Skip to contentSkip to conversation

Add procfs based memory and network tool events

PATCH Superseded Ian Rogers 3 participants 9 messages Jan 4 – Jan 15 perf events

v1 of 2 Superseded Latest is v2 →

lore
Share & tools
lore

Ian Rogers

author reviewer
#0 lore raw

cover Add procfs based memory and network tool events

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:

```
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -p $(pidof -d, chrome) -I 1000
     1.001023475                  0      net_rx_bytes
     1.001023475                  0      net_rx_compressed
     1.001023475         42,647,328      net_rx_drop
     1.001023475        463,069,152      net_rx_errors
     1.001023475                  0      net_rx_fifo
     1.001023475                  0      net_rx_frame
     1.001023475                  0      net_rx_multicast
     1.001023475    423,195,831,744      net_rx_packets
     1.001023475                  0      net_tx_bytes
     1.001023475                  0      net_tx_carrier
     1.001023475                  0      net_tx_colls
     1.001023475                  0      net_tx_compressed
     1.001023475                  0      net_tx_drop
     1.001023475                  0      net_tx_errors
     1.001023475                  0      net_tx_fifo
     1.001023475                  0      net_tx_packets
```

As the events are in the tool_pmu they can be used in metrics. The json descriptions they are exposed in perf list and the events can be seen in the python ilist application.

Note, if a process terminates then the count reading returns an error and this can expose what appear to be latent bugs in the aggregation and display code.

Ian Rogers (2):
perf tool_pmu: Add memory events
perf tool_pmu: Add network events

 tools/perf/builtin-stat.c                     |  10 +-
 .../pmu-events/arch/common/common/tool.json   | 266 ++++++++-
 tools/perf/pmu-events/empty-pmu-events.c      | 312 +++++++----
 tools/perf/util/tool_pmu.c                    | 514 +++++++++++++++++-
 tools/perf/util/tool_pmu.h                    |  44 ++
 5 files changed, 1026 insertions(+), 120 deletions(-)
Signature · 1 line
2.52.0.351.gbe84eed79e-goog
author reviewer
#1 lore raw

PATCH 1/2 perf tool_pmu: Add memory events

Add tool PMU events to report memory usage information exposed by /proc/pid/smaps_rollup and /proc/pid/statm.

The following events are added:
- memory_anon_huge_pages
- memory_anonymous
- memory_data
- memory_file_pmd_mapped
- memory_ksm
- memory_lazyfree
- memory_locked
- memory_private_clean
- memory_private_dirty
- memory_private_hugetlb
- memory_pss
- memory_pss_anon
- memory_pss_dirty
- memory_pss_file
- memory_pss_shmem
- memory_referenced
- memory_resident
- memory_rss
- memory_shared
- memory_shared_clean
- memory_shared_dirty
- memory_shared_hugetlb
- memory_shmem_pmd_mapped
- memory_size
- memory_swap
- memory_swap_pss
- memory_text
- memory_uss

The events are aggregated across processes in system-wide mode, or reported per-process when monitoring specific tasks. The tool_pmu implementation is updated to read from /proc files and tool.json is updated with the new events and detailed descriptions. The data can only be gathered for a running process except for memory_rss and memory_resident that use rusage data from wait4 to give a final value.

Examples of gathering system-wide statm and smaps_rollup based events:

```
$ perf stat -e memory_size,memory_resident,memory_shared,memory_text,memory_data -I 1000
     1.001046524 94,942,406,868,992      memory_size
     1.001046524     24,885,137,408      memory_resident
     1.001046524     15,494,561,792      memory_shared
     1.001046524     15,249,453,056      memory_text
     1.001046524     74,510,024,704      memory_data
...
$ perf stat -e memory_anon_huge_pages,memory_anonymous,memory_file_pmd_mapped,memory_ksm,memory_lazyfree,memory_locked,memory_private_clean,memory_private_dirty,memory_private_hugetlb,memory_pss,memory_pss_anon,memory_pss_dirty,memory_pss_file,memory_pss_shmem,memory_referenced,memory_rss,memory_shared,memory_shared_clean,memory_shared_dirty,memory_shared_hugetlb,memory_shmem_pmd_mapped,memory_swap,memory_swap_pss,memory_uss -I 1000
     1.001107268        981,467,136      memory_anon_huge_pages
     1.001107268      9,459,105,792      memory_anonymous
     1.001107268                  0      memory_file_pmd_mapped
     1.001107268                  0      memory_ksm
     1.001107268          1,269,760      memory_lazyfree
     1.001107268            634,880      memory_locked
     1.001107268        958,963,712      memory_private_clean
     1.001107268      8,400,080,896      memory_private_dirty
     1.001107268                  0      memory_private_hugetlb
     1.001107268     10,711,451,648      memory_pss
     1.001107268      8,387,961,856      memory_pss_anon
     1.001107268      8,642,944,000      memory_pss_dirty
     1.001107268      2,078,282,752      memory_pss_file
     1.001107268        245,125,120      memory_pss_shmem
     1.001107268     24,356,818,944      memory_referenced
     1.001107268     25,256,157,184      memory_rss
     1.001107268     15,757,029,376      memory_shared
     1.001107268     14,230,978,560      memory_shared_clean
     1.001107268      1,670,934,528      memory_shared_dirty
     1.001107268                  0      memory_shared_hugetlb
     1.001107268                  0      memory_shmem_pmd_mapped
     1.001107268                  0      memory_swap
     1.001107268                  0      memory_swap_pss
     1.001107268      9,328,041,984      memory_uss
...
```
---
 tools/perf/builtin-stat.c                     |  10 +-
 .../pmu-events/arch/common/common/tool.json   | 168 ++++++++
 tools/perf/pmu-events/empty-pmu-events.c      | 280 +++++++------
 tools/perf/util/tool_pmu.c                    | 375 +++++++++++++++++-
 tools/perf/util/tool_pmu.h                    |  28 ++
 5 files changed, 742 insertions(+), 119 deletions(-)
tools/perf/builtin-stat.c +8 -2
index ab40d85fb125..8c6290174afb 100644
@@ -107,6 +107,7 @@
 struct rusage_stats {
 	struct stats ru_utime_usec_stat;
 	struct stats ru_stime_usec_stat;
+	struct stats ru_maxrss_stat;
 };
 
 static void print_counters(struct timespec *ts, int argc, const char **argv);
@@ -288,7 +289,9 @@ static int read_single_counter(struct evsel *counter, int cpu_map_idx, int threa
 	 */
 	if (err && cpu_map_idx == 0 &&
 	    (evsel__tool_event(counter) == TOOL_PMU__EVENT_USER_TIME ||
-	     evsel__tool_event(counter) == TOOL_PMU__EVENT_SYSTEM_TIME)) {
+	     evsel__tool_event(counter) == TOOL_PMU__EVENT_SYSTEM_TIME ||
+	     evsel__tool_event(counter) == TOOL_PMU__EVENT_MEMORY_RESIDENT ||
+	     evsel__tool_event(counter) == TOOL_PMU__EVENT_MEMORY_RSS)) {
 		struct perf_counts_values *count =
 			perf_counts(counter->counts, cpu_map_idx, thread);
 		struct perf_counts_values *old_count = NULL;
@@ -299,8 +302,10 @@ static int read_single_counter(struct evsel *counter, int cpu_map_idx, int threa
 
 		if (evsel__tool_event(counter) == TOOL_PMU__EVENT_USER_TIME)
 			val = ru_stats.ru_utime_usec_stat.mean;
-		else
+		else if (evsel__tool_event(counter) == TOOL_PMU__EVENT_SYSTEM_TIME)
 			val = ru_stats.ru_stime_usec_stat.mean;
+		else
+			val = ru_stats.ru_maxrss_stat.mean;
 
 		count->val = val;
 		if (old_count) {
@@ -778,6 +783,7 @@ static void update_rusage_stats(const struct rusage *rusage)
 		(rusage->ru_utime.tv_usec * us_to_ns + rusage->ru_utime.tv_sec * s_to_ns));
 	update_stats(&ru_stats.ru_stime_usec_stat,
 		(rusage->ru_stime.tv_usec * us_to_ns + rusage->ru_stime.tv_sec * s_to_ns));
+	update_stats(&ru_stats.ru_maxrss_stat, rusage->ru_maxrss * 1024);
 }
 
 static int __run_perf_stat(int argc, const char **argv, int run_idx)
tools/perf/pmu-events/arch/common/common/tool.json +168 -0

173 lines not on this page · raw message

tools/perf/pmu-events/empty-pmu-events.c +168 -112

364 lines not on this page · raw message

tools/perf/util/tool_pmu.c +370 -5

437 lines not on this page · raw message

tools/perf/util/tool_pmu.h +28 -0
index ea343d1983d3..bf6bb196ad75 100644
@@ -24,6 +24,34 @@ enum tool_pmu_event {
 	TOOL_PMU__EVENT_SYSTEM_TSC_FREQ,
 	TOOL_PMU__EVENT_CORE_WIDE,
 	TOOL_PMU__EVENT_TARGET_CPU,
+	TOOL_PMU__EVENT_MEMORY_ANON_HUGE_PAGES,
+	TOOL_PMU__EVENT_MEMORY_ANONYMOUS,
+	TOOL_PMU__EVENT_MEMORY_DATA,
+	TOOL_PMU__EVENT_MEMORY_FILE_PMD_MAPPED,
+	TOOL_PMU__EVENT_MEMORY_KSM,
+	TOOL_PMU__EVENT_MEMORY_LAZYFREE,
+	TOOL_PMU__EVENT_MEMORY_LOCKED,
+	TOOL_PMU__EVENT_MEMORY_PRIVATE_CLEAN,
+	TOOL_PMU__EVENT_MEMORY_PRIVATE_DIRTY,
+	TOOL_PMU__EVENT_MEMORY_PRIVATE_HUGETLB,
+	TOOL_PMU__EVENT_MEMORY_PSS,
+	TOOL_PMU__EVENT_MEMORY_PSS_ANON,
+	TOOL_PMU__EVENT_MEMORY_PSS_DIRTY,
+	TOOL_PMU__EVENT_MEMORY_PSS_FILE,
+	TOOL_PMU__EVENT_MEMORY_PSS_SHMEM,
+	TOOL_PMU__EVENT_MEMORY_REFERENCED,
+	TOOL_PMU__EVENT_MEMORY_RESIDENT,
+	TOOL_PMU__EVENT_MEMORY_RSS,
+	TOOL_PMU__EVENT_MEMORY_SHARED,
+	TOOL_PMU__EVENT_MEMORY_SHARED_CLEAN,
+	TOOL_PMU__EVENT_MEMORY_SHARED_DIRTY,
+	TOOL_PMU__EVENT_MEMORY_SHARED_HUGETLB,
+	TOOL_PMU__EVENT_MEMORY_SHMEM_PMD_MAPPED,
+	TOOL_PMU__EVENT_MEMORY_SIZE,
+	TOOL_PMU__EVENT_MEMORY_SWAP,
+	TOOL_PMU__EVENT_MEMORY_SWAP_PSS,
+	TOOL_PMU__EVENT_MEMORY_TEXT,
+	TOOL_PMU__EVENT_MEMORY_USS,
 
 	TOOL_PMU__EVENT_MAX,
 };
Signature · 1 line
2.52.0.351.gbe84eed79e-goog
author reviewer
#2 lore raw

PATCH 2/2 perf tool_pmu: Add network events

In reply to Ian Rogers

Add tool PMU events to report network statistics from /proc/net/dev. The events can be read system-wide (from /proc/net/dev) or per-process (from /proc/pid/net/dev).

The following events are added for both RX (receive) and TX (transmit):
- bytes
- packets
- errors
- drop
- fifo
- frame (RX only) / colls (TX only)
- compressed
- multicast (RX only) / carrier (TX only)

Updates tool.json with the new events and descriptions. Updates tool_pmu implementation to parse the net/dev format.

Below are examples of system-wide and per-process gathering respectively:

```
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -I 1000
     1.001154872                  0      net_rx_bytes
     1.001154872                  0      net_rx_compressed
     1.001154872            444,577      net_rx_drop
     1.001154872          4,824,888      net_rx_errors
     1.001154872                  0      net_rx_fifo
     1.001154872                  0      net_rx_frame
     1.001154872                  0      net_rx_multicast
     1.001154872      4,408,889,452      net_rx_packets
     1.001154872                  0      net_tx_bytes
     1.001154872                  0      net_tx_carrier
     1.001154872                  0      net_tx_colls
     1.001154872                  0      net_tx_compressed
     1.001154872                  0      net_tx_drop
     1.001154872                  0      net_tx_errors
     1.001154872                  0      net_tx_fifo
     1.001154872                  0      net_tx_packets
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -p $(pidof -d, chrome) -I 1000
     1.001023475                  0      net_rx_bytes
     1.001023475                  0      net_rx_compressed
     1.001023475         42,647,328      net_rx_drop
     1.001023475        463,069,152      net_rx_errors
     1.001023475                  0      net_rx_fifo
     1.001023475                  0      net_rx_frame
     1.001023475                  0      net_rx_multicast
     1.001023475    423,195,831,744      net_rx_packets
     1.001023475                  0      net_tx_bytes
     1.001023475                  0      net_tx_carrier
     1.001023475                  0      net_tx_colls
     1.001023475                  0      net_tx_compressed
     1.001023475                  0      net_tx_drop
     1.001023475                  0      net_tx_errors
     1.001023475                  0      net_tx_fifo
     1.001023475                  0      net_tx_packets
...
```
---
 .../pmu-events/arch/common/common/tool.json   |  98 ++++++-
 tools/perf/pmu-events/empty-pmu-events.c      | 256 ++++++++++--------
 tools/perf/util/tool_pmu.c                    | 155 ++++++++++-
 tools/perf/util/tool_pmu.h                    |  16 ++
 4 files changed, 404 insertions(+), 121 deletions(-)
tools/perf/pmu-events/arch/common/common/tool.json +97 -1
index 4b3fce655f8a..ebd3c5a6d15d 100644
@@ -250,5 +250,101 @@
     "EventName": "memory_uss",
     "BriefDescription": "Unique Set Size (USS) in bytes",
     "ConfigCode": "42"
-  }
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_bytes",
+    "BriefDescription": "Network received bytes",
+    "ConfigCode": "43"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_packets",
+    "BriefDescription": "Network received packets",
+    "ConfigCode": "44"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_errors",
+    "BriefDescription": "Network received errors",
+    "ConfigCode": "45"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_drop",
+    "BriefDescription": "Network received dropped packets",
+    "ConfigCode": "46"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_fifo",
+    "BriefDescription": "Network received fifo overruns",
+    "ConfigCode": "47"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_frame",
+    "BriefDescription": "Network received framing errors",
+    "ConfigCode": "48"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_compressed",
+    "BriefDescription": "Network received compressed packets",
+    "ConfigCode": "49"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_rx_multicast",
+    "BriefDescription": "Network received multicast packets",
+    "ConfigCode": "50"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_bytes",
+    "BriefDescription": "Network transmitted bytes",
+    "ConfigCode": "51"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_packets",
+    "BriefDescription": "Network transmitted packets",
+    "ConfigCode": "52"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_errors",
+    "BriefDescription": "Network transmitted errors",
+    "ConfigCode": "53"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_drop",
+    "BriefDescription": "Network transmitted dropped packets",
+    "ConfigCode": "54"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_fifo",
+    "BriefDescription": "Network transmitted fifo overruns",
+    "ConfigCode": "55"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_colls",
+    "BriefDescription": "Network transmitted collisions",
+    "ConfigCode": "56"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_carrier",
+    "BriefDescription": "Network transmitted carrier losses",
+    "ConfigCode": "57"
+  },
+  {
+    "Unit": "tool",
+    "EventName": "net_tx_compressed",
+    "BriefDescription": "Network transmitted compressed packets",
+    "ConfigCode": "58"
+   }
 ]
tools/perf/pmu-events/empty-pmu-events.c +144 -112

340 lines not on this page · raw message

tools/perf/util/tool_pmu.c +147 -8

219 lines not on this page · raw message

tools/perf/util/tool_pmu.h +16 -0
index bf6bb196ad75..be8ebd9aacfb 100644
@@ -52,6 +52,22 @@ enum tool_pmu_event {
 	TOOL_PMU__EVENT_MEMORY_SWAP_PSS,
 	TOOL_PMU__EVENT_MEMORY_TEXT,
 	TOOL_PMU__EVENT_MEMORY_USS,
+	TOOL_PMU__EVENT_NET_RX_BYTES,
+	TOOL_PMU__EVENT_NET_RX_PACKETS,
+	TOOL_PMU__EVENT_NET_RX_ERRORS,
+	TOOL_PMU__EVENT_NET_RX_DROP,
+	TOOL_PMU__EVENT_NET_RX_FIFO,
+	TOOL_PMU__EVENT_NET_RX_FRAME,
+	TOOL_PMU__EVENT_NET_RX_COMPRESSED,
+	TOOL_PMU__EVENT_NET_RX_MULTICAST,
+	TOOL_PMU__EVENT_NET_TX_BYTES,
+	TOOL_PMU__EVENT_NET_TX_PACKETS,
+	TOOL_PMU__EVENT_NET_TX_ERRORS,
+	TOOL_PMU__EVENT_NET_TX_DROP,
+	TOOL_PMU__EVENT_NET_TX_FIFO,
+	TOOL_PMU__EVENT_NET_TX_COLLS,
+	TOOL_PMU__EVENT_NET_TX_CARRIER,
+	TOOL_PMU__EVENT_NET_TX_COMPRESSED,
 
 	TOOL_PMU__EVENT_MAX,
 };
Signature · 1 line
2.52.0.351.gbe84eed79e-goog
author reviewer
#3 lore raw
In reply to Ian Rogers
Quoted 25 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:

```
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -p $(pidof -d, chrome) -I 1000
     1.001023475                  0      net_rx_bytes
     1.001023475                  0      net_rx_compressed
     1.001023475         42,647,328      net_rx_drop
     1.001023475        463,069,152      net_rx_errors
     1.001023475                  0      net_rx_fifo
     1.001023475                  0      net_rx_frame
     1.001023475                  0      net_rx_multicast
     1.001023475    423,195,831,744      net_rx_packets
     1.001023475                  0      net_tx_bytes
     1.001023475                  0      net_tx_carrier
     1.001023475                  0      net_tx_colls
     1.001023475                  0      net_tx_compressed
     1.001023475                  0      net_tx_drop
     1.001023475                  0      net_tx_errors
     1.001023475                  0      net_tx_fifo
     1.001023475                  0      net_tx_packets
```
Quoted 7 lines

As the events are in the tool_pmu they can be used in metrics. The json descriptions they are exposed in perf list and the events can be seen in the python ilist application.

Note, if a process terminates then the count reading returns an error and this can expose what appear to be latent bugs in the aggregation and display code.

I forgot to mention there are also other events that could be added. For example, from /proc/self/status the signal information could be useful as could voluntary vs involuntary context switch counts. What is here is likely useful enough that a bigger patch series isn't warranted.

Thanks,
Ian

Quoted 14 lines

Ian Rogers (2):
perf tool_pmu: Add memory events
perf tool_pmu: Add network events

 tools/perf/builtin-stat.c                     |  10 +-
 .../pmu-events/arch/common/common/tool.json   | 266 ++++++++-
 tools/perf/pmu-events/empty-pmu-events.c      | 312 +++++++----
 tools/perf/util/tool_pmu.c                    | 514 +++++++++++++++++-
 tools/perf/util/tool_pmu.h                    |  44 ++
 5 files changed, 1026 insertions(+), 120 deletions(-)

--
2.52.0.351.gbe84eed79e-goog

maintainer
#4 lore raw
In reply to Ian Rogers

Hi Ian,

Quoted 15 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:

```
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -p $(pidof -d, chrome) -I 1000
     1.001023475                  0      net_rx_bytes
     1.001023475                  0      net_rx_compressed
     1.001023475         42,647,328      net_rx_drop
     1.001023475        463,069,152      net_rx_errors
     1.001023475                  0      net_rx_fifo
     1.001023475                  0      net_rx_frame
     1.001023475                  0      net_rx_multicast
     1.001023475    423,195,831,744      net_rx_packets
     1.001023475                  0      net_tx_bytes
Quoted 8 lines
     1.001023475                  0      net_tx_carrier
     1.001023475                  0      net_tx_colls
     1.001023475                  0      net_tx_compressed
     1.001023475                  0      net_tx_drop
     1.001023475                  0      net_tx_errors
     1.001023475                  0      net_tx_fifo
     1.001023475                  0      net_tx_packets
```

Interesting.

Quoted 8 lines

As the events are in the tool_pmu they can be used in metrics. The json descriptions they are exposed in perf list and the events can be seen in the python ilist application.

Note, if a process terminates then the count reading returns an error and this can expose what appear to be latent bugs in the aggregation and display code.

How do you handle system-wide mode and sampling (perf record)?

Thanks,
Namhyung

Quoted 15 lines

Ian Rogers (2):
perf tool_pmu: Add memory events
perf tool_pmu: Add network events

 tools/perf/builtin-stat.c                     |  10 +-
 .../pmu-events/arch/common/common/tool.json   | 266 ++++++++-
 tools/perf/pmu-events/empty-pmu-events.c      | 312 +++++++----
 tools/perf/util/tool_pmu.c                    | 514 +++++++++++++++++-
 tools/perf/util/tool_pmu.h                    |  44 ++
 5 files changed, 1026 insertions(+), 120 deletions(-)
Signature · 1 line
2.52.0.351.gbe84eed79e-goog
author reviewer
#5 lore raw
Quoted 28 lines · Namhyung Kim

Hi Ian,

Quoted 15 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:

```
$ perf stat -e net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets -p $(pidof -d, chrome) -I 1000
     1.001023475                  0      net_rx_bytes
     1.001023475                  0      net_rx_compressed
     1.001023475         42,647,328      net_rx_drop
     1.001023475        463,069,152      net_rx_errors
     1.001023475                  0      net_rx_fifo
     1.001023475                  0      net_rx_frame
     1.001023475                  0      net_rx_multicast
     1.001023475    423,195,831,744      net_rx_packets
     1.001023475                  0      net_tx_bytes
Quoted 8 lines
     1.001023475                  0      net_tx_carrier
     1.001023475                  0      net_tx_colls
     1.001023475                  0      net_tx_compressed
     1.001023475                  0      net_tx_drop
     1.001023475                  0      net_tx_errors
     1.001023475                  0      net_tx_fifo
     1.001023475                  0      net_tx_packets
```
Quoted 1 line

Interesting.

Thanks.

Quoted 10 lines
Quoted 8 lines

As the events are in the tool_pmu they can be used in metrics. The json descriptions they are exposed in perf list and the events can be seen in the python ilist application.

Note, if a process terminates then the count reading returns an error and this can expose what appear to be latent bugs in the aggregation and display code.

How do you handle system-wide mode and sampling (perf record)?

So tool events don't support perf record and fail at opening due to the invalid PMU type and config. This is the same as if you did perf record -e duration_time with a perf today, which looks like:

```
$ perf record -e duration_time -a sleep 1
Error:
Failure to open event 'duration_time' on PMU 'tool' which will be removed.
No fallback found for 'duration_time' for error 0
Error:
Failure to open any events for recording.
```

For system-wide the behavior is hopefully intuitive in that the memory and network counts are for the whole system rather than the given processes. For the memory events the proc directory is scanned and all processes counts aggregated. For network data /proc/net/dev is read rather than /proc/pid/net/dev. There is more detail in the individual commit messages on this.

Thanks,
Ian

Quoted 18 lines

Thanks,
Namhyung

Quoted 15 lines

Ian Rogers (2):
perf tool_pmu: Add memory events
perf tool_pmu: Add network events

 tools/perf/builtin-stat.c                     |  10 +-
 .../pmu-events/arch/common/common/tool.json   | 266 ++++++++-
 tools/perf/pmu-events/empty-pmu-events.c      | 312 +++++++----
 tools/perf/util/tool_pmu.c                    | 514 +++++++++++++++++-
 tools/perf/util/tool_pmu.h                    |  44 ++
 5 files changed, 1026 insertions(+), 120 deletions(-)

--
2.52.0.351.gbe84eed79e-goog

#6 lore raw
In reply to Ian Rogers
Quoted 8 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:
```

$ perf stat -e
net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets
-p $(pidof -d, chrome) -I 1000

But AFAIK that's for the complete network name space, not just the process, thus highly misleading in perf context because the scope is incompatible.

-Andi

author reviewer
#7 lore raw
Quoted 12 lines · Andi Kleen
Quoted 8 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:
```

$ perf stat -e
net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets
-p $(pidof -d, chrome) -I 1000
Quoted 3 lines

But AFAIK that's for the complete network name space, not just the process, thus highly misleading in perf context because the scope is incompatible.

Yeah, we can point this out in the event descriptions or just not have the events and try to do some per process BPF type thing. Given we don't have the BPF thing it is still tempting to have these counters as-is for the system-wide case.

Thanks,
Ian

maintainer
#8 lore raw
Quoted 13 lines · Ian Rogers
Quoted 12 lines · Andi Kleen
Quoted 8 lines · Ian Rogers

Add events for memory use and network activity based on data readily available in /prod/pid/statm, /proc/pid/smaps_rollup and /proc/pid/net/dev. For example the network usage of chrome processes on a system may be gathered with:
```

$ perf stat -e
net_rx_bytes,net_rx_compressed,net_rx_drop,net_rx_errors,net_rx_fifo,net_rx_frame,net_rx_multicast,net_rx_packets,net_tx_bytes,net_tx_carrier,net_tx_colls,net_tx_compressed,net_tx_drop,net_tx_errors,net_tx_fifo,net_tx_packets
-p $(pidof -d, chrome) -I 1000
Quoted 8 lines
Quoted 3 lines

But AFAIK that's for the complete network name space, not just the process, thus highly misleading in perf context because the scope is incompatible.

Yeah, we can point this out in the event descriptions or just not have the events and try to do some per process BPF type thing. Given we don't have the BPF thing it is still tempting to have these counters as-is for the system-wide case.

You may want to make it fail to open for per-process mode.

Thanks,
Namhyung