Merge tag 'sched_urgent_for_v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip

Pull scheduler fixes from Borislav Petkov:

 - Fix a performance regression when measuring the CPU time of a thread
   (clock_gettime(CLOCK_THREAD_CPUTIME_ID,...)) due to the addition of
   PSI IRQ time accounting in the hotpath

 - Fix a task_struct leak due to missing to decrement the refcount when
   the task is enqueued before the timer which is supposed to do that,
   expires

 - Revert an attempt to expedite detaching of movable tasks, as finding
   those could become very costly. Turns out the original issue wasn't
   even hit by anyone

* tag 'sched_urgent_for_v6.10' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip:
  sched: Move psi_account_irqtime() out of update_rq_clock_task() hotpath
  sched/deadline: Fix task_struct reference leak
  Revert "sched/fair: Make sure to try to detach at least one movable task"

Author: torvalds

kernel/sched/core.c

@@ -723,7 +723,6 @@ static void update_rq_clock_task(struct rq *rq, s64 delta)
 
 	rq->prev_irq_time += irq_delta;
 	delta -= irq_delta;
-	psi_account_irqtime(rq->curr, irq_delta);
 	delayacct_irq(rq->curr, irq_delta);
 #endif
 #ifdef CONFIG_PARAVIRT_TIME_ACCOUNTING
@@ -5665,7 +5664,7 @@ void sched_tick(void)
 {
 	int cpu = smp_processor_id();
 	struct rq *rq = cpu_rq(cpu);
-	struct task_struct *curr = rq->curr;
+	struct task_struct *curr;
 	struct rq_flags rf;
 	unsigned long hw_pressure;
 	u64 resched_latency;
@@ -5677,6 +5676,9 @@ void sched_tick(void)
 
 	rq_lock(rq, &rf);
 
+	curr = rq->curr;
+	psi_account_irqtime(rq, curr, NULL);
+
 	update_rq_clock(rq);
 	hw_pressure = arch_scale_hw_pressure(cpu_of(rq));
 	update_hw_load_avg(rq_clock_task(rq), rq, hw_pressure);
@@ -6737,6 +6739,7 @@ static void __sched notrace __schedule(unsigned int sched_mode)
 		++*switch_count;
 
 		migrate_disable_switch(rq, prev);
+		psi_account_irqtime(rq, prev, next);
 		psi_sched_switch(prev, next, !task_on_rq_queued(prev));
 
 		trace_sched_switch(sched_mode & SM_MASK_PREEMPT, prev, next, prev_state);

kernel/sched/deadline.c

@@ -1804,8 +1804,13 @@ static void enqueue_task_dl(struct rq *rq, struct task_struct *p, int flags)
 			 * The replenish timer needs to be canceled. No
 			 * problem if it fires concurrently: boosted threads
 			 * are ignored in dl_task_timer().
+			 *
+			 * If the timer callback was running (hrtimer_try_to_cancel == -1),
+			 * it will eventually call put_task_struct().
 			 */
-			hrtimer_try_to_cancel(&p->dl.dl_timer);
+			if (hrtimer_try_to_cancel(&p->dl.dl_timer) == 1 &&
+			    !dl_server(&p->dl))
+				put_task_struct(p);
 			p->dl.dl_throttled = 0;
 		}
 	} else if (!dl_prio(p->normal_prio)) {

kernel/sched/fair.c

@@ -9149,12 +9149,8 @@ static int detach_tasks(struct lb_env *env)
 			break;
 
 		env->loop++;
-		/*
-		 * We've more or less seen every task there is, call it quits
-		 * unless we haven't found any movable task yet.
-		 */
-		if (env->loop > env->loop_max &&
-		    !(env->flags & LBF_ALL_PINNED))
+		/* We've more or less seen every task there is, call it quits */
+		if (env->loop > env->loop_max)
 			break;
 
 		/* take a breather every nr_migrate tasks */
@@ -11393,9 +11389,7 @@ static int sched_balance_rq(int this_cpu, struct rq *this_rq,
 
 		if (env.flags & LBF_NEED_BREAK) {
 			env.flags &= ~LBF_NEED_BREAK;
-			/* Stop if we tried all running tasks */
-			if (env.loop < busiest->nr_running)
-				goto more_balance;
+			goto more_balance;
 		}
 
 		/*

kernel/sched/psi.c

@@ -773,6 +773,7 @@ static void psi_group_change(struct psi_group *group, int cpu,
 	enum psi_states s;
 	u32 state_mask;
 
+	lockdep_assert_rq_held(cpu_rq(cpu));
 	groupc = per_cpu_ptr(group->pcpu, cpu);
 
 	/*
@@ -991,22 +992,32 @@ void psi_task_switch(struct task_struct *prev, struct task_struct *next,
 }
 
 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
-void psi_account_irqtime(struct task_struct *task, u32 delta)
+void psi_account_irqtime(struct rq *rq, struct task_struct *curr, struct task_struct *prev)
 {
-	int cpu = task_cpu(task);
+	int cpu = task_cpu(curr);
 	struct psi_group *group;
 	struct psi_group_cpu *groupc;
-	u64 now;
+	u64 now, irq;
+	s64 delta;
 
 	if (static_branch_likely(&psi_disabled))
 		return;
 
-	if (!task->pid)
+	if (!curr->pid)
+		return;
+
+	lockdep_assert_rq_held(rq);
+	group = task_psi_group(curr);
+	if (prev && task_psi_group(prev) == group)
 		return;
 
 	now = cpu_clock(cpu);
+	irq = irq_time_read(cpu);
+	delta = (s64)(irq - rq->psi_irq_time);
+	if (delta < 0)
+		return;
+	rq->psi_irq_time = irq;
 
-	group = task_psi_group(task);
 	do {
 		if (!group->enabled)
 			continue;

kernel/sched/sched.h

@@ -1126,6 +1126,7 @@ struct rq {
 
 #ifdef CONFIG_IRQ_TIME_ACCOUNTING
 	u64			prev_irq_time;
+	u64			psi_irq_time;
 #endif
 #ifdef CONFIG_PARAVIRT
 	u64			prev_steal_time;

kernel/sched/stats.h

@@ -110,8 +110,12 @@ __schedstats_from_se(struct sched_entity *se)
 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);
-void psi_account_irqtime(struct task_struct *task, u32 delta);
-
+#ifdef CONFIG_IRQ_TIME_ACCOUNTING
+void psi_account_irqtime(struct rq *rq, struct task_struct *curr, struct task_struct *prev);
+#else
+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,
@@ -192,7 +196,8 @@ 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) {}
-static inline void psi_account_irqtime(struct task_struct *task, u32 delta) {}
+static inline void psi_account_irqtime(struct rq *rq, struct task_struct *curr,
+				       struct task_struct *prev) {}
 #endif /* CONFIG_PSI */
 
 #ifdef CONFIG_SCHED_INFO