[PATCH 6/6] perf cs-etm: Pass unformatted flag to decoder
James Clark
james.clark at arm.com
Wed Jul 21 01:35:53 PDT 2021
On 20/07/2021 16:45, Mathieu Poirier wrote:
> On Tue, Jul 13, 2021 at 04:40:08PM +0100, James Clark wrote:
>> The TRBE (Trace Buffer Extension) feature allows a separate trace buffer
>> for each trace source, therefore the trace wouldn't need to be
>> formatted. The driver was introduced in commit 3fbf7f011f24
>> ("coresight: sink: Add TRBE driver").
>>
>> The formatted/unformatted mode is encoded in one of the flags of the
>> AUX record. The first AUX record encountered for each event is used to
>> determine the mode, and this will persist for the remaining trace that
>> is either decoded or dumped.
>>
>> Signed-off-by: James Clark <james.clark at arm.com>
>> ---
>> tools/perf/util/cs-etm.c | 42 +++++++++++++++++++++++++++-------------
>> 1 file changed, 29 insertions(+), 13 deletions(-)
>>
>> diff --git a/tools/perf/util/cs-etm.c b/tools/perf/util/cs-etm.c
>> index 760050ea936d..62769a84a53f 100644
>> --- a/tools/perf/util/cs-etm.c
>> +++ b/tools/perf/util/cs-etm.c
>> @@ -461,13 +461,14 @@ static void cs_etm__set_trace_param_etmv4(struct cs_etm_trace_params *t_params,
>> }
>>
>> static int cs_etm__init_trace_params(struct cs_etm_trace_params *t_params,
>> - struct cs_etm_auxtrace *etm)
>> + struct cs_etm_auxtrace *etm,
>> + int decoders_per_cpu)
>> {
>> int i;
>> u32 etmidr;
>> u64 architecture;
>>
>> - for (i = 0; i < etm->num_cpu; i++) {
>> + for (i = 0; i < decoders_per_cpu; i++) {
>> architecture = etm->metadata[i][CS_ETM_MAGIC];
>>
>> switch (architecture) {
>> @@ -488,7 +489,8 @@ static int cs_etm__init_trace_params(struct cs_etm_trace_params *t_params,
>>
>> static int cs_etm__init_decoder_params(struct cs_etm_decoder_params *d_params,
>> struct cs_etm_queue *etmq,
>> - enum cs_etm_decoder_operation mode)
>> + enum cs_etm_decoder_operation mode,
>> + bool formatted)
>> {
>> int ret = -EINVAL;
>>
>> @@ -498,7 +500,7 @@ static int cs_etm__init_decoder_params(struct cs_etm_decoder_params *d_params,
>> d_params->packet_printer = cs_etm__packet_dump;
>> d_params->operation = mode;
>> d_params->data = etmq;
>> - d_params->formatted = true;
>> + d_params->formatted = formatted;
>> d_params->fsyncs = false;
>> d_params->hsyncs = false;
>> d_params->frame_aligned = true;
>> @@ -720,11 +722,13 @@ static u32 cs_etm__mem_access(struct cs_etm_queue *etmq, u8 trace_chan_id,
>> return len;
>> }
>>
>> -static struct cs_etm_queue *cs_etm__alloc_queue(struct cs_etm_auxtrace *etm)
>> +static struct cs_etm_queue *cs_etm__alloc_queue(struct cs_etm_auxtrace *etm,
>> + bool formatted)
>> {
>> struct cs_etm_decoder_params d_params;
>> struct cs_etm_trace_params *t_params = NULL;
>> struct cs_etm_queue *etmq;
>> + int decoders_per_cpu = formatted ? etm->num_cpu : 1;
>
> I really tripped on the name "decoders_per_cpu", to a point where I had to
> review the current code before looking at this patch. I find the "_per_cpu"
> part especially puzzling. In the end the variable determines the amount of
> decoders to instantiate for a specific queue...
>
> Couldn't it be just "decoders"? Or maybe it just needs a little comment to
> disambiguate things?
>
Yeah I think just decoders will be good if I add a comment explaining why there is
a difference with formatted vs unformatted. I will re-submit with the change.
Thanks for the reviews.
James
> Thanks,
> Mathieu
>
>>
>> etmq = zalloc(sizeof(*etmq));
>> if (!etmq)
>> @@ -735,21 +739,23 @@ static struct cs_etm_queue *cs_etm__alloc_queue(struct cs_etm_auxtrace *etm)
>> goto out_free;
>>
>> /* Use metadata to fill in trace parameters for trace decoder */
>> - t_params = zalloc(sizeof(*t_params) * etm->num_cpu);
>> + t_params = zalloc(sizeof(*t_params) * decoders_per_cpu);
>>
>> if (!t_params)
>> goto out_free;
>>
>> - if (cs_etm__init_trace_params(t_params, etm))
>> + if (cs_etm__init_trace_params(t_params, etm, decoders_per_cpu))
>> goto out_free;
>>
>> /* Set decoder parameters to decode trace packets */
>> if (cs_etm__init_decoder_params(&d_params, etmq,
>> dump_trace ? CS_ETM_OPERATION_PRINT :
>> - CS_ETM_OPERATION_DECODE))
>> + CS_ETM_OPERATION_DECODE,
>> + formatted))
>> goto out_free;
>>
>> - etmq->decoder = cs_etm_decoder__new(etm->num_cpu, &d_params, t_params);
>> + etmq->decoder = cs_etm_decoder__new(decoders_per_cpu, &d_params,
>> + t_params);
>>
>> if (!etmq->decoder)
>> goto out_free;
>> @@ -777,14 +783,15 @@ static struct cs_etm_queue *cs_etm__alloc_queue(struct cs_etm_auxtrace *etm)
>>
>> static int cs_etm__setup_queue(struct cs_etm_auxtrace *etm,
>> struct auxtrace_queue *queue,
>> - unsigned int queue_nr)
>> + unsigned int queue_nr,
>> + bool formatted)
>> {
>> struct cs_etm_queue *etmq = queue->priv;
>>
>> if (list_empty(&queue->head) || etmq)
>> return 0;
>>
>> - etmq = cs_etm__alloc_queue(etm);
>> + etmq = cs_etm__alloc_queue(etm, formatted);
>>
>> if (!etmq)
>> return -ENOMEM;
>> @@ -2430,8 +2437,14 @@ static int cs_etm__process_auxtrace_event(struct perf_session *session,
>> if (err)
>> return err;
>>
>> + /*
>> + * Knowing if the trace is formatted or not requires a lookup of
>> + * the aux record so only works in non-piped mode where data is
>> + * queued in cs_etm__queue_aux_records(). Always assume
>> + * formatted in piped mode (true).
>> + */
>> err = cs_etm__setup_queue(etm, &etm->queues.queue_array[idx],
>> - idx);
>> + idx, true);
>> if (err)
>> return err;
>>
>> @@ -2678,6 +2691,7 @@ static int cs_etm__queue_aux_fragment(struct perf_session *session, off_t file_o
>> union perf_event auxtrace_fragment;
>> __u64 aux_offset, aux_size;
>> __u32 idx;
>> + bool formatted;
>>
>> struct cs_etm_auxtrace *etm = container_of(session->auxtrace,
>> struct cs_etm_auxtrace,
>> @@ -2745,7 +2759,9 @@ static int cs_etm__queue_aux_fragment(struct perf_session *session, off_t file_o
>> return err;
>>
>> idx = auxtrace_event->idx;
>> - return cs_etm__setup_queue(etm, &etm->queues.queue_array[idx], idx);
>> + formatted = !(aux_event->flags & PERF_AUX_FLAG_CORESIGHT_FORMAT_RAW);
>> + return cs_etm__setup_queue(etm, &etm->queues.queue_array[idx],
>> + idx, formatted);
>> }
>>
>> /* Wasn't inside this buffer, but there were no parse errors. 1 == 'not found' */
>> --
>> 2.28.0
>>
More information about the linux-arm-kernel
mailing list