/src/openssl/crypto/bn/rsaz_exp_x2.c
Line | Count | Source (jump to first uncovered line) |
1 | | /* |
2 | | * Copyright 2020-2025 The OpenSSL Project Authors. All Rights Reserved. |
3 | | * Copyright (c) 2020-2021, Intel Corporation. All Rights Reserved. |
4 | | * |
5 | | * Licensed under the Apache License 2.0 (the "License"). You may not use |
6 | | * this file except in compliance with the License. You can obtain a copy |
7 | | * in the file LICENSE in the source distribution or at |
8 | | * https://www.openssl.org/source/license.html |
9 | | * |
10 | | * |
11 | | * Originally written by Sergey Kirillov and Andrey Matyukov. |
12 | | * Special thanks to Ilya Albrekht for his valuable hints. |
13 | | * Intel Corporation |
14 | | * |
15 | | */ |
16 | | |
17 | | #include <openssl/opensslconf.h> |
18 | | #include <openssl/crypto.h> |
19 | | #include "rsaz_exp.h" |
20 | | |
21 | | #ifndef RSAZ_ENABLED |
22 | | NON_EMPTY_TRANSLATION_UNIT |
23 | | #else |
24 | | # include <assert.h> |
25 | | # include <string.h> |
26 | | |
27 | | # define ALIGN_OF(ptr, boundary) \ |
28 | 0 | ((unsigned char *)(ptr) + (boundary - (((size_t)(ptr)) & (boundary - 1)))) |
29 | | |
30 | | /* Internal radix */ |
31 | 0 | # define DIGIT_SIZE (52) |
32 | | /* 52-bit mask */ |
33 | 0 | # define DIGIT_MASK ((uint64_t)0xFFFFFFFFFFFFF) |
34 | | |
35 | 0 | # define BITS2WORD8_SIZE(x) (((x) + 7) >> 3) |
36 | 0 | # define BITS2WORD64_SIZE(x) (((x) + 63) >> 6) |
37 | | |
38 | | /* Number of registers required to hold |digits_num| amount of qword digits */ |
39 | | # define NUMBER_OF_REGISTERS(digits_num, register_size) \ |
40 | 0 | (((digits_num) * 64 + (register_size) - 1) / (register_size)) |
41 | | |
42 | | static ossl_inline uint64_t get_digit(const uint8_t *in, int in_len); |
43 | | static ossl_inline void put_digit(uint8_t *out, int out_len, uint64_t digit); |
44 | | static void to_words52(BN_ULONG *out, int out_len, const BN_ULONG *in, |
45 | | int in_bitsize); |
46 | | static void from_words52(BN_ULONG *bn_out, int out_bitsize, const BN_ULONG *in); |
47 | | static ossl_inline void set_bit(BN_ULONG *a, int idx); |
48 | | |
49 | | /* Number of |digit_size|-bit digits in |bitsize|-bit value */ |
50 | | static ossl_inline int number_of_digits(int bitsize, int digit_size) |
51 | 0 | { |
52 | 0 | return (bitsize + digit_size - 1) / digit_size; |
53 | 0 | } |
54 | | |
55 | | /* |
56 | | * For details of the methods declared below please refer to |
57 | | * crypto/bn/asm/rsaz-avx512.pl |
58 | | * |
59 | | * Naming conventions: |
60 | | * amm = Almost Montgomery Multiplication |
61 | | * ams = Almost Montgomery Squaring |
62 | | * 52xZZ - data represented as array of ZZ digits in 52-bit radix |
63 | | * _x1_/_x2_ - 1 or 2 independent inputs/outputs |
64 | | * _ifma256 - uses 256-bit wide IFMA ISA (AVX512_IFMA256) |
65 | | * _avxifma256 - uses 256-bit wide AVXIFMA ISA (AVX_IFMA256) |
66 | | */ |
67 | | |
68 | | void ossl_rsaz_amm52x20_x1_ifma256(BN_ULONG *res, const BN_ULONG *a, |
69 | | const BN_ULONG *b, const BN_ULONG *m, |
70 | | BN_ULONG k0); |
71 | | void ossl_rsaz_amm52x20_x2_ifma256(BN_ULONG *out, const BN_ULONG *a, |
72 | | const BN_ULONG *b, const BN_ULONG *m, |
73 | | const BN_ULONG k0[2]); |
74 | | void ossl_extract_multiplier_2x20_win5(BN_ULONG *red_Y, |
75 | | const BN_ULONG *red_table, |
76 | | int red_table_idx1, int red_table_idx2); |
77 | | |
78 | | void ossl_rsaz_amm52x30_x1_ifma256(BN_ULONG *res, const BN_ULONG *a, |
79 | | const BN_ULONG *b, const BN_ULONG *m, |
80 | | BN_ULONG k0); |
81 | | void ossl_rsaz_amm52x30_x2_ifma256(BN_ULONG *out, const BN_ULONG *a, |
82 | | const BN_ULONG *b, const BN_ULONG *m, |
83 | | const BN_ULONG k0[2]); |
84 | | void ossl_extract_multiplier_2x30_win5(BN_ULONG *red_Y, |
85 | | const BN_ULONG *red_table, |
86 | | int red_table_idx1, int red_table_idx2); |
87 | | |
88 | | void ossl_rsaz_amm52x40_x1_ifma256(BN_ULONG *res, const BN_ULONG *a, |
89 | | const BN_ULONG *b, const BN_ULONG *m, |
90 | | BN_ULONG k0); |
91 | | void ossl_rsaz_amm52x40_x2_ifma256(BN_ULONG *out, const BN_ULONG *a, |
92 | | const BN_ULONG *b, const BN_ULONG *m, |
93 | | const BN_ULONG k0[2]); |
94 | | void ossl_extract_multiplier_2x40_win5(BN_ULONG *red_Y, |
95 | | const BN_ULONG *red_table, |
96 | | int red_table_idx1, int red_table_idx2); |
97 | | |
98 | | void ossl_rsaz_amm52x20_x1_avxifma256(BN_ULONG *res, const BN_ULONG *a, |
99 | | const BN_ULONG *b, const BN_ULONG *m, |
100 | | BN_ULONG k0); |
101 | | void ossl_rsaz_amm52x20_x2_avxifma256(BN_ULONG *out, const BN_ULONG *a, |
102 | | const BN_ULONG *b, const BN_ULONG *m, |
103 | | const BN_ULONG k0[2]); |
104 | | void ossl_extract_multiplier_2x20_win5_avx(BN_ULONG *red_Y, |
105 | | const BN_ULONG *red_table, |
106 | | int red_table_idx1, |
107 | | int red_table_idx2); |
108 | | |
109 | | void ossl_rsaz_amm52x30_x1_avxifma256(BN_ULONG *res, const BN_ULONG *a, |
110 | | const BN_ULONG *b, const BN_ULONG *m, |
111 | | BN_ULONG k0); |
112 | | void ossl_rsaz_amm52x30_x2_avxifma256(BN_ULONG *out, const BN_ULONG *a, |
113 | | const BN_ULONG *b, const BN_ULONG *m, |
114 | | const BN_ULONG k0[2]); |
115 | | void ossl_extract_multiplier_2x30_win5_avx(BN_ULONG *red_Y, |
116 | | const BN_ULONG *red_table, |
117 | | int red_table_idx1, |
118 | | int red_table_idx2); |
119 | | |
120 | | void ossl_rsaz_amm52x40_x1_avxifma256(BN_ULONG *res, const BN_ULONG *a, |
121 | | const BN_ULONG *b, const BN_ULONG *m, |
122 | | BN_ULONG k0); |
123 | | void ossl_rsaz_amm52x40_x2_avxifma256(BN_ULONG *out, const BN_ULONG *a, |
124 | | const BN_ULONG *b, const BN_ULONG *m, |
125 | | const BN_ULONG k0[2]); |
126 | | void ossl_extract_multiplier_2x40_win5_avx(BN_ULONG *red_Y, |
127 | | const BN_ULONG *red_table, |
128 | | int red_table_idx1, |
129 | | int red_table_idx2); |
130 | | |
131 | | typedef void (*AMM)(BN_ULONG *res, const BN_ULONG *a, const BN_ULONG *b, |
132 | | const BN_ULONG *m, BN_ULONG k0); |
133 | | |
134 | | static AMM ossl_rsaz_amm52_x1[] = { |
135 | | ossl_rsaz_amm52x20_x1_avxifma256, ossl_rsaz_amm52x20_x1_ifma256, |
136 | | ossl_rsaz_amm52x30_x1_avxifma256, ossl_rsaz_amm52x30_x1_ifma256, |
137 | | ossl_rsaz_amm52x40_x1_avxifma256, ossl_rsaz_amm52x40_x1_ifma256, |
138 | | }; |
139 | | |
140 | | typedef void (*DAMM)(BN_ULONG *res, const BN_ULONG *a, const BN_ULONG *b, |
141 | | const BN_ULONG *m, const BN_ULONG k0[2]); |
142 | | |
143 | | static DAMM ossl_rsaz_amm52_x2[] = { |
144 | | ossl_rsaz_amm52x20_x2_avxifma256, ossl_rsaz_amm52x20_x2_ifma256, |
145 | | ossl_rsaz_amm52x30_x2_avxifma256, ossl_rsaz_amm52x30_x2_ifma256, |
146 | | ossl_rsaz_amm52x40_x2_avxifma256, ossl_rsaz_amm52x40_x2_ifma256, |
147 | | }; |
148 | | |
149 | | typedef void (*DEXTRACT)(BN_ULONG *res, const BN_ULONG *red_table, |
150 | | int red_table_idx, int tbl_idx); |
151 | | |
152 | | static DEXTRACT ossl_extract_multiplier_win5[] = { |
153 | | ossl_extract_multiplier_2x20_win5_avx, ossl_extract_multiplier_2x20_win5, |
154 | | ossl_extract_multiplier_2x30_win5_avx, ossl_extract_multiplier_2x30_win5, |
155 | | ossl_extract_multiplier_2x40_win5_avx, ossl_extract_multiplier_2x40_win5, |
156 | | }; |
157 | | |
158 | | static int RSAZ_mod_exp_x2_ifma256(BN_ULONG *res, const BN_ULONG *base, |
159 | | const BN_ULONG *exp[2], const BN_ULONG *m, |
160 | | const BN_ULONG *rr, const BN_ULONG k0[2], |
161 | | int modulus_bitsize); |
162 | | |
163 | | /* |
164 | | * Dual Montgomery modular exponentiation using prime moduli of the |
165 | | * same bit size, optimized with AVX512 ISA or AVXIFMA ISA. |
166 | | * |
167 | | * Input and output parameters for each exponentiation are independent and |
168 | | * denoted here by index |i|, i = 1..2. |
169 | | * |
170 | | * Input and output are all in regular 2^64 radix. |
171 | | * |
172 | | * Each moduli shall be |factor_size| bit size. |
173 | | * |
174 | | * Supported cases: |
175 | | * - 2x1024 |
176 | | * - 2x1536 |
177 | | * - 2x2048 |
178 | | * |
179 | | * [out] res|i| - result of modular exponentiation: array of qword values |
180 | | * in regular (2^64) radix. Size of array shall be enough |
181 | | * to hold |factor_size| bits. |
182 | | * [in] base|i| - base |
183 | | * [in] exp|i| - exponent |
184 | | * [in] m|i| - moduli |
185 | | * [in] rr|i| - Montgomery parameter RR = R^2 mod m|i| |
186 | | * [in] k0_|i| - Montgomery parameter k0 = -1/m|i| mod 2^64 |
187 | | * [in] factor_size - moduli bit size |
188 | | * |
189 | | * \return 0 in case of failure, |
190 | | * 1 in case of success. |
191 | | */ |
192 | | int ossl_rsaz_mod_exp_avx512_x2(BN_ULONG *res1, |
193 | | const BN_ULONG *base1, |
194 | | const BN_ULONG *exp1, |
195 | | const BN_ULONG *m1, |
196 | | const BN_ULONG *rr1, |
197 | | BN_ULONG k0_1, |
198 | | BN_ULONG *res2, |
199 | | const BN_ULONG *base2, |
200 | | const BN_ULONG *exp2, |
201 | | const BN_ULONG *m2, |
202 | | const BN_ULONG *rr2, |
203 | | BN_ULONG k0_2, |
204 | | int factor_size) |
205 | 0 | { |
206 | 0 | int ret = 0; |
207 | | |
208 | | /* |
209 | | * Number of word-size (BN_ULONG) digits to store exponent in redundant |
210 | | * representation. |
211 | | */ |
212 | 0 | int exp_digits = number_of_digits(factor_size + 2, DIGIT_SIZE); |
213 | 0 | int coeff_pow = 4 * (DIGIT_SIZE * exp_digits - factor_size); |
214 | | |
215 | | /* Number of YMM registers required to store exponent's digits */ |
216 | 0 | int ymm_regs_num = NUMBER_OF_REGISTERS(exp_digits, 256 /* ymm bit size */); |
217 | | /* Capacity of the register set (in qwords) to store exponent */ |
218 | 0 | int regs_capacity = ymm_regs_num * 4; |
219 | |
|
220 | 0 | BN_ULONG *base1_red, *m1_red, *rr1_red; |
221 | 0 | BN_ULONG *base2_red, *m2_red, *rr2_red; |
222 | 0 | BN_ULONG *coeff_red; |
223 | 0 | BN_ULONG *storage = NULL; |
224 | 0 | BN_ULONG *storage_aligned = NULL; |
225 | 0 | int storage_len_bytes = 7 * regs_capacity * sizeof(BN_ULONG) |
226 | 0 | + 64 /* alignment */; |
227 | |
|
228 | 0 | const BN_ULONG *exp[2] = {0}; |
229 | 0 | BN_ULONG k0[2] = {0}; |
230 | | /* AMM = Almost Montgomery Multiplication */ |
231 | 0 | AMM amm = NULL; |
232 | 0 | int avx512ifma = !!ossl_rsaz_avx512ifma_eligible(); |
233 | |
|
234 | 0 | if (factor_size != 1024 && factor_size != 1536 && factor_size != 2048) |
235 | 0 | goto err; |
236 | | |
237 | 0 | amm = ossl_rsaz_amm52_x1[(factor_size / 512 - 2) * 2 + avx512ifma]; |
238 | |
|
239 | 0 | storage = (BN_ULONG *)OPENSSL_malloc(storage_len_bytes); |
240 | 0 | if (storage == NULL) |
241 | 0 | goto err; |
242 | 0 | storage_aligned = (BN_ULONG *)ALIGN_OF(storage, 64); |
243 | | |
244 | | /* Memory layout for red(undant) representations */ |
245 | 0 | base1_red = storage_aligned; |
246 | 0 | base2_red = storage_aligned + 1 * regs_capacity; |
247 | 0 | m1_red = storage_aligned + 2 * regs_capacity; |
248 | 0 | m2_red = storage_aligned + 3 * regs_capacity; |
249 | 0 | rr1_red = storage_aligned + 4 * regs_capacity; |
250 | 0 | rr2_red = storage_aligned + 5 * regs_capacity; |
251 | 0 | coeff_red = storage_aligned + 6 * regs_capacity; |
252 | | |
253 | | /* Convert base_i, m_i, rr_i, from regular to 52-bit radix */ |
254 | 0 | to_words52(base1_red, regs_capacity, base1, factor_size); |
255 | 0 | to_words52(base2_red, regs_capacity, base2, factor_size); |
256 | 0 | to_words52(m1_red, regs_capacity, m1, factor_size); |
257 | 0 | to_words52(m2_red, regs_capacity, m2, factor_size); |
258 | 0 | to_words52(rr1_red, regs_capacity, rr1, factor_size); |
259 | 0 | to_words52(rr2_red, regs_capacity, rr2, factor_size); |
260 | | |
261 | | /* |
262 | | * Compute target domain Montgomery converters RR' for each modulus |
263 | | * based on precomputed original domain's RR. |
264 | | * |
265 | | * RR -> RR' transformation steps: |
266 | | * (1) coeff = 2^k |
267 | | * (2) t = AMM(RR,RR) = RR^2 / R' mod m |
268 | | * (3) RR' = AMM(t, coeff) = RR^2 * 2^k / R'^2 mod m |
269 | | * where |
270 | | * k = 4 * (52 * digits52 - modlen) |
271 | | * R = 2^(64 * ceil(modlen/64)) mod m |
272 | | * RR = R^2 mod m |
273 | | * R' = 2^(52 * ceil(modlen/52)) mod m |
274 | | * |
275 | | * EX/ modlen = 1024: k = 64, RR = 2^2048 mod m, RR' = 2^2080 mod m |
276 | | */ |
277 | 0 | memset(coeff_red, 0, exp_digits * sizeof(BN_ULONG)); |
278 | | /* (1) in reduced domain representation */ |
279 | 0 | set_bit(coeff_red, 64 * (int)(coeff_pow / 52) + coeff_pow % 52); |
280 | |
|
281 | 0 | amm(rr1_red, rr1_red, rr1_red, m1_red, k0_1); /* (2) for m1 */ |
282 | 0 | amm(rr1_red, rr1_red, coeff_red, m1_red, k0_1); /* (3) for m1 */ |
283 | |
|
284 | 0 | amm(rr2_red, rr2_red, rr2_red, m2_red, k0_2); /* (2) for m2 */ |
285 | 0 | amm(rr2_red, rr2_red, coeff_red, m2_red, k0_2); /* (3) for m2 */ |
286 | |
|
287 | 0 | exp[0] = exp1; |
288 | 0 | exp[1] = exp2; |
289 | |
|
290 | 0 | k0[0] = k0_1; |
291 | 0 | k0[1] = k0_2; |
292 | | |
293 | | /* Dual (2-exps in parallel) exponentiation */ |
294 | 0 | ret = RSAZ_mod_exp_x2_ifma256(rr1_red, base1_red, exp, m1_red, rr1_red, |
295 | 0 | k0, factor_size); |
296 | 0 | if (!ret) |
297 | 0 | goto err; |
298 | | |
299 | | /* Convert rr_i back to regular radix */ |
300 | 0 | from_words52(res1, factor_size, rr1_red); |
301 | 0 | from_words52(res2, factor_size, rr2_red); |
302 | | |
303 | | /* bn_reduce_once_in_place expects number of BN_ULONG, not bit size */ |
304 | 0 | factor_size /= sizeof(BN_ULONG) * 8; |
305 | |
|
306 | 0 | bn_reduce_once_in_place(res1, /*carry=*/0, m1, storage, factor_size); |
307 | 0 | bn_reduce_once_in_place(res2, /*carry=*/0, m2, storage, factor_size); |
308 | |
|
309 | 0 | err: |
310 | 0 | if (storage != NULL) { |
311 | 0 | OPENSSL_cleanse(storage, storage_len_bytes); |
312 | 0 | OPENSSL_free(storage); |
313 | 0 | } |
314 | 0 | return ret; |
315 | 0 | } |
316 | | |
317 | | /* |
318 | | * Dual {1024,1536,2048}-bit w-ary modular exponentiation using prime moduli of |
319 | | * the same bit size using Almost Montgomery Multiplication, optimized with |
320 | | * AVX512_IFMA256 ISA. |
321 | | * |
322 | | * The parameter w (window size) = 5. |
323 | | * |
324 | | * [out] res - result of modular exponentiation: 2x{20,30,40} qword |
325 | | * values in 2^52 radix. |
326 | | * [in] base - base (2x{20,30,40} qword values in 2^52 radix) |
327 | | * [in] exp - array of 2 pointers to {16,24,32} qword values in 2^64 radix. |
328 | | * Exponent is not converted to redundant representation. |
329 | | * [in] m - moduli (2x{20,30,40} qword values in 2^52 radix) |
330 | | * [in] rr - Montgomery parameter for 2 moduli: |
331 | | * RR(1024) = 2^2080 mod m. |
332 | | * RR(1536) = 2^3120 mod m. |
333 | | * RR(2048) = 2^4160 mod m. |
334 | | * (2x{20,30,40} qword values in 2^52 radix) |
335 | | * [in] k0 - Montgomery parameter for 2 moduli: k0 = -1/m mod 2^64 |
336 | | * |
337 | | * \return (void). |
338 | | */ |
339 | | int RSAZ_mod_exp_x2_ifma256(BN_ULONG *out, |
340 | | const BN_ULONG *base, |
341 | | const BN_ULONG *exp[2], |
342 | | const BN_ULONG *m, |
343 | | const BN_ULONG *rr, |
344 | | const BN_ULONG k0[2], |
345 | | int modulus_bitsize) |
346 | 0 | { |
347 | 0 | int ret = 0; |
348 | 0 | int idx; |
349 | | |
350 | | /* Exponent window size */ |
351 | 0 | int exp_win_size = 5; |
352 | 0 | int exp_win_mask = (1U << exp_win_size) - 1; |
353 | | |
354 | | /* |
355 | | * Number of digits (64-bit words) in redundant representation to handle |
356 | | * modulus bits |
357 | | */ |
358 | 0 | int red_digits = 0; |
359 | 0 | int exp_digits = 0; |
360 | |
|
361 | 0 | BN_ULONG *storage = NULL; |
362 | 0 | BN_ULONG *storage_aligned = NULL; |
363 | 0 | int storage_len_bytes = 0; |
364 | | |
365 | | /* Red(undant) result Y and multiplier X */ |
366 | 0 | BN_ULONG *red_Y = NULL; /* [2][red_digits] */ |
367 | 0 | BN_ULONG *red_X = NULL; /* [2][red_digits] */ |
368 | | /* Pre-computed table of base powers */ |
369 | 0 | BN_ULONG *red_table = NULL; /* [1U << exp_win_size][2][red_digits] */ |
370 | | /* Expanded exponent */ |
371 | 0 | BN_ULONG *expz = NULL; /* [2][exp_digits + 1] */ |
372 | | |
373 | | /* Dual AMM */ |
374 | 0 | DAMM damm = NULL; |
375 | | /* Extractor from red_table */ |
376 | 0 | DEXTRACT extract = NULL; |
377 | 0 | int avx512ifma = !!ossl_rsaz_avx512ifma_eligible(); |
378 | | |
379 | | /* |
380 | | * Squaring is done using multiplication now. That can be a subject of |
381 | | * optimization in future. |
382 | | */ |
383 | 0 | # define DAMS(r,a,m,k0) damm((r),(a),(a),(m),(k0)) |
384 | |
|
385 | 0 | if (modulus_bitsize != 1024 && modulus_bitsize != 1536 && modulus_bitsize != 2048) |
386 | 0 | goto err; |
387 | | |
388 | 0 | damm = ossl_rsaz_amm52_x2[(modulus_bitsize / 512 - 2) * 2 + avx512ifma]; |
389 | 0 | extract = ossl_extract_multiplier_win5[(modulus_bitsize / 512 - 2) * 2 + avx512ifma]; |
390 | |
|
391 | 0 | switch (modulus_bitsize) { |
392 | 0 | case 1024: |
393 | 0 | red_digits = 20; |
394 | 0 | exp_digits = 16; |
395 | 0 | break; |
396 | 0 | case 1536: |
397 | | /* Extended with 2 digits padding to avoid mask ops in high YMM register */ |
398 | 0 | red_digits = 30 + 2; |
399 | 0 | exp_digits = 24; |
400 | 0 | break; |
401 | 0 | case 2048: |
402 | 0 | red_digits = 40; |
403 | 0 | exp_digits = 32; |
404 | 0 | break; |
405 | 0 | default: |
406 | 0 | goto err; |
407 | 0 | } |
408 | | |
409 | 0 | storage_len_bytes = (2 * red_digits /* red_Y */ |
410 | 0 | + 2 * red_digits /* red_X */ |
411 | 0 | + 2 * red_digits * (1U << exp_win_size) /* red_table */ |
412 | 0 | + 2 * (exp_digits + 1)) /* expz */ |
413 | 0 | * sizeof(BN_ULONG) |
414 | 0 | + 64; /* alignment */ |
415 | |
|
416 | 0 | storage = (BN_ULONG *)OPENSSL_zalloc(storage_len_bytes); |
417 | 0 | if (storage == NULL) |
418 | 0 | goto err; |
419 | 0 | storage_aligned = (BN_ULONG *)ALIGN_OF(storage, 64); |
420 | |
|
421 | 0 | red_Y = storage_aligned; |
422 | 0 | red_X = red_Y + 2 * red_digits; |
423 | 0 | red_table = red_X + 2 * red_digits; |
424 | 0 | expz = red_table + 2 * red_digits * (1U << exp_win_size); |
425 | | |
426 | | /* |
427 | | * Compute table of powers base^i, i = 0, ..., (2^EXP_WIN_SIZE) - 1 |
428 | | * table[0] = mont(x^0) = mont(1) |
429 | | * table[1] = mont(x^1) = mont(x) |
430 | | */ |
431 | 0 | red_X[0 * red_digits] = 1; |
432 | 0 | red_X[1 * red_digits] = 1; |
433 | 0 | damm(&red_table[0 * 2 * red_digits], (const BN_ULONG*)red_X, rr, m, k0); |
434 | 0 | damm(&red_table[1 * 2 * red_digits], base, rr, m, k0); |
435 | |
|
436 | 0 | for (idx = 1; idx < (int)((1U << exp_win_size) / 2); idx++) { |
437 | 0 | DAMS(&red_table[(2 * idx + 0) * 2 * red_digits], |
438 | 0 | &red_table[(1 * idx) * 2 * red_digits], m, k0); |
439 | 0 | damm(&red_table[(2 * idx + 1) * 2 * red_digits], |
440 | 0 | &red_table[(2 * idx) * 2 * red_digits], |
441 | 0 | &red_table[1 * 2 * red_digits], m, k0); |
442 | 0 | } |
443 | | |
444 | | /* Copy and expand exponents */ |
445 | 0 | memcpy(&expz[0 * (exp_digits + 1)], exp[0], exp_digits * sizeof(BN_ULONG)); |
446 | 0 | expz[1 * (exp_digits + 1) - 1] = 0; |
447 | 0 | memcpy(&expz[1 * (exp_digits + 1)], exp[1], exp_digits * sizeof(BN_ULONG)); |
448 | 0 | expz[2 * (exp_digits + 1) - 1] = 0; |
449 | | |
450 | | /* Exponentiation */ |
451 | 0 | { |
452 | 0 | const int rem = modulus_bitsize % exp_win_size; |
453 | 0 | const BN_ULONG table_idx_mask = exp_win_mask; |
454 | |
|
455 | 0 | int exp_bit_no = modulus_bitsize - rem; |
456 | 0 | int exp_chunk_no = exp_bit_no / 64; |
457 | 0 | int exp_chunk_shift = exp_bit_no % 64; |
458 | |
|
459 | 0 | BN_ULONG red_table_idx_0, red_table_idx_1; |
460 | | |
461 | | /* |
462 | | * If rem == 0, then |
463 | | * exp_bit_no = modulus_bitsize - exp_win_size |
464 | | * However, this isn't possible because rem is { 1024, 1536, 2048 } % 5 |
465 | | * which is { 4, 1, 3 } respectively. |
466 | | * |
467 | | * If this assertion ever fails the fix above is easy. |
468 | | */ |
469 | 0 | OPENSSL_assert(rem != 0); |
470 | | |
471 | | /* Process 1-st exp window - just init result */ |
472 | 0 | red_table_idx_0 = expz[exp_chunk_no + 0 * (exp_digits + 1)]; |
473 | 0 | red_table_idx_1 = expz[exp_chunk_no + 1 * (exp_digits + 1)]; |
474 | | |
475 | | /* |
476 | | * The function operates with fixed moduli sizes divisible by 64, |
477 | | * thus table index here is always in supported range [0, EXP_WIN_SIZE). |
478 | | */ |
479 | 0 | red_table_idx_0 >>= exp_chunk_shift; |
480 | 0 | red_table_idx_1 >>= exp_chunk_shift; |
481 | |
|
482 | 0 | extract(&red_Y[0 * red_digits], (const BN_ULONG*)red_table, (int)red_table_idx_0, (int)red_table_idx_1); |
483 | | |
484 | | /* Process other exp windows */ |
485 | 0 | for (exp_bit_no -= exp_win_size; exp_bit_no >= 0; exp_bit_no -= exp_win_size) { |
486 | | /* Extract pre-computed multiplier from the table */ |
487 | 0 | { |
488 | 0 | BN_ULONG T; |
489 | |
|
490 | 0 | exp_chunk_no = exp_bit_no / 64; |
491 | 0 | exp_chunk_shift = exp_bit_no % 64; |
492 | 0 | { |
493 | 0 | red_table_idx_0 = expz[exp_chunk_no + 0 * (exp_digits + 1)]; |
494 | 0 | T = expz[exp_chunk_no + 1 + 0 * (exp_digits + 1)]; |
495 | |
|
496 | 0 | red_table_idx_0 >>= exp_chunk_shift; |
497 | | /* |
498 | | * Get additional bits from then next quadword |
499 | | * when 64-bit boundaries are crossed. |
500 | | */ |
501 | 0 | if (exp_chunk_shift > 64 - exp_win_size) { |
502 | 0 | T <<= (64 - exp_chunk_shift); |
503 | 0 | red_table_idx_0 ^= T; |
504 | 0 | } |
505 | 0 | red_table_idx_0 &= table_idx_mask; |
506 | 0 | } |
507 | 0 | { |
508 | 0 | red_table_idx_1 = expz[exp_chunk_no + 1 * (exp_digits + 1)]; |
509 | 0 | T = expz[exp_chunk_no + 1 + 1 * (exp_digits + 1)]; |
510 | |
|
511 | 0 | red_table_idx_1 >>= exp_chunk_shift; |
512 | | /* |
513 | | * Get additional bits from then next quadword |
514 | | * when 64-bit boundaries are crossed. |
515 | | */ |
516 | 0 | if (exp_chunk_shift > 64 - exp_win_size) { |
517 | 0 | T <<= (64 - exp_chunk_shift); |
518 | 0 | red_table_idx_1 ^= T; |
519 | 0 | } |
520 | 0 | red_table_idx_1 &= table_idx_mask; |
521 | 0 | } |
522 | |
|
523 | 0 | extract(&red_X[0 * red_digits], (const BN_ULONG*)red_table, (int)red_table_idx_0, (int)red_table_idx_1); |
524 | 0 | } |
525 | | |
526 | | /* Series of squaring */ |
527 | 0 | DAMS((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, m, k0); |
528 | 0 | DAMS((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, m, k0); |
529 | 0 | DAMS((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, m, k0); |
530 | 0 | DAMS((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, m, k0); |
531 | 0 | DAMS((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, m, k0); |
532 | |
|
533 | 0 | damm((BN_ULONG*)red_Y, (const BN_ULONG*)red_Y, (const BN_ULONG*)red_X, m, k0); |
534 | 0 | } |
535 | 0 | } |
536 | | |
537 | | /* |
538 | | * |
539 | | * NB: After the last AMM of exponentiation in Montgomery domain, the result |
540 | | * may be (modulus_bitsize + 1), but the conversion out of Montgomery domain |
541 | | * performs an AMM(x,1) which guarantees that the final result is less than |
542 | | * |m|, so no conditional subtraction is needed here. See [1] for details. |
543 | | * |
544 | | * [1] Gueron, S. Efficient software implementations of modular exponentiation. |
545 | | * DOI: 10.1007/s13389-012-0031-5 |
546 | | */ |
547 | | |
548 | | /* Convert result back in regular 2^52 domain */ |
549 | 0 | memset(red_X, 0, 2 * red_digits * sizeof(BN_ULONG)); |
550 | 0 | red_X[0 * red_digits] = 1; |
551 | 0 | red_X[1 * red_digits] = 1; |
552 | 0 | damm(out, (const BN_ULONG*)red_Y, (const BN_ULONG*)red_X, m, k0); |
553 | |
|
554 | 0 | ret = 1; |
555 | |
|
556 | 0 | err: |
557 | 0 | if (storage != NULL) { |
558 | | /* Clear whole storage */ |
559 | 0 | OPENSSL_cleanse(storage, storage_len_bytes); |
560 | 0 | OPENSSL_free(storage); |
561 | 0 | } |
562 | |
|
563 | 0 | #undef DAMS |
564 | 0 | return ret; |
565 | 0 | } |
566 | | |
567 | | static ossl_inline uint64_t get_digit(const uint8_t *in, int in_len) |
568 | 0 | { |
569 | 0 | uint64_t digit = 0; |
570 | |
|
571 | 0 | assert(in != NULL); |
572 | 0 | assert(in_len <= 8); |
573 | | |
574 | 0 | for (; in_len > 0; in_len--) { |
575 | 0 | digit <<= 8; |
576 | 0 | digit += (uint64_t)(in[in_len - 1]); |
577 | 0 | } |
578 | 0 | return digit; |
579 | 0 | } |
580 | | |
581 | | /* |
582 | | * Convert array of words in regular (base=2^64) representation to array of |
583 | | * words in redundant (base=2^52) one. |
584 | | */ |
585 | | static void to_words52(BN_ULONG *out, int out_len, |
586 | | const BN_ULONG *in, int in_bitsize) |
587 | 0 | { |
588 | 0 | uint8_t *in_str = NULL; |
589 | |
|
590 | 0 | assert(out != NULL); |
591 | 0 | assert(in != NULL); |
592 | | /* Check destination buffer capacity */ |
593 | 0 | assert(out_len >= number_of_digits(in_bitsize, DIGIT_SIZE)); |
594 | | |
595 | 0 | in_str = (uint8_t *)in; |
596 | |
|
597 | 0 | for (; in_bitsize >= (2 * DIGIT_SIZE); in_bitsize -= (2 * DIGIT_SIZE), out += 2) { |
598 | 0 | uint64_t digit; |
599 | |
|
600 | 0 | memcpy(&digit, in_str, sizeof(digit)); |
601 | 0 | out[0] = digit & DIGIT_MASK; |
602 | 0 | in_str += 6; |
603 | 0 | memcpy(&digit, in_str, sizeof(digit)); |
604 | 0 | out[1] = (digit >> 4) & DIGIT_MASK; |
605 | 0 | in_str += 7; |
606 | 0 | out_len -= 2; |
607 | 0 | } |
608 | |
|
609 | 0 | if (in_bitsize > DIGIT_SIZE) { |
610 | 0 | uint64_t digit = get_digit(in_str, 7); |
611 | |
|
612 | 0 | out[0] = digit & DIGIT_MASK; |
613 | 0 | in_str += 6; |
614 | 0 | in_bitsize -= DIGIT_SIZE; |
615 | 0 | digit = get_digit(in_str, BITS2WORD8_SIZE(in_bitsize)); |
616 | 0 | out[1] = digit >> 4; |
617 | 0 | out += 2; |
618 | 0 | out_len -= 2; |
619 | 0 | } else if (in_bitsize > 0) { |
620 | 0 | out[0] = get_digit(in_str, BITS2WORD8_SIZE(in_bitsize)); |
621 | 0 | out++; |
622 | 0 | out_len--; |
623 | 0 | } |
624 | |
|
625 | 0 | memset(out, 0, out_len * sizeof(BN_ULONG)); |
626 | 0 | } |
627 | | |
628 | | static ossl_inline void put_digit(uint8_t *out, int out_len, uint64_t digit) |
629 | 0 | { |
630 | 0 | assert(out != NULL); |
631 | 0 | assert(out_len <= 8); |
632 | | |
633 | 0 | for (; out_len > 0; out_len--) { |
634 | 0 | *out++ = (uint8_t)(digit & 0xFF); |
635 | 0 | digit >>= 8; |
636 | 0 | } |
637 | 0 | } |
638 | | |
639 | | /* |
640 | | * Convert array of words in redundant (base=2^52) representation to array of |
641 | | * words in regular (base=2^64) one. |
642 | | */ |
643 | | static void from_words52(BN_ULONG *out, int out_bitsize, const BN_ULONG *in) |
644 | 0 | { |
645 | 0 | int i; |
646 | 0 | int out_len = BITS2WORD64_SIZE(out_bitsize); |
647 | |
|
648 | 0 | assert(out != NULL); |
649 | 0 | assert(in != NULL); |
650 | | |
651 | 0 | for (i = 0; i < out_len; i++) |
652 | 0 | out[i] = 0; |
653 | |
|
654 | 0 | { |
655 | 0 | uint8_t *out_str = (uint8_t *)out; |
656 | |
|
657 | 0 | for (; out_bitsize >= (2 * DIGIT_SIZE); |
658 | 0 | out_bitsize -= (2 * DIGIT_SIZE), in += 2) { |
659 | 0 | uint64_t digit; |
660 | |
|
661 | 0 | digit = in[0]; |
662 | 0 | memcpy(out_str, &digit, sizeof(digit)); |
663 | 0 | out_str += 6; |
664 | 0 | digit = digit >> 48 | in[1] << 4; |
665 | 0 | memcpy(out_str, &digit, sizeof(digit)); |
666 | 0 | out_str += 7; |
667 | 0 | } |
668 | |
|
669 | 0 | if (out_bitsize > DIGIT_SIZE) { |
670 | 0 | put_digit(out_str, 7, in[0]); |
671 | 0 | out_str += 6; |
672 | 0 | out_bitsize -= DIGIT_SIZE; |
673 | 0 | put_digit(out_str, BITS2WORD8_SIZE(out_bitsize), |
674 | 0 | (in[1] << 4 | in[0] >> 48)); |
675 | 0 | } else if (out_bitsize) { |
676 | 0 | put_digit(out_str, BITS2WORD8_SIZE(out_bitsize), in[0]); |
677 | 0 | } |
678 | 0 | } |
679 | 0 | } |
680 | | |
681 | | /* |
682 | | * Set bit at index |idx| in the words array |a|. |
683 | | * It does not do any boundaries checks, make sure the index is valid before |
684 | | * calling the function. |
685 | | */ |
686 | | static ossl_inline void set_bit(BN_ULONG *a, int idx) |
687 | 0 | { |
688 | 0 | assert(a != NULL); |
689 | | |
690 | 0 | { |
691 | 0 | int i, j; |
692 | |
|
693 | 0 | i = idx / BN_BITS2; |
694 | 0 | j = idx % BN_BITS2; |
695 | 0 | a[i] |= (((BN_ULONG)1) << j); |
696 | 0 | } |
697 | 0 | } |
698 | | |
699 | | #endif |