Received-SPF: softfail (google.com: domain of transitioning linux-kernel-owner@vger.kernel.org does not designate 23.128.96.19 as permitted sender) client-ip=23.128.96.19;
From:   Kyle Huey <me@kylehuey.com>
To:     linux-kernel@vger.kernel.org
Cc:     Paolo Bonzini <pbonzini@redhat.com>,
        Sean Christopherson <seanjc@google.com>,
        Vitaly Kuznetsov <vkuznets@redhat.com>,
        Wanpeng Li <wanpengli@tencent.com>,
        Jim Mattson <jmattson@google.com>,
        Joerg Roedel <joro@8bytes.org>,
        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>, kvm@vger.kernel.org,
        Robert O'Callahan <robert@ocallahan.org>,
        Keno Fischer <keno@juliacomputing.com>,
        Kyle Huey <me@kylehuey.com>
Subject: [PATCH] KVM: x86/svm: Account for family 17h event renumberings in amd_pmc_perf_hw_id
Date:   Mon,  2 May 2022 22:01:36 -0700
Message-Id: <20220503050136.86298-1-khuey@kylehuey.com>
MIME-Version: 1.0
Content-Transfer-Encoding: 8bit
Precedence: bulk

From: Kyle Huey <me@kylehuey.com>

Zen renumbered some of the performance counters that correspond to the
well known events in perf_hw_id. This code in KVM was never updated for
that, so guest that attempt to use counters on Zen that correspond to the
pre-Zen perf_hw_id values will silently receive the wrong values.

This has been observed in the wild with rr[0] when running in Zen 3
guests. rr uses the retired conditional branch counter 00d1 which is
incorrectly recognized by KVM as PERF_COUNT_HW_STALLED_CYCLES_BACKEND.

[0] https://rr-project.org/

Signed-off-by: Kyle Huey <me@kylehuey.com>
---
 arch/x86/kvm/svm/pmu.c | 28 +++++++++++++++++++++++++---
 1 file changed, 25 insertions(+), 3 deletions(-)

diff --git a/arch/x86/kvm/svm/pmu.c b/arch/x86/kvm/svm/pmu.c
index b14860863c39..43d24746c0a7 100644
--- a/arch/x86/kvm/svm/pmu.c
+++ b/arch/x86/kvm/svm/pmu.c
@@ -45,6 +45,22 @@ static struct kvm_event_hw_type_mapping amd_event_mapping[] = {
 	[7] = { 0xd1, 0x00, PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
 };
 
+/* duplicated from amd_f17h_perfmon_event_map. */
+static struct kvm_event_hw_type_mapping amd_f17h_event_mapping[] = {
+	[0] = { 0x76, 0x00, PERF_COUNT_HW_CPU_CYCLES },
+	[1] = { 0xc0, 0x00, PERF_COUNT_HW_INSTRUCTIONS },
+	[2] = { 0x60, 0xff, PERF_COUNT_HW_CACHE_REFERENCES },
+	[3] = { 0x64, 0x09, PERF_COUNT_HW_CACHE_MISSES },
+	[4] = { 0xc2, 0x00, PERF_COUNT_HW_BRANCH_INSTRUCTIONS },
+	[5] = { 0xc3, 0x00, PERF_COUNT_HW_BRANCH_MISSES },
+	[6] = { 0x87, 0x02, PERF_COUNT_HW_STALLED_CYCLES_FRONTEND },
+	[7] = { 0x87, 0x01, PERF_COUNT_HW_STALLED_CYCLES_BACKEND },
+};
+
+/* amd_pmc_perf_hw_id depends on these being the same size */
+static_assert(ARRAY_SIZE(amd_event_mapping) ==
+	     ARRAY_SIZE(amd_f17h_event_mapping));
+
 static unsigned int get_msr_base(struct kvm_pmu *pmu, enum pmu_type type)
 {
 	struct kvm_vcpu *vcpu = pmu_to_vcpu(pmu);
@@ -140,6 +156,7 @@ static inline struct kvm_pmc *get_gp_pmc_amd(struct kvm_pmu *pmu, u32 msr,
 
 static unsigned int amd_pmc_perf_hw_id(struct kvm_pmc *pmc)
 {
+	struct kvm_event_hw_type_mapping *event_mapping;
 	u8 event_select = pmc->eventsel & ARCH_PERFMON_EVENTSEL_EVENT;
 	u8 unit_mask = (pmc->eventsel & ARCH_PERFMON_EVENTSEL_UMASK) >> 8;
 	int i;
@@ -148,15 +165,20 @@ static unsigned int amd_pmc_perf_hw_id(struct kvm_pmc *pmc)
 	if (WARN_ON(pmc_is_fixed(pmc)))
 		return PERF_COUNT_HW_MAX;
 
+	if (boot_cpu_data.x86 >= 0x17)
+		event_mapping = amd_f17h_event_mapping;
+	else
+		event_mapping = amd_event_mapping;
+
 	for (i = 0; i < ARRAY_SIZE(amd_event_mapping); i++)
-		if (amd_event_mapping[i].eventsel == event_select
-		    && amd_event_mapping[i].unit_mask == unit_mask)
+		if (event_mapping[i].eventsel == event_select
+		    && event_mapping[i].unit_mask == unit_mask)
 			break;
 
 	if (i == ARRAY_SIZE(amd_event_mapping))
 		return PERF_COUNT_HW_MAX;
 
-	return amd_event_mapping[i].event_type;
+	return event_mapping[i].event_type;
 }
 
 /* check if a PMC is enabled by comparing it against global_ctrl bits. Because
-- 
2.36.0