locking/xchg/alpha: Fix xchg() and cmpxchg() memory ordering bugs

Successful RMW operations are supposed to be fully ordered, but
Alpha's xchg() and cmpxchg() do not meet this requirement.

Will Deacon noticed the bug:

  > So MP using xchg:
  >
  > WRITE_ONCE(x, 1)
  > xchg(y, 1)
  >
  > smp_load_acquire(y) == 1
  > READ_ONCE(x) == 0
  >
  > would be allowed.

... which thus violates the above requirement.

Fix it by adding a leading smp_mb() to the xchg() and cmpxchg() implementations.
Reported-by: NWill Deacon <will.deacon@arm.com>
Signed-off-by: NAndrea Parri <parri.andrea@gmail.com>
Acked-by: NPaul E. McKenney <paulmck@linux.vnet.ibm.com>
Cc: Alan Stern <stern@rowland.harvard.edu>
Cc: Andrew Morton <akpm@linux-foundation.org>
Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru>
Cc: Linus Torvalds <torvalds@linux-foundation.org>
Cc: Matt Turner <mattst88@gmail.com>
Cc: Peter Zijlstra <peterz@infradead.org>
Cc: Richard Henderson <rth@twiddle.net>
Cc: Thomas Gleixner <tglx@linutronix.de>
Cc: linux-alpha@vger.kernel.org
Link: http://lkml.kernel.org/r/1519291488-5752-1-git-send-email-parri.andrea@gmail.comSigned-off-by: NIngo Molnar <mingo@kernel.org>

arch/alpha/include/asm/xchg.h

@@ -12,6 +12,10 @@
  * Atomic exchange.
  * Since it can be used to implement critical sections
  * it must clobber "memory" (also for interrupts in UP).
+ *
+ * The leading and the trailing memory barriers guarantee that these
+ * operations are fully ordered.
+ *
  */
 
 static inline unsigned long
@@ -19,6 +23,7 @@ ____xchg(_u8, volatile char *m, unsigned long val)
 {
 	unsigned long ret, tmp, addr64;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"	andnot	%4,7,%3\n"
 	"	insbl	%1,%4,%1\n"
@@ -43,6 +48,7 @@ ____xchg(_u16, volatile short *m, unsigned long val)
 {
 	unsigned long ret, tmp, addr64;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"	andnot	%4,7,%3\n"
 	"	inswl	%1,%4,%1\n"
@@ -67,6 +73,7 @@ ____xchg(_u32, volatile int *m, unsigned long val)
 {
 	unsigned long dummy;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"1:	ldl_l %0,%4\n"
 	"	bis $31,%3,%1\n"
@@ -87,6 +94,7 @@ ____xchg(_u64, volatile long *m, unsigned long val)
 {
 	unsigned long dummy;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"1:	ldq_l %0,%4\n"
 	"	bis $31,%3,%1\n"
@@ -128,9 +136,12 @@ ____xchg(, volatile void *ptr, unsigned long x, int size)
  * store NEW in MEM.  Return the initial value in MEM.  Success is
  * indicated by comparing RETURN with OLD.
  *
- * The memory barrier is placed in SMP unconditionally, in order to
- * guarantee that dependency ordering is preserved when a dependency
- * is headed by an unsuccessful operation.
+ * The leading and the trailing memory barriers guarantee that these
+ * operations are fully ordered.
+ *
+ * The trailing memory barrier is placed in SMP unconditionally, in
+ * order to guarantee that dependency ordering is preserved when a
+ * dependency is headed by an unsuccessful operation.
  */
 
 static inline unsigned long
@@ -138,6 +149,7 @@ ____cmpxchg(_u8, volatile char *m, unsigned char old, unsigned char new)
 {
 	unsigned long prev, tmp, cmp, addr64;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"	andnot	%5,7,%4\n"
 	"	insbl	%1,%5,%1\n"
@@ -165,6 +177,7 @@ ____cmpxchg(_u16, volatile short *m, unsigned short old, unsigned short new)
 {
 	unsigned long prev, tmp, cmp, addr64;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"	andnot	%5,7,%4\n"
 	"	inswl	%1,%5,%1\n"
@@ -192,6 +205,7 @@ ____cmpxchg(_u32, volatile int *m, int old, int new)
 {
 	unsigned long prev, cmp;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"1:	ldl_l %0,%5\n"
 	"	cmpeq %0,%3,%1\n"
@@ -215,6 +229,7 @@ ____cmpxchg(_u64, volatile long *m, unsigned long old, unsigned long new)
 {
 	unsigned long prev, cmp;
 
+	smp_mb();
 	__asm__ __volatile__(
 	"1:	ldq_l %0,%5\n"
 	"	cmpeq %0,%3,%1\n"