/******************************************************************************/
/* SlunkCrypt, by LoRd_MuldeR <MuldeR2@GMX.de>                                */
/* This work has been released under the CC0 1.0 Universal license!           */
/******************************************************************************/

/* Internal */
#include "slunkcrypt.h"
#include "compiler.h"
#include "keygen.h"
#include "thread.h"
#include "version.h"

/* CRT */
#include <string.h>
#include <limits.h>

/* Version */
const uint16_t SLUNKCRYPT_VERSION_MAJOR = LIB_VERSION_MAJOR;
const uint16_t SLUNKCRYPT_VERSION_MINOR = LIB_VERSION_MINOR;
const uint16_t SLUNKCRYPT_VERSION_PATCH = LIB_VERSION_PATCH;
const char *const SLUNKCRYPT_BUILD = __DATE__ ", " __TIME__;

/* Utilities */
#define BOOLIFY(X) (!!(X))

// ==========================================================================
// Data structures
// ==========================================================================

typedef struct
{
	uint32_t x, y, z, w, v, d;
}
rand_state_t;

typedef struct
{
	int reverse_mode;
	const uint8_t (*wheel)[256U];
	uint32_t counter;
	rand_state_t random;
	uint8_t *data;
}
thread_state_t;

typedef struct
{
	uint8_t wheel[256U][256U];
	thrdpl_t thread_pool;
	size_t thread_idx;
	thread_state_t thread_data[MAX_THREADS];
}
crypt_state_t;

#define THREAD_COUNT 1U

// ==========================================================================
// Abort flag
// ==========================================================================

volatile int g_slunkcrypt_abort_flag = 0;

#define CHECK_ABORTED() do \
{ \
	if (g_slunkcrypt_abort_flag) \
	{ \
		goto aborted; \
	} \
} \
while (0)

// ==========================================================================
// Byte access (endianness agnostic)
// ==========================================================================

static INLINE uint32_t lower_u64(const uint64_t value)
{
	return (uint32_t)(value & 0xFFFFFFFF);
}

static INLINE uint32_t upper_u64(const uint64_t value)
{
	return (uint32_t)(value >> 32U);
}

// ==========================================================================
// Deterministic random bit generator
// ==========================================================================

static INLINE void random_init(rand_state_t *const state, const keydata_t *const key)
{
	slunkcrypt_bzero(state, sizeof(rand_state_t));
	state->x = lower_u64(key->a);
	state->y = upper_u64(key->a);
	state->z = lower_u64(key->b);
	state->w = upper_u64(key->b);
	state->v = lower_u64(key->c);
	state->d = upper_u64(key->c);
}

static INLINE uint32_t random_next(rand_state_t *const state)
{
	const uint32_t t = state->x ^ (state->x >> 2);
	state->x = state->y;
	state->y = state->z;
	state->z = state->w;
	state->w = state->v;
	state->v ^= (state->v << 4) ^ t ^ (t << 1);
	return (state->d += 0x000587C5) + state->v;
}

static INLINE void random_seed(rand_state_t *const state, uint64_t salt, const uint16_t pepper, const uint8_t *const passwd, const size_t passwd_len)
{
	size_t i;
	keydata_t key;
	do
	{
		slunkcrypt_keygen(&key, salt++, pepper, passwd, passwd_len);
		random_init(state, &key);
		slunkcrypt_bzero(&key, sizeof(keydata_t));
	}
	while (!(state->x || state->y || state->z || state->w || state->v));
	for (i = 0U; i < 97U; ++i)
	{
		UNUSED volatile uint32_t q = random_next(state);
	}
}

// ==========================================================================
// Initialization
// ==========================================================================

static int initialize_state(crypt_state_t *const state, const uint64_t nonce, const uint8_t *const passwd, const size_t passwd_len, const int mode, const int reset)
{
	uint8_t temp[256U][256U];
	size_t r, i;
	rand_state_t random;
	uint32_t counter;
	const int reverse_mode = BOOLIFY(mode);

	/* backup previous value */
	const thrdpl_t thread_pool = reset ? state->thread_pool : THRDPL_NULL;

	/* initialize state */
	slunkcrypt_bzero(state, sizeof(crypt_state_t));

	/* create thread-pool */
	if ((state->thread_pool = reset ? thread_pool : thrdpl_create(THREAD_COUNT)) == THRDPL_NULL)
	{
		return SLUNKCRYPT_FAILURE;
	}

	/* initialize counter */
	random_seed(&random, nonce, (uint16_t)(-1), passwd, passwd_len);
	counter = random_next(&random);

	/* set up the wheel permutations */
	for (r = 0U; r < 256U; ++r)
	{
		random_seed(&random, nonce, (uint16_t)r, passwd, passwd_len);
		for (i = 0U; i < 256U; ++i)
		{
			const size_t j = random_next(&random) % (i + 1U);
			if (j != i)
			{
				state->wheel[r][i] = state->wheel[r][j];
			}
			state->wheel[r][j] = (uint8_t)i;
		}
		CHECK_ABORTED();
	}

	/* reverse the wheels, if requested */
	if (reverse_mode)
	{
		for (r = 0U; r < 256U; ++r)
		{
			for (i = 0U; i < 256U; ++i)
			{
				temp[r][state->wheel[r][i]] = (uint8_t)i;
			}
		}
		for (r = 0U; r < 256U; ++r)
		{
			memcpy(state->wheel[255U - r], temp[r], 256U);
		}
		slunkcrypt_bzero(temp, sizeof(temp));
		CHECK_ABORTED();
	}

	/* set up thread state */
	random_seed(&random, nonce, 256U, passwd, passwd_len);
	for (i = 0U; i < THREAD_COUNT; ++i)
	{
		state->thread_data[i].reverse_mode = reverse_mode;
		state->thread_data[i].wheel = state->wheel;
		state->thread_data[i].counter = counter + ((uint32_t)i);
		memcpy(&state->thread_data[i].random, &random, sizeof(rand_state_t));
		for (r = 0U; r < i * 63U; ++r)
		{
			random_next(&state->thread_data[i].random);
		}
		CHECK_ABORTED();
	}

	slunkcrypt_bzero(&counter, sizeof(uint32_t));
	slunkcrypt_bzero(&random, sizeof(rand_state_t));

	return SLUNKCRYPT_SUCCESS;

	/* aborted */
aborted:
	thrdpl_destroy(state->thread_pool);
	slunkcrypt_bzero(state, sizeof(crypt_state_t));
	slunkcrypt_bzero(&counter, sizeof(uint32_t));
	slunkcrypt_bzero(&random, sizeof(rand_state_t));
	return SLUNKCRYPT_ABORTED;
}

// ==========================================================================
// Encrypt / Decrypt
// ==========================================================================

static INLINE void update_offset(uint8_t *const offset, uint32_t seed, rand_state_t *const state, const int reverse)
{
	size_t i;
	for (i = 0U; i < 256U; ++i, seed >>= CHAR_BIT)
	{
		if (i && (!(i & 3U)))
		{
			seed = random_next(state);
		}
		offset[reverse ? (255U - i) : i] = (uint8_t)seed;
	}
	for (i = 0U; i < 63U * (THREAD_COUNT - 1U); ++i)
	{
		random_next(state);
	}
}

static INLINE void process_next_symbol(thread_state_t *const state)
{
	uint8_t offset[256U];
	size_t i;
	update_offset(offset, state->counter, &state->random, state->reverse_mode);
	for (i = 0U; i < 256U; ++i)
	{
		*state->data = (state->wheel[i][(*state->data + offset[i]) & 0xFF] - offset[i]) & 0xFF;
	}
	state->counter += THREAD_COUNT;
}

// ==========================================================================
// Public API
// ==========================================================================

int slunkcrypt_generate_nonce(uint64_t *const nonce)
{
	if (!nonce)
	{
		return SLUNKCRYPT_FAILURE;
	}
	do
	{
		if (slunkcrypt_random_bytes((uint8_t*)nonce, sizeof(uint64_t)) != sizeof(uint64_t))
		{
			return SLUNKCRYPT_FAILURE;
		}
	}
	while (!(*nonce));
	return SLUNKCRYPT_SUCCESS;
}

slunkcrypt_t slunkcrypt_alloc(const uint64_t nonce, const uint8_t *const passwd, const size_t passwd_len, const int mode)
{
	crypt_state_t* state = NULL;
	if ((!passwd) || (passwd_len < SLUNKCRYPT_PWDLEN_MIN) || (passwd_len > SLUNKCRYPT_PWDLEN_MAX) || (mode < SLUNKCRYPT_ENCRYPT) || (mode > SLUNKCRYPT_DECRYPT))
	{
		return SLUNKCRYPT_NULL;
	}
	if (!(state = (crypt_state_t*)malloc(sizeof(crypt_state_t))))
	{
		return SLUNKCRYPT_NULL;
	}
	if (initialize_state(state, nonce, passwd, passwd_len, mode, 0) == SLUNKCRYPT_SUCCESS)
	{
		return ((slunkcrypt_t)state);
	}
	else
	{
		slunkcrypt_bzero(state, sizeof(crypt_state_t));
		return SLUNKCRYPT_NULL;
	}
}

int slunkcrypt_reset(const slunkcrypt_t context, const uint64_t nonce, const uint8_t *const passwd, const size_t passwd_len, const int mode)
{
	crypt_state_t *const state = (crypt_state_t*)context;
	int result = SLUNKCRYPT_FAILURE;
	if ((!state) || (!passwd) || (passwd_len < SLUNKCRYPT_PWDLEN_MIN) || (passwd_len > SLUNKCRYPT_PWDLEN_MAX) || (mode < SLUNKCRYPT_ENCRYPT) || (mode > SLUNKCRYPT_DECRYPT))
	{
		return SLUNKCRYPT_FAILURE;
	}
	if ((result = initialize_state(state, nonce, passwd, passwd_len, mode, 1)) != SLUNKCRYPT_SUCCESS)
	{
		slunkcrypt_bzero(state, sizeof(crypt_state_t));
	}
	return result;
}

int slunkcrypt_process(const slunkcrypt_t context, const uint8_t *const input, uint8_t *const output, size_t length)
{
	size_t i;
	crypt_state_t *const state = (crypt_state_t*)context;
	if (!state)
	{
		return SLUNKCRYPT_FAILURE;
	}

	if (length > 0U)
	{
		memcpy(output, input, length * sizeof(uint8_t));
		for (i = 0; i < length; ++i)
		{
			abort(); //process_next_symbol(state, output + i);
			CHECK_ABORTED();
		}
	}

	thrdpl_await(state->thread_pool);
	return SLUNKCRYPT_SUCCESS;

aborted:
	thrdpl_await(state->thread_pool);
	slunkcrypt_bzero(state, sizeof(crypt_state_t));
	return SLUNKCRYPT_ABORTED;
}

int slunkcrypt_inplace(const slunkcrypt_t context, uint8_t *const buffer, size_t length)
{
	size_t i;
	crypt_state_t *const state = (crypt_state_t*)context;
	if (!state)
	{
		return SLUNKCRYPT_FAILURE;
	}

	if (length > 0U)
	{
		for (i = 0; i < length; ++i)
		{
			state->thread_data[state->thread_idx].data = buffer + i;
			//process_next_symbol(&state->thread_data[state->thread_idx]);
			thrdpl_submit(state->thread_pool, process_next_symbol, &state->thread_data[state->thread_idx]);
			if (++state->thread_idx >= THREAD_COUNT)
			{
				state->thread_idx = 0U;
			}
			CHECK_ABORTED();
		}
	}

	thrdpl_await(state->thread_pool);
	return SLUNKCRYPT_SUCCESS;

aborted:
	thrdpl_await(state->thread_pool);
	slunkcrypt_bzero(state, sizeof(crypt_state_t));
	return SLUNKCRYPT_ABORTED;
}

void slunkcrypt_free(const slunkcrypt_t context)
{
	crypt_state_t *const state = (crypt_state_t*)context;
	if (state)
	{
		thrdpl_destroy(state->thread_pool);
		slunkcrypt_bzero(state, sizeof(crypt_state_t));
		free(state);
	}
}