[PATCH v2] arm64: perf: Use only exclude_kernel attribute when kernel is running in HYP

[Ganapatrao Kulkarni]

On Thu, Apr 20, 2017 at 2:19 PM, Mark Rutland <mark.rutland at arm.com> wrote:
> On Wed, Apr 19, 2017 at 11:14:06PM +0530, Ganapatrao Kulkarni wrote:
>> commit d98ecda (arm64: perf: Count EL2 events if the kernel is running in HYP)
>> is returning error for perf syscall with mixed attribute set for exclude_kernel
>> and exclude_hv. This change is breaking some applications (observed with hhvm)
>> when ran on VHE enabled platforms.
>>
>> Adding fix to consider only exclude_kernel attribute when kernel is
>> running in HYP. Also adding sysfs file to notify the bhehaviour
>> of attribute exclude_hv.
>>
>> Signed-off-by: Ganapatrao Kulkarni <ganapatrao.kulkarni at cavium.com>
>> ---
>>
>> Changelog:
>>
>> V2:
>>  - Changes as per Will Deacon's suggestion.
>>
>> V1: Initial patch
>>
>>  arch/arm64/kernel/perf_event.c | 28 ++++++++++++++++++++++++----
>>  include/linux/perf/arm_pmu.h   |  1 +
>>  2 files changed, 25 insertions(+), 4 deletions(-)
>>
>> @@ -871,14 +890,13 @@ static int armv8pmu_set_event_filter(struct hw_perf_event *event,
>>
>>       if (attr->exclude_idle)
>>               return -EPERM;
>> -     if (is_kernel_in_hyp_mode() &&
>> -         attr->exclude_kernel != attr->exclude_hv)
>> -             return -EINVAL;
>> +     if (is_kernel_in_hyp_mode() && !attr->exclude_kernel)
>> +             config_base |= ARMV8_PMU_INCLUDE_EL2;
>>       if (attr->exclude_user)
>>               config_base |= ARMV8_PMU_EXCLUDE_EL0;
>>       if (!is_kernel_in_hyp_mode() && attr->exclude_kernel)
>>               config_base |= ARMV8_PMU_EXCLUDE_EL1;
>> -     if (!attr->exclude_hv)
>> +     if (!is_kernel_in_hyp_mode() && !attr->exclude_hv)
>>               config_base |= ARMV8_PMU_INCLUDE_EL2;
>
> This isn't quite what Will suggested.
>
> The idea was that userspace would read sysfs, then use that to determine
> the correct exclusion parameters [1,2]. This logic was not expected to
> change; it correctly validates whether we can provide what the user
> requests.

OK, if you are ok with sysfs part, i can send next version with that
change only?.

>
> Thanks,
> Mark.
>
> [1] http://lists.infradead.org/pipermail/linux-arm-kernel/2017-April/499224.html
> [2] http://lists.infradead.org/pipermail/linux-arm-kernel/2017-April/499493.html
>
>>
>>       /*
>> @@ -1008,6 +1026,8 @@ static int armv8_pmuv3_init(struct arm_pmu *cpu_pmu)
>>               &armv8_pmuv3_events_attr_group;
>>       cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_FORMATS] =
>>               &armv8_pmuv3_format_attr_group;
>> +     cpu_pmu->attr_groups[ARMPMU_ATTR_GROUP_ATTR] =
>> +             &armv8_pmuv3_attr_group;
>>       return armv8pmu_probe_pmu(cpu_pmu);
>>  }
>>
>> diff --git a/include/linux/perf/arm_pmu.h b/include/linux/perf/arm_pmu.h
>> index 8462da2..a26ffc7 100644
>> --- a/include/linux/perf/arm_pmu.h
>> +++ b/include/linux/perf/arm_pmu.h
>> @@ -81,6 +81,7 @@ enum armpmu_attr_groups {
>>       ARMPMU_ATTR_GROUP_COMMON,
>>       ARMPMU_ATTR_GROUP_EVENTS,
>>       ARMPMU_ATTR_GROUP_FORMATS,
>> +     ARMPMU_ATTR_GROUP_ATTR,
>>       ARMPMU_NR_ATTR_GROUPS
>>  };
>>
>> --
>> 1.8.1.4
>>