < d0_idx) ? dd_idx1+(disks-d0_idx) : dd_idx1-d0_idx;
-	failb = (dd_idx2 < d0_idx) ? dd_idx2+(disks-d0_idx) : dd_idx2-d0_idx;
-
-	BUG_ON(faila == failb);
-	if ( failb < faila ) { int tmp = faila; faila = failb; failb = tmp; }
-
-	pr_debug("compute_block_2, stripe %llu, idx %d,%d (%d,%d)\n",
-	       (unsigned long long)sh->sector, dd_idx1, dd_idx2, faila, failb);
-
-	if ( failb == disks-1 ) {
-		/* Q disk is one of the missing disks */
-		if ( faila == disks-2 ) {
-			/* Missing P+Q, just recompute */
-			compute_parity6(sh, UPDATE_PARITY);
-			return;
-		} else {
-			/* We're missing D+Q; recompute D from P */
-			compute_block_1(sh, (dd_idx1 == qd_idx) ? dd_idx2 : dd_idx1, 0);
-			compute_parity6(sh, UPDATE_PARITY); /* Is this necessary? */
-			return;
-		}
-	}
-
-	/* We're missing D+P or D+D; build pointer table */
-	{
-		/**** FIX THIS: This could be very bad if disks is close to 256 ****/
-		void *ptrs[disks];
-
-		count = 0;
-		i = d0_idx;
-		do {
-			ptrs[count++] = page_address(sh->dev[i].page);
-			i = raid6_next_disk(i, disks);
-			if (i != dd_idx1 && i != dd_idx2 &&
-			    !test_bit(R5_UPTODATE, &sh->dev[i].flags))
-				printk("compute_2 with missing block %d/%d\n", count, i);
-		} while ( i != d0_idx );
-
-		if ( failb == disks-2 ) {
-			/* We're missing D+P. */
-			raid6_datap_recov(disks, STRIPE_SIZE, faila, ptrs);
-		} else {
-			/* We're missing D+D. */
-			raid6_2data_recov(disks, STRIPE_SIZE, faila, failb, ptrs);
-		}
-
-		/* Both the above update both missing blocks */
-		set_bit(R5_UPTODATE, &sh->dev[dd_idx1].flags);
-		set_bit(R5_UPTODATE, &sh->dev[dd_idx2].flags);
-	}
-}
-
 static void
-schedule_reconstruction5(struct stripe_head *sh, struct stripe_head_state *s,
+schedule_reconstruction(struct stripe_head *sh, struct stripe_head_state *s,
 			 int rcw, int expand)
 {
 	int i, pd_idx = sh->pd_idx, disks = sh->disks;
+	struct r5conf *conf = sh->raid_conf;
+	int level = conf->level;
 
 	if (rcw) {
-		/* if we are not expanding this is a proper write request, and
-		 * there will be bios with new data to be drained into the
-		 * stripe cache
-		 */
-		if (!expand) {
-			sh->reconstruct_state = reconstruct_state_drain_run;
-			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
-		} else
-			sh->reconstruct_state = reconstruct_state_run;
-
-		set_bit(STRIPE_OP_POSTXOR, &s->ops_request);
 
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
@@ -1649,18 +2559,29 @@ schedule_reconstruction5(struct stripe_head *sh, struct stripe_head_state *s,
 				s->locked++;
 			}
 		}
-		if (s->locked + 1 == disks)
+		/* if we are not expanding this is a proper write request, and
+		 * there will be bios with new data to be drained into the
+		 * stripe cache
+		 */
+		if (!expand) {
+			if (!s->locked)
+				/* False alarm, nothing to do */
+				return;
+			sh->reconstruct_state = reconstruct_state_drain_run;
+			set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
+		} else
+			sh->reconstruct_state = reconstruct_state_run;
+
+		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
+
+		if (s->locked + conf->max_degraded == disks)
 			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
-				atomic_inc(&sh->raid_conf->pending_full_writes);
+				atomic_inc(&conf->pending_full_writes);
 	} else {
+		BUG_ON(level == 6);
 		BUG_ON(!(test_bit(R5_UPTODATE, &sh->dev[pd_idx].flags) ||
 			test_bit(R5_Wantcompute, &sh->dev[pd_idx].flags)));
 
-		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
-		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
-		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
-		set_bit(STRIPE_OP_POSTXOR, &s->ops_request);
-
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
 			if (i == pd_idx)
@@ -1675,15 +2596,31 @@ schedule_reconstruction5(struct stripe_head *sh, struct stripe_head_state *s,
 				s->locked++;
 			}
 		}
+		if (!s->locked)
+			/* False alarm - nothing to do */
+			return;
+		sh->reconstruct_state = reconstruct_state_prexor_drain_run;
+		set_bit(STRIPE_OP_PREXOR, &s->ops_request);
+		set_bit(STRIPE_OP_BIODRAIN, &s->ops_request);
+		set_bit(STRIPE_OP_RECONSTRUCT, &s->ops_request);
 	}
 
-	/* keep the parity disk locked while asynchronous operations
+	/* keep the parity disk(s) locked while asynchronous operations
 	 * are in flight
 	 */
 	set_bit(R5_LOCKED, &sh->dev[pd_idx].flags);
 	clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
 	s->locked++;
 
+	if (level == 6) {
+		int qd_idx = sh->qd_idx;
+		struct r5dev *dev = &sh->dev[qd_idx];
+
+		set_bit(R5_LOCKED, &dev->flags);
+		clear_bit(R5_UPTODATE, &dev->flags);
+		s->locked++;
+	}
+
 	pr_debug("%s: stripe %llu locked: %d ops_request: %lx\n",
 		__func__, (unsigned long long)sh->sector,
 		s->locked, s->ops_request);
@@ -1697,95 +2634,95 @@ schedule_reconstruction5(struct stripe_head *sh, struct stripe_head_state *s,
 static int add_stripe_bio(struct stripe_head *sh, struct bio *bi, int dd_idx, int forwrite)
 {
 	struct bio **bip;
-	raid5_conf_t *conf = sh->raid_conf;
+	struct r5conf *conf = sh->raid_conf;
 	int firstwrite=0;
 
-	pr_debug("adding bh b#%llu to stripe s#%llu\n",
-		(unsigned long long)bi->bi_sector,
+	pr_debug("adding bi b#%llu to stripe s#%llu\n",
+		(unsigned long long)bi->bi_iter.bi_sector,
 		(unsigned long long)sh->sector);
 
-
-	spin_lock(&sh->lock);
-	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;
-	while (*bip && (*bip)->bi_sector < bi->bi_sector) {
-		if ((*bip)->bi_sector + ((*bip)->bi_size >> 9) > bi->bi_sector)
+	while (*bip && (*bip)->bi_iter.bi_sector < bi->bi_iter.bi_sector) {
+		if (bio_end_sector(*bip) > bi->bi_iter.bi_sector)
 			goto overlap;
 		bip = & (*bip)->bi_next;
 	}
-	if (*bip && (*bip)->bi_sector < bi->bi_sector + ((bi->bi_size)>>9))
+	if (*bip && (*bip)->bi_iter.bi_sector < bio_end_sector(bi))
 		goto overlap;
 
 	BUG_ON(*bip && bi->bi_next && (*bip) != bi->bi_next);
 	if (*bip)
 		bi->bi_next = *bip;
 	*bip = bi;
-	bi->bi_phys_segments ++;
-	spin_unlock_irq(&conf->device_lock);
-	spin_unlock(&sh->lock);
-
-	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
-		(unsigned long long)bi->bi_sector,
-		(unsigned long long)sh->sector, dd_idx);
-
-	if (conf->mddev->bitmap && firstwrite) {
-		bitmap_startwrite(conf->mddev->bitmap, sh->sector,
-				  STRIPE_SECTORS, 0);
-		sh->bm_seq = conf->seq_flush+1;
-		set_bit(STRIPE_BIT_DELAY, &sh->state);
-	}
+	raid5_inc_bi_active_stripes(bi);
 
 	if (forwrite) {
 		/* check if page is covered */
 		sector_t sector = sh->dev[dd_idx].sector;
 		for (bi=sh->dev[dd_idx].towrite;
 		     sector < sh->dev[dd_idx].sector + STRIPE_SECTORS &&
-			     bi && bi->bi_sector <= sector;
+			     bi && bi->bi_iter.bi_sector <= sector;
 		     bi = r5_next_bio(bi, sh->dev[dd_idx].sector)) {
-			if (bi->bi_sector + (bi->bi_size>>9) >= sector)
-				sector = bi->bi_sector + (bi->bi_size>>9);
+			if (bio_end_sector(bi) >= sector)
+				sector = bio_end_sector(bi);
 		}
 		if (sector >= sh->dev[dd_idx].sector + STRIPE_SECTORS)
 			set_bit(R5_OVERWRITE, &sh->dev[dd_idx].flags);
 	}
+
+	pr_debug("added bi b#%llu to stripe s#%llu, disk %d.\n",
+		(unsigned long long)(*bip)->bi_iter.bi_sector,
+		(unsigned long long)sh->sector, dd_idx);
+	spin_unlock_irq(&sh->stripe_lock);
+
+	if (conf->mddev->bitmap && firstwrite) {
+		bitmap_startwrite(conf->mddev->bitmap, sh->sector,
+				  STRIPE_SECTORS, 0);
+		sh->bm_seq = conf->seq_flush+1;
+		set_bit(STRIPE_BIT_DELAY, &sh->state);
+	}
 	return 1;
 
  overlap:
 	set_bit(R5_Overlap, &sh->dev[dd_idx].flags);
-	spin_unlock_irq(&conf->device_lock);
-	spin_unlock(&sh->lock);
+	spin_unlock_irq(&sh->stripe_lock);
 	return 0;
 }
 
-static void end_reshape(raid5_conf_t *conf);
+static void end_reshape(struct r5conf *conf);
 
-static int page_is_zero(struct page *p)
+static void stripe_set_idx(sector_t stripe, struct r5conf *conf, int previous,
+			    struct stripe_head *sh)
 {
-	char *a = page_address(p);
-	return ((*(u32*)a) == 0 &&
-		memcmp(a, a+4, STRIPE_SIZE-4)==0);
-}
-
-static int stripe_to_pdidx(sector_t stripe, raid5_conf_t *conf, int disks)
-{
-	int sectors_per_chunk = conf->chunk_size >> 9;
-	int pd_idx, dd_idx;
+	int sectors_per_chunk =
+		previous ? conf->prev_chunk_sectors : conf->chunk_sectors;
+	int dd_idx;
 	int chunk_offset = sector_div(stripe, sectors_per_chunk);
+	int disks = previous ? conf->previous_raid_disks : conf->raid_disks;
 
-	raid5_compute_sector(stripe * (disks - conf->max_degraded)
+	raid5_compute_sector(conf,
+			     stripe * (disks - conf->max_degraded)
 			     *sectors_per_chunk + chunk_offset,
-			     disks, disks - conf->max_degraded,
-			     &dd_idx, &pd_idx, conf);
-	return pd_idx;
+			     previous,
+			     &dd_idx, sh);
 }
 
 static void
-handle_failed_stripe(raid5_conf_t *conf, struct stripe_head *sh,
+handle_failed_stripe(struct r5conf *conf, struct stripe_head *sh,
 				struct stripe_head_state *s, int disks,
 				struct bio **return_bi)
 {
@@ -1795,46 +2732,63 @@ handle_failed_stripe(raid5_conf_t *conf, struct stripe_head *sh,
 		int bitmap_end = 0;
 
 		if (test_bit(R5_ReadError, &sh->dev[i].flags)) {
-			mdk_rdev_t *rdev;
+			struct md_rdev *rdev;
 			rcu_read_lock();
 			rdev = rcu_dereference(conf->disks[i].rdev);
 			if (rdev && test_bit(In_sync, &rdev->flags))
-				/* multiple read failures in one stripe */
-				md_error(conf->mddev, rdev);
+				atomic_inc(&rdev->nr_pending);
+			else
+				rdev = NULL;
 			rcu_read_unlock();
+			if (rdev) {
+				if (!rdev_set_badblocks(
+					    rdev,
+					    sh->sector,
+					    STRIPE_SECTORS, 0))
+					md_error(conf->mddev, rdev);
+				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;
-		if (bi) {
-			s->to_write--;
+		spin_unlock_irq(&sh->stripe_lock);
+		if (bi)
 			bitmap_end = 1;
-		}
 
 		if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
 			wake_up(&conf->wait_for_overlap);
 
-		while (bi && bi->bi_sector <
+		while (bi && bi->bi_iter.bi_sector <
 			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 (--bi->bi_phys_segments == 0) {
+			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;
+		if (test_and_clear_bit(R5_SkipCopy, &sh->dev[i].flags)) {
+			WARN_ON(test_bit(R5_UPTODATE, &sh->dev[i].flags));
+			sh->dev[i].page = sh->dev[i].orig_page;
+		}
+
 		if (bi) bitmap_end = 1;
-		while (bi && bi->bi_sector <
+		while (bi && bi->bi_iter.bi_sector <
 		       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 (--bi->bi_phys_segments == 0) {
+			if (!raid5_dec_bi_active_stripes(bi)) {
 				md_write_end(conf->mddev);
 				bi->bi_next = *return_bi;
 				*return_bi = bi;
@@ -1848,27 +2802,31 @@ handle_failed_stripe(raid5_conf_t *conf, struct stripe_head *sh,
 		if (!test_bit(R5_Wantfill, &sh->dev[i].flags) &&
 		    (!test_bit(R5_Insync, &sh->dev[i].flags) ||
 		      test_bit(R5_ReadError, &sh->dev[i].flags))) {
+			spin_lock_irq(&sh->stripe_lock);
 			bi = sh->dev[i].toread;
 			sh->dev[i].toread = NULL;
+			spin_unlock_irq(&sh->stripe_lock);
 			if (test_and_clear_bit(R5_Overlap, &sh->dev[i].flags))
 				wake_up(&conf->wait_for_overlap);
-			if (bi) s->to_read--;
-			while (bi && bi->bi_sector <
+			while (bi && bi->bi_iter.bi_sector <
 			       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 (--bi->bi_phys_segments == 0) {
+				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);
+		/* If we were in the middle of a write the parity block might
+		 * still be locked - so just clear all R5_LOCKED flags
+		 */
+		clear_bit(R5_LOCKED, &sh->dev[i].flags);
 	}
 
 	if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
@@ -1876,17 +2834,80 @@ handle_failed_stripe(raid5_conf_t *conf, struct stripe_head *sh,
 			md_wakeup_thread(conf->mddev->thread);
 }
 
-/* fetch_block5 - checks the given member device to see if its data needs
+static void
+handle_failed_sync(struct r5conf *conf, struct stripe_head *sh,
+		   struct stripe_head_state *s)
+{
+	int abort = 0;
+	int i;
+
+	clear_bit(STRIPE_SYNCING, &sh->state);
+	if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
+		wake_up(&conf->wait_for_overlap);
+	s->syncing = 0;
+	s->replacing = 0;
+	/* There is nothing more to do for sync/check/repair.
+	 * Don't even need to abort as that is handled elsewhere
+	 * if needed, and not always wanted e.g. if there is a known
+	 * bad block here.
+	 * For recover/replace we need to record a bad block on all
+	 * non-sync devices, or abort the recovery
+	 */
+	if (test_bit(MD_RECOVERY_RECOVER, &conf->mddev->recovery)) {
+		/* During recovery devices cannot be removed, so
+		 * locking and refcounting of rdevs is not needed
+		 */
+		for (i = 0; i < conf->raid_disks; i++) {
+			struct md_rdev *rdev = conf->disks[i].rdev;
+			if (rdev
+			    && !test_bit(Faulty, &rdev->flags)
+			    && !test_bit(In_sync, &rdev->flags)
+			    && !rdev_set_badblocks(rdev, sh->sector,
+						   STRIPE_SECTORS, 0))
+				abort = 1;
+			rdev = conf->disks[i].replacement;
+			if (rdev
+			    && !test_bit(Faulty, &rdev->flags)
+			    && !test_bit(In_sync, &rdev->flags)
+			    && !rdev_set_badblocks(rdev, sh->sector,
+						   STRIPE_SECTORS, 0))
+				abort = 1;
+		}
+		if (abort)
+			conf->recovery_disabled =
+				conf->mddev->recovery_disabled;
+	}
+	md_done_sync(conf->mddev, STRIPE_SECTORS, !abort);
+}
+
+static int want_replace(struct stripe_head *sh, int disk_idx)
+{
+	struct md_rdev *rdev;
+	int rv = 0;
+	/* Doing recovery so rcu locking not required */
+	rdev = sh->raid_conf->disks[disk_idx].replacement;
+	if (rdev
+	    && !test_bit(Faulty, &rdev->flags)
+	    && !test_bit(In_sync, &rdev->flags)
+	    && (rdev->recovery_offset <= sh->sector
+		|| rdev->mddev->recovery_cp <= sh->sector))
+		rv = 1;
+
+	return rv;
+}
+
+/* fetch_block - checks the given member device to see if its data needs
  * to be read or computed to satisfy a request.
  *
  * Returns 1 when no more member devices need to be checked, otherwise returns
- * 0 to tell the loop in handle_stripe_fill5 to continue
+ * 0 to tell the loop in handle_stripe_fill to continue
  */
-static int fetch_block5(struct stripe_head *sh, struct stripe_head_state *s,
-			int disk_idx, int disks)
+static int fetch_block(struct stripe_head *sh, struct stripe_head_state *s,
+		       int disk_idx, int disks)
 {
 	struct r5dev *dev = &sh->dev[disk_idx];
-	struct r5dev *failed_dev = &sh->dev[s->failed_num];
+	struct r5dev *fdev[2] = { &sh->dev[s->failed_num[0]],
+				  &sh->dev[s->failed_num[1]] };
 
 	/* is the data in this block needed, and can we get it? */
 	if (!test_bit(R5_LOCKED, &dev->flags) &&
@@ -1894,34 +2915,73 @@ static int fetch_block5(struct stripe_head *sh, struct stripe_head_state *s,
 	    (dev->toread ||
 	     (dev->towrite && !test_bit(R5_OVERWRITE, &dev->flags)) ||
 	     s->syncing || s->expanding ||
-	     (s->failed &&
-	      (failed_dev->toread ||
-	       (failed_dev->towrite &&
-		!test_bit(R5_OVERWRITE, &failed_dev->flags)))))) {
-		/* We would like to get this block, possibly by computing it,
+	     (s->replacing && want_replace(sh, disk_idx)) ||
+	     (s->failed >= 1 && fdev[0]->toread) ||
+	     (s->failed >= 2 && fdev[1]->toread) ||
+	     (sh->raid_conf->level <= 5 && s->failed && fdev[0]->towrite &&
+	      (!test_bit(R5_Insync, &dev->flags) || test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) &&
+	      !test_bit(R5_OVERWRITE, &fdev[0]->flags)) ||
+	     (sh->raid_conf->level == 6 && s->failed && s->to_write &&
+	      s->to_write < sh->raid_conf->raid_disks - 2 &&
+	      (!test_bit(R5_Insync, &dev->flags) || test_bit(STRIPE_PREREAD_ACTIVE, &sh->state))))) {
+		/* we would like to get this block, possibly by computing it,
 		 * otherwise read it if the backing disk is insync
 		 */
+		BUG_ON(test_bit(R5_Wantcompute, &dev->flags));
+		BUG_ON(test_bit(R5_Wantread, &dev->flags));
 		if ((s->uptodate == disks - 1) &&
-		    (s->failed && disk_idx == s->failed_num)) {
+		    (s->failed && (disk_idx == s->failed_num[0] ||
+				   disk_idx == s->failed_num[1]))) {
+			/* have disk failed, and we're requested to fetch it;
+			 * do compute it
+			 */
+			pr_debug("Computing stripe %llu block %d\n",
+			       (unsigned long long)sh->sector, disk_idx);
 			set_bit(STRIPE_COMPUTE_RUN, &sh->state);
 			set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
 			set_bit(R5_Wantcompute, &dev->flags);
 			sh->ops.target = disk_idx;
+			sh->ops.target2 = -1; /* no 2nd target */
 			s->req_compute = 1;
 			/* Careful: from this point on 'uptodate' is in the eye
-			 * of raid5_run_ops which services 'compute' operations
+			 * of raid_run_ops which services 'compute' operations
 			 * before writes. R5_Wantcompute flags a block that will
 			 * be R5_UPTODATE by the time it is needed for a
 			 * subsequent operation.
 			 */
 			s->uptodate++;
-			return 1; /* uptodate + compute == disks */
+			return 1;
+		} else if (s->uptodate == disks-2 && s->failed >= 2) {
+			/* Computing 2-failure is *very* expensive; only
+			 * do it if failed >= 2
+			 */
+			int other;
+			for (other = disks; other--; ) {
+				if (other == disk_idx)
+					continue;
+				if (!test_bit(R5_UPTODATE,
+				      &sh->dev[other].flags))
+					break;
+			}
+			BUG_ON(other < 0);
+			pr_debug("Computing stripe %llu blocks %d,%d\n",
+			       (unsigned long long)sh->sector,
+			       disk_idx, other);
+			set_bit(STRIPE_COMPUTE_RUN, &sh->state);
+			set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
+			set_bit(R5_Wantcompute, &sh->dev[disk_idx].flags);
+			set_bit(R5_Wantcompute, &sh->dev[other].flags);
+			sh->ops.target = disk_idx;
+			sh->ops.target2 = other;
+			s->uptodate += 2;
+			s->req_compute = 1;
+			return 1;
 		} else if (test_bit(R5_Insync, &dev->flags)) {
 			set_bit(R5_LOCKED, &dev->flags);
 			set_bit(R5_Wantread, &dev->flags);
 			s->locked++;
-			pr_debug("Reading block %d (sync=%d)\n", disk_idx,
-				s->syncing);
+			pr_debug("Reading block %d (sync=%d)\n",
+				disk_idx, s->syncing);
 		}
 	}
 
@@ -1929,10 +2989,11 @@ static int fetch_block5(struct stripe_head *sh, struct stripe_head_state *s,
 }
 
 /**
- * handle_stripe_fill5 - read or compute data to satisfy pending requests.
+ * handle_stripe_fill - read or compute data to satisfy pending requests.
  */
-static void handle_stripe_fill5(struct stripe_head *sh,
-			struct stripe_head_state *s, int disks)
+static void handle_stripe_fill(struct stripe_head *sh,
+			       struct stripe_head_state *s,
+			       int disks)
 {
 	int i;
 
@@ -1943,125 +3004,115 @@ static void handle_stripe_fill5(struct stripe_head *sh,
 	if (!test_bit(STRIPE_COMPUTE_RUN, &sh->state) && !sh->check_state &&
 	    !sh->reconstruct_state)
 		for (i = disks; i--; )
-			if (fetch_block5(sh, s, i, disks))
+			if (fetch_block(sh, s, i, disks))
 				break;
 	set_bit(STRIPE_HANDLE, &sh->state);
 }
 
-static void handle_stripe_fill6(struct stripe_head *sh,
-			struct stripe_head_state *s, struct r6_state *r6s,
-			int disks)
-{
-	int i;
-	for (i = disks; i--; ) {
-		struct r5dev *dev = &sh->dev[i];
-		if (!test_bit(R5_LOCKED, &dev->flags) &&
-		    !test_bit(R5_UPTODATE, &dev->flags) &&
-		    (dev->toread || (dev->towrite &&
-		     !test_bit(R5_OVERWRITE, &dev->flags)) ||
-		     s->syncing || s->expanding ||
-		     (s->failed >= 1 &&
-		      (sh->dev[r6s->failed_num[0]].toread ||
-		       s->to_write)) ||
-		     (s->failed >= 2 &&
-		      (sh->dev[r6s->failed_num[1]].toread ||
-		       s->to_write)))) {
-			/* we would like to get this block, possibly
-			 * by computing it, but we might not be able to
-			 */
-			if ((s->uptodate == disks - 1) &&
-			    (s->failed && (i == r6s->failed_num[0] ||
-					   i == r6s->failed_num[1]))) {
-				pr_debug("Computing stripe %llu block %d\n",
-				       (unsigned long long)sh->sector, i);
-				compute_block_1(sh, i, 0);
-				s->uptodate++;
-			} else if ( s->uptodate == disks-2 && s->failed >= 2 ) {
-				/* Computing 2-failure is *very* expensive; only
-				 * do it if failed >= 2
-				 */
-				int other;
-				for (other = disks; other--; ) {
-					if (other == i)
-						continue;
-					if (!test_bit(R5_UPTODATE,
-					      &sh->dev[other].flags))
-						break;
-				}
-				BUG_ON(other < 0);
-				pr_debug("Computing stripe %llu blocks %d,%d\n",
-				       (unsigned long long)sh->sector,
-				       i, other);
-				compute_block_2(sh, i, other);
-				s->uptodate += 2;
-			} else if (test_bit(R5_Insync, &dev->flags)) {
-				set_bit(R5_LOCKED, &dev->flags);
-				set_bit(R5_Wantread, &dev->flags);
-				s->locked++;
-				pr_debug("Reading block %d (sync=%d)\n",
-					i, s->syncing);
-			}
-		}
-	}
-	set_bit(STRIPE_HANDLE, &sh->state);
-}
-
 
 /* handle_stripe_clean_event
  * any written block on an uptodate or failed drive can be returned.
  * Note that if we 'wrote' to a failed drive, it will be UPTODATE, but
  * never LOCKED, so we don't need to test 'failed' directly.
  */
-static void handle_stripe_clean_event(raid5_conf_t *conf,
+static void handle_stripe_clean_event(struct r5conf *conf,
 	struct stripe_head *sh, int disks, struct bio **return_bi)
 {
 	int i;
 	struct r5dev *dev;
+	int discard_pending = 0;
 
 	for (i = disks; i--; )
 		if (sh->dev[i].written) {
 			dev = &sh->dev[i];
 			if (!test_bit(R5_LOCKED, &dev->flags) &&
-				test_bit(R5_UPTODATE, &dev->flags)) {
+			    (test_bit(R5_UPTODATE, &dev->flags) ||
+			     test_bit(R5_Discard, &dev->flags) ||
+			     test_bit(R5_SkipCopy, &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);
+				if (test_and_clear_bit(R5_Discard, &dev->flags))
+					clear_bit(R5_UPTODATE, &dev->flags);
+				if (test_and_clear_bit(R5_SkipCopy, &dev->flags)) {
+					WARN_ON(test_bit(R5_UPTODATE, &dev->flags));
+					dev->page = dev->orig_page;
+				}
 				wbi = dev->written;
 				dev->written = NULL;
-				while (wbi && wbi->bi_sector <
+				while (wbi && wbi->bi_iter.bi_sector <
 					dev->sector + STRIPE_SECTORS) {
 					wbi2 = r5_next_bio(wbi, dev->sector);
-					if (--wbi->bi_phys_segments == 0) {
+					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);
+			} else if (test_bit(R5_Discard, &dev->flags))
+				discard_pending = 1;
+			WARN_ON(test_bit(R5_SkipCopy, &dev->flags));
+			WARN_ON(dev->page != dev->orig_page);
 		}
+	if (!discard_pending &&
+	    test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags)) {
+		clear_bit(R5_Discard, &sh->dev[sh->pd_idx].flags);
+		clear_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags);
+		if (sh->qd_idx >= 0) {
+			clear_bit(R5_Discard, &sh->dev[sh->qd_idx].flags);
+			clear_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags);
+		}
+		/* now that discard is done we can proceed with any sync */
+		clear_bit(STRIPE_DISCARD, &sh->state);
+		/*
+		 * SCSI discard will change some bio fields and the stripe has
+		 * no updated data, so remove it from hash list and the stripe
+		 * will be reinitialized
+		 */
+		spin_lock_irq(&conf->device_lock);
+		remove_hash(sh);
+		spin_unlock_irq(&conf->device_lock);
+		if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state))
+			set_bit(STRIPE_HANDLE, &sh->state);
+
+	}
 
 	if (test_and_clear_bit(STRIPE_FULL_WRITE, &sh->state))
 		if (atomic_dec_and_test(&conf->pending_full_writes))
 			md_wakeup_thread(conf->mddev->thread);
 }
 
-static void handle_stripe_dirtying5(raid5_conf_t *conf,
-		struct stripe_head *sh,	struct stripe_head_state *s, int disks)
+static void handle_stripe_dirtying(struct r5conf *conf,
+				   struct stripe_head *sh,
+				   struct stripe_head_state *s,
+				   int disks)
 {
 	int rmw = 0, rcw = 0, i;
-	for (i = disks; i--; ) {
+	sector_t recovery_cp = conf->mddev->recovery_cp;
+
+	/* RAID6 requires 'rcw' in current implementation.
+	 * Otherwise, check whether resync is now happening or should start.
+	 * If yes, then the array is dirty (after unclean shutdown or
+	 * initial creation), so parity in some stripes might be inconsistent.
+	 * In this case, we need to always do reconstruct-write, to ensure
+	 * that in case of drive failure or read-error correction, we
+	 * generate correct data from the parity.
+	 */
+	if (conf->max_degraded == 2 ||
+	    (recovery_cp < MaxSector && sh->sector >= recovery_cp)) {
+		/* Calculate the real rcw later - for now make it
+		 * look like rcw is cheaper
+		 */
+		rcw = 1; rmw = 2;
+		pr_debug("force RCW max_degraded=%u, recovery_cp=%llu sh->sector=%llu\n",
+			 conf->max_degraded, (unsigned long long)recovery_cp,
+			 (unsigned long long)sh->sector);
+	} else for (i = disks; i--; ) {
 		/* would I have to read this buffer for read_modify_write */
 		struct r5dev *dev = &sh->dev[i];
 		if ((dev->towrite || i == sh->pd_idx) &&
@@ -2078,7 +3129,8 @@ static void handle_stripe_dirtying5(raid5_conf_t *conf,
 		    !test_bit(R5_LOCKED, &dev->flags) &&
 		    !(test_bit(R5_UPTODATE, &dev->flags) ||
 		    test_bit(R5_Wantcompute, &dev->flags))) {
-			if (test_bit(R5_Insync, &dev->flags)) rcw++;
+			if (test_bit(R5_Insync, &dev->flags))
+				rcw++;
 			else
 				rcw += 2*disks;
 		}
@@ -2086,8 +3138,12 @@ static void handle_stripe_dirtying5(raid5_conf_t *conf,
 	pr_debug("for sector %llu, rmw=%d rcw=%d\n",
 		(unsigned long long)sh->sector, rmw, rcw);
 	set_bit(STRIPE_HANDLE, &sh->state);
-	if (rmw < rcw && rmw > 0)
+	if (rmw < rcw && rmw > 0) {
 		/* prefer read-modify-write, but need to get some data */
+		if (conf->mddev->queue)
+			blk_add_trace_msg(conf->mddev->queue,
+					  "raid5 rmw %llu %d",
+					  (unsigned long long)sh->sector, rmw);
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
 			if ((dev->towrite || i == sh->pd_idx) &&
@@ -2095,10 +3151,10 @@ static void handle_stripe_dirtying5(raid5_conf_t *conf,
 			    !(test_bit(R5_UPTODATE, &dev->flags) ||
 			    test_bit(R5_Wantcompute, &dev->flags)) &&
 			    test_bit(R5_Insync, &dev->flags)) {
-				if (
-				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-					pr_debug("Read_old block "
-						"%d for r-m-w\n", i);
+				if (test_bit(STRIPE_PREREAD_ACTIVE,
+					     &sh->state)) {
+					pr_debug("Read_old block %d for r-m-w\n",
+						 i);
 					set_bit(R5_LOCKED, &dev->flags);
 					set_bit(R5_Wantread, &dev->flags);
 					s->locked++;
@@ -2108,146 +3164,56 @@ static void handle_stripe_dirtying5(raid5_conf_t *conf,
 				}
 			}
 		}
-	if (rcw <= rmw && rcw > 0)
+	}
+	if (rcw <= rmw && rcw > 0) {
 		/* want reconstruct write, but need to get some data */
+		int qread =0;
+		rcw = 0;
 		for (i = disks; i--; ) {
 			struct r5dev *dev = &sh->dev[i];
 			if (!test_bit(R5_OVERWRITE, &dev->flags) &&
-			    i != sh->pd_idx &&
+			    i != sh->pd_idx && i != sh->qd_idx &&
 			    !test_bit(R5_LOCKED, &dev->flags) &&
 			    !(test_bit(R5_UPTODATE, &dev->flags) ||
-			    test_bit(R5_Wantcompute, &dev->flags)) &&
-			    test_bit(R5_Insync, &dev->flags)) {
-				if (
-				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
+			      test_bit(R5_Wantcompute, &dev->flags))) {
+				rcw++;
+				if (test_bit(R5_Insync, &dev->flags) &&
+				    test_bit(STRIPE_PREREAD_ACTIVE,
+					     &sh->state)) {
 					pr_debug("Read_old block "
 						"%d for Reconstruct\n", i);
 					set_bit(R5_LOCKED, &dev->flags);
 					set_bit(R5_Wantread, &dev->flags);
 					s->locked++;
+					qread++;
 				} else {
 					set_bit(STRIPE_DELAYED, &sh->state);
 					set_bit(STRIPE_HANDLE, &sh->state);
 				}
 			}
 		}
+		if (rcw && conf->mddev->queue)
+			blk_add_trace_msg(conf->mddev->queue, "raid5 rcw %llu %d %d %d",
+					  (unsigned long long)sh->sector,
+					  rcw, qread, test_bit(STRIPE_DELAYED, &sh->state));
+	}
 	/* now if nothing is locked, and if we have enough data,
 	 * we can start a write request
 	 */
 	/* since handle_stripe can be called at any time we need to handle the
 	 * case where a compute block operation has been submitted and then a
-	 * subsequent call wants to start a write request.  raid5_run_ops only
-	 * handles the case where compute block and postxor are requested
+	 * subsequent call wants to start a write request.  raid_run_ops only
+	 * handles the case where compute block and reconstruct are requested
 	 * simultaneously.  If this is not the case then new writes need to be
 	 * held off until the compute completes.
 	 */
 	if ((s->req_compute || !test_bit(STRIPE_COMPUTE_RUN, &sh->state)) &&
 	    (s->locked == 0 && (rcw == 0 || rmw == 0) &&
 	    !test_bit(STRIPE_BIT_DELAY, &sh->state)))
-		schedule_reconstruction5(sh, s, rcw == 0, 0);
-}
-
-static void handle_stripe_dirtying6(raid5_conf_t *conf,
-		struct stripe_head *sh,	struct stripe_head_state *s,
-		struct r6_state *r6s, int disks)
-{
-	int rcw = 0, must_compute = 0, pd_idx = sh->pd_idx, i;
-	int qd_idx = r6s->qd_idx;
-	for (i = disks; i--; ) {
-		struct r5dev *dev = &sh->dev[i];
-		/* Would I have to read this buffer for reconstruct_write */
-		if (!test_bit(R5_OVERWRITE, &dev->flags)
-		    && i != pd_idx && i != qd_idx
-		    && (!test_bit(R5_LOCKED, &dev->flags)
-			    ) &&
-		    !test_bit(R5_UPTODATE, &dev->flags)) {
-			if (test_bit(R5_Insync, &dev->flags)) rcw++;
-			else {
-				pr_debug("raid6: must_compute: "
-					"disk %d flags=%#lx\n", i, dev->flags);
-				must_compute++;
-			}
-		}
-	}
-	pr_debug("for sector %llu, rcw=%d, must_compute=%d\n",
-	       (unsigned long long)sh->sector, rcw, must_compute);
-	set_bit(STRIPE_HANDLE, &sh->state);
-
-	if (rcw > 0)
-		/* want reconstruct write, but need to get some data */
-		for (i = disks; i--; ) {
-			struct r5dev *dev = &sh->dev[i];
-			if (!test_bit(R5_OVERWRITE, &dev->flags)
-			    && !(s->failed == 0 && (i == pd_idx || i == qd_idx))
-			    && !test_bit(R5_LOCKED, &dev->flags) &&
-			    !test_bit(R5_UPTODATE, &dev->flags) &&
-			    test_bit(R5_Insync, &dev->flags)) {
-				if (
-				  test_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-					pr_debug("Read_old stripe %llu "
-						"block %d for Reconstruct\n",
-					     (unsigned long long)sh->sector, i);
-					set_bit(R5_LOCKED, &dev->flags);
-					set_bit(R5_Wantread, &dev->flags);
-					s->locked++;
-				} else {
-					pr_debug("Request delayed stripe %llu "
-						"block %d for Reconstruct\n",
-					     (unsigned long long)sh->sector, i);
-					set_bit(STRIPE_DELAYED, &sh->state);
-					set_bit(STRIPE_HANDLE, &sh->state);
-				}
-			}
-		}
-	/* now if nothing is locked, and if we have enough data, we can start a
-	 * write request
-	 */
-	if (s->locked == 0 && rcw == 0 &&
-	    !test_bit(STRIPE_BIT_DELAY, &sh->state)) {
-		if (must_compute > 0) {
-			/* We have failed blocks and need to compute them */
-			switch (s->failed) {
-			case 0:
-				BUG();
-			case 1:
-				compute_block_1(sh, r6s->failed_num[0], 0);
-				break;
-			case 2:
-				compute_block_2(sh, r6s->failed_num[0],
-						r6s->failed_num[1]);
-				break;
-			default: /* This request should have been failed? */
-				BUG();
-			}
-		}
-
-		pr_debug("Computing parity for stripe %llu\n",
-			(unsigned long long)sh->sector);
-		compute_parity6(sh, RECONSTRUCT_WRITE);
-		/* now every locked buffer is ready to be written */
-		for (i = disks; i--; )
-			if (test_bit(R5_LOCKED, &sh->dev[i].flags)) {
-				pr_debug("Writing stripe %llu block %d\n",
-				       (unsigned long long)sh->sector, i);
-				s->locked++;
-				set_bit(R5_Wantwrite, &sh->dev[i].flags);
-			}
-		if (s->locked == disks)
-			if (!test_and_set_bit(STRIPE_FULL_WRITE, &sh->state))
-				atomic_inc(&conf->pending_full_writes);
-		/* after a RECONSTRUCT_WRITE, the stripe MUST be in-sync */
-		set_bit(STRIPE_INSYNC, &sh->state);
-
-		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-			atomic_dec(&conf->preread_active_stripes);
-			if (atomic_read(&conf->preread_active_stripes) <
-			    IO_THRESHOLD)
-				md_wakeup_thread(conf->mddev->thread);
-		}
-	}
+		schedule_reconstruction(sh, s, rcw == 0, 0);
 }
 
-static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
+static void handle_parity_checks5(struct r5conf *conf, struct stripe_head *sh,
 				struct stripe_head_state *s, int disks)
 {
 	struct r5dev *dev = NULL;
@@ -2265,7 +3231,7 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
 			s->uptodate--;
 			break;
 		}
-		dev = &sh->dev[s->failed_num];
+		dev = &sh->dev[s->failed_num[0]];
 		/* fall through */
 	case check_state_compute_result:
 		sh->check_state = check_state_idle;
@@ -2302,13 +3268,13 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
 		 * we are done.  Otherwise update the mismatch count and repair
 		 * parity if !MD_RECOVERY_CHECK
 		 */
-		if (sh->ops.zero_sum_result == 0)
+		if ((sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) == 0)
 			/* parity is correct (on disc,
 			 * not in buffer any more)
 			 */
 			set_bit(STRIPE_INSYNC, &sh->state);
 		else {
-			conf->mddev->resync_mismatches += STRIPE_SECTORS;
+			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
 			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
 				/* don't try to repair!! */
 				set_bit(STRIPE_INSYNC, &sh->state);
@@ -2319,6 +3285,7 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
 				set_bit(R5_Wantcompute,
 					&sh->dev[sh->pd_idx].flags);
 				sh->ops.target = sh->pd_idx;
+				sh->ops.target2 = -1;
 				s->uptodate++;
 			}
 		}
@@ -2334,88 +3301,94 @@ static void handle_parity_checks5(raid5_conf_t *conf, struct stripe_head *sh,
 }
 
 
-static void handle_parity_checks6(raid5_conf_t *conf, struct stripe_head *sh,
-				struct stripe_head_state *s,
-				struct r6_state *r6s, struct page *tmp_page,
-				int disks)
+static void handle_parity_checks6(struct r5conf *conf, struct stripe_head *sh,
+				  struct stripe_head_state *s,
+				  int disks)
 {
-	int update_p = 0, update_q = 0;
-	struct r5dev *dev;
 	int pd_idx = sh->pd_idx;
-	int qd_idx = r6s->qd_idx;
+	int qd_idx = sh->qd_idx;
+	struct r5dev *dev;
 
 	set_bit(STRIPE_HANDLE, &sh->state);
 
 	BUG_ON(s->failed > 2);
-	BUG_ON(s->uptodate < disks);
+
 	/* Want to check and possibly repair P and Q.
 	 * However there could be one 'failed' device, in which
 	 * case we can only check one of them, possibly using the
 	 * other to generate missing data
 	 */
 
-	/* If !tmp_page, we cannot do the calculations,
-	 * but as we have set STRIPE_HANDLE, we will soon be called
-	 * by stripe_handle with a tmp_page - just wait until then.
-	 */
-	if (tmp_page) {
-		if (s->failed == r6s->q_failed) {
-			/* The only possible failed device holds 'Q', so it
+	switch (sh->check_state) {
+	case check_state_idle:
+		/* start a new check operation if there are < 2 failures */
+		if (s->failed == s->q_failed) {
+			/* The only possible failed device holds Q, so it
 			 * makes sense to check P (If anything else were failed,
 			 * we would have used P to recreate it).
 			 */
-			compute_block_1(sh, pd_idx, 1);
-			if (!page_is_zero(sh->dev[pd_idx].page)) {
-				compute_block_1(sh, pd_idx, 0);
-				update_p = 1;
-			}
+			sh->check_state = check_state_run;
 		}
-		if (!r6s->q_failed && s->failed < 2) {
-			/* q is not failed, and we didn't use it to generate
+		if (!s->q_failed && s->failed < 2) {
+			/* Q is not failed, and we didn't use it to generate
 			 * anything, so it makes sense to check it
 			 */
-			memcpy(page_address(tmp_page),
-			       page_address(sh->dev[qd_idx].page),
-			       STRIPE_SIZE);
-			compute_parity6(sh, UPDATE_PARITY);
-			if (memcmp(page_address(tmp_page),
-				   page_address(sh->dev[qd_idx].page),
-				   STRIPE_SIZE) != 0) {
-				clear_bit(STRIPE_INSYNC, &sh->state);
-				update_q = 1;
-			}
+			if (sh->check_state == check_state_run)
+				sh->check_state = check_state_run_pq;
+			else
+				sh->check_state = check_state_run_q;
 		}
-		if (update_p || update_q) {
-			conf->mddev->resync_mismatches += STRIPE_SECTORS;
-			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
-				/* don't try to repair!! */
-				update_p = update_q = 0;
+
+		/* discard potentially stale zero_sum_result */
+		sh->ops.zero_sum_result = 0;
+
+		if (sh->check_state == check_state_run) {
+			/* async_xor_zero_sum destroys the contents of P */
+			clear_bit(R5_UPTODATE, &sh->dev[pd_idx].flags);
+			s->uptodate--;
 		}
+		if (sh->check_state >= check_state_run &&
+		    sh->check_state <= check_state_run_pq) {
+			/* async_syndrome_zero_sum preserves P and Q, so
+			 * no need to mark them !uptodate here
+			 */
+			set_bit(STRIPE_OP_CHECK, &s->ops_request);
+			break;
+		}
+
+		/* we have 2-disk failure */
+		BUG_ON(s->failed != 2);
+		/* fall through */
+	case check_state_compute_result:
+		sh->check_state = check_state_idle;
+
+		/* check that a write has not made the stripe insync */
+		if (test_bit(STRIPE_INSYNC, &sh->state))
+			break;
 
 		/* now write out any block on a failed drive,
-		 * or P or Q if they need it
+		 * or P or Q if they were recomputed
 		 */
-
+		BUG_ON(s->uptodate < disks - 1); /* We don't need Q to recover */
 		if (s->failed == 2) {
-			dev = &sh->dev[r6s->failed_num[1]];
+			dev = &sh->dev[s->failed_num[1]];
 			s->locked++;
 			set_bit(R5_LOCKED, &dev->flags);
 			set_bit(R5_Wantwrite, &dev->flags);
 		}
 		if (s->failed >= 1) {
-			dev = &sh->dev[r6s->failed_num[0]];
+			dev = &sh->dev[s->failed_num[0]];
 			s->locked++;
 			set_bit(R5_LOCKED, &dev->flags);
 			set_bit(R5_Wantwrite, &dev->flags);
 		}
-
-		if (update_p) {
+		if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
 			dev = &sh->dev[pd_idx];
 			s->locked++;
 			set_bit(R5_LOCKED, &dev->flags);
 			set_bit(R5_Wantwrite, &dev->flags);
 		}
-		if (update_q) {
+		if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
 			dev = &sh->dev[qd_idx];
 			s->locked++;
 			set_bit(R5_LOCKED, &dev->flags);
@@ -2424,11 +3397,74 @@ static void handle_parity_checks6(raid5_conf_t *conf, struct stripe_head *sh,
 		clear_bit(STRIPE_DEGRADED, &sh->state);
 
 		set_bit(STRIPE_INSYNC, &sh->state);
+		break;
+	case check_state_run:
+	case check_state_run_q:
+	case check_state_run_pq:
+		break; /* we will be called again upon completion */
+	case check_state_check_result:
+		sh->check_state = check_state_idle;
+
+		/* handle a successful check operation, if parity is correct
+		 * we are done.  Otherwise update the mismatch count and repair
+		 * parity if !MD_RECOVERY_CHECK
+		 */
+		if (sh->ops.zero_sum_result == 0) {
+			/* both parities are correct */
+			if (!s->failed)
+				set_bit(STRIPE_INSYNC, &sh->state);
+			else {
+				/* in contrast to the raid5 case we can validate
+				 * parity, but still have a failure to write
+				 * back
+				 */
+				sh->check_state = check_state_compute_result;
+				/* Returning at this point means that we may go
+				 * off and bring p and/or q uptodate again so
+				 * we make sure to check zero_sum_result again
+				 * to verify if p or q need writeback
+				 */
+			}
+		} else {
+			atomic64_add(STRIPE_SECTORS, &conf->mddev->resync_mismatches);
+			if (test_bit(MD_RECOVERY_CHECK, &conf->mddev->recovery))
+				/* don't try to repair!! */
+				set_bit(STRIPE_INSYNC, &sh->state);
+			else {
+				int *target = &sh->ops.target;
+
+				sh->ops.target = -1;
+				sh->ops.target2 = -1;
+				sh->check_state = check_state_compute_run;
+				set_bit(STRIPE_COMPUTE_RUN, &sh->state);
+				set_bit(STRIPE_OP_COMPUTE_BLK, &s->ops_request);
+				if (sh->ops.zero_sum_result & SUM_CHECK_P_RESULT) {
+					set_bit(R5_Wantcompute,
+						&sh->dev[pd_idx].flags);
+					*target = pd_idx;
+					target = &sh->ops.target2;
+					s->uptodate++;
+				}
+				if (sh->ops.zero_sum_result & SUM_CHECK_Q_RESULT) {
+					set_bit(R5_Wantcompute,
+						&sh->dev[qd_idx].flags);
+					*target = qd_idx;
+					s->uptodate++;
+				}
+			}
+		}
+		break;
+	case check_state_compute_run:
+		break;
+	default:
+		printk(KERN_ERR "%s: unknown check_state: %d sector: %llu\n",
+		       __func__, sh->check_state,
+		       (unsigned long long) sh->sector);
+		BUG();
 	}
 }
 
-static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh,
-				struct r6_state *r6s)
+static void handle_stripe_expansion(struct r5conf *conf, struct stripe_head *sh)
 {
 	int i;
 
@@ -2438,17 +3474,15 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh,
 	struct dma_async_tx_descriptor *tx = NULL;
 	clear_bit(STRIPE_EXPAND_SOURCE, &sh->state);
 	for (i = 0; i < sh->disks; i++)
-		if (i != sh->pd_idx && (!r6s || i != r6s->qd_idx)) {
-			int dd_idx, pd_idx, j;
+		if (i != sh->pd_idx && i != sh->qd_idx) {
+			int dd_idx, j;
 			struct stripe_head *sh2;
+			struct async_submit_ctl submit;
 
-			sector_t bn = compute_blocknr(sh, i);
-			sector_t s = raid5_compute_sector(bn, conf->raid_disks,
-						conf->raid_disks -
-						conf->max_degraded, &dd_idx,
-						&pd_idx, conf);
-			sh2 = get_active_stripe(conf, s, conf->raid_disks,
-						pd_idx, 1);
+			sector_t bn = compute_blocknr(sh, i, 1);
+			sector_t s = raid5_compute_sector(conf, bn, 0,
+							  &dd_idx, NULL);
+			sh2 = get_active_stripe(conf, s, 0, 1, 1);
 			if (sh2 == NULL)
 				/* so far only the early blocks of this stripe
 				 * have been requested.  When later blocks
@@ -2463,16 +3497,16 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh,
 			}
 
 			/* place all the copies on one channel */
+			init_async_submit(&submit, 0, tx, NULL, NULL, NULL);
 			tx = async_memcpy(sh2->dev[dd_idx].page,
-				sh->dev[i].page, 0, 0, STRIPE_SIZE,
-				ASYNC_TX_DEP_ACK, tx, NULL, NULL);
+					  sh->dev[i].page, 0, 0, STRIPE_SIZE,
+					  &submit);
 
 			set_bit(R5_Expanded, &sh2->dev[dd_idx].flags);
 			set_bit(R5_UPTODATE, &sh2->dev[dd_idx].flags);
 			for (j = 0; j < conf->raid_disks; j++)
 				if (j != sh2->pd_idx &&
-				    (!r6s || j != raid6_next_disk(sh2->pd_idx,
-								 sh2->disks)) &&
+				    j != sh2->qd_idx &&
 				    !test_bit(R5_Expanded, &sh2->dev[j].flags))
 					break;
 			if (j == conf->raid_disks) {
@@ -2483,66 +3517,52 @@ static void handle_stripe_expansion(raid5_conf_t *conf, struct stripe_head *sh,
 
 		}
 	/* done submitting copies, wait for them to complete */
-	if (tx) {
-		async_tx_ack(tx);
-		dma_wait_for_async_tx(tx);
-	}
+	async_tx_quiesce(&tx);
 }
 
-
 /*
  * handle_stripe - do things to a stripe.
  *
- * We lock the stripe and then examine the state of various bits
- * to see what needs to be done.
+ * We lock the stripe by setting STRIPE_ACTIVE and then examine the
+ * state of various bits to see what needs to be done.
  * Possible results:
- *    return some read request which now have data
- *    return some write requests which are safely on disc
+ *    return some read requests which now have data
+ *    return some write requests which are safely on storage
  *    schedule a read on some buffers
  *    schedule a write of some buffers
  *    return confirmation of parity correctness
  *
- * buffers are taken off read_list or write_list, and bh_cache buffers
- * get BH_Lock set before the stripe lock is released.
- *
  */
 
-static bool handle_stripe5(struct stripe_head *sh)
+static void analyse_stripe(struct stripe_head *sh, struct stripe_head_state *s)
 {
-	raid5_conf_t *conf = sh->raid_conf;
-	int disks = sh->disks, i;
-	struct bio *return_bi = NULL;
-	struct stripe_head_state s;
+	struct r5conf *conf = sh->raid_conf;
+	int disks = sh->disks;
 	struct r5dev *dev;
-	mdk_rdev_t *blocked_rdev = NULL;
-	int prexor;
-
-	memset(&s, 0, sizeof(s));
-	pr_debug("handling stripe %llu, state=%#lx cnt=%d, pd_idx=%d check:%d "
-		 "reconstruct:%d\n", (unsigned long long)sh->sector, sh->state,
-		 atomic_read(&sh->count), sh->pd_idx, sh->check_state,
-		 sh->reconstruct_state);
+	int i;
+	int do_recovery = 0;
 
-	spin_lock(&sh->lock);
-	clear_bit(STRIPE_HANDLE, &sh->state);
-	clear_bit(STRIPE_DELAYED, &sh->state);
+	memset(s, 0, sizeof(*s));
 
-	s.syncing = test_bit(STRIPE_SYNCING, &sh->state);
-	s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
-	s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
+	s->expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
+	s->expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
+	s->failed_num[0] = -1;
+	s->failed_num[1] = -1;
 
 	/* Now to look around and see what can be done */
 	rcu_read_lock();
 	for (i=disks; i--; ) {
-		mdk_rdev_t *rdev;
-		struct r5dev *dev = &sh->dev[i];
-		clear_bit(R5_Insync, &dev->flags);
+		struct md_rdev *rdev;
+		sector_t first_bad;
+		int bad_sectors;
+		int is_bad = 0;
 
-		pr_debug("check %d: state 0x%lx toread %p read %p write %p "
-			"written %p\n",	i, dev->flags, dev->toread, dev->read,
-			dev->towrite, dev->written);
+		dev = &sh->dev[i];
 
-		/* maybe we can request a biofill operation
+		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
+			 i, dev->flags,
+			 dev->toread, dev->towrite, dev->written);
+		/* maybe we can reply to a read
 		 *
 		 * new wantfill requests are only permitted while
 		 * ops_complete_biofill is guaranteed to be inactive
@@ -2552,50 +3572,201 @@ static bool handle_stripe5(struct stripe_head *sh)
 			set_bit(R5_Wantfill, &dev->flags);
 
 		/* now count some things */
-		if (test_bit(R5_LOCKED, &dev->flags)) s.locked++;
-		if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++;
-		if (test_bit(R5_Wantcompute, &dev->flags)) s.compute++;
+		if (test_bit(R5_LOCKED, &dev->flags))
+			s->locked++;
+		if (test_bit(R5_UPTODATE, &dev->flags))
+			s->uptodate++;
+		if (test_bit(R5_Wantcompute, &dev->flags)) {
+			s->compute++;
+			BUG_ON(s->compute > 2);
+		}
 
 		if (test_bit(R5_Wantfill, &dev->flags))
-			s.to_fill++;
+			s->to_fill++;
 		else if (dev->toread)
-			s.to_read++;
+			s->to_read++;
 		if (dev->towrite) {
-			s.to_write++;
+			s->to_write++;
 			if (!test_bit(R5_OVERWRITE, &dev->flags))
-				s.non_overwrite++;
+				s->non_overwrite++;
 		}
 		if (dev->written)
-			s.written++;
-		rdev = rcu_dereference(conf->disks[i].rdev);
-		if (blocked_rdev == NULL &&
-		    rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
-			blocked_rdev = rdev;
-			atomic_inc(&rdev->nr_pending);
+			s->written++;
+		/* Prefer to use the replacement for reads, but only
+		 * if it is recovered enough and has no bad blocks.
+		 */
+		rdev = rcu_dereference(conf->disks[i].replacement);
+		if (rdev && !test_bit(Faulty, &rdev->flags) &&
+		    rdev->recovery_offset >= sh->sector + STRIPE_SECTORS &&
+		    !is_badblock(rdev, sh->sector, STRIPE_SECTORS,
+				 &first_bad, &bad_sectors))
+			set_bit(R5_ReadRepl, &dev->flags);
+		else {
+			if (rdev)
+				set_bit(R5_NeedReplace, &dev->flags);
+			rdev = rcu_dereference(conf->disks[i].rdev);
+			clear_bit(R5_ReadRepl, &dev->flags);
+		}
+		if (rdev && test_bit(Faulty, &rdev->flags))
+			rdev = NULL;
+		if (rdev) {
+			is_bad = is_badblock(rdev, sh->sector, STRIPE_SECTORS,
+					     &first_bad, &bad_sectors);
+			if (s->blocked_rdev == NULL
+			    && (test_bit(Blocked, &rdev->flags)
+				|| is_bad < 0)) {
+				if (is_bad < 0)
+					set_bit(BlockedBadBlocks,
+						&rdev->flags);
+				s->blocked_rdev = rdev;
+				atomic_inc(&rdev->nr_pending);
+			}
+		}
+		clear_bit(R5_Insync, &dev->flags);
+		if (!rdev)
+			/* Not in-sync */;
+		else if (is_bad) {
+			/* also not in-sync */
+			if (!test_bit(WriteErrorSeen, &rdev->flags) &&
+			    test_bit(R5_UPTODATE, &dev->flags)) {
+				/* treat as in-sync, but with a read error
+				 * which we can now try to correct
+				 */
+				set_bit(R5_Insync, &dev->flags);
+				set_bit(R5_ReadError, &dev->flags);
+			}
+		} else if (test_bit(In_sync, &rdev->flags))
+			set_bit(R5_Insync, &dev->flags);
+		else if (sh->sector + STRIPE_SECTORS <= rdev->recovery_offset)
+			/* in sync if before recovery_offset */
+			set_bit(R5_Insync, &dev->flags);
+		else if (test_bit(R5_UPTODATE, &dev->flags) &&
+			 test_bit(R5_Expanded, &dev->flags))
+			/* If we've reshaped into here, we assume it is Insync.
+			 * We will shortly update recovery_offset to make
+			 * it official.
+			 */
+			set_bit(R5_Insync, &dev->flags);
+
+		if (test_bit(R5_WriteError, &dev->flags)) {
+			/* This flag does not apply to '.replacement'
+			 * only to .rdev, so make sure to check that*/
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].rdev);
+			if (rdev2 == rdev)
+				clear_bit(R5_Insync, &dev->flags);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
+				s->handle_bad_blocks = 1;
+				atomic_inc(&rdev2->nr_pending);
+			} else
+				clear_bit(R5_WriteError, &dev->flags);
+		}
+		if (test_bit(R5_MadeGood, &dev->flags)) {
+			/* This flag does not apply to '.replacement'
+			 * only to .rdev, so make sure to check that*/
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].rdev);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
+				s->handle_bad_blocks = 1;
+				atomic_inc(&rdev2->nr_pending);
+			} else
+				clear_bit(R5_MadeGood, &dev->flags);
+		}
+		if (test_bit(R5_MadeGoodRepl, &dev->flags)) {
+			struct md_rdev *rdev2 = rcu_dereference(
+				conf->disks[i].replacement);
+			if (rdev2 && !test_bit(Faulty, &rdev2->flags)) {
+				s->handle_bad_blocks = 1;
+				atomic_inc(&rdev2->nr_pending);
+			} else
+				clear_bit(R5_MadeGoodRepl, &dev->flags);
 		}
-		if (!rdev || !test_bit(In_sync, &rdev->flags)) {
+		if (!test_bit(R5_Insync, &dev->flags)) {
 			/* The ReadError flag will just be confusing now */
 			clear_bit(R5_ReadError, &dev->flags);
 			clear_bit(R5_ReWrite, &dev->flags);
 		}
-		if (!rdev || !test_bit(In_sync, &rdev->flags)
-		    || test_bit(R5_ReadError, &dev->flags)) {
-			s.failed++;
-			s.failed_num = i;
-		} else
-			set_bit(R5_Insync, &dev->flags);
+		if (test_bit(R5_ReadError, &dev->flags))
+			clear_bit(R5_Insync, &dev->flags);
+		if (!test_bit(R5_Insync, &dev->flags)) {
+			if (s->failed < 2)
+				s->failed_num[s->failed] = i;
+			s->failed++;
+			if (rdev && !test_bit(Faulty, &rdev->flags))
+				do_recovery = 1;
+		}
+	}
+	if (test_bit(STRIPE_SYNCING, &sh->state)) {
+		/* If there is a failed device being replaced,
+		 *     we must be recovering.
+		 * else if we are after recovery_cp, we must be syncing
+		 * else if MD_RECOVERY_REQUESTED is set, we also are syncing.
+		 * else we can only be replacing
+		 * sync and recovery both need to read all devices, and so
+		 * use the same flag.
+		 */
+		if (do_recovery ||
+		    sh->sector >= conf->mddev->recovery_cp ||
+		    test_bit(MD_RECOVERY_REQUESTED, &(conf->mddev->recovery)))
+			s->syncing = 1;
+		else
+			s->replacing = 1;
 	}
 	rcu_read_unlock();
+}
+
+static void handle_stripe(struct stripe_head *sh)
+{
+	struct stripe_head_state s;
+	struct r5conf *conf = sh->raid_conf;
+	int i;
+	int prexor;
+	int disks = sh->disks;
+	struct r5dev *pdev, *qdev;
+
+	clear_bit(STRIPE_HANDLE, &sh->state);
+	if (test_and_set_bit_lock(STRIPE_ACTIVE, &sh->state)) {
+		/* already being handled, ensure it gets handled
+		 * again when current action finishes */
+		set_bit(STRIPE_HANDLE, &sh->state);
+		return;
+	}
+
+	if (test_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
+		spin_lock(&sh->stripe_lock);
+		/* Cannot process 'sync' concurrently with 'discard' */
+		if (!test_bit(STRIPE_DISCARD, &sh->state) &&
+		    test_and_clear_bit(STRIPE_SYNC_REQUESTED, &sh->state)) {
+			set_bit(STRIPE_SYNCING, &sh->state);
+			clear_bit(STRIPE_INSYNC, &sh->state);
+			clear_bit(STRIPE_REPLACED, &sh->state);
+		}
+		spin_unlock(&sh->stripe_lock);
+	}
+	clear_bit(STRIPE_DELAYED, &sh->state);
 
-	if (unlikely(blocked_rdev)) {
+	pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
+		"pd_idx=%d, qd_idx=%d\n, check:%d, reconstruct:%d\n",
+	       (unsigned long long)sh->sector, sh->state,
+	       atomic_read(&sh->count), sh->pd_idx, sh->qd_idx,
+	       sh->check_state, sh->reconstruct_state);
+
+	analyse_stripe(sh, &s);
+
+	if (s.handle_bad_blocks) {
+		set_bit(STRIPE_HANDLE, &sh->state);
+		goto finish;
+	}
+
+	if (unlikely(s.blocked_rdev)) {
 		if (s.syncing || s.expanding || s.expanded ||
-		    s.to_write || s.written) {
+		    s.replacing || s.to_write || s.written) {
 			set_bit(STRIPE_HANDLE, &sh->state);
-			goto unlock;
+			goto finish;
 		}
 		/* There is nothing for the blocked_rdev to block */
-		rdev_dec_pending(blocked_rdev, conf->mddev);
-		blocked_rdev = NULL;
+		rdev_dec_pending(s.blocked_rdev, conf->mddev);
+		s.blocked_rdev = NULL;
 	}
 
 	if (s.to_fill && !test_bit(STRIPE_BIOFILL_RUN, &sh->state)) {
@@ -2604,39 +3775,21 @@ static bool handle_stripe5(struct stripe_head *sh)
 	}
 
 	pr_debug("locked=%d uptodate=%d to_read=%d"
-		" to_write=%d failed=%d failed_num=%d\n",
-		s.locked, s.uptodate, s.to_read, s.to_write,
-		s.failed, s.failed_num);
-	/* check if the array has lost two devices and, if so, some requests might
-	 * need to be failed
+	       " to_write=%d failed=%d failed_num=%d,%d\n",
+	       s.locked, s.uptodate, s.to_read, s.to_write, s.failed,
+	       s.failed_num[0], s.failed_num[1]);
+	/* check if the array has lost more than max_degraded devices and,
+	 * if so, some requests might need to be failed.
 	 */
-	if (s.failed > 1 && s.to_read+s.to_write+s.written)
-		handle_failed_stripe(conf, sh, &s, disks, &return_bi);
-	if (s.failed > 1 && s.syncing) {
-		md_done_sync(conf->mddev, STRIPE_SECTORS,0);
-		clear_bit(STRIPE_SYNCING, &sh->state);
-		s.syncing = 0;
+	if (s.failed > conf->max_degraded) {
+		sh->check_state = 0;
+		sh->reconstruct_state = 0;
+		if (s.to_read+s.to_write+s.written)
+			handle_failed_stripe(conf, sh, &s, disks, &s.return_bi);
+		if (s.syncing + s.replacing)
+			handle_failed_sync(conf, sh, &s);
 	}
 
-	/* might be able to return some write requests if the parity block
-	 * is safe, or on a failed drive
-	 */
-	dev = &sh->dev[sh->pd_idx];
-	if ( s.written &&
-	     ((test_bit(R5_Insync, &dev->flags) &&
-	       !test_bit(R5_LOCKED, &dev->flags) &&
-	       test_bit(R5_UPTODATE, &dev->flags)) ||
-	       (s.failed == 1 && s.failed_num == sh->pd_idx)))
-		handle_stripe_clean_event(conf, sh, disks, &return_bi);
-
-	/* Now we might consider reading some blocks, either to check/generate
-	 * parity, or to satisfy requests
-	 * or to load a block that is being partially written.
-	 */
-	if (s.to_read || s.non_overwrite ||
-	    (s.syncing && (s.uptodate + s.compute < disks)) || s.expanding)
-		handle_stripe_fill5(sh, &s, disks);
-
 	/* Now we check to see if any write operations have recently
 	 * completed
 	 */
@@ -2650,28 +3803,64 @@ static bool handle_stripe5(struct stripe_head *sh)
 		/* All the 'written' buffers and the parity block are ready to
 		 * be written back to disk
 		 */
-		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags));
+		BUG_ON(!test_bit(R5_UPTODATE, &sh->dev[sh->pd_idx].flags) &&
+		       !test_bit(R5_Discard, &sh->dev[sh->pd_idx].flags));
+		BUG_ON(sh->qd_idx >= 0 &&
+		       !test_bit(R5_UPTODATE, &sh->dev[sh->qd_idx].flags) &&
+		       !test_bit(R5_Discard, &sh->dev[sh->qd_idx].flags));
 		for (i = disks; i--; ) {
-			dev = &sh->dev[i];
+			struct r5dev *dev = &sh->dev[i];
 			if (test_bit(R5_LOCKED, &dev->flags) &&
-				(i == sh->pd_idx || dev->written)) {
+				(i == sh->pd_idx || i == sh->qd_idx ||
+				 dev->written)) {
 				pr_debug("Writing block %d\n", i);
 				set_bit(R5_Wantwrite, &dev->flags);
 				if (prexor)
 					continue;
 				if (!test_bit(R5_Insync, &dev->flags) ||
-				    (i == sh->pd_idx && s.failed == 0))
+				    ((i == sh->pd_idx || i == sh->qd_idx)  &&
+				     s.failed == 0))
 					set_bit(STRIPE_INSYNC, &sh->state);
 			}
 		}
-		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state)) {
-			atomic_dec(&conf->preread_active_stripes);
-			if (atomic_read(&conf->preread_active_stripes) <
-				IO_THRESHOLD)
-				md_wakeup_thread(conf->mddev->thread);
-		}
+		if (test_and_clear_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			s.dec_preread_active = 1;
 	}
 
+	/*
+	 * might be able to return some write requests if the parity blocks
+	 * are safe, or on a failed drive
+	 */
+	pdev = &sh->dev[sh->pd_idx];
+	s.p_failed = (s.failed >= 1 && s.failed_num[0] == sh->pd_idx)
+		|| (s.failed >= 2 && s.failed_num[1] == sh->pd_idx);
+	qdev = &sh->dev[sh->qd_idx];
+	s.q_failed = (s.failed >= 1 && s.failed_num[0] == sh->qd_idx)
+		|| (s.failed >= 2 && s.failed_num[1] == sh->qd_idx)
+		|| conf->level < 6;
+
+	if (s.written &&
+	    (s.p_failed || ((test_bit(R5_Insync, &pdev->flags)
+			     && !test_bit(R5_LOCKED, &pdev->flags)
+			     && (test_bit(R5_UPTODATE, &pdev->flags) ||
+				 test_bit(R5_Discard, &pdev->flags))))) &&
+	    (s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
+			     && !test_bit(R5_LOCKED, &qdev->flags)
+			     && (test_bit(R5_UPTODATE, &qdev->flags) ||
+				 test_bit(R5_Discard, &qdev->flags))))))
+		handle_stripe_clean_event(conf, sh, disks, &s.return_bi);
+
+	/* Now we might consider reading some blocks, either to check/generate
+	 * parity, or to satisfy requests
+	 * or to load a block that is being partially written.
+	 */
+	if (s.to_read || s.non_overwrite
+	    || (conf->level == 6 && s.to_write && s.failed)
+	    || (s.syncing && (s.uptodate + s.compute < disks))
+	    || s.replacing
+	    || s.expanding)
+		handle_stripe_fill(sh, &s, disks);
+
 	/* Now to consider new write requests and what else, if anything
 	 * should be read.  We do not handle new writes when:
 	 * 1/ A 'write' operation (copy+xor) is already in flight.
@@ -2679,7 +3868,7 @@ static bool handle_stripe5(struct stripe_head *sh)
 	 *    block.
 	 */
 	if (s.to_write && !sh->reconstruct_state && !sh->check_state)
-		handle_stripe_dirtying5(conf, sh, &s, disks);
+		handle_stripe_dirtying(conf, sh, &s, disks);
 
 	/* maybe we need to check and possibly fix the parity for this stripe
 	 * Any reads will already have been scheduled, so we just see if enough
@@ -2689,251 +3878,43 @@ static bool handle_stripe5(struct stripe_head *sh)
 	if (sh->check_state ||
 	    (s.syncing && s.locked == 0 &&
 	     !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
-	     !test_bit(STRIPE_INSYNC, &sh->state)))
-		handle_parity_checks5(conf, sh, &s, disks);
-
-	if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
-		md_done_sync(conf->mddev, STRIPE_SECTORS,1);
-		clear_bit(STRIPE_SYNCING, &sh->state);
-	}
-
-	/* If the failed drive is just a ReadError, then we might need to progress
-	 * the repair/check process
-	 */
-	if (s.failed == 1 && !conf->mddev->ro &&
-	    test_bit(R5_ReadError, &sh->dev[s.failed_num].flags)
-	    && !test_bit(R5_LOCKED, &sh->dev[s.failed_num].flags)
-	    && test_bit(R5_UPTODATE, &sh->dev[s.failed_num].flags)
-		) {
-		dev = &sh->dev[s.failed_num];
-		if (!test_bit(R5_ReWrite, &dev->flags)) {
-			set_bit(R5_Wantwrite, &dev->flags);
-			set_bit(R5_ReWrite, &dev->flags);
-			set_bit(R5_LOCKED, &dev->flags);
-			s.locked++;
-		} else {
-			/* let's read it back */
-			set_bit(R5_Wantread, &dev->flags);
-			set_bit(R5_LOCKED, &dev->flags);
-			s.locked++;
-		}
-	}
-
-	/* Finish reconstruct operations initiated by the expansion process */
-	if (sh->reconstruct_state == reconstruct_state_result) {
-		sh->reconstruct_state = reconstruct_state_idle;
-		clear_bit(STRIPE_EXPANDING, &sh->state);
-		for (i = conf->raid_disks; i--; ) {
-			set_bit(R5_Wantwrite, &sh->dev[i].flags);
-			set_bit(R5_LOCKED, &sh->dev[i].flags);
-			s.locked++;
-		}
-	}
-
-	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state) &&
-	    !sh->reconstruct_state) {
-		/* Need to write out all blocks after computing parity */
-		sh->disks = conf->raid_disks;
-		sh->pd_idx = stripe_to_pdidx(sh->sector, conf,
-			conf->raid_disks);
-		schedule_reconstruction5(sh, &s, 1, 1);
-	} else if (s.expanded && !sh->reconstruct_state && s.locked == 0) {
-		clear_bit(STRIPE_EXPAND_READY, &sh->state);
-		atomic_dec(&conf->reshape_stripes);
-		wake_up(&conf->wait_for_overlap);
-		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
+	     !test_bit(STRIPE_INSYNC, &sh->state))) {
+		if (conf->level == 6)
+			handle_parity_checks6(conf, sh, &s, disks);
+		else
+			handle_parity_checks5(conf, sh, &s, disks);
 	}
 
-	if (s.expanding && s.locked == 0 &&
-	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state))
-		handle_stripe_expansion(conf, sh, NULL);
-
- unlock:
-	spin_unlock(&sh->lock);
-
-	/* wait for this device to become unblocked */
-	if (unlikely(blocked_rdev))
-		md_wait_for_blocked_rdev(blocked_rdev, conf->mddev);
-
-	if (s.ops_request)
-		raid5_run_ops(sh, s.ops_request);
-
-	ops_run_io(sh, &s);
-
-	return_io(return_bi);
-
-	return blocked_rdev == NULL;
-}
-
-static bool handle_stripe6(struct stripe_head *sh, struct page *tmp_page)
-{
-	raid6_conf_t *conf = sh->raid_conf;
-	int disks = sh->disks;
-	struct bio *return_bi = NULL;
-	int i, pd_idx = sh->pd_idx;
-	struct stripe_head_state s;
-	struct r6_state r6s;
-	struct r5dev *dev, *pdev, *qdev;
-	mdk_rdev_t *blocked_rdev = NULL;
-
-	r6s.qd_idx = raid6_next_disk(pd_idx, disks);
-	pr_debug("handling stripe %llu, state=%#lx cnt=%d, "
-		"pd_idx=%d, qd_idx=%d\n",
-	       (unsigned long long)sh->sector, sh->state,
-	       atomic_read(&sh->count), pd_idx, r6s.qd_idx);
-	memset(&s, 0, sizeof(s));
-
-	spin_lock(&sh->lock);
-	clear_bit(STRIPE_HANDLE, &sh->state);
-	clear_bit(STRIPE_DELAYED, &sh->state);
-
-	s.syncing = test_bit(STRIPE_SYNCING, &sh->state);
-	s.expanding = test_bit(STRIPE_EXPAND_SOURCE, &sh->state);
-	s.expanded = test_bit(STRIPE_EXPAND_READY, &sh->state);
-	/* Now to look around and see what can be done */
-
-	rcu_read_lock();
-	for (i=disks; i--; ) {
-		mdk_rdev_t *rdev;
-		dev = &sh->dev[i];
-		clear_bit(R5_Insync, &dev->flags);
-
-		pr_debug("check %d: state 0x%lx read %p write %p written %p\n",
-			i, dev->flags, dev->toread, dev->towrite, dev->written);
-		/* maybe we can reply to a read */
-		if (test_bit(R5_UPTODATE, &dev->flags) && dev->toread) {
-			struct bio *rbi, *rbi2;
-			pr_debug("Return read for disc %d\n", i);
-			spin_lock_irq(&conf->device_lock);
-			rbi = dev->toread;
-			dev->toread = NULL;
-			if (test_and_clear_bit(R5_Overlap, &dev->flags))
-				wake_up(&conf->wait_for_overlap);
-			spin_unlock_irq(&conf->device_lock);
-			while (rbi && rbi->bi_sector < dev->sector + STRIPE_SECTORS) {
-				copy_data(0, rbi, dev->page, dev->sector);
-				rbi2 = r5_next_bio(rbi, dev->sector);
-				spin_lock_irq(&conf->device_lock);
-				if (--rbi->bi_phys_segments == 0) {
-					rbi->bi_next = return_bi;
-					return_bi = rbi;
-				}
-				spin_unlock_irq(&conf->device_lock);
-				rbi = rbi2;
+	if ((s.replacing || s.syncing) && s.locked == 0
+	    && !test_bit(STRIPE_COMPUTE_RUN, &sh->state)
+	    && !test_bit(STRIPE_REPLACED, &sh->state)) {
+		/* Write out to replacement devices where possible */
+		for (i = 0; i < conf->raid_disks; i++)
+			if (test_bit(R5_NeedReplace, &sh->dev[i].flags)) {
+				WARN_ON(!test_bit(R5_UPTODATE, &sh->dev[i].flags));
+				set_bit(R5_WantReplace, &sh->dev[i].flags);
+				set_bit(R5_LOCKED, &sh->dev[i].flags);
+				s.locked++;
 			}
-		}
-
-		/* now count some things */
-		if (test_bit(R5_LOCKED, &dev->flags)) s.locked++;
-		if (test_bit(R5_UPTODATE, &dev->flags)) s.uptodate++;
-
-
-		if (dev->toread)
-			s.to_read++;
-		if (dev->towrite) {
-			s.to_write++;
-			if (!test_bit(R5_OVERWRITE, &dev->flags))
-				s.non_overwrite++;
-		}
-		if (dev->written)
-			s.written++;
-		rdev = rcu_dereference(conf->disks[i].rdev);
-		if (blocked_rdev == NULL &&
-		    rdev && unlikely(test_bit(Blocked, &rdev->flags))) {
-			blocked_rdev = rdev;
-			atomic_inc(&rdev->nr_pending);
-		}
-		if (!rdev || !test_bit(In_sync, &rdev->flags)) {
-			/* The ReadError flag will just be confusing now */
-			clear_bit(R5_ReadError, &dev->flags);
-			clear_bit(R5_ReWrite, &dev->flags);
-		}
-		if (!rdev || !test_bit(In_sync, &rdev->flags)
-		    || test_bit(R5_ReadError, &dev->flags)) {
-			if (s.failed < 2)
-				r6s.failed_num[s.failed] = i;
-			s.failed++;
-		} else
-			set_bit(R5_Insync, &dev->flags);
-	}
-	rcu_read_unlock();
-
-	if (unlikely(blocked_rdev)) {
-		if (s.syncing || s.expanding || s.expanded ||
-		    s.to_write || s.written) {
-			set_bit(STRIPE_HANDLE, &sh->state);
-			goto unlock;
-		}
-		/* There is nothing for the blocked_rdev to block */
-		rdev_dec_pending(blocked_rdev, conf->mddev);
-		blocked_rdev = NULL;
-	}
-
-	pr_debug("locked=%d uptodate=%d to_read=%d"
-	       " to_write=%d failed=%d failed_num=%d,%d\n",
-	       s.locked, s.uptodate, s.to_read, s.to_write, s.failed,
-	       r6s.failed_num[0], r6s.failed_num[1]);
-	/* check if the array has lost >2 devices and, if so, some requests
-	 * might need to be failed
-	 */
-	if (s.failed > 2 && s.to_read+s.to_write+s.written)
-		handle_failed_stripe(conf, sh, &s, disks, &return_bi);
-	if (s.failed > 2 && s.syncing) {
-		md_done_sync(conf->mddev, STRIPE_SECTORS,0);
-		clear_bit(STRIPE_SYNCING, &sh->state);
-		s.syncing = 0;
+		if (s.replacing)
+			set_bit(STRIPE_INSYNC, &sh->state);
+		set_bit(STRIPE_REPLACED, &sh->state);
 	}
-
-	/*
-	 * might be able to return some write requests if the parity blocks
-	 * are safe, or on a failed drive
-	 */
-	pdev = &sh->dev[pd_idx];
-	r6s.p_failed = (s.failed >= 1 && r6s.failed_num[0] == pd_idx)
-		|| (s.failed >= 2 && r6s.failed_num[1] == pd_idx);
-	qdev = &sh->dev[r6s.qd_idx];
-	r6s.q_failed = (s.failed >= 1 && r6s.failed_num[0] == r6s.qd_idx)
-		|| (s.failed >= 2 && r6s.failed_num[1] == r6s.qd_idx);
-
-	if ( s.written &&
-	     ( r6s.p_failed || ((test_bit(R5_Insync, &pdev->flags)
-			     && !test_bit(R5_LOCKED, &pdev->flags)
-			     && test_bit(R5_UPTODATE, &pdev->flags)))) &&
-	     ( r6s.q_failed || ((test_bit(R5_Insync, &qdev->flags)
-			     && !test_bit(R5_LOCKED, &qdev->flags)
-			     && test_bit(R5_UPTODATE, &qdev->flags)))))
-		handle_stripe_clean_event(conf, sh, disks, &return_bi);
-
-	/* Now we might consider reading some blocks, either to check/generate
-	 * parity, or to satisfy requests
-	 * or to load a block that is being partially written.
-	 */
-	if (s.to_read || s.non_overwrite || (s.to_write && s.failed) ||
-	    (s.syncing && (s.uptodate < disks)) || s.expanding)
-		handle_stripe_fill6(sh, &s, &r6s, disks);
-
-	/* now to consider writing and what else, if anything should be read */
-	if (s.to_write)
-		handle_stripe_dirtying6(conf, sh, &s, &r6s, disks);
-
-	/* maybe we need to check and possibly fix the parity for this stripe
-	 * Any reads will already have been scheduled, so we just see if enough
-	 * data is available
-	 */
-	if (s.syncing && s.locked == 0 && !test_bit(STRIPE_INSYNC, &sh->state))
-		handle_parity_checks6(conf, sh, &s, &r6s, tmp_page, disks);
-
-	if (s.syncing && s.locked == 0 && test_bit(STRIPE_INSYNC, &sh->state)) {
-		md_done_sync(conf->mddev, STRIPE_SECTORS,1);
+	if ((s.syncing || s.replacing) && s.locked == 0 &&
+	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state) &&
+	    test_bit(STRIPE_INSYNC, &sh->state)) {
+		md_done_sync(conf->mddev, STRIPE_SECTORS, 1);
 		clear_bit(STRIPE_SYNCING, &sh->state);
+		if (test_and_clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags))
+			wake_up(&conf->wait_for_overlap);
 	}
 
 	/* If the failed drives are just a ReadError, then we might need
 	 * to progress the repair/check process
 	 */
-	if (s.failed <= 2 && !conf->mddev->ro)
+	if (s.failed <= conf->max_degraded && !conf->mddev->ro)
 		for (i = 0; i < s.failed; i++) {
-			dev = &sh->dev[r6s.failed_num[i]];
+			struct r5dev *dev = &sh->dev[s.failed_num[i]];
 			if (test_bit(R5_ReadError, &dev->flags)
 			    && !test_bit(R5_LOCKED, &dev->flags)
 			    && test_bit(R5_UPTODATE, &dev->flags)
@@ -2942,27 +3923,52 @@ static bool handle_stripe6(struct stripe_head *sh, struct page *tmp_page)
 					set_bit(R5_Wantwrite, &dev->flags);
 					set_bit(R5_ReWrite, &dev->flags);
 					set_bit(R5_LOCKED, &dev->flags);
+					s.locked++;
 				} else {
 					/* let's read it back */
 					set_bit(R5_Wantread, &dev->flags);
 					set_bit(R5_LOCKED, &dev->flags);
+					s.locked++;
 				}
 			}
 		}
 
-	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state)) {
-		/* Need to write out all blocks after computing P&Q */
-		sh->disks = conf->raid_disks;
-		sh->pd_idx = stripe_to_pdidx(sh->sector, conf,
-					     conf->raid_disks);
-		compute_parity6(sh, RECONSTRUCT_WRITE);
-		for (i = conf->raid_disks ; i-- ;  ) {
+
+	/* Finish reconstruct operations initiated by the expansion process */
+	if (sh->reconstruct_state == reconstruct_state_result) {
+		struct stripe_head *sh_src
+			= get_active_stripe(conf, sh->sector, 1, 1, 1);
+		if (sh_src && test_bit(STRIPE_EXPAND_SOURCE, &sh_src->state)) {
+			/* sh cannot be written until sh_src has been read.
+			 * so arrange for sh to be delayed a little
+			 */
+			set_bit(STRIPE_DELAYED, &sh->state);
+			set_bit(STRIPE_HANDLE, &sh->state);
+			if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE,
+					      &sh_src->state))
+				atomic_inc(&conf->preread_active_stripes);
+			release_stripe(sh_src);
+			goto finish;
+		}
+		if (sh_src)
+			release_stripe(sh_src);
+
+		sh->reconstruct_state = reconstruct_state_idle;
+		clear_bit(STRIPE_EXPANDING, &sh->state);
+		for (i = conf->raid_disks; i--; ) {
+			set_bit(R5_Wantwrite, &sh->dev[i].flags);
 			set_bit(R5_LOCKED, &sh->dev[i].flags);
 			s.locked++;
-			set_bit(R5_Wantwrite, &sh->dev[i].flags);
 		}
-		clear_bit(STRIPE_EXPANDING, &sh->state);
-	} else if (s.expanded) {
+	}
+
+	if (s.expanded && test_bit(STRIPE_EXPANDING, &sh->state) &&
+	    !sh->reconstruct_state) {
+		/* Need to write out all blocks after computing parity */
+		sh->disks = conf->raid_disks;
+		stripe_set_idx(sh->sector, conf, 0, sh);
+		schedule_reconstruction(sh, &s, 1, 1);
+	} else if (s.expanded && !sh->reconstruct_state && s.locked == 0) {
 		clear_bit(STRIPE_EXPAND_READY, &sh->state);
 		atomic_dec(&conf->reshape_stripes);
 		wake_up(&conf->wait_for_overlap);
@@ -2971,34 +3977,74 @@ static bool handle_stripe6(struct stripe_head *sh, struct page *tmp_page)
 
 	if (s.expanding && s.locked == 0 &&
 	    !test_bit(STRIPE_COMPUTE_RUN, &sh->state))
-		handle_stripe_expansion(conf, sh, &r6s);
-
- unlock:
-	spin_unlock(&sh->lock);
+		handle_stripe_expansion(conf, sh);
 
+finish:
 	/* wait for this device to become unblocked */
-	if (unlikely(blocked_rdev))
-		md_wait_for_blocked_rdev(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);
+	}
 
-	ops_run_io(sh, &s);
+	if (s.handle_bad_blocks)
+		for (i = disks; i--; ) {
+			struct md_rdev *rdev;
+			struct r5dev *dev = &sh->dev[i];
+			if (test_and_clear_bit(R5_WriteError, &dev->flags)) {
+				/* We own a safe reference to the rdev */
+				rdev = conf->disks[i].rdev;
+				if (!rdev_set_badblocks(rdev, sh->sector,
+							STRIPE_SECTORS, 0))
+					md_error(conf->mddev, rdev);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
+			if (test_and_clear_bit(R5_MadeGood, &dev->flags)) {
+				rdev = conf->disks[i].rdev;
+				rdev_clear_badblocks(rdev, sh->sector,
+						     STRIPE_SECTORS, 0);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
+			if (test_and_clear_bit(R5_MadeGoodRepl, &dev->flags)) {
+				rdev = conf->disks[i].replacement;
+				if (!rdev)
+					/* rdev have been moved down */
+					rdev = conf->disks[i].rdev;
+				rdev_clear_badblocks(rdev, sh->sector,
+						     STRIPE_SECTORS, 0);
+				rdev_dec_pending(rdev, conf->mddev);
+			}
+		}
 
-	return_io(return_bi);
+	if (s.ops_request)
+		raid_run_ops(sh, s.ops_request);
 
-	return blocked_rdev == NULL;
-}
+	ops_run_io(sh, &s);
 
-/* returns true if the stripe was handled */
-static bool handle_stripe(struct stripe_head *sh, struct page *tmp_page)
-{
-	if (sh->raid_conf->level == 6)
-		return handle_stripe6(sh, tmp_page);
-	else
-		return handle_stripe5(sh);
-}
+	if (s.dec_preread_active) {
+		/* We delay this until after ops_run_io so that if make_request
+		 * is waiting on a flush, it won't continue until the writes
+		 * have actually been submitted.
+		 */
+		atomic_dec(&conf->preread_active_stripes);
+		if (atomic_read(&conf->preread_active_stripes) <
+		    IO_THRESHOLD)
+			md_wakeup_thread(conf->mddev->thread);
+	}
 
+	return_io(s.return_bi);
 
+	clear_bit_unlock(STRIPE_ACTIVE, &sh->state);
+}
 
-static void raid5_activate_delayed(raid5_conf_t *conf)
+static void raid5_activate_delayed(struct r5conf *conf)
 {
 	if (atomic_read(&conf->preread_active_stripes) < IO_THRESHOLD) {
 		while (!list_empty(&conf->delayed_list)) {
@@ -3010,12 +4056,13 @@ static void raid5_activate_delayed(raid5_conf_t *conf)
 			if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
 				atomic_inc(&conf->preread_active_stripes);
 			list_add_tail(&sh->lru, &conf->hold_list);
+			raid5_wakeup_stripe_thread(sh);
 		}
-	} else
-		blk_plug_device(conf->mddev->queue);
+	}
 }
 
-static void activate_bit_delay(raid5_conf_t *conf)
+static void activate_bit_delay(struct r5conf *conf,
+	struct list_head *temp_inactive_list)
 {
 	/* device_lock is held */
 	struct list_head head;
@@ -3023,71 +4070,40 @@ static void activate_bit_delay(raid5_conf_t *conf)
 	list_del_init(&conf->bitmap_list);
 	while (!list_empty(&head)) {
 		struct stripe_head *sh = list_entry(head.next, struct stripe_head, lru);
+		int hash;
 		list_del_init(&sh->lru);
 		atomic_inc(&sh->count);
-		__release_stripe(conf, sh);
-	}
-}
-
-static void unplug_slaves(mddev_t *mddev)
-{
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	int i;
-
-	rcu_read_lock();
-	for (i=0; i<mddev->raid_disks; i++) {
-		mdk_rdev_t *rdev = rcu_dereference(conf->disks[i].rdev);
-		if (rdev && !test_bit(Faulty, &rdev->flags) && atomic_read(&rdev->nr_pending)) {
-			struct request_queue *r_queue = bdev_get_queue(rdev->bdev);
-
-			atomic_inc(&rdev->nr_pending);
-			rcu_read_unlock();
-
-			blk_unplug(r_queue);
-
-			rdev_dec_pending(rdev, mddev);
-			rcu_read_lock();
-		}
-	}
-	rcu_read_unlock();
-}
-
-static void raid5_unplug_device(struct request_queue *q)
-{
-	mddev_t *mddev = q->queuedata;
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	unsigned long flags;
-
-	spin_lock_irqsave(&conf->device_lock, flags);
-
-	if (blk_remove_plug(q)) {
-		conf->seq_flush++;
-		raid5_activate_delayed(conf);
+		hash = sh->hash_lock_index;
+		__release_stripe(conf, sh, &temp_inactive_list[hash]);
 	}
-	md_wakeup_thread(mddev->thread);
-
-	spin_unlock_irqrestore(&conf->device_lock, flags);
-
-	unplug_slaves(mddev);
 }
 
-static int raid5_congested(void *data, int bits)
+int md_raid5_congested(struct mddev *mddev, int bits)
 {
-	mddev_t *mddev = data;
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
 
 	/* No difference between reads and writes.  Just check
 	 * how busy the stripe_cache is
 	 */
+
 	if (conf->inactive_blocked)
 		return 1;
 	if (conf->quiesce)
 		return 1;
-	if (list_empty_careful(&conf->inactive_list))
+	if (atomic_read(&conf->empty_inactive_list_nr))
 		return 1;
 
 	return 0;
 }
+EXPORT_SYMBOL_GPL(md_raid5_congested);
+
+static int raid5_congested(void *data, int bits)
+{
+	struct mddev *mddev = data;
+
+	return mddev_congested(mddev, bits) ||
+		md_raid5_congested(mddev, bits);
+}
 
 /* We want read requests to align with chunks where possible,
  * but write requests don't need to.
@@ -3096,15 +4112,17 @@ static int raid5_mergeable_bvec(struct request_queue *q,
 				struct bvec_merge_data *bvm,
 				struct bio_vec *biovec)
 {
-	mddev_t *mddev = q->queuedata;
+	struct mddev *mddev = q->queuedata;
 	sector_t sector = bvm->bi_sector + get_start_sect(bvm->bi_bdev);
 	int max;
-	unsigned int chunk_sectors = mddev->chunk_size >> 9;
+	unsigned int chunk_sectors = mddev->chunk_sectors;
 	unsigned int bio_sectors = bvm->bi_size >> 9;
 
 	if ((bvm->bi_rw & 1) == WRITE)
 		return biovec->bv_len; /* always allow writes to be mergeable */
 
+	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
+		chunk_sectors = mddev->new_chunk_sectors;
 	max =  (chunk_sectors - ((sector & (chunk_sectors - 1)) + bio_sectors)) << 9;
 	if (max < 0) max = 0;
 	if (max <= biovec->bv_len && bio_sectors == 0)
@@ -3114,12 +4132,14 @@ static int raid5_mergeable_bvec(struct request_queue *q,
 }
 
 
-static int in_chunk_boundary(mddev_t *mddev, struct bio *bio)
+static int in_chunk_boundary(struct mddev *mddev, struct bio *bio)
 {
-	sector_t sector = bio->bi_sector + get_start_sect(bio->bi_bdev);
-	unsigned int chunk_sectors = mddev->chunk_size >> 9;
-	unsigned int bio_sectors = bio->bi_size >> 9;
+	sector_t sector = bio->bi_iter.bi_sector + get_start_sect(bio->bi_bdev);
+	unsigned int chunk_sectors = mddev->chunk_sectors;
+	unsigned int bio_sectors = bio_sectors(bio);
 
+	if (mddev->new_chunk_sectors < mddev->chunk_sectors)
+		chunk_sectors = mddev->new_chunk_sectors;
 	return  chunk_sectors >=
 		((sector & (chunk_sectors - 1)) + bio_sectors);
 }
@@ -3128,7 +4148,7 @@ static int in_chunk_boundary(mddev_t *mddev, struct bio *bio)
  *  add bio to the retry LIFO  ( in O(1) ... we are in interrupt )
  *  later sampled by raid5d.
  */
-static void add_bio_to_retry(struct bio *bi,raid5_conf_t *conf)
+static void add_bio_to_retry(struct bio *bi,struct r5conf *conf)
 {
 	unsigned long flags;
 
@@ -3142,7 +4162,7 @@ static void add_bio_to_retry(struct bio *bi,raid5_conf_t *conf)
 }
 
 
-static struct bio *remove_bio_from_retry(raid5_conf_t *conf)
+static struct bio *remove_bio_from_retry(struct r5conf *conf)
 {
 	struct bio *bi;
 
@@ -3155,8 +4175,11 @@ static struct bio *remove_bio_from_retry(raid5_conf_t *conf)
 	if(bi) {
 		conf->retry_read_aligned_list = bi->bi_next;
 		bi->bi_next = NULL;
-		bi->bi_phys_segments = 1; /* biased count of active stripes */
-		bi->bi_hw_segments = 0; /* count of processed stripes */
+		/*
+		 * this sets the active strip count to 1 and the processed
+		 * strip count to zero (upper 8 bits)
+		 */
+		raid5_set_bi_stripes(bi, 1); /* biased count of active stripes */
 	}
 
 	return bi;
@@ -3172,21 +4195,23 @@ static struct bio *remove_bio_from_retry(raid5_conf_t *conf)
 static void raid5_align_endio(struct bio *bi, int error)
 {
 	struct bio* raid_bi  = bi->bi_private;
-	mddev_t *mddev;
-	raid5_conf_t *conf;
+	struct mddev *mddev;
+	struct r5conf *conf;
 	int uptodate = test_bit(BIO_UPTODATE, &bi->bi_flags);
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
 
 	bio_put(bi);
 
-	mddev = raid_bi->bi_bdev->bd_disk->queue->queuedata;
-	conf = mddev_to_conf(mddev);
 	rdev = (void*)raid_bi->bi_next;
 	raid_bi->bi_next = NULL;
+	mddev = rdev->mddev;
+	conf = mddev->private;
 
 	rdev_dec_pending(rdev, conf->mddev);
 
 	if (!error && uptodate) {
+		trace_block_bio_complete(bdev_get_queue(raid_bi->bi_bdev),
+					 raid_bi, 0);
 		bio_endio(raid_bi, 0);
 		if (atomic_dec_and_test(&conf->active_aligned_reads))
 			wake_up(&conf->wait_for_stripe);
@@ -3203,11 +4228,10 @@ static int bio_fits_rdev(struct bio *bi)
 {
 	struct request_queue *q = bdev_get_queue(bi->bi_bdev);
 
-	if ((bi->bi_size>>9) > q->max_sectors)
+	if (bio_sectors(bi) > queue_max_sectors(q))
 		return 0;
 	blk_recount_segments(q, bi);
-	if (bi->bi_phys_segments > q->max_phys_segments ||
-	    bi->bi_hw_segments > q->max_hw_segments)
+	if (bi->bi_phys_segments > queue_max_segments(q))
 		return 0;
 
 	if (q->merge_bvec_fn)
@@ -3220,24 +4244,22 @@ static int bio_fits_rdev(struct bio *bi)
 }
 
 
-static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
+static int chunk_aligned_read(struct mddev *mddev, struct bio * raid_bio)
 {
-	mddev_t *mddev = q->queuedata;
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	const unsigned int raid_disks = conf->raid_disks;
-	const unsigned int data_disks = raid_disks - conf->max_degraded;
-	unsigned int dd_idx, pd_idx;
+	struct r5conf *conf = mddev->private;
+	int dd_idx;
 	struct bio* align_bi;
-	mdk_rdev_t *rdev;
+	struct md_rdev *rdev;
+	sector_t end_sector;
 
 	if (!in_chunk_boundary(mddev, raid_bio)) {
 		pr_debug("chunk_aligned_read : non aligned\n");
 		return 0;
 	}
 	/*
- 	 * use bio_clone to make a copy of the bio
+	 * use bio_clone_mddev to make a copy of the bio
 	 */
-	align_bi = bio_clone(raid_bio, GFP_NOIO);
+	align_bi = bio_clone_mddev(raid_bio, GFP_NOIO, mddev);
 	if (!align_bi)
 		return 0;
 	/*
@@ -3249,37 +4271,56 @@ static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
 	/*
 	 *	compute position
 	 */
-	align_bi->bi_sector =  raid5_compute_sector(raid_bio->bi_sector,
-					raid_disks,
-					data_disks,
-					&dd_idx,
-					&pd_idx,
-					conf);
+	align_bi->bi_iter.bi_sector =
+		raid5_compute_sector(conf, raid_bio->bi_iter.bi_sector,
+				     0, &dd_idx, NULL);
 
+	end_sector = bio_end_sector(align_bi);
 	rcu_read_lock();
-	rdev = rcu_dereference(conf->disks[dd_idx].rdev);
-	if (rdev && test_bit(In_sync, &rdev->flags)) {
+	rdev = rcu_dereference(conf->disks[dd_idx].replacement);
+	if (!rdev || test_bit(Faulty, &rdev->flags) ||
+	    rdev->recovery_offset < end_sector) {
+		rdev = rcu_dereference(conf->disks[dd_idx].rdev);
+		if (rdev &&
+		    (test_bit(Faulty, &rdev->flags) ||
+		    !(test_bit(In_sync, &rdev->flags) ||
+		      rdev->recovery_offset >= end_sector)))
+			rdev = NULL;
+	}
+	if (rdev) {
+		sector_t first_bad;
+		int bad_sectors;
+
 		atomic_inc(&rdev->nr_pending);
 		rcu_read_unlock();
 		raid_bio->bi_next = (void*)rdev;
 		align_bi->bi_bdev =  rdev->bdev;
 		align_bi->bi_flags &= ~(1 << BIO_SEG_VALID);
-		align_bi->bi_sector += rdev->data_offset;
 
-		if (!bio_fits_rdev(align_bi)) {
-			/* too big in some way */
+		if (!bio_fits_rdev(align_bi) ||
+		    is_badblock(rdev, align_bi->bi_iter.bi_sector,
+				bio_sectors(align_bi),
+				&first_bad, &bad_sectors)) {
+			/* too big in some way, or has a known bad block */
 			bio_put(align_bi);
 			rdev_dec_pending(rdev, mddev);
 			return 0;
 		}
 
+		/* No reshape active, so we can trust rdev->data_offset */
+		align_bi->bi_iter.bi_sector += rdev->data_offset;
+
 		spin_lock_irq(&conf->device_lock);
 		wait_event_lock_irq(conf->wait_for_stripe,
 				    conf->quiesce == 0,
-				    conf->device_lock, /* nothing */);
+				    conf->device_lock);
 		atomic_inc(&conf->active_aligned_reads);
 		spin_unlock_irq(&conf->device_lock);
 
+		if (mddev->gendisk)
+			trace_block_bio_remap(bdev_get_queue(align_bi->bi_bdev),
+					      align_bi, disk_devt(mddev->gendisk),
+					      raid_bio->bi_iter.bi_sector);
 		generic_make_request(align_bi);
 		return 1;
 	} else {
@@ -3299,18 +4340,35 @@ static int chunk_aligned_read(struct request_queue *q, struct bio * raid_bio)
  * head of the hold_list has changed, i.e. the head was promoted to the
  * handle_list.
  */
-static struct stripe_head *__get_priority_stripe(raid5_conf_t *conf)
+static struct stripe_head *__get_priority_stripe(struct r5conf *conf, int group)
 {
-	struct stripe_head *sh;
+	struct stripe_head *sh = NULL, *tmp;
+	struct list_head *handle_list = NULL;
+	struct r5worker_group *wg = NULL;
+
+	if (conf->worker_cnt_per_group == 0) {
+		handle_list = &conf->handle_list;
+	} else if (group != ANY_GROUP) {
+		handle_list = &conf->worker_groups[group].handle_list;
+		wg = &conf->worker_groups[group];
+	} else {
+		int i;
+		for (i = 0; i < conf->group_cnt; i++) {
+			handle_list = &conf->worker_groups[i].handle_list;
+			wg = &conf->worker_groups[i];
+			if (!list_empty(handle_list))
+				break;
+		}
+	}
 
 	pr_debug("%s: handle: %s hold: %s full_writes: %d bypass_count: %d\n",
 		  __func__,
-		  list_empty(&conf->handle_list) ? "empty" : "busy",
+		  list_empty(handle_list) ? "empty" : "busy",
 		  list_empty(&conf->hold_list) ? "empty" : "busy",
 		  atomic_read(&conf->pending_full_writes), conf->bypass_count);
 
-	if (!list_empty(&conf->handle_list)) {
-		sh = list_entry(conf->handle_list.next, typeof(*sh), lru);
+	if (!list_empty(handle_list)) {
+		sh = list_entry(handle_list->next, typeof(*sh), lru);
 
 		if (list_empty(&conf->hold_list))
 			conf->bypass_count = 0;
@@ -3328,92 +4386,284 @@ static struct stripe_head *__get_priority_stripe(raid5_conf_t *conf)
 		   ((conf->bypass_threshold &&
 		     conf->bypass_count > conf->bypass_threshold) ||
 		    atomic_read(&conf->pending_full_writes) == 0)) {
-		sh = list_entry(conf->hold_list.next,
-				typeof(*sh), lru);
-		conf->bypass_count -= conf->bypass_threshold;
-		if (conf->bypass_count < 0)
-			conf->bypass_count = 0;
-	} else
+
+		list_for_each_entry(tmp, &conf->hold_list,  lru) {
+			if (conf->worker_cnt_per_group == 0 ||
+			    group == ANY_GROUP ||
+			    !cpu_online(tmp->cpu) ||
+			    cpu_to_group(tmp->cpu) == group) {
+				sh = tmp;
+				break;
+			}
+		}
+
+		if (sh) {
+			conf->bypass_count -= conf->bypass_threshold;
+			if (conf->bypass_count < 0)
+				conf->bypass_count = 0;
+		}
+		wg = NULL;
+	}
+
+	if (!sh)
 		return NULL;
 
+	if (wg) {
+		wg->stripes_cnt--;
+		sh->group = NULL;
+	}
 	list_del_init(&sh->lru);
-	atomic_inc(&sh->count);
-	BUG_ON(atomic_read(&sh->count) != 1);
+	BUG_ON(atomic_inc_return(&sh->count) != 1);
 	return sh;
 }
 
-static int make_request(struct request_queue *q, struct bio * bi)
+struct raid5_plug_cb {
+	struct blk_plug_cb	cb;
+	struct list_head	list;
+	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
+};
+
+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;
+	int cnt = 0;
+	int hash;
+
+	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_atomic();
+			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
+			/*
+			 * STRIPE_ON_RELEASE_LIST could be set here. In that
+			 * case, the count is always > 1 here
+			 */
+			hash = sh->hash_lock_index;
+			__release_stripe(conf, sh, &cb->temp_inactive_list[hash]);
+			cnt++;
+		}
+		spin_unlock_irq(&conf->device_lock);
+	}
+	release_inactive_stripe_list(conf, cb->temp_inactive_list,
+				     NR_STRIPE_HASH_LOCKS);
+	if (mddev->queue)
+		trace_block_unplug(mddev->queue, cnt, !from_schedule);
+	kfree(cb);
+}
+
+static void release_stripe_plug(struct mddev *mddev,
+				struct stripe_head *sh)
 {
-	mddev_t *mddev = q->queuedata;
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	unsigned int dd_idx, pd_idx;
+	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) {
+		int i;
+		INIT_LIST_HEAD(&cb->list);
+		for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
+			INIT_LIST_HEAD(cb->temp_inactive_list + i);
+	}
+
+	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_discard_request(struct mddev *mddev, struct bio *bi)
+{
+	struct r5conf *conf = mddev->private;
+	sector_t logical_sector, last_sector;
+	struct stripe_head *sh;
+	int remaining;
+	int stripe_sectors;
+
+	if (mddev->reshape_position != MaxSector)
+		/* Skip discard while reshape is happening */
+		return;
+
+	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
+	last_sector = bi->bi_iter.bi_sector + (bi->bi_iter.bi_size>>9);
+
+	bi->bi_next = NULL;
+	bi->bi_phys_segments = 1; /* over-loaded to count active stripes */
+
+	stripe_sectors = conf->chunk_sectors *
+		(conf->raid_disks - conf->max_degraded);
+	logical_sector = DIV_ROUND_UP_SECTOR_T(logical_sector,
+					       stripe_sectors);
+	sector_div(last_sector, stripe_sectors);
+
+	logical_sector *= conf->chunk_sectors;
+	last_sector *= conf->chunk_sectors;
+
+	for (; logical_sector < last_sector;
+	     logical_sector += STRIPE_SECTORS) {
+		DEFINE_WAIT(w);
+		int d;
+	again:
+		sh = get_active_stripe(conf, logical_sector, 0, 0, 0);
+		prepare_to_wait(&conf->wait_for_overlap, &w,
+				TASK_UNINTERRUPTIBLE);
+		set_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
+		if (test_bit(STRIPE_SYNCING, &sh->state)) {
+			release_stripe(sh);
+			schedule();
+			goto again;
+		}
+		clear_bit(R5_Overlap, &sh->dev[sh->pd_idx].flags);
+		spin_lock_irq(&sh->stripe_lock);
+		for (d = 0; d < conf->raid_disks; d++) {
+			if (d == sh->pd_idx || d == sh->qd_idx)
+				continue;
+			if (sh->dev[d].towrite || sh->dev[d].toread) {
+				set_bit(R5_Overlap, &sh->dev[d].flags);
+				spin_unlock_irq(&sh->stripe_lock);
+				release_stripe(sh);
+				schedule();
+				goto again;
+			}
+		}
+		set_bit(STRIPE_DISCARD, &sh->state);
+		finish_wait(&conf->wait_for_overlap, &w);
+		for (d = 0; d < conf->raid_disks; d++) {
+			if (d == sh->pd_idx || d == sh->qd_idx)
+				continue;
+			sh->dev[d].towrite = bi;
+			set_bit(R5_OVERWRITE, &sh->dev[d].flags);
+			raid5_inc_bi_active_stripes(bi);
+		}
+		spin_unlock_irq(&sh->stripe_lock);
+		if (conf->mddev->bitmap) {
+			for (d = 0;
+			     d < conf->raid_disks - conf->max_degraded;
+			     d++)
+				bitmap_startwrite(mddev->bitmap,
+						  sh->sector,
+						  STRIPE_SECTORS,
+						  0);
+			sh->bm_seq = conf->seq_flush + 1;
+			set_bit(STRIPE_BIT_DELAY, &sh->state);
+		}
+
+		set_bit(STRIPE_HANDLE, &sh->state);
+		clear_bit(STRIPE_DELAYED, &sh->state);
+		if (!test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+			atomic_inc(&conf->preread_active_stripes);
+		release_stripe_plug(mddev, sh);
+	}
+
+	remaining = raid5_dec_bi_active_stripes(bi);
+	if (remaining == 0) {
+		md_write_end(mddev);
+		bio_endio(bi, 0);
+	}
+}
+
+static void make_request(struct mddev *mddev, struct bio * bi)
+{
+	struct r5conf *conf = mddev->private;
+	int dd_idx;
 	sector_t new_sector;
 	sector_t logical_sector, last_sector;
 	struct stripe_head *sh;
 	const int rw = bio_data_dir(bi);
 	int remaining;
+	DEFINE_WAIT(w);
+	bool do_prepare;
 
-	if (unlikely(bio_barrier(bi))) {
-		bio_endio(bi, -EOPNOTSUPP);
-		return 0;
+	if (unlikely(bi->bi_rw & REQ_FLUSH)) {
+		md_flush_request(mddev, bi);
+		return;
 	}
 
 	md_write_start(mddev, bi);
 
-	disk_stat_inc(mddev->gendisk, ios[rw]);
-	disk_stat_add(mddev->gendisk, sectors[rw], bio_sectors(bi));
-
 	if (rw == READ &&
 	     mddev->reshape_position == MaxSector &&
-	     chunk_aligned_read(q,bi))
-            	return 0;
+	     chunk_aligned_read(mddev,bi))
+		return;
+
+	if (unlikely(bi->bi_rw & REQ_DISCARD)) {
+		make_discard_request(mddev, bi);
+		return;
+	}
 
-	logical_sector = bi->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
-	last_sector = bi->bi_sector + (bi->bi_size>>9);
+	logical_sector = bi->bi_iter.bi_sector & ~((sector_t)STRIPE_SECTORS-1);
+	last_sector = bio_end_sector(bi);
 	bi->bi_next = NULL;
 	bi->bi_phys_segments = 1;	/* over-loaded to count active stripes */
 
+	prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
 	for (;logical_sector < last_sector; logical_sector += STRIPE_SECTORS) {
-		DEFINE_WAIT(w);
-		int disks, data_disks;
+		int previous;
+		int seq;
 
+		do_prepare = false;
 	retry:
-		prepare_to_wait(&conf->wait_for_overlap, &w, TASK_UNINTERRUPTIBLE);
-		if (likely(conf->expand_progress == MaxSector))
-			disks = conf->raid_disks;
-		else {
-			/* spinlock is needed as expand_progress may be
+		seq = read_seqcount_begin(&conf->gen_lock);
+		previous = 0;
+		if (do_prepare)
+			prepare_to_wait(&conf->wait_for_overlap, &w,
+				TASK_UNINTERRUPTIBLE);
+		if (unlikely(conf->reshape_progress != MaxSector)) {
+			/* spinlock is needed as reshape_progress may be
 			 * 64bit on a 32bit platform, and so it might be
 			 * possible to see a half-updated value
-			 * Ofcourse expand_progress could change after
+			 * Of course reshape_progress could change after
 			 * the lock is dropped, so once we get a reference
 			 * to the stripe that we think it is, we will have
 			 * to check again.
 			 */
 			spin_lock_irq(&conf->device_lock);
-			disks = conf->raid_disks;
-			if (logical_sector >= conf->expand_progress)
-				disks = conf->previous_raid_disks;
-			else {
-				if (logical_sector >= conf->expand_lo) {
+			if (mddev->reshape_backwards
+			    ? logical_sector < conf->reshape_progress
+			    : logical_sector >= conf->reshape_progress) {
+				previous = 1;
+			} else {
+				if (mddev->reshape_backwards
+				    ? logical_sector < conf->reshape_safe
+				    : logical_sector >= conf->reshape_safe) {
 					spin_unlock_irq(&conf->device_lock);
 					schedule();
+					do_prepare = true;
 					goto retry;
 				}
 			}
 			spin_unlock_irq(&conf->device_lock);
 		}
-		data_disks = disks - conf->max_degraded;
 
- 		new_sector = raid5_compute_sector(logical_sector, disks, data_disks,
-						  &dd_idx, &pd_idx, conf);
-		pr_debug("raid5: make_request, sector %llu logical %llu\n",
-			(unsigned long long)new_sector, 
+		new_sector = raid5_compute_sector(conf, logical_sector,
+						  previous,
+						  &dd_idx, NULL);
+		pr_debug("raid456: make_request, sector %llu logical %llu\n",
+			(unsigned long long)new_sector,
 			(unsigned long long)logical_sector);
 
-		sh = get_active_stripe(conf, new_sector, disks, pd_idx, (bi->bi_rw&RWA_MASK));
+		sh = get_active_stripe(conf, new_sector, previous,
+				       (bi->bi_rw&RWA_MASK), 0);
 		if (sh) {
-			if (unlikely(conf->expand_progress != MaxSector)) {
+			if (unlikely(previous)) {
 				/* expansion might have moved on while waiting for a
 				 * stripe, so we must do the range check again.
 				 * Expansion could still move past after this
@@ -3424,63 +4674,87 @@ static int make_request(struct request_queue *q, struct bio * bi)
 				 */
 				int must_retry = 0;
 				spin_lock_irq(&conf->device_lock);
-				if (logical_sector <  conf->expand_progress &&
-				    disks == conf->previous_raid_disks)
+				if (mddev->reshape_backwards
+				    ? logical_sector >= conf->reshape_progress
+				    : logical_sector < conf->reshape_progress)
 					/* mismatch, need to try again */
 					must_retry = 1;
 				spin_unlock_irq(&conf->device_lock);
 				if (must_retry) {
 					release_stripe(sh);
+					schedule();
+					do_prepare = true;
 					goto retry;
 				}
 			}
-			/* FIXME what if we get a false positive because these
-			 * are being updated.
-			 */
-			if (logical_sector >= mddev->suspend_lo &&
+			if (read_seqcount_retry(&conf->gen_lock, seq)) {
+				/* Might have got the wrong stripe_head
+				 * by accident
+				 */
+				release_stripe(sh);
+				goto retry;
+			}
+
+			if (rw == WRITE &&
+			    logical_sector >= mddev->suspend_lo &&
 			    logical_sector < mddev->suspend_hi) {
 				release_stripe(sh);
-				schedule();
+				/* As the suspend_* range is controlled by
+				 * userspace, we want an interruptible
+				 * wait.
+				 */
+				flush_signals(current);
+				prepare_to_wait(&conf->wait_for_overlap,
+						&w, TASK_INTERRUPTIBLE);
+				if (logical_sector >= mddev->suspend_lo &&
+				    logical_sector < mddev->suspend_hi) {
+					schedule();
+					do_prepare = true;
+				}
 				goto retry;
 			}
 
 			if (test_bit(STRIPE_EXPANDING, &sh->state) ||
-			    !add_stripe_bio(sh, bi, dd_idx, (bi->bi_rw&RW_MASK))) {
+			    !add_stripe_bio(sh, bi, dd_idx, rw)) {
 				/* Stripe is busy expanding or
 				 * add failed due to overlap.  Flush everything
 				 * and wait a while
 				 */
-				raid5_unplug_device(mddev->queue);
+				md_wakeup_thread(mddev->thread);
 				release_stripe(sh);
 				schedule();
+				do_prepare = true;
 				goto retry;
 			}
-			finish_wait(&conf->wait_for_overlap, &w);
 			set_bit(STRIPE_HANDLE, &sh->state);
 			clear_bit(STRIPE_DELAYED, &sh->state);
-			release_stripe(sh);
+			if ((bi->bi_rw & REQ_SYNC) &&
+			    !test_and_set_bit(STRIPE_PREREAD_ACTIVE, &sh->state))
+				atomic_inc(&conf->preread_active_stripes);
+			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;
 		}
-			
 	}
-	spin_lock_irq(&conf->device_lock);
-	remaining = --bi->bi_phys_segments;
-	spin_unlock_irq(&conf->device_lock);
+	finish_wait(&conf->wait_for_overlap, &w);
+
+	remaining = raid5_dec_bi_active_stripes(bi);
 	if (remaining == 0) {
 
 		if ( rw == WRITE )
 			md_write_end(mddev);
 
+		trace_block_bio_complete(bdev_get_queue(bi->bi_bdev),
+					 bi, 0);
 		bio_endio(bi, 0);
 	}
-	return 0;
 }
 
-static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped)
+static sector_t raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks);
+
+static sector_t reshape_request(struct mddev *mddev, sector_t sector_nr, int *skipped)
 {
 	/* reshaping is quite different to recovery/resync so it is
 	 * handled quite separately ... here.
@@ -3491,63 +4765,140 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
 	 * As the reads complete, handle_stripe will copy the data
 	 * into the destination stripe and release that stripe.
 	 */
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	struct r5conf *conf = mddev->private;
 	struct stripe_head *sh;
-	int pd_idx;
 	sector_t first_sector, last_sector;
 	int raid_disks = conf->previous_raid_disks;
 	int data_disks = raid_disks - conf->max_degraded;
 	int new_data_disks = conf->raid_disks - conf->max_degraded;
 	int i;
 	int dd_idx;
-	sector_t writepos, safepos, gap;
-
-	if (sector_nr == 0 &&
-	    conf->expand_progress != 0) {
-		/* restarting in the middle, skip the initial sectors */
-		sector_nr = conf->expand_progress;
+	sector_t writepos, readpos, safepos;
+	sector_t stripe_addr;
+	int reshape_sectors;
+	struct list_head stripes;
+
+	if (sector_nr == 0) {
+		/* If restarting in the middle, skip the initial sectors */
+		if (mddev->reshape_backwards &&
+		    conf->reshape_progress < raid5_size(mddev, 0, 0)) {
+			sector_nr = raid5_size(mddev, 0, 0)
+				- conf->reshape_progress;
+		} else if (!mddev->reshape_backwards &&
+			   conf->reshape_progress > 0)
+			sector_nr = conf->reshape_progress;
 		sector_div(sector_nr, new_data_disks);
-		*skipped = 1;
-		return sector_nr;
+		if (sector_nr) {
+			mddev->curr_resync_completed = sector_nr;
+			sysfs_notify(&mddev->kobj, NULL, "sync_completed");
+			*skipped = 1;
+			return sector_nr;
+		}
 	}
 
-	/* we update the metadata when there is more than 3Meg
-	 * in the block range (that is rather arbitrary, should
-	 * probably be time based) or when the data about to be
-	 * copied would over-write the source of the data at
-	 * the front of the range.
-	 * i.e. one new_stripe forward from expand_progress new_maps
-	 * to after where expand_lo old_maps to
+	/* We need to process a full chunk at a time.
+	 * If old and new chunk sizes differ, we need to process the
+	 * largest of these
 	 */
-	writepos = conf->expand_progress +
-		conf->chunk_size/512*(new_data_disks);
+	if (mddev->new_chunk_sectors > mddev->chunk_sectors)
+		reshape_sectors = mddev->new_chunk_sectors;
+	else
+		reshape_sectors = mddev->chunk_sectors;
+
+	/* We update the metadata at least every 10 seconds, or when
+	 * the data about to be copied would over-write the source of
+	 * the data at the front of the range.  i.e. one new_stripe
+	 * along from reshape_progress new_maps to after where
+	 * reshape_safe old_maps to
+	 */
+	writepos = conf->reshape_progress;
 	sector_div(writepos, new_data_disks);
-	safepos = conf->expand_lo;
+	readpos = conf->reshape_progress;
+	sector_div(readpos, data_disks);
+	safepos = conf->reshape_safe;
 	sector_div(safepos, data_disks);
-	gap = conf->expand_progress - conf->expand_lo;
+	if (mddev->reshape_backwards) {
+		writepos -= min_t(sector_t, reshape_sectors, writepos);
+		readpos += reshape_sectors;
+		safepos += reshape_sectors;
+	} else {
+		writepos += reshape_sectors;
+		readpos -= min_t(sector_t, reshape_sectors, readpos);
+		safepos -= min_t(sector_t, reshape_sectors, safepos);
+	}
+
+	/* Having calculated the 'writepos' possibly use it
+	 * to set 'stripe_addr' which is where we will write to.
+	 */
+	if (mddev->reshape_backwards) {
+		BUG_ON(conf->reshape_progress == 0);
+		stripe_addr = writepos;
+		BUG_ON((mddev->dev_sectors &
+			~((sector_t)reshape_sectors - 1))
+		       - reshape_sectors - stripe_addr
+		       != sector_nr);
+	} else {
+		BUG_ON(writepos != sector_nr + reshape_sectors);
+		stripe_addr = sector_nr;
+	}
+
+	/* 'writepos' is the most advanced device address we might write.
+	 * 'readpos' is the least advanced device address we might read.
+	 * 'safepos' is the least address recorded in the metadata as having
+	 *     been reshaped.
+	 * If there is a min_offset_diff, these are adjusted either by
+	 * increasing the safepos/readpos if diff is negative, or
+	 * increasing writepos if diff is positive.
+	 * If 'readpos' is then behind 'writepos', there is no way that we can
+	 * ensure safety in the face of a crash - that must be done by userspace
+	 * making a backup of the data.  So in that case there is no particular
+	 * rush to update metadata.
+	 * Otherwise if 'safepos' is behind 'writepos', then we really need to
+	 * update the metadata to advance 'safepos' to match 'readpos' so that
+	 * we can be safe in the event of a crash.
+	 * So we insist on updating metadata if safepos is behind writepos and
+	 * readpos is beyond writepos.
+	 * In any case, update the metadata every 10 seconds.
+	 * Maybe that number should be configurable, but I'm not sure it is
+	 * worth it.... maybe it could be a multiple of safemode_delay???
+	 */
+	if (conf->min_offset_diff < 0) {
+		safepos += -conf->min_offset_diff;
+		readpos += -conf->min_offset_diff;
+	} else
+		writepos += conf->min_offset_diff;
 
-	if (writepos >= safepos ||
-	    gap > (new_data_disks)*3000*2 /*3Meg*/) {
+	if ((mddev->reshape_backwards
+	     ? (safepos > writepos && readpos < writepos)
+	     : (safepos < writepos && readpos > writepos)) ||
+	    time_after(jiffies, conf->reshape_checkpoint + 10*HZ)) {
 		/* Cannot proceed until we've updated the superblock... */
 		wait_event(conf->wait_for_overlap,
-			   atomic_read(&conf->reshape_stripes)==0);
-		mddev->reshape_position = conf->expand_progress;
+			   atomic_read(&conf->reshape_stripes)==0
+			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (atomic_read(&conf->reshape_stripes) != 0)
+			return 0;
+		mddev->reshape_position = conf->reshape_progress;
+		mddev->curr_resync_completed = sector_nr;
+		conf->reshape_checkpoint = jiffies;
 		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		md_wakeup_thread(mddev->thread);
 		wait_event(mddev->sb_wait, mddev->flags == 0 ||
-			   kthread_should_stop());
+			   test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			return 0;
 		spin_lock_irq(&conf->device_lock);
-		conf->expand_lo = mddev->reshape_position;
+		conf->reshape_safe = mddev->reshape_position;
 		spin_unlock_irq(&conf->device_lock);
 		wake_up(&conf->wait_for_overlap);
+		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
 	}
 
-	for (i=0; i < conf->chunk_size/512; i+= STRIPE_SECTORS) {
+	INIT_LIST_HEAD(&stripes);
+	for (i = 0; i < reshape_sectors; i += STRIPE_SECTORS) {
 		int j;
-		int skipped = 0;
-		pd_idx = stripe_to_pdidx(sector_nr+i, conf, conf->raid_disks);
-		sh = get_active_stripe(conf, sector_nr+i,
-				       conf->raid_disks, pd_idx, 0);
+		int skipped_disk = 0;
+		sh = get_active_stripe(conf, stripe_addr+i, 0, 0, 1);
 		set_bit(STRIPE_EXPANDING, &sh->state);
 		atomic_inc(&conf->reshape_stripes);
 		/* If any of this stripe is beyond the end of the old
@@ -3558,25 +4909,28 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
 			if (j == sh->pd_idx)
 				continue;
 			if (conf->level == 6 &&
-			    j == raid6_next_disk(sh->pd_idx, sh->disks))
+			    j == sh->qd_idx)
 				continue;
-			s = compute_blocknr(sh, j);
-			if (s < mddev->array_sectors) {
-				skipped = 1;
+			s = compute_blocknr(sh, j, 0);
+			if (s < raid5_size(mddev, 0, 0)) {
+				skipped_disk = 1;
 				continue;
 			}
 			memset(page_address(sh->dev[j].page), 0, STRIPE_SIZE);
 			set_bit(R5_Expanded, &sh->dev[j].flags);
 			set_bit(R5_UPTODATE, &sh->dev[j].flags);
 		}
-		if (!skipped) {
+		if (!skipped_disk) {
 			set_bit(STRIPE_EXPAND_READY, &sh->state);
 			set_bit(STRIPE_HANDLE, &sh->state);
 		}
-		release_stripe(sh);
+		list_add(&sh->lru, &stripes);
 	}
 	spin_lock_irq(&conf->device_lock);
-	conf->expand_progress = (sector_nr + i) * new_data_disks;
+	if (mddev->reshape_backwards)
+		conf->reshape_progress -= reshape_sectors * new_data_disks;
+	else
+		conf->reshape_progress += reshape_sectors * new_data_disks;
 	spin_unlock_irq(&conf->device_lock);
 	/* Ok, those stripe are ready. We can start scheduling
 	 * reads on the source stripes.
@@ -3584,63 +4938,74 @@ static sector_t reshape_request(mddev_t *mddev, sector_t sector_nr, int *skipped
 	 * block on the destination stripes.
 	 */
 	first_sector =
-		raid5_compute_sector(sector_nr*(new_data_disks),
-				     raid_disks, data_disks,
-				     &dd_idx, &pd_idx, conf);
+		raid5_compute_sector(conf, stripe_addr*(new_data_disks),
+				     1, &dd_idx, NULL);
 	last_sector =
-		raid5_compute_sector((sector_nr+conf->chunk_size/512)
-				     *(new_data_disks) -1,
-				     raid_disks, data_disks,
-				     &dd_idx, &pd_idx, conf);
-	if (last_sector >= (mddev->size<<1))
-		last_sector = (mddev->size<<1)-1;
+		raid5_compute_sector(conf, ((stripe_addr+reshape_sectors)
+					    * new_data_disks - 1),
+				     1, &dd_idx, NULL);
+	if (last_sector >= mddev->dev_sectors)
+		last_sector = mddev->dev_sectors - 1;
 	while (first_sector <= last_sector) {
-		pd_idx = stripe_to_pdidx(first_sector, conf,
-					 conf->previous_raid_disks);
-		sh = get_active_stripe(conf, first_sector,
-				       conf->previous_raid_disks, pd_idx, 0);
+		sh = get_active_stripe(conf, first_sector, 1, 0, 1);
 		set_bit(STRIPE_EXPAND_SOURCE, &sh->state);
 		set_bit(STRIPE_HANDLE, &sh->state);
 		release_stripe(sh);
 		first_sector += STRIPE_SECTORS;
 	}
+	/* Now that the sources are clearly marked, we can release
+	 * the destination stripes
+	 */
+	while (!list_empty(&stripes)) {
+		sh = list_entry(stripes.next, struct stripe_head, lru);
+		list_del_init(&sh->lru);
+		release_stripe(sh);
+	}
 	/* If this takes us to the resync_max point where we have to pause,
 	 * then we need to write out the superblock.
 	 */
-	sector_nr += conf->chunk_size>>9;
-	if (sector_nr >= mddev->resync_max) {
+	sector_nr += reshape_sectors;
+	if ((sector_nr - mddev->curr_resync_completed) * 2
+	    >= mddev->resync_max - mddev->curr_resync_completed) {
 		/* Cannot proceed until we've updated the superblock... */
 		wait_event(conf->wait_for_overlap,
-			   atomic_read(&conf->reshape_stripes) == 0);
-		mddev->reshape_position = conf->expand_progress;
+			   atomic_read(&conf->reshape_stripes) == 0
+			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (atomic_read(&conf->reshape_stripes) != 0)
+			goto ret;
+		mddev->reshape_position = conf->reshape_progress;
+		mddev->curr_resync_completed = sector_nr;
+		conf->reshape_checkpoint = jiffies;
 		set_bit(MD_CHANGE_DEVS, &mddev->flags);
 		md_wakeup_thread(mddev->thread);
 		wait_event(mddev->sb_wait,
 			   !test_bit(MD_CHANGE_DEVS, &mddev->flags)
-			   || kthread_should_stop());
+			   || test_bit(MD_RECOVERY_INTR, &mddev->recovery));
+		if (test_bit(MD_RECOVERY_INTR, &mddev->recovery))
+			goto ret;
 		spin_lock_irq(&conf->device_lock);
-		conf->expand_lo = mddev->reshape_position;
+		conf->reshape_safe = mddev->reshape_position;
 		spin_unlock_irq(&conf->device_lock);
 		wake_up(&conf->wait_for_overlap);
+		sysfs_notify(&mddev->kobj, NULL, "sync_completed");
 	}
-	return conf->chunk_size>>9;
+ret:
+	return reshape_sectors;
 }
 
 /* FIXME go_faster isn't used */
-static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *skipped, int go_faster)
+static inline sector_t sync_request(struct mddev *mddev, sector_t sector_nr, int *skipped, int go_faster)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	struct r5conf *conf = mddev->private;
 	struct stripe_head *sh;
-	int pd_idx;
-	int raid_disks = conf->raid_disks;
-	sector_t max_sector = mddev->size << 1;
-	int sync_blocks;
+	sector_t max_sector = mddev->dev_sectors;
+	sector_t sync_blocks;
 	int still_degraded = 0;
 	int i;
 
 	if (sector_nr >= max_sector) {
 		/* just being told to finish up .. nothing much to do */
-		unplug_slaves(mddev);
+
 		if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery)) {
 			end_reshape(conf);
 			return 0;
@@ -3656,6 +5021,9 @@ static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *ski
 		return 0;
 	}
 
+	/* Allow raid5_quiesce to complete */
+	wait_event(conf->wait_for_overlap, conf->quiesce != 2);
+
 	if (test_bit(MD_RECOVERY_RESHAPE, &mddev->recovery))
 		return reshape_request(mddev, sector_nr, skipped);
 
@@ -3671,26 +5039,25 @@ static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *ski
 	 */
 	if (mddev->degraded >= conf->max_degraded &&
 	    test_bit(MD_RECOVERY_SYNC, &mddev->recovery)) {
-		sector_t rv = (mddev->size << 1) - sector_nr;
+		sector_t rv = mddev->dev_sectors - sector_nr;
 		*skipped = 1;
 		return rv;
 	}
-	if (!bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
-	    !test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
-	    !conf->fullsync && sync_blocks >= STRIPE_SECTORS) {
+	if (!test_bit(MD_RECOVERY_REQUESTED, &mddev->recovery) &&
+	    !conf->fullsync &&
+	    !bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, 1) &&
+	    sync_blocks >= STRIPE_SECTORS) {
 		/* we can skip this block, and probably more */
 		sync_blocks /= STRIPE_SECTORS;
 		*skipped = 1;
 		return sync_blocks * STRIPE_SECTORS; /* keep things rounded to whole stripes */
 	}
 
-
 	bitmap_cond_end_sync(mddev->bitmap, sector_nr);
 
-	pd_idx = stripe_to_pdidx(sector_nr, conf, raid_disks);
-	sh = get_active_stripe(conf, sector_nr, raid_disks, pd_idx, 1);
+	sh = get_active_stripe(conf, sector_nr, 0, 1, 0);
 	if (sh == NULL) {
-		sh = get_active_stripe(conf, sector_nr, raid_disks, pd_idx, 0);
+		sh = get_active_stripe(conf, sector_nr, 0, 0, 0);
 		/* make sure we don't swamp the stripe cache if someone else
 		 * is trying to get access
 		 */
@@ -3700,26 +5067,21 @@ static inline sector_t sync_request(mddev_t *mddev, sector_t sector_nr, int *ski
 	 * We don't need to check the 'failed' flag as when that gets set,
 	 * recovery aborts.
 	 */
-	for (i=0; i<mddev->raid_disks; i++)
+	for (i = 0; i < conf->raid_disks; i++)
 		if (conf->disks[i].rdev == NULL)
 			still_degraded = 1;
 
 	bitmap_start_sync(mddev->bitmap, sector_nr, &sync_blocks, still_degraded);
 
-	spin_lock(&sh->lock);
-	set_bit(STRIPE_SYNCING, &sh->state);
-	clear_bit(STRIPE_INSYNC, &sh->state);
-	spin_unlock(&sh->lock);
+	set_bit(STRIPE_SYNC_REQUESTED, &sh->state);
+	set_bit(STRIPE_HANDLE, &sh->state);
 
-	/* wait for any blocked device to be handled */
-	while(unlikely(!handle_stripe(sh, NULL)))
-		;
 	release_stripe(sh);
 
 	return STRIPE_SECTORS;
 }
 
-static int  retry_aligned_read(raid5_conf_t *conf, struct bio *raid_bio)
+static int  retry_aligned_read(struct r5conf *conf, struct bio *raid_bio)
 {
 	/* We may not be able to submit a whole bio at once as there
 	 * may not be enough stripe_heads available.
@@ -3732,62 +5094,135 @@ static int  retry_aligned_read(raid5_conf_t *conf, struct bio *raid_bio)
 	 * it will be only one 'dd_idx' and only need one call to raid5_compute_sector.
 	 */
 	struct stripe_head *sh;
-	int dd_idx, pd_idx;
+	int dd_idx;
 	sector_t sector, logical_sector, last_sector;
 	int scnt = 0;
 	int remaining;
 	int handled = 0;
 
-	logical_sector = raid_bio->bi_sector & ~((sector_t)STRIPE_SECTORS-1);
-	sector = raid5_compute_sector(	logical_sector,
-					conf->raid_disks,
-					conf->raid_disks - conf->max_degraded,
-					&dd_idx,
-					&pd_idx,
-					conf);
-	last_sector = raid_bio->bi_sector + (raid_bio->bi_size>>9);
+	logical_sector = raid_bio->bi_iter.bi_sector &
+		~((sector_t)STRIPE_SECTORS-1);
+	sector = raid5_compute_sector(conf, logical_sector,
+				      0, &dd_idx, NULL);
+	last_sector = bio_end_sector(raid_bio);
 
 	for (; logical_sector < last_sector;
 	     logical_sector += STRIPE_SECTORS,
 		     sector += STRIPE_SECTORS,
 		     scnt++) {
 
-		if (scnt < raid_bio->bi_hw_segments)
+		if (scnt < raid5_bi_processed_stripes(raid_bio))
 			/* already done this stripe */
 			continue;
 
-		sh = get_active_stripe(conf, sector, conf->raid_disks, pd_idx, 1);
+		sh = get_active_stripe(conf, sector, 0, 1, 1);
 
 		if (!sh) {
 			/* failed to get a stripe - must wait */
-			raid_bio->bi_hw_segments = scnt;
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
-		set_bit(R5_ReadError, &sh->dev[dd_idx].flags);
 		if (!add_stripe_bio(sh, raid_bio, dd_idx, 0)) {
 			release_stripe(sh);
-			raid_bio->bi_hw_segments = scnt;
+			raid5_set_bi_processed_stripes(raid_bio, scnt);
 			conf->retry_read_aligned = raid_bio;
 			return handled;
 		}
 
-		handle_stripe(sh, NULL);
+		set_bit(R5_ReadNoMerge, &sh->dev[dd_idx].flags);
+		handle_stripe(sh);
 		release_stripe(sh);
 		handled++;
 	}
-	spin_lock_irq(&conf->device_lock);
-	remaining = --raid_bio->bi_phys_segments;
-	spin_unlock_irq(&conf->device_lock);
-	if (remaining == 0)
+	remaining = raid5_dec_bi_active_stripes(raid_bio);
+	if (remaining == 0) {
+		trace_block_bio_complete(bdev_get_queue(raid_bio->bi_bdev),
+					 raid_bio, 0);
 		bio_endio(raid_bio, 0);
+	}
 	if (atomic_dec_and_test(&conf->active_aligned_reads))
 		wake_up(&conf->wait_for_stripe);
 	return handled;
 }
 
+static int handle_active_stripes(struct r5conf *conf, int group,
+				 struct r5worker *worker,
+				 struct list_head *temp_inactive_list)
+{
+	struct stripe_head *batch[MAX_STRIPE_BATCH], *sh;
+	int i, batch_size = 0, hash;
+	bool release_inactive = false;
+
+	while (batch_size < MAX_STRIPE_BATCH &&
+			(sh = __get_priority_stripe(conf, group)) != NULL)
+		batch[batch_size++] = sh;
+
+	if (batch_size == 0) {
+		for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
+			if (!list_empty(temp_inactive_list + i))
+				break;
+		if (i == NR_STRIPE_HASH_LOCKS)
+			return batch_size;
+		release_inactive = true;
+	}
+	spin_unlock_irq(&conf->device_lock);
+
+	release_inactive_stripe_list(conf, temp_inactive_list,
+				     NR_STRIPE_HASH_LOCKS);
+
+	if (release_inactive) {
+		spin_lock_irq(&conf->device_lock);
+		return 0;
+	}
+
+	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++) {
+		hash = batch[i]->hash_lock_index;
+		__release_stripe(conf, batch[i], &temp_inactive_list[hash]);
+	}
+	return batch_size;
+}
+
+static void raid5_do_work(struct work_struct *work)
+{
+	struct r5worker *worker = container_of(work, struct r5worker, work);
+	struct r5worker_group *group = worker->group;
+	struct r5conf *conf = group->conf;
+	int group_id = group - conf->worker_groups;
+	int handled;
+	struct blk_plug plug;
+
+	pr_debug("+++ raid5worker active\n");
+
+	blk_start_plug(&plug);
+	handled = 0;
+	spin_lock_irq(&conf->device_lock);
+	while (1) {
+		int batch_size, released;
+
+		released = release_stripe_list(conf, worker->temp_inactive_list);
+
+		batch_size = handle_active_stripes(conf, group_id, worker,
+						   worker->temp_inactive_list);
+		worker->working = false;
+		if (!batch_size && !released)
+			break;
+		handled += batch_size;
+	}
+	pr_debug("%d stripes handled\n", handled);
+
+	spin_unlock_irq(&conf->device_lock);
+	blk_finish_plug(&plug);
+
+	pr_debug("--- raid5worker inactive\n");
+}
 
 /*
  * This is our raid5 kernel thread.
@@ -3796,29 +5231,37 @@ static int  retry_aligned_read(raid5_conf_t *conf, struct bio *raid_bio)
  * During the scan, completed stripes are saved for us by the interrupt
  * handler, so that they will not have to wait for our next wakeup.
  */
-static void raid5d(mddev_t *mddev)
+static void raid5d(struct md_thread *thread)
 {
-	struct stripe_head *sh;
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct mddev *mddev = thread->mddev;
+	struct r5conf *conf = mddev->private;
 	int handled;
+	struct blk_plug plug;
 
 	pr_debug("+++ raid5d active\n");
 
 	md_check_recovery(mddev);
 
+	blk_start_plug(&plug);
 	handled = 0;
 	spin_lock_irq(&conf->device_lock);
 	while (1) {
 		struct bio *bio;
+		int batch_size, released;
 
-		if (conf->seq_flush != conf->seq_write) {
-			int seq = conf->seq_flush;
+		released = release_stripe_list(conf, conf->temp_inactive_list);
+
+		if (
+		    !list_empty(&conf->bitmap_list)) {
+			/* Now is a good time to flush some bitmap updates */
+			conf->seq_flush++;
 			spin_unlock_irq(&conf->device_lock);
 			bitmap_unplug(mddev->bitmap);
 			spin_lock_irq(&conf->device_lock);
-			conf->seq_write = seq;
-			activate_bit_delay(conf);
+			conf->seq_write = conf->seq_flush;
+			activate_bit_delay(conf, conf->temp_inactive_list);
 		}
+		raid5_activate_delayed(conf);
 
 		while ((bio = remove_bio_from_retry(conf))) {
 			int ok;
@@ -3830,68 +5273,88 @@ static void raid5d(mddev_t *mddev)
 			handled++;
 		}
 
-		sh = __get_priority_stripe(conf);
-
-		if (!sh)
+		batch_size = handle_active_stripes(conf, ANY_GROUP, NULL,
+						   conf->temp_inactive_list);
+		if (!batch_size && !released)
 			break;
-		spin_unlock_irq(&conf->device_lock);
-		
-		handled++;
-		handle_stripe(sh, conf->spare_page);
-		release_stripe(sh);
+		handled += batch_size;
 
-		spin_lock_irq(&conf->device_lock);
+		if (mddev->flags & ~(1<<MD_CHANGE_PENDING)) {
+			spin_unlock_irq(&conf->device_lock);
+			md_check_recovery(mddev);
+			spin_lock_irq(&conf->device_lock);
+		}
 	}
 	pr_debug("%d stripes handled\n", handled);
 
 	spin_unlock_irq(&conf->device_lock);
 
 	async_tx_issue_pending_all();
-	unplug_slaves(mddev);
+	blk_finish_plug(&plug);
 
 	pr_debug("--- raid5d inactive\n");
 }
 
 static ssize_t
-raid5_show_stripe_cache_size(mddev_t *mddev, char *page)
+raid5_show_stripe_cache_size(struct mddev *mddev, char *page)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
 	if (conf)
 		return sprintf(page, "%d\n", conf->max_nr_stripes);
 	else
 		return 0;
 }
 
-static ssize_t
-raid5_store_stripe_cache_size(mddev_t *mddev, const char *page, size_t len)
+int
+raid5_set_cache_size(struct mddev *mddev, int size)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	unsigned long new;
+	struct r5conf *conf = mddev->private;
 	int err;
+	int hash;
 
-	if (len >= PAGE_SIZE)
-		return -EINVAL;
-	if (!conf)
-		return -ENODEV;
-
-	if (strict_strtoul(page, 10, &new))
+	if (size <= 16 || size > 32768)
 		return -EINVAL;
-	if (new <= 16 || new > 32768)
-		return -EINVAL;
-	while (new < conf->max_nr_stripes) {
-		if (drop_one_stripe(conf))
+	hash = (conf->max_nr_stripes - 1) % NR_STRIPE_HASH_LOCKS;
+	while (size < conf->max_nr_stripes) {
+		if (drop_one_stripe(conf, hash))
 			conf->max_nr_stripes--;
 		else
 			break;
+		hash--;
+		if (hash < 0)
+			hash = NR_STRIPE_HASH_LOCKS - 1;
 	}
 	err = md_allow_write(mddev);
 	if (err)
 		return err;
-	while (new > conf->max_nr_stripes) {
-		if (grow_one_stripe(conf))
+	hash = conf->max_nr_stripes % NR_STRIPE_HASH_LOCKS;
+	while (size > conf->max_nr_stripes) {
+		if (grow_one_stripe(conf, hash))
 			conf->max_nr_stripes++;
 		else break;
+		hash = (hash + 1) % NR_STRIPE_HASH_LOCKS;
 	}
+	return 0;
+}
+EXPORT_SYMBOL(raid5_set_cache_size);
+
+static ssize_t
+raid5_store_stripe_cache_size(struct mddev *mddev, const char *page, size_t len)
+{
+	struct r5conf *conf = mddev->private;
+	unsigned long new;
+	int err;
+
+	if (len >= PAGE_SIZE)
+		return -EINVAL;
+	if (!conf)
+		return -ENODEV;
+
+	if (kstrtoul(page, 10, &new))
+		return -EINVAL;
+	err = raid5_set_cache_size(mddev, new);
+	if (err)
+		return err;
 	return len;
 }
 
@@ -3901,9 +5364,9 @@ raid5_stripecache_size = __ATTR(stripe_cache_size, S_IRUGO | S_IWUSR,
 				raid5_store_stripe_cache_size);
 
 static ssize_t
-raid5_show_preread_threshold(mddev_t *mddev, char *page)
+raid5_show_preread_threshold(struct mddev *mddev, char *page)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
 	if (conf)
 		return sprintf(page, "%d\n", conf->bypass_threshold);
 	else
@@ -3911,16 +5374,16 @@ raid5_show_preread_threshold(mddev_t *mddev, char *page)
 }
 
 static ssize_t
-raid5_store_preread_threshold(mddev_t *mddev, const char *page, size_t len)
+raid5_store_preread_threshold(struct mddev *mddev, const char *page, size_t len)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
 	unsigned long new;
 	if (len >= PAGE_SIZE)
 		return -EINVAL;
 	if (!conf)
 		return -ENODEV;
 
-	if (strict_strtoul(page, 10, &new))
+	if (kstrtoul(page, 10, &new))
 		return -EINVAL;
 	if (new > conf->max_nr_stripes)
 		return -EINVAL;
@@ -3935,9 +5398,53 @@ raid5_preread_bypass_threshold = __ATTR(preread_bypass_threshold,
 					raid5_store_preread_threshold);
 
 static ssize_t
-stripe_cache_active_show(mddev_t *mddev, char *page)
+raid5_show_skip_copy(struct mddev *mddev, char *page)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
+	if (conf)
+		return sprintf(page, "%d\n", conf->skip_copy);
+	else
+		return 0;
+}
+
+static ssize_t
+raid5_store_skip_copy(struct mddev *mddev, const char *page, size_t len)
+{
+	struct r5conf *conf = mddev->private;
+	unsigned long new;
+	if (len >= PAGE_SIZE)
+		return -EINVAL;
+	if (!conf)
+		return -ENODEV;
+
+	if (kstrtoul(page, 10, &new))
+		return -EINVAL;
+	new = !!new;
+	if (new == conf->skip_copy)
+		return len;
+
+	mddev_suspend(mddev);
+	conf->skip_copy = new;
+	if (new)
+		mddev->queue->backing_dev_info.capabilities |=
+						BDI_CAP_STABLE_WRITES;
+	else
+		mddev->queue->backing_dev_info.capabilities &=
+						~BDI_CAP_STABLE_WRITES;
+	mddev_resume(mddev);
+	return len;
+}
+
+static struct md_sysfs_entry
+raid5_skip_copy = __ATTR(skip_copy, S_IRUGO | S_IWUSR,
+					raid5_show_skip_copy,
+					raid5_store_skip_copy);
+
+
+static ssize_t
+stripe_cache_active_show(struct mddev *mddev, char *page)
+{
+	struct r5conf *conf = mddev->private;
 	if (conf)
 		return sprintf(page, "%d\n", atomic_read(&conf->active_stripes));
 	else
@@ -3947,10 +5454,79 @@ stripe_cache_active_show(mddev_t *mddev, char *page)
 static struct md_sysfs_entry
 raid5_stripecache_active = __ATTR_RO(stripe_cache_active);
 
+static ssize_t
+raid5_show_group_thread_cnt(struct mddev *mddev, char *page)
+{
+	struct r5conf *conf = mddev->private;
+	if (conf)
+		return sprintf(page, "%d\n", conf->worker_cnt_per_group);
+	else
+		return 0;
+}
+
+static int alloc_thread_groups(struct r5conf *conf, int cnt,
+			       int *group_cnt,
+			       int *worker_cnt_per_group,
+			       struct r5worker_group **worker_groups);
+static ssize_t
+raid5_store_group_thread_cnt(struct mddev *mddev, const char *page, size_t len)
+{
+	struct r5conf *conf = mddev->private;
+	unsigned long new;
+	int err;
+	struct r5worker_group *new_groups, *old_groups;
+	int group_cnt, worker_cnt_per_group;
+
+	if (len >= PAGE_SIZE)
+		return -EINVAL;
+	if (!conf)
+		return -ENODEV;
+
+	if (kstrtoul(page, 10, &new))
+		return -EINVAL;
+
+	if (new == conf->worker_cnt_per_group)
+		return len;
+
+	mddev_suspend(mddev);
+
+	old_groups = conf->worker_groups;
+	if (old_groups)
+		flush_workqueue(raid5_wq);
+
+	err = alloc_thread_groups(conf, new,
+				  &group_cnt, &worker_cnt_per_group,
+				  &new_groups);
+	if (!err) {
+		spin_lock_irq(&conf->device_lock);
+		conf->group_cnt = group_cnt;
+		conf->worker_cnt_per_group = worker_cnt_per_group;
+		conf->worker_groups = new_groups;
+		spin_unlock_irq(&conf->device_lock);
+
+		if (old_groups)
+			kfree(old_groups[0].workers);
+		kfree(old_groups);
+	}
+
+	mddev_resume(mddev);
+
+	if (err)
+		return err;
+	return len;
+}
+
+static struct md_sysfs_entry
+raid5_group_thread_cnt = __ATTR(group_thread_cnt, S_IRUGO | S_IWUSR,
+				raid5_show_group_thread_cnt,
+				raid5_store_group_thread_cnt);
+
 static struct attribute *raid5_attrs[] =  {
 	&raid5_stripecache_size.attr,
 	&raid5_stripecache_active.attr,
 	&raid5_preread_bypass_threshold.attr,
+	&raid5_group_thread_cnt.attr,
+	&raid5_skip_copy.attr,
 	NULL,
 };
 static struct attribute_group raid5_attrs_group = {
@@ -3958,441 +5534,895 @@ static struct attribute_group raid5_attrs_group = {
 	.attrs = raid5_attrs,
 };
 
-static int run(mddev_t *mddev)
+static int alloc_thread_groups(struct r5conf *conf, int cnt,
+			       int *group_cnt,
+			       int *worker_cnt_per_group,
+			       struct r5worker_group **worker_groups)
+{
+	int i, j, k;
+	ssize_t size;
+	struct r5worker *workers;
+
+	*worker_cnt_per_group = cnt;
+	if (cnt == 0) {
+		*group_cnt = 0;
+		*worker_groups = NULL;
+		return 0;
+	}
+	*group_cnt = num_possible_nodes();
+	size = sizeof(struct r5worker) * cnt;
+	workers = kzalloc(size * *group_cnt, GFP_NOIO);
+	*worker_groups = kzalloc(sizeof(struct r5worker_group) *
+				*group_cnt, GFP_NOIO);
+	if (!*worker_groups || !workers) {
+		kfree(workers);
+		kfree(*worker_groups);
+		return -ENOMEM;
+	}
+
+	for (i = 0; i < *group_cnt; i++) {
+		struct r5worker_group *group;
+
+		group = &(*worker_groups)[i];
+		INIT_LIST_HEAD(&group->handle_list);
+		group->conf = conf;
+		group->workers = workers + i * cnt;
+
+		for (j = 0; j < cnt; j++) {
+			struct r5worker *worker = group->workers + j;
+			worker->group = group;
+			INIT_WORK(&worker->work, raid5_do_work);
+
+			for (k = 0; k < NR_STRIPE_HASH_LOCKS; k++)
+				INIT_LIST_HEAD(worker->temp_inactive_list + k);
+		}
+	}
+
+	return 0;
+}
+
+static void free_thread_groups(struct r5conf *conf)
+{
+	if (conf->worker_groups)
+		kfree(conf->worker_groups[0].workers);
+	kfree(conf->worker_groups);
+	conf->worker_groups = NULL;
+}
+
+static sector_t
+raid5_size(struct mddev *mddev, sector_t sectors, int raid_disks)
+{
+	struct r5conf *conf = mddev->private;
+
+	if (!sectors)
+		sectors = mddev->dev_sectors;
+	if (!raid_disks)
+		/* size is defined by the smallest of previous and new size */
+		raid_disks = min(conf->raid_disks, conf->previous_raid_disks);
+
+	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
+	sectors &= ~((sector_t)mddev->new_chunk_sectors - 1);
+	return sectors * (raid_disks - conf->max_degraded);
+}
+
+static void free_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
+{
+	safe_put_page(percpu->spare_page);
+	kfree(percpu->scribble);
+	percpu->spare_page = NULL;
+	percpu->scribble = NULL;
+}
+
+static int alloc_scratch_buffer(struct r5conf *conf, struct raid5_percpu *percpu)
+{
+	if (conf->level == 6 && !percpu->spare_page)
+		percpu->spare_page = alloc_page(GFP_KERNEL);
+	if (!percpu->scribble)
+		percpu->scribble = kmalloc(conf->scribble_len, GFP_KERNEL);
+
+	if (!percpu->scribble || (conf->level == 6 && !percpu->spare_page)) {
+		free_scratch_buffer(conf, percpu);
+		return -ENOMEM;
+	}
+
+	return 0;
+}
+
+static void raid5_free_percpu(struct r5conf *conf)
+{
+	unsigned long cpu;
+
+	if (!conf->percpu)
+		return;
+
+#ifdef CONFIG_HOTPLUG_CPU
+	unregister_cpu_notifier(&conf->cpu_notify);
+#endif
+
+	get_online_cpus();
+	for_each_possible_cpu(cpu)
+		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
+	put_online_cpus();
+
+	free_percpu(conf->percpu);
+}
+
+static void free_conf(struct r5conf *conf)
+{
+	free_thread_groups(conf);
+	shrink_stripes(conf);
+	raid5_free_percpu(conf);
+	kfree(conf->disks);
+	kfree(conf->stripe_hashtbl);
+	kfree(conf);
+}
+
+#ifdef CONFIG_HOTPLUG_CPU
+static int raid456_cpu_notify(struct notifier_block *nfb, unsigned long action,
+			      void *hcpu)
+{
+	struct r5conf *conf = container_of(nfb, struct r5conf, cpu_notify);
+	long cpu = (long)hcpu;
+	struct raid5_percpu *percpu = per_cpu_ptr(conf->percpu, cpu);
+
+	switch (action) {
+	case CPU_UP_PREPARE:
+	case CPU_UP_PREPARE_FROZEN:
+		if (alloc_scratch_buffer(conf, percpu)) {
+			pr_err("%s: failed memory allocation for cpu%ld\n",
+			       __func__, cpu);
+			return notifier_from_errno(-ENOMEM);
+		}
+		break;
+	case CPU_DEAD:
+	case CPU_DEAD_FROZEN:
+		free_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
+		break;
+	default:
+		break;
+	}
+	return NOTIFY_OK;
+}
+#endif
+
+static int raid5_alloc_percpu(struct r5conf *conf)
+{
+	unsigned long cpu;
+	int err = 0;
+
+	conf->percpu = alloc_percpu(struct raid5_percpu);
+	if (!conf->percpu)
+		return -ENOMEM;
+
+#ifdef CONFIG_HOTPLUG_CPU
+	conf->cpu_notify.notifier_call = raid456_cpu_notify;
+	conf->cpu_notify.priority = 0;
+	err = register_cpu_notifier(&conf->cpu_notify);
+	if (err)
+		return err;
+#endif
+
+	get_online_cpus();
+	for_each_present_cpu(cpu) {
+		err = alloc_scratch_buffer(conf, per_cpu_ptr(conf->percpu, cpu));
+		if (err) {
+			pr_err("%s: failed memory allocation for cpu%ld\n",
+			       __func__, cpu);
+			break;
+		}
+	}
+	put_online_cpus();
+
+	return err;
+}
+
+static struct r5conf *setup_conf(struct mddev *mddev)
 {
-	raid5_conf_t *conf;
-	int raid_disk, memory;
-	mdk_rdev_t *rdev;
+	struct r5conf *conf;
+	int raid_disk, memory, max_disks;
+	struct md_rdev *rdev;
 	struct disk_info *disk;
-	struct list_head *tmp;
+	char pers_name[6];
+	int i;
+	int group_cnt, worker_cnt_per_group;
+	struct r5worker_group *new_group;
+
+	if (mddev->new_level != 5
+	    && mddev->new_level != 4
+	    && mddev->new_level != 6) {
+		printk(KERN_ERR "md/raid:%s: raid level not set to 4/5/6 (%d)\n",
+		       mdname(mddev), mddev->new_level);
+		return ERR_PTR(-EIO);
+	}
+	if ((mddev->new_level == 5
+	     && !algorithm_valid_raid5(mddev->new_layout)) ||
+	    (mddev->new_level == 6
+	     && !algorithm_valid_raid6(mddev->new_layout))) {
+		printk(KERN_ERR "md/raid:%s: layout %d not supported\n",
+		       mdname(mddev), mddev->new_layout);
+		return ERR_PTR(-EIO);
+	}
+	if (mddev->new_level == 6 && mddev->raid_disks < 4) {
+		printk(KERN_ERR "md/raid:%s: not enough configured devices (%d, minimum 4)\n",
+		       mdname(mddev), mddev->raid_disks);
+		return ERR_PTR(-EINVAL);
+	}
+
+	if (!mddev->new_chunk_sectors ||
+	    (mddev->new_chunk_sectors << 9) % PAGE_SIZE ||
+	    !is_power_of_2(mddev->new_chunk_sectors)) {
+		printk(KERN_ERR "md/raid:%s: invalid chunk size %d\n",
+		       mdname(mddev), mddev->new_chunk_sectors << 9);
+		return ERR_PTR(-EINVAL);
+	}
+
+	conf = kzalloc(sizeof(struct r5conf), GFP_KERNEL);
+	if (conf == NULL)
+		goto abort;
+	/* Don't enable multi-threading by default*/
+	if (!alloc_thread_groups(conf, 0, &group_cnt, &worker_cnt_per_group,
+				 &new_group)) {
+		conf->group_cnt = group_cnt;
+		conf->worker_cnt_per_group = worker_cnt_per_group;
+		conf->worker_groups = new_group;
+	} else
+		goto abort;
+	spin_lock_init(&conf->device_lock);
+	seqcount_init(&conf->gen_lock);
+	init_waitqueue_head(&conf->wait_for_stripe);
+	init_waitqueue_head(&conf->wait_for_overlap);
+	INIT_LIST_HEAD(&conf->handle_list);
+	INIT_LIST_HEAD(&conf->hold_list);
+	INIT_LIST_HEAD(&conf->delayed_list);
+	INIT_LIST_HEAD(&conf->bitmap_list);
+	init_llist_head(&conf->released_stripes);
+	atomic_set(&conf->active_stripes, 0);
+	atomic_set(&conf->preread_active_stripes, 0);
+	atomic_set(&conf->active_aligned_reads, 0);
+	conf->bypass_threshold = BYPASS_THRESHOLD;
+	conf->recovery_disabled = mddev->recovery_disabled - 1;
+
+	conf->raid_disks = mddev->raid_disks;
+	if (mddev->reshape_position == MaxSector)
+		conf->previous_raid_disks = mddev->raid_disks;
+	else
+		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
+	max_disks = max(conf->raid_disks, conf->previous_raid_disks);
+	conf->scribble_len = scribble_len(max_disks);
+
+	conf->disks = kzalloc(max_disks * sizeof(struct disk_info),
+			      GFP_KERNEL);
+	if (!conf->disks)
+		goto abort;
+
+	conf->mddev = mddev;
+
+	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
+		goto abort;
+
+	/* We init hash_locks[0] separately to that it can be used
+	 * as the reference lock in the spin_lock_nest_lock() call
+	 * in lock_all_device_hash_locks_irq in order to convince
+	 * lockdep that we know what we are doing.
+	 */
+	spin_lock_init(conf->hash_locks);
+	for (i = 1; i < NR_STRIPE_HASH_LOCKS; i++)
+		spin_lock_init(conf->hash_locks + i);
+
+	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
+		INIT_LIST_HEAD(conf->inactive_list + i);
+
+	for (i = 0; i < NR_STRIPE_HASH_LOCKS; i++)
+		INIT_LIST_HEAD(conf->temp_inactive_list + i);
+
+	conf->level = mddev->new_level;
+	if (raid5_alloc_percpu(conf) != 0)
+		goto abort;
+
+	pr_debug("raid456: run(%s) called.\n", mdname(mddev));
+
+	rdev_for_each(rdev, mddev) {
+		raid_disk = rdev->raid_disk;
+		if (raid_disk >= max_disks
+		    || raid_disk < 0)
+			continue;
+		disk = conf->disks + raid_disk;
+
+		if (test_bit(Replacement, &rdev->flags)) {
+			if (disk->replacement)
+				goto abort;
+			disk->replacement = rdev;
+		} else {
+			if (disk->rdev)
+				goto abort;
+			disk->rdev = rdev;
+		}
+
+		if (test_bit(In_sync, &rdev->flags)) {
+			char b[BDEVNAME_SIZE];
+			printk(KERN_INFO "md/raid:%s: device %s operational as raid"
+			       " disk %d\n",
+			       mdname(mddev), bdevname(rdev->bdev, b), raid_disk);
+		} else if (rdev->saved_raid_disk != raid_disk)
+			/* Cannot rely on bitmap to complete recovery */
+			conf->fullsync = 1;
+	}
+
+	conf->chunk_sectors = mddev->new_chunk_sectors;
+	conf->level = mddev->new_level;
+	if (conf->level == 6)
+		conf->max_degraded = 2;
+	else
+		conf->max_degraded = 1;
+	conf->algorithm = mddev->new_layout;
+	conf->reshape_progress = mddev->reshape_position;
+	if (conf->reshape_progress != MaxSector) {
+		conf->prev_chunk_sectors = mddev->chunk_sectors;
+		conf->prev_algo = mddev->layout;
+	}
+
+	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
+		 max_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
+	atomic_set(&conf->empty_inactive_list_nr, NR_STRIPE_HASH_LOCKS);
+	if (grow_stripes(conf, NR_STRIPES)) {
+		printk(KERN_ERR
+		       "md/raid:%s: couldn't allocate %dkB for buffers\n",
+		       mdname(mddev), memory);
+		goto abort;
+	} else
+		printk(KERN_INFO "md/raid:%s: allocated %dkB\n",
+		       mdname(mddev), memory);
+
+	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",
+		       mdname(mddev));
+		goto abort;
+	}
+
+	return conf;
+
+ abort:
+	if (conf) {
+		free_conf(conf);
+		return ERR_PTR(-EIO);
+	} else
+		return ERR_PTR(-ENOMEM);
+}
+
+
+static int only_parity(int raid_disk, int algo, int raid_disks, int max_degraded)
+{
+	switch (algo) {
+	case ALGORITHM_PARITY_0:
+		if (raid_disk < max_degraded)
+			return 1;
+		break;
+	case ALGORITHM_PARITY_N:
+		if (raid_disk >= raid_disks - max_degraded)
+			return 1;
+		break;
+	case ALGORITHM_PARITY_0_6:
+		if (raid_disk == 0 || 
+		    raid_disk == raid_disks - 1)
+			return 1;
+		break;
+	case ALGORITHM_LEFT_ASYMMETRIC_6:
+	case ALGORITHM_RIGHT_ASYMMETRIC_6:
+	case ALGORITHM_LEFT_SYMMETRIC_6:
+	case ALGORITHM_RIGHT_SYMMETRIC_6:
+		if (raid_disk == raid_disks - 1)
+			return 1;
+	}
+	return 0;
+}
+
+static int run(struct mddev *mddev)
+{
+	struct r5conf *conf;
 	int working_disks = 0;
+	int dirty_parity_disks = 0;
+	struct md_rdev *rdev;
+	sector_t reshape_offset = 0;
+	int i;
+	long long min_offset_diff = 0;
+	int first = 1;
+
+	if (mddev->recovery_cp != MaxSector)
+		printk(KERN_NOTICE "md/raid:%s: not clean"
+		       " -- starting background reconstruction\n",
+		       mdname(mddev));
 
-	if (mddev->level != 5 && mddev->level != 4 && mddev->level != 6) {
-		printk(KERN_ERR "raid5: %s: raid level not set to 4/5/6 (%d)\n",
-		       mdname(mddev), mddev->level);
-		return -EIO;
+	rdev_for_each(rdev, mddev) {
+		long long diff;
+		if (rdev->raid_disk < 0)
+			continue;
+		diff = (rdev->new_data_offset - rdev->data_offset);
+		if (first) {
+			min_offset_diff = diff;
+			first = 0;
+		} else if (mddev->reshape_backwards &&
+			 diff < min_offset_diff)
+			min_offset_diff = diff;
+		else if (!mddev->reshape_backwards &&
+			 diff > min_offset_diff)
+			min_offset_diff = diff;
 	}
 
 	if (mddev->reshape_position != MaxSector) {
 		/* Check that we can continue the reshape.
-		 * Currently only disks can change, it must
-		 * increase, and we must be past the point where
-		 * a stripe over-writes itself
+		 * Difficulties arise if the stripe we would write to
+		 * next is at or after the stripe we would read from next.
+		 * For a reshape that changes the number of devices, this
+		 * is only possible for a very short time, and mdadm makes
+		 * sure that time appears to have past before assembling
+		 * the array.  So we fail if that time hasn't passed.
+		 * For a reshape that keeps the number of devices the same
+		 * mdadm must be monitoring the reshape can keeping the
+		 * critical areas read-only and backed up.  It will start
+		 * the array in read-only mode, so we check for that.
 		 */
 		sector_t here_new, here_old;
 		int old_disks;
-		int max_degraded = (mddev->level == 5 ? 1 : 2);
+		int max_degraded = (mddev->level == 6 ? 2 : 1);
 
-		if (mddev->new_level != mddev->level ||
-		    mddev->new_layout != mddev->layout ||
-		    mddev->new_chunk != mddev->chunk_size) {
-			printk(KERN_ERR "raid5: %s: unsupported reshape "
+		if (mddev->new_level != mddev->level) {
+			printk(KERN_ERR "md/raid:%s: unsupported reshape "
 			       "required - aborting.\n",
 			       mdname(mddev));
 			return -EINVAL;
 		}
-		if (mddev->delta_disks <= 0) {
-			printk(KERN_ERR "raid5: %s: unsupported reshape "
-			       "(reduce disks) required - aborting.\n",
-			       mdname(mddev));
-			return -EINVAL;
-		}
 		old_disks = mddev->raid_disks - mddev->delta_disks;
 		/* reshape_position must be on a new-stripe boundary, and one
 		 * further up in new geometry must map after here in old
 		 * geometry.
 		 */
 		here_new = mddev->reshape_position;
-		if (sector_div(here_new, (mddev->chunk_size>>9)*
+		if (sector_div(here_new, mddev->new_chunk_sectors *
 			       (mddev->raid_disks - max_degraded))) {
-			printk(KERN_ERR "raid5: reshape_position not "
-			       "on a stripe boundary\n");
+			printk(KERN_ERR "md/raid:%s: reshape_position not "
+			       "on a stripe boundary\n", mdname(mddev));
 			return -EINVAL;
 		}
+		reshape_offset = here_new * mddev->new_chunk_sectors;
 		/* here_new is the stripe we will write to */
 		here_old = mddev->reshape_position;
-		sector_div(here_old, (mddev->chunk_size>>9)*
+		sector_div(here_old, mddev->chunk_sectors *
 			   (old_disks-max_degraded));
 		/* here_old is the first stripe that we might need to read
 		 * from */
-		if (here_new >= here_old) {
+		if (mddev->delta_disks == 0) {
+			if ((here_new * mddev->new_chunk_sectors !=
+			     here_old * mddev->chunk_sectors)) {
+				printk(KERN_ERR "md/raid:%s: reshape position is"
+				       " confused - aborting\n", mdname(mddev));
+				return -EINVAL;
+			}
+			/* We cannot be sure it is safe to start an in-place
+			 * reshape.  It is only safe if user-space is monitoring
+			 * and taking constant backups.
+			 * mdadm always starts a situation like this in
+			 * readonly mode so it can take control before
+			 * allowing any writes.  So just check for that.
+			 */
+			if (abs(min_offset_diff) >= mddev->chunk_sectors &&
+			    abs(min_offset_diff) >= mddev->new_chunk_sectors)
+				/* not really in-place - so OK */;
+			else if (mddev->ro == 0) {
+				printk(KERN_ERR "md/raid:%s: in-place reshape "
+				       "must be started in read-only mode "
+				       "- aborting\n",
+				       mdname(mddev));
+				return -EINVAL;
+			}
+		} else if (mddev->reshape_backwards
+		    ? (here_new * mddev->new_chunk_sectors + min_offset_diff <=
+		       here_old * mddev->chunk_sectors)
+		    : (here_new * mddev->new_chunk_sectors >=
+		       here_old * mddev->chunk_sectors + (-min_offset_diff))) {
 			/* Reading from the same stripe as writing to - bad */
-			printk(KERN_ERR "raid5: reshape_position too early for "
-			       "auto-recovery - aborting.\n");
+			printk(KERN_ERR "md/raid:%s: reshape_position too early for "
+			       "auto-recovery - aborting.\n",
+			       mdname(mddev));
 			return -EINVAL;
 		}
-		printk(KERN_INFO "raid5: reshape will continue\n");
+		printk(KERN_INFO "md/raid:%s: reshape will continue\n",
+		       mdname(mddev));
 		/* OK, we should be able to continue; */
-	}
-
-
-	mddev->private = kzalloc(sizeof (raid5_conf_t), GFP_KERNEL);
-	if ((conf = mddev->private) == NULL)
-		goto abort;
-	if (mddev->reshape_position == MaxSector) {
-		conf->previous_raid_disks = conf->raid_disks = mddev->raid_disks;
 	} else {
-		conf->raid_disks = mddev->raid_disks;
-		conf->previous_raid_disks = mddev->raid_disks - mddev->delta_disks;
+		BUG_ON(mddev->level != mddev->new_level);
+		BUG_ON(mddev->layout != mddev->new_layout);
+		BUG_ON(mddev->chunk_sectors != mddev->new_chunk_sectors);
+		BUG_ON(mddev->delta_disks != 0);
 	}
 
-	conf->disks = kzalloc(conf->raid_disks * sizeof(struct disk_info),
-			      GFP_KERNEL);
-	if (!conf->disks)
-		goto abort;
-
-	conf->mddev = mddev;
-
-	if ((conf->stripe_hashtbl = kzalloc(PAGE_SIZE, GFP_KERNEL)) == NULL)
-		goto abort;
+	if (mddev->private == NULL)
+		conf = setup_conf(mddev);
+	else
+		conf = mddev->private;
 
-	if (mddev->level == 6) {
-		conf->spare_page = alloc_page(GFP_KERNEL);
-		if (!conf->spare_page)
-			goto abort;
-	}
-	spin_lock_init(&conf->device_lock);
-	mddev->queue->queue_lock = &conf->device_lock;
-	init_waitqueue_head(&conf->wait_for_stripe);
-	init_waitqueue_head(&conf->wait_for_overlap);
-	INIT_LIST_HEAD(&conf->handle_list);
-	INIT_LIST_HEAD(&conf->hold_list);
-	INIT_LIST_HEAD(&conf->delayed_list);
-	INIT_LIST_HEAD(&conf->bitmap_list);
-	INIT_LIST_HEAD(&conf->inactive_list);
-	atomic_set(&conf->active_stripes, 0);
-	atomic_set(&conf->preread_active_stripes, 0);
-	atomic_set(&conf->active_aligned_reads, 0);
-	conf->bypass_threshold = BYPASS_THRESHOLD;
+	if (IS_ERR(conf))
+		return PTR_ERR(conf);
 
-	pr_debug("raid5: run(%s) called.\n", mdname(mddev));
+	conf->min_offset_diff = min_offset_diff;
+	mddev->thread = conf->thread;
+	conf->thread = NULL;
+	mddev->private = conf;
 
-	rdev_for_each(rdev, tmp, mddev) {
-		raid_disk = rdev->raid_disk;
-		if (raid_disk >= conf->raid_disks
-		    || raid_disk < 0)
+	for (i = 0; i < conf->raid_disks && conf->previous_raid_disks;
+	     i++) {
+		rdev = conf->disks[i].rdev;
+		if (!rdev && conf->disks[i].replacement) {
+			/* The replacement is all we have yet */
+			rdev = conf->disks[i].replacement;
+			conf->disks[i].replacement = NULL;
+			clear_bit(Replacement, &rdev->flags);
+			conf->disks[i].rdev = rdev;
+		}
+		if (!rdev)
 			continue;
-		disk = conf->disks + raid_disk;
-
-		disk->rdev = rdev;
-
+		if (conf->disks[i].replacement &&
+		    conf->reshape_progress != MaxSector) {
+			/* replacements and reshape simply do not mix. */
+			printk(KERN_ERR "md: cannot handle concurrent "
+			       "replacement and reshape.\n");
+			goto abort;
+		}
 		if (test_bit(In_sync, &rdev->flags)) {
-			char b[BDEVNAME_SIZE];
-			printk(KERN_INFO "raid5: device %s operational as raid"
-				" disk %d\n", bdevname(rdev->bdev,b),
-				raid_disk);
 			working_disks++;
-		} else
-			/* Cannot rely on bitmap to complete recovery */
-			conf->fullsync = 1;
+			continue;
+		}
+		/* This disc is not fully in-sync.  However if it
+		 * just stored parity (beyond the recovery_offset),
+		 * when we don't need to be concerned about the
+		 * array being dirty.
+		 * When reshape goes 'backwards', we never have
+		 * partially completed devices, so we only need
+		 * to worry about reshape going forwards.
+		 */
+		/* Hack because v0.91 doesn't store recovery_offset properly. */
+		if (mddev->major_version == 0 &&
+		    mddev->minor_version > 90)
+			rdev->recovery_offset = reshape_offset;
+
+		if (rdev->recovery_offset < reshape_offset) {
+			/* We need to check old and new layout */
+			if (!only_parity(rdev->raid_disk,
+					 conf->algorithm,
+					 conf->raid_disks,
+					 conf->max_degraded))
+				continue;
+		}
+		if (!only_parity(rdev->raid_disk,
+				 conf->prev_algo,
+				 conf->previous_raid_disks,
+				 conf->max_degraded))
+			continue;
+		dirty_parity_disks++;
 	}
 
 	/*
 	 * 0 for a fully functional array, 1 or 2 for a degraded array.
 	 */
-	mddev->degraded = conf->raid_disks - working_disks;
-	conf->mddev = mddev;
-	conf->chunk_size = mddev->chunk_size;
-	conf->level = mddev->level;
-	if (conf->level == 6)
-		conf->max_degraded = 2;
-	else
-		conf->max_degraded = 1;
-	conf->algorithm = mddev->layout;
-	conf->max_nr_stripes = NR_STRIPES;
-	conf->expand_progress = mddev->reshape_position;
+	mddev->degraded = calc_degraded(conf);
 
-	/* device size must be a multiple of chunk size */
-	mddev->size &= ~(mddev->chunk_size/1024 -1);
-	mddev->resync_max_sectors = mddev->size << 1;
-
-	if (conf->level == 6 && conf->raid_disks < 4) {
-		printk(KERN_ERR "raid6: not enough configured devices for %s (%d, minimum 4)\n",
-		       mdname(mddev), conf->raid_disks);
-		goto abort;
-	}
-	if (!conf->chunk_size || conf->chunk_size % 4) {
-		printk(KERN_ERR "raid5: invalid chunk size %d for %s\n",
-			conf->chunk_size, mdname(mddev));
-		goto abort;
-	}
-	if (conf->algorithm > ALGORITHM_RIGHT_SYMMETRIC) {
-		printk(KERN_ERR 
-			"raid5: unsupported parity algorithm %d for %s\n",
-			conf->algorithm, mdname(mddev));
-		goto abort;
-	}
-	if (mddev->degraded > conf->max_degraded) {
-		printk(KERN_ERR "raid5: not enough operational devices for %s"
+	if (has_failed(conf)) {
+		printk(KERN_ERR "md/raid:%s: not enough operational devices"
 			" (%d/%d failed)\n",
 			mdname(mddev), mddev->degraded, conf->raid_disks);
 		goto abort;
 	}
 
-	if (mddev->degraded > 0 &&
+	/* device size must be a multiple of chunk size */
+	mddev->dev_sectors &= ~(mddev->chunk_sectors - 1);
+	mddev->resync_max_sectors = mddev->dev_sectors;
+
+	if (mddev->degraded > dirty_parity_disks &&
 	    mddev->recovery_cp != MaxSector) {
 		if (mddev->ok_start_degraded)
 			printk(KERN_WARNING
-			       "raid5: starting dirty degraded array: %s"
-			       "- data corruption possible.\n",
+			       "md/raid:%s: starting dirty degraded array"
+			       " - data corruption possible.\n",
 			       mdname(mddev));
 		else {
 			printk(KERN_ERR
-			       "raid5: cannot start dirty degraded array for %s\n",
+			       "md/raid:%s: cannot start dirty degraded array.\n",
 			       mdname(mddev));
 			goto abort;
 		}
 	}
 
-	{
-		mddev->thread = md_register_thread(raid5d, mddev, "%s_raid5");
-		if (!mddev->thread) {
-			printk(KERN_ERR 
-				"raid5: couldn't allocate thread for %s\n",
-				mdname(mddev));
-			goto abort;
-		}
-	}
-	memory = conf->max_nr_stripes * (sizeof(struct stripe_head) +
-		 conf->raid_disks * ((sizeof(struct bio) + PAGE_SIZE))) / 1024;
-	if (grow_stripes(conf, conf->max_nr_stripes)) {
-		printk(KERN_ERR 
-			"raid5: couldn't allocate %dkB for buffers\n", memory);
-		shrink_stripes(conf);
-		md_unregister_thread(mddev->thread);
-		goto abort;
-	} else
-		printk(KERN_INFO "raid5: allocated %dkB for %s\n",
-			memory, mdname(mddev));
-
 	if (mddev->degraded == 0)
-		printk("raid5: raid level %d set %s active with %d out of %d"
-			" devices, algorithm %d\n", conf->level, mdname(mddev), 
-			mddev->raid_disks-mddev->degraded, mddev->raid_disks,
-			conf->algorithm);
+		printk(KERN_INFO "md/raid:%s: raid level %d active with %d out of %d"
+		       " devices, algorithm %d\n", mdname(mddev), conf->level,
+		       mddev->raid_disks-mddev->degraded, mddev->raid_disks,
+		       mddev->new_layout);
 	else
-		printk(KERN_ALERT "raid5: raid level %d set %s active with %d"
-			" out of %d devices, algorithm %d\n", conf->level,
-			mdname(mddev), mddev->raid_disks - mddev->degraded,
-			mddev->raid_disks, conf->algorithm);
+		printk(KERN_ALERT "md/raid:%s: raid level %d active with %d"
+		       " out of %d devices, algorithm %d\n",
+		       mdname(mddev), conf->level,
+		       mddev->raid_disks - mddev->degraded,
+		       mddev->raid_disks, mddev->new_layout);
 
 	print_raid5_conf(conf);
 
-	if (conf->expand_progress != MaxSector) {
-		printk("...ok start reshape thread\n");
-		conf->expand_lo = conf->expand_progress;
+	if (conf->reshape_progress != MaxSector) {
+		conf->reshape_safe = conf->reshape_progress;
 		atomic_set(&conf->reshape_stripes, 0);
 		clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
 		clear_bit(MD_RECOVERY_CHECK, &mddev->recovery);
 		set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
 		set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
 		mddev->sync_thread = md_register_thread(md_do_sync, mddev,
-							"%s_reshape");
+							"reshape");
 	}
 
-	/* read-ahead size must cover two whole stripes, which is
-	 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
-	 */
-	{
-		int data_disks = conf->previous_raid_disks - conf->max_degraded;
-		int stripe = data_disks *
-			(mddev->chunk_size / PAGE_SIZE);
-		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
-			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
-	}
 
 	/* Ok, everything is just fine now */
-	if (sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
+	if (mddev->to_remove == &raid5_attrs_group)
+		mddev->to_remove = NULL;
+	else if (mddev->kobj.sd &&
+	    sysfs_create_group(&mddev->kobj, &raid5_attrs_group))
 		printk(KERN_WARNING
 		       "raid5: failed to create sysfs attributes for %s\n",
 		       mdname(mddev));
+	md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
+
+	if (mddev->queue) {
+		int chunk_size;
+		bool discard_supported = true;
+		/* read-ahead size must cover two whole stripes, which
+		 * is 2 * (datadisks) * chunksize where 'n' is the
+		 * number of raid devices
+		 */
+		int data_disks = conf->previous_raid_disks - conf->max_degraded;
+		int stripe = data_disks *
+			((mddev->chunk_sectors << 9) / PAGE_SIZE);
+		if (mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
+			mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
 
-	mddev->queue->unplug_fn = raid5_unplug_device;
-	mddev->queue->backing_dev_info.congested_data = mddev;
-	mddev->queue->backing_dev_info.congested_fn = raid5_congested;
+		blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
 
-	mddev->array_sectors = 2 * mddev->size * (conf->previous_raid_disks -
-					    conf->max_degraded);
+		mddev->queue->backing_dev_info.congested_data = mddev;
+		mddev->queue->backing_dev_info.congested_fn = raid5_congested;
+
+		chunk_size = mddev->chunk_sectors << 9;
+		blk_queue_io_min(mddev->queue, chunk_size);
+		blk_queue_io_opt(mddev->queue, chunk_size *
+				 (conf->raid_disks - conf->max_degraded));
+		mddev->queue->limits.raid_partial_stripes_expensive = 1;
+		/*
+		 * We can only discard a whole stripe. It doesn't make sense to
+		 * discard data disk but write parity disk
+		 */
+		stripe = stripe * PAGE_SIZE;
+		/* Round up to power of 2, as discard handling
+		 * currently assumes that */
+		while ((stripe-1) & stripe)
+			stripe = (stripe | (stripe-1)) + 1;
+		mddev->queue->limits.discard_alignment = stripe;
+		mddev->queue->limits.discard_granularity = stripe;
+		/*
+		 * unaligned part of discard request will be ignored, so can't
+		 * guarantee discard_zerors_data
+		 */
+		mddev->queue->limits.discard_zeroes_data = 0;
 
-	blk_queue_merge_bvec(mddev->queue, raid5_mergeable_bvec);
+		blk_queue_max_write_same_sectors(mddev->queue, 0);
+
+		rdev_for_each(rdev, mddev) {
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->data_offset << 9);
+			disk_stack_limits(mddev->gendisk, rdev->bdev,
+					  rdev->new_data_offset << 9);
+			/*
+			 * discard_zeroes_data is required, otherwise data
+			 * could be lost. Consider a scenario: discard a stripe
+			 * (the stripe could be inconsistent if
+			 * discard_zeroes_data is 0); write one disk of the
+			 * stripe (the stripe could be inconsistent again
+			 * depending on which disks are used to calculate
+			 * parity); the disk is broken; The stripe data of this
+			 * disk is lost.
+			 */
+			if (!blk_queue_discard(bdev_get_queue(rdev->bdev)) ||
+			    !bdev_get_queue(rdev->bdev)->
+						limits.discard_zeroes_data)
+				discard_supported = false;
+		}
+
+		if (discard_supported &&
+		   mddev->queue->limits.max_discard_sectors >= stripe &&
+		   mddev->queue->limits.discard_granularity >= stripe)
+			queue_flag_set_unlocked(QUEUE_FLAG_DISCARD,
+						mddev->queue);
+		else
+			queue_flag_clear_unlocked(QUEUE_FLAG_DISCARD,
+						mddev->queue);
+	}
 
 	return 0;
 abort:
-	if (conf) {
-		print_raid5_conf(conf);
-		safe_put_page(conf->spare_page);
-		kfree(conf->disks);
-		kfree(conf->stripe_hashtbl);
-		kfree(conf);
-	}
+	md_unregister_thread(&mddev->thread);
+	print_raid5_conf(conf);
+	free_conf(conf);
 	mddev->private = NULL;
-	printk(KERN_ALERT "raid5: failed to run raid set %s\n", mdname(mddev));
+	printk(KERN_ALERT "md/raid:%s: failed to run raid set.\n", mdname(mddev));
 	return -EIO;
 }
 
-
-
-static int stop(mddev_t *mddev)
+static int stop(struct mddev *mddev)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	struct r5conf *conf = mddev->private;
 
-	md_unregister_thread(mddev->thread);
-	mddev->thread = NULL;
-	shrink_stripes(conf);
-	kfree(conf->stripe_hashtbl);
-	mddev->queue->backing_dev_info.congested_fn = NULL;
-	blk_sync_queue(mddev->queue); /* the unplug fn references 'conf'*/
-	sysfs_remove_group(&mddev->kobj, &raid5_attrs_group);
-	kfree(conf->disks);
-	kfree(conf);
+	md_unregister_thread(&mddev->thread);
+	if (mddev->queue)
+		mddev->queue->backing_dev_info.congested_fn = NULL;
+	free_conf(conf);
 	mddev->private = NULL;
+	mddev->to_remove = &raid5_attrs_group;
 	return 0;
 }
 
-#ifdef DEBUG
-static void print_sh (struct seq_file *seq, struct stripe_head *sh)
-{
-	int i;
-
-	seq_printf(seq, "sh %llu, pd_idx %d, state %ld.\n",
-		   (unsigned long long)sh->sector, sh->pd_idx, sh->state);
-	seq_printf(seq, "sh %llu,  count %d.\n",
-		   (unsigned long long)sh->sector, atomic_read(&sh->count));
-	seq_printf(seq, "sh %llu, ", (unsigned long long)sh->sector);
-	for (i = 0; i < sh->disks; i++) {
-		seq_printf(seq, "(cache%d: %p %ld) ",
-			   i, sh->dev[i].page, sh->dev[i].flags);
-	}
-	seq_printf(seq, "\n");
-}
-
-static void printall (struct seq_file *seq, raid5_conf_t *conf)
-{
-	struct stripe_head *sh;
-	struct hlist_node *hn;
-	int i;
-
-	spin_lock_irq(&conf->device_lock);
-	for (i = 0; i < NR_HASH; i++) {
-		hlist_for_each_entry(sh, hn, &conf->stripe_hashtbl[i], hash) {
-			if (sh->raid_conf != conf)
-				continue;
-			print_sh(seq, sh);
-		}
-	}
-	spin_unlock_irq(&conf->device_lock);
-}
-#endif
-
-static void status (struct seq_file *seq, mddev_t *mddev)
+static void status(struct seq_file *seq, struct mddev *mddev)
 {
-	raid5_conf_t *conf = (raid5_conf_t *) mddev->private;
+	struct r5conf *conf = mddev->private;
 	int i;
 
-	seq_printf (seq, " level %d, %dk chunk, algorithm %d", mddev->level, mddev->chunk_size >> 10, mddev->layout);
+	seq_printf(seq, " level %d, %dk chunk, algorithm %d", mddev->level,
+		mddev->chunk_sectors / 2, mddev->layout);
 	seq_printf (seq, " [%d/%d] [", conf->raid_disks, conf->raid_disks - mddev->degraded);
 	for (i = 0; i < conf->raid_disks; i++)
 		seq_printf (seq, "%s",
 			       conf->disks[i].rdev &&
 			       test_bit(In_sync, &conf->disks[i].rdev->flags) ? "U" : "_");
 	seq_printf (seq, "]");
-#ifdef DEBUG
-	seq_printf (seq, "\n");
-	printall(seq, conf);
-#endif
 }
 
-static void print_raid5_conf (raid5_conf_t *conf)
+static void print_raid5_conf (struct r5conf *conf)
 {
 	int i;
 	struct disk_info *tmp;
 
-	printk("RAID5 conf printout:\n");
+	printk(KERN_DEBUG "RAID conf printout:\n");
 	if (!conf) {
 		printk("(conf==NULL)\n");
 		return;
 	}
-	printk(" --- rd:%d wd:%d\n", conf->raid_disks,
-		 conf->raid_disks - conf->mddev->degraded);
+	printk(KERN_DEBUG " --- level:%d rd:%d wd:%d\n", conf->level,
+	       conf->raid_disks,
+	       conf->raid_disks - conf->mddev->degraded);
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		char b[BDEVNAME_SIZE];
 		tmp = conf->disks + i;
 		if (tmp->rdev)
-		printk(" disk %d, o:%d, dev:%s\n",
-			i, !test_bit(Faulty, &tmp->rdev->flags),
-			bdevname(tmp->rdev->bdev,b));
+			printk(KERN_DEBUG " disk %d, o:%d, dev:%s\n",
+			       i, !test_bit(Faulty, &tmp->rdev->flags),
+			       bdevname(tmp->rdev->bdev, b));
 	}
 }
 
-static int raid5_spare_active(mddev_t *mddev)
+static int raid5_spare_active(struct mddev *mddev)
 {
 	int i;
-	raid5_conf_t *conf = mddev->private;
+	struct r5conf *conf = mddev->private;
 	struct disk_info *tmp;
+	int count = 0;
+	unsigned long flags;
 
 	for (i = 0; i < conf->raid_disks; i++) {
 		tmp = conf->disks + i;
-		if (tmp->rdev
+		if (tmp->replacement
+		    && tmp->replacement->recovery_offset == MaxSector
+		    && !test_bit(Faulty, &tmp->replacement->flags)
+		    && !test_and_set_bit(In_sync, &tmp->replacement->flags)) {
+			/* Replacement has just become active. */
+			if (!tmp->rdev
+			    || !test_and_clear_bit(In_sync, &tmp->rdev->flags))
+				count++;
+			if (tmp->rdev) {
+				/* Replaced device not technically faulty,
+				 * but we need to be sure it gets removed
+				 * and never re-added.
+				 */
+				set_bit(Faulty, &tmp->rdev->flags);
+				sysfs_notify_dirent_safe(
+					tmp->rdev->sysfs_state);
+			}
+			sysfs_notify_dirent_safe(tmp->replacement->sysfs_state);
+		} else if (tmp->rdev
+		    && tmp->rdev->recovery_offset == MaxSector
 		    && !test_bit(Faulty, &tmp->rdev->flags)
 		    && !test_and_set_bit(In_sync, &tmp->rdev->flags)) {
-			unsigned long flags;
-			spin_lock_irqsave(&conf->device_lock, flags);
-			mddev->degraded--;
-			spin_unlock_irqrestore(&conf->device_lock, flags);
+			count++;
+			sysfs_notify_dirent_safe(tmp->rdev->sysfs_state);
 		}
 	}
+	spin_lock_irqsave(&conf->device_lock, flags);
+	mddev->degraded = calc_degraded(conf);
+	spin_unlock_irqrestore(&conf->device_lock, flags);
 	print_raid5_conf(conf);
-	return 0;
+	return count;
 }
 
-static int raid5_remove_disk(mddev_t *mddev, int number)
+static int raid5_remove_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	raid5_conf_t *conf = mddev->private;
+	struct r5conf *conf = mddev->private;
 	int err = 0;
-	mdk_rdev_t *rdev;
+	int number = rdev->raid_disk;
+	struct md_rdev **rdevp;
 	struct disk_info *p = conf->disks + number;
 
 	print_raid5_conf(conf);
-	rdev = p->rdev;
-	if (rdev) {
-		if (test_bit(In_sync, &rdev->flags) ||
-		    atomic_read(&rdev->nr_pending)) {
-			err = -EBUSY;
-			goto abort;
-		}
-		/* Only remove non-faulty devices if recovery
-		 * isn't possible.
-		 */
-		if (!test_bit(Faulty, &rdev->flags) &&
-		    mddev->degraded <= conf->max_degraded) {
-			err = -EBUSY;
-			goto abort;
-		}
-		p->rdev = NULL;
-		synchronize_rcu();
-		if (atomic_read(&rdev->nr_pending)) {
-			/* lost the race, try later */
-			err = -EBUSY;
-			p->rdev = rdev;
-		}
+	if (rdev == p->rdev)
+		rdevp = &p->rdev;
+	else if (rdev == p->replacement)
+		rdevp = &p->replacement;
+	else
+		return 0;
+
+	if (number >= conf->raid_disks &&
+	    conf->reshape_progress == MaxSector)
+		clear_bit(In_sync, &rdev->flags);
+
+	if (test_bit(In_sync, &rdev->flags) ||
+	    atomic_read(&rdev->nr_pending)) {
+		err = -EBUSY;
+		goto abort;
+	}
+	/* Only remove non-faulty devices if recovery
+	 * isn't possible.
+	 */
+	if (!test_bit(Faulty, &rdev->flags) &&
+	    mddev->recovery_disabled != conf->recovery_disabled &&
+	    !has_failed(conf) &&
+	    (!p->replacement || p->replacement == rdev) &&
+	    number < conf->raid_disks) {
+		err = -EBUSY;
+		goto abort;
 	}
+	*rdevp = NULL;
+	synchronize_rcu();
+	if (atomic_read(&rdev->nr_pending)) {
+		/* lost the race, try later */
+		err = -EBUSY;
+		*rdevp = rdev;
+	} else if (p->replacement) {
+		/* We must have just cleared 'rdev' */
+		p->rdev = p->replacement;
+		clear_bit(Replacement, &p->replacement->flags);
+		smp_mb(); /* Make sure other CPUs may see both as identical
+			   * but will never see neither - if they are careful
+			   */
+		p->replacement = NULL;
+		clear_bit(WantReplacement, &rdev->flags);
+	} else
+		/* We might have just removed the Replacement as faulty-
+		 * clear the bit just in case
+		 */
+		clear_bit(WantReplacement, &rdev->flags);
 abort:
 
 	print_raid5_conf(conf);
 	return err;
 }
 
-static int raid5_add_disk(mddev_t *mddev, mdk_rdev_t *rdev)
+static int raid5_add_disk(struct mddev *mddev, struct md_rdev *rdev)
 {
-	raid5_conf_t *conf = mddev->private;
+	struct r5conf *conf = mddev->private;
 	int err = -EEXIST;
 	int disk;
 	struct disk_info *p;
 	int first = 0;
 	int last = conf->raid_disks - 1;
 
-	if (mddev->degraded > conf->max_degraded)
+	if (mddev->recovery_disabled == conf->recovery_disabled)
+		return -EBUSY;
+
+	if (rdev->saved_raid_disk < 0 && has_failed(conf))
 		/* no point adding a device */
 		return -EINVAL;
 
@@ -4406,24 +6436,39 @@ static int raid5_add_disk(mddev_t *mddev, mdk_rdev_t *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++)
-		if ((p=conf->disks + disk)->rdev == NULL) {
+		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);
 			rdev->raid_disk = disk;
 			err = 0;
 			if (rdev->saved_raid_disk != disk)
 				conf->fullsync = 1;
 			rcu_assign_pointer(p->rdev, rdev);
+			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);
+			set_bit(Replacement, &rdev->flags);
+			rdev->raid_disk = disk;
+			err = 0;
+			conf->fullsync = 1;
+			rcu_assign_pointer(p->replacement, rdev);
 			break;
 		}
+	}
+out:
 	print_raid5_conf(conf);
 	return err;
 }
 
-static int raid5_resize(mddev_t *mddev, sector_t sectors)
+static int raid5_resize(struct mddev *mddev, sector_t sectors)
 {
 	/* no resync is happening, and there is enough space
 	 * on all devices, so we can resize.
@@ -4432,37 +6477,32 @@ static int raid5_resize(mddev_t *mddev, sector_t sectors)
 	 * any io in the removed space completes, but it hardly seems
 	 * worth it.
 	 */
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-
-	sectors &= ~((sector_t)mddev->chunk_size/512 - 1);
-	mddev->array_sectors = sectors * (mddev->raid_disks
-					  - conf->max_degraded);
+	sector_t newsize;
+	sectors &= ~((sector_t)mddev->chunk_sectors - 1);
+	newsize = raid5_size(mddev, sectors, mddev->raid_disks);
+	if (mddev->external_size &&
+	    mddev->array_sectors > newsize)
+		return -EINVAL;
+	if (mddev->bitmap) {
+		int ret = bitmap_resize(mddev->bitmap, sectors, 0, 0);
+		if (ret)
+			return ret;
+	}
+	md_set_array_sectors(mddev, newsize);
 	set_capacity(mddev->gendisk, mddev->array_sectors);
-	mddev->changed = 1;
-	if (sectors/2  > mddev->size && mddev->recovery_cp == MaxSector) {
-		mddev->recovery_cp = mddev->size << 1;
+	revalidate_disk(mddev->gendisk);
+	if (sectors > mddev->dev_sectors &&
+	    mddev->recovery_cp > mddev->dev_sectors) {
+		mddev->recovery_cp = mddev->dev_sectors;
 		set_bit(MD_RECOVERY_NEEDED, &mddev->recovery);
 	}
-	mddev->size = sectors /2;
+	mddev->dev_sectors = sectors;
 	mddev->resync_max_sectors = sectors;
 	return 0;
 }
 
-#ifdef CONFIG_MD_RAID5_RESHAPE
-static int raid5_check_reshape(mddev_t *mddev)
+static int check_stripe_cache(struct mddev *mddev)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	int err;
-
-	if (mddev->delta_disks < 0 ||
-	    mddev->new_level != mddev->level)
-		return -EINVAL; /* Cannot shrink array or change level yet */
-	if (mddev->delta_disks == 0)
-		return 0; /* nothing to do */
-	if (mddev->bitmap)
-		/* Cannot grow a bitmap yet */
-		return -EBUSY;
-
 	/* Can only proceed if there are plenty of stripe_heads.
 	 * We need a minimum of one full stripe,, and for sensible progress
 	 * it is best to have about 4 times that.
@@ -4471,39 +6511,71 @@ static int raid5_check_reshape(mddev_t *mddev)
 	 * If the chunk size is greater, user-space should request more
 	 * stripe_heads first.
 	 */
-	if ((mddev->chunk_size / STRIPE_SIZE) * 4 > conf->max_nr_stripes ||
-	    (mddev->new_chunk / STRIPE_SIZE) * 4 > conf->max_nr_stripes) {
-		printk(KERN_WARNING "raid5: reshape: not enough stripes.  Needed %lu\n",
-		       (mddev->chunk_size / STRIPE_SIZE)*4);
-		return -ENOSPC;
+	struct r5conf *conf = mddev->private;
+	if (((mddev->chunk_sectors << 9) / STRIPE_SIZE) * 4
+	    > conf->max_nr_stripes ||
+	    ((mddev->new_chunk_sectors << 9) / STRIPE_SIZE) * 4
+	    > conf->max_nr_stripes) {
+		printk(KERN_WARNING "md/raid:%s: reshape: not enough stripes.  Needed %lu\n",
+		       mdname(mddev),
+		       ((max(mddev->chunk_sectors, mddev->new_chunk_sectors) << 9)
+			/ STRIPE_SIZE)*4);
+		return 0;
 	}
+	return 1;
+}
 
-	err = resize_stripes(conf, conf->raid_disks + mddev->delta_disks);
-	if (err)
-		return err;
+static int check_reshape(struct mddev *mddev)
+{
+	struct r5conf *conf = mddev->private;
 
-	if (mddev->degraded > conf->max_degraded)
+	if (mddev->delta_disks == 0 &&
+	    mddev->new_layout == mddev->layout &&
+	    mddev->new_chunk_sectors == mddev->chunk_sectors)
+		return 0; /* nothing to do */
+	if (has_failed(conf))
 		return -EINVAL;
-	/* looks like we might be able to manage this */
-	return 0;
+	if (mddev->delta_disks < 0 && mddev->reshape_position == MaxSector) {
+		/* We might be able to shrink, but the devices must
+		 * be made bigger first.
+		 * For raid6, 4 is the minimum size.
+		 * Otherwise 2 is the minimum
+		 */
+		int min = 2;
+		if (mddev->level == 6)
+			min = 4;
+		if (mddev->raid_disks + mddev->delta_disks < min)
+			return -EINVAL;
+	}
+
+	if (!check_stripe_cache(mddev))
+		return -ENOSPC;
+
+	return resize_stripes(conf, (conf->previous_raid_disks
+				     + mddev->delta_disks));
 }
 
-static int raid5_start_reshape(mddev_t *mddev)
+static int raid5_start_reshape(struct mddev *mddev)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
-	mdk_rdev_t *rdev;
-	struct list_head *rtmp;
+	struct r5conf *conf = mddev->private;
+	struct md_rdev *rdev;
 	int spares = 0;
-	int added_devices = 0;
 	unsigned long flags;
 
 	if (test_bit(MD_RECOVERY_RUNNING, &mddev->recovery))
 		return -EBUSY;
 
-	rdev_for_each(rdev, rtmp, mddev)
-		if (rdev->raid_disk < 0 &&
-		    !test_bit(Faulty, &rdev->flags))
+	if (!check_stripe_cache(mddev))
+		return -ENOSPC;
+
+	if (has_failed(conf))
+		return -EINVAL;
+
+	rdev_for_each(rdev, mddev) {
+		if (!test_bit(In_sync, &rdev->flags)
+		    && !test_bit(Faulty, &rdev->flags))
 			spares++;
+	}
 
 	if (spares - mddev->degraded < mddev->delta_disks - conf->max_degraded)
 		/* Not enough devices even to make a degraded array
@@ -4511,41 +6583,83 @@ static int raid5_start_reshape(mddev_t *mddev)
 		 */
 		return -EINVAL;
 
+	/* Refuse to reduce size of the array.  Any reductions in
+	 * array size must be through explicit setting of array_size
+	 * attribute.
+	 */
+	if (raid5_size(mddev, 0, conf->raid_disks + mddev->delta_disks)
+	    < mddev->array_sectors) {
+		printk(KERN_ERR "md/raid:%s: array size must be reduced "
+		       "before number of disks\n", mdname(mddev));
+		return -EINVAL;
+	}
+
 	atomic_set(&conf->reshape_stripes, 0);
 	spin_lock_irq(&conf->device_lock);
+	write_seqcount_begin(&conf->gen_lock);
 	conf->previous_raid_disks = conf->raid_disks;
 	conf->raid_disks += mddev->delta_disks;
-	conf->expand_progress = 0;
-	conf->expand_lo = 0;
+	conf->prev_chunk_sectors = conf->chunk_sectors;
+	conf->chunk_sectors = mddev->new_chunk_sectors;
+	conf->prev_algo = conf->algorithm;
+	conf->algorithm = mddev->new_layout;
+	conf->generation++;
+	/* Code that selects data_offset needs to see the generation update
+	 * if reshape_progress has been set - so a memory barrier needed.
+	 */
+	smp_mb();
+	if (mddev->reshape_backwards)
+		conf->reshape_progress = raid5_size(mddev, 0, 0);
+	else
+		conf->reshape_progress = 0;
+	conf->reshape_safe = conf->reshape_progress;
+	write_seqcount_end(&conf->gen_lock);
 	spin_unlock_irq(&conf->device_lock);
 
+	/* Now make sure any requests that proceeded on the assumption
+	 * the reshape wasn't running - like Discard or Read - have
+	 * completed.
+	 */
+	mddev_suspend(mddev);
+	mddev_resume(mddev);
+
 	/* Add some new drives, as many as will fit.
 	 * We know there are enough to make the newly sized array work.
+	 * Don't add devices if we are reducing the number of
+	 * devices in the array.  This is because it is not possible
+	 * to correctly record the "partially reconstructed" state of
+	 * such devices during the reshape and confusion could result.
 	 */
-	rdev_for_each(rdev, rtmp, mddev)
-		if (rdev->raid_disk < 0 &&
-		    !test_bit(Faulty, &rdev->flags)) {
-			if (raid5_add_disk(mddev, rdev) == 0) {
-				char nm[20];
+	if (mddev->delta_disks >= 0) {
+		rdev_for_each(rdev, mddev)
+			if (rdev->raid_disk < 0 &&
+			    !test_bit(Faulty, &rdev->flags)) {
+				if (raid5_add_disk(mddev, rdev) == 0) {
+					if (rdev->raid_disk
+					    >= conf->previous_raid_disks)
+						set_bit(In_sync, &rdev->flags);
+					else
+						rdev->recovery_offset = 0;
+
+					if (sysfs_link_rdev(mddev, rdev))
+						/* Failure here is OK */;
+				}
+			} else if (rdev->raid_disk >= conf->previous_raid_disks
+				   && !test_bit(Faulty, &rdev->flags)) {
+				/* This is a spare that was manually added */
 				set_bit(In_sync, &rdev->flags);
-				added_devices++;
-				rdev->recovery_offset = 0;
-				sprintf(nm, "rd%d", rdev->raid_disk);
-				if (sysfs_create_link(&mddev->kobj,
-						      &rdev->kobj, nm))
-					printk(KERN_WARNING
-					       "raid5: failed to create "
-					       " link %s for %s\n",
-					       nm, mdname(mddev));
-			} else
-				break;
-		}
+			}
 
-	spin_lock_irqsave(&conf->device_lock, flags);
-	mddev->degraded = (conf->raid_disks - conf->previous_raid_disks) - added_devices;
-	spin_unlock_irqrestore(&conf->device_lock, flags);
+		/* When a reshape changes the number of devices,
+		 * ->degraded is measured against the larger of the
+		 * pre and post number of devices.
+		 */
+		spin_lock_irqsave(&conf->device_lock, flags);
+		mddev->degraded = calc_degraded(conf);
+		spin_unlock_irqrestore(&conf->device_lock, flags);
+	}
 	mddev->raid_disks = conf->raid_disks;
-	mddev->reshape_position = 0;
+	mddev->reshape_position = conf->reshape_progress;
 	set_bit(MD_CHANGE_DEVS, &mddev->flags);
 
 	clear_bit(MD_RECOVERY_SYNC, &mddev->recovery);
@@ -4553,60 +6667,102 @@ static int raid5_start_reshape(mddev_t *mddev)
 	set_bit(MD_RECOVERY_RESHAPE, &mddev->recovery);
 	set_bit(MD_RECOVERY_RUNNING, &mddev->recovery);
 	mddev->sync_thread = md_register_thread(md_do_sync, mddev,
-						"%s_reshape");
+						"reshape");
 	if (!mddev->sync_thread) {
 		mddev->recovery = 0;
 		spin_lock_irq(&conf->device_lock);
+		write_seqcount_begin(&conf->gen_lock);
 		mddev->raid_disks = conf->raid_disks = conf->previous_raid_disks;
-		conf->expand_progress = MaxSector;
+		mddev->new_chunk_sectors =
+			conf->chunk_sectors = conf->prev_chunk_sectors;
+		mddev->new_layout = conf->algorithm = conf->prev_algo;
+		rdev_for_each(rdev, mddev)
+			rdev->new_data_offset = rdev->data_offset;
+		smp_wmb();
+		conf->generation --;
+		conf->reshape_progress = MaxSector;
+		mddev->reshape_position = MaxSector;
+		write_seqcount_end(&conf->gen_lock);
 		spin_unlock_irq(&conf->device_lock);
 		return -EAGAIN;
 	}
+	conf->reshape_checkpoint = jiffies;
 	md_wakeup_thread(mddev->sync_thread);
 	md_new_event(mddev);
 	return 0;
 }
-#endif
 
-static void end_reshape(raid5_conf_t *conf)
+/* This is called from the reshape thread and should make any
+ * changes needed in 'conf'
+ */
+static void end_reshape(struct r5conf *conf)
 {
-	struct block_device *bdev;
 
 	if (!test_bit(MD_RECOVERY_INTR, &conf->mddev->recovery)) {
-		conf->mddev->array_sectors = 2 * conf->mddev->size *
-			(conf->raid_disks - conf->max_degraded);
-		set_capacity(conf->mddev->gendisk, conf->mddev->array_sectors);
-		conf->mddev->changed = 1;
-
-		bdev = bdget_disk(conf->mddev->gendisk, 0);
-		if (bdev) {
-			mutex_lock(&bdev->bd_inode->i_mutex);
-			i_size_write(bdev->bd_inode,
-				     (loff_t)conf->mddev->array_sectors << 9);
-			mutex_unlock(&bdev->bd_inode->i_mutex);
-			bdput(bdev);
-		}
+		struct md_rdev *rdev;
+
 		spin_lock_irq(&conf->device_lock);
-		conf->expand_progress = MaxSector;
+		conf->previous_raid_disks = conf->raid_disks;
+		rdev_for_each(rdev, conf->mddev)
+			rdev->data_offset = rdev->new_data_offset;
+		smp_wmb();
+		conf->reshape_progress = MaxSector;
 		spin_unlock_irq(&conf->device_lock);
-		conf->mddev->reshape_position = MaxSector;
+		wake_up(&conf->wait_for_overlap);
 
 		/* read-ahead size must cover two whole stripes, which is
 		 * 2 * (datadisks) * chunksize where 'n' is the number of raid devices
 		 */
-		{
-			int data_disks = conf->previous_raid_disks - conf->max_degraded;
-			int stripe = data_disks *
-				(conf->mddev->chunk_size / PAGE_SIZE);
+		if (conf->mddev->queue) {
+			int data_disks = conf->raid_disks - conf->max_degraded;
+			int stripe = data_disks * ((conf->chunk_sectors << 9)
+						   / PAGE_SIZE);
 			if (conf->mddev->queue->backing_dev_info.ra_pages < 2 * stripe)
 				conf->mddev->queue->backing_dev_info.ra_pages = 2 * stripe;
 		}
 	}
 }
 
-static void raid5_quiesce(mddev_t *mddev, int state)
+/* This is called from the raid5d thread with mddev_lock held.
+ * It makes config changes to the device.
+ */
+static void raid5_finish_reshape(struct mddev *mddev)
+{
+	struct r5conf *conf = mddev->private;
+
+	if (!test_bit(MD_RECOVERY_INTR, &mddev->recovery)) {
+
+		if (mddev->delta_disks > 0) {
+			md_set_array_sectors(mddev, raid5_size(mddev, 0, 0));
+			set_capacity(mddev->gendisk, mddev->array_sectors);
+			revalidate_disk(mddev->gendisk);
+		} else {
+			int d;
+			spin_lock_irq(&conf->device_lock);
+			mddev->degraded = calc_degraded(conf);
+			spin_unlock_irq(&conf->device_lock);
+			for (d = conf->raid_disks ;
+			     d < conf->raid_disks - mddev->delta_disks;
+			     d++) {
+				struct md_rdev *rdev = conf->disks[d].rdev;
+				if (rdev)
+					clear_bit(In_sync, &rdev->flags);
+				rdev = conf->disks[d].replacement;
+				if (rdev)
+					clear_bit(In_sync, &rdev->flags);
+			}
+		}
+		mddev->layout = conf->algorithm;
+		mddev->chunk_sectors = conf->chunk_sectors;
+		mddev->reshape_position = MaxSector;
+		mddev->delta_disks = 0;
+		mddev->reshape_backwards = 0;
+	}
+}
+
+static void raid5_quiesce(struct mddev *mddev, int state)
 {
-	raid5_conf_t *conf = mddev_to_conf(mddev);
+	struct r5conf *conf = mddev->private;
 
 	switch(state) {
 	case 2: /* resume for a suspend */
@@ -4614,26 +6770,271 @@ static void raid5_quiesce(mddev_t *mddev, int state)
 		break;
 
 	case 1: /* stop all writes */
-		spin_lock_irq(&conf->device_lock);
-		conf->quiesce = 1;
-		wait_event_lock_irq(conf->wait_for_stripe,
+		lock_all_device_hash_locks_irq(conf);
+		/* '2' tells resync/reshape to pause so that all
+		 * active stripes can drain
+		 */
+		conf->quiesce = 2;
+		wait_event_cmd(conf->wait_for_stripe,
 				    atomic_read(&conf->active_stripes) == 0 &&
 				    atomic_read(&conf->active_aligned_reads) == 0,
-				    conf->device_lock, /* nothing */);
-		spin_unlock_irq(&conf->device_lock);
+				    unlock_all_device_hash_locks_irq(conf),
+				    lock_all_device_hash_locks_irq(conf));
+		conf->quiesce = 1;
+		unlock_all_device_hash_locks_irq(conf);
+		/* allow reshape to continue */
+		wake_up(&conf->wait_for_overlap);
 		break;
 
 	case 0: /* re-enable writes */
-		spin_lock_irq(&conf->device_lock);
+		lock_all_device_hash_locks_irq(conf);
 		conf->quiesce = 0;
 		wake_up(&conf->wait_for_stripe);
 		wake_up(&conf->wait_for_overlap);
-		spin_unlock_irq(&conf->device_lock);
+		unlock_all_device_hash_locks_irq(conf);
 		break;
 	}
 }
 
-static struct mdk_personality raid6_personality =
+
+static void *raid45_takeover_raid0(struct mddev *mddev, int level)
+{
+	struct r0conf *raid0_conf = mddev->private;
+	sector_t sectors;
+
+	/* for raid0 takeover only one zone is supported */
+	if (raid0_conf->nr_strip_zones > 1) {
+		printk(KERN_ERR "md/raid:%s: cannot takeover raid0 with more than one zone.\n",
+		       mdname(mddev));
+		return ERR_PTR(-EINVAL);
+	}
+
+	sectors = raid0_conf->strip_zone[0].zone_end;
+	sector_div(sectors, raid0_conf->strip_zone[0].nb_dev);
+	mddev->dev_sectors = sectors;
+	mddev->new_level = level;
+	mddev->new_layout = ALGORITHM_PARITY_N;
+	mddev->new_chunk_sectors = mddev->chunk_sectors;
+	mddev->raid_disks += 1;
+	mddev->delta_disks = 1;
+	/* make sure it will be not marked as dirty */
+	mddev->recovery_cp = MaxSector;
+
+	return setup_conf(mddev);
+}
+
+
+static void *raid5_takeover_raid1(struct mddev *mddev)
+{
+	int chunksect;
+
+	if (mddev->raid_disks != 2 ||
+	    mddev->degraded > 1)
+		return ERR_PTR(-EINVAL);
+
+	/* Should check if there are write-behind devices? */
+
+	chunksect = 64*2; /* 64K by default */
+
+	/* The array must be an exact multiple of chunksize */
+	while (chunksect && (mddev->array_sectors & (chunksect-1)))
+		chunksect >>= 1;
+
+	if ((chunksect<<9) < STRIPE_SIZE)
+		/* array size does not allow a suitable chunk size */
+		return ERR_PTR(-EINVAL);
+
+	mddev->new_level = 5;
+	mddev->new_layout = ALGORITHM_LEFT_SYMMETRIC;
+	mddev->new_chunk_sectors = chunksect;
+
+	return setup_conf(mddev);
+}
+
+static void *raid5_takeover_raid6(struct mddev *mddev)
+{
+	int new_layout;
+
+	switch (mddev->layout) {
+	case ALGORITHM_LEFT_ASYMMETRIC_6:
+		new_layout = ALGORITHM_LEFT_ASYMMETRIC;
+		break;
+	case ALGORITHM_RIGHT_ASYMMETRIC_6:
+		new_layout = ALGORITHM_RIGHT_ASYMMETRIC;
+		break;
+	case ALGORITHM_LEFT_SYMMETRIC_6:
+		new_layout = ALGORITHM_LEFT_SYMMETRIC;
+		break;
+	case ALGORITHM_RIGHT_SYMMETRIC_6:
+		new_layout = ALGORITHM_RIGHT_SYMMETRIC;
+		break;
+	case ALGORITHM_PARITY_0_6:
+		new_layout = ALGORITHM_PARITY_0;
+		break;
+	case ALGORITHM_PARITY_N:
+		new_layout = ALGORITHM_PARITY_N;
+		break;
+	default:
+		return ERR_PTR(-EINVAL);
+	}
+	mddev->new_level = 5;
+	mddev->new_layout = new_layout;
+	mddev->delta_disks = -1;
+	mddev->raid_disks -= 1;
+	return setup_conf(mddev);
+}
+
+
+static int raid5_check_reshape(struct mddev *mddev)
+{
+	/* For a 2-drive array, the layout and chunk size can be changed
+	 * immediately as not restriping is needed.
+	 * For larger arrays we record the new value - after validation
+	 * to be used by a reshape pass.
+	 */
+	struct r5conf *conf = mddev->private;
+	int new_chunk = mddev->new_chunk_sectors;
+
+	if (mddev->new_layout >= 0 && !algorithm_valid_raid5(mddev->new_layout))
+		return -EINVAL;
+	if (new_chunk > 0) {
+		if (!is_power_of_2(new_chunk))
+			return -EINVAL;
+		if (new_chunk < (PAGE_SIZE>>9))
+			return -EINVAL;
+		if (mddev->array_sectors & (new_chunk-1))
+			/* not factor of array size */
+			return -EINVAL;
+	}
+
+	/* They look valid */
+
+	if (mddev->raid_disks == 2) {
+		/* can make the change immediately */
+		if (mddev->new_layout >= 0) {
+			conf->algorithm = mddev->new_layout;
+			mddev->layout = mddev->new_layout;
+		}
+		if (new_chunk > 0) {
+			conf->chunk_sectors = new_chunk ;
+			mddev->chunk_sectors = new_chunk;
+		}
+		set_bit(MD_CHANGE_DEVS, &mddev->flags);
+		md_wakeup_thread(mddev->thread);
+	}
+	return check_reshape(mddev);
+}
+
+static int raid6_check_reshape(struct mddev *mddev)
+{
+	int new_chunk = mddev->new_chunk_sectors;
+
+	if (mddev->new_layout >= 0 && !algorithm_valid_raid6(mddev->new_layout))
+		return -EINVAL;
+	if (new_chunk > 0) {
+		if (!is_power_of_2(new_chunk))
+			return -EINVAL;
+		if (new_chunk < (PAGE_SIZE >> 9))
+			return -EINVAL;
+		if (mddev->array_sectors & (new_chunk-1))
+			/* not factor of array size */
+			return -EINVAL;
+	}
+
+	/* They look valid */
+	return check_reshape(mddev);
+}
+
+static void *raid5_takeover(struct mddev *mddev)
+{
+	/* raid5 can take over:
+	 *  raid0 - if there is only one strip zone - make it a raid4 layout
+	 *  raid1 - if there are two drives.  We need to know the chunk size
+	 *  raid4 - trivial - just use a raid4 layout.
+	 *  raid6 - Providing it is a *_6 layout
+	 */
+	if (mddev->level == 0)
+		return raid45_takeover_raid0(mddev, 5);
+	if (mddev->level == 1)
+		return raid5_takeover_raid1(mddev);
+	if (mddev->level == 4) {
+		mddev->new_layout = ALGORITHM_PARITY_N;
+		mddev->new_level = 5;
+		return setup_conf(mddev);
+	}
+	if (mddev->level == 6)
+		return raid5_takeover_raid6(mddev);
+
+	return ERR_PTR(-EINVAL);
+}
+
+static void *raid4_takeover(struct mddev *mddev)
+{
+	/* raid4 can take over:
+	 *  raid0 - if there is only one strip zone
+	 *  raid5 - if layout is right
+	 */
+	if (mddev->level == 0)
+		return raid45_takeover_raid0(mddev, 4);
+	if (mddev->level == 5 &&
+	    mddev->layout == ALGORITHM_PARITY_N) {
+		mddev->new_layout = 0;
+		mddev->new_level = 4;
+		return setup_conf(mddev);
+	}
+	return ERR_PTR(-EINVAL);
+}
+
+static struct md_personality raid5_personality;
+
+static void *raid6_takeover(struct mddev *mddev)
+{
+	/* Currently can only take over a raid5.  We map the
+	 * personality to an equivalent raid6 personality
+	 * with the Q block at the end.
+	 */
+	int new_layout;
+
+	if (mddev->pers != &raid5_personality)
+		return ERR_PTR(-EINVAL);
+	if (mddev->degraded > 1)
+		return ERR_PTR(-EINVAL);
+	if (mddev->raid_disks > 253)
+		return ERR_PTR(-EINVAL);
+	if (mddev->raid_disks < 3)
+		return ERR_PTR(-EINVAL);
+
+	switch (mddev->layout) {
+	case ALGORITHM_LEFT_ASYMMETRIC:
+		new_layout = ALGORITHM_LEFT_ASYMMETRIC_6;
+		break;
+	case ALGORITHM_RIGHT_ASYMMETRIC:
+		new_layout = ALGORITHM_RIGHT_ASYMMETRIC_6;
+		break;
+	case ALGORITHM_LEFT_SYMMETRIC:
+		new_layout = ALGORITHM_LEFT_SYMMETRIC_6;
+		break;
+	case ALGORITHM_RIGHT_SYMMETRIC:
+		new_layout = ALGORITHM_RIGHT_SYMMETRIC_6;
+		break;
+	case ALGORITHM_PARITY_0:
+		new_layout = ALGORITHM_PARITY_0_6;
+		break;
+	case ALGORITHM_PARITY_N:
+		new_layout = ALGORITHM_PARITY_N;
+		break;
+	default:
+		return ERR_PTR(-EINVAL);
+	}
+	mddev->new_level = 6;
+	mddev->new_layout = new_layout;
+	mddev->delta_disks = 1;
+	mddev->raid_disks += 1;
+	return setup_conf(mddev);
+}
+
+
+static struct md_personality raid6_personality =
 {
 	.name		= "raid6",
 	.level		= 6,
@@ -4648,13 +7049,14 @@ static struct mdk_personality raid6_personality =
 	.spare_active	= raid5_spare_active,
 	.sync_request	= sync_request,
 	.resize		= raid5_resize,
-#ifdef CONFIG_MD_RAID5_RESHAPE
-	.check_reshape	= raid5_check_reshape,
+	.size		= raid5_size,
+	.check_reshape	= raid6_check_reshape,
 	.start_reshape  = raid5_start_reshape,
-#endif
+	.finish_reshape = raid5_finish_reshape,
 	.quiesce	= raid5_quiesce,
+	.takeover	= raid6_takeover,
 };
-static struct mdk_personality raid5_personality =
+static struct md_personality raid5_personality =
 {
 	.name		= "raid5",
 	.level		= 5,
@@ -4669,14 +7071,15 @@ static struct mdk_personality raid5_personality =
 	.spare_active	= raid5_spare_active,
 	.sync_request	= sync_request,
 	.resize		= raid5_resize,
-#ifdef CONFIG_MD_RAID5_RESHAPE
+	.size		= raid5_size,
 	.check_reshape	= raid5_check_reshape,
 	.start_reshape  = raid5_start_reshape,
-#endif
+	.finish_reshape = raid5_finish_reshape,
 	.quiesce	= raid5_quiesce,
+	.takeover	= raid5_takeover,
 };
 
-static struct mdk_personality raid4_personality =
+static struct md_personality raid4_personality =
 {
 	.name		= "raid4",
 	.level		= 4,
@@ -4691,20 +7094,20 @@ static struct mdk_personality raid4_personality =
 	.spare_active	= raid5_spare_active,
 	.sync_request	= sync_request,
 	.resize		= raid5_resize,
-#ifdef CONFIG_MD_RAID5_RESHAPE
+	.size		= raid5_size,
 	.check_reshape	= raid5_check_reshape,
 	.start_reshape  = raid5_start_reshape,
-#endif
+	.finish_reshape = raid5_finish_reshape,
 	.quiesce	= raid5_quiesce,
+	.takeover	= raid4_takeover,
 };
 
 static int __init raid5_init(void)
 {
-	int e;
-
-	e = raid6_select_algo();
-	if ( e )
-		return e;
+	raid5_wq = alloc_workqueue("raid5wq",
+		WQ_UNBOUND|WQ_MEM_RECLAIM|WQ_CPU_INTENSIVE|WQ_SYSFS, 0);
+	if (!raid5_wq)
+		return -ENOMEM;
 	register_md_personality(&raid6_personality);
 	register_md_personality(&raid5_personality);
 	register_md_personality(&raid4_personality);
@@ -4716,11 +7119,13 @@ static void raid5_exit(void)
 	unregister_md_personality(&raid6_personality);
 	unregister_md_personality(&raid5_personality);
 	unregister_md_personality(&raid4_personality);
+	destroy_workqueue(raid5_wq);
 }
 
 module_init(raid5_init);
 module_exit(raid5_exit);
 MODULE_LICENSE("GPL");
+MODULE_DESCRIPTION("RAID4/5/6 (striping with parity) personality for MD");
 MODULE_ALIAS("md-personality-4"); /* RAID5 */
 MODULE_ALIAS("md-raid5");
 MODULE_ALIAS("md-raid4");
diff --git a/drivers/md/raid5.h b/drivers/md/raid5.h
new file mode 100644
index 00000000000..bc72cd4be5f
--- /dev/null
+++ b/drivers/md/raid5.h
@@ -0,0 +1,566 @@
+#ifndef _RAID5_H
+#define _RAID5_H
+
+#include <linux/raid/xor.h>
+#include <linux/dmaengine.h>
+
+/*
+ *
+ * Each stripe contains one buffer per device.  Each buffer can be in
+ * one of a number of states stored in "flags".  Changes between
+ * these states happen *almost* exclusively under the protection of the
+ * STRIPE_ACTIVE flag.  Some very specific changes can happen in bi_end_io, and
+ * these are not protected by STRIPE_ACTIVE.
+ *
+ * The flag bits that are used to represent these states are:
+ *   R5_UPTODATE and R5_LOCKED
+ *
+ * State Empty == !UPTODATE, !LOCK
+ *        We have no data, and there is no active request
+ * State Want == !UPTODATE, LOCK
+ *        A read request is being submitted for this block
+ * State Dirty == UPTODATE, LOCK
+ *        Some new data is in this buffer, and it is being written out
+ * State Clean == UPTODATE, !LOCK
+ *        We have valid data which is the same as on disc
+ *
+ * The possible state transitions are:
+ *
+ *  Empty -> Want   - on read or write to get old data for  parity calc
+ *  Empty -> Dirty  - on compute_parity to satisfy write/sync request.
+ *  Empty -> Clean  - on compute_block when computing a block for failed drive
+ *  Want  -> Empty  - on failed read
+ *  Want  -> Clean  - on successful completion of read request
+ *  Dirty -> Clean  - on successful completion of write request
+ *  Dirty -> Clean  - on failed write
+ *  Clean -> Dirty  - on compute_parity to satisfy write/sync (RECONSTRUCT or RMW)
+ *
+ * The Want->Empty, Want->Clean, Dirty->Clean, transitions
+ * all happen in b_end_io at interrupt time.
+ * Each sets the Uptodate bit before releasing the Lock bit.
+ * This leaves one multi-stage transition:
+ *    Want->Dirty->Clean
+ * This is safe because thinking that a Clean buffer is actually dirty
+ * will at worst delay some action, and the stripe will be scheduled
+ * for attention after the transition is complete.
+ *
+ * There is one possibility that is not covered by these states.  That
+ * is if one drive has failed and there is a spare being rebuilt.  We
+ * can't distinguish between a clean block that has been generated
+ * from parity calculations, and a clean block that has been
+ * successfully written to the spare ( or to parity when resyncing).
+ * To distinguish these states we have a stripe bit STRIPE_INSYNC that
+ * is set whenever a write is scheduled to the spare, or to the parity
+ * disc if there is no spare.  A sync request clears this bit, and
+ * when we find it set with no buffers locked, we know the sync is
+ * complete.
+ *
+ * Buffers for the md device that arrive via make_request are attached
+ * to the appropriate stripe in one of two lists linked on b_reqnext.
+ * One list (bh_read) for read requests, one (bh_write) for write.
+ * There should never be more than one buffer on the two lists
+ * together, but we are not guaranteed of that so we allow for more.
+ *
+ * If a buffer is on the read list when the associated cache buffer is
+ * Uptodate, the data is copied into the read buffer and it's b_end_io
+ * routine is called.  This may happen in the end_request routine only
+ * if the buffer has just successfully been read.  end_request should
+ * remove the buffers from the list and then set the Uptodate bit on
+ * the buffer.  Other threads may do this only if they first check
+ * that the Uptodate bit is set.  Once they have checked that they may
+ * take buffers off the read queue.
+ *
+ * When a buffer on the write list is committed for write it is copied
+ * into the cache buffer, which is then marked dirty, and moved onto a
+ * third list, the written list (bh_written).  Once both the parity
+ * block and the cached buffer are successfully written, any buffer on
+ * a written list can be returned with b_end_io.
+ *
+ * The write list and read list both act as fifos.  The read list,
+ * write list and written list are protected by the device_lock.
+ * The device_lock is only for list manipulations and will only be
+ * held for a very short time.  It can be claimed from interrupts.
+ *
+ *
+ * Stripes in the stripe cache can be on one of two lists (or on
+ * neither).  The "inactive_list" contains stripes which are not
+ * currently being used for any request.  They can freely be reused
+ * for another stripe.  The "handle_list" contains stripes that need
+ * to be handled in some way.  Both of these are fifo queues.  Each
+ * stripe is also (potentially) linked to a hash bucket in the hash
+ * table so that it can be found by sector number.  Stripes that are
+ * not hashed must be on the inactive_list, and will normally be at
+ * the front.  All stripes start life this way.
+ *
+ * The inactive_list, handle_list and hash bucket lists are all protected by the
+ * device_lock.
+ *  - stripes have a reference counter. If count==0, they are on a list.
+ *  - If a stripe might need handling, STRIPE_HANDLE is set.
+ *  - When refcount reaches zero, then if STRIPE_HANDLE it is put on
+ *    handle_list else inactive_list
+ *
+ * This, combined with the fact that STRIPE_HANDLE is only ever
+ * cleared while a stripe has a non-zero count means that if the
+ * refcount is 0 and STRIPE_HANDLE is set, then it is on the
+ * handle_list and if recount is 0 and STRIPE_HANDLE is not set, then
+ * the stripe is on inactive_list.
+ *
+ * The possible transitions are:
+ *  activate an unhashed/inactive stripe (get_active_stripe())
+ *     lockdev check-hash unlink-stripe cnt++ clean-stripe hash-stripe unlockdev
+ *  activate a hashed, possibly active stripe (get_active_stripe())
+ *     lockdev check-hash if(!cnt++)unlink-stripe unlockdev
+ *  attach a request to an active stripe (add_stripe_bh())
+ *     lockdev attach-buffer unlockdev
+ *  handle a stripe (handle_stripe())
+ *     setSTRIPE_ACTIVE,  clrSTRIPE_HANDLE ...
+ *		(lockdev check-buffers unlockdev) ..
+ *		change-state ..
+ *		record io/ops needed clearSTRIPE_ACTIVE schedule io/ops
+ *  release an active stripe (release_stripe())
+ *     lockdev if (!--cnt) { if  STRIPE_HANDLE, add to handle_list else add to inactive-list } unlockdev
+ *
+ * The refcount counts each thread that have activated the stripe,
+ * plus raid5d if it is handling it, plus one for each active request
+ * on a cached buffer, and plus one if the stripe is undergoing stripe
+ * operations.
+ *
+ * The stripe operations are:
+ * -copying data between the stripe cache and user application buffers
+ * -computing blocks to save a disk access, or to recover a missing block
+ * -updating the parity on a write operation (reconstruct write and
+ *  read-modify-write)
+ * -checking parity correctness
+ * -running i/o to disk
+ * These operations are carried out by raid5_run_ops which uses the async_tx
+ * api to (optionally) offload operations to dedicated hardware engines.
+ * When requesting an operation handle_stripe sets the pending bit for the
+ * operation and increments the count.  raid5_run_ops is then run whenever
+ * the count is non-zero.
+ * There are some critical dependencies between the operations that prevent some
+ * from being requested while another is in flight.
+ * 1/ Parity check operations destroy the in cache version of the parity block,
+ *    so we prevent parity dependent operations like writes and compute_blocks
+ *    from starting while a check is in progress.  Some dma engines can perform
+ *    the check without damaging the parity block, in these cases the parity
+ *    block is re-marked up to date (assuming the check was successful) and is
+ *    not re-read from disk.
+ * 2/ When a write operation is requested we immediately lock the affected
+ *    blocks, and mark them as not up to date.  This causes new read requests
+ *    to be held off, as well as parity checks and compute block operations.
+ * 3/ Once a compute block operation has been requested handle_stripe treats
+ *    that block as if it is up to date.  raid5_run_ops guaruntees that any
+ *    operation that is dependent on the compute block result is initiated after
+ *    the compute block completes.
+ */
+
+/*
+ * Operations state - intermediate states that are visible outside of 
+ *   STRIPE_ACTIVE.
+ * In general _idle indicates nothing is running, _run indicates a data
+ * processing operation is active, and _result means the data processing result
+ * is stable and can be acted upon.  For simple operations like biofill and
+ * compute that only have an _idle and _run state they are indicated with
+ * sh->state flags (STRIPE_BIOFILL_RUN and STRIPE_COMPUTE_RUN)
+ */
+/**
+ * enum check_states - handles syncing / repairing a stripe
+ * @check_state_idle - check operations are quiesced
+ * @check_state_run - check operation is running
+ * @check_state_result - set outside lock when check result is valid
+ * @check_state_compute_run - check failed and we are repairing
+ * @check_state_compute_result - set outside lock when compute result is valid
+ */
+enum check_states {
+	check_state_idle = 0,
+	check_state_run, /* xor parity check */
+	check_state_run_q, /* q-parity check */
+	check_state_run_pq, /* pq dual parity check */
+	check_state_check_result,
+	check_state_compute_run, /* parity repair */
+	check_state_compute_result,
+};
+
+/**
+ * enum reconstruct_states - handles writing or expanding a stripe
+ */
+enum reconstruct_states {
+	reconstruct_state_idle = 0,
+	reconstruct_state_prexor_drain_run,	/* prexor-write */
+	reconstruct_state_drain_run,		/* write */
+	reconstruct_state_run,			/* expand */
+	reconstruct_state_prexor_drain_result,
+	reconstruct_state_drain_result,
+	reconstruct_state_result,
+};
+
+struct stripe_head {
+	struct hlist_node	hash;
+	struct list_head	lru;	      /* inactive_list or handle_list */
+	struct llist_node	release_list;
+	struct r5conf		*raid_conf;
+	short			generation;	/* increments with every
+						 * reshape */
+	sector_t		sector;		/* sector of this row */
+	short			pd_idx;		/* parity disk index */
+	short			qd_idx;		/* 'Q' disk index for raid6 */
+	short			ddf_layout;/* use DDF ordering to calculate Q */
+	short			hash_lock_index;
+	unsigned long		state;		/* state flags */
+	atomic_t		count;	      /* nr of active thread/requests */
+	int			bm_seq;	/* sequence number for bitmap flushes */
+	int			disks;		/* disks in stripe */
+	enum check_states	check_state;
+	enum reconstruct_states reconstruct_state;
+	spinlock_t		stripe_lock;
+	int			cpu;
+	struct r5worker_group	*group;
+	/**
+	 * struct stripe_operations
+	 * @target - STRIPE_OP_COMPUTE_BLK target
+	 * @target2 - 2nd compute target in the raid6 case
+	 * @zero_sum_result - P and Q verification flags
+	 * @request - async service request flags for raid_run_ops
+	 */
+	struct stripe_operations {
+		int 		     target, target2;
+		enum sum_check_flags zero_sum_result;
+	} ops;
+	struct r5dev {
+		/* rreq and rvec are used for the replacement device when
+		 * writing data to both devices.
+		 */
+		struct bio	req, rreq;
+		struct bio_vec	vec, rvec;
+		struct page	*page, *orig_page;
+		struct bio	*toread, *read, *towrite, *written;
+		sector_t	sector;			/* sector of this page */
+		unsigned long	flags;
+	} dev[1]; /* allocated with extra space depending of RAID geometry */
+};
+
+/* stripe_head_state - collects and tracks the dynamic state of a stripe_head
+ *     for handle_stripe.
+ */
+struct stripe_head_state {
+	/* 'syncing' means that we need to read all devices, either
+	 * to check/correct parity, or to reconstruct a missing device.
+	 * 'replacing' means we are replacing one or more drives and
+	 * the source is valid at this point so we don't need to
+	 * read all devices, just the replacement targets.
+	 */
+	int syncing, expanding, expanded, replacing;
+	int locked, uptodate, to_read, to_write, failed, written;
+	int to_fill, compute, req_compute, non_overwrite;
+	int failed_num[2];
+	int p_failed, q_failed;
+	int dec_preread_active;
+	unsigned long ops_request;
+
+	struct bio *return_bi;
+	struct md_rdev *blocked_rdev;
+	int handle_bad_blocks;
+};
+
+/* Flags for struct r5dev.flags */
+enum r5dev_flags {
+	R5_UPTODATE,	/* page contains current data */
+	R5_LOCKED,	/* IO has been submitted on "req" */
+	R5_DOUBLE_LOCKED,/* Cannot clear R5_LOCKED until 2 writes complete */
+	R5_OVERWRITE,	/* towrite covers whole page */
+/* and some that are internal to handle_stripe */
+	R5_Insync,	/* rdev && rdev->in_sync at start */
+	R5_Wantread,	/* want to schedule a read */
+	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 */
+
+	R5_Expanded,	/* This block now has post-expand data */
+	R5_Wantcompute,	/* compute_block in progress treat as
+			 * uptodate
+			 */
+	R5_Wantfill,	/* dev->toread contains a bio that needs
+			 * filling
+			 */
+	R5_Wantdrain,	/* dev->towrite needs to be drained */
+	R5_WantFUA,	/* Write should be FUA */
+	R5_SyncIO,	/* The IO is sync */
+	R5_WriteError,	/* got a write error - need to record it */
+	R5_MadeGood,	/* A bad block has been fixed by writing to it */
+	R5_ReadRepl,	/* Will/did read from replacement rather than orig */
+	R5_MadeGoodRepl,/* A bad block on the replacement device has been
+			 * fixed by writing to it */
+	R5_NeedReplace,	/* This device has a replacement which is not
+			 * up-to-date at this stripe. */
+	R5_WantReplace, /* We need to update the replacement, we have read
+			 * data in, and now is a good time to write it out.
+			 */
+	R5_Discard,	/* Discard the stripe */
+	R5_SkipCopy,	/* Don't copy data from bio to stripe cache */
+};
+
+/*
+ * Stripe state
+ */
+enum {
+	STRIPE_ACTIVE,
+	STRIPE_HANDLE,
+	STRIPE_SYNC_REQUESTED,
+	STRIPE_SYNCING,
+	STRIPE_INSYNC,
+	STRIPE_REPLACED,
+	STRIPE_PREREAD_ACTIVE,
+	STRIPE_DELAYED,
+	STRIPE_DEGRADED,
+	STRIPE_BIT_DELAY,
+	STRIPE_EXPANDING,
+	STRIPE_EXPAND_SOURCE,
+	STRIPE_EXPAND_READY,
+	STRIPE_IO_STARTED,	/* do not count towards 'bypass_count' */
+	STRIPE_FULL_WRITE,	/* all blocks are set to be overwritten */
+	STRIPE_BIOFILL_RUN,
+	STRIPE_COMPUTE_RUN,
+	STRIPE_OPS_REQ_PENDING,
+	STRIPE_ON_UNPLUG_LIST,
+	STRIPE_DISCARD,
+	STRIPE_ON_RELEASE_LIST,
+};
+
+/*
+ * Operation request flags
+ */
+enum {
+	STRIPE_OP_BIOFILL,
+	STRIPE_OP_COMPUTE_BLK,
+	STRIPE_OP_PREXOR,
+	STRIPE_OP_BIODRAIN,
+	STRIPE_OP_RECONSTRUCT,
+	STRIPE_OP_CHECK,
+};
+/*
+ * Plugging:
+ *
+ * To improve write throughput, we need to delay the handling of some
+ * stripes until there has been a chance that several write requests
+ * for the one stripe have all been collected.
+ * In particular, any write request that would require pre-reading
+ * is put on a "delayed" queue until there are no stripes currently
+ * in a pre-read phase.  Further, if the "delayed" queue is empty when
+ * a stripe is put on it then we "plug" the queue and do not process it
+ * until an unplug call is made. (the unplug_io_fn() is called).
+ *
+ * When preread is initiated on a stripe, we set PREREAD_ACTIVE and add
+ * it to the count of prereading stripes.
+ * When write is initiated, or the stripe refcnt == 0 (just in case) we
+ * clear the PREREAD_ACTIVE flag and decrement the count
+ * Whenever the 'handle' queue is empty and the device is not plugged, we
+ * move any strips from delayed to handle and clear the DELAYED flag and set
+ * PREREAD_ACTIVE.
+ * In stripe_handle, if we find pre-reading is necessary, we do it if
+ * PREREAD_ACTIVE is set, else we set DELAYED which will send it to the delayed queue.
+ * HANDLE gets cleared if stripe_handle leaves nothing locked.
+ */
+
+
+struct disk_info {
+	struct md_rdev	*rdev, *replacement;
+};
+
+/* NOTE NR_STRIPE_HASH_LOCKS must remain below 64.
+ * This is because we sometimes take all the spinlocks
+ * and creating that much locking depth can cause
+ * problems.
+ */
+#define NR_STRIPE_HASH_LOCKS 8
+#define STRIPE_HASH_LOCKS_MASK (NR_STRIPE_HASH_LOCKS - 1)
+
+struct r5worker {
+	struct work_struct work;
+	struct r5worker_group *group;
+	struct list_head temp_inactive_list[NR_STRIPE_HASH_LOCKS];
+	bool working;
+};
+
+struct r5worker_group {
+	struct list_head handle_list;
+	struct r5conf *conf;
+	struct r5worker *workers;
+	int stripes_cnt;
+};
+
+struct r5conf {
+	struct hlist_head	*stripe_hashtbl;
+	/* only protect corresponding hash list and inactive_list */
+	spinlock_t		hash_locks[NR_STRIPE_HASH_LOCKS];
+	struct mddev		*mddev;
+	int			chunk_sectors;
+	int			level, algorithm;
+	int			max_degraded;
+	int			raid_disks;
+	int			max_nr_stripes;
+
+	/* reshape_progress is the leading edge of a 'reshape'
+	 * It has value MaxSector when no reshape is happening
+	 * If delta_disks < 0, it is the last sector we started work on,
+	 * else is it the next sector to work on.
+	 */
+	sector_t		reshape_progress;
+	/* reshape_safe is the trailing edge of a reshape.  We know that
+	 * before (or after) this address, all reshape has completed.
+	 */
+	sector_t		reshape_safe;
+	int			previous_raid_disks;
+	int			prev_chunk_sectors;
+	int			prev_algo;
+	short			generation; /* increments with every reshape */
+	seqcount_t		gen_lock;	/* lock against generation changes */
+	unsigned long		reshape_checkpoint; /* Time we last updated
+						     * metadata */
+	long long		min_offset_diff; /* minimum difference between
+						  * data_offset and
+						  * new_data_offset across all
+						  * devices.  May be negative,
+						  * but is closest to zero.
+						  */
+
+	struct list_head	handle_list; /* stripes needing handling */
+	struct list_head	hold_list; /* preread ready stripes */
+	struct list_head	delayed_list; /* stripes that have plugged requests */
+	struct list_head	bitmap_list; /* stripes delaying awaiting bitmap update */
+	struct bio		*retry_read_aligned; /* currently retrying aligned bios   */
+	struct bio		*retry_read_aligned_list; /* aligned bios retry list  */
+	atomic_t		preread_active_stripes; /* stripes with scheduled io */
+	atomic_t		active_aligned_reads;
+	atomic_t		pending_full_writes; /* full write backlog */
+	int			bypass_count; /* bypassed prereads */
+	int			bypass_threshold; /* preread nice */
+	int			skip_copy; /* Don't copy data from bio to stripe cache */
+	struct list_head	*last_hold; /* detect hold_list promotions */
+
+	atomic_t		reshape_stripes; /* stripes with pending writes for reshape */
+	/* unfortunately we need two cache names as we temporarily have
+	 * two caches.
+	 */
+	int			active_name;
+	char			cache_name[2][32];
+	struct kmem_cache		*slab_cache; /* for allocating stripes */
+
+	int			seq_flush, seq_write;
+	int			quiesce;
+
+	int			fullsync;  /* set to 1 if a full sync is needed,
+					    * (fresh device added).
+					    * Cleared when a sync completes.
+					    */
+	int			recovery_disabled;
+	/* per cpu variables */
+	struct raid5_percpu {
+		struct page	*spare_page; /* Used when checking P/Q in raid6 */
+		void		*scribble;   /* space for constructing buffer
+					      * lists and performing address
+					      * conversions
+					      */
+	} __percpu *percpu;
+	size_t			scribble_len; /* size of scribble region must be
+					       * associated with conf to handle
+					       * cpu hotplug while reshaping
+					       */
+#ifdef CONFIG_HOTPLUG_CPU
+	struct notifier_block	cpu_notify;
+#endif
+
+	/*
+	 * Free stripes pool
+	 */
+	atomic_t		active_stripes;
+	struct list_head	inactive_list[NR_STRIPE_HASH_LOCKS];
+	atomic_t		empty_inactive_list_nr;
+	struct llist_head	released_stripes;
+	wait_queue_head_t	wait_for_stripe;
+	wait_queue_head_t	wait_for_overlap;
+	int			inactive_blocked;	/* release of inactive stripes blocked,
+							 * waiting for 25% to be free
+							 */
+	int			pool_size; /* number of disks in stripeheads in pool */
+	spinlock_t		device_lock;
+	struct disk_info	*disks;
+
+	/* When taking over an array from a different personality, we store
+	 * the new thread here until we fully activate the array.
+	 */
+	struct md_thread	*thread;
+	struct list_head	temp_inactive_list[NR_STRIPE_HASH_LOCKS];
+	struct r5worker_group	*worker_groups;
+	int			group_cnt;
+	int			worker_cnt_per_group;
+};
+
+/*
+ * Our supported algorithms
+ */
+#define ALGORITHM_LEFT_ASYMMETRIC	0 /* Rotating Parity N with Data Restart */
+#define ALGORITHM_RIGHT_ASYMMETRIC	1 /* Rotating Parity 0 with Data Restart */
+#define ALGORITHM_LEFT_SYMMETRIC	2 /* Rotating Parity N with Data Continuation */
+#define ALGORITHM_RIGHT_SYMMETRIC	3 /* Rotating Parity 0 with Data Continuation */
+
+/* Define non-rotating (raid4) algorithms.  These allow
+ * conversion of raid4 to raid5.
+ */
+#define ALGORITHM_PARITY_0		4 /* P or P,Q are initial devices */
+#define ALGORITHM_PARITY_N		5 /* P or P,Q are final devices. */
+
+/* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
+ * Firstly, the exact positioning of the parity block is slightly
+ * different between the 'LEFT_*' modes of md and the "_N_*" modes
+ * of DDF.
+ * Secondly, or order of datablocks over which the Q syndrome is computed
+ * is different.
+ * Consequently we have different layouts for DDF/raid6 than md/raid6.
+ * These layouts are from the DDFv1.2 spec.
+ * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
+ * leaves RLQ=3 as 'Vendor Specific'
+ */
+
+#define ALGORITHM_ROTATING_ZERO_RESTART	8 /* DDF PRL=6 RLQ=1 */
+#define ALGORITHM_ROTATING_N_RESTART	9 /* DDF PRL=6 RLQ=2 */
+#define ALGORITHM_ROTATING_N_CONTINUE	10 /*DDF PRL=6 RLQ=3 */
+
+
+/* For every RAID5 algorithm we define a RAID6 algorithm
+ * with exactly the same layout for data and parity, and
+ * with the Q block always on the last device (N-1).
+ * This allows trivial conversion from RAID5 to RAID6
+ */
+#define ALGORITHM_LEFT_ASYMMETRIC_6	16
+#define ALGORITHM_RIGHT_ASYMMETRIC_6	17
+#define ALGORITHM_LEFT_SYMMETRIC_6	18
+#define ALGORITHM_RIGHT_SYMMETRIC_6	19
+#define ALGORITHM_PARITY_0_6		20
+#define ALGORITHM_PARITY_N_6		ALGORITHM_PARITY_N
+
+static inline int algorithm_valid_raid5(int layout)
+{
+	return (layout >= 0) &&
+		(layout <= 5);
+}
+static inline int algorithm_valid_raid6(int layout)
+{
+	return (layout >= 0 && layout <= 5)
+		||
+		(layout >= 8 && layout <= 10)
+		||
+		(layout >= 16 && layout <= 20);
+}
+
+static inline int algorithm_is_DDF(int layout)
+{
+	return layout >= 8 && layout <= 10;
+}
+
+extern int md_raid5_congested(struct mddev *mddev, int bits);
+extern void md_raid5_kick_device(struct r5conf *conf);
+extern int raid5_set_cache_size(struct mddev *mddev, int size);
+#endif
diff --git a/drivers/md/raid6.h b/drivers/md/raid6.h
deleted file mode 100644
index 31cbee71365..00000000000
--- a/drivers/md/raid6.h
+++ /dev/null
@@ -1,139 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2003 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-#ifndef LINUX_RAID_RAID6_H
-#define LINUX_RAID_RAID6_H
-
-#ifdef __KERNEL__
-
-/* Set to 1 to use kernel-wide empty_zero_page */
-#define RAID6_USE_EMPTY_ZERO_PAGE 0
-
-#include <linux/module.h>
-#include <linux/stddef.h>
-#include <linux/compiler.h>
-#include <linux/types.h>
-#include <linux/kernel.h>
-#include <linux/errno.h>
-#include <linux/mempool.h>
-#include <linux/list.h>
-#include <linux/vmalloc.h>
-#include <linux/raid/md.h>
-#include <linux/raid/raid5.h>
-
-typedef raid5_conf_t raid6_conf_t; /* Same configuration */
-
-/* Additional compute_parity mode -- updates the parity w/o LOCKING */
-#define UPDATE_PARITY	4
-
-/* We need a pre-zeroed page... if we don't want to use the kernel-provided
-   one define it here */
-#if RAID6_USE_EMPTY_ZERO_PAGE
-# define raid6_empty_zero_page empty_zero_page
-#else
-extern const char raid6_empty_zero_page[PAGE_SIZE];
-#endif
-
-#else /* ! __KERNEL__ */
-/* Used for testing in user space */
-
-#include <errno.h>
-#include <inttypes.h>
-#include <limits.h>
-#include <stddef.h>
-#include <sys/mman.h>
-#include <sys/types.h>
-
-/* Not standard, but glibc defines it */
-#define BITS_PER_LONG __WORDSIZE
-
-typedef uint8_t  u8;
-typedef uint16_t u16;
-typedef uint32_t u32;
-typedef uint64_t u64;
-
-#ifndef PAGE_SIZE
-# define PAGE_SIZE 4096
-#endif
-extern const char raid6_empty_zero_page[PAGE_SIZE];
-
-#define __init
-#define __exit
-#define __attribute_const__ __attribute__((const))
-#define noinline __attribute__((noinline))
-
-#define preempt_enable()
-#define preempt_disable()
-#define cpu_has_feature(x) 1
-#define enable_kernel_altivec()
-#define disable_kernel_altivec()
-
-#endif /* __KERNEL__ */
-
-/* Routine choices */
-struct raid6_calls {
-	void (*gen_syndrome)(int, size_t, void **);
-	int  (*valid)(void);	/* Returns 1 if this routine set is usable */
-	const char *name;	/* Name of this routine set */
-	int prefer;		/* Has special performance attribute */
-};
-
-/* Selected algorithm */
-extern struct raid6_calls raid6_call;
-
-/* Algorithm list */
-extern const struct raid6_calls * const raid6_algos[];
-int raid6_select_algo(void);
-
-/* Return values from chk_syndrome */
-#define RAID6_OK	0
-#define RAID6_P_BAD	1
-#define RAID6_Q_BAD	2
-#define RAID6_PQ_BAD	3
-
-/* Galois field tables */
-extern const u8 raid6_gfmul[256][256] __attribute__((aligned(256)));
-extern const u8 raid6_gfexp[256]      __attribute__((aligned(256)));
-extern const u8 raid6_gfinv[256]      __attribute__((aligned(256)));
-extern const u8 raid6_gfexi[256]      __attribute__((aligned(256)));
-
-/* Recovery routines */
-void raid6_2data_recov(int disks, size_t bytes, int faila, int failb, void **ptrs);
-void raid6_datap_recov(int disks, size_t bytes, int faila, void **ptrs);
-void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, void **ptrs);
-
-/* Some definitions to allow code to be compiled for testing in userspace */
-#ifndef __KERNEL__
-
-# define jiffies	raid6_jiffies()
-# define printk 	printf
-# define GFP_KERNEL	0
-# define __get_free_pages(x,y)	((unsigned long)mmap(NULL, PAGE_SIZE << (y), PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, 0, 0))
-# define free_pages(x,y)	munmap((void *)(x), (y)*PAGE_SIZE)
-
-static inline void cpu_relax(void)
-{
-	/* Nothing */
-}
-
-#undef  HZ
-#define HZ 1000
-static inline uint32_t raid6_jiffies(void)
-{
-	struct timeval tv;
-	gettimeofday(&tv, NULL);
-	return tv.tv_sec*1000 + tv.tv_usec/1000;
-}
-
-#endif /* ! __KERNEL__ */
-
-#endif /* LINUX_RAID_RAID6_H */
diff --git a/drivers/md/raid6algos.c b/drivers/md/raid6algos.c
deleted file mode 100644
index 21987e3dbe6..00000000000
--- a/drivers/md/raid6algos.c
+++ /dev/null
@@ -1,154 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6algos.c
- *
- * Algorithm list and algorithm selection for RAID-6
- */
-
-#include "raid6.h"
-#ifndef __KERNEL__
-#include <sys/mman.h>
-#include <stdio.h>
-#endif
-
-struct raid6_calls raid6_call;
-
-/* Various routine sets */
-extern const struct raid6_calls raid6_intx1;
-extern const struct raid6_calls raid6_intx2;
-extern const struct raid6_calls raid6_intx4;
-extern const struct raid6_calls raid6_intx8;
-extern const struct raid6_calls raid6_intx16;
-extern const struct raid6_calls raid6_intx32;
-extern const struct raid6_calls raid6_mmxx1;
-extern const struct raid6_calls raid6_mmxx2;
-extern const struct raid6_calls raid6_sse1x1;
-extern const struct raid6_calls raid6_sse1x2;
-extern const struct raid6_calls raid6_sse2x1;
-extern const struct raid6_calls raid6_sse2x2;
-extern const struct raid6_calls raid6_sse2x4;
-extern const struct raid6_calls raid6_altivec1;
-extern const struct raid6_calls raid6_altivec2;
-extern const struct raid6_calls raid6_altivec4;
-extern const struct raid6_calls raid6_altivec8;
-
-const struct raid6_calls * const raid6_algos[] = {
-	&raid6_intx1,
-	&raid6_intx2,
-	&raid6_intx4,
-	&raid6_intx8,
-#if defined(__ia64__)
-	&raid6_intx16,
-	&raid6_intx32,
-#endif
-#if defined(__i386__) && !defined(__arch_um__)
-	&raid6_mmxx1,
-	&raid6_mmxx2,
-	&raid6_sse1x1,
-	&raid6_sse1x2,
-	&raid6_sse2x1,
-	&raid6_sse2x2,
-#endif
-#if defined(__x86_64__) && !defined(__arch_um__)
-	&raid6_sse2x1,
-	&raid6_sse2x2,
-	&raid6_sse2x4,
-#endif
-#ifdef CONFIG_ALTIVEC
-	&raid6_altivec1,
-	&raid6_altivec2,
-	&raid6_altivec4,
-	&raid6_altivec8,
-#endif
-	NULL
-};
-
-#ifdef __KERNEL__
-#define RAID6_TIME_JIFFIES_LG2	4
-#else
-/* Need more time to be stable in userspace */
-#define RAID6_TIME_JIFFIES_LG2	9
-#endif
-
-/* Try to pick the best algorithm */
-/* This code uses the gfmul table as convenient data set to abuse */
-
-int __init raid6_select_algo(void)
-{
-	const struct raid6_calls * const * algo;
-	const struct raid6_calls * best;
-	char *syndromes;
-	void *dptrs[(65536/PAGE_SIZE)+2];
-	int i, disks;
-	unsigned long perf, bestperf;
-	int bestprefer;
-	unsigned long j0, j1;
-
-	disks = (65536/PAGE_SIZE)+2;
-	for ( i = 0 ; i < disks-2 ; i++ ) {
-		dptrs[i] = ((char *)raid6_gfmul) + PAGE_SIZE*i;
-	}
-
-	/* Normal code - use a 2-page allocation to avoid D$ conflict */
-	syndromes = (void *) __get_free_pages(GFP_KERNEL, 1);
-
-	if ( !syndromes ) {
-		printk("raid6: Yikes!  No memory available.\n");
-		return -ENOMEM;
-	}
-
-	dptrs[disks-2] = syndromes;
-	dptrs[disks-1] = syndromes + PAGE_SIZE;
-
-	bestperf = 0;  bestprefer = 0;  best = NULL;
-
-	for ( algo = raid6_algos ; *algo ; algo++ ) {
-		if ( !(*algo)->valid || (*algo)->valid() ) {
-			perf = 0;
-
-			preempt_disable();
-			j0 = jiffies;
-			while ( (j1 = jiffies) == j0 )
-				cpu_relax();
-			while (time_before(jiffies,
-					    j1 + (1<<RAID6_TIME_JIFFIES_LG2))) {
-				(*algo)->gen_syndrome(disks, PAGE_SIZE, dptrs);
-				perf++;
-			}
-			preempt_enable();
-
-			if ( (*algo)->prefer > bestprefer ||
-			     ((*algo)->prefer == bestprefer &&
-			      perf > bestperf) ) {
-				best = *algo;
-				bestprefer = best->prefer;
-				bestperf = perf;
-			}
-			printk("raid6: %-8s %5ld MB/s\n", (*algo)->name,
-			       (perf*HZ) >> (20-16+RAID6_TIME_JIFFIES_LG2));
-		}
-	}
-
-	if (best) {
-		printk("raid6: using algorithm %s (%ld MB/s)\n",
-		       best->name,
-		       (bestperf*HZ) >> (20-16+RAID6_TIME_JIFFIES_LG2));
-		raid6_call = *best;
-	} else
-		printk("raid6: Yikes!  No algorithm found!\n");
-
-	free_pages((unsigned long)syndromes, 1);
-
-	return best ? 0 : -EINVAL;
-}
diff --git a/drivers/md/raid6altivec.uc b/drivers/md/raid6altivec.uc
deleted file mode 100644
index b9afd35b881..00000000000
--- a/drivers/md/raid6altivec.uc
+++ /dev/null
@@ -1,130 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002-2004 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6altivec$#.c
- *
- * $#-way unrolled portable integer math RAID-6 instruction set
- *
- * This file is postprocessed using unroll.pl
- *
- * <benh> hpa: in process,
- * you can just "steal" the vec unit with enable_kernel_altivec() (but
- * bracked this with preempt_disable/enable or in a lock)
- */
-
-#include "raid6.h"
-
-#ifdef CONFIG_ALTIVEC
-
-#include <altivec.h>
-#ifdef __KERNEL__
-# include <asm/system.h>
-# include <asm/cputable.h>
-#endif
-
-/*
- * This is the C data type to use.  We use a vector of
- * signed char so vec_cmpgt() will generate the right
- * instruction.
- */
-
-typedef vector signed char unative_t;
-
-#define NBYTES(x) ((vector signed char) {x,x,x,x, x,x,x,x, x,x,x,x, x,x,x,x})
-#define NSIZE	sizeof(unative_t)
-
-/*
- * The SHLBYTE() operation shifts each byte left by 1, *not*
- * rolling over into the next byte
- */
-static inline __attribute_const__ unative_t SHLBYTE(unative_t v)
-{
-	return vec_add(v,v);
-}
-
-/*
- * The MASK() operation returns 0xFF in any byte for which the high
- * bit is 1, 0x00 for any byte for which the high bit is 0.
- */
-static inline __attribute_const__ unative_t MASK(unative_t v)
-{
-	unative_t zv = NBYTES(0);
-
-	/* vec_cmpgt returns a vector bool char; thus the need for the cast */
-	return (unative_t)vec_cmpgt(zv, v);
-}
-
-
-/* This is noinline to make damned sure that gcc doesn't move any of the
-   Altivec code around the enable/disable code */
-static void noinline
-raid6_altivec$#_gen_syndrome_real(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	unative_t wd$$, wq$$, wp$$, w1$$, w2$$;
-	unative_t x1d = NBYTES(0x1d);
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	for ( d = 0 ; d < bytes ; d += NSIZE*$# ) {
-		wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE];
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE];
-			wp$$ = vec_xor(wp$$, wd$$);
-			w2$$ = MASK(wq$$);
-			w1$$ = SHLBYTE(wq$$);
-			w2$$ = vec_and(w2$$, x1d);
-			w1$$ = vec_xor(w1$$, w2$$);
-			wq$$ = vec_xor(w1$$, wd$$);
-		}
-		*(unative_t *)&p[d+NSIZE*$$] = wp$$;
-		*(unative_t *)&q[d+NSIZE*$$] = wq$$;
-	}
-}
-
-static void raid6_altivec$#_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	preempt_disable();
-	enable_kernel_altivec();
-
-	raid6_altivec$#_gen_syndrome_real(disks, bytes, ptrs);
-
-	preempt_enable();
-}
-
-int raid6_have_altivec(void);
-#if $# == 1
-int raid6_have_altivec(void)
-{
-	/* This assumes either all CPUs have Altivec or none does */
-# ifdef __KERNEL__
-	return cpu_has_feature(CPU_FTR_ALTIVEC);
-# else
-	return 1;
-# endif
-}
-#endif
-
-const struct raid6_calls raid6_altivec$# = {
-	raid6_altivec$#_gen_syndrome,
-	raid6_have_altivec,
-	"altivecx$#",
-	0
-};
-
-#endif /* CONFIG_ALTIVEC */
diff --git a/drivers/md/raid6int.uc b/drivers/md/raid6int.uc
deleted file mode 100644
index ad004cee0e2..00000000000
--- a/drivers/md/raid6int.uc
+++ /dev/null
@@ -1,117 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002-2004 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6int$#.c
- *
- * $#-way unrolled portable integer math RAID-6 instruction set
- *
- * This file is postprocessed using unroll.pl
- */
-
-#include "raid6.h"
-
-/*
- * This is the C data type to use
- */
-
-/* Change this from BITS_PER_LONG if there is something better... */
-#if BITS_PER_LONG == 64
-# define NBYTES(x) ((x) * 0x0101010101010101UL)
-# define NSIZE  8
-# define NSHIFT 3
-# define NSTRING "64"
-typedef u64 unative_t;
-#else
-# define NBYTES(x) ((x) * 0x01010101U)
-# define NSIZE  4
-# define NSHIFT 2
-# define NSTRING "32"
-typedef u32 unative_t;
-#endif
-
-
-
-/*
- * IA-64 wants insane amounts of unrolling.  On other architectures that
- * is just a waste of space.
- */
-#if ($# <= 8) || defined(__ia64__)
-
-
-/*
- * These sub-operations are separate inlines since they can sometimes be
- * specially optimized using architecture-specific hacks.
- */
-
-/*
- * The SHLBYTE() operation shifts each byte left by 1, *not*
- * rolling over into the next byte
- */
-static inline __attribute_const__ unative_t SHLBYTE(unative_t v)
-{
-	unative_t vv;
-
-	vv = (v << 1) & NBYTES(0xfe);
-	return vv;
-}
-
-/*
- * The MASK() operation returns 0xFF in any byte for which the high
- * bit is 1, 0x00 for any byte for which the high bit is 0.
- */
-static inline __attribute_const__ unative_t MASK(unative_t v)
-{
-	unative_t vv;
-
-	vv = v & NBYTES(0x80);
-	vv = (vv << 1) - (vv >> 7); /* Overflow on the top bit is OK */
-	return vv;
-}
-
-
-static void raid6_int$#_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	unative_t wd$$, wq$$, wp$$, w1$$, w2$$;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	for ( d = 0 ; d < bytes ; d += NSIZE*$# ) {
-		wq$$ = wp$$ = *(unative_t *)&dptr[z0][d+$$*NSIZE];
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			wd$$ = *(unative_t *)&dptr[z][d+$$*NSIZE];
-			wp$$ ^= wd$$;
-			w2$$ = MASK(wq$$);
-			w1$$ = SHLBYTE(wq$$);
-			w2$$ &= NBYTES(0x1d);
-			w1$$ ^= w2$$;
-			wq$$ = w1$$ ^ wd$$;
-		}
-		*(unative_t *)&p[d+NSIZE*$$] = wp$$;
-		*(unative_t *)&q[d+NSIZE*$$] = wq$$;
-	}
-}
-
-const struct raid6_calls raid6_intx$# = {
-	raid6_int$#_gen_syndrome,
-	NULL,		/* always valid */
-	"int" NSTRING "x$#",
-	0
-};
-
-#endif
diff --git a/drivers/md/raid6mmx.c b/drivers/md/raid6mmx.c
deleted file mode 100644
index d4e4a1bd70a..00000000000
--- a/drivers/md/raid6mmx.c
+++ /dev/null
@@ -1,142 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6mmx.c
- *
- * MMX implementation of RAID-6 syndrome functions
- */
-
-#if defined(__i386__) && !defined(__arch_um__)
-
-#include "raid6.h"
-#include "raid6x86.h"
-
-/* Shared with raid6sse1.c */
-const struct raid6_mmx_constants {
-	u64 x1d;
-} raid6_mmx_constants = {
-	0x1d1d1d1d1d1d1d1dULL,
-};
-
-static int raid6_have_mmx(void)
-{
-	/* Not really "boot_cpu" but "all_cpus" */
-	return boot_cpu_has(X86_FEATURE_MMX);
-}
-
-/*
- * Plain MMX implementation
- */
-static void raid6_mmx1_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
-	asm volatile("pxor %mm5,%mm5");	/* Zero temp */
-
-	for ( d = 0 ; d < bytes ; d += 8 ) {
-		asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
-		asm volatile("movq %mm2,%mm4");	/* Q[0] */
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			asm volatile("movq %0,%%mm6" : : "m" (dptr[z][d]));
-			asm volatile("pcmpgtb %mm4,%mm5");
-			asm volatile("paddb %mm4,%mm4");
-			asm volatile("pand %mm0,%mm5");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm5,%mm5");
-			asm volatile("pxor %mm6,%mm2");
-			asm volatile("pxor %mm6,%mm4");
-		}
-		asm volatile("movq %%mm2,%0" : "=m" (p[d]));
-		asm volatile("pxor %mm2,%mm2");
-		asm volatile("movq %%mm4,%0" : "=m" (q[d]));
-		asm volatile("pxor %mm4,%mm4");
-	}
-
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_mmxx1 = {
-	raid6_mmx1_gen_syndrome,
-	raid6_have_mmx,
-	"mmxx1",
-	0
-};
-
-/*
- * Unrolled-by-2 MMX implementation
- */
-static void raid6_mmx2_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
-	asm volatile("pxor %mm5,%mm5");	/* Zero temp */
-	asm volatile("pxor %mm7,%mm7"); /* Zero temp */
-
-	for ( d = 0 ; d < bytes ; d += 16 ) {
-		asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
-		asm volatile("movq %0,%%mm3" : : "m" (dptr[z0][d+8]));
-		asm volatile("movq %mm2,%mm4"); /* Q[0] */
-		asm volatile("movq %mm3,%mm6"); /* Q[1] */
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			asm volatile("pcmpgtb %mm4,%mm5");
-			asm volatile("pcmpgtb %mm6,%mm7");
-			asm volatile("paddb %mm4,%mm4");
-			asm volatile("paddb %mm6,%mm6");
-			asm volatile("pand %mm0,%mm5");
-			asm volatile("pand %mm0,%mm7");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm7,%mm6");
-			asm volatile("movq %0,%%mm5" : : "m" (dptr[z][d]));
-			asm volatile("movq %0,%%mm7" : : "m" (dptr[z][d+8]));
-			asm volatile("pxor %mm5,%mm2");
-			asm volatile("pxor %mm7,%mm3");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm7,%mm6");
-			asm volatile("pxor %mm5,%mm5");
-			asm volatile("pxor %mm7,%mm7");
-		}
-		asm volatile("movq %%mm2,%0" : "=m" (p[d]));
-		asm volatile("movq %%mm3,%0" : "=m" (p[d+8]));
-		asm volatile("movq %%mm4,%0" : "=m" (q[d]));
-		asm volatile("movq %%mm6,%0" : "=m" (q[d+8]));
-	}
-
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_mmxx2 = {
-	raid6_mmx2_gen_syndrome,
-	raid6_have_mmx,
-	"mmxx2",
-	0
-};
-
-#endif
diff --git a/drivers/md/raid6recov.c b/drivers/md/raid6recov.c
deleted file mode 100644
index a8c4d9451bd..00000000000
--- a/drivers/md/raid6recov.c
+++ /dev/null
@@ -1,133 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6recov.c
- *
- * RAID-6 data recovery in dual failure mode.  In single failure mode,
- * use the RAID-5 algorithm (or, in the case of Q failure, just reconstruct
- * the syndrome.)
- */
-
-#include "raid6.h"
-
-/* Recover two failed data blocks. */
-void raid6_2data_recov(int disks, size_t bytes, int faila, int failb,
-		       void **ptrs)
-{
-	u8 *p, *q, *dp, *dq;
-	u8 px, qx, db;
-	const u8 *pbmul;	/* P multiplier table for B data */
-	const u8 *qmul;		/* Q multiplier table (for both) */
-
-	p = (u8 *)ptrs[disks-2];
-	q = (u8 *)ptrs[disks-1];
-
-	/* Compute syndrome with zero for the missing data pages
-	   Use the dead data pages as temporary storage for
-	   delta p and delta q */
-	dp = (u8 *)ptrs[faila];
-	ptrs[faila] = (void *)raid6_empty_zero_page;
-	ptrs[disks-2] = dp;
-	dq = (u8 *)ptrs[failb];
-	ptrs[failb] = (void *)raid6_empty_zero_page;
-	ptrs[disks-1] = dq;
-
-	raid6_call.gen_syndrome(disks, bytes, ptrs);
-
-	/* Restore pointer table */
-	ptrs[faila]   = dp;
-	ptrs[failb]   = dq;
-	ptrs[disks-2] = p;
-	ptrs[disks-1] = q;
-
-	/* Now, pick the proper data tables */
-	pbmul = raid6_gfmul[raid6_gfexi[failb-faila]];
-	qmul  = raid6_gfmul[raid6_gfinv[raid6_gfexp[faila]^raid6_gfexp[failb]]];
-
-	/* Now do it... */
-	while ( bytes-- ) {
-		px    = *p ^ *dp;
-		qx    = qmul[*q ^ *dq];
-		*dq++ = db = pbmul[px] ^ qx; /* Reconstructed B */
-		*dp++ = db ^ px; /* Reconstructed A */
-		p++; q++;
-	}
-}
-
-
-
-
-/* Recover failure of one data block plus the P block */
-void raid6_datap_recov(int disks, size_t bytes, int faila, void **ptrs)
-{
-	u8 *p, *q, *dq;
-	const u8 *qmul;		/* Q multiplier table */
-
-	p = (u8 *)ptrs[disks-2];
-	q = (u8 *)ptrs[disks-1];
-
-	/* Compute syndrome with zero for the missing data page
-	   Use the dead data page as temporary storage for delta q */
-	dq = (u8 *)ptrs[faila];
-	ptrs[faila] = (void *)raid6_empty_zero_page;
-	ptrs[disks-1] = dq;
-
-	raid6_call.gen_syndrome(disks, bytes, ptrs);
-
-	/* Restore pointer table */
-	ptrs[faila]   = dq;
-	ptrs[disks-1] = q;
-
-	/* Now, pick the proper data tables */
-	qmul  = raid6_gfmul[raid6_gfinv[raid6_gfexp[faila]]];
-
-	/* Now do it... */
-	while ( bytes-- ) {
-		*p++ ^= *dq = qmul[*q ^ *dq];
-		q++; dq++;
-	}
-}
-
-
-#ifndef __KERNEL__		/* Testing only */
-
-/* Recover two failed blocks. */
-void raid6_dual_recov(int disks, size_t bytes, int faila, int failb, void **ptrs)
-{
-	if ( faila > failb ) {
-		int tmp = faila;
-		faila = failb;
-		failb = tmp;
-	}
-
-	if ( failb == disks-1 ) {
-		if ( faila == disks-2 ) {
-			/* P+Q failure.  Just rebuild the syndrome. */
-			raid6_call.gen_syndrome(disks, bytes, ptrs);
-		} else {
-			/* data+Q failure.  Reconstruct data from P,
-			   then rebuild syndrome. */
-			/* NOT IMPLEMENTED - equivalent to RAID-5 */
-		}
-	} else {
-		if ( failb == disks-2 ) {
-			/* data+P failure. */
-			raid6_datap_recov(disks, bytes, faila, ptrs);
-		} else {
-			/* data+data failure. */
-			raid6_2data_recov(disks, bytes, faila, failb, ptrs);
-		}
-	}
-}
-
-#endif
diff --git a/drivers/md/raid6sse1.c b/drivers/md/raid6sse1.c
deleted file mode 100644
index 0666237276f..00000000000
--- a/drivers/md/raid6sse1.c
+++ /dev/null
@@ -1,162 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6sse1.c
- *
- * SSE-1/MMXEXT implementation of RAID-6 syndrome functions
- *
- * This is really an MMX implementation, but it requires SSE-1 or
- * AMD MMXEXT for prefetch support and a few other features.  The
- * support for nontemporal memory accesses is enough to make this
- * worthwhile as a separate implementation.
- */
-
-#if defined(__i386__) && !defined(__arch_um__)
-
-#include "raid6.h"
-#include "raid6x86.h"
-
-/* Defined in raid6mmx.c */
-extern const struct raid6_mmx_constants {
-	u64 x1d;
-} raid6_mmx_constants;
-
-static int raid6_have_sse1_or_mmxext(void)
-{
-	/* Not really boot_cpu but "all_cpus" */
-	return boot_cpu_has(X86_FEATURE_MMX) &&
-		(boot_cpu_has(X86_FEATURE_XMM) ||
-		 boot_cpu_has(X86_FEATURE_MMXEXT));
-}
-
-/*
- * Plain SSE1 implementation
- */
-static void raid6_sse11_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
-	asm volatile("pxor %mm5,%mm5");	/* Zero temp */
-
-	for ( d = 0 ; d < bytes ; d += 8 ) {
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0][d]));
-		asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0-1][d]));
-		asm volatile("movq %mm2,%mm4");	/* Q[0] */
-		asm volatile("movq %0,%%mm6" : : "m" (dptr[z0-1][d]));
-		for ( z = z0-2 ; z >= 0 ; z-- ) {
-			asm volatile("prefetchnta %0" : : "m" (dptr[z][d]));
-			asm volatile("pcmpgtb %mm4,%mm5");
-			asm volatile("paddb %mm4,%mm4");
-			asm volatile("pand %mm0,%mm5");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm5,%mm5");
-			asm volatile("pxor %mm6,%mm2");
-			asm volatile("pxor %mm6,%mm4");
-			asm volatile("movq %0,%%mm6" : : "m" (dptr[z][d]));
-		}
-		asm volatile("pcmpgtb %mm4,%mm5");
-		asm volatile("paddb %mm4,%mm4");
-		asm volatile("pand %mm0,%mm5");
-		asm volatile("pxor %mm5,%mm4");
-		asm volatile("pxor %mm5,%mm5");
-		asm volatile("pxor %mm6,%mm2");
-		asm volatile("pxor %mm6,%mm4");
-
-		asm volatile("movntq %%mm2,%0" : "=m" (p[d]));
-		asm volatile("movntq %%mm4,%0" : "=m" (q[d]));
-	}
-
-	asm volatile("sfence" : : : "memory");
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_sse1x1 = {
-	raid6_sse11_gen_syndrome,
-	raid6_have_sse1_or_mmxext,
-	"sse1x1",
-	1			/* Has cache hints */
-};
-
-/*
- * Unrolled-by-2 SSE1 implementation
- */
-static void raid6_sse12_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movq %0,%%mm0" : : "m" (raid6_mmx_constants.x1d));
-	asm volatile("pxor %mm5,%mm5");	/* Zero temp */
-	asm volatile("pxor %mm7,%mm7"); /* Zero temp */
-
-	/* We uniformly assume a single prefetch covers at least 16 bytes */
-	for ( d = 0 ; d < bytes ; d += 16 ) {
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0][d]));
-		asm volatile("movq %0,%%mm2" : : "m" (dptr[z0][d])); /* P[0] */
-		asm volatile("movq %0,%%mm3" : : "m" (dptr[z0][d+8])); /* P[1] */
-		asm volatile("movq %mm2,%mm4");	/* Q[0] */
-		asm volatile("movq %mm3,%mm6"); /* Q[1] */
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			asm volatile("prefetchnta %0" : : "m" (dptr[z][d]));
-			asm volatile("pcmpgtb %mm4,%mm5");
-			asm volatile("pcmpgtb %mm6,%mm7");
-			asm volatile("paddb %mm4,%mm4");
-			asm volatile("paddb %mm6,%mm6");
-			asm volatile("pand %mm0,%mm5");
-			asm volatile("pand %mm0,%mm7");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm7,%mm6");
-			asm volatile("movq %0,%%mm5" : : "m" (dptr[z][d]));
-			asm volatile("movq %0,%%mm7" : : "m" (dptr[z][d+8]));
-			asm volatile("pxor %mm5,%mm2");
-			asm volatile("pxor %mm7,%mm3");
-			asm volatile("pxor %mm5,%mm4");
-			asm volatile("pxor %mm7,%mm6");
-			asm volatile("pxor %mm5,%mm5");
-			asm volatile("pxor %mm7,%mm7");
-		}
-		asm volatile("movntq %%mm2,%0" : "=m" (p[d]));
-		asm volatile("movntq %%mm3,%0" : "=m" (p[d+8]));
-		asm volatile("movntq %%mm4,%0" : "=m" (q[d]));
-		asm volatile("movntq %%mm6,%0" : "=m" (q[d+8]));
-	}
-
-	asm volatile("sfence" : :: "memory");
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_sse1x2 = {
-	raid6_sse12_gen_syndrome,
-	raid6_have_sse1_or_mmxext,
-	"sse1x2",
-	1			/* Has cache hints */
-};
-
-#endif
diff --git a/drivers/md/raid6sse2.c b/drivers/md/raid6sse2.c
deleted file mode 100644
index b034ad86803..00000000000
--- a/drivers/md/raid6sse2.c
+++ /dev/null
@@ -1,262 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6sse2.c
- *
- * SSE-2 implementation of RAID-6 syndrome functions
- *
- */
-
-#if (defined(__i386__) || defined(__x86_64__)) && !defined(__arch_um__)
-
-#include "raid6.h"
-#include "raid6x86.h"
-
-static const struct raid6_sse_constants {
-	u64 x1d[2];
-} raid6_sse_constants  __attribute__((aligned(16))) = {
-	{ 0x1d1d1d1d1d1d1d1dULL, 0x1d1d1d1d1d1d1d1dULL },
-};
-
-static int raid6_have_sse2(void)
-{
-	/* Not really boot_cpu but "all_cpus" */
-	return boot_cpu_has(X86_FEATURE_MMX) &&
-		boot_cpu_has(X86_FEATURE_FXSR) &&
-		boot_cpu_has(X86_FEATURE_XMM) &&
-		boot_cpu_has(X86_FEATURE_XMM2);
-}
-
-/*
- * Plain SSE2 implementation
- */
-static void raid6_sse21_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movdqa %0,%%xmm0" : : "m" (raid6_sse_constants.x1d[0]));
-	asm volatile("pxor %xmm5,%xmm5");	/* Zero temp */
-
-	for ( d = 0 ; d < bytes ; d += 16 ) {
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0][d]));
-		asm volatile("movdqa %0,%%xmm2" : : "m" (dptr[z0][d])); /* P[0] */
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0-1][d]));
-		asm volatile("movdqa %xmm2,%xmm4"); /* Q[0] */
-		asm volatile("movdqa %0,%%xmm6" : : "m" (dptr[z0-1][d]));
-		for ( z = z0-2 ; z >= 0 ; z-- ) {
-			asm volatile("prefetchnta %0" : : "m" (dptr[z][d]));
-			asm volatile("pcmpgtb %xmm4,%xmm5");
-			asm volatile("paddb %xmm4,%xmm4");
-			asm volatile("pand %xmm0,%xmm5");
-			asm volatile("pxor %xmm5,%xmm4");
-			asm volatile("pxor %xmm5,%xmm5");
-			asm volatile("pxor %xmm6,%xmm2");
-			asm volatile("pxor %xmm6,%xmm4");
-			asm volatile("movdqa %0,%%xmm6" : : "m" (dptr[z][d]));
-		}
-		asm volatile("pcmpgtb %xmm4,%xmm5");
-		asm volatile("paddb %xmm4,%xmm4");
-		asm volatile("pand %xmm0,%xmm5");
-		asm volatile("pxor %xmm5,%xmm4");
-		asm volatile("pxor %xmm5,%xmm5");
-		asm volatile("pxor %xmm6,%xmm2");
-		asm volatile("pxor %xmm6,%xmm4");
-
-		asm volatile("movntdq %%xmm2,%0" : "=m" (p[d]));
-		asm volatile("pxor %xmm2,%xmm2");
-		asm volatile("movntdq %%xmm4,%0" : "=m" (q[d]));
-		asm volatile("pxor %xmm4,%xmm4");
-	}
-
-	asm volatile("sfence" : : : "memory");
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_sse2x1 = {
-	raid6_sse21_gen_syndrome,
-	raid6_have_sse2,
-	"sse2x1",
-	1			/* Has cache hints */
-};
-
-/*
- * Unrolled-by-2 SSE2 implementation
- */
-static void raid6_sse22_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movdqa %0,%%xmm0" : : "m" (raid6_sse_constants.x1d[0]));
-	asm volatile("pxor %xmm5,%xmm5"); /* Zero temp */
-	asm volatile("pxor %xmm7,%xmm7"); /* Zero temp */
-
-	/* We uniformly assume a single prefetch covers at least 32 bytes */
-	for ( d = 0 ; d < bytes ; d += 32 ) {
-		asm volatile("prefetchnta %0" : : "m" (dptr[z0][d]));
-		asm volatile("movdqa %0,%%xmm2" : : "m" (dptr[z0][d]));    /* P[0] */
-		asm volatile("movdqa %0,%%xmm3" : : "m" (dptr[z0][d+16])); /* P[1] */
-		asm volatile("movdqa %xmm2,%xmm4"); /* Q[0] */
-		asm volatile("movdqa %xmm3,%xmm6"); /* Q[1] */
-		for ( z = z0-1 ; z >= 0 ; z-- ) {
-			asm volatile("prefetchnta %0" : : "m" (dptr[z][d]));
-			asm volatile("pcmpgtb %xmm4,%xmm5");
-			asm volatile("pcmpgtb %xmm6,%xmm7");
-			asm volatile("paddb %xmm4,%xmm4");
-			asm volatile("paddb %xmm6,%xmm6");
-			asm volatile("pand %xmm0,%xmm5");
-			asm volatile("pand %xmm0,%xmm7");
-			asm volatile("pxor %xmm5,%xmm4");
-			asm volatile("pxor %xmm7,%xmm6");
-			asm volatile("movdqa %0,%%xmm5" : : "m" (dptr[z][d]));
-			asm volatile("movdqa %0,%%xmm7" : : "m" (dptr[z][d+16]));
-			asm volatile("pxor %xmm5,%xmm2");
-			asm volatile("pxor %xmm7,%xmm3");
-			asm volatile("pxor %xmm5,%xmm4");
-			asm volatile("pxor %xmm7,%xmm6");
-			asm volatile("pxor %xmm5,%xmm5");
-			asm volatile("pxor %xmm7,%xmm7");
-		}
-		asm volatile("movntdq %%xmm2,%0" : "=m" (p[d]));
-		asm volatile("movntdq %%xmm3,%0" : "=m" (p[d+16]));
-		asm volatile("movntdq %%xmm4,%0" : "=m" (q[d]));
-		asm volatile("movntdq %%xmm6,%0" : "=m" (q[d+16]));
-	}
-
-	asm volatile("sfence" : : : "memory");
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_sse2x2 = {
-	raid6_sse22_gen_syndrome,
-	raid6_have_sse2,
-	"sse2x2",
-	1			/* Has cache hints */
-};
-
-#endif
-
-#if defined(__x86_64__) && !defined(__arch_um__)
-
-/*
- * Unrolled-by-4 SSE2 implementation
- */
-static void raid6_sse24_gen_syndrome(int disks, size_t bytes, void **ptrs)
-{
-	u8 **dptr = (u8 **)ptrs;
-	u8 *p, *q;
-	int d, z, z0;
-
-	z0 = disks - 3;		/* Highest data disk */
-	p = dptr[z0+1];		/* XOR parity */
-	q = dptr[z0+2];		/* RS syndrome */
-
-	kernel_fpu_begin();
-
-	asm volatile("movdqa %0,%%xmm0" :: "m" (raid6_sse_constants.x1d[0]));
-	asm volatile("pxor %xmm2,%xmm2");	/* P[0] */
-	asm volatile("pxor %xmm3,%xmm3");	/* P[1] */
-	asm volatile("pxor %xmm4,%xmm4"); 	/* Q[0] */
-	asm volatile("pxor %xmm5,%xmm5");	/* Zero temp */
-	asm volatile("pxor %xmm6,%xmm6"); 	/* Q[1] */
-	asm volatile("pxor %xmm7,%xmm7"); 	/* Zero temp */
-	asm volatile("pxor %xmm10,%xmm10");	/* P[2] */
-	asm volatile("pxor %xmm11,%xmm11");	/* P[3] */
-	asm volatile("pxor %xmm12,%xmm12"); 	/* Q[2] */
-	asm volatile("pxor %xmm13,%xmm13");	/* Zero temp */
-	asm volatile("pxor %xmm14,%xmm14"); 	/* Q[3] */
-	asm volatile("pxor %xmm15,%xmm15"); 	/* Zero temp */
-
-	for ( d = 0 ; d < bytes ; d += 64 ) {
-		for ( z = z0 ; z >= 0 ; z-- ) {
-			/* The second prefetch seems to improve performance... */
-			asm volatile("prefetchnta %0" :: "m" (dptr[z][d]));
-			asm volatile("prefetchnta %0" :: "m" (dptr[z][d+32]));
-			asm volatile("pcmpgtb %xmm4,%xmm5");
-			asm volatile("pcmpgtb %xmm6,%xmm7");
-			asm volatile("pcmpgtb %xmm12,%xmm13");
-			asm volatile("pcmpgtb %xmm14,%xmm15");
-			asm volatile("paddb %xmm4,%xmm4");
-			asm volatile("paddb %xmm6,%xmm6");
-			asm volatile("paddb %xmm12,%xmm12");
-			asm volatile("paddb %xmm14,%xmm14");
-			asm volatile("pand %xmm0,%xmm5");
-			asm volatile("pand %xmm0,%xmm7");
-			asm volatile("pand %xmm0,%xmm13");
-			asm volatile("pand %xmm0,%xmm15");
-			asm volatile("pxor %xmm5,%xmm4");
-			asm volatile("pxor %xmm7,%xmm6");
-			asm volatile("pxor %xmm13,%xmm12");
-			asm volatile("pxor %xmm15,%xmm14");
-			asm volatile("movdqa %0,%%xmm5" :: "m" (dptr[z][d]));
-			asm volatile("movdqa %0,%%xmm7" :: "m" (dptr[z][d+16]));
-			asm volatile("movdqa %0,%%xmm13" :: "m" (dptr[z][d+32]));
-			asm volatile("movdqa %0,%%xmm15" :: "m" (dptr[z][d+48]));
-			asm volatile("pxor %xmm5,%xmm2");
-			asm volatile("pxor %xmm7,%xmm3");
-			asm volatile("pxor %xmm13,%xmm10");
-			asm volatile("pxor %xmm15,%xmm11");
-			asm volatile("pxor %xmm5,%xmm4");
-			asm volatile("pxor %xmm7,%xmm6");
-			asm volatile("pxor %xmm13,%xmm12");
-			asm volatile("pxor %xmm15,%xmm14");
-			asm volatile("pxor %xmm5,%xmm5");
-			asm volatile("pxor %xmm7,%xmm7");
-			asm volatile("pxor %xmm13,%xmm13");
-			asm volatile("pxor %xmm15,%xmm15");
-		}
-		asm volatile("movntdq %%xmm2,%0" : "=m" (p[d]));
-		asm volatile("pxor %xmm2,%xmm2");
-		asm volatile("movntdq %%xmm3,%0" : "=m" (p[d+16]));
-		asm volatile("pxor %xmm3,%xmm3");
-		asm volatile("movntdq %%xmm10,%0" : "=m" (p[d+32]));
-		asm volatile("pxor %xmm10,%xmm10");
-		asm volatile("movntdq %%xmm11,%0" : "=m" (p[d+48]));
-		asm volatile("pxor %xmm11,%xmm11");
-		asm volatile("movntdq %%xmm4,%0" : "=m" (q[d]));
-		asm volatile("pxor %xmm4,%xmm4");
-		asm volatile("movntdq %%xmm6,%0" : "=m" (q[d+16]));
-		asm volatile("pxor %xmm6,%xmm6");
-		asm volatile("movntdq %%xmm12,%0" : "=m" (q[d+32]));
-		asm volatile("pxor %xmm12,%xmm12");
-		asm volatile("movntdq %%xmm14,%0" : "=m" (q[d+48]));
-		asm volatile("pxor %xmm14,%xmm14");
-	}
-
-	asm volatile("sfence" : : : "memory");
-	kernel_fpu_end();
-}
-
-const struct raid6_calls raid6_sse2x4 = {
-	raid6_sse24_gen_syndrome,
-	raid6_have_sse2,
-	"sse2x4",
-	1			/* Has cache hints */
-};
-
-#endif
diff --git a/drivers/md/raid6test/Makefile b/drivers/md/raid6test/Makefile
deleted file mode 100644
index 78e0396adf2..00000000000
--- a/drivers/md/raid6test/Makefile
+++ /dev/null
@@ -1,75 +0,0 @@
-#
-# This is a simple Makefile to test some of the RAID-6 code
-# from userspace.
-#
-
-CC	 = gcc
-OPTFLAGS = -O2			# Adjust as desired
-CFLAGS	 = -I.. -g $(OPTFLAGS)
-LD	 = ld
-PERL	 = perl
-AR	 = ar
-RANLIB	 = ranlib
-
-.c.o:
-	$(CC) $(CFLAGS) -c -o $@ $<
-
-%.c: ../%.c
-	cp -f $< $@
-
-%.uc: ../%.uc
-	cp -f $< $@
-
-all:	raid6.a raid6test
-
-raid6.a: raid6int1.o raid6int2.o raid6int4.o raid6int8.o raid6int16.o \
-	 raid6int32.o \
-	 raid6mmx.o raid6sse1.o raid6sse2.o \
-	 raid6altivec1.o raid6altivec2.o raid6altivec4.o raid6altivec8.o \
-	 raid6recov.o raid6algos.o \
-	 raid6tables.o
-	 rm -f $@
-	 $(AR) cq $@ $^
-	 $(RANLIB) $@
-
-raid6test: test.c raid6.a
-	$(CC) $(CFLAGS) -o raid6test $^
-
-raid6altivec1.c: raid6altivec.uc ../unroll.pl
-	$(PERL) ../unroll.pl 1 < raid6altivec.uc > $@
-
-raid6altivec2.c: raid6altivec.uc ../unroll.pl
-	$(PERL) ../unroll.pl 2 < raid6altivec.uc > $@
-
-raid6altivec4.c: raid6altivec.uc ../unroll.pl
-	$(PERL) ../unroll.pl 4 < raid6altivec.uc > $@
-
-raid6altivec8.c: raid6altivec.uc ../unroll.pl
-	$(PERL) ../unroll.pl 8 < raid6altivec.uc > $@
-
-raid6int1.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 1 < raid6int.uc > $@
-
-raid6int2.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 2 < raid6int.uc > $@
-
-raid6int4.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 4 < raid6int.uc > $@
-
-raid6int8.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 8 < raid6int.uc > $@
-
-raid6int16.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 16 < raid6int.uc > $@
-
-raid6int32.c: raid6int.uc ../unroll.pl
-	$(PERL) ../unroll.pl 32 < raid6int.uc > $@
-
-raid6tables.c: mktables
-	./mktables > raid6tables.c
-
-clean:
-	rm -f *.o *.a mktables mktables.c raid6int.uc raid6*.c raid6test
-
-spotless: clean
-	rm -f *~
diff --git a/drivers/md/raid6test/test.c b/drivers/md/raid6test/test.c
deleted file mode 100644
index 559cc41b258..00000000000
--- a/drivers/md/raid6test/test.c
+++ /dev/null
@@ -1,124 +0,0 @@
-/* -*- linux-c -*- ------------------------------------------------------- *
- *
- *   Copyright 2002-2007 H. Peter Anvin - All Rights Reserved
- *
- *   This file is part of the Linux kernel, and is made available under
- *   the terms of the GNU General Public License version 2 or (at your
- *   option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6test.c
- *
- * Test RAID-6 recovery with various algorithms
- */
-
-#include <stdlib.h>
-#include <stdio.h>
-#include <string.h>
-#include "raid6.h"
-
-#define NDISKS		16	/* Including P and Q */
-
-const char raid6_empty_zero_page[PAGE_SIZE] __attribute__((aligned(256)));
-struct raid6_calls raid6_call;
-
-char *dataptrs[NDISKS];
-char data[NDISKS][PAGE_SIZE];
-char recovi[PAGE_SIZE], recovj[PAGE_SIZE];
-
-static void makedata(void)
-{
-	int i, j;
-
-	for (i = 0; i < NDISKS; i++) {
-		for (j = 0; j < PAGE_SIZE; j++)
-			data[i][j] = rand();
-
-		dataptrs[i] = data[i];
-	}
-}
-
-static char disk_type(int d)
-{
-	switch (d) {
-	case NDISKS-2:
-		return 'P';
-	case NDISKS-1:
-		return 'Q';
-	default:
-		return 'D';
-	}
-}
-
-static int test_disks(int i, int j)
-{
-	int erra, errb;
-
-	memset(recovi, 0xf0, PAGE_SIZE);
-	memset(recovj, 0xba, PAGE_SIZE);
-
-	dataptrs[i] = recovi;
-	dataptrs[j] = recovj;
-
-	raid6_dual_recov(NDISKS, PAGE_SIZE, i, j, (void **)&dataptrs);
-
-	erra = memcmp(data[i], recovi, PAGE_SIZE);
-	errb = memcmp(data[j], recovj, PAGE_SIZE);
-
-	if (i < NDISKS-2 && j == NDISKS-1) {
-		/* We don't implement the DQ failure scenario, since it's
-		   equivalent to a RAID-5 failure (XOR, then recompute Q) */
-		erra = errb = 0;
-	} else {
-		printf("algo=%-8s  faila=%3d(%c)  failb=%3d(%c)  %s\n",
-		       raid6_call.name,
-		       i, disk_type(i),
-		       j, disk_type(j),
-		       (!erra && !errb) ? "OK" :
-		       !erra ? "ERRB" :
-		       !errb ? "ERRA" : "ERRAB");
-	}
-
-	dataptrs[i] = data[i];
-	dataptrs[j] = data[j];
-
-	return erra || errb;
-}
-
-int main(int argc, char *argv[])
-{
-	const struct raid6_calls *const *algo;
-	int i, j;
-	int err = 0;
-
-	makedata();
-
-	for (algo = raid6_algos; *algo; algo++) {
-		if (!(*algo)->valid || (*algo)->valid()) {
-			raid6_call = **algo;
-
-			/* Nuke syndromes */
-			memset(data[NDISKS-2], 0xee, 2*PAGE_SIZE);
-
-			/* Generate assumed good syndrome */
-			raid6_call.gen_syndrome(NDISKS, PAGE_SIZE,
-						(void **)&dataptrs);
-
-			for (i = 0; i < NDISKS-1; i++)
-				for (j = i+1; j < NDISKS; j++)
-					err += test_disks(i, j);
-		}
-		printf("\n");
-	}
-
-	printf("\n");
-	/* Pick the best algorithm test */
-	raid6_select_algo();
-
-	if (err)
-		printf("\n*** ERRORS FOUND ***\n");
-
-	return err;
-}
diff --git a/drivers/md/raid6x86.h b/drivers/md/raid6x86.h
deleted file mode 100644
index 99fea7a70ca..00000000000
--- a/drivers/md/raid6x86.h
+++ /dev/null
@@ -1,61 +0,0 @@
-/* ----------------------------------------------------------------------- *
- *
- *   Copyright 2002-2004 H. Peter Anvin - All Rights Reserved
- *
- *   This program is free software; you can redistribute it and/or modify
- *   it under the terms of the GNU General Public License as published by
- *   the Free Software Foundation, Inc., 53 Temple Place Ste 330,
- *   Bostom MA 02111-1307, USA; either version 2 of the License, or
- *   (at your option) any later version; incorporated herein by reference.
- *
- * ----------------------------------------------------------------------- */
-
-/*
- * raid6x86.h
- *
- * Definitions common to x86 and x86-64 RAID-6 code only
- */
-
-#ifndef LINUX_RAID_RAID6X86_H
-#define LINUX_RAID_RAID6X86_H
-
-#if (defined(__i386__) || defined(__x86_64__)) && !defined(__arch_um__)
-
-#ifdef __KERNEL__ /* Real code */
-
-#include <asm/i387.h>
-
-#else /* Dummy code for user space testing */
-
-static inline void kernel_fpu_begin(void)
-{
-}
-
-static inline void kernel_fpu_end(void)
-{
-}
-
-#define X86_FEATURE_MMX		(0*32+23) /* Multimedia Extensions */
-#define X86_FEATURE_FXSR	(0*32+24) /* FXSAVE and FXRSTOR instructions
-					   * (fast save and restore) */
-#define X86_FEATURE_XMM		(0*32+25) /* Streaming SIMD Extensions */
-#define X86_FEATURE_XMM2	(0*32+26) /* Streaming SIMD Extensions-2 */
-#define X86_FEATURE_MMXEXT	(1*32+22) /* AMD MMX extensions */
-
-/* Should work well enough on modern CPUs for testing */
-static inline int boot_cpu_has(int flag)
-{
-	u32 eax = (flag >> 5) ? 0x80000001 : 1;
-	u32 edx;
-
-	asm volatile("cpuid"
-		     : "+a" (eax), "=d" (edx)
-		     : : "ecx", "ebx");
-
-	return (edx >> (flag & 31)) & 1;
-}
-
-#endif /* ndef __KERNEL__ */
-
-#endif
-#endif
diff --git a/drivers/md/unroll.pl b/drivers/md/unroll.pl
deleted file mode 100644
index 3acc710a20e..00000000000
--- a/drivers/md/unroll.pl
+++ /dev/null
@@ -1,24 +0,0 @@
-#!/usr/bin/perl
-#
-# Take a piece of C code and for each line which contains the sequence $$
-# repeat n times with $ replaced by 0...n-1; the sequence $# is replaced
-# by the unrolling factor, and $* with a single $
-#
-
-($n) = @ARGV;
-$n += 0;
-
-while ( defined($line = <STDIN>) ) {
-    if ( $line =~ /\$\$/ ) {
-	$rep = $n;
-    } else {
-	$rep = 1;
-    }
-    for ( $i = 0 ; $i < $rep ; $i++ ) {
-	$tmp = $line;
-	$tmp =~ s/\$\$/$i/g;
-	$tmp =~ s/\$\#/$n/g;
-	$tmp =~ s/\$\*/\$/g;
-	print $tmp;
-    }
-}