sched: Rework dl_server

When a task is selected through a dl_server, it will have p->dl_server
set, such that it can account runtime to the dl_server, see
update_curr_task().

Currently p->dl_server is set in pick*task() whenever it goes through
the dl_server, clearing it is a bit of a mess though. The trivial
solution is clearing it on the final put (now that we have this
location).

However, this gives a problem when:

	p = pick_task(rq);
	if (p)
		put_prev_set_next_task(rq, prev, next);

picks the same task but through a different path, notably when it goes
from picking through the dl_server to a direct pick or vice-versa. In
that case we cannot readily determine wether we should clear or
preserve p->dl_server.

An additional complication is pick_*task() setting p->dl_server for a
remote pick, it might still need to update runtime before it schedules
the core_pick.

Close all these holes and remove all the random clearing of
p->dl_server by:

 - having pick_*task() manage rq->dl_server

 - having the final put_prev_task() clear p->dl_server

 - having the first set_next_task() set p->dl_server = rq->dl_server

 - complicate the core_sched code to save/restore rq->dl_server where
   appropriate.
Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Link: https://lore.kernel.org/r/20240813224016.259853414@infradead.org

kernel/sched/core.c

@@ -3668,8 +3668,6 @@ ttwu_do_activate(struct rq *rq, struct task_struct *p, int wake_flags,
 		rq->idle_stamp = 0;
 	}
 #endif
-
-	p->dl_server = NULL;
 }
 
 /*
@@ -5859,14 +5857,6 @@ static void prev_balance(struct rq *rq, struct task_struct *prev,
 			break;
 	}
 #endif
-
-	/*
-	 * We've updated @prev and no longer need the server link, clear it.
-	 * Must be done before ->pick_next_task() because that can (re)set
-	 * ->dl_server.
-	 */
-	if (prev->dl_server)
-		prev->dl_server = NULL;
 }
 
 /*
@@ -5878,6 +5868,8 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 	const struct sched_class *class;
 	struct task_struct *p;
 
+	rq->dl_server = NULL;
+
 	/*
 	 * Optimization: we know that if all tasks are in the fair class we can
 	 * call that function directly, but only if the @prev task wasn't of a
@@ -5897,20 +5889,6 @@ __pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 			put_prev_set_next_task(rq, prev, p);
 		}
 
-		/*
-		 * This is a normal CFS pick, but the previous could be a DL pick.
-		 * Clear it as previous is no longer picked.
-		 */
-		if (prev->dl_server)
-			prev->dl_server = NULL;
-
-		/*
-		 * This is the fast path; it cannot be a DL server pick;
-		 * therefore even if @p == @prev, ->dl_server must be NULL.
-		 */
-		if (p->dl_server)
-			p->dl_server = NULL;
-
 		return p;
 	}
 
@@ -5958,6 +5936,8 @@ static inline struct task_struct *pick_task(struct rq *rq)
 	const struct sched_class *class;
 	struct task_struct *p;
 
+	rq->dl_server = NULL;
+
 	for_each_class(class) {
 		p = class->pick_task(rq);
 		if (p)
@@ -5996,6 +5976,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 		 * another cpu during offline.
 		 */
 		rq->core_pick = NULL;
+		rq->core_dl_server = NULL;
 		return __pick_next_task(rq, prev, rf);
 	}
 
@@ -6014,7 +5995,9 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 		WRITE_ONCE(rq->core_sched_seq, rq->core->core_pick_seq);
 
 		next = rq->core_pick;
+		rq->dl_server = rq->core_dl_server;
 		rq->core_pick = NULL;
+		rq->core_dl_server = NULL;
 		goto out_set_next;
 	}
 
@@ -6059,6 +6042,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 		next = pick_task(rq);
 		if (!next->core_cookie) {
 			rq->core_pick = NULL;
+			rq->core_dl_server = NULL;
 			/*
 			 * For robustness, update the min_vruntime_fi for
 			 * unconstrained picks as well.
@@ -6086,7 +6070,9 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 		if (i != cpu && (rq_i != rq->core || !core_clock_updated))
 			update_rq_clock(rq_i);
 
-		p = rq_i->core_pick = pick_task(rq_i);
+		rq_i->core_pick = p = pick_task(rq_i);
+		rq_i->core_dl_server = rq_i->dl_server;
+
 		if (!max || prio_less(max, p, fi_before))
 			max = p;
 	}
@@ -6110,6 +6096,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 		}
 
 		rq_i->core_pick = p;
+		rq_i->core_dl_server = NULL;
 
 		if (p == rq_i->idle) {
 			if (rq_i->nr_running) {
@@ -6170,6 +6157,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 
 		if (i == cpu) {
 			rq_i->core_pick = NULL;
+			rq_i->core_dl_server = NULL;
 			continue;
 		}
 
@@ -6178,6 +6166,7 @@ pick_next_task(struct rq *rq, struct task_struct *prev, struct rq_flags *rf)
 
 		if (rq_i->curr == rq_i->core_pick) {
 			rq_i->core_pick = NULL;
+			rq_i->core_dl_server = NULL;
 			continue;
 		}
 
@@ -8401,6 +8390,7 @@ void __init sched_init(void)
 #ifdef CONFIG_SCHED_CORE
 		rq->core = rq;
 		rq->core_pick = NULL;
+		rq->core_dl_server = NULL;
 		rq->core_enabled = 0;
 		rq->core_tree = RB_ROOT;
 		rq->core_forceidle_count = 0;

kernel/sched/deadline.c

@@ -2423,7 +2423,7 @@ static struct task_struct *__pick_task_dl(struct rq *rq)
 			update_curr_dl_se(rq, dl_se, 0);
 			goto again;
 		}
-		p->dl_server = dl_se;
+		rq->dl_server = dl_se;
 	} else {
 		p = dl_task_of(dl_se);
 	}

kernel/sched/fair.c

@@ -8749,14 +8749,6 @@ static struct task_struct *pick_task_fair(struct rq *rq)
 		cfs_rq = group_cfs_rq(se);
 	} while (cfs_rq);
 
-	/*
-	 * This can be called from directly from CFS's ->pick_task() or indirectly
-	 * from DL's ->pick_task when fair server is enabled. In the indirect case,
-	 * DL will set ->dl_server just after this function is called, so its Ok to
-	 * clear. In the direct case, we are picking directly so we must clear it.
-	 */
-	task_of(se)->dl_server = NULL;
-
 	return task_of(se);
 }
 
@@ -8780,6 +8772,8 @@ pick_next_task_fair(struct rq *rq, struct task_struct *prev, struct rq_flags *rf
 	if (prev->sched_class != &fair_sched_class)
 		goto simple;
 
+	__put_prev_set_next_dl_server(rq, prev, p);
+
 	/*
 	 * Because of the set_next_buddy() in dequeue_task_fair() it is rather
 	 * likely that a next task is from the same cgroup as the current.

kernel/sched/sched.h

@@ -1066,6 +1066,7 @@ struct rq {
 	unsigned int		nr_uninterruptible;
 
 	struct task_struct __rcu	*curr;
+	struct sched_dl_entity	*dl_server;
 	struct task_struct	*idle;
 	struct task_struct	*stop;
 	unsigned long		next_balance;
@@ -1193,6 +1194,7 @@ struct rq {
 	/* per rq */
 	struct rq		*core;
 	struct task_struct	*core_pick;
+	struct sched_dl_entity	*core_dl_server;
 	unsigned int		core_enabled;
 	unsigned int		core_sched_seq;
 	struct rb_root		core_tree;
@@ -2370,12 +2372,24 @@ static inline void set_next_task(struct rq *rq, struct task_struct *next)
 	next->sched_class->set_next_task(rq, next, false);
 }
 
+static inline void
+__put_prev_set_next_dl_server(struct rq *rq,
+			      struct task_struct *prev,
+			      struct task_struct *next)
+{
+	prev->dl_server = NULL;
+	next->dl_server = rq->dl_server;
+	rq->dl_server = NULL;
+}
+
 static inline void put_prev_set_next_task(struct rq *rq,
 					  struct task_struct *prev,
 					  struct task_struct *next)
 {
 	WARN_ON_ONCE(rq->curr != prev);
 
+	__put_prev_set_next_dl_server(rq, prev, next);
+
 	if (next == prev)
 		return;