On Wed, 3 Jul 2019 at 22:13, Paolo Bonzini <pbonzini@xxxxxxxxxx> wrote: > > On 03/07/19 02:48, Wanpeng Li wrote: > > Hi Paolo, how about this patchset? Patch 2/2 is easy to take, do you > > have more concern about patch 1/2? > > I don't know. It seems somewhat hard to tune and in cyclictest it only > happens for preemption_timer=N. Are you using preemption_timer=N > together with the LAPIC-timer-on-service-CPU patches? A 25ns conservative value makes no benefit for cyclictest any more even when preemption_timer=N. Btw, maybe it is the time to merge the LAPIC-timer-on-service-CPU patches now. :) Regards, Wanpeng Li
