release/src-rt/bcmcrypto/md32_common.h

   1 /* crypto/md32_common.h */
   2 /* ====================================================================
   3  * Copyright (c) 1999-2002 The OpenSSL Project.  All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  *
   9  * 1. Redistributions of source code must retain the above copyright
  10  *    notice, this list of conditions and the following disclaimer.
  11  *
  12  * 2. Redistributions in binary form must reproduce the above copyright
  13  *    notice, this list of conditions and the following disclaimer in
  14  *    the documentation and/or other materials provided with the
  15  *    distribution.
  16  *
  17  * 3. All advertising materials mentioning features or use of this
  18  *    software must display the following acknowledgment:
  19  *    "This product includes software developed by the OpenSSL Project
  20  *    for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
  21  *
  22  * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
  23  *    endorse or promote products derived from this software without
  24  *    prior written permission. For written permission, please contact
  25  *    licensing@OpenSSL.org.
  26  *
  27  * 5. Products derived from this software may not be called "OpenSSL"
  28  *    nor may "OpenSSL" appear in their names without prior written
  29  *    permission of the OpenSSL Project.
  30  *
  31  * 6. Redistributions of any form whatsoever must retain the following
  32  *    acknowledgment:
  33  *    "This product includes software developed by the OpenSSL Project
  34  *    for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
  35  *
  36  * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
  37  * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  38  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
  39  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
  40  * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
  41  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
  42  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
  43  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  44  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
  45  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
  46  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
  47  * OF THE POSSIBILITY OF SUCH DAMAGE.
  48  * ====================================================================
  49  *
  50  * This product includes cryptographic software written by Eric Young
  51  * (eay@cryptsoft.com).  This product includes software written by Tim
  52  * Hudson (tjh@cryptsoft.com).
  53  *
  54  */
  55
  56 /*
  57  * This is a generic 32 bit "collector" for message digest algorithms.
  58  * Whenever needed it collects input character stream into chunks of
  59  * 32 bit values and invokes a block function that performs actual hash
  60  * calculations.
  61  *
  62  * Porting guide.
  63  *
  64  * Obligatory macros:
  65  *
  66  * DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
  67  *      this macro defines byte order of input stream.
  68  * HASH_CBLOCK
  69  *      size of a unit chunk HASH_BLOCK operates on.
  70  * HASH_LONG
  71  *      has to be at lest 32 bit wide, if it's wider, then
  72  *      HASH_LONG_LOG2 *has to* be defined along
  73  * HASH_CTX
  74  *      context structure that at least contains following
  75  *      members:
  76  *              typedef struct {
  77  *                      ...
  78  *                      HASH_LONG       Nl,Nh;
  79  *                      HASH_LONG       data[HASH_LBLOCK];
  80  *                      unsigned int    num;
  81  *                      ...
  82  *                      } HASH_CTX;
  83  * HASH_UPDATE
  84  *      name of "Update" function, implemented here.
  85  * HASH_TRANSFORM
  86  *      name of "Transform" function, implemented here.
  87  * HASH_FINAL
  88  *      name of "Final" function, implemented here.
  89  * HASH_BLOCK_HOST_ORDER
  90  *      name of "block" function treating *aligned* input message
  91  *      in host byte order, implemented externally.
  92  * HASH_BLOCK_DATA_ORDER
  93  *      name of "block" function treating *unaligned* input message
  94  *      in original (data) byte order, implemented externally (it
  95  *      actually is optional if data and host are of the same
  96  *      "endianess").
  97  * HASH_MAKE_STRING
  98  *      macro convering context variables to an ASCII hash string.
  99  *
 100  * Optional macros:
 101  *
 102  * B_ENDIAN or L_ENDIAN
 103  *      defines host byte-order.
 104  * HASH_LONG_LOG2
 105  *      defaults to 2 if not states otherwise.
 106  * HASH_LBLOCK
 107  *      assumed to be HASH_CBLOCK/4 if not stated otherwise.
 108  * HASH_BLOCK_DATA_ORDER_ALIGNED
 109  *      alternative "block" function capable of treating
 110  *      aligned input message in original (data) order,
 111  *      implemented externally.
 112  *
 113  * MD5 example:
 114  *
 115  *      #define DATA_ORDER_IS_LITTLE_ENDIAN
 116  *
 117  *      #define HASH_LONG               MD5_LONG
 118  *      #define HASH_LONG_LOG2          MD5_LONG_LOG2
 119  *      #define HASH_CTX                MD5_CTX
 120  *      #define HASH_CBLOCK             MD5_CBLOCK
 121  *      #define HASH_LBLOCK             MD5_LBLOCK
 122  *      #define HASH_UPDATE             MD5_Update
 123  *      #define HASH_TRANSFORM          MD5_Transform
 124  *      #define HASH_FINAL              MD5_Final
 125  *      #define HASH_BLOCK_HOST_ORDER   md5_block_host_order
 126  *      #define HASH_BLOCK_DATA_ORDER   md5_block_data_order
 127  *
 128  *                                      <appro@fy.chalmers.se>
 129  */
 130
 131 /* FILE-CSTYLED */
 132
 133 #if !defined(DATA_ORDER_IS_BIG_ENDIAN) && !defined(DATA_ORDER_IS_LITTLE_ENDIAN)
 134 #error "DATA_ORDER must be defined!"
 135 #endif
 136
 137 #ifndef HASH_CBLOCK
 138 #error "HASH_CBLOCK must be defined!"
 139 #endif
 140 #ifndef HASH_LONG
 141 #error "HASH_LONG must be defined!"
 142 #endif
 143 #ifndef HASH_CTX
 144 #error "HASH_CTX must be defined!"
 145 #endif
 146
 147 #ifndef HASH_UPDATE
 148 #error "HASH_UPDATE must be defined!"
 149 #endif
 150 #ifndef HASH_TRANSFORM
 151 #error "HASH_TRANSFORM must be defined!"
 152 #endif
 153 #ifndef HASH_FINAL
 154 #error "HASH_FINAL must be defined!"
 155 #endif
 156
 157 #ifndef HASH_BLOCK_HOST_ORDER
 158 #error "HASH_BLOCK_HOST_ORDER must be defined!"
 159 #endif
 160
 161
 162 #ifndef HASH_LBLOCK
 163 #define HASH_LBLOCK     (HASH_CBLOCK/4)
 164 #endif
 165
 166 #ifndef HASH_LONG_LOG2
 167 #define HASH_LONG_LOG2  2
 168 #endif
 169
 170 /*
 171  * Engage compiler specific rotate intrinsic function if available.
 172  */
 173 #undef ROTATE
 174 #ifndef PEDANTIC
 175 # if defined(_MSC_VER) || defined(__ICC)
 176 #  define ROTATE(a,n)   _lrotl(a,n)
 177 # elif defined(__MWERKS__)
 178 #  if defined(__POWERPC__)
 179 #   define ROTATE(a,n)  __rlwinm(a,n,0,31)
 180 #  elif defined(__MC68K__)
 181     /* Motorola specific tweak. <appro@fy.chalmers.se> */
 182 #   define ROTATE(a,n)  ( n<24 ? __rol(a,n) : __ror(a,32-n) )
 183 #  else
 184 #   define ROTATE(a,n)  __rol(a,n)
 185 #  endif
 186 # elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && \
 187         !defined(OPENSSL_NO_INLINE_ASM)
 188   /*
 189    * Some GNU C inline assembler templates. Note that these are
 190    * rotates by *constant* number of bits! But that's exactly
 191    * what we need here...
 192    *                                    <appro@fy.chalmers.se>
 193    */
 194 #  if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
 195 #   define ROTATE(a,n)  ({ register unsigned int ret;   \
 196                                 asm (                   \
 197                                 "roll %1,%0"            \
 198                                 : "=r"(ret)             \
 199                                 : "I"(n), "0"(a)        \
 200                                 : "cc");                \
 201                            ret;                         \
 202                         })
 203 #  elif defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
 204 #   define ROTATE(a,n)  ({ register unsigned int ret;   \
 205                                 asm (                   \
 206                                 "rlwinm %0,%1,%2,0,31"  \
 207                                 : "=r"(ret)             \
 208                                 : "r"(a), "I"(n));      \
 209                            ret;                         \
 210                         })
 211 #  endif
 212 # endif
 213 #endif /* PEDANTIC */
 214
 215 #if HASH_LONG_LOG2==2       /* Engage only if sizeof(HASH_LONG)== 4 */
 216 /* A nice byte order reversal from Wei Dai <weidai@eskimo.com> */
 217 #ifdef ROTATE
 218 /* 5 instructions with rotate instruction, else 9 */
 219 #define REVERSE_FETCH32(a,l)    (                                       \
 220                 l =* (const HASH_LONG *)(a),                            \
 221                 ((ROTATE(l,8)&0x00FF00FF)|(ROTATE((l&0x00FF00FF),24)))  \
 222                                 )
 223 #else
 224 /* 6 instructions with rotate instruction, else 8 */
 225 #define REVERSE_FETCH32(a,l)    (                               \
 226                 l =* (const HASH_LONG *)(a),                    \
 227                 l = (((l>>8)&0x00FF00FF)|((l&0x00FF00FF)<<8)),  \
 228                 ROTATE(l,16)                                    \
 229                                 )
 230 /*
 231  * Originally the middle line started with l=(((l&0xFF00FF00)>>8)|...
 232  * It's rewritten as above for two reasons:
 233  *      - RISCs aren't good at long constants and have to explicitely
 234  *        compose 'em with several (well, usually 2) instructions in a
 235  *        register before performing the actual operation and (as you
 236  *        already realized:-) having same constant should inspire the
 237  *        compiler to permanently allocate the only register for it;
 238  *      - most modern CPUs have two ALUs, but usually only one has
 239  *        circuitry for shifts:-( this minor tweak inspires compiler
 240  *        to schedule shift instructions in a better way...
 241  *
 242  *                              <appro@fy.chalmers.se>
 243  */
 244 #endif
 245 #endif
 246
 247 #ifndef ROTATE
 248 #define ROTATE(a,n)     (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
 249 #endif
 250
 251 /*
 252  * Make some obvious choices. E.g., HASH_BLOCK_DATA_ORDER_ALIGNED
 253  * and HASH_BLOCK_HOST_ORDER ought to be the same if input data
 254  * and host are of the same "endianess". It's possible to mask
 255  * this with blank #define HASH_BLOCK_DATA_ORDER though...
 256  *
 257  *                              <appro@fy.chalmers.se>
 258  */
 259 #if defined(B_ENDIAN)
 260 #  if defined(DATA_ORDER_IS_BIG_ENDIAN)
 261 #    if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED) && HASH_LONG_LOG2==2
 262 #      define HASH_BLOCK_DATA_ORDER_ALIGNED     HASH_BLOCK_HOST_ORDER
 263 #    endif
 264 #  endif
 265 #elif defined(L_ENDIAN)
 266 #  if defined(DATA_ORDER_IS_LITTLE_ENDIAN)
 267 #    if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED) && HASH_LONG_LOG2==2
 268 #      define HASH_BLOCK_DATA_ORDER_ALIGNED     HASH_BLOCK_HOST_ORDER
 269 #    endif
 270 #  endif
 271 #endif
 272
 273 #if !defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
 274 #ifndef HASH_BLOCK_DATA_ORDER
 275 #error "HASH_BLOCK_DATA_ORDER must be defined!"
 276 #endif
 277 #endif
 278
 279 #if defined(DATA_ORDER_IS_BIG_ENDIAN)
 280
 281 #ifndef PEDANTIC
 282 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && \
 283         !defined(OPENSSL_NO_INLINE_ASM)
 284 #  if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
 285         (defined(__x86_64) || defined(__x86_64__))
 286     /*
 287      * This gives ~30-40% performance improvement in SHA-256 compiled
 288      * with gcc [on P4]. Well, first macro to be frank. We can pull
 289      * this trick on x86* platforms only, because these CPUs can fetch
 290      * unaligned data without raising an exception.
 291      */
 292 #   define HOST_c2l(c,l)        ({ unsigned int r=*((const unsigned int *)(c)); \
 293                                    asm ("bswapl %0":"=r"(r):"0"(r));    \
 294                                    (c)+=4; (l)=r;                       })
 295 #   define HOST_l2c(l,c)        ({ unsigned int r=(l);                  \
 296                                    asm ("bswapl %0":"=r"(r):"0"(r));    \
 297                                    *((unsigned int *)(c))=r; (c)+=4; r; })
 298 #  endif
 299 # endif
 300 #endif
 301
 302 #ifndef HOST_c2l
 303 #define HOST_c2l(c,l)   (l =(((unsigned long)(*((c)++)))<<24),          \
 304                          l|=(((unsigned long)(*((c)++)))<<16),          \
 305                          l|=(((unsigned long)(*((c)++)))<< 8),          \
 306                          l|=(((unsigned long)(*((c)++)))    ),          \
 307                          l)
 308 #endif
 309 #define HOST_p_c2l(c,l,n)       {                                       \
 310                         switch (n) {                                    \
 311                         case 0: l =((unsigned long)(*((c)++)))<<24;     \
 312                         case 1: l|=((unsigned long)(*((c)++)))<<16;     \
 313                         case 2: l|=((unsigned long)(*((c)++)))<< 8;     \
 314                         case 3: l|=((unsigned long)(*((c)++)));         \
 315                                 } }
 316 #define HOST_p_c2l_p(c,l,sc,len) {                                      \
 317                         switch (sc) {                                   \
 318                         case 0: l =((unsigned long)(*((c)++)))<<24;     \
 319                                 if (--len == 0) break;                  \
 320                         case 1: l|=((unsigned long)(*((c)++)))<<16;     \
 321                                 if (--len == 0) break;                  \
 322                         case 2: l|=((unsigned long)(*((c)++)))<< 8;     \
 323                                 } }
 324 /* NOTE the pointer is not incremented at the end of this */
 325 #define HOST_c2l_p(c,l,n)       {                                       \
 326                         l=0; (c)+=n;                                    \
 327                         switch (n) {                                    \
 328                         case 3: l =((unsigned long)(*(--(c))))<< 8;     \
 329                         case 2: l|=((unsigned long)(*(--(c))))<<16;     \
 330                         case 1: l|=((unsigned long)(*(--(c))))<<24;     \
 331                                 } }
 332 #ifndef HOST_l2c
 333 #define HOST_l2c(l,c)   (*((c)++)=(unsigned char)(((l)>>24)&0xff),      \
 334                          *((c)++)=(unsigned char)(((l)>>16)&0xff),      \
 335                          *((c)++)=(unsigned char)(((l)>> 8)&0xff),      \
 336                          *((c)++)=(unsigned char)(((l)    )&0xff),      \
 337                          l)
 338 #endif
 339
 340 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
 341
 342 #if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
 343 # ifndef B_ENDIAN
 344    /* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
 345 #  define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, l)
 346 #  define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, l)
 347 # endif
 348 #endif
 349
 350 #ifndef HOST_c2l
 351 #define HOST_c2l(c,l)   (l =(((unsigned long)(*((c)++)))    ),          \
 352                          l|=(((unsigned long)(*((c)++)))<< 8),          \
 353                          l|=(((unsigned long)(*((c)++)))<<16),          \
 354                          l|=(((unsigned long)(*((c)++)))<<24),          \
 355                          l)
 356 #endif
 357 #define HOST_p_c2l(c,l,n)       {                                       \
 358                         switch (n) {                                    \
 359                         case 0: l =((unsigned long)(*((c)++)));         \
 360                         case 1: l|=((unsigned long)(*((c)++)))<< 8;     \
 361                         case 2: l|=((unsigned long)(*((c)++)))<<16;     \
 362                         case 3: l|=((unsigned long)(*((c)++)))<<24;     \
 363                                 } }
 364 #define HOST_p_c2l_p(c,l,sc,len) {                                      \
 365                         switch (sc) {                                   \
 366                         case 0: l =((unsigned long)(*((c)++)));         \
 367                                 if (--len == 0) break;                  \
 368                         case 1: l|=((unsigned long)(*((c)++)))<< 8;     \
 369                                 if (--len == 0) break;                  \
 370                         case 2: l|=((unsigned long)(*((c)++)))<<16;     \
 371                                 } }
 372 /* NOTE the pointer is not incremented at the end of this */
 373 #define HOST_c2l_p(c,l,n)       {                                       \
 374                         l=0; (c)+=n;                                    \
 375                         switch (n) {                                    \
 376                         case 3: l =((unsigned long)(*(--(c))))<<16;     \
 377                         case 2: l|=((unsigned long)(*(--(c))))<< 8;     \
 378                         case 1: l|=((unsigned long)(*(--(c))));         \
 379                                 } }
 380 #ifndef HOST_l2c
 381 #define HOST_l2c(l,c)   (*((c)++)=(unsigned char)(((l)    )&0xff),      \
 382                          *((c)++)=(unsigned char)(((l)>> 8)&0xff),      \
 383                          *((c)++)=(unsigned char)(((l)>>16)&0xff),      \
 384                          *((c)++)=(unsigned char)(((l)>>24)&0xff),      \
 385                          l)
 386 #endif
 387
 388 #endif
 389
 390 /*
 391  * Time for some action:-)
 392  */
 393
 394 int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
 395         {
 396         const unsigned char *data=data_;
 397         register HASH_LONG * p;
 398         register HASH_LONG l;
 399         size_t sw,sc,ew,ec;
 400
 401         if (len==0) return 1;
 402
 403         l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
 404         /* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
 405          * Wei Dai <weidai@eskimo.com> for pointing it out. */
 406         if (l < c->Nl) /* overflow */
 407                 c->Nh++;
 408         c->Nh+=(len>>29);       /* might cause compiler warning on 16-bit */
 409         c->Nl=l;
 410
 411         if (c->num != 0)
 412                 {
 413                 p=c->data;
 414                 sw=c->num>>2;
 415                 sc=c->num&0x03;
 416
 417                 if ((c->num+len) >= HASH_CBLOCK)
 418                         {
 419                         l=p[sw]; HOST_p_c2l(data,l,sc); p[sw++]=l;
 420                         for (; sw<HASH_LBLOCK; sw++)
 421                                 {
 422                                 HOST_c2l(data,l); p[sw]=l;
 423                                 }
 424                         HASH_BLOCK_HOST_ORDER (c,p,1);
 425                         len-=(HASH_CBLOCK-c->num);
 426                         c->num=0;
 427                         /* drop through and do the rest */
 428                         }
 429                 else
 430                         {
 431                         c->num+=(unsigned int)len;
 432                         if ((sc+len) < 4) /* ugly, add char's to a word */
 433                                 {
 434                                 l=p[sw]; HOST_p_c2l_p(data,l,sc,len); p[sw]=l;
 435                                 }
 436                         else
 437                                 {
 438                                 ew=(c->num>>2);
 439                                 ec=(c->num&0x03);
 440                                 if (sc)
 441                                         l=p[sw];
 442                                 HOST_p_c2l(data,l,sc);
 443                                 p[sw++]=l;
 444                                 for (; sw < ew; sw++)
 445                                         {
 446                                         HOST_c2l(data,l); p[sw]=l;
 447                                         }
 448                                 if (ec)
 449                                         {
 450                                         HOST_c2l_p(data,l,ec); p[sw]=l;
 451                                         }
 452                                 }
 453                         return 1;
 454                         }
 455                 }
 456
 457         sw=len/HASH_CBLOCK;
 458         if (sw > 0)
 459                 {
 460 #if defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
 461                 /*
 462                  * Note that HASH_BLOCK_DATA_ORDER_ALIGNED gets defined
 463                  * only if sizeof(HASH_LONG)==4.
 464                  */
 465                 if ((((size_t)data)%4) == 0)
 466                         {
 467                         /* data is properly aligned so that we can cast it: */
 468                         HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,sw);
 469                         sw*=HASH_CBLOCK;
 470                         data+=sw;
 471                         len-=sw;
 472                         }
 473                 else
 474 #if !defined(HASH_BLOCK_DATA_ORDER)
 475                         while (sw--)
 476                                 {
 477                                 memcpy (p=c->data,data,HASH_CBLOCK);
 478                                 HASH_BLOCK_DATA_ORDER_ALIGNED(c,p,1);
 479                                 data+=HASH_CBLOCK;
 480                                 len-=HASH_CBLOCK;
 481                                 }
 482 #endif
 483 #endif
 484 #if defined(HASH_BLOCK_DATA_ORDER)
 485                         {
 486                         HASH_BLOCK_DATA_ORDER(c,data,sw);
 487                         sw*=HASH_CBLOCK;
 488                         data+=sw;
 489                         len-=sw;
 490                         }
 491 #endif
 492                 }
 493
 494         if (len!=0)
 495                 {
 496                 p = c->data;
 497                 c->num = len;
 498                 ew=len>>2;      /* words to copy */
 499                 ec=len&0x03;
 500                 for (; ew; ew--,p++)
 501                         {
 502                         HOST_c2l(data,l); *p=l;
 503                         }
 504                 HOST_c2l_p(data,l,ec);
 505                 *p=l;
 506                 }
 507         return 1;
 508         }
 509
 510
 511 void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
 512         {
 513 #if defined(HASH_BLOCK_DATA_ORDER_ALIGNED)
 514         if ((((size_t)data)%4) == 0)
 515                 /* data is properly aligned so that we can cast it: */
 516                 HASH_BLOCK_DATA_ORDER_ALIGNED (c,(const HASH_LONG *)data,1);
 517         else
 518 #if !defined(HASH_BLOCK_DATA_ORDER)
 519                 {
 520                 memcpy (c->data,data,HASH_CBLOCK);
 521                 HASH_BLOCK_DATA_ORDER_ALIGNED (c,c->data,1);
 522                 }
 523 #endif
 524 #endif
 525 #if defined(HASH_BLOCK_DATA_ORDER)
 526         HASH_BLOCK_DATA_ORDER (c,data,1);
 527 #endif
 528         }
 529
 530
 531 int HASH_FINAL (unsigned char *md, HASH_CTX *c)
 532         {
 533         register HASH_LONG *p;
 534         register unsigned long l;
 535         register int i,j;
 536         static const unsigned char end[4]={0x80,0x00,0x00,0x00};
 537         const unsigned char *cp=end;
 538
 539         /* c->num should definitly have room for at least one more byte. */
 540         p=c->data;
 541         i=c->num>>2;
 542         j=c->num&0x03;
 543
 544         l = (j==0) ? 0 : p[i];
 545         HOST_p_c2l(cp,l,j); p[i++]=l; /* i is the next 'undefined word' */
 546
 547         if (i>(HASH_LBLOCK-2)) /* save room for Nl and Nh */
 548                 {
 549                 if (i<HASH_LBLOCK) p[i]=0;
 550                 HASH_BLOCK_HOST_ORDER (c,p,1);
 551                 i=0;
 552                 }
 553         for (; i<(HASH_LBLOCK-2); i++)
 554                 p[i]=0;
 555
 556 #if   defined(DATA_ORDER_IS_BIG_ENDIAN)
 557         p[HASH_LBLOCK-2]=c->Nh;
 558         p[HASH_LBLOCK-1]=c->Nl;
 559 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
 560         p[HASH_LBLOCK-2]=c->Nl;
 561         p[HASH_LBLOCK-1]=c->Nh;
 562 #endif
 563         HASH_BLOCK_HOST_ORDER (c,p,1);
 564
 565 #ifndef HASH_MAKE_STRING
 566 #error "HASH_MAKE_STRING must be defined!"
 567 #else
 568         HASH_MAKE_STRING(c,md);
 569 #endif
 570
 571         c->num=0;
 572         /* clear stuff, HASH_BLOCK may be leaving some stuff on the stack
 573          * but I'm not worried :-)
 574         OPENSSL_cleanse((void *)c,sizeof(HASH_CTX));
 575          */
 576         return 1;
 577         }
 578
 579 #ifndef MD32_REG_T
 580 #define MD32_REG_T long
 581 /*
 582  * This comment was originaly written for MD5, which is why it
 583  * discusses A-D. But it basically applies to all 32-bit digests,
 584  * which is why it was moved to common header file.
 585  *
 586  * In case you wonder why A-D are declared as long and not
 587  * as MD5_LONG. Doing so results in slight performance
 588  * boost on LP64 architectures. The catch is we don't
 589  * really care if 32 MSBs of a 64-bit register get polluted
 590  * with eventual overflows as we *save* only 32 LSBs in
 591  * *either* case. Now declaring 'em long excuses the compiler
 592  * from keeping 32 MSBs zeroed resulting in 13% performance
 593  * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
 594  * Well, to be honest it should say that this *prevents*
 595  * performance degradation.
 596  *                              <appro@fy.chalmers.se>
 597  * Apparently there're LP64 compilers that generate better
 598  * code if A-D are declared int. Most notably GCC-x86_64
 599  * generates better code.
 600  *                              <appro@fy.chalmers.se>
 601  */
 602 #endif