On Mon, Jun 25, 2001 at 10:09:15PM +0200, Manuel Bouyer wrote:
> On Sun, Jun 24, 2001 at 07:36:30PM -0600, Jim Bernard wrote:
> >   OK, I did this, and the resulting kernel has now been running for 4 days
> > without hanging, so I guess tagged queuing was the culprit in some way.
> > 
> >   I'm still seeing a few scsi parity errors during the execution of the
> > /etc/daily script, though.  (But there were none when I ran the script by
> > hand just after first booting the kernel.)
> 
> OK, so it's possible these parity errors degenerate in more serious SCSI
> problems when tagged queuing is enabled (highter load on the bus).
> It looks like this SCSI bus has a problem.

  Even though there has never been a scsi parity error with any older (than
early April) kernel?

  I tried tearing it apart, checking termination, etc., but nothing changed.
I haven't yet tried swapping out cables or swapping positions of hardware on
the bus.  That's largely because the kernel dependence seems so strongly
suggestive of a software problem.  But I could be persuaded to check further.

  BTW: I don't think the other person (Scott ...) who has been seeing
scsi-related hangs has been having problems with parity errors, so it's
by no means clear that the parity errors I'm seeing are at the root of the
hangs.  I.e., even if I do have a hardware problem, there does seem to be
at least one software problem as well.

  Thanks.

--Jim