OLD | NEW |
(Empty) | |
| 1 Index: nss/lib/softoken/pkcs11.c |
| 2 =================================================================== |
| 3 --- nss/lib/softoken/pkcs11.c (revision 228205) |
| 4 +++ nss/lib/softoken/pkcs11.c (working copy) |
| 5 @@ -368,6 +368,9 @@ |
| 6 {CKM_SEED_MAC, {16, 16, CKF_SN_VR}, PR_TRUE}, |
| 7 {CKM_SEED_MAC_GENERAL, {16, 16, CKF_SN_VR}, PR_TRUE}, |
| 8 {CKM_SEED_CBC_PAD, {16, 16, CKF_EN_DE_WR_UN}, PR_TRUE}
, |
| 9 + /* ------------------------- ChaCha20 Operations ---------------------- */ |
| 10 + {CKM_NSS_CHACHA20_KEY_GEN, {32, 32, CKF_GENERATE}, PR_TRUE}
, |
| 11 + {CKM_NSS_CHACHA20_POLY1305,{32, 32, CKF_EN_DE}, PR_TRUE}, |
| 12 /* ------------------------- Hashing Operations ----------------------- */ |
| 13 {CKM_MD2, {0, 0, CKF_DIGEST}, PR_FALSE}, |
| 14 {CKM_MD2_HMAC, {1, 128, CKF_SN_VR}, PR_TRUE}, |
| 15 Index: nss/lib/softoken/pkcs11c.c |
| 16 =================================================================== |
| 17 --- nss/lib/softoken/pkcs11c.c (revision 228205) |
| 18 +++ nss/lib/softoken/pkcs11c.c (working copy) |
| 19 @@ -475,6 +475,97 @@ |
| 20 maxLen, input, inputLen); |
| 21 } |
| 22 |
| 23 +static SFTKChaCha20Poly1305Info * |
| 24 +sftk_ChaCha20Poly1305_CreateContext(const unsigned char *key, |
| 25 + unsigned int keyLen, |
| 26 + const CK_NSS_AEAD_PARAMS* params) |
| 27 +{ |
| 28 + SFTKChaCha20Poly1305Info *ctx; |
| 29 + |
| 30 + if (params->ulIvLen != sizeof(ctx->nonce)) { |
| 31 + PORT_SetError(SEC_ERROR_INPUT_LEN); |
| 32 + return NULL; |
| 33 + } |
| 34 + |
| 35 + ctx = PORT_New(SFTKChaCha20Poly1305Info); |
| 36 + if (ctx == NULL) { |
| 37 + return NULL; |
| 38 + } |
| 39 + |
| 40 + if (ChaCha20Poly1305_InitContext(&ctx->freeblCtx, key, keyLen, |
| 41 + params->ulTagLen) != SECSuccess) { |
| 42 + PORT_Free(ctx); |
| 43 + return NULL; |
| 44 + } |
| 45 + |
| 46 + memcpy(ctx->nonce, params->pIv, sizeof(ctx->nonce)); |
| 47 + |
| 48 + if (params->ulAADLen > sizeof(ctx->ad)) { |
| 49 + /* Need to allocate an overflow buffer for the additional data. */ |
| 50 + ctx->adOverflow = (unsigned char *)PORT_Alloc(params->ulAADLen); |
| 51 + if (!ctx->adOverflow) { |
| 52 + PORT_Free(ctx); |
| 53 + return NULL; |
| 54 + } |
| 55 + memcpy(ctx->adOverflow, params->pAAD, params->ulAADLen); |
| 56 + } else { |
| 57 + ctx->adOverflow = NULL; |
| 58 + memcpy(ctx->ad, params->pAAD, params->ulAADLen); |
| 59 + } |
| 60 + ctx->adLen = params->ulAADLen; |
| 61 + |
| 62 + return ctx; |
| 63 +} |
| 64 + |
| 65 +static void |
| 66 +sftk_ChaCha20Poly1305_DestroyContext(SFTKChaCha20Poly1305Info *ctx, |
| 67 + PRBool freeit) |
| 68 +{ |
| 69 + ChaCha20Poly1305_DestroyContext(&ctx->freeblCtx, PR_FALSE); |
| 70 + if (ctx->adOverflow != NULL) { |
| 71 + PORT_Free(ctx->adOverflow); |
| 72 + ctx->adOverflow = NULL; |
| 73 + } |
| 74 + ctx->adLen = 0; |
| 75 + if (freeit) { |
| 76 + PORT_Free(ctx); |
| 77 + } |
| 78 +} |
| 79 + |
| 80 +static SECStatus |
| 81 +sftk_ChaCha20Poly1305_Encrypt(const SFTKChaCha20Poly1305Info *ctx, |
| 82 + unsigned char *output, unsigned int *outputLen, |
| 83 + unsigned int maxOutputLen, |
| 84 + const unsigned char *input, unsigned int inputLen) |
| 85 +{ |
| 86 + const unsigned char *ad = ctx->adOverflow; |
| 87 + |
| 88 + if (ad == NULL) { |
| 89 + ad = ctx->ad; |
| 90 + } |
| 91 + |
| 92 + return ChaCha20Poly1305_Seal(&ctx->freeblCtx, output, outputLen, |
| 93 + maxOutputLen, input, inputLen, ctx->nonce, |
| 94 + sizeof(ctx->nonce), ad, ctx->adLen); |
| 95 +} |
| 96 + |
| 97 +static SECStatus |
| 98 +sftk_ChaCha20Poly1305_Decrypt(const SFTKChaCha20Poly1305Info *ctx, |
| 99 + unsigned char *output, unsigned int *outputLen, |
| 100 + unsigned int maxOutputLen, |
| 101 + const unsigned char *input, unsigned int inputLen) |
| 102 +{ |
| 103 + const unsigned char *ad = ctx->adOverflow; |
| 104 + |
| 105 + if (ad == NULL) { |
| 106 + ad = ctx->ad; |
| 107 + } |
| 108 + |
| 109 + return ChaCha20Poly1305_Open(&ctx->freeblCtx, output, outputLen, |
| 110 + maxOutputLen, input, inputLen, ctx->nonce, |
| 111 + sizeof(ctx->nonce), ad, ctx->adLen); |
| 112 +} |
| 113 + |
| 114 /** NSC_CryptInit initializes an encryption/Decryption operation. |
| 115 * |
| 116 * Always called by NSC_EncryptInit, NSC_DecryptInit, NSC_WrapKey,NSC_UnwrapKey
. |
| 117 @@ -870,6 +961,35 @@ |
| 118 context->destroy = (SFTKDestroy) AES_DestroyContext; |
| 119 break; |
| 120 |
| 121 + case CKM_NSS_CHACHA20_POLY1305: |
| 122 + if (pMechanism->ulParameterLen != sizeof(CK_NSS_AEAD_PARAMS)) { |
| 123 + crv = CKR_MECHANISM_PARAM_INVALID; |
| 124 + break; |
| 125 + } |
| 126 + context->multi = PR_FALSE; |
| 127 + if (key_type != CKK_NSS_CHACHA20) { |
| 128 + crv = CKR_KEY_TYPE_INCONSISTENT; |
| 129 + break; |
| 130 + } |
| 131 + att = sftk_FindAttribute(key,CKA_VALUE); |
| 132 + if (att == NULL) { |
| 133 + crv = CKR_KEY_HANDLE_INVALID; |
| 134 + break; |
| 135 + } |
| 136 + context->cipherInfo = sftk_ChaCha20Poly1305_CreateContext( |
| 137 + (unsigned char*) att->attrib.pValue, att->attrib.ulValueLen, |
| 138 + (CK_NSS_AEAD_PARAMS*) pMechanism->pParameter); |
| 139 + sftk_FreeAttribute(att); |
| 140 + if (context->cipherInfo == NULL) { |
| 141 + crv = sftk_MapCryptError(PORT_GetError()); |
| 142 + break; |
| 143 + } |
| 144 + context->update = (SFTKCipher) (isEncrypt ? |
| 145 + sftk_ChaCha20Poly1305_Encrypt : |
| 146 + sftk_ChaCha20Poly1305_Decrypt); |
| 147 + context->destroy = (SFTKDestroy) sftk_ChaCha20Poly1305_DestroyContext; |
| 148 + break; |
| 149 + |
| 150 case CKM_NETSCAPE_AES_KEY_WRAP_PAD: |
| 151 context->doPad = PR_TRUE; |
| 152 /* fall thru */ |
| 153 @@ -3272,6 +3392,10 @@ |
| 154 *key_type = CKK_AES; |
| 155 if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE; |
| 156 break; |
| 157 + case CKM_NSS_CHACHA20_KEY_GEN: |
| 158 + *key_type = CKK_NSS_CHACHA20; |
| 159 + if (*key_length == 0) crv = CKR_TEMPLATE_INCOMPLETE; |
| 160 + break; |
| 161 default: |
| 162 PORT_Assert(0); |
| 163 crv = CKR_MECHANISM_INVALID; |
| 164 @@ -3516,6 +3640,7 @@ |
| 165 case CKM_SEED_KEY_GEN: |
| 166 case CKM_CAMELLIA_KEY_GEN: |
| 167 case CKM_AES_KEY_GEN: |
| 168 + case CKM_NSS_CHACHA20_KEY_GEN: |
| 169 #if NSS_SOFTOKEN_DOES_RC5 |
| 170 case CKM_RC5_KEY_GEN: |
| 171 #endif |
| 172 Index: nss/lib/softoken/pkcs11i.h |
| 173 =================================================================== |
| 174 --- nss/lib/softoken/pkcs11i.h (revision 228205) |
| 175 +++ nss/lib/softoken/pkcs11i.h (working copy) |
| 176 @@ -14,6 +14,7 @@ |
| 177 #include "pkcs11t.h" |
| 178 |
| 179 #include "sftkdbt.h" |
| 180 +#include "chacha20poly1305.h" |
| 181 #include "hasht.h" |
| 182 |
| 183 /* |
| 184 @@ -104,6 +105,7 @@ |
| 185 typedef struct SFTKOAEPEncryptInfoStr SFTKOAEPEncryptInfo; |
| 186 typedef struct SFTKOAEPDecryptInfoStr SFTKOAEPDecryptInfo; |
| 187 typedef struct SFTKSSLMACInfoStr SFTKSSLMACInfo; |
| 188 +typedef struct SFTKChaCha20Poly1305InfoStr SFTKChaCha20Poly1305Info; |
| 189 typedef struct SFTKItemTemplateStr SFTKItemTemplate; |
| 190 |
| 191 /* define function pointer typdefs for pointer tables */ |
| 192 @@ -399,6 +401,16 @@ |
| 193 unsigned int keySize; |
| 194 }; |
| 195 |
| 196 +/* SFTKChaCha20Poly1305Info saves the key, tag length, nonce, and additional |
| 197 + * data for a ChaCha20+Poly1305 AEAD operation. */ |
| 198 +struct SFTKChaCha20Poly1305InfoStr { |
| 199 + ChaCha20Poly1305Context freeblCtx; |
| 200 + unsigned char nonce[8]; |
| 201 + unsigned char ad[16]; |
| 202 + unsigned char *adOverflow; |
| 203 + unsigned int adLen; |
| 204 +}; |
| 205 + |
| 206 /* |
| 207 * Template based on SECItems, suitable for passing as arrays |
| 208 */ |
| 209 Index: nss/lib/freebl/blapit.h |
| 210 =================================================================== |
| 211 --- nss/lib/freebl/blapit.h (revision 228205) |
| 212 +++ nss/lib/freebl/blapit.h (working copy) |
| 213 @@ -222,6 +222,7 @@ |
| 214 struct SHA512ContextStr ; |
| 215 struct AESKeyWrapContextStr ; |
| 216 struct SEEDContextStr ; |
| 217 +struct ChaCha20Poly1305ContextStr; |
| 218 |
| 219 typedef struct DESContextStr DESContext; |
| 220 typedef struct RC2ContextStr RC2Context; |
| 221 @@ -240,6 +241,7 @@ |
| 222 typedef struct SHA512ContextStr SHA384Context; |
| 223 typedef struct AESKeyWrapContextStr AESKeyWrapContext; |
| 224 typedef struct SEEDContextStr SEEDContext; |
| 225 +typedef struct ChaCha20Poly1305ContextStr ChaCha20Poly1305Context; |
| 226 |
| 227 /*************************************************************************** |
| 228 ** RSA Public and Private Key structures |
| 229 Index: nss/lib/freebl/blapi.h |
| 230 =================================================================== |
| 231 --- nss/lib/freebl/blapi.h (revision 228205) |
| 232 +++ nss/lib/freebl/blapi.h (working copy) |
| 233 @@ -818,7 +818,39 @@ |
| 234 unsigned int *outputLen, unsigned int maxOutputLen, |
| 235 const unsigned char *input, unsigned int inputLen); |
| 236 |
| 237 +/******************************************/ |
| 238 +/* |
| 239 +** ChaCha20+Poly1305 AEAD |
| 240 +*/ |
| 241 |
| 242 +extern SECStatus |
| 243 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx, |
| 244 + const unsigned char *key, unsigned int keyLen, |
| 245 + unsigned int tagLen); |
| 246 + |
| 247 +extern ChaCha20Poly1305Context * |
| 248 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen, |
| 249 + unsigned int tagLen); |
| 250 + |
| 251 +extern void |
| 252 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit); |
| 253 + |
| 254 +extern SECStatus |
| 255 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx, |
| 256 + unsigned char *output, unsigned int *outputLen, |
| 257 + unsigned int maxOutputLen, |
| 258 + const unsigned char *input, unsigned int inputLen, |
| 259 + const unsigned char *nonce, unsigned int nonceLen, |
| 260 + const unsigned char *ad, unsigned int adLen); |
| 261 + |
| 262 +extern SECStatus |
| 263 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx, |
| 264 + unsigned char *output, unsigned int *outputLen, |
| 265 + unsigned int maxOutputLen, |
| 266 + const unsigned char *input, unsigned int inputLen, |
| 267 + const unsigned char *nonce, unsigned int nonceLen, |
| 268 + const unsigned char *ad, unsigned int adLen); |
| 269 + |
| 270 /******************************************/ |
| 271 /* |
| 272 ** MD5 secure hash function |
| 273 Index: nss/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c |
| 274 =================================================================== |
| 275 --- nss/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c
(revision 0) |
| 276 +++ nss/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental-source.c
(revision 0) |
| 277 @@ -0,0 +1,623 @@ |
| 278 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 279 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 280 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 281 + |
| 282 +/* This implementation of poly1305 is by Andrew Moon |
| 283 + * (https://github.com/floodyberry/poly1305-donna) and released as public |
| 284 + * domain. It implements SIMD vectorization based on the algorithm described in |
| 285 + * http://cr.yp.to/papers.html#neoncrypto. Unrolled to 2 powers, i.e. 64 byte |
| 286 + * block size. */ |
| 287 + |
| 288 +#include <emmintrin.h> |
| 289 +#include <stdint.h> |
| 290 + |
| 291 +#include "poly1305.h" |
| 292 + |
| 293 +#define ALIGN(x) __attribute__((aligned(x))) |
| 294 +#define INLINE inline |
| 295 +#define U8TO64_LE(m) (*(uint64_t*)(m)) |
| 296 +#define U8TO32_LE(m) (*(uint32_t*)(m)) |
| 297 +#define U64TO8_LE(m,v) (*(uint64_t*)(m)) = v |
| 298 + |
| 299 +typedef __m128i xmmi; |
| 300 +typedef unsigned __int128 uint128_t; |
| 301 + |
| 302 +static const uint32_t ALIGN(16) poly1305_x64_sse2_message_mask[4] = {(1 << 26)
- 1, 0, (1 << 26) - 1, 0}; |
| 303 +static const uint32_t ALIGN(16) poly1305_x64_sse2_5[4] = {5, 0, 5, 0}; |
| 304 +static const uint32_t ALIGN(16) poly1305_x64_sse2_1shl128[4] = {(1 << 24), 0, (
1 << 24), 0}; |
| 305 + |
| 306 +static uint128_t INLINE |
| 307 +add128(uint128_t a, uint128_t b) { |
| 308 + return a + b; |
| 309 +} |
| 310 + |
| 311 +static uint128_t INLINE |
| 312 +add128_64(uint128_t a, uint64_t b) { |
| 313 + return a + b; |
| 314 +} |
| 315 + |
| 316 +static uint128_t INLINE |
| 317 +mul64x64_128(uint64_t a, uint64_t b) { |
| 318 + return (uint128_t)a * b; |
| 319 +} |
| 320 + |
| 321 +static uint64_t INLINE |
| 322 +lo128(uint128_t a) { |
| 323 + return (uint64_t)a; |
| 324 +} |
| 325 + |
| 326 +static uint64_t INLINE |
| 327 +shr128(uint128_t v, const int shift) { |
| 328 + return (uint64_t)(v >> shift); |
| 329 +} |
| 330 + |
| 331 +static uint64_t INLINE |
| 332 +shr128_pair(uint64_t hi, uint64_t lo, const int shift) { |
| 333 + return (uint64_t)((((uint128_t)hi << 64) | lo) >> shift); |
| 334 +} |
| 335 + |
| 336 +typedef struct poly1305_power_t { |
| 337 + union { |
| 338 + xmmi v; |
| 339 + uint64_t u[2]; |
| 340 + uint32_t d[4]; |
| 341 + } R20,R21,R22,R23,R24,S21,S22,S23,S24; |
| 342 +} poly1305_power; |
| 343 + |
| 344 +typedef struct poly1305_state_internal_t { |
| 345 + poly1305_power P[2]; /* 288 bytes, top 32 bit halves unused = 144 by
tes of free storage */ |
| 346 + union { |
| 347 + xmmi H[5]; /* 80 bytes */ |
| 348 + uint64_t HH[10]; |
| 349 + }; |
| 350 + /* uint64_t r0,r1,r2; [24 bytes] */ |
| 351 + /* uint64_t pad0,pad1; [16 bytes] */ |
| 352 + uint64_t started; /* 8 bytes */ |
| 353 + uint64_t leftover; /* 8 bytes */ |
| 354 + uint8_t buffer[64]; /* 64 bytes */ |
| 355 +} poly1305_state_internal; /* 448 bytes total + 63 bytes for alignment = 511
bytes raw */ |
| 356 + |
| 357 +static poly1305_state_internal INLINE |
| 358 +*poly1305_aligned_state(poly1305_state *state) { |
| 359 + return (poly1305_state_internal *)(((uint64_t)state + 63) & ~63); |
| 360 +} |
| 361 + |
| 362 +/* copy 0-63 bytes */ |
| 363 +static void INLINE |
| 364 +poly1305_block_copy(uint8_t *dst, const uint8_t *src, size_t bytes) { |
| 365 + size_t offset = src - dst; |
| 366 + if (bytes & 32) { |
| 367 + _mm_storeu_si128((xmmi *)(dst + 0), _mm_loadu_si128((xmmi *)(dst
+ offset + 0))); |
| 368 + _mm_storeu_si128((xmmi *)(dst + 16), _mm_loadu_si128((xmmi *)(ds
t + offset + 16))); |
| 369 + dst += 32; |
| 370 + } |
| 371 + if (bytes & 16) { _mm_storeu_si128((xmmi *)dst, _mm_loadu_si128((xmmi *)
(dst + offset))); dst += 16; } |
| 372 + if (bytes & 8) { *(uint64_t *)dst = *(uint64_t *)(dst + offset); dst +=
8; } |
| 373 + if (bytes & 4) { *(uint32_t *)dst = *(uint32_t *)(dst + offset); dst +=
4; } |
| 374 + if (bytes & 2) { *(uint16_t *)dst = *(uint16_t *)(dst + offset); dst +=
2; } |
| 375 + if (bytes & 1) { *( uint8_t *)dst = *( uint8_t *)(dst + offset);
} |
| 376 +} |
| 377 + |
| 378 +/* zero 0-15 bytes */ |
| 379 +static void INLINE |
| 380 +poly1305_block_zero(uint8_t *dst, size_t bytes) { |
| 381 + if (bytes & 8) { *(uint64_t *)dst = 0; dst += 8; } |
| 382 + if (bytes & 4) { *(uint32_t *)dst = 0; dst += 4; } |
| 383 + if (bytes & 2) { *(uint16_t *)dst = 0; dst += 2; } |
| 384 + if (bytes & 1) { *( uint8_t *)dst = 0; } |
| 385 +} |
| 386 + |
| 387 +static size_t INLINE |
| 388 +poly1305_min(size_t a, size_t b) { |
| 389 + return (a < b) ? a : b; |
| 390 +} |
| 391 + |
| 392 +void |
| 393 +Poly1305Init(poly1305_state *state, const unsigned char key[32]) { |
| 394 + poly1305_state_internal *st = poly1305_aligned_state(state); |
| 395 + poly1305_power *p; |
| 396 + uint64_t r0,r1,r2; |
| 397 + uint64_t t0,t1; |
| 398 + |
| 399 + /* clamp key */ |
| 400 + t0 = U8TO64_LE(key + 0); |
| 401 + t1 = U8TO64_LE(key + 8); |
| 402 + r0 = t0 & 0xffc0fffffff; t0 >>= 44; t0 |= t1 << 20; |
| 403 + r1 = t0 & 0xfffffc0ffff; t1 >>= 24; |
| 404 + r2 = t1 & 0x00ffffffc0f; |
| 405 + |
| 406 + /* store r in un-used space of st->P[1] */ |
| 407 + p = &st->P[1]; |
| 408 + p->R20.d[1] = (uint32_t)(r0 ); |
| 409 + p->R20.d[3] = (uint32_t)(r0 >> 32); |
| 410 + p->R21.d[1] = (uint32_t)(r1 ); |
| 411 + p->R21.d[3] = (uint32_t)(r1 >> 32); |
| 412 + p->R22.d[1] = (uint32_t)(r2 ); |
| 413 + p->R22.d[3] = (uint32_t)(r2 >> 32); |
| 414 + |
| 415 + /* store pad */ |
| 416 + p->R23.d[1] = U8TO32_LE(key + 16); |
| 417 + p->R23.d[3] = U8TO32_LE(key + 20); |
| 418 + p->R24.d[1] = U8TO32_LE(key + 24); |
| 419 + p->R24.d[3] = U8TO32_LE(key + 28); |
| 420 + |
| 421 + /* H = 0 */ |
| 422 + st->H[0] = _mm_setzero_si128(); |
| 423 + st->H[1] = _mm_setzero_si128(); |
| 424 + st->H[2] = _mm_setzero_si128(); |
| 425 + st->H[3] = _mm_setzero_si128(); |
| 426 + st->H[4] = _mm_setzero_si128(); |
| 427 + |
| 428 + st->started = 0; |
| 429 + st->leftover = 0; |
| 430 +} |
| 431 + |
| 432 +static void |
| 433 +poly1305_first_block(poly1305_state_internal *st, const uint8_t *m) { |
| 434 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask
); |
| 435 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5); |
| 436 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128); |
| 437 + xmmi T5,T6; |
| 438 + poly1305_power *p; |
| 439 + uint128_t d[3]; |
| 440 + uint64_t r0,r1,r2; |
| 441 + uint64_t r20,r21,r22,s22; |
| 442 + uint64_t pad0,pad1; |
| 443 + uint64_t c; |
| 444 + uint64_t i; |
| 445 + |
| 446 + /* pull out stored info */ |
| 447 + p = &st->P[1]; |
| 448 + |
| 449 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1]; |
| 450 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1]; |
| 451 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1]; |
| 452 + pad0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1]; |
| 453 + pad1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1]; |
| 454 + |
| 455 + /* compute powers r^2,r^4 */ |
| 456 + r20 = r0; |
| 457 + r21 = r1; |
| 458 + r22 = r2; |
| 459 + for (i = 0; i < 2; i++) { |
| 460 + s22 = r22 * (5 << 2); |
| 461 + |
| 462 + d[0] = add128(mul64x64_128(r20, r20), mul64x64_128(r21 * 2, s22)
); |
| 463 + d[1] = add128(mul64x64_128(r22, s22), mul64x64_128(r20 * 2, r21)
); |
| 464 + d[2] = add128(mul64x64_128(r21, r21), mul64x64_128(r22 * 2, r20)
); |
| 465 + |
| 466 + r20 = lo128(d[0]) & 0xfffffffffff; c
= shr128(d[0], 44); |
| 467 + d[1] = add128_64(d[1], c); r21 = lo128(d[1]) & 0xfffffffffff; c
= shr128(d[1], 44); |
| 468 + d[2] = add128_64(d[2], c); r22 = lo128(d[2]) & 0x3ffffffffff; c
= shr128(d[2], 42); |
| 469 + r20 += c * 5; c = (r20 >> 44); r20 = r20 & 0xfffffffffff; |
| 470 + r21 += c; |
| 471 + |
| 472 + p->R20.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)( r20
) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0)); |
| 473 + p->R21.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r20 >
> 26) | (r21 << 18)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0)); |
| 474 + p->R22.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 >
> 8) ) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0)); |
| 475 + p->R23.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r21 >
> 34) | (r22 << 10)) & 0x3ffffff), _MM_SHUFFLE(1,0,1,0)); |
| 476 + p->R24.v = _mm_shuffle_epi32(_mm_cvtsi32_si128((uint32_t)((r22 >
> 16) ) ), _MM_SHUFFLE(1,0,1,0)); |
| 477 + p->S21.v = _mm_mul_epu32(p->R21.v, FIVE); |
| 478 + p->S22.v = _mm_mul_epu32(p->R22.v, FIVE); |
| 479 + p->S23.v = _mm_mul_epu32(p->R23.v, FIVE); |
| 480 + p->S24.v = _mm_mul_epu32(p->R24.v, FIVE); |
| 481 + p--; |
| 482 + } |
| 483 + |
| 484 + /* put saved info back */ |
| 485 + p = &st->P[1]; |
| 486 + p->R20.d[1] = (uint32_t)(r0 ); |
| 487 + p->R20.d[3] = (uint32_t)(r0 >> 32); |
| 488 + p->R21.d[1] = (uint32_t)(r1 ); |
| 489 + p->R21.d[3] = (uint32_t)(r1 >> 32); |
| 490 + p->R22.d[1] = (uint32_t)(r2 ); |
| 491 + p->R22.d[3] = (uint32_t)(r2 >> 32); |
| 492 + p->R23.d[1] = (uint32_t)(pad0 ); |
| 493 + p->R23.d[3] = (uint32_t)(pad0 >> 32); |
| 494 + p->R24.d[1] = (uint32_t)(pad1 ); |
| 495 + p->R24.d[3] = (uint32_t)(pad1 >> 32); |
| 496 + |
| 497 + /* H = [Mx,My] */ |
| 498 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_loadl_epi6
4((xmmi *)(m + 16))); |
| 499 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_loadl_epi6
4((xmmi *)(m + 24))); |
| 500 + st->H[0] = _mm_and_si128(MMASK, T5); |
| 501 + st->H[1] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 502 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12)); |
| 503 + st->H[2] = _mm_and_si128(MMASK, T5); |
| 504 + st->H[3] = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 505 + st->H[4] = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT); |
| 506 +} |
| 507 + |
| 508 +static void |
| 509 +poly1305_blocks(poly1305_state_internal *st, const uint8_t *m, size_t bytes) { |
| 510 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask
); |
| 511 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5); |
| 512 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128); |
| 513 + |
| 514 + poly1305_power *p; |
| 515 + xmmi H0,H1,H2,H3,H4; |
| 516 + xmmi T0,T1,T2,T3,T4,T5,T6; |
| 517 + xmmi M0,M1,M2,M3,M4; |
| 518 + xmmi C1,C2; |
| 519 + |
| 520 + H0 = st->H[0]; |
| 521 + H1 = st->H[1]; |
| 522 + H2 = st->H[2]; |
| 523 + H3 = st->H[3]; |
| 524 + H4 = st->H[4]; |
| 525 + |
| 526 + while (bytes >= 64) { |
| 527 + /* H *= [r^4,r^4] */ |
| 528 + p = &st->P[0]; |
| 529 + T0 = _mm_mul_epu32(H0, p->R20.v); |
| 530 + T1 = _mm_mul_epu32(H0, p->R21.v); |
| 531 + T2 = _mm_mul_epu32(H0, p->R22.v); |
| 532 + T3 = _mm_mul_epu32(H0, p->R23.v); |
| 533 + T4 = _mm_mul_epu32(H0, p->R24.v); |
| 534 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 535 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 536 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 537 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 538 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 539 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 540 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 541 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 542 + T5 = _mm_mul_epu32(H1, p->R23.v);
T4 = _mm_add_epi64(T4, T5); |
| 543 + T5 = _mm_mul_epu32(H2, p->R22.v);
T4 = _mm_add_epi64(T4, T5); |
| 544 + T5 = _mm_mul_epu32(H3, p->R21.v);
T4 = _mm_add_epi64(T4, T5); |
| 545 + T5 = _mm_mul_epu32(H4, p->R20.v);
T4 = _mm_add_epi64(T4, T5); |
| 546 + |
| 547 + /* H += [Mx,My]*[r^2,r^2] */ |
| 548 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_lo
adl_epi64((xmmi *)(m + 16))); |
| 549 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_lo
adl_epi64((xmmi *)(m + 24))); |
| 550 + M0 = _mm_and_si128(MMASK, T5); |
| 551 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 552 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12)
); |
| 553 + M2 = _mm_and_si128(MMASK, T5); |
| 554 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 555 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT); |
| 556 + |
| 557 + p = &st->P[1]; |
| 558 + T5 = _mm_mul_epu32(M0, p->R20.v); T6 = _mm_mul_epu32(M0, p->R21.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 559 + T5 = _mm_mul_epu32(M1, p->S24.v); T6 = _mm_mul_epu32(M1, p->R20.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 560 + T5 = _mm_mul_epu32(M2, p->S23.v); T6 = _mm_mul_epu32(M2, p->S24.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 561 + T5 = _mm_mul_epu32(M3, p->S22.v); T6 = _mm_mul_epu32(M3, p->S23.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 562 + T5 = _mm_mul_epu32(M4, p->S21.v); T6 = _mm_mul_epu32(M4, p->S22.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 563 + T5 = _mm_mul_epu32(M0, p->R22.v); T6 = _mm_mul_epu32(M0, p->R23.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 564 + T5 = _mm_mul_epu32(M1, p->R21.v); T6 = _mm_mul_epu32(M1, p->R22.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 565 + T5 = _mm_mul_epu32(M2, p->R20.v); T6 = _mm_mul_epu32(M2, p->R21.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 566 + T5 = _mm_mul_epu32(M3, p->S24.v); T6 = _mm_mul_epu32(M3, p->R20.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 567 + T5 = _mm_mul_epu32(M4, p->S23.v); T6 = _mm_mul_epu32(M4, p->S24.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 568 + T5 = _mm_mul_epu32(M0, p->R24.v);
T4 = _mm_add_epi64(T4, T5); |
| 569 + T5 = _mm_mul_epu32(M1, p->R23.v);
T4 = _mm_add_epi64(T4, T5); |
| 570 + T5 = _mm_mul_epu32(M2, p->R22.v);
T4 = _mm_add_epi64(T4, T5); |
| 571 + T5 = _mm_mul_epu32(M3, p->R21.v);
T4 = _mm_add_epi64(T4, T5); |
| 572 + T5 = _mm_mul_epu32(M4, p->R20.v);
T4 = _mm_add_epi64(T4, T5); |
| 573 + |
| 574 + /* H += [Mx,My] */ |
| 575 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 32)), _mm_l
oadl_epi64((xmmi *)(m + 48))); |
| 576 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 40)), _mm_l
oadl_epi64((xmmi *)(m + 56))); |
| 577 + M0 = _mm_and_si128(MMASK, T5); |
| 578 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 579 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12)
); |
| 580 + M2 = _mm_and_si128(MMASK, T5); |
| 581 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 582 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT); |
| 583 + |
| 584 + T0 = _mm_add_epi64(T0, M0); |
| 585 + T1 = _mm_add_epi64(T1, M1); |
| 586 + T2 = _mm_add_epi64(T2, M2); |
| 587 + T3 = _mm_add_epi64(T3, M3); |
| 588 + T4 = _mm_add_epi64(T4, M4); |
| 589 + |
| 590 + /* reduce */ |
| 591 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _
mm_and_si128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C
1); T4 = _mm_add_epi64(T4, C2); |
| 592 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _
mm_and_si128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C
1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE)); |
| 593 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _
mm_and_si128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C
1); T1 = _mm_add_epi64(T1, C2); |
| 594 + C1 = _mm_srli_epi64(T3, 26); T3 = _
mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C
1); |
| 595 + |
| 596 + /* H = (H*[r^4,r^4] + [Mx,My]*[r^2,r^2] + [Mx,My]) */ |
| 597 + H0 = T0; |
| 598 + H1 = T1; |
| 599 + H2 = T2; |
| 600 + H3 = T3; |
| 601 + H4 = T4; |
| 602 + |
| 603 + m += 64; |
| 604 + bytes -= 64; |
| 605 + } |
| 606 + |
| 607 + st->H[0] = H0; |
| 608 + st->H[1] = H1; |
| 609 + st->H[2] = H2; |
| 610 + st->H[3] = H3; |
| 611 + st->H[4] = H4; |
| 612 +} |
| 613 + |
| 614 +static size_t |
| 615 +poly1305_combine(poly1305_state_internal *st, const uint8_t *m, size_t bytes) { |
| 616 + const xmmi MMASK = _mm_load_si128((xmmi *)poly1305_x64_sse2_message_mask
); |
| 617 + const xmmi HIBIT = _mm_load_si128((xmmi*)poly1305_x64_sse2_1shl128); |
| 618 + const xmmi FIVE = _mm_load_si128((xmmi*)poly1305_x64_sse2_5); |
| 619 + |
| 620 + poly1305_power *p; |
| 621 + xmmi H0,H1,H2,H3,H4; |
| 622 + xmmi M0,M1,M2,M3,M4; |
| 623 + xmmi T0,T1,T2,T3,T4,T5,T6; |
| 624 + xmmi C1,C2; |
| 625 + |
| 626 + uint64_t r0,r1,r2; |
| 627 + uint64_t t0,t1,t2,t3,t4; |
| 628 + uint64_t c; |
| 629 + size_t consumed = 0; |
| 630 + |
| 631 + H0 = st->H[0]; |
| 632 + H1 = st->H[1]; |
| 633 + H2 = st->H[2]; |
| 634 + H3 = st->H[3]; |
| 635 + H4 = st->H[4]; |
| 636 + |
| 637 + /* p = [r^2,r^2] */ |
| 638 + p = &st->P[1]; |
| 639 + |
| 640 + if (bytes >= 32) { |
| 641 + /* H *= [r^2,r^2] */ |
| 642 + T0 = _mm_mul_epu32(H0, p->R20.v); |
| 643 + T1 = _mm_mul_epu32(H0, p->R21.v); |
| 644 + T2 = _mm_mul_epu32(H0, p->R22.v); |
| 645 + T3 = _mm_mul_epu32(H0, p->R23.v); |
| 646 + T4 = _mm_mul_epu32(H0, p->R24.v); |
| 647 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 648 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 649 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 650 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.
v); T0 = _mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 651 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 652 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 653 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 654 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.
v); T2 = _mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 655 + T5 = _mm_mul_epu32(H1, p->R23.v);
T4 = _mm_add_epi64(T4, T5); |
| 656 + T5 = _mm_mul_epu32(H2, p->R22.v);
T4 = _mm_add_epi64(T4, T5); |
| 657 + T5 = _mm_mul_epu32(H3, p->R21.v);
T4 = _mm_add_epi64(T4, T5); |
| 658 + T5 = _mm_mul_epu32(H4, p->R20.v);
T4 = _mm_add_epi64(T4, T5); |
| 659 + |
| 660 + /* H += [Mx,My] */ |
| 661 + T5 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 0)), _mm_lo
adl_epi64((xmmi *)(m + 16))); |
| 662 + T6 = _mm_unpacklo_epi64(_mm_loadl_epi64((xmmi *)(m + 8)), _mm_lo
adl_epi64((xmmi *)(m + 24))); |
| 663 + M0 = _mm_and_si128(MMASK, T5); |
| 664 + M1 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 665 + T5 = _mm_or_si128(_mm_srli_epi64(T5, 52), _mm_slli_epi64(T6, 12)
); |
| 666 + M2 = _mm_and_si128(MMASK, T5); |
| 667 + M3 = _mm_and_si128(MMASK, _mm_srli_epi64(T5, 26)); |
| 668 + M4 = _mm_or_si128(_mm_srli_epi64(T6, 40), HIBIT); |
| 669 + |
| 670 + T0 = _mm_add_epi64(T0, M0); |
| 671 + T1 = _mm_add_epi64(T1, M1); |
| 672 + T2 = _mm_add_epi64(T2, M2); |
| 673 + T3 = _mm_add_epi64(T3, M3); |
| 674 + T4 = _mm_add_epi64(T4, M4); |
| 675 + |
| 676 + /* reduce */ |
| 677 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _
mm_and_si128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C
1); T4 = _mm_add_epi64(T4, C2); |
| 678 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _
mm_and_si128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C
1); T0 = _mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE)); |
| 679 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _
mm_and_si128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C
1); T1 = _mm_add_epi64(T1, C2); |
| 680 + C1 = _mm_srli_epi64(T3, 26); T3 = _
mm_and_si128(T3, MMASK); T4 = _mm_add_epi64(T4, C
1); |
| 681 + |
| 682 + /* H = (H*[r^2,r^2] + [Mx,My]) */ |
| 683 + H0 = T0; |
| 684 + H1 = T1; |
| 685 + H2 = T2; |
| 686 + H3 = T3; |
| 687 + H4 = T4; |
| 688 + |
| 689 + consumed = 32; |
| 690 + } |
| 691 + |
| 692 + /* finalize, H *= [r^2,r] */ |
| 693 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1]; |
| 694 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1]; |
| 695 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1]; |
| 696 + |
| 697 + p->R20.d[2] = (uint32_t)( r0 ) & 0x3ffffff; |
| 698 + p->R21.d[2] = (uint32_t)((r0 >> 26) | (r1 << 18)) & 0x3ffffff; |
| 699 + p->R22.d[2] = (uint32_t)((r1 >> 8) ) & 0x3ffffff; |
| 700 + p->R23.d[2] = (uint32_t)((r1 >> 34) | (r2 << 10)) & 0x3ffffff; |
| 701 + p->R24.d[2] = (uint32_t)((r2 >> 16) ) ; |
| 702 + p->S21.d[2] = p->R21.d[2] * 5; |
| 703 + p->S22.d[2] = p->R22.d[2] * 5; |
| 704 + p->S23.d[2] = p->R23.d[2] * 5; |
| 705 + p->S24.d[2] = p->R24.d[2] * 5; |
| 706 + |
| 707 + /* H *= [r^2,r] */ |
| 708 + T0 = _mm_mul_epu32(H0, p->R20.v); |
| 709 + T1 = _mm_mul_epu32(H0, p->R21.v); |
| 710 + T2 = _mm_mul_epu32(H0, p->R22.v); |
| 711 + T3 = _mm_mul_epu32(H0, p->R23.v); |
| 712 + T4 = _mm_mul_epu32(H0, p->R24.v); |
| 713 + T5 = _mm_mul_epu32(H1, p->S24.v); T6 = _mm_mul_epu32(H1, p->R20.v); T0 =
_mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 714 + T5 = _mm_mul_epu32(H2, p->S23.v); T6 = _mm_mul_epu32(H2, p->S24.v); T0 =
_mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 715 + T5 = _mm_mul_epu32(H3, p->S22.v); T6 = _mm_mul_epu32(H3, p->S23.v); T0 =
_mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 716 + T5 = _mm_mul_epu32(H4, p->S21.v); T6 = _mm_mul_epu32(H4, p->S22.v); T0 =
_mm_add_epi64(T0, T5); T1 = _mm_add_epi64(T1, T6); |
| 717 + T5 = _mm_mul_epu32(H1, p->R21.v); T6 = _mm_mul_epu32(H1, p->R22.v); T2 =
_mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 718 + T5 = _mm_mul_epu32(H2, p->R20.v); T6 = _mm_mul_epu32(H2, p->R21.v); T2 =
_mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 719 + T5 = _mm_mul_epu32(H3, p->S24.v); T6 = _mm_mul_epu32(H3, p->R20.v); T2 =
_mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 720 + T5 = _mm_mul_epu32(H4, p->S23.v); T6 = _mm_mul_epu32(H4, p->S24.v); T2 =
_mm_add_epi64(T2, T5); T3 = _mm_add_epi64(T3, T6); |
| 721 + T5 = _mm_mul_epu32(H1, p->R23.v); T4 =
_mm_add_epi64(T4, T5); |
| 722 + T5 = _mm_mul_epu32(H2, p->R22.v); T4 =
_mm_add_epi64(T4, T5); |
| 723 + T5 = _mm_mul_epu32(H3, p->R21.v); T4 =
_mm_add_epi64(T4, T5); |
| 724 + T5 = _mm_mul_epu32(H4, p->R20.v); T4 =
_mm_add_epi64(T4, T5); |
| 725 + |
| 726 + C1 = _mm_srli_epi64(T0, 26); C2 = _mm_srli_epi64(T3, 26); T0 = _mm_and_s
i128(T0, MMASK); T3 = _mm_and_si128(T3, MMASK); T1 = _mm_add_epi64(T1, C1); T4 =
_mm_add_epi64(T4, C2); |
| 727 + C1 = _mm_srli_epi64(T1, 26); C2 = _mm_srli_epi64(T4, 26); T1 = _mm_and_s
i128(T1, MMASK); T4 = _mm_and_si128(T4, MMASK); T2 = _mm_add_epi64(T2, C1); T0 =
_mm_add_epi64(T0, _mm_mul_epu32(C2, FIVE)); |
| 728 + C1 = _mm_srli_epi64(T2, 26); C2 = _mm_srli_epi64(T0, 26); T2 = _mm_and_s
i128(T2, MMASK); T0 = _mm_and_si128(T0, MMASK); T3 = _mm_add_epi64(T3, C1); T1 =
_mm_add_epi64(T1, C2); |
| 729 + C1 = _mm_srli_epi64(T3, 26); T3 = _mm_and_s
i128(T3, MMASK); T4 = _mm_add_epi64(T4, C1); |
| 730 + |
| 731 + /* H = H[0]+H[1] */ |
| 732 + H0 = _mm_add_epi64(T0, _mm_srli_si128(T0, 8)); |
| 733 + H1 = _mm_add_epi64(T1, _mm_srli_si128(T1, 8)); |
| 734 + H2 = _mm_add_epi64(T2, _mm_srli_si128(T2, 8)); |
| 735 + H3 = _mm_add_epi64(T3, _mm_srli_si128(T3, 8)); |
| 736 + H4 = _mm_add_epi64(T4, _mm_srli_si128(T4, 8)); |
| 737 + |
| 738 + t0 = _mm_cvtsi128_si32(H0) ; c = (t0 >> 26); t0 &= 0x3ffffff; |
| 739 + t1 = _mm_cvtsi128_si32(H1) + c; c = (t1 >> 26); t1 &= 0x3ffffff; |
| 740 + t2 = _mm_cvtsi128_si32(H2) + c; c = (t2 >> 26); t2 &= 0x3ffffff; |
| 741 + t3 = _mm_cvtsi128_si32(H3) + c; c = (t3 >> 26); t3 &= 0x3ffffff; |
| 742 + t4 = _mm_cvtsi128_si32(H4) + c; c = (t4 >> 26); t4 &= 0x3ffffff; |
| 743 + t0 = t0 + (c * 5); c = (t0 >> 26); t0 &= 0x3ffffff; |
| 744 + t1 = t1 + c; |
| 745 + |
| 746 + st->HH[0] = ((t0 ) | (t1 << 26) ) & 0xfffffffffffull; |
| 747 + st->HH[1] = ((t1 >> 18) | (t2 << 8) | (t3 << 34)) & 0xfffffffffffull; |
| 748 + st->HH[2] = ((t3 >> 10) | (t4 << 16) ) & 0x3ffffffffffull; |
| 749 + |
| 750 + return consumed; |
| 751 +} |
| 752 + |
| 753 +void |
| 754 +Poly1305Update(poly1305_state *state, const unsigned char *m, size_t bytes) { |
| 755 + poly1305_state_internal *st = poly1305_aligned_state(state); |
| 756 + size_t want; |
| 757 + |
| 758 + /* need at least 32 initial bytes to start the accelerated branch */ |
| 759 + if (!st->started) { |
| 760 + if ((st->leftover == 0) && (bytes > 32)) { |
| 761 + poly1305_first_block(st, m); |
| 762 + m += 32; |
| 763 + bytes -= 32; |
| 764 + } else { |
| 765 + want = poly1305_min(32 - st->leftover, bytes); |
| 766 + poly1305_block_copy(st->buffer + st->leftover, m, want); |
| 767 + bytes -= want; |
| 768 + m += want; |
| 769 + st->leftover += want; |
| 770 + if ((st->leftover < 32) || (bytes == 0)) |
| 771 + return; |
| 772 + poly1305_first_block(st, st->buffer); |
| 773 + st->leftover = 0; |
| 774 + } |
| 775 + st->started = 1; |
| 776 + } |
| 777 + |
| 778 + /* handle leftover */ |
| 779 + if (st->leftover) { |
| 780 + want = poly1305_min(64 - st->leftover, bytes); |
| 781 + poly1305_block_copy(st->buffer + st->leftover, m, want); |
| 782 + bytes -= want; |
| 783 + m += want; |
| 784 + st->leftover += want; |
| 785 + if (st->leftover < 64) |
| 786 + return; |
| 787 + poly1305_blocks(st, st->buffer, 64); |
| 788 + st->leftover = 0; |
| 789 + } |
| 790 + |
| 791 + /* process 64 byte blocks */ |
| 792 + if (bytes >= 64) { |
| 793 + want = (bytes & ~63); |
| 794 + poly1305_blocks(st, m, want); |
| 795 + m += want; |
| 796 + bytes -= want; |
| 797 + } |
| 798 + |
| 799 + if (bytes) { |
| 800 + poly1305_block_copy(st->buffer + st->leftover, m, bytes); |
| 801 + st->leftover += bytes; |
| 802 + } |
| 803 +} |
| 804 + |
| 805 +void |
| 806 +Poly1305Finish(poly1305_state *state, unsigned char mac[16]) { |
| 807 + poly1305_state_internal *st = poly1305_aligned_state(state); |
| 808 + size_t leftover = st->leftover; |
| 809 + uint8_t *m = st->buffer; |
| 810 + uint128_t d[3]; |
| 811 + uint64_t h0,h1,h2; |
| 812 + uint64_t t0,t1; |
| 813 + uint64_t g0,g1,g2,c,nc; |
| 814 + uint64_t r0,r1,r2,s1,s2; |
| 815 + poly1305_power *p; |
| 816 + |
| 817 + if (st->started) { |
| 818 + size_t consumed = poly1305_combine(st, m, leftover); |
| 819 + leftover -= consumed; |
| 820 + m += consumed; |
| 821 + } |
| 822 + |
| 823 + /* st->HH will either be 0 or have the combined result */ |
| 824 + h0 = st->HH[0]; |
| 825 + h1 = st->HH[1]; |
| 826 + h2 = st->HH[2]; |
| 827 + |
| 828 + p = &st->P[1]; |
| 829 + r0 = ((uint64_t)p->R20.d[3] << 32) | (uint64_t)p->R20.d[1]; |
| 830 + r1 = ((uint64_t)p->R21.d[3] << 32) | (uint64_t)p->R21.d[1]; |
| 831 + r2 = ((uint64_t)p->R22.d[3] << 32) | (uint64_t)p->R22.d[1]; |
| 832 + s1 = r1 * (5 << 2); |
| 833 + s2 = r2 * (5 << 2); |
| 834 + |
| 835 + if (leftover < 16) |
| 836 + goto poly1305_donna_atmost15bytes; |
| 837 + |
| 838 +poly1305_donna_atleast16bytes: |
| 839 + t0 = U8TO64_LE(m + 0); |
| 840 + t1 = U8TO64_LE(m + 8); |
| 841 + h0 += t0 & 0xfffffffffff; |
| 842 + t0 = shr128_pair(t1, t0, 44); |
| 843 + h1 += t0 & 0xfffffffffff; |
| 844 + h2 += (t1 >> 24) | ((uint64_t)1 << 40); |
| 845 + |
| 846 +poly1305_donna_mul: |
| 847 + d[0] = add128(add128(mul64x64_128(h0, r0), mul64x64_128(h1, s2)), mul64x
64_128(h2, s1)); |
| 848 + d[1] = add128(add128(mul64x64_128(h0, r1), mul64x64_128(h1, r0)), mul64x
64_128(h2, s2)); |
| 849 + d[2] = add128(add128(mul64x64_128(h0, r2), mul64x64_128(h1, r1)), mul64x
64_128(h2, r0)); |
| 850 + h0 = lo128(d[0]) & 0xfffffffffff; c = shr128(
d[0], 44); |
| 851 + d[1] = add128_64(d[1], c); h1 = lo128(d[1]) & 0xfffffffffff; c = shr128(
d[1], 44); |
| 852 + d[2] = add128_64(d[2], c); h2 = lo128(d[2]) & 0x3ffffffffff; c = shr128(
d[2], 42); |
| 853 + h0 += c * 5; |
| 854 + |
| 855 + m += 16; |
| 856 + leftover -= 16; |
| 857 + if (leftover >= 16) goto poly1305_donna_atleast16bytes; |
| 858 + |
| 859 + /* final bytes */ |
| 860 +poly1305_donna_atmost15bytes: |
| 861 + if (!leftover) goto poly1305_donna_finish; |
| 862 + |
| 863 + m[leftover++] = 1; |
| 864 + poly1305_block_zero(m + leftover, 16 - leftover); |
| 865 + leftover = 16; |
| 866 + |
| 867 + t0 = U8TO64_LE(m+0); |
| 868 + t1 = U8TO64_LE(m+8); |
| 869 + h0 += t0 & 0xfffffffffff; t0 = shr128_pair(t1, t0, 44); |
| 870 + h1 += t0 & 0xfffffffffff; |
| 871 + h2 += (t1 >> 24); |
| 872 + |
| 873 + goto poly1305_donna_mul; |
| 874 + |
| 875 +poly1305_donna_finish: |
| 876 + c = (h0 >> 44); h0 &= 0xfffffffffff; |
| 877 + h1 += c; c = (h1 >> 44); h1 &= 0xfffffffffff; |
| 878 + h2 += c; c = (h2 >> 42); h2 &= 0x3ffffffffff; |
| 879 + h0 += c * 5; |
| 880 + |
| 881 + g0 = h0 + 5; c = (g0 >> 44); g0 &= 0xfffffffffff; |
| 882 + g1 = h1 + c; c = (g1 >> 44); g1 &= 0xfffffffffff; |
| 883 + g2 = h2 + c - ((uint64_t)1 << 42); |
| 884 + |
| 885 + c = (g2 >> 63) - 1; |
| 886 + nc = ~c; |
| 887 + h0 = (h0 & nc) | (g0 & c); |
| 888 + h1 = (h1 & nc) | (g1 & c); |
| 889 + h2 = (h2 & nc) | (g2 & c); |
| 890 + |
| 891 + /* pad */ |
| 892 + t0 = ((uint64_t)p->R23.d[3] << 32) | (uint64_t)p->R23.d[1]; |
| 893 + t1 = ((uint64_t)p->R24.d[3] << 32) | (uint64_t)p->R24.d[1]; |
| 894 + h0 += (t0 & 0xfffffffffff) ; c = (h0 >> 44); h0 &= 0xfffffffffff; t0
= shr128_pair(t1, t0, 44); |
| 895 + h1 += (t0 & 0xfffffffffff) + c; c = (h1 >> 44); h1 &= 0xfffffffffff; t1
= (t1 >> 24); |
| 896 + h2 += (t1 ) + c; |
| 897 + |
| 898 + U64TO8_LE(mac + 0, ((h0 ) | (h1 << 44))); |
| 899 + U64TO8_LE(mac + 8, ((h1 >> 20) | (h2 << 24))); |
| 900 +} |
| 901 |
| 902 Property changes on: nss/lib/freebl/poly1305/poly1305-donna-x64-sse2-incremental
-source.c |
| 903 ___________________________________________________________________ |
| 904 Added: svn:eol-style |
| 905 + LF |
| 906 |
| 907 Index: nss/lib/freebl/poly1305/poly1305.h |
| 908 =================================================================== |
| 909 --- nss/lib/freebl/poly1305/poly1305.h (revision 0) |
| 910 +++ nss/lib/freebl/poly1305/poly1305.h (revision 0) |
| 911 @@ -0,0 +1,31 @@ |
| 912 +/* |
| 913 + * poly1305.h - header file for Poly1305 implementation. |
| 914 + * |
| 915 + * This Source Code Form is subject to the terms of the Mozilla Public |
| 916 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 917 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 918 + |
| 919 +#ifndef FREEBL_POLY1305_H_ |
| 920 +#define FREEBL_POLY1305_H_ |
| 921 + |
| 922 +typedef unsigned char poly1305_state[512]; |
| 923 + |
| 924 +/* Poly1305Init sets up |state| so that it can be used to calculate an |
| 925 + * authentication tag with the one-time key |key|. Note that |key| is a |
| 926 + * one-time key and therefore there is no `reset' method because that would |
| 927 + * enable several messages to be authenticated with the same key. */ |
| 928 +extern void Poly1305Init(poly1305_state* state, |
| 929 + const unsigned char key[32]); |
| 930 + |
| 931 +/* Poly1305Update processes |in_len| bytes from |in|. It can be called zero or |
| 932 + * more times after poly1305_init. */ |
| 933 +extern void Poly1305Update(poly1305_state* state, |
| 934 + const unsigned char *in, |
| 935 + size_t inLen); |
| 936 + |
| 937 +/* Poly1305Finish completes the poly1305 calculation and writes a 16 byte |
| 938 + * authentication tag to |mac|. */ |
| 939 +extern void Poly1305Finish(poly1305_state* state, |
| 940 + unsigned char mac[16]); |
| 941 + |
| 942 +#endif /* FREEBL_POLY1305_H_ */ |
| 943 |
| 944 Property changes on: nss/lib/freebl/poly1305/poly1305.h |
| 945 ___________________________________________________________________ |
| 946 Added: svn:eol-style |
| 947 + LF |
| 948 |
| 949 Index: nss/lib/freebl/poly1305/poly1305.c |
| 950 =================================================================== |
| 951 --- nss/lib/freebl/poly1305/poly1305.c (revision 0) |
| 952 +++ nss/lib/freebl/poly1305/poly1305.c (revision 0) |
| 953 @@ -0,0 +1,254 @@ |
| 954 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 955 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 956 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 957 + |
| 958 +/* This implementation of poly1305 is by Andrew Moon |
| 959 + * (https://github.com/floodyberry/poly1305-donna) and released as public |
| 960 + * domain. */ |
| 961 + |
| 962 +#include <string.h> |
| 963 +#include <stdint.h> |
| 964 + |
| 965 +#include "poly1305.h" |
| 966 + |
| 967 +#if defined(NSS_X86) || defined(NSS_X64) |
| 968 +/* We can assume little-endian. */ |
| 969 +static uint32_t U8TO32_LE(const unsigned char *m) { |
| 970 + uint32_t r; |
| 971 + memcpy(&r, m, sizeof(r)); |
| 972 + return r; |
| 973 +} |
| 974 + |
| 975 +static void U32TO8_LE(unsigned char *m, uint32_t v) { |
| 976 + memcpy(m, &v, sizeof(v)); |
| 977 +} |
| 978 +#else |
| 979 +static void U8TO32_LE(const unsigned char *m) { |
| 980 + return (uint32_t)m[0] | |
| 981 + (uint32_t)m[1] << 8 | |
| 982 + (uint32_t)m[2] << 16 | |
| 983 + (uint32_t)m[3] << 24; |
| 984 +} |
| 985 + |
| 986 +static void U32TO8_LE(unsigned char *m, uint32_t v) { |
| 987 + m[0] = v; |
| 988 + m[1] = v >> 8; |
| 989 + m[2] = v >> 16; |
| 990 + m[3] = v >> 24; |
| 991 +} |
| 992 +#endif |
| 993 + |
| 994 +static uint64_t |
| 995 +mul32x32_64(uint32_t a, uint32_t b) { |
| 996 + return (uint64_t)a * b; |
| 997 +} |
| 998 + |
| 999 +struct poly1305_state_st { |
| 1000 + uint32_t r0,r1,r2,r3,r4; |
| 1001 + uint32_t s1,s2,s3,s4; |
| 1002 + uint32_t h0,h1,h2,h3,h4; |
| 1003 + unsigned char buf[16]; |
| 1004 + unsigned int buf_used; |
| 1005 + unsigned char key[16]; |
| 1006 +}; |
| 1007 + |
| 1008 +/* update updates |state| given some amount of input data. This function may |
| 1009 + * only be called with a |len| that is not a multiple of 16 at the end of the |
| 1010 + * data. Otherwise the input must be buffered into 16 byte blocks. */ |
| 1011 +static void update(struct poly1305_state_st *state, const unsigned char *in, |
| 1012 + size_t len) { |
| 1013 + uint32_t t0,t1,t2,t3; |
| 1014 + uint64_t t[5]; |
| 1015 + uint32_t b; |
| 1016 + uint64_t c; |
| 1017 + size_t j; |
| 1018 + unsigned char mp[16]; |
| 1019 + |
| 1020 + if (len < 16) |
| 1021 + goto poly1305_donna_atmost15bytes; |
| 1022 + |
| 1023 +poly1305_donna_16bytes: |
| 1024 + t0 = U8TO32_LE(in); |
| 1025 + t1 = U8TO32_LE(in+4); |
| 1026 + t2 = U8TO32_LE(in+8); |
| 1027 + t3 = U8TO32_LE(in+12); |
| 1028 + |
| 1029 + in += 16; |
| 1030 + len -= 16; |
| 1031 + |
| 1032 + state->h0 += t0 & 0x3ffffff; |
| 1033 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff; |
| 1034 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff; |
| 1035 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff; |
| 1036 + state->h4 += (t3 >> 8) | (1 << 24); |
| 1037 + |
| 1038 +poly1305_donna_mul: |
| 1039 + t[0] = mul32x32_64(state->h0,state->r0) + |
| 1040 + mul32x32_64(state->h1,state->s4) + |
| 1041 + mul32x32_64(state->h2,state->s3) + |
| 1042 + mul32x32_64(state->h3,state->s2) + |
| 1043 + mul32x32_64(state->h4,state->s1); |
| 1044 + t[1] = mul32x32_64(state->h0,state->r1) + |
| 1045 + mul32x32_64(state->h1,state->r0) + |
| 1046 + mul32x32_64(state->h2,state->s4) + |
| 1047 + mul32x32_64(state->h3,state->s3) + |
| 1048 + mul32x32_64(state->h4,state->s2); |
| 1049 + t[2] = mul32x32_64(state->h0,state->r2) + |
| 1050 + mul32x32_64(state->h1,state->r1) + |
| 1051 + mul32x32_64(state->h2,state->r0) + |
| 1052 + mul32x32_64(state->h3,state->s4) + |
| 1053 + mul32x32_64(state->h4,state->s3); |
| 1054 + t[3] = mul32x32_64(state->h0,state->r3) + |
| 1055 + mul32x32_64(state->h1,state->r2) + |
| 1056 + mul32x32_64(state->h2,state->r1) + |
| 1057 + mul32x32_64(state->h3,state->r0) + |
| 1058 + mul32x32_64(state->h4,state->s4); |
| 1059 + t[4] = mul32x32_64(state->h0,state->r4) + |
| 1060 + mul32x32_64(state->h1,state->r3) + |
| 1061 + mul32x32_64(state->h2,state->r2) + |
| 1062 + mul32x32_64(state->h3,state->r1) + |
| 1063 + mul32x32_64(state->h4,state->r0); |
| 1064 + |
| 1065 + state->h0 = (uint32_t)t[0] & 0x3ffffff; c = (t[0] >
> 26); |
| 1066 + t[1] += c; state->h1 = (uint32_t)t[1] & 0x3ffffff; b = (uint32_t)(t[1] >
> 26); |
| 1067 + t[2] += b; state->h2 = (uint32_t)t[2] & 0x3ffffff; b = (uint32_t)(t[2] >
> 26); |
| 1068 + t[3] += b; state->h3 = (uint32_t)t[3] & 0x3ffffff; b = (uint32_t)(t[3] >
> 26); |
| 1069 + t[4] += b; state->h4 = (uint32_t)t[4] & 0x3ffffff; b = (uint32_t)(t[4] >
> 26); |
| 1070 + state->h0 += b * 5; |
| 1071 + |
| 1072 + if (len >= 16) |
| 1073 + goto poly1305_donna_16bytes; |
| 1074 + |
| 1075 + /* final bytes */ |
| 1076 +poly1305_donna_atmost15bytes: |
| 1077 + if (!len) |
| 1078 + return; |
| 1079 + |
| 1080 + for (j = 0; j < len; j++) |
| 1081 + mp[j] = in[j]; |
| 1082 + mp[j++] = 1; |
| 1083 + for (; j < 16; j++) |
| 1084 + mp[j] = 0; |
| 1085 + len = 0; |
| 1086 + |
| 1087 + t0 = U8TO32_LE(mp+0); |
| 1088 + t1 = U8TO32_LE(mp+4); |
| 1089 + t2 = U8TO32_LE(mp+8); |
| 1090 + t3 = U8TO32_LE(mp+12); |
| 1091 + |
| 1092 + state->h0 += t0 & 0x3ffffff; |
| 1093 + state->h1 += ((((uint64_t)t1 << 32) | t0) >> 26) & 0x3ffffff; |
| 1094 + state->h2 += ((((uint64_t)t2 << 32) | t1) >> 20) & 0x3ffffff; |
| 1095 + state->h3 += ((((uint64_t)t3 << 32) | t2) >> 14) & 0x3ffffff; |
| 1096 + state->h4 += (t3 >> 8); |
| 1097 + |
| 1098 + goto poly1305_donna_mul; |
| 1099 +} |
| 1100 + |
| 1101 +void Poly1305Init(poly1305_state *statep, const unsigned char key[32]) { |
| 1102 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep; |
| 1103 + uint32_t t0,t1,t2,t3; |
| 1104 + |
| 1105 + t0 = U8TO32_LE(key+0); |
| 1106 + t1 = U8TO32_LE(key+4); |
| 1107 + t2 = U8TO32_LE(key+8); |
| 1108 + t3 = U8TO32_LE(key+12); |
| 1109 + |
| 1110 + /* precompute multipliers */ |
| 1111 + state->r0 = t0 & 0x3ffffff; t0 >>= 26; t0 |= t1 << 6; |
| 1112 + state->r1 = t0 & 0x3ffff03; t1 >>= 20; t1 |= t2 << 12; |
| 1113 + state->r2 = t1 & 0x3ffc0ff; t2 >>= 14; t2 |= t3 << 18; |
| 1114 + state->r3 = t2 & 0x3f03fff; t3 >>= 8; |
| 1115 + state->r4 = t3 & 0x00fffff; |
| 1116 + |
| 1117 + state->s1 = state->r1 * 5; |
| 1118 + state->s2 = state->r2 * 5; |
| 1119 + state->s3 = state->r3 * 5; |
| 1120 + state->s4 = state->r4 * 5; |
| 1121 + |
| 1122 + /* init state */ |
| 1123 + state->h0 = 0; |
| 1124 + state->h1 = 0; |
| 1125 + state->h2 = 0; |
| 1126 + state->h3 = 0; |
| 1127 + state->h4 = 0; |
| 1128 + |
| 1129 + state->buf_used = 0; |
| 1130 + memcpy(state->key, key + 16, sizeof(state->key)); |
| 1131 +} |
| 1132 + |
| 1133 +void Poly1305Update(poly1305_state *statep, const unsigned char *in, |
| 1134 + size_t in_len) { |
| 1135 + unsigned int i; |
| 1136 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep; |
| 1137 + |
| 1138 + if (state->buf_used) { |
| 1139 + unsigned int todo = 16 - state->buf_used; |
| 1140 + if (todo > in_len) |
| 1141 + todo = in_len; |
| 1142 + for (i = 0; i < todo; i++) |
| 1143 + state->buf[state->buf_used + i] = in[i]; |
| 1144 + state->buf_used += todo; |
| 1145 + in_len -= todo; |
| 1146 + in += todo; |
| 1147 + |
| 1148 + if (state->buf_used == 16) { |
| 1149 + update(state, state->buf, 16); |
| 1150 + state->buf_used = 0; |
| 1151 + } |
| 1152 + } |
| 1153 + |
| 1154 + if (in_len >= 16) { |
| 1155 + size_t todo = in_len & ~0xf; |
| 1156 + update(state, in, todo); |
| 1157 + in += todo; |
| 1158 + in_len &= 0xf; |
| 1159 + } |
| 1160 + |
| 1161 + if (in_len) { |
| 1162 + for (i = 0; i < in_len; i++) |
| 1163 + state->buf[i] = in[i]; |
| 1164 + state->buf_used = in_len; |
| 1165 + } |
| 1166 +} |
| 1167 + |
| 1168 +void Poly1305Finish(poly1305_state *statep, unsigned char mac[16]) { |
| 1169 + struct poly1305_state_st *state = (struct poly1305_state_st*) statep; |
| 1170 + uint64_t f0,f1,f2,f3; |
| 1171 + uint32_t g0,g1,g2,g3,g4; |
| 1172 + uint32_t b, nb; |
| 1173 + |
| 1174 + if (state->buf_used) |
| 1175 + update(state, state->buf, state->buf_used); |
| 1176 + |
| 1177 + b = state->h0 >> 26; state->h0 = state->h0 & 0x3ffff
ff; |
| 1178 + state->h1 += b; b = state->h1 >> 26; state->h1 = state->h1 & 0x3ffff
ff; |
| 1179 + state->h2 += b; b = state->h2 >> 26; state->h2 = state->h2 & 0x3ffff
ff; |
| 1180 + state->h3 += b; b = state->h3 >> 26; state->h3 = state->h3 & 0x3ffff
ff; |
| 1181 + state->h4 += b; b = state->h4 >> 26; state->h4 = state->h4 & 0x3ffff
ff; |
| 1182 + state->h0 += b * 5; |
| 1183 + |
| 1184 + g0 = state->h0 + 5; b = g0 >> 26; g0 &= 0x3ffffff; |
| 1185 + g1 = state->h1 + b; b = g1 >> 26; g1 &= 0x3ffffff; |
| 1186 + g2 = state->h2 + b; b = g2 >> 26; g2 &= 0x3ffffff; |
| 1187 + g3 = state->h3 + b; b = g3 >> 26; g3 &= 0x3ffffff; |
| 1188 + g4 = state->h4 + b - (1 << 26); |
| 1189 + |
| 1190 + b = (g4 >> 31) - 1; |
| 1191 + nb = ~b; |
| 1192 + state->h0 = (state->h0 & nb) | (g0 & b); |
| 1193 + state->h1 = (state->h1 & nb) | (g1 & b); |
| 1194 + state->h2 = (state->h2 & nb) | (g2 & b); |
| 1195 + state->h3 = (state->h3 & nb) | (g3 & b); |
| 1196 + state->h4 = (state->h4 & nb) | (g4 & b); |
| 1197 + |
| 1198 + f0 = ((state->h0 ) | (state->h1 << 26)) + (uint64_t)U8TO32_LE(&stat
e->key[0]); |
| 1199 + f1 = ((state->h1 >> 6) | (state->h2 << 20)) + (uint64_t)U8TO32_LE(&stat
e->key[4]); |
| 1200 + f2 = ((state->h2 >> 12) | (state->h3 << 14)) + (uint64_t)U8TO32_LE(&stat
e->key[8]); |
| 1201 + f3 = ((state->h3 >> 18) | (state->h4 << 8)) + (uint64_t)U8TO32_LE(&stat
e->key[12]); |
| 1202 + |
| 1203 + U32TO8_LE(&mac[ 0], f0); f1 += (f0 >> 32); |
| 1204 + U32TO8_LE(&mac[ 4], f1); f2 += (f1 >> 32); |
| 1205 + U32TO8_LE(&mac[ 8], f2); f3 += (f2 >> 32); |
| 1206 + U32TO8_LE(&mac[12], f3); |
| 1207 +} |
| 1208 |
| 1209 Property changes on: nss/lib/freebl/poly1305/poly1305.c |
| 1210 ___________________________________________________________________ |
| 1211 Added: svn:eol-style |
| 1212 + LF |
| 1213 |
| 1214 Index: nss/lib/freebl/chacha20poly1305.c |
| 1215 =================================================================== |
| 1216 --- nss/lib/freebl/chacha20poly1305.c (revision 0) |
| 1217 +++ nss/lib/freebl/chacha20poly1305.c (revision 0) |
| 1218 @@ -0,0 +1,169 @@ |
| 1219 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 1220 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 1221 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 1222 + |
| 1223 +#ifdef FREEBL_NO_DEPEND |
| 1224 +#include "stubs.h" |
| 1225 +#endif |
| 1226 + |
| 1227 +#include <string.h> |
| 1228 +#include <stdio.h> |
| 1229 + |
| 1230 +#include "seccomon.h" |
| 1231 +#include "secerr.h" |
| 1232 +#include "blapit.h" |
| 1233 +#include "poly1305/poly1305.h" |
| 1234 +#include "chacha20/chacha20.h" |
| 1235 +#include "chacha20poly1305.h" |
| 1236 + |
| 1237 +/* Poly1305Do writes the Poly1305 authenticator of the given additional data |
| 1238 + * and ciphertext to |out|. */ |
| 1239 +static void |
| 1240 +Poly1305Do(unsigned char *out, |
| 1241 + const unsigned char *ad, unsigned int adLen, |
| 1242 + const unsigned char *ciphertext, unsigned int ciphertextLen, |
| 1243 + const unsigned char key[32]) |
| 1244 +{ |
| 1245 + poly1305_state state; |
| 1246 + unsigned int j; |
| 1247 + unsigned char lengthBytes[8]; |
| 1248 + unsigned int i; |
| 1249 + |
| 1250 + Poly1305Init(&state, key); |
| 1251 + j = adLen; |
| 1252 + for (i = 0; i < sizeof(lengthBytes); i++) { |
| 1253 + lengthBytes[i] = j; |
| 1254 + j >>= 8; |
| 1255 + } |
| 1256 + Poly1305Update(&state, ad, adLen); |
| 1257 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes)); |
| 1258 + j = ciphertextLen; |
| 1259 + for (i = 0; i < sizeof(lengthBytes); i++) { |
| 1260 + lengthBytes[i] = j; |
| 1261 + j >>= 8; |
| 1262 + } |
| 1263 + Poly1305Update(&state, ciphertext, ciphertextLen); |
| 1264 + Poly1305Update(&state, lengthBytes, sizeof(lengthBytes)); |
| 1265 + Poly1305Finish(&state, out); |
| 1266 +} |
| 1267 + |
| 1268 +SECStatus |
| 1269 +ChaCha20Poly1305_InitContext(ChaCha20Poly1305Context *ctx, |
| 1270 + const unsigned char *key, unsigned int keyLen, |
| 1271 + unsigned int tagLen) |
| 1272 +{ |
| 1273 + if (keyLen != 32) { |
| 1274 + PORT_SetError(SEC_ERROR_BAD_KEY); |
| 1275 + return SECFailure; |
| 1276 + } |
| 1277 + if (tagLen == 0 || tagLen > 16) { |
| 1278 + PORT_SetError(SEC_ERROR_INPUT_LEN); |
| 1279 + return SECFailure; |
| 1280 + } |
| 1281 + |
| 1282 + memcpy(ctx->key, key, sizeof(ctx->key)); |
| 1283 + ctx->tagLen = tagLen; |
| 1284 + |
| 1285 + return SECSuccess; |
| 1286 +} |
| 1287 + |
| 1288 +ChaCha20Poly1305Context * |
| 1289 +ChaCha20Poly1305_CreateContext(const unsigned char *key, unsigned int keyLen, |
| 1290 + unsigned int tagLen) |
| 1291 +{ |
| 1292 + ChaCha20Poly1305Context *ctx; |
| 1293 + |
| 1294 + ctx = PORT_New(ChaCha20Poly1305Context); |
| 1295 + if (ctx == NULL) { |
| 1296 + return NULL; |
| 1297 + } |
| 1298 + |
| 1299 + if (ChaCha20Poly1305_InitContext(ctx, key, keyLen, tagLen) != SECSuccess) { |
| 1300 + PORT_Free(ctx); |
| 1301 + ctx = NULL; |
| 1302 + } |
| 1303 + |
| 1304 + return ctx; |
| 1305 +} |
| 1306 + |
| 1307 +void |
| 1308 +ChaCha20Poly1305_DestroyContext(ChaCha20Poly1305Context *ctx, PRBool freeit) |
| 1309 +{ |
| 1310 + memset(ctx, 0, sizeof(*ctx)); |
| 1311 + if (freeit) { |
| 1312 + PORT_Free(ctx); |
| 1313 + } |
| 1314 +} |
| 1315 + |
| 1316 +SECStatus |
| 1317 +ChaCha20Poly1305_Seal(const ChaCha20Poly1305Context *ctx, |
| 1318 + unsigned char *output, unsigned int *outputLen, |
| 1319 + unsigned int maxOutputLen, |
| 1320 + const unsigned char *input, unsigned int inputLen, |
| 1321 + const unsigned char *nonce, unsigned int nonceLen, |
| 1322 + const unsigned char *ad, unsigned int adLen) |
| 1323 +{ |
| 1324 + unsigned char block[64]; |
| 1325 + unsigned char tag[16]; |
| 1326 + |
| 1327 + if (nonceLen != 8) { |
| 1328 + PORT_SetError(SEC_ERROR_INPUT_LEN); |
| 1329 + return SECFailure; |
| 1330 + } |
| 1331 + *outputLen = inputLen + ctx->tagLen; |
| 1332 + if (maxOutputLen < *outputLen) { |
| 1333 + PORT_SetError(SEC_ERROR_OUTPUT_LEN); |
| 1334 + return SECFailure; |
| 1335 + } |
| 1336 + |
| 1337 + memset(block, 0, sizeof(block)); |
| 1338 + // Generate a block of keystream. The first 32 bytes will be the poly1305 |
| 1339 + // key. The remainder of the block is discarded. |
| 1340 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0); |
| 1341 + ChaCha20XOR(output, input, inputLen, ctx->key, nonce, 1); |
| 1342 + |
| 1343 + Poly1305Do(tag, ad, adLen, output, inputLen, block); |
| 1344 + memcpy(output + inputLen, tag, ctx->tagLen); |
| 1345 + |
| 1346 + return SECSuccess; |
| 1347 +} |
| 1348 + |
| 1349 +SECStatus |
| 1350 +ChaCha20Poly1305_Open(const ChaCha20Poly1305Context *ctx, |
| 1351 + unsigned char *output, unsigned int *outputLen, |
| 1352 + unsigned int maxOutputLen, |
| 1353 + const unsigned char *input, unsigned int inputLen, |
| 1354 + const unsigned char *nonce, unsigned int nonceLen, |
| 1355 + const unsigned char *ad, unsigned int adLen) |
| 1356 +{ |
| 1357 + unsigned char block[64]; |
| 1358 + unsigned char tag[16]; |
| 1359 + |
| 1360 + if (nonceLen != 8) { |
| 1361 + PORT_SetError(SEC_ERROR_INPUT_LEN); |
| 1362 + return SECFailure; |
| 1363 + } |
| 1364 + if (inputLen < ctx->tagLen) { |
| 1365 + PORT_SetError(SEC_ERROR_INPUT_LEN); |
| 1366 + return SECFailure; |
| 1367 + } |
| 1368 + *outputLen = inputLen - ctx->tagLen; |
| 1369 + if (maxOutputLen < *outputLen) { |
| 1370 + PORT_SetError(SEC_ERROR_OUTPUT_LEN); |
| 1371 + return SECFailure; |
| 1372 + } |
| 1373 + |
| 1374 + memset(block, 0, sizeof(block)); |
| 1375 + // Generate a block of keystream. The first 32 bytes will be the poly1305 |
| 1376 + // key. The remainder of the block is discarded. |
| 1377 + ChaCha20XOR(block, block, sizeof(block), ctx->key, nonce, 0); |
| 1378 + Poly1305Do(tag, ad, adLen, input, inputLen - ctx->tagLen, block); |
| 1379 + if (NSS_SecureMemcmp(tag, &input[inputLen - ctx->tagLen], ctx->tagLen) != 0
) { |
| 1380 + PORT_SetError(SEC_ERROR_BAD_DATA); |
| 1381 + return SECFailure; |
| 1382 + } |
| 1383 + |
| 1384 + ChaCha20XOR(output, input, inputLen - ctx->tagLen, ctx->key, nonce, 1); |
| 1385 + |
| 1386 + return SECSuccess; |
| 1387 +} |
| 1388 |
| 1389 Property changes on: nss/lib/freebl/chacha20poly1305.c |
| 1390 ___________________________________________________________________ |
| 1391 Added: svn:eol-style |
| 1392 + LF |
| 1393 |
| 1394 Index: nss/lib/freebl/chacha20/chacha20.h |
| 1395 =================================================================== |
| 1396 --- nss/lib/freebl/chacha20/chacha20.h (revision 0) |
| 1397 +++ nss/lib/freebl/chacha20/chacha20.h (revision 0) |
| 1398 @@ -0,0 +1,22 @@ |
| 1399 +/* |
| 1400 + * chacha20.h - header file for ChaCha20 implementation. |
| 1401 + * |
| 1402 + * This Source Code Form is subject to the terms of the Mozilla Public |
| 1403 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 1404 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 1405 + |
| 1406 +#ifndef FREEBL_CHACHA20_H_ |
| 1407 +#define FREEBL_CHACHA20_H_ |
| 1408 + |
| 1409 +#include <stdint.h> |
| 1410 + |
| 1411 +/* ChaCha20XOR encrypts |inLen| bytes from |in| with the given key and |
| 1412 + * nonce and writes the result to |out|, which may be equal to |in|. The |
| 1413 + * initial block counter is specified by |counter|. */ |
| 1414 +extern void ChaCha20XOR(unsigned char *out, |
| 1415 + const unsigned char *in, unsigned int inLen, |
| 1416 + const unsigned char key[32], |
| 1417 + const unsigned char nonce[8], |
| 1418 + uint64_t counter); |
| 1419 + |
| 1420 +#endif /* FREEBL_CHACHA20_H_ */ |
| 1421 |
| 1422 Property changes on: nss/lib/freebl/chacha20/chacha20.h |
| 1423 ___________________________________________________________________ |
| 1424 Added: svn:eol-style |
| 1425 + LF |
| 1426 |
| 1427 Index: nss/lib/freebl/chacha20/chacha20_vec.c |
| 1428 =================================================================== |
| 1429 --- nss/lib/freebl/chacha20/chacha20_vec.c (revision 0) |
| 1430 +++ nss/lib/freebl/chacha20/chacha20_vec.c (revision 0) |
| 1431 @@ -0,0 +1,281 @@ |
| 1432 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 1433 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 1434 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 1435 + |
| 1436 +/* This implementation is by Ted Krovetz and was submitted to SUPERCOP and |
| 1437 + * marked as public domain. It was been altered to allow for non-aligned inputs |
| 1438 + * and to allow the block counter to be passed in specifically. */ |
| 1439 + |
| 1440 +#include <string.h> |
| 1441 + |
| 1442 +#include "chacha20.h" |
| 1443 + |
| 1444 +#ifndef CHACHA_RNDS |
| 1445 +#define CHACHA_RNDS 20 /* 8 (high speed), 20 (conservative), 12 (middle) */ |
| 1446 +#endif |
| 1447 + |
| 1448 +/* Architecture-neutral way to specify 16-byte vector of ints */ |
| 1449 +typedef unsigned vec __attribute__ ((vector_size (16))); |
| 1450 + |
| 1451 +/* This implementation is designed for Neon, SSE and AltiVec machines. The |
| 1452 + * following specify how to do certain vector operations efficiently on |
| 1453 + * each architecture, using intrinsics. |
| 1454 + * This implementation supports parallel processing of multiple blocks, |
| 1455 + * including potentially using general-purpose registers. |
| 1456 + */ |
| 1457 +#if __ARM_NEON__ |
| 1458 +#include <arm_neon.h> |
| 1459 +#define GPR_TOO 1 |
| 1460 +#define VBPI 2 |
| 1461 +#define ONE (vec)vsetq_lane_u32(1,vdupq_n_u32(0),0) |
| 1462 +#define LOAD(m) (vec)(*((vec*)(m))) |
| 1463 +#define STORE(m,r) (*((vec*)(m))) = (r) |
| 1464 +#define ROTV1(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,1) |
| 1465 +#define ROTV2(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,2) |
| 1466 +#define ROTV3(x) (vec)vextq_u32((uint32x4_t)x,(uint32x4_t)x,3) |
| 1467 +#define ROTW16(x) (vec)vrev32q_u16((uint16x8_t)x) |
| 1468 +#if __clang__ |
| 1469 +#define ROTW7(x) (x << ((vec){ 7, 7, 7, 7})) ^ (x >> ((vec){25,25,25,25})) |
| 1470 +#define ROTW8(x) (x << ((vec){ 8, 8, 8, 8})) ^ (x >> ((vec){24,24,24,24})) |
| 1471 +#define ROTW12(x) (x << ((vec){12,12,12,12})) ^ (x >> ((vec){20,20,20,20})) |
| 1472 +#else |
| 1473 +#define ROTW7(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,7),(uint32x4_t)x,2
5) |
| 1474 +#define ROTW8(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,8),(uint32x4_t)x,2
4) |
| 1475 +#define ROTW12(x) (vec)vsriq_n_u32(vshlq_n_u32((uint32x4_t)x,12),(uint32x4_t)x,
20) |
| 1476 +#endif |
| 1477 +#elif __SSE2__ |
| 1478 +#include <emmintrin.h> |
| 1479 +#define GPR_TOO 0 |
| 1480 +#if __clang__ |
| 1481 +#define VBPI 4 |
| 1482 +#else |
| 1483 +#define VBPI 3 |
| 1484 +#endif |
| 1485 +#define ONE (vec)_mm_set_epi32(0,0,0,1) |
| 1486 +#define LOAD(m) (vec)_mm_loadu_si128((__m128i*)(m)) |
| 1487 +#define STORE(m,r) _mm_storeu_si128((__m128i*)(m), (__m128i) (r)) |
| 1488 +#define ROTV1(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(0,3,2,1)) |
| 1489 +#define ROTV2(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(1,0,3,2)) |
| 1490 +#define ROTV3(x) (vec)_mm_shuffle_epi32((__m128i)x,_MM_SHUFFLE(2,1,0,3)) |
| 1491 +#define ROTW7(x) (vec)(_mm_slli_epi32((__m128i)x, 7) ^ _mm_srli_epi32((__m128i
)x,25)) |
| 1492 +#define ROTW12(x) (vec)(_mm_slli_epi32((__m128i)x,12) ^ _mm_srli_epi32((__m128i
)x,20)) |
| 1493 +#if __SSSE3__ |
| 1494 +#include <tmmintrin.h> |
| 1495 +#define ROTW8(x) (vec)_mm_shuffle_epi8((__m128i)x,_mm_set_epi8(14,13,12,15,10,
9,8,11,6,5,4,7,2,1,0,3)) |
| 1496 +#define ROTW16(x) (vec)_mm_shuffle_epi8((__m128i)x,_mm_set_epi8(13,12,15,14,9,8
,11,10,5,4,7,6,1,0,3,2)) |
| 1497 +#else |
| 1498 +#define ROTW8(x) (vec)(_mm_slli_epi32((__m128i)x, 8) ^ _mm_srli_epi32((__m128i
)x,24)) |
| 1499 +#define ROTW16(x) (vec)(_mm_slli_epi32((__m128i)x,16) ^ _mm_srli_epi32((__m128i
)x,16)) |
| 1500 +#endif |
| 1501 +#else |
| 1502 +#error -- Implementation supports only machines with neon or SSE2 |
| 1503 +#endif |
| 1504 + |
| 1505 +#ifndef REVV_BE |
| 1506 +#define REVV_BE(x) (x) |
| 1507 +#endif |
| 1508 + |
| 1509 +#ifndef REVW_BE |
| 1510 +#define REVW_BE(x) (x) |
| 1511 +#endif |
| 1512 + |
| 1513 +#define BPI (VBPI + GPR_TOO) /* Blocks computed per loop iteration */ |
| 1514 + |
| 1515 +#define DQROUND_VECTORS(a,b,c,d) \ |
| 1516 + a += b; d ^= a; d = ROTW16(d); \ |
| 1517 + c += d; b ^= c; b = ROTW12(b); \ |
| 1518 + a += b; d ^= a; d = ROTW8(d); \ |
| 1519 + c += d; b ^= c; b = ROTW7(b); \ |
| 1520 + b = ROTV1(b); c = ROTV2(c); d = ROTV3(d); \ |
| 1521 + a += b; d ^= a; d = ROTW16(d); \ |
| 1522 + c += d; b ^= c; b = ROTW12(b); \ |
| 1523 + a += b; d ^= a; d = ROTW8(d); \ |
| 1524 + c += d; b ^= c; b = ROTW7(b); \ |
| 1525 + b = ROTV3(b); c = ROTV2(c); d = ROTV1(d); |
| 1526 + |
| 1527 +#define QROUND_WORDS(a,b,c,d) \ |
| 1528 + a = a+b; d ^= a; d = d<<16 | d>>16; \ |
| 1529 + c = c+d; b ^= c; b = b<<12 | b>>20; \ |
| 1530 + a = a+b; d ^= a; d = d<< 8 | d>>24; \ |
| 1531 + c = c+d; b ^= c; b = b<< 7 | b>>25; |
| 1532 + |
| 1533 +#define WRITE_XOR(in, op, d, v0, v1, v2, v3) \ |
| 1534 + STORE(op + d + 0, LOAD(in + d + 0) ^ REVV_BE(v0)); \ |
| 1535 + STORE(op + d + 4, LOAD(in + d + 4) ^ REVV_BE(v1)); \ |
| 1536 + STORE(op + d + 8, LOAD(in + d + 8) ^ REVV_BE(v2)); \ |
| 1537 + STORE(op + d +12, LOAD(in + d +12) ^ REVV_BE(v3)); |
| 1538 + |
| 1539 +void ChaCha20XOR( |
| 1540 + unsigned char *out, |
| 1541 + const unsigned char *in, |
| 1542 + unsigned int inlen, |
| 1543 + const unsigned char key[32], |
| 1544 + const unsigned char nonce[8], |
| 1545 + uint64_t counter) |
| 1546 +{ |
| 1547 + unsigned iters, i, *op=(unsigned *)out, *ip=(unsigned *)in, *kp; |
| 1548 +#if defined(__ARM_NEON__) |
| 1549 + unsigned *np; |
| 1550 +#endif |
| 1551 + vec s0, s1, s2, s3; |
| 1552 +#if !defined(__ARM_NEON__) && !defined(__SSE2__) |
| 1553 + __attribute__ ((aligned (16))) unsigned key[8], nonce[4]; |
| 1554 +#endif |
| 1555 + __attribute__ ((aligned (16))) unsigned chacha_const[] = |
| 1556 + {0x61707865,0x3320646E,0x79622D32,0x6B206574}; |
| 1557 +#if defined(__ARM_NEON__) || defined(__SSE2__) |
| 1558 + kp = (unsigned *)key; |
| 1559 +#else |
| 1560 + ((vec *)key)[0] = REVV_BE(((vec *)key)[0]); |
| 1561 + ((vec *)key)[1] = REVV_BE(((vec *)key)[1]); |
| 1562 + nonce[0] = REVW_BE(((unsigned *)nonce)[0]); |
| 1563 + nonce[1] = REVW_BE(((unsigned *)nonce)[1]); |
| 1564 + nonce[2] = REVW_BE(((unsigned *)nonce)[2]); |
| 1565 + nonce[3] = REVW_BE(((unsigned *)nonce)[3]); |
| 1566 + kp = (unsigned *)key; |
| 1567 + np = (unsigned *)nonce; |
| 1568 +#endif |
| 1569 +#if defined(__ARM_NEON__) |
| 1570 + np = (unsigned*) nonce; |
| 1571 +#endif |
| 1572 + s0 = LOAD(chacha_const); |
| 1573 + s1 = LOAD(&((vec*)kp)[0]); |
| 1574 + s2 = LOAD(&((vec*)kp)[1]); |
| 1575 + s3 = (vec) { |
| 1576 + counter & 0xffffffff, |
| 1577 + counter >> 32, |
| 1578 + ((uint32_t*)nonce)[0], |
| 1579 + ((uint32_t*)nonce)[1] |
| 1580 + }; |
| 1581 + |
| 1582 + for (iters = 0; iters < inlen/(BPI*64); iters++) { |
| 1583 +#if GPR_TOO |
| 1584 + register unsigned x0, x1, x2, x3, x4, x5, x6, x7, x8, |
| 1585 + x9, x10, x11, x12, x13, x14, x15; |
| 1586 +#endif |
| 1587 +#if VBPI > 2 |
| 1588 + vec v8,v9,v10,v11; |
| 1589 +#endif |
| 1590 +#if VBPI > 3 |
| 1591 + vec v12,v13,v14,v15; |
| 1592 +#endif |
| 1593 + |
| 1594 + vec v0,v1,v2,v3,v4,v5,v6,v7; |
| 1595 + v4 = v0 = s0; v5 = v1 = s1; v6 = v2 = s2; v3 = s3; |
| 1596 + v7 = v3 + ONE; |
| 1597 +#if VBPI > 2 |
| 1598 + v8 = v4; v9 = v5; v10 = v6; |
| 1599 + v11 = v7 + ONE; |
| 1600 +#endif |
| 1601 +#if VBPI > 3 |
| 1602 + v12 = v8; v13 = v9; v14 = v10; |
| 1603 + v15 = v11 + ONE; |
| 1604 +#endif |
| 1605 +#if GPR_TOO |
| 1606 + x0 = chacha_const[0]; x1 = chacha_const[1]; |
| 1607 + x2 = chacha_const[2]; x3 = chacha_const[3]; |
| 1608 + x4 = kp[0]; x5 = kp[1]; x6 = kp[2]; x7 = kp[3]; |
| 1609 + x8 = kp[4]; x9 = kp[5]; x10 = kp[6]; x11 = kp[7]; |
| 1610 + x12 = (counter & 0xffffffff)+BPI*iters+(BPI-1); x13 = counter >> 32; |
| 1611 + x14 = np[0]; x15 = np[1]; |
| 1612 +#endif |
| 1613 + for (i = CHACHA_RNDS/2; i; i--) { |
| 1614 + DQROUND_VECTORS(v0,v1,v2,v3) |
| 1615 + DQROUND_VECTORS(v4,v5,v6,v7) |
| 1616 +#if VBPI > 2 |
| 1617 + DQROUND_VECTORS(v8,v9,v10,v11) |
| 1618 +#endif |
| 1619 +#if VBPI > 3 |
| 1620 + DQROUND_VECTORS(v12,v13,v14,v15) |
| 1621 +#endif |
| 1622 +#if GPR_TOO |
| 1623 + QROUND_WORDS( x0, x4, x8,x12) |
| 1624 + QROUND_WORDS( x1, x5, x9,x13) |
| 1625 + QROUND_WORDS( x2, x6,x10,x14) |
| 1626 + QROUND_WORDS( x3, x7,x11,x15) |
| 1627 + QROUND_WORDS( x0, x5,x10,x15) |
| 1628 + QROUND_WORDS( x1, x6,x11,x12) |
| 1629 + QROUND_WORDS( x2, x7, x8,x13) |
| 1630 + QROUND_WORDS( x3, x4, x9,x14) |
| 1631 +#endif |
| 1632 + } |
| 1633 + |
| 1634 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3) |
| 1635 + s3 += ONE; |
| 1636 + WRITE_XOR(ip, op, 16, v4+s0, v5+s1, v6+s2, v7+s3) |
| 1637 + s3 += ONE; |
| 1638 +#if VBPI > 2 |
| 1639 + WRITE_XOR(ip, op, 32, v8+s0, v9+s1, v10+s2, v11+s3) |
| 1640 + s3 += ONE; |
| 1641 +#endif |
| 1642 +#if VBPI > 3 |
| 1643 + WRITE_XOR(ip, op, 48, v12+s0, v13+s1, v14+s2, v15+s3) |
| 1644 + s3 += ONE; |
| 1645 +#endif |
| 1646 + ip += VBPI*16; |
| 1647 + op += VBPI*16; |
| 1648 +#if GPR_TOO |
| 1649 + op[0] = REVW_BE(REVW_BE(ip[0]) ^ (x0 + chacha_const[0])); |
| 1650 + op[1] = REVW_BE(REVW_BE(ip[1]) ^ (x1 + chacha_const[1])); |
| 1651 + op[2] = REVW_BE(REVW_BE(ip[2]) ^ (x2 + chacha_const[2])); |
| 1652 + op[3] = REVW_BE(REVW_BE(ip[3]) ^ (x3 + chacha_const[3])); |
| 1653 + op[4] = REVW_BE(REVW_BE(ip[4]) ^ (x4 + kp[0])); |
| 1654 + op[5] = REVW_BE(REVW_BE(ip[5]) ^ (x5 + kp[1])); |
| 1655 + op[6] = REVW_BE(REVW_BE(ip[6]) ^ (x6 + kp[2])); |
| 1656 + op[7] = REVW_BE(REVW_BE(ip[7]) ^ (x7 + kp[3])); |
| 1657 + op[8] = REVW_BE(REVW_BE(ip[8]) ^ (x8 + kp[4])); |
| 1658 + op[9] = REVW_BE(REVW_BE(ip[9]) ^ (x9 + kp[5])); |
| 1659 + op[10] = REVW_BE(REVW_BE(ip[10]) ^ (x10 + kp[6])); |
| 1660 + op[11] = REVW_BE(REVW_BE(ip[11]) ^ (x11 + kp[7])); |
| 1661 + op[12] = REVW_BE(REVW_BE(ip[12]) ^ (x12 + (counter & 0xffffffff)+BPI*ite
rs+(BPI-1))); |
| 1662 + op[13] = REVW_BE(REVW_BE(ip[13]) ^ (x13 + (counter >> 32))); |
| 1663 + op[14] = REVW_BE(REVW_BE(ip[14]) ^ (x14 + np[0])); |
| 1664 + op[15] = REVW_BE(REVW_BE(ip[15]) ^ (x15 + np[1])); |
| 1665 + s3 += ONE; |
| 1666 + ip += 16; |
| 1667 + op += 16; |
| 1668 +#endif |
| 1669 + } |
| 1670 + |
| 1671 + for (iters = inlen%(BPI*64)/64; iters != 0; iters--) { |
| 1672 + vec v0 = s0, v1 = s1, v2 = s2, v3 = s3; |
| 1673 + for (i = CHACHA_RNDS/2; i; i--) { |
| 1674 + DQROUND_VECTORS(v0,v1,v2,v3); |
| 1675 + } |
| 1676 + WRITE_XOR(ip, op, 0, v0+s0, v1+s1, v2+s2, v3+s3) |
| 1677 + s3 += ONE; |
| 1678 + ip += 16; |
| 1679 + op += 16; |
| 1680 + } |
| 1681 + |
| 1682 + inlen = inlen % 64; |
| 1683 + if (inlen) { |
| 1684 + __attribute__ ((aligned (16))) vec buf[4]; |
| 1685 + vec v0,v1,v2,v3; |
| 1686 + v0 = s0; v1 = s1; v2 = s2; v3 = s3; |
| 1687 + for (i = CHACHA_RNDS/2; i; i--) { |
| 1688 + DQROUND_VECTORS(v0,v1,v2,v3); |
| 1689 + } |
| 1690 + |
| 1691 + if (inlen >= 16) { |
| 1692 + STORE(op + 0, LOAD(ip + 0) ^ REVV_BE(v0 + s0)); |
| 1693 + if (inlen >= 32) { |
| 1694 + STORE(op + 4, LOAD(ip + 4) ^ REVV_BE(v1 + s1)); |
| 1695 + if (inlen >= 48) { |
| 1696 + STORE(op + 8, LOAD(ip + 8) ^ REVV_BE(v2 + s2)); |
| 1697 + buf[3] = REVV_BE(v3 + s3); |
| 1698 + } else { |
| 1699 + buf[2] = REVV_BE(v2 + s2); |
| 1700 + } |
| 1701 + } else { |
| 1702 + buf[1] = REVV_BE(v1 + s1); |
| 1703 + } |
| 1704 + } else { |
| 1705 + buf[0] = REVV_BE(v0 + s0); |
| 1706 + } |
| 1707 + |
| 1708 + for (i=inlen & ~15; i<inlen; i++) { |
| 1709 + ((char *)op)[i] = ((char *)ip)[i] ^ ((char *)buf)[i]; |
| 1710 + } |
| 1711 + } |
| 1712 +} |
| 1713 |
| 1714 Property changes on: nss/lib/freebl/chacha20/chacha20_vec.c |
| 1715 ___________________________________________________________________ |
| 1716 Added: svn:eol-style |
| 1717 + LF |
| 1718 |
| 1719 Index: nss/lib/freebl/chacha20/chacha20.c |
| 1720 =================================================================== |
| 1721 --- nss/lib/freebl/chacha20/chacha20.c (revision 0) |
| 1722 +++ nss/lib/freebl/chacha20/chacha20.c (revision 0) |
| 1723 @@ -0,0 +1,108 @@ |
| 1724 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 1725 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 1726 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 1727 + |
| 1728 +/* Adopted from the public domain code in NaCl by djb. */ |
| 1729 + |
| 1730 +#include <string.h> |
| 1731 +#include <stdio.h> |
| 1732 + |
| 1733 +#include "prtypes.h" |
| 1734 +#include "chacha20.h" |
| 1735 + |
| 1736 +#define ROTL32(v, n) (((v) << (n)) | ((v) >> (32 - (n)))) |
| 1737 +#define ROTATE(v, c) ROTL32((v), (c)) |
| 1738 +#define XOR(v, w) ((v) ^ (w)) |
| 1739 +#define PLUS(x, y) ((x) + (y)) |
| 1740 + |
| 1741 +#define U32TO8_LITTLE(p, v) \ |
| 1742 + { (p)[0] = ((v) ) & 0xff; (p)[1] = ((v) >> 8) & 0xff; \ |
| 1743 + (p)[2] = ((v) >> 16) & 0xff; (p)[3] = ((v) >> 24) & 0xff; } |
| 1744 +#define U8TO32_LITTLE(p) \ |
| 1745 + (((PRUint32)((p)[0]) ) | ((PRUint32)((p)[1]) << 8) | \ |
| 1746 + ((PRUint32)((p)[2]) << 16) | ((PRUint32)((p)[3]) << 24) ) |
| 1747 + |
| 1748 +#define QUARTERROUND(a,b,c,d) \ |
| 1749 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]),16); \ |
| 1750 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]),12); \ |
| 1751 + x[a] = PLUS(x[a],x[b]); x[d] = ROTATE(XOR(x[d],x[a]), 8); \ |
| 1752 + x[c] = PLUS(x[c],x[d]); x[b] = ROTATE(XOR(x[b],x[c]), 7); |
| 1753 + |
| 1754 +static void ChaChaCore(unsigned char output[64], const PRUint32 input[16], |
| 1755 + int num_rounds) { |
| 1756 + PRUint32 x[16]; |
| 1757 + int i; |
| 1758 + |
| 1759 + memcpy(x, input, sizeof(PRUint32) * 16); |
| 1760 + for (i = num_rounds; i > 0; i -= 2) { |
| 1761 + QUARTERROUND( 0, 4, 8,12) |
| 1762 + QUARTERROUND( 1, 5, 9,13) |
| 1763 + QUARTERROUND( 2, 6,10,14) |
| 1764 + QUARTERROUND( 3, 7,11,15) |
| 1765 + QUARTERROUND( 0, 5,10,15) |
| 1766 + QUARTERROUND( 1, 6,11,12) |
| 1767 + QUARTERROUND( 2, 7, 8,13) |
| 1768 + QUARTERROUND( 3, 4, 9,14) |
| 1769 + } |
| 1770 + |
| 1771 + for (i = 0; i < 16; ++i) { |
| 1772 + x[i] = PLUS(x[i], input[i]); |
| 1773 + } |
| 1774 + for (i = 0; i < 16; ++i) { |
| 1775 + U32TO8_LITTLE(output + 4 * i, x[i]); |
| 1776 + } |
| 1777 +} |
| 1778 + |
| 1779 +static const unsigned char sigma[16] = "expand 32-byte k"; |
| 1780 + |
| 1781 +void ChaCha20XOR(unsigned char *out, const unsigned char *in, unsigned int inLe
n, |
| 1782 + const unsigned char key[32], const unsigned char nonce[8], |
| 1783 + uint64_t counter) { |
| 1784 + unsigned char block[64]; |
| 1785 + PRUint32 input[16]; |
| 1786 + unsigned int u; |
| 1787 + unsigned int i; |
| 1788 + |
| 1789 + input[4] = U8TO32_LITTLE(key + 0); |
| 1790 + input[5] = U8TO32_LITTLE(key + 4); |
| 1791 + input[6] = U8TO32_LITTLE(key + 8); |
| 1792 + input[7] = U8TO32_LITTLE(key + 12); |
| 1793 + |
| 1794 + input[8] = U8TO32_LITTLE(key + 16); |
| 1795 + input[9] = U8TO32_LITTLE(key + 20); |
| 1796 + input[10] = U8TO32_LITTLE(key + 24); |
| 1797 + input[11] = U8TO32_LITTLE(key + 28); |
| 1798 + |
| 1799 + input[0] = U8TO32_LITTLE(sigma + 0); |
| 1800 + input[1] = U8TO32_LITTLE(sigma + 4); |
| 1801 + input[2] = U8TO32_LITTLE(sigma + 8); |
| 1802 + input[3] = U8TO32_LITTLE(sigma + 12); |
| 1803 + |
| 1804 + input[12] = counter; |
| 1805 + input[13] = counter >> 32; |
| 1806 + input[14] = U8TO32_LITTLE(nonce + 0); |
| 1807 + input[15] = U8TO32_LITTLE(nonce + 4); |
| 1808 + |
| 1809 + while (inLen >= 64) { |
| 1810 + ChaChaCore(block, input, 20); |
| 1811 + for (i = 0; i < 64; i++) { |
| 1812 + out[i] = in[i] ^ block[i]; |
| 1813 + } |
| 1814 + |
| 1815 + input[12]++; |
| 1816 + if (input[12] == 0) { |
| 1817 + input[13]++; |
| 1818 + } |
| 1819 + |
| 1820 + inLen -= 64; |
| 1821 + in += 64; |
| 1822 + out += 64; |
| 1823 + } |
| 1824 + |
| 1825 + if (inLen > 0) { |
| 1826 + ChaChaCore(block, input, 20); |
| 1827 + for (i = 0; i < inLen; i++) { |
| 1828 + out[i] = in[i] ^ block[i]; |
| 1829 + } |
| 1830 + } |
| 1831 +} |
| 1832 |
| 1833 Property changes on: nss/lib/freebl/chacha20/chacha20.c |
| 1834 ___________________________________________________________________ |
| 1835 Added: svn:eol-style |
| 1836 + LF |
| 1837 |
| 1838 Index: nss/lib/freebl/chacha20poly1305.h |
| 1839 =================================================================== |
| 1840 --- nss/lib/freebl/chacha20poly1305.h (revision 0) |
| 1841 +++ nss/lib/freebl/chacha20poly1305.h (revision 0) |
| 1842 @@ -0,0 +1,15 @@ |
| 1843 +/* This Source Code Form is subject to the terms of the Mozilla Public |
| 1844 + * License, v. 2.0. If a copy of the MPL was not distributed with this |
| 1845 + * file, You can obtain one at http://mozilla.org/MPL/2.0/. */ |
| 1846 + |
| 1847 +#ifndef _CHACHA20_POLY1305_H_ |
| 1848 +#define _CHACHA20_POLY1305_H_ 1 |
| 1849 + |
| 1850 +/* ChaCha20Poly1305ContextStr saves the key and tag length for a |
| 1851 + * ChaCha20+Poly1305 AEAD operation. */ |
| 1852 +struct ChaCha20Poly1305ContextStr { |
| 1853 + unsigned char key[32]; |
| 1854 + unsigned char tagLen; |
| 1855 +}; |
| 1856 + |
| 1857 +#endif /* _CHACHA20_POLY1305_H_ */ |
| 1858 |
| 1859 Property changes on: nss/lib/freebl/chacha20poly1305.h |
| 1860 ___________________________________________________________________ |
| 1861 Added: svn:eol-style |
| 1862 + LF |
| 1863 |
| 1864 Index: nss/lib/pk11wrap/pk11mech.c |
| 1865 =================================================================== |
| 1866 --- nss/lib/pk11wrap/pk11mech.c (revision 228205) |
| 1867 +++ nss/lib/pk11wrap/pk11mech.c (working copy) |
| 1868 @@ -152,6 +152,8 @@ |
| 1869 return CKM_SEED_CBC; |
| 1870 case CKK_CAMELLIA: |
| 1871 return CKM_CAMELLIA_CBC; |
| 1872 + case CKK_NSS_CHACHA20: |
| 1873 + return CKM_NSS_CHACHA20_POLY1305; |
| 1874 case CKK_AES: |
| 1875 return CKM_AES_CBC; |
| 1876 case CKK_DES: |
| 1877 @@ -219,6 +221,8 @@ |
| 1878 case CKM_CAMELLIA_CBC_PAD: |
| 1879 case CKM_CAMELLIA_KEY_GEN: |
| 1880 return CKK_CAMELLIA; |
| 1881 + case CKM_NSS_CHACHA20_POLY1305: |
| 1882 + return CKK_NSS_CHACHA20; |
| 1883 case CKM_AES_ECB: |
| 1884 case CKM_AES_CBC: |
| 1885 case CKM_AES_CCM: |
| 1886 @@ -429,6 +433,8 @@ |
| 1887 case CKM_CAMELLIA_CBC_PAD: |
| 1888 case CKM_CAMELLIA_KEY_GEN: |
| 1889 return CKM_CAMELLIA_KEY_GEN; |
| 1890 + case CKM_NSS_CHACHA20_POLY1305: |
| 1891 + return CKM_NSS_CHACHA20_KEY_GEN; |
| 1892 case CKM_AES_ECB: |
| 1893 case CKM_AES_CBC: |
| 1894 case CKM_AES_CCM: |
| 1895 Index: nss/lib/util/pkcs11n.h |
| 1896 =================================================================== |
| 1897 --- nss/lib/util/pkcs11n.h (revision 228205) |
| 1898 +++ nss/lib/util/pkcs11n.h (working copy) |
| 1899 @@ -51,6 +51,8 @@ |
| 1900 #define CKK_NSS_JPAKE_ROUND1 (CKK_NSS + 2) |
| 1901 #define CKK_NSS_JPAKE_ROUND2 (CKK_NSS + 3) |
| 1902 |
| 1903 +#define CKK_NSS_CHACHA20 (CKK_NSS + 4) |
| 1904 + |
| 1905 /* |
| 1906 * NSS-defined certificate types |
| 1907 * |
| 1908 @@ -214,6 +216,9 @@ |
| 1909 #define CKM_NSS_TLS_KEY_AND_MAC_DERIVE_SHA256 (CKM_NSS + 23) |
| 1910 #define CKM_NSS_TLS_MASTER_KEY_DERIVE_DH_SHA256 (CKM_NSS + 24) |
| 1911 |
| 1912 +#define CKM_NSS_CHACHA20_KEY_GEN (CKM_NSS + 25) |
| 1913 +#define CKM_NSS_CHACHA20_POLY1305 (CKM_NSS + 26) |
| 1914 + |
| 1915 /* |
| 1916 * HISTORICAL: |
| 1917 * Do not attempt to use these. They are only used by NETSCAPE's internal |
| 1918 @@ -281,6 +286,14 @@ |
| 1919 CK_ULONG ulHeaderLen; /* in */ |
| 1920 } CK_NSS_MAC_CONSTANT_TIME_PARAMS; |
| 1921 |
| 1922 +typedef struct CK_NSS_AEAD_PARAMS { |
| 1923 + CK_BYTE_PTR pIv; /* This is the nonce. */ |
| 1924 + CK_ULONG ulIvLen; |
| 1925 + CK_BYTE_PTR pAAD; |
| 1926 + CK_ULONG ulAADLen; |
| 1927 + CK_ULONG ulTagLen; |
| 1928 +} CK_NSS_AEAD_PARAMS; |
| 1929 + |
| 1930 /* |
| 1931 * NSS-defined return values |
| 1932 * |
OLD | NEW |