crypto: sha512 - Avoid stack bloat on i386
[linux-2.6.git] / crypto / sha512_generic.c
1 /* SHA-512 code by Jean-Luc Cooke <jlcooke@certainkey.com>
2  *
3  * Copyright (c) Jean-Luc Cooke <jlcooke@certainkey.com>
4  * Copyright (c) Andrew McDonald <andrew@mcdonald.org.uk>
5  * Copyright (c) 2003 Kyle McMartin <kyle@debian.org>
6  *
7  * This program is free software; you can redistribute it and/or modify it
8  * under the terms of the GNU General Public License as published by the
9  * Free Software Foundation; either version 2, or (at your option) any
10  * later version.
11  *
12  */
13 #include <crypto/internal/hash.h>
14 #include <linux/kernel.h>
15 #include <linux/module.h>
16 #include <linux/mm.h>
17 #include <linux/init.h>
18 #include <linux/crypto.h>
19 #include <linux/types.h>
20 #include <crypto/sha.h>
21 #include <linux/percpu.h>
22 #include <asm/byteorder.h>
23
24 static inline u64 Ch(u64 x, u64 y, u64 z)
25 {
26         return z ^ (x & (y ^ z));
27 }
28
29 static inline u64 Maj(u64 x, u64 y, u64 z)
30 {
31         return (x & y) | (z & (x | y));
32 }
33
34 static inline u64 RORu64(u64 x, u64 y)
35 {
36         return (x >> y) | (x << (64 - y));
37 }
38
39 static const u64 sha512_K[80] = {
40         0x428a2f98d728ae22ULL, 0x7137449123ef65cdULL, 0xb5c0fbcfec4d3b2fULL,
41         0xe9b5dba58189dbbcULL, 0x3956c25bf348b538ULL, 0x59f111f1b605d019ULL,
42         0x923f82a4af194f9bULL, 0xab1c5ed5da6d8118ULL, 0xd807aa98a3030242ULL,
43         0x12835b0145706fbeULL, 0x243185be4ee4b28cULL, 0x550c7dc3d5ffb4e2ULL,
44         0x72be5d74f27b896fULL, 0x80deb1fe3b1696b1ULL, 0x9bdc06a725c71235ULL,
45         0xc19bf174cf692694ULL, 0xe49b69c19ef14ad2ULL, 0xefbe4786384f25e3ULL,
46         0x0fc19dc68b8cd5b5ULL, 0x240ca1cc77ac9c65ULL, 0x2de92c6f592b0275ULL,
47         0x4a7484aa6ea6e483ULL, 0x5cb0a9dcbd41fbd4ULL, 0x76f988da831153b5ULL,
48         0x983e5152ee66dfabULL, 0xa831c66d2db43210ULL, 0xb00327c898fb213fULL,
49         0xbf597fc7beef0ee4ULL, 0xc6e00bf33da88fc2ULL, 0xd5a79147930aa725ULL,
50         0x06ca6351e003826fULL, 0x142929670a0e6e70ULL, 0x27b70a8546d22ffcULL,
51         0x2e1b21385c26c926ULL, 0x4d2c6dfc5ac42aedULL, 0x53380d139d95b3dfULL,
52         0x650a73548baf63deULL, 0x766a0abb3c77b2a8ULL, 0x81c2c92e47edaee6ULL,
53         0x92722c851482353bULL, 0xa2bfe8a14cf10364ULL, 0xa81a664bbc423001ULL,
54         0xc24b8b70d0f89791ULL, 0xc76c51a30654be30ULL, 0xd192e819d6ef5218ULL,
55         0xd69906245565a910ULL, 0xf40e35855771202aULL, 0x106aa07032bbd1b8ULL,
56         0x19a4c116b8d2d0c8ULL, 0x1e376c085141ab53ULL, 0x2748774cdf8eeb99ULL,
57         0x34b0bcb5e19b48a8ULL, 0x391c0cb3c5c95a63ULL, 0x4ed8aa4ae3418acbULL,
58         0x5b9cca4f7763e373ULL, 0x682e6ff3d6b2b8a3ULL, 0x748f82ee5defb2fcULL,
59         0x78a5636f43172f60ULL, 0x84c87814a1f0ab72ULL, 0x8cc702081a6439ecULL,
60         0x90befffa23631e28ULL, 0xa4506cebde82bde9ULL, 0xbef9a3f7b2c67915ULL,
61         0xc67178f2e372532bULL, 0xca273eceea26619cULL, 0xd186b8c721c0c207ULL,
62         0xeada7dd6cde0eb1eULL, 0xf57d4f7fee6ed178ULL, 0x06f067aa72176fbaULL,
63         0x0a637dc5a2c898a6ULL, 0x113f9804bef90daeULL, 0x1b710b35131c471bULL,
64         0x28db77f523047d84ULL, 0x32caab7b40c72493ULL, 0x3c9ebe0a15c9bebcULL,
65         0x431d67c49c100d4cULL, 0x4cc5d4becb3e42b6ULL, 0x597f299cfc657e2aULL,
66         0x5fcb6fab3ad6faecULL, 0x6c44198c4a475817ULL,
67 };
68
69 #define e0(x)       (RORu64(x,28) ^ RORu64(x,34) ^ RORu64(x,39))
70 #define e1(x)       (RORu64(x,14) ^ RORu64(x,18) ^ RORu64(x,41))
71 #define s0(x)       (RORu64(x, 1) ^ RORu64(x, 8) ^ (x >> 7))
72 #define s1(x)       (RORu64(x,19) ^ RORu64(x,61) ^ (x >> 6))
73
74 static inline void LOAD_OP(int I, u64 *W, const u8 *input)
75 {
76         W[I] = __be64_to_cpu( ((__be64*)(input))[I] );
77 }
78
79 static inline void BLEND_OP(int I, u64 *W)
80 {
81         W[I & 15] += s1(W[(I-2) & 15]) + W[(I-7) & 15] + s0(W[(I-15) & 15]);
82 }
83
84 static void
85 sha512_transform(u64 *state, const u8 *input)
86 {
87         u64 a, b, c, d, e, f, g, h, t1, t2;
88
89         int i;
90         u64 W[16];
91
92         /* load the state into our registers */
93         a=state[0];   b=state[1];   c=state[2];   d=state[3];
94         e=state[4];   f=state[5];   g=state[6];   h=state[7];
95
96         /* now iterate */
97         for (i=0; i<80; i+=8) {
98                 if (!(i & 8)) {
99                         int j;
100
101                         if (i < 16) {
102                                 /* load the input */
103                                 for (j = 0; j < 16; j++)
104                                         LOAD_OP(i + j, W, input);
105                         } else {
106                                 for (j = 0; j < 16; j++) {
107                                         BLEND_OP(i + j, W);
108                                 }
109                         }
110                 }
111
112                 t1 = h + e1(e) + Ch(e,f,g) + sha512_K[i  ] + W[(i & 15)];
113                 t2 = e0(a) + Maj(a,b,c);    d+=t1;    h=t1+t2;
114                 t1 = g + e1(d) + Ch(d,e,f) + sha512_K[i+1] + W[(i & 15) + 1];
115                 t2 = e0(h) + Maj(h,a,b);    c+=t1;    g=t1+t2;
116                 t1 = f + e1(c) + Ch(c,d,e) + sha512_K[i+2] + W[(i & 15) + 2];
117                 t2 = e0(g) + Maj(g,h,a);    b+=t1;    f=t1+t2;
118                 t1 = e + e1(b) + Ch(b,c,d) + sha512_K[i+3] + W[(i & 15) + 3];
119                 t2 = e0(f) + Maj(f,g,h);    a+=t1;    e=t1+t2;
120                 t1 = d + e1(a) + Ch(a,b,c) + sha512_K[i+4] + W[(i & 15) + 4];
121                 t2 = e0(e) + Maj(e,f,g);    h+=t1;    d=t1+t2;
122                 t1 = c + e1(h) + Ch(h,a,b) + sha512_K[i+5] + W[(i & 15) + 5];
123                 t2 = e0(d) + Maj(d,e,f);    g+=t1;    c=t1+t2;
124                 t1 = b + e1(g) + Ch(g,h,a) + sha512_K[i+6] + W[(i & 15) + 6];
125                 t2 = e0(c) + Maj(c,d,e);    f+=t1;    b=t1+t2;
126                 t1 = a + e1(f) + Ch(f,g,h) + sha512_K[i+7] + W[(i & 15) + 7];
127                 t2 = e0(b) + Maj(b,c,d);    e+=t1;    a=t1+t2;
128         }
129
130         state[0] += a; state[1] += b; state[2] += c; state[3] += d;
131         state[4] += e; state[5] += f; state[6] += g; state[7] += h;
132
133         /* erase our data */
134         a = b = c = d = e = f = g = h = t1 = t2 = 0;
135 }
136
137 static int
138 sha512_init(struct shash_desc *desc)
139 {
140         struct sha512_state *sctx = shash_desc_ctx(desc);
141         sctx->state[0] = SHA512_H0;
142         sctx->state[1] = SHA512_H1;
143         sctx->state[2] = SHA512_H2;
144         sctx->state[3] = SHA512_H3;
145         sctx->state[4] = SHA512_H4;
146         sctx->state[5] = SHA512_H5;
147         sctx->state[6] = SHA512_H6;
148         sctx->state[7] = SHA512_H7;
149         sctx->count[0] = sctx->count[1] = 0;
150
151         return 0;
152 }
153
154 static int
155 sha384_init(struct shash_desc *desc)
156 {
157         struct sha512_state *sctx = shash_desc_ctx(desc);
158         sctx->state[0] = SHA384_H0;
159         sctx->state[1] = SHA384_H1;
160         sctx->state[2] = SHA384_H2;
161         sctx->state[3] = SHA384_H3;
162         sctx->state[4] = SHA384_H4;
163         sctx->state[5] = SHA384_H5;
164         sctx->state[6] = SHA384_H6;
165         sctx->state[7] = SHA384_H7;
166         sctx->count[0] = sctx->count[1] = 0;
167
168         return 0;
169 }
170
171 static int
172 sha512_update(struct shash_desc *desc, const u8 *data, unsigned int len)
173 {
174         struct sha512_state *sctx = shash_desc_ctx(desc);
175
176         unsigned int i, index, part_len;
177
178         /* Compute number of bytes mod 128 */
179         index = sctx->count[0] & 0x7f;
180
181         /* Update number of bytes */
182         if (!(sctx->count[0] += len))
183                 sctx->count[1]++;
184
185         part_len = 128 - index;
186
187         /* Transform as many times as possible. */
188         if (len >= part_len) {
189                 memcpy(&sctx->buf[index], data, part_len);
190                 sha512_transform(sctx->state, sctx->buf);
191
192                 for (i = part_len; i + 127 < len; i+=128)
193                         sha512_transform(sctx->state, &data[i]);
194
195                 index = 0;
196         } else {
197                 i = 0;
198         }
199
200         /* Buffer remaining input */
201         memcpy(&sctx->buf[index], &data[i], len - i);
202
203         return 0;
204 }
205
206 static int
207 sha512_final(struct shash_desc *desc, u8 *hash)
208 {
209         struct sha512_state *sctx = shash_desc_ctx(desc);
210         static u8 padding[128] = { 0x80, };
211         __be64 *dst = (__be64 *)hash;
212         __be64 bits[2];
213         unsigned int index, pad_len;
214         int i;
215
216         /* Save number of bits */
217         bits[1] = cpu_to_be64(sctx->count[0] << 3);
218         bits[0] = cpu_to_be64(sctx->count[1] << 3 | sctx->count[0] >> 61);
219
220         /* Pad out to 112 mod 128. */
221         index = sctx->count[0] & 0x7f;
222         pad_len = (index < 112) ? (112 - index) : ((128+112) - index);
223         sha512_update(desc, padding, pad_len);
224
225         /* Append length (before padding) */
226         sha512_update(desc, (const u8 *)bits, sizeof(bits));
227
228         /* Store state in digest */
229         for (i = 0; i < 8; i++)
230                 dst[i] = cpu_to_be64(sctx->state[i]);
231
232         /* Zeroize sensitive information. */
233         memset(sctx, 0, sizeof(struct sha512_state));
234
235         return 0;
236 }
237
238 static int sha384_final(struct shash_desc *desc, u8 *hash)
239 {
240         u8 D[64];
241
242         sha512_final(desc, D);
243
244         memcpy(hash, D, 48);
245         memset(D, 0, 64);
246
247         return 0;
248 }
249
250 static struct shash_alg sha512 = {
251         .digestsize     =       SHA512_DIGEST_SIZE,
252         .init           =       sha512_init,
253         .update         =       sha512_update,
254         .final          =       sha512_final,
255         .descsize       =       sizeof(struct sha512_state),
256         .base           =       {
257                 .cra_name       =       "sha512",
258                 .cra_flags      =       CRYPTO_ALG_TYPE_SHASH,
259                 .cra_blocksize  =       SHA512_BLOCK_SIZE,
260                 .cra_module     =       THIS_MODULE,
261         }
262 };
263
264 static struct shash_alg sha384 = {
265         .digestsize     =       SHA384_DIGEST_SIZE,
266         .init           =       sha384_init,
267         .update         =       sha512_update,
268         .final          =       sha384_final,
269         .descsize       =       sizeof(struct sha512_state),
270         .base           =       {
271                 .cra_name       =       "sha384",
272                 .cra_flags      =       CRYPTO_ALG_TYPE_SHASH,
273                 .cra_blocksize  =       SHA384_BLOCK_SIZE,
274                 .cra_module     =       THIS_MODULE,
275         }
276 };
277
278 static int __init sha512_generic_mod_init(void)
279 {
280         int ret = 0;
281
282         if ((ret = crypto_register_shash(&sha384)) < 0)
283                 goto out;
284         if ((ret = crypto_register_shash(&sha512)) < 0)
285                 crypto_unregister_shash(&sha384);
286 out:
287         return ret;
288 }
289
290 static void __exit sha512_generic_mod_fini(void)
291 {
292         crypto_unregister_shash(&sha384);
293         crypto_unregister_shash(&sha512);
294 }
295
296 module_init(sha512_generic_mod_init);
297 module_exit(sha512_generic_mod_fini);
298
299 MODULE_LICENSE("GPL");
300 MODULE_DESCRIPTION("SHA-512 and SHA-384 Secure Hash Algorithms");
301
302 MODULE_ALIAS("sha384");
303 MODULE_ALIAS("sha512");