rtmutex: Fix PI chain order integrity

rt_mutex_waiter::prio is a copy of task_struct::prio which is updated
during the PI chain walk, such that the PI chain order isn't messed up
by (asynchronous) task state updates.

Currently rt_mutex_waiter_less() uses task state for deadline tasks;
this is broken, since the task state can, as said above, change
asynchronously, causing the RB tree order to change without actual
tree update -> FAIL.

Fix this by also copying the deadline into the rt_mutex_waiter state
and updating it along with its prio field.

Ideally we would also force PI chain updates whenever DL tasks update
their deadline parameter, but for first approximation this is less
broken than it was.

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/20170323150216.403992539@infradead.org
Signed-off-by: Thomas Gleixner <tglx@linutronix.de>

2 files changed, 28 insertions, 2 deletions

diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c
index 00b49cdbb4e0..c6eda049ef9f 100644
--- a/kernel/locking/rtmutex.c
+++ b/kernel/locking/rtmutex.c
@@ -238,8 +238,7 @@ rt_mutex_waiter_less(struct rt_mutex_waiter *left,
 	 * then right waiter has a dl_prio() too.
 	 */
 	if (dl_prio(left->prio))
-		return dl_time_before(left->task->dl.deadline,
-				      right->task->dl.deadline);
+		return dl_time_before(left->deadline, right->deadline);
 
 	return 0;
 }
@@ -650,7 +649,26 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
 
 	/* [7] Requeue the waiter in the lock waiter tree. */
 	rt_mutex_dequeue(lock, waiter);
+
+	/*
+	 * Update the waiter prio fields now that we're dequeued.
+	 *
+	 * These values can have changed through either:
+	 *
+	 *   sys_sched_set_scheduler() / sys_sched_setattr()
+	 *
+	 * or
+	 *
+	 *   DL CBS enforcement advancing the effective deadline.
+	 *
+	 * Even though pi_waiters also uses these fields, and that tree is only
+	 * updated in [11], we can do this here, since we hold [L], which
+	 * serializes all pi_waiters access and rb_erase() does not care about
+	 * the values of the node being removed.
+	 */
 	waiter->prio = task->prio;
+	waiter->deadline = task->dl.deadline;
+
 	rt_mutex_enqueue(lock, waiter);
 
 	/* [8] Release the task */
@@ -777,6 +795,8 @@ static int rt_mutex_adjust_prio_chain(struct task_struct *task,
 static int try_to_take_rt_mutex(struct rt_mutex *lock, struct task_struct *task,
 				struct rt_mutex_waiter *waiter)
 {
+	lockdep_assert_held(&lock->wait_lock);
+
 	/*
 	 * Before testing whether we can acquire @lock, we set the
 	 * RT_MUTEX_HAS_WAITERS bit in @lock->owner. This forces all
@@ -902,6 +922,8 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
 	struct rt_mutex *next_lock;
 	int chain_walk = 0, res;
 
+	lockdep_assert_held(&lock->wait_lock);
+
 	/*
 	 * Early deadlock detection. We really don't want the task to
 	 * enqueue on itself just to untangle the mess later. It's not
@@ -919,6 +941,7 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock,
 	waiter->task = task;
 	waiter->lock = lock;
 	waiter->prio = task->prio;
+	waiter->deadline = task->dl.deadline;
 
 	/* Get the top priority waiter on the lock */
 	if (rt_mutex_has_waiters(lock))
@@ -1036,6 +1059,8 @@ static void remove_waiter(struct rt_mutex *lock,
 	struct task_struct *owner = rt_mutex_owner(lock);
 	struct rt_mutex *next_lock;
 
+	lockdep_assert_held(&lock->wait_lock);
+
 	raw_spin_lock(&current->pi_lock);
 	rt_mutex_dequeue(lock, waiter);
 	current->pi_blocked_on = NULL;
diff --git a/kernel/locking/rtmutex_common.h b/kernel/locking/rtmutex_common.h
index 9e36aeddce18..72ad45a9a794 100644
--- a/kernel/locking/rtmutex_common.h
+++ b/kernel/locking/rtmutex_common.h
@@ -34,6 +34,7 @@ struct rt_mutex_waiter {
 	struct rt_mutex		*deadlock_lock;
 #endif
 	int prio;
+	u64 deadline;
 };
 
 /*