Subject: Re: port-i386/1701: my system reproducably crashes due to ncr problems which seem to be related to simultaneous ncr and ide disk access
To: None <netbsd-bugs@NetBSD.ORG>
From: None <tsm@cs.brown.edu>
List: netbsd-bugs
Date: 10/29/1995 14:05:36
More info: I figured I'd move my swap to one of the ide disks (wd2) to see if
that helped. So, I dumped wd2a, repartitioned, newfs'd and restored it, and
then even before switching the swap to use the new partition fsck -fp no longer
reproduces the problem, and it became harder to get it with X (although still
possible). I conclude there's some kind of timing problem and there was a
different allocation of the files before the dump/restore that gave different
timings. After switching the swap it doesn't seem possible to get the same
error with X (which stands to reason because nothing in the regular system or X
accesses sd0) but I haven't tried anything else that might reproduce it yet.
One person has already suggested disabling prefetch in the bios, so I should
note that it was already disabled (due to other problems) and I've tried
disabling just about every feature in the bios I can to no avail. It might also
be worth noting that when I did the restore I got 4 checksum errors from
restore on files which have nothing wrong with them at all. The checksum error
numbers are interesting themselves because they're so similar: 5244736, 4244736
(twice), and 2244736.  This may be of relevance because the tape drive I used
is a scsi2 4mm dat drive (which has not previously had any problems) which
blinks one of its lights if there's any problem reading the media. Since it
didn't blink its light, it makes me suspect a software problem, possibly due to
ide/ncr interaction.
   Tim