On 4/6/23 18:08, Sean Christopherson wrote:
Signed-off-by: Li RongQing<lirongqing@xxxxxxxxx>
---
diff with v1: rewrite changelog and indentation
This also fails to mention my objection to querying PV_UNHALT[*]. When I said
"this needs Paolo's attention no matter what", I did not mean "post a v2 and hope
Paolo applies it", I meant we need Paolo (and others) to weigh in on the ongoing
discussion.
Quoting v1:
+ if (kvm_para_has_feature(KVM_FEATURE_PV_UNHALT) ||
Rather than have the guest rely on host KVM behavior clearing PV_UNHALT when HLT
is passed through), would it make sense to add something like KVM_HINTS_HLT_PASSTHROUGH
to more explicitly tell the guest that HLT isn't intercepted?
Yes, I agree with adding KVM_HINTS_HLT_PASSTHROUGH or
KVM_HINTS_GUEST_CSTATE (i.e. host can remain in guest mode even when
running in C1 aka hlt or possibly deeper states).
Lack of PV_UNHALT does not indicate anything about whether HLT will be
handled in host or guest. In fact the same KVM_HINTS_* value could also
be used to disable PV_UNHALT.
Paolo
