I must admit that I'm not very much further with this, but on
one or two points I can, I think, make a useful comment...

On Fri, 29 Oct 2004 [log in to unmask] wrote:

> Quoting "Alan J. Flavell" <[log in to unmask]>:
>
> > As I watched it, I finally realised that the LED was flashing Morse
> > Code, and I worked out that it is flashing out the words FATAL
> > EXCEPTION.

> The "morse code" indicated a kernel panic (fatal OS crash).  You
> might be able to obtain more information from examination of the
> system message logs (/var/log/messages).

No, this is evidently far too deep in the entrails of the kernel
to get into the message log.

My mistake, I now realise, was attempting to play with this from Gnome
windows. I should have logged on at a character mode console.

If I do that, then the outcome when things go wrong is a screenfull of
debugging stuff which ends with the lines:

 Kernel panic: Fatal exception
 In interrupt handler - not syncing

If magic sysrq is enabled, then it's possible to get some low-level
information displayed (which, however, I have to admit I don't really
understand).  It's also possible to reboot via magic sysrq without
having to resort to power cycling, although - as the message indeed
implies - it's not possible to sync the filesystem first.


One suggestion was that the problem might be associated with ACPI.
I've added "acpi=off" to the GRUB boot parameters, but it seems to
have had no particular effect on the matter.  I might remark, by the
way, that on this model of laptop (IBM R40e) it's impossible to boot
Knoppix successfully without saying acpi=off; whereas booting SL303
seems to go without incident without this parameter, and without the
log messages showing any kind of distress - nor even mention of acpi,
beyond a couple of lines in the "BIOS-provided physical RAM map" which
mention "ACPI data" and "ACPI NVS".

So I think ACPI might be a red-herring.


I should stress that I -have- got the wireless card working on
occasion, and, once it's working, it seems to stay working. These
crashes seem to be somewhere in the procedure that's configuring and
starting-up the interface.