1 /* $FreeBSD: src/sys/opencrypto/skipjack.c,v 1.1.2.1 2002/11/21 23:34:23 sam Exp $ */ 2 /* $DragonFly: src/sys/opencrypto/skipjack.c,v 1.2 2003/06/17 04:28:54 dillon Exp $ */ 3 /* $OpenBSD: skipjack.c,v 1.3 2001/05/05 00:31:34 angelos Exp $ */ 4 5 /* 6 * Further optimized test implementation of SKIPJACK algorithm 7 * Mark Tillotson <markt@chaos.org.uk>, 25 June 98 8 * Optimizations suit RISC (lots of registers) machine best. 9 * 10 * based on unoptimized implementation of 11 * Panu Rissanen <bande@lut.fi> 960624 12 * 13 * SKIPJACK and KEA Algorithm Specifications 14 * Version 2.0 15 * 29 May 1998 16 */ 17 18 #include <sys/param.h> 19 20 #include <opencrypto/skipjack.h> 21 22 static const u_int8_t ftable[0x100] = 23 { 24 0xa3, 0xd7, 0x09, 0x83, 0xf8, 0x48, 0xf6, 0xf4, 25 0xb3, 0x21, 0x15, 0x78, 0x99, 0xb1, 0xaf, 0xf9, 26 0xe7, 0x2d, 0x4d, 0x8a, 0xce, 0x4c, 0xca, 0x2e, 27 0x52, 0x95, 0xd9, 0x1e, 0x4e, 0x38, 0x44, 0x28, 28 0x0a, 0xdf, 0x02, 0xa0, 0x17, 0xf1, 0x60, 0x68, 29 0x12, 0xb7, 0x7a, 0xc3, 0xe9, 0xfa, 0x3d, 0x53, 30 0x96, 0x84, 0x6b, 0xba, 0xf2, 0x63, 0x9a, 0x19, 31 0x7c, 0xae, 0xe5, 0xf5, 0xf7, 0x16, 0x6a, 0xa2, 32 0x39, 0xb6, 0x7b, 0x0f, 0xc1, 0x93, 0x81, 0x1b, 33 0xee, 0xb4, 0x1a, 0xea, 0xd0, 0x91, 0x2f, 0xb8, 34 0x55, 0xb9, 0xda, 0x85, 0x3f, 0x41, 0xbf, 0xe0, 35 0x5a, 0x58, 0x80, 0x5f, 0x66, 0x0b, 0xd8, 0x90, 36 0x35, 0xd5, 0xc0, 0xa7, 0x33, 0x06, 0x65, 0x69, 37 0x45, 0x00, 0x94, 0x56, 0x6d, 0x98, 0x9b, 0x76, 38 0x97, 0xfc, 0xb2, 0xc2, 0xb0, 0xfe, 0xdb, 0x20, 39 0xe1, 0xeb, 0xd6, 0xe4, 0xdd, 0x47, 0x4a, 0x1d, 40 0x42, 0xed, 0x9e, 0x6e, 0x49, 0x3c, 0xcd, 0x43, 41 0x27, 0xd2, 0x07, 0xd4, 0xde, 0xc7, 0x67, 0x18, 42 0x89, 0xcb, 0x30, 0x1f, 0x8d, 0xc6, 0x8f, 0xaa, 43 0xc8, 0x74, 0xdc, 0xc9, 0x5d, 0x5c, 0x31, 0xa4, 44 0x70, 0x88, 0x61, 0x2c, 0x9f, 0x0d, 0x2b, 0x87, 45 0x50, 0x82, 0x54, 0x64, 0x26, 0x7d, 0x03, 0x40, 46 0x34, 0x4b, 0x1c, 0x73, 0xd1, 0xc4, 0xfd, 0x3b, 47 0xcc, 0xfb, 0x7f, 0xab, 0xe6, 0x3e, 0x5b, 0xa5, 48 0xad, 0x04, 0x23, 0x9c, 0x14, 0x51, 0x22, 0xf0, 49 0x29, 0x79, 0x71, 0x7e, 0xff, 0x8c, 0x0e, 0xe2, 50 0x0c, 0xef, 0xbc, 0x72, 0x75, 0x6f, 0x37, 0xa1, 51 0xec, 0xd3, 0x8e, 0x62, 0x8b, 0x86, 0x10, 0xe8, 52 0x08, 0x77, 0x11, 0xbe, 0x92, 0x4f, 0x24, 0xc5, 53 0x32, 0x36, 0x9d, 0xcf, 0xf3, 0xa6, 0xbb, 0xac, 54 0x5e, 0x6c, 0xa9, 0x13, 0x57, 0x25, 0xb5, 0xe3, 55 0xbd, 0xa8, 0x3a, 0x01, 0x05, 0x59, 0x2a, 0x46 56 }; 57 58 /* 59 * For each key byte generate a table to represent the function 60 * ftable [in ^ keybyte] 61 * 62 * These tables used to save an XOR in each stage of the G-function 63 * the tables are hopefully pointed to by register allocated variables 64 * k0, k1..k9 65 */ 66 void 67 subkey_table_gen (u_int8_t *key, u_int8_t **key_tables) 68 { 69 int i, k; 70 71 for (k = 0; k < 10; k++) { 72 u_int8_t key_byte = key [k]; 73 u_int8_t * table = key_tables[k]; 74 for (i = 0; i < 0x100; i++) 75 table [i] = ftable [i ^ key_byte]; 76 } 77 } 78 79 80 #define g(k0, k1, k2, k3, ih, il, oh, ol) \ 81 { \ 82 oh = k##k0 [il] ^ ih; \ 83 ol = k##k1 [oh] ^ il; \ 84 oh = k##k2 [ol] ^ oh; \ 85 ol = k##k3 [oh] ^ ol; \ 86 } 87 88 #define g0(ih, il, oh, ol) g(0, 1, 2, 3, ih, il, oh, ol) 89 #define g4(ih, il, oh, ol) g(4, 5, 6, 7, ih, il, oh, ol) 90 #define g8(ih, il, oh, ol) g(8, 9, 0, 1, ih, il, oh, ol) 91 #define g2(ih, il, oh, ol) g(2, 3, 4, 5, ih, il, oh, ol) 92 #define g6(ih, il, oh, ol) g(6, 7, 8, 9, ih, il, oh, ol) 93 94 95 #define g_inv(k0, k1, k2, k3, ih, il, oh, ol) \ 96 { \ 97 ol = k##k3 [ih] ^ il; \ 98 oh = k##k2 [ol] ^ ih; \ 99 ol = k##k1 [oh] ^ ol; \ 100 oh = k##k0 [ol] ^ oh; \ 101 } 102 103 104 #define g0_inv(ih, il, oh, ol) g_inv(0, 1, 2, 3, ih, il, oh, ol) 105 #define g4_inv(ih, il, oh, ol) g_inv(4, 5, 6, 7, ih, il, oh, ol) 106 #define g8_inv(ih, il, oh, ol) g_inv(8, 9, 0, 1, ih, il, oh, ol) 107 #define g2_inv(ih, il, oh, ol) g_inv(2, 3, 4, 5, ih, il, oh, ol) 108 #define g6_inv(ih, il, oh, ol) g_inv(6, 7, 8, 9, ih, il, oh, ol) 109 110 /* optimized version of Skipjack algorithm 111 * 112 * the appropriate g-function is inlined for each round 113 * 114 * the data movement is minimized by rotating the names of the 115 * variables w1..w4, not their contents (saves 3 moves per round) 116 * 117 * the loops are completely unrolled (needed to staticize choice of g) 118 * 119 * compiles to about 470 instructions on a Sparc (gcc -O) 120 * which is about 58 instructions per byte, 14 per round. 121 * gcc seems to leave in some unnecessary and with 0xFF operations 122 * but only in the latter part of the functions. Perhaps it 123 * runs out of resources to properly optimize long inlined function? 124 * in theory should get about 11 instructions per round, not 14 125 */ 126 127 void 128 skipjack_forwards(u_int8_t *plain, u_int8_t *cipher, u_int8_t **key_tables) 129 { 130 u_int8_t wh1 = plain[0]; u_int8_t wl1 = plain[1]; 131 u_int8_t wh2 = plain[2]; u_int8_t wl2 = plain[3]; 132 u_int8_t wh3 = plain[4]; u_int8_t wl3 = plain[5]; 133 u_int8_t wh4 = plain[6]; u_int8_t wl4 = plain[7]; 134 135 u_int8_t * k0 = key_tables [0]; 136 u_int8_t * k1 = key_tables [1]; 137 u_int8_t * k2 = key_tables [2]; 138 u_int8_t * k3 = key_tables [3]; 139 u_int8_t * k4 = key_tables [4]; 140 u_int8_t * k5 = key_tables [5]; 141 u_int8_t * k6 = key_tables [6]; 142 u_int8_t * k7 = key_tables [7]; 143 u_int8_t * k8 = key_tables [8]; 144 u_int8_t * k9 = key_tables [9]; 145 146 /* first 8 rounds */ 147 g0 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 1; wh4 ^= wh1; 148 g4 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 2; wh3 ^= wh4; 149 g8 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 3; wh2 ^= wh3; 150 g2 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 4; wh1 ^= wh2; 151 g6 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 5; wh4 ^= wh1; 152 g0 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 6; wh3 ^= wh4; 153 g4 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 7; wh2 ^= wh3; 154 g8 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 8; wh1 ^= wh2; 155 156 /* second 8 rounds */ 157 wh2 ^= wh1; wl2 ^= wl1 ^ 9 ; g2 (wh1,wl1, wh1,wl1); 158 wh1 ^= wh4; wl1 ^= wl4 ^ 10; g6 (wh4,wl4, wh4,wl4); 159 wh4 ^= wh3; wl4 ^= wl3 ^ 11; g0 (wh3,wl3, wh3,wl3); 160 wh3 ^= wh2; wl3 ^= wl2 ^ 12; g4 (wh2,wl2, wh2,wl2); 161 wh2 ^= wh1; wl2 ^= wl1 ^ 13; g8 (wh1,wl1, wh1,wl1); 162 wh1 ^= wh4; wl1 ^= wl4 ^ 14; g2 (wh4,wl4, wh4,wl4); 163 wh4 ^= wh3; wl4 ^= wl3 ^ 15; g6 (wh3,wl3, wh3,wl3); 164 wh3 ^= wh2; wl3 ^= wl2 ^ 16; g0 (wh2,wl2, wh2,wl2); 165 166 /* third 8 rounds */ 167 g4 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 17; wh4 ^= wh1; 168 g8 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 18; wh3 ^= wh4; 169 g2 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 19; wh2 ^= wh3; 170 g6 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 20; wh1 ^= wh2; 171 g0 (wh1,wl1, wh1,wl1); wl4 ^= wl1 ^ 21; wh4 ^= wh1; 172 g4 (wh4,wl4, wh4,wl4); wl3 ^= wl4 ^ 22; wh3 ^= wh4; 173 g8 (wh3,wl3, wh3,wl3); wl2 ^= wl3 ^ 23; wh2 ^= wh3; 174 g2 (wh2,wl2, wh2,wl2); wl1 ^= wl2 ^ 24; wh1 ^= wh2; 175 176 /* last 8 rounds */ 177 wh2 ^= wh1; wl2 ^= wl1 ^ 25; g6 (wh1,wl1, wh1,wl1); 178 wh1 ^= wh4; wl1 ^= wl4 ^ 26; g0 (wh4,wl4, wh4,wl4); 179 wh4 ^= wh3; wl4 ^= wl3 ^ 27; g4 (wh3,wl3, wh3,wl3); 180 wh3 ^= wh2; wl3 ^= wl2 ^ 28; g8 (wh2,wl2, wh2,wl2); 181 wh2 ^= wh1; wl2 ^= wl1 ^ 29; g2 (wh1,wl1, wh1,wl1); 182 wh1 ^= wh4; wl1 ^= wl4 ^ 30; g6 (wh4,wl4, wh4,wl4); 183 wh4 ^= wh3; wl4 ^= wl3 ^ 31; g0 (wh3,wl3, wh3,wl3); 184 wh3 ^= wh2; wl3 ^= wl2 ^ 32; g4 (wh2,wl2, wh2,wl2); 185 186 /* pack into byte vector */ 187 cipher [0] = wh1; cipher [1] = wl1; 188 cipher [2] = wh2; cipher [3] = wl2; 189 cipher [4] = wh3; cipher [5] = wl3; 190 cipher [6] = wh4; cipher [7] = wl4; 191 } 192 193 194 void 195 skipjack_backwards (u_int8_t *cipher, u_int8_t *plain, u_int8_t **key_tables) 196 { 197 /* setup 4 16-bit portions */ 198 u_int8_t wh1 = cipher[0]; u_int8_t wl1 = cipher[1]; 199 u_int8_t wh2 = cipher[2]; u_int8_t wl2 = cipher[3]; 200 u_int8_t wh3 = cipher[4]; u_int8_t wl3 = cipher[5]; 201 u_int8_t wh4 = cipher[6]; u_int8_t wl4 = cipher[7]; 202 203 u_int8_t * k0 = key_tables [0]; 204 u_int8_t * k1 = key_tables [1]; 205 u_int8_t * k2 = key_tables [2]; 206 u_int8_t * k3 = key_tables [3]; 207 u_int8_t * k4 = key_tables [4]; 208 u_int8_t * k5 = key_tables [5]; 209 u_int8_t * k6 = key_tables [6]; 210 u_int8_t * k7 = key_tables [7]; 211 u_int8_t * k8 = key_tables [8]; 212 u_int8_t * k9 = key_tables [9]; 213 214 /* first 8 rounds */ 215 g4_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 32; wh3 ^= wh2; 216 g0_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 31; wh4 ^= wh3; 217 g6_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 30; wh1 ^= wh4; 218 g2_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 29; wh2 ^= wh1; 219 g8_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 28; wh3 ^= wh2; 220 g4_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 27; wh4 ^= wh3; 221 g0_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 26; wh1 ^= wh4; 222 g6_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 25; wh2 ^= wh1; 223 224 /* second 8 rounds */ 225 wh1 ^= wh2; wl1 ^= wl2 ^ 24; g2_inv (wh2,wl2, wh2,wl2); 226 wh2 ^= wh3; wl2 ^= wl3 ^ 23; g8_inv (wh3,wl3, wh3,wl3); 227 wh3 ^= wh4; wl3 ^= wl4 ^ 22; g4_inv (wh4,wl4, wh4,wl4); 228 wh4 ^= wh1; wl4 ^= wl1 ^ 21; g0_inv (wh1,wl1, wh1,wl1); 229 wh1 ^= wh2; wl1 ^= wl2 ^ 20; g6_inv (wh2,wl2, wh2,wl2); 230 wh2 ^= wh3; wl2 ^= wl3 ^ 19; g2_inv (wh3,wl3, wh3,wl3); 231 wh3 ^= wh4; wl3 ^= wl4 ^ 18; g8_inv (wh4,wl4, wh4,wl4); 232 wh4 ^= wh1; wl4 ^= wl1 ^ 17; g4_inv (wh1,wl1, wh1,wl1); 233 234 /* third 8 rounds */ 235 g0_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 16; wh3 ^= wh2; 236 g6_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 15; wh4 ^= wh3; 237 g2_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 14; wh1 ^= wh4; 238 g8_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 13; wh2 ^= wh1; 239 g4_inv (wh2,wl2, wh2,wl2); wl3 ^= wl2 ^ 12; wh3 ^= wh2; 240 g0_inv (wh3,wl3, wh3,wl3); wl4 ^= wl3 ^ 11; wh4 ^= wh3; 241 g6_inv (wh4,wl4, wh4,wl4); wl1 ^= wl4 ^ 10; wh1 ^= wh4; 242 g2_inv (wh1,wl1, wh1,wl1); wl2 ^= wl1 ^ 9; wh2 ^= wh1; 243 244 /* last 8 rounds */ 245 wh1 ^= wh2; wl1 ^= wl2 ^ 8; g8_inv (wh2,wl2, wh2,wl2); 246 wh2 ^= wh3; wl2 ^= wl3 ^ 7; g4_inv (wh3,wl3, wh3,wl3); 247 wh3 ^= wh4; wl3 ^= wl4 ^ 6; g0_inv (wh4,wl4, wh4,wl4); 248 wh4 ^= wh1; wl4 ^= wl1 ^ 5; g6_inv (wh1,wl1, wh1,wl1); 249 wh1 ^= wh2; wl1 ^= wl2 ^ 4; g2_inv (wh2,wl2, wh2,wl2); 250 wh2 ^= wh3; wl2 ^= wl3 ^ 3; g8_inv (wh3,wl3, wh3,wl3); 251 wh3 ^= wh4; wl3 ^= wl4 ^ 2; g4_inv (wh4,wl4, wh4,wl4); 252 wh4 ^= wh1; wl4 ^= wl1 ^ 1; g0_inv (wh1,wl1, wh1,wl1); 253 254 /* pack into byte vector */ 255 plain [0] = wh1; plain [1] = wl1; 256 plain [2] = wh2; plain [3] = wl2; 257 plain [4] = wh3; plain [5] = wl3; 258 plain [6] = wh4; plain [7] = wl4; 259 } 260