/*
 * LITLS -- SHA-256 (FIPS 180-4 / RFC 6234)
 * Self-written implementation.
 */

#include "litls.h"
#include <string.h>

static const uint32_t K[64] = {
    0x428a2f98, 0x71374491, 0xb5c0fbcf, 0xe9b5dba5, 0x3956c25b, 0x59f111f1, 0x923f82a4, 0xab1c5ed5,
    0xd807aa98, 0x12835b01, 0x243185be, 0x550c7dc3, 0x72be5d74, 0x80deb1fe, 0x9bdc06a7, 0xc19bf174,
    0xe49b69c1, 0xefbe4786, 0x0fc19dc6, 0x240ca1cc, 0x2de92c6f, 0x4a7484aa, 0x5cb0a9dc, 0x76f988da,
    0x983e5152, 0xa831c66d, 0xb00327c8, 0xbf597fc7, 0xc6e00bf3, 0xd5a79147, 0x06ca6351, 0x14292967,
    0x27b70a85, 0x2e1b2138, 0x4d2c6dfc, 0x53380d13, 0x650a7354, 0x766a0abb, 0x81c2c92e, 0x92722c85,
    0xa2bfe8a1, 0xa81a664b, 0xc24b8b70, 0xc76c51a3, 0xd192e819, 0xd6990624, 0xf40e3585, 0x106aa070,
    0x19a4c116, 0x1e376c08, 0x2748774c, 0x34b0bcb5, 0x391c0cb3, 0x4ed8aa4a, 0x5b9cca4f, 0x682e6ff3,
    0x748f82ee, 0x78a5636f, 0x84c87814, 0x8cc70208, 0x90befffa, 0xa4506ceb, 0xbef9a3f7, 0xc67178f2};

#define ROTR32(x, n) (((x) >> (n)) | ((x) << (32 - (n))))
#define CH(x, y, z)  (((x) & (y)) ^ (~(x) & (z)))
#define MAJ(x, y, z) (((x) & (y)) ^ ((x) & (z)) ^ ((y) & (z)))
#define EP0(x)       (ROTR32(x, 2) ^ ROTR32(x, 13) ^ ROTR32(x, 22))
#define EP1(x)       (ROTR32(x, 6) ^ ROTR32(x, 11) ^ ROTR32(x, 25))
#define SIG0(x)      (ROTR32(x, 7) ^ ROTR32(x, 18) ^ ((x) >> 3))
#define SIG1(x)      (ROTR32(x, 17) ^ ROTR32(x, 19) ^ ((x) >> 10))

static void sha256_transform(litls_sha256_ctx *ctx) {
    uint32_t w[64];
    const uint8_t *d = ctx->data;

    for (int i = 0; i < 16; i++) {
        w[i] = ((uint32_t)d[i * 4] << 24) | ((uint32_t)d[i * 4 + 1] << 16) | ((uint32_t)d[i * 4 + 2] << 8) |
               (uint32_t)d[i * 4 + 3];
    }
    for (int i = 16; i < 64; i++) {
        w[i] = SIG1(w[i - 2]) + w[i - 7] + SIG0(w[i - 15]) + w[i - 16];
    }

    uint32_t a = ctx->state[0], b = ctx->state[1];
    uint32_t c = ctx->state[2], d2 = ctx->state[3];
    uint32_t e = ctx->state[4], f = ctx->state[5];
    uint32_t g = ctx->state[6], h = ctx->state[7];

    for (int i = 0; i < 64; i++) {
        uint32_t t1 = h + EP1(e) + CH(e, f, g) + K[i] + w[i];
        uint32_t t2 = EP0(a) + MAJ(a, b, c);
        h           = g;
        g           = f;
        f           = e;
        e           = d2 + t1;
        d2          = c;
        c           = b;
        b           = a;
        a           = t1 + t2;
    }

    ctx->state[0] += a;
    ctx->state[1] += b;
    ctx->state[2] += c;
    ctx->state[3] += d2;
    ctx->state[4] += e;
    ctx->state[5] += f;
    ctx->state[6] += g;
    ctx->state[7] += h;
}

void litls_sha256_init(litls_sha256_ctx *ctx) {
    ctx->datalen  = 0;
    ctx->bitlen   = 0;
    ctx->state[0] = 0x6a09e667;
    ctx->state[1] = 0xbb67ae85;
    ctx->state[2] = 0x3c6ef372;
    ctx->state[3] = 0xa54ff53a;
    ctx->state[4] = 0x510e527f;
    ctx->state[5] = 0x9b05688c;
    ctx->state[6] = 0x1f83d9ab;
    ctx->state[7] = 0x5be0cd19;
}

void litls_sha256_update(litls_sha256_ctx *ctx, const uint8_t *data, size_t len) {
    for (size_t i = 0; i < len; i++) {
        ctx->data[ctx->datalen++] = data[i];
        if (ctx->datalen == 64) {
            sha256_transform(ctx);
            ctx->bitlen += 512;
            ctx->datalen = 0;
        }
    }
}

void litls_sha256_final(litls_sha256_ctx *ctx, uint8_t out[32]) {
    uint32_t i = ctx->datalen;

    ctx->data[i++] = 0x80;
    if (i > 56) {
        while (i < 64)
            ctx->data[i++] = 0;
        sha256_transform(ctx);
        i = 0;
    }
    while (i < 56)
        ctx->data[i++] = 0;

    ctx->bitlen += (uint64_t)ctx->datalen * 8;
    ctx->data[56] = (uint8_t)(ctx->bitlen >> 56);
    ctx->data[57] = (uint8_t)(ctx->bitlen >> 48);
    ctx->data[58] = (uint8_t)(ctx->bitlen >> 40);
    ctx->data[59] = (uint8_t)(ctx->bitlen >> 32);
    ctx->data[60] = (uint8_t)(ctx->bitlen >> 24);
    ctx->data[61] = (uint8_t)(ctx->bitlen >> 16);
    ctx->data[62] = (uint8_t)(ctx->bitlen >> 8);
    ctx->data[63] = (uint8_t)(ctx->bitlen);
    sha256_transform(ctx);

    for (int j = 0; j < 8; j++) {
        out[j * 4]     = (uint8_t)(ctx->state[j] >> 24);
        out[j * 4 + 1] = (uint8_t)(ctx->state[j] >> 16);
        out[j * 4 + 2] = (uint8_t)(ctx->state[j] >> 8);
        out[j * 4 + 3] = (uint8_t)(ctx->state[j]);
    }
}

void litls_sha256(const uint8_t *data, size_t len, uint8_t out[32]) {
    litls_sha256_ctx ctx;
    litls_sha256_init(&ctx);
    litls_sha256_update(&ctx, data, len);
    litls_sha256_final(&ctx, out);
}