X-Git-Url: https://git.openssl.org/gitweb/?p=openssl.git;a=blobdiff_plain;f=crypto%2Fec%2Fecp_nistz256.c;h=6e4ca51017c5138cd75f2ad2f1e18ed404f55b8d;hp=14cf868f6640ab5663c91370ba3584a8323c4e3c;hb=f3b3d7f0033080f86ede5a53e8af2fb313091b5a;hpb=5784a52145d0062d42724d4d0fab3b4c82de35ee diff --git a/crypto/ec/ecp_nistz256.c b/crypto/ec/ecp_nistz256.c index 14cf868f66..6e4ca51017 100644 --- a/crypto/ec/ecp_nistz256.c +++ b/crypto/ec/ecp_nistz256.c @@ -1,3 +1,12 @@ +/* + * Copyright 2014-2016 The OpenSSL Project Authors. All Rights Reserved. + * + * Licensed under the OpenSSL license (the "License"). You may not use + * this file except in compliance with the License. You can obtain a copy + * in the file LICENSE in the source distribution or at + * https://www.openssl.org/source/license.html + */ + /****************************************************************************** * * * Copyright 2014 Intel Corporation * @@ -28,29 +37,26 @@ #include +#include "internal/cryptlib.h" #include "internal/bn_int.h" -#include -#include -#include "cryptlib.h" - #include "ec_lcl.h" #if BN_BITS2 != 64 -# define TOBN(hi,lo) lo,hi +# define TOBN(hi,lo) lo,hi #else -# define TOBN(hi,lo) ((BN_ULONG)hi<<32|lo) +# define TOBN(hi,lo) ((BN_ULONG)hi<<32|lo) #endif #if defined(__GNUC__) -# define ALIGN32 __attribute((aligned(32))) +# define ALIGN32 __attribute((aligned(32))) #elif defined(_MSC_VER) -# define ALIGN32 __declspec(align(32)) +# define ALIGN32 __declspec(align(32)) #else # define ALIGN32 #endif -#define ALIGNPTR(p,N) ((unsigned char *)p+N-(size_t)p%N) -#define P256_LIMBS (256/BN_BITS2) +#define ALIGNPTR(p,N) ((unsigned char *)p+N-(size_t)p%N) +#define P256_LIMBS (256/BN_BITS2) typedef unsigned short u16; @@ -68,36 +74,56 @@ typedef struct { typedef P256_POINT_AFFINE PRECOMP256_ROW[64]; /* structure for precomputed multiples of the generator */ -typedef struct ec_pre_comp_st { +struct nistz256_pre_comp_st { const EC_GROUP *group; /* Parent EC_GROUP object */ size_t w; /* Window size */ - /* Constant time access to the X and Y coordinates of the pre-computed, + /* + * Constant time access to the X and Y coordinates of the pre-computed, * generator multiplies, in the Montgomery domain. Pre-calculated - * multiplies are stored in affine form. */ + * multiplies are stored in affine form. + */ PRECOMP256_ROW *precomp; void *precomp_storage; int references; -} EC_PRE_COMP; + CRYPTO_RWLOCK *lock; +}; /* Functions implemented in assembly */ +/* + * Most of below mentioned functions *preserve* the property of inputs + * being fully reduced, i.e. being in [0, modulus) range. Simply put if + * inputs are fully reduced, then output is too. Note that reverse is + * not true, in sense that given partially reduced inputs output can be + * either, not unlikely reduced. And "most" in first sentence refers to + * the fact that given the calculations flow one can tolerate that + * addition, 1st function below, produces partially reduced result *if* + * multiplications by 2 and 3, which customarily use addition, fully + * reduce it. This effectively gives two options: a) addition produces + * fully reduced result [as long as inputs are, just like remaining + * functions]; b) addition is allowed to produce partially reduced + * result, but multiplications by 2 and 3 perform additional reduction + * step. Choice between the two can be platform-specific, but it was a) + * in all cases so far... + */ +/* Modular add: res = a+b mod P */ +void ecp_nistz256_add(BN_ULONG res[P256_LIMBS], + const BN_ULONG a[P256_LIMBS], + const BN_ULONG b[P256_LIMBS]); /* Modular mul by 2: res = 2*a mod P */ void ecp_nistz256_mul_by_2(BN_ULONG res[P256_LIMBS], const BN_ULONG a[P256_LIMBS]); -/* Modular div by 2: res = a/2 mod P */ -void ecp_nistz256_div_by_2(BN_ULONG res[P256_LIMBS], - const BN_ULONG a[P256_LIMBS]); /* Modular mul by 3: res = 3*a mod P */ void ecp_nistz256_mul_by_3(BN_ULONG res[P256_LIMBS], const BN_ULONG a[P256_LIMBS]); -/* Modular add: res = a+b mod P */ -void ecp_nistz256_add(BN_ULONG res[P256_LIMBS], - const BN_ULONG a[P256_LIMBS], - const BN_ULONG b[P256_LIMBS]); -/* Modular sub: res = a-b mod P */ + +/* Modular div by 2: res = a/2 mod P */ +void ecp_nistz256_div_by_2(BN_ULONG res[P256_LIMBS], + const BN_ULONG a[P256_LIMBS]); +/* Modular sub: res = a-b mod P */ void ecp_nistz256_sub(BN_ULONG res[P256_LIMBS], const BN_ULONG a[P256_LIMBS], const BN_ULONG b[P256_LIMBS]); -/* Modular neg: res = -a mod P */ +/* Modular neg: res = -a mod P */ void ecp_nistz256_neg(BN_ULONG res[P256_LIMBS], const BN_ULONG a[P256_LIMBS]); /* Montgomery mul: res = a*b*2^-256 mod P */ void ecp_nistz256_mul_mont(BN_ULONG res[P256_LIMBS], @@ -113,14 +139,14 @@ void ecp_nistz256_from_mont(BN_ULONG res[P256_LIMBS], void ecp_nistz256_to_mont(BN_ULONG res[P256_LIMBS], const BN_ULONG in[P256_LIMBS]); /* Functions that perform constant time access to the precomputed tables */ -void ecp_nistz256_scatter_w5(P256_POINT * val, - const P256_POINT * in_t, int index); -void ecp_nistz256_gather_w5(P256_POINT * val, - const P256_POINT * in_t, int index); -void ecp_nistz256_scatter_w7(P256_POINT_AFFINE * val, - const P256_POINT_AFFINE * in_t, int index); -void ecp_nistz256_gather_w7(P256_POINT_AFFINE * val, - const P256_POINT_AFFINE * in_t, int index); +void ecp_nistz256_scatter_w5(P256_POINT *val, + const P256_POINT *in_t, int idx); +void ecp_nistz256_gather_w5(P256_POINT *val, + const P256_POINT *in_t, int idx); +void ecp_nistz256_scatter_w7(P256_POINT_AFFINE *val, + const P256_POINT_AFFINE *in_t, int idx); +void ecp_nistz256_gather_w7(P256_POINT_AFFINE *val, + const P256_POINT_AFFINE *in_t, int idx); /* One converted into the Montgomery domain */ static const BN_ULONG ONE[P256_LIMBS] = { @@ -128,10 +154,7 @@ static const BN_ULONG ONE[P256_LIMBS] = { TOBN(0xffffffff, 0xffffffff), TOBN(0x00000000, 0xfffffffe) }; -static void *ecp_nistz256_pre_comp_dup(void *); -static void ecp_nistz256_pre_comp_free(void *); -static void ecp_nistz256_pre_comp_clear_free(void *); -static EC_PRE_COMP *ecp_nistz256_pre_comp_new(const EC_GROUP * group); +static NISTZ256_PRE_COMP *ecp_nistz256_pre_comp_new(const EC_GROUP *group); /* Precomputed tables for the default generator */ extern const PRECOMP256_ROW ecp_nistz256_precomputed[37]; @@ -164,7 +187,7 @@ static unsigned int _booth_recode_w7(unsigned int in) static void copy_conditional(BN_ULONG dst[P256_LIMBS], const BN_ULONG src[P256_LIMBS], BN_ULONG move) { - BN_ULONG mask1 = -move; + BN_ULONG mask1 = 0-move; BN_ULONG mask2 = ~mask1; dst[0] = (src[0] & mask1) ^ (dst[0] & mask2); @@ -183,7 +206,6 @@ static BN_ULONG is_zero(BN_ULONG in) { in |= (0 - in); in = ~in; - in &= BN_MASK2; in >>= BN_BITS2 - 1; return in; } @@ -207,33 +229,51 @@ static BN_ULONG is_equal(const BN_ULONG a[P256_LIMBS], return is_zero(res); } -static BN_ULONG is_one(const BN_ULONG a[P256_LIMBS]) +static BN_ULONG is_one(const BIGNUM *z) { - BN_ULONG res; - - res = a[0] ^ ONE[0]; - res |= a[1] ^ ONE[1]; - res |= a[2] ^ ONE[2]; - res |= a[3] ^ ONE[3]; - if (P256_LIMBS == 8) { - res |= a[4] ^ ONE[4]; - res |= a[5] ^ ONE[5]; - res |= a[6] ^ ONE[6]; + BN_ULONG res = 0; + BN_ULONG *a = bn_get_words(z); + + if (bn_get_top(z) == (P256_LIMBS - P256_LIMBS / 8)) { + res = a[0] ^ ONE[0]; + res |= a[1] ^ ONE[1]; + res |= a[2] ^ ONE[2]; + res |= a[3] ^ ONE[3]; + if (P256_LIMBS == 8) { + res |= a[4] ^ ONE[4]; + res |= a[5] ^ ONE[5]; + res |= a[6] ^ ONE[6]; + /* + * no check for a[7] (being zero) on 32-bit platforms, + * because value of "one" takes only 7 limbs. + */ + } + res = is_zero(res); } - return is_zero(res); + return res; } +/* + * For reference, this macro is used only when new ecp_nistz256 assembly + * module is being developed. For example, configure with + * -DECP_NISTZ256_REFERENCE_IMPLEMENTATION and implement only functions + * performing simplest arithmetic operations on 256-bit vectors. Then + * work on implementation of higher-level functions performing point + * operations. Then remove ECP_NISTZ256_REFERENCE_IMPLEMENTATION + * and never define it again. (The correct macro denoting presence of + * ecp_nistz256 module is ECP_NISTZ256_ASM.) + */ #ifndef ECP_NISTZ256_REFERENCE_IMPLEMENTATION -void ecp_nistz256_point_double(P256_POINT * r, const P256_POINT * a); -void ecp_nistz256_point_add(P256_POINT * r, - const P256_POINT * a, const P256_POINT * b); -void ecp_nistz256_point_add_affine(P256_POINT * r, - const P256_POINT * a, - const P256_POINT_AFFINE * b); +void ecp_nistz256_point_double(P256_POINT *r, const P256_POINT *a); +void ecp_nistz256_point_add(P256_POINT *r, + const P256_POINT *a, const P256_POINT *b); +void ecp_nistz256_point_add_affine(P256_POINT *r, + const P256_POINT *a, + const P256_POINT_AFFINE *b); #else /* Point double: r = 2*a */ -static void ecp_nistz256_point_double(P256_POINT * r, const P256_POINT * a) +static void ecp_nistz256_point_double(P256_POINT *r, const P256_POINT *a) { BN_ULONG S[P256_LIMBS]; BN_ULONG M[P256_LIMBS]; @@ -279,8 +319,8 @@ static void ecp_nistz256_point_double(P256_POINT * r, const P256_POINT * a) } /* Point addition: r = a+b */ -static void ecp_nistz256_point_add(P256_POINT * r, - const P256_POINT * a, const P256_POINT * b) +static void ecp_nistz256_point_add(P256_POINT *r, + const P256_POINT *a, const P256_POINT *b) { BN_ULONG U2[P256_LIMBS], S2[P256_LIMBS]; BN_ULONG U1[P256_LIMBS], S1[P256_LIMBS]; @@ -305,19 +345,16 @@ static void ecp_nistz256_point_add(P256_POINT * r, const BN_ULONG *in2_y = b->Y; const BN_ULONG *in2_z = b->Z; - /* We encode infinity as (0,0), which is not on the curve, - * so it is OK. */ - in1infty = in1_x[0] | in1_x[1] | in1_x[2] | in1_x[3] | - in1_y[0] | in1_y[1] | in1_y[2] | in1_y[3]; + /* + * Infinity in encoded as (,,0) + */ + in1infty = (in1_z[0] | in1_z[1] | in1_z[2] | in1_z[3]); if (P256_LIMBS == 8) - in1infty |= in1_x[4] | in1_x[5] | in1_x[6] | in1_x[7] | - in1_y[4] | in1_y[5] | in1_y[6] | in1_y[7]; + in1infty |= (in1_z[4] | in1_z[5] | in1_z[6] | in1_z[7]); - in2infty = in2_x[0] | in2_x[1] | in2_x[2] | in2_x[3] | - in2_y[0] | in2_y[1] | in2_y[2] | in2_y[3]; + in2infty = (in2_z[0] | in2_z[1] | in2_z[2] | in2_z[3]); if (P256_LIMBS == 8) - in2infty |= in2_x[4] | in2_x[5] | in2_x[6] | in2_x[7] | - in2_y[4] | in2_y[5] | in2_y[6] | in2_y[7]; + in2infty |= (in2_z[4] | in2_z[5] | in2_z[6] | in2_z[7]); in1infty = is_zero(in1infty); in2infty = is_zero(in2infty); @@ -336,8 +373,9 @@ static void ecp_nistz256_point_add(P256_POINT * r, ecp_nistz256_mul_mont(U2, in2_x, Z1sqr); /* U2 = X2*Z1^2 */ ecp_nistz256_sub(H, U2, U1); /* H = U2 - U1 */ - /* This should not happen during sign/ecdh, - * so no constant time violation */ + /* + * This should not happen during sign/ecdh, so no constant time violation + */ if (is_equal(U1, U2) && !in1infty && !in2infty) { if (is_equal(S1, S2)) { ecp_nistz256_point_double(r, a); @@ -380,9 +418,9 @@ static void ecp_nistz256_point_add(P256_POINT * r, } /* Point addition when b is known to be affine: r = a+b */ -static void ecp_nistz256_point_add_affine(P256_POINT * r, - const P256_POINT * a, - const P256_POINT_AFFINE * b) +static void ecp_nistz256_point_add_affine(P256_POINT *r, + const P256_POINT *a, + const P256_POINT_AFFINE *b) { BN_ULONG U2[P256_LIMBS], S2[P256_LIMBS]; BN_ULONG Z1sqr[P256_LIMBS]; @@ -404,19 +442,22 @@ static void ecp_nistz256_point_add_affine(P256_POINT * r, const BN_ULONG *in2_x = b->X; const BN_ULONG *in2_y = b->Y; - /* In affine representation we encode infty as (0,0), - * which is not on the curve, so it is OK */ - in1infty = in1_x[0] | in1_x[1] | in1_x[2] | in1_x[3] | - in1_y[0] | in1_y[1] | in1_y[2] | in1_y[3]; + /* + * Infinity in encoded as (,,0) + */ + in1infty = (in1_z[0] | in1_z[1] | in1_z[2] | in1_z[3]); if (P256_LIMBS == 8) - in1infty |= in1_x[4] | in1_x[5] | in1_x[6] | in1_x[7] | - in1_y[4] | in1_y[5] | in1_y[6] | in1_y[7]; + in1infty |= (in1_z[4] | in1_z[5] | in1_z[6] | in1_z[7]); - in2infty = in2_x[0] | in2_x[1] | in2_x[2] | in2_x[3] | - in2_y[0] | in2_y[1] | in2_y[2] | in2_y[3]; + /* + * In affine representation we encode infinity as (0,0), which is + * not on the curve, so it is OK + */ + in2infty = (in2_x[0] | in2_x[1] | in2_x[2] | in2_x[3] | + in2_y[0] | in2_y[1] | in2_y[2] | in2_y[3]); if (P256_LIMBS == 8) - in2infty |= in2_x[4] | in2_x[5] | in2_x[6] | in2_x[7] | - in2_y[4] | in2_y[5] | in2_y[6] | in2_y[7]; + in2infty |= (in2_x[4] | in2_x[5] | in2_x[6] | in2_x[7] | + in2_y[4] | in2_y[5] | in2_y[6] | in2_y[7]); in1infty = is_zero(in1infty); in2infty = is_zero(in2infty); @@ -467,8 +508,10 @@ static void ecp_nistz256_point_add_affine(P256_POINT * r, static void ecp_nistz256_mod_inverse(BN_ULONG r[P256_LIMBS], const BN_ULONG in[P256_LIMBS]) { - /* The poly is ffffffff 00000001 00000000 00000000 00000000 ffffffff ffffffff ffffffff - We use FLT and used poly-2 as exponent */ + /* + * The poly is ffffffff 00000001 00000000 00000000 00000000 ffffffff + * ffffffff ffffffff We use FLT and used poly-2 as exponent + */ BN_ULONG p2[P256_LIMBS]; BN_ULONG p4[P256_LIMBS]; BN_ULONG p8[P256_LIMBS]; @@ -538,34 +581,38 @@ static void ecp_nistz256_mod_inverse(BN_ULONG r[P256_LIMBS], memcpy(r, res, sizeof(res)); } -/* ecp_nistz256_bignum_to_field_elem copies the contents of |in| to |out| and - * returns one if it fits. Otherwise it returns zero. */ -static int ecp_nistz256_bignum_to_field_elem(BN_ULONG out[P256_LIMBS], - const BIGNUM * in) +/* + * ecp_nistz256_bignum_to_field_elem copies the contents of |in| to |out| and + * returns one if it fits. Otherwise it returns zero. + */ +__owur static int ecp_nistz256_bignum_to_field_elem(BN_ULONG out[P256_LIMBS], + const BIGNUM *in) { return bn_copy_words(out, in, P256_LIMBS); } /* r = sum(scalar[i]*point[i]) */ -static void ecp_nistz256_windowed_mul(const EC_GROUP * group, - P256_POINT * r, - const BIGNUM ** scalar, - const EC_POINT ** point, - int num, BN_CTX * ctx) +__owur static int ecp_nistz256_windowed_mul(const EC_GROUP *group, + P256_POINT *r, + const BIGNUM **scalar, + const EC_POINT **point, + size_t num, BN_CTX *ctx) { - int i, j; - unsigned int index; + size_t i; + int j, ret = 0; + unsigned int idx; unsigned char (*p_str)[33] = NULL; const unsigned int window_size = 5; const unsigned int mask = (1 << (window_size + 1)) - 1; unsigned int wvalue; - P256_POINT *temp; /* place for 5 temporary points */ + P256_POINT *temp; /* place for 5 temporary points */ const BIGNUM **scalars = NULL; - P256_POINT(*table)[16] = NULL; + P256_POINT (*table)[16] = NULL; void *table_storage = NULL; - if ((table_storage = - OPENSSL_malloc((num * 16 + 5) * sizeof(P256_POINT) + 64)) == NULL + if ((num * 16 + 6) > OPENSSL_MALLOC_MAX_NELEMS(P256_POINT) + || (table_storage = + OPENSSL_malloc((num * 16 + 5) * sizeof(P256_POINT) + 64)) == NULL || (p_str = OPENSSL_malloc(num * 33 * sizeof(unsigned char))) == NULL || (scalars = OPENSSL_malloc(num * sizeof(BIGNUM *))) == NULL) { @@ -574,11 +621,12 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, } table = (void *)ALIGNPTR(table_storage, 64); - temp = (P256_POINT *)(table + num); + temp = (P256_POINT *)(table + num); for (i = 0; i < num; i++) { P256_POINT *row = table[i]; + /* This is an unusual input, we don't guarantee constant-timeness. */ if ((BN_num_bits(scalar[i]) > 256) || BN_is_negative(scalar[i])) { BIGNUM *mod; @@ -595,16 +643,16 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, for (j = 0; j < bn_get_top(scalars[i]) * BN_BYTES; j += BN_BYTES) { BN_ULONG d = bn_get_words(scalars[i])[j / BN_BYTES]; - p_str[i][j + 0] = d & 0xff; - p_str[i][j + 1] = (d >> 8) & 0xff; - p_str[i][j + 2] = (d >> 16) & 0xff; - p_str[i][j + 3] = (d >>= 24) & 0xff; + p_str[i][j + 0] = (unsigned char)d; + p_str[i][j + 1] = (unsigned char)(d >> 8); + p_str[i][j + 2] = (unsigned char)(d >> 16); + p_str[i][j + 3] = (unsigned char)(d >>= 24); if (BN_BYTES == 8) { d >>= 8; - p_str[i][j + 4] = d & 0xff; - p_str[i][j + 5] = (d >> 8) & 0xff; - p_str[i][j + 6] = (d >> 16) & 0xff; - p_str[i][j + 7] = (d >> 24) & 0xff; + p_str[i][j + 4] = (unsigned char)d; + p_str[i][j + 5] = (unsigned char)(d >> 8); + p_str[i][j + 6] = (unsigned char)(d >> 16); + p_str[i][j + 7] = (unsigned char)(d >> 24); } } for (; j < 33; j++) @@ -613,13 +661,15 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, if (!ecp_nistz256_bignum_to_field_elem(temp[0].X, point[i]->X) || !ecp_nistz256_bignum_to_field_elem(temp[0].Y, point[i]->Y) || !ecp_nistz256_bignum_to_field_elem(temp[0].Z, point[i]->Z)) { - ECerr(EC_F_ECP_NISTZ256_WINDOWED_MUL, EC_R_COORDINATES_OUT_OF_RANGE); + ECerr(EC_F_ECP_NISTZ256_WINDOWED_MUL, + EC_R_COORDINATES_OUT_OF_RANGE); goto err; } - /* row[0] is implicitly (0,0,0) (the point at infinity), - * therefore it is not stored. All other values are actually - * stored with an offset of -1 in table. + /* + * row[0] is implicitly (0,0,0) (the point at infinity), therefore it + * is not stored. All other values are actually stored with an offset + * of -1 in table. */ ecp_nistz256_scatter_w5 (row, &temp[0], 1); @@ -655,10 +705,10 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, ecp_nistz256_scatter_w5 (row, &temp[1], 16); } - index = 255; + idx = 255; - wvalue = p_str[0][(index - 1) / 8]; - wvalue = (wvalue >> ((index - 1) % 8)) & mask; + wvalue = p_str[0][(idx - 1) / 8]; + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; /* * We gather to temp[0], because we know it's position relative @@ -667,12 +717,12 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, ecp_nistz256_gather_w5(&temp[0], table[0], _booth_recode_w5(wvalue) >> 1); memcpy(r, &temp[0], sizeof(temp[0])); - while (index >= 5) { - for (i = (index == 255 ? 1 : 0); i < num; i++) { - unsigned int off = (index - 1) / 8; + while (idx >= 5) { + for (i = (idx == 255 ? 1 : 0); i < num; i++) { + unsigned int off = (idx - 1) / 8; wvalue = p_str[i][off] | p_str[i][off + 1] << 8; - wvalue = (wvalue >> ((index - 1) % 8)) & mask; + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; wvalue = _booth_recode_w5(wvalue); @@ -684,7 +734,7 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, ecp_nistz256_point_add(r, r, &temp[0]); } - index -= window_size; + idx -= window_size; ecp_nistz256_point_double(r, r); ecp_nistz256_point_double(r, r); @@ -708,13 +758,12 @@ static void ecp_nistz256_windowed_mul(const EC_GROUP * group, ecp_nistz256_point_add(r, r, &temp[0]); } -err: - if (table_storage) - OPENSSL_free(table_storage); - if (p_str) - OPENSSL_free(p_str); - if (scalars) - OPENSSL_free(scalars); + ret = 1; + err: + OPENSSL_free(table_storage); + OPENSSL_free(p_str); + OPENSSL_free(scalars); + return ret; } /* Coordinates of G, for which we have precomputed tables */ @@ -728,39 +777,40 @@ const static BN_ULONG def_yG[P256_LIMBS] = { TOBN(0xd2e88688, 0xdd21f325), TOBN(0x8571ff18, 0x25885d85) }; -/* ecp_nistz256_is_affine_G returns one if |generator| is the standard, - * P-256 generator. */ -static int ecp_nistz256_is_affine_G(const EC_POINT * generator) +/* + * ecp_nistz256_is_affine_G returns one if |generator| is the standard, P-256 + * generator. + */ +static int ecp_nistz256_is_affine_G(const EC_POINT *generator) { return (bn_get_top(generator->X) == P256_LIMBS) && (bn_get_top(generator->Y) == P256_LIMBS) && - (bn_get_top(generator->Z) == (P256_LIMBS - P256_LIMBS / 8)) && is_equal(bn_get_words(generator->X), def_xG) && is_equal(bn_get_words(generator->Y), def_yG) && - is_one(bn_get_words(generator->Z)); + is_one(generator->Z); } -static int ecp_nistz256_mult_precompute(EC_GROUP * group, BN_CTX * ctx) +__owur static int ecp_nistz256_mult_precompute(EC_GROUP *group, BN_CTX *ctx) { - /* We precompute a table for a Booth encoded exponent (wNAF) based + /* + * We precompute a table for a Booth encoded exponent (wNAF) based * computation. Each table holds 64 values for safe access, with an - * implicit value of infinity at index zero. We use window of size 7, - * and therefore require ceil(256/7) = 37 tables. */ - BIGNUM *order; + * implicit value of infinity at index zero. We use window of size 7, and + * therefore require ceil(256/7) = 37 tables. + */ + const BIGNUM *order; EC_POINT *P = NULL, *T = NULL; const EC_POINT *generator; - EC_PRE_COMP *pre_comp; + NISTZ256_PRE_COMP *pre_comp; + BN_CTX *new_ctx = NULL; int i, j, k, ret = 0; size_t w; PRECOMP256_ROW *preComputedTable = NULL; unsigned char *precomp_storage = NULL; - /* if there is an old EC_PRE_COMP object, throw it away */ - EC_EX_DATA_free_data(&group->extra_data, ecp_nistz256_pre_comp_dup, - ecp_nistz256_pre_comp_free, - ecp_nistz256_pre_comp_clear_free); - + /* if there is an old NISTZ256_PRE_COMP object, throw it away */ + EC_pre_comp_free(group); generator = EC_GROUP_get0_generator(group); if (generator == NULL) { ECerr(EC_F_ECP_NISTZ256_MULT_PRECOMPUTE, EC_R_UNDEFINED_GENERATOR); @@ -768,8 +818,10 @@ static int ecp_nistz256_mult_precompute(EC_GROUP * group, BN_CTX * ctx) } if (ecp_nistz256_is_affine_G(generator)) { - /* No need to calculate tables for the standard generator - * because we have them statically. */ + /* + * No need to calculate tables for the standard generator because we + * have them statically. + */ return 1; } @@ -777,20 +829,17 @@ static int ecp_nistz256_mult_precompute(EC_GROUP * group, BN_CTX * ctx) return 0; if (ctx == NULL) { - ctx = BN_CTX_new(); + ctx = new_ctx = BN_CTX_new(); if (ctx == NULL) goto err; } BN_CTX_start(ctx); - order = BN_CTX_get(ctx); + order = EC_GROUP_get0_order(group); if (order == NULL) goto err; - if (!EC_GROUP_get_order(group, order, ctx)) - goto err; - if (BN_is_zero(order)) { ECerr(EC_F_ECP_NISTZ256_MULT_PRECOMPUTE, EC_R_UNKNOWN_ORDER); goto err; @@ -808,57 +857,61 @@ static int ecp_nistz256_mult_precompute(EC_GROUP * group, BN_CTX * ctx) P = EC_POINT_new(group); T = EC_POINT_new(group); + if (P == NULL || T == NULL) + goto err; - /* The zero entry is implicitly infinity, and we skip it, - * storing other values with -1 offset. */ - EC_POINT_copy(T, generator); + /* + * The zero entry is implicitly infinity, and we skip it, storing other + * values with -1 offset. + */ + if (!EC_POINT_copy(T, generator)) + goto err; for (k = 0; k < 64; k++) { - EC_POINT_copy(P, T); + if (!EC_POINT_copy(P, T)) + goto err; for (j = 0; j < 37; j++) { P256_POINT_AFFINE temp; - /* It would be faster to use - * ec_GFp_simple_points_make_affine and make multiple - * points affine at the same time. */ - ec_GFp_simple_make_affine(group, P, ctx); - ecp_nistz256_bignum_to_field_elem(temp.X, P->X); - ecp_nistz256_bignum_to_field_elem(temp.Y, P->Y); + /* + * It would be faster to use EC_POINTs_make_affine and + * make multiple points affine at the same time. + */ + if (!EC_POINT_make_affine(group, P, ctx)) + goto err; + if (!ecp_nistz256_bignum_to_field_elem(temp.X, P->X) || + !ecp_nistz256_bignum_to_field_elem(temp.Y, P->Y)) { + ECerr(EC_F_ECP_NISTZ256_MULT_PRECOMPUTE, + EC_R_COORDINATES_OUT_OF_RANGE); + goto err; + } ecp_nistz256_scatter_w7(preComputedTable[j], &temp, k); - for (i = 0; i < 7; i++) - ec_GFp_simple_dbl(group, P, P, ctx); + for (i = 0; i < 7; i++) { + if (!EC_POINT_dbl(group, P, P, ctx)) + goto err; + } } - ec_GFp_simple_add(group, T, T, generator, ctx); + if (!EC_POINT_add(group, T, T, generator, ctx)) + goto err; } pre_comp->group = group; pre_comp->w = w; pre_comp->precomp = preComputedTable; pre_comp->precomp_storage = precomp_storage; - precomp_storage = NULL; - - if (!EC_EX_DATA_set_data(&group->extra_data, pre_comp, - ecp_nistz256_pre_comp_dup, - ecp_nistz256_pre_comp_free, - ecp_nistz256_pre_comp_clear_free)) { - goto err; - } - + SETPRECOMP(group, nistz256, pre_comp); pre_comp = NULL; - ret = 1; -err: + err: if (ctx != NULL) BN_CTX_end(ctx); - if (pre_comp) - ecp_nistz256_pre_comp_free(pre_comp); - if (precomp_storage) - OPENSSL_free(precomp_storage); - if (P) - EC_POINT_free(P); - if (T) - EC_POINT_free(T); + BN_CTX_free(new_ctx); + + EC_nistz256_pre_comp_free(pre_comp); + OPENSSL_free(precomp_storage); + EC_POINT_free(P); + EC_POINT_free(T); return ret; } @@ -878,8 +931,9 @@ err: # undef ECP_NISTZ256_AVX2 # else /* Constant time access, loading four values, from four consecutive tables */ -void ecp_nistz256_avx2_multi_gather_w7(void *result, const void *in, int index0, - int index1, int index2, int index3); +void ecp_nistz256_avx2_multi_gather_w7(void *result, const void *in, + int index0, int index1, int index2, + int index3); void ecp_nistz256_avx2_transpose_convert(void *RESULTx4, const void *in); void ecp_nistz256_avx2_convert_transpose_back(void *result, const void *Ax4); void ecp_nistz256_avx2_point_add_affine_x4(void *RESULTx4, const void *Ax4, @@ -905,13 +959,14 @@ static void booth_recode_w7(unsigned char *sign, *digit = d; } -/* ecp_nistz256_avx2_mul_g performs multiplication by G, using only the +/* + * ecp_nistz256_avx2_mul_g performs multiplication by G, using only the * precomputed table. It does 4 affine point additions in parallel, - * significantly speeding up point multiplication for a fixed value. */ -static void ecp_nistz256_avx2_mul_g(P256_POINT * r, + * significantly speeding up point multiplication for a fixed value. + */ +static void ecp_nistz256_avx2_mul_g(P256_POINT *r, unsigned char p_str[33], - const - P256_POINT_AFFINE(*preComputedTable)[64]) + const P256_POINT_AFFINE(*preComputedTable)[64]) { const unsigned int window_size = 7; const unsigned int mask = (1 << (window_size + 1)) - 1; @@ -921,7 +976,7 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, unsigned char sign1, digit1; unsigned char sign2, digit2; unsigned char sign3, digit3; - unsigned int index = 0; + unsigned int idx = 0; BN_ULONG tmp[P256_LIMBS]; int i; @@ -933,19 +988,19 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, /* Initial four windows */ wvalue = *((u16 *) & p_str[0]); wvalue = (wvalue << 1) & mask; - index += window_size; + idx += window_size; booth_recode_w7(&sign0, &digit0, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign1, &digit1, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign2, &digit2, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign3, &digit3, wvalue); ecp_nistz256_avx2_multi_gather_w7(point_arr, preComputedTable[0], @@ -965,21 +1020,21 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, ecp_nistz256_avx2_to_mont(&aX4[4 * 9], &aX4[4 * 9]); ecp_nistz256_avx2_set1(&aX4[4 * 9 * 2]); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign0, &digit0, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign1, &digit1, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign2, &digit2, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign3, &digit3, wvalue); ecp_nistz256_avx2_multi_gather_w7(point_arr, preComputedTable[4 * 1], @@ -1001,21 +1056,21 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, ecp_nistz256_avx2_point_add_affines_x4(aX4, aX4, bX4); for (i = 2; i < 9; i++) { - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign0, &digit0, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign1, &digit1, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign2, &digit2, wvalue); - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; booth_recode_w7(&sign3, &digit3, wvalue); ecp_nistz256_avx2_multi_gather_w7(point_arr, @@ -1044,11 +1099,11 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, ecp_nistz256_avx2_convert_transpose_back(res_point_arr, aX4); /* Last window is performed serially */ - wvalue = *((u16 *) & p_str[(index - 1) / 8]); - wvalue = (wvalue >> ((index - 1) % 8)) & mask; + wvalue = *((u16 *) & p_str[(idx - 1) / 8]); + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; booth_recode_w7(&sign0, &digit0, wvalue); - ecp_nistz256_gather_w7((P256_POINT_AFFINE *) r, - preComputedTable[36], digit0); + ecp_nistz256_gather_w7((P256_POINT_AFFINE *)r, + preComputedTable[36], digit0); ecp_nistz256_neg(tmp, r->Y); copy_conditional(r->Y, tmp, sign0); memcpy(r->Z, ONE, sizeof(ONE)); @@ -1061,19 +1116,19 @@ static void ecp_nistz256_avx2_mul_g(P256_POINT * r, # endif #endif -static int ecp_nistz256_set_from_affine(EC_POINT * out, const EC_GROUP * group, - const P256_POINT_AFFINE * in, - BN_CTX * ctx) +__owur static int ecp_nistz256_set_from_affine(EC_POINT *out, const EC_GROUP *group, + const P256_POINT_AFFINE *in, + BN_CTX *ctx) { BIGNUM *x, *y; BN_ULONG d_x[P256_LIMBS], d_y[P256_LIMBS]; int ret = 0; x = BN_new(); - if(!x) + if (x == NULL) return 0; y = BN_new(); - if(!y) { + if (y == NULL) { BN_free(x); return 0; } @@ -1085,27 +1140,30 @@ static int ecp_nistz256_set_from_affine(EC_POINT * out, const EC_GROUP * group, ret = EC_POINT_set_affine_coordinates_GFp(group, out, x, y, ctx); - if(x) BN_free(x); - if(y) BN_free(y); + BN_free(x); + BN_free(y); return ret; } /* r = scalar*G + sum(scalars[i]*points[i]) */ -static int ecp_nistz256_points_mul(const EC_GROUP * group, - EC_POINT * r, - const BIGNUM * scalar, - size_t num, - const EC_POINT * points[], - const BIGNUM * scalars[], BN_CTX * ctx) +__owur static int ecp_nistz256_points_mul(const EC_GROUP *group, + EC_POINT *r, + const BIGNUM *scalar, + size_t num, + const EC_POINT *points[], + const BIGNUM *scalars[], BN_CTX *ctx) { int i = 0, ret = 0, no_precomp_for_generator = 0, p_is_infinity = 0; size_t j; unsigned char p_str[33] = { 0 }; const PRECOMP256_ROW *preComputedTable = NULL; - const EC_PRE_COMP *pre_comp = NULL; + const NISTZ256_PRE_COMP *pre_comp = NULL; const EC_POINT *generator = NULL; - unsigned int index = 0; + BN_CTX *new_ctx = NULL; + const BIGNUM **new_scalars = NULL; + const EC_POINT **new_points = NULL; + unsigned int idx = 0; const unsigned int window_size = 7; const unsigned int mask = (1 << (window_size + 1)) - 1; unsigned int wvalue; @@ -1115,7 +1173,7 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, } t, p; BIGNUM *tmp_scalar; - if ((num+1) == 0 || (num+1) > OPENSSL_MALLOC_MAX_NELEMS(void *)) { + if ((num + 1) == 0 || (num + 1) > OPENSSL_MALLOC_MAX_NELEMS(void *)) { ECerr(EC_F_ECP_NISTZ256_POINTS_MUL, ERR_R_MALLOC_FAILURE); return 0; } @@ -1124,6 +1182,7 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, ECerr(EC_F_ECP_NISTZ256_POINTS_MUL, EC_R_INCOMPATIBLE_OBJECTS); return 0; } + if ((scalar == NULL) && (num == 0)) return EC_POINT_set_to_infinity(group, r); @@ -1134,13 +1193,13 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, } } - /* Need 256 bits for space for all coordinates. */ - bn_wexpand(r->X, P256_LIMBS); - bn_wexpand(r->Y, P256_LIMBS); - bn_wexpand(r->Z, P256_LIMBS); - bn_set_top(r->X, P256_LIMBS); - bn_set_top(r->Y, P256_LIMBS); - bn_set_top(r->Z, P256_LIMBS); + if (ctx == NULL) { + ctx = new_ctx = BN_CTX_new(); + if (ctx == NULL) + goto err; + } + + BN_CTX_start(ctx); if (scalar) { generator = EC_GROUP_get0_generator(group); @@ -1150,23 +1209,23 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, } /* look if we can use precomputed multiples of generator */ - pre_comp = - EC_EX_DATA_get_data(group->extra_data, ecp_nistz256_pre_comp_dup, - ecp_nistz256_pre_comp_free, - ecp_nistz256_pre_comp_clear_free); + pre_comp = group->pre_comp.nistz256; if (pre_comp) { - /* If there is a precomputed table for the generator, - * check that it was generated with the same - * generator. */ + /* + * If there is a precomputed table for the generator, check that + * it was generated with the same generator. + */ EC_POINT *pre_comp_generator = EC_POINT_new(group); if (pre_comp_generator == NULL) goto err; if (!ecp_nistz256_set_from_affine(pre_comp_generator, group, pre_comp->precomp[0], - ctx)) + ctx)) { + EC_POINT_free(pre_comp_generator); goto err; + } if (0 == EC_POINT_cmp(group, generator, pre_comp_generator, ctx)) preComputedTable = (const PRECOMP256_ROW *)pre_comp->precomp; @@ -1175,10 +1234,12 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, } if (preComputedTable == NULL && ecp_nistz256_is_affine_G(generator)) { - /* If there is no precomputed data, but the generator - * is the default, a hardcoded table of precomputed - * data is used. This is because applications, such as - * Apache, do not use EC_KEY_precompute_mult. */ + /* + * If there is no precomputed data, but the generator is the + * default, a hardcoded table of precomputed data is used. This + * is because applications, such as Apache, do not use + * EC_KEY_precompute_mult. + */ preComputedTable = ecp_nistz256_precomputed; } @@ -1198,16 +1259,16 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, for (i = 0; i < bn_get_top(scalar) * BN_BYTES; i += BN_BYTES) { BN_ULONG d = bn_get_words(scalar)[i / BN_BYTES]; - p_str[i + 0] = d & 0xff; - p_str[i + 1] = (d >> 8) & 0xff; - p_str[i + 2] = (d >> 16) & 0xff; - p_str[i + 3] = (d >>= 24) & 0xff; + p_str[i + 0] = (unsigned char)d; + p_str[i + 1] = (unsigned char)(d >> 8); + p_str[i + 2] = (unsigned char)(d >> 16); + p_str[i + 3] = (unsigned char)(d >>= 24); if (BN_BYTES == 8) { d >>= 8; - p_str[i + 4] = d & 0xff; - p_str[i + 5] = (d >> 8) & 0xff; - p_str[i + 6] = (d >> 16) & 0xff; - p_str[i + 7] = (d >> 24) & 0xff; + p_str[i + 4] = (unsigned char)d; + p_str[i + 5] = (unsigned char)(d >> 8); + p_str[i + 6] = (unsigned char)(d >> 16); + p_str[i + 7] = (unsigned char)(d >> 24); } } @@ -1220,24 +1281,50 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, } else #endif { + BN_ULONG infty; + /* First window */ wvalue = (p_str[0] << 1) & mask; - index += window_size; + idx += window_size; wvalue = _booth_recode_w7(wvalue); - ecp_nistz256_gather_w7(&p.a, preComputedTable[0], wvalue >> 1); + ecp_nistz256_gather_w7(&p.a, preComputedTable[0], + wvalue >> 1); ecp_nistz256_neg(p.p.Z, p.p.Y); copy_conditional(p.p.Y, p.p.Z, wvalue & 1); - memcpy(p.p.Z, ONE, sizeof(ONE)); + /* + * Since affine infinity is encoded as (0,0) and + * Jacobian ias (,,0), we need to harmonize them + * by assigning "one" or zero to Z. + */ + infty = (p.p.X[0] | p.p.X[1] | p.p.X[2] | p.p.X[3] | + p.p.Y[0] | p.p.Y[1] | p.p.Y[2] | p.p.Y[3]); + if (P256_LIMBS == 8) + infty |= (p.p.X[4] | p.p.X[5] | p.p.X[6] | p.p.X[7] | + p.p.Y[4] | p.p.Y[5] | p.p.Y[6] | p.p.Y[7]); + + infty = 0 - is_zero(infty); + infty = ~infty; + + p.p.Z[0] = ONE[0] & infty; + p.p.Z[1] = ONE[1] & infty; + p.p.Z[2] = ONE[2] & infty; + p.p.Z[3] = ONE[3] & infty; + if (P256_LIMBS == 8) { + p.p.Z[4] = ONE[4] & infty; + p.p.Z[5] = ONE[5] & infty; + p.p.Z[6] = ONE[6] & infty; + p.p.Z[7] = ONE[7] & infty; + } for (i = 1; i < 37; i++) { - unsigned int off = (index - 1) / 8; + unsigned int off = (idx - 1) / 8; wvalue = p_str[off] | p_str[off + 1] << 8; - wvalue = (wvalue >> ((index - 1) % 8)) & mask; - index += window_size; + wvalue = (wvalue >> ((idx - 1) % 8)) & mask; + idx += window_size; wvalue = _booth_recode_w7(wvalue); @@ -1258,22 +1345,20 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, p_is_infinity = 1; if (no_precomp_for_generator) { - /* Without a precomputed table for the generator, it has to be - * handled like a normal point. */ - const BIGNUM **new_scalars; - const EC_POINT **new_points; - + /* + * Without a precomputed table for the generator, it has to be + * handled like a normal point. + */ new_scalars = OPENSSL_malloc((num + 1) * sizeof(BIGNUM *)); - if (!new_scalars) { + if (new_scalars == NULL) { ECerr(EC_F_ECP_NISTZ256_POINTS_MUL, ERR_R_MALLOC_FAILURE); - return 0; + goto err; } new_points = OPENSSL_malloc((num + 1) * sizeof(EC_POINT *)); - if (!new_points) { - OPENSSL_free(new_scalars); + if (new_points == NULL) { ECerr(EC_F_ECP_NISTZ256_POINTS_MUL, ERR_R_MALLOC_FAILURE); - return 0; + goto err; } memcpy(new_scalars, scalars, num * sizeof(BIGNUM *)); @@ -1291,39 +1376,42 @@ static int ecp_nistz256_points_mul(const EC_GROUP * group, if (p_is_infinity) out = &p.p; - ecp_nistz256_windowed_mul(group, out, scalars, points, num, ctx); + if (!ecp_nistz256_windowed_mul(group, out, scalars, points, num, ctx)) + goto err; if (!p_is_infinity) ecp_nistz256_point_add(&p.p, &p.p, out); } - if (no_precomp_for_generator) { - OPENSSL_free(points); - OPENSSL_free(scalars); + /* Not constant-time, but we're only operating on the public output. */ + if (!bn_set_words(r->X, p.p.X, P256_LIMBS) || + !bn_set_words(r->Y, p.p.Y, P256_LIMBS) || + !bn_set_words(r->Z, p.p.Z, P256_LIMBS)) { + goto err; } - - bn_set_data(r->X, p.p.X, sizeof(p.p.X)); - bn_set_data(r->Y, p.p.Y, sizeof(p.p.Y)); - bn_set_data(r->Z, p.p.Z, sizeof(p.p.Z)); - bn_correct_top(r->X); - bn_correct_top(r->Y); - bn_correct_top(r->Z); + r->Z_is_one = is_one(r->Z) & 1; ret = 1; err: + if (ctx) + BN_CTX_end(ctx); + BN_CTX_free(new_ctx); + OPENSSL_free(new_points); + OPENSSL_free(new_scalars); return ret; } -static int ecp_nistz256_get_affine(const EC_GROUP * group, - const EC_POINT * point, - BIGNUM * x, BIGNUM * y, BN_CTX * ctx) +__owur static int ecp_nistz256_get_affine(const EC_GROUP *group, + const EC_POINT *point, + BIGNUM *x, BIGNUM *y, BN_CTX *ctx) { BN_ULONG z_inv2[P256_LIMBS]; BN_ULONG z_inv3[P256_LIMBS]; BN_ULONG x_aff[P256_LIMBS]; BN_ULONG y_aff[P256_LIMBS]; BN_ULONG point_x[P256_LIMBS], point_y[P256_LIMBS], point_z[P256_LIMBS]; + BN_ULONG x_ret[P256_LIMBS], y_ret[P256_LIMBS]; if (EC_POINT_is_at_infinity(group, point)) { ECerr(EC_F_ECP_NISTZ256_GET_AFFINE, EC_R_POINT_AT_INFINITY); @@ -1342,107 +1430,87 @@ static int ecp_nistz256_get_affine(const EC_GROUP * group, ecp_nistz256_mul_mont(x_aff, z_inv2, point_x); if (x != NULL) { - bn_wexpand(x, P256_LIMBS); - bn_set_top(x, P256_LIMBS); - ecp_nistz256_from_mont(bn_get_words(x), x_aff); - bn_correct_top(x); + ecp_nistz256_from_mont(x_ret, x_aff); + if (!bn_set_words(x, x_ret, P256_LIMBS)) + return 0; } if (y != NULL) { ecp_nistz256_mul_mont(z_inv3, z_inv3, z_inv2); ecp_nistz256_mul_mont(y_aff, z_inv3, point_y); - bn_wexpand(y, P256_LIMBS); - bn_set_top(y, P256_LIMBS); - ecp_nistz256_from_mont(bn_get_words(y), y_aff); - bn_correct_top(y); + ecp_nistz256_from_mont(y_ret, y_aff); + if (!bn_set_words(y, y_ret, P256_LIMBS)) + return 0; } return 1; } -static EC_PRE_COMP *ecp_nistz256_pre_comp_new(const EC_GROUP * group) +static NISTZ256_PRE_COMP *ecp_nistz256_pre_comp_new(const EC_GROUP *group) { - EC_PRE_COMP *ret = NULL; + NISTZ256_PRE_COMP *ret = NULL; if (!group) return NULL; - ret = (EC_PRE_COMP *) OPENSSL_malloc(sizeof(EC_PRE_COMP)); + ret = OPENSSL_zalloc(sizeof(*ret)); - if (!ret) { + if (ret == NULL) { ECerr(EC_F_ECP_NISTZ256_PRE_COMP_NEW, ERR_R_MALLOC_FAILURE); return ret; } ret->group = group; ret->w = 6; /* default */ - ret->precomp = NULL; - ret->precomp_storage = NULL; ret->references = 1; - return ret; -} -static void *ecp_nistz256_pre_comp_dup(void *src_) -{ - EC_PRE_COMP *src = src_; - - /* no need to actually copy, these objects never change! */ - CRYPTO_add(&src->references, 1, CRYPTO_LOCK_EC_PRE_COMP); - - return src_; + ret->lock = CRYPTO_THREAD_lock_new(); + if (ret->lock == NULL) { + ECerr(EC_F_ECP_NISTZ256_PRE_COMP_NEW, ERR_R_MALLOC_FAILURE); + OPENSSL_free(ret); + return NULL; + } + return ret; } -static void ecp_nistz256_pre_comp_free(void *pre_) +NISTZ256_PRE_COMP *EC_nistz256_pre_comp_dup(NISTZ256_PRE_COMP *p) { int i; - EC_PRE_COMP *pre = pre_; - - if (!pre) - return; - - i = CRYPTO_add(&pre->references, -1, CRYPTO_LOCK_EC_PRE_COMP); - if (i > 0) - return; - - if (pre->precomp_storage) - OPENSSL_free(pre->precomp_storage); - - OPENSSL_free(pre); + if (p != NULL) + CRYPTO_atomic_add(&p->references, 1, &i, p->lock); + return p; } -static void ecp_nistz256_pre_comp_clear_free(void *pre_) +void EC_nistz256_pre_comp_free(NISTZ256_PRE_COMP *pre) { int i; - EC_PRE_COMP *pre = pre_; - if (!pre) + if (pre == NULL) return; - i = CRYPTO_add(&pre->references, -1, CRYPTO_LOCK_EC_PRE_COMP); + CRYPTO_atomic_add(&pre->references, -1, &i, pre->lock); + REF_PRINT_COUNT("EC_nistz256", x); if (i > 0) return; + REF_ASSERT_ISNT(i < 0); - if (pre->precomp_storage) { - OPENSSL_cleanse(pre->precomp, - 32 * sizeof(unsigned char) * (1 << pre->w) * 2 * 37); - OPENSSL_free(pre->precomp_storage); - } - OPENSSL_cleanse(pre, sizeof *pre); + OPENSSL_free(pre->precomp_storage); + CRYPTO_THREAD_lock_free(pre->lock); OPENSSL_free(pre); } -static int ecp_nistz256_window_have_precompute_mult(const EC_GROUP * group) + +static int ecp_nistz256_window_have_precompute_mult(const EC_GROUP *group) { /* There is a hard-coded table for the default generator. */ const EC_POINT *generator = EC_GROUP_get0_generator(group); + if (generator != NULL && ecp_nistz256_is_affine_G(generator)) { /* There is a hard-coded table for the default generator. */ return 1; } - return EC_EX_DATA_get_data(group->extra_data, ecp_nistz256_pre_comp_dup, - ecp_nistz256_pre_comp_free, - ecp_nistz256_pre_comp_clear_free) != NULL; + return HAVEPRECOMP(group, nistz256); } const EC_METHOD *EC_GFp_nistz256_method(void) @@ -1457,6 +1525,7 @@ const EC_METHOD *EC_GFp_nistz256_method(void) ec_GFp_mont_group_set_curve, ec_GFp_simple_group_get_curve, ec_GFp_simple_group_get_degree, + ec_group_simple_order_bits, ec_GFp_simple_group_check_discriminant, ec_GFp_simple_point_init, ec_GFp_simple_point_finish, @@ -1484,7 +1553,16 @@ const EC_METHOD *EC_GFp_nistz256_method(void) 0, /* field_div */ ec_GFp_mont_field_encode, ec_GFp_mont_field_decode, - ec_GFp_mont_field_set_to_one + ec_GFp_mont_field_set_to_one, + ec_key_simple_priv2oct, + ec_key_simple_oct2priv, + 0, /* set private */ + ec_key_simple_generate_key, + ec_key_simple_check_key, + ec_key_simple_generate_public_key, + 0, /* keycopy */ + 0, /* keyfinish */ + ecdh_simple_compute_key }; return &ret;