|
|
1.1 root 1: /*
2: ---------------------------------------------------------------------------
3: Copyright (c) 2003, Dr Brian Gladman, Worcester, UK. All rights reserved.
4:
5: LICENSE TERMS
6:
7: The free distribution and use of this software in both source and binary
8: form is allowed (with or without changes) provided that:
9:
10: 1. distributions of this source code include the above copyright
11: notice, this list of conditions and the following disclaimer;
12:
13: 2. distributions in binary form include the above copyright
14: notice, this list of conditions and the following disclaimer
15: in the documentation and/or other associated materials;
16:
17: 3. the copyright holder's name is not used to endorse products
18: built using this software without specific written permission.
19:
20: ALTERNATIVELY, provided that this notice is retained in full, this product
21: may be distributed under the terms of the GNU General Public License (GPL),
22: in which case the provisions of the GPL apply INSTEAD OF those given above.
23:
24: DISCLAIMER
25:
26: This software is provided 'as is' with no explicit or implied warranties
27: in respect of its properties, including, but not limited to, correctness
28: and/or fitness for purpose.
29: ---------------------------------------------------------------------------
1.1.1.3 ! root 30: Issue 01/08/2005
1.1 root 31:
32: This file contains the compilation options for AES (Rijndael) and code
33: that is common across encryption, key scheduling and table generation.
34:
35: OPERATION
36:
37: These source code files implement the AES algorithm Rijndael designed by
38: Joan Daemen and Vincent Rijmen. This version is designed for the standard
39: block size of 16 bytes and for key sizes of 128, 192 and 256 bits (16, 24
40: and 32 bytes).
41:
42: This version is designed for flexibility and speed using operations on
43: 32-bit words rather than operations on bytes. It can be compiled with
44: either big or little endian internal byte order but is faster when the
45: native byte order for the processor is used.
46:
47: THE CIPHER INTERFACE
48:
49: The cipher interface is implemented as an array of bytes in which lower
50: AES bit sequence indexes map to higher numeric significance within bytes.
51:
1.1.1.3 ! root 52: uint_8t (an unsigned 8-bit type)
! 53: uint_32t (an unsigned 32-bit type)
1.1 root 54: struct aes_encrypt_ctx (structure for the cipher encryption context)
55: struct aes_decrypt_ctx (structure for the cipher decryption context)
56: aes_rval the function return type
57:
58: C subroutine calls:
59:
60: aes_rval aes_encrypt_key128(const unsigned char *key, aes_encrypt_ctx cx[1]);
61: aes_rval aes_encrypt_key192(const unsigned char *key, aes_encrypt_ctx cx[1]);
62: aes_rval aes_encrypt_key256(const unsigned char *key, aes_encrypt_ctx cx[1]);
63: aes_rval aes_encrypt(const unsigned char *in, unsigned char *out,
64: const aes_encrypt_ctx cx[1]);
65:
66: aes_rval aes_decrypt_key128(const unsigned char *key, aes_decrypt_ctx cx[1]);
67: aes_rval aes_decrypt_key192(const unsigned char *key, aes_decrypt_ctx cx[1]);
68: aes_rval aes_decrypt_key256(const unsigned char *key, aes_decrypt_ctx cx[1]);
69: aes_rval aes_decrypt(const unsigned char *in, unsigned char *out,
70: const aes_decrypt_ctx cx[1]);
71:
72: IMPORTANT NOTE: If you are using this C interface with dynamic tables make sure that
73: you call genTabs() before AES is used so that the tables are initialised.
74:
75: C++ aes class subroutines:
76:
77: Class AESencrypt for encryption
78:
79: Construtors:
80: AESencrypt(void)
81: AESencrypt(const unsigned char *key) - 128 bit key
82: Members:
83: aes_rval key128(const unsigned char *key)
84: aes_rval key192(const unsigned char *key)
85: aes_rval key256(const unsigned char *key)
86: aes_rval encrypt(const unsigned char *in, unsigned char *out) const
87:
88: Class AESdecrypt for encryption
89: Construtors:
90: AESdecrypt(void)
91: AESdecrypt(const unsigned char *key) - 128 bit key
92: Members:
93: aes_rval key128(const unsigned char *key)
94: aes_rval key192(const unsigned char *key)
95: aes_rval key256(const unsigned char *key)
96: aes_rval decrypt(const unsigned char *in, unsigned char *out) const
97:
98: COMPILATION
99:
100: The files used to provide AES (Rijndael) are
101:
102: a. aes.h for the definitions needed for use in C.
103: b. aescpp.h for the definitions needed for use in C++.
104: c. aesopt.h for setting compilation options (also includes common code).
105: d. aescrypt.c for encryption and decrytpion, or
106: e. aeskey.c for key scheduling.
107: f. aestab.c for table loading or generation.
108: g. aescrypt.asm for encryption and decryption using assembler code.
109: h. aescrypt.mmx.asm for encryption and decryption using MMX assembler.
110:
111: To compile AES (Rijndael) for use in C code use aes.h and set the
112: defines here for the facilities you need (key lengths, encryption
1.1.1.3 ! root 113: and/or decryption). Do not define BUILD_DLL or AES_CPP. Set the options
1.1 root 114: for optimisations and table sizes here.
115:
116: To compile AES (Rijndael) for use in in C++ code use aescpp.h but do
1.1.1.3 ! root 117: not define BUILD_DLL
1.1 root 118:
119: To compile AES (Rijndael) in C as a Dynamic Link Library DLL) use
1.1.1.3 ! root 120: aes.h and include the BUILD_DLL define.
1.1 root 121:
122: CONFIGURATION OPTIONS (here and in aes.h)
123:
1.1.1.3 ! root 124: a. set BUILD_DLL in aes.h if AES (Rijndael) is to be compiled as a DLL
1.1 root 125: b. You may need to set PLATFORM_BYTE_ORDER to define the byte order.
126: c. If you want the code to run in a specific internal byte order, then
127: ALGORITHM_BYTE_ORDER must be set accordingly.
128: d. set other configuration options decribed below.
129: */
130:
1.1.1.3 ! root 131: /* Adapted for TrueCrypt by the TrueCrypt Foundation */
! 132:
1.1 root 133: #if !defined( _AESOPT_H )
134: #define _AESOPT_H
135:
1.1.1.3 ! root 136: #if defined( __cplusplus ) && defined( AES_CPP )
! 137: #include "aescpp.h"
! 138: #else
1.1.1.2 root 139: #include "Aes.h"
1.1.1.3 ! root 140: #endif
1.1 root 141:
1.1.1.3 ! root 142: /* PLATFORM SPECIFIC INCLUDES */
1.1 root 143:
1.1.1.2 root 144: #include "Endian.h"
1.1.1.3 ! root 145: #define IS_LITTLE_ENDIAN 1234 /* byte 0 is least significant (i386) */
! 146: #define IS_BIG_ENDIAN 4321 /* byte 0 is most significant (mc68k) */
1.1 root 147:
1.1.1.2 root 148: #if BYTE_ORDER == LITTLE_ENDIAN
1.1.1.3 ! root 149: # define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
1.1 root 150: #endif
151:
1.1.1.2 root 152: #if BYTE_ORDER == BIG_ENDIAN
1.1.1.3 ! root 153: # define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
1.1 root 154: #endif
155:
156:
1.1.1.3 ! root 157: /* CONFIGURATION - THE USE OF DEFINES
1.1 root 158:
1.1.1.3 ! root 159: Later in this section there are a number of defines that control the
! 160: operation of the code. In each section, the purpose of each define is
! 161: explained so that the relevant form can be included or excluded by
! 162: setting either 1's or 0's respectively on the branches of the related
! 163: #if clauses. The following local defines should not be changed.
1.1 root 164: */
165:
1.1.1.3 ! root 166: #define ENCRYPTION_IN_C 1
! 167: #define DECRYPTION_IN_C 2
! 168: #define ENC_KEYING_IN_C 4
! 169: #define DEC_KEYING_IN_C 8
! 170:
! 171: #define ENCRYPTION_IN_ASM 16
! 172: #define DECRYPTION_IN_ASM 32
! 173: #define ENC_KEYING_IN_ASM 64
! 174: #define DEC_KEYING_IN_ASM 128
! 175:
! 176: #define NO_TABLES 0
! 177: #define ONE_TABLE 1
! 178: #define FOUR_TABLES 4
! 179: #define NONE 0
! 180: #define PARTIAL 1
! 181: #define FULL 2
1.1 root 182:
1.1.1.3 ! root 183: /* 1. BYTE ORDER WITHIN 32 BIT WORDS
1.1 root 184:
185: The fundamental data processing units in Rijndael are 8-bit bytes. The
186: input, output and key input are all enumerated arrays of bytes in which
187: bytes are numbered starting at zero and increasing to one less than the
188: number of bytes in the array in question. This enumeration is only used
189: for naming bytes and does not imply any adjacency or order relationship
190: from one byte to another. When these inputs and outputs are considered
191: as bit sequences, bits 8*n to 8*n+7 of the bit sequence are mapped to
192: byte[n] with bit 8n+i in the sequence mapped to bit 7-i within the byte.
193: In this implementation bits are numbered from 0 to 7 starting at the
194: numerically least significant end of each byte (bit n represents 2^n).
195:
196: However, Rijndael can be implemented more efficiently using 32-bit
197: words by packing bytes into words so that bytes 4*n to 4*n+3 are placed
198: into word[n]. While in principle these bytes can be assembled into words
199: in any positions, this implementation only supports the two formats in
200: which bytes in adjacent positions within words also have adjacent byte
201: numbers. This order is called big-endian if the lowest numbered bytes
202: in words have the highest numeric significance and little-endian if the
203: opposite applies.
204:
205: This code can work in either order irrespective of the order used by the
206: machine on which it runs. Normally the internal byte order will be set
207: to the order of the processor on which the code is to be run but this
208: define can be used to reverse this in special situations
209:
1.1.1.3 ! root 210: WARNING: Assembler code versions rely on PLATFORM_BYTE_ORDER being set.
! 211: This define will hence be redefined later (in section 4) if necessary
1.1 root 212: */
1.1.1.3 ! root 213:
! 214: #if 1
1.1 root 215: #define ALGORITHM_BYTE_ORDER PLATFORM_BYTE_ORDER
216: #elif 0
1.1.1.3 ! root 217: #define ALGORITHM_BYTE_ORDER IS_LITTLE_ENDIAN
1.1 root 218: #elif 0
1.1.1.3 ! root 219: #define ALGORITHM_BYTE_ORDER IS_BIG_ENDIAN
1.1 root 220: #else
221: #error The algorithm byte order is not defined
222: #endif
223:
1.1.1.3 ! root 224: /* 2. VIA ACE SUPPORT
! 225:
! 226: Define this option if support for the VIA ACE is required. This uses
! 227: inline assembler instructions and is only implemented for the Microsoft,
! 228: Intel and GCC compilers. If VIA ACE is known to be present, then defining
! 229: ASSUME_VIA_ACE_PRESENT will remove the ordinary encryption/decryption
! 230: code. If USE_VIA_ACE_IF_PRESENT is defined then VIA ACE will be used if
! 231: it is detected (both present and enabled) but the normal AES code will
! 232: also be present.
! 233:
! 234: When VIA ACE is to be used, all AES encryption contexts MUST be 16 byte
! 235: aligned; other input/output buffers do not need to be 16 byte aligned
! 236: but there are very large performance gains if this can be arranged.
! 237: VIA ACE also requires the decryption key schedule to be in reverse
! 238: order (which the following defines ensure).
! 239: */
! 240:
! 241: #if 0 && !defined( USE_VIA_ACE_IF_PRESENT )
! 242: #define USE_VIA_ACE_IF_PRESENT
! 243: #endif
! 244:
! 245: #if 0 && !defined( ASSUME_VIA_ACE_PRESENT )
! 246: #define ASSUME_VIA_ACE_PRESENT
! 247: #endif
! 248:
! 249: #if !defined( _MSC_VER ) && !defined( __GNUC__ )
! 250: # if defined( ASSUME_VIA_ACE_PRESENT )
! 251: # undef ASSUME_VIA_ACE_PRESENT
! 252: # endif
! 253: # if defined( USE_VIA_ACE_IF_PRESENT )
! 254: # undef USE_VIA_ACE_IF_PRESENT
! 255: # endif
! 256: #endif
! 257:
! 258: #if defined( ASSUME_VIA_ACE_PRESENT ) && !defined( USE_VIA_ACE_IF_PRESENT )
! 259: #define USE_VIA_ACE_IF_PRESENT
! 260: #endif
! 261:
! 262: #if defined( USE_VIA_ACE_IF_PRESENT ) && !defined ( AES_REV_DKS )
! 263: #define AES_REV_DKS
! 264: #endif
! 265:
! 266: /* 3. ASSEMBLER SUPPORT
! 267:
! 268: This define (which can be on the command line) enables the use of the
! 269: assembler code routines for encryption, decryption and key scheduling
! 270: as follows:
! 271:
! 272: ASM_V1 uses the assembler (aescrypt1.asm) for large tables with
! 273: tables and key scheduling in C
! 274: ASM_V2 uses assembler (aescrypt2.asm) with compressed tables
! 275: and key scheduling
! 276: ASM_V2C uses assembler (aescrypt2.asm) with compressed tables
! 277: but uses key scheduling in C
! 278: */
! 279:
! 280: #if 0 && !defined( ASM_V1 )
! 281: #define ASM_V1
! 282: #elif 0 && !defined( ASM_V2 )
! 283: #define ASM_V2
! 284: #elif 0 && !defined( ASM_V2C )
! 285: #define ASM_V2C
! 286: #endif
! 287:
! 288: #if defined( ASM_V1 ) && (ALGORITHM_BYTE_ORDER != PLATFORM_BYTE_ORDER)
! 289: #undef ALGORITHM_BYTE_ORDER
! 290: #define ALGORITHM_BYTE_ORDER PLATFORM_BYTE_ORDER
! 291: #endif
! 292:
! 293: /* 4. FUNCTIONS REQUIRED
! 294:
! 295: This implementation provides subroutines for encryption, decryption
! 296: and for setting the three key lengths (separately) for encryption
! 297: and decryption. When the assembler code is not being used the following
! 298: definition blocks allow the selection of the routines that are to be
! 299: included in the compilation.
! 300: */
! 301:
! 302: #if !defined( AES_ENCRYPT )
! 303: # define EFUNCS_IN_C 0
! 304: #elif defined( USE_VIA_ACE_IF_PRESENT ) || defined( ASM_V1 )
! 305: # define EFUNCS_IN_C ENC_KEYING_IN_C
! 306: #elif defined( ASM_V2C )
! 307: # define EFUNCS_IN_C ENC_KEYING_IN_C
! 308: #elif !defined( ASM_V2 )
! 309: # define EFUNCS_IN_C ( ENCRYPTION_IN_C | ENC_KEYING_IN_C )
! 310: #else
! 311: # define EFUNCS_IN_C 0
! 312: #endif
! 313:
! 314: #if !defined( AES_DECRYPT )
! 315: # define DFUNCS_IN_C 0
! 316: #elif defined( USE_VIA_ACE_IF_PRESENT ) || defined( ASM_V1 )
! 317: # define DFUNCS_IN_C DEC_KEYING_IN_C
! 318: #elif defined( ASM_V2C )
! 319: # define DFUNCS_IN_C DEC_KEYING_IN_C
! 320: #elif !defined( ASM_V2 )
! 321: # define DFUNCS_IN_C ( DECRYPTION_IN_C | DEC_KEYING_IN_C )
! 322: #else
! 323: # define DFUNCS_IN_C 0
! 324: #endif
! 325:
! 326: #define FUNCS_IN_C ( EFUNCS_IN_C | DFUNCS_IN_C )
! 327:
! 328: /* 5. FAST INPUT/OUTPUT OPERATIONS.
1.1 root 329:
330: On some machines it is possible to improve speed by transferring the
331: bytes in the input and output arrays to and from the internal 32-bit
332: variables by addressing these arrays as if they are arrays of 32-bit
333: words. On some machines this will always be possible but there may
334: be a large performance penalty if the byte arrays are not aligned on
335: the normal word boundaries. On other machines this technique will
336: lead to memory access errors when such 32-bit word accesses are not
337: properly aligned. The option SAFE_IO avoids such problems but will
338: often be slower on those machines that support misaligned access
339: (especially so if care is taken to align the input and output byte
340: arrays on 32-bit word boundaries). If SAFE_IO is not defined it is
341: assumed that access to byte arrays as if they are arrays of 32-bit
342: words will not cause problems when such accesses are misaligned.
343: */
344: #if 1 && !defined(_MSC_VER)
345: #define SAFE_IO
346: #endif
347:
1.1.1.3 ! root 348: /* 6. LOOP UNROLLING
1.1 root 349:
350: The code for encryption and decrytpion cycles through a number of rounds
351: that can be implemented either in a loop or by expanding the code into a
352: long sequence of instructions, the latter producing a larger program but
353: one that will often be much faster. The latter is called loop unrolling.
354: There are also potential speed advantages in expanding two iterations in
355: a loop with half the number of iterations, which is called partial loop
356: unrolling. The following options allow partial or full loop unrolling
357: to be set independently for encryption and decryption
358: */
359: #if 1
360: #define ENC_UNROLL FULL
361: #elif 0
362: #define ENC_UNROLL PARTIAL
363: #else
364: #define ENC_UNROLL NONE
365: #endif
366:
367: #if 1
368: #define DEC_UNROLL FULL
369: #elif 0
370: #define DEC_UNROLL PARTIAL
371: #else
372: #define DEC_UNROLL NONE
373: #endif
374:
1.1.1.3 ! root 375: /* 7. FAST FINITE FIELD OPERATIONS
1.1 root 376:
377: If this section is included, tables are used to provide faster finite
378: field arithmetic (this has no effect if FIXED_TABLES is defined).
379: */
380: #if 1
381: #define FF_TABLES
382: #endif
383:
1.1.1.3 ! root 384: /* 8. INTERNAL STATE VARIABLE FORMAT
1.1 root 385:
386: The internal state of Rijndael is stored in a number of local 32-bit
387: word varaibles which can be defined either as an array or as individual
388: names variables. Include this section if you want to store these local
389: varaibles in arrays. Otherwise individual local variables will be used.
390: */
391: #if 1
392: #define ARRAYS
393: #endif
394:
395: /* In this implementation the columns of the state array are each held in
396: 32-bit words. The state array can be held in various ways: in an array
397: of words, in a number of individual word variables or in a number of
398: processor registers. The following define maps a variable name x and
399: a column number c to the way the state array variable is to be held.
400: The first define below maps the state into an array x[c] whereas the
401: second form maps the state into a number of individual variables x0,
402: x1, etc. Another form could map individual state colums to machine
403: register names.
404: */
405:
406: #if defined(ARRAYS)
407: #define s(x,c) x[c]
408: #else
409: #define s(x,c) x##c
410: #endif
411:
1.1.1.3 ! root 412: /* 9. FIXED OR DYNAMIC TABLES
1.1 root 413:
414: When this section is included the tables used by the code are compiled
415: statically into the binary file. Otherwise the subroutine gen_tabs()
416: must be called to compute them before the code is first used.
417: */
418: #if 1
419: #define FIXED_TABLES
420: #endif
421:
1.1.1.3 ! root 422: /* 10. TABLE ALIGNMENT
1.1 root 423:
424: On some sytsems speed will be improved by aligning the AES large lookup
425: tables on particular boundaries. This define should be set to a power of
426: two giving the desired alignment. It can be left undefined if alignment
427: is not needed. This option is specific to the Microsft VC++ compiler -
428: it seems to sometimes cause trouble for the VC++ version 6 compiler.
429: */
430:
1.1.1.3 ! root 431: #if 1 && defined(_MSC_VER) && (_MSC_VER >= 1300)
! 432: #define TABLE_ALIGN 32
1.1 root 433: #endif
434:
1.1.1.3 ! root 435: /* 11. INTERNAL TABLE CONFIGURATION
1.1 root 436:
437: This cipher proceeds by repeating in a number of cycles known as 'rounds'
438: which are implemented by a round function which can optionally be speeded
439: up using tables. The basic tables are each 256 32-bit words, with either
440: one or four tables being required for each round function depending on
441: how much speed is required. The encryption and decryption round functions
442: are different and the last encryption and decrytpion round functions are
443: different again making four different round functions in all.
444:
445: This means that:
446: 1. Normal encryption and decryption rounds can each use either 0, 1
447: or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
448: 2. The last encryption and decryption rounds can also use either 0, 1
449: or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
450:
451: Include or exclude the appropriate definitions below to set the number
452: of tables used by this implementation.
453: */
454:
455: #if 1 /* set tables for the normal encryption round */
456: #define ENC_ROUND FOUR_TABLES
457: #elif 0
458: #define ENC_ROUND ONE_TABLE
459: #else
460: #define ENC_ROUND NO_TABLES
461: #endif
462:
463: #if 1 /* set tables for the last encryption round */
464: #define LAST_ENC_ROUND FOUR_TABLES
465: #elif 0
466: #define LAST_ENC_ROUND ONE_TABLE
467: #else
468: #define LAST_ENC_ROUND NO_TABLES
469: #endif
470:
471: #if 1 /* set tables for the normal decryption round */
472: #define DEC_ROUND FOUR_TABLES
473: #elif 0
474: #define DEC_ROUND ONE_TABLE
475: #else
476: #define DEC_ROUND NO_TABLES
477: #endif
478:
479: #if 1 /* set tables for the last decryption round */
480: #define LAST_DEC_ROUND FOUR_TABLES
481: #elif 0
482: #define LAST_DEC_ROUND ONE_TABLE
483: #else
484: #define LAST_DEC_ROUND NO_TABLES
485: #endif
486:
487: /* The decryption key schedule can be speeded up with tables in the same
488: way that the round functions can. Include or exclude the following
489: defines to set this requirement.
490: */
491: #if 1
492: #define KEY_SCHED FOUR_TABLES
493: #elif 0
494: #define KEY_SCHED ONE_TABLE
495: #else
496: #define KEY_SCHED NO_TABLES
497: #endif
498:
499: /* END OF CONFIGURATION OPTIONS */
500:
501: #define RC_LENGTH (5 * (AES_BLOCK_SIZE / 4 - 2))
502:
503: /* Disable or report errors on some combinations of options */
504:
505: #if ENC_ROUND == NO_TABLES && LAST_ENC_ROUND != NO_TABLES
506: #undef LAST_ENC_ROUND
507: #define LAST_ENC_ROUND NO_TABLES
508: #elif ENC_ROUND == ONE_TABLE && LAST_ENC_ROUND == FOUR_TABLES
509: #undef LAST_ENC_ROUND
510: #define LAST_ENC_ROUND ONE_TABLE
511: #endif
512:
513: #if ENC_ROUND == NO_TABLES && ENC_UNROLL != NONE
514: #undef ENC_UNROLL
515: #define ENC_UNROLL NONE
516: #endif
517:
518: #if DEC_ROUND == NO_TABLES && LAST_DEC_ROUND != NO_TABLES
519: #undef LAST_DEC_ROUND
520: #define LAST_DEC_ROUND NO_TABLES
521: #elif DEC_ROUND == ONE_TABLE && LAST_DEC_ROUND == FOUR_TABLES
522: #undef LAST_DEC_ROUND
523: #define LAST_DEC_ROUND ONE_TABLE
524: #endif
525:
526: #if DEC_ROUND == NO_TABLES && DEC_UNROLL != NONE
527: #undef DEC_UNROLL
528: #define DEC_UNROLL NONE
529: #endif
530:
1.1.1.3 ! root 531: #if defined(bswap32)
! 532: #define aes_sw32 bswap32
! 533: #elif defined(bswap_32)
! 534: #define aes_sw32 bswap_32
! 535: #else
! 536: #define brot(x,n) (((uint_32t)(x) << n) | ((uint_32t)(x) >> (32 - n)))
! 537: #define aes_sw32(x) ((brot((x),8) & 0x00ff00ff) | (brot((x),24) & 0xff00ff00))
! 538: #endif
! 539:
1.1 root 540: /* upr(x,n): rotates bytes within words by n positions, moving bytes to
541: higher index positions with wrap around into low positions
542: ups(x,n): moves bytes by n positions to higher index positions in
543: words but without wrap around
544: bval(x,n): extracts a byte from a word
545:
1.1.1.3 ! root 546: WARNING: The definitions given here are intended only for use with
1.1 root 547: unsigned variables and with shift counts that are compile
548: time constants
549: */
550:
1.1.1.3 ! root 551: #if (ALGORITHM_BYTE_ORDER == IS_LITTLE_ENDIAN)
! 552: #define upr(x,n) (((uint_32t)(x) << (8 * (n))) | ((uint_32t)(x) >> (32 - 8 * (n))))
! 553: #define ups(x,n) ((uint_32t) (x) << (8 * (n)))
! 554: #define bval(x,n) ((uint_8t)((x) >> (8 * (n))))
1.1 root 555: #define bytes2word(b0, b1, b2, b3) \
1.1.1.3 ! root 556: (((uint_32t)(b3) << 24) | ((uint_32t)(b2) << 16) | ((uint_32t)(b1) << 8) | (b0))
1.1 root 557: #endif
558:
1.1.1.3 ! root 559: #if (ALGORITHM_BYTE_ORDER == IS_BIG_ENDIAN)
! 560: #define upr(x,n) (((uint_32t)(x) >> (8 * (n))) | ((uint_32t)(x) << (32 - 8 * (n))))
! 561: #define ups(x,n) ((uint_32t) (x) >> (8 * (n)))
! 562: #define bval(x,n) ((uint_8t)((x) >> (24 - 8 * (n))))
1.1 root 563: #define bytes2word(b0, b1, b2, b3) \
1.1.1.3 ! root 564: (((uint_32t)(b0) << 24) | ((uint_32t)(b1) << 16) | ((uint_32t)(b2) << 8) | (b3))
1.1 root 565: #endif
566:
567: #if defined(SAFE_IO)
568:
1.1.1.3 ! root 569: #define word_in(x,c) bytes2word(((const uint_8t*)(x)+4*c)[0], ((const uint_8t*)(x)+4*c)[1], \
! 570: ((const uint_8t*)(x)+4*c)[2], ((const uint_8t*)(x)+4*c)[3])
! 571: #define word_out(x,c,v) { ((uint_8t*)(x)+4*c)[0] = bval(v,0); ((uint_8t*)(x)+4*c)[1] = bval(v,1); \
! 572: ((uint_8t*)(x)+4*c)[2] = bval(v,2); ((uint_8t*)(x)+4*c)[3] = bval(v,3); }
1.1 root 573:
574: #elif (ALGORITHM_BYTE_ORDER == PLATFORM_BYTE_ORDER)
575:
1.1.1.3 ! root 576: #define word_in(x,c) (*((uint_32t*)(x)+(c)))
! 577: #define word_out(x,c,v) (*((uint_32t*)(x)+(c)) = (v))
1.1 root 578:
579: #else
580:
1.1.1.3 ! root 581: #define word_in(x,c) aes_sw32(*((uint_32t*)(x)+(c)))
! 582: #define word_out(x,c,v) (*((uint_32t*)(x)+(c)) = aes_sw32(v))
1.1 root 583:
584: #endif
585:
586: /* the finite field modular polynomial and elements */
587:
588: #define WPOLY 0x011b
589: #define BPOLY 0x1b
590:
591: /* multiply four bytes in GF(2^8) by 'x' {02} in parallel */
592:
593: #define m1 0x80808080
594: #define m2 0x7f7f7f7f
595: #define gf_mulx(x) ((((x) & m2) << 1) ^ ((((x) & m1) >> 7) * BPOLY))
596:
597: /* The following defines provide alternative definitions of gf_mulx that might
598: give improved performance if a fast 32-bit multiply is not available. Note
599: that a temporary variable u needs to be defined where gf_mulx is used.
600:
601: #define gf_mulx(x) (u = (x) & m1, u |= (u >> 1), ((x) & m2) << 1) ^ ((u >> 3) | (u >> 6))
602: #define m4 (0x01010101 * BPOLY)
603: #define gf_mulx(x) (u = (x) & m1, ((x) & m2) << 1) ^ ((u - (u >> 7)) & m4)
604: */
605:
606: /* Work out which tables are needed for the different options */
607:
1.1.1.3 ! root 608: #if defined( ASM_V1 )
1.1 root 609: #if defined( ENC_ROUND )
610: #undef ENC_ROUND
611: #endif
612: #define ENC_ROUND FOUR_TABLES
613: #if defined( LAST_ENC_ROUND )
614: #undef LAST_ENC_ROUND
615: #endif
616: #define LAST_ENC_ROUND FOUR_TABLES
617: #if defined( DEC_ROUND )
618: #undef DEC_ROUND
619: #endif
620: #define DEC_ROUND FOUR_TABLES
621: #if defined( LAST_DEC_ROUND )
622: #undef LAST_DEC_ROUND
623: #endif
624: #define LAST_DEC_ROUND FOUR_TABLES
625: #if defined( KEY_SCHED )
626: #undef KEY_SCHED
627: #define KEY_SCHED FOUR_TABLES
628: #endif
629: #endif
630:
1.1.1.3 ! root 631: #if (FUNCS_IN_C & ENCRYPTION_IN_C) || ASM_V1
1.1 root 632: #if ENC_ROUND == ONE_TABLE
633: #define FT1_SET
634: #elif ENC_ROUND == FOUR_TABLES
635: #define FT4_SET
636: #else
637: #define SBX_SET
638: #endif
639: #if LAST_ENC_ROUND == ONE_TABLE
640: #define FL1_SET
641: #elif LAST_ENC_ROUND == FOUR_TABLES
642: #define FL4_SET
643: #elif !defined(SBX_SET)
644: #define SBX_SET
645: #endif
646: #endif
647:
1.1.1.3 ! root 648: #if (FUNCS_IN_C & DECRYPTION_IN_C) || ASM_V1
1.1 root 649: #if DEC_ROUND == ONE_TABLE
650: #define IT1_SET
651: #elif DEC_ROUND == FOUR_TABLES
652: #define IT4_SET
653: #else
654: #define ISB_SET
655: #endif
656: #if LAST_DEC_ROUND == ONE_TABLE
657: #define IL1_SET
658: #elif LAST_DEC_ROUND == FOUR_TABLES
659: #define IL4_SET
660: #elif !defined(ISB_SET)
661: #define ISB_SET
662: #endif
663: #endif
664:
1.1.1.3 ! root 665: #if (FUNCS_IN_C & ENC_KEYING_IN_C) || (FUNCS_IN_C & DEC_KEYING_IN_C)
1.1 root 666: #if KEY_SCHED == ONE_TABLE
667: #define LS1_SET
668: #elif KEY_SCHED == FOUR_TABLES
669: #define LS4_SET
1.1.1.3 ! root 670: #elif !defined(SBX_SET)
! 671: #define SBX_SET
! 672: #endif
! 673: #endif
! 674:
! 675: #if (FUNCS_IN_C & DEC_KEYING_IN_C)
! 676: #if KEY_SCHED == ONE_TABLE
! 677: #define IM1_SET
! 678: #elif KEY_SCHED == FOUR_TABLES
1.1 root 679: #define IM4_SET
680: #elif !defined(SBX_SET)
681: #define SBX_SET
682: #endif
683: #endif
684:
685: /* generic definitions of Rijndael macros that use tables */
686:
687: #define no_table(x,box,vf,rf,c) bytes2word( \
688: box[bval(vf(x,0,c),rf(0,c))], \
689: box[bval(vf(x,1,c),rf(1,c))], \
690: box[bval(vf(x,2,c),rf(2,c))], \
691: box[bval(vf(x,3,c),rf(3,c))])
692:
693: #define one_table(x,op,tab,vf,rf,c) \
694: ( tab[bval(vf(x,0,c),rf(0,c))] \
695: ^ op(tab[bval(vf(x,1,c),rf(1,c))],1) \
696: ^ op(tab[bval(vf(x,2,c),rf(2,c))],2) \
697: ^ op(tab[bval(vf(x,3,c),rf(3,c))],3))
698:
699: #define four_tables(x,tab,vf,rf,c) \
700: ( tab[0][bval(vf(x,0,c),rf(0,c))] \
701: ^ tab[1][bval(vf(x,1,c),rf(1,c))] \
702: ^ tab[2][bval(vf(x,2,c),rf(2,c))] \
703: ^ tab[3][bval(vf(x,3,c),rf(3,c))])
704:
705: #define vf1(x,r,c) (x)
706: #define rf1(r,c) (r)
707: #define rf2(r,c) ((8+r-c)&3)
708:
709: /* perform forward and inverse column mix operation on four bytes in long word x in */
710: /* parallel. NOTE: x must be a simple variable, NOT an expression in these macros. */
711:
712: #if defined(FM4_SET) /* not currently used */
713: #define fwd_mcol(x) four_tables(x,t_use(f,m),vf1,rf1,0)
714: #elif defined(FM1_SET) /* not currently used */
715: #define fwd_mcol(x) one_table(x,upr,t_use(f,m),vf1,rf1,0)
716: #else
1.1.1.3 ! root 717: #define dec_fmvars uint_32t g2
1.1 root 718: #define fwd_mcol(x) (g2 = gf_mulx(x), g2 ^ upr((x) ^ g2, 3) ^ upr((x), 2) ^ upr((x), 1))
719: #endif
720:
721: #if defined(IM4_SET)
722: #define inv_mcol(x) four_tables(x,t_use(i,m),vf1,rf1,0)
723: #elif defined(IM1_SET)
724: #define inv_mcol(x) one_table(x,upr,t_use(i,m),vf1,rf1,0)
725: #else
1.1.1.3 ! root 726: #define dec_imvars uint_32t g2, g4, g9
1.1 root 727: #define inv_mcol(x) (g2 = gf_mulx(x), g4 = gf_mulx(g2), g9 = (x) ^ gf_mulx(g4), g4 ^= g9, \
728: (x) ^ g2 ^ g4 ^ upr(g2 ^ g9, 3) ^ upr(g4, 2) ^ upr(g9, 1))
729: #endif
730:
731: #if defined(FL4_SET)
732: #define ls_box(x,c) four_tables(x,t_use(f,l),vf1,rf2,c)
733: #elif defined(LS4_SET)
734: #define ls_box(x,c) four_tables(x,t_use(l,s),vf1,rf2,c)
735: #elif defined(FL1_SET)
736: #define ls_box(x,c) one_table(x,upr,t_use(f,l),vf1,rf2,c)
737: #elif defined(LS1_SET)
738: #define ls_box(x,c) one_table(x,upr,t_use(l,s),vf1,rf2,c)
739: #else
740: #define ls_box(x,c) no_table(x,t_use(s,box),vf1,rf2,c)
741: #endif
742:
1.1.1.3 ! root 743: #if defined( ASM_V1 ) && defined( AES_DECRYPT ) && !defined( ISB_SET )
! 744: #define ISB_SET
! 745: #endif
! 746:
1.1 root 747: #endif
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.