Skip to contentSkip to conversation

perf tools: Add a per tracepoint counter attribute to get raw sample

PATCH In review Frederic Weisbecker 1 participant 7 messages Aug 14, 2009 – Aug 15, 2009 perf events

lore
Share & tools
lore
#0 lore raw

PATCH 1/2 perf tools: Add a per tracepoint counter attribute to get raw sample

Add a new flag field while opening a tracepoint perf counter:

-e tracepoint_subsystem:tracepoint_name:flags

This is intended to be generic although for now it only supports the r[e[c[o[r[d]]]]] flag:

./perf record -e workqueue:workqueue_insertion:record ./perf record -e workqueue:workqueue_insertion:r

will have the same effect: enabling the raw samples record for the given tracepoint counter.

In the future, we may want to support further flags, separated by commas.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/builtin-record.c    |    2 +-
 tools/perf/util/parse-events.c |   10 ++++++++++
 2 files changed, 11 insertions(+), 1 deletions(-)
tools/perf/builtin-record.c +1 -1
index 5aeb632..3be0301 100644
@@ -380,7 +380,7 @@ static void create_counter(int counter, int cpu, pid_t pid)
 				  PERF_FORMAT_TOTAL_TIME_RUNNING |
 				  PERF_FORMAT_ID;
 
-	attr->sample_type	= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
+	attr->sample_type	|= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
 
 	if (freq) {
 		attr->sample_type	|= PERF_SAMPLE_PERIOD;
tools/perf/util/parse-events.c +10 -0
index a5d661b..4bab278 100644
@@ -379,6 +379,7 @@ static int parse_tracepoint_event(const char **strp,
 				    struct perf_counter_attr *attr)
 {
 	const char *evt_name;
+	char *flags;
 	char sys_name[MAX_EVENT_LENGTH];
 	char id_buf[4];
 	int fd;
@@ -400,6 +401,15 @@ static int parse_tracepoint_event(const char **strp,
 	strncpy(sys_name, *strp, sys_length);
 	sys_name[sys_length] = '\0';
 	evt_name = evt_name + 1;
+
+	flags = strchr(evt_name, ':');
+	if (flags) {
+		*flags = '\0';
+		flags++;
+		if (!strncmp(flags, "record", strlen(flags)))
+			attr->sample_type |= PERF_SAMPLE_RAW;
+	}
+
 	evt_length = strlen(evt_name);
 	if (evt_length >= MAX_EVENT_LENGTH)
 		return 0;
Signature · 1 line
1.6.2.3
#1 lore raw

PATCH 2/2 perf tools: Add a general option to enable raw sample records

While we can enable the perf sample records per tracepoint counter, we may also want to enable this option for every tracepoint counters to open, so that we don't need to add a :record flag for all of them.

Add the -R, --raw-samples options for this purpose.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/builtin-record.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)
tools/perf/builtin-record.c +5 -0
index 3be0301..e67c4fa 100644
@@ -35,6 +35,7 @@ static int			output;
 static const char		*output_name			= "perf.data";
 static int			group				= 0;
 static unsigned int		realtime_prio			= 0;
+static int			raw_samples			= 0;
 static int			system_wide			= 0;
 static int			profile_cpu			= -1;
 static pid_t			target_pid			= -1;
@@ -400,6 +401,8 @@ static void create_counter(int counter, int cpu, pid_t pid)
 	if (call_graph)
 		attr->sample_type	|= PERF_SAMPLE_CALLCHAIN;
 
+	if (raw_samples)
+		attr->sample_type	|= PERF_SAMPLE_RAW;
 
 	attr->mmap		= track;
 	attr->comm		= track;
@@ -632,6 +635,8 @@ static const struct option options[] = {
 		    "record events on existing pid"),
 	OPT_INTEGER('r', "realtime", &realtime_prio,
 		    "collect data with this RT SCHED_FIFO priority"),
+	OPT_BOOLEAN('R', "raw-samples", &raw_samples,
+		    "collect raw sample records from all opened counters"),
 	OPT_BOOLEAN('a', "all-cpus", &system_wide,
 			    "system-wide collection from all CPUs"),
 	OPT_BOOLEAN('A', "append", &append_file,
Signature · 1 line
1.6.2.3
#2 lore raw

Re: [PATCH 2/2] perf tools: Add a general option to enable raw sample records

Sorry, please ignore this one, it has been resent by mistake :-(

Quoted 48 lines · Frederic Weisbecker

While we can enable the perf sample records per tracepoint counter, we may also want to enable this option for every tracepoint counters to open, so that we don't need to add a :record flag for all of them.

Add the -R, --raw-samples options for this purpose.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/builtin-record.c |    5 +++++
 1 files changed, 5 insertions(+), 0 deletions(-)
tools/perf/builtin-record.c +5 -0
index 3be0301..e67c4fa 100644
@@ -35,6 +35,7 @@ static int			output;
 static const char		*output_name			= "perf.data";
 static int			group				= 0;
 static unsigned int		realtime_prio			= 0;
+static int			raw_samples			= 0;
 static int			system_wide			= 0;
 static int			profile_cpu			= -1;
 static pid_t			target_pid			= -1;
@@ -400,6 +401,8 @@ static void create_counter(int counter, int cpu, pid_t pid)
 	if (call_graph)
 		attr->sample_type	|= PERF_SAMPLE_CALLCHAIN;
 
+	if (raw_samples)
+		attr->sample_type	|= PERF_SAMPLE_RAW;
 
 	attr->mmap		= track;
 	attr->comm		= track;
@@ -632,6 +635,8 @@ static const struct option options[] = {
 		    "record events on existing pid"),
 	OPT_INTEGER('r', "realtime", &realtime_prio,
 		    "collect data with this RT SCHED_FIFO priority"),
+	OPT_BOOLEAN('R', "raw-samples", &raw_samples,
+		    "collect raw sample records from all opened counters"),
 	OPT_BOOLEAN('a', "all-cpus", &system_wide,
 			    "system-wide collection from all CPUs"),
 	OPT_BOOLEAN('A', "append", &append_file,
Signature · 1 line
1.6.2.3
#3 lore raw

PATCH 1/1 perf tools: Factorize the thread code in a dedicated file

In reply to Frederic Weisbecker

Factorize the thread management code used by perf-annotate and perf-report in dedicated source and header files.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/Makefile           |    1 +
 tools/perf/builtin-annotate.c |  173 ++++-------------------------------
 tools/perf/builtin-report.c   |  205 ++++------------------------------------
 tools/perf/util/thread.c      |  143 ++++++++++++++++++++++++++++
 tools/perf/util/thread.h      |   19 ++++
 5 files changed, 202 insertions(+), 339 deletions(-)
 create mode 100644 tools/perf/util/thread.c
 create mode 100644 tools/perf/util/thread.h
tools/perf/Makefile +1 -0
index 68218cf..0056405 100644
@@ -341,6 +341,7 @@ LIB_OBJS += util/callchain.o
 LIB_OBJS += util/values.o
 LIB_OBJS += util/debug.o
 LIB_OBJS += util/map.o
+LIB_OBJS += util/thread.o
 
 BUILTIN_OBJS += builtin-annotate.o
 BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +18 -155

234 lines not on this page · raw message

tools/perf/builtin-report.c +21 -184

271 lines not on this page · raw message

tools/perf/util/thread.c new file +143 -0

143 lines not on this page · raw message

tools/perf/util/thread.h new file +19 -0
new file mode 100644
index 0000000..d5edcba
@@ -0,0 +1,19 @@
+#include <linux/rbtree.h>
+#include <linux/list.h>
+#include <unistd.h>
+#include "symbol.h"
+
+struct thread {
+	struct rb_node	 rb_node;
+	struct list_head maps;
+	pid_t		 pid;
+	char		 *comm;
+};
+
+int thread__set_comm(struct thread *self, const char *comm);
+struct thread *
+threads__findnew(pid_t pid, struct rb_root *threads, struct thread *last_match);
+void thread__insert_map(struct thread *self, struct map *map);
+int thread__fork(struct thread *self, struct thread *parent);
+struct map *thread__find_map(struct thread *self, u64 ip);
+size_t threads__fprintf(FILE *fp, struct rb_root *threads);
Signature · 1 line
1.6.2.3
#4 lore raw

PATCH v2 1/1 perf tools: Factorize the thread code in a dedicated file

Factorize the thread management code used by perf-annotate and perf-report in dedicated source and header files.

v2: pass last_match by address so that it can actually be modified.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/Makefile           |    1 +
 tools/perf/builtin-annotate.c |  173 ++++-------------------------------
 tools/perf/builtin-report.c   |  205 ++++------------------------------------
 tools/perf/util/thread.c      |  143 ++++++++++++++++++++++++++++
 tools/perf/util/thread.h      |   19 ++++
 5 files changed, 202 insertions(+), 339 deletions(-)
 create mode 100644 tools/perf/util/thread.c
 create mode 100644 tools/perf/util/thread.h
tools/perf/Makefile +1 -0
index 68218cf..0056405 100644
@@ -341,6 +341,7 @@ LIB_OBJS += util/callchain.o
 LIB_OBJS += util/values.o
 LIB_OBJS += util/debug.o
 LIB_OBJS += util/map.o
+LIB_OBJS += util/thread.o
 
 BUILTIN_OBJS += builtin-annotate.o
 BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +18 -155

234 lines not on this page · raw message

tools/perf/builtin-report.c +21 -184

271 lines not on this page · raw message

tools/perf/util/thread.c new file +143 -0

143 lines not on this page · raw message

tools/perf/util/thread.h new file +19 -0
new file mode 100644
index 0000000..b1c6671
@@ -0,0 +1,19 @@
+#include <linux/rbtree.h>
+#include <linux/list.h>
+#include <unistd.h>
+#include "symbol.h"
+
+struct thread {
+	struct rb_node	 rb_node;
+	struct list_head maps;
+	pid_t		 pid;
+	char		 *comm;
+};
+
+int thread__set_comm(struct thread *self, const char *comm);
+struct thread *
+threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match);
+void thread__insert_map(struct thread *self, struct map *map);
+int thread__fork(struct thread *self, struct thread *parent);
+struct map *thread__find_map(struct thread *self, u64 ip);
+size_t threads__fprintf(FILE *fp, struct rb_root *threads);
Signature · 1 line
1.6.2.3
#5 lore raw

[tip:perfcounters/core] perf tools: Factorize the thread code in a dedicated file

Commit-ID:  6baa0a5ae0954fb2486c480a20556a9f1aee0965
Gitweb:     http://git.kernel.org/tip/6baa0a5ae0954fb2486c480a20556a9f1aee0965
Author:     Frederic Weisbecker <[email protected]>
AuthorDate: Fri, 14 Aug 2009 12:21:53 +0200
Committer:  Ingo Molnar <[email protected]>
CommitDate: Sat, 15 Aug 2009 16:10:19 +0200

perf tools: Factorize the thread code in a dedicated file

Factorize the thread management code used by perf-annotate and perf-report in dedicated source and header files.

v2: pass last_match by address so that it can actually be modified.

Cc Frederic Weisbecker <[email protected]>
Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>

LKML-Reference: <[email protected]>

---
 tools/perf/Makefile           |    1 +
 tools/perf/builtin-annotate.c |  173 ++++-------------------------------
 tools/perf/builtin-report.c   |  205 ++++------------------------------------
 tools/perf/util/thread.c      |  143 ++++++++++++++++++++++++++++
 tools/perf/util/thread.h      |   19 ++++
 5 files changed, 202 insertions(+), 339 deletions(-)
tools/perf/Makefile +1 -0
index 68218cf..0056405 100644
@@ -341,6 +341,7 @@ LIB_OBJS += util/callchain.o
 LIB_OBJS += util/values.o
 LIB_OBJS += util/debug.o
 LIB_OBJS += util/map.o
+LIB_OBJS += util/thread.o
 
 BUILTIN_OBJS += builtin-annotate.o
 BUILTIN_OBJS += builtin-help.o
tools/perf/builtin-annotate.c +18 -155

234 lines not on this page · raw message

tools/perf/builtin-report.c +21 -184

271 lines not on this page · raw message

tools/perf/util/thread.c new file +143 -0

143 lines not on this page · raw message

tools/perf/util/thread.h new file +19 -0
new file mode 100644
index 0000000..b1c6671
@@ -0,0 +1,19 @@
+#include <linux/rbtree.h>
+#include <linux/list.h>
+#include <unistd.h>
+#include "symbol.h"
+
+struct thread {
+	struct rb_node	 rb_node;
+	struct list_head maps;
+	pid_t		 pid;
+	char		 *comm;
+};
+
+int thread__set_comm(struct thread *self, const char *comm);
+struct thread *
+threads__findnew(pid_t pid, struct rb_root *threads, struct thread **last_match);
+void thread__insert_map(struct thread *self, struct map *map);
+int thread__fork(struct thread *self, struct thread *parent);
+struct map *thread__find_map(struct thread *self, u64 ip);
+size_t threads__fprintf(FILE *fp, struct rb_root *threads);
#6 lore raw
In reply to Frederic Weisbecker

Sorry, please ignore this one, it has been resent by mistake :-(

Quoted 69 lines · Frederic Weisbecker

Add a new flag field while opening a tracepoint perf counter:

-e tracepoint_subsystem:tracepoint_name:flags

This is intended to be generic although for now it only supports the r[e[c[o[r[d]]]]] flag:

./perf record -e workqueue:workqueue_insertion:record ./perf record -e workqueue:workqueue_insertion:r

will have the same effect: enabling the raw samples record for the given tracepoint counter.

In the future, we may want to support further flags, separated by commas.

Cc Peter Zijlstra <[email protected]>
Cc Arnaldo Carvalho de Melo <[email protected]>
Cc Mike Galbraith <[email protected]>
---
 tools/perf/builtin-record.c    |    2 +-
 tools/perf/util/parse-events.c |   10 ++++++++++
 2 files changed, 11 insertions(+), 1 deletions(-)
tools/perf/builtin-record.c +1 -1
index 5aeb632..3be0301 100644
@@ -380,7 +380,7 @@ static void create_counter(int counter, int cpu, pid_t pid)
 				  PERF_FORMAT_TOTAL_TIME_RUNNING |
 				  PERF_FORMAT_ID;
 
-	attr->sample_type	= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
+	attr->sample_type	|= PERF_SAMPLE_IP | PERF_SAMPLE_TID;
 
 	if (freq) {
 		attr->sample_type	|= PERF_SAMPLE_PERIOD;
tools/perf/util/parse-events.c +10 -0
index a5d661b..4bab278 100644
@@ -379,6 +379,7 @@ static int parse_tracepoint_event(const char **strp,
 				    struct perf_counter_attr *attr)
 {
 	const char *evt_name;
+	char *flags;
 	char sys_name[MAX_EVENT_LENGTH];
 	char id_buf[4];
 	int fd;
@@ -400,6 +401,15 @@ static int parse_tracepoint_event(const char **strp,
 	strncpy(sys_name, *strp, sys_length);
 	sys_name[sys_length] = '\0';
 	evt_name = evt_name + 1;
+
+	flags = strchr(evt_name, ':');
+	if (flags) {
+		*flags = '\0';
+		flags++;
+		if (!strncmp(flags, "record", strlen(flags)))
+			attr->sample_type |= PERF_SAMPLE_RAW;
+	}
+
 	evt_length = strlen(evt_name);
 	if (evt_length >= MAX_EVENT_LENGTH)
 		return 0;
Signature · 1 line
1.6.2.3