[ARM] barriers: improve xchg, bitops and atomic SMP barriers

Mathieu Desnoyers pointed out that the ARM barriers were lacking:

- cmpxchg, xchg and atomic add return need memory barriers on
  architectures which can reorder the relative order in which memory
  read/writes can be seen between CPUs, which seems to include recent
  ARM architectures. Those barriers are currently missing on ARM.

- test_and_xxx_bit were missing SMP barriers.

So put these barriers in.  Provide separate atomic_add/atomic_sub
operations which do not require barriers.

Reported-Reviewed-and-Acked-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>

arch/arm/include/asm/assembler.h

@@ -114,3 +114,16 @@
 	.align	3;				\
 	.long	9999b,9001f;			\
 	.previous
+
+/*
+ * SMP data memory barrier
+ */
+	.macro	smp_dmb
+#ifdef CONFIG_SMP
+#if __LINUX_ARM_ARCH__ >= 7
+	dmb
+#elif __LINUX_ARM_ARCH__ == 6
+	mcr	p15, 0, r0, c7, c10, 5	@ dmb
+#endif
+#endif
+	.endm

arch/arm/include/asm/atomic.h

@@ -44,11 +44,29 @@ static inline void atomic_set(atomic_t *v, int i)
 	: "cc");
 }
 
+static inline void atomic_add(int i, atomic_t *v)
+{
+	unsigned long tmp;
+	int result;
+
+	__asm__ __volatile__("@ atomic_add\n"
+"1:	ldrex	%0, [%2]\n"
+"	add	%0, %0, %3\n"
+"	strex	%1, %0, [%2]\n"
+"	teq	%1, #0\n"
+"	bne	1b"
+	: "=&r" (result), "=&r" (tmp)
+	: "r" (&v->counter), "Ir" (i)
+	: "cc");
+}
+
 static inline int atomic_add_return(int i, atomic_t *v)
 {
 	unsigned long tmp;
 	int result;
 
+	smp_mb();
+
 	__asm__ __volatile__("@ atomic_add_return\n"
 "1:	ldrex	%0, [%2]\n"
 "	add	%0, %0, %3\n"
@@ -59,14 +77,34 @@ static inline int atomic_add_return(int i, atomic_t *v)
 	: "r" (&v->counter), "Ir" (i)
 	: "cc");
 
+	smp_mb();
+
 	return result;
 }
 
+static inline void atomic_sub(int i, atomic_t *v)
+{
+	unsigned long tmp;
+	int result;
+
+	__asm__ __volatile__("@ atomic_sub\n"
+"1:	ldrex	%0, [%2]\n"
+"	sub	%0, %0, %3\n"
+"	strex	%1, %0, [%2]\n"
+"	teq	%1, #0\n"
+"	bne	1b"
+	: "=&r" (result), "=&r" (tmp)
+	: "r" (&v->counter), "Ir" (i)
+	: "cc");
+}
+
 static inline int atomic_sub_return(int i, atomic_t *v)
 {
 	unsigned long tmp;
 	int result;
 
+	smp_mb();
+
 	__asm__ __volatile__("@ atomic_sub_return\n"
 "1:	ldrex	%0, [%2]\n"
 "	sub	%0, %0, %3\n"
@@ -77,6 +115,8 @@ static inline int atomic_sub_return(int i, atomic_t *v)
 	: "r" (&v->counter), "Ir" (i)
 	: "cc");
 
+	smp_mb();
+
 	return result;
 }
 
@@ -84,6 +124,8 @@ static inline int atomic_cmpxchg(atomic_t *ptr, int old, int new)
 {
 	unsigned long oldval, res;
 
+	smp_mb();
+
 	do {
 		__asm__ __volatile__("@ atomic_cmpxchg\n"
 		"ldrex	%1, [%2]\n"
@@ -95,6 +137,8 @@ static inline int atomic_cmpxchg(atomic_t *ptr, int old, int new)
 		    : "cc");
 	} while (res);
 
+	smp_mb();
+
 	return oldval;
 }
 
@@ -135,6 +179,7 @@ static inline int atomic_add_return(int i, atomic_t *v)
 
 	return val;
 }
+#define atomic_add(i, v)	(void) atomic_add_return(i, v)
 
 static inline int atomic_sub_return(int i, atomic_t *v)
 {
@@ -148,6 +193,7 @@ static inline int atomic_sub_return(int i, atomic_t *v)
 
 	return val;
 }
+#define atomic_sub(i, v)	(void) atomic_sub_return(i, v)
 
 static inline int atomic_cmpxchg(atomic_t *v, int old, int new)
 {
@@ -187,10 +233,8 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 }
 #define atomic_inc_not_zero(v) atomic_add_unless((v), 1, 0)
 
-#define atomic_add(i, v)	(void) atomic_add_return(i, v)
+#define atomic_inc(v)		atomic_add(1, v)
-#define atomic_inc(v)		(void) atomic_add_return(1, v)
+#define atomic_dec(v)		atomic_sub(1, v)
-#define atomic_sub(i, v)	(void) atomic_sub_return(i, v)
-#define atomic_dec(v)		(void) atomic_sub_return(1, v)
 
 #define atomic_inc_and_test(v)	(atomic_add_return(1, v) == 0)
 #define atomic_dec_and_test(v)	(atomic_sub_return(1, v) == 0)
@@ -200,11 +244,10 @@ static inline int atomic_add_unless(atomic_t *v, int a, int u)
 
 #define atomic_add_negative(i,v) (atomic_add_return(i, v) < 0)
 
-/* Atomic operations are already serializing on ARM */
+#define smp_mb__before_atomic_dec()	smp_mb()
-#define smp_mb__before_atomic_dec()	barrier()
+#define smp_mb__after_atomic_dec()	smp_mb()
-#define smp_mb__after_atomic_dec()	barrier()
+#define smp_mb__before_atomic_inc()	smp_mb()
-#define smp_mb__before_atomic_inc()	barrier()
+#define smp_mb__after_atomic_inc()	smp_mb()
-#define smp_mb__after_atomic_inc()	barrier()
 
 #include <asm-generic/atomic.h>
 #endif

arch/arm/include/asm/system.h

@@ -248,6 +248,8 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr, int size
 	unsigned int tmp;
 #endif
 
+	smp_mb();
+
 	switch (size) {
 #if __LINUX_ARM_ARCH__ >= 6
 	case 1:
@@ -307,6 +309,7 @@ static inline unsigned long __xchg(unsigned long x, volatile void *ptr, int size
 		__bad_xchg(ptr, size), ret = 0;
 		break;
 	}
+	smp_mb();
 
 	return ret;
 }

arch/arm/kernel/entry-armv.S

@@ -815,10 +815,7 @@ __kuser_helper_start:
  */
 
 __kuser_memory_barrier:				@ 0xffff0fa0
-
+	smp_dmb
-#if __LINUX_ARM_ARCH__ >= 6 && defined(CONFIG_SMP)
-	mcr	p15, 0, r0, c7, c10, 5	@ dmb
-#endif
 	usr_ret	lr
 
 	.align	5

arch/arm/lib/bitops.h

@@ -18,12 +18,14 @@
 	mov	r2, #1
 	add	r1, r1, r0, lsr #3	@ Get byte offset
 	mov	r3, r2, lsl r3		@ create mask
+	smp_dmb
 1:	ldrexb	r2, [r1]
 	ands	r0, r2, r3		@ save old value of bit
 	\instr	r2, r2, r3			@ toggle bit
 	strexb	ip, r2, [r1]
 	cmp	ip, #0
 	bne	1b
+	smp_dmb
 	cmp	r0, #0
 	movne	r0, #1
 2:	mov	pc, lr