[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) || defined(__ICC)
#  define ROTATE(a,n)	_lrotl(a,n)
# elif defined(__MWERKS__)
#  if defined(__POWERPC__)
#   define ROTATE(a,n)	__rlwinm(a,n,0,31)
#  elif defined(__MC68K__)
    /* Motorola specific tweak. <appro@fy.chalmers.se> */
#   define ROTATE(a,n)	( n<24 ? __rol(a,n) : __ror(a,32-n) )
#  else
#   define ROTATE(a,n)	__rol(a,n)
#  endif
# 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"(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
#  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

#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	%0,0(%1)"		\
					:"=r"(l) : "r"(c));		\
				   (c)+=4; (l);				})
#   define HOST_l2c(l,c)	({ asm ("strv	%0,0(%1)"		\
					: : "r"(l),"r"(c) : "memory");	\
				   (c)+=4; (l);				})
#  endif
# 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

#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
#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>
 * Apparently there're LP64 compilers that generate better
 * code if A-D are declared int. Most notably GCC-x86_64
 * generates better code.
 *				<appro@fy.chalmers.se>
 */
#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
1a979201	145	# if defined(_MSC_VER) \|\| defined(__ICC)
1cbde6e4 AP	146	# define ROTATE(a,n) _lrotl(a,n)
1cbde6e4 AP	147	# elif defined(__MWERKS__)
1eab9a1f	148	# if defined(__POWERPC__)
a7c5241f	149	# define ROTATE(a,n) __rlwinm(a,n,0,31)
1eab9a1f AP	150	# elif defined(__MC68K__)
	151	/* Motorola specific tweak. <appro@fy.chalmers.se> */
	152	# define ROTATE(a,n) ( n<24 ? __rol(a,n) : __ror(a,32-n) )
9a1e34e5 AP	153	# else
	154	# define ROTATE(a,n) __rol(a,n)
	155	# endif
cf1b7d96	156	# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
bd3576d2 UM	157	/*
	158	* Some GNU C inline assembler templates. Note that these are
	159	* rotates by constant number of bits! But that's exactly
	160	* what we need here...
bd3576d2 UM	161	* <appro@fy.chalmers.se>
bd3576d2 UM	162	*/
2f98abbc	163	# if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
bd3576d2	164	# define ROTATE(a,n) ({ register unsigned int ret; \
0fad6cb7	165	asm ( \
bd3576d2 UM	166	"roll %1,%0" \
	167	: "=r"(ret) \
	168	: "I"(n), "0"(a) \
	169	: "cc"); \
	170	ret; \
	171	})
a9c32ace AP	172	# elif defined(_ARCH_PPC) \|\| defined(_ARCH_PPC64) \|\| \
a9c32ace AP	173	defined(__powerpc) \|\| defined(__ppc__) \|\| defined(__powerpc64__)
bd3576d2	174	# define ROTATE(a,n) ({ register unsigned int ret; \
0fad6cb7	175	asm ( \
bd3576d2 UM	176	"rlwinm %0,%1,%2,0,31" \
	177	: "=r"(ret) \
	178	: "r"(a), "I"(n)); \
	179	ret; \
	180	})
70831126	181	# elif defined(__s390x__)
b38c0add AP	182	# define ROTATE(a,n) ({ register unsigned int ret; \
	183	asm ("rll %0,%1,%2" \
	184	: "=r"(ret) \
70831126	185	: "r"(a), "I"(n)); \
b38c0add AP	186	ret; \
b38c0add AP	187	})
bd3576d2 UM	188	# endif
bd3576d2 UM	189	# endif
bd3576d2 UM	190	#endif /* PEDANTIC */
bd3576d2 UM	191
bd3576d2 UM	192	#ifndef ROTATE
	193	#define ROTATE(a,n) (((a)<<(n))\|(((a)&0xffffffff)>>(32-(n))))
	194	#endif
	195
bd3576d2 UM	196	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
bd3576d2 UM	197
a2eb9688 AP	198	#ifndef PEDANTIC
a2eb9688 AP	199	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
6d9c46b8	200	# if ((defined(__i386) \|\| defined(__i386__)) && !defined(I386_ONLY)) \|\| \
af2c2823	201	(defined(__x86_64) \|\| defined(__x86_64__))
6c1fc273	202	# if !defined(B_ENDIAN)
a2eb9688 AP	203	/*
	204	* This gives ~30-40% performance improvement in SHA-256 compiled
	205	* with gcc [on P4]. Well, first macro to be frank. We can pull
	206	* this trick on x86* platforms only, because these CPUs can fetch
	207	* unaligned data without raising an exception.
	208	*/
385c8e89 AP	209	# define HOST_c2l(c,l) ({ unsigned int r=((const unsigned int )(c)); \
	210	asm ("bswapl %0":"=r"(r):"0"(r)); \
	211	(c)+=4; (l)=r; })
a2eb9688 AP	212	# define HOST_l2c(l,c) ({ unsigned int r=(l); \
	213	asm ("bswapl %0":"=r"(r):"0"(r)); \
	214	((unsigned int )(c))=r; (c)+=4; r; })
6c1fc273	215	# endif
a2eb9688 AP	216	# endif
	217	# endif
	218	#endif
b38c0add AP	219	#if defined(__s390__) \|\| defined(__s390x__)
	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
a2eb9688 AP	223
a2eb9688 AP	224	#ifndef HOST_c2l
bd3576d2 UM	225	#define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++)))<<24), \
	226	l\|=(((unsigned long)(*((c)++)))<<16), \
	227	l\|=(((unsigned long)(*((c)++)))<< 8), \
	228	l\|=(((unsigned long)(*((c)++))) ), \
	229	l)
a2eb9688	230	#endif
a2eb9688	231	#ifndef HOST_l2c
bd3576d2 UM	232	#define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	233	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	234	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	235	*((c)++)=(unsigned char)(((l) )&0xff), \
	236	l)
a2eb9688	237	#endif
bd3576d2 UM	238
	239	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
	240
b38c0add AP	241	#ifndef PEDANTIC
	242	# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
	243	# if defined(__s390x__)
	244	# define HOST_c2l(c,l) ({ asm ("lrv %0,0(%1)" \
	245	:"=r"(l) : "r"(c)); \
	246	(c)+=4; (l); })
	247	# define HOST_l2c(l,c) ({ asm ("strv %0,0(%1)" \
	248	: : "r"(l),"r"(c) : "memory"); \
	249	(c)+=4; (l); })
	250	# endif
	251	# endif
	252	#endif
a2eb9688	253	#if defined(__i386) \|\| defined(__i386__) \|\| defined(__x86_64) \|\| defined(__x86_64__)
1110cea0 RL	254	# ifndef B_ENDIAN
	255	/* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
	256	# define HOST_c2l(c,l) ((l)=((const unsigned int )(c)), (c)+=4, l)
	257	# define HOST_l2c(l,c) (((unsigned int )(c))=(l), (c)+=4, l)
	258	# endif
a2eb9688 AP	259	#endif
	260
	261	#ifndef HOST_c2l
bd3576d2 UM	262	#define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++))) ), \
	263	l\|=(((unsigned long)(*((c)++)))<< 8), \
	264	l\|=(((unsigned long)(*((c)++)))<<16), \
	265	l\|=(((unsigned long)(*((c)++)))<<24), \
	266	l)
a2eb9688	267	#endif
a2eb9688	268	#ifndef HOST_l2c
bd3576d2 UM	269	#define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
	270	*((c)++)=(unsigned char)(((l)>> 8)&0xff), \
	271	*((c)++)=(unsigned char)(((l)>>16)&0xff), \
	272	*((c)++)=(unsigned char)(((l)>>24)&0xff), \
	273	l)
a2eb9688	274	#endif
bd3576d2 UM	275
	276	#endif
	277
	278	/*
	279	* Time for some action:-)
	280	*/
	281
9e0aad9f	282	int HASH_UPDATE (HASH_CTX c, const void data_, size_t len)
bd3576d2	283	{
bd445703	284	const unsigned char *data=data_;
c69ed6ea AP	285	unsigned char *p;
	286	HASH_LONG l;
	287	size_t n;
bd3576d2	288
2dc769a1	289	if (len==0) return 1;
bd3576d2	290
9e0aad9f	291	l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
bd3576d2 UM	292	/* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
	293	* Wei Dai <weidai@eskimo.com> for pointing it out. */
	294	if (l < c->Nl) /* overflow */
	295	c->Nh++;
5cabcf96	296	c->Nh+=(HASH_LONG)(len>>29); /* might cause compiler warning on 16-bit */
bd3576d2 UM	297	c->Nl=l;
bd3576d2 UM	298
c69ed6ea AP	299	n = c->num;
c69ed6ea AP	300	if (n != 0)
bd3576d2	301	{
c69ed6ea	302	p=(unsigned char *)c->data;
bd3576d2	303
f8d6be3f	304	if (len >= HASH_CBLOCK \|\| len+n >= HASH_CBLOCK)
bd3576d2	305	{
c69ed6ea AP	306	memcpy (p+n,data,HASH_CBLOCK-n);
	307	HASH_BLOCK_DATA_ORDER (c,p,1);
	308	n = HASH_CBLOCK-n;
	309	data += n;
	310	len -= n;
	311	c->num = 0;
	312	memset (p,0,HASH_CBLOCK); /* keep it zeroed */
bd3576d2 UM	313	}
	314	else
	315	{
c69ed6ea AP	316	memcpy (p+n,data,len);
c69ed6ea AP	317	c->num += (unsigned int)len;
2dc769a1	318	return 1;
bd3576d2 UM	319	}
	320	}
	321
c69ed6ea AP	322	n = len/HASH_CBLOCK;
c69ed6ea AP	323	if (n > 0)
bd3576d2	324	{
c69ed6ea AP	325	HASH_BLOCK_DATA_ORDER (c,data,n);
	326	n *= HASH_CBLOCK;
	327	data += n;
	328	len -= n;
bd3576d2 UM	329	}
bd3576d2 UM	330
c69ed6ea	331	if (len != 0)
bd3576d2	332	{
c69ed6ea	333	p = (unsigned char *)c->data;
5cabcf96	334	c->num = (unsigned int)len;
c69ed6ea	335	memcpy (p,data,len);
bd3576d2	336	}
2dc769a1	337	return 1;
bd3576d2 UM	338	}
	339
	340
ac7d0785	341	void HASH_TRANSFORM (HASH_CTX c, const unsigned char data)
bd3576d2	342	{
cdfb093f	343	HASH_BLOCK_DATA_ORDER (c,data,1);
bd3576d2 UM	344	}
	345
	346
2dc769a1	347	int HASH_FINAL (unsigned char md, HASH_CTX c)
bd3576d2	348	{
c69ed6ea AP	349	unsigned char p = (unsigned char )c->data;
c69ed6ea AP	350	size_t n = c->num;
bd3576d2	351
c69ed6ea AP	352	p[n] = 0x80; /* there is always room for one */
	353	n++;
	354
	355	if (n > (HASH_CBLOCK-8))
bd3576d2	356	{
20f7563f AP	357	memset (p+n,0,HASH_CBLOCK-n);
20f7563f AP	358	n=0;
c69ed6ea	359	HASH_BLOCK_DATA_ORDER (c,p,1);
bd3576d2	360	}
20f7563f	361	memset (p+n,0,HASH_CBLOCK-8-n);
bd3576d2	362
c69ed6ea	363	p += HASH_CBLOCK-8;
bd3576d2	364	#if defined(DATA_ORDER_IS_BIG_ENDIAN)
c69ed6ea AP	365	(void)HOST_l2c(c->Nh,p);
c69ed6ea AP	366	(void)HOST_l2c(c->Nl,p);
bd3576d2	367	#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
c69ed6ea AP	368	(void)HOST_l2c(c->Nl,p);
c69ed6ea AP	369	(void)HOST_l2c(c->Nh,p);
bd3576d2	370	#endif
c69ed6ea AP	371	p -= HASH_CBLOCK;
	372	HASH_BLOCK_DATA_ORDER (c,p,1);
	373	c->num=0;
	374	memset (p,0,HASH_CBLOCK);
bd3576d2	375
1cbde6e4 AP	376	#ifndef HASH_MAKE_STRING
	377	#error "HASH_MAKE_STRING must be defined!"
	378	#else
	379	HASH_MAKE_STRING(c,md);
	380	#endif
bd3576d2	381
2dc769a1	382	return 1;
bd3576d2	383	}
2f98abbc AP	384
	385	#ifndef MD32_REG_T
	386	#define MD32_REG_T long
	387	/*
	388	* This comment was originaly written for MD5, which is why it
	389	* discusses A-D. But it basically applies to all 32-bit digests,
	390	* which is why it was moved to common header file.
	391	*
	392	* In case you wonder why A-D are declared as long and not
	393	* as MD5_LONG. Doing so results in slight performance
	394	* boost on LP64 architectures. The catch is we don't
	395	* really care if 32 MSBs of a 64-bit register get polluted
	396	* with eventual overflows as we save only 32 LSBs in
	397	* either case. Now declaring 'em long excuses the compiler
	398	* from keeping 32 MSBs zeroed resulting in 13% performance
	399	* improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
	400	* Well, to be honest it should say that this prevents
	401	* performance degradation.
	402	* <appro@fy.chalmers.se>
	403	* Apparently there're LP64 compilers that generate better
	404	* code if A-D are declared int. Most notably GCC-x86_64
	405	* generates better code.
	406	* <appro@fy.chalmers.se>
	407	*/
	408	#endif