Re: [RFC PATCH 14/17] sched: Add deadline tracepoints

[Juri Lelli <juri.lelli@redhat.com>]

Hi!

On 14/08/25 17:08, Gabriele Monaco wrote:
> Add the following tracepoints:
> 
> * sched_dl_throttle(dl):
>     Called when a deadline entity is throttled
> * sched_dl_replenish(dl):
>     Called when a deadline entity's runtime is replenished
> * sched_dl_server_start(dl):
>     Called when a deadline server is started
> * sched_dl_server_stop(dl, hard):
>     Called when a deadline server is stopped (hard) or put to idle
>     waiting for the next period (!hard)
> 
> Those tracepoints can be useful to validate the deadline scheduler with
> RV and are not exported to tracefs.
> 
> Signed-off-by: Gabriele Monaco <gmonaco@redhat.com>
> ---
>  include/trace/events/sched.h | 55 ++++++++++++++++++++++++++++++++++++
>  kernel/sched/deadline.c      |  8 ++++++
>  2 files changed, 63 insertions(+)
> 
> diff --git a/include/trace/events/sched.h b/include/trace/events/sched.h
> index 7b2645b50e78..f34cc1dc4a13 100644
> --- a/include/trace/events/sched.h
> +++ b/include/trace/events/sched.h
> @@ -609,6 +609,45 @@ TRACE_EVENT(sched_pi_setprio,
>  			__entry->oldprio, __entry->newprio)
>  );
>  
> +/*
> +DECLARE_EVENT_CLASS(sched_dl_template,
> +
> +	TP_PROTO(struct sched_dl_entity *dl),
> +
> +	TP_ARGS(dl),
> +
> +	TP_STRUCT__entry(
> +		__field(  struct task_struct *,	tsk		)
> +		__string( comm,		dl->dl_server ? "server" : container_of(dl, struct task_struct, dl)->comm	)
> +		__field(  pid_t,	pid		)
> +		__field(  s64,		runtime		)
> +		__field(  u64,		deadline	)
> +		__field(  int,		dl_yielded	)

I wonder if, while we are at it, we want to print all the other fields
as well (they might turn out to be useful). That would be

 .:: static (easier to retrieve with just a trace)
 - dl_runtime
 - dl_deadline
 - dl_period

 .:: behaviour (RECLAIM)
 - flags

 .:: state
 - dl_ bool flags in addition to dl_yielded

> +	),
> +
> +	TP_fast_assign(
> +		__assign_str(comm);
> +		__entry->pid		= dl->dl_server ? -1 : container_of(dl, struct task_struct, dl)->pid;
> +		__entry->runtime	= dl->runtime;
> +		__entry->deadline	= dl->deadline;
> +		__entry->dl_yielded	= dl->dl_yielded;
> +	),
> +
> +	TP_printk("comm=%s pid=%d runtime=%lld deadline=%lld yielded=%d",
                                                        ^^^
							llu ?

> +			__get_str(comm), __entry->pid,
> +			__entry->runtime, __entry->deadline,
> +			__entry->dl_yielded)
> +);

...

> @@ -1482,6 +1486,7 @@ static void update_curr_dl_se(struct rq *rq, struct sched_dl_entity *dl_se, s64
>  
>  throttle:
>  	if (dl_runtime_exceeded(dl_se) || dl_se->dl_yielded) {
> +		trace_sched_dl_throttle_tp(dl_se);
>  		dl_se->dl_throttled = 1;

I believe we also need to trace the dl_check_constrained_dl() throttle,
please take a look.

Also - we discussed this point a little already offline - but I still
wonder if we have to do anything special for dl-server defer. Those
entities are started as throttled until 0-lag, so maybe we should still
trace them explicitly as so?

In addition, since it's related, maybe we should do something about
sched_switch event, that is currently not aware of deadlines, runtimes,
etc.

Thanks,
Juri