Re: [PATCH v3] perf/trace : Fix repetitious traces of perf on tracepoint

[chengjian (D)]

Hi, Milian


On 2018/1/16 22:33, Milian Wolff wrote:
> > perf script print the same wakeup_new event multiple times.
> >
> > These events which trigger this issue all specify a target process.
> > commit e6dab5ffab59 ("perf/trace: Add ability to set a target task
> > for events") has designed a method to trace these events. For
> > example, the sched_wakeup and sched_wakeup_new tracepoint will be
> > caught when the current task wakeup a target task.
the sched_stat_* tracepoint also specify a target process, so it will be 
reported
nrcpus times too.

for example sched_stat_sleep

         swapper     0 [002]   188.752870: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1051 delay=2133649486 [ns]
| |
| |
current task



> Oh, this sounds awesome. I don't have the setup available to compile a kernel
> with this patch applied, but I think from the description it solves a long-
> standing issue with perf's sleep-time profiling.
>
> Can someone try this please:
> https://perf.wiki.kernel.org/index.php/Tutorial#Profiling_sleep_times
when current != task.

    #echo 1 > /proc/sys/kernel/sched_schedstats
    #./perf-bin/perf record -e sched:sched_stat_sleep ./test_fork_loop

before this patch:
           :1050  1050 [000]   186.597339: sched:sched_stat_sleep: 
comm=perf pid=1051 delay=22955314 [ns]
          :1050  1050 [000]   186.597397: sched:sched_stat_sleep: 
comm=perf pid=1051 delay=22955314 [ns]
           :1050  1050 [000]   186.597406: sched:sched_stat_sleep: 
comm=perf pid=1051 delay=22955314 [ns]
           :1050  1050 [000]   186.597415: sched:sched_stat_sleep: 
comm=perf pid=1051 delay=22955314 [ns]
         swapper     0 [002]   188.752870: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1051 delay=2133649486 [ns]
         swapper     0 [002]   188.752899: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1051 delay=2133649486 [ns]
         swapper     0 [002]   188.752952: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1051 delay=2133649486 [ns]
         swapper     0 [002]   188.752965: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1051 delay=2133649486 [ns]

after this patch:
           :1041  1041 [001]    62.447062: sched:sched_stat_sleep: 
comm=perf pid=1042 delay=43035826 [ns]
         swapper     0 [001]    62.704644: sched:sched_stat_sleep: 
comm=bug_fork_loop pid=1042 delay=234613068 [ns]


> I believe your patch also fixes the sched_stat_* tracepoints to be only
> emitted once per CPU. Can you verify this? I.e. is the period finally
> correctly calculated and we get a value of roughly 1E9ns == 1s?
 Yeah, this patch fix it.

Thanks.
    CHENG Jian