^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 1) /* SPDX-License-Identifier: GPL-2.0 */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 2) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 3) * Normally compiler builtins are used, but sometimes the compiler calls out
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 4) * of line code. Based on asm-i386/string.h.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 5) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 6) * This assembly file is re-written from memmove_64.c file.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 7) * - Copyright 2011 Fenghua Yu <fenghua.yu@intel.com>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 8) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 9) #include <linux/linkage.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 10) #include <asm/cpufeatures.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 11) #include <asm/alternative-asm.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 12) #include <asm/export.h>
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 13)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 14) #undef memmove
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 15)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 16) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 17) * Implement memmove(). This can handle overlap between src and dst.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 18) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 19) * Input:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 20) * rdi: dest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 21) * rsi: src
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 22) * rdx: count
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 23) *
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 24) * Output:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 25) * rax: dest
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 26) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 27) SYM_FUNC_START_WEAK(memmove)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 28) SYM_FUNC_START(__memmove)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 29)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 30) mov %rdi, %rax
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 31)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 32) /* Decide forward/backward copy mode */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 33) cmp %rdi, %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 34) jge .Lmemmove_begin_forward
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 35) mov %rsi, %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 36) add %rdx, %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 37) cmp %rdi, %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 38) jg 2f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 39)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 40) /* FSRM implies ERMS => no length checks, do the copy directly */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 41) .Lmemmove_begin_forward:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 42) ALTERNATIVE "cmp $0x20, %rdx; jb 1f", "", X86_FEATURE_FSRM
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 43) ALTERNATIVE "", "movq %rdx, %rcx; rep movsb; retq", X86_FEATURE_ERMS
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 44)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 45) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 46) * movsq instruction have many startup latency
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 47) * so we handle small size by general register.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 48) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 49) cmp $680, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 50) jb 3f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 51) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 52) * movsq instruction is only good for aligned case.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 53) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 54)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 55) cmpb %dil, %sil
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 56) je 4f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 57) 3:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 58) sub $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 59) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 60) * We gobble 32 bytes forward in each loop.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 61) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 62) 5:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 63) sub $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 64) movq 0*8(%rsi), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 65) movq 1*8(%rsi), %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 66) movq 2*8(%rsi), %r9
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 67) movq 3*8(%rsi), %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 68) leaq 4*8(%rsi), %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 69)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 70) movq %r11, 0*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 71) movq %r10, 1*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 72) movq %r9, 2*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 73) movq %r8, 3*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 74) leaq 4*8(%rdi), %rdi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 75) jae 5b
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 76) addq $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 77) jmp 1f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 78) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 79) * Handle data forward by movsq.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 80) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 81) .p2align 4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 82) 4:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 83) movq %rdx, %rcx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 84) movq -8(%rsi, %rdx), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 85) lea -8(%rdi, %rdx), %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 86) shrq $3, %rcx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 87) rep movsq
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 88) movq %r11, (%r10)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 89) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 90) .Lmemmove_end_forward:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 91)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 92) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 93) * Handle data backward by movsq.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 94) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 95) .p2align 4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 96) 7:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 97) movq %rdx, %rcx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 98) movq (%rsi), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 99) movq %rdi, %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 100) leaq -8(%rsi, %rdx), %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 101) leaq -8(%rdi, %rdx), %rdi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 102) shrq $3, %rcx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 103) std
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 104) rep movsq
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 105) cld
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 106) movq %r11, (%r10)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 107) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 108)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 109) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 110) * Start to prepare for backward copy.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 111) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 112) .p2align 4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 113) 2:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 114) cmp $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 115) jb 1f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 116) cmp $680, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 117) jb 6f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 118) cmp %dil, %sil
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 119) je 7b
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 120) 6:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 121) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 122) * Calculate copy position to tail.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 123) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 124) addq %rdx, %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 125) addq %rdx, %rdi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 126) subq $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 127) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 128) * We gobble 32 bytes backward in each loop.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 129) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 130) 8:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 131) subq $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 132) movq -1*8(%rsi), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 133) movq -2*8(%rsi), %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 134) movq -3*8(%rsi), %r9
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 135) movq -4*8(%rsi), %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 136) leaq -4*8(%rsi), %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 137)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 138) movq %r11, -1*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 139) movq %r10, -2*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 140) movq %r9, -3*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 141) movq %r8, -4*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 142) leaq -4*8(%rdi), %rdi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 143) jae 8b
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 144) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 145) * Calculate copy position to head.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 146) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 147) addq $0x20, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 148) subq %rdx, %rsi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 149) subq %rdx, %rdi
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 150) 1:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 151) cmpq $16, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 152) jb 9f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 153) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 154) * Move data from 16 bytes to 31 bytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 155) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 156) movq 0*8(%rsi), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 157) movq 1*8(%rsi), %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 158) movq -2*8(%rsi, %rdx), %r9
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 159) movq -1*8(%rsi, %rdx), %r8
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 160) movq %r11, 0*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 161) movq %r10, 1*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 162) movq %r9, -2*8(%rdi, %rdx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 163) movq %r8, -1*8(%rdi, %rdx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 164) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 165) .p2align 4
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 166) 9:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 167) cmpq $8, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 168) jb 10f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 169) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 170) * Move data from 8 bytes to 15 bytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 171) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 172) movq 0*8(%rsi), %r11
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 173) movq -1*8(%rsi, %rdx), %r10
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 174) movq %r11, 0*8(%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 175) movq %r10, -1*8(%rdi, %rdx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 176) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 177) 10:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 178) cmpq $4, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 179) jb 11f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 180) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 181) * Move data from 4 bytes to 7 bytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 182) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 183) movl (%rsi), %r11d
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 184) movl -4(%rsi, %rdx), %r10d
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 185) movl %r11d, (%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 186) movl %r10d, -4(%rdi, %rdx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 187) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 188) 11:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 189) cmp $2, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 190) jb 12f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 191) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 192) * Move data from 2 bytes to 3 bytes.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 193) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 194) movw (%rsi), %r11w
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 195) movw -2(%rsi, %rdx), %r10w
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 196) movw %r11w, (%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 197) movw %r10w, -2(%rdi, %rdx)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 198) jmp 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 199) 12:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 200) cmp $1, %rdx
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 201) jb 13f
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 202) /*
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 203) * Move data for 1 byte.
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 204) */
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 205) movb (%rsi), %r11b
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 206) movb %r11b, (%rdi)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 207) 13:
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 208) retq
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 209) SYM_FUNC_END(__memmove)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 210) SYM_FUNC_END_ALIAS(memmove)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 211) EXPORT_SYMBOL(__memmove)
^8f3ce5b39 (kx 2023-10-28 12:00:06 +0300 212) EXPORT_SYMBOL(memmove)
Orange Pi5 kernel
Deprecated Linux kernel 5.10.110 for OrangePi 5/5B/5+ boards
3 Commits
0 Branches
0 Tags