2 * Author......: Jens Steube <jens.steube@gmail.com>
8 #include "include/constants.h"
9 #include "include/kernel_vendor.h"
28 #include "include/kernel_functions.c"
29 #include "types_amd.c"
30 #include "common_amd.c"
33 #define VECT_COMPARE_M "check_multi_vect1_comp4.c"
36 #define PUTCHAR64_BE(a,p,c) ((u8 *)(a))[(p) ^ 7] = (u8) (c)
37 #define GETCHAR64_BE(a,p) ((u8 *)(a))[(p) ^ 7]
47 __constant u64 k_sha512[80] =
49 SHA512C00, SHA512C01, SHA512C02, SHA512C03,
50 SHA512C04, SHA512C05, SHA512C06, SHA512C07,
51 SHA512C08, SHA512C09, SHA512C0a, SHA512C0b,
52 SHA512C0c, SHA512C0d, SHA512C0e, SHA512C0f,
53 SHA512C10, SHA512C11, SHA512C12, SHA512C13,
54 SHA512C14, SHA512C15, SHA512C16, SHA512C17,
55 SHA512C18, SHA512C19, SHA512C1a, SHA512C1b,
56 SHA512C1c, SHA512C1d, SHA512C1e, SHA512C1f,
57 SHA512C20, SHA512C21, SHA512C22, SHA512C23,
58 SHA512C24, SHA512C25, SHA512C26, SHA512C27,
59 SHA512C28, SHA512C29, SHA512C2a, SHA512C2b,
60 SHA512C2c, SHA512C2d, SHA512C2e, SHA512C2f,
61 SHA512C30, SHA512C31, SHA512C32, SHA512C33,
62 SHA512C34, SHA512C35, SHA512C36, SHA512C37,
63 SHA512C38, SHA512C39, SHA512C3a, SHA512C3b,
64 SHA512C3c, SHA512C3d, SHA512C3e, SHA512C3f,
65 SHA512C40, SHA512C41, SHA512C42, SHA512C43,
66 SHA512C44, SHA512C45, SHA512C46, SHA512C47,
67 SHA512C48, SHA512C49, SHA512C4a, SHA512C4b,
68 SHA512C4c, SHA512C4d, SHA512C4e, SHA512C4f,
71 static void sha512_transform (const u64x w[16], u64x digest[8])
99 #define ROUND_EXPAND() \
101 w0_t = SHA512_EXPAND (we_t, w9_t, w1_t, w0_t); \
102 w1_t = SHA512_EXPAND (wf_t, wa_t, w2_t, w1_t); \
103 w2_t = SHA512_EXPAND (w0_t, wb_t, w3_t, w2_t); \
104 w3_t = SHA512_EXPAND (w1_t, wc_t, w4_t, w3_t); \
105 w4_t = SHA512_EXPAND (w2_t, wd_t, w5_t, w4_t); \
106 w5_t = SHA512_EXPAND (w3_t, we_t, w6_t, w5_t); \
107 w6_t = SHA512_EXPAND (w4_t, wf_t, w7_t, w6_t); \
108 w7_t = SHA512_EXPAND (w5_t, w0_t, w8_t, w7_t); \
109 w8_t = SHA512_EXPAND (w6_t, w1_t, w9_t, w8_t); \
110 w9_t = SHA512_EXPAND (w7_t, w2_t, wa_t, w9_t); \
111 wa_t = SHA512_EXPAND (w8_t, w3_t, wb_t, wa_t); \
112 wb_t = SHA512_EXPAND (w9_t, w4_t, wc_t, wb_t); \
113 wc_t = SHA512_EXPAND (wa_t, w5_t, wd_t, wc_t); \
114 wd_t = SHA512_EXPAND (wb_t, w6_t, we_t, wd_t); \
115 we_t = SHA512_EXPAND (wc_t, w7_t, wf_t, we_t); \
116 wf_t = SHA512_EXPAND (wd_t, w8_t, w0_t, wf_t); \
119 #define ROUND_STEP(i) \
121 SHA512_STEP (SHA512_F0o, SHA512_F1o, a, b, c, d, e, f, g, h, w0_t, k_sha512[i + 0]); \
122 SHA512_STEP (SHA512_F0o, SHA512_F1o, h, a, b, c, d, e, f, g, w1_t, k_sha512[i + 1]); \
123 SHA512_STEP (SHA512_F0o, SHA512_F1o, g, h, a, b, c, d, e, f, w2_t, k_sha512[i + 2]); \
124 SHA512_STEP (SHA512_F0o, SHA512_F1o, f, g, h, a, b, c, d, e, w3_t, k_sha512[i + 3]); \
125 SHA512_STEP (SHA512_F0o, SHA512_F1o, e, f, g, h, a, b, c, d, w4_t, k_sha512[i + 4]); \
126 SHA512_STEP (SHA512_F0o, SHA512_F1o, d, e, f, g, h, a, b, c, w5_t, k_sha512[i + 5]); \
127 SHA512_STEP (SHA512_F0o, SHA512_F1o, c, d, e, f, g, h, a, b, w6_t, k_sha512[i + 6]); \
128 SHA512_STEP (SHA512_F0o, SHA512_F1o, b, c, d, e, f, g, h, a, w7_t, k_sha512[i + 7]); \
129 SHA512_STEP (SHA512_F0o, SHA512_F1o, a, b, c, d, e, f, g, h, w8_t, k_sha512[i + 8]); \
130 SHA512_STEP (SHA512_F0o, SHA512_F1o, h, a, b, c, d, e, f, g, w9_t, k_sha512[i + 9]); \
131 SHA512_STEP (SHA512_F0o, SHA512_F1o, g, h, a, b, c, d, e, f, wa_t, k_sha512[i + 10]); \
132 SHA512_STEP (SHA512_F0o, SHA512_F1o, f, g, h, a, b, c, d, e, wb_t, k_sha512[i + 11]); \
133 SHA512_STEP (SHA512_F0o, SHA512_F1o, e, f, g, h, a, b, c, d, wc_t, k_sha512[i + 12]); \
134 SHA512_STEP (SHA512_F0o, SHA512_F1o, d, e, f, g, h, a, b, c, wd_t, k_sha512[i + 13]); \
135 SHA512_STEP (SHA512_F0o, SHA512_F1o, c, d, e, f, g, h, a, b, we_t, k_sha512[i + 14]); \
136 SHA512_STEP (SHA512_F0o, SHA512_F1o, b, c, d, e, f, g, h, a, wf_t, k_sha512[i + 15]); \
142 for (int i = 16; i < 80; i += 16)
144 ROUND_EXPAND (); ROUND_STEP (i);
157 static void sha512_transform_workaround (const u64x w[16], u64x digest[8])
185 #define ROUND_EXPAND_WO() \
187 w0_t = SHA512_EXPAND_WO (we_t, w9_t, w1_t, w0_t); \
188 w1_t = SHA512_EXPAND_WO (wf_t, wa_t, w2_t, w1_t); \
189 w2_t = SHA512_EXPAND_WO (w0_t, wb_t, w3_t, w2_t); \
190 w3_t = SHA512_EXPAND_WO (w1_t, wc_t, w4_t, w3_t); \
191 w4_t = SHA512_EXPAND_WO (w2_t, wd_t, w5_t, w4_t); \
192 w5_t = SHA512_EXPAND_WO (w3_t, we_t, w6_t, w5_t); \
193 w6_t = SHA512_EXPAND_WO (w4_t, wf_t, w7_t, w6_t); \
194 w7_t = SHA512_EXPAND_WO (w5_t, w0_t, w8_t, w7_t); \
195 w8_t = SHA512_EXPAND_WO (w6_t, w1_t, w9_t, w8_t); \
196 w9_t = SHA512_EXPAND_WO (w7_t, w2_t, wa_t, w9_t); \
197 wa_t = SHA512_EXPAND_WO (w8_t, w3_t, wb_t, wa_t); \
198 wb_t = SHA512_EXPAND_WO (w9_t, w4_t, wc_t, wb_t); \
199 wc_t = SHA512_EXPAND_WO (wa_t, w5_t, wd_t, wc_t); \
200 wd_t = SHA512_EXPAND_WO (wb_t, w6_t, we_t, wd_t); \
201 we_t = SHA512_EXPAND_WO (wc_t, w7_t, wf_t, we_t); \
202 wf_t = SHA512_EXPAND_WO (wd_t, w8_t, w0_t, wf_t); \
207 for (int i = 16; i < 80; i += 16)
209 ROUND_EXPAND_WO (); ROUND_STEP (i);
222 static void sha512_init (sha512_ctx_t *sha512_ctx)
224 sha512_ctx->state[0] = SHA512M_A;
225 sha512_ctx->state[1] = SHA512M_B;
226 sha512_ctx->state[2] = SHA512M_C;
227 sha512_ctx->state[3] = SHA512M_D;
228 sha512_ctx->state[4] = SHA512M_E;
229 sha512_ctx->state[5] = SHA512M_F;
230 sha512_ctx->state[6] = SHA512M_G;
231 sha512_ctx->state[7] = SHA512M_H;
236 static void sha512_update (sha512_ctx_t *sha512_ctx, const u64x *buf, int len)
238 int pos = sha512_ctx->len & 0x7f;
240 sha512_ctx->len += len;
242 if ((pos + len) < 128)
244 for (int i = 0; i < len; i++)
246 PUTCHAR64_BE (sha512_ctx->buf, pos++, GETCHAR64_BE (buf, i));
254 for (int i = 0; i < cnt; i++)
256 PUTCHAR64_BE (sha512_ctx->buf, pos++, GETCHAR64_BE (buf, i));
259 sha512_transform_workaround (sha512_ctx->buf, sha512_ctx->state);
263 for (int i = 0; i < len; i++)
265 PUTCHAR64_BE (sha512_ctx->buf, i, GETCHAR64_BE (buf, cnt + i));
269 static void sha512_final (sha512_ctx_t *sha512_ctx)
271 int pos = sha512_ctx->len & 0x7f;
273 for (int i = pos; i < 128; i++)
275 PUTCHAR64_BE (sha512_ctx->buf, i, 0);
278 PUTCHAR64_BE (sha512_ctx->buf, pos, 0x80);
282 sha512_transform (sha512_ctx->buf, sha512_ctx->state);
284 sha512_ctx->buf[ 0] = 0;
285 sha512_ctx->buf[ 1] = 0;
286 sha512_ctx->buf[ 2] = 0;
287 sha512_ctx->buf[ 3] = 0;
288 sha512_ctx->buf[ 4] = 0;
289 sha512_ctx->buf[ 5] = 0;
290 sha512_ctx->buf[ 6] = 0;
291 sha512_ctx->buf[ 7] = 0;
292 sha512_ctx->buf[ 8] = 0;
293 sha512_ctx->buf[ 9] = 0;
294 sha512_ctx->buf[10] = 0;
295 sha512_ctx->buf[11] = 0;
296 sha512_ctx->buf[12] = 0;
297 sha512_ctx->buf[13] = 0;
298 sha512_ctx->buf[14] = 0;
299 sha512_ctx->buf[15] = 0;
302 sha512_ctx->buf[15] = sha512_ctx->len * 8;
304 sha512_transform (sha512_ctx->buf, sha512_ctx->state);
307 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m01800_init (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global sha512crypt_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)
313 const u32 gid = get_global_id (0);
315 if (gid >= gid_max) return;
319 w0[0] = pws[gid].i[0];
320 w0[1] = pws[gid].i[1];
321 w0[2] = pws[gid].i[2];
322 w0[3] = pws[gid].i[3];
324 const u32 pw_len = pws[gid].pw_len;
332 salt_buf[0] = salt_bufs[salt_pos].salt_buf[0];
333 salt_buf[1] = salt_bufs[salt_pos].salt_buf[1];
334 salt_buf[2] = salt_bufs[salt_pos].salt_buf[2];
335 salt_buf[3] = salt_bufs[salt_pos].salt_buf[3];
337 u32 salt_len = salt_bufs[salt_pos].salt_len;
345 pw[0] = swap_workaround (hl32_to_64 (w0[1], w0[0]));
346 pw[1] = swap_workaround (hl32_to_64 (w0[3], w0[2]));
350 salt[0] = swap_workaround (hl32_to_64 (salt_buf[1], salt_buf[0]));
351 salt[1] = swap_workaround (hl32_to_64 (salt_buf[3], salt_buf[2]));
357 sha512_ctx_t sha512_ctx;
359 sha512_init (&sha512_ctx);
361 sha512_update (&sha512_ctx, pw, pw_len);
362 sha512_update (&sha512_ctx, salt, salt_len);
363 sha512_update (&sha512_ctx, pw, pw_len);
365 sha512_final (&sha512_ctx);
369 tmp[0] = sha512_ctx.state[0];
370 tmp[1] = sha512_ctx.state[1];
371 tmp[2] = sha512_ctx.state[2];
372 tmp[3] = sha512_ctx.state[3];
373 tmp[4] = sha512_ctx.state[4];
374 tmp[5] = sha512_ctx.state[5];
375 tmp[6] = sha512_ctx.state[6];
376 tmp[7] = sha512_ctx.state[7];
378 sha512_init (&sha512_ctx);
380 sha512_update (&sha512_ctx, pw, pw_len);
381 sha512_update (&sha512_ctx, salt, salt_len);
382 sha512_update (&sha512_ctx, tmp, pw_len);
384 for (u32 j = pw_len; j; j >>= 1)
388 sha512_update (&sha512_ctx, tmp, 64);
392 sha512_update (&sha512_ctx, pw, pw_len);
396 sha512_final (&sha512_ctx);
398 tmps[gid].l_alt_result[0] = sha512_ctx.state[0];
399 tmps[gid].l_alt_result[1] = sha512_ctx.state[1];
400 tmps[gid].l_alt_result[2] = sha512_ctx.state[2];
401 tmps[gid].l_alt_result[3] = sha512_ctx.state[3];
402 tmps[gid].l_alt_result[4] = sha512_ctx.state[4];
403 tmps[gid].l_alt_result[5] = sha512_ctx.state[5];
404 tmps[gid].l_alt_result[6] = sha512_ctx.state[6];
405 tmps[gid].l_alt_result[7] = sha512_ctx.state[7];
409 sha512_init (&sha512_ctx);
411 for (u32 j = 0; j < pw_len; j++)
413 sha512_update (&sha512_ctx, pw, pw_len);
416 sha512_final (&sha512_ctx);
418 tmps[gid].l_p_bytes[0] = sha512_ctx.state[0];
419 tmps[gid].l_p_bytes[1] = sha512_ctx.state[1];
423 sha512_init (&sha512_ctx);
425 for (u32 j = 0; j < 16 + ((tmps[gid].l_alt_result[0] >> 56) & 0xff); j++)
427 sha512_update (&sha512_ctx, salt, salt_len);
430 sha512_final (&sha512_ctx);
432 tmps[gid].l_s_bytes[0] = sha512_ctx.state[0];
433 tmps[gid].l_s_bytes[1] = sha512_ctx.state[1];
436 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m01800_loop (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global sha512crypt_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)
442 const u32 gid = get_global_id (0);
444 if (gid >= gid_max) return;
448 l_p_bytes0[0] = tmps[gid].l_p_bytes[0];
449 l_p_bytes0[1] = tmps[gid].l_p_bytes[1];
451 const u32 pw_len = pws[gid].pw_len;
455 l_s_bytes0[0] = tmps[gid].l_s_bytes[0];
456 l_s_bytes0[1] = tmps[gid].l_s_bytes[1];
458 const u32 salt_len = salt_bufs[salt_pos].salt_len;
462 wpc_len[0] = 64 + 0 + 0 + pw_len;
463 wpc_len[1] = pw_len + 0 + 0 + 64;
464 wpc_len[2] = 64 + salt_len + 0 + pw_len;
465 wpc_len[3] = pw_len + salt_len + 0 + 64;
466 wpc_len[4] = 64 + 0 + pw_len + pw_len;
467 wpc_len[5] = pw_len + 0 + pw_len + 64;
468 wpc_len[6] = 64 + salt_len + pw_len + pw_len;
469 wpc_len[7] = pw_len + salt_len + pw_len + 64;
471 u64 wpc[8][16] = { 0 };
473 for (u32 i = 0; i < 8; i++)
479 for (u32 j = 0; j < pw_len; j++)
481 PUTCHAR64_BE (wpc[i], block_len++, GETCHAR64_BE (l_p_bytes0, j));
491 for (u32 j = 0; j < salt_len; j++)
493 PUTCHAR64_BE (wpc[i], block_len++, GETCHAR64_BE (l_s_bytes0, j));
499 for (u32 j = 0; j < pw_len; j++)
501 PUTCHAR64_BE (wpc[i], block_len++, GETCHAR64_BE (l_p_bytes0, j));
511 for (u32 j = 0; j < pw_len; j++)
513 PUTCHAR64_BE (wpc[i], block_len++, GETCHAR64_BE (l_p_bytes0, j));
517 PUTCHAR64_BE (wpc[i], block_len, 0x80);
519 wpc[i][15] = block_len * 8;
526 u64x l_alt_result[8];
528 l_alt_result[0] = tmps[gid].l_alt_result[0];
529 l_alt_result[1] = tmps[gid].l_alt_result[1];
530 l_alt_result[2] = tmps[gid].l_alt_result[2];
531 l_alt_result[3] = tmps[gid].l_alt_result[3];
532 l_alt_result[4] = tmps[gid].l_alt_result[4];
533 l_alt_result[5] = tmps[gid].l_alt_result[5];
534 l_alt_result[6] = tmps[gid].l_alt_result[6];
535 l_alt_result[7] = tmps[gid].l_alt_result[7];
538 /* Repeatedly run the collected hash value through SHA512 to burn
541 for (u32 i = 0, j = loop_pos; i < loop_cnt; i++, j++)
543 const u32 j1 = (j & 1) ? 1 : 0;
544 const u32 j3 = (j % 3) ? 2 : 0;
545 const u32 j7 = (j % 7) ? 4 : 0;
547 const u32 pc = j1 + j3 + j7;
551 block[ 0] = wpc[pc][ 0];
552 block[ 1] = wpc[pc][ 1];
553 block[ 2] = wpc[pc][ 2];
554 block[ 3] = wpc[pc][ 3];
555 block[ 4] = wpc[pc][ 4];
556 block[ 5] = wpc[pc][ 5];
557 block[ 6] = wpc[pc][ 6];
558 block[ 7] = wpc[pc][ 7];
559 block[ 8] = wpc[pc][ 8];
560 block[ 9] = wpc[pc][ 9];
561 block[10] = wpc[pc][10];
562 block[11] = wpc[pc][11];
563 block[12] = wpc[pc][12];
564 block[13] = wpc[pc][13];
565 block[14] = wpc[pc][14];
566 block[15] = wpc[pc][15];
570 const u32 block_len = wpc_len[pc];
573 for (u32 k = 0, p = block_len - 64; k < 64; k++, p++)
575 PUTCHAR64_BE (block, p, GETCHAR64_BE (l_alt_result, k));
580 block[0] = l_alt_result[0];
581 block[1] = l_alt_result[1];
582 block[2] = l_alt_result[2];
583 block[3] = l_alt_result[3];
584 block[4] = l_alt_result[4];
585 block[5] = l_alt_result[5];
586 block[6] = l_alt_result[6];
587 block[7] = l_alt_result[7];
590 l_alt_result[0] = SHA512M_A;
591 l_alt_result[1] = SHA512M_B;
592 l_alt_result[2] = SHA512M_C;
593 l_alt_result[3] = SHA512M_D;
594 l_alt_result[4] = SHA512M_E;
595 l_alt_result[5] = SHA512M_F;
596 l_alt_result[6] = SHA512M_G;
597 l_alt_result[7] = SHA512M_H;
599 sha512_transform (block, l_alt_result);
602 tmps[gid].l_alt_result[0] = l_alt_result[0];
603 tmps[gid].l_alt_result[1] = l_alt_result[1];
604 tmps[gid].l_alt_result[2] = l_alt_result[2];
605 tmps[gid].l_alt_result[3] = l_alt_result[3];
606 tmps[gid].l_alt_result[4] = l_alt_result[4];
607 tmps[gid].l_alt_result[5] = l_alt_result[5];
608 tmps[gid].l_alt_result[6] = l_alt_result[6];
609 tmps[gid].l_alt_result[7] = l_alt_result[7];
612 __kernel void __attribute__((reqd_work_group_size (64, 1, 1))) m01800_comp (__global pw_t *pws, __global gpu_rule_t *rules_buf, __global comb_t *combs_buf, __global bf_t *bfs_buf, __global sha512crypt_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)
618 const u32 gid = get_global_id (0);
620 if (gid >= gid_max) return;
622 const u32 lid = get_local_id (0);
624 const u64x a = swap_workaround (tmps[gid].l_alt_result[0]);
625 const u64x b = swap_workaround (tmps[gid].l_alt_result[1]);
627 const u32x r0 = l32_from_64 (a);
628 const u32x r1 = h32_from_64 (a);
629 const u32x r2 = l32_from_64 (b);
630 const u32x r3 = h32_from_64 (b);
634 #include VECT_COMPARE_M