On Thu, 2025-08-28 at 12:57 +0200, Tomas Glozar wrote: > čt 21. 8. 2025 v 5:58 odesílatel Crystal Wood <crw...@redhat.com> napsal: > > diff --git a/Documentation/tools/rtla/common_osnoise_options.rst > > b/Documentation/tools/rtla/common_osnoise_options.rst > > index d73de2d58f5f..ba8e6674c220 100644 > > --- a/Documentation/tools/rtla/common_osnoise_options.rst > > +++ b/Documentation/tools/rtla/common_osnoise_options.rst > > @@ -1,3 +1,11 @@ > > +.. |threshold| replace:: **-a/--auto**, **-s/--stop**, or > > **-S/--stop-total** > > +.. |thresharg| replace:: -s > > +.. |tracer| replace:: osnoise > > + > > +.. |actionsperf| replace:: > > + For time-sensitive actions, it is recommended to run **rtla** with > > + RT priority. > > + > > Unfortunately, real-time priority won't really help you in practice > for overflow-to-action latency, since the detection whether tracing > was stopped is only done once per second, when pulling the data. The > message should be instead just (something like): > > "Due to implementational limitations,actions might be delayed up to > one second after tracing is stopped." > > until either a trace event is implemented for stop tracing threshold > overflow (proposed in [1]), or rtla-osnoise is moved to BPF sample > collection, too. > > [1] > https://lore.kernel.org/linux-trace-kernel/CAP4=nvqwcwaoefhaqa82-vdb-00_y1-0fpopbeyetj2q1ew...@mail.gmail.com/
OK. I'm glad there's at least something we can say there, since RST didn't seem to like empty variables. :-P Looks like the workaround is to set it to a space character. I was going to ask if it should reference param->sleep_time instead, but it looks like that's not actually wired up to any config parameter... -Crystal
