On Mon, Jul 21, 2008 at 01:41:47PM +0200, Miklos Szeredi wrote:
> From: Miklos Szeredi <mszeredi@xxxxxxx>
>
> vfs_rename_dir() doesn't properly account for filesystems with
> FS_RENAME_DOES_D_MOVE. If new_dentry has a target inode attached, it
> unhashes the new_dentry prior to the rename() iop and rehashes it
> after, but doesn't account for the possibility that rename() may have
> swapped {old,new}_dentry. For FS_RENAME_DOES_D_MOVE filesystems, it
> rehashes new_dentry (now the old renamed-from name, which d_move()
> expected to go away), such that a subsequent lookup will find it.
>
> This was caught by the recently posted POSIX fstest suite, rename/10.t
> test 62 (and others) on ceph.
>
> Fix by not rehashing the new dentry. Rehashing would only make sense
> if the rename failed (which should happen extremely rarely), but we
> cannot handle that case correctly 100% of the time anyway, so...
Lovely. AFAICS, that's a fallout from
commit 349457ccf2592c14bdf13b6706170ae2e94931b1
Author: Mark Fasheh <mark.fasheh@xxxxxxxxxx>
Date: Fri Sep 8 14:22:21 2006 -0700
[PATCH] Allow file systems to manually d_move() inside of ->rename()
that had allowed that crap for directories. Note that d_rehash() used
to be needed (d_move() would unhash the source otherwise) and d_move()
used to be unconditional until the changeset above.
It's _probably_ OK now, but I'd really like to think about NFS behaviour.
There are subtle traps in that area.
BTW, failing rename() is trivial - just have a non-empty target...
--
To unsubscribe from this list: send the line "unsubscribe linux-fsdevel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at http://vger.kernel.org/majordomo-info.html
