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

From: Sean Christopherson
Date: Mon May 04 2020 - 10:29:40 EST


On Mon, May 04, 2020 at 03:12:36PM +0200, Paolo Bonzini wrote:
> On 04/05/20 14:01, Alexander Graf wrote:
> > I like the WARN_ON :). It should be almost free during execution, but
> > helps us catch problems early.
>
> Yes, it's nice. I didn't mind the "buddy" argument either, but if we're
> going to get a bool I prefer positive logic so I'd like to squash this:

I don't love need_ibpb as a param name, it doesn't provide any information
as to why the IBPB is needed. But, I can't come up with anything better
that isn't absurdly long because e.g. "different_guest" isn't necessarily
true in the vmx_vcpu_load() path.

What about going the @buddy route and adding the comment and WARN in
vmx_vcpu_load_vmcs()? E.g.

prev = per_cpu(current_vmcs, cpu);
if (prev != vmx->loaded_vmcs->vmcs) {
per_cpu(current_vmcs, cpu) = vmx->loaded_vmcs->vmcs;
vmcs_load(vmx->loaded_vmcs->vmcs);

/*
* No indirect branch prediction barrier needed when switching
* the active VMCS within a guest, e.g. on nested VM-Enter.
* The L1 VMM can protect itself with retpolines, IBPB or IBRS.
*/
if (!buddy || WARN_ON_ONCE(buddy->vmcs != prev))
indirect_branch_prediction_barrier();
}

> diff --git a/arch/x86/kvm/vmx/nested.c b/arch/x86/kvm/vmx/nested.c
> index b57420f3dd8f..299393750a18 100644
> --- a/arch/x86/kvm/vmx/nested.c
> +++ b/arch/x86/kvm/vmx/nested.c
> @@ -304,7 +304,13 @@ static void vmx_switch_vmcs(struct kvm_vcpu *vcpu, struct loaded_vmcs *vmcs)
> 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, true);
> +
> + /*
> + * This is the same guest from our point of view, so no
> + * indirect branch prediction barrier is needed. The L1
> + * guest can protect itself with retpolines, IBPB or IBRS.
> + */
> + vmx_vcpu_load_vmcs(vcpu, cpu, false);
> 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 669e14947ba9..0f9c8d2dd7f6 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, bool nested_switch)
> +void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, bool need_ibpb)
> {
> struct vcpu_vmx *vmx = to_vmx(vcpu);
> bool already_loaded = vmx->loaded_vmcs->cpu == cpu;
> @@ -1336,7 +1336,7 @@ void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, bool nested_switch)
> 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);
> - if (!nested_switch)
> + if (need_ibpb)
> indirect_branch_prediction_barrier();
> }
>
> @@ -1378,7 +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, false);
> + vmx_vcpu_load_vmcs(vcpu, cpu, true);
>
> vmx_vcpu_pi_load(vcpu, cpu);
>
> diff --git a/arch/x86/kvm/vmx/vmx.h b/arch/x86/kvm/vmx/vmx.h
> index fa61dc802183..e584ee9b3e94 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, bool nested_switch);
> +void vmx_vcpu_load_vmcs(struct kvm_vcpu *vcpu, int cpu, bool need_ibpb);
> void vmx_vcpu_load(struct kvm_vcpu *vcpu, int cpu);
> int allocate_vpid(void);
> void free_vpid(int vpid);
>
>
> Paolo
>