Re: [PATCHv4 3/5] kvm: host side for eoi optimization

[Gleb Natapov <gleb@redhat.com>]

On Thu, May 17, 2012 at 12:10:55PM +0300, Michael S. Tsirkin wrote:
> On Thu, May 17, 2012 at 10:57:33AM +0300, Avi Kivity wrote:
> > On 05/17/2012 10:49 AM, Michael S. Tsirkin wrote:
> > > On Thu, May 17, 2012 at 10:28:41AM +0300, Gleb Natapov wrote:
> > > > On Wed, May 16, 2012 at 10:07:58PM +0300, Michael S. Tsirkin wrote:
> > > > > > This code is reached from kvm_cpu_get_interrupt(), but this function will
> > > > > > not be called in above scenario.
> > > > > 
> > > > > I think I see. So this shall fix it also makes code cleaner
> > > > > (no -2 hack). Right? kvm_apic_has_interrupt is called correct?
> > > > > 
> > > > > diff --git a/arch/x86/kvm/lapic.c b/arch/x86/kvm/lapic.c
> > > > > index b4f7013..5a38e34 100644
> > > > > --- a/arch/x86/kvm/lapic.c
> > > > > +++ b/arch/x86/kvm/lapic.c
> > > > > @@ -1273,8 +1273,12 @@ int kvm_apic_has_interrupt(struct kvm_vcpu *vcpu)
> > > > >  	highest_irr = apic_find_highest_irr(apic);
> > > > >  	if (highest_irr == -1)
> > > > >  		return -1;
> > > > > -	if (((highest_irr & 0xF0) <= apic_get_reg(apic, APIC_PROCPRI)))
> > > > > -		return -2;
> > > > > +	/* Detect interrupt nesting and disable EOI optimization */
> > > > > +	if ((highest_irr & 0xF0) <= apic_get_reg(apic, APIC_PROCPRI)) {
> > > > > +		if (pv_eoi_enabled(vcpu))
> > > > > +			pv_eoi_clr_pending(vcpu);
> > > > > +		return -1;
> > > > > +	}
> > > > >  	return highest_irr;
> > > > >  }
> > > > >  
> > > > I do not like it. kvm_apic_has_interrupt() does not suppose to mutate
> > > > state.
> > >
> > > OK, so let's rename it so it's clear it can mutate state?
> > >
> > 
> > No, let's refactor this so it makes sense.  The {has|get}_interrupt
> > split is the cause of the problem, I think.  We need a single function,
> > with callbacks that are called when an event happens.  The callbacks can
> > request an irq window exit, inject an interrupt, play with pveoi, or
> > cause a #vmexit.
> 
> OK will look into this next week.
> 
What about looking into my suggestion with setting vcpu->request bit? I
think totally refactoring irq handling path should not be done as part
of this series.

--
			Gleb.