[thirdparty/openssl.git] / crypto / include / internal / md32_common.h

/* ====================================================================
 * Copyright (c) 1999-2007 The OpenSSL Project.  All rights reserved.
 *
 * Redistribution and use in source and binary forms, with or without
 * modification, are permitted provided that the following conditions
 * are met:
 *
 * 1. Redistributions of source code must retain the above copyright
 *    notice, this list of conditions and the following disclaimer.
 *
 * 2. Redistributions in binary form must reproduce the above copyright
 *    notice, this list of conditions and the following disclaimer in
 *    the documentation and/or other materials provided with the
 *    distribution.
 *
 * 3. All advertising materials mentioning features or use of this
 *    software must display the following acknowledgment:
 *    "This product includes software developed by the OpenSSL Project
 *    for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
 *
 * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
 *    endorse or promote products derived from this software without
 *    prior written permission. For written permission, please contact
 *    licensing@OpenSSL.org.
 *
 * 5. Products derived from this software may not be called "OpenSSL"
 *    nor may "OpenSSL" appear in their names without prior written
 *    permission of the OpenSSL Project.
 *
 * 6. Redistributions of any form whatsoever must retain the following
 *    acknowledgment:
 *    "This product includes software developed by the OpenSSL Project
 *    for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
 *
 * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
 * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
 * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
 * OF THE POSSIBILITY OF SUCH DAMAGE.
 * ====================================================================
 *
 */

/*-
 * This is a generic 32 bit "collector" for message digest algorithms.
 * Whenever needed it collects input character stream into chunks of
 * 32 bit values and invokes a block function that performs actual hash
 * calculations.
 *
 * Porting guide.
 *
 * Obligatory macros:
 *
 * DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
 *      this macro defines byte order of input stream.
 * HASH_CBLOCK
 *      size of a unit chunk HASH_BLOCK operates on.
 * HASH_LONG
 *      has to be at lest 32 bit wide.
 * HASH_CTX
 *      context structure that at least contains following
 *      members:
 *              typedef struct {
 *                      ...
 *                      HASH_LONG       Nl,Nh;
 *                      either {
 *                      HASH_LONG       data[HASH_LBLOCK];
 *                      unsigned char   data[HASH_CBLOCK];
 *                      };
 *                      unsigned int    num;
 *                      ...
 *                      } HASH_CTX;
 *      data[] vector is expected to be zeroed upon first call to
 *      HASH_UPDATE.
 * HASH_UPDATE
 *      name of "Update" function, implemented here.
 * HASH_TRANSFORM
 *      name of "Transform" function, implemented here.
 * HASH_FINAL
 *      name of "Final" function, implemented here.
 * HASH_BLOCK_DATA_ORDER
 *      name of "block" function capable of treating *unaligned* input
 *      message in original (data) byte order, implemented externally.
 * HASH_MAKE_STRING
 *      macro convering context variables to an ASCII hash string.
 *
 * MD5 example:
 *
 *      #define DATA_ORDER_IS_LITTLE_ENDIAN
 *
 *      #define HASH_LONG               MD5_LONG
 *      #define HASH_CTX                MD5_CTX
 *      #define HASH_CBLOCK             MD5_CBLOCK
 *      #define HASH_UPDATE             MD5_Update
 *      #define HASH_TRANSFORM          MD5_Transform
 *      #define HASH_FINAL              MD5_Final
 *      #define HASH_BLOCK_DATA_ORDER   md5_block_data_order
 *
 *                                      <appro@fy.chalmers.se>
 */

#if !defined(DATA_ORDER_IS_BIG_ENDIAN) && !defined(DATA_ORDER_IS_LITTLE_ENDIAN)
# error "DATA_ORDER must be defined!"
#endif

#ifndef HASH_CBLOCK
# error "HASH_CBLOCK must be defined!"
#endif
#ifndef HASH_LONG
# error "HASH_LONG must be defined!"
#endif
#ifndef HASH_CTX
# error "HASH_CTX must be defined!"
#endif

#ifndef HASH_UPDATE
# error "HASH_UPDATE must be defined!"
#endif
#ifndef HASH_TRANSFORM
# error "HASH_TRANSFORM must be defined!"
#endif
#ifndef HASH_FINAL
# error "HASH_FINAL must be defined!"
#endif

#ifndef HASH_BLOCK_DATA_ORDER
# error "HASH_BLOCK_DATA_ORDER must be defined!"
#endif

/*
 * Engage compiler specific rotate intrinsic function if available.
 */
#undef ROTATE
#ifndef PEDANTIC
# if defined(_MSC_VER)
#  define ROTATE(a,n)   _lrotl(a,n)
# elif defined(__ICC)
#  define ROTATE(a,n)   _rotl(a,n)
# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
  /*
   * Some GNU C inline assembler templates. Note that these are
   * rotates by *constant* number of bits! But that's exactly
   * what we need here...
   *                                    <appro@fy.chalmers.se>
   */
#  if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
#   define ROTATE(a,n)  ({ register unsigned int ret;   \
                                asm (                   \
                                "roll %1,%0"            \
                                : "=r"(ret)             \
                                : "I"(n), "0"((unsigned int)(a))        \
                                : "cc");                \
                           ret;                         \
                        })
#  elif defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
        defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
#   define ROTATE(a,n)  ({ register unsigned int ret;   \
                                asm (                   \
                                "rlwinm %0,%1,%2,0,31"  \
                                : "=r"(ret)             \
                                : "r"(a), "I"(n));      \
                           ret;                         \
                        })
#  elif defined(__s390x__)
#   define ROTATE(a,n) ({ register unsigned int ret;    \
                                asm ("rll %0,%1,%2"     \
                                : "=r"(ret)             \
                                : "r"(a), "I"(n));      \
                          ret;                          \
                        })
#  endif
# endif
#endif                          /* PEDANTIC */

#ifndef ROTATE
# define ROTATE(a,n)     (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
#endif

#if defined(DATA_ORDER_IS_BIG_ENDIAN)

# ifndef PEDANTIC
#  if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
#   if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
      (defined(__x86_64) || defined(__x86_64__))
#    if !defined(B_ENDIAN)
    /*
     * This gives ~30-40% performance improvement in SHA-256 compiled
     * with gcc [on P4]. Well, first macro to be frank. We can pull
     * this trick on x86* platforms only, because these CPUs can fetch
     * unaligned data without raising an exception.
     */
#     define HOST_c2l(c,l)        ({ unsigned int r=*((const unsigned int *)(c)); \
                                   asm ("bswapl %0":"=r"(r):"0"(r));    \
                                   (c)+=4; (l)=r;                       })
#     define HOST_l2c(l,c)        ({ unsigned int r=(l);                  \
                                   asm ("bswapl %0":"=r"(r):"0"(r));    \
                                   *((unsigned int *)(c))=r; (c)+=4; r; })
#    endif
#   elif defined(__aarch64__)
#    if defined(__BYTE_ORDER__)
#     if defined(__ORDER_LITTLE_ENDIAN__) && __BYTE_ORDER__==__ORDER_LITTLE_ENDIAN__
#      define HOST_c2l(c,l)      ({ unsigned int r;              \
                                   asm ("rev    %w0,%w1"        \
                                        :"=r"(r)                \
                                        :"r"(*((const unsigned int *)(c))));\
                                   (c)+=4; (l)=r;               })
#      define HOST_l2c(l,c)      ({ unsigned int r;              \
                                   asm ("rev    %w0,%w1"        \
                                        :"=r"(r)                \
                                        :"r"((unsigned int)(l)));\
                                   *((unsigned int *)(c))=r; (c)+=4; r; })
#     elif defined(__ORDER_BIG_ENDIAN__) && __BYTE_ORDER__==__ORDER_BIG_ENDIAN__
#      define HOST_c2l(c,l)      ((l)=*((const unsigned int *)(c)), (c)+=4, (l))
#      define HOST_l2c(l,c)      (*((unsigned int *)(c))=(l), (c)+=4, (l))
#     endif
#    endif
#   endif
#  endif
#  if defined(__s390__) || defined(__s390x__)
#   define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, (l))
#   define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, (l))
#  endif
# endif

# ifndef HOST_c2l
#  define HOST_c2l(c,l)   (l =(((unsigned long)(*((c)++)))<<24),          \
                         l|=(((unsigned long)(*((c)++)))<<16),          \
                         l|=(((unsigned long)(*((c)++)))<< 8),          \
                         l|=(((unsigned long)(*((c)++)))    )           )
# endif
# ifndef HOST_l2c
#  define HOST_l2c(l,c)   (*((c)++)=(unsigned char)(((l)>>24)&0xff),      \
                         *((c)++)=(unsigned char)(((l)>>16)&0xff),      \
                         *((c)++)=(unsigned char)(((l)>> 8)&0xff),      \
                         *((c)++)=(unsigned char)(((l)    )&0xff),      \
                         l)
# endif

#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)

# ifndef PEDANTIC
#  if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
#   if defined(__s390x__)
#    define HOST_c2l(c,l)        ({ asm ("lrv    %0,%1"                  \
                                   :"=d"(l) :"m"(*(const unsigned int *)(c)));\
                                   (c)+=4; (l);                         })
#    define HOST_l2c(l,c)        ({ asm ("strv   %1,%0"                  \
                                   :"=m"(*(unsigned int *)(c)) :"d"(l));\
                                   (c)+=4; (l);                         })
#   endif
#  endif
#  if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
#   ifndef B_ENDIAN
    /* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
#    define HOST_c2l(c,l)        ((l)=*((const unsigned int *)(c)), (c)+=4, l)
#    define HOST_l2c(l,c)        (*((unsigned int *)(c))=(l), (c)+=4, l)
#   endif
#  endif
# endif

# ifndef HOST_c2l
#  define HOST_c2l(c,l)   (l =(((unsigned long)(*((c)++)))    ),          \
                         l|=(((unsigned long)(*((c)++)))<< 8),          \
                         l|=(((unsigned long)(*((c)++)))<<16),          \
                         l|=(((unsigned long)(*((c)++)))<<24)           )
# endif
# ifndef HOST_l2c
#  define HOST_l2c(l,c)   (*((c)++)=(unsigned char)(((l)    )&0xff),      \
                         *((c)++)=(unsigned char)(((l)>> 8)&0xff),      \
                         *((c)++)=(unsigned char)(((l)>>16)&0xff),      \
                         *((c)++)=(unsigned char)(((l)>>24)&0xff),      \
                         l)
# endif

#endif

/*
 * Time for some action:-)
 */

int HASH_UPDATE(HASH_CTX *c, const void *data_, size_t len)
{
    const unsigned char *data = data_;
    unsigned char *p;
    HASH_LONG l;
    size_t n;

    if (len == 0)
        return 1;

    l = (c->Nl + (((HASH_LONG) len) << 3)) & 0xffffffffUL;
    /*
     * 95-05-24 eay Fixed a bug with the overflow handling, thanks to Wei Dai
     * <weidai@eskimo.com> for pointing it out.
     */
    if (l < c->Nl)              /* overflow */
        c->Nh++;
    c->Nh += (HASH_LONG) (len >> 29); /* might cause compiler warning on
                                       * 16-bit */
    c->Nl = l;

    n = c->num;
    if (n != 0) {
        p = (unsigned char *)c->data;

        if (len >= HASH_CBLOCK || len + n >= HASH_CBLOCK) {
            memcpy(p + n, data, HASH_CBLOCK - n);
            HASH_BLOCK_DATA_ORDER(c, p, 1);
            n = HASH_CBLOCK - n;
            data += n;
            len -= n;
            c->num = 0;
            memset(p, 0, HASH_CBLOCK); /* keep it zeroed */
        } else {
            memcpy(p + n, data, len);
            c->num += (unsigned int)len;
            return 1;
        }
    }

    n = len / HASH_CBLOCK;
    if (n > 0) {
        HASH_BLOCK_DATA_ORDER(c, data, n);
        n *= HASH_CBLOCK;
        data += n;
        len -= n;
    }

    if (len != 0) {
        p = (unsigned char *)c->data;
        c->num = (unsigned int)len;
        memcpy(p, data, len);
    }
    return 1;
}

void HASH_TRANSFORM(HASH_CTX *c, const unsigned char *data)
{
    HASH_BLOCK_DATA_ORDER(c, data, 1);
}

int HASH_FINAL(unsigned char *md, HASH_CTX *c)
{
    unsigned char *p = (unsigned char *)c->data;
    size_t n = c->num;

    p[n] = 0x80;                /* there is always room for one */
    n++;

    if (n > (HASH_CBLOCK - 8)) {
        memset(p + n, 0, HASH_CBLOCK - n);
        n = 0;
        HASH_BLOCK_DATA_ORDER(c, p, 1);
    }
    memset(p + n, 0, HASH_CBLOCK - 8 - n);

    p += HASH_CBLOCK - 8;
#if   defined(DATA_ORDER_IS_BIG_ENDIAN)
    (void)HOST_l2c(c->Nh, p);
    (void)HOST_l2c(c->Nl, p);
#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
    (void)HOST_l2c(c->Nl, p);
    (void)HOST_l2c(c->Nh, p);
#endif
    p -= HASH_CBLOCK;
    HASH_BLOCK_DATA_ORDER(c, p, 1);
    c->num = 0;
    memset(p, 0, HASH_CBLOCK);

#ifndef HASH_MAKE_STRING
# error "HASH_MAKE_STRING must be defined!"
#else
    HASH_MAKE_STRING(c, md);
#endif

    return 1;
}

#ifndef MD32_REG_T
# if defined(__alpha) || defined(__sparcv9) || defined(__mips)
#  define MD32_REG_T long
/*
 * This comment was originaly written for MD5, which is why it
 * discusses A-D. But it basically applies to all 32-bit digests,
 * which is why it was moved to common header file.
 *
 * In case you wonder why A-D are declared as long and not
 * as MD5_LONG. Doing so results in slight performance
 * boost on LP64 architectures. The catch is we don't
 * really care if 32 MSBs of a 64-bit register get polluted
 * with eventual overflows as we *save* only 32 LSBs in
 * *either* case. Now declaring 'em long excuses the compiler
 * from keeping 32 MSBs zeroed resulting in 13% performance
 * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
 * Well, to be honest it should say that this *prevents*
 * performance degradation.
 *                              <appro@fy.chalmers.se>
 */
# else
/*
 * Above is not absolute and there are LP64 compilers that
 * generate better code if MD32_REG_T is defined int. The above
 * pre-processor condition reflects the circumstances under which
 * the conclusion was made and is subject to further extension.
 *                              <appro@fy.chalmers.se>
 */
#  define MD32_REG_T int
# endif
#endif
Commit	Line	Data
bd3576d2	1	/* ====================================================================
20f7563f	2	* Copyright (c) 1999-2007 The OpenSSL Project. All rights reserved.
bd3576d2 UM	3	*
	4	* Redistribution and use in source and binary forms, with or without
	5	* modification, are permitted provided that the following conditions
	6	* are met:
	7	*
	8	* 1. Redistributions of source code must retain the above copyright
0f113f3e	9	* notice, this list of conditions and the following disclaimer.
bd3576d2 UM	10	*
	11	* 2. Redistributions in binary form must reproduce the above copyright
	12	* notice, this list of conditions and the following disclaimer in
	13	* the documentation and/or other materials provided with the
	14	* distribution.
	15	*
	16	* 3. All advertising materials mentioning features or use of this
	17	* software must display the following acknowledgment:
	18	* "This product includes software developed by the OpenSSL Project
	19	* for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
	20	*
	21	* 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
	22	* endorse or promote products derived from this software without
	23	* prior written permission. For written permission, please contact
	24	* licensing@OpenSSL.org.
	25	*
	26	* 5. Products derived from this software may not be called "OpenSSL"
	27	* nor may "OpenSSL" appear in their names without prior written
	28	* permission of the OpenSSL Project.
	29	*
	30	* 6. Redistributions of any form whatsoever must retain the following
	31	* acknowledgment:
	32	* "This product includes software developed by the OpenSSL Project
	33	* for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
	34	*
	35	* THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
	36	* EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
	37	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
	38	* PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE OpenSSL PROJECT OR
	39	* ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
	40	* SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
	41	* NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
	42	* LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
	43	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
	44	* STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
	45	* ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
	46	* OF THE POSSIBILITY OF SUCH DAMAGE.
	47	* ====================================================================
	48	*
bd3576d2 UM	49	*/
bd3576d2 UM	50
1d97c843	51	/*-
bd3576d2 UM	52	* This is a generic 32 bit "collector" for message digest algorithms.
	53	* Whenever needed it collects input character stream into chunks of
	54	* 32 bit values and invokes a block function that performs actual hash
	55	* calculations.
	56	*
	57	* Porting guide.
	58	*
	59	* Obligatory macros:
	60	*
	61	* DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
0f113f3e	62	* this macro defines byte order of input stream.
bd3576d2	63	* HASH_CBLOCK
0f113f3e	64	* size of a unit chunk HASH_BLOCK operates on.
bd3576d2	65	* HASH_LONG
04f8bcf1	66	* has to be at lest 32 bit wide.
bd3576d2	67	* HASH_CTX
0f113f3e MC	68	* context structure that at least contains following
	69	* members:
	70	* typedef struct {
	71	* ...
	72	* HASH_LONG Nl,Nh;
	73	* either {
	74	* HASH_LONG data[HASH_LBLOCK];
	75	* unsigned char data[HASH_CBLOCK];
	76	* };
	77	* unsigned int num;
	78	* ...
	79	* } HASH_CTX;
	80	* data[] vector is expected to be zeroed upon first call to
	81	* HASH_UPDATE.
bd3576d2	82	* HASH_UPDATE
0f113f3e	83	* name of "Update" function, implemented here.
bd3576d2	84	* HASH_TRANSFORM
0f113f3e	85	* name of "Transform" function, implemented here.
bd3576d2	86	* HASH_FINAL
0f113f3e	87	* name of "Final" function, implemented here.
bd3576d2	88	* HASH_BLOCK_DATA_ORDER
0f113f3e MC	89	* name of "block" function capable of treating unaligned input
0f113f3e MC	90	* message in original (data) byte order, implemented externally.
1cbde6e4	91	* HASH_MAKE_STRING
0f113f3e	92	* macro convering context variables to an ASCII hash string.
bd3576d2	93	*
bd3576d2 UM	94	* MD5 example:
bd3576d2 UM	95	*
0f113f3e	96	* #define DATA_ORDER_IS_LITTLE_ENDIAN
bd3576d2	97	*
0f113f3e	98	* #define HASH_LONG MD5_LONG
0f113f3e MC	99	* #define HASH_CTX MD5_CTX
	100	* #define HASH_CBLOCK MD5_CBLOCK
	101	* #define HASH_UPDATE MD5_Update
	102	* #define HASH_TRANSFORM MD5_Transform
	103	* #define HASH_FINAL MD5_Final
	104	* #define HASH_BLOCK_DATA_ORDER md5_block_data_order
bd3576d2	105	*
0f113f3e	106	* <appro@fy.chalmers.se>
bd3576d2 UM	107	*/
	108
	109	#if !defined(DATA_ORDER_IS_BIG_ENDIAN) && !defined(DATA_ORDER_IS_LITTLE_ENDIAN)
0f113f3e	110	# error "DATA_ORDER must be defined!"
bd3576d2 UM	111	#endif
	112
	113	#ifndef HASH_CBLOCK
0f113f3e	114	# error "HASH_CBLOCK must be defined!"
bd3576d2 UM	115	#endif
bd3576d2 UM	116	#ifndef HASH_LONG
0f113f3e	117	# error "HASH_LONG must be defined!"
bd3576d2 UM	118	#endif
bd3576d2 UM	119	#ifndef HASH_CTX
0f113f3e	120	# error "HASH_CTX must be defined!"
bd3576d2 UM	121	#endif
	122
	123	#ifndef HASH_UPDATE
0f113f3e	124	# error "HASH_UPDATE must be defined!"
bd3576d2 UM	125	#endif
bd3576d2 UM	126	#ifndef HASH_TRANSFORM
0f113f3e	127	# error "HASH_TRANSFORM must be defined!"
bd3576d2 UM	128	#endif
bd3576d2 UM	129	#ifndef HASH_FINAL
0f113f3e	130	# error "HASH_FINAL must be defined!"
bd3576d2 UM	131	#endif
bd3576d2 UM	132
bd3576d2	133	#ifndef HASH_BLOCK_DATA_ORDER
0f113f3e	134	# error "HASH_BLOCK_DATA_ORDER must be defined!"
bd3576d2	135	#endif
bd3576d2 UM	136
	137	/*
	138	* Engage compiler specific rotate intrinsic function if available.
	139	*/
	140	#undef ROTATE
	141	#ifndef PEDANTIC
cf5ecc3e	142	# if defined(_MSC_VER)
0f113f3e	143	# define ROTATE(a,n) _lrotl(a,n)
cf5ecc3e	144	# elif defined(__ICC)
0f113f3e	145	# define ROTATE(a,n) _rotl(a,n)
cf1b7d96	146	# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
bd3576d2 UM	147	/*
	148	* Some GNU C inline assembler templates. Note that these are
	149	* rotates by constant number of bits! But that's exactly
	150	* what we need here...
0f113f3e	151	* <appro@fy.chalmers.se>
bd3576d2	152	*/
2f98abbc	153	# if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
0f113f3e MC	154	# define ROTATE(a,n) ({ register unsigned int ret; \
	155	asm ( \
	156	"roll %1,%0" \
	157	: "=r"(ret) \
	158	: "I"(n), "0"((unsigned int)(a)) \
	159	: "cc"); \
	160	ret; \
	161	})
a9c32ace	162	# elif defined(_ARCH_PPC) \|\| defined(_ARCH_PPC64) \|\| \
0f113f3e MC	163	defined(__powerpc) \|\| defined(__ppc__) \|\| defined(__powerpc64__)
	164	# define ROTATE(a,n) ({ register unsigned int ret; \
	165	asm ( \
	166	"rlwinm %0,%1,%2,0,31" \
	167	: "=r"(ret) \
	168	: "r"(a), "I"(n)); \
	169	ret; \
	170	})
70831126	171	# elif defined(__s390x__)
0f113f3e MC	172	# define ROTATE(a,n) ({ register unsigned int ret; \
	173	asm ("rll %0,%1,%2" \
	174	: "=r"(ret) \
	175	: "r"(a), "I"(n)); \
	176	ret; \
	177	})
bd3576d2 UM	178	# endif
bd3576d2 UM	179	# endif
0f113f3e	180	#endif /* PEDANTIC */
bd3576d2	181
bd3576d2	182	#ifndef ROTATE
0f113f3e	183	# define ROTATE(a,n) (((a)<<(n))\|(((a)&0xffffffff)>>(32-(n))))
bd3576d2 UM	184	#endif
bd3576d2 UM	185
bd3576d2 UM	186	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
bd3576d2 UM	187
0f113f3e MC	188	# ifndef PEDANTIC
	189	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
	190	# if ((defined(__i386) \|\| defined(__i386__)) && !defined(I386_ONLY)) \|\| \
af2c2823	191	(defined(__x86_64) \|\| defined(__x86_64__))
0f113f3e	192	# if !defined(B_ENDIAN)
a2eb9688 AP	193	/*
	194	* This gives ~30-40% performance improvement in SHA-256 compiled
	195	* with gcc [on P4]. Well, first macro to be frank. We can pull
	196	* this trick on x86* platforms only, because these CPUs can fetch
	197	* unaligned data without raising an exception.
	198	*/
0f113f3e MC	199	# define HOST_c2l(c,l) ({ unsigned int r=((const unsigned int )(c)); \
	200	asm ("bswapl %0":"=r"(r):"0"(r)); \
	201	(c)+=4; (l)=r; })
	202	# define HOST_l2c(l,c) ({ unsigned int r=(l); \
	203	asm ("bswapl %0":"=r"(r):"0"(r)); \
	204	((unsigned int )(c))=r; (c)+=4; r; })
	205	# endif
	206	# elif defined(__aarch64__)
	207	# if defined(__BYTE_ORDER__)
	208	# if defined(__ORDER_LITTLE_ENDIAN__) && __BYTE_ORDER__==__ORDER_LITTLE_ENDIAN__
	209	# define HOST_c2l(c,l) ({ unsigned int r; \
	210	asm ("rev %w0,%w1" \
	211	:"=r"(r) \
	212	:"r"(((const unsigned int )(c))));\
	213	(c)+=4; (l)=r; })
	214	# define HOST_l2c(l,c) ({ unsigned int r; \
	215	asm ("rev %w0,%w1" \
	216	:"=r"(r) \
	217	:"r"((unsigned int)(l)));\
	218	((unsigned int )(c))=r; (c)+=4; r; })
	219	# elif defined(__ORDER_BIG_ENDIAN__) && __BYTE_ORDER__==__ORDER_BIG_ENDIAN__
	220	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, (l))
	221	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, (l))
	222	# endif
039081b8 AP	223	# endif
039081b8 AP	224	# endif
a2eb9688	225	# endif
0f113f3e MC	226	# if defined(__s390__) \|\| defined(__s390x__)
	227	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, (l))
	228	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, (l))
	229	# endif
a2eb9688	230	# endif
a2eb9688	231
0f113f3e MC	232	# ifndef HOST_c2l
	233	# define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++)))<<24), \
	234	l\|=(((unsigned long)(*((c)++)))<<16), \
	235	l\|=(((unsigned long)(*((c)++)))<< 8), \
	236	l\|=(((unsigned long)(*((c)++))) ) )
	237	# endif
	238	# ifndef HOST_l2c
	239	# define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	240	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	241	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	242	*((c)++)=(unsigned char)(((l) )&0xff), \
	243	l)
	244	# endif
bd3576d2 UM	245
	246	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
	247
0f113f3e MC	248	# ifndef PEDANTIC
	249	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
	250	# if defined(__s390x__)
	251	# define HOST_c2l(c,l) ({ asm ("lrv %0,%1" \
	252	:"=d"(l) :"m"((const unsigned int )(c)));\
	253	(c)+=4; (l); })
	254	# define HOST_l2c(l,c) ({ asm ("strv %1,%0" \
	255	:"=m"((unsigned int )(c)) :"d"(l));\
	256	(c)+=4; (l); })
	257	# endif
b38c0add	258	# endif
0f113f3e MC	259	# if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
0f113f3e MC	260	# ifndef B_ENDIAN
021e5043	261	/* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
0f113f3e MC	262	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, l)
	263	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, l)
	264	# endif
021e5043	265	# endif
1110cea0	266	# endif
a2eb9688	267
0f113f3e MC	268	# ifndef HOST_c2l
	269	# define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++))) ), \
	270	l\|=(((unsigned long)(*((c)++)))<< 8), \
	271	l\|=(((unsigned long)(*((c)++)))<<16), \
	272	l\|=(((unsigned long)(*((c)++)))<<24) )
	273	# endif
	274	# ifndef HOST_l2c
	275	# define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
	276	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	277	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	278	*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	279	l)
	280	# endif
bd3576d2 UM	281
	282	#endif
	283
	284	/*
	285	* Time for some action:-)
	286	*/
	287
0f113f3e MC	288	int HASH_UPDATE(HASH_CTX c, const void data_, size_t len)
	289	{
	290	const unsigned char *data = data_;
	291	unsigned char *p;
	292	HASH_LONG l;
	293	size_t n;
bd3576d2	294
0f113f3e MC	295	if (len == 0)
0f113f3e MC	296	return 1;
bd3576d2	297
0f113f3e MC	298	l = (c->Nl + (((HASH_LONG) len) << 3)) & 0xffffffffUL;
	299	/*
	300	* 95-05-24 eay Fixed a bug with the overflow handling, thanks to Wei Dai
	301	* <weidai@eskimo.com> for pointing it out.
	302	*/
	303	if (l < c->Nl) /* overflow */
	304	c->Nh++;
	305	c->Nh += (HASH_LONG) (len >> 29); /* might cause compiler warning on
	306	* 16-bit */
	307	c->Nl = l;
	308
	309	n = c->num;
	310	if (n != 0) {
	311	p = (unsigned char *)c->data;
	312
	313	if (len >= HASH_CBLOCK \|\| len + n >= HASH_CBLOCK) {
	314	memcpy(p + n, data, HASH_CBLOCK - n);
	315	HASH_BLOCK_DATA_ORDER(c, p, 1);
	316	n = HASH_CBLOCK - n;
	317	data += n;
	318	len -= n;
	319	c->num = 0;
	320	memset(p, 0, HASH_CBLOCK); /* keep it zeroed */
	321	} else {
	322	memcpy(p + n, data, len);
	323	c->num += (unsigned int)len;
	324	return 1;
	325	}
	326	}
	327
	328	n = len / HASH_CBLOCK;
	329	if (n > 0) {
	330	HASH_BLOCK_DATA_ORDER(c, data, n);
	331	n *= HASH_CBLOCK;
	332	data += n;
	333	len -= n;
	334	}
	335
	336	if (len != 0) {
	337	p = (unsigned char *)c->data;
	338	c->num = (unsigned int)len;
	339	memcpy(p, data, len);
	340	}
	341	return 1;
	342	}
	343
	344	void HASH_TRANSFORM(HASH_CTX c, const unsigned char data)
	345	{
	346	HASH_BLOCK_DATA_ORDER(c, data, 1);
	347	}
	348
	349	int HASH_FINAL(unsigned char md, HASH_CTX c)
	350	{
	351	unsigned char p = (unsigned char )c->data;
	352	size_t n = c->num;
	353
	354	p[n] = 0x80; /* there is always room for one */
	355	n++;
	356
	357	if (n > (HASH_CBLOCK - 8)) {
	358	memset(p + n, 0, HASH_CBLOCK - n);
	359	n = 0;
	360	HASH_BLOCK_DATA_ORDER(c, p, 1);
	361	}
362	memset(p + n, 0, HASH_CBLOCK - 8 - n);
363
364	p += HASH_CBLOCK - 8;
bd3576d2	365	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
0f113f3e MC	366	(void)HOST_l2c(c->Nh, p);
0f113f3e MC	367	(void)HOST_l2c(c->Nl, p);
bd3576d2	368	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
0f113f3e MC	369	(void)HOST_l2c(c->Nl, p);
0f113f3e MC	370	(void)HOST_l2c(c->Nh, p);
bd3576d2	371	#endif
0f113f3e MC	372	p -= HASH_CBLOCK;
	373	HASH_BLOCK_DATA_ORDER(c, p, 1);
	374	c->num = 0;
	375	memset(p, 0, HASH_CBLOCK);
bd3576d2	376
1cbde6e4	377	#ifndef HASH_MAKE_STRING
0f113f3e	378	# error "HASH_MAKE_STRING must be defined!"
1cbde6e4	379	#else
0f113f3e	380	HASH_MAKE_STRING(c, md);
1cbde6e4	381	#endif
bd3576d2	382
0f113f3e MC	383	return 1;
0f113f3e MC	384	}
2f98abbc AP	385
2f98abbc AP	386	#ifndef MD32_REG_T
0f113f3e MC	387	# if defined(__alpha) \|\| defined(__sparcv9) \|\| defined(__mips)
0f113f3e MC	388	# define MD32_REG_T long
2f98abbc AP	389	/*
	390	* This comment was originaly written for MD5, which is why it
	391	* discusses A-D. But it basically applies to all 32-bit digests,
	392	* which is why it was moved to common header file.
	393	*
	394	* In case you wonder why A-D are declared as long and not
	395	* as MD5_LONG. Doing so results in slight performance
	396	* boost on LP64 architectures. The catch is we don't
	397	* really care if 32 MSBs of a 64-bit register get polluted
	398	* with eventual overflows as we save only 32 LSBs in
	399	* either case. Now declaring 'em long excuses the compiler
	400	* from keeping 32 MSBs zeroed resulting in 13% performance
	401	* improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
0f113f3e	402	* Well, to be honest it should say that this prevents
2f98abbc	403	* performance degradation.
0f113f3e	404	* <appro@fy.chalmers.se>
30ab7af2	405	*/
0f113f3e	406	# else
30ab7af2 AP	407	/*
	408	* Above is not absolute and there are LP64 compilers that
	409	* generate better code if MD32_REG_T is defined int. The above
	410	* pre-processor condition reflects the circumstances under which
	411	* the conclusion was made and is subject to further extension.
0f113f3e	412	* <appro@fy.chalmers.se>
2f98abbc	413	*/
0f113f3e MC	414	# define MD32_REG_T int
0f113f3e MC	415	# endif
2f98abbc	416	#endif