[OpenAFS-port-freebsd] Re: afs port

[Tom Maher]

Robert Watson <rwatson@FreeBSD.ORG> writes:

> On Thu, 29 Mar 2001, Garance A Drosihn wrote:
> 
> > On Feb 8/2001, George.Giles@mcmail.vanderbilt.edu wrote:
> > >Has the openAFS from IBM been ported to FreeBSD yet?
> > 
> > Did anyone reply to this yet?
> > 
> > There is a mailing list at the openafs site for the freebsd port, but
> > there hasn't been a lot of activity on it.  I know someone was looking
> > over the openafs code for freebsd, but I don't know what the current
> > status is. 
> 
> Last I heard, Tom Maher had a large (130kb+) patch to get the userland
> code working.  However, that doesn't address the issue of the kernel
> module, or whether the patch was complete (unlikely given the scale of the
> task).  With regards to the kernel module, the best starting point is
> probably to select the platform with the closest VFS layer to FreeBSD's,
> and work from there.  In my view, the highest priority is really getting
> the server-side working, given that Arla provides a quite workable AFS
> client on FreeBSD already, and is well-maintained.  Unfortunately, I have
> plenty on my plate already so am unlikely to make any forward progress in
> active development.  Given semi-working code, I'd be glad to help debug
> it and attempt to track nasty kernel bugs, of which I am sure there will
> be plenty :-).

You and your nagging memory, Rob.

Yeah, I've been off-and-on working on getting the kernelspace code to
work, basing it off of the Digital Unix code, the closest VFS layer.
The problem that's been keeping me from doing much more is lack of
understanding the buffer cache, which is completely different from the
Digital Unix buffer cache.

I should probably make what code I do have up on the web.  No, I haven't
tested it.

Rob is right, though, that the priority should be getting a working
server before getting a "real" client.  However, given the sheer ammount
of stuff that OpenAFS does in kernelspace rather than userland, I'm not
sure that the server doesn't depend on the kernel module.

-- 
Tom Maher