2 * copyright (c) 2007 Michael Niedermayer <michaelni@gmx.at>
4 * some optimization ideas from aes128.c by Reimar Doeffinger
6 * This file is part of Libav.
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.
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.
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
24 #include "intreadwrite.h"
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];
43 struct AVAES *av_aes_alloc(void)
45 return av_mallocz(sizeof(struct AVAES));
48 static const uint8_t rcon[10] = {
49 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36
52 static uint8_t sbox[256];
53 static uint8_t inv_sbox[256];
55 static uint32_t enc_multbl[1][256];
56 static uint32_t dec_multbl[1][256];
58 static uint32_t enc_multbl[4][256];
59 static uint32_t dec_multbl[4][256];
63 # define ROT(x, s) ((x >> s) | (x << (32-s)))
65 # define ROT(x, s) ((x << s) | (x >> (32-s)))
68 static inline void addkey(av_aes_block *dst, const av_aes_block *src,
69 const av_aes_block *round_key)
71 dst->u64[0] = src->u64[0] ^ round_key->u64[0];
72 dst->u64[1] = src->u64[1] ^ round_key->u64[1];
75 static inline void addkey_s(av_aes_block *dst, const uint8_t *src,
76 const av_aes_block *round_key)
78 dst->u64[0] = AV_RN64(src) ^ round_key->u64[0];
79 dst->u64[1] = AV_RN64(src + 8) ^ round_key->u64[1];
82 static inline void addkey_d(uint8_t *dst, const av_aes_block *src,
83 const av_aes_block *round_key)
85 AV_WN64(dst, src->u64[0] ^ round_key->u64[0]);
86 AV_WN64(dst + 8, src->u64[1] ^ round_key->u64[1]);
89 static void subshift(av_aes_block s0[2], int s, const uint8_t *box)
91 av_aes_block *s1 = (av_aes_block *) (s0[0].u8 - s);
92 av_aes_block *s3 = (av_aes_block *) (s0[0].u8 + s);
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]];
112 static inline int mix_core(uint32_t multbl[][256], int a, int b, int c, int d)
115 return multbl[0][a] ^ ROT(multbl[0][b], 8) ^ ROT(multbl[0][c], 16) ^ ROT(multbl[0][d], 24);
117 return multbl[0][a] ^ multbl[1][b] ^ multbl[2][c] ^ multbl[3][d];
121 static inline void mix(av_aes_block state[2], uint32_t multbl[][256], int s1, int s3)
123 uint8_t (*src)[4] = state[1].u8x4;
124 state[0].u32[0] = mix_core(multbl, src[0][0], src[s1 ][1], src[2][2], src[s3 ][3]);
125 state[0].u32[1] = mix_core(multbl, src[1][0], src[s3 - 1][1], src[3][2], src[s1 - 1][3]);
126 state[0].u32[2] = mix_core(multbl, src[2][0], src[s3 ][1], src[0][2], src[s1 ][3]);
127 state[0].u32[3] = mix_core(multbl, src[3][0], src[s1 - 1][1], src[1][2], src[s3 - 1][3]);
130 static inline void crypt(AVAES *a, int s, const uint8_t *sbox,
131 uint32_t multbl[][256])
135 for (r = a->rounds - 1; r > 0; r--) {
136 mix(a->state, multbl, 3 - s, 1 + s);
137 addkey(&a->state[1], &a->state[0], &a->round_key[r]);
140 subshift(&a->state[0], s, sbox);
143 void av_aes_crypt(AVAES *a, uint8_t *dst, const uint8_t *src,
144 int count, uint8_t *iv, int decrypt)
147 addkey_s(&a->state[1], src, &a->round_key[a->rounds]);
149 crypt(a, 0, inv_sbox, dec_multbl);
151 addkey_s(&a->state[0], iv, &a->state[0]);
154 addkey_d(dst, &a->state[0], &a->round_key[0]);
157 addkey_s(&a->state[1], iv, &a->state[1]);
158 crypt(a, 2, sbox, enc_multbl);
159 addkey_d(dst, &a->state[0], &a->round_key[0]);
168 static void init_multbl2(uint32_t tbl[][256], const int c[4],
169 const uint8_t *log8, const uint8_t *alog8,
174 for (i = 0; i < 256; i++) {
179 k = alog8[x + log8[c[0]]];
180 l = alog8[x + log8[c[1]]];
181 m = alog8[x + log8[c[2]]];
182 n = alog8[x + log8[c[3]]];
183 tbl[0][i] = AV_NE(MKBETAG(k, l, m, n), MKTAG(k, l, m, n));
185 tbl[1][i] = ROT(tbl[0][i], 8);
186 tbl[2][i] = ROT(tbl[0][i], 16);
187 tbl[3][i] = ROT(tbl[0][i], 24);
193 // this is based on the reference AES code by Paulo Barreto and Vincent Rijmen
194 int av_aes_init(AVAES *a, const uint8_t *key, int key_bits, int decrypt)
196 int i, j, t, rconpointer = 0;
198 int KC = key_bits >> 5;
203 if (!enc_multbl[FF_ARRAY_ELEMS(enc_multbl) - 1][FF_ARRAY_ELEMS(enc_multbl[0]) - 1]) {
205 for (i = 0; i < 255; i++) {
206 alog8[i] = alog8[i + 255] = j;
212 for (i = 0; i < 256; i++) {
213 j = i ? alog8[255 - log8[i]] : 0;
214 j ^= (j << 1) ^ (j << 2) ^ (j << 3) ^ (j << 4);
215 j = (j ^ (j >> 8) ^ 99) & 255;
219 init_multbl2(dec_multbl, (const int[4]) { 0xe, 0x9, 0xd, 0xb },
220 log8, alog8, inv_sbox);
221 init_multbl2(enc_multbl, (const int[4]) { 0x2, 0x1, 0x1, 0x3 },
225 if (key_bits != 128 && key_bits != 192 && key_bits != 256)
230 memcpy(tk, key, KC * 4);
231 memcpy(a->round_key[0].u8, key, KC * 4);
233 for (t = KC * 4; t < (rounds + 1) * 16; t += KC * 4) {
234 for (i = 0; i < 4; i++)
235 tk[0][i] ^= sbox[tk[KC - 1][(i + 1) & 3]];
236 tk[0][0] ^= rcon[rconpointer++];
238 for (j = 1; j < KC; j++) {
239 if (KC != 8 || j != KC >> 1)
240 for (i = 0; i < 4; i++)
241 tk[j][i] ^= tk[j - 1][i];
243 for (i = 0; i < 4; i++)
244 tk[j][i] ^= sbox[tk[j - 1][i]];
247 memcpy(a->round_key[0].u8 + t, tk, KC * 4);
251 for (i = 1; i < rounds; i++) {
253 tmp[2] = a->round_key[i];
254 subshift(&tmp[1], 0, sbox);
255 mix(tmp, dec_multbl, 1, 3);
256 a->round_key[i] = tmp[0];
259 for (i = 0; i < (rounds + 1) >> 1; i++)
260 FFSWAP(av_aes_block, a->round_key[i], a->round_key[rounds - i]);