On Thu, Mar 03, 2022, Sean Christopherson wrote: > With your proposed change, KVM_SET_LAPIC will fail and we've broken a functional, > if sketchy, setup. Is there likely to be such a real-world setup that doesn't > barf on the inconsistent x2APIC ID? Probably not, but I don't see any reason to > find out. I take back the "probably not", this isn't even all that contrived. Prior to the "migration", the guest will see a consistent x2APIC ID. It's not hard to imagine a guest that snapshots the ID and never re-reads the value from "hardware".