Without this check, the following XFS_I invocations would return bad
pointers when used on non-XFS inodes (perhaps pointers into preceding
allocator chunks).
This could be used by an attacker to trick xfs_swap_extents into
performing locking operations on attacker-chosen structures in kernel
memory, potentially leading to code execution in the kernel. (I have
not investigated how likely this is to be usable for an attack in
practice.)
Signed-off-by: Jann Horn <[email protected]>
Cc: Andy Lutomirski <[email protected]>
Cc: Dave Chinner <[email protected]>
Signed-off-by: Linus Torvalds <[email protected]>
goto out_put_tmp_file;
}
+ if (f.file->f_op != &xfs_file_operations ||
+ tmp.file->f_op != &xfs_file_operations) {
+ error = -EINVAL;
+ goto out_put_tmp_file;
+ }
+
ip = XFS_I(file_inode(f.file));
tip = XFS_I(file_inode(tmp.file));