arch/arm/lib/csumpartial.S

   1 /*
   2  *  linux/arch/arm/lib/csumpartial.S
   3  *
   4  *  Copyright (C) 1995-1998 Russell King
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License version 2 as
   8  * published by the Free Software Foundation.
   9  */
  10 #include <linux/linkage.h>
  11 #include <asm/assembler.h>
  12 #include <asm/export.h>
  13
  14                 .text
  15
  16 /*
  17  * Function: __u32 csum_partial(const char *src, int len, __u32 sum)
  18  * Params  : r0 = buffer, r1 = len, r2 = checksum
  19  * Returns : r0 = new checksum
  20  */
  21
  22 buf     .req    r0
  23 len     .req    r1
  24 sum     .req    r2
  25 td0     .req    r3
  26 td1     .req    r4      @ save before use
  27 td2     .req    r5      @ save before use
  28 td3     .req    lr
  29
  30 .Lzero:         mov     r0, sum
  31                 add     sp, sp, #4
  32                 ldr     pc, [sp], #4
  33
  34                 /*
  35                  * Handle 0 to 7 bytes, with any alignment of source and
  36                  * destination pointers.  Note that when we get here, C = 0
  37                  */
  38 .Lless8:                teq     len, #0                 @ check for zero count
  39                 beq     .Lzero
  40
  41                 /* we must have at least one byte. */
  42                 tst     buf, #1                 @ odd address?
  43                 movne   sum, sum, ror #8
  44                 ldrneb  td0, [buf], #1
  45                 subne   len, len, #1
  46                 adcnes  sum, sum, td0, put_byte_1
  47
  48 .Lless4:                tst     len, #6
  49                 beq     .Lless8_byte
  50
  51                 /* we are now half-word aligned */
  52
  53 .Lless8_wordlp:
  54 #if __LINUX_ARM_ARCH__ >= 4
  55                 ldrh    td0, [buf], #2
  56                 sub     len, len, #2
  57 #else
  58                 ldrb    td0, [buf], #1
  59                 ldrb    td3, [buf], #1
  60                 sub     len, len, #2
  61 #ifndef __ARMEB__
  62                 orr     td0, td0, td3, lsl #8
  63 #else
  64                 orr     td0, td3, td0, lsl #8
  65 #endif
  66 #endif
  67                 adcs    sum, sum, td0
  68                 tst     len, #6
  69                 bne     .Lless8_wordlp
  70
  71 .Lless8_byte:   tst     len, #1                 @ odd number of bytes
  72                 ldrneb  td0, [buf], #1          @ include last byte
  73                 adcnes  sum, sum, td0, put_byte_0       @ update checksum
  74
  75 .Ldone:         adc     r0, sum, #0             @ collect up the last carry
  76                 ldr     td0, [sp], #4
  77                 tst     td0, #1                 @ check buffer alignment
  78                 movne   r0, r0, ror #8          @ rotate checksum by 8 bits
  79                 ldr     pc, [sp], #4            @ return
  80
  81 .Lnot_aligned:  tst     buf, #1                 @ odd address
  82                 ldrneb  td0, [buf], #1          @ make even
  83                 subne   len, len, #1
  84                 adcnes  sum, sum, td0, put_byte_1       @ update checksum
  85
  86                 tst     buf, #2                 @ 32-bit aligned?
  87 #if __LINUX_ARM_ARCH__ >= 4
  88                 ldrneh  td0, [buf], #2          @ make 32-bit aligned
  89                 subne   len, len, #2
  90 #else
  91                 ldrneb  td0, [buf], #1
  92                 ldrneb  ip, [buf], #1
  93                 subne   len, len, #2
  94 #ifndef __ARMEB__
  95                 orrne   td0, td0, ip, lsl #8
  96 #else
  97                 orrne   td0, ip, td0, lsl #8
  98 #endif
  99 #endif
 100                 adcnes  sum, sum, td0           @ update checksum
 101                 ret     lr
 102
 103 ENTRY(csum_partial)
 104                 stmfd   sp!, {buf, lr}
 105                 cmp     len, #8                 @ Ensure that we have at least
 106                 blo     .Lless8                 @ 8 bytes to copy.
 107
 108                 tst     buf, #1
 109                 movne   sum, sum, ror #8
 110
 111                 adds    sum, sum, #0            @ C = 0
 112                 tst     buf, #3                 @ Test destination alignment
 113                 blne    .Lnot_aligned           @ align destination, return here
 114
 115 1:              bics    ip, len, #31
 116                 beq     3f
 117
 118                 stmfd   sp!, {r4 - r5}
 119 2:              ldmia   buf!, {td0, td1, td2, td3}
 120                 adcs    sum, sum, td0
 121                 adcs    sum, sum, td1
 122                 adcs    sum, sum, td2
 123                 adcs    sum, sum, td3
 124                 ldmia   buf!, {td0, td1, td2, td3}
 125                 adcs    sum, sum, td0
 126                 adcs    sum, sum, td1
 127                 adcs    sum, sum, td2
 128                 adcs    sum, sum, td3
 129                 sub     ip, ip, #32
 130                 teq     ip, #0
 131                 bne     2b
 132                 ldmfd   sp!, {r4 - r5}
 133
 134 3:              tst     len, #0x1c              @ should not change C
 135                 beq     .Lless4
 136
 137 4:              ldr     td0, [buf], #4
 138                 sub     len, len, #4
 139                 adcs    sum, sum, td0
 140                 tst     len, #0x1c
 141                 bne     4b
 142                 b       .Lless4
 143 ENDPROC(csum_partial)
 144 EXPORT_SYMBOL(csum_partial)