02a9281b77eb33cd97e86a806e805f60adf8e33a
[libav.git] / libavutil / aes.c
1 /*
2 * copyright (c) 2007 Michael Niedermayer <michaelni@gmx.at>
3 *
4 * some optimization ideas from aes128.c by Reimar Doeffinger
5 *
6 * This file is part of Libav.
7 *
8 * Libav is free software; you can redistribute it and/or
9 * modify it under the terms of the GNU Lesser General Public
10 * License as published by the Free Software Foundation; either
11 * version 2.1 of the License, or (at your option) any later version.
12 *
13 * Libav is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
16 * Lesser General Public License for more details.
17 *
18 * You should have received a copy of the GNU Lesser General Public
19 * License along with Libav; if not, write to the Free Software
20 * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
21 */
22
23 #include "common.h"
24 #include "aes.h"
25 #include "intreadwrite.h"
26 #include "timer.h"
27
28 typedef union {
29 uint64_t u64[2];
30 uint32_t u32[4];
31 uint8_t u8x4[4][4];
32 uint8_t u8[16];
33 } av_aes_block;
34
35 typedef struct AVAES {
36 // Note: round_key[16] is accessed in the init code, but this only
37 // overwrites state, which does not matter (see also commit ba554c0).
38 av_aes_block round_key[15];
39 av_aes_block state[2];
40 int rounds;
41 } AVAES;
42
43 struct AVAES *av_aes_alloc(void)
44 {
45 return av_mallocz(sizeof(struct AVAES));
46 }
47
48 static const uint8_t rcon[10] = {
49 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36
50 };
51
52 static uint8_t sbox[256];
53 static uint8_t inv_sbox[256];
54 #if CONFIG_SMALL
55 static uint32_t enc_multbl[1][256];
56 static uint32_t dec_multbl[1][256];
57 #else
58 static uint32_t enc_multbl[4][256];
59 static uint32_t dec_multbl[4][256];
60 #endif
61
62 #if HAVE_BIGENDIAN
63 # define ROT(x, s) ((x >> s) | (x << (32-s)))
64 #else
65 # define ROT(x, s) ((x << s) | (x >> (32-s)))
66 #endif
67
68 static inline void addkey(av_aes_block *dst, const av_aes_block *src,
69 const av_aes_block *round_key)
70 {
71 dst->u64[0] = src->u64[0] ^ round_key->u64[0];
72 dst->u64[1] = src->u64[1] ^ round_key->u64[1];
73 }
74
75 static inline void addkey_s(av_aes_block *dst, const uint8_t *src,
76 const av_aes_block *round_key)
77 {
78 dst->u64[0] = AV_RN64(src) ^ round_key->u64[0];
79 dst->u64[1] = AV_RN64(src + 8) ^ round_key->u64[1];
80 }
81
82 static inline void addkey_d(uint8_t *dst, const av_aes_block *src,
83 const av_aes_block *round_key)
84 {
85 AV_WN64(dst, src->u64[0] ^ round_key->u64[0]);
86 AV_WN64(dst + 8, src->u64[1] ^ round_key->u64[1]);
87 }
88
89 static void subshift(av_aes_block s0[2], int s, const uint8_t *box)
90 {
91 av_aes_block *s1 = (av_aes_block *) (s0[0].u8 - s);
92 av_aes_block *s3 = (av_aes_block *) (s0[0].u8 + s);
93
94 s0[0].u8[ 0] = box[s0[1].u8[ 0]];
95 s0[0].u8[ 4] = box[s0[1].u8[ 4]];
96 s0[0].u8[ 8] = box[s0[1].u8[ 8]];
97 s0[0].u8[12] = box[s0[1].u8[12]];
98 s1[0].u8[ 3] = box[s1[1].u8[ 7]];
99 s1[0].u8[ 7] = box[s1[1].u8[11]];
100 s1[0].u8[11] = box[s1[1].u8[15]];
101 s1[0].u8[15] = box[s1[1].u8[ 3]];
102 s0[0].u8[ 2] = box[s0[1].u8[10]];
103 s0[0].u8[10] = box[s0[1].u8[ 2]];
104 s0[0].u8[ 6] = box[s0[1].u8[14]];
105 s0[0].u8[14] = box[s0[1].u8[ 6]];
106 s3[0].u8[ 1] = box[s3[1].u8[13]];
107 s3[0].u8[13] = box[s3[1].u8[ 9]];
108 s3[0].u8[ 9] = box[s3[1].u8[ 5]];
109 s3[0].u8[ 5] = box[s3[1].u8[ 1]];
110 }
111
112 static inline int mix_core(uint32_t multbl[][256], int a, int b, int c, int d){
113 #if CONFIG_SMALL
114 return multbl[0][a] ^ ROT(multbl[0][b], 8) ^ ROT(multbl[0][c], 16) ^ ROT(multbl[0][d], 24);
115 #else
116 return multbl[0][a] ^ multbl[1][b] ^ multbl[2][c] ^ multbl[3][d];
117 #endif
118 }
119
120 static inline void mix(av_aes_block state[2], uint32_t multbl[][256], int s1, int s3){
121 uint8_t (*src)[4] = state[1].u8x4;
122 state[0].u32[0] = mix_core(multbl, src[0][0], src[s1 ][1], src[2][2], src[s3 ][3]);
123 state[0].u32[1] = mix_core(multbl, src[1][0], src[s3-1][1], src[3][2], src[s1-1][3]);
124 state[0].u32[2] = mix_core(multbl, src[2][0], src[s3 ][1], src[0][2], src[s1 ][3]);
125 state[0].u32[3] = mix_core(multbl, src[3][0], src[s1-1][1], src[1][2], src[s3-1][3]);
126 }
127
128 static inline void crypt(AVAES *a, int s, const uint8_t *sbox,
129 uint32_t multbl[][256])
130 {
131 int r;
132
133 for (r = a->rounds - 1; r > 0; r--) {
134 mix(a->state, multbl, 3 - s, 1 + s);
135 addkey(&a->state[1], &a->state[0], &a->round_key[r]);
136 }
137
138 subshift(&a->state[0], s, sbox);
139 }
140
141 void av_aes_crypt(AVAES *a, uint8_t *dst, const uint8_t *src,
142 int count, uint8_t *iv, int decrypt)
143 {
144 while (count--) {
145 addkey_s(&a->state[1], src, &a->round_key[a->rounds]);
146 if (decrypt) {
147 crypt(a, 0, inv_sbox, dec_multbl);
148 if (iv) {
149 addkey_s(&a->state[0], iv, &a->state[0]);
150 memcpy(iv, src, 16);
151 }
152 addkey_d(dst, &a->state[0], &a->round_key[0]);
153 } else {
154 if (iv)
155 addkey_s(&a->state[1], iv, &a->state[1]);
156 crypt(a, 2, sbox, enc_multbl);
157 addkey_d(dst, &a->state[0], &a->round_key[0]);
158 if (iv)
159 memcpy(iv, dst, 16);
160 }
161 src += 16;
162 dst += 16;
163 }
164 }
165
166 static void init_multbl2(uint32_t tbl[][256], const int c[4],
167 const uint8_t *log8, const uint8_t *alog8,
168 const uint8_t *sbox)
169 {
170 int i;
171
172 for (i = 0; i < 256; i++) {
173 int x = sbox[i];
174 if (x) {
175 int k, l, m, n;
176 x = log8[x];
177 k = alog8[x + log8[c[0]]];
178 l = alog8[x + log8[c[1]]];
179 m = alog8[x + log8[c[2]]];
180 n = alog8[x + log8[c[3]]];
181 tbl[0][i] = AV_NE(MKBETAG(k,l,m,n), MKTAG(k,l,m,n));
182 #if !CONFIG_SMALL
183 tbl[1][i] = ROT(tbl[0][i], 8);
184 tbl[2][i] = ROT(tbl[0][i], 16);
185 tbl[3][i] = ROT(tbl[0][i], 24);
186 #endif
187 }
188 }
189 }
190
191 // this is based on the reference AES code by Paulo Barreto and Vincent Rijmen
192 int av_aes_init(AVAES *a, const uint8_t *key, int key_bits, int decrypt)
193 {
194 int i, j, t, rconpointer = 0;
195 uint8_t tk[8][4];
196 int KC = key_bits >> 5;
197 int rounds = KC + 6;
198 uint8_t log8[256];
199 uint8_t alog8[512];
200
201 if (!enc_multbl[FF_ARRAY_ELEMS(enc_multbl)-1][FF_ARRAY_ELEMS(enc_multbl[0])-1]) {
202 j = 1;
203 for (i = 0; i < 255; i++) {
204 alog8[i] = alog8[i + 255] = j;
205 log8[j] = i;
206 j ^= j + j;
207 if (j > 255)
208 j ^= 0x11B;
209 }
210 for (i = 0; i < 256; i++) {
211 j = i ? alog8[255 - log8[i]] : 0;
212 j ^= (j << 1) ^ (j << 2) ^ (j << 3) ^ (j << 4);
213 j = (j ^ (j >> 8) ^ 99) & 255;
214 inv_sbox[j] = i;
215 sbox[i] = j;
216 }
217 init_multbl2(dec_multbl, (const int[4]) { 0xe, 0x9, 0xd, 0xb },
218 log8, alog8, inv_sbox);
219 init_multbl2(enc_multbl, (const int[4]) { 0x2, 0x1, 0x1, 0x3 },
220 log8, alog8, sbox);
221 }
222
223 if (key_bits != 128 && key_bits != 192 && key_bits != 256)
224 return -1;
225
226 a->rounds = rounds;
227
228 memcpy(tk, key, KC * 4);
229 memcpy(a->round_key[0].u8, key, KC * 4);
230
231 for (t = KC * 4; t < (rounds + 1) * 16; t += KC * 4) {
232 for (i = 0; i < 4; i++)
233 tk[0][i] ^= sbox[tk[KC - 1][(i + 1) & 3]];
234 tk[0][0] ^= rcon[rconpointer++];
235
236 for (j = 1; j < KC; j++) {
237 if (KC != 8 || j != KC >> 1)
238 for (i = 0; i < 4; i++)
239 tk[j][i] ^= tk[j - 1][i];
240 else
241 for (i = 0; i < 4; i++)
242 tk[j][i] ^= sbox[tk[j - 1][i]];
243 }
244
245 memcpy(a->round_key[0].u8 + t, tk, KC * 4);
246 }
247
248 if (decrypt) {
249 for (i = 1; i < rounds; i++) {
250 av_aes_block tmp[3];
251 tmp[2] = a->round_key[i];
252 subshift(&tmp[1], 0, sbox);
253 mix(tmp, dec_multbl, 1, 3);
254 a->round_key[i] = tmp[0];
255 }
256 } else {
257 for (i = 0; i < (rounds + 1) >> 1; i++) {
258 FFSWAP(av_aes_block, a->round_key[i], a->round_key[rounds-i]);
259 }
260 }
261
262 return 0;
263 }
264
265 #ifdef TEST
266 #include <string.h>
267 #include "lfg.h"
268 #include "log.h"
269
270 int main(int argc, char **argv)
271 {
272 int i, j;
273 AVAES b;
274 uint8_t rkey[2][16] = {
275 { 0 },
276 { 0x10, 0xa5, 0x88, 0x69, 0xd7, 0x4b, 0xe5, 0xa3,
277 0x74, 0xcf, 0x86, 0x7c, 0xfb, 0x47, 0x38, 0x59 }
278 };
279 uint8_t pt[16], rpt[2][16]= {
280 { 0x6a, 0x84, 0x86, 0x7c, 0xd7, 0x7e, 0x12, 0xad,
281 0x07, 0xea, 0x1b, 0xe8, 0x95, 0xc5, 0x3f, 0xa3 },
282 { 0 }
283 };
284 uint8_t rct[2][16]= {
285 { 0x73, 0x22, 0x81, 0xc0, 0xa0, 0xaa, 0xb8, 0xf7,
286 0xa5, 0x4a, 0x0c, 0x67, 0xa0, 0xc4, 0x5e, 0xcf },
287 { 0x6d, 0x25, 0x1e, 0x69, 0x44, 0xb0, 0x51, 0xe0,
288 0x4e, 0xaa, 0x6f, 0xb4, 0xdb, 0xf7, 0x84, 0x65 }
289 };
290 uint8_t temp[16];
291 int err = 0;
292
293 av_log_set_level(AV_LOG_DEBUG);
294
295 for (i = 0; i < 2; i++) {
296 av_aes_init(&b, rkey[i], 128, 1);
297 av_aes_crypt(&b, temp, rct[i], 1, NULL, 1);
298 for (j = 0; j < 16; j++) {
299 if (rpt[i][j] != temp[j]) {
300 av_log(NULL, AV_LOG_ERROR, "%d %02X %02X\n",
301 j, rpt[i][j], temp[j]);
302 err = 1;
303 }
304 }
305 }
306
307 if (argc > 1 && !strcmp(argv[1], "-t")) {
308 AVAES ae, ad;
309 AVLFG prng;
310
311 av_aes_init(&ae, "PI=3.141592654..", 128, 0);
312 av_aes_init(&ad, "PI=3.141592654..", 128, 1);
313 av_lfg_init(&prng, 1);
314
315 for (i = 0; i < 10000; i++) {
316 for (j = 0; j < 16; j++) {
317 pt[j] = av_lfg_get(&prng);
318 }
319 {
320 START_TIMER;
321 av_aes_crypt(&ae, temp, pt, 1, NULL, 0);
322 if (!(i & (i - 1)))
323 av_log(NULL, AV_LOG_ERROR, "%02X %02X %02X %02X\n",
324 temp[0], temp[5], temp[10], temp[15]);
325 av_aes_crypt(&ad, temp, temp, 1, NULL, 1);
326 STOP_TIMER("aes");
327 }
328 for (j = 0; j < 16; j++) {
329 if (pt[j] != temp[j]) {
330 av_log(NULL, AV_LOG_ERROR, "%d %d %02X %02X\n",
331 i, j, pt[j], temp[j]);
332 }
333 }
334 }
335 }
336 return err;
337 }
338 #endif