xref: /freebsd/sys/contrib/openzfs/module/icp/algs/edonr/edonr.c (revision 2c449a4c5a338d2dd39a1949524e57c06ce7eff6)
1 /*
2  * IDI,NTNU
3  *
4  * CDDL HEADER START
5  *
6  * The contents of this file are subject to the terms of the
7  * Common Development and Distribution License (the "License").
8  * You may not use this file except in compliance with the License.
9  *
10  * You can obtain a copy of the license at usr/src/OPENSOLARIS.LICENSE
11  * or http://opensource.org/licenses/CDDL-1.0.
12  * See the License for the specific language governing permissions
13  * and limitations under the License.
14  *
15  * When distributing Covered Code, include this CDDL HEADER in each
16  * file and include the License file at usr/src/OPENSOLARIS.LICENSE.
17  * If applicable, add the following below this CDDL HEADER, with the
18  * fields enclosed by brackets "[]" replaced with your own identifying
19  * information: Portions Copyright [yyyy] [name of copyright owner]
20  *
21  * CDDL HEADER END
22  *
23  * Copyright (C) 2009, 2010, Jorn Amundsen <jorn.amundsen@ntnu.no>
24  * Tweaked Edon-R implementation for SUPERCOP, based on NIST API.
25  *
26  * $Id: edonr.c 517 2013-02-17 20:34:39Z joern $
27  */
28 /*
29  * Portions copyright (c) 2013, Saso Kiselkov, All rights reserved
30  */
31 
32 /*
33  * Unlike sha2 or skein, we won't expose edonr via the Kernel Cryptographic
34  * Framework (KCF), because Edon-R is *NOT* suitable for general-purpose
35  * cryptographic use. Users of Edon-R must interface directly to this module.
36  */
37 
38 #include <sys/strings.h>
39 #include <sys/edonr.h>
40 #include <sys/debug.h>
41 
42 /* big endian support, provides no-op's if run on little endian hosts */
43 #include "edonr_byteorder.h"
44 
45 #define	hashState224(x)	((x)->pipe->p256)
46 #define	hashState256(x)	((x)->pipe->p256)
47 #define	hashState384(x)	((x)->pipe->p512)
48 #define	hashState512(x)	((x)->pipe->p512)
49 
50 /* shift and rotate shortcuts */
51 #define	shl(x, n)	((x) << n)
52 #define	shr(x, n)	((x) >> n)
53 
54 #define	rotl32(x, n)	(((x) << (n)) | ((x) >> (32 - (n))))
55 #define	rotr32(x, n)	(((x) >> (n)) | ((x) << (32 - (n))))
56 
57 #define	rotl64(x, n)	(((x) << (n)) | ((x) >> (64 - (n))))
58 #define	rotr64(x, n)	(((x) >> (n)) | ((x) << (64 - (n))))
59 
60 #if !defined(__C99_RESTRICT)
61 #define	restrict	/* restrict */
62 #endif
63 
64 #define	EDONR_VALID_HASHBITLEN(x) \
65 	((x) == 512 || (x) == 384 || (x) == 256 || (x) == 224)
66 
67 /* EdonR224 initial double chaining pipe */
68 static const uint32_t i224p2[16] = {
69 	0x00010203ul, 0x04050607ul, 0x08090a0bul, 0x0c0d0e0ful,
70 	0x10111213ul, 0x14151617ul, 0x18191a1bul, 0x1c1d1e1ful,
71 	0x20212223ul, 0x24252627ul, 0x28292a2bul, 0x2c2d2e2ful,
72 	0x30313233ul, 0x34353637ul, 0x38393a3bul, 0x3c3d3e3ful,
73 };
74 
75 /* EdonR256 initial double chaining pipe */
76 static const uint32_t i256p2[16] = {
77 	0x40414243ul, 0x44454647ul, 0x48494a4bul, 0x4c4d4e4ful,
78 	0x50515253ul, 0x54555657ul, 0x58595a5bul, 0x5c5d5e5ful,
79 	0x60616263ul, 0x64656667ul, 0x68696a6bul, 0x6c6d6e6ful,
80 	0x70717273ul, 0x74757677ul, 0x78797a7bul, 0x7c7d7e7ful,
81 };
82 
83 /* EdonR384 initial double chaining pipe */
84 static const uint64_t i384p2[16] = {
85 	0x0001020304050607ull, 0x08090a0b0c0d0e0full,
86 	0x1011121314151617ull, 0x18191a1b1c1d1e1full,
87 	0x2021222324252627ull, 0x28292a2b2c2d2e2full,
88 	0x3031323334353637ull, 0x38393a3b3c3d3e3full,
89 	0x4041424344454647ull, 0x48494a4b4c4d4e4full,
90 	0x5051525354555657ull, 0x58595a5b5c5d5e5full,
91 	0x6061626364656667ull, 0x68696a6b6c6d6e6full,
92 	0x7071727374757677ull, 0x78797a7b7c7d7e7full
93 };
94 
95 /* EdonR512 initial double chaining pipe */
96 static const uint64_t i512p2[16] = {
97 	0x8081828384858687ull, 0x88898a8b8c8d8e8full,
98 	0x9091929394959697ull, 0x98999a9b9c9d9e9full,
99 	0xa0a1a2a3a4a5a6a7ull, 0xa8a9aaabacadaeafull,
100 	0xb0b1b2b3b4b5b6b7ull, 0xb8b9babbbcbdbebfull,
101 	0xc0c1c2c3c4c5c6c7ull, 0xc8c9cacbcccdcecfull,
102 	0xd0d1d2d3d4d5d6d7ull, 0xd8d9dadbdcdddedfull,
103 	0xe0e1e2e3e4e5e6e7ull, 0xe8e9eaebecedeeefull,
104 	0xf0f1f2f3f4f5f6f7ull, 0xf8f9fafbfcfdfeffull
105 };
106 
107 /*
108  * First Latin Square
109  * 0   7   1   3   2   4   6   5
110  * 4   1   7   6   3   0   5   2
111  * 7   0   4   2   5   3   1   6
112  * 1   4   0   5   6   2   7   3
113  * 2   3   6   7   1   5   0   4
114  * 5   2   3   1   7   6   4   0
115  * 3   6   5   0   4   7   2   1
116  * 6   5   2   4   0   1   3   7
117  */
118 #define	LS1_256(c, x0, x1, x2, x3, x4, x5, x6, x7)			\
119 {									\
120 	uint32_t x04, x17, x23, x56, x07, x26;				\
121 	x04 = x0+x4, x17 = x1+x7, x07 = x04+x17;			\
122 	s0 = c + x07 + x2;						\
123 	s1 = rotl32(x07 + x3, 4);					\
124 	s2 = rotl32(x07 + x6, 8);					\
125 	x23 = x2 + x3;							\
126 	s5 = rotl32(x04 + x23 + x5, 22);				\
127 	x56 = x5 + x6;							\
128 	s6 = rotl32(x17 + x56 + x0, 24);				\
129 	x26 = x23+x56;							\
130 	s3 = rotl32(x26 + x7, 13);					\
131 	s4 = rotl32(x26 + x1, 17);					\
132 	s7 = rotl32(x26 + x4, 29);					\
133 }
134 
135 #define	LS1_512(c, x0, x1, x2, x3, x4, x5, x6, x7)			\
136 {									\
137 	uint64_t x04, x17, x23, x56, x07, x26;				\
138 	x04 = x0+x4, x17 = x1+x7, x07 = x04+x17;			\
139 	s0 = c + x07 + x2;						\
140 	s1 = rotl64(x07 + x3, 5);					\
141 	s2 = rotl64(x07 + x6, 15);					\
142 	x23 = x2 + x3;							\
143 	s5 = rotl64(x04 + x23 + x5, 40);				\
144 	x56 = x5 + x6;							\
145 	s6 = rotl64(x17 + x56 + x0, 50);				\
146 	x26 = x23+x56;							\
147 	s3 = rotl64(x26 + x7, 22);					\
148 	s4 = rotl64(x26 + x1, 31);					\
149 	s7 = rotl64(x26 + x4, 59);					\
150 }
151 
152 /*
153  * Second Orthogonal Latin Square
154  * 0   4   2   3   1   6   5   7
155  * 7   6   3   2   5   4   1   0
156  * 5   3   1   6   0   2   7   4
157  * 1   0   5   4   3   7   2   6
158  * 2   1   0   7   4   5   6   3
159  * 3   5   7   0   6   1   4   2
160  * 4   7   6   1   2   0   3   5
161  * 6   2   4   5   7   3   0   1
162  */
163 #define	LS2_256(c, y0, y1, y2, y3, y4, y5, y6, y7)			\
164 {									\
165 	uint32_t y01, y25, y34, y67, y04, y05, y27, y37;		\
166 	y01 = y0+y1, y25 = y2+y5, y05 = y01+y25;			\
167 	t0  = ~c + y05 + y7;						\
168 	t2 = rotl32(y05 + y3, 9);					\
169 	y34 = y3+y4, y04 = y01+y34;					\
170 	t1 = rotl32(y04 + y6, 5);					\
171 	t4 = rotl32(y04 + y5, 15);					\
172 	y67 = y6+y7, y37 = y34+y67;					\
173 	t3 = rotl32(y37 + y2, 11);					\
174 	t7 = rotl32(y37 + y0, 27);					\
175 	y27 = y25+y67;							\
176 	t5 = rotl32(y27 + y4, 20);					\
177 	t6 = rotl32(y27 + y1, 25);					\
178 }
179 
180 #define	LS2_512(c, y0, y1, y2, y3, y4, y5, y6, y7)			\
181 {									\
182 	uint64_t y01, y25, y34, y67, y04, y05, y27, y37;		\
183 	y01 = y0+y1, y25 = y2+y5, y05 = y01+y25;			\
184 	t0  = ~c + y05 + y7;						\
185 	t2 = rotl64(y05 + y3, 19);					\
186 	y34 = y3+y4, y04 = y01+y34;					\
187 	t1 = rotl64(y04 + y6, 10);					\
188 	t4 = rotl64(y04 + y5, 36);					\
189 	y67 = y6+y7, y37 = y34+y67;					\
190 	t3 = rotl64(y37 + y2, 29);					\
191 	t7 = rotl64(y37 + y0, 55);					\
192 	y27 = y25+y67;							\
193 	t5 = rotl64(y27 + y4, 44);					\
194 	t6 = rotl64(y27 + y1, 48);					\
195 }
196 
197 #define	quasi_exform256(r0, r1, r2, r3, r4, r5, r6, r7)			\
198 {									\
199 	uint32_t s04, s17, s23, s56, t01, t25, t34, t67;		\
200 	s04 = s0 ^ s4, t01 = t0 ^ t1;					\
201 	r0 = (s04 ^ s1) + (t01 ^ t5);					\
202 	t67 = t6 ^ t7;							\
203 	r1 = (s04 ^ s7) + (t2 ^ t67);					\
204 	s23 = s2 ^ s3;							\
205 	r7 = (s23 ^ s5) + (t4 ^ t67);					\
206 	t34 = t3 ^ t4;							\
207 	r3 = (s23 ^ s4) + (t0 ^ t34);					\
208 	s56 = s5 ^ s6;							\
209 	r5 = (s3 ^ s56) + (t34 ^ t6);					\
210 	t25 = t2 ^ t5;							\
211 	r6 = (s2 ^ s56) + (t25 ^ t7);					\
212 	s17 = s1 ^ s7;							\
213 	r4 = (s0 ^ s17) + (t1 ^ t25);					\
214 	r2 = (s17 ^ s6) + (t01 ^ t3);					\
215 }
216 
217 #define	quasi_exform512(r0, r1, r2, r3, r4, r5, r6, r7)			\
218 {									\
219 	uint64_t s04, s17, s23, s56, t01, t25, t34, t67;		\
220 	s04 = s0 ^ s4, t01 = t0 ^ t1;					\
221 	r0 = (s04 ^ s1) + (t01 ^ t5);					\
222 	t67 = t6 ^ t7;							\
223 	r1 = (s04 ^ s7) + (t2 ^ t67);					\
224 	s23 = s2 ^ s3;							\
225 	r7 = (s23 ^ s5) + (t4 ^ t67);					\
226 	t34 = t3 ^ t4;							\
227 	r3 = (s23 ^ s4) + (t0 ^ t34);					\
228 	s56 = s5 ^ s6;							\
229 	r5 = (s3 ^ s56) + (t34 ^ t6);					\
230 	t25 = t2 ^ t5;							\
231 	r6 = (s2 ^ s56) + (t25 ^ t7);					\
232 	s17 = s1 ^ s7;							\
233 	r4 = (s0 ^ s17) + (t1 ^ t25);					\
234 	r2 = (s17 ^ s6) + (t01 ^ t3);					\
235 }
236 
237 static size_t
238 Q256(size_t bitlen, const uint32_t *data, uint32_t *restrict p)
239 {
240 	size_t bl;
241 
242 	for (bl = bitlen; bl >= EdonR256_BLOCK_BITSIZE;
243 	    bl -= EdonR256_BLOCK_BITSIZE, data += 16) {
244 		uint32_t s0, s1, s2, s3, s4, s5, s6, s7, t0, t1, t2, t3, t4,
245 		    t5, t6, t7;
246 		uint32_t p0, p1, p2, p3, p4, p5, p6, p7, q0, q1, q2, q3, q4,
247 		    q5, q6, q7;
248 		const uint32_t defix = 0xaaaaaaaa;
249 #if defined(MACHINE_IS_BIG_ENDIAN)
250 		uint32_t swp0, swp1, swp2, swp3, swp4, swp5, swp6, swp7, swp8,
251 		    swp9, swp10, swp11, swp12, swp13, swp14, swp15;
252 #define	d(j)	swp ## j
253 #define	s32(j)	ld_swap32((uint32_t *)data + j, swp ## j)
254 #else
255 #define	d(j)	data[j]
256 #endif
257 
258 		/* First row of quasigroup e-transformations */
259 #if defined(MACHINE_IS_BIG_ENDIAN)
260 		s32(8);
261 		s32(9);
262 		s32(10);
263 		s32(11);
264 		s32(12);
265 		s32(13);
266 		s32(14);
267 		s32(15);
268 #endif
269 		LS1_256(defix, d(15), d(14), d(13), d(12), d(11), d(10), d(9),
270 		    d(8));
271 #if defined(MACHINE_IS_BIG_ENDIAN)
272 		s32(0);
273 		s32(1);
274 		s32(2);
275 		s32(3);
276 		s32(4);
277 		s32(5);
278 		s32(6);
279 		s32(7);
280 #undef s32
281 #endif
282 		LS2_256(defix, d(0), d(1), d(2), d(3), d(4), d(5), d(6), d(7));
283 		quasi_exform256(p0, p1, p2, p3, p4, p5, p6, p7);
284 
285 		LS1_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
286 		LS2_256(defix, d(8), d(9), d(10), d(11), d(12), d(13), d(14),
287 		    d(15));
288 		quasi_exform256(q0, q1, q2, q3, q4, q5, q6, q7);
289 
290 		/* Second row of quasigroup e-transformations */
291 		LS1_256(defix, p[8], p[9], p[10], p[11], p[12], p[13], p[14],
292 		    p[15]);
293 		LS2_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
294 		quasi_exform256(p0, p1, p2, p3, p4, p5, p6, p7);
295 
296 		LS1_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
297 		LS2_256(defix, q0, q1, q2, q3, q4, q5, q6, q7);
298 		quasi_exform256(q0, q1, q2, q3, q4, q5, q6, q7);
299 
300 		/* Third row of quasigroup e-transformations */
301 		LS1_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
302 		LS2_256(defix, p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7]);
303 		quasi_exform256(p0, p1, p2, p3, p4, p5, p6, p7);
304 
305 		LS1_256(defix, q0, q1, q2, q3, q4, q5, q6, q7);
306 		LS2_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
307 		quasi_exform256(q0, q1, q2, q3, q4, q5, q6, q7);
308 
309 		/* Fourth row of quasigroup e-transformations */
310 		LS1_256(defix, d(7), d(6), d(5), d(4), d(3), d(2), d(1), d(0));
311 		LS2_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
312 		quasi_exform256(p0, p1, p2, p3, p4, p5, p6, p7);
313 
314 		LS1_256(defix, p0, p1, p2, p3, p4, p5, p6, p7);
315 		LS2_256(defix, q0, q1, q2, q3, q4, q5, q6, q7);
316 		quasi_exform256(q0, q1, q2, q3, q4, q5, q6, q7);
317 
318 		/* Edon-R tweak on the original SHA-3 Edon-R submission. */
319 		p[0] ^= d(8) ^ p0;
320 		p[1] ^= d(9) ^ p1;
321 		p[2] ^= d(10) ^ p2;
322 		p[3] ^= d(11) ^ p3;
323 		p[4] ^= d(12) ^ p4;
324 		p[5] ^= d(13) ^ p5;
325 		p[6] ^= d(14) ^ p6;
326 		p[7] ^= d(15) ^ p7;
327 		p[8] ^= d(0) ^ q0;
328 		p[9] ^= d(1) ^ q1;
329 		p[10] ^= d(2) ^ q2;
330 		p[11] ^= d(3) ^ q3;
331 		p[12] ^= d(4) ^ q4;
332 		p[13] ^= d(5) ^ q5;
333 		p[14] ^= d(6) ^ q6;
334 		p[15] ^= d(7) ^ q7;
335 	}
336 
337 #undef d
338 	return (bitlen - bl);
339 }
340 
341 /*
342  * Why is this #pragma here?
343  *
344  * Checksum functions like this one can go over the stack frame size check
345  * Linux imposes on 32-bit platforms (-Wframe-larger-than=1024).  We can
346  * safely ignore the compiler error since we know that in OpenZFS, that
347  * the function will be called from a worker thread that won't be using
348  * much stack.  The only function that goes over the 1k limit is Q512(),
349  * which only goes over it by a hair (1248 bytes on ARM32).
350  */
351 #include <sys/isa_defs.h>	/* for _ILP32 */
352 #ifdef _ILP32   /* We're 32-bit, assume small stack frames */
353 #pragma GCC diagnostic ignored "-Wframe-larger-than="
354 #endif
355 
356 #if defined(__IBMC__) && defined(_AIX) && defined(__64BIT__)
357 static inline size_t
358 #else
359 static size_t
360 #endif
361 Q512(size_t bitlen, const uint64_t *data, uint64_t *restrict p)
362 {
363 	size_t bl;
364 
365 	for (bl = bitlen; bl >= EdonR512_BLOCK_BITSIZE;
366 	    bl -= EdonR512_BLOCK_BITSIZE, data += 16) {
367 		uint64_t s0, s1, s2, s3, s4, s5, s6, s7, t0, t1, t2, t3, t4,
368 		    t5, t6, t7;
369 		uint64_t p0, p1, p2, p3, p4, p5, p6, p7, q0, q1, q2, q3, q4,
370 		    q5, q6, q7;
371 		const uint64_t defix = 0xaaaaaaaaaaaaaaaaull;
372 #if defined(MACHINE_IS_BIG_ENDIAN)
373 		uint64_t swp0, swp1, swp2, swp3, swp4, swp5, swp6, swp7, swp8,
374 		    swp9, swp10, swp11, swp12, swp13, swp14, swp15;
375 #define	d(j)	swp##j
376 #define	s64(j)	ld_swap64((uint64_t *)data+j, swp##j)
377 #else
378 #define	d(j)	data[j]
379 #endif
380 
381 		/* First row of quasigroup e-transformations */
382 #if defined(MACHINE_IS_BIG_ENDIAN)
383 		s64(8);
384 		s64(9);
385 		s64(10);
386 		s64(11);
387 		s64(12);
388 		s64(13);
389 		s64(14);
390 		s64(15);
391 #endif
392 		LS1_512(defix, d(15), d(14), d(13), d(12), d(11), d(10), d(9),
393 		    d(8));
394 #if defined(MACHINE_IS_BIG_ENDIAN)
395 		s64(0);
396 		s64(1);
397 		s64(2);
398 		s64(3);
399 		s64(4);
400 		s64(5);
401 		s64(6);
402 		s64(7);
403 #undef s64
404 #endif
405 		LS2_512(defix, d(0), d(1), d(2), d(3), d(4), d(5), d(6), d(7));
406 		quasi_exform512(p0, p1, p2, p3, p4, p5, p6, p7);
407 
408 		LS1_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
409 		LS2_512(defix, d(8), d(9), d(10), d(11), d(12), d(13), d(14),
410 		    d(15));
411 		quasi_exform512(q0, q1, q2, q3, q4, q5, q6, q7);
412 
413 		/* Second row of quasigroup e-transformations */
414 		LS1_512(defix, p[8], p[9], p[10], p[11], p[12], p[13], p[14],
415 		    p[15]);
416 		LS2_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
417 		quasi_exform512(p0, p1, p2, p3, p4, p5, p6, p7);
418 
419 		LS1_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
420 		LS2_512(defix, q0, q1, q2, q3, q4, q5, q6, q7);
421 		quasi_exform512(q0, q1, q2, q3, q4, q5, q6, q7);
422 
423 		/* Third row of quasigroup e-transformations */
424 		LS1_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
425 		LS2_512(defix, p[0], p[1], p[2], p[3], p[4], p[5], p[6], p[7]);
426 		quasi_exform512(p0, p1, p2, p3, p4, p5, p6, p7);
427 
428 		LS1_512(defix, q0, q1, q2, q3, q4, q5, q6, q7);
429 		LS2_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
430 		quasi_exform512(q0, q1, q2, q3, q4, q5, q6, q7);
431 
432 		/* Fourth row of quasigroup e-transformations */
433 		LS1_512(defix, d(7), d(6), d(5), d(4), d(3), d(2), d(1), d(0));
434 		LS2_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
435 		quasi_exform512(p0, p1, p2, p3, p4, p5, p6, p7);
436 
437 		LS1_512(defix, p0, p1, p2, p3, p4, p5, p6, p7);
438 		LS2_512(defix, q0, q1, q2, q3, q4, q5, q6, q7);
439 		quasi_exform512(q0, q1, q2, q3, q4, q5, q6, q7);
440 
441 		/* Edon-R tweak on the original SHA-3 Edon-R submission. */
442 		p[0] ^= d(8) ^ p0;
443 		p[1] ^= d(9) ^ p1;
444 		p[2] ^= d(10) ^ p2;
445 		p[3] ^= d(11) ^ p3;
446 		p[4] ^= d(12) ^ p4;
447 		p[5] ^= d(13) ^ p5;
448 		p[6] ^= d(14) ^ p6;
449 		p[7] ^= d(15) ^ p7;
450 		p[8] ^= d(0) ^ q0;
451 		p[9] ^= d(1) ^ q1;
452 		p[10] ^= d(2) ^ q2;
453 		p[11] ^= d(3) ^ q3;
454 		p[12] ^= d(4) ^ q4;
455 		p[13] ^= d(5) ^ q5;
456 		p[14] ^= d(6) ^ q6;
457 		p[15] ^= d(7) ^ q7;
458 	}
459 
460 #undef d
461 	return (bitlen - bl);
462 }
463 
464 void
465 EdonRInit(EdonRState *state, size_t hashbitlen)
466 {
467 	ASSERT(EDONR_VALID_HASHBITLEN(hashbitlen));
468 	switch (hashbitlen) {
469 	case 224:
470 		state->hashbitlen = 224;
471 		state->bits_processed = 0;
472 		state->unprocessed_bits = 0;
473 		bcopy(i224p2, hashState224(state)->DoublePipe,
474 		    16 * sizeof (uint32_t));
475 		break;
476 
477 	case 256:
478 		state->hashbitlen = 256;
479 		state->bits_processed = 0;
480 		state->unprocessed_bits = 0;
481 		bcopy(i256p2, hashState256(state)->DoublePipe,
482 		    16 * sizeof (uint32_t));
483 		break;
484 
485 	case 384:
486 		state->hashbitlen = 384;
487 		state->bits_processed = 0;
488 		state->unprocessed_bits = 0;
489 		bcopy(i384p2, hashState384(state)->DoublePipe,
490 		    16 * sizeof (uint64_t));
491 		break;
492 
493 	case 512:
494 		state->hashbitlen = 512;
495 		state->bits_processed = 0;
496 		state->unprocessed_bits = 0;
497 		bcopy(i512p2, hashState224(state)->DoublePipe,
498 		    16 * sizeof (uint64_t));
499 		break;
500 	}
501 }
502 
503 
504 void
505 EdonRUpdate(EdonRState *state, const uint8_t *data, size_t databitlen)
506 {
507 	uint32_t *data32;
508 	uint64_t *data64;
509 
510 	size_t bits_processed;
511 
512 	ASSERT(EDONR_VALID_HASHBITLEN(state->hashbitlen));
513 	switch (state->hashbitlen) {
514 	case 224:
515 	case 256:
516 		if (state->unprocessed_bits > 0) {
517 			/* LastBytes = databitlen / 8 */
518 			int LastBytes = (int)databitlen >> 3;
519 
520 			ASSERT(state->unprocessed_bits + databitlen <=
521 			    EdonR256_BLOCK_SIZE * 8);
522 
523 			bcopy(data, hashState256(state)->LastPart
524 			    + (state->unprocessed_bits >> 3), LastBytes);
525 			state->unprocessed_bits += (int)databitlen;
526 			databitlen = state->unprocessed_bits;
527 			/* LINTED E_BAD_PTR_CAST_ALIGN */
528 			data32 = (uint32_t *)hashState256(state)->LastPart;
529 		} else
530 			/* LINTED E_BAD_PTR_CAST_ALIGN */
531 			data32 = (uint32_t *)data;
532 
533 		bits_processed = Q256(databitlen, data32,
534 		    hashState256(state)->DoublePipe);
535 		state->bits_processed += bits_processed;
536 		databitlen -= bits_processed;
537 		state->unprocessed_bits = (int)databitlen;
538 		if (databitlen > 0) {
539 			/* LastBytes = Ceil(databitlen / 8) */
540 			int LastBytes =
541 			    ((~(((-(int)databitlen) >> 3) & 0x01ff)) +
542 			    1) & 0x01ff;
543 
544 			data32 += bits_processed >> 5;	/* byte size update */
545 			bcopy(data32, hashState256(state)->LastPart, LastBytes);
546 		}
547 		break;
548 
549 	case 384:
550 	case 512:
551 		if (state->unprocessed_bits > 0) {
552 			/* LastBytes = databitlen / 8 */
553 			int LastBytes = (int)databitlen >> 3;
554 
555 			ASSERT(state->unprocessed_bits + databitlen <=
556 			    EdonR512_BLOCK_SIZE * 8);
557 
558 			bcopy(data, hashState512(state)->LastPart
559 			    + (state->unprocessed_bits >> 3), LastBytes);
560 			state->unprocessed_bits += (int)databitlen;
561 			databitlen = state->unprocessed_bits;
562 			/* LINTED E_BAD_PTR_CAST_ALIGN */
563 			data64 = (uint64_t *)hashState512(state)->LastPart;
564 		} else
565 			/* LINTED E_BAD_PTR_CAST_ALIGN */
566 			data64 = (uint64_t *)data;
567 
568 		bits_processed = Q512(databitlen, data64,
569 		    hashState512(state)->DoublePipe);
570 		state->bits_processed += bits_processed;
571 		databitlen -= bits_processed;
572 		state->unprocessed_bits = (int)databitlen;
573 		if (databitlen > 0) {
574 			/* LastBytes = Ceil(databitlen / 8) */
575 			int LastBytes =
576 			    ((~(((-(int)databitlen) >> 3) & 0x03ff)) +
577 			    1) & 0x03ff;
578 
579 			data64 += bits_processed >> 6;	/* byte size update */
580 			bcopy(data64, hashState512(state)->LastPart, LastBytes);
581 		}
582 		break;
583 	}
584 }
585 
586 void
587 EdonRFinal(EdonRState *state, uint8_t *hashval)
588 {
589 	uint32_t *data32;
590 	uint64_t *data64, num_bits;
591 
592 	size_t databitlen;
593 	int LastByte, PadOnePosition;
594 
595 	num_bits = state->bits_processed + state->unprocessed_bits;
596 	ASSERT(EDONR_VALID_HASHBITLEN(state->hashbitlen));
597 	switch (state->hashbitlen) {
598 	case 224:
599 	case 256:
600 		LastByte = (int)state->unprocessed_bits >> 3;
601 		PadOnePosition = 7 - (state->unprocessed_bits & 0x07);
602 		hashState256(state)->LastPart[LastByte] =
603 		    (hashState256(state)->LastPart[LastByte]
604 		    & (0xff << (PadOnePosition + 1))) ^
605 		    (0x01 << PadOnePosition);
606 		/* LINTED E_BAD_PTR_CAST_ALIGN */
607 		data64 = (uint64_t *)hashState256(state)->LastPart;
608 
609 		if (state->unprocessed_bits < 448) {
610 			(void) memset((hashState256(state)->LastPart) +
611 			    LastByte + 1, 0x00,
612 			    EdonR256_BLOCK_SIZE - LastByte - 9);
613 			databitlen = EdonR256_BLOCK_SIZE * 8;
614 #if defined(MACHINE_IS_BIG_ENDIAN)
615 			st_swap64(num_bits, data64 + 7);
616 #else
617 			data64[7] = num_bits;
618 #endif
619 		} else {
620 			(void) memset((hashState256(state)->LastPart) +
621 			    LastByte + 1, 0x00,
622 			    EdonR256_BLOCK_SIZE * 2 - LastByte - 9);
623 			databitlen = EdonR256_BLOCK_SIZE * 16;
624 #if defined(MACHINE_IS_BIG_ENDIAN)
625 			st_swap64(num_bits, data64 + 15);
626 #else
627 			data64[15] = num_bits;
628 #endif
629 		}
630 
631 		/* LINTED E_BAD_PTR_CAST_ALIGN */
632 		data32 = (uint32_t *)hashState256(state)->LastPart;
633 		state->bits_processed += Q256(databitlen, data32,
634 		    hashState256(state)->DoublePipe);
635 		break;
636 
637 	case 384:
638 	case 512:
639 		LastByte = (int)state->unprocessed_bits >> 3;
640 		PadOnePosition = 7 - (state->unprocessed_bits & 0x07);
641 		hashState512(state)->LastPart[LastByte] =
642 		    (hashState512(state)->LastPart[LastByte]
643 		    & (0xff << (PadOnePosition + 1))) ^
644 		    (0x01 << PadOnePosition);
645 		/* LINTED E_BAD_PTR_CAST_ALIGN */
646 		data64 = (uint64_t *)hashState512(state)->LastPart;
647 
648 		if (state->unprocessed_bits < 960) {
649 			(void) memset((hashState512(state)->LastPart) +
650 			    LastByte + 1, 0x00,
651 			    EdonR512_BLOCK_SIZE - LastByte - 9);
652 			databitlen = EdonR512_BLOCK_SIZE * 8;
653 #if defined(MACHINE_IS_BIG_ENDIAN)
654 			st_swap64(num_bits, data64 + 15);
655 #else
656 			data64[15] = num_bits;
657 #endif
658 		} else {
659 			(void) memset((hashState512(state)->LastPart) +
660 			    LastByte + 1, 0x00,
661 			    EdonR512_BLOCK_SIZE * 2 - LastByte - 9);
662 			databitlen = EdonR512_BLOCK_SIZE * 16;
663 #if defined(MACHINE_IS_BIG_ENDIAN)
664 			st_swap64(num_bits, data64 + 31);
665 #else
666 			data64[31] = num_bits;
667 #endif
668 		}
669 
670 		state->bits_processed += Q512(databitlen, data64,
671 		    hashState512(state)->DoublePipe);
672 		break;
673 	}
674 
675 	switch (state->hashbitlen) {
676 	case 224: {
677 #if defined(MACHINE_IS_BIG_ENDIAN)
678 		uint32_t *d32 = (uint32_t *)hashval;
679 		uint32_t *s32 = hashState224(state)->DoublePipe + 9;
680 		int j;
681 
682 		for (j = 0; j < EdonR224_DIGEST_SIZE >> 2; j++)
683 			st_swap32(s32[j], d32 + j);
684 #else
685 		bcopy(hashState256(state)->DoublePipe + 9, hashval,
686 		    EdonR224_DIGEST_SIZE);
687 #endif
688 		break;
689 	}
690 	case 256: {
691 #if defined(MACHINE_IS_BIG_ENDIAN)
692 		uint32_t *d32 = (uint32_t *)hashval;
693 		uint32_t *s32 = hashState224(state)->DoublePipe + 8;
694 		int j;
695 
696 		for (j = 0; j < EdonR256_DIGEST_SIZE >> 2; j++)
697 			st_swap32(s32[j], d32 + j);
698 #else
699 		bcopy(hashState256(state)->DoublePipe + 8, hashval,
700 		    EdonR256_DIGEST_SIZE);
701 #endif
702 		break;
703 	}
704 	case 384: {
705 #if defined(MACHINE_IS_BIG_ENDIAN)
706 		uint64_t *d64 = (uint64_t *)hashval;
707 		uint64_t *s64 = hashState384(state)->DoublePipe + 10;
708 		int j;
709 
710 		for (j = 0; j < EdonR384_DIGEST_SIZE >> 3; j++)
711 			st_swap64(s64[j], d64 + j);
712 #else
713 		bcopy(hashState384(state)->DoublePipe + 10, hashval,
714 		    EdonR384_DIGEST_SIZE);
715 #endif
716 		break;
717 	}
718 	case 512: {
719 #if defined(MACHINE_IS_BIG_ENDIAN)
720 		uint64_t *d64 = (uint64_t *)hashval;
721 		uint64_t *s64 = hashState512(state)->DoublePipe + 8;
722 		int j;
723 
724 		for (j = 0; j < EdonR512_DIGEST_SIZE >> 3; j++)
725 			st_swap64(s64[j], d64 + j);
726 #else
727 		bcopy(hashState512(state)->DoublePipe + 8, hashval,
728 		    EdonR512_DIGEST_SIZE);
729 #endif
730 		break;
731 	}
732 	}
733 }
734 
735 
736 void
737 EdonRHash(size_t hashbitlen, const uint8_t *data, size_t databitlen,
738     uint8_t *hashval)
739 {
740 	EdonRState state;
741 
742 	EdonRInit(&state, hashbitlen);
743 	EdonRUpdate(&state, data, databitlen);
744 	EdonRFinal(&state, hashval);
745 }
746 
747 #ifdef _KERNEL
748 EXPORT_SYMBOL(EdonRInit);
749 EXPORT_SYMBOL(EdonRUpdate);
750 EXPORT_SYMBOL(EdonRHash);
751 EXPORT_SYMBOL(EdonRFinal);
752 #endif
753