On Thu, Jan 27, 2022 at 12:27 AM Dmitrii Dolgov <9erthalion6@xxxxxxxxx> wrote:
>
> Commit 82e6b1eee6a8 ("bpf: Allow to specify user-provided bpf_cookie for
> BPF perf links") introduced the concept of user specified bpf_cookie,
> which could be accessed by BPF programs using bpf_get_attach_cookie().
> For troubleshooting purposes it is convenient to expose bpf_cookie via
> bpftool as well, so there is no need to meddle with the target BPF
> program itself.
>
> $ bpftool perf
> pid 83 fd 9: prog_id 5 bpf_cookie: 123 tracepoint sched_process_exec
>
I think a more natural place to expose this would be in `bpftool link
show` output, as bpf_cookie is actually per attachment (i.e., link)
information (not a per-program). We'll need to anticipate multi-attach
use cases (e.g., multi-attach kprobe and fentry programs we are
discussing at the moment).
> Signed-off-by: Dmitrii Dolgov <9erthalion6@xxxxxxxxx>
> ---
> include/linux/trace_events.h | 4 ++--
> kernel/bpf/syscall.c | 13 +++++++------
> kernel/trace/bpf_trace.c | 3 ++-
> samples/bpf/task_fd_query_user.c | 16 ++++++++--------
> tools/bpf/bpftool/perf.c | 19 +++++++++++--------
> tools/lib/bpf/bpf.c | 3 ++-
> tools/lib/bpf/bpf.h | 2 +-
> tools/lib/bpf/libbpf.map | 1 +
> .../bpf/prog_tests/task_fd_query_rawtp.c | 10 +++++-----
> .../bpf/prog_tests/task_fd_query_tp.c | 4 ++--
> 10 files changed, 41 insertions(+), 34 deletions(-)
>
[...]
> diff --git a/tools/lib/bpf/bpf.h b/tools/lib/bpf/bpf.h
> index 079cc81ac51e..80bd705eca59 100644
> --- a/tools/lib/bpf/bpf.h
> +++ b/tools/lib/bpf/bpf.h
> @@ -315,7 +315,7 @@ LIBBPF_API int bpf_load_btf(const void *btf, __u32 btf_size, char *log_buf,
> __u32 log_buf_size, bool do_log);
> LIBBPF_API int bpf_task_fd_query(int pid, int fd, __u32 flags, char *buf,
> __u32 *buf_len, __u32 *prog_id, __u32 *fd_type,
> - __u64 *probe_offset, __u64 *probe_addr);
> + __u64 *probe_offset, __u64 *probe_addr, __u64 *bpf_cookie);
can't do this, it breaks source code compatibility.
bpf_task_fd_query() API isn't extendable.
>
> enum bpf_stats_type; /* defined in up-to-date linux/bpf.h */
> LIBBPF_API int bpf_enable_stats(enum bpf_stats_type type);
[...]
