From owner-freebsd-stable Thu Feb 5 07:17:26 1998 Return-Path: Received: (from majordom@localhost) by hub.freebsd.org (8.8.8/8.8.8) id HAA01875 for stable-outgoing; Thu, 5 Feb 1998 07:17:26 -0800 (PST) (envelope-from owner-freebsd-stable@FreeBSD.ORG) Received: from passer.osg.gov.bc.ca (passer.osg.gov.bc.ca [142.32.110.29]) by hub.freebsd.org (8.8.8/8.8.8) with ESMTP id HAA01868 for ; Thu, 5 Feb 1998 07:17:22 -0800 (PST) (envelope-from cy@cschuber.net.gov.bc.ca) Received: (from uucp@localhost) by passer.osg.gov.bc.ca (8.8.8/8.6.10) id HAA30820; Thu, 5 Feb 1998 07:17:15 -0800 (PST) Received: from cschuber.net.gov.bc.ca(142.31.240.113), claiming to be "cwsys.cwsent.com" via SMTP by passer.osg.gov.bc.ca, id smtpdaaCxya; Thu Feb 5 07:17:08 1998 Received: (from uucp@localhost) by cwsys.cwsent.com (8.8.8/8.6.10) id HAA04554; Thu, 5 Feb 1998 07:17:04 -0800 (PST) Message-Id: <199802051517.HAA04554@cwsys.cwsent.com> Received: from localhost.cwsent.com(127.0.0.1), claiming to be "cwsys" via SMTP by localhost.cwsent.com, id smtpd004545; Thu Feb 5 15:16:54 1998 X-Mailer: exmh version 2.0.1 12/23/97 Reply-to: Cy Schubert - ITSD Open Systems Group From: Cy Schubert - ITSD Open Systems Group X-Sender: cy To: "evanc-freebsd-stable@freebsd.org"@synapse.net cc: freebsd-stable@FreeBSD.ORG Subject: Re: Instant trap on make installworld In-reply-to: Your message of "01 Feb 1998 21:16:17 GMT." <19980201211617.2653.qmail@piano.synapse.net> Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Thu, 05 Feb 1998 07:16:50 -0800 Sender: owner-freebsd-stable@FreeBSD.ORG Precedence: bulk X-To-Unsubscribe: mail to majordomo@FreeBSD.org "unsubscribe stable" > (apologies if this goes out twice; I originally sent this this morning > but haven't seen it come through yet...) > > Configuration: > > 2.2.5-STABLE client mounts /usr/obj and /usr/src off a 2.2.5-STABLE server. > On the server, these directories are on /exports, and are loopback-mounted > to /usr/obj and /usr/src via NULLFS so that make won't get confused about > the pathname differences. > > The client's make world works great. The problem comes when I go to the > server, and try to make installworld based on the client's build. During > the first line of the make, the server panics: > > cd /usr/src && > PATH=/usr/obj/usr/src/tmp/sbin:/usr/obj/usr/src/tmp/usr/sbin:/usr/obj/usr/sr > c/tmp/bin:/usr/obj/usr/src/tmp/usr/bin > BISON_SIMPLE=/usr/obj/usr/src/tmp/usr/share/misc/bison.simple > COMPILER_PATH=/usr/obj/usr/src/tmp/usr/libexec:/usr/obj/usr/src/tmp/usr/bin > GCC_EXEC_PREFIX=/usr/obj/usr/src/tmp/usr/lib/ > LD_LIBRARY_PATH=/usr/obj/usr/src/tmp/usr/lib > LIBRARY_PATH=/usr/obj/usr/src/tmp/usr/lib:/usr/obj/usr/src/tmp/usr/lib > CC='cc -nostdinc' /usr/obj/usr/src/tmp/usr/bin/make reinstall > > Fatal trap 18: integer divide fault while in kernel mode > instruction pointer = 0x8:0xf01b00b8 > stack pointer = 0x10:0xefbffdfc > frame pointer = 0x10:0xefbffe48 > code segment = base 0x0, limit 0xfffff, type 0x1b > = DPL 0, pres 1, def32 1, gran 1 > processor eflags = interrupt enabled, resume, IOPL = 0 > current process = 196 (make) > interrupt mask = > panic: integer divide fault > > (IIRC, it isn't always this panic, but it always panics instantly...) > > The loopback filesystems seem to work fine for all other purposes, so I'm at > a loss to explain why the server fails. The only thing I can see is that it > points all the > > FYI, there are a couple reasons why I am not building this on the server. > The first is that I am forcing myself to treat that box like a NetApp or > somesuch -- it's just a big disk on the net which I'm not allowed to play > around with. The other is that for some reason if I do too much on that > box, it will also panic. I think that the particular -STABLE I put on there > originally isn't particularly stable :-) but as long as the box is left > alone, it runs just great as an NFS server. We've verified the RAM and are > quite sure it isn't a hardware fault. This panic is always something like: > > Fatal trap 12: page fault while in kernel mode > fault virtual address = 0x0 > fault code = supervisor read, page not present > instruction pointer = 0x8:0x0 > stack pointer = 0x10:0xefbffb48 > frame pointer = 0x10:0xefbffb84 > code segment = base 0x0, limit 0xfffff, type 0x1b > = DPL 0, pres 1, def32 1, gran 1 > processor eflags = interrupt enabled, resume, IOPL = 0 > current process = 1304 (install) > interrupt mask = > panic: page fault > > As a side note, is there a way to get this to work with symlinks instead of > NULLFS? > > Attached is my kernel config. The only "odd" thing is that I'm running the > latest dpt drivers for 2.2.5-STABLE (version 1.2.4). This box was installed > using January 22's 2.2.5-STABLE snap, and has never successfully been > updated, due to the above problems. > > Evan There is a bug in nullfs/umapfs or VM (depending on your point of view) which causes the panic you're describing when the mmap() system call is used. I put together a patch (kludge) and submitted it as a PR. This has caused a fair bit of discussion among various develpers and members of the Core Team. Bruce Evans subsequently sent me a patch, however this patch was designed for -current, which is too different in this area of VM to even attempt to retrofit it into -stable. I suspect that -current will fix this problem once the VM rewrite is complete. Regards, Phone: (250)387-8437 Cy Schubert Fax: (250)387-5766 UNIX Support OV/VM: BCSC02(CSCHUBER) ITSD BITNET: CSCHUBER@BCSC02.BITNET Government of BC Internet: cschuber@uumail.gov.bc.ca Cy.Schubert@gems8.gov.bc.ca