On Thu, Sep 2, 2021 at 10:23 AM Dave Marchevsky <davemarchevsky@xxxxxx> wrote:
>
> This corner case isn't covered by existing selftests' use of bpf_printk.
>
> Just test compilation, not output, as trace_vprintk already tests
> trace_pipe output.
>
> Signed-off-by: Dave Marchevsky <davemarchevsky@xxxxxx>
> ---
> tools/testing/selftests/bpf/progs/trace_vprintk.c | 7 +++++++
> 1 file changed, 7 insertions(+)
>
> diff --git a/tools/testing/selftests/bpf/progs/trace_vprintk.c b/tools/testing/selftests/bpf/progs/trace_vprintk.c
> index 255e2f018efe..33455e48a9ab 100644
> --- a/tools/testing/selftests/bpf/progs/trace_vprintk.c
> +++ b/tools/testing/selftests/bpf/progs/trace_vprintk.c
> @@ -23,3 +23,10 @@ int sys_enter(void *ctx)
> one, 2, three, 4, five, 6, seven, 8, nine, 10, ++trace_vprintk_ran);
> return 0;
> }
> +
> +SEC("fentry/__x64_sys_nanosleep")
> +int zero_fmt_args(void *ctx)
> +{
> + bpf_printk("\t"); // runner doesn't search for this, just ensure it compiles
C++ comments :( please use /* */
I'd probably just add this bpf_printk() in the same BPF program above
and roll it into previous patch. Doesn't seem like we need dedicated
BPF program just for this.
> + return 0;
> +}
> --
> 2.30.2
>
