On Mon, 2018-07-09 at 14:55 +0200, Peter Zijlstra wrote: > On Mon, Jul 09, 2018 at 05:34:57AM -0700, Paul E. McKenney wrote: > > But KVM defeats this by checking need_resched() before invoking > > cond_resched(). > > That's not wrong or even uncommon I think. Right. Which is precisely why I baulked at just adding the extra check in the KVM loop, and wanted to talk about making need_resched() consistent with what cond_resched() actually does. :)