On Thu, 08 Mar 2018 00:07:07 +0530
"Naveen N. Rao" <naveen.n....@linux.vnet.ibm.com> wrote:
> Yes, that's negligible.
> Though, to be honest, I will have to introduce a 'mfmsr' for the older
> -pg variant. I still think that the improved reliability far outweighs
> the minor slowdown there.
In that case, can you introduce a read_mostly variable that can be
tested before calling the mfmsr. Why punish normal ftrace tracing if
kvm is not enabled or running?
Both should probably have an #ifdef CONFIG_KVM encapsulating the code.