xref: /csrg-svn/sys/netinet/tcp_input.c (revision 5623)
1 /*	tcp_input.c	1.51	82/01/25	*/
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 "../net/in.h"
10 #include "../net/in_pcb.h"
11 #include "../net/in_systm.h"
12 #include "../net/if.h"
13 #include "../net/ip.h"
14 #include "../net/ip_var.h"
15 #include "../net/tcp.h"
16 #include "../net/tcp_fsm.h"
17 #include "../net/tcp_seq.h"
18 #include "../net/tcp_timer.h"
19 #include "../net/tcp_var.h"
20 #include "../net/tcpip.h"
21 #include "../net/tcp_debug.h"
22 #include "../errno.h"
23 
24 int	tcpprintfs = 0;
25 int	tcpcksum = 1;
26 struct	sockaddr_in tcp_in = { AF_INET };
27 struct	tcpiphdr tcp_saveti;
28 extern	tcpnodelack;
29 
30 struct	tcpcb *tcp_newtcpcb();
31 /*
32  * TCP input routine, follows pages 65-76 of the
33  * protocol specification dated September, 1981 very closely.
34  */
35 tcp_input(m0)
36 	struct mbuf *m0;
37 {
38 	register struct tcpiphdr *ti;
39 	struct inpcb *inp;
40 	register struct mbuf *m;
41 	struct mbuf *om = 0;
42 	int len, tlen, off;
43 	register struct tcpcb *tp = 0;
44 	register int tiflags;
45 	struct socket *so;
46 	int todrop, acked;
47 	short ostate;
48 
49 COUNT(TCP_INPUT);
50 	/*
51 	 * Get IP and TCP header together in first mbuf.
52 	 * Note: IP leaves IP header in first mbuf.
53 	 */
54 	m = m0;
55 	ti = mtod(m, struct tcpiphdr *);
56 	if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2))
57 		ip_stripoptions((struct ip *)ti, (struct mbuf *)0);
58 	if (m->m_off > MMAXOFF || m->m_len < sizeof (struct tcpiphdr)) {
59 		if ((m = m_pullup(m, sizeof (struct tcpiphdr))) == 0) {
60 			tcpstat.tcps_hdrops++;
61 			return;
62 		}
63 		ti = mtod(m, struct tcpiphdr *);
64 	}
65 
66 	/*
67 	 * Checksum extended TCP header and data.
68 	 */
69 	tlen = ((struct ip *)ti)->ip_len;
70 	len = sizeof (struct ip) + tlen;
71 	if (tcpcksum) {
72 		ti->ti_next = ti->ti_prev = 0;
73 		ti->ti_x1 = 0;
74 		ti->ti_len = (u_short)tlen;
75 #if vax
76 		ti->ti_len = htons(ti->ti_len);
77 #endif
78 		if (ti->ti_sum = in_cksum(m, len)) {
79 			tcpstat.tcps_badsum++;
80 			printf("tcp cksum %x\n", ti->ti_sum);
81 			goto drop;
82 		}
83 	}
84 
85 	/*
86 	 * Check that TCP offset makes sense,
87 	 * pull out TCP options and adjust length.
88 	 */
89 	off = ti->ti_off << 2;
90 	if (off < sizeof (struct tcphdr) || off > tlen) {
91 		tcpstat.tcps_badoff++;
92 		goto drop;
93 	}
94 	ti->ti_len = tlen - off;
95 	if (off > sizeof (struct tcphdr)) {
96 		if ((m = m_pullup(m, sizeof (struct ip) + off)) == 0) {
97 			tcpstat.tcps_hdrops++;
98 			goto drop;
99 		}
100 		ti = mtod(m, struct tcpiphdr *);
101 		om = m_get(M_DONTWAIT);
102 		if (om == 0)
103 			goto drop;
104 		om->m_off = MMINOFF;
105 		om->m_len = off - sizeof (struct tcphdr);
106 		{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
107 		  bcopy(op, mtod(om, caddr_t), om->m_len);
108 		  m->m_len -= om->m_len;
109 		  bcopy(op+om->m_len, op, m->m_len-sizeof (struct tcpiphdr));
110 		}
111 	}
112 	tiflags = ti->ti_flags;
113 
114 #if vax
115 	/*
116 	 * Convert TCP protocol specific fields to host format.
117 	 */
118 	ti->ti_seq = ntohl(ti->ti_seq);
119 	ti->ti_ack = ntohl(ti->ti_ack);
120 	ti->ti_win = ntohs(ti->ti_win);
121 	ti->ti_urp = ntohs(ti->ti_urp);
122 #endif
123 
124 	/*
125 	 * Locate pcb for segment.
126 	 */
127 	inp = in_pcblookup
128 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport);
129 
130 	/*
131 	 * If the state is CLOSED (i.e., TCB does not exist) then
132 	 * all data in the incoming segment is discarded.
133 	 */
134 	if (inp == 0)
135 		goto dropwithreset;
136 	tp = intotcpcb(inp);
137 	if (tp == 0)
138 		goto dropwithreset;
139 	so = inp->inp_socket;
140 	if (so->so_options & SO_DEBUG) {
141 		ostate = tp->t_state;
142 		tcp_saveti = *ti;
143 	}
144 
145 	/*
146 	 * Segment received on connection.
147 	 * Reset idle time and keep-alive timer.
148 	 */
149 	tp->t_idle = 0;
150 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
151 
152 	/*
153 	 * Process options.
154 	 */
155 	if (om) {
156 		tcp_dooptions(tp, om);
157 		om = 0;
158 	}
159 
160 	/*
161 	 * Calculate amount of space in receive window,
162 	 * and then do TCP input processing.
163 	 */
164 	tp->rcv_wnd = sbspace(&so->so_rcv);
165 	if (tp->rcv_wnd < 0)
166 		tp->rcv_wnd = 0;
167 
168 	switch (tp->t_state) {
169 
170 	/*
171 	 * If the state is LISTEN then ignore segment if it contains an RST.
172 	 * If the segment contains an ACK then it is bad and send a RST.
173 	 * If it does not contain a SYN then it is not interesting; drop it.
174 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
175 	 * tp->iss, and send a segment:
176 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
177 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
178 	 * Fill in remote peer address fields if not previously specified.
179 	 * Enter SYN_RECEIVED state, and process any other fields of this
180 	 * segment in this state.
181 	 */
182 	case TCPS_LISTEN:
183 		if (tiflags & TH_RST)
184 			goto drop;
185 		if (tiflags & TH_ACK)
186 			goto dropwithreset;
187 		if ((tiflags & TH_SYN) == 0)
188 			goto drop;
189 		tcp_in.sin_addr = ti->ti_src;
190 		tcp_in.sin_port = ti->ti_sport;
191 		if (in_pcbconnect(inp, (struct sockaddr *)&tcp_in))
192 			goto drop;
193 		tp->t_template = tcp_template(tp);
194 		if (tp->t_template == 0) {
195 			in_pcbdisconnect(inp);
196 			goto drop;
197 		}
198 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
199 		tp->irs = ti->ti_seq;
200 		tcp_sendseqinit(tp);
201 		tcp_rcvseqinit(tp);
202 		tp->t_state = TCPS_SYN_RECEIVED;
203 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
204 		goto trimthenstep6;
205 
206 	/*
207 	 * If the state is SYN_SENT:
208 	 *	if seg contains an ACK, but not for our SYN, drop the input.
209 	 *	if seg contains a RST, then drop the connection.
210 	 *	if seg does not contain SYN, then drop it.
211 	 * Otherwise this is an acceptable SYN segment
212 	 *	initialize tp->rcv_nxt and tp->irs
213 	 *	if seg contains ack then advance tp->snd_una
214 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
215 	 *	arrange for segment to be acked (eventually)
216 	 *	continue processing rest of data/controls, beginning with URG
217 	 */
218 	case TCPS_SYN_SENT:
219 		if ((tiflags & TH_ACK) &&
220 /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
221 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
222 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
223 			goto dropwithreset;
224 		if (tiflags & TH_RST) {
225 			if (tiflags & TH_ACK)
226 				tcp_drop(tp, ECONNREFUSED);
227 			goto drop;
228 		}
229 		if ((tiflags & TH_SYN) == 0)
230 			goto drop;
231 		tp->snd_una = ti->ti_ack;
232 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
233 			tp->snd_nxt = tp->snd_una;
234 		tp->t_timer[TCPT_REXMT] = 0;
235 		tp->irs = ti->ti_seq;
236 		tcp_rcvseqinit(tp);
237 		tp->t_flags |= TF_ACKNOW;
238 		if (SEQ_GT(tp->snd_una, tp->iss)) {
239 			if (so->so_options & SO_ACCEPTCONN)
240 				so->so_state |= SS_CONNAWAITING;
241 			soisconnected(so);
242 			tp->t_state = TCPS_ESTABLISHED;
243 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
244 		} else
245 			tp->t_state = TCPS_SYN_RECEIVED;
246 		goto trimthenstep6;
247 
248 trimthenstep6:
249 		/*
250 		 * Advance ti->ti_seq to correspond to first data byte.
251 		 * If data, trim to stay within window,
252 		 * dropping FIN if necessary.
253 		 */
254 		ti->ti_seq++;
255 		if (ti->ti_len > tp->rcv_wnd) {
256 			todrop = ti->ti_len - tp->rcv_wnd;
257 			m_adj(m, -todrop);
258 			ti->ti_len = tp->rcv_wnd;
259 			ti->ti_flags &= ~TH_FIN;
260 		}
261 		tp->snd_wl1 = ti->ti_seq - 1;
262 		goto step6;
263 	}
264 
265 	/*
266 	 * States other than LISTEN or SYN_SENT.
267 	 * First check that at least some bytes of segment are within
268 	 * receive window.
269 	 */
270 	if (tp->rcv_wnd == 0) {
271 		/*
272 		 * If window is closed can only take segments at
273 		 * window edge, and have to drop data and PUSH from
274 		 * incoming segments.
275 		 */
276 		if (tp->rcv_nxt != ti->ti_seq)
277 			goto dropafterack;
278 		if (ti->ti_len > 0) {
279 			ti->ti_len = 0;
280 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
281 		}
282 	} else {
283 		/*
284 		 * If segment begins before rcv_nxt, drop leading
285 		 * data (and SYN); if nothing left, just ack.
286 		 */
287 		if (SEQ_GT(tp->rcv_nxt, ti->ti_seq)) {
288 			todrop = tp->rcv_nxt - ti->ti_seq;
289 			if (tiflags & TH_SYN) {
290 				tiflags &= ~TH_SYN;
291 				ti->ti_seq++;
292 				if (ti->ti_urp > 1)
293 					ti->ti_urp--;
294 				else
295 					tiflags &= ~TH_URG;
296 				todrop--;
297 			}
298 			if (todrop > ti->ti_len)
299 				goto dropafterack;
300 			m_adj(m, todrop);
301 			ti->ti_seq += todrop;
302 			ti->ti_len -= todrop;
303 			if (ti->ti_urp > todrop)
304 				ti->ti_urp -= todrop;
305 			else {
306 				tiflags &= ~TH_URG;
307 				/* ti->ti_flags &= ~TH_URG; */
308 				/* ti->ti_urp = 0; */
309 			}
310 			/* tiflags &= ~TH_SYN; */
311 			/* ti->ti_flags &= ~TH_SYN; */
312 		}
313 		/*
314 		 * If segment ends after window, drop trailing data
315 		 * (and PUSH and FIN); if nothing left, just ACK.
316 		 */
317 		if (SEQ_GT(ti->ti_seq+ti->ti_len, tp->rcv_nxt+tp->rcv_wnd)) {
318 			todrop =
319 			     ti->ti_seq+ti->ti_len - (tp->rcv_nxt+tp->rcv_wnd);
320 			if (todrop > ti->ti_len)
321 				goto dropafterack;
322 			m_adj(m, -todrop);
323 			ti->ti_len -= todrop;
324 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
325 		}
326 	}
327 
328 	/*
329 	 * If the RST bit is set examine the state:
330 	 *    SYN_RECEIVED STATE:
331 	 *	If passive open, return to LISTEN state.
332 	 *	If active open, inform user that connection was refused.
333 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
334 	 *	Inform user that connection was reset, and close tcb.
335 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
336 	 *	Close the tcb.
337 	 */
338 	if (tiflags&TH_RST) switch (tp->t_state) {
339 
340 	case TCPS_SYN_RECEIVED:
341 		if (inp->inp_socket->so_options & SO_ACCEPTCONN) {
342 			/* a miniature tcp_close, but invisible to user */
343 			(void) m_free(dtom(tp->t_template));
344 			(void) m_free(dtom(tp));
345 			inp->inp_ppcb = 0;
346 			tp = tcp_newtcpcb(inp);
347 			tp->t_state = TCPS_LISTEN;
348 			goto drop;
349 		}
350 		tcp_drop(tp, ECONNREFUSED);
351 		goto drop;
352 
353 	case TCPS_ESTABLISHED:
354 	case TCPS_FIN_WAIT_1:
355 	case TCPS_FIN_WAIT_2:
356 	case TCPS_CLOSE_WAIT:
357 		tcp_drop(tp, ECONNRESET);
358 		goto drop;
359 
360 	case TCPS_CLOSING:
361 	case TCPS_LAST_ACK:
362 	case TCPS_TIME_WAIT:
363 		tcp_close(tp);
364 		goto drop;
365 	}
366 
367 	/*
368 	 * If a SYN is in the window, then this is an
369 	 * error and we send an RST and drop the connection.
370 	 */
371 	if (tiflags & TH_SYN) {
372 		tcp_drop(tp, ECONNRESET);
373 		goto dropwithreset;
374 	}
375 
376 	/*
377 	 * If the ACK bit is off we drop the segment and return.
378 	 */
379 	if ((tiflags & TH_ACK) == 0)
380 		goto drop;
381 
382 	/*
383 	 * Ack processing.
384 	 */
385 	switch (tp->t_state) {
386 
387 	/*
388 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
389 	 * ESTABLISHED state and continue processing, othewise
390 	 * send an RST.
391 	 */
392 	case TCPS_SYN_RECEIVED:
393 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
394 		    SEQ_GT(ti->ti_ack, tp->snd_max))
395 			goto dropwithreset;
396 		tp->snd_una++;			/* SYN acked */
397 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
398 			tp->snd_nxt = tp->snd_una;
399 		tp->t_timer[TCPT_REXMT] = 0;
400 		if (so->so_options & SO_ACCEPTCONN)
401 			so->so_state |= SS_CONNAWAITING;
402 		soisconnected(so);
403 		tp->t_state = TCPS_ESTABLISHED;
404 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
405 		tp->snd_wl1 = ti->ti_seq - 1;
406 		/* fall into ... */
407 
408 	/*
409 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
410 	 * ACKs.  If the ack is in the range
411 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
412 	 * then advance tp->snd_una to ti->ti_ack and drop
413 	 * data from the retransmission queue.  If this ACK reflects
414 	 * more up to date window information we update our window information.
415 	 */
416 	case TCPS_ESTABLISHED:
417 	case TCPS_FIN_WAIT_1:
418 	case TCPS_FIN_WAIT_2:
419 	case TCPS_CLOSE_WAIT:
420 	case TCPS_CLOSING:
421 	case TCPS_LAST_ACK:
422 	case TCPS_TIME_WAIT:
423 #define	ourfinisacked	(acked > 0)
424 
425 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
426 			break;
427 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
428 			goto dropafterack;
429 		acked = ti->ti_ack - tp->snd_una;
430 		if (ti->ti_ack == tp->snd_max)
431 			tp->t_timer[TCPT_REXMT] = 0;
432 		else {
433 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
434 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
435 			tp->t_rtt = 0;
436 			tp->t_rxtshift = 0;
437 		}
438 		if (acked > so->so_snd.sb_cc) {
439 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
440 			tp->snd_wnd -= so->so_snd.sb_cc;
441 		} else {
442 			sbdrop(&so->so_snd.sb_cc, acked);
443 			tp->snd_wnd -= acked;
444 			acked = 0;
445 		}
446 		if (so->so_snd.sb_flags & SB_WAIT)
447 			sowwakeup(so);
448 		tp->snd_una = ti->ti_ack;
449 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
450 			tp->snd_nxt = tp->snd_una;
451 
452 		/*
453 		 * If transmit timer is running and timed sequence
454 		 * number was acked, update smoothed round trip time.
455 		 */
456 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
457 			if (tp->t_srtt == 0)
458 				tp->t_srtt = tp->t_rtt;
459 			else
460 				tp->t_srtt =
461 				    tcp_alpha * tp->t_srtt +
462 				    (1 - tcp_alpha) * tp->t_rtt;
463 			tp->t_rtt = 0;
464 		}
465 
466 		switch (tp->t_state) {
467 
468 		/*
469 		 * In FIN_WAIT_1 STATE in addition to the processing
470 		 * for the ESTABLISHED state if our FIN is now acknowledged
471 		 * then enter FIN_WAIT_2.
472 		 */
473 		case TCPS_FIN_WAIT_1:
474 			if (ourfinisacked)
475 				tp->t_state = TCPS_FIN_WAIT_2;
476 			break;
477 
478 	 	/*
479 		 * In CLOSING STATE in addition to the processing for
480 		 * the ESTABLISHED state if the ACK acknowledges our FIN
481 		 * then enter the TIME-WAIT state, otherwise ignore
482 		 * the segment.
483 		 */
484 		case TCPS_CLOSING:
485 			if (ourfinisacked) {
486 				tp->t_state = TCPS_TIME_WAIT;
487 				tcp_canceltimers(tp);
488 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
489 				soisdisconnected(so);
490 			}
491 			break;
492 
493 		/*
494 		 * The only thing that can arrive in  LAST_ACK state
495 		 * is an acknowledgment of our FIN.  If our FIN is now
496 		 * acknowledged, delete the TCB, enter the closed state
497 		 * and return.
498 		 */
499 		case TCPS_LAST_ACK:
500 			if (ourfinisacked)
501 				tcp_close(tp);
502 			goto drop;
503 
504 		/*
505 		 * In TIME_WAIT state the only thing that should arrive
506 		 * is a retransmission of the remote FIN.  Acknowledge
507 		 * it and restart the finack timer.
508 		 */
509 		case TCPS_TIME_WAIT:
510 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
511 			goto dropafterack;
512 		}
513 #undef ourfinisacked
514 	}
515 
516 step6:
517 	/*
518 	 * Update window information.
519 	 */
520 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
521 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
522 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
523 /*
524 printf("wl1 %x seq %x wl2 %x ack %x win %x wnd %x\n", tp->snd_wl1, ti->ti_seq, tp->snd_wl2, ti->ti_ack, ti->ti_win, tp->snd_wnd);
525 */
526 		tp->snd_wnd = ti->ti_win;
527 		tp->snd_wl1 = ti->ti_seq;
528 		tp->snd_wl2 = ti->ti_ack;
529 		if (tp->snd_wnd > 0)
530 			tp->t_timer[TCPT_PERSIST] = 0;
531 	}
532 
533 	/*
534 	 * Process segments with URG.
535 	 */
536 	if ((tiflags & TH_URG) && TCPS_HAVERCVDFIN(tp->t_state) == 0) {
537 		/*
538 		 * If this segment advances the known urgent pointer,
539 		 * then mark the data stream.  This should not happen
540 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
541 		 * a FIN has been received from the remote side.
542 		 * In these states we ignore the URG.
543 		 */
544 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
545 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
546 			so->so_oobmark = so->so_rcv.sb_cc +
547 			    (tp->rcv_up - tp->rcv_nxt) - 1;
548 			if (so->so_oobmark == 0)
549 				so->so_state |= SS_RCVATMARK;
550 #ifdef TCPTRUEOOB
551 			if ((tp->t_flags & TF_DOOOB) == 0)
552 #endif
553 				sohasoutofband(so);
554 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
555 		}
556 		/*
557 		 * Remove out of band data so doesn't get presented to user.
558 		 * This can happen independent of advancing the URG pointer,
559 		 * but if two URG's are pending at once, some out-of-band
560 		 * data may creep in... ick.
561 		 */
562 		if (ti->ti_urp <= ti->ti_len) {
563 			tcp_pulloutofband(so, ti);
564 		}
565 	}
566 
567 	/*
568 	 * Process the segment text, merging it into the TCP sequencing queue,
569 	 * and arranging for acknowledgment of receipt if necessary.
570 	 * This process logically involves adjusting tp->rcv_wnd as data
571 	 * is presented to the user (this happens in tcp_usrreq.c,
572 	 * case PRU_RCVD).  If a FIN has already been received on this
573 	 * connection then we just ignore the text.
574 	 */
575 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
576 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
577 		off += sizeof (struct ip);		/* drop IP header */
578 		m->m_off += off;
579 		m->m_len -= off;
580 		tiflags = tcp_reass(tp, ti);
581 		if (tcpnodelack == 0)
582 			tp->t_flags |= TF_DELACK;
583 		else
584 			tp->t_flags |= TF_ACKNOW;
585 	} else {
586 		m_freem(m);
587 		tiflags &= ~TH_FIN;
588 	}
589 
590 	/*
591 	 * If FIN is received ACK the FIN and let the user know
592 	 * that the connection is closing.
593 	 */
594 	if (tiflags & TH_FIN) {
595 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
596 			socantrcvmore(so);
597 			tp->t_flags |= TF_ACKNOW;
598 			tp->rcv_nxt++;
599 		}
600 		switch (tp->t_state) {
601 
602 	 	/*
603 		 * In SYN_RECEIVED and ESTABLISHED STATES
604 		 * enter the CLOSE_WAIT state.
605 		 */
606 		case TCPS_SYN_RECEIVED:
607 		case TCPS_ESTABLISHED:
608 			tp->t_state = TCPS_CLOSE_WAIT;
609 			break;
610 
611 	 	/*
612 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
613 		 * enter the CLOSING state.
614 		 */
615 		case TCPS_FIN_WAIT_1:
616 			tp->t_state = TCPS_CLOSING;
617 			break;
618 
619 	 	/*
620 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
621 		 * starting the time-wait timer, turning off the other
622 		 * standard timers.
623 		 */
624 		case TCPS_FIN_WAIT_2:
625 			tp->t_state = TCPS_TIME_WAIT;
626 			tcp_canceltimers(tp);
627 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
628 			soisdisconnected(so);
629 			break;
630 
631 		/*
632 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
633 		 */
634 		case TCPS_TIME_WAIT:
635 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
636 			break;
637 		}
638 	}
639 	if (so->so_options & SO_DEBUG)
640 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
641 
642 	/*
643 	 * Return any desired output.
644 	 */
645 	tcp_output(tp);
646 	return;
647 
648 dropafterack:
649 	/*
650 	 * Generate an ACK dropping incoming segment.
651 	 * Make ACK reflect our state.
652 	 */
653 	if (tiflags & TH_RST)
654 		goto drop;
655 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
656 	return;
657 
658 dropwithreset:
659 	if (om)
660 		m_free(om);
661 	/*
662 	 * Generate a RST, dropping incoming segment.
663 	 * Make ACK acceptable to originator of segment.
664 	 */
665 	if (tiflags & TH_RST)
666 		goto drop;
667 	if (tiflags & TH_ACK)
668 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
669 	else {
670 		if (tiflags & TH_SYN)
671 			ti->ti_len++;
672 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK);
673 	}
674 	return;
675 
676 drop:
677 	/*
678 	 * Drop space held by incoming segment and return.
679 	 */
680 	m_freem(m);
681 	return;
682 }
683 
684 tcp_dooptions(tp, om)
685 	struct tcpcb *tp;
686 	struct mbuf *om;
687 {
688 	register u_char *cp;
689 	int opt, optlen, cnt;
690 
691 	cp = mtod(om, u_char *);
692 	cnt = om->m_len;
693 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
694 		opt = cp[0];
695 		if (opt == TCPOPT_EOL)
696 			break;
697 		if (opt == TCPOPT_NOP)
698 			optlen = 1;
699 		else
700 			optlen = cp[1];
701 		switch (opt) {
702 
703 		default:
704 			break;
705 
706 		case TCPOPT_MAXSEG:
707 			if (optlen != 4)
708 				continue;
709 			tp->t_maxseg = *(u_short *)(cp + 2);
710 #if vax
711 			tp->t_maxseg = ntohs(tp->t_maxseg);
712 #endif
713 			break;
714 
715 #ifdef TCPTRUEOOB
716 		case TCPOPT_WILLOOB:
717 			tp->t_flags |= TF_DOOOB;
718 printf("tp %x dooob\n", tp);
719 			break;
720 
721 		case TCPOPT_OOBDATA: {
722 			int seq;
723 			register struct socket *so = tp->t_inpcb->inp_socket;
724 			tcp_seq mark;
725 
726 			if (optlen != 8)
727 				continue;
728 			seq = cp[2];
729 			if (seq < tp->t_iobseq)
730 				seq += 256;
731 printf("oobdata cp[2] %d iobseq %d seq %d\n", cp[2], tp->t_iobseq, seq);
732 			if (seq - tp->t_iobseq > 128) {
733 printf("bad seq\n");
734 				tp->t_oobflags |= TCPOOB_OWEACK;
735 				break;
736 			}
737 			tp->t_iobseq = cp[2];
738 			tp->t_iobc = cp[3];
739 			mark = *(tcp_seq *)(cp + 4);
740 #if vax
741 			mark = ntohl(mark);
742 #endif
743 			so->so_oobmark = so->so_rcv.sb_cc + (mark-tp->rcv_nxt);
744 			if (so->so_oobmark == 0)
745 				so->so_state |= SS_RCVATMARK;
746 printf("take oob data %x input iobseq now %x\n", tp->t_iobc, tp->t_iobseq);
747 			sohasoutofband(so);
748 			break;
749 		}
750 
751 		case TCPOPT_OOBACK: {
752 			int seq;
753 
754 			if (optlen != 4)
755 				continue;
756 			if (tp->t_oobseq != cp[2]) {
757 printf("wrong ack\n");
758 				break;
759 			}
760 printf("take oob ack %x and cancel rexmt\n", cp[2]);
761 			tp->t_oobflags &= ~TCPOOB_NEEDACK;
762 			tp->t_timer[TCPT_OOBREXMT] = 0;
763 			break;
764 		}
765 #endif TCPTRUEOOB
766 		}
767 	}
768 	m_free(om);
769 }
770 
771 /*
772  * Pull out of band byte out of a segment so
773  * it doesn't appear in the user's data queue.
774  * It is still reflected in the segment length for
775  * sequencing purposes.
776  */
777 tcp_pulloutofband(so, ti)
778 	struct socket *so;
779 	struct tcpiphdr *ti;
780 {
781 	register struct mbuf *m;
782 	int cnt = sizeof (struct tcpiphdr) + ti->ti_urp - 1;
783 
784 	m = dtom(ti);
785 	while (cnt >= 0) {
786 		if (m->m_len > cnt) {
787 			char *cp = mtod(m, caddr_t) + cnt;
788 			struct tcpcb *tp = sototcpcb(so);
789 
790 			tp->t_iobc = *cp;
791 			tp->t_oobflags |= TCPOOB_HAVEDATA;
792 			bcopy(cp+1, cp, m->m_len - cnt - 1);
793 			m->m_len--;
794 			return;
795 		}
796 		cnt -= m->m_len;
797 		m = m->m_next;
798 		if (m == 0)
799 			break;
800 	}
801 	panic("tcp_pulloutofband");
802 }
803 
804 /*
805  * Insert segment ti into reassembly queue of tcp with
806  * control block tp.  Return TH_FIN if reassembly now includes
807  * a segment with FIN.
808  */
809 tcp_reass(tp, ti)
810 	register struct tcpcb *tp;
811 	register struct tcpiphdr *ti;
812 {
813 	register struct tcpiphdr *q;
814 	struct socket *so = tp->t_inpcb->inp_socket;
815 	struct mbuf *m;
816 	int flags;
817 COUNT(TCP_REASS);
818 
819 	/*
820 	 * Call with ti==0 after become established to
821 	 * force pre-ESTABLISHED data up to user socket.
822 	 */
823 	if (ti == 0)
824 		goto present;
825 
826 	/*
827 	 * Find a segment which begins after this one does.
828 	 */
829 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
830 	    q = (struct tcpiphdr *)q->ti_next)
831 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
832 			break;
833 
834 	/*
835 	 * If there is a preceding segment, it may provide some of
836 	 * our data already.  If so, drop the data from the incoming
837 	 * segment.  If it provides all of our data, drop us.
838 	 */
839 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
840 		register int i;
841 		q = (struct tcpiphdr *)(q->ti_prev);
842 		/* conversion to int (in i) handles seq wraparound */
843 		i = q->ti_seq + q->ti_len - ti->ti_seq;
844 		if (i > 0) {
845 			if (i >= ti->ti_len)
846 				goto drop;
847 			m_adj(dtom(tp), i);
848 			ti->ti_len -= i;
849 			ti->ti_seq += i;
850 		}
851 		q = (struct tcpiphdr *)(q->ti_next);
852 	}
853 
854 	/*
855 	 * While we overlap succeeding segments trim them or,
856 	 * if they are completely covered, dequeue them.
857 	 */
858 	while (q != (struct tcpiphdr *)tp &&
859 	    SEQ_GT(ti->ti_seq + ti->ti_len, q->ti_seq)) {
860 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
861 		if (i < q->ti_len) {
862 			q->ti_len -= i;
863 			m_adj(dtom(q), i);
864 			break;
865 		}
866 		q = (struct tcpiphdr *)q->ti_next;
867 		m = dtom(q->ti_prev);
868 		remque(q->ti_prev);
869 		m_freem(m);
870 	}
871 
872 	/*
873 	 * Stick new segment in its place.
874 	 */
875 	insque(ti, q->ti_prev);
876 
877 present:
878 	/*
879 	 * Present data to user, advancing rcv_nxt through
880 	 * completed sequence space.
881 	 */
882 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
883 		return (0);
884 	ti = tp->seg_next;
885 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
886 		return (0);
887 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
888 		return (0);
889 	do {
890 		tp->rcv_nxt += ti->ti_len;
891 		flags = ti->ti_flags & TH_FIN;
892 		remque(ti);
893 		m = dtom(ti);
894 		ti = (struct tcpiphdr *)ti->ti_next;
895 		if (so->so_state & SS_CANTRCVMORE)
896 			(void) m_freem(m);
897 		else
898 			sbappend(&so->so_rcv, m);
899 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
900 	sorwakeup(so);
901 	return (flags);
902 drop:
903 	m_freem(dtom(ti));
904 	return (0);
905 }
906