/*
* copyright (c) 2007 Michael Niedermayer <michaelni@gmx.at>
*
+ * some optimization ideas from aes128.c by Reimar Doeffinger
+ *
* This file is part of FFmpeg.
*
* FFmpeg is free software; you can redistribute it and/or
* You should have received a copy of the GNU Lesser General Public
* License along with FFmpeg; if not, write to the Free Software
* Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- *
- * some optimization ideas from aes128.c by Reimar Doeffinger
*/
#include "common.h"
#include "aes.h"
+typedef union {
+ uint64_t u64[2];
+ uint32_t u32[4];
+ uint8_t u8x4[4][4];
+ uint8_t u8[16];
+} av_aes_block;
+
typedef struct AVAES{
- uint8_t round_key[15][4][4];
- uint8_t state[2][4][4];
+ // Note: round_key[16] is accessed in the init code, but this only
+ // overwrites state, which does not matter (see also r7471).
+ av_aes_block round_key[15];
+ av_aes_block state[2];
int rounds;
}AVAES;
static uint8_t sbox[256];
static uint8_t inv_sbox[256];
-#ifdef CONFIG_SMALL
+#if CONFIG_SMALL
static uint32_t enc_multbl[1][256];
static uint32_t dec_multbl[1][256];
#else
static uint32_t dec_multbl[4][256];
#endif
-static inline void addkey(uint64_t state[2], uint64_t round_key[2]){
- state[2] = state[0] ^ round_key[0];
- state[3] = state[1] ^ round_key[1];
+static inline void addkey(av_aes_block *dst, const av_aes_block *src, const av_aes_block *round_key){
+ dst->u64[0] = src->u64[0] ^ round_key->u64[0];
+ dst->u64[1] = src->u64[1] ^ round_key->u64[1];
}
-static void subshift(uint8_t s0[2][16], uint8_t s1[2][16], uint8_t s3[2][16], uint8_t *box){
- s0[0][0]=box[s0[1][ 0]]; s0[0][ 4]=box[s0[1][ 4]]; s0[0][ 8]=box[s0[1][ 8]]; s0[0][12]=box[s0[1][12]];
- s1[0][0]=box[s1[1][ 4]]; s1[0][ 4]=box[s1[1][ 8]]; s1[0][ 8]=box[s1[1][12]]; s1[0][12]=box[s1[1][ 0]];
- s0[0][2]=box[s0[1][10]]; s0[0][10]=box[s0[1][ 2]]; s0[0][ 6]=box[s0[1][14]]; s0[0][14]=box[s0[1][ 6]];
- s3[0][0]=box[s3[1][12]]; s3[0][12]=box[s3[1][ 8]]; s3[0][ 8]=box[s3[1][ 4]]; s3[0][ 4]=box[s3[1][ 0]];
+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);
+ av_aes_block *s3= (av_aes_block *)(s0[0].u8 + s);
+ s0[0].u8[0]=box[s0[1].u8[ 0]]; s0[0].u8[ 4]=box[s0[1].u8[ 4]]; s0[0].u8[ 8]=box[s0[1].u8[ 8]]; s0[0].u8[12]=box[s0[1].u8[12]];
+ s1[0].u8[3]=box[s1[1].u8[ 7]]; s1[0].u8[ 7]=box[s1[1].u8[11]]; s1[0].u8[11]=box[s1[1].u8[15]]; s1[0].u8[15]=box[s1[1].u8[ 3]];
+ s0[0].u8[2]=box[s0[1].u8[10]]; s0[0].u8[10]=box[s0[1].u8[ 2]]; s0[0].u8[ 6]=box[s0[1].u8[14]]; s0[0].u8[14]=box[s0[1].u8[ 6]];
+ s3[0].u8[1]=box[s3[1].u8[13]]; s3[0].u8[13]=box[s3[1].u8[ 9]]; s3[0].u8[ 9]=box[s3[1].u8[ 5]]; s3[0].u8[ 5]=box[s3[1].u8[ 1]];
}
+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)))
-#if 0
-static inline void mix(uint8_t state[4][4], uint32_t multbl[4][256]){
- int i;
- for(i=0; i<4; i++)
-#ifdef CONFIG_SMALL
- ((uint32_t *)(state))[i] = multbl[0][state[i][0]] ^ ROT(multbl[0][state[i][1]], 8)
- ^ROT(multbl[0][state[i][2]],16) ^ ROT(multbl[0][state[i][3]],24);
+ return multbl[0][a] ^ ROT(multbl[0][b], 8) ^ ROT(multbl[0][c], 16) ^ ROT(multbl[0][d], 24);
#else
- ((uint32_t *)(state))[i] = multbl[0][state[i][0]] ^ multbl[1][state[i][1]]
- ^multbl[2][state[i][2]] ^ multbl[3][state[i][3]];
+ return multbl[0][a] ^ multbl[1][b] ^ multbl[2][c] ^ multbl[3][d];
#endif
}
-#endif
-static inline void mix(uint8_t state[2][4][4], uint32_t multbl[4][256], int s1, int s3){
- ((uint32_t *)(state))[0] = multbl[0][state[1][0][0]] ^ multbl[1][state[1][s1 ][1]]
- ^multbl[2][state[1][2][2]] ^ multbl[3][state[1][s3 ][3]];
- ((uint32_t *)(state))[1] = multbl[0][state[1][1][0]] ^ multbl[1][state[1][s3-1][1]]
- ^multbl[2][state[1][3][2]] ^ multbl[3][state[1][s1-1][3]];
- ((uint32_t *)(state))[2] = multbl[0][state[1][2][0]] ^ multbl[1][state[1][s3 ][1]]
- ^multbl[2][state[1][0][2]] ^ multbl[3][state[1][s1 ][3]];
- ((uint32_t *)(state))[3] = multbl[0][state[1][3][0]] ^ multbl[1][state[1][s1-1][1]]
- ^multbl[2][state[1][1][2]] ^ multbl[3][state[1][s3-1][3]];
+static inline void mix(av_aes_block state[2], uint32_t multbl[][256], int s1, int s3){
+ uint8_t (*src)[4] = state[1].u8x4;
+ state[0].u32[0] = mix_core(multbl, src[0][0], src[s1 ][1], src[2][2], src[s3 ][3]);
+ state[0].u32[1] = mix_core(multbl, src[1][0], src[s3-1][1], src[3][2], src[s1-1][3]);
+ state[0].u32[2] = mix_core(multbl, src[2][0], src[s3 ][1], src[0][2], src[s1 ][3]);
+ state[0].u32[3] = mix_core(multbl, src[3][0], src[s1-1][1], src[1][2], src[s3-1][3]);
}
-static inline void crypt(AVAES *a, int s, uint8_t *sbox, uint32_t *multbl){
+static inline void crypt(AVAES *a, int s, const uint8_t *sbox, uint32_t multbl[][256]){
int r;
- for(r=a->rounds; r>1; r--){
- addkey(a->state, a->round_key[r]);
+ for(r=a->rounds-1; r>0; r--){
mix(a->state, multbl, 3-s, 1+s);
+ addkey(&a->state[1], &a->state[0], &a->round_key[r]);
}
- addkey(a->state, a->round_key[1]);
- subshift(a->state[0][0], a->state[0][0]+3-s, a->state[0][0]+1+s, sbox);
- addkey(a->state, a->round_key[0]);
+ subshift(&a->state[0], s, sbox);
}
-static void aes_decrypt(AVAES *a){
- crypt(a, 0, inv_sbox, dec_multbl);
-}
-
-static void aes_encrypt(AVAES *a){
- crypt(a, 2, sbox, enc_multbl);
+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]);
+ if(decrypt) {
+ crypt(a, 0, inv_sbox, dec_multbl);
+ if(iv){
+ addkey(&a->state[0], &a->state[0], iv);
+ memcpy(iv, src, 16);
+ }
+ addkey(dst, &a->state[0], &a->round_key[0]);
+ }else{
+ if(iv) addkey(&a->state[1], &a->state[1], iv);
+ crypt(a, 2, sbox, enc_multbl);
+ addkey(dst, &a->state[0], &a->round_key[0]);
+ if(iv) memcpy(iv, dst, 16);
+ }
+ src++;
+ dst++;
+ }
}
-static void init_multbl2(uint8_t tbl[1024], int c[4], uint8_t *log8, uint8_t *alog8, uint8_t *sbox){
+static void init_multbl2(uint8_t tbl[1024], 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]] ];
}
-#ifndef CONFIG_SMALL
+#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];
}
// this is based on the reference AES code by Paulo Barreto and Vincent Rijmen
-int av_aes_init(AVAES *a, uint8_t *key, int key_bits, int decrypt) {
+int av_aes_init(AVAES *a, const uint8_t *key, int key_bits, int decrypt) {
int i, j, t, rconpointer = 0;
uint8_t tk[8][4];
int KC= key_bits>>5;
uint8_t log8[256];
uint8_t alog8[512];
- if(!enc_multbl[4][1023]){
+ if(!enc_multbl[FF_ARRAY_ELEMS(enc_multbl)-1][FF_ARRAY_ELEMS(enc_multbl[0])-1]){
j=1;
for(i=0; i<255; i++){
alog8[i]=
inv_sbox[j]= i;
sbox [i]= j;
}
- init_multbl2(dec_multbl[0], (int[4]){0xe, 0x9, 0xd, 0xb}, log8, alog8, inv_sbox);
- init_multbl2(enc_multbl[0], (int[4]){0x2, 0x1, 0x1, 0x3}, log8, alog8, sbox);
+ init_multbl2(dec_multbl[0], (const int[4]){0xe, 0x9, 0xd, 0xb}, log8, alog8, inv_sbox);
+ init_multbl2(enc_multbl[0], (const int[4]){0x2, 0x1, 0x1, 0x3}, log8, alog8, sbox);
}
if(key_bits!=128 && key_bits!=192 && key_bits!=256)
memcpy(tk, key, KC*4);
- for(t= 0; t < (rounds+1)*4;) {
- memcpy(a->round_key[0][t], tk, KC*4);
- t+= KC;
+ for(t= 0; t < (rounds+1)*16;) {
+ memcpy(a->round_key[0].u8+t, tk, KC*4);
+ t+= KC*4;
for(i = 0; i < 4; i++)
tk[0][i] ^= sbox[tk[KC-1][(i+1)&3]];
if(decrypt){
for(i=1; i<rounds; i++){
- uint8_t tmp[2][16];
- memcpy(tmp[1], a->round_key[i][0], 16);
- subshift(tmp[0], tmp[0]+3, tmp[0]+1, sbox);
- memcpy(tmp[1], tmp[0], 16);
+ av_aes_block tmp[3];
+ memcpy(&tmp[2], &a->round_key[i], 16);
+ subshift(&tmp[1], 0, sbox);
mix(tmp, dec_multbl, 1, 3);
- memcpy(a->round_key[i][0], tmp[0], 16);
+ memcpy(&a->round_key[i], &tmp[0], 16);
}
}else{
for(i=0; i<(rounds+1)>>1; i++){
for(j=0; j<16; j++)
- FFSWAP(int, a->round_key[i][0][j], a->round_key[rounds-i][0][j]);
+ FFSWAP(int, a->round_key[i].u8[j], a->round_key[rounds-i].u8[j]);
}
}
}
#ifdef TEST
+#include "lfg.h"
#include "log.h"
-int main(){
+int main(void){
int i,j;
AVAES ae, ad, b;
uint8_t rkey[2][16]= {
uint8_t rct[2][16]= {
{0x73, 0x22, 0x81, 0xc0, 0xa0, 0xaa, 0xb8, 0xf7, 0xa5, 0x4a, 0x0c, 0x67, 0xa0, 0xc4, 0x5e, 0xcf},
{0x6d, 0x25, 0x1e, 0x69, 0x44, 0xb0, 0x51, 0xe0, 0x4e, 0xaa, 0x6f, 0xb4, 0xdb, 0xf7, 0x84, 0x65}};
+ uint8_t temp[16];
+ AVLFG prng;
av_aes_init(&ae, "PI=3.141592654..", 128, 0);
av_aes_init(&ad, "PI=3.141592654..", 128, 1);
- av_log_level= AV_LOG_DEBUG;
+ av_log_set_level(AV_LOG_DEBUG);
+ av_lfg_init(&prng, 1);
for(i=0; i<2; i++){
av_aes_init(&b, rkey[i], 128, 1);
- memcpy(b.state, rct[i], 16);
- aes_decrypt(&b);
+ av_aes_crypt(&b, temp, rct[i], 1, NULL, 1);
for(j=0; j<16; j++)
- if(rpt[i][j] != b.state[1][0][j])
- av_log(NULL, AV_LOG_ERROR, "%d %02X %02X\n", j, rpt[i][j], b.state[1][0][j]);
+ if(rpt[i][j] != temp[j])
+ av_log(NULL, AV_LOG_ERROR, "%d %02X %02X\n", j, rpt[i][j], temp[j]);
}
for(i=0; i<10000; i++){
for(j=0; j<16; j++){
- pt[j]= random();
+ pt[j] = av_lfg_get(&prng);
}
- memcpy(ae.state, pt, 16);
{START_TIMER
- aes_encrypt(&ae);
+ av_aes_crypt(&ae, temp, pt, 1, NULL, 0);
if(!(i&(i-1)))
- av_log(NULL, AV_LOG_ERROR, "%02X %02X %02X %02X\n", ae.state[1][0][0], ae.state[1][1][1], ae.state[1][2][2], ae.state[1][3][3]);
- memcpy(ad.state[0], ae.state[1], 16);
- aes_decrypt(&ad);
+ av_log(NULL, AV_LOG_ERROR, "%02X %02X %02X %02X\n", temp[0], temp[5], temp[10], temp[15]);
+ av_aes_crypt(&ad, temp, temp, 1, NULL, 1);
STOP_TIMER("aes")}
for(j=0; j<16; j++){
- if(pt[j] != ad.state[1][0][j]){
- av_log(NULL, AV_LOG_ERROR, "%d %d %02X %02X\n", i,j, pt[j], ad.state[1][0][j]);
+ if(pt[j] != temp[j]){
+ av_log(NULL, AV_LOG_ERROR, "%d %d %02X %02X\n", i,j, pt[j], temp[j]);
}
}
}