1 // SPDX-License-Identifier: GPL-2.0 2 /* 3 * SHA-1 and HMAC-SHA1 library functions 4 */ 5 6 #include <crypto/hmac.h> 7 #include <crypto/sha1.h> 8 #include <linux/bitops.h> 9 #include <linux/export.h> 10 #include <linux/kernel.h> 11 #include <linux/module.h> 12 #include <linux/string.h> 13 #include <linux/unaligned.h> 14 #include <linux/wordpart.h> 15 #include "fips.h" 16 17 static const struct sha1_block_state sha1_iv = { 18 .h = { SHA1_H0, SHA1_H1, SHA1_H2, SHA1_H3, SHA1_H4 }, 19 }; 20 21 /* 22 * If you have 32 registers or more, the compiler can (and should) 23 * try to change the array[] accesses into registers. However, on 24 * machines with less than ~25 registers, that won't really work, 25 * and at least gcc will make an unholy mess of it. 26 * 27 * So to avoid that mess which just slows things down, we force 28 * the stores to memory to actually happen (we might be better off 29 * with a 'W(t)=(val);asm("":"+m" (W(t))' there instead, as 30 * suggested by Artur Skawina - that will also make gcc unable to 31 * try to do the silly "optimize away loads" part because it won't 32 * see what the value will be). 33 * 34 * Ben Herrenschmidt reports that on PPC, the C version comes close 35 * to the optimized asm with this (ie on PPC you don't want that 36 * 'volatile', since there are lots of registers). 37 * 38 * On ARM we get the best code generation by forcing a full memory barrier 39 * between each SHA_ROUND, otherwise gcc happily get wild with spilling and 40 * the stack frame size simply explode and performance goes down the drain. 41 */ 42 43 #ifdef CONFIG_X86 44 #define setW(x, val) (*(volatile __u32 *)&W(x) = (val)) 45 #elif defined(CONFIG_ARM) 46 #define setW(x, val) do { W(x) = (val); __asm__("":::"memory"); } while (0) 47 #else 48 #define setW(x, val) (W(x) = (val)) 49 #endif 50 51 /* This "rolls" over the 512-bit array */ 52 #define W(x) (array[(x)&15]) 53 54 /* 55 * Where do we get the source from? The first 16 iterations get it from 56 * the input data, the next mix it from the 512-bit array. 57 */ 58 #define SHA_SRC(t) get_unaligned_be32((__u32 *)data + t) 59 #define SHA_MIX(t) rol32(W(t+13) ^ W(t+8) ^ W(t+2) ^ W(t), 1) 60 61 #define SHA_ROUND(t, input, fn, constant, A, B, C, D, E) do { \ 62 __u32 TEMP = input(t); setW(t, TEMP); \ 63 E += TEMP + rol32(A,5) + (fn) + (constant); \ 64 B = ror32(B, 2); \ 65 TEMP = E; E = D; D = C; C = B; B = A; A = TEMP; } while (0) 66 67 #define T_0_15(t, A, B, C, D, E) SHA_ROUND(t, SHA_SRC, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) 68 #define T_16_19(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (((C^D)&B)^D) , 0x5a827999, A, B, C, D, E ) 69 #define T_20_39(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0x6ed9eba1, A, B, C, D, E ) 70 #define T_40_59(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, ((B&C)+(D&(B^C))) , 0x8f1bbcdc, A, B, C, D, E ) 71 #define T_60_79(t, A, B, C, D, E) SHA_ROUND(t, SHA_MIX, (B^C^D) , 0xca62c1d6, A, B, C, D, E ) 72 73 /** 74 * sha1_transform - single block SHA1 transform (deprecated) 75 * 76 * @digest: 160 bit digest to update 77 * @data: 512 bits of data to hash 78 * @array: 16 words of workspace (see note) 79 * 80 * This function executes SHA-1's internal compression function. It updates the 81 * 160-bit internal state (@digest) with a single 512-bit data block (@data). 82 * 83 * Don't use this function. SHA-1 is no longer considered secure. And even if 84 * you do have to use SHA-1, this isn't the correct way to hash something with 85 * SHA-1 as this doesn't handle padding and finalization. 86 * 87 * Note: If the hash is security sensitive, the caller should be sure 88 * to clear the workspace. This is left to the caller to avoid 89 * unnecessary clears between chained hashing operations. 90 */ 91 void sha1_transform(__u32 *digest, const char *data, __u32 *array) 92 { 93 __u32 A, B, C, D, E; 94 unsigned int i = 0; 95 96 A = digest[0]; 97 B = digest[1]; 98 C = digest[2]; 99 D = digest[3]; 100 E = digest[4]; 101 102 /* Round 1 - iterations 0-16 take their input from 'data' */ 103 for (; i < 16; ++i) 104 T_0_15(i, A, B, C, D, E); 105 106 /* Round 1 - tail. Input from 512-bit mixing array */ 107 for (; i < 20; ++i) 108 T_16_19(i, A, B, C, D, E); 109 110 /* Round 2 */ 111 for (; i < 40; ++i) 112 T_20_39(i, A, B, C, D, E); 113 114 /* Round 3 */ 115 for (; i < 60; ++i) 116 T_40_59(i, A, B, C, D, E); 117 118 /* Round 4 */ 119 for (; i < 80; ++i) 120 T_60_79(i, A, B, C, D, E); 121 122 digest[0] += A; 123 digest[1] += B; 124 digest[2] += C; 125 digest[3] += D; 126 digest[4] += E; 127 } 128 EXPORT_SYMBOL(sha1_transform); 129 130 /** 131 * sha1_init_raw - initialize the vectors for a SHA1 digest 132 * @buf: vector to initialize 133 */ 134 void sha1_init_raw(__u32 *buf) 135 { 136 buf[0] = 0x67452301; 137 buf[1] = 0xefcdab89; 138 buf[2] = 0x98badcfe; 139 buf[3] = 0x10325476; 140 buf[4] = 0xc3d2e1f0; 141 } 142 EXPORT_SYMBOL(sha1_init_raw); 143 144 static void __maybe_unused sha1_blocks_generic(struct sha1_block_state *state, 145 const u8 *data, size_t nblocks) 146 { 147 u32 workspace[SHA1_WORKSPACE_WORDS]; 148 149 do { 150 sha1_transform(state->h, data, workspace); 151 data += SHA1_BLOCK_SIZE; 152 } while (--nblocks); 153 154 memzero_explicit(workspace, sizeof(workspace)); 155 } 156 157 #ifdef CONFIG_CRYPTO_LIB_SHA1_ARCH 158 #include "sha1.h" /* $(SRCARCH)/sha1.h */ 159 #else 160 #define sha1_blocks sha1_blocks_generic 161 #endif 162 163 void sha1_init(struct sha1_ctx *ctx) 164 { 165 ctx->state = sha1_iv; 166 ctx->bytecount = 0; 167 } 168 EXPORT_SYMBOL_GPL(sha1_init); 169 170 void sha1_update(struct sha1_ctx *ctx, const u8 *data, size_t len) 171 { 172 size_t partial = ctx->bytecount % SHA1_BLOCK_SIZE; 173 174 ctx->bytecount += len; 175 176 if (partial + len >= SHA1_BLOCK_SIZE) { 177 size_t nblocks; 178 179 if (partial) { 180 size_t l = SHA1_BLOCK_SIZE - partial; 181 182 memcpy(&ctx->buf[partial], data, l); 183 data += l; 184 len -= l; 185 186 sha1_blocks(&ctx->state, ctx->buf, 1); 187 } 188 189 nblocks = len / SHA1_BLOCK_SIZE; 190 len %= SHA1_BLOCK_SIZE; 191 192 if (nblocks) { 193 sha1_blocks(&ctx->state, data, nblocks); 194 data += nblocks * SHA1_BLOCK_SIZE; 195 } 196 partial = 0; 197 } 198 if (len) 199 memcpy(&ctx->buf[partial], data, len); 200 } 201 EXPORT_SYMBOL_GPL(sha1_update); 202 203 static void __sha1_final(struct sha1_ctx *ctx, u8 out[SHA1_DIGEST_SIZE]) 204 { 205 u64 bitcount = ctx->bytecount << 3; 206 size_t partial = ctx->bytecount % SHA1_BLOCK_SIZE; 207 208 ctx->buf[partial++] = 0x80; 209 if (partial > SHA1_BLOCK_SIZE - 8) { 210 memset(&ctx->buf[partial], 0, SHA1_BLOCK_SIZE - partial); 211 sha1_blocks(&ctx->state, ctx->buf, 1); 212 partial = 0; 213 } 214 memset(&ctx->buf[partial], 0, SHA1_BLOCK_SIZE - 8 - partial); 215 *(__be64 *)&ctx->buf[SHA1_BLOCK_SIZE - 8] = cpu_to_be64(bitcount); 216 sha1_blocks(&ctx->state, ctx->buf, 1); 217 218 for (size_t i = 0; i < SHA1_DIGEST_SIZE; i += 4) 219 put_unaligned_be32(ctx->state.h[i / 4], out + i); 220 } 221 222 void sha1_final(struct sha1_ctx *ctx, u8 out[SHA1_DIGEST_SIZE]) 223 { 224 __sha1_final(ctx, out); 225 memzero_explicit(ctx, sizeof(*ctx)); 226 } 227 EXPORT_SYMBOL_GPL(sha1_final); 228 229 void sha1(const u8 *data, size_t len, u8 out[SHA1_DIGEST_SIZE]) 230 { 231 struct sha1_ctx ctx; 232 233 sha1_init(&ctx); 234 sha1_update(&ctx, data, len); 235 sha1_final(&ctx, out); 236 } 237 EXPORT_SYMBOL_GPL(sha1); 238 239 static void __hmac_sha1_preparekey(struct sha1_block_state *istate, 240 struct sha1_block_state *ostate, 241 const u8 *raw_key, size_t raw_key_len) 242 { 243 union { 244 u8 b[SHA1_BLOCK_SIZE]; 245 unsigned long w[SHA1_BLOCK_SIZE / sizeof(unsigned long)]; 246 } derived_key = { 0 }; 247 248 if (unlikely(raw_key_len > SHA1_BLOCK_SIZE)) 249 sha1(raw_key, raw_key_len, derived_key.b); 250 else 251 memcpy(derived_key.b, raw_key, raw_key_len); 252 253 for (size_t i = 0; i < ARRAY_SIZE(derived_key.w); i++) 254 derived_key.w[i] ^= REPEAT_BYTE(HMAC_IPAD_VALUE); 255 *istate = sha1_iv; 256 sha1_blocks(istate, derived_key.b, 1); 257 258 for (size_t i = 0; i < ARRAY_SIZE(derived_key.w); i++) 259 derived_key.w[i] ^= REPEAT_BYTE(HMAC_OPAD_VALUE ^ 260 HMAC_IPAD_VALUE); 261 *ostate = sha1_iv; 262 sha1_blocks(ostate, derived_key.b, 1); 263 264 memzero_explicit(&derived_key, sizeof(derived_key)); 265 } 266 267 void hmac_sha1_preparekey(struct hmac_sha1_key *key, 268 const u8 *raw_key, size_t raw_key_len) 269 { 270 __hmac_sha1_preparekey(&key->istate, &key->ostate, 271 raw_key, raw_key_len); 272 } 273 EXPORT_SYMBOL_GPL(hmac_sha1_preparekey); 274 275 void hmac_sha1_init(struct hmac_sha1_ctx *ctx, const struct hmac_sha1_key *key) 276 { 277 ctx->sha_ctx.state = key->istate; 278 ctx->sha_ctx.bytecount = SHA1_BLOCK_SIZE; 279 ctx->ostate = key->ostate; 280 } 281 EXPORT_SYMBOL_GPL(hmac_sha1_init); 282 283 void hmac_sha1_init_usingrawkey(struct hmac_sha1_ctx *ctx, 284 const u8 *raw_key, size_t raw_key_len) 285 { 286 __hmac_sha1_preparekey(&ctx->sha_ctx.state, &ctx->ostate, 287 raw_key, raw_key_len); 288 ctx->sha_ctx.bytecount = SHA1_BLOCK_SIZE; 289 } 290 EXPORT_SYMBOL_GPL(hmac_sha1_init_usingrawkey); 291 292 void hmac_sha1_final(struct hmac_sha1_ctx *ctx, u8 out[SHA1_DIGEST_SIZE]) 293 { 294 /* Generate the padded input for the outer hash in ctx->sha_ctx.buf. */ 295 __sha1_final(&ctx->sha_ctx, ctx->sha_ctx.buf); 296 memset(&ctx->sha_ctx.buf[SHA1_DIGEST_SIZE], 0, 297 SHA1_BLOCK_SIZE - SHA1_DIGEST_SIZE); 298 ctx->sha_ctx.buf[SHA1_DIGEST_SIZE] = 0x80; 299 *(__be32 *)&ctx->sha_ctx.buf[SHA1_BLOCK_SIZE - 4] = 300 cpu_to_be32(8 * (SHA1_BLOCK_SIZE + SHA1_DIGEST_SIZE)); 301 302 /* Compute the outer hash, which gives the HMAC value. */ 303 sha1_blocks(&ctx->ostate, ctx->sha_ctx.buf, 1); 304 for (size_t i = 0; i < SHA1_DIGEST_SIZE; i += 4) 305 put_unaligned_be32(ctx->ostate.h[i / 4], out + i); 306 307 memzero_explicit(ctx, sizeof(*ctx)); 308 } 309 EXPORT_SYMBOL_GPL(hmac_sha1_final); 310 311 void hmac_sha1(const struct hmac_sha1_key *key, 312 const u8 *data, size_t data_len, u8 out[SHA1_DIGEST_SIZE]) 313 { 314 struct hmac_sha1_ctx ctx; 315 316 hmac_sha1_init(&ctx, key); 317 hmac_sha1_update(&ctx, data, data_len); 318 hmac_sha1_final(&ctx, out); 319 } 320 EXPORT_SYMBOL_GPL(hmac_sha1); 321 322 void hmac_sha1_usingrawkey(const u8 *raw_key, size_t raw_key_len, 323 const u8 *data, size_t data_len, 324 u8 out[SHA1_DIGEST_SIZE]) 325 { 326 struct hmac_sha1_ctx ctx; 327 328 hmac_sha1_init_usingrawkey(&ctx, raw_key, raw_key_len); 329 hmac_sha1_update(&ctx, data, data_len); 330 hmac_sha1_final(&ctx, out); 331 } 332 EXPORT_SYMBOL_GPL(hmac_sha1_usingrawkey); 333 334 #if defined(sha1_mod_init_arch) || defined(CONFIG_CRYPTO_FIPS) 335 static int __init sha1_mod_init(void) 336 { 337 #ifdef sha1_mod_init_arch 338 sha1_mod_init_arch(); 339 #endif 340 if (fips_enabled) { 341 /* 342 * FIPS cryptographic algorithm self-test. As per the FIPS 343 * Implementation Guidance, testing HMAC-SHA1 satisfies the test 344 * requirement for SHA-1 too. 345 */ 346 u8 mac[SHA1_DIGEST_SIZE]; 347 348 hmac_sha1_usingrawkey(fips_test_key, sizeof(fips_test_key), 349 fips_test_data, sizeof(fips_test_data), 350 mac); 351 if (memcmp(fips_test_hmac_sha1_value, mac, sizeof(mac)) != 0) 352 panic("sha1: FIPS self-test failed\n"); 353 } 354 return 0; 355 } 356 subsys_initcall(sha1_mod_init); 357 358 static void __exit sha1_mod_exit(void) 359 { 360 } 361 module_exit(sha1_mod_exit); 362 #endif 363 364 MODULE_DESCRIPTION("SHA-1 and HMAC-SHA1 library functions"); 365 MODULE_LICENSE("GPL"); 366