openssl/ssl/d1_lib.c

970 lines
29 KiB
C

/*
* Copyright 2005-2022 The OpenSSL Project Authors. All Rights Reserved.
*
* Licensed under the Apache License 2.0 (the "License"). You may not use
* this file except in compliance with the License. You can obtain a copy
* in the file LICENSE in the source distribution or at
* https://www.openssl.org/source/license.html
*/
#include "internal/e_os.h"
#include <stdio.h>
#include <openssl/objects.h>
#include <openssl/rand.h>
#include "ssl_local.h"
#include "internal/time.h"
static int dtls1_handshake_write(SSL_CONNECTION *s);
static size_t dtls1_link_min_mtu(void);
/* XDTLS: figure out the right values */
static const size_t g_probable_mtu[] = { 1500, 512, 256 };
const SSL3_ENC_METHOD DTLSv1_enc_data = {
tls1_setup_key_block,
tls1_generate_master_secret,
tls1_change_cipher_state,
tls1_final_finish_mac,
TLS_MD_CLIENT_FINISH_CONST, TLS_MD_CLIENT_FINISH_CONST_SIZE,
TLS_MD_SERVER_FINISH_CONST, TLS_MD_SERVER_FINISH_CONST_SIZE,
tls1_alert_code,
tls1_export_keying_material,
SSL_ENC_FLAG_DTLS | SSL_ENC_FLAG_EXPLICIT_IV,
dtls1_set_handshake_header,
dtls1_close_construct_packet,
dtls1_handshake_write
};
const SSL3_ENC_METHOD DTLSv1_2_enc_data = {
tls1_setup_key_block,
tls1_generate_master_secret,
tls1_change_cipher_state,
tls1_final_finish_mac,
TLS_MD_CLIENT_FINISH_CONST, TLS_MD_CLIENT_FINISH_CONST_SIZE,
TLS_MD_SERVER_FINISH_CONST, TLS_MD_SERVER_FINISH_CONST_SIZE,
tls1_alert_code,
tls1_export_keying_material,
SSL_ENC_FLAG_DTLS | SSL_ENC_FLAG_EXPLICIT_IV | SSL_ENC_FLAG_SIGALGS
| SSL_ENC_FLAG_SHA256_PRF | SSL_ENC_FLAG_TLS1_2_CIPHERS,
dtls1_set_handshake_header,
dtls1_close_construct_packet,
dtls1_handshake_write
};
OSSL_TIME dtls1_default_timeout(void)
{
/*
* 2 hours, the 24 hours mentioned in the DTLSv1 spec is way too long for
* http, the cache would over fill
*/
return ossl_seconds2time(60 * 60 * 2);
}
int dtls1_new(SSL *ssl)
{
DTLS1_STATE *d1;
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return 0;
if (!DTLS_RECORD_LAYER_new(&s->rlayer)) {
return 0;
}
if (!ssl3_new(ssl))
return 0;
if ((d1 = OPENSSL_zalloc(sizeof(*d1))) == NULL) {
ssl3_free(ssl);
return 0;
}
d1->buffered_messages = pqueue_new();
d1->sent_messages = pqueue_new();
if (s->server) {
d1->cookie_len = sizeof(s->d1->cookie);
}
d1->link_mtu = 0;
d1->mtu = 0;
if (d1->buffered_messages == NULL || d1->sent_messages == NULL) {
pqueue_free(d1->buffered_messages);
pqueue_free(d1->sent_messages);
OPENSSL_free(d1);
ssl3_free(ssl);
return 0;
}
s->d1 = d1;
if (!ssl->method->ssl_clear(ssl))
return 0;
return 1;
}
static void dtls1_clear_queues(SSL_CONNECTION *s)
{
dtls1_clear_received_buffer(s);
dtls1_clear_sent_buffer(s);
}
void dtls1_clear_received_buffer(SSL_CONNECTION *s)
{
pitem *item = NULL;
hm_fragment *frag = NULL;
while ((item = pqueue_pop(s->d1->buffered_messages)) != NULL) {
frag = (hm_fragment *)item->data;
dtls1_hm_fragment_free(frag);
pitem_free(item);
}
}
void dtls1_clear_sent_buffer(SSL_CONNECTION *s)
{
pitem *item = NULL;
hm_fragment *frag = NULL;
while ((item = pqueue_pop(s->d1->sent_messages)) != NULL) {
frag = (hm_fragment *)item->data;
dtls1_hm_fragment_free(frag);
pitem_free(item);
}
}
void dtls1_free(SSL *ssl)
{
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return;
DTLS_RECORD_LAYER_free(&s->rlayer);
ssl3_free(ssl);
if (s->d1 != NULL) {
dtls1_clear_queues(s);
pqueue_free(s->d1->buffered_messages);
pqueue_free(s->d1->sent_messages);
}
OPENSSL_free(s->d1);
s->d1 = NULL;
}
int dtls1_clear(SSL *ssl)
{
pqueue *buffered_messages;
pqueue *sent_messages;
size_t mtu;
size_t link_mtu;
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return 0;
DTLS_RECORD_LAYER_clear(&s->rlayer);
if (s->d1) {
DTLS_timer_cb timer_cb = s->d1->timer_cb;
buffered_messages = s->d1->buffered_messages;
sent_messages = s->d1->sent_messages;
mtu = s->d1->mtu;
link_mtu = s->d1->link_mtu;
dtls1_clear_queues(s);
memset(s->d1, 0, sizeof(*s->d1));
/* Restore the timer callback from previous state */
s->d1->timer_cb = timer_cb;
if (s->server) {
s->d1->cookie_len = sizeof(s->d1->cookie);
}
if (SSL_get_options(ssl) & SSL_OP_NO_QUERY_MTU) {
s->d1->mtu = mtu;
s->d1->link_mtu = link_mtu;
}
s->d1->buffered_messages = buffered_messages;
s->d1->sent_messages = sent_messages;
}
if (!ssl3_clear(ssl))
return 0;
if (ssl->method->version == DTLS_ANY_VERSION)
s->version = DTLS_MAX_VERSION_INTERNAL;
#ifndef OPENSSL_NO_DTLS1_METHOD
else if (s->options & SSL_OP_CISCO_ANYCONNECT)
s->client_version = s->version = DTLS1_BAD_VER;
#endif
else
s->version = ssl->method->version;
return 1;
}
long dtls1_ctrl(SSL *ssl, int cmd, long larg, void *parg)
{
int ret = 0;
OSSL_TIME t;
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return 0;
switch (cmd) {
case DTLS_CTRL_GET_TIMEOUT:
if (dtls1_get_timeout(s, &t) != NULL) {
*(struct timeval *)parg = ossl_time_to_timeval(t);
ret = 1;
}
break;
case DTLS_CTRL_HANDLE_TIMEOUT:
ret = dtls1_handle_timeout(s);
break;
case DTLS_CTRL_SET_LINK_MTU:
if (larg < (long)dtls1_link_min_mtu())
return 0;
s->d1->link_mtu = larg;
return 1;
case DTLS_CTRL_GET_LINK_MIN_MTU:
return (long)dtls1_link_min_mtu();
case SSL_CTRL_SET_MTU:
/*
* We may not have a BIO set yet so can't call dtls1_min_mtu()
* We'll have to make do with dtls1_link_min_mtu() and max overhead
*/
if (larg < (long)dtls1_link_min_mtu() - DTLS1_MAX_MTU_OVERHEAD)
return 0;
s->d1->mtu = larg;
return larg;
default:
ret = ssl3_ctrl(ssl, cmd, larg, parg);
break;
}
return ret;
}
void dtls1_start_timer(SSL_CONNECTION *s)
{
struct timeval tv;
OSSL_TIME duration;
SSL *ssl = SSL_CONNECTION_GET_SSL(s);
#ifndef OPENSSL_NO_SCTP
/* Disable timer for SCTP */
if (BIO_dgram_is_sctp(SSL_get_wbio(ssl))) {
s->d1->next_timeout = ossl_time_zero();
return;
}
#endif
/*
* If timer is not set, initialize duration with 1 second or
* a user-specified value if the timer callback is installed.
*/
if (ossl_time_is_zero(s->d1->next_timeout)) {
if (s->d1->timer_cb != NULL)
s->d1->timeout_duration_us = s->d1->timer_cb(ssl, 0);
else
s->d1->timeout_duration_us = 1000000;
}
/* Set timeout to current time plus duration */
duration = ossl_us2time(s->d1->timeout_duration_us);
s->d1->next_timeout = ossl_time_add(ossl_time_now(), duration);
tv = ossl_time_to_timeval(s->d1->next_timeout);
BIO_ctrl(SSL_get_rbio(ssl), BIO_CTRL_DGRAM_SET_NEXT_TIMEOUT, 0, &tv);
}
OSSL_TIME *dtls1_get_timeout(SSL_CONNECTION *s, OSSL_TIME *timeleft)
{
OSSL_TIME timenow;
/* If no timeout is set, just return NULL */
if (ossl_time_is_zero(s->d1->next_timeout))
return NULL;
/* Get current time */
timenow = ossl_time_now();
/*
* If timer already expired or if remaining time is less than 15 ms,
* set it to 0 to prevent issues because of small divergences with
* socket timeouts.
*/
*timeleft = ossl_time_subtract(s->d1->next_timeout, timenow);
if (ossl_time_compare(*timeleft, ossl_ms2time(15)) <= 0)
*timeleft = ossl_time_zero();
return timeleft;
}
int dtls1_is_timer_expired(SSL_CONNECTION *s)
{
OSSL_TIME timeleft;
/* Get time left until timeout, return false if no timer running */
if (dtls1_get_timeout(s, &timeleft) == NULL) {
return 0;
}
/* Return false if timer is not expired yet */
if (!ossl_time_is_zero(timeleft)) {
return 0;
}
/* Timer expired, so return true */
return 1;
}
static void dtls1_double_timeout(SSL_CONNECTION *s)
{
s->d1->timeout_duration_us *= 2;
if (s->d1->timeout_duration_us > 60000000)
s->d1->timeout_duration_us = 60000000;
}
void dtls1_stop_timer(SSL_CONNECTION *s)
{
struct timeval tv;
/* Reset everything */
s->d1->timeout_num_alerts = 0;
s->d1->next_timeout = ossl_time_zero();
s->d1->timeout_duration_us = 1000000;
tv = ossl_time_to_timeval(s->d1->next_timeout);
BIO_ctrl(s->rbio, BIO_CTRL_DGRAM_SET_NEXT_TIMEOUT, 0, &tv);
/* Clear retransmission buffer */
dtls1_clear_sent_buffer(s);
}
int dtls1_check_timeout_num(SSL_CONNECTION *s)
{
size_t mtu;
SSL *ssl = SSL_CONNECTION_GET_SSL(s);
s->d1->timeout_num_alerts++;
/* Reduce MTU after 2 unsuccessful retransmissions */
if (s->d1->timeout_num_alerts > 2
&& !(SSL_get_options(ssl) & SSL_OP_NO_QUERY_MTU)) {
mtu =
BIO_ctrl(SSL_get_wbio(ssl), BIO_CTRL_DGRAM_GET_FALLBACK_MTU, 0, NULL);
if (mtu < s->d1->mtu)
s->d1->mtu = mtu;
}
if (s->d1->timeout_num_alerts > DTLS1_TMO_ALERT_COUNT) {
/* fail the connection, enough alerts have been sent */
SSLfatal(s, SSL_AD_NO_ALERT, SSL_R_READ_TIMEOUT_EXPIRED);
return -1;
}
return 0;
}
int dtls1_handle_timeout(SSL_CONNECTION *s)
{
/* if no timer is expired, don't do anything */
if (!dtls1_is_timer_expired(s)) {
return 0;
}
if (s->d1->timer_cb != NULL)
s->d1->timeout_duration_us = s->d1->timer_cb(SSL_CONNECTION_GET_SSL(s),
s->d1->timeout_duration_us);
else
dtls1_double_timeout(s);
if (dtls1_check_timeout_num(s) < 0) {
/* SSLfatal() already called */
return -1;
}
dtls1_start_timer(s);
/* Calls SSLfatal() if required */
return dtls1_retransmit_buffered_messages(s);
}
#define LISTEN_SUCCESS 2
#define LISTEN_SEND_VERIFY_REQUEST 1
#ifndef OPENSSL_NO_SOCK
int DTLSv1_listen(SSL *ssl, BIO_ADDR *client)
{
int next, n, ret = 0;
unsigned char cookie[DTLS1_COOKIE_LENGTH];
unsigned char seq[SEQ_NUM_SIZE];
const unsigned char *data;
unsigned char *buf = NULL, *wbuf;
size_t fragoff, fraglen, msglen;
unsigned int rectype, versmajor, msgseq, msgtype, clientvers, cookielen;
BIO *rbio, *wbio;
BIO_ADDR *tmpclient = NULL;
PACKET pkt, msgpkt, msgpayload, session, cookiepkt;
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return -1;
if (s->handshake_func == NULL) {
/* Not properly initialized yet */
SSL_set_accept_state(ssl);
}
/* Ensure there is no state left over from a previous invocation */
if (!SSL_clear(ssl))
return -1;
ERR_clear_error();
rbio = SSL_get_rbio(ssl);
wbio = SSL_get_wbio(ssl);
if (!rbio || !wbio) {
ERR_raise(ERR_LIB_SSL, SSL_R_BIO_NOT_SET);
return -1;
}
/*
* Note: This check deliberately excludes DTLS1_BAD_VER because that version
* requires the MAC to be calculated *including* the first ClientHello
* (without the cookie). Since DTLSv1_listen is stateless that cannot be
* supported. DTLS1_BAD_VER must use cookies in a stateful manner (e.g. via
* SSL_accept)
*/
if ((s->version & 0xff00) != (DTLS1_VERSION & 0xff00)) {
ERR_raise(ERR_LIB_SSL, SSL_R_UNSUPPORTED_SSL_VERSION);
return -1;
}
buf = OPENSSL_malloc(DTLS1_RT_HEADER_LENGTH + SSL3_RT_MAX_PLAIN_LENGTH);
if (buf == NULL)
return -1;
wbuf = OPENSSL_malloc(DTLS1_RT_HEADER_LENGTH + SSL3_RT_MAX_PLAIN_LENGTH);
if (wbuf == NULL) {
OPENSSL_free(buf);
return -1;
}
do {
/* Get a packet */
clear_sys_error();
n = BIO_read(rbio, buf, SSL3_RT_MAX_PLAIN_LENGTH
+ DTLS1_RT_HEADER_LENGTH);
if (n <= 0) {
if (BIO_should_retry(rbio)) {
/* Non-blocking IO */
goto end;
}
ret = -1;
goto end;
}
if (!PACKET_buf_init(&pkt, buf, n)) {
ERR_raise(ERR_LIB_SSL, ERR_R_INTERNAL_ERROR);
ret = -1;
goto end;
}
/*
* Parse the received record. If there are any problems with it we just
* dump it - with no alert. RFC6347 says this "Unlike TLS, DTLS is
* resilient in the face of invalid records (e.g., invalid formatting,
* length, MAC, etc.). In general, invalid records SHOULD be silently
* discarded, thus preserving the association; however, an error MAY be
* logged for diagnostic purposes."
*/
/* this packet contained a partial record, dump it */
if (n < DTLS1_RT_HEADER_LENGTH) {
ERR_raise(ERR_LIB_SSL, SSL_R_RECORD_TOO_SMALL);
goto end;
}
if (s->msg_callback)
s->msg_callback(0, 0, SSL3_RT_HEADER, buf,
DTLS1_RT_HEADER_LENGTH, ssl, s->msg_callback_arg);
/* Get the record header */
if (!PACKET_get_1(&pkt, &rectype)
|| !PACKET_get_1(&pkt, &versmajor)) {
ERR_raise(ERR_LIB_SSL, SSL_R_LENGTH_MISMATCH);
goto end;
}
if (rectype != SSL3_RT_HANDSHAKE) {
ERR_raise(ERR_LIB_SSL, SSL_R_UNEXPECTED_MESSAGE);
goto end;
}
/*
* Check record version number. We only check that the major version is
* the same.
*/
if (versmajor != DTLS1_VERSION_MAJOR) {
ERR_raise(ERR_LIB_SSL, SSL_R_BAD_PROTOCOL_VERSION_NUMBER);
goto end;
}
if (!PACKET_forward(&pkt, 1)
/* Save the sequence number: 64 bits, with top 2 bytes = epoch */
|| !PACKET_copy_bytes(&pkt, seq, SEQ_NUM_SIZE)
|| !PACKET_get_length_prefixed_2(&pkt, &msgpkt)) {
ERR_raise(ERR_LIB_SSL, SSL_R_LENGTH_MISMATCH);
goto end;
}
/*
* We allow data remaining at the end of the packet because there could
* be a second record (but we ignore it)
*/
/* This is an initial ClientHello so the epoch has to be 0 */
if (seq[0] != 0 || seq[1] != 0) {
ERR_raise(ERR_LIB_SSL, SSL_R_UNEXPECTED_MESSAGE);
goto end;
}
/* Get a pointer to the raw message for the later callback */
data = PACKET_data(&msgpkt);
/* Finished processing the record header, now process the message */
if (!PACKET_get_1(&msgpkt, &msgtype)
|| !PACKET_get_net_3_len(&msgpkt, &msglen)
|| !PACKET_get_net_2(&msgpkt, &msgseq)
|| !PACKET_get_net_3_len(&msgpkt, &fragoff)
|| !PACKET_get_net_3_len(&msgpkt, &fraglen)
|| !PACKET_get_sub_packet(&msgpkt, &msgpayload, fraglen)
|| PACKET_remaining(&msgpkt) != 0) {
ERR_raise(ERR_LIB_SSL, SSL_R_LENGTH_MISMATCH);
goto end;
}
if (msgtype != SSL3_MT_CLIENT_HELLO) {
ERR_raise(ERR_LIB_SSL, SSL_R_UNEXPECTED_MESSAGE);
goto end;
}
/* Message sequence number can only be 0 or 1 */
if (msgseq > 2) {
ERR_raise(ERR_LIB_SSL, SSL_R_INVALID_SEQUENCE_NUMBER);
goto end;
}
/*
* We don't support fragment reassembly for ClientHellos whilst
* listening because that would require server side state (which is
* against the whole point of the ClientHello/HelloVerifyRequest
* mechanism). Instead we only look at the first ClientHello fragment
* and require that the cookie must be contained within it.
*/
if (fragoff != 0 || fraglen > msglen) {
/* Non initial ClientHello fragment (or bad fragment) */
ERR_raise(ERR_LIB_SSL, SSL_R_FRAGMENTED_CLIENT_HELLO);
goto end;
}
if (s->msg_callback)
s->msg_callback(0, s->version, SSL3_RT_HANDSHAKE, data,
fraglen + DTLS1_HM_HEADER_LENGTH, ssl,
s->msg_callback_arg);
if (!PACKET_get_net_2(&msgpayload, &clientvers)) {
ERR_raise(ERR_LIB_SSL, SSL_R_LENGTH_MISMATCH);
goto end;
}
/*
* Verify client version is supported
*/
if (DTLS_VERSION_LT(clientvers, (unsigned int)ssl->method->version) &&
ssl->method->version != DTLS_ANY_VERSION) {
ERR_raise(ERR_LIB_SSL, SSL_R_WRONG_VERSION_NUMBER);
goto end;
}
if (!PACKET_forward(&msgpayload, SSL3_RANDOM_SIZE)
|| !PACKET_get_length_prefixed_1(&msgpayload, &session)
|| !PACKET_get_length_prefixed_1(&msgpayload, &cookiepkt)) {
/*
* Could be malformed or the cookie does not fit within the initial
* ClientHello fragment. Either way we can't handle it.
*/
ERR_raise(ERR_LIB_SSL, SSL_R_LENGTH_MISMATCH);
goto end;
}
/*
* Check if we have a cookie or not. If not we need to send a
* HelloVerifyRequest.
*/
if (PACKET_remaining(&cookiepkt) == 0) {
next = LISTEN_SEND_VERIFY_REQUEST;
} else {
/*
* We have a cookie, so lets check it.
*/
if (ssl->ctx->app_verify_cookie_cb == NULL) {
ERR_raise(ERR_LIB_SSL, SSL_R_NO_VERIFY_COOKIE_CALLBACK);
/* This is fatal */
ret = -1;
goto end;
}
if (ssl->ctx->app_verify_cookie_cb(ssl, PACKET_data(&cookiepkt),
(unsigned int)PACKET_remaining(&cookiepkt)) == 0) {
/*
* We treat invalid cookies in the same was as no cookie as
* per RFC6347
*/
next = LISTEN_SEND_VERIFY_REQUEST;
} else {
/* Cookie verification succeeded */
next = LISTEN_SUCCESS;
}
}
if (next == LISTEN_SEND_VERIFY_REQUEST) {
WPACKET wpkt;
unsigned int version;
size_t wreclen;
/*
* There was no cookie in the ClientHello so we need to send a
* HelloVerifyRequest. If this fails we do not worry about trying
* to resend, we just drop it.
*/
/* Generate the cookie */
if (ssl->ctx->app_gen_cookie_cb == NULL ||
ssl->ctx->app_gen_cookie_cb(ssl, cookie, &cookielen) == 0 ||
cookielen > 255) {
ERR_raise(ERR_LIB_SSL, SSL_R_COOKIE_GEN_CALLBACK_FAILURE);
/* This is fatal */
ret = -1;
goto end;
}
/*
* Special case: for hello verify request, client version 1.0 and we
* haven't decided which version to use yet send back using version
* 1.0 header: otherwise some clients will ignore it.
*/
version = (ssl->method->version == DTLS_ANY_VERSION) ? DTLS1_VERSION
: s->version;
/* Construct the record and message headers */
if (!WPACKET_init_static_len(&wpkt,
wbuf,
ssl_get_max_send_fragment(s)
+ DTLS1_RT_HEADER_LENGTH,
0)
|| !WPACKET_put_bytes_u8(&wpkt, SSL3_RT_HANDSHAKE)
|| !WPACKET_put_bytes_u16(&wpkt, version)
/*
* Record sequence number is always the same as in the
* received ClientHello
*/
|| !WPACKET_memcpy(&wpkt, seq, SEQ_NUM_SIZE)
/* End of record, start sub packet for message */
|| !WPACKET_start_sub_packet_u16(&wpkt)
/* Message type */
|| !WPACKET_put_bytes_u8(&wpkt,
DTLS1_MT_HELLO_VERIFY_REQUEST)
/*
* Message length - doesn't follow normal TLS convention:
* the length isn't the last thing in the message header.
* We'll need to fill this in later when we know the
* length. Set it to zero for now
*/
|| !WPACKET_put_bytes_u24(&wpkt, 0)
/*
* Message sequence number is always 0 for a
* HelloVerifyRequest
*/
|| !WPACKET_put_bytes_u16(&wpkt, 0)
/*
* We never fragment a HelloVerifyRequest, so fragment
* offset is 0
*/
|| !WPACKET_put_bytes_u24(&wpkt, 0)
/*
* Fragment length is the same as message length, but
* this *is* the last thing in the message header so we
* can just start a sub-packet. No need to come back
* later for this one.
*/
|| !WPACKET_start_sub_packet_u24(&wpkt)
/* Create the actual HelloVerifyRequest body */
|| !dtls_raw_hello_verify_request(&wpkt, cookie, cookielen)
/* Close message body */
|| !WPACKET_close(&wpkt)
/* Close record body */
|| !WPACKET_close(&wpkt)
|| !WPACKET_get_total_written(&wpkt, &wreclen)
|| !WPACKET_finish(&wpkt)) {
ERR_raise(ERR_LIB_SSL, ERR_R_INTERNAL_ERROR);
WPACKET_cleanup(&wpkt);
/* This is fatal */
ret = -1;
goto end;
}
/*
* Fix up the message len in the message header. Its the same as the
* fragment len which has been filled in by WPACKET, so just copy
* that. Destination for the message len is after the record header
* plus one byte for the message content type. The source is the
* last 3 bytes of the message header
*/
memcpy(&wbuf[DTLS1_RT_HEADER_LENGTH + 1],
&wbuf[DTLS1_RT_HEADER_LENGTH + DTLS1_HM_HEADER_LENGTH - 3],
3);
if (s->msg_callback)
s->msg_callback(1, 0, SSL3_RT_HEADER, buf,
DTLS1_RT_HEADER_LENGTH, ssl,
s->msg_callback_arg);
if ((tmpclient = BIO_ADDR_new()) == NULL) {
ERR_raise(ERR_LIB_SSL, ERR_R_BIO_LIB);
goto end;
}
/*
* This is unnecessary if rbio and wbio are one and the same - but
* maybe they're not. We ignore errors here - some BIOs do not
* support this.
*/
if (BIO_dgram_get_peer(rbio, tmpclient) > 0) {
(void)BIO_dgram_set_peer(wbio, tmpclient);
}
BIO_ADDR_free(tmpclient);
tmpclient = NULL;
if (BIO_write(wbio, wbuf, wreclen) < (int)wreclen) {
if (BIO_should_retry(wbio)) {
/*
* Non-blocking IO...but we're stateless, so we're just
* going to drop this packet.
*/
goto end;
}
ret = -1;
goto end;
}
if (BIO_flush(wbio) <= 0) {
if (BIO_should_retry(wbio)) {
/*
* Non-blocking IO...but we're stateless, so we're just
* going to drop this packet.
*/
goto end;
}
ret = -1;
goto end;
}
}
} while (next != LISTEN_SUCCESS);
/*
* Set expected sequence numbers to continue the handshake.
*/
s->d1->handshake_read_seq = 1;
s->d1->handshake_write_seq = 1;
s->d1->next_handshake_write_seq = 1;
s->rlayer.wrlmethod->increment_sequence_ctr(s->rlayer.wrl);
/*
* We are doing cookie exchange, so make sure we set that option in the
* SSL object
*/
SSL_set_options(ssl, SSL_OP_COOKIE_EXCHANGE);
/*
* Tell the state machine that we've done the initial hello verify
* exchange
*/
ossl_statem_set_hello_verify_done(s);
/*
* Some BIOs may not support this. If we fail we clear the client address
*/
if (BIO_dgram_get_peer(rbio, client) <= 0)
BIO_ADDR_clear(client);
/* Buffer the record for use by the record layer */
if (BIO_write(s->rlayer.rrlnext, buf, n) != n) {
ERR_raise(ERR_LIB_SSL, ERR_R_INTERNAL_ERROR);
ret = -1;
goto end;
}
/*
* Reset the record layer - but this time we can use the record we just
* buffered in s->rlayer.rrlnext
*/
if (!ssl_set_new_record_layer(s,
DTLS_ANY_VERSION,
OSSL_RECORD_DIRECTION_READ,
OSSL_RECORD_PROTECTION_LEVEL_NONE,
NULL, 0, NULL, 0, NULL, 0, NULL, 0,
NID_undef, NULL, NULL)) {
/* SSLfatal already called */
ret = -1;
goto end;
}
ret = 1;
end:
BIO_ADDR_free(tmpclient);
OPENSSL_free(buf);
OPENSSL_free(wbuf);
return ret;
}
#endif
static int dtls1_handshake_write(SSL_CONNECTION *s)
{
return dtls1_do_write(s, SSL3_RT_HANDSHAKE);
}
int dtls1_shutdown(SSL *s)
{
int ret;
#ifndef OPENSSL_NO_SCTP
BIO *wbio;
SSL_CONNECTION *sc = SSL_CONNECTION_FROM_SSL_ONLY(s);
if (s == NULL)
return -1;
wbio = SSL_get_wbio(s);
if (wbio != NULL && BIO_dgram_is_sctp(wbio) &&
!(sc->shutdown & SSL_SENT_SHUTDOWN)) {
ret = BIO_dgram_sctp_wait_for_dry(wbio);
if (ret < 0)
return -1;
if (ret == 0)
BIO_ctrl(SSL_get_wbio(s), BIO_CTRL_DGRAM_SCTP_SAVE_SHUTDOWN, 1,
NULL);
}
#endif
ret = ssl3_shutdown(s);
#ifndef OPENSSL_NO_SCTP
BIO_ctrl(SSL_get_wbio(s), BIO_CTRL_DGRAM_SCTP_SAVE_SHUTDOWN, 0, NULL);
#endif
return ret;
}
int dtls1_query_mtu(SSL_CONNECTION *s)
{
SSL *ssl = SSL_CONNECTION_GET_SSL(s);
if (s->d1->link_mtu) {
s->d1->mtu =
s->d1->link_mtu - BIO_dgram_get_mtu_overhead(SSL_get_wbio(ssl));
s->d1->link_mtu = 0;
}
/* AHA! Figure out the MTU, and stick to the right size */
if (s->d1->mtu < dtls1_min_mtu(s)) {
if (!(SSL_get_options(ssl) & SSL_OP_NO_QUERY_MTU)) {
s->d1->mtu =
BIO_ctrl(SSL_get_wbio(ssl), BIO_CTRL_DGRAM_QUERY_MTU, 0, NULL);
/*
* I've seen the kernel return bogus numbers when it doesn't know
* (initial write), so just make sure we have a reasonable number
*/
if (s->d1->mtu < dtls1_min_mtu(s)) {
/* Set to min mtu */
s->d1->mtu = dtls1_min_mtu(s);
BIO_ctrl(SSL_get_wbio(ssl), BIO_CTRL_DGRAM_SET_MTU,
(long)s->d1->mtu, NULL);
}
} else
return 0;
}
return 1;
}
static size_t dtls1_link_min_mtu(void)
{
return (g_probable_mtu[(sizeof(g_probable_mtu) /
sizeof(g_probable_mtu[0])) - 1]);
}
size_t dtls1_min_mtu(SSL_CONNECTION *s)
{
SSL *ssl = SSL_CONNECTION_GET_SSL(s);
return dtls1_link_min_mtu() - BIO_dgram_get_mtu_overhead(SSL_get_wbio(ssl));
}
size_t DTLS_get_data_mtu(const SSL *ssl)
{
size_t mac_overhead, int_overhead, blocksize, ext_overhead;
const SSL_CIPHER *ciph = SSL_get_current_cipher(ssl);
size_t mtu;
const SSL_CONNECTION *s = SSL_CONNECTION_FROM_CONST_SSL_ONLY(ssl);
if (s == NULL)
return 0;
mtu = s->d1->mtu;
if (ciph == NULL)
return 0;
if (!ssl_cipher_get_overhead(ciph, &mac_overhead, &int_overhead,
&blocksize, &ext_overhead))
return 0;
if (SSL_READ_ETM(s))
ext_overhead += mac_overhead;
else
int_overhead += mac_overhead;
/* Subtract external overhead (e.g. IV/nonce, separate MAC) */
if (ext_overhead + DTLS1_RT_HEADER_LENGTH >= mtu)
return 0;
mtu -= ext_overhead + DTLS1_RT_HEADER_LENGTH;
/* Round encrypted payload down to cipher block size (for CBC etc.)
* No check for overflow since 'mtu % blocksize' cannot exceed mtu. */
if (blocksize)
mtu -= (mtu % blocksize);
/* Subtract internal overhead (e.g. CBC padding len byte) */
if (int_overhead >= mtu)
return 0;
mtu -= int_overhead;
return mtu;
}
void DTLS_set_timer_cb(SSL *ssl, DTLS_timer_cb cb)
{
SSL_CONNECTION *s = SSL_CONNECTION_FROM_SSL_ONLY(ssl);
if (s == NULL)
return;
s->d1->timer_cb = cb;
}