Re: [PATCH V9 03/18] perf/x86/intel: Handle guest PEBS overflow PMI for KVM guest

["Zhu, Lingshan" <lingshan.zhu@intel.com>]

On 8/5/2021 9:15 AM, Liuxiangdong wrote:
>
>
> On 2021/7/22 13:41, Zhu Lingshan wrote:
>> From: Like Xu <like.xu@linux.intel.com>
>>
>> With PEBS virtualization, the guest PEBS records get delivered to the
>> guest DS, and the host pmi handler uses perf_guest_cbs->is_in_guest()
>> to distinguish whether the PMI comes from the guest code like Intel PT.
>>
>> No matter how many guest PEBS counters are overflowed, only triggering
>> one fake event is enough. The fake event causes the KVM PMI callback to
>> be called, thereby injecting the PEBS overflow PMI into the guest.
>>
>> KVM may inject the PMI with BUFFER_OVF set, even if the guest DS is
>> empty. That should really be harmless. Thus guest PEBS handler would
>> retrieve the correct information from its own PEBS records buffer.
>>
>> Originally-by: Andi Kleen <ak@linux.intel.com>
>> Co-developed-by: Kan Liang <kan.liang@linux.intel.com>
>> Signed-off-by: Kan Liang <kan.liang@linux.intel.com>
>> Signed-off-by: Like Xu <like.xu@linux.intel.com>
>> Signed-off-by: Zhu Lingshan <lingshan.zhu@intel.com>
>> ---
>>   arch/x86/events/intel/core.c | 45 ++++++++++++++++++++++++++++++++++++
>>   1 file changed, 45 insertions(+)
>>
>> diff --git a/arch/x86/events/intel/core.c b/arch/x86/events/intel/core.c
>> index da835f5a37e2..2eceb73cd303 100644
>> --- a/arch/x86/events/intel/core.c
>> +++ b/arch/x86/events/intel/core.c
>> @@ -2783,6 +2783,50 @@ static void intel_pmu_reset(void)
>>   }
>>     DECLARE_STATIC_CALL(x86_guest_handle_intel_pt_intr, 
>> *(perf_guest_cbs->handle_intel_pt_intr));
>> +DECLARE_STATIC_CALL(x86_guest_state, *(perf_guest_cbs->state));
>> +
>> +/*
>> + * We may be running with guest PEBS events created by KVM, and the
>> + * PEBS records are logged into the guest's DS and invisible to host.
>> + *
>> + * In the case of guest PEBS overflow, we only trigger a fake event
>> + * to emulate the PEBS overflow PMI for guest PBES counters in KVM.
>> + * The guest will then vm-entry and check the guest DS area to read
>> + * the guest PEBS records.
>> + *
>> + * The contents and other behavior of the guest event do not matter.
>> + */
>> +static void x86_pmu_handle_guest_pebs(struct pt_regs *regs,
>> +                      struct perf_sample_data *data)
>> +{
>> +    struct cpu_hw_events *cpuc = this_cpu_ptr(&cpu_hw_events);
>> +    u64 guest_pebs_idxs = cpuc->pebs_enabled & 
>> ~cpuc->intel_ctrl_host_mask;
>
> guest_pebs_idxs has been defined here.
>
>> +    struct perf_event *event = NULL;
>> +    unsigned int guest = 0;
>> +    int bit;
>> +
>> +    guest = static_call(x86_guest_state)();
>> +    if (!(guest & PERF_GUEST_ACTIVE))
>> +        return;
>> +
>> +    if (!x86_pmu.pebs_vmx || !x86_pmu.pebs_active ||
>> +        !(cpuc->pebs_enabled & ~cpuc->intel_ctrl_host_mask))
>> +        return;
>> +
> Why not use guest_pebs_idxs?
>
> +    if (!x86_pmu.pebs_vmx || !x86_pmu.pebs_active ||
> +        !guest_pebs_idxs)
> +        return;
Thanks, I have apply this change in V10

Thanks
>
>
>> + for_each_set_bit(bit, (unsigned long *)&guest_pebs_idxs,
>> +             INTEL_PMC_IDX_FIXED + x86_pmu.num_counters_fixed) {
>> +        event = cpuc->events[bit];
>> +        if (!event->attr.precise_ip)
>> +            continue;
>> +
>> +        perf_sample_data_init(data, 0, event->hw.last_period);
>> +        if (perf_event_overflow(event, data, regs))
>> +            x86_pmu_stop(event, 0);
>> +
>> +        /* Inject one fake event is enough. */
>> +        break;
>> +    }
>> +}
>>     static int handle_pmi_common(struct pt_regs *regs, u64 status)
>>   {
>> @@ -2835,6 +2879,7 @@ static int handle_pmi_common(struct pt_regs 
>> *regs, u64 status)
>>           u64 pebs_enabled = cpuc->pebs_enabled;
>>             handled++;
>> +        x86_pmu_handle_guest_pebs(regs, &data);
>>           x86_pmu.drain_pebs(regs, &data);
>>           status &= intel_ctrl | GLOBAL_STATUS_TRACE_TOPAPMI;
>