2 * Author......: Jens Steube <jens.steube@gmail.com>
8 #include "include/constants.h"
9 #include "include/kernel_vendor.h"
16 #include "include/kernel_functions.c"
17 #include "OpenCL/types_ocl.c"
18 #include "OpenCL/common.c"
20 #define COMPARE_S "OpenCL/check_single_comp4.c"
21 #define COMPARE_M "OpenCL/check_multi_comp4.c"
23 #define md5crypt_magic 0x00243124
25 static void md5_transform (const u32 w0[4], const u32 w1[4], const u32 w2[4], const u32 w3[4], u32 digest[4])
51 MD5_STEP (MD5_Fo, a, b, c, d, w0_t, MD5C00, MD5S00);
52 MD5_STEP (MD5_Fo, d, a, b, c, w1_t, MD5C01, MD5S01);
53 MD5_STEP (MD5_Fo, c, d, a, b, w2_t, MD5C02, MD5S02);
54 MD5_STEP (MD5_Fo, b, c, d, a, w3_t, MD5C03, MD5S03);
55 MD5_STEP (MD5_Fo, a, b, c, d, w4_t, MD5C04, MD5S00);
56 MD5_STEP (MD5_Fo, d, a, b, c, w5_t, MD5C05, MD5S01);
57 MD5_STEP (MD5_Fo, c, d, a, b, w6_t, MD5C06, MD5S02);
58 MD5_STEP (MD5_Fo, b, c, d, a, w7_t, MD5C07, MD5S03);
59 MD5_STEP (MD5_Fo, a, b, c, d, w8_t, MD5C08, MD5S00);
60 MD5_STEP (MD5_Fo, d, a, b, c, w9_t, MD5C09, MD5S01);
61 MD5_STEP (MD5_Fo, c, d, a, b, wa_t, MD5C0a, MD5S02);
62 MD5_STEP (MD5_Fo, b, c, d, a, wb_t, MD5C0b, MD5S03);
63 MD5_STEP (MD5_Fo, a, b, c, d, wc_t, MD5C0c, MD5S00);
64 MD5_STEP (MD5_Fo, d, a, b, c, wd_t, MD5C0d, MD5S01);
65 MD5_STEP (MD5_Fo, c, d, a, b, we_t, MD5C0e, MD5S02);
66 MD5_STEP (MD5_Fo, b, c, d, a, wf_t, MD5C0f, MD5S03);
68 MD5_STEP (MD5_Go, a, b, c, d, w1_t, MD5C10, MD5S10);
69 MD5_STEP (MD5_Go, d, a, b, c, w6_t, MD5C11, MD5S11);
70 MD5_STEP (MD5_Go, c, d, a, b, wb_t, MD5C12, MD5S12);
71 MD5_STEP (MD5_Go, b, c, d, a, w0_t, MD5C13, MD5S13);
72 MD5_STEP (MD5_Go, a, b, c, d, w5_t, MD5C14, MD5S10);
73 MD5_STEP (MD5_Go, d, a, b, c, wa_t, MD5C15, MD5S11);
74 MD5_STEP (MD5_Go, c, d, a, b, wf_t, MD5C16, MD5S12);
75 MD5_STEP (MD5_Go, b, c, d, a, w4_t, MD5C17, MD5S13);
76 MD5_STEP (MD5_Go, a, b, c, d, w9_t, MD5C18, MD5S10);
77 MD5_STEP (MD5_Go, d, a, b, c, we_t, MD5C19, MD5S11);
78 MD5_STEP (MD5_Go, c, d, a, b, w3_t, MD5C1a, MD5S12);
79 MD5_STEP (MD5_Go, b, c, d, a, w8_t, MD5C1b, MD5S13);
80 MD5_STEP (MD5_Go, a, b, c, d, wd_t, MD5C1c, MD5S10);
81 MD5_STEP (MD5_Go, d, a, b, c, w2_t, MD5C1d, MD5S11);
82 MD5_STEP (MD5_Go, c, d, a, b, w7_t, MD5C1e, MD5S12);
83 MD5_STEP (MD5_Go, b, c, d, a, wc_t, MD5C1f, MD5S13);
85 MD5_STEP (MD5_H1, a, b, c, d, w5_t, MD5C20, MD5S20);
86 MD5_STEP (MD5_H2, d, a, b, c, w8_t, MD5C21, MD5S21);
87 MD5_STEP (MD5_H1, c, d, a, b, wb_t, MD5C22, MD5S22);
88 MD5_STEP (MD5_H2, b, c, d, a, we_t, MD5C23, MD5S23);
89 MD5_STEP (MD5_H1, a, b, c, d, w1_t, MD5C24, MD5S20);
90 MD5_STEP (MD5_H2, d, a, b, c, w4_t, MD5C25, MD5S21);
91 MD5_STEP (MD5_H1, c, d, a, b, w7_t, MD5C26, MD5S22);
92 MD5_STEP (MD5_H2, b, c, d, a, wa_t, MD5C27, MD5S23);
93 MD5_STEP (MD5_H1, a, b, c, d, wd_t, MD5C28, MD5S20);
94 MD5_STEP (MD5_H2, d, a, b, c, w0_t, MD5C29, MD5S21);
95 MD5_STEP (MD5_H1, c, d, a, b, w3_t, MD5C2a, MD5S22);
96 MD5_STEP (MD5_H2, b, c, d, a, w6_t, MD5C2b, MD5S23);
97 MD5_STEP (MD5_H1, a, b, c, d, w9_t, MD5C2c, MD5S20);
98 MD5_STEP (MD5_H2, d, a, b, c, wc_t, MD5C2d, MD5S21);
99 MD5_STEP (MD5_H1, c, d, a, b, wf_t, MD5C2e, MD5S22);
100 MD5_STEP (MD5_H2, b, c, d, a, w2_t, MD5C2f, MD5S23);
102 MD5_STEP (MD5_I , a, b, c, d, w0_t, MD5C30, MD5S30);
103 MD5_STEP (MD5_I , d, a, b, c, w7_t, MD5C31, MD5S31);
104 MD5_STEP (MD5_I , c, d, a, b, we_t, MD5C32, MD5S32);
105 MD5_STEP (MD5_I , b, c, d, a, w5_t, MD5C33, MD5S33);
106 MD5_STEP (MD5_I , a, b, c, d, wc_t, MD5C34, MD5S30);
107 MD5_STEP (MD5_I , d, a, b, c, w3_t, MD5C35, MD5S31);
108 MD5_STEP (MD5_I , c, d, a, b, wa_t, MD5C36, MD5S32);
109 MD5_STEP (MD5_I , b, c, d, a, w1_t, MD5C37, MD5S33);
110 MD5_STEP (MD5_I , a, b, c, d, w8_t, MD5C38, MD5S30);
111 MD5_STEP (MD5_I , d, a, b, c, wf_t, MD5C39, MD5S31);
112 MD5_STEP (MD5_I , c, d, a, b, w6_t, MD5C3a, MD5S32);
113 MD5_STEP (MD5_I , b, c, d, a, wd_t, MD5C3b, MD5S33);
114 MD5_STEP (MD5_I , a, b, c, d, w4_t, MD5C3c, MD5S30);
115 MD5_STEP (MD5_I , d, a, b, c, wb_t, MD5C3d, MD5S31);
116 MD5_STEP (MD5_I , c, d, a, b, w2_t, MD5C3e, MD5S32);
117 MD5_STEP (MD5_I , b, c, d, a, w9_t, MD5C3f, MD5S33);
125 static void memcat16 (u32 block0[4], u32 block1[4], u32 block2[4], u32 block3[4], const u32 block_len, const u32 append[4])
135 const int offset_minus_4 = 4 - (block_len & 3);
137 tmp0 = amd_bytealign (append[0], 0, offset_minus_4);
138 tmp1 = amd_bytealign (append[1], append[0], offset_minus_4);
139 tmp2 = amd_bytealign (append[2], append[1], offset_minus_4);
140 tmp3 = amd_bytealign (append[3], append[2], offset_minus_4);
141 tmp4 = amd_bytealign ( 0, append[3], offset_minus_4);
143 const u32 mod = block_len & 3;
158 const int offset_minus_4 = 4 - (block_len & 3);
160 const int selector = (0x76543210 >> (offset_minus_4 * 4)) & 0xffff;
162 tmp0 = __byte_perm ( 0, append[0], selector);
163 tmp1 = __byte_perm (append[0], append[1], selector);
164 tmp2 = __byte_perm (append[1], append[2], selector);
165 tmp3 = __byte_perm (append[2], append[3], selector);
166 tmp4 = __byte_perm (append[3], 0, selector);
170 const u32 div = block_len / 4;
174 case 0: block0[0] |= tmp0;
180 case 1: block0[1] |= tmp0;
186 case 2: block0[2] |= tmp0;
192 case 3: block0[3] |= tmp0;
198 case 4: block1[0] |= tmp0;
204 case 5: block1[1] |= tmp0;
210 case 6: block1[2] |= tmp0;
216 case 7: block1[3] |= tmp0;
222 case 8: block2[0] |= tmp0;
228 case 9: block2[1] |= tmp0;
237 static void memcat16_x80 (u32 block0[4], u32 block1[4], u32 block2[4], u32 block3[4], const u32 block_len, const u32 append[4])
247 const int offset_minus_4 = 4 - (block_len & 3);
249 tmp0 = amd_bytealign (append[0], 0, offset_minus_4);
250 tmp1 = amd_bytealign (append[1], append[0], offset_minus_4);
251 tmp2 = amd_bytealign (append[2], append[1], offset_minus_4);
252 tmp3 = amd_bytealign (append[3], append[2], offset_minus_4);
253 tmp4 = amd_bytealign ( 0x80, append[3], offset_minus_4);
255 const u32 mod = block_len & 3;
270 const int offset_minus_4 = 4 - (block_len & 3);
272 const int selector = (0x76543210 >> (offset_minus_4 * 4)) & 0xffff;
274 tmp0 = __byte_perm ( 0, append[0], selector);
275 tmp1 = __byte_perm (append[0], append[1], selector);
276 tmp2 = __byte_perm (append[1], append[2], selector);
277 tmp3 = __byte_perm (append[2], append[3], selector);
278 tmp4 = __byte_perm (append[3], 0x80, selector);
282 const u32 div = block_len / 4;
286 case 0: block0[0] |= tmp0;
292 case 1: block0[1] |= tmp0;
298 case 2: block0[2] |= tmp0;
304 case 3: block0[3] |= tmp0;
310 case 4: block1[0] |= tmp0;
316 case 5: block1[1] |= tmp0;
322 case 6: block1[2] |= tmp0;
328 case 7: block1[3] |= tmp0;
334 case 8: block2[0] |= tmp0;
340 case 9: block2[1] |= tmp0;
349 static void memcat8 (u32 block0[4], u32 block1[4], u32 block2[4], u32 block3[4], const u32 block_len, const u32 append[2])
357 const int offset_minus_4 = 4 - (block_len & 3);
359 tmp0 = amd_bytealign (append[0], 0, offset_minus_4);
360 tmp1 = amd_bytealign (append[1], append[0], offset_minus_4);
361 tmp2 = amd_bytealign ( 0, append[1], offset_minus_4);
363 const u32 mod = block_len & 3;
376 const int offset_minus_4 = 4 - (block_len & 3);
378 const int selector = (0x76543210 >> (offset_minus_4 * 4)) & 0xffff;
380 tmp0 = __byte_perm ( 0, append[0], selector);
381 tmp1 = __byte_perm (append[0], append[1], selector);
382 tmp2 = __byte_perm (append[1], 0, selector);
386 const u32 div = block_len / 4;
390 case 0: block0[0] |= tmp0;
394 case 1: block0[1] |= tmp0;
398 case 2: block0[2] |= tmp0;
402 case 3: block0[3] |= tmp0;
406 case 4: block1[0] |= tmp0;
410 case 5: block1[1] |= tmp0;
414 case 6: block1[2] |= tmp0;
418 case 7: block1[3] |= tmp0;
422 case 8: block2[0] |= tmp0;
426 case 9: block2[1] |= tmp0;
430 case 10: block2[2] |= tmp0;
434 case 11: block2[3] |= tmp0;
441 static void append_sign (u32 block0[4], u32 block1[4], const u32 block_len)
446 block0[0] = md5crypt_magic;
450 block0[0] = block0[0] | md5crypt_magic << 8;
451 block0[1] = md5crypt_magic >> 24;
455 block0[0] = block0[0] | md5crypt_magic << 16;
456 block0[1] = md5crypt_magic >> 16;
460 block0[0] = block0[0] | md5crypt_magic << 24;
461 block0[1] = md5crypt_magic >> 8;
465 block0[1] = md5crypt_magic;
469 block0[1] = block0[1] | md5crypt_magic << 8;
470 block0[2] = md5crypt_magic >> 24;
474 block0[1] = block0[1] | md5crypt_magic << 16;
475 block0[2] = md5crypt_magic >> 16;
479 block0[1] = block0[1] | md5crypt_magic << 24;
480 block0[2] = md5crypt_magic >> 8;
484 block0[2] = md5crypt_magic;
488 block0[2] = block0[2] | md5crypt_magic << 8;
489 block0[3] = md5crypt_magic >> 24;
493 block0[2] = block0[2] | md5crypt_magic << 16;
494 block0[3] = md5crypt_magic >> 16;
498 block0[2] = block0[2] | md5crypt_magic << 24;
499 block0[3] = md5crypt_magic >> 8;
503 block0[3] = md5crypt_magic;
507 block0[3] = block0[3] | md5crypt_magic << 8;
508 block1[0] = md5crypt_magic >> 24;
512 block0[3] = block0[3] | md5crypt_magic << 16;
513 block1[0] = md5crypt_magic >> 16;
517 block0[3] = block0[3] | md5crypt_magic << 24;
518 block1[0] = md5crypt_magic >> 8;
523 static void append_1st (u32 block0[4], u32 block1[4], u32 block2[4], u32 block3[4], const u32 block_len, const u32 append)
532 block0[0] = block0[0] | append << 8;
536 block0[0] = block0[0] | append << 16;
540 block0[0] = block0[0] | append << 24;
548 block0[1] = block0[1] | append << 8;
552 block0[1] = block0[1] | append << 16;
556 block0[1] = block0[1] | append << 24;
564 block0[2] = block0[2] | append << 8;
568 block0[2] = block0[2] | append << 16;
572 block0[2] = block0[2] | append << 24;
580 block0[3] = block0[3] | append << 8;
584 block0[3] = block0[3] | append << 16;
588 block0[3] = block0[3] | append << 24;
596 block1[0] = block1[0] | append << 8;
600 block1[0] = block1[0] | append << 16;
604 block1[0] = block1[0] | append << 24;
612 block1[1] = block1[1] | append << 8;
616 block1[1] = block1[1] | append << 16;
620 block1[1] = block1[1] | append << 24;
628 block1[2] = block1[2] | append << 8;
632 block1[2] = block1[2] | append << 16;
636 block1[2] = block1[2] | append << 24;
644 block1[3] = block1[3] | append << 8;
648 block1[3] = block1[3] | append << 16;
652 block1[3] = block1[3] | append << 24;
660 block2[0] = block2[0] | append << 8;
664 block2[0] = block2[0] | append << 16;
668 block2[0] = block2[0] | append << 24;
676 block2[1] = block2[1] | append << 8;
680 block2[1] = block2[1] | append << 16;
684 block2[1] = block2[1] | append << 24;
692 block2[2] = block2[2] | append << 8;
696 block2[2] = block2[2] | append << 16;
700 block2[2] = block2[2] | append << 24;
708 block2[3] = block2[3] | append << 8;
712 block2[3] = block2[3] | append << 16;
716 block2[3] = block2[3] | append << 24;
724 block3[0] = block3[0] | append << 8;
728 block3[0] = block3[0] | append << 16;
732 block3[0] = block3[0] | append << 24;
740 block3[1] = block3[1] | append << 8;
744 block3[1] = block3[1] | append << 16;
748 block3[1] = block3[1] | append << 24;
757 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m00500_init (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global md5crypt_tmp_t *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 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
763 const u32 gid = get_global_id (0);
765 if (gid >= gid_max) return;
769 w0[0] = pws[gid].i[0];
770 w0[1] = pws[gid].i[1];
771 w0[2] = pws[gid].i[2];
772 w0[3] = pws[gid].i[3];
774 const u32 pw_len = pws[gid].pw_len;
782 salt_buf[0] = salt_bufs[salt_pos].salt_buf[0];
783 salt_buf[1] = salt_bufs[salt_pos].salt_buf[1];
785 const u32 salt_len = salt_bufs[salt_pos].salt_len;
791 //memcat16 (block0, block1, block2, block3, block_len, w0);
792 //block_len += pw_len;
794 u32 block_len = pw_len;
824 memcat8 (block0, block1, block2, block3, block_len, salt_buf);
826 block_len += salt_len;
828 memcat16 (block0, block1, block2, block3, block_len, w0);
832 append_0x80_4x4 (block0, block1, block2, block3, block_len);
834 block3[2] = block_len * 8;
843 md5_transform (block0, block1, block2, block3, digest);
845 /* The password first, since that is what is most unknown */
846 /* Then our magic string */
847 /* Then the raw salt */
848 /* Then just as many characters of the MD5(pw,salt,pw) */
850 //memcat16 (block0, block1, block2, block3, block_len, w);
851 //block_len += pw_len;
875 append_sign (block0, block1, block_len);
879 memcat8 (block0, block1, block2, block3, block_len, salt_buf);
881 block_len += salt_len;
883 truncate_block (digest, pw_len);
885 memcat16 (block0, block1, block2, block3, block_len, digest);
889 /* Then something really weird... */
891 u32 append = block0[0] & 0xFF;
893 for (u32 j = pw_len; j; j >>= 1)
897 append_1st (block0, block1, block2, block3, block_len, append);
903 append_0x80_4x4 (block0, block1, block2, block3, block_len);
905 block3[2] = block_len * 8;
912 md5_transform (block0, block1, block2, block3, digest);
914 tmps[gid].digest_buf[0] = digest[0];
915 tmps[gid].digest_buf[1] = digest[1];
916 tmps[gid].digest_buf[2] = digest[2];
917 tmps[gid].digest_buf[3] = digest[3];
920 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m00500_loop (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global md5crypt_tmp_t *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 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
926 const u32 gid = get_global_id (0);
928 if (gid >= gid_max) return;
932 w0[0] = pws[gid].i[0];
933 w0[1] = pws[gid].i[1];
934 w0[2] = pws[gid].i[2];
935 w0[3] = pws[gid].i[3];
937 const u32 pw_len = pws[gid].pw_len;
946 append_0x80_1x4 (w0_x80, pw_len);
954 salt_buf[0] = salt_bufs[salt_pos].salt_buf[0];
955 salt_buf[1] = salt_bufs[salt_pos].salt_buf[1];
957 const u32 salt_len = salt_bufs[salt_pos].salt_len;
965 digest[0] = tmps[gid].digest_buf[0];
966 digest[1] = tmps[gid].digest_buf[1];
967 digest[2] = tmps[gid].digest_buf[2];
968 digest[3] = tmps[gid].digest_buf[3];
974 /* and now, just to make sure things don't run too fast */
1006 for (u32 i = 0, j = loop_pos; i < loop_cnt; i++, j++)
1019 const u32 j1 = (j & 1) ? 1 : 0;
1020 const u32 j3 = (j % 3) ? 1 : 0;
1021 const u32 j7 = (j % 7) ? 1 : 0;
1034 memcat8 (block0, block1, block2, block3, block_len, salt_buf);
1036 block_len += salt_len;
1041 memcat16 (block0, block1, block2, block3, block_len, w0);
1043 block_len += pw_len;
1046 memcat16_x80 (block0, block1, block2, block3, block_len, digest);
1052 block0[0] = digest[0];
1053 block0[1] = digest[1];
1054 block0[2] = digest[2];
1055 block0[3] = digest[3];
1061 block1[0] = salt_buf[0];
1062 block1[1] = salt_buf[1];
1064 block_len += salt_len;
1066 memcat16 (block0, block1, block2, block3, block_len, w0);
1068 block_len += pw_len;
1072 block1[0] = salt_buf[0];
1073 block1[1] = salt_buf[1];
1075 block_len += salt_len;
1084 block_len += pw_len;
1087 memcat16 (block0, block1, block2, block3, block_len, w0_x80);
1089 block_len += pw_len;
1092 block3[2] = block_len * 8;
1099 md5_transform (block0, block1, block2, block3, digest);
1102 tmps[gid].digest_buf[0] = digest[0];
1103 tmps[gid].digest_buf[1] = digest[1];
1104 tmps[gid].digest_buf[2] = digest[2];
1105 tmps[gid].digest_buf[3] = digest[3];
1108 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m00500_comp (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global md5crypt_tmp_t *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 rules_cnt, const u32 digests_cnt, const u32 digests_offset, const u32 combs_mode, const u32 gid_max)
1114 const u32 gid = get_global_id (0);
1116 if (gid >= gid_max) return;
1118 const u32 lid = get_local_id (0);
1124 const u32 r0 = tmps[gid].digest_buf[DGST_R0];
1125 const u32 r1 = tmps[gid].digest_buf[DGST_R1];
1126 const u32 r2 = tmps[gid].digest_buf[DGST_R2];
1127 const u32 r3 = tmps[gid].digest_buf[DGST_R3];