> > I had a kernel once that I used for 4 months or so and never saw the
> > bug. But the kernels before and after that would hang fairly reliably. 
> 
> Hrm. Any thoughts and what options were or weren't in that kernel?
> Or was it code changes, you think? Any idea of the source date?

I assumed it had to do with how big the kernel was actually, since I
know there have been other bugs where large kernels overwrote OF data or
something.

> Do you know if there's any way to provoke the problem and get a
> useful crash dump out of it? Or, at least, a ddb prompt?

Oh, I remember now. You have a ddb prompt. You just can't see it or type
into it, because X is running. Presumably if you boot off a serial
console you could do it that way. (I hacked the ddb entry points to
switch back to console mode.) But then you'll find that the stack is
useless because of the OF vs kernel stack situation.

-- Dan