Received-SPF: pass (google.com: domain of linux-kernel-owner@vger.kernel.org designates 23.128.96.34 as permitted sender) client-ip=23.128.96.34;
Date:   Wed,  4 Oct 2023 20:12:37 -0700
Mime-Version: 1.0
Message-ID: <20231005031237.1652871-1-jmattson@google.com>
Subject: [PATCH v2] x86: KVM: Add feature flag for CPUID.80000021H:EAX[bit 1]
From:   Jim Mattson <jmattson@google.com>
To:     kvm@vger.kernel.org, linux-kernel@vger.kernel.org,
        Pawan Gupta <pawan.kumar.gupta@linux.intel.com>,
        Jiaxi Chen <jiaxi.chen@linux.intel.com>,
        Kim Phillips <kim.phillips@amd.com>,
        Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        "H. Peter Anvin" <hpa@zytor.com>, x86@kernel.org,
        Dave Hansen <dave.hansen@linux.intel.com>,
        Borislav Petkov <bp@alien8.de>, Ingo Molnar <mingo@redhat.com>,
        Thomas Gleixner <tglx@linutronix.de>
Cc:     Jim Mattson <jmattson@google.com>
Content-Type: text/plain; charset="UTF-8"
Precedence: bulk

Define an X86_FEATURE_* flag for CPUID.80000021H:EAX.[bit 1], and
advertise the feature to userspace via KVM_GET_SUPPORTED_CPUID.

Per AMD's "Processor Programming Reference (PPR) for AMD Family 19h
Model 61h, Revision B1 Processors (56713-B1-PUB)," this CPUID bit
indicates that a WRMSR to MSR_FS_BASE, MSR_GS_BASE, or
MSR_KERNEL_GS_BASE is non-serializing. This is a change in previously
architected behavior.

Effectively, this CPUID bit is a "defeature" bit, or a reverse
polarity feature bit. When this CPUID bit is clear, the feature
(serialization on WRMSR to any of these three MSRs) is available. When
this CPUID bit is set, the feature is not available.

KVM_GET_SUPPORTED_CPUID must pass this bit through from the underlying
hardware, if it is set. Leaving the bit clear claims that WRMSR to
these three MSRs will be serializing in a guest running under
KVM. That isn't true. Though KVM could emulate the feature by
intercepting writes to the specified MSRs, it does not do so
today. The guest is allowed direct read/write access to these MSRs
without interception, so the innate hardware behavior is preserved
under KVM.

Signed-off-by: Jim Mattson <jmattson@google.com>
---

v1 -> v2: Added justification for this change to the commit message,
          tweaked the macro name and comment in cpufeatures.h for
	  improved clarity.

 arch/x86/include/asm/cpufeatures.h | 1 +
 arch/x86/kvm/cpuid.c               | 3 ++-
 2 files changed, 3 insertions(+), 1 deletion(-)

diff --git a/arch/x86/include/asm/cpufeatures.h b/arch/x86/include/asm/cpufeatures.h
index 58cb9495e40f..4af140cf5719 100644
--- a/arch/x86/include/asm/cpufeatures.h
+++ b/arch/x86/include/asm/cpufeatures.h
@@ -443,6 +443,7 @@
 
 /* AMD-defined Extended Feature 2 EAX, CPUID level 0x80000021 (EAX), word 20 */
 #define X86_FEATURE_NO_NESTED_DATA_BP	(20*32+ 0) /* "" No Nested Data Breakpoints */
+#define X86_FEATURE_WRMSR_XX_BASE_NS	(20*32+ 1) /* "" WRMSR to {FS,GS,KERNEL_GS}_BASE is non-serializing */
 #define X86_FEATURE_LFENCE_RDTSC	(20*32+ 2) /* "" LFENCE always serializing / synchronizes RDTSC */
 #define X86_FEATURE_NULL_SEL_CLR_BASE	(20*32+ 6) /* "" Null Selector Clears Base */
 #define X86_FEATURE_AUTOIBRS		(20*32+ 8) /* "" Automatic IBRS */
diff --git a/arch/x86/kvm/cpuid.c b/arch/x86/kvm/cpuid.c
index 0544e30b4946..93241b33e36f 100644
--- a/arch/x86/kvm/cpuid.c
+++ b/arch/x86/kvm/cpuid.c
@@ -761,7 +761,8 @@ void kvm_set_cpu_caps(void)
 
 	kvm_cpu_cap_mask(CPUID_8000_0021_EAX,
 		F(NO_NESTED_DATA_BP) | F(LFENCE_RDTSC) | 0 /* SmmPgCfgLock */ |
-		F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */
+		F(NULL_SEL_CLR_BASE) | F(AUTOIBRS) | 0 /* PrefetchCtlMsr */ |
+		F(WRMSR_XX_BASE_NS)
 	);
 
 	if (cpu_feature_enabled(X86_FEATURE_SRSO_NO))
-- 
2.42.0.582.g8ccd20d70d-goog