On Wed, Oct 6, 2021 at 11:56 AM Yucong Sun <fallentree@xxxxxx> wrote:
>
> When running tests concurrently, perf rate limiting will often cause
> some events to be skipped and make some tests flaky, disabling it making
> running tests more robust.
>
> Signed-off-by: Yucong Sun <sunyucong@xxxxxxxxx>
> ---
> tools/testing/selftests/bpf/test_progs.c | 2 ++
> 1 file changed, 2 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/test_progs.c b/tools/testing/selftests/bpf/test_progs.c
> index 4e2028189471..2ac922f8aa2c 100644
> --- a/tools/testing/selftests/bpf/test_progs.c
> +++ b/tools/testing/selftests/bpf/test_progs.c
> @@ -1376,6 +1376,8 @@ int main(int argc, char **argv)
> }
> }
>
> + system("echo 0 > /proc/sys/kernel/perf_cpu_time_max_percent");
I don't think we want to do this outside of QEMU, changing the global
state of the real system. I think this might have also been relevant
to me waiting for perf_branches test for about 20 seconds or so. I've
dropped this patch for now.
> +
> /* ignore workers if we are just listing */
> if (env.get_test_cnt || env.list_test_names)
> env.workers = 0;
> --
> 2.30.2
>
