1
0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-14 03:39:09 +00:00
hashcat/OpenCL/m23900-pure.cl

455 lines
9.7 KiB
Common Lisp

/**
* 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_hash_sha256.cl)
#include M2S(INCLUDE_PATH/inc_cipher_aes.cl)
#endif
#define MIN(a,b) (((a) < (b)) ? (a) : (b))
typedef struct bestcrypt_tmp
{
u32 salt_pw_buf[33];
u32 out[8];
} bestcrypt_tmp_t;
typedef struct bestcrypt
{
u32 data[24];
} bestcrypt_t;
KERNEL_FQ void m23900_init (KERN_ATTR_TMPS_ESALT (bestcrypt_tmp_t, bestcrypt_t))
{
const u64 gid = get_global_id (0);
if (gid >= GID_CNT) return;
const int salt_pw_len = 8 + MIN (pws[gid].pw_len, 56);
u32 comb[16];
comb[ 0] = salt_bufs[SALT_POS_HOST].salt_buf[0];
comb[ 1] = salt_bufs[SALT_POS_HOST].salt_buf[1];
comb[ 2] = hc_swap32_S (pws[gid].i[ 0]); // in theory BE is faster because it
comb[ 3] = hc_swap32_S (pws[gid].i[ 1]); // avoids several other byte swaps later on
comb[ 4] = hc_swap32_S (pws[gid].i[ 2]);
comb[ 5] = hc_swap32_S (pws[gid].i[ 3]);
comb[ 6] = hc_swap32_S (pws[gid].i[ 4]);
comb[ 7] = hc_swap32_S (pws[gid].i[ 5]);
comb[ 8] = hc_swap32_S (pws[gid].i[ 6]);
comb[ 9] = hc_swap32_S (pws[gid].i[ 7]);
comb[10] = hc_swap32_S (pws[gid].i[ 8]);
comb[11] = hc_swap32_S (pws[gid].i[ 9]);
comb[12] = hc_swap32_S (pws[gid].i[10]);
comb[13] = hc_swap32_S (pws[gid].i[11]);
comb[14] = hc_swap32_S (pws[gid].i[12]);
comb[15] = hc_swap32_S (pws[gid].i[13]);
u32 salt_pw_buf[32 + 1] = { 0 }; // 8 + 56 + 64 = 128 bytes
for (int i = 0; i < 128; i += salt_pw_len)
{
const int idx = i / 4;
const int mod = i % 4;
const int full_len = MIN (salt_pw_len, 128 - i);
const int copy_len = (full_len + 3) / 4; // ceil () + convert to 4-byte block (u32)
for (int j = 0, k = idx; j < copy_len; j++, k++)
{
// salt_pw_buf[k] |= comb[j] >> (mod * 8);
// if (mod) salt_pw_buf[k + 1] |= comb[j] << ((4 - mod) * 8);
switch (mod)
{
case 0:
salt_pw_buf[k + 0] |= comb[j];
break;
case 1:
salt_pw_buf[k + 0] |= comb[j] >> 8;
salt_pw_buf[k + 1] |= comb[j] << 24;
break;
case 2:
salt_pw_buf[k + 0] |= comb[j] >> 16;
salt_pw_buf[k + 1] |= comb[j] << 16;
break;
case 3:
salt_pw_buf[k + 0] |= comb[j] >> 24;
salt_pw_buf[k + 1] |= comb[j] << 8;
break;
}
}
}
#ifdef _unroll
#pragma unroll
#endif
for (int i = 0; i < 33; i++)
{
tmps[gid].salt_pw_buf[i] = salt_pw_buf[i];
}
}
KERNEL_FQ void m23900_loop (KERN_ATTR_TMPS_ESALT (bestcrypt_tmp_t, bestcrypt_t))
{
const u64 gid = get_global_id (0);
if (gid >= GID_CNT) return;
const int salt_pw_len = 8 + MIN (pws[gid].pw_len, 56);
u32 salt_pw_buf[32 + 1]; // 8 + 56 + 64 = 128 bytes
#ifdef _unroll
#pragma unroll
#endif
for (int i = 0; i < 33; i++)
{
salt_pw_buf[i] = tmps[gid].salt_pw_buf[i];
}
u32 tbl[1024] = { 0 }; // 4 KiB lookup table
for (int i = 0; i < 64; i++)
{
const int idx = i / 4;
const int mod = i % 4;
// init:
int k = i * 16;
int l = idx;
// tbl[k] |= salt_pw_buf[l] << (mod * 8);
switch (mod)
{
case 0:
tbl[k] |= salt_pw_buf[l];
break;
case 1:
tbl[k] |= salt_pw_buf[l] << 8;
break;
case 2:
tbl[k] |= salt_pw_buf[l] << 16;
break;
case 3:
tbl[k] |= salt_pw_buf[l] << 24;
break;
}
k += 1;
l += 1;
// loop:
for (int j = 1; j < 16; j++, k++, l++)
{
// if (mod) tbl[k - 1] |= salt_pw_buf[l] >> ((4 - mod) * 8);
// tbl[k] |= salt_pw_buf[l] << (mod * 8);
switch (mod)
{
case 0:
tbl[k - 0] |= salt_pw_buf[l];
break;
case 1:
tbl[k - 0] |= salt_pw_buf[l] << 8;
tbl[k - 1] |= salt_pw_buf[l] >> 24;
break;
case 2:
tbl[k - 0] |= salt_pw_buf[l] << 16;
tbl[k - 1] |= salt_pw_buf[l] >> 16;
break;
case 3:
tbl[k - 0] |= salt_pw_buf[l] << 24;
tbl[k - 1] |= salt_pw_buf[l] >> 8;
break;
}
}
// final:
// if (mod) tbl[k - 1] |= salt_pw_buf[l] >> ((4 - mod) * 8);
switch (mod)
{
case 0:
break;
case 1:
tbl[k - 1] |= salt_pw_buf[l] >> 24;
break;
case 2:
tbl[k - 1] |= salt_pw_buf[l] >> 16;
break;
case 3:
tbl[k - 1] |= salt_pw_buf[l] >> 8;
break;
}
}
u32 digest[8];
digest[0] = SHA256M_A;
digest[1] = SHA256M_B;
digest[2] = SHA256M_C;
digest[3] = SHA256M_D;
digest[4] = SHA256M_E;
digest[5] = SHA256M_F;
digest[6] = SHA256M_G;
digest[7] = SHA256M_H;
for (int i = 0; i < 65536; i += 64)
{
const int idx = (i % salt_pw_len) * 16;
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
w0[0] = tbl[idx + 0];
w0[1] = tbl[idx + 1];
w0[2] = tbl[idx + 2];
w0[3] = tbl[idx + 3];
w1[0] = tbl[idx + 4];
w1[1] = tbl[idx + 5];
w1[2] = tbl[idx + 6];
w1[3] = tbl[idx + 7];
w2[0] = tbl[idx + 8];
w2[1] = tbl[idx + 9];
w2[2] = tbl[idx + 10];
w2[3] = tbl[idx + 11];
w3[0] = tbl[idx + 12];
w3[1] = tbl[idx + 13];
w3[2] = tbl[idx + 14];
w3[3] = tbl[idx + 15];
sha256_transform (w0, w1, w2, w3, digest);
}
tmps[gid].out[0] = digest[0];
tmps[gid].out[1] = digest[1];
tmps[gid].out[2] = digest[2];
tmps[gid].out[3] = digest[3];
tmps[gid].out[4] = digest[4];
tmps[gid].out[5] = digest[5];
tmps[gid].out[6] = digest[6];
tmps[gid].out[7] = digest[7];
}
KERNEL_FQ void m23900_comp (KERN_ATTR_TMPS_ESALT (bestcrypt_tmp_t, bestcrypt_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
if (gid >= GID_CNT) return;
// final transform of sha256:
u32 digest[8];
digest[0] = tmps[gid].out[0];
digest[1] = tmps[gid].out[1];
digest[2] = tmps[gid].out[2];
digest[3] = tmps[gid].out[3];
digest[4] = tmps[gid].out[4];
digest[5] = tmps[gid].out[5];
digest[6] = tmps[gid].out[6];
digest[7] = tmps[gid].out[7];
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
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] = 65536 * 8;
sha256_transform (w0, w1, w2, w3, digest);
/**
* AES part
*/
#define KEYLEN 60
u32 ks[KEYLEN];
AES256_set_decrypt_key (ks, digest, s_te0, s_te1, s_te2, s_te3, s_td0, s_td1, s_td2, s_td3);
u32 iv[4] = { 0 };
u32 res[20]; // full would be 24 x u32 (96 bytes)
for (u32 i = 0; i < 20; i += 4) // 96 bytes output would contain the full 32 byte checksum
{
u32 data[4];
data[0] = esalt_bufs[DIGESTS_OFFSET_HOST].data[i + 0];
data[1] = esalt_bufs[DIGESTS_OFFSET_HOST].data[i + 1];
data[2] = esalt_bufs[DIGESTS_OFFSET_HOST].data[i + 2];
data[3] = esalt_bufs[DIGESTS_OFFSET_HOST].data[i + 3];
u32 out[4];
aes256_decrypt (ks, data, out, s_td0, s_td1, s_td2, s_td3, s_td4);
res[i + 0] = hc_swap32_S (out[0] ^ iv[0]);
res[i + 1] = hc_swap32_S (out[1] ^ iv[1]);
res[i + 2] = hc_swap32_S (out[2] ^ iv[2]);
res[i + 3] = hc_swap32_S (out[3] ^ iv[3]);
iv[0] = data[0];
iv[1] = data[1];
iv[2] = data[2];
iv[3] = data[3];
}
// checksum:
// sha256_ctx_t ctx;
// sha256_init (&ctx);
// sha256_update_swap (&ctx, res, 64);
// sha256_final (&ctx);
digest[0] = SHA256M_A;
digest[1] = SHA256M_B;
digest[2] = SHA256M_C;
digest[3] = SHA256M_D;
digest[4] = SHA256M_E;
digest[5] = SHA256M_F;
digest[6] = SHA256M_G;
digest[7] = SHA256M_H;
w0[0] = res[ 0];
w0[1] = res[ 1];
w0[2] = res[ 2];
w0[3] = res[ 3];
w1[0] = res[ 4];
w1[1] = res[ 5];
w1[2] = res[ 6];
w1[3] = res[ 7];
w2[0] = res[ 8];
w2[1] = res[ 9];
w2[2] = res[10];
w2[3] = res[11];
w3[0] = res[12];
w3[1] = res[13];
w3[2] = res[14];
w3[3] = res[15];
sha256_transform (w0, w1, w2, w3, digest);
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 * 8;
sha256_transform (w0, w1, w2, w3, digest);
if ((digest[0] == res[16]) &&
(digest[1] == res[17]) &&
(digest[2] == res[18]) &&
(digest[3] == res[19]))
{
if (hc_atomic_inc (&hashes_shown[DIGESTS_OFFSET_HOST]) == 0)
{
mark_hash (plains_buf, d_return_buf, SALT_POS_HOST, DIGESTS_CNT, 0, DIGESTS_OFFSET_HOST + 0, gid, 0, 0, 0);
}
return;
}
}