On 13/04/20 23:26, Sean Christopherson wrote: > FWIW, I think the whole vcpuid thing is a bad interface, almost all the > tests end up defining an arbitrary number for the sole VCPU_ID, i.e. the > vcpuid interface just adds a pointless layer of obfuscation. I haven't > looked through all the tests, but returning the vcpu and making the struct > opaque, same as kvm_vm, seems like it would yield more readable code with > less overhead. Yes, I agree. This was in the original Google submission, I didn't like it either but I didn't feel like changing it and I wouldn't mind if someone does the work... Paolo
