On Fri, 10 Jul 2020 16:49:17 +0000
Jakub Sitnicki <jakub@xxxxxxxxxxxxxx> wrote:
> BPF_PROG_TYPE_CGROUP_SOCKOPT requires expected_attach_type to be set on
> prog load. Set it in the verifier test that checks if calling
> bpf_perf_event_output() from CGROUP_SOCKOPT is allowed so that the runner
> does skip it.
>
> Cc: Stanislav Fomichev <sdf@xxxxxxxxxx>
> Fixes: 0456ea170cd6 ("bpf: Enable more helpers for BPF_PROG_TYPE_CGROUP_{DEVICE,SYSCTL,SOCKOPT}")
> Signed-off-by: Jakub Sitnicki <jakub@xxxxxxxxxxxxxx>
> ---
> tools/testing/selftests/bpf/verifier/event_output.c | 1 +
> 1 file changed, 1 insertion(+)
>
> diff --git a/tools/testing/selftests/bpf/verifier/event_output.c b/tools/testing/selftests/bpf/verifier/event_output.c
> index 99f8f582c02b..c5e805980409 100644
> --- a/tools/testing/selftests/bpf/verifier/event_output.c
> +++ b/tools/testing/selftests/bpf/verifier/event_output.c
> @@ -112,6 +112,7 @@
> "perfevent for cgroup sockopt",
> .insns = { __PERF_EVENT_INSNS__ },
> .prog_type = BPF_PROG_TYPE_CGROUP_SOCKOPT,
> + .expected_attach_type = BPF_CGROUP_SETSOCKOPT,
> .fixup_map_event_output = { 4 },
> .result = ACCEPT,
> .retval = 1,
Just noticed same fix has been already posted a couple hours earlier:
https://lore.kernel.org/bpf/20200710150439.126627-1-jean-philippe@xxxxxxxxxx/
Please ignore.
