diff --git a/sys/arm/arm/locore.S b/sys/arm/arm/locore.S
index 1fad8d18cb0a..0a0280c5090b 100644
--- a/sys/arm/arm/locore.S
+++ b/sys/arm/arm/locore.S
@@ -95,11 +95,6 @@ do_copy:
 	sub	r0, r0, r9
 	add	r0, r0, r8
 	mov	r4, r0
-	/* Make sure _arm_memcpy is NULL */
-	ldr	r3, .L_arm_memcpy
-	ldr	r3, [r3]
-	mov	r5, #0
-	str	r5, [r3]
 	bl	memcpy
 	ldr	r0, Lram_offset
 	add	pc, r4, r0
diff --git a/sys/arm/arm/support.S b/sys/arm/arm/support.S
index 44d80f4ed097..c2602bd2b9e1 100644
--- a/sys/arm/arm/support.S
+++ b/sys/arm/arm/support.S
@@ -871,6 +871,18 @@ ENTRY(memmove)
 #if !defined(__XSCALE__)
 ENTRY(memcpy)
 	/* save leaf functions having to store this away */
+	/* Do not check arm_memcpy if we're running from flash */
+#ifdef FLASHADDR
+#if FLASHADDR > PHYSADDR
+	ldr	r3, =FLASHADDR
+	cmp	r3, pc
+	bls	.Lnormal
+#else
+	ldr	r3, =FLASHADDR
+	cmp	r3, pc
+	bhi	.Lnormal
+#endif
+#endif
 	ldr	r3, .L_arm_memcpy
 	ldr	r3, [r3]
 	cmp	r3, #0
@@ -1096,6 +1108,17 @@ ENTRY(memcpy)
 	pld	[r1]
 	cmp	r2, #0x0c
 	ble	.Lmemcpy_short		/* <= 12 bytes */
+#ifdef FLASHADDR
+#if FLASHADDR > PHYSADDR
+	ldr	r3, =FLASHADDR
+	cmp	r3, pc
+	bls	.Lnormal
+#else
+	ldr	r3, =FLASHADDR
+	cmp	r3, pc
+	bhi	.Lnormal
+#endif
+#endif
 	ldr	r3, .L_arm_memcpy
 	ldr	r3, [r3]
 	cmp	r3, #0