crypto: sha512 - Move message schedule W[80] to static percpu area
[pandora-kernel.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
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 struct sha512_ctx {
25         u64 state[8];
26         u32 count[4];
27         u8 buf[128];
28 };
29
30 static DEFINE_PER_CPU(u64[80], msg_schedule);
31
32 static inline u64 Ch(u64 x, u64 y, u64 z)
33 {
34         return z ^ (x & (y ^ z));
35 }
36
37 static inline u64 Maj(u64 x, u64 y, u64 z)
38 {
39         return (x & y) | (z & (x | y));
40 }
41
42 static inline u64 RORu64(u64 x, u64 y)
43 {
44         return (x >> y) | (x << (64 - y));
45 }
46
47 static const u64 sha512_K[80] = {
48         0x428a2f98d728ae22ULL, 0x7137449123ef65cdULL, 0xb5c0fbcfec4d3b2fULL,
49         0xe9b5dba58189dbbcULL, 0x3956c25bf348b538ULL, 0x59f111f1b605d019ULL,
50         0x923f82a4af194f9bULL, 0xab1c5ed5da6d8118ULL, 0xd807aa98a3030242ULL,
51         0x12835b0145706fbeULL, 0x243185be4ee4b28cULL, 0x550c7dc3d5ffb4e2ULL,
52         0x72be5d74f27b896fULL, 0x80deb1fe3b1696b1ULL, 0x9bdc06a725c71235ULL,
53         0xc19bf174cf692694ULL, 0xe49b69c19ef14ad2ULL, 0xefbe4786384f25e3ULL,
54         0x0fc19dc68b8cd5b5ULL, 0x240ca1cc77ac9c65ULL, 0x2de92c6f592b0275ULL,
55         0x4a7484aa6ea6e483ULL, 0x5cb0a9dcbd41fbd4ULL, 0x76f988da831153b5ULL,
56         0x983e5152ee66dfabULL, 0xa831c66d2db43210ULL, 0xb00327c898fb213fULL,
57         0xbf597fc7beef0ee4ULL, 0xc6e00bf33da88fc2ULL, 0xd5a79147930aa725ULL,
58         0x06ca6351e003826fULL, 0x142929670a0e6e70ULL, 0x27b70a8546d22ffcULL,
59         0x2e1b21385c26c926ULL, 0x4d2c6dfc5ac42aedULL, 0x53380d139d95b3dfULL,
60         0x650a73548baf63deULL, 0x766a0abb3c77b2a8ULL, 0x81c2c92e47edaee6ULL,
61         0x92722c851482353bULL, 0xa2bfe8a14cf10364ULL, 0xa81a664bbc423001ULL,
62         0xc24b8b70d0f89791ULL, 0xc76c51a30654be30ULL, 0xd192e819d6ef5218ULL,
63         0xd69906245565a910ULL, 0xf40e35855771202aULL, 0x106aa07032bbd1b8ULL,
64         0x19a4c116b8d2d0c8ULL, 0x1e376c085141ab53ULL, 0x2748774cdf8eeb99ULL,
65         0x34b0bcb5e19b48a8ULL, 0x391c0cb3c5c95a63ULL, 0x4ed8aa4ae3418acbULL,
66         0x5b9cca4f7763e373ULL, 0x682e6ff3d6b2b8a3ULL, 0x748f82ee5defb2fcULL,
67         0x78a5636f43172f60ULL, 0x84c87814a1f0ab72ULL, 0x8cc702081a6439ecULL,
68         0x90befffa23631e28ULL, 0xa4506cebde82bde9ULL, 0xbef9a3f7b2c67915ULL,
69         0xc67178f2e372532bULL, 0xca273eceea26619cULL, 0xd186b8c721c0c207ULL,
70         0xeada7dd6cde0eb1eULL, 0xf57d4f7fee6ed178ULL, 0x06f067aa72176fbaULL,
71         0x0a637dc5a2c898a6ULL, 0x113f9804bef90daeULL, 0x1b710b35131c471bULL,
72         0x28db77f523047d84ULL, 0x32caab7b40c72493ULL, 0x3c9ebe0a15c9bebcULL,
73         0x431d67c49c100d4cULL, 0x4cc5d4becb3e42b6ULL, 0x597f299cfc657e2aULL,
74         0x5fcb6fab3ad6faecULL, 0x6c44198c4a475817ULL,
75 };
76
77 #define e0(x)       (RORu64(x,28) ^ RORu64(x,34) ^ RORu64(x,39))
78 #define e1(x)       (RORu64(x,14) ^ RORu64(x,18) ^ RORu64(x,41))
79 #define s0(x)       (RORu64(x, 1) ^ RORu64(x, 8) ^ (x >> 7))
80 #define s1(x)       (RORu64(x,19) ^ RORu64(x,61) ^ (x >> 6))
81
82 static inline void LOAD_OP(int I, u64 *W, const u8 *input)
83 {
84         W[I] = __be64_to_cpu( ((__be64*)(input))[I] );
85 }
86
87 static inline void BLEND_OP(int I, u64 *W)
88 {
89         W[I] = s1(W[I-2]) + W[I-7] + s0(W[I-15]) + W[I-16];
90 }
91
92 static void
93 sha512_transform(u64 *state, const u8 *input)
94 {
95         u64 a, b, c, d, e, f, g, h, t1, t2;
96
97         int i;
98         u64 *W = get_cpu_var(msg_schedule);
99
100         /* load the input */
101         for (i = 0; i < 16; i++)
102                 LOAD_OP(i, W, input);
103
104         for (i = 16; i < 80; i++) {
105                 BLEND_OP(i, W);
106         }
107
108         /* load the state into our registers */
109         a=state[0];   b=state[1];   c=state[2];   d=state[3];
110         e=state[4];   f=state[5];   g=state[6];   h=state[7];
111
112         /* now iterate */
113         for (i=0; i<80; i+=8) {
114                 t1 = h + e1(e) + Ch(e,f,g) + sha512_K[i  ] + W[i  ];
115                 t2 = e0(a) + Maj(a,b,c);    d+=t1;    h=t1+t2;
116                 t1 = g + e1(d) + Ch(d,e,f) + sha512_K[i+1] + W[i+1];
117                 t2 = e0(h) + Maj(h,a,b);    c+=t1;    g=t1+t2;
118                 t1 = f + e1(c) + Ch(c,d,e) + sha512_K[i+2] + W[i+2];
119                 t2 = e0(g) + Maj(g,h,a);    b+=t1;    f=t1+t2;
120                 t1 = e + e1(b) + Ch(b,c,d) + sha512_K[i+3] + W[i+3];
121                 t2 = e0(f) + Maj(f,g,h);    a+=t1;    e=t1+t2;
122                 t1 = d + e1(a) + Ch(a,b,c) + sha512_K[i+4] + W[i+4];
123                 t2 = e0(e) + Maj(e,f,g);    h+=t1;    d=t1+t2;
124                 t1 = c + e1(h) + Ch(h,a,b) + sha512_K[i+5] + W[i+5];
125                 t2 = e0(d) + Maj(d,e,f);    g+=t1;    c=t1+t2;
126                 t1 = b + e1(g) + Ch(g,h,a) + sha512_K[i+6] + W[i+6];
127                 t2 = e0(c) + Maj(c,d,e);    f+=t1;    b=t1+t2;
128                 t1 = a + e1(f) + Ch(f,g,h) + sha512_K[i+7] + W[i+7];
129                 t2 = e0(b) + Maj(b,c,d);    e+=t1;    a=t1+t2;
130         }
131
132         state[0] += a; state[1] += b; state[2] += c; state[3] += d;
133         state[4] += e; state[5] += f; state[6] += g; state[7] += h;
134
135         /* erase our data */
136         a = b = c = d = e = f = g = h = t1 = t2 = 0;
137         memset(W, 0, sizeof(__get_cpu_var(msg_schedule)));
138         put_cpu_var(msg_schedule);
139 }
140
141 static void
142 sha512_init(struct crypto_tfm *tfm)
143 {
144         struct sha512_ctx *sctx = crypto_tfm_ctx(tfm);
145         sctx->state[0] = SHA512_H0;
146         sctx->state[1] = SHA512_H1;
147         sctx->state[2] = SHA512_H2;
148         sctx->state[3] = SHA512_H3;
149         sctx->state[4] = SHA512_H4;
150         sctx->state[5] = SHA512_H5;
151         sctx->state[6] = SHA512_H6;
152         sctx->state[7] = SHA512_H7;
153         sctx->count[0] = sctx->count[1] = sctx->count[2] = sctx->count[3] = 0;
154 }
155
156 static void
157 sha384_init(struct crypto_tfm *tfm)
158 {
159         struct sha512_ctx *sctx = crypto_tfm_ctx(tfm);
160         sctx->state[0] = SHA384_H0;
161         sctx->state[1] = SHA384_H1;
162         sctx->state[2] = SHA384_H2;
163         sctx->state[3] = SHA384_H3;
164         sctx->state[4] = SHA384_H4;
165         sctx->state[5] = SHA384_H5;
166         sctx->state[6] = SHA384_H6;
167         sctx->state[7] = SHA384_H7;
168         sctx->count[0] = sctx->count[1] = sctx->count[2] = sctx->count[3] = 0;
169 }
170
171 static void
172 sha512_update(struct crypto_tfm *tfm, const u8 *data, unsigned int len)
173 {
174         struct sha512_ctx *sctx = crypto_tfm_ctx(tfm);
175
176         unsigned int i, index, part_len;
177
178         /* Compute number of bytes mod 128 */
179         index = (unsigned int)((sctx->count[0] >> 3) & 0x7F);
180
181         /* Update number of bits */
182         if ((sctx->count[0] += (len << 3)) < (len << 3)) {
183                 if ((sctx->count[1] += 1) < 1)
184                         if ((sctx->count[2] += 1) < 1)
185                                 sctx->count[3]++;
186                 sctx->count[1] += (len >> 29);
187         }
188
189         part_len = 128 - index;
190
191         /* Transform as many times as possible. */
192         if (len >= part_len) {
193                 memcpy(&sctx->buf[index], data, part_len);
194                 sha512_transform(sctx->state, sctx->buf);
195
196                 for (i = part_len; i + 127 < len; i+=128)
197                         sha512_transform(sctx->state, &data[i]);
198
199                 index = 0;
200         } else {
201                 i = 0;
202         }
203
204         /* Buffer remaining input */
205         memcpy(&sctx->buf[index], &data[i], len - i);
206 }
207
208 static void
209 sha512_final(struct crypto_tfm *tfm, u8 *hash)
210 {
211         struct sha512_ctx *sctx = crypto_tfm_ctx(tfm);
212         static u8 padding[128] = { 0x80, };
213         __be64 *dst = (__be64 *)hash;
214         __be32 bits[4];
215         unsigned int index, pad_len;
216         int i;
217
218         /* Save number of bits */
219         bits[3] = cpu_to_be32(sctx->count[0]);
220         bits[2] = cpu_to_be32(sctx->count[1]);
221         bits[1] = cpu_to_be32(sctx->count[2]);
222         bits[0] = cpu_to_be32(sctx->count[3]);
223
224         /* Pad out to 112 mod 128. */
225         index = (sctx->count[0] >> 3) & 0x7f;
226         pad_len = (index < 112) ? (112 - index) : ((128+112) - index);
227         sha512_update(tfm, padding, pad_len);
228
229         /* Append length (before padding) */
230         sha512_update(tfm, (const u8 *)bits, sizeof(bits));
231
232         /* Store state in digest */
233         for (i = 0; i < 8; i++)
234                 dst[i] = cpu_to_be64(sctx->state[i]);
235
236         /* Zeroize sensitive information. */
237         memset(sctx, 0, sizeof(struct sha512_ctx));
238 }
239
240 static void sha384_final(struct crypto_tfm *tfm, u8 *hash)
241 {
242         u8 D[64];
243
244         sha512_final(tfm, D);
245
246         memcpy(hash, D, 48);
247         memset(D, 0, 64);
248 }
249
250 static struct crypto_alg sha512 = {
251         .cra_name       = "sha512",
252         .cra_flags      = CRYPTO_ALG_TYPE_DIGEST,
253         .cra_blocksize  = SHA512_BLOCK_SIZE,
254         .cra_ctxsize    = sizeof(struct sha512_ctx),
255         .cra_module     = THIS_MODULE,
256         .cra_alignmask  = 3,
257         .cra_list       = LIST_HEAD_INIT(sha512.cra_list),
258         .cra_u          = { .digest = {
259                                 .dia_digestsize = SHA512_DIGEST_SIZE,
260                                 .dia_init       = sha512_init,
261                                 .dia_update     = sha512_update,
262                                 .dia_final      = sha512_final }
263         }
264 };
265
266 static struct crypto_alg sha384 = {
267         .cra_name       = "sha384",
268         .cra_flags      = CRYPTO_ALG_TYPE_DIGEST,
269         .cra_blocksize  = SHA384_BLOCK_SIZE,
270         .cra_ctxsize    = sizeof(struct sha512_ctx),
271         .cra_alignmask  = 3,
272         .cra_module     = THIS_MODULE,
273         .cra_list       = LIST_HEAD_INIT(sha384.cra_list),
274         .cra_u          = { .digest = {
275                                 .dia_digestsize = SHA384_DIGEST_SIZE,
276                                 .dia_init       = sha384_init,
277                                 .dia_update     = sha512_update,
278                                 .dia_final      = sha384_final }
279         }
280 };
281
282 static int __init sha512_generic_mod_init(void)
283 {
284         int ret = 0;
285
286         if ((ret = crypto_register_alg(&sha384)) < 0)
287                 goto out;
288         if ((ret = crypto_register_alg(&sha512)) < 0)
289                 crypto_unregister_alg(&sha384);
290 out:
291         return ret;
292 }
293
294 static void __exit sha512_generic_mod_fini(void)
295 {
296         crypto_unregister_alg(&sha384);
297         crypto_unregister_alg(&sha512);
298 }
299
300 module_init(sha512_generic_mod_init);
301 module_exit(sha512_generic_mod_fini);
302
303 MODULE_LICENSE("GPL");
304 MODULE_DESCRIPTION("SHA-512 and SHA-384 Secure Hash Algorithms");
305
306 MODULE_ALIAS("sha384");
307 MODULE_ALIAS("sha512");