Re: [patch] Race between init_idle and reschedule_idle

[Martin J. Bligh]

> martin, we experienced the same problem way back when we brought our
> Fujitsu based Numa machine up. We also experience the problem with 
> our cpu pooling and load balancing approach. I think the ksoftirq might
> have similar problems.
> We solved it in a similar fashion, through counters.
> I strongly suggest to put this code into the main track.

It's there, in both Linus' and Alan's latest trees. 

For ages, it showed up as an undiagnosable BINT error, but in 2.4.10 it 
changed to a  diagnosable panic, making it fairly easy to resolve. Not sure 
what changed this - may have been luck ;-)

M.