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