Re: [RFC] KVM: optimize the kvm_vcpu_on_spin

[Paolo Bonzini]

On 31/07/2017 15:42, Marc Zyngier wrote:
>> If the vcpu(me) exit due to request a usermode spinlock, then
>> the spinlock-holder may be preempted in usermode or kernmode.
>> But if the vcpu(me) is in kernmode, then the holder must be
>> preempted in kernmode, so we should choose a vcpu in kernmode
>> as the most eligible candidate.
>
> That seems to preclude any form of locking between userspace and kernel
> (which probably wouldn't be Linux). Are you sure that this form of
> construct is not used anywhere? I have the feeling this patch could
> break this scenario...

It's just a heuristic; it would only be broken if you overcommit, and it
would be just as broken as if KVM didn't implement directed yield at all.

Paolo