| 1 |
/* crypto/bn/bn_lcl.h */ |
| 2 |
/* Copyright (C) 1995-1998 Eric Young ([email protected]) |
| 3 |
* All rights reserved. |
| 4 |
* |
| 5 |
* This package is an SSL implementation written |
| 6 |
* by Eric Young ([email protected]). |
| 7 |
* The implementation was written so as to conform with Netscapes SSL. |
| 8 |
* |
| 9 |
* This library is free for commercial and non-commercial use as long as |
| 10 |
* the following conditions are aheared to. The following conditions |
| 11 |
* apply to all code found in this distribution, be it the RC4, RSA, |
| 12 |
* lhash, DES, etc., code; not just the SSL code. The SSL documentation |
| 13 |
* included with this distribution is covered by the same copyright terms |
| 14 |
* except that the holder is Tim Hudson ([email protected]). |
| 15 |
* |
| 16 |
* Copyright remains Eric Young's, and as such any Copyright notices in |
| 17 |
* the code are not to be removed. |
| 18 |
* If this package is used in a product, Eric Young should be given attribution |
| 19 |
* as the author of the parts of the library used. |
| 20 |
* This can be in the form of a textual message at program startup or |
| 21 |
* in documentation (online or textual) provided with the package. |
| 22 |
* |
| 23 |
* Redistribution and use in source and binary forms, with or without |
| 24 |
* modification, are permitted provided that the following conditions |
| 25 |
* are met: |
| 26 |
* 1. Redistributions of source code must retain the copyright |
| 27 |
* notice, this list of conditions and the following disclaimer. |
| 28 |
* 2. Redistributions in binary form must reproduce the above copyright |
| 29 |
* notice, this list of conditions and the following disclaimer in the |
| 30 |
* documentation and/or other materials provided with the distribution. |
| 31 |
* 3. All advertising materials mentioning features or use of this software |
| 32 |
* must display the following acknowledgement: |
| 33 |
* "This product includes cryptographic software written by |
| 34 |
* Eric Young ([email protected])" |
| 35 |
* The word 'cryptographic' can be left out if the rouines from the library |
| 36 |
* being used are not cryptographic related :-). |
| 37 |
* 4. If you include any Windows specific code (or a derivative thereof) from |
| 38 |
* the apps directory (application code) you must include an acknowledgement: |
| 39 |
* "This product includes software written by Tim Hudson ([email protected])" |
| 40 |
* |
| 41 |
* THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND |
| 42 |
* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 43 |
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE |
| 44 |
* ARE DISCLAIMED. IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE |
| 45 |
* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL |
| 46 |
* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS |
| 47 |
* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 48 |
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT |
| 49 |
* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY |
| 50 |
* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF |
| 51 |
* SUCH DAMAGE. |
| 52 |
* |
| 53 |
* The licence and distribution terms for any publically available version or |
| 54 |
* derivative of this code cannot be changed. i.e. this code cannot simply be |
| 55 |
* copied and put under another distribution licence |
| 56 |
* [including the GNU Public Licence.] |
| 57 |
*/ |
| 58 |
/* ==================================================================== |
| 59 |
* Copyright (c) 1998-2000 The OpenSSL Project. All rights reserved. |
| 60 |
* |
| 61 |
* Redistribution and use in source and binary forms, with or without |
| 62 |
* modification, are permitted provided that the following conditions |
| 63 |
* are met: |
| 64 |
* |
| 65 |
* 1. Redistributions of source code must retain the above copyright |
| 66 |
* notice, this list of conditions and the following disclaimer. |
| 67 |
* |
| 68 |
* 2. Redistributions in binary form must reproduce the above copyright |
| 69 |
* notice, this list of conditions and the following disclaimer in |
| 70 |
* the documentation and/or other materials provided with the |
| 71 |
* distribution. |
| 72 |
* |
| 73 |
* 3. All advertising materials mentioning features or use of this |
| 74 |
* software must display the following acknowledgment: |
| 75 |
* "This product includes software developed by the OpenSSL Project |
| 76 |
* for use in the OpenSSL Toolkit. (http://www.openssl.org/)" |
| 77 |
* |
| 78 |
* 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to |
| 79 |
* endorse or promote products derived from this software without |
| 80 |
* prior written permission. For written permission, please contact |
| 81 |
* [email protected]. |
| 82 |
* |
| 83 |
* 5. Products derived from this software may not be called "OpenSSL" |
| 84 |
* nor may "OpenSSL" appear in their names without prior written |
| 85 |
* permission of the OpenSSL Project. |
| 86 |
* |
| 87 |
* 6. Redistributions of any form whatsoever must retain the following |
| 88 |
* acknowledgment: |
| 89 |
* "This product includes software developed by the OpenSSL Project |
| 90 |
* for use in the OpenSSL Toolkit (http://www.openssl.org/)" |
| 91 |
* |
| 92 |
* THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY |
| 93 |
* EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE |
| 94 |
* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR |
| 95 |
* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR |
| 96 |
* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, |
| 97 |
* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT |
| 98 |
* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; |
| 99 |
* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) |
| 100 |
* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, |
| 101 |
* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) |
| 102 |
* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED |
| 103 |
* OF THE POSSIBILITY OF SUCH DAMAGE. |
| 104 |
* ==================================================================== |
| 105 |
* |
| 106 |
* This product includes cryptographic software written by Eric Young |
| 107 |
* ([email protected]). This product includes software written by Tim |
| 108 |
* Hudson ([email protected]). |
| 109 |
* |
| 110 |
*/ |
| 111 |
|
| 112 |
#ifndef HEADER_BN_LCL_H |
| 113 |
# define HEADER_BN_LCL_H |
| 114 |
|
| 115 |
# include <openssl/bn.h> |
| 116 |
|
| 117 |
#ifdef __cplusplus |
| 118 |
extern "C" { |
| 119 |
#endif |
| 120 |
|
| 121 |
/*- |
| 122 |
* BN_window_bits_for_exponent_size -- macro for sliding window mod_exp functions |
| 123 |
* |
| 124 |
* |
| 125 |
* For window size 'w' (w >= 2) and a random 'b' bits exponent, |
| 126 |
* the number of multiplications is a constant plus on average |
| 127 |
* |
| 128 |
* 2^(w-1) + (b-w)/(w+1); |
| 129 |
* |
| 130 |
* here 2^(w-1) is for precomputing the table (we actually need |
| 131 |
* entries only for windows that have the lowest bit set), and |
| 132 |
* (b-w)/(w+1) is an approximation for the expected number of |
| 133 |
* w-bit windows, not counting the first one. |
| 134 |
* |
| 135 |
* Thus we should use |
| 136 |
* |
| 137 |
* w >= 6 if b > 671 |
| 138 |
* w = 5 if 671 > b > 239 |
| 139 |
* w = 4 if 239 > b > 79 |
| 140 |
* w = 3 if 79 > b > 23 |
| 141 |
* w <= 2 if 23 > b |
| 142 |
* |
| 143 |
* (with draws in between). Very small exponents are often selected |
| 144 |
* with low Hamming weight, so we use w = 1 for b <= 23. |
| 145 |
*/ |
| 146 |
# if 1 |
| 147 |
# define BN_window_bits_for_exponent_size(b) \ |
| 148 |
((b) > 671 ? 6 : \ |
| 149 |
(b) > 239 ? 5 : \ |
| 150 |
(b) > 79 ? 4 : \ |
| 151 |
(b) > 23 ? 3 : 1) |
| 152 |
# else |
| 153 |
/* |
| 154 |
* Old SSLeay/OpenSSL table. Maximum window size was 5, so this table differs |
| 155 |
* for b==1024; but it coincides for other interesting values (b==160, |
| 156 |
* b==512). |
| 157 |
*/ |
| 158 |
# define BN_window_bits_for_exponent_size(b) \ |
| 159 |
((b) > 255 ? 5 : \ |
| 160 |
(b) > 127 ? 4 : \ |
| 161 |
(b) > 17 ? 3 : 1) |
| 162 |
# endif |
| 163 |
|
| 164 |
/* |
| 165 |
* BN_mod_exp_mont_conttime is based on the assumption that the L1 data cache |
| 166 |
* line width of the target processor is at least the following value. |
| 167 |
*/ |
| 168 |
# define MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH ( 64 ) |
| 169 |
# define MOD_EXP_CTIME_MIN_CACHE_LINE_MASK (MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH - 1) |
| 170 |
|
| 171 |
/* |
| 172 |
* Window sizes optimized for fixed window size modular exponentiation |
| 173 |
* algorithm (BN_mod_exp_mont_consttime). To achieve the security goals of |
| 174 |
* BN_mode_exp_mont_consttime, the maximum size of the window must not exceed |
| 175 |
* log_2(MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH). Window size thresholds are |
| 176 |
* defined for cache line sizes of 32 and 64, cache line sizes where |
| 177 |
* log_2(32)=5 and log_2(64)=6 respectively. A window size of 7 should only be |
| 178 |
* used on processors that have a 128 byte or greater cache line size. |
| 179 |
*/ |
| 180 |
# if MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 64 |
| 181 |
|
| 182 |
# define BN_window_bits_for_ctime_exponent_size(b) \ |
| 183 |
((b) > 937 ? 6 : \ |
| 184 |
(b) > 306 ? 5 : \ |
| 185 |
(b) > 89 ? 4 : \ |
| 186 |
(b) > 22 ? 3 : 1) |
| 187 |
# define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (6) |
| 188 |
|
| 189 |
# elif MOD_EXP_CTIME_MIN_CACHE_LINE_WIDTH == 32 |
| 190 |
|
| 191 |
# define BN_window_bits_for_ctime_exponent_size(b) \ |
| 192 |
((b) > 306 ? 5 : \ |
| 193 |
(b) > 89 ? 4 : \ |
| 194 |
(b) > 22 ? 3 : 1) |
| 195 |
# define BN_MAX_WINDOW_BITS_FOR_CTIME_EXPONENT_SIZE (5) |
| 196 |
|
| 197 |
# endif |
| 198 |
|
| 199 |
/* Pentium pro 16,16,16,32,64 */ |
| 200 |
/* Alpha 16,16,16,16.64 */ |
| 201 |
# define BN_MULL_SIZE_NORMAL (16)/* 32 */ |
| 202 |
# define BN_MUL_RECURSIVE_SIZE_NORMAL (16)/* 32 less than */ |
| 203 |
# define BN_SQR_RECURSIVE_SIZE_NORMAL (16)/* 32 */ |
| 204 |
# define BN_MUL_LOW_RECURSIVE_SIZE_NORMAL (32)/* 32 */ |
| 205 |
# define BN_MONT_CTX_SET_SIZE_WORD (64)/* 32 */ |
| 206 |
|
| 207 |
/* |
| 208 |
* 2011-02-22 SMS. In various places, a size_t variable or a type cast to |
| 209 |
* size_t was used to perform integer-only operations on pointers. This |
| 210 |
* failed on VMS with 64-bit pointers (CC /POINTER_SIZE = 64) because size_t |
| 211 |
* is still only 32 bits. What's needed in these cases is an integer type |
| 212 |
* with the same size as a pointer, which size_t is not certain to be. The |
| 213 |
* only fix here is VMS-specific. |
| 214 |
*/ |
| 215 |
# if defined(OPENSSL_SYS_VMS) |
| 216 |
# if __INITIAL_POINTER_SIZE == 64 |
| 217 |
# define PTR_SIZE_INT long long |
| 218 |
# else /* __INITIAL_POINTER_SIZE == 64 */ |
| 219 |
# define PTR_SIZE_INT int |
| 220 |
# endif /* __INITIAL_POINTER_SIZE == 64 [else] */ |
| 221 |
# elif !defined(PTR_SIZE_INT) /* defined(OPENSSL_SYS_VMS) */ |
| 222 |
# define PTR_SIZE_INT size_t |
| 223 |
# endif /* defined(OPENSSL_SYS_VMS) [else] */ |
| 224 |
|
| 225 |
# if !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC) |
| 226 |
/* |
| 227 |
* BN_UMULT_HIGH section. |
| 228 |
* |
| 229 |
* No, I'm not trying to overwhelm you when stating that the |
| 230 |
* product of N-bit numbers is 2*N bits wide:-) No, I don't expect |
| 231 |
* you to be impressed when I say that if the compiler doesn't |
| 232 |
* support 2*N integer type, then you have to replace every N*N |
| 233 |
* multiplication with 4 (N/2)*(N/2) accompanied by some shifts |
| 234 |
* and additions which unavoidably results in severe performance |
| 235 |
* penalties. Of course provided that the hardware is capable of |
| 236 |
* producing 2*N result... That's when you normally start |
| 237 |
* considering assembler implementation. However! It should be |
| 238 |
* pointed out that some CPUs (most notably Alpha, PowerPC and |
| 239 |
* upcoming IA-64 family:-) provide *separate* instruction |
| 240 |
* calculating the upper half of the product placing the result |
| 241 |
* into a general purpose register. Now *if* the compiler supports |
| 242 |
* inline assembler, then it's not impossible to implement the |
| 243 |
* "bignum" routines (and have the compiler optimize 'em) |
| 244 |
* exhibiting "native" performance in C. That's what BN_UMULT_HIGH |
| 245 |
* macro is about:-) |
| 246 |
* |
| 247 |
* <[email protected]> |
| 248 |
*/ |
| 249 |
# if defined(__alpha) && (defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT)) |
| 250 |
# if defined(__DECC) |
| 251 |
# include <c_asm.h> |
| 252 |
# define BN_UMULT_HIGH(a,b) (BN_ULONG)asm("umulh %a0,%a1,%v0",(a),(b)) |
| 253 |
# elif defined(__GNUC__) && __GNUC__>=2 |
| 254 |
# define BN_UMULT_HIGH(a,b) ({ \ |
| 255 |
register BN_ULONG ret; \ |
| 256 |
asm ("umulh %1,%2,%0" \ |
| 257 |
: "=r"(ret) \ |
| 258 |
: "r"(a), "r"(b)); \ |
| 259 |
ret; }) |
| 260 |
# endif /* compiler */ |
| 261 |
# elif defined(_ARCH_PPC) && defined(__64BIT__) && defined(SIXTY_FOUR_BIT_LONG) |
| 262 |
# if defined(__GNUC__) && __GNUC__>=2 |
| 263 |
# define BN_UMULT_HIGH(a,b) ({ \ |
| 264 |
register BN_ULONG ret; \ |
| 265 |
asm ("mulhdu %0,%1,%2" \ |
| 266 |
: "=r"(ret) \ |
| 267 |
: "r"(a), "r"(b)); \ |
| 268 |
ret; }) |
| 269 |
# endif /* compiler */ |
| 270 |
# elif (defined(__x86_64) || defined(__x86_64__)) && \ |
| 271 |
(defined(SIXTY_FOUR_BIT_LONG) || defined(SIXTY_FOUR_BIT)) |
| 272 |
# if defined(__GNUC__) && __GNUC__>=2 |
| 273 |
# define BN_UMULT_HIGH(a,b) ({ \ |
| 274 |
register BN_ULONG ret,discard; \ |
| 275 |
asm ("mulq %3" \ |
| 276 |
: "=a"(discard),"=d"(ret) \ |
| 277 |
: "a"(a), "g"(b) \ |
| 278 |
: "cc"); \ |
| 279 |
ret; }) |
| 280 |
# define BN_UMULT_LOHI(low,high,a,b) \ |
| 281 |
asm ("mulq %3" \ |
| 282 |
: "=a"(low),"=d"(high) \ |
| 283 |
: "a"(a),"g"(b) \ |
| 284 |
: "cc"); |
| 285 |
# endif |
| 286 |
# elif (defined(_M_AMD64) || defined(_M_X64)) && defined(SIXTY_FOUR_BIT) |
| 287 |
# if defined(_MSC_VER) && _MSC_VER>=1400 |
| 288 |
unsigned __int64 __umulh(unsigned __int64 a, unsigned __int64 b); |
| 289 |
unsigned __int64 _umul128(unsigned __int64 a, unsigned __int64 b, |
| 290 |
unsigned __int64 *h); |
| 291 |
# pragma intrinsic(__umulh,_umul128) |
| 292 |
# define BN_UMULT_HIGH(a,b) __umulh((a),(b)) |
| 293 |
# define BN_UMULT_LOHI(low,high,a,b) ((low)=_umul128((a),(b),&(high))) |
| 294 |
# endif |
| 295 |
# elif defined(__mips) && (defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG)) |
| 296 |
# if defined(__GNUC__) && __GNUC__>=2 |
| 297 |
# if __GNUC__>4 || (__GNUC__>=4 && __GNUC_MINOR__>=4) |
| 298 |
/* "h" constraint is no more since 4.4 */ |
| 299 |
# define BN_UMULT_HIGH(a,b) (((__uint128_t)(a)*(b))>>64) |
| 300 |
# define BN_UMULT_LOHI(low,high,a,b) ({ \ |
| 301 |
__uint128_t ret=(__uint128_t)(a)*(b); \ |
| 302 |
(high)=ret>>64; (low)=ret; }) |
| 303 |
# else |
| 304 |
# define BN_UMULT_HIGH(a,b) ({ \ |
| 305 |
register BN_ULONG ret; \ |
| 306 |
asm ("dmultu %1,%2" \ |
| 307 |
: "=h"(ret) \ |
| 308 |
: "r"(a), "r"(b) : "l"); \ |
| 309 |
ret; }) |
| 310 |
# define BN_UMULT_LOHI(low,high,a,b)\ |
| 311 |
asm ("dmultu %2,%3" \ |
| 312 |
: "=l"(low),"=h"(high) \ |
| 313 |
: "r"(a), "r"(b)); |
| 314 |
# endif |
| 315 |
# endif |
| 316 |
# elif defined(__aarch64__) && defined(SIXTY_FOUR_BIT_LONG) |
| 317 |
# if defined(__GNUC__) && __GNUC__>=2 |
| 318 |
# define BN_UMULT_HIGH(a,b) ({ \ |
| 319 |
register BN_ULONG ret; \ |
| 320 |
asm ("umulh %0,%1,%2" \ |
| 321 |
: "=r"(ret) \ |
| 322 |
: "r"(a), "r"(b)); \ |
| 323 |
ret; }) |
| 324 |
# endif |
| 325 |
# endif /* cpu */ |
| 326 |
# endif /* OPENSSL_NO_ASM */ |
| 327 |
|
| 328 |
/************************************************************* |
| 329 |
* Using the long long type |
| 330 |
*/ |
| 331 |
# define Lw(t) (((BN_ULONG)(t))&BN_MASK2) |
| 332 |
# define Hw(t) (((BN_ULONG)((t)>>BN_BITS2))&BN_MASK2) |
| 333 |
|
| 334 |
# ifdef BN_DEBUG_RAND |
| 335 |
# define bn_clear_top2max(a) \ |
| 336 |
{ \ |
| 337 |
int ind = (a)->dmax - (a)->top; \ |
| 338 |
BN_ULONG *ftl = &(a)->d[(a)->top-1]; \ |
| 339 |
for (; ind != 0; ind--) \ |
| 340 |
*(++ftl) = 0x0; \ |
| 341 |
} |
| 342 |
# else |
| 343 |
# define bn_clear_top2max(a) |
| 344 |
# endif |
| 345 |
|
| 346 |
# ifdef BN_LLONG |
| 347 |
# define mul_add(r,a,w,c) { \ |
| 348 |
BN_ULLONG t; \ |
| 349 |
t=(BN_ULLONG)w * (a) + (r) + (c); \ |
| 350 |
(r)= Lw(t); \ |
| 351 |
(c)= Hw(t); \ |
| 352 |
} |
| 353 |
|
| 354 |
# define mul(r,a,w,c) { \ |
| 355 |
BN_ULLONG t; \ |
| 356 |
t=(BN_ULLONG)w * (a) + (c); \ |
| 357 |
(r)= Lw(t); \ |
| 358 |
(c)= Hw(t); \ |
| 359 |
} |
| 360 |
|
| 361 |
# define sqr(r0,r1,a) { \ |
| 362 |
BN_ULLONG t; \ |
| 363 |
t=(BN_ULLONG)(a)*(a); \ |
| 364 |
(r0)=Lw(t); \ |
| 365 |
(r1)=Hw(t); \ |
| 366 |
} |
| 367 |
|
| 368 |
# elif defined(BN_UMULT_LOHI) |
| 369 |
# define mul_add(r,a,w,c) { \ |
| 370 |
BN_ULONG high,low,ret,tmp=(a); \ |
| 371 |
ret = (r); \ |
| 372 |
BN_UMULT_LOHI(low,high,w,tmp); \ |
| 373 |
ret += (c); \ |
| 374 |
(c) = (ret<(c))?1:0; \ |
| 375 |
(c) += high; \ |
| 376 |
ret += low; \ |
| 377 |
(c) += (ret<low)?1:0; \ |
| 378 |
(r) = ret; \ |
| 379 |
} |
| 380 |
|
| 381 |
# define mul(r,a,w,c) { \ |
| 382 |
BN_ULONG high,low,ret,ta=(a); \ |
| 383 |
BN_UMULT_LOHI(low,high,w,ta); \ |
| 384 |
ret = low + (c); \ |
| 385 |
(c) = high; \ |
| 386 |
(c) += (ret<low)?1:0; \ |
| 387 |
(r) = ret; \ |
| 388 |
} |
| 389 |
|
| 390 |
# define sqr(r0,r1,a) { \ |
| 391 |
BN_ULONG tmp=(a); \ |
| 392 |
BN_UMULT_LOHI(r0,r1,tmp,tmp); \ |
| 393 |
} |
| 394 |
|
| 395 |
# elif defined(BN_UMULT_HIGH) |
| 396 |
# define mul_add(r,a,w,c) { \ |
| 397 |
BN_ULONG high,low,ret,tmp=(a); \ |
| 398 |
ret = (r); \ |
| 399 |
high= BN_UMULT_HIGH(w,tmp); \ |
| 400 |
ret += (c); \ |
| 401 |
low = (w) * tmp; \ |
| 402 |
(c) = (ret<(c))?1:0; \ |
| 403 |
(c) += high; \ |
| 404 |
ret += low; \ |
| 405 |
(c) += (ret<low)?1:0; \ |
| 406 |
(r) = ret; \ |
| 407 |
} |
| 408 |
|
| 409 |
# define mul(r,a,w,c) { \ |
| 410 |
BN_ULONG high,low,ret,ta=(a); \ |
| 411 |
low = (w) * ta; \ |
| 412 |
high= BN_UMULT_HIGH(w,ta); \ |
| 413 |
ret = low + (c); \ |
| 414 |
(c) = high; \ |
| 415 |
(c) += (ret<low)?1:0; \ |
| 416 |
(r) = ret; \ |
| 417 |
} |
| 418 |
|
| 419 |
# define sqr(r0,r1,a) { \ |
| 420 |
BN_ULONG tmp=(a); \ |
| 421 |
(r0) = tmp * tmp; \ |
| 422 |
(r1) = BN_UMULT_HIGH(tmp,tmp); \ |
| 423 |
} |
| 424 |
|
| 425 |
# else |
| 426 |
/************************************************************* |
| 427 |
* No long long type |
| 428 |
*/ |
| 429 |
|
| 430 |
# define LBITS(a) ((a)&BN_MASK2l) |
| 431 |
# define HBITS(a) (((a)>>BN_BITS4)&BN_MASK2l) |
| 432 |
# define L2HBITS(a) (((a)<<BN_BITS4)&BN_MASK2) |
| 433 |
|
| 434 |
# define LLBITS(a) ((a)&BN_MASKl) |
| 435 |
# define LHBITS(a) (((a)>>BN_BITS2)&BN_MASKl) |
| 436 |
# define LL2HBITS(a) ((BN_ULLONG)((a)&BN_MASKl)<<BN_BITS2) |
| 437 |
|
| 438 |
# define mul64(l,h,bl,bh) \ |
| 439 |
{ \ |
| 440 |
BN_ULONG m,m1,lt,ht; \ |
| 441 |
\ |
| 442 |
lt=l; \ |
| 443 |
ht=h; \ |
| 444 |
m =(bh)*(lt); \ |
| 445 |
lt=(bl)*(lt); \ |
| 446 |
m1=(bl)*(ht); \ |
| 447 |
ht =(bh)*(ht); \ |
| 448 |
m=(m+m1)&BN_MASK2; if (m < m1) ht+=L2HBITS((BN_ULONG)1); \ |
| 449 |
ht+=HBITS(m); \ |
| 450 |
m1=L2HBITS(m); \ |
| 451 |
lt=(lt+m1)&BN_MASK2; if (lt < m1) ht++; \ |
| 452 |
(l)=lt; \ |
| 453 |
(h)=ht; \ |
| 454 |
} |
| 455 |
|
| 456 |
# define sqr64(lo,ho,in) \ |
| 457 |
{ \ |
| 458 |
BN_ULONG l,h,m; \ |
| 459 |
\ |
| 460 |
h=(in); \ |
| 461 |
l=LBITS(h); \ |
| 462 |
h=HBITS(h); \ |
| 463 |
m =(l)*(h); \ |
| 464 |
l*=l; \ |
| 465 |
h*=h; \ |
| 466 |
h+=(m&BN_MASK2h1)>>(BN_BITS4-1); \ |
| 467 |
m =(m&BN_MASK2l)<<(BN_BITS4+1); \ |
| 468 |
l=(l+m)&BN_MASK2; if (l < m) h++; \ |
| 469 |
(lo)=l; \ |
| 470 |
(ho)=h; \ |
| 471 |
} |
| 472 |
|
| 473 |
# define mul_add(r,a,bl,bh,c) { \ |
| 474 |
BN_ULONG l,h; \ |
| 475 |
\ |
| 476 |
h= (a); \ |
| 477 |
l=LBITS(h); \ |
| 478 |
h=HBITS(h); \ |
| 479 |
mul64(l,h,(bl),(bh)); \ |
| 480 |
\ |
| 481 |
/* non-multiply part */ \ |
| 482 |
l=(l+(c))&BN_MASK2; if (l < (c)) h++; \ |
| 483 |
(c)=(r); \ |
| 484 |
l=(l+(c))&BN_MASK2; if (l < (c)) h++; \ |
| 485 |
(c)=h&BN_MASK2; \ |
| 486 |
(r)=l; \ |
| 487 |
} |
| 488 |
|
| 489 |
# define mul(r,a,bl,bh,c) { \ |
| 490 |
BN_ULONG l,h; \ |
| 491 |
\ |
| 492 |
h= (a); \ |
| 493 |
l=LBITS(h); \ |
| 494 |
h=HBITS(h); \ |
| 495 |
mul64(l,h,(bl),(bh)); \ |
| 496 |
\ |
| 497 |
/* non-multiply part */ \ |
| 498 |
l+=(c); if ((l&BN_MASK2) < (c)) h++; \ |
| 499 |
(c)=h&BN_MASK2; \ |
| 500 |
(r)=l&BN_MASK2; \ |
| 501 |
} |
| 502 |
# endif /* !BN_LLONG */ |
| 503 |
|
| 504 |
# if defined(OPENSSL_DOING_MAKEDEPEND) && defined(OPENSSL_FIPS) |
| 505 |
# undef bn_div_words |
| 506 |
# endif |
| 507 |
|
| 508 |
void bn_mul_normal(BN_ULONG *r, BN_ULONG *a, int na, BN_ULONG *b, int nb); |
| 509 |
void bn_mul_comba8(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b); |
| 510 |
void bn_mul_comba4(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b); |
| 511 |
void bn_sqr_normal(BN_ULONG *r, const BN_ULONG *a, int n, BN_ULONG *tmp); |
| 512 |
void bn_sqr_comba8(BN_ULONG *r, const BN_ULONG *a); |
| 513 |
void bn_sqr_comba4(BN_ULONG *r, const BN_ULONG *a); |
| 514 |
int bn_cmp_words(const BN_ULONG *a, const BN_ULONG *b, int n); |
| 515 |
int bn_cmp_part_words(const BN_ULONG *a, const BN_ULONG *b, int cl, int dl); |
| 516 |
void bn_mul_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, |
| 517 |
int dna, int dnb, BN_ULONG *t); |
| 518 |
void bn_mul_part_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, |
| 519 |
int n, int tna, int tnb, BN_ULONG *t); |
| 520 |
void bn_sqr_recursive(BN_ULONG *r, const BN_ULONG *a, int n2, BN_ULONG *t); |
| 521 |
void bn_mul_low_normal(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n); |
| 522 |
void bn_mul_low_recursive(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, int n2, |
| 523 |
BN_ULONG *t); |
| 524 |
void bn_mul_high(BN_ULONG *r, BN_ULONG *a, BN_ULONG *b, BN_ULONG *l, int n2, |
| 525 |
BN_ULONG *t); |
| 526 |
BN_ULONG bn_add_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, |
| 527 |
int cl, int dl); |
| 528 |
BN_ULONG bn_sub_part_words(BN_ULONG *r, const BN_ULONG *a, const BN_ULONG *b, |
| 529 |
int cl, int dl); |
| 530 |
int bn_mul_mont(BN_ULONG *rp, const BN_ULONG *ap, const BN_ULONG *bp, |
| 531 |
const BN_ULONG *np, const BN_ULONG *n0, int num); |
| 532 |
|
| 533 |
#ifdef __cplusplus |
| 534 |
} |
| 535 |
#endif |
| 536 |
|
| 537 |
#endif |