Skip to contentSkip to conversation

perf tools: Fix memory leak in read_ftrace_printk()

PATCH In review Li Zefan 3 participants 9 messages Sep 17, 2009 perf events

lore
Share & tools
lore

Li Zefan

author
#0 lore raw

PATCH 1/2 perf tools: Fix memory leak in read_ftrace_printk()

get_tracing_file() should be paired with put_tracing_file().

---
 tools/perf/util/trace-event-info.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
tools/perf/util/trace-event-info.c +4 -3
index 6c9302a..1fd824c 100644
@@ -458,7 +458,7 @@ static void read_proc_kallsyms(void)
 static void read_ftrace_printk(void)
 {
 	unsigned int size, check_size;
-	const char *path;
+	char *path;
 	struct stat st;
 	int ret;
 
@@ -468,14 +468,15 @@ static void read_ftrace_printk(void)
 		/* not found */
 		size = 0;
 		write_or_die(&size, 4);
-		return;
+		goto out;
 	}
 	size = get_size(path);
 	write_or_die(&size, 4);
 	check_size = copy_file(path);
 	if (size != check_size)
 		die("error in size of file '%s'", path);
-
+out:
+	put_tracing_file(path);
 }
 
 static struct tracepoint_path *
Signature · 1 line
1.6.3
author
#1 lore raw

PATCH 2/2 perf tools: Increase MAX_EVENT_LENGTH

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a
---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +1 -1
index 034245e..c9ef944 100644
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 40
 
 int valid_debugfs_mount(const char *debugfs)
 {
Signature · 1 line
1.6.3
#2 lore raw

Re: [PATCH 2/2] perf tools: Increase MAX_EVENT_LENGTH

Quoted 17 lines · Li Zefan

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a
---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +0 -0
index 034245e..c9ef944 100644

tools/perf/util/parse-events.c static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) · line 148

Quoted 5 lines
tools/perf/util/parse-events.c +1 -1
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 40

Ack too.

Thanks Li!

Quoted 5 lines
tools/perf/util/parse-events.c +0 -0
 int valid_debugfs_mount(const char *debugfs)
 {
Signature · 1 line
1.6.3
maintainer
#3 lore raw

Re: [PATCH 2/2] perf tools: Increase MAX_EVENT_LENGTH

Quoted 21 lines · Frederic Weisbecker
Quoted 20 lines · Li Zefan

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a
---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +0 -0
index 034245e..c9ef944 100644
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
Quoted 2 lines

tools/perf/util/parse-events.c static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) · line 148

Quoted 2 lines
tools/perf/util/parse-events.c +1 -1
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 40

This is userspace, is there any reason to be cheap with memory like this?

Why not stick in 1024 and be done for a while?

#4 lore raw

Re: [PATCH 2/2] perf tools: Increase MAX_EVENT_LENGTH

Quoted 22 lines · Peter Zijlstra
Quoted 21 lines · Frederic Weisbecker
Quoted 20 lines · Li Zefan

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a
---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +0 -0
index 034245e..c9ef944 100644
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
Quoted 8 lines
Quoted 2 lines

tools/perf/util/parse-events.c static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) · line 148

Quoted 2 lines
tools/perf/util/parse-events.c +1 -1
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 40

This is userspace, is there any reason to be cheap with memory like this?

Why not stick in 1024 and be done for a while?

Indeed, we may need to go even further than 40 in the future. But I guess 512 would be already sufficient.

author
#5 lore raw

Re: [PATCH 2/2] perf tools: Increase MAX_EVENT_LENGTH

Quoted 25 lines
Quoted 25 lines
Quoted 20 lines
Quoted 20 lines

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a
---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +0 -0
index 034245e..c9ef944 100644
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
Quoted 2 lines

tools/perf/util/parse-events.c static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir) · line 148

Quoted 2 lines
tools/perf/util/parse-events.c +1 -1
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 40

This is userspace, is there any reason to be cheap with memory like this?

Quoted 6 lines
Quoted 2 lines

Why not stick in 1024 and be done for a while?

Indeed, we may need to go even further than 40 in the future. But I guess 512 would be already sufficient.

We are not allocating memory for all tracepoints, but only those to be profiled, so a larger limit is OK.

But as the sys_name:evnt_name will be displayed using perf-list cmd, I don't think the names will be too long.

Anyway, we decide to make it 512? I'll resend the patch.

#6 lore raw

[tip:perfcounters/core] perf tools: Increase MAX_EVENT_LENGTH

In reply to Li Zefan
Commit-ID:  270bbbe80d82fad8b698d0b407eb3ad67cc3492b
Gitweb:     http://git.kernel.org/tip/270bbbe80d82fad8b698d0b407eb3ad67cc3492b
Author:     Li Zefan <[email protected]>
AuthorDate: Thu, 17 Sep 2009 16:34:51 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 18 Sep 2009 00:30:25 +0200

perf tools: Increase MAX_EVENT_LENGTH

The name length of some trace events is longer than 30, like sys_enter_sched_get_priority_max and ext4_mb_discard_preallocations.

Passing those events to perf-record will fail, try:

  # ./perf record -f -e syscalls:sys_enter_sched_get_priority_max -F 1 -a

LKML-Reference: <[email protected]>

---
 tools/perf/util/parse-events.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)
tools/perf/util/parse-events.c +1 -1
index 034245e..910283c 100644
@@ -145,7 +145,7 @@ static int tp_event_has_id(struct dirent *sys_dir, struct dirent *evt_dir)
 	   (strcmp(evt_dirent.d_name, "..")) &&				       \
 	   (!tp_event_has_id(&sys_dirent, &evt_dirent)))
 
-#define MAX_EVENT_LENGTH 30
+#define MAX_EVENT_LENGTH 512
 
 int valid_debugfs_mount(const char *debugfs)
 {
Acked-by on v1 1/2
#7 lore raw
In reply to Li Zefan
Quoted 3 lines · Li Zefan

get_tracing_file() should be paired with put_tracing_file().

Quoted 38 lines
---
 tools/perf/util/trace-event-info.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
tools/perf/util/trace-event-info.c +4 -3
index 6c9302a..1fd824c 100644
@@ -458,7 +458,7 @@ static void read_proc_kallsyms(void)
 static void read_ftrace_printk(void)
 {
 	unsigned int size, check_size;
-	const char *path;
+	char *path;
 	struct stat st;
 	int ret;
 
@@ -468,14 +468,15 @@ static void read_ftrace_printk(void)
 		/* not found */
 		size = 0;
 		write_or_die(&size, 4);
-		return;
+		goto out;
 	}
 	size = get_size(path);
 	write_or_die(&size, 4);
 	check_size = copy_file(path);
 	if (size != check_size)
 		die("error in size of file '%s'", path);
-
+out:
+	put_tracing_file(path);
 }
 
 static struct tracepoint_path *
Signature · 1 line
1.6.3
#8 lore raw
In reply to Li Zefan
Commit-ID:  6706ccf8e776e51e38ffa89fe7dd20e80eb1e860
Gitweb:     http://git.kernel.org/tip/6706ccf8e776e51e38ffa89fe7dd20e80eb1e860
Author:     Li Zefan <[email protected]>
AuthorDate: Thu, 17 Sep 2009 16:34:23 +0800
Committer:  Ingo Molnar <[email protected]>
CommitDate: Fri, 18 Sep 2009 00:30:24 +0200

perf tools: Fix memory leak in read_ftrace_printk()

get_tracing_file() should be paired with put_tracing_file().

LKML-Reference: <[email protected]>

---
 tools/perf/util/trace-event-info.c |    7 ++++---
 1 files changed, 4 insertions(+), 3 deletions(-)
tools/perf/util/trace-event-info.c +4 -3
index 6c9302a..1fd824c 100644
@@ -458,7 +458,7 @@ static void read_proc_kallsyms(void)
 static void read_ftrace_printk(void)
 {
 	unsigned int size, check_size;
-	const char *path;
+	char *path;
 	struct stat st;
 	int ret;
 
@@ -468,14 +468,15 @@ static void read_ftrace_printk(void)
 		/* not found */
 		size = 0;
 		write_or_die(&size, 4);
-		return;
+		goto out;
 	}
 	size = get_size(path);
 	write_or_die(&size, 4);
 	check_size = copy_file(path);
 	if (size != check_size)
 		die("error in size of file '%s'", path);
-
+out:
+	put_tracing_file(path);
 }
 
 static struct tracepoint_path *