xref: /linux/crypto/ecc.c (revision 16217dc79dbc599b110dda26d0421df47904bba4)
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 
33 #include "ecc.h"
34 #include "ecc_curve_defs.h"
35 
36 typedef struct {
37 	u64 m_low;
38 	u64 m_high;
39 } uint128_t;
40 
41 static inline const struct ecc_curve *ecc_get_curve(unsigned int curve_id)
42 {
43 	switch (curve_id) {
44 	/* In FIPS mode only allow P256 and higher */
45 	case ECC_CURVE_NIST_P192:
46 		return fips_enabled ? NULL : &nist_p192;
47 	case ECC_CURVE_NIST_P256:
48 		return &nist_p256;
49 	default:
50 		return NULL;
51 	}
52 }
53 
54 static u64 *ecc_alloc_digits_space(unsigned int ndigits)
55 {
56 	size_t len = ndigits * sizeof(u64);
57 
58 	if (!len)
59 		return NULL;
60 
61 	return kmalloc(len, GFP_KERNEL);
62 }
63 
64 static void ecc_free_digits_space(u64 *space)
65 {
66 	kzfree(space);
67 }
68 
69 static struct ecc_point *ecc_alloc_point(unsigned int ndigits)
70 {
71 	struct ecc_point *p = kmalloc(sizeof(*p), GFP_KERNEL);
72 
73 	if (!p)
74 		return NULL;
75 
76 	p->x = ecc_alloc_digits_space(ndigits);
77 	if (!p->x)
78 		goto err_alloc_x;
79 
80 	p->y = ecc_alloc_digits_space(ndigits);
81 	if (!p->y)
82 		goto err_alloc_y;
83 
84 	p->ndigits = ndigits;
85 
86 	return p;
87 
88 err_alloc_y:
89 	ecc_free_digits_space(p->x);
90 err_alloc_x:
91 	kfree(p);
92 	return NULL;
93 }
94 
95 static void ecc_free_point(struct ecc_point *p)
96 {
97 	if (!p)
98 		return;
99 
100 	kzfree(p->x);
101 	kzfree(p->y);
102 	kzfree(p);
103 }
104 
105 static void vli_clear(u64 *vli, unsigned int ndigits)
106 {
107 	int i;
108 
109 	for (i = 0; i < ndigits; i++)
110 		vli[i] = 0;
111 }
112 
113 /* Returns true if vli == 0, false otherwise. */
114 static bool vli_is_zero(const u64 *vli, unsigned int ndigits)
115 {
116 	int i;
117 
118 	for (i = 0; i < ndigits; i++) {
119 		if (vli[i])
120 			return false;
121 	}
122 
123 	return true;
124 }
125 
126 /* Returns nonzero if bit bit of vli is set. */
127 static u64 vli_test_bit(const u64 *vli, unsigned int bit)
128 {
129 	return (vli[bit / 64] & ((u64)1 << (bit % 64)));
130 }
131 
132 /* Counts the number of 64-bit "digits" in vli. */
133 static unsigned int vli_num_digits(const u64 *vli, unsigned int ndigits)
134 {
135 	int i;
136 
137 	/* Search from the end until we find a non-zero digit.
138 	 * We do it in reverse because we expect that most digits will
139 	 * be nonzero.
140 	 */
141 	for (i = ndigits - 1; i >= 0 && vli[i] == 0; i--);
142 
143 	return (i + 1);
144 }
145 
146 /* Counts the number of bits required for vli. */
147 static unsigned int vli_num_bits(const u64 *vli, unsigned int ndigits)
148 {
149 	unsigned int i, num_digits;
150 	u64 digit;
151 
152 	num_digits = vli_num_digits(vli, ndigits);
153 	if (num_digits == 0)
154 		return 0;
155 
156 	digit = vli[num_digits - 1];
157 	for (i = 0; digit; i++)
158 		digit >>= 1;
159 
160 	return ((num_digits - 1) * 64 + i);
161 }
162 
163 /* Sets dest = src. */
164 static void vli_set(u64 *dest, const u64 *src, unsigned int ndigits)
165 {
166 	int i;
167 
168 	for (i = 0; i < ndigits; i++)
169 		dest[i] = src[i];
170 }
171 
172 /* Returns sign of left - right. */
173 static int vli_cmp(const u64 *left, const u64 *right, unsigned int ndigits)
174 {
175 	int i;
176 
177 	for (i = ndigits - 1; i >= 0; i--) {
178 		if (left[i] > right[i])
179 			return 1;
180 		else if (left[i] < right[i])
181 			return -1;
182 	}
183 
184 	return 0;
185 }
186 
187 /* Computes result = in << c, returning carry. Can modify in place
188  * (if result == in). 0 < shift < 64.
189  */
190 static u64 vli_lshift(u64 *result, const u64 *in, unsigned int shift,
191 		      unsigned int ndigits)
192 {
193 	u64 carry = 0;
194 	int i;
195 
196 	for (i = 0; i < ndigits; i++) {
197 		u64 temp = in[i];
198 
199 		result[i] = (temp << shift) | carry;
200 		carry = temp >> (64 - shift);
201 	}
202 
203 	return carry;
204 }
205 
206 /* Computes vli = vli >> 1. */
207 static void vli_rshift1(u64 *vli, unsigned int ndigits)
208 {
209 	u64 *end = vli;
210 	u64 carry = 0;
211 
212 	vli += ndigits;
213 
214 	while (vli-- > end) {
215 		u64 temp = *vli;
216 		*vli = (temp >> 1) | carry;
217 		carry = temp << 63;
218 	}
219 }
220 
221 /* Computes result = left + right, returning carry. Can modify in place. */
222 static u64 vli_add(u64 *result, const u64 *left, const u64 *right,
223 		   unsigned int ndigits)
224 {
225 	u64 carry = 0;
226 	int i;
227 
228 	for (i = 0; i < ndigits; i++) {
229 		u64 sum;
230 
231 		sum = left[i] + right[i] + carry;
232 		if (sum != left[i])
233 			carry = (sum < left[i]);
234 
235 		result[i] = sum;
236 	}
237 
238 	return carry;
239 }
240 
241 /* Computes result = left - right, returning borrow. Can modify in place. */
242 static u64 vli_sub(u64 *result, const u64 *left, const u64 *right,
243 		   unsigned int ndigits)
244 {
245 	u64 borrow = 0;
246 	int i;
247 
248 	for (i = 0; i < ndigits; i++) {
249 		u64 diff;
250 
251 		diff = left[i] - right[i] - borrow;
252 		if (diff != left[i])
253 			borrow = (diff > left[i]);
254 
255 		result[i] = diff;
256 	}
257 
258 	return borrow;
259 }
260 
261 static uint128_t mul_64_64(u64 left, u64 right)
262 {
263 	u64 a0 = left & 0xffffffffull;
264 	u64 a1 = left >> 32;
265 	u64 b0 = right & 0xffffffffull;
266 	u64 b1 = right >> 32;
267 	u64 m0 = a0 * b0;
268 	u64 m1 = a0 * b1;
269 	u64 m2 = a1 * b0;
270 	u64 m3 = a1 * b1;
271 	uint128_t result;
272 
273 	m2 += (m0 >> 32);
274 	m2 += m1;
275 
276 	/* Overflow */
277 	if (m2 < m1)
278 		m3 += 0x100000000ull;
279 
280 	result.m_low = (m0 & 0xffffffffull) | (m2 << 32);
281 	result.m_high = m3 + (m2 >> 32);
282 
283 	return result;
284 }
285 
286 static uint128_t add_128_128(uint128_t a, uint128_t b)
287 {
288 	uint128_t result;
289 
290 	result.m_low = a.m_low + b.m_low;
291 	result.m_high = a.m_high + b.m_high + (result.m_low < a.m_low);
292 
293 	return result;
294 }
295 
296 static void vli_mult(u64 *result, const u64 *left, const u64 *right,
297 		     unsigned int ndigits)
298 {
299 	uint128_t r01 = { 0, 0 };
300 	u64 r2 = 0;
301 	unsigned int i, k;
302 
303 	/* Compute each digit of result in sequence, maintaining the
304 	 * carries.
305 	 */
306 	for (k = 0; k < ndigits * 2 - 1; k++) {
307 		unsigned int min;
308 
309 		if (k < ndigits)
310 			min = 0;
311 		else
312 			min = (k + 1) - ndigits;
313 
314 		for (i = min; i <= k && i < ndigits; i++) {
315 			uint128_t product;
316 
317 			product = mul_64_64(left[i], right[k - i]);
318 
319 			r01 = add_128_128(r01, product);
320 			r2 += (r01.m_high < product.m_high);
321 		}
322 
323 		result[k] = r01.m_low;
324 		r01.m_low = r01.m_high;
325 		r01.m_high = r2;
326 		r2 = 0;
327 	}
328 
329 	result[ndigits * 2 - 1] = r01.m_low;
330 }
331 
332 static void vli_square(u64 *result, const u64 *left, unsigned int ndigits)
333 {
334 	uint128_t r01 = { 0, 0 };
335 	u64 r2 = 0;
336 	int i, k;
337 
338 	for (k = 0; k < ndigits * 2 - 1; k++) {
339 		unsigned int min;
340 
341 		if (k < ndigits)
342 			min = 0;
343 		else
344 			min = (k + 1) - ndigits;
345 
346 		for (i = min; i <= k && i <= k - i; i++) {
347 			uint128_t product;
348 
349 			product = mul_64_64(left[i], left[k - i]);
350 
351 			if (i < k - i) {
352 				r2 += product.m_high >> 63;
353 				product.m_high = (product.m_high << 1) |
354 						 (product.m_low >> 63);
355 				product.m_low <<= 1;
356 			}
357 
358 			r01 = add_128_128(r01, product);
359 			r2 += (r01.m_high < product.m_high);
360 		}
361 
362 		result[k] = r01.m_low;
363 		r01.m_low = r01.m_high;
364 		r01.m_high = r2;
365 		r2 = 0;
366 	}
367 
368 	result[ndigits * 2 - 1] = r01.m_low;
369 }
370 
371 /* Computes result = (left + right) % mod.
372  * Assumes that left < mod and right < mod, result != mod.
373  */
374 static void vli_mod_add(u64 *result, const u64 *left, const u64 *right,
375 			const u64 *mod, unsigned int ndigits)
376 {
377 	u64 carry;
378 
379 	carry = vli_add(result, left, right, ndigits);
380 
381 	/* result > mod (result = mod + remainder), so subtract mod to
382 	 * get remainder.
383 	 */
384 	if (carry || vli_cmp(result, mod, ndigits) >= 0)
385 		vli_sub(result, result, mod, ndigits);
386 }
387 
388 /* Computes result = (left - right) % mod.
389  * Assumes that left < mod and right < mod, result != mod.
390  */
391 static void vli_mod_sub(u64 *result, const u64 *left, const u64 *right,
392 			const u64 *mod, unsigned int ndigits)
393 {
394 	u64 borrow = vli_sub(result, left, right, ndigits);
395 
396 	/* In this case, p_result == -diff == (max int) - diff.
397 	 * Since -x % d == d - x, we can get the correct result from
398 	 * result + mod (with overflow).
399 	 */
400 	if (borrow)
401 		vli_add(result, result, mod, ndigits);
402 }
403 
404 /* Computes p_result = p_product % curve_p.
405  * See algorithm 5 and 6 from
406  * http://www.isys.uni-klu.ac.at/PDF/2001-0126-MT.pdf
407  */
408 static void vli_mmod_fast_192(u64 *result, const u64 *product,
409 			      const u64 *curve_prime, u64 *tmp)
410 {
411 	const unsigned int ndigits = 3;
412 	int carry;
413 
414 	vli_set(result, product, ndigits);
415 
416 	vli_set(tmp, &product[3], ndigits);
417 	carry = vli_add(result, result, tmp, ndigits);
418 
419 	tmp[0] = 0;
420 	tmp[1] = product[3];
421 	tmp[2] = product[4];
422 	carry += vli_add(result, result, tmp, ndigits);
423 
424 	tmp[0] = tmp[1] = product[5];
425 	tmp[2] = 0;
426 	carry += vli_add(result, result, tmp, ndigits);
427 
428 	while (carry || vli_cmp(curve_prime, result, ndigits) != 1)
429 		carry -= vli_sub(result, result, curve_prime, ndigits);
430 }
431 
432 /* Computes result = product % curve_prime
433  * from http://www.nsa.gov/ia/_files/nist-routines.pdf
434  */
435 static void vli_mmod_fast_256(u64 *result, const u64 *product,
436 			      const u64 *curve_prime, u64 *tmp)
437 {
438 	int carry;
439 	const unsigned int ndigits = 4;
440 
441 	/* t */
442 	vli_set(result, product, ndigits);
443 
444 	/* s1 */
445 	tmp[0] = 0;
446 	tmp[1] = product[5] & 0xffffffff00000000ull;
447 	tmp[2] = product[6];
448 	tmp[3] = product[7];
449 	carry = vli_lshift(tmp, tmp, 1, ndigits);
450 	carry += vli_add(result, result, tmp, ndigits);
451 
452 	/* s2 */
453 	tmp[1] = product[6] << 32;
454 	tmp[2] = (product[6] >> 32) | (product[7] << 32);
455 	tmp[3] = product[7] >> 32;
456 	carry += vli_lshift(tmp, tmp, 1, ndigits);
457 	carry += vli_add(result, result, tmp, ndigits);
458 
459 	/* s3 */
460 	tmp[0] = product[4];
461 	tmp[1] = product[5] & 0xffffffff;
462 	tmp[2] = 0;
463 	tmp[3] = product[7];
464 	carry += vli_add(result, result, tmp, ndigits);
465 
466 	/* s4 */
467 	tmp[0] = (product[4] >> 32) | (product[5] << 32);
468 	tmp[1] = (product[5] >> 32) | (product[6] & 0xffffffff00000000ull);
469 	tmp[2] = product[7];
470 	tmp[3] = (product[6] >> 32) | (product[4] << 32);
471 	carry += vli_add(result, result, tmp, ndigits);
472 
473 	/* d1 */
474 	tmp[0] = (product[5] >> 32) | (product[6] << 32);
475 	tmp[1] = (product[6] >> 32);
476 	tmp[2] = 0;
477 	tmp[3] = (product[4] & 0xffffffff) | (product[5] << 32);
478 	carry -= vli_sub(result, result, tmp, ndigits);
479 
480 	/* d2 */
481 	tmp[0] = product[6];
482 	tmp[1] = product[7];
483 	tmp[2] = 0;
484 	tmp[3] = (product[4] >> 32) | (product[5] & 0xffffffff00000000ull);
485 	carry -= vli_sub(result, result, tmp, ndigits);
486 
487 	/* d3 */
488 	tmp[0] = (product[6] >> 32) | (product[7] << 32);
489 	tmp[1] = (product[7] >> 32) | (product[4] << 32);
490 	tmp[2] = (product[4] >> 32) | (product[5] << 32);
491 	tmp[3] = (product[6] << 32);
492 	carry -= vli_sub(result, result, tmp, ndigits);
493 
494 	/* d4 */
495 	tmp[0] = product[7];
496 	tmp[1] = product[4] & 0xffffffff00000000ull;
497 	tmp[2] = product[5];
498 	tmp[3] = product[6] & 0xffffffff00000000ull;
499 	carry -= vli_sub(result, result, tmp, ndigits);
500 
501 	if (carry < 0) {
502 		do {
503 			carry += vli_add(result, result, curve_prime, ndigits);
504 		} while (carry < 0);
505 	} else {
506 		while (carry || vli_cmp(curve_prime, result, ndigits) != 1)
507 			carry -= vli_sub(result, result, curve_prime, ndigits);
508 	}
509 }
510 
511 /* Computes result = product % curve_prime
512  *  from http://www.nsa.gov/ia/_files/nist-routines.pdf
513 */
514 static bool vli_mmod_fast(u64 *result, u64 *product,
515 			  const u64 *curve_prime, unsigned int ndigits)
516 {
517 	u64 tmp[2 * ndigits];
518 
519 	switch (ndigits) {
520 	case 3:
521 		vli_mmod_fast_192(result, product, curve_prime, tmp);
522 		break;
523 	case 4:
524 		vli_mmod_fast_256(result, product, curve_prime, tmp);
525 		break;
526 	default:
527 		pr_err("unsupports digits size!\n");
528 		return false;
529 	}
530 
531 	return true;
532 }
533 
534 /* Computes result = (left * right) % curve_prime. */
535 static void vli_mod_mult_fast(u64 *result, const u64 *left, const u64 *right,
536 			      const u64 *curve_prime, unsigned int ndigits)
537 {
538 	u64 product[2 * ndigits];
539 
540 	vli_mult(product, left, right, ndigits);
541 	vli_mmod_fast(result, product, curve_prime, ndigits);
542 }
543 
544 /* Computes result = left^2 % curve_prime. */
545 static void vli_mod_square_fast(u64 *result, const u64 *left,
546 				const u64 *curve_prime, unsigned int ndigits)
547 {
548 	u64 product[2 * ndigits];
549 
550 	vli_square(product, left, ndigits);
551 	vli_mmod_fast(result, product, curve_prime, ndigits);
552 }
553 
554 #define EVEN(vli) (!(vli[0] & 1))
555 /* Computes result = (1 / p_input) % mod. All VLIs are the same size.
556  * See "From Euclid's GCD to Montgomery Multiplication to the Great Divide"
557  * https://labs.oracle.com/techrep/2001/smli_tr-2001-95.pdf
558  */
559 static void vli_mod_inv(u64 *result, const u64 *input, const u64 *mod,
560 			unsigned int ndigits)
561 {
562 	u64 a[ndigits], b[ndigits];
563 	u64 u[ndigits], v[ndigits];
564 	u64 carry;
565 	int cmp_result;
566 
567 	if (vli_is_zero(input, ndigits)) {
568 		vli_clear(result, ndigits);
569 		return;
570 	}
571 
572 	vli_set(a, input, ndigits);
573 	vli_set(b, mod, ndigits);
574 	vli_clear(u, ndigits);
575 	u[0] = 1;
576 	vli_clear(v, ndigits);
577 
578 	while ((cmp_result = vli_cmp(a, b, ndigits)) != 0) {
579 		carry = 0;
580 
581 		if (EVEN(a)) {
582 			vli_rshift1(a, ndigits);
583 
584 			if (!EVEN(u))
585 				carry = vli_add(u, u, mod, ndigits);
586 
587 			vli_rshift1(u, ndigits);
588 			if (carry)
589 				u[ndigits - 1] |= 0x8000000000000000ull;
590 		} else if (EVEN(b)) {
591 			vli_rshift1(b, ndigits);
592 
593 			if (!EVEN(v))
594 				carry = vli_add(v, v, mod, ndigits);
595 
596 			vli_rshift1(v, ndigits);
597 			if (carry)
598 				v[ndigits - 1] |= 0x8000000000000000ull;
599 		} else if (cmp_result > 0) {
600 			vli_sub(a, a, b, ndigits);
601 			vli_rshift1(a, ndigits);
602 
603 			if (vli_cmp(u, v, ndigits) < 0)
604 				vli_add(u, u, mod, ndigits);
605 
606 			vli_sub(u, u, v, ndigits);
607 			if (!EVEN(u))
608 				carry = vli_add(u, u, mod, ndigits);
609 
610 			vli_rshift1(u, ndigits);
611 			if (carry)
612 				u[ndigits - 1] |= 0x8000000000000000ull;
613 		} else {
614 			vli_sub(b, b, a, ndigits);
615 			vli_rshift1(b, ndigits);
616 
617 			if (vli_cmp(v, u, ndigits) < 0)
618 				vli_add(v, v, mod, ndigits);
619 
620 			vli_sub(v, v, u, ndigits);
621 			if (!EVEN(v))
622 				carry = vli_add(v, v, mod, ndigits);
623 
624 			vli_rshift1(v, ndigits);
625 			if (carry)
626 				v[ndigits - 1] |= 0x8000000000000000ull;
627 		}
628 	}
629 
630 	vli_set(result, u, ndigits);
631 }
632 
633 /* ------ Point operations ------ */
634 
635 /* Returns true if p_point is the point at infinity, false otherwise. */
636 static bool ecc_point_is_zero(const struct ecc_point *point)
637 {
638 	return (vli_is_zero(point->x, point->ndigits) &&
639 		vli_is_zero(point->y, point->ndigits));
640 }
641 
642 /* Point multiplication algorithm using Montgomery's ladder with co-Z
643  * coordinates. From http://eprint.iacr.org/2011/338.pdf
644  */
645 
646 /* Double in place */
647 static void ecc_point_double_jacobian(u64 *x1, u64 *y1, u64 *z1,
648 				      u64 *curve_prime, unsigned int ndigits)
649 {
650 	/* t1 = x, t2 = y, t3 = z */
651 	u64 t4[ndigits];
652 	u64 t5[ndigits];
653 
654 	if (vli_is_zero(z1, ndigits))
655 		return;
656 
657 	/* t4 = y1^2 */
658 	vli_mod_square_fast(t4, y1, curve_prime, ndigits);
659 	/* t5 = x1*y1^2 = A */
660 	vli_mod_mult_fast(t5, x1, t4, curve_prime, ndigits);
661 	/* t4 = y1^4 */
662 	vli_mod_square_fast(t4, t4, curve_prime, ndigits);
663 	/* t2 = y1*z1 = z3 */
664 	vli_mod_mult_fast(y1, y1, z1, curve_prime, ndigits);
665 	/* t3 = z1^2 */
666 	vli_mod_square_fast(z1, z1, curve_prime, ndigits);
667 
668 	/* t1 = x1 + z1^2 */
669 	vli_mod_add(x1, x1, z1, curve_prime, ndigits);
670 	/* t3 = 2*z1^2 */
671 	vli_mod_add(z1, z1, z1, curve_prime, ndigits);
672 	/* t3 = x1 - z1^2 */
673 	vli_mod_sub(z1, x1, z1, curve_prime, ndigits);
674 	/* t1 = x1^2 - z1^4 */
675 	vli_mod_mult_fast(x1, x1, z1, curve_prime, ndigits);
676 
677 	/* t3 = 2*(x1^2 - z1^4) */
678 	vli_mod_add(z1, x1, x1, curve_prime, ndigits);
679 	/* t1 = 3*(x1^2 - z1^4) */
680 	vli_mod_add(x1, x1, z1, curve_prime, ndigits);
681 	if (vli_test_bit(x1, 0)) {
682 		u64 carry = vli_add(x1, x1, curve_prime, ndigits);
683 
684 		vli_rshift1(x1, ndigits);
685 		x1[ndigits - 1] |= carry << 63;
686 	} else {
687 		vli_rshift1(x1, ndigits);
688 	}
689 	/* t1 = 3/2*(x1^2 - z1^4) = B */
690 
691 	/* t3 = B^2 */
692 	vli_mod_square_fast(z1, x1, curve_prime, ndigits);
693 	/* t3 = B^2 - A */
694 	vli_mod_sub(z1, z1, t5, curve_prime, ndigits);
695 	/* t3 = B^2 - 2A = x3 */
696 	vli_mod_sub(z1, z1, t5, curve_prime, ndigits);
697 	/* t5 = A - x3 */
698 	vli_mod_sub(t5, t5, z1, curve_prime, ndigits);
699 	/* t1 = B * (A - x3) */
700 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
701 	/* t4 = B * (A - x3) - y1^4 = y3 */
702 	vli_mod_sub(t4, x1, t4, curve_prime, ndigits);
703 
704 	vli_set(x1, z1, ndigits);
705 	vli_set(z1, y1, ndigits);
706 	vli_set(y1, t4, ndigits);
707 }
708 
709 /* Modify (x1, y1) => (x1 * z^2, y1 * z^3) */
710 static void apply_z(u64 *x1, u64 *y1, u64 *z, u64 *curve_prime,
711 		    unsigned int ndigits)
712 {
713 	u64 t1[ndigits];
714 
715 	vli_mod_square_fast(t1, z, curve_prime, ndigits);    /* z^2 */
716 	vli_mod_mult_fast(x1, x1, t1, curve_prime, ndigits); /* x1 * z^2 */
717 	vli_mod_mult_fast(t1, t1, z, curve_prime, ndigits);  /* z^3 */
718 	vli_mod_mult_fast(y1, y1, t1, curve_prime, ndigits); /* y1 * z^3 */
719 }
720 
721 /* P = (x1, y1) => 2P, (x2, y2) => P' */
722 static void xycz_initial_double(u64 *x1, u64 *y1, u64 *x2, u64 *y2,
723 				u64 *p_initial_z, u64 *curve_prime,
724 				unsigned int ndigits)
725 {
726 	u64 z[ndigits];
727 
728 	vli_set(x2, x1, ndigits);
729 	vli_set(y2, y1, ndigits);
730 
731 	vli_clear(z, ndigits);
732 	z[0] = 1;
733 
734 	if (p_initial_z)
735 		vli_set(z, p_initial_z, ndigits);
736 
737 	apply_z(x1, y1, z, curve_prime, ndigits);
738 
739 	ecc_point_double_jacobian(x1, y1, z, curve_prime, ndigits);
740 
741 	apply_z(x2, y2, z, curve_prime, ndigits);
742 }
743 
744 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
745  * Output P' = (x1', y1', Z3), P + Q = (x3, y3, Z3)
746  * or P => P', Q => P + Q
747  */
748 static void xycz_add(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime,
749 		     unsigned int ndigits)
750 {
751 	/* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
752 	u64 t5[ndigits];
753 
754 	/* t5 = x2 - x1 */
755 	vli_mod_sub(t5, x2, x1, curve_prime, ndigits);
756 	/* t5 = (x2 - x1)^2 = A */
757 	vli_mod_square_fast(t5, t5, curve_prime, ndigits);
758 	/* t1 = x1*A = B */
759 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
760 	/* t3 = x2*A = C */
761 	vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits);
762 	/* t4 = y2 - y1 */
763 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
764 	/* t5 = (y2 - y1)^2 = D */
765 	vli_mod_square_fast(t5, y2, curve_prime, ndigits);
766 
767 	/* t5 = D - B */
768 	vli_mod_sub(t5, t5, x1, curve_prime, ndigits);
769 	/* t5 = D - B - C = x3 */
770 	vli_mod_sub(t5, t5, x2, curve_prime, ndigits);
771 	/* t3 = C - B */
772 	vli_mod_sub(x2, x2, x1, curve_prime, ndigits);
773 	/* t2 = y1*(C - B) */
774 	vli_mod_mult_fast(y1, y1, x2, curve_prime, ndigits);
775 	/* t3 = B - x3 */
776 	vli_mod_sub(x2, x1, t5, curve_prime, ndigits);
777 	/* t4 = (y2 - y1)*(B - x3) */
778 	vli_mod_mult_fast(y2, y2, x2, curve_prime, ndigits);
779 	/* t4 = y3 */
780 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
781 
782 	vli_set(x2, t5, ndigits);
783 }
784 
785 /* Input P = (x1, y1, Z), Q = (x2, y2, Z)
786  * Output P + Q = (x3, y3, Z3), P - Q = (x3', y3', Z3)
787  * or P => P - Q, Q => P + Q
788  */
789 static void xycz_add_c(u64 *x1, u64 *y1, u64 *x2, u64 *y2, u64 *curve_prime,
790 		       unsigned int ndigits)
791 {
792 	/* t1 = X1, t2 = Y1, t3 = X2, t4 = Y2 */
793 	u64 t5[ndigits];
794 	u64 t6[ndigits];
795 	u64 t7[ndigits];
796 
797 	/* t5 = x2 - x1 */
798 	vli_mod_sub(t5, x2, x1, curve_prime, ndigits);
799 	/* t5 = (x2 - x1)^2 = A */
800 	vli_mod_square_fast(t5, t5, curve_prime, ndigits);
801 	/* t1 = x1*A = B */
802 	vli_mod_mult_fast(x1, x1, t5, curve_prime, ndigits);
803 	/* t3 = x2*A = C */
804 	vli_mod_mult_fast(x2, x2, t5, curve_prime, ndigits);
805 	/* t4 = y2 + y1 */
806 	vli_mod_add(t5, y2, y1, curve_prime, ndigits);
807 	/* t4 = y2 - y1 */
808 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
809 
810 	/* t6 = C - B */
811 	vli_mod_sub(t6, x2, x1, curve_prime, ndigits);
812 	/* t2 = y1 * (C - B) */
813 	vli_mod_mult_fast(y1, y1, t6, curve_prime, ndigits);
814 	/* t6 = B + C */
815 	vli_mod_add(t6, x1, x2, curve_prime, ndigits);
816 	/* t3 = (y2 - y1)^2 */
817 	vli_mod_square_fast(x2, y2, curve_prime, ndigits);
818 	/* t3 = x3 */
819 	vli_mod_sub(x2, x2, t6, curve_prime, ndigits);
820 
821 	/* t7 = B - x3 */
822 	vli_mod_sub(t7, x1, x2, curve_prime, ndigits);
823 	/* t4 = (y2 - y1)*(B - x3) */
824 	vli_mod_mult_fast(y2, y2, t7, curve_prime, ndigits);
825 	/* t4 = y3 */
826 	vli_mod_sub(y2, y2, y1, curve_prime, ndigits);
827 
828 	/* t7 = (y2 + y1)^2 = F */
829 	vli_mod_square_fast(t7, t5, curve_prime, ndigits);
830 	/* t7 = x3' */
831 	vli_mod_sub(t7, t7, t6, curve_prime, ndigits);
832 	/* t6 = x3' - B */
833 	vli_mod_sub(t6, t7, x1, curve_prime, ndigits);
834 	/* t6 = (y2 + y1)*(x3' - B) */
835 	vli_mod_mult_fast(t6, t6, t5, curve_prime, ndigits);
836 	/* t2 = y3' */
837 	vli_mod_sub(y1, t6, y1, curve_prime, ndigits);
838 
839 	vli_set(x1, t7, ndigits);
840 }
841 
842 static void ecc_point_mult(struct ecc_point *result,
843 			   const struct ecc_point *point, const u64 *scalar,
844 			   u64 *initial_z, u64 *curve_prime,
845 			   unsigned int ndigits)
846 {
847 	/* R0 and R1 */
848 	u64 rx[2][ndigits];
849 	u64 ry[2][ndigits];
850 	u64 z[ndigits];
851 	int i, nb;
852 	int num_bits = vli_num_bits(scalar, ndigits);
853 
854 	vli_set(rx[1], point->x, ndigits);
855 	vli_set(ry[1], point->y, ndigits);
856 
857 	xycz_initial_double(rx[1], ry[1], rx[0], ry[0], initial_z, curve_prime,
858 			    ndigits);
859 
860 	for (i = num_bits - 2; i > 0; i--) {
861 		nb = !vli_test_bit(scalar, i);
862 		xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime,
863 			   ndigits);
864 		xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime,
865 			 ndigits);
866 	}
867 
868 	nb = !vli_test_bit(scalar, 0);
869 	xycz_add_c(rx[1 - nb], ry[1 - nb], rx[nb], ry[nb], curve_prime,
870 		   ndigits);
871 
872 	/* Find final 1/Z value. */
873 	/* X1 - X0 */
874 	vli_mod_sub(z, rx[1], rx[0], curve_prime, ndigits);
875 	/* Yb * (X1 - X0) */
876 	vli_mod_mult_fast(z, z, ry[1 - nb], curve_prime, ndigits);
877 	/* xP * Yb * (X1 - X0) */
878 	vli_mod_mult_fast(z, z, point->x, curve_prime, ndigits);
879 
880 	/* 1 / (xP * Yb * (X1 - X0)) */
881 	vli_mod_inv(z, z, curve_prime, point->ndigits);
882 
883 	/* yP / (xP * Yb * (X1 - X0)) */
884 	vli_mod_mult_fast(z, z, point->y, curve_prime, ndigits);
885 	/* Xb * yP / (xP * Yb * (X1 - X0)) */
886 	vli_mod_mult_fast(z, z, rx[1 - nb], curve_prime, ndigits);
887 	/* End 1/Z calculation */
888 
889 	xycz_add(rx[nb], ry[nb], rx[1 - nb], ry[1 - nb], curve_prime, ndigits);
890 
891 	apply_z(rx[0], ry[0], z, curve_prime, ndigits);
892 
893 	vli_set(result->x, rx[0], ndigits);
894 	vli_set(result->y, ry[0], ndigits);
895 }
896 
897 static inline void ecc_swap_digits(const u64 *in, u64 *out,
898 				   unsigned int ndigits)
899 {
900 	int i;
901 
902 	for (i = 0; i < ndigits; i++)
903 		out[i] = __swab64(in[ndigits - 1 - i]);
904 }
905 
906 int ecc_is_key_valid(unsigned int curve_id, unsigned int ndigits,
907 		     const u8 *private_key, unsigned int private_key_len)
908 {
909 	int nbytes;
910 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
911 
912 	if (!private_key)
913 		return -EINVAL;
914 
915 	nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
916 
917 	if (private_key_len != nbytes)
918 		return -EINVAL;
919 
920 	if (vli_is_zero((const u64 *)&private_key[0], ndigits))
921 		return -EINVAL;
922 
923 	/* Make sure the private key is in the range [1, n-1]. */
924 	if (vli_cmp(curve->n, (const u64 *)&private_key[0], ndigits) != 1)
925 		return -EINVAL;
926 
927 	return 0;
928 }
929 
930 int ecdh_make_pub_key(unsigned int curve_id, unsigned int ndigits,
931 		      const u8 *private_key, unsigned int private_key_len,
932 		      u8 *public_key, unsigned int public_key_len)
933 {
934 	int ret = 0;
935 	struct ecc_point *pk;
936 	u64 priv[ndigits];
937 	unsigned int nbytes;
938 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
939 
940 	if (!private_key || !curve) {
941 		ret = -EINVAL;
942 		goto out;
943 	}
944 
945 	ecc_swap_digits((const u64 *)private_key, priv, ndigits);
946 
947 	pk = ecc_alloc_point(ndigits);
948 	if (!pk) {
949 		ret = -ENOMEM;
950 		goto out;
951 	}
952 
953 	ecc_point_mult(pk, &curve->g, priv, NULL, curve->p, ndigits);
954 	if (ecc_point_is_zero(pk)) {
955 		ret = -EAGAIN;
956 		goto err_free_point;
957 	}
958 
959 	nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
960 	ecc_swap_digits(pk->x, (u64 *)public_key, ndigits);
961 	ecc_swap_digits(pk->y, (u64 *)&public_key[nbytes], ndigits);
962 
963 err_free_point:
964 	ecc_free_point(pk);
965 out:
966 	return ret;
967 }
968 
969 int crypto_ecdh_shared_secret(unsigned int curve_id, unsigned int ndigits,
970 		       const u8 *private_key, unsigned int private_key_len,
971 		       const u8 *public_key, unsigned int public_key_len,
972 		       u8 *secret, unsigned int secret_len)
973 {
974 	int ret = 0;
975 	struct ecc_point *product, *pk;
976 	u64 priv[ndigits];
977 	u64 rand_z[ndigits];
978 	unsigned int nbytes;
979 	const struct ecc_curve *curve = ecc_get_curve(curve_id);
980 
981 	if (!private_key || !public_key || !curve) {
982 		ret = -EINVAL;
983 		goto out;
984 	}
985 
986 	nbytes = ndigits << ECC_DIGITS_TO_BYTES_SHIFT;
987 
988 	get_random_bytes(rand_z, nbytes);
989 
990 	pk = ecc_alloc_point(ndigits);
991 	if (!pk) {
992 		ret = -ENOMEM;
993 		goto out;
994 	}
995 
996 	product = ecc_alloc_point(ndigits);
997 	if (!product) {
998 		ret = -ENOMEM;
999 		goto err_alloc_product;
1000 	}
1001 
1002 	ecc_swap_digits((const u64 *)public_key, pk->x, ndigits);
1003 	ecc_swap_digits((const u64 *)&public_key[nbytes], pk->y, ndigits);
1004 	ecc_swap_digits((const u64 *)private_key, priv, ndigits);
1005 
1006 	ecc_point_mult(product, pk, priv, rand_z, curve->p, ndigits);
1007 
1008 	ecc_swap_digits(product->x, (u64 *)secret, ndigits);
1009 
1010 	if (ecc_point_is_zero(product))
1011 		ret = -EFAULT;
1012 
1013 	ecc_free_point(product);
1014 err_alloc_product:
1015 	ecc_free_point(pk);
1016 out:
1017 	return ret;
1018 }
1019