openssl/crypto/bio/bss_dgram.c
Matt Caswell 9924087573 Fix DTLS failure when used in a build which has SCTP enabled
The value of BIO_CTRL_DGRAM_SET_PEEK_MODE was clashing with the value for
BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE. In an SCTP enabled build
BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE was used unconditionally with
the reasoning that it would be ignored if SCTP wasn't in use. Unfortunately
due to this clash, this wasn't the case. The BIO ended up going into peek
mode and was continually reading the same data over and over - throwing it
away as a replay.

Fixes #3723

Reviewed-by: Rich Salz <rsalz@openssl.org>
(Merged from https://github.com/openssl/openssl/pull/3724)
2017-06-21 14:56:24 +01:00

1932 lines
57 KiB
C

/*
* Copyright 2005-2016 The OpenSSL Project Authors. All Rights Reserved.
*
* Licensed under the OpenSSL license (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 <stdio.h>
#include <errno.h>
#include "bio_lcl.h"
#ifndef OPENSSL_NO_DGRAM
# if !defined(_WIN32)
# include <sys/time.h>
# endif
# ifndef OPENSSL_NO_SCTP
# include <netinet/sctp.h>
# include <fcntl.h>
# define OPENSSL_SCTP_DATA_CHUNK_TYPE 0x00
# define OPENSSL_SCTP_FORWARD_CUM_TSN_CHUNK_TYPE 0xc0
# endif
# if defined(OPENSSL_SYS_LINUX) && !defined(IP_MTU)
# define IP_MTU 14 /* linux is lame */
# endif
# if OPENSSL_USE_IPV6 && !defined(IPPROTO_IPV6)
# define IPPROTO_IPV6 41 /* windows is lame */
# endif
# if defined(__FreeBSD__) && defined(IN6_IS_ADDR_V4MAPPED)
/* Standard definition causes type-punning problems. */
# undef IN6_IS_ADDR_V4MAPPED
# define s6_addr32 __u6_addr.__u6_addr32
# define IN6_IS_ADDR_V4MAPPED(a) \
(((a)->s6_addr32[0] == 0) && \
((a)->s6_addr32[1] == 0) && \
((a)->s6_addr32[2] == htonl(0x0000ffff)))
# endif
static int dgram_write(BIO *h, const char *buf, int num);
static int dgram_read(BIO *h, char *buf, int size);
static int dgram_puts(BIO *h, const char *str);
static long dgram_ctrl(BIO *h, int cmd, long arg1, void *arg2);
static int dgram_new(BIO *h);
static int dgram_free(BIO *data);
static int dgram_clear(BIO *bio);
# ifndef OPENSSL_NO_SCTP
static int dgram_sctp_write(BIO *h, const char *buf, int num);
static int dgram_sctp_read(BIO *h, char *buf, int size);
static int dgram_sctp_puts(BIO *h, const char *str);
static long dgram_sctp_ctrl(BIO *h, int cmd, long arg1, void *arg2);
static int dgram_sctp_new(BIO *h);
static int dgram_sctp_free(BIO *data);
# ifdef SCTP_AUTHENTICATION_EVENT
static void dgram_sctp_handle_auth_free_key_event(BIO *b, union sctp_notification
*snp);
# endif
# endif
static int BIO_dgram_should_retry(int s);
static void get_current_time(struct timeval *t);
static const BIO_METHOD methods_dgramp = {
BIO_TYPE_DGRAM,
"datagram socket",
/* TODO: Convert to new style write function */
bwrite_conv,
dgram_write,
/* TODO: Convert to new style read function */
bread_conv,
dgram_read,
dgram_puts,
NULL, /* dgram_gets, */
dgram_ctrl,
dgram_new,
dgram_free,
NULL,
};
# ifndef OPENSSL_NO_SCTP
static const BIO_METHOD methods_dgramp_sctp = {
BIO_TYPE_DGRAM_SCTP,
"datagram sctp socket",
/* TODO: Convert to new style write function */
bwrite_conv,
dgram_sctp_write,
/* TODO: Convert to new style write function */
bread_conv,
dgram_sctp_read,
dgram_sctp_puts,
NULL, /* dgram_gets, */
dgram_sctp_ctrl,
dgram_sctp_new,
dgram_sctp_free,
NULL,
};
# endif
typedef struct bio_dgram_data_st {
BIO_ADDR peer;
unsigned int connected;
unsigned int _errno;
unsigned int mtu;
struct timeval next_timeout;
struct timeval socket_timeout;
unsigned int peekmode;
} bio_dgram_data;
# ifndef OPENSSL_NO_SCTP
typedef struct bio_dgram_sctp_save_message_st {
BIO *bio;
char *data;
int length;
} bio_dgram_sctp_save_message;
typedef struct bio_dgram_sctp_data_st {
BIO_ADDR peer;
unsigned int connected;
unsigned int _errno;
unsigned int mtu;
struct bio_dgram_sctp_sndinfo sndinfo;
struct bio_dgram_sctp_rcvinfo rcvinfo;
struct bio_dgram_sctp_prinfo prinfo;
void (*handle_notifications) (BIO *bio, void *context, void *buf);
void *notification_context;
int in_handshake;
int ccs_rcvd;
int ccs_sent;
int save_shutdown;
int peer_auth_tested;
} bio_dgram_sctp_data;
# endif
const BIO_METHOD *BIO_s_datagram(void)
{
return (&methods_dgramp);
}
BIO *BIO_new_dgram(int fd, int close_flag)
{
BIO *ret;
ret = BIO_new(BIO_s_datagram());
if (ret == NULL)
return (NULL);
BIO_set_fd(ret, fd, close_flag);
return (ret);
}
static int dgram_new(BIO *bi)
{
bio_dgram_data *data = OPENSSL_zalloc(sizeof(*data));
if (data == NULL)
return 0;
bi->ptr = data;
return (1);
}
static int dgram_free(BIO *a)
{
bio_dgram_data *data;
if (a == NULL)
return (0);
if (!dgram_clear(a))
return 0;
data = (bio_dgram_data *)a->ptr;
OPENSSL_free(data);
return (1);
}
static int dgram_clear(BIO *a)
{
if (a == NULL)
return (0);
if (a->shutdown) {
if (a->init) {
BIO_closesocket(a->num);
}
a->init = 0;
a->flags = 0;
}
return (1);
}
static void dgram_adjust_rcv_timeout(BIO *b)
{
# if defined(SO_RCVTIMEO)
bio_dgram_data *data = (bio_dgram_data *)b->ptr;
union {
size_t s;
int i;
} sz = {
0
};
/* Is a timer active? */
if (data->next_timeout.tv_sec > 0 || data->next_timeout.tv_usec > 0) {
struct timeval timenow, timeleft;
/* Read current socket timeout */
# ifdef OPENSSL_SYS_WINDOWS
int timeout;
sz.i = sizeof(timeout);
if (getsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
(void *)&timeout, &sz.i) < 0) {
perror("getsockopt");
} else {
data->socket_timeout.tv_sec = timeout / 1000;
data->socket_timeout.tv_usec = (timeout % 1000) * 1000;
}
# else
sz.i = sizeof(data->socket_timeout);
if (getsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
&(data->socket_timeout), (void *)&sz) < 0) {
perror("getsockopt");
} else if (sizeof(sz.s) != sizeof(sz.i) && sz.i == 0)
OPENSSL_assert(sz.s <= sizeof(data->socket_timeout));
# endif
/* Get current time */
get_current_time(&timenow);
/* Calculate time left until timer expires */
memcpy(&timeleft, &(data->next_timeout), sizeof(struct timeval));
if (timeleft.tv_usec < timenow.tv_usec) {
timeleft.tv_usec = 1000000 - timenow.tv_usec + timeleft.tv_usec;
timeleft.tv_sec--;
} else {
timeleft.tv_usec -= timenow.tv_usec;
}
if (timeleft.tv_sec < timenow.tv_sec) {
timeleft.tv_sec = 0;
timeleft.tv_usec = 1;
} else {
timeleft.tv_sec -= timenow.tv_sec;
}
/*
* Adjust socket timeout if next handshake message timer will expire
* earlier.
*/
if ((data->socket_timeout.tv_sec == 0
&& data->socket_timeout.tv_usec == 0)
|| (data->socket_timeout.tv_sec > timeleft.tv_sec)
|| (data->socket_timeout.tv_sec == timeleft.tv_sec
&& data->socket_timeout.tv_usec >= timeleft.tv_usec)) {
# ifdef OPENSSL_SYS_WINDOWS
timeout = timeleft.tv_sec * 1000 + timeleft.tv_usec / 1000;
if (setsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
(void *)&timeout, sizeof(timeout)) < 0) {
perror("setsockopt");
}
# else
if (setsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO, &timeleft,
sizeof(struct timeval)) < 0) {
perror("setsockopt");
}
# endif
}
}
# endif
}
static void dgram_reset_rcv_timeout(BIO *b)
{
# if defined(SO_RCVTIMEO)
bio_dgram_data *data = (bio_dgram_data *)b->ptr;
/* Is a timer active? */
if (data->next_timeout.tv_sec > 0 || data->next_timeout.tv_usec > 0) {
# ifdef OPENSSL_SYS_WINDOWS
int timeout = data->socket_timeout.tv_sec * 1000 +
data->socket_timeout.tv_usec / 1000;
if (setsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
(void *)&timeout, sizeof(timeout)) < 0) {
perror("setsockopt");
}
# else
if (setsockopt
(b->num, SOL_SOCKET, SO_RCVTIMEO, &(data->socket_timeout),
sizeof(struct timeval)) < 0) {
perror("setsockopt");
}
# endif
}
# endif
}
static int dgram_read(BIO *b, char *out, int outl)
{
int ret = 0;
bio_dgram_data *data = (bio_dgram_data *)b->ptr;
int flags = 0;
BIO_ADDR peer;
socklen_t len = sizeof(peer);
if (out != NULL) {
clear_socket_error();
memset(&peer, 0, sizeof(peer));
dgram_adjust_rcv_timeout(b);
if (data->peekmode)
flags = MSG_PEEK;
ret = recvfrom(b->num, out, outl, flags,
BIO_ADDR_sockaddr_noconst(&peer), &len);
if (!data->connected && ret >= 0)
BIO_ctrl(b, BIO_CTRL_DGRAM_SET_PEER, 0, &peer);
BIO_clear_retry_flags(b);
if (ret < 0) {
if (BIO_dgram_should_retry(ret)) {
BIO_set_retry_read(b);
data->_errno = get_last_socket_error();
}
}
dgram_reset_rcv_timeout(b);
}
return (ret);
}
static int dgram_write(BIO *b, const char *in, int inl)
{
int ret;
bio_dgram_data *data = (bio_dgram_data *)b->ptr;
clear_socket_error();
if (data->connected)
ret = writesocket(b->num, in, inl);
else {
int peerlen = BIO_ADDR_sockaddr_size(&data->peer);
# if defined(NETWARE_CLIB) && defined(NETWARE_BSDSOCK)
ret = sendto(b->num, (char *)in, inl, 0,
BIO_ADDR_sockaddr(&data->peer), peerlen);
# else
ret = sendto(b->num, in, inl, 0,
BIO_ADDR_sockaddr(&data->peer), peerlen);
# endif
}
BIO_clear_retry_flags(b);
if (ret <= 0) {
if (BIO_dgram_should_retry(ret)) {
BIO_set_retry_write(b);
data->_errno = get_last_socket_error();
}
}
return (ret);
}
static long dgram_get_mtu_overhead(bio_dgram_data *data)
{
long ret;
switch (BIO_ADDR_family(&data->peer)) {
case AF_INET:
/*
* Assume this is UDP - 20 bytes for IP, 8 bytes for UDP
*/
ret = 28;
break;
# ifdef AF_INET6
case AF_INET6:
{
# ifdef IN6_IS_ADDR_V4MAPPED
struct in6_addr tmp_addr;
if (BIO_ADDR_rawaddress(&data->peer, &tmp_addr, NULL)
&& IN6_IS_ADDR_V4MAPPED(&tmp_addr))
/*
* Assume this is UDP - 20 bytes for IP, 8 bytes for UDP
*/
ret = 28;
else
# endif
/*
* Assume this is UDP - 40 bytes for IP, 8 bytes for UDP
*/
ret = 48;
}
break;
# endif
default:
/* We don't know. Go with the historical default */
ret = 28;
break;
}
return ret;
}
static long dgram_ctrl(BIO *b, int cmd, long num, void *ptr)
{
long ret = 1;
int *ip;
bio_dgram_data *data = NULL;
int sockopt_val = 0;
int d_errno;
# if defined(OPENSSL_SYS_LINUX) && (defined(IP_MTU_DISCOVER) || defined(IP_MTU))
socklen_t sockopt_len; /* assume that system supporting IP_MTU is
* modern enough to define socklen_t */
socklen_t addr_len;
BIO_ADDR addr;
# endif
data = (bio_dgram_data *)b->ptr;
switch (cmd) {
case BIO_CTRL_RESET:
num = 0;
ret = 0;
break;
case BIO_CTRL_INFO:
ret = 0;
break;
case BIO_C_SET_FD:
dgram_clear(b);
b->num = *((int *)ptr);
b->shutdown = (int)num;
b->init = 1;
break;
case BIO_C_GET_FD:
if (b->init) {
ip = (int *)ptr;
if (ip != NULL)
*ip = b->num;
ret = b->num;
} else
ret = -1;
break;
case BIO_CTRL_GET_CLOSE:
ret = b->shutdown;
break;
case BIO_CTRL_SET_CLOSE:
b->shutdown = (int)num;
break;
case BIO_CTRL_PENDING:
case BIO_CTRL_WPENDING:
ret = 0;
break;
case BIO_CTRL_DUP:
case BIO_CTRL_FLUSH:
ret = 1;
break;
case BIO_CTRL_DGRAM_CONNECT:
BIO_ADDR_make(&data->peer, BIO_ADDR_sockaddr((BIO_ADDR *)ptr));
break;
/* (Linux)kernel sets DF bit on outgoing IP packets */
case BIO_CTRL_DGRAM_MTU_DISCOVER:
# if defined(OPENSSL_SYS_LINUX) && defined(IP_MTU_DISCOVER) && defined(IP_PMTUDISC_DO)
addr_len = (socklen_t) sizeof(addr);
memset(&addr, 0, sizeof(addr));
if (getsockname(b->num, &addr.sa, &addr_len) < 0) {
ret = 0;
break;
}
switch (addr.sa.sa_family) {
case AF_INET:
sockopt_val = IP_PMTUDISC_DO;
if ((ret = setsockopt(b->num, IPPROTO_IP, IP_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0)
perror("setsockopt");
break;
# if OPENSSL_USE_IPV6 && defined(IPV6_MTU_DISCOVER) && defined(IPV6_PMTUDISC_DO)
case AF_INET6:
sockopt_val = IPV6_PMTUDISC_DO;
if ((ret = setsockopt(b->num, IPPROTO_IPV6, IPV6_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0)
perror("setsockopt");
break;
# endif
default:
ret = -1;
break;
}
# else
ret = -1;
# endif
break;
case BIO_CTRL_DGRAM_QUERY_MTU:
# if defined(OPENSSL_SYS_LINUX) && defined(IP_MTU)
addr_len = (socklen_t) sizeof(addr);
memset(&addr, 0, sizeof(addr));
if (getsockname(b->num, &addr.sa, &addr_len) < 0) {
ret = 0;
break;
}
sockopt_len = sizeof(sockopt_val);
switch (addr.sa.sa_family) {
case AF_INET:
if ((ret =
getsockopt(b->num, IPPROTO_IP, IP_MTU, (void *)&sockopt_val,
&sockopt_len)) < 0 || sockopt_val < 0) {
ret = 0;
} else {
/*
* we assume that the transport protocol is UDP and no IP
* options are used.
*/
data->mtu = sockopt_val - 8 - 20;
ret = data->mtu;
}
break;
# if OPENSSL_USE_IPV6 && defined(IPV6_MTU)
case AF_INET6:
if ((ret =
getsockopt(b->num, IPPROTO_IPV6, IPV6_MTU,
(void *)&sockopt_val, &sockopt_len)) < 0
|| sockopt_val < 0) {
ret = 0;
} else {
/*
* we assume that the transport protocol is UDP and no IPV6
* options are used.
*/
data->mtu = sockopt_val - 8 - 40;
ret = data->mtu;
}
break;
# endif
default:
ret = 0;
break;
}
# else
ret = 0;
# endif
break;
case BIO_CTRL_DGRAM_GET_FALLBACK_MTU:
ret = -dgram_get_mtu_overhead(data);
switch (BIO_ADDR_family(&data->peer)) {
case AF_INET:
ret += 576;
break;
# if OPENSSL_USE_IPV6
case AF_INET6:
{
# ifdef IN6_IS_ADDR_V4MAPPED
struct in6_addr tmp_addr;
if (BIO_ADDR_rawaddress(&data->peer, &tmp_addr, NULL)
&& IN6_IS_ADDR_V4MAPPED(&tmp_addr))
ret += 576;
else
# endif
ret += 1280;
}
break;
# endif
default:
ret += 576;
break;
}
break;
case BIO_CTRL_DGRAM_GET_MTU:
return data->mtu;
case BIO_CTRL_DGRAM_SET_MTU:
data->mtu = num;
ret = num;
break;
case BIO_CTRL_DGRAM_SET_CONNECTED:
if (ptr != NULL) {
data->connected = 1;
BIO_ADDR_make(&data->peer, BIO_ADDR_sockaddr((BIO_ADDR *)ptr));
} else {
data->connected = 0;
memset(&data->peer, 0, sizeof(data->peer));
}
break;
case BIO_CTRL_DGRAM_GET_PEER:
ret = BIO_ADDR_sockaddr_size(&data->peer);
/* FIXME: if num < ret, we will only return part of an address.
That should bee an error, no? */
if (num == 0 || num > ret)
num = ret;
memcpy(ptr, &data->peer, (ret = num));
break;
case BIO_CTRL_DGRAM_SET_PEER:
BIO_ADDR_make(&data->peer, BIO_ADDR_sockaddr((BIO_ADDR *)ptr));
break;
case BIO_CTRL_DGRAM_SET_NEXT_TIMEOUT:
memcpy(&(data->next_timeout), ptr, sizeof(struct timeval));
break;
# if defined(SO_RCVTIMEO)
case BIO_CTRL_DGRAM_SET_RECV_TIMEOUT:
# ifdef OPENSSL_SYS_WINDOWS
{
struct timeval *tv = (struct timeval *)ptr;
int timeout = tv->tv_sec * 1000 + tv->tv_usec / 1000;
if (setsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
(void *)&timeout, sizeof(timeout)) < 0) {
perror("setsockopt");
ret = -1;
}
}
# else
if (setsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO, ptr,
sizeof(struct timeval)) < 0) {
perror("setsockopt");
ret = -1;
}
# endif
break;
case BIO_CTRL_DGRAM_GET_RECV_TIMEOUT:
{
union {
size_t s;
int i;
} sz = {
0
};
# ifdef OPENSSL_SYS_WINDOWS
int timeout;
struct timeval *tv = (struct timeval *)ptr;
sz.i = sizeof(timeout);
if (getsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
(void *)&timeout, &sz.i) < 0) {
perror("getsockopt");
ret = -1;
} else {
tv->tv_sec = timeout / 1000;
tv->tv_usec = (timeout % 1000) * 1000;
ret = sizeof(*tv);
}
# else
sz.i = sizeof(struct timeval);
if (getsockopt(b->num, SOL_SOCKET, SO_RCVTIMEO,
ptr, (void *)&sz) < 0) {
perror("getsockopt");
ret = -1;
} else if (sizeof(sz.s) != sizeof(sz.i) && sz.i == 0) {
OPENSSL_assert(sz.s <= sizeof(struct timeval));
ret = (int)sz.s;
} else
ret = sz.i;
# endif
}
break;
# endif
# if defined(SO_SNDTIMEO)
case BIO_CTRL_DGRAM_SET_SEND_TIMEOUT:
# ifdef OPENSSL_SYS_WINDOWS
{
struct timeval *tv = (struct timeval *)ptr;
int timeout = tv->tv_sec * 1000 + tv->tv_usec / 1000;
if (setsockopt(b->num, SOL_SOCKET, SO_SNDTIMEO,
(void *)&timeout, sizeof(timeout)) < 0) {
perror("setsockopt");
ret = -1;
}
}
# else
if (setsockopt(b->num, SOL_SOCKET, SO_SNDTIMEO, ptr,
sizeof(struct timeval)) < 0) {
perror("setsockopt");
ret = -1;
}
# endif
break;
case BIO_CTRL_DGRAM_GET_SEND_TIMEOUT:
{
union {
size_t s;
int i;
} sz = {
0
};
# ifdef OPENSSL_SYS_WINDOWS
int timeout;
struct timeval *tv = (struct timeval *)ptr;
sz.i = sizeof(timeout);
if (getsockopt(b->num, SOL_SOCKET, SO_SNDTIMEO,
(void *)&timeout, &sz.i) < 0) {
perror("getsockopt");
ret = -1;
} else {
tv->tv_sec = timeout / 1000;
tv->tv_usec = (timeout % 1000) * 1000;
ret = sizeof(*tv);
}
# else
sz.i = sizeof(struct timeval);
if (getsockopt(b->num, SOL_SOCKET, SO_SNDTIMEO,
ptr, (void *)&sz) < 0) {
perror("getsockopt");
ret = -1;
} else if (sizeof(sz.s) != sizeof(sz.i) && sz.i == 0) {
OPENSSL_assert(sz.s <= sizeof(struct timeval));
ret = (int)sz.s;
} else
ret = sz.i;
# endif
}
break;
# endif
case BIO_CTRL_DGRAM_GET_SEND_TIMER_EXP:
/* fall-through */
case BIO_CTRL_DGRAM_GET_RECV_TIMER_EXP:
# ifdef OPENSSL_SYS_WINDOWS
d_errno = (data->_errno == WSAETIMEDOUT);
# else
d_errno = (data->_errno == EAGAIN);
# endif
if (d_errno) {
ret = 1;
data->_errno = 0;
} else
ret = 0;
break;
# ifdef EMSGSIZE
case BIO_CTRL_DGRAM_MTU_EXCEEDED:
if (data->_errno == EMSGSIZE) {
ret = 1;
data->_errno = 0;
} else
ret = 0;
break;
# endif
case BIO_CTRL_DGRAM_SET_DONT_FRAG:
sockopt_val = num ? 1 : 0;
switch (data->peer.sa.sa_family) {
case AF_INET:
# if defined(IP_DONTFRAG)
if ((ret = setsockopt(b->num, IPPROTO_IP, IP_DONTFRAG,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_LINUX) && defined(IP_MTU_DISCOVER) && defined (IP_PMTUDISC_PROBE)
if ((sockopt_val = num ? IP_PMTUDISC_PROBE : IP_PMTUDISC_DONT),
(ret = setsockopt(b->num, IPPROTO_IP, IP_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_WINDOWS) && defined(IP_DONTFRAGMENT)
if ((ret = setsockopt(b->num, IPPROTO_IP, IP_DONTFRAGMENT,
(const char *)&sockopt_val,
sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# else
ret = -1;
# endif
break;
# if OPENSSL_USE_IPV6
case AF_INET6:
# if defined(IPV6_DONTFRAG)
if ((ret = setsockopt(b->num, IPPROTO_IPV6, IPV6_DONTFRAG,
(const void *)&sockopt_val,
sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# elif defined(OPENSSL_SYS_LINUX) && defined(IPV6_MTUDISCOVER)
if ((sockopt_val = num ? IP_PMTUDISC_PROBE : IP_PMTUDISC_DONT),
(ret = setsockopt(b->num, IPPROTO_IPV6, IPV6_MTU_DISCOVER,
&sockopt_val, sizeof(sockopt_val))) < 0) {
perror("setsockopt");
ret = -1;
}
# else
ret = -1;
# endif
break;
# endif
default:
ret = -1;
break;
}
break;
case BIO_CTRL_DGRAM_GET_MTU_OVERHEAD:
ret = dgram_get_mtu_overhead(data);
break;
/*
* BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE is used here for compatibility
* reasons. When BIO_CTRL_DGRAM_SET_PEEK_MODE was first defined its value
* was incorrectly clashing with BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE. The
* value has been updated to a non-clashing value. However to preserve
* binary compatiblity we now respond to both the old value and the new one
*/
case BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE:
case BIO_CTRL_DGRAM_SET_PEEK_MODE:
data->peekmode = (unsigned int)num;
break;
default:
ret = 0;
break;
}
return (ret);
}
static int dgram_puts(BIO *bp, const char *str)
{
int n, ret;
n = strlen(str);
ret = dgram_write(bp, str, n);
return (ret);
}
# ifndef OPENSSL_NO_SCTP
const BIO_METHOD *BIO_s_datagram_sctp(void)
{
return (&methods_dgramp_sctp);
}
BIO *BIO_new_dgram_sctp(int fd, int close_flag)
{
BIO *bio;
int ret, optval = 20000;
int auth_data = 0, auth_forward = 0;
unsigned char *p;
struct sctp_authchunk auth;
struct sctp_authchunks *authchunks;
socklen_t sockopt_len;
# ifdef SCTP_AUTHENTICATION_EVENT
# ifdef SCTP_EVENT
struct sctp_event event;
# else
struct sctp_event_subscribe event;
# endif
# endif
bio = BIO_new(BIO_s_datagram_sctp());
if (bio == NULL)
return (NULL);
BIO_set_fd(bio, fd, close_flag);
/* Activate SCTP-AUTH for DATA and FORWARD-TSN chunks */
auth.sauth_chunk = OPENSSL_SCTP_DATA_CHUNK_TYPE;
ret =
setsockopt(fd, IPPROTO_SCTP, SCTP_AUTH_CHUNK, &auth,
sizeof(struct sctp_authchunk));
if (ret < 0) {
BIO_vfree(bio);
BIOerr(BIO_F_BIO_NEW_DGRAM_SCTP, ERR_R_SYS_LIB);
ERR_add_error_data(1, "Ensure SCTP AUTH chunks are enabled in kernel");
return (NULL);
}
auth.sauth_chunk = OPENSSL_SCTP_FORWARD_CUM_TSN_CHUNK_TYPE;
ret =
setsockopt(fd, IPPROTO_SCTP, SCTP_AUTH_CHUNK, &auth,
sizeof(struct sctp_authchunk));
if (ret < 0) {
BIO_vfree(bio);
BIOerr(BIO_F_BIO_NEW_DGRAM_SCTP, ERR_R_SYS_LIB);
ERR_add_error_data(1, "Ensure SCTP AUTH chunks are enabled in kernel");
return (NULL);
}
/*
* Test if activation was successful. When using accept(), SCTP-AUTH has
* to be activated for the listening socket already, otherwise the
* connected socket won't use it. Similarly with connect(): the socket
* prior to connection must be activated for SCTP-AUTH
*/
sockopt_len = (socklen_t) (sizeof(sctp_assoc_t) + 256 * sizeof(uint8_t));
authchunks = OPENSSL_zalloc(sockopt_len);
if (authchunks == NULL) {
BIO_vfree(bio);
return (NULL);
}
ret = getsockopt(fd, IPPROTO_SCTP, SCTP_LOCAL_AUTH_CHUNKS, authchunks,
&sockopt_len);
if (ret < 0) {
OPENSSL_free(authchunks);
BIO_vfree(bio);
return (NULL);
}
for (p = (unsigned char *)authchunks->gauth_chunks;
p < (unsigned char *)authchunks + sockopt_len;
p += sizeof(uint8_t)) {
if (*p == OPENSSL_SCTP_DATA_CHUNK_TYPE)
auth_data = 1;
if (*p == OPENSSL_SCTP_FORWARD_CUM_TSN_CHUNK_TYPE)
auth_forward = 1;
}
OPENSSL_free(authchunks);
if (!auth_data || !auth_forward) {
BIO_vfree(bio);
BIOerr(BIO_F_BIO_NEW_DGRAM_SCTP, ERR_R_SYS_LIB);
ERR_add_error_data(1,
"Ensure SCTP AUTH chunks are enabled on the "
"underlying socket");
return NULL;
}
# ifdef SCTP_AUTHENTICATION_EVENT
# ifdef SCTP_EVENT
memset(&event, 0, sizeof(event));
event.se_assoc_id = 0;
event.se_type = SCTP_AUTHENTICATION_EVENT;
event.se_on = 1;
ret =
setsockopt(fd, IPPROTO_SCTP, SCTP_EVENT, &event,
sizeof(struct sctp_event));
if (ret < 0) {
BIO_vfree(bio);
return (NULL);
}
# else
sockopt_len = (socklen_t) sizeof(struct sctp_event_subscribe);
ret = getsockopt(fd, IPPROTO_SCTP, SCTP_EVENTS, &event, &sockopt_len);
if (ret < 0) {
BIO_vfree(bio);
return (NULL);
}
event.sctp_authentication_event = 1;
ret =
setsockopt(fd, IPPROTO_SCTP, SCTP_EVENTS, &event,
sizeof(struct sctp_event_subscribe));
if (ret < 0) {
BIO_vfree(bio);
return (NULL);
}
# endif
# endif
/*
* Disable partial delivery by setting the min size larger than the max
* record size of 2^14 + 2048 + 13
*/
ret =
setsockopt(fd, IPPROTO_SCTP, SCTP_PARTIAL_DELIVERY_POINT, &optval,
sizeof(optval));
if (ret < 0) {
BIO_vfree(bio);
return (NULL);
}
return (bio);
}
int BIO_dgram_is_sctp(BIO *bio)
{
return (BIO_method_type(bio) == BIO_TYPE_DGRAM_SCTP);
}
static int dgram_sctp_new(BIO *bi)
{
bio_dgram_sctp_data *data = NULL;
bi->init = 0;
bi->num = 0;
data = OPENSSL_zalloc(sizeof(*data));
if (data == NULL)
return 0;
# ifdef SCTP_PR_SCTP_NONE
data->prinfo.pr_policy = SCTP_PR_SCTP_NONE;
# endif
bi->ptr = data;
bi->flags = 0;
return (1);
}
static int dgram_sctp_free(BIO *a)
{
bio_dgram_sctp_data *data;
if (a == NULL)
return (0);
if (!dgram_clear(a))
return 0;
data = (bio_dgram_sctp_data *) a->ptr;
if (data != NULL)
OPENSSL_free(data);
return (1);
}
# ifdef SCTP_AUTHENTICATION_EVENT
void dgram_sctp_handle_auth_free_key_event(BIO *b,
union sctp_notification *snp)
{
int ret;
struct sctp_authkey_event *authkeyevent = &snp->sn_auth_event;
if (authkeyevent->auth_indication == SCTP_AUTH_FREE_KEY) {
struct sctp_authkeyid authkeyid;
/* delete key */
authkeyid.scact_keynumber = authkeyevent->auth_keynumber;
ret = setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_DELETE_KEY,
&authkeyid, sizeof(struct sctp_authkeyid));
}
}
# endif
static int dgram_sctp_read(BIO *b, char *out, int outl)
{
int ret = 0, n = 0, i, optval;
socklen_t optlen;
bio_dgram_sctp_data *data = (bio_dgram_sctp_data *) b->ptr;
union sctp_notification *snp;
struct msghdr msg;
struct iovec iov;
struct cmsghdr *cmsg;
char cmsgbuf[512];
if (out != NULL) {
clear_socket_error();
do {
memset(&data->rcvinfo, 0, sizeof(data->rcvinfo));
iov.iov_base = out;
iov.iov_len = outl;
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = cmsgbuf;
msg.msg_controllen = 512;
msg.msg_flags = 0;
n = recvmsg(b->num, &msg, 0);
if (n <= 0) {
if (n < 0)
ret = n;
break;
}
if (msg.msg_controllen > 0) {
for (cmsg = CMSG_FIRSTHDR(&msg); cmsg;
cmsg = CMSG_NXTHDR(&msg, cmsg)) {
if (cmsg->cmsg_level != IPPROTO_SCTP)
continue;
# ifdef SCTP_RCVINFO
if (cmsg->cmsg_type == SCTP_RCVINFO) {
struct sctp_rcvinfo *rcvinfo;
rcvinfo = (struct sctp_rcvinfo *)CMSG_DATA(cmsg);
data->rcvinfo.rcv_sid = rcvinfo->rcv_sid;
data->rcvinfo.rcv_ssn = rcvinfo->rcv_ssn;
data->rcvinfo.rcv_flags = rcvinfo->rcv_flags;
data->rcvinfo.rcv_ppid = rcvinfo->rcv_ppid;
data->rcvinfo.rcv_tsn = rcvinfo->rcv_tsn;
data->rcvinfo.rcv_cumtsn = rcvinfo->rcv_cumtsn;
data->rcvinfo.rcv_context = rcvinfo->rcv_context;
}
# endif
# ifdef SCTP_SNDRCV
if (cmsg->cmsg_type == SCTP_SNDRCV) {
struct sctp_sndrcvinfo *sndrcvinfo;
sndrcvinfo =
(struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
data->rcvinfo.rcv_sid = sndrcvinfo->sinfo_stream;
data->rcvinfo.rcv_ssn = sndrcvinfo->sinfo_ssn;
data->rcvinfo.rcv_flags = sndrcvinfo->sinfo_flags;
data->rcvinfo.rcv_ppid = sndrcvinfo->sinfo_ppid;
data->rcvinfo.rcv_tsn = sndrcvinfo->sinfo_tsn;
data->rcvinfo.rcv_cumtsn = sndrcvinfo->sinfo_cumtsn;
data->rcvinfo.rcv_context = sndrcvinfo->sinfo_context;
}
# endif
}
}
if (msg.msg_flags & MSG_NOTIFICATION) {
snp = (union sctp_notification *)out;
if (snp->sn_header.sn_type == SCTP_SENDER_DRY_EVENT) {
# ifdef SCTP_EVENT
struct sctp_event event;
# else
struct sctp_event_subscribe event;
socklen_t eventsize;
# endif
/* disable sender dry event */
# ifdef SCTP_EVENT
memset(&event, 0, sizeof(event));
event.se_assoc_id = 0;
event.se_type = SCTP_SENDER_DRY_EVENT;
event.se_on = 0;
i = setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENT, &event,
sizeof(struct sctp_event));
if (i < 0) {
ret = i;
break;
}
# else
eventsize = sizeof(struct sctp_event_subscribe);
i = getsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event,
&eventsize);
if (i < 0) {
ret = i;
break;
}
event.sctp_sender_dry_event = 0;
i = setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event,
sizeof(struct sctp_event_subscribe));
if (i < 0) {
ret = i;
break;
}
# endif
}
# ifdef SCTP_AUTHENTICATION_EVENT
if (snp->sn_header.sn_type == SCTP_AUTHENTICATION_EVENT)
dgram_sctp_handle_auth_free_key_event(b, snp);
# endif
if (data->handle_notifications != NULL)
data->handle_notifications(b, data->notification_context,
(void *)out);
memset(out, 0, outl);
} else
ret += n;
}
while ((msg.msg_flags & MSG_NOTIFICATION) && (msg.msg_flags & MSG_EOR)
&& (ret < outl));
if (ret > 0 && !(msg.msg_flags & MSG_EOR)) {
/* Partial message read, this should never happen! */
/*
* The buffer was too small, this means the peer sent a message
* that was larger than allowed.
*/
if (ret == outl)
return -1;
/*
* Test if socket buffer can handle max record size (2^14 + 2048
* + 13)
*/
optlen = (socklen_t) sizeof(int);
ret = getsockopt(b->num, SOL_SOCKET, SO_RCVBUF, &optval, &optlen);
if (ret >= 0)
OPENSSL_assert(optval >= 18445);
/*
* Test if SCTP doesn't partially deliver below max record size
* (2^14 + 2048 + 13)
*/
optlen = (socklen_t) sizeof(int);
ret =
getsockopt(b->num, IPPROTO_SCTP, SCTP_PARTIAL_DELIVERY_POINT,
&optval, &optlen);
if (ret >= 0)
OPENSSL_assert(optval >= 18445);
/*
* Partially delivered notification??? Probably a bug....
*/
OPENSSL_assert(!(msg.msg_flags & MSG_NOTIFICATION));
/*
* Everything seems ok till now, so it's most likely a message
* dropped by PR-SCTP.
*/
memset(out, 0, outl);
BIO_set_retry_read(b);
return -1;
}
BIO_clear_retry_flags(b);
if (ret < 0) {
if (BIO_dgram_should_retry(ret)) {
BIO_set_retry_read(b);
data->_errno = get_last_socket_error();
}
}
/* Test if peer uses SCTP-AUTH before continuing */
if (!data->peer_auth_tested) {
int ii, auth_data = 0, auth_forward = 0;
unsigned char *p;
struct sctp_authchunks *authchunks;
optlen =
(socklen_t) (sizeof(sctp_assoc_t) + 256 * sizeof(uint8_t));
authchunks = OPENSSL_malloc(optlen);
if (authchunks == NULL) {
BIOerr(BIO_F_DGRAM_SCTP_READ, ERR_R_MALLOC_FAILURE);
return -1;
}
memset(authchunks, 0, optlen);
ii = getsockopt(b->num, IPPROTO_SCTP, SCTP_PEER_AUTH_CHUNKS,
authchunks, &optlen);
if (ii >= 0)
for (p = (unsigned char *)authchunks->gauth_chunks;
p < (unsigned char *)authchunks + optlen;
p += sizeof(uint8_t)) {
if (*p == OPENSSL_SCTP_DATA_CHUNK_TYPE)
auth_data = 1;
if (*p == OPENSSL_SCTP_FORWARD_CUM_TSN_CHUNK_TYPE)
auth_forward = 1;
}
OPENSSL_free(authchunks);
if (!auth_data || !auth_forward) {
BIOerr(BIO_F_DGRAM_SCTP_READ, BIO_R_CONNECT_ERROR);
return -1;
}
data->peer_auth_tested = 1;
}
}
return (ret);
}
/*
* dgram_sctp_write - send message on SCTP socket
* @b: BIO to write to
* @in: data to send
* @inl: amount of bytes in @in to send
*
* Returns -1 on error or the sent amount of bytes on success
*/
static int dgram_sctp_write(BIO *b, const char *in, int inl)
{
int ret;
bio_dgram_sctp_data *data = (bio_dgram_sctp_data *) b->ptr;
struct bio_dgram_sctp_sndinfo *sinfo = &(data->sndinfo);
struct bio_dgram_sctp_prinfo *pinfo = &(data->prinfo);
struct bio_dgram_sctp_sndinfo handshake_sinfo;
struct iovec iov[1];
struct msghdr msg;
struct cmsghdr *cmsg;
# if defined(SCTP_SNDINFO) && defined(SCTP_PRINFO)
char cmsgbuf[CMSG_SPACE(sizeof(struct sctp_sndinfo)) +
CMSG_SPACE(sizeof(struct sctp_prinfo))];
struct sctp_sndinfo *sndinfo;
struct sctp_prinfo *prinfo;
# else
char cmsgbuf[CMSG_SPACE(sizeof(struct sctp_sndrcvinfo))];
struct sctp_sndrcvinfo *sndrcvinfo;
# endif
clear_socket_error();
/*
* If we're send anything else than application data, disable all user
* parameters and flags.
*/
if (in[0] != 23) {
memset(&handshake_sinfo, 0, sizeof(handshake_sinfo));
# ifdef SCTP_SACK_IMMEDIATELY
handshake_sinfo.snd_flags = SCTP_SACK_IMMEDIATELY;
# endif
sinfo = &handshake_sinfo;
}
/* We can only send a shutdown alert if the socket is dry */
if (data->save_shutdown) {
ret = BIO_dgram_sctp_wait_for_dry(b);
if (ret < 0)
return -1;
if (ret == 0) {
BIO_clear_retry_flags(b);
BIO_set_retry_write(b);
return -1;
}
}
iov[0].iov_base = (char *)in;
iov[0].iov_len = inl;
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = iov;
msg.msg_iovlen = 1;
msg.msg_control = (caddr_t) cmsgbuf;
msg.msg_controllen = 0;
msg.msg_flags = 0;
# if defined(SCTP_SNDINFO) && defined(SCTP_PRINFO)
cmsg = (struct cmsghdr *)cmsgbuf;
cmsg->cmsg_level = IPPROTO_SCTP;
cmsg->cmsg_type = SCTP_SNDINFO;
cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndinfo));
sndinfo = (struct sctp_sndinfo *)CMSG_DATA(cmsg);
memset(sndinfo, 0, sizeof(*sndinfo));
sndinfo->snd_sid = sinfo->snd_sid;
sndinfo->snd_flags = sinfo->snd_flags;
sndinfo->snd_ppid = sinfo->snd_ppid;
sndinfo->snd_context = sinfo->snd_context;
msg.msg_controllen += CMSG_SPACE(sizeof(struct sctp_sndinfo));
cmsg =
(struct cmsghdr *)&cmsgbuf[CMSG_SPACE(sizeof(struct sctp_sndinfo))];
cmsg->cmsg_level = IPPROTO_SCTP;
cmsg->cmsg_type = SCTP_PRINFO;
cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_prinfo));
prinfo = (struct sctp_prinfo *)CMSG_DATA(cmsg);
memset(prinfo, 0, sizeof(*prinfo));
prinfo->pr_policy = pinfo->pr_policy;
prinfo->pr_value = pinfo->pr_value;
msg.msg_controllen += CMSG_SPACE(sizeof(struct sctp_prinfo));
# else
cmsg = (struct cmsghdr *)cmsgbuf;
cmsg->cmsg_level = IPPROTO_SCTP;
cmsg->cmsg_type = SCTP_SNDRCV;
cmsg->cmsg_len = CMSG_LEN(sizeof(struct sctp_sndrcvinfo));
sndrcvinfo = (struct sctp_sndrcvinfo *)CMSG_DATA(cmsg);
memset(sndrcvinfo, 0, sizeof(*sndrcvinfo));
sndrcvinfo->sinfo_stream = sinfo->snd_sid;
sndrcvinfo->sinfo_flags = sinfo->snd_flags;
# ifdef __FreeBSD__
sndrcvinfo->sinfo_flags |= pinfo->pr_policy;
# endif
sndrcvinfo->sinfo_ppid = sinfo->snd_ppid;
sndrcvinfo->sinfo_context = sinfo->snd_context;
sndrcvinfo->sinfo_timetolive = pinfo->pr_value;
msg.msg_controllen += CMSG_SPACE(sizeof(struct sctp_sndrcvinfo));
# endif
ret = sendmsg(b->num, &msg, 0);
BIO_clear_retry_flags(b);
if (ret <= 0) {
if (BIO_dgram_should_retry(ret)) {
BIO_set_retry_write(b);
data->_errno = get_last_socket_error();
}
}
return (ret);
}
static long dgram_sctp_ctrl(BIO *b, int cmd, long num, void *ptr)
{
long ret = 1;
bio_dgram_sctp_data *data = NULL;
socklen_t sockopt_len = 0;
struct sctp_authkeyid authkeyid;
struct sctp_authkey *authkey = NULL;
data = (bio_dgram_sctp_data *) b->ptr;
switch (cmd) {
case BIO_CTRL_DGRAM_QUERY_MTU:
/*
* Set to maximum (2^14) and ignore user input to enable transport
* protocol fragmentation. Returns always 2^14.
*/
data->mtu = 16384;
ret = data->mtu;
break;
case BIO_CTRL_DGRAM_SET_MTU:
/*
* Set to maximum (2^14) and ignore input to enable transport
* protocol fragmentation. Returns always 2^14.
*/
data->mtu = 16384;
ret = data->mtu;
break;
case BIO_CTRL_DGRAM_SET_CONNECTED:
case BIO_CTRL_DGRAM_CONNECT:
/* Returns always -1. */
ret = -1;
break;
case BIO_CTRL_DGRAM_SET_NEXT_TIMEOUT:
/*
* SCTP doesn't need the DTLS timer Returns always 1.
*/
break;
case BIO_CTRL_DGRAM_GET_MTU_OVERHEAD:
/*
* We allow transport protocol fragmentation so this is irrelevant
*/
ret = 0;
break;
case BIO_CTRL_DGRAM_SCTP_SET_IN_HANDSHAKE:
if (num > 0)
data->in_handshake = 1;
else
data->in_handshake = 0;
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_NODELAY,
&data->in_handshake, sizeof(int));
break;
case BIO_CTRL_DGRAM_SCTP_ADD_AUTH_KEY:
/*
* New shared key for SCTP AUTH. Returns 0 on success, -1 otherwise.
*/
/* Get active key */
sockopt_len = sizeof(struct sctp_authkeyid);
ret =
getsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_ACTIVE_KEY, &authkeyid,
&sockopt_len);
if (ret < 0)
break;
/* Add new key */
sockopt_len = sizeof(struct sctp_authkey) + 64 * sizeof(uint8_t);
authkey = OPENSSL_malloc(sockopt_len);
if (authkey == NULL) {
ret = -1;
break;
}
memset(authkey, 0, sockopt_len);
authkey->sca_keynumber = authkeyid.scact_keynumber + 1;
# ifndef __FreeBSD__
/*
* This field is missing in FreeBSD 8.2 and earlier, and FreeBSD 8.3
* and higher work without it.
*/
authkey->sca_keylength = 64;
# endif
memcpy(&authkey->sca_key[0], ptr, 64 * sizeof(uint8_t));
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_KEY, authkey,
sockopt_len);
OPENSSL_free(authkey);
authkey = NULL;
if (ret < 0)
break;
/* Reset active key */
ret = setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_ACTIVE_KEY,
&authkeyid, sizeof(struct sctp_authkeyid));
if (ret < 0)
break;
break;
case BIO_CTRL_DGRAM_SCTP_NEXT_AUTH_KEY:
/* Returns 0 on success, -1 otherwise. */
/* Get active key */
sockopt_len = sizeof(struct sctp_authkeyid);
ret =
getsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_ACTIVE_KEY, &authkeyid,
&sockopt_len);
if (ret < 0)
break;
/* Set active key */
authkeyid.scact_keynumber = authkeyid.scact_keynumber + 1;
ret = setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_ACTIVE_KEY,
&authkeyid, sizeof(struct sctp_authkeyid));
if (ret < 0)
break;
/*
* CCS has been sent, so remember that and fall through to check if
* we need to deactivate an old key
*/
data->ccs_sent = 1;
case BIO_CTRL_DGRAM_SCTP_AUTH_CCS_RCVD:
/* Returns 0 on success, -1 otherwise. */
/*
* Has this command really been called or is this just a
* fall-through?
*/
if (cmd == BIO_CTRL_DGRAM_SCTP_AUTH_CCS_RCVD)
data->ccs_rcvd = 1;
/*
* CSS has been both, received and sent, so deactivate an old key
*/
if (data->ccs_rcvd == 1 && data->ccs_sent == 1) {
/* Get active key */
sockopt_len = sizeof(struct sctp_authkeyid);
ret =
getsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_ACTIVE_KEY,
&authkeyid, &sockopt_len);
if (ret < 0)
break;
/*
* Deactivate key or delete second last key if
* SCTP_AUTHENTICATION_EVENT is not available.
*/
authkeyid.scact_keynumber = authkeyid.scact_keynumber - 1;
# ifdef SCTP_AUTH_DEACTIVATE_KEY
sockopt_len = sizeof(struct sctp_authkeyid);
ret = setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_DEACTIVATE_KEY,
&authkeyid, sockopt_len);
if (ret < 0)
break;
# endif
# ifndef SCTP_AUTHENTICATION_EVENT
if (authkeyid.scact_keynumber > 0) {
authkeyid.scact_keynumber = authkeyid.scact_keynumber - 1;
ret = setsockopt(b->num, IPPROTO_SCTP, SCTP_AUTH_DELETE_KEY,
&authkeyid, sizeof(struct sctp_authkeyid));
if (ret < 0)
break;
}
# endif
data->ccs_rcvd = 0;
data->ccs_sent = 0;
}
break;
case BIO_CTRL_DGRAM_SCTP_GET_SNDINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_sndinfo))
num = sizeof(struct bio_dgram_sctp_sndinfo);
memcpy(ptr, &(data->sndinfo), num);
ret = num;
break;
case BIO_CTRL_DGRAM_SCTP_SET_SNDINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_sndinfo))
num = sizeof(struct bio_dgram_sctp_sndinfo);
memcpy(&(data->sndinfo), ptr, num);
break;
case BIO_CTRL_DGRAM_SCTP_GET_RCVINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_rcvinfo))
num = sizeof(struct bio_dgram_sctp_rcvinfo);
memcpy(ptr, &data->rcvinfo, num);
ret = num;
break;
case BIO_CTRL_DGRAM_SCTP_SET_RCVINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_rcvinfo))
num = sizeof(struct bio_dgram_sctp_rcvinfo);
memcpy(&(data->rcvinfo), ptr, num);
break;
case BIO_CTRL_DGRAM_SCTP_GET_PRINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_prinfo))
num = sizeof(struct bio_dgram_sctp_prinfo);
memcpy(ptr, &(data->prinfo), num);
ret = num;
break;
case BIO_CTRL_DGRAM_SCTP_SET_PRINFO:
/* Returns the size of the copied struct. */
if (num > (long)sizeof(struct bio_dgram_sctp_prinfo))
num = sizeof(struct bio_dgram_sctp_prinfo);
memcpy(&(data->prinfo), ptr, num);
break;
case BIO_CTRL_DGRAM_SCTP_SAVE_SHUTDOWN:
/* Returns always 1. */
if (num > 0)
data->save_shutdown = 1;
else
data->save_shutdown = 0;
break;
default:
/*
* Pass to default ctrl function to process SCTP unspecific commands
*/
ret = dgram_ctrl(b, cmd, num, ptr);
break;
}
return (ret);
}
int BIO_dgram_sctp_notification_cb(BIO *b,
void (*handle_notifications) (BIO *bio,
void
*context,
void *buf),
void *context)
{
bio_dgram_sctp_data *data = (bio_dgram_sctp_data *) b->ptr;
if (handle_notifications != NULL) {
data->handle_notifications = handle_notifications;
data->notification_context = context;
} else
return -1;
return 0;
}
/*
* BIO_dgram_sctp_wait_for_dry - Wait for SCTP SENDER_DRY event
* @b: The BIO to check for the dry event
*
* Wait until the peer confirms all packets have been received, and so that
* our kernel doesn't have anything to send anymore. This is only received by
* the peer's kernel, not the application.
*
* Returns:
* -1 on error
* 0 when not dry yet
* 1 when dry
*/
int BIO_dgram_sctp_wait_for_dry(BIO *b)
{
int is_dry = 0;
int sockflags = 0;
int n, ret;
union sctp_notification snp;
struct msghdr msg;
struct iovec iov;
# ifdef SCTP_EVENT
struct sctp_event event;
# else
struct sctp_event_subscribe event;
socklen_t eventsize;
# endif
bio_dgram_sctp_data *data = (bio_dgram_sctp_data *) b->ptr;
/* set sender dry event */
# ifdef SCTP_EVENT
memset(&event, 0, sizeof(event));
event.se_assoc_id = 0;
event.se_type = SCTP_SENDER_DRY_EVENT;
event.se_on = 1;
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENT, &event,
sizeof(struct sctp_event));
# else
eventsize = sizeof(struct sctp_event_subscribe);
ret = getsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event, &eventsize);
if (ret < 0)
return -1;
event.sctp_sender_dry_event = 1;
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event,
sizeof(struct sctp_event_subscribe));
# endif
if (ret < 0)
return -1;
/* peek for notification */
memset(&snp, 0, sizeof(snp));
iov.iov_base = (char *)&snp;
iov.iov_len = sizeof(union sctp_notification);
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = 0;
n = recvmsg(b->num, &msg, MSG_PEEK);
if (n <= 0) {
if ((n < 0) && (get_last_socket_error() != EAGAIN)
&& (get_last_socket_error() != EWOULDBLOCK))
return -1;
else
return 0;
}
/* if we find a notification, process it and try again if necessary */
while (msg.msg_flags & MSG_NOTIFICATION) {
memset(&snp, 0, sizeof(snp));
iov.iov_base = (char *)&snp;
iov.iov_len = sizeof(union sctp_notification);
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = 0;
n = recvmsg(b->num, &msg, 0);
if (n <= 0) {
if ((n < 0) && (get_last_socket_error() != EAGAIN)
&& (get_last_socket_error() != EWOULDBLOCK))
return -1;
else
return is_dry;
}
if (snp.sn_header.sn_type == SCTP_SENDER_DRY_EVENT) {
is_dry = 1;
/* disable sender dry event */
# ifdef SCTP_EVENT
memset(&event, 0, sizeof(event));
event.se_assoc_id = 0;
event.se_type = SCTP_SENDER_DRY_EVENT;
event.se_on = 0;
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENT, &event,
sizeof(struct sctp_event));
# else
eventsize = (socklen_t) sizeof(struct sctp_event_subscribe);
ret =
getsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event,
&eventsize);
if (ret < 0)
return -1;
event.sctp_sender_dry_event = 0;
ret =
setsockopt(b->num, IPPROTO_SCTP, SCTP_EVENTS, &event,
sizeof(struct sctp_event_subscribe));
# endif
if (ret < 0)
return -1;
}
# ifdef SCTP_AUTHENTICATION_EVENT
if (snp.sn_header.sn_type == SCTP_AUTHENTICATION_EVENT)
dgram_sctp_handle_auth_free_key_event(b, &snp);
# endif
if (data->handle_notifications != NULL)
data->handle_notifications(b, data->notification_context,
(void *)&snp);
/* found notification, peek again */
memset(&snp, 0, sizeof(snp));
iov.iov_base = (char *)&snp;
iov.iov_len = sizeof(union sctp_notification);
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = 0;
/* if we have seen the dry already, don't wait */
if (is_dry) {
sockflags = fcntl(b->num, F_GETFL, 0);
fcntl(b->num, F_SETFL, O_NONBLOCK);
}
n = recvmsg(b->num, &msg, MSG_PEEK);
if (is_dry) {
fcntl(b->num, F_SETFL, sockflags);
}
if (n <= 0) {
if ((n < 0) && (get_last_socket_error() != EAGAIN)
&& (get_last_socket_error() != EWOULDBLOCK))
return -1;
else
return is_dry;
}
}
/* read anything else */
return is_dry;
}
int BIO_dgram_sctp_msg_waiting(BIO *b)
{
int n, sockflags;
union sctp_notification snp;
struct msghdr msg;
struct iovec iov;
bio_dgram_sctp_data *data = (bio_dgram_sctp_data *) b->ptr;
/* Check if there are any messages waiting to be read */
do {
memset(&snp, 0, sizeof(snp));
iov.iov_base = (char *)&snp;
iov.iov_len = sizeof(union sctp_notification);
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = 0;
sockflags = fcntl(b->num, F_GETFL, 0);
fcntl(b->num, F_SETFL, O_NONBLOCK);
n = recvmsg(b->num, &msg, MSG_PEEK);
fcntl(b->num, F_SETFL, sockflags);
/* if notification, process and try again */
if (n > 0 && (msg.msg_flags & MSG_NOTIFICATION)) {
# ifdef SCTP_AUTHENTICATION_EVENT
if (snp.sn_header.sn_type == SCTP_AUTHENTICATION_EVENT)
dgram_sctp_handle_auth_free_key_event(b, &snp);
# endif
memset(&snp, 0, sizeof(snp));
iov.iov_base = (char *)&snp;
iov.iov_len = sizeof(union sctp_notification);
msg.msg_name = NULL;
msg.msg_namelen = 0;
msg.msg_iov = &iov;
msg.msg_iovlen = 1;
msg.msg_control = NULL;
msg.msg_controllen = 0;
msg.msg_flags = 0;
n = recvmsg(b->num, &msg, 0);
if (data->handle_notifications != NULL)
data->handle_notifications(b, data->notification_context,
(void *)&snp);
}
} while (n > 0 && (msg.msg_flags & MSG_NOTIFICATION));
/* Return 1 if there is a message to be read, return 0 otherwise. */
if (n > 0)
return 1;
else
return 0;
}
static int dgram_sctp_puts(BIO *bp, const char *str)
{
int n, ret;
n = strlen(str);
ret = dgram_sctp_write(bp, str, n);
return (ret);
}
# endif
static int BIO_dgram_should_retry(int i)
{
int err;
if ((i == 0) || (i == -1)) {
err = get_last_socket_error();
# if defined(OPENSSL_SYS_WINDOWS)
/*
* If the socket return value (i) is -1 and err is unexpectedly 0 at
* this point, the error code was overwritten by another system call
* before this error handling is called.
*/
# endif
return (BIO_dgram_non_fatal_error(err));
}
return (0);
}
int BIO_dgram_non_fatal_error(int err)
{
switch (err) {
# if defined(OPENSSL_SYS_WINDOWS)
# if defined(WSAEWOULDBLOCK)
case WSAEWOULDBLOCK:
# endif
# endif
# ifdef EWOULDBLOCK
# ifdef WSAEWOULDBLOCK
# if WSAEWOULDBLOCK != EWOULDBLOCK
case EWOULDBLOCK:
# endif
# else
case EWOULDBLOCK:
# endif
# endif
# ifdef EINTR
case EINTR:
# endif
# ifdef EAGAIN
# if EWOULDBLOCK != EAGAIN
case EAGAIN:
# endif
# endif
# ifdef EPROTO
case EPROTO:
# endif
# ifdef EINPROGRESS
case EINPROGRESS:
# endif
# ifdef EALREADY
case EALREADY:
# endif
return (1);
default:
break;
}
return (0);
}
static void get_current_time(struct timeval *t)
{
# if defined(_WIN32)
SYSTEMTIME st;
union {
unsigned __int64 ul;
FILETIME ft;
} now;
GetSystemTime(&st);
SystemTimeToFileTime(&st, &now.ft);
# ifdef __MINGW32__
now.ul -= 116444736000000000ULL;
# else
now.ul -= 116444736000000000UI64; /* re-bias to 1/1/1970 */
# endif
t->tv_sec = (long)(now.ul / 10000000);
t->tv_usec = ((int)(now.ul % 10000000)) / 10;
# else
gettimeofday(t, NULL);
# endif
}
#endif