md4_locl.h revision 68651
168651Skris/* crypto/md4/md4_locl.h */
268651Skris/* Copyright (C) 1995-1998 Eric Young (eay@cryptsoft.com)
368651Skris * All rights reserved.
468651Skris *
568651Skris * This package is an SSL implementation written
668651Skris * by Eric Young (eay@cryptsoft.com).
768651Skris * The implementation was written so as to conform with Netscapes SSL.
868651Skris *
968651Skris * This library is free for commercial and non-commercial use as long as
1068651Skris * the following conditions are aheared to.  The following conditions
1168651Skris * apply to all code found in this distribution, be it the RC4, RSA,
1268651Skris * lhash, DES, etc., code; not just the SSL code.  The SSL documentation
1368651Skris * included with this distribution is covered by the same copyright terms
1468651Skris * except that the holder is Tim Hudson (tjh@cryptsoft.com).
1568651Skris *
1668651Skris * Copyright remains Eric Young's, and as such any Copyright notices in
1768651Skris * the code are not to be removed.
1868651Skris * If this package is used in a product, Eric Young should be given attribution
1968651Skris * as the author of the parts of the library used.
2068651Skris * This can be in the form of a textual message at program startup or
2168651Skris * in documentation (online or textual) provided with the package.
2268651Skris *
2368651Skris * Redistribution and use in source and binary forms, with or without
2468651Skris * modification, are permitted provided that the following conditions
2568651Skris * are met:
2668651Skris * 1. Redistributions of source code must retain the copyright
2768651Skris *    notice, this list of conditions and the following disclaimer.
2868651Skris * 2. Redistributions in binary form must reproduce the above copyright
2968651Skris *    notice, this list of conditions and the following disclaimer in the
3068651Skris *    documentation and/or other materials provided with the distribution.
3168651Skris * 3. All advertising materials mentioning features or use of this software
3268651Skris *    must display the following acknowledgement:
3368651Skris *    "This product includes cryptographic software written by
3468651Skris *     Eric Young (eay@cryptsoft.com)"
3568651Skris *    The word 'cryptographic' can be left out if the rouines from the library
3668651Skris *    being used are not cryptographic related :-).
3768651Skris * 4. If you include any Windows specific code (or a derivative thereof) from
3868651Skris *    the apps directory (application code) you must include an acknowledgement:
3968651Skris *    "This product includes software written by Tim Hudson (tjh@cryptsoft.com)"
4068651Skris *
4168651Skris * THIS SOFTWARE IS PROVIDED BY ERIC YOUNG ``AS IS'' AND
4268651Skris * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
4368651Skris * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
4468651Skris * ARE DISCLAIMED.  IN NO EVENT SHALL THE AUTHOR OR CONTRIBUTORS BE LIABLE
4568651Skris * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
4668651Skris * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
4768651Skris * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
4868651Skris * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
4968651Skris * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
5068651Skris * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
5168651Skris * SUCH DAMAGE.
5268651Skris *
5368651Skris * The licence and distribution terms for any publically available version or
5468651Skris * derivative of this code cannot be changed.  i.e. this code cannot simply be
5568651Skris * copied and put under another distribution licence
5668651Skris * [including the GNU Public Licence.]
5768651Skris */
5868651Skris
5968651Skris#include <stdlib.h>
6068651Skris#include <string.h>
6168651Skris#include <openssl/opensslconf.h>
6268651Skris#include <openssl/md4.h>
6368651Skris
6468651Skris#ifndef MD4_LONG_LOG2
6568651Skris#define MD4_LONG_LOG2 2 /* default to 32 bits */
6668651Skris#endif
6768651Skris
6868651Skrisvoid md4_block_host_order (MD4_CTX *c, const void *p,int num);
6968651Skrisvoid md4_block_data_order (MD4_CTX *c, const void *p,int num);
7068651Skris
7168651Skris#if defined(__i386) || defined(_M_IX86) || defined(__INTEL__)
7268651Skris/*
7368651Skris * *_block_host_order is expected to handle aligned data while
7468651Skris * *_block_data_order - unaligned. As algorithm and host (x86)
7568651Skris * are in this case of the same "endianness" these two are
7668651Skris * otherwise indistinguishable. But normally you don't want to
7768651Skris * call the same function because unaligned access in places
7868651Skris * where alignment is expected is usually a "Bad Thing". Indeed,
7968651Skris * on RISCs you get punished with BUS ERROR signal or *severe*
8068651Skris * performance degradation. Intel CPUs are in turn perfectly
8168651Skris * capable of loading unaligned data without such drastic side
8268651Skris * effect. Yes, they say it's slower than aligned load, but no
8368651Skris * exception is generated and therefore performance degradation
8468651Skris * is *incomparable* with RISCs. What we should weight here is
8568651Skris * costs of unaligned access against costs of aligning data.
8668651Skris * According to my measurements allowing unaligned access results
8768651Skris * in ~9% performance improvement on Pentium II operating at
8868651Skris * 266MHz. I won't be surprised if the difference will be higher
8968651Skris * on faster systems:-)
9068651Skris *
9168651Skris *				<appro@fy.chalmers.se>
9268651Skris */
9368651Skris#define md4_block_data_order md4_block_host_order
9468651Skris#endif
9568651Skris
9668651Skris#define DATA_ORDER_IS_LITTLE_ENDIAN
9768651Skris
9868651Skris#define HASH_LONG		MD4_LONG
9968651Skris#define HASH_LONG_LOG2		MD4_LONG_LOG2
10068651Skris#define HASH_CTX		MD4_CTX
10168651Skris#define HASH_CBLOCK		MD4_CBLOCK
10268651Skris#define HASH_LBLOCK		MD4_LBLOCK
10368651Skris#define HASH_UPDATE		MD4_Update
10468651Skris#define HASH_TRANSFORM		MD4_Transform
10568651Skris#define HASH_FINAL		MD4_Final
10668651Skris#define	HASH_MAKE_STRING(c,s)	do {	\
10768651Skris	unsigned long ll;		\
10868651Skris	ll=(c)->A; HOST_l2c(ll,(s));	\
10968651Skris	ll=(c)->B; HOST_l2c(ll,(s));	\
11068651Skris	ll=(c)->C; HOST_l2c(ll,(s));	\
11168651Skris	ll=(c)->D; HOST_l2c(ll,(s));	\
11268651Skris	} while (0)
11368651Skris#define HASH_BLOCK_HOST_ORDER	md4_block_host_order
11468651Skris#if !defined(L_ENDIAN) || defined(md4_block_data_order)
11568651Skris#define	HASH_BLOCK_DATA_ORDER	md4_block_data_order
11668651Skris/*
11768651Skris * Little-endians (Intel and Alpha) feel better without this.
11868651Skris * It looks like memcpy does better job than generic
11968651Skris * md4_block_data_order on copying-n-aligning input data.
12068651Skris * But frankly speaking I didn't expect such result on Alpha.
12168651Skris * On the other hand I've got this with egcs-1.0.2 and if
12268651Skris * program is compiled with another (better?) compiler it
12368651Skris * might turn out other way around.
12468651Skris *
12568651Skris *				<appro@fy.chalmers.se>
12668651Skris */
12768651Skris#endif
12868651Skris
12968651Skris#include "md32_common.h"
13068651Skris
13168651Skris/*
13268651Skris#define	F(x,y,z)	(((x) & (y))  |  ((~(x)) & (z)))
13368651Skris#define	G(x,y,z)	(((x) & (y))  |  ((x) & ((z))) | ((y) & ((z))))
13468651Skris*/
13568651Skris
13668651Skris/* As pointed out by Wei Dai <weidai@eskimo.com>, the above can be
13768651Skris * simplified to the code below.  Wei attributes these optimizations
13868651Skris * to Peter Gutmann's SHS code, and he attributes it to Rich Schroeppel.
13968651Skris */
14068651Skris#define	F(b,c,d)	((((c) ^ (d)) & (b)) ^ (d))
14168651Skris#define G(b,c,d)	(((b) & (c)) | ((b) & (d)) | ((c) & (d)))
14268651Skris#define	H(b,c,d)	((b) ^ (c) ^ (d))
14368651Skris
14468651Skris#define R0(a,b,c,d,k,s,t) { \
14568651Skris	a+=((k)+(t)+F((b),(c),(d))); \
14668651Skris	a=ROTATE(a,s); };
14768651Skris
14868651Skris#define R1(a,b,c,d,k,s,t) { \
14968651Skris	a+=((k)+(t)+G((b),(c),(d))); \
15068651Skris	a=ROTATE(a,s); };\
15168651Skris
15268651Skris#define R2(a,b,c,d,k,s,t) { \
15368651Skris	a+=((k)+(t)+H((b),(c),(d))); \
15468651Skris	a=ROTATE(a,s); };
155