New kernel old nfs-utils

J. Bruce Fields bfields at fieldses.org
Tue Jan 29 17:19:22 EST 2008


On Tue, Jan 29, 2008 at 04:54:38PM -0500, Dave Quigley wrote:
> 
> On Tue, 2008-01-29 at 17:06 -0500, J. Bruce Fields wrote:
> > On Tue, Jan 29, 2008 at 04:43:04PM -0500, Dave Quigley wrote:
> > > On newer nfs-utils sec=flavor seems to be valid in the exports file. In
> > > 1.0.9 it doesn't seem to be the case. It is telling me that sec isn't a
> > > valid keyword. It could be that it is failing on the value it is set to
> > > but if that is the case then the error message is completely wrong for
> > > that kind of failure.
> > 
> > Oh, right.  The sec= thing is a new feature that requires both a new
> > nfs-utils and a new kernel to work.   So on an older setup you need to
> > do:
> > 
> > 	/exports	gss/krb5(rw)
> > 	/exports	gss/krb5i(rw)
> > 
> > intead of
> > 
> > 	/exports	*(rw,sec=krb5:krb5i)
> > 
> > --b.
> so if we made a new auth method called seclabel and we wanted both unix
> and seclabel would I do:
> 
> /exports seclabel(rw) 
> /exports unix(rw)

Well, more likely

/exports seclabel(rw)
/exports *(rw)

The "unix" is implicit in the second one.

And you'd probably want to choose something other than
"seclabel"--something less likely to be confused with a host name or
netgroup.

It may be possible to support the old syntax, but I wouldn't recommend
it; just use the new.  But, yes, that does mean you'll need a newer
nfs-utils and kernel both.

--b.


More information about the NFSv4 mailing list