1/*
2 * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org>
3 *
4 * Permission is hereby granted, free of charge, to any person obtaining
5 * a copy of this software and associated documentation files (the
6 * "Software"), to deal in the Software without restriction, including
7 * without limitation the rights to use, copy, modify, merge, publish,
8 * distribute, sublicense, and/or sell copies of the Software, and to
9 * permit persons to whom the Software is furnished to do so, subject to
10 * the following conditions:
11 *
12 * The above copyright notice and this permission notice shall be
13 * included in all copies or substantial portions of the Software.
14 *
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
16 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
17 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
18 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
19 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
20 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
21 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
22 * SOFTWARE.
23 */
24
25#include "inner.h"
26
27#define U   (1 + (BR_MAX_RSA_FACTOR >> 5))
28
29/* see bearssl_rsa.h */
30uint32_t
31br_rsa_i32_private(unsigned char *x, const br_rsa_private_key *sk)
32{
33	const unsigned char *p, *q;
34	size_t plen, qlen;
35	uint32_t tmp[6 * U];
36	uint32_t *mp, *mq, *s1, *s2, *t1, *t2, *t3;
37	uint32_t p0i, q0i;
38	size_t xlen, u;
39	uint32_t r;
40
41	/*
42	 * All our temporary buffers are from the tmp[] array.
43	 *
44	 * The mp, mq, s1, s2, t1 and t2 buffers are large enough to
45	 * contain a RSA factor. The t3 buffer can contain a complete
46	 * RSA modulus. t3 shares its storage space with s2, s1 and t1,
47	 * in that order (this is important, see below).
48	 */
49	mq = tmp;
50	mp = tmp + U;
51	t2 = tmp + 2 * U;
52	s2 = tmp + 3 * U;
53	s1 = tmp + 4 * U;
54	t1 = tmp + 5 * U;
55	t3 = s2;
56
57	/*
58	 * Compute the actual lengths (in bytes) of p and q, and check
59	 * that they fit within our stack buffers.
60	 */
61	p = sk->p;
62	plen = sk->plen;
63	while (plen > 0 && *p == 0) {
64		p ++;
65		plen --;
66	}
67	q = sk->q;
68	qlen = sk->qlen;
69	while (qlen > 0 && *q == 0) {
70		q ++;
71		qlen --;
72	}
73	if (plen > (BR_MAX_RSA_FACTOR >> 3)
74		|| qlen > (BR_MAX_RSA_FACTOR >> 3))
75	{
76		return 0;
77	}
78
79	/*
80	 * Decode p and q.
81	 */
82	br_i32_decode(mp, p, plen);
83	br_i32_decode(mq, q, qlen);
84
85	/*
86	 * Recompute modulus, to compare with the source value.
87	 */
88	br_i32_zero(t2, mp[0]);
89	br_i32_mulacc(t2, mp, mq);
90	xlen = (sk->n_bitlen + 7) >> 3;
91	br_i32_encode(t2 + 2 * U, xlen, t2);
92	u = xlen;
93	r = 0;
94	while (u > 0) {
95		uint32_t wn, wx;
96
97		u --;
98		wn = ((unsigned char *)(t2 + 2 * U))[u];
99		wx = x[u];
100		r = ((wx - (wn + r)) >> 8) & 1;
101	}
102
103	/*
104	 * Compute s1 = x^dp mod p.
105	 */
106	p0i = br_i32_ninv32(mp[1]);
107	br_i32_decode_reduce(s1, x, xlen, mp);
108	br_i32_modpow(s1, sk->dp, sk->dplen, mp, p0i, t1, t2);
109
110	/*
111	 * Compute s2 = x^dq mod q.
112	 */
113	q0i = br_i32_ninv32(mq[1]);
114	br_i32_decode_reduce(s2, x, xlen, mq);
115	br_i32_modpow(s2, sk->dq, sk->dqlen, mq, q0i, t1, t2);
116
117	/*
118	 * Compute:
119	 *   h = (s1 - s2)*(1/q) mod p
120	 * s1 is an integer modulo p, but s2 is modulo q. PKCS#1 is
121	 * unclear about whether p may be lower than q (some existing,
122	 * widely deployed implementations of RSA don't tolerate p < q),
123	 * but we want to support that occurrence, so we need to use the
124	 * reduction function.
125	 *
126	 * Since we use br_i32_decode_reduce() for iq (purportedly, the
127	 * inverse of q modulo p), we also tolerate improperly large
128	 * values for this parameter.
129	 */
130	br_i32_reduce(t2, s2, mp);
131	br_i32_add(s1, mp, br_i32_sub(s1, t2, 1));
132	br_i32_to_monty(s1, mp);
133	br_i32_decode_reduce(t1, sk->iq, sk->iqlen, mp);
134	br_i32_montymul(t2, s1, t1, mp, p0i);
135
136	/*
137	 * h is now in t2. We compute the final result:
138	 *   s = s2 + q*h
139	 * All these operations are non-modular.
140	 *
141	 * We need mq, s2 and t2. We use the t3 buffer as destination.
142	 * The buffers mp, s1 and t1 are no longer needed. Moreover,
143	 * the first step is to copy s2 into the destination buffer t3.
144	 * We thus arranged for t3 to actually share space with s2, and
145	 * to be followed by the space formerly used by s1 and t1.
146	 */
147	br_i32_mulacc(t3, mq, t2);
148
149	/*
150	 * Encode the result. Since we already checked the value of xlen,
151	 * we can just use it right away.
152	 */
153	br_i32_encode(x, xlen, t3);
154
155	/*
156	 * The only error conditions remaining at that point are invalid
157	 * values for p and q (even integers).
158	 */
159	return p0i & q0i & r;
160}
161