xref: /csrg-svn/sys/netinet/tcp_input.c (revision 10394)
1 /*	tcp_input.c	1.87	83/01/17	*/
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 "../netinet/in.h"
10 #include "../net/route.h"
11 #include "../netinet/in_pcb.h"
12 #include "../netinet/in_systm.h"
13 #include "../net/if.h"
14 #include "../netinet/ip.h"
15 #include "../netinet/ip_var.h"
16 #include "../netinet/tcp.h"
17 #include "../netinet/tcp_fsm.h"
18 #include "../netinet/tcp_seq.h"
19 #include "../netinet/tcp_timer.h"
20 #include "../netinet/tcp_var.h"
21 #include "../netinet/tcpip.h"
22 #include "../netinet/tcp_debug.h"
23 #include <errno.h>
24 
25 int	tcpprintfs = 0;
26 int	tcpcksum = 1;
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 	struct in_addr laddr;
49 
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 		ti->ti_len = htons((u_short)ti->ti_len);
76 		if (ti->ti_sum = in_cksum(m, len)) {
77 			tcpstat.tcps_badsum++;
78 			if (tcpprintfs)
79 				printf("tcp cksum %x\n", ti->ti_sum);
80 			goto drop;
81 		}
82 	}
83 
84 	/*
85 	 * Check that TCP offset makes sense,
86 	 * pull out TCP options and adjust length.
87 	 */
88 	off = ti->ti_off << 2;
89 	if (off < sizeof (struct tcphdr) || off > tlen) {
90 		tcpstat.tcps_badoff++;
91 		goto drop;
92 	}
93 	tlen -= off;
94 	ti->ti_len = tlen;
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, MT_DATA);
102 		if (om == 0)
103 			goto drop;
104 		om->m_len = off - sizeof (struct tcphdr);
105 		{ caddr_t op = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
106 		  bcopy(op, mtod(om, caddr_t), (unsigned)om->m_len);
107 		  m->m_len -= om->m_len;
108 		  bcopy(op+om->m_len, op,
109 		   (unsigned)(m->m_len-sizeof (struct tcpiphdr)));
110 		}
111 	}
112 	tiflags = ti->ti_flags;
113 
114 	/*
115 	 * Drop TCP and IP headers.
116 	 */
117 	off += sizeof (struct ip);
118 	m->m_off += off;
119 	m->m_len -= off;
120 
121 	/*
122 	 * Convert TCP protocol specific fields to host format.
123 	 */
124 	ti->ti_seq = ntohl(ti->ti_seq);
125 	ti->ti_ack = ntohl(ti->ti_ack);
126 	ti->ti_win = ntohs(ti->ti_win);
127 	ti->ti_urp = ntohs(ti->ti_urp);
128 
129 	/*
130 	 * Locate pcb for segment.
131 	 */
132 	inp = in_pcblookup
133 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport,
134 		INPLOOKUP_WILDCARD);
135 
136 	/*
137 	 * If the state is CLOSED (i.e., TCB does not exist) then
138 	 * all data in the incoming segment is discarded.
139 	 */
140 	if (inp == 0)
141 		goto dropwithreset;
142 	tp = intotcpcb(inp);
143 	if (tp == 0)
144 		goto dropwithreset;
145 	so = inp->inp_socket;
146 	if (so->so_options & SO_DEBUG) {
147 		ostate = tp->t_state;
148 		tcp_saveti = *ti;
149 	}
150 	if (so->so_options & SO_ACCEPTCONN) {
151 		so = sonewconn(so);
152 		if (so == 0)
153 			goto drop;
154 		inp = (struct inpcb *)so->so_pcb;
155 		inp->inp_laddr = ti->ti_dst;
156 		inp->inp_lport = ti->ti_dport;
157 		tp = intotcpcb(inp);
158 		tp->t_state = TCPS_LISTEN;
159 	}
160 
161 	/*
162 	 * Segment received on connection.
163 	 * Reset idle time and keep-alive timer.
164 	 */
165 	tp->t_idle = 0;
166 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
167 
168 	/*
169 	 * Process options.
170 	 */
171 	if (om) {
172 		tcp_dooptions(tp, om);
173 		om = 0;
174 	}
175 
176 	/*
177 	 * Calculate amount of space in receive window,
178 	 * and then do TCP input processing.
179 	 */
180 	tp->rcv_wnd = sbspace(&so->so_rcv);
181 	if (tp->rcv_wnd < 0)
182 		tp->rcv_wnd = 0;
183 
184 	switch (tp->t_state) {
185 
186 	/*
187 	 * If the state is LISTEN then ignore segment if it contains an RST.
188 	 * If the segment contains an ACK then it is bad and send a RST.
189 	 * If it does not contain a SYN then it is not interesting; drop it.
190 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
191 	 * tp->iss, and send a segment:
192 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
193 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
194 	 * Fill in remote peer address fields if not previously specified.
195 	 * Enter SYN_RECEIVED state, and process any other fields of this
196 	 * segment in this state.
197 	 */
198 	case TCPS_LISTEN: {
199 		struct mbuf *am;
200 		register struct sockaddr_in *sin;
201 
202 		if (tiflags & TH_RST)
203 			goto drop;
204 		if (tiflags & TH_ACK)
205 			goto dropwithreset;
206 		if ((tiflags & TH_SYN) == 0)
207 			goto drop;
208 		am = m_get(M_DONTWAIT, MT_SONAME);
209 		if (am == NULL)
210 			goto drop;
211 		am->m_len = sizeof (struct sockaddr_in);
212 		sin = mtod(am, struct sockaddr_in *);
213 		sin->sin_family = AF_INET;
214 		sin->sin_addr = ti->ti_src;
215 		sin->sin_port = ti->ti_sport;
216 		laddr = inp->inp_laddr;
217 		if (inp->inp_laddr.s_addr == INADDR_ANY)
218 			inp->inp_laddr = ti->ti_dst;
219 		if (in_pcbconnect(inp, am)) {
220 			inp->inp_laddr = laddr;
221 			(void) m_free(am);
222 			goto drop;
223 		}
224 		(void) m_free(am);
225 		tp->t_template = tcp_template(tp);
226 		if (tp->t_template == 0) {
227 			in_pcbdisconnect(inp);
228 			inp->inp_laddr = laddr;
229 			tp = 0;
230 			goto drop;
231 		}
232 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
233 		tp->irs = ti->ti_seq;
234 		tcp_sendseqinit(tp);
235 		tcp_rcvseqinit(tp);
236 		tp->t_state = TCPS_SYN_RECEIVED;
237 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
238 		goto trimthenstep6;
239 		}
240 
241 	/*
242 	 * If the state is SYN_SENT:
243 	 *	if seg contains an ACK, but not for our SYN, drop the input.
244 	 *	if seg contains a RST, then drop the connection.
245 	 *	if seg does not contain SYN, then drop it.
246 	 * Otherwise this is an acceptable SYN segment
247 	 *	initialize tp->rcv_nxt and tp->irs
248 	 *	if seg contains ack then advance tp->snd_una
249 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
250 	 *	arrange for segment to be acked (eventually)
251 	 *	continue processing rest of data/controls, beginning with URG
252 	 */
253 	case TCPS_SYN_SENT:
254 		if ((tiflags & TH_ACK) &&
255 /* this should be SEQ_LT; is SEQ_LEQ for BBN vax TCP only */
256 		    (SEQ_LT(ti->ti_ack, tp->iss) ||
257 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
258 			goto dropwithreset;
259 		if (tiflags & TH_RST) {
260 			if (tiflags & TH_ACK)
261 				tp = tcp_drop(tp, ECONNREFUSED);
262 			goto drop;
263 		}
264 		if ((tiflags & TH_SYN) == 0)
265 			goto drop;
266 		tp->snd_una = ti->ti_ack;
267 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
268 			tp->snd_nxt = tp->snd_una;
269 		tp->t_timer[TCPT_REXMT] = 0;
270 		tp->irs = ti->ti_seq;
271 		tcp_rcvseqinit(tp);
272 		tp->t_flags |= TF_ACKNOW;
273 		if (SEQ_GT(tp->snd_una, tp->iss)) {
274 			soisconnected(so);
275 			tp->t_state = TCPS_ESTABLISHED;
276 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
277 		} else
278 			tp->t_state = TCPS_SYN_RECEIVED;
279 		goto trimthenstep6;
280 
281 trimthenstep6:
282 		/*
283 		 * Advance ti->ti_seq to correspond to first data byte.
284 		 * If data, trim to stay within window,
285 		 * dropping FIN if necessary.
286 		 */
287 		ti->ti_seq++;
288 		if (ti->ti_len > tp->rcv_wnd) {
289 			todrop = ti->ti_len - tp->rcv_wnd;
290 			m_adj(m, -todrop);
291 			ti->ti_len = tp->rcv_wnd;
292 			ti->ti_flags &= ~TH_FIN;
293 		}
294 		tp->snd_wl1 = ti->ti_seq - 1;
295 		goto step6;
296 	}
297 
298 	/*
299 	 * States other than LISTEN or SYN_SENT.
300 	 * First check that at least some bytes of segment are within
301 	 * receive window.
302 	 */
303 	if (tp->rcv_wnd == 0) {
304 		/*
305 		 * If window is closed can only take segments at
306 		 * window edge, and have to drop data and PUSH from
307 		 * incoming segments.
308 		 */
309 		if (tp->rcv_nxt != ti->ti_seq)
310 			goto dropafterack;
311 		if (ti->ti_len > 0) {
312 			m_adj(m, ti->ti_len);
313 			ti->ti_len = 0;
314 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
315 		}
316 	} else {
317 		/*
318 		 * If segment begins before rcv_nxt, drop leading
319 		 * data (and SYN); if nothing left, just ack.
320 		 */
321 		todrop = tp->rcv_nxt - ti->ti_seq;
322 		if (todrop > 0) {
323 			if (tiflags & TH_SYN) {
324 				tiflags &= ~TH_SYN;
325 				ti->ti_flags &= ~TH_SYN;
326 				ti->ti_seq++;
327 				if (ti->ti_urp > 1)
328 					ti->ti_urp--;
329 				else
330 					tiflags &= ~TH_URG;
331 				todrop--;
332 			}
333 			if (todrop > ti->ti_len ||
334 			    todrop == ti->ti_len && (tiflags&TH_FIN) == 0)
335 				goto dropafterack;
336 			m_adj(m, todrop);
337 			ti->ti_seq += todrop;
338 			ti->ti_len -= todrop;
339 			if (ti->ti_urp > todrop)
340 				ti->ti_urp -= todrop;
341 			else {
342 				tiflags &= ~TH_URG;
343 				ti->ti_flags &= ~TH_URG;
344 				ti->ti_urp = 0;
345 			}
346 		}
347 		/*
348 		 * If segment ends after window, drop trailing data
349 		 * (and PUSH and FIN); if nothing left, just ACK.
350 		 */
351 		todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
352 		if (todrop > 0) {
353 			if (todrop >= ti->ti_len)
354 				goto dropafterack;
355 			m_adj(m, -todrop);
356 			ti->ti_len -= todrop;
357 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
358 		}
359 	}
360 
361 	/*
362 	 * If data is received on a connection after the
363 	 * user processes are gone, then RST the other end.
364 	 */
365 	if ((so->so_state & SS_NOFDREF) && tp->t_state > TCPS_CLOSE_WAIT &&
366 	    ti->ti_len) {
367 		tp = tcp_close(tp);
368 		goto dropwithreset;
369 	}
370 
371 	/*
372 	 * If the RST bit is set examine the state:
373 	 *    SYN_RECEIVED STATE:
374 	 *	If passive open, return to LISTEN state.
375 	 *	If active open, inform user that connection was refused.
376 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
377 	 *	Inform user that connection was reset, and close tcb.
378 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
379 	 *	Close the tcb.
380 	 */
381 	if (tiflags&TH_RST) switch (tp->t_state) {
382 
383 	case TCPS_SYN_RECEIVED:
384 		tp = tcp_drop(tp, ECONNREFUSED);
385 		goto drop;
386 
387 	case TCPS_ESTABLISHED:
388 	case TCPS_FIN_WAIT_1:
389 	case TCPS_FIN_WAIT_2:
390 	case TCPS_CLOSE_WAIT:
391 		tp = tcp_drop(tp, ECONNRESET);
392 		goto drop;
393 
394 	case TCPS_CLOSING:
395 	case TCPS_LAST_ACK:
396 	case TCPS_TIME_WAIT:
397 		tp = tcp_close(tp);
398 		goto drop;
399 	}
400 
401 	/*
402 	 * If a SYN is in the window, then this is an
403 	 * error and we send an RST and drop the connection.
404 	 */
405 	if (tiflags & TH_SYN) {
406 		tp = tcp_drop(tp, ECONNRESET);
407 		goto dropwithreset;
408 	}
409 
410 	/*
411 	 * If the ACK bit is off we drop the segment and return.
412 	 */
413 	if ((tiflags & TH_ACK) == 0)
414 		goto drop;
415 
416 	/*
417 	 * Ack processing.
418 	 */
419 	switch (tp->t_state) {
420 
421 	/*
422 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
423 	 * ESTABLISHED state and continue processing, othewise
424 	 * send an RST.
425 	 */
426 	case TCPS_SYN_RECEIVED:
427 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
428 		    SEQ_GT(ti->ti_ack, tp->snd_max))
429 			goto dropwithreset;
430 		tp->snd_una++;			/* SYN acked */
431 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
432 			tp->snd_nxt = tp->snd_una;
433 		tp->t_timer[TCPT_REXMT] = 0;
434 		soisconnected(so);
435 		tp->t_state = TCPS_ESTABLISHED;
436 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
437 		tp->snd_wl1 = ti->ti_seq - 1;
438 		/* fall into ... */
439 
440 	/*
441 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
442 	 * ACKs.  If the ack is in the range
443 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
444 	 * then advance tp->snd_una to ti->ti_ack and drop
445 	 * data from the retransmission queue.  If this ACK reflects
446 	 * more up to date window information we update our window information.
447 	 */
448 	case TCPS_ESTABLISHED:
449 	case TCPS_FIN_WAIT_1:
450 	case TCPS_FIN_WAIT_2:
451 	case TCPS_CLOSE_WAIT:
452 	case TCPS_CLOSING:
453 	case TCPS_LAST_ACK:
454 	case TCPS_TIME_WAIT:
455 #define	ourfinisacked	(acked > 0)
456 
457 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
458 			break;
459 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
460 			goto dropafterack;
461 		acked = ti->ti_ack - tp->snd_una;
462 
463 		/*
464 		 * If transmit timer is running and timed sequence
465 		 * number was acked, update smoothed round trip time.
466 		 */
467 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
468 			if (tp->t_srtt == 0)
469 				tp->t_srtt = tp->t_rtt;
470 			else
471 				tp->t_srtt =
472 				    tcp_alpha * tp->t_srtt +
473 				    (1 - tcp_alpha) * tp->t_rtt;
474 			tp->t_rtt = 0;
475 		}
476 
477 		if (ti->ti_ack == tp->snd_max)
478 			tp->t_timer[TCPT_REXMT] = 0;
479 		else {
480 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
481 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
482 			tp->t_rtt = 1;
483 			tp->t_rxtshift = 0;
484 		}
485 		if (acked > so->so_snd.sb_cc) {
486 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
487 			tp->snd_wnd -= so->so_snd.sb_cc;
488 		} else {
489 			sbdrop(&so->so_snd, acked);
490 			tp->snd_wnd -= acked;
491 			acked = 0;
492 		}
493 		if ((so->so_snd.sb_flags & SB_WAIT) || so->so_snd.sb_sel)
494 			sowwakeup(so);
495 		tp->snd_una = ti->ti_ack;
496 		if (SEQ_LT(tp->snd_nxt, tp->snd_una))
497 			tp->snd_nxt = tp->snd_una;
498 
499 		switch (tp->t_state) {
500 
501 		/*
502 		 * In FIN_WAIT_1 STATE in addition to the processing
503 		 * for the ESTABLISHED state if our FIN is now acknowledged
504 		 * then enter FIN_WAIT_2.
505 		 */
506 		case TCPS_FIN_WAIT_1:
507 			if (ourfinisacked) {
508 				/*
509 				 * If we can't receive any more
510 				 * data, then closing user can proceed.
511 				 */
512 				if (so->so_state & SS_CANTRCVMORE)
513 					soisdisconnected(so);
514 				tp->t_state = TCPS_FIN_WAIT_2;
515 			}
516 			break;
517 
518 	 	/*
519 		 * In CLOSING STATE in addition to the processing for
520 		 * the ESTABLISHED state if the ACK acknowledges our FIN
521 		 * then enter the TIME-WAIT state, otherwise ignore
522 		 * the segment.
523 		 */
524 		case TCPS_CLOSING:
525 			if (ourfinisacked) {
526 				tp->t_state = TCPS_TIME_WAIT;
527 				tcp_canceltimers(tp);
528 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
529 				soisdisconnected(so);
530 			}
531 			break;
532 
533 		/*
534 		 * The only thing that can arrive in  LAST_ACK state
535 		 * is an acknowledgment of our FIN.  If our FIN is now
536 		 * acknowledged, delete the TCB, enter the closed state
537 		 * and return.
538 		 */
539 		case TCPS_LAST_ACK:
540 			if (ourfinisacked)
541 				tp = tcp_close(tp);
542 			goto drop;
543 
544 		/*
545 		 * In TIME_WAIT state the only thing that should arrive
546 		 * is a retransmission of the remote FIN.  Acknowledge
547 		 * it and restart the finack timer.
548 		 */
549 		case TCPS_TIME_WAIT:
550 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
551 			goto dropafterack;
552 		}
553 #undef ourfinisacked
554 	}
555 
556 step6:
557 	/*
558 	 * Update window information.
559 	 */
560 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) || tp->snd_wl1 == ti->ti_seq &&
561 	    (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
562 	     tp->snd_wl2 == ti->ti_ack && ti->ti_win > tp->snd_wnd)) {
563 		tp->snd_wnd = ti->ti_win;
564 		tp->snd_wl1 = ti->ti_seq;
565 		tp->snd_wl2 = ti->ti_ack;
566 		if (tp->snd_wnd != 0)
567 			tp->t_timer[TCPT_PERSIST] = 0;
568 	}
569 
570 	/*
571 	 * Process segments with URG.
572 	 */
573 	if ((tiflags & TH_URG) && ti->ti_urp &&
574 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
575 		/*
576 		 * If this segment advances the known urgent pointer,
577 		 * then mark the data stream.  This should not happen
578 		 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
579 		 * a FIN has been received from the remote side.
580 		 * In these states we ignore the URG.
581 		 */
582 		if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
583 			tp->rcv_up = ti->ti_seq + ti->ti_urp;
584 			so->so_oobmark = so->so_rcv.sb_cc +
585 			    (tp->rcv_up - tp->rcv_nxt) - 1;
586 			if (so->so_oobmark == 0)
587 				so->so_state |= SS_RCVATMARK;
588 			sohasoutofband(so);
589 			tp->t_oobflags &= ~TCPOOB_HAVEDATA;
590 		}
591 		/*
592 		 * Remove out of band data so doesn't get presented to user.
593 		 * This can happen independent of advancing the URG pointer,
594 		 * but if two URG's are pending at once, some out-of-band
595 		 * data may creep in... ick.
596 		 */
597 		if (ti->ti_urp <= ti->ti_len)
598 			tcp_pulloutofband(so, ti);
599 	}
600 
601 	/*
602 	 * Process the segment text, merging it into the TCP sequencing queue,
603 	 * and arranging for acknowledgment of receipt if necessary.
604 	 * This process logically involves adjusting tp->rcv_wnd as data
605 	 * is presented to the user (this happens in tcp_usrreq.c,
606 	 * case PRU_RCVD).  If a FIN has already been received on this
607 	 * connection then we just ignore the text.
608 	 */
609 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
610 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
611 		tiflags = tcp_reass(tp, ti);
612 		if (tcpnodelack == 0)
613 			tp->t_flags |= TF_DELACK;
614 		else
615 			tp->t_flags |= TF_ACKNOW;
616 	} else {
617 		m_freem(m);
618 		tiflags &= ~TH_FIN;
619 	}
620 
621 	/*
622 	 * If FIN is received ACK the FIN and let the user know
623 	 * that the connection is closing.
624 	 */
625 	if (tiflags & TH_FIN) {
626 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
627 			socantrcvmore(so);
628 			tp->t_flags |= TF_ACKNOW;
629 			tp->rcv_nxt++;
630 		}
631 		switch (tp->t_state) {
632 
633 	 	/*
634 		 * In SYN_RECEIVED and ESTABLISHED STATES
635 		 * enter the CLOSE_WAIT state.
636 		 */
637 		case TCPS_SYN_RECEIVED:
638 		case TCPS_ESTABLISHED:
639 			tp->t_state = TCPS_CLOSE_WAIT;
640 			break;
641 
642 	 	/*
643 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
644 		 * enter the CLOSING state.
645 		 */
646 		case TCPS_FIN_WAIT_1:
647 			tp->t_state = TCPS_CLOSING;
648 			break;
649 
650 	 	/*
651 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
652 		 * starting the time-wait timer, turning off the other
653 		 * standard timers.
654 		 */
655 		case TCPS_FIN_WAIT_2:
656 			tp->t_state = TCPS_TIME_WAIT;
657 			tcp_canceltimers(tp);
658 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
659 			soisdisconnected(so);
660 			break;
661 
662 		/*
663 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
664 		 */
665 		case TCPS_TIME_WAIT:
666 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
667 			break;
668 		}
669 	}
670 	if (so->so_options & SO_DEBUG)
671 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
672 
673 	/*
674 	 * Return any desired output.
675 	 */
676 	(void) tcp_output(tp);
677 	return;
678 
679 dropafterack:
680 	/*
681 	 * Generate an ACK dropping incoming segment if it occupies
682 	 * sequence space, where the ACK reflects our state.
683 	 */
684 	if ((tiflags&TH_RST) ||
685 	    tlen == 0 && (tiflags&(TH_SYN|TH_FIN)) == 0)
686 		goto drop;
687 	if (tp->t_inpcb->inp_socket->so_options & SO_DEBUG)
688 		tcp_trace(TA_RESPOND, ostate, tp, &tcp_saveti, 0);
689 	tcp_respond(tp, ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
690 	return;
691 
692 dropwithreset:
693 	if (om)
694 		(void) m_free(om);
695 	/*
696 	 * Generate a RST, dropping incoming segment.
697 	 * Make ACK acceptable to originator of segment.
698 	 */
699 	if (tiflags & TH_RST)
700 		goto drop;
701 	if (tiflags & TH_ACK)
702 		tcp_respond(tp, ti, (tcp_seq)0, ti->ti_ack, TH_RST);
703 	else {
704 		if (tiflags & TH_SYN)
705 			ti->ti_len++;
706 		tcp_respond(tp, ti, ti->ti_seq+ti->ti_len, (tcp_seq)0,
707 		    TH_RST|TH_ACK);
708 	}
709 	return;
710 
711 drop:
712 	/*
713 	 * Drop space held by incoming segment and return.
714 	 */
715 	if (tp && (tp->t_inpcb->inp_socket->so_options & SO_DEBUG))
716 		tcp_trace(TA_DROP, ostate, tp, &tcp_saveti, 0);
717 	m_freem(m);
718 	return;
719 }
720 
721 tcp_dooptions(tp, om)
722 	struct tcpcb *tp;
723 	struct mbuf *om;
724 {
725 	register u_char *cp;
726 	int opt, optlen, cnt;
727 
728 	cp = mtod(om, u_char *);
729 	cnt = om->m_len;
730 	for (; cnt > 0; cnt -= optlen, cp += optlen) {
731 		opt = cp[0];
732 		if (opt == TCPOPT_EOL)
733 			break;
734 		if (opt == TCPOPT_NOP)
735 			optlen = 1;
736 		else
737 			optlen = cp[1];
738 		switch (opt) {
739 
740 		default:
741 			break;
742 
743 		case TCPOPT_MAXSEG:
744 			if (optlen != 4)
745 				continue;
746 			tp->t_maxseg = *(u_short *)(cp + 2);
747 			tp->t_maxseg = ntohs((u_short)tp->t_maxseg);
748 			break;
749 		}
750 	}
751 	(void) m_free(om);
752 }
753 
754 /*
755  * Pull out of band byte out of a segment so
756  * it doesn't appear in the user's data queue.
757  * It is still reflected in the segment length for
758  * sequencing purposes.
759  */
760 tcp_pulloutofband(so, ti)
761 	struct socket *so;
762 	struct tcpiphdr *ti;
763 {
764 	register struct mbuf *m;
765 	int cnt = ti->ti_urp - 1;
766 
767 	m = dtom(ti);
768 	while (cnt >= 0) {
769 		if (m->m_len > cnt) {
770 			char *cp = mtod(m, caddr_t) + cnt;
771 			struct tcpcb *tp = sototcpcb(so);
772 
773 			tp->t_iobc = *cp;
774 			tp->t_oobflags |= TCPOOB_HAVEDATA;
775 			bcopy(cp+1, cp, (unsigned)(m->m_len - cnt - 1));
776 			m->m_len--;
777 			return;
778 		}
779 		cnt -= m->m_len;
780 		m = m->m_next;
781 		if (m == 0)
782 			break;
783 	}
784 	panic("tcp_pulloutofband");
785 }
786 
787 /*
788  * Insert segment ti into reassembly queue of tcp with
789  * control block tp.  Return TH_FIN if reassembly now includes
790  * a segment with FIN.
791  */
792 tcp_reass(tp, ti)
793 	register struct tcpcb *tp;
794 	register struct tcpiphdr *ti;
795 {
796 	register struct tcpiphdr *q;
797 	struct socket *so = tp->t_inpcb->inp_socket;
798 	struct mbuf *m;
799 	int flags;
800 
801 	/*
802 	 * Call with ti==0 after become established to
803 	 * force pre-ESTABLISHED data up to user socket.
804 	 */
805 	if (ti == 0)
806 		goto present;
807 
808 	/*
809 	 * Find a segment which begins after this one does.
810 	 */
811 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
812 	    q = (struct tcpiphdr *)q->ti_next)
813 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
814 			break;
815 
816 	/*
817 	 * If there is a preceding segment, it may provide some of
818 	 * our data already.  If so, drop the data from the incoming
819 	 * segment.  If it provides all of our data, drop us.
820 	 */
821 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
822 		register int i;
823 		q = (struct tcpiphdr *)q->ti_prev;
824 		/* conversion to int (in i) handles seq wraparound */
825 		i = q->ti_seq + q->ti_len - ti->ti_seq;
826 		if (i > 0) {
827 			if (i >= ti->ti_len)
828 				goto drop;
829 			m_adj(dtom(ti), i);
830 			ti->ti_len -= i;
831 			ti->ti_seq += i;
832 		}
833 		q = (struct tcpiphdr *)(q->ti_next);
834 	}
835 
836 	/*
837 	 * While we overlap succeeding segments trim them or,
838 	 * if they are completely covered, dequeue them.
839 	 */
840 	while (q != (struct tcpiphdr *)tp) {
841 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
842 		if (i <= 0)
843 			break;
844 		if (i < q->ti_len) {
845 			q->ti_seq += i;
846 			q->ti_len -= i;
847 			m_adj(dtom(q), i);
848 			break;
849 		}
850 		q = (struct tcpiphdr *)q->ti_next;
851 		m = dtom(q->ti_prev);
852 		remque(q->ti_prev);
853 		m_freem(m);
854 	}
855 
856 	/*
857 	 * Stick new segment in its place.
858 	 */
859 	insque(ti, q->ti_prev);
860 
861 present:
862 	/*
863 	 * Present data to user, advancing rcv_nxt through
864 	 * completed sequence space.
865 	 */
866 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
867 		return (0);
868 	ti = tp->seg_next;
869 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
870 		return (0);
871 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
872 		return (0);
873 	do {
874 		tp->rcv_nxt += ti->ti_len;
875 		flags = ti->ti_flags & TH_FIN;
876 		remque(ti);
877 		m = dtom(ti);
878 		ti = (struct tcpiphdr *)ti->ti_next;
879 		if (so->so_state & SS_CANTRCVMORE)
880 			m_freem(m);
881 		else
882 			sbappend(&so->so_rcv, m);
883 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
884 	sorwakeup(so);
885 	return (flags);
886 drop:
887 	m_freem(dtom(ti));
888 	return (0);
889 }
890