On 04/05/2017 17:34, Radim Krčmář wrote: > > VMCS consistency is actually pretty tricky to solve and the most > reasonable way is to add direct access into the VCPU IOCTL interface and > go through userspace. Although, when you are doing that, you can just > implement add an interface for the feature that needs VMCS changes ... > and this is actually how we develop KVM. I agree. Read-only access is already debatable but I guess okay as a debug interface; write access to the VMCS is really a no-no. What do you want to do with it? Paolo
