tjh.dev / kernel
Linux kernel mirror (for testing) git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git
kernel os linux
fork atom

crypto: powerpc/aes - assembler core

The assembler AES encryption and decryption core routines.
Implemented & optimized for big endian. Nevertheless they
work on little endian too.

For most efficient reuse in (higher level) block cipher
routines they are implemented as "fast" call modules without
any stack handling or register saving. The caller must
take care of that part.

Signed-off-by: Markus Stockhausen <stockhausen@collogia.de>
Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>

authored by

Markus Stockhausen and committed by
Herbert Xu 1c201e64 0c5f9aea

+351
+351
arch/powerpc/crypto/aes-spe-core.S
··· 1 + /* 2 + * Fast AES implementation for SPE instruction set (PPC) 3 + * 4 + * This code makes use of the SPE SIMD instruction set as defined in 5 + * http://cache.freescale.com/files/32bit/doc/ref_manual/SPEPIM.pdf 6 + * Implementation is based on optimization guide notes from 7 + * http://cache.freescale.com/files/32bit/doc/app_note/AN2665.pdf 8 + * 9 + * Copyright (c) 2015 Markus Stockhausen <stockhausen@collogia.de> 10 + * 11 + * This program is free software; you can redistribute it and/or modify it 12 + * under the terms of the GNU General Public License as published by the Free 13 + * Software Foundation; either version 2 of the License, or (at your option) 14 + * any later version. 15 + * 16 + */ 17 + 18 + #include <asm/ppc_asm.h> 19 + #include "aes-spe-regs.h" 20 + 21 + #define EAD(in, bpos) \ 22 + rlwimi rT0,in,28-((bpos+3)%4)*8,20,27; 23 + 24 + #define DAD(in, bpos) \ 25 + rlwimi rT1,in,24-((bpos+3)%4)*8,24,31; 26 + 27 + #define LWH(out, off) \ 28 + evlwwsplat out,off(rT0); /* load word high */ 29 + 30 + #define LWL(out, off) \ 31 + lwz out,off(rT0); /* load word low */ 32 + 33 + #define LBZ(out, tab, off) \ 34 + lbz out,off(tab); /* load byte */ 35 + 36 + #define LAH(out, in, bpos, off) \ 37 + EAD(in, bpos) /* calc addr + load word high */ \ 38 + LWH(out, off) 39 + 40 + #define LAL(out, in, bpos, off) \ 41 + EAD(in, bpos) /* calc addr + load word low */ \ 42 + LWL(out, off) 43 + 44 + #define LAE(out, in, bpos) \ 45 + EAD(in, bpos) /* calc addr + load enc byte */ \ 46 + LBZ(out, rT0, 8) 47 + 48 + #define LBE(out) \ 49 + LBZ(out, rT0, 8) /* load enc byte */ 50 + 51 + #define LAD(out, in, bpos) \ 52 + DAD(in, bpos) /* calc addr + load dec byte */ \ 53 + LBZ(out, rT1, 0) 54 + 55 + #define LBD(out) \ 56 + LBZ(out, rT1, 0) 57 + 58 + /* 59 + * ppc_encrypt_block: The central encryption function for a single 16 bytes 60 + * block. It does no stack handling or register saving to support fast calls 61 + * via bl/blr. It expects that caller has pre-xored input data with first 62 + * 4 words of encryption key into rD0-rD3. Pointer/counter registers must 63 + * have also been set up before (rT0, rKP, CTR). Output is stored in rD0-rD3 64 + * and rW0-rW3 and caller must execute a final xor on the ouput registers. 65 + * All working registers rD0-rD3 & rW0-rW7 are overwritten during processing. 66 + * 67 + */ 68 + _GLOBAL(ppc_encrypt_block) 69 + LAH(rW4, rD1, 2, 4) 70 + LAH(rW6, rD0, 3, 0) 71 + LAH(rW3, rD0, 1, 8) 72 + ppc_encrypt_block_loop: 73 + LAH(rW0, rD3, 0, 12) 74 + LAL(rW0, rD0, 0, 12) 75 + LAH(rW1, rD1, 0, 12) 76 + LAH(rW2, rD2, 1, 8) 77 + LAL(rW2, rD3, 1, 8) 78 + LAL(rW3, rD1, 1, 8) 79 + LAL(rW4, rD2, 2, 4) 80 + LAL(rW6, rD1, 3, 0) 81 + LAH(rW5, rD3, 2, 4) 82 + LAL(rW5, rD0, 2, 4) 83 + LAH(rW7, rD2, 3, 0) 84 + evldw rD1,16(rKP) 85 + EAD(rD3, 3) 86 + evxor rW2,rW2,rW4 87 + LWL(rW7, 0) 88 + evxor rW2,rW2,rW6 89 + EAD(rD2, 0) 90 + evxor rD1,rD1,rW2 91 + LWL(rW1, 12) 92 + evxor rD1,rD1,rW0 93 + evldw rD3,24(rKP) 94 + evmergehi rD0,rD0,rD1 95 + EAD(rD1, 2) 96 + evxor rW3,rW3,rW5 97 + LWH(rW4, 4) 98 + evxor rW3,rW3,rW7 99 + EAD(rD0, 3) 100 + evxor rD3,rD3,rW3 101 + LWH(rW6, 0) 102 + evxor rD3,rD3,rW1 103 + EAD(rD0, 1) 104 + evmergehi rD2,rD2,rD3 105 + LWH(rW3, 8) 106 + LAH(rW0, rD3, 0, 12) 107 + LAL(rW0, rD0, 0, 12) 108 + LAH(rW1, rD1, 0, 12) 109 + LAH(rW2, rD2, 1, 8) 110 + LAL(rW2, rD3, 1, 8) 111 + LAL(rW3, rD1, 1, 8) 112 + LAL(rW4, rD2, 2, 4) 113 + LAL(rW6, rD1, 3, 0) 114 + LAH(rW5, rD3, 2, 4) 115 + LAL(rW5, rD0, 2, 4) 116 + LAH(rW7, rD2, 3, 0) 117 + evldw rD1,32(rKP) 118 + EAD(rD3, 3) 119 + evxor rW2,rW2,rW4 120 + LWL(rW7, 0) 121 + evxor rW2,rW2,rW6 122 + EAD(rD2, 0) 123 + evxor rD1,rD1,rW2 124 + LWL(rW1, 12) 125 + evxor rD1,rD1,rW0 126 + evldw rD3,40(rKP) 127 + evmergehi rD0,rD0,rD1 128 + EAD(rD1, 2) 129 + evxor rW3,rW3,rW5 130 + LWH(rW4, 4) 131 + evxor rW3,rW3,rW7 132 + EAD(rD0, 3) 133 + evxor rD3,rD3,rW3 134 + LWH(rW6, 0) 135 + evxor rD3,rD3,rW1 136 + EAD(rD0, 1) 137 + evmergehi rD2,rD2,rD3 138 + LWH(rW3, 8) 139 + addi rKP,rKP,32 140 + bdnz ppc_encrypt_block_loop 141 + LAH(rW0, rD3, 0, 12) 142 + LAL(rW0, rD0, 0, 12) 143 + LAH(rW1, rD1, 0, 12) 144 + LAH(rW2, rD2, 1, 8) 145 + LAL(rW2, rD3, 1, 8) 146 + LAL(rW3, rD1, 1, 8) 147 + LAL(rW4, rD2, 2, 4) 148 + LAH(rW5, rD3, 2, 4) 149 + LAL(rW6, rD1, 3, 0) 150 + LAL(rW5, rD0, 2, 4) 151 + LAH(rW7, rD2, 3, 0) 152 + evldw rD1,16(rKP) 153 + EAD(rD3, 3) 154 + evxor rW2,rW2,rW4 155 + LWL(rW7, 0) 156 + evxor rW2,rW2,rW6 157 + EAD(rD2, 0) 158 + evxor rD1,rD1,rW2 159 + LWL(rW1, 12) 160 + evxor rD1,rD1,rW0 161 + evldw rD3,24(rKP) 162 + evmergehi rD0,rD0,rD1 163 + EAD(rD1, 0) 164 + evxor rW3,rW3,rW5 165 + LBE(rW2) 166 + evxor rW3,rW3,rW7 167 + EAD(rD0, 1) 168 + evxor rD3,rD3,rW3 169 + LBE(rW6) 170 + evxor rD3,rD3,rW1 171 + EAD(rD0, 0) 172 + evmergehi rD2,rD2,rD3 173 + LBE(rW1) 174 + LAE(rW0, rD3, 0) 175 + LAE(rW1, rD0, 0) 176 + LAE(rW4, rD2, 1) 177 + LAE(rW5, rD3, 1) 178 + LAE(rW3, rD2, 0) 179 + LAE(rW7, rD1, 1) 180 + rlwimi rW0,rW4,8,16,23 181 + rlwimi rW1,rW5,8,16,23 182 + LAE(rW4, rD1, 2) 183 + LAE(rW5, rD2, 2) 184 + rlwimi rW2,rW6,8,16,23 185 + rlwimi rW3,rW7,8,16,23 186 + LAE(rW6, rD3, 2) 187 + LAE(rW7, rD0, 2) 188 + rlwimi rW0,rW4,16,8,15 189 + rlwimi rW1,rW5,16,8,15 190 + LAE(rW4, rD0, 3) 191 + LAE(rW5, rD1, 3) 192 + rlwimi rW2,rW6,16,8,15 193 + lwz rD0,32(rKP) 194 + rlwimi rW3,rW7,16,8,15 195 + lwz rD1,36(rKP) 196 + LAE(rW6, rD2, 3) 197 + LAE(rW7, rD3, 3) 198 + rlwimi rW0,rW4,24,0,7 199 + lwz rD2,40(rKP) 200 + rlwimi rW1,rW5,24,0,7 201 + lwz rD3,44(rKP) 202 + rlwimi rW2,rW6,24,0,7 203 + rlwimi rW3,rW7,24,0,7 204 + blr 205 + 206 + /* 207 + * ppc_decrypt_block: The central decryption function for a single 16 bytes 208 + * block. It does no stack handling or register saving to support fast calls 209 + * via bl/blr. It expects that caller has pre-xored input data with first 210 + * 4 words of encryption key into rD0-rD3. Pointer/counter registers must 211 + * have also been set up before (rT0, rKP, CTR). Output is stored in rD0-rD3 212 + * and rW0-rW3 and caller must execute a final xor on the ouput registers. 213 + * All working registers rD0-rD3 & rW0-rW7 are overwritten during processing. 214 + * 215 + */ 216 + _GLOBAL(ppc_decrypt_block) 217 + LAH(rW0, rD1, 0, 12) 218 + LAH(rW6, rD0, 3, 0) 219 + LAH(rW3, rD0, 1, 8) 220 + ppc_decrypt_block_loop: 221 + LAH(rW1, rD3, 0, 12) 222 + LAL(rW0, rD2, 0, 12) 223 + LAH(rW2, rD2, 1, 8) 224 + LAL(rW2, rD3, 1, 8) 225 + LAH(rW4, rD3, 2, 4) 226 + LAL(rW4, rD0, 2, 4) 227 + LAL(rW6, rD1, 3, 0) 228 + LAH(rW5, rD1, 2, 4) 229 + LAH(rW7, rD2, 3, 0) 230 + LAL(rW7, rD3, 3, 0) 231 + LAL(rW3, rD1, 1, 8) 232 + evldw rD1,16(rKP) 233 + EAD(rD0, 0) 234 + evxor rW4,rW4,rW6 235 + LWL(rW1, 12) 236 + evxor rW0,rW0,rW4 237 + EAD(rD2, 2) 238 + evxor rW0,rW0,rW2 239 + LWL(rW5, 4) 240 + evxor rD1,rD1,rW0 241 + evldw rD3,24(rKP) 242 + evmergehi rD0,rD0,rD1 243 + EAD(rD1, 0) 244 + evxor rW3,rW3,rW7 245 + LWH(rW0, 12) 246 + evxor rW3,rW3,rW1 247 + EAD(rD0, 3) 248 + evxor rD3,rD3,rW3 249 + LWH(rW6, 0) 250 + evxor rD3,rD3,rW5 251 + EAD(rD0, 1) 252 + evmergehi rD2,rD2,rD3 253 + LWH(rW3, 8) 254 + LAH(rW1, rD3, 0, 12) 255 + LAL(rW0, rD2, 0, 12) 256 + LAH(rW2, rD2, 1, 8) 257 + LAL(rW2, rD3, 1, 8) 258 + LAH(rW4, rD3, 2, 4) 259 + LAL(rW4, rD0, 2, 4) 260 + LAL(rW6, rD1, 3, 0) 261 + LAH(rW5, rD1, 2, 4) 262 + LAH(rW7, rD2, 3, 0) 263 + LAL(rW7, rD3, 3, 0) 264 + LAL(rW3, rD1, 1, 8) 265 + evldw rD1,32(rKP) 266 + EAD(rD0, 0) 267 + evxor rW4,rW4,rW6 268 + LWL(rW1, 12) 269 + evxor rW0,rW0,rW4 270 + EAD(rD2, 2) 271 + evxor rW0,rW0,rW2 272 + LWL(rW5, 4) 273 + evxor rD1,rD1,rW0 274 + evldw rD3,40(rKP) 275 + evmergehi rD0,rD0,rD1 276 + EAD(rD1, 0) 277 + evxor rW3,rW3,rW7 278 + LWH(rW0, 12) 279 + evxor rW3,rW3,rW1 280 + EAD(rD0, 3) 281 + evxor rD3,rD3,rW3 282 + LWH(rW6, 0) 283 + evxor rD3,rD3,rW5 284 + EAD(rD0, 1) 285 + evmergehi rD2,rD2,rD3 286 + LWH(rW3, 8) 287 + addi rKP,rKP,32 288 + bdnz ppc_decrypt_block_loop 289 + LAH(rW1, rD3, 0, 12) 290 + LAL(rW0, rD2, 0, 12) 291 + LAH(rW2, rD2, 1, 8) 292 + LAL(rW2, rD3, 1, 8) 293 + LAH(rW4, rD3, 2, 4) 294 + LAL(rW4, rD0, 2, 4) 295 + LAL(rW6, rD1, 3, 0) 296 + LAH(rW5, rD1, 2, 4) 297 + LAH(rW7, rD2, 3, 0) 298 + LAL(rW7, rD3, 3, 0) 299 + LAL(rW3, rD1, 1, 8) 300 + evldw rD1,16(rKP) 301 + EAD(rD0, 0) 302 + evxor rW4,rW4,rW6 303 + LWL(rW1, 12) 304 + evxor rW0,rW0,rW4 305 + EAD(rD2, 2) 306 + evxor rW0,rW0,rW2 307 + LWL(rW5, 4) 308 + evxor rD1,rD1,rW0 309 + evldw rD3,24(rKP) 310 + evmergehi rD0,rD0,rD1 311 + DAD(rD1, 0) 312 + evxor rW3,rW3,rW7 313 + LBD(rW0) 314 + evxor rW3,rW3,rW1 315 + DAD(rD0, 1) 316 + evxor rD3,rD3,rW3 317 + LBD(rW6) 318 + evxor rD3,rD3,rW5 319 + DAD(rD0, 0) 320 + evmergehi rD2,rD2,rD3 321 + LBD(rW3) 322 + LAD(rW2, rD3, 0) 323 + LAD(rW1, rD2, 0) 324 + LAD(rW4, rD2, 1) 325 + LAD(rW5, rD3, 1) 326 + LAD(rW7, rD1, 1) 327 + rlwimi rW0,rW4,8,16,23 328 + rlwimi rW1,rW5,8,16,23 329 + LAD(rW4, rD3, 2) 330 + LAD(rW5, rD0, 2) 331 + rlwimi rW2,rW6,8,16,23 332 + rlwimi rW3,rW7,8,16,23 333 + LAD(rW6, rD1, 2) 334 + LAD(rW7, rD2, 2) 335 + rlwimi rW0,rW4,16,8,15 336 + rlwimi rW1,rW5,16,8,15 337 + LAD(rW4, rD0, 3) 338 + LAD(rW5, rD1, 3) 339 + rlwimi rW2,rW6,16,8,15 340 + lwz rD0,32(rKP) 341 + rlwimi rW3,rW7,16,8,15 342 + lwz rD1,36(rKP) 343 + LAD(rW6, rD2, 3) 344 + LAD(rW7, rD3, 3) 345 + rlwimi rW0,rW4,24,0,7 346 + lwz rD2,40(rKP) 347 + rlwimi rW1,rW5,24,0,7 348 + lwz rD3,44(rKP) 349 + rlwimi rW2,rW6,24,0,7 350 + rlwimi rW3,rW7,24,0,7 351 + blr