Re: [PATCH v4 4/6] perf arm-spe: Assign kernel time to synthesized event

From: James Clark <james.clark@arm.com>
To: Leo Yan <leo.yan@linaro.org>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>,
	Al Grant <Al.Grant@arm.com>, John Garry <john.garry@huawei.com>,
	Will Deacon <will@kernel.org>,
	Mathieu Poirier <mathieu.poirier@linaro.org>,
	Peter Zijlstra <peterz@infradead.org>,
	Ingo Molnar <mingo@redhat.com>,
	Mark Rutland <mark.rutland@arm.com>,
	Alexander Shishkin <alexander.shishkin@linux.intel.com>,
	Jiri Olsa <jolsa@redhat.com>, Namhyung Kim <namhyung@kernel.org>,
	Adrian Hunter <adrian.hunter@intel.com>,
	Dave Martin <Dave.Martin@arm.com>,
	linux-arm-kernel@lists.infradead.org,
	linux-kernel@vger.kernel.org
Subject: Re: [PATCH v4 4/6] perf arm-spe: Assign kernel time to synthesized event
Date: Fri, 16 Apr 2021 15:51:25 +0300	[thread overview]
Message-ID: <c9c060d1-0884-b67d-7689-6b45ad10baa9@arm.com> (raw)
In-Reply-To: <20210415152348.GF1011890@leoy-ThinkPad-X240s>

On 15/04/2021 18:23, Leo Yan wrote:
> On Thu, Apr 15, 2021 at 05:46:31PM +0300, James Clark wrote:
>>
>>
>> On 12/04/2021 12:10, Leo Yan wrote:
>>> In current code, it assigns the arch timer counter to the synthesized
>>> samples Arm SPE trace, thus the samples don't contain the kernel time
>>> but only contain the raw counter value.
>>>
>>> To fix the issue, this patch converts the timer counter to kernel time
>>> and assigns it to sample timestamp.
>>>
>>> Signed-off-by: Leo Yan <leo.yan@linaro.org>
>>> ---
>>>  tools/perf/util/arm-spe.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
>>> index 23714cf0380e..c13a89f06ab8 100644
>>> --- a/tools/perf/util/arm-spe.c
>>> +++ b/tools/perf/util/arm-spe.c
>>> @@ -234,7 +234,7 @@ static void arm_spe_prep_sample(struct arm_spe *spe,
>>>  	struct arm_spe_record *record = &speq->decoder->record;
>>>  
>>>  	if (!spe->timeless_decoding)
>>> -		sample->time = speq->timestamp;
>>> +		sample->time = tsc_to_perf_time(record->timestamp, &spe->tc);
>>
>>
>> I noticed that in arm_spe_recording_options() the TIME sample bit is set regardless of any options.
>> I don't know of a way to remove this, and if there isn't, does that mean that all the code in this
>> file that looks at spe->timeless_decoding is untested and has never been hit?
>>
>> Unless there is a way to get a perf file with only the AUXTRACE event and no others? I think that one
>> might have no timestamp set. Otherwise other events will always have timestamps so spe->timeless_decoding
>> is always false.
> 
> Good point.  To be honest, I never noticed this issue until you
> mentioned this.
> 
> We should fix for the "timeless" flow; and it's questionable for the
> function arm_spe_recording_options(), except for setting
> PERF_SAMPLE_TIME, it also hard codes for setting
> PERF_SAMPLE_CPU and PERF_SAMPLE_TID.  Might need to carefully go
> through this function.
> 

Yeah, it's not strictly related to your change, which is definitely an improvement.
But maybe we should have a look at the SPE implementation relating to timestamps as a whole.

> Thanks,
> Leo
> 

_______________________________________________
linux-arm-kernel mailing list
linux-arm-kernel@lists.infradead.org
http://lists.infradead.org/mailman/listinfo/linux-arm-kernel