On Tue, Apr 12, 2022, Jue Wang wrote: > On Tue, Apr 12, 2022 at 7:22 AM Jue Wang <juew@xxxxxxxxxx> wrote: > > > > + return 0; > > > > + > > > > + if (lapic_in_kernel(vcpu)) > > > > > > Any reason to support UNCA injection without an in-kernel APIC? > > Even without a viable path for CMCI signaling, the UCNA errors should > still be logged in registers so it gives consistent semantics to > hardware for e.g., MCE overflow checks on an MCE signaled later. Right, what I was suggesting is that KVM reject the ioctl() if the VM doesn't have an in-kernel APIC, i.e. force userspace to use KVM's local APIC if userspace wants to also support UNCA injection. Side topic, please trim your replies, i.e. delete all the unnecessary quoted context.
