Subject: Re: anybody seen output lossage with 'de' driver?
To: John Birrell <jb@cimlogic.com.au>
From: Chris G Demetriou <Chris_G_Demetriou@ux2.sp.cs.cmu.edu>
List: port-alpha
Date: 10/15/1996 08:57:40
> 32 Mb RAM
>
> Rebooted this morning after the usual lockup; Have done lots of NFS client
> copying (NetBSD src tree from another machine to the AXPpci) since.
>
> netbsd2# pstat -s
> Device 1K-blocks Used Avail Capacity Type
> /dev/sd0b 55800 616 55184 1% Interleaved
>
> I bet if I don't run anything else (other than std cron stuff), it'll
> lock up by tomorrow morning.
Hmm.
Well, i'd suggest that I get a new kernel to you, and a copy of
savecore, so you could cause it to do a crash dump which you or I
could then look at...
However:
(1) all AXPpci systems that i've seen don't have "reset"
buttons (but then, i've only seen multias), and
(2) even if they did, right now there's no easy way to make
a machine at the monitor prompt do a crash dump...
I suppose I can work on (2), but unfortunately i've got a lot of other
active projects right now.
> Maybe I should whack the disk on the 3000-300X and see if it suffers
> the same fate?
That would be interesting to know, and could fix (1) above, to make
debugging easier if the problem happens there as well.
Also interesting (since you said you have no monitor attached, but
also no serial console) would be what would have shown up on the
monitor... Can you hook up a monitor, the next time it goes belly-up?
> > > > Could you send me (privately) your /etc/ttys?
> > >
> > > It's the same as the one you distribute in /usr/src/etc/etc.alpha/ttys.
> >
> > Which suggests that it should be edited to match the local machine's
> > configuration? 8-)
>
> It _does_ match the configuration, at least for those devices that are
> turned on (== console). Correct me if I'm wrong, but devices that are
> "off" are ignored, aren't they?!.
Yes, but there have been bugs where console devices being specified
incorrectly would cause the machine's console (and that's it) to hang
overnight. (Those bugs were on other ports, and, to the best of my
knowledge, have all been fixed... Finding such a problem on the alpha
port would surprise me, but not terribly so.)
chris