1 /*
2  *  Core bignum functions
3  *
4  *  Copyright The Mbed TLS Contributors
5  *  SPDX-License-Identifier: Apache-2.0 OR GPL-2.0-or-later
6  */
7 
8 #include "common.h"
9 
10 #if defined(MBEDTLS_BIGNUM_C)
11 
12 #include <string.h>
13 
14 #include "mbedtls/error.h"
15 #include "mbedtls/platform_util.h"
16 #include "constant_time_internal.h"
17 
18 #include "mbedtls/platform.h"
19 
20 #include "bignum_core.h"
21 #include "bn_mul.h"
22 #include "constant_time_internal.h"
23 
mbedtls_mpi_core_clz(mbedtls_mpi_uint a)24 size_t mbedtls_mpi_core_clz(mbedtls_mpi_uint a)
25 {
26 #if defined(__has_builtin)
27 #if (MBEDTLS_MPI_UINT_MAX == UINT_MAX) && __has_builtin(__builtin_clz)
28     #define core_clz __builtin_clz
29 #elif (MBEDTLS_MPI_UINT_MAX == ULONG_MAX) && __has_builtin(__builtin_clzl)
30     #define core_clz __builtin_clzl
31 #elif (MBEDTLS_MPI_UINT_MAX == ULLONG_MAX) && __has_builtin(__builtin_clzll)
32     #define core_clz __builtin_clzll
33 #endif
34 #endif
35 #if defined(core_clz)
36     return (size_t) core_clz(a);
37 #else
38     size_t j;
39     mbedtls_mpi_uint mask = (mbedtls_mpi_uint) 1 << (biL - 1);
40 
41     for (j = 0; j < biL; j++) {
42         if (a & mask) {
43             break;
44         }
45 
46         mask >>= 1;
47     }
48 
49     return j;
50 #endif
51 }
52 
mbedtls_mpi_core_bitlen(const mbedtls_mpi_uint * A,size_t A_limbs)53 size_t mbedtls_mpi_core_bitlen(const mbedtls_mpi_uint *A, size_t A_limbs)
54 {
55     int i;
56     size_t j;
57 
58     for (i = ((int) A_limbs) - 1; i >= 0; i--) {
59         if (A[i] != 0) {
60             j = biL - mbedtls_mpi_core_clz(A[i]);
61             return (i * biL) + j;
62         }
63     }
64 
65     return 0;
66 }
67 
mpi_bigendian_to_host(mbedtls_mpi_uint a)68 static mbedtls_mpi_uint mpi_bigendian_to_host(mbedtls_mpi_uint a)
69 {
70     if (MBEDTLS_IS_BIG_ENDIAN) {
71         /* Nothing to do on bigendian systems. */
72         return a;
73     } else {
74 #if defined(MBEDTLS_HAVE_INT32)
75         return (mbedtls_mpi_uint) MBEDTLS_BSWAP32(a);
76 #elif defined(MBEDTLS_HAVE_INT64)
77         return (mbedtls_mpi_uint) MBEDTLS_BSWAP64(a);
78 #endif
79     }
80 }
81 
mbedtls_mpi_core_bigendian_to_host(mbedtls_mpi_uint * A,size_t A_limbs)82 void mbedtls_mpi_core_bigendian_to_host(mbedtls_mpi_uint *A,
83                                         size_t A_limbs)
84 {
85     mbedtls_mpi_uint *cur_limb_left;
86     mbedtls_mpi_uint *cur_limb_right;
87     if (A_limbs == 0) {
88         return;
89     }
90 
91     /*
92      * Traverse limbs and
93      * - adapt byte-order in each limb
94      * - swap the limbs themselves.
95      * For that, simultaneously traverse the limbs from left to right
96      * and from right to left, as long as the left index is not bigger
97      * than the right index (it's not a problem if limbs is odd and the
98      * indices coincide in the last iteration).
99      */
100     for (cur_limb_left = A, cur_limb_right = A + (A_limbs - 1);
101          cur_limb_left <= cur_limb_right;
102          cur_limb_left++, cur_limb_right--) {
103         mbedtls_mpi_uint tmp;
104         /* Note that if cur_limb_left == cur_limb_right,
105          * this code effectively swaps the bytes only once. */
106         tmp             = mpi_bigendian_to_host(*cur_limb_left);
107         *cur_limb_left  = mpi_bigendian_to_host(*cur_limb_right);
108         *cur_limb_right = tmp;
109     }
110 }
111 
112 /* Whether min <= A, in constant time.
113  * A_limbs must be at least 1. */
mbedtls_mpi_core_uint_le_mpi(mbedtls_mpi_uint min,const mbedtls_mpi_uint * A,size_t A_limbs)114 mbedtls_ct_condition_t mbedtls_mpi_core_uint_le_mpi(mbedtls_mpi_uint min,
115                                                     const mbedtls_mpi_uint *A,
116                                                     size_t A_limbs)
117 {
118     /* min <= least significant limb? */
119     mbedtls_ct_condition_t min_le_lsl = mbedtls_ct_uint_ge(A[0], min);
120 
121     /* limbs other than the least significant one are all zero? */
122     mbedtls_ct_condition_t msll_mask = MBEDTLS_CT_FALSE;
123     for (size_t i = 1; i < A_limbs; i++) {
124         msll_mask = mbedtls_ct_bool_or(msll_mask, mbedtls_ct_bool(A[i]));
125     }
126 
127     /* min <= A iff the lowest limb of A is >= min or the other limbs
128      * are not all zero. */
129     return mbedtls_ct_bool_or(msll_mask, min_le_lsl);
130 }
131 
mbedtls_mpi_core_lt_ct(const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t limbs)132 mbedtls_ct_condition_t mbedtls_mpi_core_lt_ct(const mbedtls_mpi_uint *A,
133                                               const mbedtls_mpi_uint *B,
134                                               size_t limbs)
135 {
136     mbedtls_ct_condition_t ret = MBEDTLS_CT_FALSE, cond = MBEDTLS_CT_FALSE, done = MBEDTLS_CT_FALSE;
137 
138     for (size_t i = limbs; i > 0; i--) {
139         /*
140          * If B[i - 1] < A[i - 1] then A < B is false and the result must
141          * remain 0.
142          *
143          * Again even if we can make a decision, we just mark the result and
144          * the fact that we are done and continue looping.
145          */
146         cond = mbedtls_ct_uint_lt(B[i - 1], A[i - 1]);
147         done = mbedtls_ct_bool_or(done, cond);
148 
149         /*
150          * If A[i - 1] < B[i - 1] then A < B is true.
151          *
152          * Again even if we can make a decision, we just mark the result and
153          * the fact that we are done and continue looping.
154          */
155         cond = mbedtls_ct_uint_lt(A[i - 1], B[i - 1]);
156         ret  = mbedtls_ct_bool_or(ret, mbedtls_ct_bool_and(cond, mbedtls_ct_bool_not(done)));
157         done = mbedtls_ct_bool_or(done, cond);
158     }
159 
160     /*
161      * If all the limbs were equal, then the numbers are equal, A < B is false
162      * and leaving the result 0 is correct.
163      */
164 
165     return ret;
166 }
167 
mbedtls_mpi_core_cond_assign(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,size_t limbs,mbedtls_ct_condition_t assign)168 void mbedtls_mpi_core_cond_assign(mbedtls_mpi_uint *X,
169                                   const mbedtls_mpi_uint *A,
170                                   size_t limbs,
171                                   mbedtls_ct_condition_t assign)
172 {
173     if (X == A) {
174         return;
175     }
176 
177     /* This function is very performance-sensitive for RSA. For this reason
178      * we have the loop below, instead of calling mbedtls_ct_memcpy_if
179      * (this is more optimal since here we don't have to handle the case where
180      * we copy awkwardly sized data).
181      */
182     for (size_t i = 0; i < limbs; i++) {
183         X[i] = mbedtls_ct_mpi_uint_if(assign, A[i], X[i]);
184     }
185 }
186 
mbedtls_mpi_core_cond_swap(mbedtls_mpi_uint * X,mbedtls_mpi_uint * Y,size_t limbs,mbedtls_ct_condition_t swap)187 void mbedtls_mpi_core_cond_swap(mbedtls_mpi_uint *X,
188                                 mbedtls_mpi_uint *Y,
189                                 size_t limbs,
190                                 mbedtls_ct_condition_t swap)
191 {
192     if (X == Y) {
193         return;
194     }
195 
196     for (size_t i = 0; i < limbs; i++) {
197         mbedtls_mpi_uint tmp = X[i];
198         X[i] = mbedtls_ct_mpi_uint_if(swap, Y[i], X[i]);
199         Y[i] = mbedtls_ct_mpi_uint_if(swap, tmp, Y[i]);
200     }
201 }
202 
mbedtls_mpi_core_read_le(mbedtls_mpi_uint * X,size_t X_limbs,const unsigned char * input,size_t input_length)203 int mbedtls_mpi_core_read_le(mbedtls_mpi_uint *X,
204                              size_t X_limbs,
205                              const unsigned char *input,
206                              size_t input_length)
207 {
208     const size_t limbs = CHARS_TO_LIMBS(input_length);
209 
210     if (X_limbs < limbs) {
211         return MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL;
212     }
213 
214     if (X != NULL) {
215         memset(X, 0, X_limbs * ciL);
216 
217         for (size_t i = 0; i < input_length; i++) {
218             size_t offset = ((i % ciL) << 3);
219             X[i / ciL] |= ((mbedtls_mpi_uint) input[i]) << offset;
220         }
221     }
222 
223     return 0;
224 }
225 
mbedtls_mpi_core_read_be(mbedtls_mpi_uint * X,size_t X_limbs,const unsigned char * input,size_t input_length)226 int mbedtls_mpi_core_read_be(mbedtls_mpi_uint *X,
227                              size_t X_limbs,
228                              const unsigned char *input,
229                              size_t input_length)
230 {
231     const size_t limbs = CHARS_TO_LIMBS(input_length);
232 
233     if (X_limbs < limbs) {
234         return MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL;
235     }
236 
237     /* If X_limbs is 0, input_length must also be 0 (from previous test).
238      * Nothing to do. */
239     if (X_limbs == 0) {
240         return 0;
241     }
242 
243     memset(X, 0, X_limbs * ciL);
244 
245     /* memcpy() with (NULL, 0) is undefined behaviour */
246     if (input_length != 0) {
247         size_t overhead = (X_limbs * ciL) - input_length;
248         unsigned char *Xp = (unsigned char *) X;
249         memcpy(Xp + overhead, input, input_length);
250     }
251 
252     mbedtls_mpi_core_bigendian_to_host(X, X_limbs);
253 
254     return 0;
255 }
256 
mbedtls_mpi_core_write_le(const mbedtls_mpi_uint * A,size_t A_limbs,unsigned char * output,size_t output_length)257 int mbedtls_mpi_core_write_le(const mbedtls_mpi_uint *A,
258                               size_t A_limbs,
259                               unsigned char *output,
260                               size_t output_length)
261 {
262     size_t stored_bytes = A_limbs * ciL;
263     size_t bytes_to_copy;
264 
265     if (stored_bytes < output_length) {
266         bytes_to_copy = stored_bytes;
267     } else {
268         bytes_to_copy = output_length;
269 
270         /* The output buffer is smaller than the allocated size of A.
271          * However A may fit if its leading bytes are zero. */
272         for (size_t i = bytes_to_copy; i < stored_bytes; i++) {
273             if (GET_BYTE(A, i) != 0) {
274                 return MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL;
275             }
276         }
277     }
278 
279     for (size_t i = 0; i < bytes_to_copy; i++) {
280         output[i] = GET_BYTE(A, i);
281     }
282 
283     if (stored_bytes < output_length) {
284         /* Write trailing 0 bytes */
285         memset(output + stored_bytes, 0, output_length - stored_bytes);
286     }
287 
288     return 0;
289 }
290 
mbedtls_mpi_core_write_be(const mbedtls_mpi_uint * X,size_t X_limbs,unsigned char * output,size_t output_length)291 int mbedtls_mpi_core_write_be(const mbedtls_mpi_uint *X,
292                               size_t X_limbs,
293                               unsigned char *output,
294                               size_t output_length)
295 {
296     size_t stored_bytes;
297     size_t bytes_to_copy;
298     unsigned char *p;
299 
300     stored_bytes = X_limbs * ciL;
301 
302     if (stored_bytes < output_length) {
303         /* There is enough space in the output buffer. Write initial
304          * null bytes and record the position at which to start
305          * writing the significant bytes. In this case, the execution
306          * trace of this function does not depend on the value of the
307          * number. */
308         bytes_to_copy = stored_bytes;
309         p = output + output_length - stored_bytes;
310         memset(output, 0, output_length - stored_bytes);
311     } else {
312         /* The output buffer is smaller than the allocated size of X.
313          * However X may fit if its leading bytes are zero. */
314         bytes_to_copy = output_length;
315         p = output;
316         for (size_t i = bytes_to_copy; i < stored_bytes; i++) {
317             if (GET_BYTE(X, i) != 0) {
318                 return MBEDTLS_ERR_MPI_BUFFER_TOO_SMALL;
319             }
320         }
321     }
322 
323     for (size_t i = 0; i < bytes_to_copy; i++) {
324         p[bytes_to_copy - i - 1] = GET_BYTE(X, i);
325     }
326 
327     return 0;
328 }
329 
mbedtls_mpi_core_shift_r(mbedtls_mpi_uint * X,size_t limbs,size_t count)330 void mbedtls_mpi_core_shift_r(mbedtls_mpi_uint *X, size_t limbs,
331                               size_t count)
332 {
333     size_t i, v0, v1;
334     mbedtls_mpi_uint r0 = 0, r1;
335 
336     v0 = count /  biL;
337     v1 = count & (biL - 1);
338 
339     if (v0 > limbs || (v0 == limbs && v1 > 0)) {
340         memset(X, 0, limbs * ciL);
341         return;
342     }
343 
344     /*
345      * shift by count / limb_size
346      */
347     if (v0 > 0) {
348         for (i = 0; i < limbs - v0; i++) {
349             X[i] = X[i + v0];
350         }
351 
352         for (; i < limbs; i++) {
353             X[i] = 0;
354         }
355     }
356 
357     /*
358      * shift by count % limb_size
359      */
360     if (v1 > 0) {
361         for (i = limbs; i > 0; i--) {
362             r1 = X[i - 1] << (biL - v1);
363             X[i - 1] >>= v1;
364             X[i - 1] |= r0;
365             r0 = r1;
366         }
367     }
368 }
369 
mbedtls_mpi_core_shift_l(mbedtls_mpi_uint * X,size_t limbs,size_t count)370 void mbedtls_mpi_core_shift_l(mbedtls_mpi_uint *X, size_t limbs,
371                               size_t count)
372 {
373     size_t i, v0, v1;
374     mbedtls_mpi_uint r0 = 0, r1;
375 
376     v0 = count / (biL);
377     v1 = count & (biL - 1);
378 
379     /*
380      * shift by count / limb_size
381      */
382     if (v0 > 0) {
383         for (i = limbs; i > v0; i--) {
384             X[i - 1] = X[i - v0 - 1];
385         }
386 
387         for (; i > 0; i--) {
388             X[i - 1] = 0;
389         }
390     }
391 
392     /*
393      * shift by count % limb_size
394      */
395     if (v1 > 0) {
396         for (i = v0; i < limbs; i++) {
397             r1 = X[i] >> (biL - v1);
398             X[i] <<= v1;
399             X[i] |= r0;
400             r0 = r1;
401         }
402     }
403 }
404 
mbedtls_mpi_core_add(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t limbs)405 mbedtls_mpi_uint mbedtls_mpi_core_add(mbedtls_mpi_uint *X,
406                                       const mbedtls_mpi_uint *A,
407                                       const mbedtls_mpi_uint *B,
408                                       size_t limbs)
409 {
410     mbedtls_mpi_uint c = 0;
411 
412     for (size_t i = 0; i < limbs; i++) {
413         mbedtls_mpi_uint t = c + A[i];
414         c = (t < A[i]);
415         t += B[i];
416         c += (t < B[i]);
417         X[i] = t;
418     }
419 
420     return c;
421 }
422 
mbedtls_mpi_core_add_if(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,size_t limbs,unsigned cond)423 mbedtls_mpi_uint mbedtls_mpi_core_add_if(mbedtls_mpi_uint *X,
424                                          const mbedtls_mpi_uint *A,
425                                          size_t limbs,
426                                          unsigned cond)
427 {
428     mbedtls_mpi_uint c = 0;
429 
430     mbedtls_ct_condition_t do_add = mbedtls_ct_bool(cond);
431 
432     for (size_t i = 0; i < limbs; i++) {
433         mbedtls_mpi_uint add = mbedtls_ct_mpi_uint_if_else_0(do_add, A[i]);
434         mbedtls_mpi_uint t = c + X[i];
435         c = (t < X[i]);
436         t += add;
437         c += (t < add);
438         X[i] = t;
439     }
440 
441     return c;
442 }
443 
mbedtls_mpi_core_sub(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * B,size_t limbs)444 mbedtls_mpi_uint mbedtls_mpi_core_sub(mbedtls_mpi_uint *X,
445                                       const mbedtls_mpi_uint *A,
446                                       const mbedtls_mpi_uint *B,
447                                       size_t limbs)
448 {
449     mbedtls_mpi_uint c = 0;
450 
451     for (size_t i = 0; i < limbs; i++) {
452         mbedtls_mpi_uint z = (A[i] < c);
453         mbedtls_mpi_uint t = A[i] - c;
454         c = (t < B[i]) + z;
455         X[i] = t - B[i];
456     }
457 
458     return c;
459 }
460 
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)461 mbedtls_mpi_uint mbedtls_mpi_core_mla(mbedtls_mpi_uint *d, size_t d_len,
462                                       const mbedtls_mpi_uint *s, size_t s_len,
463                                       mbedtls_mpi_uint b)
464 {
465     mbedtls_mpi_uint c = 0; /* carry */
466     /*
467      * It is a documented precondition of this function that d_len >= s_len.
468      * If that's not the case, we swap these round: this turns what would be
469      * a buffer overflow into an incorrect result.
470      */
471     if (d_len < s_len) {
472         s_len = d_len;
473     }
474     size_t excess_len = d_len - s_len;
475     size_t steps_x8 = s_len / 8;
476     size_t steps_x1 = s_len & 7;
477 
478     while (steps_x8--) {
479         MULADDC_X8_INIT
480         MULADDC_X8_CORE
481             MULADDC_X8_STOP
482     }
483 
484     while (steps_x1--) {
485         MULADDC_X1_INIT
486         MULADDC_X1_CORE
487             MULADDC_X1_STOP
488     }
489 
490     while (excess_len--) {
491         *d += c;
492         c = (*d < c);
493         d++;
494     }
495 
496     return c;
497 }
498 
mbedtls_mpi_core_mul(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,size_t A_limbs,const mbedtls_mpi_uint * B,size_t B_limbs)499 void mbedtls_mpi_core_mul(mbedtls_mpi_uint *X,
500                           const mbedtls_mpi_uint *A, size_t A_limbs,
501                           const mbedtls_mpi_uint *B, size_t B_limbs)
502 {
503     memset(X, 0, (A_limbs + B_limbs) * ciL);
504 
505     for (size_t i = 0; i < B_limbs; i++) {
506         (void) mbedtls_mpi_core_mla(X + i, A_limbs + 1, A, A_limbs, B[i]);
507     }
508 }
509 
510 /*
511  * Fast Montgomery initialization (thanks to Tom St Denis).
512  */
mbedtls_mpi_core_montmul_init(const mbedtls_mpi_uint * N)513 mbedtls_mpi_uint mbedtls_mpi_core_montmul_init(const mbedtls_mpi_uint *N)
514 {
515     mbedtls_mpi_uint x = N[0];
516 
517     x += ((N[0] + 2) & 4) << 1;
518 
519     for (unsigned int i = biL; i >= 8; i /= 2) {
520         x *= (2 - (N[0] * x));
521     }
522 
523     return ~x + 1;
524 }
525 
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)526 void mbedtls_mpi_core_montmul(mbedtls_mpi_uint *X,
527                               const mbedtls_mpi_uint *A,
528                               const mbedtls_mpi_uint *B,
529                               size_t B_limbs,
530                               const mbedtls_mpi_uint *N,
531                               size_t AN_limbs,
532                               mbedtls_mpi_uint mm,
533                               mbedtls_mpi_uint *T)
534 {
535     memset(T, 0, (2 * AN_limbs + 1) * ciL);
536 
537     for (size_t i = 0; i < AN_limbs; i++) {
538         /* T = (T + u0*B + u1*N) / 2^biL */
539         mbedtls_mpi_uint u0 = A[i];
540         mbedtls_mpi_uint u1 = (T[0] + u0 * B[0]) * mm;
541 
542         (void) mbedtls_mpi_core_mla(T, AN_limbs + 2, B, B_limbs, u0);
543         (void) mbedtls_mpi_core_mla(T, AN_limbs + 2, N, AN_limbs, u1);
544 
545         T++;
546     }
547 
548     /*
549      * The result we want is (T >= N) ? T - N : T.
550      *
551      * For better constant-time properties in this function, we always do the
552      * subtraction, with the result in X.
553      *
554      * We also look to see if there was any carry in the final additions in the
555      * loop above.
556      */
557 
558     mbedtls_mpi_uint carry  = T[AN_limbs];
559     mbedtls_mpi_uint borrow = mbedtls_mpi_core_sub(X, T, N, AN_limbs);
560 
561     /*
562      * Using R as the Montgomery radix (auxiliary modulus) i.e. 2^(biL*AN_limbs):
563      *
564      * T can be in one of 3 ranges:
565      *
566      * 1) T < N      : (carry, borrow) = (0, 1): we want T
567      * 2) N <= T < R : (carry, borrow) = (0, 0): we want X
568      * 3) T >= R     : (carry, borrow) = (1, 1): we want X
569      *
570      * and (carry, borrow) = (1, 0) can't happen.
571      *
572      * So the correct return value is already in X if (carry ^ borrow) = 0,
573      * but is in (the lower AN_limbs limbs of) T if (carry ^ borrow) = 1.
574      */
575     mbedtls_ct_memcpy_if(mbedtls_ct_bool(carry ^ borrow),
576                          (unsigned char *) X,
577                          (unsigned char *) T,
578                          NULL,
579                          AN_limbs * sizeof(mbedtls_mpi_uint));
580 }
581 
mbedtls_mpi_core_get_mont_r2_unsafe(mbedtls_mpi * X,const mbedtls_mpi * N)582 int mbedtls_mpi_core_get_mont_r2_unsafe(mbedtls_mpi *X,
583                                         const mbedtls_mpi *N)
584 {
585     int ret = MBEDTLS_ERR_ERROR_CORRUPTION_DETECTED;
586 
587     MBEDTLS_MPI_CHK(mbedtls_mpi_lset(X, 1));
588     MBEDTLS_MPI_CHK(mbedtls_mpi_shift_l(X, N->n * 2 * biL));
589     MBEDTLS_MPI_CHK(mbedtls_mpi_mod_mpi(X, X, N));
590     MBEDTLS_MPI_CHK(mbedtls_mpi_shrink(X, N->n));
591 
592 cleanup:
593     return ret;
594 }
595 
596 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)597 void mbedtls_mpi_core_ct_uint_table_lookup(mbedtls_mpi_uint *dest,
598                                            const mbedtls_mpi_uint *table,
599                                            size_t limbs,
600                                            size_t count,
601                                            size_t index)
602 {
603     for (size_t i = 0; i < count; i++, table += limbs) {
604         mbedtls_ct_condition_t assign = mbedtls_ct_uint_eq(i, index);
605         mbedtls_mpi_core_cond_assign(dest, table, limbs, assign);
606     }
607 }
608 
609 /* Fill X with n_bytes random bytes.
610  * X must already have room for those bytes.
611  * The ordering of the bytes returned from the RNG is suitable for
612  * deterministic ECDSA (see RFC 6979 §3.3 and the specification of
613  * mbedtls_mpi_core_random()).
614  */
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)615 int mbedtls_mpi_core_fill_random(
616     mbedtls_mpi_uint *X, size_t X_limbs,
617     size_t n_bytes,
618     int (*f_rng)(void *, unsigned char *, size_t), void *p_rng)
619 {
620     int ret = MBEDTLS_ERR_ERROR_CORRUPTION_DETECTED;
621     const size_t limbs = CHARS_TO_LIMBS(n_bytes);
622     const size_t overhead = (limbs * ciL) - n_bytes;
623 
624     if (X_limbs < limbs) {
625         return MBEDTLS_ERR_MPI_BAD_INPUT_DATA;
626     }
627 
628     memset(X, 0, overhead);
629     memset((unsigned char *) X + limbs * ciL, 0, (X_limbs - limbs) * ciL);
630     MBEDTLS_MPI_CHK(f_rng(p_rng, (unsigned char *) X + overhead, n_bytes));
631     mbedtls_mpi_core_bigendian_to_host(X, limbs);
632 
633 cleanup:
634     return ret;
635 }
636 
mbedtls_mpi_core_random(mbedtls_mpi_uint * X,mbedtls_mpi_uint min,const mbedtls_mpi_uint * N,size_t limbs,int (* f_rng)(void *,unsigned char *,size_t),void * p_rng)637 int mbedtls_mpi_core_random(mbedtls_mpi_uint *X,
638                             mbedtls_mpi_uint min,
639                             const mbedtls_mpi_uint *N,
640                             size_t limbs,
641                             int (*f_rng)(void *, unsigned char *, size_t),
642                             void *p_rng)
643 {
644     mbedtls_ct_condition_t ge_lower = MBEDTLS_CT_TRUE, lt_upper = MBEDTLS_CT_FALSE;
645     size_t n_bits = mbedtls_mpi_core_bitlen(N, limbs);
646     size_t n_bytes = (n_bits + 7) / 8;
647     int ret = MBEDTLS_ERR_ERROR_CORRUPTION_DETECTED;
648 
649     /*
650      * When min == 0, each try has at worst a probability 1/2 of failing
651      * (the msb has a probability 1/2 of being 0, and then the result will
652      * be < N), so after 30 tries failure probability is a most 2**(-30).
653      *
654      * When N is just below a power of 2, as is the case when generating
655      * a random scalar on most elliptic curves, 1 try is enough with
656      * overwhelming probability. When N is just above a power of 2,
657      * as when generating a random scalar on secp224k1, each try has
658      * a probability of failing that is almost 1/2.
659      *
660      * The probabilities are almost the same if min is nonzero but negligible
661      * compared to N. This is always the case when N is crypto-sized, but
662      * it's convenient to support small N for testing purposes. When N
663      * is small, use a higher repeat count, otherwise the probability of
664      * failure is macroscopic.
665      */
666     int count = (n_bytes > 4 ? 30 : 250);
667 
668     /*
669      * Match the procedure given in RFC 6979 §3.3 (deterministic ECDSA)
670      * when f_rng is a suitably parametrized instance of HMAC_DRBG:
671      * - use the same byte ordering;
672      * - keep the leftmost n_bits bits of the generated octet string;
673      * - try until result is in the desired range.
674      * This also avoids any bias, which is especially important for ECDSA.
675      */
676     do {
677         MBEDTLS_MPI_CHK(mbedtls_mpi_core_fill_random(X, limbs,
678                                                      n_bytes,
679                                                      f_rng, p_rng));
680         mbedtls_mpi_core_shift_r(X, limbs, 8 * n_bytes - n_bits);
681 
682         if (--count == 0) {
683             ret = MBEDTLS_ERR_MPI_NOT_ACCEPTABLE;
684             goto cleanup;
685         }
686 
687         ge_lower = mbedtls_mpi_core_uint_le_mpi(min, X, limbs);
688         lt_upper = mbedtls_mpi_core_lt_ct(X, N, limbs);
689     } while (mbedtls_ct_bool_and(ge_lower, lt_upper) == MBEDTLS_CT_FALSE);
690 
691 cleanup:
692     return ret;
693 }
694 
exp_mod_get_window_size(size_t Ebits)695 static size_t exp_mod_get_window_size(size_t Ebits)
696 {
697 #if MBEDTLS_MPI_WINDOW_SIZE >= 6
698     return (Ebits > 671) ? 6 : (Ebits > 239) ? 5 : (Ebits >  79) ? 4 : 1;
699 #elif MBEDTLS_MPI_WINDOW_SIZE == 5
700     return (Ebits > 239) ? 5 : (Ebits >  79) ? 4 : 1;
701 #elif MBEDTLS_MPI_WINDOW_SIZE > 1
702     return (Ebits >  79) ? MBEDTLS_MPI_WINDOW_SIZE : 1;
703 #else
704     (void) Ebits;
705     return 1;
706 #endif
707 }
708 
mbedtls_mpi_core_exp_mod_working_limbs(size_t AN_limbs,size_t E_limbs)709 size_t mbedtls_mpi_core_exp_mod_working_limbs(size_t AN_limbs, size_t E_limbs)
710 {
711     const size_t wsize = exp_mod_get_window_size(E_limbs * biL);
712     const size_t welem = ((size_t) 1) << wsize;
713 
714     /* How big does each part of the working memory pool need to be? */
715     const size_t table_limbs   = welem * AN_limbs;
716     const size_t select_limbs  = AN_limbs;
717     const size_t temp_limbs    = 2 * AN_limbs + 1;
718 
719     return table_limbs + select_limbs + temp_limbs;
720 }
721 
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)722 static void exp_mod_precompute_window(const mbedtls_mpi_uint *A,
723                                       const mbedtls_mpi_uint *N,
724                                       size_t AN_limbs,
725                                       mbedtls_mpi_uint mm,
726                                       const mbedtls_mpi_uint *RR,
727                                       size_t welem,
728                                       mbedtls_mpi_uint *Wtable,
729                                       mbedtls_mpi_uint *temp)
730 {
731     /* W[0] = 1 (in Montgomery presentation) */
732     memset(Wtable, 0, AN_limbs * ciL);
733     Wtable[0] = 1;
734     mbedtls_mpi_core_montmul(Wtable, Wtable, RR, AN_limbs, N, AN_limbs, mm, temp);
735 
736     /* W[1] = A (already in Montgomery presentation) */
737     mbedtls_mpi_uint *W1 = Wtable + AN_limbs;
738     memcpy(W1, A, AN_limbs * ciL);
739 
740     /* W[i+1] = W[i] * W[1], i >= 2 */
741     mbedtls_mpi_uint *Wprev = W1;
742     for (size_t i = 2; i < welem; i++) {
743         mbedtls_mpi_uint *Wcur = Wprev + AN_limbs;
744         mbedtls_mpi_core_montmul(Wcur, Wprev, W1, AN_limbs, N, AN_limbs, mm, temp);
745         Wprev = Wcur;
746     }
747 }
748 
749 #if defined(MBEDTLS_TEST_HOOKS) && !defined(MBEDTLS_THREADING_C)
750 // Set to a default that is neither MBEDTLS_MPI_IS_PUBLIC nor MBEDTLS_MPI_IS_SECRET
751 int mbedtls_mpi_optionally_safe_codepath = MBEDTLS_MPI_IS_PUBLIC + MBEDTLS_MPI_IS_SECRET + 1;
752 #endif
753 
754 /*
755  * This function calculates the indices of the exponent where the exponentiation algorithm should
756  * start processing.
757  *
758  * Warning! If the parameter E_public has MBEDTLS_MPI_IS_PUBLIC as its value,
759  * this function is not constant time with respect to the exponent (parameter E).
760  */
exp_mod_calc_first_bit_optionally_safe(const mbedtls_mpi_uint * E,size_t E_limbs,int E_public,size_t * E_limb_index,size_t * E_bit_index)761 static inline void exp_mod_calc_first_bit_optionally_safe(const mbedtls_mpi_uint *E,
762                                                           size_t E_limbs,
763                                                           int E_public,
764                                                           size_t *E_limb_index,
765                                                           size_t *E_bit_index)
766 {
767     if (E_public == MBEDTLS_MPI_IS_PUBLIC) {
768         /*
769          * Skip leading zero bits.
770          */
771         size_t E_bits = mbedtls_mpi_core_bitlen(E, E_limbs);
772         if (E_bits == 0) {
773             /*
774              * If E is 0 mbedtls_mpi_core_bitlen() returns 0. Even if that is the case, we will want
775              * to represent it as a single 0 bit and as such the bitlength will be 1.
776              */
777             E_bits = 1;
778         }
779 
780         *E_limb_index = E_bits / biL;
781         *E_bit_index = E_bits % biL;
782 
783 #if defined(MBEDTLS_TEST_HOOKS) && !defined(MBEDTLS_THREADING_C)
784         mbedtls_mpi_optionally_safe_codepath = MBEDTLS_MPI_IS_PUBLIC;
785 #endif
786     } else {
787         /*
788          * Here we need to be constant time with respect to E and can't do anything better than
789          * start at the first allocated bit.
790          */
791         *E_limb_index = E_limbs;
792         *E_bit_index = 0;
793 #if defined(MBEDTLS_TEST_HOOKS) && !defined(MBEDTLS_THREADING_C)
794         // Only mark the codepath safe if there wasn't an unsafe codepath before
795         if (mbedtls_mpi_optionally_safe_codepath != MBEDTLS_MPI_IS_PUBLIC) {
796             mbedtls_mpi_optionally_safe_codepath = MBEDTLS_MPI_IS_SECRET;
797         }
798 #endif
799     }
800 }
801 
802 /*
803  * Warning! If the parameter window_public has MBEDTLS_MPI_IS_PUBLIC as its value, this function is
804  * not constant time with respect to the window parameter and consequently the exponent of the
805  * exponentiation (parameter E of mbedtls_mpi_core_exp_mod_optionally_safe).
806  */
exp_mod_table_lookup_optionally_safe(mbedtls_mpi_uint * Wselect,mbedtls_mpi_uint * Wtable,size_t AN_limbs,size_t welem,mbedtls_mpi_uint window,int window_public)807 static inline void exp_mod_table_lookup_optionally_safe(mbedtls_mpi_uint *Wselect,
808                                                         mbedtls_mpi_uint *Wtable,
809                                                         size_t AN_limbs, size_t welem,
810                                                         mbedtls_mpi_uint window,
811                                                         int window_public)
812 {
813     if (window_public == MBEDTLS_MPI_IS_PUBLIC) {
814         memcpy(Wselect, Wtable + window * AN_limbs, AN_limbs * ciL);
815 #if defined(MBEDTLS_TEST_HOOKS) && !defined(MBEDTLS_THREADING_C)
816         mbedtls_mpi_optionally_safe_codepath = MBEDTLS_MPI_IS_PUBLIC;
817 #endif
818     } else {
819         /* Select Wtable[window] without leaking window through
820          * memory access patterns. */
821         mbedtls_mpi_core_ct_uint_table_lookup(Wselect, Wtable,
822                                               AN_limbs, welem, window);
823 #if defined(MBEDTLS_TEST_HOOKS) && !defined(MBEDTLS_THREADING_C)
824         // Only mark the codepath safe if there wasn't an unsafe codepath before
825         if (mbedtls_mpi_optionally_safe_codepath != MBEDTLS_MPI_IS_PUBLIC) {
826             mbedtls_mpi_optionally_safe_codepath = MBEDTLS_MPI_IS_SECRET;
827         }
828 #endif
829     }
830 }
831 
832 /* Exponentiation: X := A^E mod N.
833  *
834  * Warning! If the parameter E_public has MBEDTLS_MPI_IS_PUBLIC as its value,
835  * this function is not constant time with respect to the exponent (parameter E).
836  *
837  * A must already be in Montgomery form.
838  *
839  * As in other bignum functions, assume that AN_limbs and E_limbs are nonzero.
840  *
841  * RR must contain 2^{2*biL} mod N.
842  *
843  * The algorithm is a variant of Left-to-right k-ary exponentiation: HAC 14.82
844  * (The difference is that the body in our loop processes a single bit instead
845  * of a full window.)
846  */
mbedtls_mpi_core_exp_mod_optionally_safe(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,int E_public,const mbedtls_mpi_uint * RR,mbedtls_mpi_uint * T)847 static void mbedtls_mpi_core_exp_mod_optionally_safe(mbedtls_mpi_uint *X,
848                                                      const mbedtls_mpi_uint *A,
849                                                      const mbedtls_mpi_uint *N,
850                                                      size_t AN_limbs,
851                                                      const mbedtls_mpi_uint *E,
852                                                      size_t E_limbs,
853                                                      int E_public,
854                                                      const mbedtls_mpi_uint *RR,
855                                                      mbedtls_mpi_uint *T)
856 {
857     /* We'll process the bits of E from most significant
858      * (limb_index=E_limbs-1, E_bit_index=biL-1) to least significant
859      * (limb_index=0, E_bit_index=0). */
860     size_t E_limb_index;
861     size_t E_bit_index;
862     exp_mod_calc_first_bit_optionally_safe(E, E_limbs, E_public,
863                                            &E_limb_index, &E_bit_index);
864 
865     const size_t wsize = exp_mod_get_window_size(E_limb_index * biL);
866     const size_t welem = ((size_t) 1) << wsize;
867 
868     /* This is how we will use the temporary storage T, which must have space
869      * for table_limbs, select_limbs and (2 * AN_limbs + 1) for montmul. */
870     const size_t table_limbs  = welem * AN_limbs;
871     const size_t select_limbs = AN_limbs;
872 
873     /* Pointers to specific parts of the temporary working memory pool */
874     mbedtls_mpi_uint *const Wtable  = T;
875     mbedtls_mpi_uint *const Wselect = Wtable  +  table_limbs;
876     mbedtls_mpi_uint *const temp    = Wselect + select_limbs;
877 
878     /*
879      * Window precomputation
880      */
881 
882     const mbedtls_mpi_uint mm = mbedtls_mpi_core_montmul_init(N);
883 
884     /* Set Wtable[i] = A^i (in Montgomery representation) */
885     exp_mod_precompute_window(A, N, AN_limbs,
886                               mm, RR,
887                               welem, Wtable, temp);
888 
889     /*
890      * Fixed window exponentiation
891      */
892 
893     /* X = 1 (in Montgomery presentation) initially */
894     memcpy(X, Wtable, AN_limbs * ciL);
895 
896     /* At any given time, window contains window_bits bits from E.
897      * window_bits can go up to wsize. */
898     size_t window_bits = 0;
899     mbedtls_mpi_uint window = 0;
900 
901     do {
902         /* Square */
903         mbedtls_mpi_core_montmul(X, X, X, AN_limbs, N, AN_limbs, mm, temp);
904 
905         /* Move to the next bit of the exponent */
906         if (E_bit_index == 0) {
907             --E_limb_index;
908             E_bit_index = biL - 1;
909         } else {
910             --E_bit_index;
911         }
912         /* Insert next exponent bit into window */
913         ++window_bits;
914         window <<= 1;
915         window |= (E[E_limb_index] >> E_bit_index) & 1;
916 
917         /* Clear window if it's full. Also clear the window at the end,
918          * when we've finished processing the exponent. */
919         if (window_bits == wsize ||
920             (E_bit_index == 0 && E_limb_index == 0)) {
921 
922             exp_mod_table_lookup_optionally_safe(Wselect, Wtable, AN_limbs, welem,
923                                                  window, E_public);
924             /* Multiply X by the selected element. */
925             mbedtls_mpi_core_montmul(X, X, Wselect, AN_limbs, N, AN_limbs, mm,
926                                      temp);
927             window = 0;
928             window_bits = 0;
929         }
930     } while (!(E_bit_index == 0 && E_limb_index == 0));
931 }
932 
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)933 void mbedtls_mpi_core_exp_mod(mbedtls_mpi_uint *X,
934                               const mbedtls_mpi_uint *A,
935                               const mbedtls_mpi_uint *N, size_t AN_limbs,
936                               const mbedtls_mpi_uint *E, size_t E_limbs,
937                               const mbedtls_mpi_uint *RR,
938                               mbedtls_mpi_uint *T)
939 {
940     mbedtls_mpi_core_exp_mod_optionally_safe(X,
941                                              A,
942                                              N,
943                                              AN_limbs,
944                                              E,
945                                              E_limbs,
946                                              MBEDTLS_MPI_IS_SECRET,
947                                              RR,
948                                              T);
949 }
950 
mbedtls_mpi_core_exp_mod_unsafe(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)951 void mbedtls_mpi_core_exp_mod_unsafe(mbedtls_mpi_uint *X,
952                                      const mbedtls_mpi_uint *A,
953                                      const mbedtls_mpi_uint *N, size_t AN_limbs,
954                                      const mbedtls_mpi_uint *E, size_t E_limbs,
955                                      const mbedtls_mpi_uint *RR,
956                                      mbedtls_mpi_uint *T)
957 {
958     mbedtls_mpi_core_exp_mod_optionally_safe(X,
959                                              A,
960                                              N,
961                                              AN_limbs,
962                                              E,
963                                              E_limbs,
964                                              MBEDTLS_MPI_IS_PUBLIC,
965                                              RR,
966                                              T);
967 }
968 
mbedtls_mpi_core_sub_int(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,mbedtls_mpi_uint c,size_t limbs)969 mbedtls_mpi_uint mbedtls_mpi_core_sub_int(mbedtls_mpi_uint *X,
970                                           const mbedtls_mpi_uint *A,
971                                           mbedtls_mpi_uint c,  /* doubles as carry */
972                                           size_t limbs)
973 {
974     for (size_t i = 0; i < limbs; i++) {
975         mbedtls_mpi_uint s = A[i];
976         mbedtls_mpi_uint t = s - c;
977         c = (t > s);
978         X[i] = t;
979     }
980 
981     return c;
982 }
983 
mbedtls_mpi_core_check_zero_ct(const mbedtls_mpi_uint * A,size_t limbs)984 mbedtls_ct_condition_t mbedtls_mpi_core_check_zero_ct(const mbedtls_mpi_uint *A,
985                                                       size_t limbs)
986 {
987     volatile const mbedtls_mpi_uint *force_read_A = A;
988     mbedtls_mpi_uint bits = 0;
989 
990     for (size_t i = 0; i < limbs; i++) {
991         bits |= force_read_A[i];
992     }
993 
994     return mbedtls_ct_bool(bits);
995 }
996 
mbedtls_mpi_core_to_mont_rep(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * N,size_t AN_limbs,mbedtls_mpi_uint mm,const mbedtls_mpi_uint * rr,mbedtls_mpi_uint * T)997 void mbedtls_mpi_core_to_mont_rep(mbedtls_mpi_uint *X,
998                                   const mbedtls_mpi_uint *A,
999                                   const mbedtls_mpi_uint *N,
1000                                   size_t AN_limbs,
1001                                   mbedtls_mpi_uint mm,
1002                                   const mbedtls_mpi_uint *rr,
1003                                   mbedtls_mpi_uint *T)
1004 {
1005     mbedtls_mpi_core_montmul(X, A, rr, AN_limbs, N, AN_limbs, mm, T);
1006 }
1007 
mbedtls_mpi_core_from_mont_rep(mbedtls_mpi_uint * X,const mbedtls_mpi_uint * A,const mbedtls_mpi_uint * N,size_t AN_limbs,mbedtls_mpi_uint mm,mbedtls_mpi_uint * T)1008 void mbedtls_mpi_core_from_mont_rep(mbedtls_mpi_uint *X,
1009                                     const mbedtls_mpi_uint *A,
1010                                     const mbedtls_mpi_uint *N,
1011                                     size_t AN_limbs,
1012                                     mbedtls_mpi_uint mm,
1013                                     mbedtls_mpi_uint *T)
1014 {
1015     const mbedtls_mpi_uint Rinv = 1;    /* 1/R in Mont. rep => 1 */
1016 
1017     mbedtls_mpi_core_montmul(X, A, &Rinv, 1, N, AN_limbs, mm, T);
1018 }
1019 
1020 #endif /* MBEDTLS_BIGNUM_C */
1021