On 8/28/2023 3:55 PM, Jiri Olsa wrote:
> Adding support to gather stats for kprobe_multi programs.
>
> We now count:
> - missed stats due to bpf_prog_active protection (always)
> - cnt/nsec of the bpf program execution (if kernel.bpf_stats_enabled=1)
>
> Signed-off-by: Jiri Olsa <jolsa@xxxxxxxxxx>
Acked-by: Hou Tao <houtao1@xxxxxxxxxx>
With one nit below.
> ---
> kernel/trace/bpf_trace.c | 8 +++++++-
> 1 file changed, 7 insertions(+), 1 deletion(-)
>
> diff --git a/kernel/trace/bpf_trace.c b/kernel/trace/bpf_trace.c
> index a7264b2c17ad..0a8685fc1eee 100644
> --- a/kernel/trace/bpf_trace.c
> +++ b/kernel/trace/bpf_trace.c
> @@ -2706,18 +2706,24 @@ kprobe_multi_link_prog_run(struct bpf_kprobe_multi_link *link,
> .link = link,
> .entry_ip = entry_ip,
> };
> + struct bpf_prog *prog = link->link.prog;
> struct bpf_run_ctx *old_run_ctx;
> + u64 start;
> int err;
>
> if (unlikely(__this_cpu_inc_return(bpf_prog_active) != 1)) {
> + bpf_prog_inc_misses_counter(prog);
> err = 0;
> goto out;
> }
>
> +
The extra empty line is not needed here.
