1 files changed, 78 insertions, 34 deletions

diff --git a/kernel/sched/stats.h b/kernel/sched/stats.h
index 3a3c826dd83a..c903f1a42891 100644
--- a/kernel/sched/stats.h
+++ b/kernel/sched/stats.h
@@ -1,4 +1,6 @@
 /* SPDX-License-Identifier: GPL-2.0 */
+#ifndef _KERNEL_STATS_H
+#define _KERNEL_STATS_H
 
 #ifdef CONFIG_SCHEDSTATS
 
@@ -105,55 +107,93 @@ __schedstats_from_se(struct sched_entity *se)
 }
 
 #ifdef CONFIG_PSI
+void psi_task_change(struct task_struct *task, int clear, int set);
+void psi_task_switch(struct task_struct *prev, struct task_struct *next,
+		     bool sleep);
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+void psi_account_irqtime(struct rq *rq, struct task_struct *curr, struct task_struct *prev);
+#else /* !CONFIG_IRQ_TIME_ACCOUNTING: */
+static inline void psi_account_irqtime(struct rq *rq, struct task_struct *curr,
+				       struct task_struct *prev) {}
+#endif /* !CONFIG_IRQ_TIME_ACCOUNTING */
 /*
  * PSI tracks state that persists across sleeps, such as iowaits and
  * memory stalls. As a result, it has to distinguish between sleeps,
- * where a task's runnable state changes, and requeues, where a task
- * and its state are being moved between CPUs and runqueues.
+ * where a task's runnable state changes, and migrations, where a task
+ * and its runnable state are being moved between CPUs and runqueues.
+ *
+ * A notable case is a task whose dequeue is delayed. PSI considers
+ * those sleeping, but because they are still on the runqueue they can
+ * go through migration requeues. In this case, *sleeping* states need
+ * to be transferred.
  */
-static inline void psi_enqueue(struct task_struct *p, bool wakeup)
+static inline void psi_enqueue(struct task_struct *p, int flags)
 {
-	int clear = 0, set = TSK_RUNNING;
+	int clear = 0, set = 0;
 
 	if (static_branch_likely(&psi_disabled))
 		return;
 
-	if (p->in_memstall)
-		set |= TSK_MEMSTALL_RUNNING;
+	/* Same runqueue, nothing changed for psi */
+	if (flags & ENQUEUE_RESTORE)
+		return;
+
+	/* psi_sched_switch() will handle the flags */
+	if (task_on_cpu(task_rq(p), p))
+		return;
 
-	if (!wakeup || p->sched_psi_wake_requeue) {
+	if (p->se.sched_delayed) {
+		/* CPU migration of "sleeping" task */
+		WARN_ON_ONCE(!(flags & ENQUEUE_MIGRATED));
 		if (p->in_memstall)
 			set |= TSK_MEMSTALL;
-		if (p->sched_psi_wake_requeue)
-			p->sched_psi_wake_requeue = 0;
+		if (p->in_iowait)
+			set |= TSK_IOWAIT;
+	} else if (flags & ENQUEUE_MIGRATED) {
+		/* CPU migration of runnable task */
+		set = TSK_RUNNING;
+		if (p->in_memstall)
+			set |= TSK_MEMSTALL | TSK_MEMSTALL_RUNNING;
 	} else {
+		/* Wakeup of new or sleeping task */
 		if (p->in_iowait)
 			clear |= TSK_IOWAIT;
+		set = TSK_RUNNING;
+		if (p->in_memstall)
+			set |= TSK_MEMSTALL_RUNNING;
 	}
 
 	psi_task_change(p, clear, set);
 }
 
-static inline void psi_dequeue(struct task_struct *p, bool sleep)
+static inline void psi_dequeue(struct task_struct *p, int flags)
 {
-	int clear = TSK_RUNNING;
-
 	if (static_branch_likely(&psi_disabled))
 		return;
 
+	/* Same runqueue, nothing changed for psi */
+	if (flags & DEQUEUE_SAVE)
+		return;
+
 	/*
 	 * A voluntary sleep is a dequeue followed by a task switch. To
 	 * avoid walking all ancestors twice, psi_task_switch() handles
 	 * TSK_RUNNING and TSK_IOWAIT for us when it moves TSK_ONCPU.
 	 * Do nothing here.
+	 *
+	 * In the SCHED_PROXY_EXECUTION case we may do sleeping
+	 * dequeues that are not followed by a task switch, so check
+	 * TSK_ONCPU is set to ensure the task switch is imminent.
+	 * Otherwise clear the flags as usual.
 	 */
-	if (sleep)
+	if ((flags & DEQUEUE_SLEEP) && (p->psi_flags & TSK_ONCPU))
 		return;
 
-	if (p->in_memstall)
-		clear |= (TSK_MEMSTALL | TSK_MEMSTALL_RUNNING);
-
-	psi_task_change(p, clear, 0);
+	/*
+	 * When migrating a task to another CPU, clear all psi
+	 * state. The enqueue callback above will work it out.
+	 */
+	psi_task_change(p, p->psi_flags, 0);
 }
 
 static inline void psi_ttwu_dequeue(struct task_struct *p)
@@ -165,20 +205,13 @@ static inline void psi_ttwu_dequeue(struct task_struct *p)
 	 * deregister its sleep-persistent psi states from the old
 	 * queue, and let psi_enqueue() know it has to requeue.
 	 */
-	if (unlikely(p->in_iowait || p->in_memstall)) {
+	if (unlikely(p->psi_flags)) {
 		struct rq_flags rf;
 		struct rq *rq;
-		int clear = 0;
-
-		if (p->in_iowait)
-			clear |= TSK_IOWAIT;
-		if (p->in_memstall)
-			clear |= TSK_MEMSTALL;
 
 		rq = __task_rq_lock(p, &rf);
-		psi_task_change(p, clear, 0);
-		p->sched_psi_wake_requeue = 1;
-		__task_rq_unlock(rq, &rf);
+		psi_task_change(p, p->psi_flags, 0);
+		__task_rq_unlock(rq, p, &rf);
 	}
 }
 
@@ -192,14 +225,16 @@ static inline void psi_sched_switch(struct task_struct *prev,
 	psi_task_switch(prev, next, sleep);
 }
 
-#else /* CONFIG_PSI */
-static inline void psi_enqueue(struct task_struct *p, bool wakeup) {}
-static inline void psi_dequeue(struct task_struct *p, bool sleep) {}
+#else /* !CONFIG_PSI: */
+static inline void psi_enqueue(struct task_struct *p, bool migrate) {}
+static inline void psi_dequeue(struct task_struct *p, bool migrate) {}
 static inline void psi_ttwu_dequeue(struct task_struct *p) {}
 static inline void psi_sched_switch(struct task_struct *prev,
 				    struct task_struct *next,
 				    bool sleep) {}
-#endif /* CONFIG_PSI */
+static inline void psi_account_irqtime(struct rq *rq, struct task_struct *curr,
+				       struct task_struct *prev) {}
+#endif /* !CONFIG_PSI */
 
 #ifdef CONFIG_SCHED_INFO
 /*
@@ -218,14 +253,17 @@ static inline void sched_info_dequeue(struct rq *rq, struct task_struct *t)
 	delta = rq_clock(rq) - t->sched_info.last_queued;
 	t->sched_info.last_queued = 0;
 	t->sched_info.run_delay += delta;
-
+	if (delta > t->sched_info.max_run_delay)
+		t->sched_info.max_run_delay = delta;
+	if (delta && (!t->sched_info.min_run_delay || delta < t->sched_info.min_run_delay))
+		t->sched_info.min_run_delay = delta;
 	rq_sched_info_dequeue(rq, delta);
 }
 
 /*
  * Called when a task finally hits the CPU.  We can now calculate how
  * long it was waiting to run.  We also note when it began so that we
- * can keep stats on how long its timeslice is.
+ * can keep stats on how long its time-slice is.
  */
 static void sched_info_arrive(struct rq *rq, struct task_struct *t)
 {
@@ -240,6 +278,10 @@ static void sched_info_arrive(struct rq *rq, struct task_struct *t)
 	t->sched_info.run_delay += delta;
 	t->sched_info.last_arrival = now;
 	t->sched_info.pcount++;
+	if (delta > t->sched_info.max_run_delay)
+		t->sched_info.max_run_delay = delta;
+	if (delta && (!t->sched_info.min_run_delay || delta < t->sched_info.min_run_delay))
+		t->sched_info.min_run_delay = delta;
 
 	rq_sched_info_arrive(rq, delta);
 }
@@ -297,4 +339,6 @@ sched_info_switch(struct rq *rq, struct task_struct *prev, struct task_struct *n
 # define sched_info_enqueue(rq, t)	do { } while (0)
 # define sched_info_dequeue(rq, t)	do { } while (0)
 # define sched_info_switch(rq, t, next)	do { } while (0)
-#endif /* CONFIG_SCHED_INFO */
+#endif /* !CONFIG_SCHED_INFO */
+
+#endif /* _KERNEL_STATS_H */