On Thu, Jan 9, 2020 at 5:16 PM Yonghong Song <yhs@xxxxxx> wrote:
>
> The test_progs send_signal() is amended to test
> bpf_send_signal_thread() as well.
>
> $ ./test_progs -n 40
> #40/1 send_signal_tracepoint:OK
> #40/2 send_signal_perf:OK
> #40/3 send_signal_nmi:OK
> #40/4 send_signal_tracepoint_thread:OK
> #40/5 send_signal_perf_thread:OK
> #40/6 send_signal_nmi_thread:OK
> #40 send_signal:OK
> Summary: 1/6 PASSED, 0 SKIPPED, 0 FAILED
>
> Signed-off-by: Yonghong Song <yhs@xxxxxx>
> ---
> tools/include/uapi/linux/bpf.h | 18 +++++++++--
maybe do tools/uapi header sync in a first patch, along the original change?
> .../selftests/bpf/prog_tests/send_signal.c | 30 ++++++++++++-------
> .../bpf/progs/test_send_signal_kern.c | 9 ++++--
> 3 files changed, 42 insertions(+), 15 deletions(-)
>
[...]
> diff --git a/tools/testing/selftests/bpf/progs/test_send_signal_kern.c b/tools/testing/selftests/bpf/progs/test_send_signal_kern.c
> index 0e6be01157e6..4a722024c32b 100644
> --- a/tools/testing/selftests/bpf/progs/test_send_signal_kern.c
> +++ b/tools/testing/selftests/bpf/progs/test_send_signal_kern.c
> @@ -23,6 +23,7 @@ int bpf_send_signal_test(void *ctx)
> {
> __u64 *info_val, *status_val;
> __u32 key = 0, pid, sig;
> + int use_signal_thread;
> int ret;
>
> status_val = bpf_map_lookup_elem(&status_map, &key);
> @@ -33,11 +34,15 @@ int bpf_send_signal_test(void *ctx)
> if (!info_val || *info_val == 0)
> return 0;
>
> - sig = *info_val >> 32;
> + use_signal_thread = *info_val >> 48;
> + sig = *info_val >> 32 & 0xFFFF;
> pid = *info_val & 0xffffFFFF;
Would you mind rewriting this test w/ BPF skeleton and global data? It
would make it cleaner without all this masking stuff?
>
> if ((bpf_get_current_pid_tgid() >> 32) == pid) {
> - ret = bpf_send_signal(sig);
> + if (use_signal_thread)
> + ret = bpf_send_signal_thread(sig);
> + else
> + ret = bpf_send_signal(sig);
> if (ret == 0)
> *status_val = 1;
> }
> --
> 2.17.1
>
