diff --git a/arch/arm/src/armv7-a/arm_memcpy.S b/arch/arm/src/armv7-a/arm_memcpy.S
index 84da4942b6e..713359d09ee 100644
--- a/arch/arm/src/armv7-a/arm_memcpy.S
+++ b/arch/arm/src/armv7-a/arm_memcpy.S
@@ -143,11 +143,26 @@ MEM_LongCopyTable:
  * Input Parameters:
  *   r0 = destination, r1 = source, r2 = length
  *
+ * Returned Value:
+ *   r0 = destination r1-r3 burned
+ *
  ************************************************************************************/
 
+	.align 4
 	.thumb_func
+
 memcpy:
 	push	{r14}
+	push    {r0}
+	bl      _do_memcpy
+	pop     {r0}
+	pop     {pc}
+
+	.align 4
+
+	.thumb_func
+_do_memcpy:
+	push    {r14}
 
 	/* This allows the inner workings to "assume" a minimum amount of bytes */
 	/* Quickly check for very short copies */
diff --git a/arch/arm/src/armv7-m/up_memcpy.S b/arch/arm/src/armv7-m/up_memcpy.S
index ea267f78fcc..84877b04009 100644
--- a/arch/arm/src/armv7-m/up_memcpy.S
+++ b/arch/arm/src/armv7-m/up_memcpy.S
@@ -142,11 +142,26 @@ MEM_LongCopyTable:
  * Input Parameters:
  *   r0 = destination, r1 = source, r2 = length
  *
+ * Returned Value:
+ *   r0 = destination r1-r3 burned
+ *
  ************************************************************************************/
 
+	.align 4
 	.thumb_func
+
 memcpy:
 	push	{r14}
+	push    {r0}
+	bl      _do_memcpy
+	pop     {r0}
+	pop     {pc}
+
+	.align 4
+
+	.thumb_func
+_do_memcpy:
+	push    {r14}
 
 	/* This allows the inner workings to "assume" a minimum amount of bytes */
 	/* Quickly check for very short copies */