On Tue, Nov 25, 2025 at 10:49 AM Tomas Glozar <[email protected]> wrote:
>
> po 3. 11. 2025 v 15:45 odesílatel Wander Lairson Costa
> <[email protected]> napsal:
> > >
> > > Executing additional BPF code on latency threshold overflow allows doing
> > > doing low-latency and in-kernel troubleshooting of the cause of the
> >
> > typo: double "doing"
> >
>
> Thanks, I'll fix that :)
>
> > > --- a/tools/tracing/rtla/src/timerlat.c
> > > +++ b/tools/tracing/rtla/src/timerlat.c
> > > @@ -48,6 +48,17 @@ timerlat_apply_config(struct osnoise_tool *tool, 
> > > struct timerlat_params *params)
> > >               }
> > >       }
> > >
> > > +     /* Check if BPF action program is requested but BPF is not 
> > > available */
> > > +     if (params->bpf_action_program) {
> > > +             if (params->mode == TRACING_MODE_TRACEFS) {
> > > +                     err_msg("BPF actions are not supported in 
> > > tracefs-only mode\n");
> >
> > I would just emit a warning to the user and proceed ignoring the bpf action 
> > argument.
> >
>
> I believe if the user explicitly requests BPF actions to be used,
> measurement should not proceed without the action. Imagine someone
> setting --bpf-action in an automated test, expecting it to report
> something. But the action never fires, because they do not notice they
> are running an old kernel that does not support this.
>
> The user can always restart/reconfigure RTLA to skip the option.
>

I was thinking in a more interactive debug session where you go back and forth
with little changes in the command line. But you have a good point.

> Tomas
>


Reply via email to