mirror of
https://github.com/hashcat/hashcat.git
synced 2024-11-27 02:18:21 +00:00
839 lines
24 KiB
Common Lisp
839 lines
24 KiB
Common Lisp
/**
|
|
* Author......: See docs/credits.txt
|
|
* License.....: MIT
|
|
*/
|
|
|
|
//#define NEW_SIMD_CODE
|
|
|
|
#ifdef KERNEL_STATIC
|
|
#include "inc_vendor.h"
|
|
#include "inc_types.h"
|
|
#include "inc_platform.cl"
|
|
#include "inc_common.cl"
|
|
#include "inc_simd.cl"
|
|
#include "inc_hash_whirlpool.cl"
|
|
#include "inc_cipher_aes.cl"
|
|
#include "inc_cipher_twofish.cl"
|
|
#include "inc_cipher_serpent.cl"
|
|
#include "inc_cipher_camellia.cl"
|
|
#include "inc_cipher_kuznyechik.cl"
|
|
#endif
|
|
|
|
typedef struct vc
|
|
{
|
|
u32 salt_buf[32];
|
|
u32 data_buf[112];
|
|
u32 keyfile_buf[16];
|
|
u32 signature;
|
|
|
|
keyboard_layout_mapping_t keyboard_layout_mapping_buf[256];
|
|
int keyboard_layout_mapping_cnt;
|
|
|
|
int pim_multi; // 2048 for boot (not SHA-512 or Whirlpool), 1000 for others
|
|
int pim_start;
|
|
int pim_stop;
|
|
|
|
} vc_t;
|
|
|
|
#ifdef KERNEL_STATIC
|
|
#include "inc_truecrypt_keyfile.cl"
|
|
#include "inc_truecrypt_crc32.cl"
|
|
#include "inc_truecrypt_xts.cl"
|
|
#include "inc_veracrypt_xts.cl"
|
|
#endif
|
|
|
|
typedef struct vc_tmp
|
|
{
|
|
u32 ipad[16];
|
|
u32 opad[16];
|
|
|
|
u32 dgst[64];
|
|
u32 out[64];
|
|
|
|
u32 pim_key[64];
|
|
int pim; // marker for cracked
|
|
|
|
} vc_tmp_t;
|
|
|
|
DECLSPEC int check_header_0512 (GLOBAL_AS const vc_t *esalt_bufs, GLOBAL_AS u32 *key, SHM_TYPE u32 *s_te0, SHM_TYPE u32 *s_te1, SHM_TYPE u32 *s_te2, SHM_TYPE u32 *s_te3, SHM_TYPE u32 *s_te4, SHM_TYPE u32 *s_td0, SHM_TYPE u32 *s_td1, SHM_TYPE u32 *s_td2, SHM_TYPE u32 *s_td3, SHM_TYPE u32 *s_td4)
|
|
{
|
|
u32 key1[8];
|
|
u32 key2[8];
|
|
|
|
key1[0] = hc_swap32_S (key[ 0]);
|
|
key1[1] = hc_swap32_S (key[ 1]);
|
|
key1[2] = hc_swap32_S (key[ 2]);
|
|
key1[3] = hc_swap32_S (key[ 3]);
|
|
key1[4] = hc_swap32_S (key[ 4]);
|
|
key1[5] = hc_swap32_S (key[ 5]);
|
|
key1[6] = hc_swap32_S (key[ 6]);
|
|
key1[7] = hc_swap32_S (key[ 7]);
|
|
key2[0] = hc_swap32_S (key[ 8]);
|
|
key2[1] = hc_swap32_S (key[ 9]);
|
|
key2[2] = hc_swap32_S (key[10]);
|
|
key2[3] = hc_swap32_S (key[11]);
|
|
key2[4] = hc_swap32_S (key[12]);
|
|
key2[5] = hc_swap32_S (key[13]);
|
|
key2[6] = hc_swap32_S (key[14]);
|
|
key2[7] = hc_swap32_S (key[15]);
|
|
|
|
if (verify_header_serpent (esalt_bufs[0].data_buf, esalt_bufs[0].signature, key1, key2) == 1) return 0;
|
|
if (verify_header_twofish (esalt_bufs[0].data_buf, esalt_bufs[0].signature, key1, key2) == 1) return 0;
|
|
if (verify_header_camellia (esalt_bufs[0].data_buf, esalt_bufs[0].signature, key1, key2) == 1) return 0;
|
|
if (verify_header_kuznyechik (esalt_bufs[0].data_buf, esalt_bufs[0].signature, key1, key2) == 1) return 0;
|
|
if (verify_header_aes (esalt_bufs[0].data_buf, esalt_bufs[0].signature, key1, key2, s_te0, s_te1, s_te2, s_te3, s_te4, s_td0, s_td1, s_td2, s_td3, s_td4) == 1) return 0;
|
|
|
|
return -1;
|
|
}
|
|
|
|
DECLSPEC void hmac_whirlpool_run_V (u32x *w0, u32x *w1, u32x *w2, u32x *w3, u32x *ipad, u32x *opad, u32x *digest, SHM_TYPE u32 (*s_Ch)[256], SHM_TYPE u32 (*s_Cl)[256])
|
|
{
|
|
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];
|
|
digest[ 8] = ipad[ 8];
|
|
digest[ 9] = ipad[ 9];
|
|
digest[10] = ipad[10];
|
|
digest[11] = ipad[11];
|
|
digest[12] = ipad[12];
|
|
digest[13] = ipad[13];
|
|
digest[14] = ipad[14];
|
|
digest[15] = ipad[15];
|
|
|
|
whirlpool_transform_vector (w0, w1, w2, w3, digest, s_Ch, s_Cl);
|
|
|
|
w0[0] = 0x80000000;
|
|
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] = (64 + 64) * 8;
|
|
|
|
whirlpool_transform_vector (w0, w1, w2, w3, digest, s_Ch, s_Cl);
|
|
|
|
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] = digest[ 8];
|
|
w2[1] = digest[ 9];
|
|
w2[2] = digest[10];
|
|
w2[3] = digest[11];
|
|
w3[0] = digest[12];
|
|
w3[1] = digest[13];
|
|
w3[2] = digest[14];
|
|
w3[3] = digest[15];
|
|
|
|
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];
|
|
digest[ 8] = opad[ 8];
|
|
digest[ 9] = opad[ 9];
|
|
digest[10] = opad[10];
|
|
digest[11] = opad[11];
|
|
digest[12] = opad[12];
|
|
digest[13] = opad[13];
|
|
digest[14] = opad[14];
|
|
digest[15] = opad[15];
|
|
|
|
whirlpool_transform_vector (w0, w1, w2, w3, digest, s_Ch, s_Cl);
|
|
|
|
w0[0] = 0x80000000;
|
|
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] = (64 + 64) * 8;
|
|
|
|
whirlpool_transform_vector (w0, w1, w2, w3, digest, s_Ch, s_Cl);
|
|
}
|
|
|
|
KERNEL_FQ void m13731_init (KERN_ATTR_TMPS_ESALT (vc_tmp_t, vc_t))
|
|
{
|
|
const u64 gid = get_global_id (0);
|
|
const u64 lid = get_local_id (0);
|
|
const u64 lsz = get_local_size (0);
|
|
|
|
/**
|
|
* keyboard layout shared
|
|
*/
|
|
|
|
const int keyboard_layout_mapping_cnt = esalt_bufs[digests_offset].keyboard_layout_mapping_cnt;
|
|
|
|
LOCAL_VK keyboard_layout_mapping_t s_keyboard_layout_mapping_buf[256];
|
|
|
|
for (u32 i = lid; i < 256; i += lsz)
|
|
{
|
|
s_keyboard_layout_mapping_buf[i] = esalt_bufs[digests_offset].keyboard_layout_mapping_buf[i];
|
|
}
|
|
|
|
SYNC_THREADS ();
|
|
|
|
/**
|
|
* Whirlpool shared
|
|
*/
|
|
|
|
#ifdef REAL_SHM
|
|
|
|
LOCAL_VK u32 s_Ch[8][256];
|
|
LOCAL_VK u32 s_Cl[8][256];
|
|
|
|
for (u32 i = lid; i < 256; i += lsz)
|
|
{
|
|
s_Ch[0][i] = Ch[0][i];
|
|
s_Ch[1][i] = Ch[1][i];
|
|
s_Ch[2][i] = Ch[2][i];
|
|
s_Ch[3][i] = Ch[3][i];
|
|
s_Ch[4][i] = Ch[4][i];
|
|
s_Ch[5][i] = Ch[5][i];
|
|
s_Ch[6][i] = Ch[6][i];
|
|
s_Ch[7][i] = Ch[7][i];
|
|
|
|
s_Cl[0][i] = Cl[0][i];
|
|
s_Cl[1][i] = Cl[1][i];
|
|
s_Cl[2][i] = Cl[2][i];
|
|
s_Cl[3][i] = Cl[3][i];
|
|
s_Cl[4][i] = Cl[4][i];
|
|
s_Cl[5][i] = Cl[5][i];
|
|
s_Cl[6][i] = Cl[6][i];
|
|
s_Cl[7][i] = Cl[7][i];
|
|
}
|
|
|
|
SYNC_THREADS ();
|
|
|
|
#else
|
|
|
|
CONSTANT_AS u32a (*s_Ch)[256] = Ch;
|
|
CONSTANT_AS u32a (*s_Cl)[256] = Cl;
|
|
|
|
#endif
|
|
|
|
if (gid >= gid_max) return;
|
|
|
|
/**
|
|
* base
|
|
*/
|
|
|
|
u32 w0[4];
|
|
u32 w1[4];
|
|
u32 w2[4];
|
|
u32 w3[4];
|
|
|
|
w0[0] = pws[gid].i[ 0];
|
|
w0[1] = pws[gid].i[ 1];
|
|
w0[2] = pws[gid].i[ 2];
|
|
w0[3] = pws[gid].i[ 3];
|
|
w1[0] = pws[gid].i[ 4];
|
|
w1[1] = pws[gid].i[ 5];
|
|
w1[2] = pws[gid].i[ 6];
|
|
w1[3] = pws[gid].i[ 7];
|
|
w2[0] = pws[gid].i[ 8];
|
|
w2[1] = pws[gid].i[ 9];
|
|
w2[2] = pws[gid].i[10];
|
|
w2[3] = pws[gid].i[11];
|
|
w3[0] = pws[gid].i[12];
|
|
w3[1] = pws[gid].i[13];
|
|
w3[2] = pws[gid].i[14];
|
|
w3[3] = pws[gid].i[15];
|
|
|
|
const u32 pw_len = pws[gid].pw_len;
|
|
|
|
hc_execute_keyboard_layout_mapping (w0, w1, w2, w3, pw_len, s_keyboard_layout_mapping_buf, keyboard_layout_mapping_cnt);
|
|
|
|
w0[0] = u8add (w0[0], esalt_bufs[digests_offset].keyfile_buf[ 0]);
|
|
w0[1] = u8add (w0[1], esalt_bufs[digests_offset].keyfile_buf[ 1]);
|
|
w0[2] = u8add (w0[2], esalt_bufs[digests_offset].keyfile_buf[ 2]);
|
|
w0[3] = u8add (w0[3], esalt_bufs[digests_offset].keyfile_buf[ 3]);
|
|
w1[0] = u8add (w1[0], esalt_bufs[digests_offset].keyfile_buf[ 4]);
|
|
w1[1] = u8add (w1[1], esalt_bufs[digests_offset].keyfile_buf[ 5]);
|
|
w1[2] = u8add (w1[2], esalt_bufs[digests_offset].keyfile_buf[ 6]);
|
|
w1[3] = u8add (w1[3], esalt_bufs[digests_offset].keyfile_buf[ 7]);
|
|
w2[0] = u8add (w2[0], esalt_bufs[digests_offset].keyfile_buf[ 8]);
|
|
w2[1] = u8add (w2[1], esalt_bufs[digests_offset].keyfile_buf[ 9]);
|
|
w2[2] = u8add (w2[2], esalt_bufs[digests_offset].keyfile_buf[10]);
|
|
w2[3] = u8add (w2[3], esalt_bufs[digests_offset].keyfile_buf[11]);
|
|
w3[0] = u8add (w3[0], esalt_bufs[digests_offset].keyfile_buf[12]);
|
|
w3[1] = u8add (w3[1], esalt_bufs[digests_offset].keyfile_buf[13]);
|
|
w3[2] = u8add (w3[2], esalt_bufs[digests_offset].keyfile_buf[14]);
|
|
w3[3] = u8add (w3[3], esalt_bufs[digests_offset].keyfile_buf[15]);
|
|
|
|
w0[0] = hc_swap32_S (w0[0]);
|
|
w0[1] = hc_swap32_S (w0[1]);
|
|
w0[2] = hc_swap32_S (w0[2]);
|
|
w0[3] = hc_swap32_S (w0[3]);
|
|
w1[0] = hc_swap32_S (w1[0]);
|
|
w1[1] = hc_swap32_S (w1[1]);
|
|
w1[2] = hc_swap32_S (w1[2]);
|
|
w1[3] = hc_swap32_S (w1[3]);
|
|
w2[0] = hc_swap32_S (w2[0]);
|
|
w2[1] = hc_swap32_S (w2[1]);
|
|
w2[2] = hc_swap32_S (w2[2]);
|
|
w2[3] = hc_swap32_S (w2[3]);
|
|
w3[0] = hc_swap32_S (w3[0]);
|
|
w3[1] = hc_swap32_S (w3[1]);
|
|
w3[2] = hc_swap32_S (w3[2]);
|
|
w3[3] = hc_swap32_S (w3[3]);
|
|
|
|
whirlpool_hmac_ctx_t whirlpool_hmac_ctx;
|
|
|
|
whirlpool_hmac_init_64 (&whirlpool_hmac_ctx, w0, w1, w2, w3, s_Ch, s_Cl);
|
|
|
|
tmps[gid].ipad[ 0] = whirlpool_hmac_ctx.ipad.h[ 0];
|
|
tmps[gid].ipad[ 1] = whirlpool_hmac_ctx.ipad.h[ 1];
|
|
tmps[gid].ipad[ 2] = whirlpool_hmac_ctx.ipad.h[ 2];
|
|
tmps[gid].ipad[ 3] = whirlpool_hmac_ctx.ipad.h[ 3];
|
|
tmps[gid].ipad[ 4] = whirlpool_hmac_ctx.ipad.h[ 4];
|
|
tmps[gid].ipad[ 5] = whirlpool_hmac_ctx.ipad.h[ 5];
|
|
tmps[gid].ipad[ 6] = whirlpool_hmac_ctx.ipad.h[ 6];
|
|
tmps[gid].ipad[ 7] = whirlpool_hmac_ctx.ipad.h[ 7];
|
|
tmps[gid].ipad[ 8] = whirlpool_hmac_ctx.ipad.h[ 8];
|
|
tmps[gid].ipad[ 9] = whirlpool_hmac_ctx.ipad.h[ 9];
|
|
tmps[gid].ipad[10] = whirlpool_hmac_ctx.ipad.h[10];
|
|
tmps[gid].ipad[11] = whirlpool_hmac_ctx.ipad.h[11];
|
|
tmps[gid].ipad[12] = whirlpool_hmac_ctx.ipad.h[12];
|
|
tmps[gid].ipad[13] = whirlpool_hmac_ctx.ipad.h[13];
|
|
tmps[gid].ipad[14] = whirlpool_hmac_ctx.ipad.h[14];
|
|
tmps[gid].ipad[15] = whirlpool_hmac_ctx.ipad.h[15];
|
|
|
|
tmps[gid].opad[ 0] = whirlpool_hmac_ctx.opad.h[ 0];
|
|
tmps[gid].opad[ 1] = whirlpool_hmac_ctx.opad.h[ 1];
|
|
tmps[gid].opad[ 2] = whirlpool_hmac_ctx.opad.h[ 2];
|
|
tmps[gid].opad[ 3] = whirlpool_hmac_ctx.opad.h[ 3];
|
|
tmps[gid].opad[ 4] = whirlpool_hmac_ctx.opad.h[ 4];
|
|
tmps[gid].opad[ 5] = whirlpool_hmac_ctx.opad.h[ 5];
|
|
tmps[gid].opad[ 6] = whirlpool_hmac_ctx.opad.h[ 6];
|
|
tmps[gid].opad[ 7] = whirlpool_hmac_ctx.opad.h[ 7];
|
|
tmps[gid].opad[ 8] = whirlpool_hmac_ctx.opad.h[ 8];
|
|
tmps[gid].opad[ 9] = whirlpool_hmac_ctx.opad.h[ 9];
|
|
tmps[gid].opad[10] = whirlpool_hmac_ctx.opad.h[10];
|
|
tmps[gid].opad[11] = whirlpool_hmac_ctx.opad.h[11];
|
|
tmps[gid].opad[12] = whirlpool_hmac_ctx.opad.h[12];
|
|
tmps[gid].opad[13] = whirlpool_hmac_ctx.opad.h[13];
|
|
tmps[gid].opad[14] = whirlpool_hmac_ctx.opad.h[14];
|
|
tmps[gid].opad[15] = whirlpool_hmac_ctx.opad.h[15];
|
|
|
|
whirlpool_hmac_update_global_swap (&whirlpool_hmac_ctx, esalt_bufs[digests_offset].salt_buf, 64);
|
|
|
|
for (u32 i = 0, j = 1; i < 16; i += 16, j += 1)
|
|
{
|
|
whirlpool_hmac_ctx_t whirlpool_hmac_ctx2 = whirlpool_hmac_ctx;
|
|
|
|
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;
|
|
|
|
whirlpool_hmac_update_64 (&whirlpool_hmac_ctx2, w0, w1, w2, w3, 4);
|
|
|
|
whirlpool_hmac_final (&whirlpool_hmac_ctx2);
|
|
|
|
tmps[gid].dgst[i + 0] = whirlpool_hmac_ctx2.opad.h[ 0];
|
|
tmps[gid].dgst[i + 1] = whirlpool_hmac_ctx2.opad.h[ 1];
|
|
tmps[gid].dgst[i + 2] = whirlpool_hmac_ctx2.opad.h[ 2];
|
|
tmps[gid].dgst[i + 3] = whirlpool_hmac_ctx2.opad.h[ 3];
|
|
tmps[gid].dgst[i + 4] = whirlpool_hmac_ctx2.opad.h[ 4];
|
|
tmps[gid].dgst[i + 5] = whirlpool_hmac_ctx2.opad.h[ 5];
|
|
tmps[gid].dgst[i + 6] = whirlpool_hmac_ctx2.opad.h[ 6];
|
|
tmps[gid].dgst[i + 7] = whirlpool_hmac_ctx2.opad.h[ 7];
|
|
tmps[gid].dgst[i + 8] = whirlpool_hmac_ctx2.opad.h[ 8];
|
|
tmps[gid].dgst[i + 9] = whirlpool_hmac_ctx2.opad.h[ 9];
|
|
tmps[gid].dgst[i + 10] = whirlpool_hmac_ctx2.opad.h[10];
|
|
tmps[gid].dgst[i + 11] = whirlpool_hmac_ctx2.opad.h[11];
|
|
tmps[gid].dgst[i + 12] = whirlpool_hmac_ctx2.opad.h[12];
|
|
tmps[gid].dgst[i + 13] = whirlpool_hmac_ctx2.opad.h[13];
|
|
tmps[gid].dgst[i + 14] = whirlpool_hmac_ctx2.opad.h[14];
|
|
tmps[gid].dgst[i + 15] = whirlpool_hmac_ctx2.opad.h[15];
|
|
|
|
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];
|
|
tmps[gid].out[i + 8] = tmps[gid].dgst[i + 8];
|
|
tmps[gid].out[i + 9] = tmps[gid].dgst[i + 9];
|
|
tmps[gid].out[i + 10] = tmps[gid].dgst[i + 10];
|
|
tmps[gid].out[i + 11] = tmps[gid].dgst[i + 11];
|
|
tmps[gid].out[i + 12] = tmps[gid].dgst[i + 12];
|
|
tmps[gid].out[i + 13] = tmps[gid].dgst[i + 13];
|
|
tmps[gid].out[i + 14] = tmps[gid].dgst[i + 14];
|
|
tmps[gid].out[i + 15] = tmps[gid].dgst[i + 15];
|
|
}
|
|
}
|
|
|
|
KERNEL_FQ void m13731_loop (KERN_ATTR_TMPS_ESALT (vc_tmp_t, vc_t))
|
|
{
|
|
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_td0[256];
|
|
LOCAL_VK u32 s_td1[256];
|
|
LOCAL_VK u32 s_td2[256];
|
|
LOCAL_VK u32 s_td3[256];
|
|
LOCAL_VK u32 s_td4[256];
|
|
|
|
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_td0[i] = td0[i];
|
|
s_td1[i] = td1[i];
|
|
s_td2[i] = td2[i];
|
|
s_td3[i] = td3[i];
|
|
s_td4[i] = td4[i];
|
|
|
|
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_td0 = td0;
|
|
CONSTANT_AS u32a *s_td1 = td1;
|
|
CONSTANT_AS u32a *s_td2 = td2;
|
|
CONSTANT_AS u32a *s_td3 = td3;
|
|
CONSTANT_AS u32a *s_td4 = td4;
|
|
|
|
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
|
|
|
|
/**
|
|
* Whirlpool shared
|
|
*/
|
|
|
|
#ifdef REAL_SHM
|
|
|
|
LOCAL_VK u32 s_Ch[8][256];
|
|
LOCAL_VK u32 s_Cl[8][256];
|
|
|
|
for (u32 i = lid; i < 256; i += lsz)
|
|
{
|
|
s_Ch[0][i] = Ch[0][i];
|
|
s_Ch[1][i] = Ch[1][i];
|
|
s_Ch[2][i] = Ch[2][i];
|
|
s_Ch[3][i] = Ch[3][i];
|
|
s_Ch[4][i] = Ch[4][i];
|
|
s_Ch[5][i] = Ch[5][i];
|
|
s_Ch[6][i] = Ch[6][i];
|
|
s_Ch[7][i] = Ch[7][i];
|
|
|
|
s_Cl[0][i] = Cl[0][i];
|
|
s_Cl[1][i] = Cl[1][i];
|
|
s_Cl[2][i] = Cl[2][i];
|
|
s_Cl[3][i] = Cl[3][i];
|
|
s_Cl[4][i] = Cl[4][i];
|
|
s_Cl[5][i] = Cl[5][i];
|
|
s_Cl[6][i] = Cl[6][i];
|
|
s_Cl[7][i] = Cl[7][i];
|
|
}
|
|
|
|
SYNC_THREADS ();
|
|
|
|
#else
|
|
|
|
CONSTANT_AS u32a (*s_Ch)[256] = Ch;
|
|
CONSTANT_AS u32a (*s_Cl)[256] = Cl;
|
|
|
|
#endif
|
|
|
|
if (gid >= gid_max) return;
|
|
|
|
// this is the pim range check
|
|
// it is guaranteed that only 0 or 1 innerloops will match a "pim" mark (each 1000 iterations)
|
|
// therefore the module limits the inner loop iteration count to 1000
|
|
// if the key_pim is set, we know that we have to save and check the key for this pim
|
|
|
|
const int pim_multi = esalt_bufs[digests_offset].pim_multi;
|
|
const int pim_start = esalt_bufs[digests_offset].pim_start;
|
|
const int pim_stop = esalt_bufs[digests_offset].pim_stop;
|
|
|
|
int pim = 0;
|
|
int pim_at = 0;
|
|
|
|
for (u32 j = 0; j < loop_cnt; j++)
|
|
{
|
|
const int iter_abs = 1 + loop_pos + j;
|
|
|
|
if ((iter_abs % pim_multi) == pim_multi - 1)
|
|
{
|
|
const int pim_cur = (iter_abs / pim_multi) + 1;
|
|
|
|
if ((pim_cur >= pim_start) && (pim_cur <= pim_stop))
|
|
{
|
|
pim = pim_cur;
|
|
|
|
pim_at = j;
|
|
}
|
|
}
|
|
}
|
|
|
|
// irregular pbkdf2 from here
|
|
|
|
u32x ipad[16];
|
|
u32x opad[16];
|
|
|
|
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);
|
|
ipad[ 8] = packv (tmps, ipad, gid, 8);
|
|
ipad[ 9] = packv (tmps, ipad, gid, 9);
|
|
ipad[10] = packv (tmps, ipad, gid, 10);
|
|
ipad[11] = packv (tmps, ipad, gid, 11);
|
|
ipad[12] = packv (tmps, ipad, gid, 12);
|
|
ipad[13] = packv (tmps, ipad, gid, 13);
|
|
ipad[14] = packv (tmps, ipad, gid, 14);
|
|
ipad[15] = packv (tmps, ipad, gid, 15);
|
|
|
|
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);
|
|
opad[ 8] = packv (tmps, opad, gid, 8);
|
|
opad[ 9] = packv (tmps, opad, gid, 9);
|
|
opad[10] = packv (tmps, opad, gid, 10);
|
|
opad[11] = packv (tmps, opad, gid, 11);
|
|
opad[12] = packv (tmps, opad, gid, 12);
|
|
opad[13] = packv (tmps, opad, gid, 13);
|
|
opad[14] = packv (tmps, opad, gid, 14);
|
|
opad[15] = packv (tmps, opad, gid, 15);
|
|
|
|
for (u32 i = 0; i < 16; i += 16)
|
|
{
|
|
u32x dgst[16];
|
|
u32x out[16];
|
|
|
|
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);
|
|
dgst[ 8] = packv (tmps, dgst, gid, i + 8);
|
|
dgst[ 9] = packv (tmps, dgst, gid, i + 9);
|
|
dgst[10] = packv (tmps, dgst, gid, i + 10);
|
|
dgst[11] = packv (tmps, dgst, gid, i + 11);
|
|
dgst[12] = packv (tmps, dgst, gid, i + 12);
|
|
dgst[13] = packv (tmps, dgst, gid, i + 13);
|
|
dgst[14] = packv (tmps, dgst, gid, i + 14);
|
|
dgst[15] = packv (tmps, dgst, gid, i + 15);
|
|
|
|
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);
|
|
out[ 8] = packv (tmps, out, gid, i + 8);
|
|
out[ 9] = packv (tmps, out, gid, i + 9);
|
|
out[10] = packv (tmps, out, gid, i + 10);
|
|
out[11] = packv (tmps, out, gid, i + 11);
|
|
out[12] = packv (tmps, out, gid, i + 12);
|
|
out[13] = packv (tmps, out, gid, i + 13);
|
|
out[14] = packv (tmps, out, gid, i + 14);
|
|
out[15] = packv (tmps, out, gid, i + 15);
|
|
|
|
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] = dgst[ 8];
|
|
w2[1] = dgst[ 9];
|
|
w2[2] = dgst[10];
|
|
w2[3] = dgst[11];
|
|
w3[0] = dgst[12];
|
|
w3[1] = dgst[13];
|
|
w3[2] = dgst[14];
|
|
w3[3] = dgst[15];
|
|
|
|
hmac_whirlpool_run_V (w0, w1, w2, w3, ipad, opad, dgst, s_Ch, s_Cl);
|
|
|
|
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];
|
|
out[ 8] ^= dgst[ 8];
|
|
out[ 9] ^= dgst[ 9];
|
|
out[10] ^= dgst[10];
|
|
out[11] ^= dgst[11];
|
|
out[12] ^= dgst[12];
|
|
out[13] ^= dgst[13];
|
|
out[14] ^= dgst[14];
|
|
out[15] ^= dgst[15];
|
|
|
|
// this iteration creates a valid pim
|
|
|
|
if (j == pim_at)
|
|
{
|
|
tmps[gid].pim_key[i + 0] = out[ 0];
|
|
tmps[gid].pim_key[i + 1] = out[ 1];
|
|
tmps[gid].pim_key[i + 2] = out[ 2];
|
|
tmps[gid].pim_key[i + 3] = out[ 3];
|
|
tmps[gid].pim_key[i + 4] = out[ 4];
|
|
tmps[gid].pim_key[i + 5] = out[ 5];
|
|
tmps[gid].pim_key[i + 6] = out[ 6];
|
|
tmps[gid].pim_key[i + 7] = out[ 7];
|
|
tmps[gid].pim_key[i + 8] = out[ 8];
|
|
tmps[gid].pim_key[i + 9] = out[ 9];
|
|
tmps[gid].pim_key[i + 10] = out[10];
|
|
tmps[gid].pim_key[i + 11] = out[11];
|
|
tmps[gid].pim_key[i + 12] = out[12];
|
|
tmps[gid].pim_key[i + 13] = out[13];
|
|
tmps[gid].pim_key[i + 14] = out[14];
|
|
tmps[gid].pim_key[i + 15] = out[15];
|
|
}
|
|
}
|
|
|
|
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, dgst, gid, i + 8, dgst[ 8]);
|
|
unpackv (tmps, dgst, gid, i + 9, dgst[ 9]);
|
|
unpackv (tmps, dgst, gid, i + 10, dgst[10]);
|
|
unpackv (tmps, dgst, gid, i + 11, dgst[11]);
|
|
unpackv (tmps, dgst, gid, i + 12, dgst[12]);
|
|
unpackv (tmps, dgst, gid, i + 13, dgst[13]);
|
|
unpackv (tmps, dgst, gid, i + 14, dgst[14]);
|
|
unpackv (tmps, dgst, gid, i + 15, dgst[15]);
|
|
|
|
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]);
|
|
unpackv (tmps, out, gid, i + 8, out[ 8]);
|
|
unpackv (tmps, out, gid, i + 9, out[ 9]);
|
|
unpackv (tmps, out, gid, i + 10, out[10]);
|
|
unpackv (tmps, out, gid, i + 11, out[11]);
|
|
unpackv (tmps, out, gid, i + 12, out[12]);
|
|
unpackv (tmps, out, gid, i + 13, out[13]);
|
|
unpackv (tmps, out, gid, i + 14, out[14]);
|
|
unpackv (tmps, out, gid, i + 15, out[15]);
|
|
}
|
|
|
|
if (pim == 0) return;
|
|
|
|
if (check_header_0512 (esalt_bufs, tmps[gid].pim_key, s_te0, s_te1, s_te2, s_te3, s_te4, s_td0, s_td1, s_td2, s_td3, s_td4) != -1) tmps[gid].pim = pim;
|
|
}
|
|
|
|
KERNEL_FQ void m13731_comp (KERN_ATTR_TMPS_ESALT (vc_tmp_t, vc_t))
|
|
{
|
|
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_td0[256];
|
|
LOCAL_VK u32 s_td1[256];
|
|
LOCAL_VK u32 s_td2[256];
|
|
LOCAL_VK u32 s_td3[256];
|
|
LOCAL_VK u32 s_td4[256];
|
|
|
|
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_td0[i] = td0[i];
|
|
s_td1[i] = td1[i];
|
|
s_td2[i] = td2[i];
|
|
s_td3[i] = td3[i];
|
|
s_td4[i] = td4[i];
|
|
|
|
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_td0 = td0;
|
|
CONSTANT_AS u32a *s_td1 = td1;
|
|
CONSTANT_AS u32a *s_td2 = td2;
|
|
CONSTANT_AS u32a *s_td3 = td3;
|
|
CONSTANT_AS u32a *s_td4 = td4;
|
|
|
|
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
|
|
|
|
/**
|
|
* Whirlpool shared
|
|
*/
|
|
|
|
#ifdef REAL_SHM
|
|
|
|
LOCAL_VK u32 s_Ch[8][256];
|
|
LOCAL_VK u32 s_Cl[8][256];
|
|
|
|
for (u32 i = lid; i < 256; i += lsz)
|
|
{
|
|
s_Ch[0][i] = Ch[0][i];
|
|
s_Ch[1][i] = Ch[1][i];
|
|
s_Ch[2][i] = Ch[2][i];
|
|
s_Ch[3][i] = Ch[3][i];
|
|
s_Ch[4][i] = Ch[4][i];
|
|
s_Ch[5][i] = Ch[5][i];
|
|
s_Ch[6][i] = Ch[6][i];
|
|
s_Ch[7][i] = Ch[7][i];
|
|
|
|
s_Cl[0][i] = Cl[0][i];
|
|
s_Cl[1][i] = Cl[1][i];
|
|
s_Cl[2][i] = Cl[2][i];
|
|
s_Cl[3][i] = Cl[3][i];
|
|
s_Cl[4][i] = Cl[4][i];
|
|
s_Cl[5][i] = Cl[5][i];
|
|
s_Cl[6][i] = Cl[6][i];
|
|
s_Cl[7][i] = Cl[7][i];
|
|
}
|
|
|
|
SYNC_THREADS ();
|
|
|
|
#else
|
|
|
|
CONSTANT_AS u32a (*s_Ch)[256] = Ch;
|
|
CONSTANT_AS u32a (*s_Cl)[256] = Cl;
|
|
|
|
#endif
|
|
|
|
if (gid >= gid_max) return;
|
|
|
|
if (tmps[gid].pim)
|
|
{
|
|
if (atomic_inc (&hashes_shown[0]) == 0)
|
|
{
|
|
mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, 0, gid, 0, 0, 0);
|
|
}
|
|
}
|
|
else
|
|
{
|
|
if (check_header_0512 (esalt_bufs, tmps[gid].out, s_te0, s_te1, s_te2, s_te3, s_te4, s_td0, s_td1, s_td2, s_td3, s_td4) != -1)
|
|
{
|
|
if (atomic_inc (&hashes_shown[0]) == 0)
|
|
{
|
|
mark_hash (plains_buf, d_return_buf, salt_pos, digests_cnt, 0, 0, gid, 0, 0, 0);
|
|
}
|
|
}
|
|
}
|
|
}
|