btrfs-progs: convert: insert a dummy inode item before inode ref for ext2/4

[BUG]
There is a report about failed btrfs-convert, which shows the following
error:

  Create btrfs metadata
  corrupt leaf: root=5 block=5001931145216 slot=1 ino=89911763, invalid previous key objectid, have 89911762 expect 89911763
  leaf 5001931145216 items 336 free space 7 generation 90 owner FS_TREE
  leaf 5001931145216 flags 0x1(WRITTEN) backref revision 1
  fs uuid 8b69f018-37c3-4b30-b859-42ccfcbe2449
  chunk uuid 448ce78c-ea41-49f6-99dc-46ad80b93da9
          item 0 key (89911762 INODE_REF 3858733) itemoff 16222 itemsize 61
                  index 171 namelen 51 name: [FILENAME1]
          item 1 key (89911763 INODE_REF 3858733) itemoff 16161 itemsize 61
                  index 103 namelen 51 name: [FILENAME2]

[CAUSE]
When iterating a directory, btrfs-convert would insert the DIR_ITEMs,
along with the INODE_REF of that inode.

This leads to above stray INODE_REFs, and trigger the tree-checker.

This can only happen for large fs, as for most cases we have all these
modified tree blocks cached, thus tree-checker won't be triggered.
But when the tree block cache is not hit, and we have to read from disk,
then such behavior can lead to above tree-checker error.

[FIX]
Insert a dummy INODE_ITEM for the INODE_REF first, the inode items would
be updated when iterating the child inode of the directory.

Issue: #731
Signed-off-by: Qu Wenruo <wqu@suse.com>
Signed-off-by: David Sterba <dsterba@suse.com>
This commit is contained in:
Qu Wenruo 2024-01-13 19:07:06 +10:30 committed by David Sterba
parent 04f3efaa0d
commit 517ba2d9e5
4 changed files with 56 additions and 25 deletions

View file

@ -167,21 +167,6 @@ static inline bool is_valid_imode(u32 imode)
int recow_extent_buffer(struct btrfs_root *root, struct extent_buffer *eb);
static inline u32 btrfs_type_to_imode(u8 type)
{
static u32 imode_by_btrfs_type[] = {
[BTRFS_FT_REG_FILE] = S_IFREG,
[BTRFS_FT_DIR] = S_IFDIR,
[BTRFS_FT_CHRDEV] = S_IFCHR,
[BTRFS_FT_BLKDEV] = S_IFBLK,
[BTRFS_FT_FIFO] = S_IFIFO,
[BTRFS_FT_SOCK] = S_IFSOCK,
[BTRFS_FT_SYMLINK] = S_IFLNK,
};
return imode_by_btrfs_type[(type)];
}
int get_extent_item_generation(u64 bytenr, u64 *gen_ret);
/*

View file

@ -22,6 +22,7 @@
#include "kerncompat.h"
#include <stdbool.h>
#include <stddef.h>
#include <sys/stat.h>
#include "kernel-lib/list.h"
#include "kernel-shared/volumes.h"
#include "common/fsfeatures.h"
@ -40,6 +41,21 @@ enum exclusive_operation {
BTRFS_EXCLOP_UNKNOWN = -1,
};
static inline u32 btrfs_type_to_imode(u8 type)
{
static u32 imode_by_btrfs_type[] = {
[BTRFS_FT_REG_FILE] = S_IFREG,
[BTRFS_FT_DIR] = S_IFDIR,
[BTRFS_FT_CHRDEV] = S_IFCHR,
[BTRFS_FT_BLKDEV] = S_IFBLK,
[BTRFS_FT_FIFO] = S_IFIFO,
[BTRFS_FT_SOCK] = S_IFSOCK,
[BTRFS_FT_SYMLINK] = S_IFLNK,
};
return imode_by_btrfs_type[(type)];
}
/* 2 for "0x", 2 for each byte, plus nul */
#define BTRFS_CSUM_STRING_LEN (2 + 2 * BTRFS_CSUM_SIZE + 1)
void btrfs_format_csum(u16 csum_type, const u8 *data, char *output);

View file

@ -857,6 +857,10 @@ static int ext2_copy_single_inode(struct btrfs_trans_handle *trans,
struct btrfs_key inode_key;
struct btrfs_path path = { 0 };
inode_key.objectid = objectid;
inode_key.type = BTRFS_INODE_ITEM_KEY;
inode_key.offset = 0;
if (ext2_inode->i_links_count == 0)
return 0;
@ -878,13 +882,23 @@ static int ext2_copy_single_inode(struct btrfs_trans_handle *trans,
ext2_convert_inode_flags(&btrfs_inode, ext2_inode);
/*
* The inode item must be inserted before any file extents/dir items/xattrs,
* or we may trigger tree-checker. File extents/dir items/xattrs require
* the previous item has the same key objectid.
* The inode may already be created (with dummy contents), in that
* case we don't need to do anything yet.
* The inode item would be updated at the end anyway.
*/
ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
if (ret < 0)
return ret;
ret = btrfs_lookup_inode(trans, root, &path, &inode_key, 1);
btrfs_release_path(&path);
if (ret > 0) {
/*
* No inode item yet, the inode item must be inserted before
* any file extents/dir items/xattrs, or we may trigger
* tree-checker. File extents/dir items/xattrs require the
* previous item to have the same key objectid.
*/
ret = btrfs_insert_inode(trans, root, objectid, &btrfs_inode);
if (ret < 0)
return ret;
}
switch (ext2_inode->i_mode & S_IFMT) {
case S_IFREG:
@ -917,10 +931,6 @@ static int ext2_copy_single_inode(struct btrfs_trans_handle *trans,
* Update the inode item, as above insert never updates the inode's
* nbytes and size.
*/
inode_key.objectid = objectid;
inode_key.type = BTRFS_INODE_ITEM_KEY;
inode_key.offset = 0;
ret = btrfs_lookup_inode(trans, root, &path, &inode_key, 1);
if (ret > 0)
ret = -ENOENT;

View file

@ -23,6 +23,8 @@
#include "kernel-shared/ctree.h"
#include "kernel-shared/disk-io.h"
#include "kernel-shared/volumes.h"
#include "kernel-shared/transaction.h"
#include "common/utils.h"
#include "common/internal.h"
#include "common/messages.h"
#include "common/extent-cache.h"
@ -183,6 +185,7 @@ int convert_insert_dirent(struct btrfs_trans_handle *trans,
{
int ret;
u64 inode_size;
struct btrfs_inode_item dummy_iitem = { 0 };
struct btrfs_key location = {
.objectid = objectid,
.offset = 0,
@ -193,6 +196,23 @@ int convert_insert_dirent(struct btrfs_trans_handle *trans,
dir, &location, file_type, index_cnt);
if (ret)
return ret;
btrfs_set_stack_inode_mode(&dummy_iitem, btrfs_type_to_imode(file_type));
btrfs_set_stack_inode_generation(&dummy_iitem, trans->transid);
btrfs_set_stack_inode_transid(&dummy_iitem, trans->transid);
/*
* We must have an INOTE_ITEM before INODE_REF, or tree-checker won't
* be happy.
* The content of the INODE_ITEM would be properly updated when iterating
* that child inode, but we should still try to make it as valid as
* possible, or we may still trigger some tree checker.
*/
ret = btrfs_insert_inode(trans, root, objectid, &dummy_iitem);
/* The inode item is already there, just skip it. */
if (ret == -EEXIST)
ret = 0;
if (ret < 0)
return ret;
ret = btrfs_insert_inode_ref(trans, root, name, name_len,
objectid, dir, index_cnt);
if (ret)