boringssl/ssl/ssl_ecdh.c

612 lines
16 KiB
C
Raw Normal View History

Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
/* Copyright (c) 2015, Google Inc.
*
* Permission to use, copy, modify, and/or distribute this software for any
* purpose with or without fee is hereby granted, provided that the above
* copyright notice and this permission notice appear in all copies.
*
* THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
* WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
* MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR ANY
* SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
* WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN ACTION
* OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF OR IN
* CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE. */
#include <openssl/ssl.h>
#include <assert.h>
#include <string.h>
#include <openssl/bn.h>
#include <openssl/bytestring.h>
#include <openssl/curve25519.h>
#include <openssl/ec.h>
#include <openssl/err.h>
#include <openssl/mem.h>
#include <openssl/newhope.h>
#include <openssl/nid.h>
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
#include "internal.h"
/* |EC_POINT| implementation. */
static void ssl_ec_point_cleanup(SSL_ECDH_CTX *ctx) {
BIGNUM *private_key = (BIGNUM *)ctx->data;
BN_clear_free(private_key);
}
static int ssl_ec_point_offer(SSL_ECDH_CTX *ctx, CBB *out) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
assert(ctx->data == NULL);
BIGNUM *private_key = BN_new();
if (private_key == NULL) {
return 0;
}
ctx->data = private_key;
/* Set up a shared |BN_CTX| for all operations. */
BN_CTX *bn_ctx = BN_CTX_new();
if (bn_ctx == NULL) {
return 0;
}
BN_CTX_start(bn_ctx);
int ret = 0;
EC_POINT *public_key = NULL;
EC_GROUP *group = EC_GROUP_new_by_curve_name(ctx->method->nid);
if (group == NULL) {
goto err;
}
/* Generate a private key. */
if (!BN_rand_range_ex(private_key, 1, EC_GROUP_get0_order(group))) {
goto err;
}
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
/* Compute the corresponding public key and serialize it. */
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
public_key = EC_POINT_new(group);
if (public_key == NULL ||
!EC_POINT_mul(group, public_key, private_key, NULL, NULL, bn_ctx) ||
!EC_POINT_point2cbb(out, group, public_key, POINT_CONVERSION_UNCOMPRESSED,
bn_ctx)) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
goto err;
}
ret = 1;
err:
EC_GROUP_free(group);
EC_POINT_free(public_key);
BN_CTX_end(bn_ctx);
BN_CTX_free(bn_ctx);
return ret;
}
static int ssl_ec_point_finish(SSL_ECDH_CTX *ctx, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
BIGNUM *private_key = (BIGNUM *)ctx->data;
assert(private_key != NULL);
*out_alert = SSL_AD_INTERNAL_ERROR;
/* Set up a shared |BN_CTX| for all operations. */
BN_CTX *bn_ctx = BN_CTX_new();
if (bn_ctx == NULL) {
return 0;
}
BN_CTX_start(bn_ctx);
int ret = 0;
EC_GROUP *group = EC_GROUP_new_by_curve_name(ctx->method->nid);
EC_POINT *peer_point = NULL, *result = NULL;
uint8_t *secret = NULL;
if (group == NULL) {
goto err;
}
/* Compute the x-coordinate of |peer_key| * |private_key|. */
peer_point = EC_POINT_new(group);
result = EC_POINT_new(group);
if (peer_point == NULL || result == NULL) {
goto err;
}
BIGNUM *x = BN_CTX_get(bn_ctx);
if (x == NULL) {
goto err;
}
if (!EC_POINT_oct2point(group, peer_point, peer_key, peer_key_len, bn_ctx)) {
*out_alert = SSL_AD_DECODE_ERROR;
goto err;
}
if (!EC_POINT_mul(group, result, NULL, peer_point, private_key, bn_ctx) ||
!EC_POINT_get_affine_coordinates_GFp(group, result, x, NULL, bn_ctx)) {
goto err;
}
/* Encode the x-coordinate left-padded with zeros. */
size_t secret_len = (EC_GROUP_get_degree(group) + 7) / 8;
secret = OPENSSL_malloc(secret_len);
if (secret == NULL || !BN_bn2bin_padded(secret, secret_len, x)) {
goto err;
}
*out_secret = secret;
*out_secret_len = secret_len;
secret = NULL;
ret = 1;
err:
EC_GROUP_free(group);
EC_POINT_free(peer_point);
EC_POINT_free(result);
BN_CTX_end(bn_ctx);
BN_CTX_free(bn_ctx);
OPENSSL_free(secret);
return ret;
}
static int ssl_ec_point_accept(SSL_ECDH_CTX *ctx, CBB *out_public_key,
uint8_t **out_secret, size_t *out_secret_len,
uint8_t *out_alert, const uint8_t *peer_key,
size_t peer_key_len) {
*out_alert = SSL_AD_INTERNAL_ERROR;
if (!ssl_ec_point_offer(ctx, out_public_key) ||
!ssl_ec_point_finish(ctx, out_secret, out_secret_len, out_alert, peer_key,
peer_key_len)) {
return 0;
}
return 1;
}
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
/* X25119 implementation. */
static void ssl_x25519_cleanup(SSL_ECDH_CTX *ctx) {
if (ctx->data == NULL) {
return;
}
OPENSSL_cleanse(ctx->data, 32);
OPENSSL_free(ctx->data);
}
static int ssl_x25519_offer(SSL_ECDH_CTX *ctx, CBB *out) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
assert(ctx->data == NULL);
ctx->data = OPENSSL_malloc(32);
if (ctx->data == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
uint8_t public_key[32];
X25519_keypair(public_key, (uint8_t *)ctx->data);
return CBB_add_bytes(out, public_key, sizeof(public_key));
}
static int ssl_x25519_finish(SSL_ECDH_CTX *ctx, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
assert(ctx->data != NULL);
*out_alert = SSL_AD_INTERNAL_ERROR;
uint8_t *secret = OPENSSL_malloc(32);
if (secret == NULL) {
return 0;
}
if (peer_key_len != 32 ||
!X25519(secret, (uint8_t *)ctx->data, peer_key)) {
OPENSSL_free(secret);
*out_alert = SSL_AD_DECODE_ERROR;
OPENSSL_PUT_ERROR(SSL, SSL_R_BAD_ECPOINT);
return 0;
}
*out_secret = secret;
*out_secret_len = 32;
return 1;
}
static int ssl_x25519_accept(SSL_ECDH_CTX *ctx, CBB *out_public_key,
uint8_t **out_secret, size_t *out_secret_len,
uint8_t *out_alert, const uint8_t *peer_key,
size_t peer_key_len) {
*out_alert = SSL_AD_INTERNAL_ERROR;
if (!ssl_x25519_offer(ctx, out_public_key) ||
!ssl_x25519_finish(ctx, out_secret, out_secret_len, out_alert, peer_key,
peer_key_len)) {
return 0;
}
return 1;
}
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
/* Combined X25119 + New Hope (post-quantum) implementation. */
typedef struct {
uint8_t x25519_key[32];
NEWHOPE_POLY *newhope_sk;
} cecpq1_data;
#define CECPQ1_OFFERMSG_LENGTH (32 + NEWHOPE_OFFERMSG_LENGTH)
#define CECPQ1_ACCEPTMSG_LENGTH (32 + NEWHOPE_ACCEPTMSG_LENGTH)
#define CECPQ1_SECRET_LENGTH (32 + SHA256_DIGEST_LENGTH)
static void ssl_cecpq1_cleanup(SSL_ECDH_CTX *ctx) {
if (ctx->data == NULL) {
return;
}
cecpq1_data *data = ctx->data;
NEWHOPE_POLY_free(data->newhope_sk);
OPENSSL_cleanse(data, sizeof(cecpq1_data));
OPENSSL_free(data);
}
static int ssl_cecpq1_offer(SSL_ECDH_CTX *ctx, CBB *out) {
assert(ctx->data == NULL);
cecpq1_data *data = OPENSSL_malloc(sizeof(cecpq1_data));
if (data == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
ctx->data = data;
data->newhope_sk = NEWHOPE_POLY_new();
if (data->newhope_sk == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
uint8_t x25519_public_key[32];
X25519_keypair(x25519_public_key, data->x25519_key);
uint8_t newhope_offermsg[NEWHOPE_OFFERMSG_LENGTH];
NEWHOPE_offer(newhope_offermsg, data->newhope_sk);
if (!CBB_add_bytes(out, x25519_public_key, sizeof(x25519_public_key)) ||
!CBB_add_bytes(out, newhope_offermsg, sizeof(newhope_offermsg))) {
return 0;
}
return 1;
}
static int ssl_cecpq1_accept(SSL_ECDH_CTX *ctx, CBB *cbb, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
if (peer_key_len != CECPQ1_OFFERMSG_LENGTH) {
*out_alert = SSL_AD_DECODE_ERROR;
return 0;
}
*out_alert = SSL_AD_INTERNAL_ERROR;
assert(ctx->data == NULL);
cecpq1_data *data = OPENSSL_malloc(sizeof(cecpq1_data));
if (data == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
data->newhope_sk = NULL;
ctx->data = data;
uint8_t *secret = OPENSSL_malloc(CECPQ1_SECRET_LENGTH);
if (secret == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
/* Generate message to server, and secret key, at once. */
uint8_t x25519_public_key[32];
X25519_keypair(x25519_public_key, data->x25519_key);
if (!X25519(secret, data->x25519_key, peer_key)) {
*out_alert = SSL_AD_DECODE_ERROR;
OPENSSL_PUT_ERROR(SSL, SSL_R_BAD_ECPOINT);
goto err;
}
uint8_t newhope_acceptmsg[NEWHOPE_ACCEPTMSG_LENGTH];
if (!NEWHOPE_accept(secret + 32, newhope_acceptmsg, peer_key + 32,
NEWHOPE_OFFERMSG_LENGTH)) {
*out_alert = SSL_AD_DECODE_ERROR;
goto err;
}
if (!CBB_add_bytes(cbb, x25519_public_key, sizeof(x25519_public_key)) ||
!CBB_add_bytes(cbb, newhope_acceptmsg, sizeof(newhope_acceptmsg))) {
goto err;
}
*out_secret = secret;
*out_secret_len = CECPQ1_SECRET_LENGTH;
return 1;
err:
OPENSSL_cleanse(secret, CECPQ1_SECRET_LENGTH);
OPENSSL_free(secret);
return 0;
}
static int ssl_cecpq1_finish(SSL_ECDH_CTX *ctx, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
if (peer_key_len != CECPQ1_ACCEPTMSG_LENGTH) {
*out_alert = SSL_AD_DECODE_ERROR;
return 0;
}
*out_alert = SSL_AD_INTERNAL_ERROR;
assert(ctx->data != NULL);
cecpq1_data *data = ctx->data;
uint8_t *secret = OPENSSL_malloc(CECPQ1_SECRET_LENGTH);
if (secret == NULL) {
OPENSSL_PUT_ERROR(SSL, ERR_R_MALLOC_FAILURE);
return 0;
}
if (!X25519(secret, data->x25519_key, peer_key)) {
*out_alert = SSL_AD_DECODE_ERROR;
OPENSSL_PUT_ERROR(SSL, SSL_R_BAD_ECPOINT);
goto err;
}
if (!NEWHOPE_finish(secret + 32, data->newhope_sk, peer_key + 32,
NEWHOPE_ACCEPTMSG_LENGTH)) {
*out_alert = SSL_AD_DECODE_ERROR;
goto err;
}
*out_secret = secret;
*out_secret_len = CECPQ1_SECRET_LENGTH;
return 1;
err:
OPENSSL_cleanse(secret, CECPQ1_SECRET_LENGTH);
OPENSSL_free(secret);
return 0;
}
/* Legacy DHE-based implementation. */
static void ssl_dhe_cleanup(SSL_ECDH_CTX *ctx) {
DH_free((DH *)ctx->data);
}
static int ssl_dhe_offer(SSL_ECDH_CTX *ctx, CBB *out) {
DH *dh = (DH *)ctx->data;
/* The group must have been initialized already, but not the key. */
assert(dh != NULL);
assert(dh->priv_key == NULL);
/* Due to a bug in yaSSL, the public key must be zero padded to the size of
* the prime. */
return DH_generate_key(dh) &&
BN_bn2cbb_padded(out, BN_num_bytes(dh->p), dh->pub_key);
}
static int ssl_dhe_finish(SSL_ECDH_CTX *ctx, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
DH *dh = (DH *)ctx->data;
assert(dh != NULL);
assert(dh->priv_key != NULL);
*out_alert = SSL_AD_INTERNAL_ERROR;
int secret_len = 0;
uint8_t *secret = NULL;
BIGNUM *peer_point = BN_bin2bn(peer_key, peer_key_len, NULL);
if (peer_point == NULL) {
goto err;
}
secret = OPENSSL_malloc(DH_size(dh));
if (secret == NULL) {
goto err;
}
secret_len = DH_compute_key(secret, peer_point, dh);
if (secret_len <= 0) {
goto err;
}
*out_secret = secret;
*out_secret_len = (size_t)secret_len;
BN_free(peer_point);
return 1;
err:
if (secret_len > 0) {
OPENSSL_cleanse(secret, (size_t)secret_len);
}
OPENSSL_free(secret);
BN_free(peer_point);
return 0;
}
static int ssl_dhe_accept(SSL_ECDH_CTX *ctx, CBB *out_public_key,
uint8_t **out_secret, size_t *out_secret_len,
uint8_t *out_alert, const uint8_t *peer_key,
size_t peer_key_len) {
*out_alert = SSL_AD_INTERNAL_ERROR;
if (!ssl_dhe_offer(ctx, out_public_key) ||
!ssl_dhe_finish(ctx, out_secret, out_secret_len, out_alert, peer_key,
peer_key_len)) {
return 0;
}
return 1;
}
static const SSL_ECDH_METHOD kDHEMethod = {
NID_undef, 0, "",
ssl_dhe_cleanup,
ssl_dhe_offer,
ssl_dhe_accept,
ssl_dhe_finish,
CBS_get_u16_length_prefixed,
CBB_add_u16_length_prefixed,
};
static const SSL_ECDH_METHOD kCECPQ1Method = {
NID_undef, 0, "",
ssl_cecpq1_cleanup,
ssl_cecpq1_offer,
ssl_cecpq1_accept,
ssl_cecpq1_finish,
CBS_get_u16_length_prefixed,
CBB_add_u16_length_prefixed,
};
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
static const SSL_ECDH_METHOD kMethods[] = {
{
NID_X9_62_prime256v1,
SSL_CURVE_SECP256R1,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
"P-256",
ssl_ec_point_cleanup,
ssl_ec_point_offer,
ssl_ec_point_accept,
ssl_ec_point_finish,
CBS_get_u8_length_prefixed,
CBB_add_u8_length_prefixed,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
},
{
NID_secp384r1,
SSL_CURVE_SECP384R1,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
"P-384",
ssl_ec_point_cleanup,
ssl_ec_point_offer,
ssl_ec_point_accept,
ssl_ec_point_finish,
CBS_get_u8_length_prefixed,
CBB_add_u8_length_prefixed,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
},
{
NID_secp521r1,
SSL_CURVE_SECP521R1,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
"P-521",
ssl_ec_point_cleanup,
ssl_ec_point_offer,
ssl_ec_point_accept,
ssl_ec_point_finish,
CBS_get_u8_length_prefixed,
CBB_add_u8_length_prefixed,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
},
{
NID_X25519,
SSL_CURVE_X25519,
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
"X25519",
ssl_x25519_cleanup,
ssl_x25519_offer,
ssl_x25519_accept,
ssl_x25519_finish,
CBS_get_u8_length_prefixed,
CBB_add_u8_length_prefixed,
},
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
};
static const SSL_ECDH_METHOD *method_from_group_id(uint16_t group_id) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
size_t i;
for (i = 0; i < sizeof(kMethods) / sizeof(kMethods[0]); i++) {
if (kMethods[i].group_id == group_id) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
return &kMethods[i];
}
}
return NULL;
}
static const SSL_ECDH_METHOD *method_from_nid(int nid) {
size_t i;
for (i = 0; i < sizeof(kMethods) / sizeof(kMethods[0]); i++) {
if (kMethods[i].nid == nid) {
return &kMethods[i];
}
}
return NULL;
}
const char* SSL_get_curve_name(uint16_t group_id) {
const SSL_ECDH_METHOD *method = method_from_group_id(group_id);
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
if (method == NULL) {
return NULL;
}
return method->name;
}
int ssl_nid_to_group_id(uint16_t *out_group_id, int nid) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
const SSL_ECDH_METHOD *method = method_from_nid(nid);
if (method == NULL) {
return 0;
}
*out_group_id = method->group_id;
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
return 1;
}
int SSL_ECDH_CTX_init(SSL_ECDH_CTX *ctx, uint16_t group_id) {
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
SSL_ECDH_CTX_cleanup(ctx);
const SSL_ECDH_METHOD *method = method_from_group_id(group_id);
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
if (method == NULL) {
OPENSSL_PUT_ERROR(SSL, SSL_R_UNSUPPORTED_ELLIPTIC_CURVE);
return 0;
}
ctx->method = method;
return 1;
}
void SSL_ECDH_CTX_init_for_dhe(SSL_ECDH_CTX *ctx, DH *params) {
SSL_ECDH_CTX_cleanup(ctx);
ctx->method = &kDHEMethod;
ctx->data = params;
}
void SSL_ECDH_CTX_init_for_cecpq1(SSL_ECDH_CTX *ctx) {
SSL_ECDH_CTX_cleanup(ctx);
ctx->method = &kCECPQ1Method;
}
void SSL_ECDH_CTX_cleanup(SSL_ECDH_CTX *ctx) {
if (ctx->method == NULL) {
return;
}
ctx->method->cleanup(ctx);
ctx->method = NULL;
ctx->data = NULL;
}
uint16_t SSL_ECDH_CTX_get_id(const SSL_ECDH_CTX *ctx) {
return ctx->method->group_id;
}
int SSL_ECDH_CTX_get_key(SSL_ECDH_CTX *ctx, CBS *cbs, CBS *out) {
if (ctx->method == NULL) {
return 0;
}
return ctx->method->get_key(cbs, out);
}
int SSL_ECDH_CTX_add_key(SSL_ECDH_CTX *ctx, CBB *cbb, CBB *out_contents) {
if (ctx->method == NULL) {
return 0;
}
return ctx->method->add_key(cbb, out_contents);
}
int SSL_ECDH_CTX_offer(SSL_ECDH_CTX *ctx, CBB *out_public_key) {
return ctx->method->offer(ctx, out_public_key);
}
int SSL_ECDH_CTX_accept(SSL_ECDH_CTX *ctx, CBB *out_public_key,
uint8_t **out_secret, size_t *out_secret_len,
uint8_t *out_alert, const uint8_t *peer_key,
size_t peer_key_len) {
return ctx->method->accept(ctx, out_public_key, out_secret, out_secret_len,
out_alert, peer_key, peer_key_len);
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
}
int SSL_ECDH_CTX_finish(SSL_ECDH_CTX *ctx, uint8_t **out_secret,
size_t *out_secret_len, uint8_t *out_alert,
const uint8_t *peer_key, size_t peer_key_len) {
return ctx->method->finish(ctx, out_secret, out_secret_len, out_alert,
peer_key, peer_key_len);
Implement draft-ietf-tls-curve25519-01 in C. The new curve is not enabled by default. As EC_GROUP/EC_POINT is a bit too complex for X25519, this introduces an SSL_ECDH_METHOD abstraction which wraps just the raw ECDH operation. It also tidies up some of the curve code which kept converting back and force between NIDs and curve IDs. Now everything transits as curve IDs except for API entry points (SSL_set1_curves) which take NIDs. Those convert immediately and act on curve IDs from then on. Note that, like the Go implementation, this slightly tweaks the order of operations. The client sees the server public key before sending its own. To keep the abstraction simple, SSL_ECDH_METHOD expects to generate a keypair before consuming the peer's public key. Instead, the client handshake stashes the serialized peer public value and defers parsing it until it comes time to send ClientKeyExchange. (This is analogous to what it was doing before where it stashed the parsed peer public value instead.) It still uses TLS 1.2 terminology everywhere, but this abstraction should also be compatible with TLS 1.3 which unifies (EC)DH-style key exchanges. (Accordingly, this abstraction intentionally does not handle parsing the ClientKeyExchange/ServerKeyExchange framing or attempt to handle asynchronous plain RSA or the authentication bits.) BUG=571231 Change-Id: Iba09dddee5bcdfeb2b70185308e8ab0632717932 Reviewed-on: https://boringssl-review.googlesource.com/6780 Reviewed-by: Adam Langley <agl@google.com>
2015-12-19 05:18:25 +00:00
}