Re: [PATCH 2/3] KVM: nVMX: Fix conditions for NMI and interrupt injection

[Gleb Natapov <gleb@redhat.com>]

On Wed, Mar 13, 2013 at 05:53:45PM +0100, Jan Kiszka wrote:
> If we are in guest mode, L0 can only inject events into L2 if L1 has
> nothing pending. Otherwise, L0 would overwrite L1's events and they
> would get lost. This check is conceptually independent of
> nested_exit_on_intr.
> 
> If L1 traps external interrupts, then we also need to look at L1's
> idt_vectoring_info_field. If it is empty, we can kick the guest from L2
> to L1, just like the previous code worked.
> 
> Finally, the logic for checking interrupt has to be applied also on NMIs
> in an analogous way. This enables NMI interception for nested guests.
> 
> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
> ---
>  arch/x86/kvm/vmx.c |   59 ++++++++++++++++++++++++++++++++++++++++++++-------
>  1 files changed, 51 insertions(+), 8 deletions(-)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index b50174d..10de336 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -4211,6 +4211,12 @@ static bool nested_exit_on_intr(struct kvm_vcpu *vcpu)
>  		PIN_BASED_EXT_INTR_MASK;
>  }
>  
> +static bool nested_exit_on_nmi(struct kvm_vcpu *vcpu)
> +{
> +	return get_vmcs12(vcpu)->pin_based_vm_exec_control &
> +		PIN_BASED_NMI_EXITING;
> +}
> +
It will take me some time to review this, but I have a small nit now.
You open code checking of this bit in your previous patch, why not move
this hunk there?

>  static void enable_irq_window(struct kvm_vcpu *vcpu)
>  {
>  	u32 cpu_based_vm_exec_control;
> @@ -4307,6 +4313,30 @@ static void vmx_inject_nmi(struct kvm_vcpu *vcpu)
>  
>  static int vmx_nmi_allowed(struct kvm_vcpu *vcpu)
>  {
> +	if (is_guest_mode(vcpu)) {
> +		struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
> +
> +		if (to_vmx(vcpu)->nested.nested_run_pending &&
> +		    (vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK))
> +			return 0;
> +		if (nested_exit_on_nmi(vcpu)) {
> +			/*
> +			 * Check if the idt_vectoring_info_field is free. We
> +			 * cannot raise EXIT_REASON_EXCEPTION_NMI if it isn't.
> +			 */
> +			if (vmcs12->idt_vectoring_info_field &
> +			    VECTORING_INFO_VALID_MASK)
> +				return 0;
> +			nested_vmx_vmexit(vcpu);
> +			vmcs12->vm_exit_reason = EXIT_REASON_EXCEPTION_NMI;
> +			vmcs12->vm_exit_intr_info = NMI_VECTOR |
> +				INTR_TYPE_NMI_INTR | INTR_INFO_VALID_MASK;
> +			/*
> +			 * fall through to normal code, but now in L1, not L2
> +			 */
> +		}
> +	}
> +
>  	if (!cpu_has_virtual_nmis() && to_vmx(vcpu)->soft_vnmi_blocked)
>  		return 0;
>  
> @@ -4346,16 +4376,29 @@ static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked)
>  
>  static int vmx_interrupt_allowed(struct kvm_vcpu *vcpu)
>  {
> -	if (is_guest_mode(vcpu) && nested_exit_on_intr(vcpu)) {
> +	if (is_guest_mode(vcpu)) {
>  		struct vmcs12 *vmcs12 = get_vmcs12(vcpu);
> -		if (to_vmx(vcpu)->nested.nested_run_pending ||
> -		    (vmcs12->idt_vectoring_info_field &
> -		     VECTORING_INFO_VALID_MASK))
> +
> +		if (to_vmx(vcpu)->nested.nested_run_pending &&
> +		    (vmcs12->vm_entry_intr_info_field & INTR_INFO_VALID_MASK))
>  			return 0;
> -		nested_vmx_vmexit(vcpu);
> -		vmcs12->vm_exit_reason = EXIT_REASON_EXTERNAL_INTERRUPT;
> -		vmcs12->vm_exit_intr_info = 0;
> -		/* fall through to normal code, but now in L1, not L2 */
> +		if (nested_exit_on_intr(vcpu)) {
> +			/*
> +			 * Check if the idt_vectoring_info_field is free. We
> +			 * cannot raise EXIT_REASON_EXTERNAL_INTERRUPT if it
> +			 * isn't.
> +			 */
> +			if (vmcs12->idt_vectoring_info_field &
> +			    VECTORING_INFO_VALID_MASK)
> +				return 0;
> +			nested_vmx_vmexit(vcpu);
> +			vmcs12->vm_exit_reason =
> +				EXIT_REASON_EXTERNAL_INTERRUPT;
> +			vmcs12->vm_exit_intr_info = 0;
> +			/*
> +			 * fall through to normal code, but now in L1, not L2
> +			 */
> +		}
>  	}
>  
>  	return (vmcs_readl(GUEST_RFLAGS) & X86_EFLAGS_IF) &&
> -- 
> 1.7.3.4

--
			Gleb.