perf lock contention: Show detail failure reason for BPF
authorNamhyung Kim <namhyung@kernel.org>
Mon, 27 Mar 2023 22:57:11 +0000 (15:57 -0700)
committerArnaldo Carvalho de Melo <acme@redhat.com>
Tue, 4 Apr 2023 12:39:56 +0000 (09:39 -0300)
commit84c3a2bb4c513ca9fd1326cf24d3729b769d9cf1
treef3e40c0eae98014a57f29d238b642b1abbf94f40
parent35bf007e2ee0447166081646c0797bfa831b062f
perf lock contention: Show detail failure reason for BPF

It can fail to collect lock stat from BPF for various reasons.  For
example, I've got a report that sometimes time calculation seems wrong
in case of contended spinlocks.  I suspect the time delta went negative
for some reason.

Count them separately and show in the output like below:

$ sudo perf lock contention -abE5 sleep 10
 contended   total wait     max wait     avg wait         type   caller

        13    785.61 us     79.36 us     60.43 us     spinlock   remove_wait_queue+0x14
        10    469.02 us     87.51 us     46.90 us     spinlock   prepare_to_wait+0x27
         9    289.09 us     69.08 us     32.12 us     spinlock   finish_wait+0x36
       114    251.05 us      8.56 us      2.20 us     spinlock   try_to_wake_up+0x1f5
       132    188.63 us      5.01 us      1.43 us     spinlock   __wake_up_common_lock+0x62

=== output for debug ===

bad: 1, total: 279
bad rate: 0.36 %
histogram of failure reasons
       task: 1
      stack: 0
       time: 0

Signed-off-by: Namhyung Kim <namhyung@kernel.org>
Cc: Adrian Hunter <adrian.hunter@intel.com>
Cc: Hao Luo <haoluo@google.com>
Cc: Ian Rogers <irogers@google.com>
Cc: Ingo Molnar <mingo@kernel.org>
Cc: Jiri Olsa <jolsa@kernel.org>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: bpf@vger.kernel.org
Link: https://lore.kernel.org/r/20230327225711.245738-1-namhyung@kernel.org
Signed-off-by: Arnaldo Carvalho de Melo <acme@redhat.com>
tools/perf/builtin-lock.c
tools/perf/util/bpf_lock_contention.c
tools/perf/util/bpf_skel/lock_contention.bpf.c
tools/perf/util/lock-contention.h