diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
index a834057..b2655d9 100644
--- a/arch/xtensa/lib/strncpy_user.S
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -25,18 +25,18 @@
 /*
  * char *__strncpy_user(char *dst, const char *src, size_t len)
  */
-.text
-.begin literal
-.align	4
-.Lmask0:
-	.byte	0xff, 0x00, 0x00, 0x00
-.Lmask1:
-	.byte	0x00, 0xff, 0x00, 0x00
-.Lmask2:
-	.byte	0x00, 0x00, 0xff, 0x00
-.Lmask3:
-	.byte	0x00, 0x00, 0x00, 0xff
-.end literal
+
+#ifdef __XTENSA_EB__
+# define MASK0 0xff000000
+# define MASK1 0x00ff0000
+# define MASK2 0x0000ff00
+# define MASK3 0x000000ff
+#else
+# define MASK0 0x000000ff
+# define MASK1 0x0000ff00
+# define MASK2 0x00ff0000
+# define MASK3 0xff000000
+#endif
 
 # Register use
 #   a0/ return address
@@ -53,6 +53,7 @@
 #   a11/ dst
 #   a12/ tmp
 
+.text
 .align	4
 .global	__strncpy_user
 .type	__strncpy_user,@function
@@ -61,10 +62,10 @@
 	# a2/ dst, a3/ src, a4/ len
 	mov	a11, a2		# leave dst in return value register
 	beqz	a4, .Lret	# if len is zero
-	l32r	a5, .Lmask0	# mask for byte 0
-	l32r	a6, .Lmask1	# mask for byte 1
-	l32r	a7, .Lmask2	# mask for byte 2
-	l32r	a8, .Lmask3	# mask for byte 3
+	movi	a5, MASK0	# mask for byte 0
+	movi	a6, MASK1	# mask for byte 1
+	movi	a7, MASK2	# mask for byte 2
+	movi	a8, MASK3	# mask for byte 3
 	bbsi.l	a3, 0, .Lsrc1mod2 # if only  8-bit aligned
 	bbsi.l	a3, 1, .Lsrc2mod4 # if only 16-bit aligned
 .Lsrcaligned:	# return here when src is word-aligned