Merge branch 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs

* 'for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/mason/linux-btrfs: (114 commits)
  Btrfs: check for a null fs root when writing to the backup root log
  Btrfs: fix race during transaction joins
  Btrfs: fix a potential btrfs_bio leak on scrub fixups
  Btrfs: rename btrfs_bio multi -> bbio for consistency
  Btrfs: stop leaking btrfs_bios on readahead
  Btrfs: stop the readahead threads on failed mount
  Btrfs: fix extent_buffer leak in the metadata IO error handling
  Btrfs: fix the new inspection ioctls for 32 bit compat
  Btrfs: fix delayed insertion reservation
  Btrfs: ClearPageError during writepage and clean_tree_block
  Btrfs: be smarter about committing the transaction in reserve_metadata_bytes
  Btrfs: make a delayed_block_rsv for the delayed item insertion
  Btrfs: add a log of past tree roots
  btrfs: separate superblock items out of fs_info
  Btrfs: use the global reserve when truncating the free space cache inode
  Btrfs: release metadata from global reserve if we have to fallback for unlink
  Btrfs: make sure to flush queued bios if write_cache_pages waits
  Btrfs: fix extent pinning bugs in the tree log
  Btrfs: make sure btrfs_remove_free_space doesn't leak EAGAIN
  Btrfs: don't wait as long for more batches during SSD log commit
  ...

fs/btrfs/Makefile

@@ -7,6 +7,7 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   extent_map.o sysfs.o struct-funcs.o xattr.o ordered-data.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
+	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
+	   reada.o backref.o
 
 btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o

fs/btrfs/acl.c

@@ -59,22 +59,19 @@ struct posix_acl *btrfs_get_acl(struct inode *inode, int type)
 		if (!value)
 			return ERR_PTR(-ENOMEM);
 		size = __btrfs_getxattr(inode, name, value, size);
-		if (size > 0) {
-			acl = posix_acl_from_xattr(value, size);
-			if (IS_ERR(acl)) {
-				kfree(value);
-				return acl;
-			}
-			set_cached_acl(inode, type, acl);
-		}
-		kfree(value);
+	}
+	if (size > 0) {
+		acl = posix_acl_from_xattr(value, size);
 	} else if (size == -ENOENT || size == -ENODATA || size == 0) {
 		/* FIXME, who returns -ENOENT?  I think nobody */
 		acl = NULL;
-		set_cached_acl(inode, type, acl);
 	} else {
 		acl = ERR_PTR(-EIO);
 	}
+	kfree(value);
+
+	if (!IS_ERR(acl))
+		set_cached_acl(inode, type, acl);
 
 	return acl;
 }

fs/btrfs/backref.c

@@ -0,0 +1,776 @@
+/*
+ * Copyright (C) 2011 STRATO.  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 "ctree.h"
+#include "disk-io.h"
+#include "backref.h"
+
+struct __data_ref {
+	struct list_head list;
+	u64 inum;
+	u64 root;
+	u64 extent_data_item_offset;
+};
+
+struct __shared_ref {
+	struct list_head list;
+	u64 disk_byte;
+};
+
+static int __inode_info(u64 inum, u64 ioff, u8 key_type,
+			struct btrfs_root *fs_root, struct btrfs_path *path,
+			struct btrfs_key *found_key)
+{
+	int ret;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+
+	key.type = key_type;
+	key.objectid = inum;
+	key.offset = ioff;
+
+	ret = btrfs_search_slot(NULL, fs_root, &key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+
+	eb = path->nodes[0];
+	if (ret && path->slots[0] >= btrfs_header_nritems(eb)) {
+		ret = btrfs_next_leaf(fs_root, path);
+		if (ret)
+			return ret;
+		eb = path->nodes[0];
+	}
+
+	btrfs_item_key_to_cpu(eb, found_key, path->slots[0]);
+	if (found_key->type != key.type || found_key->objectid != key.objectid)
+		return 1;
+
+	return 0;
+}
+
+/*
+ * this makes the path point to (inum INODE_ITEM ioff)
+ */
+int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+			struct btrfs_path *path)
+{
+	struct btrfs_key key;
+	return __inode_info(inum, ioff, BTRFS_INODE_ITEM_KEY, fs_root, path,
+				&key);
+}
+
+static int inode_ref_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+				struct btrfs_path *path,
+				struct btrfs_key *found_key)
+{
+	return __inode_info(inum, ioff, BTRFS_INODE_REF_KEY, fs_root, path,
+				found_key);
+}
+
+/*
+ * this iterates to turn a btrfs_inode_ref into a full filesystem path. elements
+ * of the path are separated by '/' and the path is guaranteed to be
+ * 0-terminated. the path is only given within the current file system.
+ * Therefore, it never starts with a '/'. the caller is responsible to provide
+ * "size" bytes in "dest". the dest buffer will be filled backwards. finally,
+ * the start point of the resulting string is returned. this pointer is within
+ * dest, normally.
+ * in case the path buffer would overflow, the pointer is decremented further
+ * as if output was written to the buffer, though no more output is actually
+ * generated. that way, the caller can determine how much space would be
+ * required for the path to fit into the buffer. in that case, the returned
+ * value will be smaller than dest. callers must check this!
+ */
+static char *iref_to_path(struct btrfs_root *fs_root, struct btrfs_path *path,
+				struct btrfs_inode_ref *iref,
+				struct extent_buffer *eb_in, u64 parent,
+				char *dest, u32 size)
+{
+	u32 len;
+	int slot;
+	u64 next_inum;
+	int ret;
+	s64 bytes_left = size - 1;
+	struct extent_buffer *eb = eb_in;
+	struct btrfs_key found_key;
+
+	if (bytes_left >= 0)
+		dest[bytes_left] = '\0';
+
+	while (1) {
+		len = btrfs_inode_ref_name_len(eb, iref);
+		bytes_left -= len;
+		if (bytes_left >= 0)
+			read_extent_buffer(eb, dest + bytes_left,
+						(unsigned long)(iref + 1), len);
+		if (eb != eb_in)
+			free_extent_buffer(eb);
+		ret = inode_ref_info(parent, 0, fs_root, path, &found_key);
+		if (ret)
+			break;
+		next_inum = found_key.offset;
+
+		/* regular exit ahead */
+		if (parent == next_inum)
+			break;
+
+		slot = path->slots[0];
+		eb = path->nodes[0];
+		/* make sure we can use eb after releasing the path */
+		if (eb != eb_in)
+			atomic_inc(&eb->refs);
+		btrfs_release_path(path);
+
+		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
+		parent = next_inum;
+		--bytes_left;
+		if (bytes_left >= 0)
+			dest[bytes_left] = '/';
+	}
+
+	btrfs_release_path(path);
+
+	if (ret)
+		return ERR_PTR(ret);
+
+	return dest + bytes_left;
+}
+
+/*
+ * this makes the path point to (logical EXTENT_ITEM *)
+ * returns BTRFS_EXTENT_FLAG_DATA for data, BTRFS_EXTENT_FLAG_TREE_BLOCK for
+ * tree blocks and <0 on error.
+ */
+int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
+			struct btrfs_path *path, struct btrfs_key *found_key)
+{
+	int ret;
+	u64 flags;
+	u32 item_size;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct btrfs_key key;
+
+	key.type = BTRFS_EXTENT_ITEM_KEY;
+	key.objectid = logical;
+	key.offset = (u64)-1;
+
+	ret = btrfs_search_slot(NULL, fs_info->extent_root, &key, path, 0, 0);
+	if (ret < 0)
+		return ret;
+	ret = btrfs_previous_item(fs_info->extent_root, path,
+					0, BTRFS_EXTENT_ITEM_KEY);
+	if (ret < 0)
+		return ret;
+
+	btrfs_item_key_to_cpu(path->nodes[0], found_key, path->slots[0]);
+	if (found_key->type != BTRFS_EXTENT_ITEM_KEY ||
+	    found_key->objectid > logical ||
+	    found_key->objectid + found_key->offset <= logical)
+		return -ENOENT;
+
+	eb = path->nodes[0];
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+	BUG_ON(item_size < sizeof(*ei));
+
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	flags = btrfs_extent_flags(eb, ei);
+
+	if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		return BTRFS_EXTENT_FLAG_TREE_BLOCK;
+	if (flags & BTRFS_EXTENT_FLAG_DATA)
+		return BTRFS_EXTENT_FLAG_DATA;
+
+	return -EIO;
+}
+
+/*
+ * helper function to iterate extent inline refs. ptr must point to a 0 value
+ * for the first call and may be modified. it is used to track state.
+ * if more refs exist, 0 is returned and the next call to
+ * __get_extent_inline_ref must pass the modified ptr parameter to get the
+ * next ref. after the last ref was processed, 1 is returned.
+ * returns <0 on error
+ */
+static int __get_extent_inline_ref(unsigned long *ptr, struct extent_buffer *eb,
+				struct btrfs_extent_item *ei, u32 item_size,
+				struct btrfs_extent_inline_ref **out_eiref,
+				int *out_type)
+{
+	unsigned long end;
+	u64 flags;
+	struct btrfs_tree_block_info *info;
+
+	if (!*ptr) {
+		/* first call */
+		flags = btrfs_extent_flags(eb, ei);
+		if (flags & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+			info = (struct btrfs_tree_block_info *)(ei + 1);
+			*out_eiref =
+				(struct btrfs_extent_inline_ref *)(info + 1);
+		} else {
+			*out_eiref = (struct btrfs_extent_inline_ref *)(ei + 1);
+		}
+		*ptr = (unsigned long)*out_eiref;
+		if ((void *)*ptr >= (void *)ei + item_size)
+			return -ENOENT;
+	}
+
+	end = (unsigned long)ei + item_size;
+	*out_eiref = (struct btrfs_extent_inline_ref *)*ptr;
+	*out_type = btrfs_extent_inline_ref_type(eb, *out_eiref);
+
+	*ptr += btrfs_extent_inline_ref_size(*out_type);
+	WARN_ON(*ptr > end);
+	if (*ptr == end)
+		return 1; /* last */
+
+	return 0;
+}
+
+/*
+ * reads the tree block backref for an extent. tree level and root are returned
+ * through out_level and out_root. ptr must point to a 0 value for the first
+ * call and may be modified (see __get_extent_inline_ref comment).
+ * returns 0 if data was provided, 1 if there was no more data to provide or
+ * <0 on error.
+ */
+int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
+				struct btrfs_extent_item *ei, u32 item_size,
+				u64 *out_root, u8 *out_level)
+{
+	int ret;
+	int type;
+	struct btrfs_tree_block_info *info;
+	struct btrfs_extent_inline_ref *eiref;
+
+	if (*ptr == (unsigned long)-1)
+		return 1;
+
+	while (1) {
+		ret = __get_extent_inline_ref(ptr, eb, ei, item_size,
+						&eiref, &type);
+		if (ret < 0)
+			return ret;
+
+		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
+		    type == BTRFS_SHARED_BLOCK_REF_KEY)
+			break;
+
+		if (ret == 1)
+			return 1;
+	}
+
+	/* we can treat both ref types equally here */
+	info = (struct btrfs_tree_block_info *)(ei + 1);
+	*out_root = btrfs_extent_inline_ref_offset(eb, eiref);
+	*out_level = btrfs_tree_block_level(eb, info);
+
+	if (ret == 1)
+		*ptr = (unsigned long)-1;
+
+	return 0;
+}
+
+static int __data_list_add(struct list_head *head, u64 inum,
+				u64 extent_data_item_offset, u64 root)
+{
+	struct __data_ref *ref;
+
+	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	if (!ref)
+		return -ENOMEM;
+
+	ref->inum = inum;
+	ref->extent_data_item_offset = extent_data_item_offset;
+	ref->root = root;
+	list_add_tail(&ref->list, head);
+
+	return 0;
+}
+
+static int __data_list_add_eb(struct list_head *head, struct extent_buffer *eb,
+				struct btrfs_extent_data_ref *dref)
+{
+	return __data_list_add(head, btrfs_extent_data_ref_objectid(eb, dref),
+				btrfs_extent_data_ref_offset(eb, dref),
+				btrfs_extent_data_ref_root(eb, dref));
+}
+
+static int __shared_list_add(struct list_head *head, u64 disk_byte)
+{
+	struct __shared_ref *ref;
+
+	ref = kmalloc(sizeof(*ref), GFP_NOFS);
+	if (!ref)
+		return -ENOMEM;
+
+	ref->disk_byte = disk_byte;
+	list_add_tail(&ref->list, head);
+
+	return 0;
+}
+
+static int __iter_shared_inline_ref_inodes(struct btrfs_fs_info *fs_info,
+					   u64 logical, u64 inum,
+					   u64 extent_data_item_offset,
+					   u64 extent_offset,
+					   struct btrfs_path *path,
+					   struct list_head *data_refs,
+					   iterate_extent_inodes_t *iterate,
+					   void *ctx)
+{
+	u64 ref_root;
+	u32 item_size;
+	struct btrfs_key key;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct btrfs_extent_inline_ref *eiref;
+	struct __data_ref *ref;
+	int ret;
+	int type;
+	int last;
+	unsigned long ptr = 0;
+
+	WARN_ON(!list_empty(data_refs));
+	ret = extent_from_logical(fs_info, logical, path, &key);
+	if (ret & BTRFS_EXTENT_FLAG_DATA)
+		ret = -EIO;
+	if (ret < 0)
+		goto out;
+
+	eb = path->nodes[0];
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+
+	ret = 0;
+	ref_root = 0;
+	/*
+	 * as done in iterate_extent_inodes, we first build a list of refs to
+	 * iterate, then free the path and then iterate them to avoid deadlocks.
+	 */
+	do {
+		last = __get_extent_inline_ref(&ptr, eb, ei, item_size,
+						&eiref, &type);
+		if (last < 0) {
+			ret = last;
+			goto out;
+		}
+		if (type == BTRFS_TREE_BLOCK_REF_KEY ||
+		    type == BTRFS_SHARED_BLOCK_REF_KEY) {
+			ref_root = btrfs_extent_inline_ref_offset(eb, eiref);
+			ret = __data_list_add(data_refs, inum,
+						extent_data_item_offset,
+						ref_root);
+		}
+	} while (!ret && !last);
+
+	btrfs_release_path(path);
+
+	if (ref_root == 0) {
+		printk(KERN_ERR "btrfs: failed to find tree block ref "
+			"for shared data backref %llu\n", logical);
+		WARN_ON(1);
+		ret = -EIO;
+	}
+
+out:
+	while (!list_empty(data_refs)) {
+		ref = list_first_entry(data_refs, struct __data_ref, list);
+		list_del(&ref->list);
+		if (!ret)
+			ret = iterate(ref->inum, extent_offset +
+					ref->extent_data_item_offset,
+					ref->root, ctx);
+		kfree(ref);
+	}
+
+	return ret;
+}
+
+static int __iter_shared_inline_ref(struct btrfs_fs_info *fs_info,
+				    u64 logical, u64 orig_extent_item_objectid,
+				    u64 extent_offset, struct btrfs_path *path,
+				    struct list_head *data_refs,
+				    iterate_extent_inodes_t *iterate,
+				    void *ctx)
+{
+	u64 disk_byte;
+	struct btrfs_key key;
+	struct btrfs_file_extent_item *fi;
+	struct extent_buffer *eb;
+	int slot;
+	int nritems;
+	int ret;
+	int found = 0;
+
+	eb = read_tree_block(fs_info->tree_root, logical,
+				fs_info->tree_root->leafsize, 0);
+	if (!eb)
+		return -EIO;
+
+	/*
+	 * from the shared data ref, we only have the leaf but we need
+	 * the key. thus, we must look into all items and see that we
+	 * find one (some) with a reference to our extent item.
+	 */
+	nritems = btrfs_header_nritems(eb);
+	for (slot = 0; slot < nritems; ++slot) {
+		btrfs_item_key_to_cpu(eb, &key, slot);
+		if (key.type != BTRFS_EXTENT_DATA_KEY)
+			continue;
+		fi = btrfs_item_ptr(eb, slot, struct btrfs_file_extent_item);
+		if (!fi) {
+			free_extent_buffer(eb);
+			return -EIO;
+		}
+		disk_byte = btrfs_file_extent_disk_bytenr(eb, fi);
+		if (disk_byte != orig_extent_item_objectid) {
+			if (found)
+				break;
+			else
+				continue;
+		}
+		++found;
+		ret = __iter_shared_inline_ref_inodes(fs_info, logical,
+							key.objectid,
+							key.offset,
+							extent_offset, path,
+							data_refs,
+							iterate, ctx);
+		if (ret)
+			break;
+	}
+
+	if (!found) {
+		printk(KERN_ERR "btrfs: failed to follow shared data backref "
+			"to parent %llu\n", logical);
+		WARN_ON(1);
+		ret = -EIO;
+	}
+
+	free_extent_buffer(eb);
+	return ret;
+}
+
+/*
+ * calls iterate() for every inode that references the extent identified by
+ * the given parameters. will use the path given as a parameter and return it
+ * released.
+ * when the iterator function returns a non-zero value, iteration stops.
+ */
+int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				u64 extent_item_objectid,
+				u64 extent_offset,
+				iterate_extent_inodes_t *iterate, void *ctx)
+{
+	unsigned long ptr = 0;
+	int last;
+	int ret;
+	int type;
+	u64 logical;
+	u32 item_size;
+	struct btrfs_extent_inline_ref *eiref;
+	struct btrfs_extent_data_ref *dref;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct btrfs_key key;
+	struct list_head data_refs = LIST_HEAD_INIT(data_refs);
+	struct list_head shared_refs = LIST_HEAD_INIT(shared_refs);
+	struct __data_ref *ref_d;
+	struct __shared_ref *ref_s;
+
+	eb = path->nodes[0];
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+
+	/* first we iterate the inline refs, ... */
+	do {
+		last = __get_extent_inline_ref(&ptr, eb, ei, item_size,
+						&eiref, &type);
+		if (last == -ENOENT) {
+			ret = 0;
+			break;
+		}
+		if (last < 0) {
+			ret = last;
+			break;
+		}
+
+		if (type == BTRFS_EXTENT_DATA_REF_KEY) {
+			dref = (struct btrfs_extent_data_ref *)(&eiref->offset);
+			ret = __data_list_add_eb(&data_refs, eb, dref);
+		} else if (type == BTRFS_SHARED_DATA_REF_KEY) {
+			logical = btrfs_extent_inline_ref_offset(eb, eiref);
+			ret = __shared_list_add(&shared_refs, logical);
+		}
+	} while (!ret && !last);
+
+	/* ... then we proceed to in-tree references and ... */
+	while (!ret) {
+		++path->slots[0];
+		if (path->slots[0] > btrfs_header_nritems(eb)) {
+			ret = btrfs_next_leaf(fs_info->extent_root, path);
+			if (ret) {
+				if (ret == 1)
+					ret = 0; /* we're done */
+				break;
+			}
+			eb = path->nodes[0];
+		}
+		btrfs_item_key_to_cpu(eb, &key, path->slots[0]);
+		if (key.objectid != extent_item_objectid)
+			break;
+		if (key.type == BTRFS_EXTENT_DATA_REF_KEY) {
+			dref = btrfs_item_ptr(eb, path->slots[0],
+						struct btrfs_extent_data_ref);
+			ret = __data_list_add_eb(&data_refs, eb, dref);
+		} else if (key.type == BTRFS_SHARED_DATA_REF_KEY) {
+			ret = __shared_list_add(&shared_refs, key.offset);
+		}
+	}
+
+	btrfs_release_path(path);
+
+	/*
+	 * ... only at the very end we can process the refs we found. this is
+	 * because the iterator function we call is allowed to make tree lookups
+	 * and we have to avoid deadlocks. additionally, we need more tree
+	 * lookups ourselves for shared data refs.
+	 */
+	while (!list_empty(&data_refs)) {
+		ref_d = list_first_entry(&data_refs, struct __data_ref, list);
+		list_del(&ref_d->list);
+		if (!ret)
+			ret = iterate(ref_d->inum, extent_offset +
+					ref_d->extent_data_item_offset,
+					ref_d->root, ctx);
+		kfree(ref_d);
+	}
+
+	while (!list_empty(&shared_refs)) {
+		ref_s = list_first_entry(&shared_refs, struct __shared_ref,
+					list);
+		list_del(&ref_s->list);
+		if (!ret)
+			ret = __iter_shared_inline_ref(fs_info,
+							ref_s->disk_byte,
+							extent_item_objectid,
+							extent_offset, path,
+							&data_refs,
+							iterate, ctx);
+		kfree(ref_s);
+	}
+
+	return ret;
+}
+
+int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				iterate_extent_inodes_t *iterate, void *ctx)
+{
+	int ret;
+	u64 offset;
+	struct btrfs_key found_key;
+
+	ret = extent_from_logical(fs_info, logical, path,
+					&found_key);
+	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		ret = -EINVAL;
+	if (ret < 0)
+		return ret;
+
+	offset = logical - found_key.objectid;
+	ret = iterate_extent_inodes(fs_info, path, found_key.objectid,
+					offset, iterate, ctx);
+
+	return ret;
+}
+
+static int iterate_irefs(u64 inum, struct btrfs_root *fs_root,
+				struct btrfs_path *path,
+				iterate_irefs_t *iterate, void *ctx)
+{
+	int ret;
+	int slot;
+	u32 cur;
+	u32 len;
+	u32 name_len;
+	u64 parent = 0;
+	int found = 0;
+	struct extent_buffer *eb;
+	struct btrfs_item *item;
+	struct btrfs_inode_ref *iref;
+	struct btrfs_key found_key;
+
+	while (1) {
+		ret = inode_ref_info(inum, parent ? parent+1 : 0, fs_root, path,
+					&found_key);
+		if (ret < 0)
+			break;
+		if (ret) {
+			ret = found ? 0 : -ENOENT;
+			break;
+		}
+		++found;
+
+		parent = found_key.offset;
+		slot = path->slots[0];
+		eb = path->nodes[0];
+		/* make sure we can use eb after releasing the path */
+		atomic_inc(&eb->refs);
+		btrfs_release_path(path);
+
+		item = btrfs_item_nr(eb, slot);
+		iref = btrfs_item_ptr(eb, slot, struct btrfs_inode_ref);
+
+		for (cur = 0; cur < btrfs_item_size(eb, item); cur += len) {
+			name_len = btrfs_inode_ref_name_len(eb, iref);
+			/* path must be released before calling iterate()! */
+			ret = iterate(parent, iref, eb, ctx);
+			if (ret) {
+				free_extent_buffer(eb);
+				break;
+			}
+			len = sizeof(*iref) + name_len;
+			iref = (struct btrfs_inode_ref *)((char *)iref + len);
+		}
+		free_extent_buffer(eb);
+	}
+
+	btrfs_release_path(path);
+
+	return ret;
+}
+
+/*
+ * returns 0 if the path could be dumped (probably truncated)
+ * returns <0 in case of an error
+ */
+static int inode_to_path(u64 inum, struct btrfs_inode_ref *iref,
+				struct extent_buffer *eb, void *ctx)
+{
+	struct inode_fs_paths *ipath = ctx;
+	char *fspath;
+	char *fspath_min;
+	int i = ipath->fspath->elem_cnt;
+	const int s_ptr = sizeof(char *);
+	u32 bytes_left;
+
+	bytes_left = ipath->fspath->bytes_left > s_ptr ?
+					ipath->fspath->bytes_left - s_ptr : 0;
+
+	fspath_min = (char *)ipath->fspath->val + (i + 1) * s_ptr;
+	fspath = iref_to_path(ipath->fs_root, ipath->btrfs_path, iref, eb,
+				inum, fspath_min, bytes_left);
+	if (IS_ERR(fspath))
+		return PTR_ERR(fspath);
+
+	if (fspath > fspath_min) {
+		ipath->fspath->val[i] = (u64)fspath;
+		++ipath->fspath->elem_cnt;
+		ipath->fspath->bytes_left = fspath - fspath_min;
+	} else {
+		++ipath->fspath->elem_missed;
+		ipath->fspath->bytes_missing += fspath_min - fspath;
+		ipath->fspath->bytes_left = 0;
+	}
+
+	return 0;
+}
+
+/*
+ * this dumps all file system paths to the inode into the ipath struct, provided
+ * is has been created large enough. each path is zero-terminated and accessed
+ * from ipath->fspath->val[i].
+ * when it returns, there are ipath->fspath->elem_cnt number of paths available
+ * in ipath->fspath->val[]. when the allocated space wasn't sufficient, the
+ * number of missed paths in recored in ipath->fspath->elem_missed, otherwise,
+ * it's zero. ipath->fspath->bytes_missing holds the number of bytes that would
+ * have been needed to return all paths.
+ */
+int paths_from_inode(u64 inum, struct inode_fs_paths *ipath)
+{
+	return iterate_irefs(inum, ipath->fs_root, ipath->btrfs_path,
+				inode_to_path, ipath);
+}
+
+/*
+ * allocates space to return multiple file system paths for an inode.
+ * total_bytes to allocate are passed, note that space usable for actual path
+ * information will be total_bytes - sizeof(struct inode_fs_paths).
+ * the returned pointer must be freed with free_ipath() in the end.
+ */
+struct btrfs_data_container *init_data_container(u32 total_bytes)
+{
+	struct btrfs_data_container *data;
+	size_t alloc_bytes;
+
+	alloc_bytes = max_t(size_t, total_bytes, sizeof(*data));
+	data = kmalloc(alloc_bytes, GFP_NOFS);
+	if (!data)
+		return ERR_PTR(-ENOMEM);
+
+	if (total_bytes >= sizeof(*data)) {
+		data->bytes_left = total_bytes - sizeof(*data);
+		data->bytes_missing = 0;
+	} else {
+		data->bytes_missing = sizeof(*data) - total_bytes;
+		data->bytes_left = 0;
+	}
+
+	data->elem_cnt = 0;
+	data->elem_missed = 0;
+
+	return data;
+}
+
+/*
+ * allocates space to return multiple file system paths for an inode.
+ * total_bytes to allocate are passed, note that space usable for actual path
+ * information will be total_bytes - sizeof(struct inode_fs_paths).
+ * the returned pointer must be freed with free_ipath() in the end.
+ */
+struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
+					struct btrfs_path *path)
+{
+	struct inode_fs_paths *ifp;
+	struct btrfs_data_container *fspath;
+
+	fspath = init_data_container(total_bytes);
+	if (IS_ERR(fspath))
+		return (void *)fspath;
+
+	ifp = kmalloc(sizeof(*ifp), GFP_NOFS);
+	if (!ifp) {
+		kfree(fspath);
+		return ERR_PTR(-ENOMEM);
+	}
+
+	ifp->btrfs_path = path;
+	ifp->fspath = fspath;
+	ifp->fs_root = fs_root;
+
+	return ifp;
+}
+
+void free_ipath(struct inode_fs_paths *ipath)
+{
+	kfree(ipath);
+}

fs/btrfs/backref.h

@@ -0,0 +1,62 @@
+/*
+ * Copyright (C) 2011 STRATO.  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.
+ */
+
+#ifndef __BTRFS_BACKREF__
+#define __BTRFS_BACKREF__
+
+#include "ioctl.h"
+
+struct inode_fs_paths {
+	struct btrfs_path		*btrfs_path;
+	struct btrfs_root		*fs_root;
+	struct btrfs_data_container	*fspath;
+};
+
+typedef int (iterate_extent_inodes_t)(u64 inum, u64 offset, u64 root,
+		void *ctx);
+typedef int (iterate_irefs_t)(u64 parent, struct btrfs_inode_ref *iref,
+				struct extent_buffer *eb, void *ctx);
+
+int inode_item_info(u64 inum, u64 ioff, struct btrfs_root *fs_root,
+			struct btrfs_path *path);
+
+int extent_from_logical(struct btrfs_fs_info *fs_info, u64 logical,
+			struct btrfs_path *path, struct btrfs_key *found_key);
+
+int tree_backref_for_extent(unsigned long *ptr, struct extent_buffer *eb,
+				struct btrfs_extent_item *ei, u32 item_size,
+				u64 *out_root, u8 *out_level);
+
+int iterate_extent_inodes(struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				u64 extent_item_objectid,
+				u64 extent_offset,
+				iterate_extent_inodes_t *iterate, void *ctx);
+
+int iterate_inodes_from_logical(u64 logical, struct btrfs_fs_info *fs_info,
+				struct btrfs_path *path,
+				iterate_extent_inodes_t *iterate, void *ctx);
+
+int paths_from_inode(u64 inum, struct inode_fs_paths *ipath);
+
+struct btrfs_data_container *init_data_container(u32 total_bytes);
+struct inode_fs_paths *init_ipath(s32 total_bytes, struct btrfs_root *fs_root,
+					struct btrfs_path *path);
+void free_ipath(struct inode_fs_paths *ipath);
+
+#endif

fs/btrfs/btrfs_inode.h

@@ -103,11 +103,6 @@ struct btrfs_inode {
 	 */
 	u64 delalloc_bytes;
 
-	/* total number of bytes that may be used for this inode for
-	 * delalloc
-	 */
-	u64 reserved_bytes;
-
 	/*
 	 * the size of the file stored in the metadata on disk.  data=ordered
 	 * means the in-memory i_size might be larger than the size on disk
@@ -115,9 +110,6 @@ struct btrfs_inode {
 	 */
 	u64 disk_i_size;
 
-	/* flags field from the on disk inode */
-	u32 flags;
-
 	/*
 	 * if this is a directory then index_cnt is the counter for the index
 	 * number for new files that are created
@@ -131,6 +123,15 @@ struct btrfs_inode {
 	 */
 	u64 last_unlink_trans;
 
+	/*
+	 * Number of bytes outstanding that are going to need csums.  This is
+	 * used in ENOSPC accounting.
+	 */
+	u64 csum_bytes;
+
+	/* flags field from the on disk inode */
+	u32 flags;
+
 	/*
 	 * Counters to keep track of the number of extent item's we may use due
 	 * to delalloc and such.  outstanding_extents is the number of extent

fs/btrfs/compression.c

@@ -85,7 +85,8 @@ struct compressed_bio {
 static inline int compressed_bio_size(struct btrfs_root *root,
 				      unsigned long disk_size)
 {
-	u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
+
 	return sizeof(struct compressed_bio) +
 		((disk_size + root->sectorsize - 1) / root->sectorsize) *
 		csum_size;

fs/btrfs/ctree.c

@@ -902,9 +902,10 @@ static noinline int balance_level(struct btrfs_trans_handle *trans,
 
 	orig_ptr = btrfs_node_blockptr(mid, orig_slot);
 
-	if (level < BTRFS_MAX_LEVEL - 1)
+	if (level < BTRFS_MAX_LEVEL - 1) {
 		parent = path->nodes[level + 1];
-	pslot = path->slots[level + 1];
+		pslot = path->slots[level + 1];
+	}
 
 	/*
 	 * deal with the case where there is only one pointer in the root
@@ -1107,9 +1108,10 @@ static noinline int push_nodes_for_insert(struct btrfs_trans_handle *trans,
 	mid = path->nodes[level];
 	WARN_ON(btrfs_header_generation(mid) != trans->transid);
 
-	if (level < BTRFS_MAX_LEVEL - 1)
+	if (level < BTRFS_MAX_LEVEL - 1) {
 		parent = path->nodes[level + 1];
-	pslot = path->slots[level + 1];
+		pslot = path->slots[level + 1];
+	}
 
 	if (!parent)
 		return 1;

fs/btrfs/ctree.h

@@ -30,6 +30,7 @@
 #include <linux/kobject.h>
 #include <trace/events/btrfs.h>
 #include <asm/kmap_types.h>
+#include <linux/pagemap.h>
 #include "extent_io.h"
 #include "extent_map.h"
 #include "async-thread.h"
@@ -359,6 +360,47 @@ struct btrfs_header {
 #define BTRFS_SYSTEM_CHUNK_ARRAY_SIZE 2048
 #define BTRFS_LABEL_SIZE 256
 
+/*
+ * just in case we somehow lose the roots and are not able to mount,
+ * we store an array of the roots from previous transactions
+ * in the super.
+ */
+#define BTRFS_NUM_BACKUP_ROOTS 4
+struct btrfs_root_backup {
+	__le64 tree_root;
+	__le64 tree_root_gen;
+
+	__le64 chunk_root;
+	__le64 chunk_root_gen;
+
+	__le64 extent_root;
+	__le64 extent_root_gen;
+
+	__le64 fs_root;
+	__le64 fs_root_gen;
+
+	__le64 dev_root;
+	__le64 dev_root_gen;
+
+	__le64 csum_root;
+	__le64 csum_root_gen;
+
+	__le64 total_bytes;
+	__le64 bytes_used;
+	__le64 num_devices;
+	/* future */
+	__le64 unsed_64[4];
+
+	u8 tree_root_level;
+	u8 chunk_root_level;
+	u8 extent_root_level;
+	u8 fs_root_level;
+	u8 dev_root_level;
+	u8 csum_root_level;
+	/* future and to align */
+	u8 unused_8[10];
+} __attribute__ ((__packed__));
+
 /*
  * the super block basically lists the main trees of the FS
  * it currently lacks any block count etc etc
@@ -405,6 +447,7 @@ struct btrfs_super_block {
 	/* future expansion */
 	__le64 reserved[31];
 	u8 sys_chunk_array[BTRFS_SYSTEM_CHUNK_ARRAY_SIZE];
+	struct btrfs_root_backup super_roots[BTRFS_NUM_BACKUP_ROOTS];
 } __attribute__ ((__packed__));
 
 /*
@@ -772,14 +815,8 @@ struct btrfs_space_info {
 struct btrfs_block_rsv {
 	u64 size;
 	u64 reserved;
-	u64 freed[2];
 	struct btrfs_space_info *space_info;
-	struct list_head list;
 	spinlock_t lock;
-	atomic_t usage;
-	unsigned int priority:8;
-	unsigned int durable:1;
-	unsigned int refill_used:1;
 	unsigned int full:1;
 };
 
@@ -840,10 +877,10 @@ struct btrfs_block_group_cache {
 	spinlock_t lock;
 	u64 pinned;
 	u64 reserved;
-	u64 reserved_pinned;
 	u64 bytes_super;
 	u64 flags;
 	u64 sectorsize;
+	u64 cache_generation;
 	unsigned int ro:1;
 	unsigned int dirty:1;
 	unsigned int iref:1;
@@ -899,6 +936,10 @@ struct btrfs_fs_info {
 	spinlock_t block_group_cache_lock;
 	struct rb_root block_group_cache_tree;
 
+	/* keep track of unallocated space */
+	spinlock_t free_chunk_lock;
+	u64 free_chunk_space;
+
 	struct extent_io_tree freed_extents[2];
 	struct extent_io_tree *pinned_extents;
 
@@ -916,14 +957,11 @@ struct btrfs_fs_info {
 	struct btrfs_block_rsv trans_block_rsv;
 	/* block reservation for chunk tree */
 	struct btrfs_block_rsv chunk_block_rsv;
+	/* block reservation for delayed operations */
+	struct btrfs_block_rsv delayed_block_rsv;
 
 	struct btrfs_block_rsv empty_block_rsv;
 
-	/* list of block reservations that cross multiple transactions */
-	struct list_head durable_block_rsv_list;
-
-	struct mutex durable_block_rsv_mutex;
-
 	u64 generation;
 	u64 last_trans_committed;
 
@@ -942,8 +980,8 @@ struct btrfs_fs_info {
 	wait_queue_head_t transaction_blocked_wait;
 	wait_queue_head_t async_submit_wait;
 
-	struct btrfs_super_block super_copy;
-	struct btrfs_super_block super_for_commit;
+	struct btrfs_super_block *super_copy;
+	struct btrfs_super_block *super_for_commit;
 	struct block_device *__bdev;
 	struct super_block *sb;
 	struct inode *btree_inode;
@@ -1036,6 +1074,7 @@ struct btrfs_fs_info {
 	struct btrfs_workers endio_freespace_worker;
 	struct btrfs_workers submit_workers;
 	struct btrfs_workers caching_workers;
+	struct btrfs_workers readahead_workers;
 
 	/*
 	 * fixup workers take dirty pages that didn't properly go through
@@ -1119,6 +1158,13 @@ struct btrfs_fs_info {
 	u64 fs_state;
 
 	struct btrfs_delayed_root *delayed_root;
+
+	/* readahead tree */
+	spinlock_t reada_lock;
+	struct radix_tree_root reada_tree;
+
+	/* next backup root to be overwritten */
+	int backup_root_index;
 };
 
 /*
@@ -1363,6 +1409,7 @@ struct btrfs_ioctl_defrag_range_args {
 #define BTRFS_MOUNT_ENOSPC_DEBUG	 (1 << 15)
 #define BTRFS_MOUNT_AUTO_DEFRAG		(1 << 16)
 #define BTRFS_MOUNT_INODE_MAP_CACHE	(1 << 17)
+#define BTRFS_MOUNT_RECOVERY		(1 << 18)
 
 #define btrfs_clear_opt(o, opt)		((o) &= ~BTRFS_MOUNT_##opt)
 #define btrfs_set_opt(o, opt)		((o) |= BTRFS_MOUNT_##opt)
@@ -1978,6 +2025,55 @@ static inline bool btrfs_root_readonly(struct btrfs_root *root)
 	return root->root_item.flags & BTRFS_ROOT_SUBVOL_RDONLY;
 }
 
+/* struct btrfs_root_backup */
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root, struct btrfs_root_backup,
+		   tree_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root_gen, struct btrfs_root_backup,
+		   tree_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_tree_root_level, struct btrfs_root_backup,
+		   tree_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root, struct btrfs_root_backup,
+		   chunk_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root_gen, struct btrfs_root_backup,
+		   chunk_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_chunk_root_level, struct btrfs_root_backup,
+		   chunk_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root, struct btrfs_root_backup,
+		   extent_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root_gen, struct btrfs_root_backup,
+		   extent_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_extent_root_level, struct btrfs_root_backup,
+		   extent_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root, struct btrfs_root_backup,
+		   fs_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root_gen, struct btrfs_root_backup,
+		   fs_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_fs_root_level, struct btrfs_root_backup,
+		   fs_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root, struct btrfs_root_backup,
+		   dev_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root_gen, struct btrfs_root_backup,
+		   dev_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_dev_root_level, struct btrfs_root_backup,
+		   dev_root_level, 8);
+
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root, struct btrfs_root_backup,
+		   csum_root, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root_gen, struct btrfs_root_backup,
+		   csum_root_gen, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_csum_root_level, struct btrfs_root_backup,
+		   csum_root_level, 8);
+BTRFS_SETGET_STACK_FUNCS(backup_total_bytes, struct btrfs_root_backup,
+		   total_bytes, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_bytes_used, struct btrfs_root_backup,
+		   bytes_used, 64);
+BTRFS_SETGET_STACK_FUNCS(backup_num_devices, struct btrfs_root_backup,
+		   num_devices, 64);
+
 /* struct btrfs_super_block */
 
 BTRFS_SETGET_STACK_FUNCS(super_bytenr, struct btrfs_super_block, bytenr, 64);
@@ -2129,6 +2225,11 @@ static inline bool btrfs_mixed_space_info(struct btrfs_space_info *space_info)
 		(space_info->flags & BTRFS_BLOCK_GROUP_DATA));
 }
 
+static inline gfp_t btrfs_alloc_write_mask(struct address_space *mapping)
+{
+	return mapping_gfp_mask(mapping) & ~__GFP_FS;
+}
+
 /* extent-tree.c */
 static inline u64 btrfs_calc_trans_metadata_size(struct btrfs_root *root,
 						 unsigned num_items)
@@ -2137,6 +2238,17 @@ static inline u64 btrfs_calc_trans_metadata_size(struct btrfs_root *root,
 		3 * num_items;
 }
 
+/*
+ * Doing a truncate won't result in new nodes or leaves, just what we need for
+ * COW.
+ */
+static inline u64 btrfs_calc_trunc_metadata_size(struct btrfs_root *root,
+						 unsigned num_items)
+{
+	return (root->leafsize + root->nodesize * (BTRFS_MAX_LEVEL - 1)) *
+		num_items;
+}
+
 void btrfs_put_block_group(struct btrfs_block_group_cache *cache);
 int btrfs_run_delayed_refs(struct btrfs_trans_handle *trans,
 			   struct btrfs_root *root, unsigned long count);
@@ -2146,6 +2258,9 @@ int btrfs_lookup_extent_info(struct btrfs_trans_handle *trans,
 			     u64 num_bytes, u64 *refs, u64 *flags);
 int btrfs_pin_extent(struct btrfs_root *root,
 		     u64 bytenr, u64 num, int reserved);
+int btrfs_pin_extent_for_log_replay(struct btrfs_trans_handle *trans,
+				    struct btrfs_root *root,
+				    u64 bytenr, u64 num_bytes);
 int btrfs_cross_ref_exist(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  u64 objectid, u64 offset, u64 bytenr);
@@ -2196,8 +2311,8 @@ int btrfs_free_extent(struct btrfs_trans_handle *trans,
 		      u64 root_objectid, u64 owner, u64 offset);
 
 int btrfs_free_reserved_extent(struct btrfs_root *root, u64 start, u64 len);
-int btrfs_update_reserved_bytes(struct btrfs_block_group_cache *cache,
-				u64 num_bytes, int reserve, int sinfo);
+int btrfs_free_and_pin_reserved_extent(struct btrfs_root *root,
+				       u64 start, u64 len);
 int btrfs_prepare_extent_commit(struct btrfs_trans_handle *trans,
 				struct btrfs_root *root);
 int btrfs_finish_extent_commit(struct btrfs_trans_handle *trans,
@@ -2240,25 +2355,23 @@ void btrfs_init_block_rsv(struct btrfs_block_rsv *rsv);
 struct btrfs_block_rsv *btrfs_alloc_block_rsv(struct btrfs_root *root);
 void btrfs_free_block_rsv(struct btrfs_root *root,
 			  struct btrfs_block_rsv *rsv);
-void btrfs_add_durable_block_rsv(struct btrfs_fs_info *fs_info,
-				 struct btrfs_block_rsv *rsv);
-int btrfs_block_rsv_add(struct btrfs_trans_handle *trans,
-			struct btrfs_root *root,
+int btrfs_block_rsv_add(struct btrfs_root *root,
 			struct btrfs_block_rsv *block_rsv,
 			u64 num_bytes);
-int btrfs_block_rsv_check(struct btrfs_trans_handle *trans,
-			  struct btrfs_root *root,
+int btrfs_block_rsv_add_noflush(struct btrfs_root *root,
+				struct btrfs_block_rsv *block_rsv,
+				u64 num_bytes);
+int btrfs_block_rsv_check(struct btrfs_root *root,
+			  struct btrfs_block_rsv *block_rsv, int min_factor);
+int btrfs_block_rsv_refill(struct btrfs_root *root,
 			  struct btrfs_block_rsv *block_rsv,
-			  u64 min_reserved, int min_factor);
+			  u64 min_reserved);
 int btrfs_block_rsv_migrate(struct btrfs_block_rsv *src_rsv,
 			    struct btrfs_block_rsv *dst_rsv,
 			    u64 num_bytes);
 void btrfs_block_rsv_release(struct btrfs_root *root,
 			     struct btrfs_block_rsv *block_rsv,
 			     u64 num_bytes);
-int btrfs_truncate_reserve_metadata(struct btrfs_trans_handle *trans,
-				    struct btrfs_root *root,
-				    struct btrfs_block_rsv *rsv);
 int btrfs_set_block_group_ro(struct btrfs_root *root,
 			     struct btrfs_block_group_cache *cache);
 int btrfs_set_block_group_rw(struct btrfs_root *root,
@@ -2379,6 +2492,18 @@ static inline int btrfs_fs_closing(struct btrfs_fs_info *fs_info)
 	smp_mb();
 	return fs_info->closing;
 }
+static inline void free_fs_info(struct btrfs_fs_info *fs_info)
+{
+	kfree(fs_info->delayed_root);
+	kfree(fs_info->extent_root);
+	kfree(fs_info->tree_root);
+	kfree(fs_info->chunk_root);
+	kfree(fs_info->dev_root);
+	kfree(fs_info->csum_root);
+	kfree(fs_info->super_copy);
+	kfree(fs_info->super_for_commit);
+	kfree(fs_info);
+}
 
 /* root-item.c */
 int btrfs_find_root_ref(struct btrfs_root *tree_root,
@@ -2579,11 +2704,6 @@ int btrfs_update_inode(struct btrfs_trans_handle *trans,
 int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode);
 int btrfs_orphan_del(struct btrfs_trans_handle *trans, struct inode *inode);
 int btrfs_orphan_cleanup(struct btrfs_root *root);
-void btrfs_orphan_pre_snapshot(struct btrfs_trans_handle *trans,
-				struct btrfs_pending_snapshot *pending,
-				u64 *bytes_to_reserve);
-void btrfs_orphan_post_snapshot(struct btrfs_trans_handle *trans,
-				struct btrfs_pending_snapshot *pending);
 void btrfs_orphan_commit_root(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root);
 int btrfs_cont_expand(struct inode *inode, loff_t oldsize, loff_t size);
@@ -2697,4 +2817,20 @@ int btrfs_scrub_cancel_devid(struct btrfs_root *root, u64 devid);
 int btrfs_scrub_progress(struct btrfs_root *root, u64 devid,
 			 struct btrfs_scrub_progress *progress);
 
+/* reada.c */
+struct reada_control {
+	struct btrfs_root	*root;		/* tree to prefetch */
+	struct btrfs_key	key_start;
+	struct btrfs_key	key_end;	/* exclusive */
+	atomic_t		elems;
+	struct kref		refcnt;
+	wait_queue_head_t	wait;
+};
+struct reada_control *btrfs_reada_add(struct btrfs_root *root,
+			      struct btrfs_key *start, struct btrfs_key *end);
+int btrfs_reada_wait(void *handle);
+void btrfs_reada_detach(void *handle);
+int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			 u64 start, int err);
+
 #endif

fs/btrfs/delayed-inode.c

@@ -591,7 +591,7 @@ static int btrfs_delayed_item_reserve_metadata(struct btrfs_trans_handle *trans,
 		return 0;
 
 	src_rsv = trans->block_rsv;
-	dst_rsv = &root->fs_info->global_block_rsv;
+	dst_rsv = &root->fs_info->delayed_block_rsv;
 
 	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
@@ -609,7 +609,7 @@ static void btrfs_delayed_item_release_metadata(struct btrfs_root *root,
 	if (!item->bytes_reserved)
 		return;
 
-	rsv = &root->fs_info->global_block_rsv;
+	rsv = &root->fs_info->delayed_block_rsv;
 	btrfs_block_rsv_release(root, rsv,
 				item->bytes_reserved);
 }
@@ -624,13 +624,36 @@ static int btrfs_delayed_inode_reserve_metadata(
 	u64 num_bytes;
 	int ret;
 
-	if (!trans->bytes_reserved)
-		return 0;
-
 	src_rsv = trans->block_rsv;
-	dst_rsv = &root->fs_info->global_block_rsv;
+	dst_rsv = &root->fs_info->delayed_block_rsv;
 
 	num_bytes = btrfs_calc_trans_metadata_size(root, 1);
+
+	/*
+	 * btrfs_dirty_inode will update the inode under btrfs_join_transaction
+	 * which doesn't reserve space for speed.  This is a problem since we
+	 * still need to reserve space for this update, so try to reserve the
+	 * space.
+	 *
+	 * Now if src_rsv == delalloc_block_rsv we'll let it just steal since
+	 * we're accounted for.
+	 */
+	if (!trans->bytes_reserved &&
+	    src_rsv != &root->fs_info->delalloc_block_rsv) {
+		ret = btrfs_block_rsv_add_noflush(root, dst_rsv, num_bytes);
+		/*
+		 * Since we're under a transaction reserve_metadata_bytes could
+		 * try to commit the transaction which will make it return
+		 * EAGAIN to make us stop the transaction we have, so return
+		 * ENOSPC instead so that btrfs_dirty_inode knows what to do.
+		 */
+		if (ret == -EAGAIN)
+			ret = -ENOSPC;
+		if (!ret)
+			node->bytes_reserved = num_bytes;
+		return ret;
+	}
+
 	ret = btrfs_block_rsv_migrate(src_rsv, dst_rsv, num_bytes);
 	if (!ret)
 		node->bytes_reserved = num_bytes;
@@ -646,7 +669,7 @@ static void btrfs_delayed_inode_release_metadata(struct btrfs_root *root,
 	if (!node->bytes_reserved)
 		return;
 
-	rsv = &root->fs_info->global_block_rsv;
+	rsv = &root->fs_info->delayed_block_rsv;
 	btrfs_block_rsv_release(root, rsv,
 				node->bytes_reserved);
 	node->bytes_reserved = 0;
@@ -1026,7 +1049,7 @@ int btrfs_run_delayed_items(struct btrfs_trans_handle *trans,
 	path->leave_spinning = 1;
 
 	block_rsv = trans->block_rsv;
-	trans->block_rsv = &root->fs_info->global_block_rsv;
+	trans->block_rsv = &root->fs_info->delayed_block_rsv;
 
 	delayed_root = btrfs_get_delayed_root(root);
 
@@ -1069,7 +1092,7 @@ static int __btrfs_commit_inode_delayed_items(struct btrfs_trans_handle *trans,
 	path->leave_spinning = 1;
 
 	block_rsv = trans->block_rsv;
-	trans->block_rsv = &node->root->fs_info->global_block_rsv;
+	trans->block_rsv = &node->root->fs_info->delayed_block_rsv;
 
 	ret = btrfs_insert_delayed_items(trans, path, node->root, node);
 	if (!ret)
@@ -1149,7 +1172,7 @@ static void btrfs_async_run_delayed_node_done(struct btrfs_work *work)
 		goto free_path;
 
 	block_rsv = trans->block_rsv;
-	trans->block_rsv = &root->fs_info->global_block_rsv;
+	trans->block_rsv = &root->fs_info->delayed_block_rsv;
 
 	ret = btrfs_insert_delayed_items(trans, path, root, delayed_node);
 	if (!ret)
@@ -1686,11 +1709,8 @@ int btrfs_delayed_update_inode(struct btrfs_trans_handle *trans,
 	}
 
 	ret = btrfs_delayed_inode_reserve_metadata(trans, root, delayed_node);
-	/*
-	 * we must reserve enough space when we start a new transaction,
-	 * so reserving metadata failure is impossible
-	 */
-	BUG_ON(ret);
+	if (ret)
+		goto release_node;
 
 	fill_stack_inode_item(trans, &delayed_node->inode_item, inode);
 	delayed_node->inode_dirty = 1;

fs/btrfs/disk-io.c

@@ -256,8 +256,7 @@ void btrfs_csum_final(u32 crc, char *result)
 static int csum_tree_block(struct btrfs_root *root, struct extent_buffer *buf,
 			   int verify)
 {
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	char *result = NULL;
 	unsigned long len;
 	unsigned long cur_len;
@@ -367,7 +366,8 @@ static int btree_read_extent_buffer_pages(struct btrfs_root *root,
 	clear_bit(EXTENT_BUFFER_CORRUPT, &eb->bflags);
 	io_tree = &BTRFS_I(root->fs_info->btree_inode)->io_tree;
 	while (1) {
-		ret = read_extent_buffer_pages(io_tree, eb, start, 1,
+		ret = read_extent_buffer_pages(io_tree, eb, start,
+					       WAIT_COMPLETE,
 					       btree_get_extent, mirror_num);
 		if (!ret &&
 		    !verify_parent_transid(io_tree, eb, parent_transid))
@@ -608,11 +608,48 @@ static int btree_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 	end = min_t(u64, eb->len, PAGE_CACHE_SIZE);
 	end = eb->start + end - 1;
 err:
+	if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
+		clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
+		btree_readahead_hook(root, eb, eb->start, ret);
+	}
+
 	free_extent_buffer(eb);
 out:
 	return ret;
 }
 
+static int btree_io_failed_hook(struct bio *failed_bio,
+			 struct page *page, u64 start, u64 end,
+			 u64 mirror_num, struct extent_state *state)
+{
+	struct extent_io_tree *tree;
+	unsigned long len;
+	struct extent_buffer *eb;
+	struct btrfs_root *root = BTRFS_I(page->mapping->host)->root;
+
+	tree = &BTRFS_I(page->mapping->host)->io_tree;
+	if (page->private == EXTENT_PAGE_PRIVATE)
+		goto out;
+	if (!page->private)
+		goto out;
+
+	len = page->private >> 2;
+	WARN_ON(len == 0);
+
+	eb = alloc_extent_buffer(tree, start, len, page);
+	if (eb == NULL)
+		goto out;
+
+	if (test_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags)) {
+		clear_bit(EXTENT_BUFFER_READAHEAD, &eb->bflags);
+		btree_readahead_hook(root, eb, eb->start, -EIO);
+	}
+	free_extent_buffer(eb);
+
+out:
+	return -EIO;	/* we fixed nothing */
+}
+
 static void end_workqueue_bio(struct bio *bio, int err)
 {
 	struct end_io_wq *end_io_wq = bio->bi_private;
@@ -908,7 +945,7 @@ static int btree_readpage(struct file *file, struct page *page)
 {
 	struct extent_io_tree *tree;
 	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	return extent_read_full_page(tree, page, btree_get_extent);
+	return extent_read_full_page(tree, page, btree_get_extent, 0);
 }
 
 static int btree_releasepage(struct page *page, gfp_t gfp_flags)
@@ -974,11 +1011,43 @@ int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
 	if (!buf)
 		return 0;
 	read_extent_buffer_pages(&BTRFS_I(btree_inode)->io_tree,
-				 buf, 0, 0, btree_get_extent, 0);
+				 buf, 0, WAIT_NONE, btree_get_extent, 0);
 	free_extent_buffer(buf);
 	return ret;
 }
 
+int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 int mirror_num, struct extent_buffer **eb)
+{
+	struct extent_buffer *buf = NULL;
+	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct extent_io_tree *io_tree = &BTRFS_I(btree_inode)->io_tree;
+	int ret;
+
+	buf = btrfs_find_create_tree_block(root, bytenr, blocksize);
+	if (!buf)
+		return 0;
+
+	set_bit(EXTENT_BUFFER_READAHEAD, &buf->bflags);
+
+	ret = read_extent_buffer_pages(io_tree, buf, 0, WAIT_PAGE_LOCK,
+				       btree_get_extent, mirror_num);
+	if (ret) {
+		free_extent_buffer(buf);
+		return ret;
+	}
+
+	if (test_bit(EXTENT_BUFFER_CORRUPT, &buf->bflags)) {
+		free_extent_buffer(buf);
+		return -EIO;
+	} else if (extent_buffer_uptodate(io_tree, buf, NULL)) {
+		*eb = buf;
+	} else {
+		free_extent_buffer(buf);
+	}
+	return 0;
+}
+
 struct extent_buffer *btrfs_find_tree_block(struct btrfs_root *root,
 					    u64 bytenr, u32 blocksize)
 {
@@ -1135,10 +1204,12 @@ static int find_and_setup_root(struct btrfs_root *tree_root,
 
 	generation = btrfs_root_generation(&root->root_item);
 	blocksize = btrfs_level_size(root, btrfs_root_level(&root->root_item));
+	root->commit_root = NULL;
 	root->node = read_tree_block(root, btrfs_root_bytenr(&root->root_item),
 				     blocksize, generation);
 	if (!root->node || !btrfs_buffer_uptodate(root->node, generation)) {
 		free_extent_buffer(root->node);
+		root->node = NULL;
 		return -EIO;
 	}
 	root->commit_root = btrfs_root_node(root);
@@ -1577,6 +1648,235 @@ sleep:
 	return 0;
 }
 
+/*
+ * this will find the highest generation in the array of
+ * root backups.  The index of the highest array is returned,
+ * or -1 if we can't find anything.
+ *
+ * We check to make sure the array is valid by comparing the
+ * generation of the latest  root in the array with the generation
+ * in the super block.  If they don't match we pitch it.
+ */
+static int find_newest_super_backup(struct btrfs_fs_info *info, u64 newest_gen)
+{
+	u64 cur;
+	int newest_index = -1;
+	struct btrfs_root_backup *root_backup;
+	int i;
+
+	for (i = 0; i < BTRFS_NUM_BACKUP_ROOTS; i++) {
+		root_backup = info->super_copy->super_roots + i;
+		cur = btrfs_backup_tree_root_gen(root_backup);
+		if (cur == newest_gen)
+			newest_index = i;
+	}
+
+	/* check to see if we actually wrapped around */
+	if (newest_index == BTRFS_NUM_BACKUP_ROOTS - 1) {
+		root_backup = info->super_copy->super_roots;
+		cur = btrfs_backup_tree_root_gen(root_backup);
+		if (cur == newest_gen)
+			newest_index = 0;
+	}
+	return newest_index;
+}
+
+
+/*
+ * find the oldest backup so we know where to store new entries
+ * in the backup array.  This will set the backup_root_index
+ * field in the fs_info struct
+ */
+static void find_oldest_super_backup(struct btrfs_fs_info *info,
+				     u64 newest_gen)
+{
+	int newest_index = -1;
+
+	newest_index = find_newest_super_backup(info, newest_gen);
+	/* if there was garbage in there, just move along */
+	if (newest_index == -1) {
+		info->backup_root_index = 0;
+	} else {
+		info->backup_root_index = (newest_index + 1) % BTRFS_NUM_BACKUP_ROOTS;
+	}
+}
+
+/*
+ * copy all the root pointers into the super backup array.
+ * this will bump the backup pointer by one when it is
+ * done
+ */
+static void backup_super_roots(struct btrfs_fs_info *info)
+{
+	int next_backup;
+	struct btrfs_root_backup *root_backup;
+	int last_backup;
+
+	next_backup = info->backup_root_index;
+	last_backup = (next_backup + BTRFS_NUM_BACKUP_ROOTS - 1) %
+		BTRFS_NUM_BACKUP_ROOTS;
+
+	/*
+	 * just overwrite the last backup if we're at the same generation
+	 * this happens only at umount
+	 */
+	root_backup = info->super_for_commit->super_roots + last_backup;
+	if (btrfs_backup_tree_root_gen(root_backup) ==
+	    btrfs_header_generation(info->tree_root->node))
+		next_backup = last_backup;
+
+	root_backup = info->super_for_commit->super_roots + next_backup;
+
+	/*
+	 * make sure all of our padding and empty slots get zero filled
+	 * regardless of which ones we use today
+	 */
+	memset(root_backup, 0, sizeof(*root_backup));
+
+	info->backup_root_index = (next_backup + 1) % BTRFS_NUM_BACKUP_ROOTS;
+
+	btrfs_set_backup_tree_root(root_backup, info->tree_root->node->start);
+	btrfs_set_backup_tree_root_gen(root_backup,
+			       btrfs_header_generation(info->tree_root->node));
+
+	btrfs_set_backup_tree_root_level(root_backup,
+			       btrfs_header_level(info->tree_root->node));
+
+	btrfs_set_backup_chunk_root(root_backup, info->chunk_root->node->start);
+	btrfs_set_backup_chunk_root_gen(root_backup,
+			       btrfs_header_generation(info->chunk_root->node));
+	btrfs_set_backup_chunk_root_level(root_backup,
+			       btrfs_header_level(info->chunk_root->node));
+
+	btrfs_set_backup_extent_root(root_backup, info->extent_root->node->start);
+	btrfs_set_backup_extent_root_gen(root_backup,
+			       btrfs_header_generation(info->extent_root->node));
+	btrfs_set_backup_extent_root_level(root_backup,
+			       btrfs_header_level(info->extent_root->node));
+
+	/*
+	 * we might commit during log recovery, which happens before we set
+	 * the fs_root.  Make sure it is valid before we fill it in.
+	 */
+	if (info->fs_root && info->fs_root->node) {
+		btrfs_set_backup_fs_root(root_backup,
+					 info->fs_root->node->start);
+		btrfs_set_backup_fs_root_gen(root_backup,
+			       btrfs_header_generation(info->fs_root->node));
+		btrfs_set_backup_fs_root_level(root_backup,
+			       btrfs_header_level(info->fs_root->node));
+	}
+
+	btrfs_set_backup_dev_root(root_backup, info->dev_root->node->start);
+	btrfs_set_backup_dev_root_gen(root_backup,
+			       btrfs_header_generation(info->dev_root->node));
+	btrfs_set_backup_dev_root_level(root_backup,
+				       btrfs_header_level(info->dev_root->node));
+
+	btrfs_set_backup_csum_root(root_backup, info->csum_root->node->start);
+	btrfs_set_backup_csum_root_gen(root_backup,
+			       btrfs_header_generation(info->csum_root->node));
+	btrfs_set_backup_csum_root_level(root_backup,
+			       btrfs_header_level(info->csum_root->node));
+
+	btrfs_set_backup_total_bytes(root_backup,
+			     btrfs_super_total_bytes(info->super_copy));
+	btrfs_set_backup_bytes_used(root_backup,
+			     btrfs_super_bytes_used(info->super_copy));
+	btrfs_set_backup_num_devices(root_backup,
+			     btrfs_super_num_devices(info->super_copy));
+
+	/*
+	 * if we don't copy this out to the super_copy, it won't get remembered
+	 * for the next commit
+	 */
+	memcpy(&info->super_copy->super_roots,
+	       &info->super_for_commit->super_roots,
+	       sizeof(*root_backup) * BTRFS_NUM_BACKUP_ROOTS);
+}
+
+/*
+ * this copies info out of the root backup array and back into
+ * the in-memory super block.  It is meant to help iterate through
+ * the array, so you send it the number of backups you've already
+ * tried and the last backup index you used.
+ *
+ * this returns -1 when it has tried all the backups
+ */
+static noinline int next_root_backup(struct btrfs_fs_info *info,
+				     struct btrfs_super_block *super,
+				     int *num_backups_tried, int *backup_index)
+{
+	struct btrfs_root_backup *root_backup;
+	int newest = *backup_index;
+
+	if (*num_backups_tried == 0) {
+		u64 gen = btrfs_super_generation(super);
+
+		newest = find_newest_super_backup(info, gen);
+		if (newest == -1)
+			return -1;
+
+		*backup_index = newest;
+		*num_backups_tried = 1;
+	} else if (*num_backups_tried == BTRFS_NUM_BACKUP_ROOTS) {
+		/* we've tried all the backups, all done */
+		return -1;
+	} else {
+		/* jump to the next oldest backup */
+		newest = (*backup_index + BTRFS_NUM_BACKUP_ROOTS - 1) %
+			BTRFS_NUM_BACKUP_ROOTS;
+		*backup_index = newest;
+		*num_backups_tried += 1;
+	}
+	root_backup = super->super_roots + newest;
+
+	btrfs_set_super_generation(super,
+				   btrfs_backup_tree_root_gen(root_backup));
+	btrfs_set_super_root(super, btrfs_backup_tree_root(root_backup));
+	btrfs_set_super_root_level(super,
+				   btrfs_backup_tree_root_level(root_backup));
+	btrfs_set_super_bytes_used(super, btrfs_backup_bytes_used(root_backup));
+
+	/*
+	 * fixme: the total bytes and num_devices need to match or we should
+	 * need a fsck
+	 */
+	btrfs_set_super_total_bytes(super, btrfs_backup_total_bytes(root_backup));
+	btrfs_set_super_num_devices(super, btrfs_backup_num_devices(root_backup));
+	return 0;
+}
+
+/* helper to cleanup tree roots */
+static void free_root_pointers(struct btrfs_fs_info *info, int chunk_root)
+{
+	free_extent_buffer(info->tree_root->node);
+	free_extent_buffer(info->tree_root->commit_root);
+	free_extent_buffer(info->dev_root->node);
+	free_extent_buffer(info->dev_root->commit_root);
+	free_extent_buffer(info->extent_root->node);
+	free_extent_buffer(info->extent_root->commit_root);
+	free_extent_buffer(info->csum_root->node);
+	free_extent_buffer(info->csum_root->commit_root);
+
+	info->tree_root->node = NULL;
+	info->tree_root->commit_root = NULL;
+	info->dev_root->node = NULL;
+	info->dev_root->commit_root = NULL;
+	info->extent_root->node = NULL;
+	info->extent_root->commit_root = NULL;
+	info->csum_root->node = NULL;
+	info->csum_root->commit_root = NULL;
+
+	if (chunk_root) {
+		free_extent_buffer(info->chunk_root->node);
+		free_extent_buffer(info->chunk_root->commit_root);
+		info->chunk_root->node = NULL;
+		info->chunk_root->commit_root = NULL;
+	}
+}
+
+
 struct btrfs_root *open_ctree(struct super_block *sb,
 			      struct btrfs_fs_devices *fs_devices,
 			      char *options)
@@ -1604,6 +1904,8 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	int ret;
 	int err = -EINVAL;
+	int num_backups_tried = 0;
+	int backup_index = 0;
 
 	struct btrfs_super_block *disk_super;
 
@@ -1648,6 +1950,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	spin_lock_init(&fs_info->fs_roots_radix_lock);
 	spin_lock_init(&fs_info->delayed_iput_lock);
 	spin_lock_init(&fs_info->defrag_inodes_lock);
+	spin_lock_init(&fs_info->free_chunk_lock);
 	mutex_init(&fs_info->reloc_mutex);
 
 	init_completion(&fs_info->kobj_unregister);
@@ -1665,8 +1968,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_init_block_rsv(&fs_info->trans_block_rsv);
 	btrfs_init_block_rsv(&fs_info->chunk_block_rsv);
 	btrfs_init_block_rsv(&fs_info->empty_block_rsv);
-	INIT_LIST_HEAD(&fs_info->durable_block_rsv_list);
-	mutex_init(&fs_info->durable_block_rsv_mutex);
+	btrfs_init_block_rsv(&fs_info->delayed_block_rsv);
 	atomic_set(&fs_info->nr_async_submits, 0);
 	atomic_set(&fs_info->async_delalloc_pages, 0);
 	atomic_set(&fs_info->async_submit_draining, 0);
@@ -1677,6 +1979,11 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	fs_info->metadata_ratio = 0;
 	fs_info->defrag_inodes = RB_ROOT;
 	fs_info->trans_no_join = 0;
+	fs_info->free_chunk_space = 0;
+
+	/* readahead state */
+	INIT_RADIX_TREE(&fs_info->reada_tree, GFP_NOFS & ~__GFP_WAIT);
+	spin_lock_init(&fs_info->reada_lock);
 
 	fs_info->thread_pool_size = min_t(unsigned long,
 					  num_online_cpus() + 2, 8);
@@ -1766,14 +2073,14 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 		goto fail_alloc;
 	}
 
-	memcpy(&fs_info->super_copy, bh->b_data, sizeof(fs_info->super_copy));
-	memcpy(&fs_info->super_for_commit, &fs_info->super_copy,
-	       sizeof(fs_info->super_for_commit));
+	memcpy(fs_info->super_copy, bh->b_data, sizeof(*fs_info->super_copy));
+	memcpy(fs_info->super_for_commit, fs_info->super_copy,
+	       sizeof(*fs_info->super_for_commit));
 	brelse(bh);
 
-	memcpy(fs_info->fsid, fs_info->super_copy.fsid, BTRFS_FSID_SIZE);
+	memcpy(fs_info->fsid, fs_info->super_copy->fsid, BTRFS_FSID_SIZE);
 
-	disk_super = &fs_info->super_copy;
+	disk_super = fs_info->super_copy;
 	if (!btrfs_super_root(disk_super))
 		goto fail_alloc;
 
@@ -1782,6 +2089,13 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	btrfs_check_super_valid(fs_info, sb->s_flags & MS_RDONLY);
 
+	/*
+	 * run through our array of backup supers and setup
+	 * our ring pointer to the oldest one
+	 */
+	generation = btrfs_super_generation(disk_super);
+	find_oldest_super_backup(fs_info, generation);
+
 	/*
 	 * In the long term, we'll store the compression type in the super
 	 * block, and it'll be used for per file compression control.
@@ -1870,6 +2184,9 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_init_workers(&fs_info->delayed_workers, "delayed-meta",
 			   fs_info->thread_pool_size,
 			   &fs_info->generic_worker);
+	btrfs_init_workers(&fs_info->readahead_workers, "readahead",
+			   fs_info->thread_pool_size,
+			   &fs_info->generic_worker);
 
 	/*
 	 * endios are largely parallel and should have a very
@@ -1880,6 +2197,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 
 	fs_info->endio_write_workers.idle_thresh = 2;
 	fs_info->endio_meta_write_workers.idle_thresh = 2;
+	fs_info->readahead_workers.idle_thresh = 2;
 
 	btrfs_start_workers(&fs_info->workers, 1);
 	btrfs_start_workers(&fs_info->generic_worker, 1);
@@ -1893,6 +2211,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	btrfs_start_workers(&fs_info->endio_freespace_worker, 1);
 	btrfs_start_workers(&fs_info->delayed_workers, 1);
 	btrfs_start_workers(&fs_info->caching_workers, 1);
+	btrfs_start_workers(&fs_info->readahead_workers, 1);
 
 	fs_info->bdi.ra_pages *= btrfs_super_num_devices(disk_super);
 	fs_info->bdi.ra_pages = max(fs_info->bdi.ra_pages,
@@ -1939,7 +2258,7 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (!test_bit(EXTENT_BUFFER_UPTODATE, &chunk_root->node->bflags)) {
 		printk(KERN_WARNING "btrfs: failed to read chunk root on %s\n",
 		       sb->s_id);
-		goto fail_chunk_root;
+		goto fail_tree_roots;
 	}
 	btrfs_set_root_node(&chunk_root->root_item, chunk_root->node);
 	chunk_root->commit_root = btrfs_root_node(chunk_root);
@@ -1954,11 +2273,12 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	if (ret) {
 		printk(KERN_WARNING "btrfs: failed to read chunk tree on %s\n",
 		       sb->s_id);
-		goto fail_chunk_root;
+		goto fail_tree_roots;
 	}
 
 	btrfs_close_extra_devices(fs_devices);
 
+retry_root_backup:
 	blocksize = btrfs_level_size(tree_root,
 				     btrfs_super_root_level(disk_super));
 	generation = btrfs_super_generation(disk_super);
@@ -1966,32 +2286,33 @@ struct btrfs_root *open_ctree(struct super_block *sb,
 	tree_root->node = read_tree_block(tree_root,
 					  btrfs_super_root(disk_super),
 					  blocksize, generation);
-	if (!tree_root->node)
-		goto fail_chunk_root;
-	if (!test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
+	if (!tree_root->node ||
+	    !test_bit(EXTENT_BUFFER_UPTODATE, &tree_root->node->bflags)) {
 		printk(KERN_WARNING "btrfs: failed to read tree root on %s\n",
 		       sb->s_id);
-		goto fail_tree_root;
+
+		goto recovery_tree_root;
 	}
+
 	btrfs_set_root_node(&tree_root->root_item, tree_root->node);
 	tree_root->commit_root = btrfs_root_node(tree_root);
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_EXTENT_TREE_OBJECTID, extent_root);
 	if (ret)
-		goto fail_tree_root;
+		goto recovery_tree_root;
 	extent_root->track_dirty = 1;
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_DEV_TREE_OBJECTID, dev_root);
 	if (ret)
-		goto fail_extent_root;
+		goto recovery_tree_root;
 	dev_root->track_dirty = 1;
 
 	ret = find_and_setup_root(tree_root, fs_info,
 				  BTRFS_CSUM_TREE_OBJECTID, csum_root);
 	if (ret)
-		goto fail_dev_root;
+		goto recovery_tree_root;
 
 	csum_root->track_dirty = 1;
 
@@ -2124,22 +2445,13 @@ fail_cleaner:
 
 fail_block_groups:
 	btrfs_free_block_groups(fs_info);
-	free_extent_buffer(csum_root->node);
-	free_extent_buffer(csum_root->commit_root);
-fail_dev_root:
-	free_extent_buffer(dev_root->node);
-	free_extent_buffer(dev_root->commit_root);
-fail_extent_root:
-	free_extent_buffer(extent_root->node);
-	free_extent_buffer(extent_root->commit_root);
-fail_tree_root:
-	free_extent_buffer(tree_root->node);
-	free_extent_buffer(tree_root->commit_root);
-fail_chunk_root:
-	free_extent_buffer(chunk_root->node);
-	free_extent_buffer(chunk_root->commit_root);
+
+fail_tree_roots:
+	free_root_pointers(fs_info, 1);
+
 fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->generic_worker);
+	btrfs_stop_workers(&fs_info->readahead_workers);
 	btrfs_stop_workers(&fs_info->fixup_workers);
 	btrfs_stop_workers(&fs_info->delalloc_workers);
 	btrfs_stop_workers(&fs_info->workers);
@@ -2152,7 +2464,6 @@ fail_sb_buffer:
 	btrfs_stop_workers(&fs_info->delayed_workers);
 	btrfs_stop_workers(&fs_info->caching_workers);
 fail_alloc:
-	kfree(fs_info->delayed_root);
 fail_iput:
 	invalidate_inode_pages2(fs_info->btree_inode->i_mapping);
 	iput(fs_info->btree_inode);
@@ -2164,13 +2475,27 @@ fail_bdi:
 fail_srcu:
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 fail:
-	kfree(extent_root);
-	kfree(tree_root);
-	kfree(fs_info);
-	kfree(chunk_root);
-	kfree(dev_root);
-	kfree(csum_root);
+	free_fs_info(fs_info);
 	return ERR_PTR(err);
+
+recovery_tree_root:
+
+	if (!btrfs_test_opt(tree_root, RECOVERY))
+		goto fail_tree_roots;
+
+	free_root_pointers(fs_info, 0);
+
+	/* don't use the log in recovery mode, it won't be valid */
+	btrfs_set_super_log_root(disk_super, 0);
+
+	/* we can't trust the free space cache either */
+	btrfs_set_opt(fs_info->mount_opt, CLEAR_CACHE);
+
+	ret = next_root_backup(fs_info, fs_info->super_copy,
+			       &num_backups_tried, &backup_index);
+	if (ret == -1)
+		goto fail_block_groups;
+	goto retry_root_backup;
 }
 
 static void btrfs_end_buffer_write_sync(struct buffer_head *bh, int uptodate)
@@ -2338,10 +2663,11 @@ int write_all_supers(struct btrfs_root *root, int max_mirrors)
 	int total_errors = 0;
 	u64 flags;
 
-	max_errors = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
+	max_errors = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
 	do_barriers = !btrfs_test_opt(root, NOBARRIER);
+	backup_super_roots(root->fs_info);
 
-	sb = &root->fs_info->super_for_commit;
+	sb = root->fs_info->super_for_commit;
 	dev_item = &sb->dev_item;
 
 	mutex_lock(&root->fs_info->fs_devices->device_list_mutex);
@@ -2545,8 +2871,6 @@ int close_ctree(struct btrfs_root *root)
 	/* clear out the rbtree of defraggable inodes */
 	btrfs_run_defrag_inodes(root->fs_info);
 
-	btrfs_put_block_group_cache(fs_info);
-
 	/*
 	 * Here come 2 situations when btrfs is broken to flip readonly:
 	 *
@@ -2572,6 +2896,8 @@ int close_ctree(struct btrfs_root *root)
 			printk(KERN_ERR "btrfs: commit super ret %d\n", ret);
 	}
 
+	btrfs_put_block_group_cache(fs_info);
+
 	kthread_stop(root->fs_info->transaction_kthread);
 	kthread_stop(root->fs_info->cleaner_kthread);
 
@@ -2603,7 +2929,6 @@ int close_ctree(struct btrfs_root *root)
 	del_fs_roots(fs_info);
 
 	iput(fs_info->btree_inode);
-	kfree(fs_info->delayed_root);
 
 	btrfs_stop_workers(&fs_info->generic_worker);
 	btrfs_stop_workers(&fs_info->fixup_workers);
@@ -2617,6 +2942,7 @@ int close_ctree(struct btrfs_root *root)
 	btrfs_stop_workers(&fs_info->submit_workers);
 	btrfs_stop_workers(&fs_info->delayed_workers);
 	btrfs_stop_workers(&fs_info->caching_workers);
+	btrfs_stop_workers(&fs_info->readahead_workers);
 
 	btrfs_close_devices(fs_info->fs_devices);
 	btrfs_mapping_tree_free(&fs_info->mapping_tree);
@@ -2624,12 +2950,7 @@ int close_ctree(struct btrfs_root *root)
 	bdi_destroy(&fs_info->bdi);
 	cleanup_srcu_struct(&fs_info->subvol_srcu);
 
-	kfree(fs_info->extent_root);
-	kfree(fs_info->tree_root);
-	kfree(fs_info->chunk_root);
-	kfree(fs_info->dev_root);
-	kfree(fs_info->csum_root);
-	kfree(fs_info);
+	free_fs_info(fs_info);
 
 	return 0;
 }
@@ -2735,7 +3056,8 @@ int btrfs_read_buffer(struct extent_buffer *buf, u64 parent_transid)
 	return ret;
 }
 
-int btree_lock_page_hook(struct page *page)
+static int btree_lock_page_hook(struct page *page, void *data,
+				void (*flush_fn)(void *))
 {
 	struct inode *inode = page->mapping->host;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
@@ -2752,7 +3074,10 @@ int btree_lock_page_hook(struct page *page)
 	if (!eb)
 		goto out;
 
-	btrfs_tree_lock(eb);
+	if (!btrfs_try_tree_write_lock(eb)) {
+		flush_fn(data);
+		btrfs_tree_lock(eb);
+	}
 	btrfs_set_header_flag(eb, BTRFS_HEADER_FLAG_WRITTEN);
 
 	if (test_and_clear_bit(EXTENT_BUFFER_DIRTY, &eb->bflags)) {
@@ -2767,7 +3092,10 @@ int btree_lock_page_hook(struct page *page)
 	btrfs_tree_unlock(eb);
 	free_extent_buffer(eb);
 out:
-	lock_page(page);
+	if (!trylock_page(page)) {
+		flush_fn(data);
+		lock_page(page);
+	}
 	return 0;
 }
 
@@ -3123,6 +3451,7 @@ static int btrfs_cleanup_transaction(struct btrfs_root *root)
 static struct extent_io_ops btree_extent_io_ops = {
 	.write_cache_pages_lock_hook = btree_lock_page_hook,
 	.readpage_end_io_hook = btree_readpage_end_io_hook,
+	.readpage_io_failed_hook = btree_io_failed_hook,
 	.submit_bio_hook = btree_submit_bio_hook,
 	/* note we're sharing with inode.c for the merge bio hook */
 	.merge_bio_hook = btrfs_merge_bio_hook,

fs/btrfs/disk-io.h

@@ -40,6 +40,8 @@ struct extent_buffer *read_tree_block(struct btrfs_root *root, u64 bytenr,
 				      u32 blocksize, u64 parent_transid);
 int readahead_tree_block(struct btrfs_root *root, u64 bytenr, u32 blocksize,
 			 u64 parent_transid);
+int reada_tree_block_flagged(struct btrfs_root *root, u64 bytenr, u32 blocksize,
+			 int mirror_num, struct extent_buffer **eb);
 struct extent_buffer *btrfs_find_create_tree_block(struct btrfs_root *root,
 						   u64 bytenr, u32 blocksize);
 int clean_tree_block(struct btrfs_trans_handle *trans,
@@ -83,8 +85,6 @@ int btrfs_init_log_root_tree(struct btrfs_trans_handle *trans,
 			     struct btrfs_fs_info *fs_info);
 int btrfs_add_log_tree(struct btrfs_trans_handle *trans,
 		       struct btrfs_root *root);
-int btree_lock_page_hook(struct page *page);
-
 
 #ifdef CONFIG_DEBUG_LOCK_ALLOC
 void btrfs_init_lockdep(void);

fs/btrfs/extent_io.c

@@ -17,6 +17,7 @@
 #include "compat.h"
 #include "ctree.h"
 #include "btrfs_inode.h"
+#include "volumes.h"
 
 static struct kmem_cache *extent_state_cache;
 static struct kmem_cache *extent_buffer_cache;
@@ -894,6 +895,194 @@ search_again:
 	goto again;
 }
 
+/**
+ * convert_extent - convert all bits in a given range from one bit to another
+ * @tree:	the io tree to search
+ * @start:	the start offset in bytes
+ * @end:	the end offset in bytes (inclusive)
+ * @bits:	the bits to set in this range
+ * @clear_bits:	the bits to clear in this range
+ * @mask:	the allocation mask
+ *
+ * This will go through and set bits for the given range.  If any states exist
+ * already in this range they are set with the given bit and cleared of the
+ * clear_bits.  This is only meant to be used by things that are mergeable, ie
+ * converting from say DELALLOC to DIRTY.  This is not meant to be used with
+ * boundary bits like LOCK.
+ */
+int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		       int bits, int clear_bits, gfp_t mask)
+{
+	struct extent_state *state;
+	struct extent_state *prealloc = NULL;
+	struct rb_node *node;
+	int err = 0;
+	u64 last_start;
+	u64 last_end;
+
+again:
+	if (!prealloc && (mask & __GFP_WAIT)) {
+		prealloc = alloc_extent_state(mask);
+		if (!prealloc)
+			return -ENOMEM;
+	}
+
+	spin_lock(&tree->lock);
+	/*
+	 * this search will find all the extents that end after
+	 * our range starts.
+	 */
+	node = tree_search(tree, start);
+	if (!node) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc)
+			return -ENOMEM;
+		err = insert_state(tree, prealloc, start, end, &bits);
+		prealloc = NULL;
+		BUG_ON(err == -EEXIST);
+		goto out;
+	}
+	state = rb_entry(node, struct extent_state, rb_node);
+hit_next:
+	last_start = state->start;
+	last_end = state->end;
+
+	/*
+	 * | ---- desired range ---- |
+	 * | state |
+	 *
+	 * Just lock what we found and keep going
+	 */
+	if (state->start == start && state->end <= end) {
+		struct rb_node *next_node;
+
+		set_state_bits(tree, state, &bits);
+		clear_state_bit(tree, state, &clear_bits, 0);
+
+		merge_state(tree, state);
+		if (last_end == (u64)-1)
+			goto out;
+
+		start = last_end + 1;
+		next_node = rb_next(&state->rb_node);
+		if (next_node && start < end && prealloc && !need_resched()) {
+			state = rb_entry(next_node, struct extent_state,
+					 rb_node);
+			if (state->start == start)
+				goto hit_next;
+		}
+		goto search_again;
+	}
+
+	/*
+	 *     | ---- desired range ---- |
+	 * | state |
+	 *   or
+	 * | ------------- state -------------- |
+	 *
+	 * We need to split the extent we found, and may flip bits on
+	 * second half.
+	 *
+	 * If the extent we found extends past our
+	 * range, we just split and search again.  It'll get split
+	 * again the next time though.
+	 *
+	 * If the extent we found is inside our range, we set the
+	 * desired bit on it.
+	 */
+	if (state->start < start) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc)
+			return -ENOMEM;
+		err = split_state(tree, state, prealloc, start);
+		BUG_ON(err == -EEXIST);
+		prealloc = NULL;
+		if (err)
+			goto out;
+		if (state->end <= end) {
+			set_state_bits(tree, state, &bits);
+			clear_state_bit(tree, state, &clear_bits, 0);
+			merge_state(tree, state);
+			if (last_end == (u64)-1)
+				goto out;
+			start = last_end + 1;
+		}
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *     | state | or               | state |
+	 *
+	 * There's a hole, we need to insert something in it and
+	 * ignore the extent we found.
+	 */
+	if (state->start > start) {
+		u64 this_end;
+		if (end < last_start)
+			this_end = end;
+		else
+			this_end = last_start - 1;
+
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc)
+			return -ENOMEM;
+
+		/*
+		 * Avoid to free 'prealloc' if it can be merged with
+		 * the later extent.
+		 */
+		err = insert_state(tree, prealloc, start, this_end,
+				   &bits);
+		BUG_ON(err == -EEXIST);
+		if (err) {
+			free_extent_state(prealloc);
+			prealloc = NULL;
+			goto out;
+		}
+		prealloc = NULL;
+		start = this_end + 1;
+		goto search_again;
+	}
+	/*
+	 * | ---- desired range ---- |
+	 *                        | state |
+	 * We need to split the extent, and set the bit
+	 * on the first half
+	 */
+	if (state->start <= end && state->end > end) {
+		prealloc = alloc_extent_state_atomic(prealloc);
+		if (!prealloc)
+			return -ENOMEM;
+
+		err = split_state(tree, state, prealloc, end + 1);
+		BUG_ON(err == -EEXIST);
+
+		set_state_bits(tree, prealloc, &bits);
+		clear_state_bit(tree, prealloc, &clear_bits, 0);
+
+		merge_state(tree, prealloc);
+		prealloc = NULL;
+		goto out;
+	}
+
+	goto search_again;
+
+out:
+	spin_unlock(&tree->lock);
+	if (prealloc)
+		free_extent_state(prealloc);
+
+	return err;
+
+search_again:
+	if (start > end)
+		goto out;
+	spin_unlock(&tree->lock);
+	if (mask & __GFP_WAIT)
+		cond_resched();
+	goto again;
+}
+
 /* wrappers around set/clear extent bit */
 int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		     gfp_t mask)
@@ -919,7 +1108,7 @@ int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
 			struct extent_state **cached_state, gfp_t mask)
 {
 	return set_extent_bit(tree, start, end,
-			      EXTENT_DELALLOC | EXTENT_DIRTY | EXTENT_UPTODATE,
+			      EXTENT_DELALLOC | EXTENT_UPTODATE,
 			      0, NULL, cached_state, mask);
 }
 
@@ -1599,6 +1788,368 @@ static int check_page_writeback(struct extent_io_tree *tree,
 	return 0;
 }
 
+/*
+ * When IO fails, either with EIO or csum verification fails, we
+ * try other mirrors that might have a good copy of the data.  This
+ * io_failure_record is used to record state as we go through all the
+ * mirrors.  If another mirror has good data, the page is set up to date
+ * and things continue.  If a good mirror can't be found, the original
+ * bio end_io callback is called to indicate things have failed.
+ */
+struct io_failure_record {
+	struct page *page;
+	u64 start;
+	u64 len;
+	u64 logical;
+	unsigned long bio_flags;
+	int this_mirror;
+	int failed_mirror;
+	int in_validation;
+};
+
+static int free_io_failure(struct inode *inode, struct io_failure_record *rec,
+				int did_repair)
+{
+	int ret;
+	int err = 0;
+	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
+
+	set_state_private(failure_tree, rec->start, 0);
+	ret = clear_extent_bits(failure_tree, rec->start,
+				rec->start + rec->len - 1,
+				EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
+	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;
+	}
+
+	kfree(rec);
+	return err;
+}
+
+static void repair_io_failure_callback(struct bio *bio, int err)
+{
+	complete(bio->bi_private);
+}
+
+/*
+ * this bypasses the standard btrfs submit functions deliberately, as
+ * the standard behavior is to write all copies in a raid setup. here we only
+ * want to write the one bad copy. so we do the mapping for ourselves and issue
+ * submit_bio directly.
+ * to avoid any synchonization issues, wait for the data after writing, which
+ * actually prevents the read that triggered the error from finishing.
+ * currently, there can be no more than two copies of every data bit. thus,
+ * exactly one rewrite is required.
+ */
+int repair_io_failure(struct btrfs_mapping_tree *map_tree, u64 start,
+			u64 length, u64 logical, struct page *page,
+			int mirror_num)
+{
+	struct bio *bio;
+	struct btrfs_device *dev;
+	DECLARE_COMPLETION_ONSTACK(compl);
+	u64 map_length = 0;
+	u64 sector;
+	struct btrfs_bio *bbio = NULL;
+	int ret;
+
+	BUG_ON(!mirror_num);
+
+	bio = bio_alloc(GFP_NOFS, 1);
+	if (!bio)
+		return -EIO;
+	bio->bi_private = &compl;
+	bio->bi_end_io = repair_io_failure_callback;
+	bio->bi_size = 0;
+	map_length = length;
+
+	ret = btrfs_map_block(map_tree, WRITE, logical,
+			      &map_length, &bbio, mirror_num);
+	if (ret) {
+		bio_put(bio);
+		return -EIO;
+	}
+	BUG_ON(mirror_num != bbio->mirror_num);
+	sector = bbio->stripes[mirror_num-1].physical >> 9;
+	bio->bi_sector = sector;
+	dev = bbio->stripes[mirror_num-1].dev;
+	kfree(bbio);
+	if (!dev || !dev->bdev || !dev->writeable) {
+		bio_put(bio);
+		return -EIO;
+	}
+	bio->bi_bdev = dev->bdev;
+	bio_add_page(bio, page, length, start-page_offset(page));
+	submit_bio(WRITE_SYNC, bio);
+	wait_for_completion(&compl);
+
+	if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) {
+		/* try to remap that extent elsewhere? */
+		bio_put(bio);
+		return -EIO;
+	}
+
+	printk(KERN_INFO "btrfs read error corrected: ino %lu off %llu (dev %s "
+			"sector %llu)\n", page->mapping->host->i_ino, start,
+			dev->name, sector);
+
+	bio_put(bio);
+	return 0;
+}
+
+/*
+ * each time an IO finishes, we do a fast check in the IO failure tree
+ * to see if we need to process or clean up an io_failure_record
+ */
+static int clean_io_failure(u64 start, struct page *page)
+{
+	u64 private;
+	u64 private_failure;
+	struct io_failure_record *failrec;
+	struct btrfs_mapping_tree *map_tree;
+	struct extent_state *state;
+	int num_copies;
+	int did_repair = 0;
+	int ret;
+	struct inode *inode = page->mapping->host;
+
+	private = 0;
+	ret = count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
+				(u64)-1, 1, EXTENT_DIRTY, 0);
+	if (!ret)
+		return 0;
+
+	ret = get_state_private(&BTRFS_I(inode)->io_failure_tree, start,
+				&private_failure);
+	if (ret)
+		return 0;
+
+	failrec = (struct io_failure_record *)(unsigned long) private_failure;
+	BUG_ON(!failrec->this_mirror);
+
+	if (failrec->in_validation) {
+		/* there was no real error, just free the record */
+		pr_debug("clean_io_failure: freeing dummy error at %llu\n",
+			 failrec->start);
+		did_repair = 1;
+		goto out;
+	}
+
+	spin_lock(&BTRFS_I(inode)->io_tree.lock);
+	state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
+					    failrec->start,
+					    EXTENT_LOCKED);
+	spin_unlock(&BTRFS_I(inode)->io_tree.lock);
+
+	if (state && state->start == failrec->start) {
+		map_tree = &BTRFS_I(inode)->root->fs_info->mapping_tree;
+		num_copies = btrfs_num_copies(map_tree, failrec->logical,
+						failrec->len);
+		if (num_copies > 1)  {
+			ret = repair_io_failure(map_tree, start, failrec->len,
+						failrec->logical, page,
+						failrec->failed_mirror);
+			did_repair = !ret;
+		}
+	}
+
+out:
+	if (!ret)
+		ret = free_io_failure(inode, failrec, did_repair);
+
+	return ret;
+}
+
+/*
+ * this is a generic handler for readpage errors (default
+ * readpage_io_failed_hook). if other copies exist, read those and write back
+ * good data to the failed position. does not investigate in remapping the
+ * failed extent elsewhere, hoping the device will be smart enough to do this as
+ * needed
+ */
+
+static int bio_readpage_error(struct bio *failed_bio, struct page *page,
+				u64 start, u64 end, int failed_mirror,
+				struct extent_state *state)
+{
+	struct io_failure_record *failrec = NULL;
+	u64 private;
+	struct extent_map *em;
+	struct inode *inode = page->mapping->host;
+	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
+	struct extent_io_tree *tree = &BTRFS_I(inode)->io_tree;
+	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
+	struct bio *bio;
+	int num_copies;
+	int ret;
+	int read_mode;
+	u64 logical;
+
+	BUG_ON(failed_bio->bi_rw & REQ_WRITE);
+
+	ret = get_state_private(failure_tree, start, &private);
+	if (ret) {
+		failrec = kzalloc(sizeof(*failrec), GFP_NOFS);
+		if (!failrec)
+			return -ENOMEM;
+		failrec->start = start;
+		failrec->len = end - start + 1;
+		failrec->this_mirror = 0;
+		failrec->bio_flags = 0;
+		failrec->in_validation = 0;
+
+		read_lock(&em_tree->lock);
+		em = lookup_extent_mapping(em_tree, start, failrec->len);
+		if (!em) {
+			read_unlock(&em_tree->lock);
+			kfree(failrec);
+			return -EIO;
+		}
+
+		if (em->start > start || em->start + em->len < start) {
+			free_extent_map(em);
+			em = NULL;
+		}
+		read_unlock(&em_tree->lock);
+
+		if (!em || IS_ERR(em)) {
+			kfree(failrec);
+			return -EIO;
+		}
+		logical = start - em->start;
+		logical = em->block_start + logical;
+		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
+			logical = em->block_start;
+			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
+			extent_set_compress_type(&failrec->bio_flags,
+						 em->compress_type);
+		}
+		pr_debug("bio_readpage_error: (new) logical=%llu, start=%llu, "
+			 "len=%llu\n", logical, start, failrec->len);
+		failrec->logical = logical;
+		free_extent_map(em);
+
+		/* set the bits in the private failure tree */
+		ret = set_extent_bits(failure_tree, start, end,
+					EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
+		if (ret >= 0)
+			ret = set_state_private(failure_tree, start,
+						(u64)(unsigned long)failrec);
+		/* set the bits in the inode's tree */
+		if (ret >= 0)
+			ret = set_extent_bits(tree, start, end, EXTENT_DAMAGED,
+						GFP_NOFS);
+		if (ret < 0) {
+			kfree(failrec);
+			return ret;
+		}
+	} else {
+		failrec = (struct io_failure_record *)(unsigned long)private;
+		pr_debug("bio_readpage_error: (found) logical=%llu, "
+			 "start=%llu, len=%llu, validation=%d\n",
+			 failrec->logical, failrec->start, failrec->len,
+			 failrec->in_validation);
+		/*
+		 * when data can be on disk more than twice, add to failrec here
+		 * (e.g. with a list for failed_mirror) to make
+		 * clean_io_failure() clean all those errors at once.
+		 */
+	}
+	num_copies = btrfs_num_copies(
+			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
+			      failrec->logical, failrec->len);
+	if (num_copies == 1) {
+		/*
+		 * we only have a single copy of the data, so don't bother with
+		 * all the retry and error correction code that follows. no
+		 * matter what the error is, it is very likely to persist.
+		 */
+		pr_debug("bio_readpage_error: cannot repair, num_copies == 1. "
+			 "state=%p, num_copies=%d, next_mirror %d, "
+			 "failed_mirror %d\n", state, num_copies,
+			 failrec->this_mirror, failed_mirror);
+		free_io_failure(inode, failrec, 0);
+		return -EIO;
+	}
+
+	if (!state) {
+		spin_lock(&tree->lock);
+		state = find_first_extent_bit_state(tree, failrec->start,
+						    EXTENT_LOCKED);
+		if (state && state->start != failrec->start)
+			state = NULL;
+		spin_unlock(&tree->lock);
+	}
+
+	/*
+	 * there are two premises:
+	 *	a) deliver good data to the caller
+	 *	b) correct the bad sectors on disk
+	 */
+	if (failed_bio->bi_vcnt > 1) {
+		/*
+		 * to fulfill b), we need to know the exact failing sectors, as
+		 * we don't want to rewrite any more than the failed ones. thus,
+		 * we need separate read requests for the failed bio
+		 *
+		 * if the following BUG_ON triggers, our validation request got
+		 * merged. we need separate requests for our algorithm to work.
+		 */
+		BUG_ON(failrec->in_validation);
+		failrec->in_validation = 1;
+		failrec->this_mirror = failed_mirror;
+		read_mode = READ_SYNC | REQ_FAILFAST_DEV;
+	} else {
+		/*
+		 * we're ready to fulfill a) and b) alongside. get a good copy
+		 * of the failed sector and if we succeed, we have setup
+		 * everything for repair_io_failure to do the rest for us.
+		 */
+		if (failrec->in_validation) {
+			BUG_ON(failrec->this_mirror != failed_mirror);
+			failrec->in_validation = 0;
+			failrec->this_mirror = 0;
+		}
+		failrec->failed_mirror = failed_mirror;
+		failrec->this_mirror++;
+		if (failrec->this_mirror == failed_mirror)
+			failrec->this_mirror++;
+		read_mode = READ_SYNC;
+	}
+
+	if (!state || failrec->this_mirror > num_copies) {
+		pr_debug("bio_readpage_error: (fail) state=%p, num_copies=%d, "
+			 "next_mirror %d, failed_mirror %d\n", state,
+			 num_copies, failrec->this_mirror, failed_mirror);
+		free_io_failure(inode, failrec, 0);
+		return -EIO;
+	}
+
+	bio = bio_alloc(GFP_NOFS, 1);
+	bio->bi_private = state;
+	bio->bi_end_io = failed_bio->bi_end_io;
+	bio->bi_sector = failrec->logical >> 9;
+	bio->bi_bdev = BTRFS_I(inode)->root->fs_info->fs_devices->latest_bdev;
+	bio->bi_size = 0;
+
+	bio_add_page(bio, page, failrec->len, start - page_offset(page));
+
+	pr_debug("bio_readpage_error: submitting new read[%#x] to "
+		 "this_mirror=%d, num_copies=%d, in_validation=%d\n", read_mode,
+		 failrec->this_mirror, num_copies, failrec->in_validation);
+
+	tree->ops->submit_bio_hook(inode, read_mode, bio, failrec->this_mirror,
+					failrec->bio_flags, 0);
+	return 0;
+}
+
 /* lots and lots of room for performance fixes in the end_bio funcs */
 
 /*
@@ -1697,6 +2248,9 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
 		struct extent_state *cached = NULL;
 		struct extent_state *state;
 
+		pr_debug("end_bio_extent_readpage: bi_vcnt=%d, idx=%d, err=%d, "
+			 "mirror=%ld\n", bio->bi_vcnt, bio->bi_idx, err,
+			 (long int)bio->bi_bdev);
 		tree = &BTRFS_I(page->mapping->host)->io_tree;
 
 		start = ((u64)page->index << PAGE_CACHE_SHIFT) +
@@ -1727,11 +2281,19 @@ static void end_bio_extent_readpage(struct bio *bio, int err)
 							      state);
 			if (ret)
 				uptodate = 0;
+			else
+				clean_io_failure(start, page);
 		}
-		if (!uptodate && tree->ops &&
-		    tree->ops->readpage_io_failed_hook) {
-			ret = tree->ops->readpage_io_failed_hook(bio, page,
-							 start, end, NULL);
+		if (!uptodate) {
+			u64 failed_mirror;
+			failed_mirror = (u64)bio->bi_bdev;
+			if (tree->ops && tree->ops->readpage_io_failed_hook)
+				ret = tree->ops->readpage_io_failed_hook(
+						bio, page, start, end,
+						failed_mirror, state);
+			else
+				ret = bio_readpage_error(bio, page, start, end,
+							 failed_mirror, NULL);
 			if (ret == 0) {
 				uptodate =
 					test_bit(BIO_UPTODATE, &bio->bi_flags);
@@ -1811,6 +2373,7 @@ static int submit_one_bio(int rw, struct bio *bio, int mirror_num,
 					   mirror_num, bio_flags, start);
 	else
 		submit_bio(rw, bio);
+
 	if (bio_flagged(bio, BIO_EOPNOTSUPP))
 		ret = -EOPNOTSUPP;
 	bio_put(bio);
@@ -2076,16 +2639,16 @@ out:
 }
 
 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
-			    get_extent_t *get_extent)
+			    get_extent_t *get_extent, int mirror_num)
 {
 	struct bio *bio = NULL;
 	unsigned long bio_flags = 0;
 	int ret;
 
-	ret = __extent_read_full_page(tree, page, get_extent, &bio, 0,
+	ret = __extent_read_full_page(tree, page, get_extent, &bio, mirror_num,
 				      &bio_flags);
 	if (bio)
-		ret = submit_one_bio(READ, bio, 0, bio_flags);
+		ret = submit_one_bio(READ, bio, mirror_num, bio_flags);
 	return ret;
 }
 
@@ -2136,6 +2699,7 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 	int compressed;
 	int write_flags;
 	unsigned long nr_written = 0;
+	bool fill_delalloc = true;
 
 	if (wbc->sync_mode == WB_SYNC_ALL)
 		write_flags = WRITE_SYNC;
@@ -2145,6 +2709,9 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 	trace___extent_writepage(page, inode, wbc);
 
 	WARN_ON(!PageLocked(page));
+
+	ClearPageError(page);
+
 	pg_offset = i_size & (PAGE_CACHE_SIZE - 1);
 	if (page->index > end_index ||
 	   (page->index == end_index && !pg_offset)) {
@@ -2166,10 +2733,13 @@ static int __extent_writepage(struct page *page, struct writeback_control *wbc,
 
 	set_page_extent_mapped(page);
 
+	if (!tree->ops || !tree->ops->fill_delalloc)
+		fill_delalloc = false;
+
 	delalloc_start = start;
 	delalloc_end = 0;
 	page_started = 0;
-	if (!epd->extent_locked) {
+	if (!epd->extent_locked && fill_delalloc) {
 		u64 delalloc_to_write = 0;
 		/*
 		 * make sure the wbc mapping index is at least updated
@@ -2421,10 +2991,16 @@ retry:
 			 * swizzled back from swapper_space to tmpfs file
 			 * mapping
 			 */
-			if (tree->ops && tree->ops->write_cache_pages_lock_hook)
-				tree->ops->write_cache_pages_lock_hook(page);
-			else
-				lock_page(page);
+			if (tree->ops &&
+			    tree->ops->write_cache_pages_lock_hook) {
+				tree->ops->write_cache_pages_lock_hook(page,
+							       data, flush_fn);
+			} else {
+				if (!trylock_page(page)) {
+					flush_fn(data);
+					lock_page(page);
+				}
+			}
 
 			if (unlikely(page->mapping != mapping)) {
 				unlock_page(page);
@@ -2926,7 +3502,7 @@ out:
 	return ret;
 }
 
-static inline struct page *extent_buffer_page(struct extent_buffer *eb,
+inline struct page *extent_buffer_page(struct extent_buffer *eb,
 					      unsigned long i)
 {
 	struct page *p;
@@ -2951,7 +3527,7 @@ static inline struct page *extent_buffer_page(struct extent_buffer *eb,
 	return p;
 }
 
-static inline unsigned long num_extent_pages(u64 start, u64 len)
+inline unsigned long num_extent_pages(u64 start, u64 len)
 {
 	return ((start + len + PAGE_CACHE_SIZE - 1) >> PAGE_CACHE_SHIFT) -
 		(start >> PAGE_CACHE_SHIFT);
@@ -3204,6 +3780,7 @@ int clear_extent_buffer_dirty(struct extent_io_tree *tree,
 						PAGECACHE_TAG_DIRTY);
 		}
 		spin_unlock_irq(&page->mapping->tree_lock);
+		ClearPageError(page);
 		unlock_page(page);
 	}
 	return 0;
@@ -3349,8 +3926,7 @@ int extent_buffer_uptodate(struct extent_io_tree *tree,
 }
 
 int read_extent_buffer_pages(struct extent_io_tree *tree,
-			     struct extent_buffer *eb,
-			     u64 start, int wait,
+			     struct extent_buffer *eb, u64 start, int wait,
 			     get_extent_t *get_extent, int mirror_num)
 {
 	unsigned long i;
@@ -3386,7 +3962,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	num_pages = num_extent_pages(eb->start, eb->len);
 	for (i = start_i; i < num_pages; i++) {
 		page = extent_buffer_page(eb, i);
-		if (!wait) {
+		if (wait == WAIT_NONE) {
 			if (!trylock_page(page))
 				goto unlock_exit;
 		} else {
@@ -3430,7 +4006,7 @@ int read_extent_buffer_pages(struct extent_io_tree *tree,
 	if (bio)
 		submit_one_bio(READ, bio, mirror_num, bio_flags);
 
-	if (ret || !wait)
+	if (ret || wait != WAIT_COMPLETE)
 		return ret;
 
 	for (i = start_i; i < num_pages; i++) {

fs/btrfs/extent_io.h

@@ -17,6 +17,8 @@
 #define EXTENT_NODATASUM (1 << 10)
 #define EXTENT_DO_ACCOUNTING (1 << 11)
 #define EXTENT_FIRST_DELALLOC (1 << 12)
+#define EXTENT_NEED_WAIT (1 << 13)
+#define EXTENT_DAMAGED (1 << 14)
 #define EXTENT_IOBITS (EXTENT_LOCKED | EXTENT_WRITEBACK)
 #define EXTENT_CTLBITS (EXTENT_DO_ACCOUNTING | EXTENT_FIRST_DELALLOC)
 
@@ -32,6 +34,7 @@
 #define EXTENT_BUFFER_BLOCKING 1
 #define EXTENT_BUFFER_DIRTY 2
 #define EXTENT_BUFFER_CORRUPT 3
+#define EXTENT_BUFFER_READAHEAD 4	/* this got triggered by readahead */
 
 /* these are flags for extent_clear_unlock_delalloc */
 #define EXTENT_CLEAR_UNLOCK_PAGE 0x1
@@ -67,7 +70,7 @@ struct extent_io_ops {
 			      unsigned long bio_flags);
 	int (*readpage_io_hook)(struct page *page, u64 start, u64 end);
 	int (*readpage_io_failed_hook)(struct bio *bio, struct page *page,
-				       u64 start, u64 end,
+				       u64 start, u64 end, u64 failed_mirror,
 				       struct extent_state *state);
 	int (*writepage_io_failed_hook)(struct bio *bio, struct page *page,
 					u64 start, u64 end,
@@ -85,7 +88,8 @@ struct extent_io_ops {
 				  struct extent_state *other);
 	void (*split_extent_hook)(struct inode *inode,
 				  struct extent_state *orig, u64 split);
-	int (*write_cache_pages_lock_hook)(struct page *page);
+	int (*write_cache_pages_lock_hook)(struct page *page, void *data,
+					   void (*flush_fn)(void *));
 };
 
 struct extent_io_tree {
@@ -185,7 +189,7 @@ int unlock_extent_cached(struct extent_io_tree *tree, u64 start, u64 end,
 int try_lock_extent(struct extent_io_tree *tree, u64 start, u64 end,
 		    gfp_t mask);
 int extent_read_full_page(struct extent_io_tree *tree, struct page *page,
-			  get_extent_t *get_extent);
+			  get_extent_t *get_extent, int mirror_num);
 int __init extent_io_init(void);
 void extent_io_exit(void);
 
@@ -214,6 +218,8 @@ int set_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		     gfp_t mask);
 int clear_extent_dirty(struct extent_io_tree *tree, u64 start, u64 end,
 		       gfp_t mask);
+int convert_extent_bit(struct extent_io_tree *tree, u64 start, u64 end,
+		       int bits, int clear_bits, gfp_t mask);
 int set_extent_delalloc(struct extent_io_tree *tree, u64 start, u64 end,
 			struct extent_state **cached_state, gfp_t mask);
 int find_first_extent_bit(struct extent_io_tree *tree, u64 start,
@@ -248,9 +254,14 @@ struct extent_buffer *alloc_extent_buffer(struct extent_io_tree *tree,
 struct extent_buffer *find_extent_buffer(struct extent_io_tree *tree,
 					 u64 start, unsigned long len);
 void free_extent_buffer(struct extent_buffer *eb);
+#define WAIT_NONE	0
+#define WAIT_COMPLETE	1
+#define WAIT_PAGE_LOCK	2
 int read_extent_buffer_pages(struct extent_io_tree *tree,
 			     struct extent_buffer *eb, u64 start, int wait,
 			     get_extent_t *get_extent, int mirror_num);
+unsigned long num_extent_pages(u64 start, u64 len);
+struct page *extent_buffer_page(struct extent_buffer *eb, unsigned long i);
 
 static inline void extent_buffer_get(struct extent_buffer *eb)
 {
@@ -300,4 +311,10 @@ int extent_clear_unlock_delalloc(struct inode *inode,
 struct bio *
 btrfs_bio_alloc(struct block_device *bdev, u64 first_sector, int nr_vecs,
 		gfp_t gfp_flags);
+
+struct btrfs_mapping_tree;
+
+int repair_io_failure(struct btrfs_mapping_tree *map_tree, u64 start,
+			u64 length, u64 logical, struct page *page,
+			int mirror_num);
 #endif

fs/btrfs/file-item.c

@@ -91,8 +91,7 @@ struct btrfs_csum_item *btrfs_lookup_csum(struct btrfs_trans_handle *trans,
 	struct btrfs_csum_item *item;
 	struct extent_buffer *leaf;
 	u64 csum_offset = 0;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	int csums_in_item;
 
 	file_key.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
@@ -162,8 +161,7 @@ static int __btrfs_lookup_bio_sums(struct btrfs_root *root,
 	u64 item_last_offset = 0;
 	u64 disk_bytenr;
 	u32 diff;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	int ret;
 	struct btrfs_path *path;
 	struct btrfs_csum_item *item = NULL;
@@ -290,7 +288,7 @@ int btrfs_lookup_csums_range(struct btrfs_root *root, u64 start, u64 end,
 	int ret;
 	size_t size;
 	u64 csum_end;
-	u16 csum_size = btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 
 	path = btrfs_alloc_path();
 	if (!path)
@@ -492,8 +490,7 @@ static noinline int truncate_one_csum(struct btrfs_trans_handle *trans,
 				      u64 bytenr, u64 len)
 {
 	struct extent_buffer *leaf;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	u64 csum_end;
 	u64 end_byte = bytenr + len;
 	u32 blocksize_bits = root->fs_info->sb->s_blocksize_bits;
@@ -549,8 +546,7 @@ int btrfs_del_csums(struct btrfs_trans_handle *trans,
 	u64 csum_end;
 	struct extent_buffer *leaf;
 	int ret;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 	int blocksize_bits = root->fs_info->sb->s_blocksize_bits;
 
 	root = root->fs_info->csum_root;
@@ -676,8 +672,7 @@ int btrfs_csum_file_blocks(struct btrfs_trans_handle *trans,
 	struct btrfs_sector_sum *sector_sum;
 	u32 nritems;
 	u32 ins_size;
-	u16 csum_size =
-		btrfs_super_csum_size(&root->fs_info->super_copy);
+	u16 csum_size = btrfs_super_csum_size(root->fs_info->super_copy);
 
 	path = btrfs_alloc_path();
 	if (!path)

fs/btrfs/file.c

@@ -1069,6 +1069,7 @@ static noinline int prepare_pages(struct btrfs_root *root, struct file *file,
 	int i;
 	unsigned long index = pos >> PAGE_CACHE_SHIFT;
 	struct inode *inode = fdentry(file)->d_inode;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 	int err = 0;
 	int faili = 0;
 	u64 start_pos;
@@ -1080,7 +1081,7 @@ static noinline int prepare_pages(struct btrfs_root *root, struct file *file,
 again:
 	for (i = 0; i < num_pages; i++) {
 		pages[i] = find_or_create_page(inode->i_mapping, index + i,
-					       GFP_NOFS);
+					       mask);
 		if (!pages[i]) {
 			faili = i - 1;
 			err = -ENOMEM;
@@ -1615,10 +1616,6 @@ static long btrfs_fallocate(struct file *file, int mode,
 			goto out;
 	}
 
-	ret = btrfs_check_data_free_space(inode, alloc_end - alloc_start);
-	if (ret)
-		goto out;
-
 	locked_end = alloc_end - 1;
 	while (1) {
 		struct btrfs_ordered_extent *ordered;
@@ -1664,11 +1661,27 @@ static long btrfs_fallocate(struct file *file, int mode,
 		if (em->block_start == EXTENT_MAP_HOLE ||
 		    (cur_offset >= inode->i_size &&
 		     !test_bit(EXTENT_FLAG_PREALLOC, &em->flags))) {
+
+			/*
+			 * Make sure we have enough space before we do the
+			 * allocation.
+			 */
+			ret = btrfs_check_data_free_space(inode, last_byte -
+							  cur_offset);
+			if (ret) {
+				free_extent_map(em);
+				break;
+			}
+
 			ret = btrfs_prealloc_file_range(inode, mode, cur_offset,
 							last_byte - cur_offset,
 							1 << inode->i_blkbits,
 							offset + len,
 							&alloc_hint);
+
+			/* Let go of our reservation. */
+			btrfs_free_reserved_data_space(inode, last_byte -
+						       cur_offset);
 			if (ret < 0) {
 				free_extent_map(em);
 				break;
@@ -1694,8 +1707,6 @@ static long btrfs_fallocate(struct file *file, int mode,
 	}
 	unlock_extent_cached(&BTRFS_I(inode)->io_tree, alloc_start, locked_end,
 			     &cached_state, GFP_NOFS);
-
-	btrfs_free_reserved_data_space(inode, alloc_end - alloc_start);
 out:
 	mutex_unlock(&inode->i_mutex);
 	return ret;

fs/btrfs/inode-map.c

@@ -465,14 +465,16 @@ again:
 	/* Just to make sure we have enough space */
 	prealloc += 8 * PAGE_CACHE_SIZE;
 
-	ret = btrfs_check_data_free_space(inode, prealloc);
+	ret = btrfs_delalloc_reserve_space(inode, prealloc);
 	if (ret)
 		goto out_put;
 
 	ret = btrfs_prealloc_file_range_trans(inode, trans, 0, 0, prealloc,
 					      prealloc, prealloc, &alloc_hint);
-	if (ret)
+	if (ret) {
+		btrfs_delalloc_release_space(inode, prealloc);
 		goto out_put;
+	}
 	btrfs_free_reserved_data_space(inode, prealloc);
 
 out_put:

fs/btrfs/inode.c

@@ -45,10 +45,10 @@
 #include "btrfs_inode.h"
 #include "ioctl.h"
 #include "print-tree.h"
-#include "volumes.h"
 #include "ordered-data.h"
 #include "xattr.h"
 #include "tree-log.h"
+#include "volumes.h"
 #include "compression.h"
 #include "locking.h"
 #include "free-space-cache.h"
@@ -393,7 +393,10 @@ again:
 	     (BTRFS_I(inode)->flags & BTRFS_INODE_COMPRESS))) {
 		WARN_ON(pages);
 		pages = kzalloc(sizeof(struct page *) * nr_pages, GFP_NOFS);
-		BUG_ON(!pages);
+		if (!pages) {
+			/* just bail out to the uncompressed code */
+			goto cont;
+		}
 
 		if (BTRFS_I(inode)->force_compress)
 			compress_type = BTRFS_I(inode)->force_compress;
@@ -424,6 +427,7 @@ again:
 			will_compress = 1;
 		}
 	}
+cont:
 	if (start == 0) {
 		trans = btrfs_join_transaction(root);
 		BUG_ON(IS_ERR(trans));
@@ -820,7 +824,7 @@ static noinline int cow_file_range(struct inode *inode,
 	}
 
 	BUG_ON(disk_num_bytes >
-	       btrfs_super_total_bytes(&root->fs_info->super_copy));
+	       btrfs_super_total_bytes(root->fs_info->super_copy));
 
 	alloc_hint = get_extent_allocation_hint(inode, start, num_bytes);
 	btrfs_drop_extent_cache(inode, start, start + num_bytes - 1, 0);
@@ -1792,12 +1796,12 @@ static int btrfs_finish_ordered_io(struct inode *inode, u64 start, u64 end)
 	}
 	ret = 0;
 out:
-	if (nolock) {
-		if (trans)
-			btrfs_end_transaction_nolock(trans, root);
-	} else {
+	if (root != root->fs_info->tree_root)
 		btrfs_delalloc_release_metadata(inode, ordered_extent->len);
-		if (trans)
+	if (trans) {
+		if (nolock)
+			btrfs_end_transaction_nolock(trans, root);
+		else
 			btrfs_end_transaction(trans, root);
 	}
 
@@ -1818,154 +1822,10 @@ static int btrfs_writepage_end_io_hook(struct page *page, u64 start, u64 end,
 	return btrfs_finish_ordered_io(page->mapping->host, start, end);
 }
 
-/*
- * When IO fails, either with EIO or csum verification fails, we
- * try other mirrors that might have a good copy of the data.  This
- * io_failure_record is used to record state as we go through all the
- * mirrors.  If another mirror has good data, the page is set up to date
- * and things continue.  If a good mirror can't be found, the original
- * bio end_io callback is called to indicate things have failed.
- */
-struct io_failure_record {
-	struct page *page;
-	u64 start;
-	u64 len;
-	u64 logical;
-	unsigned long bio_flags;
-	int last_mirror;
-};
-
-static int btrfs_io_failed_hook(struct bio *failed_bio,
-			 struct page *page, u64 start, u64 end,
-			 struct extent_state *state)
-{
-	struct io_failure_record *failrec = NULL;
-	u64 private;
-	struct extent_map *em;
-	struct inode *inode = page->mapping->host;
-	struct extent_io_tree *failure_tree = &BTRFS_I(inode)->io_failure_tree;
-	struct extent_map_tree *em_tree = &BTRFS_I(inode)->extent_tree;
-	struct bio *bio;
-	int num_copies;
-	int ret;
-	int rw;
-	u64 logical;
-
-	ret = get_state_private(failure_tree, start, &private);
-	if (ret) {
-		failrec = kmalloc(sizeof(*failrec), GFP_NOFS);
-		if (!failrec)
-			return -ENOMEM;
-		failrec->start = start;
-		failrec->len = end - start + 1;
-		failrec->last_mirror = 0;
-		failrec->bio_flags = 0;
-
-		read_lock(&em_tree->lock);
-		em = lookup_extent_mapping(em_tree, start, failrec->len);
-		if (em->start > start || em->start + em->len < start) {
-			free_extent_map(em);
-			em = NULL;
-		}
-		read_unlock(&em_tree->lock);
-
-		if (IS_ERR_OR_NULL(em)) {
-			kfree(failrec);
-			return -EIO;
-		}
-		logical = start - em->start;
-		logical = em->block_start + logical;
-		if (test_bit(EXTENT_FLAG_COMPRESSED, &em->flags)) {
-			logical = em->block_start;
-			failrec->bio_flags = EXTENT_BIO_COMPRESSED;
-			extent_set_compress_type(&failrec->bio_flags,
-						 em->compress_type);
-		}
-		failrec->logical = logical;
-		free_extent_map(em);
-		set_extent_bits(failure_tree, start, end, EXTENT_LOCKED |
-				EXTENT_DIRTY, GFP_NOFS);
-		set_state_private(failure_tree, start,
-				 (u64)(unsigned long)failrec);
-	} else {
-		failrec = (struct io_failure_record *)(unsigned long)private;
-	}
-	num_copies = btrfs_num_copies(
-			      &BTRFS_I(inode)->root->fs_info->mapping_tree,
-			      failrec->logical, failrec->len);
-	failrec->last_mirror++;
-	if (!state) {
-		spin_lock(&BTRFS_I(inode)->io_tree.lock);
-		state = find_first_extent_bit_state(&BTRFS_I(inode)->io_tree,
-						    failrec->start,
-						    EXTENT_LOCKED);
-		if (state && state->start != failrec->start)
-			state = NULL;
-		spin_unlock(&BTRFS_I(inode)->io_tree.lock);
-	}
-	if (!state || failrec->last_mirror > num_copies) {
-		set_state_private(failure_tree, failrec->start, 0);
-		clear_extent_bits(failure_tree, failrec->start,
-				  failrec->start + failrec->len - 1,
-				  EXTENT_LOCKED | EXTENT_DIRTY, GFP_NOFS);
-		kfree(failrec);
-		return -EIO;
-	}
-	bio = bio_alloc(GFP_NOFS, 1);
-	bio->bi_private = state;
-	bio->bi_end_io = failed_bio->bi_end_io;
-	bio->bi_sector = failrec->logical >> 9;
-	bio->bi_bdev = failed_bio->bi_bdev;
-	bio->bi_size = 0;
-
-	bio_add_page(bio, page, failrec->len, start - page_offset(page));
-	if (failed_bio->bi_rw & REQ_WRITE)
-		rw = WRITE;
-	else
-		rw = READ;
-
-	ret = BTRFS_I(inode)->io_tree.ops->submit_bio_hook(inode, rw, bio,
-						      failrec->last_mirror,
-						      failrec->bio_flags, 0);
-	return ret;
-}
-
-/*
- * each time an IO finishes, we do a fast check in the IO failure tree
- * to see if we need to process or clean up an io_failure_record
- */
-static int btrfs_clean_io_failures(struct inode *inode, u64 start)
-{
-	u64 private;
-	u64 private_failure;
-	struct io_failure_record *failure;
-	int ret;
-
-	private = 0;
-	if (count_range_bits(&BTRFS_I(inode)->io_failure_tree, &private,
-			     (u64)-1, 1, EXTENT_DIRTY, 0)) {
-		ret = get_state_private(&BTRFS_I(inode)->io_failure_tree,
-					start, &private_failure);
-		if (ret == 0) {
-			failure = (struct io_failure_record *)(unsigned long)
-				   private_failure;
-			set_state_private(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start, 0);
-			clear_extent_bits(&BTRFS_I(inode)->io_failure_tree,
-					  failure->start,
-					  failure->start + failure->len - 1,
-					  EXTENT_DIRTY | EXTENT_LOCKED,
-					  GFP_NOFS);
-			kfree(failure);
-		}
-	}
-	return 0;
-}
-
 /*
  * when reads are done, we need to check csums to verify the data is correct
- * if there's a match, we allow the bio to finish.  If not, we go through
- * the io_failure_record routines to find good copies
+ * if there's a match, we allow the bio to finish.  If not, the code in
+ * extent_io.c will try to find good copies for us.
  */
 static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 			       struct extent_state *state)
@@ -2011,10 +1871,6 @@ static int btrfs_readpage_end_io_hook(struct page *page, u64 start, u64 end,
 
 	kunmap_atomic(kaddr, KM_USER0);
 good:
-	/* if the io failure tree for this inode is non-empty,
-	 * check to see if we've recovered from a failed IO
-	 */
-	btrfs_clean_io_failures(inode, start);
 	return 0;
 
 zeroit:
@@ -2079,89 +1935,6 @@ void btrfs_run_delayed_iputs(struct btrfs_root *root)
 	up_read(&root->fs_info->cleanup_work_sem);
 }
 
-/*
- * calculate extra metadata reservation when snapshotting a subvolume
- * contains orphan files.
- */
-void btrfs_orphan_pre_snapshot(struct btrfs_trans_handle *trans,
-				struct btrfs_pending_snapshot *pending,
-				u64 *bytes_to_reserve)
-{
-	struct btrfs_root *root;
-	struct btrfs_block_rsv *block_rsv;
-	u64 num_bytes;
-	int index;
-
-	root = pending->root;
-	if (!root->orphan_block_rsv || list_empty(&root->orphan_list))
-		return;
-
-	block_rsv = root->orphan_block_rsv;
-
-	/* orphan block reservation for the snapshot */
-	num_bytes = block_rsv->size;
-
-	/*
-	 * after the snapshot is created, COWing tree blocks may use more
-	 * space than it frees. So we should make sure there is enough
-	 * reserved space.
-	 */
-	index = trans->transid & 0x1;
-	if (block_rsv->reserved + block_rsv->freed[index] < block_rsv->size) {
-		num_bytes += block_rsv->size -
-			     (block_rsv->reserved + block_rsv->freed[index]);
-	}
-
-	*bytes_to_reserve += num_bytes;
-}
-
-void btrfs_orphan_post_snapshot(struct btrfs_trans_handle *trans,
-				struct btrfs_pending_snapshot *pending)
-{
-	struct btrfs_root *root = pending->root;
-	struct btrfs_root *snap = pending->snap;
-	struct btrfs_block_rsv *block_rsv;
-	u64 num_bytes;
-	int index;
-	int ret;
-
-	if (!root->orphan_block_rsv || list_empty(&root->orphan_list))
-		return;
-
-	/* refill source subvolume's orphan block reservation */
-	block_rsv = root->orphan_block_rsv;
-	index = trans->transid & 0x1;
-	if (block_rsv->reserved + block_rsv->freed[index] < block_rsv->size) {
-		num_bytes = block_rsv->size -
-			    (block_rsv->reserved + block_rsv->freed[index]);
-		ret = btrfs_block_rsv_migrate(&pending->block_rsv,
-					      root->orphan_block_rsv,
-					      num_bytes);
-		BUG_ON(ret);
-	}
-
-	/* setup orphan block reservation for the snapshot */
-	block_rsv = btrfs_alloc_block_rsv(snap);
-	BUG_ON(!block_rsv);
-
-	btrfs_add_durable_block_rsv(root->fs_info, block_rsv);
-	snap->orphan_block_rsv = block_rsv;
-
-	num_bytes = root->orphan_block_rsv->size;
-	ret = btrfs_block_rsv_migrate(&pending->block_rsv,
-				      block_rsv, num_bytes);
-	BUG_ON(ret);
-
-#if 0
-	/* insert orphan item for the snapshot */
-	WARN_ON(!root->orphan_item_inserted);
-	ret = btrfs_insert_orphan_item(trans, root->fs_info->tree_root,
-				       snap->root_key.objectid);
-	BUG_ON(ret);
-	snap->orphan_item_inserted = 1;
-#endif
-}
-
 enum btrfs_orphan_cleanup_state {
 	ORPHAN_CLEANUP_STARTED	= 1,
 	ORPHAN_CLEANUP_DONE	= 2,
@@ -2247,9 +2020,6 @@ int btrfs_orphan_add(struct btrfs_trans_handle *trans, struct inode *inode)
 	}
 	spin_unlock(&root->orphan_lock);
 
-	if (block_rsv)
-		btrfs_add_durable_block_rsv(root->fs_info, block_rsv);
-
 	/* grab metadata reservation from transaction handle */
 	if (reserve) {
 		ret = btrfs_orphan_reserve_metadata(trans, inode);
@@ -2316,6 +2086,7 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
 	struct btrfs_key key, found_key;
 	struct btrfs_trans_handle *trans;
 	struct inode *inode;
+	u64 last_objectid = 0;
 	int ret = 0, nr_unlink = 0, nr_truncate = 0;
 
 	if (cmpxchg(&root->orphan_cleanup_state, 0, ORPHAN_CLEANUP_STARTED))
@@ -2367,41 +2138,49 @@ int btrfs_orphan_cleanup(struct btrfs_root *root)
 		 * crossing root thing.  we store the inode number in the
 		 * offset of the orphan item.
 		 */
+
+		if (found_key.offset == last_objectid) {
+			printk(KERN_ERR "btrfs: Error removing orphan entry, "
+			       "stopping orphan cleanup\n");
+			ret = -EINVAL;
+			goto out;
+		}
+
+		last_objectid = found_key.offset;
+
 		found_key.objectid = found_key.offset;
 		found_key.type = BTRFS_INODE_ITEM_KEY;
 		found_key.offset = 0;
 		inode = btrfs_iget(root->fs_info->sb, &found_key, root, NULL);
-		if (IS_ERR(inode)) {
-			ret = PTR_ERR(inode);
+		ret = PTR_RET(inode);
+		if (ret && ret != -ESTALE)
 			goto out;
-		}
 
 		/*
-		 * add this inode to the orphan list so btrfs_orphan_del does
-		 * the proper thing when we hit it
+		 * Inode is already gone but the orphan item is still there,
+		 * kill the orphan item.
 		 */
-		spin_lock(&root->orphan_lock);
-		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
-		spin_unlock(&root->orphan_lock);
-
-		/*
-		 * if this is a bad inode, means we actually succeeded in
-		 * removing the inode, but not the orphan record, which means
-		 * we need to manually delete the orphan since iput will just
-		 * do a destroy_inode
-		 */
-		if (is_bad_inode(inode)) {
-			trans = btrfs_start_transaction(root, 0);
+		if (ret == -ESTALE) {
+			trans = btrfs_start_transaction(root, 1);
 			if (IS_ERR(trans)) {
 				ret = PTR_ERR(trans);
 				goto out;
 			}
-			btrfs_orphan_del(trans, inode);
+			ret = btrfs_del_orphan_item(trans, root,
+						    found_key.objectid);
+			BUG_ON(ret);
 			btrfs_end_transaction(trans, root);
-			iput(inode);
 			continue;
 		}
 
+		/*
+		 * add this inode to the orphan list so btrfs_orphan_del does
+		 * the proper thing when we hit it
+		 */
+		spin_lock(&root->orphan_lock);
+		list_add(&BTRFS_I(inode)->i_orphan, &root->orphan_list);
+		spin_unlock(&root->orphan_lock);
+
 		/* if we have links, this was a truncate, lets do that */
 		if (inode->i_nlink) {
 			if (!S_ISREG(inode->i_mode)) {
@@ -2835,7 +2614,16 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
 	u64 ino = btrfs_ino(inode);
 	u64 dir_ino = btrfs_ino(dir);
 
-	trans = btrfs_start_transaction(root, 10);
+	/*
+	 * 1 for the possible orphan item
+	 * 1 for the dir item
+	 * 1 for the dir index
+	 * 1 for the inode ref
+	 * 1 for the inode ref in the tree log
+	 * 2 for the dir entries in the log
+	 * 1 for the inode
+	 */
+	trans = btrfs_start_transaction(root, 8);
 	if (!IS_ERR(trans) || PTR_ERR(trans) != -ENOSPC)
 		return trans;
 
@@ -2858,7 +2646,8 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
 		return ERR_PTR(-ENOMEM);
 	}
 
-	trans = btrfs_start_transaction(root, 0);
+	/* 1 for the orphan item */
+	trans = btrfs_start_transaction(root, 1);
 	if (IS_ERR(trans)) {
 		btrfs_free_path(path);
 		root->fs_info->enospc_unlink = 0;
@@ -2963,6 +2752,12 @@ static struct btrfs_trans_handle *__unlink_start_trans(struct inode *dir,
 	err = 0;
 out:
 	btrfs_free_path(path);
+	/* Migrate the orphan reservation over */
+	if (!err)
+		err = btrfs_block_rsv_migrate(trans->block_rsv,
+				&root->fs_info->global_block_rsv,
+				trans->bytes_reserved);
+
 	if (err) {
 		btrfs_end_transaction(trans, root);
 		root->fs_info->enospc_unlink = 0;
@@ -2977,6 +2772,9 @@ static void __unlink_end_trans(struct btrfs_trans_handle *trans,
 			       struct btrfs_root *root)
 {
 	if (trans->block_rsv == &root->fs_info->global_block_rsv) {
+		btrfs_block_rsv_release(root, trans->block_rsv,
+					trans->bytes_reserved);
+		trans->block_rsv = &root->fs_info->trans_block_rsv;
 		BUG_ON(!root->fs_info->enospc_unlink);
 		root->fs_info->enospc_unlink = 0;
 	}
@@ -3368,6 +3166,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 	pgoff_t index = from >> PAGE_CACHE_SHIFT;
 	unsigned offset = from & (PAGE_CACHE_SIZE-1);
 	struct page *page;
+	gfp_t mask = btrfs_alloc_write_mask(mapping);
 	int ret = 0;
 	u64 page_start;
 	u64 page_end;
@@ -3380,7 +3179,7 @@ static int btrfs_truncate_page(struct address_space *mapping, loff_t from)
 
 	ret = -ENOMEM;
 again:
-	page = find_or_create_page(mapping, index, GFP_NOFS);
+	page = find_or_create_page(mapping, index, mask);
 	if (!page) {
 		btrfs_delalloc_release_space(inode, PAGE_CACHE_SIZE);
 		goto out;
@@ -3613,6 +3412,8 @@ void btrfs_evict_inode(struct inode *inode)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_root *root = BTRFS_I(inode)->root;
+	struct btrfs_block_rsv *rsv, *global_rsv;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
 	unsigned long nr;
 	int ret;
 
@@ -3640,22 +3441,55 @@ void btrfs_evict_inode(struct inode *inode)
 		goto no_delete;
 	}
 
+	rsv = btrfs_alloc_block_rsv(root);
+	if (!rsv) {
+		btrfs_orphan_del(NULL, inode);
+		goto no_delete;
+	}
+	rsv->size = min_size;
+	global_rsv = &root->fs_info->global_block_rsv;
+
 	btrfs_i_size_write(inode, 0);
 
+	/*
+	 * This is a bit simpler than btrfs_truncate since
+	 *
+	 * 1) We've already reserved our space for our orphan item in the
+	 *    unlink.
+	 * 2) We're going to delete the inode item, so we don't need to update
+	 *    it at all.
+	 *
+	 * So we just need to reserve some slack space in case we add bytes when
+	 * doing the truncate.
+	 */
 	while (1) {
-		trans = btrfs_join_transaction(root);
-		BUG_ON(IS_ERR(trans));
-		trans->block_rsv = root->orphan_block_rsv;
+		ret = btrfs_block_rsv_refill(root, rsv, min_size);
+
+		/*
+		 * Try and steal from the global reserve since we will
+		 * likely not use this space anyway, we want to try as
+		 * hard as possible to get this to work.
+		 */
+		if (ret)
+			ret = btrfs_block_rsv_migrate(global_rsv, rsv, min_size);
 
-		ret = btrfs_block_rsv_check(trans, root,
-					    root->orphan_block_rsv, 0, 5);
 		if (ret) {
-			BUG_ON(ret != -EAGAIN);
-			ret = btrfs_commit_transaction(trans, root);
-			BUG_ON(ret);
-			continue;
+			printk(KERN_WARNING "Could not get space for a "
+			       "delete, will truncate on mount %d\n", ret);
+			btrfs_orphan_del(NULL, inode);
+			btrfs_free_block_rsv(root, rsv);
+			goto no_delete;
 		}
 
+		trans = btrfs_start_transaction(root, 0);
+		if (IS_ERR(trans)) {
+			btrfs_orphan_del(NULL, inode);
+			btrfs_free_block_rsv(root, rsv);
+			goto no_delete;
+		}
+
+		trans->block_rsv = rsv;
+
 		ret = btrfs_truncate_inode_items(trans, root, inode, 0, 0);
 		if (ret != -EAGAIN)
 			break;
@@ -3664,14 +3498,17 @@ void btrfs_evict_inode(struct inode *inode)
 		btrfs_end_transaction(trans, root);
 		trans = NULL;
 		btrfs_btree_balance_dirty(root, nr);
-
 	}
 
+	btrfs_free_block_rsv(root, rsv);
+
 	if (ret == 0) {
+		trans->block_rsv = root->orphan_block_rsv;
 		ret = btrfs_orphan_del(trans, inode);
 		BUG_ON(ret);
 	}
 
+	trans->block_rsv = &root->fs_info->trans_block_rsv;
 	if (!(root == root->fs_info->tree_root ||
 	      root->root_key.objectid == BTRFS_TREE_RELOC_OBJECTID))
 		btrfs_return_ino(root, btrfs_ino(inode));
@@ -5795,8 +5632,7 @@ again:
 	if (test_bit(BTRFS_ORDERED_NOCOW, &ordered->flags)) {
 		ret = btrfs_ordered_update_i_size(inode, 0, ordered);
 		if (!ret)
-			ret = btrfs_update_inode(trans, root, inode);
-		err = ret;
+			err = btrfs_update_inode(trans, root, inode);
 		goto out;
 	}
 
@@ -6289,7 +6125,7 @@ int btrfs_readpage(struct file *file, struct page *page)
 {
 	struct extent_io_tree *tree;
 	tree = &BTRFS_I(page->mapping->host)->io_tree;
-	return extent_read_full_page(tree, page, btrfs_get_extent);
+	return extent_read_full_page(tree, page, btrfs_get_extent, 0);
 }
 
 static int btrfs_writepage(struct page *page, struct writeback_control *wbc)
@@ -6541,6 +6377,7 @@ static int btrfs_truncate(struct inode *inode)
 	struct btrfs_trans_handle *trans;
 	unsigned long nr;
 	u64 mask = root->sectorsize - 1;
+	u64 min_size = btrfs_calc_trunc_metadata_size(root, 1);
 
 	ret = btrfs_truncate_page(inode->i_mapping, inode->i_size);
 	if (ret)
@@ -6588,19 +6425,23 @@ static int btrfs_truncate(struct inode *inode)
 	rsv = btrfs_alloc_block_rsv(root);
 	if (!rsv)
 		return -ENOMEM;
-	btrfs_add_durable_block_rsv(root->fs_info, rsv);
+	rsv->size = min_size;
 
+	/*
+	 * 1 for the truncate slack space
+	 * 1 for the orphan item we're going to add
+	 * 1 for the orphan item deletion
+	 * 1 for updating the inode.
+	 */
 	trans = btrfs_start_transaction(root, 4);
 	if (IS_ERR(trans)) {
 		err = PTR_ERR(trans);
 		goto out;
 	}
 
-	/*
-	 * Reserve space for the truncate process.  Truncate should be adding
-	 * space, but if there are snapshots it may end up using space.
-	 */
-	ret = btrfs_truncate_reserve_metadata(trans, root, rsv);
+	/* Migrate the slack space for the truncate to our reserve */
+	ret = btrfs_block_rsv_migrate(&root->fs_info->trans_block_rsv, rsv,
+				      min_size);
 	BUG_ON(ret);
 
 	ret = btrfs_orphan_add(trans, inode);
@@ -6609,21 +6450,6 @@ static int btrfs_truncate(struct inode *inode)
 		goto out;
 	}
 
-	nr = trans->blocks_used;
-	btrfs_end_transaction(trans, root);
-	btrfs_btree_balance_dirty(root, nr);
-
-	/*
-	 * Ok so we've already migrated our bytes over for the truncate, so here
-	 * just reserve the one slot we need for updating the inode.
-	 */
-	trans = btrfs_start_transaction(root, 1);
-	if (IS_ERR(trans)) {
-		err = PTR_ERR(trans);
-		goto out;
-	}
-	trans->block_rsv = rsv;
-
 	/*
 	 * setattr is responsible for setting the ordered_data_close flag,
 	 * but that is only tested during the last file release.  That
@@ -6645,20 +6471,30 @@ static int btrfs_truncate(struct inode *inode)
 		btrfs_add_ordered_operation(trans, root, inode);
 
 	while (1) {
+		ret = btrfs_block_rsv_refill(root, rsv, min_size);
+		if (ret) {
+			/*
+			 * This can only happen with the original transaction we
+			 * started above, every other time we shouldn't have a
+			 * transaction started yet.
+			 */
+			if (ret == -EAGAIN)
+				goto end_trans;
+			err = ret;
+			break;
+		}
+
 		if (!trans) {
-			trans = btrfs_start_transaction(root, 3);
+			/* Just need the 1 for updating the inode */
+			trans = btrfs_start_transaction(root, 1);
 			if (IS_ERR(trans)) {
 				err = PTR_ERR(trans);
 				goto out;
 			}
-
-			ret = btrfs_truncate_reserve_metadata(trans, root,
-							      rsv);
-			BUG_ON(ret);
-
-			trans->block_rsv = rsv;
 		}
 
+		trans->block_rsv = rsv;
+
 		ret = btrfs_truncate_inode_items(trans, root, inode,
 						 inode->i_size,
 						 BTRFS_EXTENT_DATA_KEY);
@@ -6673,7 +6509,7 @@ static int btrfs_truncate(struct inode *inode)
 			err = ret;
 			break;
 		}
-
+end_trans:
 		nr = trans->blocks_used;
 		btrfs_end_transaction(trans, root);
 		trans = NULL;
@@ -6755,9 +6591,9 @@ struct inode *btrfs_alloc_inode(struct super_block *sb)
 	ei->last_sub_trans = 0;
 	ei->logged_trans = 0;
 	ei->delalloc_bytes = 0;
-	ei->reserved_bytes = 0;
 	ei->disk_i_size = 0;
 	ei->flags = 0;
+	ei->csum_bytes = 0;
 	ei->index_cnt = (u64)-1;
 	ei->last_unlink_trans = 0;
 
@@ -6803,6 +6639,8 @@ void btrfs_destroy_inode(struct inode *inode)
 	WARN_ON(inode->i_data.nrpages);
 	WARN_ON(BTRFS_I(inode)->outstanding_extents);
 	WARN_ON(BTRFS_I(inode)->reserved_extents);
+	WARN_ON(BTRFS_I(inode)->delalloc_bytes);
+	WARN_ON(BTRFS_I(inode)->csum_bytes);
 
 	/*
 	 * This can happen where we create an inode, but somebody else also
@@ -7420,7 +7258,6 @@ static struct extent_io_ops btrfs_extent_io_ops = {
 	.readpage_end_io_hook = btrfs_readpage_end_io_hook,
 	.writepage_end_io_hook = btrfs_writepage_end_io_hook,
 	.writepage_start_hook = btrfs_writepage_start_hook,
-	.readpage_io_failed_hook = btrfs_io_failed_hook,
 	.set_bit_hook = btrfs_set_bit_hook,
 	.clear_bit_hook = btrfs_clear_bit_hook,
 	.merge_extent_hook = btrfs_merge_extent_hook,

fs/btrfs/ioctl.c

@@ -51,6 +51,7 @@
 #include "volumes.h"
 #include "locking.h"
 #include "inode-map.h"
+#include "backref.h"
 
 /* Mask out flags that are inappropriate for the given type of inode. */
 static inline __u32 btrfs_mask_flags(umode_t mode, __u32 flags)
@@ -117,7 +118,7 @@ void btrfs_update_iflags(struct inode *inode)
 /*
  * Inherit flags from the parent inode.
  *
- * Unlike extN we don't have any flags we don't want to inherit currently.
+ * Currently only the compression flags and the cow flags are inherited.
  */
 void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 {
@@ -128,12 +129,17 @@ void btrfs_inherit_iflags(struct inode *inode, struct inode *dir)
 
 	flags = BTRFS_I(dir)->flags;
 
-	if (S_ISREG(inode->i_mode))
-		flags &= ~BTRFS_INODE_DIRSYNC;
-	else if (!S_ISDIR(inode->i_mode))
-		flags &= (BTRFS_INODE_NODUMP | BTRFS_INODE_NOATIME);
+	if (flags & BTRFS_INODE_NOCOMPRESS) {
+		BTRFS_I(inode)->flags &= ~BTRFS_INODE_COMPRESS;
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NOCOMPRESS;
+	} else if (flags & BTRFS_INODE_COMPRESS) {
+		BTRFS_I(inode)->flags &= ~BTRFS_INODE_NOCOMPRESS;
+		BTRFS_I(inode)->flags |= BTRFS_INODE_COMPRESS;
+	}
+
+	if (flags & BTRFS_INODE_NODATACOW)
+		BTRFS_I(inode)->flags |= BTRFS_INODE_NODATACOW;
 
-	BTRFS_I(inode)->flags = flags;
 	btrfs_update_iflags(inode);
 }
 
@@ -277,6 +283,7 @@ static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 	struct fstrim_range range;
 	u64 minlen = ULLONG_MAX;
 	u64 num_devices = 0;
+	u64 total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
 	int ret;
 
 	if (!capable(CAP_SYS_ADMIN))
@@ -295,12 +302,15 @@ static noinline int btrfs_ioctl_fitrim(struct file *file, void __user *arg)
 		}
 	}
 	rcu_read_unlock();
+
 	if (!num_devices)
 		return -EOPNOTSUPP;
-
 	if (copy_from_user(&range, arg, sizeof(range)))
 		return -EFAULT;
+	if (range.start > total_bytes)
+		return -EINVAL;
 
+	range.len = min(range.len, total_bytes - range.start);
 	range.minlen = max(range.minlen, minlen);
 	ret = btrfs_trim_fs(root, &range);
 	if (ret < 0)
@@ -760,7 +770,7 @@ static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 	int ret = 1;
 
 	/*
-	 * make sure that once we start defragging and extent, we keep on
+	 * make sure that once we start defragging an extent, we keep on
 	 * defragging it
 	 */
 	if (start < *defrag_end)
@@ -805,7 +815,6 @@ static int should_defrag_range(struct inode *inode, u64 start, u64 len,
 	 * extent will force at least part of that big extent to be defragged.
 	 */
 	if (ret) {
-		*last_len += len;
 		*defrag_end = extent_map_end(em);
 	} else {
 		*last_len = 0;
@@ -843,6 +852,7 @@ static int cluster_pages_for_defrag(struct inode *inode,
 	int i_done;
 	struct btrfs_ordered_extent *ordered;
 	struct extent_state *cached_state = NULL;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 
 	if (isize == 0)
 		return 0;
@@ -860,7 +870,7 @@ again:
 	for (i = 0; i < num_pages; i++) {
 		struct page *page;
 		page = find_or_create_page(inode->i_mapping,
-					    start_index + i, GFP_NOFS);
+					    start_index + i, mask);
 		if (!page)
 			break;
 
@@ -972,18 +982,20 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 	struct btrfs_super_block *disk_super;
 	struct file_ra_state *ra = NULL;
 	unsigned long last_index;
+	u64 isize = i_size_read(inode);
 	u64 features;
 	u64 last_len = 0;
 	u64 skip = 0;
 	u64 defrag_end = 0;
 	u64 newer_off = range->start;
-	int newer_left = 0;
 	unsigned long i;
+	unsigned long ra_index = 0;
 	int ret;
 	int defrag_count = 0;
 	int compress_type = BTRFS_COMPRESS_ZLIB;
 	int extent_thresh = range->extent_thresh;
-	int newer_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
+	int max_cluster = (256 * 1024) >> PAGE_CACHE_SHIFT;
+	int cluster = max_cluster;
 	u64 new_align = ~((u64)128 * 1024 - 1);
 	struct page **pages = NULL;
 
@@ -997,7 +1009,7 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 			compress_type = range->compress_type;
 	}
 
-	if (inode->i_size == 0)
+	if (isize == 0)
 		return 0;
 
 	/*
@@ -1013,7 +1025,7 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 		ra = &file->f_ra;
 	}
 
-	pages = kmalloc(sizeof(struct page *) * newer_cluster,
+	pages = kmalloc(sizeof(struct page *) * max_cluster,
 			GFP_NOFS);
 	if (!pages) {
 		ret = -ENOMEM;
@@ -1022,10 +1034,10 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 
 	/* find the last page to defrag */
 	if (range->start + range->len > range->start) {
-		last_index = min_t(u64, inode->i_size - 1,
+		last_index = min_t(u64, isize - 1,
 			 range->start + range->len - 1) >> PAGE_CACHE_SHIFT;
 	} else {
-		last_index = (inode->i_size - 1) >> PAGE_CACHE_SHIFT;
+		last_index = (isize - 1) >> PAGE_CACHE_SHIFT;
 	}
 
 	if (newer_than) {
@@ -1038,14 +1050,13 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 			 * the extents in the file evenly spaced
 			 */
 			i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
-			newer_left = newer_cluster;
 		} else
 			goto out_ra;
 	} else {
 		i = range->start >> PAGE_CACHE_SHIFT;
 	}
 	if (!max_to_defrag)
-		max_to_defrag = last_index - 1;
+		max_to_defrag = last_index;
 
 	/*
 	 * make writeback starts from i, so the defrag range can be
@@ -1079,18 +1090,31 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 			i = max(i + 1, next);
 			continue;
 		}
+
+		if (!newer_than) {
+			cluster = (PAGE_CACHE_ALIGN(defrag_end) >>
+				   PAGE_CACHE_SHIFT) - i;
+			cluster = min(cluster, max_cluster);
+		} else {
+			cluster = max_cluster;
+		}
+
 		if (range->flags & BTRFS_DEFRAG_RANGE_COMPRESS)
 			BTRFS_I(inode)->force_compress = compress_type;
 
-		btrfs_force_ra(inode->i_mapping, ra, file, i, newer_cluster);
+		if (i + cluster > ra_index) {
+			ra_index = max(i, ra_index);
+			btrfs_force_ra(inode->i_mapping, ra, file, ra_index,
+				       cluster);
+			ra_index += max_cluster;
+		}
 
-		ret = cluster_pages_for_defrag(inode, pages, i, newer_cluster);
+		ret = cluster_pages_for_defrag(inode, pages, i, cluster);
 		if (ret < 0)
 			goto out_ra;
 
 		defrag_count += ret;
 		balance_dirty_pages_ratelimited_nr(inode->i_mapping, ret);
-		i += ret;
 
 		if (newer_than) {
 			if (newer_off == (u64)-1)
@@ -1105,12 +1129,17 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 			if (!ret) {
 				range->start = newer_off;
 				i = (newer_off & new_align) >> PAGE_CACHE_SHIFT;
-				newer_left = newer_cluster;
 			} else {
 				break;
 			}
 		} else {
-			i++;
+			if (ret > 0) {
+				i += ret;
+				last_len += ret << PAGE_CACHE_SHIFT;
+			} else {
+				i++;
+				last_len = 0;
+			}
 		}
 	}
 
@@ -1136,16 +1165,14 @@ int btrfs_defrag_file(struct inode *inode, struct file *file,
 		mutex_unlock(&inode->i_mutex);
 	}
 
-	disk_super = &root->fs_info->super_copy;
+	disk_super = root->fs_info->super_copy;
 	features = btrfs_super_incompat_flags(disk_super);
 	if (range->compress_type == BTRFS_COMPRESS_LZO) {
 		features |= BTRFS_FEATURE_INCOMPAT_COMPRESS_LZO;
 		btrfs_set_super_incompat_flags(disk_super, features);
 	}
 
-	if (!file)
-		kfree(ra);
-	return defrag_count;
+	ret = defrag_count;
 
 out_ra:
 	if (!file)
@@ -2587,7 +2614,7 @@ static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
 		return PTR_ERR(trans);
 	}
 
-	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
+	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
 	di = btrfs_lookup_dir_item(trans, root->fs_info->tree_root, path,
 				   dir_id, "default", 7, 1);
 	if (IS_ERR_OR_NULL(di)) {
@@ -2603,7 +2630,7 @@ static long btrfs_ioctl_default_subvol(struct file *file, void __user *argp)
 	btrfs_mark_buffer_dirty(path->nodes[0]);
 	btrfs_free_path(path);
 
-	disk_super = &root->fs_info->super_copy;
+	disk_super = root->fs_info->super_copy;
 	features = btrfs_super_incompat_flags(disk_super);
 	if (!(features & BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL)) {
 		features |= BTRFS_FEATURE_INCOMPAT_DEFAULT_SUBVOL;
@@ -2864,6 +2891,144 @@ static long btrfs_ioctl_scrub_progress(struct btrfs_root *root,
 	return ret;
 }
 
+static long btrfs_ioctl_ino_to_path(struct btrfs_root *root, void __user *arg)
+{
+	int ret = 0;
+	int i;
+	u64 rel_ptr;
+	int size;
+	struct btrfs_ioctl_ino_path_args *ipa = NULL;
+	struct inode_fs_paths *ipath = NULL;
+	struct btrfs_path *path;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	ipa = memdup_user(arg, sizeof(*ipa));
+	if (IS_ERR(ipa)) {
+		ret = PTR_ERR(ipa);
+		ipa = NULL;
+		goto out;
+	}
+
+	size = min_t(u32, ipa->size, 4096);
+	ipath = init_ipath(size, root, path);
+	if (IS_ERR(ipath)) {
+		ret = PTR_ERR(ipath);
+		ipath = NULL;
+		goto out;
+	}
+
+	ret = paths_from_inode(ipa->inum, ipath);
+	if (ret < 0)
+		goto out;
+
+	for (i = 0; i < ipath->fspath->elem_cnt; ++i) {
+		rel_ptr = ipath->fspath->val[i] - (u64)ipath->fspath->val;
+		ipath->fspath->val[i] = rel_ptr;
+	}
+
+	ret = copy_to_user((void *)ipa->fspath, (void *)ipath->fspath, size);
+	if (ret) {
+		ret = -EFAULT;
+		goto out;
+	}
+
+out:
+	btrfs_free_path(path);
+	free_ipath(ipath);
+	kfree(ipa);
+
+	return ret;
+}
+
+static int build_ino_list(u64 inum, u64 offset, u64 root, void *ctx)
+{
+	struct btrfs_data_container *inodes = ctx;
+	const size_t c = 3 * sizeof(u64);
+
+	if (inodes->bytes_left >= c) {
+		inodes->bytes_left -= c;
+		inodes->val[inodes->elem_cnt] = inum;
+		inodes->val[inodes->elem_cnt + 1] = offset;
+		inodes->val[inodes->elem_cnt + 2] = root;
+		inodes->elem_cnt += 3;
+	} else {
+		inodes->bytes_missing += c - inodes->bytes_left;
+		inodes->bytes_left = 0;
+		inodes->elem_missed += 3;
+	}
+
+	return 0;
+}
+
+static long btrfs_ioctl_logical_to_ino(struct btrfs_root *root,
+					void __user *arg)
+{
+	int ret = 0;
+	int size;
+	u64 extent_offset;
+	struct btrfs_ioctl_logical_ino_args *loi;
+	struct btrfs_data_container *inodes = NULL;
+	struct btrfs_path *path = NULL;
+	struct btrfs_key key;
+
+	if (!capable(CAP_SYS_ADMIN))
+		return -EPERM;
+
+	loi = memdup_user(arg, sizeof(*loi));
+	if (IS_ERR(loi)) {
+		ret = PTR_ERR(loi);
+		loi = NULL;
+		goto out;
+	}
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	size = min_t(u32, loi->size, 4096);
+	inodes = init_data_container(size);
+	if (IS_ERR(inodes)) {
+		ret = PTR_ERR(inodes);
+		inodes = NULL;
+		goto out;
+	}
+
+	ret = extent_from_logical(root->fs_info, loi->logical, path, &key);
+
+	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK)
+		ret = -ENOENT;
+	if (ret < 0)
+		goto out;
+
+	extent_offset = loi->logical - key.objectid;
+	ret = iterate_extent_inodes(root->fs_info, path, key.objectid,
+					extent_offset, build_ino_list, inodes);
+
+	if (ret < 0)
+		goto out;
+
+	ret = copy_to_user((void *)loi->inodes, (void *)inodes, size);
+	if (ret)
+		ret = -EFAULT;
+
+out:
+	btrfs_free_path(path);
+	kfree(inodes);
+	kfree(loi);
+
+	return ret;
+}
+
 long btrfs_ioctl(struct file *file, unsigned int
 		cmd, unsigned long arg)
 {
@@ -2921,6 +3086,10 @@ long btrfs_ioctl(struct file *file, unsigned int
 		return btrfs_ioctl_tree_search(file, argp);
 	case BTRFS_IOC_INO_LOOKUP:
 		return btrfs_ioctl_ino_lookup(file, argp);
+	case BTRFS_IOC_INO_PATHS:
+		return btrfs_ioctl_ino_to_path(root, argp);
+	case BTRFS_IOC_LOGICAL_INO:
+		return btrfs_ioctl_logical_to_ino(root, argp);
 	case BTRFS_IOC_SPACE_INFO:
 		return btrfs_ioctl_space_info(root, argp);
 	case BTRFS_IOC_SYNC:

fs/btrfs/ioctl.h

@@ -193,6 +193,30 @@ struct btrfs_ioctl_space_args {
 	struct btrfs_ioctl_space_info spaces[0];
 };
 
+struct btrfs_data_container {
+	__u32	bytes_left;	/* out -- bytes not needed to deliver output */
+	__u32	bytes_missing;	/* out -- additional bytes needed for result */
+	__u32	elem_cnt;	/* out */
+	__u32	elem_missed;	/* out */
+	__u64	val[0];		/* out */
+};
+
+struct btrfs_ioctl_ino_path_args {
+	__u64				inum;		/* in */
+	__u32				size;		/* in */
+	__u64				reserved[4];
+	/* struct btrfs_data_container	*fspath;	   out */
+	__u64				fspath;		/* out */
+};
+
+struct btrfs_ioctl_logical_ino_args {
+	__u64				logical;	/* in */
+	__u32				size;		/* in */
+	__u64				reserved[4];
+	/* struct btrfs_data_container	*inodes;	out   */
+	__u64				inodes;
+};
+
 #define BTRFS_IOC_SNAP_CREATE _IOW(BTRFS_IOCTL_MAGIC, 1, \
 				   struct btrfs_ioctl_vol_args)
 #define BTRFS_IOC_DEFRAG _IOW(BTRFS_IOCTL_MAGIC, 2, \
@@ -248,4 +272,9 @@ struct btrfs_ioctl_space_args {
 				 struct btrfs_ioctl_dev_info_args)
 #define BTRFS_IOC_FS_INFO _IOR(BTRFS_IOCTL_MAGIC, 31, \
 			       struct btrfs_ioctl_fs_info_args)
+#define BTRFS_IOC_INO_PATHS _IOWR(BTRFS_IOCTL_MAGIC, 35, \
+					struct btrfs_ioctl_ino_path_args)
+#define BTRFS_IOC_LOGICAL_INO _IOWR(BTRFS_IOCTL_MAGIC, 36, \
+					struct btrfs_ioctl_ino_path_args)
+
 #endif

fs/btrfs/print-tree.c

@@ -158,8 +158,7 @@ static void print_extent_ref_v0(struct extent_buffer *eb, int slot)
 void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 {
 	int i;
-	u32 type;
-	u32 nr = btrfs_header_nritems(l);
+	u32 type, nr;
 	struct btrfs_item *item;
 	struct btrfs_root_item *ri;
 	struct btrfs_dir_item *di;
@@ -172,6 +171,11 @@ void btrfs_print_leaf(struct btrfs_root *root, struct extent_buffer *l)
 	struct btrfs_key key;
 	struct btrfs_key found_key;
 
+	if (!l)
+		return;
+
+	nr = btrfs_header_nritems(l);
+
 	printk(KERN_INFO "leaf %llu total ptrs %d free space %d\n",
 		(unsigned long long)btrfs_header_bytenr(l), nr,
 		btrfs_leaf_free_space(root, l));

fs/btrfs/reada.c

@@ -0,0 +1,951 @@
+/*
+ * Copyright (C) 2011 STRATO.  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/pagemap.h>
+#include <linux/writeback.h>
+#include <linux/blkdev.h>
+#include <linux/rbtree.h>
+#include <linux/slab.h>
+#include <linux/workqueue.h>
+#include "ctree.h"
+#include "volumes.h"
+#include "disk-io.h"
+#include "transaction.h"
+
+#undef DEBUG
+
+/*
+ * This is the implementation for the generic read ahead framework.
+ *
+ * To trigger a readahead, btrfs_reada_add must be called. It will start
+ * a read ahead for the given range [start, end) on tree root. The returned
+ * handle can either be used to wait on the readahead to finish
+ * (btrfs_reada_wait), or to send it to the background (btrfs_reada_detach).
+ *
+ * The read ahead works as follows:
+ * On btrfs_reada_add, the root of the tree is inserted into a radix_tree.
+ * reada_start_machine will then search for extents to prefetch and trigger
+ * some reads. When a read finishes for a node, all contained node/leaf
+ * pointers that lie in the given range will also be enqueued. The reads will
+ * be triggered in sequential order, thus giving a big win over a naive
+ * enumeration. It will also make use of multi-device layouts. Each disk
+ * will have its on read pointer and all disks will by utilized in parallel.
+ * Also will no two disks read both sides of a mirror simultaneously, as this
+ * would waste seeking capacity. Instead both disks will read different parts
+ * of the filesystem.
+ * Any number of readaheads can be started in parallel. The read order will be
+ * determined globally, i.e. 2 parallel readaheads will normally finish faster
+ * than the 2 started one after another.
+ */
+
+#define MAX_MIRRORS 2
+#define MAX_IN_FLIGHT 6
+
+struct reada_extctl {
+	struct list_head	list;
+	struct reada_control	*rc;
+	u64			generation;
+};
+
+struct reada_extent {
+	u64			logical;
+	struct btrfs_key	top;
+	u32			blocksize;
+	int			err;
+	struct list_head	extctl;
+	struct kref		refcnt;
+	spinlock_t		lock;
+	struct reada_zone	*zones[MAX_MIRRORS];
+	int			nzones;
+	struct btrfs_device	*scheduled_for;
+};
+
+struct reada_zone {
+	u64			start;
+	u64			end;
+	u64			elems;
+	struct list_head	list;
+	spinlock_t		lock;
+	int			locked;
+	struct btrfs_device	*device;
+	struct btrfs_device	*devs[MAX_MIRRORS]; /* full list, incl self */
+	int			ndevs;
+	struct kref		refcnt;
+};
+
+struct reada_machine_work {
+	struct btrfs_work	work;
+	struct btrfs_fs_info	*fs_info;
+};
+
+static void reada_extent_put(struct btrfs_fs_info *, struct reada_extent *);
+static void reada_control_release(struct kref *kref);
+static void reada_zone_release(struct kref *kref);
+static void reada_start_machine(struct btrfs_fs_info *fs_info);
+static void __reada_start_machine(struct btrfs_fs_info *fs_info);
+
+static int reada_add_block(struct reada_control *rc, u64 logical,
+			   struct btrfs_key *top, int level, u64 generation);
+
+/* recurses */
+/* in case of err, eb might be NULL */
+static int __readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			    u64 start, int err)
+{
+	int level = 0;
+	int nritems;
+	int i;
+	u64 bytenr;
+	u64 generation;
+	struct reada_extent *re;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct list_head list;
+	unsigned long index = start >> PAGE_CACHE_SHIFT;
+	struct btrfs_device *for_dev;
+
+	if (eb)
+		level = btrfs_header_level(eb);
+
+	/* find extent */
+	spin_lock(&fs_info->reada_lock);
+	re = radix_tree_lookup(&fs_info->reada_tree, index);
+	if (re)
+		kref_get(&re->refcnt);
+	spin_unlock(&fs_info->reada_lock);
+
+	if (!re)
+		return -1;
+
+	spin_lock(&re->lock);
+	/*
+	 * just take the full list from the extent. afterwards we
+	 * don't need the lock anymore
+	 */
+	list_replace_init(&re->extctl, &list);
+	for_dev = re->scheduled_for;
+	re->scheduled_for = NULL;
+	spin_unlock(&re->lock);
+
+	if (err == 0) {
+		nritems = level ? btrfs_header_nritems(eb) : 0;
+		generation = btrfs_header_generation(eb);
+		/*
+		 * FIXME: currently we just set nritems to 0 if this is a leaf,
+		 * effectively ignoring the content. In a next step we could
+		 * trigger more readahead depending from the content, e.g.
+		 * fetch the checksums for the extents in the leaf.
+		 */
+	} else {
+		/*
+		 * this is the error case, the extent buffer has not been
+		 * read correctly. We won't access anything from it and
+		 * just cleanup our data structures. Effectively this will
+		 * cut the branch below this node from read ahead.
+		 */
+		nritems = 0;
+		generation = 0;
+	}
+
+	for (i = 0; i < nritems; i++) {
+		struct reada_extctl *rec;
+		u64 n_gen;
+		struct btrfs_key key;
+		struct btrfs_key next_key;
+
+		btrfs_node_key_to_cpu(eb, &key, i);
+		if (i + 1 < nritems)
+			btrfs_node_key_to_cpu(eb, &next_key, i + 1);
+		else
+			next_key = re->top;
+		bytenr = btrfs_node_blockptr(eb, i);
+		n_gen = btrfs_node_ptr_generation(eb, i);
+
+		list_for_each_entry(rec, &list, list) {
+			struct reada_control *rc = rec->rc;
+
+			/*
+			 * if the generation doesn't match, just ignore this
+			 * extctl. This will probably cut off a branch from
+			 * prefetch. Alternatively one could start a new (sub-)
+			 * prefetch for this branch, starting again from root.
+			 * FIXME: move the generation check out of this loop
+			 */
+#ifdef DEBUG
+			if (rec->generation != generation) {
+				printk(KERN_DEBUG "generation mismatch for "
+						"(%llu,%d,%llu) %llu != %llu\n",
+				       key.objectid, key.type, key.offset,
+				       rec->generation, generation);
+			}
+#endif
+			if (rec->generation == generation &&
+			    btrfs_comp_cpu_keys(&key, &rc->key_end) < 0 &&
+			    btrfs_comp_cpu_keys(&next_key, &rc->key_start) > 0)
+				reada_add_block(rc, bytenr, &next_key,
+						level - 1, n_gen);
+		}
+	}
+	/*
+	 * free extctl records
+	 */
+	while (!list_empty(&list)) {
+		struct reada_control *rc;
+		struct reada_extctl *rec;
+
+		rec = list_first_entry(&list, struct reada_extctl, list);
+		list_del(&rec->list);
+		rc = rec->rc;
+		kfree(rec);
+
+		kref_get(&rc->refcnt);
+		if (atomic_dec_and_test(&rc->elems)) {
+			kref_put(&rc->refcnt, reada_control_release);
+			wake_up(&rc->wait);
+		}
+		kref_put(&rc->refcnt, reada_control_release);
+
+		reada_extent_put(fs_info, re);	/* one ref for each entry */
+	}
+	reada_extent_put(fs_info, re);	/* our ref */
+	if (for_dev)
+		atomic_dec(&for_dev->reada_in_flight);
+
+	return 0;
+}
+
+/*
+ * start is passed separately in case eb in NULL, which may be the case with
+ * failed I/O
+ */
+int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
+			 u64 start, int err)
+{
+	int ret;
+
+	ret = __readahead_hook(root, eb, start, err);
+
+	reada_start_machine(root->fs_info);
+
+	return ret;
+}
+
+static struct reada_zone *reada_find_zone(struct btrfs_fs_info *fs_info,
+					  struct btrfs_device *dev, u64 logical,
+					  struct btrfs_bio *bbio)
+{
+	int ret;
+	int looped = 0;
+	struct reada_zone *zone;
+	struct btrfs_block_group_cache *cache = NULL;
+	u64 start;
+	u64 end;
+	int i;
+
+again:
+	zone = NULL;
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_gang_lookup(&dev->reada_zones, (void **)&zone,
+				     logical >> PAGE_CACHE_SHIFT, 1);
+	if (ret == 1)
+		kref_get(&zone->refcnt);
+	spin_unlock(&fs_info->reada_lock);
+
+	if (ret == 1) {
+		if (logical >= zone->start && logical < zone->end)
+			return zone;
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+
+	if (looped)
+		return NULL;
+
+	cache = btrfs_lookup_block_group(fs_info, logical);
+	if (!cache)
+		return NULL;
+
+	start = cache->key.objectid;
+	end = start + cache->key.offset - 1;
+	btrfs_put_block_group(cache);
+
+	zone = kzalloc(sizeof(*zone), GFP_NOFS);
+	if (!zone)
+		return NULL;
+
+	zone->start = start;
+	zone->end = end;
+	INIT_LIST_HEAD(&zone->list);
+	spin_lock_init(&zone->lock);
+	zone->locked = 0;
+	kref_init(&zone->refcnt);
+	zone->elems = 0;
+	zone->device = dev; /* our device always sits at index 0 */
+	for (i = 0; i < bbio->num_stripes; ++i) {
+		/* bounds have already been checked */
+		zone->devs[i] = bbio->stripes[i].dev;
+	}
+	zone->ndevs = bbio->num_stripes;
+
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_insert(&dev->reada_zones,
+				(unsigned long)zone->end >> PAGE_CACHE_SHIFT,
+				zone);
+	spin_unlock(&fs_info->reada_lock);
+
+	if (ret) {
+		kfree(zone);
+		looped = 1;
+		goto again;
+	}
+
+	return zone;
+}
+
+static struct reada_extent *reada_find_extent(struct btrfs_root *root,
+					      u64 logical,
+					      struct btrfs_key *top, int level)
+{
+	int ret;
+	int looped = 0;
+	struct reada_extent *re = NULL;
+	struct btrfs_fs_info *fs_info = root->fs_info;
+	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
+	struct btrfs_bio *bbio = NULL;
+	struct btrfs_device *dev;
+	u32 blocksize;
+	u64 length;
+	int nzones = 0;
+	int i;
+	unsigned long index = logical >> PAGE_CACHE_SHIFT;
+
+again:
+	spin_lock(&fs_info->reada_lock);
+	re = radix_tree_lookup(&fs_info->reada_tree, index);
+	if (re)
+		kref_get(&re->refcnt);
+	spin_unlock(&fs_info->reada_lock);
+
+	if (re || looped)
+		return re;
+
+	re = kzalloc(sizeof(*re), GFP_NOFS);
+	if (!re)
+		return NULL;
+
+	blocksize = btrfs_level_size(root, level);
+	re->logical = logical;
+	re->blocksize = blocksize;
+	re->top = *top;
+	INIT_LIST_HEAD(&re->extctl);
+	spin_lock_init(&re->lock);
+	kref_init(&re->refcnt);
+
+	/*
+	 * map block
+	 */
+	length = blocksize;
+	ret = btrfs_map_block(map_tree, REQ_WRITE, logical, &length, &bbio, 0);
+	if (ret || !bbio || length < blocksize)
+		goto error;
+
+	if (bbio->num_stripes > MAX_MIRRORS) {
+		printk(KERN_ERR "btrfs readahead: more than %d copies not "
+				"supported", MAX_MIRRORS);
+		goto error;
+	}
+
+	for (nzones = 0; nzones < bbio->num_stripes; ++nzones) {
+		struct reada_zone *zone;
+
+		dev = bbio->stripes[nzones].dev;
+		zone = reada_find_zone(fs_info, dev, logical, bbio);
+		if (!zone)
+			break;
+
+		re->zones[nzones] = zone;
+		spin_lock(&zone->lock);
+		if (!zone->elems)
+			kref_get(&zone->refcnt);
+		++zone->elems;
+		spin_unlock(&zone->lock);
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	re->nzones = nzones;
+	if (nzones == 0) {
+		/* not a single zone found, error and out */
+		goto error;
+	}
+
+	/* insert extent in reada_tree + all per-device trees, all or nothing */
+	spin_lock(&fs_info->reada_lock);
+	ret = radix_tree_insert(&fs_info->reada_tree, index, re);
+	if (ret) {
+		spin_unlock(&fs_info->reada_lock);
+		if (ret != -ENOMEM) {
+			/* someone inserted the extent in the meantime */
+			looped = 1;
+		}
+		goto error;
+	}
+	for (i = 0; i < nzones; ++i) {
+		dev = bbio->stripes[i].dev;
+		ret = radix_tree_insert(&dev->reada_extents, index, re);
+		if (ret) {
+			while (--i >= 0) {
+				dev = bbio->stripes[i].dev;
+				BUG_ON(dev == NULL);
+				radix_tree_delete(&dev->reada_extents, index);
+			}
+			BUG_ON(fs_info == NULL);
+			radix_tree_delete(&fs_info->reada_tree, index);
+			spin_unlock(&fs_info->reada_lock);
+			goto error;
+		}
+	}
+	spin_unlock(&fs_info->reada_lock);
+
+	kfree(bbio);
+	return re;
+
+error:
+	while (nzones) {
+		struct reada_zone *zone;
+
+		--nzones;
+		zone = re->zones[nzones];
+		kref_get(&zone->refcnt);
+		spin_lock(&zone->lock);
+		--zone->elems;
+		if (zone->elems == 0) {
+			/*
+			 * no fs_info->reada_lock needed, as this can't be
+			 * the last ref
+			 */
+			kref_put(&zone->refcnt, reada_zone_release);
+		}
+		spin_unlock(&zone->lock);
+
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	kfree(bbio);
+	kfree(re);
+	if (looped)
+		goto again;
+	return NULL;
+}
+
+static void reada_kref_dummy(struct kref *kr)
+{
+}
+
+static void reada_extent_put(struct btrfs_fs_info *fs_info,
+			     struct reada_extent *re)
+{
+	int i;
+	unsigned long index = re->logical >> PAGE_CACHE_SHIFT;
+
+	spin_lock(&fs_info->reada_lock);
+	if (!kref_put(&re->refcnt, reada_kref_dummy)) {
+		spin_unlock(&fs_info->reada_lock);
+		return;
+	}
+
+	radix_tree_delete(&fs_info->reada_tree, index);
+	for (i = 0; i < re->nzones; ++i) {
+		struct reada_zone *zone = re->zones[i];
+
+		radix_tree_delete(&zone->device->reada_extents, index);
+	}
+
+	spin_unlock(&fs_info->reada_lock);
+
+	for (i = 0; i < re->nzones; ++i) {
+		struct reada_zone *zone = re->zones[i];
+
+		kref_get(&zone->refcnt);
+		spin_lock(&zone->lock);
+		--zone->elems;
+		if (zone->elems == 0) {
+			/* no fs_info->reada_lock needed, as this can't be
+			 * the last ref */
+			kref_put(&zone->refcnt, reada_zone_release);
+		}
+		spin_unlock(&zone->lock);
+
+		spin_lock(&fs_info->reada_lock);
+		kref_put(&zone->refcnt, reada_zone_release);
+		spin_unlock(&fs_info->reada_lock);
+	}
+	if (re->scheduled_for)
+		atomic_dec(&re->scheduled_for->reada_in_flight);
+
+	kfree(re);
+}
+
+static void reada_zone_release(struct kref *kref)
+{
+	struct reada_zone *zone = container_of(kref, struct reada_zone, refcnt);
+
+	radix_tree_delete(&zone->device->reada_zones,
+			  zone->end >> PAGE_CACHE_SHIFT);
+
+	kfree(zone);
+}
+
+static void reada_control_release(struct kref *kref)
+{
+	struct reada_control *rc = container_of(kref, struct reada_control,
+						refcnt);
+
+	kfree(rc);
+}
+
+static int reada_add_block(struct reada_control *rc, u64 logical,
+			   struct btrfs_key *top, int level, u64 generation)
+{
+	struct btrfs_root *root = rc->root;
+	struct reada_extent *re;
+	struct reada_extctl *rec;
+
+	re = reada_find_extent(root, logical, top, level); /* takes one ref */
+	if (!re)
+		return -1;
+
+	rec = kzalloc(sizeof(*rec), GFP_NOFS);
+	if (!rec) {
+		reada_extent_put(root->fs_info, re);
+		return -1;
+	}
+
+	rec->rc = rc;
+	rec->generation = generation;
+	atomic_inc(&rc->elems);
+
+	spin_lock(&re->lock);
+	list_add_tail(&rec->list, &re->extctl);
+	spin_unlock(&re->lock);
+
+	/* leave the ref on the extent */
+
+	return 0;
+}
+
+/*
+ * called with fs_info->reada_lock held
+ */
+static void reada_peer_zones_set_lock(struct reada_zone *zone, int lock)
+{
+	int i;
+	unsigned long index = zone->end >> PAGE_CACHE_SHIFT;
+
+	for (i = 0; i < zone->ndevs; ++i) {
+		struct reada_zone *peer;
+		peer = radix_tree_lookup(&zone->devs[i]->reada_zones, index);
+		if (peer && peer->device != zone->device)
+			peer->locked = lock;
+	}
+}
+
+/*
+ * called with fs_info->reada_lock held
+ */
+static int reada_pick_zone(struct btrfs_device *dev)
+{
+	struct reada_zone *top_zone = NULL;
+	struct reada_zone *top_locked_zone = NULL;
+	u64 top_elems = 0;
+	u64 top_locked_elems = 0;
+	unsigned long index = 0;
+	int ret;
+
+	if (dev->reada_curr_zone) {
+		reada_peer_zones_set_lock(dev->reada_curr_zone, 0);
+		kref_put(&dev->reada_curr_zone->refcnt, reada_zone_release);
+		dev->reada_curr_zone = NULL;
+	}
+	/* pick the zone with the most elements */
+	while (1) {
+		struct reada_zone *zone;
+
+		ret = radix_tree_gang_lookup(&dev->reada_zones,
+					     (void **)&zone, index, 1);
+		if (ret == 0)
+			break;
+		index = (zone->end >> PAGE_CACHE_SHIFT) + 1;
+		if (zone->locked) {
+			if (zone->elems > top_locked_elems) {
+				top_locked_elems = zone->elems;
+				top_locked_zone = zone;
+			}
+		} else {
+			if (zone->elems > top_elems) {
+				top_elems = zone->elems;
+				top_zone = zone;
+			}
+		}
+	}
+	if (top_zone)
+		dev->reada_curr_zone = top_zone;
+	else if (top_locked_zone)
+		dev->reada_curr_zone = top_locked_zone;
+	else
+		return 0;
+
+	dev->reada_next = dev->reada_curr_zone->start;
+	kref_get(&dev->reada_curr_zone->refcnt);
+	reada_peer_zones_set_lock(dev->reada_curr_zone, 1);
+
+	return 1;
+}
+
+static int reada_start_machine_dev(struct btrfs_fs_info *fs_info,
+				   struct btrfs_device *dev)
+{
+	struct reada_extent *re = NULL;
+	int mirror_num = 0;
+	struct extent_buffer *eb = NULL;
+	u64 logical;
+	u32 blocksize;
+	int ret;
+	int i;
+	int need_kick = 0;
+
+	spin_lock(&fs_info->reada_lock);
+	if (dev->reada_curr_zone == NULL) {
+		ret = reada_pick_zone(dev);
+		if (!ret) {
+			spin_unlock(&fs_info->reada_lock);
+			return 0;
+		}
+	}
+	/*
+	 * FIXME currently we issue the reads one extent at a time. If we have
+	 * a contiguous block of extents, we could also coagulate them or use
+	 * plugging to speed things up
+	 */
+	ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
+				     dev->reada_next >> PAGE_CACHE_SHIFT, 1);
+	if (ret == 0 || re->logical >= dev->reada_curr_zone->end) {
+		ret = reada_pick_zone(dev);
+		if (!ret) {
+			spin_unlock(&fs_info->reada_lock);
+			return 0;
+		}
+		re = NULL;
+		ret = radix_tree_gang_lookup(&dev->reada_extents, (void **)&re,
+					dev->reada_next >> PAGE_CACHE_SHIFT, 1);
+	}
+	if (ret == 0) {
+		spin_unlock(&fs_info->reada_lock);
+		return 0;
+	}
+	dev->reada_next = re->logical + re->blocksize;
+	kref_get(&re->refcnt);
+
+	spin_unlock(&fs_info->reada_lock);
+
+	/*
+	 * find mirror num
+	 */
+	for (i = 0; i < re->nzones; ++i) {
+		if (re->zones[i]->device == dev) {
+			mirror_num = i + 1;
+			break;
+		}
+	}
+	logical = re->logical;
+	blocksize = re->blocksize;
+
+	spin_lock(&re->lock);
+	if (re->scheduled_for == NULL) {
+		re->scheduled_for = dev;
+		need_kick = 1;
+	}
+	spin_unlock(&re->lock);
+
+	reada_extent_put(fs_info, re);
+
+	if (!need_kick)
+		return 0;
+
+	atomic_inc(&dev->reada_in_flight);
+	ret = reada_tree_block_flagged(fs_info->extent_root, logical, blocksize,
+			 mirror_num, &eb);
+	if (ret)
+		__readahead_hook(fs_info->extent_root, NULL, logical, ret);
+	else if (eb)
+		__readahead_hook(fs_info->extent_root, eb, eb->start, ret);
+
+	if (eb)
+		free_extent_buffer(eb);
+
+	return 1;
+
+}
+
+static void reada_start_machine_worker(struct btrfs_work *work)
+{
+	struct reada_machine_work *rmw;
+	struct btrfs_fs_info *fs_info;
+
+	rmw = container_of(work, struct reada_machine_work, work);
+	fs_info = rmw->fs_info;
+
+	kfree(rmw);
+
+	__reada_start_machine(fs_info);
+}
+
+static void __reada_start_machine(struct btrfs_fs_info *fs_info)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	u64 enqueued;
+	u64 total = 0;
+	int i;
+
+	do {
+		enqueued = 0;
+		list_for_each_entry(device, &fs_devices->devices, dev_list) {
+			if (atomic_read(&device->reada_in_flight) <
+			    MAX_IN_FLIGHT)
+				enqueued += reada_start_machine_dev(fs_info,
+								    device);
+		}
+		total += enqueued;
+	} while (enqueued && total < 10000);
+
+	if (enqueued == 0)
+		return;
+
+	/*
+	 * If everything is already in the cache, this is effectively single
+	 * threaded. To a) not hold the caller for too long and b) to utilize
+	 * more cores, we broke the loop above after 10000 iterations and now
+	 * enqueue to workers to finish it. This will distribute the load to
+	 * the cores.
+	 */
+	for (i = 0; i < 2; ++i)
+		reada_start_machine(fs_info);
+}
+
+static void reada_start_machine(struct btrfs_fs_info *fs_info)
+{
+	struct reada_machine_work *rmw;
+
+	rmw = kzalloc(sizeof(*rmw), GFP_NOFS);
+	if (!rmw) {
+		/* FIXME we cannot handle this properly right now */
+		BUG();
+	}
+	rmw->work.func = reada_start_machine_worker;
+	rmw->fs_info = fs_info;
+
+	btrfs_queue_worker(&fs_info->readahead_workers, &rmw->work);
+}
+
+#ifdef DEBUG
+static void dump_devs(struct btrfs_fs_info *fs_info, int all)
+{
+	struct btrfs_device *device;
+	struct btrfs_fs_devices *fs_devices = fs_info->fs_devices;
+	unsigned long index;
+	int ret;
+	int i;
+	int j;
+	int cnt;
+
+	spin_lock(&fs_info->reada_lock);
+	list_for_each_entry(device, &fs_devices->devices, dev_list) {
+		printk(KERN_DEBUG "dev %lld has %d in flight\n", device->devid,
+			atomic_read(&device->reada_in_flight));
+		index = 0;
+		while (1) {
+			struct reada_zone *zone;
+			ret = radix_tree_gang_lookup(&device->reada_zones,
+						     (void **)&zone, index, 1);
+			if (ret == 0)
+				break;
+			printk(KERN_DEBUG "  zone %llu-%llu elems %llu locked "
+				"%d devs", zone->start, zone->end, zone->elems,
+				zone->locked);
+			for (j = 0; j < zone->ndevs; ++j) {
+				printk(KERN_CONT " %lld",
+					zone->devs[j]->devid);
+			}
+			if (device->reada_curr_zone == zone)
+				printk(KERN_CONT " curr off %llu",
+					device->reada_next - zone->start);
+			printk(KERN_CONT "\n");
+			index = (zone->end >> PAGE_CACHE_SHIFT) + 1;
+		}
+		cnt = 0;
+		index = 0;
+		while (all) {
+			struct reada_extent *re = NULL;
+
+			ret = radix_tree_gang_lookup(&device->reada_extents,
+						     (void **)&re, index, 1);
+			if (ret == 0)
+				break;
+			printk(KERN_DEBUG
+				"  re: logical %llu size %u empty %d for %lld",
+				re->logical, re->blocksize,
+				list_empty(&re->extctl), re->scheduled_for ?
+				re->scheduled_for->devid : -1);
+
+			for (i = 0; i < re->nzones; ++i) {
+				printk(KERN_CONT " zone %llu-%llu devs",
+					re->zones[i]->start,
+					re->zones[i]->end);
+				for (j = 0; j < re->zones[i]->ndevs; ++j) {
+					printk(KERN_CONT " %lld",
+						re->zones[i]->devs[j]->devid);
+				}
+			}
+			printk(KERN_CONT "\n");
+			index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+			if (++cnt > 15)
+				break;
+		}
+	}
+
+	index = 0;
+	cnt = 0;
+	while (all) {
+		struct reada_extent *re = NULL;
+
+		ret = radix_tree_gang_lookup(&fs_info->reada_tree, (void **)&re,
+					     index, 1);
+		if (ret == 0)
+			break;
+		if (!re->scheduled_for) {
+			index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+			continue;
+		}
+		printk(KERN_DEBUG
+			"re: logical %llu size %u list empty %d for %lld",
+			re->logical, re->blocksize, list_empty(&re->extctl),
+			re->scheduled_for ? re->scheduled_for->devid : -1);
+		for (i = 0; i < re->nzones; ++i) {
+			printk(KERN_CONT " zone %llu-%llu devs",
+				re->zones[i]->start,
+				re->zones[i]->end);
+			for (i = 0; i < re->nzones; ++i) {
+				printk(KERN_CONT " zone %llu-%llu devs",
+					re->zones[i]->start,
+					re->zones[i]->end);
+				for (j = 0; j < re->zones[i]->ndevs; ++j) {
+					printk(KERN_CONT " %lld",
+						re->zones[i]->devs[j]->devid);
+				}
+			}
+		}
+		printk(KERN_CONT "\n");
+		index = (re->logical >> PAGE_CACHE_SHIFT) + 1;
+	}
+	spin_unlock(&fs_info->reada_lock);
+}
+#endif
+
+/*
+ * interface
+ */
+struct reada_control *btrfs_reada_add(struct btrfs_root *root,
+			struct btrfs_key *key_start, struct btrfs_key *key_end)
+{
+	struct reada_control *rc;
+	u64 start;
+	u64 generation;
+	int level;
+	struct extent_buffer *node;
+	static struct btrfs_key max_key = {
+		.objectid = (u64)-1,
+		.type = (u8)-1,
+		.offset = (u64)-1
+	};
+
+	rc = kzalloc(sizeof(*rc), GFP_NOFS);
+	if (!rc)
+		return ERR_PTR(-ENOMEM);
+
+	rc->root = root;
+	rc->key_start = *key_start;
+	rc->key_end = *key_end;
+	atomic_set(&rc->elems, 0);
+	init_waitqueue_head(&rc->wait);
+	kref_init(&rc->refcnt);
+	kref_get(&rc->refcnt); /* one ref for having elements */
+
+	node = btrfs_root_node(root);
+	start = node->start;
+	level = btrfs_header_level(node);
+	generation = btrfs_header_generation(node);
+	free_extent_buffer(node);
+
+	reada_add_block(rc, start, &max_key, level, generation);
+
+	reada_start_machine(root->fs_info);
+
+	return rc;
+}
+
+#ifdef DEBUG
+int btrfs_reada_wait(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	while (atomic_read(&rc->elems)) {
+		wait_event_timeout(rc->wait, atomic_read(&rc->elems) == 0,
+				   5 * HZ);
+		dump_devs(rc->root->fs_info, rc->elems < 10 ? 1 : 0);
+	}
+
+	dump_devs(rc->root->fs_info, rc->elems < 10 ? 1 : 0);
+
+	kref_put(&rc->refcnt, reada_control_release);
+
+	return 0;
+}
+#else
+int btrfs_reada_wait(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	while (atomic_read(&rc->elems)) {
+		wait_event(rc->wait, atomic_read(&rc->elems) == 0);
+	}
+
+	kref_put(&rc->refcnt, reada_control_release);
+
+	return 0;
+}
+#endif
+
+void btrfs_reada_detach(void *handle)
+{
+	struct reada_control *rc = handle;
+
+	kref_put(&rc->refcnt, reada_control_release);
+}

fs/btrfs/relocation.c

@@ -2041,8 +2041,7 @@ static noinline_for_stack int merge_reloc_root(struct reloc_control *rc,
 		BUG_ON(IS_ERR(trans));
 		trans->block_rsv = rc->block_rsv;
 
-		ret = btrfs_block_rsv_check(trans, root, rc->block_rsv,
-					    min_reserved, 0);
+		ret = btrfs_block_rsv_refill(root, rc->block_rsv, min_reserved);
 		if (ret) {
 			BUG_ON(ret != -EAGAIN);
 			ret = btrfs_commit_transaction(trans, root);
@@ -2152,8 +2151,7 @@ int prepare_to_merge(struct reloc_control *rc, int err)
 again:
 	if (!err) {
 		num_bytes = rc->merging_rsv_size;
-		ret = btrfs_block_rsv_add(NULL, root, rc->block_rsv,
-					  num_bytes);
+		ret = btrfs_block_rsv_add(root, rc->block_rsv, num_bytes);
 		if (ret)
 			err = ret;
 	}
@@ -2427,7 +2425,7 @@ static int reserve_metadata_space(struct btrfs_trans_handle *trans,
 	num_bytes = calcu_metadata_size(rc, node, 1) * 2;
 
 	trans->block_rsv = rc->block_rsv;
-	ret = btrfs_block_rsv_add(trans, root, rc->block_rsv, num_bytes);
+	ret = btrfs_block_rsv_add(root, rc->block_rsv, num_bytes);
 	if (ret) {
 		if (ret == -EAGAIN)
 			rc->commit_transaction = 1;
@@ -2922,6 +2920,7 @@ static int relocate_file_extent_cluster(struct inode *inode,
 	unsigned long last_index;
 	struct page *page;
 	struct file_ra_state *ra;
+	gfp_t mask = btrfs_alloc_write_mask(inode->i_mapping);
 	int nr = 0;
 	int ret = 0;
 
@@ -2956,7 +2955,7 @@ static int relocate_file_extent_cluster(struct inode *inode,
 						  ra, NULL, index,
 						  last_index + 1 - index);
 			page = find_or_create_page(inode->i_mapping, index,
-						   GFP_NOFS);
+						   mask);
 			if (!page) {
 				btrfs_delalloc_release_metadata(inode,
 							PAGE_CACHE_SIZE);
@@ -3323,8 +3322,11 @@ static int find_data_references(struct reloc_control *rc,
 	}
 
 	key.objectid = ref_objectid;
-	key.offset = ref_offset;
 	key.type = BTRFS_EXTENT_DATA_KEY;
+	if (ref_offset > ((u64)-1 << 32))
+		key.offset = 0;
+	else
+		key.offset = ref_offset;
 
 	path->search_commit_root = 1;
 	path->skip_locking = 1;
@@ -3645,14 +3647,11 @@ int prepare_to_relocate(struct reloc_control *rc)
 	 * btrfs_init_reloc_root will use them when there
 	 * is no reservation in transaction handle.
 	 */
-	ret = btrfs_block_rsv_add(NULL, rc->extent_root, rc->block_rsv,
+	ret = btrfs_block_rsv_add(rc->extent_root, rc->block_rsv,
 				  rc->extent_root->nodesize * 256);
 	if (ret)
 		return ret;
 
-	rc->block_rsv->refill_used = 1;
-	btrfs_add_durable_block_rsv(rc->extent_root->fs_info, rc->block_rsv);
-
 	memset(&rc->cluster, 0, sizeof(rc->cluster));
 	rc->search_start = rc->block_group->key.objectid;
 	rc->extents_found = 0;
@@ -3777,8 +3776,7 @@ restart:
 			}
 		}
 
-		ret = btrfs_block_rsv_check(trans, rc->extent_root,
-					    rc->block_rsv, 0, 5);
+		ret = btrfs_block_rsv_check(rc->extent_root, rc->block_rsv, 5);
 		if (ret < 0) {
 			if (ret != -EAGAIN) {
 				err = ret;

fs/btrfs/scrub.c

@@ -17,10 +17,14 @@
  */
 
 #include <linux/blkdev.h>
+#include <linux/ratelimit.h>
 #include "ctree.h"
 #include "volumes.h"
 #include "disk-io.h"
 #include "ordered-data.h"
+#include "transaction.h"
+#include "backref.h"
+#include "extent_io.h"
 
 /*
  * This is only the first step towards a full-features scrub. It reads all
@@ -29,15 +33,12 @@
  * any can be found.
  *
  * Future enhancements:
- *  - To enhance the performance, better read-ahead strategies for the
- *    extent-tree can be employed.
  *  - In case an unrepairable extent is encountered, track which files are
  *    affected and report them
  *  - In case of a read error on files with nodatasum, map the file and read
  *    the extent to trigger a writeback of the good copy
  *  - track and record media errors, throw out bad devices
  *  - add a mode to also read unallocated space
- *  - make the prefetch cancellable
  */
 
 struct scrub_bio;
@@ -63,7 +64,7 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix);
 struct scrub_page {
 	u64			flags;  /* extent flags */
 	u64			generation;
-	u64			mirror_num;
+	int			mirror_num;
 	int			have_csum;
 	u8			csum[BTRFS_CSUM_SIZE];
 };
@@ -87,6 +88,7 @@ struct scrub_dev {
 	int			first_free;
 	int			curr;
 	atomic_t		in_flight;
+	atomic_t		fixup_cnt;
 	spinlock_t		list_lock;
 	wait_queue_head_t	list_wait;
 	u16			csum_size;
@@ -100,6 +102,27 @@ struct scrub_dev {
 	spinlock_t		stat_lock;
 };
 
+struct scrub_fixup_nodatasum {
+	struct scrub_dev	*sdev;
+	u64			logical;
+	struct btrfs_root	*root;
+	struct btrfs_work	work;
+	int			mirror_num;
+};
+
+struct scrub_warning {
+	struct btrfs_path	*path;
+	u64			extent_item_size;
+	char			*scratch_buf;
+	char			*msg_buf;
+	const char		*errstr;
+	sector_t		sector;
+	u64			logical;
+	struct btrfs_device	*dev;
+	int			msg_bufsize;
+	int			scratch_bufsize;
+};
+
 static void scrub_free_csums(struct scrub_dev *sdev)
 {
 	while (!list_empty(&sdev->csum_list)) {
@@ -175,14 +198,15 @@ struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev)
 
 		if (i != SCRUB_BIOS_PER_DEV-1)
 			sdev->bios[i]->next_free = i + 1;
-		 else
+		else
 			sdev->bios[i]->next_free = -1;
 	}
 	sdev->first_free = 0;
 	sdev->curr = -1;
 	atomic_set(&sdev->in_flight, 0);
+	atomic_set(&sdev->fixup_cnt, 0);
 	atomic_set(&sdev->cancel_req, 0);
-	sdev->csum_size = btrfs_super_csum_size(&fs_info->super_copy);
+	sdev->csum_size = btrfs_super_csum_size(fs_info->super_copy);
 	INIT_LIST_HEAD(&sdev->csum_list);
 
 	spin_lock_init(&sdev->list_lock);
@@ -195,24 +219,361 @@ nomem:
 	return ERR_PTR(-ENOMEM);
 }
 
+static int scrub_print_warning_inode(u64 inum, u64 offset, u64 root, void *ctx)
+{
+	u64 isize;
+	u32 nlink;
+	int ret;
+	int i;
+	struct extent_buffer *eb;
+	struct btrfs_inode_item *inode_item;
+	struct scrub_warning *swarn = ctx;
+	struct btrfs_fs_info *fs_info = swarn->dev->dev_root->fs_info;
+	struct inode_fs_paths *ipath = NULL;
+	struct btrfs_root *local_root;
+	struct btrfs_key root_key;
+
+	root_key.objectid = root;
+	root_key.type = BTRFS_ROOT_ITEM_KEY;
+	root_key.offset = (u64)-1;
+	local_root = btrfs_read_fs_root_no_name(fs_info, &root_key);
+	if (IS_ERR(local_root)) {
+		ret = PTR_ERR(local_root);
+		goto err;
+	}
+
+	ret = inode_item_info(inum, 0, local_root, swarn->path);
+	if (ret) {
+		btrfs_release_path(swarn->path);
+		goto err;
+	}
+
+	eb = swarn->path->nodes[0];
+	inode_item = btrfs_item_ptr(eb, swarn->path->slots[0],
+					struct btrfs_inode_item);
+	isize = btrfs_inode_size(eb, inode_item);
+	nlink = btrfs_inode_nlink(eb, inode_item);
+	btrfs_release_path(swarn->path);
+
+	ipath = init_ipath(4096, local_root, swarn->path);
+	ret = paths_from_inode(inum, ipath);
+
+	if (ret < 0)
+		goto err;
+
+	/*
+	 * we deliberately ignore the bit ipath might have been too small to
+	 * hold all of the paths here
+	 */
+	for (i = 0; i < ipath->fspath->elem_cnt; ++i)
+		printk(KERN_WARNING "btrfs: %s at logical %llu on dev "
+			"%s, sector %llu, root %llu, inode %llu, offset %llu, "
+			"length %llu, links %u (path: %s)\n", swarn->errstr,
+			swarn->logical, swarn->dev->name,
+			(unsigned long long)swarn->sector, root, inum, offset,
+			min(isize - offset, (u64)PAGE_SIZE), nlink,
+			(char *)ipath->fspath->val[i]);
+
+	free_ipath(ipath);
+	return 0;
+
+err:
+	printk(KERN_WARNING "btrfs: %s at logical %llu on dev "
+		"%s, sector %llu, root %llu, inode %llu, offset %llu: path "
+		"resolving failed with ret=%d\n", swarn->errstr,
+		swarn->logical, swarn->dev->name,
+		(unsigned long long)swarn->sector, root, inum, offset, ret);
+
+	free_ipath(ipath);
+	return 0;
+}
+
+static void scrub_print_warning(const char *errstr, struct scrub_bio *sbio,
+				int ix)
+{
+	struct btrfs_device *dev = sbio->sdev->dev;
+	struct btrfs_fs_info *fs_info = dev->dev_root->fs_info;
+	struct btrfs_path *path;
+	struct btrfs_key found_key;
+	struct extent_buffer *eb;
+	struct btrfs_extent_item *ei;
+	struct scrub_warning swarn;
+	u32 item_size;
+	int ret;
+	u64 ref_root;
+	u8 ref_level;
+	unsigned long ptr = 0;
+	const int bufsize = 4096;
+	u64 extent_offset;
+
+	path = btrfs_alloc_path();
+
+	swarn.scratch_buf = kmalloc(bufsize, GFP_NOFS);
+	swarn.msg_buf = kmalloc(bufsize, GFP_NOFS);
+	swarn.sector = (sbio->physical + ix * PAGE_SIZE) >> 9;
+	swarn.logical = sbio->logical + ix * PAGE_SIZE;
+	swarn.errstr = errstr;
+	swarn.dev = dev;
+	swarn.msg_bufsize = bufsize;
+	swarn.scratch_bufsize = bufsize;
+
+	if (!path || !swarn.scratch_buf || !swarn.msg_buf)
+		goto out;
+
+	ret = extent_from_logical(fs_info, swarn.logical, path, &found_key);
+	if (ret < 0)
+		goto out;
+
+	extent_offset = swarn.logical - found_key.objectid;
+	swarn.extent_item_size = found_key.offset;
+
+	eb = path->nodes[0];
+	ei = btrfs_item_ptr(eb, path->slots[0], struct btrfs_extent_item);
+	item_size = btrfs_item_size_nr(eb, path->slots[0]);
+
+	if (ret & BTRFS_EXTENT_FLAG_TREE_BLOCK) {
+		do {
+			ret = tree_backref_for_extent(&ptr, eb, ei, item_size,
+							&ref_root, &ref_level);
+			printk(KERN_WARNING "%s at logical %llu on dev %s, "
+				"sector %llu: metadata %s (level %d) in tree "
+				"%llu\n", errstr, swarn.logical, dev->name,
+				(unsigned long long)swarn.sector,
+				ref_level ? "node" : "leaf",
+				ret < 0 ? -1 : ref_level,
+				ret < 0 ? -1 : ref_root);
+		} while (ret != 1);
+	} else {
+		swarn.path = path;
+		iterate_extent_inodes(fs_info, path, found_key.objectid,
+					extent_offset,
+					scrub_print_warning_inode, &swarn);
+	}
+
+out:
+	btrfs_free_path(path);
+	kfree(swarn.scratch_buf);
+	kfree(swarn.msg_buf);
+}
+
+static int scrub_fixup_readpage(u64 inum, u64 offset, u64 root, void *ctx)
+{
+	struct page *page = NULL;
+	unsigned long index;
+	struct scrub_fixup_nodatasum *fixup = ctx;
+	int ret;
+	int corrected = 0;
+	struct btrfs_key key;
+	struct inode *inode = NULL;
+	u64 end = offset + PAGE_SIZE - 1;
+	struct btrfs_root *local_root;
+
+	key.objectid = root;
+	key.type = BTRFS_ROOT_ITEM_KEY;
+	key.offset = (u64)-1;
+	local_root = btrfs_read_fs_root_no_name(fixup->root->fs_info, &key);
+	if (IS_ERR(local_root))
+		return PTR_ERR(local_root);
+
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.objectid = inum;
+	key.offset = 0;
+	inode = btrfs_iget(fixup->root->fs_info->sb, &key, local_root, NULL);
+	if (IS_ERR(inode))
+		return PTR_ERR(inode);
+
+	index = offset >> PAGE_CACHE_SHIFT;
+
+	page = find_or_create_page(inode->i_mapping, index, GFP_NOFS);
+	if (!page) {
+		ret = -ENOMEM;
+		goto out;
+	}
+
+	if (PageUptodate(page)) {
+		struct btrfs_mapping_tree *map_tree;
+		if (PageDirty(page)) {
+			/*
+			 * we need to write the data to the defect sector. the
+			 * data that was in that sector is not in memory,
+			 * because the page was modified. we must not write the
+			 * modified page to that sector.
+			 *
+			 * TODO: what could be done here: wait for the delalloc
+			 *       runner to write out that page (might involve
+			 *       COW) and see whether the sector is still
+			 *       referenced afterwards.
+			 *
+			 * For the meantime, we'll treat this error
+			 * incorrectable, although there is a chance that a
+			 * later scrub will find the bad sector again and that
+			 * there's no dirty page in memory, then.
+			 */
+			ret = -EIO;
+			goto out;
+		}
+		map_tree = &BTRFS_I(inode)->root->fs_info->mapping_tree;
+		ret = repair_io_failure(map_tree, offset, PAGE_SIZE,
+					fixup->logical, page,
+					fixup->mirror_num);
+		unlock_page(page);
+		corrected = !ret;
+	} else {
+		/*
+		 * we need to get good data first. the general readpage path
+		 * will call repair_io_failure for us, we just have to make
+		 * sure we read the bad mirror.
+		 */
+		ret = set_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
+					EXTENT_DAMAGED, GFP_NOFS);
+		if (ret) {
+			/* set_extent_bits should give proper error */
+			WARN_ON(ret > 0);
+			if (ret > 0)
+				ret = -EFAULT;
+			goto out;
+		}
+
+		ret = extent_read_full_page(&BTRFS_I(inode)->io_tree, page,
+						btrfs_get_extent,
+						fixup->mirror_num);
+		wait_on_page_locked(page);
+
+		corrected = !test_range_bit(&BTRFS_I(inode)->io_tree, offset,
+						end, EXTENT_DAMAGED, 0, NULL);
+		if (!corrected)
+			clear_extent_bits(&BTRFS_I(inode)->io_tree, offset, end,
+						EXTENT_DAMAGED, GFP_NOFS);
+	}
+
+out:
+	if (page)
+		put_page(page);
+	if (inode)
+		iput(inode);
+
+	if (ret < 0)
+		return ret;
+
+	if (ret == 0 && corrected) {
+		/*
+		 * we only need to call readpage for one of the inodes belonging
+		 * to this extent. so make iterate_extent_inodes stop
+		 */
+		return 1;
+	}
+
+	return -EIO;
+}
+
+static void scrub_fixup_nodatasum(struct btrfs_work *work)
+{
+	int ret;
+	struct scrub_fixup_nodatasum *fixup;
+	struct scrub_dev *sdev;
+	struct btrfs_trans_handle *trans = NULL;
+	struct btrfs_fs_info *fs_info;
+	struct btrfs_path *path;
+	int uncorrectable = 0;
+
+	fixup = container_of(work, struct scrub_fixup_nodatasum, work);
+	sdev = fixup->sdev;
+	fs_info = fixup->root->fs_info;
+
+	path = btrfs_alloc_path();
+	if (!path) {
+		spin_lock(&sdev->stat_lock);
+		++sdev->stat.malloc_errors;
+		spin_unlock(&sdev->stat_lock);
+		uncorrectable = 1;
+		goto out;
+	}
+
+	trans = btrfs_join_transaction(fixup->root);
+	if (IS_ERR(trans)) {
+		uncorrectable = 1;
+		goto out;
+	}
+
+	/*
+	 * the idea is to trigger a regular read through the standard path. we
+	 * read a page from the (failed) logical address by specifying the
+	 * corresponding copynum of the failed sector. thus, that readpage is
+	 * expected to fail.
+	 * that is the point where on-the-fly error correction will kick in
+	 * (once it's finished) and rewrite the failed sector if a good copy
+	 * can be found.
+	 */
+	ret = iterate_inodes_from_logical(fixup->logical, fixup->root->fs_info,
+						path, scrub_fixup_readpage,
+						fixup);
+	if (ret < 0) {
+		uncorrectable = 1;
+		goto out;
+	}
+	WARN_ON(ret != 1);
+
+	spin_lock(&sdev->stat_lock);
+	++sdev->stat.corrected_errors;
+	spin_unlock(&sdev->stat_lock);
+
+out:
+	if (trans && !IS_ERR(trans))
+		btrfs_end_transaction(trans, fixup->root);
+	if (uncorrectable) {
+		spin_lock(&sdev->stat_lock);
+		++sdev->stat.uncorrectable_errors;
+		spin_unlock(&sdev->stat_lock);
+		printk_ratelimited(KERN_ERR "btrfs: unable to fixup "
+					"(nodatasum) error at logical %llu\n",
+					fixup->logical);
+	}
+
+	btrfs_free_path(path);
+	kfree(fixup);
+
+	/* see caller why we're pretending to be paused in the scrub counters */
+	mutex_lock(&fs_info->scrub_lock);
+	atomic_dec(&fs_info->scrubs_running);
+	atomic_dec(&fs_info->scrubs_paused);
+	mutex_unlock(&fs_info->scrub_lock);
+	atomic_dec(&sdev->fixup_cnt);
+	wake_up(&fs_info->scrub_pause_wait);
+	wake_up(&sdev->list_wait);
+}
+
 /*
  * scrub_recheck_error gets called when either verification of the page
  * failed or the bio failed to read, e.g. with EIO. In the latter case,
  * recheck_error gets called for every page in the bio, even though only
  * one may be bad
  */
-static void scrub_recheck_error(struct scrub_bio *sbio, int ix)
+static int scrub_recheck_error(struct scrub_bio *sbio, int ix)
 {
+	struct scrub_dev *sdev = sbio->sdev;
+	u64 sector = (sbio->physical + ix * PAGE_SIZE) >> 9;
+	static DEFINE_RATELIMIT_STATE(_rs, DEFAULT_RATELIMIT_INTERVAL,
+					DEFAULT_RATELIMIT_BURST);
+
 	if (sbio->err) {
-		if (scrub_fixup_io(READ, sbio->sdev->dev->bdev,
-				   (sbio->physical + ix * PAGE_SIZE) >> 9,
+		if (scrub_fixup_io(READ, sbio->sdev->dev->bdev, sector,
 				   sbio->bio->bi_io_vec[ix].bv_page) == 0) {
 			if (scrub_fixup_check(sbio, ix) == 0)
-				return;
+				return 0;
 		}
+		if (__ratelimit(&_rs))
+			scrub_print_warning("i/o error", sbio, ix);
+	} else {
+		if (__ratelimit(&_rs))
+			scrub_print_warning("checksum error", sbio, ix);
 	}
 
+	spin_lock(&sdev->stat_lock);
+	++sdev->stat.read_errors;
+	spin_unlock(&sdev->stat_lock);
+
 	scrub_fixup(sbio, ix);
+	return 1;
 }
 
 static int scrub_fixup_check(struct scrub_bio *sbio, int ix)
@@ -250,7 +611,8 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix)
 	struct scrub_dev *sdev = sbio->sdev;
 	struct btrfs_fs_info *fs_info = sdev->dev->dev_root->fs_info;
 	struct btrfs_mapping_tree *map_tree = &fs_info->mapping_tree;
-	struct btrfs_multi_bio *multi = NULL;
+	struct btrfs_bio *bbio = NULL;
+	struct scrub_fixup_nodatasum *fixup;
 	u64 logical = sbio->logical + ix * PAGE_SIZE;
 	u64 length;
 	int i;
@@ -259,38 +621,57 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix)
 
 	if ((sbio->spag[ix].flags & BTRFS_EXTENT_FLAG_DATA) &&
 	    (sbio->spag[ix].have_csum == 0)) {
+		fixup = kzalloc(sizeof(*fixup), GFP_NOFS);
+		if (!fixup)
+			goto uncorrectable;
+		fixup->sdev = sdev;
+		fixup->logical = logical;
+		fixup->root = fs_info->extent_root;
+		fixup->mirror_num = sbio->spag[ix].mirror_num;
 		/*
-		 * nodatasum, don't try to fix anything
-		 * FIXME: we can do better, open the inode and trigger a
-		 * writeback
+		 * increment scrubs_running to prevent cancel requests from
+		 * completing as long as a fixup worker is running. we must also
+		 * increment scrubs_paused to prevent deadlocking on pause
+		 * requests used for transactions commits (as the worker uses a
+		 * transaction context). it is safe to regard the fixup worker
+		 * as paused for all matters practical. effectively, we only
+		 * avoid cancellation requests from completing.
 		 */
-		goto uncorrectable;
+		mutex_lock(&fs_info->scrub_lock);
+		atomic_inc(&fs_info->scrubs_running);
+		atomic_inc(&fs_info->scrubs_paused);
+		mutex_unlock(&fs_info->scrub_lock);
+		atomic_inc(&sdev->fixup_cnt);
+		fixup->work.func = scrub_fixup_nodatasum;
+		btrfs_queue_worker(&fs_info->scrub_workers, &fixup->work);
+		return;
 	}
 
 	length = PAGE_SIZE;
 	ret = btrfs_map_block(map_tree, REQ_WRITE, logical, &length,
-			      &multi, 0);
-	if (ret || !multi || length < PAGE_SIZE) {
+			      &bbio, 0);
+	if (ret || !bbio || length < PAGE_SIZE) {
 		printk(KERN_ERR
 		       "scrub_fixup: btrfs_map_block failed us for %llu\n",
 		       (unsigned long long)logical);
 		WARN_ON(1);
+		kfree(bbio);
 		return;
 	}
 
-	if (multi->num_stripes == 1)
+	if (bbio->num_stripes == 1)
 		/* there aren't any replicas */
 		goto uncorrectable;
 
 	/*
 	 * first find a good copy
 	 */
-	for (i = 0; i < multi->num_stripes; ++i) {
-		if (i == sbio->spag[ix].mirror_num)
+	for (i = 0; i < bbio->num_stripes; ++i) {
+		if (i + 1 == sbio->spag[ix].mirror_num)
 			continue;
 
-		if (scrub_fixup_io(READ, multi->stripes[i].dev->bdev,
-				   multi->stripes[i].physical >> 9,
+		if (scrub_fixup_io(READ, bbio->stripes[i].dev->bdev,
+				   bbio->stripes[i].physical >> 9,
 				   sbio->bio->bi_io_vec[ix].bv_page)) {
 			/* I/O-error, this is not a good copy */
 			continue;
@@ -299,7 +680,7 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix)
 		if (scrub_fixup_check(sbio, ix) == 0)
 			break;
 	}
-	if (i == multi->num_stripes)
+	if (i == bbio->num_stripes)
 		goto uncorrectable;
 
 	if (!sdev->readonly) {
@@ -314,25 +695,23 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix)
 		}
 	}
 
-	kfree(multi);
+	kfree(bbio);
 	spin_lock(&sdev->stat_lock);
 	++sdev->stat.corrected_errors;
 	spin_unlock(&sdev->stat_lock);
 
-	if (printk_ratelimit())
-		printk(KERN_ERR "btrfs: fixed up at %llu\n",
-		       (unsigned long long)logical);
+	printk_ratelimited(KERN_ERR "btrfs: fixed up error at logical %llu\n",
+			       (unsigned long long)logical);
 	return;
 
 uncorrectable:
-	kfree(multi);
+	kfree(bbio);
 	spin_lock(&sdev->stat_lock);
 	++sdev->stat.uncorrectable_errors;
 	spin_unlock(&sdev->stat_lock);
 
-	if (printk_ratelimit())
-		printk(KERN_ERR "btrfs: unable to fixup at %llu\n",
-			 (unsigned long long)logical);
+	printk_ratelimited(KERN_ERR "btrfs: unable to fixup (regular) error at "
+				"logical %llu\n", (unsigned long long)logical);
 }
 
 static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector,
@@ -382,8 +761,14 @@ static void scrub_checksum(struct btrfs_work *work)
 	int ret;
 
 	if (sbio->err) {
+		ret = 0;
 		for (i = 0; i < sbio->count; ++i)
-			scrub_recheck_error(sbio, i);
+			ret |= scrub_recheck_error(sbio, i);
+		if (!ret) {
+			spin_lock(&sdev->stat_lock);
+			++sdev->stat.unverified_errors;
+			spin_unlock(&sdev->stat_lock);
+		}
 
 		sbio->bio->bi_flags &= ~(BIO_POOL_MASK - 1);
 		sbio->bio->bi_flags |= 1 << BIO_UPTODATE;
@@ -396,10 +781,6 @@ static void scrub_checksum(struct btrfs_work *work)
 			bi->bv_offset = 0;
 			bi->bv_len = PAGE_SIZE;
 		}
-
-		spin_lock(&sdev->stat_lock);
-		++sdev->stat.read_errors;
-		spin_unlock(&sdev->stat_lock);
 		goto out;
 	}
 	for (i = 0; i < sbio->count; ++i) {
@@ -420,8 +801,14 @@ static void scrub_checksum(struct btrfs_work *work)
 			WARN_ON(1);
 		}
 		kunmap_atomic(buffer, KM_USER0);
-		if (ret)
-			scrub_recheck_error(sbio, i);
+		if (ret) {
+			ret = scrub_recheck_error(sbio, i);
+			if (!ret) {
+				spin_lock(&sdev->stat_lock);
+				++sdev->stat.unverified_errors;
+				spin_unlock(&sdev->stat_lock);
+			}
+		}
 	}
 
 out:
@@ -604,7 +991,7 @@ nomem:
 }
 
 static int scrub_page(struct scrub_dev *sdev, u64 logical, u64 len,
-		      u64 physical, u64 flags, u64 gen, u64 mirror_num,
+		      u64 physical, u64 flags, u64 gen, int mirror_num,
 		      u8 *csum, int force)
 {
 	struct scrub_bio *sbio;
@@ -701,7 +1088,7 @@ static int scrub_find_csum(struct scrub_dev *sdev, u64 logical, u64 len,
 
 /* scrub extent tries to collect up to 64 kB for each bio */
 static int scrub_extent(struct scrub_dev *sdev, u64 logical, u64 len,
-			u64 physical, u64 flags, u64 gen, u64 mirror_num)
+			u64 physical, u64 flags, u64 gen, int mirror_num)
 {
 	int ret;
 	u8 csum[BTRFS_CSUM_SIZE];
@@ -741,13 +1128,16 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev,
 	int slot;
 	int i;
 	u64 nstripes;
-	int start_stripe;
 	struct extent_buffer *l;
 	struct btrfs_key key;
 	u64 physical;
 	u64 logical;
 	u64 generation;
-	u64 mirror_num;
+	int mirror_num;
+	struct reada_control *reada1;
+	struct reada_control *reada2;
+	struct btrfs_key key_start;
+	struct btrfs_key key_end;
 
 	u64 increment = map->stripe_len;
 	u64 offset;
@@ -758,102 +1148,88 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev,
 	if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
 		offset = map->stripe_len * num;
 		increment = map->stripe_len * map->num_stripes;
-		mirror_num = 0;
+		mirror_num = 1;
 	} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
 		int factor = map->num_stripes / map->sub_stripes;
 		offset = map->stripe_len * (num / map->sub_stripes);
 		increment = map->stripe_len * factor;
-		mirror_num = num % map->sub_stripes;
+		mirror_num = num % map->sub_stripes + 1;
 	} else if (map->type & BTRFS_BLOCK_GROUP_RAID1) {
 		increment = map->stripe_len;
-		mirror_num = num % map->num_stripes;
+		mirror_num = num % map->num_stripes + 1;
 	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
 		increment = map->stripe_len;
-		mirror_num = num % map->num_stripes;
+		mirror_num = num % map->num_stripes + 1;
 	} else {
 		increment = map->stripe_len;
-		mirror_num = 0;
+		mirror_num = 1;
 	}
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	path->reada = 2;
 	path->search_commit_root = 1;
 	path->skip_locking = 1;
 
 	/*
-	 * find all extents for each stripe and just read them to get
-	 * them into the page cache
-	 * FIXME: we can do better. build a more intelligent prefetching
+	 * trigger the readahead for extent tree csum tree and wait for
+	 * completion. During readahead, the scrub is officially paused
+	 * to not hold off transaction commits
 	 */
 	logical = base + offset;
-	physical = map->stripes[num].physical;
-	ret = 0;
-	for (i = 0; i < nstripes; ++i) {
-		key.objectid = logical;
-		key.type = BTRFS_EXTENT_ITEM_KEY;
-		key.offset = (u64)0;
-
-		ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
-		if (ret < 0)
-			goto out_noplug;
-
-		/*
-		 * we might miss half an extent here, but that doesn't matter,
-		 * as it's only the prefetch
-		 */
-		while (1) {
-			l = path->nodes[0];
-			slot = path->slots[0];
-			if (slot >= btrfs_header_nritems(l)) {
-				ret = btrfs_next_leaf(root, path);
-				if (ret == 0)
-					continue;
-				if (ret < 0)
-					goto out_noplug;
 
-				break;
-			}
-			btrfs_item_key_to_cpu(l, &key, slot);
+	wait_event(sdev->list_wait,
+		   atomic_read(&sdev->in_flight) == 0);
+	atomic_inc(&fs_info->scrubs_paused);
+	wake_up(&fs_info->scrub_pause_wait);
 
-			if (key.objectid >= logical + map->stripe_len)
-				break;
+	/* FIXME it might be better to start readahead at commit root */
+	key_start.objectid = logical;
+	key_start.type = BTRFS_EXTENT_ITEM_KEY;
+	key_start.offset = (u64)0;
+	key_end.objectid = base + offset + nstripes * increment;
+	key_end.type = BTRFS_EXTENT_ITEM_KEY;
+	key_end.offset = (u64)0;
+	reada1 = btrfs_reada_add(root, &key_start, &key_end);
+
+	key_start.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
+	key_start.type = BTRFS_EXTENT_CSUM_KEY;
+	key_start.offset = logical;
+	key_end.objectid = BTRFS_EXTENT_CSUM_OBJECTID;
+	key_end.type = BTRFS_EXTENT_CSUM_KEY;
+	key_end.offset = base + offset + nstripes * increment;
+	reada2 = btrfs_reada_add(csum_root, &key_start, &key_end);
+
+	if (!IS_ERR(reada1))
+		btrfs_reada_wait(reada1);
+	if (!IS_ERR(reada2))
+		btrfs_reada_wait(reada2);
 
-			path->slots[0]++;
-		}
-		btrfs_release_path(path);
-		logical += increment;
-		physical += map->stripe_len;
-		cond_resched();
+	mutex_lock(&fs_info->scrub_lock);
+	while (atomic_read(&fs_info->scrub_pause_req)) {
+		mutex_unlock(&fs_info->scrub_lock);
+		wait_event(fs_info->scrub_pause_wait,
+		   atomic_read(&fs_info->scrub_pause_req) == 0);
+		mutex_lock(&fs_info->scrub_lock);
 	}
+	atomic_dec(&fs_info->scrubs_paused);
+	mutex_unlock(&fs_info->scrub_lock);
+	wake_up(&fs_info->scrub_pause_wait);
 
 	/*
 	 * collect all data csums for the stripe to avoid seeking during
 	 * the scrub. This might currently (crc32) end up to be about 1MB
 	 */
-	start_stripe = 0;
 	blk_start_plug(&plug);
-again:
-	logical = base + offset + start_stripe * increment;
-	for (i = start_stripe; i < nstripes; ++i) {
-		ret = btrfs_lookup_csums_range(csum_root, logical,
-					       logical + map->stripe_len - 1,
-					       &sdev->csum_list, 1);
-		if (ret)
-			goto out;
 
-		logical += increment;
-		cond_resched();
-	}
 	/*
 	 * now find all extents for each stripe and scrub them
 	 */
-	logical = base + offset + start_stripe * increment;
-	physical = map->stripes[num].physical + start_stripe * map->stripe_len;
+	logical = base + offset;
+	physical = map->stripes[num].physical;
 	ret = 0;
-	for (i = start_stripe; i < nstripes; ++i) {
+	for (i = 0; i < nstripes; ++i) {
 		/*
 		 * canceled?
 		 */
@@ -882,11 +1258,14 @@ again:
 			atomic_dec(&fs_info->scrubs_paused);
 			mutex_unlock(&fs_info->scrub_lock);
 			wake_up(&fs_info->scrub_pause_wait);
-			scrub_free_csums(sdev);
-			start_stripe = i;
-			goto again;
 		}
 
+		ret = btrfs_lookup_csums_range(csum_root, logical,
+					       logical + map->stripe_len - 1,
+					       &sdev->csum_list, 1);
+		if (ret)
+			goto out;
+
 		key.objectid = logical;
 		key.type = BTRFS_EXTENT_ITEM_KEY;
 		key.offset = (u64)0;
@@ -982,7 +1361,6 @@ next:
 
 out:
 	blk_finish_plug(&plug);
-out_noplug:
 	btrfs_free_path(path);
 	return ret < 0 ? ret : 0;
 }
@@ -1253,10 +1631,11 @@ int btrfs_scrub_dev(struct btrfs_root *root, u64 devid, u64 start, u64 end,
 		ret = scrub_enumerate_chunks(sdev, start, end);
 
 	wait_event(sdev->list_wait, atomic_read(&sdev->in_flight) == 0);
-
 	atomic_dec(&fs_info->scrubs_running);
 	wake_up(&fs_info->scrub_pause_wait);
 
+	wait_event(sdev->list_wait, atomic_read(&sdev->fixup_cnt) == 0);
+
 	if (progress)
 		memcpy(progress, &sdev->stat, sizeof(*progress));
 

fs/btrfs/super.c

@@ -40,6 +40,7 @@
 #include <linux/magic.h>
 #include <linux/slab.h>
 #include <linux/cleancache.h>
+#include <linux/mnt_namespace.h>
 #include "compat.h"
 #include "delayed-inode.h"
 #include "ctree.h"
@@ -58,6 +59,7 @@
 #include <trace/events/btrfs.h>
 
 static const struct super_operations btrfs_super_ops;
+static struct file_system_type btrfs_fs_type;
 
 static const char *btrfs_decode_error(struct btrfs_fs_info *fs_info, int errno,
 				      char nbuf[16])
@@ -162,7 +164,7 @@ enum {
 	Opt_notreelog, Opt_ratio, Opt_flushoncommit, Opt_discard,
 	Opt_space_cache, Opt_clear_cache, Opt_user_subvol_rm_allowed,
 	Opt_enospc_debug, Opt_subvolrootid, Opt_defrag,
-	Opt_inode_cache, Opt_err,
+	Opt_inode_cache, Opt_no_space_cache, Opt_recovery, Opt_err,
 };
 
 static match_table_t tokens = {
@@ -195,6 +197,8 @@ static match_table_t tokens = {
 	{Opt_subvolrootid, "subvolrootid=%d"},
 	{Opt_defrag, "autodefrag"},
 	{Opt_inode_cache, "inode_cache"},
+	{Opt_no_space_cache, "no_space_cache"},
+	{Opt_recovery, "recovery"},
 	{Opt_err, NULL},
 };
 
@@ -206,14 +210,19 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 {
 	struct btrfs_fs_info *info = root->fs_info;
 	substring_t args[MAX_OPT_ARGS];
-	char *p, *num, *orig;
+	char *p, *num, *orig = NULL;
+	u64 cache_gen;
 	int intarg;
 	int ret = 0;
 	char *compress_type;
 	bool compress_force = false;
 
+	cache_gen = btrfs_super_cache_generation(root->fs_info->super_copy);
+	if (cache_gen)
+		btrfs_set_opt(info->mount_opt, SPACE_CACHE);
+
 	if (!options)
-		return 0;
+		goto out;
 
 	/*
 	 * strsep changes the string, duplicate it because parse_options
@@ -360,9 +369,12 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			btrfs_set_opt(info->mount_opt, DISCARD);
 			break;
 		case Opt_space_cache:
-			printk(KERN_INFO "btrfs: enabling disk space caching\n");
 			btrfs_set_opt(info->mount_opt, SPACE_CACHE);
 			break;
+		case Opt_no_space_cache:
+			printk(KERN_INFO "btrfs: disabling disk space caching\n");
+			btrfs_clear_opt(info->mount_opt, SPACE_CACHE);
+			break;
 		case Opt_inode_cache:
 			printk(KERN_INFO "btrfs: enabling inode map caching\n");
 			btrfs_set_opt(info->mount_opt, INODE_MAP_CACHE);
@@ -381,6 +393,10 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 			printk(KERN_INFO "btrfs: enabling auto defrag");
 			btrfs_set_opt(info->mount_opt, AUTO_DEFRAG);
 			break;
+		case Opt_recovery:
+			printk(KERN_INFO "btrfs: enabling auto recovery");
+			btrfs_set_opt(info->mount_opt, RECOVERY);
+			break;
 		case Opt_err:
 			printk(KERN_INFO "btrfs: unrecognized mount option "
 			       "'%s'\n", p);
@@ -391,6 +407,8 @@ int btrfs_parse_options(struct btrfs_root *root, char *options)
 		}
 	}
 out:
+	if (!ret && btrfs_test_opt(root, SPACE_CACHE))
+		printk(KERN_INFO "btrfs: disk space caching is enabled\n");
 	kfree(orig);
 	return ret;
 }
@@ -406,12 +424,12 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags,
 		u64 *subvol_rootid, struct btrfs_fs_devices **fs_devices)
 {
 	substring_t args[MAX_OPT_ARGS];
-	char *opts, *orig, *p;
+	char *device_name, *opts, *orig, *p;
 	int error = 0;
 	int intarg;
 
 	if (!options)
-		goto out;
+		return 0;
 
 	/*
 	 * strsep changes the string, duplicate it because parse_options
@@ -457,29 +475,24 @@ static int btrfs_parse_early_options(const char *options, fmode_t flags,
 			}
 			break;
 		case Opt_device:
-			error = btrfs_scan_one_device(match_strdup(&args[0]),
+			device_name = match_strdup(&args[0]);
+			if (!device_name) {
+				error = -ENOMEM;
+				goto out;
+			}
+			error = btrfs_scan_one_device(device_name,
 					flags, holder, fs_devices);
+			kfree(device_name);
 			if (error)
-				goto out_free_opts;
+				goto out;
 			break;
 		default:
 			break;
 		}
 	}
 
- out_free_opts:
+out:
 	kfree(orig);
- out:
-	/*
-	 * If no subvolume name is specified we use the default one.  Allocate
-	 * a copy of the string "." here so that code later in the
-	 * mount path doesn't care if it's the default volume or another one.
-	 */
-	if (!*subvol_name) {
-		*subvol_name = kstrdup(".", GFP_KERNEL);
-		if (!*subvol_name)
-			return -ENOMEM;
-	}
 	return error;
 }
 
@@ -492,7 +505,6 @@ static struct dentry *get_default_root(struct super_block *sb,
 	struct btrfs_path *path;
 	struct btrfs_key location;
 	struct inode *inode;
-	struct dentry *dentry;
 	u64 dir_id;
 	int new = 0;
 
@@ -517,7 +529,7 @@ static struct dentry *get_default_root(struct super_block *sb,
 	 * will mount by default if we haven't been given a specific subvolume
 	 * to mount.
 	 */
-	dir_id = btrfs_super_root_dir(&root->fs_info->super_copy);
+	dir_id = btrfs_super_root_dir(root->fs_info->super_copy);
 	di = btrfs_lookup_dir_item(NULL, root, path, dir_id, "default", 7, 0);
 	if (IS_ERR(di)) {
 		btrfs_free_path(path);
@@ -566,29 +578,7 @@ setup_root:
 		return dget(sb->s_root);
 	}
 
-	if (new) {
-		const struct qstr name = { .name = "/", .len = 1 };
-
-		/*
-		 * New inode, we need to make the dentry a sibling of s_root so
-		 * everything gets cleaned up properly on unmount.
-		 */
-		dentry = d_alloc(sb->s_root, &name);
-		if (!dentry) {
-			iput(inode);
-			return ERR_PTR(-ENOMEM);
-		}
-		d_splice_alias(inode, dentry);
-	} else {
-		/*
-		 * We found the inode in cache, just find a dentry for it and
-		 * put the reference to the inode we just got.
-		 */
-		dentry = d_find_alias(inode);
-		iput(inode);
-	}
-
-	return dentry;
+	return d_obtain_alias(inode);
 }
 
 static int btrfs_fill_super(struct super_block *sb,
@@ -719,6 +709,8 @@ static int btrfs_show_options(struct seq_file *seq, struct vfsmount *vfs)
 		seq_puts(seq, ",noacl");
 	if (btrfs_test_opt(root, SPACE_CACHE))
 		seq_puts(seq, ",space_cache");
+	else
+		seq_puts(seq, ",no_space_cache");
 	if (btrfs_test_opt(root, CLEAR_CACHE))
 		seq_puts(seq, ",clear_cache");
 	if (btrfs_test_opt(root, USER_SUBVOL_RM_ALLOWED))
@@ -753,6 +745,137 @@ static int btrfs_set_super(struct super_block *s, void *data)
 	return set_anon_super(s, data);
 }
 
+/*
+ * subvolumes are identified by ino 256
+ */
+static inline int is_subvolume_inode(struct inode *inode)
+{
+	if (inode && inode->i_ino == BTRFS_FIRST_FREE_OBJECTID)
+		return 1;
+	return 0;
+}
+
+/*
+ * This will strip out the subvol=%s argument for an argument string and add
+ * subvolid=0 to make sure we get the actual tree root for path walking to the
+ * subvol we want.
+ */
+static char *setup_root_args(char *args)
+{
+	unsigned copied = 0;
+	unsigned len = strlen(args) + 2;
+	char *pos;
+	char *ret;
+
+	/*
+	 * We need the same args as before, but minus
+	 *
+	 * subvol=a
+	 *
+	 * and add
+	 *
+	 * subvolid=0
+	 *
+	 * which is a difference of 2 characters, so we allocate strlen(args) +
+	 * 2 characters.
+	 */
+	ret = kzalloc(len * sizeof(char), GFP_NOFS);
+	if (!ret)
+		return NULL;
+	pos = strstr(args, "subvol=");
+
+	/* This shouldn't happen, but just in case.. */
+	if (!pos) {
+		kfree(ret);
+		return NULL;
+	}
+
+	/*
+	 * The subvol=<> arg is not at the front of the string, copy everybody
+	 * up to that into ret.
+	 */
+	if (pos != args) {
+		*pos = '\0';
+		strcpy(ret, args);
+		copied += strlen(args);
+		pos++;
+	}
+
+	strncpy(ret + copied, "subvolid=0", len - copied);
+
+	/* Length of subvolid=0 */
+	copied += 10;
+
+	/*
+	 * If there is no , after the subvol= option then we know there's no
+	 * other options and we can just return.
+	 */
+	pos = strchr(pos, ',');
+	if (!pos)
+		return ret;
+
+	/* Copy the rest of the arguments into our buffer */
+	strncpy(ret + copied, pos, len - copied);
+	copied += strlen(pos);
+
+	return ret;
+}
+
+static struct dentry *mount_subvol(const char *subvol_name, int flags,
+				   const char *device_name, char *data)
+{
+	struct super_block *s;
+	struct dentry *root;
+	struct vfsmount *mnt;
+	struct mnt_namespace *ns_private;
+	char *newargs;
+	struct path path;
+	int error;
+
+	newargs = setup_root_args(data);
+	if (!newargs)
+		return ERR_PTR(-ENOMEM);
+	mnt = vfs_kern_mount(&btrfs_fs_type, flags, device_name,
+			     newargs);
+	kfree(newargs);
+	if (IS_ERR(mnt))
+		return ERR_CAST(mnt);
+
+	ns_private = create_mnt_ns(mnt);
+	if (IS_ERR(ns_private)) {
+		mntput(mnt);
+		return ERR_CAST(ns_private);
+	}
+
+	/*
+	 * This will trigger the automount of the subvol so we can just
+	 * drop the mnt we have here and return the dentry that we
+	 * found.
+	 */
+	error = vfs_path_lookup(mnt->mnt_root, mnt, subvol_name,
+				LOOKUP_FOLLOW, &path);
+	put_mnt_ns(ns_private);
+	if (error)
+		return ERR_PTR(error);
+
+	if (!is_subvolume_inode(path.dentry->d_inode)) {
+		path_put(&path);
+		mntput(mnt);
+		error = -EINVAL;
+		printk(KERN_ERR "btrfs: '%s' is not a valid subvolume\n",
+				subvol_name);
+		return ERR_PTR(-EINVAL);
+	}
+
+	/* Get a ref to the sb and the dentry we found and return it */
+	s = path.mnt->mnt_sb;
+	atomic_inc(&s->s_active);
+	root = dget(path.dentry);
+	path_put(&path);
+	down_write(&s->s_umount);
+
+	return root;
+}
 
 /*
  * Find a superblock for the given device / mount point.
@@ -784,13 +907,19 @@ static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 	if (error)
 		return ERR_PTR(error);
 
+	if (subvol_name) {
+		root = mount_subvol(subvol_name, flags, device_name, data);
+		kfree(subvol_name);
+		return root;
+	}
+
 	error = btrfs_scan_one_device(device_name, mode, fs_type, &fs_devices);
 	if (error)
-		goto error_free_subvol_name;
+		return ERR_PTR(error);
 
 	error = btrfs_open_devices(fs_devices, mode, fs_type);
 	if (error)
-		goto error_free_subvol_name;
+		return ERR_PTR(error);
 
 	if (!(flags & MS_RDONLY) && fs_devices->rw_devices == 0) {
 		error = -EACCES;
@@ -813,88 +942,57 @@ static struct dentry *btrfs_mount(struct file_system_type *fs_type, int flags,
 	fs_info->fs_devices = fs_devices;
 	tree_root->fs_info = fs_info;
 
+	fs_info->super_copy = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
+	fs_info->super_for_commit = kzalloc(BTRFS_SUPER_INFO_SIZE, GFP_NOFS);
+	if (!fs_info->super_copy || !fs_info->super_for_commit) {
+		error = -ENOMEM;
+		goto error_close_devices;
+	}
+
 	bdev = fs_devices->latest_bdev;
 	s = sget(fs_type, btrfs_test_super, btrfs_set_super, tree_root);
-	if (IS_ERR(s))
-		goto error_s;
+	if (IS_ERR(s)) {
+		error = PTR_ERR(s);
+		goto error_close_devices;
+	}
 
 	if (s->s_root) {
 		if ((flags ^ s->s_flags) & MS_RDONLY) {
 			deactivate_locked_super(s);
-			error = -EBUSY;
-			goto error_close_devices;
+			return ERR_PTR(-EBUSY);
 		}
 
 		btrfs_close_devices(fs_devices);
-		kfree(fs_info);
+		free_fs_info(fs_info);
 		kfree(tree_root);
 	} else {
 		char b[BDEVNAME_SIZE];
 
 		s->s_flags = flags | MS_NOSEC;
 		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
+		btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 		error = btrfs_fill_super(s, fs_devices, data,
 					 flags & MS_SILENT ? 1 : 0);
 		if (error) {
 			deactivate_locked_super(s);
-			goto error_free_subvol_name;
+			return ERR_PTR(error);
 		}
 
-		btrfs_sb(s)->fs_info->bdev_holder = fs_type;
 		s->s_flags |= MS_ACTIVE;
 	}
 
-	/* if they gave us a subvolume name bind mount into that */
-	if (strcmp(subvol_name, ".")) {
-		struct dentry *new_root;
-
-		root = get_default_root(s, subvol_rootid);
-		if (IS_ERR(root)) {
-			error = PTR_ERR(root);
-			deactivate_locked_super(s);
-			goto error_free_subvol_name;
-		}
-
-		mutex_lock(&root->d_inode->i_mutex);
-		new_root = lookup_one_len(subvol_name, root,
-				      strlen(subvol_name));
-		mutex_unlock(&root->d_inode->i_mutex);
-
-		if (IS_ERR(new_root)) {
-			dput(root);
-			deactivate_locked_super(s);
-			error = PTR_ERR(new_root);
-			goto error_free_subvol_name;
-		}
-		if (!new_root->d_inode) {
-			dput(root);
-			dput(new_root);
-			deactivate_locked_super(s);
-			error = -ENXIO;
-			goto error_free_subvol_name;
-		}
-		dput(root);
-		root = new_root;
-	} else {
-		root = get_default_root(s, subvol_objectid);
-		if (IS_ERR(root)) {
-			error = PTR_ERR(root);
-			deactivate_locked_super(s);
-			goto error_free_subvol_name;
-		}
+	root = get_default_root(s, subvol_objectid);
+	if (IS_ERR(root)) {
+		deactivate_locked_super(s);
+		return root;
 	}
 
-	kfree(subvol_name);
 	return root;
 
-error_s:
-	error = PTR_ERR(s);
 error_close_devices:
 	btrfs_close_devices(fs_devices);
-	kfree(fs_info);
+	free_fs_info(fs_info);
 	kfree(tree_root);
-error_free_subvol_name:
-	kfree(subvol_name);
 	return ERR_PTR(error);
 }
 
@@ -919,7 +1017,7 @@ static int btrfs_remount(struct super_block *sb, int *flags, char *data)
 		if (root->fs_info->fs_devices->rw_devices == 0)
 			return -EACCES;
 
-		if (btrfs_super_log_root(&root->fs_info->super_copy) != 0)
+		if (btrfs_super_log_root(root->fs_info->super_copy) != 0)
 			return -EINVAL;
 
 		ret = btrfs_cleanup_fs_roots(root->fs_info);
@@ -1085,7 +1183,7 @@ static int btrfs_calc_avail_data_space(struct btrfs_root *root, u64 *free_bytes)
 static int btrfs_statfs(struct dentry *dentry, struct kstatfs *buf)
 {
 	struct btrfs_root *root = btrfs_sb(dentry->d_sb);
-	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
+	struct btrfs_super_block *disk_super = root->fs_info->super_copy;
 	struct list_head *head = &root->fs_info->space_info;
 	struct btrfs_space_info *found;
 	u64 total_used = 0;

fs/btrfs/transaction.c

@@ -55,6 +55,7 @@ static noinline int join_transaction(struct btrfs_root *root, int nofail)
 	struct btrfs_transaction *cur_trans;
 
 	spin_lock(&root->fs_info->trans_lock);
+loop:
 	if (root->fs_info->trans_no_join) {
 		if (!nofail) {
 			spin_unlock(&root->fs_info->trans_lock);
@@ -75,16 +76,18 @@ static noinline int join_transaction(struct btrfs_root *root, int nofail)
 	cur_trans = kmem_cache_alloc(btrfs_transaction_cachep, GFP_NOFS);
 	if (!cur_trans)
 		return -ENOMEM;
+
 	spin_lock(&root->fs_info->trans_lock);
 	if (root->fs_info->running_transaction) {
+		/*
+		 * someone started a transaction after we unlocked.  Make sure
+		 * to redo the trans_no_join checks above
+		 */
 		kmem_cache_free(btrfs_transaction_cachep, cur_trans);
 		cur_trans = root->fs_info->running_transaction;
-		atomic_inc(&cur_trans->use_count);
-		atomic_inc(&cur_trans->num_writers);
-		cur_trans->num_joined++;
-		spin_unlock(&root->fs_info->trans_lock);
-		return 0;
+		goto loop;
 	}
+
 	atomic_set(&cur_trans->num_writers, 1);
 	cur_trans->num_joined = 0;
 	init_waitqueue_head(&cur_trans->writer_wait);
@@ -275,7 +278,7 @@ static struct btrfs_trans_handle *start_transaction(struct btrfs_root *root,
 	 */
 	if (num_items > 0 && root != root->fs_info->chunk_root) {
 		num_bytes = btrfs_calc_trans_metadata_size(root, num_items);
-		ret = btrfs_block_rsv_add(NULL, root,
+		ret = btrfs_block_rsv_add(root,
 					  &root->fs_info->trans_block_rsv,
 					  num_bytes);
 		if (ret)
@@ -418,8 +421,8 @@ static int should_end_transaction(struct btrfs_trans_handle *trans,
 				  struct btrfs_root *root)
 {
 	int ret;
-	ret = btrfs_block_rsv_check(trans, root,
-				    &root->fs_info->global_block_rsv, 0, 5);
+
+	ret = btrfs_block_rsv_check(root, &root->fs_info->global_block_rsv, 5);
 	return ret ? 1 : 0;
 }
 
@@ -427,17 +430,26 @@ int btrfs_should_end_transaction(struct btrfs_trans_handle *trans,
 				 struct btrfs_root *root)
 {
 	struct btrfs_transaction *cur_trans = trans->transaction;
+	struct btrfs_block_rsv *rsv = trans->block_rsv;
 	int updates;
 
 	smp_mb();
 	if (cur_trans->blocked || cur_trans->delayed_refs.flushing)
 		return 1;
 
+	/*
+	 * We need to do this in case we're deleting csums so the global block
+	 * rsv get's used instead of the csum block rsv.
+	 */
+	trans->block_rsv = NULL;
+
 	updates = trans->delayed_ref_updates;
 	trans->delayed_ref_updates = 0;
 	if (updates)
 		btrfs_run_delayed_refs(trans, root, updates);
 
+	trans->block_rsv = rsv;
+
 	return should_end_transaction(trans, root);
 }
 
@@ -453,6 +465,8 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 		return 0;
 	}
 
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
 	while (count < 4) {
 		unsigned long cur = trans->delayed_ref_updates;
 		trans->delayed_ref_updates = 0;
@@ -473,8 +487,6 @@ static int __btrfs_end_transaction(struct btrfs_trans_handle *trans,
 		count++;
 	}
 
-	btrfs_trans_release_metadata(trans, root);
-
 	if (lock && !atomic_read(&root->fs_info->open_ioctl_trans) &&
 	    should_end_transaction(trans, root)) {
 		trans->transaction->blocked = 1;
@@ -562,50 +574,21 @@ int btrfs_end_transaction_dmeta(struct btrfs_trans_handle *trans,
 int btrfs_write_marked_extents(struct btrfs_root *root,
 			       struct extent_io_tree *dirty_pages, int mark)
 {
-	int ret;
 	int err = 0;
 	int werr = 0;
-	struct page *page;
-	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
 	u64 start = 0;
 	u64 end;
-	unsigned long index;
-
-	while (1) {
-		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
-					    mark);
-		if (ret)
-			break;
-		while (start <= end) {
-			cond_resched();
-
-			index = start >> PAGE_CACHE_SHIFT;
-			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
-			page = find_get_page(btree_inode->i_mapping, index);
-			if (!page)
-				continue;
-
-			btree_lock_page_hook(page);
-			if (!page->mapping) {
-				unlock_page(page);
-				page_cache_release(page);
-				continue;
-			}
 
-			if (PageWriteback(page)) {
-				if (PageDirty(page))
-					wait_on_page_writeback(page);
-				else {
-					unlock_page(page);
-					page_cache_release(page);
-					continue;
-				}
-			}
-			err = write_one_page(page, 0);
-			if (err)
-				werr = err;
-			page_cache_release(page);
-		}
+	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
+				      mark)) {
+		convert_extent_bit(dirty_pages, start, end, EXTENT_NEED_WAIT, mark,
+				   GFP_NOFS);
+		err = filemap_fdatawrite_range(mapping, start, end);
+		if (err)
+			werr = err;
+		cond_resched();
+		start = end + 1;
 	}
 	if (err)
 		werr = err;
@@ -621,39 +604,20 @@ int btrfs_write_marked_extents(struct btrfs_root *root,
 int btrfs_wait_marked_extents(struct btrfs_root *root,
 			      struct extent_io_tree *dirty_pages, int mark)
 {
-	int ret;
 	int err = 0;
 	int werr = 0;
-	struct page *page;
-	struct inode *btree_inode = root->fs_info->btree_inode;
+	struct address_space *mapping = root->fs_info->btree_inode->i_mapping;
 	u64 start = 0;
 	u64 end;
-	unsigned long index;
-
-	while (1) {
-		ret = find_first_extent_bit(dirty_pages, start, &start, &end,
-					    mark);
-		if (ret)
-			break;
 
-		clear_extent_bits(dirty_pages, start, end, mark, GFP_NOFS);
-		while (start <= end) {
-			index = start >> PAGE_CACHE_SHIFT;
-			start = (u64)(index + 1) << PAGE_CACHE_SHIFT;
-			page = find_get_page(btree_inode->i_mapping, index);
-			if (!page)
-				continue;
-			if (PageDirty(page)) {
-				btree_lock_page_hook(page);
-				wait_on_page_writeback(page);
-				err = write_one_page(page, 0);
-				if (err)
-					werr = err;
-			}
-			wait_on_page_writeback(page);
-			page_cache_release(page);
-			cond_resched();
-		}
+	while (!find_first_extent_bit(dirty_pages, start, &start, &end,
+				      EXTENT_NEED_WAIT)) {
+		clear_extent_bits(dirty_pages, start, end, EXTENT_NEED_WAIT, GFP_NOFS);
+		err = filemap_fdatawait_range(mapping, start, end);
+		if (err)
+			werr = err;
+		cond_resched();
+		start = end + 1;
 	}
 	if (err)
 		werr = err;
@@ -673,7 +637,12 @@ int btrfs_write_and_wait_marked_extents(struct btrfs_root *root,
 
 	ret = btrfs_write_marked_extents(root, dirty_pages, mark);
 	ret2 = btrfs_wait_marked_extents(root, dirty_pages, mark);
-	return ret || ret2;
+
+	if (ret)
+		return ret;
+	if (ret2)
+		return ret2;
+	return 0;
 }
 
 int btrfs_write_and_wait_transaction(struct btrfs_trans_handle *trans,
@@ -911,10 +880,9 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	}
 
 	btrfs_reloc_pre_snapshot(trans, pending, &to_reserve);
-	btrfs_orphan_pre_snapshot(trans, pending, &to_reserve);
 
 	if (to_reserve > 0) {
-		ret = btrfs_block_rsv_add(trans, root, &pending->block_rsv,
+		ret = btrfs_block_rsv_add(root, &pending->block_rsv,
 					  to_reserve);
 		if (ret) {
 			pending->error = ret;
@@ -1002,7 +970,6 @@ static noinline int create_pending_snapshot(struct btrfs_trans_handle *trans,
 	BUG_ON(IS_ERR(pending->snap));
 
 	btrfs_reloc_post_snapshot(trans, pending);
-	btrfs_orphan_post_snapshot(trans, pending);
 fail:
 	kfree(new_root_item);
 	trans->block_rsv = rsv;
@@ -1032,7 +999,7 @@ static void update_super_roots(struct btrfs_root *root)
 	struct btrfs_root_item *root_item;
 	struct btrfs_super_block *super;
 
-	super = &root->fs_info->super_copy;
+	super = root->fs_info->super_copy;
 
 	root_item = &root->fs_info->chunk_root->root_item;
 	super->chunk_root = root_item->bytenr;
@@ -1043,7 +1010,7 @@ static void update_super_roots(struct btrfs_root *root)
 	super->root = root_item->bytenr;
 	super->generation = root_item->generation;
 	super->root_level = root_item->level;
-	if (super->cache_generation != 0 || btrfs_test_opt(root, SPACE_CACHE))
+	if (btrfs_test_opt(root, SPACE_CACHE))
 		super->cache_generation = root_item->generation;
 }
 
@@ -1168,14 +1135,15 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 
 	btrfs_run_ordered_operations(root, 0);
 
+	btrfs_trans_release_metadata(trans, root);
+	trans->block_rsv = NULL;
+
 	/* make a pass through all the delayed refs we have so far
 	 * any runnings procs may add more while we are here
 	 */
 	ret = btrfs_run_delayed_refs(trans, root, 0);
 	BUG_ON(ret);
 
-	btrfs_trans_release_metadata(trans, root);
-
 	cur_trans = trans->transaction;
 	/*
 	 * set the flushing flag so procs in this transaction have to
@@ -1341,12 +1309,12 @@ int btrfs_commit_transaction(struct btrfs_trans_handle *trans,
 	update_super_roots(root);
 
 	if (!root->fs_info->log_root_recovering) {
-		btrfs_set_super_log_root(&root->fs_info->super_copy, 0);
-		btrfs_set_super_log_root_level(&root->fs_info->super_copy, 0);
+		btrfs_set_super_log_root(root->fs_info->super_copy, 0);
+		btrfs_set_super_log_root_level(root->fs_info->super_copy, 0);
 	}
 
-	memcpy(&root->fs_info->super_for_commit, &root->fs_info->super_copy,
-	       sizeof(root->fs_info->super_copy));
+	memcpy(root->fs_info->super_for_commit, root->fs_info->super_copy,
+	       sizeof(*root->fs_info->super_copy));
 
 	trans->transaction->blocked = 0;
 	spin_lock(&root->fs_info->trans_lock);

fs/btrfs/tree-log.c

@@ -276,8 +276,9 @@ static int process_one_buffer(struct btrfs_root *log,
 			      struct walk_control *wc, u64 gen)
 {
 	if (wc->pin)
-		btrfs_pin_extent(log->fs_info->extent_root,
-				 eb->start, eb->len, 0);
+		btrfs_pin_extent_for_log_replay(wc->trans,
+						log->fs_info->extent_root,
+						eb->start, eb->len);
 
 	if (btrfs_buffer_uptodate(eb, gen)) {
 		if (wc->write)
@@ -1760,7 +1761,7 @@ static noinline int walk_down_log_tree(struct btrfs_trans_handle *trans,
 
 				WARN_ON(root_owner !=
 					BTRFS_TREE_LOG_OBJECTID);
-				ret = btrfs_free_reserved_extent(root,
+				ret = btrfs_free_and_pin_reserved_extent(root,
 							 bytenr, blocksize);
 				BUG_ON(ret);
 			}
@@ -1828,7 +1829,7 @@ static noinline int walk_up_log_tree(struct btrfs_trans_handle *trans,
 				btrfs_tree_unlock(next);
 
 				WARN_ON(root_owner != BTRFS_TREE_LOG_OBJECTID);
-				ret = btrfs_free_reserved_extent(root,
+				ret = btrfs_free_and_pin_reserved_extent(root,
 						path->nodes[*level]->start,
 						path->nodes[*level]->len);
 				BUG_ON(ret);
@@ -1897,7 +1898,7 @@ static int walk_log_tree(struct btrfs_trans_handle *trans,
 
 			WARN_ON(log->root_key.objectid !=
 				BTRFS_TREE_LOG_OBJECTID);
-			ret = btrfs_free_reserved_extent(log, next->start,
+			ret = btrfs_free_and_pin_reserved_extent(log, next->start,
 							 next->len);
 			BUG_ON(ret);
 		}
@@ -2013,10 +2014,10 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	/* wait for previous tree log sync to complete */
 	if (atomic_read(&root->log_commit[(index1 + 1) % 2]))
 		wait_log_commit(trans, root, root->log_transid - 1);
-
 	while (1) {
 		unsigned long batch = root->log_batch;
-		if (root->log_multiple_pids) {
+		/* when we're on an ssd, just kick the log commit out */
+		if (!btrfs_test_opt(root, SSD) && root->log_multiple_pids) {
 			mutex_unlock(&root->log_mutex);
 			schedule_timeout_uninterruptible(1);
 			mutex_lock(&root->log_mutex);
@@ -2117,9 +2118,9 @@ int btrfs_sync_log(struct btrfs_trans_handle *trans,
 	BUG_ON(ret);
 	btrfs_wait_marked_extents(log, &log->dirty_log_pages, mark);
 
-	btrfs_set_super_log_root(&root->fs_info->super_for_commit,
+	btrfs_set_super_log_root(root->fs_info->super_for_commit,
 				log_root_tree->node->start);
-	btrfs_set_super_log_root_level(&root->fs_info->super_for_commit,
+	btrfs_set_super_log_root_level(root->fs_info->super_for_commit,
 				btrfs_header_level(log_root_tree->node));
 
 	log_root_tree->log_batch = 0;

fs/btrfs/volumes.c

@@ -366,6 +366,14 @@ static noinline int device_list_add(const char *path,
 		}
 		INIT_LIST_HEAD(&device->dev_alloc_list);
 
+		/* init readahead state */
+		spin_lock_init(&device->reada_lock);
+		device->reada_curr_zone = NULL;
+		atomic_set(&device->reada_in_flight, 0);
+		device->reada_next = 0;
+		INIT_RADIX_TREE(&device->reada_zones, GFP_NOFS & ~__GFP_WAIT);
+		INIT_RADIX_TREE(&device->reada_extents, GFP_NOFS & ~__GFP_WAIT);
+
 		mutex_lock(&fs_devices->device_list_mutex);
 		list_add_rcu(&device->dev_list, &fs_devices->devices);
 		mutex_unlock(&fs_devices->device_list_mutex);
@@ -597,10 +605,8 @@ static int __btrfs_open_devices(struct btrfs_fs_devices *fs_devices,
 		set_blocksize(bdev, 4096);
 
 		bh = btrfs_read_dev_super(bdev);
-		if (!bh) {
-			ret = -EINVAL;
+		if (!bh)
 			goto error_close;
-		}
 
 		disk_super = (struct btrfs_super_block *)bh->b_data;
 		devid = btrfs_stack_device_id(&disk_super->dev_item);
@@ -655,7 +661,7 @@ error:
 		continue;
 	}
 	if (fs_devices->open_devices == 0) {
-		ret = -EIO;
+		ret = -EINVAL;
 		goto out;
 	}
 	fs_devices->seeding = seeding;
@@ -1013,8 +1019,13 @@ static int btrfs_free_dev_extent(struct btrfs_trans_handle *trans,
 	}
 	BUG_ON(ret);
 
-	if (device->bytes_used > 0)
-		device->bytes_used -= btrfs_dev_extent_length(leaf, extent);
+	if (device->bytes_used > 0) {
+		u64 len = btrfs_dev_extent_length(leaf, extent);
+		device->bytes_used -= len;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += len;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	ret = btrfs_del_item(trans, root, path);
 
 out:
@@ -1356,6 +1367,11 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 	if (ret)
 		goto error_undo;
 
+	spin_lock(&root->fs_info->free_chunk_lock);
+	root->fs_info->free_chunk_space = device->total_bytes -
+		device->bytes_used;
+	spin_unlock(&root->fs_info->free_chunk_lock);
+
 	device->in_fs_metadata = 0;
 	btrfs_scrub_cancel_dev(root, device);
 
@@ -1387,8 +1403,8 @@ int btrfs_rm_device(struct btrfs_root *root, char *device_path)
 	call_rcu(&device->rcu, free_device);
 	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
-	num_devices = btrfs_super_num_devices(&root->fs_info->super_copy) - 1;
-	btrfs_set_super_num_devices(&root->fs_info->super_copy, num_devices);
+	num_devices = btrfs_super_num_devices(root->fs_info->super_copy) - 1;
+	btrfs_set_super_num_devices(root->fs_info->super_copy, num_devices);
 
 	if (cur_devices->open_devices == 0) {
 		struct btrfs_fs_devices *fs_devices;
@@ -1450,7 +1466,7 @@ static int btrfs_prepare_sprout(struct btrfs_trans_handle *trans,
 	struct btrfs_fs_devices *fs_devices = root->fs_info->fs_devices;
 	struct btrfs_fs_devices *old_devices;
 	struct btrfs_fs_devices *seed_devices;
-	struct btrfs_super_block *disk_super = &root->fs_info->super_copy;
+	struct btrfs_super_block *disk_super = root->fs_info->super_copy;
 	struct btrfs_device *device;
 	u64 super_flags;
 
@@ -1691,15 +1707,19 @@ int btrfs_init_new_device(struct btrfs_root *root, char *device_path)
 		root->fs_info->fs_devices->num_can_discard++;
 	root->fs_info->fs_devices->total_rw_bytes += device->total_bytes;
 
+	spin_lock(&root->fs_info->free_chunk_lock);
+	root->fs_info->free_chunk_space += device->total_bytes;
+	spin_unlock(&root->fs_info->free_chunk_lock);
+
 	if (!blk_queue_nonrot(bdev_get_queue(bdev)))
 		root->fs_info->fs_devices->rotating = 1;
 
-	total_bytes = btrfs_super_total_bytes(&root->fs_info->super_copy);
-	btrfs_set_super_total_bytes(&root->fs_info->super_copy,
+	total_bytes = btrfs_super_total_bytes(root->fs_info->super_copy);
+	btrfs_set_super_total_bytes(root->fs_info->super_copy,
 				    total_bytes + device->total_bytes);
 
-	total_bytes = btrfs_super_num_devices(&root->fs_info->super_copy);
-	btrfs_set_super_num_devices(&root->fs_info->super_copy,
+	total_bytes = btrfs_super_num_devices(root->fs_info->super_copy);
+	btrfs_set_super_num_devices(root->fs_info->super_copy,
 				    total_bytes + 1);
 	mutex_unlock(&root->fs_info->fs_devices->device_list_mutex);
 
@@ -1790,7 +1810,7 @@ static int __btrfs_grow_device(struct btrfs_trans_handle *trans,
 		      struct btrfs_device *device, u64 new_size)
 {
 	struct btrfs_super_block *super_copy =
-		&device->dev_root->fs_info->super_copy;
+		device->dev_root->fs_info->super_copy;
 	u64 old_total = btrfs_super_total_bytes(super_copy);
 	u64 diff = new_size - device->total_bytes;
 
@@ -1849,7 +1869,7 @@ static int btrfs_free_chunk(struct btrfs_trans_handle *trans,
 static int btrfs_del_sys_chunk(struct btrfs_root *root, u64 chunk_objectid, u64
 			chunk_offset)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct btrfs_disk_key *disk_key;
 	struct btrfs_chunk *chunk;
 	u8 *ptr;
@@ -2175,7 +2195,7 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
 	bool retried = false;
 	struct extent_buffer *l;
 	struct btrfs_key key;
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	u64 old_total = btrfs_super_total_bytes(super_copy);
 	u64 old_size = device->total_bytes;
 	u64 diff = device->total_bytes - new_size;
@@ -2192,8 +2212,12 @@ int btrfs_shrink_device(struct btrfs_device *device, u64 new_size)
 	lock_chunks(root);
 
 	device->total_bytes = new_size;
-	if (device->writeable)
+	if (device->writeable) {
 		device->fs_devices->total_rw_bytes -= diff;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space -= diff;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	unlock_chunks(root);
 
 again:
@@ -2257,6 +2281,9 @@ again:
 		device->total_bytes = old_size;
 		if (device->writeable)
 			device->fs_devices->total_rw_bytes += diff;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += diff;
+		spin_unlock(&root->fs_info->free_chunk_lock);
 		unlock_chunks(root);
 		goto done;
 	}
@@ -2292,7 +2319,7 @@ static int btrfs_add_system_chunk(struct btrfs_trans_handle *trans,
 			   struct btrfs_key *key,
 			   struct btrfs_chunk *chunk, int item_size)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct btrfs_disk_key disk_key;
 	u32 array_size;
 	u8 *ptr;
@@ -2615,6 +2642,11 @@ static int __finish_chunk_alloc(struct btrfs_trans_handle *trans,
 		index++;
 	}
 
+	spin_lock(&extent_root->fs_info->free_chunk_lock);
+	extent_root->fs_info->free_chunk_space -= (stripe_size *
+						   map->num_stripes);
+	spin_unlock(&extent_root->fs_info->free_chunk_lock);
+
 	index = 0;
 	stripe = &chunk->stripe;
 	while (index < map->num_stripes) {
@@ -2848,7 +2880,7 @@ static int find_live_mirror(struct map_lookup *map, int first, int num,
 
 static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 			     u64 logical, u64 *length,
-			     struct btrfs_multi_bio **multi_ret,
+			     struct btrfs_bio **bbio_ret,
 			     int mirror_num)
 {
 	struct extent_map *em;
@@ -2866,18 +2898,18 @@ static int __btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 	int i;
 	int num_stripes;
 	int max_errors = 0;
-	struct btrfs_multi_bio *multi = NULL;
+	struct btrfs_bio *bbio = NULL;
 
-	if (multi_ret && !(rw & (REQ_WRITE | REQ_DISCARD)))
+	if (bbio_ret && !(rw & (REQ_WRITE | REQ_DISCARD)))
 		stripes_allocated = 1;
 again:
-	if (multi_ret) {
-		multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
+	if (bbio_ret) {
+		bbio = kzalloc(btrfs_bio_size(stripes_allocated),
 				GFP_NOFS);
-		if (!multi)
+		if (!bbio)
 			return -ENOMEM;
 
-		atomic_set(&multi->error, 0);
+		atomic_set(&bbio->error, 0);
 	}
 
 	read_lock(&em_tree->lock);
@@ -2898,7 +2930,7 @@ again:
 	if (mirror_num > map->num_stripes)
 		mirror_num = 0;
 
-	/* if our multi bio struct is too small, back off and try again */
+	/* if our btrfs_bio struct is too small, back off and try again */
 	if (rw & REQ_WRITE) {
 		if (map->type & (BTRFS_BLOCK_GROUP_RAID1 |
 				 BTRFS_BLOCK_GROUP_DUP)) {
@@ -2917,11 +2949,11 @@ again:
 			stripes_required = map->num_stripes;
 		}
 	}
-	if (multi_ret && (rw & (REQ_WRITE | REQ_DISCARD)) &&
+	if (bbio_ret && (rw & (REQ_WRITE | REQ_DISCARD)) &&
 	    stripes_allocated < stripes_required) {
 		stripes_allocated = map->num_stripes;
 		free_extent_map(em);
-		kfree(multi);
+		kfree(bbio);
 		goto again;
 	}
 	stripe_nr = offset;
@@ -2950,7 +2982,7 @@ again:
 		*length = em->len - offset;
 	}
 
-	if (!multi_ret)
+	if (!bbio_ret)
 		goto out;
 
 	num_stripes = 1;
@@ -2975,13 +3007,17 @@ again:
 			stripe_index = find_live_mirror(map, 0,
 					    map->num_stripes,
 					    current->pid % map->num_stripes);
+			mirror_num = stripe_index + 1;
 		}
 
 	} else if (map->type & BTRFS_BLOCK_GROUP_DUP) {
-		if (rw & (REQ_WRITE | REQ_DISCARD))
+		if (rw & (REQ_WRITE | REQ_DISCARD)) {
 			num_stripes = map->num_stripes;
-		else if (mirror_num)
+		} else if (mirror_num) {
 			stripe_index = mirror_num - 1;
+		} else {
+			mirror_num = 1;
+		}
 
 	} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
 		int factor = map->num_stripes / map->sub_stripes;
@@ -3001,6 +3037,7 @@ again:
 			stripe_index = find_live_mirror(map, stripe_index,
 					      map->sub_stripes, stripe_index +
 					      current->pid % map->sub_stripes);
+			mirror_num = stripe_index + 1;
 		}
 	} else {
 		/*
@@ -3009,15 +3046,16 @@ again:
 		 * stripe_index is the number of our device in the stripe array
 		 */
 		stripe_index = do_div(stripe_nr, map->num_stripes);
+		mirror_num = stripe_index + 1;
 	}
 	BUG_ON(stripe_index >= map->num_stripes);
 
 	if (rw & REQ_DISCARD) {
 		for (i = 0; i < num_stripes; i++) {
-			multi->stripes[i].physical =
+			bbio->stripes[i].physical =
 				map->stripes[stripe_index].physical +
 				stripe_offset + stripe_nr * map->stripe_len;
-			multi->stripes[i].dev = map->stripes[stripe_index].dev;
+			bbio->stripes[i].dev = map->stripes[stripe_index].dev;
 
 			if (map->type & BTRFS_BLOCK_GROUP_RAID0) {
 				u64 stripes;
@@ -3038,16 +3076,16 @@ again:
 				}
 				stripes = stripe_nr_end - 1 - j;
 				do_div(stripes, map->num_stripes);
-				multi->stripes[i].length = map->stripe_len *
+				bbio->stripes[i].length = map->stripe_len *
 					(stripes - stripe_nr + 1);
 
 				if (i == 0) {
-					multi->stripes[i].length -=
+					bbio->stripes[i].length -=
 						stripe_offset;
 					stripe_offset = 0;
 				}
 				if (stripe_index == last_stripe)
-					multi->stripes[i].length -=
+					bbio->stripes[i].length -=
 						stripe_end_offset;
 			} else if (map->type & BTRFS_BLOCK_GROUP_RAID10) {
 				u64 stripes;
@@ -3072,11 +3110,11 @@ again:
 				}
 				stripes = stripe_nr_end - 1 - j;
 				do_div(stripes, factor);
-				multi->stripes[i].length = map->stripe_len *
+				bbio->stripes[i].length = map->stripe_len *
 					(stripes - stripe_nr + 1);
 
 				if (i < map->sub_stripes) {
-					multi->stripes[i].length -=
+					bbio->stripes[i].length -=
 						stripe_offset;
 					if (i == map->sub_stripes - 1)
 						stripe_offset = 0;
@@ -3084,11 +3122,11 @@ again:
 				if (stripe_index >= last_stripe &&
 				    stripe_index <= (last_stripe +
 						     map->sub_stripes - 1)) {
-					multi->stripes[i].length -=
+					bbio->stripes[i].length -=
 						stripe_end_offset;
 				}
 			} else
-				multi->stripes[i].length = *length;
+				bbio->stripes[i].length = *length;
 
 			stripe_index++;
 			if (stripe_index == map->num_stripes) {
@@ -3099,19 +3137,20 @@ again:
 		}
 	} else {
 		for (i = 0; i < num_stripes; i++) {
-			multi->stripes[i].physical =
+			bbio->stripes[i].physical =
 				map->stripes[stripe_index].physical +
 				stripe_offset +
 				stripe_nr * map->stripe_len;
-			multi->stripes[i].dev =
+			bbio->stripes[i].dev =
 				map->stripes[stripe_index].dev;
 			stripe_index++;
 		}
 	}
-	if (multi_ret) {
-		*multi_ret = multi;
-		multi->num_stripes = num_stripes;
-		multi->max_errors = max_errors;
+	if (bbio_ret) {
+		*bbio_ret = bbio;
+		bbio->num_stripes = num_stripes;
+		bbio->max_errors = max_errors;
+		bbio->mirror_num = mirror_num;
 	}
 out:
 	free_extent_map(em);
@@ -3120,9 +3159,9 @@ out:
 
 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 		      u64 logical, u64 *length,
-		      struct btrfs_multi_bio **multi_ret, int mirror_num)
+		      struct btrfs_bio **bbio_ret, int mirror_num)
 {
-	return __btrfs_map_block(map_tree, rw, logical, length, multi_ret,
+	return __btrfs_map_block(map_tree, rw, logical, length, bbio_ret,
 				 mirror_num);
 }
 
@@ -3191,28 +3230,30 @@ int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 	return 0;
 }
 
-static void end_bio_multi_stripe(struct bio *bio, int err)
+static void btrfs_end_bio(struct bio *bio, int err)
 {
-	struct btrfs_multi_bio *multi = bio->bi_private;
+	struct btrfs_bio *bbio = bio->bi_private;
 	int is_orig_bio = 0;
 
 	if (err)
-		atomic_inc(&multi->error);
+		atomic_inc(&bbio->error);
 
-	if (bio == multi->orig_bio)
+	if (bio == bbio->orig_bio)
 		is_orig_bio = 1;
 
-	if (atomic_dec_and_test(&multi->stripes_pending)) {
+	if (atomic_dec_and_test(&bbio->stripes_pending)) {
 		if (!is_orig_bio) {
 			bio_put(bio);
-			bio = multi->orig_bio;
+			bio = bbio->orig_bio;
 		}
-		bio->bi_private = multi->private;
-		bio->bi_end_io = multi->end_io;
+		bio->bi_private = bbio->private;
+		bio->bi_end_io = bbio->end_io;
+		bio->bi_bdev = (struct block_device *)
+					(unsigned long)bbio->mirror_num;
 		/* only send an error to the higher layers if it is
 		 * beyond the tolerance of the multi-bio
 		 */
-		if (atomic_read(&multi->error) > multi->max_errors) {
+		if (atomic_read(&bbio->error) > bbio->max_errors) {
 			err = -EIO;
 		} else if (err) {
 			/*
@@ -3222,7 +3263,7 @@ static void end_bio_multi_stripe(struct bio *bio, int err)
 			set_bit(BIO_UPTODATE, &bio->bi_flags);
 			err = 0;
 		}
-		kfree(multi);
+		kfree(bbio);
 
 		bio_endio(bio, err);
 	} else if (!is_orig_bio) {
@@ -3302,20 +3343,20 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 	u64 logical = (u64)bio->bi_sector << 9;
 	u64 length = 0;
 	u64 map_length;
-	struct btrfs_multi_bio *multi = NULL;
 	int ret;
 	int dev_nr = 0;
 	int total_devs = 1;
+	struct btrfs_bio *bbio = NULL;
 
 	length = bio->bi_size;
 	map_tree = &root->fs_info->mapping_tree;
 	map_length = length;
 
-	ret = btrfs_map_block(map_tree, rw, logical, &map_length, &multi,
+	ret = btrfs_map_block(map_tree, rw, logical, &map_length, &bbio,
 			      mirror_num);
 	BUG_ON(ret);
 
-	total_devs = multi->num_stripes;
+	total_devs = bbio->num_stripes;
 	if (map_length < length) {
 		printk(KERN_CRIT "mapping failed logical %llu bio len %llu "
 		       "len %llu\n", (unsigned long long)logical,
@@ -3323,25 +3364,28 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 		       (unsigned long long)map_length);
 		BUG();
 	}
-	multi->end_io = first_bio->bi_end_io;
-	multi->private = first_bio->bi_private;
-	multi->orig_bio = first_bio;
-	atomic_set(&multi->stripes_pending, multi->num_stripes);
+
+	bbio->orig_bio = first_bio;
+	bbio->private = first_bio->bi_private;
+	bbio->end_io = first_bio->bi_end_io;
+	atomic_set(&bbio->stripes_pending, bbio->num_stripes);
 
 	while (dev_nr < total_devs) {
-		if (total_devs > 1) {
-			if (dev_nr < total_devs - 1) {
-				bio = bio_clone(first_bio, GFP_NOFS);
-				BUG_ON(!bio);
-			} else {
-				bio = first_bio;
-			}
-			bio->bi_private = multi;
-			bio->bi_end_io = end_bio_multi_stripe;
+		if (dev_nr < total_devs - 1) {
+			bio = bio_clone(first_bio, GFP_NOFS);
+			BUG_ON(!bio);
+		} else {
+			bio = first_bio;
 		}
-		bio->bi_sector = multi->stripes[dev_nr].physical >> 9;
-		dev = multi->stripes[dev_nr].dev;
+		bio->bi_private = bbio;
+		bio->bi_end_io = btrfs_end_bio;
+		bio->bi_sector = bbio->stripes[dev_nr].physical >> 9;
+		dev = bbio->stripes[dev_nr].dev;
 		if (dev && dev->bdev && (rw != WRITE || dev->writeable)) {
+			pr_debug("btrfs_map_bio: rw %d, secor=%llu, dev=%lu "
+				 "(%s id %llu), size=%u\n", rw,
+				 (u64)bio->bi_sector, (u_long)dev->bdev->bd_dev,
+				 dev->name, dev->devid, bio->bi_size);
 			bio->bi_bdev = dev->bdev;
 			if (async_submit)
 				schedule_bio(root, dev, rw, bio);
@@ -3354,8 +3398,6 @@ int btrfs_map_bio(struct btrfs_root *root, int rw, struct bio *bio,
 		}
 		dev_nr++;
 	}
-	if (total_devs == 1)
-		kfree(multi);
 	return 0;
 }
 
@@ -3616,15 +3658,20 @@ static int read_one_dev(struct btrfs_root *root,
 	fill_device_from_item(leaf, dev_item, device);
 	device->dev_root = root->fs_info->dev_root;
 	device->in_fs_metadata = 1;
-	if (device->writeable)
+	if (device->writeable) {
 		device->fs_devices->total_rw_bytes += device->total_bytes;
+		spin_lock(&root->fs_info->free_chunk_lock);
+		root->fs_info->free_chunk_space += device->total_bytes -
+			device->bytes_used;
+		spin_unlock(&root->fs_info->free_chunk_lock);
+	}
 	ret = 0;
 	return ret;
 }
 
 int btrfs_read_sys_array(struct btrfs_root *root)
 {
-	struct btrfs_super_block *super_copy = &root->fs_info->super_copy;
+	struct btrfs_super_block *super_copy = root->fs_info->super_copy;
 	struct extent_buffer *sb;
 	struct btrfs_disk_key *disk_key;
 	struct btrfs_chunk *chunk;

fs/btrfs/volumes.h

@@ -92,6 +92,14 @@ struct btrfs_device {
 	struct btrfs_work work;
 	struct rcu_head rcu;
 	struct work_struct rcu_work;
+
+	/* readahead state */
+	spinlock_t reada_lock;
+	atomic_t reada_in_flight;
+	u64 reada_next;
+	struct reada_zone *reada_curr_zone;
+	struct radix_tree_root reada_zones;
+	struct radix_tree_root reada_extents;
 };
 
 struct btrfs_fs_devices {
@@ -136,7 +144,10 @@ struct btrfs_bio_stripe {
 	u64 length; /* only used for discard mappings */
 };
 
-struct btrfs_multi_bio {
+struct btrfs_bio;
+typedef void (btrfs_bio_end_io_t) (struct btrfs_bio *bio, int err);
+
+struct btrfs_bio {
 	atomic_t stripes_pending;
 	bio_end_io_t *end_io;
 	struct bio *orig_bio;
@@ -144,6 +155,7 @@ struct btrfs_multi_bio {
 	atomic_t error;
 	int max_errors;
 	int num_stripes;
+	int mirror_num;
 	struct btrfs_bio_stripe stripes[];
 };
 
@@ -171,7 +183,7 @@ struct map_lookup {
 int btrfs_account_dev_extents_size(struct btrfs_device *device, u64 start,
 				   u64 end, u64 *length);
 
-#define btrfs_multi_bio_size(n) (sizeof(struct btrfs_multi_bio) + \
+#define btrfs_bio_size(n) (sizeof(struct btrfs_bio) + \
 			    (sizeof(struct btrfs_bio_stripe) * (n)))
 
 int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
@@ -180,7 +192,7 @@ int btrfs_alloc_dev_extent(struct btrfs_trans_handle *trans,
 			   u64 chunk_offset, u64 start, u64 num_bytes);
 int btrfs_map_block(struct btrfs_mapping_tree *map_tree, int rw,
 		    u64 logical, u64 *length,
-		    struct btrfs_multi_bio **multi_ret, int mirror_num);
+		    struct btrfs_bio **bbio_ret, int mirror_num);
 int btrfs_rmap_block(struct btrfs_mapping_tree *map_tree,
 		     u64 chunk_start, u64 physical, u64 devid,
 		     u64 **logical, int *naddrs, int *stripe_len);

fs/btrfs/xattr.c

@@ -127,6 +127,17 @@ static int do_setxattr(struct btrfs_trans_handle *trans,
 again:
 	ret = btrfs_insert_xattr_item(trans, root, path, btrfs_ino(inode),
 				      name, name_len, value, size);
+	/*
+	 * If we're setting an xattr to a new value but the new value is say
+	 * exactly BTRFS_MAX_XATTR_SIZE, we could end up with EOVERFLOW getting
+	 * back from split_leaf.  This is because it thinks we'll be extending
+	 * the existing item size, but we're asking for enough space to add the
+	 * item itself.  So if we get EOVERFLOW just set ret to EEXIST and let
+	 * the rest of the function figure it out.
+	 */
+	if (ret == -EOVERFLOW)
+		ret = -EEXIST;
+
 	if (ret == -EEXIST) {
 		if (flags & XATTR_CREATE)
 			goto out;