#include <string.h>
-#include "internal/bn_int.h"
-#include <openssl/err.h>
-#include <openssl/ec.h>
#include "cryptlib.h"
-
+#include "internal/bn_int.h"
#include "ec_lcl.h"
#if BN_BITS2 != 64
typedef struct ec_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;
/* 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 */
+/* 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 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],
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);
+ 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 index);
+ const P256_POINT_AFFINE *in_t, int idx);
void ecp_nistz256_gather_w7(P256_POINT_AFFINE *val,
- const P256_POINT_AFFINE *in_t, int index);
+ const P256_POINT_AFFINE *in_t, int idx);
/* One converted into the Montgomery domain */
static const BN_ULONG ONE[P256_LIMBS] = {
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);
}
/* 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];
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);
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 */
+ /*
+ * 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]);
if (P256_LIMBS == 8)
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];
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. */
+/*
+ * 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)
+ const BIGNUM *in)
{
return bn_copy_words(out, in, P256_LIMBS);
}
P256_POINT *r,
const BIGNUM **scalar,
const EC_POINT **point,
- int num, BN_CTX *ctx)
+ size_t num, BN_CTX *ctx)
{
- int i, j;
- unsigned int index;
+ size_t i;
+ int j;
+ 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) {
}
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];
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++)
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);
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
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);
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);
TOBN(0xd2e88688, 0xdd21f325), TOBN(0x8571ff18, 0x25885d85)
};
-/* ecp_nistz256_is_affine_G returns one if |generator| is the standard,
- * P-256 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) &&
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. */
+ * implicit value of infinity at index zero. We use window of size 7, and
+ * therefore require ceil(256/7) = 37 tables.
+ */
BIGNUM *order;
EC_POINT *P = NULL, *T = NULL;
const EC_POINT *generator;
}
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;
}
P = EC_POINT_new(group);
T = EC_POINT_new(group);
- /* The zero entry is implicitly infinity, and we skip it,
- * storing other values with -1 offset. */
+ /*
+ * The zero entry is implicitly infinity, and we skip it, storing other
+ * values with -1 offset.
+ */
EC_POINT_copy(T, generator);
for (k = 0; k < 64; k++) {
EC_POINT_copy(P, T);
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. */
+ /*
+ * 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);
*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. */
+ * 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])
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;
/* 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],
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],
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,
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);
const PRECOMP256_ROW *preComputedTable = NULL;
const EC_PRE_COMP *pre_comp = NULL;
const EC_POINT *generator = NULL;
- unsigned int index = 0;
+ unsigned int idx = 0;
const unsigned int window_size = 7;
const unsigned int mask = (1 << (window_size + 1)) - 1;
unsigned int wvalue;
ecp_nistz256_pre_comp_clear_free);
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 (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;
}
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);
}
}
{
/* First window */
wvalue = (p_str[0] << 1) & mask;
- index += window_size;
+ idx += window_size;
wvalue = _booth_recode_w7(wvalue);
memcpy(p.p.Z, ONE, sizeof(ONE));
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);
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. */
+ /*
+ * 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;