Subject: bin/9972: kdump dumps core in ktrgenio function.
To: None <gnats-bugs@gnats.netbsd.org>
From: None <roskens@elfin.net>
List: netbsd-bugs
Date: 04/24/2000 10:02:13
>Number:         9972
>Category:       bin
>Synopsis:       kdump dumps core in ktrgenio function.
>Confidential:   no
>Severity:       non-critical
>Priority:       medium
>Responsible:    bin-bug-people
>State:          open
>Class:          sw-bug
>Submitter-Id:   net
>Arrival-Date:   Mon Apr 24 10:03:00 PDT 2000
>Closed-Date:
>Last-Modified:
>Originator:     Ron Roskens
>Release:        NetBSD-20000416
>Organization:

>Environment:
System: NetBSD pos 1.4X NetBSD 1.4X (POS_TOSHIBA) #4: Fri Apr 21 08:57:42 PDT 2000     root@pos:/usr/src/sys/arch/i386/compile/POS_TOSHIBA i386


>Description:
Header from kdump.c:
/*      $NetBSD: kdump.c,v 1.28 2000/04/10 09:13:45 jdolecek Exp $      */

While trying to track down a compilation problem I was using ktrace to see
why it wasn't opening a specific library. When I tried to run kdump against
the ktrace.out file, kdump would read through part of the file and then
dump core.

When I used the -m option to set maxdata (-m 2048), kdump ran without any
problems.

kdump cores in ktrgenio() at line 473. I think that dp is at the end of the
IO stream and vis() gets called to look beyond the end of it.

>How-To-Repeat:
download ftp://ftp.elfin.net/pub/roskens/kdump-core.tar.gz and run kdump on
an i386 against the ktrace.out file.  A debug kdump binary and the
resulting kdump.core file are included.
>Fix:

Unknown.
>Release-Note:
>Audit-Trail:
>Unformatted: