Re: [patch 1/2] KVM: hlt emulation should take in-kernel APIC/PIT timers into account

["Yang, Sheng" <sheng.yang@intel.com>]

On Friday 09 May 2008 22:53:00 Marcelo Tosatti wrote:
> On Fri, May 09, 2008 at 03:49:20PM +0800, Yang, Sheng wrote:
> > On Sunday 13 April 2008 17:28:22 Avi Kivity wrote:
> > > Marcelo Tosatti wrote:
> > > > On Fri, Apr 11, 2008 at 03:12:41PM +0300, Avi Kivity wrote:
> > > >> This breaks ia64 (and shouldn't s390 use this too?)
> > > >>
> > > >>> 	 * We will block until either an interrupt or a signal wakes us up
> > > >>> 	 */
> > > >>> 	while (!kvm_cpu_has_interrupt(vcpu)
> > > >>> +	       && !kvm_cpu_has_pending_timer(vcpu)
> > > >>
> > > >> I guess the fix is to stub this out for the other archs.
> > > >
> > > > Agreed. How's this.
> > >
> > > Better :); applied.
> >
> > Hi, Marcelo
> >
> > This patch got into trouble when OS don't use PIT/LAPIC timer and don't
> > disable them. Then the pending counters would keep increasing, but the
> > HLT emulation can't be executed. And this would resulted in mass a lot
> > (above 220,000 per second) halt_exit for the Windows XP that using RTC as
> > the clocksource (and keep PIT enabled after bios did, just mask the pin)
> > idle, and the cpu utilize would be about 100% of QEmu process.
> >
> > The following patch used another way to fix the issue, though not very
> > formal.
>
> Hi Sheng,
>
> Did you have kvm.git commit 8ae6dc90ac84d9734e343210c8ec709f50cd9d89
> when testing this?
>
> I believe it should fix that issue, because "ps->inject_pending" won't
> be set by kvm_pit_timer_intr_post() if the IRQ is masked. Please correct
> me if I'm wrong.

Oh, sorry, I missed that commit. But... It just solved an half of the problem. 
LAPIC suffered from it as well, and the current HLT emulation still didn't 
work... And I can't find something like inject_pending in LAPIC timer.

I have to say, I think my method is more preciously, directly and efficient... 
It also can be extended easily if we got more clock sources (though I don't 
think this would happen in near future...). In fact, I think take care of 
pending counts is some kind of *wrong concept*... We should take care of the 
window, or when the increment of pending counters happened, CMIIW. And it got 
nothing to do with the current counter number (yeah, I realized it after saw 
the hlt behaviour in XP, not before ;) ).

--
Thanks
Yang, Sheng

-------------------------------------------------------------------------
This SF.net email is sponsored by the 2008 JavaOne(SM) Conference 
Don't miss this year's exciting event. There's still time to save $100. 
Use priority code J8TL2D2. 
http://ad.doubleclick.net/clk;198757673;13503038;p?http://java.sun.com/javaone