mirror of
https://gitlab.winehq.org/wine/wine-gecko.git
synced 2024-09-13 09:24:08 -07:00
282 lines
12 KiB
C
282 lines
12 KiB
C
/*
|
|
* ***** BEGIN LICENSE BLOCK *****
|
|
* Version: MPL 1.1/GPL 2.0/LGPL 2.1
|
|
*
|
|
* The contents of this file are subject to the Mozilla Public License Version
|
|
* 1.1 (the "License"); you may not use this file except in compliance with
|
|
* the License. You may obtain a copy of the License at
|
|
* http://www.mozilla.org/MPL/
|
|
*
|
|
* Software distributed under the License is distributed on an "AS IS" basis,
|
|
* WITHOUT WARRANTY OF ANY KIND, either express or implied. See the License
|
|
* for the specific language governing rights and limitations under the
|
|
* License.
|
|
*
|
|
* The Original Code is the elliptic curve math library.
|
|
*
|
|
* The Initial Developer of the Original Code is
|
|
* Sun Microsystems, Inc.
|
|
* Portions created by the Initial Developer are Copyright (C) 2003
|
|
* the Initial Developer. All Rights Reserved.
|
|
*
|
|
* Contributor(s):
|
|
* Stephen Fung <fungstep@hotmail.com> and
|
|
* Douglas Stebila <douglas@stebila.ca>, Sun Microsystems Laboratories
|
|
*
|
|
* Alternatively, the contents of this file may be used under the terms of
|
|
* either the GNU General Public License Version 2 or later (the "GPL"), or
|
|
* the GNU Lesser General Public License Version 2.1 or later (the "LGPL"),
|
|
* in which case the provisions of the GPL or the LGPL are applicable instead
|
|
* of those above. If you wish to allow use of your version of this file only
|
|
* under the terms of either the GPL or the LGPL, and not to allow others to
|
|
* use your version of this file under the terms of the MPL, indicate your
|
|
* decision by deleting the provisions above and replace them with the notice
|
|
* and other provisions required by the GPL or the LGPL. If you do not delete
|
|
* the provisions above, a recipient may use your version of this file under
|
|
* the terms of any one of the MPL, the GPL or the LGPL.
|
|
*
|
|
* ***** END LICENSE BLOCK ***** */
|
|
|
|
#ifndef __ecl_priv_h_
|
|
#define __ecl_priv_h_
|
|
|
|
#include "ecl.h"
|
|
#include "mpi.h"
|
|
#include "mplogic.h"
|
|
|
|
/* MAX_FIELD_SIZE_DIGITS is the maximum size of field element supported */
|
|
/* the following needs to go away... */
|
|
#if defined(MP_USE_LONG_LONG_DIGIT) || defined(MP_USE_LONG_DIGIT)
|
|
#define ECL_SIXTY_FOUR_BIT
|
|
#else
|
|
#define ECL_THIRTY_TWO_BIT
|
|
#endif
|
|
|
|
#define ECL_CURVE_DIGITS(curve_size_in_bits) \
|
|
(((curve_size_in_bits)+(sizeof(mp_digit)*8-1))/(sizeof(mp_digit)*8))
|
|
#define ECL_BITS (sizeof(mp_digit)*8)
|
|
#define ECL_MAX_FIELD_SIZE_DIGITS (80/sizeof(mp_digit))
|
|
|
|
/* Gets the i'th bit in the binary representation of a. If i >= length(a),
|
|
* then return 0. (The above behaviour differs from mpl_get_bit, which
|
|
* causes an error if i >= length(a).) */
|
|
#define MP_GET_BIT(a, i) \
|
|
((i) >= mpl_significant_bits((a))) ? 0 : mpl_get_bit((a), (i))
|
|
|
|
#if !defined(MP_NO_MP_WORD) && !defined(MP_NO_ADD_WORD)
|
|
#define MP_ADD_CARRY(a1, a2, s, cin, cout) \
|
|
{ mp_word w; \
|
|
w = ((mp_word)(cin)) + (a1) + (a2); \
|
|
s = ACCUM(w); \
|
|
cout = CARRYOUT(w); }
|
|
|
|
#define MP_SUB_BORROW(a1, a2, s, bin, bout) \
|
|
{ mp_word w; \
|
|
w = ((mp_word)(a1)) - (a2) - (bin); \
|
|
s = ACCUM(w); \
|
|
bout = (w >> MP_DIGIT_BIT) & 1; }
|
|
|
|
#else
|
|
/* NOTE,
|
|
* cin and cout could be the same variable.
|
|
* bin and bout could be the same variable.
|
|
* a1 or a2 and s could be the same variable.
|
|
* don't trash those outputs until their respective inputs have
|
|
* been read. */
|
|
#define MP_ADD_CARRY(a1, a2, s, cin, cout) \
|
|
{ mp_digit tmp,sum; \
|
|
tmp = (a1); \
|
|
sum = tmp + (a2); \
|
|
tmp = (sum < tmp); /* detect overflow */ \
|
|
s = sum += (cin); \
|
|
cout = tmp + (sum < (cin)); }
|
|
|
|
#define MP_SUB_BORROW(a1, a2, s, bin, bout) \
|
|
{ mp_digit tmp; \
|
|
tmp = (a1); \
|
|
s = tmp - (a2); \
|
|
tmp = (s > tmp); /* detect borrow */ \
|
|
if ((bin) && !s--) tmp++; \
|
|
bout = tmp; }
|
|
#endif
|
|
|
|
|
|
struct GFMethodStr;
|
|
typedef struct GFMethodStr GFMethod;
|
|
struct GFMethodStr {
|
|
/* Indicates whether the structure was constructed from dynamic memory
|
|
* or statically created. */
|
|
int constructed;
|
|
/* Irreducible that defines the field. For prime fields, this is the
|
|
* prime p. For binary polynomial fields, this is the bitstring
|
|
* representation of the irreducible polynomial. */
|
|
mp_int irr;
|
|
/* For prime fields, the value irr_arr[0] is the number of bits in the
|
|
* field. For binary polynomial fields, the irreducible polynomial
|
|
* f(t) is represented as an array of unsigned int[], where f(t) is
|
|
* of the form: f(t) = t^p[0] + t^p[1] + ... + t^p[4] where m = p[0]
|
|
* > p[1] > ... > p[4] = 0. */
|
|
unsigned int irr_arr[5];
|
|
/* Field arithmetic methods. All methods (except field_enc and
|
|
* field_dec) are assumed to take field-encoded parameters and return
|
|
* field-encoded values. All methods (except field_enc and field_dec)
|
|
* are required to be implemented. */
|
|
mp_err (*field_add) (const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err (*field_neg) (const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err (*field_sub) (const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err (*field_mod) (const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err (*field_mul) (const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err (*field_sqr) (const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err (*field_div) (const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err (*field_enc) (const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err (*field_dec) (const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
/* Extra storage for implementation-specific data. Any memory
|
|
* allocated to these extra fields will be cleared by extra_free. */
|
|
void *extra1;
|
|
void *extra2;
|
|
void (*extra_free) (GFMethod *meth);
|
|
};
|
|
|
|
/* Construct generic GFMethods. */
|
|
GFMethod *GFMethod_consGFp(const mp_int *irr);
|
|
GFMethod *GFMethod_consGFp_mont(const mp_int *irr);
|
|
GFMethod *GFMethod_consGF2m(const mp_int *irr,
|
|
const unsigned int irr_arr[5]);
|
|
/* Free the memory allocated (if any) to a GFMethod object. */
|
|
void GFMethod_free(GFMethod *meth);
|
|
|
|
struct ECGroupStr {
|
|
/* Indicates whether the structure was constructed from dynamic memory
|
|
* or statically created. */
|
|
int constructed;
|
|
/* Field definition and arithmetic. */
|
|
GFMethod *meth;
|
|
/* Textual representation of curve name, if any. */
|
|
char *text;
|
|
/* Curve parameters, field-encoded. */
|
|
mp_int curvea, curveb;
|
|
/* x and y coordinates of the base point, field-encoded. */
|
|
mp_int genx, geny;
|
|
/* Order and cofactor of the base point. */
|
|
mp_int order;
|
|
int cofactor;
|
|
/* Point arithmetic methods. All methods are assumed to take
|
|
* field-encoded parameters and return field-encoded values. All
|
|
* methods (except base_point_mul and points_mul) are required to be
|
|
* implemented. */
|
|
mp_err (*point_add) (const mp_int *px, const mp_int *py,
|
|
const mp_int *qx, const mp_int *qy, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
mp_err (*point_sub) (const mp_int *px, const mp_int *py,
|
|
const mp_int *qx, const mp_int *qy, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
mp_err (*point_dbl) (const mp_int *px, const mp_int *py, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
mp_err (*point_mul) (const mp_int *n, const mp_int *px,
|
|
const mp_int *py, mp_int *rx, mp_int *ry,
|
|
const ECGroup *group);
|
|
mp_err (*base_point_mul) (const mp_int *n, mp_int *rx, mp_int *ry,
|
|
const ECGroup *group);
|
|
mp_err (*points_mul) (const mp_int *k1, const mp_int *k2,
|
|
const mp_int *px, const mp_int *py, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
mp_err (*validate_point) (const mp_int *px, const mp_int *py, const ECGroup *group);
|
|
/* Extra storage for implementation-specific data. Any memory
|
|
* allocated to these extra fields will be cleared by extra_free. */
|
|
void *extra1;
|
|
void *extra2;
|
|
void (*extra_free) (ECGroup *group);
|
|
};
|
|
|
|
/* Wrapper functions for generic prime field arithmetic. */
|
|
mp_err ec_GFp_add(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_neg(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GFp_sub(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
|
|
/* fixed length in-line adds. Count is in words */
|
|
mp_err ec_GFp_add_3(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_add_4(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_add_5(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_add_6(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sub_3(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sub_4(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sub_5(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sub_6(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
|
|
mp_err ec_GFp_mod(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GFp_mul(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sqr(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GFp_div(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
/* Wrapper functions for generic binary polynomial field arithmetic. */
|
|
mp_err ec_GF2m_add(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GF2m_neg(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GF2m_mod(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GF2m_mul(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GF2m_sqr(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GF2m_div(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
|
|
/* Montgomery prime field arithmetic. */
|
|
mp_err ec_GFp_mul_mont(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_sqr_mont(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GFp_div_mont(const mp_int *a, const mp_int *b, mp_int *r,
|
|
const GFMethod *meth);
|
|
mp_err ec_GFp_enc_mont(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
mp_err ec_GFp_dec_mont(const mp_int *a, mp_int *r, const GFMethod *meth);
|
|
void ec_GFp_extra_free_mont(GFMethod *meth);
|
|
|
|
/* point multiplication */
|
|
mp_err ec_pts_mul_basic(const mp_int *k1, const mp_int *k2,
|
|
const mp_int *px, const mp_int *py, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
mp_err ec_pts_mul_simul_w2(const mp_int *k1, const mp_int *k2,
|
|
const mp_int *px, const mp_int *py, mp_int *rx,
|
|
mp_int *ry, const ECGroup *group);
|
|
|
|
/* Computes the windowed non-adjacent-form (NAF) of a scalar. Out should
|
|
* be an array of signed char's to output to, bitsize should be the number
|
|
* of bits of out, in is the original scalar, and w is the window size.
|
|
* NAF is discussed in the paper: D. Hankerson, J. Hernandez and A.
|
|
* Menezes, "Software implementation of elliptic curve cryptography over
|
|
* binary fields", Proc. CHES 2000. */
|
|
mp_err ec_compute_wNAF(signed char *out, int bitsize, const mp_int *in,
|
|
int w);
|
|
|
|
/* Optimized field arithmetic */
|
|
mp_err ec_group_set_gfp192(ECGroup *group, ECCurveName);
|
|
mp_err ec_group_set_gfp224(ECGroup *group, ECCurveName);
|
|
mp_err ec_group_set_gfp256(ECGroup *group, ECCurveName);
|
|
mp_err ec_group_set_gfp384(ECGroup *group, ECCurveName);
|
|
mp_err ec_group_set_gfp521(ECGroup *group, ECCurveName);
|
|
mp_err ec_group_set_gf2m163(ECGroup *group, ECCurveName name);
|
|
mp_err ec_group_set_gf2m193(ECGroup *group, ECCurveName name);
|
|
mp_err ec_group_set_gf2m233(ECGroup *group, ECCurveName name);
|
|
|
|
/* Optimized floating-point arithmetic */
|
|
#ifdef ECL_USE_FP
|
|
mp_err ec_group_set_secp160r1_fp(ECGroup *group);
|
|
mp_err ec_group_set_nistp192_fp(ECGroup *group);
|
|
mp_err ec_group_set_nistp224_fp(ECGroup *group);
|
|
#endif
|
|
|
|
#endif /* __ecl_priv_h_ */
|