Subject: Re: FFS journal
To: None <>
From: Bill Studenmund <>
List: tech-kern
Date: 07/24/2006 12:57:25
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
Content-Transfer-Encoding: quoted-printable

On Sat, Jul 22, 2006 at 09:37:38AM +0200, Bernd Ernesti wrote:
> On Tue, Jul 11, 2006 at 10:59:27AM -0700, Bill Studenmund wrote:
> > On Tue, Jul 11, 2006 at 04:25:17PM +0200, Pawel Jakub Dawidek wrote:
> [..]
> > > How do you handle situations when process which removed open file is
> > > chrooted and .deleted/ is no accesable from within his context?
> >=20
> > My thought was that ".deleted" is never accessed by paths. It is only=
> > accessed/maintained by the file system and fsck when directly accesing =
> > disk; VOP_LOOKUP() never returns it. Or it never returns it for accesse=
> > that aren't ffs-internal.
> And what will happen if someone tries to create a .deleted file or direct=
> in such a filesystem?
> Right now that is an allowed file or directory name.

Please read the rest of the thread.

There is no real need for the ".deleted" directory to actually be in the=20
name hierarcy. Thus the presence or absence of a real ".deleted" directory=
won't matter.

In fact, life is easier if the ".deleted" directory isn't in the real=20
hierarcy, as we then don't need to add hooks to isolate it. Adding an=20
inode field to the superblock would let us find the directory.

If we decide we want the directory to be in the namespace, then we will=20
just have to pick a name, and users can't touch that name. The extended=20
attribute stuff is all in ROOT/.attribute so there is precedent for=20
putting these things in the namespace. I'd rather not do that, though, as=
we then have to add hooks to keep users out of the directory.

Take care,


Content-Type: application/pgp-signature
Content-Disposition: inline

Version: GnuPG v1.2.3 (NetBSD)