[OpenAFS] Re: /afs mounts RO on kernel 3.1.0-1.1

Andrew Deason adeason@sinenomine.net
Tue, 15 Nov 2011 13:13:45 -0600

On Tue, 15 Nov 2011 12:27:21 -0600
Andrew Deason <adeason@sinenomine.net> wrote:

> I think this involves mountpoints with an explicit local cell name.
> (That is, 'cell:foo' rather than 'foo' with local cell 'cell'.) I
> don't see a difference myself just between Linux 3.1 and 2.6, but I'm
> pretty sure at least different OpenAFS client versions handle this
> differently.  Sometimes mountpoints like that are treated like foreign
> mounts, where the algorithm for determining RO-ness is reset.
> I don't think it's supposed to do that, so while it is a bug, I
> thought it wasn't new with 1.6 or Linux 3.*.

Nevermind my "this was already broken" comments; I was confusing this
with -fakestat/-fakestat-all behavior, which does currently treat local
explicit-cell mountpoints as "foreign". The RO path reset thing I think
is new.

Andrew Deason