Commit f12cf3d2 authored by Michael Niedermayer's avatar Michael Niedermayer

merge inv_sbox into inv_mix

Originally committed as revision 7487 to svn://svn.ffmpeg.org/ffmpeg/trunk
parent ae557450
......@@ -23,7 +23,8 @@
#include "aes.h"
typedef struct AVAES{
uint8_t round_key[15][4][4];
uint8_t round_enc_key[15][4][4];
uint8_t round_dec_key[15][4][4];
uint8_t state[4][4];
int rounds;
}AVAES;
......@@ -74,40 +75,38 @@ static inline void mix(uint8_t state[4][4], uint32_t multbl[4][256]){
void av_aes_decrypt(AVAES *a){
int t, r;
addkey(a->state, a->round_key[a->rounds]);
for(r=a->rounds-1; r>=0; r--){
if(r!=a->rounds-1)
mix(a->state, dec_multbl);
SUBSHIFT0((a->state[0]+0), inv_sbox)
SUBSHIFT3((a->state[0]+1), inv_sbox)
SUBSHIFT2((a->state[0]+2), inv_sbox)
SUBSHIFT1((a->state[0]+3), inv_sbox)
addkey(a->state, a->round_key[r]);
addkey(a->state, a->round_enc_key[a->rounds]);
for(r=a->rounds-2; r>=0; r--){
// SUBSHIFT0((a->state[0]+0), inv_sbox)
SUBSHIFT3x((a->state[0]+1), inv_sbox)
SUBSHIFT2x((a->state[0]+2), inv_sbox)
SUBSHIFT1x((a->state[0]+3), inv_sbox)
mix(a->state, dec_multbl);
addkey(a->state, a->round_dec_key[r+1]);
}
SUBSHIFT0((a->state[0]+0), inv_sbox)
SUBSHIFT3((a->state[0]+1), inv_sbox)
SUBSHIFT2((a->state[0]+2), inv_sbox)
SUBSHIFT1((a->state[0]+3), inv_sbox)
addkey(a->state, a->round_enc_key[0]);
}
void av_aes_encrypt(AVAES *a){
int r, t;
for(r=0; r<a->rounds-1; r++){
addkey(a->state, a->round_key[r]);
addkey(a->state, a->round_enc_key[r]);
SUBSHIFT1x((a->state[0]+1), sbox)
SUBSHIFT2x((a->state[0]+2), sbox)
SUBSHIFT3x((a->state[0]+3), sbox)
mix(a->state, enc_multbl); //FIXME replace log8 by const / optimze mix as this can be simplified alot
}
addkey(a->state, a->round_key[r]);
addkey(a->state, a->round_enc_key[r]);
SUBSHIFT0((a->state[0]+0), sbox)
SUBSHIFT1((a->state[0]+1), sbox)
SUBSHIFT2((a->state[0]+2), sbox)
SUBSHIFT3((a->state[0]+3), sbox)
addkey(a->state, a->round_key[r+1]);
}
static init_multbl(uint8_t tbl[1024], int c[4], uint8_t *log8, uint8_t *alog8){
int i;
for(i=4; i<1024; i++)
tbl[i]= alog8[ log8[i/4] + log8[c[i&3]] ];
addkey(a->state, a->round_enc_key[r+1]);
}
static init_multbl2(uint8_t tbl[1024], int c[4], uint8_t *log8, uint8_t *alog8, uint8_t *sbox){
......@@ -147,11 +146,11 @@ AVAES *av_aes_init(uint8_t *key, int key_bits) {
sbox [i]= j;
// av_log(NULL, AV_LOG_ERROR, "%d, ", log8[i]);
}
init_multbl(dec_multbl[0], (int[4]){0xe, 0x9, 0xd, 0xb}, log8, alog8);
init_multbl2(dec_multbl[0], (int[4]){0xe, 0x9, 0xd, 0xb}, log8, alog8, inv_sbox);
#ifndef CONFIG_SMALL
init_multbl(dec_multbl[1], (int[4]){0xb, 0xe, 0x9, 0xd}, log8, alog8);
init_multbl(dec_multbl[2], (int[4]){0xd, 0xb, 0xe, 0x9}, log8, alog8);
init_multbl(dec_multbl[3], (int[4]){0x9, 0xd, 0xb, 0xe}, log8, alog8);
init_multbl2(dec_multbl[1], (int[4]){0xb, 0xe, 0x9, 0xd}, log8, alog8, inv_sbox);
init_multbl2(dec_multbl[2], (int[4]){0xd, 0xb, 0xe, 0x9}, log8, alog8, inv_sbox);
init_multbl2(dec_multbl[3], (int[4]){0x9, 0xd, 0xb, 0xe}, log8, alog8, inv_sbox);
#endif
init_multbl2(enc_multbl[0], (int[4]){0x2, 0x1, 0x1, 0x3}, log8, alog8, sbox);
#ifndef CONFIG_SMALL
......@@ -170,7 +169,7 @@ AVAES *av_aes_init(uint8_t *key, int key_bits) {
memcpy(tk, key, KC*4);
for(t= 0; t < (rounds+1)*4;) {
memcpy(a->round_key[0][t], tk, KC*4);
memcpy(a->round_enc_key[0][t], tk, KC*4);
t+= KC;
for(i = 0; i < 4; i++)
......@@ -184,6 +183,12 @@ AVAES *av_aes_init(uint8_t *key, int key_bits) {
for(i = 0; i < 4; i++) tk[j][i] ^= sbox[tk[j-1][i]];
}
}
for(i=0; i<sizeof(a->round_enc_key); i++)
a->round_dec_key[0][0][i]= sbox[a->round_enc_key[0][0][i]];
for(i=1; i<rounds; i++)
mix(a->round_dec_key[i], dec_multbl);
return a;
}
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment