Messages in this thread Patch in this message | | | From | Davidlohr Bueso <> | Subject | [PATCH 1/4] locking/rtmutex: Implement lockless top-waiter wakeup | Date | Tue, 19 May 2015 10:24:55 -0700 |
| |
Mark the task for later wakeup after the wait_lock has been released. This way, once the next task is awoken, it will have a better chance to of finding the wait_lock free when continuing executing in __rt_mutex_slowlock() when trying to acquire the rtmutex, calling try_to_take_rt_mutex(). Upon contended scenarios, other tasks attempting take the lock may acquire it first, right after the wait_lock is released, but (a) this can also occur with the current code, as it relies on the spinlock fairness, and (b) we are dealing with the top-waiter anyway, so it will always take the lock next.
Signed-off-by: Davidlohr Bueso <dbueso@suse.de> --- kernel/locking/rtmutex.c | 21 ++++++++++----------- 1 file changed, 10 insertions(+), 11 deletions(-)
diff --git a/kernel/locking/rtmutex.c b/kernel/locking/rtmutex.c index 36573e9..74188d8 100644 --- a/kernel/locking/rtmutex.c +++ b/kernel/locking/rtmutex.c @@ -955,14 +955,13 @@ static int task_blocks_on_rt_mutex(struct rt_mutex *lock, } /* - * Wake up the next waiter on the lock. - * * Remove the top waiter from the current tasks pi waiter list and - * wake it up. + * queue it up. * * Called with lock->wait_lock held. */ -static void wakeup_next_waiter(struct rt_mutex *lock) +static void mark_wakeup_next_waiter(struct wake_q_head *wake_q, + struct rt_mutex *lock) { struct rt_mutex_waiter *waiter; unsigned long flags; @@ -991,12 +990,7 @@ static void wakeup_next_waiter(struct rt_mutex *lock) raw_spin_unlock_irqrestore(¤t->pi_lock, flags); - /* - * It's safe to dereference waiter as it cannot go away as - * long as we hold lock->wait_lock. The waiter task needs to - * acquire it in order to dequeue the waiter. - */ - wake_up_process(waiter->task); + wake_q_add(wake_q, waiter->task); } /* @@ -1255,6 +1249,8 @@ static inline int rt_mutex_slowtrylock(struct rt_mutex *lock) static void __sched rt_mutex_slowunlock(struct rt_mutex *lock) { + WAKE_Q(wake_q); + raw_spin_lock(&lock->wait_lock); debug_rt_mutex_unlock(lock); @@ -1303,10 +1299,13 @@ rt_mutex_slowunlock(struct rt_mutex *lock) /* * The wakeup next waiter path does not suffer from the above * race. See the comments there. + * + * Queue the next waiter for wakeup once we release the wait_lock. */ - wakeup_next_waiter(lock); + mark_wakeup_next_waiter(&wake_q, lock); raw_spin_unlock(&lock->wait_lock); + wake_up_q(&wake_q); /* Undo pi boosting if necessary: */ rt_mutex_adjust_prio(current); -- 2.1.4
| |