1#if defined(__arm__) 2#include <openssl/arm_arch.h> 3 4@ Silence ARMv8 deprecated IT instruction warnings. This file is used by both 5@ ARMv7 and ARMv8 processors and does not use ARMv8 instructions. (ARMv8 PMULL 6@ instructions are in aesv8-armx.pl.) 7.arch armv7-a 8 9.text 10#if defined(__thumb2__) || defined(__clang__) 11.syntax unified 12#endif 13#if defined(__thumb2__) 14.thumb 15#else 16.code 32 17#endif 18 19#ifdef __clang__ 20#define ldrplb ldrbpl 21#define ldrneb ldrbne 22#endif 23 24.type rem_4bit,%object 25.align 5 26rem_4bit: 27.short 0x0000,0x1C20,0x3840,0x2460 28.short 0x7080,0x6CA0,0x48C0,0x54E0 29.short 0xE100,0xFD20,0xD940,0xC560 30.short 0x9180,0x8DA0,0xA9C0,0xB5E0 31.size rem_4bit,.-rem_4bit 32 33.type rem_4bit_get,%function 34rem_4bit_get: 35#if defined(__thumb2__) 36 adr r2,rem_4bit 37#else 38 sub r2,pc,#8+32 @ &rem_4bit 39#endif 40 b .Lrem_4bit_got 41 nop 42 nop 43.size rem_4bit_get,.-rem_4bit_get 44 45.globl gcm_ghash_4bit 46.hidden gcm_ghash_4bit 47.type gcm_ghash_4bit,%function 48.align 4 49gcm_ghash_4bit: 50#if defined(__thumb2__) 51 adr r12,rem_4bit 52#else 53 sub r12,pc,#8+48 @ &rem_4bit 54#endif 55 add r3,r2,r3 @ r3 to point at the end 56 stmdb sp!,{r3,r4,r5,r6,r7,r8,r9,r10,r11,lr} @ save r3/end too 57 58 ldmia r12,{r4,r5,r6,r7,r8,r9,r10,r11} @ copy rem_4bit ... 59 stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11} @ ... to stack 60 61 ldrb r12,[r2,#15] 62 ldrb r14,[r0,#15] 63.Louter: 64 eor r12,r12,r14 65 and r14,r12,#0xf0 66 and r12,r12,#0x0f 67 mov r3,#14 68 69 add r7,r1,r12,lsl#4 70 ldmia r7,{r4,r5,r6,r7} @ load Htbl[nlo] 71 add r11,r1,r14 72 ldrb r12,[r2,#14] 73 74 and r14,r4,#0xf @ rem 75 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nhi] 76 add r14,r14,r14 77 eor r4,r8,r4,lsr#4 78 ldrh r8,[sp,r14] @ rem_4bit[rem] 79 eor r4,r4,r5,lsl#28 80 ldrb r14,[r0,#14] 81 eor r5,r9,r5,lsr#4 82 eor r5,r5,r6,lsl#28 83 eor r6,r10,r6,lsr#4 84 eor r6,r6,r7,lsl#28 85 eor r7,r11,r7,lsr#4 86 eor r12,r12,r14 87 and r14,r12,#0xf0 88 and r12,r12,#0x0f 89 eor r7,r7,r8,lsl#16 90 91.Linner: 92 add r11,r1,r12,lsl#4 93 and r12,r4,#0xf @ rem 94 subs r3,r3,#1 95 add r12,r12,r12 96 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nlo] 97 eor r4,r8,r4,lsr#4 98 eor r4,r4,r5,lsl#28 99 eor r5,r9,r5,lsr#4 100 eor r5,r5,r6,lsl#28 101 ldrh r8,[sp,r12] @ rem_4bit[rem] 102 eor r6,r10,r6,lsr#4 103#ifdef __thumb2__ 104 it pl 105#endif 106 ldrplb r12,[r2,r3] 107 eor r6,r6,r7,lsl#28 108 eor r7,r11,r7,lsr#4 109 110 add r11,r1,r14 111 and r14,r4,#0xf @ rem 112 eor r7,r7,r8,lsl#16 @ ^= rem_4bit[rem] 113 add r14,r14,r14 114 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nhi] 115 eor r4,r8,r4,lsr#4 116#ifdef __thumb2__ 117 it pl 118#endif 119 ldrplb r8,[r0,r3] 120 eor r4,r4,r5,lsl#28 121 eor r5,r9,r5,lsr#4 122 ldrh r9,[sp,r14] 123 eor r5,r5,r6,lsl#28 124 eor r6,r10,r6,lsr#4 125 eor r6,r6,r7,lsl#28 126#ifdef __thumb2__ 127 it pl 128#endif 129 eorpl r12,r12,r8 130 eor r7,r11,r7,lsr#4 131#ifdef __thumb2__ 132 itt pl 133#endif 134 andpl r14,r12,#0xf0 135 andpl r12,r12,#0x0f 136 eor r7,r7,r9,lsl#16 @ ^= rem_4bit[rem] 137 bpl .Linner 138 139 ldr r3,[sp,#32] @ re-load r3/end 140 add r2,r2,#16 141 mov r14,r4 142#if __ARM_ARCH__>=7 && defined(__ARMEL__) 143 rev r4,r4 144 str r4,[r0,#12] 145#elif defined(__ARMEB__) 146 str r4,[r0,#12] 147#else 148 mov r9,r4,lsr#8 149 strb r4,[r0,#12+3] 150 mov r10,r4,lsr#16 151 strb r9,[r0,#12+2] 152 mov r11,r4,lsr#24 153 strb r10,[r0,#12+1] 154 strb r11,[r0,#12] 155#endif 156 cmp r2,r3 157#if __ARM_ARCH__>=7 && defined(__ARMEL__) 158 rev r5,r5 159 str r5,[r0,#8] 160#elif defined(__ARMEB__) 161 str r5,[r0,#8] 162#else 163 mov r9,r5,lsr#8 164 strb r5,[r0,#8+3] 165 mov r10,r5,lsr#16 166 strb r9,[r0,#8+2] 167 mov r11,r5,lsr#24 168 strb r10,[r0,#8+1] 169 strb r11,[r0,#8] 170#endif 171 172#ifdef __thumb2__ 173 it ne 174#endif 175 ldrneb r12,[r2,#15] 176#if __ARM_ARCH__>=7 && defined(__ARMEL__) 177 rev r6,r6 178 str r6,[r0,#4] 179#elif defined(__ARMEB__) 180 str r6,[r0,#4] 181#else 182 mov r9,r6,lsr#8 183 strb r6,[r0,#4+3] 184 mov r10,r6,lsr#16 185 strb r9,[r0,#4+2] 186 mov r11,r6,lsr#24 187 strb r10,[r0,#4+1] 188 strb r11,[r0,#4] 189#endif 190 191#if __ARM_ARCH__>=7 && defined(__ARMEL__) 192 rev r7,r7 193 str r7,[r0,#0] 194#elif defined(__ARMEB__) 195 str r7,[r0,#0] 196#else 197 mov r9,r7,lsr#8 198 strb r7,[r0,#0+3] 199 mov r10,r7,lsr#16 200 strb r9,[r0,#0+2] 201 mov r11,r7,lsr#24 202 strb r10,[r0,#0+1] 203 strb r11,[r0,#0] 204#endif 205 206 bne .Louter 207 208 add sp,sp,#36 209#if __ARM_ARCH__>=5 210 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,pc} 211#else 212 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,lr} 213 tst lr,#1 214 moveq pc,lr @ be binary compatible with V4, yet 215.word 0xe12fff1e @ interoperable with Thumb ISA:-) 216#endif 217.size gcm_ghash_4bit,.-gcm_ghash_4bit 218 219.globl gcm_gmult_4bit 220.hidden gcm_gmult_4bit 221.type gcm_gmult_4bit,%function 222gcm_gmult_4bit: 223 stmdb sp!,{r4,r5,r6,r7,r8,r9,r10,r11,lr} 224 ldrb r12,[r0,#15] 225 b rem_4bit_get 226.Lrem_4bit_got: 227 and r14,r12,#0xf0 228 and r12,r12,#0x0f 229 mov r3,#14 230 231 add r7,r1,r12,lsl#4 232 ldmia r7,{r4,r5,r6,r7} @ load Htbl[nlo] 233 ldrb r12,[r0,#14] 234 235 add r11,r1,r14 236 and r14,r4,#0xf @ rem 237 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nhi] 238 add r14,r14,r14 239 eor r4,r8,r4,lsr#4 240 ldrh r8,[r2,r14] @ rem_4bit[rem] 241 eor r4,r4,r5,lsl#28 242 eor r5,r9,r5,lsr#4 243 eor r5,r5,r6,lsl#28 244 eor r6,r10,r6,lsr#4 245 eor r6,r6,r7,lsl#28 246 eor r7,r11,r7,lsr#4 247 and r14,r12,#0xf0 248 eor r7,r7,r8,lsl#16 249 and r12,r12,#0x0f 250 251.Loop: 252 add r11,r1,r12,lsl#4 253 and r12,r4,#0xf @ rem 254 subs r3,r3,#1 255 add r12,r12,r12 256 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nlo] 257 eor r4,r8,r4,lsr#4 258 eor r4,r4,r5,lsl#28 259 eor r5,r9,r5,lsr#4 260 eor r5,r5,r6,lsl#28 261 ldrh r8,[r2,r12] @ rem_4bit[rem] 262 eor r6,r10,r6,lsr#4 263#ifdef __thumb2__ 264 it pl 265#endif 266 ldrplb r12,[r0,r3] 267 eor r6,r6,r7,lsl#28 268 eor r7,r11,r7,lsr#4 269 270 add r11,r1,r14 271 and r14,r4,#0xf @ rem 272 eor r7,r7,r8,lsl#16 @ ^= rem_4bit[rem] 273 add r14,r14,r14 274 ldmia r11,{r8,r9,r10,r11} @ load Htbl[nhi] 275 eor r4,r8,r4,lsr#4 276 eor r4,r4,r5,lsl#28 277 eor r5,r9,r5,lsr#4 278 ldrh r8,[r2,r14] @ rem_4bit[rem] 279 eor r5,r5,r6,lsl#28 280 eor r6,r10,r6,lsr#4 281 eor r6,r6,r7,lsl#28 282 eor r7,r11,r7,lsr#4 283#ifdef __thumb2__ 284 itt pl 285#endif 286 andpl r14,r12,#0xf0 287 andpl r12,r12,#0x0f 288 eor r7,r7,r8,lsl#16 @ ^= rem_4bit[rem] 289 bpl .Loop 290#if __ARM_ARCH__>=7 && defined(__ARMEL__) 291 rev r4,r4 292 str r4,[r0,#12] 293#elif defined(__ARMEB__) 294 str r4,[r0,#12] 295#else 296 mov r9,r4,lsr#8 297 strb r4,[r0,#12+3] 298 mov r10,r4,lsr#16 299 strb r9,[r0,#12+2] 300 mov r11,r4,lsr#24 301 strb r10,[r0,#12+1] 302 strb r11,[r0,#12] 303#endif 304 305#if __ARM_ARCH__>=7 && defined(__ARMEL__) 306 rev r5,r5 307 str r5,[r0,#8] 308#elif defined(__ARMEB__) 309 str r5,[r0,#8] 310#else 311 mov r9,r5,lsr#8 312 strb r5,[r0,#8+3] 313 mov r10,r5,lsr#16 314 strb r9,[r0,#8+2] 315 mov r11,r5,lsr#24 316 strb r10,[r0,#8+1] 317 strb r11,[r0,#8] 318#endif 319 320#if __ARM_ARCH__>=7 && defined(__ARMEL__) 321 rev r6,r6 322 str r6,[r0,#4] 323#elif defined(__ARMEB__) 324 str r6,[r0,#4] 325#else 326 mov r9,r6,lsr#8 327 strb r6,[r0,#4+3] 328 mov r10,r6,lsr#16 329 strb r9,[r0,#4+2] 330 mov r11,r6,lsr#24 331 strb r10,[r0,#4+1] 332 strb r11,[r0,#4] 333#endif 334 335#if __ARM_ARCH__>=7 && defined(__ARMEL__) 336 rev r7,r7 337 str r7,[r0,#0] 338#elif defined(__ARMEB__) 339 str r7,[r0,#0] 340#else 341 mov r9,r7,lsr#8 342 strb r7,[r0,#0+3] 343 mov r10,r7,lsr#16 344 strb r9,[r0,#0+2] 345 mov r11,r7,lsr#24 346 strb r10,[r0,#0+1] 347 strb r11,[r0,#0] 348#endif 349 350#if __ARM_ARCH__>=5 351 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,pc} 352#else 353 ldmia sp!,{r4,r5,r6,r7,r8,r9,r10,r11,lr} 354 tst lr,#1 355 moveq pc,lr @ be binary compatible with V4, yet 356.word 0xe12fff1e @ interoperable with Thumb ISA:-) 357#endif 358.size gcm_gmult_4bit,.-gcm_gmult_4bit 359#if __ARM_MAX_ARCH__>=7 360.arch armv7-a 361.fpu neon 362 363.globl gcm_init_neon 364.hidden gcm_init_neon 365.type gcm_init_neon,%function 366.align 4 367gcm_init_neon: 368 vld1.64 d7,[r1]! @ load H 369 vmov.i8 q8,#0xe1 370 vld1.64 d6,[r1] 371 vshl.i64 d17,#57 372 vshr.u64 d16,#63 @ t0=0xc2....01 373 vdup.8 q9,d7[7] 374 vshr.u64 d26,d6,#63 375 vshr.s8 q9,#7 @ broadcast carry bit 376 vshl.i64 q3,q3,#1 377 vand q8,q8,q9 378 vorr d7,d26 @ H<<<=1 379 veor q3,q3,q8 @ twisted H 380 vstmia r0,{q3} 381 382 bx lr @ bx lr 383.size gcm_init_neon,.-gcm_init_neon 384 385.globl gcm_gmult_neon 386.hidden gcm_gmult_neon 387.type gcm_gmult_neon,%function 388.align 4 389gcm_gmult_neon: 390 vld1.64 d7,[r0]! @ load Xi 391 vld1.64 d6,[r0]! 392 vmov.i64 d29,#0x0000ffffffffffff 393 vldmia r1,{d26,d27} @ load twisted H 394 vmov.i64 d30,#0x00000000ffffffff 395#ifdef __ARMEL__ 396 vrev64.8 q3,q3 397#endif 398 vmov.i64 d31,#0x000000000000ffff 399 veor d28,d26,d27 @ Karatsuba pre-processing 400 mov r3,#16 401 b .Lgmult_neon 402.size gcm_gmult_neon,.-gcm_gmult_neon 403 404.globl gcm_ghash_neon 405.hidden gcm_ghash_neon 406.type gcm_ghash_neon,%function 407.align 4 408gcm_ghash_neon: 409 vld1.64 d1,[r0]! @ load Xi 410 vld1.64 d0,[r0]! 411 vmov.i64 d29,#0x0000ffffffffffff 412 vldmia r1,{d26,d27} @ load twisted H 413 vmov.i64 d30,#0x00000000ffffffff 414#ifdef __ARMEL__ 415 vrev64.8 q0,q0 416#endif 417 vmov.i64 d31,#0x000000000000ffff 418 veor d28,d26,d27 @ Karatsuba pre-processing 419 420.Loop_neon: 421 vld1.64 d7,[r2]! @ load inp 422 vld1.64 d6,[r2]! 423#ifdef __ARMEL__ 424 vrev64.8 q3,q3 425#endif 426 veor q3,q0 @ inp^=Xi 427.Lgmult_neon: 428 vext.8 d16, d26, d26, #1 @ A1 429 vmull.p8 q8, d16, d6 @ F = A1*B 430 vext.8 d0, d6, d6, #1 @ B1 431 vmull.p8 q0, d26, d0 @ E = A*B1 432 vext.8 d18, d26, d26, #2 @ A2 433 vmull.p8 q9, d18, d6 @ H = A2*B 434 vext.8 d22, d6, d6, #2 @ B2 435 vmull.p8 q11, d26, d22 @ G = A*B2 436 vext.8 d20, d26, d26, #3 @ A3 437 veor q8, q8, q0 @ L = E + F 438 vmull.p8 q10, d20, d6 @ J = A3*B 439 vext.8 d0, d6, d6, #3 @ B3 440 veor q9, q9, q11 @ M = G + H 441 vmull.p8 q0, d26, d0 @ I = A*B3 442 veor d16, d16, d17 @ t0 = (L) (P0 + P1) << 8 443 vand d17, d17, d29 444 vext.8 d22, d6, d6, #4 @ B4 445 veor d18, d18, d19 @ t1 = (M) (P2 + P3) << 16 446 vand d19, d19, d30 447 vmull.p8 q11, d26, d22 @ K = A*B4 448 veor q10, q10, q0 @ N = I + J 449 veor d16, d16, d17 450 veor d18, d18, d19 451 veor d20, d20, d21 @ t2 = (N) (P4 + P5) << 24 452 vand d21, d21, d31 453 vext.8 q8, q8, q8, #15 454 veor d22, d22, d23 @ t3 = (K) (P6 + P7) << 32 455 vmov.i64 d23, #0 456 vext.8 q9, q9, q9, #14 457 veor d20, d20, d21 458 vmull.p8 q0, d26, d6 @ D = A*B 459 vext.8 q11, q11, q11, #12 460 vext.8 q10, q10, q10, #13 461 veor q8, q8, q9 462 veor q10, q10, q11 463 veor q0, q0, q8 464 veor q0, q0, q10 465 veor d6,d6,d7 @ Karatsuba pre-processing 466 vext.8 d16, d28, d28, #1 @ A1 467 vmull.p8 q8, d16, d6 @ F = A1*B 468 vext.8 d2, d6, d6, #1 @ B1 469 vmull.p8 q1, d28, d2 @ E = A*B1 470 vext.8 d18, d28, d28, #2 @ A2 471 vmull.p8 q9, d18, d6 @ H = A2*B 472 vext.8 d22, d6, d6, #2 @ B2 473 vmull.p8 q11, d28, d22 @ G = A*B2 474 vext.8 d20, d28, d28, #3 @ A3 475 veor q8, q8, q1 @ L = E + F 476 vmull.p8 q10, d20, d6 @ J = A3*B 477 vext.8 d2, d6, d6, #3 @ B3 478 veor q9, q9, q11 @ M = G + H 479 vmull.p8 q1, d28, d2 @ I = A*B3 480 veor d16, d16, d17 @ t0 = (L) (P0 + P1) << 8 481 vand d17, d17, d29 482 vext.8 d22, d6, d6, #4 @ B4 483 veor d18, d18, d19 @ t1 = (M) (P2 + P3) << 16 484 vand d19, d19, d30 485 vmull.p8 q11, d28, d22 @ K = A*B4 486 veor q10, q10, q1 @ N = I + J 487 veor d16, d16, d17 488 veor d18, d18, d19 489 veor d20, d20, d21 @ t2 = (N) (P4 + P5) << 24 490 vand d21, d21, d31 491 vext.8 q8, q8, q8, #15 492 veor d22, d22, d23 @ t3 = (K) (P6 + P7) << 32 493 vmov.i64 d23, #0 494 vext.8 q9, q9, q9, #14 495 veor d20, d20, d21 496 vmull.p8 q1, d28, d6 @ D = A*B 497 vext.8 q11, q11, q11, #12 498 vext.8 q10, q10, q10, #13 499 veor q8, q8, q9 500 veor q10, q10, q11 501 veor q1, q1, q8 502 veor q1, q1, q10 503 vext.8 d16, d27, d27, #1 @ A1 504 vmull.p8 q8, d16, d7 @ F = A1*B 505 vext.8 d4, d7, d7, #1 @ B1 506 vmull.p8 q2, d27, d4 @ E = A*B1 507 vext.8 d18, d27, d27, #2 @ A2 508 vmull.p8 q9, d18, d7 @ H = A2*B 509 vext.8 d22, d7, d7, #2 @ B2 510 vmull.p8 q11, d27, d22 @ G = A*B2 511 vext.8 d20, d27, d27, #3 @ A3 512 veor q8, q8, q2 @ L = E + F 513 vmull.p8 q10, d20, d7 @ J = A3*B 514 vext.8 d4, d7, d7, #3 @ B3 515 veor q9, q9, q11 @ M = G + H 516 vmull.p8 q2, d27, d4 @ I = A*B3 517 veor d16, d16, d17 @ t0 = (L) (P0 + P1) << 8 518 vand d17, d17, d29 519 vext.8 d22, d7, d7, #4 @ B4 520 veor d18, d18, d19 @ t1 = (M) (P2 + P3) << 16 521 vand d19, d19, d30 522 vmull.p8 q11, d27, d22 @ K = A*B4 523 veor q10, q10, q2 @ N = I + J 524 veor d16, d16, d17 525 veor d18, d18, d19 526 veor d20, d20, d21 @ t2 = (N) (P4 + P5) << 24 527 vand d21, d21, d31 528 vext.8 q8, q8, q8, #15 529 veor d22, d22, d23 @ t3 = (K) (P6 + P7) << 32 530 vmov.i64 d23, #0 531 vext.8 q9, q9, q9, #14 532 veor d20, d20, d21 533 vmull.p8 q2, d27, d7 @ D = A*B 534 vext.8 q11, q11, q11, #12 535 vext.8 q10, q10, q10, #13 536 veor q8, q8, q9 537 veor q10, q10, q11 538 veor q2, q2, q8 539 veor q2, q2, q10 540 veor q1,q1,q0 @ Karatsuba post-processing 541 veor q1,q1,q2 542 veor d1,d1,d2 543 veor d4,d4,d3 @ Xh|Xl - 256-bit result 544 545 @ equivalent of reduction_avx from ghash-x86_64.pl 546 vshl.i64 q9,q0,#57 @ 1st phase 547 vshl.i64 q10,q0,#62 548 veor q10,q10,q9 @ 549 vshl.i64 q9,q0,#63 550 veor q10, q10, q9 @ 551 veor d1,d1,d20 @ 552 veor d4,d4,d21 553 554 vshr.u64 q10,q0,#1 @ 2nd phase 555 veor q2,q2,q0 556 veor q0,q0,q10 @ 557 vshr.u64 q10,q10,#6 558 vshr.u64 q0,q0,#1 @ 559 veor q0,q0,q2 @ 560 veor q0,q0,q10 @ 561 562 subs r3,#16 563 bne .Loop_neon 564 565#ifdef __ARMEL__ 566 vrev64.8 q0,q0 567#endif 568 sub r0,#16 569 vst1.64 d1,[r0]! @ write out Xi 570 vst1.64 d0,[r0] 571 572 bx lr @ bx lr 573.size gcm_ghash_neon,.-gcm_ghash_neon 574#endif 575.byte 71,72,65,83,72,32,102,111,114,32,65,82,77,118,52,47,78,69,79,78,44,32,67,82,89,80,84,79,71,65,77,83,32,98,121,32,60,97,112,112,114,111,64,111,112,101,110,115,115,108,46,111,114,103,62,0 576.align 2 577.align 2 578#endif 579