Print this page
6665607 Need a SHA256/SHA384/SHA512 implementation optimized for 64-bit x86
   1 /*
   2  * Copyright 2007 Sun Microsystems, Inc.  All rights reserved.
   3  * Use is subject to license terms.
   4  */
   5 
   6 #pragma ident   "@(#)sha2.c     1.7     07/04/10 SMI"
   7 
   8 
   9 /*
  10  * The basic framework for this code came from the reference
  11  * implementation for MD5.  That implementation is Copyright (C)
  12  * 1991-2, RSA Data Security, Inc. Created 1991. All rights reserved.
  13  *
  14  * License to copy and use this software is granted provided that it
  15  * is identified as the "RSA Data Security, Inc. MD5 Message-Digest
  16  * Algorithm" in all material mentioning or referencing this software
  17  * or this function.
  18  *
  19  * License is also granted to make and use derivative works provided
  20  * that such works are identified as "derived from the RSA Data
  21  * Security, Inc. MD5 Message-Digest Algorithm" in all material
  22  * mentioning or referencing the derived work.
  23  *
  24  * RSA Data Security, Inc. makes no representations concerning either
  25  * the merchantability of this software or the suitability of this
  26  * software for any particular purpose. It is provided "as is"
  27  * without express or implied warranty of any kind.
  28  *
  29  * These notices must be retained in any copies of any part of this
  30  * documentation and/or software.
  31  *
  32  * NOTE: Cleaned-up and optimized, version of SHA2, based on the FIPS 180-2
  33  * standard, available at http://www.itl.nist.gov/div897/pubs/fip180-2.htm
  34  * Not as fast as one would like -- further optimizations are encouraged
  35  * and appreciated.
  36  */
  37 
  38 #include <sys/types.h>
  39 #include <sys/param.h>
  40 #include <sys/systm.h>
  41 #include <sys/sysmacros.h>
  42 #define _SHA2_IMPL
  43 #include <sys/sha2.h>
  44 #include <sys/sha2_consts.h>
  45 
  46 #ifndef _KERNEL

  47 

  48 #include <strings.h>
  49 #include <stdlib.h>
  50 #include <errno.h>
  51 
  52 #pragma weak SHA256Update = SHA2Update
  53 #pragma weak SHA384Update = SHA2Update
  54 #pragma weak SHA512Update = SHA2Update
  55 
  56 #pragma weak SHA256Final = SHA2Final
  57 #pragma weak SHA384Final = SHA2Final
  58 #pragma weak SHA512Final = SHA2Final
  59 
  60 #endif  /* !_KERNEL */
  61 
  62 #ifdef _KERNEL
  63 #include <sys/cmn_err.h>
  64 #endif /* _KERNEL */
  65 
  66 static void Encode(uint8_t *, uint32_t *, size_t);
  67 static void Encode64(uint8_t *, uint64_t *, size_t);









  68 static void SHA256Transform(SHA2_CTX *, const uint8_t *);
  69 static void SHA512Transform(SHA2_CTX *, const uint8_t *);

  70 
  71 static uint8_t PADDING[128] = { 0x80, /* all zeros */ };
  72 
  73 /* Ch and Maj are the basic SHA2 functions. */
  74 #define Ch(b, c, d)     (((b) & (c)) ^ ((~b) & (d)))
  75 #define Maj(b, c, d)    (((b) & (c)) ^ ((b) & (d)) ^ ((c) & (d)))
  76 
  77 /* Rotates x right n bits. */
  78 #define ROTR(x, n)      \
  79         (((x) >> (n)) | ((x) << ((sizeof (x) * NBBY)-(n))))
  80 
  81 /* Shift x right n bits */
  82 #define SHR(x, n)       ((x) >> (n))
  83 
  84 /* SHA256 Functions */
  85 #define BIGSIGMA0_256(x)        (ROTR((x), 2) ^ ROTR((x), 13) ^ ROTR((x), 22))
  86 #define BIGSIGMA1_256(x)        (ROTR((x), 6) ^ ROTR((x), 11) ^ ROTR((x), 25))
  87 #define SIGMA0_256(x)           (ROTR((x), 7) ^ ROTR((x), 18) ^ SHR((x), 3))
  88 #define SIGMA1_256(x)           (ROTR((x), 17) ^ ROTR((x), 19) ^ SHR((x), 10))
  89 


 118 #define LOAD_BIG_32(addr)       (*(uint32_t *)(addr))
 119 
 120 #else   /* little endian -- will work on big endian, but slowly */
 121 
 122 #define LOAD_BIG_32(addr)       \
 123         (((addr)[0] << 24) | ((addr)[1] << 16) | ((addr)[2] << 8) | (addr)[3])
 124 #endif
 125 
 126 
 127 #if     defined(_BIG_ENDIAN)
 128 
 129 #define LOAD_BIG_64(addr)       (*(uint64_t *)(addr))
 130 
 131 #else   /* little endian -- will work on big endian, but slowly */
 132 
 133 #define LOAD_BIG_64(addr)       \
 134         (((uint64_t)(addr)[0] << 56) | ((uint64_t)(addr)[1] << 48) |        \
 135             ((uint64_t)(addr)[2] << 40) | ((uint64_t)(addr)[3] << 32) |     \
 136             ((uint64_t)(addr)[4] << 24) | ((uint64_t)(addr)[5] << 16) |     \
 137             ((uint64_t)(addr)[6] << 8) | (uint64_t)(addr)[7])
 138 
 139 #endif
 140 
 141 

 142 /* SHA256 Transform */
 143 
 144 static void
 145 SHA256Transform(SHA2_CTX *ctx, const uint8_t *blk)
 146 {
 147 
 148         uint32_t a = ctx->state.s32[0];
 149         uint32_t b = ctx->state.s32[1];
 150         uint32_t c = ctx->state.s32[2];
 151         uint32_t d = ctx->state.s32[3];
 152         uint32_t e = ctx->state.s32[4];
 153         uint32_t f = ctx->state.s32[5];
 154         uint32_t g = ctx->state.s32[6];
 155         uint32_t h = ctx->state.s32[7];
 156 
 157         uint32_t w0, w1, w2, w3, w4, w5, w6, w7;
 158         uint32_t w8, w9, w10, w11, w12, w13, w14, w15;
 159         uint32_t T1, T2;
 160 
 161 #if     defined(__sparc)
 162         static const uint32_t sha256_consts[] = {
 163                 SHA256_CONST_0, SHA256_CONST_1, SHA256_CONST_2,
 164                 SHA256_CONST_3, SHA256_CONST_4, SHA256_CONST_5,
 165                 SHA256_CONST_6, SHA256_CONST_7, SHA256_CONST_8,
 166                 SHA256_CONST_9, SHA256_CONST_10, SHA256_CONST_11,
 167                 SHA256_CONST_12, SHA256_CONST_13, SHA256_CONST_14,
 168                 SHA256_CONST_15, SHA256_CONST_16, SHA256_CONST_17,
 169                 SHA256_CONST_18, SHA256_CONST_19, SHA256_CONST_20,
 170                 SHA256_CONST_21, SHA256_CONST_22, SHA256_CONST_23,
 171                 SHA256_CONST_24, SHA256_CONST_25, SHA256_CONST_26,
 172                 SHA256_CONST_27, SHA256_CONST_28, SHA256_CONST_29,
 173                 SHA256_CONST_30, SHA256_CONST_31, SHA256_CONST_32,
 174                 SHA256_CONST_33, SHA256_CONST_34, SHA256_CONST_35,
 175                 SHA256_CONST_36, SHA256_CONST_37, SHA256_CONST_38,
 176                 SHA256_CONST_39, SHA256_CONST_40, SHA256_CONST_41,
 177                 SHA256_CONST_42, SHA256_CONST_43, SHA256_CONST_44,
 178                 SHA256_CONST_45, SHA256_CONST_46, SHA256_CONST_47,
 179                 SHA256_CONST_48, SHA256_CONST_49, SHA256_CONST_50,
 180                 SHA256_CONST_51, SHA256_CONST_52, SHA256_CONST_53,
 181                 SHA256_CONST_54, SHA256_CONST_55, SHA256_CONST_56,
 182                 SHA256_CONST_57, SHA256_CONST_58, SHA256_CONST_59,
 183                 SHA256_CONST_60, SHA256_CONST_61, SHA256_CONST_62,
 184                 SHA256_CONST_63
 185         };
 186 #endif
 187 
 188         if ((uintptr_t)blk & 0x3) {         /* not 4-byte aligned? */
 189                 bcopy(blk, ctx->buf_un.buf32,  sizeof (ctx->buf_un.buf32));
 190                 blk = (uint8_t *)ctx->buf_un.buf32;
 191         }
 192 
 193         /* LINTED E_BAD_PTR_CAST_ALIGN */
 194         w0 =  LOAD_BIG_32(blk + 4 * 0);
 195         SHA256ROUND(a, b, c, d, e, f, g, h, 0, w0);
 196         /* LINTED E_BAD_PTR_CAST_ALIGN */
 197         w1 =  LOAD_BIG_32(blk + 4 * 1);
 198         SHA256ROUND(h, a, b, c, d, e, f, g, 1, w1);
 199         /* LINTED E_BAD_PTR_CAST_ALIGN */
 200         w2 =  LOAD_BIG_32(blk + 4 * 2);
 201         SHA256ROUND(g, h, a, b, c, d, e, f, 2, w2);
 202         /* LINTED E_BAD_PTR_CAST_ALIGN */
 203         w3 =  LOAD_BIG_32(blk + 4 * 3);
 204         SHA256ROUND(f, g, h, a, b, c, d, e, 3, w3);
 205         /* LINTED E_BAD_PTR_CAST_ALIGN */
 206         w4 =  LOAD_BIG_32(blk + 4 * 4);


 381                 SHA512_CONST_24, SHA512_CONST_25, SHA512_CONST_26,
 382                 SHA512_CONST_27, SHA512_CONST_28, SHA512_CONST_29,
 383                 SHA512_CONST_30, SHA512_CONST_31, SHA512_CONST_32,
 384                 SHA512_CONST_33, SHA512_CONST_34, SHA512_CONST_35,
 385                 SHA512_CONST_36, SHA512_CONST_37, SHA512_CONST_38,
 386                 SHA512_CONST_39, SHA512_CONST_40, SHA512_CONST_41,
 387                 SHA512_CONST_42, SHA512_CONST_43, SHA512_CONST_44,
 388                 SHA512_CONST_45, SHA512_CONST_46, SHA512_CONST_47,
 389                 SHA512_CONST_48, SHA512_CONST_49, SHA512_CONST_50,
 390                 SHA512_CONST_51, SHA512_CONST_52, SHA512_CONST_53,
 391                 SHA512_CONST_54, SHA512_CONST_55, SHA512_CONST_56,
 392                 SHA512_CONST_57, SHA512_CONST_58, SHA512_CONST_59,
 393                 SHA512_CONST_60, SHA512_CONST_61, SHA512_CONST_62,
 394                 SHA512_CONST_63, SHA512_CONST_64, SHA512_CONST_65,
 395                 SHA512_CONST_66, SHA512_CONST_67, SHA512_CONST_68,
 396                 SHA512_CONST_69, SHA512_CONST_70, SHA512_CONST_71,
 397                 SHA512_CONST_72, SHA512_CONST_73, SHA512_CONST_74,
 398                 SHA512_CONST_75, SHA512_CONST_76, SHA512_CONST_77,
 399                 SHA512_CONST_78, SHA512_CONST_79
 400         };
 401 #endif
 402 
 403 
 404         if ((uintptr_t)blk & 0x7) {         /* not 8-byte aligned? */
 405                 bcopy(blk, ctx->buf_un.buf64,  sizeof (ctx->buf_un.buf64));
 406                 blk = (uint8_t *)ctx->buf_un.buf64;
 407         }
 408 
 409         /* LINTED E_BAD_PTR_CAST_ALIGN */
 410         w0 =  LOAD_BIG_64(blk + 8 * 0);
 411         SHA512ROUND(a, b, c, d, e, f, g, h, 0, w0);
 412         /* LINTED E_BAD_PTR_CAST_ALIGN */
 413         w1 =  LOAD_BIG_64(blk + 8 * 1);
 414         SHA512ROUND(h, a, b, c, d, e, f, g, 1, w1);
 415         /* LINTED E_BAD_PTR_CAST_ALIGN */
 416         w2 =  LOAD_BIG_64(blk + 8 * 2);
 417         SHA512ROUND(g, h, a, b, c, d, e, f, 2, w2);
 418         /* LINTED E_BAD_PTR_CAST_ALIGN */
 419         w3 =  LOAD_BIG_64(blk + 8 * 3);
 420         SHA512ROUND(f, g, h, a, b, c, d, e, 3, w3);
 421         /* LINTED E_BAD_PTR_CAST_ALIGN */


 580         SHA512ROUND(f, g, h, a, b, c, d, e, 75, w11);
 581         w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
 582         SHA512ROUND(e, f, g, h, a, b, c, d, 76, w12);
 583         w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
 584         SHA512ROUND(d, e, f, g, h, a, b, c, 77, w13);
 585         w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
 586         SHA512ROUND(c, d, e, f, g, h, a, b, 78, w14);
 587         w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
 588         SHA512ROUND(b, c, d, e, f, g, h, a, 79, w15);
 589 
 590         ctx->state.s64[0] += a;
 591         ctx->state.s64[1] += b;
 592         ctx->state.s64[2] += c;
 593         ctx->state.s64[3] += d;
 594         ctx->state.s64[4] += e;
 595         ctx->state.s64[5] += f;
 596         ctx->state.s64[6] += g;
 597         ctx->state.s64[7] += h;
 598 
 599 }

 600 
 601 
 602 /*
 603  * Encode()
 604  *
 605  * purpose: to convert a list of numbers from little endian to big endian
 606  *   input: uint8_t *   : place to store the converted big endian numbers
 607  *          uint32_t *  : place to get numbers to convert from
 608  *          size_t      : the length of the input in bytes
 609  *  output: void
 610  */
 611 
 612 static void
 613 Encode(uint8_t *_RESTRICT_KYWD output, uint32_t *_RESTRICT_KYWD input,
 614     size_t len)
 615 {
 616         size_t          i, j;
 617 
 618 #if     defined(__sparc)
 619         if (IS_P2ALIGNED(output, sizeof (uint32_t))) {


 732 SHA384Init(SHA384_CTX *ctx)
 733 {
 734         SHA2Init(SHA384, ctx);
 735 }
 736 
 737 void
 738 SHA512Init(SHA512_CTX *ctx)
 739 {
 740         SHA2Init(SHA512, ctx);
 741 }
 742 
 743 #endif /* _KERNEL */
 744 
 745 /*
 746  * SHA2Update()
 747  *
 748  * purpose: continues an sha2 digest operation, using the message block
 749  *          to update the context.
 750  *   input: SHA2_CTX *  : the context to update
 751  *          void *      : the message block
 752  *          size_t    : the length of the message block in bytes
 753  *  output: void
 754  */
 755 
 756 void
 757 SHA2Update(SHA2_CTX *ctx, const void *inptr, size_t input_len)
 758 {
 759         uint32_t i, buf_index, buf_len, buf_limit;
 760         const uint8_t *input = inptr;




 761 

 762         /* check for noop */
 763         if (input_len == 0)
 764                 return;
 765 
 766         if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
 767                 buf_limit = 64;
 768 
 769                 /* compute number of bytes mod 64 */
 770                 buf_index = (ctx->count.c32[1] >> 3) & 0x3F;
 771 
 772                 /* update number of bits */
 773                 if ((ctx->count.c32[1] += (input_len << 3)) < (input_len << 3))
 774                         ctx->count.c32[0]++;
 775 
 776                 ctx->count.c32[0] += (input_len >> 29);
 777 
 778         } else {
 779                 buf_limit = 128;
 780 
 781                 /* compute number of bytes mod 128 */
 782                 buf_index = (ctx->count.c64[1] >> 3) & 0x7F;
 783 
 784                 /* update number of bits */
 785                 if ((ctx->count.c64[1] += (input_len << 3)) < (input_len << 3))
 786                         ctx->count.c64[0]++;


 788                 ctx->count.c64[0] += (input_len >> 29);
 789         }
 790 
 791         buf_len = buf_limit - buf_index;
 792 
 793         /* transform as many times as possible */
 794         i = 0;
 795         if (input_len >= buf_len) {
 796 
 797                 /*
 798                  * general optimization:
 799                  *
 800                  * only do initial bcopy() and SHA2Transform() if
 801                  * buf_index != 0.  if buf_index == 0, we're just
 802                  * wasting our time doing the bcopy() since there
 803                  * wasn't any data left over from a previous call to
 804                  * SHA2Update().
 805                  */
 806                 if (buf_index) {
 807                         bcopy(input, &ctx->buf_un.buf8[buf_index], buf_len);
 808                         if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
 809                                 SHA256Transform(ctx, ctx->buf_un.buf8);
 810                         else
 811                                 SHA512Transform(ctx, ctx->buf_un.buf8);
 812 
 813                         i = buf_len;
 814                 }
 815 
 816 

 817                 for (; i + buf_limit - 1 < input_len; i += buf_limit) {
 818                         if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
 819                                 SHA256Transform(ctx, &input[i]);
 820                         else


 821                                 SHA512Transform(ctx, &input[i]);
 822                 }

 823 


















 824                 /*
 825                  * general optimization:
 826                  *
 827                  * if i and input_len are the same, return now instead
 828                  * of calling bcopy(), since the bcopy() in this case
 829                  * will be an expensive nop.
 830                  */
 831 
 832                 if (input_len == i)
 833                         return;
 834 
 835                 buf_index = 0;
 836         }
 837 
 838         /* buffer remaining input */
 839         bcopy(&input[i], &ctx->buf_un.buf8[buf_index], input_len - i);
 840 }
 841 
 842 
 843 /*
 844  * SHA2Final()
 845  *
 846  * purpose: ends an sha2 digest operation, finalizing the message digest and
 847  *          zeroing the context.
 848  *   input: uchar_t *   : a buffer to store the digest in
 849  *                      : The function actually uses void* because many
 850  *                      : callers pass things other than uchar_t here.
 851  *          SHA2_CTX *  : the context to finalize, save, and zero
 852  *  output: void
 853  */
 854 
 855 void
 856 SHA2Final(void *digest, SHA2_CTX *ctx)
 857 {
 858         uint8_t         bitcount_be[sizeof (ctx->count.c32)];
 859         uint8_t         bitcount_be64[sizeof (ctx->count.c64)];
 860         uint32_t        index;

 861 
 862 
 863         if (ctx->algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
 864                 index  = (ctx->count.c32[1] >> 3) & 0x3f;
 865                 Encode(bitcount_be, ctx->count.c32, sizeof (bitcount_be));
 866                 SHA2Update(ctx, PADDING, ((index < 56) ? 56 : 120) - index);
 867                 SHA2Update(ctx, bitcount_be, sizeof (bitcount_be));
 868                 Encode(digest, ctx->state.s32, sizeof (ctx->state.s32));
 869 
 870         } else {
 871                 index  = (ctx->count.c64[1] >> 3) & 0x7f;
 872                 Encode64(bitcount_be64, ctx->count.c64,
 873                     sizeof (bitcount_be64));
 874                 SHA2Update(ctx, PADDING, ((index < 112) ? 112 : 240) - index);
 875                 SHA2Update(ctx, bitcount_be64, sizeof (bitcount_be64));
 876                 if (ctx->algotype <= SHA384_HMAC_GEN_MECH_INFO_TYPE) {
 877                         ctx->state.s64[6] = ctx->state.s64[7] = 0;
 878                         Encode64(digest, ctx->state.s64,
 879                             sizeof (uint64_t) * 6);
 880                 } else
 881                         Encode64(digest, ctx->state.s64,
 882                             sizeof (ctx->state.s64));
 883         }
 884 
 885         /* zeroize sensitive information */
 886         bzero(ctx, sizeof (*ctx));
 887 }
   1 /*
   2  * Copyright 2008 Sun Microsystems, Inc.  All rights reserved.
   3  * Use is subject to license terms.
   4  */
   5 
   6 #pragma ident   "@(#)sha2.c     1.8     08/03/05 SMI"
   7 

   8 /*
   9  * The basic framework for this code came from the reference
  10  * implementation for MD5.  That implementation is Copyright (C)
  11  * 1991-2, RSA Data Security, Inc. Created 1991. All rights reserved.
  12  *
  13  * License to copy and use this software is granted provided that it
  14  * is identified as the "RSA Data Security, Inc. MD5 Message-Digest
  15  * Algorithm" in all material mentioning or referencing this software
  16  * or this function.
  17  *
  18  * License is also granted to make and use derivative works provided
  19  * that such works are identified as "derived from the RSA Data
  20  * Security, Inc. MD5 Message-Digest Algorithm" in all material
  21  * mentioning or referencing the derived work.
  22  *
  23  * RSA Data Security, Inc. makes no representations concerning either
  24  * the merchantability of this software or the suitability of this
  25  * software for any particular purpose. It is provided "as is"
  26  * without express or implied warranty of any kind.
  27  *
  28  * These notices must be retained in any copies of any part of this
  29  * documentation and/or software.
  30  *
  31  * NOTE: Cleaned-up and optimized, version of SHA2, based on the FIPS 180-2
  32  * standard, available at http://www.itl.nist.gov/div897/pubs/fip180-2.htm
  33  * Not as fast as one would like -- further optimizations are encouraged
  34  * and appreciated.
  35  */
  36 
  37 #include <sys/types.h>
  38 #include <sys/param.h>
  39 #include <sys/systm.h>
  40 #include <sys/sysmacros.h>
  41 #define _SHA2_IMPL
  42 #include <sys/sha2.h>
  43 #include <sys/sha2_consts.h>
  44 
  45 #ifdef _KERNEL
  46 #include <sys/cmn_err.h>
  47 
  48 #else
  49 #include <strings.h>
  50 #include <stdlib.h>
  51 #include <errno.h>
  52 
  53 #pragma weak SHA256Update = SHA2Update
  54 #pragma weak SHA384Update = SHA2Update
  55 #pragma weak SHA512Update = SHA2Update
  56 
  57 #pragma weak SHA256Final = SHA2Final
  58 #pragma weak SHA384Final = SHA2Final
  59 #pragma weak SHA512Final = SHA2Final
  60 




  61 #endif  /* _KERNEL */
  62 
  63 static void Encode(uint8_t *, uint32_t *, size_t);
  64 static void Encode64(uint8_t *, uint64_t *, size_t);
  65 
  66 #if     defined(__amd64)
  67 #define SHA512Transform(ctx, in) SHA512TransformBlocks((ctx), (in), 1)
  68 #define SHA256Transform(ctx, in) SHA256TransformBlocks((ctx), (in), 1)
  69 
  70 void SHA512TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
  71 void SHA256TransformBlocks(SHA2_CTX *ctx, const void *in, size_t num);
  72 
  73 #else
  74 static void SHA256Transform(SHA2_CTX *, const uint8_t *);
  75 static void SHA512Transform(SHA2_CTX *, const uint8_t *);
  76 #endif  /* __amd64 */
  77 
  78 static uint8_t PADDING[128] = { 0x80, /* all zeros */ };
  79 
  80 /* Ch and Maj are the basic SHA2 functions. */
  81 #define Ch(b, c, d)     (((b) & (c)) ^ ((~b) & (d)))
  82 #define Maj(b, c, d)    (((b) & (c)) ^ ((b) & (d)) ^ ((c) & (d)))
  83 
  84 /* Rotates x right n bits. */
  85 #define ROTR(x, n)      \
  86         (((x) >> (n)) | ((x) << ((sizeof (x) * NBBY)-(n))))
  87 
  88 /* Shift x right n bits */
  89 #define SHR(x, n)       ((x) >> (n))
  90 
  91 /* SHA256 Functions */
  92 #define BIGSIGMA0_256(x)        (ROTR((x), 2) ^ ROTR((x), 13) ^ ROTR((x), 22))
  93 #define BIGSIGMA1_256(x)        (ROTR((x), 6) ^ ROTR((x), 11) ^ ROTR((x), 25))
  94 #define SIGMA0_256(x)           (ROTR((x), 7) ^ ROTR((x), 18) ^ SHR((x), 3))
  95 #define SIGMA1_256(x)           (ROTR((x), 17) ^ ROTR((x), 19) ^ SHR((x), 10))
  96 


 125 #define LOAD_BIG_32(addr)       (*(uint32_t *)(addr))
 126 
 127 #else   /* little endian -- will work on big endian, but slowly */
 128 
 129 #define LOAD_BIG_32(addr)       \
 130         (((addr)[0] << 24) | ((addr)[1] << 16) | ((addr)[2] << 8) | (addr)[3])
 131 #endif
 132 
 133 
 134 #if     defined(_BIG_ENDIAN)
 135 
 136 #define LOAD_BIG_64(addr)       (*(uint64_t *)(addr))
 137 
 138 #else   /* little endian -- will work on big endian, but slowly */
 139 
 140 #define LOAD_BIG_64(addr)       \
 141         (((uint64_t)(addr)[0] << 56) | ((uint64_t)(addr)[1] << 48) |        \
 142             ((uint64_t)(addr)[2] << 40) | ((uint64_t)(addr)[3] << 32) |     \
 143             ((uint64_t)(addr)[4] << 24) | ((uint64_t)(addr)[5] << 16) |     \
 144             ((uint64_t)(addr)[6] << 8) | (uint64_t)(addr)[7])

 145 #endif
 146 
 147 
 148 #if     !defined(__amd64)
 149 /* SHA256 Transform */
 150 
 151 static void
 152 SHA256Transform(SHA2_CTX *ctx, const uint8_t *blk)
 153 {

 154         uint32_t a = ctx->state.s32[0];
 155         uint32_t b = ctx->state.s32[1];
 156         uint32_t c = ctx->state.s32[2];
 157         uint32_t d = ctx->state.s32[3];
 158         uint32_t e = ctx->state.s32[4];
 159         uint32_t f = ctx->state.s32[5];
 160         uint32_t g = ctx->state.s32[6];
 161         uint32_t h = ctx->state.s32[7];
 162 
 163         uint32_t w0, w1, w2, w3, w4, w5, w6, w7;
 164         uint32_t w8, w9, w10, w11, w12, w13, w14, w15;
 165         uint32_t T1, T2;
 166 
 167 #if     defined(__sparc)
 168         static const uint32_t sha256_consts[] = {
 169                 SHA256_CONST_0, SHA256_CONST_1, SHA256_CONST_2,
 170                 SHA256_CONST_3, SHA256_CONST_4, SHA256_CONST_5,
 171                 SHA256_CONST_6, SHA256_CONST_7, SHA256_CONST_8,
 172                 SHA256_CONST_9, SHA256_CONST_10, SHA256_CONST_11,
 173                 SHA256_CONST_12, SHA256_CONST_13, SHA256_CONST_14,
 174                 SHA256_CONST_15, SHA256_CONST_16, SHA256_CONST_17,
 175                 SHA256_CONST_18, SHA256_CONST_19, SHA256_CONST_20,
 176                 SHA256_CONST_21, SHA256_CONST_22, SHA256_CONST_23,
 177                 SHA256_CONST_24, SHA256_CONST_25, SHA256_CONST_26,
 178                 SHA256_CONST_27, SHA256_CONST_28, SHA256_CONST_29,
 179                 SHA256_CONST_30, SHA256_CONST_31, SHA256_CONST_32,
 180                 SHA256_CONST_33, SHA256_CONST_34, SHA256_CONST_35,
 181                 SHA256_CONST_36, SHA256_CONST_37, SHA256_CONST_38,
 182                 SHA256_CONST_39, SHA256_CONST_40, SHA256_CONST_41,
 183                 SHA256_CONST_42, SHA256_CONST_43, SHA256_CONST_44,
 184                 SHA256_CONST_45, SHA256_CONST_46, SHA256_CONST_47,
 185                 SHA256_CONST_48, SHA256_CONST_49, SHA256_CONST_50,
 186                 SHA256_CONST_51, SHA256_CONST_52, SHA256_CONST_53,
 187                 SHA256_CONST_54, SHA256_CONST_55, SHA256_CONST_56,
 188                 SHA256_CONST_57, SHA256_CONST_58, SHA256_CONST_59,
 189                 SHA256_CONST_60, SHA256_CONST_61, SHA256_CONST_62,
 190                 SHA256_CONST_63
 191         };
 192 #endif  /* __sparc */
 193 
 194         if ((uintptr_t)blk & 0x3) {         /* not 4-byte aligned? */
 195                 bcopy(blk, ctx->buf_un.buf32,  sizeof (ctx->buf_un.buf32));
 196                 blk = (uint8_t *)ctx->buf_un.buf32;
 197         }
 198 
 199         /* LINTED E_BAD_PTR_CAST_ALIGN */
 200         w0 =  LOAD_BIG_32(blk + 4 * 0);
 201         SHA256ROUND(a, b, c, d, e, f, g, h, 0, w0);
 202         /* LINTED E_BAD_PTR_CAST_ALIGN */
 203         w1 =  LOAD_BIG_32(blk + 4 * 1);
 204         SHA256ROUND(h, a, b, c, d, e, f, g, 1, w1);
 205         /* LINTED E_BAD_PTR_CAST_ALIGN */
 206         w2 =  LOAD_BIG_32(blk + 4 * 2);
 207         SHA256ROUND(g, h, a, b, c, d, e, f, 2, w2);
 208         /* LINTED E_BAD_PTR_CAST_ALIGN */
 209         w3 =  LOAD_BIG_32(blk + 4 * 3);
 210         SHA256ROUND(f, g, h, a, b, c, d, e, 3, w3);
 211         /* LINTED E_BAD_PTR_CAST_ALIGN */
 212         w4 =  LOAD_BIG_32(blk + 4 * 4);


 387                 SHA512_CONST_24, SHA512_CONST_25, SHA512_CONST_26,
 388                 SHA512_CONST_27, SHA512_CONST_28, SHA512_CONST_29,
 389                 SHA512_CONST_30, SHA512_CONST_31, SHA512_CONST_32,
 390                 SHA512_CONST_33, SHA512_CONST_34, SHA512_CONST_35,
 391                 SHA512_CONST_36, SHA512_CONST_37, SHA512_CONST_38,
 392                 SHA512_CONST_39, SHA512_CONST_40, SHA512_CONST_41,
 393                 SHA512_CONST_42, SHA512_CONST_43, SHA512_CONST_44,
 394                 SHA512_CONST_45, SHA512_CONST_46, SHA512_CONST_47,
 395                 SHA512_CONST_48, SHA512_CONST_49, SHA512_CONST_50,
 396                 SHA512_CONST_51, SHA512_CONST_52, SHA512_CONST_53,
 397                 SHA512_CONST_54, SHA512_CONST_55, SHA512_CONST_56,
 398                 SHA512_CONST_57, SHA512_CONST_58, SHA512_CONST_59,
 399                 SHA512_CONST_60, SHA512_CONST_61, SHA512_CONST_62,
 400                 SHA512_CONST_63, SHA512_CONST_64, SHA512_CONST_65,
 401                 SHA512_CONST_66, SHA512_CONST_67, SHA512_CONST_68,
 402                 SHA512_CONST_69, SHA512_CONST_70, SHA512_CONST_71,
 403                 SHA512_CONST_72, SHA512_CONST_73, SHA512_CONST_74,
 404                 SHA512_CONST_75, SHA512_CONST_76, SHA512_CONST_77,
 405                 SHA512_CONST_78, SHA512_CONST_79
 406         };
 407 #endif  /* __sparc */
 408 
 409 
 410         if ((uintptr_t)blk & 0x7) {         /* not 8-byte aligned? */
 411                 bcopy(blk, ctx->buf_un.buf64,  sizeof (ctx->buf_un.buf64));
 412                 blk = (uint8_t *)ctx->buf_un.buf64;
 413         }
 414 
 415         /* LINTED E_BAD_PTR_CAST_ALIGN */
 416         w0 =  LOAD_BIG_64(blk + 8 * 0);
 417         SHA512ROUND(a, b, c, d, e, f, g, h, 0, w0);
 418         /* LINTED E_BAD_PTR_CAST_ALIGN */
 419         w1 =  LOAD_BIG_64(blk + 8 * 1);
 420         SHA512ROUND(h, a, b, c, d, e, f, g, 1, w1);
 421         /* LINTED E_BAD_PTR_CAST_ALIGN */
 422         w2 =  LOAD_BIG_64(blk + 8 * 2);
 423         SHA512ROUND(g, h, a, b, c, d, e, f, 2, w2);
 424         /* LINTED E_BAD_PTR_CAST_ALIGN */
 425         w3 =  LOAD_BIG_64(blk + 8 * 3);
 426         SHA512ROUND(f, g, h, a, b, c, d, e, 3, w3);
 427         /* LINTED E_BAD_PTR_CAST_ALIGN */


 586         SHA512ROUND(f, g, h, a, b, c, d, e, 75, w11);
 587         w12 = SIGMA1(w10) + w5 + SIGMA0(w13) + w12;
 588         SHA512ROUND(e, f, g, h, a, b, c, d, 76, w12);
 589         w13 = SIGMA1(w11) + w6 + SIGMA0(w14) + w13;
 590         SHA512ROUND(d, e, f, g, h, a, b, c, 77, w13);
 591         w14 = SIGMA1(w12) + w7 + SIGMA0(w15) + w14;
 592         SHA512ROUND(c, d, e, f, g, h, a, b, 78, w14);
 593         w15 = SIGMA1(w13) + w8 + SIGMA0(w0) + w15;
 594         SHA512ROUND(b, c, d, e, f, g, h, a, 79, w15);
 595 
 596         ctx->state.s64[0] += a;
 597         ctx->state.s64[1] += b;
 598         ctx->state.s64[2] += c;
 599         ctx->state.s64[3] += d;
 600         ctx->state.s64[4] += e;
 601         ctx->state.s64[5] += f;
 602         ctx->state.s64[6] += g;
 603         ctx->state.s64[7] += h;
 604 
 605 }
 606 #endif  /* !__amd64 */
 607 
 608 
 609 /*
 610  * Encode()
 611  *
 612  * purpose: to convert a list of numbers from little endian to big endian
 613  *   input: uint8_t *   : place to store the converted big endian numbers
 614  *          uint32_t *  : place to get numbers to convert from
 615  *          size_t      : the length of the input in bytes
 616  *  output: void
 617  */
 618 
 619 static void
 620 Encode(uint8_t *_RESTRICT_KYWD output, uint32_t *_RESTRICT_KYWD input,
 621     size_t len)
 622 {
 623         size_t          i, j;
 624 
 625 #if     defined(__sparc)
 626         if (IS_P2ALIGNED(output, sizeof (uint32_t))) {


 739 SHA384Init(SHA384_CTX *ctx)
 740 {
 741         SHA2Init(SHA384, ctx);
 742 }
 743 
 744 void
 745 SHA512Init(SHA512_CTX *ctx)
 746 {
 747         SHA2Init(SHA512, ctx);
 748 }
 749 
 750 #endif /* _KERNEL */
 751 
 752 /*
 753  * SHA2Update()
 754  *
 755  * purpose: continues an sha2 digest operation, using the message block
 756  *          to update the context.
 757  *   input: SHA2_CTX *  : the context to update
 758  *          void *      : the message block
 759  *          size_t      : the length of the message block, in bytes
 760  *  output: void
 761  */
 762 
 763 void
 764 SHA2Update(SHA2_CTX *ctx, const void *inptr, size_t input_len)
 765 {
 766         uint32_t        i, buf_index, buf_len, buf_limit;
 767         const uint8_t   *input = inptr;
 768         uint32_t        algotype = ctx->algotype;
 769 #if defined(__amd64)
 770         uint32_t        block_count;
 771 #endif  /* !__amd64 */
 772 
 773 
 774         /* check for noop */
 775         if (input_len == 0)
 776                 return;
 777 
 778         if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
 779                 buf_limit = 64;
 780 
 781                 /* compute number of bytes mod 64 */
 782                 buf_index = (ctx->count.c32[1] >> 3) & 0x3F;
 783 
 784                 /* update number of bits */
 785                 if ((ctx->count.c32[1] += (input_len << 3)) < (input_len << 3))
 786                         ctx->count.c32[0]++;
 787 
 788                 ctx->count.c32[0] += (input_len >> 29);
 789 
 790         } else {
 791                 buf_limit = 128;
 792 
 793                 /* compute number of bytes mod 128 */
 794                 buf_index = (ctx->count.c64[1] >> 3) & 0x7F;
 795 
 796                 /* update number of bits */
 797                 if ((ctx->count.c64[1] += (input_len << 3)) < (input_len << 3))
 798                         ctx->count.c64[0]++;


 800                 ctx->count.c64[0] += (input_len >> 29);
 801         }
 802 
 803         buf_len = buf_limit - buf_index;
 804 
 805         /* transform as many times as possible */
 806         i = 0;
 807         if (input_len >= buf_len) {
 808 
 809                 /*
 810                  * general optimization:
 811                  *
 812                  * only do initial bcopy() and SHA2Transform() if
 813                  * buf_index != 0.  if buf_index == 0, we're just
 814                  * wasting our time doing the bcopy() since there
 815                  * wasn't any data left over from a previous call to
 816                  * SHA2Update().
 817                  */
 818                 if (buf_index) {
 819                         bcopy(input, &ctx->buf_un.buf8[buf_index], buf_len);
 820                         if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE)
 821                                 SHA256Transform(ctx, ctx->buf_un.buf8);
 822                         else
 823                                 SHA512Transform(ctx, ctx->buf_un.buf8);
 824 
 825                         i = buf_len;
 826                 }
 827 
 828 #if !defined(__amd64)
 829                 if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
 830                         for (; i + buf_limit - 1 < input_len; i += buf_limit) {

 831                                 SHA256Transform(ctx, &input[i]);
 832                         }
 833                 } else {
 834                         for (; i + buf_limit - 1 < input_len; i += buf_limit) {
 835                                 SHA512Transform(ctx, &input[i]);
 836                         }
 837                 }
 838 
 839 #else
 840                 if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {
 841                         block_count = (input_len - i) >> 6;
 842                         if (block_count > 0) {
 843                                 SHA256TransformBlocks(ctx, &input[i],
 844                                     block_count);
 845                                 i += block_count << 6;
 846                         }
 847                 } else {
 848                         block_count = (input_len - i) >> 7;
 849                         if (block_count > 0) {
 850                                 SHA512TransformBlocks(ctx, &input[i],
 851                                     block_count);
 852                                 i += block_count << 7;
 853                         }
 854                 }
 855 #endif  /* !__amd64 */
 856 
 857                 /*
 858                  * general optimization:
 859                  *
 860                  * if i and input_len are the same, return now instead
 861                  * of calling bcopy(), since the bcopy() in this case
 862                  * will be an expensive noop.
 863                  */
 864 
 865                 if (input_len == i)
 866                         return;
 867 
 868                 buf_index = 0;
 869         }
 870 
 871         /* buffer remaining input */
 872         bcopy(&input[i], &ctx->buf_un.buf8[buf_index], input_len - i);
 873 }
 874 
 875 
 876 /*
 877  * SHA2Final()
 878  *
 879  * purpose: ends an sha2 digest operation, finalizing the message digest and
 880  *          zeroing the context.
 881  *   input: uchar_t *   : a buffer to store the digest
 882  *                      : The function actually uses void* because many
 883  *                      : callers pass things other than uchar_t here.
 884  *          SHA2_CTX *  : the context to finalize, save, and zero
 885  *  output: void
 886  */
 887 
 888 void
 889 SHA2Final(void *digest, SHA2_CTX *ctx)
 890 {
 891         uint8_t         bitcount_be[sizeof (ctx->count.c32)];
 892         uint8_t         bitcount_be64[sizeof (ctx->count.c64)];
 893         uint32_t        index;
 894         uint32_t        algotype = ctx->algotype;
 895 
 896         if (algotype <= SHA256_HMAC_GEN_MECH_INFO_TYPE) {

 897                 index  = (ctx->count.c32[1] >> 3) & 0x3f;
 898                 Encode(bitcount_be, ctx->count.c32, sizeof (bitcount_be));
 899                 SHA2Update(ctx, PADDING, ((index < 56) ? 56 : 120) - index);
 900                 SHA2Update(ctx, bitcount_be, sizeof (bitcount_be));
 901                 Encode(digest, ctx->state.s32, sizeof (ctx->state.s32));
 902 
 903         } else {
 904                 index  = (ctx->count.c64[1] >> 3) & 0x7f;
 905                 Encode64(bitcount_be64, ctx->count.c64,
 906                     sizeof (bitcount_be64));
 907                 SHA2Update(ctx, PADDING, ((index < 112) ? 112 : 240) - index);
 908                 SHA2Update(ctx, bitcount_be64, sizeof (bitcount_be64));
 909                 if (algotype <= SHA384_HMAC_GEN_MECH_INFO_TYPE) {
 910                         ctx->state.s64[6] = ctx->state.s64[7] = 0;
 911                         Encode64(digest, ctx->state.s64,
 912                             sizeof (uint64_t) * 6);
 913                 } else
 914                         Encode64(digest, ctx->state.s64,
 915                             sizeof (ctx->state.s64));
 916         }
 917 
 918         /* zeroize sensitive information */
 919         bzero(ctx, sizeof (*ctx));
 920 }