Re: [PATCH v4 5/6] KVM: nVMX: Fix conditions for NMI injection

[Jan Kiszka <jan.kiszka@web.de>]

[-- Attachment #1: Type: text/plain, Size: 2131 bytes --]

On 2013-04-14 17:23, Gleb Natapov wrote:
> On Sun, Apr 14, 2013 at 12:12:49PM +0200, Jan Kiszka wrote:
>> From: Jan Kiszka <jan.kiszka@siemens.com>
>>
>> The logic for checking if interrupts can be injected has to be applied
>> also on NMIs. The difference is that if NMI interception is on these
>> events are consumed and blocked by the VM exit.
>>
>> Signed-off-by: Jan Kiszka <jan.kiszka@siemens.com>
>> ---
>>  arch/x86/kvm/vmx.c |   28 ++++++++++++++++++++++++++++
>>  1 files changed, 28 insertions(+), 0 deletions(-)
>>
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 56e7519..ad9b4bc 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -4190,6 +4190,12 @@ 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;
>> +}
>> +
>>  static void enable_irq_window(struct kvm_vcpu *vcpu)
>>  {
>>  	u32 cpu_based_vm_exec_control;
>> @@ -4315,6 +4321,28 @@ static void vmx_set_nmi_mask(struct kvm_vcpu *vcpu, bool masked)
>>  
>>  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 ||
>> +		    vmcs_read32(GUEST_ACTIVITY_STATE) ==
>> +			   GUEST_ACTIVITY_WAIT_SIPI)
> The same is true for interrupt too,

Yes, but aren't we already waiting with interrupts disabled in that state?

> but I do not think that we should allow
> nested guest directly enter GUEST_ACTIVITY_WAIT_SIPI state or any other
> state except ACTIVE. They should be emulated instead.

Well, they aren't emulated yet but directly applied. So I think the
patch is correct in the current context at least.

What negative effects do you expect from entering those states with L2?

> From quick look at
> the spec it looks like external interrupts do not cause VMEXIT while
> vcpu is in GUEST_ACTIVITY_WAIT_SIPI.

Yes, see above.

Jan



[-- Attachment #2: OpenPGP digital signature --]
[-- Type: application/pgp-signature, Size: 263 bytes --]