1 /***********************************************************************
2 **
3 ** Implementation of the Skein block functions.
4 **
5 ** Source code author: Doug Whiting, 2008.
6 **
7 ** This algorithm and source code is released to the public domain.
8 **
9 ** Compile-time switches:
10 **
11 ** SKEIN_USE_ASM -- set bits (256/512/1024) to select which
12 ** versions use ASM code for block processing
13 ** [default: use C for all block sizes]
14 **
15 ************************************************************************/
16
17 #include <linux/string.h>
18 #include "skein.h"
19 #include "skein_block.h"
20
21 #ifndef SKEIN_USE_ASM
22 #define SKEIN_USE_ASM (0) /* default is all C code (no ASM) */
23 #endif
24
25 #ifndef SKEIN_LOOP
26 #define SKEIN_LOOP 001 /* default: unroll 256 and 512, but not 1024 */
27 #endif
28
29 #define BLK_BITS (WCNT*64) /* some useful definitions for code here */
30 #define KW_TWK_BASE (0)
31 #define KW_KEY_BASE (3)
32 #define ks (kw + KW_KEY_BASE)
33 #define ts (kw + KW_TWK_BASE)
34
35 #ifdef SKEIN_DEBUG
36 #define debug_save_tweak(ctx) { \
37 ctx->h.tweak[0] = ts[0]; ctx->h.tweak[1] = ts[1]; }
38 #else
39 #define debug_save_tweak(ctx)
40 #endif
41
42 /***************************** SKEIN_256 ******************************/
43 #if !(SKEIN_USE_ASM & 256)
skein_256_process_block(struct skein_256_ctx * ctx,const u8 * blk_ptr,size_t blk_cnt,size_t byte_cnt_add)44 void skein_256_process_block(struct skein_256_ctx *ctx, const u8 *blk_ptr,
45 size_t blk_cnt, size_t byte_cnt_add)
46 { /* do it in C */
47 enum {
48 WCNT = SKEIN_256_STATE_WORDS
49 };
50 #undef RCNT
51 #define RCNT (SKEIN_256_ROUNDS_TOTAL/8)
52
53 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
54 #define SKEIN_UNROLL_256 (((SKEIN_LOOP)/100)%10)
55 #else
56 #define SKEIN_UNROLL_256 (0)
57 #endif
58
59 #if SKEIN_UNROLL_256
60 #if (RCNT % SKEIN_UNROLL_256)
61 #error "Invalid SKEIN_UNROLL_256" /* sanity check on unroll count */
62 #endif
63 size_t r;
64 u64 kw[WCNT+4+RCNT*2]; /* key schedule: chaining vars + tweak + "rot"*/
65 #else
66 u64 kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
67 #endif
68 u64 X0, X1, X2, X3; /* local copy of context vars, for speed */
69 u64 w[WCNT]; /* local copy of input block */
70 #ifdef SKEIN_DEBUG
71 const u64 *X_ptr[4]; /* use for debugging (help cc put Xn in regs) */
72
73 X_ptr[0] = &X0; X_ptr[1] = &X1; X_ptr[2] = &X2; X_ptr[3] = &X3;
74 #endif
75 skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
76 ts[0] = ctx->h.tweak[0];
77 ts[1] = ctx->h.tweak[1];
78 do {
79 /*
80 * this implementation only supports 2**64 input bytes
81 * (no carry out here)
82 */
83 ts[0] += byte_cnt_add; /* update processed length */
84
85 /* precompute the key schedule for this block */
86 ks[0] = ctx->x[0];
87 ks[1] = ctx->x[1];
88 ks[2] = ctx->x[2];
89 ks[3] = ctx->x[3];
90 ks[4] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^ SKEIN_KS_PARITY;
91
92 ts[2] = ts[0] ^ ts[1];
93
94 /* get input block in little-endian format */
95 skein_get64_lsb_first(w, blk_ptr, WCNT);
96 debug_save_tweak(ctx);
97 skein_show_block(BLK_BITS, &ctx->h, ctx->x, blk_ptr, w, ks, ts);
98
99 X0 = w[0] + ks[0]; /* do the first full key injection */
100 X1 = w[1] + ks[1] + ts[0];
101 X2 = w[2] + ks[2] + ts[1];
102 X3 = w[3] + ks[3];
103
104 /* show starting state values */
105 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
106 x_ptr);
107
108 blk_ptr += SKEIN_256_BLOCK_BYTES;
109
110 /* run the rounds */
111
112 #define ROUND256(p0, p1, p2, p3, ROT, r_num) \
113 do { \
114 X##p0 += X##p1; X##p1 = rotl_64(X##p1, ROT##_0); X##p1 ^= X##p0; \
115 X##p2 += X##p3; X##p3 = rotl_64(X##p3, ROT##_1); X##p3 ^= X##p2; \
116 } while (0)
117
118 #if SKEIN_UNROLL_256 == 0
119 #define R256(p0, p1, p2, p3, ROT, r_num) /* fully unrolled */ \
120 do { \
121 ROUND256(p0, p1, p2, p3, ROT, r_num); \
122 skein_show_r_ptr(BLK_BITS, &ctx->h, r_num, X_ptr); \
123 } while (0)
124
125 #define I256(R) \
126 do { \
127 /* inject the key schedule value */ \
128 X0 += ks[((R)+1) % 5]; \
129 X1 += ks[((R)+2) % 5] + ts[((R)+1) % 3]; \
130 X2 += ks[((R)+3) % 5] + ts[((R)+2) % 3]; \
131 X3 += ks[((R)+4) % 5] + (R)+1; \
132 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
133 } while (0)
134 #else /* looping version */
135 #define R256(p0, p1, p2, p3, ROT, r_num) \
136 do { \
137 ROUND256(p0, p1, p2, p3, ROT, r_num); \
138 skein_show_r_ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + r_num, X_ptr); \
139 } while (0)
140
141 #define I256(R) \
142 do { \
143 /* inject the key schedule value */ \
144 X0 += ks[r+(R)+0]; \
145 X1 += ks[r+(R)+1] + ts[r+(R)+0]; \
146 X2 += ks[r+(R)+2] + ts[r+(R)+1]; \
147 X3 += ks[r+(R)+3] + r+(R); \
148 /* rotate key schedule */ \
149 ks[r + (R) + 4] = ks[r + (R) - 1]; \
150 ts[r + (R) + 2] = ts[r + (R) - 1]; \
151 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
152 } while (0)
153
154 for (r = 1; r < 2 * RCNT; r += 2 * SKEIN_UNROLL_256)
155 #endif
156 {
157 #define R256_8_ROUNDS(R) \
158 do { \
159 R256(0, 1, 2, 3, R_256_0, 8 * (R) + 1); \
160 R256(0, 3, 2, 1, R_256_1, 8 * (R) + 2); \
161 R256(0, 1, 2, 3, R_256_2, 8 * (R) + 3); \
162 R256(0, 3, 2, 1, R_256_3, 8 * (R) + 4); \
163 I256(2 * (R)); \
164 R256(0, 1, 2, 3, R_256_4, 8 * (R) + 5); \
165 R256(0, 3, 2, 1, R_256_5, 8 * (R) + 6); \
166 R256(0, 1, 2, 3, R_256_6, 8 * (R) + 7); \
167 R256(0, 3, 2, 1, R_256_7, 8 * (R) + 8); \
168 I256(2 * (R) + 1); \
169 } while (0)
170
171 R256_8_ROUNDS(0);
172
173 #define R256_UNROLL_R(NN) \
174 ((SKEIN_UNROLL_256 == 0 && \
175 SKEIN_256_ROUNDS_TOTAL/8 > (NN)) || \
176 (SKEIN_UNROLL_256 > (NN)))
177
178 #if R256_UNROLL_R(1)
179 R256_8_ROUNDS(1);
180 #endif
181 #if R256_UNROLL_R(2)
182 R256_8_ROUNDS(2);
183 #endif
184 #if R256_UNROLL_R(3)
185 R256_8_ROUNDS(3);
186 #endif
187 #if R256_UNROLL_R(4)
188 R256_8_ROUNDS(4);
189 #endif
190 #if R256_UNROLL_R(5)
191 R256_8_ROUNDS(5);
192 #endif
193 #if R256_UNROLL_R(6)
194 R256_8_ROUNDS(6);
195 #endif
196 #if R256_UNROLL_R(7)
197 R256_8_ROUNDS(7);
198 #endif
199 #if R256_UNROLL_R(8)
200 R256_8_ROUNDS(8);
201 #endif
202 #if R256_UNROLL_R(9)
203 R256_8_ROUNDS(9);
204 #endif
205 #if R256_UNROLL_R(10)
206 R256_8_ROUNDS(10);
207 #endif
208 #if R256_UNROLL_R(11)
209 R256_8_ROUNDS(11);
210 #endif
211 #if R256_UNROLL_R(12)
212 R256_8_ROUNDS(12);
213 #endif
214 #if R256_UNROLL_R(13)
215 R256_8_ROUNDS(13);
216 #endif
217 #if R256_UNROLL_R(14)
218 R256_8_ROUNDS(14);
219 #endif
220 #if (SKEIN_UNROLL_256 > 14)
221 #error "need more unrolling in skein_256_process_block"
222 #endif
223 }
224 /* do the final "feedforward" xor, update context chaining */
225 ctx->x[0] = X0 ^ w[0];
226 ctx->x[1] = X1 ^ w[1];
227 ctx->x[2] = X2 ^ w[2];
228 ctx->x[3] = X3 ^ w[3];
229
230 skein_show_round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->x);
231
232 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
233 } while (--blk_cnt);
234 ctx->h.tweak[0] = ts[0];
235 ctx->h.tweak[1] = ts[1];
236 }
237
238 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
skein_256_process_block_code_size(void)239 size_t skein_256_process_block_code_size(void)
240 {
241 return ((u8 *) skein_256_process_block_code_size) -
242 ((u8 *) skein_256_process_block);
243 }
skein_256_unroll_cnt(void)244 unsigned int skein_256_unroll_cnt(void)
245 {
246 return SKEIN_UNROLL_256;
247 }
248 #endif
249 #endif
250
251 /***************************** SKEIN_512 ******************************/
252 #if !(SKEIN_USE_ASM & 512)
skein_512_process_block(struct skein_512_ctx * ctx,const u8 * blk_ptr,size_t blk_cnt,size_t byte_cnt_add)253 void skein_512_process_block(struct skein_512_ctx *ctx, const u8 *blk_ptr,
254 size_t blk_cnt, size_t byte_cnt_add)
255 { /* do it in C */
256 enum {
257 WCNT = SKEIN_512_STATE_WORDS
258 };
259 #undef RCNT
260 #define RCNT (SKEIN_512_ROUNDS_TOTAL/8)
261
262 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
263 #define SKEIN_UNROLL_512 (((SKEIN_LOOP)/10)%10)
264 #else
265 #define SKEIN_UNROLL_512 (0)
266 #endif
267
268 #if SKEIN_UNROLL_512
269 #if (RCNT % SKEIN_UNROLL_512)
270 #error "Invalid SKEIN_UNROLL_512" /* sanity check on unroll count */
271 #endif
272 size_t r;
273 u64 kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot"*/
274 #else
275 u64 kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
276 #endif
277 u64 X0, X1, X2, X3, X4, X5, X6, X7; /* local copies, for speed */
278 u64 w[WCNT]; /* local copy of input block */
279 #ifdef SKEIN_DEBUG
280 const u64 *X_ptr[8]; /* use for debugging (help cc put Xn in regs) */
281
282 X_ptr[0] = &X0; X_ptr[1] = &X1; X_ptr[2] = &X2; X_ptr[3] = &X3;
283 X_ptr[4] = &X4; X_ptr[5] = &X5; X_ptr[6] = &X6; X_ptr[7] = &X7;
284 #endif
285
286 skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
287 ts[0] = ctx->h.tweak[0];
288 ts[1] = ctx->h.tweak[1];
289 do {
290 /*
291 * this implementation only supports 2**64 input bytes
292 * (no carry out here)
293 */
294 ts[0] += byte_cnt_add; /* update processed length */
295
296 /* precompute the key schedule for this block */
297 ks[0] = ctx->x[0];
298 ks[1] = ctx->x[1];
299 ks[2] = ctx->x[2];
300 ks[3] = ctx->x[3];
301 ks[4] = ctx->x[4];
302 ks[5] = ctx->x[5];
303 ks[6] = ctx->x[6];
304 ks[7] = ctx->x[7];
305 ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
306 ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;
307
308 ts[2] = ts[0] ^ ts[1];
309
310 /* get input block in little-endian format */
311 skein_get64_lsb_first(w, blk_ptr, WCNT);
312 debug_save_tweak(ctx);
313 skein_show_block(BLK_BITS, &ctx->h, ctx->x, blk_ptr, w, ks, ts);
314
315 X0 = w[0] + ks[0]; /* do the first full key injection */
316 X1 = w[1] + ks[1];
317 X2 = w[2] + ks[2];
318 X3 = w[3] + ks[3];
319 X4 = w[4] + ks[4];
320 X5 = w[5] + ks[5] + ts[0];
321 X6 = w[6] + ks[6] + ts[1];
322 X7 = w[7] + ks[7];
323
324 blk_ptr += SKEIN_512_BLOCK_BYTES;
325
326 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
327 X_ptr);
328 /* run the rounds */
329 #define ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
330 do { \
331 X##p0 += X##p1; X##p1 = rotl_64(X##p1, ROT##_0); X##p1 ^= X##p0; \
332 X##p2 += X##p3; X##p3 = rotl_64(X##p3, ROT##_1); X##p3 ^= X##p2; \
333 X##p4 += X##p5; X##p5 = rotl_64(X##p5, ROT##_2); X##p5 ^= X##p4; \
334 X##p6 += X##p7; X##p7 = rotl_64(X##p7, ROT##_3); X##p7 ^= X##p6; \
335 } while (0)
336
337 #if SKEIN_UNROLL_512 == 0
338 #define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) /* unrolled */ \
339 do { \
340 ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
341 skein_show_r_ptr(BLK_BITS, &ctx->h, r_num, X_ptr); \
342 } while (0)
343
344 #define I512(R) \
345 do { \
346 /* inject the key schedule value */ \
347 X0 += ks[((R) + 1) % 9]; \
348 X1 += ks[((R) + 2) % 9]; \
349 X2 += ks[((R) + 3) % 9]; \
350 X3 += ks[((R) + 4) % 9]; \
351 X4 += ks[((R) + 5) % 9]; \
352 X5 += ks[((R) + 6) % 9] + ts[((R) + 1) % 3]; \
353 X6 += ks[((R) + 7) % 9] + ts[((R) + 2) % 3]; \
354 X7 += ks[((R) + 8) % 9] + (R) + 1; \
355 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
356 } while (0)
357 #else /* looping version */
358 #define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
359 do { \
360 ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num); \
361 skein_show_r_ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + r_num, X_ptr); \
362 } while (0)
363
364 #define I512(R) \
365 do { \
366 /* inject the key schedule value */ \
367 X0 += ks[r + (R) + 0]; \
368 X1 += ks[r + (R) + 1]; \
369 X2 += ks[r + (R) + 2]; \
370 X3 += ks[r + (R) + 3]; \
371 X4 += ks[r + (R) + 4]; \
372 X5 += ks[r + (R) + 5] + ts[r + (R) + 0]; \
373 X6 += ks[r + (R) + 6] + ts[r + (R) + 1]; \
374 X7 += ks[r + (R) + 7] + r + (R); \
375 /* rotate key schedule */ \
376 ks[r + (R) + 8] = ks[r + (R) - 1]; \
377 ts[r + (R) + 2] = ts[r + (R) - 1]; \
378 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
379 } while (0)
380
381 for (r = 1; r < 2 * RCNT; r += 2 * SKEIN_UNROLL_512)
382 #endif /* end of looped code definitions */
383 {
384 #define R512_8_ROUNDS(R) /* do 8 full rounds */ \
385 do { \
386 R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_0, 8 * (R) + 1); \
387 R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_1, 8 * (R) + 2); \
388 R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_2, 8 * (R) + 3); \
389 R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_3, 8 * (R) + 4); \
390 I512(2 * (R)); \
391 R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_4, 8 * (R) + 5); \
392 R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_5, 8 * (R) + 6); \
393 R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_6, 8 * (R) + 7); \
394 R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_7, 8 * (R) + 8); \
395 I512(2 * (R) + 1); /* and key injection */ \
396 } while (0)
397
398 R512_8_ROUNDS(0);
399
400 #define R512_UNROLL_R(NN) \
401 ((SKEIN_UNROLL_512 == 0 && \
402 SKEIN_512_ROUNDS_TOTAL/8 > (NN)) || \
403 (SKEIN_UNROLL_512 > (NN)))
404
405 #if R512_UNROLL_R(1)
406 R512_8_ROUNDS(1);
407 #endif
408 #if R512_UNROLL_R(2)
409 R512_8_ROUNDS(2);
410 #endif
411 #if R512_UNROLL_R(3)
412 R512_8_ROUNDS(3);
413 #endif
414 #if R512_UNROLL_R(4)
415 R512_8_ROUNDS(4);
416 #endif
417 #if R512_UNROLL_R(5)
418 R512_8_ROUNDS(5);
419 #endif
420 #if R512_UNROLL_R(6)
421 R512_8_ROUNDS(6);
422 #endif
423 #if R512_UNROLL_R(7)
424 R512_8_ROUNDS(7);
425 #endif
426 #if R512_UNROLL_R(8)
427 R512_8_ROUNDS(8);
428 #endif
429 #if R512_UNROLL_R(9)
430 R512_8_ROUNDS(9);
431 #endif
432 #if R512_UNROLL_R(10)
433 R512_8_ROUNDS(10);
434 #endif
435 #if R512_UNROLL_R(11)
436 R512_8_ROUNDS(11);
437 #endif
438 #if R512_UNROLL_R(12)
439 R512_8_ROUNDS(12);
440 #endif
441 #if R512_UNROLL_R(13)
442 R512_8_ROUNDS(13);
443 #endif
444 #if R512_UNROLL_R(14)
445 R512_8_ROUNDS(14);
446 #endif
447 #if (SKEIN_UNROLL_512 > 14)
448 #error "need more unrolling in skein_512_process_block"
449 #endif
450 }
451
452 /* do the final "feedforward" xor, update context chaining */
453 ctx->x[0] = X0 ^ w[0];
454 ctx->x[1] = X1 ^ w[1];
455 ctx->x[2] = X2 ^ w[2];
456 ctx->x[3] = X3 ^ w[3];
457 ctx->x[4] = X4 ^ w[4];
458 ctx->x[5] = X5 ^ w[5];
459 ctx->x[6] = X6 ^ w[6];
460 ctx->x[7] = X7 ^ w[7];
461 skein_show_round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->x);
462
463 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
464 } while (--blk_cnt);
465 ctx->h.tweak[0] = ts[0];
466 ctx->h.tweak[1] = ts[1];
467 }
468
469 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
skein_512_process_block_code_size(void)470 size_t skein_512_process_block_code_size(void)
471 {
472 return ((u8 *) skein_512_process_block_code_size) -
473 ((u8 *) skein_512_process_block);
474 }
skein_512_unroll_cnt(void)475 unsigned int skein_512_unroll_cnt(void)
476 {
477 return SKEIN_UNROLL_512;
478 }
479 #endif
480 #endif
481
482 /***************************** SKEIN_1024 ******************************/
483 #if !(SKEIN_USE_ASM & 1024)
skein_1024_process_block(struct skein_1024_ctx * ctx,const u8 * blk_ptr,size_t blk_cnt,size_t byte_cnt_add)484 void skein_1024_process_block(struct skein_1024_ctx *ctx, const u8 *blk_ptr,
485 size_t blk_cnt, size_t byte_cnt_add)
486 { /* do it in C, always looping (unrolled is bigger AND slower!) */
487 enum {
488 WCNT = SKEIN_1024_STATE_WORDS
489 };
490 #undef RCNT
491 #define RCNT (SKEIN_1024_ROUNDS_TOTAL/8)
492
493 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
494 #define SKEIN_UNROLL_1024 ((SKEIN_LOOP)%10)
495 #else
496 #define SKEIN_UNROLL_1024 (0)
497 #endif
498
499 #if (SKEIN_UNROLL_1024 != 0)
500 #if (RCNT % SKEIN_UNROLL_1024)
501 #error "Invalid SKEIN_UNROLL_1024" /* sanity check on unroll count */
502 #endif
503 size_t r;
504 u64 kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot" */
505 #else
506 u64 kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
507 #endif
508
509 /* local copy of vars, for speed */
510 u64 X00, X01, X02, X03, X04, X05, X06, X07,
511 X08, X09, X10, X11, X12, X13, X14, X15;
512 u64 w[WCNT]; /* local copy of input block */
513 #ifdef SKEIN_DEBUG
514 const u64 *X_ptr[16]; /* use for debugging (help cc put Xn in regs) */
515
516 X_ptr[0] = &X00; X_ptr[1] = &X01; X_ptr[2] = &X02;
517 X_ptr[3] = &X03; X_ptr[4] = &X04; X_ptr[5] = &X05;
518 X_ptr[6] = &X06; X_ptr[7] = &X07; X_ptr[8] = &X08;
519 X_ptr[9] = &X09; X_ptr[10] = &X10; X_ptr[11] = &X11;
520 X_ptr[12] = &X12; X_ptr[13] = &X13; X_ptr[14] = &X14;
521 X_ptr[15] = &X15;
522 #endif
523
524 skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
525 ts[0] = ctx->h.tweak[0];
526 ts[1] = ctx->h.tweak[1];
527 do {
528 /*
529 * this implementation only supports 2**64 input bytes
530 * (no carry out here)
531 */
532 ts[0] += byte_cnt_add; /* update processed length */
533
534 /* precompute the key schedule for this block */
535 ks[0] = ctx->x[0];
536 ks[1] = ctx->x[1];
537 ks[2] = ctx->x[2];
538 ks[3] = ctx->x[3];
539 ks[4] = ctx->x[4];
540 ks[5] = ctx->x[5];
541 ks[6] = ctx->x[6];
542 ks[7] = ctx->x[7];
543 ks[8] = ctx->x[8];
544 ks[9] = ctx->x[9];
545 ks[10] = ctx->x[10];
546 ks[11] = ctx->x[11];
547 ks[12] = ctx->x[12];
548 ks[13] = ctx->x[13];
549 ks[14] = ctx->x[14];
550 ks[15] = ctx->x[15];
551 ks[16] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
552 ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^
553 ks[8] ^ ks[9] ^ ks[10] ^ ks[11] ^
554 ks[12] ^ ks[13] ^ ks[14] ^ ks[15] ^ SKEIN_KS_PARITY;
555
556 ts[2] = ts[0] ^ ts[1];
557
558 /* get input block in little-endian format */
559 skein_get64_lsb_first(w, blk_ptr, WCNT);
560 debug_save_tweak(ctx);
561 skein_show_block(BLK_BITS, &ctx->h, ctx->x, blk_ptr, w, ks, ts);
562
563 X00 = w[0] + ks[0]; /* do the first full key injection */
564 X01 = w[1] + ks[1];
565 X02 = w[2] + ks[2];
566 X03 = w[3] + ks[3];
567 X04 = w[4] + ks[4];
568 X05 = w[5] + ks[5];
569 X06 = w[6] + ks[6];
570 X07 = w[7] + ks[7];
571 X08 = w[8] + ks[8];
572 X09 = w[9] + ks[9];
573 X10 = w[10] + ks[10];
574 X11 = w[11] + ks[11];
575 X12 = w[12] + ks[12];
576 X13 = w[13] + ks[13] + ts[0];
577 X14 = w[14] + ks[14] + ts[1];
578 X15 = w[15] + ks[15];
579
580 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INITIAL,
581 X_ptr);
582
583 #define ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
584 pF, ROT, r_num) \
585 do { \
586 X##p0 += X##p1; X##p1 = rotl_64(X##p1, ROT##_0); X##p1 ^= X##p0; \
587 X##p2 += X##p3; X##p3 = rotl_64(X##p3, ROT##_1); X##p3 ^= X##p2; \
588 X##p4 += X##p5; X##p5 = rotl_64(X##p5, ROT##_2); X##p5 ^= X##p4; \
589 X##p6 += X##p7; X##p7 = rotl_64(X##p7, ROT##_3); X##p7 ^= X##p6; \
590 X##p8 += X##p9; X##p9 = rotl_64(X##p9, ROT##_4); X##p9 ^= X##p8; \
591 X##pA += X##pB; X##pB = rotl_64(X##pB, ROT##_5); X##pB ^= X##pA; \
592 X##pC += X##pD; X##pD = rotl_64(X##pD, ROT##_6); X##pD ^= X##pC; \
593 X##pE += X##pF; X##pF = rotl_64(X##pF, ROT##_7); X##pF ^= X##pE; \
594 } while (0)
595
596 #if SKEIN_UNROLL_1024 == 0
597 #define R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, pF, \
598 ROT, rn) \
599 do { \
600 ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
601 pF, ROT, rn); \
602 skein_show_r_ptr(BLK_BITS, &ctx->h, rn, X_ptr); \
603 } while (0)
604
605 #define I1024(R) \
606 do { \
607 /* inject the key schedule value */ \
608 X00 += ks[((R) + 1) % 17]; \
609 X01 += ks[((R) + 2) % 17]; \
610 X02 += ks[((R) + 3) % 17]; \
611 X03 += ks[((R) + 4) % 17]; \
612 X04 += ks[((R) + 5) % 17]; \
613 X05 += ks[((R) + 6) % 17]; \
614 X06 += ks[((R) + 7) % 17]; \
615 X07 += ks[((R) + 8) % 17]; \
616 X08 += ks[((R) + 9) % 17]; \
617 X09 += ks[((R) + 10) % 17]; \
618 X10 += ks[((R) + 11) % 17]; \
619 X11 += ks[((R) + 12) % 17]; \
620 X12 += ks[((R) + 13) % 17]; \
621 X13 += ks[((R) + 14) % 17] + ts[((R) + 1) % 3]; \
622 X14 += ks[((R) + 15) % 17] + ts[((R) + 2) % 3]; \
623 X15 += ks[((R) + 16) % 17] + (R) + 1; \
624 skein_show_r_ptr(BLK_BITS, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
625 } while (0)
626 #else /* looping version */
627 #define R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, pF, \
628 ROT, rn) \
629 do { \
630 ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
631 pF, ROT, rn); \
632 skein_show_r_ptr(BLK_BITS, &ctx->h, 4 * (r - 1) + rn, X_ptr); \
633 } while (0)
634
635 #define I1024(R) \
636 do { \
637 /* inject the key schedule value */ \
638 X00 += ks[r + (R) + 0]; \
639 X01 += ks[r + (R) + 1]; \
640 X02 += ks[r + (R) + 2]; \
641 X03 += ks[r + (R) + 3]; \
642 X04 += ks[r + (R) + 4]; \
643 X05 += ks[r + (R) + 5]; \
644 X06 += ks[r + (R) + 6]; \
645 X07 += ks[r + (R) + 7]; \
646 X08 += ks[r + (R) + 8]; \
647 X09 += ks[r + (R) + 9]; \
648 X10 += ks[r + (R) + 10]; \
649 X11 += ks[r + (R) + 11]; \
650 X12 += ks[r + (R) + 12]; \
651 X13 += ks[r + (R) + 13] + ts[r + (R) + 0]; \
652 X14 += ks[r + (R) + 14] + ts[r + (R) + 1]; \
653 X15 += ks[r + (R) + 15] + r + (R); \
654 /* rotate key schedule */ \
655 ks[r + (R) + 16] = ks[r + (R) - 1]; \
656 ts[r + (R) + 2] = ts[r + (R) - 1]; \
657 skein_show_r_ptr(BLK_BITSi, &ctx->h, SKEIN_RND_KEY_INJECT, X_ptr); \
658 } while (0)
659
660 for (r = 1; r <= 2 * RCNT; r += 2 * SKEIN_UNROLL_1024)
661 #endif
662 {
663 #define R1024_8_ROUNDS(R) \
664 do { \
665 R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, \
666 R1024_0, 8*(R) + 1); \
667 R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05, 08, 01, \
668 R1024_1, 8*(R) + 2); \
669 R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11, 10, 09, \
670 R1024_2, 8*(R) + 3); \
671 R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03, 12, 07, \
672 R1024_3, 8*(R) + 4); \
673 I1024(2*(R)); \
674 R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, \
675 R1024_4, 8*(R) + 5); \
676 R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05, 08, 01, \
677 R1024_5, 8*(R) + 6); \
678 R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11, 10, 09, \
679 R1024_6, 8*(R) + 7); \
680 R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03, 12, 07, \
681 R1024_7, 8*(R) + 8); \
682 I1024(2*(R)+1); \
683 } while (0)
684
685 R1024_8_ROUNDS(0);
686
687 #define R1024_UNROLL_R(NN) \
688 ((SKEIN_UNROLL_1024 == 0 && \
689 SKEIN_1024_ROUNDS_TOTAL/8 > (NN)) || \
690 (SKEIN_UNROLL_1024 > (NN)))
691
692 #if R1024_UNROLL_R(1)
693 R1024_8_ROUNDS(1);
694 #endif
695 #if R1024_UNROLL_R(2)
696 R1024_8_ROUNDS(2);
697 #endif
698 #if R1024_UNROLL_R(3)
699 R1024_8_ROUNDS(3);
700 #endif
701 #if R1024_UNROLL_R(4)
702 R1024_8_ROUNDS(4);
703 #endif
704 #if R1024_UNROLL_R(5)
705 R1024_8_ROUNDS(5);
706 #endif
707 #if R1024_UNROLL_R(6)
708 R1024_8_ROUNDS(6);
709 #endif
710 #if R1024_UNROLL_R(7)
711 R1024_8_ROUNDS(7);
712 #endif
713 #if R1024_UNROLL_R(8)
714 R1024_8_ROUNDS(8);
715 #endif
716 #if R1024_UNROLL_R(9)
717 R1024_8_ROUNDS(9);
718 #endif
719 #if R1024_UNROLL_R(10)
720 R1024_8_ROUNDS(10);
721 #endif
722 #if R1024_UNROLL_R(11)
723 R1024_8_ROUNDS(11);
724 #endif
725 #if R1024_UNROLL_R(12)
726 R1024_8_ROUNDS(12);
727 #endif
728 #if R1024_UNROLL_R(13)
729 R1024_8_ROUNDS(13);
730 #endif
731 #if R1024_UNROLL_R(14)
732 R1024_8_ROUNDS(14);
733 #endif
734 #if (SKEIN_UNROLL_1024 > 14)
735 #error "need more unrolling in Skein_1024_Process_Block"
736 #endif
737 }
738 /* do the final "feedforward" xor, update context chaining */
739
740 ctx->x[0] = X00 ^ w[0];
741 ctx->x[1] = X01 ^ w[1];
742 ctx->x[2] = X02 ^ w[2];
743 ctx->x[3] = X03 ^ w[3];
744 ctx->x[4] = X04 ^ w[4];
745 ctx->x[5] = X05 ^ w[5];
746 ctx->x[6] = X06 ^ w[6];
747 ctx->x[7] = X07 ^ w[7];
748 ctx->x[8] = X08 ^ w[8];
749 ctx->x[9] = X09 ^ w[9];
750 ctx->x[10] = X10 ^ w[10];
751 ctx->x[11] = X11 ^ w[11];
752 ctx->x[12] = X12 ^ w[12];
753 ctx->x[13] = X13 ^ w[13];
754 ctx->x[14] = X14 ^ w[14];
755 ctx->x[15] = X15 ^ w[15];
756
757 skein_show_round(BLK_BITS, &ctx->h, SKEIN_RND_FEED_FWD, ctx->x);
758
759 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
760 blk_ptr += SKEIN_1024_BLOCK_BYTES;
761 } while (--blk_cnt);
762 ctx->h.tweak[0] = ts[0];
763 ctx->h.tweak[1] = ts[1];
764 }
765
766 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
skein_1024_process_block_code_size(void)767 size_t skein_1024_process_block_code_size(void)
768 {
769 return ((u8 *) skein_1024_process_block_code_size) -
770 ((u8 *) skein_1024_process_block);
771 }
skein_1024_unroll_cnt(void)772 unsigned int skein_1024_unroll_cnt(void)
773 {
774 return SKEIN_UNROLL_1024;
775 }
776 #endif
777 #endif
778