On 13/05/20 14:49, Jon Doron wrote: >> > > To be honest from my understanding of the TLFS it states: > "The maximum input value for hypervisor CPUID information." > > So we should not expose stuff we wont "answer" to, I agree you can > always issue CPUID to any leaf and you will get zeroes but if we try to > follow TLFS it sounds like this needs to be capped here. I think Roman is right in the reading, but it's also nicer to have a capped EAX because you can just look at EAX and guess what features are there (it's simpler than looking for zeroes). Paolo
