On 01/23/2014 05:07 AM, Peter Zijlstra wrote:

On Wed, Jan 22, 2014 at 04:33:55PM -0500, Waiman Long wrote:

quoted

+/**
+ * queue_read_unlock - release read lock of a queue rwlock
+ * @lock : Pointer to queue rwlock structure
+ */
+static inline void queue_read_unlock(struct qrwlock *lock)
+{
+	/*
+	 * Atomically decrement the reader count
+	 */
+	atomic_sub(_QR_BIAS,&lock->cnts.rwa);
+}
+
+/**
+ * queue_write_unlock - release write lock of a queue rwlock
+ * @lock : Pointer to queue rwlock structure
+ */
+static inline void queue_write_unlock(struct qrwlock *lock)
+{
+	/*
+	 * If the writer field is atomic, it can be cleared directly.
+	 * Otherwise, an atomic subtraction will be used to clear it.
+	 */
+	if (__native_word(lock->cnts.writer))
+		smp_store_release(&lock->cnts.writer, 0);
+	else
+		atomic_sub(_QW_LOCKED,&lock->cnts.rwa);
+}

Both these unlocks miss a barrier; atomic_sub() doesn't imply any
barrier what so ever.

The smp_store_release() does, but the other two are invalid release ops
in generic.

I thought that all atomic RMW instructions are memory barrier. If they 
are not, what kind of barrier should be added?

-Longman