[libc] Reduce overall code size by externalising memmove()

Typical saving is 15-20 bytes in each file using memmove().

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

src/arch/x86/core/x86_string.c

@@ -35,7 +35,8 @@ FILE_LICENCE ( GPL2_OR_LATER );
  * @v len		Length
  * @ret dest		Destination address
  */
-void * __memcpy ( void *dest, const void *src, size_t len ) {
+void * __attribute__ (( noinline )) __memcpy ( void *dest, const void *src,
+					       size_t len ) {
 	void *edi = dest;
 	const void *esi = src;
 	int discard_ecx;
@@ -56,3 +57,50 @@ void * __memcpy ( void *dest, const void *src, size_t len ) {
 			       : "memory" );
 	return dest;
 }
+
+/**
+ * Copy memory area backwards
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+void * __attribute__ (( noinline )) __memcpy_reverse ( void *dest,
+						       const void *src,
+						       size_t len ) {
+	void *edi = ( dest + len - 1 );
+	const void *esi = ( src + len - 1 );
+	int discard_ecx;
+
+	/* Assume memmove() is not performance-critical, and perform a
+	 * bytewise copy for simplicity.
+	 */
+	__asm__ __volatile__ ( "std\n\t"
+			       "rep movsb\n\t"
+			       "cld\n\t"
+			       : "=&D" ( edi ), "=&S" ( esi ),
+				 "=&c" ( discard_ecx )
+			       : "0" ( edi ), "1" ( esi ),
+				 "2" ( len )
+			       : "memory" );
+	return dest;
+}
+
+
+/**
+ * Copy (possibly overlapping) memory area
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+void * __memmove ( void *dest, const void *src, size_t len ) {
+
+	if ( dest <= src ) {
+		return __memcpy ( dest, src, len );
+	} else {
+		return __memcpy_reverse ( dest, src, len );
+	}
+}

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

@@ -26,6 +26,7 @@ FILE_LICENCE ( PUBLIC_DOMAIN );
 #define __HAVE_ARCH_MEMCPY
 
 extern void * __memcpy ( void *dest, const void *src, size_t len );
+extern void * __memcpy_reverse ( void *dest, const void *src, size_t len );
 
 static inline __attribute__ (( always_inline )) void *
 __constant_memcpy ( void *dest, const void *src, size_t len ) {
@@ -144,29 +145,30 @@ __constant_memcpy ( void *dest, const void *src, size_t len ) {
 	  __memcpy ( (dest), (src), (len) ) )
 
 #define __HAVE_ARCH_MEMMOVE
-static inline void * memmove(void * dest,const void * src, size_t n)
-{
-int d0, d1, d2;
-if (dest<src)
-__asm__ __volatile__(
-	"cld\n\t"
-	"rep\n\t"
-	"movsb"
-	: "=&c" (d0), "=&S" (d1), "=&D" (d2)
-	:"0" (n),"1" (src),"2" (dest)
-	: "memory");
-else
-__asm__ __volatile__(
-	"std\n\t"
-	"rep\n\t"
-	"movsb\n\t"
-	"cld"
-	: "=&c" (d0), "=&S" (d1), "=&D" (d2)
-	:"0" (n),
-	 "1" (n-1+(const char *)src),
-	 "2" (n-1+(char *)dest)
-	:"memory");
-return dest;
+
+extern void * __memmove ( void *dest, const void *src, size_t len );
+
+/**
+ * Copy (possibly overlapping) memory area
+ *
+ * @v dest		Destination address
+ * @v src		Source address
+ * @v len		Length
+ * @ret dest		Destination address
+ */
+static inline __attribute__ (( always_inline )) void *
+memmove ( void *dest, const void *src, size_t len ) {
+	ssize_t offset = ( dest - src );
+
+	if ( __builtin_constant_p ( offset ) ) {
+		if ( offset <= 0 ) {
+			return memcpy ( dest, src, len );
+		} else {
+			return __memcpy_reverse ( dest, src, len );
+		}
+	} else {
+		return __memmove ( dest, src, len );
+	}
 }
 
 #define __HAVE_ARCH_MEMSET