Re: BPF logging infrastructure. Was: [PATCH bpf-next 4/6] tools: add new members to bpf_attr.raw_tracepoint in bpf.h

["Toke Høiland-Jørgensen" <toke@redhat.com>]

Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:

>> So myself and Toke are wearing 'bpf user' hat in that context.
>> Both of us indicated that libbpf output is too verbose.
>> Your response "just send a patch" is a sure way to turn away more users.
>>
>
> I can't find any such complaint from Toke in this thread, and can't
> really recall something like that from recent discussions. I'd rather
> have him speak for himself.

I think what I said (not in this thread, way back during some other
discussion) was that I agreed that libbpf was being too verbose by
dumping all the sections and relocations it finds when reading an ELF
file, which causes the useful error messages to get lost. I would like
to see those messages demoted to another log level, or removed
altogether.

I won't have time to look more at this right now, but I do plan to
circle back to it: I agree with you that we need to make this more
friendly. And yes, I also think this should include finding a way to
disambiguate between different conditions leading to the same error from
the kernel. I've run into a lot of the same issues as you when
supporting people who are new to BPF - thank you for the extensive list!

-Toke