1
0
mirror of https://github.com/hashcat/hashcat.git synced 2024-11-18 14:18:11 +00:00
hashcat/OpenCL/m25000-pure.cl
2022-02-07 13:31:22 +01:00

593 lines
14 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_simd.cl)
#include M2S(INCLUDE_PATH/inc_hash_md5.cl)
#include M2S(INCLUDE_PATH/inc_hash_sha1.cl)
#endif
#define COMPARE_S M2S(INCLUDE_PATH/inc_comp_single.cl)
#define COMPARE_M M2S(INCLUDE_PATH/inc_comp_multi.cl)
#define SNMPV3_SALT_MAX 1500
#define SNMPV3_ENGINEID_MAX 34
#define SNMPV3_MSG_AUTH_PARAMS_LEN 12
#define SNMPV3_ROUNDS 1048576
#define SNMPV3_MAX_PW_LENGTH 64
#define SNMPV3_TMP_ELEMS 4096 // 4096 = (256 (max pw length) * 64) / sizeof (u32)
#define SNMPV3_HASH_ELEMS_MD5 4
#define SNMPV3_HASH_ELEMS_SHA1 8 // 8 = aligned 5
#define SNMPV3_MAX_SALT_ELEMS 512 // 512 * 4 = 2048 > 1500, also has to be multiple of 64
#define SNMPV3_MAX_ENGINE_ELEMS 16 // 16 * 4 = 64 > 32, also has to be multiple of 64
#define SNMPV3_MAX_PNUM_ELEMS 4 // 4 * 4 = 16 > 9
#define SNMPV3_MAX_PW_LENGTH_OPT 64
#define SNMPV3_TMP_ELEMS_OPT ((SNMPV3_MAX_PW_LENGTH_OPT * SNMPV3_MAX_PW_LENGTH) / 4)
// (64 * 64) / 4 = 1024
// for pw length > 64 we use global memory reads
typedef struct hmac_md5_tmp
{
u32 tmp_md5[SNMPV3_TMP_ELEMS];
u32 tmp_sha1[SNMPV3_TMP_ELEMS];
u32 h_md5[SNMPV3_HASH_ELEMS_MD5];
u32 h_sha1[SNMPV3_HASH_ELEMS_SHA1];
} hmac_md5_tmp_t;
typedef struct snmpv3
{
u32 salt_buf[SNMPV3_MAX_SALT_ELEMS];
u32 salt_len;
u32 engineID_buf[SNMPV3_MAX_ENGINE_ELEMS];
u32 engineID_len;
u32 packet_number[SNMPV3_MAX_PNUM_ELEMS];
} snmpv3_t;
KERNEL_FQ void m25000_init (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* modifier
*/
const u64 gid = get_global_id (0);
if (gid >= GID_CNT) return;
/**
* base
*/
const u32 pw_len = pws[gid].pw_len;
u32 w[64] = { 0 };
for (u32 i = 0, idx = 0; i < pw_len; i += 4, idx += 1)
{
w[idx] = pws[gid].i[idx];
}
PRIVATE_AS u8 *src_ptr = (PRIVATE_AS u8 *) w;
// password 64 times, also swapped
u32 dst_buf[16];
PRIVATE_AS u8 *dst_ptr = (PRIVATE_AS u8 *) dst_buf;
int tmp_idx = 0;
for (int i = 0; i < 64; i++)
{
for (int j = 0; j < pw_len; j++)
{
const int dst_idx = tmp_idx & 63;
dst_ptr[dst_idx] = src_ptr[j];
// write to global memory every time 64 byte are written into cache
if (dst_idx == 63)
{
const int tmp_idx4 = (tmp_idx - 63) / 4;
// md5
tmps[gid].tmp_md5[tmp_idx4 + 0] = dst_buf[ 0];
tmps[gid].tmp_md5[tmp_idx4 + 1] = dst_buf[ 1];
tmps[gid].tmp_md5[tmp_idx4 + 2] = dst_buf[ 2];
tmps[gid].tmp_md5[tmp_idx4 + 3] = dst_buf[ 3];
tmps[gid].tmp_md5[tmp_idx4 + 4] = dst_buf[ 4];
tmps[gid].tmp_md5[tmp_idx4 + 5] = dst_buf[ 5];
tmps[gid].tmp_md5[tmp_idx4 + 6] = dst_buf[ 6];
tmps[gid].tmp_md5[tmp_idx4 + 7] = dst_buf[ 7];
tmps[gid].tmp_md5[tmp_idx4 + 8] = dst_buf[ 8];
tmps[gid].tmp_md5[tmp_idx4 + 9] = dst_buf[ 9];
tmps[gid].tmp_md5[tmp_idx4 + 10] = dst_buf[10];
tmps[gid].tmp_md5[tmp_idx4 + 11] = dst_buf[11];
tmps[gid].tmp_md5[tmp_idx4 + 12] = dst_buf[12];
tmps[gid].tmp_md5[tmp_idx4 + 13] = dst_buf[13];
tmps[gid].tmp_md5[tmp_idx4 + 14] = dst_buf[14];
tmps[gid].tmp_md5[tmp_idx4 + 15] = dst_buf[15];
// sha1
tmps[gid].tmp_sha1[tmp_idx4 + 0] = hc_swap32_S (dst_buf[ 0]);
tmps[gid].tmp_sha1[tmp_idx4 + 1] = hc_swap32_S (dst_buf[ 1]);
tmps[gid].tmp_sha1[tmp_idx4 + 2] = hc_swap32_S (dst_buf[ 2]);
tmps[gid].tmp_sha1[tmp_idx4 + 3] = hc_swap32_S (dst_buf[ 3]);
tmps[gid].tmp_sha1[tmp_idx4 + 4] = hc_swap32_S (dst_buf[ 4]);
tmps[gid].tmp_sha1[tmp_idx4 + 5] = hc_swap32_S (dst_buf[ 5]);
tmps[gid].tmp_sha1[tmp_idx4 + 6] = hc_swap32_S (dst_buf[ 6]);
tmps[gid].tmp_sha1[tmp_idx4 + 7] = hc_swap32_S (dst_buf[ 7]);
tmps[gid].tmp_sha1[tmp_idx4 + 8] = hc_swap32_S (dst_buf[ 8]);
tmps[gid].tmp_sha1[tmp_idx4 + 9] = hc_swap32_S (dst_buf[ 9]);
tmps[gid].tmp_sha1[tmp_idx4 + 10] = hc_swap32_S (dst_buf[10]);
tmps[gid].tmp_sha1[tmp_idx4 + 11] = hc_swap32_S (dst_buf[11]);
tmps[gid].tmp_sha1[tmp_idx4 + 12] = hc_swap32_S (dst_buf[12]);
tmps[gid].tmp_sha1[tmp_idx4 + 13] = hc_swap32_S (dst_buf[13]);
tmps[gid].tmp_sha1[tmp_idx4 + 14] = hc_swap32_S (dst_buf[14]);
tmps[gid].tmp_sha1[tmp_idx4 + 15] = hc_swap32_S (dst_buf[15]);
}
tmp_idx++;
}
}
// hash md5
tmps[gid].h_md5[0] = MD5M_A;
tmps[gid].h_md5[1] = MD5M_B;
tmps[gid].h_md5[2] = MD5M_C;
tmps[gid].h_md5[3] = MD5M_D;
// hash sha1
tmps[gid].h_sha1[0] = SHA1M_A;
tmps[gid].h_sha1[1] = SHA1M_B;
tmps[gid].h_sha1[2] = SHA1M_C;
tmps[gid].h_sha1[3] = SHA1M_D;
tmps[gid].h_sha1[4] = SHA1M_E;
}
KERNEL_FQ void m25000_loop (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* base
*/
const u64 gid = get_global_id (0);
if (gid >= GID_CNT) return;
u32 h_md5[4];
h_md5[0] = tmps[gid].h_md5[0];
h_md5[1] = tmps[gid].h_md5[1];
h_md5[2] = tmps[gid].h_md5[2];
h_md5[3] = tmps[gid].h_md5[3];
u32 h_sha1[5];
h_sha1[0] = tmps[gid].h_sha1[0];
h_sha1[1] = tmps[gid].h_sha1[1];
h_sha1[2] = tmps[gid].h_sha1[2];
h_sha1[3] = tmps[gid].h_sha1[3];
h_sha1[4] = tmps[gid].h_sha1[4];
const u32 pw_len = pws[gid].pw_len;
const int pw_len64 = pw_len * 64;
if (pw_len <= SNMPV3_MAX_PW_LENGTH_OPT)
{
u32 tmp_shared[SNMPV3_TMP_ELEMS_OPT];
// md5
for (int i = 0; i < pw_len64 / 4; i++)
{
tmp_shared[i] = tmps[gid].tmp_md5[i];
}
for (int i = 0, j = LOOP_POS; i < LOOP_CNT; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
w0[0] = tmp_shared[idx + 0];
w0[1] = tmp_shared[idx + 1];
w0[2] = tmp_shared[idx + 2];
w0[3] = tmp_shared[idx + 3];
w1[0] = tmp_shared[idx + 4];
w1[1] = tmp_shared[idx + 5];
w1[2] = tmp_shared[idx + 6];
w1[3] = tmp_shared[idx + 7];
w2[0] = tmp_shared[idx + 8];
w2[1] = tmp_shared[idx + 9];
w2[2] = tmp_shared[idx + 10];
w2[3] = tmp_shared[idx + 11];
w3[0] = tmp_shared[idx + 12];
w3[1] = tmp_shared[idx + 13];
w3[2] = tmp_shared[idx + 14];
w3[3] = tmp_shared[idx + 15];
md5_transform (w0, w1, w2, w3, h_md5);
}
// sha1
for (int i = 0; i < pw_len64 / 4; i++)
{
tmp_shared[i] = tmps[gid].tmp_sha1[i];
}
for (int i = 0, j = LOOP_POS; i < LOOP_CNT; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
w0[0] = tmp_shared[idx + 0];
w0[1] = tmp_shared[idx + 1];
w0[2] = tmp_shared[idx + 2];
w0[3] = tmp_shared[idx + 3];
w1[0] = tmp_shared[idx + 4];
w1[1] = tmp_shared[idx + 5];
w1[2] = tmp_shared[idx + 6];
w1[3] = tmp_shared[idx + 7];
w2[0] = tmp_shared[idx + 8];
w2[1] = tmp_shared[idx + 9];
w2[2] = tmp_shared[idx + 10];
w2[3] = tmp_shared[idx + 11];
w3[0] = tmp_shared[idx + 12];
w3[1] = tmp_shared[idx + 13];
w3[2] = tmp_shared[idx + 14];
w3[3] = tmp_shared[idx + 15];
sha1_transform (w0, w1, w2, w3, h_sha1);
}
}
else
{
for (int i = 0, j = LOOP_POS; i < LOOP_CNT; i += 64, j += 64)
{
const int idx = (j % pw_len64) / 4; // the optimization trick is to be able to do this
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
// md5
w0[0] = tmps[gid].tmp_md5[idx + 0];
w0[1] = tmps[gid].tmp_md5[idx + 1];
w0[2] = tmps[gid].tmp_md5[idx + 2];
w0[3] = tmps[gid].tmp_md5[idx + 3];
w1[0] = tmps[gid].tmp_md5[idx + 4];
w1[1] = tmps[gid].tmp_md5[idx + 5];
w1[2] = tmps[gid].tmp_md5[idx + 6];
w1[3] = tmps[gid].tmp_md5[idx + 7];
w2[0] = tmps[gid].tmp_md5[idx + 8];
w2[1] = tmps[gid].tmp_md5[idx + 9];
w2[2] = tmps[gid].tmp_md5[idx + 10];
w2[3] = tmps[gid].tmp_md5[idx + 11];
w3[0] = tmps[gid].tmp_md5[idx + 12];
w3[1] = tmps[gid].tmp_md5[idx + 13];
w3[2] = tmps[gid].tmp_md5[idx + 14];
w3[3] = tmps[gid].tmp_md5[idx + 15];
md5_transform (w0, w1, w2, w3, h_md5);
// sha1
w0[0] = tmps[gid].tmp_sha1[idx + 0];
w0[1] = tmps[gid].tmp_sha1[idx + 1];
w0[2] = tmps[gid].tmp_sha1[idx + 2];
w0[3] = tmps[gid].tmp_sha1[idx + 3];
w1[0] = tmps[gid].tmp_sha1[idx + 4];
w1[1] = tmps[gid].tmp_sha1[idx + 5];
w1[2] = tmps[gid].tmp_sha1[idx + 6];
w1[3] = tmps[gid].tmp_sha1[idx + 7];
w2[0] = tmps[gid].tmp_sha1[idx + 8];
w2[1] = tmps[gid].tmp_sha1[idx + 9];
w2[2] = tmps[gid].tmp_sha1[idx + 10];
w2[3] = tmps[gid].tmp_sha1[idx + 11];
w3[0] = tmps[gid].tmp_sha1[idx + 12];
w3[1] = tmps[gid].tmp_sha1[idx + 13];
w3[2] = tmps[gid].tmp_sha1[idx + 14];
w3[3] = tmps[gid].tmp_sha1[idx + 15];
sha1_transform (w0, w1, w2, w3, h_sha1);
}
}
tmps[gid].h_md5[0] = h_md5[0];
tmps[gid].h_md5[1] = h_md5[1];
tmps[gid].h_md5[2] = h_md5[2];
tmps[gid].h_md5[3] = h_md5[3];
tmps[gid].h_sha1[0] = h_sha1[0];
tmps[gid].h_sha1[1] = h_sha1[1];
tmps[gid].h_sha1[2] = h_sha1[2];
tmps[gid].h_sha1[3] = h_sha1[3];
tmps[gid].h_sha1[4] = h_sha1[4];
}
KERNEL_FQ void m25000_comp (KERN_ATTR_TMPS_ESALT (hmac_md5_tmp_t, snmpv3_t))
{
/**
* modifier
*/
const u64 gid = get_global_id (0);
if (gid >= GID_CNT) return;
u32 w0[4];
u32 w1[4];
u32 w2[4];
u32 w3[4];
// md5
w0[0] = 0x00000080;
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] = 1048576 * 8;
w3[3] = 0;
u32 h_md5[4];
h_md5[0] = tmps[gid].h_md5[0];
h_md5[1] = tmps[gid].h_md5[1];
h_md5[2] = tmps[gid].h_md5[2];
h_md5[3] = tmps[gid].h_md5[3];
md5_transform (w0, w1, w2, w3, h_md5);
// sha1
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] = 1048576 * 8;
u32 h_sha1[5];
h_sha1[0] = tmps[gid].h_sha1[0];
h_sha1[1] = tmps[gid].h_sha1[1];
h_sha1[2] = tmps[gid].h_sha1[2];
h_sha1[3] = tmps[gid].h_sha1[3];
h_sha1[4] = tmps[gid].h_sha1[4];
sha1_transform (w0, w1, w2, w3, h_sha1);
md5_ctx_t md5_ctx;
sha1_ctx_t sha1_ctx;
md5_init (&md5_ctx);
sha1_init (&sha1_ctx);
u32 w[16];
// md5
w[ 0] = h_md5[0];
w[ 1] = h_md5[1];
w[ 2] = h_md5[2];
w[ 3] = h_md5[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_update (&md5_ctx, w, 16);
// sha1
w[ 0] = h_sha1[0];
w[ 1] = h_sha1[1];
w[ 2] = h_sha1[2];
w[ 3] = h_sha1[3];
w[ 4] = h_sha1[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_update (&sha1_ctx, w, 20);
// engineID
md5_update_global (&md5_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].engineID_buf, esalt_bufs[DIGESTS_OFFSET_HOST].engineID_len);
sha1_update_global_swap (&sha1_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].engineID_buf, esalt_bufs[DIGESTS_OFFSET_HOST].engineID_len);
// md5
w[ 0] = h_md5[0];
w[ 1] = h_md5[1];
w[ 2] = h_md5[2];
w[ 3] = h_md5[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_update (&md5_ctx, w, 16);
// sha1
w[ 0] = h_sha1[0];
w[ 1] = h_sha1[1];
w[ 2] = h_sha1[2];
w[ 3] = h_sha1[3];
w[ 4] = h_sha1[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_update (&sha1_ctx, w, 20);
md5_final (&md5_ctx);
sha1_final (&sha1_ctx);
// md5
w[ 0] = md5_ctx.h[0];
w[ 1] = md5_ctx.h[1];
w[ 2] = md5_ctx.h[2];
w[ 3] = md5_ctx.h[3];
w[ 4] = 0;
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
md5_hmac_ctx_t md5_hmac_ctx;
md5_hmac_init (&md5_hmac_ctx, w, 16);
md5_hmac_update_global (&md5_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf, esalt_bufs[DIGESTS_OFFSET_HOST].salt_len);
md5_hmac_final (&md5_hmac_ctx);
{
const u32 r0 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R0]);
const u32 r1 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R1]);
const u32 r2 = hc_swap32_S (md5_hmac_ctx.opad.h[DGST_R2]);
const u32 r3 = 0;
#define il_pos 0
#ifdef KERNEL_STATIC
#include COMPARE_M
#endif
}
// sha1
w[ 0] = sha1_ctx.h[0];
w[ 1] = sha1_ctx.h[1];
w[ 2] = sha1_ctx.h[2];
w[ 3] = sha1_ctx.h[3];
w[ 4] = sha1_ctx.h[4];
w[ 5] = 0;
w[ 6] = 0;
w[ 7] = 0;
w[ 8] = 0;
w[ 9] = 0;
w[10] = 0;
w[11] = 0;
w[12] = 0;
w[13] = 0;
w[14] = 0;
w[15] = 0;
sha1_hmac_ctx_t sha1_hmac_ctx;
sha1_hmac_init (&sha1_hmac_ctx, w, 20);
sha1_hmac_update_global_swap (&sha1_hmac_ctx, esalt_bufs[DIGESTS_OFFSET_HOST].salt_buf, esalt_bufs[DIGESTS_OFFSET_HOST].salt_len);
sha1_hmac_final (&sha1_hmac_ctx);
{
const u32 r0 = sha1_hmac_ctx.opad.h[DGST_R0];
const u32 r1 = sha1_hmac_ctx.opad.h[DGST_R1];
const u32 r2 = sha1_hmac_ctx.opad.h[DGST_R2];
const u32 r3 = 0;
#define il_pos 0
#ifdef KERNEL_STATIC
#include COMPARE_M
#endif
}
}