local_t: alpha extension

Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Richard Henderson <rth@twiddle.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>

local_t: alpha extension
Signed-off-by: Mathieu Desnoyers <mathieu.desnoyers@polymtl.ca> Cc: Ivan Kokshaysky <ink@jurassic.park.msu.ru> Cc: Richard Henderson <rth@twiddle.net> Signed-off-by: Andrew Morton <akpm@linux-foundation.org> Signed-off-by: Linus Torvalds <torvalds@linux-foundation.org>
Mathieu Desnoyers · Linus Torvalds
1 parent 5e97b9309b
Showing 2 changed files with 329 additions and 24 deletions Side-by-side Diff
include/asm-alpha/local.h
include/asm-alpha/system.h
@@ -4,38 +4,116 @@
 #include <linux/percpu.h>
 #include <asm/atomic.h>
  
-typedef atomic64_t local_t;
+typedef struct
+{
+	atomic_long_t a;
+} local_t;
  
-#define LOCAL_INIT(i)	ATOMIC64_INIT(i)
-#define local_read(v)	atomic64_read(v)
-#define local_set(v,i)	atomic64_set(v,i)
+#define LOCAL_INIT(i)	{ ATOMIC_LONG_INIT(i) }
+#define local_read(l)	atomic_long_read(&(l)->a)
+#define local_set(l,i)	atomic_long_set(&(l)->a, (i))
+#define local_inc(l)	atomic_long_inc(&(l)->a)
+#define local_dec(l)	atomic_long_dec(&(l)->a)
+#define local_add(i,l)	atomic_long_add((i),(&(l)->a))
+#define local_sub(i,l)	atomic_long_sub((i),(&(l)->a))
  
-#define local_inc(v)	atomic64_inc(v)
-#define local_dec(v)	atomic64_dec(v)
-#define local_add(i, v)	atomic64_add(i, v)
-#define local_sub(i, v)	atomic64_sub(i, v)
+static __inline__ long local_add_return(long i, local_t * l)
+{
+	long temp, result;
+	__asm__ __volatile__(
+	"1:	ldq_l %0,%1\n"
+	"	addq %0,%3,%2\n"
+	"	addq %0,%3,%0\n"
+	"	stq_c %0,%1\n"
+	"	beq %0,2f\n"
+	".subsection 2\n"
+	"2:	br 1b\n"
+	".previous"
+	:"=&r" (temp), "=m" (l->a.counter), "=&r" (result)
+	:"Ir" (i), "m" (l->a.counter) : "memory");
+	return result;
+}
  
-#define __local_inc(v)		((v)->counter++)
-#define __local_dec(v)		((v)->counter++)
-#define __local_add(i,v)	((v)->counter+=(i))
-#define __local_sub(i,v)	((v)->counter-=(i))
+static __inline__ long local_sub_return(long i, local_t * l)
+{
+	long temp, result;
+	__asm__ __volatile__(
+	"1:	ldq_l %0,%1\n"
+	"	subq %0,%3,%2\n"
+	"	subq %0,%3,%0\n"
+	"	stq_c %0,%1\n"
+	"	beq %0,2f\n"
+	".subsection 2\n"
+	"2:	br 1b\n"
+	".previous"
+	:"=&r" (temp), "=m" (l->a.counter), "=&r" (result)
+	:"Ir" (i), "m" (l->a.counter) : "memory");
+	return result;
+}
  
+#define local_cmpxchg(l, o, n) \
+	(cmpxchg_local(&((l)->a.counter), (o), (n)))
+#define local_xchg(l, n) (xchg_local(&((l)->a.counter), (n)))
+
+/**
+ * local_add_unless - add unless the number is a given value
+ * @l: pointer of type local_t
+ * @a: the amount to add to l...
+ * @u: ...unless l is equal to u.
+ *
+ * Atomically adds @a to @l, so long as it was not @u.
+ * Returns non-zero if @l was not @u, and zero otherwise.
+ */
+#define local_add_unless(l, a, u)				\
+({								\
+	long c, old;						\
+	c = local_read(l);					\
+	for (;;) {						\
+		if (unlikely(c == (u)))				\
+			break;					\
+		old = local_cmpxchg((l), c, c + (a));	\
+		if (likely(old == c))				\
+			break;					\
+		c = old;					\
+	}							\
+	c != (u);						\
+})
+#define local_inc_not_zero(l) local_add_unless((l), 1, 0)
+
+#define local_add_negative(a, l) (local_add_return((a), (l)) < 0)
+
+#define local_dec_return(l) local_sub_return(1,(l))
+
+#define local_inc_return(l) local_add_return(1,(l))
+
+#define local_sub_and_test(i,l) (local_sub_return((i), (l)) == 0)
+
+#define local_inc_and_test(l) (local_add_return(1, (l)) == 0)
+
+#define local_dec_and_test(l) (local_sub_return(1, (l)) == 0)
+
+/* Verify if faster than atomic ops */
+#define __local_inc(l)		((l)->a.counter++)
+#define __local_dec(l)		((l)->a.counter++)
+#define __local_add(i,l)	((l)->a.counter+=(i))
+#define __local_sub(i,l)	((l)->a.counter-=(i))
+
 /* Use these for per-cpu local_t variables: on some archs they are
  * much more efficient than these naive implementations.  Note they take
  * a variable, not an address.
  */
-#define cpu_local_read(v)	local_read(&__get_cpu_var(v))
-#define cpu_local_set(v, i)	local_set(&__get_cpu_var(v), (i))
+#define cpu_local_read(l)	local_read(&__get_cpu_var(l))
+#define cpu_local_set(l, i)	local_set(&__get_cpu_var(l), (i))
  
-#define cpu_local_inc(v)	local_inc(&__get_cpu_var(v))
-#define cpu_local_dec(v)	local_dec(&__get_cpu_var(v))
-#define cpu_local_add(i, v)	local_add((i), &__get_cpu_var(v))
-#define cpu_local_sub(i, v)	local_sub((i), &__get_cpu_var(v))
+#define cpu_local_inc(l)	local_inc(&__get_cpu_var(l))
+#define cpu_local_dec(l)	local_dec(&__get_cpu_var(l))
+#define cpu_local_add(i, l)	local_add((i), &__get_cpu_var(l))
+#define cpu_local_sub(i, l)	local_sub((i), &__get_cpu_var(l))
  
-#define __cpu_local_inc(v)	__local_inc(&__get_cpu_var(v))
-#define __cpu_local_dec(v)	__local_dec(&__get_cpu_var(v))
-#define __cpu_local_add(i, v)	__local_add((i), &__get_cpu_var(v))
-#define __cpu_local_sub(i, v)	__local_sub((i), &__get_cpu_var(v))
+#define __cpu_local_inc(l)	__local_inc(&__get_cpu_var(l))
+#define __cpu_local_dec(l)	__local_dec(&__get_cpu_var(l))
+#define __cpu_local_add(i, l)	__local_add((i), &__get_cpu_var(l))
+#define __cpu_local_sub(i, l)	__local_sub((i), &__get_cpu_var(l))
  
 #endif /* _ALPHA_LOCAL_H */
@@ -443,6 +443,111 @@
      (__typeof__(*(ptr))) __xchg((ptr), (unsigned long)_x_, sizeof(*(ptr))); \
   })
  
+static inline unsigned long
+__xchg_u8_local(volatile char *m, unsigned long val)
+{
+	unsigned long ret, tmp, addr64;
+
+	__asm__ __volatile__(
+	"	andnot	%4,7,%3\n"
+	"	insbl	%1,%4,%1\n"
+	"1:	ldq_l	%2,0(%3)\n"
+	"	extbl	%2,%4,%0\n"
+	"	mskbl	%2,%4,%2\n"
+	"	or	%1,%2,%2\n"
+	"	stq_c	%2,0(%3)\n"
+	"	beq	%2,2f\n"
+	".subsection 2\n"
+	"2:	br	1b\n"
+	".previous"
+	: "=&r" (ret), "=&r" (val), "=&r" (tmp), "=&r" (addr64)
+	: "r" ((long)m), "1" (val) : "memory");
+
+	return ret;
+}
+
+static inline unsigned long
+__xchg_u16_local(volatile short *m, unsigned long val)
+{
+	unsigned long ret, tmp, addr64;
+
+	__asm__ __volatile__(
+	"	andnot	%4,7,%3\n"
+	"	inswl	%1,%4,%1\n"
+	"1:	ldq_l	%2,0(%3)\n"
+	"	extwl	%2,%4,%0\n"
+	"	mskwl	%2,%4,%2\n"
+	"	or	%1,%2,%2\n"
+	"	stq_c	%2,0(%3)\n"
+	"	beq	%2,2f\n"
+	".subsection 2\n"
+	"2:	br	1b\n"
+	".previous"
+	: "=&r" (ret), "=&r" (val), "=&r" (tmp), "=&r" (addr64)
+	: "r" ((long)m), "1" (val) : "memory");
+
+	return ret;
+}
+
+static inline unsigned long
+__xchg_u32_local(volatile int *m, unsigned long val)
+{
+	unsigned long dummy;
+
+	__asm__ __volatile__(
+	"1:	ldl_l %0,%4\n"
+	"	bis $31,%3,%1\n"
+	"	stl_c %1,%2\n"
+	"	beq %1,2f\n"
+	".subsection 2\n"
+	"2:	br 1b\n"
+	".previous"
+	: "=&r" (val), "=&r" (dummy), "=m" (*m)
+	: "rI" (val), "m" (*m) : "memory");
+
+	return val;
+}
+
+static inline unsigned long
+__xchg_u64_local(volatile long *m, unsigned long val)
+{
+	unsigned long dummy;
+
+	__asm__ __volatile__(
+	"1:	ldq_l %0,%4\n"
+	"	bis $31,%3,%1\n"
+	"	stq_c %1,%2\n"
+	"	beq %1,2f\n"
+	".subsection 2\n"
+	"2:	br 1b\n"
+	".previous"
+	: "=&r" (val), "=&r" (dummy), "=m" (*m)
+	: "rI" (val), "m" (*m) : "memory");
+
+	return val;
+}
+
+#define __xchg_local(ptr, x, size) \
+({ \
+	unsigned long __xchg__res; \
+	volatile void *__xchg__ptr = (ptr); \
+	switch (size) { \
+		case 1: __xchg__res = __xchg_u8_local(__xchg__ptr, x); break; \
+		case 2: __xchg__res = __xchg_u16_local(__xchg__ptr, x); break; \
+		case 4: __xchg__res = __xchg_u32_local(__xchg__ptr, x); break; \
+		case 8: __xchg__res = __xchg_u64_local(__xchg__ptr, x); break; \
+		default: __xchg_called_with_bad_pointer(); __xchg__res = x; \
+	} \
+	__xchg__res; \
+})
+
+#define xchg_local(ptr,x)						     \
+  ({									     \
+     __typeof__(*(ptr)) _x_ = (x);					     \
+     (__typeof__(*(ptr))) __xchg_local((ptr), (unsigned long)_x_,	     \
+     		sizeof(*(ptr))); \
+  })
+
 #define tas(ptr) (xchg((ptr),1))
  
  
@@ -593,6 +698,128 @@
      __typeof__(*(ptr)) _o_ = (o);					 \
      __typeof__(*(ptr)) _n_ = (n);					 \
      (__typeof__(*(ptr))) __cmpxchg((ptr), (unsigned long)_o_,		 \
+				    (unsigned long)_n_, sizeof(*(ptr))); \
+  })
+
+static inline unsigned long
+__cmpxchg_u8_local(volatile char *m, long old, long new)
+{
+	unsigned long prev, tmp, cmp, addr64;
+
+	__asm__ __volatile__(
+	"	andnot	%5,7,%4\n"
+	"	insbl	%1,%5,%1\n"
+	"1:	ldq_l	%2,0(%4)\n"
+	"	extbl	%2,%5,%0\n"
+	"	cmpeq	%0,%6,%3\n"
+	"	beq	%3,2f\n"
+	"	mskbl	%2,%5,%2\n"
+	"	or	%1,%2,%2\n"
+	"	stq_c	%2,0(%4)\n"
+	"	beq	%2,3f\n"
+	"2:\n"
+	".subsection 2\n"
+	"3:	br	1b\n"
+	".previous"
+	: "=&r" (prev), "=&r" (new), "=&r" (tmp), "=&r" (cmp), "=&r" (addr64)
+	: "r" ((long)m), "Ir" (old), "1" (new) : "memory");
+
+	return prev;
+}
+
+static inline unsigned long
+__cmpxchg_u16_local(volatile short *m, long old, long new)
+{
+	unsigned long prev, tmp, cmp, addr64;
+
+	__asm__ __volatile__(
+	"	andnot	%5,7,%4\n"
+	"	inswl	%1,%5,%1\n"
+	"1:	ldq_l	%2,0(%4)\n"
+	"	extwl	%2,%5,%0\n"
+	"	cmpeq	%0,%6,%3\n"
+	"	beq	%3,2f\n"
+	"	mskwl	%2,%5,%2\n"
+	"	or	%1,%2,%2\n"
+	"	stq_c	%2,0(%4)\n"
+	"	beq	%2,3f\n"
+	"2:\n"
+	".subsection 2\n"
+	"3:	br	1b\n"
+	".previous"
+	: "=&r" (prev), "=&r" (new), "=&r" (tmp), "=&r" (cmp), "=&r" (addr64)
+	: "r" ((long)m), "Ir" (old), "1" (new) : "memory");
+
+	return prev;
+}
+
+static inline unsigned long
+__cmpxchg_u32_local(volatile int *m, int old, int new)
+{
+	unsigned long prev, cmp;
+
+	__asm__ __volatile__(
+	"1:	ldl_l %0,%5\n"
+	"	cmpeq %0,%3,%1\n"
+	"	beq %1,2f\n"
+	"	mov %4,%1\n"
+	"	stl_c %1,%2\n"
+	"	beq %1,3f\n"
+	"2:\n"
+	".subsection 2\n"
+	"3:	br 1b\n"
+	".previous"
+	: "=&r"(prev), "=&r"(cmp), "=m"(*m)
+	: "r"((long) old), "r"(new), "m"(*m) : "memory");
+
+	return prev;
+}
+
+static inline unsigned long
+__cmpxchg_u64_local(volatile long *m, unsigned long old, unsigned long new)
+{
+	unsigned long prev, cmp;
+
+	__asm__ __volatile__(
+	"1:	ldq_l %0,%5\n"
+	"	cmpeq %0,%3,%1\n"
+	"	beq %1,2f\n"
+	"	mov %4,%1\n"
+	"	stq_c %1,%2\n"
+	"	beq %1,3f\n"
+	"2:\n"
+	".subsection 2\n"
+	"3:	br 1b\n"
+	".previous"
+	: "=&r"(prev), "=&r"(cmp), "=m"(*m)
+	: "r"((long) old), "r"(new), "m"(*m) : "memory");
+
+	return prev;
+}
+
+static __always_inline unsigned long
+__cmpxchg_local(volatile void *ptr, unsigned long old, unsigned long new,
+		int size)
+{
+	switch (size) {
+		case 1:
+			return __cmpxchg_u8_local(ptr, old, new);
+		case 2:
+			return __cmpxchg_u16_local(ptr, old, new);
+		case 4:
+			return __cmpxchg_u32_local(ptr, old, new);
+		case 8:
+			return __cmpxchg_u64_local(ptr, old, new);
+	}
+	__cmpxchg_called_with_bad_pointer();
+	return old;
+}
+
+#define cmpxchg_local(ptr,o,n)						 \
+  ({									 \
+     __typeof__(*(ptr)) _o_ = (o);					 \
+     __typeof__(*(ptr)) _n_ = (n);					 \
+     (__typeof__(*(ptr))) __cmpxchg_local((ptr), (unsigned long)_o_,	 \
 				    (unsigned long)_n_, sizeof(*(ptr))); \
   })
...	...	@@ -4,38 +4,116 @@
4	4	#include <linux/percpu.h>
5	5	#include <asm/atomic.h>
6	6
7		-typedef atomic64_t local_t;
	7	+typedef struct
	8	+{
	9	+ atomic_long_t a;
	10	+} local_t;
8	11
9		-#define LOCAL_INIT(i) ATOMIC64_INIT(i)
10		-#define local_read(v) atomic64_read(v)
11		-#define local_set(v,i) atomic64_set(v,i)
	12	+#define LOCAL_INIT(i) { ATOMIC_LONG_INIT(i) }
	13	+#define local_read(l) atomic_long_read(&(l)->a)
	14	+#define local_set(l,i) atomic_long_set(&(l)->a, (i))
	15	+#define local_inc(l) atomic_long_inc(&(l)->a)
	16	+#define local_dec(l) atomic_long_dec(&(l)->a)
	17	+#define local_add(i,l) atomic_long_add((i),(&(l)->a))
	18	+#define local_sub(i,l) atomic_long_sub((i),(&(l)->a))
12	19
13		-#define local_inc(v) atomic64_inc(v)
14		-#define local_dec(v) atomic64_dec(v)
15		-#define local_add(i, v) atomic64_add(i, v)
16		-#define local_sub(i, v) atomic64_sub(i, v)
	20	+static __inline__ long local_add_return(long i, local_t * l)
	21	+{
	22	+ long temp, result;
	23	+ __asm__ __volatile__(
	24	+ "1: ldq_l %0,%1\n"
	25	+ " addq %0,%3,%2\n"
	26	+ " addq %0,%3,%0\n"
	27	+ " stq_c %0,%1\n"
	28	+ " beq %0,2f\n"
	29	+ ".subsection 2\n"
	30	+ "2: br 1b\n"
	31	+ ".previous"
	32	+ :"=&r" (temp), "=m" (l->a.counter), "=&r" (result)
	33	+ :"Ir" (i), "m" (l->a.counter) : "memory");
	34	+ return result;
	35	+}
17	36
18		-#define __local_inc(v) ((v)->counter++)
19		-#define __local_dec(v) ((v)->counter++)
20		-#define __local_add(i,v) ((v)->counter+=(i))
21		-#define __local_sub(i,v) ((v)->counter-=(i))
	37	+static __inline__ long local_sub_return(long i, local_t * l)
	38	+{
	39	+ long temp, result;
	40	+ __asm__ __volatile__(
	41	+ "1: ldq_l %0,%1\n"
	42	+ " subq %0,%3,%2\n"
	43	+ " subq %0,%3,%0\n"
	44	+ " stq_c %0,%1\n"
	45	+ " beq %0,2f\n"
	46	+ ".subsection 2\n"
	47	+ "2: br 1b\n"
	48	+ ".previous"
	49	+ :"=&r" (temp), "=m" (l->a.counter), "=&r" (result)
	50	+ :"Ir" (i), "m" (l->a.counter) : "memory");
	51	+ return result;
	52	+}
22	53
	54	+#define local_cmpxchg(l, o, n) \
	55	+ (cmpxchg_local(&((l)->a.counter), (o), (n)))
	56	+#define local_xchg(l, n) (xchg_local(&((l)->a.counter), (n)))
	57	+
	58	+/**
	59	+ * local_add_unless - add unless the number is a given value
	60	+ * @l: pointer of type local_t
	61	+ * @a: the amount to add to l...
	62	+ * @u: ...unless l is equal to u.
	63	+ *
	64	+ * Atomically adds @a to @l, so long as it was not @u.
	65	+ * Returns non-zero if @l was not @u, and zero otherwise.
	66	+ */
	67	+#define local_add_unless(l, a, u) \
	68	+({ \
	69	+ long c, old; \
	70	+ c = local_read(l); \
	71	+ for (;;) { \
	72	+ if (unlikely(c == (u))) \
	73	+ break; \
	74	+ old = local_cmpxchg((l), c, c + (a)); \
	75	+ if (likely(old == c)) \
	76	+ break; \
	77	+ c = old; \
	78	+ } \
	79	+ c != (u); \
	80	+})
	81	+#define local_inc_not_zero(l) local_add_unless((l), 1, 0)
	82	+
	83	+#define local_add_negative(a, l) (local_add_return((a), (l)) < 0)
	84	+
	85	+#define local_dec_return(l) local_sub_return(1,(l))
	86	+
	87	+#define local_inc_return(l) local_add_return(1,(l))
	88	+
	89	+#define local_sub_and_test(i,l) (local_sub_return((i), (l)) == 0)
	90	+
	91	+#define local_inc_and_test(l) (local_add_return(1, (l)) == 0)
	92	+
	93	+#define local_dec_and_test(l) (local_sub_return(1, (l)) == 0)
	94	+
	95	+/* Verify if faster than atomic ops */
	96	+#define __local_inc(l) ((l)->a.counter++)
	97	+#define __local_dec(l) ((l)->a.counter++)
	98	+#define __local_add(i,l) ((l)->a.counter+=(i))
	99	+#define __local_sub(i,l) ((l)->a.counter-=(i))
	100	+
23	101	/* Use these for per-cpu local_t variables: on some archs they are
24	102	* much more efficient than these naive implementations. Note they take
25	103	* a variable, not an address.
26	104	*/
27		-#define cpu_local_read(v) local_read(&__get_cpu_var(v))
28		-#define cpu_local_set(v, i) local_set(&__get_cpu_var(v), (i))
	105	+#define cpu_local_read(l) local_read(&__get_cpu_var(l))
	106	+#define cpu_local_set(l, i) local_set(&__get_cpu_var(l), (i))
29	107
30		-#define cpu_local_inc(v) local_inc(&__get_cpu_var(v))
31		-#define cpu_local_dec(v) local_dec(&__get_cpu_var(v))
32		-#define cpu_local_add(i, v) local_add((i), &__get_cpu_var(v))
33		-#define cpu_local_sub(i, v) local_sub((i), &__get_cpu_var(v))
	108	+#define cpu_local_inc(l) local_inc(&__get_cpu_var(l))
	109	+#define cpu_local_dec(l) local_dec(&__get_cpu_var(l))
	110	+#define cpu_local_add(i, l) local_add((i), &__get_cpu_var(l))
	111	+#define cpu_local_sub(i, l) local_sub((i), &__get_cpu_var(l))
34	112
35		-#define __cpu_local_inc(v) __local_inc(&__get_cpu_var(v))
36		-#define __cpu_local_dec(v) __local_dec(&__get_cpu_var(v))
37		-#define __cpu_local_add(i, v) __local_add((i), &__get_cpu_var(v))
38		-#define __cpu_local_sub(i, v) __local_sub((i), &__get_cpu_var(v))
	113	+#define __cpu_local_inc(l) __local_inc(&__get_cpu_var(l))
	114	+#define __cpu_local_dec(l) __local_dec(&__get_cpu_var(l))
	115	+#define __cpu_local_add(i, l) __local_add((i), &__get_cpu_var(l))
	116	+#define __cpu_local_sub(i, l) __local_sub((i), &__get_cpu_var(l))
39	117
40	118	#endif /* _ALPHA_LOCAL_H */
...	...	@@ -443,6 +443,111 @@
443	443	(__typeof__((ptr))) __xchg((ptr), (unsigned long)_x_, sizeof((ptr))); \
444	444	})
445	445
	446	+static inline unsigned long
	447	+__xchg_u8_local(volatile char *m, unsigned long val)
	448	+{
	449	+ unsigned long ret, tmp, addr64;
	450	+
	451	+ __asm__ __volatile__(
	452	+ " andnot %4,7,%3\n"
	453	+ " insbl %1,%4,%1\n"
	454	+ "1: ldq_l %2,0(%3)\n"
	455	+ " extbl %2,%4,%0\n"
	456	+ " mskbl %2,%4,%2\n"
	457	+ " or %1,%2,%2\n"
	458	+ " stq_c %2,0(%3)\n"
	459	+ " beq %2,2f\n"
	460	+ ".subsection 2\n"
	461	+ "2: br 1b\n"
	462	+ ".previous"
	463	+ : "=&r" (ret), "=&r" (val), "=&r" (tmp), "=&r" (addr64)
	464	+ : "r" ((long)m), "1" (val) : "memory");
	465	+
	466	+ return ret;
	467	+}
	468	+
	469	+static inline unsigned long
	470	+__xchg_u16_local(volatile short *m, unsigned long val)
	471	+{
	472	+ unsigned long ret, tmp, addr64;
	473	+
	474	+ __asm__ __volatile__(
	475	+ " andnot %4,7,%3\n"
	476	+ " inswl %1,%4,%1\n"
	477	+ "1: ldq_l %2,0(%3)\n"
	478	+ " extwl %2,%4,%0\n"
	479	+ " mskwl %2,%4,%2\n"
	480	+ " or %1,%2,%2\n"
	481	+ " stq_c %2,0(%3)\n"
	482	+ " beq %2,2f\n"
	483	+ ".subsection 2\n"
	484	+ "2: br 1b\n"
	485	+ ".previous"
	486	+ : "=&r" (ret), "=&r" (val), "=&r" (tmp), "=&r" (addr64)
	487	+ : "r" ((long)m), "1" (val) : "memory");
	488	+
	489	+ return ret;
	490	+}
	491	+
	492	+static inline unsigned long
	493	+__xchg_u32_local(volatile int *m, unsigned long val)
	494	+{
	495	+ unsigned long dummy;
	496	+
	497	+ __asm__ __volatile__(
	498	+ "1: ldl_l %0,%4\n"
	499	+ " bis $31,%3,%1\n"
	500	+ " stl_c %1,%2\n"
	501	+ " beq %1,2f\n"
	502	+ ".subsection 2\n"
	503	+ "2: br 1b\n"
	504	+ ".previous"
	505	+ : "=&r" (val), "=&r" (dummy), "=m" (*m)
	506	+ : "rI" (val), "m" (*m) : "memory");
	507	+
	508	+ return val;
	509	+}
	510	+
	511	+static inline unsigned long
	512	+__xchg_u64_local(volatile long *m, unsigned long val)
	513	+{
	514	+ unsigned long dummy;
	515	+
	516	+ __asm__ __volatile__(
	517	+ "1: ldq_l %0,%4\n"
	518	+ " bis $31,%3,%1\n"
	519	+ " stq_c %1,%2\n"
	520	+ " beq %1,2f\n"
	521	+ ".subsection 2\n"
	522	+ "2: br 1b\n"
	523	+ ".previous"
	524	+ : "=&r" (val), "=&r" (dummy), "=m" (*m)
	525	+ : "rI" (val), "m" (*m) : "memory");
	526	+
	527	+ return val;
	528	+}
	529	+
	530	+#define __xchg_local(ptr, x, size) \
	531	+({ \
	532	+ unsigned long __xchg__res; \
	533	+ volatile void *__xchg__ptr = (ptr); \
	534	+ switch (size) { \
	535	+ case 1: __xchg__res = __xchg_u8_local(__xchg__ptr, x); break; \
	536	+ case 2: __xchg__res = __xchg_u16_local(__xchg__ptr, x); break; \
	537	+ case 4: __xchg__res = __xchg_u32_local(__xchg__ptr, x); break; \
	538	+ case 8: __xchg__res = __xchg_u64_local(__xchg__ptr, x); break; \
	539	+ default: __xchg_called_with_bad_pointer(); __xchg__res = x; \
	540	+ } \
	541	+ __xchg__res; \
	542	+})
	543	+
	544	+#define xchg_local(ptr,x) \
	545	+ ({ \
	546	+ __typeof__(*(ptr)) _x_ = (x); \
	547	+ (__typeof__(*(ptr))) __xchg_local((ptr), (unsigned long)_x_, \
	548	+ sizeof(*(ptr))); \
	549	+ })
	550	+
446	551	#define tas(ptr) (xchg((ptr),1))
447	552
448	553
...	...	@@ -593,6 +698,128 @@
593	698	__typeof__(*(ptr)) _o_ = (o); \
594	699	__typeof__(*(ptr)) _n_ = (n); \
595	700	(__typeof__(*(ptr))) __cmpxchg((ptr), (unsigned long)_o_, \
	701	+ (unsigned long)_n_, sizeof(*(ptr))); \
	702	+ })
	703	+
	704	+static inline unsigned long
	705	+__cmpxchg_u8_local(volatile char *m, long old, long new)
	706	+{
	707	+ unsigned long prev, tmp, cmp, addr64;
	708	+
	709	+ __asm__ __volatile__(
	710	+ " andnot %5,7,%4\n"
	711	+ " insbl %1,%5,%1\n"
	712	+ "1: ldq_l %2,0(%4)\n"
	713	+ " extbl %2,%5,%0\n"
	714	+ " cmpeq %0,%6,%3\n"
	715	+ " beq %3,2f\n"
	716	+ " mskbl %2,%5,%2\n"
	717	+ " or %1,%2,%2\n"
	718	+ " stq_c %2,0(%4)\n"
	719	+ " beq %2,3f\n"
	720	+ "2:\n"
	721	+ ".subsection 2\n"
	722	+ "3: br 1b\n"
	723	+ ".previous"
	724	+ : "=&r" (prev), "=&r" (new), "=&r" (tmp), "=&r" (cmp), "=&r" (addr64)
	725	+ : "r" ((long)m), "Ir" (old), "1" (new) : "memory");
	726	+
	727	+ return prev;
	728	+}
	729	+
	730	+static inline unsigned long
	731	+__cmpxchg_u16_local(volatile short *m, long old, long new)
	732	+{
	733	+ unsigned long prev, tmp, cmp, addr64;
	734	+
	735	+ __asm__ __volatile__(
	736	+ " andnot %5,7,%4\n"
	737	+ " inswl %1,%5,%1\n"
	738	+ "1: ldq_l %2,0(%4)\n"
	739	+ " extwl %2,%5,%0\n"
	740	+ " cmpeq %0,%6,%3\n"
	741	+ " beq %3,2f\n"
	742	+ " mskwl %2,%5,%2\n"
	743	+ " or %1,%2,%2\n"
	744	+ " stq_c %2,0(%4)\n"
	745	+ " beq %2,3f\n"
	746	+ "2:\n"
	747	+ ".subsection 2\n"
	748	+ "3: br 1b\n"
	749	+ ".previous"
	750	+ : "=&r" (prev), "=&r" (new), "=&r" (tmp), "=&r" (cmp), "=&r" (addr64)
	751	+ : "r" ((long)m), "Ir" (old), "1" (new) : "memory");
	752	+
	753	+ return prev;
	754	+}
	755	+
	756	+static inline unsigned long
	757	+__cmpxchg_u32_local(volatile int *m, int old, int new)
	758	+{
	759	+ unsigned long prev, cmp;
	760	+
	761	+ __asm__ __volatile__(
	762	+ "1: ldl_l %0,%5\n"
	763	+ " cmpeq %0,%3,%1\n"
	764	+ " beq %1,2f\n"
	765	+ " mov %4,%1\n"
	766	+ " stl_c %1,%2\n"
	767	+ " beq %1,3f\n"
	768	+ "2:\n"
	769	+ ".subsection 2\n"
	770	+ "3: br 1b\n"
	771	+ ".previous"
	772	+ : "=&r"(prev), "=&r"(cmp), "=m"(*m)
	773	+ : "r"((long) old), "r"(new), "m"(*m) : "memory");
	774	+
	775	+ return prev;
	776	+}
	777	+
	778	+static inline unsigned long
	779	+__cmpxchg_u64_local(volatile long *m, unsigned long old, unsigned long new)
	780	+{
	781	+ unsigned long prev, cmp;
	782	+
	783	+ __asm__ __volatile__(
	784	+ "1: ldq_l %0,%5\n"
	785	+ " cmpeq %0,%3,%1\n"
	786	+ " beq %1,2f\n"
	787	+ " mov %4,%1\n"
	788	+ " stq_c %1,%2\n"
	789	+ " beq %1,3f\n"
	790	+ "2:\n"
	791	+ ".subsection 2\n"
	792	+ "3: br 1b\n"
	793	+ ".previous"
	794	+ : "=&r"(prev), "=&r"(cmp), "=m"(*m)
	795	+ : "r"((long) old), "r"(new), "m"(*m) : "memory");
	796	+
	797	+ return prev;
	798	+}
	799	+
	800	+static __always_inline unsigned long
	801	+__cmpxchg_local(volatile void *ptr, unsigned long old, unsigned long new,
	802	+ int size)
	803	+{
	804	+ switch (size) {
	805	+ case 1:
	806	+ return __cmpxchg_u8_local(ptr, old, new);
	807	+ case 2:
	808	+ return __cmpxchg_u16_local(ptr, old, new);
	809	+ case 4:
	810	+ return __cmpxchg_u32_local(ptr, old, new);
	811	+ case 8:
	812	+ return __cmpxchg_u64_local(ptr, old, new);
	813	+ }
	814	+ __cmpxchg_called_with_bad_pointer();
	815	+ return old;
	816	+}
	817	+
	818	+#define cmpxchg_local(ptr,o,n) \
	819	+ ({ \
	820	+ __typeof__(*(ptr)) _o_ = (o); \
	821	+ __typeof__(*(ptr)) _n_ = (n); \
	822	+ (__typeof__(*(ptr))) __cmpxchg_local((ptr), (unsigned long)_o_, \
596	823	(unsigned long)_n_, sizeof(*(ptr))); \
597	824	})
598	825