inline void AES_set_decrypt_key_fast(AES_KEY *dkey, const AES_KEY *ekey) {
int j = 0;
int i = ROUNDS(ekey);
#if (OCB_KEY_LEN == 0)
dkey->rounds = i;
#endif
dkey->rd_key[i--] = ekey->rd_key[j++];
while (i)
dkey->rd_key[i--] = _mm_aesimc_si128(ekey->rd_key[j++]);
dkey->rd_key[i] = ekey->rd_key[j];
}
/*
* AES-128 Key Schedule
*/
void AES_128_NI::key_schedule(const byte key[], size_t)
{
#define AES_128_key_exp(K, RCON) \
aes_128_key_expansion(K, _mm_aeskeygenassist_si128(K, RCON))
__m128i K0 = _mm_loadu_si128((const __m128i*)(key));
__m128i K1 = AES_128_key_exp(K0, 0x01);
__m128i K2 = AES_128_key_exp(K1, 0x02);
__m128i K3 = AES_128_key_exp(K2, 0x04);
__m128i K4 = AES_128_key_exp(K3, 0x08);
__m128i K5 = AES_128_key_exp(K4, 0x10);
__m128i K6 = AES_128_key_exp(K5, 0x20);
__m128i K7 = AES_128_key_exp(K6, 0x40);
__m128i K8 = AES_128_key_exp(K7, 0x80);
__m128i K9 = AES_128_key_exp(K8, 0x1B);
__m128i K10 = AES_128_key_exp(K9, 0x36);
__m128i* EK_mm = (__m128i*)&EK[0];
_mm_storeu_si128(EK_mm , K0);
_mm_storeu_si128(EK_mm + 1, K1);
_mm_storeu_si128(EK_mm + 2, K2);
_mm_storeu_si128(EK_mm + 3, K3);
_mm_storeu_si128(EK_mm + 4, K4);
_mm_storeu_si128(EK_mm + 5, K5);
_mm_storeu_si128(EK_mm + 6, K6);
_mm_storeu_si128(EK_mm + 7, K7);
_mm_storeu_si128(EK_mm + 8, K8);
_mm_storeu_si128(EK_mm + 9, K9);
_mm_storeu_si128(EK_mm + 10, K10);
// Now generate decryption keys
__m128i* DK_mm = (__m128i*)&DK[0];
_mm_storeu_si128(DK_mm , K10);
_mm_storeu_si128(DK_mm + 1, _mm_aesimc_si128(K9));
_mm_storeu_si128(DK_mm + 2, _mm_aesimc_si128(K8));
_mm_storeu_si128(DK_mm + 3, _mm_aesimc_si128(K7));
_mm_storeu_si128(DK_mm + 4, _mm_aesimc_si128(K6));
_mm_storeu_si128(DK_mm + 5, _mm_aesimc_si128(K5));
_mm_storeu_si128(DK_mm + 6, _mm_aesimc_si128(K4));
_mm_storeu_si128(DK_mm + 7, _mm_aesimc_si128(K3));
_mm_storeu_si128(DK_mm + 8, _mm_aesimc_si128(K2));
_mm_storeu_si128(DK_mm + 9, _mm_aesimc_si128(K1));
_mm_storeu_si128(DK_mm + 10, K0);
}
unsigned char *Cryptor::genKeySchedule(const Key &key,
bool encryption) {
ALIGN16 unsigned char *schedule = new unsigned char[15 * 16];
int upper;
switch (key.size) {
case _128_BITS:
upper = 9;
expandKey128(key.key, schedule);
break;
case _192_BITS:
upper = 11;
expandKey192(key.key, schedule);
break;
case _256_BITS:
upper = 13;
expandKey256(key.key, schedule);
break;
}
// Generate decryption round keys by using aesimc
// instruction. This only concerns keys 1-9/11/13. And reverse
// the order for all of them!
if (!encryption) {
__m128i *keySchedule = (__m128i*) schedule;
ALIGN16 unsigned char *tempSchedule =
new unsigned char[15 * 16];
__m128i *tempKeySchedule = (__m128i*) tempSchedule;
tempKeySchedule[upper + 1] = keySchedule[0];
for (int i = 1; i <= upper; i++) {
tempKeySchedule[(upper + 1) - i] =
_mm_aesimc_si128(keySchedule[i]);
}
tempKeySchedule[0] = keySchedule[upper + 1];
// Now use the temp. instead!
delete[] schedule;
return (unsigned char*) tempKeySchedule;
}
return schedule;
}
int AESNI_set_decrypt_key (const unsigned char *userKey, const int bits, AES_KEY *key)
{
int i,nr;;
AES_KEY temp_key;
__m128i *Key_Schedule = (__m128i*)key->KEY;
__m128i *Temp_Key_Schedule = (__m128i*)temp_key.KEY;
mybits=bits;
if (!userKey || !key) return -1;
if (AESNI_set_encrypt_key(userKey,bits,&temp_key) == -2) return -2;
nr = temp_key.nr;
key->nr = nr;
Key_Schedule[nr] = Temp_Key_Schedule[0];
Key_Schedule[nr-1] = _mm_aesimc_si128(Temp_Key_Schedule[1]);
Key_Schedule[nr-2] = _mm_aesimc_si128(Temp_Key_Schedule[2]);
Key_Schedule[nr-3] = _mm_aesimc_si128(Temp_Key_Schedule[3]);
Key_Schedule[nr-4] = _mm_aesimc_si128(Temp_Key_Schedule[4]);
Key_Schedule[nr-5] = _mm_aesimc_si128(Temp_Key_Schedule[5]);
Key_Schedule[nr-6] = _mm_aesimc_si128(Temp_Key_Schedule[6]);
Key_Schedule[nr-7] = _mm_aesimc_si128(Temp_Key_Schedule[7]);
Key_Schedule[nr-8] = _mm_aesimc_si128(Temp_Key_Schedule[8]);
Key_Schedule[nr-9] = _mm_aesimc_si128(Temp_Key_Schedule[9]);
if(nr>10)
{
Key_Schedule[nr-10] = _mm_aesimc_si128(Temp_Key_Schedule[10]);
Key_Schedule[nr-11] = _mm_aesimc_si128(Temp_Key_Schedule[11]);
}
if(nr>12)
{
Key_Schedule[nr-12] = _mm_aesimc_si128(Temp_Key_Schedule[12]);
Key_Schedule[nr-13] = _mm_aesimc_si128(Temp_Key_Schedule[13]);
}
Key_Schedule[0] = Temp_Key_Schedule[nr];
return 0;
}
/*
* AES-256 Key Schedule
*/
void AES_256_NI::key_schedule(const byte key[], size_t)
{
__m128i K0 = _mm_loadu_si128((const __m128i*)(key));
__m128i K1 = _mm_loadu_si128((const __m128i*)(key + 16));
__m128i K2 = aes_128_key_expansion(K0, _mm_aeskeygenassist_si128(K1, 0x01));
__m128i K3 = aes_256_key_expansion(K1, K2);
__m128i K4 = aes_128_key_expansion(K2, _mm_aeskeygenassist_si128(K3, 0x02));
__m128i K5 = aes_256_key_expansion(K3, K4);
__m128i K6 = aes_128_key_expansion(K4, _mm_aeskeygenassist_si128(K5, 0x04));
__m128i K7 = aes_256_key_expansion(K5, K6);
__m128i K8 = aes_128_key_expansion(K6, _mm_aeskeygenassist_si128(K7, 0x08));
__m128i K9 = aes_256_key_expansion(K7, K8);
__m128i K10 = aes_128_key_expansion(K8, _mm_aeskeygenassist_si128(K9, 0x10));
__m128i K11 = aes_256_key_expansion(K9, K10);
__m128i K12 = aes_128_key_expansion(K10, _mm_aeskeygenassist_si128(K11, 0x20));
__m128i K13 = aes_256_key_expansion(K11, K12);
__m128i K14 = aes_128_key_expansion(K12, _mm_aeskeygenassist_si128(K13, 0x40));
__m128i* EK_mm = (__m128i*)&EK[0];
_mm_storeu_si128(EK_mm , K0);
_mm_storeu_si128(EK_mm + 1, K1);
_mm_storeu_si128(EK_mm + 2, K2);
_mm_storeu_si128(EK_mm + 3, K3);
_mm_storeu_si128(EK_mm + 4, K4);
_mm_storeu_si128(EK_mm + 5, K5);
_mm_storeu_si128(EK_mm + 6, K6);
_mm_storeu_si128(EK_mm + 7, K7);
_mm_storeu_si128(EK_mm + 8, K8);
_mm_storeu_si128(EK_mm + 9, K9);
_mm_storeu_si128(EK_mm + 10, K10);
_mm_storeu_si128(EK_mm + 11, K11);
_mm_storeu_si128(EK_mm + 12, K12);
_mm_storeu_si128(EK_mm + 13, K13);
_mm_storeu_si128(EK_mm + 14, K14);
// Now generate decryption keys
__m128i* DK_mm = (__m128i*)&DK[0];
_mm_storeu_si128(DK_mm , K14);
_mm_storeu_si128(DK_mm + 1, _mm_aesimc_si128(K13));
_mm_storeu_si128(DK_mm + 2, _mm_aesimc_si128(K12));
_mm_storeu_si128(DK_mm + 3, _mm_aesimc_si128(K11));
_mm_storeu_si128(DK_mm + 4, _mm_aesimc_si128(K10));
_mm_storeu_si128(DK_mm + 5, _mm_aesimc_si128(K9));
_mm_storeu_si128(DK_mm + 6, _mm_aesimc_si128(K8));
_mm_storeu_si128(DK_mm + 7, _mm_aesimc_si128(K7));
_mm_storeu_si128(DK_mm + 8, _mm_aesimc_si128(K6));
_mm_storeu_si128(DK_mm + 9, _mm_aesimc_si128(K5));
_mm_storeu_si128(DK_mm + 10, _mm_aesimc_si128(K4));
_mm_storeu_si128(DK_mm + 11, _mm_aesimc_si128(K3));
_mm_storeu_si128(DK_mm + 12, _mm_aesimc_si128(K2));
_mm_storeu_si128(DK_mm + 13, _mm_aesimc_si128(K1));
_mm_storeu_si128(DK_mm + 14, K0);
}
/*
* AES-192 Key Schedule
*/
void AES_192_NI::key_schedule(const byte key[], size_t)
{
__m128i K0 = _mm_loadu_si128((const __m128i*)(key));
__m128i K1 = _mm_loadu_si128((const __m128i*)(key + 8));
K1 = _mm_srli_si128(K1, 8);
load_le(&EK[0], key, 6);
#define AES_192_key_exp(RCON, EK_OFF) \
aes_192_key_expansion(&K0, &K1, \
_mm_aeskeygenassist_si128(K1, RCON), \
EK + EK_OFF, EK_OFF == 48)
AES_192_key_exp(0x01, 6);
AES_192_key_exp(0x02, 12);
AES_192_key_exp(0x04, 18);
AES_192_key_exp(0x08, 24);
AES_192_key_exp(0x10, 30);
AES_192_key_exp(0x20, 36);
AES_192_key_exp(0x40, 42);
AES_192_key_exp(0x80, 48);
// Now generate decryption keys
const __m128i* EK_mm = (const __m128i*)&EK[0];
__m128i* DK_mm = (__m128i*)&DK[0];
_mm_storeu_si128(DK_mm , _mm_loadu_si128(EK_mm + 12));
_mm_storeu_si128(DK_mm + 1, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 11)));
_mm_storeu_si128(DK_mm + 2, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 10)));
_mm_storeu_si128(DK_mm + 3, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 9)));
_mm_storeu_si128(DK_mm + 4, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 8)));
_mm_storeu_si128(DK_mm + 5, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 7)));
_mm_storeu_si128(DK_mm + 6, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 6)));
_mm_storeu_si128(DK_mm + 7, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 5)));
_mm_storeu_si128(DK_mm + 8, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 4)));
_mm_storeu_si128(DK_mm + 9, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 3)));
_mm_storeu_si128(DK_mm + 10, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 2)));
_mm_storeu_si128(DK_mm + 11, _mm_aesimc_si128(_mm_loadu_si128(EK_mm + 1)));
_mm_storeu_si128(DK_mm + 12, _mm_loadu_si128(EK_mm + 0));
}
void __fastcall aes_AES256_derive_decryption_keys_(
const AES_AES256_RoundKeys* encryption_keys,
AES_AES256_RoundKeys* decryption_keys)
{
decryption_keys->keys[0] = encryption_keys->keys[14];
decryption_keys->keys[1] = _mm_aesimc_si128(encryption_keys->keys[13]);
decryption_keys->keys[2] = _mm_aesimc_si128(encryption_keys->keys[12]);
decryption_keys->keys[3] = _mm_aesimc_si128(encryption_keys->keys[11]);
decryption_keys->keys[4] = _mm_aesimc_si128(encryption_keys->keys[10]);
decryption_keys->keys[5] = _mm_aesimc_si128(encryption_keys->keys[9]);
decryption_keys->keys[6] = _mm_aesimc_si128(encryption_keys->keys[8]);
decryption_keys->keys[7] = _mm_aesimc_si128(encryption_keys->keys[7]);
decryption_keys->keys[8] = _mm_aesimc_si128(encryption_keys->keys[6]);
decryption_keys->keys[9] = _mm_aesimc_si128(encryption_keys->keys[5]);
decryption_keys->keys[10] = _mm_aesimc_si128(encryption_keys->keys[4]);
decryption_keys->keys[11] = _mm_aesimc_si128(encryption_keys->keys[3]);
decryption_keys->keys[12] = _mm_aesimc_si128(encryption_keys->keys[2]);
decryption_keys->keys[13] = _mm_aesimc_si128(encryption_keys->keys[1]);
decryption_keys->keys[14] = encryption_keys->keys[0];
}
