diff --git a/arch/tile/include/asm/atomic_32.h b/arch/tile/include/asm/atomic_32.h
index 1ad4a1f..1b109fa 100644
--- a/arch/tile/include/asm/atomic_32.h
+++ b/arch/tile/include/asm/atomic_32.h
@@ -169,16 +169,6 @@
 #define atomic64_dec_and_test(v)	(atomic64_dec_return((v)) == 0)
 #define atomic64_inc_not_zero(v)	atomic64_add_unless((v), 1LL, 0LL)
 
-/*
- * We need to barrier before modifying the word, since the _atomic_xxx()
- * routines just tns the lock and then read/modify/write of the word.
- * But after the word is updated, the routine issues an "mf" before returning,
- * and since it's a function call, we don't even need a compiler barrier.
- */
-#define smp_mb__before_atomic_dec()	smp_mb()
-#define smp_mb__before_atomic_inc()	smp_mb()
-#define smp_mb__after_atomic_dec()	do { } while (0)
-#define smp_mb__after_atomic_inc()	do { } while (0)
 
 #endif /* !__ASSEMBLY__ */
 
diff --git a/arch/tile/include/asm/atomic_64.h b/arch/tile/include/asm/atomic_64.h
index ad220ee..7b11c5f 100644
--- a/arch/tile/include/asm/atomic_64.h
+++ b/arch/tile/include/asm/atomic_64.h
@@ -105,12 +105,6 @@
 
 #define atomic64_inc_not_zero(v)	atomic64_add_unless((v), 1, 0)
 
-/* Atomic dec and inc don't implement barrier, so provide them if needed. */
-#define smp_mb__before_atomic_dec()	smp_mb()
-#define smp_mb__after_atomic_dec()	smp_mb()
-#define smp_mb__before_atomic_inc()	smp_mb()
-#define smp_mb__after_atomic_inc()	smp_mb()
-
 /* Define this to indicate that cmpxchg is an efficient operation. */
 #define __HAVE_ARCH_CMPXCHG
 
diff --git a/arch/tile/include/asm/barrier.h b/arch/tile/include/asm/barrier.h
index b5a05d0..96a42ae 100644
--- a/arch/tile/include/asm/barrier.h
+++ b/arch/tile/include/asm/barrier.h
@@ -72,6 +72,20 @@
 #define mb()		fast_mb()
 #define iob()		fast_iob()
 
+#ifndef __tilegx__ /* 32 bit */
+/*
+ * We need to barrier before modifying the word, since the _atomic_xxx()
+ * routines just tns the lock and then read/modify/write of the word.
+ * But after the word is updated, the routine issues an "mf" before returning,
+ * and since it's a function call, we don't even need a compiler barrier.
+ */
+#define smp_mb__before_atomic()	smp_mb()
+#define smp_mb__after_atomic()	do { } while (0)
+#else /* 64 bit */
+#define smp_mb__before_atomic()	smp_mb()
+#define smp_mb__after_atomic()	smp_mb()
+#endif
+
 #include <asm-generic/barrier.h>
 
 #endif /* !__ASSEMBLY__ */
diff --git a/arch/tile/include/asm/bitops.h b/arch/tile/include/asm/bitops.h
index d5a2068..20caa34 100644
--- a/arch/tile/include/asm/bitops.h
+++ b/arch/tile/include/asm/bitops.h
@@ -17,6 +17,7 @@
 #define _ASM_TILE_BITOPS_H
 
 #include <linux/types.h>
+#include <asm/barrier.h>
 
 #ifndef _LINUX_BITOPS_H
 #error only <linux/bitops.h> can be included directly
diff --git a/arch/tile/include/asm/bitops_32.h b/arch/tile/include/asm/bitops_32.h
index 386865ad..bbf7b66 100644
--- a/arch/tile/include/asm/bitops_32.h
+++ b/arch/tile/include/asm/bitops_32.h
@@ -49,8 +49,8 @@
  * restricted to acting on a single-word quantity.
  *
  * clear_bit() may not contain a memory barrier, so if it is used for
- * locking purposes, you should call smp_mb__before_clear_bit() and/or
- * smp_mb__after_clear_bit() to ensure changes are visible on other cpus.
+ * locking purposes, you should call smp_mb__before_atomic() and/or
+ * smp_mb__after_atomic() to ensure changes are visible on other cpus.
  */
 static inline void clear_bit(unsigned nr, volatile unsigned long *addr)
 {
@@ -121,10 +121,6 @@
 	return (_atomic_xor(addr, mask) & mask) != 0;
 }
 
-/* See discussion at smp_mb__before_atomic_dec() in <asm/atomic_32.h>. */
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	do {} while (0)
-
 #include <asm-generic/bitops/ext2-atomic.h>
 
 #endif /* _ASM_TILE_BITOPS_32_H */
diff --git a/arch/tile/include/asm/bitops_64.h b/arch/tile/include/asm/bitops_64.h
index ad34cd0..bb1a292 100644
--- a/arch/tile/include/asm/bitops_64.h
+++ b/arch/tile/include/asm/bitops_64.h
@@ -32,10 +32,6 @@
 	__insn_fetchand((void *)(addr + nr / BITS_PER_LONG), ~mask);
 }
 
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-
-
 static inline void change_bit(unsigned nr, volatile unsigned long *addr)
 {
 	unsigned long mask = (1UL << (nr % BITS_PER_LONG));