xref: /openbmc/linux/crypto/ecc.c (revision 3b23dc52)
1 /*
2  * Copyright (c) 2013, Kenneth MacKay
3  * All rights reserved.
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 
27 #include <linux/random.h>
28 #include <linux/slab.h>
29 #include <linux/swab.h>
30 #include <linux/fips.h>
31 #include <crypto/ecdh.h>
32 #include <crypto/rng.h>
33 
34 #include "ecc.h"
35 #include "ecc_curve_defs.h"
36 
37 typedef struct {
38 	u64 m_low;
39 	u64 m_high;
40 } uint128_t;
41 
42 static inline const struct ecc_curve *ecc_get_curve(unsigned int curve_id)
43 {
44 	switch (curve_id) {
45 	/* In FIPS mode only allow P256 and higher */
46 	case ECC_CURVE_NIST_P192:
47 		return fips_enabled ? NULL : &nist_p192;
48 	case ECC_CURVE_NIST_P256:
49 		return &nist_p256;
50 	default:
51 		return NULL;
52 	}
53 }
54 
55 static u64 *ecc_alloc_digits_space(unsigned int ndigits)
56 {
57 	size_t len = ndigits * sizeof(u64);
58 
59 	if (!len)
60 		return NULL;
61 
62 	return kmalloc(len, GFP_KERNEL);
63 }
64 
65 static void ecc_free_digits_space(u64 *space)
66 {
67 	kzfree(space);
68 }
69 
70 static struct ecc_point *ecc_alloc_point(unsigned int ndigits)
71 {
72 	struct ecc_point *p = kmalloc(sizeof(*p), GFP_KERNEL);
73 
74 	if (!p)
75 		return NULL;
76 
77 	p->x = ecc_alloc_digits_space(ndigits);
78 	if (!p->x)
79 		goto err_alloc_x;
80 
81 	p->y = ecc_alloc_digits_space(ndigits);
82 	if (!p->y)
83 		goto err_alloc_y;
84 
85 	p->ndigits = ndigits;
86 
87 	return p;
88 
89 err_alloc_y:
90 	ecc_free_digits_space(p->x);
91 err_alloc_x:
92 	kfree(p);
93 	return NULL;
94 }
95 
96 static void ecc_free_point(struct ecc_point *p)
97 {
98 	if (!p)
99 		return;
100 
101 	kzfree(p->x);
102 	kzfree(p->y);
103 	kzfree(p);
104 }
105 
106 static void vli_clear(u64 *vli, unsigned int ndigits)
107 {
108 	int i;
109 
110 	for (i = 0; i < ndigits; i++)
111 		vli[i] = 0;
112 }
113 
114 /* Returns true if vli == 0, false otherwise. */
115 static bool vli_is_zero(const u64 *vli, unsigned int ndigits)
116 {
117 	int i;
118 
119 	for (i = 0; i < ndigits; i++) {
120 		if (vli[i])
121 			return false;
122 	}
123 
124 	return true;
125 }
126 
127 /* Returns nonzero if bit bit of vli is set. */
128 static u64 vli_test_bit(const u64 *vli, unsigned int bit)
129 {
130 	return (vli[bit / 64] & ((u64)1 << (bit % 64)));
131 }
132 
133 /* Counts the number of 64-bit "digits" in vli. */
134 static unsigned int vli_num_digits(const u64 *vli, unsigned int ndigits)
135 {
136 	int i;
137 
138 	/* Search from the end until we find a non-zero digit.
139 	 * We do it in reverse because we expect that most digits will
140 	 * be nonzero.
141 	 */
142 	for (i = ndigits - 1; i >= 0 && vli[i] == 0; i--);
143 
144 	return (i + 1);
145 }
146 
147 /* Counts the number of bits required for vli. */
148 static unsigned int vli_num_bits(const u64 *vli, unsigned int ndigits)
149 {
150 	unsigned int i, num_digits;
151 	u64 digit;
152 
153 	num_digits = vli_num_digits(vli, ndigits);
154 	if (num_digits == 0)
155 		return 0;
156 
157 	digit = vli[num_digits - 1];
158 	for (i = 0; digit; i++)
159 		digit >>= 1;
160 
161 	return ((num_digits - 1) * 64 + i);
162 }
163 
164 /* Sets dest = src. */
165 static void vli_set(u64 *dest, const u64 *src, unsigned int ndigits)
166 {
167 	int i;
168 
169 	for (i = 0; i < ndigits; i++)
170 		dest[i] = src[i];
171 }
172 
173 /* Returns sign of left - right. */
174 static int vli_cmp(const u64 *left, const u64 *right, unsigned int ndigits)
175 {
176 	int i;
177 
178 	for (i = ndigits - 1; i >= 0; i--) {
179 		if (left[i] > right[i])
180 			return 1;
181 		else if (left[i] < right[i])
182 			return -1;
183 	}
184 
185 	return 0;
186 }
187 
188 /* Computes result = in << c, returning carry. Can modify in place
189  * (if result == in). 0 < shift < 64.
190  */
191 static u64 vli_lshift(u64 *result, const u64 *in, unsigned int shift,
192 		      unsigned int ndigits)
193 {
194 	u64 carry = 0;
195 	int i;
196 
197 	for (i = 0; i < ndigits; i++) {
198 		u64 temp = in[i];
199 
200 		result[i] = (temp << shift) | carry;
201 		carry = temp >> (64 - shift);
202 	}
203 
204 	return carry;
205 }
206 
207 /* Computes vli = vli >> 1. */
208 static void vli_rshift1(u64 *vli, unsigned int ndigits)
209 {
210 	u64 *end = vli;
211 	u64 carry = 0;
212 
213 	vli += ndigits;
214 
215 	while (vli-- > end) {
216 		u64 temp = *vli;
217 		*vli = (temp >> 1) | carry;
218 		carry = temp << 63;
219 	}
220 }
221 
222 /* Computes result = left + right, returning carry. Can modify in place. */
223 static u64 vli_add(u64 *result, const u64 *left, const u64 *right,
224 		   unsigned int ndigits)
225 {
226 	u64 carry = 0;
227 	int i;
228 
229 	for (i = 0; i < ndigits; i++) {
230 		u64 sum;
231 
232 		sum = left[i] + right[i] + carry;
233 		if (sum != left[i])
234 			carry = (sum < left[i]);
235 
236 		result[i] = sum;
237 	}
238 
239 	return carry;
240 }
241 
242 /* Computes result = left - right, returning borrow. Can modify in place. */
243 static u64 vli_sub(u64 *result, const u64 *left, const u64 *right,
244 		   unsigned int ndigits)
245 {
246 	u64 borrow = 0;
247 	int i;
248 
249 	for (i = 0; i < ndigits; i++) {
250 		u64 diff;
251 
252 		diff = left[i] - right[i] - borrow;
253 		if (diff != left[i])
254 			borrow = (diff > left[i]);
255 
256 		result[i] = diff;
257 	}
258 
259 	return borrow;
260 }
261 
262 static uint128_t mul_64_64(u64 left, u64 right)
263 {
264 	u64 a0 = left & 0xffffffffull;
265 	u64 a1 = left >> 32;
266 	u64 b0 = right & 0xffffffffull;
267 	u64 b1 = right >> 32;
268 	u64 m0 = a0 * b0;
269 	u64 m1 = a0 * b1;
270 	u64 m2 = a1 * b0;
271 	u64 m3 = a1 * b1;
272 	uint128_t result;
273 
274 	m2 += (m0 >> 32);
275 	m2 += m1;
276 
277 	/* Overflow */
278 	if (m2 < m1)
279 		m3 += 0x100000000ull;
280 
281 	result.m_low = (m0 & 0xffffffffull) | (m2 << 32);
282 	result.m_high = m3 + (m2 >> 32);
283 
284 	return result;
285 }
286 
287 static uint128_t add_128_128(uint128_t a, uint128_t b)
288 {
289 	uint128_t result;
290 
291 	result.m_low = a.m_low + b.m_low;
292 	result.m_high = a.m_high + b.m_high + (result.m_low < a.m_low);
293 
294 	return result;
295 }
296 
297 static void vli_mult(u64 *result, const u64 *left, const u64 *right,
298 		     unsigned int ndigits)
299 {
300 	uint128_t r01 = { 0, 0 };
301 	u64 r2 = 0;
302 	unsigned int i, k;
303 
304 	/* Compute each digit of result in sequence, maintaining the
305 	 * carries.
306 	 */
307 	for (k = 0; k < ndigits * 2 - 1; k++) {
308 		unsigned int min;
309 
310 		if (k < ndigits)
311 			min = 0;
312 		else
313 			min = (k + 1) - ndigits;
314 
315 		for (i = min; i <= k && i < ndigits; i++) {
316 			uint128_t product;
317 
318 			product = mul_64_64(left[i], right[k - i]);
319 
320 			r01 = add_128_128(r01, product);
321 			r2 += (r01.m_high < product.m_high);
322 		}
323 
324 		result[k] = r01.m_low;
325 		r01.m_low = r01.m_high;
326 		r01.m_high = r2;
327 		r2 = 0;
328 	}
329 
330 	result[ndigits * 2 - 1] = r01.m_low;
331 }
332 
333 static void vli_square(u64 *result, const u64 *left, unsigned int ndigits)
334 {
335 	uint128_t r01 = { 0, 0 };
336 	u64 r2 = 0;
337 	int i, k;
338 
339 	for (k = 0; k < ndigits * 2 - 1; k++) {
340 		unsigned int min;
341 
342 		if (k < ndigits)
343 			min = 0;
344 		else
345 			min = (k + 1) - ndigits;
346 
347 		for (i = min; i <= k && i <= k - i; i++) {
348 			uint128_t product;
349 
350 			product = mul_64_64(left[i], left[k - i]);
351 
352 			if (i < k - i) {
353 				r2 += product.m_high >> 63;
354 				product.m_high = (product.m_high << 1) |
355 						 (product.m_low >> 63);
356 				product.m_low <<= 1;
357 			}
358 
359 			r01 = add_128_128(r01, product);
360 			r2 += (r01.m_high < product.m_high);
361 		}
362 
363 		result[k] = r01.m_low;
364 		r01.m_low = r01.m_high;
365 		r01.m_high = r2;
366 		r2 = 0;
367 	}
368 
369 	result[ndigits * 2 - 1] = r01.m_low;
370 }
371 
372 /* Computes result = (left + right) % mod.
373  * Assumes that left < mod and right < mod, result != mod.
374  */
375 static void vli_mod_add(u64 *result, const u64 *left, const u64 *right,
376 			const u64 *mod, unsigned int ndigits)
377 {
378 	u64 carry;
379 
380 	carry = vli_add(result, left, right, ndigits);
381 
382 	/* result > mod (result = mod + remainder), so subtract mod to
383 	 * get remainder.
384 	 */
385 	if (carry || vli_cmp(result, mod, ndigits) >= 0)
386 		vli_sub(result, result, mod, ndigits);
387 }
388 
389 /* Computes result = (left - right) % mod.
390  * Assumes that left < mod and right < mod, result != mod.
391  */
392 static void vli_mod_sub(u64 *result, const u64 *left, const u64 *right,
393 			const u64 *mod, unsigned int ndigits)
394 {
395 	u64 borrow = vli_sub(result, left, right, ndigits);
396 
397 	/* In this case, p_result == -diff == (max int) - diff.
398 	 * Since -x % d == d - x, we can get the correct result from
399 	 * result + mod (with overflow).
400 	 */
401 	if (borrow)
402 		vli_add(result, result, mod, ndigits);
403 }
404 
405 /* Computes p_result = p_product % curve_p.
406  * See algorithm 5 and 6 from
407  * http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf
408  */
409 static void vli_mmod_fast_192(u64 *result, const u64 *product,
410 			      const u64 *curve_prime, u64 *tmp)
411 {
412 	const unsigned int ndigits = 3;
413 	int carry;
414 
415 	vli_set(result, product, ndigits);
416 
417 	vli_set(tmp, &product[3], ndigits);
418 	carry = vli_add(result, result, tmp, ndigits);
419 
420 	tmp[0] = 0;
421 	tmp[1] = product[3];
422 	tmp[2] = product[4];
423 	carry += vli_add(result, result, tmp, ndigits);
424 
425 	tmp[0] = tmp[1] = product[5];
426 	tmp[2] = 0;
427 	carry += vli_add(result, result, tmp, ndigits);
428 
429 	while (carry || vli_cmp(curve_prime, result, ndigits) != 1)
430 		carry -= vli_sub(result, result, curve_prime, ndigits);
431 }
432 
433 /* Computes result = product % curve_prime
434  * from http://www.nsa.gov/ia/_files/nist-routines.pdf
435  */
436 static void vli_mmod_fast_256(u64 *result, const u64 *product,
437 			      const u64 *curve_prime, u64 *tmp)
438 {
439 	int carry;
440 	const unsigned int ndigits = 4;
441 
442 	/* t */
443 	vli_set(result, product, ndigits);
444 
445 	/* s1 */
446 	tmp[0] = 0;
447 	tmp[1] = product[5] & 0xffffffff00000000ull;
448 	tmp[2] = product[6];
449 	tmp[3] = product[7];
450 	carry = vli_lshift(tmp, tmp, 1, ndigits);
451 	carry += vli_add(result, result, tmp, ndigits);
452 
453 	/* s2 */
454 	tmp[1] = product[6] << 32;
455 	tmp[2] = (product[6] >> 32) | (product[7] << 32);
456 	tmp[3] = product[7] >> 32;
457 	carry += vli_lshift(tmp, tmp, 1, ndigits);
458 	carry += vli_add(result, result, tmp, ndigits);
459 
460 	/* s3 */
461 	tmp[0] = product[4];
462 	tmp[1] = product[5] & 0xffffffff;
463 	tmp[2] = 0;
464 	tmp[3] = product[7];
465 	carry += vli_add(result, result, tmp, ndigits);
466 
467 	/* s4 */
468 	tmp[0] = (product[4] >> 32) | (product[5] << 32);
469 	tmp[1] = (product[5] >> 32) | (product[6] & 0xffffffff00000000ull);
470 	tmp[2] = product[7];
471 	tmp[3] = (product[6] >> 32) | (product[4] << 32);
472 	carry += vli_add(result, result, tmp, ndigits);
473 
474 	/* d1 */
475 	tmp[0] = (product[5] >> 32) | (product[6] << 32);
476 	tmp[1] = (product[6] >> 32);
477 	tmp[2] = 0;
478 	tmp[3] = (product[4] & 0xffffffff) | (product[5] << 32);
479 	carry -= vli_sub(result, result, tmp, ndigits);
480 
481 	/* d2 */
482 	tmp[0] = product[6];
483 	tmp[1] = product[7];
484 	tmp[2] = 0;
485 	tmp[3] = (product[4] >> 32) | (product[5] & 0xffffffff00000000ull);
486 	carry -= vli_sub(result, result, tmp, ndigits);
487 
488 	/* d3 */
489 	tmp[0] = (product[6] >> 32) | (product[7] << 32);
490 	tmp[1] = (product[7] >> 32) | (product[4] << 32);
491 	tmp[2] = (product[4] >> 32) | (product[5] << 32);
492 	tmp[3] = (product[6] << 32);
493 	carry -= vli_sub(result, result, tmp, ndigits);
494 
495 	/* d4 */
496 	tmp[0] = product[7];
497 	tmp[1] = product[4] & 0xffffffff00000000ull;
498 	tmp[2] = product[5];
499 	tmp[3] = product[6] & 0xffffffff00000000ull;
500 	carry -= vli_sub(result, result, tmp, ndigits);
501 
502 	if (carry < 0) {
503 		do {
504 			carry += vli_add(result, result, curve_prime, ndigits);
505 		} while (carry < 0);
506 	} else {
507 		while (carry || vli_cmp(curve_prime, result, ndigits) != 1)
508 			carry -= vli_sub(result, result, curve_prime, ndigits);
509 	}
510 }
511 
512 /* Computes result = product % curve_prime
513  *  from http://www.nsa.gov/ia/_files/nist-routines.pdf
514 */
515 static bool vli_mmod_fast(u64 *result, u64 *product,
516 			  const u64 *curve_prime, unsigned int ndigits)
517 {
518 	u64 tmp[2 * ECC_MAX_DIGITS];
519 
520 	switch (ndigits) {
521 	case 3:
522 		vli_mmod_fast_192(result, product, curve_prime, tmp);
523 		break;
524 	case 4:
525 		vli_mmod_fast_256(result, product, curve_prime, tmp);
526 		break;
527 	default:
528 		pr_err("unsupports digits size!\n");
529 		return false;
530 	}
531 
532 	return true;
533 }
534 
535 /* Computes result = (left * right) % curve_prime. */
536 static void vli_mod_mult_fast(u64 *result, const u64 *left, const u64 *right,
537 			      const u64 *curve_prime, unsigned int ndigits)
538 {
539 	u64 product[2 * ECC_MAX_DIGITS];
540 
541 	vli_mult(product, left, right, ndigits);
542 	vli_mmod_fast(result, product, curve_prime, ndigits);
543 }
544 
545 /* Computes result = left^2 % curve_prime. */
546 static void vli_mod_square_fast(u64 *result, const u64 *left,
547 				const u64 *curve_prime, unsigned int ndigits)
548 {
549 	u64 product[2 * ECC_MAX_DIGITS];
550 
551 	vli_square(product, left, ndigits);
552 	vli_mmod_fast(result, product, curve_prime, ndigits);
553 }
554 
555 #define EVEN(vli) (!(vli[0] & 1))
556 /* Computes result = (1 / p_input) % mod. All VLIs are the same size.
557  * See "From Euclid's GCD to Montgomery Multiplication to the Great Divide"
558  * https://labs.oracle.com/techrep/2001/smli_tr-2001-95.pdf
559  */
560 static void vli_mod_inv(u64 *result, const u64 *input, const u64 *mod,
561 			unsigned int ndigits)
562 {
563 	u64 a[ECC_MAX_DIGITS], b[ECC_MAX_DIGITS];
564 	u64 u[ECC_MAX_DIGITS], v[ECC_MAX_DIGITS];
565 	u64 carry;
566 	int cmp_result;
567 
568 	if (vli_is_zero(input, ndigits)) {
569 		vli_clear(result, ndigits);
570 		return;
571 	}
572 
573 	vli_set(a, input, ndigits);
574 	vli_set(b, mod, ndigits);
575 	vli_clear(u, ndigits);
576 	u[0] = 1;
577 	vli_clear(v, ndigits);
578 
579 	while ((cmp_result = vli_cmp(a, b, ndigits)) != 0) {
580 		carry = 0;
581 
582 		if (EVEN(a)) {
583 			vli_rshift1(a, ndigits);
584 
585 			if (!EVEN(u))
586 				carry = vli_add(u, u, mod, ndigits);
587 
588 			vli_rshift1(u, ndigits);
589 			if (carry)
590 				u[ndigits - 1] |= 0x8000000000000000ull;
591 		} else if (EVEN(b)) {
592 			vli_rshift1(b, ndigits);
593 
594 			if (!EVEN(v))
595 				carry = vli_add(v, v, mod, ndigits);
596 
597 			vli_rshift1(v, ndigits);
598 			if (carry)
599 				v[ndigits - 1] |= 0x8000000000000000ull;
600 		} else if (cmp_result > 0) {
601 			vli_sub(a, a, b, ndigits);
602 			vli_rshift1(a, ndigits);
603 
604 			if (vli_cmp(u, v, ndigits) < 0)
605 				vli_add(u, u, mod, ndigits);
606 
607 			vli_sub(u, u, v, ndigits);
608 			if (!EVEN(u))
609 				carry = vli_add(u, u, mod, ndigits);
610 
611 			vli_rshift1(u, ndigits);
612 			if (carry)
613 				u[ndigits - 1] |= 0x8000000000000000ull;
614 		} else {
615 			vli_sub(b, b, a, ndigits);
616 			vli_rshift1(b, ndigits);
617 
618 			if (vli_cmp(v, u, ndigits) < 0)
619 				vli_add(v, v, mod, ndigits);
620 
621 			vli_sub(v, v, u, ndigits);
622 			if (!EVEN(v))
623 				carry = vli_add(v, v, mod, ndigits);
624 
625 			vli_rshift1(v, ndigits);
626 			if (carry)
627 				v[ndigits - 1] |= 0x8000000000000000ull;
628 		}
629 	}
630 
631 	vli_set(result, u, ndigits);
632 }
633 
634 /* ------ Point operations ------ */
635 
636 /* Returns true if p_point is the point at infinity, false otherwise. */
637 static bool ecc_point_is_zero(const struct ecc_point *point)
638 {
639 	return (vli_is_zero(point->x, point->ndigits) &&
640 		vli_is_zero(point->y, point->ndigits));
641 }
642 
643 /* Point multiplication algorithm using Montgomery's ladder with co-Z
644  * coordinates. From http://eprint.iacr.org/2011/338.pdf
645  */
646 
647 /* Double in place */
648 static void ecc_point_double_jacobian(u64 *x1, u64 *y1, u64 *z1,
649 				      u64 *curve_prime, unsigned int ndigits)
650 {
651 	/* t1 = x, t2 = y, t3 = z */
652 	u64 t4[ECC_MAX_DIGITS];
653 	u64 t5[ECC_MAX_DIGITS];
654 
655 	if (vli_is_zero(z1, ndigits))
656 		return;
657 
658 	/* t4 = y1^2 */
659 	vli_mod_square_fast(t4, y1, curve_prime, ndigits);
660 	/* t5 = x1*y1^2 = A */
661 	vli_mod_mult_fast(t5, x1, t4, curve_prime, ndigits);
662 	/* t4 = y1^4 */
663 	vli_mod_square_fast(t4, t4, curve_prime, ndigits);
664 	/* t2 = y1*z1 = z3 */
665 	vli_mod_mult_fast(y1, y1, z1, curve_prime, ndigits);
666 	/* t3 = z1^2 */
667 	vli_mod_square_fast(z1, z1, curve_prime, ndigits);
668 
669 	/* t1 = x1 + z1^2 */
670 	vli_mod_add(x1, x1, z1, curve_prime, ndigits);
671 	/* t3 = 2*z1^2 */
672 	vli_mod_add(z1, z1, z1, curve_prime, ndigits);
673 	/* t3 = x1 - z1^2 */
674 	vli_mod_sub(z1, x1, z1, curve_prime, ndigits);
675 	/* t1 = x1^2 - z1^4 */
676 	vli_mod_mult_fast(x1, x1, z1, curve_prime, ndigits);
677 
678 	/* t3 = 2*(x1^2 - z1^4) */
679 	vli_mod_add(z1, x1, x1, curve_prime, ndigits);
680 	/* t1 = 3*(x1^2 - z1^4) */
681 	vli_mod_add(x1, x1, z1, curve_prime, ndigits);
682 	if (vli_test_bit(x1, 0)) {
683 		u64 carry = vli_add(x1, x1, curve_prime, ndigits);
684 
685 		vli_rshift1(x1, ndigits);
686 		x1[ndigits - 1] |= carry << 63;
687 	} else {
688 		vli_rshift1(x1, ndigits);
689 	}
690 	/* t1 = 3/2*(x1^2 - z1^4) = B */
691 
692 	/* t3 = B^2 */
693 	vli_mod_square_fast(z1, x1, curve_prime, ndigits);
694 	/* t3 = B^2 - A */
695 	vli_mod_sub(z1, z1, t5, curve_prime, ndigits);
696 	/* t3 = B^2 - 2A = x3 */
697 	vli_mod_sub(z1, z1, t5, curve_prime, ndigits);
698 	/* t5 = A - x3 */
699 	vli_mod_sub(t5, t5, z1, curve_prime, ndigits);
700 	/* t1 = B * (A - x3) */
701 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
702 	/* t4 = B * (A - x3) - y1^4 = y3 */
703 	vli_mod_sub(t4, x1, t4, curve_prime, ndigits);
704 
705 	vli_set(x1, z1, ndigits);
706 	vli_set(z1, y1, ndigits);
707 	vli_set(y1, t4, ndigits);
708 }
709 
710 /* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */
711 static void apply_z(u64 *x1, u64 *y1, u64 *z, u64 *curve_prime,
712 		    unsigned int ndigits)
713 {
714 	u64 t1[ECC_MAX_DIGITS];
715 
716 	vli_mod_square_fast(t1, z, curve_prime, ndigits);    /* z^2 */
717 	vli_mod_mult_fast(x1, x1, t1, curve_prime, ndigits); /* x1 * z^2 */
718 	vli_mod_mult_fast(t1, t1, z, curve_prime, ndigits);  /* z^3 */
719 	vli_mod_mult_fast(y1, y1, t1, curve_prime, ndigits); /* y1 * z^3 */
720 }
721 
722 /* P = (x1, y1) => 2P, (x2, y2) => P' */
723 static void xycz_initial_double(u64 *x1, u64 *y1, u64 *x2, u64 *y2,
724 				u64 *p_initial_z, u64 *curve_prime,
725 				unsigned int ndigits)
726 {
727 	u64 z[ECC_MAX_DIGITS];
728 
729 	vli_set(x2, x1, ndigits);
730 	vli_set(y2, y1, ndigits);
731 
732 	vli_clear(z, ndigits);
733 	z[0] = 1;
734 
735 	if (p_initial_z)
736 		vli_set(z, p_initial_z, ndigits);
737 
738 	apply_z(x1, y1, z, curve_prime, ndigits);
739 
740 	ecc_point_double_jacobian(x1, y1, z, curve_prime, ndigits);
741 
742 	apply_z(x2, y2, z, curve_prime, ndigits);
743 }
744 
745 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
746  * Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3)
747  * or P => P', Q => P + Q
748  */
749 static void xycz_add(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime,
750 		     unsigned int ndigits)
751 {
752 	/* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
753 	u64 t5[ECC_MAX_DIGITS];
754 
755 	/* t5 = x2 - x1 */
756 	vli_mod_sub(t5, x2, x1, curve_prime, ndigits);
757 	/* t5 = (x2 - x1)^2 = A */
758 	vli_mod_square_fast(t5, t5, curve_prime, ndigits);
759 	/* t1 = x1*A = B */
760 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
761 	/* t3 = x2*A = C */
762 	vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits);
763 	/* t4 = y2 - y1 */
764 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
765 	/* t5 = (y2 - y1)^2 = D */
766 	vli_mod_square_fast(t5, y2, curve_prime, ndigits);
767 
768 	/* t5 = D - B */
769 	vli_mod_sub(t5, t5, x1, curve_prime, ndigits);
770 	/* t5 = D - B - C = x3 */
771 	vli_mod_sub(t5, t5, x2, curve_prime, ndigits);
772 	/* t3 = C - B */
773 	vli_mod_sub(x2, x2, x1, curve_prime, ndigits);
774 	/* t2 = y1*(C - B) */
775 	vli_mod_mult_fast(y1, y1, x2, curve_prime, ndigits);
776 	/* t3 = B - x3 */
777 	vli_mod_sub(x2, x1, t5, curve_prime, ndigits);
778 	/* t4 = (y2 - y1)*(B - x3) */
779 	vli_mod_mult_fast(y2, y2, x2, curve_prime, ndigits);
780 	/* t4 = y3 */
781 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
782 
783 	vli_set(x2, t5, ndigits);
784 }
785 
786 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
787  * Output P + Q = (x3, y3, Z3), P - Q = (x3', y3', Z3)
788  * or P => P - Q, Q => P + Q
789  */
790 static void xycz_add_c(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime,
791 		       unsigned int ndigits)
792 {
793 	/* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
794 	u64 t5[ECC_MAX_DIGITS];
795 	u64 t6[ECC_MAX_DIGITS];
796 	u64 t7[ECC_MAX_DIGITS];
797 
798 	/* t5 = x2 - x1 */
799 	vli_mod_sub(t5, x2, x1, curve_prime, ndigits);
800 	/* t5 = (x2 - x1)^2 = A */
801 	vli_mod_square_fast(t5, t5, curve_prime, ndigits);
802 	/* t1 = x1*A = B */
803 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
804 	/* t3 = x2*A = C */
805 	vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits);
806 	/* t4 = y2 + y1 */
807 	vli_mod_add(t5, y2, y1, curve_prime, ndigits);
808 	/* t4 = y2 - y1 */
809 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
810 
811 	/* t6 = C - B */
812 	vli_mod_sub(t6, x2, x1, curve_prime, ndigits);
813 	/* t2 = y1 * (C - B) */
814 	vli_mod_mult_fast(y1, y1, t6, curve_prime, ndigits);
815 	/* t6 = B + C */
816 	vli_mod_add(t6, x1, x2, curve_prime, ndigits);
817 	/* t3 = (y2 - y1)^2 */
818 	vli_mod_square_fast(x2, y2, curve_prime, ndigits);
819 	/* t3 = x3 */
820 	vli_mod_sub(x2, x2, t6, curve_prime, ndigits);
821 
822 	/* t7 = B - x3 */
823 	vli_mod_sub(t7, x1, x2, curve_prime, ndigits);
824 	/* t4 = (y2 - y1)*(B - x3) */
825 	vli_mod_mult_fast(y2, y2, t7, curve_prime, ndigits);
826 	/* t4 = y3 */
827 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
828 
829 	/* t7 = (y2 + y1)^2 = F */
830 	vli_mod_square_fast(t7, t5, curve_prime, ndigits);
831 	/* t7 = x3' */
832 	vli_mod_sub(t7, t7, t6, curve_prime, ndigits);
833 	/* t6 = x3' - B */
834 	vli_mod_sub(t6, t7, x1, curve_prime, ndigits);
835 	/* t6 = (y2 + y1)*(x3' - B) */
836 	vli_mod_mult_fast(t6, t6, t5, curve_prime, ndigits);
837 	/* t2 = y3' */
838 	vli_mod_sub(y1, t6, y1, curve_prime, ndigits);
839 
840 	vli_set(x1, t7, ndigits);
841 }
842 
843 static void ecc_point_mult(struct ecc_point *result,
844 			   const struct ecc_point *point, const u64 *scalar,
845 			   u64 *initial_z, u64 *curve_prime,
846 			   unsigned int ndigits)
847 {
848 	/* R0 and R1 */
849 	u64 rx[2][ECC_MAX_DIGITS];
850 	u64 ry[2][ECC_MAX_DIGITS];
851 	u64 z[ECC_MAX_DIGITS];
852 	int i, nb;
853 	int num_bits = vli_num_bits(scalar, ndigits);
854 
855 	vli_set(rx[1], point->x, ndigits);
856 	vli_set(ry[1], point->y, ndigits);
857 
858 	xycz_initial_double(rx[1], ry[1], rx[0], ry[0], initial_z, curve_prime,
859 			    ndigits);
860 
861 	for (i = num_bits - 2; i > 0; i--) {
862 		nb = !vli_test_bit(scalar, i);
863 		xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime,
864 			   ndigits);
865 		xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime,
866 			 ndigits);
867 	}
868 
869 	nb = !vli_test_bit(scalar, 0);
870 	xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime,
871 		   ndigits);
872 
873 	/* Find final 1/Z value. */
874 	/* X1 - X0 */
875 	vli_mod_sub(z, rx[1], rx[0], curve_prime, ndigits);
876 	/* Yb * (X1 - X0) */
877 	vli_mod_mult_fast(z, z, ry[1 - nb], curve_prime, ndigits);
878 	/* xP * Yb * (X1 - X0) */
879 	vli_mod_mult_fast(z, z, point->x, curve_prime, ndigits);
880 
881 	/* 1 / (xP * Yb * (X1 - X0)) */
882 	vli_mod_inv(z, z, curve_prime, point->ndigits);
883 
884 	/* yP / (xP * Yb * (X1 - X0)) */
885 	vli_mod_mult_fast(z, z, point->y, curve_prime, ndigits);
886 	/* Xb * yP / (xP * Yb * (X1 - X0)) */
887 	vli_mod_mult_fast(z, z, rx[1 - nb], curve_prime, ndigits);
888 	/* End 1/Z calculation */
889 
890 	xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime, ndigits);
891 
892 	apply_z(rx[0], ry[0], z, curve_prime, ndigits);
893 
894 	vli_set(result->x, rx[0], ndigits);
895 	vli_set(result->y, ry[0], ndigits);
896 }
897 
898 static inline void ecc_swap_digits(const u64 *in, u64 *out,
899 				   unsigned int ndigits)
900 {
901 	int i;
902 
903 	for (i = 0; i < ndigits; i++)
904 		out[i] = __swab64(in[ndigits - 1 - i]);
905 }
906 
907 int ecc_is_key_valid(unsigned int curve_id, unsigned int ndigits,
908 		     const u64 *private_key, unsigned int private_key_len)
909 {
910 	int nbytes;
911 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
912 
913 	if (!private_key)
914 		return -EINVAL;
915 
916 	nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
917 
918 	if (private_key_len != nbytes)
919 		return -EINVAL;
920 
921 	if (vli_is_zero(private_key, ndigits))
922 		return -EINVAL;
923 
924 	/* Make sure the private key is in the range [1, n-1]. */
925 	if (vli_cmp(curve->n, private_key, ndigits) != 1)
926 		return -EINVAL;
927 
928 	return 0;
929 }
930 
931 /*
932  * ECC private keys are generated using the method of extra random bits,
933  * equivalent to that described in FIPS 186-4, Appendix B.4.1.
934  *
935  * d = (c mod(n–1)) + 1    where c is a string of random bits, 64 bits longer
936  *                         than requested
937  * 0 <= c mod(n-1) <= n-2  and implies that
938  * 1 <= d <= n-1
939  *
940  * This method generates a private key uniformly distributed in the range
941  * [1, n-1].
942  */
943 int ecc_gen_privkey(unsigned int curve_id, unsigned int ndigits, u64 *privkey)
944 {
945 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
946 	u64 priv[ECC_MAX_DIGITS];
947 	unsigned int nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
948 	unsigned int nbits = vli_num_bits(curve->n, ndigits);
949 	int err;
950 
951 	/* Check that N is included in Table 1 of FIPS 186-4, section 6.1.1 */
952 	if (nbits < 160 || ndigits > ARRAY_SIZE(priv))
953 		return -EINVAL;
954 
955 	/*
956 	 * FIPS 186-4 recommends that the private key should be obtained from a
957 	 * RBG with a security strength equal to or greater than the security
958 	 * strength associated with N.
959 	 *
960 	 * The maximum security strength identified by NIST SP800-57pt1r4 for
961 	 * ECC is 256 (N >= 512).
962 	 *
963 	 * This condition is met by the default RNG because it selects a favored
964 	 * DRBG with a security strength of 256.
965 	 */
966 	if (crypto_get_default_rng())
967 		return -EFAULT;
968 
969 	err = crypto_rng_get_bytes(crypto_default_rng, (u8 *)priv, nbytes);
970 	crypto_put_default_rng();
971 	if (err)
972 		return err;
973 
974 	if (vli_is_zero(priv, ndigits))
975 		return -EINVAL;
976 
977 	/* Make sure the private key is in the range [1, n-1]. */
978 	if (vli_cmp(curve->n, priv, ndigits) != 1)
979 		return -EINVAL;
980 
981 	ecc_swap_digits(priv, privkey, ndigits);
982 
983 	return 0;
984 }
985 
986 int ecc_make_pub_key(unsigned int curve_id, unsigned int ndigits,
987 		     const u64 *private_key, u64 *public_key)
988 {
989 	int ret = 0;
990 	struct ecc_point *pk;
991 	u64 priv[ECC_MAX_DIGITS];
992 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
993 
994 	if (!private_key || !curve || ndigits > ARRAY_SIZE(priv)) {
995 		ret = -EINVAL;
996 		goto out;
997 	}
998 
999 	ecc_swap_digits(private_key, priv, ndigits);
1000 
1001 	pk = ecc_alloc_point(ndigits);
1002 	if (!pk) {
1003 		ret = -ENOMEM;
1004 		goto out;
1005 	}
1006 
1007 	ecc_point_mult(pk, &curve->g, priv, NULL, curve->p, ndigits);
1008 	if (ecc_point_is_zero(pk)) {
1009 		ret = -EAGAIN;
1010 		goto err_free_point;
1011 	}
1012 
1013 	ecc_swap_digits(pk->x, public_key, ndigits);
1014 	ecc_swap_digits(pk->y, &public_key[ndigits], ndigits);
1015 
1016 err_free_point:
1017 	ecc_free_point(pk);
1018 out:
1019 	return ret;
1020 }
1021 
1022 int crypto_ecdh_shared_secret(unsigned int curve_id, unsigned int ndigits,
1023 			      const u64 *private_key, const u64 *public_key,
1024 			      u64 *secret)
1025 {
1026 	int ret = 0;
1027 	struct ecc_point *product, *pk;
1028 	u64 priv[ECC_MAX_DIGITS];
1029 	u64 rand_z[ECC_MAX_DIGITS];
1030 	unsigned int nbytes;
1031 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
1032 
1033 	if (!private_key || !public_key || !curve ||
1034 	    ndigits > ARRAY_SIZE(priv) || ndigits > ARRAY_SIZE(rand_z)) {
1035 		ret = -EINVAL;
1036 		goto out;
1037 	}
1038 
1039 	nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
1040 
1041 	get_random_bytes(rand_z, nbytes);
1042 
1043 	pk = ecc_alloc_point(ndigits);
1044 	if (!pk) {
1045 		ret = -ENOMEM;
1046 		goto out;
1047 	}
1048 
1049 	product = ecc_alloc_point(ndigits);
1050 	if (!product) {
1051 		ret = -ENOMEM;
1052 		goto err_alloc_product;
1053 	}
1054 
1055 	ecc_swap_digits(public_key, pk->x, ndigits);
1056 	ecc_swap_digits(&public_key[ndigits], pk->y, ndigits);
1057 	ecc_swap_digits(private_key, priv, ndigits);
1058 
1059 	ecc_point_mult(product, pk, priv, rand_z, curve->p, ndigits);
1060 
1061 	ecc_swap_digits(product->x, secret, ndigits);
1062 
1063 	if (ecc_point_is_zero(product))
1064 		ret = -EFAULT;
1065 
1066 	ecc_free_point(product);
1067 err_alloc_product:
1068 	ecc_free_point(pk);
1069 out:
1070 	return ret;
1071 }
1072