[linux-cifs-client] CIFS and rename of open files

[Jeff Layton]

On Tue, 22 Jul 2008 00:44:14 +0200
Wilhelm Meier <wilhelm.meier at fh-kl.de> wrote:

> Am Montag, 30. Juni 2008 schrieb Jeff Layton:
> > On Wed, 25 Jun 2008 09:52:23 +0200
> >
> > Wilhelm Meier <wilhelm.meier at fh-kl.de> wrote:
> > > Hello Jeff,
> > >
> > > Am Freitag, 6. Juni 2008 schrieb Wilhelm Meier:
> > > > Am Freitag, 6. Juni 2008 schrieb Jeff Layton:
> > > > > On Fri, 6 Jun 2008 10:49:26 +0200
> > > > >
> > > > > Wilhelm Meier <wilhelm.meier at fh-kl.de> wrote:
> > > > > > Hello,
> > > > > >
> > > > > > are there any news to the issue of renaming open files -
> > > > > > because this missing feature is preventing the use of cifs
> > > > > > as user homes on linux systems.
> > > > >
> > > > > I was looking at the problem earlier this week, but haven't
> > > > > gotten all the way to a solution. The issue is very dependent
> > > > > on server behavior, with windows servers being particularly
> > > > > problematic. I've been doing some experimentation, but
> > > > > haven't found a solution that seems to work well for all
> > > > > cases.
> > >
> > > is there any progess or activity from your side? I'm asking
> > > because this is a show-stopper using cifs-homes for users with
> > > newer kmail-versions.
> > > Or do you know any workaround in this respect?
> >
> > I'm afraid that I've been sidetracked into other things and haven't
> > had time to touch this. The problem is (I think) that CIFS
> > currently tries to set the DELETE_ON_CLOSE bit using the create
> > flags in the open call, but this only works if the file is actually
> > being created. So we need to set the bit with a SET_FILE_INFO call
> > and then try to rename it.
> >
> > I'm probably wrong on some of this though, this turns out to be a
> > rather tricky situation to handle correctly, and it will probably
> > take some experimentation to get it right. I don't know of any
> > workarounds for this at the moment...
> 
> Is there anything I can do in the combination of samba-server and 
> linux-cifs to prebent the problem?
> 

Not that I know of. Possibly there are some server settings that can
work around this, but I'm not certain.

> Do you need a testprogram? If so, I would try to write one simpler 
> than kmail ;-)
> 

A test program would be very nice when I (or someone else) gets an
opportunity to work on this. Something like kmail is going to be
far too complex to be useful.

Thanks,
-- 
Jeff Layton <jlayton at redhat.com>