Message ID | 20190129080913.40998-1-suravee.suthikulpanit@amd.com (mailing list archive) |
---|---|
State | New, archived |
Headers | show |
Series | svm: Fix improper check when deactivate AVIC | expand |
On 29/01/19 09:09, Suthikulpanit, Suravee wrote: > The function svm_refresh_apicv_exec_ctrl() always returning prematurely > as kvm_vcpu_apicv_active() always return false when calling from > the function arch/x86/kvm/x86.c:kvm_vcpu_deactivate_apicv(). > This is because the apicv_active is set to false just before calling > refresh_apicv_exec_ctrl(). > > Also, we need to mark VMCB_AVIC bit as dirty instead of VMCB_INTR. > > So, fix svm_refresh_apicv_exec_ctrl() to properly deactivate AVIC. > > Fixes: 67034bb9dd5e ('KVM: SVM: Add irqchip_split() checks before enabling AVIC') Having the lapic_in_kernel check in your code is a bit ugly - the caller should check it. And actually, even though it doesn't check it right now, svm_refresh_apicv_exec_ctrl is only reachable if lapic_in_kernel returns true; see the KVM_CAP_HYPERV_SYNIC case of kvm_vcpu_ioctl_enable_cap. What about this simpler patch: diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c index 01b66bbcdd7a..74ceda470eae 100644 --- a/arch/x86/kvm/svm.c +++ b/arch/x86/kvm/svm.c @@ -5122,11 +5122,11 @@ static void svm_refresh_apicv_exec_ctrl( struct vcpu_svm *svm = to_svm(vcpu); struct vmcb *vmcb = svm->vmcb; - if (!kvm_vcpu_apicv_active(&svm->vcpu)) - return; - - vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; - mark_dirty(vmcb, VMCB_INTR); + if (kvm_vcpu_apicv_active(vcpu)) + vmcb->control.int_ctl |= AVIC_ENABLE_MASK; + else + vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; + mark_dirty(vmcb, VMCB_AVIC); } static void svm_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap) ? Thanks, Paolo
Paolo, On 1/30/19 11:22 PM, Paolo Bonzini wrote: > On 29/01/19 09:09, Suthikulpanit, Suravee wrote: >> The function svm_refresh_apicv_exec_ctrl() always returning prematurely >> as kvm_vcpu_apicv_active() always return false when calling from >> the function arch/x86/kvm/x86.c:kvm_vcpu_deactivate_apicv(). >> This is because the apicv_active is set to false just before calling >> refresh_apicv_exec_ctrl(). >> >> Also, we need to mark VMCB_AVIC bit as dirty instead of VMCB_INTR. >> >> So, fix svm_refresh_apicv_exec_ctrl() to properly deactivate AVIC. >> >> Fixes: 67034bb9dd5e ('KVM: SVM: Add irqchip_split() checks before enabling AVIC') > > Having the lapic_in_kernel check in your code is a bit ugly - the caller > should check it. And actually, even though it doesn't check it right > now, svm_refresh_apicv_exec_ctrl is only reachable if lapic_in_kernel > returns true; see the KVM_CAP_HYPERV_SYNIC case of > kvm_vcpu_ioctl_enable_cap. > > What about this simpler patch: > > diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c > index 01b66bbcdd7a..74ceda470eae 100644 > --- a/arch/x86/kvm/svm.c > +++ b/arch/x86/kvm/svm.c > @@ -5122,11 +5122,11 @@ static void svm_refresh_apicv_exec_ctrl( > struct vcpu_svm *svm = to_svm(vcpu); > struct vmcb *vmcb = svm->vmcb; > > - if (!kvm_vcpu_apicv_active(&svm->vcpu)) > - return; > - > - vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; > - mark_dirty(vmcb, VMCB_INTR); > + if (kvm_vcpu_apicv_active(vcpu)) > + vmcb->control.int_ctl |= AVIC_ENABLE_MASK; > + else > + vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; > + mark_dirty(vmcb, VMCB_AVIC); > } > > static void svm_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 > *eoi_exit_bitmap) > > ? > > Thanks, > > Paolo > Yes, this would work. This patch is actually a precursor of a patch series that I am working on to allow AVIC enabling/disabling during run-time. I will send out a series for RFC soon. Thanks, Suravee
Paolo, On 2/1/19 6:01 PM, Suravee Suthikulpanit wrote: > Paolo, > > On 1/30/19 11:22 PM, Paolo Bonzini wrote: >> On 29/01/19 09:09, Suthikulpanit, Suravee wrote: >>> The function svm_refresh_apicv_exec_ctrl() always returning prematurely >>> as kvm_vcpu_apicv_active() always return false when calling from >>> the function arch/x86/kvm/x86.c:kvm_vcpu_deactivate_apicv(). >>> This is because the apicv_active is set to false just before calling >>> refresh_apicv_exec_ctrl(). >>> >>> Also, we need to mark VMCB_AVIC bit as dirty instead of VMCB_INTR. >>> >>> So, fix svm_refresh_apicv_exec_ctrl() to properly deactivate AVIC. >>> >>> Fixes: 67034bb9dd5e ('KVM: SVM: Add irqchip_split() checks before enabling AVIC') >> >> Having the lapic_in_kernel check in your code is a bit ugly - the caller >> should check it. And actually, even though it doesn't check it right >> now, svm_refresh_apicv_exec_ctrl is only reachable if lapic_in_kernel >> returns true; see the KVM_CAP_HYPERV_SYNIC case of >> kvm_vcpu_ioctl_enable_cap. >> >> What about this simpler patch: >> >> diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c >> index 01b66bbcdd7a..74ceda470eae 100644 >> --- a/arch/x86/kvm/svm.c >> +++ b/arch/x86/kvm/svm.c >> @@ -5122,11 +5122,11 @@ static void svm_refresh_apicv_exec_ctrl( >> struct vcpu_svm *svm = to_svm(vcpu); >> struct vmcb *vmcb = svm->vmcb; >> >> - if (!kvm_vcpu_apicv_active(&svm->vcpu)) >> - return; >> - >> - vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; >> - mark_dirty(vmcb, VMCB_INTR); >> + if (kvm_vcpu_apicv_active(vcpu)) >> + vmcb->control.int_ctl |= AVIC_ENABLE_MASK; >> + else >> + vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; >> + mark_dirty(vmcb, VMCB_AVIC); >> } >> >> static void svm_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 >> *eoi_exit_bitmap) >> >> ? >> >> Thanks, >> >> Paolo >> > > Yes, this would work. > > This patch is actually a precursor of a patch series that I am working on > to allow AVIC enabling/disabling during run-time. I will send out a series > for RFC soon. > > Thanks, > Suravee Not sure if you have already applied this change. Anyhow, I just want to let you know that I have included the change you suggested as patch 1/8 of the RFC series "KVM: x86: svm: Enabling AVIC with in kernel irqchip" Thanks, Suravee
diff --git a/arch/x86/kvm/svm.c b/arch/x86/kvm/svm.c index d35c9002f282..59b1f68060c7 100644 --- a/arch/x86/kvm/svm.c +++ b/arch/x86/kvm/svm.c @@ -390,6 +390,8 @@ static int nested_svm_vmexit(struct vcpu_svm *svm); static int nested_svm_check_exception(struct vcpu_svm *svm, unsigned nr, bool has_error_code, u32 error_code); +static bool svm_get_enable_apicv(struct kvm_vcpu *vcpu); + enum { VMCB_INTERCEPTS, /* Intercept vectors, TSC offset, pause filter count */ @@ -478,6 +480,12 @@ static inline struct vcpu_svm *to_svm(struct kvm_vcpu *vcpu) return container_of(vcpu, struct vcpu_svm, vcpu); } +static bool is_avic_active(struct vcpu_svm *svm) +{ + return (svm_get_enable_apicv(&svm->vcpu) && + svm->vmcb->control.int_ctl & AVIC_ENABLE_MASK); +} + static inline void avic_update_vapic_bar(struct vcpu_svm *svm, u64 data) { svm->vmcb->control.avic_vapic_bar = data & VMCB_AVIC_APIC_BAR_MASK; @@ -5108,17 +5116,23 @@ static void svm_hwapic_isr_update(struct kvm_vcpu *vcpu, int max_isr) { } -/* Note: Currently only used by Hyper-V. */ -static void svm_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu) +static void svm_deactivate_avic(struct kvm_vcpu *vcpu) { struct vcpu_svm *svm = to_svm(vcpu); struct vmcb *vmcb = svm->vmcb; - if (!kvm_vcpu_apicv_active(&svm->vcpu)) + if (!lapic_in_kernel(vcpu) || !is_avic_active(svm)) return; vmcb->control.int_ctl &= ~AVIC_ENABLE_MASK; - mark_dirty(vmcb, VMCB_INTR); + mark_dirty(vmcb, VMCB_AVIC); +} + +/* Note: Currently only used by Hyper-V. */ +static void svm_refresh_apicv_exec_ctrl(struct kvm_vcpu *vcpu) +{ + if (!kvm_vcpu_apicv_active(vcpu)) + svm_deactivate_avic(vcpu); } static void svm_load_eoi_exitmap(struct kvm_vcpu *vcpu, u64 *eoi_exit_bitmap)
The function svm_refresh_apicv_exec_ctrl() always returning prematurely as kvm_vcpu_apicv_active() always return false when calling from the function arch/x86/kvm/x86.c:kvm_vcpu_deactivate_apicv(). This is because the apicv_active is set to false just before calling refresh_apicv_exec_ctrl(). Also, we need to mark VMCB_AVIC bit as dirty instead of VMCB_INTR. So, fix svm_refresh_apicv_exec_ctrl() to properly deactivate AVIC. Fixes: 67034bb9dd5e ('KVM: SVM: Add irqchip_split() checks before enabling AVIC') Cc: Radim Krčmář <rkrcmar@redhat.com> Cc: Paolo Bonzini <pbonzini@redhat.com> Signed-off-by: Suravee Suthikulpanit <suravee.suthikulpanit@amd.com> --- arch/x86/kvm/svm.c | 22 ++++++++++++++++++---- 1 file changed, 18 insertions(+), 4 deletions(-)