xref: /csrg-svn/sys/netinet/tcp_input.c (revision 5267)
1 /*	tcp_input.c	1.42	81/12/20	*/
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	tcpcksum = 1;
25 struct	sockaddr_in tcp_in = { AF_INET };
26 struct	tcpiphdr tcp_saveti;
27 
28 struct	tcpcb *tcp_newtcpcb();
29 /*
30  * TCP input routine, follows pages 65-76 of the
31  * protocol specification dated September, 1981 very closely.
32  */
33 tcp_input(m0)
34 	struct mbuf *m0;
35 {
36 	register struct tcpiphdr *ti;
37 	struct inpcb *inp;
38 	register struct mbuf *m;
39 	int len, tlen, off;
40 	register struct tcpcb *tp;
41 	register int tiflags;
42 	struct socket *so;
43 	int todrop, acked;
44 	short ostate;
45 
46 COUNT(TCP_INPUT);
47 	/*
48 	 * Get IP and TCP header together in first mbuf.
49 	 * Note: IP leaves IP header in first mbuf.
50 	 */
51 	m = m0;
52 	ti = mtod(m, struct tcpiphdr *);
53 	if (((struct ip *)ti)->ip_hl > (sizeof (struct ip) >> 2))
54 		ip_stripoptions((struct ip *)ti, (struct mbuf *)0);
55 	if (m->m_len < sizeof (struct tcpiphdr)) {
56 		if (m_pullup(m, sizeof (struct tcpiphdr)) == 0) {
57 			tcpstat.tcps_hdrops++;
58 			goto drop;
59 		}
60 		ti = mtod(m, struct tcpiphdr *);
61 	}
62 
63 	/*
64 	 * Checksum extended TCP header and data.
65 	 */
66 	tlen = ((struct ip *)ti)->ip_len;
67 	len = sizeof (struct ip) + tlen;
68 	if (tcpcksum) {
69 		ti->ti_next = ti->ti_prev = 0;
70 		ti->ti_x1 = 0;
71 		ti->ti_len = (u_short)tlen;
72 #if vax
73 		ti->ti_len = htons(ti->ti_len);
74 #endif
75 		if (ti->ti_sum = in_cksum(m, len)) {
76 			tcpstat.tcps_badsum++;
77 			printf("tcp cksum %x\n", ti->ti_sum);
78 			goto drop;
79 		}
80 	}
81 
82 	/*
83 	 * Check that TCP offset makes sense,
84 	 * process TCP options and adjust length.
85 	 */
86 	off = ti->ti_off << 2;
87 	if (off < sizeof (struct tcphdr) || off > tlen) {
88 		tcpstat.tcps_badoff++;
89 		goto drop;
90 	}
91 	ti->ti_len = tlen - off;
92 #if 0
93 	if (off > sizeof (struct tcphdr))
94 		tcp_options(ti);
95 #endif
96 	tiflags = ti->ti_flags;
97 
98 #if vax
99 	/*
100 	 * Convert TCP protocol specific fields to host format.
101 	 */
102 	ti->ti_seq = ntohl(ti->ti_seq);
103 	ti->ti_ack = ntohl(ti->ti_ack);
104 	ti->ti_win = ntohs(ti->ti_win);
105 	ti->ti_urp = ntohs(ti->ti_urp);
106 #endif
107 
108 	/*
109 	 * Locate pcb for segment.
110 	 */
111 	inp = in_pcblookup
112 		(&tcb, ti->ti_src, ti->ti_sport, ti->ti_dst, ti->ti_dport);
113 
114 	/*
115 	 * If the state is CLOSED (i.e., TCB does not exist) then
116 	 * all data in the incoming segment is discarded.
117 	 */
118 	if (inp == 0)
119 		goto dropwithreset;
120 	tp = intotcpcb(inp);
121 	if (tp == 0)
122 		goto dropwithreset;
123 	so = inp->inp_socket;
124 	if (so->so_options & SO_DEBUG) {
125 		ostate = tp->t_state;
126 		tcp_saveti = *ti;
127 	}
128 
129 
130 	/*
131 	 * Segment received on connection.
132 	 * Reset idle time and keep-alive timer.
133 	 */
134 	tp->t_idle = 0;
135 	tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
136 
137 	/*
138 	 * Calculate amount of space in receive window,
139 	 * and then do TCP input processing.
140 	 */
141 	tp->rcv_wnd = sbspace(&so->so_rcv);
142 	if (tp->rcv_wnd < 0)
143 		tp->rcv_wnd = 0;
144 
145 	switch (tp->t_state) {
146 
147 	/*
148 	 * If the state is LISTEN then ignore segment if it contains an RST.
149 	 * If the segment contains an ACK then it is bad and send a RST.
150 	 * If it does not contain a SYN then it is not interesting; drop it.
151 	 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
152 	 * tp->iss, and send a segment:
153 	 *     <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
154 	 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
155 	 * Fill in remote peer address fields if not previously specified.
156 	 * Enter SYN_RECEIVED state, and process any other fields of this
157 	 * segment in this state.
158 	 */
159 	case TCPS_LISTEN:
160 		if (tiflags & TH_RST)
161 			goto drop;
162 		if (tiflags & TH_ACK)
163 			goto dropwithreset;
164 		if ((tiflags & TH_SYN) == 0)
165 			goto drop;
166 		tcp_in.sin_addr = ti->ti_src;
167 		tcp_in.sin_port = ti->ti_sport;
168 		if (in_pcbconnect(inp, (struct sockaddr *)&tcp_in))
169 			goto drop;
170 		tp->t_template = tcp_template(tp);
171 		if (tp->t_template == 0) {
172 			in_pcbdisconnect(inp);
173 			goto drop;
174 		}
175 		tp->iss = tcp_iss; tcp_iss += TCP_ISSINCR/2;
176 		tp->irs = ti->ti_seq;
177 		tcp_sendseqinit(tp);
178 		tcp_rcvseqinit(tp);
179 		tp->t_state = TCPS_SYN_RECEIVED;
180 		tp->t_timer[TCPT_KEEP] = TCPTV_KEEP;
181 		goto trimthenstep6;
182 
183 	/*
184 	 * If the state is SYN_SENT:
185 	 *	if seg contains an ACK, but not for our SYN, drop the input.
186 	 *	if seg contains a RST, then drop the connection.
187 	 *	if seg does not contain SYN, then drop it.
188 	 * Otherwise this is an acceptable SYN segment
189 	 *	initialize tp->rcv_nxt and tp->irs
190 	 *	if seg contains ack then advance tp->snd_una
191 	 *	if SYN has been acked change to ESTABLISHED else SYN_RCVD state
192 	 *	arrange for segment to be acked (eventually)
193 	 *	continue processing rest of data/controls, beginning with URG
194 	 */
195 	case TCPS_SYN_SENT:
196 		if ((tiflags & TH_ACK) &&
197 		    (SEQ_LEQ(ti->ti_ack, tp->iss) ||
198 		     SEQ_GT(ti->ti_ack, tp->snd_max)))
199 			goto dropwithreset;
200 		if (tiflags & TH_RST) {
201 			if (tiflags & TH_ACK)
202 				tcp_drop(tp, ECONNREFUSED);
203 			goto drop;
204 		}
205 		if ((tiflags & TH_SYN) == 0)
206 			goto drop;
207 		tp->snd_una = ti->ti_ack;
208 		tp->t_timer[TCPT_REXMT] = 0;
209 		tp->irs = ti->ti_seq;
210 		tcp_rcvseqinit(tp);
211 		tp->t_flags |= TF_ACKNOW;
212 		if (SEQ_GT(tp->snd_una, tp->iss)) {
213 			so->so_state |= SS_CONNAWAITING;
214 			soisconnected(so);
215 			tp->t_state = TCPS_ESTABLISHED;
216 			(void) tcp_reass(tp, (struct tcpiphdr *)0);
217 		} else
218 			tp->t_state = TCPS_SYN_RECEIVED;
219 		goto trimthenstep6;
220 
221 trimthenstep6:
222 		/*
223 		 * Advance ti->ti_seq to correspond to first data byte.
224 		 * If data, trim to stay within window,
225 		 * dropping FIN if necessary.
226 		 */
227 		ti->ti_seq++;
228 		if (ti->ti_len > tp->rcv_wnd) {
229 			todrop = ti->ti_len - tp->rcv_wnd;
230 			m_adj(m, -todrop);
231 			ti->ti_len = tp->rcv_wnd;
232 			ti->ti_flags &= ~TH_FIN;
233 		}
234 		tp->snd_wl1 = ti->ti_seq - 1;
235 		goto step6;
236 	}
237 
238 	/*
239 	 * States other than LISTEN or SYN_SENT.
240 	 * First check that at least some bytes of segment are within
241 	 * receive window.
242 	 */
243 	if (tp->rcv_wnd == 0) {
244 		/*
245 		 * If window is closed can only take segments at
246 		 * window edge, and have to drop data and PUSH from
247 		 * incoming segments.
248 		 */
249 		if (tp->rcv_nxt != ti->ti_seq)
250 			goto dropafterack;
251 		if (ti->ti_len > 0) {
252 			ti->ti_len = 0;
253 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
254 		}
255 	} else {
256 		/*
257 		 * If segment begins before rcv_nxt, drop leading
258 		 * data (and SYN); if nothing left, just ack.
259 		 */
260 		if (SEQ_GT(tp->rcv_nxt, ti->ti_seq)) {
261 			todrop = tp->rcv_nxt - ti->ti_seq;
262 			if (tiflags & TH_SYN) {
263 				ti->ti_seq++;
264 				if (ti->ti_urp > 1)
265 					ti->ti_urp--;
266 				else
267 					tiflags &= ~TH_URG;
268 				todrop--;
269 			}
270 			if (todrop > ti->ti_len)
271 				goto dropafterack;
272 			m_adj(m, todrop);
273 			ti->ti_seq += todrop;
274 			ti->ti_len -= todrop;
275 			if (ti->ti_urp > todrop)
276 				ti->ti_urp -= todrop;
277 			else {
278 				tiflags &= ~TH_URG;
279 				/* ti->ti_flags &= ~TH_URG; */
280 				/* ti->ti_urp = 0; */
281 			}
282 			/* tiflags &= ~TH_SYN; */
283 			/* ti->ti_flags &= ~TH_SYN; */
284 		}
285 		/*
286 		 * If segment ends after window, drop trailing data
287 		 * (and PUSH and FIN); if nothing left, just ACK.
288 		 */
289 		if (SEQ_GT(ti->ti_seq+ti->ti_len, tp->rcv_nxt+tp->rcv_wnd)) {
290 			todrop =
291 			     ti->ti_seq+ti->ti_len - (tp->rcv_nxt+tp->rcv_wnd);
292 			if (todrop > ti->ti_len)
293 				goto dropafterack;
294 			m_adj(m, -todrop);
295 			ti->ti_len -= todrop;
296 			ti->ti_flags &= ~(TH_PUSH|TH_FIN);
297 		}
298 	}
299 
300 	/*
301 	 * If the RST bit is set examine the state:
302 	 *    SYN_RECEIVED STATE:
303 	 *	If passive open, return to LISTEN state.
304 	 *	If active open, inform user that connection was refused.
305 	 *    ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
306 	 *	Inform user that connection was reset, and close tcb.
307 	 *    CLOSING, LAST_ACK, TIME_WAIT STATES
308 	 *	Close the tcb.
309 	 */
310 	if (tiflags&TH_RST) switch (tp->t_state) {
311 
312 	case TCPS_SYN_RECEIVED:
313 		if (inp->inp_socket->so_options & SO_ACCEPTCONN) {
314 			/* a miniature tcp_close, but invisible to user */
315 			(void) m_free(dtom(tp->t_template));
316 			(void) m_free(dtom(tp));
317 			inp->inp_ppcb = 0;
318 			tp = tcp_newtcpcb(inp);
319 			tp->t_state = TCPS_LISTEN;
320 			goto drop;
321 		}
322 		tcp_drop(tp, ECONNREFUSED);
323 		goto drop;
324 
325 	case TCPS_ESTABLISHED:
326 	case TCPS_FIN_WAIT_1:
327 	case TCPS_FIN_WAIT_2:
328 	case TCPS_CLOSE_WAIT:
329 		tcp_drop(tp, ECONNRESET);
330 		goto drop;
331 
332 	case TCPS_CLOSING:
333 	case TCPS_LAST_ACK:
334 	case TCPS_TIME_WAIT:
335 		tcp_close(tp);
336 		goto drop;
337 	}
338 
339 	/*
340 	 * If a SYN is in the window, then this is an
341 	 * error and we send an RST and drop the connection.
342 	 */
343 	if (tiflags & TH_SYN) {
344 		tcp_drop(tp, ECONNRESET);
345 		goto dropwithreset;
346 	}
347 
348 	/*
349 	 * If the ACK bit is off we drop the segment and return.
350 	 */
351 	if ((tiflags & TH_ACK) == 0)
352 		goto drop;
353 
354 	/*
355 	 * Ack processing.
356 	 */
357 	switch (tp->t_state) {
358 
359 	/*
360 	 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
361 	 * ESTABLISHED state and continue processing, othewise
362 	 * send an RST.
363 	 */
364 	case TCPS_SYN_RECEIVED:
365 		if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
366 		    SEQ_GT(ti->ti_ack, tp->snd_max))
367 			goto dropwithreset;
368 		tp->snd_una++;			/* SYN acked */
369 		tp->t_timer[TCPT_REXMT] = 0;
370 		so->so_state |= SS_CONNAWAITING;
371 		soisconnected(so);
372 		tp->t_state = TCPS_ESTABLISHED;
373 		(void) tcp_reass(tp, (struct tcpiphdr *)0);
374 		tp->snd_wl1 = ti->ti_seq - 1;
375 		/* fall into ... */
376 
377 	/*
378 	 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
379 	 * ACKs.  If the ack is in the range
380 	 *	tp->snd_una < ti->ti_ack <= tp->snd_max
381 	 * then advance tp->snd_una to ti->ti_ack and drop
382 	 * data from the retransmission queue.  If this ACK reflects
383 	 * more up to date window information we update our window information.
384 	 */
385 	case TCPS_ESTABLISHED:
386 	case TCPS_FIN_WAIT_1:
387 	case TCPS_FIN_WAIT_2:
388 	case TCPS_CLOSE_WAIT:
389 	case TCPS_CLOSING:
390 	case TCPS_LAST_ACK:
391 	case TCPS_TIME_WAIT:
392 #define	ourfinisacked	(acked > 0)
393 
394 		if (SEQ_LEQ(ti->ti_ack, tp->snd_una))
395 			break;
396 		if (SEQ_GT(ti->ti_ack, tp->snd_max))
397 			goto dropafterack;
398 		acked = ti->ti_ack - tp->snd_una;
399 		if (acked >= so->so_snd.sb_cc) {
400 			acked -= so->so_snd.sb_cc;
401 			/* if acked > 0 our FIN is acked */
402 if (so->so_snd.sb_cc)
403 			sbdrop(&so->so_snd, so->so_snd.sb_cc);
404 			tp->t_timer[TCPT_REXMT] = 0;
405 		} else {
406 if (acked)
407 			sbdrop(&so->so_snd, acked);
408 			acked = 0;
409 			TCPT_RANGESET(tp->t_timer[TCPT_REXMT],
410 			    tcp_beta * tp->t_srtt, TCPTV_MIN, TCPTV_MAX);
411 		}
412 		tp->snd_una = ti->ti_ack;
413 
414 		/*
415 		 * If transmit timer is running and timed sequence
416 		 * number was acked, update smoothed round trip time.
417 		 */
418 		if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq)) {
419 			if (tp->t_srtt == 0)
420 				tp->t_srtt = tp->t_rtt;
421 			else
422 				tp->t_srtt =
423 				    tcp_alpha * tp->t_srtt +
424 				    (1 - tcp_alpha) * tp->t_rtt;
425 			tp->t_rtt = 0;
426 		}
427 
428 		switch (tp->t_state) {
429 
430 		/*
431 		 * In FIN_WAIT_1 STATE in addition to the processing
432 		 * for the ESTABLISHED state if our FIN is now acknowledged
433 		 * then enter FIN_WAIT_2.
434 		 */
435 		case TCPS_FIN_WAIT_1:
436 			if (ourfinisacked)
437 				tp->t_state = TCPS_FIN_WAIT_2;
438 			break;
439 
440 	 	/*
441 		 * In CLOSING STATE in addition to the processing for
442 		 * the ESTABLISHED state if the ACK acknowledges our FIN
443 		 * then enter the TIME-WAIT state, otherwise ignore
444 		 * the segment.
445 		 */
446 		case TCPS_CLOSING:
447 			if (ourfinisacked) {
448 				tp->t_state = TCPS_TIME_WAIT;
449 				tcp_canceltimers(tp);
450 				tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
451 				soisdisconnected(so);
452 			}
453 			break;
454 
455 		/*
456 		 * The only thing that can arrive in  LAST_ACK state
457 		 * is an acknowledgment of our FIN.  If our FIN is now
458 		 * acknowledged, delete the TCB, enter the closed state
459 		 * and return.
460 		 */
461 		case TCPS_LAST_ACK:
462 			if (ourfinisacked)
463 				tcp_close(tp);
464 			goto drop;
465 
466 		/*
467 		 * In TIME_WAIT state the only thing that should arrive
468 		 * is a retransmission of the remote FIN.  Acknowledge
469 		 * it and restart the finack timer.
470 		 */
471 		case TCPS_TIME_WAIT:
472 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
473 			goto dropafterack;
474 		}
475 #undef ourfinisacked
476 	}
477 
478 step6:
479 	/*
480 	 * Update window information.
481 	 */
482 	if (SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
483 	    tp->snd_wl1==ti->ti_seq && SEQ_LEQ(tp->snd_wl2,ti->ti_seq)) {
484 		tp->snd_wnd = ti->ti_win;
485 		tp->snd_wl1 = ti->ti_seq;
486 		tp->snd_wl2 = ti->ti_ack;
487 		if (tp->snd_wnd > 0)
488 			tp->t_timer[TCPT_PERSIST] = 0;
489 	}
490 
491 	/*
492 	 * If an URG bit is set in the segment and is greater than the
493 	 * current known urgent pointer, then signal the user that the
494 	 * remote side has out of band data.  This should not happen
495 	 * in CLOSE_WAIT, CLOSING, LAST-ACK or TIME_WAIT STATES since
496 	 * a FIN has been received from the remote side.  In these states
497 	 * we ignore the URG.
498 	 */
499 	if ((tiflags & TH_URG) == 0 && TCPS_HAVERCVDFIN(tp->t_state) == 0)
500 		if (SEQ_GT(ti->ti_urp, tp->rcv_up)) {
501 			tp->rcv_up = ti->ti_urp;
502 #if 0
503 			sohasoutofband(so);		/* XXX */
504 #endif
505 		}
506 
507 	/*
508 	 * Process the segment text, merging it into the TCP sequencing queue,
509 	 * and arranging for acknowledgment of receipt if necessary.
510 	 * This process logically involves adjusting tp->rcv_wnd as data
511 	 * is presented to the user (this happens in tcp_usrreq.c,
512 	 * case PRU_RCVD).  If a FIN has already been received on this
513 	 * connection then we just ignore the text.
514 	 */
515 	if ((ti->ti_len || (tiflags&TH_FIN)) &&
516 	    TCPS_HAVERCVDFIN(tp->t_state) == 0) {
517 		off += sizeof (struct ip);		/* drop IP header */
518 		m->m_off += off;
519 		m->m_len -= off;
520 		tiflags = tcp_reass(tp, ti);
521 		tp->t_flags |= TF_ACKNOW;		/* XXX TF_DELACK */
522 	} else {
523 		m_freem(m);
524 		tiflags &= ~TH_FIN;
525 	}
526 
527 	/*
528 	 * If FIN is received ACK the FIN and let the user know
529 	 * that the connection is closing.
530 	 */
531 	if (tiflags & TH_FIN) {
532 		if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
533 			socantrcvmore(so);
534 			tp->t_flags |= TF_ACKNOW;
535 			tp->rcv_nxt++;
536 		}
537 		switch (tp->t_state) {
538 
539 	 	/*
540 		 * In SYN_RECEIVED and ESTABLISHED STATES
541 		 * enter the CLOSE_WAIT state.
542 		 */
543 		case TCPS_SYN_RECEIVED:
544 		case TCPS_ESTABLISHED:
545 			tp->t_state = TCPS_CLOSE_WAIT;
546 			break;
547 
548 	 	/*
549 		 * If still in FIN_WAIT_1 STATE FIN has not been acked so
550 		 * enter the CLOSING state.
551 		 */
552 		case TCPS_FIN_WAIT_1:
553 			tp->t_state = TCPS_CLOSING;
554 			break;
555 
556 	 	/*
557 		 * In FIN_WAIT_2 state enter the TIME_WAIT state,
558 		 * starting the time-wait timer, turning off the other
559 		 * standard timers.
560 		 */
561 		case TCPS_FIN_WAIT_2:
562 			tp->t_state = TCPS_TIME_WAIT;
563 			tcp_canceltimers(tp);
564 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
565 			soisdisconnected(so);
566 			break;
567 
568 		/*
569 		 * In TIME_WAIT state restart the 2 MSL time_wait timer.
570 		 */
571 		case TCPS_TIME_WAIT:
572 			tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
573 			break;
574 		}
575 	}
576 	if (so->so_options & SO_DEBUG)
577 		tcp_trace(TA_INPUT, ostate, tp, &tcp_saveti, 0);
578 
579 	/*
580 	 * Return any desired output.
581 	 */
582 	tcp_output(tp);
583 	return;
584 
585 dropafterack:
586 	/*
587 	 * Generate an ACK dropping incoming segment.
588 	 * Make ACK reflect our state.
589 	 */
590 	if (tiflags & TH_RST)
591 		goto drop;
592 	tcp_respond(ti, tp->rcv_nxt, tp->snd_nxt, TH_ACK);
593 	return;
594 
595 dropwithreset:
596 	/*
597 	 * Generate a RST, dropping incoming segment.
598 	 * Make ACK acceptable to originator of segment.
599 	 */
600 	if (tiflags & TH_RST)
601 		goto drop;
602 	if (tiflags & TH_ACK)
603 		tcp_respond(ti, (tcp_seq)0, ti->ti_ack, TH_RST);
604 	else {
605 		if (tiflags & TH_SYN)
606 			ti->ti_len++;
607 		tcp_respond(ti, ti->ti_seq+ti->ti_len, (tcp_seq)0, TH_RST|TH_ACK);
608 	}
609 	return;
610 
611 drop:
612 	/*
613 	 * Drop space held by incoming segment and return.
614 	 */
615 	m_freem(m);
616 	return;
617 }
618 
619 /*
620  * Insert segment ti into reassembly queue of tcp with
621  * control block tp.  Return TH_FIN if reassembly now includes
622  * a segment with FIN.
623  */
624 tcp_reass(tp, ti)
625 	register struct tcpcb *tp;
626 	register struct tcpiphdr *ti;
627 {
628 	register struct tcpiphdr *q;
629 	struct socket *so = tp->t_inpcb->inp_socket;
630 	struct mbuf *m;
631 	int flags;
632 COUNT(TCP_REASS);
633 
634 	/*
635 	 * Call with ti==0 after become established to
636 	 * force pre-ESTABLISHED data up to user socket.
637 	 */
638 	if (ti == 0)
639 		goto present;
640 
641 	/*
642 	 * Find a segment which begins after this one does.
643 	 */
644 	for (q = tp->seg_next; q != (struct tcpiphdr *)tp;
645 	    q = (struct tcpiphdr *)q->ti_next)
646 		if (SEQ_GT(q->ti_seq, ti->ti_seq))
647 			break;
648 
649 	/*
650 	 * If there is a preceding segment, it may provide some of
651 	 * our data already.  If so, drop the data from the incoming
652 	 * segment.  If it provides all of our data, drop us.
653 	 */
654 	if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
655 		register int i;
656 		q = (struct tcpiphdr *)(q->ti_prev);
657 		/* conversion to int (in i) handles seq wraparound */
658 		i = q->ti_seq + q->ti_len - ti->ti_seq;
659 		if (i > 0) {
660 			if (i >= ti->ti_len)
661 				goto drop;
662 			m_adj(dtom(tp), i);
663 			ti->ti_len -= i;
664 			ti->ti_seq += i;
665 		}
666 		q = (struct tcpiphdr *)(q->ti_next);
667 	}
668 
669 	/*
670 	 * While we overlap succeeding segments trim them or,
671 	 * if they are completely covered, dequeue them.
672 	 */
673 	while (q != (struct tcpiphdr *)tp &&
674 	    SEQ_GT(ti->ti_seq + ti->ti_len, q->ti_seq)) {
675 		register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
676 		if (i < q->ti_len) {
677 			q->ti_len -= i;
678 			m_adj(dtom(q), i);
679 			break;
680 		}
681 		q = (struct tcpiphdr *)q->ti_next;
682 		m_freem(dtom(q->ti_prev));
683 		remque(q->ti_prev);
684 	}
685 
686 	/*
687 	 * Stick new segment in its place.
688 	 */
689 	insque(ti, q->ti_prev);
690 
691 present:
692 	/*
693 	 * Present data to user, advancing rcv_nxt through
694 	 * completed sequence space.
695 	 */
696 	if (TCPS_HAVERCVDSYN(tp->t_state) == 0)
697 		return (0);
698 	ti = tp->seg_next;
699 	if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
700 		return (0);
701 	if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
702 		return (0);
703 	do {
704 		tp->rcv_nxt += ti->ti_len;
705 		flags = ti->ti_flags & TH_FIN;
706 		remque(ti);
707 		m = dtom(ti);
708 		ti = (struct tcpiphdr *)ti->ti_next;
709 		if (so->so_state & SS_CANTRCVMORE)
710 			(void) m_freem(m);
711 		else
712 			sbappend(&so->so_rcv, m);
713 	} while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
714 	sorwakeup(so);
715 	return (flags);
716 drop:
717 	m_freem(dtom(ti));
718 	return (0);
719 }
720