summaryrefslogtreecommitdiff
path: root/arch/xtensa/lib/strncpy_user.S
diff options
context:
space:
mode:
Diffstat (limited to 'arch/xtensa/lib/strncpy_user.S')
-rw-r--r--arch/xtensa/lib/strncpy_user.S94
1 files changed, 43 insertions, 51 deletions
diff --git a/arch/xtensa/lib/strncpy_user.S b/arch/xtensa/lib/strncpy_user.S
index 1ad0ecf45368..9841d1694cdf 100644
--- a/arch/xtensa/lib/strncpy_user.S
+++ b/arch/xtensa/lib/strncpy_user.S
@@ -11,16 +11,10 @@
* 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 <linux/linkage.h>
+#include <asm/asmmacro.h>
+#include <asm/core.h>
/*
* char *__strncpy_user(char *dst, const char *src, size_t len)
@@ -51,14 +45,11 @@
# a9/ tmp
# a10/ tmp
# a11/ dst
-# a12/ tmp
.text
-.align 4
-.global __strncpy_user
-.type __strncpy_user,@function
-__strncpy_user:
- entry sp, 16 # minimal stack frame
+ENTRY(__strncpy_user)
+
+ abi_entry_default
# a2/ dst, a3/ src, a4/ len
mov a11, a2 # leave dst in return value register
beqz a4, .Lret # if len is zero
@@ -69,15 +60,15 @@ __strncpy_user:
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
- srli a12, a4, 2 # number of loop iterations with 4B per loop
+ srli a10, a4, 2 # number of loop iterations with 4B per loop
movi a9, 3
bnone a11, a9, .Laligned
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
@@ -85,23 +76,23 @@ __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
bnez a4, .Lsrcaligned # if len is nonzero
.Lret:
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
/*
* dst is word-aligned, src is word-aligned
@@ -110,29 +101,29 @@ __strncpy_user:
.byte 0 # (0 mod 4 alignment for LBEG)
.Laligned:
#if XCHAL_HAVE_LOOPS
- loopnez a12, .Loop1done
+ loopnez a10, .Loop1done
#else
- beqz a12, .Loop1done
- slli a12, a12, 2
- add a12, a12, a11 # a12 = end of last 4B chunck
+ beqz a10, .Loop1done
+ slli a10, a10, 2
+ add a10, a10, a11 # a10 = 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
- blt a11, a12, .Loop1
+ blt a11, a10, .Loop1
#endif
.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 +132,13 @@ __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
@@ -156,32 +147,32 @@ __strncpy_user:
.Lz3: # byte 3 is zero
addi a11, a11, 3 # advance dst pointer
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
.Lz0: # byte 0 is zero
#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
+ abi_ret_default
.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
+ abi_ret_default
.Lz2: # byte 2 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
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
+ abi_ret_default
.align 4 # 1 mod 4 alignment for LOOPNEZ
.byte 0 # (0 mod 4 alignment for LBEG)
@@ -193,22 +184,24 @@ __strncpy_user:
loopnez a4, .Lunalignedend
#else
beqz a4, .Lunalignedend
- add a12, a11, a4 # a12 = ending address
+ add a10, a11, a4 # a10 = 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
- blt a11, a12, .Lnextbyte
+ blt a11, a10, .Lnextbyte
#endif
.Lunalignedend:
sub a2, a11, a2 # compute strlen
- retw
+ abi_ret_default
+ENDPROC(__strncpy_user)
+EXPORT_SYMBOL(__strncpy_user)
.section .fixup, "ax"
.align 4
@@ -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
-
+ abi_ret_default