The Problem:
3/260 32M crashed several times over a couple of days with errors
similiar to the following:
> ECC Error Register d1<INTR,INTENA,CE_ENA,CE>
> Memory Error Register d1<INTR,INTENA,CHECK,ERR00> d1<INTR,INTENA,CE_ENA,CE>
> DVMA = 0, context = 4, virtual address = f064007
> pme = a3000032, physical address = 64007
> panic: unknown memory error
The Solution:
Presented by Kevin Thomas, Mark Morrissey, Sam Horrocks,
Dave Uraneck, and Greg Earle.
It is best summarized by Greg Earle, I include his reponse here
-------------------------------------------------------------------------------
These are correctable errors (CE). There is a bug in SunOS 4.0.3 that only
affects the Sun-3/260 and 3/280 series machines (the ones with ECC memory)
that causes the system to panic with that `panic: unknown memory error' message
every time there is a memory error, even if it is correctable and the system
should just say `I found one' or say nothing at all. Call the Sun US Answer
Center and request the 4.0.3 `bigmap' patch which fixes this and one other
bug in /usr/sys/sun3/OBJ/machdep.o.
A quick hack fix (for just the `unknown memory error' part) is
yourmachine# adb -w -k /vmunix /dev/mem
memerr+0x3a?w 0x7202
_memerr+0x3a: 7201 = 7202
memerr+0x3a/w 0x7202
_memerr+0x3a: 7201 = 7202
^D
yourmachine# adb -w /usr/sys/sun3/OBJ/machdep.o
memerr+0x3a?w 0x7202
_memerr+0x3a: 7201 = 7202
^D
You can trust me on this; I was the first person to discover this problem,
find the cause, and develop this patch. It works.
- Greg Earle
Sun Microsystems, Inc. - JPL on-site Software Support
-------------------------------------------------------------------------------
I installed the adb patch, and the system has been running fine since.
Thanks to all who responded...
Ron Vasey vasey@mcc.com
Kevin W. Thomas kwthomas@nsslsun.gcn.uoknor.edu
Mark Morrissey bit!markm@cse.ogi.edu
Sam Horrocks sam@zola.ICS.UCI.EDU
Rich Kaul kaul@icarus.eng.ohio-state.edu
Jeff Nieusma nieusma@eclipse.Colorado.EDU
Jim Ziobro JAMES_M._ZIOBRO.WBST102A@xerox.com
Dave Uraneck uunet!uranus!duran
Greg Earle earle@Sun.COM
and to those whose response is "in the mail".
--------------------------------------------------------------
William A. Cook | Image Aquisition Products Division
Mail Stop 02008 | Internet: bcook@Kodak.COM
Eastman Kodak Company |
Rochester, NY 14650-2008 |
This archive was generated by hypermail 2.1.2 : Fri Sep 28 2001 - 23:05:59 CDT