| 1 | /* |
| 2 | * Copyright (c) 2007, 2011, Oracle and/or its affiliates. All rights reserved. |
| 3 | * Use is subject to license terms. |
| 4 | * |
| 5 | * This library is free software; you can redistribute it and/or |
| 6 | * modify it under the terms of the GNU Lesser General Public |
| 7 | * License as published by the Free Software Foundation; either |
| 8 | * version 2.1 of the License, or (at your option) any later version. |
| 9 | * |
| 10 | * This library is distributed in the hope that it will be useful, |
| 11 | * but WITHOUT ANY WARRANTY; without even the implied warranty of |
| 12 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| 13 | * Lesser General Public License for more details. |
| 14 | * |
| 15 | * You should have received a copy of the GNU Lesser General Public License |
| 16 | * along with this library; if not, write to the Free Software Foundation, |
| 17 | * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA. |
| 18 | * |
| 19 | * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA |
| 20 | * or visit www.oracle.com if you need additional information or have any |
| 21 | * questions. |
| 22 | */ |
| 23 | |
| 24 | /* ********************************************************************* |
| 25 | * |
| 26 | * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library. |
| 27 | * |
| 28 | * The Initial Developer of the Original Code is |
| 29 | * Michael J. Fromberger. |
| 30 | * Portions created by the Initial Developer are Copyright (C) 1998 |
| 31 | * the Initial Developer. All Rights Reserved. |
| 32 | * |
| 33 | * Contributor(s): |
| 34 | * Netscape Communications Corporation |
| 35 | * |
| 36 | *********************************************************************** */ |
| 37 | |
| 38 | /* Arbitrary precision integer arithmetic library |
| 39 | * |
| 40 | * NOTE WELL: the content of this header file is NOT part of the "public" |
| 41 | * API for the MPI library, and may change at any time. |
| 42 | * Application programs that use libmpi should NOT include this header file. |
| 43 | */ |
| 44 | |
| 45 | #ifndef _MPI_PRIV_H |
| 46 | #define _MPI_PRIV_H |
| 47 | |
| 48 | /* $Id: mpi-priv.h,v 1.20 2005/11/22 07:16:43 relyea%netscape.com Exp $ */ |
| 49 | |
| 50 | #include "mpi.h" |
| 51 | #ifndef _KERNEL |
| 52 | #include <stdlib.h> |
| 53 | #include <string.h> |
| 54 | #include <ctype.h> |
| 55 | #endif /* _KERNEL */ |
| 56 | |
| 57 | #if MP_DEBUG |
| 58 | #include <stdio.h> |
| 59 | |
| 60 | #define DIAG(T,V) {fprintf(stderr,T);mp_print(V,stderr);fputc('\n',stderr);} |
| 61 | #else |
| 62 | #define DIAG(T,V) |
| 63 | #endif |
| 64 | |
| 65 | /* If we aren't using a wired-in logarithm table, we need to include |
| 66 | the math library to get the log() function |
| 67 | */ |
| 68 | |
| 69 | /* {{{ s_logv_2[] - log table for 2 in various bases */ |
| 70 | |
| 71 | #if MP_LOGTAB |
| 72 | /* |
| 73 | A table of the logs of 2 for various bases (the 0 and 1 entries of |
| 74 | this table are meaningless and should not be referenced). |
| 75 | |
| 76 | This table is used to compute output lengths for the mp_toradix() |
| 77 | function. Since a number n in radix r takes up about log_r(n) |
| 78 | digits, we estimate the output size by taking the least integer |
| 79 | greater than log_r(n), where: |
| 80 | |
| 81 | log_r(n) = log_2(n) * log_r(2) |
| 82 | |
| 83 | This table, therefore, is a table of log_r(2) for 2 <= r <= 36, |
| 84 | which are the output bases supported. |
| 85 | */ |
| 86 | |
| 87 | extern const float s_logv_2[]; |
| 88 | #define LOG_V_2(R) s_logv_2[(R)] |
| 89 | |
| 90 | #else |
| 91 | |
| 92 | /* |
| 93 | If MP_LOGTAB is not defined, use the math library to compute the |
| 94 | logarithms on the fly. Otherwise, use the table. |
| 95 | Pick which works best for your system. |
| 96 | */ |
| 97 | |
| 98 | #include <math.h> |
| 99 | #define LOG_V_2(R) (log(2.0)/log(R)) |
| 100 | |
| 101 | #endif /* if MP_LOGTAB */ |
| 102 | |
| 103 | /* }}} */ |
| 104 | |
| 105 | /* {{{ Digit arithmetic macros */ |
| 106 | |
| 107 | /* |
| 108 | When adding and multiplying digits, the results can be larger than |
| 109 | can be contained in an mp_digit. Thus, an mp_word is used. These |
| 110 | macros mask off the upper and lower digits of the mp_word (the |
| 111 | mp_word may be more than 2 mp_digits wide, but we only concern |
| 112 | ourselves with the low-order 2 mp_digits) |
| 113 | */ |
| 114 | |
| 115 | #define CARRYOUT(W) (mp_digit)((W)>>DIGIT_BIT) |
| 116 | #define ACCUM(W) (mp_digit)(W) |
| 117 | |
| 118 | #define MP_MIN(a,b) (((a) < (b)) ? (a) : (b)) |
| 119 | #define MP_MAX(a,b) (((a) > (b)) ? (a) : (b)) |
| 120 | #define MP_HOWMANY(a,b) (((a) + (b) - 1)/(b)) |
| 121 | #define MP_ROUNDUP(a,b) (MP_HOWMANY(a,b) * (b)) |
| 122 | |
| 123 | /* }}} */ |
| 124 | |
| 125 | /* {{{ Comparison constants */ |
| 126 | |
| 127 | #define MP_LT -1 |
| 128 | #define MP_EQ 0 |
| 129 | #define MP_GT 1 |
| 130 | |
| 131 | /* }}} */ |
| 132 | |
| 133 | /* {{{ private function declarations */ |
| 134 | |
| 135 | /* |
| 136 | If MP_MACRO is false, these will be defined as actual functions; |
| 137 | otherwise, suitable macro definitions will be used. This works |
| 138 | around the fact that ANSI C89 doesn't support an 'inline' keyword |
| 139 | (although I hear C9x will ... about bloody time). At present, the |
| 140 | macro definitions are identical to the function bodies, but they'll |
| 141 | expand in place, instead of generating a function call. |
| 142 | |
| 143 | I chose these particular functions to be made into macros because |
| 144 | some profiling showed they are called a lot on a typical workload, |
| 145 | and yet they are primarily housekeeping. |
| 146 | */ |
| 147 | #if MP_MACRO == 0 |
| 148 | void s_mp_setz(mp_digit *dp, mp_size count); /* zero digits */ |
| 149 | void s_mp_copy(const mp_digit *sp, mp_digit *dp, mp_size count); /* copy */ |
| 150 | void *s_mp_alloc(size_t nb, size_t ni, int flag); /* general allocator */ |
| 151 | void s_mp_free(void *ptr, mp_size); /* general free function */ |
| 152 | extern unsigned long mp_allocs; |
| 153 | extern unsigned long mp_frees; |
| 154 | extern unsigned long mp_copies; |
| 155 | #else |
| 156 | |
| 157 | /* Even if these are defined as macros, we need to respect the settings |
| 158 | of the MP_MEMSET and MP_MEMCPY configuration options... |
| 159 | */ |
| 160 | #if MP_MEMSET == 0 |
| 161 | #define s_mp_setz(dp, count) \ |
| 162 | {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=0;} |
| 163 | #else |
| 164 | #define s_mp_setz(dp, count) memset(dp, 0, (count) * sizeof(mp_digit)) |
| 165 | #endif /* MP_MEMSET */ |
| 166 | |
| 167 | #if MP_MEMCPY == 0 |
| 168 | #define s_mp_copy(sp, dp, count) \ |
| 169 | {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=(sp)[ix];} |
| 170 | #else |
| 171 | #define s_mp_copy(sp, dp, count) memcpy(dp, sp, (count) * sizeof(mp_digit)) |
| 172 | #endif /* MP_MEMCPY */ |
| 173 | |
| 174 | #define s_mp_alloc(nb, ni) calloc(nb, ni) |
| 175 | #define s_mp_free(ptr) {if(ptr) free(ptr);} |
| 176 | #endif /* MP_MACRO */ |
| 177 | |
| 178 | mp_err s_mp_grow(mp_int *mp, mp_size min); /* increase allocated size */ |
| 179 | mp_err s_mp_pad(mp_int *mp, mp_size min); /* left pad with zeroes */ |
| 180 | |
| 181 | #if MP_MACRO == 0 |
| 182 | void s_mp_clamp(mp_int *mp); /* clip leading zeroes */ |
| 183 | #else |
| 184 | #define s_mp_clamp(mp)\ |
| 185 | { mp_size used = MP_USED(mp); \ |
| 186 | while (used > 1 && DIGIT(mp, used - 1) == 0) --used; \ |
| 187 | MP_USED(mp) = used; \ |
| 188 | } |
| 189 | #endif /* MP_MACRO */ |
| 190 | |
| 191 | void s_mp_exch(mp_int *a, mp_int *b); /* swap a and b in place */ |
| 192 | |
| 193 | mp_err s_mp_lshd(mp_int *mp, mp_size p); /* left-shift by p digits */ |
| 194 | void s_mp_rshd(mp_int *mp, mp_size p); /* right-shift by p digits */ |
| 195 | mp_err s_mp_mul_2d(mp_int *mp, mp_digit d); /* multiply by 2^d in place */ |
| 196 | void s_mp_div_2d(mp_int *mp, mp_digit d); /* divide by 2^d in place */ |
| 197 | void s_mp_mod_2d(mp_int *mp, mp_digit d); /* modulo 2^d in place */ |
| 198 | void s_mp_div_2(mp_int *mp); /* divide by 2 in place */ |
| 199 | mp_err s_mp_mul_2(mp_int *mp); /* multiply by 2 in place */ |
| 200 | mp_err s_mp_norm(mp_int *a, mp_int *b, mp_digit *pd); |
| 201 | /* normalize for division */ |
| 202 | mp_err s_mp_add_d(mp_int *mp, mp_digit d); /* unsigned digit addition */ |
| 203 | mp_err s_mp_sub_d(mp_int *mp, mp_digit d); /* unsigned digit subtract */ |
| 204 | mp_err s_mp_mul_d(mp_int *mp, mp_digit d); /* unsigned digit multiply */ |
| 205 | mp_err s_mp_div_d(mp_int *mp, mp_digit d, mp_digit *r); |
| 206 | /* unsigned digit divide */ |
| 207 | mp_err s_mp_reduce(mp_int *x, const mp_int *m, const mp_int *mu); |
| 208 | /* Barrett reduction */ |
| 209 | mp_err s_mp_add(mp_int *a, const mp_int *b); /* magnitude addition */ |
| 210 | mp_err s_mp_add_3arg(const mp_int *a, const mp_int *b, mp_int *c); |
| 211 | mp_err s_mp_sub(mp_int *a, const mp_int *b); /* magnitude subtract */ |
| 212 | mp_err s_mp_sub_3arg(const mp_int *a, const mp_int *b, mp_int *c); |
| 213 | mp_err s_mp_add_offset(mp_int *a, mp_int *b, mp_size offset); |
| 214 | /* a += b * RADIX^offset */ |
| 215 | mp_err s_mp_mul(mp_int *a, const mp_int *b); /* magnitude multiply */ |
| 216 | #if MP_SQUARE |
| 217 | mp_err s_mp_sqr(mp_int *a); /* magnitude square */ |
| 218 | #else |
| 219 | #define s_mp_sqr(a) s_mp_mul(a, a) |
| 220 | #endif |
| 221 | mp_err s_mp_div(mp_int *rem, mp_int *div, mp_int *quot); /* magnitude div */ |
| 222 | mp_err s_mp_exptmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c); |
| 223 | mp_err s_mp_2expt(mp_int *a, mp_digit k); /* a = 2^k */ |
| 224 | int s_mp_cmp(const mp_int *a, const mp_int *b); /* magnitude comparison */ |
| 225 | int s_mp_cmp_d(const mp_int *a, mp_digit d); /* magnitude digit compare */ |
| 226 | int s_mp_ispow2(const mp_int *v); /* is v a power of 2? */ |
| 227 | int s_mp_ispow2d(mp_digit d); /* is d a power of 2? */ |
| 228 | |
| 229 | int s_mp_tovalue(char ch, int r); /* convert ch to value */ |
| 230 | char s_mp_todigit(mp_digit val, int r, int low); /* convert val to digit */ |
| 231 | int s_mp_outlen(int bits, int r); /* output length in bytes */ |
| 232 | mp_digit s_mp_invmod_radix(mp_digit P); /* returns (P ** -1) mod RADIX */ |
| 233 | mp_err s_mp_invmod_odd_m( const mp_int *a, const mp_int *m, mp_int *c); |
| 234 | mp_err s_mp_invmod_2d( const mp_int *a, mp_size k, mp_int *c); |
| 235 | mp_err s_mp_invmod_even_m(const mp_int *a, const mp_int *m, mp_int *c); |
| 236 | |
| 237 | #ifdef NSS_USE_COMBA |
| 238 | |
| 239 | #define IS_POWER_OF_2(a) ((a) && !((a) & ((a)-1))) |
| 240 | |
| 241 | void s_mp_mul_comba_4(const mp_int *A, const mp_int *B, mp_int *C); |
| 242 | void s_mp_mul_comba_8(const mp_int *A, const mp_int *B, mp_int *C); |
| 243 | void s_mp_mul_comba_16(const mp_int *A, const mp_int *B, mp_int *C); |
| 244 | void s_mp_mul_comba_32(const mp_int *A, const mp_int *B, mp_int *C); |
| 245 | |
| 246 | void s_mp_sqr_comba_4(const mp_int *A, mp_int *B); |
| 247 | void s_mp_sqr_comba_8(const mp_int *A, mp_int *B); |
| 248 | void s_mp_sqr_comba_16(const mp_int *A, mp_int *B); |
| 249 | void s_mp_sqr_comba_32(const mp_int *A, mp_int *B); |
| 250 | |
| 251 | #endif /* end NSS_USE_COMBA */ |
| 252 | |
| 253 | /* ------ mpv functions, operate on arrays of digits, not on mp_int's ------ */ |
| 254 | #if defined (__OS2__) && defined (__IBMC__) |
| 255 | #define MPI_ASM_DECL __cdecl |
| 256 | #else |
| 257 | #define MPI_ASM_DECL |
| 258 | #endif |
| 259 | |
| 260 | #ifdef MPI_AMD64 |
| 261 | |
| 262 | mp_digit MPI_ASM_DECL s_mpv_mul_set_vec64(mp_digit*, mp_digit *, mp_size, mp_digit); |
| 263 | mp_digit MPI_ASM_DECL s_mpv_mul_add_vec64(mp_digit*, const mp_digit*, mp_size, mp_digit); |
| 264 | |
| 265 | /* c = a * b */ |
| 266 | #define s_mpv_mul_d(a, a_len, b, c) \ |
| 267 | ((unsigned long*)c)[a_len] = s_mpv_mul_set_vec64(c, a, a_len, b) |
| 268 | |
| 269 | /* c += a * b */ |
| 270 | #define s_mpv_mul_d_add(a, a_len, b, c) \ |
| 271 | ((unsigned long*)c)[a_len] = s_mpv_mul_add_vec64(c, a, a_len, b) |
| 272 | |
| 273 | #else |
| 274 | |
| 275 | void MPI_ASM_DECL s_mpv_mul_d(const mp_digit *a, mp_size a_len, |
| 276 | mp_digit b, mp_digit *c); |
| 277 | void MPI_ASM_DECL s_mpv_mul_d_add(const mp_digit *a, mp_size a_len, |
| 278 | mp_digit b, mp_digit *c); |
| 279 | |
| 280 | #endif |
| 281 | |
| 282 | void MPI_ASM_DECL s_mpv_mul_d_add_prop(const mp_digit *a, |
| 283 | mp_size a_len, mp_digit b, |
| 284 | mp_digit *c); |
| 285 | void MPI_ASM_DECL s_mpv_sqr_add_prop(const mp_digit *a, |
| 286 | mp_size a_len, |
| 287 | mp_digit *sqrs); |
| 288 | |
| 289 | mp_err MPI_ASM_DECL s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo, |
| 290 | mp_digit divisor, mp_digit *quot, mp_digit *rem); |
| 291 | |
| 292 | /* c += a * b * (MP_RADIX ** offset); */ |
| 293 | #define s_mp_mul_d_add_offset(a, b, c, off) \ |
| 294 | (s_mpv_mul_d_add_prop(MP_DIGITS(a), MP_USED(a), b, MP_DIGITS(c) + off), MP_OKAY) |
| 295 | |
| 296 | typedef struct { |
| 297 | mp_int N; /* modulus N */ |
| 298 | mp_digit n0prime; /* n0' = - (n0 ** -1) mod MP_RADIX */ |
| 299 | mp_size b; /* R == 2 ** b, also b = # significant bits in N */ |
| 300 | } mp_mont_modulus; |
| 301 | |
| 302 | mp_err s_mp_mul_mont(const mp_int *a, const mp_int *b, mp_int *c, |
| 303 | mp_mont_modulus *mmm); |
| 304 | mp_err s_mp_redc(mp_int *T, mp_mont_modulus *mmm); |
| 305 | |
| 306 | /* |
| 307 | * s_mpi_getProcessorLineSize() returns the size in bytes of the cache line |
| 308 | * if a cache exists, or zero if there is no cache. If more than one |
| 309 | * cache line exists, it should return the smallest line size (which is |
| 310 | * usually the L1 cache). |
| 311 | * |
| 312 | * mp_modexp uses this information to make sure that private key information |
| 313 | * isn't being leaked through the cache. |
| 314 | * |
| 315 | * see mpcpucache.c for the implementation. |
| 316 | */ |
| 317 | unsigned long s_mpi_getProcessorLineSize(); |
| 318 | |
| 319 | /* }}} */ |
| 320 | #endif /* _MPI_PRIV_H */ |
| 321 | |