Commit | Line | Data |
---|---|---|
3c4b2390 | 1 | /* |
0d7a7864 VC |
2 | * Copyright (c) 2013, 2014 Kenneth MacKay. All rights reserved. |
3 | * Copyright (c) 2019 Vitaly Chikunov <vt@altlinux.org> | |
3c4b2390 SB |
4 | * |
5 | * Redistribution and use in source and binary forms, with or without | |
6 | * modification, are permitted provided that the following conditions are | |
7 | * met: | |
8 | * * Redistributions of source code must retain the above copyright | |
9 | * notice, this list of conditions and the following disclaimer. | |
10 | * * Redistributions in binary form must reproduce the above copyright | |
11 | * notice, this list of conditions and the following disclaimer in the | |
12 | * documentation and/or other materials provided with the distribution. | |
13 | * | |
14 | * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS | |
15 | * "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT | |
16 | * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR | |
17 | * A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT | |
18 | * HOLDER OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, | |
19 | * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT | |
20 | * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, | |
21 | * DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY | |
22 | * THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT | |
23 | * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE | |
24 | * OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. | |
25 | */ | |
26 | ||
14bb7676 | 27 | #include <crypto/ecc_curve.h> |
4a2289da | 28 | #include <linux/module.h> |
3c4b2390 SB |
29 | #include <linux/random.h> |
30 | #include <linux/slab.h> | |
31 | #include <linux/swab.h> | |
32 | #include <linux/fips.h> | |
33 | #include <crypto/ecdh.h> | |
6755fd26 | 34 | #include <crypto/rng.h> |
0d7a7864 VC |
35 | #include <asm/unaligned.h> |
36 | #include <linux/ratelimit.h> | |
3c4b2390 SB |
37 | |
38 | #include "ecc.h" | |
39 | #include "ecc_curve_defs.h" | |
40 | ||
41 | typedef struct { | |
42 | u64 m_low; | |
43 | u64 m_high; | |
44 | } uint128_t; | |
45 | ||
14bb7676 MY |
46 | |
47 | const struct ecc_curve *ecc_get_curve(unsigned int curve_id) | |
3c4b2390 SB |
48 | { |
49 | switch (curve_id) { | |
50 | /* In FIPS mode only allow P256 and higher */ | |
51 | case ECC_CURVE_NIST_P192: | |
52 | return fips_enabled ? NULL : &nist_p192; | |
53 | case ECC_CURVE_NIST_P256: | |
54 | return &nist_p256; | |
55 | default: | |
56 | return NULL; | |
57 | } | |
58 | } | |
14bb7676 | 59 | EXPORT_SYMBOL(ecc_get_curve); |
3c4b2390 SB |
60 | |
61 | static u64 *ecc_alloc_digits_space(unsigned int ndigits) | |
62 | { | |
63 | size_t len = ndigits * sizeof(u64); | |
64 | ||
65 | if (!len) | |
66 | return NULL; | |
67 | ||
68 | return kmalloc(len, GFP_KERNEL); | |
69 | } | |
70 | ||
71 | static void ecc_free_digits_space(u64 *space) | |
72 | { | |
453431a5 | 73 | kfree_sensitive(space); |
3c4b2390 SB |
74 | } |
75 | ||
76 | static struct ecc_point *ecc_alloc_point(unsigned int ndigits) | |
77 | { | |
78 | struct ecc_point *p = kmalloc(sizeof(*p), GFP_KERNEL); | |
79 | ||
80 | if (!p) | |
81 | return NULL; | |
82 | ||
83 | p->x = ecc_alloc_digits_space(ndigits); | |
84 | if (!p->x) | |
85 | goto err_alloc_x; | |
86 | ||
87 | p->y = ecc_alloc_digits_space(ndigits); | |
88 | if (!p->y) | |
89 | goto err_alloc_y; | |
90 | ||
91 | p->ndigits = ndigits; | |
92 | ||
93 | return p; | |
94 | ||
95 | err_alloc_y: | |
96 | ecc_free_digits_space(p->x); | |
97 | err_alloc_x: | |
98 | kfree(p); | |
99 | return NULL; | |
100 | } | |
101 | ||
102 | static void ecc_free_point(struct ecc_point *p) | |
103 | { | |
104 | if (!p) | |
105 | return; | |
106 | ||
453431a5 WL |
107 | kfree_sensitive(p->x); |
108 | kfree_sensitive(p->y); | |
109 | kfree_sensitive(p); | |
3c4b2390 SB |
110 | } |
111 | ||
112 | static void vli_clear(u64 *vli, unsigned int ndigits) | |
113 | { | |
114 | int i; | |
115 | ||
116 | for (i = 0; i < ndigits; i++) | |
117 | vli[i] = 0; | |
118 | } | |
119 | ||
120 | /* Returns true if vli == 0, false otherwise. */ | |
4a2289da | 121 | bool vli_is_zero(const u64 *vli, unsigned int ndigits) |
3c4b2390 SB |
122 | { |
123 | int i; | |
124 | ||
125 | for (i = 0; i < ndigits; i++) { | |
126 | if (vli[i]) | |
127 | return false; | |
128 | } | |
129 | ||
130 | return true; | |
131 | } | |
4a2289da | 132 | EXPORT_SYMBOL(vli_is_zero); |
3c4b2390 SB |
133 | |
134 | /* Returns nonzero if bit bit of vli is set. */ | |
135 | static u64 vli_test_bit(const u64 *vli, unsigned int bit) | |
136 | { | |
137 | return (vli[bit / 64] & ((u64)1 << (bit % 64))); | |
138 | } | |
139 | ||
0d7a7864 VC |
140 | static bool vli_is_negative(const u64 *vli, unsigned int ndigits) |
141 | { | |
142 | return vli_test_bit(vli, ndigits * 64 - 1); | |
143 | } | |
144 | ||
3c4b2390 SB |
145 | /* Counts the number of 64-bit "digits" in vli. */ |
146 | static unsigned int vli_num_digits(const u64 *vli, unsigned int ndigits) | |
147 | { | |
148 | int i; | |
149 | ||
150 | /* Search from the end until we find a non-zero digit. | |
151 | * We do it in reverse because we expect that most digits will | |
152 | * be nonzero. | |
153 | */ | |
154 | for (i = ndigits - 1; i >= 0 && vli[i] == 0; i--); | |
155 | ||
156 | return (i + 1); | |
157 | } | |
158 | ||
159 | /* Counts the number of bits required for vli. */ | |
160 | static unsigned int vli_num_bits(const u64 *vli, unsigned int ndigits) | |
161 | { | |
162 | unsigned int i, num_digits; | |
163 | u64 digit; | |
164 | ||
165 | num_digits = vli_num_digits(vli, ndigits); | |
166 | if (num_digits == 0) | |
167 | return 0; | |
168 | ||
169 | digit = vli[num_digits - 1]; | |
170 | for (i = 0; digit; i++) | |
171 | digit >>= 1; | |
172 | ||
173 | return ((num_digits - 1) * 64 + i); | |
174 | } | |
175 | ||
0d7a7864 VC |
176 | /* Set dest from unaligned bit string src. */ |
177 | void vli_from_be64(u64 *dest, const void *src, unsigned int ndigits) | |
178 | { | |
179 | int i; | |
180 | const u64 *from = src; | |
181 | ||
182 | for (i = 0; i < ndigits; i++) | |
183 | dest[i] = get_unaligned_be64(&from[ndigits - 1 - i]); | |
184 | } | |
185 | EXPORT_SYMBOL(vli_from_be64); | |
186 | ||
187 | void vli_from_le64(u64 *dest, const void *src, unsigned int ndigits) | |
188 | { | |
189 | int i; | |
190 | const u64 *from = src; | |
191 | ||
192 | for (i = 0; i < ndigits; i++) | |
193 | dest[i] = get_unaligned_le64(&from[i]); | |
194 | } | |
195 | EXPORT_SYMBOL(vli_from_le64); | |
196 | ||
3c4b2390 SB |
197 | /* Sets dest = src. */ |
198 | static void vli_set(u64 *dest, const u64 *src, unsigned int ndigits) | |
199 | { | |
200 | int i; | |
201 | ||
202 | for (i = 0; i < ndigits; i++) | |
203 | dest[i] = src[i]; | |
204 | } | |
205 | ||
206 | /* Returns sign of left - right. */ | |
4a2289da | 207 | int vli_cmp(const u64 *left, const u64 *right, unsigned int ndigits) |
3c4b2390 SB |
208 | { |
209 | int i; | |
210 | ||
211 | for (i = ndigits - 1; i >= 0; i--) { | |
212 | if (left[i] > right[i]) | |
213 | return 1; | |
214 | else if (left[i] < right[i]) | |
215 | return -1; | |
216 | } | |
217 | ||
218 | return 0; | |
219 | } | |
4a2289da | 220 | EXPORT_SYMBOL(vli_cmp); |
3c4b2390 SB |
221 | |
222 | /* Computes result = in << c, returning carry. Can modify in place | |
223 | * (if result == in). 0 < shift < 64. | |
224 | */ | |
225 | static u64 vli_lshift(u64 *result, const u64 *in, unsigned int shift, | |
226 | unsigned int ndigits) | |
227 | { | |
228 | u64 carry = 0; | |
229 | int i; | |
230 | ||
231 | for (i = 0; i < ndigits; i++) { | |
232 | u64 temp = in[i]; | |
233 | ||
234 | result[i] = (temp << shift) | carry; | |
235 | carry = temp >> (64 - shift); | |
236 | } | |
237 | ||
238 | return carry; | |
239 | } | |
240 | ||
241 | /* Computes vli = vli >> 1. */ | |
242 | static void vli_rshift1(u64 *vli, unsigned int ndigits) | |
243 | { | |
244 | u64 *end = vli; | |
245 | u64 carry = 0; | |
246 | ||
247 | vli += ndigits; | |
248 | ||
249 | while (vli-- > end) { | |
250 | u64 temp = *vli; | |
251 | *vli = (temp >> 1) | carry; | |
252 | carry = temp << 63; | |
253 | } | |
254 | } | |
255 | ||
256 | /* Computes result = left + right, returning carry. Can modify in place. */ | |
257 | static u64 vli_add(u64 *result, const u64 *left, const u64 *right, | |
258 | unsigned int ndigits) | |
259 | { | |
260 | u64 carry = 0; | |
261 | int i; | |
262 | ||
263 | for (i = 0; i < ndigits; i++) { | |
264 | u64 sum; | |
265 | ||
266 | sum = left[i] + right[i] + carry; | |
267 | if (sum != left[i]) | |
268 | carry = (sum < left[i]); | |
269 | ||
270 | result[i] = sum; | |
271 | } | |
272 | ||
273 | return carry; | |
274 | } | |
275 | ||
0d7a7864 VC |
276 | /* Computes result = left + right, returning carry. Can modify in place. */ |
277 | static u64 vli_uadd(u64 *result, const u64 *left, u64 right, | |
278 | unsigned int ndigits) | |
279 | { | |
280 | u64 carry = right; | |
281 | int i; | |
282 | ||
283 | for (i = 0; i < ndigits; i++) { | |
284 | u64 sum; | |
285 | ||
286 | sum = left[i] + carry; | |
287 | if (sum != left[i]) | |
288 | carry = (sum < left[i]); | |
289 | else | |
290 | carry = !!carry; | |
291 | ||
292 | result[i] = sum; | |
293 | } | |
294 | ||
295 | return carry; | |
296 | } | |
297 | ||
3c4b2390 | 298 | /* Computes result = left - right, returning borrow. Can modify in place. */ |
4a2289da | 299 | u64 vli_sub(u64 *result, const u64 *left, const u64 *right, |
3c4b2390 SB |
300 | unsigned int ndigits) |
301 | { | |
302 | u64 borrow = 0; | |
303 | int i; | |
304 | ||
305 | for (i = 0; i < ndigits; i++) { | |
306 | u64 diff; | |
307 | ||
308 | diff = left[i] - right[i] - borrow; | |
309 | if (diff != left[i]) | |
310 | borrow = (diff > left[i]); | |
311 | ||
312 | result[i] = diff; | |
313 | } | |
314 | ||
315 | return borrow; | |
316 | } | |
4a2289da | 317 | EXPORT_SYMBOL(vli_sub); |
3c4b2390 | 318 | |
0d7a7864 VC |
319 | /* Computes result = left - right, returning borrow. Can modify in place. */ |
320 | static u64 vli_usub(u64 *result, const u64 *left, u64 right, | |
321 | unsigned int ndigits) | |
322 | { | |
323 | u64 borrow = right; | |
324 | int i; | |
325 | ||
326 | for (i = 0; i < ndigits; i++) { | |
327 | u64 diff; | |
328 | ||
329 | diff = left[i] - borrow; | |
330 | if (diff != left[i]) | |
331 | borrow = (diff > left[i]); | |
332 | ||
333 | result[i] = diff; | |
334 | } | |
335 | ||
336 | return borrow; | |
337 | } | |
338 | ||
3c4b2390 SB |
339 | static uint128_t mul_64_64(u64 left, u64 right) |
340 | { | |
0d7a7864 | 341 | uint128_t result; |
c12d3362 | 342 | #if defined(CONFIG_ARCH_SUPPORTS_INT128) |
0d7a7864 VC |
343 | unsigned __int128 m = (unsigned __int128)left * right; |
344 | ||
345 | result.m_low = m; | |
346 | result.m_high = m >> 64; | |
347 | #else | |
3c4b2390 SB |
348 | u64 a0 = left & 0xffffffffull; |
349 | u64 a1 = left >> 32; | |
350 | u64 b0 = right & 0xffffffffull; | |
351 | u64 b1 = right >> 32; | |
352 | u64 m0 = a0 * b0; | |
353 | u64 m1 = a0 * b1; | |
354 | u64 m2 = a1 * b0; | |
355 | u64 m3 = a1 * b1; | |
3c4b2390 SB |
356 | |
357 | m2 += (m0 >> 32); | |
358 | m2 += m1; | |
359 | ||
360 | /* Overflow */ | |
361 | if (m2 < m1) | |
362 | m3 += 0x100000000ull; | |
363 | ||
364 | result.m_low = (m0 & 0xffffffffull) | (m2 << 32); | |
365 | result.m_high = m3 + (m2 >> 32); | |
0d7a7864 | 366 | #endif |
3c4b2390 SB |
367 | return result; |
368 | } | |
369 | ||
370 | static uint128_t add_128_128(uint128_t a, uint128_t b) | |
371 | { | |
372 | uint128_t result; | |
373 | ||
374 | result.m_low = a.m_low + b.m_low; | |
375 | result.m_high = a.m_high + b.m_high + (result.m_low < a.m_low); | |
376 | ||
377 | return result; | |
378 | } | |
379 | ||
380 | static void vli_mult(u64 *result, const u64 *left, const u64 *right, | |
381 | unsigned int ndigits) | |
382 | { | |
383 | uint128_t r01 = { 0, 0 }; | |
384 | u64 r2 = 0; | |
385 | unsigned int i, k; | |
386 | ||
387 | /* Compute each digit of result in sequence, maintaining the | |
388 | * carries. | |
389 | */ | |
390 | for (k = 0; k < ndigits * 2 - 1; k++) { | |
391 | unsigned int min; | |
392 | ||
393 | if (k < ndigits) | |
394 | min = 0; | |
395 | else | |
396 | min = (k + 1) - ndigits; | |
397 | ||
398 | for (i = min; i <= k && i < ndigits; i++) { | |
399 | uint128_t product; | |
400 | ||
401 | product = mul_64_64(left[i], right[k - i]); | |
402 | ||
403 | r01 = add_128_128(r01, product); | |
404 | r2 += (r01.m_high < product.m_high); | |
405 | } | |
406 | ||
407 | result[k] = r01.m_low; | |
408 | r01.m_low = r01.m_high; | |
409 | r01.m_high = r2; | |
410 | r2 = 0; | |
411 | } | |
412 | ||
413 | result[ndigits * 2 - 1] = r01.m_low; | |
414 | } | |
415 | ||
0d7a7864 VC |
416 | /* Compute product = left * right, for a small right value. */ |
417 | static void vli_umult(u64 *result, const u64 *left, u32 right, | |
418 | unsigned int ndigits) | |
419 | { | |
420 | uint128_t r01 = { 0 }; | |
421 | unsigned int k; | |
422 | ||
423 | for (k = 0; k < ndigits; k++) { | |
424 | uint128_t product; | |
425 | ||
426 | product = mul_64_64(left[k], right); | |
427 | r01 = add_128_128(r01, product); | |
428 | /* no carry */ | |
429 | result[k] = r01.m_low; | |
430 | r01.m_low = r01.m_high; | |
431 | r01.m_high = 0; | |
432 | } | |
433 | result[k] = r01.m_low; | |
434 | for (++k; k < ndigits * 2; k++) | |
435 | result[k] = 0; | |
436 | } | |
437 | ||
3c4b2390 SB |
438 | static void vli_square(u64 *result, const u64 *left, unsigned int ndigits) |
439 | { | |
440 | uint128_t r01 = { 0, 0 }; | |
441 | u64 r2 = 0; | |
442 | int i, k; | |
443 | ||
444 | for (k = 0; k < ndigits * 2 - 1; k++) { | |
445 | unsigned int min; | |
446 | ||
447 | if (k < ndigits) | |
448 | min = 0; | |
449 | else | |
450 | min = (k + 1) - ndigits; | |
451 | ||
452 | for (i = min; i <= k && i <= k - i; i++) { | |
453 | uint128_t product; | |
454 | ||
455 | product = mul_64_64(left[i], left[k - i]); | |
456 | ||
457 | if (i < k - i) { | |
458 | r2 += product.m_high >> 63; | |
459 | product.m_high = (product.m_high << 1) | | |
460 | (product.m_low >> 63); | |
461 | product.m_low <<= 1; | |
462 | } | |
463 | ||
464 | r01 = add_128_128(r01, product); | |
465 | r2 += (r01.m_high < product.m_high); | |
466 | } | |
467 | ||
468 | result[k] = r01.m_low; | |
469 | r01.m_low = r01.m_high; | |
470 | r01.m_high = r2; | |
471 | r2 = 0; | |
472 | } | |
473 | ||
474 | result[ndigits * 2 - 1] = r01.m_low; | |
475 | } | |
476 | ||
477 | /* Computes result = (left + right) % mod. | |
478 | * Assumes that left < mod and right < mod, result != mod. | |
479 | */ | |
480 | static void vli_mod_add(u64 *result, const u64 *left, const u64 *right, | |
481 | const u64 *mod, unsigned int ndigits) | |
482 | { | |
483 | u64 carry; | |
484 | ||
485 | carry = vli_add(result, left, right, ndigits); | |
486 | ||
487 | /* result > mod (result = mod + remainder), so subtract mod to | |
488 | * get remainder. | |
489 | */ | |
490 | if (carry || vli_cmp(result, mod, ndigits) >= 0) | |
491 | vli_sub(result, result, mod, ndigits); | |
492 | } | |
493 | ||
494 | /* Computes result = (left - right) % mod. | |
495 | * Assumes that left < mod and right < mod, result != mod. | |
496 | */ | |
497 | static void vli_mod_sub(u64 *result, const u64 *left, const u64 *right, | |
498 | const u64 *mod, unsigned int ndigits) | |
499 | { | |
500 | u64 borrow = vli_sub(result, left, right, ndigits); | |
501 | ||
502 | /* In this case, p_result == -diff == (max int) - diff. | |
503 | * Since -x % d == d - x, we can get the correct result from | |
504 | * result + mod (with overflow). | |
505 | */ | |
506 | if (borrow) | |
507 | vli_add(result, result, mod, ndigits); | |
508 | } | |
509 | ||
0d7a7864 VC |
510 | /* |
511 | * Computes result = product % mod | |
512 | * for special form moduli: p = 2^k-c, for small c (note the minus sign) | |
513 | * | |
514 | * References: | |
515 | * R. Crandall, C. Pomerance. Prime Numbers: A Computational Perspective. | |
516 | * 9 Fast Algorithms for Large-Integer Arithmetic. 9.2.3 Moduli of special form | |
517 | * Algorithm 9.2.13 (Fast mod operation for special-form moduli). | |
518 | */ | |
519 | static void vli_mmod_special(u64 *result, const u64 *product, | |
520 | const u64 *mod, unsigned int ndigits) | |
521 | { | |
522 | u64 c = -mod[0]; | |
523 | u64 t[ECC_MAX_DIGITS * 2]; | |
524 | u64 r[ECC_MAX_DIGITS * 2]; | |
525 | ||
526 | vli_set(r, product, ndigits * 2); | |
527 | while (!vli_is_zero(r + ndigits, ndigits)) { | |
528 | vli_umult(t, r + ndigits, c, ndigits); | |
529 | vli_clear(r + ndigits, ndigits); | |
530 | vli_add(r, r, t, ndigits * 2); | |
531 | } | |
532 | vli_set(t, mod, ndigits); | |
533 | vli_clear(t + ndigits, ndigits); | |
534 | while (vli_cmp(r, t, ndigits * 2) >= 0) | |
535 | vli_sub(r, r, t, ndigits * 2); | |
536 | vli_set(result, r, ndigits); | |
537 | } | |
538 | ||
539 | /* | |
540 | * Computes result = product % mod | |
541 | * for special form moduli: p = 2^{k-1}+c, for small c (note the plus sign) | |
542 | * where k-1 does not fit into qword boundary by -1 bit (such as 255). | |
543 | ||
544 | * References (loosely based on): | |
545 | * A. Menezes, P. van Oorschot, S. Vanstone. Handbook of Applied Cryptography. | |
546 | * 14.3.4 Reduction methods for moduli of special form. Algorithm 14.47. | |
547 | * URL: http://cacr.uwaterloo.ca/hac/about/chap14.pdf | |
548 | * | |
549 | * H. Cohen, G. Frey, R. Avanzi, C. Doche, T. Lange, K. Nguyen, F. Vercauteren. | |
550 | * Handbook of Elliptic and Hyperelliptic Curve Cryptography. | |
551 | * Algorithm 10.25 Fast reduction for special form moduli | |
552 | */ | |
553 | static void vli_mmod_special2(u64 *result, const u64 *product, | |
554 | const u64 *mod, unsigned int ndigits) | |
555 | { | |
556 | u64 c2 = mod[0] * 2; | |
557 | u64 q[ECC_MAX_DIGITS]; | |
558 | u64 r[ECC_MAX_DIGITS * 2]; | |
559 | u64 m[ECC_MAX_DIGITS * 2]; /* expanded mod */ | |
560 | int carry; /* last bit that doesn't fit into q */ | |
561 | int i; | |
562 | ||
563 | vli_set(m, mod, ndigits); | |
564 | vli_clear(m + ndigits, ndigits); | |
565 | ||
566 | vli_set(r, product, ndigits); | |
567 | /* q and carry are top bits */ | |
568 | vli_set(q, product + ndigits, ndigits); | |
569 | vli_clear(r + ndigits, ndigits); | |
570 | carry = vli_is_negative(r, ndigits); | |
571 | if (carry) | |
572 | r[ndigits - 1] &= (1ull << 63) - 1; | |
573 | for (i = 1; carry || !vli_is_zero(q, ndigits); i++) { | |
574 | u64 qc[ECC_MAX_DIGITS * 2]; | |
575 | ||
576 | vli_umult(qc, q, c2, ndigits); | |
577 | if (carry) | |
578 | vli_uadd(qc, qc, mod[0], ndigits * 2); | |
579 | vli_set(q, qc + ndigits, ndigits); | |
580 | vli_clear(qc + ndigits, ndigits); | |
581 | carry = vli_is_negative(qc, ndigits); | |
582 | if (carry) | |
583 | qc[ndigits - 1] &= (1ull << 63) - 1; | |
584 | if (i & 1) | |
585 | vli_sub(r, r, qc, ndigits * 2); | |
586 | else | |
587 | vli_add(r, r, qc, ndigits * 2); | |
588 | } | |
589 | while (vli_is_negative(r, ndigits * 2)) | |
590 | vli_add(r, r, m, ndigits * 2); | |
591 | while (vli_cmp(r, m, ndigits * 2) >= 0) | |
592 | vli_sub(r, r, m, ndigits * 2); | |
593 | ||
594 | vli_set(result, r, ndigits); | |
595 | } | |
596 | ||
597 | /* | |
598 | * Computes result = product % mod, where product is 2N words long. | |
599 | * Reference: Ken MacKay's micro-ecc. | |
600 | * Currently only designed to work for curve_p or curve_n. | |
601 | */ | |
602 | static void vli_mmod_slow(u64 *result, u64 *product, const u64 *mod, | |
603 | unsigned int ndigits) | |
604 | { | |
605 | u64 mod_m[2 * ECC_MAX_DIGITS]; | |
606 | u64 tmp[2 * ECC_MAX_DIGITS]; | |
607 | u64 *v[2] = { tmp, product }; | |
608 | u64 carry = 0; | |
609 | unsigned int i; | |
610 | /* Shift mod so its highest set bit is at the maximum position. */ | |
611 | int shift = (ndigits * 2 * 64) - vli_num_bits(mod, ndigits); | |
612 | int word_shift = shift / 64; | |
613 | int bit_shift = shift % 64; | |
614 | ||
615 | vli_clear(mod_m, word_shift); | |
616 | if (bit_shift > 0) { | |
617 | for (i = 0; i < ndigits; ++i) { | |
618 | mod_m[word_shift + i] = (mod[i] << bit_shift) | carry; | |
619 | carry = mod[i] >> (64 - bit_shift); | |
620 | } | |
621 | } else | |
622 | vli_set(mod_m + word_shift, mod, ndigits); | |
623 | ||
624 | for (i = 1; shift >= 0; --shift) { | |
625 | u64 borrow = 0; | |
626 | unsigned int j; | |
627 | ||
628 | for (j = 0; j < ndigits * 2; ++j) { | |
629 | u64 diff = v[i][j] - mod_m[j] - borrow; | |
630 | ||
631 | if (diff != v[i][j]) | |
632 | borrow = (diff > v[i][j]); | |
633 | v[1 - i][j] = diff; | |
634 | } | |
635 | i = !(i ^ borrow); /* Swap the index if there was no borrow */ | |
636 | vli_rshift1(mod_m, ndigits); | |
637 | mod_m[ndigits - 1] |= mod_m[ndigits] << (64 - 1); | |
638 | vli_rshift1(mod_m + ndigits, ndigits); | |
639 | } | |
640 | vli_set(result, v[i], ndigits); | |
641 | } | |
642 | ||
643 | /* Computes result = product % mod using Barrett's reduction with precomputed | |
644 | * value mu appended to the mod after ndigits, mu = (2^{2w} / mod) and have | |
645 | * length ndigits + 1, where mu * (2^w - 1) should not overflow ndigits | |
646 | * boundary. | |
647 | * | |
648 | * Reference: | |
649 | * R. Brent, P. Zimmermann. Modern Computer Arithmetic. 2010. | |
650 | * 2.4.1 Barrett's algorithm. Algorithm 2.5. | |
651 | */ | |
652 | static void vli_mmod_barrett(u64 *result, u64 *product, const u64 *mod, | |
653 | unsigned int ndigits) | |
654 | { | |
655 | u64 q[ECC_MAX_DIGITS * 2]; | |
656 | u64 r[ECC_MAX_DIGITS * 2]; | |
657 | const u64 *mu = mod + ndigits; | |
658 | ||
659 | vli_mult(q, product + ndigits, mu, ndigits); | |
660 | if (mu[ndigits]) | |
661 | vli_add(q + ndigits, q + ndigits, product + ndigits, ndigits); | |
662 | vli_mult(r, mod, q + ndigits, ndigits); | |
663 | vli_sub(r, product, r, ndigits * 2); | |
664 | while (!vli_is_zero(r + ndigits, ndigits) || | |
665 | vli_cmp(r, mod, ndigits) != -1) { | |
666 | u64 carry; | |
667 | ||
668 | carry = vli_sub(r, r, mod, ndigits); | |
669 | vli_usub(r + ndigits, r + ndigits, carry, ndigits); | |
670 | } | |
671 | vli_set(result, r, ndigits); | |
672 | } | |
673 | ||
3c4b2390 SB |
674 | /* Computes p_result = p_product % curve_p. |
675 | * See algorithm 5 and 6 from | |
676 | * http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf | |
677 | */ | |
678 | static void vli_mmod_fast_192(u64 *result, const u64 *product, | |
679 | const u64 *curve_prime, u64 *tmp) | |
680 | { | |
681 | const unsigned int ndigits = 3; | |
682 | int carry; | |
683 | ||
684 | vli_set(result, product, ndigits); | |
685 | ||
686 | vli_set(tmp, &product[3], ndigits); | |
687 | carry = vli_add(result, result, tmp, ndigits); | |
688 | ||
689 | tmp[0] = 0; | |
690 | tmp[1] = product[3]; | |
691 | tmp[2] = product[4]; | |
692 | carry += vli_add(result, result, tmp, ndigits); | |
693 | ||
694 | tmp[0] = tmp[1] = product[5]; | |
695 | tmp[2] = 0; | |
696 | carry += vli_add(result, result, tmp, ndigits); | |
697 | ||
698 | while (carry || vli_cmp(curve_prime, result, ndigits) != 1) | |
699 | carry -= vli_sub(result, result, curve_prime, ndigits); | |
700 | } | |
701 | ||
702 | /* Computes result = product % curve_prime | |
703 | * from http://www.nsa.gov/ia/_files/nist-routines.pdf | |
704 | */ | |
705 | static void vli_mmod_fast_256(u64 *result, const u64 *product, | |
706 | const u64 *curve_prime, u64 *tmp) | |
707 | { | |
708 | int carry; | |
709 | const unsigned int ndigits = 4; | |
710 | ||
711 | /* t */ | |
712 | vli_set(result, product, ndigits); | |
713 | ||
714 | /* s1 */ | |
715 | tmp[0] = 0; | |
716 | tmp[1] = product[5] & 0xffffffff00000000ull; | |
717 | tmp[2] = product[6]; | |
718 | tmp[3] = product[7]; | |
719 | carry = vli_lshift(tmp, tmp, 1, ndigits); | |
720 | carry += vli_add(result, result, tmp, ndigits); | |
721 | ||
722 | /* s2 */ | |
723 | tmp[1] = product[6] << 32; | |
724 | tmp[2] = (product[6] >> 32) | (product[7] << 32); | |
725 | tmp[3] = product[7] >> 32; | |
726 | carry += vli_lshift(tmp, tmp, 1, ndigits); | |
727 | carry += vli_add(result, result, tmp, ndigits); | |
728 | ||
729 | /* s3 */ | |
730 | tmp[0] = product[4]; | |
731 | tmp[1] = product[5] & 0xffffffff; | |
732 | tmp[2] = 0; | |
733 | tmp[3] = product[7]; | |
734 | carry += vli_add(result, result, tmp, ndigits); | |
735 | ||
736 | /* s4 */ | |
737 | tmp[0] = (product[4] >> 32) | (product[5] << 32); | |
738 | tmp[1] = (product[5] >> 32) | (product[6] & 0xffffffff00000000ull); | |
739 | tmp[2] = product[7]; | |
740 | tmp[3] = (product[6] >> 32) | (product[4] << 32); | |
741 | carry += vli_add(result, result, tmp, ndigits); | |
742 | ||
743 | /* d1 */ | |
744 | tmp[0] = (product[5] >> 32) | (product[6] << 32); | |
745 | tmp[1] = (product[6] >> 32); | |
746 | tmp[2] = 0; | |
747 | tmp[3] = (product[4] & 0xffffffff) | (product[5] << 32); | |
748 | carry -= vli_sub(result, result, tmp, ndigits); | |
749 | ||
750 | /* d2 */ | |
751 | tmp[0] = product[6]; | |
752 | tmp[1] = product[7]; | |
753 | tmp[2] = 0; | |
754 | tmp[3] = (product[4] >> 32) | (product[5] & 0xffffffff00000000ull); | |
755 | carry -= vli_sub(result, result, tmp, ndigits); | |
756 | ||
757 | /* d3 */ | |
758 | tmp[0] = (product[6] >> 32) | (product[7] << 32); | |
759 | tmp[1] = (product[7] >> 32) | (product[4] << 32); | |
760 | tmp[2] = (product[4] >> 32) | (product[5] << 32); | |
761 | tmp[3] = (product[6] << 32); | |
762 | carry -= vli_sub(result, result, tmp, ndigits); | |
763 | ||
764 | /* d4 */ | |
765 | tmp[0] = product[7]; | |
766 | tmp[1] = product[4] & 0xffffffff00000000ull; | |
767 | tmp[2] = product[5]; | |
768 | tmp[3] = product[6] & 0xffffffff00000000ull; | |
769 | carry -= vli_sub(result, result, tmp, ndigits); | |
770 | ||
771 | if (carry < 0) { | |
772 | do { | |
773 | carry += vli_add(result, result, curve_prime, ndigits); | |
774 | } while (carry < 0); | |
775 | } else { | |
776 | while (carry || vli_cmp(curve_prime, result, ndigits) != 1) | |
777 | carry -= vli_sub(result, result, curve_prime, ndigits); | |
778 | } | |
779 | } | |
780 | ||
0d7a7864 VC |
781 | /* Computes result = product % curve_prime for different curve_primes. |
782 | * | |
783 | * Note that curve_primes are distinguished just by heuristic check and | |
784 | * not by complete conformance check. | |
785 | */ | |
3c4b2390 SB |
786 | static bool vli_mmod_fast(u64 *result, u64 *product, |
787 | const u64 *curve_prime, unsigned int ndigits) | |
788 | { | |
d5c3b178 | 789 | u64 tmp[2 * ECC_MAX_DIGITS]; |
3c4b2390 | 790 | |
0d7a7864 VC |
791 | /* Currently, both NIST primes have -1 in lowest qword. */ |
792 | if (curve_prime[0] != -1ull) { | |
793 | /* Try to handle Pseudo-Marsenne primes. */ | |
794 | if (curve_prime[ndigits - 1] == -1ull) { | |
795 | vli_mmod_special(result, product, curve_prime, | |
796 | ndigits); | |
797 | return true; | |
798 | } else if (curve_prime[ndigits - 1] == 1ull << 63 && | |
799 | curve_prime[ndigits - 2] == 0) { | |
800 | vli_mmod_special2(result, product, curve_prime, | |
801 | ndigits); | |
802 | return true; | |
803 | } | |
804 | vli_mmod_barrett(result, product, curve_prime, ndigits); | |
805 | return true; | |
806 | } | |
807 | ||
3c4b2390 SB |
808 | switch (ndigits) { |
809 | case 3: | |
810 | vli_mmod_fast_192(result, product, curve_prime, tmp); | |
811 | break; | |
812 | case 4: | |
813 | vli_mmod_fast_256(result, product, curve_prime, tmp); | |
814 | break; | |
815 | default: | |
0d7a7864 | 816 | pr_err_ratelimited("ecc: unsupported digits size!\n"); |
3c4b2390 SB |
817 | return false; |
818 | } | |
819 | ||
820 | return true; | |
821 | } | |
822 | ||
0d7a7864 VC |
823 | /* Computes result = (left * right) % mod. |
824 | * Assumes that mod is big enough curve order. | |
825 | */ | |
826 | void vli_mod_mult_slow(u64 *result, const u64 *left, const u64 *right, | |
827 | const u64 *mod, unsigned int ndigits) | |
828 | { | |
829 | u64 product[ECC_MAX_DIGITS * 2]; | |
830 | ||
831 | vli_mult(product, left, right, ndigits); | |
832 | vli_mmod_slow(result, product, mod, ndigits); | |
833 | } | |
834 | EXPORT_SYMBOL(vli_mod_mult_slow); | |
835 | ||
3c4b2390 SB |
836 | /* Computes result = (left * right) % curve_prime. */ |
837 | static void vli_mod_mult_fast(u64 *result, const u64 *left, const u64 *right, | |
838 | const u64 *curve_prime, unsigned int ndigits) | |
839 | { | |
d5c3b178 | 840 | u64 product[2 * ECC_MAX_DIGITS]; |
3c4b2390 SB |
841 | |
842 | vli_mult(product, left, right, ndigits); | |
843 | vli_mmod_fast(result, product, curve_prime, ndigits); | |
844 | } | |
845 | ||
846 | /* Computes result = left^2 % curve_prime. */ | |
847 | static void vli_mod_square_fast(u64 *result, const u64 *left, | |
848 | const u64 *curve_prime, unsigned int ndigits) | |
849 | { | |
d5c3b178 | 850 | u64 product[2 * ECC_MAX_DIGITS]; |
3c4b2390 SB |
851 | |
852 | vli_square(product, left, ndigits); | |
853 | vli_mmod_fast(result, product, curve_prime, ndigits); | |
854 | } | |
855 | ||
856 | #define EVEN(vli) (!(vli[0] & 1)) | |
857 | /* Computes result = (1 / p_input) % mod. All VLIs are the same size. | |
858 | * See "From Euclid's GCD to Montgomery Multiplication to the Great Divide" | |
859 | * https://labs.oracle.com/techrep/2001/smli_tr-2001-95.pdf | |
860 | */ | |
4a2289da | 861 | void vli_mod_inv(u64 *result, const u64 *input, const u64 *mod, |
3c4b2390 SB |
862 | unsigned int ndigits) |
863 | { | |
d5c3b178 KC |
864 | u64 a[ECC_MAX_DIGITS], b[ECC_MAX_DIGITS]; |
865 | u64 u[ECC_MAX_DIGITS], v[ECC_MAX_DIGITS]; | |
3c4b2390 SB |
866 | u64 carry; |
867 | int cmp_result; | |
868 | ||
869 | if (vli_is_zero(input, ndigits)) { | |
870 | vli_clear(result, ndigits); | |
871 | return; | |
872 | } | |
873 | ||
874 | vli_set(a, input, ndigits); | |
875 | vli_set(b, mod, ndigits); | |
876 | vli_clear(u, ndigits); | |
877 | u[0] = 1; | |
878 | vli_clear(v, ndigits); | |
879 | ||
880 | while ((cmp_result = vli_cmp(a, b, ndigits)) != 0) { | |
881 | carry = 0; | |
882 | ||
883 | if (EVEN(a)) { | |
884 | vli_rshift1(a, ndigits); | |
885 | ||
886 | if (!EVEN(u)) | |
887 | carry = vli_add(u, u, mod, ndigits); | |
888 | ||
889 | vli_rshift1(u, ndigits); | |
890 | if (carry) | |
891 | u[ndigits - 1] |= 0x8000000000000000ull; | |
892 | } else if (EVEN(b)) { | |
893 | vli_rshift1(b, ndigits); | |
894 | ||
895 | if (!EVEN(v)) | |
896 | carry = vli_add(v, v, mod, ndigits); | |
897 | ||
898 | vli_rshift1(v, ndigits); | |
899 | if (carry) | |
900 | v[ndigits - 1] |= 0x8000000000000000ull; | |
901 | } else if (cmp_result > 0) { | |
902 | vli_sub(a, a, b, ndigits); | |
903 | vli_rshift1(a, ndigits); | |
904 | ||
905 | if (vli_cmp(u, v, ndigits) < 0) | |
906 | vli_add(u, u, mod, ndigits); | |
907 | ||
908 | vli_sub(u, u, v, ndigits); | |
909 | if (!EVEN(u)) | |
910 | carry = vli_add(u, u, mod, ndigits); | |
911 | ||
912 | vli_rshift1(u, ndigits); | |
913 | if (carry) | |
914 | u[ndigits - 1] |= 0x8000000000000000ull; | |
915 | } else { | |
916 | vli_sub(b, b, a, ndigits); | |
917 | vli_rshift1(b, ndigits); | |
918 | ||
919 | if (vli_cmp(v, u, ndigits) < 0) | |
920 | vli_add(v, v, mod, ndigits); | |
921 | ||
922 | vli_sub(v, v, u, ndigits); | |
923 | if (!EVEN(v)) | |
924 | carry = vli_add(v, v, mod, ndigits); | |
925 | ||
926 | vli_rshift1(v, ndigits); | |
927 | if (carry) | |
928 | v[ndigits - 1] |= 0x8000000000000000ull; | |
929 | } | |
930 | } | |
931 | ||
932 | vli_set(result, u, ndigits); | |
933 | } | |
4a2289da | 934 | EXPORT_SYMBOL(vli_mod_inv); |
3c4b2390 SB |
935 | |
936 | /* ------ Point operations ------ */ | |
937 | ||
938 | /* Returns true if p_point is the point at infinity, false otherwise. */ | |
939 | static bool ecc_point_is_zero(const struct ecc_point *point) | |
940 | { | |
941 | return (vli_is_zero(point->x, point->ndigits) && | |
942 | vli_is_zero(point->y, point->ndigits)); | |
943 | } | |
944 | ||
945 | /* Point multiplication algorithm using Montgomery's ladder with co-Z | |
9332a9e7 | 946 | * coordinates. From https://eprint.iacr.org/2011/338.pdf |
3c4b2390 SB |
947 | */ |
948 | ||
949 | /* Double in place */ | |
950 | static void ecc_point_double_jacobian(u64 *x1, u64 *y1, u64 *z1, | |
951 | u64 *curve_prime, unsigned int ndigits) | |
952 | { | |
953 | /* t1 = x, t2 = y, t3 = z */ | |
d5c3b178 KC |
954 | u64 t4[ECC_MAX_DIGITS]; |
955 | u64 t5[ECC_MAX_DIGITS]; | |
3c4b2390 SB |
956 | |
957 | if (vli_is_zero(z1, ndigits)) | |
958 | return; | |
959 | ||
960 | /* t4 = y1^2 */ | |
961 | vli_mod_square_fast(t4, y1, curve_prime, ndigits); | |
962 | /* t5 = x1*y1^2 = A */ | |
963 | vli_mod_mult_fast(t5, x1, t4, curve_prime, ndigits); | |
964 | /* t4 = y1^4 */ | |
965 | vli_mod_square_fast(t4, t4, curve_prime, ndigits); | |
966 | /* t2 = y1*z1 = z3 */ | |
967 | vli_mod_mult_fast(y1, y1, z1, curve_prime, ndigits); | |
968 | /* t3 = z1^2 */ | |
969 | vli_mod_square_fast(z1, z1, curve_prime, ndigits); | |
970 | ||
971 | /* t1 = x1 + z1^2 */ | |
972 | vli_mod_add(x1, x1, z1, curve_prime, ndigits); | |
973 | /* t3 = 2*z1^2 */ | |
974 | vli_mod_add(z1, z1, z1, curve_prime, ndigits); | |
975 | /* t3 = x1 - z1^2 */ | |
976 | vli_mod_sub(z1, x1, z1, curve_prime, ndigits); | |
977 | /* t1 = x1^2 - z1^4 */ | |
978 | vli_mod_mult_fast(x1, x1, z1, curve_prime, ndigits); | |
979 | ||
980 | /* t3 = 2*(x1^2 - z1^4) */ | |
981 | vli_mod_add(z1, x1, x1, curve_prime, ndigits); | |
982 | /* t1 = 3*(x1^2 - z1^4) */ | |
983 | vli_mod_add(x1, x1, z1, curve_prime, ndigits); | |
984 | if (vli_test_bit(x1, 0)) { | |
985 | u64 carry = vli_add(x1, x1, curve_prime, ndigits); | |
986 | ||
987 | vli_rshift1(x1, ndigits); | |
988 | x1[ndigits - 1] |= carry << 63; | |
989 | } else { | |
990 | vli_rshift1(x1, ndigits); | |
991 | } | |
992 | /* t1 = 3/2*(x1^2 - z1^4) = B */ | |
993 | ||
994 | /* t3 = B^2 */ | |
995 | vli_mod_square_fast(z1, x1, curve_prime, ndigits); | |
996 | /* t3 = B^2 - A */ | |
997 | vli_mod_sub(z1, z1, t5, curve_prime, ndigits); | |
998 | /* t3 = B^2 - 2A = x3 */ | |
999 | vli_mod_sub(z1, z1, t5, curve_prime, ndigits); | |
1000 | /* t5 = A - x3 */ | |
1001 | vli_mod_sub(t5, t5, z1, curve_prime, ndigits); | |
1002 | /* t1 = B * (A - x3) */ | |
1003 | vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits); | |
1004 | /* t4 = B * (A - x3) - y1^4 = y3 */ | |
1005 | vli_mod_sub(t4, x1, t4, curve_prime, ndigits); | |
1006 | ||
1007 | vli_set(x1, z1, ndigits); | |
1008 | vli_set(z1, y1, ndigits); | |
1009 | vli_set(y1, t4, ndigits); | |
1010 | } | |
1011 | ||
1012 | /* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */ | |
1013 | static void apply_z(u64 *x1, u64 *y1, u64 *z, u64 *curve_prime, | |
1014 | unsigned int ndigits) | |
1015 | { | |
d5c3b178 | 1016 | u64 t1[ECC_MAX_DIGITS]; |
3c4b2390 SB |
1017 | |
1018 | vli_mod_square_fast(t1, z, curve_prime, ndigits); /* z^2 */ | |
1019 | vli_mod_mult_fast(x1, x1, t1, curve_prime, ndigits); /* x1 * z^2 */ | |
1020 | vli_mod_mult_fast(t1, t1, z, curve_prime, ndigits); /* z^3 */ | |
1021 | vli_mod_mult_fast(y1, y1, t1, curve_prime, ndigits); /* y1 * z^3 */ | |
1022 | } | |
1023 | ||
1024 | /* P = (x1, y1) => 2P, (x2, y2) => P' */ | |
1025 | static void xycz_initial_double(u64 *x1, u64 *y1, u64 *x2, u64 *y2, | |
1026 | u64 *p_initial_z, u64 *curve_prime, | |
1027 | unsigned int ndigits) | |
1028 | { | |
d5c3b178 | 1029 | u64 z[ECC_MAX_DIGITS]; |
3c4b2390 SB |
1030 | |
1031 | vli_set(x2, x1, ndigits); | |
1032 | vli_set(y2, y1, ndigits); | |
1033 | ||
1034 | vli_clear(z, ndigits); | |
1035 | z[0] = 1; | |
1036 | ||
1037 | if (p_initial_z) | |
1038 | vli_set(z, p_initial_z, ndigits); | |
1039 | ||
1040 | apply_z(x1, y1, z, curve_prime, ndigits); | |
1041 | ||
1042 | ecc_point_double_jacobian(x1, y1, z, curve_prime, ndigits); | |
1043 | ||
1044 | apply_z(x2, y2, z, curve_prime, ndigits); | |
1045 | } | |
1046 | ||
1047 | /* Input P = (x1, y1, Z), Q = (x2, y2, Z) | |
1048 | * Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3) | |
1049 | * or P => P', Q => P + Q | |
1050 | */ | |
1051 | static void xycz_add(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime, | |
1052 | unsigned int ndigits) | |
1053 | { | |
1054 | /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */ | |
d5c3b178 | 1055 | u64 t5[ECC_MAX_DIGITS]; |
3c4b2390 SB |
1056 | |
1057 | /* t5 = x2 - x1 */ | |
1058 | vli_mod_sub(t5, x2, x1, curve_prime, ndigits); | |
1059 | /* t5 = (x2 - x1)^2 = A */ | |
1060 | vli_mod_square_fast(t5, t5, curve_prime, ndigits); | |
1061 | /* t1 = x1*A = B */ | |
1062 | vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits); | |
1063 | /* t3 = x2*A = C */ | |
1064 | vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits); | |
1065 | /* t4 = y2 - y1 */ | |
1066 | vli_mod_sub(y2, y2, y1, curve_prime, ndigits); | |
1067 | /* t5 = (y2 - y1)^2 = D */ | |
1068 | vli_mod_square_fast(t5, y2, curve_prime, ndigits); | |
1069 | ||
1070 | /* t5 = D - B */ | |
1071 | vli_mod_sub(t5, t5, x1, curve_prime, ndigits); | |
1072 | /* t5 = D - B - C = x3 */ | |
1073 | vli_mod_sub(t5, t5, x2, curve_prime, ndigits); | |
1074 | /* t3 = C - B */ | |
1075 | vli_mod_sub(x2, x2, x1, curve_prime, ndigits); | |
1076 | /* t2 = y1*(C - B) */ | |
1077 | vli_mod_mult_fast(y1, y1, x2, curve_prime, ndigits); | |
1078 | /* t3 = B - x3 */ | |
1079 | vli_mod_sub(x2, x1, t5, curve_prime, ndigits); | |
1080 | /* t4 = (y2 - y1)*(B - x3) */ | |
1081 | vli_mod_mult_fast(y2, y2, x2, curve_prime, ndigits); | |
1082 | /* t4 = y3 */ | |
1083 | vli_mod_sub(y2, y2, y1, curve_prime, ndigits); | |
1084 | ||
1085 | vli_set(x2, t5, ndigits); | |
1086 | } | |
1087 | ||
1088 | /* Input P = (x1, y1, Z), Q = (x2, y2, Z) | |
1089 | * Output P + Q = (x3, y3, Z3), P - Q = (x3', y3', Z3) | |
1090 | * or P => P - Q, Q => P + Q | |
1091 | */ | |
1092 | static void xycz_add_c(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime, | |
1093 | unsigned int ndigits) | |
1094 | { | |
1095 | /* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */ | |
d5c3b178 KC |
1096 | u64 t5[ECC_MAX_DIGITS]; |
1097 | u64 t6[ECC_MAX_DIGITS]; | |
1098 | u64 t7[ECC_MAX_DIGITS]; | |
3c4b2390 SB |
1099 | |
1100 | /* t5 = x2 - x1 */ | |
1101 | vli_mod_sub(t5, x2, x1, curve_prime, ndigits); | |
1102 | /* t5 = (x2 - x1)^2 = A */ | |
1103 | vli_mod_square_fast(t5, t5, curve_prime, ndigits); | |
1104 | /* t1 = x1*A = B */ | |
1105 | vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits); | |
1106 | /* t3 = x2*A = C */ | |
1107 | vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits); | |
1108 | /* t4 = y2 + y1 */ | |
1109 | vli_mod_add(t5, y2, y1, curve_prime, ndigits); | |
1110 | /* t4 = y2 - y1 */ | |
1111 | vli_mod_sub(y2, y2, y1, curve_prime, ndigits); | |
1112 | ||
1113 | /* t6 = C - B */ | |
1114 | vli_mod_sub(t6, x2, x1, curve_prime, ndigits); | |
1115 | /* t2 = y1 * (C - B) */ | |
1116 | vli_mod_mult_fast(y1, y1, t6, curve_prime, ndigits); | |
1117 | /* t6 = B + C */ | |
1118 | vli_mod_add(t6, x1, x2, curve_prime, ndigits); | |
1119 | /* t3 = (y2 - y1)^2 */ | |
1120 | vli_mod_square_fast(x2, y2, curve_prime, ndigits); | |
1121 | /* t3 = x3 */ | |
1122 | vli_mod_sub(x2, x2, t6, curve_prime, ndigits); | |
1123 | ||
1124 | /* t7 = B - x3 */ | |
1125 | vli_mod_sub(t7, x1, x2, curve_prime, ndigits); | |
1126 | /* t4 = (y2 - y1)*(B - x3) */ | |
1127 | vli_mod_mult_fast(y2, y2, t7, curve_prime, ndigits); | |
1128 | /* t4 = y3 */ | |
1129 | vli_mod_sub(y2, y2, y1, curve_prime, ndigits); | |
1130 | ||
1131 | /* t7 = (y2 + y1)^2 = F */ | |
1132 | vli_mod_square_fast(t7, t5, curve_prime, ndigits); | |
1133 | /* t7 = x3' */ | |
1134 | vli_mod_sub(t7, t7, t6, curve_prime, ndigits); | |
1135 | /* t6 = x3' - B */ | |
1136 | vli_mod_sub(t6, t7, x1, curve_prime, ndigits); | |
1137 | /* t6 = (y2 + y1)*(x3' - B) */ | |
1138 | vli_mod_mult_fast(t6, t6, t5, curve_prime, ndigits); | |
1139 | /* t2 = y3' */ | |
1140 | vli_mod_sub(y1, t6, y1, curve_prime, ndigits); | |
1141 | ||
1142 | vli_set(x1, t7, ndigits); | |
1143 | } | |
1144 | ||
1145 | static void ecc_point_mult(struct ecc_point *result, | |
1146 | const struct ecc_point *point, const u64 *scalar, | |
3da2c1df | 1147 | u64 *initial_z, const struct ecc_curve *curve, |
3c4b2390 SB |
1148 | unsigned int ndigits) |
1149 | { | |
1150 | /* R0 and R1 */ | |
d5c3b178 KC |
1151 | u64 rx[2][ECC_MAX_DIGITS]; |
1152 | u64 ry[2][ECC_MAX_DIGITS]; | |
1153 | u64 z[ECC_MAX_DIGITS]; | |
3da2c1df VC |
1154 | u64 sk[2][ECC_MAX_DIGITS]; |
1155 | u64 *curve_prime = curve->p; | |
3c4b2390 | 1156 | int i, nb; |
3da2c1df VC |
1157 | int num_bits; |
1158 | int carry; | |
1159 | ||
1160 | carry = vli_add(sk[0], scalar, curve->n, ndigits); | |
1161 | vli_add(sk[1], sk[0], curve->n, ndigits); | |
1162 | scalar = sk[!carry]; | |
1163 | num_bits = sizeof(u64) * ndigits * 8 + 1; | |
3c4b2390 SB |
1164 | |
1165 | vli_set(rx[1], point->x, ndigits); | |
1166 | vli_set(ry[1], point->y, ndigits); | |
1167 | ||
1168 | xycz_initial_double(rx[1], ry[1], rx[0], ry[0], initial_z, curve_prime, | |
1169 | ndigits); | |
1170 | ||
1171 | for (i = num_bits - 2; i > 0; i--) { | |
1172 | nb = !vli_test_bit(scalar, i); | |
1173 | xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime, | |
1174 | ndigits); | |
1175 | xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime, | |
1176 | ndigits); | |
1177 | } | |
1178 | ||
1179 | nb = !vli_test_bit(scalar, 0); | |
1180 | xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime, | |
1181 | ndigits); | |
1182 | ||
1183 | /* Find final 1/Z value. */ | |
1184 | /* X1 - X0 */ | |
1185 | vli_mod_sub(z, rx[1], rx[0], curve_prime, ndigits); | |
1186 | /* Yb * (X1 - X0) */ | |
1187 | vli_mod_mult_fast(z, z, ry[1 - nb], curve_prime, ndigits); | |
1188 | /* xP * Yb * (X1 - X0) */ | |
1189 | vli_mod_mult_fast(z, z, point->x, curve_prime, ndigits); | |
1190 | ||
1191 | /* 1 / (xP * Yb * (X1 - X0)) */ | |
1192 | vli_mod_inv(z, z, curve_prime, point->ndigits); | |
1193 | ||
1194 | /* yP / (xP * Yb * (X1 - X0)) */ | |
1195 | vli_mod_mult_fast(z, z, point->y, curve_prime, ndigits); | |
1196 | /* Xb * yP / (xP * Yb * (X1 - X0)) */ | |
1197 | vli_mod_mult_fast(z, z, rx[1 - nb], curve_prime, ndigits); | |
1198 | /* End 1/Z calculation */ | |
1199 | ||
1200 | xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime, ndigits); | |
1201 | ||
1202 | apply_z(rx[0], ry[0], z, curve_prime, ndigits); | |
1203 | ||
1204 | vli_set(result->x, rx[0], ndigits); | |
1205 | vli_set(result->y, ry[0], ndigits); | |
1206 | } | |
1207 | ||
0d7a7864 VC |
1208 | /* Computes R = P + Q mod p */ |
1209 | static void ecc_point_add(const struct ecc_point *result, | |
1210 | const struct ecc_point *p, const struct ecc_point *q, | |
1211 | const struct ecc_curve *curve) | |
1212 | { | |
1213 | u64 z[ECC_MAX_DIGITS]; | |
1214 | u64 px[ECC_MAX_DIGITS]; | |
1215 | u64 py[ECC_MAX_DIGITS]; | |
1216 | unsigned int ndigits = curve->g.ndigits; | |
1217 | ||
1218 | vli_set(result->x, q->x, ndigits); | |
1219 | vli_set(result->y, q->y, ndigits); | |
1220 | vli_mod_sub(z, result->x, p->x, curve->p, ndigits); | |
1221 | vli_set(px, p->x, ndigits); | |
1222 | vli_set(py, p->y, ndigits); | |
1223 | xycz_add(px, py, result->x, result->y, curve->p, ndigits); | |
1224 | vli_mod_inv(z, z, curve->p, ndigits); | |
1225 | apply_z(result->x, result->y, z, curve->p, ndigits); | |
1226 | } | |
1227 | ||
1228 | /* Computes R = u1P + u2Q mod p using Shamir's trick. | |
1229 | * Based on: Kenneth MacKay's micro-ecc (2014). | |
1230 | */ | |
1231 | void ecc_point_mult_shamir(const struct ecc_point *result, | |
1232 | const u64 *u1, const struct ecc_point *p, | |
1233 | const u64 *u2, const struct ecc_point *q, | |
1234 | const struct ecc_curve *curve) | |
1235 | { | |
1236 | u64 z[ECC_MAX_DIGITS]; | |
1237 | u64 sump[2][ECC_MAX_DIGITS]; | |
1238 | u64 *rx = result->x; | |
1239 | u64 *ry = result->y; | |
1240 | unsigned int ndigits = curve->g.ndigits; | |
1241 | unsigned int num_bits; | |
1242 | struct ecc_point sum = ECC_POINT_INIT(sump[0], sump[1], ndigits); | |
1243 | const struct ecc_point *points[4]; | |
1244 | const struct ecc_point *point; | |
1245 | unsigned int idx; | |
1246 | int i; | |
1247 | ||
1248 | ecc_point_add(&sum, p, q, curve); | |
1249 | points[0] = NULL; | |
1250 | points[1] = p; | |
1251 | points[2] = q; | |
1252 | points[3] = ∑ | |
1253 | ||
1254 | num_bits = max(vli_num_bits(u1, ndigits), | |
1255 | vli_num_bits(u2, ndigits)); | |
1256 | i = num_bits - 1; | |
1257 | idx = (!!vli_test_bit(u1, i)) | ((!!vli_test_bit(u2, i)) << 1); | |
1258 | point = points[idx]; | |
1259 | ||
1260 | vli_set(rx, point->x, ndigits); | |
1261 | vli_set(ry, point->y, ndigits); | |
1262 | vli_clear(z + 1, ndigits - 1); | |
1263 | z[0] = 1; | |
1264 | ||
1265 | for (--i; i >= 0; i--) { | |
1266 | ecc_point_double_jacobian(rx, ry, z, curve->p, ndigits); | |
1267 | idx = (!!vli_test_bit(u1, i)) | ((!!vli_test_bit(u2, i)) << 1); | |
1268 | point = points[idx]; | |
1269 | if (point) { | |
1270 | u64 tx[ECC_MAX_DIGITS]; | |
1271 | u64 ty[ECC_MAX_DIGITS]; | |
1272 | u64 tz[ECC_MAX_DIGITS]; | |
1273 | ||
1274 | vli_set(tx, point->x, ndigits); | |
1275 | vli_set(ty, point->y, ndigits); | |
1276 | apply_z(tx, ty, z, curve->p, ndigits); | |
1277 | vli_mod_sub(tz, rx, tx, curve->p, ndigits); | |
1278 | xycz_add(tx, ty, rx, ry, curve->p, ndigits); | |
1279 | vli_mod_mult_fast(z, z, tz, curve->p, ndigits); | |
1280 | } | |
1281 | } | |
1282 | vli_mod_inv(z, z, curve->p, ndigits); | |
1283 | apply_z(rx, ry, z, curve->p, ndigits); | |
1284 | } | |
1285 | EXPORT_SYMBOL(ecc_point_mult_shamir); | |
1286 | ||
3c4b2390 SB |
1287 | static inline void ecc_swap_digits(const u64 *in, u64 *out, |
1288 | unsigned int ndigits) | |
1289 | { | |
f398243e | 1290 | const __be64 *src = (__force __be64 *)in; |
3c4b2390 SB |
1291 | int i; |
1292 | ||
1293 | for (i = 0; i < ndigits; i++) | |
f398243e | 1294 | out[i] = be64_to_cpu(src[ndigits - 1 - i]); |
3c4b2390 SB |
1295 | } |
1296 | ||
2eb4942b VC |
1297 | static int __ecc_is_key_valid(const struct ecc_curve *curve, |
1298 | const u64 *private_key, unsigned int ndigits) | |
3c4b2390 | 1299 | { |
2eb4942b VC |
1300 | u64 one[ECC_MAX_DIGITS] = { 1, }; |
1301 | u64 res[ECC_MAX_DIGITS]; | |
3c4b2390 SB |
1302 | |
1303 | if (!private_key) | |
1304 | return -EINVAL; | |
1305 | ||
2eb4942b | 1306 | if (curve->g.ndigits != ndigits) |
3c4b2390 SB |
1307 | return -EINVAL; |
1308 | ||
2eb4942b VC |
1309 | /* Make sure the private key is in the range [2, n-3]. */ |
1310 | if (vli_cmp(one, private_key, ndigits) != -1) | |
3c4b2390 | 1311 | return -EINVAL; |
2eb4942b VC |
1312 | vli_sub(res, curve->n, one, ndigits); |
1313 | vli_sub(res, res, one, ndigits); | |
1314 | if (vli_cmp(res, private_key, ndigits) != 1) | |
3c4b2390 SB |
1315 | return -EINVAL; |
1316 | ||
1317 | return 0; | |
1318 | } | |
1319 | ||
2eb4942b VC |
1320 | int ecc_is_key_valid(unsigned int curve_id, unsigned int ndigits, |
1321 | const u64 *private_key, unsigned int private_key_len) | |
1322 | { | |
1323 | int nbytes; | |
1324 | const struct ecc_curve *curve = ecc_get_curve(curve_id); | |
1325 | ||
1326 | nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT; | |
1327 | ||
1328 | if (private_key_len != nbytes) | |
1329 | return -EINVAL; | |
1330 | ||
1331 | return __ecc_is_key_valid(curve, private_key, ndigits); | |
1332 | } | |
4a2289da | 1333 | EXPORT_SYMBOL(ecc_is_key_valid); |
2eb4942b | 1334 | |
6755fd26 TDA |
1335 | /* |
1336 | * ECC private keys are generated using the method of extra random bits, | |
1337 | * equivalent to that described in FIPS 186-4, Appendix B.4.1. | |
1338 | * | |
1339 | * d = (c mod(n–1)) + 1 where c is a string of random bits, 64 bits longer | |
1340 | * than requested | |
1341 | * 0 <= c mod(n-1) <= n-2 and implies that | |
1342 | * 1 <= d <= n-1 | |
1343 | * | |
1344 | * This method generates a private key uniformly distributed in the range | |
1345 | * [1, n-1]. | |
1346 | */ | |
1347 | int ecc_gen_privkey(unsigned int curve_id, unsigned int ndigits, u64 *privkey) | |
1348 | { | |
1349 | const struct ecc_curve *curve = ecc_get_curve(curve_id); | |
d5c3b178 | 1350 | u64 priv[ECC_MAX_DIGITS]; |
6755fd26 TDA |
1351 | unsigned int nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT; |
1352 | unsigned int nbits = vli_num_bits(curve->n, ndigits); | |
1353 | int err; | |
1354 | ||
1355 | /* Check that N is included in Table 1 of FIPS 186-4, section 6.1.1 */ | |
d5c3b178 | 1356 | if (nbits < 160 || ndigits > ARRAY_SIZE(priv)) |
6755fd26 TDA |
1357 | return -EINVAL; |
1358 | ||
1359 | /* | |
1360 | * FIPS 186-4 recommends that the private key should be obtained from a | |
1361 | * RBG with a security strength equal to or greater than the security | |
1362 | * strength associated with N. | |
1363 | * | |
1364 | * The maximum security strength identified by NIST SP800-57pt1r4 for | |
1365 | * ECC is 256 (N >= 512). | |
1366 | * | |
1367 | * This condition is met by the default RNG because it selects a favored | |
1368 | * DRBG with a security strength of 256. | |
1369 | */ | |
1370 | if (crypto_get_default_rng()) | |
4c0e22c9 | 1371 | return -EFAULT; |
6755fd26 TDA |
1372 | |
1373 | err = crypto_rng_get_bytes(crypto_default_rng, (u8 *)priv, nbytes); | |
1374 | crypto_put_default_rng(); | |
1375 | if (err) | |
1376 | return err; | |
1377 | ||
2eb4942b VC |
1378 | /* Make sure the private key is in the valid range. */ |
1379 | if (__ecc_is_key_valid(curve, priv, ndigits)) | |
6755fd26 TDA |
1380 | return -EINVAL; |
1381 | ||
1382 | ecc_swap_digits(priv, privkey, ndigits); | |
1383 | ||
1384 | return 0; | |
1385 | } | |
4a2289da | 1386 | EXPORT_SYMBOL(ecc_gen_privkey); |
6755fd26 | 1387 | |
7380c56d TDA |
1388 | int ecc_make_pub_key(unsigned int curve_id, unsigned int ndigits, |
1389 | const u64 *private_key, u64 *public_key) | |
3c4b2390 SB |
1390 | { |
1391 | int ret = 0; | |
1392 | struct ecc_point *pk; | |
d5c3b178 | 1393 | u64 priv[ECC_MAX_DIGITS]; |
3c4b2390 SB |
1394 | const struct ecc_curve *curve = ecc_get_curve(curve_id); |
1395 | ||
d5c3b178 | 1396 | if (!private_key || !curve || ndigits > ARRAY_SIZE(priv)) { |
3c4b2390 SB |
1397 | ret = -EINVAL; |
1398 | goto out; | |
1399 | } | |
1400 | ||
ad269597 | 1401 | ecc_swap_digits(private_key, priv, ndigits); |
3c4b2390 SB |
1402 | |
1403 | pk = ecc_alloc_point(ndigits); | |
1404 | if (!pk) { | |
1405 | ret = -ENOMEM; | |
1406 | goto out; | |
1407 | } | |
1408 | ||
3da2c1df | 1409 | ecc_point_mult(pk, &curve->g, priv, NULL, curve, ndigits); |
6914dd53 SM |
1410 | |
1411 | /* SP800-56A rev 3 5.6.2.1.3 key check */ | |
1412 | if (ecc_is_pubkey_valid_full(curve, pk)) { | |
3c4b2390 SB |
1413 | ret = -EAGAIN; |
1414 | goto err_free_point; | |
1415 | } | |
1416 | ||
ad269597 TDA |
1417 | ecc_swap_digits(pk->x, public_key, ndigits); |
1418 | ecc_swap_digits(pk->y, &public_key[ndigits], ndigits); | |
3c4b2390 SB |
1419 | |
1420 | err_free_point: | |
1421 | ecc_free_point(pk); | |
1422 | out: | |
1423 | return ret; | |
1424 | } | |
4a2289da | 1425 | EXPORT_SYMBOL(ecc_make_pub_key); |
3c4b2390 | 1426 | |
ea169a30 | 1427 | /* SP800-56A section 5.6.2.3.4 partial verification: ephemeral keys only */ |
4a2289da VC |
1428 | int ecc_is_pubkey_valid_partial(const struct ecc_curve *curve, |
1429 | struct ecc_point *pk) | |
ea169a30 SM |
1430 | { |
1431 | u64 yy[ECC_MAX_DIGITS], xxx[ECC_MAX_DIGITS], w[ECC_MAX_DIGITS]; | |
1432 | ||
0d7a7864 VC |
1433 | if (WARN_ON(pk->ndigits != curve->g.ndigits)) |
1434 | return -EINVAL; | |
1435 | ||
ea169a30 SM |
1436 | /* Check 1: Verify key is not the zero point. */ |
1437 | if (ecc_point_is_zero(pk)) | |
1438 | return -EINVAL; | |
1439 | ||
1440 | /* Check 2: Verify key is in the range [1, p-1]. */ | |
1441 | if (vli_cmp(curve->p, pk->x, pk->ndigits) != 1) | |
1442 | return -EINVAL; | |
1443 | if (vli_cmp(curve->p, pk->y, pk->ndigits) != 1) | |
1444 | return -EINVAL; | |
1445 | ||
1446 | /* Check 3: Verify that y^2 == (x^3 + a·x + b) mod p */ | |
1447 | vli_mod_square_fast(yy, pk->y, curve->p, pk->ndigits); /* y^2 */ | |
1448 | vli_mod_square_fast(xxx, pk->x, curve->p, pk->ndigits); /* x^2 */ | |
1449 | vli_mod_mult_fast(xxx, xxx, pk->x, curve->p, pk->ndigits); /* x^3 */ | |
1450 | vli_mod_mult_fast(w, curve->a, pk->x, curve->p, pk->ndigits); /* a·x */ | |
1451 | vli_mod_add(w, w, curve->b, curve->p, pk->ndigits); /* a·x + b */ | |
1452 | vli_mod_add(w, w, xxx, curve->p, pk->ndigits); /* x^3 + a·x + b */ | |
1453 | if (vli_cmp(yy, w, pk->ndigits) != 0) /* Equation */ | |
1454 | return -EINVAL; | |
1455 | ||
1456 | return 0; | |
ea169a30 | 1457 | } |
4a2289da | 1458 | EXPORT_SYMBOL(ecc_is_pubkey_valid_partial); |
ea169a30 | 1459 | |
6914dd53 SM |
1460 | /* SP800-56A section 5.6.2.3.3 full verification */ |
1461 | int ecc_is_pubkey_valid_full(const struct ecc_curve *curve, | |
1462 | struct ecc_point *pk) | |
1463 | { | |
1464 | struct ecc_point *nQ; | |
1465 | ||
1466 | /* Checks 1 through 3 */ | |
1467 | int ret = ecc_is_pubkey_valid_partial(curve, pk); | |
1468 | ||
1469 | if (ret) | |
1470 | return ret; | |
1471 | ||
1472 | /* Check 4: Verify that nQ is the zero point. */ | |
1473 | nQ = ecc_alloc_point(pk->ndigits); | |
1474 | if (!nQ) | |
1475 | return -ENOMEM; | |
1476 | ||
1477 | ecc_point_mult(nQ, pk, curve->n, NULL, curve, pk->ndigits); | |
1478 | if (!ecc_point_is_zero(nQ)) | |
1479 | ret = -EINVAL; | |
1480 | ||
1481 | ecc_free_point(nQ); | |
1482 | ||
1483 | return ret; | |
1484 | } | |
1485 | EXPORT_SYMBOL(ecc_is_pubkey_valid_full); | |
1486 | ||
8f44df15 | 1487 | int crypto_ecdh_shared_secret(unsigned int curve_id, unsigned int ndigits, |
ad269597 TDA |
1488 | const u64 *private_key, const u64 *public_key, |
1489 | u64 *secret) | |
3c4b2390 SB |
1490 | { |
1491 | int ret = 0; | |
1492 | struct ecc_point *product, *pk; | |
d5c3b178 KC |
1493 | u64 priv[ECC_MAX_DIGITS]; |
1494 | u64 rand_z[ECC_MAX_DIGITS]; | |
1495 | unsigned int nbytes; | |
3c4b2390 SB |
1496 | const struct ecc_curve *curve = ecc_get_curve(curve_id); |
1497 | ||
d5c3b178 KC |
1498 | if (!private_key || !public_key || !curve || |
1499 | ndigits > ARRAY_SIZE(priv) || ndigits > ARRAY_SIZE(rand_z)) { | |
3c4b2390 SB |
1500 | ret = -EINVAL; |
1501 | goto out; | |
1502 | } | |
1503 | ||
d5c3b178 | 1504 | nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT; |
3c4b2390 | 1505 | |
d5c3b178 | 1506 | get_random_bytes(rand_z, nbytes); |
3c4b2390 SB |
1507 | |
1508 | pk = ecc_alloc_point(ndigits); | |
1509 | if (!pk) { | |
1510 | ret = -ENOMEM; | |
d5c3b178 | 1511 | goto out; |
3c4b2390 SB |
1512 | } |
1513 | ||
ea169a30 SM |
1514 | ecc_swap_digits(public_key, pk->x, ndigits); |
1515 | ecc_swap_digits(&public_key[ndigits], pk->y, ndigits); | |
1516 | ret = ecc_is_pubkey_valid_partial(curve, pk); | |
1517 | if (ret) | |
1518 | goto err_alloc_product; | |
1519 | ||
1520 | ecc_swap_digits(private_key, priv, ndigits); | |
1521 | ||
3c4b2390 SB |
1522 | product = ecc_alloc_point(ndigits); |
1523 | if (!product) { | |
1524 | ret = -ENOMEM; | |
1525 | goto err_alloc_product; | |
1526 | } | |
1527 | ||
3da2c1df | 1528 | ecc_point_mult(product, pk, priv, rand_z, curve, ndigits); |
3c4b2390 | 1529 | |
e7d2b41e | 1530 | if (ecc_point_is_zero(product)) { |
3c4b2390 | 1531 | ret = -EFAULT; |
e7d2b41e SM |
1532 | goto err_validity; |
1533 | } | |
1534 | ||
1535 | ecc_swap_digits(product->x, secret, ndigits); | |
3c4b2390 | 1536 | |
e7d2b41e SM |
1537 | err_validity: |
1538 | memzero_explicit(priv, sizeof(priv)); | |
1539 | memzero_explicit(rand_z, sizeof(rand_z)); | |
3c4b2390 SB |
1540 | ecc_free_point(product); |
1541 | err_alloc_product: | |
1542 | ecc_free_point(pk); | |
1543 | out: | |
1544 | return ret; | |
1545 | } | |
4a2289da VC |
1546 | EXPORT_SYMBOL(crypto_ecdh_shared_secret); |
1547 | ||
1548 | MODULE_LICENSE("Dual BSD/GPL"); |