xref: /freebsd/sys/netinet/tcp_usrreq.c (revision ef53690bb495367dfad069994f4e295769e9f162)
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
34ef53690bSGarrett Wollman  *	$Id$
35df8bae1dSRodney W. Grimes  */
36df8bae1dSRodney W. Grimes 
37df8bae1dSRodney W. Grimes #include <sys/param.h>
382ee45d7dSDavid Greenman #include <sys/queue.h>
39df8bae1dSRodney W. Grimes #include <sys/systm.h>
40c7a82f90SGarrett Wollman #include <sys/kernel.h>
4198163b98SPoul-Henning Kamp #include <sys/sysctl.h>
42df8bae1dSRodney W. Grimes #include <sys/malloc.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 #include <sys/errno.h>
48df8bae1dSRodney W. Grimes #include <sys/stat.h>
49df8bae1dSRodney W. Grimes 
50df8bae1dSRodney W. Grimes #include <net/if.h>
51df8bae1dSRodney W. Grimes #include <net/route.h>
52df8bae1dSRodney W. Grimes 
53df8bae1dSRodney W. Grimes #include <netinet/in.h>
54df8bae1dSRodney W. Grimes #include <netinet/in_systm.h>
55df8bae1dSRodney W. Grimes #include <netinet/ip.h>
56df8bae1dSRodney W. Grimes #include <netinet/in_pcb.h>
57b5e8ce9fSBruce Evans #include <netinet/in_var.h>
58df8bae1dSRodney W. Grimes #include <netinet/ip_var.h>
59df8bae1dSRodney W. Grimes #include <netinet/tcp.h>
60df8bae1dSRodney W. Grimes #include <netinet/tcp_fsm.h>
61df8bae1dSRodney W. Grimes #include <netinet/tcp_seq.h>
62df8bae1dSRodney W. Grimes #include <netinet/tcp_timer.h>
63df8bae1dSRodney W. Grimes #include <netinet/tcp_var.h>
64df8bae1dSRodney W. Grimes #include <netinet/tcpip.h>
65610ee2f9SDavid Greenman #ifdef TCPDEBUG
66df8bae1dSRodney W. Grimes #include <netinet/tcp_debug.h>
67610ee2f9SDavid Greenman #endif
68df8bae1dSRodney W. Grimes 
69df8bae1dSRodney W. Grimes /*
70df8bae1dSRodney W. Grimes  * TCP protocol interface to socket abstraction.
71df8bae1dSRodney W. Grimes  */
72117bcae7SGarrett Wollman extern	char *tcpstates[];	/* XXX ??? */
73df8bae1dSRodney W. Grimes 
740312fbe9SPoul-Henning Kamp static int	tcp_attach __P((struct socket *));
750312fbe9SPoul-Henning Kamp static int	tcp_connect __P((struct tcpcb *, struct mbuf *));
760312fbe9SPoul-Henning Kamp static struct tcpcb *
770312fbe9SPoul-Henning Kamp 		tcp_disconnect __P((struct tcpcb *));
780312fbe9SPoul-Henning Kamp static struct tcpcb *
790312fbe9SPoul-Henning Kamp 		tcp_usrclosed __P((struct tcpcb *));
802c37256eSGarrett Wollman 
812c37256eSGarrett Wollman #ifdef TCPDEBUG
822c37256eSGarrett Wollman #define	TCPDEBUG0	int ostate
832c37256eSGarrett Wollman #define	TCPDEBUG1()	ostate = tp ? tp->t_state : 0
84af7a2999SDavid Greenman #define	TCPDEBUG2(req)	if (tp && (so->so_options & SO_DEBUG)) \
852c37256eSGarrett Wollman 				tcp_trace(TA_USER, ostate, tp, 0, req)
862c37256eSGarrett Wollman #else
872c37256eSGarrett Wollman #define	TCPDEBUG0
882c37256eSGarrett Wollman #define	TCPDEBUG1()
892c37256eSGarrett Wollman #define	TCPDEBUG2(req)
902c37256eSGarrett Wollman #endif
912c37256eSGarrett Wollman 
922c37256eSGarrett Wollman /*
932c37256eSGarrett Wollman  * TCP attaches to socket via pru_attach(), reserving space,
942c37256eSGarrett Wollman  * and an internet control block.
952c37256eSGarrett Wollman  */
962c37256eSGarrett Wollman static int
972c37256eSGarrett Wollman tcp_usr_attach(struct socket *so, int proto)
982c37256eSGarrett Wollman {
992c37256eSGarrett Wollman 	int s = splnet();
1002c37256eSGarrett Wollman 	int error;
1012c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1022c37256eSGarrett Wollman 	struct tcpcb *tp = 0;
1032c37256eSGarrett Wollman 	TCPDEBUG0;
1042c37256eSGarrett Wollman 
1052c37256eSGarrett Wollman 	TCPDEBUG1();
1062c37256eSGarrett Wollman 	if (inp) {
1072c37256eSGarrett Wollman 		error = EISCONN;
1082c37256eSGarrett Wollman 		goto out;
1092c37256eSGarrett Wollman 	}
1102c37256eSGarrett Wollman 
1112c37256eSGarrett Wollman 	error = tcp_attach(so);
1122c37256eSGarrett Wollman 	if (error)
1132c37256eSGarrett Wollman 		goto out;
1142c37256eSGarrett Wollman 
1152c37256eSGarrett Wollman 	if ((so->so_options & SO_LINGER) && so->so_linger == 0)
1162c37256eSGarrett Wollman 		so->so_linger = TCP_LINGERTIME * hz;
1172c37256eSGarrett Wollman 	tp = sototcpcb(so);
1182c37256eSGarrett Wollman out:
1192c37256eSGarrett Wollman 	TCPDEBUG2(PRU_ATTACH);
1202c37256eSGarrett Wollman 	splx(s);
1212c37256eSGarrett Wollman 	return error;
1222c37256eSGarrett Wollman }
1232c37256eSGarrett Wollman 
1242c37256eSGarrett Wollman /*
1252c37256eSGarrett Wollman  * pru_detach() detaches the TCP protocol from the socket.
1262c37256eSGarrett Wollman  * If the protocol state is non-embryonic, then can't
1272c37256eSGarrett Wollman  * do this directly: have to initiate a pru_disconnect(),
1282c37256eSGarrett Wollman  * which may finish later; embryonic TCB's can just
1292c37256eSGarrett Wollman  * be discarded here.
1302c37256eSGarrett Wollman  */
1312c37256eSGarrett Wollman static int
1322c37256eSGarrett Wollman tcp_usr_detach(struct socket *so)
1332c37256eSGarrett Wollman {
1342c37256eSGarrett Wollman 	int s = splnet();
1352c37256eSGarrett Wollman 	int error = 0;
1362c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1372c37256eSGarrett Wollman 	struct tcpcb *tp;
1382c37256eSGarrett Wollman 	TCPDEBUG0;
1392c37256eSGarrett Wollman 
1402c37256eSGarrett Wollman 	if (inp == 0) {
1412c37256eSGarrett Wollman 		splx(s);
1422c37256eSGarrett Wollman 		return EINVAL;	/* XXX */
1432c37256eSGarrett Wollman 	}
1442c37256eSGarrett Wollman 	tp = intotcpcb(inp);
1452c37256eSGarrett Wollman 	TCPDEBUG1();
1462c37256eSGarrett Wollman 	if (tp->t_state > TCPS_LISTEN)
1472c37256eSGarrett Wollman 		tp = tcp_disconnect(tp);
1482c37256eSGarrett Wollman 	else
1492c37256eSGarrett Wollman 		tp = tcp_close(tp);
1502c37256eSGarrett Wollman 
1512c37256eSGarrett Wollman 	TCPDEBUG2(PRU_DETACH);
1522c37256eSGarrett Wollman 	splx(s);
1532c37256eSGarrett Wollman 	return error;
1542c37256eSGarrett Wollman }
1552c37256eSGarrett Wollman 
1562c37256eSGarrett Wollman #define	COMMON_START()	TCPDEBUG0; \
1572c37256eSGarrett Wollman 			do { \
1582c37256eSGarrett Wollman 				     if (inp == 0) { \
1592c37256eSGarrett Wollman 					     splx(s); \
1602c37256eSGarrett Wollman 					     return EINVAL; \
1612c37256eSGarrett Wollman 				     } \
1622c37256eSGarrett Wollman 				     tp = intotcpcb(inp); \
1632c37256eSGarrett Wollman 				     TCPDEBUG1(); \
1642c37256eSGarrett Wollman 		     } while(0)
1652c37256eSGarrett Wollman 
1662c37256eSGarrett Wollman #define COMMON_END(req)	out: TCPDEBUG2(req); splx(s); return error; goto out
1672c37256eSGarrett Wollman 
1682c37256eSGarrett Wollman 
1692c37256eSGarrett Wollman /*
1702c37256eSGarrett Wollman  * Give the socket an address.
1712c37256eSGarrett Wollman  */
1722c37256eSGarrett Wollman static int
1732c37256eSGarrett Wollman tcp_usr_bind(struct socket *so, struct mbuf *nam)
1742c37256eSGarrett Wollman {
1752c37256eSGarrett Wollman 	int s = splnet();
1762c37256eSGarrett Wollman 	int error = 0;
1772c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
1782c37256eSGarrett Wollman 	struct tcpcb *tp;
1792c37256eSGarrett Wollman 	struct sockaddr_in *sinp;
1802c37256eSGarrett Wollman 
1812c37256eSGarrett Wollman 	COMMON_START();
1822c37256eSGarrett Wollman 
1832c37256eSGarrett Wollman 	/*
1842c37256eSGarrett Wollman 	 * Must check for multicast addresses and disallow binding
1852c37256eSGarrett Wollman 	 * to them.
1862c37256eSGarrett Wollman 	 */
1872c37256eSGarrett Wollman 	sinp = mtod(nam, struct sockaddr_in *);
1882c37256eSGarrett Wollman 	if (sinp->sin_family == AF_INET &&
1892c37256eSGarrett Wollman 	    IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
1902c37256eSGarrett Wollman 		error = EAFNOSUPPORT;
1912c37256eSGarrett Wollman 		goto out;
1922c37256eSGarrett Wollman 	}
1932c37256eSGarrett Wollman 	error = in_pcbbind(inp, nam);
1942c37256eSGarrett Wollman 	if (error)
1952c37256eSGarrett Wollman 		goto out;
1962c37256eSGarrett Wollman 	COMMON_END(PRU_BIND);
1972c37256eSGarrett Wollman 
1982c37256eSGarrett Wollman }
1992c37256eSGarrett Wollman 
2002c37256eSGarrett Wollman /*
2012c37256eSGarrett Wollman  * Prepare to accept connections.
2022c37256eSGarrett Wollman  */
2032c37256eSGarrett Wollman static int
2042c37256eSGarrett Wollman tcp_usr_listen(struct socket *so)
2052c37256eSGarrett Wollman {
2062c37256eSGarrett Wollman 	int s = splnet();
2072c37256eSGarrett Wollman 	int error = 0;
2082c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2092c37256eSGarrett Wollman 	struct tcpcb *tp;
2102c37256eSGarrett Wollman 
2112c37256eSGarrett Wollman 	COMMON_START();
2122c37256eSGarrett Wollman 	if (inp->inp_lport == 0)
2132c37256eSGarrett Wollman 		error = in_pcbbind(inp, NULL);
2142c37256eSGarrett Wollman 	if (error == 0)
2152c37256eSGarrett Wollman 		tp->t_state = TCPS_LISTEN;
2162c37256eSGarrett Wollman 	COMMON_END(PRU_LISTEN);
2172c37256eSGarrett Wollman }
2182c37256eSGarrett Wollman 
2192c37256eSGarrett Wollman /*
2202c37256eSGarrett Wollman  * Initiate connection to peer.
2212c37256eSGarrett Wollman  * Create a template for use in transmissions on this connection.
2222c37256eSGarrett Wollman  * Enter SYN_SENT state, and mark socket as connecting.
2232c37256eSGarrett Wollman  * Start keep-alive timer, and seed output sequence space.
2242c37256eSGarrett Wollman  * Send initial segment on connection.
2252c37256eSGarrett Wollman  */
2262c37256eSGarrett Wollman static int
2272c37256eSGarrett Wollman tcp_usr_connect(struct socket *so, struct mbuf *nam)
2282c37256eSGarrett Wollman {
2292c37256eSGarrett Wollman 	int s = splnet();
2302c37256eSGarrett Wollman 	int error = 0;
2312c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2322c37256eSGarrett Wollman 	struct tcpcb *tp;
2332c37256eSGarrett Wollman 	struct sockaddr_in *sinp;
2342c37256eSGarrett Wollman 
2352c37256eSGarrett Wollman 	COMMON_START();
2362c37256eSGarrett Wollman 
2372c37256eSGarrett Wollman 	/*
2382c37256eSGarrett Wollman 	 * Must disallow TCP ``connections'' to multicast addresses.
2392c37256eSGarrett Wollman 	 */
2402c37256eSGarrett Wollman 	sinp = mtod(nam, struct sockaddr_in *);
2412c37256eSGarrett Wollman 	if (sinp->sin_family == AF_INET
2422c37256eSGarrett Wollman 	    && IN_MULTICAST(ntohl(sinp->sin_addr.s_addr))) {
2432c37256eSGarrett Wollman 		error = EAFNOSUPPORT;
2442c37256eSGarrett Wollman 		goto out;
2452c37256eSGarrett Wollman 	}
2462c37256eSGarrett Wollman 
2472c37256eSGarrett Wollman 	if ((error = tcp_connect(tp, nam)) != 0)
2482c37256eSGarrett Wollman 		goto out;
2492c37256eSGarrett Wollman 	error = tcp_output(tp);
2502c37256eSGarrett Wollman 	COMMON_END(PRU_CONNECT);
2512c37256eSGarrett Wollman }
2522c37256eSGarrett Wollman 
2532c37256eSGarrett Wollman /*
2542c37256eSGarrett Wollman  * Initiate disconnect from peer.
2552c37256eSGarrett Wollman  * If connection never passed embryonic stage, just drop;
2562c37256eSGarrett Wollman  * else if don't need to let data drain, then can just drop anyways,
2572c37256eSGarrett Wollman  * else have to begin TCP shutdown process: mark socket disconnecting,
2582c37256eSGarrett Wollman  * drain unread data, state switch to reflect user close, and
2592c37256eSGarrett Wollman  * send segment (e.g. FIN) to peer.  Socket will be really disconnected
2602c37256eSGarrett Wollman  * when peer sends FIN and acks ours.
2612c37256eSGarrett Wollman  *
2622c37256eSGarrett Wollman  * SHOULD IMPLEMENT LATER PRU_CONNECT VIA REALLOC TCPCB.
2632c37256eSGarrett Wollman  */
2642c37256eSGarrett Wollman static int
2652c37256eSGarrett Wollman tcp_usr_disconnect(struct socket *so)
2662c37256eSGarrett Wollman {
2672c37256eSGarrett Wollman 	int s = splnet();
2682c37256eSGarrett Wollman 	int error = 0;
2692c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2702c37256eSGarrett Wollman 	struct tcpcb *tp;
2712c37256eSGarrett Wollman 
2722c37256eSGarrett Wollman 	COMMON_START();
2732c37256eSGarrett Wollman 	tp = tcp_disconnect(tp);
2742c37256eSGarrett Wollman 	COMMON_END(PRU_DISCONNECT);
2752c37256eSGarrett Wollman }
2762c37256eSGarrett Wollman 
2772c37256eSGarrett Wollman /*
2782c37256eSGarrett Wollman  * Accept a connection.  Essentially all the work is
2792c37256eSGarrett Wollman  * done at higher levels; just return the address
2802c37256eSGarrett Wollman  * of the peer, storing through addr.
2812c37256eSGarrett Wollman  */
2822c37256eSGarrett Wollman static int
2832c37256eSGarrett Wollman tcp_usr_accept(struct socket *so, struct mbuf *nam)
2842c37256eSGarrett Wollman {
2852c37256eSGarrett Wollman 	int s = splnet();
2862c37256eSGarrett Wollman 	int error = 0;
2872c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
2882c37256eSGarrett Wollman 	struct tcpcb *tp;
2892c37256eSGarrett Wollman 
2902c37256eSGarrett Wollman 	COMMON_START();
291117bcae7SGarrett Wollman 	in_setpeeraddr(so, nam);
2922c37256eSGarrett Wollman 	COMMON_END(PRU_ACCEPT);
2932c37256eSGarrett Wollman }
2942c37256eSGarrett Wollman 
2952c37256eSGarrett Wollman /*
2962c37256eSGarrett Wollman  * Mark the connection as being incapable of further output.
2972c37256eSGarrett Wollman  */
2982c37256eSGarrett Wollman static int
2992c37256eSGarrett Wollman tcp_usr_shutdown(struct socket *so)
3002c37256eSGarrett Wollman {
3012c37256eSGarrett Wollman 	int s = splnet();
3022c37256eSGarrett Wollman 	int error = 0;
3032c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
3042c37256eSGarrett Wollman 	struct tcpcb *tp;
3052c37256eSGarrett Wollman 
3062c37256eSGarrett Wollman 	COMMON_START();
3072c37256eSGarrett Wollman 	socantsendmore(so);
3082c37256eSGarrett Wollman 	tp = tcp_usrclosed(tp);
3092c37256eSGarrett Wollman 	if (tp)
3102c37256eSGarrett Wollman 		error = tcp_output(tp);
3112c37256eSGarrett Wollman 	COMMON_END(PRU_SHUTDOWN);
3122c37256eSGarrett Wollman }
3132c37256eSGarrett Wollman 
3142c37256eSGarrett Wollman /*
3152c37256eSGarrett Wollman  * After a receive, possibly send window update to peer.
3162c37256eSGarrett Wollman  */
3172c37256eSGarrett Wollman static int
3182c37256eSGarrett Wollman tcp_usr_rcvd(struct socket *so, int flags)
3192c37256eSGarrett Wollman {
3202c37256eSGarrett Wollman 	int s = splnet();
3212c37256eSGarrett Wollman 	int error = 0;
3222c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
3232c37256eSGarrett Wollman 	struct tcpcb *tp;
3242c37256eSGarrett Wollman 
3252c37256eSGarrett Wollman 	COMMON_START();
3262c37256eSGarrett Wollman 	tcp_output(tp);
3272c37256eSGarrett Wollman 	COMMON_END(PRU_RCVD);
3282c37256eSGarrett Wollman }
3292c37256eSGarrett Wollman 
3302c37256eSGarrett Wollman /*
3312c37256eSGarrett Wollman  * Do a send by putting data in output queue and updating urgent
3322c37256eSGarrett Wollman  * marker if URG set.  Possibly send more data.
3332c37256eSGarrett Wollman  */
3342c37256eSGarrett Wollman static int
3352c37256eSGarrett Wollman tcp_usr_send(struct socket *so, int flags, struct mbuf *m, struct mbuf *nam,
3362c37256eSGarrett Wollman 	     struct mbuf *control)
3372c37256eSGarrett Wollman {
3382c37256eSGarrett Wollman 	int s = splnet();
3392c37256eSGarrett Wollman 	int error = 0;
3402c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
3412c37256eSGarrett Wollman 	struct tcpcb *tp;
3422c37256eSGarrett Wollman 
3432c37256eSGarrett Wollman 	COMMON_START();
3442c37256eSGarrett Wollman 	if (control && control->m_len) {
3452c37256eSGarrett Wollman 		m_freem(control); /* XXX shouldn't caller do this??? */
3462c37256eSGarrett Wollman 		if (m)
3472c37256eSGarrett Wollman 			m_freem(m);
3482c37256eSGarrett Wollman 		return EINVAL;
3492c37256eSGarrett Wollman 	}
3502c37256eSGarrett Wollman 
3512c37256eSGarrett Wollman 	if(!(flags & PRUS_OOB)) {
3522c37256eSGarrett Wollman 		sbappend(&so->so_snd, m);
3532c37256eSGarrett Wollman 		if (nam && tp->t_state < TCPS_SYN_SENT) {
3542c37256eSGarrett Wollman 			/*
3552c37256eSGarrett Wollman 			 * Do implied connect if not yet connected,
3562c37256eSGarrett Wollman 			 * initialize window to default value, and
3572c37256eSGarrett Wollman 			 * initialize maxseg/maxopd using peer's cached
3582c37256eSGarrett Wollman 			 * MSS.
3592c37256eSGarrett Wollman 			 */
3602c37256eSGarrett Wollman 			error = tcp_connect(tp, nam);
3612c37256eSGarrett Wollman 			if (error)
3622c37256eSGarrett Wollman 				goto out;
3632c37256eSGarrett Wollman 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
3642c37256eSGarrett Wollman 			tcp_mss(tp, -1);
3652c37256eSGarrett Wollman 		}
3662c37256eSGarrett Wollman 
3672c37256eSGarrett Wollman 		if (flags & PRUS_EOF) {
3682c37256eSGarrett Wollman 			/*
3692c37256eSGarrett Wollman 			 * Close the send side of the connection after
3702c37256eSGarrett Wollman 			 * the data is sent.
3712c37256eSGarrett Wollman 			 */
3722c37256eSGarrett Wollman 			socantsendmore(so);
3732c37256eSGarrett Wollman 			tp = tcp_usrclosed(tp);
3742c37256eSGarrett Wollman 		}
3752c37256eSGarrett Wollman 		if (tp != NULL)
3762c37256eSGarrett Wollman 			error = tcp_output(tp);
3772c37256eSGarrett Wollman 	} else {
3782c37256eSGarrett Wollman 		if (sbspace(&so->so_snd) < -512) {
3792c37256eSGarrett Wollman 			m_freem(m);
3802c37256eSGarrett Wollman 			error = ENOBUFS;
3812c37256eSGarrett Wollman 			goto out;
3822c37256eSGarrett Wollman 		}
3832c37256eSGarrett Wollman 		/*
3842c37256eSGarrett Wollman 		 * According to RFC961 (Assigned Protocols),
3852c37256eSGarrett Wollman 		 * the urgent pointer points to the last octet
3862c37256eSGarrett Wollman 		 * of urgent data.  We continue, however,
3872c37256eSGarrett Wollman 		 * to consider it to indicate the first octet
3882c37256eSGarrett Wollman 		 * of data past the urgent section.
3892c37256eSGarrett Wollman 		 * Otherwise, snd_up should be one lower.
3902c37256eSGarrett Wollman 		 */
3912c37256eSGarrett Wollman 		sbappend(&so->so_snd, m);
392ef53690bSGarrett Wollman 		if (nam && tp->t_state < TCPS_SYN_SENT) {
393ef53690bSGarrett Wollman 			/*
394ef53690bSGarrett Wollman 			 * Do implied connect if not yet connected,
395ef53690bSGarrett Wollman 			 * initialize window to default value, and
396ef53690bSGarrett Wollman 			 * initialize maxseg/maxopd using peer's cached
397ef53690bSGarrett Wollman 			 * MSS.
398ef53690bSGarrett Wollman 			 */
399ef53690bSGarrett Wollman 			error = tcp_connect(tp, nam);
400ef53690bSGarrett Wollman 			if (error)
401ef53690bSGarrett Wollman 				goto out;
402ef53690bSGarrett Wollman 			tp->snd_wnd = TTCP_CLIENT_SND_WND;
403ef53690bSGarrett Wollman 			tcp_mss(tp, -1);
404ef53690bSGarrett Wollman 		}
4052c37256eSGarrett Wollman 		tp->snd_up = tp->snd_una + so->so_snd.sb_cc;
4062c37256eSGarrett Wollman 		tp->t_force = 1;
4072c37256eSGarrett Wollman 		error = tcp_output(tp);
4082c37256eSGarrett Wollman 		tp->t_force = 0;
4092c37256eSGarrett Wollman 	}
4102c37256eSGarrett Wollman 	COMMON_END((flags & PRUS_OOB) ? PRU_SENDOOB :
4112c37256eSGarrett Wollman 		   ((flags & PRUS_EOF) ? PRU_SEND_EOF : PRU_SEND));
4122c37256eSGarrett Wollman }
4132c37256eSGarrett Wollman 
4142c37256eSGarrett Wollman /*
4152c37256eSGarrett Wollman  * Abort the TCP.
4162c37256eSGarrett Wollman  */
4172c37256eSGarrett Wollman static int
4182c37256eSGarrett Wollman tcp_usr_abort(struct socket *so)
4192c37256eSGarrett Wollman {
4202c37256eSGarrett Wollman 	int s = splnet();
4212c37256eSGarrett Wollman 	int error = 0;
4222c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
4232c37256eSGarrett Wollman 	struct tcpcb *tp;
4242c37256eSGarrett Wollman 
4252c37256eSGarrett Wollman 	COMMON_START();
4262c37256eSGarrett Wollman 	tp = tcp_drop(tp, ECONNABORTED);
4272c37256eSGarrett Wollman 	COMMON_END(PRU_ABORT);
4282c37256eSGarrett Wollman }
4292c37256eSGarrett Wollman 
4302c37256eSGarrett Wollman /*
4312c37256eSGarrett Wollman  * Receive out-of-band data.
4322c37256eSGarrett Wollman  */
4332c37256eSGarrett Wollman static int
4342c37256eSGarrett Wollman tcp_usr_rcvoob(struct socket *so, struct mbuf *m, int flags)
4352c37256eSGarrett Wollman {
4362c37256eSGarrett Wollman 	int s = splnet();
4372c37256eSGarrett Wollman 	int error = 0;
4382c37256eSGarrett Wollman 	struct inpcb *inp = sotoinpcb(so);
4392c37256eSGarrett Wollman 	struct tcpcb *tp;
4402c37256eSGarrett Wollman 
4412c37256eSGarrett Wollman 	COMMON_START();
4422c37256eSGarrett Wollman 	if ((so->so_oobmark == 0 &&
4432c37256eSGarrett Wollman 	     (so->so_state & SS_RCVATMARK) == 0) ||
4442c37256eSGarrett Wollman 	    so->so_options & SO_OOBINLINE ||
4452c37256eSGarrett Wollman 	    tp->t_oobflags & TCPOOB_HADDATA) {
4462c37256eSGarrett Wollman 		error = EINVAL;
4472c37256eSGarrett Wollman 		goto out;
4482c37256eSGarrett Wollman 	}
4492c37256eSGarrett Wollman 	if ((tp->t_oobflags & TCPOOB_HAVEDATA) == 0) {
4502c37256eSGarrett Wollman 		error = EWOULDBLOCK;
4512c37256eSGarrett Wollman 		goto out;
4522c37256eSGarrett Wollman 	}
4532c37256eSGarrett Wollman 	m->m_len = 1;
4542c37256eSGarrett Wollman 	*mtod(m, caddr_t) = tp->t_iobc;
4552c37256eSGarrett Wollman 	if ((flags & MSG_PEEK) == 0)
4562c37256eSGarrett Wollman 		tp->t_oobflags ^= (TCPOOB_HAVEDATA | TCPOOB_HADDATA);
4572c37256eSGarrett Wollman 	COMMON_END(PRU_RCVOOB);
4582c37256eSGarrett Wollman }
4592c37256eSGarrett Wollman 
4602c37256eSGarrett Wollman /* xxx - should be const */
4612c37256eSGarrett Wollman struct pr_usrreqs tcp_usrreqs = {
4622c37256eSGarrett Wollman 	tcp_usr_abort, tcp_usr_accept, tcp_usr_attach, tcp_usr_bind,
463117bcae7SGarrett Wollman 	tcp_usr_connect, pru_connect2_notsupp, in_control, tcp_usr_detach,
464117bcae7SGarrett Wollman 	tcp_usr_disconnect, tcp_usr_listen, in_setpeeraddr, tcp_usr_rcvd,
465117bcae7SGarrett Wollman 	tcp_usr_rcvoob, tcp_usr_send, pru_sense_null, tcp_usr_shutdown,
466117bcae7SGarrett Wollman 	in_setsockaddr
4672c37256eSGarrett Wollman };
468df8bae1dSRodney W. Grimes 
469a0292f23SGarrett Wollman /*
470a0292f23SGarrett Wollman  * Common subroutine to open a TCP connection to remote host specified
471a0292f23SGarrett Wollman  * by struct sockaddr_in in mbuf *nam.  Call in_pcbbind to assign a local
472a0292f23SGarrett Wollman  * port number if needed.  Call in_pcbladdr to do the routing and to choose
473a0292f23SGarrett Wollman  * a local host address (interface).  If there is an existing incarnation
474a0292f23SGarrett Wollman  * of the same connection in TIME-WAIT state and if the remote host was
475a0292f23SGarrett Wollman  * sending CC options and if the connection duration was < MSL, then
476a0292f23SGarrett Wollman  * truncate the previous TIME-WAIT state and proceed.
477a0292f23SGarrett Wollman  * Initialize connection parameters and enter SYN-SENT state.
478a0292f23SGarrett Wollman  */
4790312fbe9SPoul-Henning Kamp static int
480a0292f23SGarrett Wollman tcp_connect(tp, nam)
481a0292f23SGarrett Wollman 	register struct tcpcb *tp;
482a0292f23SGarrett Wollman 	struct mbuf *nam;
483a0292f23SGarrett Wollman {
484a0292f23SGarrett Wollman 	struct inpcb *inp = tp->t_inpcb, *oinp;
485a0292f23SGarrett Wollman 	struct socket *so = inp->inp_socket;
486a0292f23SGarrett Wollman 	struct tcpcb *otp;
487a0292f23SGarrett Wollman 	struct sockaddr_in *sin = mtod(nam, struct sockaddr_in *);
488a0292f23SGarrett Wollman 	struct sockaddr_in *ifaddr;
489a0292f23SGarrett Wollman 	int error;
490a45d2726SAndras Olah 	struct rmxp_tao *taop;
491a45d2726SAndras Olah 	struct rmxp_tao tao_noncached;
492a0292f23SGarrett Wollman 
493a0292f23SGarrett Wollman 	if (inp->inp_lport == 0) {
494a0292f23SGarrett Wollman 		error = in_pcbbind(inp, NULL);
495a0292f23SGarrett Wollman 		if (error)
496a0292f23SGarrett Wollman 			return error;
497a0292f23SGarrett Wollman 	}
498a0292f23SGarrett Wollman 
499a0292f23SGarrett Wollman 	/*
500a0292f23SGarrett Wollman 	 * Cannot simply call in_pcbconnect, because there might be an
501a0292f23SGarrett Wollman 	 * earlier incarnation of this same connection still in
502a0292f23SGarrett Wollman 	 * TIME_WAIT state, creating an ADDRINUSE error.
503a0292f23SGarrett Wollman 	 */
504a0292f23SGarrett Wollman 	error = in_pcbladdr(inp, nam, &ifaddr);
505d3628763SRodney W. Grimes 	if (error)
506d3628763SRodney W. Grimes 		return error;
5076d6a026bSDavid Greenman 	oinp = in_pcblookuphash(inp->inp_pcbinfo,
508a0292f23SGarrett Wollman 	    sin->sin_addr, sin->sin_port,
509a0292f23SGarrett Wollman 	    inp->inp_laddr.s_addr != INADDR_ANY ? inp->inp_laddr
510a0292f23SGarrett Wollman 						: ifaddr->sin_addr,
511a0292f23SGarrett Wollman 	    inp->inp_lport,  0);
512a0292f23SGarrett Wollman 	if (oinp) {
513a0292f23SGarrett Wollman 		if (oinp != inp && (otp = intotcpcb(oinp)) != NULL &&
514a0292f23SGarrett Wollman 		otp->t_state == TCPS_TIME_WAIT &&
515a0292f23SGarrett Wollman 		    otp->t_duration < TCPTV_MSL &&
516a0292f23SGarrett Wollman 		    (otp->t_flags & TF_RCVD_CC))
517a0292f23SGarrett Wollman 			otp = tcp_close(otp);
518a0292f23SGarrett Wollman 		else
519a0292f23SGarrett Wollman 			return EADDRINUSE;
520a0292f23SGarrett Wollman 	}
521a0292f23SGarrett Wollman 	if (inp->inp_laddr.s_addr == INADDR_ANY)
522a0292f23SGarrett Wollman 		inp->inp_laddr = ifaddr->sin_addr;
523a0292f23SGarrett Wollman 	inp->inp_faddr = sin->sin_addr;
524a0292f23SGarrett Wollman 	inp->inp_fport = sin->sin_port;
52515bd2b43SDavid Greenman 	in_pcbrehash(inp);
526a0292f23SGarrett Wollman 
527a0292f23SGarrett Wollman 	tp->t_template = tcp_template(tp);
528a0292f23SGarrett Wollman 	if (tp->t_template == 0) {
529a0292f23SGarrett Wollman 		in_pcbdisconnect(inp);
530a0292f23SGarrett Wollman 		return ENOBUFS;
531a0292f23SGarrett Wollman 	}
532a0292f23SGarrett Wollman 
533a0292f23SGarrett Wollman 	/* Compute window scaling to request.  */
534a0292f23SGarrett Wollman 	while (tp->request_r_scale < TCP_MAX_WINSHIFT &&
535a0292f23SGarrett Wollman 	    (TCP_MAXWIN << tp->request_r_scale) < so->so_rcv.sb_hiwat)
536a0292f23SGarrett Wollman 		tp->request_r_scale++;
537a0292f23SGarrett Wollman 
538a0292f23SGarrett Wollman 	soisconnecting(so);
539a0292f23SGarrett Wollman 	tcpstat.tcps_connattempt++;
540a0292f23SGarrett Wollman 	tp->t_state = TCPS_SYN_SENT;
5417b40aa32SPaul Traina 	tp->t_timer[TCPT_KEEP] = tcp_keepinit;
542a0292f23SGarrett Wollman 	tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
543a0292f23SGarrett Wollman 	tcp_sendseqinit(tp);
544a45d2726SAndras Olah 
545a45d2726SAndras Olah 	/*
546a45d2726SAndras Olah 	 * Generate a CC value for this connection and
547a45d2726SAndras Olah 	 * check whether CC or CCnew should be used.
548a45d2726SAndras Olah 	 */
549a45d2726SAndras Olah 	if ((taop = tcp_gettaocache(tp->t_inpcb)) == NULL) {
550a45d2726SAndras Olah 		taop = &tao_noncached;
551a45d2726SAndras Olah 		bzero(taop, sizeof(*taop));
552a45d2726SAndras Olah 	}
553a45d2726SAndras Olah 
554a0292f23SGarrett Wollman 	tp->cc_send = CC_INC(tcp_ccgen);
555a45d2726SAndras Olah 	if (taop->tao_ccsent != 0 &&
556a45d2726SAndras Olah 	    CC_GEQ(tp->cc_send, taop->tao_ccsent)) {
557a45d2726SAndras Olah 		taop->tao_ccsent = tp->cc_send;
558a45d2726SAndras Olah 	} else {
559a45d2726SAndras Olah 		taop->tao_ccsent = 0;
560a45d2726SAndras Olah 		tp->t_flags |= TF_SENDCCNEW;
561a45d2726SAndras Olah 	}
562a0292f23SGarrett Wollman 
563a0292f23SGarrett Wollman 	return 0;
564a0292f23SGarrett Wollman }
565a0292f23SGarrett Wollman 
566df8bae1dSRodney W. Grimes int
567df8bae1dSRodney W. Grimes tcp_ctloutput(op, so, level, optname, mp)
568df8bae1dSRodney W. Grimes 	int op;
569df8bae1dSRodney W. Grimes 	struct socket *so;
570df8bae1dSRodney W. Grimes 	int level, optname;
571df8bae1dSRodney W. Grimes 	struct mbuf **mp;
572df8bae1dSRodney W. Grimes {
573df8bae1dSRodney W. Grimes 	int error = 0, s;
574df8bae1dSRodney W. Grimes 	struct inpcb *inp;
575df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
576df8bae1dSRodney W. Grimes 	register struct mbuf *m;
577df8bae1dSRodney W. Grimes 	register int i;
578df8bae1dSRodney W. Grimes 
579df8bae1dSRodney W. Grimes 	s = splnet();
580df8bae1dSRodney W. Grimes 	inp = sotoinpcb(so);
581df8bae1dSRodney W. Grimes 	if (inp == NULL) {
582df8bae1dSRodney W. Grimes 		splx(s);
583df8bae1dSRodney W. Grimes 		if (op == PRCO_SETOPT && *mp)
584df8bae1dSRodney W. Grimes 			(void) m_free(*mp);
585df8bae1dSRodney W. Grimes 		return (ECONNRESET);
586df8bae1dSRodney W. Grimes 	}
587df8bae1dSRodney W. Grimes 	if (level != IPPROTO_TCP) {
588df8bae1dSRodney W. Grimes 		error = ip_ctloutput(op, so, level, optname, mp);
589df8bae1dSRodney W. Grimes 		splx(s);
590df8bae1dSRodney W. Grimes 		return (error);
591df8bae1dSRodney W. Grimes 	}
592df8bae1dSRodney W. Grimes 	tp = intotcpcb(inp);
593df8bae1dSRodney W. Grimes 
594df8bae1dSRodney W. Grimes 	switch (op) {
595df8bae1dSRodney W. Grimes 
596df8bae1dSRodney W. Grimes 	case PRCO_SETOPT:
597df8bae1dSRodney W. Grimes 		m = *mp;
598df8bae1dSRodney W. Grimes 		switch (optname) {
599df8bae1dSRodney W. Grimes 
600df8bae1dSRodney W. Grimes 		case TCP_NODELAY:
601df8bae1dSRodney W. Grimes 			if (m == NULL || m->m_len < sizeof (int))
602df8bae1dSRodney W. Grimes 				error = EINVAL;
603df8bae1dSRodney W. Grimes 			else if (*mtod(m, int *))
604df8bae1dSRodney W. Grimes 				tp->t_flags |= TF_NODELAY;
605df8bae1dSRodney W. Grimes 			else
606df8bae1dSRodney W. Grimes 				tp->t_flags &= ~TF_NODELAY;
607df8bae1dSRodney W. Grimes 			break;
608df8bae1dSRodney W. Grimes 
609df8bae1dSRodney W. Grimes 		case TCP_MAXSEG:
610df8bae1dSRodney W. Grimes 			if (m && (i = *mtod(m, int *)) > 0 && i <= tp->t_maxseg)
611df8bae1dSRodney W. Grimes 				tp->t_maxseg = i;
612df8bae1dSRodney W. Grimes 			else
613df8bae1dSRodney W. Grimes 				error = EINVAL;
614df8bae1dSRodney W. Grimes 			break;
615df8bae1dSRodney W. Grimes 
616a0292f23SGarrett Wollman 		case TCP_NOOPT:
617a0292f23SGarrett Wollman 			if (m == NULL || m->m_len < sizeof (int))
618a0292f23SGarrett Wollman 				error = EINVAL;
619a0292f23SGarrett Wollman 			else if (*mtod(m, int *))
620a0292f23SGarrett Wollman 				tp->t_flags |= TF_NOOPT;
621a0292f23SGarrett Wollman 			else
622a0292f23SGarrett Wollman 				tp->t_flags &= ~TF_NOOPT;
623a0292f23SGarrett Wollman 			break;
624a0292f23SGarrett Wollman 
625a0292f23SGarrett Wollman 		case TCP_NOPUSH:
626a0292f23SGarrett Wollman 			if (m == NULL || m->m_len < sizeof (int))
627a0292f23SGarrett Wollman 				error = EINVAL;
628a0292f23SGarrett Wollman 			else if (*mtod(m, int *))
629a0292f23SGarrett Wollman 				tp->t_flags |= TF_NOPUSH;
630a0292f23SGarrett Wollman 			else
631a0292f23SGarrett Wollman 				tp->t_flags &= ~TF_NOPUSH;
632a0292f23SGarrett Wollman 			break;
633a0292f23SGarrett Wollman 
634df8bae1dSRodney W. Grimes 		default:
635df8bae1dSRodney W. Grimes 			error = ENOPROTOOPT;
636df8bae1dSRodney W. Grimes 			break;
637df8bae1dSRodney W. Grimes 		}
638df8bae1dSRodney W. Grimes 		if (m)
639df8bae1dSRodney W. Grimes 			(void) m_free(m);
640df8bae1dSRodney W. Grimes 		break;
641df8bae1dSRodney W. Grimes 
642df8bae1dSRodney W. Grimes 	case PRCO_GETOPT:
643df8bae1dSRodney W. Grimes 		*mp = m = m_get(M_WAIT, MT_SOOPTS);
644df8bae1dSRodney W. Grimes 		m->m_len = sizeof(int);
645df8bae1dSRodney W. Grimes 
646df8bae1dSRodney W. Grimes 		switch (optname) {
647df8bae1dSRodney W. Grimes 		case TCP_NODELAY:
648df8bae1dSRodney W. Grimes 			*mtod(m, int *) = tp->t_flags & TF_NODELAY;
649df8bae1dSRodney W. Grimes 			break;
650df8bae1dSRodney W. Grimes 		case TCP_MAXSEG:
651df8bae1dSRodney W. Grimes 			*mtod(m, int *) = tp->t_maxseg;
652df8bae1dSRodney W. Grimes 			break;
653a0292f23SGarrett Wollman 		case TCP_NOOPT:
654a0292f23SGarrett Wollman 			*mtod(m, int *) = tp->t_flags & TF_NOOPT;
655a0292f23SGarrett Wollman 			break;
656a0292f23SGarrett Wollman 		case TCP_NOPUSH:
657a0292f23SGarrett Wollman 			*mtod(m, int *) = tp->t_flags & TF_NOPUSH;
658a0292f23SGarrett Wollman 			break;
659df8bae1dSRodney W. Grimes 		default:
660df8bae1dSRodney W. Grimes 			error = ENOPROTOOPT;
661df8bae1dSRodney W. Grimes 			break;
662df8bae1dSRodney W. Grimes 		}
663df8bae1dSRodney W. Grimes 		break;
664df8bae1dSRodney W. Grimes 	}
665df8bae1dSRodney W. Grimes 	splx(s);
666df8bae1dSRodney W. Grimes 	return (error);
667df8bae1dSRodney W. Grimes }
668df8bae1dSRodney W. Grimes 
66926e30fbbSDavid Greenman /*
67026e30fbbSDavid Greenman  * tcp_sendspace and tcp_recvspace are the default send and receive window
67126e30fbbSDavid Greenman  * sizes, respectively.  These are obsolescent (this information should
67226e30fbbSDavid Greenman  * be set by the route).
67326e30fbbSDavid Greenman  */
67426e30fbbSDavid Greenman u_long	tcp_sendspace = 1024*16;
67598163b98SPoul-Henning Kamp SYSCTL_INT(_net_inet_tcp, TCPCTL_SENDSPACE, sendspace,
67698163b98SPoul-Henning Kamp 	CTLFLAG_RW, &tcp_sendspace , 0, "");
67726e30fbbSDavid Greenman u_long	tcp_recvspace = 1024*16;
67898163b98SPoul-Henning Kamp SYSCTL_INT(_net_inet_tcp, TCPCTL_RECVSPACE, recvspace,
67998163b98SPoul-Henning Kamp 	CTLFLAG_RW, &tcp_recvspace , 0, "");
680df8bae1dSRodney W. Grimes 
681df8bae1dSRodney W. Grimes /*
682df8bae1dSRodney W. Grimes  * Attach TCP protocol to socket, allocating
683df8bae1dSRodney W. Grimes  * internet protocol control block, tcp control block,
684df8bae1dSRodney W. Grimes  * bufer space, and entering LISTEN state if to accept connections.
685df8bae1dSRodney W. Grimes  */
6860312fbe9SPoul-Henning Kamp static int
687df8bae1dSRodney W. Grimes tcp_attach(so)
688df8bae1dSRodney W. Grimes 	struct socket *so;
689df8bae1dSRodney W. Grimes {
690df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
691df8bae1dSRodney W. Grimes 	struct inpcb *inp;
692df8bae1dSRodney W. Grimes 	int error;
693df8bae1dSRodney W. Grimes 
694df8bae1dSRodney W. Grimes 	if (so->so_snd.sb_hiwat == 0 || so->so_rcv.sb_hiwat == 0) {
695df8bae1dSRodney W. Grimes 		error = soreserve(so, tcp_sendspace, tcp_recvspace);
696df8bae1dSRodney W. Grimes 		if (error)
697df8bae1dSRodney W. Grimes 			return (error);
698df8bae1dSRodney W. Grimes 	}
69915bd2b43SDavid Greenman 	error = in_pcballoc(so, &tcbinfo);
700df8bae1dSRodney W. Grimes 	if (error)
701df8bae1dSRodney W. Grimes 		return (error);
702df8bae1dSRodney W. Grimes 	inp = sotoinpcb(so);
703df8bae1dSRodney W. Grimes 	tp = tcp_newtcpcb(inp);
704df8bae1dSRodney W. Grimes 	if (tp == 0) {
705df8bae1dSRodney W. Grimes 		int nofd = so->so_state & SS_NOFDREF;	/* XXX */
706df8bae1dSRodney W. Grimes 
707df8bae1dSRodney W. Grimes 		so->so_state &= ~SS_NOFDREF;	/* don't free the socket yet */
708df8bae1dSRodney W. Grimes 		in_pcbdetach(inp);
709df8bae1dSRodney W. Grimes 		so->so_state |= nofd;
710df8bae1dSRodney W. Grimes 		return (ENOBUFS);
711df8bae1dSRodney W. Grimes 	}
712df8bae1dSRodney W. Grimes 	tp->t_state = TCPS_CLOSED;
713df8bae1dSRodney W. Grimes 	return (0);
714df8bae1dSRodney W. Grimes }
715df8bae1dSRodney W. Grimes 
716df8bae1dSRodney W. Grimes /*
717df8bae1dSRodney W. Grimes  * Initiate (or continue) disconnect.
718df8bae1dSRodney W. Grimes  * If embryonic state, just send reset (once).
719df8bae1dSRodney W. Grimes  * If in ``let data drain'' option and linger null, just drop.
720df8bae1dSRodney W. Grimes  * Otherwise (hard), mark socket disconnecting and drop
721df8bae1dSRodney W. Grimes  * current input data; switch states based on user close, and
722df8bae1dSRodney W. Grimes  * send segment to peer (with FIN).
723df8bae1dSRodney W. Grimes  */
7240312fbe9SPoul-Henning Kamp static struct tcpcb *
725df8bae1dSRodney W. Grimes tcp_disconnect(tp)
726df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
727df8bae1dSRodney W. Grimes {
728df8bae1dSRodney W. Grimes 	struct socket *so = tp->t_inpcb->inp_socket;
729df8bae1dSRodney W. Grimes 
730df8bae1dSRodney W. Grimes 	if (tp->t_state < TCPS_ESTABLISHED)
731df8bae1dSRodney W. Grimes 		tp = tcp_close(tp);
732df8bae1dSRodney W. Grimes 	else if ((so->so_options & SO_LINGER) && so->so_linger == 0)
733df8bae1dSRodney W. Grimes 		tp = tcp_drop(tp, 0);
734df8bae1dSRodney W. Grimes 	else {
735df8bae1dSRodney W. Grimes 		soisdisconnecting(so);
736df8bae1dSRodney W. Grimes 		sbflush(&so->so_rcv);
737df8bae1dSRodney W. Grimes 		tp = tcp_usrclosed(tp);
738df8bae1dSRodney W. Grimes 		if (tp)
739df8bae1dSRodney W. Grimes 			(void) tcp_output(tp);
740df8bae1dSRodney W. Grimes 	}
741df8bae1dSRodney W. Grimes 	return (tp);
742df8bae1dSRodney W. Grimes }
743df8bae1dSRodney W. Grimes 
744df8bae1dSRodney W. Grimes /*
745df8bae1dSRodney W. Grimes  * User issued close, and wish to trail through shutdown states:
746df8bae1dSRodney W. Grimes  * if never received SYN, just forget it.  If got a SYN from peer,
747df8bae1dSRodney W. Grimes  * but haven't sent FIN, then go to FIN_WAIT_1 state to send peer a FIN.
748df8bae1dSRodney W. Grimes  * If already got a FIN from peer, then almost done; go to LAST_ACK
749df8bae1dSRodney W. Grimes  * state.  In all other cases, have already sent FIN to peer (e.g.
750df8bae1dSRodney W. Grimes  * after PRU_SHUTDOWN), and just have to play tedious game waiting
751df8bae1dSRodney W. Grimes  * for peer to send FIN or not respond to keep-alives, etc.
752df8bae1dSRodney W. Grimes  * We can let the user exit from the close as soon as the FIN is acked.
753df8bae1dSRodney W. Grimes  */
7540312fbe9SPoul-Henning Kamp static struct tcpcb *
755df8bae1dSRodney W. Grimes tcp_usrclosed(tp)
756df8bae1dSRodney W. Grimes 	register struct tcpcb *tp;
757df8bae1dSRodney W. Grimes {
758df8bae1dSRodney W. Grimes 
759df8bae1dSRodney W. Grimes 	switch (tp->t_state) {
760df8bae1dSRodney W. Grimes 
761df8bae1dSRodney W. Grimes 	case TCPS_CLOSED:
762df8bae1dSRodney W. Grimes 	case TCPS_LISTEN:
763df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_CLOSED;
764df8bae1dSRodney W. Grimes 		tp = tcp_close(tp);
765df8bae1dSRodney W. Grimes 		break;
766df8bae1dSRodney W. Grimes 
767a0292f23SGarrett Wollman 	case TCPS_SYN_SENT:
768df8bae1dSRodney W. Grimes 	case TCPS_SYN_RECEIVED:
769a0292f23SGarrett Wollman 		tp->t_flags |= TF_NEEDFIN;
770a0292f23SGarrett Wollman 		break;
771a0292f23SGarrett Wollman 
772df8bae1dSRodney W. Grimes 	case TCPS_ESTABLISHED:
773df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_FIN_WAIT_1;
774df8bae1dSRodney W. Grimes 		break;
775df8bae1dSRodney W. Grimes 
776df8bae1dSRodney W. Grimes 	case TCPS_CLOSE_WAIT:
777df8bae1dSRodney W. Grimes 		tp->t_state = TCPS_LAST_ACK;
778df8bae1dSRodney W. Grimes 		break;
779df8bae1dSRodney W. Grimes 	}
780b6239c4aSAndras Olah 	if (tp && tp->t_state >= TCPS_FIN_WAIT_2) {
781df8bae1dSRodney W. Grimes 		soisdisconnected(tp->t_inpcb->inp_socket);
782b6239c4aSAndras Olah 		/* To prevent the connection hanging in FIN_WAIT_2 forever. */
783b6239c4aSAndras Olah 		if (tp->t_state == TCPS_FIN_WAIT_2)
784b6239c4aSAndras Olah 			tp->t_timer[TCPT_2MSL] = tcp_maxidle;
785b6239c4aSAndras Olah 	}
786df8bae1dSRodney W. Grimes 	return (tp);
787df8bae1dSRodney W. Grimes }
788a0292f23SGarrett Wollman 
789