1 /* crypto/md32_common.h */
2 /* ====================================================================
3 * Copyright (c) 1999-2006 The OpenSSL Project. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
9 * 1. Redistributions of source code must retain the above copyright
10 * notice, this list of conditions and the following disclaimer.
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
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/)"
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.
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.
31 * 6. Redistributions of any form whatsoever must retain the following
33 * "This product includes software developed by the OpenSSL Project
34 * for use in the OpenSSL Toolkit (http://www.OpenSSL.org/)"
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 * ====================================================================
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
62 * DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
63 * this macro defines byte order of input stream.
65 * size of a unit chunk HASH_BLOCK operates on.
67 * has to be at lest 32 bit wide.
69 * context structure that at least contains following
75 * HASH_LONG data[HASH_LBLOCK];
76 * unsigned char data[HASH_CBLOCK];
81 * data[] vector is expected to be zeroed upon first call to
84 * name of "Update" function, implemented here.
86 * name of "Transform" function, implemented here.
88 * name of "Final" function, implemented here.
89 * HASH_BLOCK_DATA_ORDER
90 * name of "block" function capable of treating *unaligned* input
91 * message in original (data) byte order, implemented externally.
93 * macro convering context variables to an ASCII hash string.
97 * #define DATA_ORDER_IS_LITTLE_ENDIAN
99 * #define HASH_LONG MD5_LONG
100 * #define HASH_CTX MD5_CTX
101 * #define HASH_CBLOCK MD5_CBLOCK
102 * #define HASH_UPDATE MD5_Update
103 * #define HASH_TRANSFORM MD5_Transform
104 * #define HASH_FINAL MD5_Final
105 * #define HASH_BLOCK_DATA_ORDER md5_block_data_order
107 * <appro@fy.chalmers.se>
110 #if !defined(DATA_ORDER_IS_BIG_ENDIAN) && !defined(DATA_ORDER_IS_LITTLE_ENDIAN)
111 #error "DATA_ORDER must be defined!"
115 #error "HASH_CBLOCK must be defined!"
118 #error "HASH_LONG must be defined!"
121 #error "HASH_CTX must be defined!"
125 #error "HASH_UPDATE must be defined!"
127 #ifndef HASH_TRANSFORM
128 #error "HASH_TRANSFORM must be defined!"
131 #error "HASH_FINAL must be defined!"
134 #ifndef HASH_BLOCK_DATA_ORDER
135 #error "HASH_BLOCK_DATA_ORDER must be defined!"
139 * Engage compiler specific rotate intrinsic function if available.
143 # if defined(_MSC_VER) || defined(__ICC)
144 # define ROTATE(a,n) _lrotl(a,n)
145 # elif defined(__MWERKS__)
146 # if defined(__POWERPC__)
147 # define ROTATE(a,n) __rlwinm(a,n,0,31)
148 # elif defined(__MC68K__)
149 /* Motorola specific tweak. <appro@fy.chalmers.se> */
150 # define ROTATE(a,n) ( n<24 ? __rol(a,n) : __ror(a,32-n) )
152 # define ROTATE(a,n) __rol(a,n)
154 # elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
156 * Some GNU C inline assembler templates. Note that these are
157 * rotates by *constant* number of bits! But that's exactly
158 * what we need here...
159 * <appro@fy.chalmers.se>
161 # if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
162 # define ROTATE(a,n) ({ register unsigned int ret; \
170 # elif defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
171 defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
172 # define ROTATE(a,n) ({ register unsigned int ret; \
174 "rlwinm %0,%1,%2,0,31" \
179 # elif defined(__s390x__)
180 # define ROTATE(a,n) ({ register unsigned int ret; \
181 asm ("rll %0,%1,%2" \
188 #endif /* PEDANTIC */
191 #define ROTATE(a,n) (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
194 #if defined(DATA_ORDER_IS_BIG_ENDIAN)
197 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
198 # if ((defined(__i386) || defined(__i386__)) && !defined(I386_ONLY)) || \
199 (defined(__x86_64) || defined(__x86_64__))
200 # if !defined(B_ENDIAN)
202 * This gives ~30-40% performance improvement in SHA-256 compiled
203 * with gcc [on P4]. Well, first macro to be frank. We can pull
204 * this trick on x86* platforms only, because these CPUs can fetch
205 * unaligned data without raising an exception.
207 # define HOST_c2l(c,l) ({ unsigned int r=*((const unsigned int *)(c)); \
208 asm ("bswapl %0":"=r"(r):"0"(r)); \
210 # define HOST_l2c(l,c) ({ unsigned int r=(l); \
211 asm ("bswapl %0":"=r"(r):"0"(r)); \
212 *((unsigned int *)(c))=r; (c)+=4; r; })
217 #if defined(__s390__) || defined(__s390x__)
218 # define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, (l))
219 # define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, (l))
223 #define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++)))<<24), \
224 l|=(((unsigned long)(*((c)++)))<<16), \
225 l|=(((unsigned long)(*((c)++)))<< 8), \
226 l|=(((unsigned long)(*((c)++))) ), \
230 #define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l)>>24)&0xff), \
231 *((c)++)=(unsigned char)(((l)>>16)&0xff), \
232 *((c)++)=(unsigned char)(((l)>> 8)&0xff), \
233 *((c)++)=(unsigned char)(((l) )&0xff), \
237 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
240 # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
241 # if defined(__s390x__)
242 # define HOST_c2l(c,l) ({ asm ("lrv %0,0(%1)" \
243 :"=r"(l) : "r"(c)); \
245 # define HOST_l2c(l,c) ({ asm ("strv %0,0(%1)" \
246 : : "r"(l),"r"(c) : "memory"); \
251 #if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
253 /* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
254 # define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, l)
255 # define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, l)
260 #define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++))) ), \
261 l|=(((unsigned long)(*((c)++)))<< 8), \
262 l|=(((unsigned long)(*((c)++)))<<16), \
263 l|=(((unsigned long)(*((c)++)))<<24), \
267 #define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
268 *((c)++)=(unsigned char)(((l)>> 8)&0xff), \
269 *((c)++)=(unsigned char)(((l)>>16)&0xff), \
270 *((c)++)=(unsigned char)(((l)>>24)&0xff), \
277 * Time for some action:-)
280 int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
282 const unsigned char *data=data_;
287 if (len==0) return 1;
289 l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
290 /* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
291 * Wei Dai <weidai@eskimo.com> for pointing it out. */
292 if (l < c->Nl) /* overflow */
294 c->Nh+=(len>>29); /* might cause compiler warning on 16-bit */
300 p=(unsigned char *)c->data;
302 if ((n+len) >= HASH_CBLOCK)
304 memcpy (p+n,data,HASH_CBLOCK-n);
305 HASH_BLOCK_DATA_ORDER (c,p,1);
310 memset (p,0,HASH_CBLOCK); /* keep it zeroed */
314 memcpy (p+n,data,len);
315 c->num += (unsigned int)len;
323 HASH_BLOCK_DATA_ORDER (c,data,n);
331 p = (unsigned char *)c->data;
339 void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
341 HASH_BLOCK_DATA_ORDER (c,data,1);
345 int HASH_FINAL (unsigned char *md, HASH_CTX *c)
347 unsigned char *p = (unsigned char *)c->data;
350 p[n] = 0x80; /* there is always room for one */
353 if (n > (HASH_CBLOCK-8))
355 HASH_BLOCK_DATA_ORDER (c,p,1);
356 memset (p,0,HASH_CBLOCK);
360 #if defined(DATA_ORDER_IS_BIG_ENDIAN)
361 (void)HOST_l2c(c->Nh,p);
362 (void)HOST_l2c(c->Nl,p);
363 #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
364 (void)HOST_l2c(c->Nl,p);
365 (void)HOST_l2c(c->Nh,p);
368 HASH_BLOCK_DATA_ORDER (c,p,1);
370 memset (p,0,HASH_CBLOCK);
372 #ifndef HASH_MAKE_STRING
373 #error "HASH_MAKE_STRING must be defined!"
375 HASH_MAKE_STRING(c,md);
382 #define MD32_REG_T long
384 * This comment was originaly written for MD5, which is why it
385 * discusses A-D. But it basically applies to all 32-bit digests,
386 * which is why it was moved to common header file.
388 * In case you wonder why A-D are declared as long and not
389 * as MD5_LONG. Doing so results in slight performance
390 * boost on LP64 architectures. The catch is we don't
391 * really care if 32 MSBs of a 64-bit register get polluted
392 * with eventual overflows as we *save* only 32 LSBs in
393 * *either* case. Now declaring 'em long excuses the compiler
394 * from keeping 32 MSBs zeroed resulting in 13% performance
395 * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
396 * Well, to be honest it should say that this *prevents*
397 * performance degradation.
398 * <appro@fy.chalmers.se>
399 * Apparently there're LP64 compilers that generate better
400 * code if A-D are declared int. Most notably GCC-x86_64
401 * generates better code.
402 * <appro@fy.chalmers.se>