} EC_PRE_COMP;
/* 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 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],
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 = z->d;
+
+ if (z->top == (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;
+}
+
+static int ecp_nistz256_set_words(BIGNUM *a, BN_ULONG words[P256_LIMBS])
+ {
+ if (bn_wexpand(a, P256_LIMBS) == NULL) {
+ ECerr(EC_F_ECP_NISTZ256_SET_WORDS, ERR_R_MALLOC_FAILURE);
+ return 0;
+ }
+ memcpy(a->d, words, sizeof(BN_ULONG) * P256_LIMBS);
+ a->top = P256_LIMBS;
+ bn_correct_top(a);
+ return 1;
}
#ifndef ECP_NISTZ256_REFERENCE_IMPLEMENTATION
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);
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
+ * Infinity in encoded as (,,0)
*/
- 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]);
+ 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]);
+ /*
+ * 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)
}
/* r = sum(scalar[i]*point[i]) */
-static void ecp_nistz256_windowed_mul(const EC_GROUP *group,
+static int ecp_nistz256_windowed_mul(const EC_GROUP *group,
P256_POINT *r,
const BIGNUM **scalar,
const EC_POINT **point,
int num, BN_CTX *ctx)
{
- int i, j;
+
+ int i, j, ret = 0;
unsigned int index;
unsigned char (*p_str)[33] = NULL;
const unsigned int window_size = 5;
ecp_nistz256_point_add(r, r, &h);
}
+ ret = 1;
err:
if (table_storage)
OPENSSL_free(table_storage);
OPENSSL_free(p_str);
if (scalars)
OPENSSL_free(scalars);
+ return ret;
}
/* Coordinates of G, for which we have precomputed tables */
{
return (generator->X.top == P256_LIMBS) &&
(generator->Y.top == P256_LIMBS) &&
- (generator->Z.top == (P256_LIMBS - P256_LIMBS / 8)) &&
is_equal(generator->X.d, def_xG) &&
- is_equal(generator->Y.d, def_yG) && is_one(generator->Z.d);
+ is_equal(generator->Y.d, def_yG) && is_one(&generator->Z);
}
static int ecp_nistz256_mult_precompute(EC_GROUP *group, BN_CTX *ctx)
goto err;
for (j = 0; j < 37; j++) {
/*
- * It would be faster to use
- * ec_GFp_simple_points_make_affine and make multiple
- * points affine at the same time.
+ * It would be faster to use EC_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(preComputedTable[j]
- [k].X, &P->X);
- ecp_nistz256_bignum_to_field_elem(preComputedTable[j]
- [k].Y, &P->Y);
- for (i = 0; i < 7; i++)
- ec_GFp_simple_dbl(group, P, P, ctx);
+ if (!EC_POINT_make_affine(group, P, ctx))
+ goto err;
+ if (!ecp_nistz256_bignum_to_field_elem(preComputedTable[j][k].X,
+ &P->X) ||
+ !ecp_nistz256_bignum_to_field_elem(preComputedTable[j][k].Y,
+ &P->Y)) {
+ ECerr(EC_F_ECP_NISTZ256_MULT_PRECOMPUTE,
+ EC_R_COORDINATES_OUT_OF_RANGE);
+ goto err;
+ }
+ 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;
const EC_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;
const unsigned int window_size = 7;
const unsigned int mask = (1 << (window_size + 1)) - 1;
unsigned int wvalue;
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);
}
}
- /* 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);
- r->X.top = P256_LIMBS;
- r->Y.top = P256_LIMBS;
- r->Z.top = 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);
goto err;
if (!ecp_nistz256_set_from_affine
- (pre_comp_generator, group, pre_comp->precomp[0], ctx))
+ (pre_comp_generator, group, pre_comp->precomp[0], 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;
} else
#endif
{
+ BN_ULONG infty;
+
/* First window */
wvalue = (p_str[0] << 1) & mask;
index += window_size;
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;
* 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;
-
new_scalars = OPENSSL_malloc((num + 1) * sizeof(BIGNUM *));
if (!new_scalars) {
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);
ECerr(EC_F_ECP_NISTZ256_POINTS_MUL, ERR_R_MALLOC_FAILURE);
- return 0;
+ goto err;
}
memcpy(new_scalars, scalars, num * sizeof(BIGNUM *));
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);
- }
-
- memcpy(r->X.d, p.p.X, sizeof(p.p.X));
- memcpy(r->Y.d, p.p.Y, sizeof(p.p.Y));
- memcpy(r->Z.d, p.p.Z, sizeof(p.p.Z));
/* Not constant-time, but we're only operating on the public output. */
- bn_correct_top(&r->X);
- bn_correct_top(&r->Y);
- bn_correct_top(&r->Z);
- r->Z_is_one = is_one(p.p.Z);
+ if (!ecp_nistz256_set_words(&r->X, p.p.X) ||
+ !ecp_nistz256_set_words(&r->Y, p.p.Y) ||
+ !ecp_nistz256_set_words(&r->Z, p.p.Z)) {
+ goto err;
+ }
+ r->Z_is_one = is_one(&r->Z) & 1;
ret = 1;
- err:
+err:
+ if (ctx)
+ BN_CTX_end(ctx);
+ BN_CTX_free(new_ctx);
+ if (new_points)
+ OPENSSL_free(new_points);
+ if (new_scalars)
+ OPENSSL_free(new_scalars);
return ret;
}
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);
ecp_nistz256_mul_mont(x_aff, z_inv2, point_x);
if (x != NULL) {
- bn_wexpand(x, P256_LIMBS);
- x->top = P256_LIMBS;
- ecp_nistz256_from_mont(x->d, x_aff);
- bn_correct_top(x);
+ ecp_nistz256_from_mont(x_ret, x_aff);
+ if (!ecp_nistz256_set_words(x, x_ret))
+ 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);
- y->top = P256_LIMBS;
- ecp_nistz256_from_mont(y->d, y_aff);
- bn_correct_top(y);
+ ecp_nistz256_from_mont(y_ret, y_aff);
+ if (!ecp_nistz256_set_words(y, y_ret))
+ return 0;
}
return 1;