arch/xtensa/lib/strncpy_user.S

root/arch/xtensa/lib/strncpy_user.S
/*
 *  arch/xtensa/lib/strncpy_user.S
 *
 *  This file is subject to the terms and conditions of the GNU General
 *  Public License.  See the file "COPYING" in the main directory of
 *  this archive for more details.
 *
 *  Returns: -EFAULT if exception before terminator, N if the entire
 *  buffer filled, else strlen.
 *
 *  Copyright (C) 2002 Tensilica Inc.
 */

#include <linux/errno.h>
#include <linux/linkage.h>
#include <asm/asmmacro.h>
#include <asm/core.h>

/*
 * char *__strncpy_user(char *dst, const char *src, size_t len)
 */

#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
#   a1/ stack pointer
#   a2/ return value
#   a3/ src
#   a4/ len
#   a5/ mask0
#   a6/ mask1
#   a7/ mask2
#   a8/ mask3
#   a9/ tmp
#   a10/ tmp
#   a11/ dst

.text
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
        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
        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(11f) l8ui    a9, a3, 0               # get byte 0
        addi    a3, a3, 1               # advance src pointer
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
        bbci.l  a3, 1, .Lsrcaligned     # if src is now word-aligned

.Lsrc2mod4:     # src address is 2 mod 4
EX(11f) l8ui    a9, a3, 0               # get byte 0
        /* 1-cycle interlock */
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(11f) l8ui    a9, a3, 1               # get byte 0
        addi    a3, a3, 2               # advance src pointer
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
        abi_ret_default

/*
 * dst is word-aligned, src is word-aligned
 */
        .align  4               # 1 mod 4 alignment for LOOPNEZ
        .byte   0               # (0 mod 4 alignment for LBEG)
.Laligned:
#if XCHAL_HAVE_LOOPS
        loopnez a10, .Loop1done
#else
        beqz    a10, .Loop1done
        slli    a10, a10, 2
        add     a10, a10, a11   # a10 = end of last 4B chunck
#endif
.Loop1:
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(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, a10, .Loop1
#endif

.Loop1done:
        bbci.l  a4, 1, .L100
        # copy 2 bytes
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
        bnone   a9, a8, .Lz1            # if byte 3 is zero
#else
        bnone   a9, a5, .Lz0            # if byte 0 is zero
        bnone   a9, a6, .Lz1            # if byte 1 is zero
#endif
EX(10f) s16i    a9, a11, 0
        addi    a11, a11, 2             # advance dst pointer
.L100:
        bbci.l  a4, 0, .Lret
EX(11f) l8ui    a9, a3, 0
        /* slot */
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
        /* fall thru to .Lz3 and "retw" */

.Lz3:   # byte 3 is zero
        addi    a11, a11, 3             # advance dst pointer
        sub     a2, a11, a2             # compute strlen
        abi_ret_default
.Lz0:   # byte 0 is zero
#ifdef __XTENSA_EB__
        movi    a9, 0
#endif /* __XTENSA_EB__ */
EX(10f) s8i     a9, a11, 0
        sub     a2, a11, a2             # compute strlen
        abi_ret_default
.Lz1:   # byte 1 is zero
#ifdef __XTENSA_EB__
        extui   a9, a9, 16, 16
#endif /* __XTENSA_EB__ */
EX(10f) s16i    a9, a11, 0
        addi    a11, a11, 1             # advance dst pointer
        sub     a2, a11, a2             # compute strlen
        abi_ret_default
.Lz2:   # byte 2 is zero
#ifdef __XTENSA_EB__
        extui   a9, a9, 16, 16
#endif /* __XTENSA_EB__ */
EX(10f) s16i    a9, a11, 0
        movi    a9, 0
EX(10f) s8i     a9, a11, 2
        addi    a11, a11, 2             # advance dst pointer
        sub     a2, a11, a2             # compute strlen
        abi_ret_default

        .align  4               # 1 mod 4 alignment for LOOPNEZ
        .byte   0               # (0 mod 4 alignment for LBEG)
.Ldstunaligned:
/*
 * for now just use byte copy loop
 */
#if XCHAL_HAVE_LOOPS
        loopnez a4, .Lunalignedend
#else
        beqz    a4, .Lunalignedend
        add     a10, a11, a4            # a10 = ending address
#endif /* XCHAL_HAVE_LOOPS */
.Lnextbyte:
EX(11f) l8ui    a9, a3, 0
        addi    a3, a3, 1
EX(10f) s8i     a9, a11, 0
        beqz    a9, .Lunalignedend
        addi    a11, a11, 1
#if !XCHAL_HAVE_LOOPS
        blt     a11, a10, .Lnextbyte
#endif

.Lunalignedend:
        sub     a2, a11, a2             # compute strlen
        abi_ret_default

ENDPROC(__strncpy_user)
EXPORT_SYMBOL(__strncpy_user)

        .section .fixup, "ax"
        .align  4

        /* For now, just return -EFAULT.  Future implementations might
         * like to clear remaining kernel space, like the fixup
         * implementation in memset().  Thus, we differentiate between
         * load/store fixups. */

10:
11:
        movi    a2, -EFAULT
        abi_ret_default
Linux
