Not any more, no - the earlier kernel panics were similarly randomly
timed, perhaps the 3c59x driver formerly handled an APM-related
interrupt badly (see below).
> > makes me think there's an APM problem but (i) it's disabled in the
> > BIOS and (ii) the time at which the hangs occur is not consistent
>
> it sounds like an interrupt controller crash
I was wondering about that. The BIOS is supposed to do an API wake-up
(whatever you want to call it) on various interrupts. So, I've fiddled
with the BIOS APM settings, that is to say (i) enabled it in 'User
defined' mode, (ii) disabled the individual when-to-go-to-sleep times,
(iii) enabled the disc and ethernet card IRQ wake-ups and (iv) turned
off APM again. With the same kernel the machine has made it through
the night without any hang and the clock still says the right time. It
looks like this BIOS is remembering the individual APM options and
possibly implementing some of them even when APM is supposed to be
off. I had APM support in my kernel, still do, but dmesg reports 'APM
BIOS not found'...
The BIOS is Award v4.50PG (c) 1984-1995. Maybe I need a new one? Maybe
its interaction with the motherboard (dunno what brand) is just bad.
Now that I think of it this machine's clock always used to lose time
before, but never had any panics or lost its ethernet connectivity,
when it had a cheap ne2000 clone ethernet card. It never bothered me
before, it didn't seem related, I guess I forgot about it. Hey ho...
> I think you have a different (and _very_ weird) problem to the rest
Never mind my personal problems ;-)
I think the machine problems are sorted now. Thanks to all members of
this list who gave me inspiration and made me think, even if my
thinking was wrong I seem to have got things working with your help.
John. _ _
-- _ | |___| |_ _ _
John Robinson 46 Bank Street, Dumfries DG1 2PA, UK | |_| / . \ ' \| ' \
+44 1387 247249 \___/\___/_||_|_||_|
How I had to keep on trying, and get better every day... --Queen