1/* $NetBSD: skipjack.c,v 1.4 2014/01/01 15:18:57 pgoyette Exp $ */
2/* $OpenBSD: skipjack.c,v 1.3 2001/05/05 00:31:34 angelos Exp $ */
3
4/*
5 * Further optimized test implementation of SKIPJACK algorithm
6 * Mark Tillotson <markt@chaos.org.uk>, 25 June 98
7 * Optimizations suit RISC (lots of registers) machine best.
8 *
9 * based on unoptimized implementation of
10 * Panu Rissanen <bande@lut.fi> 960624
11 *
12 * SKIPJACK and KEA Algorithm Specifications
13 * Version 2.0
14 * 29 May 1998
15*/
16
17#include <sys/cdefs.h>
18__KERNEL_RCSID(0, "$NetBSD: skipjack.c,v 1.4 2014/01/01 15:18:57 pgoyette Exp $");
19
20#include <sys/errno.h>
21#include <sys/malloc.h>
22#include <sys/module.h>
23#include <sys/param.h>
24
25#include <crypto/skipjack/skipjack.h>
26#include <opencrypto/cryptodev.h>
27
28static const u_int8_t ftable[0x100] =
29{
30 0xa3, 0xd7, 0x09, 0x83, 0xf8, 0x48, 0xf6, 0xf4,
31 0xb3, 0x21, 0x15, 0x78, 0x99, 0xb1, 0xaf, 0xf9,
32 0xe7, 0x2d, 0x4d, 0x8a, 0xce, 0x4c, 0xca, 0x2e,
33 0x52, 0x95, 0xd9, 0x1e, 0x4e, 0x38, 0x44, 0x28,
34 0x0a, 0xdf, 0x02, 0xa0, 0x17, 0xf1, 0x60, 0x68,
35 0x12, 0xb7, 0x7a, 0xc3, 0xe9, 0xfa, 0x3d, 0x53,
36 0x96, 0x84, 0x6b, 0xba, 0xf2, 0x63, 0x9a, 0x19,
37 0x7c, 0xae, 0xe5, 0xf5, 0xf7, 0x16, 0x6a, 0xa2,
38 0x39, 0xb6, 0x7b, 0x0f, 0xc1, 0x93, 0x81, 0x1b,
39 0xee, 0xb4, 0x1a, 0xea, 0xd0, 0x91, 0x2f, 0xb8,
40 0x55, 0xb9, 0xda, 0x85, 0x3f, 0x41, 0xbf, 0xe0,
41 0x5a, 0x58, 0x80, 0x5f, 0x66, 0x0b, 0xd8, 0x90,
42 0x35, 0xd5, 0xc0, 0xa7, 0x33, 0x06, 0x65, 0x69,
43 0x45, 0x00, 0x94, 0x56, 0x6d, 0x98, 0x9b, 0x76,
44 0x97, 0xfc, 0xb2, 0xc2, 0xb0, 0xfe, 0xdb, 0x20,
45 0xe1, 0xeb, 0xd6, 0xe4, 0xdd, 0x47, 0x4a, 0x1d,
46 0x42, 0xed, 0x9e, 0x6e, 0x49, 0x3c, 0xcd, 0x43,
47 0x27, 0xd2, 0x07, 0xd4, 0xde, 0xc7, 0x67, 0x18,
48 0x89, 0xcb, 0x30, 0x1f, 0x8d, 0xc6, 0x8f, 0xaa,
49 0xc8, 0x74, 0xdc, 0xc9, 0x5d, 0x5c, 0x31, 0xa4,
50 0x70, 0x88, 0x61, 0x2c, 0x9f, 0x0d, 0x2b, 0x87,
51 0x50, 0x82, 0x54, 0x64, 0x26, 0x7d, 0x03, 0x40,
52 0x34, 0x4b, 0x1c, 0x73, 0xd1, 0xc4, 0xfd, 0x3b,
53 0xcc, 0xfb, 0x7f, 0xab, 0xe6, 0x3e, 0x5b, 0xa5,
54 0xad, 0x04, 0x23, 0x9c, 0x14, 0x51, 0x22, 0xf0,
55 0x29, 0x79, 0x71, 0x7e, 0xff, 0x8c, 0x0e, 0xe2,
56 0x0c, 0xef, 0xbc, 0x72, 0x75, 0x6f, 0x37, 0xa1,
57 0xec, 0xd3, 0x8e, 0x62, 0x8b, 0x86, 0x10, 0xe8,
58 0x08, 0x77, 0x11, 0xbe, 0x92, 0x4f, 0x24, 0xc5,
59 0x32, 0x36, 0x9d, 0xcf, 0xf3, 0xa6, 0xbb, 0xac,
60 0x5e, 0x6c, 0xa9, 0x13, 0x57, 0x25, 0xb5, 0xe3,
61 0xbd, 0xa8, 0x3a, 0x01, 0x05, 0x59, 0x2a, 0x46
62};
63
64/*
65 * For each key byte generate a table to represent the function
66 * ftable [in ^ keybyte]
67 *
68 * These tables used to save an XOR in each stage of the G-function
69 * the tables are hopefully pointed to by register allocated variables
70 * k0, k1..k9
71 */
72void
73subkey_table_gen (const u_int8_t *key, u_int8_t **key_tables)
74{
75 int i, k;
76
77 for (k = 0; k < 10; k++) {
78 u_int8_t key_byte = key [k];
79 u_int8_t * table = key_tables[k];
80 for (i = 0; i < 0x100; i++)
81 table [i] = ftable [i ^ key_byte];
82 }
83}
84
85
86#define g(k0, k1, k2, k3, ih, il, oh, ol) \
87{ \
88 oh = k##k0 [il] ^ ih; \
89 ol = k##k1 [oh] ^ il; \
90 oh = k##k2 [ol] ^ oh; \
91 ol = k##k3 [oh] ^ ol; \
92}
93
94#define g0(ih, il, oh, ol) g(0, 1, 2, 3, ih, il, oh, ol)
95#define g4(ih, il, oh, ol) g(4, 5, 6, 7, ih, il, oh, ol)
96#define g8(ih, il, oh, ol) g(8, 9, 0, 1, ih, il, oh, ol)
97#define g2(ih, il, oh, ol) g(2, 3, 4, 5, ih, il, oh, ol)
98#define g6(ih, il, oh, ol) g(6, 7, 8, 9, ih, il, oh, ol)
99
100
101#define g_inv(k0, k1, k2, k3, ih, il, oh, ol) \
102{ \
103 ol = k##k3 [ih] ^ il; \
104 oh = k##k2 [ol] ^ ih; \
105 ol = k##k1 [oh] ^ ol; \
106 oh = k##k0 [ol] ^ oh; \
107}
108
109
110#define g0_inv(ih, il, oh, ol) g_inv(0, 1, 2, 3, ih, il, oh, ol)
111#define g4_inv(ih, il, oh, ol) g_inv(4, 5, 6, 7, ih, il, oh, ol)
112#define g8_inv(ih, il, oh, ol) g_inv(8, 9, 0, 1, ih, il, oh, ol)
113#define g2_inv(ih, il, oh, ol) g_inv(2, 3, 4, 5, ih, il, oh, ol)
114#define g6_inv(ih, il, oh, ol) g_inv(6, 7, 8, 9, ih, il, oh, ol)
115
116/* optimized version of Skipjack algorithm
117 *
118 * the appropriate g-function is inlined for each round
119 *
120 * the data movement is minimized by rotating the names of the
121 * variables w1..w4, not their contents (saves 3 moves per round)
122 *
123 * the loops are completely unrolled (needed to staticize choice of g)
124 *
125 * compiles to about 470 instructions on a Sparc (gcc -O)
126 * which is about 58 instructions per byte, 14 per round.
127 * gcc seems to leave in some unnecessary and with 0xFF operations
128 * but only in the latter part of the functions. Perhaps it
129 * runs out of resources to properly optimize long inlined function?
130 * in theory should get about 11 instructions per round, not 14
131 */
132
133void
134skipjack_forwards(u_int8_t *plain, u_int8_t *cipher, u_int8_t **key_tables)
135{
136 u_int8_t wh1 = plain[0]; u_int8_t wl1 = plain[1];
137 u_int8_t wh2 = plain[2]; u_int8_t wl2 = plain[3];
138 u_int8_t wh3 = plain[4]; u_int8_t wl3 = plain[5];
139 u_int8_t wh4 = plain[6]; u_int8_t wl4 = plain[7];
140
141 u_int8_t * k0 = key_tables [0];
142 u_int8_t * k1 = key_tables [1];
143 u_int8_t * k2 = key_tables [2];
144 u_int8_t * k3 = key_tables [3];
145 u_int8_t * k4 = key_tables [4];
146 u_int8_t * k5 = key_tables [5];
147 u_int8_t * k6 = key_tables [6];
148 u_int8_t * k7 = key_tables [7];
149 u_int8_t * k8 = key_tables [8];
150 u_int8_t * k9 = key_tables [9];
151
152 /* first 8 rounds */
153 g0 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 1; wh4 ^= wh1;
154 g4 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 2; wh3 ^= wh4;
155 g8 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 3; wh2 ^= wh3;
156 g2 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 4; wh1 ^= wh2;
157 g6 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 5; wh4 ^= wh1;
158 g0 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 6; wh3 ^= wh4;
159 g4 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 7; wh2 ^= wh3;
160 g8 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 8; wh1 ^= wh2;
161
162 /* second 8 rounds */
163 wh2 ^= wh1; wl2 ^= wl1 ^ 9 ; g2 (wh1,wl1, wh1,wl1);
164 wh1 ^= wh4; wl1 ^= wl4 ^ 10; g6 (wh4,wl4, wh4,wl4);
165 wh4 ^= wh3; wl4 ^= wl3 ^ 11; g0 (wh3,wl3, wh3,wl3);
166 wh3 ^= wh2; wl3 ^= wl2 ^ 12; g4 (wh2,wl2, wh2,wl2);
167 wh2 ^= wh1; wl2 ^= wl1 ^ 13; g8 (wh1,wl1, wh1,wl1);
168 wh1 ^= wh4; wl1 ^= wl4 ^ 14; g2 (wh4,wl4, wh4,wl4);
169 wh4 ^= wh3; wl4 ^= wl3 ^ 15; g6 (wh3,wl3, wh3,wl3);
170 wh3 ^= wh2; wl3 ^= wl2 ^ 16; g0 (wh2,wl2, wh2,wl2);
171
172 /* third 8 rounds */
173 g4 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 17; wh4 ^= wh1;
174 g8 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 18; wh3 ^= wh4;
175 g2 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 19; wh2 ^= wh3;
176 g6 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 20; wh1 ^= wh2;
177 g0 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 21; wh4 ^= wh1;
178 g4 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 22; wh3 ^= wh4;
179 g8 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 23; wh2 ^= wh3;
180 g2 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 24; wh1 ^= wh2;
181
182 /* last 8 rounds */
183 wh2 ^= wh1; wl2 ^= wl1 ^ 25; g6 (wh1,wl1, wh1,wl1);
184 wh1 ^= wh4; wl1 ^= wl4 ^ 26; g0 (wh4,wl4, wh4,wl4);
185 wh4 ^= wh3; wl4 ^= wl3 ^ 27; g4 (wh3,wl3, wh3,wl3);
186 wh3 ^= wh2; wl3 ^= wl2 ^ 28; g8 (wh2,wl2, wh2,wl2);
187 wh2 ^= wh1; wl2 ^= wl1 ^ 29; g2 (wh1,wl1, wh1,wl1);
188 wh1 ^= wh4; wl1 ^= wl4 ^ 30; g6 (wh4,wl4, wh4,wl4);
189 wh4 ^= wh3; wl4 ^= wl3 ^ 31; g0 (wh3,wl3, wh3,wl3);
190 wh3 ^= wh2; wl3 ^= wl2 ^ 32; g4 (wh2,wl2, wh2,wl2);
191
192 /* pack into byte vector */
193 cipher [0] = wh1; cipher [1] = wl1;
194 cipher [2] = wh2; cipher [3] = wl2;
195 cipher [4] = wh3; cipher [5] = wl3;
196 cipher [6] = wh4; cipher [7] = wl4;
197}
198
199
200void
201skipjack_backwards (u_int8_t *cipher, u_int8_t *plain, u_int8_t **key_tables)
202{
203 /* setup 4 16-bit portions */
204 u_int8_t wh1 = cipher[0]; u_int8_t wl1 = cipher[1];
205 u_int8_t wh2 = cipher[2]; u_int8_t wl2 = cipher[3];
206 u_int8_t wh3 = cipher[4]; u_int8_t wl3 = cipher[5];
207 u_int8_t wh4 = cipher[6]; u_int8_t wl4 = cipher[7];
208
209 u_int8_t * k0 = key_tables [0];
210 u_int8_t * k1 = key_tables [1];
211 u_int8_t * k2 = key_tables [2];
212 u_int8_t * k3 = key_tables [3];
213 u_int8_t * k4 = key_tables [4];
214 u_int8_t * k5 = key_tables [5];
215 u_int8_t * k6 = key_tables [6];
216 u_int8_t * k7 = key_tables [7];
217 u_int8_t * k8 = key_tables [8];
218 u_int8_t * k9 = key_tables [9];
219
220 /* first 8 rounds */
221 g4_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 32; wh3 ^= wh2;
222 g0_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 31; wh4 ^= wh3;
223 g6_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 30; wh1 ^= wh4;
224 g2_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 29; wh2 ^= wh1;
225 g8_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 28; wh3 ^= wh2;
226 g4_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 27; wh4 ^= wh3;
227 g0_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 26; wh1 ^= wh4;
228 g6_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 25; wh2 ^= wh1;
229
230 /* second 8 rounds */
231 wh1 ^= wh2; wl1 ^= wl2 ^ 24; g2_inv (wh2,wl2, wh2,wl2);
232 wh2 ^= wh3; wl2 ^= wl3 ^ 23; g8_inv (wh3,wl3, wh3,wl3);
233 wh3 ^= wh4; wl3 ^= wl4 ^ 22; g4_inv (wh4,wl4, wh4,wl4);
234 wh4 ^= wh1; wl4 ^= wl1 ^ 21; g0_inv (wh1,wl1, wh1,wl1);
235 wh1 ^= wh2; wl1 ^= wl2 ^ 20; g6_inv (wh2,wl2, wh2,wl2);
236 wh2 ^= wh3; wl2 ^= wl3 ^ 19; g2_inv (wh3,wl3, wh3,wl3);
237 wh3 ^= wh4; wl3 ^= wl4 ^ 18; g8_inv (wh4,wl4, wh4,wl4);
238 wh4 ^= wh1; wl4 ^= wl1 ^ 17; g4_inv (wh1,wl1, wh1,wl1);
239
240 /* third 8 rounds */
241 g0_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 16; wh3 ^= wh2;
242 g6_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 15; wh4 ^= wh3;
243 g2_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 14; wh1 ^= wh4;
244 g8_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 13; wh2 ^= wh1;
245 g4_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 12; wh3 ^= wh2;
246 g0_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 11; wh4 ^= wh3;
247 g6_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 10; wh1 ^= wh4;
248 g2_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 9; wh2 ^= wh1;
249
250 /* last 8 rounds */
251 wh1 ^= wh2; wl1 ^= wl2 ^ 8; g8_inv (wh2,wl2, wh2,wl2);
252 wh2 ^= wh3; wl2 ^= wl3 ^ 7; g4_inv (wh3,wl3, wh3,wl3);
253 wh3 ^= wh4; wl3 ^= wl4 ^ 6; g0_inv (wh4,wl4, wh4,wl4);
254 wh4 ^= wh1; wl4 ^= wl1 ^ 5; g6_inv (wh1,wl1, wh1,wl1);
255 wh1 ^= wh2; wl1 ^= wl2 ^ 4; g2_inv (wh2,wl2, wh2,wl2);
256 wh2 ^= wh3; wl2 ^= wl3 ^ 3; g8_inv (wh3,wl3, wh3,wl3);
257 wh3 ^= wh4; wl3 ^= wl4 ^ 2; g4_inv (wh4,wl4, wh4,wl4);
258 wh4 ^= wh1; wl4 ^= wl1 ^ 1; g0_inv (wh1,wl1, wh1,wl1);
259
260 /* pack into byte vector */
261 plain [0] = wh1; plain [1] = wl1;
262 plain [2] = wh2; plain [3] = wl2;
263 plain [4] = wh3; plain [5] = wl3;
264 plain [6] = wh4; plain [7] = wl4;
265}
266
267MODULE(MODULE_CLASS_MISC, skipjack, NULL);
268
269static int
270skipjack_modcmd(modcmd_t cmd, void *opaque)
271{
272
273 switch (cmd) {
274 case MODULE_CMD_INIT:
275 return 0;
276 case MODULE_CMD_FINI:
277 return 0;
278 default:
279 return ENOTTY;
280 }
281}
282