Messages in this thread | | | Date | Sun, 18 Jan 2009 09:37:44 +0100 | From | Ingo Molnar <> | Subject | Re: [git pull] scheduler fixes |
| |
* Avi Kivity <avi@redhat.com> wrote:
> Ingo Molnar wrote: >> * Mike Galbraith <efault@gmx.de> wrote: >> >> >>> On Sat, 2009-01-17 at 04:43 -0800, Andrew Morton wrote: >>> >>>> http://bugzilla.kernel.org/show_bug.cgi?id=12465 just popped up - another >>>> scheduler regression. It has been bisected. >>>> >>> Seems pretty clear. I'd suggest reverting it. >>> >> >> We can revert it (and will revert it if no solution is found), but i'd >> also like to understand why it happens, because that kind of regression >> from this change is unexpected - we might be hiding some bug that could >> pop up under less debuggable circumstances, so we need to understand it >> while we have a chance. >> >> Below is the commit in question. Avi, any ideas what makes KVM special >> here? Perhaps its use of "preempt notifiers" is causing a problem >> somehow? >> > > preempt notifiers use should cause additional context switch costs of a > few thousand cycles and possible an IPI (if a vcpu was migrated). So > I'd suspect scheduling latency here. > > Is it possible to trace this (the time between a wake up and actual > scheduling of a task)?
Can you reproduce those latencies? We didnt get similar reports from elsewhere so there seems to be a KVM angle.
Ingo
| |