On 23.04.2010, at 16:13, Joerg Roedel wrote: > On Fri, Apr 23, 2010 at 03:46:07PM +0200, Alexander Graf wrote: >> >> On 22.04.2010, at 12:33, Joerg Roedel wrote: >> >>> The patch introducing nested nmi handling had a bug. The >>> check does not belong to enable_nmi_window but must be in >>> nmi_allowed. This patch fixes this. >>> >>> Signed-off-by: Joerg Roedel <joerg.roedel@xxxxxxx> >>> --- >>> arch/x86/kvm/svm.c | 16 +++++++++------- >>> 1 files changed, 9 insertions(+), 7 deletions(-) >>> >>> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c >>> index ab78eb8..ec20584 100644 >>> --- a/arch/x86/kvm/svm.c >>> +++ b/arch/x86/kvm/svm.c >>> @@ -2771,8 +2771,12 @@ static int svm_nmi_allowed(struct kvm_vcpu *vcpu) >>> { >>> struct vcpu_svm *svm = to_svm(vcpu); >>> struct vmcb *vmcb = svm->vmcb; >>> - return !(vmcb->control.int_state & SVM_INTERRUPT_SHADOW_MASK) && >>> - !(svm->vcpu.arch.hflags & HF_NMI_MASK); >>> + int ret; >>> + ret = !(vmcb->control.int_state & SVM_INTERRUPT_SHADOW_MASK) && >>> + !(svm->vcpu.arch.hflags & HF_NMI_MASK); >>> + ret = ret && gif_set(svm) && nested_svm_nmi(svm); >>> + >>> + return ret; >>> } >>> >>> static bool svm_get_nmi_mask(struct kvm_vcpu *vcpu) >>> @@ -2841,11 +2845,9 @@ static void enable_nmi_window(struct kvm_vcpu *vcpu) >>> * Something prevents NMI from been injected. Single step over possible >>> * problem (IRET or exception injection or interrupt shadow) >>> */ >>> - if (gif_set(svm) && nested_svm_nmi(svm)) { >>> - svm->nmi_singlestep = true; >>> - svm->vmcb->save.rflags |= (X86_EFLAGS_TF | X86_EFLAGS_RF); >>> - update_db_intercept(vcpu); >>> - } >>> + svm->nmi_singlestep = true; >>> + svm->vmcb->save.rflags |= (X86_EFLAGS_TF | X86_EFLAGS_RF); >>> + update_db_intercept(vcpu); >> >> So we're always messing with the nested guest state when the host >> wants to inject an nmi into the l1 guest? Is that safe? > > Why not? We can't inject an NMI directly into L2 if the nested > hypervisor intercepts it. So where did the code go that does the #vmexit in case the nested hypervisor does intercept it? It used to be nested_svm_nmi(), right? Alex -- To unsubscribe from this list: send the line "unsubscribe kvm" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html