1 | /* |
2 | * Copyright 2004-2018 The OpenSSL Project Authors. All Rights Reserved. |
3 | * |
4 | * Licensed under the Apache License 2.0 (the "License"). You may not use |
5 | * this file except in compliance with the License. You can obtain a copy |
6 | * in the file LICENSE in the source distribution or at |
7 | * https://www.openssl.org/source/license.html |
8 | */ |
9 | |
10 | #include <openssl/opensslconf.h> |
11 | /*- |
12 | * IMPLEMENTATION NOTES. |
13 | * |
14 | * As you might have noticed 32-bit hash algorithms: |
15 | * |
16 | * - permit SHA_LONG to be wider than 32-bit |
17 | * - optimized versions implement two transform functions: one operating |
18 | * on [aligned] data in host byte order and one - on data in input |
19 | * stream byte order; |
20 | * - share common byte-order neutral collector and padding function |
21 | * implementations, ../md32_common.h; |
22 | * |
23 | * Neither of the above applies to this SHA-512 implementations. Reasons |
24 | * [in reverse order] are: |
25 | * |
26 | * - it's the only 64-bit hash algorithm for the moment of this writing, |
27 | * there is no need for common collector/padding implementation [yet]; |
28 | * - by supporting only one transform function [which operates on |
29 | * *aligned* data in input stream byte order, big-endian in this case] |
30 | * we minimize burden of maintenance in two ways: a) collector/padding |
31 | * function is simpler; b) only one transform function to stare at; |
32 | * - SHA_LONG64 is required to be exactly 64-bit in order to be able to |
33 | * apply a number of optimizations to mitigate potential performance |
34 | * penalties caused by previous design decision; |
35 | * |
36 | * Caveat lector. |
37 | * |
38 | * Implementation relies on the fact that "long long" is 64-bit on |
39 | * both 32- and 64-bit platforms. If some compiler vendor comes up |
40 | * with 128-bit long long, adjustment to sha.h would be required. |
41 | * As this implementation relies on 64-bit integer type, it's totally |
42 | * inappropriate for platforms which don't support it, most notably |
43 | * 16-bit platforms. |
44 | */ |
45 | #include <stdlib.h> |
46 | #include <string.h> |
47 | |
48 | #include <openssl/crypto.h> |
49 | #include <openssl/sha.h> |
50 | #include <openssl/opensslv.h> |
51 | |
52 | #include "internal/cryptlib.h" |
53 | #include "crypto/sha.h" |
54 | |
55 | #if defined(__i386) || defined(__i386__) || defined(_M_IX86) || \ |
56 | defined(__x86_64) || defined(_M_AMD64) || defined(_M_X64) || \ |
57 | defined(__s390__) || defined(__s390x__) || \ |
58 | defined(__aarch64__) || \ |
59 | defined(SHA512_ASM) |
60 | # define SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA |
61 | #endif |
62 | |
63 | #if (defined(_WIN32) || defined(_WIN64)) && !defined(__MINGW32__) |
64 | # define U64(C) C##UI64 |
65 | #elif defined(__arch64__) |
66 | # define U64(C) C##UL |
67 | #else |
68 | # define U64(C) C##ULL |
69 | #endif |
70 | |
71 | int sha512_224_init(SHA512_CTX *c) |
72 | { |
73 | c->h[0] = U64(0x8c3d37c819544da2); |
74 | c->h[1] = U64(0x73e1996689dcd4d6); |
75 | c->h[2] = U64(0x1dfab7ae32ff9c82); |
76 | c->h[3] = U64(0x679dd514582f9fcf); |
77 | c->h[4] = U64(0x0f6d2b697bd44da8); |
78 | c->h[5] = U64(0x77e36f7304c48942); |
79 | c->h[6] = U64(0x3f9d85a86a1d36c8); |
80 | c->h[7] = U64(0x1112e6ad91d692a1); |
81 | |
82 | c->Nl = 0; |
83 | c->Nh = 0; |
84 | c->num = 0; |
85 | c->md_len = SHA224_DIGEST_LENGTH; |
86 | return 1; |
87 | } |
88 | |
89 | int sha512_256_init(SHA512_CTX *c) |
90 | { |
91 | c->h[0] = U64(0x22312194fc2bf72c); |
92 | c->h[1] = U64(0x9f555fa3c84c64c2); |
93 | c->h[2] = U64(0x2393b86b6f53b151); |
94 | c->h[3] = U64(0x963877195940eabd); |
95 | c->h[4] = U64(0x96283ee2a88effe3); |
96 | c->h[5] = U64(0xbe5e1e2553863992); |
97 | c->h[6] = U64(0x2b0199fc2c85b8aa); |
98 | c->h[7] = U64(0x0eb72ddc81c52ca2); |
99 | |
100 | c->Nl = 0; |
101 | c->Nh = 0; |
102 | c->num = 0; |
103 | c->md_len = SHA256_DIGEST_LENGTH; |
104 | return 1; |
105 | } |
106 | |
107 | int SHA384_Init(SHA512_CTX *c) |
108 | { |
109 | c->h[0] = U64(0xcbbb9d5dc1059ed8); |
110 | c->h[1] = U64(0x629a292a367cd507); |
111 | c->h[2] = U64(0x9159015a3070dd17); |
112 | c->h[3] = U64(0x152fecd8f70e5939); |
113 | c->h[4] = U64(0x67332667ffc00b31); |
114 | c->h[5] = U64(0x8eb44a8768581511); |
115 | c->h[6] = U64(0xdb0c2e0d64f98fa7); |
116 | c->h[7] = U64(0x47b5481dbefa4fa4); |
117 | |
118 | c->Nl = 0; |
119 | c->Nh = 0; |
120 | c->num = 0; |
121 | c->md_len = SHA384_DIGEST_LENGTH; |
122 | return 1; |
123 | } |
124 | |
125 | int SHA512_Init(SHA512_CTX *c) |
126 | { |
127 | c->h[0] = U64(0x6a09e667f3bcc908); |
128 | c->h[1] = U64(0xbb67ae8584caa73b); |
129 | c->h[2] = U64(0x3c6ef372fe94f82b); |
130 | c->h[3] = U64(0xa54ff53a5f1d36f1); |
131 | c->h[4] = U64(0x510e527fade682d1); |
132 | c->h[5] = U64(0x9b05688c2b3e6c1f); |
133 | c->h[6] = U64(0x1f83d9abfb41bd6b); |
134 | c->h[7] = U64(0x5be0cd19137e2179); |
135 | |
136 | c->Nl = 0; |
137 | c->Nh = 0; |
138 | c->num = 0; |
139 | c->md_len = SHA512_DIGEST_LENGTH; |
140 | return 1; |
141 | } |
142 | |
143 | #ifndef SHA512_ASM |
144 | static |
145 | #endif |
146 | void sha512_block_data_order(SHA512_CTX *ctx, const void *in, size_t num); |
147 | |
148 | int SHA512_Final(unsigned char *md, SHA512_CTX *c) |
149 | { |
150 | unsigned char *p = (unsigned char *)c->u.p; |
151 | size_t n = c->num; |
152 | |
153 | p[n] = 0x80; /* There always is a room for one */ |
154 | n++; |
155 | if (n > (sizeof(c->u) - 16)) { |
156 | memset(p + n, 0, sizeof(c->u) - n); |
157 | n = 0; |
158 | sha512_block_data_order(c, p, 1); |
159 | } |
160 | |
161 | memset(p + n, 0, sizeof(c->u) - 16 - n); |
162 | #ifdef B_ENDIAN |
163 | c->u.d[SHA_LBLOCK - 2] = c->Nh; |
164 | c->u.d[SHA_LBLOCK - 1] = c->Nl; |
165 | #else |
166 | p[sizeof(c->u) - 1] = (unsigned char)(c->Nl); |
167 | p[sizeof(c->u) - 2] = (unsigned char)(c->Nl >> 8); |
168 | p[sizeof(c->u) - 3] = (unsigned char)(c->Nl >> 16); |
169 | p[sizeof(c->u) - 4] = (unsigned char)(c->Nl >> 24); |
170 | p[sizeof(c->u) - 5] = (unsigned char)(c->Nl >> 32); |
171 | p[sizeof(c->u) - 6] = (unsigned char)(c->Nl >> 40); |
172 | p[sizeof(c->u) - 7] = (unsigned char)(c->Nl >> 48); |
173 | p[sizeof(c->u) - 8] = (unsigned char)(c->Nl >> 56); |
174 | p[sizeof(c->u) - 9] = (unsigned char)(c->Nh); |
175 | p[sizeof(c->u) - 10] = (unsigned char)(c->Nh >> 8); |
176 | p[sizeof(c->u) - 11] = (unsigned char)(c->Nh >> 16); |
177 | p[sizeof(c->u) - 12] = (unsigned char)(c->Nh >> 24); |
178 | p[sizeof(c->u) - 13] = (unsigned char)(c->Nh >> 32); |
179 | p[sizeof(c->u) - 14] = (unsigned char)(c->Nh >> 40); |
180 | p[sizeof(c->u) - 15] = (unsigned char)(c->Nh >> 48); |
181 | p[sizeof(c->u) - 16] = (unsigned char)(c->Nh >> 56); |
182 | #endif |
183 | |
184 | sha512_block_data_order(c, p, 1); |
185 | |
186 | if (md == 0) |
187 | return 0; |
188 | |
189 | switch (c->md_len) { |
190 | /* Let compiler decide if it's appropriate to unroll... */ |
191 | case SHA224_DIGEST_LENGTH: |
192 | for (n = 0; n < SHA224_DIGEST_LENGTH / 8; n++) { |
193 | SHA_LONG64 t = c->h[n]; |
194 | |
195 | *(md++) = (unsigned char)(t >> 56); |
196 | *(md++) = (unsigned char)(t >> 48); |
197 | *(md++) = (unsigned char)(t >> 40); |
198 | *(md++) = (unsigned char)(t >> 32); |
199 | *(md++) = (unsigned char)(t >> 24); |
200 | *(md++) = (unsigned char)(t >> 16); |
201 | *(md++) = (unsigned char)(t >> 8); |
202 | *(md++) = (unsigned char)(t); |
203 | } |
204 | /* |
205 | * For 224 bits, there are four bytes left over that have to be |
206 | * processed separately. |
207 | */ |
208 | { |
209 | SHA_LONG64 t = c->h[SHA224_DIGEST_LENGTH / 8]; |
210 | |
211 | *(md++) = (unsigned char)(t >> 56); |
212 | *(md++) = (unsigned char)(t >> 48); |
213 | *(md++) = (unsigned char)(t >> 40); |
214 | *(md++) = (unsigned char)(t >> 32); |
215 | } |
216 | break; |
217 | case SHA256_DIGEST_LENGTH: |
218 | for (n = 0; n < SHA256_DIGEST_LENGTH / 8; n++) { |
219 | SHA_LONG64 t = c->h[n]; |
220 | |
221 | *(md++) = (unsigned char)(t >> 56); |
222 | *(md++) = (unsigned char)(t >> 48); |
223 | *(md++) = (unsigned char)(t >> 40); |
224 | *(md++) = (unsigned char)(t >> 32); |
225 | *(md++) = (unsigned char)(t >> 24); |
226 | *(md++) = (unsigned char)(t >> 16); |
227 | *(md++) = (unsigned char)(t >> 8); |
228 | *(md++) = (unsigned char)(t); |
229 | } |
230 | break; |
231 | case SHA384_DIGEST_LENGTH: |
232 | for (n = 0; n < SHA384_DIGEST_LENGTH / 8; n++) { |
233 | SHA_LONG64 t = c->h[n]; |
234 | |
235 | *(md++) = (unsigned char)(t >> 56); |
236 | *(md++) = (unsigned char)(t >> 48); |
237 | *(md++) = (unsigned char)(t >> 40); |
238 | *(md++) = (unsigned char)(t >> 32); |
239 | *(md++) = (unsigned char)(t >> 24); |
240 | *(md++) = (unsigned char)(t >> 16); |
241 | *(md++) = (unsigned char)(t >> 8); |
242 | *(md++) = (unsigned char)(t); |
243 | } |
244 | break; |
245 | case SHA512_DIGEST_LENGTH: |
246 | for (n = 0; n < SHA512_DIGEST_LENGTH / 8; n++) { |
247 | SHA_LONG64 t = c->h[n]; |
248 | |
249 | *(md++) = (unsigned char)(t >> 56); |
250 | *(md++) = (unsigned char)(t >> 48); |
251 | *(md++) = (unsigned char)(t >> 40); |
252 | *(md++) = (unsigned char)(t >> 32); |
253 | *(md++) = (unsigned char)(t >> 24); |
254 | *(md++) = (unsigned char)(t >> 16); |
255 | *(md++) = (unsigned char)(t >> 8); |
256 | *(md++) = (unsigned char)(t); |
257 | } |
258 | break; |
259 | /* ... as well as make sure md_len is not abused. */ |
260 | default: |
261 | return 0; |
262 | } |
263 | |
264 | return 1; |
265 | } |
266 | |
267 | int SHA384_Final(unsigned char *md, SHA512_CTX *c) |
268 | { |
269 | return SHA512_Final(md, c); |
270 | } |
271 | |
272 | int SHA512_Update(SHA512_CTX *c, const void *_data, size_t len) |
273 | { |
274 | SHA_LONG64 l; |
275 | unsigned char *p = c->u.p; |
276 | const unsigned char *data = (const unsigned char *)_data; |
277 | |
278 | if (len == 0) |
279 | return 1; |
280 | |
281 | l = (c->Nl + (((SHA_LONG64) len) << 3)) & U64(0xffffffffffffffff); |
282 | if (l < c->Nl) |
283 | c->Nh++; |
284 | if (sizeof(len) >= 8) |
285 | c->Nh += (((SHA_LONG64) len) >> 61); |
286 | c->Nl = l; |
287 | |
288 | if (c->num != 0) { |
289 | size_t n = sizeof(c->u) - c->num; |
290 | |
291 | if (len < n) { |
292 | memcpy(p + c->num, data, len), c->num += (unsigned int)len; |
293 | return 1; |
294 | } else { |
295 | memcpy(p + c->num, data, n), c->num = 0; |
296 | len -= n, data += n; |
297 | sha512_block_data_order(c, p, 1); |
298 | } |
299 | } |
300 | |
301 | if (len >= sizeof(c->u)) { |
302 | #ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA |
303 | if ((size_t)data % sizeof(c->u.d[0]) != 0) |
304 | while (len >= sizeof(c->u)) |
305 | memcpy(p, data, sizeof(c->u)), |
306 | sha512_block_data_order(c, p, 1), |
307 | len -= sizeof(c->u), data += sizeof(c->u); |
308 | else |
309 | #endif |
310 | sha512_block_data_order(c, data, len / sizeof(c->u)), |
311 | data += len, len %= sizeof(c->u), data -= len; |
312 | } |
313 | |
314 | if (len != 0) |
315 | memcpy(p, data, len), c->num = (int)len; |
316 | |
317 | return 1; |
318 | } |
319 | |
320 | int SHA384_Update(SHA512_CTX *c, const void *data, size_t len) |
321 | { |
322 | return SHA512_Update(c, data, len); |
323 | } |
324 | |
325 | void SHA512_Transform(SHA512_CTX *c, const unsigned char *data) |
326 | { |
327 | #ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA |
328 | if ((size_t)data % sizeof(c->u.d[0]) != 0) |
329 | memcpy(c->u.p, data, sizeof(c->u.p)), data = c->u.p; |
330 | #endif |
331 | sha512_block_data_order(c, data, 1); |
332 | } |
333 | |
334 | unsigned char *SHA384(const unsigned char *d, size_t n, unsigned char *md) |
335 | { |
336 | SHA512_CTX c; |
337 | static unsigned char m[SHA384_DIGEST_LENGTH]; |
338 | |
339 | if (md == NULL) |
340 | md = m; |
341 | SHA384_Init(&c); |
342 | SHA512_Update(&c, d, n); |
343 | SHA512_Final(md, &c); |
344 | OPENSSL_cleanse(&c, sizeof(c)); |
345 | return md; |
346 | } |
347 | |
348 | unsigned char *SHA512(const unsigned char *d, size_t n, unsigned char *md) |
349 | { |
350 | SHA512_CTX c; |
351 | static unsigned char m[SHA512_DIGEST_LENGTH]; |
352 | |
353 | if (md == NULL) |
354 | md = m; |
355 | SHA512_Init(&c); |
356 | SHA512_Update(&c, d, n); |
357 | SHA512_Final(md, &c); |
358 | OPENSSL_cleanse(&c, sizeof(c)); |
359 | return md; |
360 | } |
361 | |
362 | #ifndef SHA512_ASM |
363 | static const SHA_LONG64 K512[80] = { |
364 | U64(0x428a2f98d728ae22), U64(0x7137449123ef65cd), |
365 | U64(0xb5c0fbcfec4d3b2f), U64(0xe9b5dba58189dbbc), |
366 | U64(0x3956c25bf348b538), U64(0x59f111f1b605d019), |
367 | U64(0x923f82a4af194f9b), U64(0xab1c5ed5da6d8118), |
368 | U64(0xd807aa98a3030242), U64(0x12835b0145706fbe), |
369 | U64(0x243185be4ee4b28c), U64(0x550c7dc3d5ffb4e2), |
370 | U64(0x72be5d74f27b896f), U64(0x80deb1fe3b1696b1), |
371 | U64(0x9bdc06a725c71235), U64(0xc19bf174cf692694), |
372 | U64(0xe49b69c19ef14ad2), U64(0xefbe4786384f25e3), |
373 | U64(0x0fc19dc68b8cd5b5), U64(0x240ca1cc77ac9c65), |
374 | U64(0x2de92c6f592b0275), U64(0x4a7484aa6ea6e483), |
375 | U64(0x5cb0a9dcbd41fbd4), U64(0x76f988da831153b5), |
376 | U64(0x983e5152ee66dfab), U64(0xa831c66d2db43210), |
377 | U64(0xb00327c898fb213f), U64(0xbf597fc7beef0ee4), |
378 | U64(0xc6e00bf33da88fc2), U64(0xd5a79147930aa725), |
379 | U64(0x06ca6351e003826f), U64(0x142929670a0e6e70), |
380 | U64(0x27b70a8546d22ffc), U64(0x2e1b21385c26c926), |
381 | U64(0x4d2c6dfc5ac42aed), U64(0x53380d139d95b3df), |
382 | U64(0x650a73548baf63de), U64(0x766a0abb3c77b2a8), |
383 | U64(0x81c2c92e47edaee6), U64(0x92722c851482353b), |
384 | U64(0xa2bfe8a14cf10364), U64(0xa81a664bbc423001), |
385 | U64(0xc24b8b70d0f89791), U64(0xc76c51a30654be30), |
386 | U64(0xd192e819d6ef5218), U64(0xd69906245565a910), |
387 | U64(0xf40e35855771202a), U64(0x106aa07032bbd1b8), |
388 | U64(0x19a4c116b8d2d0c8), U64(0x1e376c085141ab53), |
389 | U64(0x2748774cdf8eeb99), U64(0x34b0bcb5e19b48a8), |
390 | U64(0x391c0cb3c5c95a63), U64(0x4ed8aa4ae3418acb), |
391 | U64(0x5b9cca4f7763e373), U64(0x682e6ff3d6b2b8a3), |
392 | U64(0x748f82ee5defb2fc), U64(0x78a5636f43172f60), |
393 | U64(0x84c87814a1f0ab72), U64(0x8cc702081a6439ec), |
394 | U64(0x90befffa23631e28), U64(0xa4506cebde82bde9), |
395 | U64(0xbef9a3f7b2c67915), U64(0xc67178f2e372532b), |
396 | U64(0xca273eceea26619c), U64(0xd186b8c721c0c207), |
397 | U64(0xeada7dd6cde0eb1e), U64(0xf57d4f7fee6ed178), |
398 | U64(0x06f067aa72176fba), U64(0x0a637dc5a2c898a6), |
399 | U64(0x113f9804bef90dae), U64(0x1b710b35131c471b), |
400 | U64(0x28db77f523047d84), U64(0x32caab7b40c72493), |
401 | U64(0x3c9ebe0a15c9bebc), U64(0x431d67c49c100d4c), |
402 | U64(0x4cc5d4becb3e42b6), U64(0x597f299cfc657e2a), |
403 | U64(0x5fcb6fab3ad6faec), U64(0x6c44198c4a475817) |
404 | }; |
405 | |
406 | # ifndef PEDANTIC |
407 | # if defined(__GNUC__) && __GNUC__>=2 && \ |
408 | !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) |
409 | # if defined(__x86_64) || defined(__x86_64__) |
410 | # define ROTR(a,n) ({ SHA_LONG64 ret; \ |
411 | asm ("rorq %1,%0" \ |
412 | : "=r"(ret) \ |
413 | : "J"(n),"0"(a) \ |
414 | : "cc"); ret; }) |
415 | # if !defined(B_ENDIAN) |
416 | # define PULL64(x) ({ SHA_LONG64 ret=*((const SHA_LONG64 *)(&(x))); \ |
417 | asm ("bswapq %0" \ |
418 | : "=r"(ret) \ |
419 | : "0"(ret)); ret; }) |
420 | # endif |
421 | # elif (defined(__i386) || defined(__i386__)) && !defined(B_ENDIAN) |
422 | # if defined(I386_ONLY) |
423 | # define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\ |
424 | unsigned int hi=p[0],lo=p[1]; \ |
425 | asm("xchgb %%ah,%%al;xchgb %%dh,%%dl;"\ |
426 | "roll $16,%%eax; roll $16,%%edx; "\ |
427 | "xchgb %%ah,%%al;xchgb %%dh,%%dl;"\ |
428 | : "=a"(lo),"=d"(hi) \ |
429 | : "0"(lo),"1"(hi) : "cc"); \ |
430 | ((SHA_LONG64)hi)<<32|lo; }) |
431 | # else |
432 | # define PULL64(x) ({ const unsigned int *p=(const unsigned int *)(&(x));\ |
433 | unsigned int hi=p[0],lo=p[1]; \ |
434 | asm ("bswapl %0; bswapl %1;" \ |
435 | : "=r"(lo),"=r"(hi) \ |
436 | : "0"(lo),"1"(hi)); \ |
437 | ((SHA_LONG64)hi)<<32|lo; }) |
438 | # endif |
439 | # elif (defined(_ARCH_PPC) && defined(__64BIT__)) || defined(_ARCH_PPC64) |
440 | # define ROTR(a,n) ({ SHA_LONG64 ret; \ |
441 | asm ("rotrdi %0,%1,%2" \ |
442 | : "=r"(ret) \ |
443 | : "r"(a),"K"(n)); ret; }) |
444 | # elif defined(__aarch64__) |
445 | # define ROTR(a,n) ({ SHA_LONG64 ret; \ |
446 | asm ("ror %0,%1,%2" \ |
447 | : "=r"(ret) \ |
448 | : "r"(a),"I"(n)); ret; }) |
449 | # if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && \ |
450 | __BYTE_ORDER__==__ORDER_LITTLE_ENDIAN__ |
451 | # define PULL64(x) ({ SHA_LONG64 ret; \ |
452 | asm ("rev %0,%1" \ |
453 | : "=r"(ret) \ |
454 | : "r"(*((const SHA_LONG64 *)(&(x))))); ret; }) |
455 | # endif |
456 | # endif |
457 | # elif defined(_MSC_VER) |
458 | # if defined(_WIN64) /* applies to both IA-64 and AMD64 */ |
459 | # pragma intrinsic(_rotr64) |
460 | # define ROTR(a,n) _rotr64((a),n) |
461 | # endif |
462 | # if defined(_M_IX86) && !defined(OPENSSL_NO_ASM) && \ |
463 | !defined(OPENSSL_NO_INLINE_ASM) |
464 | # if defined(I386_ONLY) |
465 | static SHA_LONG64 __fastcall __pull64be(const void *x) |
466 | { |
467 | _asm mov edx,[ecx + 0] |
468 | _asm mov eax,[ecx + 4] |
469 | _asm xchg dh, dl |
470 | _asm xchg ah, al |
471 | _asm rol edx, 16 |
472 | _asm rol eax, 16 |
473 | _asm xchg dh, dl |
474 | _asm xchg ah, al |
475 | } |
476 | # else |
477 | static SHA_LONG64 __fastcall __pull64be(const void *x) |
478 | { |
479 | _asm mov edx,[ecx + 0] |
480 | _asm mov eax,[ecx + 4] |
481 | _asm bswap edx |
482 | _asm bswap eax |
483 | } |
484 | # endif |
485 | # define PULL64(x) __pull64be(&(x)) |
486 | # endif |
487 | # endif |
488 | # endif |
489 | # ifndef PULL64 |
490 | # define B(x,j) (((SHA_LONG64)(*(((const unsigned char *)(&x))+j)))<<((7-j)*8)) |
491 | # define PULL64(x) (B(x,0)|B(x,1)|B(x,2)|B(x,3)|B(x,4)|B(x,5)|B(x,6)|B(x,7)) |
492 | # endif |
493 | # ifndef ROTR |
494 | # define ROTR(x,s) (((x)>>s) | (x)<<(64-s)) |
495 | # endif |
496 | # define Sigma0(x) (ROTR((x),28) ^ ROTR((x),34) ^ ROTR((x),39)) |
497 | # define Sigma1(x) (ROTR((x),14) ^ ROTR((x),18) ^ ROTR((x),41)) |
498 | # define sigma0(x) (ROTR((x),1) ^ ROTR((x),8) ^ ((x)>>7)) |
499 | # define sigma1(x) (ROTR((x),19) ^ ROTR((x),61) ^ ((x)>>6)) |
500 | # define Ch(x,y,z) (((x) & (y)) ^ ((~(x)) & (z))) |
501 | # define Maj(x,y,z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z))) |
502 | |
503 | # if defined(__i386) || defined(__i386__) || defined(_M_IX86) |
504 | /* |
505 | * This code should give better results on 32-bit CPU with less than |
506 | * ~24 registers, both size and performance wise... |
507 | */ |
508 | |
509 | static void sha512_block_data_order(SHA512_CTX *ctx, const void *in, |
510 | size_t num) |
511 | { |
512 | const SHA_LONG64 *W = in; |
513 | SHA_LONG64 A, E, T; |
514 | SHA_LONG64 X[9 + 80], *F; |
515 | int i; |
516 | |
517 | while (num--) { |
518 | |
519 | F = X + 80; |
520 | A = ctx->h[0]; |
521 | F[1] = ctx->h[1]; |
522 | F[2] = ctx->h[2]; |
523 | F[3] = ctx->h[3]; |
524 | E = ctx->h[4]; |
525 | F[5] = ctx->h[5]; |
526 | F[6] = ctx->h[6]; |
527 | F[7] = ctx->h[7]; |
528 | |
529 | for (i = 0; i < 16; i++, F--) { |
530 | # ifdef B_ENDIAN |
531 | T = W[i]; |
532 | # else |
533 | T = PULL64(W[i]); |
534 | # endif |
535 | F[0] = A; |
536 | F[4] = E; |
537 | F[8] = T; |
538 | T += F[7] + Sigma1(E) + Ch(E, F[5], F[6]) + K512[i]; |
539 | E = F[3] + T; |
540 | A = T + Sigma0(A) + Maj(A, F[1], F[2]); |
541 | } |
542 | |
543 | for (; i < 80; i++, F--) { |
544 | T = sigma0(F[8 + 16 - 1]); |
545 | T += sigma1(F[8 + 16 - 14]); |
546 | T += F[8 + 16] + F[8 + 16 - 9]; |
547 | |
548 | F[0] = A; |
549 | F[4] = E; |
550 | F[8] = T; |
551 | T += F[7] + Sigma1(E) + Ch(E, F[5], F[6]) + K512[i]; |
552 | E = F[3] + T; |
553 | A = T + Sigma0(A) + Maj(A, F[1], F[2]); |
554 | } |
555 | |
556 | ctx->h[0] += A; |
557 | ctx->h[1] += F[1]; |
558 | ctx->h[2] += F[2]; |
559 | ctx->h[3] += F[3]; |
560 | ctx->h[4] += E; |
561 | ctx->h[5] += F[5]; |
562 | ctx->h[6] += F[6]; |
563 | ctx->h[7] += F[7]; |
564 | |
565 | W += SHA_LBLOCK; |
566 | } |
567 | } |
568 | |
569 | # elif defined(OPENSSL_SMALL_FOOTPRINT) |
570 | |
571 | static void sha512_block_data_order(SHA512_CTX *ctx, const void *in, |
572 | size_t num) |
573 | { |
574 | const SHA_LONG64 *W = in; |
575 | SHA_LONG64 a, b, c, d, e, f, g, h, s0, s1, T1, T2; |
576 | SHA_LONG64 X[16]; |
577 | int i; |
578 | |
579 | while (num--) { |
580 | |
581 | a = ctx->h[0]; |
582 | b = ctx->h[1]; |
583 | c = ctx->h[2]; |
584 | d = ctx->h[3]; |
585 | e = ctx->h[4]; |
586 | f = ctx->h[5]; |
587 | g = ctx->h[6]; |
588 | h = ctx->h[7]; |
589 | |
590 | for (i = 0; i < 16; i++) { |
591 | # ifdef B_ENDIAN |
592 | T1 = X[i] = W[i]; |
593 | # else |
594 | T1 = X[i] = PULL64(W[i]); |
595 | # endif |
596 | T1 += h + Sigma1(e) + Ch(e, f, g) + K512[i]; |
597 | T2 = Sigma0(a) + Maj(a, b, c); |
598 | h = g; |
599 | g = f; |
600 | f = e; |
601 | e = d + T1; |
602 | d = c; |
603 | c = b; |
604 | b = a; |
605 | a = T1 + T2; |
606 | } |
607 | |
608 | for (; i < 80; i++) { |
609 | s0 = X[(i + 1) & 0x0f]; |
610 | s0 = sigma0(s0); |
611 | s1 = X[(i + 14) & 0x0f]; |
612 | s1 = sigma1(s1); |
613 | |
614 | T1 = X[i & 0xf] += s0 + s1 + X[(i + 9) & 0xf]; |
615 | T1 += h + Sigma1(e) + Ch(e, f, g) + K512[i]; |
616 | T2 = Sigma0(a) + Maj(a, b, c); |
617 | h = g; |
618 | g = f; |
619 | f = e; |
620 | e = d + T1; |
621 | d = c; |
622 | c = b; |
623 | b = a; |
624 | a = T1 + T2; |
625 | } |
626 | |
627 | ctx->h[0] += a; |
628 | ctx->h[1] += b; |
629 | ctx->h[2] += c; |
630 | ctx->h[3] += d; |
631 | ctx->h[4] += e; |
632 | ctx->h[5] += f; |
633 | ctx->h[6] += g; |
634 | ctx->h[7] += h; |
635 | |
636 | W += SHA_LBLOCK; |
637 | } |
638 | } |
639 | |
640 | # else |
641 | # define ROUND_00_15(i,a,b,c,d,e,f,g,h) do { \ |
642 | T1 += h + Sigma1(e) + Ch(e,f,g) + K512[i]; \ |
643 | h = Sigma0(a) + Maj(a,b,c); \ |
644 | d += T1; h += T1; } while (0) |
645 | |
646 | # define ROUND_16_80(i,j,a,b,c,d,e,f,g,h,X) do { \ |
647 | s0 = X[(j+1)&0x0f]; s0 = sigma0(s0); \ |
648 | s1 = X[(j+14)&0x0f]; s1 = sigma1(s1); \ |
649 | T1 = X[(j)&0x0f] += s0 + s1 + X[(j+9)&0x0f]; \ |
650 | ROUND_00_15(i+j,a,b,c,d,e,f,g,h); } while (0) |
651 | |
652 | static void sha512_block_data_order(SHA512_CTX *ctx, const void *in, |
653 | size_t num) |
654 | { |
655 | const SHA_LONG64 *W = in; |
656 | SHA_LONG64 a, b, c, d, e, f, g, h, s0, s1, T1; |
657 | SHA_LONG64 X[16]; |
658 | int i; |
659 | |
660 | while (num--) { |
661 | |
662 | a = ctx->h[0]; |
663 | b = ctx->h[1]; |
664 | c = ctx->h[2]; |
665 | d = ctx->h[3]; |
666 | e = ctx->h[4]; |
667 | f = ctx->h[5]; |
668 | g = ctx->h[6]; |
669 | h = ctx->h[7]; |
670 | |
671 | # ifdef B_ENDIAN |
672 | T1 = X[0] = W[0]; |
673 | ROUND_00_15(0, a, b, c, d, e, f, g, h); |
674 | T1 = X[1] = W[1]; |
675 | ROUND_00_15(1, h, a, b, c, d, e, f, g); |
676 | T1 = X[2] = W[2]; |
677 | ROUND_00_15(2, g, h, a, b, c, d, e, f); |
678 | T1 = X[3] = W[3]; |
679 | ROUND_00_15(3, f, g, h, a, b, c, d, e); |
680 | T1 = X[4] = W[4]; |
681 | ROUND_00_15(4, e, f, g, h, a, b, c, d); |
682 | T1 = X[5] = W[5]; |
683 | ROUND_00_15(5, d, e, f, g, h, a, b, c); |
684 | T1 = X[6] = W[6]; |
685 | ROUND_00_15(6, c, d, e, f, g, h, a, b); |
686 | T1 = X[7] = W[7]; |
687 | ROUND_00_15(7, b, c, d, e, f, g, h, a); |
688 | T1 = X[8] = W[8]; |
689 | ROUND_00_15(8, a, b, c, d, e, f, g, h); |
690 | T1 = X[9] = W[9]; |
691 | ROUND_00_15(9, h, a, b, c, d, e, f, g); |
692 | T1 = X[10] = W[10]; |
693 | ROUND_00_15(10, g, h, a, b, c, d, e, f); |
694 | T1 = X[11] = W[11]; |
695 | ROUND_00_15(11, f, g, h, a, b, c, d, e); |
696 | T1 = X[12] = W[12]; |
697 | ROUND_00_15(12, e, f, g, h, a, b, c, d); |
698 | T1 = X[13] = W[13]; |
699 | ROUND_00_15(13, d, e, f, g, h, a, b, c); |
700 | T1 = X[14] = W[14]; |
701 | ROUND_00_15(14, c, d, e, f, g, h, a, b); |
702 | T1 = X[15] = W[15]; |
703 | ROUND_00_15(15, b, c, d, e, f, g, h, a); |
704 | # else |
705 | T1 = X[0] = PULL64(W[0]); |
706 | ROUND_00_15(0, a, b, c, d, e, f, g, h); |
707 | T1 = X[1] = PULL64(W[1]); |
708 | ROUND_00_15(1, h, a, b, c, d, e, f, g); |
709 | T1 = X[2] = PULL64(W[2]); |
710 | ROUND_00_15(2, g, h, a, b, c, d, e, f); |
711 | T1 = X[3] = PULL64(W[3]); |
712 | ROUND_00_15(3, f, g, h, a, b, c, d, e); |
713 | T1 = X[4] = PULL64(W[4]); |
714 | ROUND_00_15(4, e, f, g, h, a, b, c, d); |
715 | T1 = X[5] = PULL64(W[5]); |
716 | ROUND_00_15(5, d, e, f, g, h, a, b, c); |
717 | T1 = X[6] = PULL64(W[6]); |
718 | ROUND_00_15(6, c, d, e, f, g, h, a, b); |
719 | T1 = X[7] = PULL64(W[7]); |
720 | ROUND_00_15(7, b, c, d, e, f, g, h, a); |
721 | T1 = X[8] = PULL64(W[8]); |
722 | ROUND_00_15(8, a, b, c, d, e, f, g, h); |
723 | T1 = X[9] = PULL64(W[9]); |
724 | ROUND_00_15(9, h, a, b, c, d, e, f, g); |
725 | T1 = X[10] = PULL64(W[10]); |
726 | ROUND_00_15(10, g, h, a, b, c, d, e, f); |
727 | T1 = X[11] = PULL64(W[11]); |
728 | ROUND_00_15(11, f, g, h, a, b, c, d, e); |
729 | T1 = X[12] = PULL64(W[12]); |
730 | ROUND_00_15(12, e, f, g, h, a, b, c, d); |
731 | T1 = X[13] = PULL64(W[13]); |
732 | ROUND_00_15(13, d, e, f, g, h, a, b, c); |
733 | T1 = X[14] = PULL64(W[14]); |
734 | ROUND_00_15(14, c, d, e, f, g, h, a, b); |
735 | T1 = X[15] = PULL64(W[15]); |
736 | ROUND_00_15(15, b, c, d, e, f, g, h, a); |
737 | # endif |
738 | |
739 | for (i = 16; i < 80; i += 16) { |
740 | ROUND_16_80(i, 0, a, b, c, d, e, f, g, h, X); |
741 | ROUND_16_80(i, 1, h, a, b, c, d, e, f, g, X); |
742 | ROUND_16_80(i, 2, g, h, a, b, c, d, e, f, X); |
743 | ROUND_16_80(i, 3, f, g, h, a, b, c, d, e, X); |
744 | ROUND_16_80(i, 4, e, f, g, h, a, b, c, d, X); |
745 | ROUND_16_80(i, 5, d, e, f, g, h, a, b, c, X); |
746 | ROUND_16_80(i, 6, c, d, e, f, g, h, a, b, X); |
747 | ROUND_16_80(i, 7, b, c, d, e, f, g, h, a, X); |
748 | ROUND_16_80(i, 8, a, b, c, d, e, f, g, h, X); |
749 | ROUND_16_80(i, 9, h, a, b, c, d, e, f, g, X); |
750 | ROUND_16_80(i, 10, g, h, a, b, c, d, e, f, X); |
751 | ROUND_16_80(i, 11, f, g, h, a, b, c, d, e, X); |
752 | ROUND_16_80(i, 12, e, f, g, h, a, b, c, d, X); |
753 | ROUND_16_80(i, 13, d, e, f, g, h, a, b, c, X); |
754 | ROUND_16_80(i, 14, c, d, e, f, g, h, a, b, X); |
755 | ROUND_16_80(i, 15, b, c, d, e, f, g, h, a, X); |
756 | } |
757 | |
758 | ctx->h[0] += a; |
759 | ctx->h[1] += b; |
760 | ctx->h[2] += c; |
761 | ctx->h[3] += d; |
762 | ctx->h[4] += e; |
763 | ctx->h[5] += f; |
764 | ctx->h[6] += g; |
765 | ctx->h[7] += h; |
766 | |
767 | W += SHA_LBLOCK; |
768 | } |
769 | } |
770 | |
771 | # endif |
772 | |
773 | #endif /* SHA512_ASM */ |
774 | |