Re: [RFC][PATCH] sched: Make class_schedulers avoid pushing current, and get rid of proxy_tag_curr()

[Peter Zijlstra]

On Sat, Mar 07, 2026 at 07:39:29AM +0000, John Stultz wrote:
> With proxy-execution, the scheduler selects the donor, but for
> blocked donors, we end up running the lock owner.
> 
> This caused some complexity, because the class schedulers make
> sure to remove the task they pick from their pushable task
> lists, which prevents the donor from being migrated, but there
> wasn't then anything to prevent rq->curr from being migrated
> if rq->curr != rq->donor.
> 
> This was sort of hacked around by calling proxy_tag_curr() on
> the rq->curr task if we were running something other then the
> donor. proxy_tag_curr() did a dequeue/enqueue pair on the
> rq->curr task, allowing the class schedulers to remove it from
> their pushable list.
> 
> The dequeue/enqueue pair was wasteful, and additionally K Prateek
> highlighted that we didn't properly undo things when we stopped
> proxying, leaving the lock owner off the pushable list.
> 
> After some alternative approaches were considered, Peter
> suggested just having the RT/DL classes just avoid migrating
> when task_on_cpu().
> 
> So rework pick_next_pushable_dl_task() and the rt
> pick_next_pushable_task() functions so that they skip over the
> first pushable task if it is on_cpu.
> 
> Then just drop all of the proxy_tag_curr() logic.
> 
> Fixes: be39617e38e0 ("sched: Fix proxy/current (push,pull)ability")
> Reported-by: K Prateek Nayak <kprateek.nayak@xxxxxxx>
> Closes: https://lore.kernel.org/lkml/e735cae0-2cc9-4bae-b761-fcb082ed3e94@xxxxxxx/
> Suggested-by: Peter Zijlstra <peterz@xxxxxxxxxxxxx>
> Signed-off-by: John Stultz <jstultz@xxxxxxxxxx>

Right, that works for me ;-)

Sone bits I also had in my 'patch' that didn't make it, and quite
frankly don't belong in the same patch anyway, is the below.

Compilers are really bad (as in they utterly refuse) optimizing (even
when marked with __pure) the static branch things, and will happily emit
multiple identical in a row.

So pull out the one obvious sched_proxy_exec() branch in __schedule()
and remove some of the 'implicit' ones in that path.

Hmm?

---
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -6597,11 +6597,7 @@ find_proxy_task(struct rq *rq, struct ta
 	struct mutex *mutex;
 
 	/* Follow blocked_on chain. */
-	for (p = donor; task_is_blocked(p); p = owner) {
-		mutex = p->blocked_on;
-		/* Something changed in the chain, so pick again */
-		if (!mutex)
-			return NULL;
+	for (p = donor; (mutex = p->blocked_on); p = owner) {
 		/*
 		 * By taking mutex->wait_lock we hold off concurrent mutex_unlock()
 		 * and ensure @owner sticks around.
@@ -6829,14 +6825,16 @@ static void __sched notrace __schedule(i
 
 pick_again:
 	next = pick_next_task(rq, rq->donor, &rf);
-	rq_set_donor(rq, next);
 	rq->next_class = next->sched_class;
-	if (unlikely(task_is_blocked(next))) {
-		next = find_proxy_task(rq, next, &rf);
-		if (!next)
-			goto pick_again;
-		if (next == rq->idle)
-			goto keep_resched;
+	if (sched_proxy_exec()) {
+		rq_set_donor(rq, next);
+		if (p->blocked_on) {
+			next = find_proxy_task(rq, next, &rf);
+			if (!next)
+				goto pick_again;
+			if (next == rq->idle)
+				goto keep_resched;
+		}
 	}
 picked:
 	clear_tsk_need_resched(prev);
@@ -6886,10 +6884,6 @@ static void __sched notrace __schedule(i
 		/* Also unlocks the rq: */
 		rq = context_switch(rq, prev, next, &rf);
 	} else {
-		/* In case next was already curr but just got blocked_donor */
-		if (!task_current_donor(rq, next))
-			proxy_tag_curr(rq, next);
-
 		rq_unpin_lock(rq, &rf);
 		__balance_callbacks(rq, NULL);
 		raw_spin_rq_unlock_irq(rq);