/**
 * 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_sha1.cl)
#include M2S(INCLUDE_PATH/inc_cipher_aes.cl)
#endif

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

typedef struct pbkdf_sha1_tmp
{
  u32  ipad[5];
  u32  opad[5];

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

} pbkdf_sha1_tmp_t;

typedef struct terra
{
  u32 salt_buf[8];
  u32 ct[16]; // 16 * 4 = 64 bytes (we have extra 16 bytes in digest: 64 + 16 = 80)
  u32 iv[4];
} terra_t;

#define FIXED_SALT_SIZE 16


DECLSPEC void hmac_sha1_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];

  sha1_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] = 0x80000000;
  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] = (64 + 20) * 8;

  digest[0] = opad[0];
  digest[1] = opad[1];
  digest[2] = opad[2];
  digest[3] = opad[3];
  digest[4] = opad[4];

  sha1_transform_vector (w0, w1, w2, w3, digest);
}

KERNEL_FQ void m29600_init (KERN_ATTR_TMPS_ESALT (pbkdf_sha1_tmp_t,terra_t))
{
  const u64 gid = get_global_id (0);
  if (gid >= GID_CNT) return;

  // Salt needs to be copied into a big enough buffer to call hmac init
  u32 s[16] = { 0 };


  s[0] = hc_swap32_S(esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf[0]);
  s[1] = hc_swap32_S(esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf[1]);
  s[2] = hc_swap32_S(esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf[2]);
  s[3] = hc_swap32_S(esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf[3]);


  sha1_hmac_ctx_t sha1_hmac_ctx;
  sha1_hmac_init_global_swap (&sha1_hmac_ctx, pws[gid].i, pws[gid].pw_len);


  tmps[gid].ipad[0] = sha1_hmac_ctx.ipad.h[0];
  tmps[gid].ipad[1] = sha1_hmac_ctx.ipad.h[1];
  tmps[gid].ipad[2] = sha1_hmac_ctx.ipad.h[2];
  tmps[gid].ipad[3] = sha1_hmac_ctx.ipad.h[3];
  tmps[gid].ipad[4] = sha1_hmac_ctx.ipad.h[4];

  tmps[gid].opad[0] = sha1_hmac_ctx.opad.h[0];
  tmps[gid].opad[1] = sha1_hmac_ctx.opad.h[1];
  tmps[gid].opad[2] = sha1_hmac_ctx.opad.h[2];
  tmps[gid].opad[3] = sha1_hmac_ctx.opad.h[3];
  tmps[gid].opad[4] = sha1_hmac_ctx.opad.h[4];


  sha1_hmac_update (&sha1_hmac_ctx, s, FIXED_SALT_SIZE);


  // First iteration
  for (u32 i = 0, j = 1; i < 8; i += 5, j += 1)
  {
    sha1_hmac_ctx_t sha1_hmac_ctx2 = sha1_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;

    sha1_hmac_update_64 (&sha1_hmac_ctx2, w0, w1, w2, w3, 4);
    sha1_hmac_final (&sha1_hmac_ctx2);


    tmps[gid].dgst[i + 0] = sha1_hmac_ctx2.opad.h[0];
    tmps[gid].dgst[i + 1] = sha1_hmac_ctx2.opad.h[1];
    tmps[gid].dgst[i + 2] = sha1_hmac_ctx2.opad.h[2];
    tmps[gid].dgst[i + 3] = sha1_hmac_ctx2.opad.h[3];
    tmps[gid].dgst[i + 4] = sha1_hmac_ctx2.opad.h[4];

    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];
  }
}

KERNEL_FQ void m29600_loop (KERN_ATTR_TMPS_ESALT (pbkdf_sha1_tmp_t,terra_t))
{
  const u64 gid = get_global_id (0);

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

  u32x ipad[5];
  u32x opad[5];

  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);

  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);

  for (u32 i = 0; i < 8; i += 5)
  {
    u32x dgst[5];
    u32x out[5];

    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);

    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);

    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] = 0x80000000;
      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] = (64 + 20) * 8;

      hmac_sha1_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];
      }

    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, 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]);
  }
}

KERNEL_FQ void m29600_comp (KERN_ATTR_TMPS_ESALT (pbkdf_sha1_tmp_t,terra_t))
{
  /**
   * base
   */

  const u64 gid = get_global_id (0);
  const u64 lid = get_local_id(0);
  const u64 lsz = get_local_size(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;

  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 ks[60];
  u32 d[4];
  d[0] = esalt_bufs[DIGESTS_OFFSET_HOST].ct[0];
  d[1] = esalt_bufs[DIGESTS_OFFSET_HOST].ct[1];
  d[2] = esalt_bufs[DIGESTS_OFFSET_HOST].ct[2];
  d[3] = esalt_bufs[DIGESTS_OFFSET_HOST].ct[3];
  u32 CT[4] = { 0 };
  AES256_set_encrypt_key (ks, ukey, s_te0, s_te1, s_te2, s_te3);
  AES256_encrypt(ks, d, CT, s_te0, s_te1, s_te2, s_te3, s_te4);
  // check


  const u32 r0 = CT[0];
  const u32 r1 = CT[1];
  const u32 r2 = CT[2];
  const u32 r3 = CT[3];

  #define il_pos 0

  #ifdef KERNEL_STATIC
  #include COMPARE_M
  #endif
}