1 |
747 |
jeremybenn |
// Copyright 2012 The Go Authors. All rights reserved.
|
2 |
|
|
// Use of this source code is governed by a BSD-style
|
3 |
|
|
// license that can be found in the LICENSE file.
|
4 |
|
|
|
5 |
|
|
package elliptic
|
6 |
|
|
|
7 |
|
|
// This is a constant-time, 32-bit implementation of P224. See FIPS 186-3,
|
8 |
|
|
// section D.2.2.
|
9 |
|
|
//
|
10 |
|
|
// See http://www.imperialviolet.org/2010/12/04/ecc.html ([1]) for background.
|
11 |
|
|
|
12 |
|
|
import (
|
13 |
|
|
"math/big"
|
14 |
|
|
)
|
15 |
|
|
|
16 |
|
|
var p224 p224Curve
|
17 |
|
|
|
18 |
|
|
type p224Curve struct {
|
19 |
|
|
*CurveParams
|
20 |
|
|
gx, gy, b p224FieldElement
|
21 |
|
|
}
|
22 |
|
|
|
23 |
|
|
func initP224() {
|
24 |
|
|
// See FIPS 186-3, section D.2.2
|
25 |
|
|
p224.CurveParams = new(CurveParams)
|
26 |
|
|
p224.P, _ = new(big.Int).SetString("26959946667150639794667015087019630673557916260026308143510066298881", 10)
|
27 |
|
|
p224.N, _ = new(big.Int).SetString("26959946667150639794667015087019625940457807714424391721682722368061", 10)
|
28 |
|
|
p224.B, _ = new(big.Int).SetString("b4050a850c04b3abf54132565044b0b7d7bfd8ba270b39432355ffb4", 16)
|
29 |
|
|
p224.Gx, _ = new(big.Int).SetString("b70e0cbd6bb4bf7f321390b94a03c1d356c21122343280d6115c1d21", 16)
|
30 |
|
|
p224.Gy, _ = new(big.Int).SetString("bd376388b5f723fb4c22dfe6cd4375a05a07476444d5819985007e34", 16)
|
31 |
|
|
p224.BitSize = 224
|
32 |
|
|
|
33 |
|
|
p224FromBig(&p224.gx, p224.Gx)
|
34 |
|
|
p224FromBig(&p224.gy, p224.Gy)
|
35 |
|
|
p224FromBig(&p224.b, p224.B)
|
36 |
|
|
}
|
37 |
|
|
|
38 |
|
|
// P224 returns a Curve which implements P-224 (see FIPS 186-3, section D.2.2)
|
39 |
|
|
func P224() Curve {
|
40 |
|
|
initonce.Do(initAll)
|
41 |
|
|
return p224
|
42 |
|
|
}
|
43 |
|
|
|
44 |
|
|
func (curve p224Curve) Params() *CurveParams {
|
45 |
|
|
return curve.CurveParams
|
46 |
|
|
}
|
47 |
|
|
|
48 |
|
|
func (curve p224Curve) IsOnCurve(bigX, bigY *big.Int) bool {
|
49 |
|
|
var x, y p224FieldElement
|
50 |
|
|
p224FromBig(&x, bigX)
|
51 |
|
|
p224FromBig(&y, bigY)
|
52 |
|
|
|
53 |
|
|
// y² = x³ - 3x + b
|
54 |
|
|
var tmp p224LargeFieldElement
|
55 |
|
|
var x3 p224FieldElement
|
56 |
|
|
p224Square(&x3, &x, &tmp)
|
57 |
|
|
p224Mul(&x3, &x3, &x, &tmp)
|
58 |
|
|
|
59 |
|
|
for i := 0; i < 8; i++ {
|
60 |
|
|
x[i] *= 3
|
61 |
|
|
}
|
62 |
|
|
p224Sub(&x3, &x3, &x)
|
63 |
|
|
p224Reduce(&x3)
|
64 |
|
|
p224Add(&x3, &x3, &curve.b)
|
65 |
|
|
p224Contract(&x3, &x3)
|
66 |
|
|
|
67 |
|
|
p224Square(&y, &y, &tmp)
|
68 |
|
|
p224Contract(&y, &y)
|
69 |
|
|
|
70 |
|
|
for i := 0; i < 8; i++ {
|
71 |
|
|
if y[i] != x3[i] {
|
72 |
|
|
return false
|
73 |
|
|
}
|
74 |
|
|
}
|
75 |
|
|
return true
|
76 |
|
|
}
|
77 |
|
|
|
78 |
|
|
func (p224Curve) Add(bigX1, bigY1, bigX2, bigY2 *big.Int) (x, y *big.Int) {
|
79 |
|
|
var x1, y1, z1, x2, y2, z2, x3, y3, z3 p224FieldElement
|
80 |
|
|
|
81 |
|
|
p224FromBig(&x1, bigX1)
|
82 |
|
|
p224FromBig(&y1, bigY1)
|
83 |
|
|
z1[0] = 1
|
84 |
|
|
p224FromBig(&x2, bigX2)
|
85 |
|
|
p224FromBig(&y2, bigY2)
|
86 |
|
|
z2[0] = 1
|
87 |
|
|
|
88 |
|
|
p224AddJacobian(&x3, &y3, &z3, &x1, &y1, &z1, &x2, &y2, &z2)
|
89 |
|
|
return p224ToAffine(&x3, &y3, &z3)
|
90 |
|
|
}
|
91 |
|
|
|
92 |
|
|
func (p224Curve) Double(bigX1, bigY1 *big.Int) (x, y *big.Int) {
|
93 |
|
|
var x1, y1, z1, x2, y2, z2 p224FieldElement
|
94 |
|
|
|
95 |
|
|
p224FromBig(&x1, bigX1)
|
96 |
|
|
p224FromBig(&y1, bigY1)
|
97 |
|
|
z1[0] = 1
|
98 |
|
|
|
99 |
|
|
p224DoubleJacobian(&x2, &y2, &z2, &x1, &y1, &z1)
|
100 |
|
|
return p224ToAffine(&x2, &y2, &z2)
|
101 |
|
|
}
|
102 |
|
|
|
103 |
|
|
func (p224Curve) ScalarMult(bigX1, bigY1 *big.Int, scalar []byte) (x, y *big.Int) {
|
104 |
|
|
var x1, y1, z1, x2, y2, z2 p224FieldElement
|
105 |
|
|
|
106 |
|
|
p224FromBig(&x1, bigX1)
|
107 |
|
|
p224FromBig(&y1, bigY1)
|
108 |
|
|
z1[0] = 1
|
109 |
|
|
|
110 |
|
|
p224ScalarMult(&x2, &y2, &z2, &x1, &y1, &z1, scalar)
|
111 |
|
|
return p224ToAffine(&x2, &y2, &z2)
|
112 |
|
|
}
|
113 |
|
|
|
114 |
|
|
func (curve p224Curve) ScalarBaseMult(scalar []byte) (x, y *big.Int) {
|
115 |
|
|
var z1, x2, y2, z2 p224FieldElement
|
116 |
|
|
|
117 |
|
|
z1[0] = 1
|
118 |
|
|
p224ScalarMult(&x2, &y2, &z2, &curve.gx, &curve.gy, &z1, scalar)
|
119 |
|
|
return p224ToAffine(&x2, &y2, &z2)
|
120 |
|
|
}
|
121 |
|
|
|
122 |
|
|
// Field element functions.
|
123 |
|
|
//
|
124 |
|
|
// The field that we're dealing with is ℤ/pℤ where p = 2**224 - 2**96 + 1.
|
125 |
|
|
//
|
126 |
|
|
// Field elements are represented by a FieldElement, which is a typedef to an
|
127 |
|
|
// array of 8 uint32's. The value of a FieldElement, a, is:
|
128 |
|
|
// a[0] + 2**28·a[1] + 2**56·a[1] + ... + 2**196·a[7]
|
129 |
|
|
//
|
130 |
|
|
// Using 28-bit limbs means that there's only 4 bits of headroom, which is less
|
131 |
|
|
// than we would really like. But it has the useful feature that we hit 2**224
|
132 |
|
|
// exactly, making the reflections during a reduce much nicer.
|
133 |
|
|
type p224FieldElement [8]uint32
|
134 |
|
|
|
135 |
|
|
// p224Add computes *out = a+b
|
136 |
|
|
//
|
137 |
|
|
// a[i] + b[i] < 2**32
|
138 |
|
|
func p224Add(out, a, b *p224FieldElement) {
|
139 |
|
|
for i := 0; i < 8; i++ {
|
140 |
|
|
out[i] = a[i] + b[i]
|
141 |
|
|
}
|
142 |
|
|
}
|
143 |
|
|
|
144 |
|
|
const two31p3 = 1<<31 + 1<<3
|
145 |
|
|
const two31m3 = 1<<31 - 1<<3
|
146 |
|
|
const two31m15m3 = 1<<31 - 1<<15 - 1<<3
|
147 |
|
|
|
148 |
|
|
// p224ZeroModP31 is 0 mod p where bit 31 is set in all limbs so that we can
|
149 |
|
|
// subtract smaller amounts without underflow. See the section "Subtraction" in
|
150 |
|
|
// [1] for reasoning.
|
151 |
|
|
var p224ZeroModP31 = []uint32{two31p3, two31m3, two31m3, two31m15m3, two31m3, two31m3, two31m3, two31m3}
|
152 |
|
|
|
153 |
|
|
// p224Sub computes *out = a-b
|
154 |
|
|
//
|
155 |
|
|
// a[i], b[i] < 2**30
|
156 |
|
|
// out[i] < 2**32
|
157 |
|
|
func p224Sub(out, a, b *p224FieldElement) {
|
158 |
|
|
for i := 0; i < 8; i++ {
|
159 |
|
|
out[i] = a[i] + p224ZeroModP31[i] - b[i]
|
160 |
|
|
}
|
161 |
|
|
}
|
162 |
|
|
|
163 |
|
|
// LargeFieldElement also represents an element of the field. The limbs are
|
164 |
|
|
// still spaced 28-bits apart and in little-endian order. So the limbs are at
|
165 |
|
|
// 0, 28, 56, ..., 392 bits, each 64-bits wide.
|
166 |
|
|
type p224LargeFieldElement [15]uint64
|
167 |
|
|
|
168 |
|
|
const two63p35 = 1<<63 + 1<<35
|
169 |
|
|
const two63m35 = 1<<63 - 1<<35
|
170 |
|
|
const two63m35m19 = 1<<63 - 1<<35 - 1<<19
|
171 |
|
|
|
172 |
|
|
// p224ZeroModP63 is 0 mod p where bit 63 is set in all limbs. See the section
|
173 |
|
|
// "Subtraction" in [1] for why.
|
174 |
|
|
var p224ZeroModP63 = [8]uint64{two63p35, two63m35, two63m35, two63m35, two63m35m19, two63m35, two63m35, two63m35}
|
175 |
|
|
|
176 |
|
|
const bottom12Bits = 0xfff
|
177 |
|
|
const bottom28Bits = 0xfffffff
|
178 |
|
|
|
179 |
|
|
// p224Mul computes *out = a*b
|
180 |
|
|
//
|
181 |
|
|
// a[i] < 2**29, b[i] < 2**30 (or vice versa)
|
182 |
|
|
// out[i] < 2**29
|
183 |
|
|
func p224Mul(out, a, b *p224FieldElement, tmp *p224LargeFieldElement) {
|
184 |
|
|
for i := 0; i < 15; i++ {
|
185 |
|
|
tmp[i] = 0
|
186 |
|
|
}
|
187 |
|
|
|
188 |
|
|
for i := 0; i < 8; i++ {
|
189 |
|
|
for j := 0; j < 8; j++ {
|
190 |
|
|
tmp[i+j] += uint64(a[i]) * uint64(b[j])
|
191 |
|
|
}
|
192 |
|
|
}
|
193 |
|
|
|
194 |
|
|
p224ReduceLarge(out, tmp)
|
195 |
|
|
}
|
196 |
|
|
|
197 |
|
|
// Square computes *out = a*a
|
198 |
|
|
//
|
199 |
|
|
// a[i] < 2**29
|
200 |
|
|
// out[i] < 2**29
|
201 |
|
|
func p224Square(out, a *p224FieldElement, tmp *p224LargeFieldElement) {
|
202 |
|
|
for i := 0; i < 15; i++ {
|
203 |
|
|
tmp[i] = 0
|
204 |
|
|
}
|
205 |
|
|
|
206 |
|
|
for i := 0; i < 8; i++ {
|
207 |
|
|
for j := 0; j <= i; j++ {
|
208 |
|
|
r := uint64(a[i]) * uint64(a[j])
|
209 |
|
|
if i == j {
|
210 |
|
|
tmp[i+j] += r
|
211 |
|
|
} else {
|
212 |
|
|
tmp[i+j] += r << 1
|
213 |
|
|
}
|
214 |
|
|
}
|
215 |
|
|
}
|
216 |
|
|
|
217 |
|
|
p224ReduceLarge(out, tmp)
|
218 |
|
|
}
|
219 |
|
|
|
220 |
|
|
// ReduceLarge converts a p224LargeFieldElement to a p224FieldElement.
|
221 |
|
|
//
|
222 |
|
|
// in[i] < 2**62
|
223 |
|
|
func p224ReduceLarge(out *p224FieldElement, in *p224LargeFieldElement) {
|
224 |
|
|
for i := 0; i < 8; i++ {
|
225 |
|
|
in[i] += p224ZeroModP63[i]
|
226 |
|
|
}
|
227 |
|
|
|
228 |
|
|
// Eliminate the coefficients at 2**224 and greater.
|
229 |
|
|
for i := 14; i >= 8; i-- {
|
230 |
|
|
in[i-8] -= in[i]
|
231 |
|
|
in[i-5] += (in[i] & 0xffff) << 12
|
232 |
|
|
in[i-4] += in[i] >> 16
|
233 |
|
|
}
|
234 |
|
|
in[8] = 0
|
235 |
|
|
// in[0..8] < 2**64
|
236 |
|
|
|
237 |
|
|
// As the values become small enough, we start to store them in |out|
|
238 |
|
|
// and use 32-bit operations.
|
239 |
|
|
for i := 1; i < 8; i++ {
|
240 |
|
|
in[i+1] += in[i] >> 28
|
241 |
|
|
out[i] = uint32(in[i] & bottom28Bits)
|
242 |
|
|
}
|
243 |
|
|
in[0] -= in[8]
|
244 |
|
|
out[3] += uint32(in[8]&0xffff) << 12
|
245 |
|
|
out[4] += uint32(in[8] >> 16)
|
246 |
|
|
// in[0] < 2**64
|
247 |
|
|
// out[3] < 2**29
|
248 |
|
|
// out[4] < 2**29
|
249 |
|
|
// out[1,2,5..7] < 2**28
|
250 |
|
|
|
251 |
|
|
out[0] = uint32(in[0] & bottom28Bits)
|
252 |
|
|
out[1] += uint32((in[0] >> 28) & bottom28Bits)
|
253 |
|
|
out[2] += uint32(in[0] >> 56)
|
254 |
|
|
// out[0] < 2**28
|
255 |
|
|
// out[1..4] < 2**29
|
256 |
|
|
// out[5..7] < 2**28
|
257 |
|
|
}
|
258 |
|
|
|
259 |
|
|
// Reduce reduces the coefficients of a to smaller bounds.
|
260 |
|
|
//
|
261 |
|
|
// On entry: a[i] < 2**31 + 2**30
|
262 |
|
|
// On exit: a[i] < 2**29
|
263 |
|
|
func p224Reduce(a *p224FieldElement) {
|
264 |
|
|
for i := 0; i < 7; i++ {
|
265 |
|
|
a[i+1] += a[i] >> 28
|
266 |
|
|
a[i] &= bottom28Bits
|
267 |
|
|
}
|
268 |
|
|
top := a[7] >> 28
|
269 |
|
|
a[7] &= bottom28Bits
|
270 |
|
|
|
271 |
|
|
// top < 2**4
|
272 |
|
|
mask := top
|
273 |
|
|
mask |= mask >> 2
|
274 |
|
|
mask |= mask >> 1
|
275 |
|
|
mask <<= 31
|
276 |
|
|
mask = uint32(int32(mask) >> 31)
|
277 |
|
|
// Mask is all ones if top != 0, all zero otherwise
|
278 |
|
|
|
279 |
|
|
a[0] -= top
|
280 |
|
|
a[3] += top << 12
|
281 |
|
|
|
282 |
|
|
// We may have just made a[0] negative but, if we did, then we must
|
283 |
|
|
// have added something to a[3], this it's > 2**12. Therefore we can
|
284 |
|
|
// carry down to a[0].
|
285 |
|
|
a[3] -= 1 & mask
|
286 |
|
|
a[2] += mask & (1<<28 - 1)
|
287 |
|
|
a[1] += mask & (1<<28 - 1)
|
288 |
|
|
a[0] += mask & (1 << 28)
|
289 |
|
|
}
|
290 |
|
|
|
291 |
|
|
// p224Invert calculates *out = in**-1 by computing in**(2**224 - 2**96 - 1),
|
292 |
|
|
// i.e. Fermat's little theorem.
|
293 |
|
|
func p224Invert(out, in *p224FieldElement) {
|
294 |
|
|
var f1, f2, f3, f4 p224FieldElement
|
295 |
|
|
var c p224LargeFieldElement
|
296 |
|
|
|
297 |
|
|
p224Square(&f1, in, &c) // 2
|
298 |
|
|
p224Mul(&f1, &f1, in, &c) // 2**2 - 1
|
299 |
|
|
p224Square(&f1, &f1, &c) // 2**3 - 2
|
300 |
|
|
p224Mul(&f1, &f1, in, &c) // 2**3 - 1
|
301 |
|
|
p224Square(&f2, &f1, &c) // 2**4 - 2
|
302 |
|
|
p224Square(&f2, &f2, &c) // 2**5 - 4
|
303 |
|
|
p224Square(&f2, &f2, &c) // 2**6 - 8
|
304 |
|
|
p224Mul(&f1, &f1, &f2, &c) // 2**6 - 1
|
305 |
|
|
p224Square(&f2, &f1, &c) // 2**7 - 2
|
306 |
|
|
for i := 0; i < 5; i++ { // 2**12 - 2**6
|
307 |
|
|
p224Square(&f2, &f2, &c)
|
308 |
|
|
}
|
309 |
|
|
p224Mul(&f2, &f2, &f1, &c) // 2**12 - 1
|
310 |
|
|
p224Square(&f3, &f2, &c) // 2**13 - 2
|
311 |
|
|
for i := 0; i < 11; i++ { // 2**24 - 2**12
|
312 |
|
|
p224Square(&f3, &f3, &c)
|
313 |
|
|
}
|
314 |
|
|
p224Mul(&f2, &f3, &f2, &c) // 2**24 - 1
|
315 |
|
|
p224Square(&f3, &f2, &c) // 2**25 - 2
|
316 |
|
|
for i := 0; i < 23; i++ { // 2**48 - 2**24
|
317 |
|
|
p224Square(&f3, &f3, &c)
|
318 |
|
|
}
|
319 |
|
|
p224Mul(&f3, &f3, &f2, &c) // 2**48 - 1
|
320 |
|
|
p224Square(&f4, &f3, &c) // 2**49 - 2
|
321 |
|
|
for i := 0; i < 47; i++ { // 2**96 - 2**48
|
322 |
|
|
p224Square(&f4, &f4, &c)
|
323 |
|
|
}
|
324 |
|
|
p224Mul(&f3, &f3, &f4, &c) // 2**96 - 1
|
325 |
|
|
p224Square(&f4, &f3, &c) // 2**97 - 2
|
326 |
|
|
for i := 0; i < 23; i++ { // 2**120 - 2**24
|
327 |
|
|
p224Square(&f4, &f4, &c)
|
328 |
|
|
}
|
329 |
|
|
p224Mul(&f2, &f4, &f2, &c) // 2**120 - 1
|
330 |
|
|
for i := 0; i < 6; i++ { // 2**126 - 2**6
|
331 |
|
|
p224Square(&f2, &f2, &c)
|
332 |
|
|
}
|
333 |
|
|
p224Mul(&f1, &f1, &f2, &c) // 2**126 - 1
|
334 |
|
|
p224Square(&f1, &f1, &c) // 2**127 - 2
|
335 |
|
|
p224Mul(&f1, &f1, in, &c) // 2**127 - 1
|
336 |
|
|
for i := 0; i < 97; i++ { // 2**224 - 2**97
|
337 |
|
|
p224Square(&f1, &f1, &c)
|
338 |
|
|
}
|
339 |
|
|
p224Mul(out, &f1, &f3, &c) // 2**224 - 2**96 - 1
|
340 |
|
|
}
|
341 |
|
|
|
342 |
|
|
// p224Contract converts a FieldElement to its unique, minimal form.
|
343 |
|
|
//
|
344 |
|
|
// On entry, in[i] < 2**29
|
345 |
|
|
// On exit, in[i] < 2**28
|
346 |
|
|
func p224Contract(out, in *p224FieldElement) {
|
347 |
|
|
copy(out[:], in[:])
|
348 |
|
|
|
349 |
|
|
for i := 0; i < 7; i++ {
|
350 |
|
|
out[i+1] += out[i] >> 28
|
351 |
|
|
out[i] &= bottom28Bits
|
352 |
|
|
}
|
353 |
|
|
top := out[7] >> 28
|
354 |
|
|
out[7] &= bottom28Bits
|
355 |
|
|
|
356 |
|
|
out[0] -= top
|
357 |
|
|
out[3] += top << 12
|
358 |
|
|
|
359 |
|
|
// We may just have made out[i] negative. So we carry down. If we made
|
360 |
|
|
// out[0] negative then we know that out[3] is sufficiently positive
|
361 |
|
|
// because we just added to it.
|
362 |
|
|
for i := 0; i < 3; i++ {
|
363 |
|
|
mask := uint32(int32(out[i]) >> 31)
|
364 |
|
|
out[i] += (1 << 28) & mask
|
365 |
|
|
out[i+1] -= 1 & mask
|
366 |
|
|
}
|
367 |
|
|
|
368 |
|
|
// We might have pushed out[3] over 2**28 so we perform another, partial,
|
369 |
|
|
// carry chain.
|
370 |
|
|
for i := 3; i < 7; i++ {
|
371 |
|
|
out[i+1] += out[i] >> 28
|
372 |
|
|
out[i] &= bottom28Bits
|
373 |
|
|
}
|
374 |
|
|
top = out[7] >> 28
|
375 |
|
|
out[7] &= bottom28Bits
|
376 |
|
|
|
377 |
|
|
// Eliminate top while maintaining the same value mod p.
|
378 |
|
|
out[0] -= top
|
379 |
|
|
out[3] += top << 12
|
380 |
|
|
|
381 |
|
|
// There are two cases to consider for out[3]:
|
382 |
|
|
// 1) The first time that we eliminated top, we didn't push out[3] over
|
383 |
|
|
// 2**28. In this case, the partial carry chain didn't change any values
|
384 |
|
|
// and top is zero.
|
385 |
|
|
// 2) We did push out[3] over 2**28 the first time that we eliminated top.
|
386 |
|
|
// The first value of top was in [0..16), therefore, prior to eliminating
|
387 |
|
|
// the first top, 0xfff1000 <= out[3] <= 0xfffffff. Therefore, after
|
388 |
|
|
// overflowing and being reduced by the second carry chain, out[3] <=
|
389 |
|
|
// 0xf000. Thus it cannot have overflowed when we eliminated top for the
|
390 |
|
|
// second time.
|
391 |
|
|
|
392 |
|
|
// Again, we may just have made out[0] negative, so do the same carry down.
|
393 |
|
|
// As before, if we made out[0] negative then we know that out[3] is
|
394 |
|
|
// sufficiently positive.
|
395 |
|
|
for i := 0; i < 3; i++ {
|
396 |
|
|
mask := uint32(int32(out[i]) >> 31)
|
397 |
|
|
out[i] += (1 << 28) & mask
|
398 |
|
|
out[i+1] -= 1 & mask
|
399 |
|
|
}
|
400 |
|
|
|
401 |
|
|
// Now we see if the value is >= p and, if so, subtract p.
|
402 |
|
|
|
403 |
|
|
// First we build a mask from the top four limbs, which must all be
|
404 |
|
|
// equal to bottom28Bits if the whole value is >= p. If top4AllOnes
|
405 |
|
|
// ends up with any zero bits in the bottom 28 bits, then this wasn't
|
406 |
|
|
// true.
|
407 |
|
|
top4AllOnes := uint32(0xffffffff)
|
408 |
|
|
for i := 4; i < 8; i++ {
|
409 |
|
|
top4AllOnes &= (out[i] & bottom28Bits) - 1
|
410 |
|
|
}
|
411 |
|
|
top4AllOnes |= 0xf0000000
|
412 |
|
|
// Now we replicate any zero bits to all the bits in top4AllOnes.
|
413 |
|
|
top4AllOnes &= top4AllOnes >> 16
|
414 |
|
|
top4AllOnes &= top4AllOnes >> 8
|
415 |
|
|
top4AllOnes &= top4AllOnes >> 4
|
416 |
|
|
top4AllOnes &= top4AllOnes >> 2
|
417 |
|
|
top4AllOnes &= top4AllOnes >> 1
|
418 |
|
|
top4AllOnes = uint32(int32(top4AllOnes<<31) >> 31)
|
419 |
|
|
|
420 |
|
|
// Now we test whether the bottom three limbs are non-zero.
|
421 |
|
|
bottom3NonZero := out[0] | out[1] | out[2]
|
422 |
|
|
bottom3NonZero |= bottom3NonZero >> 16
|
423 |
|
|
bottom3NonZero |= bottom3NonZero >> 8
|
424 |
|
|
bottom3NonZero |= bottom3NonZero >> 4
|
425 |
|
|
bottom3NonZero |= bottom3NonZero >> 2
|
426 |
|
|
bottom3NonZero |= bottom3NonZero >> 1
|
427 |
|
|
bottom3NonZero = uint32(int32(bottom3NonZero<<31) >> 31)
|
428 |
|
|
|
429 |
|
|
// Everything depends on the value of out[3].
|
430 |
|
|
// If it's > 0xffff000 and top4AllOnes != 0 then the whole value is >= p
|
431 |
|
|
// If it's = 0xffff000 and top4AllOnes != 0 and bottom3NonZero != 0,
|
432 |
|
|
// then the whole value is >= p
|
433 |
|
|
// If it's < 0xffff000, then the whole value is < p
|
434 |
|
|
n := out[3] - 0xffff000
|
435 |
|
|
out3Equal := n
|
436 |
|
|
out3Equal |= out3Equal >> 16
|
437 |
|
|
out3Equal |= out3Equal >> 8
|
438 |
|
|
out3Equal |= out3Equal >> 4
|
439 |
|
|
out3Equal |= out3Equal >> 2
|
440 |
|
|
out3Equal |= out3Equal >> 1
|
441 |
|
|
out3Equal = ^uint32(int32(out3Equal<<31) >> 31)
|
442 |
|
|
|
443 |
|
|
// If out[3] > 0xffff000 then n's MSB will be zero.
|
444 |
|
|
out3GT := ^uint32(int32(n<<31) >> 31)
|
445 |
|
|
|
446 |
|
|
mask := top4AllOnes & ((out3Equal & bottom3NonZero) | out3GT)
|
447 |
|
|
out[0] -= 1 & mask
|
448 |
|
|
out[3] -= 0xffff000 & mask
|
449 |
|
|
out[4] -= 0xfffffff & mask
|
450 |
|
|
out[5] -= 0xfffffff & mask
|
451 |
|
|
out[6] -= 0xfffffff & mask
|
452 |
|
|
out[7] -= 0xfffffff & mask
|
453 |
|
|
}
|
454 |
|
|
|
455 |
|
|
// Group element functions.
|
456 |
|
|
//
|
457 |
|
|
// These functions deal with group elements. The group is an elliptic curve
|
458 |
|
|
// group with a = -3 defined in FIPS 186-3, section D.2.2.
|
459 |
|
|
|
460 |
|
|
// p224AddJacobian computes *out = a+b where a != b.
|
461 |
|
|
func p224AddJacobian(x3, y3, z3, x1, y1, z1, x2, y2, z2 *p224FieldElement) {
|
462 |
|
|
// See http://hyperelliptic.org/EFD/g1p/auto-shortw-jacobian-3.html#addition-p224Add-2007-bl
|
463 |
|
|
var z1z1, z2z2, u1, u2, s1, s2, h, i, j, r, v p224FieldElement
|
464 |
|
|
var c p224LargeFieldElement
|
465 |
|
|
|
466 |
|
|
// Z1Z1 = Z1²
|
467 |
|
|
p224Square(&z1z1, z1, &c)
|
468 |
|
|
// Z2Z2 = Z2²
|
469 |
|
|
p224Square(&z2z2, z2, &c)
|
470 |
|
|
// U1 = X1*Z2Z2
|
471 |
|
|
p224Mul(&u1, x1, &z2z2, &c)
|
472 |
|
|
// U2 = X2*Z1Z1
|
473 |
|
|
p224Mul(&u2, x2, &z1z1, &c)
|
474 |
|
|
// S1 = Y1*Z2*Z2Z2
|
475 |
|
|
p224Mul(&s1, z2, &z2z2, &c)
|
476 |
|
|
p224Mul(&s1, y1, &s1, &c)
|
477 |
|
|
// S2 = Y2*Z1*Z1Z1
|
478 |
|
|
p224Mul(&s2, z1, &z1z1, &c)
|
479 |
|
|
p224Mul(&s2, y2, &s2, &c)
|
480 |
|
|
// H = U2-U1
|
481 |
|
|
p224Sub(&h, &u2, &u1)
|
482 |
|
|
p224Reduce(&h)
|
483 |
|
|
// I = (2*H)²
|
484 |
|
|
for j := 0; j < 8; j++ {
|
485 |
|
|
i[j] = h[j] << 1
|
486 |
|
|
}
|
487 |
|
|
p224Reduce(&i)
|
488 |
|
|
p224Square(&i, &i, &c)
|
489 |
|
|
// J = H*I
|
490 |
|
|
p224Mul(&j, &h, &i, &c)
|
491 |
|
|
// r = 2*(S2-S1)
|
492 |
|
|
p224Sub(&r, &s2, &s1)
|
493 |
|
|
p224Reduce(&r)
|
494 |
|
|
for i := 0; i < 8; i++ {
|
495 |
|
|
r[i] <<= 1
|
496 |
|
|
}
|
497 |
|
|
p224Reduce(&r)
|
498 |
|
|
// V = U1*I
|
499 |
|
|
p224Mul(&v, &u1, &i, &c)
|
500 |
|
|
// Z3 = ((Z1+Z2)²-Z1Z1-Z2Z2)*H
|
501 |
|
|
p224Add(&z1z1, &z1z1, &z2z2)
|
502 |
|
|
p224Add(&z2z2, z1, z2)
|
503 |
|
|
p224Reduce(&z2z2)
|
504 |
|
|
p224Square(&z2z2, &z2z2, &c)
|
505 |
|
|
p224Sub(z3, &z2z2, &z1z1)
|
506 |
|
|
p224Reduce(z3)
|
507 |
|
|
p224Mul(z3, z3, &h, &c)
|
508 |
|
|
// X3 = r²-J-2*V
|
509 |
|
|
for i := 0; i < 8; i++ {
|
510 |
|
|
z1z1[i] = v[i] << 1
|
511 |
|
|
}
|
512 |
|
|
p224Add(&z1z1, &j, &z1z1)
|
513 |
|
|
p224Reduce(&z1z1)
|
514 |
|
|
p224Square(x3, &r, &c)
|
515 |
|
|
p224Sub(x3, x3, &z1z1)
|
516 |
|
|
p224Reduce(x3)
|
517 |
|
|
// Y3 = r*(V-X3)-2*S1*J
|
518 |
|
|
for i := 0; i < 8; i++ {
|
519 |
|
|
s1[i] <<= 1
|
520 |
|
|
}
|
521 |
|
|
p224Mul(&s1, &s1, &j, &c)
|
522 |
|
|
p224Sub(&z1z1, &v, x3)
|
523 |
|
|
p224Reduce(&z1z1)
|
524 |
|
|
p224Mul(&z1z1, &z1z1, &r, &c)
|
525 |
|
|
p224Sub(y3, &z1z1, &s1)
|
526 |
|
|
p224Reduce(y3)
|
527 |
|
|
}
|
528 |
|
|
|
529 |
|
|
// p224DoubleJacobian computes *out = a+a.
|
530 |
|
|
func p224DoubleJacobian(x3, y3, z3, x1, y1, z1 *p224FieldElement) {
|
531 |
|
|
var delta, gamma, beta, alpha, t p224FieldElement
|
532 |
|
|
var c p224LargeFieldElement
|
533 |
|
|
|
534 |
|
|
p224Square(&delta, z1, &c)
|
535 |
|
|
p224Square(&gamma, y1, &c)
|
536 |
|
|
p224Mul(&beta, x1, &gamma, &c)
|
537 |
|
|
|
538 |
|
|
// alpha = 3*(X1-delta)*(X1+delta)
|
539 |
|
|
p224Add(&t, x1, &delta)
|
540 |
|
|
for i := 0; i < 8; i++ {
|
541 |
|
|
t[i] += t[i] << 1
|
542 |
|
|
}
|
543 |
|
|
p224Reduce(&t)
|
544 |
|
|
p224Sub(&alpha, x1, &delta)
|
545 |
|
|
p224Reduce(&alpha)
|
546 |
|
|
p224Mul(&alpha, &alpha, &t, &c)
|
547 |
|
|
|
548 |
|
|
// Z3 = (Y1+Z1)²-gamma-delta
|
549 |
|
|
p224Add(z3, y1, z1)
|
550 |
|
|
p224Reduce(z3)
|
551 |
|
|
p224Square(z3, z3, &c)
|
552 |
|
|
p224Sub(z3, z3, &gamma)
|
553 |
|
|
p224Reduce(z3)
|
554 |
|
|
p224Sub(z3, z3, &delta)
|
555 |
|
|
p224Reduce(z3)
|
556 |
|
|
|
557 |
|
|
// X3 = alpha²-8*beta
|
558 |
|
|
for i := 0; i < 8; i++ {
|
559 |
|
|
delta[i] = beta[i] << 3
|
560 |
|
|
}
|
561 |
|
|
p224Reduce(&delta)
|
562 |
|
|
p224Square(x3, &alpha, &c)
|
563 |
|
|
p224Sub(x3, x3, &delta)
|
564 |
|
|
p224Reduce(x3)
|
565 |
|
|
|
566 |
|
|
// Y3 = alpha*(4*beta-X3)-8*gamma²
|
567 |
|
|
for i := 0; i < 8; i++ {
|
568 |
|
|
beta[i] <<= 2
|
569 |
|
|
}
|
570 |
|
|
p224Sub(&beta, &beta, x3)
|
571 |
|
|
p224Reduce(&beta)
|
572 |
|
|
p224Square(&gamma, &gamma, &c)
|
573 |
|
|
for i := 0; i < 8; i++ {
|
574 |
|
|
gamma[i] <<= 3
|
575 |
|
|
}
|
576 |
|
|
p224Reduce(&gamma)
|
577 |
|
|
p224Mul(y3, &alpha, &beta, &c)
|
578 |
|
|
p224Sub(y3, y3, &gamma)
|
579 |
|
|
p224Reduce(y3)
|
580 |
|
|
}
|
581 |
|
|
|
582 |
|
|
// p224CopyConditional sets *out = *in iff the least-significant-bit of control
|
583 |
|
|
// is true, and it runs in constant time.
|
584 |
|
|
func p224CopyConditional(out, in *p224FieldElement, control uint32) {
|
585 |
|
|
control <<= 31
|
586 |
|
|
control = uint32(int32(control) >> 31)
|
587 |
|
|
|
588 |
|
|
for i := 0; i < 8; i++ {
|
589 |
|
|
out[i] ^= (out[i] ^ in[i]) & control
|
590 |
|
|
}
|
591 |
|
|
}
|
592 |
|
|
|
593 |
|
|
func p224ScalarMult(outX, outY, outZ, inX, inY, inZ *p224FieldElement, scalar []byte) {
|
594 |
|
|
var xx, yy, zz p224FieldElement
|
595 |
|
|
for i := 0; i < 8; i++ {
|
596 |
|
|
outZ[i] = 0
|
597 |
|
|
}
|
598 |
|
|
|
599 |
|
|
firstBit := uint32(1)
|
600 |
|
|
for _, byte := range scalar {
|
601 |
|
|
for bitNum := uint(0); bitNum < 8; bitNum++ {
|
602 |
|
|
p224DoubleJacobian(outX, outY, outZ, outX, outY, outZ)
|
603 |
|
|
bit := uint32((byte >> (7 - bitNum)) & 1)
|
604 |
|
|
p224AddJacobian(&xx, &yy, &zz, inX, inY, inZ, outX, outY, outZ)
|
605 |
|
|
p224CopyConditional(outX, inX, firstBit&bit)
|
606 |
|
|
p224CopyConditional(outY, inY, firstBit&bit)
|
607 |
|
|
p224CopyConditional(outZ, inZ, firstBit&bit)
|
608 |
|
|
p224CopyConditional(outX, &xx, ^firstBit&bit)
|
609 |
|
|
p224CopyConditional(outY, &yy, ^firstBit&bit)
|
610 |
|
|
p224CopyConditional(outZ, &zz, ^firstBit&bit)
|
611 |
|
|
firstBit = firstBit & ^bit
|
612 |
|
|
}
|
613 |
|
|
}
|
614 |
|
|
}
|
615 |
|
|
|
616 |
|
|
// p224ToAffine converts from Jacobian to affine form.
|
617 |
|
|
func p224ToAffine(x, y, z *p224FieldElement) (*big.Int, *big.Int) {
|
618 |
|
|
var zinv, zinvsq, outx, outy p224FieldElement
|
619 |
|
|
var tmp p224LargeFieldElement
|
620 |
|
|
|
621 |
|
|
isPointAtInfinity := true
|
622 |
|
|
for i := 0; i < 8; i++ {
|
623 |
|
|
if z[i] != 0 {
|
624 |
|
|
isPointAtInfinity = false
|
625 |
|
|
break
|
626 |
|
|
}
|
627 |
|
|
}
|
628 |
|
|
|
629 |
|
|
if isPointAtInfinity {
|
630 |
|
|
return nil, nil
|
631 |
|
|
}
|
632 |
|
|
|
633 |
|
|
p224Invert(&zinv, z)
|
634 |
|
|
p224Square(&zinvsq, &zinv, &tmp)
|
635 |
|
|
p224Mul(x, x, &zinvsq, &tmp)
|
636 |
|
|
p224Mul(&zinvsq, &zinvsq, &zinv, &tmp)
|
637 |
|
|
p224Mul(y, y, &zinvsq, &tmp)
|
638 |
|
|
|
639 |
|
|
p224Contract(&outx, x)
|
640 |
|
|
p224Contract(&outy, y)
|
641 |
|
|
return p224ToBig(&outx), p224ToBig(&outy)
|
642 |
|
|
}
|
643 |
|
|
|
644 |
|
|
// get28BitsFromEnd returns the least-significant 28 bits from buf>>shift,
|
645 |
|
|
// where buf is interpreted as a big-endian number.
|
646 |
|
|
func get28BitsFromEnd(buf []byte, shift uint) (uint32, []byte) {
|
647 |
|
|
var ret uint32
|
648 |
|
|
|
649 |
|
|
for i := uint(0); i < 4; i++ {
|
650 |
|
|
var b byte
|
651 |
|
|
if l := len(buf); l > 0 {
|
652 |
|
|
b = buf[l-1]
|
653 |
|
|
// We don't remove the byte if we're about to return and we're not
|
654 |
|
|
// reading all of it.
|
655 |
|
|
if i != 3 || shift == 4 {
|
656 |
|
|
buf = buf[:l-1]
|
657 |
|
|
}
|
658 |
|
|
}
|
659 |
|
|
ret |= uint32(b) << (8 * i) >> shift
|
660 |
|
|
}
|
661 |
|
|
ret &= bottom28Bits
|
662 |
|
|
return ret, buf
|
663 |
|
|
}
|
664 |
|
|
|
665 |
|
|
// p224FromBig sets *out = *in.
|
666 |
|
|
func p224FromBig(out *p224FieldElement, in *big.Int) {
|
667 |
|
|
bytes := in.Bytes()
|
668 |
|
|
out[0], bytes = get28BitsFromEnd(bytes, 0)
|
669 |
|
|
out[1], bytes = get28BitsFromEnd(bytes, 4)
|
670 |
|
|
out[2], bytes = get28BitsFromEnd(bytes, 0)
|
671 |
|
|
out[3], bytes = get28BitsFromEnd(bytes, 4)
|
672 |
|
|
out[4], bytes = get28BitsFromEnd(bytes, 0)
|
673 |
|
|
out[5], bytes = get28BitsFromEnd(bytes, 4)
|
674 |
|
|
out[6], bytes = get28BitsFromEnd(bytes, 0)
|
675 |
|
|
out[7], bytes = get28BitsFromEnd(bytes, 4)
|
676 |
|
|
}
|
677 |
|
|
|
678 |
|
|
// p224ToBig returns in as a big.Int.
|
679 |
|
|
func p224ToBig(in *p224FieldElement) *big.Int {
|
680 |
|
|
var buf [28]byte
|
681 |
|
|
buf[27] = byte(in[0])
|
682 |
|
|
buf[26] = byte(in[0] >> 8)
|
683 |
|
|
buf[25] = byte(in[0] >> 16)
|
684 |
|
|
buf[24] = byte(((in[0] >> 24) & 0x0f) | (in[1]<<4)&0xf0)
|
685 |
|
|
|
686 |
|
|
buf[23] = byte(in[1] >> 4)
|
687 |
|
|
buf[22] = byte(in[1] >> 12)
|
688 |
|
|
buf[21] = byte(in[1] >> 20)
|
689 |
|
|
|
690 |
|
|
buf[20] = byte(in[2])
|
691 |
|
|
buf[19] = byte(in[2] >> 8)
|
692 |
|
|
buf[18] = byte(in[2] >> 16)
|
693 |
|
|
buf[17] = byte(((in[2] >> 24) & 0x0f) | (in[3]<<4)&0xf0)
|
694 |
|
|
|
695 |
|
|
buf[16] = byte(in[3] >> 4)
|
696 |
|
|
buf[15] = byte(in[3] >> 12)
|
697 |
|
|
buf[14] = byte(in[3] >> 20)
|
698 |
|
|
|
699 |
|
|
buf[13] = byte(in[4])
|
700 |
|
|
buf[12] = byte(in[4] >> 8)
|
701 |
|
|
buf[11] = byte(in[4] >> 16)
|
702 |
|
|
buf[10] = byte(((in[4] >> 24) & 0x0f) | (in[5]<<4)&0xf0)
|
703 |
|
|
|
704 |
|
|
buf[9] = byte(in[5] >> 4)
|
705 |
|
|
buf[8] = byte(in[5] >> 12)
|
706 |
|
|
buf[7] = byte(in[5] >> 20)
|
707 |
|
|
|
708 |
|
|
buf[6] = byte(in[6])
|
709 |
|
|
buf[5] = byte(in[6] >> 8)
|
710 |
|
|
buf[4] = byte(in[6] >> 16)
|
711 |
|
|
buf[3] = byte(((in[6] >> 24) & 0x0f) | (in[7]<<4)&0xf0)
|
712 |
|
|
|
713 |
|
|
buf[2] = byte(in[7] >> 4)
|
714 |
|
|
buf[1] = byte(in[7] >> 12)
|
715 |
|
|
buf[0] = byte(in[7] >> 20)
|
716 |
|
|
|
717 |
|
|
return new(big.Int).SetBytes(buf[:])
|
718 |
|
|
}
|