NetBSD Problem Report #3183
Received: (qmail-queue invoked from smtpd); 2 Feb 1997 17:52:14 -0000
Message-Id: <199702021748.MAA05880@anduin.eldar.org>
Date: Sun, 2 Feb 1997 12:48:19 -0500 (EST)
From: Brad Spencer <brad@anduin.eldar.org>
Reply-To: brad@anduin.eldar.org
To: gnats-bugs@gnats.netbsd.org
Subject: fatal kernel trap
X-Send-Pr-Version: 3.95
>Number: 3183
>Category: port-alpha
>Synopsis: fatal kernel trap
>Confidential: no
>Severity: critical
>Priority: high
>Responsible: gnats-admin
>State: closed
>Class: sw-bug
>Submitter-Id: net
>Arrival-Date: Sun Feb 02 10:05:00 +0000 1997
>Closed-Date: Thu Apr 10 01:51:42 +0000 1997
>Last-Modified: Thu Apr 10 01:52:06 +0000 1997
>Originator: Brad Spencer
>Release: -current supped and compiled on 2/1/97
>Organization:
Sitting at home
>Environment:
NetBSD valinor.eldar.org 1.2C NetBSD 1.2C (VALINOR) #1: Sun Feb 2 09:54:49 EST 1997 brad@valinor.eldar.org:/usr/src/sys/arch/alpha/compile/VALINOR alpha
>Description:
Wanted to get NCSA's http daemon compiled up for my Multia. The
compile went fine, after compiling without optimization [which
appeared to cause the compiler and friends to go into a infinite
loop], but when the daemon is contacted the following panic occures:
fatal kernel trap:
trap entry = 0x4 (unaligned access fault)
a0 = 0xfffffe004a5f472c
a1 = 0x2d
a2 = 0x4
pc = 0xfffffc0000351ad4
ra = 0xfffffc00003507dc
curproc = 0xfffffe004a5eae00
pid = 144, comm = httpd
panic: trap
The machine, in question, does not have a monitor, so it is unclear if
there was more information printed before the reboot.
A crash dump was made, but it was unclear how one might go about
debugging with it, as the alpha gdb does not appear to deal with the
idea of kernel debugging.
>How-To-Repeat:
Acquire NCSA's http daemon, version 1.4.2 [ya, I know this is probably
old], do a 'make netbsd' after making the target not compile with
optimzation.
>Fix:
Unknown...
>Release-Note:
>Audit-Trail:
From: "Chris G. Demetriou" <cgd@cs.cmu.edu>
To: brad@anduin.eldar.org
Cc: gnats-bugs@gnats.netbsd.org, netbsd-bugs@netbsd.org,
cgd@auchentoshan.pdl.cs.cmu.edu
Subject: Re: port-alpha/3183: fatal kernel trap
Date: Sun, 02 Feb 1997 13:22:02 -0500
> fatal kernel trap:
>
> trap entry = 0x4 (unaligned access fault)
> a0 = 0xfffffe004a5f472c
> a1 = 0x2d
> a2 = 0x4
> pc = 0xfffffc0000351ad4
> ra = 0xfffffc00003507dc
> curproc = 0xfffffe004a5eae00
> pid = 144, comm = httpd
>
> panic: trap
"oh cool!" I've not seen an unaligned access fault in the kernel for
... a long time. 8-S
could you run gdb on the kernel binary with debugging symbols that
corresponds to the kernel you're running? (When your kernel was
built, a 'netbsd.gdb' was built at the same time as your 'netbsd'.)
In gdb, 'list *0xfffffc0000351ad4' and 'list *0xfffffc00003507dc', and
reply to this message with the results.
This is probably pretty repeatable, so you can probably cause the same
result on a new kernel if you need to build one from scratch to get a
'reasonable' (i.e. fitting your sources and kernel binary) netbsd.gdb.
If you have to build a new kernel and make it crash (so you have an
up-to-date netbsd.gdb), use the pc and ra values reported by the fatal
trap in those 'list' commands.
chris
From: "Chris G. Demetriou" <cgd@cs.cmu.edu>
To: brad@anduin.eldar.org
Cc: gnats-bugs@gnats.netbsd.org, netbsd-bugs@netbsd.org,
cgd@auchentoshan.pdl.cs.cmu.edu
Subject: Re: port-alpha/3183: fatal kernel trap
Date: Sun, 02 Feb 1997 13:27:53 -0500
> "oh cool!" I've not seen an unaligned access fault in the kernel for
> ... a long time. 8-S
>
> [ ... ]
Alternately, if the bug is as repeatable as i think it is, it should
be easily triggered by you sending me the mosaic binary, and me
running it on one of my systems.
If you'd like to do that, send me e-mail privately, and i'll tell you
where to put a copy of the binary.
cgd
From: Brad Spencer <brad@anduin.eldar.org>
To: cgd@cs.cmu.edu
Cc: gnats-bugs@gnats.netbsd.org, netbsd-bugs@netbsd.org,
cgd@auchentoshan.pdl.cs.cmu.edu
Subject: Re: port-alpha/3183: fatal kernel trap
Date: Sun, 2 Feb 1997 13:33:17 -0500 (EST)
> fatal kernel trap:
>
> trap entry = 0x4 (unaligned access fault)
> a0 = 0xfffffe004a5f472c
> a1 = 0x2d
> a2 = 0x4
> pc = 0xfffffc0000351ad4
> ra = 0xfffffc00003507dc
> curproc = 0xfffffe004a5eae00
> pid = 144, comm = httpd
>
> panic: trap
"oh cool!" I've not seen an unaligned access fault in the kernel for
... a long time. 8-S
could you run gdb on the kernel binary with debugging symbols that
corresponds to the kernel you're running? (When your kernel was
built, a 'netbsd.gdb' was built at the same time as your 'netbsd'.)
Yes, not a problem.....
In gdb, 'list *0xfffffc0000351ad4' and 'list *0xfffffc00003507dc', and
reply to this message with the results.
(gdb) list *0xfffffc0000351ad4
0xfffffc0000351ad4 is in unp_internalize (../../../../kern/uipc_usrreq.c:682).
677 }
678 rp = (struct file **)(cm + 1);
679 for (i = 0; i < oldfds; i++) {
680 fp = fdp->fd_ofiles[*(int *)rp];
681 *rp++ = fp;
682 fp->f_count++;
683 fp->f_msgcount++;
684 unp_rights++;
685 }
686 return (0);
(gdb) list *0xfffffc00003507dc
0xfffffc00003507dc is in uipc_usrreq (../../../../kern/uipc_usrreq.c:219).
214 panic("uipc 2");
215 }
216 break;
217
218 case PRU_SEND:
219 if (control && (error = unp_internalize(control, p)))
220 break;
221 switch (so->so_type) {
222
223 case SOCK_DGRAM: {
Hopefully, that is helpful...
This is probably pretty repeatable, so you can probably cause the same
result on a new kernel if you need to build one from scratch to get a
'reasonable' (i.e. fitting your sources and kernel binary) netbsd.gdb.
It is completely repeatable...
If you have to build a new kernel and make it crash (so you have an
up-to-date netbsd.gdb), use the pc and ra values reported by the fatal
trap in those 'list' commands.
I actually found the panic on a system build over a month ago, so the
first thing I did was sup and build a new system. Of course, the same
thing happened.
chris
Brad Spencer - brad@anduin.eldar.org http://anduin.eldar.org
From: "Chris G. Demetriou" <cgd@cs.cmu.edu>
To: Brad Spencer <brad@anduin.eldar.org>
Cc: cgd@cs.cmu.edu, gnats-bugs@gnats.netbsd.org, netbsd-bugs@netbsd.org
Subject: Re: port-alpha/3183: fatal kernel trap
Date: Sun, 02 Feb 1997 13:52:15 -0500
> [ ... ]
>
> Hopefully, that is helpful...
Yup.
A long time ago, when i was working in my office rather than the lab
that i typically work in now, i wrote on my white board that i thought
that passing file descriptors over local-domain sockets might be
buggy.
"whee! I guess i was right!"
This won't be a trivial fix, unfortunately. hopefully i'll kill it
today, though.
cgd
State-Changed-From-To: open->feedback
State-Changed-By: cgd
State-Changed-When: Sun Feb 2 12:44:00 PST 1997
State-Changed-Why:
a patch which appears to fix the problem has been sent to the user
but needs code review before it can be checked into the master sources.
State-Changed-From-To: feedback->closed
State-Changed-By: cgd
State-Changed-When: Wed Apr 9 18:51:42 PDT 1997
State-Changed-Why:
aforementioned patch has been checked in.
>Unformatted:
(Contact us)
$NetBSD: query-full-pr,v 1.51 2026/08/10 02:28:17 riastradh Exp $
$NetBSD: gnats_config.sh,v 1.10 2026/05/13 22:00:09 riastradh Exp $
Copyright © 1994-2026
The NetBSD Foundation, Inc. ALL RIGHTS RESERVED.