Skip site navigation (1)Skip section navigation (2)
Date:      Sat, 2 Oct 2004 03:05:50 -0400 (EDT)
From:      "David E. Cross" <crossd@cs.rpi.edu>
To:        "Jason C. Wells" <jcw@highperformance.net>
Cc:        port-freebsd@openafs.org
Subject:   Re: FreeBSD 5.2.1 Client Success!
Message-ID:  <20041002030159.A84436@monica.cs.rpi.edu>
In-Reply-To: <F99BE7989AEABA2B6ABCEF48@[192.168.1.16]>
References:  <66F8A41B3F4164470D9446F2@[192.168.1.16]> <F99BE7989AEABA2B6ABCEF48@[192.168.1.16]>

next in thread | previous in thread | raw e-mail | index | archive | help


On Fri, 1 Oct 2004, Jason C. Wells wrote:

> --On Friday, October 01, 2004 8:21 PM -0700 "Jason C. Wells"
> <jcw@highperformance.net> wrote:
>
> > I have been bugging folks for FreeBSD status reports for a while now.  It
> > is my turn to make one.  I have a working FreeBSD OpenAFS 1.3.71 client.
> > I plan to do some stress testing to see if it pukes.
>
> Yeah, it puked. :|
>
> Writing from and to AFS even simultaneously worked fine.  Deleting was
> problematic.  When running OpenAFS 1.3.71 on FreeBSD 5.2.1 an attempt to
> delete an AFS file results in:
>
> 	panic: lockmgr:  locking against myself
>
> If anyone is working on the port anymore, I will be glad to work this issue
> in greater detail.  I don't have the skill to hack a fix.


Ugh, its been awhile since I've played with this, or the sections of the
VFS that AFS likes to (ab)use.  As I recall many of the lockmgr errors are
the fallout of AFS having its own vnodes.  If you can get a stack trace
from that, or a debugging output of what it was trying to lock, it could
be helpful.

It _used_ to be that AFS kept an artificially incremente VNODE count to
prevent "its" vnodes from re-entering the system pool, there were more
than a couple of places that this cause various OS assumptions to be
false.  Though I believe this was replaced with a VNODE flag that
basically indicated that the VNODE in question was "foreign" and told the
kernel to keep its hands off; its possible this abstraction broke down at
some point either in the FreeBSD or OpenAFS side of the equation...
aforementioned stack trace/dump would be instrumental in pinning that
down.

-- 
David E. Cross



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20041002030159.A84436>