2 ---------------------------------------------------------------------------
3 Copyright (c) 2003, Dr Brian Gladman <brg@gladman.me.uk>, Worcester, UK.
8 The free distribution and use of this software in both source and binary
9 form is allowed (with or without changes) provided that:
11 1. distributions of this source code include the above copyright
12 notice, this list of conditions and the following disclaimer;
14 2. distributions in binary form include the above copyright
15 notice, this list of conditions and the following disclaimer
16 in the documentation and/or other associated materials;
18 3. the copyright holder's name is not used to endorse products
19 built using this software without specific written permission.
21 ALTERNATIVELY, provided that this notice is retained in full, this product
22 may be distributed under the terms of the GNU General Public License (GPL),
23 in which case the provisions of the GPL apply INSTEAD OF those given above.
27 This software is provided 'as is' with no explicit or implied warranties
28 in respect of its properties, including, but not limited to, correctness
29 and/or fitness for purpose.
30 ---------------------------------------------------------------------------
31 Issue Date: 26/08/2003
33 My thanks go to Dag Arne Osvik for devising the schemes used here for key
34 length derivation from the form of the key schedule
36 This file contains the compilation options for AES (Rijndael) and code
37 that is common across encryption, key scheduling and table generation.
41 These source code files implement the AES algorithm Rijndael designed by
42 Joan Daemen and Vincent Rijmen. This version is designed for the standard
43 block size of 16 bytes and for key sizes of 128, 192 and 256 bits (16, 24
46 This version is designed for flexibility and speed using operations on
47 32-bit words rather than operations on bytes. It can be compiled with
48 either big or little endian internal byte order but is faster when the
49 native byte order for the processor is used.
53 The cipher interface is implemented as an array of bytes in which lower
54 AES bit sequence indexes map to higher numeric significance within bytes.
56 aes_08t (an unsigned 8-bit type)
57 aes_32t (an unsigned 32-bit type)
58 struct aes_encrypt_ctx (structure for the cipher encryption context)
59 struct aes_decrypt_ctx (structure for the cipher decryption context)
60 aes_rval the function return type
64 aes_rval aes_encrypt_key128(const void *in_key, aes_encrypt_ctx cx[1]);
65 aes_rval aes_encrypt_key192(const void *in_key, aes_encrypt_ctx cx[1]);
66 aes_rval aes_encrypt_key256(const void *in_key, aes_encrypt_ctx cx[1]);
67 aes_rval aes_encrypt(const void *in_blk,
68 void *out_blk, const aes_encrypt_ctx cx[1]);
70 aes_rval aes_decrypt_key128(const void *in_key, aes_decrypt_ctx cx[1]);
71 aes_rval aes_decrypt_key192(const void *in_key, aes_decrypt_ctx cx[1]);
72 aes_rval aes_decrypt_key256(const void *in_key, aes_decrypt_ctx cx[1]);
73 aes_rval aes_decrypt(const void *in_blk,
74 void *out_blk, const aes_decrypt_ctx cx[1]);
76 IMPORTANT NOTE: If you are using this C interface with dynamic tables make sure that
77 you call genTabs() before AES is used so that the tables are initialised.
79 C++ aes class subroutines:
81 Class AESencrypt for encryption
85 AESencrypt(const void *in_key) - 128 bit key
87 void key128(const void *in_key)
88 void key192(const void *in_key)
89 void key256(const void *in_key)
90 void encrypt(const void *in_blk, void *out_blk) const
92 Class AESdecrypt for encryption
95 AESdecrypt(const void *in_key) - 128 bit key
97 void key128(const void *in_key)
98 void key192(const void *in_key)
99 void key256(const void *in_key)
100 void decrypt(const void *in_blk, void *out_blk) const
104 The files used to provide AES (Rijndael) are
106 a. aes.h for the definitions needed for use in C.
107 b. aescpp.h for the definitions needed for use in C++.
108 c. aesopt.h for setting compilation options (also includes common code).
109 d. aescrypt.c for encryption and decrytpion, or
110 e. aeskey.c for key scheduling.
111 f. aestab.c for table loading or generation.
112 g. aescrypt.asm for encryption and decryption using assembler code.
113 h. aescrypt.mmx.asm for encryption and decryption using MMX assembler.
115 To compile AES (Rijndael) for use in C code use aes.h and set the
116 defines here for the facilities you need (key lengths, encryption
117 and/or decryption). Do not define AES_DLL or AES_CPP. Set the options
118 for optimisations and table sizes here.
120 To compile AES (Rijndael) for use in in C++ code use aescpp.h but do
123 To compile AES (Rijndael) in C as a Dynamic Link Library DLL) use
124 aes.h and include the AES_DLL define.
126 CONFIGURATION OPTIONS (here and in aes.h)
128 a. set AES_DLL in aes.h if AES (Rijndael) is to be compiled as a DLL
129 b. You may need to set PLATFORM_BYTE_ORDER to define the byte order.
130 c. If you want the code to run in a specific internal byte order, then
131 ALGORITHM_BYTE_ORDER must be set accordingly.
132 d. set other configuration options decribed below.
138 #include <asterisk/aes.h>
140 /* CONFIGURATION - USE OF DEFINES
142 Later in this section there are a number of defines that control the
143 operation of the code. In each section, the purpose of each define is
144 explained so that the relevant form can be included or excluded by
145 setting either 1's or 0's respectively on the branches of the related
149 /* PLATFORM SPECIFIC INCLUDES */
151 #if defined( __OpenBSD__ )
152 # include <machine/types.h>
153 # include <sys/endian.h>
154 #elif defined( __FreeBSD__ )
155 # include <sys/types.h>
156 # include <sys/endian.h>
157 #elif defined( BSD ) && ( BSD >= 199103 ) || defined(__APPLE__)
158 # include <machine/endian.h>
159 #elif defined( __GNUC__ ) || defined( __GNU_LIBRARY__ )
161 #if !defined(__APPLE__)
162 # include <byteswap.h>
164 #elif defined( linux )
168 /* BYTE ORDER IN 32-BIT WORDS
170 To obtain the highest speed on processors with 32-bit words, this code
171 needs to determine the byte order of the target machine. The following
172 block of code is an attempt to capture the most obvious ways in which
173 various environemnts define byte order. It may well fail, in which case
174 the definitions will need to be set by editing at the points marked
175 **** EDIT HERE IF NECESSARY **** below. My thanks to Peter Gutmann for
176 some of these defines (from cryptlib).
179 #define BRG_LITTLE_ENDIAN 1234 /* byte 0 is least significant (i386) */
180 #define BRG_BIG_ENDIAN 4321 /* byte 0 is most significant (mc68k) */
182 #if defined( __alpha__ ) || defined( __alpha ) || defined( i386 ) || \
183 defined( __i386__ ) || defined( _M_I86 ) || defined( _M_IX86 ) || \
184 defined( __OS2__ ) || defined( sun386 ) || defined( __TURBOC__ ) || \
185 defined( vax ) || defined( vms ) || defined( VMS ) || \
188 #define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
192 #if defined( AMIGA ) || defined( applec ) || defined( __AS400__ ) || \
193 defined( _CRAY ) || defined( __hppa ) || defined( __hp9000 ) || \
194 defined( ibm370 ) || defined( mc68000 ) || defined( m68k ) || \
195 defined( __MRC__ ) || defined( __MVS__ ) || defined( __MWERKS__ ) || \
196 defined( sparc ) || defined( __sparc) || defined( SYMANTEC_C ) || \
197 defined( __TANDEM ) || defined( THINK_C ) || defined( __VMCMS__ )
199 #define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
203 /* if the platform is still not known, try to find its byte order */
204 /* from commonly used definitions in the headers included earlier */
206 #if !defined(PLATFORM_BYTE_ORDER)
208 #if defined(LITTLE_ENDIAN) || defined(BIG_ENDIAN)
209 # if defined(LITTLE_ENDIAN) && !defined(BIG_ENDIAN)
210 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
211 # elif !defined(LITTLE_ENDIAN) && defined(BIG_ENDIAN)
212 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
213 # elif defined(BYTE_ORDER) && (BYTE_ORDER == LITTLE_ENDIAN)
214 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
215 # elif defined(BYTE_ORDER) && (BYTE_ORDER == BIG_ENDIAN)
216 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
219 #elif defined(_LITTLE_ENDIAN) || defined(_BIG_ENDIAN)
220 # if defined(_LITTLE_ENDIAN) && !defined(_BIG_ENDIAN)
221 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
222 # elif !defined(_LITTLE_ENDIAN) && defined(_BIG_ENDIAN)
223 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
224 # elif defined(_BYTE_ORDER) && (_BYTE_ORDER == _LITTLE_ENDIAN)
225 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
226 # elif defined(_BYTE_ORDER) && (_BYTE_ORDER == _BIG_ENDIAN)
227 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
230 #elif defined(__LITTLE_ENDIAN__) || defined(__BIG_ENDIAN__)
231 # if defined(__LITTLE_ENDIAN__) && !defined(__BIG_ENDIAN__)
232 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
233 # elif !defined(__LITTLE_ENDIAN__) && defined(__BIG_ENDIAN__)
234 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
235 # elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __LITTLE_ENDIAN__)
236 # define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
237 # elif defined(__BYTE_ORDER__) && (__BYTE_ORDER__ == __BIG_ENDIAN__)
238 # define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
241 #elif 0 /* **** EDIT HERE IF NECESSARY **** */
242 #define PLATFORM_BYTE_ORDER BRG_LITTLE_ENDIAN
244 #elif 0 /* **** EDIT HERE IF NECESSARY **** */
245 #define PLATFORM_BYTE_ORDER BRG_BIG_ENDIAN
248 #error Please edit aesopt.h (line 235 or 238) to set the platform byte order
253 /* SOME LOCAL DEFINITIONS */
257 #define FOUR_TABLES 4
263 #define aes_sw32 bswap32
264 #elif defined(bswap_32)
265 #define aes_sw32 bswap_32
267 #define brot(x,n) (((aes_32t)(x) << n) | ((aes_32t)(x) >> (32 - n)))
268 #define aes_sw32(x) ((brot((x),8) & 0x00ff00ff) | (brot((x),24) & 0xff00ff00))
271 /* 1. FUNCTIONS REQUIRED
273 This implementation provides subroutines for encryption, decryption
274 and for setting the three key lengths (separately) for encryption
275 and decryption. When the assembler code is not being used the following
276 definition blocks allow the selection of the routines that are to be
277 included in the compilation.
281 #define ENCRYPTION_KEY_SCHEDULE
286 #define DECRYPTION_KEY_SCHEDULE
289 /* 2. ASSEMBLER SUPPORT
291 This define (which can be on the command line) enables the use of the
292 assembler code routines for encryption and decryption with the C code
293 only providing key scheduling
299 /* 3. BYTE ORDER WITHIN 32 BIT WORDS
301 The fundamental data processing units in Rijndael are 8-bit bytes. The
302 input, output and key input are all enumerated arrays of bytes in which
303 bytes are numbered starting at zero and increasing to one less than the
304 number of bytes in the array in question. This enumeration is only used
305 for naming bytes and does not imply any adjacency or order relationship
306 from one byte to another. When these inputs and outputs are considered
307 as bit sequences, bits 8*n to 8*n+7 of the bit sequence are mapped to
308 byte[n] with bit 8n+i in the sequence mapped to bit 7-i within the byte.
309 In this implementation bits are numbered from 0 to 7 starting at the
310 numerically least significant end of each byte (bit n represents 2^n).
312 However, Rijndael can be implemented more efficiently using 32-bit
313 words by packing bytes into words so that bytes 4*n to 4*n+3 are placed
314 into word[n]. While in principle these bytes can be assembled into words
315 in any positions, this implementation only supports the two formats in
316 which bytes in adjacent positions within words also have adjacent byte
317 numbers. This order is called big-endian if the lowest numbered bytes
318 in words have the highest numeric significance and little-endian if the
321 This code can work in either order irrespective of the order used by the
322 machine on which it runs. Normally the internal byte order will be set
323 to the order of the processor on which the code is to be run but this
324 define can be used to reverse this in special situations
326 NOTE: Assembler code versions rely on PLATFORM_BYTE_ORDER being set
328 #if 1 || defined(AES_ASM)
329 #define ALGORITHM_BYTE_ORDER PLATFORM_BYTE_ORDER
331 #define ALGORITHM_BYTE_ORDER BRG_LITTLE_ENDIAN
333 #define ALGORITHM_BYTE_ORDER BRG_BIG_ENDIAN
335 #error The algorithm byte order is not defined
338 /* 4. FAST INPUT/OUTPUT OPERATIONS.
340 On some machines it is possible to improve speed by transferring the
341 bytes in the input and output arrays to and from the internal 32-bit
342 variables by addressing these arrays as if they are arrays of 32-bit
343 words. On some machines this will always be possible but there may
344 be a large performance penalty if the byte arrays are not aligned on
345 the normal word boundaries. On other machines this technique will
346 lead to memory access errors when such 32-bit word accesses are not
347 properly aligned. The option SAFE_IO avoids such problems but will
348 often be slower on those machines that support misaligned access
349 (especially so if care is taken to align the input and output byte
350 arrays on 32-bit word boundaries). If SAFE_IO is not defined it is
351 assumed that access to byte arrays as if they are arrays of 32-bit
352 words will not cause problems when such accesses are misaligned.
354 #if 1 && !defined(_MSC_VER)
360 The code for encryption and decrytpion cycles through a number of rounds
361 that can be implemented either in a loop or by expanding the code into a
362 long sequence of instructions, the latter producing a larger program but
363 one that will often be much faster. The latter is called loop unrolling.
364 There are also potential speed advantages in expanding two iterations in
365 a loop with half the number of iterations, which is called partial loop
366 unrolling. The following options allow partial or full loop unrolling
367 to be set independently for encryption and decryption
370 #define ENC_UNROLL FULL
372 #define ENC_UNROLL PARTIAL
374 #define ENC_UNROLL NONE
378 #define DEC_UNROLL FULL
380 #define DEC_UNROLL PARTIAL
382 #define DEC_UNROLL NONE
385 /* 6. FAST FINITE FIELD OPERATIONS
387 If this section is included, tables are used to provide faster finite
388 field arithmetic (this has no effect if FIXED_TABLES is defined).
394 /* 7. INTERNAL STATE VARIABLE FORMAT
396 The internal state of Rijndael is stored in a number of local 32-bit
397 word varaibles which can be defined either as an array or as individual
398 names variables. Include this section if you want to store these local
399 varaibles in arrays. Otherwise individual local variables will be used.
405 /* In this implementation the columns of the state array are each held in
406 32-bit words. The state array can be held in various ways: in an array
407 of words, in a number of individual word variables or in a number of
408 processor registers. The following define maps a variable name x and
409 a column number c to the way the state array variable is to be held.
410 The first define below maps the state into an array x[c] whereas the
411 second form maps the state into a number of individual variables x0,
412 x1, etc. Another form could map individual state colums to machine
422 /* 8. FIXED OR DYNAMIC TABLES
424 When this section is included the tables used by the code are compiled
425 statically into the binary file. Otherwise the subroutine gen_tabs()
426 must be called to compute them before the code is first used.
432 /* 9. TABLE ALIGNMENT
434 On some sytsems speed will be improved by aligning the AES large lookup
435 tables on particular boundaries. This define should be set to a power of
436 two giving the desired alignment. It can be left undefined if alignment
437 is not needed. This option is specific to the Microsft VC++ compiler -
438 it seems to sometimes cause trouble for the VC++ version 6 compiler.
441 #if 0 && defined(_MSC_VER) && (_MSC_VER >= 1300)
442 #define TABLE_ALIGN 64
445 /* 10. INTERNAL TABLE CONFIGURATION
447 This cipher proceeds by repeating in a number of cycles known as 'rounds'
448 which are implemented by a round function which can optionally be speeded
449 up using tables. The basic tables are each 256 32-bit words, with either
450 one or four tables being required for each round function depending on
451 how much speed is required. The encryption and decryption round functions
452 are different and the last encryption and decrytpion round functions are
453 different again making four different round functions in all.
456 1. Normal encryption and decryption rounds can each use either 0, 1
457 or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
458 2. The last encryption and decryption rounds can also use either 0, 1
459 or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
461 Include or exclude the appropriate definitions below to set the number
462 of tables used by this implementation.
465 #if 1 /* set tables for the normal encryption round */
466 #define ENC_ROUND FOUR_TABLES
468 #define ENC_ROUND ONE_TABLE
470 #define ENC_ROUND NO_TABLES
473 #if 1 /* set tables for the last encryption round */
474 #define LAST_ENC_ROUND FOUR_TABLES
476 #define LAST_ENC_ROUND ONE_TABLE
478 #define LAST_ENC_ROUND NO_TABLES
481 #if 1 /* set tables for the normal decryption round */
482 #define DEC_ROUND FOUR_TABLES
484 #define DEC_ROUND ONE_TABLE
486 #define DEC_ROUND NO_TABLES
489 #if 1 /* set tables for the last decryption round */
490 #define LAST_DEC_ROUND FOUR_TABLES
492 #define LAST_DEC_ROUND ONE_TABLE
494 #define LAST_DEC_ROUND NO_TABLES
497 /* The decryption key schedule can be speeded up with tables in the same
498 way that the round functions can. Include or exclude the following
499 defines to set this requirement.
502 #define KEY_SCHED FOUR_TABLES
504 #define KEY_SCHED ONE_TABLE
506 #define KEY_SCHED NO_TABLES
509 /* END OF CONFIGURATION OPTIONS */
511 #define RC_LENGTH (5 * (AES_BLOCK_SIZE / 4 - 2))
513 /* Disable or report errors on some combinations of options */
515 #if ENC_ROUND == NO_TABLES && LAST_ENC_ROUND != NO_TABLES
516 #undef LAST_ENC_ROUND
517 #define LAST_ENC_ROUND NO_TABLES
518 #elif ENC_ROUND == ONE_TABLE && LAST_ENC_ROUND == FOUR_TABLES
519 #undef LAST_ENC_ROUND
520 #define LAST_ENC_ROUND ONE_TABLE
523 #if ENC_ROUND == NO_TABLES && ENC_UNROLL != NONE
525 #define ENC_UNROLL NONE
528 #if DEC_ROUND == NO_TABLES && LAST_DEC_ROUND != NO_TABLES
529 #undef LAST_DEC_ROUND
530 #define LAST_DEC_ROUND NO_TABLES
531 #elif DEC_ROUND == ONE_TABLE && LAST_DEC_ROUND == FOUR_TABLES
532 #undef LAST_DEC_ROUND
533 #define LAST_DEC_ROUND ONE_TABLE
536 #if DEC_ROUND == NO_TABLES && DEC_UNROLL != NONE
538 #define DEC_UNROLL NONE
541 /* upr(x,n): rotates bytes within words by n positions, moving bytes to
542 higher index positions with wrap around into low positions
543 ups(x,n): moves bytes by n positions to higher index positions in
544 words but without wrap around
545 bval(x,n): extracts a byte from a word
547 NOTE: The definitions given here are intended only for use with
548 unsigned variables and with shift counts that are compile
552 #if (ALGORITHM_BYTE_ORDER == BRG_LITTLE_ENDIAN)
553 #define upr(x,n) (((aes_32t)(x) << (8 * (n))) | ((aes_32t)(x) >> (32 - 8 * (n))))
554 #define ups(x,n) ((aes_32t) (x) << (8 * (n)))
555 #define bval(x,n) ((aes_08t)((x) >> (8 * (n))))
556 #define bytes2word(b0, b1, b2, b3) \
557 (((aes_32t)(b3) << 24) | ((aes_32t)(b2) << 16) | ((aes_32t)(b1) << 8) | (b0))
560 #if (ALGORITHM_BYTE_ORDER == BRG_BIG_ENDIAN)
561 #define upr(x,n) (((aes_32t)(x) >> (8 * (n))) | ((aes_32t)(x) << (32 - 8 * (n))))
562 #define ups(x,n) ((aes_32t) (x) >> (8 * (n))))
563 #define bval(x,n) ((aes_08t)((x) >> (24 - 8 * (n))))
564 #define bytes2word(b0, b1, b2, b3) \
565 (((aes_32t)(b0) << 24) | ((aes_32t)(b1) << 16) | ((aes_32t)(b2) << 8) | (b3))
570 #define word_in(x,c) bytes2word(((aes_08t*)(x)+4*c)[0], ((aes_08t*)(x)+4*c)[1], \
571 ((aes_08t*)(x)+4*c)[2], ((aes_08t*)(x)+4*c)[3])
572 #define word_out(x,c,v) { ((aes_08t*)(x)+4*c)[0] = bval(v,0); ((aes_08t*)(x)+4*c)[1] = bval(v,1); \
573 ((aes_08t*)(x)+4*c)[2] = bval(v,2); ((aes_08t*)(x)+4*c)[3] = bval(v,3); }
575 #elif (ALGORITHM_BYTE_ORDER == PLATFORM_BYTE_ORDER)
577 #define word_in(x,c) (*((aes_32t*)(x)+(c)))
578 #define word_out(x,c,v) (*((aes_32t*)(x)+(c)) = (v))
582 #define word_in(x,c) aes_sw32(*((aes_32t*)(x)+(c)))
583 #define word_out(x,c,v) (*((aes_32t*)(x)+(c)) = aes_sw32(v))
587 /* the finite field modular polynomial and elements */
592 /* multiply four bytes in GF(2^8) by 'x' {02} in parallel */
594 #define m1 0x80808080
595 #define m2 0x7f7f7f7f
596 #define gf_mulx(x) ((((x) & m2) << 1) ^ ((((x) & m1) >> 7) * BPOLY))
598 /* The following defines provide alternative definitions of gf_mulx that might
599 give improved performance if a fast 32-bit multiply is not available. Note
600 that a temporary variable u needs to be defined where gf_mulx is used.
602 #define gf_mulx(x) (u = (x) & m1, u |= (u >> 1), ((x) & m2) << 1) ^ ((u >> 3) | (u >> 6))
603 #define m4 (0x01010101 * BPOLY)
604 #define gf_mulx(x) (u = (x) & m1, ((x) & m2) << 1) ^ ((u - (u >> 7)) & m4)
607 /* Work out which tables are needed for the different options */
613 #define ENC_ROUND FOUR_TABLES
614 #ifdef LAST_ENC_ROUND
615 #undef LAST_ENC_ROUND
617 #define LAST_ENC_ROUND FOUR_TABLES
621 #define DEC_ROUND FOUR_TABLES
622 #ifdef LAST_DEC_ROUND
623 #undef LAST_DEC_ROUND
625 #define LAST_DEC_ROUND FOUR_TABLES
628 #define KEY_SCHED FOUR_TABLES
632 #if defined(ENCRYPTION) || defined(AES_ASM)
633 #if ENC_ROUND == ONE_TABLE
635 #elif ENC_ROUND == FOUR_TABLES
640 #if LAST_ENC_ROUND == ONE_TABLE
642 #elif LAST_ENC_ROUND == FOUR_TABLES
644 #elif !defined(SBX_SET)
649 #if defined(DECRYPTION) || defined(AES_ASM)
650 #if DEC_ROUND == ONE_TABLE
652 #elif DEC_ROUND == FOUR_TABLES
657 #if LAST_DEC_ROUND == ONE_TABLE
659 #elif LAST_DEC_ROUND == FOUR_TABLES
661 #elif !defined(ISB_SET)
666 #if defined(ENCRYPTION_KEY_SCHEDULE) || defined(DECRYPTION_KEY_SCHEDULE)
667 #if KEY_SCHED == ONE_TABLE
670 #elif KEY_SCHED == FOUR_TABLES
673 #elif !defined(SBX_SET)
678 /* generic definitions of Rijndael macros that use tables */
680 #define no_table(x,box,vf,rf,c) bytes2word( \
681 box[bval(vf(x,0,c),rf(0,c))], \
682 box[bval(vf(x,1,c),rf(1,c))], \
683 box[bval(vf(x,2,c),rf(2,c))], \
684 box[bval(vf(x,3,c),rf(3,c))])
686 #define one_table(x,op,tab,vf,rf,c) \
687 ( tab[bval(vf(x,0,c),rf(0,c))] \
688 ^ op(tab[bval(vf(x,1,c),rf(1,c))],1) \
689 ^ op(tab[bval(vf(x,2,c),rf(2,c))],2) \
690 ^ op(tab[bval(vf(x,3,c),rf(3,c))],3))
692 #define four_tables(x,tab,vf,rf,c) \
693 ( tab[0][bval(vf(x,0,c),rf(0,c))] \
694 ^ tab[1][bval(vf(x,1,c),rf(1,c))] \
695 ^ tab[2][bval(vf(x,2,c),rf(2,c))] \
696 ^ tab[3][bval(vf(x,3,c),rf(3,c))])
698 #define vf1(x,r,c) (x)
700 #define rf2(r,c) ((8+r-c)&3)
702 /* perform forward and inverse column mix operation on four bytes in long word x in */
703 /* parallel. NOTE: x must be a simple variable, NOT an expression in these macros. */
705 #if defined(FM4_SET) /* not currently used */
706 #define fwd_mcol(x) four_tables(x,t_use(f,m),vf1,rf1,0)
707 #elif defined(FM1_SET) /* not currently used */
708 #define fwd_mcol(x) one_table(x,upr,t_use(f,m),vf1,rf1,0)
710 #define dec_fmvars aes_32t g2
711 #define fwd_mcol(x) (g2 = gf_mulx(x), g2 ^ upr((x) ^ g2, 3) ^ upr((x), 2) ^ upr((x), 1))
715 #define inv_mcol(x) four_tables(x,t_use(i,m),vf1,rf1,0)
716 #elif defined(IM1_SET)
717 #define inv_mcol(x) one_table(x,upr,t_use(i,m),vf1,rf1,0)
719 #define dec_imvars aes_32t g2, g4, g9
720 #define inv_mcol(x) (g2 = gf_mulx(x), g4 = gf_mulx(g2), g9 = (x) ^ gf_mulx(g4), g4 ^= g9, \
721 (x) ^ g2 ^ g4 ^ upr(g2 ^ g9, 3) ^ upr(g4, 2) ^ upr(g9, 1))
725 #define ls_box(x,c) four_tables(x,t_use(f,l),vf1,rf2,c)
726 #elif defined(LS4_SET)
727 #define ls_box(x,c) four_tables(x,t_use(l,s),vf1,rf2,c)
728 #elif defined(FL1_SET)
729 #define ls_box(x,c) one_table(x,upr,t_use(f,l),vf1,rf2,c)
730 #elif defined(LS1_SET)
731 #define ls_box(x,c) one_table(x,upr,t_use(l,s),vf1,rf2,c)
733 #define ls_box(x,c) no_table(x,t_use(s,box),vf1,rf2,c)
736 #if defined(__cplusplus)
741 /* If there are no global variables, the definitions here can be
742 used to put the AES tables in a structure so that a pointer
743 can then be added to the AES context to pass them to the AES
744 routines that need them. If this facility is used, the calling
745 program has to ensure that this pointer is managed appropriately.
746 In particular, the value of the t_dec(in,it) item in the table
747 structure must be set to zero in order to ensure that the tables
748 are initialised. In practice the three code sequences in aeskey.c
749 that control the calls to gen_tabs() and the gen_tabs() routine
750 itself will have to be changed for a specific implementation. If
751 global variables are available it will generally be preferable to
752 use them with the precomputed FIXED_TABLES option that uses static
755 The following defines can be used to control the way the tables
756 are defined, initialised and used in embedded environments that
757 require special features for these purposes
759 the 't_dec' construction is used to declare fixed table arrays
760 the 't_set' construction is used to set fixed table values
761 the 't_use' construction is used to access fixed table values
765 t_xxx(s,box) => forward S box
766 t_xxx(i,box) => inverse S box
768 256 32-bit word OR 4 x 256 32-bit word tables:
770 t_xxx(f,n) => forward normal round
771 t_xxx(f,l) => forward last round
772 t_xxx(i,n) => inverse normal round
773 t_xxx(i,l) => inverse last round
774 t_xxx(l,s) => key schedule table
775 t_xxx(i,m) => key schedule table
777 Other variables and tables:
779 t_xxx(r,c) => the rcon table
782 #define t_dec(m,n) t_##m##n
783 #define t_set(m,n) t_##m##n
784 #define t_use(m,n) t_##m##n
786 #if defined(DO_TABLES) /* declare and instantiate tables */
788 /* finite field arithmetic operations for table generation */
790 #if defined(FIXED_TABLES) || !defined(FF_TABLES)
792 #define f2(x) ((x<<1) ^ (((x>>7) & 1) * WPOLY))
793 #define f4(x) ((x<<2) ^ (((x>>6) & 1) * WPOLY) ^ (((x>>6) & 2) * WPOLY))
794 #define f8(x) ((x<<3) ^ (((x>>5) & 1) * WPOLY) ^ (((x>>5) & 2) * WPOLY) \
795 ^ (((x>>5) & 4) * WPOLY))
796 #define f3(x) (f2(x) ^ x)
797 #define f9(x) (f8(x) ^ x)
798 #define fb(x) (f8(x) ^ f2(x) ^ x)
799 #define fd(x) (f8(x) ^ f4(x) ^ x)
800 #define fe(x) (f8(x) ^ f4(x) ^ f2(x))
804 #define f2(x) ((x) ? pow[log[x] + 0x19] : 0)
805 #define f3(x) ((x) ? pow[log[x] + 0x01] : 0)
806 #define f9(x) ((x) ? pow[log[x] + 0xc7] : 0)
807 #define fb(x) ((x) ? pow[log[x] + 0x68] : 0)
808 #define fd(x) ((x) ? pow[log[x] + 0xee] : 0)
809 #define fe(x) ((x) ? pow[log[x] + 0xdf] : 0)
810 #define fi(x) ((x) ? pow[ 255 - log[x]] : 0)
814 #if defined(FIXED_TABLES) /* declare and set values for static tables */
817 w(0x63), w(0x7c), w(0x77), w(0x7b), w(0xf2), w(0x6b), w(0x6f), w(0xc5),\
818 w(0x30), w(0x01), w(0x67), w(0x2b), w(0xfe), w(0xd7), w(0xab), w(0x76),\
819 w(0xca), w(0x82), w(0xc9), w(0x7d), w(0xfa), w(0x59), w(0x47), w(0xf0),\
820 w(0xad), w(0xd4), w(0xa2), w(0xaf), w(0x9c), w(0xa4), w(0x72), w(0xc0),\
821 w(0xb7), w(0xfd), w(0x93), w(0x26), w(0x36), w(0x3f), w(0xf7), w(0xcc),\
822 w(0x34), w(0xa5), w(0xe5), w(0xf1), w(0x71), w(0xd8), w(0x31), w(0x15),\
823 w(0x04), w(0xc7), w(0x23), w(0xc3), w(0x18), w(0x96), w(0x05), w(0x9a),\
824 w(0x07), w(0x12), w(0x80), w(0xe2), w(0xeb), w(0x27), w(0xb2), w(0x75),\
825 w(0x09), w(0x83), w(0x2c), w(0x1a), w(0x1b), w(0x6e), w(0x5a), w(0xa0),\
826 w(0x52), w(0x3b), w(0xd6), w(0xb3), w(0x29), w(0xe3), w(0x2f), w(0x84),\
827 w(0x53), w(0xd1), w(0x00), w(0xed), w(0x20), w(0xfc), w(0xb1), w(0x5b),\
828 w(0x6a), w(0xcb), w(0xbe), w(0x39), w(0x4a), w(0x4c), w(0x58), w(0xcf),\
829 w(0xd0), w(0xef), w(0xaa), w(0xfb), w(0x43), w(0x4d), w(0x33), w(0x85),\
830 w(0x45), w(0xf9), w(0x02), w(0x7f), w(0x50), w(0x3c), w(0x9f), w(0xa8),\
831 w(0x51), w(0xa3), w(0x40), w(0x8f), w(0x92), w(0x9d), w(0x38), w(0xf5),\
832 w(0xbc), w(0xb6), w(0xda), w(0x21), w(0x10), w(0xff), w(0xf3), w(0xd2),\
833 w(0xcd), w(0x0c), w(0x13), w(0xec), w(0x5f), w(0x97), w(0x44), w(0x17),\
834 w(0xc4), w(0xa7), w(0x7e), w(0x3d), w(0x64), w(0x5d), w(0x19), w(0x73),\
835 w(0x60), w(0x81), w(0x4f), w(0xdc), w(0x22), w(0x2a), w(0x90), w(0x88),\
836 w(0x46), w(0xee), w(0xb8), w(0x14), w(0xde), w(0x5e), w(0x0b), w(0xdb),\
837 w(0xe0), w(0x32), w(0x3a), w(0x0a), w(0x49), w(0x06), w(0x24), w(0x5c),\
838 w(0xc2), w(0xd3), w(0xac), w(0x62), w(0x91), w(0x95), w(0xe4), w(0x79),\
839 w(0xe7), w(0xc8), w(0x37), w(0x6d), w(0x8d), w(0xd5), w(0x4e), w(0xa9),\
840 w(0x6c), w(0x56), w(0xf4), w(0xea), w(0x65), w(0x7a), w(0xae), w(0x08),\
841 w(0xba), w(0x78), w(0x25), w(0x2e), w(0x1c), w(0xa6), w(0xb4), w(0xc6),\
842 w(0xe8), w(0xdd), w(0x74), w(0x1f), w(0x4b), w(0xbd), w(0x8b), w(0x8a),\
843 w(0x70), w(0x3e), w(0xb5), w(0x66), w(0x48), w(0x03), w(0xf6), w(0x0e),\
844 w(0x61), w(0x35), w(0x57), w(0xb9), w(0x86), w(0xc1), w(0x1d), w(0x9e),\
845 w(0xe1), w(0xf8), w(0x98), w(0x11), w(0x69), w(0xd9), w(0x8e), w(0x94),\
846 w(0x9b), w(0x1e), w(0x87), w(0xe9), w(0xce), w(0x55), w(0x28), w(0xdf),\
847 w(0x8c), w(0xa1), w(0x89), w(0x0d), w(0xbf), w(0xe6), w(0x42), w(0x68),\
848 w(0x41), w(0x99), w(0x2d), w(0x0f), w(0xb0), w(0x54), w(0xbb), w(0x16)
850 #define isb_data(w) \
851 w(0x52), w(0x09), w(0x6a), w(0xd5), w(0x30), w(0x36), w(0xa5), w(0x38),\
852 w(0xbf), w(0x40), w(0xa3), w(0x9e), w(0x81), w(0xf3), w(0xd7), w(0xfb),\
853 w(0x7c), w(0xe3), w(0x39), w(0x82), w(0x9b), w(0x2f), w(0xff), w(0x87),\
854 w(0x34), w(0x8e), w(0x43), w(0x44), w(0xc4), w(0xde), w(0xe9), w(0xcb),\
855 w(0x54), w(0x7b), w(0x94), w(0x32), w(0xa6), w(0xc2), w(0x23), w(0x3d),\
856 w(0xee), w(0x4c), w(0x95), w(0x0b), w(0x42), w(0xfa), w(0xc3), w(0x4e),\
857 w(0x08), w(0x2e), w(0xa1), w(0x66), w(0x28), w(0xd9), w(0x24), w(0xb2),\
858 w(0x76), w(0x5b), w(0xa2), w(0x49), w(0x6d), w(0x8b), w(0xd1), w(0x25),\
859 w(0x72), w(0xf8), w(0xf6), w(0x64), w(0x86), w(0x68), w(0x98), w(0x16),\
860 w(0xd4), w(0xa4), w(0x5c), w(0xcc), w(0x5d), w(0x65), w(0xb6), w(0x92),\
861 w(0x6c), w(0x70), w(0x48), w(0x50), w(0xfd), w(0xed), w(0xb9), w(0xda),\
862 w(0x5e), w(0x15), w(0x46), w(0x57), w(0xa7), w(0x8d), w(0x9d), w(0x84),\
863 w(0x90), w(0xd8), w(0xab), w(0x00), w(0x8c), w(0xbc), w(0xd3), w(0x0a),\
864 w(0xf7), w(0xe4), w(0x58), w(0x05), w(0xb8), w(0xb3), w(0x45), w(0x06),\
865 w(0xd0), w(0x2c), w(0x1e), w(0x8f), w(0xca), w(0x3f), w(0x0f), w(0x02),\
866 w(0xc1), w(0xaf), w(0xbd), w(0x03), w(0x01), w(0x13), w(0x8a), w(0x6b),\
867 w(0x3a), w(0x91), w(0x11), w(0x41), w(0x4f), w(0x67), w(0xdc), w(0xea),\
868 w(0x97), w(0xf2), w(0xcf), w(0xce), w(0xf0), w(0xb4), w(0xe6), w(0x73),\
869 w(0x96), w(0xac), w(0x74), w(0x22), w(0xe7), w(0xad), w(0x35), w(0x85),\
870 w(0xe2), w(0xf9), w(0x37), w(0xe8), w(0x1c), w(0x75), w(0xdf), w(0x6e),\
871 w(0x47), w(0xf1), w(0x1a), w(0x71), w(0x1d), w(0x29), w(0xc5), w(0x89),\
872 w(0x6f), w(0xb7), w(0x62), w(0x0e), w(0xaa), w(0x18), w(0xbe), w(0x1b),\
873 w(0xfc), w(0x56), w(0x3e), w(0x4b), w(0xc6), w(0xd2), w(0x79), w(0x20),\
874 w(0x9a), w(0xdb), w(0xc0), w(0xfe), w(0x78), w(0xcd), w(0x5a), w(0xf4),\
875 w(0x1f), w(0xdd), w(0xa8), w(0x33), w(0x88), w(0x07), w(0xc7), w(0x31),\
876 w(0xb1), w(0x12), w(0x10), w(0x59), w(0x27), w(0x80), w(0xec), w(0x5f),\
877 w(0x60), w(0x51), w(0x7f), w(0xa9), w(0x19), w(0xb5), w(0x4a), w(0x0d),\
878 w(0x2d), w(0xe5), w(0x7a), w(0x9f), w(0x93), w(0xc9), w(0x9c), w(0xef),\
879 w(0xa0), w(0xe0), w(0x3b), w(0x4d), w(0xae), w(0x2a), w(0xf5), w(0xb0),\
880 w(0xc8), w(0xeb), w(0xbb), w(0x3c), w(0x83), w(0x53), w(0x99), w(0x61),\
881 w(0x17), w(0x2b), w(0x04), w(0x7e), w(0xba), w(0x77), w(0xd6), w(0x26),\
882 w(0xe1), w(0x69), w(0x14), w(0x63), w(0x55), w(0x21), w(0x0c), w(0x7d),
885 w(0x00), w(0x01), w(0x02), w(0x03), w(0x04), w(0x05), w(0x06), w(0x07),\
886 w(0x08), w(0x09), w(0x0a), w(0x0b), w(0x0c), w(0x0d), w(0x0e), w(0x0f),\
887 w(0x10), w(0x11), w(0x12), w(0x13), w(0x14), w(0x15), w(0x16), w(0x17),\
888 w(0x18), w(0x19), w(0x1a), w(0x1b), w(0x1c), w(0x1d), w(0x1e), w(0x1f),\
889 w(0x20), w(0x21), w(0x22), w(0x23), w(0x24), w(0x25), w(0x26), w(0x27),\
890 w(0x28), w(0x29), w(0x2a), w(0x2b), w(0x2c), w(0x2d), w(0x2e), w(0x2f),\
891 w(0x30), w(0x31), w(0x32), w(0x33), w(0x34), w(0x35), w(0x36), w(0x37),\
892 w(0x38), w(0x39), w(0x3a), w(0x3b), w(0x3c), w(0x3d), w(0x3e), w(0x3f),\
893 w(0x40), w(0x41), w(0x42), w(0x43), w(0x44), w(0x45), w(0x46), w(0x47),\
894 w(0x48), w(0x49), w(0x4a), w(0x4b), w(0x4c), w(0x4d), w(0x4e), w(0x4f),\
895 w(0x50), w(0x51), w(0x52), w(0x53), w(0x54), w(0x55), w(0x56), w(0x57),\
896 w(0x58), w(0x59), w(0x5a), w(0x5b), w(0x5c), w(0x5d), w(0x5e), w(0x5f),\
897 w(0x60), w(0x61), w(0x62), w(0x63), w(0x64), w(0x65), w(0x66), w(0x67),\
898 w(0x68), w(0x69), w(0x6a), w(0x6b), w(0x6c), w(0x6d), w(0x6e), w(0x6f),\
899 w(0x70), w(0x71), w(0x72), w(0x73), w(0x74), w(0x75), w(0x76), w(0x77),\
900 w(0x78), w(0x79), w(0x7a), w(0x7b), w(0x7c), w(0x7d), w(0x7e), w(0x7f),\
901 w(0x80), w(0x81), w(0x82), w(0x83), w(0x84), w(0x85), w(0x86), w(0x87),\
902 w(0x88), w(0x89), w(0x8a), w(0x8b), w(0x8c), w(0x8d), w(0x8e), w(0x8f),\
903 w(0x90), w(0x91), w(0x92), w(0x93), w(0x94), w(0x95), w(0x96), w(0x97),\
904 w(0x98), w(0x99), w(0x9a), w(0x9b), w(0x9c), w(0x9d), w(0x9e), w(0x9f),\
905 w(0xa0), w(0xa1), w(0xa2), w(0xa3), w(0xa4), w(0xa5), w(0xa6), w(0xa7),\
906 w(0xa8), w(0xa9), w(0xaa), w(0xab), w(0xac), w(0xad), w(0xae), w(0xaf),\
907 w(0xb0), w(0xb1), w(0xb2), w(0xb3), w(0xb4), w(0xb5), w(0xb6), w(0xb7),\
908 w(0xb8), w(0xb9), w(0xba), w(0xbb), w(0xbc), w(0xbd), w(0xbe), w(0xbf),\
909 w(0xc0), w(0xc1), w(0xc2), w(0xc3), w(0xc4), w(0xc5), w(0xc6), w(0xc7),\
910 w(0xc8), w(0xc9), w(0xca), w(0xcb), w(0xcc), w(0xcd), w(0xce), w(0xcf),\
911 w(0xd0), w(0xd1), w(0xd2), w(0xd3), w(0xd4), w(0xd5), w(0xd6), w(0xd7),\
912 w(0xd8), w(0xd9), w(0xda), w(0xdb), w(0xdc), w(0xdd), w(0xde), w(0xdf),\
913 w(0xe0), w(0xe1), w(0xe2), w(0xe3), w(0xe4), w(0xe5), w(0xe6), w(0xe7),\
914 w(0xe8), w(0xe9), w(0xea), w(0xeb), w(0xec), w(0xed), w(0xee), w(0xef),\
915 w(0xf0), w(0xf1), w(0xf2), w(0xf3), w(0xf4), w(0xf5), w(0xf6), w(0xf7),\
916 w(0xf8), w(0xf9), w(0xfa), w(0xfb), w(0xfc), w(0xfd), w(0xfe), w(0xff)
920 /* These defines are used to ensure tables are generated in the
921 right format depending on the internal byte order required
924 #define w0(p) bytes2word(p, 0, 0, 0)
925 #define w1(p) bytes2word(0, p, 0, 0)
926 #define w2(p) bytes2word(0, 0, p, 0)
927 #define w3(p) bytes2word(0, 0, 0, p)
929 #define u0(p) bytes2word(f2(p), p, p, f3(p))
930 #define u1(p) bytes2word(f3(p), f2(p), p, p)
931 #define u2(p) bytes2word(p, f3(p), f2(p), p)
932 #define u3(p) bytes2word(p, p, f3(p), f2(p))
934 #define v0(p) bytes2word(fe(p), f9(p), fd(p), fb(p))
935 #define v1(p) bytes2word(fb(p), fe(p), f9(p), fd(p))
936 #define v2(p) bytes2word(fd(p), fb(p), fe(p), f9(p))
937 #define v3(p) bytes2word(f9(p), fd(p), fb(p), fe(p))
939 const aes_32t t_dec(r,c)[RC_LENGTH] =
941 w0(0x01), w0(0x02), w0(0x04), w0(0x08), w0(0x10),
942 w0(0x20), w0(0x40), w0(0x80), w0(0x1b), w0(0x36)
945 #define d_1(t,n,b,v) const t n[256] = { b(v##0) }
946 #define d_4(t,n,b,v) const t n[4][256] = { { b(v##0) }, { b(v##1) }, { b(v##2) }, { b(v##3) } }
948 #else /* declare and instantiate tables for dynamic value generation in in tab.c */
950 aes_32t t_dec(r,c)[RC_LENGTH];
952 #define d_1(t,n,b,v) t n[256]
953 #define d_4(t,n,b,v) t n[4][256]
957 #else /* declare tables without instantiation */
959 #if defined(FIXED_TABLES)
961 extern const aes_32t t_dec(r,c)[RC_LENGTH];
963 #if defined(_MSC_VER) && defined(TABLE_ALIGN)
964 #define d_1(t,n,b,v) extern __declspec(align(TABLE_ALIGN)) const t n[256]
965 #define d_4(t,n,b,v) extern __declspec(align(TABLE_ALIGN)) const t n[4][256]
967 #define d_1(t,n,b,v) extern const t n[256]
968 #define d_4(t,n,b,v) extern const t n[4][256]
972 extern aes_32t t_dec(r,c)[RC_LENGTH];
974 #if defined(_MSC_VER) && defined(TABLE_ALIGN)
975 #define d_1(t,n,b,v) extern __declspec(align(TABLE_ALIGN)) t n[256]
976 #define d_4(t,n,b,v) extern __declspec(align(TABLE_ALIGN)) t n[4][256]
978 #define d_1(t,n,b,v) extern t n[256]
979 #define d_4(t,n,b,v) extern t n[4][256]
986 d_1(aes_08t, t_dec(s,box), sb_data, h);
989 d_1(aes_08t, t_dec(i,box), isb_data, h);
993 d_1(aes_32t, t_dec(f,n), sb_data, u);
996 d_4(aes_32t, t_dec(f,n), sb_data, u);
1000 d_1(aes_32t, t_dec(f,l), sb_data, w);
1003 d_4(aes_32t, t_dec(f,l), sb_data, w);
1007 d_1(aes_32t, t_dec(i,n), isb_data, v);
1010 d_4(aes_32t, t_dec(i,n), isb_data, v);
1014 d_1(aes_32t, t_dec(i,l), isb_data, w);
1017 d_4(aes_32t, t_dec(i,l), isb_data, w);
1024 d_1(aes_32t, t_dec(l,s), sb_data, w);
1032 d_4(aes_32t, t_dec(l,s), sb_data, w);
1037 d_1(aes_32t, t_dec(i,m), mm_data, v);
1040 d_4(aes_32t, t_dec(i,m), mm_data, v);
1043 #if defined(__cplusplus)