From b75d8e421243371fa43f83b72ff68aa37342b84a Mon Sep 17 00:00:00 2001 From: "rtenneti@chromium.org" Date: Fri, 8 Mar 2013 23:40:42 +0000 Subject: Added Curve25519-donna changes. Added a wrapper class that implements the following API calls which for Curve25519. + ScalarMult to compute the shared key. + ScalarBaseMult to get public key. + ConvertToPrivateKey returns a private key from random bytes. Per agl/wtc, grabbed the rev 234205ff from the git repo (https://github.com/agl/curve25519-donna/tree/234205ff1ecaf6b3c1dc76798a462c4293f31fdb) and checked it in to crypto/ because that version has pure Google copyright. R=wtc@chromium.org,agl@chromium.org,rsleevi@chromium.org TEST=crypto unit tests Review URL: https://chromiumcodereview.appspot.com/12457004 git-svn-id: svn://svn.chromium.org/chrome/trunk/src@187074 0039d316-1c4b-4281-b951-d872f2087c98 --- crypto/curve25519-donna.c | 592 ++++++++++++++++++++++++++++++++++++++++++++++ 1 file changed, 592 insertions(+) create mode 100644 crypto/curve25519-donna.c (limited to 'crypto/curve25519-donna.c') diff --git a/crypto/curve25519-donna.c b/crypto/curve25519-donna.c new file mode 100644 index 0000000..f141ac0 --- /dev/null +++ b/crypto/curve25519-donna.c @@ -0,0 +1,592 @@ +// Copyright (c) 2013 The Chromium Authors. All rights reserved. +// Use of this source code is governed by a BSD-style license that can be +// found in the LICENSE file. + +/* + * curve25519-donna: Curve25519 elliptic curve, public key function + * + * http://code.google.com/p/curve25519-donna/ + * + * Adam Langley + * + * Derived from public domain C code by Daniel J. Bernstein + * + * More information about curve25519 can be found here + * http://cr.yp.to/ecdh.html + * + * djb's sample implementation of curve25519 is written in a special assembly + * language called qhasm and uses the floating point registers. + * + * This is, almost, a clean room reimplementation from the curve25519 paper. It + * uses many of the tricks described therein. Only the crecip function is taken + * from the sample implementation. + */ + +#include +#include + +typedef uint8_t u8; +typedef int32_t s32; +typedef int64_t limb; + +/* Field element representation: + * + * Field elements are written as an array of signed, 64-bit limbs, least + * significant first. The value of the field element is: + * x[0] + 2^26·x[1] + x^51·x[2] + 2^102·x[3] + ... + * + * i.e. the limbs are 26, 25, 26, 25, ... bits wide. + */ + +/* Sum two numbers: output += in */ +static void fsum(limb *output, const limb *in) { + unsigned i; + for (i = 0; i < 10; i += 2) { + output[0+i] = (output[0+i] + in[0+i]); + output[1+i] = (output[1+i] + in[1+i]); + } +} + +/* Find the difference of two numbers: output = in - output + * (note the order of the arguments!) + */ +static void fdifference(limb *output, const limb *in) { + unsigned i; + for (i = 0; i < 10; ++i) { + output[i] = (in[i] - output[i]); + } +} + +/* Multiply a number my a scalar: output = in * scalar */ +static void fscalar_product(limb *output, const limb *in, const limb scalar) { + unsigned i; + for (i = 0; i < 10; ++i) { + output[i] = in[i] * scalar; + } +} + +/* Multiply two numbers: output = in2 * in + * + * output must be distinct to both inputs. The inputs are reduced coefficient + * form, the output is not. + */ +static void fproduct(limb *output, const limb *in2, const limb *in) { + output[0] = ((limb) ((s32) in2[0])) * ((s32) in[0]); + output[1] = ((limb) ((s32) in2[0])) * ((s32) in[1]) + + ((limb) ((s32) in2[1])) * ((s32) in[0]); + output[2] = 2 * ((limb) ((s32) in2[1])) * ((s32) in[1]) + + ((limb) ((s32) in2[0])) * ((s32) in[2]) + + ((limb) ((s32) in2[2])) * ((s32) in[0]); + output[3] = ((limb) ((s32) in2[1])) * ((s32) in[2]) + + ((limb) ((s32) in2[2])) * ((s32) in[1]) + + ((limb) ((s32) in2[0])) * ((s32) in[3]) + + ((limb) ((s32) in2[3])) * ((s32) in[0]); + output[4] = ((limb) ((s32) in2[2])) * ((s32) in[2]) + + 2 * (((limb) ((s32) in2[1])) * ((s32) in[3]) + + ((limb) ((s32) in2[3])) * ((s32) in[1])) + + ((limb) ((s32) in2[0])) * ((s32) in[4]) + + ((limb) ((s32) in2[4])) * ((s32) in[0]); + output[5] = ((limb) ((s32) in2[2])) * ((s32) in[3]) + + ((limb) ((s32) in2[3])) * ((s32) in[2]) + + ((limb) ((s32) in2[1])) * ((s32) in[4]) + + ((limb) ((s32) in2[4])) * ((s32) in[1]) + + ((limb) ((s32) in2[0])) * ((s32) in[5]) + + ((limb) ((s32) in2[5])) * ((s32) in[0]); + output[6] = 2 * (((limb) ((s32) in2[3])) * ((s32) in[3]) + + ((limb) ((s32) in2[1])) * ((s32) in[5]) + + ((limb) ((s32) in2[5])) * ((s32) in[1])) + + ((limb) ((s32) in2[2])) * ((s32) in[4]) + + ((limb) ((s32) in2[4])) * ((s32) in[2]) + + ((limb) ((s32) in2[0])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[0]); + output[7] = ((limb) ((s32) in2[3])) * ((s32) in[4]) + + ((limb) ((s32) in2[4])) * ((s32) in[3]) + + ((limb) ((s32) in2[2])) * ((s32) in[5]) + + ((limb) ((s32) in2[5])) * ((s32) in[2]) + + ((limb) ((s32) in2[1])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[1]) + + ((limb) ((s32) in2[0])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[0]); + output[8] = ((limb) ((s32) in2[4])) * ((s32) in[4]) + + 2 * (((limb) ((s32) in2[3])) * ((s32) in[5]) + + ((limb) ((s32) in2[5])) * ((s32) in[3]) + + ((limb) ((s32) in2[1])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[1])) + + ((limb) ((s32) in2[2])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[2]) + + ((limb) ((s32) in2[0])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[0]); + output[9] = ((limb) ((s32) in2[4])) * ((s32) in[5]) + + ((limb) ((s32) in2[5])) * ((s32) in[4]) + + ((limb) ((s32) in2[3])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[3]) + + ((limb) ((s32) in2[2])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[2]) + + ((limb) ((s32) in2[1])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[1]) + + ((limb) ((s32) in2[0])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[0]); + output[10] = 2 * (((limb) ((s32) in2[5])) * ((s32) in[5]) + + ((limb) ((s32) in2[3])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[3]) + + ((limb) ((s32) in2[1])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[1])) + + ((limb) ((s32) in2[4])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[4]) + + ((limb) ((s32) in2[2])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[2]); + output[11] = ((limb) ((s32) in2[5])) * ((s32) in[6]) + + ((limb) ((s32) in2[6])) * ((s32) in[5]) + + ((limb) ((s32) in2[4])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[4]) + + ((limb) ((s32) in2[3])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[3]) + + ((limb) ((s32) in2[2])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[2]); + output[12] = ((limb) ((s32) in2[6])) * ((s32) in[6]) + + 2 * (((limb) ((s32) in2[5])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[5]) + + ((limb) ((s32) in2[3])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[3])) + + ((limb) ((s32) in2[4])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[4]); + output[13] = ((limb) ((s32) in2[6])) * ((s32) in[7]) + + ((limb) ((s32) in2[7])) * ((s32) in[6]) + + ((limb) ((s32) in2[5])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[5]) + + ((limb) ((s32) in2[4])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[4]); + output[14] = 2 * (((limb) ((s32) in2[7])) * ((s32) in[7]) + + ((limb) ((s32) in2[5])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[5])) + + ((limb) ((s32) in2[6])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[6]); + output[15] = ((limb) ((s32) in2[7])) * ((s32) in[8]) + + ((limb) ((s32) in2[8])) * ((s32) in[7]) + + ((limb) ((s32) in2[6])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[6]); + output[16] = ((limb) ((s32) in2[8])) * ((s32) in[8]) + + 2 * (((limb) ((s32) in2[7])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[7])); + output[17] = ((limb) ((s32) in2[8])) * ((s32) in[9]) + + ((limb) ((s32) in2[9])) * ((s32) in[8]); + output[18] = 2 * ((limb) ((s32) in2[9])) * ((s32) in[9]); +} + +/* Reduce a long form to a short form by taking the input mod 2^255 - 19. */ +static void freduce_degree(limb *output) { + /* Each of these shifts and adds ends up multiplying the value by 19. */ + output[8] += output[18] << 4; + output[8] += output[18] << 1; + output[8] += output[18]; + output[7] += output[17] << 4; + output[7] += output[17] << 1; + output[7] += output[17]; + output[6] += output[16] << 4; + output[6] += output[16] << 1; + output[6] += output[16]; + output[5] += output[15] << 4; + output[5] += output[15] << 1; + output[5] += output[15]; + output[4] += output[14] << 4; + output[4] += output[14] << 1; + output[4] += output[14]; + output[3] += output[13] << 4; + output[3] += output[13] << 1; + output[3] += output[13]; + output[2] += output[12] << 4; + output[2] += output[12] << 1; + output[2] += output[12]; + output[1] += output[11] << 4; + output[1] += output[11] << 1; + output[1] += output[11]; + output[0] += output[10] << 4; + output[0] += output[10] << 1; + output[0] += output[10]; +} + +/* Reduce all coefficients of the short form input so that |x| < 2^26. + * + * On entry: |output[i]| < 2^62 + */ +static void freduce_coefficients(limb *output) { + unsigned i; + do { + output[10] = 0; + + for (i = 0; i < 10; i += 2) { + limb over = output[i] / 0x4000000l; + output[i+1] += over; + output[i] -= over * 0x4000000l; + + over = output[i+1] / 0x2000000; + output[i+2] += over; + output[i+1] -= over * 0x2000000; + } + output[0] += 19 * output[10]; + } while (output[10]); +} + +/* A helpful wrapper around fproduct: output = in * in2. + * + * output must be distinct to both inputs. The output is reduced degree and + * reduced coefficient. + */ +static void +fmul(limb *output, const limb *in, const limb *in2) { + limb t[19]; + fproduct(t, in, in2); + freduce_degree(t); + freduce_coefficients(t); + memcpy(output, t, sizeof(limb) * 10); +} + +static void fsquare_inner(limb *output, const limb *in) { + output[0] = ((limb) ((s32) in[0])) * ((s32) in[0]); + output[1] = 2 * ((limb) ((s32) in[0])) * ((s32) in[1]); + output[2] = 2 * (((limb) ((s32) in[1])) * ((s32) in[1]) + + ((limb) ((s32) in[0])) * ((s32) in[2])); + output[3] = 2 * (((limb) ((s32) in[1])) * ((s32) in[2]) + + ((limb) ((s32) in[0])) * ((s32) in[3])); + output[4] = ((limb) ((s32) in[2])) * ((s32) in[2]) + + 4 * ((limb) ((s32) in[1])) * ((s32) in[3]) + + 2 * ((limb) ((s32) in[0])) * ((s32) in[4]); + output[5] = 2 * (((limb) ((s32) in[2])) * ((s32) in[3]) + + ((limb) ((s32) in[1])) * ((s32) in[4]) + + ((limb) ((s32) in[0])) * ((s32) in[5])); + output[6] = 2 * (((limb) ((s32) in[3])) * ((s32) in[3]) + + ((limb) ((s32) in[2])) * ((s32) in[4]) + + ((limb) ((s32) in[0])) * ((s32) in[6]) + + 2 * ((limb) ((s32) in[1])) * ((s32) in[5])); + output[7] = 2 * (((limb) ((s32) in[3])) * ((s32) in[4]) + + ((limb) ((s32) in[2])) * ((s32) in[5]) + + ((limb) ((s32) in[1])) * ((s32) in[6]) + + ((limb) ((s32) in[0])) * ((s32) in[7])); + output[8] = ((limb) ((s32) in[4])) * ((s32) in[4]) + + 2 * (((limb) ((s32) in[2])) * ((s32) in[6]) + + ((limb) ((s32) in[0])) * ((s32) in[8]) + + 2 * (((limb) ((s32) in[1])) * ((s32) in[7]) + + ((limb) ((s32) in[3])) * ((s32) in[5]))); + output[9] = 2 * (((limb) ((s32) in[4])) * ((s32) in[5]) + + ((limb) ((s32) in[3])) * ((s32) in[6]) + + ((limb) ((s32) in[2])) * ((s32) in[7]) + + ((limb) ((s32) in[1])) * ((s32) in[8]) + + ((limb) ((s32) in[0])) * ((s32) in[9])); + output[10] = 2 * (((limb) ((s32) in[5])) * ((s32) in[5]) + + ((limb) ((s32) in[4])) * ((s32) in[6]) + + ((limb) ((s32) in[2])) * ((s32) in[8]) + + 2 * (((limb) ((s32) in[3])) * ((s32) in[7]) + + ((limb) ((s32) in[1])) * ((s32) in[9]))); + output[11] = 2 * (((limb) ((s32) in[5])) * ((s32) in[6]) + + ((limb) ((s32) in[4])) * ((s32) in[7]) + + ((limb) ((s32) in[3])) * ((s32) in[8]) + + ((limb) ((s32) in[2])) * ((s32) in[9])); + output[12] = ((limb) ((s32) in[6])) * ((s32) in[6]) + + 2 * (((limb) ((s32) in[4])) * ((s32) in[8]) + + 2 * (((limb) ((s32) in[5])) * ((s32) in[7]) + + ((limb) ((s32) in[3])) * ((s32) in[9]))); + output[13] = 2 * (((limb) ((s32) in[6])) * ((s32) in[7]) + + ((limb) ((s32) in[5])) * ((s32) in[8]) + + ((limb) ((s32) in[4])) * ((s32) in[9])); + output[14] = 2 * (((limb) ((s32) in[7])) * ((s32) in[7]) + + ((limb) ((s32) in[6])) * ((s32) in[8]) + + 2 * ((limb) ((s32) in[5])) * ((s32) in[9])); + output[15] = 2 * (((limb) ((s32) in[7])) * ((s32) in[8]) + + ((limb) ((s32) in[6])) * ((s32) in[9])); + output[16] = ((limb) ((s32) in[8])) * ((s32) in[8]) + + 4 * ((limb) ((s32) in[7])) * ((s32) in[9]); + output[17] = 2 * ((limb) ((s32) in[8])) * ((s32) in[9]); + output[18] = 2 * ((limb) ((s32) in[9])) * ((s32) in[9]); +} + +static void +fsquare(limb *output, const limb *in) { + limb t[19]; + fsquare_inner(t, in); + freduce_degree(t); + freduce_coefficients(t); + memcpy(output, t, sizeof(limb) * 10); +} + +/* Take a little-endian, 32-byte number and expand it into polynomial form */ +static void +fexpand(limb *output, const u8 *input) { +#define F(n,start,shift,mask) \ + output[n] = ((((limb) input[start + 0]) | \ + ((limb) input[start + 1]) << 8 | \ + ((limb) input[start + 2]) << 16 | \ + ((limb) input[start + 3]) << 24) >> shift) & mask; + F(0, 0, 0, 0x3ffffff); + F(1, 3, 2, 0x1ffffff); + F(2, 6, 3, 0x3ffffff); + F(3, 9, 5, 0x1ffffff); + F(4, 12, 6, 0x3ffffff); + F(5, 16, 0, 0x1ffffff); + F(6, 19, 1, 0x3ffffff); + F(7, 22, 3, 0x1ffffff); + F(8, 25, 4, 0x3ffffff); + F(9, 28, 6, 0x1ffffff); +#undef F +} + +/* Take a fully reduced polynomial form number and contract it into a + * little-endian, 32-byte array + */ +static void +fcontract(u8 *output, limb *input) { + int i; + + do { + for (i = 0; i < 9; ++i) { + if ((i & 1) == 1) { + while (input[i] < 0) { + input[i] += 0x2000000; + input[i + 1]--; + } + } else { + while (input[i] < 0) { + input[i] += 0x4000000; + input[i + 1]--; + } + } + } + while (input[9] < 0) { + input[9] += 0x2000000; + input[0] -= 19; + } + } while (input[0] < 0); + + input[1] <<= 2; + input[2] <<= 3; + input[3] <<= 5; + input[4] <<= 6; + input[6] <<= 1; + input[7] <<= 3; + input[8] <<= 4; + input[9] <<= 6; +#define F(i, s) \ + output[s+0] |= input[i] & 0xff; \ + output[s+1] = (input[i] >> 8) & 0xff; \ + output[s+2] = (input[i] >> 16) & 0xff; \ + output[s+3] = (input[i] >> 24) & 0xff; + output[0] = 0; + output[16] = 0; + F(0,0); + F(1,3); + F(2,6); + F(3,9); + F(4,12); + F(5,16); + F(6,19); + F(7,22); + F(8,25); + F(9,28); +#undef F +} + +/* Input: Q, Q', Q-Q' + * Output: 2Q, Q+Q' + * + * x2 z3: long form + * x3 z3: long form + * x z: short form, destroyed + * xprime zprime: short form, destroyed + * qmqp: short form, preserved + */ +static void fmonty(limb *x2, limb *z2, /* output 2Q */ + limb *x3, limb *z3, /* output Q + Q' */ + limb *x, limb *z, /* input Q */ + limb *xprime, limb *zprime, /* input Q' */ + const limb *qmqp /* input Q - Q' */) { + limb origx[10], origxprime[10], zzz[19], xx[19], zz[19], xxprime[19], + zzprime[19], zzzprime[19], xxxprime[19]; + + memcpy(origx, x, 10 * sizeof(limb)); + fsum(x, z); + fdifference(z, origx); // does x - z + + memcpy(origxprime, xprime, sizeof(limb) * 10); + fsum(xprime, zprime); + fdifference(zprime, origxprime); + fproduct(xxprime, xprime, z); + fproduct(zzprime, x, zprime); + freduce_degree(xxprime); + freduce_coefficients(xxprime); + freduce_degree(zzprime); + freduce_coefficients(zzprime); + memcpy(origxprime, xxprime, sizeof(limb) * 10); + fsum(xxprime, zzprime); + fdifference(zzprime, origxprime); + fsquare(xxxprime, xxprime); + fsquare(zzzprime, zzprime); + fproduct(zzprime, zzzprime, qmqp); + freduce_degree(zzprime); + freduce_coefficients(zzprime); + memcpy(x3, xxxprime, sizeof(limb) * 10); + memcpy(z3, zzprime, sizeof(limb) * 10); + + fsquare(xx, x); + fsquare(zz, z); + fproduct(x2, xx, zz); + freduce_degree(x2); + freduce_coefficients(x2); + fdifference(zz, xx); // does zz = xx - zz + memset(zzz + 10, 0, sizeof(limb) * 9); + fscalar_product(zzz, zz, 121665); + freduce_degree(zzz); + freduce_coefficients(zzz); + fsum(zzz, xx); + fproduct(z2, zz, zzz); + freduce_degree(z2); + freduce_coefficients(z2); +} + +/* Calculates nQ where Q is the x-coordinate of a point on the curve + * + * resultx/resultz: the x coordinate of the resulting curve point (short form) + * n: a little endian, 32-byte number + * q: a point of the curve (short form) + */ +static void +cmult(limb *resultx, limb *resultz, const u8 *n, const limb *q) { + limb a[19] = {0}, b[19] = {1}, c[19] = {1}, d[19] = {0}; + limb *nqpqx = a, *nqpqz = b, *nqx = c, *nqz = d, *t; + limb e[19] = {0}, f[19] = {1}, g[19] = {0}, h[19] = {1}; + limb *nqpqx2 = e, *nqpqz2 = f, *nqx2 = g, *nqz2 = h; + + unsigned i, j; + + memcpy(nqpqx, q, sizeof(limb) * 10); + + for (i = 0; i < 32; ++i) { + u8 byte = n[31 - i]; + for (j = 0; j < 8; ++j) { + if (byte & 0x80) { + fmonty(nqpqx2, nqpqz2, + nqx2, nqz2, + nqpqx, nqpqz, + nqx, nqz, + q); + } else { + fmonty(nqx2, nqz2, + nqpqx2, nqpqz2, + nqx, nqz, + nqpqx, nqpqz, + q); + } + + t = nqx; + nqx = nqx2; + nqx2 = t; + t = nqz; + nqz = nqz2; + nqz2 = t; + t = nqpqx; + nqpqx = nqpqx2; + nqpqx2 = t; + t = nqpqz; + nqpqz = nqpqz2; + nqpqz2 = t; + + byte <<= 1; + } + } + + memcpy(resultx, nqx, sizeof(limb) * 10); + memcpy(resultz, nqz, sizeof(limb) * 10); +} + +// ----------------------------------------------------------------------------- +// Shamelessly copied from djb's code +// ----------------------------------------------------------------------------- +static void +crecip(limb *out, const limb *z) { + limb z2[10]; + limb z9[10]; + limb z11[10]; + limb z2_5_0[10]; + limb z2_10_0[10]; + limb z2_20_0[10]; + limb z2_50_0[10]; + limb z2_100_0[10]; + limb t0[10]; + limb t1[10]; + int i; + + /* 2 */ fsquare(z2,z); + /* 4 */ fsquare(t1,z2); + /* 8 */ fsquare(t0,t1); + /* 9 */ fmul(z9,t0,z); + /* 11 */ fmul(z11,z9,z2); + /* 22 */ fsquare(t0,z11); + /* 2^5 - 2^0 = 31 */ fmul(z2_5_0,t0,z9); + + /* 2^6 - 2^1 */ fsquare(t0,z2_5_0); + /* 2^7 - 2^2 */ fsquare(t1,t0); + /* 2^8 - 2^3 */ fsquare(t0,t1); + /* 2^9 - 2^4 */ fsquare(t1,t0); + /* 2^10 - 2^5 */ fsquare(t0,t1); + /* 2^10 - 2^0 */ fmul(z2_10_0,t0,z2_5_0); + + /* 2^11 - 2^1 */ fsquare(t0,z2_10_0); + /* 2^12 - 2^2 */ fsquare(t1,t0); + /* 2^20 - 2^10 */ + for (i = 2;i < 10;i += 2) { fsquare(t0,t1); fsquare(t1,t0); } + /* 2^20 - 2^0 */ fmul(z2_20_0,t1,z2_10_0); + + /* 2^21 - 2^1 */ fsquare(t0,z2_20_0); + /* 2^22 - 2^2 */ fsquare(t1,t0); + /* 2^40 - 2^20 */ + for (i = 2;i < 20;i += 2) { fsquare(t0,t1); fsquare(t1,t0); } + /* 2^40 - 2^0 */ fmul(t0,t1,z2_20_0); + + /* 2^41 - 2^1 */ fsquare(t1,t0); + /* 2^42 - 2^2 */ fsquare(t0,t1); + /* 2^50 - 2^10 */ + for (i = 2;i < 10;i += 2) { fsquare(t1,t0); fsquare(t0,t1); } + /* 2^50 - 2^0 */ fmul(z2_50_0,t0,z2_10_0); + + /* 2^51 - 2^1 */ fsquare(t0,z2_50_0); + /* 2^52 - 2^2 */ fsquare(t1,t0); + /* 2^100 - 2^50 */ + for (i = 2;i < 50;i += 2) { fsquare(t0,t1); fsquare(t1,t0); } + /* 2^100 - 2^0 */ fmul(z2_100_0,t1,z2_50_0); + + /* 2^101 - 2^1 */ fsquare(t1,z2_100_0); + /* 2^102 - 2^2 */ fsquare(t0,t1); + /* 2^200 - 2^100 */ + for (i = 2;i < 100;i += 2) { fsquare(t1,t0); fsquare(t0,t1); } + /* 2^200 - 2^0 */ fmul(t1,t0,z2_100_0); + + /* 2^201 - 2^1 */ fsquare(t0,t1); + /* 2^202 - 2^2 */ fsquare(t1,t0); + /* 2^250 - 2^50 */ + for (i = 2;i < 50;i += 2) { fsquare(t0,t1); fsquare(t1,t0); } + /* 2^250 - 2^0 */ fmul(t0,t1,z2_50_0); + + /* 2^251 - 2^1 */ fsquare(t1,t0); + /* 2^252 - 2^2 */ fsquare(t0,t1); + /* 2^253 - 2^3 */ fsquare(t1,t0); + /* 2^254 - 2^4 */ fsquare(t0,t1); + /* 2^255 - 2^5 */ fsquare(t1,t0); + /* 2^255 - 21 */ fmul(out,t1,z11); +} + +int +curve25519_donna(u8 *mypublic, const u8 *secret, const u8 *basepoint) { + limb bp[10], x[10], z[10], zmone[10]; + uint8_t e[32]; + int i; + + for (i = 0; i < 32; ++i) e[i] = secret[i]; + e[0] &= 248; + e[31] &= 127; + e[31] |= 64; + + fexpand(bp, basepoint); + cmult(x, z, e, bp); + crecip(zmone, z); + fmul(z, x, zmone); + fcontract(mypublic, z); + return 0; +} -- cgit v1.1