ia64: Fix _raw_read_lock() to not switch text sections. Tidy it up with the

	help of ia64_fetchadd() macro.  Ditto for _raw_read_unlock().

include/asm-ia64/intrinsics.h

@@ -17,16 +17,16 @@
 extern unsigned long __bad_size_for_ia64_fetch_and_add (void);
 extern unsigned long __bad_increment_for_ia64_fetch_and_add (void);
 
-#define IA64_FETCHADD(tmp,v,n,sz)						\
+#define IA64_FETCHADD(tmp,v,n,sz,sem)						\
 ({										\
 	switch (sz) {								\
 	      case 4:								\
-		__asm__ __volatile__ ("fetchadd4.rel %0=[%1],%2"		\
+		__asm__ __volatile__ ("fetchadd4."sem" %0=[%1],%2"		\
 				      : "=r"(tmp) : "r"(v), "i"(n) : "memory");	\
 		break;								\
 										\
 	      case 8:								\
-		__asm__ __volatile__ ("fetchadd8.rel %0=[%1],%2"		\
+		__asm__ __volatile__ ("fetchadd8."sem" %0=[%1],%2"		\
 				      : "=r"(tmp) : "r"(v), "i"(n) : "memory");	\
 		break;								\
 										\
@@ -35,32 +35,34 @@ extern unsigned long __bad_increment_for_ia64_fetch_and_add (void);
 	}									\
 })
 
-#define ia64_fetch_and_add(i,v)								\
+#define ia64_fetchadd(i,v,sem)								\
 ({											\
 	__u64 _tmp;									\
 	volatile __typeof__(*(v)) *_v = (v);						\
 	/* Can't use a switch () here: gcc isn't always smart enough for that... */	\
 	if ((i) == -16)									\
-		IA64_FETCHADD(_tmp, _v, -16, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, -16, sizeof(*(v)), sem);			\
 	else if ((i) == -8)								\
-		IA64_FETCHADD(_tmp, _v, -8, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, -8, sizeof(*(v)), sem);				\
 	else if ((i) == -4)								\
-		IA64_FETCHADD(_tmp, _v, -4, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, -4, sizeof(*(v)), sem);				\
 	else if ((i) == -1)								\
-		IA64_FETCHADD(_tmp, _v, -1, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, -1, sizeof(*(v)), sem);				\
 	else if ((i) == 1)								\
-		IA64_FETCHADD(_tmp, _v, 1, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, 1, sizeof(*(v)), sem);				\
 	else if ((i) == 4)								\
-		IA64_FETCHADD(_tmp, _v, 4, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, 4, sizeof(*(v)), sem);				\
 	else if ((i) == 8)								\
-		IA64_FETCHADD(_tmp, _v, 8, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, 8, sizeof(*(v)), sem);				\
 	else if ((i) == 16)								\
-		IA64_FETCHADD(_tmp, _v, 16, sizeof(*(v)));				\
+		IA64_FETCHADD(_tmp, _v, 16, sizeof(*(v)), sem);				\
 	else										\
 		_tmp = __bad_increment_for_ia64_fetch_and_add();			\
-	(__typeof__(*(v))) (_tmp + (i));	/* return new value */			\
+	(__typeof__(*(v))) (_tmp);	/* return old value */				\
 })
 
+#define ia64_fetch_and_add(i,v)	(ia64_fetchadd(i, v, "rel") + (i)) /* return new value */
+
 /*
  * This function doesn't exist, so you'll get a linker error if
  * something tries to do an invalid xchg().

include/asm-ia64/spinlock.h

@@ -72,43 +72,31 @@ do {											\
 #define spin_unlock_wait(x)	do { barrier(); } while ((x)->lock)
 
 typedef struct {
-	volatile int read_counter:31;
-	volatile int write_lock:1;
+	volatile int read_counter	: 31;
+	volatile int write_lock		:  1;
 } rwlock_t;
 #define RW_LOCK_UNLOCKED (rwlock_t) { 0, 0 }
 
 #define rwlock_init(x)		do { *(x) = RW_LOCK_UNLOCKED; } while(0)
 #define rwlock_is_locked(x)	(*(volatile int *) (x) != 0)
 
-#define _raw_read_lock(rw)							\
-do {										\
-	int __read_lock_tmp = 0;						\
-	__asm__ __volatile__ ("1:\tfetchadd4.acq %0 = [%1], 1\n"		\
-			      ";;\n"						\
-			      "tbit.nz p6,p0 = %0, 31\n"			\
-			      "(p6) br.cond.sptk.few 2f\n"			\
-			      ".section .text.lock,\"ax\"\n"			\
-			      "2:\tfetchadd4.rel %0 = [%1], -1\n"		\
-			      ";;\n"						\
-			      "3:\tld4.acq %0 = [%1]\n"				\
-			      ";;\n"						\
-			      "tbit.nz p6,p0 = %0, 31\n"			\
-			      "(p6) br.cond.sptk.few 3b\n"			\
-			      "br.cond.sptk.few 1b\n"				\
-			      ";;\n"						\
-			      ".previous\n"					\
-			      : "=&r" (__read_lock_tmp)				\
-			      : "r" (rw) : "p6", "memory");			\
-} while(0)
+#define _raw_read_lock(rw)								\
+do {											\
+	rwlock_t *__read_lock_ptr = (rw);						\
+											\
+	while (unlikely(ia64_fetchadd(1, (int *) __read_lock_ptr, "acq") < 0)) {	\
+		ia64_fetchadd(-1, (int *) __read_lock_ptr, "rel");			\
+		while (*(volatile int *)__read_lock_ptr < 0)				\
+			barrier();							\
+											\
+	}										\
+} while (0)
 
-#define _raw_read_unlock(rw)							\
-do {										\
-	int __read_unlock_tmp = 0;						\
-	__asm__ __volatile__ ("fetchadd4.rel %0 = [%1], -1\n"			\
-			      : "=r" (__read_unlock_tmp)			\
-			      : "r" (rw)					\
-			      : "memory");					\
-} while(0)
+#define _raw_read_unlock(rw)					\
+do {								\
+	rwlock_t *__read_lock_ptr = (rw);			\
+	ia64_fetchadd(-1, (int *) __read_lock_ptr, "rel");	\
+} while (0)
 
 #define _raw_write_lock(rw)							\
 do {										\