[linux-cifs-client] Re: new regression from cifs open on lookup patches

[Jeff Layton]

On Sat, 23 May 2009 11:10:09 -0500
Steve French <smfrench at gmail.com> wrote:

> On Sat, May 23, 2009 at 11:02 AM, Christoph Hellwig <hch at infradead.org> wrote:
> > On Fri, May 22, 2009 at 10:59:10AM -0400, Jeff Layton wrote:
> >> At this point, we need to consider what to do for 2.6.30. We're looking
> >> at the 3rd regression from these patches. I'm strongly of the opinion
> >> that committing them was premature, and that they should be pulled and
> >> deferred until they are better tested.
> >>
> >> Open on lookup is a nice optimization but ultimately this is something
> >> that we can live without until it's working correctly.
> >
> > Personally I would prefer this to be reverted.  The VFS intent support in
> > it's current form is a really fragile mess and the best thing would be
> > to wait for Al to get this sorted out.
> I am glad that the intent support is being straightened out.
> but isn't the use of these flags for (only) create the same path that
> NFS has used for years.  This seems safest (and in this patch series
> and my testing was area that had few problems).
> 

NFS has used these flags, but the whole concept of lookup/open intents
is pretty ugly. A better scheme for this sort of thing would be welcome.

> 
> I am not sure that it can be as easily reverted since the patch series
> includes changes to three other areas which call cifs_posix_open and
> cifs_posix_open itself
> 

It certainly could be reverted if we choose to do so. I've been
regularly reverting these patches in order to test other things since
they've been committed.

-- 
Jeff Layton <jlayton at redhat.com>