[PATCH 0/2] perf tests: Fix Track with sched_switch test for hybrid case

[PATCH 0/2] perf tests: Fix Track with sched_switch test for hybrid case

[Adrian Hunter]

Hi

Here are a couple of fixes related to the Track with sched_switch test.


Adrian Hunter (2):
      perf parse-events: Fix segfault when event parser gets an error
      perf tests: Fix Track with sched_switch test for hybrid case

 tools/perf/tests/switch-tracking.c | 18 +++++++++++++-----
 tools/perf/util/parse-events.c     | 14 +++++++++++---
 2 files changed, 24 insertions(+), 8 deletions(-)


Regards
Adrian

[PATCH 1/2] perf parse-events: Fix segfault when event parser gets an error

[Adrian Hunter]

parse_events() is often called with parse_events_error set to NULL.
Make parse_events_error__handle() not segfault in that case.

Fixes: 43eb05d06679 ("perf tests: Support 'Track with sched_switch' test for hybrid")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/util/parse-events.c | 14 +++++++++++---
 1 file changed, 11 insertions(+), 3 deletions(-)

diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
index 7ed235740431..700c95eafd62 100644
--- a/tools/perf/util/parse-events.c
+++ b/tools/perf/util/parse-events.c
@@ -2391,9 +2391,12 @@ void parse_events_error__exit(struct parse_events_error *err)
 void parse_events_error__handle(struct parse_events_error *err, int idx,
 				char *str, char *help)
 {
-	if (WARN(!str, "WARNING: failed to provide error string\n")) {
-		free(help);
-		return;
+	if (WARN(!str, "WARNING: failed to provide error string\n"))
+		goto out_free;
+	if (!err) {
+		/* Assume caller does not want message printed */
+		pr_debug("event syntax error: %s\n", str);
+		goto out_free;
 	}
 	switch (err->num_errors) {
 	case 0:
@@ -2419,6 +2422,11 @@ void parse_events_error__handle(struct parse_events_error *err, int idx,
 		break;
 	}
 	err->num_errors++;
+	return;
+
+out_free:
+	free(str);
+	free(help);
 }
 
 #define MAX_WIDTH 1000
-- 
2.25.1

[PATCH 2/2] perf tests: Fix Track with sched_switch test for hybrid case

[Adrian Hunter]

If cpu_core PMU event fails to parse, try also cpu_atom PMU event when
parsing cycles event.

Fixes: 43eb05d06679 ("perf tests: Support 'Track with sched_switch' test for hybrid")
Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
---
 tools/perf/tests/switch-tracking.c | 18 +++++++++++++-----
 1 file changed, 13 insertions(+), 5 deletions(-)

diff --git a/tools/perf/tests/switch-tracking.c b/tools/perf/tests/switch-tracking.c
index 0c0c2328bf4e..6f53bee33f7c 100644
--- a/tools/perf/tests/switch-tracking.c
+++ b/tools/perf/tests/switch-tracking.c
@@ -324,6 +324,7 @@ static int process_events(struct evlist *evlist,
 static int test__switch_tracking(struct test_suite *test __maybe_unused, int subtest __maybe_unused)
 {
 	const char *sched_switch = "sched:sched_switch";
+	const char *cycles = "cycles:u";
 	struct switch_tracking switch_tracking = { .tids = NULL, };
 	struct record_opts opts = {
 		.mmap_pages	     = UINT_MAX,
@@ -372,12 +373,19 @@ static int test__switch_tracking(struct test_suite *test __maybe_unused, int sub
 	cpu_clocks_evsel = evlist__last(evlist);
 
 	/* Second event */
-	if (perf_pmu__has_hybrid())
-		err = parse_events(evlist, "cpu_core/cycles/u", NULL);
-	else
-		err = parse_events(evlist, "cycles:u", NULL);
+	if (perf_pmu__has_hybrid()) {
+		cycles = "cpu_core/cycles/u";
+		err = parse_events(evlist, cycles, NULL);
+		if (err) {
+			cycles = "cpu_atom/cycles/u";
+			pr_debug("Trying %s\n", cycles);
+			err = parse_events(evlist, cycles, NULL);
+		}
+	} else {
+		err = parse_events(evlist, cycles, NULL);
+	}
 	if (err) {
-		pr_debug("Failed to parse event cycles:u\n");
+		pr_debug("Failed to parse event %s\n", cycles);
 		goto out_err;
 	}
 
-- 
2.25.1

Re: [PATCH 1/2] perf parse-events: Fix segfault when event parser gets an error

[Ian Rogers]

On Tue, Jul 12, 2022 at 4:28 AM Adrian Hunter <adrian.hunter@intel.com> wrote:
>
> parse_events() is often called with parse_events_error set to NULL.
> Make parse_events_error__handle() not segfault in that case.
>
> Fixes: 43eb05d06679 ("perf tests: Support 'Track with sched_switch' test for hybrid")
> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
> ---
>  tools/perf/util/parse-events.c | 14 +++++++++++---
>  1 file changed, 11 insertions(+), 3 deletions(-)
>
> diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
> index 7ed235740431..700c95eafd62 100644
> --- a/tools/perf/util/parse-events.c
> +++ b/tools/perf/util/parse-events.c
> @@ -2391,9 +2391,12 @@ void parse_events_error__exit(struct parse_events_error *err)
>  void parse_events_error__handle(struct parse_events_error *err, int idx,
>                                 char *str, char *help)
>  {
> -       if (WARN(!str, "WARNING: failed to provide error string\n")) {
> -               free(help);
> -               return;
> +       if (WARN(!str, "WARNING: failed to provide error string\n"))
> +               goto out_free;
> +       if (!err) {
> +               /* Assume caller does not want message printed */
> +               pr_debug("event syntax error: %s\n", str);
> +               goto out_free;


It feels like a simpler invariant (as done elsewhere) to have the
caller always pass err and then in the caller to call
parse_events_error__exit. Any reason for behavior change?

Thanks,
Ian

>         }
>         switch (err->num_errors) {
>         case 0:
> @@ -2419,6 +2422,11 @@ void parse_events_error__handle(struct parse_events_error *err, int idx,
>                 break;
>         }
>         err->num_errors++;
> +       return;
> +
> +out_free:
> +       free(str);
> +       free(help);
>  }
>
>  #define MAX_WIDTH 1000
> --
> 2.25.1
>

Re: [PATCH 1/2] perf parse-events: Fix segfault when event parser gets an error

[Adrian Hunter]

On 19/07/22 04:10, Ian Rogers wrote:
> On Tue, Jul 12, 2022 at 4:28 AM Adrian Hunter <adrian.hunter@intel.com> wrote:
>>
>> parse_events() is often called with parse_events_error set to NULL.
>> Make parse_events_error__handle() not segfault in that case.
>>
>> Fixes: 43eb05d06679 ("perf tests: Support 'Track with sched_switch' test for hybrid")
>> Signed-off-by: Adrian Hunter <adrian.hunter@intel.com>
>> ---
>>  tools/perf/util/parse-events.c | 14 +++++++++++---
>>  1 file changed, 11 insertions(+), 3 deletions(-)
>>
>> diff --git a/tools/perf/util/parse-events.c b/tools/perf/util/parse-events.c
>> index 7ed235740431..700c95eafd62 100644
>> --- a/tools/perf/util/parse-events.c
>> +++ b/tools/perf/util/parse-events.c
>> @@ -2391,9 +2391,12 @@ void parse_events_error__exit(struct parse_events_error *err)
>>  void parse_events_error__handle(struct parse_events_error *err, int idx,
>>                                 char *str, char *help)
>>  {
>> -       if (WARN(!str, "WARNING: failed to provide error string\n")) {
>> -               free(help);
>> -               return;
>> +       if (WARN(!str, "WARNING: failed to provide error string\n"))
>> +               goto out_free;
>> +       if (!err) {
>> +               /* Assume caller does not want message printed */
>> +               pr_debug("event syntax error: %s\n", str);
>> +               goto out_free;
> 
> 
> It feels like a simpler invariant (as done elsewhere) to have the
> caller always pass err and then in the caller to call
> parse_events_error__exit. Any reason for behavior change?

This is a bug fix. More complicated changes can be done separately,
if desired, but it is wrong to segfault here in any case.

> 
> Thanks,
> Ian
> 
>>         }
>>         switch (err->num_errors) {
>>         case 0:
>> @@ -2419,6 +2422,11 @@ void parse_events_error__handle(struct parse_events_error *err, int idx,
>>                 break;
>>         }
>>         err->num_errors++;
>> +       return;
>> +
>> +out_free:
>> +       free(str);
>> +       free(help);
>>  }
>>
>>  #define MAX_WIDTH 1000
>> --
>> 2.25.1
>>