| 1 |
/*************************************************************************** |
| 2 |
TWOFISH2.C -- Optimized C API calls for TWOFISH AES submission |
| 3 |
|
| 4 |
Submitters: |
| 5 |
Bruce Schneier, Counterpane Systems |
| 6 |
Doug Whiting, Hi/fn |
| 7 |
John Kelsey, Counterpane Systems |
| 8 |
Chris Hall, Counterpane Systems |
| 9 |
David Wagner, UC Berkeley |
| 10 |
|
| 11 |
Code Author: Doug Whiting, Hi/fn |
| 12 |
|
| 13 |
Version 1.00 April 1998 |
| 14 |
|
| 15 |
Copyright 1998, Hi/fn and Counterpane Systems. All rights reserved. |
| 16 |
|
| 17 |
Notes: |
| 18 |
* Optimized version |
| 19 |
* Tab size is set to 4 characters in this file |
| 20 |
|
| 21 |
***************************************************************************/ |
| 22 |
#include "aes.h" |
| 23 |
#include "table.h" |
| 24 |
|
| 25 |
#include <memory.h> |
| 26 |
/*#include <assert.h>*/ |
| 27 |
|
| 28 |
#if defined(min_key) && !defined(MIN_KEY) |
| 29 |
#define MIN_KEY 1 /* toupper() */ |
| 30 |
#elif defined(part_key) && !defined(PART_KEY) |
| 31 |
#define PART_KEY 1 |
| 32 |
#elif defined(zero_key) && !defined(ZERO_KEY) |
| 33 |
#define ZERO_KEY 1 |
| 34 |
#endif |
| 35 |
|
| 36 |
|
| 37 |
#ifdef USE_ASM |
| 38 |
extern int useAsm; /* ok to use ASM code? */ |
| 39 |
|
| 40 |
typedef int cdecl CipherProc |
| 41 |
(cipherInstance *cipher, keyInstance *key,BYTE *input,int inputLen,BYTE *outBuffer); |
| 42 |
typedef int cdecl KeySetupProc(keyInstance *key); |
| 43 |
|
| 44 |
extern CipherProc *blockEncrypt_86; /* ptr to ASM functions */ |
| 45 |
extern CipherProc *blockDecrypt_86; |
| 46 |
extern KeySetupProc *reKey_86; |
| 47 |
extern DWORD cdecl TwofishAsmCodeSize(void); |
| 48 |
#endif |
| 49 |
|
| 50 |
/* |
| 51 |
+***************************************************************************** |
| 52 |
* Constants/Macros/Tables |
| 53 |
-****************************************************************************/ |
| 54 |
|
| 55 |
#define CONST /* help syntax from C++, NOP here */ |
| 56 |
|
| 57 |
static CONST fullSbox MDStab; /* not actually const. Initialized ONE time */ |
| 58 |
static int needToBuildMDS=1; /* is MDStab initialized yet? */ |
| 59 |
|
| 60 |
#define BIG_TAB 0 |
| 61 |
|
| 62 |
#if BIG_TAB |
| 63 |
static BYTE bigTab[4][256][256]; /* pre-computed S-box */ |
| 64 |
#endif |
| 65 |
|
| 66 |
/* number of rounds for various key sizes: 128, 192, 256 */ |
| 67 |
/* (ignored for now in optimized code!) */ |
| 68 |
static CONST int numRounds[4]= {0,ROUNDS_128,ROUNDS_192,ROUNDS_256}; |
| 69 |
|
| 70 |
#if REENTRANT |
| 71 |
#define _sBox_ key->sBox8x32 |
| 72 |
#else |
| 73 |
static fullSbox _sBox_; /* permuted MDStab based on keys */ |
| 74 |
#endif |
| 75 |
#define _sBox8_(N) (((BYTE *) _sBox_) + (N)*256) |
| 76 |
|
| 77 |
/*------- see what level of S-box precomputation we need to do -----*/ |
| 78 |
#if defined(ZERO_KEY) |
| 79 |
#define MOD_STRING "(Zero S-box keying)" |
| 80 |
#define Fe32_128(x,R) \ |
| 81 |
( MDStab[0][p8(01)[p8(02)[_b(x,R )]^b0(SKEY[1])]^b0(SKEY[0])] ^ \ |
| 82 |
MDStab[1][p8(11)[p8(12)[_b(x,R+1)]^b1(SKEY[1])]^b1(SKEY[0])] ^ \ |
| 83 |
MDStab[2][p8(21)[p8(22)[_b(x,R+2)]^b2(SKEY[1])]^b2(SKEY[0])] ^ \ |
| 84 |
MDStab[3][p8(31)[p8(32)[_b(x,R+3)]^b3(SKEY[1])]^b3(SKEY[0])] ) |
| 85 |
#define Fe32_192(x,R) \ |
| 86 |
( MDStab[0][p8(01)[p8(02)[p8(03)[_b(x,R )]^b0(SKEY[2])]^b0(SKEY[1])]^b0(SKEY[0])] ^ \ |
| 87 |
MDStab[1][p8(11)[p8(12)[p8(13)[_b(x,R+1)]^b1(SKEY[2])]^b1(SKEY[1])]^b1(SKEY[0])] ^ \ |
| 88 |
MDStab[2][p8(21)[p8(22)[p8(23)[_b(x,R+2)]^b2(SKEY[2])]^b2(SKEY[1])]^b2(SKEY[0])] ^ \ |
| 89 |
MDStab[3][p8(31)[p8(32)[p8(33)[_b(x,R+3)]^b3(SKEY[2])]^b3(SKEY[1])]^b3(SKEY[0])] ) |
| 90 |
#define Fe32_256(x,R) \ |
| 91 |
( MDStab[0][p8(01)[p8(02)[p8(03)[p8(04)[_b(x,R )]^b0(SKEY[3])]^b0(SKEY[2])]^b0(SKEY[1])]^b0(SKEY[0])] ^ \ |
| 92 |
MDStab[1][p8(11)[p8(12)[p8(13)[p8(14)[_b(x,R+1)]^b1(SKEY[3])]^b1(SKEY[2])]^b1(SKEY[1])]^b1(SKEY[0])] ^ \ |
| 93 |
MDStab[2][p8(21)[p8(22)[p8(23)[p8(24)[_b(x,R+2)]^b2(SKEY[3])]^b2(SKEY[2])]^b2(SKEY[1])]^b2(SKEY[0])] ^ \ |
| 94 |
MDStab[3][p8(31)[p8(32)[p8(33)[p8(34)[_b(x,R+3)]^b3(SKEY[3])]^b3(SKEY[2])]^b3(SKEY[1])]^b3(SKEY[0])] ) |
| 95 |
|
| 96 |
#define GetSboxKey DWORD SKEY[4]; /* local copy */ \ |
| 97 |
memcpy(SKEY,key->sboxKeys,sizeof(SKEY)); |
| 98 |
/*----------------------------------------------------------------*/ |
| 99 |
#elif defined(MIN_KEY) |
| 100 |
#define MOD_STRING "(Minimal keying)" |
| 101 |
#define Fe32_(x,R)(MDStab[0][p8(01)[_sBox8_(0)[_b(x,R )]] ^ b0(SKEY0)] ^ \ |
| 102 |
MDStab[1][p8(11)[_sBox8_(1)[_b(x,R+1)]] ^ b1(SKEY0)] ^ \ |
| 103 |
MDStab[2][p8(21)[_sBox8_(2)[_b(x,R+2)]] ^ b2(SKEY0)] ^ \ |
| 104 |
MDStab[3][p8(31)[_sBox8_(3)[_b(x,R+3)]] ^ b3(SKEY0)]) |
| 105 |
#define sbSet(N,i,J,v) { _sBox8_(N)[i+J] = v; } |
| 106 |
#define GetSboxKey DWORD SKEY0 = key->sboxKeys[0] /* local copy */ |
| 107 |
/*----------------------------------------------------------------*/ |
| 108 |
#elif defined(PART_KEY) |
| 109 |
#define MOD_STRING "(Partial keying)" |
| 110 |
#define Fe32_(x,R)(MDStab[0][_sBox8_(0)[_b(x,R )]] ^ \ |
| 111 |
MDStab[1][_sBox8_(1)[_b(x,R+1)]] ^ \ |
| 112 |
MDStab[2][_sBox8_(2)[_b(x,R+2)]] ^ \ |
| 113 |
MDStab[3][_sBox8_(3)[_b(x,R+3)]]) |
| 114 |
#define sbSet(N,i,J,v) { _sBox8_(N)[i+J] = v; } |
| 115 |
#define GetSboxKey |
| 116 |
/*----------------------------------------------------------------*/ |
| 117 |
#else /* default is FULL_KEY */ |
| 118 |
#ifndef FULL_KEY |
| 119 |
#define FULL_KEY 1 |
| 120 |
#endif |
| 121 |
#if BIG_TAB |
| 122 |
#define TAB_STR " (Big table)" |
| 123 |
#else |
| 124 |
#define TAB_STR |
| 125 |
#endif |
| 126 |
#ifdef COMPILE_KEY |
| 127 |
#define MOD_STRING "(Compiled subkeys)" TAB_STR |
| 128 |
#else |
| 129 |
#define MOD_STRING "(Full keying)" TAB_STR |
| 130 |
#endif |
| 131 |
/* Fe32_ does a full S-box + MDS lookup. Need to #define _sBox_ before use. |
| 132 |
Note that we "interleave" 0,1, and 2,3 to avoid cache bank collisions |
| 133 |
in optimized assembly language. |
| 134 |
*/ |
| 135 |
#define Fe32_(x,R) (_sBox_[0][2*_b(x,R )] ^ _sBox_[0][2*_b(x,R+1)+1] ^ \ |
| 136 |
_sBox_[2][2*_b(x,R+2)] ^ _sBox_[2][2*_b(x,R+3)+1]) |
| 137 |
/* set a single S-box value, given the input byte */ |
| 138 |
//#define sbSet(N,i,J,v) { _sBox_[N&2][2*i+(N&1)+2*J]=MDStab[N][v]; } |
| 139 |
#define sbSet(N,i,J,v) { *((DWORD *)_sBox_ + (N&2)*256 + 2*i + (N&1) + 2*J) = MDStab[N][v]; } |
| 140 |
#define GetSboxKey |
| 141 |
#endif |
| 142 |
|
| 143 |
/* macro(s) for debugging help */ |
| 144 |
#define CHECK_TABLE 0 /* nonzero --> compare against "slow" table */ |
| 145 |
#define VALIDATE_PARMS 0 /* disable for full speed */ |
| 146 |
|
| 147 |
/* end of debug macros */ |
| 148 |
|
| 149 |
#ifdef GetCodeSize |
| 150 |
static extern DWORD Here(DWORD x); /* return caller's address! */ |
| 151 |
static DWORD TwofishCodeStart(void) { return Here(0); } |
| 152 |
#endif |
| 153 |
|
| 154 |
/* |
| 155 |
+***************************************************************************** |
| 156 |
* |
| 157 |
* Function Name: TableOp |
| 158 |
* |
| 159 |
* Function: Handle table use checking |
| 160 |
* |
| 161 |
* Arguments: op = what to do (see TAB_* defns in AES.H) |
| 162 |
* |
| 163 |
* Return: TRUE --> done (for TAB_QUERY) |
| 164 |
* |
| 165 |
* Notes: This routine is for use in generating the tables KAT file. |
| 166 |
* For this optimized version, we don't actually track table usage, |
| 167 |
* since it would make the macros incredibly ugly. Instead we just |
| 168 |
* run for a fixed number of queries and then say we're done. |
| 169 |
* |
| 170 |
-****************************************************************************/ |
| 171 |
static int TableOp(int op) |
| 172 |
{ |
| 173 |
static int queryCnt=0; |
| 174 |
|
| 175 |
switch (op) |
| 176 |
{ |
| 177 |
case TAB_DISABLE: |
| 178 |
break; |
| 179 |
case TAB_ENABLE: |
| 180 |
break; |
| 181 |
case TAB_RESET: |
| 182 |
queryCnt=0; |
| 183 |
break; |
| 184 |
case TAB_QUERY: |
| 185 |
queryCnt++; |
| 186 |
if (queryCnt < TAB_MIN_QUERY) |
| 187 |
return FALSE; |
| 188 |
} |
| 189 |
return TRUE; |
| 190 |
} |
| 191 |
|
| 192 |
|
| 193 |
#if CHECK_TABLE |
| 194 |
/* |
| 195 |
+***************************************************************************** |
| 196 |
* |
| 197 |
* Function Name: f32 |
| 198 |
* |
| 199 |
* Function: Run four bytes through keyed S-boxes and apply MDS matrix |
| 200 |
* |
| 201 |
* Arguments: x = input to f function |
| 202 |
* k32 = pointer to key dwords |
| 203 |
* keyLen = total key length (k32 --> keyLey/2 bits) |
| 204 |
* |
| 205 |
* Return: The output of the keyed permutation applied to x. |
| 206 |
* |
| 207 |
* Notes: |
| 208 |
* This function is a keyed 32-bit permutation. It is the major building |
| 209 |
* block for the Twofish round function, including the four keyed 8x8 |
| 210 |
* permutations and the 4x4 MDS matrix multiply. This function is used |
| 211 |
* both for generating round subkeys and within the round function on the |
| 212 |
* block being encrypted. |
| 213 |
* |
| 214 |
* This version is fairly slow and pedagogical, although a smartcard would |
| 215 |
* probably perform the operation exactly this way in firmware. For |
| 216 |
* ultimate performance, the entire operation can be completed with four |
| 217 |
* lookups into four 256x32-bit tables, with three dword xors. |
| 218 |
* |
| 219 |
* The MDS matrix is defined in TABLE.H. To multiply by Mij, just use the |
| 220 |
* macro Mij(x). |
| 221 |
* |
| 222 |
-****************************************************************************/ |
| 223 |
static DWORD f32(DWORD x,CONST DWORD *k32,int keyLen) |
| 224 |
{ |
| 225 |
BYTE b[4]; |
| 226 |
|
| 227 |
/* Run each byte thru 8x8 S-boxes, xoring with key byte at each stage. */ |
| 228 |
/* Note that each byte goes through a different combination of S-boxes.*/ |
| 229 |
|
| 230 |
*((DWORD *)b) = Bswap(x); /* make b[0] = LSB, b[3] = MSB */ |
| 231 |
switch (((keyLen + 63)/64) & 3) |
| 232 |
{ |
| 233 |
case 0: /* 256 bits of key */ |
| 234 |
b[0] = p8(04)[b[0]] ^ b0(k32[3]); |
| 235 |
b[1] = p8(14)[b[1]] ^ b1(k32[3]); |
| 236 |
b[2] = p8(24)[b[2]] ^ b2(k32[3]); |
| 237 |
b[3] = p8(34)[b[3]] ^ b3(k32[3]); |
| 238 |
/* fall thru, having pre-processed b[0]..b[3] with k32[3] */ |
| 239 |
case 3: /* 192 bits of key */ |
| 240 |
b[0] = p8(03)[b[0]] ^ b0(k32[2]); |
| 241 |
b[1] = p8(13)[b[1]] ^ b1(k32[2]); |
| 242 |
b[2] = p8(23)[b[2]] ^ b2(k32[2]); |
| 243 |
b[3] = p8(33)[b[3]] ^ b3(k32[2]); |
| 244 |
/* fall thru, having pre-processed b[0]..b[3] with k32[2] */ |
| 245 |
case 2: /* 128 bits of key */ |
| 246 |
b[0] = p8(00)[p8(01)[p8(02)[b[0]] ^ b0(k32[1])] ^ b0(k32[0])]; |
| 247 |
b[1] = p8(10)[p8(11)[p8(12)[b[1]] ^ b1(k32[1])] ^ b1(k32[0])]; |
| 248 |
b[2] = p8(20)[p8(21)[p8(22)[b[2]] ^ b2(k32[1])] ^ b2(k32[0])]; |
| 249 |
b[3] = p8(30)[p8(31)[p8(32)[b[3]] ^ b3(k32[1])] ^ b3(k32[0])]; |
| 250 |
} |
| 251 |
|
| 252 |
/* Now perform the MDS matrix multiply inline. */ |
| 253 |
return ((M00(b[0]) ^ M01(b[1]) ^ M02(b[2]) ^ M03(b[3])) ) ^ |
| 254 |
((M10(b[0]) ^ M11(b[1]) ^ M12(b[2]) ^ M13(b[3])) << 8) ^ |
| 255 |
((M20(b[0]) ^ M21(b[1]) ^ M22(b[2]) ^ M23(b[3])) << 16) ^ |
| 256 |
((M30(b[0]) ^ M31(b[1]) ^ M32(b[2]) ^ M33(b[3])) << 24) ; |
| 257 |
} |
| 258 |
#endif /* CHECK_TABLE */ |
| 259 |
|
| 260 |
|
| 261 |
/* |
| 262 |
+***************************************************************************** |
| 263 |
* |
| 264 |
* Function Name: RS_MDS_encode |
| 265 |
* |
| 266 |
* Function: Use (12,8) Reed-Solomon code over GF(256) to produce |
| 267 |
* a key S-box dword from two key material dwords. |
| 268 |
* |
| 269 |
* Arguments: k0 = 1st dword |
| 270 |
* k1 = 2nd dword |
| 271 |
* |
| 272 |
* Return: Remainder polynomial generated using RS code |
| 273 |
* |
| 274 |
* Notes: |
| 275 |
* Since this computation is done only once per reKey per 64 bits of key, |
| 276 |
* the performance impact of this routine is imperceptible. The RS code |
| 277 |
* chosen has "simple" coefficients to allow smartcard/hardware implementation |
| 278 |
* without lookup tables. |
| 279 |
* |
| 280 |
-****************************************************************************/ |
| 281 |
static DWORD RS_MDS_Encode(DWORD k0,DWORD k1) |
| 282 |
{ |
| 283 |
int i,j; |
| 284 |
DWORD r; |
| 285 |
|
| 286 |
for (i=r=0;i<2;i++) |
| 287 |
{ |
| 288 |
r ^= (i) ? k0 : k1; /* merge in 32 more key bits */ |
| 289 |
for (j=0;j<4;j++) /* shift one byte at a time */ |
| 290 |
RS_rem(r); |
| 291 |
} |
| 292 |
return r; |
| 293 |
} |
| 294 |
|
| 295 |
|
| 296 |
/* |
| 297 |
+***************************************************************************** |
| 298 |
* |
| 299 |
* Function Name: BuildMDS |
| 300 |
* |
| 301 |
* Function: Initialize the MDStab array |
| 302 |
* |
| 303 |
* Arguments: None. |
| 304 |
* |
| 305 |
* Return: None. |
| 306 |
* |
| 307 |
* Notes: |
| 308 |
* Here we precompute all the fixed MDS table. This only needs to be done |
| 309 |
* one time at initialization, after which the table is "CONST". |
| 310 |
* |
| 311 |
-****************************************************************************/ |
| 312 |
static void BuildMDS(void) |
| 313 |
{ |
| 314 |
int i; |
| 315 |
DWORD d; |
| 316 |
BYTE m1[2],mX[2],mY[4]; |
| 317 |
|
| 318 |
for (i=0;i<256;i++) |
| 319 |
{ |
| 320 |
m1[0]=P8x8[0][i]; /* compute all the matrix elements */ |
| 321 |
mX[0]=(BYTE) Mul_X(m1[0]); |
| 322 |
mY[0]=(BYTE) Mul_Y(m1[0]); |
| 323 |
|
| 324 |
m1[1]=P8x8[1][i]; |
| 325 |
mX[1]=(BYTE) Mul_X(m1[1]); |
| 326 |
mY[1]=(BYTE) Mul_Y(m1[1]); |
| 327 |
|
| 328 |
#undef Mul_1 /* change what the pre-processor does with Mij */ |
| 329 |
#undef Mul_X |
| 330 |
#undef Mul_Y |
| 331 |
#define Mul_1 m1 /* It will now access m01[], m5B[], and mEF[] */ |
| 332 |
#define Mul_X mX |
| 333 |
#define Mul_Y mY |
| 334 |
|
| 335 |
#define SetMDS(N) \ |
| 336 |
b0(d) = M0##N[P_##N##0]; \ |
| 337 |
b1(d) = M1##N[P_##N##0]; \ |
| 338 |
b2(d) = M2##N[P_##N##0]; \ |
| 339 |
b3(d) = M3##N[P_##N##0]; \ |
| 340 |
MDStab[N][i] = d; |
| 341 |
|
| 342 |
SetMDS(0); /* fill in the matrix with elements computed above */ |
| 343 |
SetMDS(1); |
| 344 |
SetMDS(2); |
| 345 |
SetMDS(3); |
| 346 |
} |
| 347 |
#undef Mul_1 |
| 348 |
#undef Mul_X |
| 349 |
#undef Mul_Y |
| 350 |
#define Mul_1 Mx_1 /* re-enable true multiply */ |
| 351 |
#define Mul_X Mx_X |
| 352 |
#define Mul_Y Mx_Y |
| 353 |
|
| 354 |
#if BIG_TAB |
| 355 |
{ |
| 356 |
int j,k; |
| 357 |
BYTE *q0,*q1; |
| 358 |
|
| 359 |
for (i=0;i<4;i++) |
| 360 |
{ |
| 361 |
switch (i) |
| 362 |
{ |
| 363 |
case 0: q0=p8(01); q1=p8(02); break; |
| 364 |
case 1: q0=p8(11); q1=p8(12); break; |
| 365 |
case 2: q0=p8(21); q1=p8(22); break; |
| 366 |
case 3: q0=p8(31); q1=p8(32); break; |
| 367 |
} |
| 368 |
for (j=0;j<256;j++) |
| 369 |
for (k=0;k<256;k++) |
| 370 |
bigTab[i][j][k]=q0[q1[k]^j]; |
| 371 |
} |
| 372 |
} |
| 373 |
#endif |
| 374 |
|
| 375 |
needToBuildMDS=0; /* NEVER modify the table again! */ |
| 376 |
} |
| 377 |
|
| 378 |
/* |
| 379 |
+***************************************************************************** |
| 380 |
* |
| 381 |
* Function Name: ReverseRoundSubkeys |
| 382 |
* |
| 383 |
* Function: Reverse order of round subkeys to switch between encrypt/decrypt |
| 384 |
* |
| 385 |
* Arguments: key = ptr to keyInstance to be reversed |
| 386 |
* newDir = new direction value |
| 387 |
* |
| 388 |
* Return: None. |
| 389 |
* |
| 390 |
* Notes: |
| 391 |
* This optimization allows both blockEncrypt and blockDecrypt to use the same |
| 392 |
* "fallthru" switch statement based on the number of rounds. |
| 393 |
* Note that key->numRounds must be even and >= 2 here. |
| 394 |
* |
| 395 |
-****************************************************************************/ |
| 396 |
static void ReverseRoundSubkeys(keyInstance *key,BYTE newDir) |
| 397 |
{ |
| 398 |
DWORD t0,t1; |
| 399 |
register DWORD *r0=key->subKeys+ROUND_SUBKEYS; |
| 400 |
register DWORD *r1=r0 + 2*key->numRounds - 2; |
| 401 |
|
| 402 |
for (;r0 < r1;r0+=2,r1-=2) |
| 403 |
{ |
| 404 |
t0=r0[0]; /* swap the order */ |
| 405 |
t1=r0[1]; |
| 406 |
r0[0]=r1[0]; /* but keep relative order within pairs */ |
| 407 |
r0[1]=r1[1]; |
| 408 |
r1[0]=t0; |
| 409 |
r1[1]=t1; |
| 410 |
} |
| 411 |
|
| 412 |
key->direction=newDir; |
| 413 |
} |
| 414 |
|
| 415 |
/* |
| 416 |
+***************************************************************************** |
| 417 |
* |
| 418 |
* Function Name: Xor256 |
| 419 |
* |
| 420 |
* Function: Copy an 8-bit permutation (256 bytes), xoring with a byte |
| 421 |
* |
| 422 |
* Arguments: dst = where to put result |
| 423 |
* src = where to get data (can be same asa dst) |
| 424 |
* b = byte to xor |
| 425 |
* |
| 426 |
* Return: None |
| 427 |
* |
| 428 |
* Notes: |
| 429 |
* BorlandC's optimization is terrible! When we put the code inline, |
| 430 |
* it generates fairly good code in the *following* segment (not in the Xor256 |
| 431 |
* code itself). If the call is made, the code following the call is awful! |
| 432 |
* The penalty is nearly 50%! So we take the code size hit for inlining for |
| 433 |
* Borland, while Microsoft happily works with a call. |
| 434 |
* |
| 435 |
-****************************************************************************/ |
| 436 |
#if defined(__BORLANDC__) /* do it inline */ |
| 437 |
#define Xor32(dst,src,i) { ((DWORD *)dst)[i] = ((DWORD *)src)[i] ^ tmpX; } |
| 438 |
#define Xor256(dst,src,b) \ |
| 439 |
{ \ |
| 440 |
register DWORD tmpX=0x01010101u * b;\ |
| 441 |
for (i=0;i<64;i+=4) \ |
| 442 |
{ Xor32(dst,src,i ); Xor32(dst,src,i+1); Xor32(dst,src,i+2); Xor32(dst,src,i+3); } \ |
| 443 |
} |
| 444 |
#else /* do it as a function call */ |
| 445 |
static void Xor256(void *dst,void *src,BYTE b) |
| 446 |
{ |
| 447 |
register DWORD x=b*0x01010101u; /* replicate byte to all four bytes */ |
| 448 |
register DWORD *d=(DWORD *)dst; |
| 449 |
register DWORD *s=(DWORD *)src; |
| 450 |
#define X_8(N) { d[N]=s[N] ^ x; d[N+1]=s[N+1] ^ x; } |
| 451 |
#define X_32(N) { X_8(N); X_8(N+2); X_8(N+4); X_8(N+6); } |
| 452 |
X_32(0 ); X_32( 8); X_32(16); X_32(24); /* all inline */ |
| 453 |
d+=32; /* keep offsets small! */ |
| 454 |
s+=32; |
| 455 |
X_32(0 ); X_32( 8); X_32(16); X_32(24); /* all inline */ |
| 456 |
} |
| 457 |
#endif |
| 458 |
|
| 459 |
/* |
| 460 |
+***************************************************************************** |
| 461 |
* |
| 462 |
* Function Name: reKey |
| 463 |
* |
| 464 |
* Function: Initialize the Twofish key schedule from key32 |
| 465 |
* |
| 466 |
* Arguments: key = ptr to keyInstance to be initialized |
| 467 |
* |
| 468 |
* Return: TRUE on success |
| 469 |
* |
| 470 |
* Notes: |
| 471 |
* Here we precompute all the round subkeys, although that is not actually |
| 472 |
* required. For example, on a smartcard, the round subkeys can |
| 473 |
* be generated on-the-fly using f32() |
| 474 |
* |
| 475 |
-****************************************************************************/ |
| 476 |
static int reKey(keyInstance *key) |
| 477 |
{ |
| 478 |
int i,j,k64Cnt,keyLen; |
| 479 |
int subkeyCnt; |
| 480 |
DWORD A=0,B=0,q; |
| 481 |
DWORD sKey[MAX_KEY_BITS/64],k32e[MAX_KEY_BITS/64],k32o[MAX_KEY_BITS/64]; |
| 482 |
BYTE L0[256],L1[256]; /* small local 8-bit permutations */ |
| 483 |
|
| 484 |
#if VALIDATE_PARMS |
| 485 |
#if ALIGN32 |
| 486 |
if (((int)key) & 3) |
| 487 |
return BAD_ALIGN32; |
| 488 |
if ((key->keyLen % 64) || (key->keyLen < MIN_KEY_BITS)) |
| 489 |
return BAD_KEY_INSTANCE; |
| 490 |
#endif |
| 491 |
#endif |
| 492 |
|
| 493 |
if (needToBuildMDS) /* do this one time only */ |
| 494 |
BuildMDS(); |
| 495 |
|
| 496 |
#define F32(res,x,k32) \ |
| 497 |
{ \ |
| 498 |
DWORD t=x; \ |
| 499 |
switch (k64Cnt & 3) \ |
| 500 |
{ \ |
| 501 |
case 0: /* same as 4 */ \ |
| 502 |
b0(t) = p8(04)[b0(t)] ^ b0(k32[3]); \ |
| 503 |
b1(t) = p8(14)[b1(t)] ^ b1(k32[3]); \ |
| 504 |
b2(t) = p8(24)[b2(t)] ^ b2(k32[3]); \ |
| 505 |
b3(t) = p8(34)[b3(t)] ^ b3(k32[3]); \ |
| 506 |
/* fall thru, having pre-processed t */ \ |
| 507 |
case 3: b0(t) = p8(03)[b0(t)] ^ b0(k32[2]); \ |
| 508 |
b1(t) = p8(13)[b1(t)] ^ b1(k32[2]); \ |
| 509 |
b2(t) = p8(23)[b2(t)] ^ b2(k32[2]); \ |
| 510 |
b3(t) = p8(33)[b3(t)] ^ b3(k32[2]); \ |
| 511 |
/* fall thru, having pre-processed t */ \ |
| 512 |
case 2: /* 128-bit keys (optimize for this case) */ \ |
| 513 |
res= MDStab[0][p8(01)[p8(02)[b0(t)] ^ b0(k32[1])] ^ b0(k32[0])] ^ \ |
| 514 |
MDStab[1][p8(11)[p8(12)[b1(t)] ^ b1(k32[1])] ^ b1(k32[0])] ^ \ |
| 515 |
MDStab[2][p8(21)[p8(22)[b2(t)] ^ b2(k32[1])] ^ b2(k32[0])] ^ \ |
| 516 |
MDStab[3][p8(31)[p8(32)[b3(t)] ^ b3(k32[1])] ^ b3(k32[0])] ; \ |
| 517 |
} \ |
| 518 |
} |
| 519 |
|
| 520 |
|
| 521 |
#if !CHECK_TABLE |
| 522 |
#if defined(USE_ASM) /* only do this if not using assember */ |
| 523 |
if (!(useAsm & 4)) |
| 524 |
#endif |
| 525 |
#endif |
| 526 |
{ |
| 527 |
subkeyCnt = ROUND_SUBKEYS + 2*key->numRounds; |
| 528 |
keyLen=key->keyLen; |
| 529 |
k64Cnt=(keyLen+63)/64; /* number of 64-bit key words */ |
| 530 |
for (i=0,j=k64Cnt-1;i<k64Cnt;i++,j--) |
| 531 |
{ /* split into even/odd key dwords */ |
| 532 |
k32e[i]=key->key32[2*i ]; |
| 533 |
k32o[i]=key->key32[2*i+1]; |
| 534 |
/* compute S-box keys using (12,8) Reed-Solomon code over GF(256) */ |
| 535 |
sKey[j]=key->sboxKeys[j]=RS_MDS_Encode(k32e[i],k32o[i]); /* reverse order */ |
| 536 |
} |
| 537 |
} |
| 538 |
|
| 539 |
#ifdef USE_ASM |
| 540 |
if (useAsm & 4) |
| 541 |
{ |
| 542 |
#if defined(COMPILE_KEY) && defined(USE_ASM) |
| 543 |
key->keySig = VALID_SIG; /* show that we are initialized */ |
| 544 |
key->codeSize = sizeof(key->compiledCode); /* set size */ |
| 545 |
#endif |
| 546 |
reKey_86(key); |
| 547 |
} |
| 548 |
else |
| 549 |
#endif |
| 550 |
{ |
| 551 |
for (i=q=0;i<subkeyCnt/2;i++,q+=SK_STEP) |
| 552 |
{ /* compute round subkeys for PHT */ |
| 553 |
F32(A,q ,k32e); /* A uses even key dwords */ |
| 554 |
F32(B,q+SK_BUMP,k32o); /* B uses odd key dwords */ |
| 555 |
B = ROL(B,8); |
| 556 |
key->subKeys[2*i ] = A+B; /* combine with a PHT */ |
| 557 |
B = A + 2*B; |
| 558 |
key->subKeys[2*i+1] = ROL(B,SK_ROTL); |
| 559 |
} |
| 560 |
#if !defined(ZERO_KEY) |
| 561 |
switch (keyLen) /* case out key length for speed in generating S-boxes */ |
| 562 |
{ |
| 563 |
case 128: |
| 564 |
#if defined(FULL_KEY) || defined(PART_KEY) |
| 565 |
#if BIG_TAB |
| 566 |
#define one128(N,J) sbSet(N,i,J,L0[i+J]) |
| 567 |
#define sb128(N) { \ |
| 568 |
BYTE *qq=bigTab[N][b##N(sKey[1])]; \ |
| 569 |
Xor256(L0,qq,b##N(sKey[0])); \ |
| 570 |
for (i=0;i<256;i+=2) { one128(N,0); one128(N,1); } } |
| 571 |
#else |
| 572 |
#define one128(N,J) sbSet(N,i,J,p8(N##1)[L0[i+J]]^k0) |
| 573 |
#define sb128(N) { \ |
| 574 |
Xor256(L0,p8(N##2),b##N(sKey[1])); \ |
| 575 |
{ register DWORD k0=b##N(sKey[0]); \ |
| 576 |
for (i=0;i<256;i+=2) { one128(N,0); one128(N,1); } } } |
| 577 |
#endif |
| 578 |
#elif defined(MIN_KEY) |
| 579 |
#define sb128(N) Xor256(_sBox8_(N),p8(N##2),b##N(sKey[1])) |
| 580 |
#endif |
| 581 |
sb128(0); sb128(1); sb128(2); sb128(3); |
| 582 |
break; |
| 583 |
case 192: |
| 584 |
#if defined(FULL_KEY) || defined(PART_KEY) |
| 585 |
#define one192(N,J) sbSet(N,i,J,p8(N##1)[p8(N##2)[L0[i+J]]^k1]^k0) |
| 586 |
#define sb192(N) { \ |
| 587 |
Xor256(L0,p8(N##3),b##N(sKey[2])); \ |
| 588 |
{ register DWORD k0=b##N(sKey[0]); \ |
| 589 |
register DWORD k1=b##N(sKey[1]); \ |
| 590 |
for (i=0;i<256;i+=2) { one192(N,0); one192(N,1); } } } |
| 591 |
#elif defined(MIN_KEY) |
| 592 |
#define one192(N,J) sbSet(N,i,J,p8(N##2)[L0[i+J]]^k1) |
| 593 |
#define sb192(N) { \ |
| 594 |
Xor256(L0,p8(N##3),b##N(sKey[2])); \ |
| 595 |
{ register DWORD k1=b##N(sKey[1]); \ |
| 596 |
for (i=0;i<256;i+=2) { one192(N,0); one192(N,1); } } } |
| 597 |
#endif |
| 598 |
sb192(0); sb192(1); sb192(2); sb192(3); |
| 599 |
break; |
| 600 |
case 256: |
| 601 |
#if defined(FULL_KEY) || defined(PART_KEY) |
| 602 |
#define one256(N,J) sbSet(N,i,J,p8(N##1)[p8(N##2)[L0[i+J]]^k1]^k0) |
| 603 |
#define sb256(N) { \ |
| 604 |
Xor256(L1,p8(N##4),b##N(sKey[3])); \ |
| 605 |
for (i=0;i<256;i+=2) {L0[i ]=p8(N##3)[L1[i]]; \ |
| 606 |
L0[i+1]=p8(N##3)[L1[i+1]]; } \ |
| 607 |
Xor256(L0,L0,b##N(sKey[2])); \ |
| 608 |
{ register DWORD k0=b##N(sKey[0]); \ |
| 609 |
register DWORD k1=b##N(sKey[1]); \ |
| 610 |
for (i=0;i<256;i+=2) { one256(N,0); one256(N,1); } } } |
| 611 |
#elif defined(MIN_KEY) |
| 612 |
#define one256(N,J) sbSet(N,i,J,p8(N##2)[L0[i+J]]^k1) |
| 613 |
#define sb256(N) { \ |
| 614 |
Xor256(L1,p8(N##4),b##N(sKey[3])); \ |
| 615 |
for (i=0;i<256;i+=2) {L0[i ]=p8(N##3)[L1[i]]; \ |
| 616 |
L0[i+1]=p8(N##3)[L1[i+1]]; } \ |
| 617 |
Xor256(L0,L0,b##N(sKey[2])); \ |
| 618 |
{ register DWORD k1=b##N(sKey[1]); \ |
| 619 |
for (i=0;i<256;i+=2) { one256(N,0); one256(N,1); } } } |
| 620 |
#endif |
| 621 |
sb256(0); sb256(1); sb256(2); sb256(3); |
| 622 |
break; |
| 623 |
} |
| 624 |
#endif |
| 625 |
} |
| 626 |
|
| 627 |
#if CHECK_TABLE /* sanity check vs. pedagogical code*/ |
| 628 |
{ |
| 629 |
GetSboxKey; |
| 630 |
for (i=0;i<subkeyCnt/2;i++) |
| 631 |
{ |
| 632 |
A = f32(i*SK_STEP ,k32e,keyLen); /* A uses even key dwords */ |
| 633 |
B = f32(i*SK_STEP+SK_BUMP,k32o,keyLen); /* B uses odd key dwords */ |
| 634 |
B = ROL(B,8); |
| 635 |
assert(key->subKeys[2*i ] == A+ B); |
| 636 |
assert(key->subKeys[2*i+1] == ROL(A+2*B,SK_ROTL)); |
| 637 |
} |
| 638 |
#if !defined(ZERO_KEY) /* any S-boxes to check? */ |
| 639 |
for (i=q=0;i<256;i++,q+=0x01010101) |
| 640 |
assert(f32(q,key->sboxKeys,keyLen) == Fe32_(q,0)); |
| 641 |
#endif |
| 642 |
} |
| 643 |
#endif /* CHECK_TABLE */ |
| 644 |
|
| 645 |
if (key->direction == DIR_ENCRYPT) |
| 646 |
ReverseRoundSubkeys(key,DIR_ENCRYPT); /* reverse the round subkey order */ |
| 647 |
|
| 648 |
return TRUE; |
| 649 |
} |
| 650 |
/* |
| 651 |
+***************************************************************************** |
| 652 |
* |
| 653 |
* Function Name: makeKey |
| 654 |
* |
| 655 |
* Function: Initialize the Twofish key schedule |
| 656 |
* |
| 657 |
* Arguments: key = ptr to keyInstance to be initialized |
| 658 |
* direction = DIR_ENCRYPT or DIR_DECRYPT |
| 659 |
* keyLen = # bits of key text at *keyMaterial |
| 660 |
* keyMaterial = ptr to hex ASCII chars representing key bits |
| 661 |
* |
| 662 |
* Return: TRUE on success |
| 663 |
* else error code (e.g., BAD_KEY_DIR) |
| 664 |
* |
| 665 |
* Notes: This parses the key bits from keyMaterial. Zeroes out unused key bits |
| 666 |
* |
| 667 |
-****************************************************************************/ |
| 668 |
static int makeKey(keyInstance *key, BYTE direction, int keyLen,CONST char *keyMaterial) |
| 669 |
{ |
| 670 |
int i; |
| 671 |
|
| 672 |
#if VALIDATE_PARMS /* first, sanity check on parameters */ |
| 673 |
if (key == NULL) |
| 674 |
return BAD_KEY_INSTANCE;/* must have a keyInstance to initialize */ |
| 675 |
if ((direction != DIR_ENCRYPT) && (direction != DIR_DECRYPT)) |
| 676 |
return BAD_KEY_DIR; /* must have valid direction */ |
| 677 |
if ((keyLen > MAX_KEY_BITS) || (keyLen < 8) || (keyLen & 0x3F)) |
| 678 |
return BAD_KEY_MAT; /* length must be valid */ |
| 679 |
key->keySig = VALID_SIG; /* show that we are initialized */ |
| 680 |
#if ALIGN32 |
| 681 |
if ((((int)key) & 3) || (((int)key->key32) & 3)) |
| 682 |
return BAD_ALIGN32; |
| 683 |
#endif |
| 684 |
#endif |
| 685 |
|
| 686 |
key->direction = direction;/* set our cipher direction */ |
| 687 |
key->keyLen = (keyLen+63) & ~63; /* round up to multiple of 64 */ |
| 688 |
key->numRounds = numRounds[(keyLen-1)/64]; |
| 689 |
memset(key->key32,0,sizeof(key->key32)); /* zero unused bits */ |
| 690 |
|
| 691 |
if (keyMaterial == NULL) |
| 692 |
return TRUE; /* allow a "dummy" call */ |
| 693 |
|
| 694 |
for (i=0;i<keyLen/32;i++) /* make byte-oriented copy for CFB1 */ |
| 695 |
key->key32[i] = (((unsigned char *)keyMaterial)[i*4+0] << 0) |
| 696 |
| (((unsigned char *)keyMaterial)[i*4+1] << 8) |
| 697 |
| (((unsigned char *)keyMaterial)[i*4+2] << 16) |
| 698 |
| (((unsigned char *)keyMaterial)[i*4+3] << 24); |
| 699 |
|
| 700 |
return reKey(key); /* generate round subkeys */ |
| 701 |
} |
| 702 |
|
| 703 |
|
| 704 |
/* |
| 705 |
+***************************************************************************** |
| 706 |
* |
| 707 |
* Function Name: cipherInit |
| 708 |
* |
| 709 |
* Function: Initialize the Twofish cipher in a given mode |
| 710 |
* |
| 711 |
* Arguments: cipher = ptr to cipherInstance to be initialized |
| 712 |
* mode = MODE_ECB, MODE_CBC, or MODE_CFB1 |
| 713 |
* IV = ptr to hex ASCII test representing IV bytes |
| 714 |
* |
| 715 |
* Return: TRUE on success |
| 716 |
* else error code (e.g., BAD_CIPHER_MODE) |
| 717 |
* |
| 718 |
-****************************************************************************/ |
| 719 |
static int cipherInit(cipherInstance *cipher, BYTE mode,CONST char *IV) |
| 720 |
{ |
| 721 |
int i; |
| 722 |
#if VALIDATE_PARMS /* first, sanity check on parameters */ |
| 723 |
if (cipher == NULL) |
| 724 |
return BAD_PARAMS; /* must have a cipherInstance to initialize */ |
| 725 |
if ((mode != MODE_ECB) && (mode != MODE_CBC) && (mode != MODE_CFB1)) |
| 726 |
return BAD_CIPHER_MODE; /* must have valid cipher mode */ |
| 727 |
cipher->cipherSig = VALID_SIG; |
| 728 |
#if ALIGN32 |
| 729 |
if ((((int)cipher) & 3) || (((int)cipher->IV) & 3) || (((int)cipher->iv32) & 3)) |
| 730 |
return BAD_ALIGN32; |
| 731 |
#endif |
| 732 |
#endif |
| 733 |
|
| 734 |
if ((mode != MODE_ECB) && (IV)) /* parse the IV */ |
| 735 |
{ |
| 736 |
memcpy (cipher->iv32, IV, BLOCK_SIZE/32); |
| 737 |
for (i=0;i<BLOCK_SIZE/32;i++) /* make byte-oriented copy for CFB1 */ |
| 738 |
((DWORD *)cipher->IV)[i] = Bswap(cipher->iv32[i]); |
| 739 |
} |
| 740 |
|
| 741 |
cipher->mode = mode; |
| 742 |
|
| 743 |
return TRUE; |
| 744 |
} |
| 745 |
|
| 746 |
/* |
| 747 |
+***************************************************************************** |
| 748 |
* |
| 749 |
* Function Name: blockEncrypt |
| 750 |
* |
| 751 |
* Function: Encrypt block(s) of data using Twofish |
| 752 |
* |
| 753 |
* Arguments: cipher = ptr to already initialized cipherInstance |
| 754 |
* key = ptr to already initialized keyInstance |
| 755 |
* input = ptr to data blocks to be encrypted |
| 756 |
* inputLen = # bits to encrypt (multiple of blockSize) |
| 757 |
* outBuffer = ptr to where to put encrypted blocks |
| 758 |
* |
| 759 |
* Return: # bits ciphered (>= 0) |
| 760 |
* else error code (e.g., BAD_CIPHER_STATE, BAD_KEY_MATERIAL) |
| 761 |
* |
| 762 |
* Notes: The only supported block size for ECB/CBC modes is BLOCK_SIZE bits. |
| 763 |
* If inputLen is not a multiple of BLOCK_SIZE bits in those modes, |
| 764 |
* an error BAD_INPUT_LEN is returned. In CFB1 mode, all block |
| 765 |
* sizes can be supported. |
| 766 |
* |
| 767 |
-****************************************************************************/ |
| 768 |
static int blockEncrypt(cipherInstance *cipher, keyInstance *key,CONST BYTE *input, |
| 769 |
int inputLen, BYTE *outBuffer) |
| 770 |
{ |
| 771 |
int i,n; /* loop counters */ |
| 772 |
DWORD x[BLOCK_SIZE/32]; /* block being encrypted */ |
| 773 |
DWORD t0,t1; /* temp variables */ |
| 774 |
int rounds=key->numRounds; /* number of rounds */ |
| 775 |
BYTE bit,bit0,ctBit,carry; /* temps for CFB */ |
| 776 |
|
| 777 |
/* make local copies of things for faster access */ |
| 778 |
int mode = cipher->mode; |
| 779 |
DWORD sk[TOTAL_SUBKEYS]; |
| 780 |
DWORD IV[BLOCK_SIZE/32]; |
| 781 |
|
| 782 |
GetSboxKey; |
| 783 |
|
| 784 |
#if VALIDATE_PARMS |
| 785 |
if ((cipher == NULL) || (cipher->cipherSig != VALID_SIG)) |
| 786 |
return BAD_CIPHER_STATE; |
| 787 |
if ((key == NULL) || (key->keySig != VALID_SIG)) |
| 788 |
return BAD_KEY_INSTANCE; |
| 789 |
if ((rounds < 2) || (rounds > MAX_ROUNDS) || (rounds&1)) |
| 790 |
return BAD_KEY_INSTANCE; |
| 791 |
if ((mode != MODE_CFB1) && (inputLen % BLOCK_SIZE)) |
| 792 |
return BAD_INPUT_LEN; |
| 793 |
#if ALIGN32 |
| 794 |
if ( (((int)cipher) & 3) || (((int)key ) & 3) || |
| 795 |
(((int)input ) & 3) || (((int)outBuffer) & 3)) |
| 796 |
return BAD_ALIGN32; |
| 797 |
#endif |
| 798 |
#endif |
| 799 |
|
| 800 |
if (mode == MODE_CFB1) |
| 801 |
{ /* use recursion here to handle CFB, one block at a time */ |
| 802 |
cipher->mode = MODE_ECB; /* do encryption in ECB */ |
| 803 |
for (n=0;n<inputLen;n++) |
| 804 |
{ |
| 805 |
blockEncrypt(cipher,key,cipher->IV,BLOCK_SIZE,(BYTE *)x); |
| 806 |
bit0 = 0x80 >> (n & 7);/* which bit position in byte */ |
| 807 |
ctBit = (input[n/8] & bit0) ^ ((((BYTE *) x)[0] & 0x80) >> (n&7)); |
| 808 |
outBuffer[n/8] = (outBuffer[n/8] & ~ bit0) | ctBit; |
| 809 |
carry = ctBit >> (7 - (n&7)); |
| 810 |
for (i=BLOCK_SIZE/8-1;i>=0;i--) |
| 811 |
{ |
| 812 |
bit = cipher->IV[i] >> 7; /* save next "carry" from shift */ |
| 813 |
cipher->IV[i] = (cipher->IV[i] << 1) ^ carry; |
| 814 |
carry = bit; |
| 815 |
} |
| 816 |
} |
| 817 |
cipher->mode = MODE_CFB1; /* restore mode for next time */ |
| 818 |
return inputLen; |
| 819 |
} |
| 820 |
|
| 821 |
/* here for ECB, CBC modes */ |
| 822 |
if (key->direction != DIR_ENCRYPT) |
| 823 |
ReverseRoundSubkeys(key,DIR_ENCRYPT); /* reverse the round subkey order */ |
| 824 |
|
| 825 |
#ifdef USE_ASM |
| 826 |
if ((useAsm & 1) && (inputLen)) |
| 827 |
#ifdef COMPILE_KEY |
| 828 |
if (key->keySig == VALID_SIG) |
| 829 |
return ((CipherProc *)(key->encryptFuncPtr))(cipher,key,input,inputLen,outBuffer); |
| 830 |
#else |
| 831 |
return (*blockEncrypt_86)(cipher,key,input,inputLen,outBuffer); |
| 832 |
#endif |
| 833 |
#endif |
| 834 |
/* make local copy of subkeys for speed */ |
| 835 |
memcpy(sk,key->subKeys,sizeof(DWORD)*(ROUND_SUBKEYS+2*rounds)); |
| 836 |
if (mode == MODE_CBC) |
| 837 |
BlockCopy(IV,cipher->iv32) |
| 838 |
else |
| 839 |
IV[0]=IV[1]=IV[2]=IV[3]=0; |
| 840 |
|
| 841 |
for (n=0;n<inputLen;n+=BLOCK_SIZE,input+=BLOCK_SIZE/8,outBuffer+=BLOCK_SIZE/8) |
| 842 |
{ |
| 843 |
#define LoadBlockE(N) x[N]=Bswap(((DWORD *)input)[N]) ^ sk[INPUT_WHITEN+N] ^ IV[N] |
| 844 |
LoadBlockE(0); LoadBlockE(1); LoadBlockE(2); LoadBlockE(3); |
| 845 |
#define EncryptRound(K,R,id) \ |
| 846 |
t0 = Fe32##id(x[K ],0); \ |
| 847 |
t1 = Fe32##id(x[K^1],3); \ |
| 848 |
x[K^3] = ROL(x[K^3],1); \ |
| 849 |
x[K^2]^= t0 + t1 + sk[ROUND_SUBKEYS+2*(R) ]; \ |
| 850 |
x[K^3]^= t0 + 2*t1 + sk[ROUND_SUBKEYS+2*(R)+1]; \ |
| 851 |
x[K^2] = ROR(x[K^2],1); |
| 852 |
#define Encrypt2(R,id) { EncryptRound(0,R+1,id); EncryptRound(2,R,id); } |
| 853 |
|
| 854 |
#if defined(ZERO_KEY) |
| 855 |
switch (key->keyLen) |
| 856 |
{ |
| 857 |
case 128: |
| 858 |
for (i=rounds-2;i>=0;i-=2) |
| 859 |
Encrypt2(i,_128); |
| 860 |
break; |
| 861 |
case 192: |
| 862 |
for (i=rounds-2;i>=0;i-=2) |
| 863 |
Encrypt2(i,_192); |
| 864 |
break; |
| 865 |
case 256: |
| 866 |
for (i=rounds-2;i>=0;i-=2) |
| 867 |
Encrypt2(i,_256); |
| 868 |
break; |
| 869 |
} |
| 870 |
#else |
| 871 |
Encrypt2(14,_); |
| 872 |
Encrypt2(12,_); |
| 873 |
Encrypt2(10,_); |
| 874 |
Encrypt2( 8,_); |
| 875 |
Encrypt2( 6,_); |
| 876 |
Encrypt2( 4,_); |
| 877 |
Encrypt2( 2,_); |
| 878 |
Encrypt2( 0,_); |
| 879 |
#endif |
| 880 |
|
| 881 |
/* need to do (or undo, depending on your point of view) final swap */ |
| 882 |
#if LittleEndian |
| 883 |
#define StoreBlockE(N) ((DWORD *)outBuffer)[N]=x[N^2] ^ sk[OUTPUT_WHITEN+N] |
| 884 |
#else |
| 885 |
#define StoreBlockE(N) { t0=x[N^2] ^ sk[OUTPUT_WHITEN+N]; ((DWORD *)outBuffer)[N]=Bswap(t0); } |
| 886 |
#endif |
| 887 |
StoreBlockE(0); StoreBlockE(1); StoreBlockE(2); StoreBlockE(3); |
| 888 |
if (mode == MODE_CBC) |
| 889 |
{ |
| 890 |
IV[0]=Bswap(((DWORD *)outBuffer)[0]); |
| 891 |
IV[1]=Bswap(((DWORD *)outBuffer)[1]); |
| 892 |
IV[2]=Bswap(((DWORD *)outBuffer)[2]); |
| 893 |
IV[3]=Bswap(((DWORD *)outBuffer)[3]); |
| 894 |
} |
| 895 |
} |
| 896 |
|
| 897 |
if (mode == MODE_CBC) |
| 898 |
BlockCopy(cipher->iv32,IV); |
| 899 |
|
| 900 |
return inputLen; |
| 901 |
} |
| 902 |
|
| 903 |
/* |
| 904 |
+***************************************************************************** |
| 905 |
* |
| 906 |
* Function Name: blockDecrypt |
| 907 |
* |
| 908 |
* Function: Decrypt block(s) of data using Twofish |
| 909 |
* |
| 910 |
* Arguments: cipher = ptr to already initialized cipherInstance |
| 911 |
* key = ptr to already initialized keyInstance |
| 912 |
* input = ptr to data blocks to be decrypted |
| 913 |
* inputLen = # bits to encrypt (multiple of blockSize) |
| 914 |
* outBuffer = ptr to where to put decrypted blocks |
| 915 |
* |
| 916 |
* Return: # bits ciphered (>= 0) |
| 917 |
* else error code (e.g., BAD_CIPHER_STATE, BAD_KEY_MATERIAL) |
| 918 |
* |
| 919 |
* Notes: The only supported block size for ECB/CBC modes is BLOCK_SIZE bits. |
| 920 |
* If inputLen is not a multiple of BLOCK_SIZE bits in those modes, |
| 921 |
* an error BAD_INPUT_LEN is returned. In CFB1 mode, all block |
| 922 |
* sizes can be supported. |
| 923 |
* |
| 924 |
-****************************************************************************/ |
| 925 |
static int blockDecrypt(cipherInstance *cipher, keyInstance *key,CONST BYTE *input, |
| 926 |
int inputLen, BYTE *outBuffer) |
| 927 |
{ |
| 928 |
int i,n; /* loop counters */ |
| 929 |
DWORD x[BLOCK_SIZE/32]; /* block being encrypted */ |
| 930 |
DWORD t0,t1; /* temp variables */ |
| 931 |
int rounds=key->numRounds; /* number of rounds */ |
| 932 |
BYTE bit,bit0,ctBit,carry; /* temps for CFB */ |
| 933 |
|
| 934 |
/* make local copies of things for faster access */ |
| 935 |
int mode = cipher->mode; |
| 936 |
DWORD sk[TOTAL_SUBKEYS]; |
| 937 |
DWORD IV[BLOCK_SIZE/32]; |
| 938 |
|
| 939 |
GetSboxKey; |
| 940 |
|
| 941 |
#if VALIDATE_PARMS |
| 942 |
if ((cipher == NULL) || (cipher->cipherSig != VALID_SIG)) |
| 943 |
return BAD_CIPHER_STATE; |
| 944 |
if ((key == NULL) || (key->keySig != VALID_SIG)) |
| 945 |
return BAD_KEY_INSTANCE; |
| 946 |
if ((rounds < 2) || (rounds > MAX_ROUNDS) || (rounds&1)) |
| 947 |
return BAD_KEY_INSTANCE; |
| 948 |
if ((cipher->mode != MODE_CFB1) && (inputLen % BLOCK_SIZE)) |
| 949 |
return BAD_INPUT_LEN; |
| 950 |
#if ALIGN32 |
| 951 |
if ( (((int)cipher) & 3) || (((int)key ) & 3) || |
| 952 |
(((int)input) & 3) || (((int)outBuffer) & 3)) |
| 953 |
return BAD_ALIGN32; |
| 954 |
#endif |
| 955 |
#endif |
| 956 |
|
| 957 |
if (cipher->mode == MODE_CFB1) |
| 958 |
{ /* use blockEncrypt here to handle CFB, one block at a time */ |
| 959 |
cipher->mode = MODE_ECB; /* do encryption in ECB */ |
| 960 |
for (n=0;n<inputLen;n++) |
| 961 |
{ |
| 962 |
blockEncrypt(cipher,key,cipher->IV,BLOCK_SIZE,(BYTE *)x); |
| 963 |
bit0 = 0x80 >> (n & 7); |
| 964 |
ctBit = input[n/8] & bit0; |
| 965 |
outBuffer[n/8] = (outBuffer[n/8] & ~ bit0) | |
| 966 |
(ctBit ^ ((((BYTE *) x)[0] & 0x80) >> (n&7))); |
| 967 |
carry = ctBit >> (7 - (n&7)); |
| 968 |
for (i=BLOCK_SIZE/8-1;i>=0;i--) |
| 969 |
{ |
| 970 |
bit = cipher->IV[i] >> 7; /* save next "carry" from shift */ |
| 971 |
cipher->IV[i] = (cipher->IV[i] << 1) ^ carry; |
| 972 |
carry = bit; |
| 973 |
} |
| 974 |
} |
| 975 |
cipher->mode = MODE_CFB1; /* restore mode for next time */ |
| 976 |
return inputLen; |
| 977 |
} |
| 978 |
|
| 979 |
/* here for ECB, CBC modes */ |
| 980 |
if (key->direction != DIR_DECRYPT) |
| 981 |
ReverseRoundSubkeys(key,DIR_DECRYPT); /* reverse the round subkey order */ |
| 982 |
#ifdef USE_ASM |
| 983 |
if ((useAsm & 2) && (inputLen)) |
| 984 |
#ifdef COMPILE_KEY |
| 985 |
if (key->keySig == VALID_SIG) |
| 986 |
return ((CipherProc *)(key->decryptFuncPtr))(cipher,key,input,inputLen,outBuffer); |
| 987 |
#else |
| 988 |
return (*blockDecrypt_86)(cipher,key,input,inputLen,outBuffer); |
| 989 |
#endif |
| 990 |
#endif |
| 991 |
/* make local copy of subkeys for speed */ |
| 992 |
memcpy(sk,key->subKeys,sizeof(DWORD)*(ROUND_SUBKEYS+2*rounds)); |
| 993 |
if (mode == MODE_CBC) |
| 994 |
BlockCopy(IV,cipher->iv32) |
| 995 |
else |
| 996 |
IV[0]=IV[1]=IV[2]=IV[3]=0; |
| 997 |
|
| 998 |
for (n=0;n<inputLen;n+=BLOCK_SIZE,input+=BLOCK_SIZE/8,outBuffer+=BLOCK_SIZE/8) |
| 999 |
{ |
| 1000 |
#define LoadBlockD(N) x[N^2]=Bswap(((DWORD *)input)[N]) ^ sk[OUTPUT_WHITEN+N] |
| 1001 |
LoadBlockD(0); LoadBlockD(1); LoadBlockD(2); LoadBlockD(3); |
| 1002 |
|
| 1003 |
#define DecryptRound(K,R,id) \ |
| 1004 |
t0 = Fe32##id(x[K ],0); \ |
| 1005 |
t1 = Fe32##id(x[K^1],3); \ |
| 1006 |
x[K^2] = ROL (x[K^2],1); \ |
| 1007 |
x[K^2]^= t0 + t1 + sk[ROUND_SUBKEYS+2*(R) ]; \ |
| 1008 |
x[K^3]^= t0 + 2*t1 + sk[ROUND_SUBKEYS+2*(R)+1]; \ |
| 1009 |
x[K^3] = ROR (x[K^3],1); |
| 1010 |
|
| 1011 |
#define Decrypt2(R,id) { DecryptRound(2,R+1,id); DecryptRound(0,R,id); } |
| 1012 |
|
| 1013 |
#if defined(ZERO_KEY) |
| 1014 |
switch (key->keyLen) |
| 1015 |
{ |
| 1016 |
case 128: |
| 1017 |
for (i=rounds-2;i>=0;i-=2) |
| 1018 |
Decrypt2(i,_128); |
| 1019 |
break; |
| 1020 |
case 192: |
| 1021 |
for (i=rounds-2;i>=0;i-=2) |
| 1022 |
Decrypt2(i,_192); |
| 1023 |
break; |
| 1024 |
case 256: |
| 1025 |
for (i=rounds-2;i>=0;i-=2) |
| 1026 |
Decrypt2(i,_256); |
| 1027 |
break; |
| 1028 |
} |
| 1029 |
#else |
| 1030 |
{ |
| 1031 |
Decrypt2(14,_); |
| 1032 |
Decrypt2(12,_); |
| 1033 |
Decrypt2(10,_); |
| 1034 |
Decrypt2( 8,_); |
| 1035 |
Decrypt2( 6,_); |
| 1036 |
Decrypt2( 4,_); |
| 1037 |
Decrypt2( 2,_); |
| 1038 |
Decrypt2( 0,_); |
| 1039 |
} |
| 1040 |
#endif |
| 1041 |
if (cipher->mode == MODE_ECB) |
| 1042 |
{ |
| 1043 |
#if LittleEndian |
| 1044 |
#define StoreBlockD(N) ((DWORD *)outBuffer)[N] = x[N] ^ sk[INPUT_WHITEN+N] |
| 1045 |
#else |
| 1046 |
#define StoreBlockD(N) { t0=x[N]^sk[INPUT_WHITEN+N]; ((DWORD *)outBuffer)[N] = Bswap(t0); } |
| 1047 |
#endif |
| 1048 |
StoreBlockD(0); StoreBlockD(1); StoreBlockD(2); StoreBlockD(3); |
| 1049 |
#undef StoreBlockD |
| 1050 |
continue; |
| 1051 |
} |
| 1052 |
else |
| 1053 |
{ |
| 1054 |
#define StoreBlockD(N) x[N] ^= sk[INPUT_WHITEN+N] ^ IV[N]; \ |
| 1055 |
IV[N] = Bswap(((DWORD *)input)[N]); \ |
| 1056 |
((DWORD *)outBuffer)[N] = Bswap(x[N]); |
| 1057 |
StoreBlockD(0); StoreBlockD(1); StoreBlockD(2); StoreBlockD(3); |
| 1058 |
#undef StoreBlockD |
| 1059 |
} |
| 1060 |
} |
| 1061 |
if (mode == MODE_CBC) /* restore iv32 to cipher */ |
| 1062 |
BlockCopy(cipher->iv32,IV) |
| 1063 |
|
| 1064 |
return inputLen; |
| 1065 |
} |
| 1066 |
|
| 1067 |
#ifdef GetCodeSize |
| 1068 |
static DWORD TwofishCodeSize(void) |
| 1069 |
{ |
| 1070 |
DWORD x= Here(0); |
| 1071 |
#ifdef USE_ASM |
| 1072 |
if (useAsm & 3) |
| 1073 |
return TwofishAsmCodeSize(); |
| 1074 |
#endif |
| 1075 |
return x - TwofishCodeStart(); |
| 1076 |
}; |
| 1077 |
#endif |
