xref: /csrg-svn/sys/netinet/tcp_input.c (revision 12441)
1 /*	tcp_input.c	1.94	83/05/14	*/
2 
3 #include "../h/param.h"
4 #include "../h/systm.h"
5 #include "../h/mbuf.h"
6 #include "../h/protosw.h"
7 #include "../h/socket.h"
8 #include "../h/socketvar.h"
9 #include "../h/errno.h"
10 
11 #include "../net/if.h"
12 #include "../net/route.h"
13 
14 #include "../netinet/in.h"
15 #include "../netinet/in_pcb.h"
16 #include "../netinet/in_systm.h"
17 #include "../netinet/ip.h"
18 #include "../netinet/ip_var.h"
19 #include "../netinet/tcp.h"
20 #include "../netinet/tcp_fsm.h"
21 #include "../netinet/tcp_seq.h"
22 #include "../netinet/tcp_timer.h"
23 #include "../netinet/tcp_var.h"
24 #include "../netinet/tcpip.h"
25 #include "../netinet/tcp_debug.h"
26 
27 int	tcpprintfs = 0;
28 int	tcpcksum = 1;
29 struct	tcpiphdr tcp_saveti;
30 extern	tcpnodelack;
31 
32 struct	tcpcb *tcp_newtcpcb();
33 /*
34  * TCP input routine, follows pages 65-76 of the
35  * protocol specification dated September, 1981 very closely.
36  */
37 tcp_input(m0)
38 	struct mbuf *m0;
39 {
40 	register struct tcpiphdr *ti;
41 	struct inpcb *inp;
42 	register struct mbuf *m;
43 	struct mbuf *om = 0;
44 	int len, tlen, off;
45 	register struct tcpcb *tp = 0;
46 	register int tiflags;
47 	struct socket *so;
48 	int todrop, acked;
49 	short ostate;
50 	struct in_addr laddr;
51 	int dropsocket = 0;
52 
53 	/*
54 	 * Get IP and TCP header together in first mbuf.
55 	 * Note: IP leaves IP header in first mbuf.
56 	 */
57 	m = m0;
58 	ti = mtod(m, struct tcpiphdr *);
59 	if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2))
60 		ip_stripoptions((struct ip *)ti, (struct mbuf *)0);
61 	if (m->m_off > MMAXOFF || m->m_len < sizeof (struct tcpiphdr)) {
62 		if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) {
63 			tcpstat.tcps_hdrops++;
64 			return;
65 		}
66 		ti = mtod(m, struct tcpiphdr *);
67 	}
68 
69 	/*
70 	 * Checksum extended TCP header and data.
71 	 */
72 	tlen = ((struct ip *)ti)->ip_len;
73 	len = sizeof (struct ip) + tlen;
74 	if (tcpcksum) {
75 		ti->ti_next = ti->ti_prev = 0;
76 		ti->ti_x1 = 0;
77 		ti->ti_len = (u_short)tlen;
78 		ti->ti_len = htons((u_short)ti->ti_len);
79 		if (ti->ti_sum = in_cksum(m, len)) {
80 			if (tcpprintfs)
81 				printf("tcp sum: src %x\n", ti->ti_src);
82 			tcpstat.tcps_badsum++;
83 			goto drop;
84 		}
85 	}
86 
87 	/*
88 	 * Check that TCP offset makes sense,
89 	 * pull out TCP options and adjust length.
90 	 */
91 	off = ti->ti_off << 2;
92 	if (off < sizeof (struct tcphdr) || off > tlen) {
93 		if (tcpprintfs)
94 			printf("tcp off: src %x off %d\n", ti->ti_src, off);
95 		tcpstat.tcps_badoff++;
96 		goto drop;
97 	}
98 	tlen -= off;
99 	ti->ti_len = tlen;
100 	if (off > sizeof (struct tcphdr)) {
101 		if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
102 			tcpstat.tcps_hdrops++;
103 			return;
104 		}
105 		ti = mtod(m, struct tcpiphdr *);
106 		om = m_get(M_DONTWAIT, MT_DATA);
107 		if (om == 0)
108 			goto drop;
109 		om->m_len = off - sizeof (struct tcphdr);
110 		{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
111 		  bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len);
112 		  m->m_len -= om->m_len;
113 		  bcopy(op+om->m_len, op,
114 		   (unsigned)(m->m_len-sizeof (struct tcpiphdr)));
115 		}
116 	}
117 	tiflags = ti->ti_flags;
118 
119 	/*
120 	 * Drop TCP and IP headers.
121 	 */
122 	off += sizeof (struct ip);
123 	m->m_off += off;
124 	m->m_len -= off;
125 
126 	/*
127 	 * Convert TCP protocol specific fields to host format.
128 	 */
129 	ti->ti_seq = ntohl(ti->ti_seq);
130 	ti->ti_ack = ntohl(ti->ti_ack);
131 	ti->ti_win = ntohs(ti->ti_win);
132 	ti->ti_urp = ntohs(ti->ti_urp);
133 
134 	/*
135 	 * Locate pcb for segment.
136 	 */
137 	inp = in_pcblookup
138 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport,
139 		INPLOOKUP_WILDCARD);
140 
141 	/*
142 	 * If the state is CLOSED (i.e., TCB does not exist) then
143 	 * all data in the incoming segment is discarded.
144 	 */
145 	if (inp == 0)
146 		goto dropwithreset;
147 	tp = intotcpcb(inp);
148 	if (tp == 0)
149 		goto dropwithreset;
150 	so = inp->inp_socket;
151 	if (so->so_options & SO_DEBUG) {
152 		ostate = tp->t_state;
153 		tcp_saveti = *ti;
154 	}
155 	if (so->so_options & SO_ACCEPTCONN) {
156 		so = sonewconn(so);
157 		if (so == 0)
158 			goto drop;
159 		/*
160 		 * This is ugly, but ....
161 		 *
162 		 * Mark socket as temporary until we're
163 		 * committed to keeping it.  The code at
164 		 * ``drop'' and ``dropwithreset'' check the
165 		 * flag dropsocket to see if the temporary
166 		 * socket created here should be discarded.
167 		 * We mark the socket as discardable until
168 		 * we're committed to it below in TCPS_LISTEN.
169 		 */
170 		dropsocket++;
171 		inp = (struct inpcb *)so->so_pcb;
172 		inp->inp_laddr = ti->ti_dst;
173 		inp->inp_lport = ti->ti_dport;
174 		tp = intotcpcb(inp);
175 		tp->t_state = TCPS_LISTEN;
176 	}
177 
178 	/*
179 	 * Segment received on connection.
180 	 * Reset idle time and keep-alive timer.
181 	 */
182 	tp->t_idle = 0;
183 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
184 
185 	/*
186 	 * Process options.
187 	 */
188 	if (om) {
189 		tcp_dooptions(tp, om);
190 		om = 0;
191 	}
192 
193 	/*
194 	 * Calculate amount of space in receive window,
195 	 * and then do TCP input processing.
196 	 */
197 	tp->rcv_wnd = sbspace(&so->so_rcv);
198 	if (tp->rcv_wnd < 0)
199 		tp->rcv_wnd = 0;
200 
201 	switch (tp->t_state) {
202 
203 	/*
204 	 * If the state is LISTEN then ignore segment if it contains an RST.
205 	 * If the segment contains an ACK then it is bad and send a RST.
206 	 * If it does not contain a SYN then it is not interesting; drop it.
207 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
208 	 * tp->iss, and send a segment:
209 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
210 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
211 	 * Fill in remote peer address fields if not previously specified.
212 	 * Enter SYN_RECEIVED state, and process any other fields of this
213 	 * segment in this state.
214 	 */
215 	case TCPS_LISTEN: {
216 		struct mbuf *am;
217 		register struct sockaddr_in *sin;
218 
219 		if (tiflags & TH_RST)
220 			goto drop;
221 		if (tiflags & TH_ACK)
222 			goto dropwithreset;
223 		if ((tiflags & TH_SYN) == 0)
224 			goto drop;
225 		am = m_get(M_DONTWAIT, MT_SONAME);
226 		if (am == NULL)
227 			goto drop;
228 		am->m_len = sizeof (struct sockaddr_in);
229 		sin = mtod(am, struct sockaddr_in *);
230 		sin->sin_family = AF_INET;
231 		sin->sin_addr = ti->ti_src;
232 		sin->sin_port = ti->ti_sport;
233 		laddr = inp->inp_laddr;
234 		if (inp->inp_laddr.s_addr == INADDR_ANY)
235 			inp->inp_laddr = ti->ti_dst;
236 		if (in_pcbconnect(inp, am)) {
237 			inp->inp_laddr = laddr;
238 			(void) m_free(am);
239 			goto drop;
240 		}
241 		(void) m_free(am);
242 		tp->t_template = tcp_template(tp);
243 		if (tp->t_template == 0) {
244 			in_pcbdisconnect(inp);
245 			inp->inp_laddr = laddr;
246 			tp = 0;
247 			goto drop;
248 		}
249 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
250 		tp->irs = ti->ti_seq;
251 		tcp_sendseqinit(tp);
252 		tcp_rcvseqinit(tp);
253 		tp->t_state = TCPS_SYN_RECEIVED;
254 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
255 		dropsocket = 0;		/* committed to socket */
256 		goto trimthenstep6;
257 		}
258 
259 	/*
260 	 * If the state is SYN_SENT:
261 	 *	if seg contains an ACK, but not for our SYN, drop the input.
262 	 *	if seg contains a RST, then drop the connection.
263 	 *	if seg does not contain SYN, then drop it.
264 	 * Otherwise this is an acceptable SYN segment
265 	 *	initialize tp->rcv_nxt and tp->irs
266 	 *	if seg contains ack then advance tp->snd_una
267 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
268 	 *	arrange for segment to be acked (eventually)
269 	 *	continue processing rest of data/controls, beginning with URG
270 	 */
271 	case TCPS_SYN_SENT:
272 		if ((tiflags & TH_ACK) &&
273 /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
274 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
275 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
276 			goto dropwithreset;
277 		if (tiflags & TH_RST) {
278 			if (tiflags & TH_ACK)
279 				tp = tcp_drop(tp, ECONNREFUSED);
280 			goto drop;
281 		}
282 		if ((tiflags & TH_SYN) == 0)
283 			goto drop;
284 		tp->snd_una = ti->ti_ack;
285 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
286 			tp->snd_nxt = tp->snd_una;
287 		tp->t_timer[TCPT_REXMT] = 0;
288 		tp->irs = ti->ti_seq;
289 		tcp_rcvseqinit(tp);
290 		tp->t_flags |= TF_ACKNOW;
291 		if (SEQ_GT(tp->snd_una, tp->iss)) {
292 			soisconnected(so);
293 			tp->t_state = TCPS_ESTABLISHED;
294 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
295 		} else
296 			tp->t_state = TCPS_SYN_RECEIVED;
297 		goto trimthenstep6;
298 
299 trimthenstep6:
300 		/*
301 		 * Advance ti->ti_seq to correspond to first data byte.
302 		 * If data, trim to stay within window,
303 		 * dropping FIN if necessary.
304 		 */
305 		ti->ti_seq++;
306 		if (ti->ti_len > tp->rcv_wnd) {
307 			todrop = ti->ti_len - tp->rcv_wnd;
308 			m_adj(m, -todrop);
309 			ti->ti_len = tp->rcv_wnd;
310 			ti->ti_flags &= ~TH_FIN;
311 		}
312 		tp->snd_wl1 = ti->ti_seq - 1;
313 		goto step6;
314 	}
315 
316 	/*
317 	 * States other than LISTEN or SYN_SENT.
318 	 * First check that at least some bytes of segment are within
319 	 * receive window.
320 	 */
321 	if (tp->rcv_wnd == 0) {
322 		/*
323 		 * If window is closed can only take segments at
324 		 * window edge, and have to drop data and PUSH from
325 		 * incoming segments.
326 		 */
327 		if (tp->rcv_nxt != ti->ti_seq)
328 			goto dropafterack;
329 		if (ti->ti_len > 0) {
330 			m_adj(m, ti->ti_len);
331 			ti->ti_len = 0;
332 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
333 		}
334 	} else {
335 		/*
336 		 * If segment begins before rcv_nxt, drop leading
337 		 * data (and SYN); if nothing left, just ack.
338 		 */
339 		todrop = tp->rcv_nxt - ti->ti_seq;
340 		if (todrop > 0) {
341 			if (tiflags & TH_SYN) {
342 				tiflags &= ~TH_SYN;
343 				ti->ti_flags &= ~TH_SYN;
344 				ti->ti_seq++;
345 				if (ti->ti_urp > 1)
346 					ti->ti_urp--;
347 				else
348 					tiflags &= ~TH_URG;
349 				todrop--;
350 			}
351 			if (todrop > ti->ti_len ||
352 			    todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
353 				goto dropafterack;
354 			m_adj(m, todrop);
355 			ti->ti_seq += todrop;
356 			ti->ti_len -= todrop;
357 			if (ti->ti_urp > todrop)
358 				ti->ti_urp -= todrop;
359 			else {
360 				tiflags &= ~TH_URG;
361 				ti->ti_flags &= ~TH_URG;
362 				ti->ti_urp = 0;
363 			}
364 		}
365 		/*
366 		 * If segment ends after window, drop trailing data
367 		 * (and PUSH and FIN); if nothing left, just ACK.
368 		 */
369 		todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
370 		if (todrop > 0) {
371 			if (todrop >= ti->ti_len)
372 				goto dropafterack;
373 			m_adj(m, -todrop);
374 			ti->ti_len -= todrop;
375 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
376 		}
377 	}
378 
379 	/*
380 	 * If data is received on a connection after the
381 	 * user processes are gone, then RST the other end.
382 	 */
383 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
384 	    ti->ti_len) {
385 		tp = tcp_close(tp);
386 		goto dropwithreset;
387 	}
388 
389 	/*
390 	 * If the RST bit is set examine the state:
391 	 *    SYN_RECEIVED STATE:
392 	 *	If passive open, return to LISTEN state.
393 	 *	If active open, inform user that connection was refused.
394 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
395 	 *	Inform user that connection was reset, and close tcb.
396 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
397 	 *	Close the tcb.
398 	 */
399 	if (tiflags&TH_RST) switch (tp->t_state) {
400 
401 	case TCPS_SYN_RECEIVED:
402 		tp = tcp_drop(tp, ECONNREFUSED);
403 		goto drop;
404 
405 	case TCPS_ESTABLISHED:
406 	case TCPS_FIN_WAIT_1:
407 	case TCPS_FIN_WAIT_2:
408 	case TCPS_CLOSE_WAIT:
409 		tp = tcp_drop(tp, ECONNRESET);
410 		goto drop;
411 
412 	case TCPS_CLOSING:
413 	case TCPS_LAST_ACK:
414 	case TCPS_TIME_WAIT:
415 		tp = tcp_close(tp);
416 		goto drop;
417 	}
418 
419 	/*
420 	 * If a SYN is in the window, then this is an
421 	 * error and we send an RST and drop the connection.
422 	 */
423 	if (tiflags & TH_SYN) {
424 		tp = tcp_drop(tp, ECONNRESET);
425 		goto dropwithreset;
426 	}
427 
428 	/*
429 	 * If the ACK bit is off we drop the segment and return.
430 	 */
431 	if ((tiflags & TH_ACK) == 0)
432 		goto drop;
433 
434 	/*
435 	 * Ack processing.
436 	 */
437 	switch (tp->t_state) {
438 
439 	/*
440 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
441 	 * ESTABLISHED state and continue processing, othewise
442 	 * send an RST.
443 	 */
444 	case TCPS_SYN_RECEIVED:
445 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
446 		    SEQ_GT(ti->ti_ack, tp->snd_max))
447 			goto dropwithreset;
448 		tp->snd_una++;			/* SYN acked */
449 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
450 			tp->snd_nxt = tp->snd_una;
451 		tp->t_timer[TCPT_REXMT] = 0;
452 		soisconnected(so);
453 		tp->t_state = TCPS_ESTABLISHED;
454 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
455 		tp->snd_wl1 = ti->ti_seq - 1;
456 		/* fall into ... */
457 
458 	/*
459 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
460 	 * ACKs.  If the ack is in the range
461 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
462 	 * then advance tp->snd_una to ti->ti_ack and drop
463 	 * data from the retransmission queue.  If this ACK reflects
464 	 * more up to date window information we update our window information.
465 	 */
466 	case TCPS_ESTABLISHED:
467 	case TCPS_FIN_WAIT_1:
468 	case TCPS_FIN_WAIT_2:
469 	case TCPS_CLOSE_WAIT:
470 	case TCPS_CLOSING:
471 	case TCPS_LAST_ACK:
472 	case TCPS_TIME_WAIT:
473 #define	ourfinisacked	(acked > 0)
474 
475 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
476 			break;
477 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
478 			goto dropafterack;
479 		acked = ti->ti_ack - tp->snd_una;
480 
481 		/*
482 		 * If transmit timer is running and timed sequence
483 		 * number was acked, update smoothed round trip time.
484 		 */
485 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
486 			if (tp->t_srtt == 0)
487 				tp->t_srtt = tp->t_rtt;
488 			else
489 				tp->t_srtt =
490 				    tcp_alpha * tp->t_srtt +
491 				    (1 - tcp_alpha) * tp->t_rtt;
492 			tp->t_rtt = 0;
493 		}
494 
495 		if (ti->ti_ack == tp->snd_max)
496 			tp->t_timer[TCPT_REXMT] = 0;
497 		else {
498 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
499 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
500 			tp->t_rtt = 1;
501 			tp->t_rxtshift = 0;
502 		}
503 		if (acked > so->so_snd.sb_cc) {
504 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
505 			tp->snd_wnd -= so->so_snd.sb_cc;
506 		} else {
507 			sbdrop(&so->so_snd, acked);
508 			tp->snd_wnd -= acked;
509 			acked = 0;
510 		}
511 		if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
512 			sowwakeup(so);
513 		tp->snd_una = ti->ti_ack;
514 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
515 			tp->snd_nxt = tp->snd_una;
516 
517 		switch (tp->t_state) {
518 
519 		/*
520 		 * In FIN_WAIT_1 STATE in addition to the processing
521 		 * for the ESTABLISHED state if our FIN is now acknowledged
522 		 * then enter FIN_WAIT_2.
523 		 */
524 		case TCPS_FIN_WAIT_1:
525 			if (ourfinisacked) {
526 				/*
527 				 * If we can't receive any more
528 				 * data, then closing user can proceed.
529 				 */
530 				if (so->so_state & SS_CANTRCVMORE)
531 					soisdisconnected(so);
532 				tp->t_state = TCPS_FIN_WAIT_2;
533 			}
534 			break;
535 
536 	 	/*
537 		 * In CLOSING STATE in addition to the processing for
538 		 * the ESTABLISHED state if the ACK acknowledges our FIN
539 		 * then enter the TIME-WAIT state, otherwise ignore
540 		 * the segment.
541 		 */
542 		case TCPS_CLOSING:
543 			if (ourfinisacked) {
544 				tp->t_state = TCPS_TIME_WAIT;
545 				tcp_canceltimers(tp);
546 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
547 				soisdisconnected(so);
548 			}
549 			break;
550 
551 		/*
552 		 * The only thing that can arrive in  LAST_ACK state
553 		 * is an acknowledgment of our FIN.  If our FIN is now
554 		 * acknowledged, delete the TCB, enter the closed state
555 		 * and return.
556 		 */
557 		case TCPS_LAST_ACK:
558 			if (ourfinisacked)
559 				tp = tcp_close(tp);
560 			goto drop;
561 
562 		/*
563 		 * In TIME_WAIT state the only thing that should arrive
564 		 * is a retransmission of the remote FIN.  Acknowledge
565 		 * it and restart the finack timer.
566 		 */
567 		case TCPS_TIME_WAIT:
568 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
569 			goto dropafterack;
570 		}
571 #undef ourfinisacked
572 	}
573 
574 step6:
575 	/*
576 	 * Update window information.
577 	 */
578 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
579 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
580 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
581 		tp->snd_wnd = ti->ti_win;
582 		tp->snd_wl1 = ti->ti_seq;
583 		tp->snd_wl2 = ti->ti_ack;
584 		if (tp->snd_wnd != 0)
585 			tp->t_timer[TCPT_PERSIST] = 0;
586 	}
587 
588 	/*
589 	 * Process segments with URG.
590 	 */
591 	if ((tiflags & TH_URG) && ti->ti_urp &&
592 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
593 		/*
594 		 * This is a kludge, but the BBN C70 TCP
595 		 * randomly sends packets with the urgent flag
596 		 * set and random ti_urp values.  This crashes
597 		 * the system because so_oobmark ends up being
598 		 * interpreted as a negative number in soreceive.
599 		 */
600 		if (ti->ti_urp > tp->t_maxseg) {	/* XXX */
601 			ti->ti_urp = 0;			/* XXX */
602 			tiflags &= ~TH_URG;		/* XXX */
603 			ti->ti_flags &= ~TH_URG;	/* XXX */
604 			goto bbnurp;			/* XXX */
605 		}
606 		/*
607 		 * If this segment advances the known urgent pointer,
608 		 * then mark the data stream.  This should not happen
609 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
610 		 * a FIN has been received from the remote side.
611 		 * In these states we ignore the URG.
612 		 */
613 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
614 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
615 			so->so_oobmark = so->so_rcv.sb_cc +
616 			    (tp->rcv_up - tp->rcv_nxt) - 1;
617 			if (so->so_oobmark == 0)
618 				so->so_state |= SS_RCVATMARK;
619 			sohasoutofband(so);
620 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
621 		}
622 		/*
623 		 * Remove out of band data so doesn't get presented to user.
624 		 * This can happen independent of advancing the URG pointer,
625 		 * but if two URG's are pending at once, some out-of-band
626 		 * data may creep in... ick.
627 		 */
628 		if (ti->ti_urp <= ti->ti_len)
629 			tcp_pulloutofband(so, ti);
630 	}
631 bbnurp:							/* XXX */
632 
633 	/*
634 	 * Process the segment text, merging it into the TCP sequencing queue,
635 	 * and arranging for acknowledgment of receipt if necessary.
636 	 * This process logically involves adjusting tp->rcv_wnd as data
637 	 * is presented to the user (this happens in tcp_usrreq.c,
638 	 * case PRU_RCVD).  If a FIN has already been received on this
639 	 * connection then we just ignore the text.
640 	 */
641 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
642 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
643 		tiflags = tcp_reass(tp, ti);
644 		if (tcpnodelack == 0)
645 			tp->t_flags |= TF_DELACK;
646 		else
647 			tp->t_flags |= TF_ACKNOW;
648 	} else {
649 		m_freem(m);
650 		tiflags &= ~TH_FIN;
651 	}
652 
653 	/*
654 	 * If FIN is received ACK the FIN and let the user know
655 	 * that the connection is closing.
656 	 */
657 	if (tiflags & TH_FIN) {
658 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
659 			socantrcvmore(so);
660 			tp->t_flags |= TF_ACKNOW;
661 			tp->rcv_nxt++;
662 		}
663 		switch (tp->t_state) {
664 
665 	 	/*
666 		 * In SYN_RECEIVED and ESTABLISHED STATES
667 		 * enter the CLOSE_WAIT state.
668 		 */
669 		case TCPS_SYN_RECEIVED:
670 		case TCPS_ESTABLISHED:
671 			tp->t_state = TCPS_CLOSE_WAIT;
672 			break;
673 
674 	 	/*
675 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
676 		 * enter the CLOSING state.
677 		 */
678 		case TCPS_FIN_WAIT_1:
679 			tp->t_state = TCPS_CLOSING;
680 			break;
681 
682 	 	/*
683 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
684 		 * starting the time-wait timer, turning off the other
685 		 * standard timers.
686 		 */
687 		case TCPS_FIN_WAIT_2:
688 			tp->t_state = TCPS_TIME_WAIT;
689 			tcp_canceltimers(tp);
690 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
691 			soisdisconnected(so);
692 			break;
693 
694 		/*
695 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
696 		 */
697 		case TCPS_TIME_WAIT:
698 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
699 			break;
700 		}
701 	}
702 	if (so->so_options & SO_DEBUG)
703 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
704 
705 	/*
706 	 * Return any desired output.
707 	 */
708 	(void) tcp_output(tp);
709 	return;
710 
711 dropafterack:
712 	/*
713 	 * Generate an ACK dropping incoming segment if it occupies
714 	 * sequence space, where the ACK reflects our state.
715 	 */
716 	if ((tiflags&TH_RST) ||
717 	    tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
718 		goto drop;
719 	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
720 		tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
721 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
722 	return;
723 
724 dropwithreset:
725 	if (om) {
726 		(void) m_free(om);
727 		om = 0;
728 	}
729 	/*
730 	 * Generate a RST, dropping incoming segment.
731 	 * Make ACK acceptable to originator of segment.
732 	 */
733 	if (tiflags & TH_RST)
734 		goto drop;
735 	if (tiflags & TH_ACK)
736 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
737 	else {
738 		if (tiflags & TH_SYN)
739 			ti->ti_len++;
740 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
741 		    TH_RST|TH_ACK);
742 	}
743 	/* destroy temporarily created socket */
744 	if (dropsocket)
745 		(void) soabort(so);
746 	return;
747 
748 drop:
749 	if (om)
750 		(void) m_free(om);
751 	/*
752 	 * Drop space held by incoming segment and return.
753 	 */
754 	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
755 		tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
756 	m_freem(m);
757 	/* destroy temporarily created socket */
758 	if (dropsocket)
759 		(void) soabort(so);
760 	return;
761 }
762 
763 tcp_dooptions(tp, om)
764 	struct tcpcb *tp;
765 	struct mbuf *om;
766 {
767 	register u_char *cp;
768 	int opt, optlen, cnt;
769 
770 	cp = mtod(om, u_char *);
771 	cnt = om->m_len;
772 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
773 		opt = cp[0];
774 		if (opt == TCPOPT_EOL)
775 			break;
776 		if (opt == TCPOPT_NOP)
777 			optlen = 1;
778 		else {
779 			optlen = cp[1];
780 			if (optlen <= 0)
781 				break;
782 		}
783 		switch (opt) {
784 
785 		default:
786 			break;
787 
788 		case TCPOPT_MAXSEG:
789 			if (optlen != 4)
790 				continue;
791 			tp->t_maxseg = *(u_short *)(cp + 2);
792 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
793 			break;
794 		}
795 	}
796 	(void) m_free(om);
797 }
798 
799 /*
800  * Pull out of band byte out of a segment so
801  * it doesn't appear in the user's data queue.
802  * It is still reflected in the segment length for
803  * sequencing purposes.
804  */
805 tcp_pulloutofband(so, ti)
806 	struct socket *so;
807 	struct tcpiphdr *ti;
808 {
809 	register struct mbuf *m;
810 	int cnt = ti->ti_urp - 1;
811 
812 	m = dtom(ti);
813 	while (cnt >= 0) {
814 		if (m->m_len > cnt) {
815 			char *cp = mtod(m, caddr_t) + cnt;
816 			struct tcpcb *tp = sototcpcb(so);
817 
818 			tp->t_iobc = *cp;
819 			tp->t_oobflags |= TCPOOB_HAVEDATA;
820 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
821 			m->m_len--;
822 			return;
823 		}
824 		cnt -= m->m_len;
825 		m = m->m_next;
826 		if (m == 0)
827 			break;
828 	}
829 	panic("tcp_pulloutofband");
830 }
831 
832 /*
833  * Insert segment ti into reassembly queue of tcp with
834  * control block tp.  Return TH_FIN if reassembly now includes
835  * a segment with FIN.
836  */
837 tcp_reass(tp, ti)
838 	register struct tcpcb *tp;
839 	register struct tcpiphdr *ti;
840 {
841 	register struct tcpiphdr *q;
842 	struct socket *so = tp->t_inpcb->inp_socket;
843 	struct mbuf *m;
844 	int flags;
845 
846 	/*
847 	 * Call with ti==0 after become established to
848 	 * force pre-ESTABLISHED data up to user socket.
849 	 */
850 	if (ti == 0)
851 		goto present;
852 
853 	/*
854 	 * Find a segment which begins after this one does.
855 	 */
856 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
857 	    q = (struct tcpiphdr *)q->ti_next)
858 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
859 			break;
860 
861 	/*
862 	 * If there is a preceding segment, it may provide some of
863 	 * our data already.  If so, drop the data from the incoming
864 	 * segment.  If it provides all of our data, drop us.
865 	 */
866 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
867 		register int i;
868 		q = (struct tcpiphdr *)q->ti_prev;
869 		/* conversion to int (in i) handles seq wraparound */
870 		i = q->ti_seq + q->ti_len - ti->ti_seq;
871 		if (i > 0) {
872 			if (i >= ti->ti_len)
873 				goto drop;
874 			m_adj(dtom(ti), i);
875 			ti->ti_len -= i;
876 			ti->ti_seq += i;
877 		}
878 		q = (struct tcpiphdr *)(q->ti_next);
879 	}
880 
881 	/*
882 	 * While we overlap succeeding segments trim them or,
883 	 * if they are completely covered, dequeue them.
884 	 */
885 	while (q != (struct tcpiphdr *)tp) {
886 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
887 		if (i <= 0)
888 			break;
889 		if (i < q->ti_len) {
890 			q->ti_seq += i;
891 			q->ti_len -= i;
892 			m_adj(dtom(q), i);
893 			break;
894 		}
895 		q = (struct tcpiphdr *)q->ti_next;
896 		m = dtom(q->ti_prev);
897 		remque(q->ti_prev);
898 		m_freem(m);
899 	}
900 
901 	/*
902 	 * Stick new segment in its place.
903 	 */
904 	insque(ti, q->ti_prev);
905 
906 present:
907 	/*
908 	 * Present data to user, advancing rcv_nxt through
909 	 * completed sequence space.
910 	 */
911 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
912 		return (0);
913 	ti = tp->seg_next;
914 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
915 		return (0);
916 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
917 		return (0);
918 	do {
919 		tp->rcv_nxt += ti->ti_len;
920 		flags = ti->ti_flags & TH_FIN;
921 		remque(ti);
922 		m = dtom(ti);
923 		ti = (struct tcpiphdr *)ti->ti_next;
924 		if (so->so_state & SS_CANTRCVMORE)
925 			m_freem(m);
926 		else
927 			sbappend(&so->so_rcv, m);
928 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
929 	sorwakeup(so);
930 	return (flags);
931 drop:
932 	m_freem(dtom(ti));
933 	return (0);
934 }
935