1/*	$NetBSD: aes_sse2_4x32.c,v 1.1 2025/11/23 22:48:26 riastradh Exp $	*/
2
3/*
4 * Copyright (c) 2016 Thomas Pornin <pornin@bolet.org>
5 *
6 * Permission is hereby granted, free of charge, to any person obtaining
7 * a copy of this software and associated documentation files (the
8 * "Software"), to deal in the Software without restriction, including
9 * without limitation the rights to use, copy, modify, merge, publish,
10 * distribute, sublicense, and/or sell copies of the Software, and to
11 * permit persons to whom the Software is furnished to do so, subject to
12 * the following conditions:
13 *
14 * The above copyright notice and this permission notice shall be
15 * included in all copies or substantial portions of the Software.
16 *
17 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
18 * EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
19 * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
20 * NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
21 * BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
22 * ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
23 * CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
24 * SOFTWARE.
25 */
26
27#include <sys/cdefs.h>
28__KERNEL_RCSID(1, "$NetBSD: aes_sse2_4x32.c,v 1.1 2025/11/23 22:48:26 riastradh Exp $");
29
30#include <sys/types.h>
31
32#ifdef _KERNEL
33#include <lib/libkern/libkern.h>
34#else
35#include <string.h>
36#endif
37
38#include "aes_sse2_4x32_impl.h"
39
40/* see inner.h */
41void
42aes_sse2_4x32_bitslice_Sbox(__m128i q[static 8])
43{
44	/*
45	 * This S-box implementation is a straightforward translation of
46	 * the circuit described by Boyar and Peralta in "A new
47	 * combinational logic minimization technique with applications
48	 * to cryptology" (https://eprint.iacr.org/2009/191.pdf).
49	 *
50	 * Note that variables x* (input) and s* (output) are numbered
51	 * in "reverse" order (x0 is the high bit, x7 is the low bit).
52	 */
53
54	__m128i x0, x1, x2, x3, x4, x5, x6, x7;
55	__m128i y1, y2, y3, y4, y5, y6, y7, y8, y9;
56	__m128i y10, y11, y12, y13, y14, y15, y16, y17, y18, y19;
57	__m128i y20, y21;
58	__m128i z0, z1, z2, z3, z4, z5, z6, z7, z8, z9;
59	__m128i z10, z11, z12, z13, z14, z15, z16, z17;
60	__m128i t0, t1, t2, t3, t4, t5, t6, t7, t8, t9;
61	__m128i t10, t11, t12, t13, t14, t15, t16, t17, t18, t19;
62	__m128i t20, t21, t22, t23, t24, t25, t26, t27, t28, t29;
63	__m128i t30, t31, t32, t33, t34, t35, t36, t37, t38, t39;
64	__m128i t40, t41, t42, t43, t44, t45, t46, t47, t48, t49;
65	__m128i t50, t51, t52, t53, t54, t55, t56, t57, t58, t59;
66	__m128i t60, t61, t62, t63, t64, t65, t66, t67;
67	__m128i s0, s1, s2, s3, s4, s5, s6, s7;
68
69	x0 = q[7];
70	x1 = q[6];
71	x2 = q[5];
72	x3 = q[4];
73	x4 = q[3];
74	x5 = q[2];
75	x6 = q[1];
76	x7 = q[0];
77
78	/*
79	 * Top linear transformation.
80	 */
81	y14 = x3 ^ x5;
82	y13 = x0 ^ x6;
83	y9 = x0 ^ x3;
84	y8 = x0 ^ x5;
85	t0 = x1 ^ x2;
86	y1 = t0 ^ x7;
87	y4 = y1 ^ x3;
88	y12 = y13 ^ y14;
89	y2 = y1 ^ x0;
90	y5 = y1 ^ x6;
91	y3 = y5 ^ y8;
92	t1 = x4 ^ y12;
93	y15 = t1 ^ x5;
94	y20 = t1 ^ x1;
95	y6 = y15 ^ x7;
96	y10 = y15 ^ t0;
97	y11 = y20 ^ y9;
98	y7 = x7 ^ y11;
99	y17 = y10 ^ y11;
100	y19 = y10 ^ y8;
101	y16 = t0 ^ y11;
102	y21 = y13 ^ y16;
103	y18 = x0 ^ y16;
104
105	/*
106	 * Non-linear section.
107	 */
108	t2 = y12 & y15;
109	t3 = y3 & y6;
110	t4 = t3 ^ t2;
111	t5 = y4 & x7;
112	t6 = t5 ^ t2;
113	t7 = y13 & y16;
114	t8 = y5 & y1;
115	t9 = t8 ^ t7;
116	t10 = y2 & y7;
117	t11 = t10 ^ t7;
118	t12 = y9 & y11;
119	t13 = y14 & y17;
120	t14 = t13 ^ t12;
121	t15 = y8 & y10;
122	t16 = t15 ^ t12;
123	t17 = t4 ^ t14;
124	t18 = t6 ^ t16;
125	t19 = t9 ^ t14;
126	t20 = t11 ^ t16;
127	t21 = t17 ^ y20;
128	t22 = t18 ^ y19;
129	t23 = t19 ^ y21;
130	t24 = t20 ^ y18;
131
132	t25 = t21 ^ t22;
133	t26 = t21 & t23;
134	t27 = t24 ^ t26;
135	t28 = t25 & t27;
136	t29 = t28 ^ t22;
137	t30 = t23 ^ t24;
138	t31 = t22 ^ t26;
139	t32 = t31 & t30;
140	t33 = t32 ^ t24;
141	t34 = t23 ^ t33;
142	t35 = t27 ^ t33;
143	t36 = t24 & t35;
144	t37 = t36 ^ t34;
145	t38 = t27 ^ t36;
146	t39 = t29 & t38;
147	t40 = t25 ^ t39;
148
149	t41 = t40 ^ t37;
150	t42 = t29 ^ t33;
151	t43 = t29 ^ t40;
152	t44 = t33 ^ t37;
153	t45 = t42 ^ t41;
154	z0 = t44 & y15;
155	z1 = t37 & y6;
156	z2 = t33 & x7;
157	z3 = t43 & y16;
158	z4 = t40 & y1;
159	z5 = t29 & y7;
160	z6 = t42 & y11;
161	z7 = t45 & y17;
162	z8 = t41 & y10;
163	z9 = t44 & y12;
164	z10 = t37 & y3;
165	z11 = t33 & y4;
166	z12 = t43 & y13;
167	z13 = t40 & y5;
168	z14 = t29 & y2;
169	z15 = t42 & y9;
170	z16 = t45 & y14;
171	z17 = t41 & y8;
172
173	/*
174	 * Bottom linear transformation.
175	 */
176	t46 = z15 ^ z16;
177	t47 = z10 ^ z11;
178	t48 = z5 ^ z13;
179	t49 = z9 ^ z10;
180	t50 = z2 ^ z12;
181	t51 = z2 ^ z5;
182	t52 = z7 ^ z8;
183	t53 = z0 ^ z3;
184	t54 = z6 ^ z7;
185	t55 = z16 ^ z17;
186	t56 = z12 ^ t48;
187	t57 = t50 ^ t53;
188	t58 = z4 ^ t46;
189	t59 = z3 ^ t54;
190	t60 = t46 ^ t57;
191	t61 = z14 ^ t57;
192	t62 = t52 ^ t58;
193	t63 = t49 ^ t58;
194	t64 = z4 ^ t59;
195	t65 = t61 ^ t62;
196	t66 = z1 ^ t63;
197	s0 = t59 ^ t63;
198	s6 = t56 ^ ~t62;
199	s7 = t48 ^ ~t60;
200	t67 = t64 ^ t65;
201	s3 = t53 ^ t66;
202	s4 = t51 ^ t66;
203	s5 = t47 ^ t65;
204	s1 = t64 ^ ~s3;
205	s2 = t55 ^ ~t67;
206
207	q[7] = s0;
208	q[6] = s1;
209	q[5] = s2;
210	q[4] = s3;
211	q[3] = s4;
212	q[2] = s5;
213	q[1] = s6;
214	q[0] = s7;
215}
216
217/* see inner.h */
218void
219aes_sse2_4x32_ortho(__m128i q[static 8])
220{
221#define SWAPN(cl, ch, s, x, y)   do { \
222		__m128i cl128 = _mm_set1_epi32(cl); \
223		__m128i ch128 = _mm_set1_epi32(ch); \
224		__m128i a, b; \
225		a = _mm_load_si128(&(x)); \
226		b = _mm_load_si128(&(y)); \
227		_mm_store_si128(&(x), \
228		    (a & cl128) | _mm_slli_epi32((b & cl128), (s))); \
229		_mm_store_si128(&(y), \
230		    _mm_srli_epi32((a & ch128), (s)) | (b & ch128)); \
231	} while (0)
232
233#define SWAP2(x, y)   SWAPN(0x55555555, 0xAAAAAAAA, 1, x, y)
234#define SWAP4(x, y)   SWAPN(0x33333333, 0xCCCCCCCC, 2, x, y)
235#define SWAP8(x, y)   SWAPN(0x0F0F0F0F, 0xF0F0F0F0, 4, x, y)
236
237	SWAP2(q[0], q[1]);
238	SWAP2(q[2], q[3]);
239	SWAP2(q[4], q[5]);
240	SWAP2(q[6], q[7]);
241
242	SWAP4(q[0], q[2]);
243	SWAP4(q[1], q[3]);
244	SWAP4(q[4], q[6]);
245	SWAP4(q[5], q[7]);
246
247	SWAP8(q[0], q[4]);
248	SWAP8(q[1], q[5]);
249	SWAP8(q[2], q[6]);
250	SWAP8(q[3], q[7]);
251}
252
253static const unsigned char Rcon[] = {
254	0x01, 0x02, 0x04, 0x08, 0x10, 0x20, 0x40, 0x80, 0x1B, 0x36
255};
256
257static uint32_t
258sub_word(uint32_t x)
259{
260	__m128i q[8];
261	uint32_t y;
262
263	memset(q, 0, sizeof(q));
264	q[0] = _mm_loadu_si32(&x);
265	aes_sse2_4x32_ortho(q);
266	aes_sse2_4x32_bitslice_Sbox(q);
267	aes_sse2_4x32_ortho(q);
268	_mm_storeu_si32(&y, q[0]);
269	return y;
270}
271
272/* see inner.h */
273unsigned
274aes_sse2_4x32_keysched(uint32_t comp_skey[static 60], const void *key,
275    size_t key_len)
276{
277	unsigned num_rounds;
278	int i, j, k, nk, nkf;
279	uint32_t tmp;
280	uint32_t skey[120];
281
282	switch (key_len) {
283	case 16:
284		num_rounds = 10;
285		break;
286	case 24:
287		num_rounds = 12;
288		break;
289	case 32:
290		num_rounds = 14;
291		break;
292	default:
293		/* abort(); */
294		return 0;
295	}
296	nk = (int)(key_len >> 2);
297	nkf = (int)((num_rounds + 1) << 2);
298	tmp = 0;
299	for (i = 0; i < nk; i ++) {
300		tmp = br_dec32le((const unsigned char *)key + (i << 2));
301		skey[(i << 1) + 0] = tmp;
302		skey[(i << 1) + 1] = tmp;
303	}
304	for (i = nk, j = 0, k = 0; i < nkf; i ++) {
305		if (j == 0) {
306			tmp = (tmp << 24) | (tmp >> 8);
307			tmp = sub_word(tmp) ^ Rcon[k];
308		} else if (nk > 6 && j == 4) {
309			tmp = sub_word(tmp);
310		}
311		tmp ^= skey[(i - nk) << 1];
312		skey[(i << 1) + 0] = tmp;
313		skey[(i << 1) + 1] = tmp;
314		if (++ j == nk) {
315			j = 0;
316			k ++;
317		}
318	}
319	for (i = 0; i < nkf; i += 4) {
320		__m128i q[8];
321
322		for (j = 0; j < 8; j++)
323			q[j] = _mm_loadu_si32(&skey[(i << 1) + j]);
324		aes_sse2_4x32_ortho(q);
325		for (j = 0; j < 8; j++)
326			_mm_storeu_si32(&skey[(i << 1) + j], q[j]);
327	}
328	for (i = 0, j = 0; i < nkf; i ++, j += 2) {
329		comp_skey[i] = (skey[j + 0] & 0x55555555)
330			| (skey[j + 1] & 0xAAAAAAAA);
331	}
332	return num_rounds;
333}
334
335/* see inner.h */
336void
337aes_sse2_4x32_skey_expand(uint32_t skey[static 120],
338	unsigned num_rounds, const uint32_t comp_skey[static 60])
339{
340	unsigned u, v, n;
341
342	n = (num_rounds + 1) << 2;
343	for (u = 0, v = 0; u < n; u ++, v += 2) {
344		uint32_t x, y;
345
346		x = y = comp_skey[u];
347		x &= 0x55555555;
348		skey[v + 0] = x | (x << 1);
349		y &= 0xAAAAAAAA;
350		skey[v + 1] = y | (y >> 1);
351	}
352}
353