Sun 4/280, Exabyte 8200, Sun OS 4.1, Writeback error
cdp>From: Chris Peckham <cdp@hertz.njit.edu>
cdp>Date: Fri, 6 Sep 1991 17:35:50 -0400
cdp>To:sun-managers@eecs.nwu.edu
cdp>Subject: Sun 4/280, Exabyte 8200, Sun OS 4.1, Writeback error question
cdp>Cc:cdp@hertz.njit.edu,brw@hertz.njit.edu
cdp>Status: O
cdp>
cdp>We have a problem with our Sun 4/280.
cdp>
cdp>We are running Sun OS 4.1. The system has 96 MB of RAM. We just installed an
cdp>Exabyte 8200 8mm tape drive attached to the system through a SCSI3 board.
cdp>The only other device on the SCSI chain is a Sun CD-ROM reader. There are
cdp>no software patches currently installed on the system. This system has
cdp>several filesystems NFS exported.
cdp>
cdp>When we are in the middle of performing a dump to the 8mm drive, the following
cdp>error occurs and the machine crashes :-(
cdp>
cdp> Memory Error Register d4<INTR,INTENA,CE_ENA,WBACKERR>
cdp> DVMA = 0, context = 0, virtual address = fff5fbd8
cdp> pme = 0, physical address = 1bd8
cdp> panic: writeback error
cdp> syncing file systems... [2] 23 [2] 23
cdp> and so on...
cdp>
cdp>This has happened three times with basically the same error each time. The
cdp>first time that it happened, the following information was also given:
cdp>
cdp> Correctable Memory Error on Board 0,
cdp> at Physical Address 0x00C78390, Syndrome Register 00.
cdp>and so on....
and the answer,
From: "Fuat C. Baran" <fuat@cunixf.cc.columbia.edu>
From: John DiMarco <jdd@db.toronto.edu>
From: markets!keith@uunet.UU.NET (Keith Farrar)
From:shj@ultra.com (Steve Jay)
fuat> Check yout /etc/fstab file. Are you mentioning the config'ed swap
fuat> partition there? If so, remove it. Only mention additional swaps in
fuat> /etc/fstab. Otherwise, when you first swap (typically during a dump)
fuat> you'll crash.
jdd> This is a known problem with SunOS 4.1 and greater.
keith> 2. SunOS doesn't know better than to remount a swap partition declared
keith> in /etc/fstab if it's already mounted. I had the following entry in
keith> /etc/fstab: "/dev/xd0b /swap1 swap rw 0 0". This caused the kernel to
keith> apparently attempt to map xd0b into virtual memory a second time when
keith> /etc/rc called 'swapon -a'. The kernel automatically uses partition
keith> b on the boot drive as swap and swapon didn't check before using it
keith> again. Dump just happened to be the first program to get far enough
keith> out into virtual memory to trip over this mess.
Keith also mentioned that he had posted this solution before (the beginning of
August). Sorry I did not remember it/pay enough attention to my mail :-( .
The problem was that I had both swap partitions in /etc/fstab. We just lost
the root file system and I added the second swap partition to /etc/fstab when
I restored the root filesystem.
Thanks for the quick help -
Chris Peckham
Unix Systems Programmer
