/**
 * Author......: See docs/credits.txt
 * License.....: MIT
 */

#define NEW_SIMD_CODE

#ifdef KERNEL_STATIC
#include M2S(INCLUDE_PATH/inc_vendor.h)
#include M2S(INCLUDE_PATH/inc_types.h)
#include M2S(INCLUDE_PATH/inc_platform.cl)
#include M2S(INCLUDE_PATH/inc_common.cl)
#include M2S(INCLUDE_PATH/inc_simd.cl)
#include M2S(INCLUDE_PATH/inc_hash_sha256.cl)
#include M2S(INCLUDE_PATH/inc_cipher_aes.cl)
#include M2S(INCLUDE_PATH/inc_cipher_aes-gcm.cl)
#endif

#define COMPARE_S M2S(INCLUDE_PATH/inc_comp_single.cl)
#define COMPARE_M M2S(INCLUDE_PATH/inc_comp_multi.cl)

typedef struct pbkdf2_sha256_tmp
{
  u32  ipad[8];
  u32  opad[8];

  u32  dgst[32];
  u32  out[32];

} pbkdf2_sha256_tmp_t;

typedef struct pbkdf2_sha256_aes_gcm
{
  u32 salt_buf[64];
  u32 iv_buf[4];
  u32 iv_len;
  u32 ct_buf[784]; // TODO this can be smaller and would speedup the attack, only 64 bytes of ciphertext are allowed
  u32 ct_len;

} pbkdf2_sha256_aes_gcm_t;

DECLSPEC void hmac_sha256_run_V (PRIVATE_AS u32x *w0, PRIVATE_AS u32x *w1, PRIVATE_AS u32x *w2, PRIVATE_AS u32x *w3, PRIVATE_AS u32x *ipad, PRIVATE_AS u32x *opad, PRIVATE_AS u32x *digest)
{
  digest[0] = ipad[0];
  digest[1] = ipad[1];
  digest[2] = ipad[2];
  digest[3] = ipad[3];
  digest[4] = ipad[4];
  digest[5] = ipad[5];
  digest[6] = ipad[6];
  digest[7] = ipad[7];

  sha256_transform_vector (w0, w1, w2, w3, digest);

  w0[0] = digest[0];
  w0[1] = digest[1];
  w0[2] = digest[2];
  w0[3] = digest[3];
  w1[0] = digest[4];
  w1[1] = digest[5];
  w1[2] = digest[6];
  w1[3] = digest[7];
  w2[0] = 0x80000000;
  w2[1] = 0;
  w2[2] = 0;
  w2[3] = 0;
  w3[0] = 0;
  w3[1] = 0;
  w3[2] = 0;
  w3[3] = (64 + 32) * 8;

  digest[0] = opad[0];
  digest[1] = opad[1];
  digest[2] = opad[2];
  digest[3] = opad[3];
  digest[4] = opad[4];
  digest[5] = opad[5];
  digest[6] = opad[6];
  digest[7] = opad[7];

  sha256_transform_vector (w0, w1, w2, w3, digest);
}

KERNEL_FQ void m26610_init (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, pbkdf2_sha256_aes_gcm_t))
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);

  if (gid >= GID_CNT) return;

  sha256_hmac_ctx_t sha256_hmac_ctx;

  sha256_hmac_init_global_swap (&sha256_hmac_ctx, pws[gid].i, pws[gid].pw_len);

  tmps[gid].ipad[0] = sha256_hmac_ctx.ipad.h[0];
  tmps[gid].ipad[1] = sha256_hmac_ctx.ipad.h[1];
  tmps[gid].ipad[2] = sha256_hmac_ctx.ipad.h[2];
  tmps[gid].ipad[3] = sha256_hmac_ctx.ipad.h[3];
  tmps[gid].ipad[4] = sha256_hmac_ctx.ipad.h[4];
  tmps[gid].ipad[5] = sha256_hmac_ctx.ipad.h[5];
  tmps[gid].ipad[6] = sha256_hmac_ctx.ipad.h[6];
  tmps[gid].ipad[7] = sha256_hmac_ctx.ipad.h[7];

  tmps[gid].opad[0] = sha256_hmac_ctx.opad.h[0];
  tmps[gid].opad[1] = sha256_hmac_ctx.opad.h[1];
  tmps[gid].opad[2] = sha256_hmac_ctx.opad.h[2];
  tmps[gid].opad[3] = sha256_hmac_ctx.opad.h[3];
  tmps[gid].opad[4] = sha256_hmac_ctx.opad.h[4];
  tmps[gid].opad[5] = sha256_hmac_ctx.opad.h[5];
  tmps[gid].opad[6] = sha256_hmac_ctx.opad.h[6];
  tmps[gid].opad[7] = sha256_hmac_ctx.opad.h[7];

  sha256_hmac_update_global_swap (&sha256_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf, salt_bufs[SALT_POS_HOST].salt_len);

  for (u32 i = 0, j = 1; i < 8; i += 8, j += 1)
  {
    sha256_hmac_ctx_t sha256_hmac_ctx2 = sha256_hmac_ctx;

    u32 w0[4];
    u32 w1[4];
    u32 w2[4];
    u32 w3[4];

    w0[0] = j;
    w0[1] = 0;
    w0[2] = 0;
    w0[3] = 0;
    w1[0] = 0;
    w1[1] = 0;
    w1[2] = 0;
    w1[3] = 0;
    w2[0] = 0;
    w2[1] = 0;
    w2[2] = 0;
    w2[3] = 0;
    w3[0] = 0;
    w3[1] = 0;
    w3[2] = 0;
    w3[3] = 0;

    sha256_hmac_update_64 (&sha256_hmac_ctx2, w0, w1, w2, w3, 4);

    sha256_hmac_final (&sha256_hmac_ctx2);

    tmps[gid].dgst[i + 0] = sha256_hmac_ctx2.opad.h[0];
    tmps[gid].dgst[i + 1] = sha256_hmac_ctx2.opad.h[1];
    tmps[gid].dgst[i + 2] = sha256_hmac_ctx2.opad.h[2];
    tmps[gid].dgst[i + 3] = sha256_hmac_ctx2.opad.h[3];
    tmps[gid].dgst[i + 4] = sha256_hmac_ctx2.opad.h[4];
    tmps[gid].dgst[i + 5] = sha256_hmac_ctx2.opad.h[5];
    tmps[gid].dgst[i + 6] = sha256_hmac_ctx2.opad.h[6];
    tmps[gid].dgst[i + 7] = sha256_hmac_ctx2.opad.h[7];

    tmps[gid].out[i + 0] = tmps[gid].dgst[i + 0];
    tmps[gid].out[i + 1] = tmps[gid].dgst[i + 1];
    tmps[gid].out[i + 2] = tmps[gid].dgst[i + 2];
    tmps[gid].out[i + 3] = tmps[gid].dgst[i + 3];
    tmps[gid].out[i + 4] = tmps[gid].dgst[i + 4];
    tmps[gid].out[i + 5] = tmps[gid].dgst[i + 5];
    tmps[gid].out[i + 6] = tmps[gid].dgst[i + 6];
    tmps[gid].out[i + 7] = tmps[gid].dgst[i + 7];
  }
}

KERNEL_FQ void m26610_loop (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, pbkdf2_sha256_aes_gcm_t))
{
  const u64 gid = get_global_id (0);

  if ((gid * VECT_SIZE) >= GID_CNT) return;

  u32x ipad[8];
  u32x opad[8];

  ipad[0] = packv (tmps, ipad, gid, 0);
  ipad[1] = packv (tmps, ipad, gid, 1);
  ipad[2] = packv (tmps, ipad, gid, 2);
  ipad[3] = packv (tmps, ipad, gid, 3);
  ipad[4] = packv (tmps, ipad, gid, 4);
  ipad[5] = packv (tmps, ipad, gid, 5);
  ipad[6] = packv (tmps, ipad, gid, 6);
  ipad[7] = packv (tmps, ipad, gid, 7);

  opad[0] = packv (tmps, opad, gid, 0);
  opad[1] = packv (tmps, opad, gid, 1);
  opad[2] = packv (tmps, opad, gid, 2);
  opad[3] = packv (tmps, opad, gid, 3);
  opad[4] = packv (tmps, opad, gid, 4);
  opad[5] = packv (tmps, opad, gid, 5);
  opad[6] = packv (tmps, opad, gid, 6);
  opad[7] = packv (tmps, opad, gid, 7);

  for (u32 i = 0; i < 8; i += 8)
  {
    u32x dgst[8];
    u32x out[8];

    dgst[0] = packv (tmps, dgst, gid, i + 0);
    dgst[1] = packv (tmps, dgst, gid, i + 1);
    dgst[2] = packv (tmps, dgst, gid, i + 2);
    dgst[3] = packv (tmps, dgst, gid, i + 3);
    dgst[4] = packv (tmps, dgst, gid, i + 4);
    dgst[5] = packv (tmps, dgst, gid, i + 5);
    dgst[6] = packv (tmps, dgst, gid, i + 6);
    dgst[7] = packv (tmps, dgst, gid, i + 7);

    out[0] = packv (tmps, out, gid, i + 0);
    out[1] = packv (tmps, out, gid, i + 1);
    out[2] = packv (tmps, out, gid, i + 2);
    out[3] = packv (tmps, out, gid, i + 3);
    out[4] = packv (tmps, out, gid, i + 4);
    out[5] = packv (tmps, out, gid, i + 5);
    out[6] = packv (tmps, out, gid, i + 6);
    out[7] = packv (tmps, out, gid, i + 7);

    for (u32 j = 0; j < LOOP_CNT; j++)
    {
      u32x w0[4];
      u32x w1[4];
      u32x w2[4];
      u32x w3[4];

      w0[0] = dgst[0];
      w0[1] = dgst[1];
      w0[2] = dgst[2];
      w0[3] = dgst[3];
      w1[0] = dgst[4];
      w1[1] = dgst[5];
      w1[2] = dgst[6];
      w1[3] = dgst[7];
      w2[0] = 0x80000000;
      w2[1] = 0;
      w2[2] = 0;
      w2[3] = 0;
      w3[0] = 0;
      w3[1] = 0;
      w3[2] = 0;
      w3[3] = (64 + 32) * 8;

      hmac_sha256_run_V (w0, w1, w2, w3, ipad, opad, dgst);

      out[0] ^= dgst[0];
      out[1] ^= dgst[1];
      out[2] ^= dgst[2];
      out[3] ^= dgst[3];
      out[4] ^= dgst[4];
      out[5] ^= dgst[5];
      out[6] ^= dgst[6];
      out[7] ^= dgst[7];
    }

    unpackv (tmps, dgst, gid, i + 0, dgst[0]);
    unpackv (tmps, dgst, gid, i + 1, dgst[1]);
    unpackv (tmps, dgst, gid, i + 2, dgst[2]);
    unpackv (tmps, dgst, gid, i + 3, dgst[3]);
    unpackv (tmps, dgst, gid, i + 4, dgst[4]);
    unpackv (tmps, dgst, gid, i + 5, dgst[5]);
    unpackv (tmps, dgst, gid, i + 6, dgst[6]);
    unpackv (tmps, dgst, gid, i + 7, dgst[7]);

    unpackv (tmps, out, gid, i + 0, out[0]);
    unpackv (tmps, out, gid, i + 1, out[1]);
    unpackv (tmps, out, gid, i + 2, out[2]);
    unpackv (tmps, out, gid, i + 3, out[3]);
    unpackv (tmps, out, gid, i + 4, out[4]);
    unpackv (tmps, out, gid, i + 5, out[5]);
    unpackv (tmps, out, gid, i + 6, out[6]);
    unpackv (tmps, out, gid, i + 7, out[7]);
  }
}

KERNEL_FQ void m26610_comp (KERN_ATTR_TMPS_ESALT (pbkdf2_sha256_tmp_t, pbkdf2_sha256_aes_gcm_t))
{
  const u64 gid = get_global_id (0);
  const u64 lid = get_local_id (0);
  const u64 lsz = get_local_size (0);
  #define il_pos 0


  /**
   * aes shared
   */

  #ifdef REAL_SHM

  LOCAL_VK u32 s_te0[256];
  LOCAL_VK u32 s_te1[256];
  LOCAL_VK u32 s_te2[256];
  LOCAL_VK u32 s_te3[256];
  LOCAL_VK u32 s_te4[256];

  for (u32 i = lid; i < 256; i += lsz)
  {
    s_te0[i] = te0[i];
    s_te1[i] = te1[i];
    s_te2[i] = te2[i];
    s_te3[i] = te3[i];
    s_te4[i] = te4[i];
  }

  SYNC_THREADS ();

  #else

  CONSTANT_AS u32a *s_te0 = te0;
  CONSTANT_AS u32a *s_te1 = te1;
  CONSTANT_AS u32a *s_te2 = te2;
  CONSTANT_AS u32a *s_te3 = te3;
  CONSTANT_AS u32a *s_te4 = te4;

  #endif

  if (gid >= GID_CNT) return;

  // keys

  u32 ukey[8];

  ukey[0] = tmps[gid].out[0];
  ukey[1] = tmps[gid].out[1];
  ukey[2] = tmps[gid].out[2];
  ukey[3] = tmps[gid].out[3];
  ukey[4] = tmps[gid].out[4];
  ukey[5] = tmps[gid].out[5];
  ukey[6] = tmps[gid].out[6];
  ukey[7] = tmps[gid].out[7];

  u32 key_len = 32 * 8;

  u32 key[60] = { 0 };
  u32 subKey[4] = { 0 };

  AES_GCM_Init (ukey, key_len, key, subKey, s_te0, s_te1, s_te2, s_te3, s_te4);

  // iv
  const u32 iv[4] = {
    esalt_bufs[DIGESTS_OFFSET_HOST].iv_buf[0],
    esalt_bufs[DIGESTS_OFFSET_HOST].iv_buf[1],
    esalt_bufs[DIGESTS_OFFSET_HOST].iv_buf[2],
    esalt_bufs[DIGESTS_OFFSET_HOST].iv_buf[3]
  };

  const u32 iv_len = esalt_bufs[DIGESTS_OFFSET_HOST].iv_len;

  u32 J0[4] = { 0 };

  AES_GCM_Prepare_J0 (iv, iv_len, subKey, J0);

  //ct
  u32 ct[4] = { 
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[0], 
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[1],
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[2],
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[3] 
  };
  
  u32 pt[4] = { 0 };

  // we try to decrypt the ciphertext
  // TODO this can be moved to a separate decryption function in inc_cipher_aes-gcm.cl
  AES_GCM_inc32(J0); // the first ctr is used to compute the tag, only the second is used for decryption: https://en.wikipedia.org/wiki/Galois/Counter_Mode#/media/File:GCM-Galois_Counter_Mode_with_IV.svg
  AES_GCM_GCTR (key, J0, ct, 16, pt, s_te0, s_te1, s_te2, s_te3, s_te4); // decrypt the ciphertext

  // if ((gid == 0) && (lid == 0)) printf ("pt[0]=%08x\n", pt[0]); // should be 5b7b2274 or [{"type"

  // cast plaintext buffer to byte such that we can do a byte per byte comparison
  PRIVATE_AS const u32 *u32OutBufPtr = (PRIVATE_AS u32 *) pt;
  PRIVATE_AS const u8  *u8OutBufPtr  = (PRIVATE_AS u8  *) u32OutBufPtr;

  // the best comparison I can think of is checking each byte
  //  whether it's ASCII, if so we're good,
  //  if not, decryption was not successful
  bool correct = true;
  
  for(int i=0;i<16;i++)
  { 
        if(u8OutBufPtr[i] >=20 && u8OutBufPtr[i] <= 0x7e) {
          //if ((gid == 0) && (lid == 0)) printf("correct ASCII byte[%d]=0x%02x\n", i, u8OutBufPtr[i]);
        } 
        else {
          //if ((gid == 0) && (lid == 0)) printf("NOT correct!   byte[%d]=0x%02x\n", i, u8OutBufPtr[i]);
          correct = false;
          break;
        }
  }

  const u32 digest[4] =
  {
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[0],
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[1],
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[2],
    esalt_bufs[DIGESTS_OFFSET_HOST].ct_buf[3],
  };
  
  //if ((gid == 0) && (lid == 0)) printf ("ct[0]=%08x\n", ct[0]);
  //if ((gid == 0) && (lid == 0)) printf ("ct[1]=%08x\n", ct[1]);
  //if ((gid == 0) && (lid == 0)) printf ("ct[2]=%08x\n", ct[2]);
  //if ((gid == 0) && (lid == 0)) printf ("ct[3]=%08x\n", ct[3]);

  if (correct)
  {
    int digest_pos = find_hash (digest, DIGESTS_CNT, &digests_buf[DIGESTS_OFFSET_HOST]);

    if (digest_pos != -1)
    {
      const u32 final_hash_pos = DIGESTS_OFFSET_HOST + digest_pos;

      if (hc_atomic_inc (&hashes_shown[final_hash_pos]) == 0)
      {
        mark_hash (plains_buf, d_return_buf, SALT_POS_HOST, DIGESTS_CNT, digest_pos, final_hash_pos, gid, il_pos, 0, 0);
      }
    }
  }
}