xref: /original-bsd/sys/netinet/tcp_input.c (revision 950ddd82)
1 /*	tcp_input.c	1.89	83/02/10	*/
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 			tcpstat.tcps_badsum++;
81 			if (tcpprintfs)
82 				printf("tcp cksum %x\n", ti->ti_sum);
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 		tcpstat.tcps_badoff++;
94 		goto drop;
95 	}
96 	tlen -= off;
97 	ti->ti_len = tlen;
98 	if (off > sizeof (struct tcphdr)) {
99 		if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
100 			tcpstat.tcps_hdrops++;
101 			goto drop;
102 		}
103 		ti = mtod(m, struct tcpiphdr *);
104 		om = m_get(M_DONTWAIT, MT_DATA);
105 		if (om == 0)
106 			goto drop;
107 		om->m_len = off - sizeof (struct tcphdr);
108 		{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
109 		  bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len);
110 		  m->m_len -= om->m_len;
111 		  bcopy(op+om->m_len, op,
112 		   (unsigned)(m->m_len-sizeof (struct tcpiphdr)));
113 		}
114 	}
115 	tiflags = ti->ti_flags;
116 
117 	/*
118 	 * Drop TCP and IP headers.
119 	 */
120 	off += sizeof (struct ip);
121 	m->m_off += off;
122 	m->m_len -= off;
123 
124 	/*
125 	 * Convert TCP protocol specific fields to host format.
126 	 */
127 	ti->ti_seq = ntohl(ti->ti_seq);
128 	ti->ti_ack = ntohl(ti->ti_ack);
129 	ti->ti_win = ntohs(ti->ti_win);
130 	ti->ti_urp = ntohs(ti->ti_urp);
131 
132 	/*
133 	 * Locate pcb for segment.
134 	 */
135 	inp = in_pcblookup
136 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport,
137 		INPLOOKUP_WILDCARD);
138 
139 	/*
140 	 * If the state is CLOSED (i.e., TCB does not exist) then
141 	 * all data in the incoming segment is discarded.
142 	 */
143 	if (inp == 0)
144 		goto dropwithreset;
145 	tp = intotcpcb(inp);
146 	if (tp == 0)
147 		goto dropwithreset;
148 	so = inp->inp_socket;
149 	if (so->so_options & SO_DEBUG) {
150 		ostate = tp->t_state;
151 		tcp_saveti = *ti;
152 	}
153 	if (so->so_options & SO_ACCEPTCONN) {
154 		so = sonewconn(so);
155 		if (so == 0)
156 			goto drop;
157 		/*
158 		 * This is ugly, but ....
159 		 *
160 		 * Mark socket as temporary until we're
161 		 * committed to keeping it.  The code at
162 		 * ``drop'' and ``dropwithreset'' check the
163 		 * flag dropsocket to see if the temporary
164 		 * socket created here should be discarded.
165 		 * We mark the socket as discardable until
166 		 * we're committed to it below in TCPS_LISTEN.
167 		 */
168 		dropsocket++;
169 		inp = (struct inpcb *)so->so_pcb;
170 		inp->inp_laddr = ti->ti_dst;
171 		inp->inp_lport = ti->ti_dport;
172 		tp = intotcpcb(inp);
173 		tp->t_state = TCPS_LISTEN;
174 	}
175 
176 	/*
177 	 * Segment received on connection.
178 	 * Reset idle time and keep-alive timer.
179 	 */
180 	tp->t_idle = 0;
181 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
182 
183 	/*
184 	 * Process options.
185 	 */
186 	if (om) {
187 		tcp_dooptions(tp, om);
188 		om = 0;
189 	}
190 
191 	/*
192 	 * Calculate amount of space in receive window,
193 	 * and then do TCP input processing.
194 	 */
195 	tp->rcv_wnd = sbspace(&so->so_rcv);
196 	if (tp->rcv_wnd < 0)
197 		tp->rcv_wnd = 0;
198 
199 	switch (tp->t_state) {
200 
201 	/*
202 	 * If the state is LISTEN then ignore segment if it contains an RST.
203 	 * If the segment contains an ACK then it is bad and send a RST.
204 	 * If it does not contain a SYN then it is not interesting; drop it.
205 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
206 	 * tp->iss, and send a segment:
207 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
208 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
209 	 * Fill in remote peer address fields if not previously specified.
210 	 * Enter SYN_RECEIVED state, and process any other fields of this
211 	 * segment in this state.
212 	 */
213 	case TCPS_LISTEN: {
214 		struct mbuf *am;
215 		register struct sockaddr_in *sin;
216 
217 		if (tiflags & TH_RST)
218 			goto drop;
219 		if (tiflags & TH_ACK)
220 			goto dropwithreset;
221 		if ((tiflags & TH_SYN) == 0)
222 			goto drop;
223 		am = m_get(M_DONTWAIT, MT_SONAME);
224 		if (am == NULL)
225 			goto drop;
226 		am->m_len = sizeof (struct sockaddr_in);
227 		sin = mtod(am, struct sockaddr_in *);
228 		sin->sin_family = AF_INET;
229 		sin->sin_addr = ti->ti_src;
230 		sin->sin_port = ti->ti_sport;
231 		laddr = inp->inp_laddr;
232 		if (inp->inp_laddr.s_addr == INADDR_ANY)
233 			inp->inp_laddr = ti->ti_dst;
234 		if (in_pcbconnect(inp, am)) {
235 			inp->inp_laddr = laddr;
236 			(void) m_free(am);
237 			goto drop;
238 		}
239 		(void) m_free(am);
240 		tp->t_template = tcp_template(tp);
241 		if (tp->t_template == 0) {
242 			in_pcbdisconnect(inp);
243 			inp->inp_laddr = laddr;
244 			tp = 0;
245 			goto drop;
246 		}
247 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
248 		tp->irs = ti->ti_seq;
249 		tcp_sendseqinit(tp);
250 		tcp_rcvseqinit(tp);
251 		tp->t_state = TCPS_SYN_RECEIVED;
252 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
253 		dropsocket = 0;		/* committed to socket */
254 		goto trimthenstep6;
255 		}
256 
257 	/*
258 	 * If the state is SYN_SENT:
259 	 *	if seg contains an ACK, but not for our SYN, drop the input.
260 	 *	if seg contains a RST, then drop the connection.
261 	 *	if seg does not contain SYN, then drop it.
262 	 * Otherwise this is an acceptable SYN segment
263 	 *	initialize tp->rcv_nxt and tp->irs
264 	 *	if seg contains ack then advance tp->snd_una
265 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
266 	 *	arrange for segment to be acked (eventually)
267 	 *	continue processing rest of data/controls, beginning with URG
268 	 */
269 	case TCPS_SYN_SENT:
270 		if ((tiflags & TH_ACK) &&
271 /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
272 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
273 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
274 			goto dropwithreset;
275 		if (tiflags & TH_RST) {
276 			if (tiflags & TH_ACK)
277 				tp = tcp_drop(tp, ECONNREFUSED);
278 			goto drop;
279 		}
280 		if ((tiflags & TH_SYN) == 0)
281 			goto drop;
282 		tp->snd_una = ti->ti_ack;
283 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
284 			tp->snd_nxt = tp->snd_una;
285 		tp->t_timer[TCPT_REXMT] = 0;
286 		tp->irs = ti->ti_seq;
287 		tcp_rcvseqinit(tp);
288 		tp->t_flags |= TF_ACKNOW;
289 		if (SEQ_GT(tp->snd_una, tp->iss)) {
290 			soisconnected(so);
291 			tp->t_state = TCPS_ESTABLISHED;
292 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
293 		} else
294 			tp->t_state = TCPS_SYN_RECEIVED;
295 		goto trimthenstep6;
296 
297 trimthenstep6:
298 		/*
299 		 * Advance ti->ti_seq to correspond to first data byte.
300 		 * If data, trim to stay within window,
301 		 * dropping FIN if necessary.
302 		 */
303 		ti->ti_seq++;
304 		if (ti->ti_len > tp->rcv_wnd) {
305 			todrop = ti->ti_len - tp->rcv_wnd;
306 			m_adj(m, -todrop);
307 			ti->ti_len = tp->rcv_wnd;
308 			ti->ti_flags &= ~TH_FIN;
309 		}
310 		tp->snd_wl1 = ti->ti_seq - 1;
311 		goto step6;
312 	}
313 
314 	/*
315 	 * States other than LISTEN or SYN_SENT.
316 	 * First check that at least some bytes of segment are within
317 	 * receive window.
318 	 */
319 	if (tp->rcv_wnd == 0) {
320 		/*
321 		 * If window is closed can only take segments at
322 		 * window edge, and have to drop data and PUSH from
323 		 * incoming segments.
324 		 */
325 		if (tp->rcv_nxt != ti->ti_seq)
326 			goto dropafterack;
327 		if (ti->ti_len > 0) {
328 			m_adj(m, ti->ti_len);
329 			ti->ti_len = 0;
330 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
331 		}
332 	} else {
333 		/*
334 		 * If segment begins before rcv_nxt, drop leading
335 		 * data (and SYN); if nothing left, just ack.
336 		 */
337 		todrop = tp->rcv_nxt - ti->ti_seq;
338 		if (todrop > 0) {
339 			if (tiflags & TH_SYN) {
340 				tiflags &= ~TH_SYN;
341 				ti->ti_flags &= ~TH_SYN;
342 				ti->ti_seq++;
343 				if (ti->ti_urp > 1)
344 					ti->ti_urp--;
345 				else
346 					tiflags &= ~TH_URG;
347 				todrop--;
348 			}
349 			if (todrop > ti->ti_len ||
350 			    todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
351 				goto dropafterack;
352 			m_adj(m, todrop);
353 			ti->ti_seq += todrop;
354 			ti->ti_len -= todrop;
355 			if (ti->ti_urp > todrop)
356 				ti->ti_urp -= todrop;
357 			else {
358 				tiflags &= ~TH_URG;
359 				ti->ti_flags &= ~TH_URG;
360 				ti->ti_urp = 0;
361 			}
362 		}
363 		/*
364 		 * If segment ends after window, drop trailing data
365 		 * (and PUSH and FIN); if nothing left, just ACK.
366 		 */
367 		todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
368 		if (todrop > 0) {
369 			if (todrop >= ti->ti_len)
370 				goto dropafterack;
371 			m_adj(m, -todrop);
372 			ti->ti_len -= todrop;
373 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
374 		}
375 	}
376 
377 	/*
378 	 * If data is received on a connection after the
379 	 * user processes are gone, then RST the other end.
380 	 */
381 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
382 	    ti->ti_len) {
383 		tp = tcp_close(tp);
384 		goto dropwithreset;
385 	}
386 
387 	/*
388 	 * If the RST bit is set examine the state:
389 	 *    SYN_RECEIVED STATE:
390 	 *	If passive open, return to LISTEN state.
391 	 *	If active open, inform user that connection was refused.
392 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
393 	 *	Inform user that connection was reset, and close tcb.
394 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
395 	 *	Close the tcb.
396 	 */
397 	if (tiflags&TH_RST) switch (tp->t_state) {
398 
399 	case TCPS_SYN_RECEIVED:
400 		tp = tcp_drop(tp, ECONNREFUSED);
401 		goto drop;
402 
403 	case TCPS_ESTABLISHED:
404 	case TCPS_FIN_WAIT_1:
405 	case TCPS_FIN_WAIT_2:
406 	case TCPS_CLOSE_WAIT:
407 		tp = tcp_drop(tp, ECONNRESET);
408 		goto drop;
409 
410 	case TCPS_CLOSING:
411 	case TCPS_LAST_ACK:
412 	case TCPS_TIME_WAIT:
413 		tp = tcp_close(tp);
414 		goto drop;
415 	}
416 
417 	/*
418 	 * If a SYN is in the window, then this is an
419 	 * error and we send an RST and drop the connection.
420 	 */
421 	if (tiflags & TH_SYN) {
422 		tp = tcp_drop(tp, ECONNRESET);
423 		goto dropwithreset;
424 	}
425 
426 	/*
427 	 * If the ACK bit is off we drop the segment and return.
428 	 */
429 	if ((tiflags & TH_ACK) == 0)
430 		goto drop;
431 
432 	/*
433 	 * Ack processing.
434 	 */
435 	switch (tp->t_state) {
436 
437 	/*
438 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
439 	 * ESTABLISHED state and continue processing, othewise
440 	 * send an RST.
441 	 */
442 	case TCPS_SYN_RECEIVED:
443 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
444 		    SEQ_GT(ti->ti_ack, tp->snd_max))
445 			goto dropwithreset;
446 		tp->snd_una++;			/* SYN acked */
447 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
448 			tp->snd_nxt = tp->snd_una;
449 		tp->t_timer[TCPT_REXMT] = 0;
450 		soisconnected(so);
451 		tp->t_state = TCPS_ESTABLISHED;
452 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
453 		tp->snd_wl1 = ti->ti_seq - 1;
454 		/* fall into ... */
455 
456 	/*
457 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
458 	 * ACKs.  If the ack is in the range
459 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
460 	 * then advance tp->snd_una to ti->ti_ack and drop
461 	 * data from the retransmission queue.  If this ACK reflects
462 	 * more up to date window information we update our window information.
463 	 */
464 	case TCPS_ESTABLISHED:
465 	case TCPS_FIN_WAIT_1:
466 	case TCPS_FIN_WAIT_2:
467 	case TCPS_CLOSE_WAIT:
468 	case TCPS_CLOSING:
469 	case TCPS_LAST_ACK:
470 	case TCPS_TIME_WAIT:
471 #define	ourfinisacked	(acked > 0)
472 
473 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
474 			break;
475 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
476 			goto dropafterack;
477 		acked = ti->ti_ack - tp->snd_una;
478 
479 		/*
480 		 * If transmit timer is running and timed sequence
481 		 * number was acked, update smoothed round trip time.
482 		 */
483 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
484 			if (tp->t_srtt == 0)
485 				tp->t_srtt = tp->t_rtt;
486 			else
487 				tp->t_srtt =
488 				    tcp_alpha * tp->t_srtt +
489 				    (1 - tcp_alpha) * tp->t_rtt;
490 			tp->t_rtt = 0;
491 		}
492 
493 		if (ti->ti_ack == tp->snd_max)
494 			tp->t_timer[TCPT_REXMT] = 0;
495 		else {
496 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
497 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
498 			tp->t_rtt = 1;
499 			tp->t_rxtshift = 0;
500 		}
501 		if (acked > so->so_snd.sb_cc) {
502 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
503 			tp->snd_wnd -= so->so_snd.sb_cc;
504 		} else {
505 			sbdrop(&so->so_snd, acked);
506 			tp->snd_wnd -= acked;
507 			acked = 0;
508 		}
509 		if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
510 			sowwakeup(so);
511 		tp->snd_una = ti->ti_ack;
512 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
513 			tp->snd_nxt = tp->snd_una;
514 
515 		switch (tp->t_state) {
516 
517 		/*
518 		 * In FIN_WAIT_1 STATE in addition to the processing
519 		 * for the ESTABLISHED state if our FIN is now acknowledged
520 		 * then enter FIN_WAIT_2.
521 		 */
522 		case TCPS_FIN_WAIT_1:
523 			if (ourfinisacked) {
524 				/*
525 				 * If we can't receive any more
526 				 * data, then closing user can proceed.
527 				 */
528 				if (so->so_state & SS_CANTRCVMORE)
529 					soisdisconnected(so);
530 				tp->t_state = TCPS_FIN_WAIT_2;
531 			}
532 			break;
533 
534 	 	/*
535 		 * In CLOSING STATE in addition to the processing for
536 		 * the ESTABLISHED state if the ACK acknowledges our FIN
537 		 * then enter the TIME-WAIT state, otherwise ignore
538 		 * the segment.
539 		 */
540 		case TCPS_CLOSING:
541 			if (ourfinisacked) {
542 				tp->t_state = TCPS_TIME_WAIT;
543 				tcp_canceltimers(tp);
544 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
545 				soisdisconnected(so);
546 			}
547 			break;
548 
549 		/*
550 		 * The only thing that can arrive in  LAST_ACK state
551 		 * is an acknowledgment of our FIN.  If our FIN is now
552 		 * acknowledged, delete the TCB, enter the closed state
553 		 * and return.
554 		 */
555 		case TCPS_LAST_ACK:
556 			if (ourfinisacked)
557 				tp = tcp_close(tp);
558 			goto drop;
559 
560 		/*
561 		 * In TIME_WAIT state the only thing that should arrive
562 		 * is a retransmission of the remote FIN.  Acknowledge
563 		 * it and restart the finack timer.
564 		 */
565 		case TCPS_TIME_WAIT:
566 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
567 			goto dropafterack;
568 		}
569 #undef ourfinisacked
570 	}
571 
572 step6:
573 	/*
574 	 * Update window information.
575 	 */
576 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
577 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
578 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
579 		tp->snd_wnd = ti->ti_win;
580 		tp->snd_wl1 = ti->ti_seq;
581 		tp->snd_wl2 = ti->ti_ack;
582 		if (tp->snd_wnd != 0)
583 			tp->t_timer[TCPT_PERSIST] = 0;
584 	}
585 
586 	/*
587 	 * Process segments with URG.
588 	 */
589 	if ((tiflags & TH_URG) && ti->ti_urp &&
590 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
591 		/*
592 		 * If this segment advances the known urgent pointer,
593 		 * then mark the data stream.  This should not happen
594 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
595 		 * a FIN has been received from the remote side.
596 		 * In these states we ignore the URG.
597 		 */
598 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
599 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
600 			so->so_oobmark = so->so_rcv.sb_cc +
601 			    (tp->rcv_up - tp->rcv_nxt) - 1;
602 			if (so->so_oobmark == 0)
603 				so->so_state |= SS_RCVATMARK;
604 			sohasoutofband(so);
605 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
606 		}
607 		/*
608 		 * Remove out of band data so doesn't get presented to user.
609 		 * This can happen independent of advancing the URG pointer,
610 		 * but if two URG's are pending at once, some out-of-band
611 		 * data may creep in... ick.
612 		 */
613 		if (ti->ti_urp <= ti->ti_len)
614 			tcp_pulloutofband(so, ti);
615 	}
616 
617 	/*
618 	 * Process the segment text, merging it into the TCP sequencing queue,
619 	 * and arranging for acknowledgment of receipt if necessary.
620 	 * This process logically involves adjusting tp->rcv_wnd as data
621 	 * is presented to the user (this happens in tcp_usrreq.c,
622 	 * case PRU_RCVD).  If a FIN has already been received on this
623 	 * connection then we just ignore the text.
624 	 */
625 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
626 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
627 		tiflags = tcp_reass(tp, ti);
628 		if (tcpnodelack == 0)
629 			tp->t_flags |= TF_DELACK;
630 		else
631 			tp->t_flags |= TF_ACKNOW;
632 	} else {
633 		m_freem(m);
634 		tiflags &= ~TH_FIN;
635 	}
636 
637 	/*
638 	 * If FIN is received ACK the FIN and let the user know
639 	 * that the connection is closing.
640 	 */
641 	if (tiflags & TH_FIN) {
642 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
643 			socantrcvmore(so);
644 			tp->t_flags |= TF_ACKNOW;
645 			tp->rcv_nxt++;
646 		}
647 		switch (tp->t_state) {
648 
649 	 	/*
650 		 * In SYN_RECEIVED and ESTABLISHED STATES
651 		 * enter the CLOSE_WAIT state.
652 		 */
653 		case TCPS_SYN_RECEIVED:
654 		case TCPS_ESTABLISHED:
655 			tp->t_state = TCPS_CLOSE_WAIT;
656 			break;
657 
658 	 	/*
659 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
660 		 * enter the CLOSING state.
661 		 */
662 		case TCPS_FIN_WAIT_1:
663 			tp->t_state = TCPS_CLOSING;
664 			break;
665 
666 	 	/*
667 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
668 		 * starting the time-wait timer, turning off the other
669 		 * standard timers.
670 		 */
671 		case TCPS_FIN_WAIT_2:
672 			tp->t_state = TCPS_TIME_WAIT;
673 			tcp_canceltimers(tp);
674 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
675 			soisdisconnected(so);
676 			break;
677 
678 		/*
679 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
680 		 */
681 		case TCPS_TIME_WAIT:
682 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
683 			break;
684 		}
685 	}
686 	if (so->so_options & SO_DEBUG)
687 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
688 
689 	/*
690 	 * Return any desired output.
691 	 */
692 	(void) tcp_output(tp);
693 	return;
694 
695 dropafterack:
696 	/*
697 	 * Generate an ACK dropping incoming segment if it occupies
698 	 * sequence space, where the ACK reflects our state.
699 	 */
700 	if ((tiflags&TH_RST) ||
701 	    tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
702 		goto drop;
703 	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
704 		tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
705 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
706 	return;
707 
708 dropwithreset:
709 	if (om)
710 		(void) m_free(om);
711 	/*
712 	 * Generate a RST, dropping incoming segment.
713 	 * Make ACK acceptable to originator of segment.
714 	 */
715 	if (tiflags & TH_RST)
716 		goto drop;
717 	if (tiflags & TH_ACK)
718 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
719 	else {
720 		if (tiflags & TH_SYN)
721 			ti->ti_len++;
722 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
723 		    TH_RST|TH_ACK);
724 	}
725 	/* destroy temporarily created socket */
726 	if (dropsocket)
727 		(void) soabort(so);
728 	return;
729 
730 drop:
731 	/*
732 	 * Drop space held by incoming segment and return.
733 	 */
734 	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
735 		tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
736 	m_freem(m);
737 	/* destroy temporarily created socket */
738 	if (dropsocket)
739 		(void) soabort(so);
740 	return;
741 }
742 
743 tcp_dooptions(tp, om)
744 	struct tcpcb *tp;
745 	struct mbuf *om;
746 {
747 	register u_char *cp;
748 	int opt, optlen, cnt;
749 
750 	cp = mtod(om, u_char *);
751 	cnt = om->m_len;
752 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
753 		opt = cp[0];
754 		if (opt == TCPOPT_EOL)
755 			break;
756 		if (opt == TCPOPT_NOP)
757 			optlen = 1;
758 		else
759 			optlen = cp[1];
760 		switch (opt) {
761 
762 		default:
763 			break;
764 
765 		case TCPOPT_MAXSEG:
766 			if (optlen != 4)
767 				continue;
768 			tp->t_maxseg = *(u_short *)(cp + 2);
769 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
770 			break;
771 		}
772 	}
773 	(void) m_free(om);
774 }
775 
776 /*
777  * Pull out of band byte out of a segment so
778  * it doesn't appear in the user's data queue.
779  * It is still reflected in the segment length for
780  * sequencing purposes.
781  */
782 tcp_pulloutofband(so, ti)
783 	struct socket *so;
784 	struct tcpiphdr *ti;
785 {
786 	register struct mbuf *m;
787 	int cnt = ti->ti_urp - 1;
788 
789 	m = dtom(ti);
790 	while (cnt >= 0) {
791 		if (m->m_len > cnt) {
792 			char *cp = mtod(m, caddr_t) + cnt;
793 			struct tcpcb *tp = sototcpcb(so);
794 
795 			tp->t_iobc = *cp;
796 			tp->t_oobflags |= TCPOOB_HAVEDATA;
797 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
798 			m->m_len--;
799 			return;
800 		}
801 		cnt -= m->m_len;
802 		m = m->m_next;
803 		if (m == 0)
804 			break;
805 	}
806 	panic("tcp_pulloutofband");
807 }
808 
809 /*
810  * Insert segment ti into reassembly queue of tcp with
811  * control block tp.  Return TH_FIN if reassembly now includes
812  * a segment with FIN.
813  */
814 tcp_reass(tp, ti)
815 	register struct tcpcb *tp;
816 	register struct tcpiphdr *ti;
817 {
818 	register struct tcpiphdr *q;
819 	struct socket *so = tp->t_inpcb->inp_socket;
820 	struct mbuf *m;
821 	int flags;
822 
823 	/*
824 	 * Call with ti==0 after become established to
825 	 * force pre-ESTABLISHED data up to user socket.
826 	 */
827 	if (ti == 0)
828 		goto present;
829 
830 	/*
831 	 * Find a segment which begins after this one does.
832 	 */
833 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
834 	    q = (struct tcpiphdr *)q->ti_next)
835 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
836 			break;
837 
838 	/*
839 	 * If there is a preceding segment, it may provide some of
840 	 * our data already.  If so, drop the data from the incoming
841 	 * segment.  If it provides all of our data, drop us.
842 	 */
843 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
844 		register int i;
845 		q = (struct tcpiphdr *)q->ti_prev;
846 		/* conversion to int (in i) handles seq wraparound */
847 		i = q->ti_seq + q->ti_len - ti->ti_seq;
848 		if (i > 0) {
849 			if (i >= ti->ti_len)
850 				goto drop;
851 			m_adj(dtom(ti), i);
852 			ti->ti_len -= i;
853 			ti->ti_seq += i;
854 		}
855 		q = (struct tcpiphdr *)(q->ti_next);
856 	}
857 
858 	/*
859 	 * While we overlap succeeding segments trim them or,
860 	 * if they are completely covered, dequeue them.
861 	 */
862 	while (q != (struct tcpiphdr *)tp) {
863 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
864 		if (i <= 0)
865 			break;
866 		if (i < q->ti_len) {
867 			q->ti_seq += i;
868 			q->ti_len -= i;
869 			m_adj(dtom(q), i);
870 			break;
871 		}
872 		q = (struct tcpiphdr *)q->ti_next;
873 		m = dtom(q->ti_prev);
874 		remque(q->ti_prev);
875 		m_freem(m);
876 	}
877 
878 	/*
879 	 * Stick new segment in its place.
880 	 */
881 	insque(ti, q->ti_prev);
882 
883 present:
884 	/*
885 	 * Present data to user, advancing rcv_nxt through
886 	 * completed sequence space.
887 	 */
888 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
889 		return (0);
890 	ti = tp->seg_next;
891 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
892 		return (0);
893 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
894 		return (0);
895 	do {
896 		tp->rcv_nxt += ti->ti_len;
897 		flags = ti->ti_flags & TH_FIN;
898 		remque(ti);
899 		m = dtom(ti);
900 		ti = (struct tcpiphdr *)ti->ti_next;
901 		if (so->so_state & SS_CANTRCVMORE)
902 			m_freem(m);
903 		else
904 			sbappend(&so->so_rcv, m);
905 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
906 	sorwakeup(so);
907 	return (flags);
908 drop:
909 	m_freem(dtom(ti));
910 	return (0);
911 }
912