1 /*
2 * mpi-priv.h - Private header file for MPI
3 * Arbitrary precision integer arithmetic library
4 *
5 * NOTE WELL: the content of this header file is NOT part of the "public"
6 * API for the MPI library, and may change at any time.
7 * Application programs that use libmpi should NOT include this header file.
8 *
9 * ***** BEGIN LICENSE BLOCK *****
10 * Version: MPL 1.1/GPL 2.0/LGPL 2.1
11 *
12 * The contents of this file are subject to the Mozilla Public License Version
13 * 1.1 (the "License"); you may not use this file except in compliance with
14 * the License. You may obtain a copy of the License at
15 * http://www.mozilla.org/MPL/
16 *
17 * Software distributed under the License is distributed on an "AS IS" basis,
18 * WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
19 * for the specific language governing rights and limitations under the
20 * License.
21 *
22 * The Original Code is the MPI Arbitrary Precision Integer Arithmetic library.
23 *
24 * The Initial Developer of the Original Code is
25 * Michael J. Fromberger.
26 * Portions created by the Initial Developer are Copyright (C) 1998
27 * the Initial Developer. All Rights Reserved.
28 *
29 * Contributor(s):
30 * Netscape Communications Corporation
31 *
32 * Alternatively, the contents of this file may be used under the terms of
33 * either the GNU General Public License Version 2 or later (the "GPL"), or
34 * the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
35 * in which case the provisions of the GPL or the LGPL are applicable instead
36 * of those above. If you wish to allow use of your version of this file only
37 * under the terms of either the GPL or the LGPL, and not to allow others to
38 * use your version of this file under the terms of the MPL, indicate your
39 * decision by deleting the provisions above and replace them with the notice
40 * and other provisions required by the GPL or the LGPL. If you do not delete
41 * the provisions above, a recipient may use your version of this file under
42 * the terms of any one of the MPL, the GPL or the LGPL.
43 *
44 * ***** END LICENSE BLOCK ***** */
45 /*
46 * Copyright 2007 Sun Microsystems, Inc. All rights reserved.
47 * Use is subject to license terms.
48 *
49 * Sun elects to use this software under the MPL license.
50 */
51
52 #ifndef _MPI_PRIV_H
53 #define _MPI_PRIV_H
54
55 #pragma ident "%Z%%M% %I% %E% SMI"
56
57 /* $Id: mpi-priv.h,v 1.20 2005/11/22 07:16:43 relyea%netscape.com Exp $ */
58
59 #include "mpi.h"
60 #ifndef _KERNEL
61 #include <stdlib.h>
62 #include <string.h>
63 #include <ctype.h>
64 #endif /* _KERNEL */
65
66 #if MP_DEBUG
67 #include <stdio.h>
68
69 #define DIAG(T,V) {fprintf(stderr,T);mp_print(V,stderr);fputc('\n',stderr);}
70 #else
71 #define DIAG(T,V)
72 #endif
73
74 /* If we aren't using a wired-in logarithm table, we need to include
75 the math library to get the log() function
76 */
77
78 /* {{{ s_logv_2[] - log table for 2 in various bases */
79
80 #if MP_LOGTAB
81 /*
82 A table of the logs of 2 for various bases (the 0 and 1 entries of
83 this table are meaningless and should not be referenced).
84
85 This table is used to compute output lengths for the mp_toradix()
86 function. Since a number n in radix r takes up about log_r(n)
87 digits, we estimate the output size by taking the least integer
88 greater than log_r(n), where:
89
90 log_r(n) = log_2(n) * log_r(2)
91
92 This table, therefore, is a table of log_r(2) for 2 <= r <= 36,
93 which are the output bases supported.
94 */
95
96 extern const float s_logv_2[];
97 #define LOG_V_2(R) s_logv_2[(R)]
98
99 #else
100
101 /*
102 If MP_LOGTAB is not defined, use the math library to compute the
103 logarithms on the fly. Otherwise, use the table.
104 Pick which works best for your system.
105 */
106
107 #include <math.h>
108 #define LOG_V_2(R) (log(2.0)/log(R))
109
110 #endif /* if MP_LOGTAB */
111
112 /* }}} */
113
114 /* {{{ Digit arithmetic macros */
115
116 /*
117 When adding and multiplying digits, the results can be larger than
118 can be contained in an mp_digit. Thus, an mp_word is used. These
119 macros mask off the upper and lower digits of the mp_word (the
120 mp_word may be more than 2 mp_digits wide, but we only concern
121 ourselves with the low-order 2 mp_digits)
122 */
123
124 #define CARRYOUT(W) (mp_digit)((W)>>DIGIT_BIT)
125 #define ACCUM(W) (mp_digit)(W)
126
127 #define MP_MIN(a,b) (((a) < (b)) ? (a) : (b))
128 #define MP_MAX(a,b) (((a) > (b)) ? (a) : (b))
129 #define MP_HOWMANY(a,b) (((a) + (b) - 1)/(b))
130 #define MP_ROUNDUP(a,b) (MP_HOWMANY(a,b) * (b))
131
132 /* }}} */
133
134 /* {{{ Comparison constants */
135
136 #define MP_LT -1
137 #define MP_EQ 0
138 #define MP_GT 1
139
140 /* }}} */
141
142 /* {{{ private function declarations */
143
144 /*
145 If MP_MACRO is false, these will be defined as actual functions;
146 otherwise, suitable macro definitions will be used. This works
147 around the fact that ANSI C89 doesn't support an 'inline' keyword
148 (although I hear C9x will ... about bloody time). At present, the
149 macro definitions are identical to the function bodies, but they'll
150 expand in place, instead of generating a function call.
151
152 I chose these particular functions to be made into macros because
153 some profiling showed they are called a lot on a typical workload,
154 and yet they are primarily housekeeping.
155 */
156 #if MP_MACRO == 0
157 void s_mp_setz(mp_digit *dp, mp_size count); /* zero digits */
158 void s_mp_copy(const mp_digit *sp, mp_digit *dp, mp_size count); /* copy */
159 void *s_mp_alloc(size_t nb, size_t ni, int flag); /* general allocator */
160 void s_mp_free(void *ptr, mp_size); /* general free function */
161 extern unsigned long mp_allocs;
162 extern unsigned long mp_frees;
163 extern unsigned long mp_copies;
164 #else
165
166 /* Even if these are defined as macros, we need to respect the settings
167 of the MP_MEMSET and MP_MEMCPY configuration options...
168 */
169 #if MP_MEMSET == 0
170 #define s_mp_setz(dp, count) \
171 {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=0;}
172 #else
173 #define s_mp_setz(dp, count) memset(dp, 0, (count) * sizeof(mp_digit))
174 #endif /* MP_MEMSET */
175
176 #if MP_MEMCPY == 0
177 #define s_mp_copy(sp, dp, count) \
178 {int ix;for(ix=0;ix<(count);ix++)(dp)[ix]=(sp)[ix];}
179 #else
180 #define s_mp_copy(sp, dp, count) memcpy(dp, sp, (count) * sizeof(mp_digit))
181 #endif /* MP_MEMCPY */
182
183 #define s_mp_alloc(nb, ni) calloc(nb, ni)
184 #define s_mp_free(ptr) {if(ptr) free(ptr);}
185 #endif /* MP_MACRO */
186
187 mp_err s_mp_grow(mp_int *mp, mp_size min); /* increase allocated size */
188 mp_err s_mp_pad(mp_int *mp, mp_size min); /* left pad with zeroes */
189
190 #if MP_MACRO == 0
191 void s_mp_clamp(mp_int *mp); /* clip leading zeroes */
192 #else
193 #define s_mp_clamp(mp)\
194 { mp_size used = MP_USED(mp); \
195 while (used > 1 && DIGIT(mp, used - 1) == 0) --used; \
196 MP_USED(mp) = used; \
197 }
198 #endif /* MP_MACRO */
199
200 void s_mp_exch(mp_int *a, mp_int *b); /* swap a and b in place */
201
202 mp_err s_mp_lshd(mp_int *mp, mp_size p); /* left-shift by p digits */
203 void s_mp_rshd(mp_int *mp, mp_size p); /* right-shift by p digits */
204 mp_err s_mp_mul_2d(mp_int *mp, mp_digit d); /* multiply by 2^d in place */
205 void s_mp_div_2d(mp_int *mp, mp_digit d); /* divide by 2^d in place */
206 void s_mp_mod_2d(mp_int *mp, mp_digit d); /* modulo 2^d in place */
207 void s_mp_div_2(mp_int *mp); /* divide by 2 in place */
208 mp_err s_mp_mul_2(mp_int *mp); /* multiply by 2 in place */
209 mp_err s_mp_norm(mp_int *a, mp_int *b, mp_digit *pd);
210 /* normalize for division */
211 mp_err s_mp_add_d(mp_int *mp, mp_digit d); /* unsigned digit addition */
212 mp_err s_mp_sub_d(mp_int *mp, mp_digit d); /* unsigned digit subtract */
213 mp_err s_mp_mul_d(mp_int *mp, mp_digit d); /* unsigned digit multiply */
214 mp_err s_mp_div_d(mp_int *mp, mp_digit d, mp_digit *r);
215 /* unsigned digit divide */
216 mp_err s_mp_reduce(mp_int *x, const mp_int *m, const mp_int *mu);
217 /* Barrett reduction */
218 mp_err s_mp_add(mp_int *a, const mp_int *b); /* magnitude addition */
219 mp_err s_mp_add_3arg(const mp_int *a, const mp_int *b, mp_int *c);
220 mp_err s_mp_sub(mp_int *a, const mp_int *b); /* magnitude subtract */
221 mp_err s_mp_sub_3arg(const mp_int *a, const mp_int *b, mp_int *c);
222 mp_err s_mp_add_offset(mp_int *a, mp_int *b, mp_size offset);
223 /* a += b * RADIX^offset */
224 mp_err s_mp_mul(mp_int *a, const mp_int *b); /* magnitude multiply */
225 #if MP_SQUARE
226 mp_err s_mp_sqr(mp_int *a); /* magnitude square */
227 #else
228 #define s_mp_sqr(a) s_mp_mul(a, a)
229 #endif
230 mp_err s_mp_div(mp_int *rem, mp_int *div, mp_int *quot); /* magnitude div */
231 mp_err s_mp_exptmod(const mp_int *a, const mp_int *b, const mp_int *m, mp_int *c);
232 mp_err s_mp_2expt(mp_int *a, mp_digit k); /* a = 2^k */
233 int s_mp_cmp(const mp_int *a, const mp_int *b); /* magnitude comparison */
234 int s_mp_cmp_d(const mp_int *a, mp_digit d); /* magnitude digit compare */
235 int s_mp_ispow2(const mp_int *v); /* is v a power of 2? */
236 int s_mp_ispow2d(mp_digit d); /* is d a power of 2? */
237
238 int s_mp_tovalue(char ch, int r); /* convert ch to value */
239 char s_mp_todigit(mp_digit val, int r, int low); /* convert val to digit */
240 int s_mp_outlen(int bits, int r); /* output length in bytes */
241 mp_digit s_mp_invmod_radix(mp_digit P); /* returns (P ** -1) mod RADIX */
242 mp_err s_mp_invmod_odd_m( const mp_int *a, const mp_int *m, mp_int *c);
243 mp_err s_mp_invmod_2d( const mp_int *a, mp_size k, mp_int *c);
244 mp_err s_mp_invmod_even_m(const mp_int *a, const mp_int *m, mp_int *c);
245
246 #ifdef NSS_USE_COMBA
247
248 #define IS_POWER_OF_2(a) ((a) && !((a) & ((a)-1)))
249
250 void s_mp_mul_comba_4(const mp_int *A, const mp_int *B, mp_int *C);
251 void s_mp_mul_comba_8(const mp_int *A, const mp_int *B, mp_int *C);
252 void s_mp_mul_comba_16(const mp_int *A, const mp_int *B, mp_int *C);
253 void s_mp_mul_comba_32(const mp_int *A, const mp_int *B, mp_int *C);
254
255 void s_mp_sqr_comba_4(const mp_int *A, mp_int *B);
256 void s_mp_sqr_comba_8(const mp_int *A, mp_int *B);
257 void s_mp_sqr_comba_16(const mp_int *A, mp_int *B);
258 void s_mp_sqr_comba_32(const mp_int *A, mp_int *B);
259
260 #endif /* end NSS_USE_COMBA */
261
262 /* ------ mpv functions, operate on arrays of digits, not on mp_int's ------ */
263 #if defined (__OS2__) && defined (__IBMC__)
264 #define MPI_ASM_DECL __cdecl
265 #else
266 #define MPI_ASM_DECL
267 #endif
268
269 #ifdef MPI_AMD64
270
271 mp_digit MPI_ASM_DECL s_mpv_mul_set_vec64(mp_digit*, mp_digit *, mp_size, mp_digit);
272 mp_digit MPI_ASM_DECL s_mpv_mul_add_vec64(mp_digit*, const mp_digit*, mp_size, mp_digit);
273
274 /* c = a * b */
275 #define s_mpv_mul_d(a, a_len, b, c) \
276 ((unsigned long*)c)[a_len] = s_mpv_mul_set_vec64(c, a, a_len, b)
277
278 /* c += a * b */
279 #define s_mpv_mul_d_add(a, a_len, b, c) \
280 ((unsigned long*)c)[a_len] = s_mpv_mul_add_vec64(c, a, a_len, b)
281
282 #else
283
284 void MPI_ASM_DECL s_mpv_mul_d(const mp_digit *a, mp_size a_len,
285 mp_digit b, mp_digit *c);
286 void MPI_ASM_DECL s_mpv_mul_d_add(const mp_digit *a, mp_size a_len,
287 mp_digit b, mp_digit *c);
288
289 #endif
290
291 void MPI_ASM_DECL s_mpv_mul_d_add_prop(const mp_digit *a,
292 mp_size a_len, mp_digit b,
293 mp_digit *c);
294 void MPI_ASM_DECL s_mpv_sqr_add_prop(const mp_digit *a,
295 mp_size a_len,
296 mp_digit *sqrs);
297
298 mp_err MPI_ASM_DECL s_mpv_div_2dx1d(mp_digit Nhi, mp_digit Nlo,
299 mp_digit divisor, mp_digit *quot, mp_digit *rem);
300
301 /* c += a * b * (MP_RADIX ** offset); */
302 #define s_mp_mul_d_add_offset(a, b, c, off) \
303 (s_mpv_mul_d_add_prop(MP_DIGITS(a), MP_USED(a), b, MP_DIGITS(c) + off), MP_OKAY)
304
305 typedef struct {
306 mp_int N; /* modulus N */
307 mp_digit n0prime; /* n0' = - (n0 ** -1) mod MP_RADIX */
308 mp_size b; /* R == 2 ** b, also b = # significant bits in N */
309 } mp_mont_modulus;
310
311 mp_err s_mp_mul_mont(const mp_int *a, const mp_int *b, mp_int *c,
312 mp_mont_modulus *mmm);
313 mp_err s_mp_redc(mp_int *T, mp_mont_modulus *mmm);
314
315 /*
316 * s_mpi_getProcessorLineSize() returns the size in bytes of the cache line
317 * if a cache exists, or zero if there is no cache. If more than one
318 * cache line exists, it should return the smallest line size (which is
319 * usually the L1 cache).
320 *
321 * mp_modexp uses this information to make sure that private key information
322 * isn't being leaked through the cache.
323 *
324 * see mpcpucache.c for the implementation.
325 */
326 unsigned long s_mpi_getProcessorLineSize();
327
328 /* }}} */
329 #endif /* _MPI_PRIV_H */