On 24/06/20 3:42 pm, Jiri Olsa wrote:
> On Tue, Jun 23, 2020 at 07:48:50PM -0400, Gaurav Singh wrote:
>> intel_pt_evsel cannot be NULL here since its already being
>> dereferenced above. Remove this redundant check.
>>
>> Signed-off-by: Gaurav Singh <[email protected]>
>> ---
>> tools/perf/arch/x86/util/intel-pt.c | 24 +++++++++++-------------
>> 1 file changed, 11 insertions(+), 13 deletions(-)
>>
>> diff --git a/tools/perf/arch/x86/util/intel-pt.c
>> b/tools/perf/arch/x86/util/intel-pt.c
>> index 839ef52c1ac2..8cc87fd2dc6f 100644
>> --- a/tools/perf/arch/x86/util/intel-pt.c
>> +++ b/tools/perf/arch/x86/util/intel-pt.c
>> @@ -836,19 +836,17 @@ static int intel_pt_recording_options(struct
>> auxtrace_record *itr,
>> }
>> }
>>
>
> I wonder we sould put check before that above usage of intel_pt_evsel,
> because there's no check for that.. but yes, if intel_pt_evsel is NULL
> it will fail earlier.. Adrian?
The code was written when sampling-mode did not require an intel_pt_evsel.
Now it does, so intel_pt_evsel will not be NULL. I think it is OK as is.
>
> thanks,
> jirka
>
>> - if (intel_pt_evsel) {
>> - /*
>> - * To obtain the auxtrace buffer file descriptor, the auxtrace
>> - * event must come first.
>> - */
>> - perf_evlist__to_front(evlist, intel_pt_evsel);
>> - /*
>> - * In the case of per-cpu mmaps, we need the CPU on the
>> - * AUX event.
>> - */
>> - if (!perf_cpu_map__empty(cpus))
>> - evsel__set_sample_bit(intel_pt_evsel, CPU);
>> - }
>> + /*
>> + * To obtain the auxtrace buffer file descriptor, the auxtrace
>> + * event must come first.
>> + */
>> + perf_evlist__to_front(evlist, intel_pt_evsel);
>> + /*
>> + * In the case of per-cpu mmaps, we need the CPU on the
>> + * AUX event.
>> + */
>> + if (!perf_cpu_map__empty(cpus))
>> + evsel__set_sample_bit(intel_pt_evsel, CPU);
>>
>> /* Add dummy event to keep tracking */
>> if (opts->full_auxtrace) {
>> --
>> 2.17.1
>>
>
