[RFC][PATCH -v2 4/4] locking/mutex: Add lock handoff to avoid starvation

[Peter Zijlstra]

Now that we have an atomic owner field, we can do explicit lock
handoff. Use this to avoid starvation.

Signed-off-by: Peter Zijlstra (Intel) <peterz@xxxxxxxxxxxxx>
---
 kernel/locking/mutex.c |   52 +++++++++++++++++++++++++++++++++++++++++++------
 1 file changed, 46 insertions(+), 6 deletions(-)

--- a/kernel/locking/mutex.c
+++ b/kernel/locking/mutex.c
@@ -52,6 +52,7 @@ __mutex_init(struct mutex *lock, const c
 EXPORT_SYMBOL(__mutex_init);
 
 #define MUTEX_FLAG_WAITERS	0x01
+#define MUTEX_FLAG_HANDOFF	0x02
 
 #define MUTEX_FLAGS		0x03
 
@@ -117,6 +118,33 @@ static inline void __mutex_clear_flag(st
 	atomic_long_andnot(flag, &lock->owner);
 }
 
+static inline bool __mutex_waiter_is_first(struct mutex *lock, struct mutex_waiter *waiter)
+{
+	return list_first_entry(&lock->wait_list, struct mutex_waiter, list) == waiter;
+}
+
+static void __mutex_handoff(struct mutex *lock, struct task_struct *task)
+{
+	unsigned long owner = atomic_long_read(&lock->owner);
+
+	for (;;) {
+		unsigned long old, new;
+
+#ifdef CONFIG_DEBUG_MUTEXES
+		DEBUG_LOCKS_WARN_ON(__owner_task(owner) != current);
+#endif
+
+		new = (owner & MUTEX_FLAG_WAITERS);
+		new |= (unsigned long)task;
+
+		old = atomic_long_cmpxchg(&lock->owner, owner, new);
+		if (old == owner)
+			break;
+
+		owner = old;
+	}
+}
+
 #ifndef CONFIG_DEBUG_LOCK_ALLOC
 /*
  * We split the mutex lock/unlock logic into separate fastpath and
@@ -447,7 +475,7 @@ static bool mutex_optimistic_spin(struct
 }
 #endif
 
-static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock);
+static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock, unsigned long owner);
 
 /**
  * mutex_unlock - release the mutex
@@ -468,9 +496,12 @@ void __sched mutex_unlock(struct mutex *
 	DEBUG_LOCKS_WARN_ON(__mutex_owner(lock) != current);
 #endif
 
-	owner = atomic_long_fetch_and(MUTEX_FLAGS, &lock->owner);
+	owner = atomic_long_read(&lock->owner);
+	if (!(owner & MUTEX_FLAG_HANDOFF))
+		owner = atomic_long_fetch_and(MUTEX_FLAGS, &lock->owner);
+
 	if (__owner_flags(owner))
-		__mutex_unlock_slowpath(lock);
+		__mutex_unlock_slowpath(lock, owner);
 }
 EXPORT_SYMBOL(mutex_unlock);
 
@@ -568,7 +599,7 @@ __mutex_lock_common(struct mutex *lock,
 	list_add_tail(&waiter.list, &lock->wait_list);
 	waiter.task = task;
 
-	if (list_first_entry(&lock->wait_list, struct mutex_waiter, list) == &waiter) {
+	if (__mutex_waiter_is_first(lock, &waiter)) {
 		__mutex_set_flag(lock, MUTEX_FLAG_WAITERS);
 		/*
 		 * We must be sure to set WAITERS before attempting the trylock
@@ -605,13 +636,16 @@ __mutex_lock_common(struct mutex *lock,
 		spin_unlock_mutex(&lock->wait_lock, flags);
 		schedule_preempt_disabled();
 		spin_lock_mutex(&lock->wait_lock, flags);
+
+		if (__mutex_waiter_is_first(lock, &waiter))
+			__mutex_set_flag(lock, MUTEX_FLAG_HANDOFF);
 	}
 	__set_task_state(task, TASK_RUNNING);
 
 	mutex_remove_waiter(lock, &waiter, task);
 	/* set it to 0 if there are no waiters left: */
 	if (likely(list_empty(&lock->wait_list)))
-		__mutex_clear_flag(lock, MUTEX_FLAG_WAITERS);
+		__mutex_clear_flag(lock, MUTEX_FLAGS);
 
 	debug_mutex_free_waiter(&waiter);
 
@@ -737,8 +771,9 @@ EXPORT_SYMBOL_GPL(__ww_mutex_lock_interr
 /*
  * Release the lock, slowpath:
  */
-static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock)
+static noinline void __sched __mutex_unlock_slowpath(struct mutex *lock, unsigned long owner)
 {
+	struct task_struct *next = NULL;
 	unsigned long flags;
 	WAKE_Q(wake_q);
 
@@ -752,10 +787,15 @@ static noinline void __sched __mutex_unl
 				list_entry(lock->wait_list.next,
 					   struct mutex_waiter, list);
 
+		next = waiter->task;
+
 		debug_mutex_wake_waiter(lock, waiter);
 		wake_q_add(&wake_q, waiter->task);
 	}
 
+	if (owner & MUTEX_FLAG_HANDOFF)
+		__mutex_handoff(lock, next);
+
 	spin_unlock_mutex(&lock->wait_lock, flags);
 	wake_up_q(&wake_q);
 }