[PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Ladi Prosek]

The PML feature is not exposed to guests so we should not be forwarding
the vmexit either.

This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
enabled Windows Server 2016 in L1 on hardware that supports PML.

Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
Signed-off-by: Ladi Prosek <lprosek@redhat.com>
---
 arch/x86/kvm/vmx.c | 3 +++
 1 file changed, 3 insertions(+)

diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
index 2ee00db..6051832 100644
--- a/arch/x86/kvm/vmx.c
+++ b/arch/x86/kvm/vmx.c
@@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
 		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
 	case EXIT_REASON_PREEMPTION_TIMER:
 		return false;
+	case EXIT_REASON_PML_FULL:
+		/* We don't expose PML support to L1. */
+		return false;
 	default:
 		return true;
 	}
-- 
2.9.3

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Radim Krčmář]

2017-03-31 10:19+0200, Ladi Prosek:
> The PML feature is not exposed to guests so we should not be forwarding
> the vmexit either.
> 
> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
> enabled Windows Server 2016 in L1 on hardware that supports PML.
> 
> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
> Signed-off-by: Ladi Prosek <lprosek@redhat.com>
> ---

Good catch,

Reviewed-by: Radim Krčmář <rkrcmar@redhat.com>

>  arch/x86/kvm/vmx.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 2ee00db..6051832 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>  		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
>  	case EXIT_REASON_PREEMPTION_TIMER:
>  		return false;
> +	case EXIT_REASON_PML_FULL:
> +		/* We don't expose PML support to L1. */
> +		return false;
>  	default:
>  		return true;
>  	}
> -- 
> 2.9.3
> 

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[David Hildenbrand]

On 31.03.2017 10:19, Ladi Prosek wrote:
> The PML feature is not exposed to guests so we should not be forwarding
> the vmexit either.
> 
> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
> enabled Windows Server 2016 in L1 on hardware that supports PML.
> 
> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
> Signed-off-by: Ladi Prosek <lprosek@redhat.com>
> ---
>  arch/x86/kvm/vmx.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 2ee00db..6051832 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>  		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
>  	case EXIT_REASON_PREEMPTION_TIMER:
>  		return false;
> +	case EXIT_REASON_PML_FULL:
> +		/* We don't expose PML support to L1. */
> +		return false;
>  	default:
>  		return true;
>  	}
> 

Reviewed-by: David Hildenbrand <david@redhat.com>

-- 

Thanks,

David

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[David Hildenbrand]

On 31.03.2017 10:19, Ladi Prosek wrote:
> The PML feature is not exposed to guests so we should not be forwarding
> the vmexit either.

I wonder if a whitelist of vmexits would make more sense, compared to
defaulting to passing them through. But I don't know the history of this

> 
> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
> enabled Windows Server 2016 in L1 on hardware that supports PML.
> 
> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
> Signed-off-by: Ladi Prosek <lprosek@redhat.com>


-- 

Thanks,

David

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Paolo Bonzini]

On 03/04/2017 13:27, David Hildenbrand wrote:
>> The PML feature is not exposed to guests so we should not be forwarding
>> the vmexit either.
> 
> I wonder if a whitelist of vmexits would make more sense, compared to
> defaulting to passing them through. But I don't know the history of this

I checked the other day and it's really 50-50 between "return true" and
"return false" vmexits.  Adding a WARNing would probably make even more
sense. :)

Thanks,

Paolo

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[David Hildenbrand]

On 03.04.2017 13:43, Paolo Bonzini wrote:
> 
> 
> On 03/04/2017 13:27, David Hildenbrand wrote:
>>> The PML feature is not exposed to guests so we should not be forwarding
>>> the vmexit either.
>>
>> I wonder if a whitelist of vmexits would make more sense, compared to
>> defaulting to passing them through. But I don't know the history of this
> 
> I checked the other day and it's really 50-50 between "return true" and
> "return false" vmexits.  Adding a WARNing would probably make even more
> sense. :)
> 

Well, a whitelist will make sure that L1 really only sees what you want
it to see. As an alternative, specify all of them explicitly
(white+black list - should that currently already be done? haven't
checked it), default to "false" and print a WARNing (unhandled exit
code). But don't expose it to L1. (just what I have in mind).

> Thanks,
> 
> Paolo
> 


-- 

Thanks,

David

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Radim Krčmář]

2017-03-31 10:19+0200, Ladi Prosek:
> The PML feature is not exposed to guests so we should not be forwarding
> the vmexit either.
> 
> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
> enabled Windows Server 2016 in L1 on hardware that supports PML.
> 
> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
> Signed-off-by: Ladi Prosek <lprosek@redhat.com>
> ---

Applied to kvm/master, thanks.

>  arch/x86/kvm/vmx.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 2ee00db..6051832 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>  		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
>  	case EXIT_REASON_PREEMPTION_TIMER:
>  		return false;
> +	case EXIT_REASON_PML_FULL:
> +		/* We don't expose PML support to L1. */
> +		return false;
>  	default:
>  		return true;
>  	}
> -- 
> 2.9.3
> 

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Paolo Bonzini]

On 31/03/2017 10:19, Ladi Prosek wrote:
> The PML feature is not exposed to guests so we should not be forwarding
> the vmexit either.
> 
> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
> enabled Windows Server 2016 in L1 on hardware that supports PML.
> 
> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
> Signed-off-by: Ladi Prosek <lprosek@redhat.com>
> ---
>  arch/x86/kvm/vmx.c | 3 +++
>  1 file changed, 3 insertions(+)
> 
> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
> index 2ee00db..6051832 100644
> --- a/arch/x86/kvm/vmx.c
> +++ b/arch/x86/kvm/vmx.c
> @@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>  		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
>  	case EXIT_REASON_PREEMPTION_TIMER:
>  		return false;
> +	case EXIT_REASON_PML_FULL:
> +		/* We don't expose PML support to L1. */
> +		return false;
>  	default:
>  		return true;
>  	}
> 

This fell through the cracks, I'm queueing it now.

Paolo

Re: [PATCH] KVM: nVMX: do not leak PML full vmexit to L1

[Radim Krčmář]

2017-05-02 15:05+0200, Paolo Bonzini:
> On 31/03/2017 10:19, Ladi Prosek wrote:
>> The PML feature is not exposed to guests so we should not be forwarding
>> the vmexit either.
>> 
>> This commit fixes BSOD 0x20001 (HYPERVISOR_ERROR) when running Hyper-V
>> enabled Windows Server 2016 in L1 on hardware that supports PML.
>> 
>> Fixes: 843e4330573c ("KVM: VMX: Add PML support in VMX")
>> Signed-off-by: Ladi Prosek <lprosek@redhat.com>
>> ---
>>  arch/x86/kvm/vmx.c | 3 +++
>>  1 file changed, 3 insertions(+)
>> 
>> diff --git a/arch/x86/kvm/vmx.c b/arch/x86/kvm/vmx.c
>> index 2ee00db..6051832 100644
>> --- a/arch/x86/kvm/vmx.c
>> +++ b/arch/x86/kvm/vmx.c
>> @@ -8198,6 +8198,9 @@ static bool nested_vmx_exit_handled(struct kvm_vcpu *vcpu)
>>  		return nested_cpu_has2(vmcs12, SECONDARY_EXEC_XSAVES);
>>  	case EXIT_REASON_PREEMPTION_TIMER:
>>  		return false;
>> +	case EXIT_REASON_PML_FULL:
>> +		/* We don't expose PML support to L1. */
>> +		return false;
>>  	default:
>>  		return true;
>>  	}
>> 
> 
> This fell through the cracks, I'm queueing it now.

It is in Linus' tree as ab007cc94ff9, we are just missing it in queue
and next as they haven't merged yet.