Messages in this thread | | | Date | Tue, 12 May 2015 06:08:24 -0700 | From | "Paul E. McKenney" <> | Subject | Re: [PATCH 3/6] dataplane nohz: run softirqs synchronously on user entry |
| |
On Tue, May 12, 2015 at 11:32:02AM +0200, Peter Zijlstra wrote: > On Mon, May 11, 2015 at 04:13:16PM -0400, Chris Metcalf wrote: > > The thing you want to avoid is having two processes both > > runnable at once > > Right, because as soon as nr_running > 1 we kill the entire nohz_full > thing. RT or not for ksoftirqd doesn't matter. > > Then again, like interrupts, you basically want to avoid softirqs in > this mode. > > So I think the right solution is to figure out why the softirqs get > raised and cure that.
Makes sense, but it also makes sense to have something that detects when that cure fails and clean up. And, in a test/debug environment, also issuing some sort of diagnostic in that case.
Thanx, Paul
| |