release/src-rt/bcmcrypto/bn_lcl.h

   1 /*
   2  * bn_lcl.h: Big Number local header file.
   3  *
   4  * Code copied from openssl distribution and
   5  * Modified just enough so that compiles and runs standalone
   6  *
   7  * Copyright (C) 2010, Broadcom Corporation. All Rights Reserved.
   8  *
   9  * Permission to use, copy, modify, and/or distribute this software for any
  10  * purpose with or without fee is hereby granted, provided that the above
  11  * copyright notice and this permission notice appear in all copies.
  12  *
  13  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
  14  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
  15  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
  16  * SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
  17  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
  18  * OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
  19  * CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
  20  *
  21  * $Id: bn_lcl.h,v 1.3 2006-06-15 10:09:43 Exp $
  22  */
  23 /* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
  24  * All rights reserved.
  25  *
  26  * This package is an SSL implementation written
  27  * by Eric Young (eay@cryptsoft.com).
  28  * The implementation was written so as to conform with Netscapes SSL.
  29  *
  30  * This library is free for commercial and non-commercial use as long as
  31  * the following conditions are aheared to.  The following conditions
  32  * apply to all code found in this distribution, be it the RC4, RSA,
  33  * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
  34  * included with this distribution is covered by the same copyright terms
  35  * except that the holder is Tim Hudson (tjh@cryptsoft.com).
  36  *
  37  * Copyright remains Eric Young's, and as such any Copyright notices in
  38  * the code are not to be removed.
  39  * If this package is used in a product, Eric Young should be given attribution
  40  * as the author of the parts of the library used.
  41  * This can be in the form of a textual message at program startup or
  42  * in documentation (online or textual) provided with the package.
  43  *
  44  * Redistribution and use in source and binary forms, with or without
  45  * modification, are permitted provided that the following conditions
  46  * are met:
  47  * 1. Redistributions of source code must retain the copyright
  48  *    notice, this list of conditions and the following disclaimer.
  49  * 2. Redistributions in binary form must reproduce the above copyright
  50  *    notice, this list of conditions and the following disclaimer in the
  51  *    documentation and/or other materials provided with the distribution.
  52  * 3. All advertising materials mentioning features or use of this software
  53  *    must display the following acknowledgement:
  54  *    "This product includes cryptographic software written by
  55  *     Eric Young (eay@cryptsoft.com)"
  56  *    The word 'cryptographic' can be left out if the rouines from the library
  57  *    being used are not cryptographic related :-).
  58  * 4. If you include any Windows specific code (or a derivative thereof) from
  59  *    the apps directory (application code) you must include an acknowledgement:
  60  *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
  61  *
  62  * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
  63  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  64  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  65  * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
  66  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  67  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  68  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  69  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  70  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  71  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  72  * SUCH DAMAGE.
  73  *
  74  * The licence and distribution terms for any publically available version or
  75  * derivative of this code cannot be changed.  i.e. this code cannot simply be
  76  * copied and put under another distribution licence
  77  * [including the GNU Public Licence.]
  78  */
  79 /* ====================================================================
  80  * Copyright (c) 1998-2000 The OpenSSL Project.  All rights reserved.
  81  *
  82  * Redistribution and use in source and binary forms, with or without
  83  * modification, are permitted provided that the following conditions
  84  * are met:
  85  *
  86  * 1. Redistributions of source code must retain the above copyright
  87  *    notice, this list of conditions and the following disclaimer.
  88  *
  89  * 2. Redistributions in binary form must reproduce the above copyright
  90  *    notice, this list of conditions and the following disclaimer in
  91  *    the documentation and/or other materials provided with the
  92  *    distribution.
  93  *
  94  * 3. All advertising materials mentioning features or use of this
  95  *    software must display the following acknowledgment:
  96  *    "This product includes software developed by the OpenSSL Project
  97  *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
  98  *
  99  * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
 100  *    endorse or promote products derived from this software without
 101  *    prior written permission. For written permission, please contact
 102  *    openssl-core@openssl.org.
 103  *
 104  * 5. Products derived from this software may not be called "OpenSSL"
 105  *    nor may "OpenSSL" appear in their names without prior written
 106  *    permission of the OpenSSL Project.
 107  *
 108  * 6. Redistributions of any form whatsoever must retain the following
 109  *    acknowledgment:
 110  *    "This product includes software developed by the OpenSSL Project
 111  *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
 112  *
 113  * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
 114  * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 115  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 116  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
 117  * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 118  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 119  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 120  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 121  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
 122  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 123  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
 124  * OF THE POSSIBILITY OF SUCH DAMAGE.
 125  * ====================================================================
 126  *
 127  * This product includes cryptographic software written by Eric Young
 128  * (eay@cryptsoft.com).  This product includes software written by Tim
 129  * Hudson (tjh@cryptsoft.com).
 130  *
 131  */
 132
 133 #ifndef HEADER_BN_LCL_H
 134 #define HEADER_BN_LCL_H
 135
 136 #ifdef  __cplusplus
 137 extern "C" {
 138 #endif
 139
 140
 141 /* Used for temp variables */
 142 #define BN_CTX_NUM      32
 143 #define BN_CTX_NUM_POS  12
 144 struct bignum_ctx
 145         {
 146         int tos;
 147         BIGNUM bn[BN_CTX_NUM];
 148         int flags;
 149         int depth;
 150         int pos[BN_CTX_NUM_POS];
 151         int too_many;
 152         } /* BN_CTX */;
 153
 154
 155 /*
 156  * BN_window_bits_for_exponent_size -- macro for sliding window mod_exp functions
 157  *
 158  *
 159  * For window size 'w' (w >= 2) and a random 'b' bits exponent,
 160  * the number of multiplications is a constant plus on average
 161  *
 162  *    2^(w-1) + (b-w)/(w+1);
 163  *
 164  * here  2^(w-1)  is for precomputing the table (we actually need
 165  * entries only for windows that have the lowest bit set), and
 166  * (b-w)/(w+1)  is an approximation for the expected number of
 167  * w-bit windows, not counting the first one.
 168  *
 169  * Thus we should use
 170  *
 171  *    w >= 6  if        b > 671
 172  *     w = 5  if  671 > b > 239
 173  *     w = 4  if  239 > b >  79
 174  *     w = 3  if   79 > b >  23
 175  *    w <= 2  if   23 > b
 176  *
 177  * (with draws in between).  Very small exponents are often selected
 178  * with low Hamming weight, so we use  w = 1  for b <= 23.
 179  */
 180 #define BN_window_bits_for_exponent_size(b) \
 181                 ((b) > 671 ? 6 : \
 182                  (b) > 239 ? 5 : \
 183                  (b) >  79 ? 4 : \
 184                  (b) >  23 ? 3 : 1)
 185
 186
 187 /* Pentium pro 16, 16, 16, 32, 64 */
 188 /* Alpha       16, 16, 16, 16.64 */
 189 #define BN_MULL_SIZE_NORMAL                     (16) /* 32 */
 190 #define BN_MUL_RECURSIVE_SIZE_NORMAL            (16) /* 32 less than */
 191 #define BN_SQR_RECURSIVE_SIZE_NORMAL            (16) /* 32 */
 192 #define BN_MUL_LOW_RECURSIVE_SIZE_NORMAL        (32) /* 32 */
 193 #define BN_MONT_CTX_SET_SIZE_WORD               (64) /* 32 */
 194
 195 #if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC)
 196 /*
 197  * BN_UMULT_HIGH section.
 198  *
 199  * No, I'm not trying to overwhelm you when stating that the
 200  * product of N-bit numbers is 2*N bits wide:-) No, I don't expect
 201  * you to be impressed when I say that if the compiler doesn't
 202  * support 2*N integer type, then you have to replace every N*N
 203  * multiplication with 4 (N/2)*(N/2) accompanied by some shifts
 204  * and additions which unavoidably results in severe performance
 205  * penalties. Of course provided that the hardware is capable of
 206  * producing 2*N result... That's when you normally start
 207  * considering assembler implementation. However! It should be
 208  * pointed out that some CPUs (most notably Alpha, PowerPC and
 209  * upcoming IA-64 family:-) provide *separate* instruction
 210  * calculating the upper half of the product placing the result
 211  * into a general purpose register. Now *if* the compiler supports
 212  * inline assembler, then it's not impossible to implement the
 213  * "bignum" routines (and have the compiler optimize 'em)
 214  * exhibiting "native" performance in C. That's what BN_UMULT_HIGH
 215  * macro is about:-)
 216  *
 217  *                                      <appro@fy.chalmers.se>
 218  */
 219 # if defined(__alpha) && (defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT))
 220 #  if defined(__DECC)
 221 #   include <c_asm.h>
 222 #   define BN_UMULT_HIGH(a, b)  (BN_ULONG)asm("umulh %a0, %a1, %v0", (a), (b))
 223 #  elif defined(__GNUC__)
 224 #   define BN_UMULT_HIGH(a, b)  ({      \
 225         register BN_ULONG ret;          \
 226         asm("umulh      %1, %2, %0"     \
 227                 : "=r"(ret)             \
 228                 : "r"(a), "r"(b));      \
 229         ret;                    })
 230 #  endif        /* compiler */
 231 # elif defined(_ARCH_PPC) && defined(__64BIT__) && defined(SIXTY_FOUR_BIT_LONG)
 232 #  if defined(__GNUC__)
 233 #   define BN_UMULT_HIGH(a, b)  ({      \
 234         register BN_ULONG ret;          \
 235         asm("mulhdu     %0, %1, %2"     \
 236              : "=r"(ret)                \
 237              : "r"(a), "r"(b));         \
 238         ret;                    })
 239 #  endif        /* compiler */
 240 # elif defined(__x86_64) && defined(SIXTY_FOUR_BIT_LONG)
 241 #  if defined(__GNUC__)
 242 #   define BN_UMULT_HIGH(a, b)  ({      \
 243         register BN_ULONG ret, discard; \
 244         asm("mulq       %3"             \
 245              : "=a"(discard), "=d"(ret) \
 246              : "a"(a), "g"(b)           \
 247              : "cc");                   \
 248         ret;                    })
 249 #   define BN_UMULT_LOHI(low, high, a, b) \
 250         asm("mulq       %3"             \
 251                 : "=a"(low), "=d"(high) \
 252                 : "a"(a), "g"(b)        \
 253                 : "cc");
 254 #  endif
 255 # endif         /* cpu */
 256 #endif          /* OPENSSL_NO_ASM */
 257
 258 /*************************************************************
 259  * Using the long long type
 260  */
 261 #define Lw(t)    (((BN_ULONG)(t))&BN_MASK2)
 262 #define Hw(t)    (((BN_ULONG)((t)>>BN_BITS2))&BN_MASK2)
 263
 264 /* This is used for internal error checking and is not normally used */
 265 #ifdef BN_DEBUG
 266 # include <assert.h>
 267 # define bn_check_top(a) assert ((a)->top >= 0 && (a)->top <= (a)->dmax);
 268 #else
 269 # define bn_check_top(a)
 270 #endif
 271
 272 /* This macro is to add extra stuff for development checking */
 273 #ifdef BN_DEBUG
 274 #define bn_set_max(r) ((r)->max = (r)->top, BN_set_flags((r), BN_FLG_STATIC_DATA))
 275 #else
 276 #define bn_set_max(r)
 277 #endif
 278
 279 /* These macros are used to 'take' a section of a bignum for read only use */
 280 #define bn_set_low(r, a, n) { \
 281         (r)->top = ((a)->top > (n))?(n):(a)->top; \
 282         (r)->d = (a)->d; \
 283         (r)->neg = (a)->neg; \
 284         (r)->flags |= BN_FLG_STATIC_DATA; \
 285         bn_set_max(r); \
 286 }
 287
 288 #define bn_set_high(r, a, n) { \
 289         if ((a)->top > (n)) { \
 290                 (r)->top = (a)->top-n; \
 291                 (r)->d = &((a)->d[n]); \
 292         } else \
 293                 (r)->top = 0; \
 294         (r)->neg = (a)->neg; \
 295         (r)->flags |= BN_FLG_STATIC_DATA; \
 296         bn_set_max(r); \
 297 }
 298
 299 #ifdef BN_LLONG
 300 #define mul_add(r, a, w, c) { \
 301         BN_ULLONG t; \
 302         t = (BN_ULLONG)w * (a) + (r) + (c); \
 303         (r) = Lw(t); \
 304         (c) = Hw(t); \
 305 }
 306
 307 #define mul(r, a, w, c) { \
 308         BN_ULLONG t; \
 309         t = (BN_ULLONG)w * (a) + (c); \
 310         (r) = Lw(t); \
 311         (c) = Hw(t); \
 312 }
 313
 314 #define sqr(r0, r1, a) { \
 315         BN_ULLONG t; \
 316         t = (BN_ULLONG)(a)*(a); \
 317         (r0) = Lw(t); \
 318         (r1) = Hw(t); \
 319 }
 320
 321 #elif defined(BN_UMULT_HIGH)
 322 #define mul_add(r, a, w, c) {           \
 323         BN_ULONG high, low, ret, tmp = (a); \
 324         ret = (r);                      \
 325         high = BN_UMULT_HIGH(w, tmp);   \
 326         ret += (c);                     \
 327         low = (w) * tmp;                \
 328         (c) = (ret < (c)) ? 1 : 0;      \
 329         (c) += high;                    \
 330         ret += low;                     \
 331         (c) += (ret < low) ? 1 : 0;     \
 332         (r) = ret;                      \
 333 }
 334
 335 #define mul(r, a, w, c) {               \
 336         BN_ULONG high, low, ret, ta = (a); \
 337         low = (w) * ta;                 \
 338         high = BN_UMULT_HIGH(w, ta);    \
 339         ret = low + (c);                \
 340         (c) = high;                     \
 341         (c) += (ret < low) ? 1 : 0;     \
 342         (r) =  ret;                     \
 343 }
 344
 345 #define sqr(r0, r1, a)  {               \
 346         BN_ULONG tmp = (a);             \
 347         (r0) = tmp * tmp;               \
 348         (r1) = BN_UMULT_HIGH(tmp, tmp); \
 349 }
 350
 351 #else
 352 /* No long long type */
 353
 354 #define LBITS(a)        ((a) & BN_MASK2l)
 355 #define HBITS(a)        (((a) >> BN_BITS4) & BN_MASK2l)
 356 #define L2HBITS(a)      (((a) << BN_BITS4) & BN_MASK2)
 357
 358 #define LLBITS(a)       ((a) & BN_MASKl)
 359 #define LHBITS(a)       (((a) >> BN_BITS2) & BN_MASKl)
 360 #define LL2HBITS(a)     ((BN_ULLONG)((a) & BN_MASKl) << BN_BITS2)
 361
 362 #define mul64(l, h, bl, bh) { \
 363         BN_ULONG m, m1, lt, ht; \
 364         lt = l; \
 365         ht = h; \
 366         m = (bh) * (lt); \
 367         lt = (bl) * (lt); \
 368         m1 = (bl) * (ht); \
 369         ht = (bh) * (ht); \
 370         m = (m + m1) & BN_MASK2; \
 371         if (m < m1) \
 372                 ht += L2HBITS((BN_ULONG)1); \
 373         ht += HBITS(m); \
 374         m1 = L2HBITS(m); \
 375         lt = (lt + m1) & BN_MASK2; \
 376         if (lt < m1) \
 377                 ht++; \
 378         (l) = lt; \
 379         (h) = ht; \
 380 }
 381
 382 #define sqr64(lo, ho, in) { \
 383         BN_ULONG l, h, m; \
 384         h = (in); \
 385         l = LBITS(h); \
 386         h = HBITS(h); \
 387         m = (l) * (h); \
 388         l *= l; \
 389         h *= h; \
 390         h += (m & BN_MASK2h1) >> (BN_BITS4 - 1); \
 391         m = (m & BN_MASK2l) << (BN_BITS4 + 1); \
 392         l = (l + m) & BN_MASK2; \
 393         if (l < m) \
 394                 h++; \
 395         (lo) = l; \
 396         (ho) = h; \
 397 }
 398
 399 #define mul_add(r, a, bl, bh, c) { \
 400         BN_ULONG l, h; \
 401         h = (a); \
 402         l = LBITS(h); \
 403         h = HBITS(h); \
 404         mul64(l, h, (bl), (bh)); \
 405         /* non-multiply part */ \
 406         l = (l + (c)) & BN_MASK2; \
 407         if (l < (c)) \
 408                 h++; \
 409         (c) = (r); \
 410         l = (l + (c)) & BN_MASK2; \
 411         if (l < (c)) \
 412                 h++; \
 413         (c) = h & BN_MASK2; \
 414         (r) = l; \
 415 }
 416
 417 #define mul(r, a, bl, bh, c) { \
 418         BN_ULONG l, h; \
 419         h = (a); \
 420         l = LBITS(h); \
 421         h = HBITS(h); \
 422         mul64(l, h, (bl), (bh)); \
 423         /* non-multiply part */ \
 424         l += (c); \
 425         if ((l & BN_MASK2) < (c)) \
 426                 h++; \
 427         (c) = h & BN_MASK2; \
 428         (r) = l & BN_MASK2; \
 429 }
 430 #endif /* !BN_LLONG */
 431
 432 void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb);
 433 void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
 434 void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b);
 435 void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp);
 436 void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a);
 437 void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a);
 438 int bn_cmp_words(const BN_ULONG *a, const BN_ULONG *b, int n);
 439 int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b, int cl, int dl);
 440 void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t);
 441 void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n);
 442 void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2,
 443         BN_ULONG *t);
 444 void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2,
 445         BN_ULONG *t);
 446
 447 #ifdef  __cplusplus
 448 }
 449 #endif
 450
 451 #endif  /* HEADER_BN_LCL_H */