[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20251009140925.GD4068168@noisy.programming.kicks-ass.net>
Date: Thu, 9 Oct 2025 16:09:25 +0200
From: Peter Zijlstra <peterz@...radead.org>
To: Dietmar Eggemann <dietmar.eggemann@....com>
Cc: tj@...nel.org, linux-kernel@...r.kernel.org, mingo@...nel.org,
juri.lelli@...hat.com, vincent.guittot@...aro.org,
rostedt@...dmis.org, bsegall@...gle.com, mgorman@...e.de,
vschneid@...hat.com, longman@...hat.com, hannes@...xchg.org,
mkoutny@...e.com, void@...ifault.com, arighi@...dia.com,
changwoo@...lia.com, cgroups@...r.kernel.org,
sched-ext@...ts.linux.dev, liuwenfang@...or.com, tglx@...utronix.de
Subject: Re: [PATCH 03/12] sched: Fold
sched_class::switch{ing,ed}_{to,from}() into the change pattern
On Thu, Oct 09, 2025 at 03:54:08PM +0200, Peter Zijlstra wrote:
> On Thu, Oct 09, 2025 at 03:30:02PM +0200, Dietmar Eggemann wrote:
> > On 06.10.25 12:44, Peter Zijlstra wrote:
> > > Add {DE,EN}QUEUE_CLASS and fold the sched_class::switch* methods into
> > > the change pattern. This completes and makes the pattern more
> > > symmetric.
> > >
> > > This changes the order of callbacks slightly:
> > >
> > > |
> > > | switching_from()
> > > dequeue_task(); | dequeue_task()
> > > put_prev_task(); | put_prev_task()
> > > | switched_from()
> > > |
> > > ... change task ... | ... change task ...
> > > |
> > > switching_to(); | switching_to()
> > > enqueue_task(); | enqueue_task()
> > > set_next_task(); | set_next_task()
> > > prev_class->switched_from() |
> > > switched_to() | switched_to()
> > > |
> > >
> > > Notably, it moves the switched_from() callback right after the
> > > dequeue/put. Existing implementations don't appear to be affected by
> > > this change in location -- specifically the task isn't enqueued on the
> > > class in question in either location.
> > >
> > > Make (CLASS)^(SAVE|MOVE), because there is nothing to save-restore
> > > when changing scheduling classes.
> >
> > This one causes a DL bw related warning when I run a simple 1 DL task
> > rt-app workload:
>
> > Not sure yet how this is related to switched_from_dl() being now called earlier?
>
> Ooh, I might see a problem. task_non_contending() uses dl_task(), which
> uses p->prio. The move above means it is now called using the 'old'
> prio, whereas it used to run with the 'new' prio.
>
> Let me see if I can figure out something for this.
Does this help? /me goes find rt-app.
diff --git a/kernel/sched/deadline.c b/kernel/sched/deadline.c
index 615411a0a881..fe2272c812b2 100644
--- a/kernel/sched/deadline.c
+++ b/kernel/sched/deadline.c
@@ -405,7 +405,7 @@ static void __dl_clear_params(struct sched_dl_entity *dl_se);
* up, and checks if the task is still in the "ACTIVE non contending"
* state or not (in the second case, it updates running_bw).
*/
-static void task_non_contending(struct sched_dl_entity *dl_se)
+static void task_non_contending(struct sched_dl_entity *dl_se, bool dl_task)
{
struct hrtimer *timer = &dl_se->inactive_timer;
struct rq *rq = rq_of_dl_se(dl_se);
@@ -444,10 +444,10 @@ static void task_non_contending(struct sched_dl_entity *dl_se)
} else {
struct task_struct *p = dl_task_of(dl_se);
- if (dl_task(p))
+ if (dl_task)
sub_running_bw(dl_se, dl_rq);
- if (!dl_task(p) || READ_ONCE(p->__state) == TASK_DEAD) {
+ if (!dl_task || READ_ONCE(p->__state) == TASK_DEAD) {
struct dl_bw *dl_b = dl_bw_of(task_cpu(p));
if (READ_ONCE(p->__state) == TASK_DEAD)
@@ -2045,7 +2045,7 @@ static void dequeue_dl_entity(struct sched_dl_entity *dl_se, int flags)
* or "inactive")
*/
if (flags & DEQUEUE_SLEEP)
- task_non_contending(dl_se);
+ task_non_contending(dl_se, true);
}
static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
@@ -2970,7 +2970,7 @@ static void switched_from_dl(struct rq *rq, struct task_struct *p)
* will reset the task parameters.
*/
if (task_on_rq_queued(p) && p->dl.dl_runtime)
- task_non_contending(&p->dl);
+ task_non_contending(&p->dl, false);
/*
* In case a task is setscheduled out from SCHED_DEADLINE we need to
Powered by blists - more mailing lists