From owner-freebsd-current Tue Sep 12 18:00:13 1995 Return-Path: current-owner Received: (from majordom@localhost) by freefall.freebsd.org (8.6.12/8.6.6) id SAA19124 for current-outgoing; Tue, 12 Sep 1995 18:00:13 -0700 Received: from Root.COM (implode.Root.COM [198.145.90.17]) by freefall.freebsd.org (8.6.12/8.6.6) with ESMTP id SAA19116 for ; Tue, 12 Sep 1995 18:00:12 -0700 Received: from corbin.Root.COM (corbin [198.145.90.34]) by Root.COM (8.6.12/8.6.5) with ESMTP id RAA10199; Tue, 12 Sep 1995 17:59:01 -0700 Received: from localhost (localhost [127.0.0.1]) by corbin.Root.COM (8.6.12/8.6.5) with SMTP id SAA00136; Tue, 12 Sep 1995 18:01:15 -0700 Message-Id: <199509130101.SAA00136@corbin.Root.COM> To: "Rodney W. Grimes" cc: terry@lambert.org, current@freebsd.org Subject: Re: Potentially serious NFS problem In-reply-to: Your message of "Tue, 12 Sep 95 09:58:33 PDT." <199509121658.JAA03776@GndRsh.aac.dev.com> From: David Greenman Reply-To: davidg@Root.COM Date: Tue, 12 Sep 1995 18:01:14 -0700 Sender: current-owner@freebsd.org Precedence: bulk >While your in there poking around see if why when you kill -1 mountd >on the server if a client is doing I/O at that time it causes an >error to be returned and causes what ever was running to die. > >This is a sick bug, it makes FreeBSD unsutable for production NFS >server applications as adding or removing or changing the /etc/exports >file and hupping mountd is a quite standard procedure. It's caused by the need to replace the kernel export list whenever a change is made to the exports file. While the kernel export list is being rebuilt, there is a window where NFS I/O will fail if the appropriate entry hasn't yet been restored. I think the main reason the code is written this way has to do with the way the table is ordered, but I may be mistaken. It's a design flaw that might not be easily fixed. I've known about this for about 6 months, but I seem to never get enough of a break to do anything about it. -DG