Re: [PATCH v7 13/15] coresight: events: PERF_RECORD_AUX_OUTPUT_HW_ID used for Trace ID

Hi Suzuki

On Thu, 19 Jan 2023 at 11:02, Suzuki K Poulose suzuki.poulose@arm.com wrote:

Hi Mike

On 16/01/2023 12:49, Mike Leach wrote:

...

Use the perf_report_aux_output_id() call to output the CoreSight trace ID and associated CPU as a PERF_RECORD_AUX_OUTPUT_HW_ID record in the perf.data file.

Signed-off-by: Mike Leach mike.leach@linaro.org Reviewed-by: Suzuki K Poulose suzuki.poulose@arm.com

---

drivers/hwtracing/coresight/coresight-etm-perf.c | 7 +++++++ include/linux/coresight-pmu.h | 14 ++++++++++++++ 2 files changed, 21 insertions(+)

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c index bdb9ab86173a..12fff661456e 100644 --- a/drivers/hwtracing/coresight/coresight-etm-perf.c +++ b/drivers/hwtracing/coresight/coresight-etm-perf.c @@ -4,6 +4,7 @@

• Author: Mathieu Poirier mathieu.poirier@linaro.org

*/

+#include <linux/bitfield.h> #include <linux/coresight.h> #include <linux/coresight-pmu.h> #include <linux/cpumask.h> @@ -448,6 +449,7 @@ static void etm_event_start(struct perf_event *event, int flags) struct perf_output_handle *handle = &ctxt->handle; struct coresight_device *sink, *csdev = per_cpu(csdev_src, cpu); struct list_head *path;

• u64 hw_id;
  
  if (!csdev)
          goto fail;
  

@@ -493,6 +495,11 @@ static void etm_event_start(struct perf_event *event, int flags) if (source_ops(csdev)->enable(csdev, event, CS_MODE_PERF)) goto fail_disable_path;

• /* output cpu / trace ID in perf record */
  

• hw_id = FIELD_PREP(CS_AUX_HW_ID_VERSION_MASK, CS_AUX_HW_ID_CURR_VERSION);
  

• hw_id |= FIELD_PREP(CS_AUX_HW_ID_TRACE_ID_MASK, coresight_trace_id_read_cpu_id(cpu));
  

• perf_report_aux_output_id(event, hw_id);
  

One minor nit:

This ends up putting the HWID packet for every single time the event is scheduled on the CPU and thus we have :

$ perf report -D | grep OUTPUT_HW_ID ... AUX_OUTPUT_HW_ID events: 55 (18.3%)

We could do the above step only once per CPU for the lifetime of the event by book keeping the info in event_data. With that we get, one record per CPU and:

$ perf report -D | grep OUTPUT_HW_ID ... AUX_OUTPUT_HW_ID events: 5 ( 1.9%)

Here is a quick patch. We can queue this separately if you all are Ok with this. Thoughts ?

Seems reasonable given we guarantee the ID <=> CPU association for the lifetime of the perf session.

Regards

Mike

---8>---

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.c b/drivers/hwtracing/coresight/coresight-etm-perf.c index 12fff661456e..5335f3a2b010 100644 --- a/drivers/hwtracing/coresight/coresight-etm-perf.c +++ b/drivers/hwtracing/coresight/coresight-etm-perf.c @@ -495,10 +495,16 @@ static void etm_event_start(struct perf_event *event, int flags) if (source_ops(csdev)->enable(csdev, event, CS_MODE_PERF)) goto fail_disable_path;

•   /* output cpu / trace ID in perf record */
  

•   hw_id = FIELD_PREP(CS_AUX_HW_ID_VERSION_MASK,
  

CS_AUX_HW_ID_CURR_VERSION);

•   hw_id |= FIELD_PREP(CS_AUX_HW_ID_TRACE_ID_MASK,
  

coresight_trace_id_read_cpu_id(cpu));

•   perf_report_aux_output_id(event, hw_id);
  

•   /*
  

•    * output cpu / trace ID in perf record, once for the lifetime
  

•    * of the event.
  

•    */
  

•   if (!cpumask_test_cpu(cpu, &event_data->aux_hwid_done)) {
  

•           cpumask_set_cpu(cpu, &event_data->aux_hwid_done);
  

•           hw_id = FIELD_PREP(CS_AUX_HW_ID_VERSION_MASK,
  

CS_AUX_HW_ID_CURR_VERSION);

•           hw_id |= FIELD_PREP(CS_AUX_HW_ID_TRACE_ID_MASK,
  

coresight_trace_id_read_cpu_id(cpu));

•           perf_report_aux_output_id(event, hw_id);
  

•   }
  
  

  out: /* Tell the perf core the event is alive */

diff --git a/drivers/hwtracing/coresight/coresight-etm-perf.h b/drivers/hwtracing/coresight/coresight-etm-perf.h index 468f7799ab4f..9b11bb3d96d9 100644 --- a/drivers/hwtracing/coresight/coresight-etm-perf.h +++ b/drivers/hwtracing/coresight/coresight-etm-perf.h @@ -55,6 +55,7 @@ struct etm_filters { struct etm_event_data { struct work_struct work; cpumask_t mask;

•   cpumask_t aux_hwid_done;
     void *snk_config;
     u32 cfg_hash;
     struct list_head * __percpu *path;
  
  
  
  

...

out: /* Tell the perf core the event is alive */ event->hw.state = 0; diff --git a/include/linux/coresight-pmu.h b/include/linux/coresight-pmu.h index 624f4843453e..51ac441a37c3 100644 --- a/include/linux/coresight-pmu.h +++ b/include/linux/coresight-pmu.h @@ -7,6 +7,8 @@ #ifndef _LINUX_CORESIGHT_PMU_H #define _LINUX_CORESIGHT_PMU_H

+#include <linux/bits.h>

• #define CORESIGHT_ETM_PMU_NAME "cs_etm"

  /*

@@ -43,4 +45,16 @@ #define ETM4_CFG_BIT_RETSTK 12 #define ETM4_CFG_BIT_VMID_OPT 15

+/*

• • Interpretation of the PERF_RECORD_AUX_OUTPUT_HW_ID payload.
• • Used to associate a CPU with the CoreSight Trace ID.
• • [07:00] - Trace ID - uses 8 bits to make value easy to read in file.
• • [59:08] - Unused (SBZ)
• • [63:60] - Version
• */

+#define CS_AUX_HW_ID_TRACE_ID_MASK GENMASK_ULL(7, 0) +#define CS_AUX_HW_ID_VERSION_MASK GENMASK_ULL(63, 60)

+#define CS_AUX_HW_ID_CURR_VERSION 0

• #endif