Subject: pmap panic, ed timeout, interrupts
To: None <>
From: Frank van der Linden <>
List: current-users
Date: 02/02/1994 23:19:34
Here at the U. of Amsterdam, we had (after they were not needed
for some project anymore) a small network of 6 386s and one 486
lying around. A friend and myself decided to 'hijack' these
machines for our own use.

Now, one of them is a 2 year old 486/33, with an 1.2gb ESDI drive +
Ultrastor Ultra-12 controller. It's got 64Mb. This one is going to
be the fileserver. And, since the minicluster is on a subnet, it
should also serve as router. So far, so good. Now, my friend
is more of a Linux person, so he installed it on the server.
We had some trouble in getting Linux to recognize 2 ethernet cards,
and, eventually, we did not get it to act as a gateway. I thought:
this is my chance to prove that NetBSD is so much more suited for
these kind of tasks :) I build a kernel with the right config,
and produced a boot floppy + root filesystem, enabling the
machine to at least act as a gateway. Unfortunately, it panics
at boottime. Somewhere after config, when it prints 'changing
root device to fd0a', it panics with 'panic: ptdi <hex number>'.
Booting from floppy, the kernel has no swap space, of course..
Could this be the problem? The GENERICAHA kernel from the 0.9
distribution boots fine, btw. Also, the kernel that does not boot
on the 486, works OK on one of the 386s (it does print a warning
about no swap space at about the time were the 486 panics, but
I believe that this is normal).

What could the problem be in pmap() ?

Secondly, when booting the kernel in question on a 386, after having
put the 2 ethernet cards into it, it keeps producing 'ed0: device timeout'.
ed1 works fine, though. The possible trouble here is, that com1 is
configured at irq3, just like ed0. Although both devices are recognized,
and the serial ports are not being used. Linux does not seem to
have any trouble with the devices on the same IRQ (haven't tested that
when actually using the serial port, though). Anyway: could this
be the problem with the 'ed0: device timeout' messages? I hear that
multiple devices sharing an irq is going to be possible in NetBSD
in the near future (it is already in 'magnum').

Thanks in advance,

- Frank