> On Mar 8, 2022, at 12:19 PM, sunyucong@xxxxxxxxx wrote: > > On Wed, Mar 2, 2022 at 3:53 PM Mykola Lysenko <mykolal@xxxxxx> wrote: >> >> In send_signal, replace sleep with dummy cpu intensive computation >> to increase probability of child process being scheduled. Add few >> more asserts. >> >> In find_vma, reduce sample_freq as higher values may be rejected in >> some qemu setups, remove usleep and increase length of cpu intensive >> computation. >> >> In bpf_cookie, perf_link and perf_branches, reduce sample_freq as >> higher values may be rejected in some qemu setups >> >> Signed-off-by: Mykola Lysenko <mykolal@xxxxxx> >> Acked-by: Yonghong Song <yhs@xxxxxx> >> --- >> .../selftests/bpf/prog_tests/bpf_cookie.c | 2 +- >> .../testing/selftests/bpf/prog_tests/find_vma.c | 13 ++++++++++--- >> .../selftests/bpf/prog_tests/perf_branches.c | 4 ++-- >> .../selftests/bpf/prog_tests/perf_link.c | 2 +- >> .../selftests/bpf/prog_tests/send_signal.c | 17 ++++++++++------- >> .../selftests/bpf/progs/test_send_signal_kern.c | 2 +- >> 6 files changed, 25 insertions(+), 15 deletions(-) >> >> diff --git a/tools/testing/selftests/bpf/prog_tests/bpf_cookie.c b/tools/testing/selftests/bpf/prog_tests/bpf_cookie.c >> index cd10df6cd0fc..0612e79a9281 100644 >> --- a/tools/testing/selftests/bpf/prog_tests/bpf_cookie.c >> +++ b/tools/testing/selftests/bpf/prog_tests/bpf_cookie.c >> @@ -199,7 +199,7 @@ static void pe_subtest(struct test_bpf_cookie *skel) >> attr.type = PERF_TYPE_SOFTWARE; >> attr.config = PERF_COUNT_SW_CPU_CLOCK; >> attr.freq = 1; >> - attr.sample_freq = 4000; >> + attr.sample_freq = 1000; >> pfd = syscall(__NR_perf_event_open, &attr, -1, 0, -1, PERF_FLAG_FD_CLOEXEC); >> if (!ASSERT_GE(pfd, 0, "perf_fd")) >> goto cleanup; >> diff --git a/tools/testing/selftests/bpf/prog_tests/find_vma.c b/tools/testing/selftests/bpf/prog_tests/find_vma.c >> index b74b3c0c555a..7cf4feb6464c 100644 >> --- a/tools/testing/selftests/bpf/prog_tests/find_vma.c >> +++ b/tools/testing/selftests/bpf/prog_tests/find_vma.c >> @@ -30,12 +30,20 @@ static int open_pe(void) >> attr.type = PERF_TYPE_HARDWARE; >> attr.config = PERF_COUNT_HW_CPU_CYCLES; >> attr.freq = 1; >> - attr.sample_freq = 4000; >> + attr.sample_freq = 1000; > > I think It's actually better to modify sysctl. > perf_event_max_sample_rate through test_progs, I had a patch do that > before, but Andrii didn't apply it. (I've forgotten why) , but this is > a recurring issue when running in VM in parallel mode. I thought about this. But why to do it if sample_freq = 1000 is enough for our tests to function? > >> pfd = syscall(__NR_perf_event_open, &attr, 0, -1, -1, PERF_FLAG_FD_CLOEXEC); >> >> return pfd >= 0 ? pfd : -errno; >> }
