Re: [PATCH v2] KVM: nVMX: Skip IBPB when switching between vmcs01 and vmcs02

[Alexander Graf <graf@amazon.com>]

On 01.05.20 18:31, Sean Christopherson wrote:
> 
> Skip the Indirect Branch Prediction Barrier that is triggered on a VMCS
> switch when running with spectre_v2_user=on/auto if the switch is
> between two VMCSes in the same guest, i.e. between vmcs01 and vmcs02.
> The IBPB is intended to prevent one guest from attacking another, which
> is unnecessary in the nested case as it's the same guest from KVM's
> perspective.
> 
> This all but eliminates the overhead observed for nested VMX transitions
> when running with CONFIG_RETPOLINE=y and spectre_v2_user=on/auto, which
> can be significant, e.g. roughly 3x on current systems.
> 
> Reported-by: Alexander Graf <graf@amazon.com>
> Cc: KarimAllah Raslan <karahmed@amazon.de>
> Cc: stable@vger.kernel.org
> Fixes: 15d45071523d ("KVM/x86: Add IBPB support")
> Signed-off-by: Sean Christopherson <sean.j.christopherson@intel.com>
> ---
> 
> v2: Pass a boolean to indicate a nested VMCS switch and instead WARN if
>      the buddy VMCS is not already loaded.  [Alex]
> 
> Paolo, feel free to drop the WARN_ON_ONCE() if you think it's overkill.
> I'm 50/50 as to whether it's useful or just a waste of cycles.  Figured
> it'd be easier for you to delete a line of code while applying than to add
> and test a new WARN.

I like the WARN_ON :). It should be almost free during execution, but 
helps us catch problems early.

> 
>   arch/x86/kvm/vmx/nested.c | 3 ++-
>   arch/x86/kvm/vmx/vmx.c    | 7 ++++---
>   arch/x86/kvm/vmx/vmx.h    | 2 +-
>   3 files changed, 7 insertions(+), 5 deletions(-)
> 
> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index 2c36f3f53108..b57420f3dd8f 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -302,8 +302,9 @@ static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
> 
>          cpu = get_cpu();
>          prev = vmx->loaded_vmcs;
> +       WARN_ON_ONCE(prev->cpu != cpu || prev->vmcs != per_cpu(current_vmcs, cpu));
>          vmx->loaded_vmcs = vmcs;
> -       vmx_vcpu_load_vmcs(vcpu, cpu);
> +       vmx_vcpu_load_vmcs(vcpu, cpu, true);
>          vmx_sync_vmcs_host_state(vmx, prev);
>          put_cpu();
> 
> diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> index 3ab6ca6062ce..d3d57b7a67bd 100644
> --- a/arch/x86/kvm/vmx/vmx.c
> +++ b/arch/x86/kvm/vmx/vmx.c
> @@ -1311,7 +1311,7 @@ static void vmx_vcpu_pi_load(struct kvm_vcpu *vcpu, int cpu)
>                  pi_set_on(pi_desc);
>   }
> 
> -void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu)
> +void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, bool nested_switch)
>   {
>          struct vcpu_vmx *vmx = to_vmx(vcpu);
>          bool already_loaded = vmx->loaded_vmcs->cpu == cpu;
> @@ -1336,7 +1336,8 @@ void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu)
>          if (per_cpu(current_vmcs, cpu) != vmx->loaded_vmcs->vmcs) {
>                  per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs;
>                  vmcs_load(vmx->loaded_vmcs->vmcs);
> -               indirect_branch_prediction_barrier();

... however, this really needs an in-code comment to explain why it's 
safe not to flush the branch predictor cache here.


Alex

> +               if (!nested_switch)
> +                       indirect_branch_prediction_barrier();
>          }
> 
>          if (!already_loaded) {
> @@ -1377,7 +1378,7 @@ void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu)
>   {
>          struct vcpu_vmx *vmx = to_vmx(vcpu);
> 
> -       vmx_vcpu_load_vmcs(vcpu, cpu);
> +       vmx_vcpu_load_vmcs(vcpu, cpu, false);
> 
>          vmx_vcpu_pi_load(vcpu, cpu);
> 
> diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
> index b5e773267abe..fa61dc802183 100644
> --- a/arch/x86/kvm/vmx/vmx.h
> +++ b/arch/x86/kvm/vmx/vmx.h
> @@ -320,7 +320,7 @@ struct kvm_vmx {
>   };
> 
>   bool nested_vmx_allowed(struct kvm_vcpu *vcpu);
> -void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu);
> +void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, bool nested_switch);
>   void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
>   int allocate_vpid(void);
>   void free_vpid(int vpid);
> --
> 2.26.0
> 



Amazon Development Center Germany GmbH
Krausenstr. 38
10117 Berlin
Geschaeftsfuehrung: Christian Schlaeger, Jonathan Weiss
Eingetragen am Amtsgericht Charlottenburg unter HRB 149173 B
Sitz: Berlin
Ust-ID: DE 289 237 879