| 1 |
/* ==================================================================== |
| 2 |
* Copyright (c) 2008 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 |
* [email protected]. |
| 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 |
#include <openssl/base.h> |
| 50 |
|
| 51 |
#include <assert.h> |
| 52 |
#include <string.h> |
| 53 |
|
| 54 |
#include <openssl/mem.h> |
| 55 |
#include <openssl/cpu.h> |
| 56 |
|
| 57 |
#include "internal.h" |
| 58 |
#include "../internal.h" |
| 59 |
|
| 60 |
|
| 61 |
#if !defined(OPENSSL_NO_ASM) && \ |
| 62 |
(defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \ |
| 63 |
defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64)) |
| 64 |
#define GHASH_ASM |
| 65 |
#endif |
| 66 |
|
| 67 |
#if defined(BSWAP4) && STRICT_ALIGNMENT == 1 |
| 68 |
/* redefine, because alignment is ensured */ |
| 69 |
#undef GETU32 |
| 70 |
#define GETU32(p) BSWAP4(*(const uint32_t *)(p)) |
| 71 |
#undef PUTU32 |
| 72 |
#define PUTU32(p, v) *(uint32_t *)(p) = BSWAP4(v) |
| 73 |
#endif |
| 74 |
|
| 75 |
#define PACK(s) ((size_t)(s) << (sizeof(size_t) * 8 - 16)) |
| 76 |
#define REDUCE1BIT(V) \ |
| 77 |
do { \ |
| 78 |
if (sizeof(size_t) == 8) { \ |
| 79 |
uint64_t T = UINT64_C(0xe100000000000000) & (0 - (V.lo & 1)); \ |
| 80 |
V.lo = (V.hi << 63) | (V.lo >> 1); \ |
| 81 |
V.hi = (V.hi >> 1) ^ T; \ |
| 82 |
} else { \ |
| 83 |
uint32_t T = 0xe1000000U & (0 - (uint32_t)(V.lo & 1)); \ |
| 84 |
V.lo = (V.hi << 63) | (V.lo >> 1); \ |
| 85 |
V.hi = (V.hi >> 1) ^ ((uint64_t)T << 32); \ |
| 86 |
} \ |
| 87 |
} while (0) |
| 88 |
|
| 89 |
// kSizeTWithoutLower4Bits is a mask that can be used to zero the lower four |
| 90 |
// bits of a |size_t|. |
| 91 |
static const size_t kSizeTWithoutLower4Bits = (size_t) -16; |
| 92 |
|
| 93 |
static void gcm_init_4bit(u128 Htable[16], uint64_t H[2]) { |
| 94 |
u128 V; |
| 95 |
|
| 96 |
Htable[0].hi = 0; |
| 97 |
Htable[0].lo = 0; |
| 98 |
V.hi = H[0]; |
| 99 |
V.lo = H[1]; |
| 100 |
|
| 101 |
Htable[8] = V; |
| 102 |
REDUCE1BIT(V); |
| 103 |
Htable[4] = V; |
| 104 |
REDUCE1BIT(V); |
| 105 |
Htable[2] = V; |
| 106 |
REDUCE1BIT(V); |
| 107 |
Htable[1] = V; |
| 108 |
Htable[3].hi = V.hi ^ Htable[2].hi, Htable[3].lo = V.lo ^ Htable[2].lo; |
| 109 |
V = Htable[4]; |
| 110 |
Htable[5].hi = V.hi ^ Htable[1].hi, Htable[5].lo = V.lo ^ Htable[1].lo; |
| 111 |
Htable[6].hi = V.hi ^ Htable[2].hi, Htable[6].lo = V.lo ^ Htable[2].lo; |
| 112 |
Htable[7].hi = V.hi ^ Htable[3].hi, Htable[7].lo = V.lo ^ Htable[3].lo; |
| 113 |
V = Htable[8]; |
| 114 |
Htable[9].hi = V.hi ^ Htable[1].hi, Htable[9].lo = V.lo ^ Htable[1].lo; |
| 115 |
Htable[10].hi = V.hi ^ Htable[2].hi, Htable[10].lo = V.lo ^ Htable[2].lo; |
| 116 |
Htable[11].hi = V.hi ^ Htable[3].hi, Htable[11].lo = V.lo ^ Htable[3].lo; |
| 117 |
Htable[12].hi = V.hi ^ Htable[4].hi, Htable[12].lo = V.lo ^ Htable[4].lo; |
| 118 |
Htable[13].hi = V.hi ^ Htable[5].hi, Htable[13].lo = V.lo ^ Htable[5].lo; |
| 119 |
Htable[14].hi = V.hi ^ Htable[6].hi, Htable[14].lo = V.lo ^ Htable[6].lo; |
| 120 |
Htable[15].hi = V.hi ^ Htable[7].hi, Htable[15].lo = V.lo ^ Htable[7].lo; |
| 121 |
|
| 122 |
#if defined(GHASH_ASM) && defined(OPENSSL_ARM) |
| 123 |
/* ARM assembler expects specific dword order in Htable. */ |
| 124 |
{ |
| 125 |
int j; |
| 126 |
const union { |
| 127 |
long one; |
| 128 |
char little; |
| 129 |
} is_endian = {1}; |
| 130 |
|
| 131 |
if (is_endian.little) { |
| 132 |
for (j = 0; j < 16; ++j) { |
| 133 |
V = Htable[j]; |
| 134 |
Htable[j].hi = V.lo; |
| 135 |
Htable[j].lo = V.hi; |
| 136 |
} |
| 137 |
} else { |
| 138 |
for (j = 0; j < 16; ++j) { |
| 139 |
V = Htable[j]; |
| 140 |
Htable[j].hi = V.lo << 32 | V.lo >> 32; |
| 141 |
Htable[j].lo = V.hi << 32 | V.hi >> 32; |
| 142 |
} |
| 143 |
} |
| 144 |
} |
| 145 |
#endif |
| 146 |
} |
| 147 |
|
| 148 |
#if !defined(GHASH_ASM) || defined(OPENSSL_AARCH64) |
| 149 |
static const size_t rem_4bit[16] = { |
| 150 |
PACK(0x0000), PACK(0x1C20), PACK(0x3840), PACK(0x2460), |
| 151 |
PACK(0x7080), PACK(0x6CA0), PACK(0x48C0), PACK(0x54E0), |
| 152 |
PACK(0xE100), PACK(0xFD20), PACK(0xD940), PACK(0xC560), |
| 153 |
PACK(0x9180), PACK(0x8DA0), PACK(0xA9C0), PACK(0xB5E0)}; |
| 154 |
|
| 155 |
static void gcm_gmult_4bit(uint64_t Xi[2], const u128 Htable[16]) { |
| 156 |
u128 Z; |
| 157 |
int cnt = 15; |
| 158 |
size_t rem, nlo, nhi; |
| 159 |
const union { |
| 160 |
long one; |
| 161 |
char little; |
| 162 |
} is_endian = {1}; |
| 163 |
|
| 164 |
nlo = ((const uint8_t *)Xi)[15]; |
| 165 |
nhi = nlo >> 4; |
| 166 |
nlo &= 0xf; |
| 167 |
|
| 168 |
Z.hi = Htable[nlo].hi; |
| 169 |
Z.lo = Htable[nlo].lo; |
| 170 |
|
| 171 |
while (1) { |
| 172 |
rem = (size_t)Z.lo & 0xf; |
| 173 |
Z.lo = (Z.hi << 60) | (Z.lo >> 4); |
| 174 |
Z.hi = (Z.hi >> 4); |
| 175 |
if (sizeof(size_t) == 8) { |
| 176 |
Z.hi ^= rem_4bit[rem]; |
| 177 |
} else { |
| 178 |
Z.hi ^= (uint64_t)rem_4bit[rem] << 32; |
| 179 |
} |
| 180 |
|
| 181 |
Z.hi ^= Htable[nhi].hi; |
| 182 |
Z.lo ^= Htable[nhi].lo; |
| 183 |
|
| 184 |
if (--cnt < 0) { |
| 185 |
break; |
| 186 |
} |
| 187 |
|
| 188 |
nlo = ((const uint8_t *)Xi)[cnt]; |
| 189 |
nhi = nlo >> 4; |
| 190 |
nlo &= 0xf; |
| 191 |
|
| 192 |
rem = (size_t)Z.lo & 0xf; |
| 193 |
Z.lo = (Z.hi << 60) | (Z.lo >> 4); |
| 194 |
Z.hi = (Z.hi >> 4); |
| 195 |
if (sizeof(size_t) == 8) { |
| 196 |
Z.hi ^= rem_4bit[rem]; |
| 197 |
} else { |
| 198 |
Z.hi ^= (uint64_t)rem_4bit[rem] << 32; |
| 199 |
} |
| 200 |
|
| 201 |
Z.hi ^= Htable[nlo].hi; |
| 202 |
Z.lo ^= Htable[nlo].lo; |
| 203 |
} |
| 204 |
|
| 205 |
if (is_endian.little) { |
| 206 |
#ifdef BSWAP8 |
| 207 |
Xi[0] = BSWAP8(Z.hi); |
| 208 |
Xi[1] = BSWAP8(Z.lo); |
| 209 |
#else |
| 210 |
uint8_t *p = (uint8_t *)Xi; |
| 211 |
uint32_t v; |
| 212 |
v = (uint32_t)(Z.hi >> 32); |
| 213 |
PUTU32(p, v); |
| 214 |
v = (uint32_t)(Z.hi); |
| 215 |
PUTU32(p + 4, v); |
| 216 |
v = (uint32_t)(Z.lo >> 32); |
| 217 |
PUTU32(p + 8, v); |
| 218 |
v = (uint32_t)(Z.lo); |
| 219 |
PUTU32(p + 12, v); |
| 220 |
#endif |
| 221 |
} else { |
| 222 |
Xi[0] = Z.hi; |
| 223 |
Xi[1] = Z.lo; |
| 224 |
} |
| 225 |
} |
| 226 |
|
| 227 |
/* Streamed gcm_mult_4bit, see CRYPTO_gcm128_[en|de]crypt for |
| 228 |
* details... Compiler-generated code doesn't seem to give any |
| 229 |
* performance improvement, at least not on x86[_64]. It's here |
| 230 |
* mostly as reference and a placeholder for possible future |
| 231 |
* non-trivial optimization[s]... */ |
| 232 |
static void gcm_ghash_4bit(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 233 |
size_t len) { |
| 234 |
u128 Z; |
| 235 |
int cnt; |
| 236 |
size_t rem, nlo, nhi; |
| 237 |
const union { |
| 238 |
long one; |
| 239 |
char little; |
| 240 |
} is_endian = {1}; |
| 241 |
|
| 242 |
do { |
| 243 |
cnt = 15; |
| 244 |
nlo = ((const uint8_t *)Xi)[15]; |
| 245 |
nlo ^= inp[15]; |
| 246 |
nhi = nlo >> 4; |
| 247 |
nlo &= 0xf; |
| 248 |
|
| 249 |
Z.hi = Htable[nlo].hi; |
| 250 |
Z.lo = Htable[nlo].lo; |
| 251 |
|
| 252 |
while (1) { |
| 253 |
rem = (size_t)Z.lo & 0xf; |
| 254 |
Z.lo = (Z.hi << 60) | (Z.lo >> 4); |
| 255 |
Z.hi = (Z.hi >> 4); |
| 256 |
if (sizeof(size_t) == 8) { |
| 257 |
Z.hi ^= rem_4bit[rem]; |
| 258 |
} else { |
| 259 |
Z.hi ^= (uint64_t)rem_4bit[rem] << 32; |
| 260 |
} |
| 261 |
|
| 262 |
Z.hi ^= Htable[nhi].hi; |
| 263 |
Z.lo ^= Htable[nhi].lo; |
| 264 |
|
| 265 |
if (--cnt < 0) { |
| 266 |
break; |
| 267 |
} |
| 268 |
|
| 269 |
nlo = ((const uint8_t *)Xi)[cnt]; |
| 270 |
nlo ^= inp[cnt]; |
| 271 |
nhi = nlo >> 4; |
| 272 |
nlo &= 0xf; |
| 273 |
|
| 274 |
rem = (size_t)Z.lo & 0xf; |
| 275 |
Z.lo = (Z.hi << 60) | (Z.lo >> 4); |
| 276 |
Z.hi = (Z.hi >> 4); |
| 277 |
if (sizeof(size_t) == 8) { |
| 278 |
Z.hi ^= rem_4bit[rem]; |
| 279 |
} else { |
| 280 |
Z.hi ^= (uint64_t)rem_4bit[rem] << 32; |
| 281 |
} |
| 282 |
|
| 283 |
Z.hi ^= Htable[nlo].hi; |
| 284 |
Z.lo ^= Htable[nlo].lo; |
| 285 |
} |
| 286 |
|
| 287 |
if (is_endian.little) { |
| 288 |
#ifdef BSWAP8 |
| 289 |
Xi[0] = BSWAP8(Z.hi); |
| 290 |
Xi[1] = BSWAP8(Z.lo); |
| 291 |
#else |
| 292 |
uint8_t *p = (uint8_t *)Xi; |
| 293 |
uint32_t v; |
| 294 |
v = (uint32_t)(Z.hi >> 32); |
| 295 |
PUTU32(p, v); |
| 296 |
v = (uint32_t)(Z.hi); |
| 297 |
PUTU32(p + 4, v); |
| 298 |
v = (uint32_t)(Z.lo >> 32); |
| 299 |
PUTU32(p + 8, v); |
| 300 |
v = (uint32_t)(Z.lo); |
| 301 |
PUTU32(p + 12, v); |
| 302 |
#endif |
| 303 |
} else { |
| 304 |
Xi[0] = Z.hi; |
| 305 |
Xi[1] = Z.lo; |
| 306 |
} |
| 307 |
} while (inp += 16, len -= 16); |
| 308 |
} |
| 309 |
#else /* GHASH_ASM */ |
| 310 |
void gcm_gmult_4bit(uint64_t Xi[2], const u128 Htable[16]); |
| 311 |
void gcm_ghash_4bit(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 312 |
size_t len); |
| 313 |
#endif |
| 314 |
|
| 315 |
#define GCM_MUL(ctx, Xi) gcm_gmult_4bit(ctx->Xi.u, ctx->Htable) |
| 316 |
#if defined(GHASH_ASM) |
| 317 |
#define GHASH(ctx, in, len) gcm_ghash_4bit((ctx)->Xi.u, (ctx)->Htable, in, len) |
| 318 |
/* GHASH_CHUNK is "stride parameter" missioned to mitigate cache |
| 319 |
* trashing effect. In other words idea is to hash data while it's |
| 320 |
* still in L1 cache after encryption pass... */ |
| 321 |
#define GHASH_CHUNK (3 * 1024) |
| 322 |
#endif |
| 323 |
|
| 324 |
|
| 325 |
#if defined(GHASH_ASM) |
| 326 |
#if defined(OPENSSL_X86) || defined(OPENSSL_X86_64) |
| 327 |
#define GHASH_ASM_X86_OR_64 |
| 328 |
#define GCM_FUNCREF_4BIT |
| 329 |
void gcm_init_clmul(u128 Htable[16], const uint64_t Xi[2]); |
| 330 |
void gcm_gmult_clmul(uint64_t Xi[2], const u128 Htable[16]); |
| 331 |
void gcm_ghash_clmul(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 332 |
size_t len); |
| 333 |
|
| 334 |
#if defined(OPENSSL_X86) |
| 335 |
#define gcm_init_avx gcm_init_clmul |
| 336 |
#define gcm_gmult_avx gcm_gmult_clmul |
| 337 |
#define gcm_ghash_avx gcm_ghash_clmul |
| 338 |
#else |
| 339 |
void gcm_init_avx(u128 Htable[16], const uint64_t Xi[2]); |
| 340 |
void gcm_gmult_avx(uint64_t Xi[2], const u128 Htable[16]); |
| 341 |
void gcm_ghash_avx(uint64_t Xi[2], const u128 Htable[16], const uint8_t *in, |
| 342 |
size_t len); |
| 343 |
#define AESNI_GCM |
| 344 |
static int aesni_gcm_enabled(GCM128_CONTEXT *ctx, ctr128_f stream) { |
| 345 |
return stream == aesni_ctr32_encrypt_blocks && |
| 346 |
ctx->ghash == gcm_ghash_avx; |
| 347 |
} |
| 348 |
|
| 349 |
size_t aesni_gcm_encrypt(const uint8_t *in, uint8_t *out, size_t len, |
| 350 |
const void *key, uint8_t ivec[16], uint64_t *Xi); |
| 351 |
size_t aesni_gcm_decrypt(const uint8_t *in, uint8_t *out, size_t len, |
| 352 |
const void *key, uint8_t ivec[16], uint64_t *Xi); |
| 353 |
#endif |
| 354 |
|
| 355 |
#if defined(OPENSSL_X86) |
| 356 |
#define GHASH_ASM_X86 |
| 357 |
void gcm_gmult_4bit_mmx(uint64_t Xi[2], const u128 Htable[16]); |
| 358 |
void gcm_ghash_4bit_mmx(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 359 |
size_t len); |
| 360 |
|
| 361 |
void gcm_gmult_4bit_x86(uint64_t Xi[2], const u128 Htable[16]); |
| 362 |
void gcm_ghash_4bit_x86(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 363 |
size_t len); |
| 364 |
#endif |
| 365 |
#elif defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64) |
| 366 |
#include <openssl/arm_arch.h> |
| 367 |
#if __ARM_ARCH__ >= 7 |
| 368 |
#define GHASH_ASM_ARM |
| 369 |
#define GCM_FUNCREF_4BIT |
| 370 |
|
| 371 |
static int pmull_capable(void) { |
| 372 |
return CRYPTO_is_ARMv8_PMULL_capable(); |
| 373 |
} |
| 374 |
|
| 375 |
void gcm_init_v8(u128 Htable[16], const uint64_t Xi[2]); |
| 376 |
void gcm_gmult_v8(uint64_t Xi[2], const u128 Htable[16]); |
| 377 |
void gcm_ghash_v8(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 378 |
size_t len); |
| 379 |
|
| 380 |
#if defined(OPENSSL_ARM) |
| 381 |
/* 32-bit ARM also has support for doing GCM with NEON instructions. */ |
| 382 |
static int neon_capable(void) { |
| 383 |
return CRYPTO_is_NEON_capable(); |
| 384 |
} |
| 385 |
|
| 386 |
void gcm_init_neon(u128 Htable[16], const uint64_t Xi[2]); |
| 387 |
void gcm_gmult_neon(uint64_t Xi[2], const u128 Htable[16]); |
| 388 |
void gcm_ghash_neon(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 389 |
size_t len); |
| 390 |
#else |
| 391 |
/* AArch64 only has the ARMv8 versions of functions. */ |
| 392 |
static int neon_capable(void) { |
| 393 |
return 0; |
| 394 |
} |
| 395 |
static void gcm_init_neon(u128 Htable[16], const uint64_t Xi[2]) { |
| 396 |
abort(); |
| 397 |
} |
| 398 |
static void gcm_gmult_neon(uint64_t Xi[2], const u128 Htable[16]) { |
| 399 |
abort(); |
| 400 |
} |
| 401 |
static void gcm_ghash_neon(uint64_t Xi[2], const u128 Htable[16], |
| 402 |
const uint8_t *inp, size_t len) { |
| 403 |
abort(); |
| 404 |
} |
| 405 |
#endif |
| 406 |
|
| 407 |
#endif |
| 408 |
#endif |
| 409 |
#endif |
| 410 |
|
| 411 |
#ifdef GCM_FUNCREF_4BIT |
| 412 |
#undef GCM_MUL |
| 413 |
#define GCM_MUL(ctx, Xi) (*gcm_gmult_p)(ctx->Xi.u, ctx->Htable) |
| 414 |
#ifdef GHASH |
| 415 |
#undef GHASH |
| 416 |
#define GHASH(ctx, in, len) (*gcm_ghash_p)(ctx->Xi.u, ctx->Htable, in, len) |
| 417 |
#endif |
| 418 |
#endif |
| 419 |
|
| 420 |
void CRYPTO_gcm128_init(GCM128_CONTEXT *ctx, const void *key, |
| 421 |
block128_f block) { |
| 422 |
const union { |
| 423 |
long one; |
| 424 |
char little; |
| 425 |
} is_endian = {1}; |
| 426 |
|
| 427 |
memset(ctx, 0, sizeof(*ctx)); |
| 428 |
ctx->block = block; |
| 429 |
|
| 430 |
(*block)(ctx->H.c, ctx->H.c, key); |
| 431 |
|
| 432 |
if (is_endian.little) { |
| 433 |
/* H is stored in host byte order */ |
| 434 |
#ifdef BSWAP8 |
| 435 |
ctx->H.u[0] = BSWAP8(ctx->H.u[0]); |
| 436 |
ctx->H.u[1] = BSWAP8(ctx->H.u[1]); |
| 437 |
#else |
| 438 |
uint8_t *p = ctx->H.c; |
| 439 |
uint64_t hi, lo; |
| 440 |
hi = (uint64_t)GETU32(p) << 32 | GETU32(p + 4); |
| 441 |
lo = (uint64_t)GETU32(p + 8) << 32 | GETU32(p + 12); |
| 442 |
ctx->H.u[0] = hi; |
| 443 |
ctx->H.u[1] = lo; |
| 444 |
#endif |
| 445 |
} |
| 446 |
|
| 447 |
#if defined(GHASH_ASM_X86_OR_64) |
| 448 |
if (crypto_gcm_clmul_enabled()) { |
| 449 |
if (((OPENSSL_ia32cap_P[1] >> 22) & 0x41) == 0x41) { /* AVX+MOVBE */ |
| 450 |
gcm_init_avx(ctx->Htable, ctx->H.u); |
| 451 |
ctx->gmult = gcm_gmult_avx; |
| 452 |
ctx->ghash = gcm_ghash_avx; |
| 453 |
} else { |
| 454 |
gcm_init_clmul(ctx->Htable, ctx->H.u); |
| 455 |
ctx->gmult = gcm_gmult_clmul; |
| 456 |
ctx->ghash = gcm_ghash_clmul; |
| 457 |
} |
| 458 |
return; |
| 459 |
} |
| 460 |
gcm_init_4bit(ctx->Htable, ctx->H.u); |
| 461 |
#if defined(GHASH_ASM_X86) /* x86 only */ |
| 462 |
if (OPENSSL_ia32cap_P[0] & (1 << 25)) { /* check SSE bit */ |
| 463 |
ctx->gmult = gcm_gmult_4bit_mmx; |
| 464 |
ctx->ghash = gcm_ghash_4bit_mmx; |
| 465 |
} else { |
| 466 |
ctx->gmult = gcm_gmult_4bit_x86; |
| 467 |
ctx->ghash = gcm_ghash_4bit_x86; |
| 468 |
} |
| 469 |
#else |
| 470 |
ctx->gmult = gcm_gmult_4bit; |
| 471 |
ctx->ghash = gcm_ghash_4bit; |
| 472 |
#endif |
| 473 |
#elif defined(GHASH_ASM_ARM) |
| 474 |
if (pmull_capable()) { |
| 475 |
gcm_init_v8(ctx->Htable, ctx->H.u); |
| 476 |
ctx->gmult = gcm_gmult_v8; |
| 477 |
ctx->ghash = gcm_ghash_v8; |
| 478 |
} else if (neon_capable()) { |
| 479 |
gcm_init_neon(ctx->Htable,ctx->H.u); |
| 480 |
ctx->gmult = gcm_gmult_neon; |
| 481 |
ctx->ghash = gcm_ghash_neon; |
| 482 |
} else { |
| 483 |
gcm_init_4bit(ctx->Htable, ctx->H.u); |
| 484 |
ctx->gmult = gcm_gmult_4bit; |
| 485 |
ctx->ghash = gcm_ghash_4bit; |
| 486 |
} |
| 487 |
#else |
| 488 |
gcm_init_4bit(ctx->Htable, ctx->H.u); |
| 489 |
ctx->gmult = gcm_gmult_4bit; |
| 490 |
ctx->ghash = gcm_ghash_4bit; |
| 491 |
#endif |
| 492 |
} |
| 493 |
|
| 494 |
void CRYPTO_gcm128_setiv(GCM128_CONTEXT *ctx, const void *key, |
| 495 |
const uint8_t *iv, size_t len) { |
| 496 |
const union { |
| 497 |
long one; |
| 498 |
char little; |
| 499 |
} is_endian = {1}; |
| 500 |
unsigned int ctr; |
| 501 |
#ifdef GCM_FUNCREF_4BIT |
| 502 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 503 |
#endif |
| 504 |
|
| 505 |
ctx->Yi.u[0] = 0; |
| 506 |
ctx->Yi.u[1] = 0; |
| 507 |
ctx->Xi.u[0] = 0; |
| 508 |
ctx->Xi.u[1] = 0; |
| 509 |
ctx->len.u[0] = 0; /* AAD length */ |
| 510 |
ctx->len.u[1] = 0; /* message length */ |
| 511 |
ctx->ares = 0; |
| 512 |
ctx->mres = 0; |
| 513 |
|
| 514 |
if (len == 12) { |
| 515 |
memcpy(ctx->Yi.c, iv, 12); |
| 516 |
ctx->Yi.c[15] = 1; |
| 517 |
ctr = 1; |
| 518 |
} else { |
| 519 |
size_t i; |
| 520 |
uint64_t len0 = len; |
| 521 |
|
| 522 |
while (len >= 16) { |
| 523 |
for (i = 0; i < 16; ++i) { |
| 524 |
ctx->Yi.c[i] ^= iv[i]; |
| 525 |
} |
| 526 |
GCM_MUL(ctx, Yi); |
| 527 |
iv += 16; |
| 528 |
len -= 16; |
| 529 |
} |
| 530 |
if (len) { |
| 531 |
for (i = 0; i < len; ++i) { |
| 532 |
ctx->Yi.c[i] ^= iv[i]; |
| 533 |
} |
| 534 |
GCM_MUL(ctx, Yi); |
| 535 |
} |
| 536 |
len0 <<= 3; |
| 537 |
if (is_endian.little) { |
| 538 |
#ifdef BSWAP8 |
| 539 |
ctx->Yi.u[1] ^= BSWAP8(len0); |
| 540 |
#else |
| 541 |
ctx->Yi.c[8] ^= (uint8_t)(len0 >> 56); |
| 542 |
ctx->Yi.c[9] ^= (uint8_t)(len0 >> 48); |
| 543 |
ctx->Yi.c[10] ^= (uint8_t)(len0 >> 40); |
| 544 |
ctx->Yi.c[11] ^= (uint8_t)(len0 >> 32); |
| 545 |
ctx->Yi.c[12] ^= (uint8_t)(len0 >> 24); |
| 546 |
ctx->Yi.c[13] ^= (uint8_t)(len0 >> 16); |
| 547 |
ctx->Yi.c[14] ^= (uint8_t)(len0 >> 8); |
| 548 |
ctx->Yi.c[15] ^= (uint8_t)(len0); |
| 549 |
#endif |
| 550 |
} else { |
| 551 |
ctx->Yi.u[1] ^= len0; |
| 552 |
} |
| 553 |
|
| 554 |
GCM_MUL(ctx, Yi); |
| 555 |
|
| 556 |
if (is_endian.little) { |
| 557 |
ctr = GETU32(ctx->Yi.c + 12); |
| 558 |
} else { |
| 559 |
ctr = ctx->Yi.d[3]; |
| 560 |
} |
| 561 |
} |
| 562 |
|
| 563 |
(*ctx->block)(ctx->Yi.c, ctx->EK0.c, key); |
| 564 |
++ctr; |
| 565 |
if (is_endian.little) { |
| 566 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 567 |
} else { |
| 568 |
ctx->Yi.d[3] = ctr; |
| 569 |
} |
| 570 |
} |
| 571 |
|
| 572 |
int CRYPTO_gcm128_aad(GCM128_CONTEXT *ctx, const uint8_t *aad, size_t len) { |
| 573 |
size_t i; |
| 574 |
unsigned int n; |
| 575 |
uint64_t alen = ctx->len.u[0]; |
| 576 |
#ifdef GCM_FUNCREF_4BIT |
| 577 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 578 |
#ifdef GHASH |
| 579 |
void (*gcm_ghash_p)(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 580 |
size_t len) = ctx->ghash; |
| 581 |
#endif |
| 582 |
#endif |
| 583 |
|
| 584 |
if (ctx->len.u[1]) { |
| 585 |
return 0; |
| 586 |
} |
| 587 |
|
| 588 |
alen += len; |
| 589 |
if (alen > (UINT64_C(1) << 61) || (sizeof(len) == 8 && alen < len)) { |
| 590 |
return 0; |
| 591 |
} |
| 592 |
ctx->len.u[0] = alen; |
| 593 |
|
| 594 |
n = ctx->ares; |
| 595 |
if (n) { |
| 596 |
while (n && len) { |
| 597 |
ctx->Xi.c[n] ^= *(aad++); |
| 598 |
--len; |
| 599 |
n = (n + 1) % 16; |
| 600 |
} |
| 601 |
if (n == 0) { |
| 602 |
GCM_MUL(ctx, Xi); |
| 603 |
} else { |
| 604 |
ctx->ares = n; |
| 605 |
return 1; |
| 606 |
} |
| 607 |
} |
| 608 |
|
| 609 |
#ifdef GHASH |
| 610 |
i = len & kSizeTWithoutLower4Bits; |
| 611 |
if (i != 0) { |
| 612 |
GHASH(ctx, aad, i); |
| 613 |
aad += i; |
| 614 |
len -= i; |
| 615 |
} |
| 616 |
#else |
| 617 |
while (len >= 16) { |
| 618 |
for (i = 0; i < 16; ++i) { |
| 619 |
ctx->Xi.c[i] ^= aad[i]; |
| 620 |
} |
| 621 |
GCM_MUL(ctx, Xi); |
| 622 |
aad += 16; |
| 623 |
len -= 16; |
| 624 |
} |
| 625 |
#endif |
| 626 |
if (len) { |
| 627 |
n = (unsigned int)len; |
| 628 |
for (i = 0; i < len; ++i) { |
| 629 |
ctx->Xi.c[i] ^= aad[i]; |
| 630 |
} |
| 631 |
} |
| 632 |
|
| 633 |
ctx->ares = n; |
| 634 |
return 1; |
| 635 |
} |
| 636 |
|
| 637 |
int CRYPTO_gcm128_encrypt(GCM128_CONTEXT *ctx, const void *key, |
| 638 |
const unsigned char *in, unsigned char *out, |
| 639 |
size_t len) { |
| 640 |
const union { |
| 641 |
long one; |
| 642 |
char little; |
| 643 |
} is_endian = {1}; |
| 644 |
unsigned int n, ctr; |
| 645 |
size_t i; |
| 646 |
uint64_t mlen = ctx->len.u[1]; |
| 647 |
block128_f block = ctx->block; |
| 648 |
#ifdef GCM_FUNCREF_4BIT |
| 649 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 650 |
#ifdef GHASH |
| 651 |
void (*gcm_ghash_p)(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 652 |
size_t len) = ctx->ghash; |
| 653 |
#endif |
| 654 |
#endif |
| 655 |
|
| 656 |
mlen += len; |
| 657 |
if (mlen > ((UINT64_C(1) << 36) - 32) || |
| 658 |
(sizeof(len) == 8 && mlen < len)) { |
| 659 |
return 0; |
| 660 |
} |
| 661 |
ctx->len.u[1] = mlen; |
| 662 |
|
| 663 |
if (ctx->ares) { |
| 664 |
/* First call to encrypt finalizes GHASH(AAD) */ |
| 665 |
GCM_MUL(ctx, Xi); |
| 666 |
ctx->ares = 0; |
| 667 |
} |
| 668 |
|
| 669 |
if (is_endian.little) { |
| 670 |
ctr = GETU32(ctx->Yi.c + 12); |
| 671 |
} else { |
| 672 |
ctr = ctx->Yi.d[3]; |
| 673 |
} |
| 674 |
|
| 675 |
n = ctx->mres; |
| 676 |
if (n) { |
| 677 |
while (n && len) { |
| 678 |
ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n]; |
| 679 |
--len; |
| 680 |
n = (n + 1) % 16; |
| 681 |
} |
| 682 |
if (n == 0) { |
| 683 |
GCM_MUL(ctx, Xi); |
| 684 |
} else { |
| 685 |
ctx->mres = n; |
| 686 |
return 1; |
| 687 |
} |
| 688 |
} |
| 689 |
if (STRICT_ALIGNMENT && ((size_t)in | (size_t)out) % sizeof(size_t) != 0) { |
| 690 |
for (i = 0; i < len; ++i) { |
| 691 |
if (n == 0) { |
| 692 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 693 |
++ctr; |
| 694 |
if (is_endian.little) { |
| 695 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 696 |
} else { |
| 697 |
ctx->Yi.d[3] = ctr; |
| 698 |
} |
| 699 |
} |
| 700 |
ctx->Xi.c[n] ^= out[i] = in[i] ^ ctx->EKi.c[n]; |
| 701 |
n = (n + 1) % 16; |
| 702 |
if (n == 0) { |
| 703 |
GCM_MUL(ctx, Xi); |
| 704 |
} |
| 705 |
} |
| 706 |
|
| 707 |
ctx->mres = n; |
| 708 |
return 1; |
| 709 |
} |
| 710 |
#if defined(GHASH) && defined(GHASH_CHUNK) |
| 711 |
while (len >= GHASH_CHUNK) { |
| 712 |
size_t j = GHASH_CHUNK; |
| 713 |
|
| 714 |
while (j) { |
| 715 |
size_t *out_t = (size_t *)out; |
| 716 |
const size_t *in_t = (const size_t *)in; |
| 717 |
|
| 718 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 719 |
++ctr; |
| 720 |
if (is_endian.little) { |
| 721 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 722 |
} else { |
| 723 |
ctx->Yi.d[3] = ctr; |
| 724 |
} |
| 725 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 726 |
out_t[i] = in_t[i] ^ ctx->EKi.t[i]; |
| 727 |
} |
| 728 |
out += 16; |
| 729 |
in += 16; |
| 730 |
j -= 16; |
| 731 |
} |
| 732 |
GHASH(ctx, out - GHASH_CHUNK, GHASH_CHUNK); |
| 733 |
len -= GHASH_CHUNK; |
| 734 |
} |
| 735 |
if ((i = (len & (size_t) - 16))) { |
| 736 |
size_t j = i; |
| 737 |
|
| 738 |
while (len >= 16) { |
| 739 |
size_t *out_t = (size_t *)out; |
| 740 |
const size_t *in_t = (const size_t *)in; |
| 741 |
|
| 742 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 743 |
++ctr; |
| 744 |
if (is_endian.little) { |
| 745 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 746 |
} else { |
| 747 |
ctx->Yi.d[3] = ctr; |
| 748 |
} |
| 749 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 750 |
out_t[i] = in_t[i] ^ ctx->EKi.t[i]; |
| 751 |
} |
| 752 |
out += 16; |
| 753 |
in += 16; |
| 754 |
len -= 16; |
| 755 |
} |
| 756 |
GHASH(ctx, out - j, j); |
| 757 |
} |
| 758 |
#else |
| 759 |
while (len >= 16) { |
| 760 |
size_t *out_t = (size_t *)out; |
| 761 |
const size_t *in_t = (const size_t *)in; |
| 762 |
|
| 763 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 764 |
++ctr; |
| 765 |
if (is_endian.little) { |
| 766 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 767 |
} else { |
| 768 |
ctx->Yi.d[3] = ctr; |
| 769 |
} |
| 770 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 771 |
ctx->Xi.t[i] ^= out_t[i] = in_t[i] ^ ctx->EKi.t[i]; |
| 772 |
} |
| 773 |
GCM_MUL(ctx, Xi); |
| 774 |
out += 16; |
| 775 |
in += 16; |
| 776 |
len -= 16; |
| 777 |
} |
| 778 |
#endif |
| 779 |
if (len) { |
| 780 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 781 |
++ctr; |
| 782 |
if (is_endian.little) { |
| 783 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 784 |
} else { |
| 785 |
ctx->Yi.d[3] = ctr; |
| 786 |
} |
| 787 |
while (len--) { |
| 788 |
ctx->Xi.c[n] ^= out[n] = in[n] ^ ctx->EKi.c[n]; |
| 789 |
++n; |
| 790 |
} |
| 791 |
} |
| 792 |
|
| 793 |
ctx->mres = n; |
| 794 |
return 1; |
| 795 |
} |
| 796 |
|
| 797 |
int CRYPTO_gcm128_decrypt(GCM128_CONTEXT *ctx, const void *key, |
| 798 |
const unsigned char *in, unsigned char *out, |
| 799 |
size_t len) { |
| 800 |
const union { |
| 801 |
long one; |
| 802 |
char little; |
| 803 |
} is_endian = {1}; |
| 804 |
unsigned int n, ctr; |
| 805 |
size_t i; |
| 806 |
uint64_t mlen = ctx->len.u[1]; |
| 807 |
block128_f block = ctx->block; |
| 808 |
#ifdef GCM_FUNCREF_4BIT |
| 809 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 810 |
#ifdef GHASH |
| 811 |
void (*gcm_ghash_p)(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 812 |
size_t len) = ctx->ghash; |
| 813 |
#endif |
| 814 |
#endif |
| 815 |
|
| 816 |
mlen += len; |
| 817 |
if (mlen > ((UINT64_C(1) << 36) - 32) || |
| 818 |
(sizeof(len) == 8 && mlen < len)) { |
| 819 |
return 0; |
| 820 |
} |
| 821 |
ctx->len.u[1] = mlen; |
| 822 |
|
| 823 |
if (ctx->ares) { |
| 824 |
/* First call to decrypt finalizes GHASH(AAD) */ |
| 825 |
GCM_MUL(ctx, Xi); |
| 826 |
ctx->ares = 0; |
| 827 |
} |
| 828 |
|
| 829 |
if (is_endian.little) { |
| 830 |
ctr = GETU32(ctx->Yi.c + 12); |
| 831 |
} else { |
| 832 |
ctr = ctx->Yi.d[3]; |
| 833 |
} |
| 834 |
|
| 835 |
n = ctx->mres; |
| 836 |
if (n) { |
| 837 |
while (n && len) { |
| 838 |
uint8_t c = *(in++); |
| 839 |
*(out++) = c ^ ctx->EKi.c[n]; |
| 840 |
ctx->Xi.c[n] ^= c; |
| 841 |
--len; |
| 842 |
n = (n + 1) % 16; |
| 843 |
} |
| 844 |
if (n == 0) { |
| 845 |
GCM_MUL(ctx, Xi); |
| 846 |
} else { |
| 847 |
ctx->mres = n; |
| 848 |
return 1; |
| 849 |
} |
| 850 |
} |
| 851 |
if (STRICT_ALIGNMENT && ((size_t)in | (size_t)out) % sizeof(size_t) != 0) { |
| 852 |
for (i = 0; i < len; ++i) { |
| 853 |
uint8_t c; |
| 854 |
if (n == 0) { |
| 855 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 856 |
++ctr; |
| 857 |
if (is_endian.little) { |
| 858 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 859 |
} else { |
| 860 |
ctx->Yi.d[3] = ctr; |
| 861 |
} |
| 862 |
} |
| 863 |
c = in[i]; |
| 864 |
out[i] = c ^ ctx->EKi.c[n]; |
| 865 |
ctx->Xi.c[n] ^= c; |
| 866 |
n = (n + 1) % 16; |
| 867 |
if (n == 0) { |
| 868 |
GCM_MUL(ctx, Xi); |
| 869 |
} |
| 870 |
} |
| 871 |
|
| 872 |
ctx->mres = n; |
| 873 |
return 1; |
| 874 |
} |
| 875 |
#if defined(GHASH) && defined(GHASH_CHUNK) |
| 876 |
while (len >= GHASH_CHUNK) { |
| 877 |
size_t j = GHASH_CHUNK; |
| 878 |
|
| 879 |
GHASH(ctx, in, GHASH_CHUNK); |
| 880 |
while (j) { |
| 881 |
size_t *out_t = (size_t *)out; |
| 882 |
const size_t *in_t = (const size_t *)in; |
| 883 |
|
| 884 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 885 |
++ctr; |
| 886 |
if (is_endian.little) { |
| 887 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 888 |
} else { |
| 889 |
ctx->Yi.d[3] = ctr; |
| 890 |
} |
| 891 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 892 |
out_t[i] = in_t[i] ^ ctx->EKi.t[i]; |
| 893 |
} |
| 894 |
out += 16; |
| 895 |
in += 16; |
| 896 |
j -= 16; |
| 897 |
} |
| 898 |
len -= GHASH_CHUNK; |
| 899 |
} |
| 900 |
i = len & kSizeTWithoutLower4Bits; |
| 901 |
if (i != 0) { |
| 902 |
GHASH(ctx, in, i); |
| 903 |
while (len >= 16) { |
| 904 |
size_t *out_t = (size_t *)out; |
| 905 |
const size_t *in_t = (const size_t *)in; |
| 906 |
|
| 907 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 908 |
++ctr; |
| 909 |
if (is_endian.little) { |
| 910 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 911 |
} else { |
| 912 |
ctx->Yi.d[3] = ctr; |
| 913 |
} |
| 914 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 915 |
out_t[i] = in_t[i] ^ ctx->EKi.t[i]; |
| 916 |
} |
| 917 |
out += 16; |
| 918 |
in += 16; |
| 919 |
len -= 16; |
| 920 |
} |
| 921 |
} |
| 922 |
#else |
| 923 |
while (len >= 16) { |
| 924 |
size_t *out_t = (size_t *)out; |
| 925 |
const size_t *in_t = (const size_t *)in; |
| 926 |
|
| 927 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 928 |
++ctr; |
| 929 |
if (is_endian.little) { |
| 930 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 931 |
} else { |
| 932 |
ctx->Yi.d[3] = ctr; |
| 933 |
} |
| 934 |
for (i = 0; i < 16 / sizeof(size_t); ++i) { |
| 935 |
size_t c = in_t[i]; |
| 936 |
out_t[i] = c ^ ctx->EKi.t[i]; |
| 937 |
ctx->Xi.t[i] ^= c; |
| 938 |
} |
| 939 |
GCM_MUL(ctx, Xi); |
| 940 |
out += 16; |
| 941 |
in += 16; |
| 942 |
len -= 16; |
| 943 |
} |
| 944 |
#endif |
| 945 |
if (len) { |
| 946 |
(*block)(ctx->Yi.c, ctx->EKi.c, key); |
| 947 |
++ctr; |
| 948 |
if (is_endian.little) { |
| 949 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 950 |
} else { |
| 951 |
ctx->Yi.d[3] = ctr; |
| 952 |
} |
| 953 |
while (len--) { |
| 954 |
uint8_t c = in[n]; |
| 955 |
ctx->Xi.c[n] ^= c; |
| 956 |
out[n] = c ^ ctx->EKi.c[n]; |
| 957 |
++n; |
| 958 |
} |
| 959 |
} |
| 960 |
|
| 961 |
ctx->mres = n; |
| 962 |
return 1; |
| 963 |
} |
| 964 |
|
| 965 |
int CRYPTO_gcm128_encrypt_ctr32(GCM128_CONTEXT *ctx, const void *key, |
| 966 |
const uint8_t *in, uint8_t *out, size_t len, |
| 967 |
ctr128_f stream) { |
| 968 |
const union { |
| 969 |
long one; |
| 970 |
char little; |
| 971 |
} is_endian = {1}; |
| 972 |
unsigned int n, ctr; |
| 973 |
uint64_t mlen = ctx->len.u[1]; |
| 974 |
#ifdef GCM_FUNCREF_4BIT |
| 975 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 976 |
#ifdef GHASH |
| 977 |
void (*gcm_ghash_p)(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 978 |
size_t len) = ctx->ghash; |
| 979 |
#endif |
| 980 |
#endif |
| 981 |
|
| 982 |
mlen += len; |
| 983 |
if (mlen > ((UINT64_C(1) << 36) - 32) || |
| 984 |
(sizeof(len) == 8 && mlen < len)) { |
| 985 |
return 0; |
| 986 |
} |
| 987 |
ctx->len.u[1] = mlen; |
| 988 |
|
| 989 |
if (ctx->ares) { |
| 990 |
/* First call to encrypt finalizes GHASH(AAD) */ |
| 991 |
GCM_MUL(ctx, Xi); |
| 992 |
ctx->ares = 0; |
| 993 |
} |
| 994 |
|
| 995 |
n = ctx->mres; |
| 996 |
if (n) { |
| 997 |
while (n && len) { |
| 998 |
ctx->Xi.c[n] ^= *(out++) = *(in++) ^ ctx->EKi.c[n]; |
| 999 |
--len; |
| 1000 |
n = (n + 1) % 16; |
| 1001 |
} |
| 1002 |
if (n == 0) { |
| 1003 |
GCM_MUL(ctx, Xi); |
| 1004 |
} else { |
| 1005 |
ctx->mres = n; |
| 1006 |
return 1; |
| 1007 |
} |
| 1008 |
} |
| 1009 |
|
| 1010 |
#if defined(AESNI_GCM) |
| 1011 |
if (aesni_gcm_enabled(ctx, stream)) { |
| 1012 |
/* |aesni_gcm_encrypt| may not process all the input given to it. It may |
| 1013 |
* not process *any* of its input if it is deemed too small. */ |
| 1014 |
size_t bulk = aesni_gcm_encrypt(in, out, len, key, ctx->Yi.c, ctx->Xi.u); |
| 1015 |
in += bulk; |
| 1016 |
out += bulk; |
| 1017 |
len -= bulk; |
| 1018 |
} |
| 1019 |
#endif |
| 1020 |
|
| 1021 |
if (is_endian.little) { |
| 1022 |
ctr = GETU32(ctx->Yi.c + 12); |
| 1023 |
} else { |
| 1024 |
ctr = ctx->Yi.d[3]; |
| 1025 |
} |
| 1026 |
|
| 1027 |
#if defined(GHASH) |
| 1028 |
while (len >= GHASH_CHUNK) { |
| 1029 |
(*stream)(in, out, GHASH_CHUNK / 16, key, ctx->Yi.c); |
| 1030 |
ctr += GHASH_CHUNK / 16; |
| 1031 |
if (is_endian.little) { |
| 1032 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1033 |
} else { |
| 1034 |
ctx->Yi.d[3] = ctr; |
| 1035 |
} |
| 1036 |
GHASH(ctx, out, GHASH_CHUNK); |
| 1037 |
out += GHASH_CHUNK; |
| 1038 |
in += GHASH_CHUNK; |
| 1039 |
len -= GHASH_CHUNK; |
| 1040 |
} |
| 1041 |
#endif |
| 1042 |
size_t i = len & kSizeTWithoutLower4Bits; |
| 1043 |
if (i != 0) { |
| 1044 |
size_t j = i / 16; |
| 1045 |
|
| 1046 |
(*stream)(in, out, j, key, ctx->Yi.c); |
| 1047 |
ctr += (unsigned int)j; |
| 1048 |
if (is_endian.little) { |
| 1049 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1050 |
} else { |
| 1051 |
ctx->Yi.d[3] = ctr; |
| 1052 |
} |
| 1053 |
in += i; |
| 1054 |
len -= i; |
| 1055 |
#if defined(GHASH) |
| 1056 |
GHASH(ctx, out, i); |
| 1057 |
out += i; |
| 1058 |
#else |
| 1059 |
while (j--) { |
| 1060 |
for (i = 0; i < 16; ++i) { |
| 1061 |
ctx->Xi.c[i] ^= out[i]; |
| 1062 |
} |
| 1063 |
GCM_MUL(ctx, Xi); |
| 1064 |
out += 16; |
| 1065 |
} |
| 1066 |
#endif |
| 1067 |
} |
| 1068 |
if (len) { |
| 1069 |
(*ctx->block)(ctx->Yi.c, ctx->EKi.c, key); |
| 1070 |
++ctr; |
| 1071 |
if (is_endian.little) { |
| 1072 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1073 |
} else { |
| 1074 |
ctx->Yi.d[3] = ctr; |
| 1075 |
} |
| 1076 |
while (len--) { |
| 1077 |
ctx->Xi.c[n] ^= out[n] = in[n] ^ ctx->EKi.c[n]; |
| 1078 |
++n; |
| 1079 |
} |
| 1080 |
} |
| 1081 |
|
| 1082 |
ctx->mres = n; |
| 1083 |
return 1; |
| 1084 |
} |
| 1085 |
|
| 1086 |
int CRYPTO_gcm128_decrypt_ctr32(GCM128_CONTEXT *ctx, const void *key, |
| 1087 |
const uint8_t *in, uint8_t *out, size_t len, |
| 1088 |
ctr128_f stream) { |
| 1089 |
const union { |
| 1090 |
long one; |
| 1091 |
char little; |
| 1092 |
} is_endian = {1}; |
| 1093 |
unsigned int n, ctr; |
| 1094 |
uint64_t mlen = ctx->len.u[1]; |
| 1095 |
#ifdef GCM_FUNCREF_4BIT |
| 1096 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 1097 |
#ifdef GHASH |
| 1098 |
void (*gcm_ghash_p)(uint64_t Xi[2], const u128 Htable[16], const uint8_t *inp, |
| 1099 |
size_t len) = ctx->ghash; |
| 1100 |
#endif |
| 1101 |
#endif |
| 1102 |
|
| 1103 |
mlen += len; |
| 1104 |
if (mlen > ((UINT64_C(1) << 36) - 32) || |
| 1105 |
(sizeof(len) == 8 && mlen < len)) { |
| 1106 |
return 0; |
| 1107 |
} |
| 1108 |
ctx->len.u[1] = mlen; |
| 1109 |
|
| 1110 |
if (ctx->ares) { |
| 1111 |
/* First call to decrypt finalizes GHASH(AAD) */ |
| 1112 |
GCM_MUL(ctx, Xi); |
| 1113 |
ctx->ares = 0; |
| 1114 |
} |
| 1115 |
|
| 1116 |
n = ctx->mres; |
| 1117 |
if (n) { |
| 1118 |
while (n && len) { |
| 1119 |
uint8_t c = *(in++); |
| 1120 |
*(out++) = c ^ ctx->EKi.c[n]; |
| 1121 |
ctx->Xi.c[n] ^= c; |
| 1122 |
--len; |
| 1123 |
n = (n + 1) % 16; |
| 1124 |
} |
| 1125 |
if (n == 0) { |
| 1126 |
GCM_MUL(ctx, Xi); |
| 1127 |
} else { |
| 1128 |
ctx->mres = n; |
| 1129 |
return 1; |
| 1130 |
} |
| 1131 |
} |
| 1132 |
|
| 1133 |
#if defined(AESNI_GCM) |
| 1134 |
if (aesni_gcm_enabled(ctx, stream)) { |
| 1135 |
/* |aesni_gcm_decrypt| may not process all the input given to it. It may |
| 1136 |
* not process *any* of its input if it is deemed too small. */ |
| 1137 |
size_t bulk = aesni_gcm_decrypt(in, out, len, key, ctx->Yi.c, ctx->Xi.u); |
| 1138 |
in += bulk; |
| 1139 |
out += bulk; |
| 1140 |
len -= bulk; |
| 1141 |
} |
| 1142 |
#endif |
| 1143 |
|
| 1144 |
if (is_endian.little) { |
| 1145 |
ctr = GETU32(ctx->Yi.c + 12); |
| 1146 |
} else { |
| 1147 |
ctr = ctx->Yi.d[3]; |
| 1148 |
} |
| 1149 |
|
| 1150 |
#if defined(GHASH) |
| 1151 |
while (len >= GHASH_CHUNK) { |
| 1152 |
GHASH(ctx, in, GHASH_CHUNK); |
| 1153 |
(*stream)(in, out, GHASH_CHUNK / 16, key, ctx->Yi.c); |
| 1154 |
ctr += GHASH_CHUNK / 16; |
| 1155 |
if (is_endian.little) { |
| 1156 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1157 |
} else { |
| 1158 |
ctx->Yi.d[3] = ctr; |
| 1159 |
} |
| 1160 |
out += GHASH_CHUNK; |
| 1161 |
in += GHASH_CHUNK; |
| 1162 |
len -= GHASH_CHUNK; |
| 1163 |
} |
| 1164 |
#endif |
| 1165 |
size_t i = len & kSizeTWithoutLower4Bits; |
| 1166 |
if (i != 0) { |
| 1167 |
size_t j = i / 16; |
| 1168 |
|
| 1169 |
#if defined(GHASH) |
| 1170 |
GHASH(ctx, in, i); |
| 1171 |
#else |
| 1172 |
while (j--) { |
| 1173 |
size_t k; |
| 1174 |
for (k = 0; k < 16; ++k) { |
| 1175 |
ctx->Xi.c[k] ^= in[k]; |
| 1176 |
} |
| 1177 |
GCM_MUL(ctx, Xi); |
| 1178 |
in += 16; |
| 1179 |
} |
| 1180 |
j = i / 16; |
| 1181 |
in -= i; |
| 1182 |
#endif |
| 1183 |
(*stream)(in, out, j, key, ctx->Yi.c); |
| 1184 |
ctr += (unsigned int)j; |
| 1185 |
if (is_endian.little) { |
| 1186 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1187 |
} else { |
| 1188 |
ctx->Yi.d[3] = ctr; |
| 1189 |
} |
| 1190 |
out += i; |
| 1191 |
in += i; |
| 1192 |
len -= i; |
| 1193 |
} |
| 1194 |
if (len) { |
| 1195 |
(*ctx->block)(ctx->Yi.c, ctx->EKi.c, key); |
| 1196 |
++ctr; |
| 1197 |
if (is_endian.little) { |
| 1198 |
PUTU32(ctx->Yi.c + 12, ctr); |
| 1199 |
} else { |
| 1200 |
ctx->Yi.d[3] = ctr; |
| 1201 |
} |
| 1202 |
while (len--) { |
| 1203 |
uint8_t c = in[n]; |
| 1204 |
ctx->Xi.c[n] ^= c; |
| 1205 |
out[n] = c ^ ctx->EKi.c[n]; |
| 1206 |
++n; |
| 1207 |
} |
| 1208 |
} |
| 1209 |
|
| 1210 |
ctx->mres = n; |
| 1211 |
return 1; |
| 1212 |
} |
| 1213 |
|
| 1214 |
int CRYPTO_gcm128_finish(GCM128_CONTEXT *ctx, const uint8_t *tag, size_t len) { |
| 1215 |
const union { |
| 1216 |
long one; |
| 1217 |
char little; |
| 1218 |
} is_endian = {1}; |
| 1219 |
uint64_t alen = ctx->len.u[0] << 3; |
| 1220 |
uint64_t clen = ctx->len.u[1] << 3; |
| 1221 |
#ifdef GCM_FUNCREF_4BIT |
| 1222 |
void (*gcm_gmult_p)(uint64_t Xi[2], const u128 Htable[16]) = ctx->gmult; |
| 1223 |
#endif |
| 1224 |
|
| 1225 |
if (ctx->mres || ctx->ares) { |
| 1226 |
GCM_MUL(ctx, Xi); |
| 1227 |
} |
| 1228 |
|
| 1229 |
if (is_endian.little) { |
| 1230 |
#ifdef BSWAP8 |
| 1231 |
alen = BSWAP8(alen); |
| 1232 |
clen = BSWAP8(clen); |
| 1233 |
#else |
| 1234 |
uint8_t *p = ctx->len.c; |
| 1235 |
|
| 1236 |
ctx->len.u[0] = alen; |
| 1237 |
ctx->len.u[1] = clen; |
| 1238 |
|
| 1239 |
alen = (uint64_t)GETU32(p) << 32 | GETU32(p + 4); |
| 1240 |
clen = (uint64_t)GETU32(p + 8) << 32 | GETU32(p + 12); |
| 1241 |
#endif |
| 1242 |
} |
| 1243 |
|
| 1244 |
ctx->Xi.u[0] ^= alen; |
| 1245 |
ctx->Xi.u[1] ^= clen; |
| 1246 |
GCM_MUL(ctx, Xi); |
| 1247 |
|
| 1248 |
ctx->Xi.u[0] ^= ctx->EK0.u[0]; |
| 1249 |
ctx->Xi.u[1] ^= ctx->EK0.u[1]; |
| 1250 |
|
| 1251 |
if (tag && len <= sizeof(ctx->Xi)) { |
| 1252 |
return CRYPTO_memcmp(ctx->Xi.c, tag, len) == 0; |
| 1253 |
} else { |
| 1254 |
return 0; |
| 1255 |
} |
| 1256 |
} |
| 1257 |
|
| 1258 |
void CRYPTO_gcm128_tag(GCM128_CONTEXT *ctx, unsigned char *tag, size_t len) { |
| 1259 |
CRYPTO_gcm128_finish(ctx, NULL, 0); |
| 1260 |
memcpy(tag, ctx->Xi.c, len <= sizeof(ctx->Xi.c) ? len : sizeof(ctx->Xi.c)); |
| 1261 |
} |
| 1262 |
|
| 1263 |
#if defined(OPENSSL_X86) || defined(OPENSSL_X86_64) |
| 1264 |
int crypto_gcm_clmul_enabled(void) { |
| 1265 |
#ifdef GHASH_ASM |
| 1266 |
return OPENSSL_ia32cap_P[0] & (1 << 24) && /* check FXSR bit */ |
| 1267 |
OPENSSL_ia32cap_P[1] & (1 << 1); /* check PCLMULQDQ bit */ |
| 1268 |
#else |
| 1269 |
return 0; |
| 1270 |
#endif |
| 1271 |
} |
| 1272 |
#endif |