drivers/staging/skein/skein_block.c

   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_base.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 {                                   \
  38         ctx->h.tweak[0] = ts[0];    \
  39         ctx->h.tweak[1] = ts[1];    \
  40 }
  41 #else
  42 #define debug_save_tweak(ctx)
  43 #endif
  44
  45 #if !(SKEIN_USE_ASM & 256)
  46 #undef  RCNT
  47 #define RCNT (SKEIN_256_ROUNDS_TOTAL / 8)
  48 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
  49 #define SKEIN_UNROLL_256 (((SKEIN_LOOP) / 100) % 10)
  50 #else
  51 #define SKEIN_UNROLL_256 (0)
  52 #endif
  53
  54 #if SKEIN_UNROLL_256
  55 #if (RCNT % SKEIN_UNROLL_256)
  56 #error "Invalid SKEIN_UNROLL_256" /* sanity check on unroll count */
  57 #endif
  58 #endif
  59 #define ROUND256(p0, p1, p2, p3, ROT, r_num) \
  60 do {                                         \
  61         X##p0 += X##p1;                      \
  62         X##p1 = rotl_64(X##p1, ROT##_0);     \
  63         X##p1 ^= X##p0;                      \
  64         X##p2 += X##p3;                      \
  65         X##p3 = rotl_64(X##p3, ROT##_1);     \
  66         X##p3 ^= X##p2;                      \
  67 } while (0)
  68
  69 #if SKEIN_UNROLL_256 == 0
  70 #define R256(p0, p1, p2, p3, ROT, r_num) /* fully unrolled */ \
  71 do {                                                          \
  72         ROUND256(p0, p1, p2, p3, ROT, r_num);                 \
  73 } while (0)
  74
  75 #define I256(R)                                                           \
  76 do {                                                                      \
  77         /* inject the key schedule value */                               \
  78         X0   += ks[((R) + 1) % 5];                                        \
  79         X1   += ks[((R) + 2) % 5] + ts[((R) + 1) % 3];                    \
  80         X2   += ks[((R) + 3) % 5] + ts[((R) + 2) % 3];                    \
  81         X3   += ks[((R) + 4) % 5] + (R) + 1;                              \
  82 } while (0)
  83 #else
  84 /* looping version */
  85 #define R256(p0, p1, p2, p3, ROT, r_num) \
  86 do { \
  87         ROUND256(p0, p1, p2, p3, ROT, r_num); \
  88 } while (0)
  89
  90 #define I256(R) \
  91 do { \
  92         /* inject the key schedule value */ \
  93         X0 += ks[r + (R) + 0]; \
  94         X1 += ks[r + (R) + 1] + ts[r + (R) + 0];                          \
  95         X2 += ks[r + (R) + 2] + ts[r + (R) + 1];                          \
  96         X3 += ks[r + (R) + 3] + r + (R);                                  \
  97         /* rotate key schedule */                                         \
  98         ks[r + (R) + 4] = ks[r + (R) - 1];                                \
  99         ts[r + (R) + 2] = ts[r + (R) - 1];                                \
 100 } while (0)
 101 #endif
 102 #define R256_8_ROUNDS(R)                                 \
 103 do {                                                     \
 104                 R256(0, 1, 2, 3, R_256_0, 8 * (R) + 1);  \
 105                 R256(0, 3, 2, 1, R_256_1, 8 * (R) + 2);  \
 106                 R256(0, 1, 2, 3, R_256_2, 8 * (R) + 3);  \
 107                 R256(0, 3, 2, 1, R_256_3, 8 * (R) + 4);  \
 108                 I256(2 * (R));                           \
 109                 R256(0, 1, 2, 3, R_256_4, 8 * (R) + 5);  \
 110                 R256(0, 3, 2, 1, R_256_5, 8 * (R) + 6);  \
 111                 R256(0, 1, 2, 3, R_256_6, 8 * (R) + 7);  \
 112                 R256(0, 3, 2, 1, R_256_7, 8 * (R) + 8);  \
 113                 I256(2 * (R) + 1);                       \
 114 } while (0)
 115
 116 #define R256_UNROLL_R(NN)                     \
 117         ((SKEIN_UNROLL_256 == 0 &&            \
 118         SKEIN_256_ROUNDS_TOTAL / 8 > (NN)) || \
 119         (SKEIN_UNROLL_256 > (NN)))
 120
 121 #if  (SKEIN_UNROLL_256 > 14)
 122 #error  "need more unrolling in skein_256_process_block"
 123 #endif
 124 #endif
 125
 126 #if !(SKEIN_USE_ASM & 512)
 127 #undef  RCNT
 128 #define RCNT  (SKEIN_512_ROUNDS_TOTAL/8)
 129
 130 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
 131 #define SKEIN_UNROLL_512 (((SKEIN_LOOP)/10)%10)
 132 #else
 133 #define SKEIN_UNROLL_512 (0)
 134 #endif
 135
 136 #if SKEIN_UNROLL_512
 137 #if (RCNT % SKEIN_UNROLL_512)
 138 #error "Invalid SKEIN_UNROLL_512" /* sanity check on unroll count */
 139 #endif
 140 #endif
 141 #define ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) \
 142 do {                                                         \
 143         X##p0 += X##p1;                                      \
 144         X##p1 = rotl_64(X##p1, ROT##_0);                     \
 145         X##p1 ^= X##p0;                                      \
 146         X##p2 += X##p3;                                      \
 147         X##p3 = rotl_64(X##p3, ROT##_1);                     \
 148         X##p3 ^= X##p2;                                      \
 149         X##p4 += X##p5;                                      \
 150         X##p5 = rotl_64(X##p5, ROT##_2);                     \
 151         X##p5 ^= X##p4;                                      \
 152         X##p6 += X##p7; X##p7 = rotl_64(X##p7, ROT##_3);     \
 153         X##p7 ^= X##p6;                                      \
 154 } while (0)
 155
 156 #if SKEIN_UNROLL_512 == 0
 157 #define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num) /* unrolled */ \
 158 do {                                                                    \
 159         ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num);           \
 160 } while (0)
 161
 162 #define I512(R)                                                           \
 163 do {                                                                      \
 164         /* inject the key schedule value */                               \
 165         X0   += ks[((R) + 1) % 9];                                        \
 166         X1   += ks[((R) + 2) % 9];                                        \
 167         X2   += ks[((R) + 3) % 9];                                        \
 168         X3   += ks[((R) + 4) % 9];                                        \
 169         X4   += ks[((R) + 5) % 9];                                        \
 170         X5   += ks[((R) + 6) % 9] + ts[((R) + 1) % 3];                    \
 171         X6   += ks[((R) + 7) % 9] + ts[((R) + 2) % 3];                    \
 172         X7   += ks[((R) + 8) % 9] + (R) + 1;                              \
 173 } while (0)
 174
 175 #else /* looping version */
 176 #define R512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num)                 \
 177 do {                                                                     \
 178         ROUND512(p0, p1, p2, p3, p4, p5, p6, p7, ROT, r_num);            \
 179 } while (0)
 180
 181 #define I512(R)                                                           \
 182 do {                                                                      \
 183         /* inject the key schedule value */                               \
 184         X0   += ks[r + (R) + 0];                                          \
 185         X1   += ks[r + (R) + 1];                                          \
 186         X2   += ks[r + (R) + 2];                                          \
 187         X3   += ks[r + (R) + 3];                                          \
 188         X4   += ks[r + (R) + 4];                                          \
 189         X5   += ks[r + (R) + 5] + ts[r + (R) + 0];                        \
 190         X6   += ks[r + (R) + 6] + ts[r + (R) + 1];                        \
 191         X7   += ks[r + (R) + 7] + r + (R);                                \
 192         /* rotate key schedule */                                         \
 193         ks[r + (R) + 8] = ks[r + (R) - 1];                                \
 194         ts[r + (R) + 2] = ts[r + (R) - 1];                                \
 195 } while (0)
 196 #endif /* end of looped code definitions */
 197 #define R512_8_ROUNDS(R)  /* do 8 full rounds */                      \
 198 do {                                                                  \
 199                 R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_0, 8 * (R) + 1);   \
 200                 R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_1, 8 * (R) + 2);   \
 201                 R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_2, 8 * (R) + 3);   \
 202                 R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_3, 8 * (R) + 4);   \
 203                 I512(2 * (R));                              \
 204                 R512(0, 1, 2, 3, 4, 5, 6, 7, R_512_4, 8 * (R) + 5);   \
 205                 R512(2, 1, 4, 7, 6, 5, 0, 3, R_512_5, 8 * (R) + 6);   \
 206                 R512(4, 1, 6, 3, 0, 5, 2, 7, R_512_6, 8 * (R) + 7);   \
 207                 R512(6, 1, 0, 7, 2, 5, 4, 3, R_512_7, 8 * (R) + 8);   \
 208                 I512(2 * (R) + 1);        /* and key injection */     \
 209 } while (0)
 210 #define R512_UNROLL_R(NN)                             \
 211                 ((SKEIN_UNROLL_512 == 0 &&            \
 212                 SKEIN_512_ROUNDS_TOTAL/8 > (NN)) ||   \
 213                 (SKEIN_UNROLL_512 > (NN)))
 214
 215 #if  (SKEIN_UNROLL_512 > 14)
 216 #error  "need more unrolling in skein_512_process_block"
 217 #endif
 218 #endif
 219
 220 #if !(SKEIN_USE_ASM & 1024)
 221 #undef  RCNT
 222 #define RCNT  (SKEIN_1024_ROUNDS_TOTAL/8)
 223 #ifdef SKEIN_LOOP /* configure how much to unroll the loop */
 224 #define SKEIN_UNROLL_1024 ((SKEIN_LOOP)%10)
 225 #else
 226 #define SKEIN_UNROLL_1024 (0)
 227 #endif
 228
 229 #if (SKEIN_UNROLL_1024 != 0)
 230 #if (RCNT % SKEIN_UNROLL_1024)
 231 #error "Invalid SKEIN_UNROLL_1024" /* sanity check on unroll count */
 232 #endif
 233 #endif
 234 #define ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
 235                   pF, ROT, r_num)                                             \
 236 do {                                                                          \
 237         X##p0 += X##p1;                                                       \
 238         X##p1 = rotl_64(X##p1, ROT##_0);                                      \
 239         X##p1 ^= X##p0;                                                       \
 240         X##p2 += X##p3;                                                       \
 241         X##p3 = rotl_64(X##p3, ROT##_1);                                      \
 242         X##p3 ^= X##p2;                                                       \
 243         X##p4 += X##p5;                                                       \
 244         X##p5 = rotl_64(X##p5, ROT##_2);                                      \
 245         X##p5 ^= X##p4;                                                       \
 246         X##p6 += X##p7;                                                       \
 247         X##p7 = rotl_64(X##p7, ROT##_3);                                      \
 248         X##p7 ^= X##p6;                                                       \
 249         X##p8 += X##p9;                                                       \
 250         X##p9 = rotl_64(X##p9, ROT##_4);                                      \
 251         X##p9 ^= X##p8;                                                       \
 252         X##pA += X##pB;                                                       \
 253         X##pB = rotl_64(X##pB, ROT##_5);                                      \
 254         X##pB ^= X##pA;                                                       \
 255         X##pC += X##pD;                                                       \
 256         X##pD = rotl_64(X##pD, ROT##_6);                                      \
 257         X##pD ^= X##pC;                                                       \
 258         X##pE += X##pF;                                                       \
 259         X##pF = rotl_64(X##pF, ROT##_7);                                      \
 260         X##pF ^= X##pE;                                                       \
 261 } while (0)
 262
 263 #if SKEIN_UNROLL_1024 == 0
 264 #define R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, pF, \
 265               ROT, rn)                                                        \
 266 do {                                                                          \
 267         ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
 268                   pF, ROT, rn);                                               \
 269 } while (0)
 270
 271 #define I1024(R)                                                          \
 272 do {                                                                      \
 273         /* inject the key schedule value */                               \
 274         X00 += ks[((R) + 1) % 17];                                        \
 275         X01 += ks[((R) + 2) % 17];                                        \
 276         X02 += ks[((R) + 3) % 17];                                        \
 277         X03 += ks[((R) + 4) % 17];                                        \
 278         X04 += ks[((R) + 5) % 17];                                        \
 279         X05 += ks[((R) + 6) % 17];                                        \
 280         X06 += ks[((R) + 7) % 17];                                        \
 281         X07 += ks[((R) + 8) % 17];                                        \
 282         X08 += ks[((R) + 9) % 17];                                        \
 283         X09 += ks[((R) + 10) % 17];                                       \
 284         X10 += ks[((R) + 11) % 17];                                       \
 285         X11 += ks[((R) + 12) % 17];                                       \
 286         X12 += ks[((R) + 13) % 17];                                       \
 287         X13 += ks[((R) + 14) % 17] + ts[((R) + 1) % 3];                   \
 288         X14 += ks[((R) + 15) % 17] + ts[((R) + 2) % 3];                   \
 289         X15 += ks[((R) + 16) % 17] + (R) + 1;                             \
 290 } while (0)
 291 #else /* looping version */
 292 #define R1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, pF, \
 293               ROT, rn)                                                        \
 294 do {                                                                          \
 295         ROUND1024(p0, p1, p2, p3, p4, p5, p6, p7, p8, p9, pA, pB, pC, pD, pE, \
 296                   pF, ROT, rn);                                               \
 297 } while (0)
 298
 299 #define I1024(R)                                                           \
 300 do {                                                                       \
 301         /* inject the key schedule value */                                \
 302         X00 += ks[r + (R) + 0];                                            \
 303         X01 += ks[r + (R) + 1];                                            \
 304         X02 += ks[r + (R) + 2];                                            \
 305         X03 += ks[r + (R) + 3];                                            \
 306         X04 += ks[r + (R) + 4];                                            \
 307         X05 += ks[r + (R) + 5];                                            \
 308         X06 += ks[r + (R) + 6];                                            \
 309         X07 += ks[r + (R) + 7];                                            \
 310         X08 += ks[r + (R) + 8];                                            \
 311         X09 += ks[r + (R) + 9];                                            \
 312         X10 += ks[r + (R) + 10];                                           \
 313         X11 += ks[r + (R) + 11];                                           \
 314         X12 += ks[r + (R) + 12];                                           \
 315         X13 += ks[r + (R) + 13] + ts[r + (R) + 0];                         \
 316         X14 += ks[r + (R) + 14] + ts[r + (R) + 1];                         \
 317         X15 += ks[r + (R) + 15] + r + (R);                                 \
 318         /* rotate key schedule */                                          \
 319         ks[r + (R) + 16] = ks[r + (R) - 1];                                \
 320         ts[r + (R) + 2] = ts[r + (R) - 1];                                 \
 321 } while (0)
 322
 323 #endif
 324 #define R1024_8_ROUNDS(R)                                                     \
 325 do {                                                                          \
 326         R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, \
 327               R1024_0, 8*(R) + 1);                                            \
 328         R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05, 08, 01, \
 329               R1024_1, 8*(R) + 2);                                            \
 330         R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11, 10, 09, \
 331               R1024_2, 8*(R) + 3);                                            \
 332         R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03, 12, 07, \
 333               R1024_3, 8*(R) + 4);                                            \
 334         I1024(2*(R));                                                         \
 335         R1024(00, 01, 02, 03, 04, 05, 06, 07, 08, 09, 10, 11, 12, 13, 14, 15, \
 336               R1024_4, 8*(R) + 5);                                            \
 337         R1024(00, 09, 02, 13, 06, 11, 04, 15, 10, 07, 12, 03, 14, 05, 08, 01, \
 338               R1024_5, 8*(R) + 6);                                            \
 339         R1024(00, 07, 02, 05, 04, 03, 06, 01, 12, 15, 14, 13, 08, 11, 10, 09, \
 340               R1024_6, 8*(R) + 7);                                            \
 341         R1024(00, 15, 02, 11, 06, 13, 04, 09, 14, 01, 08, 05, 10, 03, 12, 07, \
 342               R1024_7, 8*(R) + 8);                                            \
 343         I1024(2*(R)+1);                                                       \
 344 } while (0)
 345
 346 #define R1024_UNROLL_R(NN)                              \
 347                 ((SKEIN_UNROLL_1024 == 0 &&             \
 348                 SKEIN_1024_ROUNDS_TOTAL/8 > (NN)) ||  \
 349                 (SKEIN_UNROLL_1024 > (NN)))
 350
 351 #if  (SKEIN_UNROLL_1024 > 14)
 352 #error  "need more unrolling in Skein_1024_Process_Block"
 353 #endif
 354 #endif
 355
 356 /*****************************  SKEIN_256 ******************************/
 357 #if !(SKEIN_USE_ASM & 256)
 358 void skein_256_process_block(struct skein_256_ctx *ctx, const u8 *blk_ptr,
 359                              size_t blk_cnt, size_t byte_cnt_add)
 360 { /* do it in C */
 361         enum {
 362                 WCNT = SKEIN_256_STATE_WORDS
 363         };
 364         size_t r;
 365 #if SKEIN_UNROLL_256
 366         /* key schedule: chaining vars + tweak + "rot"*/
 367         u64  kw[WCNT+4+RCNT*2];
 368 #else
 369         /* key schedule words : chaining vars + tweak */
 370         u64  kw[WCNT+4];
 371 #endif
 372         u64  X0, X1, X2, X3; /* local copy of context vars, for speed */
 373         u64  w[WCNT]; /* local copy of input block */
 374 #ifdef SKEIN_DEBUG
 375         const u64 *X_ptr[4]; /* use for debugging (help cc put Xn in regs) */
 376
 377         X_ptr[0] = &X0;
 378         X_ptr[1] = &X1;
 379         X_ptr[2] = &X2;
 380         X_ptr[3] = &X3;
 381 #endif
 382         skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
 383         ts[0] = ctx->h.tweak[0];
 384         ts[1] = ctx->h.tweak[1];
 385         do  {
 386                 /*
 387                  * this implementation only supports 2**64 input bytes
 388                  * (no carry out here)
 389                  */
 390                 ts[0] += byte_cnt_add; /* update processed length */
 391
 392                 /* precompute the key schedule for this block */
 393                 ks[0] = ctx->x[0];
 394                 ks[1] = ctx->x[1];
 395                 ks[2] = ctx->x[2];
 396                 ks[3] = ctx->x[3];
 397                 ks[4] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^ SKEIN_KS_PARITY;
 398
 399                 ts[2] = ts[0] ^ ts[1];
 400
 401                 /* get input block in little-endian format */
 402                 skein_get64_lsb_first(w, blk_ptr, WCNT);
 403                 debug_save_tweak(ctx);
 404
 405                 /* do the first full key injection */
 406                 X0 = w[0] + ks[0];
 407                 X1 = w[1] + ks[1] + ts[0];
 408                 X2 = w[2] + ks[2] + ts[1];
 409                 X3 = w[3] + ks[3];
 410
 411                 blk_ptr += SKEIN_256_BLOCK_BYTES;
 412
 413                 /* run the rounds */
 414                 for (r = 1;
 415                         r < (SKEIN_UNROLL_256 ? 2 * RCNT : 2);
 416                         r += (SKEIN_UNROLL_256 ? 2 * SKEIN_UNROLL_256 : 1)) {
 417                         R256_8_ROUNDS(0);
 418 #if   R256_UNROLL_R(1)
 419                         R256_8_ROUNDS(1);
 420 #endif
 421 #if   R256_UNROLL_R(2)
 422                         R256_8_ROUNDS(2);
 423 #endif
 424 #if   R256_UNROLL_R(3)
 425                         R256_8_ROUNDS(3);
 426 #endif
 427 #if   R256_UNROLL_R(4)
 428                         R256_8_ROUNDS(4);
 429 #endif
 430 #if   R256_UNROLL_R(5)
 431                         R256_8_ROUNDS(5);
 432 #endif
 433 #if   R256_UNROLL_R(6)
 434                         R256_8_ROUNDS(6);
 435 #endif
 436 #if   R256_UNROLL_R(7)
 437                         R256_8_ROUNDS(7);
 438 #endif
 439 #if   R256_UNROLL_R(8)
 440                         R256_8_ROUNDS(8);
 441 #endif
 442 #if   R256_UNROLL_R(9)
 443                         R256_8_ROUNDS(9);
 444 #endif
 445 #if   R256_UNROLL_R(10)
 446                         R256_8_ROUNDS(10);
 447 #endif
 448 #if   R256_UNROLL_R(11)
 449                         R256_8_ROUNDS(11);
 450 #endif
 451 #if   R256_UNROLL_R(12)
 452                         R256_8_ROUNDS(12);
 453 #endif
 454 #if   R256_UNROLL_R(13)
 455                         R256_8_ROUNDS(13);
 456 #endif
 457 #if   R256_UNROLL_R(14)
 458                         R256_8_ROUNDS(14);
 459 #endif
 460                 }
 461                 /* do the final "feedforward" xor, update context chaining */
 462                 ctx->x[0] = X0 ^ w[0];
 463                 ctx->x[1] = X1 ^ w[1];
 464                 ctx->x[2] = X2 ^ w[2];
 465                 ctx->x[3] = X3 ^ w[3];
 466
 467                 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
 468         } while (--blk_cnt);
 469         ctx->h.tweak[0] = ts[0];
 470         ctx->h.tweak[1] = ts[1];
 471 }
 472
 473 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
 474 size_t skein_256_process_block_code_size(void)
 475 {
 476         return ((u8 *) skein_256_process_block_code_size) -
 477                 ((u8 *) skein_256_process_block);
 478 }
 479 unsigned int skein_256_unroll_cnt(void)
 480 {
 481         return SKEIN_UNROLL_256;
 482 }
 483 #endif
 484 #endif
 485
 486 /*****************************  SKEIN_512 ******************************/
 487 #if !(SKEIN_USE_ASM & 512)
 488 void skein_512_process_block(struct skein_512_ctx *ctx, const u8 *blk_ptr,
 489                              size_t blk_cnt, size_t byte_cnt_add)
 490 { /* do it in C */
 491         enum {
 492                 WCNT = SKEIN_512_STATE_WORDS
 493         };
 494         size_t  r;
 495 #if SKEIN_UNROLL_512
 496         u64  kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot"*/
 497 #else
 498         u64  kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
 499 #endif
 500         u64  X0, X1, X2, X3, X4, X5, X6, X7; /* local copies, for speed */
 501         u64  w[WCNT]; /* local copy of input block */
 502 #ifdef SKEIN_DEBUG
 503         const u64 *X_ptr[8]; /* use for debugging (help cc put Xn in regs) */
 504
 505         X_ptr[0] = &X0;
 506         X_ptr[1] = &X1;
 507         X_ptr[2] = &X2;
 508         X_ptr[3] = &X3;
 509         X_ptr[4] = &X4;
 510         X_ptr[5] = &X5;
 511         X_ptr[6] = &X6;
 512         X_ptr[7] = &X7;
 513 #endif
 514
 515         skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
 516         ts[0] = ctx->h.tweak[0];
 517         ts[1] = ctx->h.tweak[1];
 518         do  {
 519                 /*
 520                  * this implementation only supports 2**64 input bytes
 521                  * (no carry out here)
 522                  */
 523                 ts[0] += byte_cnt_add; /* update processed length */
 524
 525                 /* precompute the key schedule for this block */
 526                 ks[0] = ctx->x[0];
 527                 ks[1] = ctx->x[1];
 528                 ks[2] = ctx->x[2];
 529                 ks[3] = ctx->x[3];
 530                 ks[4] = ctx->x[4];
 531                 ks[5] = ctx->x[5];
 532                 ks[6] = ctx->x[6];
 533                 ks[7] = ctx->x[7];
 534                 ks[8] = ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
 535                         ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^ SKEIN_KS_PARITY;
 536
 537                 ts[2] = ts[0] ^ ts[1];
 538
 539                 /* get input block in little-endian format */
 540                 skein_get64_lsb_first(w, blk_ptr, WCNT);
 541                 debug_save_tweak(ctx);
 542
 543                 /* do the first full key injection */
 544                 X0 = w[0] + ks[0];
 545                 X1 = w[1] + ks[1];
 546                 X2 = w[2] + ks[2];
 547                 X3 = w[3] + ks[3];
 548                 X4 = w[4] + ks[4];
 549                 X5 = w[5] + ks[5] + ts[0];
 550                 X6 = w[6] + ks[6] + ts[1];
 551                 X7 = w[7] + ks[7];
 552
 553                 blk_ptr += SKEIN_512_BLOCK_BYTES;
 554
 555                 /* run the rounds */
 556                 for (r = 1;
 557                         r < (SKEIN_UNROLL_512 ? 2 * RCNT : 2);
 558                         r += (SKEIN_UNROLL_512 ? 2 * SKEIN_UNROLL_512 : 1)) {
 559
 560                         R512_8_ROUNDS(0);
 561
 562 #if   R512_UNROLL_R(1)
 563                         R512_8_ROUNDS(1);
 564 #endif
 565 #if   R512_UNROLL_R(2)
 566                         R512_8_ROUNDS(2);
 567 #endif
 568 #if   R512_UNROLL_R(3)
 569                         R512_8_ROUNDS(3);
 570 #endif
 571 #if   R512_UNROLL_R(4)
 572                         R512_8_ROUNDS(4);
 573 #endif
 574 #if   R512_UNROLL_R(5)
 575                         R512_8_ROUNDS(5);
 576 #endif
 577 #if   R512_UNROLL_R(6)
 578                         R512_8_ROUNDS(6);
 579 #endif
 580 #if   R512_UNROLL_R(7)
 581                         R512_8_ROUNDS(7);
 582 #endif
 583 #if   R512_UNROLL_R(8)
 584                         R512_8_ROUNDS(8);
 585 #endif
 586 #if   R512_UNROLL_R(9)
 587                         R512_8_ROUNDS(9);
 588 #endif
 589 #if   R512_UNROLL_R(10)
 590                         R512_8_ROUNDS(10);
 591 #endif
 592 #if   R512_UNROLL_R(11)
 593                         R512_8_ROUNDS(11);
 594 #endif
 595 #if   R512_UNROLL_R(12)
 596                         R512_8_ROUNDS(12);
 597 #endif
 598 #if   R512_UNROLL_R(13)
 599                         R512_8_ROUNDS(13);
 600 #endif
 601 #if   R512_UNROLL_R(14)
 602                         R512_8_ROUNDS(14);
 603 #endif
 604                 }
 605
 606                 /* do the final "feedforward" xor, update context chaining */
 607                 ctx->x[0] = X0 ^ w[0];
 608                 ctx->x[1] = X1 ^ w[1];
 609                 ctx->x[2] = X2 ^ w[2];
 610                 ctx->x[3] = X3 ^ w[3];
 611                 ctx->x[4] = X4 ^ w[4];
 612                 ctx->x[5] = X5 ^ w[5];
 613                 ctx->x[6] = X6 ^ w[6];
 614                 ctx->x[7] = X7 ^ w[7];
 615
 616                 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
 617         } while (--blk_cnt);
 618         ctx->h.tweak[0] = ts[0];
 619         ctx->h.tweak[1] = ts[1];
 620 }
 621
 622 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
 623 size_t skein_512_process_block_code_size(void)
 624 {
 625         return ((u8 *) skein_512_process_block_code_size) -
 626                 ((u8 *) skein_512_process_block);
 627 }
 628 unsigned int skein_512_unroll_cnt(void)
 629 {
 630         return SKEIN_UNROLL_512;
 631 }
 632 #endif
 633 #endif
 634
 635 /*****************************  SKEIN_1024 ******************************/
 636 #if !(SKEIN_USE_ASM & 1024)
 637 void skein_1024_process_block(struct skein_1024_ctx *ctx, const u8 *blk_ptr,
 638                               size_t blk_cnt, size_t byte_cnt_add)
 639 { /* do it in C, always looping (unrolled is bigger AND slower!) */
 640         enum {
 641                 WCNT = SKEIN_1024_STATE_WORDS
 642         };
 643         size_t  r;
 644 #if (SKEIN_UNROLL_1024 != 0)
 645         u64  kw[WCNT+4+RCNT*2]; /* key sched: chaining vars + tweak + "rot" */
 646 #else
 647         u64  kw[WCNT+4]; /* key schedule words : chaining vars + tweak */
 648 #endif
 649
 650         /* local copy of vars, for speed */
 651         u64  X00, X01, X02, X03, X04, X05, X06, X07,
 652              X08, X09, X10, X11, X12, X13, X14, X15;
 653         u64  w[WCNT]; /* local copy of input block */
 654
 655         skein_assert(blk_cnt != 0); /* never call with blk_cnt == 0! */
 656         ts[0] = ctx->h.tweak[0];
 657         ts[1] = ctx->h.tweak[1];
 658         do  {
 659                 /*
 660                  * this implementation only supports 2**64 input bytes
 661                  * (no carry out here)
 662                  */
 663                 ts[0] += byte_cnt_add; /* update processed length */
 664
 665                 /* precompute the key schedule for this block */
 666                 ks[0]  = ctx->x[0];
 667                 ks[1]  = ctx->x[1];
 668                 ks[2]  = ctx->x[2];
 669                 ks[3]  = ctx->x[3];
 670                 ks[4]  = ctx->x[4];
 671                 ks[5]  = ctx->x[5];
 672                 ks[6]  = ctx->x[6];
 673                 ks[7]  = ctx->x[7];
 674                 ks[8]  = ctx->x[8];
 675                 ks[9]  = ctx->x[9];
 676                 ks[10] = ctx->x[10];
 677                 ks[11] = ctx->x[11];
 678                 ks[12] = ctx->x[12];
 679                 ks[13] = ctx->x[13];
 680                 ks[14] = ctx->x[14];
 681                 ks[15] = ctx->x[15];
 682                 ks[16] =  ks[0] ^ ks[1] ^ ks[2] ^ ks[3] ^
 683                           ks[4] ^ ks[5] ^ ks[6] ^ ks[7] ^
 684                           ks[8] ^ ks[9] ^ ks[10] ^ ks[11] ^
 685                           ks[12] ^ ks[13] ^ ks[14] ^ ks[15] ^ SKEIN_KS_PARITY;
 686
 687                 ts[2] = ts[0] ^ ts[1];
 688
 689                 /* get input block in little-endian format */
 690                 skein_get64_lsb_first(w, blk_ptr, WCNT);
 691                 debug_save_tweak(ctx);
 692
 693                 /* do the first full key injection */
 694                 X00 = w[0] + ks[0];
 695                 X01 = w[1] + ks[1];
 696                 X02 = w[2] + ks[2];
 697                 X03 = w[3] + ks[3];
 698                 X04 = w[4] + ks[4];
 699                 X05 = w[5] + ks[5];
 700                 X06 = w[6] + ks[6];
 701                 X07 = w[7] + ks[7];
 702                 X08 = w[8] + ks[8];
 703                 X09 = w[9] + ks[9];
 704                 X10 = w[10] + ks[10];
 705                 X11 = w[11] + ks[11];
 706                 X12 = w[12] + ks[12];
 707                 X13 = w[13] + ks[13] + ts[0];
 708                 X14 = w[14] + ks[14] + ts[1];
 709                 X15 = w[15] + ks[15];
 710
 711                 for (r = 1;
 712                         r < (SKEIN_UNROLL_1024 ? 2 * RCNT : 2);
 713                         r += (SKEIN_UNROLL_1024 ? 2 * SKEIN_UNROLL_1024 : 1)) {
 714                         R1024_8_ROUNDS(0);
 715 #if   R1024_UNROLL_R(1)
 716                         R1024_8_ROUNDS(1);
 717 #endif
 718 #if   R1024_UNROLL_R(2)
 719                         R1024_8_ROUNDS(2);
 720 #endif
 721 #if   R1024_UNROLL_R(3)
 722                         R1024_8_ROUNDS(3);
 723 #endif
 724 #if   R1024_UNROLL_R(4)
 725                         R1024_8_ROUNDS(4);
 726 #endif
 727 #if   R1024_UNROLL_R(5)
 728                         R1024_8_ROUNDS(5);
 729 #endif
 730 #if   R1024_UNROLL_R(6)
 731                         R1024_8_ROUNDS(6);
 732 #endif
 733 #if   R1024_UNROLL_R(7)
 734                         R1024_8_ROUNDS(7);
 735 #endif
 736 #if   R1024_UNROLL_R(8)
 737                         R1024_8_ROUNDS(8);
 738 #endif
 739 #if   R1024_UNROLL_R(9)
 740                         R1024_8_ROUNDS(9);
 741 #endif
 742 #if   R1024_UNROLL_R(10)
 743                         R1024_8_ROUNDS(10);
 744 #endif
 745 #if   R1024_UNROLL_R(11)
 746                         R1024_8_ROUNDS(11);
 747 #endif
 748 #if   R1024_UNROLL_R(12)
 749                         R1024_8_ROUNDS(12);
 750 #endif
 751 #if   R1024_UNROLL_R(13)
 752                         R1024_8_ROUNDS(13);
 753 #endif
 754 #if   R1024_UNROLL_R(14)
 755                         R1024_8_ROUNDS(14);
 756 #endif
 757                 }
 758                 /* do the final "feedforward" xor, update context chaining */
 759
 760                 ctx->x[0] = X00 ^ w[0];
 761                 ctx->x[1] = X01 ^ w[1];
 762                 ctx->x[2] = X02 ^ w[2];
 763                 ctx->x[3] = X03 ^ w[3];
 764                 ctx->x[4] = X04 ^ w[4];
 765                 ctx->x[5] = X05 ^ w[5];
 766                 ctx->x[6] = X06 ^ w[6];
 767                 ctx->x[7] = X07 ^ w[7];
 768                 ctx->x[8] = X08 ^ w[8];
 769                 ctx->x[9] = X09 ^ w[9];
 770                 ctx->x[10] = X10 ^ w[10];
 771                 ctx->x[11] = X11 ^ w[11];
 772                 ctx->x[12] = X12 ^ w[12];
 773                 ctx->x[13] = X13 ^ w[13];
 774                 ctx->x[14] = X14 ^ w[14];
 775                 ctx->x[15] = X15 ^ w[15];
 776
 777                 ts[1] &= ~SKEIN_T1_FLAG_FIRST;
 778                 blk_ptr += SKEIN_1024_BLOCK_BYTES;
 779         } while (--blk_cnt);
 780         ctx->h.tweak[0] = ts[0];
 781         ctx->h.tweak[1] = ts[1];
 782 }
 783
 784 #if defined(SKEIN_CODE_SIZE) || defined(SKEIN_PERF)
 785 size_t skein_1024_process_block_code_size(void)
 786 {
 787         return ((u8 *) skein_1024_process_block_code_size) -
 788                 ((u8 *) skein_1024_process_block);
 789 }
 790 unsigned int skein_1024_unroll_cnt(void)
 791 {
 792         return SKEIN_UNROLL_1024;
 793 }
 794 #endif
 795 #endif