[PATCH 0/6] More generic inode nlink repair function

[PATCH 0/6] More generic inode nlink repair function

[Qu Wenruo]

Update on patch 4 and 6, other is not changed.
This nlink repair function is more generic than the original one.

The old one can only handle a specific case that the inode_ref is
invalid, either point to a non-exist parent inode or point to a invalid
inode(not dir or conflicting index/name).

The new one will reset all the backref, no matter it is valid or not,
and re-add all the valid backref, this make the nlink handles more
corrupt cases.

Qu Wenruo (6):
  btrfs-progs: print root dir verbose error in fsck
  btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.
  btrfs-progs: Import lookup/del_inode_ref() function.
  btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.
  btrfs-progs: Add btrfs_mkdir() function for the incoming 'lost+found' 
       fsck mechanism.
  btrfs-progs: Add fixing function for inodes whose nlink dismatch

 Makefile     |   2 +-
 cmds-check.c | 311 ++++++++++++++++++++++++++++++++++++--
 ctree.c      |   6 +
 ctree.h      |  38 +++++
 inode-item.c | 318 +++++++++++++++++++++++++++++++++++++++
 inode.c      | 484 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 6 files changed, 1148 insertions(+), 11 deletions(-)
 create mode 100644 inode.c

-- 
2.1.3

[PATCH RESEND 1/6] btrfs-progs: print root dir verbose error in fsck

[Qu Wenruo]

Before this patch, when btrfsck found an error in root dir, it will only
output the following message "root %llu root dir %llu error" without any
detailed error.

Just add print_inode_error() to print out the whole error.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 cmds-check.c | 1 +
 1 file changed, 1 insertion(+)

diff --git a/cmds-check.c b/cmds-check.c
index 389674f..9fc1410 100644
--- a/cmds-check.c
+++ b/cmds-check.c
@@ -1984,6 +1984,7 @@ static int check_inode_recs(struct btrfs_root *root,
 			fprintf(stderr, "root %llu root dir %llu error\n",
 				(unsigned long long)root->root_key.objectid,
 				(unsigned long long)root_dirid);
+			print_inode_error(root, rec);
 			error++;
 		}
 	} else {
-- 
2.1.3

[PATCH RESEND 2/6] btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.

[Qu Wenruo]

Import btrfs_insert/del/lookup_extref() functions form kernel for the
incoming btrfs_add_link() and btrfs_unlink() functions.

As the base of incoming btrfs 'lost+found' recovery mechanism.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 ctree.h      |  14 ++++
 inode-item.c | 206 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 220 insertions(+)

diff --git a/ctree.h b/ctree.h
index 89036de..e185e7a 100644
--- a/ctree.h
+++ b/ctree.h
@@ -2389,6 +2389,20 @@ int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 int btrfs_lookup_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 		       *root, struct btrfs_path *path,
 		       struct btrfs_key *location, int mod);
+struct btrfs_inode_extref
+*btrfs_lookup_inode_extref(struct btrfs_trans_handle *trans,
+			   struct btrfs_path *path, struct btrfs_root *root,
+			   u64 ino, u64 parent_ino, u64 index, const char *name,
+			   int namelen, int ins_len);
+int btrfs_del_inode_extref(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   const char *name, int name_len,
+			   u64 inode_objectid, u64 ref_objectid,
+			   u64 *index);
+int btrfs_insert_inode_extref(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
+			      const char *name, int name_len,
+			      u64 inode_objectid, u64 ref_objectid, u64 index);
 
 /* file-item.c */
 int btrfs_del_csums(struct btrfs_trans_handle *trans,
diff --git a/inode-item.c b/inode-item.c
index 8cc98c6..7337ac9 100644
--- a/inode-item.c
+++ b/inode-item.c
@@ -19,6 +19,7 @@
 #include "ctree.h"
 #include "disk-io.h"
 #include "transaction.h"
+#include "crc32c.h"
 
 static int find_name_in_backref(struct btrfs_path *path, const char * name,
 			 int name_len, struct btrfs_inode_ref **ref_ret)
@@ -145,3 +146,208 @@ int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 				sizeof(*inode_item));
 	return ret;
 }
+
+static inline u64 btrfs_extref_hash(u64 parent_ino, const char *name,
+				    int namelen)
+{
+	return (u64) btrfs_crc32c(parent_ino, name, namelen);
+}
+
+static int
+btrfs_find_name_in_ext_backref(struct btrfs_path *path,
+			       u64 parent_ino, const char *name,
+			       int namelen,
+			       struct btrfs_inode_extref **extref_ret)
+{
+	struct extent_buffer *node;
+	struct btrfs_inode_extref *extref;
+	unsigned long ptr;
+	unsigned long name_ptr;
+	u32 item_size;
+	u32 cur_offset = 0;
+	int ref_name_len;
+	int slot;
+
+	node = path->nodes[0];
+	slot = path->slots[0];
+	item_size = btrfs_item_size_nr(node, slot);
+	ptr = btrfs_item_ptr_offset(node, slot);
+
+	/*
+	 * Search all extended backrefs in this item. We're only
+	 * looking through any collisions so most of the time this is
+	 * just going to compare against one buffer. If all is well,
+	 * we'll return success and the inode ref object.
+	 */
+	while (cur_offset < item_size) {
+		extref = (struct btrfs_inode_extref *) (ptr + cur_offset);
+		name_ptr = (unsigned long)(&extref->name);
+		ref_name_len = btrfs_inode_extref_name_len(node, extref);
+
+		if (ref_name_len == namelen &&
+		    btrfs_inode_extref_parent(node, extref) == parent_ino &&
+		    (memcmp_extent_buffer(node, name, name_ptr,
+					  namelen) == 0)) {
+			if (extref_ret)
+				*extref_ret = extref;
+			return 1;
+		}
+
+		cur_offset += ref_name_len + sizeof(*extref);
+	}
+	return 0;
+}
+
+struct btrfs_inode_extref
+*btrfs_lookup_inode_extref(struct btrfs_trans_handle *trans,
+			   struct btrfs_path *path, struct btrfs_root *root,
+			   u64 ino, u64 parent_ino, u64 index, const char *name,
+			   int namelen, int ins_len)
+{
+	struct btrfs_key key;
+	struct btrfs_inode_extref *extref;
+	int ret = 0;
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_EXTREF_KEY;
+	key.offset = btrfs_extref_hash(parent_ino, name, namelen);
+
+	ret = btrfs_search_slot(trans, root, &key, path, ins_len,
+				ins_len ? 1 : 0);
+	if (ret < 0)
+		return ERR_PTR(ret);
+	if (ret > 0)
+		return NULL;
+	if (!btrfs_find_name_in_ext_backref(path, parent_ino, name,
+					    namelen, &extref))
+		return NULL;
+	return extref;
+}
+
+int btrfs_del_inode_extref(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   const char *name, int name_len,
+			   u64 inode_objectid, u64 ref_objectid,
+			   u64 *index)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_extref *extref;
+	struct extent_buffer *leaf;
+	int ret;
+	int del_len = name_len + sizeof(*extref);
+	unsigned long ptr;
+	unsigned long item_start;
+	u32 item_size;
+
+	key.objectid = inode_objectid;
+	key.type = BTRFS_INODE_EXTREF_KEY;
+	key.offset = btrfs_extref_hash(ref_objectid, name, name_len);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret > 0)
+		ret = -ENOENT;
+	if (ret < 0)
+		goto out;
+
+	/*
+	 * Sanity check - did we find the right item for this name?
+	 * This should always succeed so error here will make the FS
+	 * readonly.
+	 */
+	if (!btrfs_find_name_in_ext_backref(path, ref_objectid,
+					    name, name_len, &extref)) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	leaf = path->nodes[0];
+	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+	if (index)
+		*index = btrfs_inode_extref_index(leaf, extref);
+
+	if (del_len == item_size) {
+		/*
+		 * Common case only one ref in the item, remove the
+		 * whole item.
+		 */
+		ret = btrfs_del_item(trans, root, path);
+		goto out;
+	}
+
+	ptr = (unsigned long)extref;
+	item_start = btrfs_item_ptr_offset(leaf, path->slots[0]);
+
+	memmove_extent_buffer(leaf, ptr, ptr + del_len,
+			      item_size - (ptr + del_len - item_start));
+
+	btrfs_truncate_item(trans, root, path, item_size - del_len, 1);
+
+out:
+	btrfs_free_path(path);
+
+	return ret;
+}
+
+/*
+ * btrfs_insert_inode_extref() - Inserts an extended inode ref into a tree.
+ *
+ * The caller must have checked against BTRFS_LINK_MAX already.
+ */
+int btrfs_insert_inode_extref(struct btrfs_trans_handle *trans,
+			      struct btrfs_root *root,
+			      const char *name, int name_len,
+			      u64 inode_objectid, u64 ref_objectid, u64 index)
+{
+	struct btrfs_inode_extref *extref;
+	int ret;
+	int ins_len = name_len + sizeof(*extref);
+	unsigned long ptr;
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct extent_buffer *leaf;
+	struct btrfs_item *item;
+
+	key.objectid = inode_objectid;
+	key.type = BTRFS_INODE_EXTREF_KEY;
+	key.offset = btrfs_extref_hash(ref_objectid, name, name_len);
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key,
+				      ins_len);
+	if (ret == -EEXIST) {
+		if (btrfs_find_name_in_ext_backref(path, ref_objectid,
+						   name, name_len, NULL))
+			goto out;
+
+		btrfs_extend_item(trans, root, path, ins_len);
+		ret = 0;
+	}
+	if (ret < 0)
+		goto out;
+
+	leaf = path->nodes[0];
+	item = btrfs_item_nr(path->slots[0]);
+	ptr = (unsigned long)btrfs_item_ptr(leaf, path->slots[0], char);
+	ptr += btrfs_item_size(leaf, item) - ins_len;
+	extref = (struct btrfs_inode_extref *)ptr;
+
+	btrfs_set_inode_extref_name_len(path->nodes[0], extref, name_len);
+	btrfs_set_inode_extref_index(path->nodes[0], extref, index);
+	btrfs_set_inode_extref_parent(path->nodes[0], extref, ref_objectid);
+
+	ptr = (unsigned long)&extref->name;
+	write_extent_buffer(path->nodes[0], name, ptr, name_len);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
-- 
2.1.3

[PATCH RESEND 3/6] btrfs-progs: Import lookup/del_inode_ref() function.

[Qu Wenruo]

Import lookup/del_inode_ref() function in inode-item.c, as base functions
for the incoming btrfs_add_link() and btrfs_unlink() functions.

Also modify btrfs_insert_inode_ref() and split_leaf() making them able
to deal with EXTENT_IREF incompat flag.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
 ctree.c      |   6 ++++
 ctree.h      |  10 ++++++
 inode-item.c | 112 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 128 insertions(+)

diff --git a/ctree.c b/ctree.c
index 23399e2..bd6cb12 100644
--- a/ctree.c
+++ b/ctree.c
@@ -2015,6 +2015,12 @@ static noinline int split_leaf(struct btrfs_trans_handle *trans,
 	int split;
 	int num_doubles = 0;
 
+	l = path->nodes[0];
+	slot = path->slots[0];
+	if (extend && data_size + btrfs_item_size_nr(l, slot) +
+	    sizeof(struct btrfs_item) > BTRFS_LEAF_DATA_SIZE(root))
+		return -EOVERFLOW;
+
 	/* first try to make some room by pushing left and right */
 	if (data_size && ins_key->type != BTRFS_DIR_ITEM_KEY) {
 		wret = push_leaf_right(trans, root, path, data_size, 0);
diff --git a/ctree.h b/ctree.h
index e185e7a..32b1286 100644
--- a/ctree.h
+++ b/ctree.h
@@ -2403,6 +2403,16 @@ int btrfs_insert_inode_extref(struct btrfs_trans_handle *trans,
 			      struct btrfs_root *root,
 			      const char *name, int name_len,
 			      u64 inode_objectid, u64 ref_objectid, u64 index);
+struct btrfs_inode_ref
+*btrfs_lookup_inode_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root,
+			struct btrfs_path *path,
+			const char *name, int namelen,
+			u64 ino, u64 parent_ino, u64 index, int ins_len);
+int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root,
+			const char *name, int name_len,
+			u64 ino, u64 parent_ino, u64 *index);
 
 /* file-item.c */
 int btrfs_del_csums(struct btrfs_trans_handle *trans,
diff --git a/inode-item.c b/inode-item.c
index 7337ac9..5a9b675 100644
--- a/inode-item.c
+++ b/inode-item.c
@@ -89,6 +89,8 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 		ptr = (unsigned long)(ref + 1);
 		ret = 0;
 	} else if (ret < 0) {
+		if (ret == EOVERFLOW)
+			ret = -EMLINK;
 		goto out;
 	} else {
 		ref = btrfs_item_ptr(path->nodes[0], path->slots[0],
@@ -102,6 +104,15 @@ int btrfs_insert_inode_ref(struct btrfs_trans_handle *trans,
 
 out:
 	btrfs_free_path(path);
+
+	if (ret == -EMLINK) {
+		if (btrfs_fs_incompat(root->fs_info,
+				      BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF))
+			ret = btrfs_insert_inode_extref(trans, root, name,
+							name_len,
+							inode_objectid,
+							ref_objectid, index);
+	}
 	return ret;
 }
 
@@ -147,6 +158,34 @@ int btrfs_insert_inode(struct btrfs_trans_handle *trans, struct btrfs_root
 	return ret;
 }
 
+struct btrfs_inode_ref
+*btrfs_lookup_inode_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root,
+			struct btrfs_path *path,
+			const char *name, int namelen,
+			u64 ino, u64 parent_ino, u64 index, int ins_len)
+{
+	struct btrfs_key key;
+	struct btrfs_inode_ref *ret_inode_ref = NULL;
+	int ret = 0;
+
+	key.objectid = ino;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = parent_ino;
+
+	ret = btrfs_search_slot(trans, root, &key, path, ins_len,
+				ins_len ? 1 : 0);
+	if (ret)
+		goto out;
+
+	find_name_in_backref(path, name, namelen, &ret_inode_ref);
+out:
+	if (ret < 0)
+		return ERR_PTR(ret);
+	else
+		return ret_inode_ref;
+}
+
 static inline u64 btrfs_extref_hash(u64 parent_ino, const char *name,
 				    int namelen)
 {
@@ -351,3 +390,76 @@ out:
 	btrfs_free_path(path);
 	return ret;
 }
+
+int btrfs_del_inode_ref(struct btrfs_trans_handle *trans,
+			struct btrfs_root *root,
+			const char *name, int name_len,
+			u64 ino, u64 parent_ino, u64 *index)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_ref *ref;
+	struct extent_buffer *leaf;
+	unsigned long ptr;
+	unsigned long item_start;
+	u32 item_size;
+	u32 sub_item_len;
+	int ret;
+	int search_ext_refs = 0;
+	int del_len = name_len + sizeof(*ref);
+
+	key.objectid = ino;
+	key.offset = parent_ino;
+	key.type = BTRFS_INODE_REF_KEY;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+	if (ret > 0) {
+		ret = -ENOENT;
+		search_ext_refs = 1;
+		goto out;
+	} else if (ret < 0) {
+		goto out;
+	}
+	if (!find_name_in_backref(path, name, name_len, &ref)) {
+		ret = -ENOENT;
+		search_ext_refs = 1;
+		goto out;
+	}
+	leaf = path->nodes[0];
+	item_size = btrfs_item_size_nr(leaf, path->slots[0]);
+
+	if (index)
+		*index = btrfs_inode_ref_index(leaf, ref);
+
+	if (del_len == item_size) {
+		ret = btrfs_del_item(trans, root, path);
+		goto out;
+	}
+	ptr = (unsigned long)ref;
+	sub_item_len = name_len + sizeof(*ref);
+	item_start = btrfs_item_ptr_offset(leaf, path->slots[0]);
+	memmove_extent_buffer(leaf, ptr, ptr + sub_item_len,
+			      item_size - (ptr + sub_item_len - item_start));
+	btrfs_truncate_item(trans, root, path, item_size - sub_item_len, 1);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+out:
+	btrfs_free_path(path);
+
+	if (search_ext_refs &&
+	    btrfs_fs_incompat(root->fs_info,
+		    BTRFS_FEATURE_INCOMPAT_EXTENDED_IREF)) {
+		/*
+		 * No refs were found, or we could not find the
+		 * name in our ref array. Find and remove the extended
+		 * inode ref then.
+		 */
+		return btrfs_del_inode_extref(trans, root, name, name_len,
+					      ino, parent_ino, index);
+	}
+
+	return ret;
+}
-- 
2.1.3

[PATCH v3 4/6] btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.

[Qu Wenruo]

Add btrfs_unlink() and btrfs_add_link() functions in inode.c,
for the incoming btrfs_mkdir() and later inode operations functions.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
Changlog:
v2:
   Do dir name conflicting check before adding inode_backref or
   dir_item/index.
v3:
   Allow btrfs_add_link() to add dir_index with given index, iff the
   index doesn't conflict with existing one.
   And export the check_dir_conflict() function in inode.c, and fix a
   false alert case in it.
---
 Makefile     |   2 +-
 cmds-check.c |   7 +-
 ctree.h      |  15 ++
 inode.c      | 438 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 4 files changed, 455 insertions(+), 7 deletions(-)
 create mode 100644 inode.c

diff --git a/Makefile b/Makefile
index 4cae30c..d7a5cbe 100644
--- a/Makefile
+++ b/Makefile
@@ -10,7 +10,7 @@ objects = ctree.o disk-io.o radix-tree.o extent-tree.o print-tree.o \
 	  root-tree.o dir-item.o file-item.o inode-item.o inode-map.o \
 	  extent-cache.o extent_io.o volumes.o utils.o repair.o \
 	  qgroup.o raid6.o free-space-cache.o list_sort.o props.o \
-	  ulist.o qgroup-verify.o backref.o
+	  ulist.o qgroup-verify.o backref.o inode.o
 cmds_objects = cmds-subvolume.o cmds-filesystem.o cmds-device.o cmds-scrub.o \
 	       cmds-inspect.o cmds-balance.o cmds-send.o cmds-receive.o \
 	       cmds-quota.o cmds-qgroup.o cmds-replace.o cmds-check.o \
diff --git a/cmds-check.c b/cmds-check.c
index 9fc1410..6419caf 100644
--- a/cmds-check.c
+++ b/cmds-check.c
@@ -1599,14 +1599,9 @@ static int repair_inode_orphan_item(struct btrfs_trans_handle *trans,
 				    struct btrfs_path *path,
 				    struct inode_record *rec)
 {
-	struct btrfs_key key;
 	int ret;
 
-	key.objectid = BTRFS_ORPHAN_OBJECTID;
-	key.type = BTRFS_ORPHAN_ITEM_KEY;
-	key.offset = rec->ino;
-
-	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
+	ret = btrfs_add_orphan_item(trans, root, path, rec->ino);
 	btrfs_release_path(path);
 	if (!ret)
 		rec->errors &= ~I_ERR_NO_ORPHAN_ITEM;
diff --git a/ctree.h b/ctree.h
index 32b1286..21bafd2 100644
--- a/ctree.h
+++ b/ctree.h
@@ -2444,4 +2444,19 @@ static inline int is_fstree(u64 rootid)
 		return 1;
 	return 0;
 }
+
+/* inode.c */
+int check_dir_conflict(struct btrfs_root *root,
+		       char *name, int namelen,
+		       u64 dir, u64 index);
+int btrfs_add_link(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   u64 ino, u64 parent_ino, char *name, int namelen,
+		   u8 type, u64 *index, int add_backref);
+int btrfs_unlink(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		 u64 ino, u64 parent_ino, u64 index, const char *name,
+		 int namelen, int add_orphan);
+int btrfs_add_orphan_item(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  struct btrfs_path *path,
+			  u64 ino);
 #endif
diff --git a/inode.c b/inode.c
new file mode 100644
index 0000000..aeb0a5c
--- /dev/null
+++ b/inode.c
@@ -0,0 +1,438 @@
+/*
+ * Copyright (C) 2014 Fujitsu.  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.
+ */
+
+/*
+ * Unlike inode.c in kernel, which can use most of the kernel infrastructure
+ * like inode/dentry things, in user-land, we can only use inode number to
+ * do directly operation on extent buffer, which may cause extra searching,
+ * but should not be a huge problem since progs is less performence sensitive.
+ */
+#include <sys/stat.h>
+#include "ctree.h"
+#include "transaction.h"
+#include "disk-io.h"
+#include "time.h"
+
+/*
+ * Find a free inode index for later btrfs_add_link().
+ * Currently just search from the largest dir_index and +1.
+ */
+static int btrfs_find_free_dir_index(struct btrfs_root *root, u64 dir_ino,
+				     u64 *ret_ino)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	u64 ret_val = 2;
+	int ret = 0;
+
+	if (!ret_ino)
+		return 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	key.objectid = dir_ino;
+	key.type = BTRFS_DIR_INDEX_KEY;
+	key.offset = (u64)-1;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	ret = 0;
+	if (path->slots[0] == 0) {
+		ret = btrfs_prev_leaf(root, path);
+		if (ret < 0)
+			goto out;
+		if (ret > 0) {
+			/*
+			 * This shouldn't happen since there must be a leaf
+			 * containing the DIR_ITEM.
+			 * Can only happen when the previous leaf is corrupted.
+			 */
+			ret = -EIO;
+			goto out;
+		}
+	} else {
+		path->slots[0]--;
+	}
+	btrfs_item_key_to_cpu(path->nodes[0], &found_key, path->slots[0]);
+	if (found_key.objectid != dir_ino ||
+	    found_key.type != BTRFS_DIR_INDEX_KEY)
+		goto out;
+	ret_val = found_key.offset + 1;
+out:
+	btrfs_free_path(path);
+	if (ret == 0)
+		*ret_ino = ret_val;
+	return ret;
+}
+
+/* Check the dir_item/index conflicts before insert */
+int check_dir_conflict(struct btrfs_root *root,
+		       char *name, int namelen,
+		       u64 dir, u64 index)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_item *inode_item;
+	struct btrfs_dir_item *dir_item;
+	int ret = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	/* Given dir exists? */
+	key.objectid = dir;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+
+	/* Is it a dir? */
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_inode_item);
+	if (!(btrfs_inode_mode(path->nodes[0], inode_item) &
+	      S_IFDIR)) {
+		ret = -ENOTDIR;
+		goto out;
+	}
+	btrfs_release_path(path);
+
+	/* Name conflicting? */
+	dir_item = btrfs_lookup_dir_item(NULL, root, path, dir, name,
+					 namelen, 0);
+	if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		goto out;
+	}
+	if (dir_item) {
+		ret = -EEXIST;
+		goto out;
+	}
+	btrfs_release_path(path);
+
+	/* Index conflicting? */
+	dir_item = btrfs_lookup_dir_index(NULL, root, path, dir, name,
+					  namelen, index, 0);
+	if (IS_ERR(dir_item) && PTR_ERR(dir_item) == -ENOENT)
+		dir_item = NULL;
+	if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		goto out;
+	}
+	if (dir_item) {
+		ret = -EEXIST;
+		goto out;
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
+
+/*
+ * Add dir_item/index for 'parent_ino'
+ * if add_backref is true, also insert a backref from the ino to parent dir
+ * and update the nlink(Kernel version does not do this thing)
+ *
+ * Currently only supports adding link from an inode to another inode.
+ */
+
+int btrfs_add_link(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		   u64 ino, u64 parent_ino, char *name, int namelen,
+		   u8 type, u64 *index, int add_backref)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_item *inode_item;
+	u32 nlink;
+	u64 inode_size;
+	u64 ret_index = 0;
+	int ret = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	if (index && *index) {
+		ret_index = *index;
+	} else {
+		ret = btrfs_find_free_dir_index(root, parent_ino, &ret_index);
+		if (ret < 0)
+			goto out;
+	}
+
+	ret = check_dir_conflict(root, name, namelen, parent_ino, ret_index);
+	if (ret < 0)
+		goto out;
+
+	/* Add inode ref */
+	if (add_backref) {
+		ret = btrfs_insert_inode_ref(trans, root, name, namelen,
+					     ino, parent_ino, ret_index);
+		if (ret < 0)
+			goto out;
+
+		/* Update nlinks for the inode */
+		key.objectid = ino;
+		key.type = BTRFS_INODE_ITEM_KEY;
+		key.offset = 0;
+		ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
+		if (ret) {
+			if (ret > 0)
+				ret = -ENOENT;
+			goto out;
+		}
+		inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_inode_item);
+		nlink = btrfs_inode_nlink(path->nodes[0], inode_item);
+		nlink++;
+		btrfs_set_inode_nlink(path->nodes[0], inode_item, nlink);
+		btrfs_mark_buffer_dirty(path->nodes[0]);
+		btrfs_release_path(path);
+	}
+
+	/* Add dir_item and dir_index */
+	key.objectid = ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_insert_dir_item(trans, root, name, namelen, parent_ino,
+				    &key, type, ret_index);
+	if (ret < 0)
+		goto out;
+
+	/* Update inode size of the parent inode */
+	key.objectid = parent_ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
+	if (ret)
+		goto out;
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_inode_item);
+	inode_size = btrfs_inode_size(path->nodes[0], inode_item);
+	inode_size += namelen * 2;
+	btrfs_set_inode_size(path->nodes[0], inode_item, inode_size);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_release_path(path);
+
+out:
+	btrfs_free_path(path);
+	if (ret == 0 && index)
+		*index = ret_index;
+	return ret;
+}
+
+int btrfs_add_orphan_item(struct btrfs_trans_handle *trans,
+			  struct btrfs_root *root,
+			  struct btrfs_path *path,
+			  u64 ino)
+{
+	struct btrfs_key key;
+	int ret = 0;
+
+	key.objectid = BTRFS_ORPHAN_OBJECTID;
+	key.type = BTRFS_ORPHAN_ITEM_KEY;
+	key.offset = ino;
+
+	ret = btrfs_insert_empty_item(trans, root, path, &key, 0);
+	return ret;
+}
+
+/*
+ * Unlink an inode, which will remove its backref and corresponding dir_index/
+ * dir_item if any of them exists.
+ *
+ * If an inode's nlink is reduced to 0 and 'add_orphan' is true, it will be
+ * added to orphan inode and wairing to be deleted by next kernel mount.
+ */
+int btrfs_unlink(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		 u64 ino, u64 parent_ino, u64 index, const char *name,
+		 int namelen, int add_orphan)
+{
+	struct btrfs_path *path;
+	struct btrfs_key key;
+	struct btrfs_inode_item *inode_item;
+	struct btrfs_inode_ref *inode_ref;
+	struct btrfs_dir_item *dir_item;
+	u64 inode_size;
+	u32 nlinks;
+	int del_inode_ref = 0;
+	int del_dir_item = 0;
+	int del_dir_index = 0;
+	int ret = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	/* check the ref and backref exists */
+	inode_ref = btrfs_lookup_inode_ref(trans, root, path, name, namelen,
+					   ino, parent_ino, index, 0);
+	if (IS_ERR(inode_ref)) {
+		ret = PTR_ERR(inode_ref);
+		goto out;
+	}
+	if (inode_ref)
+		del_inode_ref = 1;
+	btrfs_release_path(path);
+
+	dir_item = btrfs_lookup_dir_item(NULL, root, path, parent_ino,
+					 name, namelen, 0);
+	if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		goto out;
+	}
+	if (dir_item)
+		del_dir_item = 1;
+	btrfs_release_path(path);
+
+	dir_item = btrfs_lookup_dir_index(NULL, root, path, parent_ino,
+					  name, namelen, index, 0);
+	/*
+	 * Since lookup_dir_index() will return -ENOENT when not found,
+	 * we need to do extra check.
+	 */
+	if (IS_ERR(dir_item) && PTR_ERR(dir_item) == -ENOENT)
+		dir_item = NULL;
+	if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		goto out;
+	}
+	if (dir_item)
+		del_dir_index = 1;
+	btrfs_release_path(path);
+
+	if (!del_inode_ref && !del_dir_item && !del_dir_index) {
+		/* All not found, shouldn't happen */
+		ret = -ENOENT;
+		goto out;
+	}
+
+	if (del_inode_ref) {
+		/* Only decrease nlink when deleting inode_ref */
+		key.objectid = ino;
+		key.type = BTRFS_INODE_ITEM_KEY;
+		key.offset = 0;
+		ret = btrfs_search_slot(trans, root, &key, path, -1, 1);
+		if (ret) {
+			if (ret > 0)
+				ret = -ENOENT;
+			goto out;
+		}
+		inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					    struct btrfs_inode_item);
+		nlinks = btrfs_inode_nlink(path->nodes[0], inode_item);
+		if (nlinks > 0)
+			nlinks--;
+		btrfs_set_inode_nlink(path->nodes[0], inode_item, nlinks);
+		btrfs_mark_buffer_dirty(path->nodes[0]);
+		btrfs_release_path(path);
+
+		/* For nlinks == 0, add it to orphan list if needed */
+		if (nlinks == 0 && add_orphan) {
+			ret = btrfs_add_orphan_item(trans, root, path, ino);
+			btrfs_mark_buffer_dirty(path->nodes[0]);
+			btrfs_release_path(path);
+		}
+
+		ret = btrfs_del_inode_ref(trans, root, name, namelen, ino,
+					  parent_ino, &index);
+		if (ret < 0)
+			goto out;
+	}
+
+	if (del_dir_index) {
+		dir_item = btrfs_lookup_dir_index(trans, root, path,
+						  parent_ino, name, namelen,
+						  index, -1);
+		if (IS_ERR(dir_item)) {
+			ret = PTR_ERR(dir_item);
+			goto out;
+		}
+		if (!dir_item) {
+			ret = -ENOENT;
+			goto out;
+		}
+		ret = btrfs_delete_one_dir_name(trans, root, path, dir_item);
+		if (ret)
+			goto out;
+		btrfs_release_path(path);
+
+		/* Update inode size of the parent inode */
+		key.objectid = parent_ino;
+		key.type = BTRFS_INODE_ITEM_KEY;
+		key.offset = 0;
+		ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
+		if (ret)
+			goto out;
+		inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					    struct btrfs_inode_item);
+		inode_size = btrfs_inode_size(path->nodes[0], inode_item);
+		if (inode_size >= namelen)
+			inode_size -= namelen;
+		btrfs_set_inode_size(path->nodes[0], inode_item, inode_size);
+		btrfs_mark_buffer_dirty(path->nodes[0]);
+		btrfs_release_path(path);
+	}
+
+	if (del_dir_item) {
+		dir_item = btrfs_lookup_dir_item(trans, root, path, parent_ino,
+						 name, namelen, -1);
+		if (IS_ERR(dir_item)) {
+			ret = PTR_ERR(dir_item);
+			goto out;
+		}
+		if (!dir_item) {
+			ret = -ENOENT;
+			goto out;
+		}
+		ret = btrfs_delete_one_dir_name(trans, root, path, dir_item);
+		if (ret < 0)
+			goto out;
+		btrfs_release_path(path);
+
+		/* Update inode size of the parent inode */
+		key.objectid = parent_ino;
+		key.type = BTRFS_INODE_ITEM_KEY;
+		key.offset = 0;
+		ret = btrfs_search_slot(trans, root, &key, path, 1, 1);
+		if (ret)
+			goto out;
+		inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+					    struct btrfs_inode_item);
+		inode_size = btrfs_inode_size(path->nodes[0], inode_item);
+		if (inode_size >= namelen)
+			inode_size -= namelen;
+		btrfs_set_inode_size(path->nodes[0], inode_item, inode_size);
+		btrfs_mark_buffer_dirty(path->nodes[0]);
+		btrfs_release_path(path);
+	}
+
+out:
+	btrfs_free_path(path);
+	return ret;
+}
-- 
2.1.3

[PATCH RESEND v2 5/6] btrfs-progs: Add btrfs_mkdir() function for the incoming 'lost+found' fsck mechanism.

[Qu Wenruo]

With the previous btrfs inode operations patches, now we can use
btrfs_mkdir() to create the 'lost+found' dir to do some data salvage in
btrfsck.

This patch along with previous ones will make data salvage easier.

Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
---
Changlog:
v2:
   Fix a bug that returns the parent ino other than the existing dir
   ino.
---
 ctree.h |  2 ++
 inode.c | 92 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 94 insertions(+)

diff --git a/ctree.h b/ctree.h
index 21bafd2..682255c 100644
--- a/ctree.h
+++ b/ctree.h
@@ -2459,4 +2459,6 @@ int btrfs_add_orphan_item(struct btrfs_trans_handle *trans,
 			  struct btrfs_root *root,
 			  struct btrfs_path *path,
 			  u64 ino);
+int btrfs_mkdir(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		char *name, int namelen, u64 parent_ino, u64 *ino, int mode);
 #endif
diff --git a/inode.c b/inode.c
index aeb0a5c..b354f5a 100644
--- a/inode.c
+++ b/inode.c
@@ -436,3 +436,95 @@ out:
 	btrfs_free_path(path);
 	return ret;
 }
+
+/* Fill inode item with 'mode'. Uid/gid to root/root */
+static void fill_inode_item(struct btrfs_trans_handle *trans,
+			    struct btrfs_inode_item *inode_item,
+			    u32 mode, u32 nlink)
+{
+	time_t now = time(NULL);
+
+	btrfs_set_stack_inode_generation(inode_item, trans->transid);
+	btrfs_set_stack_inode_uid(inode_item, 0);
+	btrfs_set_stack_inode_gid(inode_item, 0);
+	btrfs_set_stack_inode_size(inode_item, 0);
+	btrfs_set_stack_inode_mode(inode_item, mode);
+	btrfs_set_stack_inode_nlink(inode_item, nlink);
+	btrfs_set_stack_timespec_sec(&inode_item->atime, now);
+	btrfs_set_stack_timespec_nsec(&inode_item->atime, 0);
+	btrfs_set_stack_timespec_sec(&inode_item->mtime, now);
+	btrfs_set_stack_timespec_nsec(&inode_item->mtime, 0);
+	btrfs_set_stack_timespec_sec(&inode_item->ctime, now);
+	btrfs_set_stack_timespec_nsec(&inode_item->ctime, 0);
+}
+
+/*
+ * Unlike kernel btrfs_new_inode(), we only create the INODE_ITEM, without
+ * its backref.
+ * The backref is added by btrfs_add_link().
+ */
+static int btrfs_new_inode(struct btrfs_trans_handle *trans,
+			   struct btrfs_root *root,
+			   u64 ino, u32 mode)
+{
+	struct btrfs_inode_item inode_item = {0};
+	int ret = 0;
+
+	fill_inode_item(trans, &inode_item, mode, 0);
+	ret = btrfs_insert_inode(trans, root, ino, &inode_item);
+	return ret;
+}
+
+/*
+ * Make a dir under the parent inode 'parent_ino' with 'name'
+ * and 'mode', The owner will be root/root.
+ */
+int btrfs_mkdir(struct btrfs_trans_handle *trans, struct btrfs_root *root,
+		char *name, int namelen, u64 parent_ino, u64 *ino, int mode)
+{
+	struct btrfs_dir_item *dir_item;
+	struct btrfs_path *path;
+	u64 ret_ino;
+	int ret = 0;
+
+	path = btrfs_alloc_path();
+	if (!path)
+		return -ENOMEM;
+
+	dir_item = btrfs_lookup_dir_item(NULL, root, path, parent_ino,
+					 name, namelen, 0);
+	if (IS_ERR(dir_item)) {
+		ret = PTR_ERR(dir_item);
+		goto out;
+	}
+
+	if (dir_item) {
+		struct btrfs_key found_key;
+
+		/*
+		 * Already have conflicting name, check if it is a dir.
+		 * Either way, no need to continue.
+		 */
+		btrfs_dir_item_key_to_cpu(path->nodes[0], dir_item, &found_key);
+		ret_ino = found_key.objectid;
+		if (btrfs_dir_type(path->nodes[0], dir_item) != BTRFS_FT_DIR)
+			ret = -EEXIST;
+		goto out;
+	}
+
+	ret = btrfs_find_free_objectid(NULL, root, parent_ino, &ret_ino);
+	if (ret)
+		goto out;
+	ret = btrfs_new_inode(trans, root, ret_ino, mode | S_IFDIR);
+	if (ret)
+		goto out;
+	ret = btrfs_add_link(trans, root, ret_ino, parent_ino, name, namelen,
+			     BTRFS_FT_DIR, NULL, 1);
+	if (ret)
+		goto out;
+out:
+	btrfs_free_path(path);
+	if (ret == 0 && ino)
+		*ino = ret_ino;
+	return ret;
+}
-- 
2.1.3

[PATCH v2 6/6] btrfs-progs: Add fixing function for inodes whose nlink dismatch

[Qu Wenruo]

[BUG]
At least two users have already hit a bug in btrfs causing file
missing(Chromium config file).
The missing file's nlink is still 1 but its backref points to non-exist
parent inode.

This should be a kernel bug, but btrfsck fix is needed anyway.

[FIX]
For such nlink mismatch inode, we will delete all the invalid backref,
if there is no valid backref for it, create 'lost+found' under the root
of the subvolume and add link to the directory.

Reported-by: Mike Gavrilov <mikhail.v.gavrilov@gmail.com>
Reported-by: Ed Tomlinson <edt@aei.ca>
Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>

---
changelog:
v2:
   Use the a more generic fucntion, reset_nlink(), to repair the inode
   nlink.
   It will remove all, including valid, backref(along with
   dir_item/index) and set nlink to 0, and add valid ones back.
   This reset_nlink() method can handle more nlink error, not only
   invalid inode_ref but also pure nlinks mismatch(2 valid inode_ref but
   nlink is 1)
---
 cmds-check.c | 236 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
 1 file changed, 232 insertions(+), 4 deletions(-)

diff --git a/cmds-check.c b/cmds-check.c
index 6419caf..627b794 100644
--- a/cmds-check.c
+++ b/cmds-check.c
@@ -27,6 +27,7 @@
 #include <unistd.h>
 #include <getopt.h>
 #include <uuid/uuid.h>
+#include <math.h>
 #include "ctree.h"
 #include "volumes.h"
 #include "repair.h"
@@ -1837,6 +1838,18 @@ static int repair_inode_backrefs(struct btrfs_root *root,
 			struct btrfs_trans_handle *trans;
 			struct btrfs_key location;
 
+			ret = check_dir_conflict(root, backref->name,
+						 backref->namelen,
+						 backref->dir,
+						 backref->index);
+			if (ret) {
+				/*
+				 * let nlink fixing routine to handle it,
+				 * which can do it better.
+				 */
+				ret = 0;
+				break;
+			}
 			location.objectid = rec->ino;
 			location.type = BTRFS_INODE_ITEM_KEY;
 			location.offset = 0;
@@ -1874,20 +1887,233 @@ static int repair_inode_backrefs(struct btrfs_root *root,
 	return ret ? ret : repaired;
 }
 
+/*
+ * Search for the inode_ref of given 'ino' to get the filename and
+ * btrfs type.
+ * This will only use the first inode_ref.
+ */
+static int find_file_name_type(struct btrfs_root *root,
+			       struct btrfs_path *path,
+			       u64 ino, char *buf,
+			       int *namelen, u8 *type)
+{
+	struct btrfs_key key;
+	struct btrfs_key found_key;
+	struct btrfs_inode_ref *inode_ref;
+	struct btrfs_inode_item *inode_item;
+	struct extent_buffer *node;
+	u32 ret_namelen;
+	int slot;
+	int ret = 0;
+
+	/* Search for name in backref(Use the first one) */
+	key.objectid = ino;
+	key.type = BTRFS_INODE_REF_KEY;
+	key.offset = 0;
+
+	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
+	if (ret < 0)
+		goto out;
+	node = path->nodes[0];
+	slot = path->slots[0];
+	btrfs_item_key_to_cpu(node, &found_key, slot);
+	if (found_key.objectid != ino ||
+	    found_key.type != BTRFS_INODE_REF_KEY) {
+		ret = -ENOENT;
+		goto out;
+	}
+	inode_ref = btrfs_item_ptr(node, slot, struct btrfs_inode_ref);
+	ret_namelen = btrfs_inode_ref_name_len(node, inode_ref);
+	read_extent_buffer(node, buf, (unsigned long)(inode_ref + 1),
+			   ret_namelen);
+	/* Search for inode type */
+	ret = btrfs_previous_item(root, path, ino, BTRFS_INODE_ITEM_KEY);
+	if (ret) {
+		if (ret > 0)
+			ret = -ENOENT;
+		goto out;
+	}
+	node = path->nodes[0];
+	slot = path->slots[0];
+	inode_item = btrfs_item_ptr(node, slot, struct btrfs_inode_item);
+
+	if (namelen)
+		*namelen = ret_namelen;
+	if (type)
+		*type = imode_to_type(btrfs_inode_mode(node, inode_item));
+out:
+	btrfs_release_path(path);
+	return ret;
+}
+
+/* Reset the nlink of the inode to the correct one */
+static int reset_nlink(struct btrfs_trans_handle *trans,
+		       struct btrfs_root *root,
+		       struct btrfs_path *path,
+		       struct inode_record *rec)
+{
+	struct inode_backref *backref;
+	struct inode_backref *tmp;
+	struct btrfs_key key;
+	struct btrfs_inode_item *inode_item;
+	int ret = 0;
+
+	/* Remove all backref including the valid ones */
+	list_for_each_entry_safe(backref, tmp, &rec->backrefs, list) {
+		ret = btrfs_unlink(trans, root, rec->ino, backref->dir,
+				   backref->index, backref->name,
+				   backref->namelen, 0);
+		if (ret < 0)
+			goto out;
+
+		/* remove invalid backref, so it won't be added back */
+		if (!(backref->found_dir_index &&
+		      backref->found_dir_item &&
+		      backref->found_inode_ref)) {
+			list_del(&backref->list);
+			free(backref);
+		}
+	}
+
+	/* Set nlink to 0 */
+	key.objectid = rec->ino;
+	key.type = BTRFS_INODE_ITEM_KEY;
+	key.offset = 0;
+	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
+	if (ret < 0)
+		goto out;
+	if (ret > 0) {
+		ret = -ENOENT;
+		goto out;
+	}
+	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
+				    struct btrfs_inode_item);
+	btrfs_set_inode_nlink(path->nodes[0], inode_item, 0);
+	btrfs_mark_buffer_dirty(path->nodes[0]);
+	btrfs_release_path(path);
+
+	/*
+	 * Add back valid inode_ref/dir_item/dir_index,
+	 * add_link() will handle the nlink inc, so new nlink must be correct
+	 */
+	list_for_each_entry(backref, &rec->backrefs, list) {
+		ret = btrfs_add_link(trans, root, rec->ino, backref->dir,
+				     backref->name, backref->namelen,
+				     backref->ref_type, &backref->index, 1);
+		if (ret < 0)
+			goto out;
+	}
+out:
+	btrfs_release_path(path);
+	return ret;
+}
+
+static int repair_inode_nlinks(struct btrfs_trans_handle *trans,
+			       struct btrfs_root *root,
+			       struct btrfs_path *path,
+			       struct inode_record *rec)
+{
+	char *dir_name = "lost+found";
+	char namebuf[BTRFS_NAME_LEN] = {0};
+	u64 lost_found_ino;
+	u32 mode = 0700;
+	u8 type = 0;
+	int namelen = 0;
+	int ret = 0;
+
+	/*
+	 * Get file name and type first before these invalid inode ref
+	 * are deleted by remove_all_invalid_backref()
+	 */
+	ret = find_file_name_type(root, path, rec->ino, namebuf,
+				  &namelen, &type);
+	if (ret < 0) {
+		fprintf(stderr,
+			"Fail to get file name of inode %llu: %s\n",
+			rec->ino, strerror(-ret));
+		goto out;
+	}
+	ret = reset_nlink(trans, root, path, rec);
+	if (ret < 0) {
+		fprintf(stderr,
+			"Fail to reset nlink for inode %llu: %s\n",
+			rec->ino, strerror(-ret));
+		goto out;
+	}
+
+	if (rec->found_link == 0) {
+		ret = btrfs_mkdir(trans, root, dir_name, strlen(dir_name),
+				  BTRFS_FIRST_FREE_OBJECTID, &lost_found_ino,
+				  mode);
+		if (ret < 0) {
+			fprintf(stderr, "Failed to create '%s' dir: %s",
+				dir_name, strerror(-ret));
+			goto out;
+		}
+		ret = btrfs_add_link(trans, root, rec->ino, lost_found_ino,
+				     namebuf, namelen, type, NULL, 1);
+		if (ret == -EEXIST) {
+			/*
+			 * Conflicting file name, add ".INO" as suffix
+			 * +1 for '.' and +1 for log10
+			 */
+			if (namelen + log10(rec->ino) + 2 > BTRFS_NAME_LEN) {
+				ret = -EFBIG;
+				goto out;
+			}
+			snprintf(namebuf + namelen, BTRFS_NAME_LEN - namelen,
+				 ".%llu", rec->ino);
+			namelen += (log10(rec->ino) + 2);
+			ret = btrfs_add_link(trans, root, rec->ino,
+					     lost_found_ino, namebuf,
+					     namelen, type, NULL, 1);
+		}
+		if (ret < 0) {
+			fprintf(stderr,
+				"Fail to link the inode %llu to %s dir: %s",
+				rec->ino, dir_name, strerror(-ret));
+			goto out;
+		}
+		/*
+		 * Just increase the found_link, don't actually add the
+		 * backref. This will make things easiler and this inode
+		 * record will be freed after the repair is done.
+		 * So fsck will not report problem about this inode.
+		 */
+		rec->found_link++;
+		printf("Moving file '%.*s' to '%s' dir since it has no valid backref\n",
+		       namelen, namebuf, dir_name);
+	}
+	rec->errors &= ~I_ERR_LINK_COUNT_WRONG;
+	printf("Fixed the nlink of inode %llu\n", rec->ino);
+out:
+	btrfs_release_path(path);
+	return ret;
+}
+
 static int try_repair_inode(struct btrfs_root *root, struct inode_record *rec)
 {
 	struct btrfs_trans_handle *trans;
 	struct btrfs_path *path;
 	int ret = 0;
 
-	if (!(rec->errors & (I_ERR_DIR_ISIZE_WRONG | I_ERR_NO_ORPHAN_ITEM)))
+	if (!(rec->errors & (I_ERR_DIR_ISIZE_WRONG |
+			     I_ERR_NO_ORPHAN_ITEM |
+			     I_ERR_LINK_COUNT_WRONG)))
 		return rec->errors;
 
 	path = btrfs_alloc_path();
 	if (!path)
 		return -ENOMEM;
 
-	trans = btrfs_start_transaction(root, 1);
+	/*
+	 * For nlink repair, it may create a dir and add link, so
+	 * 2 for parent(256)'s dir_index and dir_item
+	 * 2 for lost+found dir's inode_item and inode_ref
+	 * 1 for the new inode_ref of the file
+	 * 2 for lost+found dir's dir_index and dir_item for the file
+	 */
+	trans = btrfs_start_transaction(root, 7);
 	if (IS_ERR(trans)) {
 		btrfs_free_path(path);
 		return PTR_ERR(trans);
@@ -1897,6 +2123,8 @@ static int try_repair_inode(struct btrfs_root *root, struct inode_record *rec)
 		ret = repair_inode_isize(trans, root, path, rec);
 	if (!ret && rec->errors & I_ERR_NO_ORPHAN_ITEM)
 		ret = repair_inode_orphan_item(trans, root, path, rec);
+	if (!ret && rec->errors & I_ERR_LINK_COUNT_WRONG)
+		ret = repair_inode_nlinks(trans, root, path, rec);
 	btrfs_commit_transaction(trans, root);
 	btrfs_free_path(path);
 	return ret;
@@ -2032,6 +2260,8 @@ static int check_inode_recs(struct btrfs_root *root,
 			}
 		}
 
+		if (rec->found_link != rec->nlink)
+			rec->errors |= I_ERR_LINK_COUNT_WRONG;
 		if (repair) {
 			ret = try_repair_inode(root, rec);
 			if (ret == 0 && can_free_inode_rec(rec)) {
@@ -2044,8 +2274,6 @@ static int check_inode_recs(struct btrfs_root *root,
 		error++;
 		if (!rec->found_inode_item)
 			rec->errors |= I_ERR_NO_INODE_ITEM;
-		if (rec->found_link != rec->nlink)
-			rec->errors |= I_ERR_LINK_COUNT_WRONG;
 		print_inode_error(root, rec);
 		list_for_each_entry(backref, &rec->backrefs, list) {
 			if (!backref->found_dir_item)
-- 
2.1.3

Re: [PATCH v2 6/6] btrfs-progs: Add fixing function for inodes whose nlink dismatch

[Ed Tomlinson]

Hi,

Looks like 

Reported-by:  Daniel Miranda <danielkza2@gmail.com>

also needs to be added see: "Re: Apparent metadata corruption (file that simultaneously does/does not exist) on kernel 3.17.3"
Where Daniel reports these patches fixed his fs too.  

I expect an fsck with --repair specified to change my fs and creating and populating an lost+found directory is not an unexpected result.
While btrfs is not supposed to need this - its obvious that it does and that this corruption is not that uncommon.

Three victims repaired in less than two weeks - how many more are out there?

Thanks
Ed

PS. The other important question is how do we find the bug causing this and fix it?

On Wednesday 03 December 2014 12:18:32 Qu Wenruo wrote:
> [BUG]
> At least two users have already hit a bug in btrfs causing file
> missing(Chromium config file).
> The missing file's nlink is still 1 but its backref points to non-exist
> parent inode.
> 
> This should be a kernel bug, but btrfsck fix is needed anyway.
> 
> [FIX]
> For such nlink mismatch inode, we will delete all the invalid backref,
> if there is no valid backref for it, create 'lost+found' under the root
> of the subvolume and add link to the directory.
> 
> Reported-by: Mike Gavrilov <mikhail.v.gavrilov@gmail.com>
> Reported-by: Ed Tomlinson <edt@aei.ca>
> Signed-off-by: Qu Wenruo <quwenruo@cn.fujitsu.com>
> 
> ---
> changelog:
> v2:
>    Use the a more generic fucntion, reset_nlink(), to repair the inode
>    nlink.
>    It will remove all, including valid, backref(along with
>    dir_item/index) and set nlink to 0, and add valid ones back.
>    This reset_nlink() method can handle more nlink error, not only
>    invalid inode_ref but also pure nlinks mismatch(2 valid inode_ref but
>    nlink is 1)
> ---
>  cmds-check.c | 236 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
>  1 file changed, 232 insertions(+), 4 deletions(-)
> 
> diff --git a/cmds-check.c b/cmds-check.c
> index 6419caf..627b794 100644
> --- a/cmds-check.c
> +++ b/cmds-check.c
> @@ -27,6 +27,7 @@
>  #include <unistd.h>
>  #include <getopt.h>
>  #include <uuid/uuid.h>
> +#include <math.h>
>  #include "ctree.h"
>  #include "volumes.h"
>  #include "repair.h"
> @@ -1837,6 +1838,18 @@ static int repair_inode_backrefs(struct btrfs_root *root,
>  			struct btrfs_trans_handle *trans;
>  			struct btrfs_key location;
>  
> +			ret = check_dir_conflict(root, backref->name,
> +						 backref->namelen,
> +						 backref->dir,
> +						 backref->index);
> +			if (ret) {
> +				/*
> +				 * let nlink fixing routine to handle it,
> +				 * which can do it better.
> +				 */
> +				ret = 0;
> +				break;
> +			}
>  			location.objectid = rec->ino;
>  			location.type = BTRFS_INODE_ITEM_KEY;
>  			location.offset = 0;
> @@ -1874,20 +1887,233 @@ static int repair_inode_backrefs(struct btrfs_root *root,
>  	return ret ? ret : repaired;
>  }
>  
> +/*
> + * Search for the inode_ref of given 'ino' to get the filename and
> + * btrfs type.
> + * This will only use the first inode_ref.
> + */
> +static int find_file_name_type(struct btrfs_root *root,
> +			       struct btrfs_path *path,
> +			       u64 ino, char *buf,
> +			       int *namelen, u8 *type)
> +{
> +	struct btrfs_key key;
> +	struct btrfs_key found_key;
> +	struct btrfs_inode_ref *inode_ref;
> +	struct btrfs_inode_item *inode_item;
> +	struct extent_buffer *node;
> +	u32 ret_namelen;
> +	int slot;
> +	int ret = 0;
> +
> +	/* Search for name in backref(Use the first one) */
> +	key.objectid = ino;
> +	key.type = BTRFS_INODE_REF_KEY;
> +	key.offset = 0;
> +
> +	ret = btrfs_search_slot(NULL, root, &key, path, 0, 0);
> +	if (ret < 0)
> +		goto out;
> +	node = path->nodes[0];
> +	slot = path->slots[0];
> +	btrfs_item_key_to_cpu(node, &found_key, slot);
> +	if (found_key.objectid != ino ||
> +	    found_key.type != BTRFS_INODE_REF_KEY) {
> +		ret = -ENOENT;
> +		goto out;
> +	}
> +	inode_ref = btrfs_item_ptr(node, slot, struct btrfs_inode_ref);
> +	ret_namelen = btrfs_inode_ref_name_len(node, inode_ref);
> +	read_extent_buffer(node, buf, (unsigned long)(inode_ref + 1),
> +			   ret_namelen);
> +	/* Search for inode type */
> +	ret = btrfs_previous_item(root, path, ino, BTRFS_INODE_ITEM_KEY);
> +	if (ret) {
> +		if (ret > 0)
> +			ret = -ENOENT;
> +		goto out;
> +	}
> +	node = path->nodes[0];
> +	slot = path->slots[0];
> +	inode_item = btrfs_item_ptr(node, slot, struct btrfs_inode_item);
> +
> +	if (namelen)
> +		*namelen = ret_namelen;
> +	if (type)
> +		*type = imode_to_type(btrfs_inode_mode(node, inode_item));
> +out:
> +	btrfs_release_path(path);
> +	return ret;
> +}
> +
> +/* Reset the nlink of the inode to the correct one */
> +static int reset_nlink(struct btrfs_trans_handle *trans,
> +		       struct btrfs_root *root,
> +		       struct btrfs_path *path,
> +		       struct inode_record *rec)
> +{
> +	struct inode_backref *backref;
> +	struct inode_backref *tmp;
> +	struct btrfs_key key;
> +	struct btrfs_inode_item *inode_item;
> +	int ret = 0;
> +
> +	/* Remove all backref including the valid ones */
> +	list_for_each_entry_safe(backref, tmp, &rec->backrefs, list) {
> +		ret = btrfs_unlink(trans, root, rec->ino, backref->dir,
> +				   backref->index, backref->name,
> +				   backref->namelen, 0);
> +		if (ret < 0)
> +			goto out;
> +
> +		/* remove invalid backref, so it won't be added back */
> +		if (!(backref->found_dir_index &&
> +		      backref->found_dir_item &&
> +		      backref->found_inode_ref)) {
> +			list_del(&backref->list);
> +			free(backref);
> +		}
> +	}
> +
> +	/* Set nlink to 0 */
> +	key.objectid = rec->ino;
> +	key.type = BTRFS_INODE_ITEM_KEY;
> +	key.offset = 0;
> +	ret = btrfs_search_slot(trans, root, &key, path, 0, 1);
> +	if (ret < 0)
> +		goto out;
> +	if (ret > 0) {
> +		ret = -ENOENT;
> +		goto out;
> +	}
> +	inode_item = btrfs_item_ptr(path->nodes[0], path->slots[0],
> +				    struct btrfs_inode_item);
> +	btrfs_set_inode_nlink(path->nodes[0], inode_item, 0);
> +	btrfs_mark_buffer_dirty(path->nodes[0]);
> +	btrfs_release_path(path);
> +
> +	/*
> +	 * Add back valid inode_ref/dir_item/dir_index,
> +	 * add_link() will handle the nlink inc, so new nlink must be correct
> +	 */
> +	list_for_each_entry(backref, &rec->backrefs, list) {
> +		ret = btrfs_add_link(trans, root, rec->ino, backref->dir,
> +				     backref->name, backref->namelen,
> +				     backref->ref_type, &backref->index, 1);
> +		if (ret < 0)
> +			goto out;
> +	}
> +out:
> +	btrfs_release_path(path);
> +	return ret;
> +}
> +
> +static int repair_inode_nlinks(struct btrfs_trans_handle *trans,
> +			       struct btrfs_root *root,
> +			       struct btrfs_path *path,
> +			       struct inode_record *rec)
> +{
> +	char *dir_name = "lost+found";
> +	char namebuf[BTRFS_NAME_LEN] = {0};
> +	u64 lost_found_ino;
> +	u32 mode = 0700;
> +	u8 type = 0;
> +	int namelen = 0;
> +	int ret = 0;
> +
> +	/*
> +	 * Get file name and type first before these invalid inode ref
> +	 * are deleted by remove_all_invalid_backref()
> +	 */
> +	ret = find_file_name_type(root, path, rec->ino, namebuf,
> +				  &namelen, &type);
> +	if (ret < 0) {
> +		fprintf(stderr,
> +			"Fail to get file name of inode %llu: %s\n",
> +			rec->ino, strerror(-ret));
> +		goto out;
> +	}
> +	ret = reset_nlink(trans, root, path, rec);
> +	if (ret < 0) {
> +		fprintf(stderr,
> +			"Fail to reset nlink for inode %llu: %s\n",
> +			rec->ino, strerror(-ret));
> +		goto out;
> +	}
> +
> +	if (rec->found_link == 0) {
> +		ret = btrfs_mkdir(trans, root, dir_name, strlen(dir_name),
> +				  BTRFS_FIRST_FREE_OBJECTID, &lost_found_ino,
> +				  mode);
> +		if (ret < 0) {
> +			fprintf(stderr, "Failed to create '%s' dir: %s",
> +				dir_name, strerror(-ret));
> +			goto out;
> +		}
> +		ret = btrfs_add_link(trans, root, rec->ino, lost_found_ino,
> +				     namebuf, namelen, type, NULL, 1);
> +		if (ret == -EEXIST) {
> +			/*
> +			 * Conflicting file name, add ".INO" as suffix
> +			 * +1 for '.' and +1 for log10
> +			 */
> +			if (namelen + log10(rec->ino) + 2 > BTRFS_NAME_LEN) {
> +				ret = -EFBIG;
> +				goto out;
> +			}
> +			snprintf(namebuf + namelen, BTRFS_NAME_LEN - namelen,
> +				 ".%llu", rec->ino);
> +			namelen += (log10(rec->ino) + 2);
> +			ret = btrfs_add_link(trans, root, rec->ino,
> +					     lost_found_ino, namebuf,
> +					     namelen, type, NULL, 1);
> +		}
> +		if (ret < 0) {
> +			fprintf(stderr,
> +				"Fail to link the inode %llu to %s dir: %s",
> +				rec->ino, dir_name, strerror(-ret));
> +			goto out;
> +		}
> +		/*
> +		 * Just increase the found_link, don't actually add the
> +		 * backref. This will make things easiler and this inode
> +		 * record will be freed after the repair is done.
> +		 * So fsck will not report problem about this inode.
> +		 */
> +		rec->found_link++;
> +		printf("Moving file '%.*s' to '%s' dir since it has no valid backref\n",
> +		       namelen, namebuf, dir_name);
> +	}
> +	rec->errors &= ~I_ERR_LINK_COUNT_WRONG;
> +	printf("Fixed the nlink of inode %llu\n", rec->ino);
> +out:
> +	btrfs_release_path(path);
> +	return ret;
> +}
> +
>  static int try_repair_inode(struct btrfs_root *root, struct inode_record *rec)
>  {
>  	struct btrfs_trans_handle *trans;
>  	struct btrfs_path *path;
>  	int ret = 0;
>  
> -	if (!(rec->errors & (I_ERR_DIR_ISIZE_WRONG | I_ERR_NO_ORPHAN_ITEM)))
> +	if (!(rec->errors & (I_ERR_DIR_ISIZE_WRONG |
> +			     I_ERR_NO_ORPHAN_ITEM |
> +			     I_ERR_LINK_COUNT_WRONG)))
>  		return rec->errors;
>  
>  	path = btrfs_alloc_path();
>  	if (!path)
>  		return -ENOMEM;
>  
> -	trans = btrfs_start_transaction(root, 1);
> +	/*
> +	 * For nlink repair, it may create a dir and add link, so
> +	 * 2 for parent(256)'s dir_index and dir_item
> +	 * 2 for lost+found dir's inode_item and inode_ref
> +	 * 1 for the new inode_ref of the file
> +	 * 2 for lost+found dir's dir_index and dir_item for the file
> +	 */
> +	trans = btrfs_start_transaction(root, 7);
>  	if (IS_ERR(trans)) {
>  		btrfs_free_path(path);
>  		return PTR_ERR(trans);
> @@ -1897,6 +2123,8 @@ static int try_repair_inode(struct btrfs_root *root, struct inode_record *rec)
>  		ret = repair_inode_isize(trans, root, path, rec);
>  	if (!ret && rec->errors & I_ERR_NO_ORPHAN_ITEM)
>  		ret = repair_inode_orphan_item(trans, root, path, rec);
> +	if (!ret && rec->errors & I_ERR_LINK_COUNT_WRONG)
> +		ret = repair_inode_nlinks(trans, root, path, rec);
>  	btrfs_commit_transaction(trans, root);
>  	btrfs_free_path(path);
>  	return ret;
> @@ -2032,6 +2260,8 @@ static int check_inode_recs(struct btrfs_root *root,
>  			}
>  		}
>  
> +		if (rec->found_link != rec->nlink)
> +			rec->errors |= I_ERR_LINK_COUNT_WRONG;
>  		if (repair) {
>  			ret = try_repair_inode(root, rec);
>  			if (ret == 0 && can_free_inode_rec(rec)) {
> @@ -2044,8 +2274,6 @@ static int check_inode_recs(struct btrfs_root *root,
>  		error++;
>  		if (!rec->found_inode_item)
>  			rec->errors |= I_ERR_NO_INODE_ITEM;
> -		if (rec->found_link != rec->nlink)
> -			rec->errors |= I_ERR_LINK_COUNT_WRONG;
>  		print_inode_error(root, rec);
>  		list_for_each_entry(backref, &rec->backrefs, list) {
>  			if (!backref->found_dir_item)
> 

Re: [PATCH 0/6] More generic inode nlink repair function

[Ed Tomlinson]

Hi,

I'd really like to see these patches included in btrfsck - they repaired my fs.   Once
Qu got them working they found additional corruptions.  This time there was no crash or stall
just an umount that left (chromium) files unlinked...  The bug with these files has been
hitting me for a while - just did not recognize what was causing it or notice the corruption.

The only objection I have seen to these patches is that they may create a "lost+found" 
directory.  I submit this is an expected behavior for a fsck utility.  When --repair is specified 
I expect a fsck to make changes to my fs one of which may be adding and populating a 
lost+found directory.

Thanks
Ed Tomlinson

PS. It would be very interesting to find out WHY these files are ending up unlinked.  Ideas?


On Wednesday 03 December 2014 12:18:26 you wrote:
> Update on patch 4 and 6, other is not changed.
> This nlink repair function is more generic than the original one.
> 
> The old one can only handle a specific case that the inode_ref is
> invalid, either point to a non-exist parent inode or point to a invalid
> inode(not dir or conflicting index/name).
> 
> The new one will reset all the backref, no matter it is valid or not,
> and re-add all the valid backref, this make the nlink handles more
> corrupt cases.
> 
> Qu Wenruo (6):
>   btrfs-progs: print root dir verbose error in fsck
>   btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.
>   btrfs-progs: Import lookup/del_inode_ref() function.
>   btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.
>   btrfs-progs: Add btrfs_mkdir() function for the incoming 'lost+found' 
>        fsck mechanism.
>   btrfs-progs: Add fixing function for inodes whose nlink dismatch
> 
>  Makefile     |   2 +-
>  cmds-check.c | 311 ++++++++++++++++++++++++++++++++++++--
>  ctree.c      |   6 +
>  ctree.h      |  38 +++++
>  inode-item.c | 318 +++++++++++++++++++++++++++++++++++++++
>  inode.c      | 484 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>  6 files changed, 1148 insertions(+), 11 deletions(-)
>  create mode 100644 inode.c
> 
> 

Re: [PATCH 0/6] More generic inode nlink repair function

[Satoru Takeuchi]

Hi,

(2014/12/03 14:03), Ed Tomlinson wrote:
> Hi,
>
> I'd really like to see these patches included in btrfsck - they repaired my fs.   Once
> Qu got them working they found additional corruptions.  This time there was no crash or stall
> just an umount that left (chromium) files unlinked...  The bug with these files has been
> hitting me for a while - just did not recognize what was causing it or notice the corruption.
>
> The only objection I have seen to these patches is that they may create a "lost+found"
> directory.  I submit this is an expected behavior for a fsck utility.  When --repair is specified
> I expect a fsck to make changes to my fs one of which may be adding and populating a
> lost+found directory.

How about making lost+found on mkfs.btrfs like ext4?

Thanks,
Satoru

>
> Thanks
> Ed Tomlinson
>
> PS. It would be very interesting to find out WHY these files are ending up unlinked.  Ideas?
>
>
> On Wednesday 03 December 2014 12:18:26 you wrote:
>> Update on patch 4 and 6, other is not changed.
>> This nlink repair function is more generic than the original one.
>>
>> The old one can only handle a specific case that the inode_ref is
>> invalid, either point to a non-exist parent inode or point to a invalid
>> inode(not dir or conflicting index/name).
>>
>> The new one will reset all the backref, no matter it is valid or not,
>> and re-add all the valid backref, this make the nlink handles more
>> corrupt cases.
>>
>> Qu Wenruo (6):
>>    btrfs-progs: print root dir verbose error in fsck
>>    btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.
>>    btrfs-progs: Import lookup/del_inode_ref() function.
>>    btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.
>>    btrfs-progs: Add btrfs_mkdir() function for the incoming 'lost+found'
>>         fsck mechanism.
>>    btrfs-progs: Add fixing function for inodes whose nlink dismatch
>>
>>   Makefile     |   2 +-
>>   cmds-check.c | 311 ++++++++++++++++++++++++++++++++++++--
>>   ctree.c      |   6 +
>>   ctree.h      |  38 +++++
>>   inode-item.c | 318 +++++++++++++++++++++++++++++++++++++++
>>   inode.c      | 484 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>   6 files changed, 1148 insertions(+), 11 deletions(-)
>>   create mode 100644 inode.c
>>
>>
>
> --
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
>

Re: [PATCH 0/6] More generic inode nlink repair function

[Qu Wenruo]

-------- Original Message --------
Subject: Re: [PATCH 0/6] More generic inode nlink repair function
From: Satoru Takeuchi <takeuchi_satoru@jp.fujitsu.com>
To: Ed Tomlinson <edt@aei.ca>, <linux-btrfs@vger.kernel.org>
Date: 2014年12月03日 15:12
> Hi,
>
> (2014/12/03 14:03), Ed Tomlinson wrote:
>> Hi,
>>
>> I'd really like to see these patches included in btrfsck - they 
>> repaired my fs.   Once
>> Qu got them working they found additional corruptions.  This time 
>> there was no crash or stall
>> just an umount that left (chromium) files unlinked...  The bug with 
>> these files has been
>> hitting me for a while - just did not recognize what was causing it 
>> or notice the corruption.
>>
>> The only objection I have seen to these patches is that they may 
>> create a "lost+found"
>> directory.  I submit this is an expected behavior for a fsck 
>> utility.  When --repair is specified
>> I expect a fsck to make changes to my fs one of which may be adding 
>> and populating a
>> lost+found directory.
>
> How about making lost+found on mkfs.btrfs like ext4?
>
> Thanks,
> Satoru
>
I hope most user won't see the lost+found dir.

Due to the COW feature especially on all the metadata,
btrfs is *supposed* to be fsck free and self repairing fs, make 
lost+found at mkfs time
may make btrfs not so special among the old time ext* fs.

IMHO, most user will not see the lost+found dir, if hardware and kernel 
works as they should be,
usual bit flip can be recovered by the default DUP or higher RAID 
metadata profile.
The lost+found dir will only occur if the fs is really badly damaged or 
kernel bug.
(or intently damaged by btrfs-corrupt-block for QA reason).

With the maturing of btrfs, such case should be more and more rare and 
most user will forget
the fact that btrfsck will create lost+found dir, but it will still be 
the last hope.
So lost+found should only occur when it is needed, and when it is needed 
it will be created.

Thanks,
Qu
>>
>> Thanks
>> Ed Tomlinson
>>
>> PS. It would be very interesting to find out WHY these files are 
>> ending up unlinked.  Ideas?
>>
>>
>> On Wednesday 03 December 2014 12:18:26 you wrote:
>>> Update on patch 4 and 6, other is not changed.
>>> This nlink repair function is more generic than the original one.
>>>
>>> The old one can only handle a specific case that the inode_ref is
>>> invalid, either point to a non-exist parent inode or point to a invalid
>>> inode(not dir or conflicting index/name).
>>>
>>> The new one will reset all the backref, no matter it is valid or not,
>>> and re-add all the valid backref, this make the nlink handles more
>>> corrupt cases.
>>>
>>> Qu Wenruo (6):
>>>    btrfs-progs: print root dir verbose error in fsck
>>>    btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.
>>>    btrfs-progs: Import lookup/del_inode_ref() function.
>>>    btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.
>>>    btrfs-progs: Add btrfs_mkdir() function for the incoming 
>>> 'lost+found'
>>>         fsck mechanism.
>>>    btrfs-progs: Add fixing function for inodes whose nlink dismatch
>>>
>>>   Makefile     |   2 +-
>>>   cmds-check.c | 311 ++++++++++++++++++++++++++++++++++++--
>>>   ctree.c      |   6 +
>>>   ctree.h      |  38 +++++
>>>   inode-item.c | 318 +++++++++++++++++++++++++++++++++++++++
>>>   inode.c      | 484 
>>> +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
>>>   6 files changed, 1148 insertions(+), 11 deletions(-)
>>>   create mode 100644 inode.c
>>>
>>>
>>
>> -- 
>> To unsubscribe from this list: send the line "unsubscribe 
>> linux-btrfs" in
>> the body of a message to majordomo@vger.kernel.org
>> More majordomo info at http://vger.kernel.org/majordomo-info.html
>>
>
> -- 
> To unsubscribe from this list: send the line "unsubscribe linux-btrfs" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html

Re: [PATCH 0/6] More generic inode nlink repair function

[David Sterba]

On Wed, Dec 03, 2014 at 03:41:19PM +0800, Qu Wenruo wrote:
> > How about making lost+found on mkfs.btrfs like ext4?
> >
> I hope most user won't see the lost+found dir.
...
> So lost+found should only occur when it is needed, and when it is needed 
> it will be created.

Ack, as was mentioned before, lost+found is local to the containing
subvolume so creating one for the whole filesystem at mkfs time is of no
use.

The usecase for creating the directory (compared to copying the files
somewhere else) is to give a direct access to the files once the
repaired filesystem is mounted again. Both should be implemented in the
end.

Re: [PATCH 0/6] More generic inode nlink repair function

[Qu Wenruo]

-------- Original Message --------
Subject: Re: [PATCH 0/6] More generic inode nlink repair function
From: David Sterba <dsterba@suse.cz>
To: Qu Wenruo <quwenruo@cn.fujitsu.com>
Date: 2014年12月05日 01:20
> On Wed, Dec 03, 2014 at 03:41:19PM +0800, Qu Wenruo wrote:
>>> How about making lost+found on mkfs.btrfs like ext4?
>>>
>> I hope most user won't see the lost+found dir.
> ...
>> So lost+found should only occur when it is needed, and when it is needed
>> it will be created.
> Ack, as was mentioned before, lost+found is local to the containing
> subvolume so creating one for the whole filesystem at mkfs time is of no
> use.
>
> The usecase for creating the directory (compared to copying the files
> somewhere else) is to give a direct access to the files once the
> repaired filesystem is mounted again. Both should be implemented in the
> end.
Understood.
I'll also try if there is anything that can help btrfs-restore and 
backport it to restore later,
hoping later btrfsck patches can help both restore and btrfsck --repair.

Thanks,
Qu

Re: [PATCH 0/6] More generic inode nlink repair function

[David Sterba]

On Wed, Dec 03, 2014 at 12:18:26PM +0800, Qu Wenruo wrote:
> Update on patch 4 and 6, other is not changed.
> This nlink repair function is more generic than the original one.
> 
> The old one can only handle a specific case that the inode_ref is
> invalid, either point to a non-exist parent inode or point to a invalid
> inode(not dir or conflicting index/name).
> 
> The new one will reset all the backref, no matter it is valid or not,
> and re-add all the valid backref, this make the nlink handles more
> corrupt cases.
> 
> Qu Wenruo (6):
>   btrfs-progs: print root dir verbose error in fsck
>   btrfs-progs: Import btrfs_insert/del/lookup_extref() functions.
>   btrfs-progs: Import lookup/del_inode_ref() function.
>   btrfs-progs: Add btrfs_unlink() and btrfs_add_link() functions.
>   btrfs-progs: Add btrfs_mkdir() function for the incoming 'lost+found' 
>        fsck mechanism.
>   btrfs-progs: Add fixing function for inodes whose nlink dismatch

Thanks.  All patches queued for 3.18, I made some trivial formatting
chnages.