aesopt.h

老外写的加密库cryptlib(版本3.1)

    variables by addressing these arrays as if they are arrays of 32-bit
    words.  On some machines this will always be possible but there may
    be a large performance penalty if the byte arrays are not aligned on
    the normal word boundaries. On other machines this technique will
    lead to memory access errors when such 32-bit word accesses are not
    properly aligned. The option SAFE_IO avoids such problems but will
    often be slower on those machines that support misaligned access
    (especially so if care is taken to align the input  and output byte
    arrays on 32-bit word boundaries). If SAFE_IO is not defined it is
    assumed that access to byte arrays as if they are arrays of 32-bit
    words will not cause problems when such accesses are misaligned.
*/
#if 1 && !defined(_MSC_VER)
#define SAFE_IO
#endif

/*  5. LOOP UNROLLING

    The code for encryption and decrytpion cycles through a number of rounds
    that can be implemented either in a loop or by expanding the code into a
    long sequence of instructions, the latter producing a larger program but
    one that will often be much faster. The latter is called loop unrolling.
    There are also potential speed advantages in expanding two iterations in
    a loop with half the number of iterations, which is called partial loop
    unrolling.  The following options allow partial or full loop unrolling
    to be set independently for encryption and decryption
*/
#if 1
#define ENC_UNROLL  FULL
#elif 0
#define ENC_UNROLL  PARTIAL
#else
#define ENC_UNROLL  NONE
#endif

#if 1
#define DEC_UNROLL  FULL
#elif 0
#define DEC_UNROLL  PARTIAL
#else
#define DEC_UNROLL  NONE
#endif

/*  6. FAST FINITE FIELD OPERATIONS

    If this section is included, tables are used to provide faster finite
    field arithmetic (this has no effect if FIXED_TABLES is defined).
*/
#if 1
#define FF_TABLES
#endif

/*  7. INTERNAL STATE VARIABLE FORMAT

    The internal state of Rijndael is stored in a number of local 32-bit
    word varaibles which can be defined either as an array or as individual
    names variables. Include this section if you want to store these local
    varaibles in arrays. Otherwise individual local variables will be used.
*/
#if 1
#define ARRAYS
#endif

/* In this implementation the columns of the state array are each held in
   32-bit words. The state array can be held in various ways: in an array
   of words, in a number of individual word variables or in a number of
   processor registers. The following define maps a variable name x and
   a column number c to the way the state array variable is to be held.
   The first define below maps the state into an array x[c] whereas the
   second form maps the state into a number of individual variables x0,
   x1, etc.  Another form could map individual state colums to machine
   register names.
*/

#if defined(ARRAYS)
#define s(x,c) x[c]
#else
#define s(x,c) x##c
#endif

/*  8. FIXED OR DYNAMIC TABLES

    When this section is included the tables used by the code are compiled
    statically into the binary file.  Otherwise the subroutine gen_tabs()
    must be called to compute them before the code is first used.
*/
#if 1
#define FIXED_TABLES
#endif

/*  9. TABLE ALIGNMENT

    On some sytsems speed will be improved by aligning the AES large lookup
    tables on particular boundaries. This define should be set to a power of
    two giving the desired alignment. It can be left undefined if alignment
    is not needed.  This option is specific to the Microsft VC++ compiler -
    it seems to sometimes cause trouble for the VC++ version 6 compiler.
*/

#if 0 && defined(_MSC_VER) && (_MSC_VER >= 1300)
#define TABLE_ALIGN 64
#endif

/*  10. INTERNAL TABLE CONFIGURATION

    This cipher proceeds by repeating in a number of cycles known as 'rounds'
    which are implemented by a round function which can optionally be speeded
    up using tables.  The basic tables are each 256 32-bit words, with either
    one or four tables being required for each round function depending on
    how much speed is required. The encryption and decryption round functions
    are different and the last encryption and decrytpion round functions are
    different again making four different round functions in all.

    This means that:
      1. Normal encryption and decryption rounds can each use either 0, 1
         or 4 tables and table spaces of 0, 1024 or 4096 bytes each.
      2. The last encryption and decryption rounds can also use either 0, 1
         or 4 tables and table spaces of 0, 1024 or 4096 bytes each.

    Include or exclude the appropriate definitions below to set the number
    of tables used by this implementation.
*/

#if 1   /* set tables for the normal encryption round */
#define ENC_ROUND   FOUR_TABLES
#elif 0
#define ENC_ROUND   ONE_TABLE
#else
#define ENC_ROUND   NO_TABLES
#endif

#if 1   /* set tables for the last encryption round */
#define LAST_ENC_ROUND  FOUR_TABLES
#elif 0
#define LAST_ENC_ROUND  ONE_TABLE
#else
#define LAST_ENC_ROUND  NO_TABLES
#endif

#if 1   /* set tables for the normal decryption round */
#define DEC_ROUND   FOUR_TABLES
#elif 0
#define DEC_ROUND   ONE_TABLE
#else
#define DEC_ROUND   NO_TABLES
#endif

#if 1   /* set tables for the last decryption round */
#define LAST_DEC_ROUND  FOUR_TABLES
#elif 0
#define LAST_DEC_ROUND  ONE_TABLE
#else
#define LAST_DEC_ROUND  NO_TABLES
#endif

/*  The decryption key schedule can be speeded up with tables in the same
    way that the round functions can.  Include or exclude the following
    defines to set this requirement.
*/
#if 1
#define KEY_SCHED   FOUR_TABLES
#elif 0
#define KEY_SCHED   ONE_TABLE
#else
#define KEY_SCHED   NO_TABLES
#endif

/* END OF CONFIGURATION OPTIONS */

#define RC_LENGTH   (5 * (AES_BLOCK_SIZE / 4 - 2))

/* Disable or report errors on some combinations of options */

#if ENC_ROUND == NO_TABLES && LAST_ENC_ROUND != NO_TABLES
#undef  LAST_ENC_ROUND
#define LAST_ENC_ROUND  NO_TABLES
#elif ENC_ROUND == ONE_TABLE && LAST_ENC_ROUND == FOUR_TABLES
#undef  LAST_ENC_ROUND
#define LAST_ENC_ROUND  ONE_TABLE
#endif

#if ENC_ROUND == NO_TABLES && ENC_UNROLL != NONE
#undef  ENC_UNROLL
#define ENC_UNROLL  NONE
#endif

#if DEC_ROUND == NO_TABLES && LAST_DEC_ROUND != NO_TABLES
#undef  LAST_DEC_ROUND
#define LAST_DEC_ROUND  NO_TABLES
#elif DEC_ROUND == ONE_TABLE && LAST_DEC_ROUND == FOUR_TABLES
#undef  LAST_DEC_ROUND
#define LAST_DEC_ROUND  ONE_TABLE
#endif

#if DEC_ROUND == NO_TABLES && DEC_UNROLL != NONE
#undef  DEC_UNROLL
#define DEC_UNROLL  NONE
#endif

/*  upr(x,n):  rotates bytes within words by n positions, moving bytes to
               higher index positions with wrap around into low positions
    ups(x,n):  moves bytes by n positions to higher index positions in
               words but without wrap around
    bval(x,n): extracts a byte from a word

    NOTE:      The definitions given here are intended only for use with
               unsigned variables and with shift counts that are compile
               time constants
*/

#if (ALGORITHM_BYTE_ORDER == BRG_LITTLE_ENDIAN)
#define upr(x,n)        (((aes_32t)(x) << (8 * (n))) | ((aes_32t)(x) >> (32 - 8 * (n))))
#define ups(x,n)        ((aes_32t) (x) << (8 * (n)))
#define bval(x,n)       ((aes_08t)((x) >> (8 * (n))))
#define bytes2word(b0, b1, b2, b3)  \
        (((aes_32t)(b3) << 24) | ((aes_32t)(b2) << 16) | ((aes_32t)(b1) << 8) | (b0))
#endif

#if (ALGORITHM_BYTE_ORDER == BRG_BIG_ENDIAN)
#define upr(x,n)        (((aes_32t)(x) >> (8 * (n))) | ((aes_32t)(x) << (32 - 8 * (n))))
#define ups(x,n)        ((aes_32t) (x) >> (8 * (n))))
#define bval(x,n)       ((aes_08t)((x) >> (24 - 8 * (n))))
#define bytes2word(b0, b1, b2, b3)  \
        (((aes_32t)(b0) << 24) | ((aes_32t)(b1) << 16) | ((aes_32t)(b2) << 8) | (b3))
#endif

#if defined(SAFE_IO)

#define word_in(x,c)    bytes2word(((aes_08t*)(x)+4*c)[0], ((aes_08t*)(x)+4*c)[1], \
                                   ((aes_08t*)(x)+4*c)[2], ((aes_08t*)(x)+4*c)[3])
#define word_out(x,c,v) { ((aes_08t*)(x)+4*c)[0] = bval(v,0); ((aes_08t*)(x)+4*c)[1] = bval(v,1); \
                          ((aes_08t*)(x)+4*c)[2] = bval(v,2); ((aes_08t*)(x)+4*c)[3] = bval(v,3); }

#elif (ALGORITHM_BYTE_ORDER == PLATFORM_BYTE_ORDER)

#define word_in(x,c)    (*((aes_32t*)(x)+(c)))
#define word_out(x,c,v) (*((aes_32t*)(x)+(c)) = (v))

#else

#define word_in(x,c)    aes_sw32(*((aes_32t*)(x)+(c)))
#define word_out(x,c,v) (*((aes_32t*)(x)+(c)) = aes_sw32(v))

#endif

/* the finite field modular polynomial and elements */

#define WPOLY   0x011b
#define BPOLY     0x1b

/* multiply four bytes in GF(2^8) by 'x' {02} in parallel */

#define m1  0x80808080
#define m2  0x7f7f7f7f
#define gf_mulx(x)  ((((x) & m2) << 1) ^ ((((x) & m1) >> 7) * BPOLY))

/* The following defines provide alternative definitions of gf_mulx that might
   give improved performance if a fast 32-bit multiply is not available. Note
   that a temporary variable u needs to be defined where gf_mulx is used.

#define gf_mulx(x) (u = (x) & m1, u |= (u >> 1), ((x) & m2) << 1) ^ ((u >> 3) | (u >> 6))
#define m4  (0x01010101 * BPOLY)
#define gf_mulx(x) (u = (x) & m1, ((x) & m2) << 1) ^ ((u - (u >> 7)) & m4)
*/

/* Work out which tables are needed for the different options   */

#ifdef  AES_ASM
#ifdef  ENC_ROUND
#undef  ENC_ROUND
#endif
#define ENC_ROUND   FOUR_TABLES
#ifdef  LAST_ENC_ROUND
#undef  LAST_ENC_ROUND
#endif
#define LAST_ENC_ROUND  FOUR_TABLES
#ifdef  DEC_ROUND
#undef  DEC_ROUND
#endif
#define DEC_ROUND   FOUR_TABLES
#ifdef  LAST_DEC_ROUND
#undef  LAST_DEC_ROUND
#endif
#define LAST_DEC_ROUND  FOUR_TABLES
#ifdef  KEY_SCHED
#undef  KEY_SCHED
#define KEY_SCHED   FOUR_TABLES
#endif
#endif

#if defined(ENCRYPTION) || defined(AES_ASM)
#if ENC_ROUND == ONE_TABLE
#define FT1_SET
#elif ENC_ROUND == FOUR_TABLES
#define FT4_SET
#else
#define SBX_SET
#endif
#if LAST_ENC_ROUND == ONE_TABLE
#define FL1_SET
#elif LAST_ENC_ROUND == FOUR_TABLES
#define FL4_SET
#elif !defined(SBX_SET)
#define SBX_SET
#endif
#endif

#if defined(DECRYPTION) || defined(AES_ASM)
#if DEC_ROUND == ONE_TABLE
#define IT1_SET
#elif DEC_ROUND == FOUR_TABLES
#define IT4_SET
#else
#define ISB_SET
#endif
#if LAST_DEC_ROUND == ONE_TABLE
#define IL1_SET
#elif LAST_DEC_ROUND == FOUR_TABLES
#define IL4_SET
#elif !defined(ISB_SET)
#define ISB_SET
#endif
#endif

#if defined(ENCRYPTION_KEY_SCHEDULE) || defined(DECRYPTION_KEY_SCHEDULE)
#if KEY_SCHED == ONE_TABLE
#define LS1_SET
#define IM1_SET
#elif KEY_SCHED == FOUR_TABLES
#define LS4_SET
#define IM4_SET
#elif !defined(SBX_SET)
#define SBX_SET
#endif
#endif

/* generic definitions of Rijndael macros that use tables    */

#define no_table(x,box,vf,rf,c) bytes2word( \
    box[bval(vf(x,0,c),rf(0,c))], \
    box[bval(vf(x,1,c),rf(1,c))], \
    box[bval(vf(x,2,c),rf(2,c))], \
    box[bval(vf(x,3,c),rf(3,c))])

#define one_table(x,op,tab,vf,rf,c) \
 (     tab[bval(vf(x,0,c),rf(0,c))] \
  ^ op(tab[bval(vf(x,1,c),rf(1,c))],1) \
  ^ op(tab[bval(vf(x,2,c),rf(2,c))],2) \
  ^ op(tab[bval(vf(x,3,c),rf(3,c))],3))

#define four_tables(x,tab,vf,rf,c) \
 (  tab[0][bval(vf(x,0,c),rf(0,c))] \
  ^ tab[1][bval(vf(x,1,c),rf(1,c))] \
  ^ tab[2][bval(vf(x,2,c),rf(2,c))] \