[PATCH 0/1] perf arm-spe: Inject SPE samples in perf-inject

perf-inject is currently not working for Arm SPE. When you try to run
perf-inject and perf-report with a perf.data file that contains SPE
traces, the tool reports a "Bad address" error:

  # ./perf record -e arm_spe_0/ts_enable=1,store_filter=1,branch_filter=1,load_filter=1/ -a -- sleep 1
  # ./perf inject -i perf.data -o perf.inject.data --itrace
  # ./perf report -i perf.inject.data --stdio
  
  0x42c00 [0x8]: failed to process type: 9 [Bad address]
  Error:
  failed to process sample

As far as I know, the issue was first spotted in [1], but perf-inject
was not yet injecting the samples. This patch does something similar to
what cs_etm does for injecting the samples [2], but for SPE.

[1] https://patchwork.kernel.org/project/linux-arm-kernel/cover/20210412091006.468557-1-leo.yan@linaro.org/#24117339
[2] https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/util/cs-etm.c?h=perf/core&id=133fe2e617e48ca0948983329f43877064ffda3e#n1196

German Gomez (1):
  perf arm-spe: Inject SPE samples in perf-inject

 tools/perf/util/arm-spe.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

-- 
2.25.1

Inject synthesized SPE samples during perf-inject run.

Signed-off-by: German Gomez <redacted>
---
 tools/perf/util/arm-spe.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 58b7069c5..7054f2315 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c

@@ -51,6 +51,7 @@ struct arm_spe {
 	u8				timeless_decoding;
 	u8				data_queued;
 
+	u64				sample_type;
 	u8				sample_flc;
 	u8				sample_llc;
 	u8				sample_tlb;

@@ -248,6 +249,12 @@ static void arm_spe_prep_sample(struct arm_spe *spe,
 	event->sample.header.size = sizeof(struct perf_event_header);
 }
 
+static int arm_spe__inject_event(union perf_event *event, struct perf_sample *sample, u64 type)
+{
+	event->header.size = perf_event__sample_event_size(sample, type, 0);
+	return perf_event__synthesize_sample(event, type, 0, sample);
+}
+
 static inline int
 arm_spe_deliver_synth_event(struct arm_spe *spe,
 			    struct arm_spe_queue *speq __maybe_unused,

@@ -256,6 +263,12 @@ arm_spe_deliver_synth_event(struct arm_spe *spe,
 {
 	int ret;
 
+	if (spe->synth_opts.inject) {
+		ret = arm_spe__inject_event(event, sample, spe->sample_type);
+		if (ret)
+			return ret;
+	}
+
 	ret = perf_session__deliver_synth_event(spe->session, event, sample);
 	if (ret)
 		pr_err("ARM SPE: failed to deliver event, error %d\n", ret);

@@ -920,6 +933,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
 	else
 		attr.sample_type |= PERF_SAMPLE_TIME;
 
+	spe->sample_type = attr.sample_type;
+
 	attr.exclude_user = evsel->core.attr.exclude_user;
 	attr.exclude_kernel = evsel->core.attr.exclude_kernel;
 	attr.exclude_hv = evsel->core.attr.exclude_hv;

-- 
2.25.1

On 05/11/2021 10:41, German Gomez wrote:

Inject synthesized SPE samples during perf-inject run.

Signed-off-by: German Gomez <redacted>

Reviewed-by: James Clark <redacted>

Em Mon, Nov 08, 2021 at 08:39:58PM +0000, James Clark escreveu:

On 05/11/2021 10:41, German Gomez wrote:

quoted

Inject synthesized SPE samples during perf-inject run.

Signed-off-by: German Gomez <redacted>

Reviewed-by: James Clark <redacted>

Thanks, applied.

- Arnaldo

quoted

---
 tools/perf/util/arm-spe.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 58b7069c5..7054f2315 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c

@@ -51,6 +51,7 @@ struct arm_spe {
 	u8				timeless_decoding;
 	u8				data_queued;
 
+	u64				sample_type;
 	u8				sample_flc;
 	u8				sample_llc;
 	u8				sample_tlb;

@@ -248,6 +249,12 @@ static void arm_spe_prep_sample(struct arm_spe *spe,
 	event->sample.header.size = sizeof(struct perf_event_header);
 }
 
+static int arm_spe__inject_event(union perf_event *event, struct perf_sample *sample, u64 type)
+{
+	event->header.size = perf_event__sample_event_size(sample, type, 0);
+	return perf_event__synthesize_sample(event, type, 0, sample);
+}
+
 static inline int
 arm_spe_deliver_synth_event(struct arm_spe *spe,
 			    struct arm_spe_queue *speq __maybe_unused,

@@ -256,6 +263,12 @@ arm_spe_deliver_synth_event(struct arm_spe *spe,
 {
 	int ret;
 
+	if (spe->synth_opts.inject) {
+		ret = arm_spe__inject_event(event, sample, spe->sample_type);
+		if (ret)
+			return ret;
+	}
+
 	ret = perf_session__deliver_synth_event(spe->session, event, sample);
 	if (ret)
 		pr_err("ARM SPE: failed to deliver event, error %d\n", ret);

@@ -920,6 +933,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
 	else
 		attr.sample_type |= PERF_SAMPLE_TIME;
 
+	spe->sample_type = attr.sample_type;
+
 	attr.exclude_user = evsel->core.attr.exclude_user;
 	attr.exclude_kernel = evsel->core.attr.exclude_kernel;
 	attr.exclude_hv = evsel->core.attr.exclude_hv;

-- 

- Arnaldo

Em Mon, Nov 08, 2021 at 08:39:58PM +0000, James Clark escreveu:

On 05/11/2021 10:41, German Gomez wrote:

quoted

Inject synthesized SPE samples during perf-inject run.

Signed-off-by: German Gomez <redacted>

Reviewed-by: James Clark <redacted>

Next time please expand this explanations a bit more: why should we
inject such samples? Is this enabling some new mode of operation, fixing
something, what is an example of output before this patch and after it?

- Arnaldo

quoted

---
 tools/perf/util/arm-spe.c | 15 +++++++++++++++
 1 file changed, 15 insertions(+)

diff --git a/tools/perf/util/arm-spe.c b/tools/perf/util/arm-spe.c
index 58b7069c5..7054f2315 100644
--- a/tools/perf/util/arm-spe.c
+++ b/tools/perf/util/arm-spe.c

@@ -51,6 +51,7 @@ struct arm_spe {
 	u8				timeless_decoding;
 	u8				data_queued;
 
+	u64				sample_type;
 	u8				sample_flc;
 	u8				sample_llc;
 	u8				sample_tlb;

@@ -248,6 +249,12 @@ static void arm_spe_prep_sample(struct arm_spe *spe,
 	event->sample.header.size = sizeof(struct perf_event_header);
 }
 
+static int arm_spe__inject_event(union perf_event *event, struct perf_sample *sample, u64 type)
+{
+	event->header.size = perf_event__sample_event_size(sample, type, 0);
+	return perf_event__synthesize_sample(event, type, 0, sample);
+}
+
 static inline int
 arm_spe_deliver_synth_event(struct arm_spe *spe,
 			    struct arm_spe_queue *speq __maybe_unused,

@@ -256,6 +263,12 @@ arm_spe_deliver_synth_event(struct arm_spe *spe,
 {
 	int ret;
 
+	if (spe->synth_opts.inject) {
+		ret = arm_spe__inject_event(event, sample, spe->sample_type);
+		if (ret)
+			return ret;
+	}
+
 	ret = perf_session__deliver_synth_event(spe->session, event, sample);
 	if (ret)
 		pr_err("ARM SPE: failed to deliver event, error %d\n", ret);

@@ -920,6 +933,8 @@ arm_spe_synth_events(struct arm_spe *spe, struct perf_session *session)
 	else
 		attr.sample_type |= PERF_SAMPLE_TIME;
 
+	spe->sample_type = attr.sample_type;
+
 	attr.exclude_user = evsel->core.attr.exclude_user;
 	attr.exclude_kernel = evsel->core.attr.exclude_kernel;
 	attr.exclude_hv = evsel->core.attr.exclude_hv;

-- 

- Arnaldo

Hi Arnaldo,

On 17/11/2021 15:29, Arnaldo Carvalho de Melo wrote:

[...]
Next time please expand this explanations a bit more: why should we
inject such samples? Is this enabling some new mode of operation, fixing
something, what is an example of output before this patch and after it?

I will keep this in mind, thanks. In this case the support was missing
and I included the context in the cover letter but not the commit msg.

Thanks,
German

- Arnaldo

[...]

Em Wed, Nov 17, 2021 at 04:27:27PM +0000, German Gomez escreveu:

Hi Arnaldo,

On 17/11/2021 15:29, Arnaldo Carvalho de Melo wrote:

quoted

[...]
Next time please expand this explanations a bit more: why should we
inject such samples? Is this enabling some new mode of operation, fixing
something, what is an example of output before this patch and after it?

I will keep this in mind, thanks. In this case the support was missing
and I included the context in the cover letter but not the commit msg.

Got it, I'm adding it to the cset comment:

-------------------------------------------------------------------------------

perf-inject is currently not working for Arm SPE. When you try to run
perf-inject and perf-report with a perf.data file that contains SPE
traces, the tool reports a "Bad address" error:

  # ./perf record -e arm_spe_0/ts_enable=1,store_filter=1,branch_filter=1,load_filter=1/ -a -- sleep 1
  # ./perf inject -i perf.data -o perf.inject.data --itrace
  # ./perf report -i perf.inject.data --stdio

  0x42c00 [0x8]: failed to process type: 9 [Bad address]
  Error:
  failed to process sample

As far as I know, the issue was first spotted in [1], but perf-inject
was not yet injecting the samples. This patch does something similar to
what cs_etm does for injecting the samples [2], but for SPE.

[1] https://patchwork.kernel.org/project/linux-arm-kernel/cover/20210412091006.468557-1-leo.yan@linaro.org/#24117339
[2] https://git.kernel.org/pub/scm/linux/kernel/git/acme/linux.git/tree/tools/perf/util/cs-etm.c?h=perf/core&id=133fe2e617e48ca0948983329f43877064ffda3e#n1196
-------------------------------------------------------------------------------