Remove fips_constseg references.
[openssl.git] / crypto / bn / bn_nist.c
1 /* crypto/bn/bn_nist.c */
2 /*
3  * Written by Nils Larsch for the OpenSSL project
4  */
5 /* ====================================================================
6  * Copyright (c) 1998-2005 The OpenSSL Project.  All rights reserved.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  *
12  * 1. Redistributions of source code must retain the above copyright
13  *    notice, this list of conditions and the following disclaimer. 
14  *
15  * 2. Redistributions in binary form must reproduce the above copyright
16  *    notice, this list of conditions and the following disclaimer in
17  *    the documentation and/or other materials provided with the
18  *    distribution.
19  *
20  * 3. All advertising materials mentioning features or use of this
21  *    software must display the following acknowledgment:
22  *    "This product includes software developed by the OpenSSL Project
23  *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
24  *
25  * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
26  *    endorse or promote products derived from this software without
27  *    prior written permission. For written permission, please contact
28  *    openssl-core@openssl.org.
29  *
30  * 5. Products derived from this software may not be called "OpenSSL"
31  *    nor may "OpenSSL" appear in their names without prior written
32  *    permission of the OpenSSL Project.
33  *
34  * 6. Redistributions of any form whatsoever must retain the following
35  *    acknowledgment:
36  *    "This product includes software developed by the OpenSSL Project
37  *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
38  *
39  * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
40  * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
41  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
42  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
43  * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
44  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
45  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
46  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
48  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
49  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
50  * OF THE POSSIBILITY OF SUCH DAMAGE.
51  * ====================================================================
52  *
53  * This product includes cryptographic software written by Eric Young
54  * (eay@cryptsoft.com).  This product includes software written by Tim
55  * Hudson (tjh@cryptsoft.com).
56  *
57  */
58
59 #include "bn_lcl.h"
60 #include "cryptlib.h"
61
62
63 #define BN_NIST_192_TOP (192+BN_BITS2-1)/BN_BITS2
64 #define BN_NIST_224_TOP (224+BN_BITS2-1)/BN_BITS2
65 #define BN_NIST_256_TOP (256+BN_BITS2-1)/BN_BITS2
66 #define BN_NIST_384_TOP (384+BN_BITS2-1)/BN_BITS2
67 #define BN_NIST_521_TOP (521+BN_BITS2-1)/BN_BITS2
68
69 /* pre-computed tables are "carry-less" values of modulus*(i+1) */
70 #if BN_BITS2 == 64
71 static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
72         {0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFEULL,0xFFFFFFFFFFFFFFFFULL},
73         {0xFFFFFFFFFFFFFFFEULL,0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFFULL},
74         {0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFCULL,0xFFFFFFFFFFFFFFFFULL}
75         };
76 static const BN_ULONG _nist_p_192_sqr[] = {
77         0x0000000000000001ULL,0x0000000000000002ULL,0x0000000000000001ULL,
78         0xFFFFFFFFFFFFFFFEULL,0xFFFFFFFFFFFFFFFDULL,0xFFFFFFFFFFFFFFFFULL
79         };
80 static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
81         {0x0000000000000001ULL,0xFFFFFFFF00000000ULL,
82          0xFFFFFFFFFFFFFFFFULL,0x00000000FFFFFFFFULL},
83         {0x0000000000000002ULL,0xFFFFFFFE00000000ULL,
84          0xFFFFFFFFFFFFFFFFULL,0x00000001FFFFFFFFULL} /* this one is "carry-full" */
85         };
86 static const BN_ULONG _nist_p_224_sqr[] = {
87         0x0000000000000001ULL,0xFFFFFFFE00000000ULL,
88         0xFFFFFFFFFFFFFFFFULL,0x0000000200000000ULL,
89         0x0000000000000000ULL,0xFFFFFFFFFFFFFFFEULL,
90         0xFFFFFFFFFFFFFFFFULL
91         };
92 static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
93         {0xFFFFFFFFFFFFFFFFULL,0x00000000FFFFFFFFULL,
94          0x0000000000000000ULL,0xFFFFFFFF00000001ULL},
95         {0xFFFFFFFFFFFFFFFEULL,0x00000001FFFFFFFFULL,
96          0x0000000000000000ULL,0xFFFFFFFE00000002ULL},
97         {0xFFFFFFFFFFFFFFFDULL,0x00000002FFFFFFFFULL,
98          0x0000000000000000ULL,0xFFFFFFFD00000003ULL},
99         {0xFFFFFFFFFFFFFFFCULL,0x00000003FFFFFFFFULL,
100          0x0000000000000000ULL,0xFFFFFFFC00000004ULL},
101         {0xFFFFFFFFFFFFFFFBULL,0x00000004FFFFFFFFULL,
102          0x0000000000000000ULL,0xFFFFFFFB00000005ULL},
103         };
104 static const BN_ULONG _nist_p_256_sqr[] = {
105         0x0000000000000001ULL,0xFFFFFFFE00000000ULL,
106         0xFFFFFFFFFFFFFFFFULL,0x00000001FFFFFFFEULL,
107         0x00000001FFFFFFFEULL,0x00000001FFFFFFFEULL,
108         0xFFFFFFFE00000001ULL,0xFFFFFFFE00000002ULL
109         };
110 static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
111         {0x00000000FFFFFFFFULL,0xFFFFFFFF00000000ULL,0xFFFFFFFFFFFFFFFEULL,
112          0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
113         {0x00000001FFFFFFFEULL,0xFFFFFFFE00000000ULL,0xFFFFFFFFFFFFFFFDULL,
114          0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
115         {0x00000002FFFFFFFDULL,0xFFFFFFFD00000000ULL,0xFFFFFFFFFFFFFFFCULL,
116          0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
117         {0x00000003FFFFFFFCULL,0xFFFFFFFC00000000ULL,0xFFFFFFFFFFFFFFFBULL,
118          0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
119         {0x00000004FFFFFFFBULL,0xFFFFFFFB00000000ULL,0xFFFFFFFFFFFFFFFAULL,
120          0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL},
121         };
122 static const BN_ULONG _nist_p_384_sqr[] = {
123         0xFFFFFFFE00000001ULL,0x0000000200000000ULL,0xFFFFFFFE00000000ULL,
124         0x0000000200000000ULL,0x0000000000000001ULL,0x0000000000000000ULL,
125         0x00000001FFFFFFFEULL,0xFFFFFFFE00000000ULL,0xFFFFFFFFFFFFFFFDULL,
126         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL
127         };
128 static const BN_ULONG _nist_p_521[] =
129         {0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
130         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
131         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
132         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
133         0x00000000000001FFULL};
134 static const BN_ULONG _nist_p_521_sqr[] = {
135         0x0000000000000001ULL,0x0000000000000000ULL,0x0000000000000000ULL,
136         0x0000000000000000ULL,0x0000000000000000ULL,0x0000000000000000ULL,
137         0x0000000000000000ULL,0x0000000000000000ULL,0xFFFFFFFFFFFFFC00ULL,
138         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
139         0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,0xFFFFFFFFFFFFFFFFULL,
140         0xFFFFFFFFFFFFFFFFULL,0x000000000003FFFFULL
141         };
142 #elif BN_BITS2 == 32
143 static const BN_ULONG _nist_p_192[][BN_NIST_192_TOP] = {
144         {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
145         {0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
146         {0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFC,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF}
147         };
148 static const BN_ULONG _nist_p_192_sqr[] = {
149         0x00000001,0x00000000,0x00000002,0x00000000,0x00000001,0x00000000,
150         0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF
151         };
152 static const BN_ULONG _nist_p_224[][BN_NIST_224_TOP] = {
153         {0x00000001,0x00000000,0x00000000,0xFFFFFFFF,
154          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
155         {0x00000002,0x00000000,0x00000000,0xFFFFFFFE,
156          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF}
157         };
158 static const BN_ULONG _nist_p_224_sqr[] = {
159         0x00000001,0x00000000,0x00000000,0xFFFFFFFE,
160         0xFFFFFFFF,0xFFFFFFFF,0x00000000,0x00000002,
161         0x00000000,0x00000000,0xFFFFFFFE,0xFFFFFFFF,
162         0xFFFFFFFF,0xFFFFFFFF
163         };
164 static const BN_ULONG _nist_p_256[][BN_NIST_256_TOP] = {
165         {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0x00000000,
166          0x00000000,0x00000000,0x00000001,0xFFFFFFFF},
167         {0xFFFFFFFE,0xFFFFFFFF,0xFFFFFFFF,0x00000001,
168          0x00000000,0x00000000,0x00000002,0xFFFFFFFE},
169         {0xFFFFFFFD,0xFFFFFFFF,0xFFFFFFFF,0x00000002,
170          0x00000000,0x00000000,0x00000003,0xFFFFFFFD},
171         {0xFFFFFFFC,0xFFFFFFFF,0xFFFFFFFF,0x00000003,
172          0x00000000,0x00000000,0x00000004,0xFFFFFFFC},
173         {0xFFFFFFFB,0xFFFFFFFF,0xFFFFFFFF,0x00000004,
174          0x00000000,0x00000000,0x00000005,0xFFFFFFFB},
175         };
176 static const BN_ULONG _nist_p_256_sqr[] = {
177         0x00000001,0x00000000,0x00000000,0xFFFFFFFE,
178         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFE,0x00000001,
179         0xFFFFFFFE,0x00000001,0xFFFFFFFE,0x00000001,
180         0x00000001,0xFFFFFFFE,0x00000002,0xFFFFFFFE
181         };
182 static const BN_ULONG _nist_p_384[][BN_NIST_384_TOP] = {
183         {0xFFFFFFFF,0x00000000,0x00000000,0xFFFFFFFF,0xFFFFFFFE,0xFFFFFFFF,
184          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
185         {0xFFFFFFFE,0x00000001,0x00000000,0xFFFFFFFE,0xFFFFFFFD,0xFFFFFFFF,
186          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
187         {0xFFFFFFFD,0x00000002,0x00000000,0xFFFFFFFD,0xFFFFFFFC,0xFFFFFFFF,
188          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
189         {0xFFFFFFFC,0x00000003,0x00000000,0xFFFFFFFC,0xFFFFFFFB,0xFFFFFFFF,
190          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
191         {0xFFFFFFFB,0x00000004,0x00000000,0xFFFFFFFB,0xFFFFFFFA,0xFFFFFFFF,
192          0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF},
193         };
194 static const BN_ULONG _nist_p_384_sqr[] = {
195         0x00000001,0xFFFFFFFE,0x00000000,0x00000002,0x00000000,0xFFFFFFFE,
196         0x00000000,0x00000002,0x00000001,0x00000000,0x00000000,0x00000000,
197         0xFFFFFFFE,0x00000001,0x00000000,0xFFFFFFFE,0xFFFFFFFD,0xFFFFFFFF,
198         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF
199         };
200 static const BN_ULONG _nist_p_521[] = {0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
201         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
202         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
203         0xFFFFFFFF,0x000001FF};
204 static const BN_ULONG _nist_p_521_sqr[] = {
205         0x00000001,0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,
206         0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,0x00000000,
207         0x00000000,0x00000000,0x00000000,0x00000000,0xFFFFFC00,0xFFFFFFFF,
208         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
209         0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,0xFFFFFFFF,
210         0xFFFFFFFF,0xFFFFFFFF,0x0003FFFF
211         };
212 #else
213 #error "unsupported BN_BITS2"
214 #endif
215
216
217 static const BIGNUM _bignum_nist_p_192 =
218         {
219         (BN_ULONG *)_nist_p_192[0],
220         BN_NIST_192_TOP,
221         BN_NIST_192_TOP,
222         0,
223         BN_FLG_STATIC_DATA
224         };
225
226 static const BIGNUM _bignum_nist_p_224 =
227         {
228         (BN_ULONG *)_nist_p_224[0],
229         BN_NIST_224_TOP,
230         BN_NIST_224_TOP,
231         0,
232         BN_FLG_STATIC_DATA
233         };
234
235 static const BIGNUM _bignum_nist_p_256 =
236         {
237         (BN_ULONG *)_nist_p_256[0],
238         BN_NIST_256_TOP,
239         BN_NIST_256_TOP,
240         0,
241         BN_FLG_STATIC_DATA
242         };
243
244 static const BIGNUM _bignum_nist_p_384 =
245         {
246         (BN_ULONG *)_nist_p_384[0],
247         BN_NIST_384_TOP,
248         BN_NIST_384_TOP,
249         0,
250         BN_FLG_STATIC_DATA
251         };
252
253 static const BIGNUM _bignum_nist_p_521 =
254         {
255         (BN_ULONG *)_nist_p_521,
256         BN_NIST_521_TOP,
257         BN_NIST_521_TOP,
258         0,
259         BN_FLG_STATIC_DATA
260         };
261
262
263 const BIGNUM *BN_get0_nist_prime_192(void)
264         {
265         return &_bignum_nist_p_192;
266         }
267
268 const BIGNUM *BN_get0_nist_prime_224(void)
269         {
270         return &_bignum_nist_p_224;
271         }
272
273 const BIGNUM *BN_get0_nist_prime_256(void)
274         {
275         return &_bignum_nist_p_256;
276         }
277
278 const BIGNUM *BN_get0_nist_prime_384(void)
279         {
280         return &_bignum_nist_p_384;
281         }
282
283 const BIGNUM *BN_get0_nist_prime_521(void)
284         {
285         return &_bignum_nist_p_521;
286         }
287
288
289 static void nist_cp_bn_0(BN_ULONG *dst, const BN_ULONG *src, int top, int max)
290         {
291         int i;
292
293 #ifdef BN_DEBUG
294         OPENSSL_assert(top <= max);
295 #endif
296         for (i = 0; i < top; i++)
297                 dst[i] = src[i];
298         for (; i < max; i++)
299                 dst[i] = 0;
300         }
301
302 static void nist_cp_bn(BN_ULONG *dst, const BN_ULONG *src, int top)
303         { 
304         int i;
305
306         for (i = 0; i < top; i++)
307                 dst[i] = src[i];
308         }
309
310 #if BN_BITS2 == 64
311 #define bn_cp_64(to, n, from, m)        (to)[n] = (m>=0)?((from)[m]):0;
312 #define bn_64_set_0(to, n)              (to)[n] = (BN_ULONG)0;
313 /*
314  * two following macros are implemented under assumption that they
315  * are called in a sequence with *ascending* n, i.e. as they are...
316  */
317 #define bn_cp_32_naked(to, n, from, m)  (((n)&1)?(to[(n)/2]|=((m)&1)?(from[(m)/2]&BN_MASK2h):(from[(m)/2]<<32))\
318                                                 :(to[(n)/2] =((m)&1)?(from[(m)/2]>>32):(from[(m)/2]&BN_MASK2l)))
319 #define bn_32_set_0(to, n)              (((n)&1)?(to[(n)/2]&=BN_MASK2l):(to[(n)/2]=0));
320 #define bn_cp_32(to,n,from,m)           ((m)>=0)?bn_cp_32_naked(to,n,from,m):bn_32_set_0(to,n)
321 # if defined(L_ENDIAN)
322 #  if defined(__arch64__)
323 #   define NIST_INT64 long
324 #  else
325 #   define NIST_INT64 long long
326 #  endif
327 # endif
328 #else
329 #define bn_cp_64(to, n, from, m) \
330         { \
331         bn_cp_32(to, (n)*2, from, (m)*2); \
332         bn_cp_32(to, (n)*2+1, from, (m)*2+1); \
333         }
334 #define bn_64_set_0(to, n) \
335         { \
336         bn_32_set_0(to, (n)*2); \
337         bn_32_set_0(to, (n)*2+1); \
338         }
339 #define bn_cp_32(to, n, from, m)        (to)[n] = (m>=0)?((from)[m]):0;
340 #define bn_32_set_0(to, n)              (to)[n] = (BN_ULONG)0;
341 # if defined(_WIN32) && !defined(__GNUC__)
342 #  define NIST_INT64 __int64
343 # elif defined(BN_LLONG)
344 #  define NIST_INT64 long long
345 # endif
346 #endif /* BN_BITS2 != 64 */
347
348 #define nist_set_192(to, from, a1, a2, a3) \
349         { \
350         bn_cp_64(to, 0, from, (a3) - 3) \
351         bn_cp_64(to, 1, from, (a2) - 3) \
352         bn_cp_64(to, 2, from, (a1) - 3) \
353         }
354
355 int BN_nist_mod_192(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
356         BN_CTX *ctx)
357         {
358         int      top = a->top, i;
359         int      carry;
360         register BN_ULONG *r_d, *a_d = a->d;
361         union   {
362                 BN_ULONG        bn[BN_NIST_192_TOP];
363                 unsigned int    ui[BN_NIST_192_TOP*sizeof(BN_ULONG)/sizeof(unsigned int)];
364                 } buf;
365         BN_ULONG c_d[BN_NIST_192_TOP],
366                 *res;
367         PTR_SIZE_INT mask;
368         static const BIGNUM _bignum_nist_p_192_sqr = {
369                 (BN_ULONG *)_nist_p_192_sqr,
370                 sizeof(_nist_p_192_sqr)/sizeof(_nist_p_192_sqr[0]),
371                 sizeof(_nist_p_192_sqr)/sizeof(_nist_p_192_sqr[0]),
372                 0,BN_FLG_STATIC_DATA };
373
374         field = &_bignum_nist_p_192; /* just to make sure */
375
376         if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_192_sqr)>=0)
377                 return BN_nnmod(r, a, field, ctx);
378
379         i = BN_ucmp(field, a);
380         if (i == 0)
381                 {
382                 BN_zero(r);
383                 return 1;
384                 }
385         else if (i > 0)
386                 return (r == a) ? 1 : (BN_copy(r ,a) != NULL);
387
388         if (r != a)
389                 {
390                 if (!bn_wexpand(r, BN_NIST_192_TOP))
391                         return 0;
392                 r_d = r->d;
393                 nist_cp_bn(r_d, a_d, BN_NIST_192_TOP);
394                 }
395         else
396                 r_d = a_d;
397
398         nist_cp_bn_0(buf.bn, a_d + BN_NIST_192_TOP, top - BN_NIST_192_TOP, BN_NIST_192_TOP);
399
400 #if defined(NIST_INT64)
401         {
402         NIST_INT64              acc;    /* accumulator */
403         unsigned int            *rp=(unsigned int *)r_d;
404         const unsigned int      *bp=(const unsigned int *)buf.ui;
405
406         acc  = rp[0];   acc += bp[3*2-6];
407                         acc += bp[5*2-6]; rp[0] = (unsigned int)acc; acc >>= 32;
408
409         acc += rp[1];   acc += bp[3*2-5];
410                         acc += bp[5*2-5]; rp[1] = (unsigned int)acc; acc >>= 32;
411
412         acc += rp[2];   acc += bp[3*2-6];
413                         acc += bp[4*2-6];
414                         acc += bp[5*2-6]; rp[2] = (unsigned int)acc; acc >>= 32;
415
416         acc += rp[3];   acc += bp[3*2-5];
417                         acc += bp[4*2-5];
418                         acc += bp[5*2-5]; rp[3] = (unsigned int)acc; acc >>= 32;
419
420         acc += rp[4];   acc += bp[4*2-6];
421                         acc += bp[5*2-6]; rp[4] = (unsigned int)acc; acc >>= 32;
422
423         acc += rp[5];   acc += bp[4*2-5];
424                         acc += bp[5*2-5]; rp[5] = (unsigned int)acc;
425
426         carry = (int)(acc>>32);
427         }
428 #else
429         {
430         BN_ULONG t_d[BN_NIST_192_TOP];
431
432         nist_set_192(t_d, buf.bn, 0, 3, 3);
433         carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
434         nist_set_192(t_d, buf.bn, 4, 4, 0);
435         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
436         nist_set_192(t_d, buf.bn, 5, 5, 5)
437         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_192_TOP);
438         }
439 #endif
440         if (carry > 0)
441                 carry = (int)bn_sub_words(r_d,r_d,_nist_p_192[carry-1],BN_NIST_192_TOP);
442         else
443                 carry = 1;
444
445         /*
446          * we need 'if (carry==0 || result>=modulus) result-=modulus;'
447          * as comparison implies subtraction, we can write
448          * 'tmp=result-modulus; if (!carry || !borrow) result=tmp;'
449          * this is what happens below, but without explicit if:-) a.
450          */
451         mask  = 0-(PTR_SIZE_INT)bn_sub_words(c_d,r_d,_nist_p_192[0],BN_NIST_192_TOP);
452         mask &= 0-(PTR_SIZE_INT)carry;
453         res   = c_d;
454         res   = (BN_ULONG *)
455          (((PTR_SIZE_INT)res&~mask) | ((PTR_SIZE_INT)r_d&mask));
456         nist_cp_bn(r_d, res, BN_NIST_192_TOP);
457         r->top = BN_NIST_192_TOP;
458         bn_correct_top(r);
459
460         return 1;
461         }
462
463 typedef BN_ULONG (*bn_addsub_f)(BN_ULONG *,const BN_ULONG *,const BN_ULONG *,int);
464
465 #define nist_set_224(to, from, a1, a2, a3, a4, a5, a6, a7) \
466         { \
467         bn_cp_32(to, 0, from, (a7) - 7) \
468         bn_cp_32(to, 1, from, (a6) - 7) \
469         bn_cp_32(to, 2, from, (a5) - 7) \
470         bn_cp_32(to, 3, from, (a4) - 7) \
471         bn_cp_32(to, 4, from, (a3) - 7) \
472         bn_cp_32(to, 5, from, (a2) - 7) \
473         bn_cp_32(to, 6, from, (a1) - 7) \
474         }
475
476 int BN_nist_mod_224(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
477         BN_CTX *ctx)
478         {
479         int     top = a->top, i;
480         int     carry;
481         BN_ULONG *r_d, *a_d = a->d;
482         union   {
483                 BN_ULONG        bn[BN_NIST_224_TOP];
484                 unsigned int    ui[BN_NIST_224_TOP*sizeof(BN_ULONG)/sizeof(unsigned int)];
485                 } buf;
486         BN_ULONG c_d[BN_NIST_224_TOP],
487                 *res;
488         PTR_SIZE_INT mask;
489         union { bn_addsub_f f; PTR_SIZE_INT p; } u;
490         static const BIGNUM _bignum_nist_p_224_sqr = {
491                 (BN_ULONG *)_nist_p_224_sqr,
492                 sizeof(_nist_p_224_sqr)/sizeof(_nist_p_224_sqr[0]),
493                 sizeof(_nist_p_224_sqr)/sizeof(_nist_p_224_sqr[0]),
494                 0,BN_FLG_STATIC_DATA };
495
496
497         field = &_bignum_nist_p_224; /* just to make sure */
498
499         if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_224_sqr)>=0)
500                 return BN_nnmod(r, a, field, ctx);
501
502         i = BN_ucmp(field, a);
503         if (i == 0)
504                 {
505                 BN_zero(r);
506                 return 1;
507                 }
508         else if (i > 0)
509                 return (r == a)? 1 : (BN_copy(r ,a) != NULL);
510
511         if (r != a)
512                 {
513                 if (!bn_wexpand(r, BN_NIST_224_TOP))
514                         return 0;
515                 r_d = r->d;
516                 nist_cp_bn(r_d, a_d, BN_NIST_224_TOP);
517                 }
518         else
519                 r_d = a_d;
520
521 #if BN_BITS2==64
522         /* copy upper 256 bits of 448 bit number ... */
523         nist_cp_bn_0(c_d, a_d + (BN_NIST_224_TOP-1), top - (BN_NIST_224_TOP-1), BN_NIST_224_TOP);
524         /* ... and right shift by 32 to obtain upper 224 bits */
525         nist_set_224(buf.bn, c_d, 14, 13, 12, 11, 10, 9, 8);
526         /* truncate lower part to 224 bits too */
527         r_d[BN_NIST_224_TOP-1] &= BN_MASK2l;
528 #else
529         nist_cp_bn_0(buf.bn, a_d + BN_NIST_224_TOP, top - BN_NIST_224_TOP, BN_NIST_224_TOP);
530 #endif
531
532 #if defined(NIST_INT64) && BN_BITS2!=64
533         {
534         NIST_INT64              acc;    /* accumulator */
535         unsigned int            *rp=(unsigned int *)r_d;
536         const unsigned int      *bp=(const unsigned int *)buf.ui;
537
538         acc  = rp[0];   acc -= bp[7-7];
539                         acc -= bp[11-7]; rp[0] = (unsigned int)acc; acc >>= 32;
540
541         acc += rp[1];   acc -= bp[8-7];
542                         acc -= bp[12-7]; rp[1] = (unsigned int)acc; acc >>= 32;
543
544         acc += rp[2];   acc -= bp[9-7];
545                         acc -= bp[13-7]; rp[2] = (unsigned int)acc; acc >>= 32;
546
547         acc += rp[3];   acc += bp[7-7];
548                         acc += bp[11-7];
549                         acc -= bp[10-7]; rp[3] = (unsigned int)acc; acc>>= 32;
550
551         acc += rp[4];   acc += bp[8-7];
552                         acc += bp[12-7];
553                         acc -= bp[11-7]; rp[4] = (unsigned int)acc; acc >>= 32;
554
555         acc += rp[5];   acc += bp[9-7];
556                         acc += bp[13-7];
557                         acc -= bp[12-7]; rp[5] = (unsigned int)acc; acc >>= 32;
558
559         acc += rp[6];   acc += bp[10-7];
560                         acc -= bp[13-7]; rp[6] = (unsigned int)acc;
561
562         carry = (int)(acc>>32);
563 # if BN_BITS2==64
564         rp[7] = carry;
565 # endif
566         }       
567 #else
568         {
569         BN_ULONG t_d[BN_NIST_224_TOP];
570
571         nist_set_224(t_d, buf.bn, 10, 9, 8, 7, 0, 0, 0);
572         carry = (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
573         nist_set_224(t_d, buf.bn, 0, 13, 12, 11, 0, 0, 0);
574         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_224_TOP);
575         nist_set_224(t_d, buf.bn, 13, 12, 11, 10, 9, 8, 7);
576         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
577         nist_set_224(t_d, buf.bn, 0, 0, 0, 0, 13, 12, 11);
578         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_224_TOP);
579
580 #if BN_BITS2==64
581         carry = (int)(r_d[BN_NIST_224_TOP-1]>>32);
582 #endif
583         }
584 #endif
585         u.f = bn_sub_words;
586         if (carry > 0)
587                 {
588                 carry = (int)bn_sub_words(r_d,r_d,_nist_p_224[carry-1],BN_NIST_224_TOP);
589 #if BN_BITS2==64
590                 carry=(int)(~(r_d[BN_NIST_224_TOP-1]>>32))&1;
591 #endif
592                 }
593         else if (carry < 0)
594                 {
595                 /* it's a bit more comlicated logic in this case.
596                  * if bn_add_words yields no carry, then result
597                  * has to be adjusted by unconditionally *adding*
598                  * the modulus. but if it does, then result has
599                  * to be compared to the modulus and conditionally
600                  * adjusted by *subtracting* the latter. */
601                 carry = (int)bn_add_words(r_d,r_d,_nist_p_224[-carry-1],BN_NIST_224_TOP);
602                 mask = 0-(PTR_SIZE_INT)carry;
603                 u.p = ((PTR_SIZE_INT)bn_sub_words&mask) |
604                  ((PTR_SIZE_INT)bn_add_words&~mask);
605                 }
606         else
607                 carry = 1;
608
609         /* otherwise it's effectively same as in BN_nist_mod_192... */
610         mask  = 0-(PTR_SIZE_INT)(*u.f)(c_d,r_d,_nist_p_224[0],BN_NIST_224_TOP);
611         mask &= 0-(PTR_SIZE_INT)carry;
612         res   = c_d;
613         res   = (BN_ULONG *)(((PTR_SIZE_INT)res&~mask) |
614          ((PTR_SIZE_INT)r_d&mask));
615         nist_cp_bn(r_d, res, BN_NIST_224_TOP);
616         r->top = BN_NIST_224_TOP;
617         bn_correct_top(r);
618
619         return 1;
620         }
621
622 #define nist_set_256(to, from, a1, a2, a3, a4, a5, a6, a7, a8) \
623         { \
624         bn_cp_32(to, 0, from, (a8) - 8) \
625         bn_cp_32(to, 1, from, (a7) - 8) \
626         bn_cp_32(to, 2, from, (a6) - 8) \
627         bn_cp_32(to, 3, from, (a5) - 8) \
628         bn_cp_32(to, 4, from, (a4) - 8) \
629         bn_cp_32(to, 5, from, (a3) - 8) \
630         bn_cp_32(to, 6, from, (a2) - 8) \
631         bn_cp_32(to, 7, from, (a1) - 8) \
632         }
633
634 int BN_nist_mod_256(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
635         BN_CTX *ctx)
636         {
637         int     i, top = a->top;
638         int     carry = 0;
639         register BN_ULONG *a_d = a->d, *r_d;
640         union   {
641                 BN_ULONG bn[BN_NIST_256_TOP];
642                 unsigned int ui[BN_NIST_256_TOP*sizeof(BN_ULONG)/sizeof(unsigned int)];
643                 } buf;
644         BN_ULONG c_d[BN_NIST_256_TOP],
645                 *res;
646         PTR_SIZE_INT mask;
647         union { bn_addsub_f f; PTR_SIZE_INT p; } u;
648         static const BIGNUM _bignum_nist_p_256_sqr = {
649                 (BN_ULONG *)_nist_p_256_sqr,
650                 sizeof(_nist_p_256_sqr)/sizeof(_nist_p_256_sqr[0]),
651                 sizeof(_nist_p_256_sqr)/sizeof(_nist_p_256_sqr[0]),
652                 0,BN_FLG_STATIC_DATA };
653
654         field = &_bignum_nist_p_256; /* just to make sure */
655
656         if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_256_sqr)>=0)
657                 return BN_nnmod(r, a, field, ctx);
658
659         i = BN_ucmp(field, a);
660         if (i == 0)
661                 {
662                 BN_zero(r);
663                 return 1;
664                 }
665         else if (i > 0)
666                 return (r == a)? 1 : (BN_copy(r ,a) != NULL);
667
668         if (r != a)
669                 {
670                 if (!bn_wexpand(r, BN_NIST_256_TOP))
671                         return 0;
672                 r_d = r->d;
673                 nist_cp_bn(r_d, a_d, BN_NIST_256_TOP);
674                 }
675         else
676                 r_d = a_d;
677
678         nist_cp_bn_0(buf.bn, a_d + BN_NIST_256_TOP, top - BN_NIST_256_TOP, BN_NIST_256_TOP);
679
680 #if defined(NIST_INT64)
681         {
682         NIST_INT64              acc;    /* accumulator */
683         unsigned int            *rp=(unsigned int *)r_d;
684         const unsigned int      *bp=(const unsigned int *)buf.ui;
685
686         acc = rp[0];    acc += bp[8-8];
687                         acc += bp[9-8];
688                         acc -= bp[11-8];
689                         acc -= bp[12-8];
690                         acc -= bp[13-8];
691                         acc -= bp[14-8]; rp[0] = (unsigned int)acc; acc >>= 32;
692
693         acc += rp[1];   acc += bp[9-8];
694                         acc += bp[10-8];
695                         acc -= bp[12-8];
696                         acc -= bp[13-8];
697                         acc -= bp[14-8];
698                         acc -= bp[15-8]; rp[1] = (unsigned int)acc; acc >>= 32;
699
700         acc += rp[2];   acc += bp[10-8];
701                         acc += bp[11-8];
702                         acc -= bp[13-8];
703                         acc -= bp[14-8];
704                         acc -= bp[15-8]; rp[2] = (unsigned int)acc; acc >>= 32;
705
706         acc += rp[3];   acc += bp[11-8];
707                         acc += bp[11-8];
708                         acc += bp[12-8];
709                         acc += bp[12-8];
710                         acc += bp[13-8];
711                         acc -= bp[15-8];
712                         acc -= bp[8-8];
713                         acc -= bp[9-8];  rp[3] = (unsigned int)acc; acc >>= 32;
714
715         acc += rp[4];   acc += bp[12-8];
716                         acc += bp[12-8];
717                         acc += bp[13-8];
718                         acc += bp[13-8];
719                         acc += bp[14-8];
720                         acc -= bp[9-8];
721                         acc -= bp[10-8]; rp[4] = (unsigned int)acc; acc >>= 32;
722
723         acc += rp[5];   acc += bp[13-8];
724                         acc += bp[13-8];
725                         acc += bp[14-8];
726                         acc += bp[14-8];
727                         acc += bp[15-8];
728                         acc -= bp[10-8];
729                         acc -= bp[11-8]; rp[5] = (unsigned int)acc; acc >>= 32;
730
731         acc += rp[6];   acc += bp[14-8];
732                         acc += bp[14-8];
733                         acc += bp[15-8];
734                         acc += bp[15-8];
735                         acc += bp[14-8];
736                         acc += bp[13-8];
737                         acc -= bp[8-8];
738                         acc -= bp[9-8];  rp[6] = (unsigned int)acc; acc >>= 32;
739
740         acc += rp[7];   acc += bp[15-8];
741                         acc += bp[15-8];
742                         acc += bp[15-8];
743                         acc += bp[8 -8];
744                         acc -= bp[10-8];
745                         acc -= bp[11-8];
746                         acc -= bp[12-8];
747                         acc -= bp[13-8]; rp[7] = (unsigned int)acc;
748
749         carry = (int)(acc>>32);
750         }
751 #else
752         {
753         BN_ULONG t_d[BN_NIST_256_TOP];
754
755         /*S1*/
756         nist_set_256(t_d, buf.bn, 15, 14, 13, 12, 11, 0, 0, 0);
757         /*S2*/
758         nist_set_256(c_d, buf.bn, 0, 15, 14, 13, 12, 0, 0, 0);
759         carry = (int)bn_add_words(t_d, t_d, c_d, BN_NIST_256_TOP);
760         /* left shift */
761                 {
762                 register BN_ULONG *ap,t,c;
763                 ap = t_d;
764                 c=0;
765                 for (i = BN_NIST_256_TOP; i != 0; --i)
766                         {
767                         t= *ap;
768                         *(ap++)=((t<<1)|c)&BN_MASK2;
769                         c=(t & BN_TBIT)?1:0;
770                         }
771                 carry <<= 1;
772                 carry  |= c;
773                 }
774         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
775         /*S3*/
776         nist_set_256(t_d, buf.bn, 15, 14, 0, 0, 0, 10, 9, 8);
777         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
778         /*S4*/
779         nist_set_256(t_d, buf.bn, 8, 13, 15, 14, 13, 11, 10, 9);
780         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_256_TOP);
781         /*D1*/
782         nist_set_256(t_d, buf.bn, 10, 8, 0, 0, 0, 13, 12, 11);
783         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
784         /*D2*/
785         nist_set_256(t_d, buf.bn, 11, 9, 0, 0, 15, 14, 13, 12);
786         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
787         /*D3*/
788         nist_set_256(t_d, buf.bn, 12, 0, 10, 9, 8, 15, 14, 13);
789         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
790         /*D4*/
791         nist_set_256(t_d, buf.bn, 13, 0, 11, 10, 9, 0, 15, 14);
792         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_256_TOP);
793
794         }
795 #endif
796         /* see BN_nist_mod_224 for explanation */
797         u.f = bn_sub_words;
798         if (carry > 0)
799                 carry = (int)bn_sub_words(r_d,r_d,_nist_p_256[carry-1],BN_NIST_256_TOP);
800         else if (carry < 0)
801                 {
802                 carry = (int)bn_add_words(r_d,r_d,_nist_p_256[-carry-1],BN_NIST_256_TOP);
803                 mask = 0-(PTR_SIZE_INT)carry;
804                 u.p = ((PTR_SIZE_INT)bn_sub_words&mask) |
805                  ((PTR_SIZE_INT)bn_add_words&~mask);
806                 }
807         else
808                 carry = 1;
809
810         mask  = 0-(PTR_SIZE_INT)(*u.f)(c_d,r_d,_nist_p_256[0],BN_NIST_256_TOP);
811         mask &= 0-(PTR_SIZE_INT)carry;
812         res   = c_d;
813         res   = (BN_ULONG *)(((PTR_SIZE_INT)res&~mask) |
814          ((PTR_SIZE_INT)r_d&mask));
815         nist_cp_bn(r_d, res, BN_NIST_256_TOP);
816         r->top = BN_NIST_256_TOP;
817         bn_correct_top(r);
818
819         return 1;
820         }
821
822 #define nist_set_384(to,from,a1,a2,a3,a4,a5,a6,a7,a8,a9,a10,a11,a12) \
823         { \
824         bn_cp_32(to, 0, from,  (a12) - 12) \
825         bn_cp_32(to, 1, from,  (a11) - 12) \
826         bn_cp_32(to, 2, from,  (a10) - 12) \
827         bn_cp_32(to, 3, from,  (a9) - 12)  \
828         bn_cp_32(to, 4, from,  (a8) - 12)  \
829         bn_cp_32(to, 5, from,  (a7) - 12)  \
830         bn_cp_32(to, 6, from,  (a6) - 12)  \
831         bn_cp_32(to, 7, from,  (a5) - 12)  \
832         bn_cp_32(to, 8, from,  (a4) - 12)  \
833         bn_cp_32(to, 9, from,  (a3) - 12)  \
834         bn_cp_32(to, 10, from, (a2) - 12)  \
835         bn_cp_32(to, 11, from, (a1) - 12)  \
836         }
837
838 int BN_nist_mod_384(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
839         BN_CTX *ctx)
840         {
841         int     i, top = a->top;
842         int     carry = 0;
843         register BN_ULONG *r_d, *a_d = a->d;
844         union   {
845                 BN_ULONG bn[BN_NIST_384_TOP];
846                 unsigned int ui[BN_NIST_384_TOP*sizeof(BN_ULONG)/sizeof(unsigned int)];
847                 } buf;
848         BN_ULONG c_d[BN_NIST_384_TOP],
849                 *res;
850         PTR_SIZE_INT mask;
851         union { bn_addsub_f f; PTR_SIZE_INT p; } u;
852         static const BIGNUM _bignum_nist_p_384_sqr = {
853                 (BN_ULONG *)_nist_p_384_sqr,
854                 sizeof(_nist_p_384_sqr)/sizeof(_nist_p_384_sqr[0]),
855                 sizeof(_nist_p_384_sqr)/sizeof(_nist_p_384_sqr[0]),
856                 0,BN_FLG_STATIC_DATA };
857
858
859         field = &_bignum_nist_p_384; /* just to make sure */
860
861         if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_384_sqr)>=0)
862                 return BN_nnmod(r, a, field, ctx);
863
864         i = BN_ucmp(field, a);
865         if (i == 0)
866                 {
867                 BN_zero(r);
868                 return 1;
869                 }
870         else if (i > 0)
871                 return (r == a)? 1 : (BN_copy(r ,a) != NULL);
872
873         if (r != a)
874                 {
875                 if (!bn_wexpand(r, BN_NIST_384_TOP))
876                         return 0;
877                 r_d = r->d;
878                 nist_cp_bn(r_d, a_d, BN_NIST_384_TOP);
879                 }
880         else
881                 r_d = a_d;
882
883         nist_cp_bn_0(buf.bn, a_d + BN_NIST_384_TOP, top - BN_NIST_384_TOP, BN_NIST_384_TOP);
884
885 #if defined(NIST_INT64)
886         {
887         NIST_INT64              acc;    /* accumulator */
888         unsigned int            *rp=(unsigned int *)r_d;
889         const unsigned int      *bp=(const unsigned int *)buf.ui;
890
891         acc = rp[0];    acc += bp[12-12];
892                         acc += bp[21-12];
893                         acc += bp[20-12];
894                         acc -= bp[23-12]; rp[0] = (unsigned int)acc; acc >>= 32;
895
896         acc += rp[1];   acc += bp[13-12];
897                         acc += bp[22-12];
898                         acc += bp[23-12];
899                         acc -= bp[12-12];
900                         acc -= bp[20-12]; rp[1] = (unsigned int)acc; acc >>= 32;
901
902         acc += rp[2];   acc += bp[14-12];
903                         acc += bp[23-12];
904                         acc -= bp[13-12];
905                         acc -= bp[21-12]; rp[2] = (unsigned int)acc; acc >>= 32;
906
907         acc += rp[3];   acc += bp[15-12];
908                         acc += bp[12-12];
909                         acc += bp[20-12];
910                         acc += bp[21-12];
911                         acc -= bp[14-12];
912                         acc -= bp[22-12];
913                         acc -= bp[23-12]; rp[3] = (unsigned int)acc; acc >>= 32;
914
915         acc += rp[4];   acc += bp[21-12];
916                         acc += bp[21-12];
917                         acc += bp[16-12];
918                         acc += bp[13-12];
919                         acc += bp[12-12];
920                         acc += bp[20-12];
921                         acc += bp[22-12];
922                         acc -= bp[15-12];
923                         acc -= bp[23-12];
924                         acc -= bp[23-12]; rp[4] = (unsigned int)acc; acc >>= 32;
925
926         acc += rp[5];   acc += bp[22-12];
927                         acc += bp[22-12];
928                         acc += bp[17-12];
929                         acc += bp[14-12];
930                         acc += bp[13-12];
931                         acc += bp[21-12];
932                         acc += bp[23-12];
933                         acc -= bp[16-12]; rp[5] = (unsigned int)acc; acc >>= 32;
934                         
935         acc += rp[6];   acc += bp[23-12];
936                         acc += bp[23-12];
937                         acc += bp[18-12];
938                         acc += bp[15-12];
939                         acc += bp[14-12];
940                         acc += bp[22-12];
941                         acc -= bp[17-12]; rp[6] = (unsigned int)acc; acc >>= 32;
942                         
943         acc += rp[7];   acc += bp[19-12];
944                         acc += bp[16-12];
945                         acc += bp[15-12];
946                         acc += bp[23-12];
947                         acc -= bp[18-12]; rp[7] = (unsigned int)acc; acc >>= 32;
948                         
949         acc += rp[8];   acc += bp[20-12];
950                         acc += bp[17-12];
951                         acc += bp[16-12];
952                         acc -= bp[19-12]; rp[8] = (unsigned int)acc; acc >>= 32;
953                         
954         acc += rp[9];   acc += bp[21-12];
955                         acc += bp[18-12];
956                         acc += bp[17-12];
957                         acc -= bp[20-12]; rp[9] = (unsigned int)acc; acc >>= 32;
958                         
959         acc += rp[10];  acc += bp[22-12];
960                         acc += bp[19-12];
961                         acc += bp[18-12];
962                         acc -= bp[21-12]; rp[10] = (unsigned int)acc; acc >>= 32;
963                         
964         acc += rp[11];  acc += bp[23-12];
965                         acc += bp[20-12];
966                         acc += bp[19-12];
967                         acc -= bp[22-12]; rp[11] = (unsigned int)acc;
968
969         carry = (int)(acc>>32);
970         }
971 #else
972         {
973         BN_ULONG t_d[BN_NIST_384_TOP];
974
975         /*S1*/
976         nist_set_256(t_d, buf.bn, 0, 0, 0, 0, 0, 23-4, 22-4, 21-4);
977                 /* left shift */
978                 {
979                 register BN_ULONG *ap,t,c;
980                 ap = t_d;
981                 c=0;
982                 for (i = 3; i != 0; --i)
983                         {
984                         t= *ap;
985                         *(ap++)=((t<<1)|c)&BN_MASK2;
986                         c=(t & BN_TBIT)?1:0;
987                         }
988                 *ap=c;
989                 }
990         carry = (int)bn_add_words(r_d+(128/BN_BITS2), r_d+(128/BN_BITS2), 
991                 t_d, BN_NIST_256_TOP);
992         /*S2 */
993         carry += (int)bn_add_words(r_d, r_d, buf.bn, BN_NIST_384_TOP);
994         /*S3*/
995         nist_set_384(t_d,buf.bn,20,19,18,17,16,15,14,13,12,23,22,21);
996         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
997         /*S4*/
998         nist_set_384(t_d,buf.bn,19,18,17,16,15,14,13,12,20,0,23,0);
999         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1000         /*S5*/
1001         nist_set_384(t_d, buf.bn,0,0,0,0,23,22,21,20,0,0,0,0);
1002         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1003         /*S6*/
1004         nist_set_384(t_d,buf.bn,0,0,0,0,0,0,23,22,21,0,0,20);
1005         carry += (int)bn_add_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1006         /*D1*/
1007         nist_set_384(t_d,buf.bn,22,21,20,19,18,17,16,15,14,13,12,23);
1008         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1009         /*D2*/
1010         nist_set_384(t_d,buf.bn,0,0,0,0,0,0,0,23,22,21,20,0);
1011         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1012         /*D3*/
1013         nist_set_384(t_d,buf.bn,0,0,0,0,0,0,0,23,23,0,0,0);
1014         carry -= (int)bn_sub_words(r_d, r_d, t_d, BN_NIST_384_TOP);
1015
1016         }
1017 #endif
1018         /* see BN_nist_mod_224 for explanation */
1019         u.f = bn_sub_words;
1020         if (carry > 0)
1021                 carry = (int)bn_sub_words(r_d,r_d,_nist_p_384[carry-1],BN_NIST_384_TOP);
1022         else if (carry < 0)
1023                 {
1024                 carry = (int)bn_add_words(r_d,r_d,_nist_p_384[-carry-1],BN_NIST_384_TOP);
1025                 mask = 0-(PTR_SIZE_INT)carry;
1026                 u.p = ((PTR_SIZE_INT)bn_sub_words&mask) |
1027                  ((PTR_SIZE_INT)bn_add_words&~mask);
1028                 }
1029         else
1030                 carry = 1;
1031
1032         mask  = 0-(PTR_SIZE_INT)(*u.f)(c_d,r_d,_nist_p_384[0],BN_NIST_384_TOP);
1033         mask &= 0-(PTR_SIZE_INT)carry;
1034         res   = c_d;
1035         res   = (BN_ULONG *)(((PTR_SIZE_INT)res&~mask) |
1036          ((PTR_SIZE_INT)r_d&mask));
1037         nist_cp_bn(r_d, res, BN_NIST_384_TOP);
1038         r->top = BN_NIST_384_TOP;
1039         bn_correct_top(r);
1040
1041         return 1;
1042         }
1043
1044 #define BN_NIST_521_RSHIFT      (521%BN_BITS2)
1045 #define BN_NIST_521_LSHIFT      (BN_BITS2-BN_NIST_521_RSHIFT)
1046 #define BN_NIST_521_TOP_MASK    ((BN_ULONG)BN_MASK2>>BN_NIST_521_LSHIFT)
1047
1048 int BN_nist_mod_521(BIGNUM *r, const BIGNUM *a, const BIGNUM *field,
1049         BN_CTX *ctx)
1050         {
1051         int     top = a->top, i;
1052         BN_ULONG *r_d, *a_d = a->d,
1053                  t_d[BN_NIST_521_TOP],
1054                  val,tmp,*res;
1055         PTR_SIZE_INT mask;
1056         static const BIGNUM _bignum_nist_p_521_sqr = {
1057                 (BN_ULONG *)_nist_p_521_sqr,
1058                 sizeof(_nist_p_521_sqr)/sizeof(_nist_p_521_sqr[0]),
1059                 sizeof(_nist_p_521_sqr)/sizeof(_nist_p_521_sqr[0]),
1060                 0,BN_FLG_STATIC_DATA };
1061
1062         field = &_bignum_nist_p_521; /* just to make sure */
1063
1064         if (BN_is_negative(a) || BN_ucmp(a,&_bignum_nist_p_521_sqr)>=0)
1065                 return BN_nnmod(r, a, field, ctx);
1066
1067         i = BN_ucmp(field, a);
1068         if (i == 0)
1069                 {
1070                 BN_zero(r);
1071                 return 1;
1072                 }
1073         else if (i > 0)
1074                 return (r == a)? 1 : (BN_copy(r ,a) != NULL);
1075
1076         if (r != a)
1077                 {
1078                 if (!bn_wexpand(r,BN_NIST_521_TOP))
1079                         return 0;
1080                 r_d = r->d;
1081                 nist_cp_bn(r_d,a_d, BN_NIST_521_TOP);
1082                 }
1083         else
1084                 r_d = a_d;
1085
1086         /* upper 521 bits, copy ... */
1087         nist_cp_bn_0(t_d,a_d + (BN_NIST_521_TOP-1), top - (BN_NIST_521_TOP-1),BN_NIST_521_TOP);
1088         /* ... and right shift */
1089         for (val=t_d[0],i=0; i<BN_NIST_521_TOP-1; i++)
1090                 {
1091 #if 0
1092                 /*
1093                  * MSC ARM compiler [version 2013, presumably even earlier,
1094                  * much earlier] miscompiles this code, but not one in
1095                  * #else section. See RT#3541.
1096                  */
1097                 tmp = val>>BN_NIST_521_RSHIFT;
1098                 val = t_d[i+1];
1099                 t_d[i] = (tmp | val<<BN_NIST_521_LSHIFT) & BN_MASK2;
1100 #else
1101                 t_d[i] = ( val>>BN_NIST_521_RSHIFT |
1102                           (tmp=t_d[i+1])<<BN_NIST_521_LSHIFT ) & BN_MASK2;
1103                 val=tmp;
1104 #endif
1105                 }
1106         t_d[i] = val>>BN_NIST_521_RSHIFT;
1107         /* lower 521 bits */
1108         r_d[i] &= BN_NIST_521_TOP_MASK;
1109
1110         bn_add_words(r_d,r_d,t_d,BN_NIST_521_TOP);
1111         mask = 0-(PTR_SIZE_INT)bn_sub_words(t_d,r_d,_nist_p_521,BN_NIST_521_TOP);
1112         res  = t_d;
1113         res  = (BN_ULONG *)(((PTR_SIZE_INT)res&~mask) |
1114          ((PTR_SIZE_INT)r_d&mask));
1115         nist_cp_bn(r_d,res,BN_NIST_521_TOP);
1116         r->top = BN_NIST_521_TOP;
1117         bn_correct_top(r);
1118
1119         return 1;
1120         }
1121
1122 int (*BN_nist_mod_func(const BIGNUM *p))(BIGNUM *r, const BIGNUM *a, const BIGNUM *field, BN_CTX *ctx)
1123         {
1124         if (BN_ucmp(&_bignum_nist_p_192, p) == 0)
1125                 return BN_nist_mod_192;
1126         if (BN_ucmp(&_bignum_nist_p_224, p) == 0)
1127                 return BN_nist_mod_224;
1128         if (BN_ucmp(&_bignum_nist_p_256, p) == 0)
1129                 return BN_nist_mod_256;
1130         if (BN_ucmp(&_bignum_nist_p_384, p) == 0)
1131                 return BN_nist_mod_384;
1132         if (BN_ucmp(&_bignum_nist_p_521, p) == 0)
1133                 return BN_nist_mod_521;
1134         return 0;
1135         }