xtensa: clean up fixups in assembly code

Remove duplicate definitions of EX() and similar TRY/CATCH and SRC/DST
macros from assembly sources and put single definition into asm/asmmacro.h

Signed-off-by: Max Filippov <jcmvbkbc@gmail.com>

arch/xtensa/include/asm/asmmacro.h

@@ -150,5 +150,12 @@
 		__endl	\ar \as
 	.endm
 
+/* Load or store instructions that may cause exceptions use the EX macro. */
+
+#define EX(handler)				\
+	.section __ex_table, "a";		\
+	.word	97f, handler;			\
+	.previous				\
+97:
 
 #endif /* _XTENSA_ASMMACRO_H */

arch/xtensa/kernel/entry.S

@@ -14,6 +14,7 @@
 
 #include <linux/linkage.h>
 #include <asm/asm-offsets.h>
+#include <asm/asmmacro.h>
 #include <asm/processor.h>
 #include <asm/coprocessor.h>
 #include <asm/thread_info.h>
@@ -1094,35 +1095,12 @@ ENDPROC(fast_syscall_unrecoverable)
  *	     <  VALID_DOUBLE_EXCEPTION_ADDRESS: regular exception
  *
  * Note: we don't have to save a2; a2 holds the return value
- *
- * We use the two macros TRY and CATCH:
- *
- * TRY	 adds an entry to the __ex_table fixup table for the immediately
- *	 following instruction.
- *
- * CATCH catches any exception that occurred at one of the preceding TRY
- *       statements and continues from there
- *
- * Usage TRY	l32i	a0, a1, 0
- *		<other code>
- *	 done:	rfe
- *	 CATCH	<set return code>
- *		j done
  */
 
 	.literal_position
 
 #ifdef CONFIG_FAST_SYSCALL_XTENSA
 
-#define TRY								\
-	.section __ex_table, "a";					\
-	.word	66f, 67f;						\
-	.text;								\
-66:
-
-#define CATCH								\
-67:
-
 ENTRY(fast_syscall_xtensa)
 
 	s32i	a7, a2, PT_AREG7	# we need an additional register
@@ -1136,9 +1114,9 @@ ENTRY(fast_syscall_xtensa)
 
 .Lswp:	/* Atomic compare and swap */
 
-TRY	l32i	a0, a3, 0		# read old value
+EX(.Leac) l32i	a0, a3, 0		# read old value
 	bne	a0, a4, 1f		# same as old value? jump
-TRY	s32i	a5, a3, 0		# different, modify value
+EX(.Leac) s32i	a5, a3, 0		# different, modify value
 	l32i	a7, a2, PT_AREG7	# restore a7
 	l32i	a0, a2, PT_AREG0	# restore a0
 	movi	a2, 1			# and return 1
@@ -1151,12 +1129,12 @@ TRY	s32i	a5, a3, 0		# different, modify value
 
 .Lnswp:	/* Atomic set, add, and exg_add. */
 
-TRY	l32i	a7, a3, 0		# orig
+EX(.Leac) l32i	a7, a3, 0		# orig
 	addi	a6, a6, -SYS_XTENSA_ATOMIC_SET
 	add	a0, a4, a7		# + arg
 	moveqz	a0, a4, a6		# set
 	addi	a6, a6, SYS_XTENSA_ATOMIC_SET
-TRY	s32i	a0, a3, 0		# write new value
+EX(.Leac) s32i	a0, a3, 0		# write new value
 
 	mov	a0, a2
 	mov	a2, a7
@@ -1164,7 +1142,6 @@ TRY	s32i	a0, a3, 0		# write new value
 	l32i	a0, a0, PT_AREG0	# restore a0
 	rfe
 
-CATCH
 .Leac:	l32i	a7, a2, PT_AREG7	# restore a7
 	l32i	a0, a2, PT_AREG0	# restore a0
 	movi	a2, -EFAULT

arch/xtensa/lib/checksum.S

@@ -14,9 +14,10 @@
  *		2 of the License, or (at your option) any later version.
  */
 
-#include <asm/errno.h>
+#include <linux/errno.h>
 #include <linux/linkage.h>
 #include <variant/core.h>
+#include <asm/asmmacro.h>
 
 /*
  * computes a partial checksum, e.g. for TCP/UDP fragments
@@ -175,23 +176,8 @@ ENDPROC(csum_partial)
 
 /*
  * Copy from ds while checksumming, otherwise like csum_partial
- *
- * The macros SRC and DST specify the type of access for the instruction.
- * thus we can call a custom exception handler for each access type.
  */
 
-#define SRC(y...)			\
-	9999: y;			\
-	.section __ex_table, "a";	\
-	.long 9999b, 6001f	;	\
-	.previous
-
-#define DST(y...)			\
-	9999: y;			\
-	.section __ex_table, "a";	\
-	.long 9999b, 6002f	;	\
-	.previous
-
 /*
 unsigned int csum_partial_copy_generic (const char *src, char *dst, int len,
 					int sum, int *src_err_ptr, int *dst_err_ptr)
@@ -244,28 +230,28 @@ ENTRY(csum_partial_copy_generic)
 	add	a10, a10, a2	/* a10 = end of last 32-byte src chunk */
 .Loop5:
 #endif
-SRC(	l32i	a9, a2, 0	)
-SRC(	l32i	a8, a2, 4	)
-DST(	s32i	a9, a3, 0	)
-DST(	s32i	a8, a3, 4	)
+EX(10f)	l32i	a9, a2, 0
+EX(10f)	l32i	a8, a2, 4
+EX(11f)	s32i	a9, a3, 0
+EX(11f)	s32i	a8, a3, 4
 	ONES_ADD(a5, a9)
 	ONES_ADD(a5, a8)
-SRC(	l32i	a9, a2, 8	)
-SRC(	l32i	a8, a2, 12	)
-DST(	s32i	a9, a3, 8	)
-DST(	s32i	a8, a3, 12	)
+EX(10f)	l32i	a9, a2, 8
+EX(10f)	l32i	a8, a2, 12
+EX(11f)	s32i	a9, a3, 8
+EX(11f)	s32i	a8, a3, 12
 	ONES_ADD(a5, a9)
 	ONES_ADD(a5, a8)
-SRC(	l32i	a9, a2, 16	)
-SRC(	l32i	a8, a2, 20	)
-DST(	s32i	a9, a3, 16	)
-DST(	s32i	a8, a3, 20	)
+EX(10f)	l32i	a9, a2, 16
+EX(10f)	l32i	a8, a2, 20
+EX(11f)	s32i	a9, a3, 16
+EX(11f)	s32i	a8, a3, 20
 	ONES_ADD(a5, a9)
 	ONES_ADD(a5, a8)
-SRC(	l32i	a9, a2, 24	)
-SRC(	l32i	a8, a2, 28	)
-DST(	s32i	a9, a3, 24	)
-DST(	s32i	a8, a3, 28	)
+EX(10f)	l32i	a9, a2, 24
+EX(10f)	l32i	a8, a2, 28
+EX(11f)	s32i	a9, a3, 24
+EX(11f)	s32i	a8, a3, 28
 	ONES_ADD(a5, a9)
 	ONES_ADD(a5, a8)
 	addi	a2, a2, 32
@@ -284,8 +270,8 @@ DST(	s32i	a8, a3, 28	)
 	add	a10, a10, a2	/* a10 = end of last 4-byte src chunk */
 .Loop6:
 #endif
-SRC(	l32i	a9, a2, 0	)
-DST(	s32i	a9, a3, 0	)
+EX(10f)	l32i	a9, a2, 0
+EX(11f)	s32i	a9, a3, 0
 	ONES_ADD(a5, a9)
 	addi	a2, a2, 4
 	addi	a3, a3, 4
@@ -315,8 +301,8 @@ DST(	s32i	a9, a3, 0	)
 	add	a10, a10, a2	/* a10 = end of last 2-byte src chunk */
 .Loop7:
 #endif
-SRC(	l16ui	a9, a2, 0	)
-DST(	s16i	a9, a3, 0	)
+EX(10f)	l16ui	a9, a2, 0
+EX(11f)	s16i	a9, a3, 0
 	ONES_ADD(a5, a9)
 	addi	a2, a2, 2
 	addi	a3, a3, 2
@@ -326,8 +312,8 @@ DST(	s16i	a9, a3, 0	)
 4:
 	/* This section processes a possible trailing odd byte. */
 	_bbci.l	a4, 0, 8f	/* 1-byte chunk */
-SRC(	l8ui	a9, a2, 0	)
-DST(	s8i	a9, a3, 0	)
+EX(10f)	l8ui	a9, a2, 0
+EX(11f)	s8i	a9, a3, 0
 #ifdef __XTENSA_EB__
 	slli	a9, a9, 8	/* shift byte to bits 8..15 */
 #endif
@@ -350,10 +336,10 @@ DST(	s8i	a9, a3, 0	)
 	add	a10, a10, a2	/* a10 = end of last odd-aligned, 2-byte src chunk */
 .Loop8:
 #endif
-SRC(	l8ui	a9, a2, 0	)
-SRC(	l8ui	a8, a2, 1	)
-DST(	s8i	a9, a3, 0	)
-DST(	s8i	a8, a3, 1	)
+EX(10f)	l8ui	a9, a2, 0
+EX(10f)	l8ui	a8, a2, 1
+EX(11f)	s8i	a9, a3, 0
+EX(11f)	s8i	a8, a3, 1
 #ifdef __XTENSA_EB__
 	slli	a9, a9, 8	/* combine into a single 16-bit value */
 #else				/* for checksum computation */
@@ -381,7 +367,7 @@ ENDPROC(csum_partial_copy_generic)
 	a12 = original dst for exception handling
 */
 
-6001:
+10:
 	_movi	a2, -EFAULT
 	s32i	a2, a6, 0	/* src_err_ptr */
 
@@ -403,7 +389,7 @@ ENDPROC(csum_partial_copy_generic)
 2:
 	retw
 
-6002:
+11:
 	movi	a2, -EFAULT
 	s32i	a2, a7, 0	/* dst_err_ptr */
 	movi	a2, 0

arch/xtensa/lib/memset.S

@@ -12,6 +12,7 @@
  */
 
 #include <variant/core.h>
+#include <asm/asmmacro.h>
 
 /*
  * void *memset(void *dst, int c, size_t length)
@@ -28,15 +29,6 @@
  *     the alignment labels).
  */
 
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler)	\
-9:	insn	reg1, reg2, offset;		\
-	.section __ex_table, "a";		\
-	.word	9b, handler;			\
-	.previous
-
-
 .text
 .align	4
 .global	memset
@@ -73,10 +65,10 @@ memset:
 	add	a6, a6, a5	# a6 = end of last 16B chunk
 #endif /* !XCHAL_HAVE_LOOPS */
 .Loop1:
-	EX(s32i, a3, a5,  0, memset_fixup)
-	EX(s32i, a3, a5,  4, memset_fixup)
-	EX(s32i, a3, a5,  8, memset_fixup)
-	EX(s32i, a3, a5, 12, memset_fixup)
+EX(10f) s32i	a3, a5,  0
+EX(10f) s32i	a3, a5,  4
+EX(10f) s32i	a3, a5,  8
+EX(10f) s32i	a3, a5, 12
 	addi	a5, a5, 16
 #if !XCHAL_HAVE_LOOPS
 	blt	a5, a6, .Loop1
@@ -84,23 +76,23 @@ memset:
 .Loop1done:
 	bbci.l	a4, 3, .L2
 	# set 8 bytes
-	EX(s32i, a3, a5,  0, memset_fixup)
-	EX(s32i, a3, a5,  4, memset_fixup)
+EX(10f) s32i	a3, a5,  0
+EX(10f) s32i	a3, a5,  4
 	addi	a5, a5,  8
 .L2:
 	bbci.l	a4, 2, .L3
 	# set 4 bytes
-	EX(s32i, a3, a5,  0, memset_fixup)
+EX(10f) s32i	a3, a5,  0
 	addi	a5, a5,  4
 .L3:
 	bbci.l	a4, 1, .L4
 	# set 2 bytes
-	EX(s16i, a3, a5,  0, memset_fixup)
+EX(10f) s16i	a3, a5,  0
 	addi	a5, a5,  2
 .L4:
 	bbci.l	a4, 0, .L5
 	# set 1 byte
-	EX(s8i, a3, a5,  0, memset_fixup)
+EX(10f) s8i	a3, a5,  0
 .L5:
 .Lret1:
 	retw
@@ -114,7 +106,7 @@ memset:
 	bbci.l	a5, 0, .L20		# branch if dst alignment half-aligned
 	# dst is only byte aligned
 	# set 1 byte
-	EX(s8i, a3, a5,  0, memset_fixup)
+EX(10f) s8i	a3, a5,  0
 	addi	a5, a5,  1
 	addi	a4, a4, -1
 	# now retest if dst aligned
@@ -122,7 +114,7 @@ memset:
 .L20:
 	# dst half-aligned
 	# set 2 bytes
-	EX(s16i, a3, a5,  0, memset_fixup)
+EX(10f) s16i	a3, a5,  0
 	addi	a5, a5,  2
 	addi	a4, a4, -2
 	j	.L0		# dst is now aligned, return to main algorithm
@@ -141,7 +133,7 @@ memset:
 	add	a6, a5, a4	# a6 = ending address
 #endif /* !XCHAL_HAVE_LOOPS */
 .Lbyteloop:
-	EX(s8i, a3, a5, 0, memset_fixup)
+EX(10f) s8i	a3, a5, 0
 	addi	a5, a5, 1
 #if !XCHAL_HAVE_LOOPS
 	blt	a5, a6, .Lbyteloop
@@ -155,6 +147,6 @@ memset:
 
 /* We return zero if a failure occurred. */
 
-memset_fixup:
+10:
 	movi	a2, 0
 	retw

arch/xtensa/lib/strncpy_user.S

@@ -11,16 +11,9 @@
  *  Copyright (C) 2002 Tensilica Inc.
  */
 
-#include <variant/core.h>
 #include <linux/errno.h>
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler)	\
-9:	insn	reg1, reg2, offset;		\
-	.section __ex_table, "a";		\
-	.word	9b, handler;			\
-	.previous
+#include <variant/core.h>
+#include <asm/asmmacro.h>
 
 /*
  * char *__strncpy_user(char *dst, const char *src, size_t len)
@@ -75,9 +68,9 @@ __strncpy_user:
 	j	.Ldstunaligned
 
 .Lsrc1mod2:	# src address is odd
-	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
+EX(11f)	l8ui	a9, a3, 0		# get byte 0
 	addi	a3, a3, 1		# advance src pointer
-	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+EX(10f)	s8i	a9, a11, 0		# store byte 0
 	beqz	a9, .Lret		# if byte 0 is zero
 	addi	a11, a11, 1		# advance dst pointer
 	addi	a4, a4, -1		# decrement len
@@ -85,16 +78,16 @@ __strncpy_user:
 	bbci.l	a3, 1, .Lsrcaligned	# if src is now word-aligned
 
 .Lsrc2mod4:	# src address is 2 mod 4
-	EX(l8ui, a9, a3, 0, fixup_l)	# get byte 0
+EX(11f)	l8ui	a9, a3, 0		# get byte 0
 	/* 1-cycle interlock */
-	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+EX(10f)	s8i	a9, a11, 0		# store byte 0
 	beqz	a9, .Lret		# if byte 0 is zero
 	addi	a11, a11, 1		# advance dst pointer
 	addi	a4, a4, -1		# decrement len
 	beqz	a4, .Lret		# if len is zero
-	EX(l8ui, a9, a3, 1, fixup_l)	# get byte 0
+EX(11f)	l8ui	a9, a3, 1		# get byte 0
 	addi	a3, a3, 2		# advance src pointer
-	EX(s8i, a9, a11, 0, fixup_s)	# store byte 0
+EX(10f)	s8i	a9, a11, 0		# store byte 0
 	beqz	a9, .Lret		# if byte 0 is zero
 	addi	a11, a11, 1		# advance dst pointer
 	addi	a4, a4, -1		# decrement len
@@ -117,12 +110,12 @@ __strncpy_user:
 	add	a12, a12, a11	# a12 = end of last 4B chunck
 #endif
 .Loop1:
-	EX(l32i, a9, a3, 0, fixup_l)	# get word from src
+EX(11f)	l32i	a9, a3, 0		# get word from src
 	addi	a3, a3, 4		# advance src pointer
 	bnone	a9, a5, .Lz0		# if byte 0 is zero
 	bnone	a9, a6, .Lz1		# if byte 1 is zero
 	bnone	a9, a7, .Lz2		# if byte 2 is zero
-	EX(s32i, a9, a11, 0, fixup_s)	# store word to dst
+EX(10f)	s32i	a9, a11, 0		# store word to dst
 	bnone	a9, a8, .Lz3		# if byte 3 is zero
 	addi	a11, a11, 4		# advance dst pointer
 #if !XCHAL_HAVE_LOOPS
@@ -132,7 +125,7 @@ __strncpy_user:
 .Loop1done:
 	bbci.l	a4, 1, .L100
 	# copy 2 bytes
-	EX(l16ui, a9, a3, 0, fixup_l)
+EX(11f)	l16ui	a9, a3, 0
 	addi	a3, a3, 2		# advance src pointer
 #ifdef __XTENSA_EB__
 	bnone	a9, a7, .Lz0		# if byte 2 is zero
@@ -141,13 +134,13 @@ __strncpy_user:
 	bnone	a9, a5, .Lz0		# if byte 0 is zero
 	bnone	a9, a6, .Lz1		# if byte 1 is zero
 #endif
-	EX(s16i, a9, a11, 0, fixup_s)
+EX(10f)	s16i	a9, a11, 0
 	addi	a11, a11, 2		# advance dst pointer
 .L100:
 	bbci.l	a4, 0, .Lret
-	EX(l8ui, a9, a3, 0, fixup_l)
+EX(11f)	l8ui	a9, a3, 0
 	/* slot */
-	EX(s8i, a9, a11, 0, fixup_s)
+EX(10f)	s8i	a9, a11, 0
 	beqz	a9, .Lret		# if byte is zero
 	addi	a11, a11, 1-3		# advance dst ptr 1, but also cancel
 					# the effect of adding 3 in .Lz3 code
@@ -161,14 +154,14 @@ __strncpy_user:
 #ifdef __XTENSA_EB__
 	movi	a9, 0
 #endif /* __XTENSA_EB__ */
-	EX(s8i, a9, a11, 0, fixup_s)
+EX(10f)	s8i	a9, a11, 0
 	sub	a2, a11, a2		# compute strlen
 	retw
 .Lz1:	# byte 1 is zero
 #ifdef __XTENSA_EB__
 	extui   a9, a9, 16, 16
 #endif /* __XTENSA_EB__ */
-	EX(s16i, a9, a11, 0, fixup_s)
+EX(10f)	s16i	a9, a11, 0
 	addi	a11, a11, 1		# advance dst pointer
 	sub	a2, a11, a2		# compute strlen
 	retw
@@ -176,9 +169,9 @@ __strncpy_user:
 #ifdef __XTENSA_EB__
 	extui   a9, a9, 16, 16
 #endif /* __XTENSA_EB__ */
-	EX(s16i, a9, a11, 0, fixup_s)
+EX(10f)	s16i	a9, a11, 0
 	movi	a9, 0
-	EX(s8i, a9, a11, 2, fixup_s)
+EX(10f)	s8i	a9, a11, 2
 	addi	a11, a11, 2		# advance dst pointer
 	sub	a2, a11, a2		# compute strlen
 	retw
@@ -196,9 +189,9 @@ __strncpy_user:
 	add	a12, a11, a4		# a12 = ending address
 #endif /* XCHAL_HAVE_LOOPS */
 .Lnextbyte:
-	EX(l8ui, a9, a3, 0, fixup_l)
+EX(11f)	l8ui	a9, a3, 0
 	addi	a3, a3, 1
-	EX(s8i, a9, a11, 0, fixup_s)
+EX(10f)	s8i	a9, a11, 0
 	beqz	a9, .Lunalignedend
 	addi	a11, a11, 1
 #if !XCHAL_HAVE_LOOPS
@@ -218,8 +211,7 @@ __strncpy_user:
 	 * implementation in memset().  Thus, we differentiate between
 	 * load/store fixups. */
 
-fixup_s:
-fixup_l:
+10:
+11:
 	movi	a2, -EFAULT
 	retw
-

arch/xtensa/lib/strnlen_user.S

@@ -12,14 +12,7 @@
  */
 
 #include <variant/core.h>
-
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler)	\
-9:	insn	reg1, reg2, offset;		\
-	.section __ex_table, "a";		\
-	.word	9b, handler;			\
-	.previous
+#include <asm/asmmacro.h>
 
 /*
  * size_t __strnlen_user(const char *s, size_t len)
@@ -77,7 +70,7 @@ __strnlen_user:
 	add	a10, a10, a4	# a10 = end of last 4B chunk
 #endif /* XCHAL_HAVE_LOOPS */
 .Loop:
-	EX(l32i, a9, a4, 4, lenfixup)	# get next word of string
+EX(10f)	l32i	a9, a4, 4		# get next word of string
 	addi	a4, a4, 4		# advance string pointer
 	bnone	a9, a5, .Lz0		# if byte 0 is zero
 	bnone	a9, a6, .Lz1		# if byte 1 is zero
@@ -88,7 +81,7 @@ __strnlen_user:
 #endif
 
 .Ldone:
-	EX(l32i, a9, a4, 4, lenfixup)	# load 4 bytes for remaining checks
+EX(10f)	l32i	a9, a4, 4	# load 4 bytes for remaining checks
 
 	bbci.l	a3, 1, .L100
 	# check two more bytes (bytes 0, 1 of word)
@@ -125,14 +118,14 @@ __strnlen_user:
 	retw
 
 .L1mod2:	# address is odd
-	EX(l8ui, a9, a4, 4, lenfixup)	# get byte 0
+EX(10f)	l8ui	a9, a4, 4		# get byte 0
 	addi	a4, a4, 1		# advance string pointer
 	beqz	a9, .Lz3		# if byte 0 is zero
 	bbci.l	a4, 1, .Laligned	# if string pointer is now word-aligned
 
 .L2mod4:	# address is 2 mod 4
 	addi	a4, a4, 2	# advance ptr for aligned access
-	EX(l32i, a9, a4, 0, lenfixup)	# get word with first two bytes of string
+EX(10f)	l32i	a9, a4, 0	# get word with first two bytes of string
 	bnone	a9, a7, .Lz2	# if byte 2 (of word, not string) is zero
 	bany	a9, a8, .Laligned # if byte 3 (of word, not string) is nonzero
 	# byte 3 is zero
@@ -142,6 +135,6 @@ __strnlen_user:
 
 	.section .fixup, "ax"
 	.align	4
-lenfixup:
+10:
 	movi	a2, 0
 	retw

arch/xtensa/lib/usercopy.S

@@ -54,6 +54,7 @@
  */
 
 #include <variant/core.h>
+#include <asm/asmmacro.h>
 
 #ifdef __XTENSA_EB__
 #define ALIGN(R, W0, W1) src	R, W0, W1
@@ -63,15 +64,6 @@
 #define SSA8(R)	ssa8l R
 #endif
 
-/* Load or store instructions that may cause exceptions use the EX macro. */
-
-#define EX(insn,reg1,reg2,offset,handler)	\
-9:	insn	reg1, reg2, offset;		\
-	.section __ex_table, "a";		\
-	.word	9b, handler;			\
-	.previous
-
-
 	.text
 	.align	4
 	.global	__xtensa_copy_user
@@ -102,9 +94,9 @@ __xtensa_copy_user:
 	bltui	a4, 7, .Lbytecopy	# do short copies byte by byte
 
 	# copy 1 byte
-	EX(l8ui, a6, a3, 0, fixup)
+EX(10f)	l8ui	a6, a3, 0
 	addi	a3, a3,  1
-	EX(s8i, a6, a5,  0, fixup)
+EX(10f)	s8i	a6, a5,  0
 	addi	a5, a5,  1
 	addi	a4, a4, -1
 	bbci.l	a5, 1, .Ldstaligned	# if dst is now aligned, then
@@ -112,11 +104,11 @@ __xtensa_copy_user:
 .Ldst2mod4:	# dst 16-bit aligned
 	# copy 2 bytes
 	bltui	a4, 6, .Lbytecopy	# do short copies byte by byte
-	EX(l8ui, a6, a3, 0, fixup)
-	EX(l8ui, a7, a3, 1, fixup)
+EX(10f)	l8ui	a6, a3, 0
+EX(10f)	l8ui	a7, a3, 1
 	addi	a3, a3,  2
-	EX(s8i, a6, a5,  0, fixup)
-	EX(s8i, a7, a5,  1, fixup)
+EX(10f)	s8i	a6, a5,  0
+EX(10f)	s8i	a7, a5,  1
 	addi	a5, a5,  2
 	addi	a4, a4, -2
 	j	.Ldstaligned	# dst is now aligned, return to main algorithm
@@ -135,9 +127,9 @@ __xtensa_copy_user:
 	add	a7, a3, a4	# a7 = end address for source
 #endif /* !XCHAL_HAVE_LOOPS */
 .Lnextbyte:
-	EX(l8ui, a6, a3, 0, fixup)
+EX(10f)	l8ui	a6, a3, 0
 	addi	a3, a3, 1
-	EX(s8i, a6, a5, 0, fixup)
+EX(10f)	s8i	a6, a5, 0
 	addi	a5, a5, 1
 #if !XCHAL_HAVE_LOOPS
 	blt	a3, a7, .Lnextbyte
@@ -161,15 +153,15 @@ __xtensa_copy_user:
 	add	a8, a8, a3	# a8 = end of last 16B source chunk
 #endif /* !XCHAL_HAVE_LOOPS */
 .Loop1:
-	EX(l32i, a6, a3,  0, fixup)
-	EX(l32i, a7, a3,  4, fixup)
-	EX(s32i, a6, a5,  0, fixup)
-	EX(l32i, a6, a3,  8, fixup)
-	EX(s32i, a7, a5,  4, fixup)
-	EX(l32i, a7, a3, 12, fixup)
-	EX(s32i, a6, a5,  8, fixup)
+EX(10f)	l32i	a6, a3,  0
+EX(10f)	l32i	a7, a3,  4
+EX(10f)	s32i	a6, a5,  0
+EX(10f)	l32i	a6, a3,  8
+EX(10f)	s32i	a7, a5,  4
+EX(10f)	l32i	a7, a3, 12
+EX(10f)	s32i	a6, a5,  8
 	addi	a3, a3, 16
-	EX(s32i, a7, a5, 12, fixup)
+EX(10f)	s32i	a7, a5, 12
 	addi	a5, a5, 16
 #if !XCHAL_HAVE_LOOPS
 	blt	a3, a8, .Loop1
@@ -177,31 +169,31 @@ __xtensa_copy_user:
 .Loop1done:
 	bbci.l	a4, 3, .L2
 	# copy 8 bytes
-	EX(l32i, a6, a3,  0, fixup)
-	EX(l32i, a7, a3,  4, fixup)
+EX(10f)	l32i	a6, a3,  0
+EX(10f)	l32i	a7, a3,  4
 	addi	a3, a3,  8
-	EX(s32i, a6, a5,  0, fixup)
-	EX(s32i, a7, a5,  4, fixup)
+EX(10f)	s32i	a6, a5,  0
+EX(10f)	s32i	a7, a5,  4
 	addi	a5, a5,  8
 .L2:
 	bbci.l	a4, 2, .L3
 	# copy 4 bytes
-	EX(l32i, a6, a3,  0, fixup)
+EX(10f)	l32i	a6, a3,  0
 	addi	a3, a3,  4
-	EX(s32i, a6, a5,  0, fixup)
+EX(10f)	s32i	a6, a5,  0
 	addi	a5, a5,  4
 .L3:
 	bbci.l	a4, 1, .L4
 	# copy 2 bytes
-	EX(l16ui, a6, a3,  0, fixup)
+EX(10f)	l16ui	a6, a3,  0
 	addi	a3, a3,  2
-	EX(s16i,  a6, a5,  0, fixup)
+EX(10f)	s16i	a6, a5,  0
 	addi	a5, a5,  2
 .L4:
 	bbci.l	a4, 0, .L5
 	# copy 1 byte
-	EX(l8ui, a6, a3,  0, fixup)
-	EX(s8i,  a6, a5,  0, fixup)
+EX(10f)	l8ui	a6, a3,  0
+EX(10f)	s8i	a6, a5,  0
 .L5:
 	movi	a2, 0		# return success for len bytes copied
 	retw
@@ -217,7 +209,7 @@ __xtensa_copy_user:
 	# copy 16 bytes per iteration for word-aligned dst and unaligned src
 	and	a10, a3, a8	# save unalignment offset for below
 	sub	a3, a3, a10	# align a3 (to avoid sim warnings only; not needed for hardware)
-	EX(l32i, a6, a3, 0, fixup)	# load first word
+EX(10f)	l32i	a6, a3, 0	# load first word
 #if XCHAL_HAVE_LOOPS
 	loopnez	a7, .Loop2done
 #else /* !XCHAL_HAVE_LOOPS */
@@ -226,19 +218,19 @@ __xtensa_copy_user:
 	add	a12, a12, a3	# a12 = end of last 16B source chunk
 #endif /* !XCHAL_HAVE_LOOPS */
 .Loop2:
-	EX(l32i, a7, a3,  4, fixup)
-	EX(l32i, a8, a3,  8, fixup)
+EX(10f)	l32i	a7, a3,  4
+EX(10f)	l32i	a8, a3,  8
 	ALIGN(	a6, a6, a7)
-	EX(s32i, a6, a5,  0, fixup)
-	EX(l32i, a9, a3, 12, fixup)
+EX(10f)	s32i	a6, a5,  0
+EX(10f)	l32i	a9, a3, 12
 	ALIGN(	a7, a7, a8)
-	EX(s32i, a7, a5,  4, fixup)
-	EX(l32i, a6, a3, 16, fixup)
+EX(10f)	s32i	a7, a5,  4
+EX(10f)	l32i	a6, a3, 16
 	ALIGN(	a8, a8, a9)
-	EX(s32i, a8, a5,  8, fixup)
+EX(10f)	s32i	a8, a5,  8
 	addi	a3, a3, 16
 	ALIGN(	a9, a9, a6)
-	EX(s32i, a9, a5, 12, fixup)
+EX(10f)	s32i	a9, a5, 12
 	addi	a5, a5, 16
 #if !XCHAL_HAVE_LOOPS
 	blt	a3, a12, .Loop2
@@ -246,39 +238,39 @@ __xtensa_copy_user:
 .Loop2done:
 	bbci.l	a4, 3, .L12
 	# copy 8 bytes
-	EX(l32i, a7, a3,  4, fixup)
-	EX(l32i, a8, a3,  8, fixup)
+EX(10f)	l32i	a7, a3,  4
+EX(10f)	l32i	a8, a3,  8
 	ALIGN(	a6, a6, a7)
-	EX(s32i, a6, a5,  0, fixup)
+EX(10f)	s32i	a6, a5,  0
 	addi	a3, a3,  8
 	ALIGN(	a7, a7, a8)
-	EX(s32i, a7, a5,  4, fixup)
+EX(10f)	s32i	a7, a5,  4
 	addi	a5, a5,  8
 	mov	a6, a8
 .L12:
 	bbci.l	a4, 2, .L13
 	# copy 4 bytes
-	EX(l32i, a7, a3,  4, fixup)
+EX(10f)	l32i	a7, a3,  4
 	addi	a3, a3,  4
 	ALIGN(	a6, a6, a7)
-	EX(s32i, a6, a5,  0, fixup)
+EX(10f)	s32i	a6, a5,  0
 	addi	a5, a5,  4
 	mov	a6, a7
 .L13:
 	add	a3, a3, a10	# readjust a3 with correct misalignment
 	bbci.l	a4, 1, .L14
 	# copy 2 bytes
-	EX(l8ui, a6, a3,  0, fixup)
-	EX(l8ui, a7, a3,  1, fixup)
+EX(10f)	l8ui	a6, a3,  0
+EX(10f)	l8ui	a7, a3,  1
 	addi	a3, a3,  2
-	EX(s8i, a6, a5,  0, fixup)
-	EX(s8i, a7, a5,  1, fixup)
+EX(10f)	s8i	a6, a5,  0
+EX(10f)	s8i	a7, a5,  1
 	addi	a5, a5,  2
 .L14:
 	bbci.l	a4, 0, .L15
 	# copy 1 byte
-	EX(l8ui, a6, a3,  0, fixup)
-	EX(s8i,  a6, a5,  0, fixup)
+EX(10f)	l8ui	a6, a3,  0
+EX(10f)	s8i	a6, a5,  0
 .L15:
 	movi	a2, 0		# return success for len bytes copied
 	retw
@@ -294,7 +286,7 @@ __xtensa_copy_user:
  */
 
 
-fixup:
+10:
 	sub	a2, a5, a2	/* a2 <-- bytes copied */
 	sub	a2, a11, a2	/* a2 <-- bytes not copied */
 	retw