Re: [PATCH] If interrupt injection is not possible do not scan IRR.

[Gleb Natapov <gleb@redhat.com>]

On Mon, May 18, 2009 at 12:00:40PM +0300, Avi Kivity wrote:
> Gleb Natapov wrote:
>> @@ -3174,10 +3174,11 @@ static void inject_pending_irq(struct kvm_vcpu *vcpu, struct kvm_run *kvm_run)
>>  			vcpu->arch.nmi_injected = true;
>>  			kvm_x86_ops->set_nmi(vcpu);
>>  		}
>> -	} else if (kvm_cpu_has_interrupt(vcpu)) {
>> -		if (kvm_x86_ops->interrupt_allowed(vcpu)) {
>> -			kvm_queue_interrupt(vcpu, kvm_cpu_get_interrupt(vcpu),
>> -					    false);
>> +	} else if (kvm_x86_ops->interrupt_allowed(vcpu)) {
>> +		int vec = kvm_cpu_get_interrupt(vcpu);
>> +		if (vec != -1) {
>> +			printk(KERN_ERR"inject %d rip=%x\n", vec, kvm_rip_read(vcpu));
>> +			kvm_queue_interrupt(vcpu, vec, false);
>>  			kvm_x86_ops->set_irq(vcpu);
>>  		} 	}
>>   
>
> Not sure if this is a win.  Usually interrupts are allowed, so we'll  
> have to do both calculations.  On vmx ->interrupt_allowed() reads the  
> vmcs, which is slow.
>
Depend on how slow vmcs read is. In case when interrupt is available we
also scan IRR twice with current code: first time in kvm_cpu_has_interrupt()
and another in kvm_cpu_get_interrupt().

If we want to optimize vmcs access, and for nested virtualization we may
want to do it, we can read common ones on exit and use in memory copy.

--
			Gleb.