David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 1 | /* SPDX-License-Identifier: GPL-2.0-only */ |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 2 | /* |
| 3 | * linux/arch/arm/lib/csumpartialcopygeneric.S |
| 4 | * |
| 5 | * Copyright (C) 1995-2001 Russell King |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 6 | */ |
| 7 | #include <asm/assembler.h> |
| 8 | |
| 9 | /* |
| 10 | * unsigned int |
| 11 | * csum_partial_copy_xxx(const char *src, char *dst, int len, int sum, ) |
| 12 | * r0 = src, r1 = dst, r2 = len, r3 = sum |
| 13 | * Returns : r0 = checksum |
| 14 | * |
| 15 | * Note that 'tst' and 'teq' preserve the carry flag. |
| 16 | */ |
| 17 | |
| 18 | src .req r0 |
| 19 | dst .req r1 |
| 20 | len .req r2 |
| 21 | sum .req r3 |
| 22 | |
| 23 | .Lzero: mov r0, sum |
| 24 | load_regs |
| 25 | |
| 26 | /* |
| 27 | * Align an unaligned destination pointer. We know that |
| 28 | * we have >= 8 bytes here, so we don't need to check |
| 29 | * the length. Note that the source pointer hasn't been |
| 30 | * aligned yet. |
| 31 | */ |
| 32 | .Ldst_unaligned: |
| 33 | tst dst, #1 |
| 34 | beq .Ldst_16bit |
| 35 | |
| 36 | load1b ip |
| 37 | sub len, len, #1 |
| 38 | adcs sum, sum, ip, put_byte_1 @ update checksum |
| 39 | strb ip, [dst], #1 |
| 40 | tst dst, #2 |
| 41 | reteq lr @ dst is now 32bit aligned |
| 42 | |
| 43 | .Ldst_16bit: load2b r8, ip |
| 44 | sub len, len, #2 |
| 45 | adcs sum, sum, r8, put_byte_0 |
| 46 | strb r8, [dst], #1 |
| 47 | adcs sum, sum, ip, put_byte_1 |
| 48 | strb ip, [dst], #1 |
| 49 | ret lr @ dst is now 32bit aligned |
| 50 | |
| 51 | /* |
| 52 | * Handle 0 to 7 bytes, with any alignment of source and |
| 53 | * destination pointers. Note that when we get here, C = 0 |
| 54 | */ |
| 55 | .Lless8: teq len, #0 @ check for zero count |
| 56 | beq .Lzero |
| 57 | |
| 58 | /* we must have at least one byte. */ |
| 59 | tst dst, #1 @ dst 16-bit aligned |
| 60 | beq .Lless8_aligned |
| 61 | |
| 62 | /* Align dst */ |
| 63 | load1b ip |
| 64 | sub len, len, #1 |
| 65 | adcs sum, sum, ip, put_byte_1 @ update checksum |
| 66 | strb ip, [dst], #1 |
| 67 | tst len, #6 |
| 68 | beq .Lless8_byteonly |
| 69 | |
| 70 | 1: load2b r8, ip |
| 71 | sub len, len, #2 |
| 72 | adcs sum, sum, r8, put_byte_0 |
| 73 | strb r8, [dst], #1 |
| 74 | adcs sum, sum, ip, put_byte_1 |
| 75 | strb ip, [dst], #1 |
| 76 | .Lless8_aligned: |
| 77 | tst len, #6 |
| 78 | bne 1b |
| 79 | .Lless8_byteonly: |
| 80 | tst len, #1 |
| 81 | beq .Ldone |
| 82 | load1b r8 |
| 83 | adcs sum, sum, r8, put_byte_0 @ update checksum |
| 84 | strb r8, [dst], #1 |
| 85 | b .Ldone |
| 86 | |
| 87 | FN_ENTRY |
| 88 | save_regs |
| 89 | |
| 90 | cmp len, #8 @ Ensure that we have at least |
| 91 | blo .Lless8 @ 8 bytes to copy. |
| 92 | |
| 93 | adds sum, sum, #0 @ C = 0 |
| 94 | tst dst, #3 @ Test destination alignment |
| 95 | blne .Ldst_unaligned @ align destination, return here |
| 96 | |
| 97 | /* |
| 98 | * Ok, the dst pointer is now 32bit aligned, and we know |
| 99 | * that we must have more than 4 bytes to copy. Note |
| 100 | * that C contains the carry from the dst alignment above. |
| 101 | */ |
| 102 | |
| 103 | tst src, #3 @ Test source alignment |
| 104 | bne .Lsrc_not_aligned |
| 105 | |
| 106 | /* Routine for src & dst aligned */ |
| 107 | |
| 108 | bics ip, len, #15 |
| 109 | beq 2f |
| 110 | |
| 111 | 1: load4l r4, r5, r6, r7 |
| 112 | stmia dst!, {r4, r5, r6, r7} |
| 113 | adcs sum, sum, r4 |
| 114 | adcs sum, sum, r5 |
| 115 | adcs sum, sum, r6 |
| 116 | adcs sum, sum, r7 |
| 117 | sub ip, ip, #16 |
| 118 | teq ip, #0 |
| 119 | bne 1b |
| 120 | |
| 121 | 2: ands ip, len, #12 |
| 122 | beq 4f |
| 123 | tst ip, #8 |
| 124 | beq 3f |
| 125 | load2l r4, r5 |
| 126 | stmia dst!, {r4, r5} |
| 127 | adcs sum, sum, r4 |
| 128 | adcs sum, sum, r5 |
| 129 | tst ip, #4 |
| 130 | beq 4f |
| 131 | |
| 132 | 3: load1l r4 |
| 133 | str r4, [dst], #4 |
| 134 | adcs sum, sum, r4 |
| 135 | |
| 136 | 4: ands len, len, #3 |
| 137 | beq .Ldone |
| 138 | load1l r4 |
| 139 | tst len, #2 |
| 140 | mov r5, r4, get_byte_0 |
| 141 | beq .Lexit |
| 142 | adcs sum, sum, r4, lspush #16 |
| 143 | strb r5, [dst], #1 |
| 144 | mov r5, r4, get_byte_1 |
| 145 | strb r5, [dst], #1 |
| 146 | mov r5, r4, get_byte_2 |
| 147 | .Lexit: tst len, #1 |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 148 | strbne r5, [dst], #1 |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 149 | andne r5, r5, #255 |
David Brazdil | 0f672f6 | 2019-12-10 10:32:29 +0000 | [diff] [blame^] | 150 | adcsne sum, sum, r5, put_byte_0 |
Andrew Scull | b4b6d4a | 2019-01-02 15:54:55 +0000 | [diff] [blame] | 151 | |
| 152 | /* |
| 153 | * If the dst pointer was not 16-bit aligned, we |
| 154 | * need to rotate the checksum here to get around |
| 155 | * the inefficient byte manipulations in the |
| 156 | * architecture independent code. |
| 157 | */ |
| 158 | .Ldone: adc r0, sum, #0 |
| 159 | ldr sum, [sp, #0] @ dst |
| 160 | tst sum, #1 |
| 161 | movne r0, r0, ror #8 |
| 162 | load_regs |
| 163 | |
| 164 | .Lsrc_not_aligned: |
| 165 | adc sum, sum, #0 @ include C from dst alignment |
| 166 | and ip, src, #3 |
| 167 | bic src, src, #3 |
| 168 | load1l r5 |
| 169 | cmp ip, #2 |
| 170 | beq .Lsrc2_aligned |
| 171 | bhi .Lsrc3_aligned |
| 172 | mov r4, r5, lspull #8 @ C = 0 |
| 173 | bics ip, len, #15 |
| 174 | beq 2f |
| 175 | 1: load4l r5, r6, r7, r8 |
| 176 | orr r4, r4, r5, lspush #24 |
| 177 | mov r5, r5, lspull #8 |
| 178 | orr r5, r5, r6, lspush #24 |
| 179 | mov r6, r6, lspull #8 |
| 180 | orr r6, r6, r7, lspush #24 |
| 181 | mov r7, r7, lspull #8 |
| 182 | orr r7, r7, r8, lspush #24 |
| 183 | stmia dst!, {r4, r5, r6, r7} |
| 184 | adcs sum, sum, r4 |
| 185 | adcs sum, sum, r5 |
| 186 | adcs sum, sum, r6 |
| 187 | adcs sum, sum, r7 |
| 188 | mov r4, r8, lspull #8 |
| 189 | sub ip, ip, #16 |
| 190 | teq ip, #0 |
| 191 | bne 1b |
| 192 | 2: ands ip, len, #12 |
| 193 | beq 4f |
| 194 | tst ip, #8 |
| 195 | beq 3f |
| 196 | load2l r5, r6 |
| 197 | orr r4, r4, r5, lspush #24 |
| 198 | mov r5, r5, lspull #8 |
| 199 | orr r5, r5, r6, lspush #24 |
| 200 | stmia dst!, {r4, r5} |
| 201 | adcs sum, sum, r4 |
| 202 | adcs sum, sum, r5 |
| 203 | mov r4, r6, lspull #8 |
| 204 | tst ip, #4 |
| 205 | beq 4f |
| 206 | 3: load1l r5 |
| 207 | orr r4, r4, r5, lspush #24 |
| 208 | str r4, [dst], #4 |
| 209 | adcs sum, sum, r4 |
| 210 | mov r4, r5, lspull #8 |
| 211 | 4: ands len, len, #3 |
| 212 | beq .Ldone |
| 213 | mov r5, r4, get_byte_0 |
| 214 | tst len, #2 |
| 215 | beq .Lexit |
| 216 | adcs sum, sum, r4, lspush #16 |
| 217 | strb r5, [dst], #1 |
| 218 | mov r5, r4, get_byte_1 |
| 219 | strb r5, [dst], #1 |
| 220 | mov r5, r4, get_byte_2 |
| 221 | b .Lexit |
| 222 | |
| 223 | .Lsrc2_aligned: mov r4, r5, lspull #16 |
| 224 | adds sum, sum, #0 |
| 225 | bics ip, len, #15 |
| 226 | beq 2f |
| 227 | 1: load4l r5, r6, r7, r8 |
| 228 | orr r4, r4, r5, lspush #16 |
| 229 | mov r5, r5, lspull #16 |
| 230 | orr r5, r5, r6, lspush #16 |
| 231 | mov r6, r6, lspull #16 |
| 232 | orr r6, r6, r7, lspush #16 |
| 233 | mov r7, r7, lspull #16 |
| 234 | orr r7, r7, r8, lspush #16 |
| 235 | stmia dst!, {r4, r5, r6, r7} |
| 236 | adcs sum, sum, r4 |
| 237 | adcs sum, sum, r5 |
| 238 | adcs sum, sum, r6 |
| 239 | adcs sum, sum, r7 |
| 240 | mov r4, r8, lspull #16 |
| 241 | sub ip, ip, #16 |
| 242 | teq ip, #0 |
| 243 | bne 1b |
| 244 | 2: ands ip, len, #12 |
| 245 | beq 4f |
| 246 | tst ip, #8 |
| 247 | beq 3f |
| 248 | load2l r5, r6 |
| 249 | orr r4, r4, r5, lspush #16 |
| 250 | mov r5, r5, lspull #16 |
| 251 | orr r5, r5, r6, lspush #16 |
| 252 | stmia dst!, {r4, r5} |
| 253 | adcs sum, sum, r4 |
| 254 | adcs sum, sum, r5 |
| 255 | mov r4, r6, lspull #16 |
| 256 | tst ip, #4 |
| 257 | beq 4f |
| 258 | 3: load1l r5 |
| 259 | orr r4, r4, r5, lspush #16 |
| 260 | str r4, [dst], #4 |
| 261 | adcs sum, sum, r4 |
| 262 | mov r4, r5, lspull #16 |
| 263 | 4: ands len, len, #3 |
| 264 | beq .Ldone |
| 265 | mov r5, r4, get_byte_0 |
| 266 | tst len, #2 |
| 267 | beq .Lexit |
| 268 | adcs sum, sum, r4 |
| 269 | strb r5, [dst], #1 |
| 270 | mov r5, r4, get_byte_1 |
| 271 | strb r5, [dst], #1 |
| 272 | tst len, #1 |
| 273 | beq .Ldone |
| 274 | load1b r5 |
| 275 | b .Lexit |
| 276 | |
| 277 | .Lsrc3_aligned: mov r4, r5, lspull #24 |
| 278 | adds sum, sum, #0 |
| 279 | bics ip, len, #15 |
| 280 | beq 2f |
| 281 | 1: load4l r5, r6, r7, r8 |
| 282 | orr r4, r4, r5, lspush #8 |
| 283 | mov r5, r5, lspull #24 |
| 284 | orr r5, r5, r6, lspush #8 |
| 285 | mov r6, r6, lspull #24 |
| 286 | orr r6, r6, r7, lspush #8 |
| 287 | mov r7, r7, lspull #24 |
| 288 | orr r7, r7, r8, lspush #8 |
| 289 | stmia dst!, {r4, r5, r6, r7} |
| 290 | adcs sum, sum, r4 |
| 291 | adcs sum, sum, r5 |
| 292 | adcs sum, sum, r6 |
| 293 | adcs sum, sum, r7 |
| 294 | mov r4, r8, lspull #24 |
| 295 | sub ip, ip, #16 |
| 296 | teq ip, #0 |
| 297 | bne 1b |
| 298 | 2: ands ip, len, #12 |
| 299 | beq 4f |
| 300 | tst ip, #8 |
| 301 | beq 3f |
| 302 | load2l r5, r6 |
| 303 | orr r4, r4, r5, lspush #8 |
| 304 | mov r5, r5, lspull #24 |
| 305 | orr r5, r5, r6, lspush #8 |
| 306 | stmia dst!, {r4, r5} |
| 307 | adcs sum, sum, r4 |
| 308 | adcs sum, sum, r5 |
| 309 | mov r4, r6, lspull #24 |
| 310 | tst ip, #4 |
| 311 | beq 4f |
| 312 | 3: load1l r5 |
| 313 | orr r4, r4, r5, lspush #8 |
| 314 | str r4, [dst], #4 |
| 315 | adcs sum, sum, r4 |
| 316 | mov r4, r5, lspull #24 |
| 317 | 4: ands len, len, #3 |
| 318 | beq .Ldone |
| 319 | mov r5, r4, get_byte_0 |
| 320 | tst len, #2 |
| 321 | beq .Lexit |
| 322 | strb r5, [dst], #1 |
| 323 | adcs sum, sum, r4 |
| 324 | load1l r4 |
| 325 | mov r5, r4, get_byte_0 |
| 326 | strb r5, [dst], #1 |
| 327 | adcs sum, sum, r4, lspush #24 |
| 328 | mov r5, r4, get_byte_1 |
| 329 | b .Lexit |
| 330 | FN_EXIT |