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

/* crypto/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, if it's wider, then
 *	HASH_LONG_LOG2 *has to* be defined along
 * 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_LONG_LOG2		MD5_LONG_LOG2
 *	#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)++)))    ),		\
			 l)
#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),		\
			 l)
#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 UM	1	/* crypto/md32_common.h */
bd3576d2 UM	2	/* ====================================================================
20f7563f	3	* Copyright (c) 1999-2007 The OpenSSL Project. All rights reserved.
bd3576d2 UM	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	*
bd3576d2 UM	50	*/
	51
	52	/*
	53	* This is a generic 32 bit "collector" for message digest algorithms.
	54	* Whenever needed it collects input character stream into chunks of
	55	* 32 bit values and invokes a block function that performs actual hash
	56	* calculations.
	57	*
	58	* Porting guide.
	59	*
	60	* Obligatory macros:
	61	*
	62	* DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
	63	* this macro defines byte order of input stream.
	64	* HASH_CBLOCK
	65	* size of a unit chunk HASH_BLOCK operates on.
	66	* HASH_LONG
f6fb2c95 AP	67	* has to be at lest 32 bit wide, if it's wider, then
f6fb2c95 AP	68	* HASH_LONG_LOG2 has to be defined along
bd3576d2 UM	69	* HASH_CTX
	70	* context structure that at least contains following
	71	* members:
	72	* typedef struct {
	73	* ...
	74	* HASH_LONG Nl,Nh;
c69ed6ea	75	* either {
f6fb2c95 AP	76	* HASH_LONG data[HASH_LBLOCK];
f6fb2c95 AP	77	* unsigned char data[HASH_CBLOCK];
c69ed6ea	78	* };
16ab8a93	79	* unsigned int num;
bd3576d2 UM	80	* ...
bd3576d2 UM	81	* } HASH_CTX;
c69ed6ea AP	82	* data[] vector is expected to be zeroed upon first call to
c69ed6ea AP	83	* HASH_UPDATE.
bd3576d2 UM	84	* HASH_UPDATE
	85	* name of "Update" function, implemented here.
	86	* HASH_TRANSFORM
	87	* name of "Transform" function, implemented here.
	88	* HASH_FINAL
	89	* name of "Final" function, implemented here.
bd3576d2	90	* HASH_BLOCK_DATA_ORDER
c69ed6ea AP	91	* name of "block" function capable of treating unaligned input
c69ed6ea AP	92	* message in original (data) byte order, implemented externally.
1cbde6e4 AP	93	* HASH_MAKE_STRING
1cbde6e4 AP	94	* macro convering context variables to an ASCII hash string.
bd3576d2	95	*
bd3576d2 UM	96	* MD5 example:
	97	*
	98	* #define DATA_ORDER_IS_LITTLE_ENDIAN
	99	*
	100	* #define HASH_LONG MD5_LONG
f6fb2c95	101	* #define HASH_LONG_LOG2 MD5_LONG_LOG2
bd3576d2 UM	102	* #define HASH_CTX MD5_CTX
bd3576d2 UM	103	* #define HASH_CBLOCK MD5_CBLOCK
bd3576d2 UM	104	* #define HASH_UPDATE MD5_Update
	105	* #define HASH_TRANSFORM MD5_Transform
	106	* #define HASH_FINAL MD5_Final
bd3576d2 UM	107	* #define HASH_BLOCK_DATA_ORDER md5_block_data_order
	108	*
	109	* <appro@fy.chalmers.se>
	110	*/
	111
	112	#if !defined(DATA_ORDER_IS_BIG_ENDIAN) && !defined(DATA_ORDER_IS_LITTLE_ENDIAN)
	113	#error "DATA_ORDER must be defined!"
	114	#endif
	115
	116	#ifndef HASH_CBLOCK
	117	#error "HASH_CBLOCK must be defined!"
	118	#endif
	119	#ifndef HASH_LONG
	120	#error "HASH_LONG must be defined!"
	121	#endif
	122	#ifndef HASH_CTX
	123	#error "HASH_CTX must be defined!"
	124	#endif
	125
	126	#ifndef HASH_UPDATE
	127	#error "HASH_UPDATE must be defined!"
	128	#endif
	129	#ifndef HASH_TRANSFORM
	130	#error "HASH_TRANSFORM must be defined!"
	131	#endif
	132	#ifndef HASH_FINAL
	133	#error "HASH_FINAL must be defined!"
	134	#endif
	135
bd3576d2 UM	136	#ifndef HASH_BLOCK_DATA_ORDER
	137	#error "HASH_BLOCK_DATA_ORDER must be defined!"
	138	#endif
bd3576d2 UM	139
	140	/*
	141	* Engage compiler specific rotate intrinsic function if available.
	142	*/
	143	#undef ROTATE
	144	#ifndef PEDANTIC
cf5ecc3e	145	# if defined(_MSC_VER)
1cbde6e4	146	# define ROTATE(a,n) _lrotl(a,n)
cf5ecc3e AP	147	# elif defined(__ICC)
cf5ecc3e AP	148	# define ROTATE(a,n) _rotl(a,n)
cf1b7d96	149	# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
bd3576d2 UM	150	/*
	151	* Some GNU C inline assembler templates. Note that these are
	152	* rotates by constant number of bits! But that's exactly
	153	* what we need here...
bd3576d2 UM	154	* <appro@fy.chalmers.se>
bd3576d2 UM	155	*/
2f98abbc	156	# if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
bd3576d2	157	# define ROTATE(a,n) ({ register unsigned int ret; \
0fad6cb7	158	asm ( \
bd3576d2 UM	159	"roll %1,%0" \
bd3576d2 UM	160	: "=r"(ret) \
30ab7af2	161	: "I"(n), "0"((unsigned int)(a)) \
bd3576d2 UM	162	: "cc"); \
	163	ret; \
	164	})
a9c32ace AP	165	# elif defined(_ARCH_PPC) \|\| defined(_ARCH_PPC64) \|\| \
a9c32ace AP	166	defined(__powerpc) \|\| defined(__ppc__) \|\| defined(__powerpc64__)
bd3576d2	167	# define ROTATE(a,n) ({ register unsigned int ret; \
0fad6cb7	168	asm ( \
bd3576d2 UM	169	"rlwinm %0,%1,%2,0,31" \
	170	: "=r"(ret) \
	171	: "r"(a), "I"(n)); \
	172	ret; \
	173	})
70831126	174	# elif defined(__s390x__)
b38c0add AP	175	# define ROTATE(a,n) ({ register unsigned int ret; \
	176	asm ("rll %0,%1,%2" \
	177	: "=r"(ret) \
70831126	178	: "r"(a), "I"(n)); \
b38c0add AP	179	ret; \
b38c0add AP	180	})
bd3576d2 UM	181	# endif
bd3576d2 UM	182	# endif
bd3576d2 UM	183	#endif /* PEDANTIC */
bd3576d2 UM	184
bd3576d2 UM	185	#ifndef ROTATE
	186	#define ROTATE(a,n) (((a)<<(n))\|(((a)&0xffffffff)>>(32-(n))))
	187	#endif
	188
bd3576d2 UM	189	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
bd3576d2 UM	190
a2eb9688 AP	191	#ifndef PEDANTIC
a2eb9688 AP	192	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
6d9c46b8	193	# if ((defined(__i386) \|\| defined(__i386__)) && !defined(I386_ONLY)) \|\| \
af2c2823	194	(defined(__x86_64) \|\| defined(__x86_64__))
6c1fc273	195	# if !defined(B_ENDIAN)
a2eb9688 AP	196	/*
	197	* This gives ~30-40% performance improvement in SHA-256 compiled
	198	* with gcc [on P4]. Well, first macro to be frank. We can pull
	199	* this trick on x86* platforms only, because these CPUs can fetch
	200	* unaligned data without raising an exception.
	201	*/
385c8e89 AP	202	# define HOST_c2l(c,l) ({ unsigned int r=((const unsigned int )(c)); \
	203	asm ("bswapl %0":"=r"(r):"0"(r)); \
	204	(c)+=4; (l)=r; })
a2eb9688 AP	205	# define HOST_l2c(l,c) ({ unsigned int r=(l); \
	206	asm ("bswapl %0":"=r"(r):"0"(r)); \
	207	((unsigned int )(c))=r; (c)+=4; r; })
6c1fc273	208	# endif
039081b8 AP	209	# elif defined(__aarch64__)
	210	# if defined(__BYTE_ORDER__)
	211	# if defined(__ORDER_LITTLE_ENDIAN__) && __BYTE_ORDER__==__ORDER_LITTLE_ENDIAN__
	212	# define HOST_c2l(c,l) ({ unsigned int r; \
	213	asm ("rev %w0,%w1" \
	214	:"=r"(r) \
	215	:"r"(((const unsigned int )(c))));\
	216	(c)+=4; (l)=r; })
	217	# define HOST_l2c(l,c) ({ unsigned int r; \
	218	asm ("rev %w0,%w1" \
	219	:"=r"(r) \
	220	:"r"((unsigned int)(l)));\
	221	((unsigned int )(c))=r; (c)+=4; r; })
	222	# elif defined(__ORDER_BIG_ENDIAN__) && __BYTE_ORDER__==__ORDER_BIG_ENDIAN__
	223	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, (l))
	224	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, (l))
	225	# endif
	226	# endif
a2eb9688 AP	227	# endif
a2eb9688 AP	228	# endif
021e5043 AP	229	# if defined(__s390__) \|\| defined(__s390x__)
	230	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, (l))
	231	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, (l))
	232	# endif
b38c0add	233	#endif
a2eb9688 AP	234
a2eb9688 AP	235	#ifndef HOST_c2l
bd3576d2 UM	236	#define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++)))<<24), \
	237	l\|=(((unsigned long)(*((c)++)))<<16), \
	238	l\|=(((unsigned long)(*((c)++)))<< 8), \
	239	l\|=(((unsigned long)(*((c)++))) ), \
	240	l)
a2eb9688	241	#endif
a2eb9688	242	#ifndef HOST_l2c
bd3576d2 UM	243	#define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	244	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	245	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	246	*((c)++)=(unsigned char)(((l) )&0xff), \
	247	l)
a2eb9688	248	#endif
bd3576d2 UM	249
	250	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
	251
b38c0add AP	252	#ifndef PEDANTIC
	253	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
	254	# if defined(__s390x__)
d8c7bd6e AP	255	# define HOST_c2l(c,l) ({ asm ("lrv %0,%1" \
d8c7bd6e AP	256	:"=d"(l) :"m"((const unsigned int )(c)));\
b38c0add	257	(c)+=4; (l); })
d8c7bd6e AP	258	# define HOST_l2c(l,c) ({ asm ("strv %1,%0" \
d8c7bd6e AP	259	:"=m"((unsigned int )(c)) :"d"(l));\
b38c0add AP	260	(c)+=4; (l); })
	261	# endif
	262	# endif
021e5043 AP	263	# if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
	264	# ifndef B_ENDIAN
	265	/* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
	266	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, l)
	267	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, l)
	268	# endif
1110cea0	269	# endif
a2eb9688 AP	270	#endif
	271
	272	#ifndef HOST_c2l
bd3576d2 UM	273	#define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++))) ), \
	274	l\|=(((unsigned long)(*((c)++)))<< 8), \
	275	l\|=(((unsigned long)(*((c)++)))<<16), \
	276	l\|=(((unsigned long)(*((c)++)))<<24), \
	277	l)
a2eb9688	278	#endif
a2eb9688	279	#ifndef HOST_l2c
bd3576d2 UM	280	#define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
	281	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	282	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	283	*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	284	l)
a2eb9688	285	#endif
bd3576d2 UM	286
	287	#endif
	288
	289	/*
	290	* Time for some action:-)
	291	*/
	292
9e0aad9f	293	int HASH_UPDATE (HASH_CTX c, const void data_, size_t len)
bd3576d2	294	{
bd445703	295	const unsigned char *data=data_;
c69ed6ea AP	296	unsigned char *p;
	297	HASH_LONG l;
	298	size_t n;
bd3576d2	299
2dc769a1	300	if (len==0) return 1;
bd3576d2	301
9e0aad9f	302	l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
bd3576d2 UM	303	/* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
	304	* Wei Dai <weidai@eskimo.com> for pointing it out. */
	305	if (l < c->Nl) /* overflow */
	306	c->Nh++;
5cabcf96	307	c->Nh+=(HASH_LONG)(len>>29); /* might cause compiler warning on 16-bit */
bd3576d2 UM	308	c->Nl=l;
bd3576d2 UM	309
c69ed6ea AP	310	n = c->num;
c69ed6ea AP	311	if (n != 0)
bd3576d2	312	{
c69ed6ea	313	p=(unsigned char *)c->data;
bd3576d2	314
f8d6be3f	315	if (len >= HASH_CBLOCK \|\| len+n >= HASH_CBLOCK)
bd3576d2	316	{
c69ed6ea AP	317	memcpy (p+n,data,HASH_CBLOCK-n);
	318	HASH_BLOCK_DATA_ORDER (c,p,1);
	319	n = HASH_CBLOCK-n;
	320	data += n;
	321	len -= n;
	322	c->num = 0;
	323	memset (p,0,HASH_CBLOCK); /* keep it zeroed */
bd3576d2 UM	324	}
	325	else
	326	{
c69ed6ea AP	327	memcpy (p+n,data,len);
c69ed6ea AP	328	c->num += (unsigned int)len;
2dc769a1	329	return 1;
bd3576d2 UM	330	}
	331	}
	332
c69ed6ea AP	333	n = len/HASH_CBLOCK;
c69ed6ea AP	334	if (n > 0)
bd3576d2	335	{
c69ed6ea AP	336	HASH_BLOCK_DATA_ORDER (c,data,n);
	337	n *= HASH_CBLOCK;
	338	data += n;
	339	len -= n;
bd3576d2 UM	340	}
bd3576d2 UM	341
c69ed6ea	342	if (len != 0)
bd3576d2	343	{
c69ed6ea	344	p = (unsigned char *)c->data;
5cabcf96	345	c->num = (unsigned int)len;
c69ed6ea	346	memcpy (p,data,len);
bd3576d2	347	}
2dc769a1	348	return 1;
bd3576d2 UM	349	}
	350
	351
ac7d0785	352	void HASH_TRANSFORM (HASH_CTX c, const unsigned char data)
bd3576d2	353	{
cdfb093f	354	HASH_BLOCK_DATA_ORDER (c,data,1);
bd3576d2 UM	355	}
	356
	357
2dc769a1	358	int HASH_FINAL (unsigned char md, HASH_CTX c)
bd3576d2	359	{
c69ed6ea AP	360	unsigned char p = (unsigned char )c->data;
c69ed6ea AP	361	size_t n = c->num;
bd3576d2	362
c69ed6ea AP	363	p[n] = 0x80; /* there is always room for one */
	364	n++;
	365
	366	if (n > (HASH_CBLOCK-8))
bd3576d2	367	{
20f7563f AP	368	memset (p+n,0,HASH_CBLOCK-n);
20f7563f AP	369	n=0;
c69ed6ea	370	HASH_BLOCK_DATA_ORDER (c,p,1);
bd3576d2	371	}
20f7563f	372	memset (p+n,0,HASH_CBLOCK-8-n);
bd3576d2	373
c69ed6ea	374	p += HASH_CBLOCK-8;
bd3576d2	375	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
c69ed6ea AP	376	(void)HOST_l2c(c->Nh,p);
c69ed6ea AP	377	(void)HOST_l2c(c->Nl,p);
bd3576d2	378	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
c69ed6ea AP	379	(void)HOST_l2c(c->Nl,p);
c69ed6ea AP	380	(void)HOST_l2c(c->Nh,p);
bd3576d2	381	#endif
c69ed6ea AP	382	p -= HASH_CBLOCK;
	383	HASH_BLOCK_DATA_ORDER (c,p,1);
	384	c->num=0;
	385	memset (p,0,HASH_CBLOCK);
bd3576d2	386
1cbde6e4 AP	387	#ifndef HASH_MAKE_STRING
	388	#error "HASH_MAKE_STRING must be defined!"
	389	#else
	390	HASH_MAKE_STRING(c,md);
	391	#endif
bd3576d2	392
2dc769a1	393	return 1;
bd3576d2	394	}
2f98abbc AP	395
2f98abbc AP	396	#ifndef MD32_REG_T
30ab7af2	397	#if defined(__alpha) \|\| defined(__sparcv9) \|\| defined(__mips)
2f98abbc AP	398	#define MD32_REG_T long
	399	/*
	400	* This comment was originaly written for MD5, which is why it
	401	* discusses A-D. But it basically applies to all 32-bit digests,
	402	* which is why it was moved to common header file.
	403	*
	404	* In case you wonder why A-D are declared as long and not
	405	* as MD5_LONG. Doing so results in slight performance
	406	* boost on LP64 architectures. The catch is we don't
	407	* really care if 32 MSBs of a 64-bit register get polluted
	408	* with eventual overflows as we save only 32 LSBs in
	409	* either case. Now declaring 'em long excuses the compiler
	410	* from keeping 32 MSBs zeroed resulting in 13% performance
	411	* improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
	412	* Well, to be honest it should say that this prevents
	413	* performance degradation.
	414	* <appro@fy.chalmers.se>
30ab7af2 AP	415	*/
	416	#else
	417	/*
	418	* Above is not absolute and there are LP64 compilers that
	419	* generate better code if MD32_REG_T is defined int. The above
	420	* pre-processor condition reflects the circumstances under which
	421	* the conclusion was made and is subject to further extension.
2f98abbc AP	422	* <appro@fy.chalmers.se>
2f98abbc AP	423	*/
30ab7af2 AP	424	#define MD32_REG_T int
30ab7af2 AP	425	#endif
2f98abbc	426	#endif