From 5c6b3fa7ab5c8d2baf637514286ad9388e700f74 Mon Sep 17 00:00:00 2001
From: jsteube <jens.steube@gmail.com>
Date: Mon, 7 Aug 2017 15:22:18 +0200
Subject: [PATCH] Add pure kernels for Kerberos 5 TGS-REP etype 23

---
 OpenCL/m13100_a0.cl | 516 +++++++++++++++++++++++++++++++++++++++++++
 OpenCL/m13100_a1.cl | 492 +++++++++++++++++++++++++++++++++++++++++
 OpenCL/m13100_a3.cl | 526 ++++++++++++++++++++++++++++++++++++++++++++
 3 files changed, 1534 insertions(+)
 create mode 100644 OpenCL/m13100_a0.cl
 create mode 100644 OpenCL/m13100_a1.cl
 create mode 100644 OpenCL/m13100_a3.cl

diff --git a/OpenCL/m13100_a0.cl b/OpenCL/m13100_a0.cl
new file mode 100644
index 000000000..d01619666
--- /dev/null
+++ b/OpenCL/m13100_a0.cl
@@ -0,0 +1,516 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+//shared mem too small
+//#define NEW_SIMD_CODE
+
+#include "inc_vendor.cl"
+#include "inc_hash_constants.h"
+#include "inc_hash_functions.cl"
+#include "inc_types.cl"
+#include "inc_common.cl"
+#include "inc_rp.h"
+#include "inc_rp.cl"
+#include "inc_hash_md4.cl"
+#include "inc_hash_md5.cl"
+
+typedef struct
+{
+  u8 S[256];
+
+  u32 wtf_its_faster;
+
+} RC4_KEY;
+
+void swap (__local RC4_KEY *rc4_key, const u8 i, const u8 j)
+{
+  u8 tmp;
+
+  tmp           = rc4_key->S[i];
+  rc4_key->S[i] = rc4_key->S[j];
+  rc4_key->S[j] = tmp;
+}
+
+void rc4_init_16 (__local RC4_KEY *rc4_key, const u32 data[4])
+{
+  u32 v = 0x03020100;
+  u32 a = 0x04040404;
+
+  __local u32 *ptr = (__local u32 *) rc4_key->S;
+
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 i = 0; i < 64; i++)
+  {
+    *ptr++ = v; v += a;
+  }
+
+  u32 j = 0;
+
+  for (u32 i = 0; i < 16; i++)
+  {
+    u32 idx = i * 16;
+
+    u32 v;
+
+    v = data[0];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[1];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[2];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[3];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+  }
+}
+
+u8 rc4_next_16 (__local RC4_KEY *rc4_key, u8 i, u8 j, __global const u32 in[4], u32 out[4])
+{
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 k = 0; k < 4; k++)
+  {
+    u32 xor4 = 0;
+
+    u8 idx;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  0;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  8;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 16;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 24;
+
+    out[k] = in[k] ^ xor4;
+  }
+
+  return j;
+}
+
+int decrypt_and_check (__local RC4_KEY *rc4_key, u32 data[4], __global const u32 *edata2, const u32 edata2_len, const u32 K2[4], const u32 checksum[4])
+{
+  rc4_init_16 (rc4_key, data);
+
+  u32 out0[4];
+  u32 out1[4];
+
+  u8 i = 0;
+  u8 j = 0;
+
+  /*
+    8 first bytes are nonce, then ASN1 structs (DER encoding: type-length-data)
+
+    if length >= 128 bytes:
+        length is on 2 bytes and type is \x63\x82 (encode_krb5_enc_tkt_part) and data is an ASN1 sequence \x30\x82
+    else:
+        length is on 1 byte and type is \x63\x81 and data is an ASN1 sequence \x30\x81
+
+    next headers follow the same ASN1 "type-length-data" scheme
+  */
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 0, out0); i += 16;
+
+  if (((out0[2] & 0xff00ffff) != 0x30008163) && ((out0[2] & 0x0000ffff) != 0x00008263)) return 0;
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 4, out1); i += 16;
+
+  if (((out1[0] & 0x00ffffff) != 0x00000503) && (out1[0] != 0x050307A0)) return 0;
+
+  rc4_init_16 (rc4_key, data);
+
+  i = 0;
+  j = 0;
+
+  // init hmac
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K2[0];
+  w0[1] = K2[1];
+  w0[2] = K2[2];
+  w0[3] = K2[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  int edata2_left;
+
+  for (edata2_left = edata2_len; edata2_left >= 64; edata2_left -= 64)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 64);
+  }
+
+  w0[0] = 0;
+  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;
+
+  if (edata2_left < 16)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w0, edata2_left & 0xf);
+  }
+  else if (edata2_left < 32)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w1, edata2_left & 0xf);
+  }
+  else if (edata2_left < 48)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w2, edata2_left & 0xf);
+  }
+  else
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w3, edata2_left & 0xf);
+  }
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, edata2_left);
+
+  md5_hmac_final (&ctx);
+
+  if (checksum[0] != ctx.opad.h[0]) return 0;
+  if (checksum[1] != ctx.opad.h[1]) return 0;
+  if (checksum[2] != ctx.opad.h[2]) return 0;
+  if (checksum[3] != ctx.opad.h[3]) return 0;
+
+  return 1;
+}
+
+void kerb_prepare (const u32 K[4], const u32 checksum[4], u32 digest[4], u32 K2[4])
+{
+  // K1=MD5_HMAC(K,1); with 1 encoded as little indian on 4 bytes (01000000 in hexa);
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K[0];
+  w0[1] = K[1];
+  w0[2] = K[2];
+  w0[3] = K[3];
+  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;
+
+  md5_hmac_ctx_t ctx1;
+
+  md5_hmac_init_64 (&ctx1, w0, w1, w2, w3);
+
+  w0[0] = 2;
+  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;
+
+  md5_hmac_update_64 (&ctx1, w0, w1, w2, w3, 4);
+
+  md5_hmac_final (&ctx1);
+
+  w0[0] = ctx1.opad.h[0];
+  w0[1] = ctx1.opad.h[1];
+  w0[2] = ctx1.opad.h[2];
+  w0[3] = ctx1.opad.h[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  w0[0] = checksum[0];
+  w0[1] = checksum[1];
+  w0[2] = checksum[2];
+  w0[3] = checksum[3];
+  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;
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 16);
+
+  md5_hmac_final (&ctx);
+
+  digest[0] = ctx.opad.h[0];
+  digest[1] = ctx.opad.h[1];
+  digest[2] = ctx.opad.h[2];
+  digest[3] = ctx.opad.h[3];
+
+  K2[0] = ctx1.opad.h[0];
+  K2[1] = ctx1.opad.h[1];
+  K2[2] = ctx1.opad.h[2];
+  K2[3] = ctx1.opad.h[3];
+}
+
+__kernel void m13100_mxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32 w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    // todo: add rules engine
+
+    md4_ctx_t ctx;
+
+    md4_init (&ctx);
+
+    md4_update_utf16le (&ctx, w, pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}
+
+__kernel void m13100_sxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32 w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    // todo: add rules engine
+
+    md4_ctx_t ctx;
+
+    md4_init (&ctx);
+
+    md4_update_utf16le (&ctx, w, pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}
diff --git a/OpenCL/m13100_a1.cl b/OpenCL/m13100_a1.cl
new file mode 100644
index 000000000..38b1ed643
--- /dev/null
+++ b/OpenCL/m13100_a1.cl
@@ -0,0 +1,492 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+//shared mem too small
+//#define NEW_SIMD_CODE
+
+#include "inc_vendor.cl"
+#include "inc_hash_constants.h"
+#include "inc_hash_functions.cl"
+#include "inc_types.cl"
+#include "inc_common.cl"
+#include "inc_hash_md4.cl"
+#include "inc_hash_md5.cl"
+
+typedef struct
+{
+  u8 S[256];
+
+  u32 wtf_its_faster;
+
+} RC4_KEY;
+
+void swap (__local RC4_KEY *rc4_key, const u8 i, const u8 j)
+{
+  u8 tmp;
+
+  tmp           = rc4_key->S[i];
+  rc4_key->S[i] = rc4_key->S[j];
+  rc4_key->S[j] = tmp;
+}
+
+void rc4_init_16 (__local RC4_KEY *rc4_key, const u32 data[4])
+{
+  u32 v = 0x03020100;
+  u32 a = 0x04040404;
+
+  __local u32 *ptr = (__local u32 *) rc4_key->S;
+
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 i = 0; i < 64; i++)
+  {
+    *ptr++ = v; v += a;
+  }
+
+  u32 j = 0;
+
+  for (u32 i = 0; i < 16; i++)
+  {
+    u32 idx = i * 16;
+
+    u32 v;
+
+    v = data[0];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[1];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[2];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[3];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+  }
+}
+
+u8 rc4_next_16 (__local RC4_KEY *rc4_key, u8 i, u8 j, __global const u32 in[4], u32 out[4])
+{
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 k = 0; k < 4; k++)
+  {
+    u32 xor4 = 0;
+
+    u8 idx;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  0;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  8;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 16;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 24;
+
+    out[k] = in[k] ^ xor4;
+  }
+
+  return j;
+}
+
+int decrypt_and_check (__local RC4_KEY *rc4_key, u32 data[4], __global const u32 *edata2, const u32 edata2_len, const u32 K2[4], const u32 checksum[4])
+{
+  rc4_init_16 (rc4_key, data);
+
+  u32 out0[4];
+  u32 out1[4];
+
+  u8 i = 0;
+  u8 j = 0;
+
+  /*
+    8 first bytes are nonce, then ASN1 structs (DER encoding: type-length-data)
+
+    if length >= 128 bytes:
+        length is on 2 bytes and type is \x63\x82 (encode_krb5_enc_tkt_part) and data is an ASN1 sequence \x30\x82
+    else:
+        length is on 1 byte and type is \x63\x81 and data is an ASN1 sequence \x30\x81
+
+    next headers follow the same ASN1 "type-length-data" scheme
+  */
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 0, out0); i += 16;
+
+  if (((out0[2] & 0xff00ffff) != 0x30008163) && ((out0[2] & 0x0000ffff) != 0x00008263)) return 0;
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 4, out1); i += 16;
+
+  if (((out1[0] & 0x00ffffff) != 0x00000503) && (out1[0] != 0x050307A0)) return 0;
+
+  rc4_init_16 (rc4_key, data);
+
+  i = 0;
+  j = 0;
+
+  // init hmac
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K2[0];
+  w0[1] = K2[1];
+  w0[2] = K2[2];
+  w0[3] = K2[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  int edata2_left;
+
+  for (edata2_left = edata2_len; edata2_left >= 64; edata2_left -= 64)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 64);
+  }
+
+  w0[0] = 0;
+  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;
+
+  if (edata2_left < 16)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w0, edata2_left & 0xf);
+  }
+  else if (edata2_left < 32)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w1, edata2_left & 0xf);
+  }
+  else if (edata2_left < 48)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w2, edata2_left & 0xf);
+  }
+  else
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w3, edata2_left & 0xf);
+  }
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, edata2_left);
+
+  md5_hmac_final (&ctx);
+
+  if (checksum[0] != ctx.opad.h[0]) return 0;
+  if (checksum[1] != ctx.opad.h[1]) return 0;
+  if (checksum[2] != ctx.opad.h[2]) return 0;
+  if (checksum[3] != ctx.opad.h[3]) return 0;
+
+  return 1;
+}
+
+void kerb_prepare (const u32 K[4], const u32 checksum[4], u32 digest[4], u32 K2[4])
+{
+  // K1=MD5_HMAC(K,1); with 1 encoded as little indian on 4 bytes (01000000 in hexa);
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K[0];
+  w0[1] = K[1];
+  w0[2] = K[2];
+  w0[3] = K[3];
+  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;
+
+  md5_hmac_ctx_t ctx1;
+
+  md5_hmac_init_64 (&ctx1, w0, w1, w2, w3);
+
+  w0[0] = 2;
+  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;
+
+  md5_hmac_update_64 (&ctx1, w0, w1, w2, w3, 4);
+
+  md5_hmac_final (&ctx1);
+
+  w0[0] = ctx1.opad.h[0];
+  w0[1] = ctx1.opad.h[1];
+  w0[2] = ctx1.opad.h[2];
+  w0[3] = ctx1.opad.h[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  w0[0] = checksum[0];
+  w0[1] = checksum[1];
+  w0[2] = checksum[2];
+  w0[3] = checksum[3];
+  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;
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 16);
+
+  md5_hmac_final (&ctx);
+
+  digest[0] = ctx.opad.h[0];
+  digest[1] = ctx.opad.h[1];
+  digest[2] = ctx.opad.h[2];
+  digest[3] = ctx.opad.h[3];
+
+  K2[0] = ctx1.opad.h[0];
+  K2[1] = ctx1.opad.h[1];
+  K2[2] = ctx1.opad.h[2];
+  K2[3] = ctx1.opad.h[3];
+}
+
+__kernel void m13100_mxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  md4_ctx_t ctx0;
+
+  md4_init (&ctx0);
+
+  md4_update_global_utf16le (&ctx0, pws[gid].i, pws[gid].pw_len);
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    md4_ctx_t ctx = ctx0;
+
+    md4_update_global_utf16le (&ctx, combs_buf[il_pos].i, combs_buf[il_pos].pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}
+
+__kernel void m13100_sxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __global const bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  md4_ctx_t ctx0;
+
+  md4_init (&ctx0);
+
+  md4_update_global_utf16le (&ctx0, pws[gid].i, pws[gid].pw_len);
+
+  /**
+   * loop
+   */
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos++)
+  {
+    md4_ctx_t ctx = ctx0;
+
+    md4_update_global_utf16le (&ctx, combs_buf[il_pos].i, combs_buf[il_pos].pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}
diff --git a/OpenCL/m13100_a3.cl b/OpenCL/m13100_a3.cl
new file mode 100644
index 000000000..7b1ac5a05
--- /dev/null
+++ b/OpenCL/m13100_a3.cl
@@ -0,0 +1,526 @@
+/**
+ * Author......: See docs/credits.txt
+ * License.....: MIT
+ */
+
+//shared mem too small
+//#define NEW_SIMD_CODE
+
+#include "inc_vendor.cl"
+#include "inc_hash_constants.h"
+#include "inc_hash_functions.cl"
+#include "inc_types.cl"
+#include "inc_common.cl"
+#include "inc_hash_md4.cl"
+#include "inc_hash_md5.cl"
+
+typedef struct
+{
+  u8 S[256];
+
+  u32 wtf_its_faster;
+
+} RC4_KEY;
+
+void swap (__local RC4_KEY *rc4_key, const u8 i, const u8 j)
+{
+  u8 tmp;
+
+  tmp           = rc4_key->S[i];
+  rc4_key->S[i] = rc4_key->S[j];
+  rc4_key->S[j] = tmp;
+}
+
+void rc4_init_16 (__local RC4_KEY *rc4_key, const u32 data[4])
+{
+  u32 v = 0x03020100;
+  u32 a = 0x04040404;
+
+  __local u32 *ptr = (__local u32 *) rc4_key->S;
+
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 i = 0; i < 64; i++)
+  {
+    *ptr++ = v; v += a;
+  }
+
+  u32 j = 0;
+
+  for (u32 i = 0; i < 16; i++)
+  {
+    u32 idx = i * 16;
+
+    u32 v;
+
+    v = data[0];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[1];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[2];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+
+    v = data[3];
+
+    j += rc4_key->S[idx] + (v >>  0); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >>  8); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 16); swap (rc4_key, idx, j); idx++;
+    j += rc4_key->S[idx] + (v >> 24); swap (rc4_key, idx, j); idx++;
+  }
+}
+
+u8 rc4_next_16 (__local RC4_KEY *rc4_key, u8 i, u8 j, __global const u32 in[4], u32 out[4])
+{
+  #ifdef _unroll
+  #pragma unroll
+  #endif
+  for (u32 k = 0; k < 4; k++)
+  {
+    u32 xor4 = 0;
+
+    u8 idx;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  0;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] <<  8;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 16;
+
+    i += 1;
+    j += rc4_key->S[i];
+
+    swap (rc4_key, i, j);
+
+    idx = rc4_key->S[i] + rc4_key->S[j];
+
+    xor4 |= rc4_key->S[idx] << 24;
+
+    out[k] = in[k] ^ xor4;
+  }
+
+  return j;
+}
+
+int decrypt_and_check (__local RC4_KEY *rc4_key, u32 data[4], __global const u32 *edata2, const u32 edata2_len, const u32 K2[4], const u32 checksum[4])
+{
+  rc4_init_16 (rc4_key, data);
+
+  u32 out0[4];
+  u32 out1[4];
+
+  u8 i = 0;
+  u8 j = 0;
+
+  /*
+    8 first bytes are nonce, then ASN1 structs (DER encoding: type-length-data)
+
+    if length >= 128 bytes:
+        length is on 2 bytes and type is \x63\x82 (encode_krb5_enc_tkt_part) and data is an ASN1 sequence \x30\x82
+    else:
+        length is on 1 byte and type is \x63\x81 and data is an ASN1 sequence \x30\x81
+
+    next headers follow the same ASN1 "type-length-data" scheme
+  */
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 0, out0); i += 16;
+
+  if (((out0[2] & 0xff00ffff) != 0x30008163) && ((out0[2] & 0x0000ffff) != 0x00008263)) return 0;
+
+  j = rc4_next_16 (rc4_key, i, j, edata2 + 4, out1); i += 16;
+
+  if (((out1[0] & 0x00ffffff) != 0x00000503) && (out1[0] != 0x050307A0)) return 0;
+
+  rc4_init_16 (rc4_key, data);
+
+  i = 0;
+  j = 0;
+
+  // init hmac
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K2[0];
+  w0[1] = K2[1];
+  w0[2] = K2[2];
+  w0[3] = K2[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  int edata2_left;
+
+  for (edata2_left = edata2_len; edata2_left >= 64; edata2_left -= 64)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 64);
+  }
+
+  w0[0] = 0;
+  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;
+
+  if (edata2_left < 16)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w0, edata2_left & 0xf);
+  }
+  else if (edata2_left < 32)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w1, edata2_left & 0xf);
+  }
+  else if (edata2_left < 48)
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w2, edata2_left & 0xf);
+  }
+  else
+  {
+    j = rc4_next_16 (rc4_key, i, j, edata2, w0); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w1); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w2); i += 16; edata2 += 4;
+    j = rc4_next_16 (rc4_key, i, j, edata2, w3); i += 16; edata2 += 4;
+
+    truncate_block_4x4_le (w3, edata2_left & 0xf);
+  }
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, edata2_left);
+
+  md5_hmac_final (&ctx);
+
+  if (checksum[0] != ctx.opad.h[0]) return 0;
+  if (checksum[1] != ctx.opad.h[1]) return 0;
+  if (checksum[2] != ctx.opad.h[2]) return 0;
+  if (checksum[3] != ctx.opad.h[3]) return 0;
+
+  return 1;
+}
+
+void kerb_prepare (const u32 K[4], const u32 checksum[4], u32 digest[4], u32 K2[4])
+{
+  // K1=MD5_HMAC(K,1); with 1 encoded as little indian on 4 bytes (01000000 in hexa);
+
+  u32 w0[4];
+  u32 w1[4];
+  u32 w2[4];
+  u32 w3[4];
+
+  w0[0] = K[0];
+  w0[1] = K[1];
+  w0[2] = K[2];
+  w0[3] = K[3];
+  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;
+
+  md5_hmac_ctx_t ctx1;
+
+  md5_hmac_init_64 (&ctx1, w0, w1, w2, w3);
+
+  w0[0] = 2;
+  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;
+
+  md5_hmac_update_64 (&ctx1, w0, w1, w2, w3, 4);
+
+  md5_hmac_final (&ctx1);
+
+  w0[0] = ctx1.opad.h[0];
+  w0[1] = ctx1.opad.h[1];
+  w0[2] = ctx1.opad.h[2];
+  w0[3] = ctx1.opad.h[3];
+  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;
+
+  md5_hmac_ctx_t ctx;
+
+  md5_hmac_init_64 (&ctx, w0, w1, w2, w3);
+
+  w0[0] = checksum[0];
+  w0[1] = checksum[1];
+  w0[2] = checksum[2];
+  w0[3] = checksum[3];
+  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;
+
+  md5_hmac_update_64 (&ctx, w0, w1, w2, w3, 16);
+
+  md5_hmac_final (&ctx);
+
+  digest[0] = ctx.opad.h[0];
+  digest[1] = ctx.opad.h[1];
+  digest[2] = ctx.opad.h[2];
+  digest[3] = ctx.opad.h[3];
+
+  K2[0] = ctx1.opad.h[0];
+  K2[1] = ctx1.opad.h[1];
+  K2[2] = ctx1.opad.h[2];
+  K2[3] = ctx1.opad.h[3];
+}
+
+__kernel void m13100_mxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __constant const u32x *words_buf_r, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32x w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    md4_ctx_t ctx;
+
+    md4_init (&ctx);
+
+    md4_update_utf16le (&ctx, w, pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}
+
+__kernel void m13100_sxx (__global pw_t *pws, __global const kernel_rule_t *rules_buf, __global const pw_t *combs_buf, __constant const u32x *words_buf_r, __global void *tmps, __global void *hooks, __global const u32 *bitmaps_buf_s1_a, __global const u32 *bitmaps_buf_s1_b, __global const u32 *bitmaps_buf_s1_c, __global const u32 *bitmaps_buf_s1_d, __global const u32 *bitmaps_buf_s2_a, __global const u32 *bitmaps_buf_s2_b, __global const u32 *bitmaps_buf_s2_c, __global const u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global const digest_t *digests_buf, __global u32 *hashes_shown, __global const salt_t *salt_bufs, __global const krb5tgs_t *krb5tgs_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV0_buf, __global u32 *d_scryptV1_buf, __global u32 *d_scryptV2_buf, __global u32 *d_scryptV3_buf, const u32 bitmap_mask, const u32 bitmap_shift1, const u32 bitmap_shift2, const u32 salt_pos, const u32 loop_pos, const u32 loop_cnt, const u32 il_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
+{
+  /**
+   * modifier
+   */
+
+  const u32 lid = get_local_id (0);
+  const u32 gid = get_global_id (0);
+
+  if (gid >= gid_max) return;
+
+  /**
+   * base
+   */
+
+  const u32 pw_len = pws[gid].pw_len;
+
+  const u32 pw_lenv = ceil ((float) pw_len / 4);
+
+  u32x w[64] = { 0 };
+
+  for (int idx = 0; idx < pw_lenv; idx++)
+  {
+    w[idx] = pws[gid].i[idx];
+
+    barrier (CLK_GLOBAL_MEM_FENCE);
+  }
+
+  __local RC4_KEY rc4_keys[64];
+
+  u32 checksum[4];
+
+  checksum[0] = krb5tgs_bufs[digests_offset].checksum[0];
+  checksum[1] = krb5tgs_bufs[digests_offset].checksum[1];
+  checksum[2] = krb5tgs_bufs[digests_offset].checksum[2];
+  checksum[3] = krb5tgs_bufs[digests_offset].checksum[3];
+
+  /**
+   * loop
+   */
+
+  u32x w0l = w[0];
+
+  for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
+  {
+    const u32x w0r = words_buf_r[il_pos / VECT_SIZE];
+
+    const u32x w0 = w0l | w0r;
+
+    w[0] = w0;
+
+    md4_ctx_t ctx;
+
+    md4_init (&ctx);
+
+    md4_update_utf16le (&ctx, w, pw_len);
+
+    md4_final (&ctx);
+
+    u32 digest[4];
+
+    u32 K2[4];
+
+    kerb_prepare (ctx.h, checksum, digest, K2);
+
+    if (decrypt_and_check (&rc4_keys[lid], digest, krb5tgs_bufs[digests_offset].edata2, krb5tgs_bufs[digests_offset].edata2_len, K2, checksum) == 1)
+    {
+      if (atomic_inc (&hashes_shown[digests_offset]) == 0)
+      {
+        mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, digests_offset + 0, gid, il_pos);
+      }
+    }
+  }
+}