Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 2620:137:e000::1:20 as permitted sender) client-ip=2620:137:e000::1:20;
Message-ID: <f9be9f7293ed366d750fa3fa7bb257d4f1b03891.camel@redhat.com>
Subject: Re: [PATCH v3 11/25] KVM: VMX: Get rid of eVMCS specific VMX
 controls sanitization
From:   Maxim Levitsky <mlevitsk@redhat.com>
To:     Vitaly Kuznetsov <vkuznets@redhat.com>, kvm@vger.kernel.org
Cc:     Anirudh Rayabharam <anrayabh@linux.microsoft.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        linux-hyperv@vger.kernel.org, linux-kernel@vger.kernel.org,
        Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>
Date:   Tue, 12 Jul 2022 15:16:31 +0300
In-Reply-To: <877d4iplyp.fsf@redhat.com>
References: <20220708144223.610080-1-vkuznets@redhat.com>
         <20220708144223.610080-12-vkuznets@redhat.com>
         <f1d030d7db4aaf3075fe625799b99ae335fc9f60.camel@redhat.com>
         <877d4iplyp.fsf@redhat.com>
Content-Type: text/plain; charset="UTF-8"
User-Agent: Evolution 3.40.4 (3.40.4-5.fc34) 
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

On Tue, 2022-07-12 at 14:14 +0200, Vitaly Kuznetsov wrote:
> Maxim Levitsky <mlevitsk@redhat.com> writes:
> 
> > On Fri, 2022-07-08 at 16:42 +0200, Vitaly Kuznetsov wrote:
> > > With the updated eVMCSv1 definition, there's no known 'problematic'
> > > controls which are exposed in VMX control MSRs but are not present in
> > > eVMCSv1. Get rid of VMX control MSRs filtering for KVM on Hyper-V.
> > 
> > If I understand correctly we are taking about running KVM as a nested guest of Hyper-V here:
> > 
> > Don't we need to check the new CPUID bit and only then use the new fields of eVMCS,
> > aka check that the 'cpu' supports the updated eVMCS version?
> > 
> 
> I've checked various Hyper-V versions available around and it seems
> there's no need for that: these new features are exposed in VMX control
> MSRs only when the updated eVMCS is supported.

Makes sense now. Might be worth a comment somewhere.

Best regards,
	Maxim Levitsky

> 
> We can, in theory, preserve the filtering for non-updated eVMCS verison
> but I'd vote for putting a WARN_ON() or something around: we can
> eventually get rid of it in case we don't get any reports.
> 
> > Best regards,
> >         Maxim Levitsky
> > 
> > 
> > 
> > > 
> > > Note: VMX control MSRs filtering for Hyper-V on KVM
> > > (nested_evmcs_filter_control_msr()) stays as even the updated eVMCSv1
> > > definition doesn't have all the features implemented by KVM and some
> > > fields are still missing. Moreover, nested_evmcs_filter_control_msr()
> > > has to support the original eVMCSv1 version when VMM wishes so.
> > > 
> > > Signed-off-by: Vitaly Kuznetsov <vkuznets@redhat.com>
> > > ---
> > >  arch/x86/kvm/vmx/evmcs.c | 13 -------------
> > >  arch/x86/kvm/vmx/evmcs.h |  1 -
> > >  arch/x86/kvm/vmx/vmx.c   |  5 -----
> > >  3 files changed, 19 deletions(-)
> > > 
> > > diff --git a/arch/x86/kvm/vmx/evmcs.c b/arch/x86/kvm/vmx/evmcs.c
> > > index 52a53debd806..b5cfbf7d487b 100644
> > > --- a/arch/x86/kvm/vmx/evmcs.c
> > > +++ b/arch/x86/kvm/vmx/evmcs.c
> > > @@ -320,19 +320,6 @@ const struct evmcs_field vmcs_field_to_evmcs_1[] = {
> > >  };
> > >  const unsigned int nr_evmcs_1_fields = ARRAY_SIZE(vmcs_field_to_evmcs_1);
> > >  
> > > -#if IS_ENABLED(CONFIG_HYPERV)
> > > -__init void evmcs_sanitize_exec_ctrls(struct vmcs_config *vmcs_conf)
> > > -{
> > > -       vmcs_conf->cpu_based_exec_ctrl &= ~EVMCS1_UNSUPPORTED_EXEC_CTRL;
> > > -       vmcs_conf->pin_based_exec_ctrl &= ~EVMCS1_UNSUPPORTED_PINCTRL;
> > > -       vmcs_conf->cpu_based_2nd_exec_ctrl &= ~EVMCS1_UNSUPPORTED_2NDEXEC;
> > > -       vmcs_conf->cpu_based_3rd_exec_ctrl = 0;
> > > -
> > > -       vmcs_conf->vmexit_ctrl &= ~EVMCS1_UNSUPPORTED_VMEXIT_CTRL;
> > > -       vmcs_conf->vmentry_ctrl &= ~EVMCS1_UNSUPPORTED_VMENTRY_CTRL;
> > > -}
> > > -#endif
> > > -
> > >  bool nested_enlightened_vmentry(struct kvm_vcpu *vcpu, u64 *evmcs_gpa)
> > >  {
> > >         struct hv_vp_assist_page assist_page;
> > > diff --git a/arch/x86/kvm/vmx/evmcs.h b/arch/x86/kvm/vmx/evmcs.h
> > > index 4b809c79ae63..0feac101cce4 100644
> > > --- a/arch/x86/kvm/vmx/evmcs.h
> > > +++ b/arch/x86/kvm/vmx/evmcs.h
> > > @@ -203,7 +203,6 @@ static inline void evmcs_load(u64 phys_addr)
> > >         vp_ap->enlighten_vmentry = 1;
> > >  }
> > >  
> > > -__init void evmcs_sanitize_exec_ctrls(struct vmcs_config *vmcs_conf);
> > >  #else /* !IS_ENABLED(CONFIG_HYPERV) */
> > >  static __always_inline void evmcs_write64(unsigned long field, u64 value) {}
> > >  static inline void evmcs_write32(unsigned long field, u32 value) {}
> > > diff --git a/arch/x86/kvm/vmx/vmx.c b/arch/x86/kvm/vmx/vmx.c
> > > index b4915d841357..dd905ad72637 100644
> > > --- a/arch/x86/kvm/vmx/vmx.c
> > > +++ b/arch/x86/kvm/vmx/vmx.c
> > > @@ -2689,11 +2689,6 @@ static __init int setup_vmcs_config(struct vmcs_config *vmcs_conf,
> > >         vmcs_conf->vmexit_ctrl         = _vmexit_control;
> > >         vmcs_conf->vmentry_ctrl        = _vmentry_control;
> > >  
> > > -#if IS_ENABLED(CONFIG_HYPERV)
> > > -       if (enlightened_vmcs)
> > > -               evmcs_sanitize_exec_ctrls(vmcs_conf);
> > > -#endif
> > > -
> > >         return 0;
> > >  }
> > >  
> > 
> > 
>