Re: [PATCH v5 1/3] KVM: X86: Save&restore the triple fault request

[Chenyi Qiang <chenyi.qiang@intel.com>]

On 4/6/2022 7:31 AM, Sean Christopherson wrote:
> On Fri, Mar 18, 2022, Chenyi Qiang wrote:
>> For the triple fault sythesized by KVM, e.g. the RSM path or
>> nested_vmx_abort(), if KVM exits to userspace before the request is
>> serviced, userspace could migrate the VM and lose the triple fault.
>> Fix this issue by adding a new event KVM_VCPUEVENT_TRIPLE_FAULT in
>> get/set_vcpu_events() to track the triple fault request.
>>
>> Signed-off-by: Chenyi Qiang <chenyi.qiang@intel.com>
>> ---
>>   Documentation/virt/kvm/api.rst  | 6 ++++++
>>   arch/x86/include/uapi/asm/kvm.h | 1 +
>>   arch/x86/kvm/x86.c              | 9 ++++++++-
>>   3 files changed, 15 insertions(+), 1 deletion(-)
>>
>> diff --git a/Documentation/virt/kvm/api.rst b/Documentation/virt/kvm/api.rst
>> index 691ff84444bd..9682b0a438bd 100644
>> --- a/Documentation/virt/kvm/api.rst
>> +++ b/Documentation/virt/kvm/api.rst
>> @@ -1146,6 +1146,9 @@ The following bits are defined in the flags field:
>>     fields contain a valid state. This bit will be set whenever
>>     KVM_CAP_EXCEPTION_PAYLOAD is enabled.
>>   
>> +- KVM_VCPUEVENT_TRIPLE_FAULT may be set to signal that there's a
>> +  triple fault request waiting to be serviced.
> 
> Please avoid "request" in the docs, as before, that's a KVM implemenation detail.
> For this one, maybe "there's a pending triple fault event"?
> 
>> +
>>   ARM/ARM64:
>>   ^^^^^^^^^^
>>   
>> @@ -1241,6 +1244,9 @@ can be set in the flags field to signal that the
>>   exception_has_payload, exception_payload, and exception.pending fields
>>   contain a valid state and shall be written into the VCPU.
>>   
>> +KVM_VCPUEVENT_TRIPLE_FAULT can be set in flags field to signal that a
>> +triple fault request should be made.
> 
> 
> And here, "to signal that KVM should synthesize a triple fault for the guest"?
> 
>> +
>>   ARM/ARM64:
>>   ^^^^^^^^^^
>>   
>> diff --git a/arch/x86/include/uapi/asm/kvm.h b/arch/x86/include/uapi/asm/kvm.h
>> index bf6e96011dfe..d8ef0d993e86 100644
>> --- a/arch/x86/include/uapi/asm/kvm.h
>> +++ b/arch/x86/include/uapi/asm/kvm.h
>> @@ -325,6 +325,7 @@ struct kvm_reinject_control {
>>   #define KVM_VCPUEVENT_VALID_SHADOW	0x00000004
>>   #define KVM_VCPUEVENT_VALID_SMM		0x00000008
>>   #define KVM_VCPUEVENT_VALID_PAYLOAD	0x00000010
>> +#define KVM_VCPUEVENT_TRIPLE_FAULT	0x00000020
>>   
>>   /* Interrupt shadow states */
>>   #define KVM_X86_SHADOW_INT_MOV_SS	0x01
>> diff --git a/arch/x86/kvm/x86.c b/arch/x86/kvm/x86.c
>> index 4fa4d8269e5b..fee402a700df 100644
>> --- a/arch/x86/kvm/x86.c
>> +++ b/arch/x86/kvm/x86.c
>> @@ -4891,6 +4891,9 @@ static void kvm_vcpu_ioctl_x86_get_vcpu_events(struct kvm_vcpu *vcpu,
>>   	if (vcpu->kvm->arch.exception_payload_enabled)
>>   		events->flags |= KVM_VCPUEVENT_VALID_PAYLOAD;
>>   
>> +	if (kvm_check_request(KVM_REQ_TRIPLE_FAULT, vcpu))
>> +		events->flags |= KVM_VCPUEVENT_TRIPLE_FAULT;
>> +
>>   	memset(&events->reserved, 0, sizeof(events->reserved));
>>   }
>>   
>> @@ -4903,7 +4906,8 @@ static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu,
>>   			      | KVM_VCPUEVENT_VALID_SIPI_VECTOR
>>   			      | KVM_VCPUEVENT_VALID_SHADOW
>>   			      | KVM_VCPUEVENT_VALID_SMM
>> -			      | KVM_VCPUEVENT_VALID_PAYLOAD))
>> +			      | KVM_VCPUEVENT_VALID_PAYLOAD
>> +			      | KVM_VCPUEVENT_TRIPLE_FAULT))
>>   		return -EINVAL;
>>   
>>   	if (events->flags & KVM_VCPUEVENT_VALID_PAYLOAD) {
>> @@ -4976,6 +4980,9 @@ static int kvm_vcpu_ioctl_x86_set_vcpu_events(struct kvm_vcpu *vcpu,
>>   		}
>>   	}
>>   
>> +	if (events->flags & KVM_VCPUEVENT_TRIPLE_FAULT)
>> +		kvm_make_request(KVM_REQ_TRIPLE_FAULT, vcpu);
>> +
>>   	kvm_make_request(KVM_REQ_EVENT, vcpu);
> 
> Looks correct, but this really needs a selftest, at least for the SET path since
> the intent is to use that for the NOTIFY handling.  Doesn't need to be super fancy,
> e.g. do port I/O from L2, inject a triple fault, and verify L1 sees the appropriate
> exit.
> 
> Aha!  And for the GET path, abuse KVM_X86_SET_MCE with CR4.MCE=0 to coerce KVM into
> making a KVM_REQ_TRIPLE_FAULT, that way there's no need to try and hit a timing
> window to intercept the request.

OK, will cook a selftest to verify it.