Thanks for the quick response, Steve.
I'd be happy to tackle the problem if you have an idea which module(s)
changed. Maybe nfsd would have more luck detecting XFS than the other way
around.
Ken
----- Original Message -----
From: "Steve Lord" <lord@xxxxxxx>
To: "Ken Cross" <kcross@xxxxxxxxx>
Cc: "Linux XFS" <linux-xfs@xxxxxxxxxxx>
Sent: Wednesday, August 15, 2001 10:56 AM
Subject: Re: SPEC failures
> > I've been trying to run SPEC sfs tests, but during validation it fails
with:
> >
> > validating create file file_en.00000 ...
> > file_en.00000: create attribute mismatch
> > mode: returned = 100644, specified = 100666
> >
> > validating create file file_en.00001 ...
> > file_en.00001: create attribute mismatch
> > mode: returned = 100644, specified = 100666
> >
> > Later on in the validation pass, I get:
> >
> > validating setattr file_en.00000 ...
> > file_en.00000: setattr attribute mismatch
> > mode: returned = 100666, specified = 100644
> >
> > validating setattr file_en.00001 ...
> > file_en.00001: setattr attribute mismatch
> > mode: returned = 100666, specified = 100644
> >
> >
> > This is on a system with the latest (8/15/01) CVS code, using egcs
1.1.2.
> > It works fine using chmod manually, but SPEC sfs fails very consistently
> > with different clients and different hosts.
> >
> > It does *not* fail with the Linux-XFS 1.0.1 ISO image.
> >
> > Any ideas?
> >
> > Thanks,
> > Ken
> >
>
> Yes, Andrew Tridgell hit the same thing, this has to do with the code
> added to xfs for posix acl support and when we do and do not apply
> the process umask to file create permissions. When and how these
> masks are applied in the main kernel has changed since the 1.0.1
> release.
>
> Our problem right now is we need a way in the create code in xfs to
> tell we are being called from NFS, and apart from looking at the
> name of the running process I don't have a way of doing this.
>
> So basically the status is : broken until someone has a brainwave
> on how to detect a create/mkdir/mknod from nfs neatly.
>
> Steve
|