Date: Sun, 15 Sep 2002 14:48:42 -0400 (EDT) From: Robert Watson <rwatson@freebsd.org> To: Kris Kennaway <kris@obsecurity.org> Cc: "Andrew P. Lentvorski" <bsder@mail.allcaps.org>, Matthew Jacob <mjacob@feral.com>, current@freebsd.org Subject: Re: NFS lock failure (was Re: FreeBSD 5.0 as a desktop 'failure' report) Message-ID: <Pine.NEB.3.96L.1020915144701.10061A-100000@fledge.watson.org> In-Reply-To: <20020915012837.GA943@xor.obsecurity.org>
next in thread | previous in thread | raw e-mail | index | archive | help
On Sat, 14 Sep 2002, Kris Kennaway wrote: > On Sat, Sep 14, 2002 at 05:51:32PM -0700, Andrew P. Lentvorski wrote: > > Folks, > > > > Could someone *please* take a look at the NFS lock packets *on the wire*? > > Without this, it becomes significantly harder to debug the failure. > > Well, there's also the apparent problem of a non-blocking flock() > operation blocking on the named pipe..but thanks for your suggestions, > I'll go and try them out now. The way the NFS locking code currently does an upcall to userland probably needs to be re-done before release. Right now it does some pretty funky things when used with chroot, for example, since the lookup of the named pipe is relative to the root of the process performing the lock operation, not necessarily the root that actually has the lockd process. Alfred and I have talked a little about re-write options, but my feeling is that using a pseudo-device would probably be best, even though it introduces a little more machinery. It would also avoid some problems with the use of inappropriate credentials in the lookup, etc. Robert N M Watson FreeBSD Core Team, TrustedBSD Projects robert@fledge.watson.org Network Associates Laboratories To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-current" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.NEB.3.96L.1020915144701.10061A-100000>