On Wed, Jan 05, 2022, Yang Zhong wrote:
> @@ -6399,6 +6424,26 @@ static void handle_interrupt_nmi_irqoff(struct kvm_vcpu *vcpu,
> kvm_after_interrupt(vcpu);
> }
>
> +static void handle_nm_fault_irqoff(struct kvm_vcpu *vcpu)
> +{
> + /*
> + * Save xfd_err to guest_fpu before interrupt is enabled, so the
> + * MSR value is not clobbered by the host activity before the guest
> + * has chance to consume it.
> + *
> + * We should not blindly read xfd_err here, since this exception
Nit, avoid "we", and explain what KVM does (or doesn't) do, not what KVM "should"
do, e.g. just say
* Do not blindly read ...
> + * might be caused by L1 interception on a platform which doesn't
> + * support xfd at all.
> + *
> + * Do it conditionally upon guest_fpu::xfd. xfd_err matters
> + * only when xfd contains a non-zero value.
> + *
> + * Queuing exception is done in vmx_handle_exit. See comment there.
Another nit, it's worth explaining why XFD_ERR needs to be read here regardless
of is_guest_mode(). E.g.
* Injecting the #NM back into the guest is handled in the standard path
* as an #NM in L2 may be reflected into L1 as a VM-Exit. Read XFD_ERR
* even if the #NM is from L2, as L1 may have exposed XFD to L2.
Side topic, in a follow up series/patch, it's probably worth adding support in
nested_vmx_prepare_msr_bitmap() to allow passthrough of the MSRs to L2.
> + */
> + if (vcpu->arch.guest_fpu.fpstate->xfd)
> + rdmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err);
> +}
> +
> static void handle_exception_nmi_irqoff(struct vcpu_vmx *vmx)
> {
> const unsigned long nmi_entry = (unsigned long)asm_exc_nmi_noist;
> @@ -6407,6 +6452,9 @@ static void handle_exception_nmi_irqoff(struct vcpu_vmx *vmx)
> /* if exit due to PF check for async PF */
> if (is_page_fault(intr_info))
> vmx->vcpu.arch.apf.host_apf_flags = kvm_read_and_reset_apf_flags();
> + /* if exit due to NM, handle before interrupts are enabled */
Nit, drop this comment, it's slightly misleading since the #NM isn't fully handled
here. The comment in handle_nm_fault_irqoff() is more than sufficient.
> + else if (is_nm_fault(intr_info))
> + handle_nm_fault_irqoff(&vmx->vcpu);
> /* Handle machine checks before interrupts are enabled */
> else if (is_machine_check(intr_info))
> kvm_machine_check();
> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
> index 21ce65220e38..2c988f8ca616 100644
> --- a/arch/x86/kvm/x86.c
> +++ b/arch/x86/kvm/x86.c
> @@ -9953,6 +9953,9 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
> if (test_thread_flag(TIF_NEED_FPU_LOAD))
> switch_fpu_return();
>
> + if (vcpu->arch.guest_fpu.xfd_err)
> + wrmsrl(MSR_IA32_XFD_ERR, vcpu->arch.guest_fpu.xfd_err);
> +
> if (unlikely(vcpu->arch.switch_db_regs)) {
> set_debugreg(0, 7);
> set_debugreg(vcpu->arch.eff_db[0], 0);
> @@ -10016,6 +10019,9 @@ static int vcpu_enter_guest(struct kvm_vcpu *vcpu)
>
> static_call(kvm_x86_handle_exit_irqoff)(vcpu);
>
> + if (vcpu->arch.guest_fpu.xfd_err)
> + wrmsrl(MSR_IA32_XFD_ERR, 0);
> +
> /*
> * Consume any pending interrupts, including the possible source of
> * VM-Exit on SVM and any ticks that occur between VM-Exit and now.
