Merge branch 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip into next

Pull core locking updates from Ingo Molnar:
 "The main changes in this cycle were:

   - reduced/streamlined smp_mb__*() interface that allows more usecases
     and makes the existing ones less buggy, especially in rarer
     architectures

   - add rwsem implementation comments

   - bump up lockdep limits"

* 'locking-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tip/tip: (33 commits)
  rwsem: Add comments to explain the meaning of the rwsem's count field
  lockdep: Increase static allocations
  arch: Mass conversion of smp_mb__*()
  arch,doc: Convert smp_mb__*()
  arch,xtensa: Convert smp_mb__*()
  arch,x86: Convert smp_mb__*()
  arch,tile: Convert smp_mb__*()
  arch,sparc: Convert smp_mb__*()
  arch,sh: Convert smp_mb__*()
  arch,score: Convert smp_mb__*()
  arch,s390: Convert smp_mb__*()
  arch,powerpc: Convert smp_mb__*()
  arch,parisc: Convert smp_mb__*()
  arch,openrisc: Convert smp_mb__*()
  arch,mn10300: Convert smp_mb__*()
  arch,mips: Convert smp_mb__*()
  arch,metag: Convert smp_mb__*()
  arch,m68k: Convert smp_mb__*()
  arch,m32r: Convert smp_mb__*()
  arch,ia64: Convert smp_mb__*()
  ...

Documentation/atomic_ops.txt

@@ -285,15 +285,13 @@ If a caller requires memory barrier semantics around an atomic_t
 operation which does not return a value, a set of interfaces are
 defined which accomplish this:
 
-	void smp_mb__before_atomic_dec(void);
-	void smp_mb__after_atomic_dec(void);
-	void smp_mb__before_atomic_inc(void);
-	void smp_mb__after_atomic_inc(void);
+	void smp_mb__before_atomic(void);
+	void smp_mb__after_atomic(void);
 
-For example, smp_mb__before_atomic_dec() can be used like so:
+For example, smp_mb__before_atomic() can be used like so:
 
 	obj->dead = 1;
-	smp_mb__before_atomic_dec();
+	smp_mb__before_atomic();
 	atomic_dec(&obj->ref_count);
 
 It makes sure that all memory operations preceding the atomic_dec()
@@ -302,15 +300,10 @@ operation.  In the above example, it guarantees that the assignment of
 "1" to obj->dead will be globally visible to other cpus before the
 atomic counter decrement.
 
-Without the explicit smp_mb__before_atomic_dec() call, the
+Without the explicit smp_mb__before_atomic() call, the
 implementation could legally allow the atomic counter update visible
 to other cpus before the "obj->dead = 1;" assignment.
 
-The other three interfaces listed are used to provide explicit
-ordering with respect to memory operations after an atomic_dec() call
-(smp_mb__after_atomic_dec()) and around atomic_inc() calls
-(smp_mb__{before,after}_atomic_inc()).
-
 A missing memory barrier in the cases where they are required by the
 atomic_t implementation above can have disastrous results.  Here is
 an example, which follows a pattern occurring frequently in the Linux
@@ -487,12 +480,12 @@ Finally there is the basic operation:
 Which returns a boolean indicating if bit "nr" is set in the bitmask
 pointed to by "addr".
 
-If explicit memory barriers are required around clear_bit() (which
-does not return a value, and thus does not need to provide memory
-barrier semantics), two interfaces are provided:
+If explicit memory barriers are required around {set,clear}_bit() (which do
+not return a value, and thus does not need to provide memory barrier
+semantics), two interfaces are provided:
 
-	void smp_mb__before_clear_bit(void);
-	void smp_mb__after_clear_bit(void);
+	void smp_mb__before_atomic(void);
+	void smp_mb__after_atomic(void);
 
 They are used as follows, and are akin to their atomic_t operation
 brothers:
@@ -500,13 +493,13 @@ brothers:
 	/* All memory operations before this call will
 	 * be globally visible before the clear_bit().
 	 */
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit( ... );
 
 	/* The clear_bit() will be visible before all
 	 * subsequent memory operations.
 	 */
-	 smp_mb__after_clear_bit();
+	 smp_mb__after_atomic();
 
 There are two special bitops with lock barrier semantics (acquire/release,
 same as spinlocks). These operate in the same way as their non-_lock/unlock

Documentation/memory-barriers.txt

@@ -1583,20 +1583,21 @@ There are some more advanced barrier functions:
      insert anything more than a compiler barrier in a UP compilation.
 
 
- (*) smp_mb__before_atomic_dec();
- (*) smp_mb__after_atomic_dec();
- (*) smp_mb__before_atomic_inc();
- (*) smp_mb__after_atomic_inc();
+ (*) smp_mb__before_atomic();
+ (*) smp_mb__after_atomic();
 
-     These are for use with atomic add, subtract, increment and decrement
-     functions that don't return a value, especially when used for reference
-     counting.  These functions do not imply memory barriers.
+     These are for use with atomic (such as add, subtract, increment and
+     decrement) functions that don't return a value, especially when used for
+     reference counting.  These functions do not imply memory barriers.
+
+     These are also used for atomic bitop functions that do not return a
+     value (such as set_bit and clear_bit).
 
      As an example, consider a piece of code that marks an object as being dead
      and then decrements the object's reference count:
 
 	obj->dead = 1;
-	smp_mb__before_atomic_dec();
+	smp_mb__before_atomic();
 	atomic_dec(&obj->ref_count);
 
      This makes sure that the death mark on the object is perceived to be set
@@ -1606,27 +1607,6 @@ There are some more advanced barrier functions:
      operations" subsection for information on where to use these.
 
 
- (*) smp_mb__before_clear_bit(void);
- (*) smp_mb__after_clear_bit(void);
-
-     These are for use similar to the atomic inc/dec barriers.  These are
-     typically used for bitwise unlocking operations, so care must be taken as
-     there are no implicit memory barriers here either.
-
-     Consider implementing an unlock operation of some nature by clearing a
-     locking bit.  The clear_bit() would then need to be barriered like this:
-
-	smp_mb__before_clear_bit();
-	clear_bit( ... );
-
-     This prevents memory operations before the clear leaking to after it.  See
-     the subsection on "Locking Functions" with reference to RELEASE operation
-     implications.
-
-     See Documentation/atomic_ops.txt for more information.  See the "Atomic
-     operations" subsection for information on where to use these.
-
-
 MMIO WRITE BARRIER
 ------------------
 
@@ -2283,11 +2263,11 @@ operations:
 	change_bit();
 
 With these the appropriate explicit memory barrier should be used if necessary
-(smp_mb__before_clear_bit() for instance).
+(smp_mb__before_atomic() for instance).
 
 
 The following also do _not_ imply memory barriers, and so may require explicit
-memory barriers under some circumstances (smp_mb__before_atomic_dec() for
+memory barriers under some circumstances (smp_mb__before_atomic() for
 instance):
 
 	atomic_add();

arch/alpha/include/asm/atomic.h

@@ -292,9 +292,4 @@ static inline long atomic64_dec_if_positive(atomic64_t *v)
 #define atomic_dec(v) atomic_sub(1,(v))
 #define atomic64_dec(v) atomic64_sub(1,(v))
 
-#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()
-
 #endif /* _ALPHA_ATOMIC_H */

arch/alpha/include/asm/bitops.h

@@ -53,9 +53,6 @@ __set_bit(unsigned long nr, volatile void * addr)
 	*m |= 1 << (nr & 31);
 }
 
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-
 static inline void
 clear_bit(unsigned long nr, volatile void * addr)
 {

arch/arc/include/asm/atomic.h

@@ -190,11 +190,6 @@ static inline void atomic_clear_mask(unsigned long mask, unsigned long *addr)
 
 #endif /* !CONFIG_ARC_HAS_LLSC */
 
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 /**
  * __atomic_add_unless - add unless the number is a given value
  * @v: pointer of type atomic_t

arch/arc/include/asm/bitops.h

@@ -19,6 +19,7 @@
 
 #include <linux/types.h>
 #include <linux/compiler.h>
+#include <asm/barrier.h>
 
 /*
  * Hardware assisted read-modify-write using ARC700 LLOCK/SCOND insns.
@@ -496,10 +497,6 @@ static inline __attribute__ ((const)) int __ffs(unsigned long word)
  */
 #define ffz(x)	__ffs(~(x))
 
-/* TODO does this affect uni-processor code */
-#define smp_mb__before_clear_bit()  barrier()
-#define smp_mb__after_clear_bit()   barrier()
-
 #include <asm-generic/bitops/hweight.h>
 #include <asm-generic/bitops/fls64.h>
 #include <asm-generic/bitops/sched.h>

arch/arm/include/asm/atomic.h

@@ -241,11 +241,6 @@ 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)
 
-#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()
-
 #ifndef CONFIG_GENERIC_ATOMIC64
 typedef struct {
 	long long counter;

arch/arm/include/asm/barrier.h

@@ -79,5 +79,8 @@ do {									\
 
 #define set_mb(var, value)	do { var = value; smp_mb(); } while (0)
 
+#define smp_mb__before_atomic()	smp_mb()
+#define smp_mb__after_atomic()	smp_mb()
+
 #endif /* !__ASSEMBLY__ */
 #endif /* __ASM_BARRIER_H */

arch/arm/include/asm/bitops.h

@@ -25,9 +25,7 @@
 
 #include <linux/compiler.h>
 #include <linux/irqflags.h>
-
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
+#include <asm/barrier.h>
 
 /*
  * These functions are the basis of our bit ops.

arch/arm64/include/asm/atomic.h

@@ -152,11 +152,6 @@ 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)
 
-#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()
-
 /*
  * 64-bit atomic operations.
  */

arch/arm64/include/asm/barrier.h

@@ -98,6 +98,9 @@ do {									\
 #define set_mb(var, value)	do { var = value; smp_mb(); } while (0)
 #define nop()		asm volatile("nop");
 
+#define smp_mb__before_atomic()	smp_mb()
+#define smp_mb__after_atomic()	smp_mb()
+
 #endif	/* __ASSEMBLY__ */
 
 #endif	/* __ASM_BARRIER_H */

arch/arm64/include/asm/bitops.h

@@ -17,17 +17,8 @@
 #define __ASM_BITOPS_H
 
 #include <linux/compiler.h>
-
 #include <asm/barrier.h>
 
-/*
- * clear_bit may not imply a memory barrier
- */
-#ifndef smp_mb__before_clear_bit
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-#endif
-
 #ifndef _LINUX_BITOPS_H
 #error only <linux/bitops.h> can be included directly
 #endif

arch/avr32/include/asm/atomic.h

@@ -183,9 +183,4 @@ static inline int atomic_sub_if_positive(int i, atomic_t *v)
 
 #define atomic_dec_if_positive(v) atomic_sub_if_positive(1, v)
 
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /*  __ASM_AVR32_ATOMIC_H */

arch/avr32/include/asm/bitops.h

@@ -13,12 +13,7 @@
 #endif
 
 #include <asm/byteorder.h>
-
-/*
- * clear_bit() doesn't provide any barrier for the compiler
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
+#include <asm/barrier.h>
 
 /*
  * set_bit - Atomically set a bit in memory
@@ -67,7 +62,7 @@ static inline void set_bit(int nr, volatile void * addr)
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static inline void clear_bit(int nr, volatile void * addr)

arch/blackfin/include/asm/barrier.h

@@ -27,6 +27,9 @@
 
 #endif /* !CONFIG_SMP */
 
+#define smp_mb__before_atomic()	barrier()
+#define smp_mb__after_atomic()	barrier()
+
 #include <asm-generic/barrier.h>
 
 #endif /* _BLACKFIN_BARRIER_H */

arch/blackfin/include/asm/bitops.h

@@ -27,21 +27,17 @@
 
 #include <asm-generic/bitops/ext2-atomic.h>
 
+#include <asm/barrier.h>
+
 #ifndef CONFIG_SMP
 #include <linux/irqflags.h>
-
 /*
  * clear_bit may not imply a memory barrier
  */
-#ifndef smp_mb__before_clear_bit
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-#endif
 #include <asm-generic/bitops/atomic.h>
 #include <asm-generic/bitops/non-atomic.h>
 #else
 
-#include <asm/barrier.h>
 #include <asm/byteorder.h>	/* swab32 */
 #include <linux/linkage.h>
 
@@ -101,12 +97,6 @@ static inline int test_and_change_bit(int nr, volatile unsigned long *addr)
 	return __raw_bit_test_toggle_asm(a, nr & 0x1f);
 }
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 #define test_bit __skip_test_bit
 #include <asm-generic/bitops/non-atomic.h>
 #undef test_bit

arch/c6x/include/asm/bitops.h

@@ -14,14 +14,8 @@
 #ifdef __KERNEL__
 
 #include <linux/bitops.h>
-
 #include <asm/byteorder.h>
-
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit() barrier()
-#define smp_mb__after_clear_bit()  barrier()
+#include <asm/barrier.h>
 
 /*
  * We are lucky, DSP is perfect for bitops: do it in 3 cycles

arch/cris/include/asm/atomic.h

@@ -7,6 +7,8 @@
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
 #include <arch/atomic.h>
+#include <arch/system.h>
+#include <asm/barrier.h>
 
 /*
  * Atomic operations that C can't guarantee us.  Useful for
@@ -151,10 +153,4 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u)
 	return ret;
 }
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()    barrier()
-#define smp_mb__after_atomic_dec()     barrier()
-#define smp_mb__before_atomic_inc()    barrier()
-#define smp_mb__after_atomic_inc()     barrier()
-
 #endif

arch/cris/include/asm/bitops.h

@@ -21,6 +21,7 @@
 #include <arch/bitops.h>
 #include <linux/atomic.h>
 #include <linux/compiler.h>
+#include <asm/barrier.h>
 
 /*
  * set_bit - Atomically set a bit in memory
@@ -42,7 +43,7 @@
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 
@@ -84,12 +85,6 @@ static inline int test_and_set_bit(int nr, volatile unsigned long *addr)
 	return retval;
 }
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()      barrier()
-#define smp_mb__after_clear_bit()       barrier()
-
 /**
  * test_and_clear_bit - Clear a bit and return its old value
  * @nr: Bit to clear

arch/frv/include/asm/atomic.h

@@ -17,6 +17,7 @@
 #include <linux/types.h>
 #include <asm/spr-regs.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #ifdef CONFIG_SMP
 #error not SMP safe
@@ -29,12 +30,6 @@
  * We do not have SMP systems, so we don't have to deal with that.
  */
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #define ATOMIC_INIT(i)		{ (i) }
 #define atomic_read(v)		(*(volatile int *)&(v)->counter)
 #define atomic_set(v, i)	(((v)->counter) = (i))

arch/frv/include/asm/bitops.h

@@ -25,12 +25,6 @@
 
 #include <asm-generic/bitops/ffz.h>
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 #ifndef CONFIG_FRV_OUTOFLINE_ATOMIC_OPS
 static inline
 unsigned long atomic_test_and_ANDNOT_mask(unsigned long mask, volatile unsigned long *v)

arch/hexagon/include/asm/atomic.h

@@ -24,6 +24,7 @@
 
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #define ATOMIC_INIT(i)		{ (i) }
 
@@ -176,9 +177,4 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u)
 #define atomic_inc_return(v) (atomic_add_return(1, v))
 #define atomic_dec_return(v) (atomic_sub_return(1, v))
 
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif

arch/hexagon/include/asm/bitops.h

@@ -25,12 +25,10 @@
 #include <linux/compiler.h>
 #include <asm/byteorder.h>
 #include <asm/atomic.h>
+#include <asm/barrier.h>
 
 #ifdef __KERNEL__
 
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 /*
  * The offset calculations for these are based on BITS_PER_LONG == 32
  * (i.e. I get to shift by #5-2 (32 bits per long, 4 bytes per access),

arch/ia64/include/asm/atomic.h

@@ -15,6 +15,7 @@
 #include <linux/types.h>
 
 #include <asm/intrinsics.h>
+#include <asm/barrier.h>
 
 
 #define ATOMIC_INIT(i)		{ (i) }
@@ -208,10 +209,4 @@ atomic64_add_negative (__s64 i, atomic64_t *v)
 #define atomic64_inc(v)			atomic64_add(1, (v))
 #define atomic64_dec(v)			atomic64_sub(1, (v))
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* _ASM_IA64_ATOMIC_H */

arch/ia64/include/asm/barrier.h

@@ -55,6 +55,9 @@
 
 #endif
 
+#define smp_mb__before_atomic()	barrier()
+#define smp_mb__after_atomic()	barrier()
+
 /*
  * IA64 GCC turns volatile stores into st.rel and volatile loads into ld.acq no
  * need for asm trickery!

arch/ia64/include/asm/bitops.h

@@ -16,6 +16,7 @@
 #include <linux/compiler.h>
 #include <linux/types.h>
 #include <asm/intrinsics.h>
+#include <asm/barrier.h>
 
 /**
  * set_bit - Atomically set a bit in memory
@@ -65,12 +66,6 @@ __set_bit (int nr, volatile void *addr)
 	*((__u32 *) addr + (nr >> 5)) |= (1 << (nr & 31));
 }
 
-/*
- * clear_bit() has "acquire" semantics.
- */
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	do { /* skip */; } while (0)
-
 /**
  * clear_bit - Clears a bit in memory
  * @nr: Bit to clear
@@ -78,7 +73,7 @@ __set_bit (int nr, volatile void *addr)
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static __inline__ void

arch/ia64/include/uapi/asm/cmpxchg.h

@@ -118,6 +118,15 @@ extern long ia64_cmpxchg_called_with_bad_pointer(void);
 #define cmpxchg_rel(ptr, o, n)	\
 	ia64_cmpxchg(rel, (ptr), (o), (n), sizeof(*(ptr)))
 
+/*
+ * Worse still - early processor implementations actually just ignored
+ * the acquire/release and did a full fence all the time.  Unfortunately
+ * this meant a lot of badly written code that used .acq when they really
+ * wanted .rel became legacy out in the wild - so when we made a cpu
+ * that strictly did the .acq or .rel ... all that code started breaking - so
+ * we had to back-pedal and keep the "legacy" behavior of a full fence :-(
+ */
+
 /* for compatibility with other platforms: */
 #define cmpxchg(ptr, o, n)	cmpxchg_acq((ptr), (o), (n))
 #define cmpxchg64(ptr, o, n)	cmpxchg_acq((ptr), (o), (n))

arch/m32r/include/asm/atomic.h

@@ -13,6 +13,7 @@
 #include <asm/assembler.h>
 #include <asm/cmpxchg.h>
 #include <asm/dcache_clear.h>
+#include <asm/barrier.h>
 
 /*
  * Atomic operations that C can't guarantee us.  Useful for
@@ -308,10 +309,4 @@ static __inline__ void atomic_set_mask(unsigned long  mask, atomic_t *addr)
 	local_irq_restore(flags);
 }
 
-/* Atomic operations are already serializing on m32r */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif	/* _ASM_M32R_ATOMIC_H */

arch/m32r/include/asm/bitops.h

@@ -21,6 +21,7 @@
 #include <asm/byteorder.h>
 #include <asm/dcache_clear.h>
 #include <asm/types.h>
+#include <asm/barrier.h>
 
 /*
  * These have to be done with inline assembly: that way the bit-setting
@@ -73,7 +74,7 @@ static __inline__ void set_bit(int nr, volatile void * addr)
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static __inline__ void clear_bit(int nr, volatile void * addr)
@@ -103,9 +104,6 @@ static __inline__ void clear_bit(int nr, volatile void * addr)
 	local_irq_restore(flags);
 }
 
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 /**
  * change_bit - Toggle a bit in memory
  * @nr: Bit to clear

arch/m68k/include/asm/atomic.h

@@ -4,6 +4,7 @@
 #include <linux/types.h>
 #include <linux/irqflags.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 /*
  * Atomic operations that C can't guarantee us.  Useful for
@@ -209,11 +210,4 @@ static __inline__ int __atomic_add_unless(atomic_t *v, int a, int u)
 	return c;
 }
 
-
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* __ARCH_M68K_ATOMIC __ */

arch/m68k/include/asm/bitops.h

@@ -13,6 +13,7 @@
 #endif
 
 #include <linux/compiler.h>
+#include <asm/barrier.h>
 
 /*
  *	Bit access functions vary across the ColdFire and 68k families.
@@ -67,12 +68,6 @@ static inline void bfset_mem_set_bit(int nr, volatile unsigned long *vaddr)
 #define __set_bit(nr, vaddr)	set_bit(nr, vaddr)
 
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 static inline void bclr_reg_clear_bit(int nr, volatile unsigned long *vaddr)
 {
 	char *p = (char *)vaddr + (nr ^ 31) / 8;

arch/metag/include/asm/atomic.h

@@ -4,6 +4,7 @@
 #include <linux/compiler.h>
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #if defined(CONFIG_METAG_ATOMICITY_IRQSOFF)
 /* The simple UP case. */
@@ -39,11 +40,6 @@
 
 #define atomic_inc_not_zero(v) atomic_add_unless((v), 1, 0)
 
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif
 
 #define atomic_dec_if_positive(v)       atomic_sub_if_positive(1, v)

arch/metag/include/asm/barrier.h

@@ -100,4 +100,7 @@ do {									\
 	___p1;								\
 })
 
+#define smp_mb__before_atomic()	barrier()
+#define smp_mb__after_atomic()	barrier()
+
 #endif /* _ASM_METAG_BARRIER_H */

arch/metag/include/asm/bitops.h

@@ -5,12 +5,6 @@
 #include <asm/barrier.h>
 #include <asm/global_lock.h>
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 #ifdef CONFIG_SMP
 /*
  * These functions are the basis of our bit ops.

arch/mips/include/asm/atomic.h

@@ -761,13 +761,4 @@ static __inline__ int atomic64_add_unless(atomic64_t *v, long a, long u)
 
 #endif /* CONFIG_64BIT */
 
-/*
- * atomic*_return operations are serializing but not the non-*_return
- * versions.
- */
-#define smp_mb__before_atomic_dec()	smp_mb__before_llsc()
-#define smp_mb__after_atomic_dec()	smp_llsc_mb()
-#define smp_mb__before_atomic_inc()	smp_mb__before_llsc()
-#define smp_mb__after_atomic_inc()	smp_llsc_mb()
-
 #endif /* _ASM_ATOMIC_H */

arch/mips/include/asm/barrier.h

@@ -195,4 +195,7 @@ do {									\
 	___p1;								\
 })
 
+#define smp_mb__before_atomic()	smp_mb__before_llsc()
+#define smp_mb__after_atomic()	smp_llsc_mb()
+
 #endif /* __ASM_BARRIER_H */

arch/mips/include/asm/bitops.h

@@ -37,13 +37,6 @@
 #define __EXT		"dext	 "
 #endif
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	smp_mb__before_llsc()
-#define smp_mb__after_clear_bit()	smp_llsc_mb()
-
-
 /*
  * These are the "slower" versions of the functions and are in bitops.c.
  * These functions call raw_local_irq_{save,restore}().
@@ -120,7 +113,7 @@ static inline void set_bit(unsigned long nr, volatile unsigned long *addr)
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static inline void clear_bit(unsigned long nr, volatile unsigned long *addr)
@@ -175,7 +168,7 @@ static inline void clear_bit(unsigned long nr, volatile unsigned long *addr)
  */
 static inline void clear_bit_unlock(unsigned long nr, volatile unsigned long *addr)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(nr, addr);
 }
 

arch/mips/kernel/irq.c

@@ -62,9 +62,9 @@ void __init alloc_legacy_irqno(void)
 
 void free_irqno(unsigned int irq)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(irq, irq_map);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 /*

arch/mn10300/include/asm/atomic.h

@@ -13,6 +13,7 @@
 
 #include <asm/irqflags.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #ifndef CONFIG_SMP
 #include <asm-generic/atomic.h>
@@ -234,12 +235,6 @@ static inline void atomic_set_mask(unsigned long mask, unsigned long *addr)
 #endif
 }
 
-/* Atomic operations are already serializing on MN10300??? */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* __KERNEL__ */
 #endif /* CONFIG_SMP */
 #endif /* _ASM_ATOMIC_H */

arch/mn10300/include/asm/bitops.h

@@ -18,9 +18,7 @@
 #define __ASM_BITOPS_H
 
 #include <asm/cpu-regs.h>
-
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
+#include <asm/barrier.h>
 
 /*
  * set bit

arch/mn10300/mm/tlb-smp.c

@@ -78,9 +78,9 @@ void smp_flush_tlb(void *unused)
 	else
 		local_flush_tlb_page(flush_mm, flush_va);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	cpumask_clear_cpu(cpu_id, &flush_cpumask);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 out:
 	put_cpu();
 }

arch/openrisc/include/asm/bitops.h

@@ -27,14 +27,7 @@
 
 #include <linux/irqflags.h>
 #include <linux/compiler.h>
-
-/*
- * clear_bit may not imply a memory barrier
- */
-#ifndef smp_mb__before_clear_bit
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-#endif
+#include <asm/barrier.h>
 
 #include <asm/bitops/__ffs.h>
 #include <asm-generic/bitops/ffz.h>

arch/parisc/include/asm/atomic.h

@@ -7,6 +7,7 @@
 
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 /*
  * Atomic operations that C can't guarantee us.  Useful for
@@ -143,11 +144,6 @@ static __inline__ int __atomic_add_unless(atomic_t *v, int a, int u)
 
 #define ATOMIC_INIT(i)	{ (i) }
 
-#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()
-
 #ifdef CONFIG_64BIT
 
 #define ATOMIC64_INIT(i) { (i) }

arch/parisc/include/asm/bitops.h

@@ -8,6 +8,7 @@
 #include <linux/compiler.h>
 #include <asm/types.h>		/* for BITS_PER_LONG/SHIFT_PER_LONG */
 #include <asm/byteorder.h>
+#include <asm/barrier.h>
 #include <linux/atomic.h>
 
 /*
@@ -19,9 +20,6 @@
 #define CHOP_SHIFTCOUNT(x) (((unsigned long) (x)) & (BITS_PER_LONG - 1))
 
 
-#define smp_mb__before_clear_bit()      smp_mb()
-#define smp_mb__after_clear_bit()       smp_mb()
-
 /* See http://marc.theaimsgroup.com/?t=108826637900003 for discussion
  * on use of volatile and __*_bit() (set/clear/change):
  *	*_bit() want use of volatile.

arch/powerpc/include/asm/atomic.h

@@ -8,6 +8,7 @@
 #ifdef __KERNEL__
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #define ATOMIC_INIT(i)		{ (i) }
 
@@ -270,11 +271,6 @@ static __inline__ int atomic_dec_if_positive(atomic_t *v)
 }
 #define atomic_dec_if_positive atomic_dec_if_positive
 
-#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()
-
 #ifdef __powerpc64__
 
 #define ATOMIC64_INIT(i)	{ (i) }

arch/powerpc/include/asm/barrier.h

@@ -84,4 +84,7 @@ do {									\
 	___p1;								\
 })
 
+#define smp_mb__before_atomic()     smp_mb()
+#define smp_mb__after_atomic()      smp_mb()
+
 #endif /* _ASM_POWERPC_BARRIER_H */

arch/powerpc/include/asm/bitops.h

@@ -51,11 +51,7 @@
 #define PPC_BIT(bit)		(1UL << PPC_BITLSHIFT(bit))
 #define PPC_BITMASK(bs, be)	((PPC_BIT(bs) - PPC_BIT(be)) | PPC_BIT(bs))
 
-/*
- * clear_bit doesn't imply a memory barrier
- */
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
+#include <asm/barrier.h>
 
 /* Macro for generating the ***_bits() functions */
 #define DEFINE_BITOP(fn, op, prefix)		\

arch/powerpc/kernel/crash.c

@@ -81,7 +81,7 @@ void crash_ipi_callback(struct pt_regs *regs)
 	}
 
 	atomic_inc(&cpus_in_crash);
-	smp_mb__after_atomic_inc();
+	smp_mb__after_atomic();
 
 	/*
 	 * Starting the kdump boot.

arch/s390/include/asm/atomic.h

@@ -412,9 +412,4 @@ static inline long long atomic64_dec_if_positive(atomic64_t *v)
 #define atomic64_dec_and_test(_v)	(atomic64_sub_return(1, _v) == 0)
 #define atomic64_inc_not_zero(v)	atomic64_add_unless((v), 1, 0)
 
-#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()
-
 #endif /* __ARCH_S390_ATOMIC__  */

arch/s390/include/asm/barrier.h

@@ -27,8 +27,9 @@
 #define smp_rmb()			rmb()
 #define smp_wmb()			wmb()
 #define smp_read_barrier_depends()	read_barrier_depends()
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
+
+#define smp_mb__before_atomic()		smp_mb()
+#define smp_mb__after_atomic()		smp_mb()
 
 #define set_mb(var, value)		do { var = value; mb(); } while (0)
 

arch/score/include/asm/bitops.h

@@ -2,12 +2,7 @@
 #define _ASM_SCORE_BITOPS_H
 
 #include <asm/byteorder.h> /* swab32 */
-
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
+#include <asm/barrier.h>
 
 #include <asm-generic/bitops.h>
 #include <asm-generic/bitops/__fls.h>

arch/sh/include/asm/atomic.h

@@ -10,6 +10,7 @@
 #include <linux/compiler.h>
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #define ATOMIC_INIT(i)	{ (i) }
 
@@ -62,9 +63,4 @@ static inline int __atomic_add_unless(atomic_t *v, int a, int u)
 	return c;
 }
 
-#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()
-
 #endif /* __ASM_SH_ATOMIC_H */

arch/sh/include/asm/bitops.h

@@ -9,6 +9,7 @@
 
 /* For __swab32 */
 #include <asm/byteorder.h>
+#include <asm/barrier.h>
 
 #ifdef CONFIG_GUSA_RB
 #include <asm/bitops-grb.h>
@@ -22,12 +23,6 @@
 #include <asm-generic/bitops/non-atomic.h>
 #endif
 
-/*
- * clear_bit() doesn't provide any barrier for the compiler.
- */
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
-
 #ifdef CONFIG_SUPERH32
 static inline unsigned long ffz(unsigned long word)
 {

arch/sparc/include/asm/atomic_32.h

@@ -14,6 +14,7 @@
 #include <linux/types.h>
 
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 #include <asm-generic/atomic64.h>
 
 
@@ -52,10 +53,4 @@ extern void atomic_set(atomic_t *, int);
 #define atomic_dec_and_test(v) (atomic_dec_return(v) == 0)
 #define atomic_sub_and_test(i, v) (atomic_sub_return(i, v) == 0)
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* !(__ARCH_SPARC_ATOMIC__) */

arch/sparc/include/asm/atomic_64.h

@@ -9,6 +9,7 @@
 
 #include <linux/types.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #define ATOMIC_INIT(i)		{ (i) }
 #define ATOMIC64_INIT(i)	{ (i) }
@@ -108,10 +109,4 @@ static inline long atomic64_add_unless(atomic64_t *v, long a, long u)
 
 extern long atomic64_dec_if_positive(atomic64_t *v);
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* !(__ARCH_SPARC64_ATOMIC__) */

arch/sparc/include/asm/barrier_64.h

@@ -68,4 +68,7 @@ do {									\
 	___p1;								\
 })
 
+#define smp_mb__before_atomic()	barrier()
+#define smp_mb__after_atomic()	barrier()
+
 #endif /* !(__SPARC64_BARRIER_H) */

arch/sparc/include/asm/bitops_32.h

@@ -90,9 +90,6 @@ static inline void change_bit(unsigned long nr, volatile unsigned long *addr)
 
 #include <asm-generic/bitops/non-atomic.h>
 
-#define smp_mb__before_clear_bit()	do { } while(0)
-#define smp_mb__after_clear_bit()	do { } while(0)
-
 #include <asm-generic/bitops/ffz.h>
 #include <asm-generic/bitops/__ffs.h>
 #include <asm-generic/bitops/sched.h>

arch/sparc/include/asm/bitops_64.h

@@ -13,6 +13,7 @@
 
 #include <linux/compiler.h>
 #include <asm/byteorder.h>
+#include <asm/barrier.h>
 
 extern int test_and_set_bit(unsigned long nr, volatile unsigned long *addr);
 extern int test_and_clear_bit(unsigned long nr, volatile unsigned long *addr);
@@ -23,9 +24,6 @@ extern void change_bit(unsigned long nr, volatile unsigned long *addr);
 
 #include <asm-generic/bitops/non-atomic.h>
 
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 #include <asm-generic/bitops/fls.h>
 #include <asm-generic/bitops/__fls.h>
 #include <asm-generic/bitops/fls64.h>

arch/tile/include/asm/atomic_32.h

@@ -169,16 +169,6 @@ static inline void atomic64_set(atomic64_t *v, long long n)
 #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__ */
 

arch/tile/include/asm/atomic_64.h

@@ -105,12 +105,6 @@ static inline long atomic64_add_unless(atomic64_t *v, long a, long u)
 
 #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
 

arch/tile/include/asm/barrier.h

@@ -72,6 +72,20 @@ mb_incoherent(void)
 #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__ */

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

arch/tile/include/asm/bitops_32.h

@@ -49,8 +49,8 @@ static inline void set_bit(unsigned nr, volatile unsigned long *addr)
  * 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 @@ static inline int test_and_change_bit(unsigned nr,
 	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 */

arch/tile/include/asm/bitops_64.h

@@ -32,10 +32,6 @@ static inline void clear_bit(unsigned nr, volatile unsigned long *addr)
 	__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));

arch/x86/include/asm/atomic.h

@@ -7,6 +7,7 @@
 #include <asm/alternative.h>
 #include <asm/cmpxchg.h>
 #include <asm/rmwcc.h>
+#include <asm/barrier.h>
 
 /*
  * Atomic operations that C can't guarantee us.  Useful for
@@ -243,12 +244,6 @@ static inline void atomic_or_long(unsigned long *v1, unsigned long v2)
 		     : : "r" ((unsigned)(mask)), "m" (*(addr))	\
 		     : "memory")
 
-/* Atomic operations are already serializing on x86 */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #ifdef CONFIG_X86_32
 # include <asm/atomic64_32.h>
 #else

arch/x86/include/asm/barrier.h

@@ -137,6 +137,10 @@ do {									\
 
 #endif
 
+/* Atomic operations are already serializing on x86 */
+#define smp_mb__before_atomic()	barrier()
+#define smp_mb__after_atomic()	barrier()
+
 /*
  * Stop RDTSC speculation. This is needed when you need to use RDTSC
  * (or get_cycles or vread that possibly accesses the TSC) in a defined

arch/x86/include/asm/bitops.h

@@ -15,6 +15,7 @@
 #include <linux/compiler.h>
 #include <asm/alternative.h>
 #include <asm/rmwcc.h>
+#include <asm/barrier.h>
 
 #if BITS_PER_LONG == 32
 # define _BITOPS_LONG_SHIFT 5
@@ -102,7 +103,7 @@ static inline void __set_bit(long nr, volatile unsigned long *addr)
  *
  * clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static __always_inline void
@@ -156,9 +157,6 @@ static inline void __clear_bit_unlock(long nr, volatile unsigned long *addr)
 	__clear_bit(nr, addr);
 }
 
-#define smp_mb__before_clear_bit()	barrier()
-#define smp_mb__after_clear_bit()	barrier()
-
 /**
  * __change_bit - Toggle a bit in memory
  * @nr: the bit to change

arch/x86/include/asm/sync_bitops.h

@@ -41,7 +41,7 @@ static inline void sync_set_bit(long nr, volatile unsigned long *addr)
  *
  * sync_clear_bit() is atomic and may not be reordered.  However, it does
  * 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()
+ * you should call smp_mb__before_atomic() and/or smp_mb__after_atomic()
  * in order to ensure changes are visible on other processors.
  */
 static inline void sync_clear_bit(long nr, volatile unsigned long *addr)

arch/x86/kernel/apic/hw_nmi.c

@@ -57,7 +57,7 @@ void arch_trigger_all_cpu_backtrace(void)
 	}
 
 	clear_bit(0, &backtrace_flag);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 static int __kprobes

arch/xtensa/include/asm/atomic.h

@@ -19,6 +19,7 @@
 #ifdef __KERNEL__
 #include <asm/processor.h>
 #include <asm/cmpxchg.h>
+#include <asm/barrier.h>
 
 #define ATOMIC_INIT(i)	{ (i) }
 
@@ -387,12 +388,6 @@ static inline void atomic_set_mask(unsigned int mask, atomic_t *v)
 #endif
 }
 
-/* Atomic operations are already serializing */
-#define smp_mb__before_atomic_dec()	barrier()
-#define smp_mb__after_atomic_dec()	barrier()
-#define smp_mb__before_atomic_inc()	barrier()
-#define smp_mb__after_atomic_inc()	barrier()
-
 #endif /* __KERNEL__ */
 
 #endif /* _XTENSA_ATOMIC_H */

arch/xtensa/include/asm/barrier.h

@@ -13,6 +13,9 @@
 #define rmb() barrier()
 #define wmb() mb()
 
+#define smp_mb__before_atomic()		barrier()
+#define smp_mb__after_atomic()		barrier()
+
 #include <asm-generic/barrier.h>
 
 #endif /* _XTENSA_SYSTEM_H */

arch/xtensa/include/asm/bitops.h

@@ -21,9 +21,7 @@
 
 #include <asm/processor.h>
 #include <asm/byteorder.h>
-
-#define smp_mb__before_clear_bit()	smp_mb()
-#define smp_mb__after_clear_bit()	smp_mb()
+#include <asm/barrier.h>
 
 #include <asm-generic/bitops/non-atomic.h>
 

block/blk-iopoll.c

@@ -49,7 +49,7 @@ EXPORT_SYMBOL(blk_iopoll_sched);
 void __blk_iopoll_complete(struct blk_iopoll *iop)
 {
 	list_del(&iop->list);
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit_unlock(IOPOLL_F_SCHED, &iop->state);
 }
 EXPORT_SYMBOL(__blk_iopoll_complete);
@@ -161,7 +161,7 @@ EXPORT_SYMBOL(blk_iopoll_disable);
 void blk_iopoll_enable(struct blk_iopoll *iop)
 {
 	BUG_ON(!test_bit(IOPOLL_F_SCHED, &iop->state));
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit_unlock(IOPOLL_F_SCHED, &iop->state);
 }
 EXPORT_SYMBOL(blk_iopoll_enable);

crypto/chainiv.c

@@ -126,7 +126,7 @@ static int async_chainiv_schedule_work(struct async_chainiv_ctx *ctx)
 	int err = ctx->err;
 
 	if (!ctx->queue.qlen) {
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		clear_bit(CHAINIV_STATE_INUSE, &ctx->state);
 
 		if (!ctx->queue.qlen ||

drivers/base/power/domain.c

@@ -105,7 +105,7 @@ static bool genpd_sd_counter_dec(struct generic_pm_domain *genpd)
 static void genpd_sd_counter_inc(struct generic_pm_domain *genpd)
 {
 	atomic_inc(&genpd->sd_count);
-	smp_mb__after_atomic_inc();
+	smp_mb__after_atomic();
 }
 
 static void genpd_acquire_lock(struct generic_pm_domain *genpd)

drivers/cpuidle/coupled.c

@@ -159,7 +159,7 @@ void cpuidle_coupled_parallel_barrier(struct cpuidle_device *dev, atomic_t *a)
 {
 	int n = dev->coupled->online_count;
 
-	smp_mb__before_atomic_inc();
+	smp_mb__before_atomic();
 	atomic_inc(a);
 
 	while (atomic_read(a) < n)

drivers/firewire/ohci.c

@@ -3498,7 +3498,7 @@ static int ohci_flush_iso_completions(struct fw_iso_context *base)
 		}
 
 		clear_bit_unlock(0, &ctx->flushing_completions);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 	}
 
 	tasklet_enable(&ctx->context.tasklet);

drivers/gpu/drm/drm_irq.c

@@ -156,7 +156,7 @@ static void vblank_disable_and_save(struct drm_device *dev, int crtc)
 	 */
 	if ((vblrc > 0) && (abs64(diff_ns) > 1000000)) {
 		atomic_inc(&dev->vblank[crtc].count);
-		smp_mb__after_atomic_inc();
+		smp_mb__after_atomic();
 	}
 
 	/* Invalidate all timestamps while vblank irq's are off. */
@@ -864,9 +864,9 @@ static void drm_update_vblank_count(struct drm_device *dev, int crtc)
 		vblanktimestamp(dev, crtc, tslot) = t_vblank;
 	}
 
-	smp_mb__before_atomic_inc();
+	smp_mb__before_atomic();
 	atomic_add(diff, &dev->vblank[crtc].count);
-	smp_mb__after_atomic_inc();
+	smp_mb__after_atomic();
 }
 
 /**
@@ -1330,9 +1330,9 @@ bool drm_handle_vblank(struct drm_device *dev, int crtc)
 		/* Increment cooked vblank count. This also atomically commits
 		 * the timestamp computed above.
 		 */
-		smp_mb__before_atomic_inc();
+		smp_mb__before_atomic();
 		atomic_inc(&dev->vblank[crtc].count);
-		smp_mb__after_atomic_inc();
+		smp_mb__after_atomic();
 	} else {
 		DRM_DEBUG("crtc %d: Redundant vblirq ignored. diff_ns = %d\n",
 			  crtc, (int) diff_ns);

drivers/gpu/drm/i915/i915_irq.c

@@ -2157,7 +2157,7 @@ static void i915_error_work_func(struct work_struct *work)
 			 * updates before
 			 * the counter increment.
 			 */
-			smp_mb__before_atomic_inc();
+			smp_mb__before_atomic();
 			atomic_inc(&dev_priv->gpu_error.reset_counter);
 
 			kobject_uevent_env(&dev->primary->kdev->kobj,

drivers/md/bcache/bcache.h

@@ -828,7 +828,7 @@ static inline bool cached_dev_get(struct cached_dev *dc)
 		return false;
 
 	/* Paired with the mb in cached_dev_attach */
-	smp_mb__after_atomic_inc();
+	smp_mb__after_atomic();
 	return true;
 }
 

drivers/md/bcache/closure.h

@@ -243,7 +243,7 @@ static inline void set_closure_fn(struct closure *cl, closure_fn *fn,
 	cl->fn = fn;
 	cl->wq = wq;
 	/* between atomic_dec() in closure_put() */
-	smp_mb__before_atomic_dec();
+	smp_mb__before_atomic();
 }
 
 static inline void closure_queue(struct closure *cl)

drivers/md/dm-bufio.c

@@ -607,9 +607,9 @@ static void write_endio(struct bio *bio, int error)
 
 	BUG_ON(!test_bit(B_WRITING, &b->state));
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(B_WRITING, &b->state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	wake_up_bit(&b->state, B_WRITING);
 }
@@ -997,9 +997,9 @@ static void read_endio(struct bio *bio, int error)
 
 	BUG_ON(!test_bit(B_READING, &b->state));
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(B_READING, &b->state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	wake_up_bit(&b->state, B_READING);
 }

drivers/md/dm-snap.c

@@ -642,7 +642,7 @@ static void free_pending_exception(struct dm_snap_pending_exception *pe)
 	struct dm_snapshot *s = pe->snap;
 
 	mempool_free(pe, s->pending_pool);
-	smp_mb__before_atomic_dec();
+	smp_mb__before_atomic();
 	atomic_dec(&s->pending_exceptions_count);
 }
 
@@ -783,7 +783,7 @@ static int init_hash_tables(struct dm_snapshot *s)
 static void merge_shutdown(struct dm_snapshot *s)
 {
 	clear_bit_unlock(RUNNING_MERGE, &s->state_bits);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	wake_up_bit(&s->state_bits, RUNNING_MERGE);
 }
 

drivers/md/dm.c

@@ -2446,7 +2446,7 @@ static void dm_wq_work(struct work_struct *work)
 static void dm_queue_flush(struct mapped_device *md)
 {
 	clear_bit(DMF_BLOCK_IO_FOR_SUSPEND, &md->flags);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	queue_work(md->wq, &md->work);
 }
 

drivers/md/raid5.c

@@ -4400,7 +4400,7 @@ static void raid5_unplug(struct blk_plug_cb *blk_cb, bool from_schedule)
 			 * STRIPE_ON_UNPLUG_LIST clear but the stripe
 			 * is still in our list
 			 */
-			smp_mb__before_clear_bit();
+			smp_mb__before_atomic();
 			clear_bit(STRIPE_ON_UNPLUG_LIST, &sh->state);
 			/*
 			 * STRIPE_ON_RELEASE_LIST could be set here. In that

drivers/media/usb/dvb-usb-v2/dvb_usb_core.c

@@ -399,7 +399,7 @@ static int dvb_usb_stop_feed(struct dvb_demux_feed *dvbdmxfeed)
 
 	/* clear 'streaming' status bit */
 	clear_bit(ADAP_STREAMING, &adap->state_bits);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	wake_up_bit(&adap->state_bits, ADAP_STREAMING);
 skip_feed_stop:
 
@@ -550,7 +550,7 @@ static int dvb_usb_fe_init(struct dvb_frontend *fe)
 err:
 	if (!adap->suspend_resume_active) {
 		clear_bit(ADAP_INIT, &adap->state_bits);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 		wake_up_bit(&adap->state_bits, ADAP_INIT);
 	}
 
@@ -591,7 +591,7 @@ err:
 	if (!adap->suspend_resume_active) {
 		adap->active_fe = -1;
 		clear_bit(ADAP_SLEEP, &adap->state_bits);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 		wake_up_bit(&adap->state_bits, ADAP_SLEEP);
 	}
 

drivers/net/ethernet/broadcom/bnx2x/bnx2x_cmn.c

@@ -2781,7 +2781,7 @@ int bnx2x_nic_load(struct bnx2x *bp, int load_mode)
 
 	case LOAD_OPEN:
 		netif_tx_start_all_queues(bp->dev);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 		break;
 
 	case LOAD_DIAG:
@@ -4939,9 +4939,9 @@ void bnx2x_update_coalesce_sb_index(struct bnx2x *bp, u8 fw_sb_id,
 void bnx2x_schedule_sp_rtnl(struct bnx2x *bp, enum sp_rtnl_flag flag,
 			    u32 verbose)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	set_bit(flag, &bp->sp_rtnl_state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	DP((BNX2X_MSG_SP | verbose), "Scheduling sp_rtnl task [Flag: %d]\n",
 	   flag);
 	schedule_delayed_work(&bp->sp_rtnl_task, 0);

drivers/net/ethernet/broadcom/bnx2x/bnx2x_main.c

@@ -1858,10 +1858,10 @@ void bnx2x_sp_event(struct bnx2x_fastpath *fp, union eth_rx_cqe *rr_cqe)
 		return;
 #endif
 
-	smp_mb__before_atomic_inc();
+	smp_mb__before_atomic();
 	atomic_inc(&bp->cq_spq_left);
 	/* push the change in bp->spq_left and towards the memory */
-	smp_mb__after_atomic_inc();
+	smp_mb__after_atomic();
 
 	DP(BNX2X_MSG_SP, "bp->cq_spq_left %x\n", atomic_read(&bp->cq_spq_left));
 
@@ -1876,11 +1876,11 @@ void bnx2x_sp_event(struct bnx2x_fastpath *fp, union eth_rx_cqe *rr_cqe)
 		 * sp_state is cleared, and this order prevents
 		 * races
 		 */
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		set_bit(BNX2X_AFEX_PENDING_VIFSET_MCP_ACK, &bp->sp_state);
 		wmb();
 		clear_bit(BNX2X_AFEX_FCOE_Q_UPDATE_PENDING, &bp->sp_state);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 
 		/* schedule the sp task as mcp ack is required */
 		bnx2x_schedule_sp_task(bp);
@@ -5272,9 +5272,9 @@ static void bnx2x_after_function_update(struct bnx2x *bp)
 		__clear_bit(RAMROD_COMP_WAIT, &queue_params.ramrod_flags);
 
 		/* mark latest Q bit */
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		set_bit(BNX2X_AFEX_FCOE_Q_UPDATE_PENDING, &bp->sp_state);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 
 		/* send Q update ramrod for FCoE Q */
 		rc = bnx2x_queue_state_change(bp, &queue_params);
@@ -5500,7 +5500,7 @@ next_spqe:
 		spqe_cnt++;
 	} /* for */
 
-	smp_mb__before_atomic_inc();
+	smp_mb__before_atomic();
 	atomic_add(spqe_cnt, &bp->eq_spq_left);
 
 	bp->eq_cons = sw_cons;
@@ -13875,9 +13875,9 @@ static int bnx2x_drv_ctl(struct net_device *dev, struct drv_ctl_info *ctl)
 	case DRV_CTL_RET_L2_SPQ_CREDIT_CMD: {
 		int count = ctl->data.credit.credit_count;
 
-		smp_mb__before_atomic_inc();
+		smp_mb__before_atomic();
 		atomic_add(count, &bp->cq_spq_left);
-		smp_mb__after_atomic_inc();
+		smp_mb__after_atomic();
 		break;
 	}
 	case DRV_CTL_ULP_REGISTER_CMD: {

drivers/net/ethernet/broadcom/bnx2x/bnx2x_sp.c

@@ -258,16 +258,16 @@ static bool bnx2x_raw_check_pending(struct bnx2x_raw_obj *o)
 
 static void bnx2x_raw_clear_pending(struct bnx2x_raw_obj *o)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(o->state, o->pstate);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 static void bnx2x_raw_set_pending(struct bnx2x_raw_obj *o)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	set_bit(o->state, o->pstate);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 /**
@@ -2131,7 +2131,7 @@ static int bnx2x_set_rx_mode_e1x(struct bnx2x *bp,
 
 	/* The operation is completed */
 	clear_bit(p->state, p->pstate);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	return 0;
 }
@@ -3576,16 +3576,16 @@ error_exit1:
 
 static void bnx2x_mcast_clear_sched(struct bnx2x_mcast_obj *o)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(o->sched_state, o->raw.pstate);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 static void bnx2x_mcast_set_sched(struct bnx2x_mcast_obj *o)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	set_bit(o->sched_state, o->raw.pstate);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 static bool bnx2x_mcast_check_sched(struct bnx2x_mcast_obj *o)
@@ -4200,7 +4200,7 @@ int bnx2x_queue_state_change(struct bnx2x *bp,
 		if (rc) {
 			o->next_state = BNX2X_Q_STATE_MAX;
 			clear_bit(pending_bit, pending);
-			smp_mb__after_clear_bit();
+			smp_mb__after_atomic();
 			return rc;
 		}
 
@@ -4288,7 +4288,7 @@ static int bnx2x_queue_comp_cmd(struct bnx2x *bp,
 	wmb();
 
 	clear_bit(cmd, &o->pending);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	return 0;
 }
@@ -5279,7 +5279,7 @@ static inline int bnx2x_func_state_change_comp(struct bnx2x *bp,
 	wmb();
 
 	clear_bit(cmd, &o->pending);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	return 0;
 }
@@ -5926,7 +5926,7 @@ int bnx2x_func_state_change(struct bnx2x *bp,
 		if (rc) {
 			o->next_state = BNX2X_F_STATE_MAX;
 			clear_bit(cmd, pending);
-			smp_mb__after_clear_bit();
+			smp_mb__after_atomic();
 			return rc;
 		}
 

drivers/net/ethernet/broadcom/bnx2x/bnx2x_sriov.c

@@ -1650,9 +1650,9 @@ static
 void bnx2x_vf_handle_filters_eqe(struct bnx2x *bp,
 				 struct bnx2x_virtf *vf)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(BNX2X_FILTER_RX_MODE_PENDING, &vf->filter_state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 }
 
 static void bnx2x_vf_handle_rss_update_eqe(struct bnx2x *bp,
@@ -2990,9 +2990,9 @@ void bnx2x_iov_task(struct work_struct *work)
 
 void bnx2x_schedule_iov_task(struct bnx2x *bp, enum bnx2x_iov_flag flag)
 {
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	set_bit(flag, &bp->iov_task_state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	DP(BNX2X_MSG_IOV, "Scheduling iov task [Flag: %d]\n", flag);
 	queue_delayed_work(bnx2x_iov_wq, &bp->iov_task, 0);
 }

drivers/net/ethernet/broadcom/cnic.c

@@ -436,7 +436,7 @@ static int cnic_offld_prep(struct cnic_sock *csk)
 static int cnic_close_prep(struct cnic_sock *csk)
 {
 	clear_bit(SK_F_CONNECT_START, &csk->flags);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	if (test_and_clear_bit(SK_F_OFFLD_COMPLETE, &csk->flags)) {
 		while (test_and_set_bit(SK_F_OFFLD_SCHED, &csk->flags))
@@ -450,7 +450,7 @@ static int cnic_close_prep(struct cnic_sock *csk)
 static int cnic_abort_prep(struct cnic_sock *csk)
 {
 	clear_bit(SK_F_CONNECT_START, &csk->flags);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	while (test_and_set_bit(SK_F_OFFLD_SCHED, &csk->flags))
 		msleep(1);
@@ -3646,7 +3646,7 @@ static int cnic_cm_destroy(struct cnic_sock *csk)
 
 	csk_hold(csk);
 	clear_bit(SK_F_INUSE, &csk->flags);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 	while (atomic_read(&csk->ref_count) != 1)
 		msleep(1);
 	cnic_cm_cleanup(csk);
@@ -4026,7 +4026,7 @@ static void cnic_cm_process_kcqe(struct cnic_dev *dev, struct kcqe *kcqe)
 			 L4_KCQE_COMPLETION_STATUS_PARITY_ERROR)
 			set_bit(SK_F_HW_ERR, &csk->flags);
 
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		clear_bit(SK_F_OFFLD_SCHED, &csk->flags);
 		cnic_cm_upcall(cp, csk, opcode);
 		break;

drivers/net/ethernet/brocade/bna/bnad.c

@@ -249,7 +249,7 @@ bnad_tx_complete(struct bnad *bnad, struct bna_tcb *tcb)
 	if (likely(test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags)))
 		bna_ib_ack(tcb->i_dbell, sent);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
 
 	return sent;
@@ -1126,7 +1126,7 @@ bnad_tx_cleanup(struct delayed_work *work)
 
 		bnad_txq_cleanup(bnad, tcb);
 
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
 	}
 
@@ -2992,7 +2992,7 @@ bnad_start_xmit(struct sk_buff *skb, struct net_device *netdev)
 			sent = bnad_txcmpl_process(bnad, tcb);
 			if (likely(test_bit(BNAD_TXQ_TX_STARTED, &tcb->flags)))
 				bna_ib_ack(tcb->i_dbell, sent);
-			smp_mb__before_clear_bit();
+			smp_mb__before_atomic();
 			clear_bit(BNAD_TXQ_FREE_SENT, &tcb->flags);
 		} else {
 			netif_stop_queue(netdev);

drivers/net/ethernet/chelsio/cxgb/cxgb2.c

@@ -281,7 +281,7 @@ static int cxgb_close(struct net_device *dev)
 	if (adapter->params.stats_update_period &&
 	    !(adapter->open_device_map & PORT_MASK)) {
 		/* Stop statistics accumulation. */
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 		spin_lock(&adapter->work_lock);   /* sync with update task */
 		spin_unlock(&adapter->work_lock);
 		cancel_mac_stats_update(adapter);

drivers/net/ethernet/chelsio/cxgb3/sge.c

@@ -1379,7 +1379,7 @@ static inline int check_desc_avail(struct adapter *adap, struct sge_txq *q,
 		struct sge_qset *qs = txq_to_qset(q, qid);
 
 		set_bit(qid, &qs->txq_stopped);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 
 		if (should_restart_tx(q) &&
 		    test_and_clear_bit(qid, &qs->txq_stopped))
@@ -1492,7 +1492,7 @@ static void restart_ctrlq(unsigned long data)
 
 	if (!skb_queue_empty(&q->sendq)) {
 		set_bit(TXQ_CTRL, &qs->txq_stopped);
-		smp_mb__after_clear_bit();
+		smp_mb__after_atomic();
 
 		if (should_restart_tx(q) &&
 		    test_and_clear_bit(TXQ_CTRL, &qs->txq_stopped))
@@ -1697,7 +1697,7 @@ again:	reclaim_completed_tx(adap, q, TX_RECLAIM_CHUNK);
 
 		if (unlikely(q->size - q->in_use < ndesc)) {
 			set_bit(TXQ_OFLD, &qs->txq_stopped);
-			smp_mb__after_clear_bit();
+			smp_mb__after_atomic();
 
 			if (should_restart_tx(q) &&
 			    test_and_clear_bit(TXQ_OFLD, &qs->txq_stopped))

drivers/net/ethernet/chelsio/cxgb4/sge.c

@@ -2031,7 +2031,7 @@ static void sge_rx_timer_cb(unsigned long data)
 			struct sge_fl *fl = s->egr_map[id];
 
 			clear_bit(id, s->starving_fl);
-			smp_mb__after_clear_bit();
+			smp_mb__after_atomic();
 
 			if (fl_starving(fl)) {
 				rxq = container_of(fl, struct sge_eth_rxq, fl);

drivers/net/ethernet/chelsio/cxgb4vf/sge.c

@@ -1951,7 +1951,7 @@ static void sge_rx_timer_cb(unsigned long data)
 			struct sge_fl *fl = s->egr_map[id];
 
 			clear_bit(id, s->starving_fl);
-			smp_mb__after_clear_bit();
+			smp_mb__after_atomic();
 
 			/*
 			 * Since we are accessing fl without a lock there's a

drivers/net/ethernet/freescale/gianfar.c

@@ -1798,9 +1798,9 @@ void stop_gfar(struct net_device *dev)
 
 	netif_tx_stop_all_queues(dev);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	set_bit(GFAR_DOWN, &priv->state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	disable_napi(priv);
 
@@ -2043,9 +2043,9 @@ int startup_gfar(struct net_device *ndev)
 
 	gfar_init_tx_rx_base(priv);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(GFAR_DOWN, &priv->state);
-	smp_mb__after_clear_bit();
+	smp_mb__after_atomic();
 
 	/* Start Rx/Tx DMA and enable the interrupts */
 	gfar_start(priv);

drivers/net/ethernet/intel/i40e/i40e_main.c

@@ -4676,7 +4676,7 @@ static void i40e_service_event_complete(struct i40e_pf *pf)
 	BUG_ON(!test_bit(__I40E_SERVICE_SCHED, &pf->state));
 
 	/* flush memory to make sure state is correct before next watchog */
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__I40E_SERVICE_SCHED, &pf->state);
 }
 

drivers/net/ethernet/intel/ixgbe/ixgbe_main.c

@@ -376,7 +376,7 @@ static void ixgbe_service_event_complete(struct ixgbe_adapter *adapter)
 	BUG_ON(!test_bit(__IXGBE_SERVICE_SCHED, &adapter->state));
 
 	/* flush memory to make sure state is correct before next watchdog */
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBE_SERVICE_SCHED, &adapter->state);
 }
 
@@ -4672,7 +4672,7 @@ static void ixgbe_up_complete(struct ixgbe_adapter *adapter)
 	if (hw->mac.ops.enable_tx_laser)
 		hw->mac.ops.enable_tx_laser(hw);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBE_DOWN, &adapter->state);
 	ixgbe_napi_enable_all(adapter);
 
@@ -5568,7 +5568,7 @@ static int ixgbe_resume(struct pci_dev *pdev)
 		e_dev_err("Cannot enable PCI device from suspend\n");
 		return err;
 	}
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBE_DISABLED, &adapter->state);
 	pci_set_master(pdev);
 
@@ -8542,7 +8542,7 @@ static pci_ers_result_t ixgbe_io_slot_reset(struct pci_dev *pdev)
 		e_err(probe, "Cannot re-enable PCI device after reset.\n");
 		result = PCI_ERS_RESULT_DISCONNECT;
 	} else {
-		smp_mb__before_clear_bit();
+		smp_mb__before_atomic();
 		clear_bit(__IXGBE_DISABLED, &adapter->state);
 		adapter->hw.hw_addr = adapter->io_addr;
 		pci_set_master(pdev);

drivers/net/ethernet/intel/ixgbevf/ixgbevf_main.c

@@ -1668,7 +1668,7 @@ static void ixgbevf_up_complete(struct ixgbevf_adapter *adapter)
 
 	spin_unlock_bh(&adapter->mbx_lock);
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBEVF_DOWN, &adapter->state);
 	ixgbevf_napi_enable_all(adapter);
 
@@ -3354,7 +3354,7 @@ static int ixgbevf_resume(struct pci_dev *pdev)
 		dev_err(&pdev->dev, "Cannot enable PCI device from suspend\n");
 		return err;
 	}
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBEVF_DISABLED, &adapter->state);
 	pci_set_master(pdev);
 
@@ -3712,7 +3712,7 @@ static pci_ers_result_t ixgbevf_io_slot_reset(struct pci_dev *pdev)
 		return PCI_ERS_RESULT_DISCONNECT;
 	}
 
-	smp_mb__before_clear_bit();
+	smp_mb__before_atomic();
 	clear_bit(__IXGBEVF_DISABLED, &adapter->state);
 	pci_set_master(pdev);