Btrfs: RAID5 and RAID6

This builds on David Woodhouse's original Btrfs raid5/6 implementation.
The code has changed quite a bit, blame Chris Mason for any bugs.

Read/modify/write is done after the higher levels of the filesystem have
prepared a given bio.  This means the higher layers are not responsible
for building full stripes, and they don't need to query for the topology
of the extents that may get allocated during delayed allocation runs.
It also means different files can easily share the same stripe.

But, it does expose us to incorrect parity if we crash or lose power
while doing a read/modify/write cycle.  This will be addressed in a
later commit.

Scrub is unable to repair crc errors on raid5/6 chunks.

Discard does not work on raid5/6 (yet)

The stripe size is fixed at 64KiB per disk.  This will be tunable
in a later commit.

Signed-off-by: Chris Mason <chris.mason@fusionio.com>

15 files changed, 2283 insertions, 102 deletions

diff --git a/fs/btrfs/Kconfig b/fs/btrfs/Kconfig
index d33f01c08b60..4f5dc93fa2f8 100644
--- a/fs/btrfs/Kconfig
+++ b/fs/btrfs/Kconfig
@@ -6,6 +6,8 @@ config BTRFS_FS
 	select ZLIB_DEFLATE
 	select LZO_COMPRESS
 	select LZO_DECOMPRESS
+	select RAID6_PQ
+
 	help
 	  Btrfs is a new filesystem with extents, writable snapshotting,
 	  support for multiple devices and many more features.
diff --git a/fs/btrfs/Makefile b/fs/btrfs/Makefile
index 7df3e0f0ee51..3932224f99e9 100644
--- a/fs/btrfs/Makefile
+++ b/fs/btrfs/Makefile
@@ -8,7 +8,7 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
 	   export.o tree-log.o free-space-cache.o zlib.o lzo.o \
 	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
-	   reada.o backref.o ulist.o qgroup.o send.o dev-replace.o
+	   reada.o backref.o ulist.o qgroup.o send.o dev-replace.o raid56.o
 
 btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o
 btrfs-$(CONFIG_BTRFS_FS_CHECK_INTEGRITY) += check-integrity.o
diff --git a/fs/btrfs/ctree.h b/fs/btrfs/ctree.h
index 0ab51be6879f..0cce3aafbd62 100644
--- a/fs/btrfs/ctree.h
+++ b/fs/btrfs/ctree.h
@@ -502,6 +502,7 @@ struct btrfs_super_block {
 #define BTRFS_FEATURE_INCOMPAT_BIG_METADATA	(1ULL << 5)
 
 #define BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF	(1ULL << 6)
+#define BTRFS_FEATURE_INCOMPAT_RAID56		(1ULL << 7)
 
 #define BTRFS_FEATURE_COMPAT_SUPP		0ULL
 #define BTRFS_FEATURE_COMPAT_RO_SUPP		0ULL
@@ -511,6 +512,7 @@ struct btrfs_super_block {
 	 BTRFS_FEATURE_INCOMPAT_MIXED_GROUPS |		\
 	 BTRFS_FEATURE_INCOMPAT_BIG_METADATA |		\
 	 BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO |		\
+	 BTRFS_FEATURE_INCOMPAT_RAID56 |		\
 	 BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)
 
 /*
@@ -952,8 +954,10 @@ struct btrfs_dev_replace_item {
 #define BTRFS_BLOCK_GROUP_RAID1		(1ULL << 4)
 #define BTRFS_BLOCK_GROUP_DUP		(1ULL << 5)
 #define BTRFS_BLOCK_GROUP_RAID10	(1ULL << 6)
+#define BTRFS_BLOCK_GROUP_RAID5    (1 << 7)
+#define BTRFS_BLOCK_GROUP_RAID6    (1 << 8)
 #define BTRFS_BLOCK_GROUP_RESERVED	BTRFS_AVAIL_ALLOC_BIT_SINGLE
-#define BTRFS_NR_RAID_TYPES		5
+#define BTRFS_NR_RAID_TYPES		7
 
 #define BTRFS_BLOCK_GROUP_TYPE_MASK	(BTRFS_BLOCK_GROUP_DATA |    \
 					 BTRFS_BLOCK_GROUP_SYSTEM |  \
@@ -961,6 +965,8 @@ struct btrfs_dev_replace_item {
 
 #define BTRFS_BLOCK_GROUP_PROFILE_MASK	(BTRFS_BLOCK_GROUP_RAID0 |   \
 					 BTRFS_BLOCK_GROUP_RAID1 |   \
+					 BTRFS_BLOCK_GROUP_RAID5 |   \
+					 BTRFS_BLOCK_GROUP_RAID6 |   \
 					 BTRFS_BLOCK_GROUP_DUP |     \
 					 BTRFS_BLOCK_GROUP_RAID10)
 /*
@@ -1185,6 +1191,10 @@ struct btrfs_block_group_cache {
 	u64 flags;
 	u64 sectorsize;
 	u64 cache_generation;
+
+	/* for raid56, this is a full stripe, without parity */
+	unsigned long full_stripe_len;
+
 	unsigned int ro:1;
 	unsigned int dirty:1;
 	unsigned int iref:1;
@@ -1225,6 +1235,20 @@ struct seq_list {
 	u64 seq;
 };
 
+/* used by the raid56 code to lock stripes for read/modify/write */
+struct btrfs_stripe_hash {
+	struct list_head hash_list;
+	wait_queue_head_t wait;
+	spinlock_t lock;
+};
+
+/* used by the raid56 code to lock stripes for read/modify/write */
+struct btrfs_stripe_hash_table {
+	struct btrfs_stripe_hash *table;
+};
+
+#define BTRFS_STRIPE_HASH_TABLE_BITS 11
+
 /* fs_info */
 struct reloc_control;
 struct btrfs_device;
@@ -1307,6 +1331,13 @@ struct btrfs_fs_info {
 	struct mutex cleaner_mutex;
 	struct mutex chunk_mutex;
 	struct mutex volume_mutex;
+
+	/* this is used during read/modify/write to make sure
+	 * no two ios are trying to mod the same stripe at the same
+	 * time
+	 */
+	struct btrfs_stripe_hash_table *stripe_hash_table;
+
 	/*
 	 * this protects the ordered operations list only while we are
 	 * processing all of the entries on it.  This way we make
@@ -1395,6 +1426,8 @@ struct btrfs_fs_info {
 	struct btrfs_workers flush_workers;
 	struct btrfs_workers endio_workers;
 	struct btrfs_workers endio_meta_workers;
+	struct btrfs_workers endio_raid56_workers;
+	struct btrfs_workers rmw_workers;
 	struct btrfs_workers endio_meta_write_workers;
 	struct btrfs_workers endio_write_workers;
 	struct btrfs_workers endio_freespace_worker;
diff --git a/fs/btrfs/disk-io.c b/fs/btrfs/disk-io.c
index 65f03670a952..e9fa7b4d18e3 100644
--- a/fs/btrfs/disk-io.c
+++ b/fs/btrfs/disk-io.c
@@ -46,6 +46,7 @@
 #include "check-integrity.h"
 #include "rcu-string.h"
 #include "dev-replace.h"
+#include "raid56.h"
 
 #ifdef CONFIG_X86
 #include <asm/cpufeature.h>
@@ -639,8 +640,15 @@ err:
 		btree_readahead_hook(root, eb, eb->start, ret);
 	}
 
-	if (ret)
+	if (ret) {
+		/*
+		 * our io error hook is going to dec the io pages
+		 * again, we have to make sure it has something
+		 * to decrement
+		 */
+		atomic_inc(&eb->io_pages);
 		clear_extent_buffer_uptodate(eb);
+	}
 	free_extent_buffer(eb);
 out:
 	return ret;
@@ -654,6 +662,7 @@ static int btree_io_failed_hook(struct page *page, int failed_mirror)
 	eb = (struct extent_buffer *)page->private;
 	set_bit(EXTENT_BUFFER_IOERR, &eb->bflags);
 	eb->read_mirror = failed_mirror;
+	atomic_dec(&eb->io_pages);
 	if (test_and_clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags))
 		btree_readahead_hook(root, eb, eb->start, -EIO);
 	return -EIO;	/* we fixed nothing */
@@ -670,17 +679,23 @@ static void end_workqueue_bio(struct bio *bio, int err)
 	end_io_wq->work.flags = 0;
 
 	if (bio->bi_rw & REQ_WRITE) {
-		if (end_io_wq->metadata == 1)
+		if (end_io_wq->metadata == BTRFS_WQ_ENDIO_METADATA)
 			btrfs_queue_worker(&fs_info->endio_meta_write_workers,
 					   &end_io_wq->work);
-		else if (end_io_wq->metadata == 2)
+		else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_FREE_SPACE)
 			btrfs_queue_worker(&fs_info->endio_freespace_worker,
 					   &end_io_wq->work);
+		else if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
+			btrfs_queue_worker(&fs_info->endio_raid56_workers,
+					   &end_io_wq->work);
 		else
 			btrfs_queue_worker(&fs_info->endio_write_workers,
 					   &end_io_wq->work);
 	} else {
-		if (end_io_wq->metadata)
+		if (end_io_wq->metadata == BTRFS_WQ_ENDIO_RAID56)
+			btrfs_queue_worker(&fs_info->endio_raid56_workers,
+					   &end_io_wq->work);
+		else if (end_io_wq->metadata)
 			btrfs_queue_worker(&fs_info->endio_meta_workers,
 					   &end_io_wq->work);
 		else
@@ -695,6 +710,7 @@ static void end_workqueue_bio(struct bio *bio, int err)
  * 0 - if data
  * 1 - if normal metadta
  * 2 - if writing to the free space cache area
+ * 3 - raid parity work
  */
 int btrfs_bio_wq_end_io(struct btrfs_fs_info *info, struct bio *bio,
 			int metadata)
@@ -2165,6 +2181,12 @@ int open_ctree(struct super_block *sb,
 	init_waitqueue_head(&fs_info->transaction_blocked_wait);
 	init_waitqueue_head(&fs_info->async_submit_wait);
 
+	ret = btrfs_alloc_stripe_hash_table(fs_info);
+	if (ret) {
+		err = -ENOMEM;
+		goto fail_alloc;
+	}
+
 	__setup_root(4096, 4096, 4096, 4096, tree_root,
 		     fs_info, BTRFS_ROOT_TREE_OBJECTID);
 
@@ -2332,6 +2354,12 @@ int open_ctree(struct super_block *sb,
 	btrfs_init_workers(&fs_info->endio_meta_write_workers,
 			   "endio-meta-write", fs_info->thread_pool_size,
 			   &fs_info->generic_worker);
+	btrfs_init_workers(&fs_info->endio_raid56_workers,
+			   "endio-raid56", fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
+	btrfs_init_workers(&fs_info->rmw_workers,
+			   "rmw", fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 	btrfs_init_workers(&fs_info->endio_write_workers, "endio-write",
 			   fs_info->thread_pool_size,
 			   &fs_info->generic_worker);
@@ -2350,6 +2378,8 @@ int open_ctree(struct super_block *sb,
 	 */
 	fs_info->endio_workers.idle_thresh = 4;
 	fs_info->endio_meta_workers.idle_thresh = 4;
+	fs_info->endio_raid56_workers.idle_thresh = 4;
+	fs_info->rmw_workers.idle_thresh = 2;
 
 	fs_info->endio_write_workers.idle_thresh = 2;
 	fs_info->endio_meta_write_workers.idle_thresh = 2;
@@ -2366,6 +2396,8 @@ int open_ctree(struct super_block *sb,
 	ret |= btrfs_start_workers(&fs_info->fixup_workers);
 	ret |= btrfs_start_workers(&fs_info->endio_workers);
 	ret |= btrfs_start_workers(&fs_info->endio_meta_workers);
+	ret |= btrfs_start_workers(&fs_info->rmw_workers);
+	ret |= btrfs_start_workers(&fs_info->endio_raid56_workers);
 	ret |= btrfs_start_workers(&fs_info->endio_meta_write_workers);
 	ret |= btrfs_start_workers(&fs_info->endio_write_workers);
 	ret |= btrfs_start_workers(&fs_info->endio_freespace_worker);
@@ -2710,6 +2742,8 @@ fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->workers);
 	btrfs_stop_workers(&fs_info->endio_workers);
 	btrfs_stop_workers(&fs_info->endio_meta_workers);
+	btrfs_stop_workers(&fs_info->endio_raid56_workers);
+	btrfs_stop_workers(&fs_info->rmw_workers);
 	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
 	btrfs_stop_workers(&fs_info->endio_write_workers);
 	btrfs_stop_workers(&fs_info->endio_freespace_worker);
@@ -2728,6 +2762,7 @@ fail_bdi:
 fail_srcu:
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 fail:
+	btrfs_free_stripe_hash_table(fs_info);
 	btrfs_close_devices(fs_info->fs_devices);
 	return err;
 
@@ -3076,11 +3111,16 @@ int btrfs_calc_num_tolerated_disk_barrier_failures(
 				     ((flags & BTRFS_BLOCK_GROUP_PROFILE_MASK)
 				      == 0)))
 					num_tolerated_disk_barrier_failures = 0;
-				else if (num_tolerated_disk_barrier_failures > 1
-					 &&
-					 (flags & (BTRFS_BLOCK_GROUP_RAID1 |
-						   BTRFS_BLOCK_GROUP_RAID10)))
-					num_tolerated_disk_barrier_failures = 1;
+				else if (num_tolerated_disk_barrier_failures > 1) {
+					if (flags & (BTRFS_BLOCK_GROUP_RAID1 |
+					    BTRFS_BLOCK_GROUP_RAID5 |
+					    BTRFS_BLOCK_GROUP_RAID10)) {
+						num_tolerated_disk_barrier_failures = 1;
+					} else if (flags &
+						   BTRFS_BLOCK_GROUP_RAID5) {
+						num_tolerated_disk_barrier_failures = 2;
+					}
+				}
 			}
 		}
 		up_read(&sinfo->groups_sem);
@@ -3384,6 +3424,8 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_stop_workers(&fs_info->workers);
 	btrfs_stop_workers(&fs_info->endio_workers);
 	btrfs_stop_workers(&fs_info->endio_meta_workers);
+	btrfs_stop_workers(&fs_info->endio_raid56_workers);
+	btrfs_stop_workers(&fs_info->rmw_workers);
 	btrfs_stop_workers(&fs_info->endio_meta_write_workers);
 	btrfs_stop_workers(&fs_info->endio_write_workers);
 	btrfs_stop_workers(&fs_info->endio_freespace_worker);
@@ -3404,6 +3446,8 @@ int close_ctree(struct btrfs_root *root)
 	bdi_destroy(&fs_info->bdi);
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 
+	btrfs_free_stripe_hash_table(fs_info);
+
 	return 0;
 }
 
diff --git a/fs/btrfs/disk-io.h b/fs/btrfs/disk-io.h
index 305c33efb0e3..034d7dc552b2 100644
--- a/fs/btrfs/disk-io.h
+++ b/fs/btrfs/disk-io.h
@@ -25,6 +25,13 @@
 #define BTRFS_SUPER_MIRROR_MAX	 3
 #define BTRFS_SUPER_MIRROR_SHIFT 12
 
+enum {
+	BTRFS_WQ_ENDIO_DATA = 0,
+	BTRFS_WQ_ENDIO_METADATA = 1,
+	BTRFS_WQ_ENDIO_FREE_SPACE = 2,
+	BTRFS_WQ_ENDIO_RAID56 = 3,
+};
+
 static inline u64 btrfs_sb_offset(int mirror)
 {
 	u64 start = 16 * 1024;
diff --git a/fs/btrfs/extent-tree.c b/fs/btrfs/extent-tree.c
index d133edfcd449..3345f68fc64b 100644
--- a/fs/btrfs/extent-tree.c
+++ b/fs/btrfs/extent-tree.c
@@ -31,6 +31,7 @@
 #include "print-tree.h"
 #include "transaction.h"
 #include "volumes.h"
+#include "raid56.h"
 #include "locking.h"
 #include "free-space-cache.h"
 #include "math.h"
@@ -1852,6 +1853,8 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
 		*actual_bytes = discarded_bytes;
 
 
+	if (ret == -EOPNOTSUPP)
+		ret = 0;
 	return ret;
 }
 
@@ -3276,6 +3279,7 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
 	u64 num_devices = root->fs_info->fs_devices->rw_devices +
 		root->fs_info->fs_devices->missing_devices;
 	u64 target;
+	u64 tmp;
 
 	/*
 	 * see if restripe for this chunk_type is in progress, if so
@@ -3292,30 +3296,32 @@ u64 btrfs_reduce_alloc_profile(struct btrfs_root *root, u64 flags)
 	}
 	spin_unlock(&root->fs_info->balance_lock);
 
+	/* First, mask out the RAID levels which aren't possible */
 	if (num_devices == 1)
-		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0);
+		flags &= ~(BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID0 |
+			   BTRFS_BLOCK_GROUP_RAID5);
+	if (num_devices < 3)
+		flags &= ~BTRFS_BLOCK_GROUP_RAID6;
 	if (num_devices < 4)
 		flags &= ~BTRFS_BLOCK_GROUP_RAID10;
 
-	if ((flags & BTRFS_BLOCK_GROUP_DUP) &&
-	    (flags & (BTRFS_BLOCK_GROUP_RAID1 |
-		      BTRFS_BLOCK_GROUP_RAID10))) {
-		flags &= ~BTRFS_BLOCK_GROUP_DUP;
-	}
-
-	if ((flags & BTRFS_BLOCK_GROUP_RAID1) &&
-	    (flags & BTRFS_BLOCK_GROUP_RAID10)) {
-		flags &= ~BTRFS_BLOCK_GROUP_RAID1;
-	}
+	tmp = flags & (BTRFS_BLOCK_GROUP_DUP | BTRFS_BLOCK_GROUP_RAID0 |
+		       BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID5 |
+		       BTRFS_BLOCK_GROUP_RAID6 | BTRFS_BLOCK_GROUP_RAID10);
+	flags &= ~tmp;
 
-	if ((flags & BTRFS_BLOCK_GROUP_RAID0) &&
-	    ((flags & BTRFS_BLOCK_GROUP_RAID1) |
-	     (flags & BTRFS_BLOCK_GROUP_RAID10) |
-	     (flags & BTRFS_BLOCK_GROUP_DUP))) {
-		flags &= ~BTRFS_BLOCK_GROUP_RAID0;
-	}
+	if (tmp & BTRFS_BLOCK_GROUP_RAID6)
+		tmp = BTRFS_BLOCK_GROUP_RAID6;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID5)
+		tmp = BTRFS_BLOCK_GROUP_RAID5;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID10)
+		tmp = BTRFS_BLOCK_GROUP_RAID10;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID1)
+		tmp = BTRFS_BLOCK_GROUP_RAID1;
+	else if (tmp & BTRFS_BLOCK_GROUP_RAID0)
+		tmp = BTRFS_BLOCK_GROUP_RAID0;
 
-	return extended_to_chunk(flags);
+	return extended_to_chunk(flags | tmp);
 }
 
 static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
@@ -3333,6 +3339,7 @@ static u64 get_alloc_profile(struct btrfs_root *root, u64 flags)
 u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
 {
 	u64 flags;
+	u64 ret;
 
 	if (data)
 		flags = BTRFS_BLOCK_GROUP_DATA;
@@ -3341,7 +3348,8 @@ u64 btrfs_get_alloc_profile(struct btrfs_root *root, int data)
 	else
 		flags = BTRFS_BLOCK_GROUP_METADATA;
 
-	return get_alloc_profile(root, flags);
+	ret = get_alloc_profile(root, flags);
+	return ret;
 }
 
 /*
@@ -3516,8 +3524,10 @@ static u64 get_system_chunk_thresh(struct btrfs_root *root, u64 type)
 {
 	u64 num_dev;
 
-	if (type & BTRFS_BLOCK_GROUP_RAID10 ||
-	    type & BTRFS_BLOCK_GROUP_RAID0)
+	if (type & (BTRFS_BLOCK_GROUP_RAID10 |
+		    BTRFS_BLOCK_GROUP_RAID0 |
+		    BTRFS_BLOCK_GROUP_RAID5 |
+		    BTRFS_BLOCK_GROUP_RAID6))
 		num_dev = root->fs_info->fs_devices->rw_devices;
 	else if (type & BTRFS_BLOCK_GROUP_RAID1)
 		num_dev = 2;
@@ -3667,7 +3677,9 @@ static int can_overcommit(struct btrfs_root *root,
 
 	/*
 	 * If we have dup, raid1 or raid10 then only half of the free
-	 * space is actually useable.
+	 * space is actually useable.  For raid56, the space info used
+	 * doesn't include the parity drive, so we don't have to
+	 * change the math
 	 */
 	if (profile & (BTRFS_BLOCK_GROUP_DUP |
 		       BTRFS_BLOCK_GROUP_RAID1 |
@@ -5455,10 +5467,14 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans, struct btrfs_root *root,
 	return ret;
 }
 
-static u64 stripe_align(struct btrfs_root *root, u64 val)
+static u64 stripe_align(struct btrfs_root *root,
+			struct btrfs_block_group_cache *cache,
+			u64 val, u64 num_bytes)
 {
-	u64 mask = ((u64)root->stripesize - 1);
-	u64 ret = (val + mask) & ~mask;
+	u64 mask;
+	u64 ret;
+	mask = ((u64)root->stripesize - 1);
+	ret = (val + mask) & ~mask;
 	return ret;
 }
 
@@ -5519,9 +5535,12 @@ int __get_raid_index(u64 flags)
 		index = 2;
 	else if (flags & BTRFS_BLOCK_GROUP_RAID0)
 		index = 3;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID5)
+		index = 5;
+	else if (flags & BTRFS_BLOCK_GROUP_RAID6)
+		index = 6;
 	else
-		index = 4;
-
+		index = 4; /* BTRFS_BLOCK_GROUP_SINGLE */
 	return index;
 }
 
@@ -5665,6 +5684,8 @@ search:
 		if (!block_group_bits(block_group, data)) {
 		    u64 extra = BTRFS_BLOCK_GROUP_DUP |
 				BTRFS_BLOCK_GROUP_RAID1 |
+				BTRFS_BLOCK_GROUP_RAID5 |
+				BTRFS_BLOCK_GROUP_RAID6 |
 				BTRFS_BLOCK_GROUP_RAID10;
 
 			/*
@@ -5835,7 +5856,8 @@ unclustered_alloc:
 			goto loop;
 		}
 checks:
-		search_start = stripe_align(root, offset);
+		search_start = stripe_align(root, used_block_group,
+					    offset, num_bytes);
 
 		/* move on to the next group */
 		if (search_start + num_bytes >
@@ -7203,6 +7225,7 @@ static u64 update_block_group_flags(struct btrfs_root *root, u64 flags)
 		root->fs_info->fs_devices->missing_devices;
 
 	stripped = BTRFS_BLOCK_GROUP_RAID0 |
+		BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6 |
 		BTRFS_BLOCK_GROUP_RAID1 | BTRFS_BLOCK_GROUP_RAID10;
 
 	if (num_devices == 1) {
@@ -7754,7 +7777,9 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		btrfs_release_path(path);
 		cache->flags = btrfs_block_group_flags(&cache->item);
 		cache->sectorsize = root->sectorsize;
-
+		cache->full_stripe_len = btrfs_full_stripe_len(root,
+					       &root->fs_info->mapping_tree,
+					       found_key.objectid);
 		btrfs_init_free_space_ctl(cache);
 
 		/*
@@ -7808,6 +7833,8 @@ int btrfs_read_block_groups(struct btrfs_root *root)
 		if (!(get_alloc_profile(root, space_info->flags) &
 		      (BTRFS_BLOCK_GROUP_RAID10 |
 		       BTRFS_BLOCK_GROUP_RAID1 |
+		       BTRFS_BLOCK_GROUP_RAID5 |
+		       BTRFS_BLOCK_GROUP_RAID6 |
 		       BTRFS_BLOCK_GROUP_DUP)))
 			continue;
 		/*
@@ -7883,6 +7910,9 @@ int btrfs_make_block_group(struct btrfs_trans_handle *trans,
 	cache->key.type = BTRFS_BLOCK_GROUP_ITEM_KEY;
 	cache->sectorsize = root->sectorsize;
 	cache->fs_info = root->fs_info;
+	cache->full_stripe_len = btrfs_full_stripe_len(root,
+					       &root->fs_info->mapping_tree,
+					       chunk_offset);
 
 	atomic_set(&cache->count, 1);
 	spin_lock_init(&cache->lock);
diff --git a/fs/btrfs/extent_io.c b/fs/btrfs/extent_io.c
index 86ecca48c604..3b9fb478b0d1 100644
--- a/fs/btrfs/extent_io.c
+++ b/fs/btrfs/extent_io.c
@@ -1895,13 +1895,11 @@ static int free_io_failure(struct inode *inode, struct io_failure_record *rec,
 	if (ret)
 		err = ret;
 
-	if (did_repair) {
-		ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
-					rec->start + rec->len - 1,
-					EXTENT_DAMAGED, GFP_NOFS);
-		if (ret && !err)
-			err = ret;
-	}
+	ret = clear_extent_bits(&BTRFS_I(inode)->io_tree, rec->start,
+				rec->start + rec->len - 1,
+				EXTENT_DAMAGED, GFP_NOFS);
+	if (ret && !err)
+		err = ret;
 
 	kfree(rec);
 	return err;
@@ -1932,10 +1930,15 @@ int repair_io_failure(struct btrfs_fs_info *fs_info, u64 start,
 	u64 map_length = 0;
 	u64 sector;
 	struct btrfs_bio *bbio = NULL;
+	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
 	int ret;
 
 	BUG_ON(!mirror_num);
 
+	/* we can't repair anything in raid56 yet */
+	if (btrfs_is_parity_mirror(map_tree, logical, length, mirror_num))
+		return 0;
+
 	bio = bio_alloc(GFP_NOFS, 1);
 	if (!bio)
 		return -EIO;
@@ -2052,6 +2055,7 @@ static int clean_io_failure(u64 start, struct page *page)
 						failrec->failed_mirror);
 			did_repair = !ret;
 		}
+		ret = 0;
 	}
 
 out:
diff --git a/fs/btrfs/free-space-cache.c b/fs/btrfs/free-space-cache.c
index 59ea2e4349c9..62020b7f7036 100644
--- a/fs/btrfs/free-space-cache.c
+++ b/fs/btrfs/free-space-cache.c
@@ -1463,10 +1463,14 @@ static int search_bitmap(struct btrfs_free_space_ctl *ctl,
 }
 
 static struct btrfs_free_space *
-find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes)
+find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes,
+		unsigned long align)
 {
 	struct btrfs_free_space *entry;
 	struct rb_node *node;
+	u64 ctl_off;
+	u64 tmp;
+	u64 align_off;
 	int ret;
 
 	if (!ctl->free_space_offset.rb_node)
@@ -1481,15 +1485,34 @@ find_free_space(struct btrfs_free_space_ctl *ctl, u64 *offset, u64 *bytes)
 		if (entry->bytes < *bytes)
 			continue;
 
+		/* make sure the space returned is big enough
+		 * to match our requested alignment
+		 */
+		if (*bytes >= align) {
+			ctl_off = entry->offset - ctl->start;
+			tmp = ctl_off + align - 1;;
+			do_div(tmp, align);
+			tmp = tmp * align + ctl->start;
+			align_off = tmp - entry->offset;
+		} else {
+			align_off = 0;
+			tmp = entry->offset;
+		}
+
+		if (entry->bytes < *bytes + align_off)
+			continue;
+
 		if (entry->bitmap) {
-			ret = search_bitmap(ctl, entry, offset, bytes);
-			if (!ret)
+			ret = search_bitmap(ctl, entry, &tmp, bytes);
+			if (!ret) {
+				*offset = tmp;
 				return entry;
+			}
 			continue;
 		}
 
-		*offset = entry->offset;
-		*bytes = entry->bytes;
+		*offset = tmp;
+		*bytes = entry->bytes - align_off;
 		return entry;
 	}
 
@@ -2091,9 +2114,12 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
 	struct btrfs_free_space *entry = NULL;
 	u64 bytes_search = bytes + empty_size;
 	u64 ret = 0;
+	u64 align_gap = 0;
+	u64 align_gap_len = 0;
 
 	spin_lock(&ctl->tree_lock);
-	entry = find_free_space(ctl, &offset, &bytes_search);
+	entry = find_free_space(ctl, &offset, &bytes_search,
+				block_group->full_stripe_len);
 	if (!entry)
 		goto out;
 
@@ -2103,9 +2129,15 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
 		if (!entry->bytes)
 			free_bitmap(ctl, entry);
 	} else {
+
 		unlink_free_space(ctl, entry);
-		entry->offset += bytes;
-		entry->bytes -= bytes;
+		align_gap_len = offset - entry->offset;
+		align_gap = entry->offset;
+
+		entry->offset = offset + bytes;
+		WARN_ON(entry->bytes < bytes + align_gap_len);
+
+		entry->bytes -= bytes + align_gap_len;
 		if (!entry->bytes)
 			kmem_cache_free(btrfs_free_space_cachep, entry);
 		else
@@ -2115,6 +2147,8 @@ u64 btrfs_find_space_for_alloc(struct btrfs_block_group_cache *block_group,
 out:
 	spin_unlock(&ctl->tree_lock);
 
+	if (align_gap_len)
+		__btrfs_add_free_space(ctl, align_gap, align_gap_len);
 	return ret;
 }
 
diff --git a/fs/btrfs/inode.c b/fs/btrfs/inode.c
index 1b98c4ce3c6f..6f4e41dca970 100644
--- a/fs/btrfs/inode.c
+++ b/fs/btrfs/inode.c
@@ -39,6 +39,7 @@
 #include <linux/slab.h>
 #include <linux/ratelimit.h>
 #include <linux/mount.h>
+#include <linux/blkdev.h>
 #include "compat.h"
 #include "ctree.h"
 #include "disk-io.h"
@@ -6386,19 +6387,24 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 	int async_submit = 0;
 
 	map_length = orig_bio->bi_size;
-	ret = btrfs_map_block(root->fs_info, READ, start_sector << 9,
+	ret = btrfs_map_block(root->fs_info, rw, start_sector << 9,
 			      &map_length, NULL, 0);
 	if (ret) {
 		bio_put(orig_bio);
 		return -EIO;
 	}
-
 	if (map_length >= orig_bio->bi_size) {
 		bio = orig_bio;
 		goto submit;
 	}
 
-	async_submit = 1;
+	/* async crcs make it difficult to collect full stripe writes. */
+	if (btrfs_get_alloc_profile(root, 1) &
+	    (BTRFS_BLOCK_GROUP_RAID5 | BTRFS_BLOCK_GROUP_RAID6))
+		async_submit = 0;
+	else
+		async_submit = 1;
+
 	bio = btrfs_dio_bio_alloc(orig_bio->bi_bdev, start_sector, GFP_NOFS);
 	if (!bio)
 		return -ENOMEM;
@@ -6440,7 +6446,7 @@ static int btrfs_submit_direct_hook(int rw, struct btrfs_dio_private *dip,
 			bio->bi_end_io = btrfs_end_dio_bio;
 
 			map_length = orig_bio->bi_size;
-			ret = btrfs_map_block(root->fs_info, READ,
+			ret = btrfs_map_block(root->fs_info, rw,
 					      start_sector << 9,
 					      &map_length, NULL, 0);
 			if (ret) {
@@ -6583,15 +6589,17 @@ static ssize_t btrfs_direct_IO(int rw, struct kiocb *iocb,
 {
 	struct file *file = iocb->ki_filp;
 	struct inode *inode = file->f_mapping->host;
+	ssize_t ret;
 
 	if (check_direct_IO(BTRFS_I(inode)->root, rw, iocb, iov,
 			    offset, nr_segs))
 		return 0;
 
-	return __blockdev_direct_IO(rw, iocb, inode,
+	ret = __blockdev_direct_IO(rw, iocb, inode,
 		   BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev,
 		   iov, offset, nr_segs, btrfs_get_blocks_direct, NULL,
 		   btrfs_submit_direct, 0);
+	return ret;
 }
 
 #define BTRFS_FIEMAP_FLAGS	(FIEMAP_FLAG_SYNC)
diff --git a/fs/btrfs/raid56.c b/fs/btrfs/raid56.c
new file mode 100644
index 000000000000..d02510f34936
--- /dev/null
+++ b/fs/btrfs/raid56.c
@@ -0,0 +1,1647 @@
+/*
+ * Copyright (C) 2012 Fusion-io  All rights reserved.
+ * Copyright (C) 2012 Intel Corp. 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 v2 as published by the Free Software Foundation.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
+ * General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public
+ * License along with this program; if not, write to the
+ * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
+ * Boston, MA 021110-1307, USA.
+ */
+#include <linux/sched.h>
+#include <linux/wait.h>
+#include <linux/bio.h>
+#include <linux/slab.h>
+#include <linux/buffer_head.h>
+#include <linux/blkdev.h>
+#include <linux/random.h>
+#include <linux/iocontext.h>
+#include <linux/capability.h>
+#include <linux/ratelimit.h>
+#include <linux/kthread.h>
+#include <linux/raid/pq.h>
+#include <linux/hash.h>
+#include <linux/list_sort.h>
+#include <linux/raid/xor.h>
+#include <asm/div64.h>
+#include "compat.h"
+#include "ctree.h"
+#include "extent_map.h"
+#include "disk-io.h"
+#include "transaction.h"
+#include "print-tree.h"
+#include "volumes.h"
+#include "raid56.h"
+#include "async-thread.h"
+#include "check-integrity.h"
+#include "rcu-string.h"
+
+/* set when additional merges to this rbio are not allowed */
+#define RBIO_RMW_LOCKED_BIT	1
+
+struct btrfs_raid_bio {
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_bio *bbio;
+
+	/*
+	 * logical block numbers for the start of each stripe
+	 * The last one or two are p/q.  These are sorted,
+	 * so raid_map[0] is the start of our full stripe
+	 */
+	u64 *raid_map;
+
+	/* while we're doing rmw on a stripe
+	 * we put it into a hash table so we can
+	 * lock the stripe and merge more rbios
+	 * into it.
+	 */
+	struct list_head hash_list;
+
+	/*
+	 * for scheduling work in the helper threads
+	 */
+	struct btrfs_work work;
+
+	/*
+	 * bio list and bio_list_lock are used
+	 * to add more bios into the stripe
+	 * in hopes of avoiding the full rmw
+	 */
+	struct bio_list bio_list;
+	spinlock_t bio_list_lock;
+
+	/*
+	 * also protected by the bio_list_lock, the
+	 * stripe locking code uses plug_list to hand off
+	 * the stripe lock to the next pending IO
+	 */
+	struct list_head plug_list;
+
+	/*
+	 * flags that tell us if it is safe to
+	 * merge with this bio
+	 */
+	unsigned long flags;
+
+	/* size of each individual stripe on disk */
+	int stripe_len;
+
+	/* number of data stripes (no p/q) */
+	int nr_data;
+
+	/*
+	 * set if we're doing a parity rebuild
+	 * for a read from higher up, which is handled
+	 * differently from a parity rebuild as part of
+	 * rmw
+	 */
+	int read_rebuild;
+
+	/* first bad stripe */
+	int faila;
+
+	/* second bad stripe (for raid6 use) */
+	int failb;
+
+	/*
+	 * number of pages needed to represent the full
+	 * stripe
+	 */
+	int nr_pages;
+
+	/*
+	 * size of all the bios in the bio_list.  This
+	 * helps us decide if the rbio maps to a full
+	 * stripe or not
+	 */
+	int bio_list_bytes;
+
+	atomic_t refs;
+
+	/*
+	 * these are two arrays of pointers.  We allocate the
+	 * rbio big enough to hold them both and setup their
+	 * locations when the rbio is allocated
+	 */
+
+	/* pointers to pages that we allocated for
+	 * reading/writing stripes directly from the disk (including P/Q)
+	 */
+	struct page **stripe_pages;
+
+	/*
+	 * pointers to the pages in the bio_list.  Stored
+	 * here for faster lookup
+	 */
+	struct page **bio_pages;
+};
+
+static int __raid56_parity_recove