xref: /freebsd/sys/netinet/tcp_usrreq.c (revision cfe8b629f1f13b06fa1aec801248b7e431e48aba)
1df8bae1dSRodney W. Grimes /*
2df8bae1dSRodney W. Grimes  * Copyright (c) 1982, 1986, 1988, 1993
3df8bae1dSRodney W. Grimes  *	The Regents of the University of California.  All rights reserved.
4df8bae1dSRodney W. Grimes  *
5df8bae1dSRodney W. Grimes  * Redistribution and use in source and binary forms, with or without
6df8bae1dSRodney W. Grimes  * modification, are permitted provided that the following conditions
7df8bae1dSRodney W. Grimes  * are met:
8df8bae1dSRodney W. Grimes  * 1. Redistributions of source code must retain the above copyright
9df8bae1dSRodney W. Grimes  *    notice, this list of conditions and the following disclaimer.
10df8bae1dSRodney W. Grimes  * 2. Redistributions in binary form must reproduce the above copyright
11df8bae1dSRodney W. Grimes  *    notice, this list of conditions and the following disclaimer in the
12df8bae1dSRodney W. Grimes  *    documentation and/or other materials provided with the distribution.
13df8bae1dSRodney W. Grimes  * 3. All advertising materials mentioning features or use of this software
14df8bae1dSRodney W. Grimes  *    must display the following acknowledgement:
15df8bae1dSRodney W. Grimes  *	This product includes software developed by the University of
16df8bae1dSRodney W. Grimes  *	California, Berkeley and its contributors.
17df8bae1dSRodney W. Grimes  * 4. Neither the name of the University nor the names of its contributors
18df8bae1dSRodney W. Grimes  *    may be used to endorse or promote products derived from this software
19df8bae1dSRodney W. Grimes  *    without specific prior written permission.
20df8bae1dSRodney W. Grimes  *
21df8bae1dSRodney W. Grimes  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22df8bae1dSRodney W. Grimes  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23df8bae1dSRodney W. Grimes  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24df8bae1dSRodney W. Grimes  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25df8bae1dSRodney W. Grimes  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26df8bae1dSRodney W. Grimes  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27df8bae1dSRodney W. Grimes  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28df8bae1dSRodney W. Grimes  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29df8bae1dSRodney W. Grimes  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30df8bae1dSRodney W. Grimes  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31df8bae1dSRodney W. Grimes  * SUCH DAMAGE.
32df8bae1dSRodney W. Grimes  *
331fdbc7aeSGarrett Wollman  *	From: @(#)tcp_usrreq.c	8.2 (Berkeley) 1/3/94
34cfe8b629SGarrett Wollman  *	$Id: tcp_usrreq.c,v 1.37 1998/01/27 09:15:11 davidg Exp $
35df8bae1dSRodney W. Grimes  */
36df8bae1dSRodney W. Grimes 
370cc12cc5SJoerg Wunsch #include "opt_tcpdebug.h"
380cc12cc5SJoerg Wunsch 
39df8bae1dSRodney W. Grimes #include <sys/param.h>
40df8bae1dSRodney W. Grimes #include <sys/systm.h>
41c7a82f90SGarrett Wollman #include <sys/kernel.h>
4298163b98SPoul-Henning Kamp #include <sys/sysctl.h>
43df8bae1dSRodney W. Grimes #include <sys/mbuf.h>
44df8bae1dSRodney W. Grimes #include <sys/socket.h>
45df8bae1dSRodney W. Grimes #include <sys/socketvar.h>
46df8bae1dSRodney W. Grimes #include <sys/protosw.h>
47df8bae1dSRodney W. Grimes 
48df8bae1dSRodney W. Grimes #include <net/if.h>
49df8bae1dSRodney W. Grimes #include <net/route.h>
50df8bae1dSRodney W. Grimes 
51df8bae1dSRodney W. Grimes #include <netinet/in.h>
52df8bae1dSRodney W. Grimes #include <netinet/in_systm.h>
53df8bae1dSRodney W. Grimes #include <netinet/in_pcb.h>
54b5e8ce9fSBruce Evans #include <netinet/in_var.h>
55df8bae1dSRodney W. Grimes #include <netinet/ip_var.h>
56df8bae1dSRodney W. Grimes #include <netinet/tcp.h>
57df8bae1dSRodney W. Grimes #include <netinet/tcp_fsm.h>
58df8bae1dSRodney W. Grimes #include <netinet/tcp_seq.h>
59df8bae1dSRodney W. Grimes #include <netinet/tcp_timer.h>
60df8bae1dSRodney W. Grimes #include <netinet/tcp_var.h>
61df8bae1dSRodney W. Grimes #include <netinet/tcpip.h>
62610ee2f9SDavid Greenman #ifdef TCPDEBUG
63df8bae1dSRodney W. Grimes #include <netinet/tcp_debug.h>
64610ee2f9SDavid Greenman #endif
65df8bae1dSRodney W. Grimes 
66df8bae1dSRodney W. Grimes /*
67df8bae1dSRodney W. Grimes  * TCP protocol interface to socket abstraction.
68df8bae1dSRodney W. Grimes  */
69117bcae7SGarrett Wollman extern	char *tcpstates[];	/* XXX ??? */
70df8bae1dSRodney W. Grimes 
71a29f300eSGarrett Wollman static int	tcp_attach __P((struct socket *, struct proc *));
7257bf258eSGarrett Wollman static int	tcp_connect __P((struct tcpcb *, struct sockaddr *,
73a29f300eSGarrett Wollman 				 struct proc *));
740312fbe9SPoul-Henning Kamp static struct tcpcb *
750312fbe9SPoul-Henning Kamp 		tcp_disconnect __P((struct tcpcb *));
760312fbe9SPoul-Henning Kamp static struct tcpcb *
770312fbe9SPoul-Henning Kamp 		tcp_usrclosed __P((struct tcpcb *));
782c37256eSGarrett Wollman 
792c37256eSGarrett Wollman #ifdef TCPDEBUG
802c37256eSGarrett Wollman #define	TCPDEBUG0	int ostate
812c37256eSGarrett Wollman #define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
82af7a2999SDavid Greenman #define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
832c37256eSGarrett Wollman 				tcp_trace(TA_USER, ostate, tp, 0, req)
842c37256eSGarrett Wollman #else
852c37256eSGarrett Wollman #define	TCPDEBUG0
862c37256eSGarrett Wollman #define	TCPDEBUG1()
872c37256eSGarrett Wollman #define	TCPDEBUG2(req)
882c37256eSGarrett Wollman #endif
892c37256eSGarrett Wollman 
902c37256eSGarrett Wollman /*
912c37256eSGarrett Wollman  * TCP attaches to socket via pru_attach(), reserving space,
922c37256eSGarrett Wollman  * and an internet control block.
932c37256eSGarrett Wollman  */
942c37256eSGarrett Wollman static int
95a29f300eSGarrett Wollman tcp_usr_attach(struct socket *so, int proto, struct proc *p)
962c37256eSGarrett Wollman {
972c37256eSGarrett Wollman 	int s = splnet();
982c37256eSGarrett Wollman 	int error;
992c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1002c37256eSGarrett Wollman 	struct tcpcb *tp = 0;
1012c37256eSGarrett Wollman 	TCPDEBUG0;
1022c37256eSGarrett Wollman 
1032c37256eSGarrett Wollman 	TCPDEBUG1();
1042c37256eSGarrett Wollman 	if (inp) {
1052c37256eSGarrett Wollman 		error = EISCONN;
1062c37256eSGarrett Wollman 		goto out;
1072c37256eSGarrett Wollman 	}
1082c37256eSGarrett Wollman 
109a29f300eSGarrett Wollman 	error = tcp_attach(so, p);
1102c37256eSGarrett Wollman 	if (error)
1112c37256eSGarrett Wollman 		goto out;
1122c37256eSGarrett Wollman 
1132c37256eSGarrett Wollman 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1142c37256eSGarrett Wollman 		so->so_linger = TCP_LINGERTIME * hz;
1152c37256eSGarrett Wollman 	tp = sototcpcb(so);
1162c37256eSGarrett Wollman out:
1172c37256eSGarrett Wollman 	TCPDEBUG2(PRU_ATTACH);
1182c37256eSGarrett Wollman 	splx(s);
1192c37256eSGarrett Wollman 	return error;
1202c37256eSGarrett Wollman }
1212c37256eSGarrett Wollman 
1222c37256eSGarrett Wollman /*
1232c37256eSGarrett Wollman  * pru_detach() detaches the TCP protocol from the socket.
1242c37256eSGarrett Wollman  * If the protocol state is non-embryonic, then can't
1252c37256eSGarrett Wollman  * do this directly: have to initiate a pru_disconnect(),
1262c37256eSGarrett Wollman  * which may finish later; embryonic TCB's can just
1272c37256eSGarrett Wollman  * be discarded here.
1282c37256eSGarrett Wollman  */
1292c37256eSGarrett Wollman static int
1302c37256eSGarrett Wollman tcp_usr_detach(struct socket *so)
1312c37256eSGarrett Wollman {
1322c37256eSGarrett Wollman 	int s = splnet();
1332c37256eSGarrett Wollman 	int error = 0;
1342c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1352c37256eSGarrett Wollman 	struct tcpcb *tp;
1362c37256eSGarrett Wollman 	TCPDEBUG0;
1372c37256eSGarrett Wollman 
1382c37256eSGarrett Wollman 	if (inp == 0) {
1392c37256eSGarrett Wollman 		splx(s);
1402c37256eSGarrett Wollman 		return EINVAL;	/* XXX */
1412c37256eSGarrett Wollman 	}
1422c37256eSGarrett Wollman 	tp = intotcpcb(inp);
1432c37256eSGarrett Wollman 	TCPDEBUG1();
1442c37256eSGarrett Wollman 	tp = tcp_disconnect(tp);
1452c37256eSGarrett Wollman 
1462c37256eSGarrett Wollman 	TCPDEBUG2(PRU_DETACH);
1472c37256eSGarrett Wollman 	splx(s);
1482c37256eSGarrett Wollman 	return error;
1492c37256eSGarrett Wollman }
1502c37256eSGarrett Wollman 
1512c37256eSGarrett Wollman #define	COMMON_START()	TCPDEBUG0; \
1522c37256eSGarrett Wollman 			do { \
1532c37256eSGarrett Wollman 				     if (inp == 0) { \
1542c37256eSGarrett Wollman 					     splx(s); \
1552c37256eSGarrett Wollman 					     return EINVAL; \
1562c37256eSGarrett Wollman 				     } \
1572c37256eSGarrett Wollman 				     tp = intotcpcb(inp); \
1582c37256eSGarrett Wollman 				     TCPDEBUG1(); \
1592c37256eSGarrett Wollman 		     } while(0)
1602c37256eSGarrett Wollman 
1612c37256eSGarrett Wollman #define COMMON_END(req)	out: TCPDEBUG2(req); splx(s); return error; goto out
1622c37256eSGarrett Wollman 
1632c37256eSGarrett Wollman 
1642c37256eSGarrett Wollman /*
1652c37256eSGarrett Wollman  * Give the socket an address.
1662c37256eSGarrett Wollman  */
1672c37256eSGarrett Wollman static int
16857bf258eSGarrett Wollman tcp_usr_bind(struct socket *so, struct sockaddr *nam, struct proc *p)
1692c37256eSGarrett Wollman {
1702c37256eSGarrett Wollman 	int s = splnet();
1712c37256eSGarrett Wollman 	int error = 0;
1722c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1732c37256eSGarrett Wollman 	struct tcpcb *tp;
1742c37256eSGarrett Wollman 	struct sockaddr_in *sinp;
1752c37256eSGarrett Wollman 
1762c37256eSGarrett Wollman 	COMMON_START();
1772c37256eSGarrett Wollman 
1782c37256eSGarrett Wollman 	/*
1792c37256eSGarrett Wollman 	 * Must check for multicast addresses and disallow binding
1802c37256eSGarrett Wollman 	 * to them.
1812c37256eSGarrett Wollman 	 */
18257bf258eSGarrett Wollman 	sinp = (struct sockaddr_in *)nam;
1832c37256eSGarrett Wollman 	if (sinp->sin_family == AF_INET &&
1842c37256eSGarrett Wollman 	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
1852c37256eSGarrett Wollman 		error = EAFNOSUPPORT;
1862c37256eSGarrett Wollman 		goto out;
1872c37256eSGarrett Wollman 	}
188a29f300eSGarrett Wollman 	error = in_pcbbind(inp, nam, p);
1892c37256eSGarrett Wollman 	if (error)
1902c37256eSGarrett Wollman 		goto out;
1912c37256eSGarrett Wollman 	COMMON_END(PRU_BIND);
1922c37256eSGarrett Wollman 
1932c37256eSGarrett Wollman }
1942c37256eSGarrett Wollman 
1952c37256eSGarrett Wollman /*
1962c37256eSGarrett Wollman  * Prepare to accept connections.
1972c37256eSGarrett Wollman  */
1982c37256eSGarrett Wollman static int
199a29f300eSGarrett Wollman tcp_usr_listen(struct socket *so, struct proc *p)
2002c37256eSGarrett Wollman {
2012c37256eSGarrett Wollman 	int s = splnet();
2022c37256eSGarrett Wollman 	int error = 0;
2032c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2042c37256eSGarrett Wollman 	struct tcpcb *tp;
2052c37256eSGarrett Wollman 
2062c37256eSGarrett Wollman 	COMMON_START();
2072c37256eSGarrett Wollman 	if (inp->inp_lport == 0)
20857bf258eSGarrett Wollman 		error = in_pcbbind(inp, (struct sockaddr *)0, p);
2092c37256eSGarrett Wollman 	if (error == 0)
2102c37256eSGarrett Wollman 		tp->t_state = TCPS_LISTEN;
2112c37256eSGarrett Wollman 	COMMON_END(PRU_LISTEN);
2122c37256eSGarrett Wollman }
2132c37256eSGarrett Wollman 
2142c37256eSGarrett Wollman /*
2152c37256eSGarrett Wollman  * Initiate connection to peer.
2162c37256eSGarrett Wollman  * Create a template for use in transmissions on this connection.
2172c37256eSGarrett Wollman  * Enter SYN_SENT state, and mark socket as connecting.
2182c37256eSGarrett Wollman  * Start keep-alive timer, and seed output sequence space.
2192c37256eSGarrett Wollman  * Send initial segment on connection.
2202c37256eSGarrett Wollman  */
2212c37256eSGarrett Wollman static int
22257bf258eSGarrett Wollman tcp_usr_connect(struct socket *so, struct sockaddr *nam, struct proc *p)
2232c37256eSGarrett Wollman {
2242c37256eSGarrett Wollman 	int s = splnet();
2252c37256eSGarrett Wollman 	int error = 0;
2262c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2272c37256eSGarrett Wollman 	struct tcpcb *tp;
2282c37256eSGarrett Wollman 	struct sockaddr_in *sinp;
2292c37256eSGarrett Wollman 
2302c37256eSGarrett Wollman 	COMMON_START();
2312c37256eSGarrett Wollman 
2322c37256eSGarrett Wollman 	/*
2332c37256eSGarrett Wollman 	 * Must disallow TCP ``connections'' to multicast addresses.
2342c37256eSGarrett Wollman 	 */
23557bf258eSGarrett Wollman 	sinp = (struct sockaddr_in *)nam;
2362c37256eSGarrett Wollman 	if (sinp->sin_family == AF_INET
2372c37256eSGarrett Wollman 	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
2382c37256eSGarrett Wollman 		error = EAFNOSUPPORT;
2392c37256eSGarrett Wollman 		goto out;
2402c37256eSGarrett Wollman 	}
2412c37256eSGarrett Wollman 
242a29f300eSGarrett Wollman 	if ((error = tcp_connect(tp, nam, p)) != 0)
2432c37256eSGarrett Wollman 		goto out;
2442c37256eSGarrett Wollman 	error = tcp_output(tp);
2452c37256eSGarrett Wollman 	COMMON_END(PRU_CONNECT);
2462c37256eSGarrett Wollman }
2472c37256eSGarrett Wollman 
2482c37256eSGarrett Wollman /*
2492c37256eSGarrett Wollman  * Initiate disconnect from peer.
2502c37256eSGarrett Wollman  * If connection never passed embryonic stage, just drop;
2512c37256eSGarrett Wollman  * else if don't need to let data drain, then can just drop anyways,
2522c37256eSGarrett Wollman  * else have to begin TCP shutdown process: mark socket disconnecting,
2532c37256eSGarrett Wollman  * drain unread data, state switch to reflect user close, and
2542c37256eSGarrett Wollman  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
2552c37256eSGarrett Wollman  * when peer sends FIN and acks ours.
2562c37256eSGarrett Wollman  *
2572c37256eSGarrett Wollman  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
2582c37256eSGarrett Wollman  */
2592c37256eSGarrett Wollman static int
2602c37256eSGarrett Wollman tcp_usr_disconnect(struct socket *so)
2612c37256eSGarrett Wollman {
2622c37256eSGarrett Wollman 	int s = splnet();
2632c37256eSGarrett Wollman 	int error = 0;
2642c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2652c37256eSGarrett Wollman 	struct tcpcb *tp;
2662c37256eSGarrett Wollman 
2672c37256eSGarrett Wollman 	COMMON_START();
2682c37256eSGarrett Wollman 	tp = tcp_disconnect(tp);
2692c37256eSGarrett Wollman 	COMMON_END(PRU_DISCONNECT);
2702c37256eSGarrett Wollman }
2712c37256eSGarrett Wollman 
2722c37256eSGarrett Wollman /*
2732c37256eSGarrett Wollman  * Accept a connection.  Essentially all the work is
2742c37256eSGarrett Wollman  * done at higher levels; just return the address
2752c37256eSGarrett Wollman  * of the peer, storing through addr.
2762c37256eSGarrett Wollman  */
2772c37256eSGarrett Wollman static int
27857bf258eSGarrett Wollman tcp_usr_accept(struct socket *so, struct sockaddr **nam)
2792c37256eSGarrett Wollman {
2802c37256eSGarrett Wollman 	int s = splnet();
2812c37256eSGarrett Wollman 	int error = 0;
2822c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2832c37256eSGarrett Wollman 	struct tcpcb *tp;
2842c37256eSGarrett Wollman 
2852c37256eSGarrett Wollman 	COMMON_START();
286117bcae7SGarrett Wollman 	in_setpeeraddr(so, nam);
2872c37256eSGarrett Wollman 	COMMON_END(PRU_ACCEPT);
2882c37256eSGarrett Wollman }
2892c37256eSGarrett Wollman 
2902c37256eSGarrett Wollman /*
2912c37256eSGarrett Wollman  * Mark the connection as being incapable of further output.
2922c37256eSGarrett Wollman  */
2932c37256eSGarrett Wollman static int
2942c37256eSGarrett Wollman tcp_usr_shutdown(struct socket *so)
2952c37256eSGarrett Wollman {
2962c37256eSGarrett Wollman 	int s = splnet();
2972c37256eSGarrett Wollman 	int error = 0;
2982c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2992c37256eSGarrett Wollman 	struct tcpcb *tp;
3002c37256eSGarrett Wollman 
3012c37256eSGarrett Wollman 	COMMON_START();
3022c37256eSGarrett Wollman 	socantsendmore(so);
3032c37256eSGarrett Wollman 	tp = tcp_usrclosed(tp);
3042c37256eSGarrett Wollman 	if (tp)
3052c37256eSGarrett Wollman 		error = tcp_output(tp);
3062c37256eSGarrett Wollman 	COMMON_END(PRU_SHUTDOWN);
3072c37256eSGarrett Wollman }
3082c37256eSGarrett Wollman 
3092c37256eSGarrett Wollman /*
3102c37256eSGarrett Wollman  * After a receive, possibly send window update to peer.
3112c37256eSGarrett Wollman  */
3122c37256eSGarrett Wollman static int
3132c37256eSGarrett Wollman tcp_usr_rcvd(struct socket *so, int flags)
3142c37256eSGarrett Wollman {
3152c37256eSGarrett Wollman 	int s = splnet();
3162c37256eSGarrett Wollman 	int error = 0;
3172c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
3182c37256eSGarrett Wollman 	struct tcpcb *tp;
3192c37256eSGarrett Wollman 
3202c37256eSGarrett Wollman 	COMMON_START();
3212c37256eSGarrett Wollman 	tcp_output(tp);
3222c37256eSGarrett Wollman 	COMMON_END(PRU_RCVD);
3232c37256eSGarrett Wollman }
3242c37256eSGarrett Wollman 
3252c37256eSGarrett Wollman /*
3262c37256eSGarrett Wollman  * Do a send by putting data in output queue and updating urgent
3272c37256eSGarrett Wollman  * marker if URG set.  Possibly send more data.
3282c37256eSGarrett Wollman  */
3292c37256eSGarrett Wollman static int
33057bf258eSGarrett Wollman tcp_usr_send(struct socket *so, int flags, struct mbuf *m,
33157bf258eSGarrett Wollman 	     struct sockaddr *nam, struct mbuf *control, struct proc *p)
3322c37256eSGarrett Wollman {
3332c37256eSGarrett Wollman 	int s = splnet();
3342c37256eSGarrett Wollman 	int error = 0;
3352c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
3362c37256eSGarrett Wollman 	struct tcpcb *tp;
3372c37256eSGarrett Wollman 
3382c37256eSGarrett Wollman 	COMMON_START();
3392c37256eSGarrett Wollman 	if (control && control->m_len) {
3402c37256eSGarrett Wollman 		m_freem(control); /* XXX shouldn't caller do this??? */
3412c37256eSGarrett Wollman 		if (m)
3422c37256eSGarrett Wollman 			m_freem(m);
343744f87eaSDavid Greenman 		error = EINVAL;
344744f87eaSDavid Greenman 		goto out;
3452c37256eSGarrett Wollman 	}
3462c37256eSGarrett Wollman 
3472c37256eSGarrett Wollman 	if(!(flags & PRUS_OOB)) {
3482c37256eSGarrett Wollman 		sbappend(&so->so_snd, m);
3492c37256eSGarrett Wollman 		if (nam && tp->t_state < TCPS_SYN_SENT) {
3502c37256eSGarrett Wollman 			/*
3512c37256eSGarrett Wollman 			 * Do implied connect if not yet connected,
3522c37256eSGarrett Wollman 			 * initialize window to default value, and
3532c37256eSGarrett Wollman 			 * initialize maxseg/maxopd using peer's cached
3542c37256eSGarrett Wollman 			 * MSS.
3552c37256eSGarrett Wollman 			 */
356a29f300eSGarrett Wollman 			error = tcp_connect(tp, nam, p);
3572c37256eSGarrett Wollman 			if (error)
3582c37256eSGarrett Wollman 				goto out;
3592c37256eSGarrett Wollman 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
3602c37256eSGarrett Wollman 			tcp_mss(tp, -1);
3612c37256eSGarrett Wollman 		}
3622c37256eSGarrett Wollman 
3632c37256eSGarrett Wollman 		if (flags & PRUS_EOF) {
3642c37256eSGarrett Wollman 			/*
3652c37256eSGarrett Wollman 			 * Close the send side of the connection after
3662c37256eSGarrett Wollman 			 * the data is sent.
3672c37256eSGarrett Wollman 			 */
3682c37256eSGarrett Wollman 			socantsendmore(so);
3692c37256eSGarrett Wollman 			tp = tcp_usrclosed(tp);
3702c37256eSGarrett Wollman 		}
3712c37256eSGarrett Wollman 		if (tp != NULL)
3722c37256eSGarrett Wollman 			error = tcp_output(tp);
3732c37256eSGarrett Wollman 	} else {
3742c37256eSGarrett Wollman 		if (sbspace(&so->so_snd) < -512) {
3752c37256eSGarrett Wollman 			m_freem(m);
3762c37256eSGarrett Wollman 			error = ENOBUFS;
3772c37256eSGarrett Wollman 			goto out;
3782c37256eSGarrett Wollman 		}
3792c37256eSGarrett Wollman 		/*
3802c37256eSGarrett Wollman 		 * According to RFC961 (Assigned Protocols),
3812c37256eSGarrett Wollman 		 * the urgent pointer points to the last octet
3822c37256eSGarrett Wollman 		 * of urgent data.  We continue, however,
3832c37256eSGarrett Wollman 		 * to consider it to indicate the first octet
3842c37256eSGarrett Wollman 		 * of data past the urgent section.
3852c37256eSGarrett Wollman 		 * Otherwise, snd_up should be one lower.
3862c37256eSGarrett Wollman 		 */
3872c37256eSGarrett Wollman 		sbappend(&so->so_snd, m);
388ef53690bSGarrett Wollman 		if (nam && tp->t_state < TCPS_SYN_SENT) {
389ef53690bSGarrett Wollman 			/*
390ef53690bSGarrett Wollman 			 * Do implied connect if not yet connected,
391ef53690bSGarrett Wollman 			 * initialize window to default value, and
392ef53690bSGarrett Wollman 			 * initialize maxseg/maxopd using peer's cached
393ef53690bSGarrett Wollman 			 * MSS.
394ef53690bSGarrett Wollman 			 */
395a29f300eSGarrett Wollman 			error = tcp_connect(tp, nam, p);
396ef53690bSGarrett Wollman 			if (error)
397ef53690bSGarrett Wollman 				goto out;
398ef53690bSGarrett Wollman 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
399ef53690bSGarrett Wollman 			tcp_mss(tp, -1);
400ef53690bSGarrett Wollman 		}
4012c37256eSGarrett Wollman 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
4022c37256eSGarrett Wollman 		tp->t_force = 1;
4032c37256eSGarrett Wollman 		error = tcp_output(tp);
4042c37256eSGarrett Wollman 		tp->t_force = 0;
4052c37256eSGarrett Wollman 	}
4062c37256eSGarrett Wollman 	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
4072c37256eSGarrett Wollman 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
4082c37256eSGarrett Wollman }
4092c37256eSGarrett Wollman 
4102c37256eSGarrett Wollman /*
4112c37256eSGarrett Wollman  * Abort the TCP.
4122c37256eSGarrett Wollman  */
4132c37256eSGarrett Wollman static int
4142c37256eSGarrett Wollman tcp_usr_abort(struct socket *so)
4152c37256eSGarrett Wollman {
4162c37256eSGarrett Wollman 	int s = splnet();
4172c37256eSGarrett Wollman 	int error = 0;
4182c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
4192c37256eSGarrett Wollman 	struct tcpcb *tp;
4202c37256eSGarrett Wollman 
4212c37256eSGarrett Wollman 	COMMON_START();
4222c37256eSGarrett Wollman 	tp = tcp_drop(tp, ECONNABORTED);
4232c37256eSGarrett Wollman 	COMMON_END(PRU_ABORT);
4242c37256eSGarrett Wollman }
4252c37256eSGarrett Wollman 
4262c37256eSGarrett Wollman /*
4272c37256eSGarrett Wollman  * Receive out-of-band data.
4282c37256eSGarrett Wollman  */
4292c37256eSGarrett Wollman static int
4302c37256eSGarrett Wollman tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
4312c37256eSGarrett Wollman {
4322c37256eSGarrett Wollman 	int s = splnet();
4332c37256eSGarrett Wollman 	int error = 0;
4342c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
4352c37256eSGarrett Wollman 	struct tcpcb *tp;
4362c37256eSGarrett Wollman 
4372c37256eSGarrett Wollman 	COMMON_START();
4382c37256eSGarrett Wollman 	if ((so->so_oobmark == 0 &&
4392c37256eSGarrett Wollman 	     (so->so_state & SS_RCVATMARK) == 0) ||
4402c37256eSGarrett Wollman 	    so->so_options & SO_OOBINLINE ||
4412c37256eSGarrett Wollman 	    tp->t_oobflags & TCPOOB_HADDATA) {
4422c37256eSGarrett Wollman 		error = EINVAL;
4432c37256eSGarrett Wollman 		goto out;
4442c37256eSGarrett Wollman 	}
4452c37256eSGarrett Wollman 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
4462c37256eSGarrett Wollman 		error = EWOULDBLOCK;
4472c37256eSGarrett Wollman 		goto out;
4482c37256eSGarrett Wollman 	}
4492c37256eSGarrett Wollman 	m->m_len = 1;
4502c37256eSGarrett Wollman 	*mtod(m, caddr_t) = tp->t_iobc;
4512c37256eSGarrett Wollman 	if ((flags & MSG_PEEK) == 0)
4522c37256eSGarrett Wollman 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
4532c37256eSGarrett Wollman 	COMMON_END(PRU_RCVOOB);
4542c37256eSGarrett Wollman }
4552c37256eSGarrett Wollman 
4562c37256eSGarrett Wollman /* xxx - should be const */
4572c37256eSGarrett Wollman struct pr_usrreqs tcp_usrreqs = {
4582c37256eSGarrett Wollman 	tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
459117bcae7SGarrett Wollman 	tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
460117bcae7SGarrett Wollman 	tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
461117bcae7SGarrett Wollman 	tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
462f8f6cbbaSPeter Wemm 	in_setsockaddr, sosend, soreceive, sopoll
4632c37256eSGarrett Wollman };
464df8bae1dSRodney W. Grimes 
465a0292f23SGarrett Wollman /*
466a0292f23SGarrett Wollman  * Common subroutine to open a TCP connection to remote host specified
467a0292f23SGarrett Wollman  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
468a0292f23SGarrett Wollman  * port number if needed.  Call in_pcbladdr to do the routing and to choose
469a0292f23SGarrett Wollman  * a local host address (interface).  If there is an existing incarnation
470a0292f23SGarrett Wollman  * of the same connection in TIME-WAIT state and if the remote host was
471a0292f23SGarrett Wollman  * sending CC options and if the connection duration was < MSL, then
472a0292f23SGarrett Wollman  * truncate the previous TIME-WAIT state and proceed.
473a0292f23SGarrett Wollman  * Initialize connection parameters and enter SYN-SENT state.
474a0292f23SGarrett Wollman  */
4750312fbe9SPoul-Henning Kamp static int
476a29f300eSGarrett Wollman tcp_connect(tp, nam, p)
477a0292f23SGarrett Wollman 	register struct tcpcb *tp;
47857bf258eSGarrett Wollman 	struct sockaddr *nam;
479a29f300eSGarrett Wollman 	struct proc *p;
480a0292f23SGarrett Wollman {
481a0292f23SGarrett Wollman 	struct inpcb *inp = tp->t_inpcb, *oinp;
482a0292f23SGarrett Wollman 	struct socket *so = inp->inp_socket;
483a0292f23SGarrett Wollman 	struct tcpcb *otp;
48457bf258eSGarrett Wollman 	struct sockaddr_in *sin = (struct sockaddr_in *)nam;
485a0292f23SGarrett Wollman 	struct sockaddr_in *ifaddr;
486a45d2726SAndras Olah 	struct rmxp_tao *taop;
487a45d2726SAndras Olah 	struct rmxp_tao tao_noncached;
488c3229e05SDavid Greenman 	int error;
489a0292f23SGarrett Wollman 
490a0292f23SGarrett Wollman 	if (inp->inp_lport == 0) {
49157bf258eSGarrett Wollman 		error = in_pcbbind(inp, (struct sockaddr *)0, p);
492a0292f23SGarrett Wollman 		if (error)
493a0292f23SGarrett Wollman 			return error;
494a0292f23SGarrett Wollman 	}
495a0292f23SGarrett Wollman 
496a0292f23SGarrett Wollman 	/*
497a0292f23SGarrett Wollman 	 * Cannot simply call in_pcbconnect, because there might be an
498a0292f23SGarrett Wollman 	 * earlier incarnation of this same connection still in
499a0292f23SGarrett Wollman 	 * TIME_WAIT state, creating an ADDRINUSE error.
500a0292f23SGarrett Wollman 	 */
501a0292f23SGarrett Wollman 	error = in_pcbladdr(inp, nam, &ifaddr);
502d3628763SRodney W. Grimes 	if (error)
503d3628763SRodney W. Grimes 		return error;
504c3229e05SDavid Greenman 	oinp = in_pcblookup_hash(inp->inp_pcbinfo,
505a0292f23SGarrett Wollman 	    sin->sin_addr, sin->sin_port,
506a0292f23SGarrett Wollman 	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
507a0292f23SGarrett Wollman 						: ifaddr->sin_addr,
508a0292f23SGarrett Wollman 	    inp->inp_lport,  0);
509a0292f23SGarrett Wollman 	if (oinp) {
510a0292f23SGarrett Wollman 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
511a0292f23SGarrett Wollman 		otp->t_state == TCPS_TIME_WAIT &&
512a0292f23SGarrett Wollman 		    otp->t_duration < TCPTV_MSL &&
513a0292f23SGarrett Wollman 		    (otp->t_flags & TF_RCVD_CC))
514a0292f23SGarrett Wollman 			otp = tcp_close(otp);
515a0292f23SGarrett Wollman 		else
516a0292f23SGarrett Wollman 			return EADDRINUSE;
517a0292f23SGarrett Wollman 	}
518a0292f23SGarrett Wollman 	if (inp->inp_laddr.s_addr == INADDR_ANY)
519a0292f23SGarrett Wollman 		inp->inp_laddr = ifaddr->sin_addr;
520a0292f23SGarrett Wollman 	inp->inp_faddr = sin->sin_addr;
521a0292f23SGarrett Wollman 	inp->inp_fport = sin->sin_port;
52215bd2b43SDavid Greenman 	in_pcbrehash(inp);
523a0292f23SGarrett Wollman 
524a0292f23SGarrett Wollman 	tp->t_template = tcp_template(tp);
525a0292f23SGarrett Wollman 	if (tp->t_template == 0) {
526a0292f23SGarrett Wollman 		in_pcbdisconnect(inp);
527a0292f23SGarrett Wollman 		return ENOBUFS;
528a0292f23SGarrett Wollman 	}
529a0292f23SGarrett Wollman 
530a0292f23SGarrett Wollman 	/* Compute window scaling to request.  */
531a0292f23SGarrett Wollman 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
532a0292f23SGarrett Wollman 	    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
533a0292f23SGarrett Wollman 		tp->request_r_scale++;
534a0292f23SGarrett Wollman 
535a0292f23SGarrett Wollman 	soisconnecting(so);
536a0292f23SGarrett Wollman 	tcpstat.tcps_connattempt++;
537a0292f23SGarrett Wollman 	tp->t_state = TCPS_SYN_SENT;
5387b40aa32SPaul Traina 	tp->t_timer[TCPT_KEEP] = tcp_keepinit;
539a0292f23SGarrett Wollman 	tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
540a0292f23SGarrett Wollman 	tcp_sendseqinit(tp);
541a45d2726SAndras Olah 
542a45d2726SAndras Olah 	/*
543a45d2726SAndras Olah 	 * Generate a CC value for this connection and
544a45d2726SAndras Olah 	 * check whether CC or CCnew should be used.
545a45d2726SAndras Olah 	 */
546a45d2726SAndras Olah 	if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
547a45d2726SAndras Olah 		taop = &tao_noncached;
548a45d2726SAndras Olah 		bzero(taop, sizeof(*taop));
549a45d2726SAndras Olah 	}
550a45d2726SAndras Olah 
551a0292f23SGarrett Wollman 	tp->cc_send = CC_INC(tcp_ccgen);
552a45d2726SAndras Olah 	if (taop->tao_ccsent != 0 &&
553a45d2726SAndras Olah 	    CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
554a45d2726SAndras Olah 		taop->tao_ccsent = tp->cc_send;
555a45d2726SAndras Olah 	} else {
556a45d2726SAndras Olah 		taop->tao_ccsent = 0;
557a45d2726SAndras Olah 		tp->t_flags |= TF_SENDCCNEW;
558a45d2726SAndras Olah 	}
559a0292f23SGarrett Wollman 
560a0292f23SGarrett Wollman 	return 0;
561a0292f23SGarrett Wollman }
562a0292f23SGarrett Wollman 
563cfe8b629SGarrett Wollman /*
564cfe8b629SGarrett Wollman  * The new sockopt interface makes it possible for us to block in the
565cfe8b629SGarrett Wollman  * copyin/out step (if we take a page fault).  Taking a page fault at
566cfe8b629SGarrett Wollman  * splnet() is probably a Bad Thing.  (Since sockets and pcbs both now
567cfe8b629SGarrett Wollman  * use TSM, there probably isn't any need for this function to run at
568cfe8b629SGarrett Wollman  * splnet() any more.  This needs more examination.)
569cfe8b629SGarrett Wollman  */
570df8bae1dSRodney W. Grimes int
571cfe8b629SGarrett Wollman tcp_ctloutput(so, sopt)
572df8bae1dSRodney W. Grimes 	struct socket *so;
573cfe8b629SGarrett Wollman 	struct sockopt *sopt;
574df8bae1dSRodney W. Grimes {
575cfe8b629SGarrett Wollman 	int	error, opt, optval, s;
576df8bae1dSRodney W. Grimes 	struct	inpcb *inp;
577cfe8b629SGarrett Wollman 	struct	tcpcb *tp;
578cfe8b629SGarrett Wollman 	struct	mbuf *m;
579df8bae1dSRodney W. Grimes 
580cfe8b629SGarrett Wollman 	error = 0;
581cfe8b629SGarrett Wollman 	s = splnet();		/* XXX */
582df8bae1dSRodney W. Grimes 	inp = sotoinpcb(so);
583df8bae1dSRodney W. Grimes 	if (inp == NULL) {
584df8bae1dSRodney W. Grimes 		splx(s);
585df8bae1dSRodney W. Grimes 		return (ECONNRESET);
586df8bae1dSRodney W. Grimes 	}
587cfe8b629SGarrett Wollman 	if (sopt->sopt_level != IPPROTO_TCP) {
588cfe8b629SGarrett Wollman 		error = ip_ctloutput(so, sopt);
589df8bae1dSRodney W. Grimes 		splx(s);
590df8bae1dSRodney W. Grimes 		return (error);
591df8bae1dSRodney W. Grimes 	}
592df8bae1dSRodney W. Grimes 	tp = intotcpcb(inp);
593df8bae1dSRodney W. Grimes 
594cfe8b629SGarrett Wollman 	switch (sopt->sopt_dir) {
595cfe8b629SGarrett Wollman 	case SOPT_SET:
596cfe8b629SGarrett Wollman 		switch (sopt->sopt_name) {
597df8bae1dSRodney W. Grimes 		case TCP_NODELAY:
598cfe8b629SGarrett Wollman 		case TCP_NOOPT:
599cfe8b629SGarrett Wollman 		case TCP_NOPUSH:
600cfe8b629SGarrett Wollman 			error = sooptcopyin(sopt, &optval, sizeof optval,
601cfe8b629SGarrett Wollman 					    sizeof optval);
602cfe8b629SGarrett Wollman 			if (error)
603cfe8b629SGarrett Wollman 				break;
604cfe8b629SGarrett Wollman 
605cfe8b629SGarrett Wollman 			switch (sopt->sopt_name) {
606cfe8b629SGarrett Wollman 			case TCP_NODELAY:
607cfe8b629SGarrett Wollman 				opt = TF_NODELAY;
608cfe8b629SGarrett Wollman 				break;
609cfe8b629SGarrett Wollman 			case TCP_NOOPT:
610cfe8b629SGarrett Wollman 				opt = TF_NOOPT;
611cfe8b629SGarrett Wollman 				break;
612cfe8b629SGarrett Wollman 			case TCP_NOPUSH:
613cfe8b629SGarrett Wollman 				opt = TF_NOPUSH;
614cfe8b629SGarrett Wollman 				break;
615cfe8b629SGarrett Wollman 			default:
616cfe8b629SGarrett Wollman 				opt = 0; /* dead code to fool gcc */
617cfe8b629SGarrett Wollman 				break;
618cfe8b629SGarrett Wollman 			}
619cfe8b629SGarrett Wollman 
620cfe8b629SGarrett Wollman 			if (optval)
621cfe8b629SGarrett Wollman 				tp->t_flags |= opt;
622df8bae1dSRodney W. Grimes 			else
623cfe8b629SGarrett Wollman 				tp->t_flags &= ~opt;
624df8bae1dSRodney W. Grimes 			break;
625df8bae1dSRodney W. Grimes 
626df8bae1dSRodney W. Grimes 		case TCP_MAXSEG:
627cfe8b629SGarrett Wollman 			error = sooptcopyin(sopt, &optval, sizeof optval,
628cfe8b629SGarrett Wollman 					    sizeof optval);
629cfe8b629SGarrett Wollman 			if (error)
630df8bae1dSRodney W. Grimes 				break;
631df8bae1dSRodney W. Grimes 
632cfe8b629SGarrett Wollman 			if (optval > 0 && optval <= tp->t_maxseg)
633cfe8b629SGarrett Wollman 				tp->t_maxseg = optval;
634a0292f23SGarrett Wollman 			else
635a0292f23SGarrett Wollman 				error = EINVAL;
636a0292f23SGarrett Wollman 			break;
637a0292f23SGarrett Wollman 
638df8bae1dSRodney W. Grimes 		default:
639df8bae1dSRodney W. Grimes 			error = ENOPROTOOPT;
640df8bae1dSRodney W. Grimes 			break;
641df8bae1dSRodney W. Grimes 		}
642df8bae1dSRodney W. Grimes 		break;
643df8bae1dSRodney W. Grimes 
644cfe8b629SGarrett Wollman 	case SOPT_GET:
645cfe8b629SGarrett Wollman 		switch (sopt->sopt_name) {
646df8bae1dSRodney W. Grimes 		case TCP_NODELAY:
647cfe8b629SGarrett Wollman 			optval = tp->t_flags & TF_NODELAY;
648df8bae1dSRodney W. Grimes 			break;
649df8bae1dSRodney W. Grimes 		case TCP_MAXSEG:
650cfe8b629SGarrett Wollman 			optval = tp->t_maxseg;
651df8bae1dSRodney W. Grimes 			break;
652a0292f23SGarrett Wollman 		case TCP_NOOPT:
653cfe8b629SGarrett Wollman 			optval = tp->t_flags & TF_NOOPT;
654a0292f23SGarrett Wollman 			break;
655a0292f23SGarrett Wollman 		case TCP_NOPUSH:
656cfe8b629SGarrett Wollman 			optval = tp->t_flags & TF_NOPUSH;
657a0292f23SGarrett Wollman 			break;
658df8bae1dSRodney W. Grimes 		default:
659df8bae1dSRodney W. Grimes 			error = ENOPROTOOPT;
660df8bae1dSRodney W. Grimes 			break;
661df8bae1dSRodney W. Grimes 		}
662cfe8b629SGarrett Wollman 		if (error == 0)
663cfe8b629SGarrett Wollman 			error = sooptcopyout(sopt, &optval, sizeof optval);
664df8bae1dSRodney W. Grimes 		break;
665df8bae1dSRodney W. Grimes 	}
666df8bae1dSRodney W. Grimes 	splx(s);
667df8bae1dSRodney W. Grimes 	return (error);
668df8bae1dSRodney W. Grimes }
669df8bae1dSRodney W. Grimes 
67026e30fbbSDavid Greenman /*
67126e30fbbSDavid Greenman  * tcp_sendspace and tcp_recvspace are the default send and receive window
67226e30fbbSDavid Greenman  * sizes, respectively.  These are obsolescent (this information should
67326e30fbbSDavid Greenman  * be set by the route).
67426e30fbbSDavid Greenman  */
67526e30fbbSDavid Greenman u_long	tcp_sendspace = 1024*16;
67698163b98SPoul-Henning Kamp SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
67798163b98SPoul-Henning Kamp 	CTLFLAG_RW, &tcp_sendspace , 0, "");
67826e30fbbSDavid Greenman u_long	tcp_recvspace = 1024*16;
67998163b98SPoul-Henning Kamp SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
68098163b98SPoul-Henning Kamp 	CTLFLAG_RW, &tcp_recvspace , 0, "");
681df8bae1dSRodney W. Grimes 
682df8bae1dSRodney W. Grimes /*
683df8bae1dSRodney W. Grimes  * Attach TCP protocol to socket, allocating
684df8bae1dSRodney W. Grimes  * internet protocol control block, tcp control block,
685df8bae1dSRodney W. Grimes  * bufer space, and entering LISTEN state if to accept connections.
686df8bae1dSRodney W. Grimes  */
6870312fbe9SPoul-Henning Kamp static int
688a29f300eSGarrett Wollman tcp_attach(so, p)
689df8bae1dSRodney W. Grimes 	struct socket *so;
690a29f300eSGarrett Wollman 	struct proc *p;
691df8bae1dSRodney W. Grimes {
692df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
693df8bae1dSRodney W. Grimes 	struct inpcb *inp;
694df8bae1dSRodney W. Grimes 	int error;
695df8bae1dSRodney W. Grimes 
696df8bae1dSRodney W. Grimes 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
697df8bae1dSRodney W. Grimes 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
698df8bae1dSRodney W. Grimes 		if (error)
699df8bae1dSRodney W. Grimes 			return (error);
700df8bae1dSRodney W. Grimes 	}
701a29f300eSGarrett Wollman 	error = in_pcballoc(so, &tcbinfo, p);
702df8bae1dSRodney W. Grimes 	if (error)
703df8bae1dSRodney W. Grimes 		return (error);
704df8bae1dSRodney W. Grimes 	inp = sotoinpcb(so);
705df8bae1dSRodney W. Grimes 	tp = tcp_newtcpcb(inp);
706df8bae1dSRodney W. Grimes 	if (tp == 0) {
707df8bae1dSRodney W. Grimes 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
708df8bae1dSRodney W. Grimes 
709df8bae1dSRodney W. Grimes 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
710df8bae1dSRodney W. Grimes 		in_pcbdetach(inp);
711df8bae1dSRodney W. Grimes 		so->so_state |= nofd;
712df8bae1dSRodney W. Grimes 		return (ENOBUFS);
713df8bae1dSRodney W. Grimes 	}
714df8bae1dSRodney W. Grimes 	tp->t_state = TCPS_CLOSED;
715df8bae1dSRodney W. Grimes 	return (0);
716df8bae1dSRodney W. Grimes }
717df8bae1dSRodney W. Grimes 
718df8bae1dSRodney W. Grimes /*
719df8bae1dSRodney W. Grimes  * Initiate (or continue) disconnect.
720df8bae1dSRodney W. Grimes  * If embryonic state, just send reset (once).
721df8bae1dSRodney W. Grimes  * If in ``let data drain'' option and linger null, just drop.
722df8bae1dSRodney W. Grimes  * Otherwise (hard), mark socket disconnecting and drop
723df8bae1dSRodney W. Grimes  * current input data; switch states based on user close, and
724df8bae1dSRodney W. Grimes  * send segment to peer (with FIN).
725df8bae1dSRodney W. Grimes  */
7260312fbe9SPoul-Henning Kamp static struct tcpcb *
727df8bae1dSRodney W. Grimes tcp_disconnect(tp)
728df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
729df8bae1dSRodney W. Grimes {
730df8bae1dSRodney W. Grimes 	struct socket *so = tp->t_inpcb->inp_socket;
731df8bae1dSRodney W. Grimes 
732df8bae1dSRodney W. Grimes 	if (tp->t_state < TCPS_ESTABLISHED)
733df8bae1dSRodney W. Grimes 		tp = tcp_close(tp);
734df8bae1dSRodney W. Grimes 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
735df8bae1dSRodney W. Grimes 		tp = tcp_drop(tp, 0);
736df8bae1dSRodney W. Grimes 	else {
737df8bae1dSRodney W. Grimes 		soisdisconnecting(so);
738df8bae1dSRodney W. Grimes 		sbflush(&so->so_rcv);
739df8bae1dSRodney W. Grimes 		tp = tcp_usrclosed(tp);
740df8bae1dSRodney W. Grimes 		if (tp)
741df8bae1dSRodney W. Grimes 			(void) tcp_output(tp);
742df8bae1dSRodney W. Grimes 	}
743df8bae1dSRodney W. Grimes 	return (tp);
744df8bae1dSRodney W. Grimes }
745df8bae1dSRodney W. Grimes 
746df8bae1dSRodney W. Grimes /*
747df8bae1dSRodney W. Grimes  * User issued close, and wish to trail through shutdown states:
748df8bae1dSRodney W. Grimes  * if never received SYN, just forget it.  If got a SYN from peer,
749df8bae1dSRodney W. Grimes  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
750df8bae1dSRodney W. Grimes  * If already got a FIN from peer, then almost done; go to LAST_ACK
751df8bae1dSRodney W. Grimes  * state.  In all other cases, have already sent FIN to peer (e.g.
752df8bae1dSRodney W. Grimes  * after PRU_SHUTDOWN), and just have to play tedious game waiting
753df8bae1dSRodney W. Grimes  * for peer to send FIN or not respond to keep-alives, etc.
754df8bae1dSRodney W. Grimes  * We can let the user exit from the close as soon as the FIN is acked.
755df8bae1dSRodney W. Grimes  */
7560312fbe9SPoul-Henning Kamp static struct tcpcb *
757df8bae1dSRodney W. Grimes tcp_usrclosed(tp)
758df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
759df8bae1dSRodney W. Grimes {
760df8bae1dSRodney W. Grimes 
761df8bae1dSRodney W. Grimes 	switch (tp->t_state) {
762df8bae1dSRodney W. Grimes 
763df8bae1dSRodney W. Grimes 	case TCPS_CLOSED:
764df8bae1dSRodney W. Grimes 	case TCPS_LISTEN:
765df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_CLOSED;
766df8bae1dSRodney W. Grimes 		tp = tcp_close(tp);
767df8bae1dSRodney W. Grimes 		break;
768df8bae1dSRodney W. Grimes 
769a0292f23SGarrett Wollman 	case TCPS_SYN_SENT:
770df8bae1dSRodney W. Grimes 	case TCPS_SYN_RECEIVED:
771a0292f23SGarrett Wollman 		tp->t_flags |= TF_NEEDFIN;
772a0292f23SGarrett Wollman 		break;
773a0292f23SGarrett Wollman 
774df8bae1dSRodney W. Grimes 	case TCPS_ESTABLISHED:
775df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_FIN_WAIT_1;
776df8bae1dSRodney W. Grimes 		break;
777df8bae1dSRodney W. Grimes 
778df8bae1dSRodney W. Grimes 	case TCPS_CLOSE_WAIT:
779df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_LAST_ACK;
780df8bae1dSRodney W. Grimes 		break;
781df8bae1dSRodney W. Grimes 	}
782b6239c4aSAndras Olah 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
783df8bae1dSRodney W. Grimes 		soisdisconnected(tp->t_inpcb->inp_socket);
784b6239c4aSAndras Olah 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
785b6239c4aSAndras Olah 		if (tp->t_state == TCPS_FIN_WAIT_2)
786b6239c4aSAndras Olah 			tp->t_timer[TCPT_2MSL] = tcp_maxidle;
787b6239c4aSAndras Olah 	}
788df8bae1dSRodney W. Grimes 	return (tp);
789df8bae1dSRodney W. Grimes }
790a0292f23SGarrett Wollman 
791