github/hashcat
1
Fork 0
mirror of https://github.com/hashcat/hashcat synced 2024-12-12 17:13:55 +01:00
Code Releases Activity
dad03e394d
hashcat/OpenCL/m11100_a0.cl
jsteube dad03e394d Fixed two major problems 
1) SIMD code for all attack-mode

Macro vector_accessible() was not refactored and missing completely.
Had to rename variables rules_cnt, combs_cnt and bfs_cnt into il_cnt which was a good thing anyway as with new SIMD code they all act in the same way.

2) SIMD code for attack-mode 0

With new SIMD code, apply_rules_vect() has to return u32 not u32x.
This has massive impact on all *_a0 kernels.

I've rewritten most of them. Deep testing using test.sh is still required.

Some kernel need more fixes:

- Some are kind of completely incompatible like m10400 but they still use old check_* includes, we should get rid of them as they are no longer neccessary as we have simd.c
- Some have a chance but require additional effort like m11500. We can use commented out "#define NEW_SIMD_CODE" to find them

This change can have negative impact on -a0 performance for device that require vectorization. That is mostly CPU devices. New GPU's are all scalar, so they wont get hurt by this.
This change also proofes that there's no way to efficiently vectorize kernel rules with new SIMD code, but it enables the addition of the rule functions like @ that we were missing for some long time. This is a TODO.
2016-02-27 17:18:54 +01:00

742 lines
30 KiB
Common Lisp
Raw Blame History

/**
* Authors.....: Jens Steube <jens.steube@gmail.com>
* magnum <john.magnum@hushmail.com>
*
* License.....: MIT
*/
#define _MD5_
#define NEW_SIMD_CODE
#include "include/constants.h"
#include "include/kernel_vendor.h"
#define DGST_R0 0
#define DGST_R1 3
#define DGST_R2 2
#define DGST_R3 1
#include "include/kernel_functions.c"
#include "OpenCL/types_ocl.c"
#include "OpenCL/common.c"
#include "include/rp_kernel.h"
#include "OpenCL/rp.c"
#include "OpenCL/simd.c"
#if VECT_SIZE == 1
#define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i)])
#elif VECT_SIZE == 2
#define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i).s0], l_bin2asc[(i).s1])
#elif VECT_SIZE == 4
#define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i).s0], l_bin2asc[(i).s1], l_bin2asc[(i).s2], l_bin2asc[(i).s3])
#elif VECT_SIZE == 8
#define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i).s0], l_bin2asc[(i).s1], l_bin2asc[(i).s2], l_bin2asc[(i).s3], l_bin2asc[(i).s4], l_bin2asc[(i).s5], l_bin2asc[(i).s6], l_bin2asc[(i).s7])
#elif VECT_SIZE == 16
#define uint_to_hex_lower8(i) (u32x) (l_bin2asc[(i).s0], l_bin2asc[(i).s1], l_bin2asc[(i).s2], l_bin2asc[(i).s3], l_bin2asc[(i).s4], l_bin2asc[(i).s5], l_bin2asc[(i).s6], l_bin2asc[(i).s7], l_bin2asc[(i).s8], l_bin2asc[(i).s9], l_bin2asc[(i).sa], l_bin2asc[(i).sb], l_bin2asc[(i).sc], l_bin2asc[(i).sd], l_bin2asc[(i).se], l_bin2asc[(i).sf])
#endif
__kernel void m11100_m04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
/**
* base
*/
const u32 gid = get_global_id (0);
const u32 lid = get_local_id (0);
const u32 lsz = get_local_size (0);
/**
* bin2asc table
*/
__local u32 l_bin2asc[256];
for (u32 i = lid; i < 256; i += lsz)
{
const u32 i0 = (i >> 0) & 15;
const u32 i1 = (i >> 4) & 15;
l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
| ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
}
barrier (CLK_LOCAL_MEM_FENCE);
if (gid >= gid_max) return;
/**
* base
*/
u32 pw_buf0[4];
pw_buf0[0] = pws[gid].i[ 0];
pw_buf0[1] = pws[gid].i[ 1];
pw_buf0[2] = pws[gid].i[ 2];
pw_buf0[3] = pws[gid].i[ 3];
u32 pw_buf1[4];
pw_buf1[0] = pws[gid].i[ 4];
pw_buf1[1] = pws[gid].i[ 5];
pw_buf1[2] = pws[gid].i[ 6];
pw_buf1[3] = pws[gid].i[ 7];
const u32 pw_len = pws[gid].pw_len;
/**
* challenge
*/
u32 challenge;
challenge = salt_bufs[salt_pos].salt_buf[0];
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 1]; // not a bug
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 3];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 4];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 7];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 8];
const u32 salt_len = salt_bufs[salt_pos].salt_len - 4;
/**
* loop
*/
for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
{
u32x w0[4] = { 0 };
u32x w1[4] = { 0 };
u32x w2[4] = { 0 };
u32x w3[4] = { 0 };
const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1);
u32x w0_t[4];
w0_t[0] = salt_buf0[0];
w0_t[1] = salt_buf0[1];
w0_t[2] = salt_buf0[2];
w0_t[3] = salt_buf0[3];
u32x w1_t[4];
w1_t[0] = salt_buf1[0];
w1_t[1] = salt_buf1[1];
w1_t[2] = salt_buf1[2];
w1_t[3] = salt_buf1[3];
u32x w2_t[4];
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
u32x w3_t[4];
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 0;
w3_t[3] = 0;
/*
* append the salt
*/
switch_buffer_by_offset_le_VV (w0_t, w1_t, w2_t, w3_t, out_len);
const u32x pw_salt_len = out_len + salt_len;
w0_t[0] |= w0[0];
w0_t[1] |= w0[1];
w0_t[2] |= w0[2];
w0_t[3] |= w0[3];
w1_t[0] |= w1[0];
w1_t[1] |= w1[1];
w1_t[2] |= w1[2];
w1_t[3] |= w1[3];
w2_t[0] |= w2[0];
w2_t[1] |= w2[1];
w2_t[2] |= w2[2];
w2_t[3] |= w2[3];
w3_t[0] |= w3[0];
w3_t[1] |= w3[1];
w3_t[2] = pw_salt_len * 8;
w3_t[3] = 0;
/*
* md5 ($pass.$salt)
*/
u32x a = MD5M_A;
u32x b = MD5M_B;
u32x c = MD5M_C;
u32x d = MD5M_D;
MD5_STEP (MD5_Fo, a, b, c, d, w0_t[0], MD5C00, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w0_t[1], MD5C01, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w0_t[2], MD5C02, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w0_t[3], MD5C03, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w1_t[0], MD5C04, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w1_t[1], MD5C05, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w1_t[2], MD5C06, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w1_t[3], MD5C07, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w2_t[0], MD5C08, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w2_t[1], MD5C09, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w2_t[2], MD5C0a, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w2_t[3], MD5C0b, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w3_t[0], MD5C0c, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w3_t[1], MD5C0d, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w3_t[2], MD5C0e, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w3_t[3], MD5C0f, MD5S03);
MD5_STEP (MD5_Go, a, b, c, d, w0_t[1], MD5C10, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w1_t[2], MD5C11, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w2_t[3], MD5C12, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w0_t[0], MD5C13, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w1_t[1], MD5C14, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w2_t[2], MD5C15, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w3_t[3], MD5C16, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w1_t[0], MD5C17, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w2_t[1], MD5C18, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w3_t[2], MD5C19, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w0_t[3], MD5C1a, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w2_t[0], MD5C1b, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w3_t[1], MD5C1c, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w0_t[2], MD5C1d, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w1_t[3], MD5C1e, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w3_t[0], MD5C1f, MD5S13);
MD5_STEP (MD5_H , a, b, c, d, w1_t[1], MD5C20, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w2_t[0], MD5C21, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w2_t[3], MD5C22, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w3_t[2], MD5C23, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w0_t[1], MD5C24, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w1_t[0], MD5C25, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w1_t[3], MD5C26, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w2_t[2], MD5C27, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w3_t[1], MD5C28, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w0_t[0], MD5C29, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w0_t[3], MD5C2a, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w1_t[2], MD5C2b, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w2_t[1], MD5C2c, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w3_t[0], MD5C2d, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w3_t[3], MD5C2e, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w0_t[2], MD5C2f, MD5S23);
MD5_STEP (MD5_I , a, b, c, d, w0_t[0], MD5C30, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w1_t[3], MD5C31, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w3_t[2], MD5C32, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w1_t[1], MD5C33, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w3_t[0], MD5C34, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w0_t[3], MD5C35, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w2_t[2], MD5C36, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w0_t[1], MD5C37, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w2_t[0], MD5C38, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w3_t[3], MD5C39, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w1_t[2], MD5C3a, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w3_t[1], MD5C3b, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w1_t[0], MD5C3c, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w2_t[3], MD5C3d, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w0_t[2], MD5C3e, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w2_t[1], MD5C3f, MD5S33);
a += MD5M_A;
b += MD5M_B;
c += MD5M_C;
d += MD5M_D;
w0_t[0] = uint_to_hex_lower8 ((a >> 0) & 255) << 0
| uint_to_hex_lower8 ((a >> 8) & 255) << 16;
w0_t[1] = uint_to_hex_lower8 ((a >> 16) & 255) << 0
| uint_to_hex_lower8 ((a >> 24) & 255) << 16;
w0_t[2] = uint_to_hex_lower8 ((b >> 0) & 255) << 0
| uint_to_hex_lower8 ((b >> 8) & 255) << 16;
w0_t[3] = uint_to_hex_lower8 ((b >> 16) & 255) << 0
| uint_to_hex_lower8 ((b >> 24) & 255) << 16;
w1_t[0] = uint_to_hex_lower8 ((c >> 0) & 255) << 0
| uint_to_hex_lower8 ((c >> 8) & 255) << 16;
w1_t[1] = uint_to_hex_lower8 ((c >> 16) & 255) << 0
| uint_to_hex_lower8 ((c >> 24) & 255) << 16;
w1_t[2] = uint_to_hex_lower8 ((d >> 0) & 255) << 0
| uint_to_hex_lower8 ((d >> 8) & 255) << 16;
w1_t[3] = uint_to_hex_lower8 ((d >> 16) & 255) << 0
| uint_to_hex_lower8 ((d >> 24) & 255) << 16;
// add the 4 byte challenge here
w2_t[0] = challenge;
w2_t[1] = 0x00000080;
w2_t[2] = 0;
w2_t[3] = 0;
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = (32 + 4) * 8;
w3_t[3] = 0;
/**
* md5 ($hash.$challenge)
*/
a = MD5M_A;
b = MD5M_B;
c = MD5M_C;
d = MD5M_D;
MD5_STEP (MD5_Fo, a, b, c, d, w0_t[0], MD5C00, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w0_t[1], MD5C01, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w0_t[2], MD5C02, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w0_t[3], MD5C03, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w1_t[0], MD5C04, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w1_t[1], MD5C05, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w1_t[2], MD5C06, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w1_t[3], MD5C07, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w2_t[0], MD5C08, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w2_t[1], MD5C09, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w2_t[2], MD5C0a, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w2_t[3], MD5C0b, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w3_t[0], MD5C0c, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w3_t[1], MD5C0d, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w3_t[2], MD5C0e, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w3_t[3], MD5C0f, MD5S03);
MD5_STEP (MD5_Go, a, b, c, d, w0_t[1], MD5C10, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w1_t[2], MD5C11, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w2_t[3], MD5C12, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w0_t[0], MD5C13, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w1_t[1], MD5C14, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w2_t[2], MD5C15, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w3_t[3], MD5C16, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w1_t[0], MD5C17, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w2_t[1], MD5C18, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w3_t[2], MD5C19, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w0_t[3], MD5C1a, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w2_t[0], MD5C1b, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w3_t[1], MD5C1c, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w0_t[2], MD5C1d, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w1_t[3], MD5C1e, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w3_t[0], MD5C1f, MD5S13);
MD5_STEP (MD5_H , a, b, c, d, w1_t[1], MD5C20, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w2_t[0], MD5C21, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w2_t[3], MD5C22, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w3_t[2], MD5C23, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w0_t[1], MD5C24, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w1_t[0], MD5C25, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w1_t[3], MD5C26, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w2_t[2], MD5C27, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w3_t[1], MD5C28, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w0_t[0], MD5C29, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w0_t[3], MD5C2a, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w1_t[2], MD5C2b, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w2_t[1], MD5C2c, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w3_t[0], MD5C2d, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w3_t[3], MD5C2e, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w0_t[2], MD5C2f, MD5S23);
MD5_STEP (MD5_I , a, b, c, d, w0_t[0], MD5C30, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w1_t[3], MD5C31, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w3_t[2], MD5C32, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w1_t[1], MD5C33, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w3_t[0], MD5C34, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w0_t[3], MD5C35, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w2_t[2], MD5C36, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w0_t[1], MD5C37, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w2_t[0], MD5C38, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w3_t[3], MD5C39, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w1_t[2], MD5C3a, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w3_t[1], MD5C3b, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w1_t[0], MD5C3c, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w2_t[3], MD5C3d, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w0_t[2], MD5C3e, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w2_t[1], MD5C3f, MD5S33);
COMPARE_M_SIMD (a, d, c, b);
}
}
__kernel void m11100_m08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
}
__kernel void m11100_m16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
}
__kernel void m11100_s04 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
/**
* base
*/
const u32 gid = get_global_id (0);
const u32 lid = get_local_id (0);
const u32 lsz = get_local_size (0);
/**
* bin2asc table
*/
__local u32 l_bin2asc[256];
for (u32 i = lid; i < 256; i += lsz)
{
const u32 i0 = (i >> 0) & 15;
const u32 i1 = (i >> 4) & 15;
l_bin2asc[i] = ((i0 < 10) ? '0' + i0 : 'a' - 10 + i0) << 8
| ((i1 < 10) ? '0' + i1 : 'a' - 10 + i1) << 0;
}
barrier (CLK_LOCAL_MEM_FENCE);
if (gid >= gid_max) return;
/**
* base
*/
u32 pw_buf0[4];
pw_buf0[0] = pws[gid].i[ 0];
pw_buf0[1] = pws[gid].i[ 1];
pw_buf0[2] = pws[gid].i[ 2];
pw_buf0[3] = pws[gid].i[ 3];
u32 pw_buf1[4];
pw_buf1[0] = pws[gid].i[ 4];
pw_buf1[1] = pws[gid].i[ 5];
pw_buf1[2] = pws[gid].i[ 6];
pw_buf1[3] = pws[gid].i[ 7];
const u32 pw_len = pws[gid].pw_len;
/**
* challenge
*/
u32 challenge;
challenge = salt_bufs[salt_pos].salt_buf[0];
/**
* salt
*/
u32 salt_buf0[4];
salt_buf0[0] = salt_bufs[salt_pos].salt_buf[ 1]; // not a bug
salt_buf0[1] = salt_bufs[salt_pos].salt_buf[ 2];
salt_buf0[2] = salt_bufs[salt_pos].salt_buf[ 3];
salt_buf0[3] = salt_bufs[salt_pos].salt_buf[ 4];
u32 salt_buf1[4];
salt_buf1[0] = salt_bufs[salt_pos].salt_buf[ 5];
salt_buf1[1] = salt_bufs[salt_pos].salt_buf[ 6];
salt_buf1[2] = salt_bufs[salt_pos].salt_buf[ 7];
salt_buf1[3] = salt_bufs[salt_pos].salt_buf[ 8];
const u32 salt_len = salt_bufs[salt_pos].salt_len - 4;
/**
* digest
*/
const u32 search[4] =
{
digests_buf[digests_offset].digest_buf[DGST_R0],
digests_buf[digests_offset].digest_buf[DGST_R1],
digests_buf[digests_offset].digest_buf[DGST_R2],
digests_buf[digests_offset].digest_buf[DGST_R3]
};
/**
* loop
*/
for (u32 il_pos = 0; il_pos < il_cnt; il_pos += VECT_SIZE)
{
u32x w0[4] = { 0 };
u32x w1[4] = { 0 };
u32x w2[4] = { 0 };
u32x w3[4] = { 0 };
const u32x out_len = apply_rules_vect (pw_buf0, pw_buf1, pw_len, rules_buf, il_pos, w0, w1);
u32x w0_t[4];
w0_t[0] = salt_buf0[0];
w0_t[1] = salt_buf0[1];
w0_t[2] = salt_buf0[2];
w0_t[3] = salt_buf0[3];
u32x w1_t[4];
w1_t[0] = salt_buf1[0];
w1_t[1] = salt_buf1[1];
w1_t[2] = salt_buf1[2];
w1_t[3] = salt_buf1[3];
u32x w2_t[4];
w2_t[0] = 0;
w2_t[1] = 0;
w2_t[2] = 0;
w2_t[3] = 0;
u32x w3_t[4];
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = 0;
w3_t[3] = 0;
/*
* append the salt
*/
switch_buffer_by_offset_le_VV (w0_t, w1_t, w2_t, w3_t, out_len);
const u32x pw_salt_len = out_len + salt_len;
w0_t[0] |= w0[0];
w0_t[1] |= w0[1];
w0_t[2] |= w0[2];
w0_t[3] |= w0[3];
w1_t[0] |= w1[0];
w1_t[1] |= w1[1];
w1_t[2] |= w1[2];
w1_t[3] |= w1[3];
w2_t[0] |= w2[0];
w2_t[1] |= w2[1];
w2_t[2] |= w2[2];
w2_t[3] |= w2[3];
w3_t[0] |= w3[0];
w3_t[1] |= w3[1];
w3_t[2] = pw_salt_len * 8;
w3_t[3] = 0;
/*
* md5 ($pass.$salt)
*/
u32x a = MD5M_A;
u32x b = MD5M_B;
u32x c = MD5M_C;
u32x d = MD5M_D;
MD5_STEP (MD5_Fo, a, b, c, d, w0_t[0], MD5C00, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w0_t[1], MD5C01, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w0_t[2], MD5C02, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w0_t[3], MD5C03, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w1_t[0], MD5C04, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w1_t[1], MD5C05, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w1_t[2], MD5C06, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w1_t[3], MD5C07, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w2_t[0], MD5C08, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w2_t[1], MD5C09, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w2_t[2], MD5C0a, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w2_t[3], MD5C0b, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w3_t[0], MD5C0c, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w3_t[1], MD5C0d, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w3_t[2], MD5C0e, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w3_t[3], MD5C0f, MD5S03);
MD5_STEP (MD5_Go, a, b, c, d, w0_t[1], MD5C10, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w1_t[2], MD5C11, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w2_t[3], MD5C12, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w0_t[0], MD5C13, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w1_t[1], MD5C14, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w2_t[2], MD5C15, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w3_t[3], MD5C16, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w1_t[0], MD5C17, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w2_t[1], MD5C18, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w3_t[2], MD5C19, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w0_t[3], MD5C1a, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w2_t[0], MD5C1b, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w3_t[1], MD5C1c, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w0_t[2], MD5C1d, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w1_t[3], MD5C1e, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w3_t[0], MD5C1f, MD5S13);
MD5_STEP (MD5_H , a, b, c, d, w1_t[1], MD5C20, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w2_t[0], MD5C21, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w2_t[3], MD5C22, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w3_t[2], MD5C23, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w0_t[1], MD5C24, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w1_t[0], MD5C25, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w1_t[3], MD5C26, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w2_t[2], MD5C27, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w3_t[1], MD5C28, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w0_t[0], MD5C29, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w0_t[3], MD5C2a, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w1_t[2], MD5C2b, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w2_t[1], MD5C2c, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w3_t[0], MD5C2d, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w3_t[3], MD5C2e, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w0_t[2], MD5C2f, MD5S23);
MD5_STEP (MD5_I , a, b, c, d, w0_t[0], MD5C30, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w1_t[3], MD5C31, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w3_t[2], MD5C32, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w1_t[1], MD5C33, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w3_t[0], MD5C34, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w0_t[3], MD5C35, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w2_t[2], MD5C36, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w0_t[1], MD5C37, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w2_t[0], MD5C38, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w3_t[3], MD5C39, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w1_t[2], MD5C3a, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w3_t[1], MD5C3b, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w1_t[0], MD5C3c, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w2_t[3], MD5C3d, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w0_t[2], MD5C3e, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w2_t[1], MD5C3f, MD5S33);
a += MD5M_A;
b += MD5M_B;
c += MD5M_C;
d += MD5M_D;
w0_t[0] = uint_to_hex_lower8 ((a >> 0) & 255) << 0
| uint_to_hex_lower8 ((a >> 8) & 255) << 16;
w0_t[1] = uint_to_hex_lower8 ((a >> 16) & 255) << 0
| uint_to_hex_lower8 ((a >> 24) & 255) << 16;
w0_t[2] = uint_to_hex_lower8 ((b >> 0) & 255) << 0
| uint_to_hex_lower8 ((b >> 8) & 255) << 16;
w0_t[3] = uint_to_hex_lower8 ((b >> 16) & 255) << 0
| uint_to_hex_lower8 ((b >> 24) & 255) << 16;
w1_t[0] = uint_to_hex_lower8 ((c >> 0) & 255) << 0
| uint_to_hex_lower8 ((c >> 8) & 255) << 16;
w1_t[1] = uint_to_hex_lower8 ((c >> 16) & 255) << 0
| uint_to_hex_lower8 ((c >> 24) & 255) << 16;
w1_t[2] = uint_to_hex_lower8 ((d >> 0) & 255) << 0
| uint_to_hex_lower8 ((d >> 8) & 255) << 16;
w1_t[3] = uint_to_hex_lower8 ((d >> 16) & 255) << 0
| uint_to_hex_lower8 ((d >> 24) & 255) << 16;
// add the 4 byte challenge here
w2_t[0] = challenge;
w2_t[1] = 0x00000080;
w2_t[2] = 0;
w2_t[3] = 0;
w3_t[0] = 0;
w3_t[1] = 0;
w3_t[2] = (32 + 4) * 8;
w3_t[3] = 0;
/**
* md5 ($hash.$challenge)
*/
a = MD5M_A;
b = MD5M_B;
c = MD5M_C;
d = MD5M_D;
MD5_STEP (MD5_Fo, a, b, c, d, w0_t[0], MD5C00, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w0_t[1], MD5C01, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w0_t[2], MD5C02, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w0_t[3], MD5C03, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w1_t[0], MD5C04, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w1_t[1], MD5C05, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w1_t[2], MD5C06, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w1_t[3], MD5C07, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w2_t[0], MD5C08, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w2_t[1], MD5C09, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w2_t[2], MD5C0a, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w2_t[3], MD5C0b, MD5S03);
MD5_STEP (MD5_Fo, a, b, c, d, w3_t[0], MD5C0c, MD5S00);
MD5_STEP (MD5_Fo, d, a, b, c, w3_t[1], MD5C0d, MD5S01);
MD5_STEP (MD5_Fo, c, d, a, b, w3_t[2], MD5C0e, MD5S02);
MD5_STEP (MD5_Fo, b, c, d, a, w3_t[3], MD5C0f, MD5S03);
MD5_STEP (MD5_Go, a, b, c, d, w0_t[1], MD5C10, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w1_t[2], MD5C11, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w2_t[3], MD5C12, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w0_t[0], MD5C13, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w1_t[1], MD5C14, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w2_t[2], MD5C15, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w3_t[3], MD5C16, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w1_t[0], MD5C17, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w2_t[1], MD5C18, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w3_t[2], MD5C19, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w0_t[3], MD5C1a, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w2_t[0], MD5C1b, MD5S13);
MD5_STEP (MD5_Go, a, b, c, d, w3_t[1], MD5C1c, MD5S10);
MD5_STEP (MD5_Go, d, a, b, c, w0_t[2], MD5C1d, MD5S11);
MD5_STEP (MD5_Go, c, d, a, b, w1_t[3], MD5C1e, MD5S12);
MD5_STEP (MD5_Go, b, c, d, a, w3_t[0], MD5C1f, MD5S13);
MD5_STEP (MD5_H , a, b, c, d, w1_t[1], MD5C20, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w2_t[0], MD5C21, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w2_t[3], MD5C22, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w3_t[2], MD5C23, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w0_t[1], MD5C24, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w1_t[0], MD5C25, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w1_t[3], MD5C26, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w2_t[2], MD5C27, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w3_t[1], MD5C28, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w0_t[0], MD5C29, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w0_t[3], MD5C2a, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w1_t[2], MD5C2b, MD5S23);
MD5_STEP (MD5_H , a, b, c, d, w2_t[1], MD5C2c, MD5S20);
MD5_STEP (MD5_H , d, a, b, c, w3_t[0], MD5C2d, MD5S21);
MD5_STEP (MD5_H , c, d, a, b, w3_t[3], MD5C2e, MD5S22);
MD5_STEP (MD5_H , b, c, d, a, w0_t[2], MD5C2f, MD5S23);
MD5_STEP (MD5_I , a, b, c, d, w0_t[0], MD5C30, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w1_t[3], MD5C31, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w3_t[2], MD5C32, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w1_t[1], MD5C33, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w3_t[0], MD5C34, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w0_t[3], MD5C35, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w2_t[2], MD5C36, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w0_t[1], MD5C37, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w2_t[0], MD5C38, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w3_t[3], MD5C39, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w1_t[2], MD5C3a, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w3_t[1], MD5C3b, MD5S33);
MD5_STEP (MD5_I , a, b, c, d, w1_t[0], MD5C3c, MD5S30);
MD5_STEP (MD5_I , d, a, b, c, w2_t[3], MD5C3d, MD5S31);
MD5_STEP (MD5_I , c, d, a, b, w0_t[2], MD5C3e, MD5S32);
MD5_STEP (MD5_I , b, c, d, a, w2_t[1], MD5C3f, MD5S33);
COMPARE_S_SIMD (a, d, c, b);
}
}
__kernel void m11100_s08 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
}
__kernel void m11100_s16 (__global pw_t *pws, __global kernel_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global void *tmps, __global void *hooks, __global u32 *bitmaps_buf_s1_a, __global u32 *bitmaps_buf_s1_b, __global u32 *bitmaps_buf_s1_c, __global u32 *bitmaps_buf_s1_d, __global u32 *bitmaps_buf_s2_a, __global u32 *bitmaps_buf_s2_b, __global u32 *bitmaps_buf_s2_c, __global u32 *bitmaps_buf_s2_d, __global plain_t *plains_buf, __global digest_t *digests_buf, __global u32 *hashes_shown, __global salt_t *salt_bufs, __global void *esalt_bufs, __global u32 *d_return_buf, __global u32 *d_scryptV_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)
{
}
View Git Blame Copy Permalink