Thread (31 messages) 31 messages, 7 authors, 2011-03-30

Re: [PATCH 2/2] mutex: Apply adaptive spinning on mutex_trylock()

From: Peter Zijlstra <peterz@infradead.org>
Date: 2011-03-29 17:35:25
Also in: lkml

On Tue, 2011-03-29 at 19:09 +0200, Tejun Heo wrote:
Here's the combined patch I was planning on testing but didn't get to
(yet).  It implements two things - hard limit on spin duration and
early break if the owner also is spinning on a mutex.
This is going to give massive conflicts with

 https://lkml.org/lkml/2011/3/2/286
 https://lkml.org/lkml/2011/3/2/282

which I was planning to stuff into .40

quoted hunk ↗ jump to hunk
@@ -4021,16 +4025,44 @@ EXPORT_SYMBOL(schedule);
 
 #ifdef CONFIG_MUTEX_SPIN_ON_OWNER
 /*
+ * Maximum mutex owner spin duration in nsecs.  Don't spin more then
+ * DEF_TIMESLICE.
+ */
+#define MAX_MUTEX_SPIN_NS	(DEF_TIMESLICE * 1000000000LLU / HZ)
DEF_TIMESLICE is SCHED_RR only, so its use here is dubious at best, also
I bet we have something like NSEC_PER_SEC to avoid counting '0's.
+
+/**
+ * mutex_spin_on_owner - optimistic adaptive spinning on locked mutex
+ * @lock: the mutex to spin on
+ * @owner: the current owner (speculative pointer)
+ *
+ * The caller is trying to acquire @lock held by @owner.  If @owner is
+ * currently running, it might get unlocked soon and spinning on it can
+ * save the overhead of sleeping and waking up.
+ *
+ * Note that @owner is completely speculative and may be completely
+ * invalid.  It should be accessed very carefully.
+ *
+ * Forward progress is guaranteed regardless of locking ordering by never
+ * spinning longer than MAX_MUTEX_SPIN_NS.  This is necessary because
+ * mutex_trylock(), which doesn't have to follow the usual locking
+ * ordering, also uses this function.
While that puts a limit on things it'll still waste time. I'd much
rather pass an trylock argument to mutex_spin_on_owner() and then bail
on owner also spinning.
+ * CONTEXT:
+ * Preemption disabled.
+ *
+ * RETURNS:
+ * %true if the lock was released and the caller should retry locking.
+ * %false if the caller better go sleeping.
  */
-int mutex_spin_on_owner(struct mutex *lock, struct thread_info *owner)
+bool mutex_spin_on_owner(struct mutex *lock, struct thread_info *owner)
 {
quoted hunk ↗ jump to hunk
@@ -4070,21 +4104,30 @@ int mutex_spin_on_owner(struct mutex *lo
 			 * we likely have heavy contention. Return 0 to quit
 			 * optimistic spinning and not contend further:
 			 */
+			ret = !lock->owner;
 			break;
 		}
 
 		/*
-		 * Is that owner really running on that cpu?
+		 * Quit spinning if any of the followings is true.
+		 *
+		 * - The owner isn't running on that cpu.
+		 * - The owner also is spinning on a mutex.
+		 * - Someone else wants to use this cpu.
+		 * - We've been spinning for too long.
 		 */
+		if (task_thread_info(rq->curr) != owner ||
+		    rq->spinning_on_mutex || need_resched() ||
+		    local_clock() > start + MAX_MUTEX_SPIN_NS) {
While we did our best with making local_clock() cheap, I'm still fairly
uncomfortable with putting it in such a tight loop.
+			ret = false;
+			break;
+		}
 
 		arch_mutex_cpu_relax();
 	}
 
+	this_rq()->spinning_on_mutex = false;
+	return ret;
 }
 #endif
 
Keyboard shortcuts
hback out one level
jnext message in thread
kprevious message in thread
ldrill in
Escclose help / fold thread tree
?toggle this help