Added AES+GHASH implementation using AES-NI opcodes; also ARM-Thumb assembly for...
[BearSSL] / src / rsa / rsa_i15_priv.c
1 /*
2 * Copyright (c) 2017 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 (2 + ((BR_MAX_RSA_FACTOR + 14) / 15))
28 #define TLEN (8 * U)
29
30 /* obsolete
31 static void
32 print_int(const char *name, const uint16_t *x)
33 {
34 extern int printf(const char *fmt, ...);
35 unsigned char tmp[1000];
36 size_t u, len;
37
38 len = (x[0] - (x[0] >> 4) + 7) >> 3;
39 br_i15_encode(tmp, len, x);
40 printf("%s = ", name);
41 for (u = 0; u < len; u ++) {
42 printf("%02X", tmp[u]);
43 }
44 printf("\n");
45 }
46 */
47
48 /* see bearssl_rsa.h */
49 uint32_t
50 br_rsa_i15_private(unsigned char *x, const br_rsa_private_key *sk)
51 {
52 const unsigned char *p, *q;
53 size_t plen, qlen;
54 size_t fwlen;
55 uint16_t p0i, q0i;
56 size_t xlen;
57 uint16_t tmp[1 + TLEN];
58 long z;
59 uint16_t *mp, *mq, *s1, *s2, *t1, *t2, *t3;
60 uint32_t r;
61
62 /*
63 * Compute the actual lengths of p and q, in bytes.
64 * These lengths are not considered secret (we cannot really hide
65 * them anyway in constant-time code).
66 */
67 p = sk->p;
68 plen = sk->plen;
69 while (plen > 0 && *p == 0) {
70 p ++;
71 plen --;
72 }
73 q = sk->q;
74 qlen = sk->qlen;
75 while (qlen > 0 && *q == 0) {
76 q ++;
77 qlen --;
78 }
79
80 /*
81 * Compute the maximum factor length, in words.
82 */
83 z = (long)(plen > qlen ? plen : qlen) << 3;
84 fwlen = 1;
85 while (z > 0) {
86 z -= 15;
87 fwlen ++;
88 }
89 /*
90 * Round up the word length to an even number.
91 */
92 fwlen += (fwlen & 1);
93
94 /*
95 * We need to fit at least 6 values in the stack buffer.
96 */
97 if (6 * fwlen > TLEN) {
98 return 0;
99 }
100
101 /*
102 * Compute signature length (in bytes).
103 */
104 xlen = (sk->n_bitlen + 7) >> 3;
105
106 /*
107 * Ensure 32-bit alignment for value words.
108 */
109 mq = tmp;
110 if (((uintptr_t)mq & 2) == 0) {
111 mq ++;
112 }
113
114 /*
115 * Decode q.
116 */
117 br_i15_decode(mq, q, qlen);
118
119 /*
120 * Compute s2 = x^dq mod q.
121 */
122 q0i = br_i15_ninv15(mq[1]);
123 s2 = mq + fwlen;
124 br_i15_decode_reduce(s2, x, xlen, mq);
125 r = br_i15_modpow_opt(s2, sk->dq, sk->dqlen, mq, q0i,
126 mq + 2 * fwlen, TLEN - 2 * fwlen);
127
128 /*
129 * Decode p.
130 */
131 mp = mq + 2 * fwlen;
132 br_i15_decode(mp, p, plen);
133
134 /*
135 * Compute s1 = x^dq mod q.
136 */
137 p0i = br_i15_ninv15(mp[1]);
138 s1 = mq + 3 * fwlen;
139 br_i15_decode_reduce(s1, x, xlen, mp);
140 r &= br_i15_modpow_opt(s1, sk->dp, sk->dplen, mp, p0i,
141 mq + 4 * fwlen, TLEN - 4 * fwlen);
142
143 /*
144 * Compute:
145 * h = (s1 - s2)*(1/q) mod p
146 * s1 is an integer modulo p, but s2 is modulo q. PKCS#1 is
147 * unclear about whether p may be lower than q (some existing,
148 * widely deployed implementations of RSA don't tolerate p < q),
149 * but we want to support that occurrence, so we need to use the
150 * reduction function.
151 *
152 * Since we use br_i15_decode_reduce() for iq (purportedly, the
153 * inverse of q modulo p), we also tolerate improperly large
154 * values for this parameter.
155 */
156 t1 = mq + 4 * fwlen;
157 t2 = mq + 5 * fwlen;
158 br_i15_reduce(t2, s2, mp);
159 br_i15_add(s1, mp, br_i15_sub(s1, t2, 1));
160 br_i15_to_monty(s1, mp);
161 br_i15_decode_reduce(t1, sk->iq, sk->iqlen, mp);
162 br_i15_montymul(t2, s1, t1, mp, p0i);
163
164 /*
165 * h is now in t2. We compute the final result:
166 * s = s2 + q*h
167 * All these operations are non-modular.
168 *
169 * We need mq, s2 and t2. We use the t3 buffer as destination.
170 * The buffers mp, s1 and t1 are no longer needed, so we can
171 * reuse them for t3. Moreover, the first step of the computation
172 * is to copy s2 into t3, after which s2 is not needed. Right
173 * now, mq is in slot 0, s2 is in slot 1, and t2 in slot 5.
174 * Therefore, we have ample room for t3 by simply using s2.
175 */
176 t3 = s2;
177 br_i15_mulacc(t3, mq, t2);
178
179 /*
180 * Encode the result. Since we already checked the value of xlen,
181 * we can just use it right away.
182 */
183 br_i15_encode(x, xlen, t3);
184
185 /*
186 * The only error conditions remaining at that point are invalid
187 * values for p and q (even integers).
188 */
189 return p0i & q0i & r;
190 }