X-Git-Url: https://git.sesse.net/?a=blobdiff_plain;f=libavutil%2Faes.c;h=d3a271c379d052a6f34fa5aa32453261027e7927;hb=d691da95082d96f25a5f470ede9fa737e2d70c6c;hp=78683b7aca0cae1ca82b67658931557ca727b3fa;hpb=cf5773e95f784514a3e69c4cb3fec3027162ad0b;p=ffmpeg diff --git a/libavutil/aes.c b/libavutil/aes.c index 78683b7aca0..d3a271c379d 100644 --- a/libavutil/aes.c +++ b/libavutil/aes.c @@ -1,5 +1,7 @@ /* - * copyright (c) 2007 Michael Niedermayer and Reimar Doeffinger + * copyright (c) 2007 Michael Niedermayer + * + * some optimization ideas from aes128.c by Reimar Doeffinger * * This file is part of FFmpeg. * @@ -19,22 +21,25 @@ */ #include "common.h" -#include "log.h" #include "aes.h" 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). uint8_t round_key[15][4][4]; - uint8_t state[4][4]; + uint8_t state[2][4][4]; int rounds; }AVAES; -static const uint8_t rcon[11] = { - 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36, 0x6c +const int av_aes_size= sizeof(AVAES); + +static const uint8_t rcon[10] = { + 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1b, 0x36 }; 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 @@ -42,80 +47,90 @@ static uint32_t enc_multbl[4][256]; static uint32_t dec_multbl[4][256]; #endif -static inline void addkey(uint64_t state[2], uint64_t round_key[2]){ - state[0] ^= round_key[0]; - state[1] ^= round_key[1]; +static inline void addkey(uint64_t dst[2], const uint64_t src[2], const uint64_t round_key[2]){ + dst[0] = src[0] ^ round_key[0]; + dst[1] = src[1] ^ round_key[1]; } -#define SUBSHIFT0(s, box) s[0]=box[s[ 0]]; s[ 4]=box[s[ 4]]; s[ 8]=box[s[ 8]]; s[12]=box[s[12]]; -#define SUBSHIFT1(s, box) t=s[0]; s[0]=box[s[ 4]]; s[ 4]=box[s[ 8]]; s[ 8]=box[s[12]]; s[12]=box[t]; -#define SUBSHIFT2(s, box) t=s[0]; s[0]=box[s[ 8]]; s[ 8]=box[ t]; t=s[ 4]; s[ 4]=box[s[12]]; s[12]=box[t]; -#define SUBSHIFT3(s, box) t=s[0]; s[0]=box[s[12]]; s[12]=box[s[ 8]]; s[ 8]=box[s[ 4]]; s[ 4]=box[t]; - -#define ROT(x,s) ((x>>s)|(x<<(32-s)) - -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); +static void subshift(uint8_t s0[2][16], int s, const uint8_t *box){ + uint8_t (*s1)[16]= s0[0] - s; + uint8_t (*s3)[16]= s0[0] + s; + 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][3]=box[s1[1][ 7]]; s1[0][ 7]=box[s1[1][11]]; s1[0][11]=box[s1[1][15]]; s1[0][15]=box[s1[1][ 3]]; + 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][1]=box[s3[1][13]]; s3[0][13]=box[s3[1][ 9]]; s3[0][ 9]=box[s3[1][ 5]]; s3[0][ 5]=box[s3[1][ 1]]; +} +static inline int mix_core(uint32_t multbl[4][256], int a, int b, int c, int d){ +#if CONFIG_SMALL +#define ROT(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 - ((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 } -void av_aes_decrypt(AVAES *a){ - int t, r; - - for(r=a->rounds-1; r>=0; r--){ - if(r==a->rounds-1) - addkey(a->state, a->round_key[r+1]); - else - 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]); +static inline void mix(uint8_t state[2][4][4], uint32_t multbl[4][256], int s1, int s3){ + ((uint32_t *)(state))[0] = mix_core(multbl, state[1][0][0], state[1][s1 ][1], state[1][2][2], state[1][s3 ][3]); + ((uint32_t *)(state))[1] = mix_core(multbl, state[1][1][0], state[1][s3-1][1], state[1][3][2], state[1][s1-1][3]); + ((uint32_t *)(state))[2] = mix_core(multbl, state[1][2][0], state[1][s3 ][1], state[1][0][2], state[1][s1 ][3]); + ((uint32_t *)(state))[3] = mix_core(multbl, state[1][3][0], state[1][s1-1][1], state[1][1][2], state[1][s3-1][3]); +} + +static inline void crypt(AVAES *a, int s, const uint8_t *sbox, const uint32_t *multbl){ + int 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]); } + subshift(a->state[0][0], s, sbox); } -void av_aes_encrypt(AVAES *a){ - int r, t; - - for(r=0; rrounds; r++){ - addkey(a->state, a->round_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) - if(r==a->rounds-1) - addkey(a->state, a->round_key[r+1]); - else - mix(a->state, enc_multbl); //FIXME replace log8 by const / optimze mix as this can be simplified alot +void av_aes_crypt(AVAES *a, uint8_t *dst, const uint8_t *src, int count, uint8_t *iv, int decrypt){ + 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+=16; + dst+=16; } } -static init_multbl(uint8_t tbl[256], 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]] ]; +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]] ]; + } +#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]; +#endif } // this is based on the reference AES code by Paulo Barreto and Vincent Rijmen -AVAES *av_aes_init(uint8_t *key, int key_bits) { - AVAES *a; +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/32; + int KC= key_bits>>5; int rounds= KC + 6; uint8_t log8[256]; uint8_t alog8[512]; - if(!sbox[255]){ + if(!enc_multbl[0][sizeof(enc_multbl)/sizeof(enc_multbl[0][0])-1]){ j=1; for(i=0; i<255; i++){ alog8[i]= @@ -124,74 +139,103 @@ AVAES *av_aes_init(uint8_t *key, int key_bits) { j^= j+j; if(j>255) j^= 0x11B; } - log8[0]= 255; for(i=0; i<256; i++){ j= i ? alog8[255-log8[i]] : 0; j ^= (j<<1) ^ (j<<2) ^ (j<<3) ^ (j<<4); j = (j ^ (j>>8) ^ 99) & 255; inv_sbox[j]= i; 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_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_multbl(enc_multbl[0], (int[4]){0x2, 0x1, 0x1, 0x3}, log8, alog8); - init_multbl(enc_multbl[1], (int[4]){0x3, 0x2, 0x1, 0x1}, log8, alog8); - init_multbl(enc_multbl[2], (int[4]){0x1, 0x3, 0x2, 0x1}, log8, alog8); - init_multbl(enc_multbl[3], (int[4]){0x1, 0x1, 0x3, 0x2}, log8, alog8); + 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) - return NULL; + return -1; - a= av_malloc(sizeof(AVAES)); a->rounds= rounds; 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][0]+t, tk, KC*4); + 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(j = 1; j < KC; j++){ - if(KC != 8 || j != KC/2) + if(KC != 8 || j != KC>>1) for(i = 0; i < 4; i++) tk[j][i] ^= tk[j-1][i]; else for(i = 0; i < 4; i++) tk[j][i] ^= sbox[tk[j-1][i]]; } } - return a; + + if(decrypt){ + for(i=1; iround_key[i][0], 16); + subshift(tmp[1], 0, sbox); + mix(tmp, dec_multbl, 1, 3); + memcpy(a->round_key[i][0], 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]); + } + } + + return 0; } #ifdef TEST +#include "lfg.h" +#include "log.h" -int main(){ - int i,j,k; - AVAES *a= av_aes_init("PI=3.141592654..", 128); - uint8_t ct[16], pt[16], key[32]; - - av_log_level= AV_LOG_DEBUG; +int main(void){ + int i,j; + AVAES ae, ad, b; + uint8_t rkey[2][16]= { + {0}, + {0x10, 0xa5, 0x88, 0x69, 0xd7, 0x4b, 0xe5, 0xa3, 0x74, 0xcf, 0x86, 0x7c, 0xfb, 0x47, 0x38, 0x59}}; + uint8_t pt[16], rpt[2][16]= { + {0x6a, 0x84, 0x86, 0x7c, 0xd7, 0x7e, 0x12, 0xad, 0x07, 0xea, 0x1b, 0xe8, 0x95, 0xc5, 0x3f, 0xa3}, + {0}}; + 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_set_level(AV_LOG_DEBUG); + av_lfg_init(&prng, 1); + + for(i=0; i<2; i++){ + av_aes_init(&b, rkey[i], 128, 1); + av_aes_crypt(&b, temp, rct[i], 1, NULL, 1); + for(j=0; j<16; 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(a->state, pt, 16); {START_TIMER - av_aes_encrypt(a); + av_aes_crypt(&ae, temp, pt, 1, NULL, 0); if(!(i&(i-1))) - av_log(NULL, AV_LOG_ERROR, "%02X %02X %02X %02X\n", a->state[0][0], a->state[1][1], a->state[2][2], a->state[3][3]); - av_aes_decrypt(a); + 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] != a->state[0][j]){ - av_log(NULL, AV_LOG_ERROR, "%d %d %02X %02X\n", i,j, pt[j], a->state[0][j]); + if(pt[j] != temp[j]){ + av_log(NULL, AV_LOG_ERROR, "%d %d %02X %02X\n", i,j, pt[j], temp[j]); } } }