>> Hm, do you need this twice? > > In my opinion this has only value if we assume that HW and/or KVM is buggy and > we are running host model (or it's expansion). >
The "sanity" checks in KVM sensing code don't really hurt. But I agree, sane KVM should not produce this. > And even the we would get a warning, and nothing bad would happen with a linux > guest. > > While I'm not strongly opposing this, I would not mind it dropped. If we want > to make sure things are consistent I would prefer the consistency check being > generating an error (instead of a warning). > We use a warning as it is helpful for development (e.g. under TCG you can enable msa5, although we yield a warning due to a failing consistency check). -- Thanks, David / dhildenb
