Skip to contentSkip to conversation

perf tools: Fix use of wrong event when processing exit events

PATCH In review Adrian Hunter 3 participants 5 messages Aug 18, 2015 – Sep 5, 2015 perf events

lore
Share & tools
lore

Adrian Hunter

author reviewer
#0 lore raw

PATCH 1/1 perf tools: Fix use of wrong event when processing exit events

In a couple of cases the 'comm' member of 'union event' has been used instead of the correct member ('fork') when processing exit events.

In the cases where it has been used incorrectly, only the 'pid' and 'tid' are affected. The 'pid' value would be correct anyway because it is in the same position in 'comm' and 'fork' events, but the 'tid' would have been incorrectly assigned from 'ppid'. However, for exit events, the kernel puts the current task in the 'ppid' and 'ttid' which is the same as the exiting task. That is 'ppid' == 'pid' and if the task is not multi-threaded, 'pid' == 'tid' i.e. the data goes wrong only when tracing multi-threaded programs.

It is hard to find an example of how this would produce an error in practice. There are 3 occurences of the fix:
1. perf script is only affected if !sample_id_all which only happens on old kernels
2. intel_pt is only affected when decoding without timestamps and would probably still decode correctly - the exit event is only used to flush out data which anyway gets flushed at the end of the session
3. intel_bts also uses the exit event to flush data which would probably not cause errors as it would get flushed at the end of the session instead

---
 tools/perf/builtin-script.c | 4 ++--
 tools/perf/util/intel-bts.c | 2 +-
 tools/perf/util/intel-pt.c  | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +2 -2
index 105332e950a9..17030c6c64b6 100644
@@ -739,8 +739,8 @@ static int process_exit_event(struct perf_tool *tool,
 	if (!evsel->attr.sample_id_all) {
 		sample->cpu = 0;
 		sample->time = 0;
-		sample->tid = event->comm.tid;
-		sample->pid = event->comm.pid;
+		sample->tid = event->fork.tid;
+		sample->pid = event->fork.pid;
 	}
 	print_sample_start(sample, thread, evsel);
 	perf_event__fprintf(event, stdout);
tools/perf/util/intel-bts.c +1 -1
index dce99cfb1309..09d142b362fa 100644
@@ -623,7 +623,7 @@ static int intel_bts_process_event(struct perf_session *session,
 	if (err)
 		return err;
 	if (event->header.type == PERF_RECORD_EXIT) {
-		err = intel_bts_process_tid_exit(bts, event->comm.tid);
+		err = intel_bts_process_tid_exit(bts, event->fork.tid);
 		if (err)
 			return err;
 	}
tools/perf/util/intel-pt.c +1 -1
index 2a4a4120473b..ef2fb13f1dfa 100644
@@ -1464,7 +1464,7 @@ static int intel_pt_process_event(struct perf_session *session,
 	if (pt->timeless_decoding) {
 		if (event->header.type == PERF_RECORD_EXIT) {
 			err = intel_pt_process_timeless_queues(pt,
-							       event->comm.tid,
+							       event->fork.tid,
 							       sample->time);
 		}
 	} else if (timestamp) {
Signature · 1 line
1.9.1
author reviewer
#1 lore raw
Quoted 26 lines · Adrian Hunter

In a couple of cases the 'comm' member of 'union event' has been used instead of the correct member ('fork') when processing exit events.

In the cases where it has been used incorrectly, only the 'pid' and 'tid' are affected. The 'pid' value would be correct anyway because it is in the same position in 'comm' and 'fork' events, but the 'tid' would have been incorrectly assigned from 'ppid'. However, for exit events, the kernel puts the current task in the 'ppid' and 'ttid' which is the same as the exiting task. That is 'ppid' == 'pid' and if the task is not multi-threaded, 'pid' == 'tid' i.e. the data goes wrong only when tracing multi-threaded programs.

It is hard to find an example of how this would produce an error in practice. There are 3 occurences of the fix:
1. perf script is only affected if !sample_id_all which only happens on old kernels
2. intel_pt is only affected when decoding without timestamps and would probably still decode correctly - the exit event is only used to flush out data which anyway gets flushed at the end of the session
3. intel_bts also uses the exit event to flush data which would probably not cause errors as it would get flushed at the end of the session instead

Quoted 1 line

I think it would be worth picking this one up for 4.3

Quoted 48 lines
---
 tools/perf/builtin-script.c | 4 ++--
 tools/perf/util/intel-bts.c | 2 +-
 tools/perf/util/intel-pt.c  | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +2 -2
index 105332e950a9..17030c6c64b6 100644
@@ -739,8 +739,8 @@ static int process_exit_event(struct perf_tool *tool,
 	if (!evsel->attr.sample_id_all) {
 		sample->cpu = 0;
 		sample->time = 0;
-		sample->tid = event->comm.tid;
-		sample->pid = event->comm.pid;
+		sample->tid = event->fork.tid;
+		sample->pid = event->fork.pid;
 	}
 	print_sample_start(sample, thread, evsel);
 	perf_event__fprintf(event, stdout);
tools/perf/util/intel-bts.c +1 -1
index dce99cfb1309..09d142b362fa 100644
@@ -623,7 +623,7 @@ static int intel_bts_process_event(struct perf_session *session,
 	if (err)
 		return err;
 	if (event->header.type == PERF_RECORD_EXIT) {
-		err = intel_bts_process_tid_exit(bts, event->comm.tid);
+		err = intel_bts_process_tid_exit(bts, event->fork.tid);
 		if (err)
 			return err;
 	}
tools/perf/util/intel-pt.c +1 -1
index 2a4a4120473b..ef2fb13f1dfa 100644
@@ -1464,7 +1464,7 @@ static int intel_pt_process_event(struct perf_session *session,
 	if (pt->timeless_decoding) {
 		if (event->header.type == PERF_RECORD_EXIT) {
 			err = intel_pt_process_timeless_queues(pt,
-							       event->comm.tid,
+							       event->fork.tid,
 							       sample->time);
 		}
 	} else if (timestamp) {
#2 lore raw

Em Wed, Sep 02, 2015 at 04:00:08PM +0300, Adrian Hunter escreveu:

Quoted 27 lines · Adrian Hunter
Quoted 26 lines

In a couple of cases the 'comm' member of 'union event' has been used instead of the correct member ('fork') when processing exit events.

In the cases where it has been used incorrectly, only the 'pid' and 'tid' are affected. The 'pid' value would be correct anyway because it is in the same position in 'comm' and 'fork' events, but the 'tid' would have been incorrectly assigned from 'ppid'. However, for exit events, the kernel puts the current task in the 'ppid' and 'ttid' which is the same as the exiting task. That is 'ppid' == 'pid' and if the task is not multi-threaded, 'pid' == 'tid' i.e. the data goes wrong only when tracing multi-threaded programs.

It is hard to find an example of how this would produce an error in practice. There are 3 occurences of the fix:
1. perf script is only affected if !sample_id_all which only happens on old kernels
2. intel_pt is only affected when decoding without timestamps and would probably still decode correctly - the exit event is only used to flush out data which anyway gets flushed at the end of the session
3. intel_bts also uses the exit event to flush data which would probably not cause errors as it would get flushed at the end of the session instead

Quoted 3 lines
Quoted 1 line

I think it would be worth picking this one up for 4.3

I guess I had this picked up, is there anything else from you that is outstanding?

- Arnaldo

Quoted 48 lines
Quoted 48 lines
---
 tools/perf/builtin-script.c | 4 ++--
 tools/perf/util/intel-bts.c | 2 +-
 tools/perf/util/intel-pt.c  | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +2 -2
index 105332e950a9..17030c6c64b6 100644
@@ -739,8 +739,8 @@ static int process_exit_event(struct perf_tool *tool,
 	if (!evsel->attr.sample_id_all) {
 		sample->cpu = 0;
 		sample->time = 0;
-		sample->tid = event->comm.tid;
-		sample->pid = event->comm.pid;
+		sample->tid = event->fork.tid;
+		sample->pid = event->fork.pid;
 	}
 	print_sample_start(sample, thread, evsel);
 	perf_event__fprintf(event, stdout);
tools/perf/util/intel-bts.c +1 -1
index dce99cfb1309..09d142b362fa 100644
@@ -623,7 +623,7 @@ static int intel_bts_process_event(struct perf_session *session,
 	if (err)
 		return err;
 	if (event->header.type == PERF_RECORD_EXIT) {
-		err = intel_bts_process_tid_exit(bts, event->comm.tid);
+		err = intel_bts_process_tid_exit(bts, event->fork.tid);
 		if (err)
 			return err;
 	}
tools/perf/util/intel-pt.c +1 -1
index 2a4a4120473b..ef2fb13f1dfa 100644
@@ -1464,7 +1464,7 @@ static int intel_pt_process_event(struct perf_session *session,
 	if (pt->timeless_decoding) {
 		if (event->header.type == PERF_RECORD_EXIT) {
 			err = intel_pt_process_timeless_queues(pt,
-							       event->comm.tid,
+							       event->fork.tid,
 							       sample->time);
 		}
 	} else if (timestamp) {
author reviewer
#3 lore raw
Quoted 28 lines · Arnaldo Carvalho de Melo

Em Wed, Sep 02, 2015 at 04:00:08PM +0300, Adrian Hunter escreveu:

Quoted 27 lines · Adrian Hunter
Quoted 26 lines

In a couple of cases the 'comm' member of 'union event' has been used instead of the correct member ('fork') when processing exit events.

In the cases where it has been used incorrectly, only the 'pid' and 'tid' are affected. The 'pid' value would be correct anyway because it is in the same position in 'comm' and 'fork' events, but the 'tid' would have been incorrectly assigned from 'ppid'. However, for exit events, the kernel puts the current task in the 'ppid' and 'ttid' which is the same as the exiting task. That is 'ppid' == 'pid' and if the task is not multi-threaded, 'pid' == 'tid' i.e. the data goes wrong only when tracing multi-threaded programs.

It is hard to find an example of how this would produce an error in practice. There are 3 occurences of the fix:
1. perf script is only affected if !sample_id_all which only happens on old kernels
2. intel_pt is only affected when decoding without timestamps and would probably still decode correctly - the exit event is only used to flush out data which anyway gets flushed at the end of the session
3. intel_bts also uses the exit event to flush data which would probably not cause errors as it would get flushed at the end of the session instead

Quoted 6 lines
Quoted 3 lines
Quoted 1 line

I think it would be worth picking this one up for 4.3

I guess I had this picked up, is there anything else from you that is outstanding?

Only "perf tools: Add support for PERF_RECORD_SWITCH to Intel PT" but that is a feature not a bug.

#4 lore raw
In reply to Adrian Hunter
perf tools: Fix use of wrong event when processing exit events
Commit-ID:  53ff6bc37be449f546158a39c528d7814dfb15a1
Gitweb:     http://git.kernel.org/tip/53ff6bc37be449f546158a39c528d7814dfb15a1
Author:     Adrian Hunter <[email protected]>
AuthorDate: Tue, 18 Aug 2015 12:07:05 +0300
Committer:  Arnaldo Carvalho de Melo <[email protected]>
CommitDate: Wed, 2 Sep 2015 17:46:26 -0300

perf tools: Fix use of wrong event when processing exit events

In a couple of cases the 'comm' member of 'union event' has been used instead of the correct member ('fork') when processing exit events.

In the cases where it has been used incorrectly, only the 'pid' and 'tid' are affected. The 'pid' value would be correct anyway because it is in the same position in 'comm' and 'fork' events, but the 'tid' would have been incorrectly assigned from 'ppid'.

However, for exit events, the kernel puts the current task in the 'ppid' and 'ttid' which is the same as the exiting task. That is 'ppid' == 'pid' and if the task is not multi-threaded, 'pid' == 'tid' i.e. the data goes wrong only when tracing multi-threaded programs.

It is hard to find an example of how this would produce an error in practice. There are 3 occurences of the fix:

1. perf script is only affected if !sample_id_all which only happens on old kernels.

2. intel_pt is only affected when decoding without timestamps and would probably still decode correctly - the exit event is only used to flush out data which anyway gets flushed at the end of the session

3. intel_bts also uses the exit event to flush data which would probably not cause errors as it would get flushed at the end of the session instead

---
 tools/perf/builtin-script.c | 4 ++--
 tools/perf/util/intel-bts.c | 2 +-
 tools/perf/util/intel-pt.c  | 2 +-
 3 files changed, 4 insertions(+), 4 deletions(-)
tools/perf/builtin-script.c +2 -2
index eb51325..284a76e 100644
@@ -768,8 +768,8 @@ static int process_exit_event(struct perf_tool *tool,
 	if (!evsel->attr.sample_id_all) {
 		sample->cpu = 0;
 		sample->time = 0;
-		sample->tid = event->comm.tid;
-		sample->pid = event->comm.pid;
+		sample->tid = event->fork.tid;
+		sample->pid = event->fork.pid;
 	}
 	print_sample_start(sample, thread, evsel);
 	perf_event__fprintf(event, stdout);
tools/perf/util/intel-bts.c +1 -1
index ea76862..eb0e7f8 100644
@@ -623,7 +623,7 @@ static int intel_bts_process_event(struct perf_session *session,
 	if (err)
 		return err;
 	if (event->header.type == PERF_RECORD_EXIT) {
-		err = intel_bts_process_tid_exit(bts, event->comm.tid);
+		err = intel_bts_process_tid_exit(bts, event->fork.tid);
 		if (err)
 			return err;
 	}
tools/perf/util/intel-pt.c +1 -1
index bb41c20..535d86f 100644
@@ -1494,7 +1494,7 @@ static int intel_pt_process_event(struct perf_session *session,
 	if (pt->timeless_decoding) {
 		if (event->header.type == PERF_RECORD_EXIT) {
 			err = intel_pt_process_timeless_queues(pt,
-							       event->comm.tid,
+							       event->fork.tid,
 							       sample->time);
 		}
 	} else if (timestamp) {