Re: [PATCH 2/2] sched/debug: add sched_update_nr_running tracepoint

[Alexei Starovoitov]

On Wed, Sep 4, 2019 at 8:40 AM Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
>
> On Wed, Sep 04, 2019 at 08:25:27AM -0700, Alexei Starovoitov wrote:
> > On Wed, Sep 4, 2019 at 6:10 AM Joel Fernandes <joel@xxxxxxxxxxxxxxxxx> wrote:
> > >
> > > I wonder if this distinction of "tracepoint" being non-ABI can be documented
> > > somewhere. I would be happy to do that if there is a place for the same. I
> > > really want some general "policy" in the kernel on where we draw a line in
> > > the sand with respect to tracepoints and ABI :).
> >
> > It's been discussed millions times. tracepoints are not abi.
> > Example: android folks started abusing tracepoints inside bpf core
> > and we _deleted_ them.
>
> This is news to me, which ones?

those that your android teammates abused!

> > Same thing can be done with _any_ tracepoint.
> > Do not abuse them and stop the fud about abi.
>
> I don't know what FUD you are referring to. At least it is not coming from
> me. This thread is dealing with the issue about ABI specifically, I jumped in
> just now. As I was saying earlier, I don't have a strong opinion about this.
> I just want to know what is the agreed upon approach so that we can stick to
> it.
>
> It sounds like the agreement here is tracepoints can be added and used
> without ABI guarantees, however the same is not true with trace events.
> Where's the FUD in that?

Anything in tracing can be deleted.
Tracing is about debugging and introspection.
When underlying kernel code changes the introspection points change as well.