[libc] Fix and externalise memswap()

Make memswap() behave correctly if called with a length of zero.

Signed-off-by: Michael Brown <mcb30@ipxe.org>

src/arch/x86/core/x86_string.c

@@ -105,6 +105,34 @@ void * __memmove ( void *dest, const void *src, size_t len ) {
 	}
 }
 
+/**
+ * Swap memory areas
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+void * memswap ( void *dest, void *src, size_t len ) {
+	size_t discard_c;
+	int discard;
+
+	__asm__ __volatile__ ( "\n1:\n\t"
+			       "dec %2\n\t"
+			       "js 2f\n\t"
+			       "movb (%0,%2), %b3\n\t"
+			       "xchgb (%1,%2), %b3\n\t"
+			       "movb %b3, (%0,%2)\n\t"
+			       "jmp 1b\n\t"
+			       "2:\n\t"
+			       : "=r" ( src ), "=r" ( dest ),
+				 "=&c" ( discard_c ), "=&q" ( discard )
+			       : "0" ( src ), "1" ( dest ), "2" ( len )
+			       : "memory" );
+
+	return dest;
+}
+
 /**
  * Calculate length of string
  *

src/arch/x86/include/bits/string.h

@@ -213,21 +213,8 @@ static inline void * memset ( void *dest, int fill, size_t len ) {
 }
 
 #define __HAVE_ARCH_MEMSWAP
-static inline void * memswap(void *dest, void *src, size_t n)
+
-{
+extern void * memswap ( void *dest, void *src, size_t len );
-long d0, d1, d2, d3;
-__asm__ __volatile__(
-	"\n1:\t"
-	"movb (%2),%%al\n\t"
-	"xchgb (%1),%%al\n\t"
-	"inc %1\n\t"
-	"stosb\n\t"
-	"loop 1b"
-	: "=&c" (d0), "=&S" (d1), "=&D" (d2), "=&a" (d3)
-	: "0" (n), "1" (src), "2" (dest)
-	: "memory" );
-return dest;
-}
 
 #define __HAVE_ARCH_STRNCMP