On 3/11/2021 7:34 AM, Peter Zijlstra wrote:
On Wed, Mar 10, 2021 at 08:37:52AM -0800, [email protected] wrote:@@ -2092,9 +2105,37 @@ static int __init init_hw_perf_events(void) if (err) goto out1;- err = perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW);- if (err) - goto out2; + if (!is_hybrid()) { + err = perf_pmu_register(&pmu, "cpu", PERF_TYPE_RAW); + if (err) + goto out2; + } else { + u8 cpu_type = get_hybrid_cpu_type(smp_processor_id()); + struct x86_hybrid_pmu *hybrid_pmu; + int i; + + for (i = 0; i < x86_pmu.num_hybrid_pmus; i++) { + hybrid_pmu = &x86_pmu.hybrid_pmu[i]; + + hybrid_pmu->pmu = pmu; + hybrid_pmu->pmu.type = -1; + hybrid_pmu->pmu.attr_update = x86_pmu.attr_update; + hybrid_pmu->pmu.capabilities |= PERF_PMU_CAP_HETEROGENEOUS_CPUS; + + /* Only register the PMU for the boot CPU */Why ?! > AFAICT we could register them all here. That instantly fixes that CPU_STARTING / CPU_DEAD fail elsewhere in this patch.
It's possible that all CPUs of a certain type all offline, but I cannot know the information here, because the boot CPU is the only online CPU. I don't know the status of the other CPUs.
If we unconditionally register all PMUs, users may see a PMU in /sys/devices, but they cannot use it, because there is no available CPU.
Is it acceptable that registering an empty PMU? Thanks, Kan
