overhaul arm atomics for new atomics framework

switch to ll/sc model so that new atomic.h can provide optimized
versions of all the atomic primitives without needing an ll/sc loop
written in asm for each one.

all isa levels which use ldrex/strex now use the inline ll/sc model
even if the type of barrier to use is not known until runtime (v6).
the cas model is only used for arm v5 and earlier, and it has been
optimized to make the call via inline asm with custom constraints
rather than as a C function call.

arch/arm/atomic_arch.h

@@ -1,168 +1,64 @@
-#if __ARM_ARCH_7A__ || __ARM_ARCH_7R__ ||  __ARM_ARCH >= 7
+__attribute__((__visibility__("hidden")))
+extern const void *__arm_atomics[3]; /* gettp, cas, barrier */
 
-#define a_barrier a_barrier
-static inline void a_barrier()
-{
-	__asm__ __volatile__("dmb ish");
-}
+#if ((__ARM_ARCH_6__ || __ARM_ARCH_6K__ || __ARM_ARCH_6ZK__) && !__thumb__) \
+ || __ARM_ARCH_7A__ || __ARM_ARCH_7R__ ||  __ARM_ARCH >= 7
 
-#define a_cas a_cas
-static inline int a_cas(volatile int *p, int t, int s)
+#define a_ll a_ll
+static inline int a_ll(volatile int *p)
 {
-	int old;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%3\n"
-		"	cmp %0,%1\n"
-		"	bne 1f\n"
-		"	strex %0,%2,%3\n"
-		"	cmp %0, #0\n"
-		"	bne 1b\n"
-		"	mov %0, %1\n"
-		"1:	dmb ish\n"
-		: "=&r"(old)
-		: "r"(t), "r"(s), "Q"(*p)
-		: "memory", "cc" );
-	return old;
+	int v;
+	__asm__ __volatile__ ("ldrex %0, %1" : "=r"(v) : "Q"(*p));
+	return v;
 }
 
-#define a_swap a_swap
-static inline int a_swap(volatile int *x, int v)
+#define a_sc a_sc
+static inline int a_sc(volatile int *p, int v)
 {
-	int old, tmp;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%3\n"
-		"	strex %1,%2,%3\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(old), "=&r"(tmp)
-		: "r"(v), "Q"(*x)
-		: "memory", "cc" );
-	return old;
+	int r;
+	__asm__ __volatile__ ("strex %0,%1,%2" : "=&r"(r) : "r"(v), "Q"(*p) : "memory");
+	return !r;
 }
 
-#define a_fetch_add a_fetch_add
-static inline int a_fetch_add(volatile int *x, int v)
-{
-	int old, tmp;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%3\n"
-		"	add %0,%0,%2\n"
-		"	strex %1,%0,%3\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(old), "=&r"(tmp)
-		: "r"(v), "Q"(*x)
-		: "memory", "cc" );
-	return old-v;
-}
+#if __ARM_ARCH_7A__ || __ARM_ARCH_7R__ ||  __ARM_ARCH >= 7
 
-#define a_inc a_inc
-static inline void a_inc(volatile int *x)
+#define a_barrier a_barrier
+static inline void a_barrier()
 {
-	int tmp, tmp2;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%2\n"
-		"	add %0,%0,#1\n"
-		"	strex %1,%0,%2\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(tmp), "=&r"(tmp2)
-		: "Q"(*x)
-		: "memory", "cc" );
+	__asm__ __volatile__ ("dmb ish" : : : "memory");
 }
 
-#define a_dec a_dec
-static inline void a_dec(volatile int *x)
-{
-	int tmp, tmp2;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%2\n"
-		"	sub %0,%0,#1\n"
-		"	strex %1,%0,%2\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(tmp), "=&r"(tmp2)
-		: "Q"(*x)
-		: "memory", "cc" );
-}
+#endif
 
-#define a_and a_and
-static inline void a_and(volatile int *x, int v)
-{
-	int tmp, tmp2;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%3\n"
-		"	and %0,%0,%2\n"
-		"	strex %1,%0,%3\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(tmp), "=&r"(tmp2)
-		: "r"(v), "Q"(*x)
-		: "memory", "cc" );
-}
+#define a_pre_llsc a_barrier
+#define a_post_llsc a_barrier
 
-#define a_or a_or
-static inline void a_or(volatile int *x, int v)
-{
-	int tmp, tmp2;
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"1:	ldrex %0,%3\n"
-		"	orr %0,%0,%2\n"
-		"	strex %1,%0,%3\n"
-		"	cmp %1, #0\n"
-		"	bne 1b\n"
-		"	dmb ish\n"
-		: "=&r"(tmp), "=&r"(tmp2)
-		: "r"(v), "Q"(*x)
-		: "memory", "cc" );
-}
+#else
 
-#define a_store a_store
-static inline void a_store(volatile int *p, int x)
+#define a_cas a_cas
+static inline int a_cas(volatile int *p, int t, int s)
 {
-	__asm__ __volatile__(
-		"	dmb ish\n"
-		"	str %1,%0\n"
-		"	dmb ish\n"
-		: "=m"(*p)
-		: "r"(x)
-		: "memory", "cc" );
+	for (;;) {
+		register int r0 __asm__("r0") = t;
+		register int r1 __asm__("r1") = s;
+		register volatile int *r2 __asm__("r2") = p;
+		int old;
+		__asm__ __volatile__ (
+			"bl __a_cas"
+			: "+r"(r0) : "r"(r1), "r"(r2)
+			: "memory", "r3", "lr", "ip", "cc" );
+		if (!r0) return t;
+		if ((old=*p)!=t) return old;
+	}
 }
 
-#else
-
-int __a_cas(int, int, volatile int *) __attribute__((__visibility__("hidden")));
-#define __k_cas __a_cas
+#endif
 
+#ifndef a_barrier
 #define a_barrier a_barrier
 static inline void a_barrier()
 {
 	__asm__ __volatile__("bl __a_barrier"
 		: : : "memory", "cc", "ip", "lr" );
 }
-
-#define a_cas a_cas
-static inline int a_cas(volatile int *p, int t, int s)
-{
-	int old;
-	for (;;) {
-		if (!__k_cas(t, s, p))
-			return t;
-		if ((old=*p) != t)
-			return old;
-	}
-}
-
 #endif