1 /*
2 * Core bignum functions
3 *
4 * Copyright The Mbed TLS Contributors
5 * SPDX-License-Identifier: Apache-2.0
6 *
7 * Licensed under the Apache License, Version 2.0 (the "License"); you may
8 * not use this file except in compliance with the License.
9 * You may obtain a copy of the License at
10 *
11 * http://www.apache.org/licenses/LICENSE-2.0
12 *
13 * Unless required by applicable law or agreed to in writing, software
14 * distributed under the License is distributed on an "AS IS" BASIS, WITHOUT
15 * WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
16 * See the License for the specific language governing permissions and
17 * limitations under the License.
18 */
19
20 #include "common.h"
21
22 #if defined(MBEDTLS_BIGNUM_C)
23
24 #include <string.h>
25
26 #include "mbedtls/error.h"
27 #include "mbedtls/platform_util.h"
28 #include "constant_time_internal.h"
29
30 #include "mbedtls/platform.h"
31
32 #include "bignum_core.h"
33 #include "bn_mul.h"
34 #include "constant_time_internal.h"
35
mbedtls_mpi_core_clz(mbedtls_mpi_uint a)36 size_t mbedtls_mpi_core_clz( mbedtls_mpi_uint a )
37 {
38 size_t j;
39 mbedtls_mpi_uint mask = (mbedtls_mpi_uint) 1 << (biL - 1);
40
41 for( j = 0; j < biL; j++ )
42 {
43 if( a & mask ) break;
44
45 mask >>= 1;
46 }
47
48 return( j );
49 }
50
mbedtls_mpi_core_bitlen(const mbedtls_mpi_uint * A,size_t A_limbs)51 size_t mbedtls_mpi_core_bitlen( const mbedtls_mpi_uint *A, size_t A_limbs )
52 {
53 size_t i, j;
54
55 if( A_limbs == 0 )
56 return( 0 );
57
58 for( i = A_limbs - 1; i > 0; i-- )
59 if( A[i] != 0 )
60 break;
61
62 j = biL - mbedtls_mpi_core_clz( A[i] );
63
64 return( ( i * biL ) + j );
65 }
66
67 /* Convert a big-endian byte array aligned to the size of mbedtls_mpi_uint
68 * into the storage form used by mbedtls_mpi. */
mpi_bigendian_to_host_c(mbedtls_mpi_uint a)69 static mbedtls_mpi_uint mpi_bigendian_to_host_c( mbedtls_mpi_uint a )
70 {
71 uint8_t i;
72 unsigned char *a_ptr;
73 mbedtls_mpi_uint tmp = 0;
74
75 for( i = 0, a_ptr = (unsigned char *) &a; i < ciL; i++, a_ptr++ )
76 {
77 tmp <<= CHAR_BIT;
78 tmp |= (mbedtls_mpi_uint) *a_ptr;
79 }
80
81 return( tmp );
82 }
83
mpi_bigendian_to_host(mbedtls_mpi_uint a)84 static mbedtls_mpi_uint mpi_bigendian_to_host( mbedtls_mpi_uint a )
85 {
86 #if defined(__BYTE_ORDER__)
87
88 /* Nothing to do on bigendian systems. */
89 #if ( __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ )
90 return( a );
91 #endif /* __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__ */
92
93 #if ( __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ )
94
95 /* For GCC and Clang, have builtins for byte swapping. */
96 #if defined(__GNUC__) && defined(__GNUC_PREREQ)
97 #if __GNUC_PREREQ(4,3)
98 #define have_bswap
99 #endif
100 #endif
101
102 #if defined(__clang__) && defined(__has_builtin)
103 #if __has_builtin(__builtin_bswap32) && \
104 __has_builtin(__builtin_bswap64)
105 #define have_bswap
106 #endif
107 #endif
108
109 #if defined(have_bswap)
110 /* The compiler is hopefully able to statically evaluate this! */
111 switch( sizeof(mbedtls_mpi_uint) )
112 {
113 case 4:
114 return( __builtin_bswap32(a) );
115 case 8:
116 return( __builtin_bswap64(a) );
117 }
118 #endif
119 #endif /* __BYTE_ORDER__ == __ORDER_LITTLE_ENDIAN__ */
120 #endif /* __BYTE_ORDER__ */
121
122 /* Fall back to C-based reordering if we don't know the byte order
123 * or we couldn't use a compiler-specific builtin. */
124 return( mpi_bigendian_to_host_c( a ) );
125 }
126
mbedtls_mpi_core_bigendian_to_host(mbedtls_mpi_uint * A,size_t A_limbs)127 void mbedtls_mpi_core_bigendian_to_host( mbedtls_mpi_uint *A,
128 size_t A_limbs )
129 {
130 mbedtls_mpi_uint *cur_limb_left;
131 mbedtls_mpi_uint *cur_limb_right;
132 if( A_limbs == 0 )
133 return;
134
135 /*
136 * Traverse limbs and
137 * - adapt byte-order in each limb
138 * - swap the limbs themselves.
139 * For that, simultaneously traverse the limbs from left to right
140 * and from right to left, as long as the left index is not bigger
141 * than the right index (it's not a problem if limbs is odd and the
142 * indices coincide in the last iteration).
143 */
144 for( cur_limb_left = A, cur_limb_right = A + ( A_limbs - 1 );
145 cur_limb_left <= cur_limb_right;
146 cur_limb_left++, cur_limb_right-- )
147 {
148 mbedtls_mpi_uint tmp;
149 /* Note that if cur_limb_left == cur_limb_right,
150 * this code effectively swaps the bytes only once. */
151 tmp = mpi_bigendian_to_host( *cur_limb_left );
152 *cur_limb_left = mpi_bigendian_to_host( *cur_limb_right );
153 *cur_limb_right = tmp;
154 }
155 }
156
mbedtls_mpi_core_cond_assign(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,size_t limbs,unsigned char assign)157 void mbedtls_mpi_core_cond_assign( mbedtls_mpi_uint *X,
158 const mbedtls_mpi_uint *A,
159 size_t limbs,
160 unsigned char assign )
161 {
162 if( X == A )
163 return;
164
165 mbedtls_ct_mpi_uint_cond_assign( limbs, X, A, assign );
166 }
167
mbedtls_mpi_core_cond_swap(mbedtls_mpi_uint * X,mbedtls_mpi_uint * Y,size_t limbs,unsigned char swap)168 void mbedtls_mpi_core_cond_swap( mbedtls_mpi_uint *X,
169 mbedtls_mpi_uint *Y,
170 size_t limbs,
171 unsigned char swap )
172 {
173 if( X == Y )
174 return;
175
176 /* all-bits 1 if swap is 1, all-bits 0 if swap is 0 */
177 mbedtls_mpi_uint limb_mask = mbedtls_ct_mpi_uint_mask( swap );
178
179 for( size_t i = 0; i < limbs; i++ )
180 {
181 mbedtls_mpi_uint tmp = X[i];
182 X[i] = ( X[i] & ~limb_mask ) | ( Y[i] & limb_mask );
183 Y[i] = ( Y[i] & ~limb_mask ) | ( tmp & limb_mask );
184 }
185 }
186
mbedtls_mpi_core_read_le(mbedtls_mpi_uint * X,size_t X_limbs,const unsigned char * input,size_t input_length)187 int mbedtls_mpi_core_read_le( mbedtls_mpi_uint *X,
188 size_t X_limbs,
189 const unsigned char *input,
190 size_t input_length )
191 {
192 const size_t limbs = CHARS_TO_LIMBS( input_length );
193
194 if( X_limbs < limbs )
195 return( MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL );
196
197 if( X != NULL )
198 {
199 memset( X, 0, X_limbs * ciL );
200
201 for( size_t i = 0; i < input_length; i++ )
202 {
203 size_t offset = ( ( i % ciL ) << 3 );
204 X[i / ciL] |= ( (mbedtls_mpi_uint) input[i] ) << offset;
205 }
206 }
207
208 return( 0 );
209 }
210
mbedtls_mpi_core_read_be(mbedtls_mpi_uint * X,size_t X_limbs,const unsigned char * input,size_t input_length)211 int mbedtls_mpi_core_read_be( mbedtls_mpi_uint *X,
212 size_t X_limbs,
213 const unsigned char *input,
214 size_t input_length )
215 {
216 const size_t limbs = CHARS_TO_LIMBS( input_length );
217
218 if( X_limbs < limbs )
219 return( MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL );
220
221 /* If X_limbs is 0, input_length must also be 0 (from previous test).
222 * Nothing to do. */
223 if( X_limbs == 0 )
224 return( 0 );
225
226 memset( X, 0, X_limbs * ciL );
227
228 /* memcpy() with (NULL, 0) is undefined behaviour */
229 if( input_length != 0 )
230 {
231 size_t overhead = ( X_limbs * ciL ) - input_length;
232 unsigned char *Xp = (unsigned char *) X;
233 memcpy( Xp + overhead, input, input_length );
234 }
235
236 mbedtls_mpi_core_bigendian_to_host( X, X_limbs );
237
238 return( 0 );
239 }
240
mbedtls_mpi_core_write_le(const mbedtls_mpi_uint * A,size_t A_limbs,unsigned char * output,size_t output_length)241 int mbedtls_mpi_core_write_le( const mbedtls_mpi_uint *A,
242 size_t A_limbs,
243 unsigned char *output,
244 size_t output_length )
245 {
246 size_t stored_bytes = A_limbs * ciL;
247 size_t bytes_to_copy;
248
249 if( stored_bytes < output_length )
250 {
251 bytes_to_copy = stored_bytes;
252 }
253 else
254 {
255 bytes_to_copy = output_length;
256
257 /* The output buffer is smaller than the allocated size of A.
258 * However A may fit if its leading bytes are zero. */
259 for( size_t i = bytes_to_copy; i < stored_bytes; i++ )
260 {
261 if( GET_BYTE( A, i ) != 0 )
262 return( MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL );
263 }
264 }
265
266 for( size_t i = 0; i < bytes_to_copy; i++ )
267 output[i] = GET_BYTE( A, i );
268
269 if( stored_bytes < output_length )
270 {
271 /* Write trailing 0 bytes */
272 memset( output + stored_bytes, 0, output_length - stored_bytes );
273 }
274
275 return( 0 );
276 }
277
mbedtls_mpi_core_write_be(const mbedtls_mpi_uint * X,size_t X_limbs,unsigned char * output,size_t output_length)278 int mbedtls_mpi_core_write_be( const mbedtls_mpi_uint *X,
279 size_t X_limbs,
280 unsigned char *output,
281 size_t output_length )
282 {
283 size_t stored_bytes;
284 size_t bytes_to_copy;
285 unsigned char *p;
286
287 stored_bytes = X_limbs * ciL;
288
289 if( stored_bytes < output_length )
290 {
291 /* There is enough space in the output buffer. Write initial
292 * null bytes and record the position at which to start
293 * writing the significant bytes. In this case, the execution
294 * trace of this function does not depend on the value of the
295 * number. */
296 bytes_to_copy = stored_bytes;
297 p = output + output_length - stored_bytes;
298 memset( output, 0, output_length - stored_bytes );
299 }
300 else
301 {
302 /* The output buffer is smaller than the allocated size of X.
303 * However X may fit if its leading bytes are zero. */
304 bytes_to_copy = output_length;
305 p = output;
306 for( size_t i = bytes_to_copy; i < stored_bytes; i++ )
307 {
308 if( GET_BYTE( X, i ) != 0 )
309 return( MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL );
310 }
311 }
312
313 for( size_t i = 0; i < bytes_to_copy; i++ )
314 p[bytes_to_copy - i - 1] = GET_BYTE( X, i );
315
316 return( 0 );
317 }
318
mbedtls_mpi_core_shift_r(mbedtls_mpi_uint * X,size_t limbs,size_t count)319 void mbedtls_mpi_core_shift_r( mbedtls_mpi_uint *X, size_t limbs,
320 size_t count )
321 {
322 size_t i, v0, v1;
323 mbedtls_mpi_uint r0 = 0, r1;
324
325 v0 = count / biL;
326 v1 = count & (biL - 1);
327
328 if( v0 > limbs || ( v0 == limbs && v1 > 0 ) )
329 {
330 memset( X, 0, limbs * ciL );
331 return;
332 }
333
334 /*
335 * shift by count / limb_size
336 */
337 if( v0 > 0 )
338 {
339 for( i = 0; i < limbs - v0; i++ )
340 X[i] = X[i + v0];
341
342 for( ; i < limbs; i++ )
343 X[i] = 0;
344 }
345
346 /*
347 * shift by count % limb_size
348 */
349 if( v1 > 0 )
350 {
351 for( i = limbs; i > 0; i-- )
352 {
353 r1 = X[i - 1] << (biL - v1);
354 X[i - 1] >>= v1;
355 X[i - 1] |= r0;
356 r0 = r1;
357 }
358 }
359 }
360
mbedtls_mpi_core_add(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t limbs)361 mbedtls_mpi_uint mbedtls_mpi_core_add( mbedtls_mpi_uint *X,
362 const mbedtls_mpi_uint *A,
363 const mbedtls_mpi_uint *B,
364 size_t limbs )
365 {
366 mbedtls_mpi_uint c = 0;
367
368 for( size_t i = 0; i < limbs; i++ )
369 {
370 mbedtls_mpi_uint t = c + A[i];
371 c = ( t < A[i] );
372 t += B[i];
373 c += ( t < B[i] );
374 X[i] = t;
375 }
376
377 return( c );
378 }
379
mbedtls_mpi_core_add_if(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,size_t limbs,unsigned cond)380 mbedtls_mpi_uint mbedtls_mpi_core_add_if( mbedtls_mpi_uint *X,
381 const mbedtls_mpi_uint *A,
382 size_t limbs,
383 unsigned cond )
384 {
385 mbedtls_mpi_uint c = 0;
386
387 /* all-bits 0 if cond is 0, all-bits 1 if cond is non-0 */
388 const mbedtls_mpi_uint mask = mbedtls_ct_mpi_uint_mask( cond );
389
390 for( size_t i = 0; i < limbs; i++ )
391 {
392 mbedtls_mpi_uint add = mask & A[i];
393 mbedtls_mpi_uint t = c + X[i];
394 c = ( t < X[i] );
395 t += add;
396 c += ( t < add );
397 X[i] = t;
398 }
399
400 return( c );
401 }
402
mbedtls_mpi_core_sub(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t limbs)403 mbedtls_mpi_uint mbedtls_mpi_core_sub( mbedtls_mpi_uint *X,
404 const mbedtls_mpi_uint *A,
405 const mbedtls_mpi_uint *B,
406 size_t limbs )
407 {
408 mbedtls_mpi_uint c = 0;
409
410 for( size_t i = 0; i < limbs; i++ )
411 {
412 mbedtls_mpi_uint z = ( A[i] < c );
413 mbedtls_mpi_uint t = A[i] - c;
414 c = ( t < B[i] ) + z;
415 X[i] = t - B[i];
416 }
417
418 return( c );
419 }
420
mbedtls_mpi_core_mla(mbedtls_mpi_uint * d,size_t d_len,const mbedtls_mpi_uint * s,size_t s_len,mbedtls_mpi_uint b)421 mbedtls_mpi_uint mbedtls_mpi_core_mla( mbedtls_mpi_uint *d, size_t d_len,
422 const mbedtls_mpi_uint *s, size_t s_len,
423 mbedtls_mpi_uint b )
424 {
425 mbedtls_mpi_uint c = 0; /* carry */
426 /*
427 * It is a documented precondition of this function that d_len >= s_len.
428 * If that's not the case, we swap these round: this turns what would be
429 * a buffer overflow into an incorrect result.
430 */
431 if( d_len < s_len )
432 s_len = d_len;
433 size_t excess_len = d_len - s_len;
434 size_t steps_x8 = s_len / 8;
435 size_t steps_x1 = s_len & 7;
436
437 while( steps_x8-- )
438 {
439 MULADDC_X8_INIT
440 MULADDC_X8_CORE
441 MULADDC_X8_STOP
442 }
443
444 while( steps_x1-- )
445 {
446 MULADDC_X1_INIT
447 MULADDC_X1_CORE
448 MULADDC_X1_STOP
449 }
450
451 while( excess_len-- )
452 {
453 *d += c;
454 c = ( *d < c );
455 d++;
456 }
457
458 return( c );
459 }
460
461 /*
462 * Fast Montgomery initialization (thanks to Tom St Denis).
463 */
mbedtls_mpi_core_montmul_init(const mbedtls_mpi_uint * N)464 mbedtls_mpi_uint mbedtls_mpi_core_montmul_init( const mbedtls_mpi_uint *N )
465 {
466 mbedtls_mpi_uint x = N[0];
467
468 x += ( ( N[0] + 2 ) & 4 ) << 1;
469
470 for( unsigned int i = biL; i >= 8; i /= 2 )
471 x *= ( 2 - ( N[0] * x ) );
472
473 return( ~x + 1 );
474 }
475
mbedtls_mpi_core_montmul(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t B_limbs,const mbedtls_mpi_uint * N,size_t AN_limbs,mbedtls_mpi_uint mm,mbedtls_mpi_uint * T)476 void mbedtls_mpi_core_montmul( mbedtls_mpi_uint *X,
477 const mbedtls_mpi_uint *A,
478 const mbedtls_mpi_uint *B,
479 size_t B_limbs,
480 const mbedtls_mpi_uint *N,
481 size_t AN_limbs,
482 mbedtls_mpi_uint mm,
483 mbedtls_mpi_uint *T )
484 {
485 memset( T, 0, ( 2 * AN_limbs + 1 ) * ciL );
486
487 for( size_t i = 0; i < AN_limbs; i++ )
488 {
489 /* T = (T + u0*B + u1*N) / 2^biL */
490 mbedtls_mpi_uint u0 = A[i];
491 mbedtls_mpi_uint u1 = ( T[0] + u0 * B[0] ) * mm;
492
493 (void) mbedtls_mpi_core_mla( T, AN_limbs + 2, B, B_limbs, u0 );
494 (void) mbedtls_mpi_core_mla( T, AN_limbs + 2, N, AN_limbs, u1 );
495
496 T++;
497 }
498
499 /*
500 * The result we want is (T >= N) ? T - N : T.
501 *
502 * For better constant-time properties in this function, we always do the
503 * subtraction, with the result in X.
504 *
505 * We also look to see if there was any carry in the final additions in the
506 * loop above.
507 */
508
509 mbedtls_mpi_uint carry = T[AN_limbs];
510 mbedtls_mpi_uint borrow = mbedtls_mpi_core_sub( X, T, N, AN_limbs );
511
512 /*
513 * Using R as the Montgomery radix (auxiliary modulus) i.e. 2^(biL*AN_limbs):
514 *
515 * T can be in one of 3 ranges:
516 *
517 * 1) T < N : (carry, borrow) = (0, 1): we want T
518 * 2) N <= T < R : (carry, borrow) = (0, 0): we want X
519 * 3) T >= R : (carry, borrow) = (1, 1): we want X
520 *
521 * and (carry, borrow) = (1, 0) can't happen.
522 *
523 * So the correct return value is already in X if (carry ^ borrow) = 0,
524 * but is in (the lower AN_limbs limbs of) T if (carry ^ borrow) = 1.
525 */
526 mbedtls_ct_mpi_uint_cond_assign( AN_limbs, X, T, (unsigned char) ( carry ^ borrow ) );
527 }
528
mbedtls_mpi_core_get_mont_r2_unsafe(mbedtls_mpi * X,const mbedtls_mpi * N)529 int mbedtls_mpi_core_get_mont_r2_unsafe( mbedtls_mpi *X,
530 const mbedtls_mpi *N )
531 {
532 int ret = MBEDTLS_ERR_ERROR_CORRUPTION_DETECTED;
533
534 MBEDTLS_MPI_CHK( mbedtls_mpi_lset( X, 1 ) );
535 MBEDTLS_MPI_CHK( mbedtls_mpi_shift_l( X, N->n * 2 * biL ) );
536 MBEDTLS_MPI_CHK( mbedtls_mpi_mod_mpi( X, X, N ) );
537 MBEDTLS_MPI_CHK( mbedtls_mpi_shrink( X, N->n ) );
538
539 cleanup:
540 return( ret );
541 }
542
543 MBEDTLS_STATIC_TESTABLE
mbedtls_mpi_core_ct_uint_table_lookup(mbedtls_mpi_uint * dest,const mbedtls_mpi_uint * table,size_t limbs,size_t count,size_t index)544 void mbedtls_mpi_core_ct_uint_table_lookup( mbedtls_mpi_uint *dest,
545 const mbedtls_mpi_uint *table,
546 size_t limbs,
547 size_t count,
548 size_t index )
549 {
550 for( size_t i = 0; i < count; i++, table += limbs )
551 {
552 unsigned char assign = mbedtls_ct_size_bool_eq( i, index );
553 mbedtls_mpi_core_cond_assign( dest, table, limbs, assign );
554 }
555 }
556
557 /* Fill X with n_bytes random bytes.
558 * X must already have room for those bytes.
559 * The ordering of the bytes returned from the RNG is suitable for
560 * deterministic ECDSA (see RFC 6979 §3.3 and the specification of
561 * mbedtls_mpi_core_random()).
562 */
mbedtls_mpi_core_fill_random(mbedtls_mpi_uint * X,size_t X_limbs,size_t n_bytes,int (* f_rng)(void *,unsigned char *,size_t),void * p_rng)563 int mbedtls_mpi_core_fill_random(
564 mbedtls_mpi_uint *X, size_t X_limbs,
565 size_t n_bytes,
566 int (*f_rng)(void *, unsigned char *, size_t), void *p_rng )
567 {
568 int ret = MBEDTLS_ERR_ERROR_CORRUPTION_DETECTED;
569 const size_t limbs = CHARS_TO_LIMBS( n_bytes );
570 const size_t overhead = ( limbs * ciL ) - n_bytes;
571
572 if( X_limbs < limbs )
573 return( MBEDTLS_ERR_MPI_BAD_INPUT_DATA );
574
575 memset( X, 0, overhead );
576 memset( (unsigned char *) X + limbs * ciL, 0, ( X_limbs - limbs ) * ciL );
577 MBEDTLS_MPI_CHK( f_rng( p_rng, (unsigned char *) X + overhead, n_bytes ) );
578 mbedtls_mpi_core_bigendian_to_host( X, limbs );
579
580 cleanup:
581 return( ret );
582 }
583
584 /* BEGIN MERGE SLOT 1 */
585
exp_mod_get_window_size(size_t Ebits)586 static size_t exp_mod_get_window_size( size_t Ebits )
587 {
588 size_t wsize = ( Ebits > 671 ) ? 6 : ( Ebits > 239 ) ? 5 :
589 ( Ebits > 79 ) ? 4 : 1;
590
591 #if( MBEDTLS_MPI_WINDOW_SIZE < 6 )
592 if( wsize > MBEDTLS_MPI_WINDOW_SIZE )
593 wsize = MBEDTLS_MPI_WINDOW_SIZE;
594 #endif
595
596 return( wsize );
597 }
598
mbedtls_mpi_core_exp_mod_working_limbs(size_t AN_limbs,size_t E_limbs)599 size_t mbedtls_mpi_core_exp_mod_working_limbs( size_t AN_limbs, size_t E_limbs )
600 {
601 const size_t wsize = exp_mod_get_window_size( E_limbs * biL );
602 const size_t welem = ( (size_t) 1 ) << wsize;
603
604 /* How big does each part of the working memory pool need to be? */
605 const size_t table_limbs = welem * AN_limbs;
606 const size_t select_limbs = AN_limbs;
607 const size_t temp_limbs = 2 * AN_limbs + 1;
608
609 return( table_limbs + select_limbs + temp_limbs );
610 }
611
exp_mod_precompute_window(const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * N,size_t AN_limbs,mbedtls_mpi_uint mm,const mbedtls_mpi_uint * RR,size_t welem,mbedtls_mpi_uint * Wtable,mbedtls_mpi_uint * temp)612 static void exp_mod_precompute_window( const mbedtls_mpi_uint *A,
613 const mbedtls_mpi_uint *N,
614 size_t AN_limbs,
615 mbedtls_mpi_uint mm,
616 const mbedtls_mpi_uint *RR,
617 size_t welem,
618 mbedtls_mpi_uint *Wtable,
619 mbedtls_mpi_uint *temp )
620 {
621 /* W[0] = 1 (in Montgomery presentation) */
622 memset( Wtable, 0, AN_limbs * ciL );
623 Wtable[0] = 1;
624 mbedtls_mpi_core_montmul( Wtable, Wtable, RR, AN_limbs, N, AN_limbs, mm, temp );
625
626 /* W[1] = A (already in Montgomery presentation) */
627 mbedtls_mpi_uint *W1 = Wtable + AN_limbs;
628 memcpy( W1, A, AN_limbs * ciL );
629
630 /* W[i+1] = W[i] * W[1], i >= 2 */
631 mbedtls_mpi_uint *Wprev = W1;
632 for( size_t i = 2; i < welem; i++ )
633 {
634 mbedtls_mpi_uint *Wcur = Wprev + AN_limbs;
635 mbedtls_mpi_core_montmul( Wcur, Wprev, W1, AN_limbs, N, AN_limbs, mm, temp );
636 Wprev = Wcur;
637 }
638 }
639
640 /* Exponentiation: X := A^E mod N.
641 *
642 * A must already be in Montgomery form.
643 *
644 * As in other bignum functions, assume that AN_limbs and E_limbs are nonzero.
645 *
646 * RR must contain 2^{2*biL} mod N.
647 *
648 * The algorithm is a variant of Left-to-right k-ary exponentiation: HAC 14.82
649 * (The difference is that the body in our loop processes a single bit instead
650 * of a full window.)
651 */
mbedtls_mpi_core_exp_mod(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * N,size_t AN_limbs,const mbedtls_mpi_uint * E,size_t E_limbs,const mbedtls_mpi_uint * RR,mbedtls_mpi_uint * T)652 void mbedtls_mpi_core_exp_mod( mbedtls_mpi_uint *X,
653 const mbedtls_mpi_uint *A,
654 const mbedtls_mpi_uint *N,
655 size_t AN_limbs,
656 const mbedtls_mpi_uint *E,
657 size_t E_limbs,
658 const mbedtls_mpi_uint *RR,
659 mbedtls_mpi_uint *T )
660 {
661 const size_t wsize = exp_mod_get_window_size( E_limbs * biL );
662 const size_t welem = ( (size_t) 1 ) << wsize;
663
664 /* This is how we will use the temporary storage T, which must have space
665 * for table_limbs, select_limbs and (2 * AN_limbs + 1) for montmul. */
666 const size_t table_limbs = welem * AN_limbs;
667 const size_t select_limbs = AN_limbs;
668
669 /* Pointers to specific parts of the temporary working memory pool */
670 mbedtls_mpi_uint *const Wtable = T;
671 mbedtls_mpi_uint *const Wselect = Wtable + table_limbs;
672 mbedtls_mpi_uint *const temp = Wselect + select_limbs;
673
674 /*
675 * Window precomputation
676 */
677
678 const mbedtls_mpi_uint mm = mbedtls_mpi_core_montmul_init( N );
679
680 /* Set Wtable[i] = A^(2^i) (in Montgomery representation) */
681 exp_mod_precompute_window( A, N, AN_limbs,
682 mm, RR,
683 welem, Wtable, temp );
684
685 /*
686 * Fixed window exponentiation
687 */
688
689 /* X = 1 (in Montgomery presentation) initially */
690 memcpy( X, Wtable, AN_limbs * ciL );
691
692 /* We'll process the bits of E from most significant
693 * (limb_index=E_limbs-1, E_bit_index=biL-1) to least significant
694 * (limb_index=0, E_bit_index=0). */
695 size_t E_limb_index = E_limbs;
696 size_t E_bit_index = 0;
697 /* At any given time, window contains window_bits bits from E.
698 * window_bits can go up to wsize. */
699 size_t window_bits = 0;
700 mbedtls_mpi_uint window = 0;
701
702 do
703 {
704 /* Square */
705 mbedtls_mpi_core_montmul( X, X, X, AN_limbs, N, AN_limbs, mm, temp );
706
707 /* Move to the next bit of the exponent */
708 if( E_bit_index == 0 )
709 {
710 --E_limb_index;
711 E_bit_index = biL - 1;
712 }
713 else
714 {
715 --E_bit_index;
716 }
717 /* Insert next exponent bit into window */
718 ++window_bits;
719 window <<= 1;
720 window |= ( E[E_limb_index] >> E_bit_index ) & 1;
721
722 /* Clear window if it's full. Also clear the window at the end,
723 * when we've finished processing the exponent. */
724 if( window_bits == wsize ||
725 ( E_bit_index == 0 && E_limb_index == 0 ) )
726 {
727 /* Select Wtable[window] without leaking window through
728 * memory access patterns. */
729 mbedtls_mpi_core_ct_uint_table_lookup( Wselect, Wtable,
730 AN_limbs, welem, window );
731 /* Multiply X by the selected element. */
732 mbedtls_mpi_core_montmul( X, X, Wselect, AN_limbs, N, AN_limbs, mm,
733 temp );
734 window = 0;
735 window_bits = 0;
736 }
737 }
738 while( ! ( E_bit_index == 0 && E_limb_index == 0 ) );
739 }
740
741 /* END MERGE SLOT 1 */
742
743 /* BEGIN MERGE SLOT 2 */
744
745 /* END MERGE SLOT 2 */
746
747 /* BEGIN MERGE SLOT 3 */
748
mbedtls_mpi_core_sub_int(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,mbedtls_mpi_uint c,size_t limbs)749 mbedtls_mpi_uint mbedtls_mpi_core_sub_int( mbedtls_mpi_uint *X,
750 const mbedtls_mpi_uint *A,
751 mbedtls_mpi_uint c, /* doubles as carry */
752 size_t limbs )
753 {
754 for( size_t i = 0; i < limbs; i++ )
755 {
756 mbedtls_mpi_uint s = A[i];
757 mbedtls_mpi_uint t = s - c;
758 c = ( t > s );
759 X[i] = t;
760 }
761
762 return( c );
763 }
764
765 /* END MERGE SLOT 3 */
766
767 /* BEGIN MERGE SLOT 4 */
768
769 /* END MERGE SLOT 4 */
770
771 /* BEGIN MERGE SLOT 5 */
772
773 /* END MERGE SLOT 5 */
774
775 /* BEGIN MERGE SLOT 6 */
776
777 /* END MERGE SLOT 6 */
778
779 /* BEGIN MERGE SLOT 7 */
780
781 /* END MERGE SLOT 7 */
782
783 /* BEGIN MERGE SLOT 8 */
784
785 /* END MERGE SLOT 8 */
786
787 /* BEGIN MERGE SLOT 9 */
788
789 /* END MERGE SLOT 9 */
790
791 /* BEGIN MERGE SLOT 10 */
792
793 /* END MERGE SLOT 10 */
794
795 #endif /* MBEDTLS_BIGNUM_C */
796