2
0

md32_common.h 12 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380381382383384385386387388389390391392393394395396397398399400401402403404405406407408409410411412413414415
  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. * This is a generic 32 bit "collector" for message digest algorithms.
  53. * Whenever needed it collects input character stream into chunks of
  54. * 32 bit values and invokes a block function that performs actual hash
  55. * calculations.
  56. *
  57. * Porting guide.
  58. *
  59. * Obligatory macros:
  60. *
  61. * DATA_ORDER_IS_BIG_ENDIAN or DATA_ORDER_IS_LITTLE_ENDIAN
  62. * this macro defines byte order of input stream.
  63. * HASH_CBLOCK
  64. * size of a unit chunk HASH_BLOCK operates on.
  65. * HASH_LONG
  66. * has to be at lest 32 bit wide, if it's wider, then
  67. * HASH_LONG_LOG2 *has to* be defined along
  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_LONG_LOG2 MD5_LONG_LOG2
  101. * #define HASH_CTX MD5_CTX
  102. * #define HASH_CBLOCK MD5_CBLOCK
  103. * #define HASH_UPDATE MD5_Update
  104. * #define HASH_TRANSFORM MD5_Transform
  105. * #define HASH_FINAL MD5_Final
  106. * #define HASH_BLOCK_DATA_ORDER md5_block_data_order
  107. *
  108. * <appro@fy.chalmers.se>
  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. #ifndef HASH_CBLOCK
  114. #error "HASH_CBLOCK must be defined!"
  115. #endif
  116. #ifndef HASH_LONG
  117. #error "HASH_LONG must be defined!"
  118. #endif
  119. #ifndef HASH_CTX
  120. #error "HASH_CTX must be defined!"
  121. #endif
  122. #ifndef HASH_UPDATE
  123. #error "HASH_UPDATE must be defined!"
  124. #endif
  125. #ifndef HASH_TRANSFORM
  126. #error "HASH_TRANSFORM must be defined!"
  127. #endif
  128. #ifndef HASH_FINAL
  129. #error "HASH_FINAL must be defined!"
  130. #endif
  131. #ifndef HASH_BLOCK_DATA_ORDER
  132. #error "HASH_BLOCK_DATA_ORDER must be defined!"
  133. #endif
  134. /*
  135. * Engage compiler specific rotate intrinsic function if available.
  136. */
  137. #undef ROTATE
  138. #ifndef PEDANTIC
  139. # if defined(_MSC_VER) || defined(__ICC)
  140. # define ROTATE(a,n) _lrotl(a,n)
  141. # elif defined(__MWERKS__)
  142. # if defined(__POWERPC__)
  143. # define ROTATE(a,n) __rlwinm(a,n,0,31)
  144. # elif defined(__MC68K__)
  145. /* Motorola specific tweak. <appro@fy.chalmers.se> */
  146. # define ROTATE(a,n) ( n<24 ? __rol(a,n) : __ror(a,32-n) )
  147. # else
  148. # define ROTATE(a,n) __rol(a,n)
  149. # endif
  150. # elif defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
  151. /*
  152. * Some GNU C inline assembler templates. Note that these are
  153. * rotates by *constant* number of bits! But that's exactly
  154. * what we need here...
  155. * <appro@fy.chalmers.se>
  156. */
  157. # if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
  158. # define ROTATE(a,n) ({ register unsigned int ret; \
  159. asm ( \
  160. "roll %1,%0" \
  161. : "=r"(ret) \
  162. : "I"(n), "0"((unsigned int)(a)) \
  163. : "cc"); \
  164. ret; \
  165. })
  166. # elif defined(_ARCH_PPC) || defined(_ARCH_PPC64) || \
  167. defined(__powerpc) || defined(__ppc__) || defined(__powerpc64__)
  168. # define ROTATE(a,n) ({ register unsigned int ret; \
  169. asm ( \
  170. "rlwinm %0,%1,%2,0,31" \
  171. : "=r"(ret) \
  172. : "r"(a), "I"(n)); \
  173. ret; \
  174. })
  175. # elif defined(__s390x__)
  176. # define ROTATE(a,n) ({ register unsigned int ret; \
  177. asm ("rll %0,%1,%2" \
  178. : "=r"(ret) \
  179. : "r"(a), "I"(n)); \
  180. ret; \
  181. })
  182. # endif
  183. # endif
  184. #endif /* PEDANTIC */
  185. #ifndef ROTATE
  186. #define ROTATE(a,n) (((a)<<(n))|(((a)&0xffffffff)>>(32-(n))))
  187. #endif
  188. #if defined(DATA_ORDER_IS_BIG_ENDIAN)
  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. # endif
  208. # endif
  209. #endif
  210. #if defined(__s390__) || defined(__s390x__)
  211. # define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, (l))
  212. # define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, (l))
  213. #endif
  214. #ifndef HOST_c2l
  215. #define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++)))<<24), \
  216. l|=(((unsigned long)(*((c)++)))<<16), \
  217. l|=(((unsigned long)(*((c)++)))<< 8), \
  218. l|=(((unsigned long)(*((c)++))) ), \
  219. l)
  220. #endif
  221. #ifndef HOST_l2c
  222. #define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l)>>24)&0xff), \
  223. *((c)++)=(unsigned char)(((l)>>16)&0xff), \
  224. *((c)++)=(unsigned char)(((l)>> 8)&0xff), \
  225. *((c)++)=(unsigned char)(((l) )&0xff), \
  226. l)
  227. #endif
  228. #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
  229. #ifndef PEDANTIC
  230. # if defined(__GNUC__) && __GNUC__>=2 && !defined(OPENSSL_NO_ASM) && !defined(OPENSSL_NO_INLINE_ASM)
  231. # if defined(__s390x__)
  232. # define HOST_c2l(c,l) ({ asm ("lrv %0,%1" \
  233. :"=d"(l) :"m"(*(const unsigned int *)(c)));\
  234. (c)+=4; (l); })
  235. # define HOST_l2c(l,c) ({ asm ("strv %1,%0" \
  236. :"=m"(*(unsigned int *)(c)) :"d"(l));\
  237. (c)+=4; (l); })
  238. # endif
  239. # endif
  240. #endif
  241. #if defined(__i386) || defined(__i386__) || defined(__x86_64) || defined(__x86_64__)
  242. # ifndef B_ENDIAN
  243. /* See comment in DATA_ORDER_IS_BIG_ENDIAN section. */
  244. # define HOST_c2l(c,l) ((l)=*((const unsigned int *)(c)), (c)+=4, l)
  245. # define HOST_l2c(l,c) (*((unsigned int *)(c))=(l), (c)+=4, l)
  246. # endif
  247. #endif
  248. #ifndef HOST_c2l
  249. #define HOST_c2l(c,l) (l =(((unsigned long)(*((c)++))) ), \
  250. l|=(((unsigned long)(*((c)++)))<< 8), \
  251. l|=(((unsigned long)(*((c)++)))<<16), \
  252. l|=(((unsigned long)(*((c)++)))<<24), \
  253. l)
  254. #endif
  255. #ifndef HOST_l2c
  256. #define HOST_l2c(l,c) (*((c)++)=(unsigned char)(((l) )&0xff), \
  257. *((c)++)=(unsigned char)(((l)>> 8)&0xff), \
  258. *((c)++)=(unsigned char)(((l)>>16)&0xff), \
  259. *((c)++)=(unsigned char)(((l)>>24)&0xff), \
  260. l)
  261. #endif
  262. #endif
  263. /*
  264. * Time for some action:-)
  265. */
  266. int HASH_UPDATE (HASH_CTX *c, const void *data_, size_t len)
  267. {
  268. const unsigned char *data=data_;
  269. unsigned char *p;
  270. HASH_LONG l;
  271. size_t n;
  272. if (len==0) return 1;
  273. l=(c->Nl+(((HASH_LONG)len)<<3))&0xffffffffUL;
  274. /* 95-05-24 eay Fixed a bug with the overflow handling, thanks to
  275. * Wei Dai <weidai@eskimo.com> for pointing it out. */
  276. if (l < c->Nl) /* overflow */
  277. c->Nh++;
  278. c->Nh+=(HASH_LONG)(len>>29); /* might cause compiler warning on 16-bit */
  279. c->Nl=l;
  280. n = c->num;
  281. if (n != 0)
  282. {
  283. p=(unsigned char *)c->data;
  284. if (len >= HASH_CBLOCK || len+n >= HASH_CBLOCK)
  285. {
  286. memcpy (p+n,data,HASH_CBLOCK-n);
  287. HASH_BLOCK_DATA_ORDER (c,p,1);
  288. n = HASH_CBLOCK-n;
  289. data += n;
  290. len -= n;
  291. c->num = 0;
  292. memset (p,0,HASH_CBLOCK); /* keep it zeroed */
  293. }
  294. else
  295. {
  296. memcpy (p+n,data,len);
  297. c->num += (unsigned int)len;
  298. return 1;
  299. }
  300. }
  301. n = len/HASH_CBLOCK;
  302. if (n > 0)
  303. {
  304. HASH_BLOCK_DATA_ORDER (c,data,n);
  305. n *= HASH_CBLOCK;
  306. data += n;
  307. len -= n;
  308. }
  309. if (len != 0)
  310. {
  311. p = (unsigned char *)c->data;
  312. c->num = (unsigned int)len;
  313. memcpy (p,data,len);
  314. }
  315. return 1;
  316. }
  317. void HASH_TRANSFORM (HASH_CTX *c, const unsigned char *data)
  318. {
  319. HASH_BLOCK_DATA_ORDER (c,data,1);
  320. }
  321. int HASH_FINAL (unsigned char *md, HASH_CTX *c)
  322. {
  323. unsigned char *p = (unsigned char *)c->data;
  324. size_t n = c->num;
  325. p[n] = 0x80; /* there is always room for one */
  326. n++;
  327. if (n > (HASH_CBLOCK-8))
  328. {
  329. memset (p+n,0,HASH_CBLOCK-n);
  330. n=0;
  331. HASH_BLOCK_DATA_ORDER (c,p,1);
  332. }
  333. memset (p+n,0,HASH_CBLOCK-8-n);
  334. p += HASH_CBLOCK-8;
  335. #if defined(DATA_ORDER_IS_BIG_ENDIAN)
  336. (void)HOST_l2c(c->Nh,p);
  337. (void)HOST_l2c(c->Nl,p);
  338. #elif defined(DATA_ORDER_IS_LITTLE_ENDIAN)
  339. (void)HOST_l2c(c->Nl,p);
  340. (void)HOST_l2c(c->Nh,p);
  341. #endif
  342. p -= HASH_CBLOCK;
  343. HASH_BLOCK_DATA_ORDER (c,p,1);
  344. c->num=0;
  345. memset (p,0,HASH_CBLOCK);
  346. #ifndef HASH_MAKE_STRING
  347. #error "HASH_MAKE_STRING must be defined!"
  348. #else
  349. HASH_MAKE_STRING(c,md);
  350. #endif
  351. return 1;
  352. }
  353. #ifndef MD32_REG_T
  354. #if defined(__alpha) || defined(__sparcv9) || defined(__mips)
  355. #define MD32_REG_T long
  356. /*
  357. * This comment was originaly written for MD5, which is why it
  358. * discusses A-D. But it basically applies to all 32-bit digests,
  359. * which is why it was moved to common header file.
  360. *
  361. * In case you wonder why A-D are declared as long and not
  362. * as MD5_LONG. Doing so results in slight performance
  363. * boost on LP64 architectures. The catch is we don't
  364. * really care if 32 MSBs of a 64-bit register get polluted
  365. * with eventual overflows as we *save* only 32 LSBs in
  366. * *either* case. Now declaring 'em long excuses the compiler
  367. * from keeping 32 MSBs zeroed resulting in 13% performance
  368. * improvement under SPARC Solaris7/64 and 5% under AlphaLinux.
  369. * Well, to be honest it should say that this *prevents*
  370. * performance degradation.
  371. * <appro@fy.chalmers.se>
  372. */
  373. #else
  374. /*
  375. * Above is not absolute and there are LP64 compilers that
  376. * generate better code if MD32_REG_T is defined int. The above
  377. * pre-processor condition reflects the circumstances under which
  378. * the conclusion was made and is subject to further extension.
  379. * <appro@fy.chalmers.se>
  380. */
  381. #define MD32_REG_T int
  382. #endif
  383. #endif