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/* see inner.h */
28void
29br_aes_ct_bitslice_invSbox(uint32_t *q)
30{
31	/*
32	 * AES S-box is:
33	 *   S(x) = A(I(x)) ^ 0x63
34	 * where I() is inversion in GF(256), and A() is a linear
35	 * transform (0 is formally defined to be its own inverse).
36	 * Since inversion is an involution, the inverse S-box can be
37	 * computed from the S-box as:
38	 *   iS(x) = B(S(B(x ^ 0x63)) ^ 0x63)
39	 * where B() is the inverse of A(). Indeed, for any y in GF(256):
40	 *   iS(S(y)) = B(A(I(B(A(I(y)) ^ 0x63 ^ 0x63))) ^ 0x63 ^ 0x63) = y
41	 *
42	 * Note: we reuse the implementation of the forward S-box,
43	 * instead of duplicating it here, so that total code size is
44	 * lower. By merging the B() transforms into the S-box circuit
45	 * we could make faster CBC decryption, but CBC decryption is
46	 * already quite faster than CBC encryption because we can
47	 * process two blocks in parallel.
48	 */
49	uint32_t q0, q1, q2, q3, q4, q5, q6, q7;
50
51	q0 = ~q[0];
52	q1 = ~q[1];
53	q2 = q[2];
54	q3 = q[3];
55	q4 = q[4];
56	q5 = ~q[5];
57	q6 = ~q[6];
58	q7 = q[7];
59	q[7] = q1 ^ q4 ^ q6;
60	q[6] = q0 ^ q3 ^ q5;
61	q[5] = q7 ^ q2 ^ q4;
62	q[4] = q6 ^ q1 ^ q3;
63	q[3] = q5 ^ q0 ^ q2;
64	q[2] = q4 ^ q7 ^ q1;
65	q[1] = q3 ^ q6 ^ q0;
66	q[0] = q2 ^ q5 ^ q7;
67
68	br_aes_ct_bitslice_Sbox(q);
69
70	q0 = ~q[0];
71	q1 = ~q[1];
72	q2 = q[2];
73	q3 = q[3];
74	q4 = q[4];
75	q5 = ~q[5];
76	q6 = ~q[6];
77	q7 = q[7];
78	q[7] = q1 ^ q4 ^ q6;
79	q[6] = q0 ^ q3 ^ q5;
80	q[5] = q7 ^ q2 ^ q4;
81	q[4] = q6 ^ q1 ^ q3;
82	q[3] = q5 ^ q0 ^ q2;
83	q[2] = q4 ^ q7 ^ q1;
84	q[1] = q3 ^ q6 ^ q0;
85	q[0] = q2 ^ q5 ^ q7;
86}
87
88static void
89add_round_key(uint32_t *q, const uint32_t *sk)
90{
91	int i;
92
93	for (i = 0; i < 8; i ++) {
94		q[i] ^= sk[i];
95	}
96}
97
98static void
99inv_shift_rows(uint32_t *q)
100{
101	int i;
102
103	for (i = 0; i < 8; i ++) {
104		uint32_t x;
105
106		x = q[i];
107		q[i] = (x & 0x000000FF)
108			| ((x & 0x00003F00) << 2) | ((x & 0x0000C000) >> 6)
109			| ((x & 0x000F0000) << 4) | ((x & 0x00F00000) >> 4)
110			| ((x & 0x03000000) << 6) | ((x & 0xFC000000) >> 2);
111	}
112}
113
114static inline uint32_t
115rotr16(uint32_t x)
116{
117	return (x << 16) | (x >> 16);
118}
119
120static void
121inv_mix_columns(uint32_t *q)
122{
123	uint32_t q0, q1, q2, q3, q4, q5, q6, q7;
124	uint32_t r0, r1, r2, r3, r4, r5, r6, r7;
125
126	q0 = q[0];
127	q1 = q[1];
128	q2 = q[2];
129	q3 = q[3];
130	q4 = q[4];
131	q5 = q[5];
132	q6 = q[6];
133	q7 = q[7];
134	r0 = (q0 >> 8) | (q0 << 24);
135	r1 = (q1 >> 8) | (q1 << 24);
136	r2 = (q2 >> 8) | (q2 << 24);
137	r3 = (q3 >> 8) | (q3 << 24);
138	r4 = (q4 >> 8) | (q4 << 24);
139	r5 = (q5 >> 8) | (q5 << 24);
140	r6 = (q6 >> 8) | (q6 << 24);
141	r7 = (q7 >> 8) | (q7 << 24);
142
143	q[0] = q5 ^ q6 ^ q7 ^ r0 ^ r5 ^ r7 ^ rotr16(q0 ^ q5 ^ q6 ^ r0 ^ r5);
144	q[1] = q0 ^ q5 ^ r0 ^ r1 ^ r5 ^ r6 ^ r7 ^ rotr16(q1 ^ q5 ^ q7 ^ r1 ^ r5 ^ r6);
145	q[2] = q0 ^ q1 ^ q6 ^ r1 ^ r2 ^ r6 ^ r7 ^ rotr16(q0 ^ q2 ^ q6 ^ r2 ^ r6 ^ r7);
146	q[3] = q0 ^ q1 ^ q2 ^ q5 ^ q6 ^ r0 ^ r2 ^ r3 ^ r5 ^ rotr16(q0 ^ q1 ^ q3 ^ q5 ^ q6 ^ q7 ^ r0 ^ r3 ^ r5 ^ r7);
147	q[4] = q1 ^ q2 ^ q3 ^ q5 ^ r1 ^ r3 ^ r4 ^ r5 ^ r6 ^ r7 ^ rotr16(q1 ^ q2 ^ q4 ^ q5 ^ q7 ^ r1 ^ r4 ^ r5 ^ r6);
148	q[5] = q2 ^ q3 ^ q4 ^ q6 ^ r2 ^ r4 ^ r5 ^ r6 ^ r7 ^ rotr16(q2 ^ q3 ^ q5 ^ q6 ^ r2 ^ r5 ^ r6 ^ r7);
149	q[6] = q3 ^ q4 ^ q5 ^ q7 ^ r3 ^ r5 ^ r6 ^ r7 ^ rotr16(q3 ^ q4 ^ q6 ^ q7 ^ r3 ^ r6 ^ r7);
150	q[7] = q4 ^ q5 ^ q6 ^ r4 ^ r6 ^ r7 ^ rotr16(q4 ^ q5 ^ q7 ^ r4 ^ r7);
151}
152
153/* see inner.h */
154void
155br_aes_ct_bitslice_decrypt(unsigned num_rounds,
156	const uint32_t *skey, uint32_t *q)
157{
158	unsigned u;
159
160	add_round_key(q, skey + (num_rounds << 3));
161	for (u = num_rounds - 1; u > 0; u --) {
162		inv_shift_rows(q);
163		br_aes_ct_bitslice_invSbox(q);
164		add_round_key(q, skey + (u << 3));
165		inv_mix_columns(q);
166	}
167	inv_shift_rows(q);
168	br_aes_ct_bitslice_invSbox(q);
169	add_round_key(q, skey);
170}
171