xref: /csrg-svn/sys/netinet/tcp_input.c (revision 17316)
1 /*	tcp_input.c	6.7	84/11/01	*/
2 
3 #include "param.h"
4 #include "systm.h"
5 #include "mbuf.h"
6 #include "protosw.h"
7 #include "socket.h"
8 #include "socketvar.h"
9 #include "errno.h"
10 
11 #include "../net/if.h"
12 #include "../net/route.h"
13 
14 #include "in.h"
15 #include "in_pcb.h"
16 #include "in_systm.h"
17 #include "ip.h"
18 #include "ip_var.h"
19 #include "tcp.h"
20 #include "tcp_fsm.h"
21 #include "tcp_seq.h"
22 #include "tcp_timer.h"
23 #include "tcp_var.h"
24 #include "tcpip.h"
25 #include "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 if not in LISTEN state,
187 	 * else do it below (after getting remote address).
188 	 */
189 	if (om && tp->t_state != TCPS_LISTEN) {
190 		tcp_dooptions(tp, om, ti);
191 		om = 0;
192 	}
193 
194 	/*
195 	 * Calculate amount of space in receive window,
196 	 * and then do TCP input processing.
197 	 */
198 	tp->rcv_wnd = sbspace(&so->so_rcv);
199 	if (tp->rcv_wnd < 0)
200 		tp->rcv_wnd = 0;
201 
202 	switch (tp->t_state) {
203 
204 	/*
205 	 * If the state is LISTEN then ignore segment if it contains an RST.
206 	 * If the segment contains an ACK then it is bad and send a RST.
207 	 * If it does not contain a SYN then it is not interesting; drop it.
208 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
209 	 * tp->iss, and send a segment:
210 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
211 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
212 	 * Fill in remote peer address fields if not previously specified.
213 	 * Enter SYN_RECEIVED state, and process any other fields of this
214 	 * segment in this state.
215 	 */
216 	case TCPS_LISTEN: {
217 		struct mbuf *am;
218 		register struct sockaddr_in *sin;
219 
220 		if (tiflags & TH_RST)
221 			goto drop;
222 		if (tiflags & TH_ACK)
223 			goto dropwithreset;
224 		if ((tiflags & TH_SYN) == 0)
225 			goto drop;
226 		am = m_get(M_DONTWAIT, MT_SONAME);
227 		if (am == NULL)
228 			goto drop;
229 		am->m_len = sizeof (struct sockaddr_in);
230 		sin = mtod(am, struct sockaddr_in *);
231 		sin->sin_family = AF_INET;
232 		sin->sin_addr = ti->ti_src;
233 		sin->sin_port = ti->ti_sport;
234 		laddr = inp->inp_laddr;
235 		if (inp->inp_laddr.s_addr == INADDR_ANY)
236 			inp->inp_laddr = ti->ti_dst;
237 		if (in_pcbconnect(inp, am)) {
238 			inp->inp_laddr = laddr;
239 			(void) m_free(am);
240 			goto drop;
241 		}
242 		(void) m_free(am);
243 		tp->t_template = tcp_template(tp);
244 		if (tp->t_template == 0) {
245 			in_pcbdisconnect(inp);
246 			dropsocket = 0;		/* socket is already gone */
247 			inp->inp_laddr = laddr;
248 			tp = 0;
249 			goto drop;
250 		}
251 		if (om) {
252 			tcp_dooptions(tp, om, ti);
253 			om = 0;
254 		}
255 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
256 		tp->irs = ti->ti_seq;
257 		tcp_sendseqinit(tp);
258 		tcp_rcvseqinit(tp);
259 		tp->t_state = TCPS_SYN_RECEIVED;
260 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
261 		dropsocket = 0;		/* committed to socket */
262 		goto trimthenstep6;
263 		}
264 
265 	/*
266 	 * If the state is SYN_SENT:
267 	 *	if seg contains an ACK, but not for our SYN, drop the input.
268 	 *	if seg contains a RST, then drop the connection.
269 	 *	if seg does not contain SYN, then drop it.
270 	 * Otherwise this is an acceptable SYN segment
271 	 *	initialize tp->rcv_nxt and tp->irs
272 	 *	if seg contains ack then advance tp->snd_una
273 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
274 	 *	arrange for segment to be acked (eventually)
275 	 *	continue processing rest of data/controls, beginning with URG
276 	 */
277 	case TCPS_SYN_SENT:
278 		if ((tiflags & TH_ACK) &&
279 /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
280 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
281 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
282 			goto dropwithreset;
283 		if (tiflags & TH_RST) {
284 			if (tiflags & TH_ACK)
285 				tp = tcp_drop(tp, ECONNREFUSED);
286 			goto drop;
287 		}
288 		if ((tiflags & TH_SYN) == 0)
289 			goto drop;
290 		tp->snd_una = ti->ti_ack;
291 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
292 			tp->snd_nxt = tp->snd_una;
293 		tp->t_timer[TCPT_REXMT] = 0;
294 		tp->irs = ti->ti_seq;
295 		tcp_rcvseqinit(tp);
296 		tp->t_flags |= TF_ACKNOW;
297 		if (SEQ_GT(tp->snd_una, tp->iss)) {
298 			soisconnected(so);
299 			tp->t_state = TCPS_ESTABLISHED;
300 			tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
301 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
302 		} else
303 			tp->t_state = TCPS_SYN_RECEIVED;
304 		goto trimthenstep6;
305 
306 trimthenstep6:
307 		/*
308 		 * Advance ti->ti_seq to correspond to first data byte.
309 		 * If data, trim to stay within window,
310 		 * dropping FIN if necessary.
311 		 */
312 		ti->ti_seq++;
313 		if (ti->ti_len > tp->rcv_wnd) {
314 			todrop = ti->ti_len - tp->rcv_wnd;
315 			m_adj(m, -todrop);
316 			ti->ti_len = tp->rcv_wnd;
317 			ti->ti_flags &= ~TH_FIN;
318 		}
319 		tp->snd_wl1 = ti->ti_seq - 1;
320 		goto step6;
321 	}
322 
323 	/*
324 	 * If data is received on a connection after the
325 	 * user processes are gone, then RST the other end.
326 	 */
327 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
328 	    ti->ti_len) {
329 		tp = tcp_close(tp);
330 		goto dropwithreset;
331 	}
332 
333 	/*
334 	 * States other than LISTEN or SYN_SENT.
335 	 * First check that at least some bytes of segment are within
336 	 * receive window.
337 	 */
338 	if (tp->rcv_wnd == 0) {
339 		/*
340 		 * If window is closed can only take segments at
341 		 * window edge, and have to drop data and PUSH from
342 		 * incoming segments.
343 		 */
344 		if (tp->rcv_nxt != ti->ti_seq)
345 			goto dropafterack;
346 		if (ti->ti_len > 0) {
347 			m_adj(m, ti->ti_len);
348 			ti->ti_len = 0;
349 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
350 		}
351 	} else {
352 		/*
353 		 * If segment begins before rcv_nxt, drop leading
354 		 * data (and SYN); if nothing left, just ack.
355 		 */
356 		todrop = tp->rcv_nxt - ti->ti_seq;
357 		if (todrop > 0) {
358 			if (tiflags & TH_SYN) {
359 				tiflags &= ~TH_SYN;
360 				ti->ti_flags &= ~TH_SYN;
361 				ti->ti_seq++;
362 				if (ti->ti_urp > 1)
363 					ti->ti_urp--;
364 				else
365 					tiflags &= ~TH_URG;
366 				todrop--;
367 			}
368 			if (todrop > ti->ti_len ||
369 			    todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
370 				goto dropafterack;
371 			m_adj(m, todrop);
372 			ti->ti_seq += todrop;
373 			ti->ti_len -= todrop;
374 			if (ti->ti_urp > todrop)
375 				ti->ti_urp -= todrop;
376 			else {
377 				tiflags &= ~TH_URG;
378 				ti->ti_flags &= ~TH_URG;
379 				ti->ti_urp = 0;
380 			}
381 		}
382 		/*
383 		 * If segment ends after window, drop trailing data
384 		 * (and PUSH and FIN); if nothing left, just ACK.
385 		 */
386 		todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
387 		if (todrop > 0) {
388 			if (todrop >= ti->ti_len)
389 				goto dropafterack;
390 			m_adj(m, -todrop);
391 			ti->ti_len -= todrop;
392 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
393 		}
394 	}
395 
396 	/*
397 	 * If the RST bit is set examine the state:
398 	 *    SYN_RECEIVED STATE:
399 	 *	If passive open, return to LISTEN state.
400 	 *	If active open, inform user that connection was refused.
401 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
402 	 *	Inform user that connection was reset, and close tcb.
403 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
404 	 *	Close the tcb.
405 	 */
406 	if (tiflags&TH_RST) switch (tp->t_state) {
407 
408 	case TCPS_SYN_RECEIVED:
409 		tp = tcp_drop(tp, ECONNREFUSED);
410 		goto drop;
411 
412 	case TCPS_ESTABLISHED:
413 	case TCPS_FIN_WAIT_1:
414 	case TCPS_FIN_WAIT_2:
415 	case TCPS_CLOSE_WAIT:
416 		tp = tcp_drop(tp, ECONNRESET);
417 		goto drop;
418 
419 	case TCPS_CLOSING:
420 	case TCPS_LAST_ACK:
421 	case TCPS_TIME_WAIT:
422 		tp = tcp_close(tp);
423 		goto drop;
424 	}
425 
426 	/*
427 	 * If a SYN is in the window, then this is an
428 	 * error and we send an RST and drop the connection.
429 	 */
430 	if (tiflags & TH_SYN) {
431 		tp = tcp_drop(tp, ECONNRESET);
432 		goto dropwithreset;
433 	}
434 
435 	/*
436 	 * If the ACK bit is off we drop the segment and return.
437 	 */
438 	if ((tiflags & TH_ACK) == 0)
439 		goto drop;
440 
441 	/*
442 	 * Ack processing.
443 	 */
444 	switch (tp->t_state) {
445 
446 	/*
447 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
448 	 * ESTABLISHED state and continue processing, othewise
449 	 * send an RST.
450 	 */
451 	case TCPS_SYN_RECEIVED:
452 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
453 		    SEQ_GT(ti->ti_ack, tp->snd_max))
454 			goto dropwithreset;
455 		tp->snd_una++;			/* SYN acked */
456 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
457 			tp->snd_nxt = tp->snd_una;
458 		tp->t_timer[TCPT_REXMT] = 0;
459 		soisconnected(so);
460 		tp->t_state = TCPS_ESTABLISHED;
461 		tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
462 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
463 		tp->snd_wl1 = ti->ti_seq - 1;
464 		/* fall into ... */
465 
466 	/*
467 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
468 	 * ACKs.  If the ack is in the range
469 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
470 	 * then advance tp->snd_una to ti->ti_ack and drop
471 	 * data from the retransmission queue.  If this ACK reflects
472 	 * more up to date window information we update our window information.
473 	 */
474 	case TCPS_ESTABLISHED:
475 	case TCPS_FIN_WAIT_1:
476 	case TCPS_FIN_WAIT_2:
477 	case TCPS_CLOSE_WAIT:
478 	case TCPS_CLOSING:
479 	case TCPS_LAST_ACK:
480 	case TCPS_TIME_WAIT:
481 #define	ourfinisacked	(acked > 0)
482 
483 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
484 			break;
485 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
486 			goto dropafterack;
487 		acked = ti->ti_ack - tp->snd_una;
488 
489 		/*
490 		 * If transmit timer is running and timed sequence
491 		 * number was acked, update smoothed round trip time.
492 		 */
493 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
494 			if (tp->t_srtt == 0)
495 				tp->t_srtt = tp->t_rtt;
496 			else
497 				tp->t_srtt =
498 				    tcp_alpha * tp->t_srtt +
499 				    (1 - tcp_alpha) * tp->t_rtt;
500 			tp->t_rtt = 0;
501 		}
502 
503 		if (ti->ti_ack == tp->snd_max)
504 			tp->t_timer[TCPT_REXMT] = 0;
505 		else {
506 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
507 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
508 			tp->t_rxtshift = 0;
509 		}
510 		if (acked > so->so_snd.sb_cc) {
511 			tp->snd_wnd -= so->so_snd.sb_cc;
512 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
513 		} else {
514 			sbdrop(&so->so_snd, acked);
515 			tp->snd_wnd -= acked;
516 			acked = 0;
517 		}
518 		if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
519 			sowwakeup(so);
520 		tp->snd_una = ti->ti_ack;
521 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
522 			tp->snd_nxt = tp->snd_una;
523 
524 		switch (tp->t_state) {
525 
526 		/*
527 		 * In FIN_WAIT_1 STATE in addition to the processing
528 		 * for the ESTABLISHED state if our FIN is now acknowledged
529 		 * then enter FIN_WAIT_2.
530 		 */
531 		case TCPS_FIN_WAIT_1:
532 			if (ourfinisacked) {
533 				/*
534 				 * If we can't receive any more
535 				 * data, then closing user can proceed.
536 				 */
537 				if (so->so_state & SS_CANTRCVMORE)
538 					soisdisconnected(so);
539 				tp->t_state = TCPS_FIN_WAIT_2;
540 				/*
541 				 * This is contrary to the specification,
542 				 * but if we haven't gotten our FIN in
543 				 * 5 minutes, it's not forthcoming.
544 				 */
545 				tp->t_timer[TCPT_2MSL] = 5 * 60 * PR_SLOWHZ;
546 			}
547 			break;
548 
549 	 	/*
550 		 * In CLOSING STATE in addition to the processing for
551 		 * the ESTABLISHED state if the ACK acknowledges our FIN
552 		 * then enter the TIME-WAIT state, otherwise ignore
553 		 * the segment.
554 		 */
555 		case TCPS_CLOSING:
556 			if (ourfinisacked) {
557 				tp->t_state = TCPS_TIME_WAIT;
558 				tcp_canceltimers(tp);
559 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
560 				soisdisconnected(so);
561 			}
562 			break;
563 
564 		/*
565 		 * The only thing that can arrive in  LAST_ACK state
566 		 * is an acknowledgment of our FIN.  If our FIN is now
567 		 * acknowledged, delete the TCB, enter the closed state
568 		 * and return.
569 		 */
570 		case TCPS_LAST_ACK:
571 			if (ourfinisacked)
572 				tp = tcp_close(tp);
573 			goto drop;
574 
575 		/*
576 		 * In TIME_WAIT state the only thing that should arrive
577 		 * is a retransmission of the remote FIN.  Acknowledge
578 		 * it and restart the finack timer.
579 		 */
580 		case TCPS_TIME_WAIT:
581 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
582 			goto dropafterack;
583 		}
584 #undef ourfinisacked
585 	}
586 
587 step6:
588 	/*
589 	 * Update window information.
590 	 */
591 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
592 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
593 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
594 		tp->snd_wnd = ti->ti_win;
595 		tp->snd_wl1 = ti->ti_seq;
596 		tp->snd_wl2 = ti->ti_ack;
597 		if (tp->snd_wnd != 0)
598 			tp->t_timer[TCPT_PERSIST] = 0;
599 	}
600 
601 	/*
602 	 * Process segments with URG.
603 	 */
604 	if ((tiflags & TH_URG) && ti->ti_urp &&
605 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
606 		/*
607 		 * This is a kludge, but if we receive accept
608 		 * random urgent pointers, we'll crash in
609 		 * soreceive.  It's hard to imagine someone
610 		 * actually wanting to send this much urgent data.
611 		 */
612 		if (ti->ti_urp > tp->rcv_wnd + 1) {	/* XXX */
613 			ti->ti_urp = 0;			/* XXX */
614 			tiflags &= ~TH_URG;		/* XXX */
615 			ti->ti_flags &= ~TH_URG;	/* XXX */
616 			goto badurp;			/* XXX */
617 		}
618 		/*
619 		 * If this segment advances the known urgent pointer,
620 		 * then mark the data stream.  This should not happen
621 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
622 		 * a FIN has been received from the remote side.
623 		 * In these states we ignore the URG.
624 		 */
625 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
626 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
627 			so->so_oobmark = so->so_rcv.sb_cc +
628 			    (tp->rcv_up - tp->rcv_nxt) - 1;
629 			if (so->so_oobmark == 0)
630 				so->so_state |= SS_RCVATMARK;
631 			sohasoutofband(so);
632 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
633 		}
634 		/*
635 		 * Remove out of band data so doesn't get presented to user.
636 		 * This can happen independent of advancing the URG pointer,
637 		 * but if two URG's are pending at once, some out-of-band
638 		 * data may creep in... ick.
639 		 */
640 		if (ti->ti_urp <= ti->ti_len)
641 			tcp_pulloutofband(so, ti);
642 	}
643 badurp:							/* XXX */
644 
645 	/*
646 	 * Process the segment text, merging it into the TCP sequencing queue,
647 	 * and arranging for acknowledgment of receipt if necessary.
648 	 * This process logically involves adjusting tp->rcv_wnd as data
649 	 * is presented to the user (this happens in tcp_usrreq.c,
650 	 * case PRU_RCVD).  If a FIN has already been received on this
651 	 * connection then we just ignore the text.
652 	 */
653 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
654 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
655 		tiflags = tcp_reass(tp, ti);
656 		if (tcpnodelack == 0)
657 			tp->t_flags |= TF_DELACK;
658 		else
659 			tp->t_flags |= TF_ACKNOW;
660 	} else {
661 		m_freem(m);
662 		tiflags &= ~TH_FIN;
663 	}
664 
665 	/*
666 	 * If FIN is received ACK the FIN and let the user know
667 	 * that the connection is closing.
668 	 */
669 	if (tiflags & TH_FIN) {
670 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
671 			socantrcvmore(so);
672 			tp->t_flags |= TF_ACKNOW;
673 			tp->rcv_nxt++;
674 		}
675 		switch (tp->t_state) {
676 
677 	 	/*
678 		 * In SYN_RECEIVED and ESTABLISHED STATES
679 		 * enter the CLOSE_WAIT state.
680 		 */
681 		case TCPS_SYN_RECEIVED:
682 		case TCPS_ESTABLISHED:
683 			tp->t_state = TCPS_CLOSE_WAIT;
684 			break;
685 
686 	 	/*
687 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
688 		 * enter the CLOSING state.
689 		 */
690 		case TCPS_FIN_WAIT_1:
691 			tp->t_state = TCPS_CLOSING;
692 			break;
693 
694 	 	/*
695 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
696 		 * starting the time-wait timer, turning off the other
697 		 * standard timers.
698 		 */
699 		case TCPS_FIN_WAIT_2:
700 			tp->t_state = TCPS_TIME_WAIT;
701 			tcp_canceltimers(tp);
702 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
703 			soisdisconnected(so);
704 			break;
705 
706 		/*
707 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
708 		 */
709 		case TCPS_TIME_WAIT:
710 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
711 			break;
712 		}
713 	}
714 	if (so->so_options & SO_DEBUG)
715 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
716 
717 	/*
718 	 * Return any desired output.
719 	 */
720 	(void) tcp_output(tp);
721 	return;
722 
723 dropafterack:
724 	/*
725 	 * Generate an ACK dropping incoming segment if it occupies
726 	 * sequence space, where the ACK reflects our state.
727 	 */
728 	if ((tiflags&TH_RST) ||
729 	    tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
730 		goto drop;
731 	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
732 		tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
733 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
734 	return;
735 
736 dropwithreset:
737 	if (om) {
738 		(void) m_free(om);
739 		om = 0;
740 	}
741 	/*
742 	 * Generate a RST, dropping incoming segment.
743 	 * Make ACK acceptable to originator of segment.
744 	 */
745 	if (tiflags & TH_RST)
746 		goto drop;
747 	if (tiflags & TH_ACK)
748 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
749 	else {
750 		if (tiflags & TH_SYN)
751 			ti->ti_len++;
752 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
753 		    TH_RST|TH_ACK);
754 	}
755 	/* destroy temporarily created socket */
756 	if (dropsocket)
757 		(void) soabort(so);
758 	return;
759 
760 drop:
761 	if (om)
762 		(void) m_free(om);
763 	/*
764 	 * Drop space held by incoming segment and return.
765 	 */
766 	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
767 		tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
768 	m_freem(m);
769 	/* destroy temporarily created socket */
770 	if (dropsocket)
771 		(void) soabort(so);
772 	return;
773 }
774 
775 tcp_dooptions(tp, om, ti)
776 	struct tcpcb *tp;
777 	struct mbuf *om;
778 	struct tcpiphdr *ti;
779 {
780 	register u_char *cp;
781 	int opt, optlen, cnt;
782 
783 	cp = mtod(om, u_char *);
784 	cnt = om->m_len;
785 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
786 		opt = cp[0];
787 		if (opt == TCPOPT_EOL)
788 			break;
789 		if (opt == TCPOPT_NOP)
790 			optlen = 1;
791 		else {
792 			optlen = cp[1];
793 			if (optlen <= 0)
794 				break;
795 		}
796 		switch (opt) {
797 
798 		default:
799 			break;
800 
801 		case TCPOPT_MAXSEG:
802 			if (optlen != 4)
803 				continue;
804 			if (!(ti->ti_flags & TH_SYN))
805 				continue;
806 			tp->t_maxseg = *(u_short *)(cp + 2);
807 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
808 			tp->t_maxseg = MIN(tp->t_maxseg, tcp_mss(tp));
809 			break;
810 		}
811 	}
812 	(void) m_free(om);
813 }
814 
815 /*
816  * Pull out of band byte out of a segment so
817  * it doesn't appear in the user's data queue.
818  * It is still reflected in the segment length for
819  * sequencing purposes.
820  */
821 tcp_pulloutofband(so, ti)
822 	struct socket *so;
823 	struct tcpiphdr *ti;
824 {
825 	register struct mbuf *m;
826 	int cnt = ti->ti_urp - 1;
827 
828 	m = dtom(ti);
829 	while (cnt >= 0) {
830 		if (m->m_len > cnt) {
831 			char *cp = mtod(m, caddr_t) + cnt;
832 			struct tcpcb *tp = sototcpcb(so);
833 
834 			tp->t_iobc = *cp;
835 			tp->t_oobflags |= TCPOOB_HAVEDATA;
836 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
837 			m->m_len--;
838 			return;
839 		}
840 		cnt -= m->m_len;
841 		m = m->m_next;
842 		if (m == 0)
843 			break;
844 	}
845 	panic("tcp_pulloutofband");
846 }
847 
848 /*
849  * Insert segment ti into reassembly queue of tcp with
850  * control block tp.  Return TH_FIN if reassembly now includes
851  * a segment with FIN.
852  */
853 tcp_reass(tp, ti)
854 	register struct tcpcb *tp;
855 	register struct tcpiphdr *ti;
856 {
857 	register struct tcpiphdr *q;
858 	struct socket *so = tp->t_inpcb->inp_socket;
859 	struct mbuf *m;
860 	int flags;
861 
862 	/*
863 	 * Call with ti==0 after become established to
864 	 * force pre-ESTABLISHED data up to user socket.
865 	 */
866 	if (ti == 0)
867 		goto present;
868 
869 	/*
870 	 * Find a segment which begins after this one does.
871 	 */
872 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
873 	    q = (struct tcpiphdr *)q->ti_next)
874 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
875 			break;
876 
877 	/*
878 	 * If there is a preceding segment, it may provide some of
879 	 * our data already.  If so, drop the data from the incoming
880 	 * segment.  If it provides all of our data, drop us.
881 	 */
882 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
883 		register int i;
884 		q = (struct tcpiphdr *)q->ti_prev;
885 		/* conversion to int (in i) handles seq wraparound */
886 		i = q->ti_seq + q->ti_len - ti->ti_seq;
887 		if (i > 0) {
888 			if (i >= ti->ti_len)
889 				goto drop;
890 			m_adj(dtom(ti), i);
891 			ti->ti_len -= i;
892 			ti->ti_seq += i;
893 		}
894 		q = (struct tcpiphdr *)(q->ti_next);
895 	}
896 
897 	/*
898 	 * While we overlap succeeding segments trim them or,
899 	 * if they are completely covered, dequeue them.
900 	 */
901 	while (q != (struct tcpiphdr *)tp) {
902 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
903 		if (i <= 0)
904 			break;
905 		if (i < q->ti_len) {
906 			q->ti_seq += i;
907 			q->ti_len -= i;
908 			m_adj(dtom(q), i);
909 			break;
910 		}
911 		q = (struct tcpiphdr *)q->ti_next;
912 		m = dtom(q->ti_prev);
913 		remque(q->ti_prev);
914 		m_freem(m);
915 	}
916 
917 	/*
918 	 * Stick new segment in its place.
919 	 */
920 	insque(ti, q->ti_prev);
921 
922 present:
923 	/*
924 	 * Present data to user, advancing rcv_nxt through
925 	 * completed sequence space.
926 	 */
927 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
928 		return (0);
929 	ti = tp->seg_next;
930 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
931 		return (0);
932 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
933 		return (0);
934 	do {
935 		tp->rcv_nxt += ti->ti_len;
936 		flags = ti->ti_flags & TH_FIN;
937 		remque(ti);
938 		m = dtom(ti);
939 		ti = (struct tcpiphdr *)ti->ti_next;
940 		if (so->so_state & SS_CANTRCVMORE)
941 			m_freem(m);
942 		else
943 			sbappend(&so->so_rcv, m);
944 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
945 	sorwakeup(so);
946 	return (flags);
947 drop:
948 	m_freem(dtom(ti));
949 	return (0);
950 }
951 
952 /*
953  *  Determine a reasonable value for maxseg size.
954  *  If the route is known, use one that can be handled
955  *  on the given interface without forcing IP to fragment.
956  *  If bigger than a page (CLSIZE), round down to nearest pagesize
957  *  to utilize pagesize mbufs.
958  *  If interface pointer is unavailable, or the destination isn't local,
959  *  use a conservative size (512 or the default IP max size),
960  *  as we can't discover anything about intervening gateways or networks.
961  *
962  *  This is ugly, and doesn't belong at this level, but has to happen somehow.
963  */
964 tcp_mss(tp)
965 register struct tcpcb *tp;
966 {
967 	struct route *ro;
968 	struct ifnet *ifp;
969 	int mss;
970 	struct inpcb *inp;
971 
972 	inp = tp->t_inpcb;
973 	ro = &inp->inp_route;
974 	if ((ro->ro_rt == (struct rtentry *)0) ||
975 	    (ifp = ro->ro_rt->rt_ifp) == (struct ifnet *)0) {
976 		/* No route yet, so try to acquire one */
977 		if (inp->inp_faddr.s_addr != INADDR_ANY) {
978 			ro->ro_dst.sa_family = AF_INET;
979 			((struct sockaddr_in *) &ro->ro_dst)->sin_addr =
980 				inp->inp_faddr;
981 			rtalloc(ro);
982 		}
983 		if ((ro->ro_rt == 0) || (ifp = ro->ro_rt->rt_ifp) == 0)
984 			return (TCP_MSS);
985 	}
986 
987 	mss = ifp->if_mtu - sizeof(struct tcpiphdr);
988 #if	(CLBYTES & (CLBYTES - 1)) == 0
989 	if (mss > CLBYTES)
990 		mss &= ~(CLBYTES-1);
991 #else
992 	if (mss > CLBYTES)
993 		mss = mss / CLBYTES * CLBYTES;
994 #endif
995 	if (in_localaddr(tp->t_inpcb->inp_faddr))
996 		return(mss);
997 	return (MIN(mss, TCP_MSS));
998 }
999