[PATCH 4/7] spinlock.h: Move smp_mb__after_unlock_lock to spinlock.h

[Manfred Spraul]

Right now, there are two places for special memory barriers
around lock/unlock:
- <include/linux/spinlock.h> defines smp_mb__before_spinlock()
- <kernel/rcu/tree.h> defines smb_mb_after_unlock_lock(),
  with hardcoded CONFIG_PPC as arch override.

To improve transparency, move the define from kernel/rcu/tree.h
to <include/linux>, and add the standard approach for overrides.

v3: If smp_mb__after_unlock_lock() is in barrier.h, then
for arm64, kernel/rcu/tree.c doesn't compile because barrier.h
is not included in kernel/rcu/tree.c

(v2 was: add example from Paul, something that can happen on real HW)

spin_unlock() + spin_lock() together do not form a full memory barrier:
(everything initialized to 0)

CPU1:
  a=1;
  spin_unlock(&b);
  spin_lock(&c);
+ smp_mb__after_unlock_lock();
  r1=d;

CPU2:
  d=1;
  smp_mb();
  r2=a;

Without the smp_mb__after_unlock_lock(), r1==0 && r2==0 would
be possible.

Signed-off-by: Manfred Spraul <manfred@xxxxxxxxxxxxxxxx>
--
---
 include/linux/spinlock.h | 16 ++++++++++++++++
 kernel/rcu/tree.h        | 12 ------------
 2 files changed, 16 insertions(+), 12 deletions(-)

diff --git a/include/linux/spinlock.h b/include/linux/spinlock.h
index 47dd0ce..6436a66 100644
--- a/include/linux/spinlock.h
+++ b/include/linux/spinlock.h
@@ -130,6 +130,22 @@ do {								\
 #define smp_mb__before_spinlock()	smp_wmb()
 #endif
 
+#ifndef smp_mb__after_unlock_lock
+/**
+ * smp_mb__after_unlock_lock() - Provide smp_mb() after unlock+lock
+ *
+ * Place this after a lock-acquisition primitive to guarantee that
+ * an UNLOCK+LOCK pair act as a full barrier.  This guarantee applies
+ * if the UNLOCK and LOCK are executed by the same CPU or if the
+ * UNLOCK and LOCK operate on the same lock variable.
+ */
+#ifdef CONFIG_PPC
+#define smp_mb__after_unlock_lock()	smp_mb()  /* Full ordering for lock. */
+#else /* #ifdef CONFIG_PPC */
+#define smp_mb__after_unlock_lock()	do { } while (0)
+#endif /* #else #ifdef CONFIG_PPC */
+#endif
+
 /**
  * raw_spin_unlock_wait - wait until the spinlock gets unlocked
  * @lock: the spinlock in question.
diff --git a/kernel/rcu/tree.h b/kernel/rcu/tree.h
index e99a523..a0cd9ab 100644
--- a/kernel/rcu/tree.h
+++ b/kernel/rcu/tree.h
@@ -687,18 +687,6 @@ static inline void rcu_nocb_q_lengths(struct rcu_data *rdp, long *ql, long *qll)
 #endif /* #ifdef CONFIG_RCU_TRACE */
 
 /*
- * Place this after a lock-acquisition primitive to guarantee that
- * an UNLOCK+LOCK pair act as a full barrier.  This guarantee applies
- * if the UNLOCK and LOCK are executed by the same CPU or if the
- * UNLOCK and LOCK operate on the same lock variable.
- */
-#ifdef CONFIG_PPC
-#define smp_mb__after_unlock_lock()	smp_mb()  /* Full ordering for lock. */
-#else /* #ifdef CONFIG_PPC */
-#define smp_mb__after_unlock_lock()	do { } while (0)
-#endif /* #else #ifdef CONFIG_PPC */
-
-/*
  * Wrappers for the rcu_node::lock acquire and release.
  *
  * Because the rcu_nodes form a tree, the tree traversal locking will observe
-- 
2.7.4