1 /* 2 * Copyright 1995-2016 The OpenSSL Project Authors. All Rights Reserved. 3 * 4 * Licensed under the Apache License 2.0 (the "License"). You may not use 5 * this file except in compliance with the License. You can obtain a copy 6 * in the file LICENSE in the source distribution or at 7 * https://www.openssl.org/source/license.html 8 */ 9 10 #ifndef OSSL_CRYPTO_DES_LOCAL_H 11 # define OSSL_CRYPTO_DES_LOCAL_H 12 13 # include <openssl/e_os2.h> 14 15 # include <stdio.h> 16 # include <stdlib.h> 17 # include <string.h> 18 19 # include <openssl/des.h> 20 21 # ifdef OPENSSL_BUILD_SHLIBCRYPTO 22 # undef OPENSSL_EXTERN 23 # define OPENSSL_EXTERN OPENSSL_EXPORT 24 # endif 25 26 # define ITERATIONS 16 27 # define HALF_ITERATIONS 8 28 29 # define c2l(c,l) (l =((DES_LONG)(*((c)++))) , \ 30 l|=((DES_LONG)(*((c)++)))<< 8L, \ 31 l|=((DES_LONG)(*((c)++)))<<16L, \ 32 l|=((DES_LONG)(*((c)++)))<<24L) 33 34 /* NOTE - c is not incremented as per c2l */ 35 # define c2ln(c,l1,l2,n) { \ 36 c+=n; \ 37 l1=l2=0; \ 38 switch (n) { \ 39 case 8: l2 =((DES_LONG)(*(--(c))))<<24L; \ 40 /* fall thru */ \ 41 case 7: l2|=((DES_LONG)(*(--(c))))<<16L; \ 42 /* fall thru */ \ 43 case 6: l2|=((DES_LONG)(*(--(c))))<< 8L; \ 44 /* fall thru */ \ 45 case 5: l2|=((DES_LONG)(*(--(c)))); \ 46 /* fall thru */ \ 47 case 4: l1 =((DES_LONG)(*(--(c))))<<24L; \ 48 /* fall thru */ \ 49 case 3: l1|=((DES_LONG)(*(--(c))))<<16L; \ 50 /* fall thru */ \ 51 case 2: l1|=((DES_LONG)(*(--(c))))<< 8L; \ 52 /* fall thru */ \ 53 case 1: l1|=((DES_LONG)(*(--(c)))); \ 54 } \ 55 } 56 57 # define l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \ 58 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \ 59 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \ 60 *((c)++)=(unsigned char)(((l)>>24L)&0xff)) 61 62 /* 63 * replacements for htonl and ntohl since I have no idea what to do when 64 * faced with machines with 8 byte longs. 65 */ 66 67 # define n2l(c,l) (l =((DES_LONG)(*((c)++)))<<24L, \ 68 l|=((DES_LONG)(*((c)++)))<<16L, \ 69 l|=((DES_LONG)(*((c)++)))<< 8L, \ 70 l|=((DES_LONG)(*((c)++)))) 71 72 # define l2n(l,c) (*((c)++)=(unsigned char)(((l)>>24L)&0xff), \ 73 *((c)++)=(unsigned char)(((l)>>16L)&0xff), \ 74 *((c)++)=(unsigned char)(((l)>> 8L)&0xff), \ 75 *((c)++)=(unsigned char)(((l) )&0xff)) 76 77 /* NOTE - c is not incremented as per l2c */ 78 # define l2cn(l1,l2,c,n) { \ 79 c+=n; \ 80 switch (n) { \ 81 case 8: *(--(c))=(unsigned char)(((l2)>>24L)&0xff); \ 82 /* fall thru */ \ 83 case 7: *(--(c))=(unsigned char)(((l2)>>16L)&0xff); \ 84 /* fall thru */ \ 85 case 6: *(--(c))=(unsigned char)(((l2)>> 8L)&0xff); \ 86 /* fall thru */ \ 87 case 5: *(--(c))=(unsigned char)(((l2) )&0xff); \ 88 /* fall thru */ \ 89 case 4: *(--(c))=(unsigned char)(((l1)>>24L)&0xff); \ 90 /* fall thru */ \ 91 case 3: *(--(c))=(unsigned char)(((l1)>>16L)&0xff); \ 92 /* fall thru */ \ 93 case 2: *(--(c))=(unsigned char)(((l1)>> 8L)&0xff); \ 94 /* fall thru */ \ 95 case 1: *(--(c))=(unsigned char)(((l1) )&0xff); \ 96 } \ 97 } 98 99 # if defined(_MSC_VER) 100 # define ROTATE(a,n) (_lrotr(a,n)) 101 # elif defined(__ICC) 102 # define ROTATE(a,n) (_rotr(a,n)) 103 # elif defined(__GNUC__) && __GNUC__>=2 && !defined(__STRICT_ANSI__) && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM) && !defined(PEDANTIC) 104 # if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__) 105 # define ROTATE(a,n) ({ register unsigned int ret; \ 106 asm ("rorl %1,%0" \ 107 : "=r"(ret) \ 108 : "I"(n),"0"(a) \ 109 : "cc"); \ 110 ret; \ 111 }) 112 # elif defined(__riscv_zbb) || defined(__riscv_zbkb) 113 # if __riscv_xlen == 64 114 # define ROTATE(x, n) ({ register unsigned int ret; \ 115 asm ("roriw %0, %1, %2" \ 116 : "=r"(ret) \ 117 : "r"(x), "i"(n)); ret; }) 118 # endif 119 # if __riscv_xlen == 32 120 # define ROTATE(x, n) ({ register unsigned int ret; \ 121 asm ("rori %0, %1, %2" \ 122 : "=r"(ret) \ 123 : "r"(x), "i"(n)); ret; }) 124 # endif 125 # endif 126 # endif 127 # ifndef ROTATE 128 # define ROTATE(a,n) (((a)>>(n))+((a)<<(32-(n)))) 129 # endif 130 131 /* 132 * Don't worry about the LOAD_DATA() stuff, that is used by fcrypt() to add 133 * it's little bit to the front 134 */ 135 136 # ifdef DES_FCRYPT 137 138 # define LOAD_DATA_tmp(R,S,u,t,E0,E1) \ 139 { DES_LONG tmp; LOAD_DATA(R,S,u,t,E0,E1,tmp); } 140 141 # define LOAD_DATA(R,S,u,t,E0,E1,tmp) \ 142 t=R^(R>>16L); \ 143 u=t&E0; t&=E1; \ 144 tmp=(u<<16); u^=R^s[S ]; u^=tmp; \ 145 tmp=(t<<16); t^=R^s[S+1]; t^=tmp 146 # else 147 # define LOAD_DATA_tmp(a,b,c,d,e,f) LOAD_DATA(a,b,c,d,e,f,g) 148 # define LOAD_DATA(R,S,u,t,E0,E1,tmp) \ 149 u=R^s[S ]; \ 150 t=R^s[S+1] 151 # endif 152 153 /* 154 * It recently occurred to me that 0^0^0^0^0^0^0 == 0, so there is no reason 155 * to not xor all the sub items together. This potentially saves a register 156 * since things can be xored directly into L 157 */ 158 159 # define D_ENCRYPT(LL,R,S) { \ 160 LOAD_DATA_tmp(R,S,u,t,E0,E1); \ 161 t=ROTATE(t,4); \ 162 LL^= \ 163 DES_SPtrans[0][(u>> 2L)&0x3f]^ \ 164 DES_SPtrans[2][(u>>10L)&0x3f]^ \ 165 DES_SPtrans[4][(u>>18L)&0x3f]^ \ 166 DES_SPtrans[6][(u>>26L)&0x3f]^ \ 167 DES_SPtrans[1][(t>> 2L)&0x3f]^ \ 168 DES_SPtrans[3][(t>>10L)&0x3f]^ \ 169 DES_SPtrans[5][(t>>18L)&0x3f]^ \ 170 DES_SPtrans[7][(t>>26L)&0x3f]; } 171 172 /*- 173 * IP and FP 174 * The problem is more of a geometric problem that random bit fiddling. 175 0 1 2 3 4 5 6 7 62 54 46 38 30 22 14 6 176 8 9 10 11 12 13 14 15 60 52 44 36 28 20 12 4 177 16 17 18 19 20 21 22 23 58 50 42 34 26 18 10 2 178 24 25 26 27 28 29 30 31 to 56 48 40 32 24 16 8 0 179 180 32 33 34 35 36 37 38 39 63 55 47 39 31 23 15 7 181 40 41 42 43 44 45 46 47 61 53 45 37 29 21 13 5 182 48 49 50 51 52 53 54 55 59 51 43 35 27 19 11 3 183 56 57 58 59 60 61 62 63 57 49 41 33 25 17 9 1 184 185 The output has been subject to swaps of the form 186 0 1 -> 3 1 but the odd and even bits have been put into 187 2 3 2 0 188 different words. The main trick is to remember that 189 t=((l>>size)^r)&(mask); 190 r^=t; 191 l^=(t<<size); 192 can be used to swap and move bits between words. 193 194 So l = 0 1 2 3 r = 16 17 18 19 195 4 5 6 7 20 21 22 23 196 8 9 10 11 24 25 26 27 197 12 13 14 15 28 29 30 31 198 becomes (for size == 2 and mask == 0x3333) 199 t = 2^16 3^17 -- -- l = 0 1 16 17 r = 2 3 18 19 200 6^20 7^21 -- -- 4 5 20 21 6 7 22 23 201 10^24 11^25 -- -- 8 9 24 25 10 11 24 25 202 14^28 15^29 -- -- 12 13 28 29 14 15 28 29 203 204 Thanks for hints from Richard Outerbridge - he told me IP&FP 205 could be done in 15 xor, 10 shifts and 5 ands. 206 When I finally started to think of the problem in 2D 207 I first got ~42 operations without xors. When I remembered 208 how to use xors :-) I got it to its final state. 209 */ 210 # define PERM_OP(a,b,t,n,m) ((t)=((((a)>>(n))^(b))&(m)),\ 211 (b)^=(t),\ 212 (a)^=((t)<<(n))) 213 214 # define IP(l,r) \ 215 { \ 216 register DES_LONG tt; \ 217 PERM_OP(r,l,tt, 4,0x0f0f0f0fL); \ 218 PERM_OP(l,r,tt,16,0x0000ffffL); \ 219 PERM_OP(r,l,tt, 2,0x33333333L); \ 220 PERM_OP(l,r,tt, 8,0x00ff00ffL); \ 221 PERM_OP(r,l,tt, 1,0x55555555L); \ 222 } 223 224 # define FP(l,r) \ 225 { \ 226 register DES_LONG tt; \ 227 PERM_OP(l,r,tt, 1,0x55555555L); \ 228 PERM_OP(r,l,tt, 8,0x00ff00ffL); \ 229 PERM_OP(l,r,tt, 2,0x33333333L); \ 230 PERM_OP(r,l,tt,16,0x0000ffffL); \ 231 PERM_OP(l,r,tt, 4,0x0f0f0f0fL); \ 232 } 233 234 extern const DES_LONG DES_SPtrans[8][64]; 235 236 void fcrypt_body(DES_LONG *out, DES_key_schedule *ks, 237 DES_LONG Eswap0, DES_LONG Eswap1); 238 239 #endif 240