Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.35 as permitted sender) client-ip=23.128.96.35;
Date:   Thu, 16 Nov 2023 14:29:02 -0800
In-Reply-To: <c9f65fc1-ab55-4959-a8ec-390aee51ee3a@intel.com>
Mime-Version: 1.0
References: <20231110235528.1561679-1-seanjc@google.com> <20231110235528.1561679-4-seanjc@google.com>
 <c9f65fc1-ab55-4959-a8ec-390aee51ee3a@intel.com>
Message-ID: <ZVaXroTZQi1IcTvm@google.com>
Subject: Re: [PATCH 3/9] KVM: x86: Initialize guest cpu_caps based on guest CPUID
From:   Sean Christopherson <seanjc@google.com>
To:     Weijiang Yang <weijiang.yang@intel.com>
Cc:     Paolo Bonzini <pbonzini@redhat.com>,
        "kvm@vger.kernel.org" <kvm@vger.kernel.org>,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
        Maxim Levitsky <mlevitsk@redhat.com>
Content-Type: text/plain; charset="utf-8"
Content-Transfer-Encoding: quoted-printable
Precedence: bulk

On Thu, Nov 16, 2023, Weijiang Yang wrote:
> On 11/11/2023 7:55 AM, Sean Christopherson wrote:
>=20
> [...]
>=20
> > -static __always_inline void guest_cpu_cap_check_and_set(struct kvm_vcp=
u *vcpu,
> > -							unsigned int x86_feature)
> > +static __always_inline void guest_cpu_cap_clear(struct kvm_vcpu *vcpu,
> > +						unsigned int x86_feature)
> >   {
> > -	if (kvm_cpu_cap_has(x86_feature) && guest_cpuid_has(vcpu, x86_feature=
))
> > +	unsigned int x86_leaf =3D __feature_leaf(x86_feature);
> > +
> > +	reverse_cpuid_check(x86_leaf);
> > +	vcpu->arch.cpu_caps[x86_leaf] &=3D ~__feature_bit(x86_feature);
> > +}
> > +
> > +static __always_inline void guest_cpu_cap_change(struct kvm_vcpu *vcpu=
,
> > +						 unsigned int x86_feature,
> > +						 bool guest_has_cap)
> > +{
> > +	if (guest_has_cap)
> >   		guest_cpu_cap_set(vcpu, x86_feature);
> > +	else
> > +		guest_cpu_cap_clear(vcpu, x86_feature);
> > +}
>=20
> I don't see any necessity to add 3 functions, i.e., guest_cpu_cap_{set, c=
lear, change}, for

I want to have equivalents to the cpuid_entry_*() APIs so that we don't end=
 up
with two different sets of names.  And the clear() API already has a second=
 user.

> guest_cpu_cap update. IMHO one function is enough, e.g,:

Hrm, I open coded the OR/AND logic in cpuid_entry_change() to try to force =
CMOV
instead of Jcc.  That honestly seems like a pointless optimization.  I woul=
d
rather use the helpers, which is less code.

> static __always_inline void guest_cpu_cap_update(struct kvm_vcpu *vcpu,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 unsigned=
 int x86_feature,
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=
=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 bool gue=
st_has_cap)
> {
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 unsigned int x86_leaf =3D __fe=
ature_leaf(x86_feature);
>=20
> reverse_cpuid_check(x86_leaf);
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0 if (guest_has_cap)
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0 vcpu->arch.cpu_caps[x86_leaf] |=3D __feature_bit(x86_fea=
ture);
> else
> =C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=C2=A0=
=C2=A0=C2=A0=C2=A0 vcpu->arch.cpu_caps[x86_leaf] &=3D ~__feature_bit(x86_fe=
ature);
> }
>=20
> > +
> > +static __always_inline void guest_cpu_cap_restrict(struct kvm_vcpu *vc=
pu,
> > +						   unsigned int x86_feature)
> > +{
> > +	if (!kvm_cpu_cap_has(x86_feature))
> > +		guest_cpu_cap_clear(vcpu, x86_feature);
> >   }
>=20
> _restrict is not clear to me for what the function actually does -- it
> conditionally clears guest cap depending on KVM support of the feature.
>=20
> How about renaming it to guest_cpu_cap_sync()?

"sync" isn't correct because it's not synchronizing with KVM's capabilitiy,=
 e.g.
the guest capability will remaing unset if the guest CPUID bit is clear but=
 the
KVM capability is available.

How about constrain()?