xref: /csrg-svn/sys/netinet/tcp_input.c (revision 5244)
1*5244Sroot /*	tcp_input.c	1.39	81/12/12	*/
24601Swnj 
34601Swnj #include "../h/param.h"
44601Swnj #include "../h/systm.h"
54663Swnj #include "../h/mbuf.h"
65085Swnj #include "../h/protosw.h"
74663Swnj #include "../h/socket.h"
84803Swnj #include "../h/socketvar.h"
95085Swnj #include "../net/in.h"
105085Swnj #include "../net/in_pcb.h"
115085Swnj #include "../net/in_systm.h"
125085Swnj #include "../net/if.h"
134803Swnj #include "../net/ip.h"
144899Swnj #include "../net/ip_var.h"
154803Swnj #include "../net/tcp.h"
16*5244Sroot #define TCPSTATES
174803Swnj #include "../net/tcp_fsm.h"
185085Swnj #include "../net/tcp_seq.h"
195085Swnj #include "../net/tcp_timer.h"
204803Swnj #include "../net/tcp_var.h"
215085Swnj #include "../net/tcpip.h"
225109Swnj #include "../errno.h"
234601Swnj 
244679Swnj int	tcpcksum = 1;
25*5244Sroot struct	sockaddr_in tcp_in = { AF_INET };
264601Swnj 
275065Swnj /*
285065Swnj  * TCP input routine, follows pages 65-76 of the
295065Swnj  * protocol specification dated September, 1981 very closely.
305065Swnj  */
314924Swnj tcp_input(m0)
324924Swnj 	struct mbuf *m0;
334601Swnj {
344924Swnj 	register struct tcpiphdr *ti;
354924Swnj 	struct inpcb *inp;
364924Swnj 	register struct mbuf *m;
374924Swnj 	int len, tlen, off;
384924Swnj 	register struct tcpcb *tp;
394924Swnj 	register int tiflags;
404803Swnj 	struct socket *so;
415109Swnj 	int todrop, acked;
424924Swnj 
434601Swnj COUNT(TCP_INPUT);
444924Swnj 	/*
45*5244Sroot 	 * Get IP and TCP header together in first mbuf.
46*5244Sroot 	 * Note: IP leaves IP header in first mbuf.
474924Swnj 	 */
484924Swnj 	m = m0;
495020Sroot 	ti = mtod(m, struct tcpiphdr *);
50*5244Sroot 	if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2))
515208Swnj 		ip_stripoptions((struct ip *)ti, (struct mbuf *)0);
525085Swnj 	if (m->m_len < sizeof (struct tcpiphdr)) {
53*5244Sroot printf("m->m_len %d\n", m->m_len);
545085Swnj 		if (m_pullup(m, sizeof (struct tcpiphdr)) == 0) {
55*5244Sroot printf("tcp_input: header drop\n");
565085Swnj 			tcpstat.tcps_hdrops++;
575085Swnj 			goto drop;
585085Swnj 		}
595085Swnj 		ti = mtod(m, struct tcpiphdr *);
605085Swnj 	}
614601Swnj 
624601Swnj 	/*
63*5244Sroot 	 * Checksum extended TCP header and data.
644601Swnj 	 */
654924Swnj 	tlen = ((struct ip *)ti)->ip_len;
664924Swnj 	len = sizeof (struct ip) + tlen;
674679Swnj 	if (tcpcksum) {
684924Swnj 		ti->ti_next = ti->ti_prev = 0;
694924Swnj 		ti->ti_x1 = 0;
705223Swnj 		ti->ti_len = (u_short)tlen;
715223Swnj #if vax
725223Swnj 		ti->ti_len = htons(ti->ti_len);
735223Swnj #endif
745231Swnj 		if (ti->ti_sum = in_cksum(m, len)) {
754924Swnj 			tcpstat.tcps_badsum++;
765065Swnj 			printf("tcp cksum %x\n", ti->ti_sum);
775085Swnj 			goto drop;
784601Swnj 		}
794601Swnj 	}
804601Swnj 
814601Swnj 	/*
82*5244Sroot 	 * Check that TCP offset makes sense,
83*5244Sroot 	 * process TCP options and adjust length.
844601Swnj 	 */
854924Swnj 	off = ti->ti_off << 2;
865231Swnj 	if (off < sizeof (struct tcphdr) || off > tlen) {
87*5244Sroot printf("tcp_input: bad offset\n");
884924Swnj 		tcpstat.tcps_badoff++;
895085Swnj 		goto drop;
904924Swnj 	}
914924Swnj 	ti->ti_len = tlen - off;
925085Swnj #if 0
935231Swnj 	if (off > sizeof (struct tcphdr))
945085Swnj 		tcp_options(ti);
955085Swnj #endif
965065Swnj 	tiflags = ti->ti_flags;
974924Swnj 
985231Swnj #if vax
994924Swnj 	/*
100*5244Sroot 	 * Convert TCP protocol specific fields to host format.
1015085Swnj 	 */
1025085Swnj 	ti->ti_seq = ntohl(ti->ti_seq);
1035085Swnj 	ti->ti_ack = ntohl(ti->ti_ack);
1045085Swnj 	ti->ti_win = ntohs(ti->ti_win);
1055085Swnj 	ti->ti_urp = ntohs(ti->ti_urp);
1065231Swnj #endif
1075085Swnj 
1085085Swnj 	/*
1094924Swnj 	 * Locate pcb for segment.
1104924Swnj 	 */
1115065Swnj 	inp = in_pcblookup
1125065Swnj 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport);
1135065Swnj 
1145065Swnj 	/*
1155065Swnj 	 * If the state is CLOSED (i.e., TCB does not exist) then
116*5244Sroot 	 * all data in the incoming segment is discarded.
1175065Swnj 	 */
1184884Swnj 	if (inp == 0)
1195085Swnj 		goto dropwithreset;
1205065Swnj 	tp = intotcpcb(inp);
1215065Swnj 	if (tp == 0)
1225085Swnj 		goto dropwithreset;
123*5244Sroot printf("tcp_input: segment seq %x ack %x win %x inp %x flags",
124*5244Sroot ti->ti_seq, ti->ti_ack, ti->ti_win, inp);
125*5244Sroot if (ti->ti_flags & TH_FIN) printf(" FIN");
126*5244Sroot if (ti->ti_flags & TH_SYN) printf(" SYN");
127*5244Sroot if (ti->ti_flags & TH_RST) printf(" RST");
128*5244Sroot if (ti->ti_flags & TH_PUSH) printf(" PUSH");
129*5244Sroot if (ti->ti_flags & TH_ACK) printf(" ACK");
130*5244Sroot if (ti->ti_flags & TH_URG) printf(" URG");
131*5244Sroot printf("\n");
132*5244Sroot printf("tcp_input: "); pseqno(tp);
1335109Swnj 	so = inp->inp_socket;
1344601Swnj 
1354601Swnj 	/*
1365162Swnj 	 * Segment received on connection.
1375162Swnj 	 * Reset idle time and keep-alive timer.
1385162Swnj 	 */
1395162Swnj 	tp->t_idle = 0;
1405162Swnj 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
1415162Swnj 
1425162Swnj 	/*
1435085Swnj 	 * Calculate amount of space in receive window,
1445085Swnj 	 * and then do TCP input processing.
1454601Swnj 	 */
1465085Swnj 	tp->rcv_wnd = sbspace(&so->so_rcv);
1475231Swnj 	if (tp->rcv_wnd < 0)
1485231Swnj 		tp->rcv_wnd = 0;
1494601Swnj 
1504601Swnj 	switch (tp->t_state) {
1514601Swnj 
1525065Swnj 	/*
1535065Swnj 	 * If the state is LISTEN then ignore segment if it contains an RST.
1545065Swnj 	 * If the segment contains an ACK then it is bad and send a RST.
1555065Swnj 	 * If it does not contain a SYN then it is not interesting; drop it.
1565085Swnj 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
1575065Swnj 	 * tp->iss, and send a segment:
1585085Swnj 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
1595065Swnj 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
1605065Swnj 	 * Fill in remote peer address fields if not previously specified.
1615065Swnj 	 * Enter SYN_RECEIVED state, and process any other fields of this
162*5244Sroot 	 * segment in this state.
1635065Swnj 	 */
1645065Swnj 	case TCPS_LISTEN:
1655065Swnj 		if (tiflags & TH_RST)
1665065Swnj 			goto drop;
1675065Swnj 		if (tiflags & TH_ACK)
1685085Swnj 			goto dropwithreset;
1695065Swnj 		if ((tiflags & TH_SYN) == 0)
1705065Swnj 			goto drop;
171*5244Sroot 		tcp_in.sin_addr = ti->ti_src;
172*5244Sroot 		tcp_in.sin_port = ti->ti_sport;
173*5244Sroot 		if (in_pcbconnect(inp, (struct sockaddr *)&tcp_in))
174*5244Sroot 			goto drop;
175*5244Sroot 		tp->t_template = tcp_template(tp);
176*5244Sroot 		if (tp->t_template == 0) {
177*5244Sroot 			in_pcbdisconnect(inp);
178*5244Sroot 			goto drop;
179*5244Sroot 		}
1805085Swnj 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
1815065Swnj 		tp->irs = ti->ti_seq;
1825085Swnj 		tcp_sendseqinit(tp);
1835085Swnj 		tcp_rcvseqinit(tp);
1845065Swnj 		tp->t_state = TCPS_SYN_RECEIVED;
185*5244Sroot 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
186*5244Sroot printf("tcp_input: out of LISTEN: ");
1875085Swnj 		goto trimthenstep6;
1884601Swnj 
1895065Swnj 	/*
1905065Swnj 	 * If the state is SYN_SENT:
1915065Swnj 	 *	if seg contains an ACK, but not for our SYN, drop the input.
1925065Swnj 	 *	if seg contains a RST, then drop the connection.
1935065Swnj 	 *	if seg does not contain SYN, then drop it.
1945065Swnj 	 * Otherwise this is an acceptable SYN segment
1955065Swnj 	 *	initialize tp->rcv_nxt and tp->irs
1965065Swnj 	 *	if seg contains ack then advance tp->snd_una
1975065Swnj 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
1985065Swnj 	 *	arrange for segment to be acked (eventually)
1995065Swnj 	 *	continue processing rest of data/controls, beginning with URG
2005065Swnj 	 */
2015065Swnj 	case TCPS_SYN_SENT:
2025065Swnj 		if ((tiflags & TH_ACK) &&
2035065Swnj 		    (SEQ_LEQ(ti->ti_ack, tp->iss) ||
2045231Swnj 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
2055085Swnj 			goto dropwithreset;
2065065Swnj 		if (tiflags & TH_RST) {
2075065Swnj 			if (tiflags & TH_ACK)
2085085Swnj 				tcp_drop(tp, ECONNRESET);
2095065Swnj 			goto drop;
2104601Swnj 		}
2115065Swnj 		if ((tiflags & TH_SYN) == 0)
2125065Swnj 			goto drop;
2135231Swnj 		tp->snd_una = ti->ti_ack;
214*5244Sroot 		tp->t_timer[TCPT_REXMT] = 0;
2155065Swnj 		tp->irs = ti->ti_seq;
2165085Swnj 		tcp_rcvseqinit(tp);
2175085Swnj 		tp->t_flags |= TF_ACKNOW;
2185162Swnj 		if (SEQ_GT(tp->snd_una, tp->iss)) {
219*5244Sroot 			soisconnected(so);
2205065Swnj 			tp->t_state = TCPS_ESTABLISHED;
2215162Swnj 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
2225231Swnj 			tp->snd_wl1 = ti->ti_seq;
2235162Swnj 		} else
2245085Swnj 			tp->t_state = TCPS_SYN_RECEIVED;
225*5244Sroot printf("tcp_input: out of SYN_SENT: ");
2265085Swnj 		goto trimthenstep6;
2275085Swnj 
2285085Swnj trimthenstep6:
2295085Swnj 		/*
2305231Swnj 		 * Advance ti->ti_seq to correspond to first data byte.
2315085Swnj 		 * If data, trim to stay within window,
2325085Swnj 		 * dropping FIN if necessary.
2335085Swnj 		 */
2345231Swnj 		ti->ti_seq++;
2355085Swnj 		if (ti->ti_len > tp->rcv_wnd) {
2365085Swnj 			todrop = ti->ti_len - tp->rcv_wnd;
2375085Swnj 			m_adj(m, -todrop);
2385085Swnj 			ti->ti_len = tp->rcv_wnd;
2395085Swnj 			ti->ti_flags &= ~TH_FIN;
2405065Swnj 		}
241*5244Sroot printf("ti->ti_len %d\n", ti->ti_len);
242*5244Sroot pseqno(tp);
2435085Swnj 		goto step6;
2445065Swnj 	}
2454601Swnj 
2465065Swnj 	/*
2475065Swnj 	 * States other than LISTEN or SYN_SENT.
2485065Swnj 	 * First check that at least some bytes of segment are within
2495065Swnj 	 * receive window.
2505065Swnj 	 */
2515065Swnj 	if (tp->rcv_wnd == 0) {
2525065Swnj 		/*
2535065Swnj 		 * If window is closed can only take segments at
2545231Swnj 		 * window edge, and have to drop data and PUSH from
2555065Swnj 		 * incoming segments.
2565065Swnj 		 */
2575065Swnj 		if (tp->rcv_nxt != ti->ti_seq)
2585065Swnj 			goto dropafterack;
2595085Swnj 		if (ti->ti_len > 0) {
2605085Swnj 			ti->ti_len = 0;
2615085Swnj 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
2625065Swnj 		}
263*5244Sroot printf("tcp_input %x: window 0, drop text and FIN\n", tp);
2645065Swnj 	} else {
2655065Swnj 		/*
2665231Swnj 		 * If segment begins before rcv_nxt, drop leading
2675065Swnj 		 * data (and SYN); if nothing left, just ack.
2685065Swnj 		 */
2695065Swnj 		if (SEQ_GT(tp->rcv_nxt, ti->ti_seq)) {
2705085Swnj 			todrop = tp->rcv_nxt - ti->ti_seq;
271*5244Sroot printf("tcp_input %x: drop %d dup bytes\n", tp, todrop);
2725085Swnj 			if (tiflags & TH_SYN) {
2735085Swnj 				ti->ti_seq++;
2745085Swnj 				if (ti->ti_urp > 1)
2755085Swnj 					ti->ti_urp--;
2765085Swnj 				else
2775085Swnj 					tiflags &= ~TH_URG;
2785085Swnj 				todrop--;
2795085Swnj 			}
2805065Swnj 			if (todrop > ti->ti_len)
2815065Swnj 				goto dropafterack;
2825065Swnj 			m_adj(m, todrop);
2835065Swnj 			ti->ti_seq += todrop;
2845065Swnj 			ti->ti_len -= todrop;
2855085Swnj 			if (ti->ti_urp > todrop)
2865085Swnj 				ti->ti_urp -= todrop;
2875085Swnj 			else {
2885085Swnj 				tiflags &= ~TH_URG;
2895085Swnj 				/* ti->ti_flags &= ~TH_URG; */
2905085Swnj 				/* ti->ti_urp = 0; */
2915085Swnj 			}
2925085Swnj 			/* tiflags &= ~TH_SYN; */
2935085Swnj 			/* ti->ti_flags &= ~TH_SYN; */
2945065Swnj 		}
2955065Swnj 		/*
2965065Swnj 		 * If segment ends after window, drop trailing data
2975085Swnj 		 * (and PUSH and FIN); if nothing left, just ACK.
2985065Swnj 		 */
2995065Swnj 		if (SEQ_GT(ti->ti_seq+ti->ti_len, tp->rcv_nxt+tp->rcv_wnd)) {
3005085Swnj 			todrop =
3015065Swnj 			     ti->ti_seq+ti->ti_len - (tp->rcv_nxt+tp->rcv_wnd);
3025065Swnj 			if (todrop > ti->ti_len)
3035065Swnj 				goto dropafterack;
3045065Swnj 			m_adj(m, -todrop);
3055065Swnj 			ti->ti_len -= todrop;
3065085Swnj 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
3075065Swnj 		}
3085065Swnj 	}
3094601Swnj 
3105065Swnj 	/*
3115065Swnj 	 * If the RST bit is set examine the state:
3125065Swnj 	 *    SYN_RECEIVED STATE:
3135065Swnj 	 *	If passive open, return to LISTEN state.
3145065Swnj 	 *	If active open, inform user that connection was refused.
3155065Swnj 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
3165065Swnj 	 *	Inform user that connection was reset, and close tcb.
3175065Swnj 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
3185065Swnj 	 *	Close the tcb.
3195065Swnj 	 */
3205065Swnj 	if (tiflags&TH_RST) switch (tp->t_state) {
3215065Swnj 
3225065Swnj 	case TCPS_SYN_RECEIVED:
3235065Swnj 		if (inp->inp_socket->so_options & SO_ACCEPTCONN) {
3245085Swnj 			tp->t_state = TCPS_LISTEN;
325*5244Sroot 			tp->t_timer[TCPT_KEEP] = 0;
326*5244Sroot 			(void) m_free(dtom(tp->t_template));
327*5244Sroot 			tp->t_template = 0;
3285231Swnj 			in_pcbdisconnect(inp);
3295065Swnj 			goto drop;
3304601Swnj 		}
3315085Swnj 		tcp_drop(tp, ECONNREFUSED);
3325065Swnj 		goto drop;
3334601Swnj 
3345065Swnj 	case TCPS_ESTABLISHED:
3355065Swnj 	case TCPS_FIN_WAIT_1:
3365065Swnj 	case TCPS_FIN_WAIT_2:
3375065Swnj 	case TCPS_CLOSE_WAIT:
3385065Swnj 		tcp_drop(tp, ECONNRESET);
3395065Swnj 		goto drop;
3405065Swnj 
3415065Swnj 	case TCPS_CLOSING:
3425065Swnj 	case TCPS_LAST_ACK:
3435065Swnj 	case TCPS_TIME_WAIT:
3445065Swnj 		tcp_close(tp);
3455065Swnj 		goto drop;
3464601Swnj 	}
3474601Swnj 
3484601Swnj 	/*
3495065Swnj 	 * If a SYN is in the window, then this is an
3505065Swnj 	 * error and we send an RST and drop the connection.
3514601Swnj 	 */
3525065Swnj 	if (tiflags & TH_SYN) {
3535231Swnj 		tcp_drop(tp, ECONNRESET);
3545085Swnj 		goto dropwithreset;
3554601Swnj 	}
3564601Swnj 
3574601Swnj 	/*
3585065Swnj 	 * If the ACK bit is off we drop the segment and return.
3594601Swnj 	 */
3605085Swnj 	if ((tiflags & TH_ACK) == 0)
3615065Swnj 		goto drop;
3625065Swnj 
3635065Swnj 	/*
3645065Swnj 	 * Ack processing.
3655065Swnj 	 */
3664601Swnj 	switch (tp->t_state) {
3674601Swnj 
3685065Swnj 	/*
3695065Swnj 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
3705065Swnj 	 * ESTABLISHED state and continue processing, othewise
3715065Swnj 	 * send an RST.
3725065Swnj 	 */
3735065Swnj 	case TCPS_SYN_RECEIVED:
3745085Swnj 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
3755231Swnj 		    SEQ_GT(ti->ti_ack, tp->snd_max))
3765085Swnj 			goto dropwithreset;
377*5244Sroot 		tp->snd_una++;			/* SYN acked */
378*5244Sroot 		tp->t_timer[TCPT_REXMT] = 0;
3795085Swnj 		soisconnected(so);
3805085Swnj 		tp->t_state = TCPS_ESTABLISHED;
3815162Swnj 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
382*5244Sroot 		tp->snd_wl1 = ti->ti_seq - 1;
383*5244Sroot printf("tcp_input: to ESTAB:\n"); pseqno(tp);
3845085Swnj 		/* fall into ... */
3854601Swnj 
3865065Swnj 	/*
3875065Swnj 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
3885065Swnj 	 * ACKs.  If the ack is in the range
3895231Swnj 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
3905065Swnj 	 * then advance tp->snd_una to ti->ti_ack and drop
3915065Swnj 	 * data from the retransmission queue.  If this ACK reflects
3925065Swnj 	 * more up to date window information we update our window information.
3935065Swnj 	 */
3945065Swnj 	case TCPS_ESTABLISHED:
3955065Swnj 	case TCPS_FIN_WAIT_1:
3965065Swnj 	case TCPS_FIN_WAIT_2:
3975065Swnj 	case TCPS_CLOSE_WAIT:
3985065Swnj 	case TCPS_CLOSING:
399*5244Sroot 	case TCPS_LAST_ACK:
400*5244Sroot 	case TCPS_TIME_WAIT:
4015085Swnj #define	ourfinisacked	(acked > 0)
4025085Swnj 
403*5244Sroot 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
4045065Swnj 			break;
4055231Swnj 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
4065065Swnj 			goto dropafterack;
4075085Swnj 		acked = ti->ti_ack - tp->snd_una;
408*5244Sroot printf("tcp_input: got ack of %d bytes\n", acked);
409*5244Sroot 		if (acked >= so->so_snd.sb_cc) {
4105085Swnj 			acked -= so->so_snd.sb_cc;
411*5244Sroot 			/* if acked > 0 our FIN is acked */
412*5244Sroot 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
413*5244Sroot 			tp->t_timer[TCPT_REXMT] = 0;
4145085Swnj 		} else {
4155085Swnj 			sbdrop(&so->so_snd, acked);
4165085Swnj 			acked = 0;
417*5244Sroot 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
418*5244Sroot 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
4195085Swnj 		}
4205231Swnj 		tp->snd_una = ti->ti_ack;
4215162Swnj 
4225162Swnj 		/*
4235162Swnj 		 * If transmit timer is running and timed sequence
4245162Swnj 		 * number was acked, update smoothed round trip time.
4255162Swnj 		 */
4265162Swnj 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
427*5244Sroot 			if (tp->t_srtt == 0)
428*5244Sroot 				tp->t_srtt = tp->t_rtt;
429*5244Sroot 			else
430*5244Sroot 				tp->t_srtt =
431*5244Sroot 				    tcp_alpha * tp->t_srtt +
432*5244Sroot 				    (1 - tcp_alpha) * tp->t_rtt;
433*5244Sroot printf("tcp_input: rtt sampled %d, srtt now %d\n", tp->t_rtt, (int)(100* tp->t_srtt));
4345162Swnj 			tp->t_rtt = 0;
4355162Swnj 		}
4365162Swnj 
4374601Swnj 		switch (tp->t_state) {
4384601Swnj 
4395065Swnj 		/*
4405065Swnj 		 * In FIN_WAIT_1 STATE in addition to the processing
4415065Swnj 		 * for the ESTABLISHED state if our FIN is now acknowledged
4425085Swnj 		 * then enter FIN_WAIT_2.
4435065Swnj 		 */
4445065Swnj 		case TCPS_FIN_WAIT_1:
4455085Swnj 			if (ourfinisacked)
4465085Swnj 				tp->t_state = TCPS_FIN_WAIT_2;
4474601Swnj 			break;
4484601Swnj 
4495065Swnj 	 	/*
4505065Swnj 		 * In CLOSING STATE in addition to the processing for
4515065Swnj 		 * the ESTABLISHED state if the ACK acknowledges our FIN
4525065Swnj 		 * then enter the TIME-WAIT state, otherwise ignore
4535065Swnj 		 * the segment.
4545065Swnj 		 */
4555065Swnj 		case TCPS_CLOSING:
456*5244Sroot 			if (ourfinisacked) {
4575065Swnj 				tp->t_state = TCPS_TIME_WAIT;
458*5244Sroot 				tcp_canceltimers(tp);
459*5244Sroot 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
460*5244Sroot 				soisdisconnected(so);
461*5244Sroot 			}
462*5244Sroot 			break;
4634601Swnj 
4645065Swnj 		/*
4655085Swnj 		 * The only thing that can arrive in  LAST_ACK state
4665085Swnj 		 * is an acknowledgment of our FIN.  If our FIN is now
4675085Swnj 		 * acknowledged, delete the TCB, enter the closed state
4685085Swnj 		 * and return.
4695065Swnj 		 */
4705065Swnj 		case TCPS_LAST_ACK:
471*5244Sroot 			if (ourfinisacked) {
472*5244Sroot printf("tcp_input: LAST ACK close\n");
4735065Swnj 				tcp_close(tp);
474*5244Sroot 			}
4755065Swnj 			goto drop;
4764601Swnj 
4775065Swnj 		/*
4785065Swnj 		 * In TIME_WAIT state the only thing that should arrive
4795065Swnj 		 * is a retransmission of the remote FIN.  Acknowledge
4805065Swnj 		 * it and restart the finack timer.
4815065Swnj 		 */
4825065Swnj 		case TCPS_TIME_WAIT:
483*5244Sroot printf("tcp_input: TIME_WAIT restart timer\n");
4845162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
4855065Swnj 			goto dropafterack;
4864601Swnj 		}
4875085Swnj #undef ourfinisacked
4885085Swnj 	}
4894601Swnj 
4905065Swnj step6:
4915065Swnj 	/*
492*5244Sroot 	 * Update window information.
493*5244Sroot 	 */
494*5244Sroot printf("update win wl1 %x ti->ti_seq %x wl2 %x?", tp->snd_wl1, ti->ti_seq, tp->snd_wl2);
495*5244Sroot 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
496*5244Sroot 	    tp->snd_wl1==ti->ti_seq && SEQ_LEQ(tp->snd_wl2,ti->ti_seq)) {
497*5244Sroot printf("yes\n");
498*5244Sroot 		tp->snd_wnd = ti->ti_win;
499*5244Sroot 		tp->snd_wl1 = ti->ti_seq;
500*5244Sroot 		tp->snd_wl2 = ti->ti_ack;
501*5244Sroot 		if (tp->snd_wnd > 0)
502*5244Sroot 			tp->t_timer[TCPT_PERSIST] = 0;
503*5244Sroot 	}
504*5244Sroot else printf("no\n");
505*5244Sroot 
506*5244Sroot 	/*
5075065Swnj 	 * If an URG bit is set in the segment and is greater than the
5085065Swnj 	 * current known urgent pointer, then signal the user that the
509*5244Sroot 	 * remote side has out of band data.  This should not happen
5105065Swnj 	 * in CLOSE_WAIT, CLOSING, LAST-ACK or TIME_WAIT STATES since
5115065Swnj 	 * a FIN has been received from the remote side.  In these states
5125065Swnj 	 * we ignore the URG.
5135065Swnj 	 */
5145085Swnj 	if ((tiflags & TH_URG) == 0 && TCPS_HAVERCVDFIN(tp->t_state) == 0)
5155085Swnj 		if (SEQ_GT(ti->ti_urp, tp->rcv_up)) {
5165065Swnj 			tp->rcv_up = ti->ti_urp;
5175085Swnj #if 0
518*5244Sroot 			sohasoutofband(so);		/* XXX */
5195085Swnj #endif
5204601Swnj 		}
5214601Swnj 
5224601Swnj 	/*
5235065Swnj 	 * Process the segment text, merging it into the TCP sequencing queue,
5245065Swnj 	 * and arranging for acknowledgment of receipt if necessary.
5255065Swnj 	 * This process logically involves adjusting tp->rcv_wnd as data
5265065Swnj 	 * is presented to the user (this happens in tcp_usrreq.c,
5275065Swnj 	 * case PRU_RCVD).  If a FIN has already been received on this
5285065Swnj 	 * connection then we just ignore the text.
5294601Swnj 	 */
530*5244Sroot 	if (ti->ti_len && TCPS_HAVERCVDFIN(tp->t_state) == 0) {
5315085Swnj 		off += sizeof (struct ip);		/* drop IP header */
5325085Swnj 		m->m_off += off;
5335085Swnj 		m->m_len -= off;
5345065Swnj 		tiflags = tcp_reass(tp, ti);
5355085Swnj 		tp->t_flags |= TF_ACKNOW;		/* XXX TF_DELACK */
536*5244Sroot 	} else {
5374924Swnj 		m_freem(m);
538*5244Sroot 	}
5394601Swnj 
5404601Swnj 	/*
5415065Swnj 	 * If FIN is received then if we haven't received SYN and
5425065Swnj 	 * therefore can't validate drop the segment.  Otherwise ACK
5435065Swnj 	 * the FIN and let the user know that the connection is closing.
5444601Swnj 	 */
5455085Swnj 	if ((tiflags & TH_FIN)) {
5465074Swnj 		if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
5475074Swnj 			goto drop;
548*5244Sroot 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
549*5244Sroot 			socantrcvmore(so);
550*5244Sroot 			tp->t_flags |= TF_ACKNOW;
551*5244Sroot 			tp->rcv_nxt++;
552*5244Sroot 		}
553*5244Sroot printf("tcp_input: %x got FIN\n", tp);
5545065Swnj 		switch (tp->t_state) {
5554601Swnj 
5565065Swnj 	 	/*
5575065Swnj 		 * In SYN_RECEIVED and ESTABLISHED STATES
5585065Swnj 		 * enter the CLOSE_WAIT state.
5594884Swnj 		 */
5605065Swnj 		case TCPS_SYN_RECEIVED:
5615065Swnj 		case TCPS_ESTABLISHED:
5625065Swnj 			tp->t_state = TCPS_CLOSE_WAIT;
5635065Swnj 			break;
5644884Swnj 
5655065Swnj 	 	/*
5665085Swnj 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
5675085Swnj 		 * enter the CLOSING state.
5684884Swnj 		 */
5695065Swnj 		case TCPS_FIN_WAIT_1:
5705085Swnj 			tp->t_state = TCPS_CLOSING;
5715065Swnj 			break;
5724601Swnj 
5735065Swnj 	 	/*
5745065Swnj 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
5755065Swnj 		 * starting the time-wait timer, turning off the other
5765065Swnj 		 * standard timers.
5775065Swnj 		 */
5785065Swnj 		case TCPS_FIN_WAIT_2:
579*5244Sroot 			tp->t_state = TCPS_TIME_WAIT;
5805074Swnj 			tcp_canceltimers(tp);
5815162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
582*5244Sroot 			soisdisconnected(so);
5835065Swnj 			break;
5845065Swnj 
5854884Swnj 		/*
5865065Swnj 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
5874884Swnj 		 */
5885065Swnj 		case TCPS_TIME_WAIT:
5895162Swnj 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
5905065Swnj 			break;
5915085Swnj 		}
5924601Swnj 	}
5935085Swnj 
5945085Swnj 	/*
5955085Swnj 	 * Return any desired output.
5965085Swnj 	 */
5975085Swnj 	tcp_output(tp);
5985065Swnj 	return;
5995085Swnj 
6005065Swnj dropafterack:
6015085Swnj 	/*
602*5244Sroot 	 * Generate an ACK dropping incoming segment.
6035085Swnj 	 * Make ACK reflect our state.
6045085Swnj 	 */
605*5244Sroot printf("tcp_input: dropafterack\n");
6065085Swnj 	if (tiflags & TH_RST)
6075085Swnj 		goto drop;
6085085Swnj 	tcp_respond(ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
6095231Swnj 	return;
6105085Swnj 
6115085Swnj dropwithreset:
612*5244Sroot printf("tcp_input: dropwithreset\n");
6135085Swnj 	/*
614*5244Sroot 	 * Generate a RST, dropping incoming segment.
6155085Swnj 	 * Make ACK acceptable to originator of segment.
6165085Swnj 	 */
6175085Swnj 	if (tiflags & TH_RST)
6185085Swnj 		goto drop;
6195085Swnj 	if (tiflags & TH_ACK)
6205109Swnj 		tcp_respond(ti, (tcp_seq)0, ti->ti_ack, TH_RST);
6215085Swnj 	else {
6225085Swnj 		if (tiflags & TH_SYN)
6235085Swnj 			ti->ti_len++;
6245109Swnj 		tcp_respond(ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK);
6255085Swnj 	}
6265231Swnj 	return;
6275085Swnj 
6285065Swnj drop:
629*5244Sroot printf("tcp_input: drop\n");
6305085Swnj 	/*
6315085Swnj 	 * Drop space held by incoming segment and return.
6325085Swnj 	 */
6335065Swnj 	m_freem(m);
6345065Swnj }
6355065Swnj 
6365065Swnj /*
6375065Swnj  * Insert segment ti into reassembly queue of tcp with
6385065Swnj  * control block tp.  Return TH_FIN if reassembly now includes
6395065Swnj  * a segment with FIN.
6405065Swnj  */
6415109Swnj tcp_reass(tp, ti)
6425065Swnj 	register struct tcpcb *tp;
6435065Swnj 	register struct tcpiphdr *ti;
6445065Swnj {
6455065Swnj 	register struct tcpiphdr *q;
6465085Swnj 	struct socket *so = tp->t_inpcb->inp_socket;
6475065Swnj 	int flags = 0;		/* no FIN */
6485085Swnj COUNT(TCP_REASS);
6495065Swnj 
6505065Swnj 	/*
6515162Swnj 	 * Call with ti==0 after become established to
6525162Swnj 	 * force pre-ESTABLISHED data up to user socket.
6535065Swnj 	 */
6545162Swnj 	if (ti == 0)
6555065Swnj 		goto present;
6564601Swnj 
6575065Swnj 	/*
6585065Swnj 	 * Find a segment which begins after this one does.
6595065Swnj 	 */
6605065Swnj 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
6615065Swnj 	    q = (struct tcpiphdr *)q->ti_next)
6625065Swnj 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
6635065Swnj 			break;
6644601Swnj 
6655065Swnj 	/*
6665065Swnj 	 * If there is a preceding segment, it may provide some of
6675065Swnj 	 * our data already.  If so, drop the data from the incoming
6685065Swnj 	 * segment.  If it provides all of our data, drop us.
6695065Swnj 	 */
6705065Swnj 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
6715065Swnj 		register int i;
6725065Swnj 		q = (struct tcpiphdr *)(q->ti_prev);
6735065Swnj 		/* conversion to int (in i) handles seq wraparound */
6745065Swnj 		i = q->ti_seq + q->ti_len - ti->ti_seq;
6755065Swnj 		if (i > 0) {
6764924Swnj 			if (i >= ti->ti_len)
6775065Swnj 				goto drop;
6785065Swnj 			m_adj(dtom(tp), i);
6795065Swnj 			ti->ti_len -= i;
6804924Swnj 			ti->ti_seq += i;
6814601Swnj 		}
6825065Swnj 		q = (struct tcpiphdr *)(q->ti_next);
6835065Swnj 	}
6844601Swnj 
6855065Swnj 	/*
6865065Swnj 	 * While we overlap succeeding segments trim them or,
6875065Swnj 	 * if they are completely covered, dequeue them.
6885065Swnj 	 */
6895065Swnj 	while (q != (struct tcpiphdr *)tp &&
6905065Swnj 	    SEQ_GT(ti->ti_seq + ti->ti_len, q->ti_seq)) {
6915065Swnj 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
6925065Swnj 		if (i < q->ti_len) {
6935065Swnj 			q->ti_len -= i;
6945065Swnj 			m_adj(dtom(q), i);
6955065Swnj 			break;
6964601Swnj 		}
6975065Swnj 		q = (struct tcpiphdr *)q->ti_next;
6985065Swnj 		m_freem(dtom(q->ti_prev));
6995065Swnj 		remque(q->ti_prev);
7005065Swnj 	}
7014601Swnj 
7025065Swnj 	/*
7035065Swnj 	 * Stick new segment in its place.
7045065Swnj 	 */
7055065Swnj 	insque(ti, q->ti_prev);
7064601Swnj 
7075065Swnj present:
7085065Swnj 	/*
709*5244Sroot 	 * Present data to user, advancing rcv_nxt through
710*5244Sroot 	 * completed sequence space.
7115065Swnj 	 */
7125085Swnj 	if (tp->t_state < TCPS_ESTABLISHED)
713*5244Sroot 		return (0);
7144924Swnj 	ti = tp->seg_next;
715*5244Sroot 	while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt) {
716*5244Sroot 		tp->rcv_nxt += ti->ti_len;
717*5244Sroot 		flags = ti->ti_flags & TH_FIN;
718*5244Sroot printf("move %d bytes to user; rcv_nxt now %x\n", ti->ti_len, tp->rcv_nxt);
7194924Swnj 		remque(ti);
7204924Swnj 		sbappend(&so->so_rcv, dtom(ti));
7214924Swnj 		ti = (struct tcpiphdr *)ti->ti_next;
7224601Swnj 	}
7235074Swnj 	if (so->so_state & SS_CANTRCVMORE)
7245074Swnj 		sbflush(&so->so_rcv);
7255074Swnj 	else
7265074Swnj 		sorwakeup(so);
7275065Swnj 	return (flags);
7285065Swnj drop:
729*5244Sroot printf("tcp_reass drop\n");
7305065Swnj 	m_freem(dtom(ti));
7315065Swnj 	return (flags);
7324601Swnj }
733