This is a minimal patch series to enable '--bpf-counters' options for perf tool in 5.10 kernel version. How to compile? =============== 1. Download openEuler kernel source code: git clone https://gitee.com/openeuler/kernel.git 2. Download these patches, apply them in kernel/ directory: cd kernel/ git am --reject /path/to/*.patch 3. Prepare some requirements before compiling perf tool: Not list one by one here, please refer to [1]. [1] https://medium.com/@manas.marwah/building-perf-tool-fc838f084f71 4. Compile perf tool: cd tools/perf make BUILD_BPF_SKEL=1 If successful, you will see a binary file named 'perf' under tools/perf/ directory. How to use? =========== 1. Use bperf to sample a process: $ perf stat --bpf-counters -e cycles,instructions \ perf bench sched messaging -g 1 -l 100 -t # Running 'sched/messaging' benchmark: # 20 sender and receiver threads per group # 1 groups == 40 threads run Total time: 0.011 [sec] Performance counter stats for 'perf bench sched messaging -g 1 -l 100 -t': 639,816,606 cycles 284,102,919 instructions # 0.44 insn per cycle 0.018093452 seconds time elapsed 0.023612000 seconds user 0.255643000 seconds sys 2. Use bperf to sample a container: $ perf stat --bpf-counters -e cycles,instructions --timeout 1000 \ --for-each-cgroup docker/66166578dd7d17655923862792a0b9d60c8785fdc1bb8da8b76daedb205c4b10 Performance counter stats for 'system wide': 7,402,983 cycles docker/66166578dd7d17655923862792a0b9d60c8785fdc1bb8da8b76daedb205c4b10 6,036,926 instructions docker/66166578dd7d17655923862792a0b9d60c8785fdc1bb8da8b76daedb205c4b10 # 0.82 insn per cycle 1.004999462 seconds time elapsed How to integrate via bpf? ======================== If you are interested in process sampling, please check the following bpf progs: tools/perf/util/bpf_skel/bperf_leader.bpf.c tools/perf/util/bpf_skel/bperf_follower.bpf.c And if you are interested in container sampling, please check the following bpf prog: tools/perf/util/bpf_skel/bperf_cgroup.bpf.c There have some documents in tools/perf/util/bpf_counter.c, search the keywords 'bperf: share hardware PMCs with BPF' in it and see how bperf works. That's all. Enjoy. Namhyung Kim (9): bpf: Allow bpf_get_current_ancestor_cgroup_id for tracing perf core: Factor out __perf_sw_event_sched perf core: Add PERF_COUNT_SW_CGROUP_SWITCHES event perf tools: Add 'cgroup-switches' software event perf tools: Add read_cgroup_id() function perf tools: Add cgroup_is_v2() helper perf bpf_counter: Move common functions to bpf_counter.h perf stat: Enable BPF counter with --for-each-cgroup perf stat: Fix BPF program section name Song Liu (12): bpftool: Add Makefile target bootstrap perf build: Support build BPF skeletons with perf perf stat: Enable counting events for BPF programs perf stat: Introduce 'bperf' to share hardware PMCs with BPF perf stat: Measure 't0' and 'ref_time' after enable_counters() perf util: Move bpf_perf definitions to a libperf header perf bpf: check perf_attr_map is compatible with the perf binary perf stat: Introduce config stat.bpf-counter-events perf stat: Introduce ':b' modifier perf stat: Introduce bpf_counter_ops->disable() perf bpf: Fix building perf with BUILD_BPF_SKEL=1 by default in more distros perf bpf_skel: Do not use typedef to avoid error on old clang Tengda Wu (1): perf stat: Support inherit events during fork() for bperf Xiaomeng Zhang (1): perf stat: Increase perf_attr_map entries include/linux/perf_event.h | 40 +- include/uapi/linux/perf_event.h | 1 + kernel/trace/bpf_trace.c | 2 + tools/bpf/bpftool/Makefile | 2 + tools/build/Makefile.feature | 4 +- tools/include/uapi/linux/perf_event.h | 7 + tools/lib/perf/include/perf/bpf_perf.h | 31 + tools/perf/Documentation/perf-stat.txt | 31 + tools/perf/Makefile.config | 9 + tools/perf/Makefile.perf | 65 +- tools/perf/builtin-stat.c | 111 ++- tools/perf/util/Build | 2 + tools/perf/util/bpf_counter.c | 828 ++++++++++++++++++ tools/perf/util/bpf_counter.h | 131 +++ tools/perf/util/bpf_counter_cgroup.c | 307 +++++++ tools/perf/util/bpf_skel/.gitignore | 3 + tools/perf/util/bpf_skel/bperf_cgroup.bpf.c | 191 ++++ tools/perf/util/bpf_skel/bperf_follower.bpf.c | 162 ++++ tools/perf/util/bpf_skel/bperf_leader.bpf.c | 55 ++ tools/perf/util/bpf_skel/bperf_u.h | 19 + .../util/bpf_skel/bpf_prog_profiler.bpf.c | 93 ++ tools/perf/util/cgroup.c | 47 + tools/perf/util/cgroup.h | 13 + tools/perf/util/config.c | 4 + tools/perf/util/evlist.c | 4 + tools/perf/util/evsel.c | 27 + tools/perf/util/evsel.h | 32 + tools/perf/util/parse-events.c | 12 +- tools/perf/util/parse-events.l | 3 +- tools/perf/util/python.c | 27 + tools/perf/util/stat-display.c | 4 +- tools/perf/util/stat.c | 2 +- tools/perf/util/target.c | 34 +- tools/perf/util/target.h | 8 + tools/scripts/Makefile.include | 1 + 35 files changed, 2266 insertions(+), 46 deletions(-) create mode 100644 tools/lib/perf/include/perf/bpf_perf.h create mode 100644 tools/perf/util/bpf_counter.c create mode 100644 tools/perf/util/bpf_counter.h create mode 100644 tools/perf/util/bpf_counter_cgroup.c create mode 100644 tools/perf/util/bpf_skel/.gitignore create mode 100644 tools/perf/util/bpf_skel/bperf_cgroup.bpf.c create mode 100644 tools/perf/util/bpf_skel/bperf_follower.bpf.c create mode 100644 tools/perf/util/bpf_skel/bperf_leader.bpf.c create mode 100644 tools/perf/util/bpf_skel/bperf_u.h create mode 100644 tools/perf/util/bpf_skel/bpf_prog_profiler.bpf.c -- 2.34.1