-int EC_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)
- {
- BN_CTX *new_ctx = NULL;
- EC_POINT *generator = NULL;
- EC_POINT *tmp = NULL;
- size_t totalnum;
- size_t i, j;
- int k, t;
- int r_is_at_infinity = 1;
- size_t max_bits = 0;
- size_t *wsize = NULL; /* individual window sizes */
- unsigned long *wbits = NULL; /* individual window contents */
- int *wpos = NULL; /* position of bottom bit of current individual windows
- * (wpos[i] is valid if wbits[i] != 0) */
- size_t num_val;
- EC_POINT **val = NULL; /* precomputation */
- EC_POINT **v;
- EC_POINT ***val_sub = NULL; /* pointers to sub-arrays of 'val' */
- int ret = 0;
-
- if (scalar != NULL)
- {
- generator = EC_GROUP_get0_generator(group);
- if (generator == NULL)
- {
- ECerr(EC_F_EC_POINTS_MUL, EC_R_UNDEFINED_GENERATOR);
- return 0;
- }
- }
-
- for (i = 0; i < num; i++)
- {
- if (group->meth != points[i]->meth)
- {
- ECerr(EC_F_EC_POINTS_MUL, EC_R_INCOMPATIBLE_OBJECTS);
- return 0;
- }
- }
-
- totalnum = num + (scalar != NULL);
-
- wsize = OPENSSL_malloc(totalnum * sizeof wsize[0]);
- wbits = OPENSSL_malloc(totalnum * sizeof wbits[0]);
- wpos = OPENSSL_malloc(totalnum * sizeof wpos[0]);
- if (wsize == NULL || wbits == NULL || wpos == NULL) goto err;
-
- /* num_val := total number of points to precompute */
- num_val = 0;
- for (i = 0; i < totalnum; i++)
- {
- size_t bits;
-
- bits = i < num ? BN_num_bits(scalars[i]) : BN_num_bits(scalar);
- wsize[i] = EC_window_bits_for_scalar_size(bits);
- num_val += 1u << (wsize[i] - 1);
- if (bits > max_bits)
- max_bits = bits;
- wbits[i] = 0;
- wpos[i] = 0;
- }
-
- /* all precomputed points go into a single array 'val',
- * 'val_sub[i]' is a pointer to the subarray for the i-th point */
- val = OPENSSL_malloc((num_val + 1) * sizeof val[0]);
- if (val == NULL) goto err;
- val[num_val] = NULL; /* pivot element */
-
- val_sub = OPENSSL_malloc(totalnum * sizeof val_sub[0]);
- if (val_sub == NULL) goto err;
-
- /* allocate points for precomputation */
- v = val;
- for (i = 0; i < totalnum; i++)
- {
- val_sub[i] = v;
- for (j = 0; j < (1u << (wsize[i] - 1)); j++)
- {
- *v = EC_POINT_new(group);
- if (*v == NULL) goto err;
- v++;
- }
- }
- if (!(v == val + num_val))
- {
- ECerr(EC_F_EC_POINTS_MUL, ERR_R_INTERNAL_ERROR);
- goto err;
- }
-
- if (ctx == NULL)
- {
- ctx = new_ctx = BN_CTX_new();
- if (ctx == NULL)
- goto err;
- }
-
- tmp = EC_POINT_new(group);
- if (tmp == NULL) goto err;
-
- /* prepare precomputed values:
- * val_sub[i][0] := points[i]
- * val_sub[i][1] := 3 * points[i]
- * val_sub[i][2] := 5 * points[i]
- * ...
- */
- for (i = 0; i < totalnum; i++)
- {
- if (i < num)
- {
- if (!EC_POINT_copy(val_sub[i][0], points[i])) goto err;
- if (scalars[i]->neg)
- {
- if (!EC_POINT_invert(group, val_sub[i][0], ctx)) goto err;
- }
- }
- else
- {
- if (!EC_POINT_copy(val_sub[i][0], generator)) goto err;
- if (scalar->neg)
- {
- if (!EC_POINT_invert(group, val_sub[i][0], ctx)) goto err;
- }
- }
-
- if (wsize[i] > 1)
- {
- if (!EC_POINT_dbl(group, tmp, val_sub[i][0], ctx)) goto err;
- for (j = 1; j < (1u << (wsize[i] - 1)); j++)
- {
- if (!EC_POINT_add(group, val_sub[i][j], val_sub[i][j - 1], tmp, ctx)) goto err;
- }
- }
- }
-
-#if 1 /* optional; EC_window_bits_for_scalar_size assumes we do this step */
- if (!EC_POINTs_make_affine(group, num_val, val, ctx)) goto err;
-#endif
-
- r_is_at_infinity = 1;
-
- for (k = max_bits - 1; k >= 0; k--)
- {
- if (!r_is_at_infinity)
- {
- if (!EC_POINT_dbl(group, r, r, ctx)) goto err;
- }
-
- for (i = 0; i < totalnum; i++)
- {
- if (wbits[i] == 0)
- {
- const BIGNUM *s;
-
- s = i < num ? scalars[i] : scalar;
-
- if (BN_is_bit_set(s, k))
- {
- /* look at bits k - wsize[i] + 1 .. k for this window */
- t = k - wsize[i] + 1;
- while (!BN_is_bit_set(s, t)) /* BN_is_bit_set is false for t < 0 */
- t++;
- wpos[i] = t;
- wbits[i] = 1;
- for (t = k - 1; t >= wpos[i]; t--)
- {
- wbits[i] <<= 1;
- if (BN_is_bit_set(s, t))
- wbits[i]++;
- }
- /* now wbits[i] is the odd bit pattern at bits wpos[i] .. k */
- }
- }
-
- if ((wbits[i] != 0) && (wpos[i] == k))
- {
- if (r_is_at_infinity)
- {
- if (!EC_POINT_copy(r, val_sub[i][wbits[i] >> 1])) goto err;
- r_is_at_infinity = 0;
- }
- else
- {
- if (!EC_POINT_add(group, r, r, val_sub[i][wbits[i] >> 1], ctx)) goto err;
- }
- wbits[i] = 0;
- }
- }
- }
-
- if (r_is_at_infinity)
- if (!EC_POINT_set_to_infinity(group, r)) goto err;
-
- ret = 1;
+int ec_wNAF_mul(const EC_GROUP *group, EC_POINT *r, const BIGNUM *scalar,
+ size_t num, const EC_POINT *points[], const BIGNUM *scalars[],
+ BN_CTX *ctx)
+{
+ BN_CTX *new_ctx = NULL;
+ const EC_POINT *generator = NULL;
+ EC_POINT *tmp = NULL;
+ size_t totalnum;
+ size_t blocksize = 0, numblocks = 0; /* for wNAF splitting */
+ size_t pre_points_per_block = 0;
+ size_t i, j;
+ int k;
+ int r_is_inverted = 0;
+ int r_is_at_infinity = 1;
+ size_t *wsize = NULL; /* individual window sizes */
+ signed char **wNAF = NULL; /* individual wNAFs */
+ size_t *wNAF_len = NULL;
+ size_t max_len = 0;
+ size_t num_val;
+ EC_POINT **val = NULL; /* precomputation */
+ EC_POINT **v;
+ EC_POINT ***val_sub = NULL; /* pointers to sub-arrays of 'val' or
+ * 'pre_comp->points' */
+ const EC_PRE_COMP *pre_comp = NULL;
+ int num_scalar = 0; /* flag: will be set to 1 if 'scalar' must be
+ * treated like other scalars, i.e.
+ * precomputation is not available */
+ int ret = 0;
+
+ if (group->meth != r->meth) {
+ ECerr(EC_F_EC_WNAF_MUL, EC_R_INCOMPATIBLE_OBJECTS);
+ return 0;
+ }
+
+ if ((scalar == NULL) && (num == 0)) {
+ return EC_POINT_set_to_infinity(group, r);
+ }
+
+ for (i = 0; i < num; i++) {
+ if (group->meth != points[i]->meth) {
+ ECerr(EC_F_EC_WNAF_MUL, EC_R_INCOMPATIBLE_OBJECTS);
+ return 0;
+ }
+ }
+
+ if (ctx == NULL) {
+ ctx = new_ctx = BN_CTX_new();
+ if (ctx == NULL)
+ goto err;
+ }
+
+ if (scalar != NULL) {
+ generator = EC_GROUP_get0_generator(group);
+ if (generator == NULL) {
+ ECerr(EC_F_EC_WNAF_MUL, EC_R_UNDEFINED_GENERATOR);
+ goto err;
+ }
+
+ /* look if we can use precomputed multiples of generator */
+
+ pre_comp = group->pre_comp.ec;
+ if (pre_comp && pre_comp->numblocks
+ && (EC_POINT_cmp(group, generator, pre_comp->points[0], ctx) ==
+ 0)) {
+ blocksize = pre_comp->blocksize;
+
+ /*
+ * determine maximum number of blocks that wNAF splitting may
+ * yield (NB: maximum wNAF length is bit length plus one)
+ */
+ numblocks = (BN_num_bits(scalar) / blocksize) + 1;
+
+ /*
+ * we cannot use more blocks than we have precomputation for
+ */
+ if (numblocks > pre_comp->numblocks)
+ numblocks = pre_comp->numblocks;
+
+ pre_points_per_block = (size_t)1 << (pre_comp->w - 1);
+
+ /* check that pre_comp looks sane */
+ if (pre_comp->num != (pre_comp->numblocks * pre_points_per_block)) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ } else {
+ /* can't use precomputation */
+ pre_comp = NULL;
+ numblocks = 1;
+ num_scalar = 1; /* treat 'scalar' like 'num'-th element of
+ * 'scalars' */
+ }
+ }
+
+ totalnum = num + numblocks;
+
+ wsize = OPENSSL_malloc(totalnum * sizeof wsize[0]);
+ wNAF_len = OPENSSL_malloc(totalnum * sizeof wNAF_len[0]);
+ wNAF = OPENSSL_malloc((totalnum + 1) * sizeof wNAF[0]); /* includes space
+ * for pivot */
+ val_sub = OPENSSL_malloc(totalnum * sizeof val_sub[0]);
+
+ /* Ensure wNAF is initialised in case we end up going to err */
+ if (wNAF != NULL)
+ wNAF[0] = NULL; /* preliminary pivot */
+
+ if (wsize == NULL || wNAF_len == NULL || wNAF == NULL || val_sub == NULL) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+
+ /*
+ * num_val will be the total number of temporarily precomputed points
+ */
+ num_val = 0;
+
+ for (i = 0; i < num + num_scalar; i++) {
+ size_t bits;
+
+ bits = i < num ? BN_num_bits(scalars[i]) : BN_num_bits(scalar);
+ wsize[i] = EC_window_bits_for_scalar_size(bits);
+ num_val += (size_t)1 << (wsize[i] - 1);
+ wNAF[i + 1] = NULL; /* make sure we always have a pivot */
+ wNAF[i] =
+ bn_compute_wNAF((i < num ? scalars[i] : scalar), wsize[i],
+ &wNAF_len[i]);
+ if (wNAF[i] == NULL)
+ goto err;
+ if (wNAF_len[i] > max_len)
+ max_len = wNAF_len[i];
+ }
+
+ if (numblocks) {
+ /* we go here iff scalar != NULL */
+
+ if (pre_comp == NULL) {
+ if (num_scalar != 1) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ /* we have already generated a wNAF for 'scalar' */
+ } else {
+ signed char *tmp_wNAF = NULL;
+ size_t tmp_len = 0;
+
+ if (num_scalar != 0) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+
+ /*
+ * use the window size for which we have precomputation
+ */
+ wsize[num] = pre_comp->w;
+ tmp_wNAF = bn_compute_wNAF(scalar, wsize[num], &tmp_len);
+ if (!tmp_wNAF)
+ goto err;
+
+ if (tmp_len <= max_len) {
+ /*
+ * One of the other wNAFs is at least as long as the wNAF
+ * belonging to the generator, so wNAF splitting will not buy
+ * us anything.
+ */
+
+ numblocks = 1;
+ totalnum = num + 1; /* don't use wNAF splitting */
+ wNAF[num] = tmp_wNAF;
+ wNAF[num + 1] = NULL;
+ wNAF_len[num] = tmp_len;
+ if (tmp_len > max_len)
+ max_len = tmp_len;
+ /*
+ * pre_comp->points starts with the points that we need here:
+ */
+ val_sub[num] = pre_comp->points;
+ } else {
+ /*
+ * don't include tmp_wNAF directly into wNAF array - use wNAF
+ * splitting and include the blocks
+ */
+
+ signed char *pp;
+ EC_POINT **tmp_points;
+
+ if (tmp_len < numblocks * blocksize) {
+ /*
+ * possibly we can do with fewer blocks than estimated
+ */
+ numblocks = (tmp_len + blocksize - 1) / blocksize;
+ if (numblocks > pre_comp->numblocks) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ totalnum = num + numblocks;
+ }
+
+ /* split wNAF in 'numblocks' parts */
+ pp = tmp_wNAF;
+ tmp_points = pre_comp->points;
+
+ for (i = num; i < totalnum; i++) {
+ if (i < totalnum - 1) {
+ wNAF_len[i] = blocksize;
+ if (tmp_len < blocksize) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+ tmp_len -= blocksize;
+ } else
+ /*
+ * last block gets whatever is left (this could be
+ * more or less than 'blocksize'!)
+ */
+ wNAF_len[i] = tmp_len;
+
+ wNAF[i + 1] = NULL;
+ wNAF[i] = OPENSSL_malloc(wNAF_len[i]);
+ if (wNAF[i] == NULL) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_MALLOC_FAILURE);
+ OPENSSL_free(tmp_wNAF);
+ goto err;
+ }
+ memcpy(wNAF[i], pp, wNAF_len[i]);
+ if (wNAF_len[i] > max_len)
+ max_len = wNAF_len[i];
+
+ if (*tmp_points == NULL) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ OPENSSL_free(tmp_wNAF);
+ goto err;
+ }
+ val_sub[i] = tmp_points;
+ tmp_points += pre_points_per_block;
+ pp += blocksize;
+ }
+ OPENSSL_free(tmp_wNAF);
+ }
+ }
+ }
+
+ /*
+ * All points we precompute now go into a single array 'val'.
+ * 'val_sub[i]' is a pointer to the subarray for the i-th point, or to a
+ * subarray of 'pre_comp->points' if we already have precomputation.
+ */
+ val = OPENSSL_malloc((num_val + 1) * sizeof val[0]);
+ if (val == NULL) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_MALLOC_FAILURE);
+ goto err;
+ }
+ val[num_val] = NULL; /* pivot element */
+
+ /* allocate points for precomputation */
+ v = val;
+ for (i = 0; i < num + num_scalar; i++) {
+ val_sub[i] = v;
+ for (j = 0; j < ((size_t)1 << (wsize[i] - 1)); j++) {
+ *v = EC_POINT_new(group);
+ if (*v == NULL)
+ goto err;
+ v++;
+ }
+ }
+ if (!(v == val + num_val)) {
+ ECerr(EC_F_EC_WNAF_MUL, ERR_R_INTERNAL_ERROR);
+ goto err;
+ }
+
+ if ((tmp = EC_POINT_new(group)) == NULL)
+ goto err;
+
+ /*-
+ * prepare precomputed values:
+ * val_sub[i][0] := points[i]
+ * val_sub[i][1] := 3 * points[i]
+ * val_sub[i][2] := 5 * points[i]
+ * ...
+ */
+ for (i = 0; i < num + num_scalar; i++) {
+ if (i < num) {
+ if (!EC_POINT_copy(val_sub[i][0], points[i]))
+ goto err;
+ } else {
+ if (!EC_POINT_copy(val_sub[i][0], generator))
+ goto err;
+ }
+
+ if (wsize[i] > 1) {
+ if (!EC_POINT_dbl(group, tmp, val_sub[i][0], ctx))
+ goto err;
+ for (j = 1; j < ((size_t)1 << (wsize[i] - 1)); j++) {
+ if (!EC_POINT_add
+ (group, val_sub[i][j], val_sub[i][j - 1], tmp, ctx))
+ goto err;
+ }
+ }
+ }
+
+ if (!EC_POINTs_make_affine(group, num_val, val, ctx))
+ goto err;
+
+ r_is_at_infinity = 1;
+
+ for (k = max_len - 1; k >= 0; k--) {
+ if (!r_is_at_infinity) {
+ if (!EC_POINT_dbl(group, r, r, ctx))
+ goto err;
+ }
+
+ for (i = 0; i < totalnum; i++) {
+ if (wNAF_len[i] > (size_t)k) {
+ int digit = wNAF[i][k];
+ int is_neg;
+
+ if (digit) {
+ is_neg = digit < 0;
+
+ if (is_neg)
+ digit = -digit;
+
+ if (is_neg != r_is_inverted) {
+ if (!r_is_at_infinity) {
+ if (!EC_POINT_invert(group, r, ctx))
+ goto err;
+ }
+ r_is_inverted = !r_is_inverted;
+ }
+
+ /* digit > 0 */
+
+ if (r_is_at_infinity) {
+ if (!EC_POINT_copy(r, val_sub[i][digit >> 1]))
+ goto err;
+ r_is_at_infinity = 0;
+ } else {
+ if (!EC_POINT_add
+ (group, r, r, val_sub[i][digit >> 1], ctx))
+ goto err;
+ }
+ }
+ }
+ }
+ }
+
+ if (r_is_at_infinity) {
+ if (!EC_POINT_set_to_infinity(group, r))
+ goto err;
+ } else {
+ if (r_is_inverted)
+ if (!EC_POINT_invert(group, r, ctx))
+ goto err;
+ }
+
+ ret = 1;