On 21/02/20 14:52, Vitaly Kuznetsov wrote: >> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c >> index fbabb2f06273..72f7ca4baa6d 100644 >> --- a/arch/x86/kvm/x86.c >> +++ b/arch/x86/kvm/x86.c >> @@ -2675,7 +2675,7 @@ static void record_steal_time(struct kvm_vcpu *vcpu) >> trace_kvm_pv_tlb_flush(vcpu->vcpu_id, >> st->preempted & KVM_VCPU_FLUSH_TLB); >> if (xchg(&st->preempted, 0) & KVM_VCPU_FLUSH_TLB) >> - kvm_vcpu_flush_tlb(vcpu, false); >> + kvm_x86_ops->tlb_flush_guest(vcpu); >> >> vcpu->arch.st.preempted = 0; > There is one additional place in hyperv.c where we do TLB flush on > behalf of the guest, kvm_hv_flush_tlb(). Currently, it does > KVM_REQ_TLB_FLUSH (resulting in kvm_x86_ops->tlb_flush()), do we need > something like KVM_REQ_TLB_FLUSH_GUEST instead? Yes, that would be better since INVEPT does not flush linear mappings. So, when EPT and VPID is enabled, KVM_REQ_TLB_FLUSH would not flush the guest's translations. Paolo
