//#############################################################################################################################################################################################################
#include "AES.h"
//=============================================================================================================================================================================================================
const uint8_t AES_INV_SBOX[256] PROGMEM= {
	 0x52, 0x09, 0x6A, 0xD5, 0x30, 0x36, 0xA5, 0x38, 0xBF, 0x40, 0xA3, 0x9E, 0x81, 0xF3, 0xD7, 0xFB ,
	 0x7C, 0xE3, 0x39, 0x82, 0x9B, 0x2F, 0xFF, 0x87, 0x34, 0x8E, 0x43, 0x44, 0xC4, 0xDE, 0xE9, 0xCB ,
	 0x54, 0x7B, 0x94, 0x32, 0xA6, 0xC2, 0x23, 0x3D, 0xEE, 0x4C, 0x95, 0x0B, 0x42, 0xFA, 0xC3, 0x4E ,
	 0x08, 0x2E, 0xA1, 0x66, 0x28, 0xD9, 0x24, 0xB2, 0x76, 0x5B, 0xA2, 0x49, 0x6D, 0x8B, 0xD1, 0x25 ,
	 0x72, 0xF8, 0xF6, 0x64, 0x86, 0x68, 0x98, 0x16, 0xD4, 0xA4, 0x5C, 0xCC, 0x5D, 0x65, 0xB6, 0x92 ,
	 0x6C, 0x70, 0x48, 0x50, 0xFD, 0xED, 0xB9, 0xDA, 0x5E, 0x15, 0x46, 0x57, 0xA7, 0x8D, 0x9D, 0x84 ,
	 0x90, 0xD8, 0xAB, 0x00, 0x8C, 0xBC, 0xD3, 0x0A, 0xF7, 0xE4, 0x58, 0x05, 0xB8, 0xB3, 0x45, 0x06 ,
	 0xD0, 0x2C, 0x1E, 0x8F, 0xCA, 0x3F, 0x0F, 0x02, 0xC1, 0xAF, 0xBD, 0x03, 0x01, 0x13, 0x8A, 0x6B ,
	 0x3A, 0x91, 0x11, 0x41, 0x4F, 0x67, 0xDC, 0xEA, 0x97, 0xF2, 0xCF, 0xCE, 0xF0, 0xB4, 0xE6, 0x73 ,
	 0x96, 0xAC, 0x74, 0x22, 0xE7, 0xAD, 0x35, 0x85, 0xE2, 0xF9, 0x37, 0xE8, 0x1C, 0x75, 0xDF, 0x6E ,
	 0x47, 0xF1, 0x1A, 0x71, 0x1D, 0x29, 0xC5, 0x89, 0x6F, 0xB7, 0x62, 0x0E, 0xAA, 0x18, 0xBE, 0x1B ,
	 0xFC, 0x56, 0x3E, 0x4B, 0xC6, 0xD2, 0x79, 0x20, 0x9A, 0xDB, 0xC0, 0xFE, 0x78, 0xCD, 0x5A, 0xF4 ,
	 0x1F, 0xDD, 0xA8, 0x33, 0x88, 0x07, 0xC7, 0x31, 0xB1, 0x12, 0x10, 0x59, 0x27, 0x80, 0xEC, 0x5F ,
	 0x60, 0x51, 0x7F, 0xA9, 0x19, 0xB5, 0x4A, 0x0D, 0x2D, 0xE5, 0x7A, 0x9F, 0x93, 0xC9, 0x9C, 0xEF ,
	 0xA0, 0xE0, 0x3B, 0x4D, 0xAE, 0x2A, 0xF5, 0xB0, 0xC8, 0xEB, 0xBB, 0x3C, 0x83, 0x53, 0x99, 0x61 ,
	 0x17, 0x2B, 0x04, 0x7E, 0xBA, 0x77, 0xD6, 0x26, 0xE1, 0x69, 0x14, 0x63, 0x55, 0x21, 0x0C, 0x7D 
};
//=============================================================================================================================================================================================================
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//=============================================================================================================================================================================================================
//Static functions

inline uint8_t AES_RotWord(uint8_t *Word);
uint8_t AES_SubWord (uint8_t Word);
inline void AES_KeyExpansion(uint8_t *AES_KEY, uint8_t *w);

//uint8_t AES_xTime(uint8_t Value);
void 	AES_AddRoundKey (uint8_t *State, uint8_t *Key);

void    AES_InvSubBytes(uint8_t *state);
void    AES_InvShiftRows(uint8_t *state);
inline void    AES_InvMixColumns(uint8_t *state);

//void	AES_ApplyIV(uint8_t *state, uint8_t *IV);
//=============================================================================================================================================================================================================
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//=============================================================================================================================================================================================================
inline uint8_t AES_RotWord(uint8_t *Word){
	Word[0+4] = Word[1];
	Word[1+4] = Word[2];
	Word[2+4] = Word[3];
	Word[3+4] = Word[0];
	return Word[1];
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
uint8_t AES_SubWord (uint8_t Word)
{
    //for (uint8_t i = 0; i < 256; i++){	//FIND BETTER SOLUTION (with out using precomputed values)
	//			if (flash_read_byte((AES_INV_SBOX),i) == Word)
	//				return i;
	//}
	//return 0;
	uint8_t j,i = 0;
	do {
		if (flash_read_byte((AES_INV_SBOX),i) == Word)
			j = i;
		i++;	
	} while (i != 0);
	return j;
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
const uint8_t Rcon[7] PROGMEM= { 0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40};//, 0x80};//, 0x1B, 0x36 };

inline void AES_KeyExpansion(uint8_t *AES_KEY, uint8_t *w)
{
	uint8_t KeyTmp;
	uint8_t I;

	//for (I = 0; I < AES_NK_256BIT*4; I++)
	//{
	//	w[I] = pgm_read_byte(&(AES_KEY[I]));
	//}
	_memcopy(AES_KEY,w,AES_NK_256BIT*4);

	for (I = AES_NK_256BIT*4; I < (uint8_t)(AES_NB*(uint8_t)(AES_NR_256BIT+1U)*4); I++)
	{
		KeyTmp = w[I-4];

		if ((I & 0b11100) == 0 ) { 
			if ((I & 0b11) == 0 ) {
				KeyTmp = AES_RotWord(&w[I-4]); 
				KeyTmp = AES_SubWord(KeyTmp);
				//KeyTmp = KeyTmp ^ (1<<(I>>5)); 
				KeyTmp = KeyTmp ^ flash_read_byte((Rcon),(I>>5)-1); 
			} else
				KeyTmp = AES_SubWord(w[I]); 	
		}
		else if ( (I & 0b11100) == 16 ) 
			KeyTmp = AES_SubWord(w[I-4]);

		w[I] = w[I-(AES_NK_256BIT*4)] ^ KeyTmp;
	}
}
//=============================================================================================================================================================================================================
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//=============================================================================================================================================================================================================
//uint8_t AES_xTime(uint8_t Value) { return ( (Value & 0x80) ? ((uint8_t)(Value<<1)^0x1B) : (uint8_t)(Value<<1) ); }
//
//#define AES_MULTIPLY_02(X)			AES_xTime((X))
//#define AES_MULTIPLY_03(X)			( (X) ^ AES_xTime((X)) )
//#define AES_MULTIPLY_09(X)			( (X) ^ AES_xTime(AES_xTime(AES_xTime((X)))) )
//#define AES_MULTIPLY_0B(X)			( (X) ^ AES_xTime((X)) ^ AES_xTime(AES_xTime(AES_xTime((X)))) )
//#define AES_MULTIPLY_0E(X)			( AES_xTime((X)) ^ AES_xTime(AES_xTime((X))) ^ AES_xTime(AES_xTime(AES_xTime((X)))) )
//#define AES_MULTIPLY_0D(X)			( (X) ^ AES_xTime(AES_xTime((X))) ^ AES_xTime(AES_xTime(AES_xTime((X)))) )
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//static uint8_t rstate[4*AES_NB];

#define GF28_MUL2(a) { asm ("lsl %0 \n brcc .+2 \n eor %0, %1 \n" : "=r" (a) : "r" (c0x1b), "0" (a)); }
//#define GF28_MUL2(a) { if (a&0x80) { a = (a<<1) ^ 0x1b; } else { a <<= 1; } }
//#define GF28_MUL2(a) {a=AES_xTime(a);}
 
inline void AES_InvMixColumns(uint8_t *state)
{
	//for (uint8_t i = 0; i<16; i++)
	//	rstate[i]=state[i];
	//_memcopy(state,rstate,16);

	const uint8_t c0x1b = 0x1b;
	uint8_t i, a0, a1, a2, a3, sum, b, c, d, e;
	uint8_t* ptr = state;
	for(i = 0; i < 16; i+=4)	//https://github.com/kildom/AVR-AES/blob/master/aes.c	I don't fully understand this, but it works really fast
	{
		a0 = ptr[0];
		a1 = ptr[1];
		a2 = ptr[2];
		a3 = ptr[3];
		sum = a0 ^ a1 ^ a2 ^ a3;
		c = sum; GF28_MUL2(c);
		d = c ^ a0 ^ a2; GF28_MUL2(d); GF28_MUL2(d);
		d ^= sum;
		e = c ^ a1 ^ a3; GF28_MUL2(e); GF28_MUL2(e);
		e ^= sum;
		b = a0^a1; GF28_MUL2(b);
		*ptr++ = d ^ a0 ^ b;
		b = a1^a2; GF28_MUL2(b);
		*ptr++ = e ^ a1 ^ b;
		b = a2^a3; GF28_MUL2(b);
		*ptr++ = d ^ a2 ^ b;
		b = a3^a0; GF28_MUL2(b);
		*ptr++ = e ^ a3 ^ b;
	}

	//for (uint8_t i = 0; i < 16; i++){
	//			state[i] =	  AES_MULTIPLY_0E(rstate[(i&0xFC)+((i+0)&0b11)]) 
	//						^ AES_MULTIPLY_0B(rstate[(i&0xFC)+((i+1)&0b11)]) 
	//						^ AES_MULTIPLY_0D(rstate[(i&0xFC)+((i+2)&0b11)]) 
	//						^ AES_MULTIPLY_09(rstate[(i&0xFC)+((i+3)&0b11)]);
	//}
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
void AES_AddRoundKey (uint8_t *State, uint8_t *_key)
{
	for (uint8_t i = 0; i < 16; i++)
		State[i] ^=  (_key[i]);
}
//=============================================================================================================================================================================================================
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//=============================================================================================================================================================================================================
void AES_InvShiftRows(uint8_t *state)
{
	uint8_t ByteTmp1;
	uint8_t ByteTmp2;

	//Do not loop!!! Otherwise bad performance
	ByteTmp1 = state[13];
	state[13] = state[9];
	state[9] = state[5];
	state[5] = state[1];
	state[1] = ByteTmp1;

	ByteTmp1 = state[14];
	ByteTmp2 = state[10];
	state[14] = state[6];
	state[10] = state[2];
	state[6] = ByteTmp1;
	state[2] = ByteTmp2;

	ByteTmp1 = state[3];
	state[3] = state[7];
	state[7] = state[11];
	state[11] = state[15];
	state[15] = ByteTmp1;
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
void AES_InvSubBytes(uint8_t *state)
{
	for (uint8_t i = 0; i < 16; i++)
			state[i]  = flash_read_byte((AES_INV_SBOX),state[i]);
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
static uint8_t NextIV[16];

void AES_InvCipher(uint8_t *IN, uint8_t *w, uint8_t *IV)
{
	uint8_t *state = IN;

	_memcopy(IN,NextIV,16);

	AES_AddRoundKey(state, &w[AES_NR_256BIT*AES_NB*4]);
	for (uint8_t I = (uint8_t)(AES_NR_256BIT-1); I > 0; I--)
	{
		AES_InvShiftRows(state);
		AES_InvSubBytes(state);
		AES_AddRoundKey(state, &w[I<<4]);
		AES_InvMixColumns(state);
	}
	AES_InvShiftRows(state);
	AES_InvSubBytes(state);
	AES_AddRoundKey(state, &w[0]);

	AES_AddRoundKey(state, IV);		//XOR output with IV

	_memcopy(NextIV,IV,16);
}
//=============================================================================================================================================================================================================
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//=============================================================================================================================================================================================================
uint8_t w[(uint8_t)(AES_NB * (uint8_t)(AES_NR_256BIT+1U))*4];

uint8_t IVstored[16];

void AES_Decrypt(uint8_t *InpData, uint16_t InpDataLength)
{
	//if (KeyLength != AES_256) return;		//Only AES CBC 256 is supported
	if ( (InpDataLength & 0b1111) != 0 ) { return; }

	uint16_t DataPos = 0;
	while ( InpDataLength > 0 )
	{
		AES_InvCipher(&InpData[DataPos], w, IVstored);
		DataPos = (uint16_t)(DataPos + AES_BLOCK_SIZE);
		InpDataLength = (uint16_t)(InpDataLength - AES_BLOCK_SIZE);
	}
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
void AES_PrepareKeys(uint8_t *Key, uint8_t *IV){
	AES_KeyExpansion(Key, w);
	//AES_Copy(IV,IVstored);
	//for (uint8_t i = 0; i < 16; i++){
	//	IVstored[i] = pgm_read_byte(&(IV[i]));
	//}
	_memcopy(IV,IVstored,16);
}
//-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
//#############################################################################################################################################################################################################