1/* crypto/md32_common.h */
2/* ====================================================================
3 * Copyright (c) 1999-2007 The OpenSSL Project.  All rights reserved.
4 *
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
7 * are met:
8 *
9 * 1. Redistributions of source code must retain the above copyright
10 *    notice, this list of conditions and the following disclaimer.
11 *
12 * 2. Redistributions in binary form must reproduce the above copyright
13 *    notice, this list of conditions and the following disclaimer in
14 *    the documentation and/or other materials provided with the
15 *    distribution.
16 *
17 * 3. All advertising materials mentioning features or use of this
18 *    software must display the following acknowledgment:
19 *    "This product includes software developed by the OpenSSL Project
20 *    for use in the OpenSSL Toolkit. (http://www.OpenSSL.org/)"
21 *
22 * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
23 *    endorse or promote products derived from this software without
24 *    prior written permission. For written permission, please contact
25 *    licensing@OpenSSL.org.
26 *
27 * 5. Products derived from this software may not be called "OpenSSL"
28 *    nor may "OpenSSL" appear in their names without prior written
29 *    permission of the OpenSSL Project.
30 *
31 * 6. Redistributions of any form whatsoever must retain the following
32 *    acknowledgment:
33 *    "This product includes software developed by the OpenSSL Project
34 *    for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
35 *
36 * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
37 * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
38 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
39 * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
40 * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
41 * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
42 * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
43 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
44 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
45 * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
46 * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
47 * OF THE POSSIBILITY OF SUCH DAMAGE.
48 * ====================================================================
49 *
50 */
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
67 *	has to be at lest 32 bit wide, if it's wider, then
68 *	HASH_LONG_LOG2 *has to* be defined along
69 * HASH_CTX
70 *	context structure that at least contains following
71 *	members:
72 *		typedef struct {
73 *			...
74 *			HASH_LONG	Nl,Nh;
75 *			either {
76 *			HASH_LONG	data[HASH_LBLOCK];
77 *			unsigned char	data[HASH_CBLOCK];
78 *			};
79 *			unsigned int	num;
80 *			...
81 *			} HASH_CTX;
82 *	data[] vector is expected to be zeroed upon first call to
83 *	HASH_UPDATE.
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.
90 * HASH_BLOCK_DATA_ORDER
91 *	name of "block" function capable of treating *unaligned* input
92 *	message in original (data) byte order, implemented externally.
93 * HASH_MAKE_STRING
94 *	macro convering context variables to an ASCII hash string.
95 *
96 * MD5 example:
97 *
98 *	#define DATA_ORDER_IS_LITTLE_ENDIAN
99 *
100 *	#define HASH_LONG		MD5_LONG
101 *	#define HASH_LONG_LOG2		MD5_LONG_LOG2
102 *	#define HASH_CTX		MD5_CTX
103 *	#define HASH_CBLOCK		MD5_CBLOCK
104 *	#define HASH_UPDATE		MD5_Update
105 *	#define HASH_TRANSFORM		MD5_Transform
106 *	#define HASH_FINAL		MD5_Final
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
136#ifndef HASH_BLOCK_DATA_ORDER
137#error "HASH_BLOCK_DATA_ORDER must be defined!"
138#endif
139
140/*
141 * Engage compiler specific rotate intrinsic function if available.
142 */
143#undef ROTATE
144#ifndef PEDANTIC
145# if defined(_MSC_VER) || defined(__ICC)
146#  define ROTATE(a,n)	_lrotl(a,n)
147# elif defined(__MWERKS__)
148#  if defined(__POWERPC__)
149#   define ROTATE(a,n)	__rlwinm(a,n,0,31)
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) )
153#  else
154#   define ROTATE(a,n)	__rol(a,n)
155#  endif
156# elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
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...
161   * 					<appro@fy.chalmers.se>
162   */
163#  if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
164#   define ROTATE(a,n)	({ register unsigned int ret;	\
165				asm (			\
166				"roll %1,%0"		\
167				: "=r"(ret)		\
168				: "I"(n), "0"((unsigned int)(a))	\
169				: "cc");		\
170			   ret;				\
171			})
172#  elif defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
173	defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
174#   define ROTATE(a,n)	({ register unsigned int ret;	\
175				asm (			\
176				"rlwinm %0,%1,%2,0,31"	\
177				: "=r"(ret)		\
178				: "r"(a), "I"(n));	\
179			   ret;				\
180			})
181#  elif defined(__s390x__)
182#   define ROTATE(a,n) ({ register unsigned int ret;	\
183				asm ("rll %0,%1,%2"	\
184				: "=r"(ret)		\
185				: "r"(a), "I"(n));	\
186			  ret;				\
187			})
188#  endif
189# endif
190#endif /* PEDANTIC */
191
192#ifndef ROTATE
193#define ROTATE(a,n)     (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
194#endif
195
196#if defined(DATA_ORDER_IS_BIG_ENDIAN)
197
198#ifndef PEDANTIC
199# if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
200#  if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
201      (defined(__x86_64) || defined(__x86_64__))
202#   if !defined(B_ENDIAN)
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     */
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;			})
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;	})
215#   endif
216#  endif
217# endif
218#endif
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
223
224#ifndef HOST_c2l
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)
230#endif
231#ifndef HOST_l2c
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)
237#endif
238
239#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
240
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,%1"			\
245				   :"=d"(l) :"m"(*(const unsigned int *)(c)));\
246				   (c)+=4; (l);				})
247#   define HOST_l2c(l,c)	({ asm ("strv	%1,%0"			\
248				   :"=m"(*(unsigned int *)(c)) :"d"(l));\
249				   (c)+=4; (l);				})
250#  endif
251# endif
252#endif
253#if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
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
259#endif
260
261#ifndef HOST_c2l
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)
267#endif
268#ifndef HOST_l2c
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)
274#endif
275
276#endif
277
278/*
279 * Time for some action:-)
280 */
281
282int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
283	{
284	const unsigned char *data=data_;
285	unsigned char *p;
286	HASH_LONG l;
287	size_t n;
288
289	if (len==0) return 1;
290
291	l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
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++;
296	c->Nh+=(HASH_LONG)(len>>29);	/* might cause compiler warning on 16-bit */
297	c->Nl=l;
298
299	n = c->num;
300	if (n != 0)
301		{
302		p=(unsigned char *)c->data;
303
304		if (len >= HASH_CBLOCK || len+n >= HASH_CBLOCK)
305			{
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 */
313			}
314		else
315			{
316			memcpy (p+n,data,len);
317			c->num += (unsigned int)len;
318			return 1;
319			}
320		}
321
322	n = len/HASH_CBLOCK;
323	if (n > 0)
324		{
325		HASH_BLOCK_DATA_ORDER (c,data,n);
326		n    *= HASH_CBLOCK;
327		data += n;
328		len  -= n;
329		}
330
331	if (len != 0)
332		{
333		p = (unsigned char *)c->data;
334		c->num = (unsigned int)len;
335		memcpy (p,data,len);
336		}
337	return 1;
338	}
339
340
341void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
342	{
343	HASH_BLOCK_DATA_ORDER (c,data,1);
344	}
345
346
347int HASH_FINAL (unsigned char *md, HASH_CTX *c)
348	{
349	unsigned char *p = (unsigned char *)c->data;
350	size_t n = c->num;
351
352	p[n] = 0x80; /* there is always room for one */
353	n++;
354
355	if (n > (HASH_CBLOCK-8))
356		{
357		memset (p+n,0,HASH_CBLOCK-n);
358		n=0;
359		HASH_BLOCK_DATA_ORDER (c,p,1);
360		}
361	memset (p+n,0,HASH_CBLOCK-8-n);
362
363	p += HASH_CBLOCK-8;
364#if   defined(DATA_ORDER_IS_BIG_ENDIAN)
365	(void)HOST_l2c(c->Nh,p);
366	(void)HOST_l2c(c->Nl,p);
367#elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
368	(void)HOST_l2c(c->Nl,p);
369	(void)HOST_l2c(c->Nh,p);
370#endif
371	p -= HASH_CBLOCK;
372	HASH_BLOCK_DATA_ORDER (c,p,1);
373	c->num=0;
374	memset (p,0,HASH_CBLOCK);
375
376#ifndef HASH_MAKE_STRING
377#error "HASH_MAKE_STRING must be defined!"
378#else
379	HASH_MAKE_STRING(c,md);
380#endif
381
382	return 1;
383	}
384
385#ifndef MD32_REG_T
386#if defined(__alpha) || defined(__sparcv9) || defined(__mips)
387#define MD32_REG_T long
388/*
389 * This comment was originaly written for MD5, which is why it
390 * discusses A-D. But it basically applies to all 32-bit digests,
391 * which is why it was moved to common header file.
392 *
393 * In case you wonder why A-D are declared as long and not
394 * as MD5_LONG. Doing so results in slight performance
395 * boost on LP64 architectures. The catch is we don't
396 * really care if 32 MSBs of a 64-bit register get polluted
397 * with eventual overflows as we *save* only 32 LSBs in
398 * *either* case. Now declaring 'em long excuses the compiler
399 * from keeping 32 MSBs zeroed resulting in 13% performance
400 * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
401 * Well, to be honest it should say that this *prevents*
402 * performance degradation.
403 *				<appro@fy.chalmers.se>
404 */
405#else
406/*
407 * Above is not absolute and there are LP64 compilers that
408 * generate better code if MD32_REG_T is defined int. The above
409 * pre-processor condition reflects the circumstances under which
410 * the conclusion was made and is subject to further extension.
411 *				<appro@fy.chalmers.se>
412 */
413#define MD32_REG_T int
414#endif
415#endif
416