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