sl@0: /* crypto/bn/bn_gf2m.c */
sl@0: /* ====================================================================
sl@0:  * Copyright 2002 Sun Microsystems, Inc. ALL RIGHTS RESERVED.
sl@0:  *
sl@0:  * The Elliptic Curve Public-Key Crypto Library (ECC Code) included
sl@0:  * herein is developed by SUN MICROSYSTEMS, INC., and is contributed
sl@0:  * to the OpenSSL project.
sl@0:  *
sl@0:  * The ECC Code is licensed pursuant to the OpenSSL open source
sl@0:  * license provided below.
sl@0:  *
sl@0:  * In addition, Sun covenants to all licensees who provide a reciprocal
sl@0:  * covenant with respect to their own patents if any, not to sue under
sl@0:  * current and future patent claims necessarily infringed by the making,
sl@0:  * using, practicing, selling, offering for sale and/or otherwise
sl@0:  * disposing of the ECC Code as delivered hereunder (or portions thereof),
sl@0:  * provided that such covenant shall not apply:
sl@0:  *  1) for code that a licensee deletes from the ECC Code;
sl@0:  *  2) separates from the ECC Code; or
sl@0:  *  3) for infringements caused by:
sl@0:  *       i) the modification of the ECC Code or
sl@0:  *      ii) the combination of the ECC Code with other software or
sl@0:  *          devices where such combination causes the infringement.
sl@0:  *
sl@0:  * The software is originally written by Sheueling Chang Shantz and
sl@0:  * Douglas Stebila of Sun Microsystems Laboratories.
sl@0:  *
sl@0:  */
sl@0: 
sl@0: /* NOTE: This file is licensed pursuant to the OpenSSL license below
sl@0:  * and may be modified; but after modifications, the above covenant
sl@0:  * may no longer apply!  In such cases, the corresponding paragraph
sl@0:  * ["In addition, Sun covenants ... causes the infringement."] and
sl@0:  * this note can be edited out; but please keep the Sun copyright
sl@0:  * notice and attribution. */
sl@0: 
sl@0: /* ====================================================================
sl@0:  * Copyright (c) 1998-2002 The OpenSSL Project.  All rights reserved.
sl@0:  *
sl@0:  * Redistribution and use in source and binary forms, with or without
sl@0:  * modification, are permitted provided that the following conditions
sl@0:  * are met:
sl@0:  *
sl@0:  * 1. Redistributions of source code must retain the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer. 
sl@0:  *
sl@0:  * 2. Redistributions in binary form must reproduce the above copyright
sl@0:  *    notice, this list of conditions and the following disclaimer in
sl@0:  *    the documentation and/or other materials provided with the
sl@0:  *    distribution.
sl@0:  *
sl@0:  * 3. All advertising materials mentioning features or use of this
sl@0:  *    software must display the following acknowledgment:
sl@0:  *    "This product includes software developed by the OpenSSL Project
sl@0:  *    for use in the OpenSSL Toolkit. (http://www.openssl.org/)"
sl@0:  *
sl@0:  * 4. The names "OpenSSL Toolkit" and "OpenSSL Project" must not be used to
sl@0:  *    endorse or promote products derived from this software without
sl@0:  *    prior written permission. For written permission, please contact
sl@0:  *    openssl-core@openssl.org.
sl@0:  *
sl@0:  * 5. Products derived from this software may not be called "OpenSSL"
sl@0:  *    nor may "OpenSSL" appear in their names without prior written
sl@0:  *    permission of the OpenSSL Project.
sl@0:  *
sl@0:  * 6. Redistributions of any form whatsoever must retain the following
sl@0:  *    acknowledgment:
sl@0:  *    "This product includes software developed by the OpenSSL Project
sl@0:  *    for use in the OpenSSL Toolkit (http://www.openssl.org/)"
sl@0:  *
sl@0:  * THIS SOFTWARE IS PROVIDED BY THE OpenSSL PROJECT ``AS IS'' AND ANY
sl@0:  * EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
sl@0:  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
sl@0:  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE OpenSSL PROJECT OR
sl@0:  * ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL,
sl@0:  * SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT
sl@0:  * NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
sl@0:  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
sl@0:  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT,
sl@0:  * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
sl@0:  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED
sl@0:  * OF THE POSSIBILITY OF SUCH DAMAGE.
sl@0:  * ====================================================================
sl@0:  *
sl@0:  * This product includes cryptographic software written by Eric Young
sl@0:  * (eay@cryptsoft.com).  This product includes software written by Tim
sl@0:  * Hudson (tjh@cryptsoft.com).
sl@0:  *
sl@0:  */
sl@0: 
sl@0: #include <assert.h>
sl@0: #include <limits.h>
sl@0: #include <stdio.h>
sl@0: #include "cryptlib.h"
sl@0: #include "bn_lcl.h"
sl@0: 
sl@0: /* Maximum number of iterations before BN_GF2m_mod_solve_quad_arr should fail. */
sl@0: #define MAX_ITERATIONS 50
sl@0: 
sl@0: static const BN_ULONG SQR_tb[16] =
sl@0:   {     0,     1,     4,     5,    16,    17,    20,    21,
sl@0:        64,    65,    68,    69,    80,    81,    84,    85 };
sl@0: /* Platform-specific macros to accelerate squaring. */
sl@0: #if defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG)
sl@0: #define SQR1(w) \
sl@0:     SQR_tb[(w) >> 60 & 0xF] << 56 | SQR_tb[(w) >> 56 & 0xF] << 48 | \
sl@0:     SQR_tb[(w) >> 52 & 0xF] << 40 | SQR_tb[(w) >> 48 & 0xF] << 32 | \
sl@0:     SQR_tb[(w) >> 44 & 0xF] << 24 | SQR_tb[(w) >> 40 & 0xF] << 16 | \
sl@0:     SQR_tb[(w) >> 36 & 0xF] <<  8 | SQR_tb[(w) >> 32 & 0xF]
sl@0: #define SQR0(w) \
sl@0:     SQR_tb[(w) >> 28 & 0xF] << 56 | SQR_tb[(w) >> 24 & 0xF] << 48 | \
sl@0:     SQR_tb[(w) >> 20 & 0xF] << 40 | SQR_tb[(w) >> 16 & 0xF] << 32 | \
sl@0:     SQR_tb[(w) >> 12 & 0xF] << 24 | SQR_tb[(w) >>  8 & 0xF] << 16 | \
sl@0:     SQR_tb[(w) >>  4 & 0xF] <<  8 | SQR_tb[(w)       & 0xF]
sl@0: #endif
sl@0: #ifdef THIRTY_TWO_BIT
sl@0: #define SQR1(w) \
sl@0:     SQR_tb[(w) >> 28 & 0xF] << 24 | SQR_tb[(w) >> 24 & 0xF] << 16 | \
sl@0:     SQR_tb[(w) >> 20 & 0xF] <<  8 | SQR_tb[(w) >> 16 & 0xF]
sl@0: #define SQR0(w) \
sl@0:     SQR_tb[(w) >> 12 & 0xF] << 24 | SQR_tb[(w) >>  8 & 0xF] << 16 | \
sl@0:     SQR_tb[(w) >>  4 & 0xF] <<  8 | SQR_tb[(w)       & 0xF]
sl@0: #endif
sl@0: #ifdef SIXTEEN_BIT
sl@0: #define SQR1(w) \
sl@0:     SQR_tb[(w) >> 12 & 0xF] <<  8 | SQR_tb[(w) >>  8 & 0xF]
sl@0: #define SQR0(w) \
sl@0:     SQR_tb[(w) >>  4 & 0xF] <<  8 | SQR_tb[(w)       & 0xF]
sl@0: #endif
sl@0: #ifdef EIGHT_BIT
sl@0: #define SQR1(w) \
sl@0:     SQR_tb[(w) >>  4 & 0xF]
sl@0: #define SQR0(w) \
sl@0:     SQR_tb[(w)       & 15]
sl@0: #endif
sl@0: 
sl@0: /* Product of two polynomials a, b each with degree < BN_BITS2 - 1,
sl@0:  * result is a polynomial r with degree < 2 * BN_BITS - 1
sl@0:  * The caller MUST ensure that the variables have the right amount
sl@0:  * of space allocated.
sl@0:  */
sl@0: #ifdef EIGHT_BIT
sl@0: static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a, const BN_ULONG b)
sl@0: 	{
sl@0: 	register BN_ULONG h, l, s;
sl@0: 	BN_ULONG tab[4], top1b = a >> 7;
sl@0: 	register BN_ULONG a1, a2;
sl@0: 
sl@0: 	a1 = a & (0x7F); a2 = a1 << 1;
sl@0: 
sl@0: 	tab[0] = 0; tab[1] = a1; tab[2] = a2; tab[3] = a1^a2;
sl@0: 
sl@0: 	s = tab[b      & 0x3]; l  = s;
sl@0: 	s = tab[b >> 2 & 0x3]; l ^= s << 2; h  = s >> 6;
sl@0: 	s = tab[b >> 4 & 0x3]; l ^= s << 4; h ^= s >> 4;
sl@0: 	s = tab[b >> 6      ]; l ^= s << 6; h ^= s >> 2;
sl@0: 	
sl@0: 	/* compensate for the top bit of a */
sl@0: 
sl@0: 	if (top1b & 01) { l ^= b << 7; h ^= b >> 1; } 
sl@0: 
sl@0: 	*r1 = h; *r0 = l;
sl@0: 	} 
sl@0: #endif
sl@0: #ifdef SIXTEEN_BIT
sl@0: static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a, const BN_ULONG b)
sl@0: 	{
sl@0: 	register BN_ULONG h, l, s;
sl@0: 	BN_ULONG tab[4], top1b = a >> 15; 
sl@0: 	register BN_ULONG a1, a2;
sl@0: 
sl@0: 	a1 = a & (0x7FFF); a2 = a1 << 1;
sl@0: 
sl@0: 	tab[0] = 0; tab[1] = a1; tab[2] = a2; tab[3] = a1^a2;
sl@0: 
sl@0: 	s = tab[b      & 0x3]; l  = s;
sl@0: 	s = tab[b >> 2 & 0x3]; l ^= s <<  2; h  = s >> 14;
sl@0: 	s = tab[b >> 4 & 0x3]; l ^= s <<  4; h ^= s >> 12;
sl@0: 	s = tab[b >> 6 & 0x3]; l ^= s <<  6; h ^= s >> 10;
sl@0: 	s = tab[b >> 8 & 0x3]; l ^= s <<  8; h ^= s >>  8;
sl@0: 	s = tab[b >>10 & 0x3]; l ^= s << 10; h ^= s >>  6;
sl@0: 	s = tab[b >>12 & 0x3]; l ^= s << 12; h ^= s >>  4;
sl@0: 	s = tab[b >>14      ]; l ^= s << 14; h ^= s >>  2;
sl@0: 
sl@0: 	/* compensate for the top bit of a */
sl@0: 
sl@0: 	if (top1b & 01) { l ^= b << 15; h ^= b >> 1; } 
sl@0: 
sl@0: 	*r1 = h; *r0 = l;
sl@0: 	} 
sl@0: #endif
sl@0: #ifdef THIRTY_TWO_BIT
sl@0: static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a, const BN_ULONG b)
sl@0: 	{
sl@0: 	register BN_ULONG h, l, s;
sl@0: 	BN_ULONG tab[8], top2b = a >> 30; 
sl@0: 	register BN_ULONG a1, a2, a4;
sl@0: 
sl@0: 	a1 = a & (0x3FFFFFFF); a2 = a1 << 1; a4 = a2 << 1;
sl@0: 
sl@0: 	tab[0] =  0; tab[1] = a1;    tab[2] = a2;    tab[3] = a1^a2;
sl@0: 	tab[4] = a4; tab[5] = a1^a4; tab[6] = a2^a4; tab[7] = a1^a2^a4;
sl@0: 
sl@0: 	s = tab[b       & 0x7]; l  = s;
sl@0: 	s = tab[b >>  3 & 0x7]; l ^= s <<  3; h  = s >> 29;
sl@0: 	s = tab[b >>  6 & 0x7]; l ^= s <<  6; h ^= s >> 26;
sl@0: 	s = tab[b >>  9 & 0x7]; l ^= s <<  9; h ^= s >> 23;
sl@0: 	s = tab[b >> 12 & 0x7]; l ^= s << 12; h ^= s >> 20;
sl@0: 	s = tab[b >> 15 & 0x7]; l ^= s << 15; h ^= s >> 17;
sl@0: 	s = tab[b >> 18 & 0x7]; l ^= s << 18; h ^= s >> 14;
sl@0: 	s = tab[b >> 21 & 0x7]; l ^= s << 21; h ^= s >> 11;
sl@0: 	s = tab[b >> 24 & 0x7]; l ^= s << 24; h ^= s >>  8;
sl@0: 	s = tab[b >> 27 & 0x7]; l ^= s << 27; h ^= s >>  5;
sl@0: 	s = tab[b >> 30      ]; l ^= s << 30; h ^= s >>  2;
sl@0: 
sl@0: 	/* compensate for the top two bits of a */
sl@0: 
sl@0: 	if (top2b & 01) { l ^= b << 30; h ^= b >> 2; } 
sl@0: 	if (top2b & 02) { l ^= b << 31; h ^= b >> 1; } 
sl@0: 
sl@0: 	*r1 = h; *r0 = l;
sl@0: 	} 
sl@0: #endif
sl@0: #if defined(SIXTY_FOUR_BIT) || defined(SIXTY_FOUR_BIT_LONG)
sl@0: static void bn_GF2m_mul_1x1(BN_ULONG *r1, BN_ULONG *r0, const BN_ULONG a, const BN_ULONG b)
sl@0: 	{
sl@0: 	register BN_ULONG h, l, s;
sl@0: 	BN_ULONG tab[16], top3b = a >> 61;
sl@0: 	register BN_ULONG a1, a2, a4, a8;
sl@0: 
sl@0: 	a1 = a & (0x1FFFFFFFFFFFFFFFULL); a2 = a1 << 1; a4 = a2 << 1; a8 = a4 << 1;
sl@0: 
sl@0: 	tab[ 0] = 0;     tab[ 1] = a1;       tab[ 2] = a2;       tab[ 3] = a1^a2;
sl@0: 	tab[ 4] = a4;    tab[ 5] = a1^a4;    tab[ 6] = a2^a4;    tab[ 7] = a1^a2^a4;
sl@0: 	tab[ 8] = a8;    tab[ 9] = a1^a8;    tab[10] = a2^a8;    tab[11] = a1^a2^a8;
sl@0: 	tab[12] = a4^a8; tab[13] = a1^a4^a8; tab[14] = a2^a4^a8; tab[15] = a1^a2^a4^a8;
sl@0: 
sl@0: 	s = tab[b       & 0xF]; l  = s;
sl@0: 	s = tab[b >>  4 & 0xF]; l ^= s <<  4; h  = s >> 60;
sl@0: 	s = tab[b >>  8 & 0xF]; l ^= s <<  8; h ^= s >> 56;
sl@0: 	s = tab[b >> 12 & 0xF]; l ^= s << 12; h ^= s >> 52;
sl@0: 	s = tab[b >> 16 & 0xF]; l ^= s << 16; h ^= s >> 48;
sl@0: 	s = tab[b >> 20 & 0xF]; l ^= s << 20; h ^= s >> 44;
sl@0: 	s = tab[b >> 24 & 0xF]; l ^= s << 24; h ^= s >> 40;
sl@0: 	s = tab[b >> 28 & 0xF]; l ^= s << 28; h ^= s >> 36;
sl@0: 	s = tab[b >> 32 & 0xF]; l ^= s << 32; h ^= s >> 32;
sl@0: 	s = tab[b >> 36 & 0xF]; l ^= s << 36; h ^= s >> 28;
sl@0: 	s = tab[b >> 40 & 0xF]; l ^= s << 40; h ^= s >> 24;
sl@0: 	s = tab[b >> 44 & 0xF]; l ^= s << 44; h ^= s >> 20;
sl@0: 	s = tab[b >> 48 & 0xF]; l ^= s << 48; h ^= s >> 16;
sl@0: 	s = tab[b >> 52 & 0xF]; l ^= s << 52; h ^= s >> 12;
sl@0: 	s = tab[b >> 56 & 0xF]; l ^= s << 56; h ^= s >>  8;
sl@0: 	s = tab[b >> 60      ]; l ^= s << 60; h ^= s >>  4;
sl@0: 
sl@0: 	/* compensate for the top three bits of a */
sl@0: 
sl@0: 	if (top3b & 01) { l ^= b << 61; h ^= b >> 3; } 
sl@0: 	if (top3b & 02) { l ^= b << 62; h ^= b >> 2; } 
sl@0: 	if (top3b & 04) { l ^= b << 63; h ^= b >> 1; } 
sl@0: 
sl@0: 	*r1 = h; *r0 = l;
sl@0: 	} 
sl@0: #endif
sl@0: 
sl@0: /* Product of two polynomials a, b each with degree < 2 * BN_BITS2 - 1,
sl@0:  * result is a polynomial r with degree < 4 * BN_BITS2 - 1
sl@0:  * The caller MUST ensure that the variables have the right amount
sl@0:  * of space allocated.
sl@0:  */
sl@0: static void bn_GF2m_mul_2x2(BN_ULONG *r, const BN_ULONG a1, const BN_ULONG a0, const BN_ULONG b1, const BN_ULONG b0)
sl@0: 	{
sl@0: 	BN_ULONG m1, m0;
sl@0: 	/* r[3] = h1, r[2] = h0; r[1] = l1; r[0] = l0 */
sl@0: 	bn_GF2m_mul_1x1(r+3, r+2, a1, b1);
sl@0: 	bn_GF2m_mul_1x1(r+1, r, a0, b0);
sl@0: 	bn_GF2m_mul_1x1(&m1, &m0, a0 ^ a1, b0 ^ b1);
sl@0: 	/* Correction on m1 ^= l1 ^ h1; m0 ^= l0 ^ h0; */
sl@0: 	r[2] ^= m1 ^ r[1] ^ r[3];  /* h0 ^= m1 ^ l1 ^ h1; */
sl@0: 	r[1] = r[3] ^ r[2] ^ r[0] ^ m1 ^ m0;  /* l1 ^= l0 ^ h0 ^ m0; */
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Add polynomials a and b and store result in r; r could be a or b, a and b 
sl@0:  * could be equal; r is the bitwise XOR of a and b.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_add(BIGNUM *r, const BIGNUM *a, const BIGNUM *b)
sl@0: 	{
sl@0: 	int i;
sl@0: 	const BIGNUM *at, *bt;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(b);
sl@0: 
sl@0: 	if (a->top < b->top) { at = b; bt = a; }
sl@0: 	else { at = a; bt = b; }
sl@0: 
sl@0: 	bn_wexpand(r, at->top);
sl@0: 
sl@0: 	for (i = 0; i < bt->top; i++)
sl@0: 		{
sl@0: 		r->d[i] = at->d[i] ^ bt->d[i];
sl@0: 		}
sl@0: 	for (; i < at->top; i++)
sl@0: 		{
sl@0: 		r->d[i] = at->d[i];
sl@0: 		}
sl@0: 	
sl@0: 	r->top = at->top;
sl@0: 	bn_correct_top(r);
sl@0: 	
sl@0: 	return 1;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Some functions allow for representation of the irreducible polynomials
sl@0:  * as an int[], say p.  The irreducible f(t) is then of the form:
sl@0:  *     t^p[0] + t^p[1] + ... + t^p[k]
sl@0:  * where m = p[0] > p[1] > ... > p[k] = 0.
sl@0:  */
sl@0: 
sl@0: 
sl@0: /* Performs modular reduction of a and store result in r.  r could be a. */
sl@0: EXPORT_C int BN_GF2m_mod_arr(BIGNUM *r, const BIGNUM *a, const unsigned int p[])
sl@0: 	{
sl@0: 	int j, k;
sl@0: 	int n, dN, d0, d1;
sl@0: 	BN_ULONG zz, *z;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 
sl@0: 	if (!p[0])
sl@0: 		{
sl@0: 		/* reduction mod 1 => return 0 */
sl@0: 		BN_zero(r);
sl@0: 		return 1;
sl@0: 		}
sl@0: 
sl@0: 	/* Since the algorithm does reduction in the r value, if a != r, copy
sl@0: 	 * the contents of a into r so we can do reduction in r. 
sl@0: 	 */
sl@0: 	if (a != r)
sl@0: 		{
sl@0: 		if (!bn_wexpand(r, a->top)) return 0;
sl@0: 		for (j = 0; j < a->top; j++)
sl@0: 			{
sl@0: 			r->d[j] = a->d[j];
sl@0: 			}
sl@0: 		r->top = a->top;
sl@0: 		}
sl@0: 	z = r->d;
sl@0: 
sl@0: 	/* start reduction */
sl@0: 	dN = p[0] / BN_BITS2;  
sl@0: 	for (j = r->top - 1; j > dN;)
sl@0: 		{
sl@0: 		zz = z[j];
sl@0: 		if (z[j] == 0) { j--; continue; }
sl@0: 		z[j] = 0;
sl@0: 
sl@0: 		for (k = 1; p[k] != 0; k++)
sl@0: 			{
sl@0: 			/* reducing component t^p[k] */
sl@0: 			n = p[0] - p[k];
sl@0: 			d0 = n % BN_BITS2;  d1 = BN_BITS2 - d0;
sl@0: 			n /= BN_BITS2; 
sl@0: 			z[j-n] ^= (zz>>d0);
sl@0: 			if (d0) z[j-n-1] ^= (zz<<d1);
sl@0: 			}
sl@0: 
sl@0: 		/* reducing component t^0 */
sl@0: 		n = dN;  
sl@0: 		d0 = p[0] % BN_BITS2;
sl@0: 		d1 = BN_BITS2 - d0;
sl@0: 		z[j-n] ^= (zz >> d0);
sl@0: 		if (d0) z[j-n-1] ^= (zz << d1);
sl@0: 		}
sl@0: 
sl@0: 	/* final round of reduction */
sl@0: 	while (j == dN)
sl@0: 		{
sl@0: 
sl@0: 		d0 = p[0] % BN_BITS2;
sl@0: 		zz = z[dN] >> d0;
sl@0: 		if (zz == 0) break;
sl@0: 		d1 = BN_BITS2 - d0;
sl@0: 		
sl@0: 		if (d0) z[dN] = (z[dN] << d1) >> d1; /* clear up the top d1 bits */
sl@0: 		z[0] ^= zz; /* reduction t^0 component */
sl@0: 
sl@0: 		for (k = 1; p[k] != 0; k++)
sl@0: 			{
sl@0: 			BN_ULONG tmp_ulong;
sl@0: 
sl@0: 			/* reducing component t^p[k]*/
sl@0: 			n = p[k] / BN_BITS2;   
sl@0: 			d0 = p[k] % BN_BITS2;
sl@0: 			d1 = BN_BITS2 - d0;
sl@0: 			z[n] ^= (zz << d0);
sl@0: 			tmp_ulong = zz >> d1;
sl@0:                         if (d0 && tmp_ulong)
sl@0:                                 z[n+1] ^= tmp_ulong;
sl@0: 			}
sl@0: 
sl@0: 		
sl@0: 		}
sl@0: 
sl@0: 	bn_correct_top(r);
sl@0: 	return 1;
sl@0: 	}
sl@0: 
sl@0: /* Performs modular reduction of a by p and store result in r.  r could be a.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_arr function.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod(BIGNUM *r, const BIGNUM *a, const BIGNUM *p)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) * max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_arr(r, a, arr);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Compute the product of two polynomials a and b, reduce modulo p, and store
sl@0:  * the result in r.  r could be a or b; a could be b.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod_mul_arr(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	int zlen, i, j, k, ret = 0;
sl@0: 	BIGNUM *s;
sl@0: 	BN_ULONG x1, x0, y1, y0, zz[4];
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(b);
sl@0: 
sl@0: 	if (a == b)
sl@0: 		{
sl@0: 		return BN_GF2m_mod_sqr_arr(r, a, p, ctx);
sl@0: 		}
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((s = BN_CTX_get(ctx)) == NULL) goto err;
sl@0: 	
sl@0: 	zlen = a->top + b->top + 4;
sl@0: 	if (!bn_wexpand(s, zlen)) goto err;
sl@0: 	s->top = zlen;
sl@0: 
sl@0: 	for (i = 0; i < zlen; i++) s->d[i] = 0;
sl@0: 
sl@0: 	for (j = 0; j < b->top; j += 2)
sl@0: 		{
sl@0: 		y0 = b->d[j];
sl@0: 		y1 = ((j+1) == b->top) ? 0 : b->d[j+1];
sl@0: 		for (i = 0; i < a->top; i += 2)
sl@0: 			{
sl@0: 			x0 = a->d[i];
sl@0: 			x1 = ((i+1) == a->top) ? 0 : a->d[i+1];
sl@0: 			bn_GF2m_mul_2x2(zz, x1, x0, y1, y0);
sl@0: 			for (k = 0; k < 4; k++) s->d[i+j+k] ^= zz[k];
sl@0: 			}
sl@0: 		}
sl@0: 
sl@0: 	bn_correct_top(s);
sl@0: 	if (BN_GF2m_mod_arr(r, s, p))
sl@0: 		ret = 1;
sl@0: 	bn_check_top(r);
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Compute the product of two polynomials a and b, reduce modulo p, and store
sl@0:  * the result in r.  r could be a or b; a could equal b.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_mul_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_mul_arr function.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod_mul(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(b);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) * max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_MUL,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_mul_arr(r, a, b, arr, ctx);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Square a, reduce the result mod p, and store it in a.  r could be a. */
sl@0: EXPORT_C int	BN_GF2m_mod_sqr_arr(BIGNUM *r, const BIGNUM *a, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	int i, ret = 0;
sl@0: 	BIGNUM *s;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((s = BN_CTX_get(ctx)) == NULL) return 0;
sl@0: 	if (!bn_wexpand(s, 2 * a->top)) goto err;
sl@0: 
sl@0: 	for (i = a->top - 1; i >= 0; i--)
sl@0: 		{
sl@0: 		s->d[2*i+1] = SQR1(a->d[i]);
sl@0: 		s->d[2*i  ] = SQR0(a->d[i]);
sl@0: 		}
sl@0: 
sl@0: 	s->top = 2 * a->top;
sl@0: 	bn_correct_top(s);
sl@0: 	if (!BN_GF2m_mod_arr(r, s, p)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 	ret = 1;
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Square a, reduce the result mod p, and store it in a.  r could be a.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_sqr_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_sqr_arr function.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod_sqr(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) * max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_SQR,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_sqr_arr(r, a, arr, ctx);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Invert a, reduce modulo p, and store the result in r. r could be a. 
sl@0:  * Uses Modified Almost Inverse Algorithm (Algorithm 10) from
sl@0:  *     Hankerson, D., Hernandez, J.L., and Menezes, A.  "Software Implementation
sl@0:  *     of Elliptic Curve Cryptography Over Binary Fields".
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_inv(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	BIGNUM *b, *c, *u, *v, *tmp;
sl@0: 	int ret = 0;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(p);
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	
sl@0: 	b = BN_CTX_get(ctx);
sl@0: 	c = BN_CTX_get(ctx);
sl@0: 	u = BN_CTX_get(ctx);
sl@0: 	v = BN_CTX_get(ctx);
sl@0: 	if (v == NULL) goto err;
sl@0: 
sl@0: 	if (!BN_one(b)) goto err;
sl@0: 	if (!BN_GF2m_mod(u, a, p)) goto err;
sl@0: 	if (!BN_copy(v, p)) goto err;
sl@0: 
sl@0: 	if (BN_is_zero(u)) goto err;
sl@0: 
sl@0: 	while (1)
sl@0: 		{
sl@0: 		while (!BN_is_odd(u))
sl@0: 			{
sl@0: 			if (!BN_rshift1(u, u)) goto err;
sl@0: 			if (BN_is_odd(b))
sl@0: 				{
sl@0: 				if (!BN_GF2m_add(b, b, p)) goto err;
sl@0: 				}
sl@0: 			if (!BN_rshift1(b, b)) goto err;
sl@0: 			}
sl@0: 
sl@0: 		if (BN_abs_is_word(u, 1)) break;
sl@0: 
sl@0: 		if (BN_num_bits(u) < BN_num_bits(v))
sl@0: 			{
sl@0: 			tmp = u; u = v; v = tmp;
sl@0: 			tmp = b; b = c; c = tmp;
sl@0: 			}
sl@0: 		
sl@0: 		if (!BN_GF2m_add(u, u, v)) goto err;
sl@0: 		if (!BN_GF2m_add(b, b, c)) goto err;
sl@0: 		}
sl@0: 
sl@0: 
sl@0: 	if (!BN_copy(r, b)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 	ret = 1;
sl@0: 
sl@0: err:
sl@0:   	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Invert xx, reduce modulo p, and store the result in r. r could be xx. 
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_inv implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_inv function.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_inv_arr(BIGNUM *r, const BIGNUM *xx, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	BIGNUM *field;
sl@0: 	int ret = 0;
sl@0: 
sl@0: 	bn_check_top(xx);
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((field = BN_CTX_get(ctx)) == NULL) goto err;
sl@0: 	if (!BN_GF2m_arr2poly(p, field)) goto err;
sl@0: 	
sl@0: 	ret = BN_GF2m_mod_inv(r, xx, field, ctx);
sl@0: 	bn_check_top(r);
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: #ifndef OPENSSL_SUN_GF2M_DIV
sl@0: /* Divide y by x, reduce modulo p, and store the result in r. r could be x 
sl@0:  * or y, x could equal y.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_div(BIGNUM *r, const BIGNUM *y, const BIGNUM *x, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	BIGNUM *xinv = NULL;
sl@0: 	int ret = 0;
sl@0: 
sl@0: 	bn_check_top(y);
sl@0: 	bn_check_top(x);
sl@0: 	bn_check_top(p);
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	xinv = BN_CTX_get(ctx);
sl@0: 	if (xinv == NULL) goto err;
sl@0: 	
sl@0: 	if (!BN_GF2m_mod_inv(xinv, x, p, ctx)) goto err;
sl@0: 	if (!BN_GF2m_mod_mul(r, y, xinv, p, ctx)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 	ret = 1;
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: #else
sl@0: /* Divide y by x, reduce modulo p, and store the result in r. r could be x 
sl@0:  * or y, x could equal y.
sl@0:  * Uses algorithm Modular_Division_GF(2^m) from 
sl@0:  *     Chang-Shantz, S.  "From Euclid's GCD to Montgomery Multiplication to 
sl@0:  *     the Great Divide".
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_div(BIGNUM *r, const BIGNUM *y, const BIGNUM *x, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	BIGNUM *a, *b, *u, *v;
sl@0: 	int ret = 0;
sl@0: 
sl@0: 	bn_check_top(y);
sl@0: 	bn_check_top(x);
sl@0: 	bn_check_top(p);
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	
sl@0: 	a = BN_CTX_get(ctx);
sl@0: 	b = BN_CTX_get(ctx);
sl@0: 	u = BN_CTX_get(ctx);
sl@0: 	v = BN_CTX_get(ctx);
sl@0: 	if (v == NULL) goto err;
sl@0: 
sl@0: 	/* reduce x and y mod p */
sl@0: 	if (!BN_GF2m_mod(u, y, p)) goto err;
sl@0: 	if (!BN_GF2m_mod(a, x, p)) goto err;
sl@0: 	if (!BN_copy(b, p)) goto err;
sl@0: 	
sl@0: 	while (!BN_is_odd(a))
sl@0: 		{
sl@0: 		if (!BN_rshift1(a, a)) goto err;
sl@0: 		if (BN_is_odd(u)) if (!BN_GF2m_add(u, u, p)) goto err;
sl@0: 		if (!BN_rshift1(u, u)) goto err;
sl@0: 		}
sl@0: 
sl@0: 	do
sl@0: 		{
sl@0: 		if (BN_GF2m_cmp(b, a) > 0)
sl@0: 			{
sl@0: 			if (!BN_GF2m_add(b, b, a)) goto err;
sl@0: 			if (!BN_GF2m_add(v, v, u)) goto err;
sl@0: 			do
sl@0: 				{
sl@0: 				if (!BN_rshift1(b, b)) goto err;
sl@0: 				if (BN_is_odd(v)) if (!BN_GF2m_add(v, v, p)) goto err;
sl@0: 				if (!BN_rshift1(v, v)) goto err;
sl@0: 				} while (!BN_is_odd(b));
sl@0: 			}
sl@0: 		else if (BN_abs_is_word(a, 1))
sl@0: 			break;
sl@0: 		else
sl@0: 			{
sl@0: 			if (!BN_GF2m_add(a, a, b)) goto err;
sl@0: 			if (!BN_GF2m_add(u, u, v)) goto err;
sl@0: 			do
sl@0: 				{
sl@0: 				if (!BN_rshift1(a, a)) goto err;
sl@0: 				if (BN_is_odd(u)) if (!BN_GF2m_add(u, u, p)) goto err;
sl@0: 				if (!BN_rshift1(u, u)) goto err;
sl@0: 				} while (!BN_is_odd(a));
sl@0: 			}
sl@0: 		} while (1);
sl@0: 
sl@0: 	if (!BN_copy(r, u)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 	ret = 1;
sl@0: 
sl@0: err:
sl@0:   	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: #endif
sl@0: 
sl@0: /* Divide yy by xx, reduce modulo p, and store the result in r. r could be xx 
sl@0:  * or yy, xx could equal yy.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_div implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_div function.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_div_arr(BIGNUM *r, const BIGNUM *yy, const BIGNUM *xx, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	BIGNUM *field;
sl@0: 	int ret = 0;
sl@0: 
sl@0: 	bn_check_top(yy);
sl@0: 	bn_check_top(xx);
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((field = BN_CTX_get(ctx)) == NULL) goto err;
sl@0: 	if (!BN_GF2m_arr2poly(p, field)) goto err;
sl@0: 	
sl@0: 	ret = BN_GF2m_mod_div(r, yy, xx, field, ctx);
sl@0: 	bn_check_top(r);
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: 
sl@0: /* Compute the bth power of a, reduce modulo p, and store
sl@0:  * the result in r.  r could be a.
sl@0:  * Uses simple square-and-multiply algorithm A.5.1 from IEEE P1363.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod_exp_arr(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0, i, n;
sl@0: 	BIGNUM *u;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(b);
sl@0: 
sl@0: 	if (BN_is_zero(b))
sl@0: 		return(BN_one(r));
sl@0: 
sl@0: 	if (BN_abs_is_word(b, 1))
sl@0: 		return (BN_copy(r, a) != NULL);
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((u = BN_CTX_get(ctx)) == NULL) goto err;
sl@0: 	
sl@0: 	if (!BN_GF2m_mod_arr(u, a, p)) goto err;
sl@0: 	
sl@0: 	n = BN_num_bits(b) - 1;
sl@0: 	for (i = n - 1; i >= 0; i--)
sl@0: 		{
sl@0: 		if (!BN_GF2m_mod_sqr_arr(u, u, p, ctx)) goto err;
sl@0: 		if (BN_is_bit_set(b, i))
sl@0: 			{
sl@0: 			if (!BN_GF2m_mod_mul_arr(u, u, a, p, ctx)) goto err;
sl@0: 			}
sl@0: 		}
sl@0: 	if (!BN_copy(r, u)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 	ret = 1;
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Compute the bth power of a, reduce modulo p, and store
sl@0:  * the result in r.  r could be a.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_exp_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_exp_arr function.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_exp(BIGNUM *r, const BIGNUM *a, const BIGNUM *b, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(b);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) * max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_EXP,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_exp_arr(r, a, b, arr, ctx);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Compute the square root of a, reduce modulo p, and store
sl@0:  * the result in r.  r could be a.
sl@0:  * Uses exponentiation as in algorithm A.4.1 from IEEE P1363.
sl@0:  */
sl@0: EXPORT_C int	BN_GF2m_mod_sqrt_arr(BIGNUM *r, const BIGNUM *a, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	BIGNUM *u;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 
sl@0: 	if (!p[0])
sl@0: 		{
sl@0: 		/* reduction mod 1 => return 0 */
sl@0: 		BN_zero(r);
sl@0: 		return 1;
sl@0: 		}
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	if ((u = BN_CTX_get(ctx)) == NULL) goto err;
sl@0: 	
sl@0: 	if (!BN_set_bit(u, p[0] - 1)) goto err;
sl@0: 	ret = BN_GF2m_mod_exp_arr(r, a, u, p, ctx);
sl@0: 	bn_check_top(r);
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Compute the square root of a, reduce modulo p, and store
sl@0:  * the result in r.  r could be a.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_sqrt_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_sqrt_arr function.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_sqrt(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) * max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_SQRT,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_sqrt_arr(r, a, arr, ctx);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Find r such that r^2 + r = a mod p.  r could be a. If no r exists returns 0.
sl@0:  * Uses algorithms A.4.7 and A.4.6 from IEEE P1363.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_solve_quad_arr(BIGNUM *r, const BIGNUM *a_, const unsigned int p[], BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0, count = 0;
sl@0: 	unsigned int j;
sl@0: 	BIGNUM *a, *z, *rho, *w, *w2, *tmp;
sl@0: 
sl@0: 	bn_check_top(a_);
sl@0: 
sl@0: 	if (!p[0])
sl@0: 		{
sl@0: 		/* reduction mod 1 => return 0 */
sl@0: 		BN_zero(r);
sl@0: 		return 1;
sl@0: 		}
sl@0: 
sl@0: 	BN_CTX_start(ctx);
sl@0: 	a = BN_CTX_get(ctx);
sl@0: 	z = BN_CTX_get(ctx);
sl@0: 	w = BN_CTX_get(ctx);
sl@0: 	if (w == NULL) goto err;
sl@0: 
sl@0: 	if (!BN_GF2m_mod_arr(a, a_, p)) goto err;
sl@0: 	
sl@0: 	if (BN_is_zero(a))
sl@0: 		{
sl@0: 		BN_zero(r);
sl@0: 		ret = 1;
sl@0: 		goto err;
sl@0: 		}
sl@0: 
sl@0: 	if (p[0] & 0x1) /* m is odd */
sl@0: 		{
sl@0: 		/* compute half-trace of a */
sl@0: 		if (!BN_copy(z, a)) goto err;
sl@0: 		for (j = 1; j <= (p[0] - 1) / 2; j++)
sl@0: 			{
sl@0: 			if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx)) goto err;
sl@0: 			if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx)) goto err;
sl@0: 			if (!BN_GF2m_add(z, z, a)) goto err;
sl@0: 			}
sl@0: 		
sl@0: 		}
sl@0: 	else /* m is even */
sl@0: 		{
sl@0: 		rho = BN_CTX_get(ctx);
sl@0: 		w2 = BN_CTX_get(ctx);
sl@0: 		tmp = BN_CTX_get(ctx);
sl@0: 		if (tmp == NULL) goto err;
sl@0: 		do
sl@0: 			{
sl@0: 			if (!BN_rand(rho, p[0], 0, 0)) goto err;
sl@0: 			if (!BN_GF2m_mod_arr(rho, rho, p)) goto err;
sl@0: 			BN_zero(z);
sl@0: 			if (!BN_copy(w, rho)) goto err;
sl@0: 			for (j = 1; j <= p[0] - 1; j++)
sl@0: 				{
sl@0: 				if (!BN_GF2m_mod_sqr_arr(z, z, p, ctx)) goto err;
sl@0: 				if (!BN_GF2m_mod_sqr_arr(w2, w, p, ctx)) goto err;
sl@0: 				if (!BN_GF2m_mod_mul_arr(tmp, w2, a, p, ctx)) goto err;
sl@0: 				if (!BN_GF2m_add(z, z, tmp)) goto err;
sl@0: 				if (!BN_GF2m_add(w, w2, rho)) goto err;
sl@0: 				}
sl@0: 			count++;
sl@0: 			} while (BN_is_zero(w) && (count < MAX_ITERATIONS));
sl@0: 		if (BN_is_zero(w))
sl@0: 			{
sl@0: 			BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD_ARR,BN_R_TOO_MANY_ITERATIONS);
sl@0: 			goto err;
sl@0: 			}
sl@0: 		}
sl@0: 	
sl@0: 	if (!BN_GF2m_mod_sqr_arr(w, z, p, ctx)) goto err;
sl@0: 	if (!BN_GF2m_add(w, z, w)) goto err;
sl@0: 	if (BN_GF2m_cmp(w, a))
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD_ARR, BN_R_NO_SOLUTION);
sl@0: 		goto err;
sl@0: 		}
sl@0: 
sl@0: 	if (!BN_copy(r, z)) goto err;
sl@0: 	bn_check_top(r);
sl@0: 
sl@0: 	ret = 1;
sl@0: 
sl@0: err:
sl@0: 	BN_CTX_end(ctx);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Find r such that r^2 + r = a mod p.  r could be a. If no r exists returns 0.
sl@0:  *
sl@0:  * This function calls down to the BN_GF2m_mod_solve_quad_arr implementation; this wrapper
sl@0:  * function is only provided for convenience; for best performance, use the 
sl@0:  * BN_GF2m_mod_solve_quad_arr function.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_mod_solve_quad(BIGNUM *r, const BIGNUM *a, const BIGNUM *p, BN_CTX *ctx)
sl@0: 	{
sl@0: 	int ret = 0;
sl@0: 	const int max = BN_num_bits(p);
sl@0: 	unsigned int *arr=NULL;
sl@0: 	bn_check_top(a);
sl@0: 	bn_check_top(p);
sl@0: 	if ((arr = (unsigned int *)OPENSSL_malloc(sizeof(unsigned int) *
sl@0: 						max)) == NULL) goto err;
sl@0: 	ret = BN_GF2m_poly2arr(p, arr, max);
sl@0: 	if (!ret || ret > max)
sl@0: 		{
sl@0: 		BNerr(BN_F_BN_GF2M_MOD_SOLVE_QUAD,BN_R_INVALID_LENGTH);
sl@0: 		goto err;
sl@0: 		}
sl@0: 	ret = BN_GF2m_mod_solve_quad_arr(r, a, arr, ctx);
sl@0: 	bn_check_top(r);
sl@0: err:
sl@0: 	if (arr) OPENSSL_free(arr);
sl@0: 	return ret;
sl@0: 	}
sl@0: 
sl@0: /* Convert the bit-string representation of a polynomial
sl@0:  * ( \sum_{i=0}^n a_i * x^i , where a_0 is *not* zero) into an array
sl@0:  * of integers corresponding to the bits with non-zero coefficient.
sl@0:  * Up to max elements of the array will be filled.  Return value is total
sl@0:  * number of coefficients that would be extracted if array was large enough.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_poly2arr(const BIGNUM *a, unsigned int p[], int max)
sl@0: 	{
sl@0: 	int i, j, k = 0;
sl@0: 	BN_ULONG mask;
sl@0: 
sl@0: 	if (BN_is_zero(a) || !BN_is_bit_set(a, 0))
sl@0: 		/* a_0 == 0 => return error (the unsigned int array
sl@0: 		 * must be terminated by 0)
sl@0: 		 */
sl@0: 		return 0;
sl@0: 
sl@0: 	for (i = a->top - 1; i >= 0; i--)
sl@0: 		{
sl@0: 		if (!a->d[i])
sl@0: 			/* skip word if a->d[i] == 0 */
sl@0: 			continue;
sl@0: 		mask = BN_TBIT;
sl@0: 		for (j = BN_BITS2 - 1; j >= 0; j--)
sl@0: 			{
sl@0: 			if (a->d[i] & mask) 
sl@0: 				{
sl@0: 				if (k < max) p[k] = BN_BITS2 * i + j;
sl@0: 				k++;
sl@0: 				}
sl@0: 			mask >>= 1;
sl@0: 			}
sl@0: 		}
sl@0: 
sl@0: 	return k;
sl@0: 	}
sl@0: 
sl@0: /* Convert the coefficient array representation of a polynomial to a 
sl@0:  * bit-string.  The array must be terminated by 0.
sl@0:  */
sl@0: EXPORT_C int BN_GF2m_arr2poly(const unsigned int p[], BIGNUM *a)
sl@0: 	{
sl@0: 	int i;
sl@0: 
sl@0: 	bn_check_top(a);
sl@0: 	BN_zero(a);
sl@0: 	for (i = 0; p[i] != 0; i++)
sl@0: 		{
sl@0:   		if (BN_set_bit(a, p[i]) == 0)
sl@0: 			return 0;
sl@0: 
sl@0: 		}
sl@0: 	BN_set_bit(a, 0);
sl@0: 	bn_check_top(a);
sl@0: 
sl@0: 	return 1;
sl@0: 	}
sl@0: