Received-SPF: pass (google.com: domain of linux-kernel+bounces-99352-linux.lists.archive=gmail.com@vger.kernel.org designates 139.178.88.99 as permitted sender) client-ip=139.178.88.99;
Precedence: bulk
MIME-Version: 1.0
References: <20240309132710.1055941-1-retpolanne@posteo.net>
In-Reply-To: <20240309132710.1055941-1-retpolanne@posteo.net>
From: Namhyung Kim <namhyung@kernel.org>
Date: Mon, 11 Mar 2024 11:17:01 -0700
Message-ID: <CAM9d7cjhiua5rBj=CTDJJC-XJN6PzKxQ5DsooJGEz0QcQAry7w@mail.gmail.com>
Subject: Re: [PATCH v1] perf lock contention: skip traceiter functions
To: Anne Macedo <retpolanne@posteo.net>
Cc: Arnaldo Carvalho de Melo <acme@kernel.org>, Peter Zijlstra <peterz@infradead.org>, 
	Ingo Molnar <mingo@redhat.com>, Mark Rutland <mark.rutland@arm.com>, 
	Alexander Shishkin <alexander.shishkin@linux.intel.com>, Jiri Olsa <jolsa@kernel.org>, 
	Ian Rogers <irogers@google.com>, Adrian Hunter <adrian.hunter@intel.com>, 
	linux-perf-users@vger.kernel.org, linux-kernel@vger.kernel.org
Content-Type: text/plain; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

Hello Anne,

On Sat, Mar 9, 2024 at 5:27=E2=80=AFAM Anne Macedo <retpolanne@posteo.net> =
wrote:
>
> The perf lock contention program currently shows the caller of the locks
> as __traceiter_contention_begin+0x??. This caller can be ignored, as it i=
s
> from the traceiter itself. Instead, it should show the real callers for
> the locks.
>
> When fiddling with the --stack-skip parameter, the actual callers for
> the locks start to show up. However, just ignore the
> __traceiter_contention_begin and the __traceiter_contention_end symbols
> so the actual callers will show up.
>
> Before this patch is applied:
>
> sudo perf lock con -a -b -- sleep 3
>  contended   total wait     max wait     avg wait         type   caller
>
>          8      2.33 s       2.28 s     291.18 ms     rwlock:W   __tracei=
ter_contention_begin+0x44
>          4      2.33 s       2.28 s     582.35 ms     rwlock:W   __tracei=
ter_contention_begin+0x44
>          7    140.30 ms     46.77 ms     20.04 ms     rwlock:W   __tracei=
ter_contention_begin+0x44
>          2     63.35 ms     33.76 ms     31.68 ms        mutex   trace_co=
ntention_begin+0x84
>          2     46.74 ms     46.73 ms     23.37 ms     rwlock:W   __tracei=
ter_contention_begin+0x44
>          1     13.54 us     13.54 us     13.54 us        mutex   trace_co=
ntention_begin+0x84
>          1      3.67 us      3.67 us      3.67 us      rwsem:R   __tracei=
ter_contention_begin+0x44
>
> Before this patch is applied - using --stack-skip 5
>
> sudo perf lock con --stack-skip 5 -a -b -- sleep 3
>  contended   total wait     max wait     avg wait         type   caller
>
>          2      2.24 s       2.24 s       1.12 s      rwlock:W   do_epoll=
_wait+0x5a0
>          4      1.65 s     824.21 ms    412.08 ms     rwlock:W   do_exit+=
0x338
>          2    824.35 ms    824.29 ms    412.17 ms     spinlock   get_sign=
al+0x108
>          2    824.14 ms    824.14 ms    412.07 ms     rwlock:W   release_=
task+0x68
>          1     25.22 ms     25.22 ms     25.22 ms        mutex   cgroup_k=
n_lock_live+0x58
>          1     24.71 us     24.71 us     24.71 us     spinlock   do_exit+=
0x44
>          1     22.04 us     22.04 us     22.04 us      rwsem:R   lock_mm_=
and_find_vma+0xb0
>
> After this patch is applied:
>
> sudo ./perf lock con -a -b -- sleep 3
>  contended   total wait     max wait     avg wait         type   caller
>
>          4      4.13 s       2.07 s       1.03 s      rwlock:W   release_=
task+0x68
>          2      2.07 s       2.07 s       1.03 s      rwlock:R   mm_updat=
e_next_owner+0x50
>          2      2.07 s       2.07 s       1.03 s      rwlock:W   do_exit+=
0x338
>          1     41.56 ms     41.56 ms     41.56 ms        mutex   cgroup_k=
n_lock_live+0x58
>          2     36.12 us     18.83 us     18.06 us     rwlock:W   do_exit+=
0x338
>
> changes since v0:
>
> - skip trace_contention functions
> - use sym->end instead of __traceiter_contention_end for text_end
>
> Signed-off-by: Anne Macedo <retpolanne@posteo.net>
> ---
>  tools/perf/util/machine.c | 17 +++++++++++++++++
>  tools/perf/util/machine.h |  2 +-
>  2 files changed, 18 insertions(+), 1 deletion(-)
>
> diff --git a/tools/perf/util/machine.c b/tools/perf/util/machine.c
> index 527517db3182..db443947ccd1 100644
> --- a/tools/perf/util/machine.c
> +++ b/tools/perf/util/machine.c
> @@ -3266,6 +3266,14 @@ bool machine__is_lock_function(struct machine *mac=
hine, u64 addr)
>
>                 sym =3D machine__find_kernel_symbol_by_name(machine, "__l=
ock_text_end", &kmap);
>                 machine->lock.text_end =3D map__unmap_ip(kmap, sym->start=
);
> +
> +               sym =3D machine__find_kernel_symbol_by_name(machine, "__t=
raceiter_contention_begin", &kmap);

Unlike __sched_text_{start,end} and __lock_text_{start,end} I guess
this traceiter thing is optional so it might not be there.  You need to
handle if it's NULL.

I think it depends on the kernel version and configuration.  I remember
I saw a different symbol on old kernels.  But it'd be hard to handle all
the cases.  Let's have a single trace text section in the struct machine
and use __traceiter_contention_begin only.  If it's not found you can
fallback to trace_contention_begin.

Thanks,
Namhyung


> +               machine->traceiter.text_start =3D map__unmap_ip(kmap, sym=
->start);
> +               machine->traceiter.text_end =3D map__unmap_ip(kmap, sym->=
end);
> +
> +               sym =3D machine__find_kernel_symbol_by_name(machine, "tra=
ce_contention_begin", &kmap);
> +               machine->trace.text_start =3D map__unmap_ip(kmap, sym->st=
art);
> +               machine->trace.text_end =3D map__unmap_ip(kmap, sym->end)=
;
>         }
>
>         /* failed to get kernel symbols */
> @@ -3280,5 +3288,14 @@ bool machine__is_lock_function(struct machine *mac=
hine, u64 addr)
>         if (machine->lock.text_start <=3D addr && addr < machine->lock.te=
xt_end)
>                 return true;
>
> +       /* traceiter functions currently don't have their own section
> +        * but we consider them lock functions
> +        */
> +       if (machine->traceiter.text_start <=3D addr && addr < machine->tr=
aceiter.text_end)
> +               return true;
> +
> +       if (machine->trace.text_start <=3D addr && addr < machine->trace.=
text_end)
> +               return true;
> +
>         return false;
>  }
> diff --git a/tools/perf/util/machine.h b/tools/perf/util/machine.h
> index e28c787616fe..4312f6db6de0 100644
> --- a/tools/perf/util/machine.h
> +++ b/tools/perf/util/machine.h
> @@ -49,7 +49,7 @@ struct machine {
>         struct {
>                 u64       text_start;
>                 u64       text_end;
> -       } sched, lock;
> +       } sched, lock, traceiter, trace;
>         pid_t             *current_tid;
>         size_t            current_tid_sz;
>         union { /* Tool specific area */
> --
> 2.39.2
>