1 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
2  * All rights reserved.
3  *
4  * This package is an SSL implementation written
5  * by Eric Young (eay@cryptsoft.com).
6  * The implementation was written so as to conform with Netscapes SSL.
7  *
8  * This library is free for commercial and non-commercial use as long as
9  * the following conditions are aheared to.  The following conditions
10  * apply to all code found in this distribution, be it the RC4, RSA,
11  * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
12  * included with this distribution is covered by the same copyright terms
13  * except that the holder is Tim Hudson (tjh@cryptsoft.com).
14  *
15  * Copyright remains Eric Young's, and as such any Copyright notices in
16  * the code are not to be removed.
17  * If this package is used in a product, Eric Young should be given attribution
18  * as the author of the parts of the library used.
19  * This can be in the form of a textual message at program startup or
20  * in documentation (online or textual) provided with the package.
21  *
22  * Redistribution and use in source and binary forms, with or without
23  * modification, are permitted provided that the following conditions
24  * are met:
25  * 1. Redistributions of source code must retain the copyright
26  *    notice, this list of conditions and the following disclaimer.
27  * 2. Redistributions in binary form must reproduce the above copyright
28  *    notice, this list of conditions and the following disclaimer in the
29  *    documentation and/or other materials provided with the distribution.
30  * 3. All advertising materials mentioning features or use of this software
31  *    must display the following acknowledgement:
32  *    "This product includes cryptographic software written by
33  *     Eric Young (eay@cryptsoft.com)"
34  *    The word 'cryptographic' can be left out if the rouines from the library
35  *    being used are not cryptographic related :-).
36  * 4. If you include any Windows specific code (or a derivative thereof) from
37  *    the apps directory (application code) you must include an acknowledgement:
38  *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
39  *
40  * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
41  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
44  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
50  * SUCH DAMAGE.
51  *
52  * The licence and distribution terms for any publically available version or
53  * derivative of this code cannot be changed.  i.e. this code cannot simply be
54  * copied and put under another distribution licence
55  * [including the GNU Public Licence.] */
56 
57 #include <openssl/sha.h>
58 
59 #include <string.h>
60 
61 #include <openssl/mem.h>
62 
63 #include "../internal.h"
64 
65 
66 /* IMPLEMENTATION NOTES.
67  *
68  * The 32-bit hash algorithms share a common byte-order neutral collector and
69  * padding function implementations that operate on unaligned data,
70  * ../md32_common.h. This SHA-512 implementation does not. Reasons
71  * [in reverse order] are:
72  *
73  * - It's the only 64-bit hash algorithm for the moment of this writing,
74  *   there is no need for common collector/padding implementation [yet];
75  * - By supporting only a transform function that operates on *aligned* data
76  *   the collector/padding function is simpler and easier to optimize. */
77 
78 #if !defined(OPENSSL_NO_ASM) &&                         \
79     (defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \
80      defined(OPENSSL_ARM) || defined(OPENSSL_AARCH64))
81 #define SHA512_ASM
82 #endif
83 
84 #if defined(OPENSSL_X86) || defined(OPENSSL_X86_64) || \
85     defined(__ARM_FEATURE_UNALIGNED)
86 #define SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
87 #endif
88 
SHA384_Init(SHA512_CTX * sha)89 int SHA384_Init(SHA512_CTX *sha) {
90   sha->h[0] = UINT64_C(0xcbbb9d5dc1059ed8);
91   sha->h[1] = UINT64_C(0x629a292a367cd507);
92   sha->h[2] = UINT64_C(0x9159015a3070dd17);
93   sha->h[3] = UINT64_C(0x152fecd8f70e5939);
94   sha->h[4] = UINT64_C(0x67332667ffc00b31);
95   sha->h[5] = UINT64_C(0x8eb44a8768581511);
96   sha->h[6] = UINT64_C(0xdb0c2e0d64f98fa7);
97   sha->h[7] = UINT64_C(0x47b5481dbefa4fa4);
98 
99   sha->Nl = 0;
100   sha->Nh = 0;
101   sha->num = 0;
102   sha->md_len = SHA384_DIGEST_LENGTH;
103   return 1;
104 }
105 
106 
SHA512_Init(SHA512_CTX * sha)107 int SHA512_Init(SHA512_CTX *sha) {
108   sha->h[0] = UINT64_C(0x6a09e667f3bcc908);
109   sha->h[1] = UINT64_C(0xbb67ae8584caa73b);
110   sha->h[2] = UINT64_C(0x3c6ef372fe94f82b);
111   sha->h[3] = UINT64_C(0xa54ff53a5f1d36f1);
112   sha->h[4] = UINT64_C(0x510e527fade682d1);
113   sha->h[5] = UINT64_C(0x9b05688c2b3e6c1f);
114   sha->h[6] = UINT64_C(0x1f83d9abfb41bd6b);
115   sha->h[7] = UINT64_C(0x5be0cd19137e2179);
116 
117   sha->Nl = 0;
118   sha->Nh = 0;
119   sha->num = 0;
120   sha->md_len = SHA512_DIGEST_LENGTH;
121   return 1;
122 }
123 
SHA384(const uint8_t * data,size_t len,uint8_t * out)124 uint8_t *SHA384(const uint8_t *data, size_t len, uint8_t *out) {
125   SHA512_CTX ctx;
126   static uint8_t buf[SHA384_DIGEST_LENGTH];
127 
128   /* TODO(fork): remove this static buffer. */
129   if (out == NULL) {
130     out = buf;
131   }
132 
133   SHA384_Init(&ctx);
134   SHA384_Update(&ctx, data, len);
135   SHA384_Final(out, &ctx);
136   OPENSSL_cleanse(&ctx, sizeof(ctx));
137   return out;
138 }
139 
SHA512(const uint8_t * data,size_t len,uint8_t * out)140 uint8_t *SHA512(const uint8_t *data, size_t len, uint8_t *out) {
141   SHA512_CTX ctx;
142   static uint8_t buf[SHA512_DIGEST_LENGTH];
143 
144   /* TODO(fork): remove this static buffer. */
145   if (out == NULL) {
146     out = buf;
147   }
148   SHA512_Init(&ctx);
149   SHA512_Update(&ctx, data, len);
150   SHA512_Final(out, &ctx);
151   OPENSSL_cleanse(&ctx, sizeof(ctx));
152   return out;
153 }
154 
155 #if !defined(SHA512_ASM)
156 static
157 #endif
158 void sha512_block_data_order(uint64_t *state, const uint64_t *W, size_t num);
159 
160 
SHA384_Final(uint8_t * md,SHA512_CTX * sha)161 int SHA384_Final(uint8_t *md, SHA512_CTX *sha) {
162   return SHA512_Final(md, sha);
163 }
164 
SHA384_Update(SHA512_CTX * sha,const void * data,size_t len)165 int SHA384_Update(SHA512_CTX *sha, const void *data, size_t len) {
166   return SHA512_Update(sha, data, len);
167 }
168 
SHA512_Transform(SHA512_CTX * c,const uint8_t * block)169 void SHA512_Transform(SHA512_CTX *c, const uint8_t *block) {
170 #ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
171   if ((size_t)block % sizeof(c->u.d[0]) != 0) {
172     OPENSSL_memcpy(c->u.p, block, sizeof(c->u.p));
173     block = c->u.p;
174   }
175 #endif
176   sha512_block_data_order(c->h, (uint64_t *)block, 1);
177 }
178 
SHA512_Update(SHA512_CTX * c,const void * in_data,size_t len)179 int SHA512_Update(SHA512_CTX *c, const void *in_data, size_t len) {
180   uint64_t l;
181   uint8_t *p = c->u.p;
182   const uint8_t *data = (const uint8_t *)in_data;
183 
184   if (len == 0) {
185     return 1;
186   }
187 
188   l = (c->Nl + (((uint64_t)len) << 3)) & UINT64_C(0xffffffffffffffff);
189   if (l < c->Nl) {
190     c->Nh++;
191   }
192   if (sizeof(len) >= 8) {
193     c->Nh += (((uint64_t)len) >> 61);
194   }
195   c->Nl = l;
196 
197   if (c->num != 0) {
198     size_t n = sizeof(c->u) - c->num;
199 
200     if (len < n) {
201       OPENSSL_memcpy(p + c->num, data, len);
202       c->num += (unsigned int)len;
203       return 1;
204     } else {
205       OPENSSL_memcpy(p + c->num, data, n), c->num = 0;
206       len -= n;
207       data += n;
208       sha512_block_data_order(c->h, (uint64_t *)p, 1);
209     }
210   }
211 
212   if (len >= sizeof(c->u)) {
213 #ifndef SHA512_BLOCK_CAN_MANAGE_UNALIGNED_DATA
214     if ((size_t)data % sizeof(c->u.d[0]) != 0) {
215       while (len >= sizeof(c->u)) {
216         OPENSSL_memcpy(p, data, sizeof(c->u));
217         sha512_block_data_order(c->h, (uint64_t *)p, 1);
218         len -= sizeof(c->u);
219         data += sizeof(c->u);
220       }
221     } else
222 #endif
223     {
224       sha512_block_data_order(c->h, (uint64_t *)data, len / sizeof(c->u));
225       data += len;
226       len %= sizeof(c->u);
227       data -= len;
228     }
229   }
230 
231   if (len != 0) {
232     OPENSSL_memcpy(p, data, len);
233     c->num = (int)len;
234   }
235 
236   return 1;
237 }
238 
SHA512_Final(uint8_t * md,SHA512_CTX * sha)239 int SHA512_Final(uint8_t *md, SHA512_CTX *sha) {
240   uint8_t *p = (uint8_t *)sha->u.p;
241   size_t n = sha->num;
242 
243   p[n] = 0x80; /* There always is a room for one */
244   n++;
245   if (n > (sizeof(sha->u) - 16)) {
246     OPENSSL_memset(p + n, 0, sizeof(sha->u) - n);
247     n = 0;
248     sha512_block_data_order(sha->h, (uint64_t *)p, 1);
249   }
250 
251   OPENSSL_memset(p + n, 0, sizeof(sha->u) - 16 - n);
252   p[sizeof(sha->u) - 1] = (uint8_t)(sha->Nl);
253   p[sizeof(sha->u) - 2] = (uint8_t)(sha->Nl >> 8);
254   p[sizeof(sha->u) - 3] = (uint8_t)(sha->Nl >> 16);
255   p[sizeof(sha->u) - 4] = (uint8_t)(sha->Nl >> 24);
256   p[sizeof(sha->u) - 5] = (uint8_t)(sha->Nl >> 32);
257   p[sizeof(sha->u) - 6] = (uint8_t)(sha->Nl >> 40);
258   p[sizeof(sha->u) - 7] = (uint8_t)(sha->Nl >> 48);
259   p[sizeof(sha->u) - 8] = (uint8_t)(sha->Nl >> 56);
260   p[sizeof(sha->u) - 9] = (uint8_t)(sha->Nh);
261   p[sizeof(sha->u) - 10] = (uint8_t)(sha->Nh >> 8);
262   p[sizeof(sha->u) - 11] = (uint8_t)(sha->Nh >> 16);
263   p[sizeof(sha->u) - 12] = (uint8_t)(sha->Nh >> 24);
264   p[sizeof(sha->u) - 13] = (uint8_t)(sha->Nh >> 32);
265   p[sizeof(sha->u) - 14] = (uint8_t)(sha->Nh >> 40);
266   p[sizeof(sha->u) - 15] = (uint8_t)(sha->Nh >> 48);
267   p[sizeof(sha->u) - 16] = (uint8_t)(sha->Nh >> 56);
268 
269   sha512_block_data_order(sha->h, (uint64_t *)p, 1);
270 
271   if (md == NULL) {
272     /* TODO(davidben): This NULL check is absent in other low-level hash 'final'
273      * functions and is one of the few places one can fail. */
274     return 0;
275   }
276 
277   switch (sha->md_len) {
278     /* Let compiler decide if it's appropriate to unroll... */
279     case SHA384_DIGEST_LENGTH:
280       for (n = 0; n < SHA384_DIGEST_LENGTH / 8; n++) {
281         uint64_t t = sha->h[n];
282 
283         *(md++) = (uint8_t)(t >> 56);
284         *(md++) = (uint8_t)(t >> 48);
285         *(md++) = (uint8_t)(t >> 40);
286         *(md++) = (uint8_t)(t >> 32);
287         *(md++) = (uint8_t)(t >> 24);
288         *(md++) = (uint8_t)(t >> 16);
289         *(md++) = (uint8_t)(t >> 8);
290         *(md++) = (uint8_t)(t);
291       }
292       break;
293     case SHA512_DIGEST_LENGTH:
294       for (n = 0; n < SHA512_DIGEST_LENGTH / 8; n++) {
295         uint64_t t = sha->h[n];
296 
297         *(md++) = (uint8_t)(t >> 56);
298         *(md++) = (uint8_t)(t >> 48);
299         *(md++) = (uint8_t)(t >> 40);
300         *(md++) = (uint8_t)(t >> 32);
301         *(md++) = (uint8_t)(t >> 24);
302         *(md++) = (uint8_t)(t >> 16);
303         *(md++) = (uint8_t)(t >> 8);
304         *(md++) = (uint8_t)(t);
305       }
306       break;
307     /* ... as well as make sure md_len is not abused. */
308     default:
309       /* TODO(davidben): This bad |md_len| case is one of the few places a
310        * low-level hash 'final' function can fail. This should never happen. */
311       return 0;
312   }
313 
314   return 1;
315 }
316 
317 #ifndef SHA512_ASM
318 static const uint64_t K512[80] = {
319     UINT64_C(0x428a2f98d728ae22), UINT64_C(0x7137449123ef65cd),
320     UINT64_C(0xb5c0fbcfec4d3b2f), UINT64_C(0xe9b5dba58189dbbc),
321     UINT64_C(0x3956c25bf348b538), UINT64_C(0x59f111f1b605d019),
322     UINT64_C(0x923f82a4af194f9b), UINT64_C(0xab1c5ed5da6d8118),
323     UINT64_C(0xd807aa98a3030242), UINT64_C(0x12835b0145706fbe),
324     UINT64_C(0x243185be4ee4b28c), UINT64_C(0x550c7dc3d5ffb4e2),
325     UINT64_C(0x72be5d74f27b896f), UINT64_C(0x80deb1fe3b1696b1),
326     UINT64_C(0x9bdc06a725c71235), UINT64_C(0xc19bf174cf692694),
327     UINT64_C(0xe49b69c19ef14ad2), UINT64_C(0xefbe4786384f25e3),
328     UINT64_C(0x0fc19dc68b8cd5b5), UINT64_C(0x240ca1cc77ac9c65),
329     UINT64_C(0x2de92c6f592b0275), UINT64_C(0x4a7484aa6ea6e483),
330     UINT64_C(0x5cb0a9dcbd41fbd4), UINT64_C(0x76f988da831153b5),
331     UINT64_C(0x983e5152ee66dfab), UINT64_C(0xa831c66d2db43210),
332     UINT64_C(0xb00327c898fb213f), UINT64_C(0xbf597fc7beef0ee4),
333     UINT64_C(0xc6e00bf33da88fc2), UINT64_C(0xd5a79147930aa725),
334     UINT64_C(0x06ca6351e003826f), UINT64_C(0x142929670a0e6e70),
335     UINT64_C(0x27b70a8546d22ffc), UINT64_C(0x2e1b21385c26c926),
336     UINT64_C(0x4d2c6dfc5ac42aed), UINT64_C(0x53380d139d95b3df),
337     UINT64_C(0x650a73548baf63de), UINT64_C(0x766a0abb3c77b2a8),
338     UINT64_C(0x81c2c92e47edaee6), UINT64_C(0x92722c851482353b),
339     UINT64_C(0xa2bfe8a14cf10364), UINT64_C(0xa81a664bbc423001),
340     UINT64_C(0xc24b8b70d0f89791), UINT64_C(0xc76c51a30654be30),
341     UINT64_C(0xd192e819d6ef5218), UINT64_C(0xd69906245565a910),
342     UINT64_C(0xf40e35855771202a), UINT64_C(0x106aa07032bbd1b8),
343     UINT64_C(0x19a4c116b8d2d0c8), UINT64_C(0x1e376c085141ab53),
344     UINT64_C(0x2748774cdf8eeb99), UINT64_C(0x34b0bcb5e19b48a8),
345     UINT64_C(0x391c0cb3c5c95a63), UINT64_C(0x4ed8aa4ae3418acb),
346     UINT64_C(0x5b9cca4f7763e373), UINT64_C(0x682e6ff3d6b2b8a3),
347     UINT64_C(0x748f82ee5defb2fc), UINT64_C(0x78a5636f43172f60),
348     UINT64_C(0x84c87814a1f0ab72), UINT64_C(0x8cc702081a6439ec),
349     UINT64_C(0x90befffa23631e28), UINT64_C(0xa4506cebde82bde9),
350     UINT64_C(0xbef9a3f7b2c67915), UINT64_C(0xc67178f2e372532b),
351     UINT64_C(0xca273eceea26619c), UINT64_C(0xd186b8c721c0c207),
352     UINT64_C(0xeada7dd6cde0eb1e), UINT64_C(0xf57d4f7fee6ed178),
353     UINT64_C(0x06f067aa72176fba), UINT64_C(0x0a637dc5a2c898a6),
354     UINT64_C(0x113f9804bef90dae), UINT64_C(0x1b710b35131c471b),
355     UINT64_C(0x28db77f523047d84), UINT64_C(0x32caab7b40c72493),
356     UINT64_C(0x3c9ebe0a15c9bebc), UINT64_C(0x431d67c49c100d4c),
357     UINT64_C(0x4cc5d4becb3e42b6), UINT64_C(0x597f299cfc657e2a),
358     UINT64_C(0x5fcb6fab3ad6faec), UINT64_C(0x6c44198c4a475817),
359 };
360 
361 #if defined(__GNUC__) && __GNUC__ >= 2 && !defined(OPENSSL_NO_ASM)
362 #if defined(__x86_64) || defined(__x86_64__)
363 #define ROTR(a, n)                                              \
364   ({                                                            \
365     uint64_t ret;                                               \
366     __asm__("rorq %1, %0" : "=r"(ret) : "J"(n), "0"(a) : "cc"); \
367     ret;                                                        \
368   })
369 #define PULL64(x)                                \
370   ({                                             \
371     uint64_t ret = *((const uint64_t *)(&(x)));  \
372     __asm__("bswapq %0" : "=r"(ret) : "0"(ret)); \
373     ret;                                         \
374   })
375 #elif(defined(__i386) || defined(__i386__))
376 #define PULL64(x)                                                             \
377   ({                                                                          \
378     const unsigned int *p = (const unsigned int *)(&(x));                     \
379     unsigned int hi = p[0], lo = p[1];                                        \
380     __asm__("bswapl %0; bswapl %1;" : "=r"(lo), "=r"(hi) : "0"(lo), "1"(hi)); \
381     ((uint64_t)hi) << 32 | lo;                                                \
382   })
383 #elif(defined(_ARCH_PPC) && defined(__64BIT__)) || defined(_ARCH_PPC64)
384 #define ROTR(a, n)                                             \
385   ({                                                           \
386     uint64_t ret;                                              \
387     __asm__("rotrdi %0, %1, %2" : "=r"(ret) : "r"(a), "K"(n)); \
388     ret;                                                       \
389   })
390 #elif defined(__aarch64__)
391 #define ROTR(a, n)                                          \
392   ({                                                        \
393     uint64_t ret;                                           \
394     __asm__("ror %0, %1, %2" : "=r"(ret) : "r"(a), "I"(n)); \
395     ret;                                                    \
396   })
397 #if defined(__BYTE_ORDER__) && defined(__ORDER_LITTLE_ENDIAN__) && \
398     __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__
399 #define PULL64(x)                                                         \
400   ({                                                                      \
401     uint64_t ret;                                                         \
402     __asm__("rev %0, %1" : "=r"(ret) : "r"(*((const uint64_t *)(&(x))))); \
403     ret;                                                                  \
404   })
405 #endif
406 #endif
407 #elif defined(_MSC_VER)
408 #if defined(_WIN64) /* applies to both IA-64 and AMD64 */
409 #pragma intrinsic(_rotr64)
410 #define ROTR(a, n) _rotr64((a), n)
411 #endif
412 #if defined(_M_IX86) && !defined(OPENSSL_NO_ASM)
__pull64be(const void * x)413 static uint64_t __fastcall __pull64be(const void *x) {
414   _asm mov edx, [ecx + 0]
415   _asm mov eax, [ecx + 4]
416   _asm bswap edx
417   _asm bswap eax
418 }
419 #define PULL64(x) __pull64be(&(x))
420 #if _MSC_VER <= 1200
421 #pragma inline_depth(0)
422 #endif
423 #endif
424 #endif
425 
426 #ifndef PULL64
427 #define B(x, j) \
428   (((uint64_t)(*(((const uint8_t *)(&x)) + j))) << ((7 - j) * 8))
429 #define PULL64(x)                                                        \
430   (B(x, 0) | B(x, 1) | B(x, 2) | B(x, 3) | B(x, 4) | B(x, 5) | B(x, 6) | \
431    B(x, 7))
432 #endif
433 
434 #ifndef ROTR
435 #define ROTR(x, s) (((x) >> s) | (x) << (64 - s))
436 #endif
437 
438 #define Sigma0(x) (ROTR((x), 28) ^ ROTR((x), 34) ^ ROTR((x), 39))
439 #define Sigma1(x) (ROTR((x), 14) ^ ROTR((x), 18) ^ ROTR((x), 41))
440 #define sigma0(x) (ROTR((x), 1) ^ ROTR((x), 8) ^ ((x) >> 7))
441 #define sigma1(x) (ROTR((x), 19) ^ ROTR((x), 61) ^ ((x) >> 6))
442 
443 #define Ch(x, y, z) (((x) & (y)) ^ ((~(x)) & (z)))
444 #define Maj(x, y, z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z)))
445 
446 
447 #if defined(__i386) || defined(__i386__) || defined(_M_IX86)
448 /*
449  * This code should give better results on 32-bit CPU with less than
450  * ~24 registers, both size and performance wise...
451  */
sha512_block_data_order(uint64_t * state,const uint64_t * W,size_t num)452 static void sha512_block_data_order(uint64_t *state, const uint64_t *W,
453                                     size_t num) {
454   uint64_t A, E, T;
455   uint64_t X[9 + 80], *F;
456   int i;
457 
458   while (num--) {
459     F = X + 80;
460     A = state[0];
461     F[1] = state[1];
462     F[2] = state[2];
463     F[3] = state[3];
464     E = state[4];
465     F[5] = state[5];
466     F[6] = state[6];
467     F[7] = state[7];
468 
469     for (i = 0; i < 16; i++, F--) {
470       T = PULL64(W[i]);
471       F[0] = A;
472       F[4] = E;
473       F[8] = T;
474       T += F[7] + Sigma1(E) + Ch(E, F[5], F[6]) + K512[i];
475       E = F[3] + T;
476       A = T + Sigma0(A) + Maj(A, F[1], F[2]);
477     }
478 
479     for (; i < 80; i++, F--) {
480       T = sigma0(F[8 + 16 - 1]);
481       T += sigma1(F[8 + 16 - 14]);
482       T += F[8 + 16] + F[8 + 16 - 9];
483 
484       F[0] = A;
485       F[4] = E;
486       F[8] = T;
487       T += F[7] + Sigma1(E) + Ch(E, F[5], F[6]) + K512[i];
488       E = F[3] + T;
489       A = T + Sigma0(A) + Maj(A, F[1], F[2]);
490     }
491 
492     state[0] += A;
493     state[1] += F[1];
494     state[2] += F[2];
495     state[3] += F[3];
496     state[4] += E;
497     state[5] += F[5];
498     state[6] += F[6];
499     state[7] += F[7];
500 
501     W += 16;
502   }
503 }
504 
505 #else
506 
507 #define ROUND_00_15(i, a, b, c, d, e, f, g, h)   \
508   do {                                           \
509     T1 += h + Sigma1(e) + Ch(e, f, g) + K512[i]; \
510     h = Sigma0(a) + Maj(a, b, c);                \
511     d += T1;                                     \
512     h += T1;                                     \
513   } while (0)
514 
515 #define ROUND_16_80(i, j, a, b, c, d, e, f, g, h, X)   \
516   do {                                                 \
517     s0 = X[(j + 1) & 0x0f];                            \
518     s0 = sigma0(s0);                                   \
519     s1 = X[(j + 14) & 0x0f];                           \
520     s1 = sigma1(s1);                                   \
521     T1 = X[(j) & 0x0f] += s0 + s1 + X[(j + 9) & 0x0f]; \
522     ROUND_00_15(i + j, a, b, c, d, e, f, g, h);        \
523   } while (0)
524 
sha512_block_data_order(uint64_t * state,const uint64_t * W,size_t num)525 static void sha512_block_data_order(uint64_t *state, const uint64_t *W,
526                                     size_t num) {
527   uint64_t a, b, c, d, e, f, g, h, s0, s1, T1;
528   uint64_t X[16];
529   int i;
530 
531   while (num--) {
532 
533     a = state[0];
534     b = state[1];
535     c = state[2];
536     d = state[3];
537     e = state[4];
538     f = state[5];
539     g = state[6];
540     h = state[7];
541 
542     T1 = X[0] = PULL64(W[0]);
543     ROUND_00_15(0, a, b, c, d, e, f, g, h);
544     T1 = X[1] = PULL64(W[1]);
545     ROUND_00_15(1, h, a, b, c, d, e, f, g);
546     T1 = X[2] = PULL64(W[2]);
547     ROUND_00_15(2, g, h, a, b, c, d, e, f);
548     T1 = X[3] = PULL64(W[3]);
549     ROUND_00_15(3, f, g, h, a, b, c, d, e);
550     T1 = X[4] = PULL64(W[4]);
551     ROUND_00_15(4, e, f, g, h, a, b, c, d);
552     T1 = X[5] = PULL64(W[5]);
553     ROUND_00_15(5, d, e, f, g, h, a, b, c);
554     T1 = X[6] = PULL64(W[6]);
555     ROUND_00_15(6, c, d, e, f, g, h, a, b);
556     T1 = X[7] = PULL64(W[7]);
557     ROUND_00_15(7, b, c, d, e, f, g, h, a);
558     T1 = X[8] = PULL64(W[8]);
559     ROUND_00_15(8, a, b, c, d, e, f, g, h);
560     T1 = X[9] = PULL64(W[9]);
561     ROUND_00_15(9, h, a, b, c, d, e, f, g);
562     T1 = X[10] = PULL64(W[10]);
563     ROUND_00_15(10, g, h, a, b, c, d, e, f);
564     T1 = X[11] = PULL64(W[11]);
565     ROUND_00_15(11, f, g, h, a, b, c, d, e);
566     T1 = X[12] = PULL64(W[12]);
567     ROUND_00_15(12, e, f, g, h, a, b, c, d);
568     T1 = X[13] = PULL64(W[13]);
569     ROUND_00_15(13, d, e, f, g, h, a, b, c);
570     T1 = X[14] = PULL64(W[14]);
571     ROUND_00_15(14, c, d, e, f, g, h, a, b);
572     T1 = X[15] = PULL64(W[15]);
573     ROUND_00_15(15, b, c, d, e, f, g, h, a);
574 
575     for (i = 16; i < 80; i += 16) {
576       ROUND_16_80(i, 0, a, b, c, d, e, f, g, h, X);
577       ROUND_16_80(i, 1, h, a, b, c, d, e, f, g, X);
578       ROUND_16_80(i, 2, g, h, a, b, c, d, e, f, X);
579       ROUND_16_80(i, 3, f, g, h, a, b, c, d, e, X);
580       ROUND_16_80(i, 4, e, f, g, h, a, b, c, d, X);
581       ROUND_16_80(i, 5, d, e, f, g, h, a, b, c, X);
582       ROUND_16_80(i, 6, c, d, e, f, g, h, a, b, X);
583       ROUND_16_80(i, 7, b, c, d, e, f, g, h, a, X);
584       ROUND_16_80(i, 8, a, b, c, d, e, f, g, h, X);
585       ROUND_16_80(i, 9, h, a, b, c, d, e, f, g, X);
586       ROUND_16_80(i, 10, g, h, a, b, c, d, e, f, X);
587       ROUND_16_80(i, 11, f, g, h, a, b, c, d, e, X);
588       ROUND_16_80(i, 12, e, f, g, h, a, b, c, d, X);
589       ROUND_16_80(i, 13, d, e, f, g, h, a, b, c, X);
590       ROUND_16_80(i, 14, c, d, e, f, g, h, a, b, X);
591       ROUND_16_80(i, 15, b, c, d, e, f, g, h, a, X);
592     }
593 
594     state[0] += a;
595     state[1] += b;
596     state[2] += c;
597     state[3] += d;
598     state[4] += e;
599     state[5] += f;
600     state[6] += g;
601     state[7] += h;
602 
603     W += 16;
604   }
605 }
606 
607 #endif
608 
609 #endif /* SHA512_ASM */
610