1 // SPDX-License-Identifier: GPL-2.0-or-later 2 /* 3 * SM4 Cipher Algorithm, AES-NI/AVX2 optimized. 4 * as specified in 5 * https://tools.ietf.org/id/draft-ribose-cfrg-sm4-10.html 6 * 7 * Copyright (C) 2018 Markku-Juhani O. Saarinen <mjos@iki.fi> 8 * Copyright (C) 2020 Jussi Kivilinna <jussi.kivilinna@iki.fi> 9 * Copyright (c) 2021 Tianjia Zhang <tianjia.zhang@linux.alibaba.com> 10 */ 11 12 /* Based on SM4 AES-NI work by libgcrypt and Markku-Juhani O. Saarinen at: 13 * https://github.com/mjosaarinen/sm4ni 14 */ 15 16 #include <linux/linkage.h> 17 #include <linux/cfi_types.h> 18 #include <asm/frame.h> 19 20 #define rRIP (%rip) 21 22 /* vector registers */ 23 #define RX0 %ymm0 24 #define RX1 %ymm1 25 #define MASK_4BIT %ymm2 26 #define RTMP0 %ymm3 27 #define RTMP1 %ymm4 28 #define RTMP2 %ymm5 29 #define RTMP3 %ymm6 30 #define RTMP4 %ymm7 31 32 #define RA0 %ymm8 33 #define RA1 %ymm9 34 #define RA2 %ymm10 35 #define RA3 %ymm11 36 37 #define RB0 %ymm12 38 #define RB1 %ymm13 39 #define RB2 %ymm14 40 #define RB3 %ymm15 41 42 #define RNOT %ymm0 43 #define RBSWAP %ymm1 44 45 #define RX0x %xmm0 46 #define RX1x %xmm1 47 #define MASK_4BITx %xmm2 48 49 #define RNOTx %xmm0 50 #define RBSWAPx %xmm1 51 52 #define RTMP0x %xmm3 53 #define RTMP1x %xmm4 54 #define RTMP2x %xmm5 55 #define RTMP3x %xmm6 56 #define RTMP4x %xmm7 57 58 59 /* helper macros */ 60 61 /* Transpose four 32-bit words between 128-bit vector lanes. */ 62 #define transpose_4x4(x0, x1, x2, x3, t1, t2) \ 63 vpunpckhdq x1, x0, t2; \ 64 vpunpckldq x1, x0, x0; \ 65 \ 66 vpunpckldq x3, x2, t1; \ 67 vpunpckhdq x3, x2, x2; \ 68 \ 69 vpunpckhqdq t1, x0, x1; \ 70 vpunpcklqdq t1, x0, x0; \ 71 \ 72 vpunpckhqdq x2, t2, x3; \ 73 vpunpcklqdq x2, t2, x2; 74 75 /* post-SubByte transform. */ 76 #define transform_pre(x, lo_t, hi_t, mask4bit, tmp0) \ 77 vpand x, mask4bit, tmp0; \ 78 vpandn x, mask4bit, x; \ 79 vpsrld $4, x, x; \ 80 \ 81 vpshufb tmp0, lo_t, tmp0; \ 82 vpshufb x, hi_t, x; \ 83 vpxor tmp0, x, x; 84 85 /* post-SubByte transform. Note: x has been XOR'ed with mask4bit by 86 * 'vaeslastenc' instruction. */ 87 #define transform_post(x, lo_t, hi_t, mask4bit, tmp0) \ 88 vpandn mask4bit, x, tmp0; \ 89 vpsrld $4, x, x; \ 90 vpand x, mask4bit, x; \ 91 \ 92 vpshufb tmp0, lo_t, tmp0; \ 93 vpshufb x, hi_t, x; \ 94 vpxor tmp0, x, x; 95 96 97 .section .rodata.cst16, "aM", @progbits, 16 98 .align 16 99 100 /* 101 * Following four affine transform look-up tables are from work by 102 * Markku-Juhani O. Saarinen, at https://github.com/mjosaarinen/sm4ni 103 * 104 * These allow exposing SM4 S-Box from AES SubByte. 105 */ 106 107 /* pre-SubByte affine transform, from SM4 field to AES field. */ 108 .Lpre_tf_lo_s: 109 .quad 0x9197E2E474720701, 0xC7C1B4B222245157 110 .Lpre_tf_hi_s: 111 .quad 0xE240AB09EB49A200, 0xF052B91BF95BB012 112 113 /* post-SubByte affine transform, from AES field to SM4 field. */ 114 .Lpost_tf_lo_s: 115 .quad 0x5B67F2CEA19D0834, 0xEDD14478172BBE82 116 .Lpost_tf_hi_s: 117 .quad 0xAE7201DD73AFDC00, 0x11CDBE62CC1063BF 118 119 /* For isolating SubBytes from AESENCLAST, inverse shift row */ 120 .Linv_shift_row: 121 .byte 0x00, 0x0d, 0x0a, 0x07, 0x04, 0x01, 0x0e, 0x0b 122 .byte 0x08, 0x05, 0x02, 0x0f, 0x0c, 0x09, 0x06, 0x03 123 124 /* Inverse shift row + Rotate left by 8 bits on 32-bit words with vpshufb */ 125 .Linv_shift_row_rol_8: 126 .byte 0x07, 0x00, 0x0d, 0x0a, 0x0b, 0x04, 0x01, 0x0e 127 .byte 0x0f, 0x08, 0x05, 0x02, 0x03, 0x0c, 0x09, 0x06 128 129 /* Inverse shift row + Rotate left by 16 bits on 32-bit words with vpshufb */ 130 .Linv_shift_row_rol_16: 131 .byte 0x0a, 0x07, 0x00, 0x0d, 0x0e, 0x0b, 0x04, 0x01 132 .byte 0x02, 0x0f, 0x08, 0x05, 0x06, 0x03, 0x0c, 0x09 133 134 /* Inverse shift row + Rotate left by 24 bits on 32-bit words with vpshufb */ 135 .Linv_shift_row_rol_24: 136 .byte 0x0d, 0x0a, 0x07, 0x00, 0x01, 0x0e, 0x0b, 0x04 137 .byte 0x05, 0x02, 0x0f, 0x08, 0x09, 0x06, 0x03, 0x0c 138 139 /* For CTR-mode IV byteswap */ 140 .Lbswap128_mask: 141 .byte 15, 14, 13, 12, 11, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0 142 143 /* For input word byte-swap */ 144 .Lbswap32_mask: 145 .byte 3, 2, 1, 0, 7, 6, 5, 4, 11, 10, 9, 8, 15, 14, 13, 12 146 147 .align 4 148 /* 4-bit mask */ 149 .L0f0f0f0f: 150 .long 0x0f0f0f0f 151 152 /* 12 bytes, only for padding */ 153 .Lpadding_deadbeef: 154 .long 0xdeadbeef, 0xdeadbeef, 0xdeadbeef 155 156 .text 157 SYM_FUNC_START_LOCAL(__sm4_crypt_blk16) 158 /* input: 159 * %rdi: round key array, CTX 160 * RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3: sixteen parallel 161 * plaintext blocks 162 * output: 163 * RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3: sixteen parallel 164 * ciphertext blocks 165 */ 166 FRAME_BEGIN 167 168 vbroadcasti128 .Lbswap32_mask rRIP, RTMP2; 169 vpshufb RTMP2, RA0, RA0; 170 vpshufb RTMP2, RA1, RA1; 171 vpshufb RTMP2, RA2, RA2; 172 vpshufb RTMP2, RA3, RA3; 173 vpshufb RTMP2, RB0, RB0; 174 vpshufb RTMP2, RB1, RB1; 175 vpshufb RTMP2, RB2, RB2; 176 vpshufb RTMP2, RB3, RB3; 177 178 vpbroadcastd .L0f0f0f0f rRIP, MASK_4BIT; 179 transpose_4x4(RA0, RA1, RA2, RA3, RTMP0, RTMP1); 180 transpose_4x4(RB0, RB1, RB2, RB3, RTMP0, RTMP1); 181 182 #define ROUND(round, s0, s1, s2, s3, r0, r1, r2, r3) \ 183 vpbroadcastd (4*(round))(%rdi), RX0; \ 184 vbroadcasti128 .Lpre_tf_lo_s rRIP, RTMP4; \ 185 vbroadcasti128 .Lpre_tf_hi_s rRIP, RTMP1; \ 186 vmovdqa RX0, RX1; \ 187 vpxor s1, RX0, RX0; \ 188 vpxor s2, RX0, RX0; \ 189 vpxor s3, RX0, RX0; /* s1 ^ s2 ^ s3 ^ rk */ \ 190 vbroadcasti128 .Lpost_tf_lo_s rRIP, RTMP2; \ 191 vbroadcasti128 .Lpost_tf_hi_s rRIP, RTMP3; \ 192 vpxor r1, RX1, RX1; \ 193 vpxor r2, RX1, RX1; \ 194 vpxor r3, RX1, RX1; /* r1 ^ r2 ^ r3 ^ rk */ \ 195 \ 196 /* sbox, non-linear part */ \ 197 transform_pre(RX0, RTMP4, RTMP1, MASK_4BIT, RTMP0); \ 198 transform_pre(RX1, RTMP4, RTMP1, MASK_4BIT, RTMP0); \ 199 vextracti128 $1, RX0, RTMP4x; \ 200 vextracti128 $1, RX1, RTMP0x; \ 201 vaesenclast MASK_4BITx, RX0x, RX0x; \ 202 vaesenclast MASK_4BITx, RTMP4x, RTMP4x; \ 203 vaesenclast MASK_4BITx, RX1x, RX1x; \ 204 vaesenclast MASK_4BITx, RTMP0x, RTMP0x; \ 205 vinserti128 $1, RTMP4x, RX0, RX0; \ 206 vbroadcasti128 .Linv_shift_row rRIP, RTMP4; \ 207 vinserti128 $1, RTMP0x, RX1, RX1; \ 208 transform_post(RX0, RTMP2, RTMP3, MASK_4BIT, RTMP0); \ 209 transform_post(RX1, RTMP2, RTMP3, MASK_4BIT, RTMP0); \ 210 \ 211 /* linear part */ \ 212 vpshufb RTMP4, RX0, RTMP0; \ 213 vpxor RTMP0, s0, s0; /* s0 ^ x */ \ 214 vpshufb RTMP4, RX1, RTMP2; \ 215 vbroadcasti128 .Linv_shift_row_rol_8 rRIP, RTMP4; \ 216 vpxor RTMP2, r0, r0; /* r0 ^ x */ \ 217 vpshufb RTMP4, RX0, RTMP1; \ 218 vpxor RTMP1, RTMP0, RTMP0; /* x ^ rol(x,8) */ \ 219 vpshufb RTMP4, RX1, RTMP3; \ 220 vbroadcasti128 .Linv_shift_row_rol_16 rRIP, RTMP4; \ 221 vpxor RTMP3, RTMP2, RTMP2; /* x ^ rol(x,8) */ \ 222 vpshufb RTMP4, RX0, RTMP1; \ 223 vpxor RTMP1, RTMP0, RTMP0; /* x ^ rol(x,8) ^ rol(x,16) */ \ 224 vpshufb RTMP4, RX1, RTMP3; \ 225 vbroadcasti128 .Linv_shift_row_rol_24 rRIP, RTMP4; \ 226 vpxor RTMP3, RTMP2, RTMP2; /* x ^ rol(x,8) ^ rol(x,16) */ \ 227 vpshufb RTMP4, RX0, RTMP1; \ 228 vpxor RTMP1, s0, s0; /* s0 ^ x ^ rol(x,24) */ \ 229 vpslld $2, RTMP0, RTMP1; \ 230 vpsrld $30, RTMP0, RTMP0; \ 231 vpxor RTMP0, s0, s0; \ 232 /* s0 ^ x ^ rol(x,2) ^ rol(x,10) ^ rol(x,18) ^ rol(x,24) */ \ 233 vpxor RTMP1, s0, s0; \ 234 vpshufb RTMP4, RX1, RTMP3; \ 235 vpxor RTMP3, r0, r0; /* r0 ^ x ^ rol(x,24) */ \ 236 vpslld $2, RTMP2, RTMP3; \ 237 vpsrld $30, RTMP2, RTMP2; \ 238 vpxor RTMP2, r0, r0; \ 239 /* r0 ^ x ^ rol(x,2) ^ rol(x,10) ^ rol(x,18) ^ rol(x,24) */ \ 240 vpxor RTMP3, r0, r0; 241 242 leaq (32*4)(%rdi), %rax; 243 .align 16 244 .Lroundloop_blk8: 245 ROUND(0, RA0, RA1, RA2, RA3, RB0, RB1, RB2, RB3); 246 ROUND(1, RA1, RA2, RA3, RA0, RB1, RB2, RB3, RB0); 247 ROUND(2, RA2, RA3, RA0, RA1, RB2, RB3, RB0, RB1); 248 ROUND(3, RA3, RA0, RA1, RA2, RB3, RB0, RB1, RB2); 249 leaq (4*4)(%rdi), %rdi; 250 cmpq %rax, %rdi; 251 jne .Lroundloop_blk8; 252 253 #undef ROUND 254 255 vbroadcasti128 .Lbswap128_mask rRIP, RTMP2; 256 257 transpose_4x4(RA0, RA1, RA2, RA3, RTMP0, RTMP1); 258 transpose_4x4(RB0, RB1, RB2, RB3, RTMP0, RTMP1); 259 vpshufb RTMP2, RA0, RA0; 260 vpshufb RTMP2, RA1, RA1; 261 vpshufb RTMP2, RA2, RA2; 262 vpshufb RTMP2, RA3, RA3; 263 vpshufb RTMP2, RB0, RB0; 264 vpshufb RTMP2, RB1, RB1; 265 vpshufb RTMP2, RB2, RB2; 266 vpshufb RTMP2, RB3, RB3; 267 268 FRAME_END 269 RET; 270 SYM_FUNC_END(__sm4_crypt_blk16) 271 272 #define inc_le128(x, minus_one, tmp) \ 273 vpcmpeqq minus_one, x, tmp; \ 274 vpsubq minus_one, x, x; \ 275 vpslldq $8, tmp, tmp; \ 276 vpsubq tmp, x, x; 277 278 /* 279 * void sm4_aesni_avx2_ctr_enc_blk16(const u32 *rk, u8 *dst, 280 * const u8 *src, u8 *iv) 281 */ 282 SYM_TYPED_FUNC_START(sm4_aesni_avx2_ctr_enc_blk16) 283 /* input: 284 * %rdi: round key array, CTX 285 * %rsi: dst (16 blocks) 286 * %rdx: src (16 blocks) 287 * %rcx: iv (big endian, 128bit) 288 */ 289 FRAME_BEGIN 290 291 movq 8(%rcx), %rax; 292 bswapq %rax; 293 294 vzeroupper; 295 296 vbroadcasti128 .Lbswap128_mask rRIP, RTMP3; 297 vpcmpeqd RNOT, RNOT, RNOT; 298 vpsrldq $8, RNOT, RNOT; /* ab: -1:0 ; cd: -1:0 */ 299 vpaddq RNOT, RNOT, RTMP2; /* ab: -2:0 ; cd: -2:0 */ 300 301 /* load IV and byteswap */ 302 vmovdqu (%rcx), RTMP4x; 303 vpshufb RTMP3x, RTMP4x, RTMP4x; 304 vmovdqa RTMP4x, RTMP0x; 305 inc_le128(RTMP4x, RNOTx, RTMP1x); 306 vinserti128 $1, RTMP4x, RTMP0, RTMP0; 307 vpshufb RTMP3, RTMP0, RA0; /* +1 ; +0 */ 308 309 /* check need for handling 64-bit overflow and carry */ 310 cmpq $(0xffffffffffffffff - 16), %rax; 311 ja .Lhandle_ctr_carry; 312 313 /* construct IVs */ 314 vpsubq RTMP2, RTMP0, RTMP0; /* +3 ; +2 */ 315 vpshufb RTMP3, RTMP0, RA1; 316 vpsubq RTMP2, RTMP0, RTMP0; /* +5 ; +4 */ 317 vpshufb RTMP3, RTMP0, RA2; 318 vpsubq RTMP2, RTMP0, RTMP0; /* +7 ; +6 */ 319 vpshufb RTMP3, RTMP0, RA3; 320 vpsubq RTMP2, RTMP0, RTMP0; /* +9 ; +8 */ 321 vpshufb RTMP3, RTMP0, RB0; 322 vpsubq RTMP2, RTMP0, RTMP0; /* +11 ; +10 */ 323 vpshufb RTMP3, RTMP0, RB1; 324 vpsubq RTMP2, RTMP0, RTMP0; /* +13 ; +12 */ 325 vpshufb RTMP3, RTMP0, RB2; 326 vpsubq RTMP2, RTMP0, RTMP0; /* +15 ; +14 */ 327 vpshufb RTMP3, RTMP0, RB3; 328 vpsubq RTMP2, RTMP0, RTMP0; /* +16 */ 329 vpshufb RTMP3x, RTMP0x, RTMP0x; 330 331 jmp .Lctr_carry_done; 332 333 .Lhandle_ctr_carry: 334 /* construct IVs */ 335 inc_le128(RTMP0, RNOT, RTMP1); 336 inc_le128(RTMP0, RNOT, RTMP1); 337 vpshufb RTMP3, RTMP0, RA1; /* +3 ; +2 */ 338 inc_le128(RTMP0, RNOT, RTMP1); 339 inc_le128(RTMP0, RNOT, RTMP1); 340 vpshufb RTMP3, RTMP0, RA2; /* +5 ; +4 */ 341 inc_le128(RTMP0, RNOT, RTMP1); 342 inc_le128(RTMP0, RNOT, RTMP1); 343 vpshufb RTMP3, RTMP0, RA3; /* +7 ; +6 */ 344 inc_le128(RTMP0, RNOT, RTMP1); 345 inc_le128(RTMP0, RNOT, RTMP1); 346 vpshufb RTMP3, RTMP0, RB0; /* +9 ; +8 */ 347 inc_le128(RTMP0, RNOT, RTMP1); 348 inc_le128(RTMP0, RNOT, RTMP1); 349 vpshufb RTMP3, RTMP0, RB1; /* +11 ; +10 */ 350 inc_le128(RTMP0, RNOT, RTMP1); 351 inc_le128(RTMP0, RNOT, RTMP1); 352 vpshufb RTMP3, RTMP0, RB2; /* +13 ; +12 */ 353 inc_le128(RTMP0, RNOT, RTMP1); 354 inc_le128(RTMP0, RNOT, RTMP1); 355 vpshufb RTMP3, RTMP0, RB3; /* +15 ; +14 */ 356 inc_le128(RTMP0, RNOT, RTMP1); 357 vextracti128 $1, RTMP0, RTMP0x; 358 vpshufb RTMP3x, RTMP0x, RTMP0x; /* +16 */ 359 360 .align 4 361 .Lctr_carry_done: 362 /* store new IV */ 363 vmovdqu RTMP0x, (%rcx); 364 365 call __sm4_crypt_blk16; 366 367 vpxor (0 * 32)(%rdx), RA0, RA0; 368 vpxor (1 * 32)(%rdx), RA1, RA1; 369 vpxor (2 * 32)(%rdx), RA2, RA2; 370 vpxor (3 * 32)(%rdx), RA3, RA3; 371 vpxor (4 * 32)(%rdx), RB0, RB0; 372 vpxor (5 * 32)(%rdx), RB1, RB1; 373 vpxor (6 * 32)(%rdx), RB2, RB2; 374 vpxor (7 * 32)(%rdx), RB3, RB3; 375 376 vmovdqu RA0, (0 * 32)(%rsi); 377 vmovdqu RA1, (1 * 32)(%rsi); 378 vmovdqu RA2, (2 * 32)(%rsi); 379 vmovdqu RA3, (3 * 32)(%rsi); 380 vmovdqu RB0, (4 * 32)(%rsi); 381 vmovdqu RB1, (5 * 32)(%rsi); 382 vmovdqu RB2, (6 * 32)(%rsi); 383 vmovdqu RB3, (7 * 32)(%rsi); 384 385 vzeroall; 386 FRAME_END 387 RET; 388 SYM_FUNC_END(sm4_aesni_avx2_ctr_enc_blk16) 389 390 /* 391 * void sm4_aesni_avx2_cbc_dec_blk16(const u32 *rk, u8 *dst, 392 * const u8 *src, u8 *iv) 393 */ 394 SYM_TYPED_FUNC_START(sm4_aesni_avx2_cbc_dec_blk16) 395 /* input: 396 * %rdi: round key array, CTX 397 * %rsi: dst (16 blocks) 398 * %rdx: src (16 blocks) 399 * %rcx: iv 400 */ 401 FRAME_BEGIN 402 403 vzeroupper; 404 405 vmovdqu (0 * 32)(%rdx), RA0; 406 vmovdqu (1 * 32)(%rdx), RA1; 407 vmovdqu (2 * 32)(%rdx), RA2; 408 vmovdqu (3 * 32)(%rdx), RA3; 409 vmovdqu (4 * 32)(%rdx), RB0; 410 vmovdqu (5 * 32)(%rdx), RB1; 411 vmovdqu (6 * 32)(%rdx), RB2; 412 vmovdqu (7 * 32)(%rdx), RB3; 413 414 call __sm4_crypt_blk16; 415 416 vmovdqu (%rcx), RNOTx; 417 vinserti128 $1, (%rdx), RNOT, RNOT; 418 vpxor RNOT, RA0, RA0; 419 vpxor (0 * 32 + 16)(%rdx), RA1, RA1; 420 vpxor (1 * 32 + 16)(%rdx), RA2, RA2; 421 vpxor (2 * 32 + 16)(%rdx), RA3, RA3; 422 vpxor (3 * 32 + 16)(%rdx), RB0, RB0; 423 vpxor (4 * 32 + 16)(%rdx), RB1, RB1; 424 vpxor (5 * 32 + 16)(%rdx), RB2, RB2; 425 vpxor (6 * 32 + 16)(%rdx), RB3, RB3; 426 vmovdqu (7 * 32 + 16)(%rdx), RNOTx; 427 vmovdqu RNOTx, (%rcx); /* store new IV */ 428 429 vmovdqu RA0, (0 * 32)(%rsi); 430 vmovdqu RA1, (1 * 32)(%rsi); 431 vmovdqu RA2, (2 * 32)(%rsi); 432 vmovdqu RA3, (3 * 32)(%rsi); 433 vmovdqu RB0, (4 * 32)(%rsi); 434 vmovdqu RB1, (5 * 32)(%rsi); 435 vmovdqu RB2, (6 * 32)(%rsi); 436 vmovdqu RB3, (7 * 32)(%rsi); 437 438 vzeroall; 439 FRAME_END 440 RET; 441 SYM_FUNC_END(sm4_aesni_avx2_cbc_dec_blk16)
Linux® is a registered trademark of Linus Torvalds in the United States and other countries.
TOMOYO® is a registered trademark of NTT DATA CORPORATION.