Merge git://git.jan-o-sch.net/btrfs-unstable into integration

Conflicts:
	fs/btrfs/Makefile
	fs/btrfs/extent_io.c
	fs/btrfs/extent_io.h
	fs/btrfs/scrub.c

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

fs/btrfs/Makefile

@@ -8,6 +8,6 @@ btrfs-y += super.o ctree.o extent-tree.o print-tree.o root-tree.o dir-item.o \
 	   extent_io.o volumes.o async-thread.o ioctl.o locking.o orphan.o \
 	   export.o tree-log.o free-space-cache.o zlib.o lzo.o \
 	   compression.o delayed-ref.o relocation.o delayed-inode.o scrub.o \
-	   reada.o
+	   reada.o backref.o
 
 btrfs-$(CONFIG_BTRFS_FS_POSIX_ACL) += acl.o

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->str + (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->str[i] = 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->str[i].
+ * when it returns, there are ipath->fspath->elem_cnt number of paths available
+ * in ipath->fspath->str[]. 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/disk-io.c

@@ -620,7 +620,7 @@ out:
 
 static int btree_io_failed_hook(struct bio *failed_bio,
 			 struct page *page, u64 start, u64 end,
-			 struct extent_state *state)
+			 u64 mirror_num, struct extent_state *state)
 {
 	struct extent_io_tree *tree;
 	unsigned long len;
@@ -944,7 +944,7 @@ 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)

fs/btrfs/extent-tree.c

@@ -1788,18 +1788,18 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
 {
 	int ret;
 	u64 discarded_bytes = 0;
-	struct btrfs_multi_bio *multi = NULL;
+	struct btrfs_bio *bbio = NULL;
 
 
 	/* Tell the block device(s) that the sectors can be discarded */
 	ret = btrfs_map_block(&root->fs_info->mapping_tree, REQ_DISCARD,
-			      bytenr, &num_bytes, &multi, 0);
+			      bytenr, &num_bytes, &bbio, 0);
 	if (!ret) {
-		struct btrfs_bio_stripe *stripe = multi->stripes;
+		struct btrfs_bio_stripe *stripe = bbio->stripes;
 		int i;
 
 
-		for (i = 0; i < multi->num_stripes; i++, stripe++) {
+		for (i = 0; i < bbio->num_stripes; i++, stripe++) {
 			if (!stripe->dev->can_discard)
 				continue;
 
@@ -1818,7 +1818,7 @@ static int btrfs_discard_extent(struct btrfs_root *root, u64 bytenr,
 			 */
 			ret = 0;
 		}
-		kfree(multi);
+		kfree(bbio);
 	}
 
 	if (actual_bytes)

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;
@@ -1787,6 +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 */
 
 /*
@@ -1885,6 +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) +
@@ -1915,11 +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 &&
+		if (!uptodate) {
-		    tree->ops->readpage_io_failed_hook) {
+			u64 failed_mirror;
-			ret = tree->ops->readpage_io_failed_hook(bio, page,
+			failed_mirror = (u64)bio->bi_bdev;
-							 start, end, state);
+			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);
@@ -1999,6 +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);
@@ -2264,16 +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;
 }
 
@@ -3127,7 +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;
@@ -3152,7 +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);

fs/btrfs/extent_io.h

@@ -18,6 +18,7 @@
 #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)
 
@@ -69,7 +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,
@@ -188,7 +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);
 
@@ -259,6 +260,8 @@ void free_extent_buffer(struct extent_buffer *eb);
 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)
 {
@@ -308,4 +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/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"
@@ -1822,154 +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
+ * if there's a match, we allow the bio to finish.  If not, the code in
- * the io_failure_record routines to find good copies
+ * 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)
@@ -2015,10 +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:
@@ -6273,7 +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)
@@ -7406,7 +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)
@@ -2890,6 +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;
+	unsigned long 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->str[i] - (char *)ipath->fspath->str;
+		ipath->fspath->str[i] = (void *)rel_ptr;
+	}
+
+	ret = copy_to_user(ipa->fspath, 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(loi->inodes, 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)
 {
@@ -2947,6 +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 +193,31 @@ 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 */
+	union {
+		char	*str[0];	/* 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 */
+};
+
+struct btrfs_ioctl_logical_ino_args {
+	__u64				logical;	/* in */
+	__u32				size;		/* in */
+	__u64				reserved[4];
+	struct btrfs_data_container	*inodes;	/* out */
+};
+
 #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 +273,9 @@ 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/reada.c

@@ -247,7 +247,7 @@ int btree_readahead_hook(struct btrfs_root *root, struct extent_buffer *eb,
 
 static struct reada_zone *reada_find_zone(struct btrfs_fs_info *fs_info,
 					  struct btrfs_device *dev, u64 logical,
-					  struct btrfs_multi_bio *multi)
+					  struct btrfs_bio *multi)
 {
 	int ret;
 	int looped = 0;
@@ -327,7 +327,7 @@ static struct reada_extent *reada_find_extent(struct btrfs_root *root,
 	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_multi_bio *multi = NULL;
+	struct btrfs_bio *multi = NULL;
 	struct btrfs_device *dev;
 	u32 blocksize;
 	u64 length;

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
@@ -60,7 +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];
 };
@@ -84,6 +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;
@@ -97,6 +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)) {
@@ -178,6 +204,7 @@ struct scrub_dev *scrub_setup_dev(struct btrfs_device *dev)
 	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);
 	INIT_LIST_HEAD(&sdev->csum_list);
@@ -192,24 +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,
+			ipath->fspath->str[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,
+		if (scrub_fixup_io(READ, sbio->sdev->dev->bdev, sector,
-				   (sbio->physical + ix * PAGE_SIZE) >> 9,
 				   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)
@@ -247,7 +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;
@@ -256,18 +621,36 @@ 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);
-		 * nodatasum, don't try to fix anything
+		if (!fixup)
-		 * FIXME: we can do better, open the inode and trigger a
-		 * writeback
-		 */
 			goto uncorrectable;
+		fixup->sdev = sdev;
+		fixup->logical = logical;
+		fixup->root = fs_info->extent_root;
+		fixup->mirror_num = sbio->spag[ix].mirror_num;
+		/*
+		 * 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.
+		 */
+		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);
+			      &bbio, 0);
-	if (ret || !multi || length < PAGE_SIZE) {
+	if (ret || !bbio || length < PAGE_SIZE) {
 		printk(KERN_ERR
 		       "scrub_fixup: btrfs_map_block failed us for %llu\n",
 		       (unsigned long long)logical);
@@ -275,19 +658,19 @@ static void scrub_fixup(struct scrub_bio *sbio, int ix)
 		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) {
+	for (i = 0; i < bbio->num_stripes; ++i) {
-		if (i == sbio->spag[ix].mirror_num)
+		if (i + 1 == sbio->spag[ix].mirror_num)
 			continue;
 
-		if (scrub_fixup_io(READ, multi->stripes[i].dev->bdev,
+		if (scrub_fixup_io(READ, bbio->stripes[i].dev->bdev,
-				   multi->stripes[i].physical >> 9,
+				   bbio->stripes[i].physical >> 9,
 				   sbio->bio->bi_io_vec[ix].bv_page)) {
 			/* I/O-error, this is not a good copy */
 			continue;
@@ -296,7 +679,7 @@ 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) {
@@ -311,25 +694,23 @@ 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_ratelimited(KERN_ERR "btrfs: fixed up error at logical %llu\n",
-		printk(KERN_ERR "btrfs: fixed up at %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_ratelimited(KERN_ERR "btrfs: unable to fixup (regular) error at "
-		printk(KERN_ERR "btrfs: unable to fixup at %llu\n",
+				"logical %llu\n", (unsigned long long)logical);
-			 (unsigned long long)logical);
 }
 
 static int scrub_fixup_io(int rw, struct block_device *bdev, sector_t sector,
@@ -379,8 +760,14 @@ 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;
@@ -393,10 +780,6 @@ 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) {
@@ -417,8 +800,14 @@ static void scrub_checksum(struct btrfs_work *work)
 			WARN_ON(1);
 		}
 		kunmap_atomic(buffer, KM_USER0);
-		if (ret)
+		if (ret) {
-			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);
+			}
+		}
 	}
 
 out:
@@ -601,7 +990,7 @@ 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;
@@ -698,7 +1087,7 @@ 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];
@@ -743,7 +1132,7 @@ static noinline_for_stack int scrub_stripe(struct scrub_dev *sdev,
 	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;
@@ -758,21 +1147,21 @@ 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();
@@ -1241,10 +1630,11 @@ 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/volumes.c

@@ -2880,7 +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;
@@ -2898,18 +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) {
+	if (bbio_ret) {
-		multi = kzalloc(btrfs_multi_bio_size(stripes_allocated),
+		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);
@@ -2930,7 +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)) {
@@ -2949,11 +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;
@@ -2982,7 +2982,7 @@ again:
 		*length = em->len - offset;
 	}
 
-	if (!multi_ret)
+	if (!bbio_ret)
 		goto out;
 
 	num_stripes = 1;
@@ -3007,13 +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;
@@ -3033,6 +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 {
 		/*
@@ -3041,15 +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;
@@ -3070,16 +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;
@@ -3104,11 +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;
@@ -3116,11 +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) {
@@ -3131,19 +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) {
+	if (bbio_ret) {
-		*multi_ret = multi;
+		*bbio_ret = bbio;
-		multi->num_stripes = num_stripes;
+		bbio->num_stripes = num_stripes;
-		multi->max_errors = max_errors;
+		bbio->max_errors = max_errors;
+		bbio->mirror_num = mirror_num;
 	}
 out:
 	free_extent_map(em);
@@ -3152,9 +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);
 }
 
@@ -3223,28 +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_private = bbio->private;
-		bio->bi_end_io = multi->end_io;
+		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) {
 			/*
@@ -3254,7 +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) {
@@ -3334,20 +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,
@@ -3355,25 +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;
+	bbio->orig_bio = first_bio;
-	multi->orig_bio = first_bio;
+	bbio->private = first_bio->bi_private;
-	atomic_set(&multi->stripes_pending, multi->num_stripes);
+	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_private = bbio;
-			bio->bi_end_io = end_bio_multi_stripe;
+		bio->bi_end_io = btrfs_end_bio;
-		}
+		bio->bi_sector = bbio->stripes[dev_nr].physical >> 9;
-		bio->bi_sector = multi->stripes[dev_nr].physical >> 9;
+		dev = bbio->stripes[dev_nr].dev;
-		dev = multi->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);
@@ -3386,8 +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;
 }
 

fs/btrfs/volumes.h

@@ -144,7 +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;
@@ -152,6 +155,7 @@ struct btrfs_multi_bio {
 	atomic_t error;
 	int max_errors;
 	int num_stripes;
+	int mirror_num;
 	struct btrfs_bio_stripe stripes[];
 };
 
@@ -179,7 +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,
@@ -188,7 +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);