xref: /csrg-svn/sys/netinet/tcp_input.c (revision 17062)
1*17062Sbloom /*	tcp_input.c	6.4	84/08/29	*/
24601Swnj 
3*17062Sbloom #include "param.h"
4*17062Sbloom #include "systm.h"
5*17062Sbloom #include "mbuf.h"
6*17062Sbloom #include "protosw.h"
7*17062Sbloom #include "socket.h"
8*17062Sbloom #include "socketvar.h"
9*17062Sbloom #include "errno.h"
1010894Ssam 
1110894Ssam #include "../net/if.h"
1210894Ssam #include "../net/route.h"
1310894Ssam 
14*17062Sbloom #include "in.h"
15*17062Sbloom #include "in_pcb.h"
16*17062Sbloom #include "in_systm.h"
17*17062Sbloom #include "ip.h"
18*17062Sbloom #include "ip_var.h"
19*17062Sbloom #include "tcp.h"
20*17062Sbloom #include "tcp_fsm.h"
21*17062Sbloom #include "tcp_seq.h"
22*17062Sbloom #include "tcp_timer.h"
23*17062Sbloom #include "tcp_var.h"
24*17062Sbloom #include "tcpip.h"
25*17062Sbloom #include "tcp_debug.h"
264601Swnj 
275300Sroot int	tcpprintfs = 0;
284679Swnj int	tcpcksum = 1;
295267Sroot struct	tcpiphdr tcp_saveti;
305440Swnj extern	tcpnodelack;
314601Swnj 
325267Sroot struct	tcpcb *tcp_newtcpcb();
335065Swnj /*
345065Swnj  * TCP input routine, follows pages 65-76 of the
355065Swnj  * protocol specification dated September, 1981 very closely.
365065Swnj  */
374924Swnj tcp_input(m0)
384924Swnj 	struct mbuf *m0;
394601Swnj {
404924Swnj 	register struct tcpiphdr *ti;
414924Swnj 	struct inpcb *inp;
424924Swnj 	register struct mbuf *m;
435440Swnj 	struct mbuf *om = 0;
444924Swnj 	int len, tlen, off;
455391Swnj 	register struct tcpcb *tp = 0;
464924Swnj 	register int tiflags;
474803Swnj 	struct socket *so;
485109Swnj 	int todrop, acked;
495267Sroot 	short ostate;
506028Sroot 	struct in_addr laddr;
5110769Ssam 	int dropsocket = 0;
524924Swnj 
534924Swnj 	/*
545244Sroot 	 * Get IP and TCP header together in first mbuf.
555244Sroot 	 * Note: IP leaves IP header in first mbuf.
564924Swnj 	 */
574924Swnj 	m = m0;
585020Sroot 	ti = mtod(m, struct tcpiphdr *);
595244Sroot 	if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2))
605208Swnj 		ip_stripoptions((struct ip *)ti, (struct mbuf *)0);
615307Sroot 	if (m->m_off > MMAXOFF || m->m_len < sizeof (struct tcpiphdr)) {
625307Sroot 		if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) {
635085Swnj 			tcpstat.tcps_hdrops++;
645307Sroot 			return;
655085Swnj 		}
665085Swnj 		ti = mtod(m, struct tcpiphdr *);
675085Swnj 	}
684601Swnj 
694601Swnj 	/*
705244Sroot 	 * Checksum extended TCP header and data.
714601Swnj 	 */
724924Swnj 	tlen = ((struct ip *)ti)->ip_len;
734924Swnj 	len = sizeof (struct ip) + tlen;
744679Swnj 	if (tcpcksum) {
754924Swnj 		ti->ti_next = ti->ti_prev = 0;
764924Swnj 		ti->ti_x1 = 0;
775223Swnj 		ti->ti_len = (u_short)tlen;
786161Ssam 		ti->ti_len = htons((u_short)ti->ti_len);
795231Swnj 		if (ti->ti_sum = in_cksum(m, len)) {
8011830Ssam 			if (tcpprintfs)
8111830Ssam 				printf("tcp sum: src %x\n", ti->ti_src);
824924Swnj 			tcpstat.tcps_badsum++;
835085Swnj 			goto drop;
844601Swnj 		}
854601Swnj 	}
864601Swnj 
874601Swnj 	/*
885244Sroot 	 * Check that TCP offset makes sense,
895440Swnj 	 * pull out TCP options and adjust length.
904601Swnj 	 */
914924Swnj 	off = ti->ti_off << 2;
925231Swnj 	if (off < sizeof (struct tcphdr) || off > tlen) {
9311830Ssam 		if (tcpprintfs)
9411830Ssam 			printf("tcp off: src %x off %d\n", ti->ti_src, off);
954924Swnj 		tcpstat.tcps_badoff++;
965085Swnj 		goto drop;
974924Swnj 	}
986211Swnj 	tlen -= off;
996211Swnj 	ti->ti_len = tlen;
1005440Swnj 	if (off > sizeof (struct tcphdr)) {
1015440Swnj 		if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
1025440Swnj 			tcpstat.tcps_hdrops++;
10311730Ssam 			return;
1045440Swnj 		}
1055440Swnj 		ti = mtod(m, struct tcpiphdr *);
1069642Ssam 		om = m_get(M_DONTWAIT, MT_DATA);
1075440Swnj 		if (om == 0)
1085440Swnj 			goto drop;
1095440Swnj 		om->m_len = off - sizeof (struct tcphdr);
1105440Swnj 		{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
1116161Ssam 		  bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len);
1125440Swnj 		  m->m_len -= om->m_len;
1136161Ssam 		  bcopy(op+om->m_len, op,
1146161Ssam 		   (unsigned)(m->m_len-sizeof (struct tcpiphdr)));
1155440Swnj 		}
1165440Swnj 	}
1175065Swnj 	tiflags = ti->ti_flags;
1184924Swnj 
1196093Sroot 	/*
1206211Swnj 	 * Drop TCP and IP headers.
1216093Sroot 	 */
1226093Sroot 	off += sizeof (struct ip);
1236093Sroot 	m->m_off += off;
1246093Sroot 	m->m_len -= off;
1256093Sroot 
1264924Swnj 	/*
1275244Sroot 	 * Convert TCP protocol specific fields to host format.
1285085Swnj 	 */
1295085Swnj 	ti->ti_seq = ntohl(ti->ti_seq);
1305085Swnj 	ti->ti_ack = ntohl(ti->ti_ack);
1315085Swnj 	ti->ti_win = ntohs(ti->ti_win);
1325085Swnj 	ti->ti_urp = ntohs(ti->ti_urp);
1335085Swnj 
1345085Swnj 	/*
1358271Sroot 	 * Locate pcb for segment.
1364924Swnj 	 */
1375065Swnj 	inp = in_pcblookup
1386028Sroot 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport,
1396028Sroot 		INPLOOKUP_WILDCARD);
1405065Swnj 
1415065Swnj 	/*
1425065Swnj 	 * If the state is CLOSED (i.e., TCB does not exist) then
1435244Sroot 	 * all data in the incoming segment is discarded.
1445065Swnj 	 */
1455300Sroot 	if (inp == 0)
1465085Swnj 		goto dropwithreset;
1475065Swnj 	tp = intotcpcb(inp);
1485300Sroot 	if (tp == 0)
1495085Swnj 		goto dropwithreset;
1505109Swnj 	so = inp->inp_socket;
1515267Sroot 	if (so->so_options & SO_DEBUG) {
1525267Sroot 		ostate = tp->t_state;
1535267Sroot 		tcp_saveti = *ti;
1545267Sroot 	}
1557510Sroot 	if (so->so_options & SO_ACCEPTCONN) {
1567510Sroot 		so = sonewconn(so);
1577510Sroot 		if (so == 0)
1587510Sroot 			goto drop;
15910769Ssam 		/*
16010769Ssam 		 * This is ugly, but ....
16110769Ssam 		 *
16210769Ssam 		 * Mark socket as temporary until we're
16310769Ssam 		 * committed to keeping it.  The code at
16410769Ssam 		 * ``drop'' and ``dropwithreset'' check the
16510769Ssam 		 * flag dropsocket to see if the temporary
16610769Ssam 		 * socket created here should be discarded.
16710769Ssam 		 * We mark the socket as discardable until
16810769Ssam 		 * we're committed to it below in TCPS_LISTEN.
16910769Ssam 		 */
17010769Ssam 		dropsocket++;
1717510Sroot 		inp = (struct inpcb *)so->so_pcb;
1727510Sroot 		inp->inp_laddr = ti->ti_dst;
1737510Sroot 		inp->inp_lport = ti->ti_dport;
1747510Sroot 		tp = intotcpcb(inp);
1757510Sroot 		tp->t_state = TCPS_LISTEN;
1767510Sroot 	}
1774601Swnj 
1784601Swnj 	/*
1795162Swnj 	 * Segment received on connection.
1805162Swnj 	 * Reset idle time and keep-alive timer.
1815162Swnj 	 */
1825162Swnj 	tp->t_idle = 0;
1835162Swnj 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
1845162Swnj 
1855162Swnj 	/*
1865440Swnj 	 * Process options.
1875440Swnj 	 */
1885440Swnj 	if (om) {
1895440Swnj 		tcp_dooptions(tp, om);
1905440Swnj 		om = 0;
1915440Swnj 	}
1925440Swnj 
1935440Swnj 	/*
1945085Swnj 	 * Calculate amount of space in receive window,
1955085Swnj 	 * and then do TCP input processing.
1964601Swnj 	 */
1975085Swnj 	tp->rcv_wnd = sbspace(&so->so_rcv);
1985231Swnj 	if (tp->rcv_wnd < 0)
1995231Swnj 		tp->rcv_wnd = 0;
2004601Swnj 
2014601Swnj 	switch (tp->t_state) {
2024601Swnj 
2035065Swnj 	/*
2045065Swnj 	 * If the state is LISTEN then ignore segment if it contains an RST.
2055065Swnj 	 * If the segment contains an ACK then it is bad and send a RST.
2065065Swnj 	 * If it does not contain a SYN then it is not interesting; drop it.
2075085Swnj 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
2085065Swnj 	 * tp->iss, and send a segment:
2095085Swnj 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
2105065Swnj 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
2115065Swnj 	 * Fill in remote peer address fields if not previously specified.
2125065Swnj 	 * Enter SYN_RECEIVED state, and process any other fields of this
2135244Sroot 	 * segment in this state.
2145065Swnj 	 */
2158271Sroot 	case TCPS_LISTEN: {
21610145Ssam 		struct mbuf *am;
2178271Sroot 		register struct sockaddr_in *sin;
2188271Sroot 
2195065Swnj 		if (tiflags & TH_RST)
2205065Swnj 			goto drop;
2215300Sroot 		if (tiflags & TH_ACK)
2225085Swnj 			goto dropwithreset;
2235300Sroot 		if ((tiflags & TH_SYN) == 0)
2245065Swnj 			goto drop;
22510145Ssam 		am = m_get(M_DONTWAIT, MT_SONAME);
22610145Ssam 		if (am == NULL)
22710145Ssam 			goto drop;
22810145Ssam 		am->m_len = sizeof (struct sockaddr_in);
2298599Sroot 		sin = mtod(am, struct sockaddr_in *);
2308271Sroot 		sin->sin_family = AF_INET;
2318271Sroot 		sin->sin_addr = ti->ti_src;
2328271Sroot 		sin->sin_port = ti->ti_sport;
2336028Sroot 		laddr = inp->inp_laddr;
23410145Ssam 		if (inp->inp_laddr.s_addr == INADDR_ANY)
2356028Sroot 			inp->inp_laddr = ti->ti_dst;
2368599Sroot 		if (in_pcbconnect(inp, am)) {
2376028Sroot 			inp->inp_laddr = laddr;
2388716Sroot 			(void) m_free(am);
2395244Sroot 			goto drop;
2406028Sroot 		}
2418716Sroot 		(void) m_free(am);
2425244Sroot 		tp->t_template = tcp_template(tp);
2435244Sroot 		if (tp->t_template == 0) {
2445244Sroot 			in_pcbdisconnect(inp);
2456028Sroot 			inp->inp_laddr = laddr;
2466320Swnj 			tp = 0;
2475244Sroot 			goto drop;
2485244Sroot 		}
2495085Swnj 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
2505065Swnj 		tp->irs = ti->ti_seq;
2515085Swnj 		tcp_sendseqinit(tp);
2525085Swnj 		tcp_rcvseqinit(tp);
2535065Swnj 		tp->t_state = TCPS_SYN_RECEIVED;
2545244Sroot 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
25510769Ssam 		dropsocket = 0;		/* committed to socket */
2565085Swnj 		goto trimthenstep6;
2578271Sroot 		}
2584601Swnj 
2595065Swnj 	/*
2605065Swnj 	 * If the state is SYN_SENT:
2615065Swnj 	 *	if seg contains an ACK, but not for our SYN, drop the input.
2625065Swnj 	 *	if seg contains a RST, then drop the connection.
2635065Swnj 	 *	if seg does not contain SYN, then drop it.
2645065Swnj 	 * Otherwise this is an acceptable SYN segment
2655065Swnj 	 *	initialize tp->rcv_nxt and tp->irs
2665065Swnj 	 *	if seg contains ack then advance tp->snd_una
2675065Swnj 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
2685065Swnj 	 *	arrange for segment to be acked (eventually)
2695065Swnj 	 *	continue processing rest of data/controls, beginning with URG
2705065Swnj 	 */
2715065Swnj 	case TCPS_SYN_SENT:
2725065Swnj 		if ((tiflags & TH_ACK) &&
2735300Sroot /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
2745300Sroot 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
2755231Swnj 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
2765085Swnj 			goto dropwithreset;
2775065Swnj 		if (tiflags & TH_RST) {
27810394Ssam 			if (tiflags & TH_ACK)
27910394Ssam 				tp = tcp_drop(tp, ECONNREFUSED);
2805065Swnj 			goto drop;
2814601Swnj 		}
2825065Swnj 		if ((tiflags & TH_SYN) == 0)
2835065Swnj 			goto drop;
2845231Swnj 		tp->snd_una = ti->ti_ack;
2855357Sroot 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
2865357Sroot 			tp->snd_nxt = tp->snd_una;
2875244Sroot 		tp->t_timer[TCPT_REXMT] = 0;
2885065Swnj 		tp->irs = ti->ti_seq;
2895085Swnj 		tcp_rcvseqinit(tp);
2905085Swnj 		tp->t_flags |= TF_ACKNOW;
2915162Swnj 		if (SEQ_GT(tp->snd_una, tp->iss)) {
2925244Sroot 			soisconnected(so);
2935065Swnj 			tp->t_state = TCPS_ESTABLISHED;
2945162Swnj 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
2955162Swnj 		} else
2965085Swnj 			tp->t_state = TCPS_SYN_RECEIVED;
2975085Swnj 		goto trimthenstep6;
2985085Swnj 
2995085Swnj trimthenstep6:
3005085Swnj 		/*
3015231Swnj 		 * Advance ti->ti_seq to correspond to first data byte.
3025085Swnj 		 * If data, trim to stay within window,
3035085Swnj 		 * dropping FIN if necessary.
3045085Swnj 		 */
3055231Swnj 		ti->ti_seq++;
3065085Swnj 		if (ti->ti_len > tp->rcv_wnd) {
3075085Swnj 			todrop = ti->ti_len - tp->rcv_wnd;
3085085Swnj 			m_adj(m, -todrop);
3095085Swnj 			ti->ti_len = tp->rcv_wnd;
3105085Swnj 			ti->ti_flags &= ~TH_FIN;
3115065Swnj 		}
3125263Swnj 		tp->snd_wl1 = ti->ti_seq - 1;
3135085Swnj 		goto step6;
3145065Swnj 	}
3154601Swnj 
3165065Swnj 	/*
31716222Skarels 	 * If data is received on a connection after the
31816222Skarels 	 * user processes are gone, then RST the other end.
31916222Skarels 	 */
32016222Skarels 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
32116222Skarels 	    ti->ti_len) {
32216222Skarels 		tp = tcp_close(tp);
32316222Skarels 		goto dropwithreset;
32416222Skarels 	}
32516222Skarels 
32616222Skarels 	/*
3275065Swnj 	 * States other than LISTEN or SYN_SENT.
3285065Swnj 	 * First check that at least some bytes of segment are within
3295065Swnj 	 * receive window.
3305065Swnj 	 */
3315065Swnj 	if (tp->rcv_wnd == 0) {
3325065Swnj 		/*
3335065Swnj 		 * If window is closed can only take segments at
3345231Swnj 		 * window edge, and have to drop data and PUSH from
3355065Swnj 		 * incoming segments.
3365065Swnj 		 */
3375300Sroot 		if (tp->rcv_nxt != ti->ti_seq)
3385065Swnj 			goto dropafterack;
3395085Swnj 		if (ti->ti_len > 0) {
3405690Swnj 			m_adj(m, ti->ti_len);
3415085Swnj 			ti->ti_len = 0;
3425085Swnj 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
3435065Swnj 		}
3445065Swnj 	} else {
3455065Swnj 		/*
3465231Swnj 		 * If segment begins before rcv_nxt, drop leading
3475065Swnj 		 * data (and SYN); if nothing left, just ack.
3485065Swnj 		 */
3495690Swnj 		todrop = tp->rcv_nxt - ti->ti_seq;
3505690Swnj 		if (todrop > 0) {
3515085Swnj 			if (tiflags & TH_SYN) {
3525300Sroot 				tiflags &= ~TH_SYN;
3535690Swnj 				ti->ti_flags &= ~TH_SYN;
3545085Swnj 				ti->ti_seq++;
3555085Swnj 				if (ti->ti_urp > 1)
3565085Swnj 					ti->ti_urp--;
3575085Swnj 				else
3585085Swnj 					tiflags &= ~TH_URG;
3595085Swnj 				todrop--;
3605085Swnj 			}
3616211Swnj 			if (todrop > ti->ti_len ||
3626211Swnj 			    todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
3635065Swnj 				goto dropafterack;
3645065Swnj 			m_adj(m, todrop);
3655065Swnj 			ti->ti_seq += todrop;
3665065Swnj 			ti->ti_len -= todrop;
3675085Swnj 			if (ti->ti_urp > todrop)
3685085Swnj 				ti->ti_urp -= todrop;
3695085Swnj 			else {
3705085Swnj 				tiflags &= ~TH_URG;
3715690Swnj 				ti->ti_flags &= ~TH_URG;
3725690Swnj 				ti->ti_urp = 0;
3735085Swnj 			}
3745065Swnj 		}
3755065Swnj 		/*
3765065Swnj 		 * If segment ends after window, drop trailing data
3775085Swnj 		 * (and PUSH and FIN); if nothing left, just ACK.
3785065Swnj 		 */
3795690Swnj 		todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
3805690Swnj 		if (todrop > 0) {
3816211Swnj 			if (todrop >= ti->ti_len)
3825065Swnj 				goto dropafterack;
3835065Swnj 			m_adj(m, -todrop);
3845065Swnj 			ti->ti_len -= todrop;
3855085Swnj 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
3865065Swnj 		}
3875065Swnj 	}
3884601Swnj 
3895065Swnj 	/*
3905065Swnj 	 * If the RST bit is set examine the state:
3915065Swnj 	 *    SYN_RECEIVED STATE:
3925065Swnj 	 *	If passive open, return to LISTEN state.
3935065Swnj 	 *	If active open, inform user that connection was refused.
3945065Swnj 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
3955065Swnj 	 *	Inform user that connection was reset, and close tcb.
3965065Swnj 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
3975065Swnj 	 *	Close the tcb.
3985065Swnj 	 */
3995065Swnj 	if (tiflags&TH_RST) switch (tp->t_state) {
4005267Sroot 
4015065Swnj 	case TCPS_SYN_RECEIVED:
40210394Ssam 		tp = tcp_drop(tp, ECONNREFUSED);
4035065Swnj 		goto drop;
4044601Swnj 
4055065Swnj 	case TCPS_ESTABLISHED:
4065065Swnj 	case TCPS_FIN_WAIT_1:
4075065Swnj 	case TCPS_FIN_WAIT_2:
4085065Swnj 	case TCPS_CLOSE_WAIT:
40910394Ssam 		tp = tcp_drop(tp, ECONNRESET);
4105065Swnj 		goto drop;
4115065Swnj 
4125065Swnj 	case TCPS_CLOSING:
4135065Swnj 	case TCPS_LAST_ACK:
4145065Swnj 	case TCPS_TIME_WAIT:
41510394Ssam 		tp = tcp_close(tp);
4165065Swnj 		goto drop;
4174601Swnj 	}
4184601Swnj 
4194601Swnj 	/*
4205065Swnj 	 * If a SYN is in the window, then this is an
4215065Swnj 	 * error and we send an RST and drop the connection.
4224601Swnj 	 */
4235065Swnj 	if (tiflags & TH_SYN) {
42410394Ssam 		tp = tcp_drop(tp, ECONNRESET);
4255085Swnj 		goto dropwithreset;
4264601Swnj 	}
4274601Swnj 
4284601Swnj 	/*
4295065Swnj 	 * If the ACK bit is off we drop the segment and return.
4304601Swnj 	 */
4315085Swnj 	if ((tiflags & TH_ACK) == 0)
4325065Swnj 		goto drop;
4335065Swnj 
4345065Swnj 	/*
4355065Swnj 	 * Ack processing.
4365065Swnj 	 */
4374601Swnj 	switch (tp->t_state) {
4384601Swnj 
4395065Swnj 	/*
4405065Swnj 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
4415065Swnj 	 * ESTABLISHED state and continue processing, othewise
4425065Swnj 	 * send an RST.
4435065Swnj 	 */
4445065Swnj 	case TCPS_SYN_RECEIVED:
4455085Swnj 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
4465231Swnj 		    SEQ_GT(ti->ti_ack, tp->snd_max))
4475085Swnj 			goto dropwithreset;
4485244Sroot 		tp->snd_una++;			/* SYN acked */
4495357Sroot 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
4505357Sroot 			tp->snd_nxt = tp->snd_una;
4515244Sroot 		tp->t_timer[TCPT_REXMT] = 0;
4525085Swnj 		soisconnected(so);
4535085Swnj 		tp->t_state = TCPS_ESTABLISHED;
4545162Swnj 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
4555244Sroot 		tp->snd_wl1 = ti->ti_seq - 1;
4565085Swnj 		/* fall into ... */
4574601Swnj 
4585065Swnj 	/*
4595065Swnj 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
4605065Swnj 	 * ACKs.  If the ack is in the range
4615231Swnj 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
4625065Swnj 	 * then advance tp->snd_una to ti->ti_ack and drop
4635065Swnj 	 * data from the retransmission queue.  If this ACK reflects
4645065Swnj 	 * more up to date window information we update our window information.
4655065Swnj 	 */
4665065Swnj 	case TCPS_ESTABLISHED:
4675065Swnj 	case TCPS_FIN_WAIT_1:
4685065Swnj 	case TCPS_FIN_WAIT_2:
4695065Swnj 	case TCPS_CLOSE_WAIT:
4705065Swnj 	case TCPS_CLOSING:
4715244Sroot 	case TCPS_LAST_ACK:
4725244Sroot 	case TCPS_TIME_WAIT:
4735085Swnj #define	ourfinisacked	(acked > 0)
4745085Swnj 
4755244Sroot 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
4765065Swnj 			break;
4775300Sroot 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
4785065Swnj 			goto dropafterack;
4795085Swnj 		acked = ti->ti_ack - tp->snd_una;
4805951Swnj 
4815951Swnj 		/*
4825951Swnj 		 * If transmit timer is running and timed sequence
4835951Swnj 		 * number was acked, update smoothed round trip time.
4845951Swnj 		 */
4855951Swnj 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
4865951Swnj 			if (tp->t_srtt == 0)
4875951Swnj 				tp->t_srtt = tp->t_rtt;
4885951Swnj 			else
4895951Swnj 				tp->t_srtt =
4905951Swnj 				    tcp_alpha * tp->t_srtt +
4915951Swnj 				    (1 - tcp_alpha) * tp->t_rtt;
4925951Swnj 			tp->t_rtt = 0;
4935951Swnj 		}
4945951Swnj 
4955307Sroot 		if (ti->ti_ack == tp->snd_max)
4965244Sroot 			tp->t_timer[TCPT_REXMT] = 0;
4975307Sroot 		else {
4985244Sroot 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
4995244Sroot 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
5005951Swnj 			tp->t_rtt = 1;
5015300Sroot 			tp->t_rxtshift = 0;
5025085Swnj 		}
5035307Sroot 		if (acked > so->so_snd.sb_cc) {
50415386Ssam 			tp->snd_wnd -= so->so_snd.sb_cc;
5055307Sroot 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
5065307Sroot 		} else {
5076161Ssam 			sbdrop(&so->so_snd, acked);
5085307Sroot 			tp->snd_wnd -= acked;
5095307Sroot 			acked = 0;
5105307Sroot 		}
5116434Swnj 		if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
5125300Sroot 			sowwakeup(so);
5135231Swnj 		tp->snd_una = ti->ti_ack;
5145357Sroot 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
5155357Sroot 			tp->snd_nxt = tp->snd_una;
5165162Swnj 
5174601Swnj 		switch (tp->t_state) {
5184601Swnj 
5195065Swnj 		/*
5205065Swnj 		 * In FIN_WAIT_1 STATE in addition to the processing
5215065Swnj 		 * for the ESTABLISHED state if our FIN is now acknowledged
5225085Swnj 		 * then enter FIN_WAIT_2.
5235065Swnj 		 */
5245065Swnj 		case TCPS_FIN_WAIT_1:
5255896Swnj 			if (ourfinisacked) {
5265896Swnj 				/*
5275896Swnj 				 * If we can't receive any more
5285896Swnj 				 * data, then closing user can proceed.
5295896Swnj 				 */
5305896Swnj 				if (so->so_state & SS_CANTRCVMORE)
5315896Swnj 					soisdisconnected(so);
5325085Swnj 				tp->t_state = TCPS_FIN_WAIT_2;
5335896Swnj 			}
5344601Swnj 			break;
5354601Swnj 
5365065Swnj 	 	/*
5375065Swnj 		 * In CLOSING STATE in addition to the processing for
5385065Swnj 		 * the ESTABLISHED state if the ACK acknowledges our FIN
5395065Swnj 		 * then enter the TIME-WAIT state, otherwise ignore
5405065Swnj 		 * the segment.
5415065Swnj 		 */
5425065Swnj 		case TCPS_CLOSING:
5435244Sroot 			if (ourfinisacked) {
5445065Swnj 				tp->t_state = TCPS_TIME_WAIT;
5455244Sroot 				tcp_canceltimers(tp);
5465244Sroot 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
5475244Sroot 				soisdisconnected(so);
5485244Sroot 			}
5495244Sroot 			break;
5504601Swnj 
5515065Swnj 		/*
5525085Swnj 		 * The only thing that can arrive in  LAST_ACK state
5535085Swnj 		 * is an acknowledgment of our FIN.  If our FIN is now
5545085Swnj 		 * acknowledged, delete the TCB, enter the closed state
5555085Swnj 		 * and return.
5565065Swnj 		 */
5575065Swnj 		case TCPS_LAST_ACK:
55810394Ssam 			if (ourfinisacked)
55910394Ssam 				tp = tcp_close(tp);
5605065Swnj 			goto drop;
5614601Swnj 
5625065Swnj 		/*
5635065Swnj 		 * In TIME_WAIT state the only thing that should arrive
5645065Swnj 		 * is a retransmission of the remote FIN.  Acknowledge
5655065Swnj 		 * it and restart the finack timer.
5665065Swnj 		 */
5675065Swnj 		case TCPS_TIME_WAIT:
5685162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
5695065Swnj 			goto dropafterack;
5704601Swnj 		}
5715085Swnj #undef ourfinisacked
5725085Swnj 	}
5734601Swnj 
5745065Swnj step6:
5755065Swnj 	/*
5765244Sroot 	 * Update window information.
5775244Sroot 	 */
5785300Sroot 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
5795391Swnj 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
5805300Sroot 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
5815244Sroot 		tp->snd_wnd = ti->ti_win;
5825244Sroot 		tp->snd_wl1 = ti->ti_seq;
5835244Sroot 		tp->snd_wl2 = ti->ti_ack;
5848599Sroot 		if (tp->snd_wnd != 0)
5855244Sroot 			tp->t_timer[TCPT_PERSIST] = 0;
5865244Sroot 	}
5875244Sroot 
5885244Sroot 	/*
5895547Swnj 	 * Process segments with URG.
5905065Swnj 	 */
5917267Swnj 	if ((tiflags & TH_URG) && ti->ti_urp &&
5927267Swnj 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
5935547Swnj 		/*
59413121Ssam 		 * This is a kludge, but if we receive accept
59513121Ssam 		 * random urgent pointers, we'll crash in
59613121Ssam 		 * soreceive.  It's hard to imagine someone
59713121Ssam 		 * actually wanting to send this much urgent data.
59812441Ssam 		 */
59912441Ssam 		if (ti->ti_urp > tp->t_maxseg) {	/* XXX */
60012441Ssam 			ti->ti_urp = 0;			/* XXX */
60112441Ssam 			tiflags &= ~TH_URG;		/* XXX */
60212441Ssam 			ti->ti_flags &= ~TH_URG;	/* XXX */
60313121Ssam 			goto badurp;			/* XXX */
60412441Ssam 		}
60512441Ssam 		/*
6065547Swnj 		 * If this segment advances the known urgent pointer,
6075547Swnj 		 * then mark the data stream.  This should not happen
6085547Swnj 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
6095547Swnj 		 * a FIN has been received from the remote side.
6105547Swnj 		 * In these states we ignore the URG.
6115547Swnj 		 */
6125547Swnj 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
6135547Swnj 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
6145547Swnj 			so->so_oobmark = so->so_rcv.sb_cc +
6155547Swnj 			    (tp->rcv_up - tp->rcv_nxt) - 1;
6165547Swnj 			if (so->so_oobmark == 0)
6175547Swnj 				so->so_state |= SS_RCVATMARK;
6188313Sroot 			sohasoutofband(so);
6195547Swnj 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
6205440Swnj 		}
6215547Swnj 		/*
6225547Swnj 		 * Remove out of band data so doesn't get presented to user.
6235547Swnj 		 * This can happen independent of advancing the URG pointer,
6245547Swnj 		 * but if two URG's are pending at once, some out-of-band
6255547Swnj 		 * data may creep in... ick.
6265547Swnj 		 */
6277510Sroot 		if (ti->ti_urp <= ti->ti_len)
6285547Swnj 			tcp_pulloutofband(so, ti);
6295419Swnj 	}
63013121Ssam badurp:							/* XXX */
6314601Swnj 
6324601Swnj 	/*
6335065Swnj 	 * Process the segment text, merging it into the TCP sequencing queue,
6345065Swnj 	 * and arranging for acknowledgment of receipt if necessary.
6355065Swnj 	 * This process logically involves adjusting tp->rcv_wnd as data
6365065Swnj 	 * is presented to the user (this happens in tcp_usrreq.c,
6375065Swnj 	 * case PRU_RCVD).  If a FIN has already been received on this
6385065Swnj 	 * connection then we just ignore the text.
6394601Swnj 	 */
6405263Swnj 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
6415263Swnj 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
6425065Swnj 		tiflags = tcp_reass(tp, ti);
6435440Swnj 		if (tcpnodelack == 0)
6445440Swnj 			tp->t_flags |= TF_DELACK;
6455440Swnj 		else
6465440Swnj 			tp->t_flags |= TF_ACKNOW;
6475244Sroot 	} else {
6484924Swnj 		m_freem(m);
6495263Swnj 		tiflags &= ~TH_FIN;
6505244Sroot 	}
6514601Swnj 
6524601Swnj 	/*
6535263Swnj 	 * If FIN is received ACK the FIN and let the user know
6545263Swnj 	 * that the connection is closing.
6554601Swnj 	 */
6565263Swnj 	if (tiflags & TH_FIN) {
6575244Sroot 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
6585244Sroot 			socantrcvmore(so);
6595244Sroot 			tp->t_flags |= TF_ACKNOW;
6605244Sroot 			tp->rcv_nxt++;
6615244Sroot 		}
6625065Swnj 		switch (tp->t_state) {
6634601Swnj 
6645065Swnj 	 	/*
6655065Swnj 		 * In SYN_RECEIVED and ESTABLISHED STATES
6665065Swnj 		 * enter the CLOSE_WAIT state.
6674884Swnj 		 */
6685065Swnj 		case TCPS_SYN_RECEIVED:
6695065Swnj 		case TCPS_ESTABLISHED:
6705065Swnj 			tp->t_state = TCPS_CLOSE_WAIT;
6715065Swnj 			break;
6724884Swnj 
6735065Swnj 	 	/*
6745085Swnj 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
6755085Swnj 		 * enter the CLOSING state.
6764884Swnj 		 */
6775065Swnj 		case TCPS_FIN_WAIT_1:
6785085Swnj 			tp->t_state = TCPS_CLOSING;
6795065Swnj 			break;
6804601Swnj 
6815065Swnj 	 	/*
6825065Swnj 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
6835065Swnj 		 * starting the time-wait timer, turning off the other
6845065Swnj 		 * standard timers.
6855065Swnj 		 */
6865065Swnj 		case TCPS_FIN_WAIT_2:
6875244Sroot 			tp->t_state = TCPS_TIME_WAIT;
6885074Swnj 			tcp_canceltimers(tp);
6895162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
6905244Sroot 			soisdisconnected(so);
6915065Swnj 			break;
6925065Swnj 
6934884Swnj 		/*
6945065Swnj 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
6954884Swnj 		 */
6965065Swnj 		case TCPS_TIME_WAIT:
6975162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
6985065Swnj 			break;
6995085Swnj 		}
7004601Swnj 	}
7015267Sroot 	if (so->so_options & SO_DEBUG)
7025267Sroot 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
7035085Swnj 
7045085Swnj 	/*
7055085Swnj 	 * Return any desired output.
7065085Swnj 	 */
7076161Ssam 	(void) tcp_output(tp);
7085065Swnj 	return;
7095085Swnj 
7105065Swnj dropafterack:
7115085Swnj 	/*
7126211Swnj 	 * Generate an ACK dropping incoming segment if it occupies
7136211Swnj 	 * sequence space, where the ACK reflects our state.
7145085Swnj 	 */
7156211Swnj 	if ((tiflags&TH_RST) ||
7166211Swnj 	    tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
7175085Swnj 		goto drop;
7186303Sroot 	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
7196303Sroot 		tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
7205391Swnj 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
7215231Swnj 	return;
7225085Swnj 
7235085Swnj dropwithreset:
72411731Ssam 	if (om) {
7256161Ssam 		(void) m_free(om);
72611731Ssam 		om = 0;
72711731Ssam 	}
7285085Swnj 	/*
7295244Sroot 	 * Generate a RST, dropping incoming segment.
7305085Swnj 	 * Make ACK acceptable to originator of segment.
7315085Swnj 	 */
7325085Swnj 	if (tiflags & TH_RST)
7335085Swnj 		goto drop;
7345085Swnj 	if (tiflags & TH_ACK)
7355391Swnj 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
7365085Swnj 	else {
7375085Swnj 		if (tiflags & TH_SYN)
7385085Swnj 			ti->ti_len++;
7396211Swnj 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
7406211Swnj 		    TH_RST|TH_ACK);
7415085Swnj 	}
74210769Ssam 	/* destroy temporarily created socket */
74310769Ssam 	if (dropsocket)
74410769Ssam 		(void) soabort(so);
7455231Swnj 	return;
7465085Swnj 
7475065Swnj drop:
74811730Ssam 	if (om)
74911730Ssam 		(void) m_free(om);
7505085Swnj 	/*
7515085Swnj 	 * Drop space held by incoming segment and return.
7525085Swnj 	 */
7536303Sroot 	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
7546303Sroot 		tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
7555065Swnj 	m_freem(m);
75610769Ssam 	/* destroy temporarily created socket */
75710769Ssam 	if (dropsocket)
75810769Ssam 		(void) soabort(so);
7595267Sroot 	return;
7605065Swnj }
7615065Swnj 
7625440Swnj tcp_dooptions(tp, om)
7635440Swnj 	struct tcpcb *tp;
7645440Swnj 	struct mbuf *om;
7655419Swnj {
7665440Swnj 	register u_char *cp;
7675440Swnj 	int opt, optlen, cnt;
7685419Swnj 
7695440Swnj 	cp = mtod(om, u_char *);
7705440Swnj 	cnt = om->m_len;
7715440Swnj 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
7725440Swnj 		opt = cp[0];
7735440Swnj 		if (opt == TCPOPT_EOL)
7745440Swnj 			break;
7755440Swnj 		if (opt == TCPOPT_NOP)
7765440Swnj 			optlen = 1;
77712169Ssam 		else {
7785440Swnj 			optlen = cp[1];
77912169Ssam 			if (optlen <= 0)
78012169Ssam 				break;
78112169Ssam 		}
7825440Swnj 		switch (opt) {
7835440Swnj 
7845440Swnj 		default:
7855440Swnj 			break;
7865440Swnj 
7875440Swnj 		case TCPOPT_MAXSEG:
7885440Swnj 			if (optlen != 4)
7895440Swnj 				continue;
7905440Swnj 			tp->t_maxseg = *(u_short *)(cp + 2);
7916161Ssam 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
7925440Swnj 			break;
7935419Swnj 		}
7945419Swnj 	}
7956161Ssam 	(void) m_free(om);
7965419Swnj }
7975419Swnj 
7985419Swnj /*
7995547Swnj  * Pull out of band byte out of a segment so
8005547Swnj  * it doesn't appear in the user's data queue.
8015547Swnj  * It is still reflected in the segment length for
8025547Swnj  * sequencing purposes.
8035547Swnj  */
8045547Swnj tcp_pulloutofband(so, ti)
8055547Swnj 	struct socket *so;
8065547Swnj 	struct tcpiphdr *ti;
8075547Swnj {
8085547Swnj 	register struct mbuf *m;
8096116Swnj 	int cnt = ti->ti_urp - 1;
8105547Swnj 
8115547Swnj 	m = dtom(ti);
8125547Swnj 	while (cnt >= 0) {
8135547Swnj 		if (m->m_len > cnt) {
8145547Swnj 			char *cp = mtod(m, caddr_t) + cnt;
8155547Swnj 			struct tcpcb *tp = sototcpcb(so);
8165547Swnj 
8175547Swnj 			tp->t_iobc = *cp;
8185547Swnj 			tp->t_oobflags |= TCPOOB_HAVEDATA;
8196161Ssam 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
8205547Swnj 			m->m_len--;
8215547Swnj 			return;
8225547Swnj 		}
8235547Swnj 		cnt -= m->m_len;
8245547Swnj 		m = m->m_next;
8255547Swnj 		if (m == 0)
8265547Swnj 			break;
8275547Swnj 	}
8285547Swnj 	panic("tcp_pulloutofband");
8295547Swnj }
8305547Swnj 
8315547Swnj /*
8325065Swnj  * Insert segment ti into reassembly queue of tcp with
8335065Swnj  * control block tp.  Return TH_FIN if reassembly now includes
8345065Swnj  * a segment with FIN.
8355065Swnj  */
8365109Swnj tcp_reass(tp, ti)
8375065Swnj 	register struct tcpcb *tp;
8385065Swnj 	register struct tcpiphdr *ti;
8395065Swnj {
8405065Swnj 	register struct tcpiphdr *q;
8415085Swnj 	struct socket *so = tp->t_inpcb->inp_socket;
8425263Swnj 	struct mbuf *m;
8435263Swnj 	int flags;
8445065Swnj 
8455065Swnj 	/*
8465162Swnj 	 * Call with ti==0 after become established to
8475162Swnj 	 * force pre-ESTABLISHED data up to user socket.
8485065Swnj 	 */
8495162Swnj 	if (ti == 0)
8505065Swnj 		goto present;
8514601Swnj 
8525065Swnj 	/*
8535065Swnj 	 * Find a segment which begins after this one does.
8545065Swnj 	 */
8555065Swnj 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
8565065Swnj 	    q = (struct tcpiphdr *)q->ti_next)
8575065Swnj 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
8585065Swnj 			break;
8594601Swnj 
8605065Swnj 	/*
8615065Swnj 	 * If there is a preceding segment, it may provide some of
8625065Swnj 	 * our data already.  If so, drop the data from the incoming
8635065Swnj 	 * segment.  If it provides all of our data, drop us.
8645065Swnj 	 */
8655065Swnj 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
8665065Swnj 		register int i;
8675690Swnj 		q = (struct tcpiphdr *)q->ti_prev;
8685065Swnj 		/* conversion to int (in i) handles seq wraparound */
8695065Swnj 		i = q->ti_seq + q->ti_len - ti->ti_seq;
8705065Swnj 		if (i > 0) {
8714924Swnj 			if (i >= ti->ti_len)
8725065Swnj 				goto drop;
8737338Swnj 			m_adj(dtom(ti), i);
8745065Swnj 			ti->ti_len -= i;
8754924Swnj 			ti->ti_seq += i;
8764601Swnj 		}
8775065Swnj 		q = (struct tcpiphdr *)(q->ti_next);
8785065Swnj 	}
8794601Swnj 
8805065Swnj 	/*
8815065Swnj 	 * While we overlap succeeding segments trim them or,
8825065Swnj 	 * if they are completely covered, dequeue them.
8835065Swnj 	 */
8845690Swnj 	while (q != (struct tcpiphdr *)tp) {
8855065Swnj 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
8865690Swnj 		if (i <= 0)
8875690Swnj 			break;
8885065Swnj 		if (i < q->ti_len) {
8895690Swnj 			q->ti_seq += i;
8905065Swnj 			q->ti_len -= i;
8915065Swnj 			m_adj(dtom(q), i);
8925065Swnj 			break;
8934601Swnj 		}
8945065Swnj 		q = (struct tcpiphdr *)q->ti_next;
8955623Swnj 		m = dtom(q->ti_prev);
8965065Swnj 		remque(q->ti_prev);
8975623Swnj 		m_freem(m);
8985065Swnj 	}
8994601Swnj 
9005065Swnj 	/*
9015065Swnj 	 * Stick new segment in its place.
9025065Swnj 	 */
9035065Swnj 	insque(ti, q->ti_prev);
9044601Swnj 
9055065Swnj present:
9065065Swnj 	/*
9075244Sroot 	 * Present data to user, advancing rcv_nxt through
9085244Sroot 	 * completed sequence space.
9095065Swnj 	 */
9105263Swnj 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
9115244Sroot 		return (0);
9124924Swnj 	ti = tp->seg_next;
9135263Swnj 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
9145263Swnj 		return (0);
9155263Swnj 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
9165263Swnj 		return (0);
9175263Swnj 	do {
9185244Sroot 		tp->rcv_nxt += ti->ti_len;
9195244Sroot 		flags = ti->ti_flags & TH_FIN;
9204924Swnj 		remque(ti);
9215263Swnj 		m = dtom(ti);
9224924Swnj 		ti = (struct tcpiphdr *)ti->ti_next;
9235263Swnj 		if (so->so_state & SS_CANTRCVMORE)
9246161Ssam 			m_freem(m);
92510145Ssam 		else
9265263Swnj 			sbappend(&so->so_rcv, m);
9275263Swnj 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
9285263Swnj 	sorwakeup(so);
9295065Swnj 	return (flags);
9305065Swnj drop:
9315065Swnj 	m_freem(dtom(ti));
9325263Swnj 	return (0);
9334601Swnj }
934