/*
* Mark the connection as being incapable of further output.
*/
static int
tcp_usr_shutdown(struct socket *so)
{
int error = 0;
struct inpcb *inp;
struct tcpcb *tp = NULL;
TCPDEBUG0;
INP_INFO_WLOCK(&V_tcbinfo);
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
error = ECONNRESET;
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
socantsendmore(so);
tcp_usrclosed(tp);
if (!(inp->inp_flags & INP_DROPPED))
error = tcp_output_disconnect(tp);
out:
TCPDEBUG2(PRU_SHUTDOWN);
INP_WUNLOCK(inp);
INP_INFO_WUNLOCK(&V_tcbinfo);
return (error);
}
/*
* Initiate (or continue) disconnect.
* If embryonic state, just send reset (once).
* If in ``let data drain'' option and linger null, just drop.
* Otherwise (hard), mark socket disconnecting and drop
* current input data; switch states based on user close, and
* send segment to peer (with FIN).
*/
static void
tcp_disconnect(struct tcpcb *tp)
{
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
INP_WLOCK_ASSERT(inp);
/*
* Neither tcp_close() nor tcp_drop() should return NULL, as the
* socket is still open.
*/
if (tp->t_state < TCPS_ESTABLISHED) {
tp = tcp_close(tp);
KASSERT(tp != NULL,
("tcp_disconnect: tcp_close() returned NULL"));
} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
tp = tcp_drop(tp, 0);
KASSERT(tp != NULL,
("tcp_disconnect: tcp_drop() returned NULL"));
} else {
soisdisconnecting(so);
sbflush(&so->so_rcv);
tcp_usrclosed(tp);
if (!(inp->inp_flags & INP_DROPPED))
tcp_output_disconnect(tp);
}
}
/*
* Mark the connection as being incapable of further output.
*/
static void
tcp_usr_shutdown(netmsg_t msg)
{
struct socket *so = msg->shutdown.base.nm_so;
int error = 0;
struct inpcb *inp;
struct tcpcb *tp;
COMMON_START(so, inp, 0);
socantsendmore(so);
tp = tcp_usrclosed(tp);
if (tp)
error = tcp_output(tp);
COMMON_END(PRU_SHUTDOWN);
}
/*
* Mark the connection as being incapable of further output.
*/
static int
tcp_usr_shutdown(struct socket *so)
{
int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
COMMON_START();
socantsendmore(so);
tp = tcp_usrclosed(tp);
if (tp)
error = tcp_output(tp);
COMMON_END(PRU_SHUTDOWN);
}
struct tcpcb *
tcp_disconnect(struct tcpcb * tp)
{
struct socket * so = tp->t_inpcb->inp_socket;
if (tp->t_state < TCPS_ESTABLISHED)
tp = tcp_close(tp);
else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
tp = tcp_drop(tp, 0);
else
{
soisdisconnecting(so);
sbflush(&so->so_rcv);
tp = tcp_usrclosed(tp);
if (tp)
(void) tcp_output(tp);
}
return (tp);
}
/*
* Initiate (or continue) disconnect.
* If embryonic state, just send reset (once).
* If in ``let data drain'' option and linger null, just drop.
* Otherwise (hard), mark socket disconnecting and drop
* current input data; switch states based on user close, and
* send segment to peer (with FIN).
*/
static struct tcpcb *
tcp_disconnect(struct tcpcb *tp)
{
struct socket *so = tp->t_inpcb->inp_socket;
if (tp->t_state < TCPS_ESTABLISHED) {
tp = tcp_close(tp);
} else if ((so->so_options & SO_LINGER) && so->so_linger == 0) {
tp = tcp_drop(tp, 0);
} else {
lwkt_gettoken(&so->so_rcv.ssb_token);
soisdisconnecting(so);
sbflush(&so->so_rcv.sb);
tp = tcp_usrclosed(tp);
if (tp)
tcp_output(tp);
lwkt_reltoken(&so->so_rcv.ssb_token);
}
return (tp);
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static int
tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
struct sockaddr *nam, struct mbuf *control, struct thread *td)
{
int error = 0;
struct inpcb *inp;
struct tcpcb *tp = NULL;
int headlocked = 0;
#ifdef INET6
int isipv6;
#endif
TCPDEBUG0;
/*
* We require the pcbinfo lock in two cases:
*
* (1) An implied connect is taking place, which can result in
* binding IPs and ports and hence modification of the pcb hash
* chains.
*
* (2) PRUS_EOF is set, resulting in explicit close on the send.
*/
if ((nam != NULL) || (flags & PRUS_EOF)) {
INP_INFO_WLOCK(&V_tcbinfo);
headlocked = 1;
}
inp = sotoinpcb(so);
KASSERT(inp != NULL, ("tcp_usr_send: inp == NULL"));
INP_WLOCK(inp);
if (inp->inp_flags & (INP_TIMEWAIT | INP_DROPPED)) {
if (control)
m_freem(control);
if (m)
m_freem(m);
error = ECONNRESET;
goto out;
}
#ifdef INET6
isipv6 = nam && nam->sa_family == AF_INET6;
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control);
if (m)
m_freem(m);
error = EINVAL;
goto out;
}
m_freem(control); /* empty control, just free it */
}
if (!(flags & PRUS_OOB)) {
sbappendstream(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, td);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, td);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
socantsendmore(so);
tcp_usrclosed(tp);
}
if (headlocked) {
INP_INFO_WUNLOCK(&V_tcbinfo);
headlocked = 0;
}
if (!(inp->inp_flags & INP_DROPPED)) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output_send(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
} else {
/*
* XXXRW: PRUS_EOF not implemented with PRUS_OOB?
*/
SOCKBUF_LOCK(&so->so_snd);
if (sbspace(&so->so_snd) < -512) {
SOCKBUF_UNLOCK(&so->so_snd);
m_freem(m);
error = ENOBUFS;
goto out;
}
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
sbappendstream_locked(&so->so_snd, m);
SOCKBUF_UNLOCK(&so->so_snd);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
INP_INFO_WLOCK_ASSERT(&V_tcbinfo);
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, td);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, td);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
INP_INFO_WUNLOCK(&V_tcbinfo);
headlocked = 0;
} else if (nam) {
INP_INFO_WUNLOCK(&V_tcbinfo);
headlocked = 0;
}
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_flags |= TF_FORCEDATA;
error = tcp_output_send(tp);
tp->t_flags &= ~TF_FORCEDATA;
}
out:
TCPDEBUG2((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
INP_WUNLOCK(inp);
if (headlocked)
INP_INFO_WUNLOCK(&V_tcbinfo);
return (error);
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static int
tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
struct sockaddr *nam, struct mbuf *control, struct proc *p)
{
int s = splnet();
int error = 0;
struct inpcb *inp = sotoinpcb(so);
struct tcpcb *tp;
#ifdef INET6
int isipv6;
#endif
TCPDEBUG0;
if (inp == NULL) {
/*
* OOPS! we lost a race, the TCP session got reset after
* we checked SS_CANTSENDMORE, eg: while doing uiomove or a
* network interrupt in the non-splnet() section of sosend().
*/
if (m)
m_freem(m);
if (control)
m_freem(control);
error = ECONNRESET; /* XXX EPIPE? */
tp = NULL;
TCPDEBUG1();
goto out;
}
#ifdef INET6
isipv6 = nam && nam->sa_family == AF_INET6;
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control);
if (m)
m_freem(m);
error = EINVAL;
goto out;
}
m_freem(control); /* empty control, just free it */
}
if(!(flags & PRUS_OOB)) {
sbappend(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, p);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, p);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
if (tp != NULL) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
} else {
if (sbspace(&so->so_snd) < -512) {
m_freem(m);
error = ENOBUFS;
goto out;
}
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
sbappend(&so->so_snd, m);
if (nam && tp->t_state < TCPS_SYN_SENT) {
/*
* Do implied connect if not yet connected,
* initialize window to default value, and
* initialize maxseg/maxopd using peer's cached
* MSS.
*/
#ifdef INET6
if (isipv6)
error = tcp6_connect(tp, nam, p);
else
#endif /* INET6 */
error = tcp_connect(tp, nam, p);
if (error)
goto out;
tp->snd_wnd = TTCP_CLIENT_SND_WND;
tcp_mss(tp, -1);
}
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_force = 1;
error = tcp_output(tp);
tp->t_force = 0;
}
COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
}
static int
tcp_connect_oncpu(struct tcpcb *tp, int flags, struct mbuf *m,
struct sockaddr_in *sin, struct sockaddr_in *if_sin)
{
struct inpcb *inp = tp->t_inpcb, *oinp;
struct socket *so = inp->inp_socket;
struct route *ro = &inp->inp_route;
oinp = in_pcblookup_hash(&tcbinfo[mycpu->gd_cpuid],
sin->sin_addr, sin->sin_port,
(inp->inp_laddr.s_addr != INADDR_ANY ?
inp->inp_laddr : if_sin->sin_addr),
inp->inp_lport, 0, NULL);
if (oinp != NULL) {
m_freem(m);
return (EADDRINUSE);
}
if (inp->inp_laddr.s_addr == INADDR_ANY)
inp->inp_laddr = if_sin->sin_addr;
inp->inp_faddr = sin->sin_addr;
inp->inp_fport = sin->sin_port;
inp->inp_cpcbinfo = &tcbinfo[mycpu->gd_cpuid];
in_pcbinsconnhash(inp);
/*
* We are now on the inpcb's owner CPU, if the cached route was
* freed because the rtentry's owner CPU is not the current CPU
* (e.g. in tcp_connect()), then we try to reallocate it here with
* the hope that a rtentry may be cloned from a RTF_PRCLONING
* rtentry.
*/
if (!(inp->inp_socket->so_options & SO_DONTROUTE) && /*XXX*/
ro->ro_rt == NULL) {
bzero(&ro->ro_dst, sizeof(struct sockaddr_in));
ro->ro_dst.sa_family = AF_INET;
ro->ro_dst.sa_len = sizeof(struct sockaddr_in);
((struct sockaddr_in *)&ro->ro_dst)->sin_addr =
sin->sin_addr;
rtalloc(ro);
}
/*
* Now that no more errors can occur, change the protocol processing
* port to the current thread (which is the correct thread).
*
* Create TCP timer message now; we are on the tcpcb's owner
* CPU/thread.
*/
tcp_create_timermsg(tp, &curthread->td_msgport);
/*
* Compute window scaling to request. Use a larger scaling then
* needed for the initial receive buffer in case the receive buffer
* gets expanded.
*/
if (tp->request_r_scale < TCP_MIN_WINSHIFT)
tp->request_r_scale = TCP_MIN_WINSHIFT;
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat
) {
tp->request_r_scale++;
}
soisconnecting(so);
tcpstat.tcps_connattempt++;
tp->t_state = TCPS_SYN_SENT;
tcp_callout_reset(tp, tp->tt_keep, tcp_keepinit, tcp_timer_keep);
tp->iss = tcp_new_isn(tp);
tcp_sendseqinit(tp);
if (m) {
ssb_appendstream(&so->so_snd, m);
m = NULL;
if (flags & PRUS_OOB)
tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
}
/*
* Close the send side of the connection after
* the data is sent if flagged.
*/
if ((flags & (PRUS_OOB|PRUS_EOF)) == PRUS_EOF) {
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
return (tcp_output(tp));
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static void
tcp_usr_send(netmsg_t msg)
{
struct socket *so = msg->send.base.nm_so;
int flags = msg->send.nm_flags;
struct mbuf *m = msg->send.nm_m;
struct sockaddr *nam = msg->send.nm_addr;
struct mbuf *control = msg->send.nm_control;
struct thread *td = msg->send.nm_td;
int error = 0;
struct inpcb *inp;
struct tcpcb *tp;
#ifdef INET6
int isipv6;
#endif
TCPDEBUG0;
inp = so->so_pcb;
if (inp == NULL) {
/*
* OOPS! we lost a race, the TCP session got reset after
* we checked SS_CANTSENDMORE, eg: while doing uiomove or a
* network interrupt in the non-critical section of sosend().
*/
m_freem(m);
if (control)
m_freem(control);
error = ECONNRESET; /* XXX EPIPE? */
tp = NULL;
TCPDEBUG1();
goto out;
}
#ifdef INET6
isipv6 = nam && nam->sa_family == AF_INET6;
#endif /* INET6 */
tp = intotcpcb(inp);
TCPDEBUG1();
if (control) {
/* TCP doesn't do control messages (rights, creds, etc) */
if (control->m_len) {
m_freem(control);
m_freem(m);
error = EINVAL;
goto out;
}
m_freem(control); /* empty control, just free it */
}
/*
* Don't let too much OOB data build up
*/
if (flags & PRUS_OOB) {
if (ssb_space(&so->so_snd) < -512) {
m_freem(m);
error = ENOBUFS;
goto out;
}
}
/*
* Do implied connect if not yet connected. Any data sent
* with the connect is handled by tcp_connect() and friends.
*
* NOTE! PROTOCOL THREAD MAY BE CHANGED BY THE CONNECT!
*/
if (nam && tp->t_state < TCPS_SYN_SENT) {
kprintf("implied fallback\n");
msg->connect.nm_nam = nam;
msg->connect.nm_td = td;
msg->connect.nm_m = m;
msg->connect.nm_flags = flags;
msg->connect.nm_reconnect = NMSG_RECONNECT_FALLBACK;
#ifdef INET6
if (isipv6)
tcp6_connect(msg);
else
#endif /* INET6 */
tcp_connect(msg);
/* msg invalid now */
return;
}
/*
* Pump the data into the socket.
*/
if (m)
ssb_appendstream(&so->so_snd, m);
if (flags & PRUS_OOB) {
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
tp->t_flags |= TF_FORCE;
error = tcp_output(tp);
tp->t_flags &= ~TF_FORCE;
} else {
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
if (tp != NULL) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
}
COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
}
static int
tcp6_connect_oncpu(struct tcpcb *tp, int flags, struct mbuf **mp,
struct sockaddr_in6 *sin6, struct in6_addr *addr6)
{
struct mbuf *m = *mp;
struct inpcb *inp = tp->t_inpcb;
struct socket *so = inp->inp_socket;
struct inpcb *oinp;
/*
* 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.
*/
oinp = in6_pcblookup_hash(inp->inp_cpcbinfo,
&sin6->sin6_addr, sin6->sin6_port,
(IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr) ?
addr6 : &inp->in6p_laddr),
inp->inp_lport, 0, NULL);
if (oinp)
return (EADDRINUSE);
if (IN6_IS_ADDR_UNSPECIFIED(&inp->in6p_laddr))
inp->in6p_laddr = *addr6;
inp->in6p_faddr = sin6->sin6_addr;
inp->inp_fport = sin6->sin6_port;
if ((sin6->sin6_flowinfo & IPV6_FLOWINFO_MASK) != 0)
inp->in6p_flowinfo = sin6->sin6_flowinfo;
in_pcbinsconnhash(inp);
/*
* Now that no more errors can occur, change the protocol processing
* port to the current thread (which is the correct thread).
*
* Create TCP timer message now; we are on the tcpcb's owner
* CPU/thread.
*/
tcp_create_timermsg(tp, &curthread->td_msgport);
/* Compute window scaling to request. */
if (tp->request_r_scale < TCP_MIN_WINSHIFT)
tp->request_r_scale = TCP_MIN_WINSHIFT;
while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
(TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.ssb_hiwat) {
tp->request_r_scale++;
}
soisconnecting(so);
tcpstat.tcps_connattempt++;
tp->t_state = TCPS_SYN_SENT;
tcp_callout_reset(tp, tp->tt_keep, tcp_keepinit, tcp_timer_keep);
tp->iss = tcp_new_isn(tp);
tcp_sendseqinit(tp);
if (m) {
ssb_appendstream(&so->so_snd, m);
*mp = NULL;
if (flags & PRUS_OOB)
tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
}
/*
* Close the send side of the connection after
* the data is sent if flagged.
*/
if ((flags & (PRUS_OOB|PRUS_EOF)) == PRUS_EOF) {
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
return (tcp_output(tp));
}
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data. Unlike the other
* pru_*() routines, the mbuf chains are our responsibility. We
* must either enqueue them or free them. The other pru_* routines
* generally are caller-frees.
*/
static void
tcp_usr_send(netmsg_t msg)
{
struct socket *so = msg->send.base.nm_so;
int flags = msg->send.nm_flags;
struct mbuf *m = msg->send.nm_m;
int error = 0;
struct inpcb *inp;
struct tcpcb *tp;
TCPDEBUG0;
KKASSERT(msg->send.nm_control == NULL);
KKASSERT(msg->send.nm_addr == NULL);
KKASSERT((flags & PRUS_FREEADDR) == 0);
inp = so->so_pcb;
if (inp == NULL) {
/*
* OOPS! we lost a race, the TCP session got reset after
* we checked SS_CANTSENDMORE, eg: while doing uiomove or a
* network interrupt in the non-critical section of sosend().
*/
m_freem(m);
error = ECONNRESET; /* XXX EPIPE? */
tp = NULL;
TCPDEBUG1();
goto out;
}
tp = intotcpcb(inp);
TCPDEBUG1();
#ifdef foo
/*
* This is no longer necessary, since:
* - sosendtcp() has already checked it for us
* - It does not work with asynchronized send
*/
/*
* Don't let too much OOB data build up
*/
if (flags & PRUS_OOB) {
if (ssb_space(&so->so_snd) < -512) {
m_freem(m);
error = ENOBUFS;
goto out;
}
}
#endif
/*
* Pump the data into the socket.
*/
if (m) {
ssb_appendstream(&so->so_snd, m);
sowwakeup(so);
}
if (flags & PRUS_OOB) {
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
tp->snd_up = tp->snd_una + so->so_snd.ssb_cc;
tp->t_flags |= TF_FORCE;
error = tcp_output(tp);
tp->t_flags &= ~TF_FORCE;
} else {
if (flags & PRUS_EOF) {
/*
* Close the send side of the connection after
* the data is sent.
*/
socantsendmore(so);
tp = tcp_usrclosed(tp);
}
if (tp != NULL && !tcp_output_pending(tp)) {
if (flags & PRUS_MORETOCOME)
tp->t_flags |= TF_MORETOCOME;
error = tcp_output_fair(tp);
if (flags & PRUS_MORETOCOME)
tp->t_flags &= ~TF_MORETOCOME;
}
}
COMMON_END1((flags & PRUS_OOB) ? PRU_SENDOOB :
((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND),
(flags & PRUS_NOREPLY));
}
int
tcp_usrreq(struct socket * so,
struct mbuf * m,
struct mbuf * nam)
{
struct inpcb * inp;
struct tcpcb * tp;
int error = 0;
int req;
#ifdef DO_TCPTRACE
int ostate;
#endif
req = so->so_req; /* get request from socket struct */
inp = sotoinpcb(so);
/*
* When a TCP is attached to a socket, then there will be
* a (struct inpcb) pointed at by the socket, and this
* structure will point at a subsidary (struct tcpcb).
*/
if (inp == 0 && req != PRU_ATTACH)
{
return (EINVAL);
}
if (inp)
tp = intotcpcb(inp);
else /* inp and tp not set, make sure this is OK: */
{
if (req == PRU_ATTACH)
tp = NULL; /* stifle compiler warnings about using unassigned tp*/
else
{
dtrap(); /* programming error? */
return EINVAL;
}
}
switch (req)
{
/*
* TCP attaches to socket via PRU_ATTACH, reserving space,
* and an internet control block.
*/
case PRU_ATTACH:
if (inp)
{
error = EISCONN;
break;
}
error = tcp_attach(so);
if (error)
break;
if ((so->so_options & SO_LINGER) && so->so_linger == 0)
so->so_linger = TCP_LINGERTIME;
#ifdef DO_TCPTRACE
SETTP(tp, sototcpcb(so));
#endif
break;
/*
* PRU_DETACH detaches the TCP protocol from the socket.
* If the protocol state is non-embryonic, then can't
* do this directly: have to initiate a PRU_DISCONNECT,
* which may finish later; embryonic TCB's can just
* be discarded here.
*/
case PRU_DETACH:
if (tp->t_state > TCPS_LISTEN)
SETTP(tp, tcp_disconnect(tp));
else
SETTP(tp, tcp_close(tp));
break;
/*
* Give the socket an address.
*/
case PRU_BIND:
/* bind is quite different for IPv4 and v6, so we use two
* seperate pcbbind routines. so_domain was checked for
* validity way up in t_bind()
*/
#ifdef IP_V4
if(inp->inp_socket->so_domain == AF_INET)
{
error = in_pcbbind(inp, nam);
break;
}
#endif /* IP_V4 */
#ifdef IP_V6
if(inp->inp_socket->so_domain == AF_INET6)
{
error = ip6_pcbbind(inp, nam);
break;
}
#endif /* IP_V6 */
dtrap(); /* not v4 or v6? */
error = EINVAL;
break;
/*
* Prepare to accept connections.
*/
case PRU_LISTEN:
if (inp->inp_lport == 0)
error = in_pcbbind(inp, (struct mbuf *)0);
if (error == 0)
tp->t_state = TCPS_LISTEN;
break;
/*
* Initiate connection to peer.
* Create a template for use in transmissions on this connection.
* Enter SYN_SENT state, and mark socket as connecting.
* Start keep-alive timer, and seed output sequence space.
* Send initial segment on connection.
*/
case PRU_CONNECT:
if (inp->inp_lport == 0)
{
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
error = in_pcbbind(inp, (struct mbuf *)0);
#else /* dual mode */
if(so->so_domain == AF_INET)
error = in_pcbbind(inp, (struct mbuf *)0);
else
error = ip6_pcbbind(inp, (struct mbuf *)0);
#endif /* end dual mode code */
#else /* no v4, v6 only */
error = ip6_pcbbind(inp, (struct mbuf *)0);
#endif /* end v6 only */
if (error)
break;
}
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
error = in_pcbconnect(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET)
error = in_pcbconnect(inp, nam);
else
error = ip6_pcbconnect(inp, nam);
#endif /* end dual mode code */
#else /* no v4, v6 only */
error = ip6_pcbconnect(inp, nam);
#endif /* end v6 only */
if (error)
break;
tp->t_template = tcp_template(tp);
if (tp->t_template == 0)
{
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_pcbdisconnect(inp);
#else /* dual mode */
if(so->so_domain == AF_INET)
in_pcbdisconnect(inp);
else
ip6_pcbdisconnect(inp);
#endif /* end dual mode code */
#else /* no v4, v6 only */
ip6_pcbdisconnect(inp);
#endif /* end v6 only */
error = ENOBUFS;
break;
}
soisconnecting(so);
tcpstat.tcps_connattempt++;
tp->t_state = TCPS_SYN_SENT;
tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
tp->iss = tcp_iss;
tcp_iss += (tcp_seq)(TCP_ISSINCR/2);
tcp_sendseqinit(tp);
error = tcp_output(tp);
if (!error)
TCP_MIB_INC(tcpActiveOpens); /* keep MIB stats */
break;
/*
* Create a TCP connection between two sockets.
*/
case PRU_CONNECT2:
error = EOPNOTSUPP;
break;
/*
* Initiate disconnect from peer.
* If connection never passed embryonic stage, just drop;
* else if don't need to let data drain, then can just drop anyways,
* else have to begin TCP shutdown process: mark socket disconnecting,
* drain unread data, state switch to reflect user close, and
* send segment (e.g. FIN) to peer. Socket will be really disconnected
* when peer sends FIN and acks ours.
*
* SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
*/
case PRU_DISCONNECT:
SETTP(tp, tcp_disconnect(tp));
break;
/*
* Accept a connection. Essentially all the work is
* done at higher levels; just return the address
* of the peer, storing through addr.
*/
case PRU_ACCEPT:
{
struct sockaddr_in * sin = mtod(nam, struct sockaddr_in *);
#ifdef IP_V6
struct sockaddr_in6 * sin6 = mtod(nam, struct sockaddr_in6 *);
#endif
#ifdef IP_V6
if (so->so_domain == AF_INET6)
{
nam->m_len = sizeof (struct sockaddr_in6);
sin6->sin6_port = inp->inp_fport;
sin6->sin6_family = AF_INET6;
IP6CPY(&sin6->sin6_addr, &inp->ip6_faddr);
}
#endif
#ifdef IP_V4
if (so->so_domain == AF_INET)
{
nam->m_len = sizeof (struct sockaddr_in);
sin->sin_family = AF_INET;
sin->sin_port = inp->inp_fport;
sin->sin_addr = inp->inp_faddr;
}
#endif
if ( !(so->so_domain == AF_INET) &&
!(so->so_domain == AF_INET6)
)
{
dprintf("*** PRU_ACCEPT bad domain = %d\n", so->so_domain);
dtrap();
}
TCP_MIB_INC(tcpPassiveOpens); /* keep MIB stats */
break;
}
/*
* Mark the connection as being incapable of further output.
*/
case PRU_SHUTDOWN:
socantsendmore(so);
tp = tcp_usrclosed(tp);
if (tp)
error = tcp_output(tp);
break;
/*
* After a receive, possibly send window update to peer.
*/
case PRU_RCVD:
(void) tcp_output(tp);
break;
/*
* Do a send by putting data in output queue and updating urgent
* marker if URG set. Possibly send more data.
*/
case PRU_SEND:
if (so->so_pcb == NULL)
{ /* Return EPIPE error if socket is not connected */
error = EPIPE;
break;
}
sbappend(&so->so_snd, m);
error = tcp_output(tp);
if (error == ENOBUFS)
sbdropend(&so->so_snd,m); /* Remove data from socket buffer */
break;
/*
* Abort the TCP.
*/
case PRU_ABORT:
SETTP(tp, tcp_drop(tp, ECONNABORTED));
break;
case PRU_SENSE:
/* ((struct stat *) m)->st_blksize = so->so_snd.sb_hiwat; */
dtrap(); /* does this ever happen? */
return (0);
case PRU_RCVOOB:
if ((so->so_oobmark == 0 &&
(so->so_state & SS_RCVATMARK) == 0) ||
#ifdef SO_OOBINLINE
so->so_options & SO_OOBINLINE ||
#endif
tp->t_oobflags & TCPOOB_HADDATA)
{
error = EINVAL;
break;
}
if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0)
{
error = EWOULDBLOCK;
break;
}
m->m_len = 1;
*mtod(m, char *) = tp->t_iobc;
if ((MBUF2LONG(nam) & MSG_PEEK) == 0)
tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
break;
case PRU_SENDOOB:
if (so->so_pcb == NULL)
{ /* Return EPIPE error if socket is not connected */
error = EPIPE;
break;
}
if (sbspace(&so->so_snd) == 0)
{
m_freem(m);
error = ENOBUFS;
break;
}
/*
* According to RFC961 (Assigned Protocols),
* the urgent pointer points to the last octet
* of urgent data. We continue, however,
* to consider it to indicate the first octet
* of data past the urgent section.
* Otherwise, snd_up should be one lower.
*/
sbappend(&so->so_snd, m);
tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
tp->t_force = 1;
error = tcp_output(tp);
if (error == ENOBUFS)
sbdropend(&so->so_snd,m); /* Remove data from socket buffer */
tp->t_force = 0;
break;
case PRU_SOCKADDR:
/* sockaddr and peeraddr have to switch based on IP type */
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_setsockaddr(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET6)
ip6_setsockaddr(inp, nam);
else
in_setsockaddr(inp, nam);
#endif /* dual mode */
#else /* IP_V6 */
ip6_setsockaddr(inp, nam);
#endif
break;
case PRU_PEERADDR:
#ifdef IP_V4
#ifndef IP_V6 /* v4 only */
in_setpeeraddr(inp, nam);
#else /* dual mode */
if(so->so_domain == AF_INET6)
ip6_setpeeraddr(inp, nam);
else
in_setpeeraddr(inp, nam);
#endif /* dual mode */
#else /* IP_V6 */
ip6_setpeeraddr(inp, nam);
#endif
break;
case PRU_SLOWTIMO:
SETTP(tp, tcp_timers(tp, (int)MBUF2LONG(nam)));
#ifdef DO_TCPTRACE
req |= (long)nam << 8; /* for debug's sake */
#endif
break;
default:
panic("tcp_usrreq");
}
#ifdef DO_TCPTRACE
if (tp && (so->so_options & SO_DEBUG))
tcp_trace("usrreq: state: %d, tcpcb: %x, req: %d",
ostate, tp, req);
#endif
return (error);
}
