[RESEND] perf cs-etm: add error logging to cs_etm_is_etmv4()

Commit Message

Junhao He March 16, 2023, 1:32 p.m. UTC

On the HiSilicon platform, there is one ETM per logic core.
When setting up SMT, not every process has an ETM. So the path
".../cs_etm/cpux/trcidr/trcidr0" does not exist, the function
perf_pmu__scan_file() return an error. However, the command
'perf record' will returns silently and don't print.

After this patch, log a error when read fails that makes it easy
for users to debug.

  root@localhost:/# perf record -e /cs_etm/@ultra_smb0/ -C 3 uname -a
  cs_etm: can't read file cpu3/trcidr/trcidr0
  root@localhost:/#

Signed-off-by: Junhao He <[e-mail]>
---
 tools/perf/arch/arm/util/cs-etm.c | 6 +++---
 1 file changed, 3 insertions(+), 3 deletions(-)

Patch

diff --git a/tools/perf/arch/arm/util/cs-etm.c b/tools/perf/arch/arm/util/cs-etm.c
index 7f71c8a237ff..19ea17430399 100644
--- a/tools/perf/arch/arm/util/cs-etm.c
+++ b/tools/perf/arch/arm/util/cs-etm.c
@@ -585,7 +585,6 @@  cs_etm_info_priv_size(struct auxtrace_record *itr __maybe_unused,
 
 static bool cs_etm_is_etmv4(struct auxtrace_record *itr, int cpu)
 {
-	bool ret = false;
 	char path[PATH_MAX];
 	int scan;
 	unsigned int val;
@@ -600,9 +599,10 @@  static bool cs_etm_is_etmv4(struct auxtrace_record *itr, int cpu)
 
 	/* The file was read successfully, we have a winner */
 	if (scan == 1)
-		ret = true;
+		return true;
 
-	return ret;
+	pr_err("%s: can't read file %s\n", CORESIGHT_ETM_PMU_NAME, path);
+	return false;
 }
 
 static int cs_etm_get_ro(struct perf_pmu *pmu, int cpu, const char *path)