| 1 | /* Copyright (c) 2014, Google Inc. |
| 2 | * |
| 3 | * Permission to use, copy, modify, and/or distribute this software for any |
| 4 | * purpose with or without fee is hereby granted, provided that the above |
| 5 | * copyright notice and this permission notice appear in all copies. |
| 6 | * |
| 7 | * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES |
| 8 | * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF |
| 9 | * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY |
| 10 | * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES |
| 11 | * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION |
| 12 | * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN |
| 13 | * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */ |
| 14 | |
| 15 | #include <openssl/aead.h> |
| 16 | |
| 17 | #include <string.h> |
| 18 | |
| 19 | #include <openssl/chacha.h> |
| 20 | #include <openssl/cipher.h> |
| 21 | #include <openssl/cpu.h> |
| 22 | #include <openssl/err.h> |
| 23 | #include <openssl/mem.h> |
| 24 | #include <openssl/poly1305.h> |
| 25 | #include <openssl/type_check.h> |
| 26 | |
| 27 | #include "../fipsmodule/cipher/internal.h" |
| 28 | #include "../internal.h" |
| 29 | #include "../chacha/internal.h" |
| 30 | |
| 31 | |
| 32 | #define POLY1305_TAG_LEN 16 |
| 33 | |
| 34 | struct aead_chacha20_poly1305_ctx { |
| 35 | uint8_t key[32]; |
| 36 | }; |
| 37 | |
| 38 | OPENSSL_STATIC_ASSERT(sizeof(((EVP_AEAD_CTX *)NULL)->state) >= |
| 39 | sizeof(struct aead_chacha20_poly1305_ctx), |
| 40 | "AEAD state is too small" ); |
| 41 | #if defined(__GNUC__) || defined(__clang__) |
| 42 | OPENSSL_STATIC_ASSERT(alignof(union evp_aead_ctx_st_state) >= |
| 43 | alignof(struct aead_chacha20_poly1305_ctx), |
| 44 | "AEAD state has insufficient alignment" ); |
| 45 | #endif |
| 46 | |
| 47 | // For convenience (the x86_64 calling convention allows only six parameters in |
| 48 | // registers), the final parameter for the assembly functions is both an input |
| 49 | // and output parameter. |
| 50 | union open_data { |
| 51 | struct { |
| 52 | alignas(16) uint8_t key[32]; |
| 53 | uint32_t counter; |
| 54 | uint8_t nonce[12]; |
| 55 | } in; |
| 56 | struct { |
| 57 | uint8_t tag[POLY1305_TAG_LEN]; |
| 58 | } out; |
| 59 | }; |
| 60 | |
| 61 | union seal_data { |
| 62 | struct { |
| 63 | alignas(16) uint8_t key[32]; |
| 64 | uint32_t counter; |
| 65 | uint8_t nonce[12]; |
| 66 | const uint8_t *; |
| 67 | size_t ; |
| 68 | } in; |
| 69 | struct { |
| 70 | uint8_t tag[POLY1305_TAG_LEN]; |
| 71 | } out; |
| 72 | }; |
| 73 | |
| 74 | #if defined(OPENSSL_X86_64) && !defined(OPENSSL_NO_ASM) && \ |
| 75 | !defined(OPENSSL_WINDOWS) |
| 76 | static int asm_capable(void) { |
| 77 | const int sse41_capable = (OPENSSL_ia32cap_P[1] & (1 << 19)) != 0; |
| 78 | return sse41_capable; |
| 79 | } |
| 80 | |
| 81 | OPENSSL_STATIC_ASSERT(sizeof(union open_data) == 48, "wrong open_data size" ); |
| 82 | OPENSSL_STATIC_ASSERT(sizeof(union seal_data) == 48 + 8 + 8, |
| 83 | "wrong seal_data size" ); |
| 84 | |
| 85 | // chacha20_poly1305_open is defined in chacha20_poly1305_x86_64.pl. It decrypts |
| 86 | // |plaintext_len| bytes from |ciphertext| and writes them to |out_plaintext|. |
| 87 | // Additional input parameters are passed in |aead_data->in|. On exit, it will |
| 88 | // write calculated tag value to |aead_data->out.tag|, which the caller must |
| 89 | // check. |
| 90 | extern void chacha20_poly1305_open(uint8_t *out_plaintext, |
| 91 | const uint8_t *ciphertext, |
| 92 | size_t plaintext_len, const uint8_t *ad, |
| 93 | size_t ad_len, union open_data *aead_data); |
| 94 | |
| 95 | // chacha20_poly1305_open is defined in chacha20_poly1305_x86_64.pl. It encrypts |
| 96 | // |plaintext_len| bytes from |plaintext| and writes them to |out_ciphertext|. |
| 97 | // Additional input parameters are passed in |aead_data->in|. The calculated tag |
| 98 | // value is over the computed ciphertext concatenated with |extra_ciphertext| |
| 99 | // and written to |aead_data->out.tag|. |
| 100 | extern void chacha20_poly1305_seal(uint8_t *out_ciphertext, |
| 101 | const uint8_t *plaintext, |
| 102 | size_t plaintext_len, const uint8_t *ad, |
| 103 | size_t ad_len, union seal_data *aead_data); |
| 104 | #else |
| 105 | static int asm_capable(void) { return 0; } |
| 106 | |
| 107 | |
| 108 | static void chacha20_poly1305_open(uint8_t *out_plaintext, |
| 109 | const uint8_t *ciphertext, |
| 110 | size_t plaintext_len, const uint8_t *ad, |
| 111 | size_t ad_len, union open_data *aead_data) {} |
| 112 | |
| 113 | static void chacha20_poly1305_seal(uint8_t *out_ciphertext, |
| 114 | const uint8_t *plaintext, |
| 115 | size_t plaintext_len, const uint8_t *ad, |
| 116 | size_t ad_len, union seal_data *aead_data) {} |
| 117 | #endif |
| 118 | |
| 119 | static int aead_chacha20_poly1305_init(EVP_AEAD_CTX *ctx, const uint8_t *key, |
| 120 | size_t key_len, size_t tag_len) { |
| 121 | struct aead_chacha20_poly1305_ctx *c20_ctx = |
| 122 | (struct aead_chacha20_poly1305_ctx *)&ctx->state; |
| 123 | |
| 124 | if (tag_len == 0) { |
| 125 | tag_len = POLY1305_TAG_LEN; |
| 126 | } |
| 127 | |
| 128 | if (tag_len > POLY1305_TAG_LEN) { |
| 129 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_TOO_LARGE); |
| 130 | return 0; |
| 131 | } |
| 132 | |
| 133 | if (key_len != sizeof(c20_ctx->key)) { |
| 134 | return 0; // internal error - EVP_AEAD_CTX_init should catch this. |
| 135 | } |
| 136 | |
| 137 | OPENSSL_memcpy(c20_ctx->key, key, key_len); |
| 138 | ctx->tag_len = tag_len; |
| 139 | |
| 140 | return 1; |
| 141 | } |
| 142 | |
| 143 | static void aead_chacha20_poly1305_cleanup(EVP_AEAD_CTX *ctx) {} |
| 144 | |
| 145 | static void poly1305_update_length(poly1305_state *poly1305, size_t data_len) { |
| 146 | uint8_t length_bytes[8]; |
| 147 | |
| 148 | for (unsigned i = 0; i < sizeof(length_bytes); i++) { |
| 149 | length_bytes[i] = data_len; |
| 150 | data_len >>= 8; |
| 151 | } |
| 152 | |
| 153 | CRYPTO_poly1305_update(poly1305, length_bytes, sizeof(length_bytes)); |
| 154 | } |
| 155 | |
| 156 | // calc_tag fills |tag| with the authentication tag for the given inputs. |
| 157 | static void calc_tag(uint8_t tag[POLY1305_TAG_LEN], const uint8_t *key, |
| 158 | const uint8_t nonce[12], const uint8_t *ad, size_t ad_len, |
| 159 | const uint8_t *ciphertext, size_t ciphertext_len, |
| 160 | const uint8_t *, |
| 161 | size_t ) { |
| 162 | alignas(16) uint8_t poly1305_key[32]; |
| 163 | OPENSSL_memset(poly1305_key, 0, sizeof(poly1305_key)); |
| 164 | CRYPTO_chacha_20(poly1305_key, poly1305_key, sizeof(poly1305_key), key, nonce, |
| 165 | 0); |
| 166 | |
| 167 | static const uint8_t padding[16] = { 0 }; // Padding is all zeros. |
| 168 | poly1305_state ctx; |
| 169 | CRYPTO_poly1305_init(&ctx, poly1305_key); |
| 170 | CRYPTO_poly1305_update(&ctx, ad, ad_len); |
| 171 | if (ad_len % 16 != 0) { |
| 172 | CRYPTO_poly1305_update(&ctx, padding, sizeof(padding) - (ad_len % 16)); |
| 173 | } |
| 174 | CRYPTO_poly1305_update(&ctx, ciphertext, ciphertext_len); |
| 175 | CRYPTO_poly1305_update(&ctx, ciphertext_extra, ciphertext_extra_len); |
| 176 | const size_t ciphertext_total = ciphertext_len + ciphertext_extra_len; |
| 177 | if (ciphertext_total % 16 != 0) { |
| 178 | CRYPTO_poly1305_update(&ctx, padding, |
| 179 | sizeof(padding) - (ciphertext_total % 16)); |
| 180 | } |
| 181 | poly1305_update_length(&ctx, ad_len); |
| 182 | poly1305_update_length(&ctx, ciphertext_total); |
| 183 | CRYPTO_poly1305_finish(&ctx, tag); |
| 184 | } |
| 185 | |
| 186 | static int chacha20_poly1305_seal_scatter( |
| 187 | const uint8_t *key, uint8_t *out, uint8_t *out_tag, |
| 188 | size_t *out_tag_len, size_t max_out_tag_len, const uint8_t *nonce, |
| 189 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *, |
| 190 | size_t , const uint8_t *ad, size_t ad_len, size_t tag_len) { |
| 191 | if (extra_in_len + tag_len < tag_len) { |
| 192 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_TOO_LARGE); |
| 193 | return 0; |
| 194 | } |
| 195 | if (max_out_tag_len < tag_len + extra_in_len) { |
| 196 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_BUFFER_TOO_SMALL); |
| 197 | return 0; |
| 198 | } |
| 199 | if (nonce_len != 12) { |
| 200 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_UNSUPPORTED_NONCE_SIZE); |
| 201 | return 0; |
| 202 | } |
| 203 | |
| 204 | // |CRYPTO_chacha_20| uses a 32-bit block counter. Therefore we disallow |
| 205 | // individual operations that work on more than 256GB at a time. |
| 206 | // |in_len_64| is needed because, on 32-bit platforms, size_t is only |
| 207 | // 32-bits and this produces a warning because it's always false. |
| 208 | // Casting to uint64_t inside the conditional is not sufficient to stop |
| 209 | // the warning. |
| 210 | const uint64_t in_len_64 = in_len; |
| 211 | if (in_len_64 >= (UINT64_C(1) << 32) * 64 - 64) { |
| 212 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_TOO_LARGE); |
| 213 | return 0; |
| 214 | } |
| 215 | |
| 216 | if (max_out_tag_len < tag_len) { |
| 217 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_BUFFER_TOO_SMALL); |
| 218 | return 0; |
| 219 | } |
| 220 | |
| 221 | // The the extra input is given, it is expected to be very short and so is |
| 222 | // encrypted byte-by-byte first. |
| 223 | if (extra_in_len) { |
| 224 | static const size_t kChaChaBlockSize = 64; |
| 225 | uint32_t block_counter = 1 + (in_len / kChaChaBlockSize); |
| 226 | size_t offset = in_len % kChaChaBlockSize; |
| 227 | uint8_t block[64 /* kChaChaBlockSize */]; |
| 228 | |
| 229 | for (size_t done = 0; done < extra_in_len; block_counter++) { |
| 230 | memset(block, 0, sizeof(block)); |
| 231 | CRYPTO_chacha_20(block, block, sizeof(block), key, nonce, |
| 232 | block_counter); |
| 233 | for (size_t i = offset; i < sizeof(block) && done < extra_in_len; |
| 234 | i++, done++) { |
| 235 | out_tag[done] = extra_in[done] ^ block[i]; |
| 236 | } |
| 237 | offset = 0; |
| 238 | } |
| 239 | } |
| 240 | |
| 241 | union seal_data data; |
| 242 | if (asm_capable()) { |
| 243 | OPENSSL_memcpy(data.in.key, key, 32); |
| 244 | data.in.counter = 0; |
| 245 | OPENSSL_memcpy(data.in.nonce, nonce, 12); |
| 246 | data.in.extra_ciphertext = out_tag; |
| 247 | data.in.extra_ciphertext_len = extra_in_len; |
| 248 | chacha20_poly1305_seal(out, in, in_len, ad, ad_len, &data); |
| 249 | } else { |
| 250 | CRYPTO_chacha_20(out, in, in_len, key, nonce, 1); |
| 251 | calc_tag(data.out.tag, key, nonce, ad, ad_len, out, in_len, out_tag, |
| 252 | extra_in_len); |
| 253 | } |
| 254 | |
| 255 | OPENSSL_memcpy(out_tag + extra_in_len, data.out.tag, tag_len); |
| 256 | *out_tag_len = extra_in_len + tag_len; |
| 257 | return 1; |
| 258 | } |
| 259 | |
| 260 | static int aead_chacha20_poly1305_seal_scatter( |
| 261 | const EVP_AEAD_CTX *ctx, uint8_t *out, uint8_t *out_tag, |
| 262 | size_t *out_tag_len, size_t max_out_tag_len, const uint8_t *nonce, |
| 263 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *, |
| 264 | size_t , const uint8_t *ad, size_t ad_len) { |
| 265 | const struct aead_chacha20_poly1305_ctx *c20_ctx = |
| 266 | (struct aead_chacha20_poly1305_ctx *)&ctx->state; |
| 267 | |
| 268 | return chacha20_poly1305_seal_scatter( |
| 269 | c20_ctx->key, out, out_tag, out_tag_len, max_out_tag_len, nonce, |
| 270 | nonce_len, in, in_len, extra_in, extra_in_len, ad, ad_len, ctx->tag_len); |
| 271 | } |
| 272 | |
| 273 | static int aead_xchacha20_poly1305_seal_scatter( |
| 274 | const EVP_AEAD_CTX *ctx, uint8_t *out, uint8_t *out_tag, |
| 275 | size_t *out_tag_len, size_t max_out_tag_len, const uint8_t *nonce, |
| 276 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *, |
| 277 | size_t , const uint8_t *ad, size_t ad_len) { |
| 278 | const struct aead_chacha20_poly1305_ctx *c20_ctx = |
| 279 | (struct aead_chacha20_poly1305_ctx *)&ctx->state; |
| 280 | |
| 281 | if (nonce_len != 24) { |
| 282 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_UNSUPPORTED_NONCE_SIZE); |
| 283 | return 0; |
| 284 | } |
| 285 | |
| 286 | alignas(4) uint8_t derived_key[32]; |
| 287 | alignas(4) uint8_t derived_nonce[12]; |
| 288 | CRYPTO_hchacha20(derived_key, c20_ctx->key, nonce); |
| 289 | OPENSSL_memset(derived_nonce, 0, 4); |
| 290 | OPENSSL_memcpy(&derived_nonce[4], &nonce[16], 8); |
| 291 | |
| 292 | return chacha20_poly1305_seal_scatter( |
| 293 | derived_key, out, out_tag, out_tag_len, max_out_tag_len, |
| 294 | derived_nonce, sizeof(derived_nonce), in, in_len, extra_in, extra_in_len, |
| 295 | ad, ad_len, ctx->tag_len); |
| 296 | } |
| 297 | |
| 298 | static int chacha20_poly1305_open_gather( |
| 299 | const uint8_t *key, uint8_t *out, const uint8_t *nonce, |
| 300 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *in_tag, |
| 301 | size_t in_tag_len, const uint8_t *ad, size_t ad_len, size_t tag_len) { |
| 302 | if (nonce_len != 12) { |
| 303 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_UNSUPPORTED_NONCE_SIZE); |
| 304 | return 0; |
| 305 | } |
| 306 | |
| 307 | if (in_tag_len != tag_len) { |
| 308 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_BAD_DECRYPT); |
| 309 | return 0; |
| 310 | } |
| 311 | |
| 312 | // |CRYPTO_chacha_20| uses a 32-bit block counter. Therefore we disallow |
| 313 | // individual operations that work on more than 256GB at a time. |
| 314 | // |in_len_64| is needed because, on 32-bit platforms, size_t is only |
| 315 | // 32-bits and this produces a warning because it's always false. |
| 316 | // Casting to uint64_t inside the conditional is not sufficient to stop |
| 317 | // the warning. |
| 318 | const uint64_t in_len_64 = in_len; |
| 319 | if (in_len_64 >= (UINT64_C(1) << 32) * 64 - 64) { |
| 320 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_TOO_LARGE); |
| 321 | return 0; |
| 322 | } |
| 323 | |
| 324 | union open_data data; |
| 325 | if (asm_capable()) { |
| 326 | OPENSSL_memcpy(data.in.key, key, 32); |
| 327 | data.in.counter = 0; |
| 328 | OPENSSL_memcpy(data.in.nonce, nonce, 12); |
| 329 | chacha20_poly1305_open(out, in, in_len, ad, ad_len, &data); |
| 330 | } else { |
| 331 | calc_tag(data.out.tag, key, nonce, ad, ad_len, in, in_len, NULL, 0); |
| 332 | CRYPTO_chacha_20(out, in, in_len, key, nonce, 1); |
| 333 | } |
| 334 | |
| 335 | if (CRYPTO_memcmp(data.out.tag, in_tag, tag_len) != 0) { |
| 336 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_BAD_DECRYPT); |
| 337 | return 0; |
| 338 | } |
| 339 | |
| 340 | return 1; |
| 341 | } |
| 342 | |
| 343 | static int aead_chacha20_poly1305_open_gather( |
| 344 | const EVP_AEAD_CTX *ctx, uint8_t *out, const uint8_t *nonce, |
| 345 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *in_tag, |
| 346 | size_t in_tag_len, const uint8_t *ad, size_t ad_len) { |
| 347 | const struct aead_chacha20_poly1305_ctx *c20_ctx = |
| 348 | (struct aead_chacha20_poly1305_ctx *)&ctx->state; |
| 349 | |
| 350 | return chacha20_poly1305_open_gather(c20_ctx->key, out, nonce, nonce_len, in, |
| 351 | in_len, in_tag, in_tag_len, ad, ad_len, |
| 352 | ctx->tag_len); |
| 353 | } |
| 354 | |
| 355 | static int aead_xchacha20_poly1305_open_gather( |
| 356 | const EVP_AEAD_CTX *ctx, uint8_t *out, const uint8_t *nonce, |
| 357 | size_t nonce_len, const uint8_t *in, size_t in_len, const uint8_t *in_tag, |
| 358 | size_t in_tag_len, const uint8_t *ad, size_t ad_len) { |
| 359 | const struct aead_chacha20_poly1305_ctx *c20_ctx = |
| 360 | (struct aead_chacha20_poly1305_ctx *)&ctx->state; |
| 361 | |
| 362 | if (nonce_len != 24) { |
| 363 | OPENSSL_PUT_ERROR(CIPHER, CIPHER_R_UNSUPPORTED_NONCE_SIZE); |
| 364 | return 0; |
| 365 | } |
| 366 | |
| 367 | alignas(4) uint8_t derived_key[32]; |
| 368 | alignas(4) uint8_t derived_nonce[12]; |
| 369 | CRYPTO_hchacha20(derived_key, c20_ctx->key, nonce); |
| 370 | OPENSSL_memset(derived_nonce, 0, 4); |
| 371 | OPENSSL_memcpy(&derived_nonce[4], &nonce[16], 8); |
| 372 | |
| 373 | return chacha20_poly1305_open_gather( |
| 374 | derived_key, out, derived_nonce, sizeof(derived_nonce), in, in_len, |
| 375 | in_tag, in_tag_len, ad, ad_len, ctx->tag_len); |
| 376 | } |
| 377 | |
| 378 | static const EVP_AEAD aead_chacha20_poly1305 = { |
| 379 | 32, // key len |
| 380 | 12, // nonce len |
| 381 | POLY1305_TAG_LEN, // overhead |
| 382 | POLY1305_TAG_LEN, // max tag length |
| 383 | 1, // seal_scatter_supports_extra_in |
| 384 | |
| 385 | aead_chacha20_poly1305_init, |
| 386 | NULL, // init_with_direction |
| 387 | aead_chacha20_poly1305_cleanup, |
| 388 | NULL /* open */, |
| 389 | aead_chacha20_poly1305_seal_scatter, |
| 390 | aead_chacha20_poly1305_open_gather, |
| 391 | NULL, // get_iv |
| 392 | NULL, // tag_len |
| 393 | }; |
| 394 | |
| 395 | static const EVP_AEAD aead_xchacha20_poly1305 = { |
| 396 | 32, // key len |
| 397 | 24, // nonce len |
| 398 | POLY1305_TAG_LEN, // overhead |
| 399 | POLY1305_TAG_LEN, // max tag length |
| 400 | 1, // seal_scatter_supports_extra_in |
| 401 | |
| 402 | aead_chacha20_poly1305_init, |
| 403 | NULL, // init_with_direction |
| 404 | aead_chacha20_poly1305_cleanup, |
| 405 | NULL /* open */, |
| 406 | aead_xchacha20_poly1305_seal_scatter, |
| 407 | aead_xchacha20_poly1305_open_gather, |
| 408 | NULL, // get_iv |
| 409 | NULL, // tag_len |
| 410 | }; |
| 411 | |
| 412 | const EVP_AEAD *EVP_aead_chacha20_poly1305(void) { |
| 413 | return &aead_chacha20_poly1305; |
| 414 | } |
| 415 | |
| 416 | const EVP_AEAD *EVP_aead_xchacha20_poly1305(void) { |
| 417 | return &aead_xchacha20_poly1305; |
| 418 | } |
| 419 | |