yuezonghe | 824eb0c | 2024-06-27 02:32:26 -0700 | [diff] [blame] | 1 | /* |
| 2 | * Copyright 2011-2019 The OpenSSL Project Authors. All Rights Reserved. |
| 3 | * |
| 4 | * Licensed under the OpenSSL license (the "License"). You may not use |
| 5 | * this file except in compliance with the License. You can obtain a copy |
| 6 | * in the file LICENSE in the source distribution or at |
| 7 | * https://www.openssl.org/source/license.html |
| 8 | */ |
| 9 | |
| 10 | /* Copyright 2011 Google Inc. |
| 11 | * |
| 12 | * Licensed under the Apache License, Version 2.0 (the "License"); |
| 13 | * |
| 14 | * you may not use this file except in compliance with the License. |
| 15 | * You may obtain a copy of the License at |
| 16 | * |
| 17 | * http://www.apache.org/licenses/LICENSE-2.0 |
| 18 | * |
| 19 | * Unless required by applicable law or agreed to in writing, software |
| 20 | * distributed under the License is distributed on an "AS IS" BASIS, |
| 21 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 22 | * See the License for the specific language governing permissions and |
| 23 | * limitations under the License. |
| 24 | */ |
| 25 | |
| 26 | #include <openssl/opensslconf.h> |
| 27 | #ifdef OPENSSL_NO_EC_NISTP_64_GCC_128 |
| 28 | NON_EMPTY_TRANSLATION_UNIT |
| 29 | #else |
| 30 | |
| 31 | /* |
| 32 | * Common utility functions for ecp_nistp224.c, ecp_nistp256.c, ecp_nistp521.c. |
| 33 | */ |
| 34 | |
| 35 | # include <stddef.h> |
| 36 | # include "ec_local.h" |
| 37 | |
| 38 | /* |
| 39 | * Convert an array of points into affine coordinates. (If the point at |
| 40 | * infinity is found (Z = 0), it remains unchanged.) This function is |
| 41 | * essentially an equivalent to EC_POINTs_make_affine(), but works with the |
| 42 | * internal representation of points as used by ecp_nistp###.c rather than |
| 43 | * with (BIGNUM-based) EC_POINT data structures. point_array is the |
| 44 | * input/output buffer ('num' points in projective form, i.e. three |
| 45 | * coordinates each), based on an internal representation of field elements |
| 46 | * of size 'felem_size'. tmp_felems needs to point to a temporary array of |
| 47 | * 'num'+1 field elements for storage of intermediate values. |
| 48 | */ |
| 49 | void ec_GFp_nistp_points_make_affine_internal(size_t num, void *point_array, |
| 50 | size_t felem_size, |
| 51 | void *tmp_felems, |
| 52 | void (*felem_one) (void *out), |
| 53 | int (*felem_is_zero) (const void |
| 54 | *in), |
| 55 | void (*felem_assign) (void *out, |
| 56 | const void |
| 57 | *in), |
| 58 | void (*felem_square) (void *out, |
| 59 | const void |
| 60 | *in), |
| 61 | void (*felem_mul) (void *out, |
| 62 | const void |
| 63 | *in1, |
| 64 | const void |
| 65 | *in2), |
| 66 | void (*felem_inv) (void *out, |
| 67 | const void |
| 68 | *in), |
| 69 | void (*felem_contract) (void |
| 70 | *out, |
| 71 | const |
| 72 | void |
| 73 | *in)) |
| 74 | { |
| 75 | int i = 0; |
| 76 | |
| 77 | # define tmp_felem(I) (&((char *)tmp_felems)[(I) * felem_size]) |
| 78 | # define X(I) (&((char *)point_array)[3*(I) * felem_size]) |
| 79 | # define Y(I) (&((char *)point_array)[(3*(I) + 1) * felem_size]) |
| 80 | # define Z(I) (&((char *)point_array)[(3*(I) + 2) * felem_size]) |
| 81 | |
| 82 | if (!felem_is_zero(Z(0))) |
| 83 | felem_assign(tmp_felem(0), Z(0)); |
| 84 | else |
| 85 | felem_one(tmp_felem(0)); |
| 86 | for (i = 1; i < (int)num; i++) { |
| 87 | if (!felem_is_zero(Z(i))) |
| 88 | felem_mul(tmp_felem(i), tmp_felem(i - 1), Z(i)); |
| 89 | else |
| 90 | felem_assign(tmp_felem(i), tmp_felem(i - 1)); |
| 91 | } |
| 92 | /* |
| 93 | * Now each tmp_felem(i) is the product of Z(0) .. Z(i), skipping any |
| 94 | * zero-valued factors: if Z(i) = 0, we essentially pretend that Z(i) = 1 |
| 95 | */ |
| 96 | |
| 97 | felem_inv(tmp_felem(num - 1), tmp_felem(num - 1)); |
| 98 | for (i = num - 1; i >= 0; i--) { |
| 99 | if (i > 0) |
| 100 | /* |
| 101 | * tmp_felem(i-1) is the product of Z(0) .. Z(i-1), tmp_felem(i) |
| 102 | * is the inverse of the product of Z(0) .. Z(i) |
| 103 | */ |
| 104 | /* 1/Z(i) */ |
| 105 | felem_mul(tmp_felem(num), tmp_felem(i - 1), tmp_felem(i)); |
| 106 | else |
| 107 | felem_assign(tmp_felem(num), tmp_felem(0)); /* 1/Z(0) */ |
| 108 | |
| 109 | if (!felem_is_zero(Z(i))) { |
| 110 | if (i > 0) |
| 111 | /* |
| 112 | * For next iteration, replace tmp_felem(i-1) by its inverse |
| 113 | */ |
| 114 | felem_mul(tmp_felem(i - 1), tmp_felem(i), Z(i)); |
| 115 | |
| 116 | /* |
| 117 | * Convert point (X, Y, Z) into affine form (X/(Z^2), Y/(Z^3), 1) |
| 118 | */ |
| 119 | felem_square(Z(i), tmp_felem(num)); /* 1/(Z^2) */ |
| 120 | felem_mul(X(i), X(i), Z(i)); /* X/(Z^2) */ |
| 121 | felem_mul(Z(i), Z(i), tmp_felem(num)); /* 1/(Z^3) */ |
| 122 | felem_mul(Y(i), Y(i), Z(i)); /* Y/(Z^3) */ |
| 123 | felem_contract(X(i), X(i)); |
| 124 | felem_contract(Y(i), Y(i)); |
| 125 | felem_one(Z(i)); |
| 126 | } else { |
| 127 | if (i > 0) |
| 128 | /* |
| 129 | * For next iteration, replace tmp_felem(i-1) by its inverse |
| 130 | */ |
| 131 | felem_assign(tmp_felem(i - 1), tmp_felem(i)); |
| 132 | } |
| 133 | } |
| 134 | } |
| 135 | |
| 136 | /*- |
| 137 | * This function looks at 5+1 scalar bits (5 current, 1 adjacent less |
| 138 | * significant bit), and recodes them into a signed digit for use in fast point |
| 139 | * multiplication: the use of signed rather than unsigned digits means that |
| 140 | * fewer points need to be precomputed, given that point inversion is easy |
| 141 | * (a precomputed point dP makes -dP available as well). |
| 142 | * |
| 143 | * BACKGROUND: |
| 144 | * |
| 145 | * Signed digits for multiplication were introduced by Booth ("A signed binary |
| 146 | * multiplication technique", Quart. Journ. Mech. and Applied Math., vol. IV, |
| 147 | * pt. 2 (1951), pp. 236-240), in that case for multiplication of integers. |
| 148 | * Booth's original encoding did not generally improve the density of nonzero |
| 149 | * digits over the binary representation, and was merely meant to simplify the |
| 150 | * handling of signed factors given in two's complement; but it has since been |
| 151 | * shown to be the basis of various signed-digit representations that do have |
| 152 | * further advantages, including the wNAF, using the following general approach: |
| 153 | * |
| 154 | * (1) Given a binary representation |
| 155 | * |
| 156 | * b_k ... b_2 b_1 b_0, |
| 157 | * |
| 158 | * of a nonnegative integer (b_k in {0, 1}), rewrite it in digits 0, 1, -1 |
| 159 | * by using bit-wise subtraction as follows: |
| 160 | * |
| 161 | * b_k b_(k-1) ... b_2 b_1 b_0 |
| 162 | * - b_k ... b_3 b_2 b_1 b_0 |
| 163 | * ----------------------------------------- |
| 164 | * s_(k+1) s_k ... s_3 s_2 s_1 s_0 |
| 165 | * |
| 166 | * A left-shift followed by subtraction of the original value yields a new |
| 167 | * representation of the same value, using signed bits s_i = b_(i-1) - b_i. |
| 168 | * This representation from Booth's paper has since appeared in the |
| 169 | * literature under a variety of different names including "reversed binary |
| 170 | * form", "alternating greedy expansion", "mutual opposite form", and |
| 171 | * "sign-alternating {+-1}-representation". |
| 172 | * |
| 173 | * An interesting property is that among the nonzero bits, values 1 and -1 |
| 174 | * strictly alternate. |
| 175 | * |
| 176 | * (2) Various window schemes can be applied to the Booth representation of |
| 177 | * integers: for example, right-to-left sliding windows yield the wNAF |
| 178 | * (a signed-digit encoding independently discovered by various researchers |
| 179 | * in the 1990s), and left-to-right sliding windows yield a left-to-right |
| 180 | * equivalent of the wNAF (independently discovered by various researchers |
| 181 | * around 2004). |
| 182 | * |
| 183 | * To prevent leaking information through side channels in point multiplication, |
| 184 | * we need to recode the given integer into a regular pattern: sliding windows |
| 185 | * as in wNAFs won't do, we need their fixed-window equivalent -- which is a few |
| 186 | * decades older: we'll be using the so-called "modified Booth encoding" due to |
| 187 | * MacSorley ("High-speed arithmetic in binary computers", Proc. IRE, vol. 49 |
| 188 | * (1961), pp. 67-91), in a radix-2^5 setting. That is, we always combine five |
| 189 | * signed bits into a signed digit: |
| 190 | * |
| 191 | * s_(5j + 4) s_(5j + 3) s_(5j + 2) s_(5j + 1) s_(5j) |
| 192 | * |
| 193 | * The sign-alternating property implies that the resulting digit values are |
| 194 | * integers from -16 to 16. |
| 195 | * |
| 196 | * Of course, we don't actually need to compute the signed digits s_i as an |
| 197 | * intermediate step (that's just a nice way to see how this scheme relates |
| 198 | * to the wNAF): a direct computation obtains the recoded digit from the |
| 199 | * six bits b_(5j + 4) ... b_(5j - 1). |
| 200 | * |
| 201 | * This function takes those six bits as an integer (0 .. 63), writing the |
| 202 | * recoded digit to *sign (0 for positive, 1 for negative) and *digit (absolute |
| 203 | * value, in the range 0 .. 16). Note that this integer essentially provides |
| 204 | * the input bits "shifted to the left" by one position: for example, the input |
| 205 | * to compute the least significant recoded digit, given that there's no bit |
| 206 | * b_-1, has to be b_4 b_3 b_2 b_1 b_0 0. |
| 207 | * |
| 208 | */ |
| 209 | void ec_GFp_nistp_recode_scalar_bits(unsigned char *sign, |
| 210 | unsigned char *digit, unsigned char in) |
| 211 | { |
| 212 | unsigned char s, d; |
| 213 | |
| 214 | s = ~((in >> 5) - 1); /* sets all bits to MSB(in), 'in' seen as |
| 215 | * 6-bit value */ |
| 216 | d = (1 << 6) - in - 1; |
| 217 | d = (d & s) | (in & ~s); |
| 218 | d = (d >> 1) + (d & 1); |
| 219 | |
| 220 | *sign = s & 1; |
| 221 | *digit = d; |
| 222 | } |
| 223 | #endif |