> I guess that spontaneous reboots are mostly caused by kernel stack
> overflow. As it seems to be pretty widespread problem, it might be
> nice to produce patch to detect such overflows (and panic() machine so
> that messages are not lost).
>
> [With one version of kernel, all you had to do to reboot the machine
> was to press t and <TAB> twice in bash$ prompt ;-)]
you might want to give ktrace a go, it detects stack overflows as well.
Alternatively you also might want to try EGCS's --check-stack, i dont know
what it does, it might need some tweaking to suit kernel needs.
-- mingo