e_aes_cbc_hmac_sha256.c revision 325335
1/* ==================================================================== 2 * Copyright (c) 2011-2013 The OpenSSL Project. All rights reserved. 3 * 4 * Redistribution and use in source and binary forms, with or without 5 * modification, are permitted provided that the following conditions 6 * are met: 7 * 8 * 1. Redistributions of source code must retain the above copyright 9 * notice, this list of conditions and the following disclaimer. 10 * 11 * 2. Redistributions in binary form must reproduce the above copyright 12 * notice, this list of conditions and the following disclaimer in 13 * the documentation and/or other materials provided with the 14 * distribution. 15 * 16 * 3. All advertising materials mentioning features or use of this 17 * software must display the following acknowledgment: 18 * "This product includes software developed by the OpenSSL Project 19 * for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)" 20 * 21 * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to 22 * endorse or promote products derived from this software without 23 * prior written permission. For written permission, please contact 24 * licensing@OpenSSL.org. 25 * 26 * 5. Products derived from this software may not be called "OpenSSL" 27 * nor may "OpenSSL" appear in their names without prior written 28 * permission of the OpenSSL Project. 29 * 30 * 6. Redistributions of any form whatsoever must retain the following 31 * acknowledgment: 32 * "This product includes software developed by the OpenSSL Project 33 * for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)" 34 * 35 * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY 36 * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE 37 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 38 * PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR 39 * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, 40 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT 41 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 42 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) 43 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, 44 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) 45 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED 46 * OF THE POSSIBILITY OF SUCH DAMAGE. 47 * ==================================================================== 48 */ 49 50#include <openssl/opensslconf.h> 51 52#include <stdio.h> 53#include <string.h> 54 55#if !defined(OPENSSL_NO_AES) && !defined(OPENSSL_NO_SHA256) 56 57# include <openssl/evp.h> 58# include <openssl/objects.h> 59# include <openssl/aes.h> 60# include <openssl/sha.h> 61# include <openssl/rand.h> 62# include "modes_lcl.h" 63# include "constant_time_locl.h" 64 65# ifndef EVP_CIPH_FLAG_AEAD_CIPHER 66# define EVP_CIPH_FLAG_AEAD_CIPHER 0x200000 67# define EVP_CTRL_AEAD_TLS1_AAD 0x16 68# define EVP_CTRL_AEAD_SET_MAC_KEY 0x17 69# endif 70 71# if !defined(EVP_CIPH_FLAG_DEFAULT_ASN1) 72# define EVP_CIPH_FLAG_DEFAULT_ASN1 0 73# endif 74 75# if !defined(EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK) 76# define EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK 0 77# endif 78 79# define TLS1_1_VERSION 0x0302 80 81typedef struct { 82 AES_KEY ks; 83 SHA256_CTX head, tail, md; 84 size_t payload_length; /* AAD length in decrypt case */ 85 union { 86 unsigned int tls_ver; 87 unsigned char tls_aad[16]; /* 13 used */ 88 } aux; 89} EVP_AES_HMAC_SHA256; 90 91# define NO_PAYLOAD_LENGTH ((size_t)-1) 92 93# if defined(AES_ASM) && ( \ 94 defined(__x86_64) || defined(__x86_64__) || \ 95 defined(_M_AMD64) || defined(_M_X64) || \ 96 defined(__INTEL__) ) 97 98extern unsigned int OPENSSL_ia32cap_P[]; 99# define AESNI_CAPABLE (1<<(57-32)) 100 101int aesni_set_encrypt_key(const unsigned char *userKey, int bits, 102 AES_KEY *key); 103int aesni_set_decrypt_key(const unsigned char *userKey, int bits, 104 AES_KEY *key); 105 106void aesni_cbc_encrypt(const unsigned char *in, 107 unsigned char *out, 108 size_t length, 109 const AES_KEY *key, unsigned char *ivec, int enc); 110 111int aesni_cbc_sha256_enc(const void *inp, void *out, size_t blocks, 112 const AES_KEY *key, unsigned char iv[16], 113 SHA256_CTX *ctx, const void *in0); 114 115# define data(ctx) ((EVP_AES_HMAC_SHA256 *)(ctx)->cipher_data) 116 117static int aesni_cbc_hmac_sha256_init_key(EVP_CIPHER_CTX *ctx, 118 const unsigned char *inkey, 119 const unsigned char *iv, int enc) 120{ 121 EVP_AES_HMAC_SHA256 *key = data(ctx); 122 int ret; 123 124 if (enc) 125 memset(&key->ks, 0, sizeof(key->ks.rd_key)), 126 ret = aesni_set_encrypt_key(inkey, ctx->key_len * 8, &key->ks); 127 else 128 ret = aesni_set_decrypt_key(inkey, ctx->key_len * 8, &key->ks); 129 130 SHA256_Init(&key->head); /* handy when benchmarking */ 131 key->tail = key->head; 132 key->md = key->head; 133 134 key->payload_length = NO_PAYLOAD_LENGTH; 135 136 return ret < 0 ? 0 : 1; 137} 138 139# define STITCHED_CALL 140 141# if !defined(STITCHED_CALL) 142# define aes_off 0 143# endif 144 145void sha256_block_data_order(void *c, const void *p, size_t len); 146 147static void sha256_update(SHA256_CTX *c, const void *data, size_t len) 148{ 149 const unsigned char *ptr = data; 150 size_t res; 151 152 if ((res = c->num)) { 153 res = SHA256_CBLOCK - res; 154 if (len < res) 155 res = len; 156 SHA256_Update(c, ptr, res); 157 ptr += res; 158 len -= res; 159 } 160 161 res = len % SHA256_CBLOCK; 162 len -= res; 163 164 if (len) { 165 sha256_block_data_order(c, ptr, len / SHA256_CBLOCK); 166 167 ptr += len; 168 c->Nh += len >> 29; 169 c->Nl += len <<= 3; 170 if (c->Nl < (unsigned int)len) 171 c->Nh++; 172 } 173 174 if (res) 175 SHA256_Update(c, ptr, res); 176} 177 178# ifdef SHA256_Update 179# undef SHA256_Update 180# endif 181# define SHA256_Update sha256_update 182 183# if !defined(OPENSSL_NO_MULTIBLOCK) && EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK 184 185typedef struct { 186 unsigned int A[8], B[8], C[8], D[8], E[8], F[8], G[8], H[8]; 187} SHA256_MB_CTX; 188typedef struct { 189 const unsigned char *ptr; 190 int blocks; 191} HASH_DESC; 192 193void sha256_multi_block(SHA256_MB_CTX *, const HASH_DESC *, int); 194 195typedef struct { 196 const unsigned char *inp; 197 unsigned char *out; 198 int blocks; 199 u64 iv[2]; 200} CIPH_DESC; 201 202void aesni_multi_cbc_encrypt(CIPH_DESC *, void *, int); 203 204static size_t tls1_1_multi_block_encrypt(EVP_AES_HMAC_SHA256 *key, 205 unsigned char *out, 206 const unsigned char *inp, 207 size_t inp_len, int n4x) 208{ /* n4x is 1 or 2 */ 209 HASH_DESC hash_d[8], edges[8]; 210 CIPH_DESC ciph_d[8]; 211 unsigned char storage[sizeof(SHA256_MB_CTX) + 32]; 212 union { 213 u64 q[16]; 214 u32 d[32]; 215 u8 c[128]; 216 } blocks[8]; 217 SHA256_MB_CTX *ctx; 218 unsigned int frag, last, packlen, i, x4 = 4 * n4x, minblocks, processed = 219 0; 220 size_t ret = 0; 221 u8 *IVs; 222# if defined(BSWAP8) 223 u64 seqnum; 224# endif 225 226 /* ask for IVs in bulk */ 227 if (RAND_bytes((IVs = blocks[0].c), 16 * x4) <= 0) 228 return 0; 229 230 /* align */ 231 ctx = (SHA256_MB_CTX *) (storage + 32 - ((size_t)storage % 32)); 232 233 frag = (unsigned int)inp_len >> (1 + n4x); 234 last = (unsigned int)inp_len + frag - (frag << (1 + n4x)); 235 if (last > frag && ((last + 13 + 9) % 64) < (x4 - 1)) { 236 frag++; 237 last -= x4 - 1; 238 } 239 240 packlen = 5 + 16 + ((frag + 32 + 16) & -16); 241 242 /* populate descriptors with pointers and IVs */ 243 hash_d[0].ptr = inp; 244 ciph_d[0].inp = inp; 245 /* 5+16 is place for header and explicit IV */ 246 ciph_d[0].out = out + 5 + 16; 247 memcpy(ciph_d[0].out - 16, IVs, 16); 248 memcpy(ciph_d[0].iv, IVs, 16); 249 IVs += 16; 250 251 for (i = 1; i < x4; i++) { 252 ciph_d[i].inp = hash_d[i].ptr = hash_d[i - 1].ptr + frag; 253 ciph_d[i].out = ciph_d[i - 1].out + packlen; 254 memcpy(ciph_d[i].out - 16, IVs, 16); 255 memcpy(ciph_d[i].iv, IVs, 16); 256 IVs += 16; 257 } 258 259# if defined(BSWAP8) 260 memcpy(blocks[0].c, key->md.data, 8); 261 seqnum = BSWAP8(blocks[0].q[0]); 262# endif 263 for (i = 0; i < x4; i++) { 264 unsigned int len = (i == (x4 - 1) ? last : frag); 265# if !defined(BSWAP8) 266 unsigned int carry, j; 267# endif 268 269 ctx->A[i] = key->md.h[0]; 270 ctx->B[i] = key->md.h[1]; 271 ctx->C[i] = key->md.h[2]; 272 ctx->D[i] = key->md.h[3]; 273 ctx->E[i] = key->md.h[4]; 274 ctx->F[i] = key->md.h[5]; 275 ctx->G[i] = key->md.h[6]; 276 ctx->H[i] = key->md.h[7]; 277 278 /* fix seqnum */ 279# if defined(BSWAP8) 280 blocks[i].q[0] = BSWAP8(seqnum + i); 281# else 282 for (carry = i, j = 8; j--;) { 283 blocks[i].c[j] = ((u8 *)key->md.data)[j] + carry; 284 carry = (blocks[i].c[j] - carry) >> (sizeof(carry) * 8 - 1); 285 } 286# endif 287 blocks[i].c[8] = ((u8 *)key->md.data)[8]; 288 blocks[i].c[9] = ((u8 *)key->md.data)[9]; 289 blocks[i].c[10] = ((u8 *)key->md.data)[10]; 290 /* fix length */ 291 blocks[i].c[11] = (u8)(len >> 8); 292 blocks[i].c[12] = (u8)(len); 293 294 memcpy(blocks[i].c + 13, hash_d[i].ptr, 64 - 13); 295 hash_d[i].ptr += 64 - 13; 296 hash_d[i].blocks = (len - (64 - 13)) / 64; 297 298 edges[i].ptr = blocks[i].c; 299 edges[i].blocks = 1; 300 } 301 302 /* hash 13-byte headers and first 64-13 bytes of inputs */ 303 sha256_multi_block(ctx, edges, n4x); 304 /* hash bulk inputs */ 305# define MAXCHUNKSIZE 2048 306# if MAXCHUNKSIZE%64 307# error "MAXCHUNKSIZE is not divisible by 64" 308# elif MAXCHUNKSIZE 309 /* 310 * goal is to minimize pressure on L1 cache by moving in shorter steps, 311 * so that hashed data is still in the cache by the time we encrypt it 312 */ 313 minblocks = ((frag <= last ? frag : last) - (64 - 13)) / 64; 314 if (minblocks > MAXCHUNKSIZE / 64) { 315 for (i = 0; i < x4; i++) { 316 edges[i].ptr = hash_d[i].ptr; 317 edges[i].blocks = MAXCHUNKSIZE / 64; 318 ciph_d[i].blocks = MAXCHUNKSIZE / 16; 319 } 320 do { 321 sha256_multi_block(ctx, edges, n4x); 322 aesni_multi_cbc_encrypt(ciph_d, &key->ks, n4x); 323 324 for (i = 0; i < x4; i++) { 325 edges[i].ptr = hash_d[i].ptr += MAXCHUNKSIZE; 326 hash_d[i].blocks -= MAXCHUNKSIZE / 64; 327 edges[i].blocks = MAXCHUNKSIZE / 64; 328 ciph_d[i].inp += MAXCHUNKSIZE; 329 ciph_d[i].out += MAXCHUNKSIZE; 330 ciph_d[i].blocks = MAXCHUNKSIZE / 16; 331 memcpy(ciph_d[i].iv, ciph_d[i].out - 16, 16); 332 } 333 processed += MAXCHUNKSIZE; 334 minblocks -= MAXCHUNKSIZE / 64; 335 } while (minblocks > MAXCHUNKSIZE / 64); 336 } 337# endif 338# undef MAXCHUNKSIZE 339 sha256_multi_block(ctx, hash_d, n4x); 340 341 memset(blocks, 0, sizeof(blocks)); 342 for (i = 0; i < x4; i++) { 343 unsigned int len = (i == (x4 - 1) ? last : frag), 344 off = hash_d[i].blocks * 64; 345 const unsigned char *ptr = hash_d[i].ptr + off; 346 347 off = (len - processed) - (64 - 13) - off; /* remainder actually */ 348 memcpy(blocks[i].c, ptr, off); 349 blocks[i].c[off] = 0x80; 350 len += 64 + 13; /* 64 is HMAC header */ 351 len *= 8; /* convert to bits */ 352 if (off < (64 - 8)) { 353# ifdef BSWAP4 354 blocks[i].d[15] = BSWAP4(len); 355# else 356 PUTU32(blocks[i].c + 60, len); 357# endif 358 edges[i].blocks = 1; 359 } else { 360# ifdef BSWAP4 361 blocks[i].d[31] = BSWAP4(len); 362# else 363 PUTU32(blocks[i].c + 124, len); 364# endif 365 edges[i].blocks = 2; 366 } 367 edges[i].ptr = blocks[i].c; 368 } 369 370 /* hash input tails and finalize */ 371 sha256_multi_block(ctx, edges, n4x); 372 373 memset(blocks, 0, sizeof(blocks)); 374 for (i = 0; i < x4; i++) { 375# ifdef BSWAP4 376 blocks[i].d[0] = BSWAP4(ctx->A[i]); 377 ctx->A[i] = key->tail.h[0]; 378 blocks[i].d[1] = BSWAP4(ctx->B[i]); 379 ctx->B[i] = key->tail.h[1]; 380 blocks[i].d[2] = BSWAP4(ctx->C[i]); 381 ctx->C[i] = key->tail.h[2]; 382 blocks[i].d[3] = BSWAP4(ctx->D[i]); 383 ctx->D[i] = key->tail.h[3]; 384 blocks[i].d[4] = BSWAP4(ctx->E[i]); 385 ctx->E[i] = key->tail.h[4]; 386 blocks[i].d[5] = BSWAP4(ctx->F[i]); 387 ctx->F[i] = key->tail.h[5]; 388 blocks[i].d[6] = BSWAP4(ctx->G[i]); 389 ctx->G[i] = key->tail.h[6]; 390 blocks[i].d[7] = BSWAP4(ctx->H[i]); 391 ctx->H[i] = key->tail.h[7]; 392 blocks[i].c[32] = 0x80; 393 blocks[i].d[15] = BSWAP4((64 + 32) * 8); 394# else 395 PUTU32(blocks[i].c + 0, ctx->A[i]); 396 ctx->A[i] = key->tail.h[0]; 397 PUTU32(blocks[i].c + 4, ctx->B[i]); 398 ctx->B[i] = key->tail.h[1]; 399 PUTU32(blocks[i].c + 8, ctx->C[i]); 400 ctx->C[i] = key->tail.h[2]; 401 PUTU32(blocks[i].c + 12, ctx->D[i]); 402 ctx->D[i] = key->tail.h[3]; 403 PUTU32(blocks[i].c + 16, ctx->E[i]); 404 ctx->E[i] = key->tail.h[4]; 405 PUTU32(blocks[i].c + 20, ctx->F[i]); 406 ctx->F[i] = key->tail.h[5]; 407 PUTU32(blocks[i].c + 24, ctx->G[i]); 408 ctx->G[i] = key->tail.h[6]; 409 PUTU32(blocks[i].c + 28, ctx->H[i]); 410 ctx->H[i] = key->tail.h[7]; 411 blocks[i].c[32] = 0x80; 412 PUTU32(blocks[i].c + 60, (64 + 32) * 8); 413# endif 414 edges[i].ptr = blocks[i].c; 415 edges[i].blocks = 1; 416 } 417 418 /* finalize MACs */ 419 sha256_multi_block(ctx, edges, n4x); 420 421 for (i = 0; i < x4; i++) { 422 unsigned int len = (i == (x4 - 1) ? last : frag), pad, j; 423 unsigned char *out0 = out; 424 425 memcpy(ciph_d[i].out, ciph_d[i].inp, len - processed); 426 ciph_d[i].inp = ciph_d[i].out; 427 428 out += 5 + 16 + len; 429 430 /* write MAC */ 431 PUTU32(out + 0, ctx->A[i]); 432 PUTU32(out + 4, ctx->B[i]); 433 PUTU32(out + 8, ctx->C[i]); 434 PUTU32(out + 12, ctx->D[i]); 435 PUTU32(out + 16, ctx->E[i]); 436 PUTU32(out + 20, ctx->F[i]); 437 PUTU32(out + 24, ctx->G[i]); 438 PUTU32(out + 28, ctx->H[i]); 439 out += 32; 440 len += 32; 441 442 /* pad */ 443 pad = 15 - len % 16; 444 for (j = 0; j <= pad; j++) 445 *(out++) = pad; 446 len += pad + 1; 447 448 ciph_d[i].blocks = (len - processed) / 16; 449 len += 16; /* account for explicit iv */ 450 451 /* arrange header */ 452 out0[0] = ((u8 *)key->md.data)[8]; 453 out0[1] = ((u8 *)key->md.data)[9]; 454 out0[2] = ((u8 *)key->md.data)[10]; 455 out0[3] = (u8)(len >> 8); 456 out0[4] = (u8)(len); 457 458 ret += len + 5; 459 inp += frag; 460 } 461 462 aesni_multi_cbc_encrypt(ciph_d, &key->ks, n4x); 463 464 OPENSSL_cleanse(blocks, sizeof(blocks)); 465 OPENSSL_cleanse(ctx, sizeof(*ctx)); 466 467 return ret; 468} 469# endif 470 471static int aesni_cbc_hmac_sha256_cipher(EVP_CIPHER_CTX *ctx, 472 unsigned char *out, 473 const unsigned char *in, size_t len) 474{ 475 EVP_AES_HMAC_SHA256 *key = data(ctx); 476 unsigned int l; 477 size_t plen = key->payload_length, iv = 0, /* explicit IV in TLS 1.1 and 478 * later */ 479 sha_off = 0; 480# if defined(STITCHED_CALL) 481 size_t aes_off = 0, blocks; 482 483 sha_off = SHA256_CBLOCK - key->md.num; 484# endif 485 486 key->payload_length = NO_PAYLOAD_LENGTH; 487 488 if (len % AES_BLOCK_SIZE) 489 return 0; 490 491 if (ctx->encrypt) { 492 if (plen == NO_PAYLOAD_LENGTH) 493 plen = len; 494 else if (len != 495 ((plen + SHA256_DIGEST_LENGTH + 496 AES_BLOCK_SIZE) & -AES_BLOCK_SIZE)) 497 return 0; 498 else if (key->aux.tls_ver >= TLS1_1_VERSION) 499 iv = AES_BLOCK_SIZE; 500 501# if defined(STITCHED_CALL) 502 /* 503 * Assembly stitch handles AVX-capable processors, but its 504 * performance is not optimal on AMD Jaguar, ~40% worse, for 505 * unknown reasons. Incidentally processor in question supports 506 * AVX, but not AMD-specific XOP extension, which can be used 507 * to identify it and avoid stitch invocation. So that after we 508 * establish that current CPU supports AVX, we even see if it's 509 * either even XOP-capable Bulldozer-based or GenuineIntel one. 510 */ 511 if (OPENSSL_ia32cap_P[1] & (1 << (60 - 32)) && /* AVX? */ 512 ((OPENSSL_ia32cap_P[1] & (1 << (43 - 32))) /* XOP? */ 513 | (OPENSSL_ia32cap_P[0] & (1<<30))) && /* "Intel CPU"? */ 514 plen > (sha_off + iv) && 515 (blocks = (plen - (sha_off + iv)) / SHA256_CBLOCK)) { 516 SHA256_Update(&key->md, in + iv, sha_off); 517 518 (void)aesni_cbc_sha256_enc(in, out, blocks, &key->ks, 519 ctx->iv, &key->md, in + iv + sha_off); 520 blocks *= SHA256_CBLOCK; 521 aes_off += blocks; 522 sha_off += blocks; 523 key->md.Nh += blocks >> 29; 524 key->md.Nl += blocks <<= 3; 525 if (key->md.Nl < (unsigned int)blocks) 526 key->md.Nh++; 527 } else { 528 sha_off = 0; 529 } 530# endif 531 sha_off += iv; 532 SHA256_Update(&key->md, in + sha_off, plen - sha_off); 533 534 if (plen != len) { /* "TLS" mode of operation */ 535 if (in != out) 536 memcpy(out + aes_off, in + aes_off, plen - aes_off); 537 538 /* calculate HMAC and append it to payload */ 539 SHA256_Final(out + plen, &key->md); 540 key->md = key->tail; 541 SHA256_Update(&key->md, out + plen, SHA256_DIGEST_LENGTH); 542 SHA256_Final(out + plen, &key->md); 543 544 /* pad the payload|hmac */ 545 plen += SHA256_DIGEST_LENGTH; 546 for (l = len - plen - 1; plen < len; plen++) 547 out[plen] = l; 548 /* encrypt HMAC|padding at once */ 549 aesni_cbc_encrypt(out + aes_off, out + aes_off, len - aes_off, 550 &key->ks, ctx->iv, 1); 551 } else { 552 aesni_cbc_encrypt(in + aes_off, out + aes_off, len - aes_off, 553 &key->ks, ctx->iv, 1); 554 } 555 } else { 556 union { 557 unsigned int u[SHA256_DIGEST_LENGTH / sizeof(unsigned int)]; 558 unsigned char c[64 + SHA256_DIGEST_LENGTH]; 559 } mac, *pmac; 560 561 /* arrange cache line alignment */ 562 pmac = (void *)(((size_t)mac.c + 63) & ((size_t)0 - 64)); 563 564 /* decrypt HMAC|padding at once */ 565 aesni_cbc_encrypt(in, out, len, &key->ks, ctx->iv, 0); 566 567 if (plen != NO_PAYLOAD_LENGTH) { /* "TLS" mode of operation */ 568 size_t inp_len, mask, j, i; 569 unsigned int res, maxpad, pad, bitlen; 570 int ret = 1; 571 union { 572 unsigned int u[SHA_LBLOCK]; 573 unsigned char c[SHA256_CBLOCK]; 574 } *data = (void *)key->md.data; 575 576 if ((key->aux.tls_aad[plen - 4] << 8 | key->aux.tls_aad[plen - 3]) 577 >= TLS1_1_VERSION) 578 iv = AES_BLOCK_SIZE; 579 580 if (len < (iv + SHA256_DIGEST_LENGTH + 1)) 581 return 0; 582 583 /* omit explicit iv */ 584 out += iv; 585 len -= iv; 586 587 /* figure out payload length */ 588 pad = out[len - 1]; 589 maxpad = len - (SHA256_DIGEST_LENGTH + 1); 590 maxpad |= (255 - maxpad) >> (sizeof(maxpad) * 8 - 8); 591 maxpad &= 255; 592 593 ret &= constant_time_ge(maxpad, pad); 594 595 inp_len = len - (SHA256_DIGEST_LENGTH + pad + 1); 596 mask = (0 - ((inp_len - len) >> (sizeof(inp_len) * 8 - 1))); 597 inp_len &= mask; 598 ret &= (int)mask; 599 600 key->aux.tls_aad[plen - 2] = inp_len >> 8; 601 key->aux.tls_aad[plen - 1] = inp_len; 602 603 /* calculate HMAC */ 604 key->md = key->head; 605 SHA256_Update(&key->md, key->aux.tls_aad, plen); 606 607# if 1 608 len -= SHA256_DIGEST_LENGTH; /* amend mac */ 609 if (len >= (256 + SHA256_CBLOCK)) { 610 j = (len - (256 + SHA256_CBLOCK)) & (0 - SHA256_CBLOCK); 611 j += SHA256_CBLOCK - key->md.num; 612 SHA256_Update(&key->md, out, j); 613 out += j; 614 len -= j; 615 inp_len -= j; 616 } 617 618 /* but pretend as if we hashed padded payload */ 619 bitlen = key->md.Nl + (inp_len << 3); /* at most 18 bits */ 620# ifdef BSWAP4 621 bitlen = BSWAP4(bitlen); 622# else 623 mac.c[0] = 0; 624 mac.c[1] = (unsigned char)(bitlen >> 16); 625 mac.c[2] = (unsigned char)(bitlen >> 8); 626 mac.c[3] = (unsigned char)bitlen; 627 bitlen = mac.u[0]; 628# endif 629 630 pmac->u[0] = 0; 631 pmac->u[1] = 0; 632 pmac->u[2] = 0; 633 pmac->u[3] = 0; 634 pmac->u[4] = 0; 635 pmac->u[5] = 0; 636 pmac->u[6] = 0; 637 pmac->u[7] = 0; 638 639 for (res = key->md.num, j = 0; j < len; j++) { 640 size_t c = out[j]; 641 mask = (j - inp_len) >> (sizeof(j) * 8 - 8); 642 c &= mask; 643 c |= 0x80 & ~mask & ~((inp_len - j) >> (sizeof(j) * 8 - 8)); 644 data->c[res++] = (unsigned char)c; 645 646 if (res != SHA256_CBLOCK) 647 continue; 648 649 /* j is not incremented yet */ 650 mask = 0 - ((inp_len + 7 - j) >> (sizeof(j) * 8 - 1)); 651 data->u[SHA_LBLOCK - 1] |= bitlen & mask; 652 sha256_block_data_order(&key->md, data, 1); 653 mask &= 0 - ((j - inp_len - 72) >> (sizeof(j) * 8 - 1)); 654 pmac->u[0] |= key->md.h[0] & mask; 655 pmac->u[1] |= key->md.h[1] & mask; 656 pmac->u[2] |= key->md.h[2] & mask; 657 pmac->u[3] |= key->md.h[3] & mask; 658 pmac->u[4] |= key->md.h[4] & mask; 659 pmac->u[5] |= key->md.h[5] & mask; 660 pmac->u[6] |= key->md.h[6] & mask; 661 pmac->u[7] |= key->md.h[7] & mask; 662 res = 0; 663 } 664 665 for (i = res; i < SHA256_CBLOCK; i++, j++) 666 data->c[i] = 0; 667 668 if (res > SHA256_CBLOCK - 8) { 669 mask = 0 - ((inp_len + 8 - j) >> (sizeof(j) * 8 - 1)); 670 data->u[SHA_LBLOCK - 1] |= bitlen & mask; 671 sha256_block_data_order(&key->md, data, 1); 672 mask &= 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1)); 673 pmac->u[0] |= key->md.h[0] & mask; 674 pmac->u[1] |= key->md.h[1] & mask; 675 pmac->u[2] |= key->md.h[2] & mask; 676 pmac->u[3] |= key->md.h[3] & mask; 677 pmac->u[4] |= key->md.h[4] & mask; 678 pmac->u[5] |= key->md.h[5] & mask; 679 pmac->u[6] |= key->md.h[6] & mask; 680 pmac->u[7] |= key->md.h[7] & mask; 681 682 memset(data, 0, SHA256_CBLOCK); 683 j += 64; 684 } 685 data->u[SHA_LBLOCK - 1] = bitlen; 686 sha256_block_data_order(&key->md, data, 1); 687 mask = 0 - ((j - inp_len - 73) >> (sizeof(j) * 8 - 1)); 688 pmac->u[0] |= key->md.h[0] & mask; 689 pmac->u[1] |= key->md.h[1] & mask; 690 pmac->u[2] |= key->md.h[2] & mask; 691 pmac->u[3] |= key->md.h[3] & mask; 692 pmac->u[4] |= key->md.h[4] & mask; 693 pmac->u[5] |= key->md.h[5] & mask; 694 pmac->u[6] |= key->md.h[6] & mask; 695 pmac->u[7] |= key->md.h[7] & mask; 696 697# ifdef BSWAP4 698 pmac->u[0] = BSWAP4(pmac->u[0]); 699 pmac->u[1] = BSWAP4(pmac->u[1]); 700 pmac->u[2] = BSWAP4(pmac->u[2]); 701 pmac->u[3] = BSWAP4(pmac->u[3]); 702 pmac->u[4] = BSWAP4(pmac->u[4]); 703 pmac->u[5] = BSWAP4(pmac->u[5]); 704 pmac->u[6] = BSWAP4(pmac->u[6]); 705 pmac->u[7] = BSWAP4(pmac->u[7]); 706# else 707 for (i = 0; i < 8; i++) { 708 res = pmac->u[i]; 709 pmac->c[4 * i + 0] = (unsigned char)(res >> 24); 710 pmac->c[4 * i + 1] = (unsigned char)(res >> 16); 711 pmac->c[4 * i + 2] = (unsigned char)(res >> 8); 712 pmac->c[4 * i + 3] = (unsigned char)res; 713 } 714# endif 715 len += SHA256_DIGEST_LENGTH; 716# else 717 SHA256_Update(&key->md, out, inp_len); 718 res = key->md.num; 719 SHA256_Final(pmac->c, &key->md); 720 721 { 722 unsigned int inp_blocks, pad_blocks; 723 724 /* but pretend as if we hashed padded payload */ 725 inp_blocks = 726 1 + ((SHA256_CBLOCK - 9 - res) >> (sizeof(res) * 8 - 1)); 727 res += (unsigned int)(len - inp_len); 728 pad_blocks = res / SHA256_CBLOCK; 729 res %= SHA256_CBLOCK; 730 pad_blocks += 731 1 + ((SHA256_CBLOCK - 9 - res) >> (sizeof(res) * 8 - 1)); 732 for (; inp_blocks < pad_blocks; inp_blocks++) 733 sha1_block_data_order(&key->md, data, 1); 734 } 735# endif 736 key->md = key->tail; 737 SHA256_Update(&key->md, pmac->c, SHA256_DIGEST_LENGTH); 738 SHA256_Final(pmac->c, &key->md); 739 740 /* verify HMAC */ 741 out += inp_len; 742 len -= inp_len; 743# if 1 744 { 745 unsigned char *p = 746 out + len - 1 - maxpad - SHA256_DIGEST_LENGTH; 747 size_t off = out - p; 748 unsigned int c, cmask; 749 750 maxpad += SHA256_DIGEST_LENGTH; 751 for (res = 0, i = 0, j = 0; j < maxpad; j++) { 752 c = p[j]; 753 cmask = 754 ((int)(j - off - SHA256_DIGEST_LENGTH)) >> 755 (sizeof(int) * 8 - 1); 756 res |= (c ^ pad) & ~cmask; /* ... and padding */ 757 cmask &= ((int)(off - 1 - j)) >> (sizeof(int) * 8 - 1); 758 res |= (c ^ pmac->c[i]) & cmask; 759 i += 1 & cmask; 760 } 761 maxpad -= SHA256_DIGEST_LENGTH; 762 763 res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1)); 764 ret &= (int)~res; 765 } 766# else 767 for (res = 0, i = 0; i < SHA256_DIGEST_LENGTH; i++) 768 res |= out[i] ^ pmac->c[i]; 769 res = 0 - ((0 - res) >> (sizeof(res) * 8 - 1)); 770 ret &= (int)~res; 771 772 /* verify padding */ 773 pad = (pad & ~res) | (maxpad & res); 774 out = out + len - 1 - pad; 775 for (res = 0, i = 0; i < pad; i++) 776 res |= out[i] ^ pad; 777 778 res = (0 - res) >> (sizeof(res) * 8 - 1); 779 ret &= (int)~res; 780# endif 781 return ret; 782 } else { 783 SHA256_Update(&key->md, out, len); 784 } 785 } 786 787 return 1; 788} 789 790static int aesni_cbc_hmac_sha256_ctrl(EVP_CIPHER_CTX *ctx, int type, int arg, 791 void *ptr) 792{ 793 EVP_AES_HMAC_SHA256 *key = data(ctx); 794 795 switch (type) { 796 case EVP_CTRL_AEAD_SET_MAC_KEY: 797 { 798 unsigned int i; 799 unsigned char hmac_key[64]; 800 801 memset(hmac_key, 0, sizeof(hmac_key)); 802 803 if (arg > (int)sizeof(hmac_key)) { 804 SHA256_Init(&key->head); 805 SHA256_Update(&key->head, ptr, arg); 806 SHA256_Final(hmac_key, &key->head); 807 } else { 808 memcpy(hmac_key, ptr, arg); 809 } 810 811 for (i = 0; i < sizeof(hmac_key); i++) 812 hmac_key[i] ^= 0x36; /* ipad */ 813 SHA256_Init(&key->head); 814 SHA256_Update(&key->head, hmac_key, sizeof(hmac_key)); 815 816 for (i = 0; i < sizeof(hmac_key); i++) 817 hmac_key[i] ^= 0x36 ^ 0x5c; /* opad */ 818 SHA256_Init(&key->tail); 819 SHA256_Update(&key->tail, hmac_key, sizeof(hmac_key)); 820 821 OPENSSL_cleanse(hmac_key, sizeof(hmac_key)); 822 823 return 1; 824 } 825 case EVP_CTRL_AEAD_TLS1_AAD: 826 { 827 unsigned char *p = ptr; 828 unsigned int len; 829 830 if (arg != EVP_AEAD_TLS1_AAD_LEN) 831 return -1; 832 833 len = p[arg - 2] << 8 | p[arg - 1]; 834 835 if (ctx->encrypt) { 836 key->payload_length = len; 837 if ((key->aux.tls_ver = 838 p[arg - 4] << 8 | p[arg - 3]) >= TLS1_1_VERSION) { 839 if (len < AES_BLOCK_SIZE) 840 return 0; 841 len -= AES_BLOCK_SIZE; 842 p[arg - 2] = len >> 8; 843 p[arg - 1] = len; 844 } 845 key->md = key->head; 846 SHA256_Update(&key->md, p, arg); 847 848 return (int)(((len + SHA256_DIGEST_LENGTH + 849 AES_BLOCK_SIZE) & -AES_BLOCK_SIZE) 850 - len); 851 } else { 852 memcpy(key->aux.tls_aad, ptr, arg); 853 key->payload_length = arg; 854 855 return SHA256_DIGEST_LENGTH; 856 } 857 } 858# if !defined(OPENSSL_NO_MULTIBLOCK) && EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK 859 case EVP_CTRL_TLS1_1_MULTIBLOCK_MAX_BUFSIZE: 860 return (int)(5 + 16 + ((arg + 32 + 16) & -16)); 861 case EVP_CTRL_TLS1_1_MULTIBLOCK_AAD: 862 { 863 EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param = 864 (EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *) ptr; 865 unsigned int n4x = 1, x4; 866 unsigned int frag, last, packlen, inp_len; 867 868 if (arg < (int)sizeof(EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM)) 869 return -1; 870 871 inp_len = param->inp[11] << 8 | param->inp[12]; 872 873 if (ctx->encrypt) { 874 if ((param->inp[9] << 8 | param->inp[10]) < TLS1_1_VERSION) 875 return -1; 876 877 if (inp_len) { 878 if (inp_len < 4096) 879 return 0; /* too short */ 880 881 if (inp_len >= 8192 && OPENSSL_ia32cap_P[2] & (1 << 5)) 882 n4x = 2; /* AVX2 */ 883 } else if ((n4x = param->interleave / 4) && n4x <= 2) 884 inp_len = param->len; 885 else 886 return -1; 887 888 key->md = key->head; 889 SHA256_Update(&key->md, param->inp, 13); 890 891 x4 = 4 * n4x; 892 n4x += 1; 893 894 frag = inp_len >> n4x; 895 last = inp_len + frag - (frag << n4x); 896 if (last > frag && ((last + 13 + 9) % 64 < (x4 - 1))) { 897 frag++; 898 last -= x4 - 1; 899 } 900 901 packlen = 5 + 16 + ((frag + 32 + 16) & -16); 902 packlen = (packlen << n4x) - packlen; 903 packlen += 5 + 16 + ((last + 32 + 16) & -16); 904 905 param->interleave = x4; 906 907 return (int)packlen; 908 } else 909 return -1; /* not yet */ 910 } 911 case EVP_CTRL_TLS1_1_MULTIBLOCK_ENCRYPT: 912 { 913 EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *param = 914 (EVP_CTRL_TLS1_1_MULTIBLOCK_PARAM *) ptr; 915 916 return (int)tls1_1_multi_block_encrypt(key, param->out, 917 param->inp, param->len, 918 param->interleave / 4); 919 } 920 case EVP_CTRL_TLS1_1_MULTIBLOCK_DECRYPT: 921# endif 922 default: 923 return -1; 924 } 925} 926 927static EVP_CIPHER aesni_128_cbc_hmac_sha256_cipher = { 928# ifdef NID_aes_128_cbc_hmac_sha256 929 NID_aes_128_cbc_hmac_sha256, 930# else 931 NID_undef, 932# endif 933 16, 16, 16, 934 EVP_CIPH_CBC_MODE | EVP_CIPH_FLAG_DEFAULT_ASN1 | 935 EVP_CIPH_FLAG_AEAD_CIPHER | EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK, 936 aesni_cbc_hmac_sha256_init_key, 937 aesni_cbc_hmac_sha256_cipher, 938 NULL, 939 sizeof(EVP_AES_HMAC_SHA256), 940 EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_set_asn1_iv, 941 EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_get_asn1_iv, 942 aesni_cbc_hmac_sha256_ctrl, 943 NULL 944}; 945 946static EVP_CIPHER aesni_256_cbc_hmac_sha256_cipher = { 947# ifdef NID_aes_256_cbc_hmac_sha256 948 NID_aes_256_cbc_hmac_sha256, 949# else 950 NID_undef, 951# endif 952 16, 32, 16, 953 EVP_CIPH_CBC_MODE | EVP_CIPH_FLAG_DEFAULT_ASN1 | 954 EVP_CIPH_FLAG_AEAD_CIPHER | EVP_CIPH_FLAG_TLS1_1_MULTIBLOCK, 955 aesni_cbc_hmac_sha256_init_key, 956 aesni_cbc_hmac_sha256_cipher, 957 NULL, 958 sizeof(EVP_AES_HMAC_SHA256), 959 EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_set_asn1_iv, 960 EVP_CIPH_FLAG_DEFAULT_ASN1 ? NULL : EVP_CIPHER_get_asn1_iv, 961 aesni_cbc_hmac_sha256_ctrl, 962 NULL 963}; 964 965const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void) 966{ 967 return ((OPENSSL_ia32cap_P[1] & AESNI_CAPABLE) && 968 aesni_cbc_sha256_enc(NULL, NULL, 0, NULL, NULL, NULL, NULL) ? 969 &aesni_128_cbc_hmac_sha256_cipher : NULL); 970} 971 972const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void) 973{ 974 return ((OPENSSL_ia32cap_P[1] & AESNI_CAPABLE) && 975 aesni_cbc_sha256_enc(NULL, NULL, 0, NULL, NULL, NULL, NULL) ? 976 &aesni_256_cbc_hmac_sha256_cipher : NULL); 977} 978# else 979const EVP_CIPHER *EVP_aes_128_cbc_hmac_sha256(void) 980{ 981 return NULL; 982} 983 984const EVP_CIPHER *EVP_aes_256_cbc_hmac_sha256(void) 985{ 986 return NULL; 987} 988# endif 989#endif 990