Re: ARAnyM VMs with Debian hanging at 100% CPU usage
On Thu, 2 Aug 2012 17:13:49 +0000 (UTC), Thorsten Glaser <email@example.com> wrote:
> Geert Uytterhoeven dixit:
> >If I'm not mistaken, the warning was just a false positive, so I didn't submit
> >the fix for stable?
> Was it just the warning issued, or was there really something
> wrong with interrupt handling on Atari, that's the question
> here (which *I* cannot answer). I was under the impression it
> fixed a warning, too, but Andreas' mail suggested it may have
> been a fix for interrupt handling or somesuch...
Although the kernel warning turned out to be a false positive, the
underlying cause was a real bug with real consequences: a CPU-bound
process could block nfeth interrupts for unbounded periods of time.
(I did describe that in the patch message, but perhaps I didn't
emphasize it enough.)
I've seen it happen too, with gcc bootstraps being killed because log
output via a pipe and nfeth couldn't be drained entually SIGPIPEing