Skip to contentSkip to conversation

perf tools: intel-pt: fix endless record after being terminated

PATCH In review Wei Li 3 participants 4 messages Jan 2, 2020 – Jan 14, 2020 perf events

lore
Share & tools
lore

Wei Li

author
#0 lore raw

PATCH 1/1 perf tools: intel-pt: fix endless record after being terminated

In __cmd_record(), when receiving SIGINT(ctrl + c), a done flag will be set and the event list will be disabled by evlist__disable() once.

While in auxtrace_record.read_finish(), the related events will be enabled again, if they are continuous, the recording seems to be endless.

If the intel_pt event is disabled, we don't enable it again here.

Before the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 46803
^C^C^C^C^C^C

After the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 48591
^C[ perf record: Woken up 0 times to write data ]
Warning:
AUX data lost 504 times out of 4816!

[ perf record: Captured and wrote 2024.405 MB perf.data ]

---
 tools/perf/arch/x86/util/intel-pt.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
tools/perf/arch/x86/util/intel-pt.c +7 -3
index 20df442fdf36..1e96afcd8646 100644
@@ -1173,9 +1173,13 @@ static int intel_pt_read_finish(struct auxtrace_record *itr, int idx)
 	struct evsel *evsel;
 
 	evlist__for_each_entry(ptr->evlist, evsel) {
-		if (evsel->core.attr.type == ptr->intel_pt_pmu->type)
-			return perf_evlist__enable_event_idx(ptr->evlist, evsel,
-							     idx);
+		if (evsel->core.attr.type == ptr->intel_pt_pmu->type) {
+			if (evsel->disabled)
+				return 0;
+			else
+				return perf_evlist__enable_event_idx(
+						ptr->evlist, evsel, idx);
+		}
 	}
 	return -EINVAL;
 }
Signature · 1 line
2.17.1
#1 lore raw
Quoted 34 lines · Wei Li

In __cmd_record(), when receiving SIGINT(ctrl + c), a done flag will be set and the event list will be disabled by evlist__disable() once.

While in auxtrace_record.read_finish(), the related events will be enabled again, if they are continuous, the recording seems to be endless.

If the intel_pt event is disabled, we don't enable it again here.

Before the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 46803
^C^C^C^C^C^C

After the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 48591
^C[ perf record: Woken up 0 times to write data ]
Warning:
AUX data lost 504 times out of 4816!

[ perf record: Captured and wrote 2024.405 MB perf.data ]

---
 tools/perf/arch/x86/util/intel-pt.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
tools/perf/arch/x86/util/intel-pt.c +0 -0
index 20df442fdf36..1e96afcd8646 100644
@@ -1173,9 +1173,13 @@ static int intel_pt_read_finish(struct auxtrace_record *itr, int idx)
 	struct evsel *evsel;
 

tools/perf/arch/x86/util/intel-pt.c static int intel_pt_read_finish(struct auxtrace_record *itr, int idx) · line 1181

Quoted 10 lines
tools/perf/arch/x86/util/intel-pt.c +6 -3
 	evlist__for_each_entry(ptr->evlist, evsel) {
-		if (evsel->core.attr.type == ptr->intel_pt_pmu->type)
-			return perf_evlist__enable_event_idx(ptr->evlist, evsel,
-							     idx);
+		if (evsel->core.attr.type == ptr->intel_pt_pmu->type) {
+			if (evsel->disabled)
+				return 0;
+			else
+				return perf_evlist__enable_event_idx(
+						ptr->evlist, evsel, idx);

what's the logic behind enabling the event in here?

thanks,
jirka

reviewer
#2 lore raw
Quoted 35 lines · Jiri Olsa
Quoted 34 lines · Wei Li

In __cmd_record(), when receiving SIGINT(ctrl + c), a done flag will be set and the event list will be disabled by evlist__disable() once.

While in auxtrace_record.read_finish(), the related events will be enabled again, if they are continuous, the recording seems to be endless.

If the intel_pt event is disabled, we don't enable it again here.

Before the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 46803
^C^C^C^C^C^C

After the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 48591
^C[ perf record: Woken up 0 times to write data ]
Warning:
AUX data lost 504 times out of 4816!

[ perf record: Captured and wrote 2024.405 MB perf.data ]

---
 tools/perf/arch/x86/util/intel-pt.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
tools/perf/arch/x86/util/intel-pt.c +0 -0
index 20df442fdf36..1e96afcd8646 100644
@@ -1173,9 +1173,13 @@ static int intel_pt_read_finish(struct auxtrace_record *itr, int idx)
 	struct evsel *evsel;
 
Quoted 12 lines

tools/perf/arch/x86/util/intel-pt.c static int intel_pt_read_finish(struct auxtrace_record *itr, int idx) · line 1181

Quoted 10 lines
tools/perf/arch/x86/util/intel-pt.c +6 -3
 	evlist__for_each_entry(ptr->evlist, evsel) {
-		if (evsel->core.attr.type == ptr->intel_pt_pmu->type)
-			return perf_evlist__enable_event_idx(ptr->evlist, evsel,
-							     idx);
+		if (evsel->core.attr.type == ptr->intel_pt_pmu->type) {
+			if (evsel->disabled)
+				return 0;
+			else
+				return perf_evlist__enable_event_idx(
+						ptr->evlist, evsel, idx);

what's the logic behind enabling the event in here?

Tracing stops when the auxtrace buffer is full and won't start again until the event is scheduled in (which is never for system-wide events) but enabling here will start tracing again immediately if possible.

reviewer
#3 lore raw
In reply to Wei Li
Quoted 20 lines · Wei Li

In __cmd_record(), when receiving SIGINT(ctrl + c), a done flag will be set and the event list will be disabled by evlist__disable() once.

While in auxtrace_record.read_finish(), the related events will be enabled again, if they are continuous, the recording seems to be endless.

If the intel_pt event is disabled, we don't enable it again here.

Before the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 46803
^C^C^C^C^C^C

After the patch:

huawei@huawei-2288H-V5:~/linux-5.5-rc4/tools/perf$ ./perf record -e \
intel_pt//u -p 48591
^C[ perf record: Woken up 0 times to write data ]
Warning:
AUX data lost 504 times out of 4816!
Quoted 3 lines

[ perf record: Captured and wrote 2024.405 MB perf.data ]

All ->read_finish() implementations seem the same, so can we make a common helper auxtrace_read_finish() and set all .read_finish = to that.

Quoted 26 lines
---
 tools/perf/arch/x86/util/intel-pt.c | 10 +++++++---
 1 file changed, 7 insertions(+), 3 deletions(-)
tools/perf/arch/x86/util/intel-pt.c +7 -3
index 20df442fdf36..1e96afcd8646 100644
@@ -1173,9 +1173,13 @@ static int intel_pt_read_finish(struct auxtrace_record *itr, int idx)
 	struct evsel *evsel;
 
 	evlist__for_each_entry(ptr->evlist, evsel) {
-		if (evsel->core.attr.type == ptr->intel_pt_pmu->type)
-			return perf_evlist__enable_event_idx(ptr->evlist, evsel,
-							     idx);
+		if (evsel->core.attr.type == ptr->intel_pt_pmu->type) {
+			if (evsel->disabled)
+				return 0;
+			else
+				return perf_evlist__enable_event_idx(
+						ptr->evlist, evsel, idx);
+		}
 	}
 	return -EINVAL;
 }