From d48555b36ac512161b81f9b6bca7bea16a0cd806 Mon Sep 17 00:00:00 2001 From: Pieter Wuille Date: Tue, 18 Nov 2014 18:06:36 +0100 Subject: Squashed 'src/secp256k1/' content from commit ad2028f git-subtree-dir: src/secp256k1 git-subtree-split: ad2028f9890ca40bdd32055aa0fe5c1c9af0e485 --- src/bench_inv.c | 41 ++ src/bench_sign.c | 49 ++ src/bench_verify.c | 44 ++ src/ecdsa.h | 23 + src/ecdsa_impl.h | 183 +++++ src/eckey.h | 25 + src/eckey_impl.h | 200 ++++++ src/ecmult.h | 19 + src/ecmult_gen.h | 19 + src/ecmult_gen_impl.h | 118 ++++ src/ecmult_impl.h | 222 ++++++ src/field.h | 114 +++ src/field_10x26.h | 21 + src/field_10x26_impl.h | 884 +++++++++++++++++++++++ src/field_5x52.h | 21 + src/field_5x52_asm.asm | 469 +++++++++++++ src/field_5x52_asm_impl.h | 13 + src/field_5x52_impl.h | 260 +++++++ src/field_5x52_int128_impl.h | 279 ++++++++ src/field_gmp.h | 18 + src/field_gmp_impl.h | 163 +++++ src/field_impl.h | 293 ++++++++ src/group.h | 128 ++++ src/group_impl.h | 519 ++++++++++++++ src/java/org/bitcoin/NativeSecp256k1.java | 60 ++ src/java/org_bitcoin_NativeSecp256k1.c | 23 + src/java/org_bitcoin_NativeSecp256k1.h | 21 + src/num.h | 100 +++ src/num_gmp.h | 20 + src/num_gmp_impl.h | 376 ++++++++++ src/num_impl.h | 22 + src/scalar.h | 63 ++ src/scalar_4x64.h | 17 + src/scalar_4x64_impl.h | 359 ++++++++++ src/scalar_8x32.h | 17 + src/scalar_8x32_impl.h | 572 +++++++++++++++ src/scalar_impl.h | 184 +++++ src/secp256k1.c | 305 ++++++++ src/testrand.h | 26 + src/testrand_impl.h | 60 ++ src/tests.c | 1080 +++++++++++++++++++++++++++++ src/util.h | 64 ++ 42 files changed, 7494 insertions(+) create mode 100644 src/bench_inv.c create mode 100644 src/bench_sign.c create mode 100644 src/bench_verify.c create mode 100644 src/ecdsa.h create mode 100644 src/ecdsa_impl.h create mode 100644 src/eckey.h create mode 100644 src/eckey_impl.h create mode 100644 src/ecmult.h create mode 100644 src/ecmult_gen.h create mode 100644 src/ecmult_gen_impl.h create mode 100644 src/ecmult_impl.h create mode 100644 src/field.h create mode 100644 src/field_10x26.h create mode 100644 src/field_10x26_impl.h create mode 100644 src/field_5x52.h create mode 100644 src/field_5x52_asm.asm create mode 100644 src/field_5x52_asm_impl.h create mode 100644 src/field_5x52_impl.h create mode 100644 src/field_5x52_int128_impl.h create mode 100644 src/field_gmp.h create mode 100644 src/field_gmp_impl.h create mode 100644 src/field_impl.h create mode 100644 src/group.h create mode 100644 src/group_impl.h create mode 100644 src/java/org/bitcoin/NativeSecp256k1.java create mode 100644 src/java/org_bitcoin_NativeSecp256k1.c create mode 100644 src/java/org_bitcoin_NativeSecp256k1.h create mode 100644 src/num.h create mode 100644 src/num_gmp.h create mode 100644 src/num_gmp_impl.h create mode 100644 src/num_impl.h create mode 100644 src/scalar.h create mode 100644 src/scalar_4x64.h create mode 100644 src/scalar_4x64_impl.h create mode 100644 src/scalar_8x32.h create mode 100644 src/scalar_8x32_impl.h create mode 100644 src/scalar_impl.h create mode 100644 src/secp256k1.c create mode 100644 src/testrand.h create mode 100644 src/testrand_impl.h create mode 100644 src/tests.c create mode 100644 src/util.h (limited to 'src') diff --git a/src/bench_inv.c b/src/bench_inv.c new file mode 100644 index 0000000000..d6f664333f --- /dev/null +++ b/src/bench_inv.c @@ -0,0 +1,41 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ +#include + +#include "include/secp256k1.h" + +#include "util.h" +#include "num_impl.h" +#include "field_impl.h" +#include "group_impl.h" +#include "scalar_impl.h" + +int main(void) { + static const unsigned char init[32] = { + 0x02, 0x03, 0x05, 0x07, 0x0b, 0x0d, 0x11, 0x13, + 0x17, 0x1d, 0x1f, 0x25, 0x29, 0x2b, 0x2f, 0x35, + 0x3b, 0x3d, 0x43, 0x47, 0x49, 0x4f, 0x53, 0x59, + 0x61, 0x65, 0x67, 0x6b, 0x6d, 0x71, 0x7f, 0x83 + }; + static const unsigned char fini[32] = { + 0xba, 0x28, 0x58, 0xd8, 0xaa, 0x11, 0xd6, 0xf2, + 0xfa, 0xce, 0x50, 0xb1, 0x67, 0x19, 0xb1, 0xa6, + 0xe0, 0xaa, 0x84, 0x53, 0xf6, 0x80, 0xfc, 0x23, + 0x88, 0x3c, 0xd6, 0x74, 0x9f, 0x27, 0x09, 0x03 + }; + secp256k1_ge_start(); + secp256k1_scalar_t base, x; + secp256k1_scalar_set_b32(&base, init, NULL); + secp256k1_scalar_set_b32(&x, init, NULL); + for (int i=0; i<1000000; i++) { + secp256k1_scalar_inverse(&x, &x); + secp256k1_scalar_add(&x, &x, &base); + } + unsigned char res[32]; + secp256k1_scalar_get_b32(res, &x); + CHECK(memcmp(res, fini, 32) == 0); + return 0; +} diff --git a/src/bench_sign.c b/src/bench_sign.c new file mode 100644 index 0000000000..f01f11d689 --- /dev/null +++ b/src/bench_sign.c @@ -0,0 +1,49 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ +#include +#include + +#include "include/secp256k1.h" +#include "util.h" + +int main(void) { + secp256k1_start(SECP256K1_START_SIGN); + + unsigned char msg[32]; + unsigned char nonce[32]; + unsigned char key[32]; + + for (int i = 0; i < 32; i++) msg[i] = i + 1; + for (int i = 0; i < 32; i++) nonce[i] = i + 33; + for (int i = 0; i < 32; i++) key[i] = i + 65; + + unsigned char sig[64]; + + for (int i=0; i<1000000; i++) { + int recid = 0; + CHECK(secp256k1_ecdsa_sign_compact(msg, 32, sig, key, nonce, &recid)); + for (int j = 0; j < 32; j++) { + nonce[j] = key[j]; /* Move former key to nonce */ + msg[j] = sig[j]; /* Move former R to message. */ + key[j] = sig[j + 32]; /* Move former S to key. */ + } + } + + static const unsigned char fini[64] = { + 0x92, 0x03, 0xef, 0xf1, 0x58, 0x0b, 0x49, 0x8d, + 0x22, 0x3d, 0x49, 0x0e, 0xbf, 0x26, 0x50, 0x0e, + 0x2d, 0x62, 0x90, 0xd7, 0x82, 0xbd, 0x3d, 0x5c, + 0xa9, 0x10, 0xa5, 0x49, 0xb1, 0xd8, 0x8c, 0xc0, + 0x5b, 0x5e, 0x9e, 0x68, 0x51, 0x3d, 0xe8, 0xec, + 0x82, 0x30, 0x82, 0x88, 0x8c, 0xfd, 0xe7, 0x71, + 0x15, 0x92, 0xfc, 0x14, 0x59, 0x78, 0x31, 0xb3, + 0xf6, 0x07, 0x91, 0x18, 0x00, 0x8d, 0x4c, 0xb2 + }; + CHECK(memcmp(sig, fini, 64) == 0); + + secp256k1_stop(); + return 0; +} diff --git a/src/bench_verify.c b/src/bench_verify.c new file mode 100644 index 0000000000..690595516d --- /dev/null +++ b/src/bench_verify.c @@ -0,0 +1,44 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#include +#include + +#include "include/secp256k1.h" +#include "util.h" + +int main(void) { + secp256k1_start(SECP256K1_START_VERIFY); + + unsigned char msg[32]; + unsigned char sig[64]; + + for (int i = 0; i < 32; i++) msg[i] = 1 + i; + for (int i = 0; i < 64; i++) sig[i] = 65 + i; + + unsigned char pubkey[33]; + for (int i=0; i<1000000; i++) { + int pubkeylen = 33; + CHECK(secp256k1_ecdsa_recover_compact(msg, 32, sig, pubkey, &pubkeylen, 1, i % 2)); + for (int j = 0; j < 32; j++) { + sig[j + 32] = msg[j]; /* Move former message to S. */ + msg[j] = sig[j]; /* Move former R to message. */ + sig[j] = pubkey[j + 1]; /* Move recovered pubkey X coordinate to R (which must be a valid X coordinate). */ + } + } + + static const unsigned char fini[33] = { + 0x02, + 0x52, 0x63, 0xae, 0x9a, 0x9d, 0x47, 0x1f, 0x1a, + 0xb2, 0x36, 0x65, 0x89, 0x11, 0xe7, 0xcc, 0x86, + 0xa3, 0xab, 0x97, 0xb6, 0xf1, 0xaf, 0xfd, 0x8f, + 0x9b, 0x38, 0xb6, 0x18, 0x55, 0xe5, 0xc2, 0x43 + }; + CHECK(memcmp(fini, pubkey, 33) == 0); + + secp256k1_stop(); + return 0; +} diff --git a/src/ecdsa.h b/src/ecdsa.h new file mode 100644 index 0000000000..3b1e0484ea --- /dev/null +++ b/src/ecdsa.h @@ -0,0 +1,23 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECDSA_ +#define _SECP256K1_ECDSA_ + +#include "num.h" + +typedef struct { + secp256k1_num_t r, s; +} secp256k1_ecdsa_sig_t; + +static int secp256k1_ecdsa_sig_parse(secp256k1_ecdsa_sig_t *r, const unsigned char *sig, int size); +static int secp256k1_ecdsa_sig_serialize(unsigned char *sig, int *size, const secp256k1_ecdsa_sig_t *a); +static int secp256k1_ecdsa_sig_verify(const secp256k1_ecdsa_sig_t *sig, const secp256k1_ge_t *pubkey, const secp256k1_num_t *message); +static int secp256k1_ecdsa_sig_sign(secp256k1_ecdsa_sig_t *sig, const secp256k1_scalar_t *seckey, const secp256k1_scalar_t *message, const secp256k1_scalar_t *nonce, int *recid); +static int secp256k1_ecdsa_sig_recover(const secp256k1_ecdsa_sig_t *sig, secp256k1_ge_t *pubkey, const secp256k1_num_t *message, int recid); +static void secp256k1_ecdsa_sig_set_rs(secp256k1_ecdsa_sig_t *sig, const secp256k1_num_t *r, const secp256k1_num_t *s); + +#endif diff --git a/src/ecdsa_impl.h b/src/ecdsa_impl.h new file mode 100644 index 0000000000..4c05ec39f8 --- /dev/null +++ b/src/ecdsa_impl.h @@ -0,0 +1,183 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + + +#ifndef _SECP256K1_ECDSA_IMPL_H_ +#define _SECP256K1_ECDSA_IMPL_H_ + +#include "num.h" +#include "field.h" +#include "group.h" +#include "ecmult.h" +#include "ecmult_gen.h" +#include "ecdsa.h" + +static int secp256k1_ecdsa_sig_parse(secp256k1_ecdsa_sig_t *r, const unsigned char *sig, int size) { + if (sig[0] != 0x30) return 0; + int lenr = sig[3]; + if (5+lenr >= size) return 0; + int lens = sig[lenr+5]; + if (sig[1] != lenr+lens+4) return 0; + if (lenr+lens+6 > size) return 0; + if (sig[2] != 0x02) return 0; + if (lenr == 0) return 0; + if (sig[lenr+4] != 0x02) return 0; + if (lens == 0) return 0; + secp256k1_num_set_bin(&r->r, sig+4, lenr); + secp256k1_num_set_bin(&r->s, sig+6+lenr, lens); + return 1; +} + +static int secp256k1_ecdsa_sig_serialize(unsigned char *sig, int *size, const secp256k1_ecdsa_sig_t *a) { + int lenR = (secp256k1_num_bits(&a->r) + 7)/8; + if (lenR == 0 || secp256k1_num_get_bit(&a->r, lenR*8-1)) + lenR++; + int lenS = (secp256k1_num_bits(&a->s) + 7)/8; + if (lenS == 0 || secp256k1_num_get_bit(&a->s, lenS*8-1)) + lenS++; + if (*size < 6+lenS+lenR) + return 0; + *size = 6 + lenS + lenR; + sig[0] = 0x30; + sig[1] = 4 + lenS + lenR; + sig[2] = 0x02; + sig[3] = lenR; + secp256k1_num_get_bin(sig+4, lenR, &a->r); + sig[4+lenR] = 0x02; + sig[5+lenR] = lenS; + secp256k1_num_get_bin(sig+lenR+6, lenS, &a->s); + return 1; +} + +static int secp256k1_ecdsa_sig_recompute(secp256k1_num_t *r2, const secp256k1_ecdsa_sig_t *sig, const secp256k1_ge_t *pubkey, const secp256k1_num_t *message) { + const secp256k1_ge_consts_t *c = secp256k1_ge_consts; + + if (secp256k1_num_is_neg(&sig->r) || secp256k1_num_is_neg(&sig->s)) + return 0; + if (secp256k1_num_is_zero(&sig->r) || secp256k1_num_is_zero(&sig->s)) + return 0; + if (secp256k1_num_cmp(&sig->r, &c->order) >= 0 || secp256k1_num_cmp(&sig->s, &c->order) >= 0) + return 0; + + int ret = 0; + secp256k1_num_t sn, u1, u2; + secp256k1_num_init(&sn); + secp256k1_num_init(&u1); + secp256k1_num_init(&u2); + secp256k1_num_mod_inverse(&sn, &sig->s, &c->order); + secp256k1_num_mod_mul(&u1, &sn, message, &c->order); + secp256k1_num_mod_mul(&u2, &sn, &sig->r, &c->order); + secp256k1_gej_t pubkeyj; secp256k1_gej_set_ge(&pubkeyj, pubkey); + secp256k1_gej_t pr; secp256k1_ecmult(&pr, &pubkeyj, &u2, &u1); + if (!secp256k1_gej_is_infinity(&pr)) { + secp256k1_fe_t xr; secp256k1_gej_get_x_var(&xr, &pr); + secp256k1_fe_normalize(&xr); + unsigned char xrb[32]; secp256k1_fe_get_b32(xrb, &xr); + secp256k1_num_set_bin(r2, xrb, 32); + secp256k1_num_mod(r2, &c->order); + ret = 1; + } + secp256k1_num_free(&sn); + secp256k1_num_free(&u1); + secp256k1_num_free(&u2); + return ret; +} + +static int secp256k1_ecdsa_sig_recover(const secp256k1_ecdsa_sig_t *sig, secp256k1_ge_t *pubkey, const secp256k1_num_t *message, int recid) { + const secp256k1_ge_consts_t *c = secp256k1_ge_consts; + + if (secp256k1_num_is_neg(&sig->r) || secp256k1_num_is_neg(&sig->s)) + return 0; + if (secp256k1_num_is_zero(&sig->r) || secp256k1_num_is_zero(&sig->s)) + return 0; + if (secp256k1_num_cmp(&sig->r, &c->order) >= 0 || secp256k1_num_cmp(&sig->s, &c->order) >= 0) + return 0; + + secp256k1_num_t rx; + secp256k1_num_init(&rx); + secp256k1_num_copy(&rx, &sig->r); + if (recid & 2) { + secp256k1_num_add(&rx, &rx, &c->order); + if (secp256k1_num_cmp(&rx, &secp256k1_fe_consts->p) >= 0) + return 0; + } + unsigned char brx[32]; + secp256k1_num_get_bin(brx, 32, &rx); + secp256k1_num_free(&rx); + secp256k1_fe_t fx; + secp256k1_fe_set_b32(&fx, brx); + secp256k1_ge_t x; + if (!secp256k1_ge_set_xo(&x, &fx, recid & 1)) + return 0; + secp256k1_gej_t xj; + secp256k1_gej_set_ge(&xj, &x); + secp256k1_num_t rn, u1, u2; + secp256k1_num_init(&rn); + secp256k1_num_init(&u1); + secp256k1_num_init(&u2); + secp256k1_num_mod_inverse(&rn, &sig->r, &c->order); + secp256k1_num_mod_mul(&u1, &rn, message, &c->order); + secp256k1_num_sub(&u1, &c->order, &u1); + secp256k1_num_mod_mul(&u2, &rn, &sig->s, &c->order); + secp256k1_gej_t qj; + secp256k1_ecmult(&qj, &xj, &u2, &u1); + secp256k1_ge_set_gej_var(pubkey, &qj); + secp256k1_num_free(&rn); + secp256k1_num_free(&u1); + secp256k1_num_free(&u2); + return !secp256k1_gej_is_infinity(&qj); +} + +static int secp256k1_ecdsa_sig_verify(const secp256k1_ecdsa_sig_t *sig, const secp256k1_ge_t *pubkey, const secp256k1_num_t *message) { + secp256k1_num_t r2; + secp256k1_num_init(&r2); + int ret = 0; + ret = secp256k1_ecdsa_sig_recompute(&r2, sig, pubkey, message) && secp256k1_num_cmp(&sig->r, &r2) == 0; + secp256k1_num_free(&r2); + return ret; +} + +static int secp256k1_ecdsa_sig_sign(secp256k1_ecdsa_sig_t *sig, const secp256k1_scalar_t *seckey, const secp256k1_scalar_t *message, const secp256k1_scalar_t *nonce, int *recid) { + secp256k1_gej_t rp; + secp256k1_ecmult_gen(&rp, nonce); + secp256k1_ge_t r; + secp256k1_ge_set_gej(&r, &rp); + unsigned char b[32]; + secp256k1_fe_normalize(&r.x); + secp256k1_fe_normalize(&r.y); + secp256k1_fe_get_b32(b, &r.x); + int overflow = 0; + secp256k1_scalar_t sigr; + secp256k1_scalar_set_b32(&sigr, b, &overflow); + if (recid) + *recid = (overflow ? 2 : 0) | (secp256k1_fe_is_odd(&r.y) ? 1 : 0); + secp256k1_scalar_t n; + secp256k1_scalar_mul(&n, &sigr, seckey); + secp256k1_scalar_add(&n, &n, message); + secp256k1_scalar_t sigs; + secp256k1_scalar_inverse(&sigs, nonce); + secp256k1_scalar_mul(&sigs, &sigs, &n); + secp256k1_scalar_clear(&n); + secp256k1_gej_clear(&rp); + secp256k1_ge_clear(&r); + if (secp256k1_scalar_is_zero(&sigs)) + return 0; + if (secp256k1_scalar_is_high(&sigs)) { + secp256k1_scalar_negate(&sigs, &sigs); + if (recid) + *recid ^= 1; + } + secp256k1_scalar_get_num(&sig->s, &sigs); + secp256k1_scalar_get_num(&sig->r, &sigr); + return 1; +} + +static void secp256k1_ecdsa_sig_set_rs(secp256k1_ecdsa_sig_t *sig, const secp256k1_num_t *r, const secp256k1_num_t *s) { + secp256k1_num_copy(&sig->r, r); + secp256k1_num_copy(&sig->s, s); +} + +#endif diff --git a/src/eckey.h b/src/eckey.h new file mode 100644 index 0000000000..024c8b821b --- /dev/null +++ b/src/eckey.h @@ -0,0 +1,25 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECKEY_ +#define _SECP256K1_ECKEY_ + +#include "group.h" +#include "scalar.h" +#include "num.h" + +static int secp256k1_eckey_pubkey_parse(secp256k1_ge_t *elem, const unsigned char *pub, int size); +static int secp256k1_eckey_pubkey_serialize(secp256k1_ge_t *elem, unsigned char *pub, int *size, int compressed); + +static int secp256k1_eckey_privkey_parse(secp256k1_scalar_t *key, const unsigned char *privkey, int privkeylen); +static int secp256k1_eckey_privkey_serialize(unsigned char *privkey, int *privkeylen, const secp256k1_scalar_t *key, int compressed); + +static int secp256k1_eckey_privkey_tweak_add(secp256k1_scalar_t *key, const secp256k1_scalar_t *tweak); +static int secp256k1_eckey_pubkey_tweak_add(secp256k1_ge_t *key, const secp256k1_num_t *tweak); +static int secp256k1_eckey_privkey_tweak_mul(secp256k1_scalar_t *key, const secp256k1_scalar_t *tweak); +static int secp256k1_eckey_pubkey_tweak_mul(secp256k1_ge_t *key, const secp256k1_num_t *tweak); + +#endif diff --git a/src/eckey_impl.h b/src/eckey_impl.h new file mode 100644 index 0000000000..290b1f0900 --- /dev/null +++ b/src/eckey_impl.h @@ -0,0 +1,200 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECKEY_IMPL_H_ +#define _SECP256K1_ECKEY_IMPL_H_ + +#include "eckey.h" + +#include "num.h" +#include "field.h" +#include "group.h" +#include "ecmult_gen.h" + +static int secp256k1_eckey_pubkey_parse(secp256k1_ge_t *elem, const unsigned char *pub, int size) { + if (size == 33 && (pub[0] == 0x02 || pub[0] == 0x03)) { + secp256k1_fe_t x; + secp256k1_fe_set_b32(&x, pub+1); + return secp256k1_ge_set_xo(elem, &x, pub[0] == 0x03); + } else if (size == 65 && (pub[0] == 0x04 || pub[0] == 0x06 || pub[0] == 0x07)) { + secp256k1_fe_t x, y; + secp256k1_fe_set_b32(&x, pub+1); + secp256k1_fe_set_b32(&y, pub+33); + secp256k1_ge_set_xy(elem, &x, &y); + if ((pub[0] == 0x06 || pub[0] == 0x07) && secp256k1_fe_is_odd(&y) != (pub[0] == 0x07)) + return 0; + return secp256k1_ge_is_valid(elem); + } else { + return 0; + } +} + +static int secp256k1_eckey_pubkey_serialize(secp256k1_ge_t *elem, unsigned char *pub, int *size, int compressed) { + if (secp256k1_ge_is_infinity(elem)) { + return 0; + } + secp256k1_fe_normalize(&elem->x); + secp256k1_fe_normalize(&elem->y); + secp256k1_fe_get_b32(&pub[1], &elem->x); + if (compressed) { + *size = 33; + pub[0] = 0x02 | (secp256k1_fe_is_odd(&elem->y) ? 0x01 : 0x00); + } else { + *size = 65; + pub[0] = 0x04; + secp256k1_fe_get_b32(&pub[33], &elem->y); + } + return 1; +} + +static int secp256k1_eckey_privkey_parse(secp256k1_scalar_t *key, const unsigned char *privkey, int privkeylen) { + const unsigned char *end = privkey + privkeylen; + /* sequence header */ + if (end < privkey+1 || *privkey != 0x30) + return 0; + privkey++; + /* sequence length constructor */ + int lenb = 0; + if (end < privkey+1 || !(*privkey & 0x80)) + return 0; + lenb = *privkey & ~0x80; privkey++; + if (lenb < 1 || lenb > 2) + return 0; + if (end < privkey+lenb) + return 0; + /* sequence length */ + int len = 0; + len = privkey[lenb-1] | (lenb > 1 ? privkey[lenb-2] << 8 : 0); + privkey += lenb; + if (end < privkey+len) + return 0; + /* sequence element 0: version number (=1) */ + if (end < privkey+3 || privkey[0] != 0x02 || privkey[1] != 0x01 || privkey[2] != 0x01) + return 0; + privkey += 3; + /* sequence element 1: octet string, up to 32 bytes */ + if (end < privkey+2 || privkey[0] != 0x04 || privkey[1] > 0x20 || end < privkey+2+privkey[1]) + return 0; + int overflow = 0; + unsigned char c[32] = {0}; + memcpy(c + 32 - privkey[1], privkey + 2, privkey[1]); + secp256k1_scalar_set_b32(key, c, &overflow); + memset(c, 0, 32); + return !overflow; +} + +static int secp256k1_eckey_privkey_serialize(unsigned char *privkey, int *privkeylen, const secp256k1_scalar_t *key, int compressed) { + secp256k1_gej_t rp; + secp256k1_ecmult_gen(&rp, key); + secp256k1_ge_t r; + secp256k1_ge_set_gej(&r, &rp); + if (compressed) { + static const unsigned char begin[] = { + 0x30,0x81,0xD3,0x02,0x01,0x01,0x04,0x20 + }; + static const unsigned char middle[] = { + 0xA0,0x81,0x85,0x30,0x81,0x82,0x02,0x01,0x01,0x30,0x2C,0x06,0x07,0x2A,0x86,0x48, + 0xCE,0x3D,0x01,0x01,0x02,0x21,0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F,0x30,0x06,0x04,0x01,0x00,0x04,0x01,0x07,0x04, + 0x21,0x02,0x79,0xBE,0x66,0x7E,0xF9,0xDC,0xBB,0xAC,0x55,0xA0,0x62,0x95,0xCE,0x87, + 0x0B,0x07,0x02,0x9B,0xFC,0xDB,0x2D,0xCE,0x28,0xD9,0x59,0xF2,0x81,0x5B,0x16,0xF8, + 0x17,0x98,0x02,0x21,0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFE,0xBA,0xAE,0xDC,0xE6,0xAF,0x48,0xA0,0x3B,0xBF,0xD2,0x5E, + 0x8C,0xD0,0x36,0x41,0x41,0x02,0x01,0x01,0xA1,0x24,0x03,0x22,0x00 + }; + unsigned char *ptr = privkey; + memcpy(ptr, begin, sizeof(begin)); ptr += sizeof(begin); + secp256k1_scalar_get_b32(ptr, key); ptr += 32; + memcpy(ptr, middle, sizeof(middle)); ptr += sizeof(middle); + int pubkeylen = 0; + if (!secp256k1_eckey_pubkey_serialize(&r, ptr, &pubkeylen, 1)) { + return 0; + } + ptr += pubkeylen; + *privkeylen = ptr - privkey; + } else { + static const unsigned char begin[] = { + 0x30,0x82,0x01,0x13,0x02,0x01,0x01,0x04,0x20 + }; + static const unsigned char middle[] = { + 0xA0,0x81,0xA5,0x30,0x81,0xA2,0x02,0x01,0x01,0x30,0x2C,0x06,0x07,0x2A,0x86,0x48, + 0xCE,0x3D,0x01,0x01,0x02,0x21,0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F,0x30,0x06,0x04,0x01,0x00,0x04,0x01,0x07,0x04, + 0x41,0x04,0x79,0xBE,0x66,0x7E,0xF9,0xDC,0xBB,0xAC,0x55,0xA0,0x62,0x95,0xCE,0x87, + 0x0B,0x07,0x02,0x9B,0xFC,0xDB,0x2D,0xCE,0x28,0xD9,0x59,0xF2,0x81,0x5B,0x16,0xF8, + 0x17,0x98,0x48,0x3A,0xDA,0x77,0x26,0xA3,0xC4,0x65,0x5D,0xA4,0xFB,0xFC,0x0E,0x11, + 0x08,0xA8,0xFD,0x17,0xB4,0x48,0xA6,0x85,0x54,0x19,0x9C,0x47,0xD0,0x8F,0xFB,0x10, + 0xD4,0xB8,0x02,0x21,0x00,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFE,0xBA,0xAE,0xDC,0xE6,0xAF,0x48,0xA0,0x3B,0xBF,0xD2,0x5E, + 0x8C,0xD0,0x36,0x41,0x41,0x02,0x01,0x01,0xA1,0x44,0x03,0x42,0x00 + }; + unsigned char *ptr = privkey; + memcpy(ptr, begin, sizeof(begin)); ptr += sizeof(begin); + secp256k1_scalar_get_b32(ptr, key); ptr += 32; + memcpy(ptr, middle, sizeof(middle)); ptr += sizeof(middle); + int pubkeylen = 0; + if (!secp256k1_eckey_pubkey_serialize(&r, ptr, &pubkeylen, 0)) { + return 0; + } + ptr += pubkeylen; + *privkeylen = ptr - privkey; + } + return 1; +} + +static int secp256k1_eckey_privkey_tweak_add(secp256k1_scalar_t *key, const secp256k1_scalar_t *tweak) { + secp256k1_scalar_add(key, key, tweak); + if (secp256k1_scalar_is_zero(key)) + return 0; + return 1; +} + +static int secp256k1_eckey_pubkey_tweak_add(secp256k1_ge_t *key, const secp256k1_num_t *tweak) { + if (secp256k1_num_cmp(tweak, &secp256k1_ge_consts->order) >= 0) + return 0; + + secp256k1_gej_t pt; + secp256k1_gej_set_ge(&pt, key); + secp256k1_num_t one; + secp256k1_num_init(&one); + secp256k1_num_set_int(&one, 1); + secp256k1_ecmult(&pt, &pt, &one, tweak); + secp256k1_num_free(&one); + + if (secp256k1_gej_is_infinity(&pt)) + return 0; + secp256k1_ge_set_gej(key, &pt); + return 1; +} + +static int secp256k1_eckey_privkey_tweak_mul(secp256k1_scalar_t *key, const secp256k1_scalar_t *tweak) { + if (secp256k1_scalar_is_zero(tweak)) + return 0; + + secp256k1_scalar_mul(key, key, tweak); + return 1; +} + +static int secp256k1_eckey_pubkey_tweak_mul(secp256k1_ge_t *key, const secp256k1_num_t *tweak) { + if (secp256k1_num_is_zero(tweak)) + return 0; + if (secp256k1_num_cmp(tweak, &secp256k1_ge_consts->order) >= 0) + return 0; + + secp256k1_num_t zero; + secp256k1_num_init(&zero); + secp256k1_num_set_int(&zero, 0); + secp256k1_gej_t pt; + secp256k1_gej_set_ge(&pt, key); + secp256k1_ecmult(&pt, &pt, tweak, &zero); + secp256k1_num_free(&zero); + secp256k1_ge_set_gej(key, &pt); + return 1; +} + +#endif diff --git a/src/ecmult.h b/src/ecmult.h new file mode 100644 index 0000000000..e3cf18b680 --- /dev/null +++ b/src/ecmult.h @@ -0,0 +1,19 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECMULT_ +#define _SECP256K1_ECMULT_ + +#include "num.h" +#include "group.h" + +static void secp256k1_ecmult_start(void); +static void secp256k1_ecmult_stop(void); + +/** Double multiply: R = na*A + ng*G */ +static void secp256k1_ecmult(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_num_t *na, const secp256k1_num_t *ng); + +#endif diff --git a/src/ecmult_gen.h b/src/ecmult_gen.h new file mode 100644 index 0000000000..42f822f9ce --- /dev/null +++ b/src/ecmult_gen.h @@ -0,0 +1,19 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECMULT_GEN_ +#define _SECP256K1_ECMULT_GEN_ + +#include "scalar.h" +#include "group.h" + +static void secp256k1_ecmult_gen_start(void); +static void secp256k1_ecmult_gen_stop(void); + +/** Multiply with the generator: R = a*G */ +static void secp256k1_ecmult_gen(secp256k1_gej_t *r, const secp256k1_scalar_t *a); + +#endif diff --git a/src/ecmult_gen_impl.h b/src/ecmult_gen_impl.h new file mode 100644 index 0000000000..07859ab04b --- /dev/null +++ b/src/ecmult_gen_impl.h @@ -0,0 +1,118 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECMULT_GEN_IMPL_H_ +#define _SECP256K1_ECMULT_GEN_IMPL_H_ + +#include "scalar.h" +#include "group.h" +#include "ecmult_gen.h" + +typedef struct { + /* For accelerating the computation of a*G: + * To harden against timing attacks, use the following mechanism: + * * Break up the multiplicand into groups of 4 bits, called n_0, n_1, n_2, ..., n_63. + * * Compute sum(n_i * 16^i * G + U_i, i=0..63), where: + * * U_i = U * 2^i (for i=0..62) + * * U_i = U * (1-2^63) (for i=63) + * where U is a point with no known corresponding scalar. Note that sum(U_i, i=0..63) = 0. + * For each i, and each of the 16 possible values of n_i, (n_i * 16^i * G + U_i) is + * precomputed (call it prec(i, n_i)). The formula now becomes sum(prec(i, n_i), i=0..63). + * None of the resulting prec group elements have a known scalar, and neither do any of + * the intermediate sums while computing a*G. + * To make memory access uniform, the bytes of prec(i, n_i) are sliced per value of n_i. */ + unsigned char prec[64][sizeof(secp256k1_ge_t)][16]; /* prec[j][k][i] = k'th byte of (16^j * i * G + U_i) */ +} secp256k1_ecmult_gen_consts_t; + +static const secp256k1_ecmult_gen_consts_t *secp256k1_ecmult_gen_consts = NULL; + +static void secp256k1_ecmult_gen_start(void) { + if (secp256k1_ecmult_gen_consts != NULL) + return; + + /* Allocate the precomputation table. */ + secp256k1_ecmult_gen_consts_t *ret = (secp256k1_ecmult_gen_consts_t*)malloc(sizeof(secp256k1_ecmult_gen_consts_t)); + + /* get the generator */ + const secp256k1_ge_t *g = &secp256k1_ge_consts->g; + secp256k1_gej_t gj; secp256k1_gej_set_ge(&gj, g); + + /* Construct a group element with no known corresponding scalar (nothing up my sleeve). */ + secp256k1_gej_t nums_gej; + { + static const unsigned char nums_b32[32] = "The scalar for this x is unknown"; + secp256k1_fe_t nums_x; + secp256k1_fe_set_b32(&nums_x, nums_b32); + secp256k1_ge_t nums_ge; + VERIFY_CHECK(secp256k1_ge_set_xo(&nums_ge, &nums_x, 0)); + secp256k1_gej_set_ge(&nums_gej, &nums_ge); + /* Add G to make the bits in x uniformly distributed. */ + secp256k1_gej_add_ge_var(&nums_gej, &nums_gej, g); + } + + /* compute prec. */ + secp256k1_ge_t prec[1024]; + { + secp256k1_gej_t precj[1024]; /* Jacobian versions of prec. */ + secp256k1_gej_t gbase; gbase = gj; /* 16^j * G */ + secp256k1_gej_t numsbase; numsbase = nums_gej; /* 2^j * nums. */ + for (int j=0; j<64; j++) { + /* Set precj[j*16 .. j*16+15] to (numsbase, numsbase + gbase, ..., numsbase + 15*gbase). */ + precj[j*16] = numsbase; + for (int i=1; i<16; i++) { + secp256k1_gej_add_var(&precj[j*16 + i], &precj[j*16 + i - 1], &gbase); + } + /* Multiply gbase by 16. */ + for (int i=0; i<4; i++) { + secp256k1_gej_double_var(&gbase, &gbase); + } + /* Multiply numbase by 2. */ + secp256k1_gej_double_var(&numsbase, &numsbase); + if (j == 62) { + /* In the last iteration, numsbase is (1 - 2^j) * nums instead. */ + secp256k1_gej_neg(&numsbase, &numsbase); + secp256k1_gej_add_var(&numsbase, &numsbase, &nums_gej); + } + } + secp256k1_ge_set_all_gej_var(1024, prec, precj); + } + for (int j=0; j<64; j++) { + for (int i=0; i<16; i++) { + const unsigned char* raw = (const unsigned char*)(&prec[j*16 + i]); + for (size_t k=0; kprec[j][k][i] = raw[k]; + } + } + + /* Set the global pointer to the precomputation table. */ + secp256k1_ecmult_gen_consts = ret; +} + +static void secp256k1_ecmult_gen_stop(void) { + if (secp256k1_ecmult_gen_consts == NULL) + return; + + secp256k1_ecmult_gen_consts_t *c = (secp256k1_ecmult_gen_consts_t*)secp256k1_ecmult_gen_consts; + secp256k1_ecmult_gen_consts = NULL; + free(c); +} + +static void secp256k1_ecmult_gen(secp256k1_gej_t *r, const secp256k1_scalar_t *gn) { + const secp256k1_ecmult_gen_consts_t *c = secp256k1_ecmult_gen_consts; + secp256k1_gej_set_infinity(r); + secp256k1_ge_t add; + int bits; + for (int j=0; j<64; j++) { + bits = secp256k1_scalar_get_bits(gn, j * 4, 4); + for (size_t k=0; kprec[j][k][bits]; + secp256k1_gej_add_ge(r, r, &add); + } + bits = 0; + secp256k1_ge_clear(&add); +} + +#endif diff --git a/src/ecmult_impl.h b/src/ecmult_impl.h new file mode 100644 index 0000000000..508902564e --- /dev/null +++ b/src/ecmult_impl.h @@ -0,0 +1,222 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_ECMULT_IMPL_H_ +#define _SECP256K1_ECMULT_IMPL_H_ + +#include "num.h" +#include "group.h" +#include "ecmult.h" + +/* optimal for 128-bit and 256-bit exponents. */ +#define WINDOW_A 5 + +/** larger numbers may result in slightly better performance, at the cost of + exponentially larger precomputed tables. WINDOW_G == 14 results in 640 KiB. */ +#define WINDOW_G 14 + +/** Fill a table 'pre' with precomputed odd multiples of a. W determines the size of the table. + * pre will contains the values [1*a,3*a,5*a,...,(2^(w-1)-1)*a], so it needs place for + * 2^(w-2) entries. + * + * There are two versions of this function: + * - secp256k1_ecmult_precomp_wnaf_gej, which operates on group elements in jacobian notation, + * fast to precompute, but slower to use in later additions. + * - secp256k1_ecmult_precomp_wnaf_ge, which operates on group elements in affine notations, + * (much) slower to precompute, but a bit faster to use in later additions. + * To compute a*P + b*G, we use the jacobian version for P, and the affine version for G, as + * G is constant, so it only needs to be done once in advance. + */ +static void secp256k1_ecmult_table_precomp_gej_var(secp256k1_gej_t *pre, const secp256k1_gej_t *a, int w) { + pre[0] = *a; + secp256k1_gej_t d; secp256k1_gej_double_var(&d, &pre[0]); + for (int i=1; i<(1 << (w-2)); i++) + secp256k1_gej_add_var(&pre[i], &d, &pre[i-1]); +} + +static void secp256k1_ecmult_table_precomp_ge_var(secp256k1_ge_t *pre, const secp256k1_gej_t *a, int w) { + const int table_size = 1 << (w-2); + secp256k1_gej_t prej[table_size]; + prej[0] = *a; + secp256k1_gej_t d; secp256k1_gej_double_var(&d, a); + for (int i=1; i= -((1 << ((w)-1)) - 1)); \ + VERIFY_CHECK((n) <= ((1 << ((w)-1)) - 1)); \ + if ((n) > 0) \ + *(r) = (pre)[((n)-1)/2]; \ + else \ + (neg)((r), &(pre)[(-(n)-1)/2]); \ +} while(0) + +#define ECMULT_TABLE_GET_GEJ(r,pre,n,w) ECMULT_TABLE_GET((r),(pre),(n),(w),secp256k1_gej_neg) +#define ECMULT_TABLE_GET_GE(r,pre,n,w) ECMULT_TABLE_GET((r),(pre),(n),(w),secp256k1_ge_neg) + +typedef struct { + /* For accelerating the computation of a*P + b*G: */ + secp256k1_ge_t pre_g[ECMULT_TABLE_SIZE(WINDOW_G)]; /* odd multiples of the generator */ + secp256k1_ge_t pre_g_128[ECMULT_TABLE_SIZE(WINDOW_G)]; /* odd multiples of 2^128*generator */ +} secp256k1_ecmult_consts_t; + +static const secp256k1_ecmult_consts_t *secp256k1_ecmult_consts = NULL; + +static void secp256k1_ecmult_start(void) { + if (secp256k1_ecmult_consts != NULL) + return; + + /* Allocate the precomputation table. */ + secp256k1_ecmult_consts_t *ret = (secp256k1_ecmult_consts_t*)malloc(sizeof(secp256k1_ecmult_consts_t)); + + /* get the generator */ + const secp256k1_ge_t *g = &secp256k1_ge_consts->g; + secp256k1_gej_t gj; secp256k1_gej_set_ge(&gj, g); + + /* calculate 2^128*generator */ + secp256k1_gej_t g_128j = gj; + for (int i=0; i<128; i++) + secp256k1_gej_double_var(&g_128j, &g_128j); + + /* precompute the tables with odd multiples */ + secp256k1_ecmult_table_precomp_ge_var(ret->pre_g, &gj, WINDOW_G); + secp256k1_ecmult_table_precomp_ge_var(ret->pre_g_128, &g_128j, WINDOW_G); + + /* Set the global pointer to the precomputation table. */ + secp256k1_ecmult_consts = ret; +} + +static void secp256k1_ecmult_stop(void) { + if (secp256k1_ecmult_consts == NULL) + return; + + secp256k1_ecmult_consts_t *c = (secp256k1_ecmult_consts_t*)secp256k1_ecmult_consts; + secp256k1_ecmult_consts = NULL; + free(c); +} + +/** Convert a number to WNAF notation. The number becomes represented by sum(2^i * wnaf[i], i=0..bits), + * with the following guarantees: + * - each wnaf[i] is either 0, or an odd integer between -(1<<(w-1) - 1) and (1<<(w-1) - 1) + * - two non-zero entries in wnaf are separated by at least w-1 zeroes. + * - the index of the highest non-zero entry in wnaf (=return value-1) is at most bits, where + * bits is the number of bits necessary to represent the absolute value of the input. + */ +static int secp256k1_ecmult_wnaf(int *wnaf, const secp256k1_num_t *a, int w) { + int ret = 0; + int zeroes = 0; + secp256k1_num_t x; + secp256k1_num_copy(&x, a); + int sign = 1; + if (secp256k1_num_is_neg(&x)) { + sign = -1; + secp256k1_num_negate(&x); + } + while (!secp256k1_num_is_zero(&x)) { + while (!secp256k1_num_is_odd(&x)) { + zeroes++; + secp256k1_num_shift(&x, 1); + } + int word = secp256k1_num_shift(&x, w); + while (zeroes) { + wnaf[ret++] = 0; + zeroes--; + } + if (word & (1 << (w-1))) { + secp256k1_num_inc(&x); + wnaf[ret++] = sign * (word - (1 << w)); + } else { + wnaf[ret++] = sign * word; + } + zeroes = w-1; + } + return ret; +} + +static void secp256k1_ecmult(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_num_t *na, const secp256k1_num_t *ng) { + const secp256k1_ecmult_consts_t *c = secp256k1_ecmult_consts; + +#ifdef USE_ENDOMORPHISM + secp256k1_num_t na_1, na_lam; + /* split na into na_1 and na_lam (where na = na_1 + na_lam*lambda, and na_1 and na_lam are ~128 bit) */ + secp256k1_gej_split_exp_var(&na_1, &na_lam, na); + + /* build wnaf representation for na_1 and na_lam. */ + int wnaf_na_1[129]; int bits_na_1 = secp256k1_ecmult_wnaf(wnaf_na_1, &na_1, WINDOW_A); + int wnaf_na_lam[129]; int bits_na_lam = secp256k1_ecmult_wnaf(wnaf_na_lam, &na_lam, WINDOW_A); + int bits = bits_na_1; + if (bits_na_lam > bits) bits = bits_na_lam; +#else + /* build wnaf representation for na. */ + int wnaf_na[257]; int bits_na = secp256k1_ecmult_wnaf(wnaf_na, na, WINDOW_A); + int bits = bits_na; +#endif + + /* calculate odd multiples of a */ + secp256k1_gej_t pre_a[ECMULT_TABLE_SIZE(WINDOW_A)]; + secp256k1_ecmult_table_precomp_gej_var(pre_a, a, WINDOW_A); + +#ifdef USE_ENDOMORPHISM + secp256k1_gej_t pre_a_lam[ECMULT_TABLE_SIZE(WINDOW_A)]; + for (int i=0; i bits) bits = bits_ng_1; + if (bits_ng_128 > bits) bits = bits_ng_128; + + secp256k1_gej_set_infinity(r); + secp256k1_gej_t tmpj; + secp256k1_ge_t tmpa; + + for (int i=bits-1; i>=0; i--) { + secp256k1_gej_double_var(r, r); + int n; +#ifdef USE_ENDOMORPHISM + if (i < bits_na_1 && (n = wnaf_na_1[i])) { + ECMULT_TABLE_GET_GEJ(&tmpj, pre_a, n, WINDOW_A); + secp256k1_gej_add_var(r, r, &tmpj); + } + if (i < bits_na_lam && (n = wnaf_na_lam[i])) { + ECMULT_TABLE_GET_GEJ(&tmpj, pre_a_lam, n, WINDOW_A); + secp256k1_gej_add_var(r, r, &tmpj); + } +#else + if (i < bits_na && (n = wnaf_na[i])) { + ECMULT_TABLE_GET_GEJ(&tmpj, pre_a, n, WINDOW_A); + secp256k1_gej_add_var(r, r, &tmpj); + } +#endif + if (i < bits_ng_1 && (n = wnaf_ng_1[i])) { + ECMULT_TABLE_GET_GE(&tmpa, c->pre_g, n, WINDOW_G); + secp256k1_gej_add_ge_var(r, r, &tmpa); + } + if (i < bits_ng_128 && (n = wnaf_ng_128[i])) { + ECMULT_TABLE_GET_GE(&tmpa, c->pre_g_128, n, WINDOW_G); + secp256k1_gej_add_ge_var(r, r, &tmpa); + } + } +} + +#endif diff --git a/src/field.h b/src/field.h new file mode 100644 index 0000000000..c7feead900 --- /dev/null +++ b/src/field.h @@ -0,0 +1,114 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_ +#define _SECP256K1_FIELD_ + +/** Field element module. + * + * Field elements can be represented in several ways, but code accessing + * it (and implementations) need to take certain properaties into account: + * - Each field element can be normalized or not. + * - Each field element has a magnitude, which represents how far away + * its representation is away from normalization. Normalized elements + * always have a magnitude of 1, but a magnitude of 1 doesn't imply + * normality. + */ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#if defined(USE_FIELD_GMP) +#include "field_gmp.h" +#elif defined(USE_FIELD_10X26) +#include "field_10x26.h" +#elif defined(USE_FIELD_5X52) +#include "field_5x52.h" +#else +#error "Please select field implementation" +#endif + +typedef struct { + secp256k1_num_t p; +} secp256k1_fe_consts_t; + +static const secp256k1_fe_consts_t *secp256k1_fe_consts = NULL; + +/** Initialize field element precomputation data. */ +static void secp256k1_fe_start(void); + +/** Unload field element precomputation data. */ +static void secp256k1_fe_stop(void); + +/** Normalize a field element. */ +static void secp256k1_fe_normalize(secp256k1_fe_t *r); + +/** Set a field element equal to a small integer. Resulting field element is normalized. */ +static void secp256k1_fe_set_int(secp256k1_fe_t *r, int a); + +/** Verify whether a field element is zero. Requires the input to be normalized. */ +static int secp256k1_fe_is_zero(const secp256k1_fe_t *a); + +/** Check the "oddness" of a field element. Requires the input to be normalized. */ +static int secp256k1_fe_is_odd(const secp256k1_fe_t *a); + +/** Compare two field elements. Requires both inputs to be normalized */ +static int secp256k1_fe_equal(const secp256k1_fe_t *a, const secp256k1_fe_t *b); + +/** Set a field element equal to 32-byte big endian value. Resulting field element is normalized. */ +static void secp256k1_fe_set_b32(secp256k1_fe_t *r, const unsigned char *a); + +/** Convert a field element to a 32-byte big endian value. Requires the input to be normalized */ +static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe_t *a); + +/** Set a field element equal to the additive inverse of another. Takes a maximum magnitude of the input + * as an argument. The magnitude of the output is one higher. */ +static void secp256k1_fe_negate(secp256k1_fe_t *r, const secp256k1_fe_t *a, int m); + +/** Multiplies the passed field element with a small integer constant. Multiplies the magnitude by that + * small integer. */ +static void secp256k1_fe_mul_int(secp256k1_fe_t *r, int a); + +/** Adds a field element to another. The result has the sum of the inputs' magnitudes as magnitude. */ +static void secp256k1_fe_add(secp256k1_fe_t *r, const secp256k1_fe_t *a); + +/** Sets a field element to be the product of two others. Requires the inputs' magnitudes to be at most 8. + * The output magnitude is 1 (but not guaranteed to be normalized). */ +static void secp256k1_fe_mul(secp256k1_fe_t *r, const secp256k1_fe_t *a, const secp256k1_fe_t *b); + +/** Sets a field element to be the square of another. Requires the input's magnitude to be at most 8. + * The output magnitude is 1 (but not guaranteed to be normalized). */ +static void secp256k1_fe_sqr(secp256k1_fe_t *r, const secp256k1_fe_t *a); + +/** Sets a field element to be the (modular) square root (if any exist) of another. Requires the + * input's magnitude to be at most 8. The output magnitude is 1 (but not guaranteed to be + * normalized). Return value indicates whether a square root was found. */ +static int secp256k1_fe_sqrt(secp256k1_fe_t *r, const secp256k1_fe_t *a); + +/** Sets a field element to be the (modular) inverse of another. Requires the input's magnitude to be + * at most 8. The output magnitude is 1 (but not guaranteed to be normalized). */ +static void secp256k1_fe_inv(secp256k1_fe_t *r, const secp256k1_fe_t *a); + +/** Potentially faster version of secp256k1_fe_inv, without constant-time guarantee. */ +static void secp256k1_fe_inv_var(secp256k1_fe_t *r, const secp256k1_fe_t *a); + +/** Calculate the (modular) inverses of a batch of field elements. Requires the inputs' magnitudes to be + * at most 8. The output magnitudes are 1 (but not guaranteed to be normalized). The inputs and + * outputs must not overlap in memory. */ +static void secp256k1_fe_inv_all(size_t len, secp256k1_fe_t r[len], const secp256k1_fe_t a[len]); + +/** Potentially faster version of secp256k1_fe_inv_all, without constant-time guarantee. */ +static void secp256k1_fe_inv_all_var(size_t len, secp256k1_fe_t r[len], const secp256k1_fe_t a[len]); + + +/** Convert a field element to a hexadecimal string. */ +static void secp256k1_fe_get_hex(char *r, int *rlen, const secp256k1_fe_t *a); + +/** Convert a hexadecimal string to a field element. */ +static void secp256k1_fe_set_hex(secp256k1_fe_t *r, const char *a, int alen); + +#endif diff --git a/src/field_10x26.h b/src/field_10x26.h new file mode 100644 index 0000000000..66fb3f2563 --- /dev/null +++ b/src/field_10x26.h @@ -0,0 +1,21 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_ +#define _SECP256K1_FIELD_REPR_ + +#include + +typedef struct { + /* X = sum(i=0..9, elem[i]*2^26) mod n */ + uint32_t n[10]; +#ifdef VERIFY + int magnitude; + int normalized; +#endif +} secp256k1_fe_t; + +#endif diff --git a/src/field_10x26_impl.h b/src/field_10x26_impl.h new file mode 100644 index 0000000000..c0f1be0b2d --- /dev/null +++ b/src/field_10x26_impl.h @@ -0,0 +1,884 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_IMPL_H_ +#define _SECP256K1_FIELD_REPR_IMPL_H_ + +#include +#include +#include "util.h" +#include "num.h" +#include "field.h" + +static void secp256k1_fe_inner_start(void) {} +static void secp256k1_fe_inner_stop(void) {} + +#ifdef VERIFY +static void secp256k1_fe_verify(const secp256k1_fe_t *a) { + const uint32_t *d = a->n; + int m = a->normalized ? 1 : 2 * a->magnitude, r = 1; + r &= (d[0] <= 0x3FFFFFFUL * m); + r &= (d[1] <= 0x3FFFFFFUL * m); + r &= (d[2] <= 0x3FFFFFFUL * m); + r &= (d[3] <= 0x3FFFFFFUL * m); + r &= (d[4] <= 0x3FFFFFFUL * m); + r &= (d[5] <= 0x3FFFFFFUL * m); + r &= (d[6] <= 0x3FFFFFFUL * m); + r &= (d[7] <= 0x3FFFFFFUL * m); + r &= (d[8] <= 0x3FFFFFFUL * m); + r &= (d[9] <= 0x03FFFFFUL * m); + r &= (a->magnitude >= 0); + if (a->normalized) { + r &= (a->magnitude <= 1); + if (r && (d[9] == 0x03FFFFFUL)) { + uint32_t mid = d[8] & d[7] & d[6] & d[5] & d[4] & d[3] & d[2]; + if (mid == 0x3FFFFFFUL) { + r &= ((d[1] + 0x40UL + ((d[0] + 0x3D1UL) >> 26)) <= 0x3FFFFFFUL); + } + } + } + VERIFY_CHECK(r == 1); +} +#else +static void secp256k1_fe_verify(const secp256k1_fe_t *a) { + (void)a; +} +#endif + +static void secp256k1_fe_normalize(secp256k1_fe_t *r) { + uint32_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4], + t5 = r->n[5], t6 = r->n[6], t7 = r->n[7], t8 = r->n[8], t9 = r->n[9]; + + /* Reduce t9 at the start so there will be at most a single carry from the first pass */ + uint32_t x = t9 >> 22; t9 &= 0x03FFFFFUL; + uint32_t m; + + /* The first pass ensures the magnitude is 1, ... */ + t0 += x * 0x3D1UL; t1 += (x << 6); + t1 += (t0 >> 26); t0 &= 0x3FFFFFFUL; + t2 += (t1 >> 26); t1 &= 0x3FFFFFFUL; + t3 += (t2 >> 26); t2 &= 0x3FFFFFFUL; m = t2; + t4 += (t3 >> 26); t3 &= 0x3FFFFFFUL; m &= t3; + t5 += (t4 >> 26); t4 &= 0x3FFFFFFUL; m &= t4; + t6 += (t5 >> 26); t5 &= 0x3FFFFFFUL; m &= t5; + t7 += (t6 >> 26); t6 &= 0x3FFFFFFUL; m &= t6; + t8 += (t7 >> 26); t7 &= 0x3FFFFFFUL; m &= t7; + t9 += (t8 >> 26); t8 &= 0x3FFFFFFUL; m &= t8; + + /* ... except for a possible carry at bit 22 of t9 (i.e. bit 256 of the field element) */ + VERIFY_CHECK(t9 >> 23 == 0); + + /* At most a single final reduction is needed; check if the value is >= the field characteristic */ + x = (t9 >> 22) | ((t9 == 0x03FFFFFUL) & (m == 0x3FFFFFFUL) + & ((t1 + 0x40UL + ((t0 + 0x3D1UL) >> 26)) > 0x3FFFFFFUL)); + + /* Apply the final reduction (for constant-time behaviour, we do it always) */ + t0 += x * 0x3D1UL; t1 += (x << 6); + t1 += (t0 >> 26); t0 &= 0x3FFFFFFUL; + t2 += (t1 >> 26); t1 &= 0x3FFFFFFUL; + t3 += (t2 >> 26); t2 &= 0x3FFFFFFUL; + t4 += (t3 >> 26); t3 &= 0x3FFFFFFUL; + t5 += (t4 >> 26); t4 &= 0x3FFFFFFUL; + t6 += (t5 >> 26); t5 &= 0x3FFFFFFUL; + t7 += (t6 >> 26); t6 &= 0x3FFFFFFUL; + t8 += (t7 >> 26); t7 &= 0x3FFFFFFUL; + t9 += (t8 >> 26); t8 &= 0x3FFFFFFUL; + + /* If t9 didn't carry to bit 22 already, then it should have after any final reduction */ + VERIFY_CHECK(t9 >> 22 == x); + + /* Mask off the possible multiple of 2^256 from the final reduction */ + t9 &= 0x03FFFFFUL; + + r->n[0] = t0; r->n[1] = t1; r->n[2] = t2; r->n[3] = t3; r->n[4] = t4; + r->n[5] = t5; r->n[6] = t6; r->n[7] = t7; r->n[8] = t8; r->n[9] = t9; + +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_set_int(secp256k1_fe_t *r, int a) { + r->n[0] = a; + r->n[1] = r->n[2] = r->n[3] = r->n[4] = r->n[5] = r->n[6] = r->n[7] = r->n[8] = r->n[9] = 0; +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static int secp256k1_fe_is_zero(const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + const uint32_t *t = a->n; + return (t[0] | t[1] | t[2] | t[3] | t[4] | t[5] | t[6] | t[7] | t[8] | t[9]) == 0; +} + +SECP256K1_INLINE static int secp256k1_fe_is_odd(const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + return a->n[0] & 1; +} + +SECP256K1_INLINE static void secp256k1_fe_clear(secp256k1_fe_t *a) { +#ifdef VERIFY + a->magnitude = 0; + a->normalized = 1; +#endif + for (int i=0; i<10; i++) { + a->n[i] = 0; + } +} + +SECP256K1_INLINE static int secp256k1_fe_equal(const secp256k1_fe_t *a, const secp256k1_fe_t *b) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + VERIFY_CHECK(b->normalized); + secp256k1_fe_verify(a); + secp256k1_fe_verify(b); +#endif + const uint32_t *t = a->n, *u = b->n; + return ((t[0]^u[0]) | (t[1]^u[1]) | (t[2]^u[2]) | (t[3]^u[3]) | (t[4]^u[4]) + | (t[5]^u[5]) | (t[6]^u[6]) | (t[7]^u[7]) | (t[8]^u[8]) | (t[9]^u[9])) == 0; +} + +static void secp256k1_fe_set_b32(secp256k1_fe_t *r, const unsigned char *a) { + r->n[0] = r->n[1] = r->n[2] = r->n[3] = r->n[4] = 0; + r->n[5] = r->n[6] = r->n[7] = r->n[8] = r->n[9] = 0; + for (int i=0; i<32; i++) { + for (int j=0; j<4; j++) { + int limb = (8*i+2*j)/26; + int shift = (8*i+2*j)%26; + r->n[limb] |= (uint32_t)((a[31-i] >> (2*j)) & 0x3) << shift; + } + } +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +/** Convert a field element to a 32-byte big endian value. Requires the input to be normalized */ +static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + for (int i=0; i<32; i++) { + int c = 0; + for (int j=0; j<4; j++) { + int limb = (8*i+2*j)/26; + int shift = (8*i+2*j)%26; + c |= ((a->n[limb] >> shift) & 0x3) << (2 * j); + } + r[31-i] = c; + } +} + +SECP256K1_INLINE static void secp256k1_fe_negate(secp256k1_fe_t *r, const secp256k1_fe_t *a, int m) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= m); + secp256k1_fe_verify(a); +#endif + r->n[0] = 0x3FFFC2FUL * 2 * (m + 1) - a->n[0]; + r->n[1] = 0x3FFFFBFUL * 2 * (m + 1) - a->n[1]; + r->n[2] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[2]; + r->n[3] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[3]; + r->n[4] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[4]; + r->n[5] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[5]; + r->n[6] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[6]; + r->n[7] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[7]; + r->n[8] = 0x3FFFFFFUL * 2 * (m + 1) - a->n[8]; + r->n[9] = 0x03FFFFFUL * 2 * (m + 1) - a->n[9]; +#ifdef VERIFY + r->magnitude = m + 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_mul_int(secp256k1_fe_t *r, int a) { + r->n[0] *= a; + r->n[1] *= a; + r->n[2] *= a; + r->n[3] *= a; + r->n[4] *= a; + r->n[5] *= a; + r->n[6] *= a; + r->n[7] *= a; + r->n[8] *= a; + r->n[9] *= a; +#ifdef VERIFY + r->magnitude *= a; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_add(secp256k1_fe_t *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + secp256k1_fe_verify(a); +#endif + r->n[0] += a->n[0]; + r->n[1] += a->n[1]; + r->n[2] += a->n[2]; + r->n[3] += a->n[3]; + r->n[4] += a->n[4]; + r->n[5] += a->n[5]; + r->n[6] += a->n[6]; + r->n[7] += a->n[7]; + r->n[8] += a->n[8]; + r->n[9] += a->n[9]; +#ifdef VERIFY + r->magnitude += a->magnitude; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +#ifdef VERIFY +#define VERIFY_BITS(x, n) VERIFY_CHECK(((x) >> (n)) == 0) +#else +#define VERIFY_BITS(x, n) do { } while(0) +#endif + +SECP256K1_INLINE static void secp256k1_fe_mul_inner(const uint32_t *a, const uint32_t *b, uint32_t *r) { + VERIFY_BITS(a[0], 30); + VERIFY_BITS(a[1], 30); + VERIFY_BITS(a[2], 30); + VERIFY_BITS(a[3], 30); + VERIFY_BITS(a[4], 30); + VERIFY_BITS(a[5], 30); + VERIFY_BITS(a[6], 30); + VERIFY_BITS(a[7], 30); + VERIFY_BITS(a[8], 30); + VERIFY_BITS(a[9], 26); + VERIFY_BITS(b[0], 30); + VERIFY_BITS(b[1], 30); + VERIFY_BITS(b[2], 30); + VERIFY_BITS(b[3], 30); + VERIFY_BITS(b[4], 30); + VERIFY_BITS(b[5], 30); + VERIFY_BITS(b[6], 30); + VERIFY_BITS(b[7], 30); + VERIFY_BITS(b[8], 30); + VERIFY_BITS(b[9], 26); + + const uint32_t M = 0x3FFFFFFUL, R0 = 0x3D10UL, R1 = 0x400UL; + /** [... a b c] is a shorthand for ... + a<<52 + b<<26 + c<<0 mod n. + * px is a shorthand for sum(a[i]*b[x-i], i=0..x). + * Note that [x 0 0 0 0 0 0 0 0 0 0] = [x*R1 x*R0]. + */ + + uint64_t c, d; + + d = (uint64_t)a[0] * b[9] + + (uint64_t)a[1] * b[8] + + (uint64_t)a[2] * b[7] + + (uint64_t)a[3] * b[6] + + (uint64_t)a[4] * b[5] + + (uint64_t)a[5] * b[4] + + (uint64_t)a[6] * b[3] + + (uint64_t)a[7] * b[2] + + (uint64_t)a[8] * b[1] + + (uint64_t)a[9] * b[0]; + /* VERIFY_BITS(d, 64); */ + /* [d 0 0 0 0 0 0 0 0 0] = [p9 0 0 0 0 0 0 0 0 0] */ + uint32_t t9 = d & M; d >>= 26; + VERIFY_BITS(t9, 26); + VERIFY_BITS(d, 38); + /* [d t9 0 0 0 0 0 0 0 0 0] = [p9 0 0 0 0 0 0 0 0 0] */ + + c = (uint64_t)a[0] * b[0]; + VERIFY_BITS(c, 60); + /* [d t9 0 0 0 0 0 0 0 0 c] = [p9 0 0 0 0 0 0 0 0 p0] */ + d += (uint64_t)a[1] * b[9] + + (uint64_t)a[2] * b[8] + + (uint64_t)a[3] * b[7] + + (uint64_t)a[4] * b[6] + + (uint64_t)a[5] * b[5] + + (uint64_t)a[6] * b[4] + + (uint64_t)a[7] * b[3] + + (uint64_t)a[8] * b[2] + + (uint64_t)a[9] * b[1]; + VERIFY_BITS(d, 63); + /* [d t9 0 0 0 0 0 0 0 0 c] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + uint64_t u0 = d & M; d >>= 26; c += u0 * R0; + VERIFY_BITS(u0, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 61); + /* [d u0 t9 0 0 0 0 0 0 0 0 c-u0*R0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + uint32_t t0 = c & M; c >>= 26; c += u0 * R1; + VERIFY_BITS(t0, 26); + VERIFY_BITS(c, 37); + /* [d u0 t9 0 0 0 0 0 0 0 c-u0*R1 t0-u0*R0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + + c += (uint64_t)a[0] * b[1] + + (uint64_t)a[1] * b[0]; + VERIFY_BITS(c, 62); + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p10 p9 0 0 0 0 0 0 0 p1 p0] */ + d += (uint64_t)a[2] * b[9] + + (uint64_t)a[3] * b[8] + + (uint64_t)a[4] * b[7] + + (uint64_t)a[5] * b[6] + + (uint64_t)a[6] * b[5] + + (uint64_t)a[7] * b[4] + + (uint64_t)a[8] * b[3] + + (uint64_t)a[9] * b[2]; + VERIFY_BITS(d, 63); + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + uint64_t u1 = d & M; d >>= 26; c += u1 * R0; + VERIFY_BITS(u1, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 63); + /* [d u1 0 t9 0 0 0 0 0 0 0 c-u1*R0 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + uint32_t t1 = c & M; c >>= 26; c += u1 * R1; + VERIFY_BITS(t1, 26); + VERIFY_BITS(c, 38); + /* [d u1 0 t9 0 0 0 0 0 0 c-u1*R1 t1-u1*R0 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + + c += (uint64_t)a[0] * b[2] + + (uint64_t)a[1] * b[1] + + (uint64_t)a[2] * b[0]; + VERIFY_BITS(c, 62); + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + d += (uint64_t)a[3] * b[9] + + (uint64_t)a[4] * b[8] + + (uint64_t)a[5] * b[7] + + (uint64_t)a[6] * b[6] + + (uint64_t)a[7] * b[5] + + (uint64_t)a[8] * b[4] + + (uint64_t)a[9] * b[3]; + VERIFY_BITS(d, 63); + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + uint64_t u2 = d & M; d >>= 26; c += u2 * R0; + VERIFY_BITS(u2, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 63); + /* [d u2 0 0 t9 0 0 0 0 0 0 c-u2*R0 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + uint32_t t2 = c & M; c >>= 26; c += u2 * R1; + VERIFY_BITS(t2, 26); + VERIFY_BITS(c, 38); + /* [d u2 0 0 t9 0 0 0 0 0 c-u2*R1 t2-u2*R0 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[3] + + (uint64_t)a[1] * b[2] + + (uint64_t)a[2] * b[1] + + (uint64_t)a[3] * b[0]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + d += (uint64_t)a[4] * b[9] + + (uint64_t)a[5] * b[8] + + (uint64_t)a[6] * b[7] + + (uint64_t)a[7] * b[6] + + (uint64_t)a[8] * b[5] + + (uint64_t)a[9] * b[4]; + VERIFY_BITS(d, 63); + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + uint64_t u3 = d & M; d >>= 26; c += u3 * R0; + VERIFY_BITS(u3, 26); + VERIFY_BITS(d, 37); + /* VERIFY_BITS(c, 64); */ + /* [d u3 0 0 0 t9 0 0 0 0 0 c-u3*R0 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + uint32_t t3 = c & M; c >>= 26; c += u3 * R1; + VERIFY_BITS(t3, 26); + VERIFY_BITS(c, 39); + /* [d u3 0 0 0 t9 0 0 0 0 c-u3*R1 t3-u3*R0 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[4] + + (uint64_t)a[1] * b[3] + + (uint64_t)a[2] * b[2] + + (uint64_t)a[3] * b[1] + + (uint64_t)a[4] * b[0]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[5] * b[9] + + (uint64_t)a[6] * b[8] + + (uint64_t)a[7] * b[7] + + (uint64_t)a[8] * b[6] + + (uint64_t)a[9] * b[5]; + VERIFY_BITS(d, 62); + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + uint64_t u4 = d & M; d >>= 26; c += u4 * R0; + VERIFY_BITS(u4, 26); + VERIFY_BITS(d, 36); + /* VERIFY_BITS(c, 64); */ + /* [d u4 0 0 0 0 t9 0 0 0 0 c-u4*R0 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + uint32_t t4 = c & M; c >>= 26; c += u4 * R1; + VERIFY_BITS(t4, 26); + VERIFY_BITS(c, 39); + /* [d u4 0 0 0 0 t9 0 0 0 c-u4*R1 t4-u4*R0 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[5] + + (uint64_t)a[1] * b[4] + + (uint64_t)a[2] * b[3] + + (uint64_t)a[3] * b[2] + + (uint64_t)a[4] * b[1] + + (uint64_t)a[5] * b[0]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[6] * b[9] + + (uint64_t)a[7] * b[8] + + (uint64_t)a[8] * b[7] + + (uint64_t)a[9] * b[6]; + VERIFY_BITS(d, 62); + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + uint64_t u5 = d & M; d >>= 26; c += u5 * R0; + VERIFY_BITS(u5, 26); + VERIFY_BITS(d, 36); + /* VERIFY_BITS(c, 64); */ + /* [d u5 0 0 0 0 0 t9 0 0 0 c-u5*R0 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + uint32_t t5 = c & M; c >>= 26; c += u5 * R1; + VERIFY_BITS(t5, 26); + VERIFY_BITS(c, 39); + /* [d u5 0 0 0 0 0 t9 0 0 c-u5*R1 t5-u5*R0 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[6] + + (uint64_t)a[1] * b[5] + + (uint64_t)a[2] * b[4] + + (uint64_t)a[3] * b[3] + + (uint64_t)a[4] * b[2] + + (uint64_t)a[5] * b[1] + + (uint64_t)a[6] * b[0]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[7] * b[9] + + (uint64_t)a[8] * b[8] + + (uint64_t)a[9] * b[7]; + VERIFY_BITS(d, 61); + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u6 = d & M; d >>= 26; c += u6 * R0; + VERIFY_BITS(u6, 26); + VERIFY_BITS(d, 35); + /* VERIFY_BITS(c, 64); */ + /* [d u6 0 0 0 0 0 0 t9 0 0 c-u6*R0 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + uint32_t t6 = c & M; c >>= 26; c += u6 * R1; + VERIFY_BITS(t6, 26); + VERIFY_BITS(c, 39); + /* [d u6 0 0 0 0 0 0 t9 0 c-u6*R1 t6-u6*R0 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[7] + + (uint64_t)a[1] * b[6] + + (uint64_t)a[2] * b[5] + + (uint64_t)a[3] * b[4] + + (uint64_t)a[4] * b[3] + + (uint64_t)a[5] * b[2] + + (uint64_t)a[6] * b[1] + + (uint64_t)a[7] * b[0]; + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x8000007C00000007ULL); + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[8] * b[9] + + (uint64_t)a[9] * b[8]; + VERIFY_BITS(d, 58); + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u7 = d & M; d >>= 26; c += u7 * R0; + VERIFY_BITS(u7, 26); + VERIFY_BITS(d, 32); + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x800001703FFFC2F7ULL); + /* [d u7 0 0 0 0 0 0 0 t9 0 c-u7*R0 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint32_t t7 = c & M; c >>= 26; c += u7 * R1; + VERIFY_BITS(t7, 26); + VERIFY_BITS(c, 38); + /* [d u7 0 0 0 0 0 0 0 t9 c-u7*R1 t7-u7*R0 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)a[0] * b[8] + + (uint64_t)a[1] * b[7] + + (uint64_t)a[2] * b[6] + + (uint64_t)a[3] * b[5] + + (uint64_t)a[4] * b[4] + + (uint64_t)a[5] * b[3] + + (uint64_t)a[6] * b[2] + + (uint64_t)a[7] * b[1] + + (uint64_t)a[8] * b[0]; + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x9000007B80000008ULL); + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[9] * b[9]; + VERIFY_BITS(d, 57); + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u8 = d & M; d >>= 26; c += u8 * R0; + VERIFY_BITS(u8, 26); + VERIFY_BITS(d, 31); + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x9000016FBFFFC2F8ULL); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 t4 t3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + r[3] = t3; + VERIFY_BITS(r[3], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 t4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[4] = t4; + VERIFY_BITS(r[4], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[5] = t5; + VERIFY_BITS(r[5], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[6] = t6; + VERIFY_BITS(r[6], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[7] = t7; + VERIFY_BITS(r[7], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + r[8] = c & M; c >>= 26; c += u8 * R1; + VERIFY_BITS(r[8], 26); + VERIFY_BITS(c, 39); + /* [d u8 0 0 0 0 0 0 0 0 t9+c-u8*R1 r8-u8*R0 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 0 0 t9+c r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += d * R0 + t9; + VERIFY_BITS(c, 45); + /* [d 0 0 0 0 0 0 0 0 0 c-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[9] = c & (M >> 4); c >>= 22; c += d * (R1 << 4); + VERIFY_BITS(r[9], 22); + VERIFY_BITS(c, 46); + /* [d 0 0 0 0 0 0 0 0 r9+((c-d*R1<<4)<<22)-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 -d*R1 r9+(c<<22)-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + d = c * (R0 >> 4) + t0; + VERIFY_BITS(d, 56); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1 d-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[0] = d & M; d >>= 26; + VERIFY_BITS(r[0], 26); + VERIFY_BITS(d, 30); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1+d r0-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += c * (R1 >> 4) + t1; + VERIFY_BITS(d, 53); + VERIFY_CHECK(d <= 0x10000003FFFFBFULL); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 d-c*R1>>4 r0-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [r9 r8 r7 r6 r5 r4 r3 t2 d r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[1] = d & M; d >>= 26; + VERIFY_BITS(r[1], 26); + VERIFY_BITS(d, 27); + VERIFY_CHECK(d <= 0x4000000ULL); + /* [r9 r8 r7 r6 r5 r4 r3 t2+d r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += t2; + VERIFY_BITS(d, 27); + /* [r9 r8 r7 r6 r5 r4 r3 d r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[2] = d; + VERIFY_BITS(r[2], 27); + /* [r9 r8 r7 r6 r5 r4 r3 r2 r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ +} + +SECP256K1_INLINE static void secp256k1_fe_sqr_inner(const uint32_t *a, uint32_t *r) { + VERIFY_BITS(a[0], 30); + VERIFY_BITS(a[1], 30); + VERIFY_BITS(a[2], 30); + VERIFY_BITS(a[3], 30); + VERIFY_BITS(a[4], 30); + VERIFY_BITS(a[5], 30); + VERIFY_BITS(a[6], 30); + VERIFY_BITS(a[7], 30); + VERIFY_BITS(a[8], 30); + VERIFY_BITS(a[9], 26); + + const uint32_t M = 0x3FFFFFFUL, R0 = 0x3D10UL, R1 = 0x400UL; + /** [... a b c] is a shorthand for ... + a<<52 + b<<26 + c<<0 mod n. + * px is a shorthand for sum(a[i]*a[x-i], i=0..x). + * Note that [x 0 0 0 0 0 0 0 0 0 0] = [x*R1 x*R0]. + */ + + uint64_t c, d; + + d = (uint64_t)(a[0]*2) * a[9] + + (uint64_t)(a[1]*2) * a[8] + + (uint64_t)(a[2]*2) * a[7] + + (uint64_t)(a[3]*2) * a[6] + + (uint64_t)(a[4]*2) * a[5]; + /* VERIFY_BITS(d, 64); */ + /* [d 0 0 0 0 0 0 0 0 0] = [p9 0 0 0 0 0 0 0 0 0] */ + uint32_t t9 = d & M; d >>= 26; + VERIFY_BITS(t9, 26); + VERIFY_BITS(d, 38); + /* [d t9 0 0 0 0 0 0 0 0 0] = [p9 0 0 0 0 0 0 0 0 0] */ + + c = (uint64_t)a[0] * a[0]; + VERIFY_BITS(c, 60); + /* [d t9 0 0 0 0 0 0 0 0 c] = [p9 0 0 0 0 0 0 0 0 p0] */ + d += (uint64_t)(a[1]*2) * a[9] + + (uint64_t)(a[2]*2) * a[8] + + (uint64_t)(a[3]*2) * a[7] + + (uint64_t)(a[4]*2) * a[6] + + (uint64_t)a[5] * a[5]; + VERIFY_BITS(d, 63); + /* [d t9 0 0 0 0 0 0 0 0 c] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + uint64_t u0 = d & M; d >>= 26; c += u0 * R0; + VERIFY_BITS(u0, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 61); + /* [d u0 t9 0 0 0 0 0 0 0 0 c-u0*R0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + uint32_t t0 = c & M; c >>= 26; c += u0 * R1; + VERIFY_BITS(t0, 26); + VERIFY_BITS(c, 37); + /* [d u0 t9 0 0 0 0 0 0 0 c-u0*R1 t0-u0*R0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p10 p9 0 0 0 0 0 0 0 0 p0] */ + + c += (uint64_t)(a[0]*2) * a[1]; + VERIFY_BITS(c, 62); + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p10 p9 0 0 0 0 0 0 0 p1 p0] */ + d += (uint64_t)(a[2]*2) * a[9] + + (uint64_t)(a[3]*2) * a[8] + + (uint64_t)(a[4]*2) * a[7] + + (uint64_t)(a[5]*2) * a[6]; + VERIFY_BITS(d, 63); + /* [d 0 t9 0 0 0 0 0 0 0 c t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + uint64_t u1 = d & M; d >>= 26; c += u1 * R0; + VERIFY_BITS(u1, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 63); + /* [d u1 0 t9 0 0 0 0 0 0 0 c-u1*R0 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + uint32_t t1 = c & M; c >>= 26; c += u1 * R1; + VERIFY_BITS(t1, 26); + VERIFY_BITS(c, 38); + /* [d u1 0 t9 0 0 0 0 0 0 c-u1*R1 t1-u1*R0 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p11 p10 p9 0 0 0 0 0 0 0 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[2] + + (uint64_t)a[1] * a[1]; + VERIFY_BITS(c, 62); + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + d += (uint64_t)(a[3]*2) * a[9] + + (uint64_t)(a[4]*2) * a[8] + + (uint64_t)(a[5]*2) * a[7] + + (uint64_t)a[6] * a[6]; + VERIFY_BITS(d, 63); + /* [d 0 0 t9 0 0 0 0 0 0 c t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + uint64_t u2 = d & M; d >>= 26; c += u2 * R0; + VERIFY_BITS(u2, 26); + VERIFY_BITS(d, 37); + VERIFY_BITS(c, 63); + /* [d u2 0 0 t9 0 0 0 0 0 0 c-u2*R0 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + uint32_t t2 = c & M; c >>= 26; c += u2 * R1; + VERIFY_BITS(t2, 26); + VERIFY_BITS(c, 38); + /* [d u2 0 0 t9 0 0 0 0 0 c-u2*R1 t2-u2*R0 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 0 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[3] + + (uint64_t)(a[1]*2) * a[2]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + d += (uint64_t)(a[4]*2) * a[9] + + (uint64_t)(a[5]*2) * a[8] + + (uint64_t)(a[6]*2) * a[7]; + VERIFY_BITS(d, 63); + /* [d 0 0 0 t9 0 0 0 0 0 c t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + uint64_t u3 = d & M; d >>= 26; c += u3 * R0; + VERIFY_BITS(u3, 26); + VERIFY_BITS(d, 37); + /* VERIFY_BITS(c, 64); */ + /* [d u3 0 0 0 t9 0 0 0 0 0 c-u3*R0 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + uint32_t t3 = c & M; c >>= 26; c += u3 * R1; + VERIFY_BITS(t3, 26); + VERIFY_BITS(c, 39); + /* [d u3 0 0 0 t9 0 0 0 0 c-u3*R1 t3-u3*R0 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 0 p3 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[4] + + (uint64_t)(a[1]*2) * a[3] + + (uint64_t)a[2] * a[2]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + d += (uint64_t)(a[5]*2) * a[9] + + (uint64_t)(a[6]*2) * a[8] + + (uint64_t)a[7] * a[7]; + VERIFY_BITS(d, 62); + /* [d 0 0 0 0 t9 0 0 0 0 c t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + uint64_t u4 = d & M; d >>= 26; c += u4 * R0; + VERIFY_BITS(u4, 26); + VERIFY_BITS(d, 36); + /* VERIFY_BITS(c, 64); */ + /* [d u4 0 0 0 0 t9 0 0 0 0 c-u4*R0 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + uint32_t t4 = c & M; c >>= 26; c += u4 * R1; + VERIFY_BITS(t4, 26); + VERIFY_BITS(c, 39); + /* [d u4 0 0 0 0 t9 0 0 0 c-u4*R1 t4-u4*R0 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 0 p4 p3 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[5] + + (uint64_t)(a[1]*2) * a[4] + + (uint64_t)(a[2]*2) * a[3]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)(a[6]*2) * a[9] + + (uint64_t)(a[7]*2) * a[8]; + VERIFY_BITS(d, 62); + /* [d 0 0 0 0 0 t9 0 0 0 c t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + uint64_t u5 = d & M; d >>= 26; c += u5 * R0; + VERIFY_BITS(u5, 26); + VERIFY_BITS(d, 36); + /* VERIFY_BITS(c, 64); */ + /* [d u5 0 0 0 0 0 t9 0 0 0 c-u5*R0 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + uint32_t t5 = c & M; c >>= 26; c += u5 * R1; + VERIFY_BITS(t5, 26); + VERIFY_BITS(c, 39); + /* [d u5 0 0 0 0 0 t9 0 0 c-u5*R1 t5-u5*R0 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 0 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[6] + + (uint64_t)(a[1]*2) * a[5] + + (uint64_t)(a[2]*2) * a[4] + + (uint64_t)a[3] * a[3]; + VERIFY_BITS(c, 63); + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)(a[7]*2) * a[9] + + (uint64_t)a[8] * a[8]; + VERIFY_BITS(d, 61); + /* [d 0 0 0 0 0 0 t9 0 0 c t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u6 = d & M; d >>= 26; c += u6 * R0; + VERIFY_BITS(u6, 26); + VERIFY_BITS(d, 35); + /* VERIFY_BITS(c, 64); */ + /* [d u6 0 0 0 0 0 0 t9 0 0 c-u6*R0 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + uint32_t t6 = c & M; c >>= 26; c += u6 * R1; + VERIFY_BITS(t6, 26); + VERIFY_BITS(c, 39); + /* [d u6 0 0 0 0 0 0 t9 0 c-u6*R1 t6-u6*R0 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 0 p6 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[7] + + (uint64_t)(a[1]*2) * a[6] + + (uint64_t)(a[2]*2) * a[5] + + (uint64_t)(a[3]*2) * a[4]; + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x8000007C00000007ULL); + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)(a[8]*2) * a[9]; + VERIFY_BITS(d, 58); + /* [d 0 0 0 0 0 0 0 t9 0 c t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u7 = d & M; d >>= 26; c += u7 * R0; + VERIFY_BITS(u7, 26); + VERIFY_BITS(d, 32); + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x800001703FFFC2F7ULL); + /* [d u7 0 0 0 0 0 0 0 t9 0 c-u7*R0 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint32_t t7 = c & M; c >>= 26; c += u7 * R1; + VERIFY_BITS(t7, 26); + VERIFY_BITS(c, 38); + /* [d u7 0 0 0 0 0 0 0 t9 c-u7*R1 t7-u7*R0 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 0 p7 p6 p5 p4 p3 p2 p1 p0] */ + + c += (uint64_t)(a[0]*2) * a[8] + + (uint64_t)(a[1]*2) * a[7] + + (uint64_t)(a[2]*2) * a[6] + + (uint64_t)(a[3]*2) * a[5] + + (uint64_t)a[4] * a[4]; + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x9000007B80000008ULL); + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += (uint64_t)a[9] * a[9]; + VERIFY_BITS(d, 57); + /* [d 0 0 0 0 0 0 0 0 t9 c t7 t6 t5 t4 t3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + uint64_t u8 = d & M; d >>= 26; c += u8 * R0; + VERIFY_BITS(u8, 26); + VERIFY_BITS(d, 31); + /* VERIFY_BITS(c, 64); */ + VERIFY_CHECK(c <= 0x9000016FBFFFC2F8ULL); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 t4 t3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + r[3] = t3; + VERIFY_BITS(r[3], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 t4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[4] = t4; + VERIFY_BITS(r[4], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 t5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[5] = t5; + VERIFY_BITS(r[5], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 t6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[6] = t6; + VERIFY_BITS(r[6], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 t7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[7] = t7; + VERIFY_BITS(r[7], 26); + /* [d u8 0 0 0 0 0 0 0 0 t9 c-u8*R0 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + r[8] = c & M; c >>= 26; c += u8 * R1; + VERIFY_BITS(r[8], 26); + VERIFY_BITS(c, 39); + /* [d u8 0 0 0 0 0 0 0 0 t9+c-u8*R1 r8-u8*R0 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 0 0 t9+c r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += d * R0 + t9; + VERIFY_BITS(c, 45); + /* [d 0 0 0 0 0 0 0 0 0 c-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[9] = c & (M >> 4); c >>= 22; c += d * (R1 << 4); + VERIFY_BITS(r[9], 22); + VERIFY_BITS(c, 46); + /* [d 0 0 0 0 0 0 0 0 r9+((c-d*R1<<4)<<22)-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [d 0 0 0 0 0 0 0 -d*R1 r9+(c<<22)-d*R0 r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1 t0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + d = c * (R0 >> 4) + t0; + VERIFY_BITS(d, 56); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1 d-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[0] = d & M; d >>= 26; + VERIFY_BITS(r[0], 26); + VERIFY_BITS(d, 30); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 t1+d r0-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += c * (R1 >> 4) + t1; + VERIFY_BITS(d, 53); + VERIFY_CHECK(d <= 0x10000003FFFFBFULL); + /* [r9+(c<<22) r8 r7 r6 r5 r4 r3 t2 d-c*R1>>4 r0-c*R0>>4] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + /* [r9 r8 r7 r6 r5 r4 r3 t2 d r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[1] = d & M; d >>= 26; + VERIFY_BITS(r[1], 26); + VERIFY_BITS(d, 27); + VERIFY_CHECK(d <= 0x4000000ULL); + /* [r9 r8 r7 r6 r5 r4 r3 t2+d r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + d += t2; + VERIFY_BITS(d, 27); + /* [r9 r8 r7 r6 r5 r4 r3 d r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[2] = d; + VERIFY_BITS(r[2], 27); + /* [r9 r8 r7 r6 r5 r4 r3 r2 r1 r0] = [p18 p17 p16 p15 p14 p13 p12 p11 p10 p9 p8 p7 p6 p5 p4 p3 p2 p1 p0] */ +} + + +static void secp256k1_fe_mul(secp256k1_fe_t *r, const secp256k1_fe_t *a, const secp256k1_fe_t *b) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= 8); + VERIFY_CHECK(b->magnitude <= 8); + secp256k1_fe_verify(a); + secp256k1_fe_verify(b); +#endif + secp256k1_fe_mul_inner(a->n, b->n, r->n); +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +static void secp256k1_fe_sqr(secp256k1_fe_t *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= 8); + secp256k1_fe_verify(a); +#endif + secp256k1_fe_sqr_inner(a->n, r->n); +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +#endif diff --git a/src/field_5x52.h b/src/field_5x52.h new file mode 100644 index 0000000000..aeb0a6a1e8 --- /dev/null +++ b/src/field_5x52.h @@ -0,0 +1,21 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_ +#define _SECP256K1_FIELD_REPR_ + +#include + +typedef struct { + /* X = sum(i=0..4, elem[i]*2^52) mod n */ + uint64_t n[5]; +#ifdef VERIFY + int magnitude; + int normalized; +#endif +} secp256k1_fe_t; + +#endif diff --git a/src/field_5x52_asm.asm b/src/field_5x52_asm.asm new file mode 100644 index 0000000000..5e785f7630 --- /dev/null +++ b/src/field_5x52_asm.asm @@ -0,0 +1,469 @@ + ;; Added by Diederik Huys, March 2013 + ;; + ;; Provided public procedures: + ;; secp256k1_fe_mul_inner + ;; secp256k1_fe_sqr_inner + ;; + ;; Needed tools: YASM (http://yasm.tortall.net) + ;; + ;; + + BITS 64 + +%ifidn __OUTPUT_FORMAT__,macho64 +%define SYM(x) _ %+ x +%else +%define SYM(x) x +%endif + + ;; Procedure ExSetMult + ;; Register Layout: + ;; INPUT: rdi = a->n + ;; rsi = b->n + ;; rdx = r->a + ;; + ;; INTERNAL: rdx:rax = multiplication accumulator + ;; r9:r8 = c + ;; r10-r13 = t0-t3 + ;; r14 = b.n[0] / t4 + ;; r15 = b.n[1] / t5 + ;; rbx = b.n[2] / t6 + ;; rcx = b.n[3] / t7 + ;; rbp = Constant 0FFFFFFFFFFFFFh / t8 + ;; rsi = b.n / b.n[4] / t9 + + GLOBAL SYM(secp256k1_fe_mul_inner) + ALIGN 32 +SYM(secp256k1_fe_mul_inner): + push rbp + push rbx + push r12 + push r13 + push r14 + push r15 + push rdx + mov r14,[rsi+8*0] ; preload b.n[0]. This will be the case until + ; b.n[0] is no longer needed, then we reassign + ; r14 to t4 + ;; c=a.n[0] * b.n[0] + mov rax,[rdi+0*8] ; load a.n[0] + mov rbp,0FFFFFFFFFFFFFh + mul r14 ; rdx:rax=a.n[0]*b.n[0] + mov r15,[rsi+1*8] + mov r10,rbp ; load modulus into target register for t0 + mov r8,rax + and r10,rax ; only need lower qword of c + shrd r8,rdx,52 + xor r9,r9 ; c < 2^64, so we ditch the HO part + + ;; c+=a.n[0] * b.n[1] + a.n[1] * b.n[0] + mov rax,[rdi+0*8] + mul r15 + add r8,rax + adc r9,rdx + + mov rax,[rdi+1*8] + mul r14 + mov r11,rbp + mov rbx,[rsi+2*8] + add r8,rax + adc r9,rdx + and r11,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[0 1 2] * b.n[2 1 0] + mov rax,[rdi+0*8] + mul rbx + add r8,rax + adc r9,rdx + + mov rax,[rdi+1*8] + mul r15 + add r8,rax + adc r9,rdx + + mov rax,[rdi+2*8] + mul r14 + mov r12,rbp + mov rcx,[rsi+3*8] + add r8,rax + adc r9,rdx + and r12,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[0 1 2 3] * b.n[3 2 1 0] + mov rax,[rdi+0*8] + mul rcx + add r8,rax + adc r9,rdx + + mov rax,[rdi+1*8] + mul rbx + add r8,rax + adc r9,rdx + + mov rax,[rdi+2*8] + mul r15 + add r8,rax + adc r9,rdx + + mov rax,[rdi+3*8] + mul r14 + mov r13,rbp + mov rsi,[rsi+4*8] ; load b.n[4] and destroy pointer + add r8,rax + adc r9,rdx + and r13,r8 + + shrd r8,r9,52 + xor r9,r9 + + + ;; c+=a.n[0 1 2 3 4] * b.n[4 3 2 1 0] + mov rax,[rdi+0*8] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,[rdi+1*8] + mul rcx + add r8,rax + adc r9,rdx + + mov rax,[rdi+2*8] + mul rbx + add r8,rax + adc r9,rdx + + mov rax,[rdi+3*8] + mul r15 + add r8,rax + adc r9,rdx + + mov rax,[rdi+4*8] + mul r14 + mov r14,rbp ; load modulus into t4 and destroy a.n[0] + add r8,rax + adc r9,rdx + and r14,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[1 2 3 4] * b.n[4 3 2 1] + mov rax,[rdi+1*8] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,[rdi+2*8] + mul rcx + add r8,rax + adc r9,rdx + + mov rax,[rdi+3*8] + mul rbx + add r8,rax + adc r9,rdx + + mov rax,[rdi+4*8] + mul r15 + mov r15,rbp + add r8,rax + adc r9,rdx + + and r15,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[2 3 4] * b.n[4 3 2] + mov rax,[rdi+2*8] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,[rdi+3*8] + mul rcx + add r8,rax + adc r9,rdx + + mov rax,[rdi+4*8] + mul rbx + mov rbx,rbp + add r8,rax + adc r9,rdx + + and rbx,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[3 4] * b.n[4 3] + mov rax,[rdi+3*8] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,[rdi+4*8] + mul rcx + mov rcx,rbp + add r8,rax + adc r9,rdx + and rcx,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[4] * b.n[4] + mov rax,[rdi+4*8] + mul rsi + ;; mov rbp,rbp ; modulus already there! + add r8,rax + adc r9,rdx + and rbp,r8 + shrd r8,r9,52 + xor r9,r9 + + mov rsi,r8 ; load c into t9 and destroy b.n[4] + + ;; ******************************************************* +common_exit_norm: + mov rdi,01000003D10h ; load constant + + mov rax,r15 ; get t5 + mul rdi + add rax,r10 ; +t0 + adc rdx,0 + mov r10,0FFFFFFFFFFFFFh ; modulus. Sadly, we ran out of registers! + mov r8,rax ; +c + and r10,rax + shrd r8,rdx,52 + xor r9,r9 + + mov rax,rbx ; get t6 + mul rdi + add rax,r11 ; +t1 + adc rdx,0 + mov r11,0FFFFFFFFFFFFFh ; modulus + add r8,rax ; +c + adc r9,rdx + and r11,r8 + shrd r8,r9,52 + xor r9,r9 + + mov rax,rcx ; get t7 + mul rdi + add rax,r12 ; +t2 + adc rdx,0 + pop rbx ; retrieve pointer to this.n + mov r12,0FFFFFFFFFFFFFh ; modulus + add r8,rax ; +c + adc r9,rdx + and r12,r8 + mov [rbx+2*8],r12 ; mov into this.n[2] + shrd r8,r9,52 + xor r9,r9 + + mov rax,rbp ; get t8 + mul rdi + add rax,r13 ; +t3 + adc rdx,0 + mov r13,0FFFFFFFFFFFFFh ; modulus + add r8,rax ; +c + adc r9,rdx + and r13,r8 + mov [rbx+3*8],r13 ; -> this.n[3] + shrd r8,r9,52 + xor r9,r9 + + mov rax,rsi ; get t9 + mul rdi + add rax,r14 ; +t4 + adc rdx,0 + mov r14,0FFFFFFFFFFFFh ; !!! + add r8,rax ; +c + adc r9,rdx + and r14,r8 + mov [rbx+4*8],r14 ; -> this.n[4] + shrd r8,r9,48 ; !!! + xor r9,r9 + + mov rax,01000003D1h + mul r8 + add rax,r10 + adc rdx,0 + mov r10,0FFFFFFFFFFFFFh ; modulus + mov r8,rax + and rax,r10 + shrd r8,rdx,52 + mov [rbx+0*8],rax ; -> this.n[0] + add r8,r11 + mov [rbx+1*8],r8 ; -> this.n[1] + + pop r15 + pop r14 + pop r13 + pop r12 + pop rbx + pop rbp + ret + + + ;; PROC ExSetSquare + ;; Register Layout: + ;; INPUT: rdi = a.n + ;; rsi = this.a + ;; INTERNAL: rdx:rax = multiplication accumulator + ;; r9:r8 = c + ;; r10-r13 = t0-t3 + ;; r14 = a.n[0] / t4 + ;; r15 = a.n[1] / t5 + ;; rbx = a.n[2] / t6 + ;; rcx = a.n[3] / t7 + ;; rbp = 0FFFFFFFFFFFFFh / t8 + ;; rsi = a.n[4] / t9 + GLOBAL SYM(secp256k1_fe_sqr_inner) + ALIGN 32 +SYM(secp256k1_fe_sqr_inner): + push rbp + push rbx + push r12 + push r13 + push r14 + push r15 + push rsi + mov rbp,0FFFFFFFFFFFFFh + + ;; c=a.n[0] * a.n[0] + mov r14,[rdi+0*8] ; r14=a.n[0] + mov r10,rbp ; modulus + mov rax,r14 + mul rax + mov r15,[rdi+1*8] ; a.n[1] + add r14,r14 ; r14=2*a.n[0] + mov r8,rax + and r10,rax ; only need lower qword + shrd r8,rdx,52 + xor r9,r9 + + ;; c+=2*a.n[0] * a.n[1] + mov rax,r14 ; r14=2*a.n[0] + mul r15 + mov rbx,[rdi+2*8] ; rbx=a.n[2] + mov r11,rbp ; modulus + add r8,rax + adc r9,rdx + and r11,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[0]*a.n[2]+a.n[1]*a.n[1] + mov rax,r14 + mul rbx + add r8,rax + adc r9,rdx + + mov rax,r15 + mov r12,rbp ; modulus + mul rax + mov rcx,[rdi+3*8] ; rcx=a.n[3] + add r15,r15 ; r15=a.n[1]*2 + add r8,rax + adc r9,rdx + and r12,r8 ; only need lower dword + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[0]*a.n[3]+2*a.n[1]*a.n[2] + mov rax,r14 + mul rcx + add r8,rax + adc r9,rdx + + mov rax,r15 ; rax=2*a.n[1] + mov r13,rbp ; modulus + mul rbx + mov rsi,[rdi+4*8] ; rsi=a.n[4] + add r8,rax + adc r9,rdx + and r13,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[0]*a.n[4]+2*a.n[1]*a.n[3]+a.n[2]*a.n[2] + mov rax,r14 ; last time we need 2*a.n[0] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,r15 + mul rcx + mov r14,rbp ; modulus + add r8,rax + adc r9,rdx + + mov rax,rbx + mul rax + add rbx,rbx ; rcx=2*a.n[2] + add r8,rax + adc r9,rdx + and r14,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[1]*a.n[4]+2*a.n[2]*a.n[3] + mov rax,r15 ; last time we need 2*a.n[1] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,rbx + mul rcx + mov r15,rbp ; modulus + add r8,rax + adc r9,rdx + and r15,r8 + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[2]*a.n[4]+a.n[3]*a.n[3] + mov rax,rbx ; last time we need 2*a.n[2] + mul rsi + add r8,rax + adc r9,rdx + + mov rax,rcx ; a.n[3] + mul rax + mov rbx,rbp ; modulus + add r8,rax + adc r9,rdx + and rbx,r8 ; only need lower dword + lea rax,[2*rcx] + shrd r8,r9,52 + xor r9,r9 + + ;; c+=2*a.n[3]*a.n[4] + mul rsi + mov rcx,rbp ; modulus + add r8,rax + adc r9,rdx + and rcx,r8 ; only need lower dword + shrd r8,r9,52 + xor r9,r9 + + ;; c+=a.n[4]*a.n[4] + mov rax,rsi + mul rax + ;; mov rbp,rbp ; modulus is already there! + add r8,rax + adc r9,rdx + and rbp,r8 + shrd r8,r9,52 + xor r9,r9 + + mov rsi,r8 + + ;; ******************************************************* + jmp common_exit_norm + end + + diff --git a/src/field_5x52_asm_impl.h b/src/field_5x52_asm_impl.h new file mode 100644 index 0000000000..f29605b11b --- /dev/null +++ b/src/field_5x52_asm_impl.h @@ -0,0 +1,13 @@ +/********************************************************************** + * Copyright (c) 2013 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_INNER5X52_IMPL_H_ +#define _SECP256K1_FIELD_INNER5X52_IMPL_H_ + +void __attribute__ ((sysv_abi)) secp256k1_fe_mul_inner(const uint64_t *a, const uint64_t *b, uint64_t *r); +void __attribute__ ((sysv_abi)) secp256k1_fe_sqr_inner(const uint64_t *a, uint64_t *r); + +#endif diff --git a/src/field_5x52_impl.h b/src/field_5x52_impl.h new file mode 100644 index 0000000000..d1b06d05a4 --- /dev/null +++ b/src/field_5x52_impl.h @@ -0,0 +1,260 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_IMPL_H_ +#define _SECP256K1_FIELD_REPR_IMPL_H_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#include +#include "util.h" +#include "num.h" +#include "field.h" + +#if defined(USE_FIELD_5X52_ASM) +#include "field_5x52_asm_impl.h" +#elif defined(USE_FIELD_5X52_INT128) +#include "field_5x52_int128_impl.h" +#else +#error "Please select field_5x52 implementation" +#endif + +/** Implements arithmetic modulo FFFFFFFF FFFFFFFF FFFFFFFF FFFFFFFF FFFFFFFF FFFFFFFF FFFFFFFE FFFFFC2F, + * represented as 5 uint64_t's in base 2^52. The values are allowed to contain >52 each. In particular, + * each FieldElem has a 'magnitude' associated with it. Internally, a magnitude M means each element + * is at most M*(2^53-1), except the most significant one, which is limited to M*(2^49-1). All operations + * accept any input with magnitude at most M, and have different rules for propagating magnitude to their + * output. + */ + +static void secp256k1_fe_inner_start(void) {} +static void secp256k1_fe_inner_stop(void) {} + +#ifdef VERIFY +static void secp256k1_fe_verify(const secp256k1_fe_t *a) { + const uint64_t *d = a->n; + int m = a->normalized ? 1 : 2 * a->magnitude, r = 1; + r &= (d[0] <= 0xFFFFFFFFFFFFFULL * m); + r &= (d[1] <= 0xFFFFFFFFFFFFFULL * m); + r &= (d[2] <= 0xFFFFFFFFFFFFFULL * m); + r &= (d[3] <= 0xFFFFFFFFFFFFFULL * m); + r &= (d[4] <= 0x0FFFFFFFFFFFFULL * m); + r &= (a->magnitude >= 0); + if (a->normalized) { + r &= (a->magnitude <= 1); + if (r && (d[4] == 0x0FFFFFFFFFFFFULL) && ((d[3] & d[2] & d[1]) == 0xFFFFFFFFFFFFFULL)) { + r &= (d[0] < 0xFFFFEFFFFFC2FULL); + } + } + VERIFY_CHECK(r == 1); +} +#else +static void secp256k1_fe_verify(const secp256k1_fe_t *a) { + (void)a; +} +#endif + +static void secp256k1_fe_normalize(secp256k1_fe_t *r) { + uint64_t t0 = r->n[0], t1 = r->n[1], t2 = r->n[2], t3 = r->n[3], t4 = r->n[4]; + + /* Reduce t4 at the start so there will be at most a single carry from the first pass */ + uint64_t x = t4 >> 48; t4 &= 0x0FFFFFFFFFFFFULL; + uint64_t m; + + /* The first pass ensures the magnitude is 1, ... */ + t0 += x * 0x1000003D1ULL; + t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL; + t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; m = t1; + t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; m &= t2; + t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; m &= t3; + + /* ... except for a possible carry at bit 48 of t4 (i.e. bit 256 of the field element) */ + VERIFY_CHECK(t4 >> 49 == 0); + + /* At most a single final reduction is needed; check if the value is >= the field characteristic */ + x = (t4 >> 48) | ((t4 == 0x0FFFFFFFFFFFFULL) & (m == 0xFFFFFFFFFFFFFULL) + & (t0 >= 0xFFFFEFFFFFC2FULL)); + + /* Apply the final reduction (for constant-time behaviour, we do it always) */ + t0 += x * 0x1000003D1ULL; + t1 += (t0 >> 52); t0 &= 0xFFFFFFFFFFFFFULL; + t2 += (t1 >> 52); t1 &= 0xFFFFFFFFFFFFFULL; + t3 += (t2 >> 52); t2 &= 0xFFFFFFFFFFFFFULL; + t4 += (t3 >> 52); t3 &= 0xFFFFFFFFFFFFFULL; + + /* If t4 didn't carry to bit 48 already, then it should have after any final reduction */ + VERIFY_CHECK(t4 >> 48 == x); + + /* Mask off the possible multiple of 2^256 from the final reduction */ + t4 &= 0x0FFFFFFFFFFFFULL; + + r->n[0] = t0; r->n[1] = t1; r->n[2] = t2; r->n[3] = t3; r->n[4] = t4; + +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_set_int(secp256k1_fe_t *r, int a) { + r->n[0] = a; + r->n[1] = r->n[2] = r->n[3] = r->n[4] = 0; +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static int secp256k1_fe_is_zero(const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + const uint64_t *t = a->n; + return (t[0] | t[1] | t[2] | t[3] | t[4]) == 0; +} + +SECP256K1_INLINE static int secp256k1_fe_is_odd(const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + return a->n[0] & 1; +} + +SECP256K1_INLINE static void secp256k1_fe_clear(secp256k1_fe_t *a) { +#ifdef VERIFY + a->magnitude = 0; + a->normalized = 1; +#endif + for (int i=0; i<5; i++) { + a->n[i] = 0; + } +} + +SECP256K1_INLINE static int secp256k1_fe_equal(const secp256k1_fe_t *a, const secp256k1_fe_t *b) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + VERIFY_CHECK(b->normalized); + secp256k1_fe_verify(a); + secp256k1_fe_verify(b); +#endif + const uint64_t *t = a->n, *u = b->n; + return ((t[0]^u[0]) | (t[1]^u[1]) | (t[2]^u[2]) | (t[3]^u[3]) | (t[4]^u[4])) == 0; +} + +static void secp256k1_fe_set_b32(secp256k1_fe_t *r, const unsigned char *a) { + r->n[0] = r->n[1] = r->n[2] = r->n[3] = r->n[4] = 0; + for (int i=0; i<32; i++) { + for (int j=0; j<2; j++) { + int limb = (8*i+4*j)/52; + int shift = (8*i+4*j)%52; + r->n[limb] |= (uint64_t)((a[31-i] >> (4*j)) & 0xF) << shift; + } + } +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 1; + secp256k1_fe_verify(r); +#endif +} + +/** Convert a field element to a 32-byte big endian value. Requires the input to be normalized */ +static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->normalized); + secp256k1_fe_verify(a); +#endif + for (int i=0; i<32; i++) { + int c = 0; + for (int j=0; j<2; j++) { + int limb = (8*i+4*j)/52; + int shift = (8*i+4*j)%52; + c |= ((a->n[limb] >> shift) & 0xF) << (4 * j); + } + r[31-i] = c; + } +} + +SECP256K1_INLINE static void secp256k1_fe_negate(secp256k1_fe_t *r, const secp256k1_fe_t *a, int m) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= m); + secp256k1_fe_verify(a); +#endif + r->n[0] = 0xFFFFEFFFFFC2FULL * 2 * (m + 1) - a->n[0]; + r->n[1] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[1]; + r->n[2] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[2]; + r->n[3] = 0xFFFFFFFFFFFFFULL * 2 * (m + 1) - a->n[3]; + r->n[4] = 0x0FFFFFFFFFFFFULL * 2 * (m + 1) - a->n[4]; +#ifdef VERIFY + r->magnitude = m + 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_mul_int(secp256k1_fe_t *r, int a) { + r->n[0] *= a; + r->n[1] *= a; + r->n[2] *= a; + r->n[3] *= a; + r->n[4] *= a; +#ifdef VERIFY + r->magnitude *= a; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_add(secp256k1_fe_t *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + secp256k1_fe_verify(a); +#endif + r->n[0] += a->n[0]; + r->n[1] += a->n[1]; + r->n[2] += a->n[2]; + r->n[3] += a->n[3]; + r->n[4] += a->n[4]; +#ifdef VERIFY + r->magnitude += a->magnitude; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +static void secp256k1_fe_mul(secp256k1_fe_t *r, const secp256k1_fe_t *a, const secp256k1_fe_t *b) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= 8); + VERIFY_CHECK(b->magnitude <= 8); + secp256k1_fe_verify(a); + secp256k1_fe_verify(b); +#endif + secp256k1_fe_mul_inner(a->n, b->n, r->n); +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +static void secp256k1_fe_sqr(secp256k1_fe_t *r, const secp256k1_fe_t *a) { +#ifdef VERIFY + VERIFY_CHECK(a->magnitude <= 8); + secp256k1_fe_verify(a); +#endif + secp256k1_fe_sqr_inner(a->n, r->n); +#ifdef VERIFY + r->magnitude = 1; + r->normalized = 0; + secp256k1_fe_verify(r); +#endif +} + +#endif diff --git a/src/field_5x52_int128_impl.h b/src/field_5x52_int128_impl.h new file mode 100644 index 0000000000..c476428672 --- /dev/null +++ b/src/field_5x52_int128_impl.h @@ -0,0 +1,279 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_INNER5X52_IMPL_H_ +#define _SECP256K1_FIELD_INNER5X52_IMPL_H_ + +#include + +#ifdef VERIFY +#define VERIFY_BITS(x, n) VERIFY_CHECK(((x) >> (n)) == 0) +#else +#define VERIFY_BITS(x, n) do { } while(0) +#endif + +SECP256K1_INLINE static void secp256k1_fe_mul_inner(const uint64_t *a, const uint64_t *b, uint64_t *r) { + VERIFY_BITS(a[0], 56); + VERIFY_BITS(a[1], 56); + VERIFY_BITS(a[2], 56); + VERIFY_BITS(a[3], 56); + VERIFY_BITS(a[4], 52); + VERIFY_BITS(b[0], 56); + VERIFY_BITS(b[1], 56); + VERIFY_BITS(b[2], 56); + VERIFY_BITS(b[3], 56); + VERIFY_BITS(b[4], 52); + + const uint64_t M = 0xFFFFFFFFFFFFFULL, R = 0x1000003D10ULL; + /* [... a b c] is a shorthand for ... + a<<104 + b<<52 + c<<0 mod n. + * px is a shorthand for sum(a[i]*b[x-i], i=0..x). + * Note that [x 0 0 0 0 0] = [x*R]. + */ + + __int128 c, d; + + d = (__int128)a[0] * b[3] + + (__int128)a[1] * b[2] + + (__int128)a[2] * b[1] + + (__int128)a[3] * b[0]; + VERIFY_BITS(d, 114); + /* [d 0 0 0] = [p3 0 0 0] */ + c = (__int128)a[4] * b[4]; + VERIFY_BITS(c, 112); + /* [c 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + d += (c & M) * R; c >>= 52; + VERIFY_BITS(d, 115); + VERIFY_BITS(c, 60); + /* [c 0 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + uint64_t t3 = d & M; d >>= 52; + VERIFY_BITS(t3, 52); + VERIFY_BITS(d, 63); + /* [c 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + + d += (__int128)a[0] * b[4] + + (__int128)a[1] * b[3] + + (__int128)a[2] * b[2] + + (__int128)a[3] * b[1] + + (__int128)a[4] * b[0]; + VERIFY_BITS(d, 115); + /* [c 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + d += c * R; + VERIFY_BITS(d, 116); + /* [d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + uint64_t t4 = d & M; d >>= 52; + VERIFY_BITS(t4, 52); + VERIFY_BITS(d, 64); + /* [d t4 t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + uint64_t tx = (t4 >> 48); t4 &= (M >> 4); + VERIFY_BITS(tx, 4); + VERIFY_BITS(t4, 48); + /* [d t4+(tx<<48) t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + + c = (__int128)a[0] * b[0]; + VERIFY_BITS(c, 112); + /* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 0 p4 p3 0 0 p0] */ + d += (__int128)a[1] * b[4] + + (__int128)a[2] * b[3] + + (__int128)a[3] * b[2] + + (__int128)a[4] * b[1]; + VERIFY_BITS(d, 115); + /* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + uint64_t u0 = d & M; d >>= 52; + VERIFY_BITS(u0, 52); + VERIFY_BITS(d, 63); + /* [d u0 t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + /* [d 0 t4+(tx<<48)+(u0<<52) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + u0 = (u0 << 4) | tx; + VERIFY_BITS(u0, 56); + /* [d 0 t4+(u0<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + c += (__int128)u0 * (R >> 4); + VERIFY_BITS(c, 115); + /* [d 0 t4 t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + uint64_t t0 = c & M; c >>= 52; + VERIFY_BITS(t0, 52); + VERIFY_BITS(c, 61); + /* [d 0 t4 t3 0 c t0] = [p8 0 0 p5 p4 p3 0 0 p0] */ + + c += (__int128)a[0] * b[1] + + (__int128)a[1] * b[0]; + VERIFY_BITS(c, 114); + /* [d 0 t4 t3 0 c t0] = [p8 0 0 p5 p4 p3 0 p1 p0] */ + d += (__int128)a[2] * b[4] + + (__int128)a[3] * b[3] + + (__int128)a[4] * b[2]; + VERIFY_BITS(d, 114); + /* [d 0 t4 t3 0 c t0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + c += (d & M) * R; d >>= 52; + VERIFY_BITS(c, 115); + VERIFY_BITS(d, 62); + /* [d 0 0 t4 t3 0 c t0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + uint64_t t1 = c & M; c >>= 52; + VERIFY_BITS(t1, 52); + VERIFY_BITS(c, 63); + /* [d 0 0 t4 t3 c t1 t0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + + c += (__int128)a[0] * b[2] + + (__int128)a[1] * b[1] + + (__int128)a[2] * b[0]; + VERIFY_BITS(c, 114); + /* [d 0 0 t4 t3 c t1 t0] = [p8 0 p6 p5 p4 p3 p2 p1 p0] */ + d += (__int128)a[3] * b[4] + + (__int128)a[4] * b[3]; + VERIFY_BITS(d, 114); + /* [d 0 0 t4 t3 c t1 t0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += (d & M) * R; d >>= 52; + VERIFY_BITS(c, 115); + VERIFY_BITS(d, 62); + /* [d 0 0 0 t4 t3 c t1 t0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + r[0] = t0; + VERIFY_BITS(r[0], 52); + /* [d 0 0 0 t4 t3 c t1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[1] = t1; + VERIFY_BITS(r[1], 52); + /* [d 0 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[2] = c & M; c >>= 52; + VERIFY_BITS(r[2], 52); + VERIFY_BITS(c, 63); + /* [d 0 0 0 t4 t3+c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += d * R + t3;; + VERIFY_BITS(c, 100); + /* [t4 c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[3] = c & M; c >>= 52; + VERIFY_BITS(r[3], 52); + VERIFY_BITS(c, 48); + /* [t4+c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += t4; + VERIFY_BITS(c, 49); + /* [c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[4] = c; + VERIFY_BITS(r[4], 49); + /* [r4 r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ +} + +SECP256K1_INLINE static void secp256k1_fe_sqr_inner(const uint64_t *a, uint64_t *r) { + VERIFY_BITS(a[0], 56); + VERIFY_BITS(a[1], 56); + VERIFY_BITS(a[2], 56); + VERIFY_BITS(a[3], 56); + VERIFY_BITS(a[4], 52); + + const uint64_t M = 0xFFFFFFFFFFFFFULL, R = 0x1000003D10ULL; + /** [... a b c] is a shorthand for ... + a<<104 + b<<52 + c<<0 mod n. + * px is a shorthand for sum(a[i]*a[x-i], i=0..x). + * Note that [x 0 0 0 0 0] = [x*R]. + */ + + __int128 c, d; + + uint64_t a0 = a[0], a1 = a[1], a2 = a[2], a3 = a[3], a4 = a[4]; + + d = (__int128)(a0*2) * a3 + + (__int128)(a1*2) * a2; + VERIFY_BITS(d, 114); + /* [d 0 0 0] = [p3 0 0 0] */ + c = (__int128)a4 * a4; + VERIFY_BITS(c, 112); + /* [c 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + d += (c & M) * R; c >>= 52; + VERIFY_BITS(d, 115); + VERIFY_BITS(c, 60); + /* [c 0 0 0 0 0 d 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + uint64_t t3 = d & M; d >>= 52; + VERIFY_BITS(t3, 52); + VERIFY_BITS(d, 63); + /* [c 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 0 p3 0 0 0] */ + + a4 *= 2; + d += (__int128)a0 * a4 + + (__int128)(a1*2) * a3 + + (__int128)a2 * a2; + VERIFY_BITS(d, 115); + /* [c 0 0 0 0 d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + d += c * R; + VERIFY_BITS(d, 116); + /* [d t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + uint64_t t4 = d & M; d >>= 52; + VERIFY_BITS(t4, 52); + VERIFY_BITS(d, 64); + /* [d t4 t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + uint64_t tx = (t4 >> 48); t4 &= (M >> 4); + VERIFY_BITS(tx, 4); + VERIFY_BITS(t4, 48); + /* [d t4+(tx<<48) t3 0 0 0] = [p8 0 0 0 p4 p3 0 0 0] */ + + c = (__int128)a0 * a0; + VERIFY_BITS(c, 112); + /* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 0 p4 p3 0 0 p0] */ + d += (__int128)a1 * a4 + + (__int128)(a2*2) * a3; + VERIFY_BITS(d, 114); + /* [d t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + uint64_t u0 = d & M; d >>= 52; + VERIFY_BITS(u0, 52); + VERIFY_BITS(d, 62); + /* [d u0 t4+(tx<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + /* [d 0 t4+(tx<<48)+(u0<<52) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + u0 = (u0 << 4) | tx; + VERIFY_BITS(u0, 56); + /* [d 0 t4+(u0<<48) t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + c += (__int128)u0 * (R >> 4); + VERIFY_BITS(c, 113); + /* [d 0 t4 t3 0 0 c] = [p8 0 0 p5 p4 p3 0 0 p0] */ + r[0] = c & M; c >>= 52; + VERIFY_BITS(r[0], 52); + VERIFY_BITS(c, 61); + /* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 0 p0] */ + + a0 *= 2; + c += (__int128)a0 * a1; + VERIFY_BITS(c, 114); + /* [d 0 t4 t3 0 c r0] = [p8 0 0 p5 p4 p3 0 p1 p0] */ + d += (__int128)a2 * a4 + + (__int128)a3 * a3; + VERIFY_BITS(d, 114); + /* [d 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + c += (d & M) * R; d >>= 52; + VERIFY_BITS(c, 115); + VERIFY_BITS(d, 62); + /* [d 0 0 t4 t3 0 c r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + r[1] = c & M; c >>= 52; + VERIFY_BITS(r[1], 52); + VERIFY_BITS(c, 63); + /* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 0 p1 p0] */ + + c += (__int128)a0 * a2 + + (__int128)a1 * a1; + VERIFY_BITS(c, 114); + /* [d 0 0 t4 t3 c r1 r0] = [p8 0 p6 p5 p4 p3 p2 p1 p0] */ + d += (__int128)a3 * a4; + VERIFY_BITS(d, 114); + /* [d 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += (d & M) * R; d >>= 52; + VERIFY_BITS(c, 115); + VERIFY_BITS(d, 62); + /* [d 0 0 0 t4 t3 c r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[2] = c & M; c >>= 52; + VERIFY_BITS(r[2], 52); + VERIFY_BITS(c, 63); + /* [d 0 0 0 t4 t3+c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + + c += d * R + t3;; + VERIFY_BITS(c, 100); + /* [t4 c r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[3] = c & M; c >>= 52; + VERIFY_BITS(r[3], 52); + VERIFY_BITS(c, 48); + /* [t4+c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + c += t4; + VERIFY_BITS(c, 49); + /* [c r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ + r[4] = c; + VERIFY_BITS(r[4], 49); + /* [r4 r3 r2 r1 r0] = [p8 p7 p6 p5 p4 p3 p2 p1 p0] */ +} + +#endif diff --git a/src/field_gmp.h b/src/field_gmp.h new file mode 100644 index 0000000000..b390fd9de8 --- /dev/null +++ b/src/field_gmp.h @@ -0,0 +1,18 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_ +#define _SECP256K1_FIELD_REPR_ + +#include + +#define FIELD_LIMBS ((256 + GMP_NUMB_BITS - 1) / GMP_NUMB_BITS) + +typedef struct { + mp_limb_t n[FIELD_LIMBS+1]; +} secp256k1_fe_t; + +#endif diff --git a/src/field_gmp_impl.h b/src/field_gmp_impl.h new file mode 100644 index 0000000000..af4728e5b4 --- /dev/null +++ b/src/field_gmp_impl.h @@ -0,0 +1,163 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_REPR_IMPL_H_ +#define _SECP256K1_FIELD_REPR_IMPL_H_ + +#include +#include +#include "num.h" +#include "field.h" + +static mp_limb_t secp256k1_field_p[FIELD_LIMBS]; +static mp_limb_t secp256k1_field_pc[(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS]; + +static void secp256k1_fe_inner_start(void) { + for (int i=0; i<(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS; i++) + secp256k1_field_pc[i] = 0; + secp256k1_field_pc[0] += 0x3D1UL; + secp256k1_field_pc[32/GMP_NUMB_BITS] += (((mp_limb_t)1) << (32 % GMP_NUMB_BITS)); + for (int i=0; in[FIELD_LIMBS] != 0) { +#if (GMP_NUMB_BITS >= 40) + mp_limb_t carry = mpn_add_1(r->n, r->n, FIELD_LIMBS, 0x1000003D1ULL * r->n[FIELD_LIMBS]); + mpn_add_1(r->n, r->n, FIELD_LIMBS, 0x1000003D1ULL * carry); +#else + mp_limb_t carry = mpn_add_1(r->n, r->n, FIELD_LIMBS, 0x3D1UL * r->n[FIELD_LIMBS]) + + mpn_add_1(r->n+(32/GMP_NUMB_BITS), r->n+(32/GMP_NUMB_BITS), FIELD_LIMBS-(32/GMP_NUMB_BITS), r->n[FIELD_LIMBS] << (32 % GMP_NUMB_BITS)); + mpn_add_1(r->n, r->n, FIELD_LIMBS, 0x3D1UL * carry); + mpn_add_1(r->n+(32/GMP_NUMB_BITS), r->n+(32/GMP_NUMB_BITS), FIELD_LIMBS-(32/GMP_NUMB_BITS), carry << (32%GMP_NUMB_BITS)); +#endif + r->n[FIELD_LIMBS] = 0; + } + if (mpn_cmp(r->n, secp256k1_field_p, FIELD_LIMBS) >= 0) + mpn_sub(r->n, r->n, FIELD_LIMBS, secp256k1_field_p, FIELD_LIMBS); +} + +SECP256K1_INLINE static void secp256k1_fe_set_int(secp256k1_fe_t *r, int a) { + r->n[0] = a; + for (int i=1; in[i] = 0; +} + +SECP256K1_INLINE static void secp256k1_fe_clear(secp256k1_fe_t *r) { + for (int i=0; in[i] = 0; +} + +SECP256K1_INLINE static int secp256k1_fe_is_zero(const secp256k1_fe_t *a) { + int ret = 1; + for (int i=0; in[i] == 0); + return ret; +} + +SECP256K1_INLINE static int secp256k1_fe_is_odd(const secp256k1_fe_t *a) { + return a->n[0] & 1; +} + +SECP256K1_INLINE static int secp256k1_fe_equal(const secp256k1_fe_t *a, const secp256k1_fe_t *b) { + int ret = 1; + for (int i=0; in[i] == b->n[i]); + return ret; +} + +static void secp256k1_fe_set_b32(secp256k1_fe_t *r, const unsigned char *a) { + for (int i=0; in[i] = 0; + for (int i=0; i<256; i++) { + int limb = i/GMP_NUMB_BITS; + int shift = i%GMP_NUMB_BITS; + r->n[limb] |= (mp_limb_t)((a[31-i/8] >> (i%8)) & 0x1) << shift; + } +} + +/** Convert a field element to a 32-byte big endian value. Requires the input to be normalized */ +static void secp256k1_fe_get_b32(unsigned char *r, const secp256k1_fe_t *a) { + for (int i=0; i<32; i++) { + int c = 0; + for (int j=0; j<8; j++) { + int limb = (8*i+j)/GMP_NUMB_BITS; + int shift = (8*i+j)%GMP_NUMB_BITS; + c |= ((a->n[limb] >> shift) & 0x1) << j; + } + r[31-i] = c; + } +} + +SECP256K1_INLINE static void secp256k1_fe_negate(secp256k1_fe_t *r, const secp256k1_fe_t *a, int m) { + (void)m; + *r = *a; + secp256k1_fe_normalize(r); + for (int i=0; in[i] = ~(r->n[i]); +#if (GMP_NUMB_BITS >= 33) + mpn_sub_1(r->n, r->n, FIELD_LIMBS, 0x1000003D0ULL); +#else + mpn_sub_1(r->n, r->n, FIELD_LIMBS, 0x3D0UL); + mpn_sub_1(r->n+(32/GMP_NUMB_BITS), r->n+(32/GMP_NUMB_BITS), FIELD_LIMBS-(32/GMP_NUMB_BITS), 0x1UL << (32%GMP_NUMB_BITS)); +#endif +} + +SECP256K1_INLINE static void secp256k1_fe_mul_int(secp256k1_fe_t *r, int a) { + mpn_mul_1(r->n, r->n, FIELD_LIMBS+1, a); +} + +SECP256K1_INLINE static void secp256k1_fe_add(secp256k1_fe_t *r, const secp256k1_fe_t *a) { + mpn_add(r->n, r->n, FIELD_LIMBS+1, a->n, FIELD_LIMBS+1); +} + +static void secp256k1_fe_reduce(secp256k1_fe_t *r, mp_limb_t *tmp) { + /** + * B1 B2 B3 B4 + * + C * A1 A2 A3 A4 + * + A1 A2 A3 A4 + */ + +#if (GMP_NUMB_BITS >= 33) + mp_limb_t o = mpn_addmul_1(tmp, tmp+FIELD_LIMBS, FIELD_LIMBS, 0x1000003D1ULL); +#else + mp_limb_t o = mpn_addmul_1(tmp, tmp+FIELD_LIMBS, FIELD_LIMBS, 0x3D1UL) + + mpn_addmul_1(tmp+(32/GMP_NUMB_BITS), tmp+FIELD_LIMBS, FIELD_LIMBS-(32/GMP_NUMB_BITS), 0x1UL << (32%GMP_NUMB_BITS)); +#endif + mp_limb_t q[1+(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS]; + q[(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS] = mpn_mul_1(q, secp256k1_field_pc, (33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS, o); +#if (GMP_NUMB_BITS <= 32) + mp_limb_t o2 = tmp[2*FIELD_LIMBS-(32/GMP_NUMB_BITS)] << (32%GMP_NUMB_BITS); + q[(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS] += mpn_addmul_1(q, secp256k1_field_pc, (33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS, o2); +#endif + r->n[FIELD_LIMBS] = mpn_add(r->n, tmp, FIELD_LIMBS, q, 1+(33+GMP_NUMB_BITS-1)/GMP_NUMB_BITS); +} + +static void secp256k1_fe_mul(secp256k1_fe_t *r, const secp256k1_fe_t *a, const secp256k1_fe_t *b) { + secp256k1_fe_t ac = *a; + secp256k1_fe_t bc = *b; + secp256k1_fe_normalize(&ac); + secp256k1_fe_normalize(&bc); + mp_limb_t tmp[2*FIELD_LIMBS]; + mpn_mul_n(tmp, ac.n, bc.n, FIELD_LIMBS); + secp256k1_fe_reduce(r, tmp); +} + +static void secp256k1_fe_sqr(secp256k1_fe_t *r, const secp256k1_fe_t *a) { + secp256k1_fe_t ac = *a; + secp256k1_fe_normalize(&ac); + mp_limb_t tmp[2*FIELD_LIMBS]; + mpn_sqr(tmp, ac.n, FIELD_LIMBS); + secp256k1_fe_reduce(r, tmp); +} + +#endif diff --git a/src/field_impl.h b/src/field_impl.h new file mode 100644 index 0000000000..3a31e1844e --- /dev/null +++ b/src/field_impl.h @@ -0,0 +1,293 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_FIELD_IMPL_H_ +#define _SECP256K1_FIELD_IMPL_H_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#include "util.h" + +#if defined(USE_FIELD_GMP) +#include "field_gmp_impl.h" +#elif defined(USE_FIELD_10X26) +#include "field_10x26_impl.h" +#elif defined(USE_FIELD_5X52) +#include "field_5x52_impl.h" +#else +#error "Please select field implementation" +#endif + +static void secp256k1_fe_get_hex(char *r, int *rlen, const secp256k1_fe_t *a) { + if (*rlen < 65) { + *rlen = 65; + return; + } + *rlen = 65; + unsigned char tmp[32]; + secp256k1_fe_t b = *a; + secp256k1_fe_normalize(&b); + secp256k1_fe_get_b32(tmp, &b); + for (int i=0; i<32; i++) { + static const char *c = "0123456789ABCDEF"; + r[2*i] = c[(tmp[i] >> 4) & 0xF]; + r[2*i+1] = c[(tmp[i]) & 0xF]; + } + r[64] = 0x00; +} + +static void secp256k1_fe_set_hex(secp256k1_fe_t *r, const char *a, int alen) { + unsigned char tmp[32] = {}; + static const int cvt[256] = {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, 0,0,0,0,0,0,0,0,0,0, + 0, 1, 2, 3, 4, 5, 6,7,8,9,0,0,0,0,0,0, + 0,10,11,12,13,14,15,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 0,10,11,12,13,14,15,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 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, 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, 0,0,0,0,0,0,0,0,0,0, + 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}; + for (int i=0; i<32; i++) { + if (alen > i*2) + tmp[32 - alen/2 + i] = (cvt[(unsigned char)a[2*i]] << 4) + cvt[(unsigned char)a[2*i+1]]; + } + secp256k1_fe_set_b32(r, tmp); +} + +static int secp256k1_fe_sqrt(secp256k1_fe_t *r, const secp256k1_fe_t *a) { + + /** The binary representation of (p + 1)/4 has 3 blocks of 1s, with lengths in + * { 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block: + * 1, [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223] + */ + + secp256k1_fe_t x2; + secp256k1_fe_sqr(&x2, a); + secp256k1_fe_mul(&x2, &x2, a); + + secp256k1_fe_t x3; + secp256k1_fe_sqr(&x3, &x2); + secp256k1_fe_mul(&x3, &x3, a); + + secp256k1_fe_t x6 = x3; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x6, &x6); + secp256k1_fe_mul(&x6, &x6, &x3); + + secp256k1_fe_t x9 = x6; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x9, &x9); + secp256k1_fe_mul(&x9, &x9, &x3); + + secp256k1_fe_t x11 = x9; + for (int j=0; j<2; j++) secp256k1_fe_sqr(&x11, &x11); + secp256k1_fe_mul(&x11, &x11, &x2); + + secp256k1_fe_t x22 = x11; + for (int j=0; j<11; j++) secp256k1_fe_sqr(&x22, &x22); + secp256k1_fe_mul(&x22, &x22, &x11); + + secp256k1_fe_t x44 = x22; + for (int j=0; j<22; j++) secp256k1_fe_sqr(&x44, &x44); + secp256k1_fe_mul(&x44, &x44, &x22); + + secp256k1_fe_t x88 = x44; + for (int j=0; j<44; j++) secp256k1_fe_sqr(&x88, &x88); + secp256k1_fe_mul(&x88, &x88, &x44); + + secp256k1_fe_t x176 = x88; + for (int j=0; j<88; j++) secp256k1_fe_sqr(&x176, &x176); + secp256k1_fe_mul(&x176, &x176, &x88); + + secp256k1_fe_t x220 = x176; + for (int j=0; j<44; j++) secp256k1_fe_sqr(&x220, &x220); + secp256k1_fe_mul(&x220, &x220, &x44); + + secp256k1_fe_t x223 = x220; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x223, &x223); + secp256k1_fe_mul(&x223, &x223, &x3); + + /* The final result is then assembled using a sliding window over the blocks. */ + + secp256k1_fe_t t1 = x223; + for (int j=0; j<23; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(&t1, &t1, &x22); + for (int j=0; j<6; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(&t1, &t1, &x2); + secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_sqr(r, &t1); + + /* Check that a square root was actually calculated */ + + secp256k1_fe_sqr(&t1, r); + secp256k1_fe_negate(&t1, &t1, 1); + secp256k1_fe_add(&t1, a); + secp256k1_fe_normalize(&t1); + return secp256k1_fe_is_zero(&t1); +} + +static void secp256k1_fe_inv(secp256k1_fe_t *r, const secp256k1_fe_t *a) { + + /** The binary representation of (p - 2) has 5 blocks of 1s, with lengths in + * { 1, 2, 22, 223 }. Use an addition chain to calculate 2^n - 1 for each block: + * [1], [2], 3, 6, 9, 11, [22], 44, 88, 176, 220, [223] + */ + + secp256k1_fe_t x2; + secp256k1_fe_sqr(&x2, a); + secp256k1_fe_mul(&x2, &x2, a); + + secp256k1_fe_t x3; + secp256k1_fe_sqr(&x3, &x2); + secp256k1_fe_mul(&x3, &x3, a); + + secp256k1_fe_t x6 = x3; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x6, &x6); + secp256k1_fe_mul(&x6, &x6, &x3); + + secp256k1_fe_t x9 = x6; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x9, &x9); + secp256k1_fe_mul(&x9, &x9, &x3); + + secp256k1_fe_t x11 = x9; + for (int j=0; j<2; j++) secp256k1_fe_sqr(&x11, &x11); + secp256k1_fe_mul(&x11, &x11, &x2); + + secp256k1_fe_t x22 = x11; + for (int j=0; j<11; j++) secp256k1_fe_sqr(&x22, &x22); + secp256k1_fe_mul(&x22, &x22, &x11); + + secp256k1_fe_t x44 = x22; + for (int j=0; j<22; j++) secp256k1_fe_sqr(&x44, &x44); + secp256k1_fe_mul(&x44, &x44, &x22); + + secp256k1_fe_t x88 = x44; + for (int j=0; j<44; j++) secp256k1_fe_sqr(&x88, &x88); + secp256k1_fe_mul(&x88, &x88, &x44); + + secp256k1_fe_t x176 = x88; + for (int j=0; j<88; j++) secp256k1_fe_sqr(&x176, &x176); + secp256k1_fe_mul(&x176, &x176, &x88); + + secp256k1_fe_t x220 = x176; + for (int j=0; j<44; j++) secp256k1_fe_sqr(&x220, &x220); + secp256k1_fe_mul(&x220, &x220, &x44); + + secp256k1_fe_t x223 = x220; + for (int j=0; j<3; j++) secp256k1_fe_sqr(&x223, &x223); + secp256k1_fe_mul(&x223, &x223, &x3); + + /* The final result is then assembled using a sliding window over the blocks. */ + + secp256k1_fe_t t1 = x223; + for (int j=0; j<23; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(&t1, &t1, &x22); + for (int j=0; j<5; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(&t1, &t1, a); + for (int j=0; j<3; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(&t1, &t1, &x2); + for (int j=0; j<2; j++) secp256k1_fe_sqr(&t1, &t1); + secp256k1_fe_mul(r, &t1, a); +} + +static void secp256k1_fe_inv_var(secp256k1_fe_t *r, const secp256k1_fe_t *a) { +#if defined(USE_FIELD_INV_BUILTIN) + secp256k1_fe_inv(r, a); +#elif defined(USE_FIELD_INV_NUM) + unsigned char b[32]; + secp256k1_fe_t c = *a; + secp256k1_fe_normalize(&c); + secp256k1_fe_get_b32(b, &c); + secp256k1_num_t n; + secp256k1_num_set_bin(&n, b, 32); + secp256k1_num_mod_inverse(&n, &n, &secp256k1_fe_consts->p); + secp256k1_num_get_bin(b, 32, &n); + secp256k1_fe_set_b32(r, b); +#else +#error "Please select field inverse implementation" +#endif +} + +static void secp256k1_fe_inv_all(size_t len, secp256k1_fe_t r[len], const secp256k1_fe_t a[len]) { + if (len < 1) + return; + + VERIFY_CHECK((r + len <= a) || (a + len <= r)); + + r[0] = a[0]; + + size_t i = 0; + while (++i < len) { + secp256k1_fe_mul(&r[i], &r[i - 1], &a[i]); + } + + secp256k1_fe_t u; secp256k1_fe_inv(&u, &r[--i]); + + while (i > 0) { + int j = i--; + secp256k1_fe_mul(&r[j], &r[i], &u); + secp256k1_fe_mul(&u, &u, &a[j]); + } + + r[0] = u; +} + +static void secp256k1_fe_inv_all_var(size_t len, secp256k1_fe_t r[len], const secp256k1_fe_t a[len]) { + if (len < 1) + return; + + VERIFY_CHECK((r + len <= a) || (a + len <= r)); + + r[0] = a[0]; + + size_t i = 0; + while (++i < len) { + secp256k1_fe_mul(&r[i], &r[i - 1], &a[i]); + } + + secp256k1_fe_t u; secp256k1_fe_inv_var(&u, &r[--i]); + + while (i > 0) { + int j = i--; + secp256k1_fe_mul(&r[j], &r[i], &u); + secp256k1_fe_mul(&u, &u, &a[j]); + } + + r[0] = u; +} + +static void secp256k1_fe_start(void) { + static const unsigned char secp256k1_fe_consts_p[] = { + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFE,0xFF,0xFF,0xFC,0x2F + }; + if (secp256k1_fe_consts == NULL) { + secp256k1_fe_inner_start(); + secp256k1_fe_consts_t *ret = (secp256k1_fe_consts_t*)malloc(sizeof(secp256k1_fe_consts_t)); + secp256k1_num_set_bin(&ret->p, secp256k1_fe_consts_p, sizeof(secp256k1_fe_consts_p)); + secp256k1_fe_consts = ret; + } +} + +static void secp256k1_fe_stop(void) { + if (secp256k1_fe_consts != NULL) { + secp256k1_fe_consts_t *c = (secp256k1_fe_consts_t*)secp256k1_fe_consts; + free((void*)c); + secp256k1_fe_consts = NULL; + secp256k1_fe_inner_stop(); + } +} + +#endif diff --git a/src/group.h b/src/group.h new file mode 100644 index 0000000000..ba02549821 --- /dev/null +++ b/src/group.h @@ -0,0 +1,128 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_GROUP_ +#define _SECP256K1_GROUP_ + +#include "num.h" +#include "field.h" + +/** A group element of the secp256k1 curve, in affine coordinates. */ +typedef struct { + secp256k1_fe_t x; + secp256k1_fe_t y; + int infinity; /* whether this represents the point at infinity */ +} secp256k1_ge_t; + +/** A group element of the secp256k1 curve, in jacobian coordinates. */ +typedef struct { + secp256k1_fe_t x; /* actual X: x/z^2 */ + secp256k1_fe_t y; /* actual Y: y/z^3 */ + secp256k1_fe_t z; + int infinity; /* whether this represents the point at infinity */ +} secp256k1_gej_t; + +/** Global constants related to the group */ +typedef struct { + secp256k1_num_t order; /* the order of the curve (= order of its generator) */ + secp256k1_num_t half_order; /* half the order of the curve (= order of its generator) */ + secp256k1_ge_t g; /* the generator point */ + +#ifdef USE_ENDOMORPHISM + /* constants related to secp256k1's efficiently computable endomorphism */ + secp256k1_fe_t beta; + secp256k1_num_t lambda, a1b2, b1, a2; +#endif +} secp256k1_ge_consts_t; + +static const secp256k1_ge_consts_t *secp256k1_ge_consts = NULL; + +/** Initialize the group module. */ +static void secp256k1_ge_start(void); + +/** De-initialize the group module. */ +static void secp256k1_ge_stop(void); + +/** Set a group element equal to the point at infinity */ +static void secp256k1_ge_set_infinity(secp256k1_ge_t *r); + +/** Set a group element equal to the point with given X and Y coordinates */ +static void secp256k1_ge_set_xy(secp256k1_ge_t *r, const secp256k1_fe_t *x, const secp256k1_fe_t *y); + +/** Set a group element (affine) equal to the point with the given X coordinate, and given oddness + * for Y. Return value indicates whether the result is valid. */ +static int secp256k1_ge_set_xo(secp256k1_ge_t *r, const secp256k1_fe_t *x, int odd); + +/** Check whether a group element is the point at infinity. */ +static int secp256k1_ge_is_infinity(const secp256k1_ge_t *a); + +/** Check whether a group element is valid (i.e., on the curve). */ +static int secp256k1_ge_is_valid(const secp256k1_ge_t *a); + +static void secp256k1_ge_neg(secp256k1_ge_t *r, const secp256k1_ge_t *a); + +/** Get a hex representation of a point. *rlen will be overwritten with the real length. */ +static void secp256k1_ge_get_hex(char *r, int *rlen, const secp256k1_ge_t *a); + +/** Set a group element equal to another which is given in jacobian coordinates */ +static void secp256k1_ge_set_gej(secp256k1_ge_t *r, secp256k1_gej_t *a); + +/** Set a batch of group elements equal to the inputs given in jacobian coordinates */ +static void secp256k1_ge_set_all_gej_var(size_t len, secp256k1_ge_t r[len], const secp256k1_gej_t a[len]); + + +/** Set a group element (jacobian) equal to the point at infinity. */ +static void secp256k1_gej_set_infinity(secp256k1_gej_t *r); + +/** Set a group element (jacobian) equal to the point with given X and Y coordinates. */ +static void secp256k1_gej_set_xy(secp256k1_gej_t *r, const secp256k1_fe_t *x, const secp256k1_fe_t *y); + +/** Set a group element (jacobian) equal to another which is given in affine coordinates. */ +static void secp256k1_gej_set_ge(secp256k1_gej_t *r, const secp256k1_ge_t *a); + +/** Get the X coordinate of a group element (jacobian). */ +static void secp256k1_gej_get_x_var(secp256k1_fe_t *r, const secp256k1_gej_t *a); + +/** Set r equal to the inverse of a (i.e., mirrored around the X axis) */ +static void secp256k1_gej_neg(secp256k1_gej_t *r, const secp256k1_gej_t *a); + +/** Check whether a group element is the point at infinity. */ +static int secp256k1_gej_is_infinity(const secp256k1_gej_t *a); + +/** Set r equal to the double of a. */ +static void secp256k1_gej_double_var(secp256k1_gej_t *r, const secp256k1_gej_t *a); + +/** Set r equal to the sum of a and b. */ +static void secp256k1_gej_add_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_gej_t *b); + +/** Set r equal to the sum of a and b (with b given in affine coordinates, and not infinity). */ +static void secp256k1_gej_add_ge(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b); + +/** Set r equal to the sum of a and b (with b given in affine coordinates). This is more efficient + than secp256k1_gej_add_var. It is identical to secp256k1_gej_add_ge but without constant-time + guarantee, and b is allowed to be infinity. */ +static void secp256k1_gej_add_ge_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b); + +/** Get a hex representation of a point. *rlen will be overwritten with the real length. */ +static void secp256k1_gej_get_hex(char *r, int *rlen, const secp256k1_gej_t *a); + +#ifdef USE_ENDOMORPHISM +/** Set r to be equal to lambda times a, where lambda is chosen in a way such that this is very fast. */ +static void secp256k1_gej_mul_lambda(secp256k1_gej_t *r, const secp256k1_gej_t *a); + +/** Find r1 and r2 such that r1+r2*lambda = a, and r1 and r2 are maximum 128 bits long (given that a is + not more than 256 bits). */ +static void secp256k1_gej_split_exp_var(secp256k1_num_t *r1, secp256k1_num_t *r2, const secp256k1_num_t *a); +#endif + +/** Clear a secp256k1_gej_t to prevent leaking sensitive information. */ +static void secp256k1_gej_clear(secp256k1_gej_t *r); + +/** Clear a secp256k1_ge_t to prevent leaking sensitive information. */ +static void secp256k1_ge_clear(secp256k1_ge_t *r); + + +#endif diff --git a/src/group_impl.h b/src/group_impl.h new file mode 100644 index 0000000000..1edbc6e099 --- /dev/null +++ b/src/group_impl.h @@ -0,0 +1,519 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_GROUP_IMPL_H_ +#define _SECP256K1_GROUP_IMPL_H_ + +#include + +#include "num.h" +#include "field.h" +#include "group.h" + +static void secp256k1_ge_set_infinity(secp256k1_ge_t *r) { + r->infinity = 1; +} + +static void secp256k1_ge_set_xy(secp256k1_ge_t *r, const secp256k1_fe_t *x, const secp256k1_fe_t *y) { + r->infinity = 0; + r->x = *x; + r->y = *y; +} + +static int secp256k1_ge_is_infinity(const secp256k1_ge_t *a) { + return a->infinity; +} + +static void secp256k1_ge_neg(secp256k1_ge_t *r, const secp256k1_ge_t *a) { + r->infinity = a->infinity; + r->x = a->x; + r->y = a->y; + secp256k1_fe_normalize(&r->y); + secp256k1_fe_negate(&r->y, &r->y, 1); +} + +static void secp256k1_ge_get_hex(char *r, int *rlen, const secp256k1_ge_t *a) { + char cx[65]; int lx=65; + char cy[65]; int ly=65; + secp256k1_fe_get_hex(cx, &lx, &a->x); + secp256k1_fe_get_hex(cy, &ly, &a->y); + lx = strlen(cx); + ly = strlen(cy); + int len = lx + ly + 3 + 1; + if (*rlen < len) { + *rlen = len; + return; + } + *rlen = len; + r[0] = '('; + memcpy(r+1, cx, lx); + r[1+lx] = ','; + memcpy(r+2+lx, cy, ly); + r[2+lx+ly] = ')'; + r[3+lx+ly] = 0; +} + +static void secp256k1_ge_set_gej(secp256k1_ge_t *r, secp256k1_gej_t *a) { + r->infinity = a->infinity; + secp256k1_fe_inv(&a->z, &a->z); + secp256k1_fe_t z2; secp256k1_fe_sqr(&z2, &a->z); + secp256k1_fe_t z3; secp256k1_fe_mul(&z3, &a->z, &z2); + secp256k1_fe_mul(&a->x, &a->x, &z2); + secp256k1_fe_mul(&a->y, &a->y, &z3); + secp256k1_fe_set_int(&a->z, 1); + r->x = a->x; + r->y = a->y; +} + +static void secp256k1_ge_set_gej_var(secp256k1_ge_t *r, secp256k1_gej_t *a) { + r->infinity = a->infinity; + if (a->infinity) { + return; + } + secp256k1_fe_inv_var(&a->z, &a->z); + secp256k1_fe_t z2; secp256k1_fe_sqr(&z2, &a->z); + secp256k1_fe_t z3; secp256k1_fe_mul(&z3, &a->z, &z2); + secp256k1_fe_mul(&a->x, &a->x, &z2); + secp256k1_fe_mul(&a->y, &a->y, &z3); + secp256k1_fe_set_int(&a->z, 1); + r->x = a->x; + r->y = a->y; +} + +static void secp256k1_ge_set_all_gej_var(size_t len, secp256k1_ge_t r[len], const secp256k1_gej_t a[len]) { + size_t count = 0; + secp256k1_fe_t az[len]; + for (size_t i=0; iinfinity = 1; + secp256k1_fe_set_int(&r->x, 0); + secp256k1_fe_set_int(&r->y, 0); + secp256k1_fe_set_int(&r->z, 0); +} + +static void secp256k1_gej_set_xy(secp256k1_gej_t *r, const secp256k1_fe_t *x, const secp256k1_fe_t *y) { + r->infinity = 0; + r->x = *x; + r->y = *y; + secp256k1_fe_set_int(&r->z, 1); +} + +static void secp256k1_gej_clear(secp256k1_gej_t *r) { + r->infinity = 0; + secp256k1_fe_clear(&r->x); + secp256k1_fe_clear(&r->y); + secp256k1_fe_clear(&r->z); +} + +static void secp256k1_ge_clear(secp256k1_ge_t *r) { + r->infinity = 0; + secp256k1_fe_clear(&r->x); + secp256k1_fe_clear(&r->y); +} + +static int secp256k1_ge_set_xo(secp256k1_ge_t *r, const secp256k1_fe_t *x, int odd) { + r->x = *x; + secp256k1_fe_t x2; secp256k1_fe_sqr(&x2, x); + secp256k1_fe_t x3; secp256k1_fe_mul(&x3, x, &x2); + r->infinity = 0; + secp256k1_fe_t c; secp256k1_fe_set_int(&c, 7); + secp256k1_fe_add(&c, &x3); + if (!secp256k1_fe_sqrt(&r->y, &c)) + return 0; + secp256k1_fe_normalize(&r->y); + if (secp256k1_fe_is_odd(&r->y) != odd) + secp256k1_fe_negate(&r->y, &r->y, 1); + return 1; +} + +static void secp256k1_gej_set_ge(secp256k1_gej_t *r, const secp256k1_ge_t *a) { + r->infinity = a->infinity; + r->x = a->x; + r->y = a->y; + secp256k1_fe_set_int(&r->z, 1); +} + +static void secp256k1_gej_get_x_var(secp256k1_fe_t *r, const secp256k1_gej_t *a) { + secp256k1_fe_t zi2; secp256k1_fe_inv_var(&zi2, &a->z); secp256k1_fe_sqr(&zi2, &zi2); + secp256k1_fe_mul(r, &a->x, &zi2); +} + +static void secp256k1_gej_neg(secp256k1_gej_t *r, const secp256k1_gej_t *a) { + r->infinity = a->infinity; + r->x = a->x; + r->y = a->y; + r->z = a->z; + secp256k1_fe_normalize(&r->y); + secp256k1_fe_negate(&r->y, &r->y, 1); +} + +static int secp256k1_gej_is_infinity(const secp256k1_gej_t *a) { + return a->infinity; +} + +static int secp256k1_gej_is_valid(const secp256k1_gej_t *a) { + if (a->infinity) + return 0; + /** y^2 = x^3 + 7 + * (Y/Z^3)^2 = (X/Z^2)^3 + 7 + * Y^2 / Z^6 = X^3 / Z^6 + 7 + * Y^2 = X^3 + 7*Z^6 + */ + secp256k1_fe_t y2; secp256k1_fe_sqr(&y2, &a->y); + secp256k1_fe_t x3; secp256k1_fe_sqr(&x3, &a->x); secp256k1_fe_mul(&x3, &x3, &a->x); + secp256k1_fe_t z2; secp256k1_fe_sqr(&z2, &a->z); + secp256k1_fe_t z6; secp256k1_fe_sqr(&z6, &z2); secp256k1_fe_mul(&z6, &z6, &z2); + secp256k1_fe_mul_int(&z6, 7); + secp256k1_fe_add(&x3, &z6); + secp256k1_fe_normalize(&y2); + secp256k1_fe_normalize(&x3); + return secp256k1_fe_equal(&y2, &x3); +} + +static int secp256k1_ge_is_valid(const secp256k1_ge_t *a) { + if (a->infinity) + return 0; + /* y^2 = x^3 + 7 */ + secp256k1_fe_t y2; secp256k1_fe_sqr(&y2, &a->y); + secp256k1_fe_t x3; secp256k1_fe_sqr(&x3, &a->x); secp256k1_fe_mul(&x3, &x3, &a->x); + secp256k1_fe_t c; secp256k1_fe_set_int(&c, 7); + secp256k1_fe_add(&x3, &c); + secp256k1_fe_normalize(&y2); + secp256k1_fe_normalize(&x3); + return secp256k1_fe_equal(&y2, &x3); +} + +static void secp256k1_gej_double_var(secp256k1_gej_t *r, const secp256k1_gej_t *a) { + if (a->infinity) { + r->infinity = 1; + return; + } + + secp256k1_fe_t t5 = a->y; + secp256k1_fe_normalize(&t5); + if (secp256k1_fe_is_zero(&t5)) { + r->infinity = 1; + return; + } + + secp256k1_fe_t t1,t2,t3,t4; + secp256k1_fe_mul(&r->z, &t5, &a->z); + secp256k1_fe_mul_int(&r->z, 2); /* Z' = 2*Y*Z (2) */ + secp256k1_fe_sqr(&t1, &a->x); + secp256k1_fe_mul_int(&t1, 3); /* T1 = 3*X^2 (3) */ + secp256k1_fe_sqr(&t2, &t1); /* T2 = 9*X^4 (1) */ + secp256k1_fe_sqr(&t3, &t5); + secp256k1_fe_mul_int(&t3, 2); /* T3 = 2*Y^2 (2) */ + secp256k1_fe_sqr(&t4, &t3); + secp256k1_fe_mul_int(&t4, 2); /* T4 = 8*Y^4 (2) */ + secp256k1_fe_mul(&t3, &a->x, &t3); /* T3 = 2*X*Y^2 (1) */ + r->x = t3; + secp256k1_fe_mul_int(&r->x, 4); /* X' = 8*X*Y^2 (4) */ + secp256k1_fe_negate(&r->x, &r->x, 4); /* X' = -8*X*Y^2 (5) */ + secp256k1_fe_add(&r->x, &t2); /* X' = 9*X^4 - 8*X*Y^2 (6) */ + secp256k1_fe_negate(&t2, &t2, 1); /* T2 = -9*X^4 (2) */ + secp256k1_fe_mul_int(&t3, 6); /* T3 = 12*X*Y^2 (6) */ + secp256k1_fe_add(&t3, &t2); /* T3 = 12*X*Y^2 - 9*X^4 (8) */ + secp256k1_fe_mul(&r->y, &t1, &t3); /* Y' = 36*X^3*Y^2 - 27*X^6 (1) */ + secp256k1_fe_negate(&t2, &t4, 2); /* T2 = -8*Y^4 (3) */ + secp256k1_fe_add(&r->y, &t2); /* Y' = 36*X^3*Y^2 - 27*X^6 - 8*Y^4 (4) */ + r->infinity = 0; +} + +static void secp256k1_gej_add_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_gej_t *b) { + if (a->infinity) { + *r = *b; + return; + } + if (b->infinity) { + *r = *a; + return; + } + r->infinity = 0; + secp256k1_fe_t z22; secp256k1_fe_sqr(&z22, &b->z); + secp256k1_fe_t z12; secp256k1_fe_sqr(&z12, &a->z); + secp256k1_fe_t u1; secp256k1_fe_mul(&u1, &a->x, &z22); + secp256k1_fe_t u2; secp256k1_fe_mul(&u2, &b->x, &z12); + secp256k1_fe_t s1; secp256k1_fe_mul(&s1, &a->y, &z22); secp256k1_fe_mul(&s1, &s1, &b->z); + secp256k1_fe_t s2; secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z); + secp256k1_fe_normalize(&u1); + secp256k1_fe_normalize(&u2); + if (secp256k1_fe_equal(&u1, &u2)) { + secp256k1_fe_normalize(&s1); + secp256k1_fe_normalize(&s2); + if (secp256k1_fe_equal(&s1, &s2)) { + secp256k1_gej_double_var(r, a); + } else { + r->infinity = 1; + } + return; + } + secp256k1_fe_t h; secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2); + secp256k1_fe_t i; secp256k1_fe_negate(&i, &s1, 1); secp256k1_fe_add(&i, &s2); + secp256k1_fe_t i2; secp256k1_fe_sqr(&i2, &i); + secp256k1_fe_t h2; secp256k1_fe_sqr(&h2, &h); + secp256k1_fe_t h3; secp256k1_fe_mul(&h3, &h, &h2); + secp256k1_fe_mul(&r->z, &a->z, &b->z); secp256k1_fe_mul(&r->z, &r->z, &h); + secp256k1_fe_t t; secp256k1_fe_mul(&t, &u1, &h2); + r->x = t; secp256k1_fe_mul_int(&r->x, 2); secp256k1_fe_add(&r->x, &h3); secp256k1_fe_negate(&r->x, &r->x, 3); secp256k1_fe_add(&r->x, &i2); + secp256k1_fe_negate(&r->y, &r->x, 5); secp256k1_fe_add(&r->y, &t); secp256k1_fe_mul(&r->y, &r->y, &i); + secp256k1_fe_mul(&h3, &h3, &s1); secp256k1_fe_negate(&h3, &h3, 1); + secp256k1_fe_add(&r->y, &h3); +} + +static void secp256k1_gej_add_ge_var(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b) { + if (a->infinity) { + r->infinity = b->infinity; + r->x = b->x; + r->y = b->y; + secp256k1_fe_set_int(&r->z, 1); + return; + } + if (b->infinity) { + *r = *a; + return; + } + r->infinity = 0; + secp256k1_fe_t z12; secp256k1_fe_sqr(&z12, &a->z); + secp256k1_fe_t u1 = a->x; secp256k1_fe_normalize(&u1); + secp256k1_fe_t u2; secp256k1_fe_mul(&u2, &b->x, &z12); + secp256k1_fe_t s1 = a->y; secp256k1_fe_normalize(&s1); + secp256k1_fe_t s2; secp256k1_fe_mul(&s2, &b->y, &z12); secp256k1_fe_mul(&s2, &s2, &a->z); + secp256k1_fe_normalize(&u1); + secp256k1_fe_normalize(&u2); + if (secp256k1_fe_equal(&u1, &u2)) { + secp256k1_fe_normalize(&s1); + secp256k1_fe_normalize(&s2); + if (secp256k1_fe_equal(&s1, &s2)) { + secp256k1_gej_double_var(r, a); + } else { + r->infinity = 1; + } + return; + } + secp256k1_fe_t h; secp256k1_fe_negate(&h, &u1, 1); secp256k1_fe_add(&h, &u2); + secp256k1_fe_t i; secp256k1_fe_negate(&i, &s1, 1); secp256k1_fe_add(&i, &s2); + secp256k1_fe_t i2; secp256k1_fe_sqr(&i2, &i); + secp256k1_fe_t h2; secp256k1_fe_sqr(&h2, &h); + secp256k1_fe_t h3; secp256k1_fe_mul(&h3, &h, &h2); + r->z = a->z; secp256k1_fe_mul(&r->z, &r->z, &h); + secp256k1_fe_t t; secp256k1_fe_mul(&t, &u1, &h2); + r->x = t; secp256k1_fe_mul_int(&r->x, 2); secp256k1_fe_add(&r->x, &h3); secp256k1_fe_negate(&r->x, &r->x, 3); secp256k1_fe_add(&r->x, &i2); + secp256k1_fe_negate(&r->y, &r->x, 5); secp256k1_fe_add(&r->y, &t); secp256k1_fe_mul(&r->y, &r->y, &i); + secp256k1_fe_mul(&h3, &h3, &s1); secp256k1_fe_negate(&h3, &h3, 1); + secp256k1_fe_add(&r->y, &h3); +} + +static void secp256k1_gej_add_ge(secp256k1_gej_t *r, const secp256k1_gej_t *a, const secp256k1_ge_t *b) { + VERIFY_CHECK(!b->infinity); + VERIFY_CHECK(a->infinity == 0 || a->infinity == 1); + + /** In: + * Eric Brier and Marc Joye, Weierstrass Elliptic Curves and Side-Channel Attacks. + * In D. Naccache and P. Paillier, Eds., Public Key Cryptography, vol. 2274 of Lecture Notes in Computer Science, pages 335-345. Springer-Verlag, 2002. + * we find as solution for a unified addition/doubling formula: + * lambda = ((x1 + x2)^2 - x1 * x2 + a) / (y1 + y2), with a = 0 for secp256k1's curve equation. + * x3 = lambda^2 - (x1 + x2) + * 2*y3 = lambda * (x1 + x2 - 2 * x3) - (y1 + y2). + * + * Substituting x_i = Xi / Zi^2 and yi = Yi / Zi^3, for i=1,2,3, gives: + * U1 = X1*Z2^2, U2 = X2*Z1^2 + * S1 = X1*Z2^3, S2 = X2*Z2^3 + * Z = Z1*Z2 + * T = U1+U2 + * M = S1+S2 + * Q = T*M^2 + * R = T^2-U1*U2 + * X3 = 4*(R^2-Q) + * Y3 = 4*(R*(3*Q-2*R^2)-M^4) + * Z3 = 2*M*Z + * (Note that the paper uses xi = Xi / Zi and yi = Yi / Zi instead.) + */ + + secp256k1_fe_t zz; secp256k1_fe_sqr(&zz, &a->z); /* z = Z1^2 */ + secp256k1_fe_t u1 = a->x; secp256k1_fe_normalize(&u1); /* u1 = U1 = X1*Z2^2 (1) */ + secp256k1_fe_t u2; secp256k1_fe_mul(&u2, &b->x, &zz); /* u2 = U2 = X2*Z1^2 (1) */ + secp256k1_fe_t s1 = a->y; secp256k1_fe_normalize(&s1); /* s1 = S1 = Y1*Z2^3 (1) */ + secp256k1_fe_t s2; secp256k1_fe_mul(&s2, &b->y, &zz); /* s2 = Y2*Z2^2 (1) */ + secp256k1_fe_mul(&s2, &s2, &a->z); /* s2 = S2 = Y2*Z1^3 (1) */ + secp256k1_fe_t z = a->z; /* z = Z = Z1*Z2 (8) */ + secp256k1_fe_t t = u1; secp256k1_fe_add(&t, &u2); /* t = T = U1+U2 (2) */ + secp256k1_fe_t m = s1; secp256k1_fe_add(&m, &s2); /* m = M = S1+S2 (2) */ + secp256k1_fe_t n; secp256k1_fe_sqr(&n, &m); /* n = M^2 (1) */ + secp256k1_fe_t q; secp256k1_fe_mul(&q, &n, &t); /* q = Q = T*M^2 (1) */ + secp256k1_fe_sqr(&n, &n); /* n = M^4 (1) */ + secp256k1_fe_t rr; secp256k1_fe_sqr(&rr, &t); /* rr = T^2 (1) */ + secp256k1_fe_mul(&t, &u1, &u2); secp256k1_fe_negate(&t, &t, 1); /* t = -U1*U2 (2) */ + secp256k1_fe_add(&rr, &t); /* rr = R = T^2-U1*U2 (3) */ + secp256k1_fe_sqr(&t, &rr); /* t = R^2 (1) */ + secp256k1_fe_mul(&r->z, &m, &z); /* r->z = M*Z (1) */ + secp256k1_fe_normalize(&r->z); + int infinity = secp256k1_fe_is_zero(&r->z) * (1 - a->infinity); + secp256k1_fe_mul_int(&r->z, 2 * (1 - a->infinity)); /* r->z = Z3 = 2*M*Z (2) */ + r->x = t; /* r->x = R^2 (1) */ + secp256k1_fe_negate(&q, &q, 1); /* q = -Q (2) */ + secp256k1_fe_add(&r->x, &q); /* r->x = R^2-Q (3) */ + secp256k1_fe_normalize(&r->x); + secp256k1_fe_mul_int(&q, 3); /* q = -3*Q (6) */ + secp256k1_fe_mul_int(&t, 2); /* t = 2*R^2 (2) */ + secp256k1_fe_add(&t, &q); /* t = 2*R^2-3*Q (8) */ + secp256k1_fe_mul(&t, &t, &rr); /* t = R*(2*R^2-3*Q) (1) */ + secp256k1_fe_add(&t, &n); /* t = R*(2*R^2-3*Q)+M^4 (2) */ + secp256k1_fe_negate(&r->y, &t, 2); /* r->y = R*(3*Q-2*R^2)-M^4 (3) */ + secp256k1_fe_normalize(&r->y); + secp256k1_fe_mul_int(&r->x, 4 * (1 - a->infinity)); /* r->x = X3 = 4*(R^2-Q) */ + secp256k1_fe_mul_int(&r->y, 4 * (1 - a->infinity)); /* r->y = Y3 = 4*R*(3*Q-2*R^2)-4*M^4 (4) */ + + /** In case a->infinity == 1, the above code results in r->x, r->y, and r->z all equal to 0. + * Add b->x to x, b->y to y, and 1 to z in that case. + */ + t = b->x; secp256k1_fe_mul_int(&t, a->infinity); + secp256k1_fe_add(&r->x, &t); + t = b->y; secp256k1_fe_mul_int(&t, a->infinity); + secp256k1_fe_add(&r->y, &t); + secp256k1_fe_set_int(&t, a->infinity); + secp256k1_fe_add(&r->z, &t); + r->infinity = infinity; +} + + + +static void secp256k1_gej_get_hex(char *r, int *rlen, const secp256k1_gej_t *a) { + secp256k1_gej_t c = *a; + secp256k1_ge_t t; secp256k1_ge_set_gej(&t, &c); + secp256k1_ge_get_hex(r, rlen, &t); +} + +#ifdef USE_ENDOMORPHISM +static void secp256k1_gej_mul_lambda(secp256k1_gej_t *r, const secp256k1_gej_t *a) { + const secp256k1_fe_t *beta = &secp256k1_ge_consts->beta; + *r = *a; + secp256k1_fe_mul(&r->x, &r->x, beta); +} + +static void secp256k1_gej_split_exp_var(secp256k1_num_t *r1, secp256k1_num_t *r2, const secp256k1_num_t *a) { + const secp256k1_ge_consts_t *c = secp256k1_ge_consts; + secp256k1_num_t bnc1, bnc2, bnt1, bnt2, bnn2; + + secp256k1_num_copy(&bnn2, &c->order); + secp256k1_num_shift(&bnn2, 1); + + secp256k1_num_mul(&bnc1, a, &c->a1b2); + secp256k1_num_add(&bnc1, &bnc1, &bnn2); + secp256k1_num_div(&bnc1, &bnc1, &c->order); + + secp256k1_num_mul(&bnc2, a, &c->b1); + secp256k1_num_add(&bnc2, &bnc2, &bnn2); + secp256k1_num_div(&bnc2, &bnc2, &c->order); + + secp256k1_num_mul(&bnt1, &bnc1, &c->a1b2); + secp256k1_num_mul(&bnt2, &bnc2, &c->a2); + secp256k1_num_add(&bnt1, &bnt1, &bnt2); + secp256k1_num_sub(r1, a, &bnt1); + secp256k1_num_mul(&bnt1, &bnc1, &c->b1); + secp256k1_num_mul(&bnt2, &bnc2, &c->a1b2); + secp256k1_num_sub(r2, &bnt1, &bnt2); +} +#endif + + +static void secp256k1_ge_start(void) { + static const unsigned char secp256k1_ge_consts_order[] = { + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF, + 0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFF,0xFE, + 0xBA,0xAE,0xDC,0xE6,0xAF,0x48,0xA0,0x3B, + 0xBF,0xD2,0x5E,0x8C,0xD0,0x36,0x41,0x41 + }; + static const unsigned char secp256k1_ge_consts_g_x[] = { + 0x79,0xBE,0x66,0x7E,0xF9,0xDC,0xBB,0xAC, + 0x55,0xA0,0x62,0x95,0xCE,0x87,0x0B,0x07, + 0x02,0x9B,0xFC,0xDB,0x2D,0xCE,0x28,0xD9, + 0x59,0xF2,0x81,0x5B,0x16,0xF8,0x17,0x98 + }; + static const unsigned char secp256k1_ge_consts_g_y[] = { + 0x48,0x3A,0xDA,0x77,0x26,0xA3,0xC4,0x65, + 0x5D,0xA4,0xFB,0xFC,0x0E,0x11,0x08,0xA8, + 0xFD,0x17,0xB4,0x48,0xA6,0x85,0x54,0x19, + 0x9C,0x47,0xD0,0x8F,0xFB,0x10,0xD4,0xB8 + }; +#ifdef USE_ENDOMORPHISM + /* properties of secp256k1's efficiently computable endomorphism */ + static const unsigned char secp256k1_ge_consts_lambda[] = { + 0x53,0x63,0xad,0x4c,0xc0,0x5c,0x30,0xe0, + 0xa5,0x26,0x1c,0x02,0x88,0x12,0x64,0x5a, + 0x12,0x2e,0x22,0xea,0x20,0x81,0x66,0x78, + 0xdf,0x02,0x96,0x7c,0x1b,0x23,0xbd,0x72 + }; + static const unsigned char secp256k1_ge_consts_beta[] = { + 0x7a,0xe9,0x6a,0x2b,0x65,0x7c,0x07,0x10, + 0x6e,0x64,0x47,0x9e,0xac,0x34,0x34,0xe9, + 0x9c,0xf0,0x49,0x75,0x12,0xf5,0x89,0x95, + 0xc1,0x39,0x6c,0x28,0x71,0x95,0x01,0xee + }; + static const unsigned char secp256k1_ge_consts_a1b2[] = { + 0x30,0x86,0xd2,0x21,0xa7,0xd4,0x6b,0xcd, + 0xe8,0x6c,0x90,0xe4,0x92,0x84,0xeb,0x15 + }; + static const unsigned char secp256k1_ge_consts_b1[] = { + 0xe4,0x43,0x7e,0xd6,0x01,0x0e,0x88,0x28, + 0x6f,0x54,0x7f,0xa9,0x0a,0xbf,0xe4,0xc3 + }; + static const unsigned char secp256k1_ge_consts_a2[] = { + 0x01, + 0x14,0xca,0x50,0xf7,0xa8,0xe2,0xf3,0xf6, + 0x57,0xc1,0x10,0x8d,0x9d,0x44,0xcf,0xd8 + }; +#endif + if (secp256k1_ge_consts == NULL) { + secp256k1_ge_consts_t *ret = (secp256k1_ge_consts_t*)malloc(sizeof(secp256k1_ge_consts_t)); + secp256k1_num_set_bin(&ret->order, secp256k1_ge_consts_order, sizeof(secp256k1_ge_consts_order)); + secp256k1_num_copy(&ret->half_order, &ret->order); + secp256k1_num_shift(&ret->half_order, 1); +#ifdef USE_ENDOMORPHISM + secp256k1_num_set_bin(&ret->lambda, secp256k1_ge_consts_lambda, sizeof(secp256k1_ge_consts_lambda)); + secp256k1_num_set_bin(&ret->a1b2, secp256k1_ge_consts_a1b2, sizeof(secp256k1_ge_consts_a1b2)); + secp256k1_num_set_bin(&ret->a2, secp256k1_ge_consts_a2, sizeof(secp256k1_ge_consts_a2)); + secp256k1_num_set_bin(&ret->b1, secp256k1_ge_consts_b1, sizeof(secp256k1_ge_consts_b1)); + secp256k1_fe_set_b32(&ret->beta, secp256k1_ge_consts_beta); +#endif + secp256k1_fe_t g_x, g_y; + secp256k1_fe_set_b32(&g_x, secp256k1_ge_consts_g_x); + secp256k1_fe_set_b32(&g_y, secp256k1_ge_consts_g_y); + secp256k1_ge_set_xy(&ret->g, &g_x, &g_y); + secp256k1_ge_consts = ret; + } +} + +static void secp256k1_ge_stop(void) { + if (secp256k1_ge_consts != NULL) { + secp256k1_ge_consts_t *c = (secp256k1_ge_consts_t*)secp256k1_ge_consts; + free((void*)c); + secp256k1_ge_consts = NULL; + } +} + +#endif diff --git a/src/java/org/bitcoin/NativeSecp256k1.java b/src/java/org/bitcoin/NativeSecp256k1.java new file mode 100644 index 0000000000..90a498eaa2 --- /dev/null +++ b/src/java/org/bitcoin/NativeSecp256k1.java @@ -0,0 +1,60 @@ +package org.bitcoin; + +import java.nio.ByteBuffer; +import java.nio.ByteOrder; + +import com.google.common.base.Preconditions; + + +/** + * This class holds native methods to handle ECDSA verification. + * You can find an example library that can be used for this at + * https://github.com/sipa/secp256k1 + */ +public class NativeSecp256k1 { + public static final boolean enabled; + static { + boolean isEnabled = true; + try { + System.loadLibrary("javasecp256k1"); + } catch (UnsatisfiedLinkError e) { + isEnabled = false; + } + enabled = isEnabled; + } + + private static ThreadLocal nativeECDSABuffer = new ThreadLocal(); + /** + * Verifies the given secp256k1 signature in native code. + * Calling when enabled == false is undefined (probably library not loaded) + * + * @param data The data which was signed, must be exactly 32 bytes + * @param signature The signature + * @param pub The public key which did the signing + */ + public static boolean verify(byte[] data, byte[] signature, byte[] pub) { + Preconditions.checkArgument(data.length == 32 && signature.length <= 520 && pub.length <= 520); + + ByteBuffer byteBuff = nativeECDSABuffer.get(); + if (byteBuff == null) { + byteBuff = ByteBuffer.allocateDirect(32 + 8 + 520 + 520); + byteBuff.order(ByteOrder.nativeOrder()); + nativeECDSABuffer.set(byteBuff); + } + byteBuff.rewind(); + byteBuff.put(data); + byteBuff.putInt(signature.length); + byteBuff.putInt(pub.length); + byteBuff.put(signature); + byteBuff.put(pub); + return secp256k1_ecdsa_verify(byteBuff) == 1; + } + + /** + * @param byteBuff signature format is byte[32] data, + * native-endian int signatureLength, native-endian int pubkeyLength, + * byte[signatureLength] signature, byte[pubkeyLength] pub + * @returns 1 for valid signature, anything else for invalid + */ + private static native int secp256k1_ecdsa_verify(ByteBuffer byteBuff); +} diff --git a/src/java/org_bitcoin_NativeSecp256k1.c b/src/java/org_bitcoin_NativeSecp256k1.c new file mode 100644 index 0000000000..bb4cd70728 --- /dev/null +++ b/src/java/org_bitcoin_NativeSecp256k1.c @@ -0,0 +1,23 @@ +#include "org_bitcoin_NativeSecp256k1.h" +#include "include/secp256k1.h" + +JNIEXPORT jint JNICALL Java_org_bitcoin_NativeSecp256k1_secp256k1_1ecdsa_1verify + (JNIEnv* env, jclass classObject, jobject byteBufferObject) +{ + unsigned char* data = (unsigned char*) (*env)->GetDirectBufferAddress(env, byteBufferObject); + int sigLen = *((int*)(data + 32)); + int pubLen = *((int*)(data + 32 + 4)); + + return secp256k1_ecdsa_verify(data, 32, data+32+8, sigLen, data+32+8+sigLen, pubLen); +} + +static void __javasecp256k1_attach(void) __attribute__((constructor)); +static void __javasecp256k1_detach(void) __attribute__((destructor)); + +static void __javasecp256k1_attach(void) { + secp256k1_start(SECP256K1_START_VERIFY); +} + +static void __javasecp256k1_detach(void) { + secp256k1_stop(); +} diff --git a/src/java/org_bitcoin_NativeSecp256k1.h b/src/java/org_bitcoin_NativeSecp256k1.h new file mode 100644 index 0000000000..d7fb004fa8 --- /dev/null +++ b/src/java/org_bitcoin_NativeSecp256k1.h @@ -0,0 +1,21 @@ +/* DO NOT EDIT THIS FILE - it is machine generated */ +#include +/* Header for class org_bitcoin_NativeSecp256k1 */ + +#ifndef _Included_org_bitcoin_NativeSecp256k1 +#define _Included_org_bitcoin_NativeSecp256k1 +#ifdef __cplusplus +extern "C" { +#endif +/* + * Class: org_bitcoin_NativeSecp256k1 + * Method: secp256k1_ecdsa_verify + * Signature: (Ljava/nio/ByteBuffer;)I + */ +JNIEXPORT jint JNICALL Java_org_bitcoin_NativeSecp256k1_secp256k1_1ecdsa_1verify + (JNIEnv *, jclass, jobject); + +#ifdef __cplusplus +} +#endif +#endif diff --git a/src/num.h b/src/num.h new file mode 100644 index 0000000000..c86f847858 --- /dev/null +++ b/src/num.h @@ -0,0 +1,100 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_NUM_ +#define _SECP256K1_NUM_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#if defined(USE_NUM_GMP) +#include "num_gmp.h" +#else +#error "Please select num implementation" +#endif + +/** Clear a number to prevent the leak of sensitive data. */ +static void secp256k1_num_clear(secp256k1_num_t *r); + +/** Copy a number. */ +static void secp256k1_num_copy(secp256k1_num_t *r, const secp256k1_num_t *a); + +/** Convert a number's absolute value to a binary big-endian string. + * There must be enough place. */ +static void secp256k1_num_get_bin(unsigned char *r, unsigned int rlen, const secp256k1_num_t *a); + +/** Set a number to the value of a binary big-endian string. */ +static void secp256k1_num_set_bin(secp256k1_num_t *r, const unsigned char *a, unsigned int alen); + +/** Set a number equal to a (signed) integer. */ +static void secp256k1_num_set_int(secp256k1_num_t *r, int a); + +/** Compute a modular inverse. The input must be less than the modulus. */ +static void secp256k1_num_mod_inverse(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *m); + +/** Multiply two numbers modulo another. */ +static void secp256k1_num_mod_mul(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b, const secp256k1_num_t *m); + +/** Compare the absolute value of two numbers. */ +static int secp256k1_num_cmp(const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Test whether two number are equal (including sign). */ +static int secp256k1_num_eq(const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Add two (signed) numbers. */ +static void secp256k1_num_add(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Subtract two (signed) numbers. */ +static void secp256k1_num_sub(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Multiply two (signed) numbers. */ +static void secp256k1_num_mul(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Divide two (signed) numbers. */ +static void secp256k1_num_div(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b); + +/** Replace a number by its remainder modulo m. M's sign is ignored. The result is a number between 0 and m-1, + even if r was negative. */ +static void secp256k1_num_mod(secp256k1_num_t *r, const secp256k1_num_t *m); + +/** Calculate the number of bits in (the absolute value of) a number. */ +static int secp256k1_num_bits(const secp256k1_num_t *a); + +/** Right-shift the passed number by bits bits, and return those bits. */ +static int secp256k1_num_shift(secp256k1_num_t *r, int bits); + +/** Check whether a number is zero. */ +static int secp256k1_num_is_zero(const secp256k1_num_t *a); + +/** Check whether a number is odd. */ +static int secp256k1_num_is_odd(const secp256k1_num_t *a); + +/** Check whether a number is strictly negative. */ +static int secp256k1_num_is_neg(const secp256k1_num_t *a); + +/** Check whether a particular bit is set in a number. */ +static int secp256k1_num_get_bit(const secp256k1_num_t *a, int pos); + +/** Increase a number by 1. */ +static void secp256k1_num_inc(secp256k1_num_t *r); + +/** Set a number equal to the value of a hex string (unsigned). */ +static void secp256k1_num_set_hex(secp256k1_num_t *r, const char *a, int alen); + +/** Convert (the absolute value of) a number to a hexadecimal string. */ +static void secp256k1_num_get_hex(char *r, int rlen, const secp256k1_num_t *a); + +/** Split a number into a low and high part. */ +static void secp256k1_num_split(secp256k1_num_t *rl, secp256k1_num_t *rh, const secp256k1_num_t *a, int bits); + +/** Change a number's sign. */ +static void secp256k1_num_negate(secp256k1_num_t *r); + +/** Get a bunch of bits from a number. */ +static int secp256k1_num_get_bits(const secp256k1_num_t *a, int offset, int count); + +#endif diff --git a/src/num_gmp.h b/src/num_gmp.h new file mode 100644 index 0000000000..baa1f2bf2e --- /dev/null +++ b/src/num_gmp.h @@ -0,0 +1,20 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_NUM_REPR_ +#define _SECP256K1_NUM_REPR_ + +#include + +#define NUM_LIMBS ((256+GMP_NUMB_BITS-1)/GMP_NUMB_BITS) + +typedef struct { + mp_limb_t data[2*NUM_LIMBS]; + int neg; + int limbs; +} secp256k1_num_t; + +#endif diff --git a/src/num_gmp_impl.h b/src/num_gmp_impl.h new file mode 100644 index 0000000000..e45a59e0cd --- /dev/null +++ b/src/num_gmp_impl.h @@ -0,0 +1,376 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_NUM_REPR_IMPL_H_ +#define _SECP256K1_NUM_REPR_IMPL_H_ + +#include +#include +#include + +#include "util.h" +#include "num.h" + +#ifdef VERIFY +static void secp256k1_num_sanity(const secp256k1_num_t *a) { + VERIFY_CHECK(a->limbs == 1 || (a->limbs > 1 && a->data[a->limbs-1] != 0)); +} +#else +#define secp256k1_num_sanity(a) do { } while(0) +#endif + +static void secp256k1_num_init(secp256k1_num_t *r) { + r->neg = 0; + r->limbs = 1; + r->data[0] = 0; +} + +static void secp256k1_num_clear(secp256k1_num_t *r) { + memset(r, 0, sizeof(*r)); +} + +static void secp256k1_num_free(secp256k1_num_t *r) { + (void)r; +} + +static void secp256k1_num_copy(secp256k1_num_t *r, const secp256k1_num_t *a) { + *r = *a; +} + +static int secp256k1_num_bits(const secp256k1_num_t *a) { + int ret=(a->limbs-1)*GMP_NUMB_BITS; + mp_limb_t x=a->data[a->limbs-1]; + while (x) { + x >>= 1; + ret++; + } + return ret; +} + + +static void secp256k1_num_get_bin(unsigned char *r, unsigned int rlen, const secp256k1_num_t *a) { + unsigned char tmp[65]; + int len = 0; + if (a->limbs>1 || a->data[0] != 0) { + len = mpn_get_str(tmp, 256, (mp_limb_t*)a->data, a->limbs); + } + int shift = 0; + while (shift < len && tmp[shift] == 0) shift++; + VERIFY_CHECK(len-shift <= (int)rlen); + memset(r, 0, rlen - len + shift); + if (len > shift) { + memcpy(r + rlen - len + shift, tmp + shift, len - shift); + } + memset(tmp, 0, sizeof(tmp)); +} + +static void secp256k1_num_set_bin(secp256k1_num_t *r, const unsigned char *a, unsigned int alen) { + VERIFY_CHECK(alen > 0); + VERIFY_CHECK(alen <= 64); + int len = mpn_set_str(r->data, a, alen, 256); + VERIFY_CHECK(len <= NUM_LIMBS*2); + r->limbs = len; + r->neg = 0; + while (r->limbs > 1 && r->data[r->limbs-1]==0) r->limbs--; +} + +static void secp256k1_num_set_int(secp256k1_num_t *r, int a) { + r->limbs = 1; + r->neg = (a < 0); + r->data[0] = (a < 0) ? -a : a; +} + +static void secp256k1_num_add_abs(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + mp_limb_t c = mpn_add(r->data, a->data, a->limbs, b->data, b->limbs); + r->limbs = a->limbs; + if (c != 0) { + VERIFY_CHECK(r->limbs < 2*NUM_LIMBS); + r->data[r->limbs++] = c; + } +} + +static void secp256k1_num_sub_abs(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + mp_limb_t c = mpn_sub(r->data, a->data, a->limbs, b->data, b->limbs); + VERIFY_CHECK(c == 0); + r->limbs = a->limbs; + while (r->limbs > 1 && r->data[r->limbs-1]==0) r->limbs--; +} + +static void secp256k1_num_mod(secp256k1_num_t *r, const secp256k1_num_t *m) { + secp256k1_num_sanity(r); + secp256k1_num_sanity(m); + + if (r->limbs >= m->limbs) { + mp_limb_t t[2*NUM_LIMBS]; + mpn_tdiv_qr(t, r->data, 0, r->data, r->limbs, m->data, m->limbs); + memset(t, 0, sizeof(t)); + r->limbs = m->limbs; + while (r->limbs > 1 && r->data[r->limbs-1]==0) r->limbs--; + } + + if (r->neg && (r->limbs > 1 || r->data[0] != 0)) { + secp256k1_num_sub_abs(r, m, r); + r->neg = 0; + } +} + +static void secp256k1_num_mod_inverse(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *m) { + secp256k1_num_sanity(a); + secp256k1_num_sanity(m); + + /** mpn_gcdext computes: (G,S) = gcdext(U,V), where + * * G = gcd(U,V) + * * G = U*S + V*T + * * U has equal or more limbs than V, and V has no padding + * If we set U to be (a padded version of) a, and V = m: + * G = a*S + m*T + * G = a*S mod m + * Assuming G=1: + * S = 1/a mod m + */ + VERIFY_CHECK(m->limbs <= NUM_LIMBS); + VERIFY_CHECK(m->data[m->limbs-1] != 0); + mp_limb_t g[NUM_LIMBS+1]; + mp_limb_t u[NUM_LIMBS+1]; + mp_limb_t v[NUM_LIMBS+1]; + for (int i=0; i < m->limbs; i++) { + u[i] = (i < a->limbs) ? a->data[i] : 0; + v[i] = m->data[i]; + } + mp_size_t sn = NUM_LIMBS+1; + mp_size_t gn = mpn_gcdext(g, r->data, &sn, u, m->limbs, v, m->limbs); + VERIFY_CHECK(gn == 1); + VERIFY_CHECK(g[0] == 1); + r->neg = a->neg ^ m->neg; + if (sn < 0) { + mpn_sub(r->data, m->data, m->limbs, r->data, -sn); + r->limbs = m->limbs; + while (r->limbs > 1 && r->data[r->limbs-1]==0) r->limbs--; + } else { + r->limbs = sn; + } + memset(g, 0, sizeof(g)); + memset(u, 0, sizeof(u)); + memset(v, 0, sizeof(v)); +} + +static int secp256k1_num_is_zero(const secp256k1_num_t *a) { + return (a->limbs == 1 && a->data[0] == 0); +} + +static int secp256k1_num_is_odd(const secp256k1_num_t *a) { + return a->data[0] & 1; +} + +static int secp256k1_num_is_neg(const secp256k1_num_t *a) { + return (a->limbs > 1 || a->data[0] != 0) && a->neg; +} + +static int secp256k1_num_cmp(const secp256k1_num_t *a, const secp256k1_num_t *b) { + if (a->limbs > b->limbs) return 1; + if (a->limbs < b->limbs) return -1; + return mpn_cmp(a->data, b->data, a->limbs); +} + +static int secp256k1_num_eq(const secp256k1_num_t *a, const secp256k1_num_t *b) { + if (a->limbs > b->limbs) return 0; + if (a->limbs < b->limbs) return 0; + if ((a->neg && !secp256k1_num_is_zero(a)) != (b->neg && !secp256k1_num_is_zero(b))) return 0; + return mpn_cmp(a->data, b->data, a->limbs) == 0; +} + +static void secp256k1_num_subadd(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b, int bneg) { + if (!(b->neg ^ bneg ^ a->neg)) { /* a and b have the same sign */ + r->neg = a->neg; + if (a->limbs >= b->limbs) { + secp256k1_num_add_abs(r, a, b); + } else { + secp256k1_num_add_abs(r, b, a); + } + } else { + if (secp256k1_num_cmp(a, b) > 0) { + r->neg = a->neg; + secp256k1_num_sub_abs(r, a, b); + } else { + r->neg = b->neg ^ bneg; + secp256k1_num_sub_abs(r, b, a); + } + } +} + +static void secp256k1_num_add(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + secp256k1_num_sanity(a); + secp256k1_num_sanity(b); + secp256k1_num_subadd(r, a, b, 0); +} + +static void secp256k1_num_sub(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + secp256k1_num_sanity(a); + secp256k1_num_sanity(b); + secp256k1_num_subadd(r, a, b, 1); +} + +static void secp256k1_num_mul(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + secp256k1_num_sanity(a); + secp256k1_num_sanity(b); + + mp_limb_t tmp[2*NUM_LIMBS+1]; + VERIFY_CHECK(a->limbs + b->limbs <= 2*NUM_LIMBS+1); + if ((a->limbs==1 && a->data[0]==0) || (b->limbs==1 && b->data[0]==0)) { + r->limbs = 1; + r->neg = 0; + r->data[0] = 0; + return; + } + if (a->limbs >= b->limbs) + mpn_mul(tmp, a->data, a->limbs, b->data, b->limbs); + else + mpn_mul(tmp, b->data, b->limbs, a->data, a->limbs); + r->limbs = a->limbs + b->limbs; + if (r->limbs > 1 && tmp[r->limbs - 1]==0) r->limbs--; + VERIFY_CHECK(r->limbs <= 2*NUM_LIMBS); + mpn_copyi(r->data, tmp, r->limbs); + r->neg = a->neg ^ b->neg; + memset(tmp, 0, sizeof(tmp)); +} + +static void secp256k1_num_div(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b) { + secp256k1_num_sanity(a); + secp256k1_num_sanity(b); + if (b->limbs > a->limbs) { + r->limbs = 1; + r->data[0] = 0; + r->neg = 0; + return; + } + + mp_limb_t quo[2*NUM_LIMBS+1]; + mp_limb_t rem[2*NUM_LIMBS+1]; + mpn_tdiv_qr(quo, rem, 0, a->data, a->limbs, b->data, b->limbs); + mpn_copyi(r->data, quo, a->limbs - b->limbs + 1); + r->limbs = a->limbs - b->limbs + 1; + while (r->limbs > 1 && r->data[r->limbs - 1]==0) r->limbs--; + r->neg = a->neg ^ b->neg; +} + +static void secp256k1_num_mod_mul(secp256k1_num_t *r, const secp256k1_num_t *a, const secp256k1_num_t *b, const secp256k1_num_t *m) { + secp256k1_num_mul(r, a, b); + secp256k1_num_mod(r, m); +} + + +static int secp256k1_num_shift(secp256k1_num_t *r, int bits) { + VERIFY_CHECK(bits <= GMP_NUMB_BITS); + mp_limb_t ret = mpn_rshift(r->data, r->data, r->limbs, bits); + if (r->limbs>1 && r->data[r->limbs-1]==0) r->limbs--; + ret >>= (GMP_NUMB_BITS - bits); + return ret; +} + +static int secp256k1_num_get_bit(const secp256k1_num_t *a, int pos) { + return (a->limbs*GMP_NUMB_BITS > pos) && ((a->data[pos/GMP_NUMB_BITS] >> (pos % GMP_NUMB_BITS)) & 1); +} + +static void secp256k1_num_inc(secp256k1_num_t *r) { + mp_limb_t ret = mpn_add_1(r->data, r->data, r->limbs, (mp_limb_t)1); + if (ret) { + VERIFY_CHECK(r->limbs < 2*NUM_LIMBS); + r->data[r->limbs++] = ret; + } +} + +static void secp256k1_num_set_hex(secp256k1_num_t *r, const char *a, int alen) { + static const unsigned char cvt[256] = { + 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, 0,0,0,0,0,0,0,0,0,0, + 0, 1, 2, 3, 4, 5, 6,7,8,9,0,0,0,0,0,0, + 0,10,11,12,13,14,15,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 0,10,11,12,13,14,15,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 0, 0, 0, 0, 0, 0, 0,0,0,0,0,0,0,0,0,0, + 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, 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, 0,0,0,0,0,0,0,0,0,0, + 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 + }; + unsigned char num[257] = {}; + for (int i=0; ilimbs = mpn_set_str(r->data, num, alen, 16); + r->neg = 0; + while (r->limbs > 1 && r->data[r->limbs-1] == 0) r->limbs--; +} + +static void secp256k1_num_get_hex(char *r, int rlen, const secp256k1_num_t *a) { + static const unsigned char cvt[16] = {'0', '1', '2', '3', '4', '5', '6', '7', '8', '9', 'A', 'B', 'C', 'D', 'E', 'F'}; + unsigned char *tmp = malloc(257); + mp_size_t len = mpn_get_str(tmp, 16, (mp_limb_t*)a->data, a->limbs); + VERIFY_CHECK(len <= rlen); + for (int i=0; i= 0); + VERIFY_CHECK(rlen-len+i < rlen); + VERIFY_CHECK(tmp[i] < 16); + r[rlen-len+i] = cvt[tmp[i]]; + } + for (int i=0; i= 0); + VERIFY_CHECK(i < rlen); + r[i] = cvt[0]; + } + free(tmp); +} + +static void secp256k1_num_split(secp256k1_num_t *rl, secp256k1_num_t *rh, const secp256k1_num_t *a, int bits) { + VERIFY_CHECK(bits > 0); + rh->neg = a->neg; + if (bits >= a->limbs * GMP_NUMB_BITS) { + *rl = *a; + rh->limbs = 1; + rh->data[0] = 0; + return; + } + rl->limbs = 0; + rl->neg = a->neg; + int left = bits; + while (left >= GMP_NUMB_BITS) { + rl->data[rl->limbs] = a->data[rl->limbs]; + rl->limbs++; + left -= GMP_NUMB_BITS; + } + if (left == 0) { + mpn_copyi(rh->data, a->data + rl->limbs, a->limbs - rl->limbs); + rh->limbs = a->limbs - rl->limbs; + } else { + mpn_rshift(rh->data, a->data + rl->limbs, a->limbs - rl->limbs, left); + rh->limbs = a->limbs - rl->limbs; + while (rh->limbs>1 && rh->data[rh->limbs-1]==0) rh->limbs--; + } + if (left > 0) { + rl->data[rl->limbs] = a->data[rl->limbs] & ((((mp_limb_t)1) << left) - 1); + rl->limbs++; + } + while (rl->limbs>1 && rl->data[rl->limbs-1]==0) rl->limbs--; +} + +static void secp256k1_num_negate(secp256k1_num_t *r) { + r->neg ^= 1; +} + +static int secp256k1_num_get_bits(const secp256k1_num_t *a, int offset, int count) { + int ret = 0; + for (int i = 0; i < count; i++) { + ret |= ((a->data[(offset + i) / GMP_NUMB_BITS] >> ((offset + i) % GMP_NUMB_BITS)) & 1) << i; + } + return ret; +} + +#endif diff --git a/src/num_impl.h b/src/num_impl.h new file mode 100644 index 0000000000..f73d3ceea8 --- /dev/null +++ b/src/num_impl.h @@ -0,0 +1,22 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_NUM_IMPL_H_ +#define _SECP256K1_NUM_IMPL_H_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#include "num.h" + +#if defined(USE_NUM_GMP) +#include "num_gmp_impl.h" +#else +#error "Please select num implementation" +#endif + +#endif diff --git a/src/scalar.h b/src/scalar.h new file mode 100644 index 0000000000..3baacb3721 --- /dev/null +++ b/src/scalar.h @@ -0,0 +1,63 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_ +#define _SECP256K1_SCALAR_ + +#include "num.h" + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#if defined(USE_SCALAR_4X64) +#include "scalar_4x64.h" +#elif defined(USE_SCALAR_8X32) +#include "scalar_8x32.h" +#else +#error "Please select scalar implementation" +#endif + +/** Clear a scalar to prevent the leak of sensitive data. */ +static void secp256k1_scalar_clear(secp256k1_scalar_t *r); + +/** Access bits from a scalar. */ +static int secp256k1_scalar_get_bits(const secp256k1_scalar_t *a, int offset, int count); + +/** Set a scalar from a big endian byte array. */ +static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char *bin, int *overflow); + +/** Convert a scalar to a byte array. */ +static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar_t* a); + +/** Add two scalars together (modulo the group order). */ +static void secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b); + +/** Multiply two scalars (modulo the group order). */ +static void secp256k1_scalar_mul(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b); + +/** Compute the square of a scalar (modulo the group order). */ +static void secp256k1_scalar_sqr(secp256k1_scalar_t *r, const secp256k1_scalar_t *a); + +/** Compute the inverse of a scalar (modulo the group order). */ +static void secp256k1_scalar_inverse(secp256k1_scalar_t *r, const secp256k1_scalar_t *a); + +/** Compute the complement of a scalar (modulo the group order). */ +static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scalar_t *a); + +/** Check whether a scalar equals zero. */ +static int secp256k1_scalar_is_zero(const secp256k1_scalar_t *a); + +/** Check whether a scalar equals one. */ +static int secp256k1_scalar_is_one(const secp256k1_scalar_t *a); + +/** Check whether a scalar is higher than the group order divided by 2. */ +static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a); + +/** Convert a scalar to a number. */ +static void secp256k1_scalar_get_num(secp256k1_num_t *r, const secp256k1_scalar_t *a); + +#endif diff --git a/src/scalar_4x64.h b/src/scalar_4x64.h new file mode 100644 index 0000000000..5a751c6862 --- /dev/null +++ b/src/scalar_4x64.h @@ -0,0 +1,17 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_REPR_ +#define _SECP256K1_SCALAR_REPR_ + +#include + +/** A scalar modulo the group order of the secp256k1 curve. */ +typedef struct { + uint64_t d[4]; +} secp256k1_scalar_t; + +#endif diff --git a/src/scalar_4x64_impl.h b/src/scalar_4x64_impl.h new file mode 100644 index 0000000000..f78718234f --- /dev/null +++ b/src/scalar_4x64_impl.h @@ -0,0 +1,359 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_REPR_IMPL_H_ +#define _SECP256K1_SCALAR_REPR_IMPL_H_ + +typedef unsigned __int128 uint128_t; + +/* Limbs of the secp256k1 order. */ +#define SECP256K1_N_0 ((uint64_t)0xBFD25E8CD0364141ULL) +#define SECP256K1_N_1 ((uint64_t)0xBAAEDCE6AF48A03BULL) +#define SECP256K1_N_2 ((uint64_t)0xFFFFFFFFFFFFFFFEULL) +#define SECP256K1_N_3 ((uint64_t)0xFFFFFFFFFFFFFFFFULL) + +/* Limbs of 2^256 minus the secp256k1 order. */ +#define SECP256K1_N_C_0 (~SECP256K1_N_0 + 1) +#define SECP256K1_N_C_1 (~SECP256K1_N_1) +#define SECP256K1_N_C_2 (1) + +/* Limbs of half the secp256k1 order. */ +#define SECP256K1_N_H_0 ((uint64_t)0xDFE92F46681B20A0ULL) +#define SECP256K1_N_H_1 ((uint64_t)0x5D576E7357A4501DULL) +#define SECP256K1_N_H_2 ((uint64_t)0xFFFFFFFFFFFFFFFFULL) +#define SECP256K1_N_H_3 ((uint64_t)0x7FFFFFFFFFFFFFFFULL) + +SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar_t *r) { + r->d[0] = 0; + r->d[1] = 0; + r->d[2] = 0; + r->d[3] = 0; +} + +SECP256K1_INLINE static int secp256k1_scalar_get_bits(const secp256k1_scalar_t *a, int offset, int count) { + VERIFY_CHECK((offset + count - 1) / 64 == offset / 64); + return (a->d[offset / 64] >> (offset % 64)) & ((((uint64_t)1) << count) - 1); +} + +SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar_t *a) { + int yes = 0; + int no = 0; + no |= (a->d[3] < SECP256K1_N_3); /* No need for a > check. */ + no |= (a->d[2] < SECP256K1_N_2); + yes |= (a->d[2] > SECP256K1_N_2) & ~no; + no |= (a->d[1] < SECP256K1_N_1); + yes |= (a->d[1] > SECP256K1_N_1) & ~no; + yes |= (a->d[0] >= SECP256K1_N_0) & ~no; + return yes; +} + +SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar_t *r, unsigned int overflow) { + VERIFY_CHECK(overflow <= 1); + uint128_t t = (uint128_t)r->d[0] + overflow * SECP256K1_N_C_0; + r->d[0] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint128_t)r->d[1] + overflow * SECP256K1_N_C_1; + r->d[1] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint128_t)r->d[2] + overflow * SECP256K1_N_C_2; + r->d[2] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint64_t)r->d[3]; + r->d[3] = t & 0xFFFFFFFFFFFFFFFFULL; + return overflow; +} + +static void secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { + uint128_t t = (uint128_t)a->d[0] + b->d[0]; + r->d[0] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint128_t)a->d[1] + b->d[1]; + r->d[1] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint128_t)a->d[2] + b->d[2]; + r->d[2] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + t += (uint128_t)a->d[3] + b->d[3]; + r->d[3] = t & 0xFFFFFFFFFFFFFFFFULL; t >>= 64; + secp256k1_scalar_reduce(r, t + secp256k1_scalar_check_overflow(r)); +} + +static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char *b32, int *overflow) { + r->d[0] = (uint64_t)b32[31] | (uint64_t)b32[30] << 8 | (uint64_t)b32[29] << 16 | (uint64_t)b32[28] << 24 | (uint64_t)b32[27] << 32 | (uint64_t)b32[26] << 40 | (uint64_t)b32[25] << 48 | (uint64_t)b32[24] << 56; + r->d[1] = (uint64_t)b32[23] | (uint64_t)b32[22] << 8 | (uint64_t)b32[21] << 16 | (uint64_t)b32[20] << 24 | (uint64_t)b32[19] << 32 | (uint64_t)b32[18] << 40 | (uint64_t)b32[17] << 48 | (uint64_t)b32[16] << 56; + r->d[2] = (uint64_t)b32[15] | (uint64_t)b32[14] << 8 | (uint64_t)b32[13] << 16 | (uint64_t)b32[12] << 24 | (uint64_t)b32[11] << 32 | (uint64_t)b32[10] << 40 | (uint64_t)b32[9] << 48 | (uint64_t)b32[8] << 56; + r->d[3] = (uint64_t)b32[7] | (uint64_t)b32[6] << 8 | (uint64_t)b32[5] << 16 | (uint64_t)b32[4] << 24 | (uint64_t)b32[3] << 32 | (uint64_t)b32[2] << 40 | (uint64_t)b32[1] << 48 | (uint64_t)b32[0] << 56; + int over = secp256k1_scalar_reduce(r, secp256k1_scalar_check_overflow(r)); + if (overflow) { + *overflow = over; + } +} + +static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar_t* a) { + bin[0] = a->d[3] >> 56; bin[1] = a->d[3] >> 48; bin[2] = a->d[3] >> 40; bin[3] = a->d[3] >> 32; bin[4] = a->d[3] >> 24; bin[5] = a->d[3] >> 16; bin[6] = a->d[3] >> 8; bin[7] = a->d[3]; + bin[8] = a->d[2] >> 56; bin[9] = a->d[2] >> 48; bin[10] = a->d[2] >> 40; bin[11] = a->d[2] >> 32; bin[12] = a->d[2] >> 24; bin[13] = a->d[2] >> 16; bin[14] = a->d[2] >> 8; bin[15] = a->d[2]; + bin[16] = a->d[1] >> 56; bin[17] = a->d[1] >> 48; bin[18] = a->d[1] >> 40; bin[19] = a->d[1] >> 32; bin[20] = a->d[1] >> 24; bin[21] = a->d[1] >> 16; bin[22] = a->d[1] >> 8; bin[23] = a->d[1]; + bin[24] = a->d[0] >> 56; bin[25] = a->d[0] >> 48; bin[26] = a->d[0] >> 40; bin[27] = a->d[0] >> 32; bin[28] = a->d[0] >> 24; bin[29] = a->d[0] >> 16; bin[30] = a->d[0] >> 8; bin[31] = a->d[0]; +} + +SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar_t *a) { + return (a->d[0] | a->d[1] | a->d[2] | a->d[3]) == 0; +} + +static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { + uint64_t nonzero = 0xFFFFFFFFFFFFFFFFULL * (secp256k1_scalar_is_zero(a) == 0); + uint128_t t = (uint128_t)(~a->d[0]) + SECP256K1_N_0 + 1; + r->d[0] = t & nonzero; t >>= 64; + t += (uint128_t)(~a->d[1]) + SECP256K1_N_1; + r->d[1] = t & nonzero; t >>= 64; + t += (uint128_t)(~a->d[2]) + SECP256K1_N_2; + r->d[2] = t & nonzero; t >>= 64; + t += (uint128_t)(~a->d[3]) + SECP256K1_N_3; + r->d[3] = t & nonzero; +} + +SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar_t *a) { + return ((a->d[0] ^ 1) | a->d[1] | a->d[2] | a->d[3]) == 0; +} + +static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) { + int yes = 0; + int no = 0; + no |= (a->d[3] < SECP256K1_N_H_3); + yes |= (a->d[3] > SECP256K1_N_H_3) & ~no; + no |= (a->d[2] < SECP256K1_N_H_2) & ~yes; /* No need for a > check. */ + no |= (a->d[1] < SECP256K1_N_H_1) & ~yes; + yes |= (a->d[1] > SECP256K1_N_H_1) & ~no; + yes |= (a->d[0] > SECP256K1_N_H_0) & ~no; + return yes; +} + +/* Inspired by the macros in OpenSSL's crypto/bn/asm/x86_64-gcc.c. */ + +/** Add a*b to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define muladd(a,b) { \ + uint64_t tl, th; \ + { \ + uint128_t t = (uint128_t)a * b; \ + th = t >> 64; /* at most 0xFFFFFFFFFFFFFFFE */ \ + tl = t; \ + } \ + c0 += tl; /* overflow is handled on the next line */ \ + th += (c0 < tl) ? 1 : 0; /* at most 0xFFFFFFFFFFFFFFFF */ \ + c1 += th; /* overflow is handled on the next line */ \ + c2 += (c1 < th) ? 1 : 0; /* never overflows by contract (verified in the next line) */ \ + VERIFY_CHECK((c1 >= th) || (c2 != 0)); \ +} + +/** Add a*b to the number defined by (c0,c1). c1 must never overflow. */ +#define muladd_fast(a,b) { \ + uint64_t tl, th; \ + { \ + uint128_t t = (uint128_t)a * b; \ + th = t >> 64; /* at most 0xFFFFFFFFFFFFFFFE */ \ + tl = t; \ + } \ + c0 += tl; /* overflow is handled on the next line */ \ + th += (c0 < tl) ? 1 : 0; /* at most 0xFFFFFFFFFFFFFFFF */ \ + c1 += th; /* never overflows by contract (verified in the next line) */ \ + VERIFY_CHECK(c1 >= th); \ +} + +/** Add 2*a*b to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define muladd2(a,b) { \ + uint64_t tl, th; \ + { \ + uint128_t t = (uint128_t)a * b; \ + th = t >> 64; /* at most 0xFFFFFFFFFFFFFFFE */ \ + tl = t; \ + } \ + uint64_t th2 = th + th; /* at most 0xFFFFFFFFFFFFFFFE (in case th was 0x7FFFFFFFFFFFFFFF) */ \ + c2 += (th2 < th) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((th2 >= th) || (c2 != 0)); \ + uint64_t tl2 = tl + tl; /* at most 0xFFFFFFFFFFFFFFFE (in case the lowest 63 bits of tl were 0x7FFFFFFFFFFFFFFF) */ \ + th2 += (tl2 < tl) ? 1 : 0; /* at most 0xFFFFFFFFFFFFFFFF */ \ + c0 += tl2; /* overflow is handled on the next line */ \ + th2 += (c0 < tl2) ? 1 : 0; /* second overflow is handled on the next line */ \ + c2 += (c0 < tl2) & (th2 == 0); /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c0 >= tl2) || (th2 != 0) || (c2 != 0)); \ + c1 += th2; /* overflow is handled on the next line */ \ + c2 += (c1 < th2) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c1 >= th2) || (c2 != 0)); \ +} + +/** Add a to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define sumadd(a) { \ + c0 += (a); /* overflow is handled on the next line */ \ + unsigned int over = (c0 < (a)) ? 1 : 0; \ + c1 += over; /* overflow is handled on the next line */ \ + c2 += (c1 < over) ? 1 : 0; /* never overflows by contract */ \ +} + +/** Add a to the number defined by (c0,c1). c1 must never overflow, c2 must be zero. */ +#define sumadd_fast(a) { \ + c0 += (a); /* overflow is handled on the next line */ \ + c1 += (c0 < (a)) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c1 != 0) | (c0 >= (a))); \ + VERIFY_CHECK(c2 == 0); \ +} + +/** Extract the lowest 64 bits of (c0,c1,c2) into n, and left shift the number 64 bits. */ +#define extract(n) { \ + (n) = c0; \ + c0 = c1; \ + c1 = c2; \ + c2 = 0; \ +} + +/** Extract the lowest 64 bits of (c0,c1,c2) into n, and left shift the number 64 bits. c2 is required to be zero. */ +#define extract_fast(n) { \ + (n) = c0; \ + c0 = c1; \ + c1 = 0; \ + VERIFY_CHECK(c2 == 0); \ +} + +static void secp256k1_scalar_reduce_512(secp256k1_scalar_t *r, const uint64_t *l) { + uint64_t n0 = l[4], n1 = l[5], n2 = l[6], n3 = l[7]; + + /* 160 bit accumulator. */ + uint64_t c0, c1; + uint32_t c2; + + /* Reduce 512 bits into 385. */ + /* m[0..6] = l[0..3] + n[0..3] * SECP256K1_N_C. */ + c0 = l[0]; c1 = 0; c2 = 0; + muladd_fast(n0, SECP256K1_N_C_0); + uint64_t m0; extract_fast(m0); + sumadd_fast(l[1]); + muladd(n1, SECP256K1_N_C_0); + muladd(n0, SECP256K1_N_C_1); + uint64_t m1; extract(m1); + sumadd(l[2]); + muladd(n2, SECP256K1_N_C_0); + muladd(n1, SECP256K1_N_C_1); + sumadd(n0); + uint64_t m2; extract(m2); + sumadd(l[3]); + muladd(n3, SECP256K1_N_C_0); + muladd(n2, SECP256K1_N_C_1); + sumadd(n1); + uint64_t m3; extract(m3); + muladd(n3, SECP256K1_N_C_1); + sumadd(n2); + uint64_t m4; extract(m4); + sumadd_fast(n3); + uint64_t m5; extract_fast(m5); + VERIFY_CHECK(c0 <= 1); + uint32_t m6 = c0; + + /* Reduce 385 bits into 258. */ + /* p[0..4] = m[0..3] + m[4..6] * SECP256K1_N_C. */ + c0 = m0; c1 = 0; c2 = 0; + muladd_fast(m4, SECP256K1_N_C_0); + uint64_t p0; extract_fast(p0); + sumadd_fast(m1); + muladd(m5, SECP256K1_N_C_0); + muladd(m4, SECP256K1_N_C_1); + uint64_t p1; extract(p1); + sumadd(m2); + muladd(m6, SECP256K1_N_C_0); + muladd(m5, SECP256K1_N_C_1); + sumadd(m4); + uint64_t p2; extract(p2); + sumadd_fast(m3); + muladd_fast(m6, SECP256K1_N_C_1); + sumadd_fast(m5); + uint64_t p3; extract_fast(p3); + uint32_t p4 = c0 + m6; + VERIFY_CHECK(p4 <= 2); + + /* Reduce 258 bits into 256. */ + /* r[0..3] = p[0..3] + p[4] * SECP256K1_N_C. */ + uint128_t c = p0 + (uint128_t)SECP256K1_N_C_0 * p4; + r->d[0] = c & 0xFFFFFFFFFFFFFFFFULL; c >>= 64; + c += p1 + (uint128_t)SECP256K1_N_C_1 * p4; + r->d[1] = c & 0xFFFFFFFFFFFFFFFFULL; c >>= 64; + c += p2 + (uint128_t)p4; + r->d[2] = c & 0xFFFFFFFFFFFFFFFFULL; c >>= 64; + c += p3; + r->d[3] = c & 0xFFFFFFFFFFFFFFFFULL; c >>= 64; + + /* Final reduction of r. */ + secp256k1_scalar_reduce(r, c + secp256k1_scalar_check_overflow(r)); +} + +static void secp256k1_scalar_mul(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { + /* 160 bit accumulator. */ + uint64_t c0 = 0, c1 = 0; + uint32_t c2 = 0; + + uint64_t l[8]; + + /* l[0..7] = a[0..3] * b[0..3]. */ + muladd_fast(a->d[0], b->d[0]); + extract_fast(l[0]); + muladd(a->d[0], b->d[1]); + muladd(a->d[1], b->d[0]); + extract(l[1]); + muladd(a->d[0], b->d[2]); + muladd(a->d[1], b->d[1]); + muladd(a->d[2], b->d[0]); + extract(l[2]); + muladd(a->d[0], b->d[3]); + muladd(a->d[1], b->d[2]); + muladd(a->d[2], b->d[1]); + muladd(a->d[3], b->d[0]); + extract(l[3]); + muladd(a->d[1], b->d[3]); + muladd(a->d[2], b->d[2]); + muladd(a->d[3], b->d[1]); + extract(l[4]); + muladd(a->d[2], b->d[3]); + muladd(a->d[3], b->d[2]); + extract(l[5]); + muladd_fast(a->d[3], b->d[3]); + extract_fast(l[6]); + VERIFY_CHECK(c1 <= 0); + l[7] = c0; + + secp256k1_scalar_reduce_512(r, l); +} + +static void secp256k1_scalar_sqr(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { + /* 160 bit accumulator. */ + uint64_t c0 = 0, c1 = 0; + uint32_t c2 = 0; + + uint64_t l[8]; + + /* l[0..7] = a[0..3] * b[0..3]. */ + muladd_fast(a->d[0], a->d[0]); + extract_fast(l[0]); + muladd2(a->d[0], a->d[1]); + extract(l[1]); + muladd2(a->d[0], a->d[2]); + muladd(a->d[1], a->d[1]); + extract(l[2]); + muladd2(a->d[0], a->d[3]); + muladd2(a->d[1], a->d[2]); + extract(l[3]); + muladd2(a->d[1], a->d[3]); + muladd(a->d[2], a->d[2]); + extract(l[4]); + muladd2(a->d[2], a->d[3]); + extract(l[5]); + muladd_fast(a->d[3], a->d[3]); + extract_fast(l[6]); + VERIFY_CHECK(c1 == 0); + l[7] = c0; + + secp256k1_scalar_reduce_512(r, l); +} + +#undef sumadd +#undef sumadd_fast +#undef muladd +#undef muladd_fast +#undef muladd2 +#undef extract +#undef extract_fast + +#endif diff --git a/src/scalar_8x32.h b/src/scalar_8x32.h new file mode 100644 index 0000000000..f70328cfc9 --- /dev/null +++ b/src/scalar_8x32.h @@ -0,0 +1,17 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_REPR_ +#define _SECP256K1_SCALAR_REPR_ + +#include + +/** A scalar modulo the group order of the secp256k1 curve. */ +typedef struct { + uint32_t d[8]; +} secp256k1_scalar_t; + +#endif diff --git a/src/scalar_8x32_impl.h b/src/scalar_8x32_impl.h new file mode 100644 index 0000000000..e58be1365f --- /dev/null +++ b/src/scalar_8x32_impl.h @@ -0,0 +1,572 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_REPR_IMPL_H_ +#define _SECP256K1_SCALAR_REPR_IMPL_H_ + +/* Limbs of the secp256k1 order. */ +#define SECP256K1_N_0 ((uint32_t)0xD0364141UL) +#define SECP256K1_N_1 ((uint32_t)0xBFD25E8CUL) +#define SECP256K1_N_2 ((uint32_t)0xAF48A03BUL) +#define SECP256K1_N_3 ((uint32_t)0xBAAEDCE6UL) +#define SECP256K1_N_4 ((uint32_t)0xFFFFFFFEUL) +#define SECP256K1_N_5 ((uint32_t)0xFFFFFFFFUL) +#define SECP256K1_N_6 ((uint32_t)0xFFFFFFFFUL) +#define SECP256K1_N_7 ((uint32_t)0xFFFFFFFFUL) + +/* Limbs of 2^256 minus the secp256k1 order. */ +#define SECP256K1_N_C_0 (~SECP256K1_N_0 + 1) +#define SECP256K1_N_C_1 (~SECP256K1_N_1) +#define SECP256K1_N_C_2 (~SECP256K1_N_2) +#define SECP256K1_N_C_3 (~SECP256K1_N_3) +#define SECP256K1_N_C_4 (1) + +/* Limbs of half the secp256k1 order. */ +#define SECP256K1_N_H_0 ((uint32_t)0x681B20A0UL) +#define SECP256K1_N_H_1 ((uint32_t)0xDFE92F46UL) +#define SECP256K1_N_H_2 ((uint32_t)0x57A4501DUL) +#define SECP256K1_N_H_3 ((uint32_t)0x5D576E73UL) +#define SECP256K1_N_H_4 ((uint32_t)0xFFFFFFFFUL) +#define SECP256K1_N_H_5 ((uint32_t)0xFFFFFFFFUL) +#define SECP256K1_N_H_6 ((uint32_t)0xFFFFFFFFUL) +#define SECP256K1_N_H_7 ((uint32_t)0x7FFFFFFFUL) + +SECP256K1_INLINE static void secp256k1_scalar_clear(secp256k1_scalar_t *r) { + r->d[0] = 0; + r->d[1] = 0; + r->d[2] = 0; + r->d[3] = 0; + r->d[4] = 0; + r->d[5] = 0; + r->d[6] = 0; + r->d[7] = 0; +} + +SECP256K1_INLINE static int secp256k1_scalar_get_bits(const secp256k1_scalar_t *a, int offset, int count) { + VERIFY_CHECK((offset + count - 1) / 32 == offset / 32); + return (a->d[offset / 32] >> (offset % 32)) & ((1 << count) - 1); +} + +SECP256K1_INLINE static int secp256k1_scalar_check_overflow(const secp256k1_scalar_t *a) { + int yes = 0; + int no = 0; + no |= (a->d[7] < SECP256K1_N_7); /* No need for a > check. */ + no |= (a->d[6] < SECP256K1_N_6); /* No need for a > check. */ + no |= (a->d[5] < SECP256K1_N_5); /* No need for a > check. */ + no |= (a->d[4] < SECP256K1_N_4); + yes |= (a->d[4] > SECP256K1_N_4) & ~no; + no |= (a->d[3] < SECP256K1_N_3) & ~yes; + yes |= (a->d[3] > SECP256K1_N_3) & ~no; + no |= (a->d[2] < SECP256K1_N_2) & ~yes; + yes |= (a->d[2] > SECP256K1_N_2) & ~no; + no |= (a->d[1] < SECP256K1_N_1) & ~yes; + yes |= (a->d[1] > SECP256K1_N_1) & ~no; + yes |= (a->d[0] >= SECP256K1_N_0) & ~no; + return yes; +} + +SECP256K1_INLINE static int secp256k1_scalar_reduce(secp256k1_scalar_t *r, uint32_t overflow) { + VERIFY_CHECK(overflow <= 1); + uint64_t t = (uint64_t)r->d[0] + overflow * SECP256K1_N_C_0; + r->d[0] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[1] + overflow * SECP256K1_N_C_1; + r->d[1] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[2] + overflow * SECP256K1_N_C_2; + r->d[2] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[3] + overflow * SECP256K1_N_C_3; + r->d[3] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[4] + overflow * SECP256K1_N_C_4; + r->d[4] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[5]; + r->d[5] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[6]; + r->d[6] = t & 0xFFFFFFFFUL; t >>= 32; + t += (uint64_t)r->d[7]; + r->d[7] = t & 0xFFFFFFFFUL; + return overflow; +} + +static void secp256k1_scalar_add(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { + uint64_t t = (uint64_t)a->d[0] + b->d[0]; + r->d[0] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[1] + b->d[1]; + r->d[1] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[2] + b->d[2]; + r->d[2] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[3] + b->d[3]; + r->d[3] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[4] + b->d[4]; + r->d[4] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[5] + b->d[5]; + r->d[5] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[6] + b->d[6]; + r->d[6] = t & 0xFFFFFFFFULL; t >>= 32; + t += (uint64_t)a->d[7] + b->d[7]; + r->d[7] = t & 0xFFFFFFFFULL; t >>= 32; + secp256k1_scalar_reduce(r, t + secp256k1_scalar_check_overflow(r)); +} + +static void secp256k1_scalar_set_b32(secp256k1_scalar_t *r, const unsigned char *b32, int *overflow) { + r->d[0] = (uint32_t)b32[31] | (uint32_t)b32[30] << 8 | (uint32_t)b32[29] << 16 | (uint32_t)b32[28] << 24; + r->d[1] = (uint32_t)b32[27] | (uint32_t)b32[26] << 8 | (uint32_t)b32[25] << 16 | (uint32_t)b32[24] << 24; + r->d[2] = (uint32_t)b32[23] | (uint32_t)b32[22] << 8 | (uint32_t)b32[21] << 16 | (uint32_t)b32[20] << 24; + r->d[3] = (uint32_t)b32[19] | (uint32_t)b32[18] << 8 | (uint32_t)b32[17] << 16 | (uint32_t)b32[16] << 24; + r->d[4] = (uint32_t)b32[15] | (uint32_t)b32[14] << 8 | (uint32_t)b32[13] << 16 | (uint32_t)b32[12] << 24; + r->d[5] = (uint32_t)b32[11] | (uint32_t)b32[10] << 8 | (uint32_t)b32[9] << 16 | (uint32_t)b32[8] << 24; + r->d[6] = (uint32_t)b32[7] | (uint32_t)b32[6] << 8 | (uint32_t)b32[5] << 16 | (uint32_t)b32[4] << 24; + r->d[7] = (uint32_t)b32[3] | (uint32_t)b32[2] << 8 | (uint32_t)b32[1] << 16 | (uint32_t)b32[0] << 24; + int over = secp256k1_scalar_reduce(r, secp256k1_scalar_check_overflow(r)); + if (overflow) { + *overflow = over; + } +} + +static void secp256k1_scalar_get_b32(unsigned char *bin, const secp256k1_scalar_t* a) { + bin[0] = a->d[7] >> 24; bin[1] = a->d[7] >> 16; bin[2] = a->d[7] >> 8; bin[3] = a->d[7]; + bin[4] = a->d[6] >> 24; bin[5] = a->d[6] >> 16; bin[6] = a->d[6] >> 8; bin[7] = a->d[6]; + bin[8] = a->d[5] >> 24; bin[9] = a->d[5] >> 16; bin[10] = a->d[5] >> 8; bin[11] = a->d[5]; + bin[12] = a->d[4] >> 24; bin[13] = a->d[4] >> 16; bin[14] = a->d[4] >> 8; bin[15] = a->d[4]; + bin[16] = a->d[3] >> 24; bin[17] = a->d[3] >> 16; bin[18] = a->d[3] >> 8; bin[19] = a->d[3]; + bin[20] = a->d[2] >> 24; bin[21] = a->d[2] >> 16; bin[22] = a->d[2] >> 8; bin[23] = a->d[2]; + bin[24] = a->d[1] >> 24; bin[25] = a->d[1] >> 16; bin[26] = a->d[1] >> 8; bin[27] = a->d[1]; + bin[28] = a->d[0] >> 24; bin[29] = a->d[0] >> 16; bin[30] = a->d[0] >> 8; bin[31] = a->d[0]; +} + +SECP256K1_INLINE static int secp256k1_scalar_is_zero(const secp256k1_scalar_t *a) { + return (a->d[0] | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0; +} + +static void secp256k1_scalar_negate(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { + uint32_t nonzero = 0xFFFFFFFFUL * (secp256k1_scalar_is_zero(a) == 0); + uint64_t t = (uint64_t)(~a->d[0]) + SECP256K1_N_0 + 1; + r->d[0] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[1]) + SECP256K1_N_1; + r->d[1] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[2]) + SECP256K1_N_2; + r->d[2] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[3]) + SECP256K1_N_3; + r->d[3] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[4]) + SECP256K1_N_4; + r->d[4] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[5]) + SECP256K1_N_5; + r->d[5] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[6]) + SECP256K1_N_6; + r->d[6] = t & nonzero; t >>= 32; + t += (uint64_t)(~a->d[7]) + SECP256K1_N_7; + r->d[7] = t & nonzero; +} + +SECP256K1_INLINE static int secp256k1_scalar_is_one(const secp256k1_scalar_t *a) { + return ((a->d[0] ^ 1) | a->d[1] | a->d[2] | a->d[3] | a->d[4] | a->d[5] | a->d[6] | a->d[7]) == 0; +} + +static int secp256k1_scalar_is_high(const secp256k1_scalar_t *a) { + int yes = 0; + int no = 0; + no |= (a->d[7] < SECP256K1_N_H_7); + yes |= (a->d[7] > SECP256K1_N_H_7) & ~no; + no |= (a->d[6] < SECP256K1_N_H_6) & ~yes; /* No need for a > check. */ + no |= (a->d[5] < SECP256K1_N_H_5) & ~yes; /* No need for a > check. */ + no |= (a->d[4] < SECP256K1_N_H_4) & ~yes; /* No need for a > check. */ + no |= (a->d[3] < SECP256K1_N_H_3) & ~yes; + yes |= (a->d[3] > SECP256K1_N_H_3) & ~no; + no |= (a->d[2] < SECP256K1_N_H_2) & ~yes; + yes |= (a->d[2] > SECP256K1_N_H_2) & ~no; + no |= (a->d[1] < SECP256K1_N_H_1) & ~yes; + yes |= (a->d[1] > SECP256K1_N_H_1) & ~no; + yes |= (a->d[0] > SECP256K1_N_H_0) & ~no; + return yes; +} + +/* Inspired by the macros in OpenSSL's crypto/bn/asm/x86_64-gcc.c. */ + +/** Add a*b to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define muladd(a,b) { \ + uint32_t tl, th; \ + { \ + uint64_t t = (uint64_t)a * b; \ + th = t >> 32; /* at most 0xFFFFFFFE */ \ + tl = t; \ + } \ + c0 += tl; /* overflow is handled on the next line */ \ + th += (c0 < tl) ? 1 : 0; /* at most 0xFFFFFFFF */ \ + c1 += th; /* overflow is handled on the next line */ \ + c2 += (c1 < th) ? 1 : 0; /* never overflows by contract (verified in the next line) */ \ + VERIFY_CHECK((c1 >= th) || (c2 != 0)); \ +} + +/** Add a*b to the number defined by (c0,c1). c1 must never overflow. */ +#define muladd_fast(a,b) { \ + uint32_t tl, th; \ + { \ + uint64_t t = (uint64_t)a * b; \ + th = t >> 32; /* at most 0xFFFFFFFE */ \ + tl = t; \ + } \ + c0 += tl; /* overflow is handled on the next line */ \ + th += (c0 < tl) ? 1 : 0; /* at most 0xFFFFFFFF */ \ + c1 += th; /* never overflows by contract (verified in the next line) */ \ + VERIFY_CHECK(c1 >= th); \ +} + +/** Add 2*a*b to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define muladd2(a,b) { \ + uint32_t tl, th; \ + { \ + uint64_t t = (uint64_t)a * b; \ + th = t >> 32; /* at most 0xFFFFFFFE */ \ + tl = t; \ + } \ + uint32_t th2 = th + th; /* at most 0xFFFFFFFE (in case th was 0x7FFFFFFF) */ \ + c2 += (th2 < th) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((th2 >= th) || (c2 != 0)); \ + uint32_t tl2 = tl + tl; /* at most 0xFFFFFFFE (in case the lowest 63 bits of tl were 0x7FFFFFFF) */ \ + th2 += (tl2 < tl) ? 1 : 0; /* at most 0xFFFFFFFF */ \ + c0 += tl2; /* overflow is handled on the next line */ \ + th2 += (c0 < tl2) ? 1 : 0; /* second overflow is handled on the next line */ \ + c2 += (c0 < tl2) & (th2 == 0); /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c0 >= tl2) || (th2 != 0) || (c2 != 0)); \ + c1 += th2; /* overflow is handled on the next line */ \ + c2 += (c1 < th2) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c1 >= th2) || (c2 != 0)); \ +} + +/** Add a to the number defined by (c0,c1,c2). c2 must never overflow. */ +#define sumadd(a) { \ + c0 += (a); /* overflow is handled on the next line */ \ + unsigned int over = (c0 < (a)) ? 1 : 0; \ + c1 += over; /* overflow is handled on the next line */ \ + c2 += (c1 < over) ? 1 : 0; /* never overflows by contract */ \ +} + +/** Add a to the number defined by (c0,c1). c1 must never overflow, c2 must be zero. */ +#define sumadd_fast(a) { \ + c0 += (a); /* overflow is handled on the next line */ \ + c1 += (c0 < (a)) ? 1 : 0; /* never overflows by contract (verified the next line) */ \ + VERIFY_CHECK((c1 != 0) | (c0 >= (a))); \ + VERIFY_CHECK(c2 == 0); \ +} + +/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. */ +#define extract(n) { \ + (n) = c0; \ + c0 = c1; \ + c1 = c2; \ + c2 = 0; \ +} + +/** Extract the lowest 32 bits of (c0,c1,c2) into n, and left shift the number 32 bits. c2 is required to be zero. */ +#define extract_fast(n) { \ + (n) = c0; \ + c0 = c1; \ + c1 = 0; \ + VERIFY_CHECK(c2 == 0); \ +} + +static void secp256k1_scalar_reduce_512(secp256k1_scalar_t *r, const uint32_t *l) { + uint32_t n0 = l[8], n1 = l[9], n2 = l[10], n3 = l[11], n4 = l[12], n5 = l[13], n6 = l[14], n7 = l[15]; + + /* 96 bit accumulator. */ + uint32_t c0, c1, c2; + + /* Reduce 512 bits into 385. */ + /* m[0..12] = l[0..7] + n[0..7] * SECP256K1_N_C. */ + c0 = l[0]; c1 = 0; c2 = 0; + muladd_fast(n0, SECP256K1_N_C_0); + uint32_t m0; extract_fast(m0); + sumadd_fast(l[1]); + muladd(n1, SECP256K1_N_C_0); + muladd(n0, SECP256K1_N_C_1); + uint32_t m1; extract(m1); + sumadd(l[2]); + muladd(n2, SECP256K1_N_C_0); + muladd(n1, SECP256K1_N_C_1); + muladd(n0, SECP256K1_N_C_2); + uint32_t m2; extract(m2); + sumadd(l[3]); + muladd(n3, SECP256K1_N_C_0); + muladd(n2, SECP256K1_N_C_1); + muladd(n1, SECP256K1_N_C_2); + muladd(n0, SECP256K1_N_C_3); + uint32_t m3; extract(m3); + sumadd(l[4]); + muladd(n4, SECP256K1_N_C_0); + muladd(n3, SECP256K1_N_C_1); + muladd(n2, SECP256K1_N_C_2); + muladd(n1, SECP256K1_N_C_3); + sumadd(n0); + uint32_t m4; extract(m4); + sumadd(l[5]); + muladd(n5, SECP256K1_N_C_0); + muladd(n4, SECP256K1_N_C_1); + muladd(n3, SECP256K1_N_C_2); + muladd(n2, SECP256K1_N_C_3); + sumadd(n1); + uint32_t m5; extract(m5); + sumadd(l[6]); + muladd(n6, SECP256K1_N_C_0); + muladd(n5, SECP256K1_N_C_1); + muladd(n4, SECP256K1_N_C_2); + muladd(n3, SECP256K1_N_C_3); + sumadd(n2); + uint32_t m6; extract(m6); + sumadd(l[7]); + muladd(n7, SECP256K1_N_C_0); + muladd(n6, SECP256K1_N_C_1); + muladd(n5, SECP256K1_N_C_2); + muladd(n4, SECP256K1_N_C_3); + sumadd(n3); + uint32_t m7; extract(m7); + muladd(n7, SECP256K1_N_C_1); + muladd(n6, SECP256K1_N_C_2); + muladd(n5, SECP256K1_N_C_3); + sumadd(n4); + uint32_t m8; extract(m8); + muladd(n7, SECP256K1_N_C_2); + muladd(n6, SECP256K1_N_C_3); + sumadd(n5); + uint32_t m9; extract(m9); + muladd(n7, SECP256K1_N_C_3); + sumadd(n6); + uint32_t m10; extract(m10); + sumadd_fast(n7); + uint32_t m11; extract_fast(m11); + VERIFY_CHECK(c0 <= 1); + uint32_t m12 = c0; + + /* Reduce 385 bits into 258. */ + /* p[0..8] = m[0..7] + m[8..12] * SECP256K1_N_C. */ + c0 = m0; c1 = 0; c2 = 0; + muladd_fast(m8, SECP256K1_N_C_0); + uint32_t p0; extract_fast(p0); + sumadd_fast(m1); + muladd(m9, SECP256K1_N_C_0); + muladd(m8, SECP256K1_N_C_1); + uint32_t p1; extract(p1); + sumadd(m2); + muladd(m10, SECP256K1_N_C_0); + muladd(m9, SECP256K1_N_C_1); + muladd(m8, SECP256K1_N_C_2); + uint32_t p2; extract(p2); + sumadd(m3); + muladd(m11, SECP256K1_N_C_0); + muladd(m10, SECP256K1_N_C_1); + muladd(m9, SECP256K1_N_C_2); + muladd(m8, SECP256K1_N_C_3); + uint32_t p3; extract(p3); + sumadd(m4); + muladd(m12, SECP256K1_N_C_0); + muladd(m11, SECP256K1_N_C_1); + muladd(m10, SECP256K1_N_C_2); + muladd(m9, SECP256K1_N_C_3); + sumadd(m8); + uint32_t p4; extract(p4); + sumadd(m5); + muladd(m12, SECP256K1_N_C_1); + muladd(m11, SECP256K1_N_C_2); + muladd(m10, SECP256K1_N_C_3); + sumadd(m9); + uint32_t p5; extract(p5); + sumadd(m6); + muladd(m12, SECP256K1_N_C_2); + muladd(m11, SECP256K1_N_C_3); + sumadd(m10); + uint32_t p6; extract(p6); + sumadd_fast(m7); + muladd_fast(m12, SECP256K1_N_C_3); + sumadd_fast(m11); + uint32_t p7; extract_fast(p7); + uint32_t p8 = c0 + m12; + VERIFY_CHECK(p8 <= 2); + + /* Reduce 258 bits into 256. */ + /* r[0..7] = p[0..7] + p[8] * SECP256K1_N_C. */ + uint64_t c = p0 + (uint64_t)SECP256K1_N_C_0 * p8; + r->d[0] = c & 0xFFFFFFFFUL; c >>= 32; + c += p1 + (uint64_t)SECP256K1_N_C_1 * p8; + r->d[1] = c & 0xFFFFFFFFUL; c >>= 32; + c += p2 + (uint64_t)SECP256K1_N_C_2 * p8; + r->d[2] = c & 0xFFFFFFFFUL; c >>= 32; + c += p3 + (uint64_t)SECP256K1_N_C_3 * p8; + r->d[3] = c & 0xFFFFFFFFUL; c >>= 32; + c += p4 + (uint64_t)p8; + r->d[4] = c & 0xFFFFFFFFUL; c >>= 32; + c += p5; + r->d[5] = c & 0xFFFFFFFFUL; c >>= 32; + c += p6; + r->d[6] = c & 0xFFFFFFFFUL; c >>= 32; + c += p7; + r->d[7] = c & 0xFFFFFFFFUL; c >>= 32; + + /* Final reduction of r. */ + secp256k1_scalar_reduce(r, c + secp256k1_scalar_check_overflow(r)); +} + +static void secp256k1_scalar_mul(secp256k1_scalar_t *r, const secp256k1_scalar_t *a, const secp256k1_scalar_t *b) { + /* 96 bit accumulator. */ + uint32_t c0 = 0, c1 = 0, c2 = 0; + + uint32_t l[16]; + + /* l[0..15] = a[0..7] * b[0..7]. */ + muladd_fast(a->d[0], b->d[0]); + extract_fast(l[0]); + muladd(a->d[0], b->d[1]); + muladd(a->d[1], b->d[0]); + extract(l[1]); + muladd(a->d[0], b->d[2]); + muladd(a->d[1], b->d[1]); + muladd(a->d[2], b->d[0]); + extract(l[2]); + muladd(a->d[0], b->d[3]); + muladd(a->d[1], b->d[2]); + muladd(a->d[2], b->d[1]); + muladd(a->d[3], b->d[0]); + extract(l[3]); + muladd(a->d[0], b->d[4]); + muladd(a->d[1], b->d[3]); + muladd(a->d[2], b->d[2]); + muladd(a->d[3], b->d[1]); + muladd(a->d[4], b->d[0]); + extract(l[4]); + muladd(a->d[0], b->d[5]); + muladd(a->d[1], b->d[4]); + muladd(a->d[2], b->d[3]); + muladd(a->d[3], b->d[2]); + muladd(a->d[4], b->d[1]); + muladd(a->d[5], b->d[0]); + extract(l[5]); + muladd(a->d[0], b->d[6]); + muladd(a->d[1], b->d[5]); + muladd(a->d[2], b->d[4]); + muladd(a->d[3], b->d[3]); + muladd(a->d[4], b->d[2]); + muladd(a->d[5], b->d[1]); + muladd(a->d[6], b->d[0]); + extract(l[6]); + muladd(a->d[0], b->d[7]); + muladd(a->d[1], b->d[6]); + muladd(a->d[2], b->d[5]); + muladd(a->d[3], b->d[4]); + muladd(a->d[4], b->d[3]); + muladd(a->d[5], b->d[2]); + muladd(a->d[6], b->d[1]); + muladd(a->d[7], b->d[0]); + extract(l[7]); + muladd(a->d[1], b->d[7]); + muladd(a->d[2], b->d[6]); + muladd(a->d[3], b->d[5]); + muladd(a->d[4], b->d[4]); + muladd(a->d[5], b->d[3]); + muladd(a->d[6], b->d[2]); + muladd(a->d[7], b->d[1]); + extract(l[8]); + muladd(a->d[2], b->d[7]); + muladd(a->d[3], b->d[6]); + muladd(a->d[4], b->d[5]); + muladd(a->d[5], b->d[4]); + muladd(a->d[6], b->d[3]); + muladd(a->d[7], b->d[2]); + extract(l[9]); + muladd(a->d[3], b->d[7]); + muladd(a->d[4], b->d[6]); + muladd(a->d[5], b->d[5]); + muladd(a->d[6], b->d[4]); + muladd(a->d[7], b->d[3]); + extract(l[10]); + muladd(a->d[4], b->d[7]); + muladd(a->d[5], b->d[6]); + muladd(a->d[6], b->d[5]); + muladd(a->d[7], b->d[4]); + extract(l[11]); + muladd(a->d[5], b->d[7]); + muladd(a->d[6], b->d[6]); + muladd(a->d[7], b->d[5]); + extract(l[12]); + muladd(a->d[6], b->d[7]); + muladd(a->d[7], b->d[6]); + extract(l[13]); + muladd_fast(a->d[7], b->d[7]); + extract_fast(l[14]); + VERIFY_CHECK(c1 == 0); + l[15] = c0; + + secp256k1_scalar_reduce_512(r, l); +} + +static void secp256k1_scalar_sqr(secp256k1_scalar_t *r, const secp256k1_scalar_t *a) { + /* 96 bit accumulator. */ + uint32_t c0 = 0, c1 = 0, c2 = 0; + + uint32_t l[16]; + + /* l[0..15] = a[0..7]^2. */ + muladd_fast(a->d[0], a->d[0]); + extract_fast(l[0]); + muladd2(a->d[0], a->d[1]); + extract(l[1]); + muladd2(a->d[0], a->d[2]); + muladd(a->d[1], a->d[1]); + extract(l[2]); + muladd2(a->d[0], a->d[3]); + muladd2(a->d[1], a->d[2]); + extract(l[3]); + muladd2(a->d[0], a->d[4]); + muladd2(a->d[1], a->d[3]); + muladd(a->d[2], a->d[2]); + extract(l[4]); + muladd2(a->d[0], a->d[5]); + muladd2(a->d[1], a->d[4]); + muladd2(a->d[2], a->d[3]); + extract(l[5]); + muladd2(a->d[0], a->d[6]); + muladd2(a->d[1], a->d[5]); + muladd2(a->d[2], a->d[4]); + muladd(a->d[3], a->d[3]); + extract(l[6]); + muladd2(a->d[0], a->d[7]); + muladd2(a->d[1], a->d[6]); + muladd2(a->d[2], a->d[5]); + muladd2(a->d[3], a->d[4]); + extract(l[7]); + muladd2(a->d[1], a->d[7]); + muladd2(a->d[2], a->d[6]); + muladd2(a->d[3], a->d[5]); + muladd(a->d[4], a->d[4]); + extract(l[8]); + muladd2(a->d[2], a->d[7]); + muladd2(a->d[3], a->d[6]); + muladd2(a->d[4], a->d[5]); + extract(l[9]); + muladd2(a->d[3], a->d[7]); + muladd2(a->d[4], a->d[6]); + muladd(a->d[5], a->d[5]); + extract(l[10]); + muladd2(a->d[4], a->d[7]); + muladd2(a->d[5], a->d[6]); + extract(l[11]); + muladd2(a->d[5], a->d[7]); + muladd(a->d[6], a->d[6]); + extract(l[12]); + muladd2(a->d[6], a->d[7]); + extract(l[13]); + muladd_fast(a->d[7], a->d[7]); + extract_fast(l[14]); + VERIFY_CHECK(c1 == 0); + l[15] = c0; + + secp256k1_scalar_reduce_512(r, l); +} + +#undef sumadd +#undef sumadd_fast +#undef muladd +#undef muladd_fast +#undef muladd2 +#undef extract +#undef extract_fast + +#endif diff --git a/src/scalar_impl.h b/src/scalar_impl.h new file mode 100644 index 0000000000..ddc5061c76 --- /dev/null +++ b/src/scalar_impl.h @@ -0,0 +1,184 @@ +/********************************************************************** + * Copyright (c) 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_SCALAR_IMPL_H_ +#define _SECP256K1_SCALAR_IMPL_H_ + +#include + +#include "scalar.h" + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#if defined(USE_SCALAR_4X64) +#include "scalar_4x64_impl.h" +#elif defined(USE_SCALAR_8X32) +#include "scalar_8x32_impl.h" +#else +#error "Please select scalar implementation" +#endif + +static void secp256k1_scalar_get_num(secp256k1_num_t *r, const secp256k1_scalar_t *a) { + unsigned char c[32]; + secp256k1_scalar_get_b32(c, a); + secp256k1_num_set_bin(r, c, 32); +} + + +static void secp256k1_scalar_inverse(secp256k1_scalar_t *r, const secp256k1_scalar_t *x) { + /* First compute x ^ (2^N - 1) for some values of N. */ + secp256k1_scalar_t x2, x3, x4, x6, x7, x8, x15, x30, x60, x120, x127; + + secp256k1_scalar_sqr(&x2, x); + secp256k1_scalar_mul(&x2, &x2, x); + + secp256k1_scalar_sqr(&x3, &x2); + secp256k1_scalar_mul(&x3, &x3, x); + + secp256k1_scalar_sqr(&x4, &x3); + secp256k1_scalar_mul(&x4, &x4, x); + + secp256k1_scalar_sqr(&x6, &x4); + secp256k1_scalar_sqr(&x6, &x6); + secp256k1_scalar_mul(&x6, &x6, &x2); + + secp256k1_scalar_sqr(&x7, &x6); + secp256k1_scalar_mul(&x7, &x7, x); + + secp256k1_scalar_sqr(&x8, &x7); + secp256k1_scalar_mul(&x8, &x8, x); + + secp256k1_scalar_sqr(&x15, &x8); + for (int i=0; i<6; i++) + secp256k1_scalar_sqr(&x15, &x15); + secp256k1_scalar_mul(&x15, &x15, &x7); + + secp256k1_scalar_sqr(&x30, &x15); + for (int i=0; i<14; i++) + secp256k1_scalar_sqr(&x30, &x30); + secp256k1_scalar_mul(&x30, &x30, &x15); + + secp256k1_scalar_sqr(&x60, &x30); + for (int i=0; i<29; i++) + secp256k1_scalar_sqr(&x60, &x60); + secp256k1_scalar_mul(&x60, &x60, &x30); + + secp256k1_scalar_sqr(&x120, &x60); + for (int i=0; i<59; i++) + secp256k1_scalar_sqr(&x120, &x120); + secp256k1_scalar_mul(&x120, &x120, &x60); + + secp256k1_scalar_sqr(&x127, &x120); + for (int i=0; i<6; i++) + secp256k1_scalar_sqr(&x127, &x127); + secp256k1_scalar_mul(&x127, &x127, &x7); + + /* Then accumulate the final result (t starts at x127). */ + secp256k1_scalar_t *t = &x127; + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<4; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<4; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<3; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<4; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<5; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<4; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<5; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x4); /* 1111 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<3; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<4; i++) /* 000 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<10; i++) /* 0000000 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<4; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x3); /* 111 */ + for (int i=0; i<9; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x8); /* 11111111 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<3; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<3; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<5; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x4); /* 1111 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<5; i++) /* 000 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<4; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<2; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<8; i++) /* 000000 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<3; i++) /* 0 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, &x2); /* 11 */ + for (int i=0; i<3; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<6; i++) /* 00000 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(t, t, x); /* 1 */ + for (int i=0; i<8; i++) /* 00 */ + secp256k1_scalar_sqr(t, t); + secp256k1_scalar_mul(r, t, &x6); /* 111111 */ +} + +#endif diff --git a/src/secp256k1.c b/src/secp256k1.c new file mode 100644 index 0000000000..1ab5b3722c --- /dev/null +++ b/src/secp256k1.c @@ -0,0 +1,305 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#define SECP256K1_BUILD (1) + +#include "include/secp256k1.h" + +#include "util.h" +#include "num_impl.h" +#include "field_impl.h" +#include "scalar_impl.h" +#include "group_impl.h" +#include "ecmult_impl.h" +#include "ecmult_gen_impl.h" +#include "ecdsa_impl.h" +#include "eckey_impl.h" + +void secp256k1_start(unsigned int flags) { + secp256k1_fe_start(); + secp256k1_ge_start(); + if (flags & SECP256K1_START_SIGN) { + secp256k1_ecmult_gen_start(); + } + if (flags & SECP256K1_START_VERIFY) { + secp256k1_ecmult_start(); + } +} + +void secp256k1_stop(void) { + secp256k1_ecmult_stop(); + secp256k1_ecmult_gen_stop(); + secp256k1_ge_stop(); + secp256k1_fe_stop(); +} + +int secp256k1_ecdsa_verify(const unsigned char *msg, int msglen, const unsigned char *sig, int siglen, const unsigned char *pubkey, int pubkeylen) { + DEBUG_CHECK(secp256k1_ecmult_consts != NULL); + DEBUG_CHECK(msg != NULL); + DEBUG_CHECK(msglen <= 32); + DEBUG_CHECK(sig != NULL); + DEBUG_CHECK(pubkey != NULL); + + int ret = -3; + secp256k1_num_t m; + secp256k1_ecdsa_sig_t s; + secp256k1_ge_t q; + secp256k1_num_set_bin(&m, msg, msglen); + + if (!secp256k1_eckey_pubkey_parse(&q, pubkey, pubkeylen)) { + ret = -1; + goto end; + } + if (!secp256k1_ecdsa_sig_parse(&s, sig, siglen)) { + ret = -2; + goto end; + } + if (!secp256k1_ecdsa_sig_verify(&s, &q, &m)) { + ret = 0; + goto end; + } + ret = 1; +end: + return ret; +} + +int secp256k1_ecdsa_sign(const unsigned char *message, int messagelen, unsigned char *signature, int *signaturelen, const unsigned char *seckey, const unsigned char *nonce) { + DEBUG_CHECK(secp256k1_ecmult_gen_consts != NULL); + DEBUG_CHECK(message != NULL); + DEBUG_CHECK(messagelen <= 32); + DEBUG_CHECK(signature != NULL); + DEBUG_CHECK(signaturelen != NULL); + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(nonce != NULL); + + secp256k1_scalar_t sec, non, msg; + secp256k1_scalar_set_b32(&sec, seckey, NULL); + int overflow = 0; + secp256k1_scalar_set_b32(&non, nonce, &overflow); + { + unsigned char c[32] = {0}; + memcpy(c + 32 - messagelen, message, messagelen); + secp256k1_scalar_set_b32(&msg, c, NULL); + memset(c, 0, 32); + } + int ret = !secp256k1_scalar_is_zero(&non) && !overflow; + secp256k1_ecdsa_sig_t sig; + if (ret) { + ret = secp256k1_ecdsa_sig_sign(&sig, &sec, &msg, &non, NULL); + } + if (ret) { + secp256k1_ecdsa_sig_serialize(signature, signaturelen, &sig); + } + secp256k1_scalar_clear(&msg); + secp256k1_scalar_clear(&non); + secp256k1_scalar_clear(&sec); + return ret; +} + +int secp256k1_ecdsa_sign_compact(const unsigned char *message, int messagelen, unsigned char *sig64, const unsigned char *seckey, const unsigned char *nonce, int *recid) { + DEBUG_CHECK(secp256k1_ecmult_gen_consts != NULL); + DEBUG_CHECK(message != NULL); + DEBUG_CHECK(messagelen <= 32); + DEBUG_CHECK(sig64 != NULL); + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(nonce != NULL); + + secp256k1_scalar_t sec, non, msg; + secp256k1_scalar_set_b32(&sec, seckey, NULL); + int overflow = 0; + secp256k1_scalar_set_b32(&non, nonce, &overflow); + { + unsigned char c[32] = {0}; + memcpy(c + 32 - messagelen, message, messagelen); + secp256k1_scalar_set_b32(&msg, c, NULL); + memset(c, 0, 32); + } + int ret = !secp256k1_scalar_is_zero(&non) && !overflow; + secp256k1_ecdsa_sig_t sig; + if (ret) { + ret = secp256k1_ecdsa_sig_sign(&sig, &sec, &msg, &non, recid); + } + if (ret) { + secp256k1_num_get_bin(sig64, 32, &sig.r); + secp256k1_num_get_bin(sig64 + 32, 32, &sig.s); + } + secp256k1_scalar_clear(&msg); + secp256k1_scalar_clear(&non); + secp256k1_scalar_clear(&sec); + return ret; +} + +int secp256k1_ecdsa_recover_compact(const unsigned char *msg, int msglen, const unsigned char *sig64, unsigned char *pubkey, int *pubkeylen, int compressed, int recid) { + DEBUG_CHECK(secp256k1_ecmult_consts != NULL); + DEBUG_CHECK(msg != NULL); + DEBUG_CHECK(msglen <= 32); + DEBUG_CHECK(sig64 != NULL); + DEBUG_CHECK(pubkey != NULL); + DEBUG_CHECK(pubkeylen != NULL); + DEBUG_CHECK(recid >= 0 && recid <= 3); + + int ret = 0; + secp256k1_num_t m; + secp256k1_ecdsa_sig_t sig; + secp256k1_num_set_bin(&sig.r, sig64, 32); + secp256k1_num_set_bin(&sig.s, sig64 + 32, 32); + secp256k1_num_set_bin(&m, msg, msglen); + + secp256k1_ge_t q; + if (secp256k1_ecdsa_sig_recover(&sig, &q, &m, recid)) { + ret = secp256k1_eckey_pubkey_serialize(&q, pubkey, pubkeylen, compressed); + } + return ret; +} + +int secp256k1_ec_seckey_verify(const unsigned char *seckey) { + DEBUG_CHECK(seckey != NULL); + + secp256k1_scalar_t sec; + int overflow; + secp256k1_scalar_set_b32(&sec, seckey, &overflow); + int ret = !secp256k1_scalar_is_zero(&sec) && !overflow; + secp256k1_scalar_clear(&sec); + return ret; +} + +int secp256k1_ec_pubkey_verify(const unsigned char *pubkey, int pubkeylen) { + DEBUG_CHECK(pubkey != NULL); + + secp256k1_ge_t q; + return secp256k1_eckey_pubkey_parse(&q, pubkey, pubkeylen); +} + +int secp256k1_ec_pubkey_create(unsigned char *pubkey, int *pubkeylen, const unsigned char *seckey, int compressed) { + DEBUG_CHECK(secp256k1_ecmult_gen_consts != NULL); + DEBUG_CHECK(pubkey != NULL); + DEBUG_CHECK(pubkeylen != NULL); + DEBUG_CHECK(seckey != NULL); + + secp256k1_scalar_t sec; + secp256k1_scalar_set_b32(&sec, seckey, NULL); + secp256k1_gej_t pj; + secp256k1_ecmult_gen(&pj, &sec); + secp256k1_scalar_clear(&sec); + secp256k1_ge_t p; + secp256k1_ge_set_gej(&p, &pj); + return secp256k1_eckey_pubkey_serialize(&p, pubkey, pubkeylen, compressed); +} + +int secp256k1_ec_pubkey_decompress(unsigned char *pubkey, int *pubkeylen) { + DEBUG_CHECK(pubkey != NULL); + DEBUG_CHECK(pubkeylen != NULL); + + secp256k1_ge_t p; + if (!secp256k1_eckey_pubkey_parse(&p, pubkey, *pubkeylen)) + return 0; + return secp256k1_eckey_pubkey_serialize(&p, pubkey, pubkeylen, 0); +} + +int secp256k1_ec_privkey_tweak_add(unsigned char *seckey, const unsigned char *tweak) { + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(tweak != NULL); + + secp256k1_scalar_t term; + int overflow = 0; + secp256k1_scalar_set_b32(&term, tweak, &overflow); + secp256k1_scalar_t sec; + secp256k1_scalar_set_b32(&sec, seckey, NULL); + + int ret = secp256k1_eckey_privkey_tweak_add(&sec, &term) && !overflow; + if (ret) { + secp256k1_scalar_get_b32(seckey, &sec); + } + + secp256k1_scalar_clear(&sec); + secp256k1_scalar_clear(&term); + return ret; +} + +int secp256k1_ec_pubkey_tweak_add(unsigned char *pubkey, int pubkeylen, const unsigned char *tweak) { + DEBUG_CHECK(secp256k1_ecmult_consts != NULL); + DEBUG_CHECK(pubkey != NULL); + DEBUG_CHECK(tweak != NULL); + + secp256k1_num_t term; + secp256k1_num_set_bin(&term, tweak, 32); + secp256k1_ge_t p; + int ret = secp256k1_eckey_pubkey_parse(&p, pubkey, pubkeylen); + if (ret) { + ret = secp256k1_eckey_pubkey_tweak_add(&p, &term); + } + if (ret) { + int oldlen = pubkeylen; + ret = secp256k1_eckey_pubkey_serialize(&p, pubkey, &pubkeylen, oldlen <= 33); + VERIFY_CHECK(pubkeylen == oldlen); + } + + return ret; +} + +int secp256k1_ec_privkey_tweak_mul(unsigned char *seckey, const unsigned char *tweak) { + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(tweak != NULL); + + secp256k1_scalar_t factor; + int overflow = 0; + secp256k1_scalar_set_b32(&factor, tweak, &overflow); + secp256k1_scalar_t sec; + secp256k1_scalar_set_b32(&sec, seckey, NULL); + int ret = secp256k1_eckey_privkey_tweak_mul(&sec, &factor) && !overflow; + if (ret) { + secp256k1_scalar_get_b32(seckey, &sec); + } + + secp256k1_scalar_clear(&sec); + secp256k1_scalar_clear(&factor); + return ret; +} + +int secp256k1_ec_pubkey_tweak_mul(unsigned char *pubkey, int pubkeylen, const unsigned char *tweak) { + DEBUG_CHECK(secp256k1_ecmult_consts != NULL); + DEBUG_CHECK(pubkey != NULL); + DEBUG_CHECK(tweak != NULL); + + secp256k1_num_t factor; + secp256k1_num_set_bin(&factor, tweak, 32); + secp256k1_ge_t p; + int ret = secp256k1_eckey_pubkey_parse(&p, pubkey, pubkeylen); + if (ret) { + ret = secp256k1_eckey_pubkey_tweak_mul(&p, &factor); + } + if (ret) { + int oldlen = pubkeylen; + ret = secp256k1_eckey_pubkey_serialize(&p, pubkey, &pubkeylen, oldlen <= 33); + VERIFY_CHECK(pubkeylen == oldlen); + } + + return ret; +} + +int secp256k1_ec_privkey_export(const unsigned char *seckey, unsigned char *privkey, int *privkeylen, int compressed) { + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(privkey != NULL); + DEBUG_CHECK(privkeylen != NULL); + + secp256k1_scalar_t key; + secp256k1_scalar_set_b32(&key, seckey, NULL); + int ret = secp256k1_eckey_privkey_serialize(privkey, privkeylen, &key, compressed); + secp256k1_scalar_clear(&key); + return ret; +} + +int secp256k1_ec_privkey_import(unsigned char *seckey, const unsigned char *privkey, int privkeylen) { + DEBUG_CHECK(seckey != NULL); + DEBUG_CHECK(privkey != NULL); + + secp256k1_scalar_t key; + int ret = secp256k1_eckey_privkey_parse(&key, privkey, privkeylen); + if (ret) + secp256k1_scalar_get_b32(seckey, &key); + secp256k1_scalar_clear(&key); + return ret; +} diff --git a/src/testrand.h b/src/testrand.h new file mode 100644 index 0000000000..018b65cd53 --- /dev/null +++ b/src/testrand.h @@ -0,0 +1,26 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_TESTRAND_H_ +#define _SECP256K1_TESTRAND_H_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +/** Seed the pseudorandom number generator. */ +SECP256K1_INLINE static void secp256k1_rand_seed(uint64_t v); + +/** Generate a pseudorandom 32-bit number. */ +static uint32_t secp256k1_rand32(void); + +/** Generate a pseudorandom 32-byte array. */ +static void secp256k1_rand256(unsigned char *b32); + +/** Generate a pseudorandom 32-byte array with long sequences of zero and one bits. */ +static void secp256k1_rand256_test(unsigned char *b32); + +#endif diff --git a/src/testrand_impl.h b/src/testrand_impl.h new file mode 100644 index 0000000000..677c4b9a0e --- /dev/null +++ b/src/testrand_impl.h @@ -0,0 +1,60 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_TESTRAND_IMPL_H_ +#define _SECP256K1_TESTRAND_IMPL_H_ + +#include +#include + +#include "testrand.h" + +static uint32_t secp256k1_Rz = 11, secp256k1_Rw = 11; + +SECP256K1_INLINE static void secp256k1_rand_seed(uint64_t v) { + secp256k1_Rz = v >> 32; + secp256k1_Rw = v; + + if (secp256k1_Rz == 0 || secp256k1_Rz == 0x9068ffffU) { + secp256k1_Rz = 111; + } + if (secp256k1_Rw == 0 || secp256k1_Rw == 0x464fffffU) { + secp256k1_Rw = 111; + } +} + +SECP256K1_INLINE static uint32_t secp256k1_rand32(void) { + secp256k1_Rz = 36969 * (secp256k1_Rz & 0xFFFF) + (secp256k1_Rz >> 16); + secp256k1_Rw = 18000 * (secp256k1_Rw & 0xFFFF) + (secp256k1_Rw >> 16); + return (secp256k1_Rw << 16) + (secp256k1_Rw >> 16) + secp256k1_Rz; +} + +static void secp256k1_rand256(unsigned char *b32) { + for (int i=0; i<8; i++) { + uint32_t r = secp256k1_rand32(); + b32[i*4 + 0] = (r >> 0) & 0xFF; + b32[i*4 + 1] = (r >> 8) & 0xFF; + b32[i*4 + 2] = (r >> 16) & 0xFF; + b32[i*4 + 3] = (r >> 24) & 0xFF; + } +} + +static void secp256k1_rand256_test(unsigned char *b32) { + int bits=0; + memset(b32, 0, 32); + while (bits < 256) { + uint32_t ent = secp256k1_rand32(); + int now = 1 + ((ent % 64)*((ent >> 6) % 32)+16)/31; + uint32_t val = 1 & (ent >> 11); + while (now > 0 && bits < 256) { + b32[bits / 8] |= val << (bits % 8); + now--; + bits++; + } + } +} + +#endif diff --git a/src/tests.c b/src/tests.c new file mode 100644 index 0000000000..5d9b8344d9 --- /dev/null +++ b/src/tests.c @@ -0,0 +1,1080 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#include +#include + +#include "secp256k1.c" +#include "testrand_impl.h" + +#ifdef ENABLE_OPENSSL_TESTS +#include "openssl/bn.h" +#include "openssl/ec.h" +#include "openssl/ecdsa.h" +#include "openssl/obj_mac.h" +#endif + +static int count = 64; + +/***** NUM TESTS *****/ + +void random_num_negate(secp256k1_num_t *num) { + if (secp256k1_rand32() & 1) + secp256k1_num_negate(num); +} + +void random_field_element_test(secp256k1_fe_t *fe) { + do { + unsigned char b32[32]; + secp256k1_rand256_test(b32); + secp256k1_num_t num; + secp256k1_num_set_bin(&num, b32, 32); + if (secp256k1_num_cmp(&num, &secp256k1_fe_consts->p) >= 0) + continue; + secp256k1_fe_set_b32(fe, b32); + break; + } while(1); +} + +void random_field_element_magnitude(secp256k1_fe_t *fe) { + secp256k1_fe_normalize(fe); + int n = secp256k1_rand32() % 4; + for (int i = 0; i < n; i++) { + secp256k1_fe_negate(fe, fe, 1 + 2*i); + secp256k1_fe_negate(fe, fe, 2 + 2*i); + } +} + +void random_group_element_test(secp256k1_ge_t *ge) { + secp256k1_fe_t fe; + do { + random_field_element_test(&fe); + if (secp256k1_ge_set_xo(ge, &fe, secp256k1_rand32() & 1)) + break; + } while(1); +} + +void random_group_element_jacobian_test(secp256k1_gej_t *gej, const secp256k1_ge_t *ge) { + do { + random_field_element_test(&gej->z); + if (!secp256k1_fe_is_zero(&gej->z)) { + break; + } + } while(1); + secp256k1_fe_t z2; secp256k1_fe_sqr(&z2, &gej->z); + secp256k1_fe_t z3; secp256k1_fe_mul(&z3, &z2, &gej->z); + secp256k1_fe_mul(&gej->x, &ge->x, &z2); + secp256k1_fe_mul(&gej->y, &ge->y, &z3); + gej->infinity = ge->infinity; +} + +void random_num_order_test(secp256k1_num_t *num) { + do { + unsigned char b32[32]; + secp256k1_rand256_test(b32); + secp256k1_num_set_bin(num, b32, 32); + if (secp256k1_num_is_zero(num)) + continue; + if (secp256k1_num_cmp(num, &secp256k1_ge_consts->order) >= 0) + continue; + break; + } while(1); +} + +void random_scalar_order_test(secp256k1_scalar_t *num) { + do { + unsigned char b32[32]; + secp256k1_rand256_test(b32); + int overflow = 0; + secp256k1_scalar_set_b32(num, b32, &overflow); + if (overflow || secp256k1_scalar_is_zero(num)) + continue; + break; + } while(1); +} + +void random_num_order(secp256k1_num_t *num) { + do { + unsigned char b32[32]; + secp256k1_rand256(b32); + secp256k1_num_set_bin(num, b32, 32); + if (secp256k1_num_is_zero(num)) + continue; + if (secp256k1_num_cmp(num, &secp256k1_ge_consts->order) >= 0) + continue; + break; + } while(1); +} + +void test_num_copy_inc_cmp(void) { + secp256k1_num_t n1,n2; + random_num_order(&n1); + secp256k1_num_copy(&n2, &n1); + CHECK(secp256k1_num_eq(&n1, &n2)); + CHECK(secp256k1_num_eq(&n2, &n1)); + secp256k1_num_inc(&n2); + CHECK(!secp256k1_num_eq(&n1, &n2)); + CHECK(!secp256k1_num_eq(&n2, &n1)); +} + + +void test_num_get_set_hex(void) { + secp256k1_num_t n1,n2; + random_num_order_test(&n1); + char c[64]; + secp256k1_num_get_hex(c, 64, &n1); + secp256k1_num_set_hex(&n2, c, 64); + CHECK(secp256k1_num_eq(&n1, &n2)); + for (int i=0; i<64; i++) { + /* check whether the lower 4 bits correspond to the last hex character */ + int low1 = secp256k1_num_shift(&n1, 4); + int lowh = c[63]; + int low2 = ((lowh>>6)*9+(lowh-'0'))&15; + CHECK(low1 == low2); + /* shift bits off the hex representation, and compare */ + memmove(c+1, c, 63); + c[0] = '0'; + secp256k1_num_set_hex(&n2, c, 64); + CHECK(secp256k1_num_eq(&n1, &n2)); + } +} + +void test_num_get_set_bin(void) { + secp256k1_num_t n1,n2; + random_num_order_test(&n1); + unsigned char c[32]; + secp256k1_num_get_bin(c, 32, &n1); + secp256k1_num_set_bin(&n2, c, 32); + CHECK(secp256k1_num_eq(&n1, &n2)); + for (int i=0; i<32; i++) { + /* check whether the lower 8 bits correspond to the last byte */ + int low1 = secp256k1_num_shift(&n1, 8); + int low2 = c[31]; + CHECK(low1 == low2); + /* shift bits off the byte representation, and compare */ + memmove(c+1, c, 31); + c[0] = 0; + secp256k1_num_set_bin(&n2, c, 32); + CHECK(secp256k1_num_eq(&n1, &n2)); + } +} + +void run_num_int(void) { + secp256k1_num_t n1; + for (int i=-255; i<256; i++) { + unsigned char c1[3] = {}; + c1[2] = abs(i); + unsigned char c2[3] = {0x11,0x22,0x33}; + secp256k1_num_set_int(&n1, i); + secp256k1_num_get_bin(c2, 3, &n1); + CHECK(memcmp(c1, c2, 3) == 0); + } +} + +void test_num_negate(void) { + secp256k1_num_t n1; + secp256k1_num_t n2; + random_num_order_test(&n1); /* n1 = R */ + random_num_negate(&n1); + secp256k1_num_copy(&n2, &n1); /* n2 = R */ + secp256k1_num_sub(&n1, &n2, &n1); /* n1 = n2-n1 = 0 */ + CHECK(secp256k1_num_is_zero(&n1)); + secp256k1_num_copy(&n1, &n2); /* n1 = R */ + secp256k1_num_negate(&n1); /* n1 = -R */ + CHECK(!secp256k1_num_is_zero(&n1)); + secp256k1_num_add(&n1, &n2, &n1); /* n1 = n2+n1 = 0 */ + CHECK(secp256k1_num_is_zero(&n1)); + secp256k1_num_copy(&n1, &n2); /* n1 = R */ + secp256k1_num_negate(&n1); /* n1 = -R */ + CHECK(secp256k1_num_is_neg(&n1) != secp256k1_num_is_neg(&n2)); + secp256k1_num_negate(&n1); /* n1 = R */ + CHECK(secp256k1_num_eq(&n1, &n2)); +} + +void test_num_add_sub(void) { + int r = secp256k1_rand32(); + secp256k1_num_t n1; + secp256k1_num_t n2; + random_num_order_test(&n1); /* n1 = R1 */ + if (r & 1) { + random_num_negate(&n1); + } + random_num_order_test(&n2); /* n2 = R2 */ + if (r & 2) { + random_num_negate(&n2); + } + secp256k1_num_t n1p2, n2p1, n1m2, n2m1; + secp256k1_num_add(&n1p2, &n1, &n2); /* n1p2 = R1 + R2 */ + secp256k1_num_add(&n2p1, &n2, &n1); /* n2p1 = R2 + R1 */ + secp256k1_num_sub(&n1m2, &n1, &n2); /* n1m2 = R1 - R2 */ + secp256k1_num_sub(&n2m1, &n2, &n1); /* n2m1 = R2 - R1 */ + CHECK(secp256k1_num_eq(&n1p2, &n2p1)); + CHECK(!secp256k1_num_eq(&n1p2, &n1m2)); + secp256k1_num_negate(&n2m1); /* n2m1 = -R2 + R1 */ + CHECK(secp256k1_num_eq(&n2m1, &n1m2)); + CHECK(!secp256k1_num_eq(&n2m1, &n1)); + secp256k1_num_add(&n2m1, &n2m1, &n2); /* n2m1 = -R2 + R1 + R2 = R1 */ + CHECK(secp256k1_num_eq(&n2m1, &n1)); + CHECK(!secp256k1_num_eq(&n2p1, &n1)); + secp256k1_num_sub(&n2p1, &n2p1, &n2); /* n2p1 = R2 + R1 - R2 = R1 */ + CHECK(secp256k1_num_eq(&n2p1, &n1)); +} + +void run_num_smalltests(void) { + for (int i=0; i<100*count; i++) { + test_num_copy_inc_cmp(); + test_num_get_set_hex(); + test_num_get_set_bin(); + test_num_negate(); + test_num_add_sub(); + } + run_num_int(); +} + +/***** SCALAR TESTS *****/ + +int secp256k1_scalar_eq(const secp256k1_scalar_t *s1, const secp256k1_scalar_t *s2) { + secp256k1_scalar_t t; + secp256k1_scalar_negate(&t, s2); + secp256k1_scalar_add(&t, &t, s1); + int ret = secp256k1_scalar_is_zero(&t); + return ret; +} + +void scalar_test(void) { + unsigned char c[32]; + + /* Set 's' to a random scalar, with value 'snum'. */ + secp256k1_rand256_test(c); + secp256k1_scalar_t s; + secp256k1_scalar_set_b32(&s, c, NULL); + secp256k1_num_t snum; + secp256k1_num_set_bin(&snum, c, 32); + secp256k1_num_mod(&snum, &secp256k1_ge_consts->order); + + /* Set 's1' to a random scalar, with value 's1num'. */ + secp256k1_rand256_test(c); + secp256k1_scalar_t s1; + secp256k1_scalar_set_b32(&s1, c, NULL); + secp256k1_num_t s1num; + secp256k1_num_set_bin(&s1num, c, 32); + secp256k1_num_mod(&s1num, &secp256k1_ge_consts->order); + + /* Set 's2' to a random scalar, with value 'snum2', and byte array representation 'c'. */ + secp256k1_rand256_test(c); + secp256k1_scalar_t s2; + int overflow = 0; + secp256k1_scalar_set_b32(&s2, c, &overflow); + secp256k1_num_t s2num; + secp256k1_num_set_bin(&s2num, c, 32); + secp256k1_num_mod(&s2num, &secp256k1_ge_consts->order); + + { + /* Test that fetching groups of 4 bits from a scalar and recursing n(i)=16*n(i-1)+p(i) reconstructs it. */ + secp256k1_num_t n, t, m; + secp256k1_num_set_int(&n, 0); + secp256k1_num_set_int(&m, 16); + for (int i = 0; i < 256; i += 4) { + secp256k1_num_set_int(&t, secp256k1_scalar_get_bits(&s, 256 - 4 - i, 4)); + secp256k1_num_mul(&n, &n, &m); + secp256k1_num_add(&n, &n, &t); + } + CHECK(secp256k1_num_eq(&n, &snum)); + } + + { + /* Test that get_b32 returns the same as get_bin on the number. */ + unsigned char r1[32]; + secp256k1_scalar_get_b32(r1, &s2); + unsigned char r2[32]; + secp256k1_num_get_bin(r2, 32, &s2num); + CHECK(memcmp(r1, r2, 32) == 0); + /* If no overflow occurred when assigning, it should also be equal to the original byte array. */ + CHECK((memcmp(r1, c, 32) == 0) == (overflow == 0)); + } + + { + /* Test that adding the scalars together is equal to adding their numbers together modulo the order. */ + secp256k1_num_t rnum; + secp256k1_num_add(&rnum, &snum, &s2num); + secp256k1_num_mod(&rnum, &secp256k1_ge_consts->order); + secp256k1_scalar_t r; + secp256k1_scalar_add(&r, &s, &s2); + secp256k1_num_t r2num; + secp256k1_scalar_get_num(&r2num, &r); + CHECK(secp256k1_num_eq(&rnum, &r2num)); + } + + { + /* Test that multipying the scalars is equal to multiplying their numbers modulo the order. */ + secp256k1_num_t rnum; + secp256k1_num_mul(&rnum, &snum, &s2num); + secp256k1_num_mod(&rnum, &secp256k1_ge_consts->order); + secp256k1_scalar_t r; + secp256k1_scalar_mul(&r, &s, &s2); + secp256k1_num_t r2num; + secp256k1_scalar_get_num(&r2num, &r); + CHECK(secp256k1_num_eq(&rnum, &r2num)); + /* The result can only be zero if at least one of the factors was zero. */ + CHECK(secp256k1_scalar_is_zero(&r) == (secp256k1_scalar_is_zero(&s) || secp256k1_scalar_is_zero(&s2))); + /* The results can only be equal to one of the factors if that factor was zero, or the other factor was one. */ + CHECK(secp256k1_num_eq(&rnum, &snum) == (secp256k1_scalar_is_zero(&s) || secp256k1_scalar_is_one(&s2))); + CHECK(secp256k1_num_eq(&rnum, &s2num) == (secp256k1_scalar_is_zero(&s2) || secp256k1_scalar_is_one(&s))); + } + + { + /* Check that comparison with zero matches comparison with zero on the number. */ + CHECK(secp256k1_num_is_zero(&snum) == secp256k1_scalar_is_zero(&s)); + /* Check that comparison with the half order is equal to testing for high scalar. */ + CHECK(secp256k1_scalar_is_high(&s) == (secp256k1_num_cmp(&snum, &secp256k1_ge_consts->half_order) > 0)); + secp256k1_scalar_t neg; + secp256k1_scalar_negate(&neg, &s); + secp256k1_num_t negnum; + secp256k1_num_sub(&negnum, &secp256k1_ge_consts->order, &snum); + secp256k1_num_mod(&negnum, &secp256k1_ge_consts->order); + /* Check that comparison with the half order is equal to testing for high scalar after negation. */ + CHECK(secp256k1_scalar_is_high(&neg) == (secp256k1_num_cmp(&negnum, &secp256k1_ge_consts->half_order) > 0)); + /* Negating should change the high property, unless the value was already zero. */ + CHECK((secp256k1_scalar_is_high(&s) == secp256k1_scalar_is_high(&neg)) == secp256k1_scalar_is_zero(&s)); + secp256k1_num_t negnum2; + secp256k1_scalar_get_num(&negnum2, &neg); + /* Negating a scalar should be equal to (order - n) mod order on the number. */ + CHECK(secp256k1_num_eq(&negnum, &negnum2)); + secp256k1_scalar_add(&neg, &neg, &s); + /* Adding a number to its negation should result in zero. */ + CHECK(secp256k1_scalar_is_zero(&neg)); + secp256k1_scalar_negate(&neg, &neg); + /* Negating zero should still result in zero. */ + CHECK(secp256k1_scalar_is_zero(&neg)); + } + + { + /* Test that scalar inverses are equal to the inverse of their number modulo the order. */ + if (!secp256k1_scalar_is_zero(&s)) { + secp256k1_scalar_t inv; + secp256k1_scalar_inverse(&inv, &s); + secp256k1_num_t invnum; + secp256k1_num_mod_inverse(&invnum, &snum, &secp256k1_ge_consts->order); + secp256k1_num_t invnum2; + secp256k1_scalar_get_num(&invnum2, &inv); + CHECK(secp256k1_num_eq(&invnum, &invnum2)); + secp256k1_scalar_mul(&inv, &inv, &s); + /* Multiplying a scalar with its inverse must result in one. */ + CHECK(secp256k1_scalar_is_one(&inv)); + secp256k1_scalar_inverse(&inv, &inv); + /* Inverting one must result in one. */ + CHECK(secp256k1_scalar_is_one(&inv)); + } + } + + { + /* Test commutativity of add. */ + secp256k1_scalar_t r1, r2; + secp256k1_scalar_add(&r1, &s1, &s2); + secp256k1_scalar_add(&r2, &s2, &s1); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } + + { + /* Test commutativity of mul. */ + secp256k1_scalar_t r1, r2; + secp256k1_scalar_mul(&r1, &s1, &s2); + secp256k1_scalar_mul(&r2, &s2, &s1); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } + + { + /* Test associativity of add. */ + secp256k1_scalar_t r1, r2; + secp256k1_scalar_add(&r1, &s1, &s2); + secp256k1_scalar_add(&r1, &r1, &s); + secp256k1_scalar_add(&r2, &s2, &s); + secp256k1_scalar_add(&r2, &s1, &r2); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } + + { + /* Test associativity of mul. */ + secp256k1_scalar_t r1, r2; + secp256k1_scalar_mul(&r1, &s1, &s2); + secp256k1_scalar_mul(&r1, &r1, &s); + secp256k1_scalar_mul(&r2, &s2, &s); + secp256k1_scalar_mul(&r2, &s1, &r2); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } + + { + /* Test distributitivity of mul over add. */ + secp256k1_scalar_t r1, r2, t; + secp256k1_scalar_add(&r1, &s1, &s2); + secp256k1_scalar_mul(&r1, &r1, &s); + secp256k1_scalar_mul(&r2, &s1, &s); + secp256k1_scalar_mul(&t, &s2, &s); + secp256k1_scalar_add(&r2, &r2, &t); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } + + { + /* Test square. */ + secp256k1_scalar_t r1, r2; + secp256k1_scalar_sqr(&r1, &s1); + secp256k1_scalar_mul(&r2, &s1, &s1); + CHECK(secp256k1_scalar_eq(&r1, &r2)); + } +} + +void run_scalar_tests(void) { + for (int i = 0; i < 128 * count; i++) { + scalar_test(); + } +} + +/***** FIELD TESTS *****/ + +void random_fe(secp256k1_fe_t *x) { + unsigned char bin[32]; + secp256k1_rand256(bin); + secp256k1_fe_set_b32(x, bin); +} + +void random_fe_non_zero(secp256k1_fe_t *nz) { + int tries = 10; + while (--tries >= 0) { + random_fe(nz); + secp256k1_fe_normalize(nz); + if (!secp256k1_fe_is_zero(nz)) + break; + } + /* Infinitesimal probability of spurious failure here */ + CHECK(tries >= 0); +} + +void random_fe_non_square(secp256k1_fe_t *ns) { + random_fe_non_zero(ns); + secp256k1_fe_t r; + if (secp256k1_fe_sqrt(&r, ns)) { + secp256k1_fe_negate(ns, ns, 1); + } +} + +int check_fe_equal(const secp256k1_fe_t *a, const secp256k1_fe_t *b) { + secp256k1_fe_t an = *a; secp256k1_fe_normalize(&an); + secp256k1_fe_t bn = *b; secp256k1_fe_normalize(&bn); + return secp256k1_fe_equal(&an, &bn); +} + +int check_fe_inverse(const secp256k1_fe_t *a, const secp256k1_fe_t *ai) { + secp256k1_fe_t x; secp256k1_fe_mul(&x, a, ai); + secp256k1_fe_t one; secp256k1_fe_set_int(&one, 1); + return check_fe_equal(&x, &one); +} + +void run_field_inv(void) { + secp256k1_fe_t x, xi, xii; + for (int i=0; i<10*count; i++) { + random_fe_non_zero(&x); + secp256k1_fe_inv(&xi, &x); + CHECK(check_fe_inverse(&x, &xi)); + secp256k1_fe_inv(&xii, &xi); + CHECK(check_fe_equal(&x, &xii)); + } +} + +void run_field_inv_var(void) { + secp256k1_fe_t x, xi, xii; + for (int i=0; i<10*count; i++) { + random_fe_non_zero(&x); + secp256k1_fe_inv_var(&xi, &x); + CHECK(check_fe_inverse(&x, &xi)); + secp256k1_fe_inv_var(&xii, &xi); + CHECK(check_fe_equal(&x, &xii)); + } +} + +void run_field_inv_all(void) { + secp256k1_fe_t x[16], xi[16], xii[16]; + /* Check it's safe to call for 0 elements */ + secp256k1_fe_inv_all(0, xi, x); + for (int i=0; iinfinity && b->infinity) + return 1; + return check_fe_equal(&a->x, &b->x) && check_fe_equal(&a->y, &b->y); +} + +void ge_equals_gej(const secp256k1_ge_t *a, const secp256k1_gej_t *b) { + secp256k1_ge_t bb; + secp256k1_gej_t bj = *b; + secp256k1_ge_set_gej_var(&bb, &bj); + CHECK(ge_equals_ge(a, &bb)); +} + +void gej_equals_gej(const secp256k1_gej_t *a, const secp256k1_gej_t *b) { + secp256k1_ge_t aa, bb; + secp256k1_gej_t aj = *a, bj = *b; + secp256k1_ge_set_gej_var(&aa, &aj); + secp256k1_ge_set_gej_var(&bb, &bj); + CHECK(ge_equals_ge(&aa, &bb)); +} + +void test_ge(void) { + secp256k1_ge_t a, b, i, n; + random_group_element_test(&a); + random_group_element_test(&b); + n = a; + secp256k1_fe_normalize(&a.y); + secp256k1_fe_negate(&n.y, &a.y, 1); + secp256k1_ge_set_infinity(&i); + random_field_element_magnitude(&a.x); + random_field_element_magnitude(&a.y); + random_field_element_magnitude(&b.x); + random_field_element_magnitude(&b.y); + random_field_element_magnitude(&n.x); + random_field_element_magnitude(&n.y); + + secp256k1_gej_t aj, bj, ij, nj; + random_group_element_jacobian_test(&aj, &a); + random_group_element_jacobian_test(&bj, &b); + secp256k1_gej_set_infinity(&ij); + random_group_element_jacobian_test(&nj, &n); + random_field_element_magnitude(&aj.x); + random_field_element_magnitude(&aj.y); + random_field_element_magnitude(&aj.z); + random_field_element_magnitude(&bj.x); + random_field_element_magnitude(&bj.y); + random_field_element_magnitude(&bj.z); + random_field_element_magnitude(&nj.x); + random_field_element_magnitude(&nj.y); + random_field_element_magnitude(&nj.z); + + /* gej + gej adds */ + secp256k1_gej_t aaj; secp256k1_gej_add_var(&aaj, &aj, &aj); + secp256k1_gej_t abj; secp256k1_gej_add_var(&abj, &aj, &bj); + secp256k1_gej_t aij; secp256k1_gej_add_var(&aij, &aj, &ij); + secp256k1_gej_t anj; secp256k1_gej_add_var(&anj, &aj, &nj); + secp256k1_gej_t iaj; secp256k1_gej_add_var(&iaj, &ij, &aj); + secp256k1_gej_t iij; secp256k1_gej_add_var(&iij, &ij, &ij); + + /* gej + ge adds */ + secp256k1_gej_t aa; secp256k1_gej_add_ge_var(&aa, &aj, &a); + secp256k1_gej_t ab; secp256k1_gej_add_ge_var(&ab, &aj, &b); + secp256k1_gej_t ai; secp256k1_gej_add_ge_var(&ai, &aj, &i); + secp256k1_gej_t an; secp256k1_gej_add_ge_var(&an, &aj, &n); + secp256k1_gej_t ia; secp256k1_gej_add_ge_var(&ia, &ij, &a); + secp256k1_gej_t ii; secp256k1_gej_add_ge_var(&ii, &ij, &i); + + /* const gej + ge adds */ + secp256k1_gej_t aac; secp256k1_gej_add_ge(&aac, &aj, &a); + secp256k1_gej_t abc; secp256k1_gej_add_ge(&abc, &aj, &b); + secp256k1_gej_t anc; secp256k1_gej_add_ge(&anc, &aj, &n); + secp256k1_gej_t iac; secp256k1_gej_add_ge(&iac, &ij, &a); + + CHECK(secp256k1_gej_is_infinity(&an)); + CHECK(secp256k1_gej_is_infinity(&anj)); + CHECK(secp256k1_gej_is_infinity(&anc)); + gej_equals_gej(&aa, &aaj); + gej_equals_gej(&aa, &aac); + gej_equals_gej(&ab, &abj); + gej_equals_gej(&ab, &abc); + gej_equals_gej(&an, &anj); + gej_equals_gej(&an, &anc); + gej_equals_gej(&ia, &iaj); + gej_equals_gej(&ai, &aij); + gej_equals_gej(&ii, &iij); + ge_equals_gej(&a, &ai); + ge_equals_gej(&a, &ai); + ge_equals_gej(&a, &iaj); + ge_equals_gej(&a, &iaj); + ge_equals_gej(&a, &iac); +} + +void run_ge(void) { + for (int i = 0; i < 2000*count; i++) { + test_ge(); + } +} + +/***** ECMULT TESTS *****/ + +void run_ecmult_chain(void) { + /* random starting point A (on the curve) */ + secp256k1_fe_t ax; secp256k1_fe_set_hex(&ax, "8b30bbe9ae2a990696b22f670709dff3727fd8bc04d3362c6c7bf458e2846004", 64); + secp256k1_fe_t ay; secp256k1_fe_set_hex(&ay, "a357ae915c4a65281309edf20504740f0eb3343990216b4f81063cb65f2f7e0f", 64); + secp256k1_gej_t a; secp256k1_gej_set_xy(&a, &ax, &ay); + /* two random initial factors xn and gn */ + secp256k1_num_t xn; + secp256k1_num_set_hex(&xn, "84cc5452f7fde1edb4d38a8ce9b1b84ccef31f146e569be9705d357a42985407", 64); + secp256k1_num_t gn; + secp256k1_num_set_hex(&gn, "a1e58d22553dcd42b23980625d4c57a96e9323d42b3152e5ca2c3990edc7c9de", 64); + /* two small multipliers to be applied to xn and gn in every iteration: */ + secp256k1_num_t xf; + secp256k1_num_set_hex(&xf, "1337", 4); + secp256k1_num_t gf; + secp256k1_num_set_hex(&gf, "7113", 4); + /* accumulators with the resulting coefficients to A and G */ + secp256k1_num_t ae; + secp256k1_num_set_int(&ae, 1); + secp256k1_num_t ge; + secp256k1_num_set_int(&ge, 0); + /* the point being computed */ + secp256k1_gej_t x = a; + const secp256k1_num_t *order = &secp256k1_ge_consts->order; + for (int i=0; i<200*count; i++) { + /* in each iteration, compute X = xn*X + gn*G; */ + secp256k1_ecmult(&x, &x, &xn, &gn); + /* also compute ae and ge: the actual accumulated factors for A and G */ + /* if X was (ae*A+ge*G), xn*X + gn*G results in (xn*ae*A + (xn*ge+gn)*G) */ + secp256k1_num_mod_mul(&ae, &ae, &xn, order); + secp256k1_num_mod_mul(&ge, &ge, &xn, order); + secp256k1_num_add(&ge, &ge, &gn); + secp256k1_num_mod(&ge, order); + /* modify xn and gn */ + secp256k1_num_mod_mul(&xn, &xn, &xf, order); + secp256k1_num_mod_mul(&gn, &gn, &gf, order); + + /* verify */ + if (i == 19999) { + char res[132]; int resl = 132; + secp256k1_gej_get_hex(res, &resl, &x); + CHECK(strcmp(res, "(D6E96687F9B10D092A6F35439D86CEBEA4535D0D409F53586440BD74B933E830,B95CBCA2C77DA786539BE8FD53354D2D3B4F566AE658045407ED6015EE1B2A88)") == 0); + } + } + /* redo the computation, but directly with the resulting ae and ge coefficients: */ + secp256k1_gej_t x2; secp256k1_ecmult(&x2, &a, &ae, &ge); + char res[132]; int resl = 132; + char res2[132]; int resl2 = 132; + secp256k1_gej_get_hex(res, &resl, &x); + secp256k1_gej_get_hex(res2, &resl2, &x2); + CHECK(strcmp(res, res2) == 0); + CHECK(strlen(res) == 131); +} + +void test_point_times_order(const secp256k1_gej_t *point) { + /* multiplying a point by the order results in O */ + const secp256k1_num_t *order = &secp256k1_ge_consts->order; + secp256k1_num_t zero; + secp256k1_num_set_int(&zero, 0); + secp256k1_gej_t res; + secp256k1_ecmult(&res, point, order, order); /* calc res = order * point + order * G; */ + CHECK(secp256k1_gej_is_infinity(&res)); +} + +void run_point_times_order(void) { + secp256k1_fe_t x; secp256k1_fe_set_hex(&x, "02", 2); + for (int i=0; i<500; i++) { + secp256k1_ge_t p; + if (secp256k1_ge_set_xo(&p, &x, 1)) { + CHECK(secp256k1_ge_is_valid(&p)); + secp256k1_gej_t j; + secp256k1_gej_set_ge(&j, &p); + CHECK(secp256k1_gej_is_valid(&j)); + test_point_times_order(&j); + } + secp256k1_fe_sqr(&x, &x); + } + char c[65]; int cl=65; + secp256k1_fe_get_hex(c, &cl, &x); + CHECK(strcmp(c, "7603CB59B0EF6C63FE6084792A0C378CDB3233A80F8A9A09A877DEAD31B38C45") == 0); +} + +void test_wnaf(const secp256k1_num_t *number, int w) { + secp256k1_num_t x, two, t; + secp256k1_num_set_int(&x, 0); + secp256k1_num_set_int(&two, 2); + int wnaf[257]; + int bits = secp256k1_ecmult_wnaf(wnaf, number, w); + int zeroes = -1; + for (int i=bits-1; i>=0; i--) { + secp256k1_num_mul(&x, &x, &two); + int v = wnaf[i]; + if (v) { + CHECK(zeroes == -1 || zeroes >= w-1); /* check that distance between non-zero elements is at least w-1 */ + zeroes=0; + CHECK((v & 1) == 1); /* check non-zero elements are odd */ + CHECK(v <= (1 << (w-1)) - 1); /* check range below */ + CHECK(v >= -(1 << (w-1)) - 1); /* check range above */ + } else { + CHECK(zeroes != -1); /* check that no unnecessary zero padding exists */ + zeroes++; + } + secp256k1_num_set_int(&t, v); + secp256k1_num_add(&x, &x, &t); + } + CHECK(secp256k1_num_eq(&x, number)); /* check that wnaf represents number */ +} + +void run_wnaf(void) { + secp256k1_num_t n; + for (int i=0; i 1) { + count = strtol(argv[1], NULL, 0); + } + + /* find random seed */ + uint64_t seed; + if (argc > 2) { + seed = strtoull(argv[2], NULL, 0); + } else { + FILE *frand = fopen("/dev/urandom", "r"); + if (!frand || !fread(&seed, sizeof(seed), 1, frand)) { + seed = time(NULL) * 1337; + } + fclose(frand); + } + secp256k1_rand_seed(seed); + + printf("test count = %i\n", count); + printf("random seed = %llu\n", (unsigned long long)seed); + + /* initialize */ + secp256k1_start(SECP256K1_START_SIGN | SECP256K1_START_VERIFY); + + /* num tests */ + run_num_smalltests(); + + /* scalar tests */ + run_scalar_tests(); + + /* field tests */ + run_field_inv(); + run_field_inv_var(); + run_field_inv_all(); + run_field_inv_all_var(); + run_sqr(); + run_sqrt(); + + /* group tests */ + run_ge(); + + /* ecmult tests */ + run_wnaf(); + run_point_times_order(); + run_ecmult_chain(); + + /* ecdsa tests */ + run_ecdsa_sign_verify(); + run_ecdsa_end_to_end(); + run_ecdsa_infinity(); +#ifdef ENABLE_OPENSSL_TESTS + run_ecdsa_openssl(); +#endif + + printf("random run = %llu\n", (unsigned long long)secp256k1_rand32() + ((unsigned long long)secp256k1_rand32() << 32)); + + /* shutdown */ + secp256k1_stop(); + return 0; +} diff --git a/src/util.h b/src/util.h new file mode 100644 index 0000000000..96b47057c0 --- /dev/null +++ b/src/util.h @@ -0,0 +1,64 @@ +/********************************************************************** + * Copyright (c) 2013, 2014 Pieter Wuille * + * Distributed under the MIT software license, see the accompanying * + * file COPYING or http://www.opensource.org/licenses/mit-license.php.* + **********************************************************************/ + +#ifndef _SECP256K1_UTIL_H_ +#define _SECP256K1_UTIL_H_ + +#if defined HAVE_CONFIG_H +#include "libsecp256k1-config.h" +#endif + +#include +#include +#include + +#ifdef DETERMINISTIC +#define TEST_FAILURE(msg) do { \ + fprintf(stderr, "%s\n", msg); \ + abort(); \ +} while(0); +#else +#define TEST_FAILURE(msg) do { \ + fprintf(stderr, "%s:%d: %s\n", __FILE__, __LINE__, msg); \ + abort(); \ +} while(0) +#endif + +#ifndef HAVE_BUILTIN_EXPECT +#define EXPECT(x,c) __builtin_expect((x),(c)) +#else +#define EXPECT(x,c) (x) +#endif + +#ifdef DETERMINISTIC +#define CHECK(cond) do { \ + if (EXPECT(!(cond), 0)) { \ + TEST_FAILURE("test condition failed"); \ + } \ +} while(0) +#else +#define CHECK(cond) do { \ + if (EXPECT(!(cond), 0)) { \ + TEST_FAILURE("test condition failed: " #cond); \ + } \ +} while(0) +#endif + +/* Like assert(), but safe to use on expressions with side effects. */ +#ifndef NDEBUG +#define DEBUG_CHECK CHECK +#else +#define DEBUG_CHECK(cond) do { (void)(cond); } while(0) +#endif + +/* Like DEBUG_CHECK(), but when VERIFY is defined instead of NDEBUG not defined. */ +#ifdef VERIFY +#define VERIFY_CHECK CHECK +#else +#define VERIFY_CHECK(cond) do { (void)(cond); } while(0) +#endif + +#endif -- cgit v1.2.3