#include "common.h"
#include "aes.h"
+#include "intreadwrite.h"
typedef union {
uint64_t u64[2];
typedef struct AVAES {
// Note: round_key[16] is accessed in the init code, but this only
- // overwrites state, which does not matter (see also r7471).
+ // overwrites state, which does not matter (see also commit ba554c0).
av_aes_block round_key[15];
av_aes_block state[2];
int rounds;
static uint32_t dec_multbl[4][256];
#endif
+#if HAVE_BIGENDIAN
+# define ROT(x, s) ((x >> s) | (x << (32-s)))
+#else
+# define ROT(x, s) ((x << s) | (x >> (32-s)))
+#endif
+
static inline void addkey(av_aes_block *dst, const av_aes_block *src,
const av_aes_block *round_key)
{
dst->u64[1] = src->u64[1] ^ round_key->u64[1];
}
+static inline void addkey_s(av_aes_block *dst, const uint8_t *src,
+ const av_aes_block *round_key)
+{
+ dst->u64[0] = AV_RN64(src) ^ round_key->u64[0];
+ dst->u64[1] = AV_RN64(src + 8) ^ round_key->u64[1];
+}
+
+static inline void addkey_d(uint8_t *dst, const av_aes_block *src,
+ const av_aes_block *round_key)
+{
+ AV_WN64(dst, src->u64[0] ^ round_key->u64[0]);
+ AV_WN64(dst + 8, src->u64[1] ^ round_key->u64[1]);
+}
+
static void subshift(av_aes_block s0[2], int s, const uint8_t *box)
{
av_aes_block *s1 = (av_aes_block *) (s0[0].u8 - s);
static inline int mix_core(uint32_t multbl[][256], int a, int b, int c, int d){
#if CONFIG_SMALL
-#define ROT(x,s) ((x<<s)|(x>>(32-s)))
return multbl[0][a] ^ ROT(multbl[0][b], 8) ^ ROT(multbl[0][c], 16) ^ ROT(multbl[0][d], 24);
#else
return multbl[0][a] ^ multbl[1][b] ^ multbl[2][c] ^ multbl[3][d];
subshift(&a->state[0], s, sbox);
}
-void av_aes_crypt(AVAES *a, uint8_t *dst_, const uint8_t *src_,
- int count, uint8_t *iv_, int decrypt)
+void av_aes_crypt(AVAES *a, uint8_t *dst, const uint8_t *src,
+ int count, uint8_t *iv, int decrypt)
{
- av_aes_block *dst = (av_aes_block *) dst_;
- const av_aes_block *src = (const av_aes_block *) src_;
- av_aes_block *iv = (av_aes_block *) iv_;
-
while (count--) {
- addkey(&a->state[1], src, &a->round_key[a->rounds]);
+ addkey_s(&a->state[1], src, &a->round_key[a->rounds]);
if (decrypt) {
crypt(a, 0, inv_sbox, dec_multbl);
if (iv) {
- addkey(&a->state[0], &a->state[0], iv);
+ addkey_s(&a->state[0], iv, &a->state[0]);
memcpy(iv, src, 16);
}
- addkey(dst, &a->state[0], &a->round_key[0]);
+ addkey_d(dst, &a->state[0], &a->round_key[0]);
} else {
if (iv)
- addkey(&a->state[1], &a->state[1], iv);
+ addkey_s(&a->state[1], iv, &a->state[1]);
crypt(a, 2, sbox, enc_multbl);
- addkey(dst, &a->state[0], &a->round_key[0]);
+ addkey_d(dst, &a->state[0], &a->round_key[0]);
if (iv)
memcpy(iv, dst, 16);
}
- src++;
- dst++;
+ src += 16;
+ dst += 16;
}
}
-static void init_multbl2(uint8_t tbl[1024], const int c[4],
+static void init_multbl2(uint32_t tbl[][256], const int c[4],
const uint8_t *log8, const uint8_t *alog8,
const uint8_t *sbox)
{
- int i, j;
-
- for (i = 0; i < 1024; i++) {
- int x = sbox[i >> 2];
- if (x)
- tbl[i] = alog8[log8[x] + log8[c[i & 3]]];
- }
+ int i;
+
+ for (i = 0; i < 256; i++) {
+ int x = sbox[i];
+ if (x) {
+ int k, l, m, n;
+ x = log8[x];
+ k = alog8[x + log8[c[0]]];
+ l = alog8[x + log8[c[1]]];
+ m = alog8[x + log8[c[2]]];
+ n = alog8[x + log8[c[3]]];
+ tbl[0][i] = AV_NE(MKBETAG(k,l,m,n), MKTAG(k,l,m,n));
#if !CONFIG_SMALL
- for (j = 256; j < 1024; j++)
- for (i = 0; i < 4; i++)
- tbl[4*j + i] = tbl[4*j + ((i - 1) & 3) - 1024];
+ tbl[1][i] = ROT(tbl[0][i], 8);
+ tbl[2][i] = ROT(tbl[0][i], 16);
+ tbl[3][i] = ROT(tbl[0][i], 24);
#endif
+ }
+ }
}
// this is based on the reference AES code by Paulo Barreto and Vincent Rijmen
inv_sbox[j] = i;
sbox[i] = j;
}
- init_multbl2(dec_multbl[0], (const int[4]) { 0xe, 0x9, 0xd, 0xb },
+ init_multbl2(dec_multbl, (const int[4]) { 0xe, 0x9, 0xd, 0xb },
log8, alog8, inv_sbox);
- init_multbl2(enc_multbl[0], (const int[4]) { 0x2, 0x1, 0x1, 0x3 },
+ init_multbl2(enc_multbl, (const int[4]) { 0x2, 0x1, 0x1, 0x3 },
log8, alog8, sbox);
}
a->rounds = rounds;
memcpy(tk, key, KC * 4);
+ memcpy(a->round_key[0].u8, key, KC * 4);
- for (t = 0; t < (rounds + 1) * 16;) {
- memcpy(a->round_key[0].u8 + t, tk, KC * 4);
- t += KC * 4;
-
+ for (t = KC * 4; t < (rounds + 1) * 16; t += KC * 4) {
for (i = 0; i < 4; i++)
tk[0][i] ^= sbox[tk[KC - 1][(i + 1) & 3]];
tk[0][0] ^= rcon[rconpointer++];
for (i = 0; i < 4; i++)
tk[j][i] ^= sbox[tk[j - 1][i]];
}
+
+ memcpy(a->round_key[0].u8 + t, tk, KC * 4);
}
if (decrypt) {
for (i = 1; i < rounds; i++) {
av_aes_block tmp[3];
- memcpy(&tmp[2], &a->round_key[i], 16);
+ tmp[2] = a->round_key[i];
subshift(&tmp[1], 0, sbox);
mix(tmp, dec_multbl, 1, 3);
- memcpy(&a->round_key[i], &tmp[0], 16);
+ a->round_key[i] = tmp[0];
}
} else {
for (i = 0; i < (rounds + 1) >> 1; i++) {
- for (j = 0; j < 16; j++)
- FFSWAP(int, a->round_key[i].u8[j], a->round_key[rounds-i].u8[j]);
+ FFSWAP(av_aes_block, a->round_key[i], a->round_key[rounds-i]);
}
}