/*
* Common subroutine to open a TCP connection to remote host specified
* by struct sockaddr_in in mbuf *nam. Call in_pcbbind to assign a local
* port number if needed. Call in_pcbconnect_setup to do the routing and
* to choose a local host address (interface). If there is an existing
* incarnation of the same connection in TIME-WAIT state and if the remote
* host was sending CC options and if the connection duration was < MSL, then
* truncate the previous TIME-WAIT state and proceed.
* Initialize connection parameters and enter SYN-SENT state.
*/
static int
tcp_connect(struct tcpcb *tp, struct sockaddr *nam, struct thread *td)
{
struct inpcb *inp = tp->t_inpcb, *oinp;
struct socket *so = inp->inp_socket;
struct in_addr laddr;
u_short lport;
int error;
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
INP_WLOCK_ASSERT(inp);
if (inp->inp_lport == 0) {
error = in_pcbbind(inp, (struct sockaddr *)0, td->td_ucred);
if (error)
return error;
}
/*
* Cannot simply call in_pcbconnect, because there might be an
* earlier incarnation of this same connection still in
* TIME_WAIT state, creating an ADDRINUSE error.
*/
laddr = inp->inp_laddr;
lport = inp->inp_lport;
error = in_pcbconnect_setup(inp, nam, &laddr.s_addr, &lport,
&inp->inp_faddr.s_addr, &inp->inp_fport, &oinp, td->td_ucred);
if (error && oinp == NULL)
return error;
if (oinp)
return EADDRINUSE;
inp->inp_laddr = laddr;
in_pcbrehash(inp);
/*
* Compute window scaling to request:
* Scale to fit into sweet spot. See tcp_syncache.c.
* XXX: This should move to tcp_output().
*/
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < sb_max)
tp->request_r_scale++;
soisconnecting(so);
TCPSTAT_INC(tcps_connattempt);
tp->t_state = TCPS_SYN_SENT;
tcp_timer_activate(tp, TT_KEEP, tcp_keepinit);
tp->iss = tcp_new_isn(tp);
tp->t_bw_rtseq = tp->iss;
tcp_sendseqinit(tp);
return 0;
}
static int
udp_output(struct inpcb *inp, struct mbuf *m, struct sockaddr *addr,
struct mbuf *control, struct thread *td)
{
struct udpiphdr *ui;
int len = m->m_pkthdr.len;
struct in_addr faddr, laddr;
struct cmsghdr *cm;
struct inpcbinfo *pcbinfo;
struct sockaddr_in *sin, src;
int cscov_partial = 0;
int error = 0;
int ipflags;
u_short fport, lport;
int unlock_udbinfo;
u_char tos;
uint8_t pr;
uint16_t cscov = 0;
/*
* udp_output() may need to temporarily bind or connect the current
* inpcb. As such, we don't know up front whether we will need the
* pcbinfo lock or not. Do any work to decide what is needed up
* front before acquiring any locks.
*/
if (len + sizeof(struct udpiphdr) > IP_MAXPACKET) {
if (control)
m_freem(control);
m_freem(m);
return (EMSGSIZE);
}
src.sin_family = 0;
INP_RLOCK(inp);
tos = inp->inp_ip_tos;
if (control != NULL) {
/*
* XXX: Currently, we assume all the optional information is
* stored in a single mbuf.
*/
if (control->m_next) {
INP_RUNLOCK(inp);
m_freem(control);
m_freem(m);
return (EINVAL);
}
for (; control->m_len > 0;
control->m_data += CMSG_ALIGN(cm->cmsg_len),
control->m_len -= CMSG_ALIGN(cm->cmsg_len)) {
cm = mtod(control, struct cmsghdr *);
if (control->m_len < sizeof(*cm) || cm->cmsg_len == 0
|| cm->cmsg_len > control->m_len) {
error = EINVAL;
break;
}
if (cm->cmsg_level != IPPROTO_IP)
continue;
switch (cm->cmsg_type) {
case IP_SENDSRCADDR:
if (cm->cmsg_len !=
CMSG_LEN(sizeof(struct in_addr))) {
error = EINVAL;
break;
}
bzero(&src, sizeof(src));
src.sin_family = AF_INET;
src.sin_len = sizeof(src);
src.sin_port = inp->inp_lport;
src.sin_addr =
*(struct in_addr *)CMSG_DATA(cm);
break;
case IP_TOS:
if (cm->cmsg_len != CMSG_LEN(sizeof(u_char))) {
error = EINVAL;
break;
}
tos = *(u_char *)CMSG_DATA(cm);
break;
default:
error = ENOPROTOOPT;
break;
}
if (error)
break;
}
m_freem(control);
}
if (error) {
INP_RUNLOCK(inp);
m_freem(m);
return (error);
}
/*
* Depending on whether or not the application has bound or connected
* the socket, we may have to do varying levels of work. The optimal
* case is for a connected UDP socket, as a global lock isn't
* required at all.
*
* In order to decide which we need, we require stability of the
* inpcb binding, which we ensure by acquiring a read lock on the
* inpcb. This doesn't strictly follow the lock order, so we play
* the trylock and retry game; note that we may end up with more
* conservative locks than required the second time around, so later
* assertions have to accept that. Further analysis of the number of
* misses under contention is required.
*
* XXXRW: Check that hash locking update here is correct.
*/
pr = inp->inp_socket->so_proto->pr_protocol;
pcbinfo = get_inpcbinfo(pr);
sin = (struct sockaddr_in *)addr;
if (sin != NULL &&
(inp->inp_laddr.s_addr == INADDR_ANY && inp->inp_lport == 0)) {
INP_RUNLOCK(inp);
INP_WLOCK(inp);
INP_HASH_WLOCK(pcbinfo);
unlock_udbinfo = UH_WLOCKED;
} else if ((sin != NULL && (
(sin->sin_addr.s_addr == INADDR_ANY) ||
(sin->sin_addr.s_addr == INADDR_BROADCAST) ||
(inp->inp_laddr.s_addr == INADDR_ANY) ||
(inp->inp_lport == 0))) ||
(src.sin_family == AF_INET)) {
INP_HASH_RLOCK(pcbinfo);
unlock_udbinfo = UH_RLOCKED;
} else
unlock_udbinfo = UH_UNLOCKED;
/*
* If the IP_SENDSRCADDR control message was specified, override the
* source address for this datagram. Its use is invalidated if the
* address thus specified is incomplete or clobbers other inpcbs.
*/
laddr = inp->inp_laddr;
lport = inp->inp_lport;
if (src.sin_family == AF_INET) {
INP_HASH_LOCK_ASSERT(pcbinfo);
if ((lport == 0) ||
(laddr.s_addr == INADDR_ANY &&
src.sin_addr.s_addr == INADDR_ANY)) {
error = EINVAL;
goto release;
}
error = in_pcbbind_setup(inp, (struct sockaddr *)&src,
&laddr.s_addr, &lport, td->td_ucred);
if (error)
goto release;
}
/*
* If a UDP socket has been connected, then a local address/port will
* have been selected and bound.
*
* If a UDP socket has not been connected to, then an explicit
* destination address must be used, in which case a local
* address/port may not have been selected and bound.
*/
if (sin != NULL) {
INP_LOCK_ASSERT(inp);
if (inp->inp_faddr.s_addr != INADDR_ANY) {
error = EISCONN;
goto release;
}
/*
* Jail may rewrite the destination address, so let it do
* that before we use it.
*/
error = prison_remote_ip4(td->td_ucred, &sin->sin_addr);
if (error)
goto release;
/*
* If a local address or port hasn't yet been selected, or if
* the destination address needs to be rewritten due to using
* a special INADDR_ constant, invoke in_pcbconnect_setup()
* to do the heavy lifting. Once a port is selected, we
* commit the binding back to the socket; we also commit the
* binding of the address if in jail.
*
* If we already have a valid binding and we're not
* requesting a destination address rewrite, use a fast path.
*/
if (inp->inp_laddr.s_addr == INADDR_ANY ||
inp->inp_lport == 0 ||
sin->sin_addr.s_addr == INADDR_ANY ||
sin->sin_addr.s_addr == INADDR_BROADCAST) {
INP_HASH_LOCK_ASSERT(pcbinfo);
error = in_pcbconnect_setup(inp, addr, &laddr.s_addr,
&lport, &faddr.s_addr, &fport, NULL,
td->td_ucred);
if (error)
goto release;
/*
* XXXRW: Why not commit the port if the address is
* !INADDR_ANY?
*/
/* Commit the local port if newly assigned. */
if (inp->inp_laddr.s_addr == INADDR_ANY &&
inp->inp_lport == 0) {
INP_WLOCK_ASSERT(inp);
INP_HASH_WLOCK_ASSERT(pcbinfo);
/*
* Remember addr if jailed, to prevent
* rebinding.
*/
if (prison_flag(td->td_ucred, PR_IP4))
inp->inp_laddr = laddr;
inp->inp_lport = lport;
if (in_pcbinshash(inp) != 0) {
inp->inp_lport = 0;
error = EAGAIN;
goto release;
}
inp->inp_flags |= INP_ANONPORT;
}
} else {
faddr = sin->sin_addr;
fport = sin->sin_port;
}
} else {
INP_LOCK_ASSERT(inp);
faddr = inp->inp_faddr;
fport = inp->inp_fport;
if (faddr.s_addr == INADDR_ANY) {
error = ENOTCONN;
goto release;
}
}
/*
* Calculate data length and get a mbuf for UDP, IP, and possible
* link-layer headers. Immediate slide the data pointer back forward
* since we won't use that space at this layer.
*/
M_PREPEND(m, sizeof(struct udpiphdr) + max_linkhdr, M_NOWAIT);
if (m == NULL) {
error = ENOBUFS;
goto release;
}
m->m_data += max_linkhdr;
m->m_len -= max_linkhdr;
m->m_pkthdr.len -= max_linkhdr;
/*
* Fill in mbuf with extended UDP header and addresses and length put
* into network format.
*/
ui = mtod(m, struct udpiphdr *);
bzero(ui->ui_x1, sizeof(ui->ui_x1)); /* XXX still needed? */
ui->ui_pr = pr;
ui->ui_src = laddr;
ui->ui_dst = faddr;
ui->ui_sport = lport;
ui->ui_dport = fport;
ui->ui_ulen = htons((u_short)len + sizeof(struct udphdr));
if (pr == IPPROTO_UDPLITE) {
struct udpcb *up;
uint16_t plen;
up = intoudpcb(inp);
cscov = up->u_txcslen;
plen = (u_short)len + sizeof(struct udphdr);
if (cscov >= plen)
cscov = 0;
ui->ui_len = htons(plen);
ui->ui_ulen = htons(cscov);
/*
* For UDP-Lite, checksum coverage length of zero means
* the entire UDPLite packet is covered by the checksum.
*/
cscov_partial = (cscov == 0) ? 0 : 1;
} else
ui->ui_v = IPVERSION << 4;
/*
* Set the Don't Fragment bit in the IP header.
*/
if (inp->inp_flags & INP_DONTFRAG) {
struct ip *ip;
ip = (struct ip *)&ui->ui_i;
ip->ip_off |= htons(IP_DF);
}
ipflags = 0;
if (inp->inp_socket->so_options & SO_DONTROUTE)
ipflags |= IP_ROUTETOIF;
if (inp->inp_socket->so_options & SO_BROADCAST)
ipflags |= IP_ALLOWBROADCAST;
if (inp->inp_flags & INP_ONESBCAST)
ipflags |= IP_SENDONES;
#ifdef MAC
mac_inpcb_create_mbuf(inp, m);
#endif
/*
* Set up checksum and output datagram.
*/
ui->ui_sum = 0;
if (cscov_partial) {
if (inp->inp_flags & INP_ONESBCAST)
faddr.s_addr = INADDR_BROADCAST;
if ((ui->ui_sum = in_cksum(m, sizeof(struct ip) + cscov)) == 0)
ui->ui_sum = 0xffff;
} else if (V_udp_cksum || !cscov_partial) {
if (inp->inp_flags & INP_ONESBCAST)
faddr.s_addr = INADDR_BROADCAST;
ui->ui_sum = in_pseudo(ui->ui_src.s_addr, faddr.s_addr,
htons((u_short)len + sizeof(struct udphdr) + pr));
m->m_pkthdr.csum_flags = CSUM_UDP;
m->m_pkthdr.csum_data = offsetof(struct udphdr, uh_sum);
}
((struct ip *)ui)->ip_len = htons(sizeof(struct udpiphdr) + len);
((struct ip *)ui)->ip_ttl = inp->inp_ip_ttl; /* XXX */
((struct ip *)ui)->ip_tos = tos; /* XXX */
UDPSTAT_INC(udps_opackets);
if (unlock_udbinfo == UH_WLOCKED)
INP_HASH_WUNLOCK(pcbinfo);
else if (unlock_udbinfo == UH_RLOCKED)
INP_HASH_RUNLOCK(pcbinfo);
UDP_PROBE(send, NULL, inp, &ui->ui_i, inp, &ui->ui_u);
error = ip_output(m, inp->inp_options, NULL, ipflags,
inp->inp_moptions, inp);
if (unlock_udbinfo == UH_WLOCKED)
INP_WUNLOCK(inp);
else
INP_RUNLOCK(inp);
return (error);
release:
if (unlock_udbinfo == UH_WLOCKED) {
INP_HASH_WUNLOCK(pcbinfo);
INP_WUNLOCK(inp);
} else if (unlock_udbinfo == UH_RLOCKED) {
INP_HASH_RUNLOCK(pcbinfo);
INP_RUNLOCK(inp);
} else
INP_RUNLOCK(inp);
m_freem(m);
return (error);
}
