Use IFUNC on x86-64 memset
[glibc.git] / crypt / sha512.c
blob02127476fe21c2fd914bf06007754a887feaca64
1 /* Functions to compute SHA512 message digest of files or memory blocks.
2 according to the definition of SHA512 in FIPS 180-2.
3 Copyright (C) 2007 Free Software Foundation, Inc.
4 This file is part of the GNU C Library.
6 The GNU C Library is free software; you can redistribute it and/or
7 modify it under the terms of the GNU Lesser General Public
8 License as published by the Free Software Foundation; either
9 version 2.1 of the License, or (at your option) any later version.
11 The GNU C Library is distributed in the hope that it will be useful,
12 but WITHOUT ANY WARRANTY; without even the implied warranty of
13 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
14 Lesser General Public License for more details.
16 You should have received a copy of the GNU Lesser General Public
17 License along with the GNU C Library; if not, write to the Free
18 Software Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA
19 02111-1307 USA. */
21 /* Written by Ulrich Drepper <drepper@redhat.com>, 2007. */
23 #ifdef HAVE_CONFIG_H
24 # include <config.h>
25 #endif
27 #include <endian.h>
28 #include <stdlib.h>
29 #include <string.h>
30 #include <sys/types.h>
32 #include "sha512.h"
34 #if __BYTE_ORDER == __LITTLE_ENDIAN
35 # ifdef _LIBC
36 # include <byteswap.h>
37 # define SWAP(n) bswap_64 (n)
38 # else
39 # define SWAP(n) \
40 (((n) << 56) \
41 | (((n) & 0xff00) << 40) \
42 | (((n) & 0xff0000) << 24) \
43 | (((n) & 0xff000000) << 8) \
44 | (((n) >> 8) & 0xff000000) \
45 | (((n) >> 24) & 0xff0000) \
46 | (((n) >> 40) & 0xff00) \
47 | ((n) >> 56))
48 # endif
49 #else
50 # define SWAP(n) (n)
51 #endif
54 /* This array contains the bytes used to pad the buffer to the next
55 64-byte boundary. (FIPS 180-2:5.1.2) */
56 static const unsigned char fillbuf[128] = { 0x80, 0 /* , 0, 0, ... */ };
59 /* Constants for SHA512 from FIPS 180-2:4.2.3. */
60 static const uint64_t K[80] =
62 UINT64_C (0x428a2f98d728ae22), UINT64_C (0x7137449123ef65cd),
63 UINT64_C (0xb5c0fbcfec4d3b2f), UINT64_C (0xe9b5dba58189dbbc),
64 UINT64_C (0x3956c25bf348b538), UINT64_C (0x59f111f1b605d019),
65 UINT64_C (0x923f82a4af194f9b), UINT64_C (0xab1c5ed5da6d8118),
66 UINT64_C (0xd807aa98a3030242), UINT64_C (0x12835b0145706fbe),
67 UINT64_C (0x243185be4ee4b28c), UINT64_C (0x550c7dc3d5ffb4e2),
68 UINT64_C (0x72be5d74f27b896f), UINT64_C (0x80deb1fe3b1696b1),
69 UINT64_C (0x9bdc06a725c71235), UINT64_C (0xc19bf174cf692694),
70 UINT64_C (0xe49b69c19ef14ad2), UINT64_C (0xefbe4786384f25e3),
71 UINT64_C (0x0fc19dc68b8cd5b5), UINT64_C (0x240ca1cc77ac9c65),
72 UINT64_C (0x2de92c6f592b0275), UINT64_C (0x4a7484aa6ea6e483),
73 UINT64_C (0x5cb0a9dcbd41fbd4), UINT64_C (0x76f988da831153b5),
74 UINT64_C (0x983e5152ee66dfab), UINT64_C (0xa831c66d2db43210),
75 UINT64_C (0xb00327c898fb213f), UINT64_C (0xbf597fc7beef0ee4),
76 UINT64_C (0xc6e00bf33da88fc2), UINT64_C (0xd5a79147930aa725),
77 UINT64_C (0x06ca6351e003826f), UINT64_C (0x142929670a0e6e70),
78 UINT64_C (0x27b70a8546d22ffc), UINT64_C (0x2e1b21385c26c926),
79 UINT64_C (0x4d2c6dfc5ac42aed), UINT64_C (0x53380d139d95b3df),
80 UINT64_C (0x650a73548baf63de), UINT64_C (0x766a0abb3c77b2a8),
81 UINT64_C (0x81c2c92e47edaee6), UINT64_C (0x92722c851482353b),
82 UINT64_C (0xa2bfe8a14cf10364), UINT64_C (0xa81a664bbc423001),
83 UINT64_C (0xc24b8b70d0f89791), UINT64_C (0xc76c51a30654be30),
84 UINT64_C (0xd192e819d6ef5218), UINT64_C (0xd69906245565a910),
85 UINT64_C (0xf40e35855771202a), UINT64_C (0x106aa07032bbd1b8),
86 UINT64_C (0x19a4c116b8d2d0c8), UINT64_C (0x1e376c085141ab53),
87 UINT64_C (0x2748774cdf8eeb99), UINT64_C (0x34b0bcb5e19b48a8),
88 UINT64_C (0x391c0cb3c5c95a63), UINT64_C (0x4ed8aa4ae3418acb),
89 UINT64_C (0x5b9cca4f7763e373), UINT64_C (0x682e6ff3d6b2b8a3),
90 UINT64_C (0x748f82ee5defb2fc), UINT64_C (0x78a5636f43172f60),
91 UINT64_C (0x84c87814a1f0ab72), UINT64_C (0x8cc702081a6439ec),
92 UINT64_C (0x90befffa23631e28), UINT64_C (0xa4506cebde82bde9),
93 UINT64_C (0xbef9a3f7b2c67915), UINT64_C (0xc67178f2e372532b),
94 UINT64_C (0xca273eceea26619c), UINT64_C (0xd186b8c721c0c207),
95 UINT64_C (0xeada7dd6cde0eb1e), UINT64_C (0xf57d4f7fee6ed178),
96 UINT64_C (0x06f067aa72176fba), UINT64_C (0x0a637dc5a2c898a6),
97 UINT64_C (0x113f9804bef90dae), UINT64_C (0x1b710b35131c471b),
98 UINT64_C (0x28db77f523047d84), UINT64_C (0x32caab7b40c72493),
99 UINT64_C (0x3c9ebe0a15c9bebc), UINT64_C (0x431d67c49c100d4c),
100 UINT64_C (0x4cc5d4becb3e42b6), UINT64_C (0x597f299cfc657e2a),
101 UINT64_C (0x5fcb6fab3ad6faec), UINT64_C (0x6c44198c4a475817)
105 /* Process LEN bytes of BUFFER, accumulating context into CTX.
106 It is assumed that LEN % 128 == 0. */
107 static void
108 sha512_process_block (const void *buffer, size_t len, struct sha512_ctx *ctx)
110 const uint64_t *words = buffer;
111 size_t nwords = len / sizeof (uint64_t);
112 uint64_t a = ctx->H[0];
113 uint64_t b = ctx->H[1];
114 uint64_t c = ctx->H[2];
115 uint64_t d = ctx->H[3];
116 uint64_t e = ctx->H[4];
117 uint64_t f = ctx->H[5];
118 uint64_t g = ctx->H[6];
119 uint64_t h = ctx->H[7];
121 /* First increment the byte count. FIPS 180-2 specifies the possible
122 length of the file up to 2^128 bits. Here we only compute the
123 number of bytes. Do a double word increment. */
124 ctx->total[0] += len;
125 if (ctx->total[0] < len)
126 ++ctx->total[1];
128 /* Process all bytes in the buffer with 128 bytes in each round of
129 the loop. */
130 while (nwords > 0)
132 uint64_t W[80];
133 uint64_t a_save = a;
134 uint64_t b_save = b;
135 uint64_t c_save = c;
136 uint64_t d_save = d;
137 uint64_t e_save = e;
138 uint64_t f_save = f;
139 uint64_t g_save = g;
140 uint64_t h_save = h;
142 /* Operators defined in FIPS 180-2:4.1.2. */
143 #define Ch(x, y, z) ((x & y) ^ (~x & z))
144 #define Maj(x, y, z) ((x & y) ^ (x & z) ^ (y & z))
145 #define S0(x) (CYCLIC (x, 28) ^ CYCLIC (x, 34) ^ CYCLIC (x, 39))
146 #define S1(x) (CYCLIC (x, 14) ^ CYCLIC (x, 18) ^ CYCLIC (x, 41))
147 #define R0(x) (CYCLIC (x, 1) ^ CYCLIC (x, 8) ^ (x >> 7))
148 #define R1(x) (CYCLIC (x, 19) ^ CYCLIC (x, 61) ^ (x >> 6))
150 /* It is unfortunate that C does not provide an operator for
151 cyclic rotation. Hope the C compiler is smart enough. */
152 #define CYCLIC(w, s) ((w >> s) | (w << (64 - s)))
154 /* Compute the message schedule according to FIPS 180-2:6.3.2 step 2. */
155 for (unsigned int t = 0; t < 16; ++t)
157 W[t] = SWAP (*words);
158 ++words;
160 for (unsigned int t = 16; t < 80; ++t)
161 W[t] = R1 (W[t - 2]) + W[t - 7] + R0 (W[t - 15]) + W[t - 16];
163 /* The actual computation according to FIPS 180-2:6.3.2 step 3. */
164 for (unsigned int t = 0; t < 80; ++t)
166 uint64_t T1 = h + S1 (e) + Ch (e, f, g) + K[t] + W[t];
167 uint64_t T2 = S0 (a) + Maj (a, b, c);
168 h = g;
169 g = f;
170 f = e;
171 e = d + T1;
172 d = c;
173 c = b;
174 b = a;
175 a = T1 + T2;
178 /* Add the starting values of the context according to FIPS 180-2:6.3.2
179 step 4. */
180 a += a_save;
181 b += b_save;
182 c += c_save;
183 d += d_save;
184 e += e_save;
185 f += f_save;
186 g += g_save;
187 h += h_save;
189 /* Prepare for the next round. */
190 nwords -= 16;
193 /* Put checksum in context given as argument. */
194 ctx->H[0] = a;
195 ctx->H[1] = b;
196 ctx->H[2] = c;
197 ctx->H[3] = d;
198 ctx->H[4] = e;
199 ctx->H[5] = f;
200 ctx->H[6] = g;
201 ctx->H[7] = h;
205 /* Initialize structure containing state of computation.
206 (FIPS 180-2:5.3.3) */
207 void
208 __sha512_init_ctx (ctx)
209 struct sha512_ctx *ctx;
211 ctx->H[0] = UINT64_C (0x6a09e667f3bcc908);
212 ctx->H[1] = UINT64_C (0xbb67ae8584caa73b);
213 ctx->H[2] = UINT64_C (0x3c6ef372fe94f82b);
214 ctx->H[3] = UINT64_C (0xa54ff53a5f1d36f1);
215 ctx->H[4] = UINT64_C (0x510e527fade682d1);
216 ctx->H[5] = UINT64_C (0x9b05688c2b3e6c1f);
217 ctx->H[6] = UINT64_C (0x1f83d9abfb41bd6b);
218 ctx->H[7] = UINT64_C (0x5be0cd19137e2179);
220 ctx->total[0] = ctx->total[1] = 0;
221 ctx->buflen = 0;
225 /* Process the remaining bytes in the internal buffer and the usual
226 prolog according to the standard and write the result to RESBUF.
228 IMPORTANT: On some systems it is required that RESBUF is correctly
229 aligned for a 32 bits value. */
230 void *
231 __sha512_finish_ctx (ctx, resbuf)
232 struct sha512_ctx *ctx;
233 void *resbuf;
235 /* Take yet unprocessed bytes into account. */
236 uint64_t bytes = ctx->buflen;
237 size_t pad;
239 /* Now count remaining bytes. */
240 ctx->total[0] += bytes;
241 if (ctx->total[0] < bytes)
242 ++ctx->total[1];
244 pad = bytes >= 112 ? 128 + 112 - bytes : 112 - bytes;
245 memcpy (&ctx->buffer[bytes], fillbuf, pad);
247 /* Put the 128-bit file length in *bits* at the end of the buffer. */
248 *(uint64_t *) &ctx->buffer[bytes + pad + 8] = SWAP (ctx->total[0] << 3);
249 *(uint64_t *) &ctx->buffer[bytes + pad] = SWAP ((ctx->total[1] << 3) |
250 (ctx->total[0] >> 61));
252 /* Process last bytes. */
253 sha512_process_block (ctx->buffer, bytes + pad + 16, ctx);
255 /* Put result from CTX in first 64 bytes following RESBUF. */
256 for (unsigned int i = 0; i < 8; ++i)
257 ((uint64_t *) resbuf)[i] = SWAP (ctx->H[i]);
259 return resbuf;
263 void
264 __sha512_process_bytes (buffer, len, ctx)
265 const void *buffer;
266 size_t len;
267 struct sha512_ctx *ctx;
269 /* When we already have some bits in our internal buffer concatenate
270 both inputs first. */
271 if (ctx->buflen != 0)
273 size_t left_over = ctx->buflen;
274 size_t add = 256 - left_over > len ? len : 256 - left_over;
276 memcpy (&ctx->buffer[left_over], buffer, add);
277 ctx->buflen += add;
279 if (ctx->buflen > 128)
281 sha512_process_block (ctx->buffer, ctx->buflen & ~127, ctx);
283 ctx->buflen &= 127;
284 /* The regions in the following copy operation cannot overlap. */
285 memcpy (ctx->buffer, &ctx->buffer[(left_over + add) & ~127],
286 ctx->buflen);
289 buffer = (const char *) buffer + add;
290 len -= add;
293 /* Process available complete blocks. */
294 if (len >= 128)
296 #if !_STRING_ARCH_unaligned
297 /* To check alignment gcc has an appropriate operator. Other
298 compilers don't. */
299 # if __GNUC__ >= 2
300 # define UNALIGNED_P(p) (((uintptr_t) p) % __alignof__ (uint64_t) != 0)
301 # else
302 # define UNALIGNED_P(p) (((uintptr_t) p) % sizeof (uint64_t) != 0)
303 # endif
304 if (UNALIGNED_P (buffer))
305 while (len > 128)
307 sha512_process_block (memcpy (ctx->buffer, buffer, 128), 128,
308 ctx);
309 buffer = (const char *) buffer + 128;
310 len -= 128;
312 else
313 #endif
315 sha512_process_block (buffer, len & ~127, ctx);
316 buffer = (const char *) buffer + (len & ~127);
317 len &= 127;
321 /* Move remaining bytes into internal buffer. */
322 if (len > 0)
324 size_t left_over = ctx->buflen;
326 memcpy (&ctx->buffer[left_over], buffer, len);
327 left_over += len;
328 if (left_over >= 128)
330 sha512_process_block (ctx->buffer, 128, ctx);
331 left_over -= 128;
332 memcpy (ctx->buffer, &ctx->buffer[128], left_over);
334 ctx->buflen = left_over;