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;
MIME-Version: 1.0
References: <20220203194308.2469117-1-jmattson@google.com> <20220203194308.2469117-2-jmattson@google.com>
 <e313b068-b5d6-dbc4-9894-02d759d2fe55@intel.com>
In-Reply-To: <e313b068-b5d6-dbc4-9894-02d759d2fe55@intel.com>
From:   Jim Mattson <jmattson@google.com>
Date:   Thu, 3 Feb 2022 12:50:53 -0800
Message-ID: <CALMp9eQ+vRQhbmTh1j4YT7AhGcvRxnmzttTbjN3UKHaijwpqAw@mail.gmail.com>
Subject: Re: [PATCH 2/2] x86/cpufeatures: Add macros for Intel's new fast rep
 string features
To:     Dave Hansen <dave.hansen@intel.com>
Cc:     Thomas Gleixner <tglx@linutronix.de>,
        Ingo Molnar <mingo@redhat.com>, Borislav Petkov <bp@alien8.de>,
        Dave Hansen <dave.hansen@linux.intel.com>, x86@kernel.org,
        "H. Peter Anvin" <hpa@zytor.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Tony Luck <tony.luck@intel.com>,
        Peter Zijlstra <peterz@infradead.org>,
        "Chang S. Bae" <chang.seok.bae@intel.com>,
        Juergen Gross <jgross@suse.com>,
        Andrew Cooper <andrew.cooper3@citrix.com>,
        Babu Moger <Babu.Moger@amd.com>,
        Jing Liu <jing2.liu@intel.com>,
        Sean Christopherson <seanjc@google.com>,
        linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

On Thu, Feb 3, 2022 at 12:25 PM Dave Hansen <dave.hansen@intel.com> wrote:
>
> On 2/3/22 11:43, Jim Mattson wrote:
> > Even if no one else cares, these features should be exposed to kvm
> > guests, and the code for the KVM_GET_SUPPORTED_CPUID ioctl is more
> > readable if the bits have corresponding X86_FEATURE macros.
>
> I went digging around KVM_GET_SUPPORTED_CPUID and didn't see any obvious
> unreadable things resulting from not having these defines.  Maybe I'm
> looking in the wrong spot.  Do you have a slightly more specific pointer?

These features are not currently enumerated by
KVM_GET_SUPPORTED_CPUID. They would be added as follows [gmail mangles
diffs; sorry]:

diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index 28be02adc669..11e621b6faac 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -599,7 +599,7 @@ void kvm_set_cpu_caps(void)
                kvm_cpu_cap_set(X86_FEATURE_SPEC_CTRL_SSBD);

        kvm_cpu_cap_mask(CPUID_7_1_EAX,
-               F(AVX_VNNI) | F(AVX512_BF16)
+                F(AVX_VNNI) | F(AVX512_BF16) | F(FZRM) | F(FSRS) | F(FSRC)
        );

        kvm_cpu_cap_mask(CPUID_D_1_EAX,

(Of course, we can always define these macros in that file, if that's
preferable.)

> > +#define X86_FEATURE_FZRM             (12*32+10) /* Fast zero-length REP MOVSB */
> > +#define X86_FEATURE_FSRS             (12*32+11) /* Fast short REP STOSB */
> > +#define X86_FEATURE_FSRC             (12*32+12) /* Fast short REP {CMPSB,SCASB} */
>
> If the use really is for inside the kernel, maybe we should hide them
> from cpuinfo:
>
> #define X86_FEATURE_FSRC                (12*32+12) /* "" Fast short REP {CMPSB,SCASB} */

That sounds good to me.