Skip to contentSkip to conversation

perf record: Add support for limit perf output file size

PATCH Superseded Jiwei Sun 2 participants 5 messages Sep 25, 2019 – Oct 22, 2019 perf events

v4 of 5 Superseded Latest is v5 →

lore
Share & tools
lore

Jiwei Sun

author
#0 lore raw

PATCH v4 1/1 perf record: Add support for limit perf output file size

The patch adds a new option to limit the output file size, then based on it, we can create a wrapper of the perf command that uses the option to avoid exhausting the disk space by the unconscious user.

In order to make the perf.data parsable, we just limit the sample data size, since the perf.data consists of many headers and sample data and other data, the actual size of the recorded file will bigger than the setting value.

Testing it:

 # ./perf record -a -g --max-size=10M
Couldn't synthesize bpf events.
WARNING: The perf data has already reached the limit, stop recording!
[ perf record: Woken up 30 times to write data ]
[ perf record: Captured and wrote 10.233 MB perf.data (175650 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 11M Jul 17 14:01 perf.data
 # ./perf record -a -g --max-size=10K
WARNING: The perf data has already reached the limit, stop recording!
Couldn't synthesize bpf events.
[ perf record: Woken up 0 times to write data ]
[ perf record: Captured and wrote 1.824 MB perf.data (67 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 1.9M Jul 17 14:05 perf.data

---
v4 changes:
- Just show one WARNING message after reached the limit.

v3 changes:
- add a test result
- add the new option to tools/perf/Documentation/perf-record.txt

v2 changes:
- make patch based on latest Arnaldo's perf/core,
- display warning message when reached the limit.

---
 tools/perf/Documentation/perf-record.txt |  4 +++
 tools/perf/builtin-record.c              | 42 ++++++++++++++++++++++++
 2 files changed, 46 insertions(+)
tools/perf/Documentation/perf-record.txt +4 -0
index c6f9f31b6039..f1c6113fbc82 100644
@@ -571,6 +571,10 @@ config terms. For example: 'cycles/overwrite/' and 'instructions/no-overwrite/'.
 
 Implies --tail-synthesize.
 
+--max-size=<size>::
+Limit the sample data max size, <size> is expected to be a number with
+appended unit character - B/K/M/G
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/builtin-record.c +42 -0

72 lines not on this page · raw message

Signature · 1 line
2.20.1
author
#1 lore raw

Hello Arnaldo & Jirka,

Do you have any other suggestions regarding the patch? Any suggestions are welcome. Thank you very much.

Regards,
Jiwei

Quoted 145 lines · Jiwei Sun

The patch adds a new option to limit the output file size, then based on it, we can create a wrapper of the perf command that uses the option to avoid exhausting the disk space by the unconscious user.

In order to make the perf.data parsable, we just limit the sample data size, since the perf.data consists of many headers and sample data and other data, the actual size of the recorded file will bigger than the setting value.

Testing it:

 # ./perf record -a -g --max-size=10M
Couldn't synthesize bpf events.
WARNING: The perf data has already reached the limit, stop recording!
[ perf record: Woken up 30 times to write data ]
[ perf record: Captured and wrote 10.233 MB perf.data (175650 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 11M Jul 17 14:01 perf.data
 # ./perf record -a -g --max-size=10K
WARNING: The perf data has already reached the limit, stop recording!
Couldn't synthesize bpf events.
[ perf record: Woken up 0 times to write data ]
[ perf record: Captured and wrote 1.824 MB perf.data (67 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 1.9M Jul 17 14:05 perf.data

---
v4 changes:
- Just show one WARNING message after reached the limit.

v3 changes:
- add a test result
- add the new option to tools/perf/Documentation/perf-record.txt

v2 changes:
- make patch based on latest Arnaldo's perf/core,
- display warning message when reached the limit.

---
 tools/perf/Documentation/perf-record.txt |  4 +++
 tools/perf/builtin-record.c              | 42 ++++++++++++++++++++++++
 2 files changed, 46 insertions(+)
tools/perf/Documentation/perf-record.txt +4 -0
index c6f9f31b6039..f1c6113fbc82 100644
@@ -571,6 +571,10 @@ config terms. For example: 'cycles/overwrite/' and 'instructions/no-overwrite/'.
 
 Implies --tail-synthesize.
 
+--max-size=<size>::
+Limit the sample data max size, <size> is expected to be a number with
+appended unit character - B/K/M/G
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/builtin-record.c +42 -0
index 48600c90cc7e..30904d2a3407 100644
@@ -91,6 +91,7 @@ struct record {
 	struct switch_output	switch_output;
 	unsigned long long	samples;
 	cpu_set_t		affinity_mask;
+	unsigned long		output_max_size;	/* = 0: unlimited */
 };
 
 static volatile int auxtrace_record__snapshot_started;
@@ -120,6 +121,12 @@ static bool switch_output_time(struct record *rec)
 	       trigger_is_ready(&switch_output_trigger);
 }
 
+static bool record__output_max_size_exceeded(struct record *rec)
+{
+	return rec->output_max_size &&
+	       (rec->bytes_written >= rec->output_max_size);
+}
+
 static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 			 void *bf, size_t size)
 {
@@ -132,6 +139,12 @@ static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 
 	rec->bytes_written += size;
 
+	if (record__output_max_size_exceeded(rec)) {
+		WARN_ONCE(1, "WARNING: The perf data has already reached "
+			     "the limit, stop recording!\n");
+		raise(SIGTERM);
+	}
+
 	if (switch_output_size(rec))
 		trigger_hit(&switch_output_trigger);
 
@@ -1936,6 +1949,33 @@ static int record__parse_affinity(const struct option *opt, const char *str, int
 	return 0;
 }
 
+static int parse_output_max_size(const struct option *opt,
+				 const char *str, int unset)
+{
+	unsigned long *s = (unsigned long *)opt->value;
+	static struct parse_tag tags_size[] = {
+		{ .tag  = 'B', .mult = 1       },
+		{ .tag  = 'K', .mult = 1 << 10 },
+		{ .tag  = 'M', .mult = 1 << 20 },
+		{ .tag  = 'G', .mult = 1 << 30 },
+		{ .tag  = 0 },
+	};
+	unsigned long val;
+
+	if (unset) {
+		*s = 0;
+		return 0;
+	}
+
+	val = parse_tag_value(str, tags_size);
+	if (val != (unsigned long) -1) {
+		*s = val;
+		return 0;
+	}
+
+	return -1;
+}
+
 static int record__parse_mmap_pages(const struct option *opt,
 				    const char *str,
 				    int unset __maybe_unused)
@@ -2262,6 +2302,8 @@ static struct option __record_options[] = {
 			    "n", "Compressed records using specified level (default: 1 - fastest compression, 22 - greatest compression)",
 			    record__parse_comp_level),
 #endif
+	OPT_CALLBACK(0, "max-size", &record.output_max_size,
+		     "size", "Limit the maximum size of the output file", parse_output_max_size),
 	OPT_END()
 };
 
#2 lore raw
Quoted 4 lines · Jiwei Sun

Hello Arnaldo & Jirka,

Do you have any other suggestions regarding the patch? Any suggestions are welcome. Thank you very much.

oops, overlooked this one, sry.. I'll check on it today

jirka

Quoted 150 lines

Regards,
Jiwei

Quoted 145 lines · Jiwei Sun

The patch adds a new option to limit the output file size, then based on it, we can create a wrapper of the perf command that uses the option to avoid exhausting the disk space by the unconscious user.

In order to make the perf.data parsable, we just limit the sample data size, since the perf.data consists of many headers and sample data and other data, the actual size of the recorded file will bigger than the setting value.

Testing it:

 # ./perf record -a -g --max-size=10M
Couldn't synthesize bpf events.
WARNING: The perf data has already reached the limit, stop recording!
[ perf record: Woken up 30 times to write data ]
[ perf record: Captured and wrote 10.233 MB perf.data (175650 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 11M Jul 17 14:01 perf.data
 # ./perf record -a -g --max-size=10K
WARNING: The perf data has already reached the limit, stop recording!
Couldn't synthesize bpf events.
[ perf record: Woken up 0 times to write data ]
[ perf record: Captured and wrote 1.824 MB perf.data (67 samples) ]
Terminated
 # ls -lh perf.data
-rw------- 1 root root 1.9M Jul 17 14:05 perf.data

---
v4 changes:
- Just show one WARNING message after reached the limit.

v3 changes:
- add a test result
- add the new option to tools/perf/Documentation/perf-record.txt

v2 changes:
- make patch based on latest Arnaldo's perf/core,
- display warning message when reached the limit.

---
 tools/perf/Documentation/perf-record.txt |  4 +++
 tools/perf/builtin-record.c              | 42 ++++++++++++++++++++++++
 2 files changed, 46 insertions(+)
tools/perf/Documentation/perf-record.txt +4 -0
index c6f9f31b6039..f1c6113fbc82 100644
@@ -571,6 +571,10 @@ config terms. For example: 'cycles/overwrite/' and 'instructions/no-overwrite/'.
 
 Implies --tail-synthesize.
 
+--max-size=<size>::
+Limit the sample data max size, <size> is expected to be a number with
+appended unit character - B/K/M/G
+
 SEE ALSO
 --------
 linkperf:perf-stat[1], linkperf:perf-list[1]
tools/perf/builtin-record.c +42 -0
index 48600c90cc7e..30904d2a3407 100644
@@ -91,6 +91,7 @@ struct record {
 	struct switch_output	switch_output;
 	unsigned long long	samples;
 	cpu_set_t		affinity_mask;
+	unsigned long		output_max_size;	/* = 0: unlimited */
 };
 
 static volatile int auxtrace_record__snapshot_started;
@@ -120,6 +121,12 @@ static bool switch_output_time(struct record *rec)
 	       trigger_is_ready(&switch_output_trigger);
 }
 
+static bool record__output_max_size_exceeded(struct record *rec)
+{
+	return rec->output_max_size &&
+	       (rec->bytes_written >= rec->output_max_size);
+}
+
 static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 			 void *bf, size_t size)
 {
@@ -132,6 +139,12 @@ static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 
 	rec->bytes_written += size;
 
+	if (record__output_max_size_exceeded(rec)) {
+		WARN_ONCE(1, "WARNING: The perf data has already reached "
+			     "the limit, stop recording!\n");
+		raise(SIGTERM);
+	}
+
 	if (switch_output_size(rec))
 		trigger_hit(&switch_output_trigger);
 
@@ -1936,6 +1949,33 @@ static int record__parse_affinity(const struct option *opt, const char *str, int
 	return 0;
 }
 
+static int parse_output_max_size(const struct option *opt,
+				 const char *str, int unset)
+{
+	unsigned long *s = (unsigned long *)opt->value;
+	static struct parse_tag tags_size[] = {
+		{ .tag  = 'B', .mult = 1       },
+		{ .tag  = 'K', .mult = 1 << 10 },
+		{ .tag  = 'M', .mult = 1 << 20 },
+		{ .tag  = 'G', .mult = 1 << 30 },
+		{ .tag  = 0 },
+	};
+	unsigned long val;
+
+	if (unset) {
+		*s = 0;
+		return 0;
+	}
+
+	val = parse_tag_value(str, tags_size);
+	if (val != (unsigned long) -1) {
+		*s = val;
+		return 0;
+	}
+
+	return -1;
+}
+
 static int record__parse_mmap_pages(const struct option *opt,
 				    const char *str,
 				    int unset __maybe_unused)
@@ -2262,6 +2302,8 @@ static struct option __record_options[] = {
 			    "n", "Compressed records using specified level (default: 1 - fastest compression, 22 - greatest compression)",
 			    record__parse_comp_level),
 #endif
+	OPT_CALLBACK(0, "max-size", &record.output_max_size,
+		     "size", "Limit the maximum size of the output file", parse_output_max_size),
 	OPT_END()
 };
 
#3 lore raw
In reply to Jiwei Sun

On Wed, Sep 25, 2019 at 03:06:37PM +0800, Jiwei Sun wrote:

SNIP

Quoted 29 lines

SEE ALSO
--------
linkperf:perf-stat[1], linkperf:perf-list[1]

tools/perf/builtin-record.c +7 -0
index 48600c90cc7e..30904d2a3407 100644
@@ -91,6 +91,7 @@ struct record {
 	struct switch_output	switch_output;
 	unsigned long long	samples;
 	cpu_set_t		affinity_mask;
+	unsigned long		output_max_size;	/* = 0: unlimited */
 };
 
 static volatile int auxtrace_record__snapshot_started;
@@ -120,6 +121,12 @@ static bool switch_output_time(struct record *rec)
 	       trigger_is_ready(&switch_output_trigger);
 }
 
+static bool record__output_max_size_exceeded(struct record *rec)
+{
+	return rec->output_max_size &&
+	       (rec->bytes_written >= rec->output_max_size);
+}
+
 static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 			 void *bf, size_t size)
 {

tools/perf/builtin-record.c static int record__write(struct record *rec, struct mmap *map __maybe_unused, · line 144

Quoted 6 lines
tools/perf/builtin-record.c +3 -0
@@ -132,6 +139,12 @@ static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 
 	rec->bytes_written += size;
 
+	if (record__output_max_size_exceeded(rec)) {
+		WARN_ONCE(1, "WARNING: The perf data has already reached "
+			     "the limit, stop recording!\n");

I think the message whouldn't be a warning, the user asked for that, maybe something more like:

[ perf record: perf size limit reached (XXMB), stopping session ]

tools/perf/builtin-record.c static int record__write(struct record *rec, struct mmap *map __maybe_unused, · line 145

Quoted 1 line
tools/perf/builtin-record.c +1 -0
+		raise(SIGTERM);

could we just set 'done = 1' what's the benefit in killing perf?

thanks,
jirka

author
#4 lore raw

Hi Jirka,

Quoted 43 lines · Jiri Olsa

On Wed, Sep 25, 2019 at 03:06:37PM +0800, Jiwei Sun wrote:

SNIP

Quoted 29 lines

SEE ALSO
--------
linkperf:perf-stat[1], linkperf:perf-list[1]

tools/perf/builtin-record.c +7 -0
index 48600c90cc7e..30904d2a3407 100644
@@ -91,6 +91,7 @@ struct record {
 	struct switch_output	switch_output;
 	unsigned long long	samples;
 	cpu_set_t		affinity_mask;
+	unsigned long		output_max_size;	/* = 0: unlimited */
 };
 
 static volatile int auxtrace_record__snapshot_started;
@@ -120,6 +121,12 @@ static bool switch_output_time(struct record *rec)
 	       trigger_is_ready(&switch_output_trigger);
 }
 
+static bool record__output_max_size_exceeded(struct record *rec)
+{
+	return rec->output_max_size &&
+	       (rec->bytes_written >= rec->output_max_size);
+}
+
 static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 			 void *bf, size_t size)
 {

tools/perf/builtin-record.c static int record__write(struct record *rec, struct mmap *map __maybe_unused, · line 144

Quoted 6 lines
tools/perf/builtin-record.c +3 -0
@@ -132,6 +139,12 @@ static int record__write(struct record *rec, struct mmap *map __maybe_unused,
 
 	rec->bytes_written += size;
 
+	if (record__output_max_size_exceeded(rec)) {
+		WARN_ONCE(1, "WARNING: The perf data has already reached "
+			     "the limit, stop recording!\n");

I think the message whouldn't be a warning, the user asked for that, maybe something more like:

Quoted 5 lines

[ perf record: perf size limit reached (XXMB), stopping session ]

tools/perf/builtin-record.c static int record__write(struct record *rec, struct mmap *map __maybe_unused, · line 145

Quoted 1 line
tools/perf/builtin-record.c +1 -0
+		raise(SIGTERM);

could we just set 'done = 1' what's the benefit in killing perf?

Thanks for your suggestions. Yes, if just set "done == 1" is more efficient and more concise. And I will modify it and the output format, and then send a v5 patch.
Thanks again.

Regards,
Jiwei

Quoted 5 lines

thanks,
jirka