On 2/25/22, 5:48 AM, "Michal Prívozník" <mprivozn@redhat.com> wrote:

> Since the argument is going to be required I think it makes sense to

> require it for qemu 6.2.0. Otherwise we could have yet another

> capability to reflect whether .node attribute is needed or not. But IMO

> that's pointless because the capability would be set only for one

> particular release of QEMU (not to mention it's probably impossible to

> detect whether .node attribute is needed or not).

Really appreciate your comments!

We are re-designing our solution to support NUMA case, and might

also need to refactor current code because of under reviewing SGX

QEMU patch. Will push for reviewing once all of them are done.

Thanks,

Lin.