On Wed, 2025-07-16 at 14:38 +0200, Peter Zijlstra wrote:
> On Tue, Jul 15, 2025 at 09:14:29AM +0200, Gabriele Monaco wrote:
> > Add the following tracepoints:
> > * sched_set_need_resched(tsk, cpu, tif)
> > Called when a task is set the need resched [lazy] flag
> > * sched_switch_vain(preempt, tsk, tsk_state)
> > Called when a task is selected again during __schedule
> > i.e. prev == next == tsk : no real context switch
>
> > @@ -6592,6 +6598,7 @@ static bool try_to_block_task(struct rq *rq,
> > struct task_struct *p,
> > int flags = DEQUEUE_NOCLOCK;
> >
> > if (signal_pending_state(task_state, p)) {
> > + trace_sched_set_state_tp(p, TASK_RUNNING, true);
> > WRITE_ONCE(p->__state, TASK_RUNNING);
> > *task_state_p = TASK_RUNNING;
> > return false;
>
> I'm confused on the purpose of this. How does this relate to say the
> wakeup in signal_wake_up_state() ?
>
> > @@ -6786,6 +6793,7 @@ static void __sched notrace __schedule(int
> > sched_mode)
> > rq = context_switch(rq, prev, next, &rf);
> > } else {
> > rq_unpin_lock(rq, &rf);
> > + trace_sched_switch_vain_tp(preempt, prev,
> > prev_state);
> > __balance_callbacks(rq);
> > raw_spin_rq_unlock_irq(rq);
> > }
>
> Hurmph... don't you already have this covered by:
> trace_sched_exit_tp() ?
>
> Specifically, the only case where is_switch := false, is this case.
Mostly, it may work in some cases, but sched_exit happens with
interrupt enabled while all types of switches (including the vain ones)
must occur with interrupt disabled.
Some assumptions don't stand without this tracepoint, but I guess I
could adapt monitors to live without this if you believe it's not worth
adding a new tracepoint there.
Thanks,
Gabriele
