On 2/13/2025 2:56 AM, Sean Christopherson wrote:
On Wed, Feb 12, 2025, Binbin Wu wrote:
On 2/12/2025 8:46 AM, Sean Christopherson wrote:
I am completely comfortable saying that KVM doesn't care about STI/SS shadows
outside of the HALTED case, and so unless I'm missing something, I think it makes
sense for tdx_protected_apic_has_interrupt() to not check RVI outside of the HALTED
case, because it's impossible to know if the interrupt is actually unmasked, and
statistically it's far, far more likely that it _is_ masked.
OK. Will update tdx_protected_apic_has_interrupt() in "TDX interrupts" part.
And use kvm_vcpu_has_events() to replace the open code in this patch.
Something to keep an eye on: kvm_vcpu_has_events() returns true if pv_unhalted
is set, and pv_unhalted is only cleared on transitions KVM_MP_STATE_RUNNABLE.
If the guest initiates a spurious wakeup, pv_unhalted could be left set in
perpetuity.
Oh, yes.
KVM_HC_KICK_CPU is allowed in TDX guests.
The change below looks good to me.
One minor issue is when guest initiates a spurious wakeup, pv_unhalted is
left set, then later when the guest want to halt the vcpu, in
__kvm_emulate_halt(), since pv_unhalted is still set and the state will not
transit to KVM_MP_STATE_HALTED.
But I guess it's guests' responsibility to not initiate spurious wakeup,
guests need to bear the fact that HLT could fail due to a previous
spurious wakeup?
I _think_ this would work and is generally desirable?
diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
index 8e77e61d4fbd..435ca2782c3c 100644
--- a/arch/x86/kvm/x86.c
+++ b/arch/x86/kvm/x86.c
@@ -11114,9 +11114,6 @@ static bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu)
kvm_apic_init_sipi_allowed(vcpu))
return true;
- if (vcpu->arch.pv.pv_unhalted)
- return true;
-
if (kvm_is_exception_pending(vcpu))
return true;
@@ -11157,7 +11154,8 @@ static bool kvm_vcpu_has_events(struct kvm_vcpu *vcpu)
int kvm_arch_vcpu_runnable(struct kvm_vcpu *vcpu)
{
- return kvm_vcpu_running(vcpu) || kvm_vcpu_has_events(vcpu);
+ return kvm_vcpu_running(vcpu) || vcpu->arch.pv.pv_unhalted ||
+ kvm_vcpu_has_events(vcpu);
}
/* Called within kvm->srcu read side. */
@@ -11293,7 +11291,7 @@ static int __kvm_emulate_halt(struct kvm_vcpu *vcpu, int state, int reason)
*/
++vcpu->stat.halt_exits;
if (lapic_in_kernel(vcpu)) {
- if (kvm_vcpu_has_events(vcpu))
+ if (kvm_vcpu_has_events(vcpu) || vcpu->arch.pv.pv_unhalted)
vcpu->arch.pv.pv_unhalted = false;
else
vcpu->arch.mp_state = state;
