|
|
1.1 root 1: /*
2: ---------------------------------------------------------------------------
1.1.1.4 ! root 3: Copyright (c) 1998-2006, Brian Gladman, Worcester, UK. All rights reserved.
1.1 root 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: ---------------------------------------------------------------------------
1.1.1.4 ! root 30: Issue 09/09/2006
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)
1.1.1.4 ! root 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)
1.1.1.4 ! root 56: AES_RETURN the function return type
1.1 root 57:
58: C subroutine calls:
59:
1.1.1.4 ! root 60: AES_RETURN aes_encrypt_key128(const unsigned char *key, aes_encrypt_ctx cx[1]);
! 61: AES_RETURN aes_encrypt_key192(const unsigned char *key, aes_encrypt_ctx cx[1]);
! 62: AES_RETURN aes_encrypt_key256(const unsigned char *key, aes_encrypt_ctx cx[1]);
! 63: AES_RETURN aes_encrypt(const unsigned char *in, unsigned char *out,
1.1 root 64: const aes_encrypt_ctx cx[1]);
65:
1.1.1.4 ! root 66: AES_RETURN aes_decrypt_key128(const unsigned char *key, aes_decrypt_ctx cx[1]);
! 67: AES_RETURN aes_decrypt_key192(const unsigned char *key, aes_decrypt_ctx cx[1]);
! 68: AES_RETURN aes_decrypt_key256(const unsigned char *key, aes_decrypt_ctx cx[1]);
! 69: AES_RETURN aes_decrypt(const unsigned char *in, unsigned char *out,
1.1 root 70: const aes_decrypt_ctx cx[1]);
71:
72: IMPORTANT NOTE: If you are using this C interface with dynamic tables make sure that
1.1.1.4 ! root 73: you call gen_tabs() before AES is used so that the tables are initialised.
1.1 root 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:
1.1.1.4 ! root 83: AES_RETURN key128(const unsigned char *key)
! 84: AES_RETURN key192(const unsigned char *key)
! 85: AES_RETURN key256(const unsigned char *key)
! 86: AES_RETURN encrypt(const unsigned char *in, unsigned char *out) const
1.1 root 87:
88: Class AESdecrypt for encryption
89: Construtors:
90: AESdecrypt(void)
91: AESdecrypt(const unsigned char *key) - 128 bit key
92: Members:
1.1.1.4 ! root 93: AES_RETURN key128(const unsigned char *key)
! 94: AES_RETURN key192(const unsigned char *key)
! 95: AES_RETURN key256(const unsigned char *key)
! 96: AES_RETURN decrypt(const unsigned char *in, unsigned char *out) const
1.1 root 97: */
98:
1.1.1.3 root 99: /* Adapted for TrueCrypt by the TrueCrypt Foundation */
100:
1.1 root 101: #if !defined( _AESOPT_H )
102: #define _AESOPT_H
103:
1.1.1.4 ! root 104: #if defined( __cplusplus )
! 105: #include "Aescpp.h"
1.1.1.3 root 106: #else
1.1.1.2 root 107: #include "Aes.h"
1.1.1.3 root 108: #endif
1.1 root 109:
110:
1.1.1.4 ! root 111: #include "Common/Endian.h"
1.1.1.3 root 112: #define IS_LITTLE_ENDIAN 1234 /* byte 0 is least significant (i386) */
113: #define IS_BIG_ENDIAN 4321 /* byte 0 is most significant (mc68k) */
1.1 root 114:
1.1.1.2 root 115: #if BYTE_ORDER == LITTLE_ENDIAN
1.1.1.3 root 116: # define PLATFORM_BYTE_ORDER IS_LITTLE_ENDIAN
1.1 root 117: #endif
118:
1.1.1.2 root 119: #if BYTE_ORDER == BIG_ENDIAN
1.1.1.3 root 120: # define PLATFORM_BYTE_ORDER IS_BIG_ENDIAN
1.1 root 121: #endif
122:
123:
1.1.1.3 root 124: /* CONFIGURATION - THE USE OF DEFINES
1.1 root 125:
1.1.1.3 root 126: Later in this section there are a number of defines that control the
127: operation of the code. In each section, the purpose of each define is
128: explained so that the relevant form can be included or excluded by
129: setting either 1's or 0's respectively on the branches of the related
130: #if clauses. The following local defines should not be changed.
1.1 root 131: */
132:
1.1.1.3 root 133: #define ENCRYPTION_IN_C 1
134: #define DECRYPTION_IN_C 2
135: #define ENC_KEYING_IN_C 4
136: #define DEC_KEYING_IN_C 8
137:
138: #define NO_TABLES 0
139: #define ONE_TABLE 1
140: #define FOUR_TABLES 4
141: #define NONE 0
142: #define PARTIAL 1
143: #define FULL 2
1.1 root 144:
1.1.1.4 ! root 145: /* --- START OF USER CONFIGURED OPTIONS --- */
! 146:
1.1.1.3 root 147: /* 1. BYTE ORDER WITHIN 32 BIT WORDS
1.1 root 148:
149: The fundamental data processing units in Rijndael are 8-bit bytes. The
150: input, output and key input are all enumerated arrays of bytes in which
151: bytes are numbered starting at zero and increasing to one less than the
152: number of bytes in the array in question. This enumeration is only used
153: for naming bytes and does not imply any adjacency or order relationship
154: from one byte to another. When these inputs and outputs are considered
155: as bit sequences, bits 8*n to 8*n+7 of the bit sequence are mapped to
156: byte[n] with bit 8n+i in the sequence mapped to bit 7-i within the byte.
157: In this implementation bits are numbered from 0 to 7 starting at the
158: numerically least significant end of each byte (bit n represents 2^n).
159:
160: However, Rijndael can be implemented more efficiently using 32-bit
161: words by packing bytes into words so that bytes 4*n to 4*n+3 are placed
162: into word[n]. While in principle these bytes can be assembled into words
163: in any positions, this implementation only supports the two formats in
164: which bytes in adjacent positions within words also have adjacent byte
165: numbers. This order is called big-endian if the lowest numbered bytes
166: in words have the highest numeric significance and little-endian if the
167: opposite applies.
168:
169: This code can work in either order irrespective of the order used by the
170: machine on which it runs. Normally the internal byte order will be set
171: to the order of the processor on which the code is to be run but this
172: define can be used to reverse this in special situations
173:
1.1.1.3 root 174: WARNING: Assembler code versions rely on PLATFORM_BYTE_ORDER being set.
175: This define will hence be redefined later (in section 4) if necessary
1.1 root 176: */
1.1.1.3 root 177:
1.1.1.4 ! root 178: #if 1
1.1 root 179: #define ALGORITHM_BYTE_ORDER PLATFORM_BYTE_ORDER
180: #elif 0
1.1.1.3 root 181: #define ALGORITHM_BYTE_ORDER IS_LITTLE_ENDIAN
1.1 root 182: #elif 0
1.1.1.3 root 183: #define ALGORITHM_BYTE_ORDER IS_BIG_ENDIAN
1.1 root 184: #else
185: #error The algorithm byte order is not defined
186: #endif
187:
1.1.1.3 root 188: /* 2. VIA ACE SUPPORT
189:
1.1.1.4 ! root 190: Define this option if support for the VIA ACE is required. This uses
! 191: inline assembler instructions and is only implemented for the Microsoft,
1.1.1.3 root 192: Intel and GCC compilers. If VIA ACE is known to be present, then defining
1.1.1.4 ! root 193: ASSUME_VIA_ACE_PRESENT will remove the ordinary encryption/decryption
1.1.1.3 root 194: code. If USE_VIA_ACE_IF_PRESENT is defined then VIA ACE will be used if
1.1.1.4 ! root 195: it is detected (both present and enabled) but the normal AES code will
! 196: also be present.
! 197:
! 198: When VIA ACE is to be used, all AES encryption contexts MUST be 16 byte
! 199: aligned; other input/output buffers do not need to be 16 byte aligned
! 200: but there are very large performance gains if this can be arranged.
! 201: VIA ACE also requires the decryption key schedule to be in reverse
! 202: order (which later checks below ensure).
1.1.1.3 root 203: */
204:
205: #if 0 && !defined( USE_VIA_ACE_IF_PRESENT )
1.1.1.4 ! root 206: # define USE_VIA_ACE_IF_PRESENT
1.1.1.3 root 207: #endif
208:
209: #if 0 && !defined( ASSUME_VIA_ACE_PRESENT )
1.1.1.4 ! root 210: # define ASSUME_VIA_ACE_PRESENT
1.1.1.3 root 211: # endif
1.1.1.4 ! root 212:
! 213: #if defined ( _WIN64 ) || defined( _WIN32_WCE ) || \
! 214: defined( _MSC_VER ) && ( _MSC_VER <= 800 )
1.1.1.3 root 215: # if defined( USE_VIA_ACE_IF_PRESENT )
216: # undef USE_VIA_ACE_IF_PRESENT
217: # endif
1.1.1.4 ! root 218: # if defined( ASSUME_VIA_ACE_PRESENT )
! 219: # undef ASSUME_VIA_ACE_PRESENT
! 220: # endif
1.1.1.3 root 221: #endif
222:
223: /* 3. ASSEMBLER SUPPORT
224:
225: This define (which can be on the command line) enables the use of the
226: assembler code routines for encryption, decryption and key scheduling
227: as follows:
228:
1.1.1.4 ! root 229: ASM_X86_V1C uses the assembler (aes_x86_v1.asm) with large tables for
! 230: encryption and decryption and but with key scheduling in C
! 231: ASM_X86_V2 uses assembler (aes_x86_v2.asm) with compressed tables for
! 232: encryption, decryption and key scheduling
! 233: ASM_X86_V2C uses assembler (aes_x86_v2.asm) with compressed tables for
! 234: encryption and decryption and but with key scheduling in C
! 235: ASM_AMD64_C uses assembler (aes_amd64.asm) with compressed tables for
! 236: encryption and decryption and but with key scheduling in C
! 237:
! 238: Change one 'if 0' below to 'if 1' to select the version or define
! 239: as a compilation option.
! 240: */
! 241:
! 242: #if 0 && !defined( ASM_X86_V1C )
! 243: # define ASM_X86_V1C
! 244: #elif 0 && !defined( ASM_X86_V2 )
! 245: # define ASM_X86_V2
! 246: #elif 0 && !defined( ASM_X86_V2C )
! 247: # define ASM_X86_V2C
! 248: #elif 0 && !defined( ASM_AMD64_C )
! 249: # define ASM_AMD64_C
! 250: #endif
! 251:
! 252: #if (defined ( ASM_X86_V1C ) || defined( ASM_X86_V2 ) || defined( ASM_X86_V2C )) \
! 253: && !defined( _M_IX86 ) || defined( ASM_AMD64_C ) && !defined( _M_X64 )
! 254: # error Assembler code is only available for x86 and AMD64 systems
1.1.1.3 root 255: #endif
256:
1.1.1.4 ! root 257: /* 4. FAST INPUT/OUTPUT OPERATIONS.
1.1 root 258:
259: On some machines it is possible to improve speed by transferring the
260: bytes in the input and output arrays to and from the internal 32-bit
261: variables by addressing these arrays as if they are arrays of 32-bit
262: words. On some machines this will always be possible but there may
263: be a large performance penalty if the byte arrays are not aligned on
264: the normal word boundaries. On other machines this technique will
265: lead to memory access errors when such 32-bit word accesses are not
266: properly aligned. The option SAFE_IO avoids such problems but will
267: often be slower on those machines that support misaligned access
268: (especially so if care is taken to align the input and output byte
269: arrays on 32-bit word boundaries). If SAFE_IO is not defined it is
270: assumed that access to byte arrays as if they are arrays of 32-bit
271: words will not cause problems when such accesses are misaligned.
272: */
1.1.1.4 ! root 273: #if 1 && !defined( _MSC_VER )
1.1 root 274: #define SAFE_IO
275: #endif
276:
1.1.1.4 ! root 277: /* 5. LOOP UNROLLING
1.1 root 278:
279: The code for encryption and decrytpion cycles through a number of rounds
280: that can be implemented either in a loop or by expanding the code into a
281: long sequence of instructions, the latter producing a larger program but
282: one that will often be much faster. The latter is called loop unrolling.
283: There are also potential speed advantages in expanding two iterations in
284: a loop with half the number of iterations, which is called partial loop
285: unrolling. The following options allow partial or full loop unrolling
286: to be set independently for encryption and decryption
287: */
288: #if 1
289: #define ENC_UNROLL FULL
290: #elif 0
291: #define ENC_UNROLL PARTIAL
292: #else
293: #define ENC_UNROLL NONE
294: #endif
295:
296: #if 1
297: #define DEC_UNROLL FULL
298: #elif 0
299: #define DEC_UNROLL PARTIAL
300: #else
301: #define DEC_UNROLL NONE
302: #endif
303:
1.1.1.4 ! root 304: /* 6. FAST FINITE FIELD OPERATIONS
1.1 root 305:
306: If this section is included, tables are used to provide faster finite
307: field arithmetic (this has no effect if FIXED_TABLES is defined).
308: */
309: #if 1
310: #define FF_TABLES
311: #endif
312:
1.1.1.4 ! root 313: /* 7. INTERNAL STATE VARIABLE FORMAT
1.1 root 314:
315: The internal state of Rijndael is stored in a number of local 32-bit
316: word varaibles which can be defined either as an array or as individual
317: names variables. Include this section if you want to store these local
318: varaibles in arrays. Otherwise individual local variables will be used.
319: */
320: #if 1
321: #define ARRAYS
322: #endif
323:
1.1.1.4 ! root 324: /* 8. FIXED OR DYNAMIC TABLES
1.1 root 325:
326: When this section is included the tables used by the code are compiled
327: statically into the binary file. Otherwise the subroutine gen_tabs()
328: must be called to compute them before the code is first used.
329: */
1.1.1.4 ! root 330: #if 1 && !(defined( _MSC_VER ) && ( _MSC_VER <= 800 ))
1.1 root 331: #define FIXED_TABLES
332: #endif
333:
1.1.1.4 ! root 334: /* 9. TABLE ALIGNMENT
1.1 root 335:
336: On some sytsems speed will be improved by aligning the AES large lookup
337: tables on particular boundaries. This define should be set to a power of
338: two giving the desired alignment. It can be left undefined if alignment
339: is not needed. This option is specific to the Microsft VC++ compiler -
340: it seems to sometimes cause trouble for the VC++ version 6 compiler.
341: */
342:
1.1.1.4 ! root 343: #if 1 && defined( _MSC_VER ) && ( _MSC_VER >= 1300 )
1.1.1.3 root 344: #define TABLE_ALIGN 32
1.1 root 345: #endif
346:
1.1.1.4 ! root 347: /* 10. TABLE OPTIONS
1.1 root 348:
349: This cipher proceeds by repeating in a number of cycles known as 'rounds'
350: which are implemented by a round function which can optionally be speeded
351: up using tables. The basic tables are each 256 32-bit words, with either
352: one or four tables being required for each round function depending on
353: how much speed is required. The encryption and decryption round functions
354: are different and the last encryption and decrytpion round functions are
355: different again making four different round functions in all.
356:
357: This means that:
358: 1. Normal encryption and decryption rounds can each use either 0, 1
359: or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
360: 2. The last encryption and decryption rounds can also use either 0, 1
361: or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
362:
363: Include or exclude the appropriate definitions below to set the number
364: of tables used by this implementation.
365: */
366:
367: #if 1 /* set tables for the normal encryption round */
368: #define ENC_ROUND FOUR_TABLES
369: #elif 0
370: #define ENC_ROUND ONE_TABLE
371: #else
372: #define ENC_ROUND NO_TABLES
373: #endif
374:
375: #if 1 /* set tables for the last encryption round */
376: #define LAST_ENC_ROUND FOUR_TABLES
377: #elif 0
378: #define LAST_ENC_ROUND ONE_TABLE
379: #else
380: #define LAST_ENC_ROUND NO_TABLES
381: #endif
382:
383: #if 1 /* set tables for the normal decryption round */
384: #define DEC_ROUND FOUR_TABLES
385: #elif 0
386: #define DEC_ROUND ONE_TABLE
387: #else
388: #define DEC_ROUND NO_TABLES
389: #endif
390:
391: #if 1 /* set tables for the last decryption round */
392: #define LAST_DEC_ROUND FOUR_TABLES
393: #elif 0
394: #define LAST_DEC_ROUND ONE_TABLE
395: #else
396: #define LAST_DEC_ROUND NO_TABLES
397: #endif
398:
399: /* The decryption key schedule can be speeded up with tables in the same
400: way that the round functions can. Include or exclude the following
401: defines to set this requirement.
402: */
403: #if 1
404: #define KEY_SCHED FOUR_TABLES
405: #elif 0
406: #define KEY_SCHED ONE_TABLE
407: #else
408: #define KEY_SCHED NO_TABLES
409: #endif
410:
1.1.1.4 ! root 411: /* ---- END OF USER CONFIGURED OPTIONS ---- */
! 412:
! 413: /* VIA ACE support is only available for VC++ and GCC */
! 414:
! 415: #if !defined( _MSC_VER ) && !defined( __GNUC__ )
! 416: # if defined( ASSUME_VIA_ACE_PRESENT )
! 417: # undef ASSUME_VIA_ACE_PRESENT
! 418: # endif
! 419: # if defined( USE_VIA_ACE_IF_PRESENT )
! 420: # undef USE_VIA_ACE_IF_PRESENT
! 421: # endif
! 422: #endif
! 423:
! 424: #if defined( ASSUME_VIA_ACE_PRESENT ) && !defined( USE_VIA_ACE_IF_PRESENT )
! 425: #define USE_VIA_ACE_IF_PRESENT
! 426: #endif
! 427:
! 428: #if defined( USE_VIA_ACE_IF_PRESENT ) && !defined ( AES_REV_DKS )
! 429: #define AES_REV_DKS
! 430: #endif
! 431:
! 432: /* Assembler support requires the use of platform byte order */
! 433:
! 434: #if ( defined( ASM_X86_V1C ) || defined( ASM_X86_V2C ) || defined( ASM_AMD64_C ) ) \
! 435: && (ALGORITHM_BYTE_ORDER != PLATFORM_BYTE_ORDER)
! 436: #undef ALGORITHM_BYTE_ORDER
! 437: #define ALGORITHM_BYTE_ORDER PLATFORM_BYTE_ORDER
! 438: #endif
! 439:
! 440: /* In this implementation the columns of the state array are each held in
! 441: 32-bit words. The state array can be held in various ways: in an array
! 442: of words, in a number of individual word variables or in a number of
! 443: processor registers. The following define maps a variable name x and
! 444: a column number c to the way the state array variable is to be held.
! 445: The first define below maps the state into an array x[c] whereas the
! 446: second form maps the state into a number of individual variables x0,
! 447: x1, etc. Another form could map individual state colums to machine
! 448: register names.
! 449: */
! 450:
! 451: #if defined( ARRAYS )
! 452: #define s(x,c) x[c]
! 453: #else
! 454: #define s(x,c) x##c
! 455: #endif
! 456:
! 457: /* This implementation provides subroutines for encryption, decryption
! 458: and for setting the three key lengths (separately) for encryption
! 459: and decryption. Since not all functions are needed, masks are set
! 460: up here to determine which will be implemented in C
! 461: */
! 462:
! 463: #if !defined( AES_ENCRYPT )
! 464: # define EFUNCS_IN_C 0
! 465: #elif defined( ASSUME_VIA_ACE_PRESENT ) || defined( ASM_X86_V1C ) \
! 466: || defined( ASM_X86_V2C ) || defined( ASM_AMD64_C )
! 467: # define EFUNCS_IN_C ENC_KEYING_IN_C
! 468: #elif !defined( ASM_X86_V2 )
! 469: # define EFUNCS_IN_C ( ENCRYPTION_IN_C | ENC_KEYING_IN_C )
! 470: #else
! 471: # define EFUNCS_IN_C 0
! 472: #endif
! 473:
! 474: #if !defined( AES_DECRYPT )
! 475: # define DFUNCS_IN_C 0
! 476: #elif defined( ASSUME_VIA_ACE_PRESENT ) || defined( ASM_X86_V1C ) \
! 477: || defined( ASM_X86_V2C ) || defined( ASM_AMD64_C )
! 478: # define DFUNCS_IN_C DEC_KEYING_IN_C
! 479: #elif !defined( ASM_X86_V2 )
! 480: # define DFUNCS_IN_C ( DECRYPTION_IN_C | DEC_KEYING_IN_C )
! 481: #else
! 482: # define DFUNCS_IN_C 0
! 483: #endif
! 484:
! 485: #define FUNCS_IN_C ( EFUNCS_IN_C | DFUNCS_IN_C )
! 486:
1.1 root 487: /* END OF CONFIGURATION OPTIONS */
488:
489: #define RC_LENGTH (5 * (AES_BLOCK_SIZE / 4 - 2))
490:
491: /* Disable or report errors on some combinations of options */
492:
493: #if ENC_ROUND == NO_TABLES && LAST_ENC_ROUND != NO_TABLES
494: #undef LAST_ENC_ROUND
495: #define LAST_ENC_ROUND NO_TABLES
496: #elif ENC_ROUND == ONE_TABLE && LAST_ENC_ROUND == FOUR_TABLES
497: #undef LAST_ENC_ROUND
498: #define LAST_ENC_ROUND ONE_TABLE
499: #endif
500:
501: #if ENC_ROUND == NO_TABLES && ENC_UNROLL != NONE
502: #undef ENC_UNROLL
503: #define ENC_UNROLL NONE
504: #endif
505:
506: #if DEC_ROUND == NO_TABLES && LAST_DEC_ROUND != NO_TABLES
507: #undef LAST_DEC_ROUND
508: #define LAST_DEC_ROUND NO_TABLES
509: #elif DEC_ROUND == ONE_TABLE && LAST_DEC_ROUND == FOUR_TABLES
510: #undef LAST_DEC_ROUND
511: #define LAST_DEC_ROUND ONE_TABLE
512: #endif
513:
514: #if DEC_ROUND == NO_TABLES && DEC_UNROLL != NONE
515: #undef DEC_UNROLL
516: #define DEC_UNROLL NONE
517: #endif
518:
1.1.1.4 ! root 519: #if defined( bswap32 )
1.1.1.3 root 520: #define aes_sw32 bswap32
1.1.1.4 ! root 521: #elif defined( bswap_32 )
1.1.1.3 root 522: #define aes_sw32 bswap_32
523: #else
524: #define brot(x,n) (((uint_32t)(x) << n) | ((uint_32t)(x) >> (32 - n)))
525: #define aes_sw32(x) ((brot((x),8) & 0x00ff00ff) | (brot((x),24) & 0xff00ff00))
526: #endif
527:
1.1 root 528: /* upr(x,n): rotates bytes within words by n positions, moving bytes to
529: higher index positions with wrap around into low positions
530: ups(x,n): moves bytes by n positions to higher index positions in
531: words but without wrap around
532: bval(x,n): extracts a byte from a word
533:
1.1.1.3 root 534: WARNING: The definitions given here are intended only for use with
1.1 root 535: unsigned variables and with shift counts that are compile
536: time constants
537: */
538:
1.1.1.4 ! root 539: #if ( ALGORITHM_BYTE_ORDER == IS_LITTLE_ENDIAN )
1.1.1.3 root 540: #define upr(x,n) (((uint_32t)(x) << (8 * (n))) | ((uint_32t)(x) >> (32 - 8 * (n))))
541: #define ups(x,n) ((uint_32t) (x) << (8 * (n)))
542: #define bval(x,n) ((uint_8t)((x) >> (8 * (n))))
1.1 root 543: #define bytes2word(b0, b1, b2, b3) \
1.1.1.3 root 544: (((uint_32t)(b3) << 24) | ((uint_32t)(b2) << 16) | ((uint_32t)(b1) << 8) | (b0))
1.1 root 545: #endif
546:
1.1.1.4 ! root 547: #if ( ALGORITHM_BYTE_ORDER == IS_BIG_ENDIAN )
1.1.1.3 root 548: #define upr(x,n) (((uint_32t)(x) >> (8 * (n))) | ((uint_32t)(x) << (32 - 8 * (n))))
549: #define ups(x,n) ((uint_32t) (x) >> (8 * (n)))
550: #define bval(x,n) ((uint_8t)((x) >> (24 - 8 * (n))))
1.1 root 551: #define bytes2word(b0, b1, b2, b3) \
1.1.1.3 root 552: (((uint_32t)(b0) << 24) | ((uint_32t)(b1) << 16) | ((uint_32t)(b2) << 8) | (b3))
1.1 root 553: #endif
554:
1.1.1.4 ! root 555: #if defined( SAFE_IO )
1.1 root 556:
1.1.1.3 root 557: #define word_in(x,c) bytes2word(((const uint_8t*)(x)+4*c)[0], ((const uint_8t*)(x)+4*c)[1], \
558: ((const uint_8t*)(x)+4*c)[2], ((const uint_8t*)(x)+4*c)[3])
559: #define word_out(x,c,v) { ((uint_8t*)(x)+4*c)[0] = bval(v,0); ((uint_8t*)(x)+4*c)[1] = bval(v,1); \
560: ((uint_8t*)(x)+4*c)[2] = bval(v,2); ((uint_8t*)(x)+4*c)[3] = bval(v,3); }
1.1 root 561:
1.1.1.4 ! root 562: #elif ( ALGORITHM_BYTE_ORDER == PLATFORM_BYTE_ORDER )
1.1 root 563:
1.1.1.3 root 564: #define word_in(x,c) (*((uint_32t*)(x)+(c)))
565: #define word_out(x,c,v) (*((uint_32t*)(x)+(c)) = (v))
1.1 root 566:
567: #else
568:
1.1.1.3 root 569: #define word_in(x,c) aes_sw32(*((uint_32t*)(x)+(c)))
570: #define word_out(x,c,v) (*((uint_32t*)(x)+(c)) = aes_sw32(v))
1.1 root 571:
572: #endif
573:
574: /* the finite field modular polynomial and elements */
575:
576: #define WPOLY 0x011b
577: #define BPOLY 0x1b
578:
579: /* multiply four bytes in GF(2^8) by 'x' {02} in parallel */
580:
581: #define m1 0x80808080
582: #define m2 0x7f7f7f7f
583: #define gf_mulx(x) ((((x) & m2) << 1) ^ ((((x) & m1) >> 7) * BPOLY))
584:
585: /* The following defines provide alternative definitions of gf_mulx that might
586: give improved performance if a fast 32-bit multiply is not available. Note
587: that a temporary variable u needs to be defined where gf_mulx is used.
588:
589: #define gf_mulx(x) (u = (x) & m1, u |= (u >> 1), ((x) & m2) << 1) ^ ((u >> 3) | (u >> 6))
590: #define m4 (0x01010101 * BPOLY)
591: #define gf_mulx(x) (u = (x) & m1, ((x) & m2) << 1) ^ ((u - (u >> 7)) & m4)
592: */
593:
594: /* Work out which tables are needed for the different options */
595:
1.1.1.4 ! root 596: #if defined( ASM_X86_V1C )
1.1 root 597: #if defined( ENC_ROUND )
598: #undef ENC_ROUND
599: #endif
600: #define ENC_ROUND FOUR_TABLES
601: #if defined( LAST_ENC_ROUND )
602: #undef LAST_ENC_ROUND
603: #endif
604: #define LAST_ENC_ROUND FOUR_TABLES
605: #if defined( DEC_ROUND )
606: #undef DEC_ROUND
607: #endif
608: #define DEC_ROUND FOUR_TABLES
609: #if defined( LAST_DEC_ROUND )
610: #undef LAST_DEC_ROUND
611: #endif
612: #define LAST_DEC_ROUND FOUR_TABLES
613: #if defined( KEY_SCHED )
614: #undef KEY_SCHED
615: #define KEY_SCHED FOUR_TABLES
616: #endif
617: #endif
618:
1.1.1.4 ! root 619: #if ( FUNCS_IN_C & ENCRYPTION_IN_C ) || defined( ASM_X86_V1C )
1.1 root 620: #if ENC_ROUND == ONE_TABLE
621: #define FT1_SET
622: #elif ENC_ROUND == FOUR_TABLES
623: #define FT4_SET
624: #else
625: #define SBX_SET
626: #endif
627: #if LAST_ENC_ROUND == ONE_TABLE
628: #define FL1_SET
629: #elif LAST_ENC_ROUND == FOUR_TABLES
630: #define FL4_SET
1.1.1.4 ! root 631: #elif !defined( SBX_SET )
1.1 root 632: #define SBX_SET
633: #endif
634: #endif
635:
1.1.1.4 ! root 636: #if ( FUNCS_IN_C & DECRYPTION_IN_C ) || defined( ASM_X86_V1C )
1.1 root 637: #if DEC_ROUND == ONE_TABLE
638: #define IT1_SET
639: #elif DEC_ROUND == FOUR_TABLES
640: #define IT4_SET
641: #else
642: #define ISB_SET
643: #endif
644: #if LAST_DEC_ROUND == ONE_TABLE
645: #define IL1_SET
646: #elif LAST_DEC_ROUND == FOUR_TABLES
647: #define IL4_SET
648: #elif !defined(ISB_SET)
649: #define ISB_SET
650: #endif
651: #endif
652:
1.1.1.3 root 653: #if (FUNCS_IN_C & ENC_KEYING_IN_C) || (FUNCS_IN_C & DEC_KEYING_IN_C)
1.1 root 654: #if KEY_SCHED == ONE_TABLE
655: #define LS1_SET
656: #elif KEY_SCHED == FOUR_TABLES
657: #define LS4_SET
1.1.1.4 ! root 658: #elif !defined( SBX_SET )
1.1.1.3 root 659: #define SBX_SET
660: #endif
661: #endif
662:
663: #if (FUNCS_IN_C & DEC_KEYING_IN_C)
664: #if KEY_SCHED == ONE_TABLE
665: #define IM1_SET
666: #elif KEY_SCHED == FOUR_TABLES
1.1 root 667: #define IM4_SET
1.1.1.4 ! root 668: #elif !defined( SBX_SET )
1.1 root 669: #define SBX_SET
670: #endif
671: #endif
672:
673: /* generic definitions of Rijndael macros that use tables */
674:
675: #define no_table(x,box,vf,rf,c) bytes2word( \
676: box[bval(vf(x,0,c),rf(0,c))], \
677: box[bval(vf(x,1,c),rf(1,c))], \
678: box[bval(vf(x,2,c),rf(2,c))], \
679: box[bval(vf(x,3,c),rf(3,c))])
680:
681: #define one_table(x,op,tab,vf,rf,c) \
682: ( tab[bval(vf(x,0,c),rf(0,c))] \
683: ^ op(tab[bval(vf(x,1,c),rf(1,c))],1) \
684: ^ op(tab[bval(vf(x,2,c),rf(2,c))],2) \
685: ^ op(tab[bval(vf(x,3,c),rf(3,c))],3))
686:
687: #define four_tables(x,tab,vf,rf,c) \
688: ( tab[0][bval(vf(x,0,c),rf(0,c))] \
689: ^ tab[1][bval(vf(x,1,c),rf(1,c))] \
690: ^ tab[2][bval(vf(x,2,c),rf(2,c))] \
691: ^ tab[3][bval(vf(x,3,c),rf(3,c))])
692:
693: #define vf1(x,r,c) (x)
694: #define rf1(r,c) (r)
695: #define rf2(r,c) ((8+r-c)&3)
696:
697: /* perform forward and inverse column mix operation on four bytes in long word x in */
698: /* parallel. NOTE: x must be a simple variable, NOT an expression in these macros. */
699:
1.1.1.4 ! root 700: #if defined( FM4_SET ) /* not currently used */
! 701: #define fwd_mcol(x) four_tables(x,t_use(f,m),vf1,rf1,0)
! 702: #elif defined( FM1_SET ) /* not currently used */
! 703: #define fwd_mcol(x) one_table(x,upr,t_use(f,m),vf1,rf1,0)
! 704: #else
! 705: #define dec_fmvars uint_32t g2
! 706: #define fwd_mcol(x) (g2 = gf_mulx(x), g2 ^ upr((x) ^ g2, 3) ^ upr((x), 2) ^ upr((x), 1))
! 707: #endif
! 708:
! 709: #if defined( IM4_SET )
! 710: #define inv_mcol(x) four_tables(x,t_use(i,m),vf1,rf1,0)
! 711: #elif defined( IM1_SET )
! 712: #define inv_mcol(x) one_table(x,upr,t_use(i,m),vf1,rf1,0)
! 713: #else
! 714: #define dec_imvars uint_32t g2, g4, g9
! 715: #define inv_mcol(x) (g2 = gf_mulx(x), g4 = gf_mulx(g2), g9 = (x) ^ gf_mulx(g4), g4 ^= g9, \
! 716: (x) ^ g2 ^ g4 ^ upr(g2 ^ g9, 3) ^ upr(g4, 2) ^ upr(g9, 1))
! 717: #endif
! 718:
! 719: #if defined( FL4_SET )
! 720: #define ls_box(x,c) four_tables(x,t_use(f,l),vf1,rf2,c)
! 721: #elif defined( LS4_SET )
! 722: #define ls_box(x,c) four_tables(x,t_use(l,s),vf1,rf2,c)
! 723: #elif defined( FL1_SET )
! 724: #define ls_box(x,c) one_table(x,upr,t_use(f,l),vf1,rf2,c)
! 725: #elif defined( LS1_SET )
! 726: #define ls_box(x,c) one_table(x,upr,t_use(l,s),vf1,rf2,c)
1.1 root 727: #else
728: #define ls_box(x,c) no_table(x,t_use(s,box),vf1,rf2,c)
729: #endif
730:
1.1.1.4 ! root 731: #if defined( ASM_X86_V1C ) && defined( AES_DECRYPT ) && !defined( ISB_SET )
1.1.1.3 root 732: #define ISB_SET
733: #endif
734:
1.1 root 735: #endif
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.