[PATCHSET v5 0/5] perf stat: Expand events for each cgroup

[Namhyung Kim]

Hello,

When we profile cgroup events with perf stat, it's very annoying to
specify events and cgroups on the command line as it requires the
mapping between events and cgroups.  (Note that perf record can use
cgroup sampling but it's not usable for perf stat).

I guess most cases we just want to use a same set of events (N) for
all cgroups (M), but we need to specify NxM events and NxM cgroups.
This is not good especially when profiling large number of cgroups:
say M=200.

So I added --for-each-cgroup option to make it easy for that case.  It
will create NxM events from N events and M cgroups.  One more upside
is that it can handle metrics too.

For example, the following example measures IPC metric for 3 cgroups

  $ cat perf-expand-cgrp.sh
  #!/bin/sh
  
  METRIC=${1:-IPC}
  CGROUP_DIR=/sys/fs/cgroup/perf_event
  
  sudo mkdir $CGROUP_DIR/A $CGROUP_DIR/B $CGROUP_DIR/C
  
  # add backgroupd workload for each cgroup
  echo $$ | sudo tee $CGROUP_DIR/A/cgroup.procs > /dev/null
  yes > /dev/null &
  echo $$ | sudo tee $CGROUP_DIR/B/cgroup.procs > /dev/null
  yes > /dev/null &
  echo $$ | sudo tee $CGROUP_DIR/C/cgroup.procs > /dev/null
  yes > /dev/null &

  # run 'perf stat' in the root cgroup
  echo $$ | sudo tee $CGROUP_DIR/cgroup.procs > /dev/null
  perf stat -a -M $METRIC --for-each-cgroup A,B,C sleep 1
  
  kill %1 %2 %3
  sudo rmdir $CGROUP_DIR/A $CGROUP_DIR/B $CGROUP_DIR/C

  
  $ ./perf-expand-cgrp.sh IPC
  
   Performance counter stats for 'system wide':
  
      11,284,850,010      inst_retired.any          A #     2.71 IPC                    
       4,157,915,982      cpu_clk_unhalted.thread   A                                   
      11,342,188,640      inst_retired.any          B #     2.72 IPC                    
       4,173,014,732      cpu_clk_unhalted.thread   B                                   
      11,135,863,604      inst_retired.any          C #     2.67 IPC                    
       4,171,375,184      cpu_clk_unhalted.thread   C                                   
  
         1.011948803 seconds time elapsed


* Changes from v4:
 - fix python import error  (Kernel test robot)
 - make metric-related function calls conditional  (Jiri)

* Changes from v3:
 - rename to evlist__find_evsel  (Jiri)
 - add documentation  (Arnaldo)
 - check -G option together with --for-each-cgroup

* Changes from v2:
 - put relevant fields in evsel together  (Jiri)
 - add various error checks  (Jiri)
 - split cgroup open patch  (Jiri)

* Changes from v1:
 - rename the option to --for-each-cgroup  (Jiri)
 - copy evsel fields explicitly  (Jiri)
 - add libpfm4 test  (Ian)


The code is available at 'perf/cgroup-multiply-v5' branch on

  git://git.kernel.org/pub/scm/linux/kernel/git/namhyung/linux-perf.git

Thanks
Namhyung


Namhyung Kim (5):
  perf evsel: Add evsel__clone() function
  perf stat: Add --for-each-cgroup option
  perf tools: Copy metric events properly when expand cgroups
  perf tools: Allow creation of cgroup without open
  perf test: Add expand cgroup event test

 tools/perf/Documentation/perf-stat.txt |   5 +
 tools/perf/builtin-stat.c              |  28 ++-
 tools/perf/tests/Build                 |   1 +
 tools/perf/tests/builtin-test.c        |   4 +
 tools/perf/tests/expand-cgroup.c       | 241 +++++++++++++++++++++++++
 tools/perf/tests/tests.h               |   1 +
 tools/perf/util/cgroup.c               | 115 +++++++++++-
 tools/perf/util/cgroup.h               |   3 +
 tools/perf/util/evlist.c               |  11 ++
 tools/perf/util/evlist.h               |   1 +
 tools/perf/util/evsel.c                | 104 +++++++++++
 tools/perf/util/evsel.h                |  93 ++++++----
 tools/perf/util/metricgroup.c          |  85 +++++++++
 tools/perf/util/metricgroup.h          |   6 +
 tools/perf/util/python.c               |  19 ++
 tools/perf/util/stat.h                 |   1 +
 16 files changed, 673 insertions(+), 45 deletions(-)
 create mode 100644 tools/perf/tests/expand-cgroup.c

-- 
2.28.0.681.g6f77f65b4e-goog