On Sat, Nov 09, 2002 at 04:11:18AM -0500, Rafal Boni wrote:
> Anyone else seeing this?

I'm seeing RED state exceptions, which either lock up the system completely
or bring me to the ok prompt (not ddb).

They are timinig dependend, and tend to convert into random core dumps
when using DIAGNOSTIC and DEBUG options (but not always).

My firmware is always more or less completely wedged when I reach the OK prompt
(if I get that far), and all debugging attempts have been futile yet.

It looks like we run out of hardware contexts, invalidate them all and flush
caches. Then we hit a window fill trap and something else in the wrong order
and RED state.

Which would mean: the bug has been there for a long time, but hidden by many
superflous cache flushes. Now due to Chucks performance improvements we see
this more likely due to changed timing.

Interestingly it happens to me on a 360MHz/4MB ecache machine too, but not 
on any other.

Can you verify you get the RED state messages too? I can send you some 
debugging stuff/workaround I've been experimenting with then.

Martin