1/*****************************************************************************
2*                                                                            *
3*  Copyright (c) 2012, Intel Corporation                                     *
4*                                                                            *
5*  All rights reserved.                                                      *
6*                                                                            *
7*  Redistribution and use in source and binary forms, with or without        *
8*  modification, are permitted provided that the following conditions are    *
9*  met:                                                                      *
10*                                                                            *
11*  *  Redistributions of source code must retain the above copyright         *
12*     notice, this list of conditions and the following disclaimer.          *
13*                                                                            *
14*  *  Redistributions in binary form must reproduce the above copyright      *
15*     notice, this list of conditions and the following disclaimer in the    *
16*     documentation and/or other materials provided with the                 *
17*     distribution.                                                          *
18*                                                                            *
19*  *  Neither the name of the Intel Corporation nor the names of its         *
20*     contributors may be used to endorse or promote products derived from   *
21*     this software without specific prior written permission.               *
22*                                                                            *
23*                                                                            *
24*  THIS SOFTWARE IS PROVIDED BY INTEL CORPORATION ""AS IS"" AND ANY          *
25*  EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE         *
26*  IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR        *
27*  PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL INTEL CORPORATION OR            *
28*  CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,     *
29*  EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,       *
30*  PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR        *
31*  PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF    *
32*  LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING      *
33*  NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS        *
34*  SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.              *
35*                                                                            *
36******************************************************************************
37* Developers and authors:                                                    *
38* Shay Gueron (1, 2), and Vlad Krasnov (1)                                   *
39* (1) Intel Corporation, Israel Development Center, Haifa, Israel            *
40* (2) University of Haifa, Israel                                            *
41*****************************************************************************/
42
43#include <openssl/base.h>
44
45#if !defined(OPENSSL_NO_ASM) && defined(OPENSSL_X86_64)
46
47#include "rsaz_exp.h"
48
49#include <openssl/mem.h>
50
51/*
52 * See crypto/bn/asm/rsaz-avx2.pl for further details.
53 */
54void rsaz_1024_norm2red_avx2(void *red,const void *norm);
55void rsaz_1024_mul_avx2(void *ret,const void *a,const void *b,const void *n,BN_ULONG k);
56void rsaz_1024_sqr_avx2(void *ret,const void *a,const void *n,BN_ULONG k,int cnt);
57void rsaz_1024_scatter5_avx2(void *tbl,const void *val,int i);
58void rsaz_1024_gather5_avx2(void *val,const void *tbl,int i);
59void rsaz_1024_red2norm_avx2(void *norm,const void *red);
60
61#if defined(__GNUC__)
62# define ALIGN64	__attribute__((aligned(64)))
63#elif defined(_MSC_VER)
64# define ALIGN64	__declspec(align(64))
65#elif defined(__SUNPRO_C)
66# define ALIGN64
67# pragma align 64(one,two80)
68#else
69# define ALIGN64	/* not fatal, might hurt performance a little */
70#endif
71
72ALIGN64 static const BN_ULONG one[40] =
73	{1,0,0,    0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
74ALIGN64 static const BN_ULONG two80[40] =
75	{0,0,1<<22,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0,0};
76
77void RSAZ_1024_mod_exp_avx2(BN_ULONG result_norm[16],
78	const BN_ULONG base_norm[16], const BN_ULONG exponent[16],
79	const BN_ULONG m_norm[16], const BN_ULONG RR[16], BN_ULONG k0)
80{
81	unsigned char	 storage[320*3+32*9*16+64];	/* 5.5KB */
82	unsigned char	*p_str = storage + (64-((size_t)storage%64));
83	unsigned char	*a_inv, *m, *result,
84			*table_s = p_str+320*3,
85			*R2      = table_s;	/* borrow */
86	int index;
87	int wvalue;
88
89	if ((((size_t)p_str&4095)+320)>>12) {
90		result = p_str;
91		a_inv = p_str + 320;
92		m = p_str + 320*2;	/* should not cross page */
93	} else {
94		m = p_str;		/* should not cross page */
95		result = p_str + 320;
96		a_inv = p_str + 320*2;
97	}
98
99	rsaz_1024_norm2red_avx2(m, m_norm);
100	rsaz_1024_norm2red_avx2(a_inv, base_norm);
101	rsaz_1024_norm2red_avx2(R2, RR);
102
103	rsaz_1024_mul_avx2(R2, R2, R2, m, k0);
104	rsaz_1024_mul_avx2(R2, R2, two80, m, k0);
105
106	/* table[0] = 1 */
107	rsaz_1024_mul_avx2(result, R2, one, m, k0);
108	/* table[1] = a_inv^1 */
109	rsaz_1024_mul_avx2(a_inv, a_inv, R2, m, k0);
110
111	rsaz_1024_scatter5_avx2(table_s,result,0);
112	rsaz_1024_scatter5_avx2(table_s,a_inv,1);
113
114	/* table[2] = a_inv^2 */
115	rsaz_1024_sqr_avx2(result, a_inv, m, k0, 1);
116	rsaz_1024_scatter5_avx2(table_s,result,2);
117#if 0
118	/* this is almost 2x smaller and less than 1% slower */
119	for (index=3; index<32; index++) {
120		rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
121		rsaz_1024_scatter5_avx2(table_s,result,index);
122	}
123#else
124	/* table[4] = a_inv^4 */
125	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
126	rsaz_1024_scatter5_avx2(table_s,result,4);
127	/* table[8] = a_inv^8 */
128	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
129	rsaz_1024_scatter5_avx2(table_s,result,8);
130	/* table[16] = a_inv^16 */
131	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
132	rsaz_1024_scatter5_avx2(table_s,result,16);
133	/* table[17] = a_inv^17 */
134	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
135	rsaz_1024_scatter5_avx2(table_s,result,17);
136
137	/* table[3] */
138	rsaz_1024_gather5_avx2(result,table_s,2);
139	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
140	rsaz_1024_scatter5_avx2(table_s,result,3);
141	/* table[6] */
142	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
143	rsaz_1024_scatter5_avx2(table_s,result,6);
144	/* table[12] */
145	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
146	rsaz_1024_scatter5_avx2(table_s,result,12);
147 	/* table[24] */
148	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
149	rsaz_1024_scatter5_avx2(table_s,result,24);
150	/* table[25] */
151	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
152	rsaz_1024_scatter5_avx2(table_s,result,25);
153
154	/* table[5] */
155	rsaz_1024_gather5_avx2(result,table_s,4);
156	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
157	rsaz_1024_scatter5_avx2(table_s,result,5);
158	/* table[10] */
159	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
160	rsaz_1024_scatter5_avx2(table_s,result,10);
161	/* table[20] */
162	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
163	rsaz_1024_scatter5_avx2(table_s,result,20);
164	/* table[21] */
165	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
166	rsaz_1024_scatter5_avx2(table_s,result,21);
167
168	/* table[7] */
169	rsaz_1024_gather5_avx2(result,table_s,6);
170	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
171	rsaz_1024_scatter5_avx2(table_s,result,7);
172	/* table[14] */
173	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
174	rsaz_1024_scatter5_avx2(table_s,result,14);
175	/* table[28] */
176	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
177	rsaz_1024_scatter5_avx2(table_s,result,28);
178	/* table[29] */
179	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
180	rsaz_1024_scatter5_avx2(table_s,result,29);
181
182	/* table[9] */
183	rsaz_1024_gather5_avx2(result,table_s,8);
184	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
185	rsaz_1024_scatter5_avx2(table_s,result,9);
186	/* table[18] */
187	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
188	rsaz_1024_scatter5_avx2(table_s,result,18);
189	/* table[19] */
190	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
191	rsaz_1024_scatter5_avx2(table_s,result,19);
192
193	/* table[11] */
194	rsaz_1024_gather5_avx2(result,table_s,10);
195	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
196	rsaz_1024_scatter5_avx2(table_s,result,11);
197	/* table[22] */
198	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
199	rsaz_1024_scatter5_avx2(table_s,result,22);
200	/* table[23] */
201	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
202	rsaz_1024_scatter5_avx2(table_s,result,23);
203
204	/* table[13] */
205	rsaz_1024_gather5_avx2(result,table_s,12);
206	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
207	rsaz_1024_scatter5_avx2(table_s,result,13);
208	/* table[26] */
209	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
210	rsaz_1024_scatter5_avx2(table_s,result,26);
211	/* table[27] */
212	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
213	rsaz_1024_scatter5_avx2(table_s,result,27);
214
215	/* table[15] */
216	rsaz_1024_gather5_avx2(result,table_s,14);
217	rsaz_1024_mul_avx2(result,result,a_inv,m,k0);
218	rsaz_1024_scatter5_avx2(table_s,result,15);
219	/* table[30] */
220	rsaz_1024_sqr_avx2(result, result, m, k0, 1);
221	rsaz_1024_scatter5_avx2(table_s,result,30);
222	/* table[31] */
223	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
224	rsaz_1024_scatter5_avx2(table_s,result,31);
225#endif
226
227	/* load first window */
228	p_str = (unsigned char*)exponent;
229	wvalue = p_str[127] >> 3;
230	rsaz_1024_gather5_avx2(result,table_s,wvalue);
231
232	index = 1014;
233
234	while(index > -1) {	/* loop for the remaining 127 windows */
235
236		rsaz_1024_sqr_avx2(result, result, m, k0, 5);
237
238		wvalue = *((unsigned short*)&p_str[index/8]);
239		wvalue = (wvalue>> (index%8)) & 31;
240		index-=5;
241
242		rsaz_1024_gather5_avx2(a_inv,table_s,wvalue);	/* borrow a_inv */
243		rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
244	}
245
246	/* square four times */
247	rsaz_1024_sqr_avx2(result, result, m, k0, 4);
248
249	wvalue = p_str[0] & 15;
250
251	rsaz_1024_gather5_avx2(a_inv,table_s,wvalue);	/* borrow a_inv */
252	rsaz_1024_mul_avx2(result, result, a_inv, m, k0);
253
254	/* from Montgomery */
255	rsaz_1024_mul_avx2(result, result, one, m, k0);
256
257	rsaz_1024_red2norm_avx2(result_norm, result);
258
259	OPENSSL_cleanse(storage,sizeof(storage));
260}
261
262/*
263 * See crypto/bn/rsaz-x86_64.pl for further details.
264 */
265void rsaz_512_mul(void *ret,const void *a,const void *b,const void *n,BN_ULONG k);
266void rsaz_512_mul_scatter4(void *ret,const void *a,const void *n,BN_ULONG k,const void *tbl,unsigned int power);
267void rsaz_512_mul_gather4(void *ret,const void *a,const void *tbl,const void *n,BN_ULONG k,unsigned int power);
268void rsaz_512_mul_by_one(void *ret,const void *a,const void *n,BN_ULONG k);
269void rsaz_512_sqr(void *ret,const void *a,const void *n,BN_ULONG k,int cnt);
270void rsaz_512_scatter4(void *tbl, const BN_ULONG *val, int power);
271void rsaz_512_gather4(BN_ULONG *val, const void *tbl, int power);
272
273void RSAZ_512_mod_exp(BN_ULONG result[8],
274	const BN_ULONG base[8], const BN_ULONG exponent[8],
275	const BN_ULONG m[8], BN_ULONG k0, const BN_ULONG RR[8])
276{
277	unsigned char	 storage[16*8*8+64*2+64];	/* 1.2KB */
278	unsigned char	*table = storage + (64-((size_t)storage%64));
279	BN_ULONG	*a_inv = (BN_ULONG *)(table+16*8*8),
280			*temp  = (BN_ULONG *)(table+16*8*8+8*8);
281	unsigned char	*p_str = (unsigned char*)exponent;
282	int index;
283	unsigned int wvalue;
284
285	/* table[0] = 1_inv */
286	temp[0] = 0-m[0];	temp[1] = ~m[1];
287	temp[2] = ~m[2];	temp[3] = ~m[3];
288	temp[4] = ~m[4];	temp[5] = ~m[5];
289	temp[6] = ~m[6];	temp[7] = ~m[7];
290	rsaz_512_scatter4(table, temp, 0);
291
292	/* table [1] = a_inv^1 */
293	rsaz_512_mul(a_inv, base, RR, m, k0);
294	rsaz_512_scatter4(table, a_inv, 1);
295
296	/* table [2] = a_inv^2 */
297	rsaz_512_sqr(temp, a_inv, m, k0, 1);
298	rsaz_512_scatter4(table, temp, 2);
299
300	for (index=3; index<16; index++)
301		rsaz_512_mul_scatter4(temp, a_inv, m, k0, table, index);
302
303	/* load first window */
304	wvalue = p_str[63];
305
306	rsaz_512_gather4(temp, table, wvalue>>4);
307	rsaz_512_sqr(temp, temp, m, k0, 4);
308	rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue&0xf);
309
310	for (index=62; index>=0; index--) {
311		wvalue = p_str[index];
312
313		rsaz_512_sqr(temp, temp, m, k0, 4);
314		rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue>>4);
315
316		rsaz_512_sqr(temp, temp, m, k0, 4);
317		rsaz_512_mul_gather4(temp, temp, table, m, k0, wvalue&0x0f);
318	}
319
320	/* from Montgomery */
321	rsaz_512_mul_by_one(result, temp, m, k0);
322
323	OPENSSL_cleanse(storage,sizeof(storage));
324}
325
326#endif  /* OPENSSL_X86_64 */
327