Re: [PATCH RFC 1/2] KVM: don't check for PF_VCPU when yielding

[David Hildenbrand <dahi@linux.vnet.ibm.com>]

> On 11/28/2014 04:28 PM, Christian Borntraeger wrote:
> > Am 28.11.2014 um 11:08 schrieb Raghavendra KT:
> >> Was able to test the patch, here is the result: I have not tested with
> >> bigger VMs though. Results make it difficult to talk about any side
> >> effect of
> >> patch if any.
> >
> > Thanks a log.
> >
> > If our assumption is correct, then this patch should have no side effect on x86. Do you have any confidence guess if the numbers below mean: no-change vs. regression vs improvement?
> >
> 
> I am seeing very small improvement in <= 1x commit cases
> and for >1x overcommit, a very slight regression. But considering the
> test environment noises, I do not see much effect from the
> patch.
> 
> But I admit, I have not explored deeply about,
> 1. assumption of preempted approximately equals PF_VCPU case logic,

PF_VCPU is only a hint whether the target vcpu is executing the guest.
If preemption is off or !s390, PF_VCPU means that the target vcpu is running
and can't be preempted.

Although for preemption on and s390, this statement is false. Therefore this
check is not always right.

> 2. whether it helps for any future usages of yield_to against current
> sole usage of virtualization.
> 
> 
> 
Thanks for your test!