sched/rtmutex: Refactor rt_mutex_setprio()

With the introduction of SCHED_DEADLINE the whole notion that priority
is a single number is gone, therefore the @prio argument to
rt_mutex_setprio() doesn't make sense anymore.

So rework the code to pass a pi_task instead.

Note this also fixes a problem with pi_top_task caching; previously we
would not set the pointer (call rt_mutex_update_top_task) if the
priority didn't change, this could lead to a stale pointer.

As for the XXX, I think its fine to use pi_task->prio, because if it
differs from waiter->prio, a PI chain update is immenent.

Signed-off-by: Peter Zijlstra (Intel) <peterz@infradead.org>
Cc: juri.lelli@arm.com
Cc: bigeasy@linutronix.de
Cc: xlpang@redhat.com
Cc: rostedt@goodmis.org
Cc: mathieu.desnoyers@efficios.com
Cc: jdesfossez@efficios.com
Cc: bristot@redhat.com
Link: http://lkml.kernel.org/r/20170323150216.303827095@infradead.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

2 files changed, 83 insertions, 95 deletions

diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 4b1015ef0dc7..00b49cdbb4e0 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -322,67 +322,16 @@ rt_mutex_dequeue_pi(struct task_struct *task, struct rt_mutex_waiter *waiter)
 	RB_CLEAR_NODE(&waiter->pi_tree_entry);
 }
 
-/*
- * Must hold both p->pi_lock and task_rq(p)->lock.
- */
-void rt_mutex_update_top_task(struct task_struct *p)
-{
-	if (!task_has_pi_waiters(p)) {
-		p->pi_top_task = NULL;
-		return;
-	}
-
-	p->pi_top_task = task_top_pi_waiter(p)->task;
-}
-
-/*
- * Calculate task priority from the waiter tree priority
- *
- * Return task->normal_prio when the waiter tree is empty or when
- * the waiter is not allowed to do priority boosting
- */
-int rt_mutex_getprio(struct task_struct *task)
-{
-	if (likely(!task_has_pi_waiters(task)))
-		return task->normal_prio;
-
-	return min(task_top_pi_waiter(task)->prio,
-		   task->normal_prio);
-}
-
-/*
- * Must hold either p->pi_lock or task_rq(p)->lock.
- */
-struct task_struct *rt_mutex_get_top_task(struct task_struct *task)
-{
-	return task->pi_top_task;
-}
-
-/*
- * Called by sched_setscheduler() to get the priority which will be
- * effective after the change.
- */
-int rt_mutex_get_effective_prio(struct task_struct *task, int newprio)
+static void rt_mutex_adjust_prio(struct task_struct *p)
 {
-	struct task_struct *top_task = rt_mutex_get_top_task(task);
+	struct task_struct *pi_task = NULL;
 
-	if (!top_task)
-		return newprio;
+	lockdep_assert_held(&p->pi_lock);
 
-	return min(top_task->prio, newprio);
-}
+	if (task_has_pi_waiters(p))
+		pi_task = task_top_pi_waiter(p)->task;
 
-/*
- * Adjust the priority of a task, after its pi_waiters got modified.
- *
- * This can be both boosting and unboosting. task->pi_lock must be held.
- */
-static void __rt_mutex_adjust_prio(struct task_struct *task)
-{
-	int prio = rt_mutex_getprio(task);
-
-	if (task->prio != prio || dl_prio(prio))
-		rt_mutex_setprio(task, prio);
+	rt_mutex_setprio(p, pi_task);
 }
 
 /*
@@ -742,7 +691,7 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
 		 */
 		rt_mutex_dequeue_pi(task, prerequeue_top_waiter);
 		rt_mutex_enqueue_pi(task, waiter);
-		__rt_mutex_adjust_prio(task);
+		rt_mutex_adjust_prio(task);
 
 	} else if (prerequeue_top_waiter == waiter) {
 		/*
@@ -758,7 +707,7 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
 		rt_mutex_dequeue_pi(task, waiter);
 		waiter = rt_mutex_top_waiter(lock);
 		rt_mutex_enqueue_pi(task, waiter);
-		__rt_mutex_adjust_prio(task);
+		rt_mutex_adjust_prio(task);
 	} else {
 		/*
 		 * Nothing changed. No need to do any priority
@@ -966,7 +915,7 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
 		return -EDEADLK;
 
 	raw_spin_lock(&task->pi_lock);
-	__rt_mutex_adjust_prio(task);
+	rt_mutex_adjust_prio(task);
 	waiter->task = task;
 	waiter->lock = lock;
 	waiter->prio = task->prio;
@@ -988,7 +937,7 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
 		rt_mutex_dequeue_pi(owner, top_waiter);
 		rt_mutex_enqueue_pi(owner, waiter);
 
-		__rt_mutex_adjust_prio(owner);
+		rt_mutex_adjust_prio(owner);
 		if (owner->pi_blocked_on)
 			chain_walk = 1;
 	} else if (rt_mutex_cond_detect_deadlock(waiter, chwalk)) {
@@ -1040,13 +989,14 @@ static void mark_wakeup_next_waiter(struct wake_q_head *wake_q,
 	waiter = rt_mutex_top_waiter(lock);
 
 	/*
-	 * Remove it from current->pi_waiters. We do not adjust a
-	 * possible priority boost right now. We execute wakeup in the
-	 * boosted mode and go back to normal after releasing
-	 * lock->wait_lock.
+	 * Remove it from current->pi_waiters and deboost.
+	 *
+	 * We must in fact deboost here in order to ensure we call
+	 * rt_mutex_setprio() to update p->pi_top_task before the
+	 * task unblocks.
 	 */
 	rt_mutex_dequeue_pi(current, waiter);
-	__rt_mutex_adjust_prio(current);
+	rt_mutex_adjust_prio(current);
 
 	/*
 	 * As we are waking up the top waiter, and the waiter stays
@@ -1058,9 +1008,19 @@ static void mark_wakeup_next_waiter(struct wake_q_head *wake_q,
 	 */
 	lock->owner = (void *) RT_MUTEX_HAS_WAITERS;
 
-	raw_spin_unlock(&current->pi_lock);
-
+	/*
+	 * We deboosted before waking the top waiter task such that we don't
+	 * run two tasks with the 'same' priority (and ensure the
+	 * p->pi_top_task pointer points to a blocked task). This however can
+	 * lead to priority inversion if we would get preempted after the
+	 * deboost but before waking our donor task, hence the preempt_disable()
+	 * before unlock.
+	 *
+	 * Pairs with preempt_enable() in rt_mutex_postunlock();
+	 */
+	preempt_disable();
 	wake_q_add(wake_q, waiter->task);
+	raw_spin_unlock(&current->pi_lock);
 }
 
 /*
@@ -1095,7 +1055,7 @@ static void remove_waiter(struct rt_mutex *lock,
 	if (rt_mutex_has_waiters(lock))
 		rt_mutex_enqueue_pi(owner, rt_mutex_top_waiter(lock));
 
-	__rt_mutex_adjust_prio(owner);
+	rt_mutex_adjust_prio(owner);
 
 	/* Store the lock on which owner is blocked or NULL */
 	next_lock = task_blocked_on_lock(owner);
@@ -1134,8 +1094,7 @@ void rt_mutex_adjust_pi(struct task_struct *task)
 	raw_spin_lock_irqsave(&task->pi_lock, flags);
 
 	waiter = task->pi_blocked_on;
-	if (!waiter || (waiter->prio == task->prio &&
-			!dl_prio(task->prio))) {
+	if (!waiter || (waiter->prio == task->prio && !dl_prio(task->prio))) {
 		raw_spin_unlock_irqrestore(&task->pi_lock, flags);
 		return;
 	}
@@ -1389,17 +1348,6 @@ static bool __sched rt_mutex_slowunlock(struct rt_mutex *lock,
 	 * Queue the next waiter for wakeup once we release the wait_lock.
 	 */
 	mark_wakeup_next_waiter(wake_q, lock);
-
-	/*
-	 * We should deboost before waking the top waiter task such that
-	 * we don't run two tasks with the 'same' priority. This however
-	 * can lead to prio-inversion if we would get preempted after
-	 * the deboost but before waking our high-prio task, hence the
-	 * preempt_disable before unlock. Pairs with preempt_enable() in
-	 * rt_mutex_postunlock();
-	 */
-	preempt_disable();
-
 	raw_spin_unlock_irqrestore(&lock->wait_lock, flags);
 
 	return true; /* call rt_mutex_postunlock() */
diff --git a/kernel/sched/core.c b/kernel/sched/core.c
index e1f44ec701c8..37b152a782ae 100644
--- a/kernel/sched/core.c
+++ b/kernel/sched/core.c
@@ -3671,10 +3671,25 @@ EXPORT_SYMBOL(default_wake_function);
 
 #ifdef CONFIG_RT_MUTEXES
 
+static inline int __rt_effective_prio(struct task_struct *pi_task, int prio)
+{
+	if (pi_task)
+		prio = min(prio, pi_task->prio);
+
+	return prio;
+}
+
+static inline int rt_effective_prio(struct task_struct *p, int prio)
+{
+	struct task_struct *pi_task = rt_mutex_get_top_task(p);
+
+	return __rt_effective_prio(pi_task, prio);
+}
+
 /*
  * rt_mutex_setprio - set the current priority of a task
- * @p: task
- * @prio: prio value (kernel-internal form)
+ * @p: task to boost
+ * @pi_task: donor task
  *
  * This function changes the 'effective' priority of a task. It does
  * not touch ->normal_prio like __setscheduler().
@@ -3682,18 +3697,42 @@ EXPORT_SYMBOL(default_wake_function);
  * Used by the rt_mutex code to implement priority inheritance
  * logic. Call site only calls if the priority of the task changed.
  */
-void rt_mutex_setprio(struct task_struct *p, int prio)
+void rt_mutex_setprio(struct task_struct *p, struct task_struct *pi_task)
 {
-	int oldprio, queued, running, queue_flag =
+	int prio, oldprio, queued, running, queue_flag =
 		DEQUEUE_SAVE | DEQUEUE_MOVE | DEQUEUE_NOCLOCK;
 	const struct sched_class *prev_class;
 	struct rq_flags rf;
 	struct rq *rq;
 
-	BUG_ON(prio > MAX_PRIO);
+	/* XXX used to be waiter->prio, not waiter->task->prio */
+	prio = __rt_effective_prio(pi_task, p->normal_prio);
+
+	/*
+	 * If nothing changed; bail early.
+	 */
+	if (p->pi_top_task == pi_task && prio == p->prio && !dl_prio(prio))
+		return;
 
 	rq = __task_rq_lock(p, &rf);
 	update_rq_clock(rq);
+	/*
+	 * Set under pi_lock && rq->lock, such that the value can be used under
+	 * either lock.
+	 *
+	 * Note that there is loads of tricky to make this pointer cache work
+	 * right. rt_mutex_slowunlock()+rt_mutex_postunlock() work together to
+	 * ensure a task is de-boosted (pi_task is set to NULL) before the
+	 * task is allowed to run again (and can exit). This ensures the pointer
+	 * points to a blocked task -- which guaratees the task is present.
+	 */
+	p->pi_top_task = pi_task;
+
+	/*
+	 * For FIFO/RR we only need to set prio, if that matches we're done.
+	 */
+	if (prio == p->prio && !dl_prio(prio))
+		goto out_unlock;
 
 	/*
 	 * Idle task boosting is a nono in general. There is one
@@ -3713,9 +3752,7 @@ void rt_mutex_setprio(struct task_struct *p, int prio)
 		goto out_unlock;
 	}
 
-	rt_mutex_update_top_task(p);
-
-	trace_sched_pi_setprio(p, prio);
+	trace_sched_pi_setprio(p, prio); /* broken */
 	oldprio = p->prio;
 
 	if (oldprio == prio)
@@ -3739,7 +3776,6 @@ void rt_mutex_setprio(struct task_struct *p, int prio)
 	 *          running task
 	 */
 	if (dl_prio(prio)) {
-		struct task_struct *pi_task = rt_mutex_get_top_task(p);
 		if (!dl_prio(p->normal_prio) ||
 		    (pi_task && dl_entity_preempt(&pi_task->dl, &p->dl))) {
 			p->dl.dl_boosted = 1;
@@ -3777,6 +3813,11 @@ out_unlock:
 	balance_callback(rq);
 	preempt_enable();
 }
+#else
+static inline int rt_effective_prio(struct task_struct *p, int prio)
+{
+	return prio;
+}
 #endif
 
 void set_user_nice(struct task_struct *p, long nice)
@@ -4023,10 +4064,9 @@ static void __setscheduler(struct rq *rq, struct task_struct *p,
 	 * Keep a potential priority boosting if called from
 	 * sched_setscheduler().
 	 */
+	p->prio = normal_prio(p);
 	if (keep_boost)
-		p->prio = rt_mutex_get_effective_prio(p, normal_prio(p));
-	else
-		p->prio = normal_prio(p);
+		p->prio = rt_effective_prio(p, p->prio);
 
 	if (dl_prio(p->prio))
 		p->sched_class = &dl_sched_class;
@@ -4313,7 +4353,7 @@ change:
 		 * the runqueue. This will be done when the task deboost
 		 * itself.
 		 */
-		new_effective_prio = rt_mutex_get_effective_prio(p, newprio);
+		new_effective_prio = rt_effective_prio(p, newprio);
 		if (new_effective_prio == oldprio)
 			queue_flags &= ~DEQUEUE_MOVE;
 	}