[tip: perf/core] perf/x86/intel: Check perf metrics feature for each CPU

From: tip-bot2 for Kan Liang
Date: Mon Oct 05 2020 - 03:54:04 EST


The following commit has been merged into the perf/core branch of tip:

Commit-ID: 80a5ce116fc084e8a25d5a936617699e2931b611
Gitweb: https://git.kernel.org/tip/80a5ce116fc084e8a25d5a936617699e2931b611
Author: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
AuthorDate: Thu, 01 Oct 2020 14:17:11 -07:00
Committer: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
CommitterDate: Sat, 03 Oct 2020 16:30:56 +02:00

perf/x86/intel: Check perf metrics feature for each CPU

It might be possible that different CPUs have different CPU metrics on a
platform. In this case, writing the GLOBAL_CTRL_EN_PERF_METRICS bit to
the GLOBAL_CTRL register of a CPU, which doesn't support the TopDown
perf metrics feature, causes MSR access error.

Current TopDown perf metrics feature is enumerated using the boot CPU's
PERF_CAPABILITIES MSR. The MSR only indicates the boot CPU supports this
feature.

Check the PERF_CAPABILITIES MSR for each CPU. If any CPU doesn't support
the perf metrics feature, disable the feature globally.

Fixes: 59a854e2f3b9 ("perf/x86/intel: Support TopDown metrics on Ice Lake")
Signed-off-by: Kan Liang <kan.liang@xxxxxxxxxxxxxxx>
Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
Link: https://lkml.kernel.org/r/20201001211711.25708-1-kan.liang@xxxxxxxxxxxxxxx
---
arch/x86/events/intel/core.c | 11 +++++++++++
1 file changed, 11 insertions(+)

diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
index bdf28d2..7186098 100644
--- a/arch/x86/events/intel/core.c
+++ b/arch/x86/events/intel/core.c
@@ -4083,6 +4083,17 @@ static void intel_pmu_cpu_starting(int cpu)
if (x86_pmu.counter_freezing)
enable_counter_freeze();

+ /* Disable perf metrics if any added CPU doesn't support it. */
+ if (x86_pmu.intel_cap.perf_metrics) {
+ union perf_capabilities perf_cap;
+
+ rdmsrl(MSR_IA32_PERF_CAPABILITIES, perf_cap.capabilities);
+ if (!perf_cap.perf_metrics) {
+ x86_pmu.intel_cap.perf_metrics = 0;
+ x86_pmu.intel_ctrl &= ~(1ULL << GLOBAL_CTRL_EN_PERF_METRICS);
+ }
+ }
+
if (!cpuc->shared_regs)
return;