Re: [PATCH] ocfs2: validate inline xattr header before listxattr walks it

On Wed, Apr 29, 2026 at 10:44 AM ZhengYuan Huang <[email protected]> wrote:
>
> [BUG]
> listxattr() can walk off the end of a dinode block when the inode-body
> xattr header is corrupted:
>
> BUG: KASAN: use-after-free in ocfs2_xattr_list_entry+0x1bd/0x370 fs/ocfs2/xattr.c:918
> Read of size 13 at addr ffff88800ab5c0c0 by task syz.0.1231/3756
>
> Call Trace:
>  ...
>  ocfs2_xattr_list_entry+0x1bd/0x370 fs/ocfs2/xattr.c:918
>  ocfs2_xattr_list_entries+0x1e1/0x320 fs/ocfs2/xattr.c:938
>  ocfs2_xattr_ibody_list fs/ocfs2/xattr.c:982 [inline]
>  ocfs2_listxattr+0x4fb/0x980 fs/ocfs2/xattr.c:1044
>  vfs_listxattr+0xb4/0x120 fs/xattr.c:493
>  listxattr+0x76/0x170 fs/xattr.c:924
>  filename_listxattr fs/xattr.c:958 [inline]
>  path_listxattrat+0x137/0x320 fs/xattr.c:988
>  __do_sys_listxattr fs/xattr.c:1001 [inline]
>  ...
>
> [CAUSE]
> ocfs2_xattr_ibody_list() computes the inline xattr header from
> di->i_xattr_inline_size and passes it straight to
> ocfs2_xattr_list_entries(). If corruption inflates xh_count, the list
> walk steps past the inline xattr area and eventually past the end of the
> 4K dinode block. The xe_type load in ocfs2_xattr_get_type() then reads
> poisoned memory.
>
> [FIX]
> Validate di->i_xattr_inline_size before locating the header, then bound
> xh_count by the number of ocfs2_xattr_entry records that fit inside the
> claimed inline area. Reject corrupted metadata with ocfs2_error() and
> -EFSCORRUPTED instead of iterating past the dinode block.
>
> Signed-off-by: ZhengYuan Huang <[email protected]>
> ---
>  fs/ocfs2/xattr.c | 22 ++++++++++++++++++++++
>  1 file changed, 22 insertions(+)
>
> diff --git a/fs/ocfs2/xattr.c b/fs/ocfs2/xattr.c
> --- a/fs/ocfs2/xattr.c
> +++ b/fs/ocfs2/xattr.c
> @@
>  static int ocfs2_xattr_ibody_list(struct inode *inode,
>                                   struct ocfs2_dinode *di,
>                                   char *buffer,
>                                   size_t buffer_size)
>  {
>         struct ocfs2_xattr_header *header = NULL;
>         struct ocfs2_inode_info *oi = OCFS2_I(inode);
>         int ret = 0;
> +       u16 xattr_count;
> +       size_t max_entries;
> +       u16 inline_size;
>
>         if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
>                 return ret;
>
> +       inline_size = le16_to_cpu(di->i_xattr_inline_size);
> +
> +       if (inline_size > inode->i_sb->s_blocksize ||
> +           inline_size < sizeof(struct ocfs2_xattr_header)) {
> +               ocfs2_error(inode->i_sb,
> +                           "Invalid xattr inline size %u in inode %llu\n",
> +                           inline_size,
> +                           (unsigned long long)OCFS2_I(inode)->ip_blkno);
> +               return -EFSCORRUPTED;
> +       }
> +
>         header = (struct ocfs2_xattr_header *)
>                  ((void *)di + inode->i_sb->s_blocksize -
> -                le16_to_cpu(di->i_xattr_inline_size));
> +                inline_size);
> +
> +       xattr_count = le16_to_cpu(header->xh_count);
> +       max_entries = (inline_size - sizeof(struct ocfs2_xattr_header)) /
> +                     sizeof(struct ocfs2_xattr_entry);
> +       if (xattr_count > max_entries) {
> +               ocfs2_error(inode->i_sb,
> +                           "Invalid xattr entry count %u (max %zu) in inode %llu\n",
> +                           xattr_count, max_entries,
> +                           (unsigned long long)OCFS2_I(inode)->ip_blkno);
> +               return -EFSCORRUPTED;
> +       }
>
>         ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size);
>
>         return ret;
>  }
> --
> 2.43.0

This bug has already been fixed by commit 2a4f33430e96. Sorry for the noise.

Thanks,
ZhengYuan Huang