From owner-freebsd-current  Sun Aug 24 14:29:43 1997
Return-Path: <owner-freebsd-current>
Received: (from root@localhost)
          by hub.freebsd.org (8.8.7/8.8.7) id OAA22620
          for current-outgoing; Sun, 24 Aug 1997 14:29:43 -0700 (PDT)
Received: from dyson.iquest.net (dyson.iquest.net [198.70.144.127])
          by hub.freebsd.org (8.8.7/8.8.7) with ESMTP id OAA22607
          for <current@FreeBSD.ORG>; Sun, 24 Aug 1997 14:29:38 -0700 (PDT)
Received: (from root@localhost)
	by dyson.iquest.net (8.8.6/8.8.5) id QAA05817;
	Sun, 24 Aug 1997 16:26:09 -0500 (EST)
From: "John S. Dyson" <toor@dyson.iquest.net>
Message-Id: <199708242126.QAA05817@dyson.iquest.net>
Subject: Re: UUCP and lpr services don't work anymore
In-Reply-To: <19970824220845.43690@klemm.gtn.com> from Andreas Klemm at "Aug 24, 97 10:08:45 pm"
To: andreas@klemm.gtn.com (Andreas Klemm)
Date: Sun, 24 Aug 1997 16:26:09 -0500 (EST)
Cc: roberto@keltia.freenix.fr, current@FreeBSD.ORG
X-Mailer: ELM [version 2.4ME+ PL31 (25)]
MIME-Version: 1.0
Content-Type: text/plain; charset=US-ASCII
Content-Transfer-Encoding: 7bit
Sender: owner-freebsd-current@FreeBSD.ORG
X-Loop: FreeBSD.org
Precedence: bulk

Andreas Klemm said:
> On Sun, Aug 24, 1997 at 07:03:36PM +0200, Ollivier Robert wrote:
> > According to Andreas Klemm:
> > > Outgoing mails are mailed via UUCP. The D files get spooled
> > > ok in the D. directory, but the control files in the C. 
> > > subdir (UUCP spool dir), has the size of 0 bytes although
> > > the vfilesystem has pleanty of Mbytes free.
> >
> > I've not seen this yet on my home machine, still at July, 20th current (but
> > I'm worried, all my mail is through UUCP).
> 
> If you are not using FreeBSD-SMP then everything should be ok.
> I can send mail now using the non-SMP kernel.
> 
> So some of the changes in SMP code are causing this ...
> 
The bug that I caused in the VFS code was kind-of non-deterministic (or
seemed so.)   The problems with file truncation were not the SMP code,
even though timing issues might make the problem worse.

We still have a problem with the VFS layer (I haven't manifested it yet.)  I
am doing a careful code review, and found some places in the filesystem code
where there are some dependencies on the v_usecount field of the vnode, and
I think the current problem that I caused is somewhere around there.

John