On Tue, Feb 05, 2002 at 03:04:21PM -0700, Ronald G Minnich wrote:
> Andi, we'll try to look at your patch as soon as we can. Thanks for
> confirming the possibility of a problem. Right now we are going to patch
> supermon first to see if we can work around the problem, since we can't
> ask everybody to apply that patch.
It's not really a proposed patch, but just to confirm that I diagnosed the
problem correctly. The eventual goal (if that should be the problem) would
be to fix the fast path to work for your workload too.
You could also alternatively give supermon realtime priority to let it
schedule sooner, make it process user context TCP faster.