2 * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
8 * 1. Redistributions of source code must retain the above copyright
9 * notice, this list of conditions and the following disclaimer.
10 * 2. Redistributions in binary form must reproduce the above copyright
11 * notice, this list of conditions and the following disclaimer in the
12 * documentation and/or other materials provided with the distribution.
13 * 3. All advertising materials mentioning features or use of this software
14 * must display the following acknowledgement:
15 * This product includes software developed by the University of
16 * California, Berkeley and its contributors.
17 * 4. Neither the name of the University nor the names of its contributors
18 * may be used to endorse or promote products derived from this software
19 * without specific prior written permission.
21 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
33 * @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
34 * tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
38 * Changes and additions relating to SLiRP
39 * Copyright (c) 1995 Danny Gasparovski.
41 * Please read the file COPYRIGHT for the
42 * terms and conditions of the copyright.
50 #define min(x,y) ((x) < (y) ? (x) : (y))
51 #define max(x,y) ((x) > (y) ? (x) : (y))
53 int tcprexmtthresh
= 3;
54 struct socket
*tcp_last_so
= &tcb
;
56 tcp_seq tcp_iss
; /* tcp initial send seq # */
58 #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
60 /* for modulo comparisons of timestamps */
61 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
62 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
65 * Insert segment ti into reassembly queue of tcp with
66 * control block tp. Return TH_FIN if reassembly now includes
67 * a segment with FIN. The macro form does the common case inline
68 * (segment is the next to be received on an established connection,
69 * and the queue is empty), avoiding linkage into and removal
70 * from the queue and repetition of various conversions.
71 * Set DELACK for segments received in order, but ack immediately
72 * when segments are out of order (so fast retransmit can work).
75 #define TCP_REASS(tp, ti, m, so, flags) {\
76 if ((ti)->ti_seq == (tp)->rcv_nxt && \
77 (tp)->seg_next == (tcpiphdrp_32)(tp) && \
78 (tp)->t_state == TCPS_ESTABLISHED) {\
79 if (ti->ti_flags & TH_PUSH) \
80 tp->t_flags |= TF_ACKNOW; \
82 tp->t_flags |= TF_DELACK; \
83 (tp)->rcv_nxt += (ti)->ti_len; \
84 flags = (ti)->ti_flags & TH_FIN; \
85 tcpstat.tcps_rcvpack++;\
86 tcpstat.tcps_rcvbyte += (ti)->ti_len;\
88 if (tcp_emu((so),(m))) sbappend((so), (m)); \
90 sbappend((so), (m)); \
91 /* sorwakeup(so); */ \
93 (flags) = tcp_reass((tp), (ti), (m)); \
94 tp->t_flags |= TF_ACKNOW; \
98 #define TCP_REASS(tp, ti, m, so, flags) { \
99 if ((ti)->ti_seq == (tp)->rcv_nxt && \
100 (tp)->seg_next == (tcpiphdrp_32)(tp) && \
101 (tp)->t_state == TCPS_ESTABLISHED) { \
102 tp->t_flags |= TF_DELACK; \
103 (tp)->rcv_nxt += (ti)->ti_len; \
104 flags = (ti)->ti_flags & TH_FIN; \
105 tcpstat.tcps_rcvpack++;\
106 tcpstat.tcps_rcvbyte += (ti)->ti_len;\
108 if (tcp_emu((so),(m))) sbappend(so, (m)); \
110 sbappend((so), (m)); \
111 /* sorwakeup(so); */ \
113 (flags) = tcp_reass((tp), (ti), (m)); \
114 tp->t_flags |= TF_ACKNOW; \
121 register struct tcpcb
*tp
;
122 register struct tcpiphdr
*ti
;
125 register struct tcpiphdr
*q
;
126 struct socket
*so
= tp
->t_socket
;
130 * Call with ti==0 after become established to
131 * force pre-ESTABLISHED data up to user socket.
137 * Find a segment which begins after this one does.
139 for (q
= (struct tcpiphdr
*)tp
->seg_next
; q
!= (struct tcpiphdr
*)tp
;
140 q
= (struct tcpiphdr
*)q
->ti_next
)
141 if (SEQ_GT(q
->ti_seq
, ti
->ti_seq
))
145 * If there is a preceding segment, it may provide some of
146 * our data already. If so, drop the data from the incoming
147 * segment. If it provides all of our data, drop us.
149 if ((struct tcpiphdr
*)q
->ti_prev
!= (struct tcpiphdr
*)tp
) {
151 q
= (struct tcpiphdr
*)q
->ti_prev
;
152 /* conversion to int (in i) handles seq wraparound */
153 i
= q
->ti_seq
+ q
->ti_len
- ti
->ti_seq
;
155 if (i
>= ti
->ti_len
) {
156 tcpstat
.tcps_rcvduppack
++;
157 tcpstat
.tcps_rcvdupbyte
+= ti
->ti_len
;
160 * Try to present any queued data
161 * at the left window edge to the user.
162 * This is needed after the 3-WHS
165 goto present
; /* ??? */
171 q
= (struct tcpiphdr
*)(q
->ti_next
);
173 tcpstat
.tcps_rcvoopack
++;
174 tcpstat
.tcps_rcvoobyte
+= ti
->ti_len
;
175 REASS_MBUF(ti
) = (mbufp_32
) m
; /* XXX */
178 * While we overlap succeeding segments trim them or,
179 * if they are completely covered, dequeue them.
181 while (q
!= (struct tcpiphdr
*)tp
) {
182 register int i
= (ti
->ti_seq
+ ti
->ti_len
) - q
->ti_seq
;
188 m_adj((struct mbuf
*) REASS_MBUF(q
), i
);
191 q
= (struct tcpiphdr
*)q
->ti_next
;
192 m
= (struct mbuf
*) REASS_MBUF((struct tcpiphdr
*)q
->ti_prev
);
193 remque_32((void *)(q
->ti_prev
));
198 * Stick new segment in its place.
200 insque_32(ti
, (void *)(q
->ti_prev
));
204 * Present data to user, advancing rcv_nxt through
205 * completed sequence space.
207 if (!TCPS_HAVEESTABLISHED(tp
->t_state
))
209 ti
= (struct tcpiphdr
*) tp
->seg_next
;
210 if (ti
== (struct tcpiphdr
*)tp
|| ti
->ti_seq
!= tp
->rcv_nxt
)
212 if (tp
->t_state
== TCPS_SYN_RECEIVED
&& ti
->ti_len
)
215 tp
->rcv_nxt
+= ti
->ti_len
;
216 flags
= ti
->ti_flags
& TH_FIN
;
218 m
= (struct mbuf
*) REASS_MBUF(ti
); /* XXX */
219 ti
= (struct tcpiphdr
*)ti
->ti_next
;
220 /* if (so->so_state & SS_FCANTRCVMORE) */
221 if (so
->so_state
& SS_FCANTSENDMORE
)
225 if (tcp_emu(so
,m
)) sbappend(so
, m
);
229 } while (ti
!= (struct tcpiphdr
*)tp
&& ti
->ti_seq
== tp
->rcv_nxt
);
235 * TCP input routine, follows pages 65-76 of the
236 * protocol specification dated September, 1981 very closely.
239 tcp_input(m
, iphlen
, inso
)
240 register struct mbuf
*m
;
244 struct ip save_ip
, *ip
;
245 register struct tcpiphdr
*ti
;
249 register struct tcpcb
*tp
= 0;
250 register int tiflags
;
251 struct socket
*so
= 0;
252 int todrop
, acked
, ourfinisacked
, needoutput
= 0;
253 /* int dropsocket = 0; */
257 /* int ts_present = 0; */
259 DEBUG_CALL("tcp_input");
260 DEBUG_ARGS((dfd
," m = %8lx iphlen = %2d inso = %lx\n",
261 (long )m
, iphlen
, (long )inso
));
264 * If called with m == 0, then we're continuing the connect
269 /* Re-set a few variables */
275 tiflags
= ti
->ti_flags
;
281 tcpstat
.tcps_rcvtotal
++;
283 * Get IP and TCP header together in first mbuf.
284 * Note: IP leaves IP header in first mbuf.
286 ti
= mtod(m
, struct tcpiphdr
*);
287 if (iphlen
> sizeof(struct ip
)) {
288 ip_stripoptions(m
, (struct mbuf
*)0);
289 iphlen
=sizeof(struct ip
);
291 /* XXX Check if too short */
295 * Save a copy of the IP header in case we want restore it
296 * for sending an ICMP error message in response.
298 ip
=mtod(m
, struct ip
*);
300 save_ip
.ip_len
+= iphlen
;
303 * Checksum extended TCP header and data.
305 tlen
= ((struct ip
*)ti
)->ip_len
;
306 ti
->ti_next
= ti
->ti_prev
= 0;
308 ti
->ti_len
= htons((u_int16_t
)tlen
);
309 len
= sizeof(struct ip
) + tlen
;
310 /* keep checksum for ICMP reply
311 * ti->ti_sum = cksum(m, len);
312 * if (ti->ti_sum) { */
314 tcpstat
.tcps_rcvbadsum
++;
319 * Check that TCP offset makes sense,
320 * pull out TCP options and adjust length. XXX
322 off
= ti
->ti_off
<< 2;
323 if (off
< sizeof (struct tcphdr
) || off
> tlen
) {
324 tcpstat
.tcps_rcvbadoff
++;
329 if (off
> sizeof (struct tcphdr
)) {
330 optlen
= off
- sizeof (struct tcphdr
);
331 optp
= mtod(m
, caddr_t
) + sizeof (struct tcpiphdr
);
334 * Do quick retrieval of timestamp options ("options
335 * prediction?"). If timestamp is the only option and it's
336 * formatted as recommended in RFC 1323 appendix A, we
337 * quickly get the values now and not bother calling
338 * tcp_dooptions(), etc.
340 /* if ((optlen == TCPOLEN_TSTAMP_APPA ||
341 * (optlen > TCPOLEN_TSTAMP_APPA &&
342 * optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
343 * *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
344 * (ti->ti_flags & TH_SYN) == 0) {
346 * ts_val = ntohl(*(u_int32_t *)(optp + 4));
347 * ts_ecr = ntohl(*(u_int32_t *)(optp + 8));
348 * optp = NULL; / * we've parsed the options * /
352 tiflags
= ti
->ti_flags
;
355 * Convert TCP protocol specific fields to host format.
363 * Drop TCP, IP headers and TCP options.
365 m
->m_data
+= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
366 m
->m_len
-= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
369 * Locate pcb for segment.
373 if (so
->so_fport
!= ti
->ti_dport
||
374 so
->so_lport
!= ti
->ti_sport
||
375 so
->so_laddr
.s_addr
!= ti
->ti_src
.s_addr
||
376 so
->so_faddr
.s_addr
!= ti
->ti_dst
.s_addr
) {
377 so
= solookup(&tcb
, ti
->ti_src
, ti
->ti_sport
,
378 ti
->ti_dst
, ti
->ti_dport
);
381 ++tcpstat
.tcps_socachemiss
;
385 * If the state is CLOSED (i.e., TCB does not exist) then
386 * all data in the incoming segment is discarded.
387 * If the TCB exists but is in CLOSED state, it is embryonic,
388 * but should either do a listen or a connect soon.
390 * state == CLOSED means we've done socreate() but haven't
391 * attached it to a protocol yet...
393 * XXX If a TCB does not exist, and the TH_SYN flag is
394 * the only flag set, then create a session, mark it
395 * as if it was LISTENING, and continue...
398 if ((tiflags
& (TH_SYN
|TH_FIN
|TH_RST
|TH_URG
|TH_ACK
)) != TH_SYN
)
401 if ((so
= socreate()) == NULL
)
403 if (tcp_attach(so
) < 0) {
404 free(so
); /* Not sofree (if it failed, it's not insqued) */
408 sbreserve(&so
->so_snd
, tcp_sndspace
);
409 sbreserve(&so
->so_rcv
, tcp_rcvspace
);
411 /* tcp_last_so = so; */ /* XXX ? */
412 /* tp = sototcpcb(so); */
414 so
->so_laddr
= ti
->ti_src
;
415 so
->so_lport
= ti
->ti_sport
;
416 so
->so_faddr
= ti
->ti_dst
;
417 so
->so_fport
= ti
->ti_dport
;
419 if ((so
->so_iptos
= tcp_tos(so
)) == 0)
420 so
->so_iptos
= ((struct ip
*)ti
)->ip_tos
;
423 tp
->t_state
= TCPS_LISTEN
;
427 * If this is a still-connecting socket, this probably
428 * a retransmit of the SYN. Whether it's a retransmit SYN
429 * or something else, we nuke it.
431 if (so
->so_state
& SS_ISFCONNECTING
)
435 /* XXX Should never fail */
438 if (tp
->t_state
== TCPS_CLOSED
)
441 /* Unscale the window into a 32-bit value. */
442 /* if ((tiflags & TH_SYN) == 0)
443 * tiwin = ti->ti_win << tp->snd_scale;
449 * Segment received on connection.
450 * Reset idle time and keep-alive timer.
454 tp
->t_timer
[TCPT_KEEP
] = tcp_keepintvl
;
456 tp
->t_timer
[TCPT_KEEP
] = tcp_keepidle
;
459 * Process options if not in LISTEN state,
460 * else do it below (after getting remote address).
462 if (optp
&& tp
->t_state
!= TCPS_LISTEN
)
463 tcp_dooptions(tp
, (u_char
*)optp
, optlen
, ti
);
465 /* &ts_present, &ts_val, &ts_ecr); */
468 * Header prediction: check for the two common cases
469 * of a uni-directional data xfer. If the packet has
470 * no control flags, is in-sequence, the window didn't
471 * change and we're not retransmitting, it's a
472 * candidate. If the length is zero and the ack moved
473 * forward, we're the sender side of the xfer. Just
474 * free the data acked & wake any higher level process
475 * that was blocked waiting for space. If the length
476 * is non-zero and the ack didn't move, we're the
477 * receiver side. If we're getting packets in-order
478 * (the reassembly queue is empty), add the data to
479 * the socket buffer and note that we need a delayed ack.
481 * XXX Some of these tests are not needed
482 * eg: the tiwin == tp->snd_wnd prevents many more
483 * predictions.. with no *real* advantage..
485 if (tp
->t_state
== TCPS_ESTABLISHED
&&
486 (tiflags
& (TH_SYN
|TH_FIN
|TH_RST
|TH_URG
|TH_ACK
)) == TH_ACK
&&
487 /* (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) && */
488 ti
->ti_seq
== tp
->rcv_nxt
&&
489 tiwin
&& tiwin
== tp
->snd_wnd
&&
490 tp
->snd_nxt
== tp
->snd_max
) {
492 * If last ACK falls within this segment's sequence numbers,
493 * record the timestamp.
495 /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
496 * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
497 * tp->ts_recent_age = tcp_now;
498 * tp->ts_recent = ts_val;
501 if (ti
->ti_len
== 0) {
502 if (SEQ_GT(ti
->ti_ack
, tp
->snd_una
) &&
503 SEQ_LEQ(ti
->ti_ack
, tp
->snd_max
) &&
504 tp
->snd_cwnd
>= tp
->snd_wnd
) {
506 * this is a pure ack for outstanding data.
508 ++tcpstat
.tcps_predack
;
510 * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
513 SEQ_GT(ti
->ti_ack
, tp
->t_rtseq
))
514 tcp_xmit_timer(tp
, tp
->t_rtt
);
515 acked
= ti
->ti_ack
- tp
->snd_una
;
516 tcpstat
.tcps_rcvackpack
++;
517 tcpstat
.tcps_rcvackbyte
+= acked
;
518 sbdrop(&so
->so_snd
, acked
);
519 tp
->snd_una
= ti
->ti_ack
;
523 * If all outstanding data are acked, stop
524 * retransmit timer, otherwise restart timer
525 * using current (possibly backed-off) value.
526 * If process is waiting for space,
527 * wakeup/selwakeup/signal. If data
528 * are ready to send, let tcp_output
529 * decide between more output or persist.
531 if (tp
->snd_una
== tp
->snd_max
)
532 tp
->t_timer
[TCPT_REXMT
] = 0;
533 else if (tp
->t_timer
[TCPT_PERSIST
] == 0)
534 tp
->t_timer
[TCPT_REXMT
] = tp
->t_rxtcur
;
537 * There's room in so_snd, sowwakup will read()
538 * from the socket if we can
540 /* if (so->so_snd.sb_flags & SB_NOTIFY)
544 * This is called because sowwakeup might have
545 * put data into so_snd. Since we don't so sowwakeup,
546 * we don't need this.. XXX???
548 if (so
->so_snd
.sb_cc
)
549 (void) tcp_output(tp
);
553 } else if (ti
->ti_ack
== tp
->snd_una
&&
554 tp
->seg_next
== (tcpiphdrp_32
)tp
&&
555 ti
->ti_len
<= sbspace(&so
->so_rcv
)) {
557 * this is a pure, in-sequence data packet
558 * with nothing on the reassembly queue and
559 * we have enough buffer space to take it.
561 ++tcpstat
.tcps_preddat
;
562 tp
->rcv_nxt
+= ti
->ti_len
;
563 tcpstat
.tcps_rcvpack
++;
564 tcpstat
.tcps_rcvbyte
+= ti
->ti_len
;
566 * Add data to socket buffer.
569 if (tcp_emu(so
,m
)) sbappend(so
, m
);
574 * XXX This is called when data arrives. Later, check
575 * if we can actually write() to the socket
576 * XXX Need to check? It's be NON_BLOCKING
581 * If this is a short packet, then ACK now - with Nagel
582 * congestion avoidance sender won't send more until
585 * Here are 3 interpretations of what should happen.
586 * The best (for me) is to delay-ack everything except
587 * if it's a one-byte packet containing an ESC
588 * (this means it's an arrow key (or similar) sent using
589 * Nagel, hence there will be no echo)
590 * The first of these is the original, the second is the
591 * middle ground between the other 2
593 /* if (((unsigned)ti->ti_len < tp->t_maxseg)) {
595 /* if (((unsigned)ti->ti_len < tp->t_maxseg &&
596 * (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
597 * ((so->so_iptos & IPTOS_LOWDELAY) &&
598 * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
600 if ((unsigned)ti
->ti_len
== 1 &&
601 ((struct tcpiphdr_2
*)ti
)->first_char
== (char)27) {
602 tp
->t_flags
|= TF_ACKNOW
;
605 tp
->t_flags
|= TF_DELACK
;
609 } /* header prediction */
611 * Calculate amount of space in receive window,
612 * and then do TCP input processing.
613 * Receive window is amount of space in rcv queue,
614 * but not less than advertised window.
617 win
= sbspace(&so
->so_rcv
);
620 tp
->rcv_wnd
= max(win
, (int)(tp
->rcv_adv
- tp
->rcv_nxt
));
623 switch (tp
->t_state
) {
626 * If the state is LISTEN then ignore segment if it contains an RST.
627 * If the segment contains an ACK then it is bad and send a RST.
628 * If it does not contain a SYN then it is not interesting; drop it.
629 * Don't bother responding if the destination was a broadcast.
630 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
631 * tp->iss, and send a segment:
632 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
633 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
634 * Fill in remote peer address fields if not previously specified.
635 * Enter SYN_RECEIVED state, and process any other fields of this
636 * segment in this state.
640 if (tiflags
& TH_RST
)
642 if (tiflags
& TH_ACK
)
644 if ((tiflags
& TH_SYN
) == 0)
648 * This has way too many gotos...
649 * But a bit of spaghetti code never hurt anybody :)
653 * If this is destined for the control address, then flag to
654 * tcp_ctl once connected, otherwise connect
656 if ((so
->so_faddr
.s_addr
&htonl(0xffffff00)) == special_addr
.s_addr
) {
657 int lastbyte
=ntohl(so
->so_faddr
.s_addr
) & 0xff;
658 if (lastbyte
!=CTL_ALIAS
&& lastbyte
!=CTL_DNS
) {
660 if(lastbyte
==CTL_CMD
|| lastbyte
==CTL_EXEC
) {
661 /* Command or exec adress */
662 so
->so_state
|= SS_CTL
;
664 /* May be an add exec */
665 struct ex_list
*ex_ptr
;
667 for(ex_ptr
= exec_list
; ex_ptr
; ex_ptr
= ex_ptr
->ex_next
) {
668 if(ex_ptr
->ex_fport
== so
->so_fport
&&
669 lastbyte
== ex_ptr
->ex_addr
) {
670 so
->so_state
|= SS_CTL
;
675 if(so
->so_state
& SS_CTL
) goto cont_input
;
678 /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
681 if (so
->so_emu
& EMU_NOCONNECT
) {
682 so
->so_emu
&= ~EMU_NOCONNECT
;
686 if(tcp_fconnect(so
) == -1 && errno
!= EINPROGRESS
) {
687 u_char code
=ICMP_UNREACH_NET
;
688 DEBUG_MISC((dfd
," tcp fconnect errno = %d-%s\n",
689 errno
,strerror(errno
)));
690 if(errno
== ECONNREFUSED
) {
691 /* ACK the SYN, send RST to refuse the connection */
692 tcp_respond(tp
, ti
, m
, ti
->ti_seq
+1, (tcp_seq
)0,
695 if(errno
== EHOSTUNREACH
) code
=ICMP_UNREACH_HOST
;
696 HTONL(ti
->ti_seq
); /* restore tcp header */
700 m
->m_data
-= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
701 m
->m_len
+= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
703 icmp_error(m
, ICMP_UNREACH
,code
, 0,strerror(errno
));
709 * Haven't connected yet, save the current mbuf
711 * XXX Some OS's don't tell us whether the connect()
712 * succeeded or not. So we must time it out.
716 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEP_INIT
;
717 tp
->t_state
= TCPS_SYN_RECEIVED
;
723 * Check if the connect succeeded
725 if (so
->so_state
& SS_NOFDREF
) {
733 tcp_dooptions(tp
, (u_char
*)optp
, optlen
, ti
);
735 /* &ts_present, &ts_val, &ts_ecr); */
741 tcp_iss
+= TCP_ISSINCR
/2;
742 tp
->irs
= ti
->ti_seq
;
745 tp
->t_flags
|= TF_ACKNOW
;
746 tp
->t_state
= TCPS_SYN_RECEIVED
;
747 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEP_INIT
;
748 tcpstat
.tcps_accepts
++;
750 } /* case TCPS_LISTEN */
753 * If the state is SYN_SENT:
754 * if seg contains an ACK, but not for our SYN, drop the input.
755 * if seg contains a RST, then drop the connection.
756 * if seg does not contain SYN, then drop it.
757 * Otherwise this is an acceptable SYN segment
758 * initialize tp->rcv_nxt and tp->irs
759 * if seg contains ack then advance tp->snd_una
760 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state
761 * arrange for segment to be acked (eventually)
762 * continue processing rest of data/controls, beginning with URG
765 if ((tiflags
& TH_ACK
) &&
766 (SEQ_LEQ(ti
->ti_ack
, tp
->iss
) ||
767 SEQ_GT(ti
->ti_ack
, tp
->snd_max
)))
770 if (tiflags
& TH_RST
) {
771 if (tiflags
& TH_ACK
)
772 tp
= tcp_drop(tp
,0); /* XXX Check t_softerror! */
776 if ((tiflags
& TH_SYN
) == 0)
778 if (tiflags
& TH_ACK
) {
779 tp
->snd_una
= ti
->ti_ack
;
780 if (SEQ_LT(tp
->snd_nxt
, tp
->snd_una
))
781 tp
->snd_nxt
= tp
->snd_una
;
784 tp
->t_timer
[TCPT_REXMT
] = 0;
785 tp
->irs
= ti
->ti_seq
;
787 tp
->t_flags
|= TF_ACKNOW
;
788 if (tiflags
& TH_ACK
&& SEQ_GT(tp
->snd_una
, tp
->iss
)) {
789 tcpstat
.tcps_connects
++;
791 tp
->t_state
= TCPS_ESTABLISHED
;
793 /* Do window scaling on this connection? */
794 /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
795 * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
796 * tp->snd_scale = tp->requested_s_scale;
797 * tp->rcv_scale = tp->request_r_scale;
800 (void) tcp_reass(tp
, (struct tcpiphdr
*)0,
803 * if we didn't have to retransmit the SYN,
804 * use its rtt as our initial srtt & rtt var.
807 tcp_xmit_timer(tp
, tp
->t_rtt
);
809 tp
->t_state
= TCPS_SYN_RECEIVED
;
813 * Advance ti->ti_seq to correspond to first data byte.
814 * If data, trim to stay within window,
815 * dropping FIN if necessary.
818 if (ti
->ti_len
> tp
->rcv_wnd
) {
819 todrop
= ti
->ti_len
- tp
->rcv_wnd
;
821 ti
->ti_len
= tp
->rcv_wnd
;
823 tcpstat
.tcps_rcvpackafterwin
++;
824 tcpstat
.tcps_rcvbyteafterwin
+= todrop
;
826 tp
->snd_wl1
= ti
->ti_seq
- 1;
827 tp
->rcv_up
= ti
->ti_seq
;
829 } /* switch tp->t_state */
831 * States other than LISTEN or SYN_SENT.
832 * First check timestamp, if present.
833 * Then check that at least some bytes of segment are within
834 * receive window. If segment begins before rcv_nxt,
835 * drop leading data (and SYN); if nothing left, just ack.
837 * RFC 1323 PAWS: If we have a timestamp reply on this segment
838 * and it's less than ts_recent, drop it.
840 /* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
841 * TSTMP_LT(ts_val, tp->ts_recent)) {
843 */ /* Check to see if ts_recent is over 24 days old. */
844 /* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
846 * * Invalidate ts_recent. If this segment updates
847 * * ts_recent, the age will be reset later and ts_recent
848 * * will get a valid value. If it does not, setting
849 * * ts_recent to zero will at least satisfy the
850 * * requirement that zero be placed in the timestamp
851 * * echo reply when ts_recent isn't valid. The
852 * * age isn't reset until we get a valid ts_recent
853 * * because we don't want out-of-order segments to be
854 * * dropped when ts_recent is old.
856 /* tp->ts_recent = 0;
858 * tcpstat.tcps_rcvduppack++;
859 * tcpstat.tcps_rcvdupbyte += ti->ti_len;
860 * tcpstat.tcps_pawsdrop++;
866 todrop
= tp
->rcv_nxt
- ti
->ti_seq
;
868 if (tiflags
& TH_SYN
) {
878 * Following if statement from Stevens, vol. 2, p. 960.
880 if (todrop
> ti
->ti_len
881 || (todrop
== ti
->ti_len
&& (tiflags
& TH_FIN
) == 0)) {
883 * Any valid FIN must be to the left of the window.
884 * At this point the FIN must be a duplicate or out
885 * of sequence; drop it.
890 * Send an ACK to resynchronize and drop any data.
891 * But keep on processing for RST or ACK.
893 tp
->t_flags
|= TF_ACKNOW
;
895 tcpstat
.tcps_rcvduppack
++;
896 tcpstat
.tcps_rcvdupbyte
+= todrop
;
898 tcpstat
.tcps_rcvpartduppack
++;
899 tcpstat
.tcps_rcvpartdupbyte
+= todrop
;
902 ti
->ti_seq
+= todrop
;
903 ti
->ti_len
-= todrop
;
904 if (ti
->ti_urp
> todrop
)
905 ti
->ti_urp
-= todrop
;
912 * If new data are received on a connection after the
913 * user processes are gone, then RST the other end.
915 if ((so
->so_state
& SS_NOFDREF
) &&
916 tp
->t_state
> TCPS_CLOSE_WAIT
&& ti
->ti_len
) {
918 tcpstat
.tcps_rcvafterclose
++;
923 * If segment ends after window, drop trailing data
924 * (and PUSH and FIN); if nothing left, just ACK.
926 todrop
= (ti
->ti_seq
+ti
->ti_len
) - (tp
->rcv_nxt
+tp
->rcv_wnd
);
928 tcpstat
.tcps_rcvpackafterwin
++;
929 if (todrop
>= ti
->ti_len
) {
930 tcpstat
.tcps_rcvbyteafterwin
+= ti
->ti_len
;
932 * If a new connection request is received
933 * while in TIME_WAIT, drop the old connection
934 * and start over if the sequence numbers
935 * are above the previous ones.
937 if (tiflags
& TH_SYN
&&
938 tp
->t_state
== TCPS_TIME_WAIT
&&
939 SEQ_GT(ti
->ti_seq
, tp
->rcv_nxt
)) {
940 iss
= tp
->rcv_nxt
+ TCP_ISSINCR
;
945 * If window is closed can only take segments at
946 * window edge, and have to drop data and PUSH from
947 * incoming segments. Continue processing, but
948 * remember to ack. Otherwise, drop segment
951 if (tp
->rcv_wnd
== 0 && ti
->ti_seq
== tp
->rcv_nxt
) {
952 tp
->t_flags
|= TF_ACKNOW
;
953 tcpstat
.tcps_rcvwinprobe
++;
957 tcpstat
.tcps_rcvbyteafterwin
+= todrop
;
959 ti
->ti_len
-= todrop
;
960 tiflags
&= ~(TH_PUSH
|TH_FIN
);
964 * If last ACK falls within this segment's sequence numbers,
965 * record its timestamp.
967 /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
968 * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
969 * ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
970 * tp->ts_recent_age = tcp_now;
971 * tp->ts_recent = ts_val;
976 * If the RST bit is set examine the state:
977 * SYN_RECEIVED STATE:
978 * If passive open, return to LISTEN state.
979 * If active open, inform user that connection was refused.
980 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
981 * Inform user that connection was reset, and close tcb.
982 * CLOSING, LAST_ACK, TIME_WAIT STATES
985 if (tiflags
&TH_RST
) switch (tp
->t_state
) {
987 case TCPS_SYN_RECEIVED
:
988 /* so->so_error = ECONNREFUSED; */
991 case TCPS_ESTABLISHED
:
992 case TCPS_FIN_WAIT_1
:
993 case TCPS_FIN_WAIT_2
:
994 case TCPS_CLOSE_WAIT
:
995 /* so->so_error = ECONNRESET; */
997 tp
->t_state
= TCPS_CLOSED
;
998 tcpstat
.tcps_drops
++;
1004 case TCPS_TIME_WAIT
:
1010 * If a SYN is in the window, then this is an
1011 * error and we send an RST and drop the connection.
1013 if (tiflags
& TH_SYN
) {
1014 tp
= tcp_drop(tp
,0);
1019 * If the ACK bit is off we drop the segment and return.
1021 if ((tiflags
& TH_ACK
) == 0) goto drop
;
1026 switch (tp
->t_state
) {
1028 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
1029 * ESTABLISHED state and continue processing, otherwise
1030 * send an RST. una<=ack<=max
1032 case TCPS_SYN_RECEIVED
:
1034 if (SEQ_GT(tp
->snd_una
, ti
->ti_ack
) ||
1035 SEQ_GT(ti
->ti_ack
, tp
->snd_max
))
1037 tcpstat
.tcps_connects
++;
1038 tp
->t_state
= TCPS_ESTABLISHED
;
1040 * The sent SYN is ack'ed with our sequence number +1
1041 * The first data byte already in the buffer will get
1042 * lost if no correction is made. This is only needed for
1043 * SS_CTL since the buffer is empty otherwise.
1044 * tp->snd_una++; or:
1046 tp
->snd_una
=ti
->ti_ack
;
1047 if (so
->so_state
& SS_CTL
) {
1048 /* So tcp_ctl reports the right state */
1052 so
->so_state
&= ~SS_CTL
; /* success XXX */
1053 } else if (ret
== 2) {
1054 so
->so_state
= SS_NOFDREF
; /* CTL_CMD */
1057 tp
->t_state
= TCPS_FIN_WAIT_1
;
1063 /* Do window scaling? */
1064 /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
1065 * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
1066 * tp->snd_scale = tp->requested_s_scale;
1067 * tp->rcv_scale = tp->request_r_scale;
1070 (void) tcp_reass(tp
, (struct tcpiphdr
*)0, (struct mbuf
*)0);
1071 tp
->snd_wl1
= ti
->ti_seq
- 1;
1072 /* Avoid ack processing; snd_una==ti_ack => dup ack */
1077 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1078 * ACKs. If the ack is in the range
1079 * tp->snd_una < ti->ti_ack <= tp->snd_max
1080 * then advance tp->snd_una to ti->ti_ack and drop
1081 * data from the retransmission queue. If this ACK reflects
1082 * more up to date window information we update our window information.
1084 case TCPS_ESTABLISHED
:
1085 case TCPS_FIN_WAIT_1
:
1086 case TCPS_FIN_WAIT_2
:
1087 case TCPS_CLOSE_WAIT
:
1090 case TCPS_TIME_WAIT
:
1092 if (SEQ_LEQ(ti
->ti_ack
, tp
->snd_una
)) {
1093 if (ti
->ti_len
== 0 && tiwin
== tp
->snd_wnd
) {
1094 tcpstat
.tcps_rcvdupack
++;
1095 DEBUG_MISC((dfd
," dup ack m = %lx so = %lx \n",
1096 (long )m
, (long )so
));
1098 * If we have outstanding data (other than
1099 * a window probe), this is a completely
1100 * duplicate ack (ie, window info didn't
1101 * change), the ack is the biggest we've
1102 * seen and we've seen exactly our rexmt
1103 * threshold of them, assume a packet
1104 * has been dropped and retransmit it.
1105 * Kludge snd_nxt & the congestion
1106 * window so we send only this one
1109 * We know we're losing at the current
1110 * window size so do congestion avoidance
1111 * (set ssthresh to half the current window
1112 * and pull our congestion window back to
1113 * the new ssthresh).
1115 * Dup acks mean that packets have left the
1116 * network (they're now cached at the receiver)
1117 * so bump cwnd by the amount in the receiver
1118 * to keep a constant cwnd packets in the
1121 if (tp
->t_timer
[TCPT_REXMT
] == 0 ||
1122 ti
->ti_ack
!= tp
->snd_una
)
1124 else if (++tp
->t_dupacks
== tcprexmtthresh
) {
1125 tcp_seq onxt
= tp
->snd_nxt
;
1127 min(tp
->snd_wnd
, tp
->snd_cwnd
) / 2 /
1132 tp
->snd_ssthresh
= win
* tp
->t_maxseg
;
1133 tp
->t_timer
[TCPT_REXMT
] = 0;
1135 tp
->snd_nxt
= ti
->ti_ack
;
1136 tp
->snd_cwnd
= tp
->t_maxseg
;
1137 (void) tcp_output(tp
);
1138 tp
->snd_cwnd
= tp
->snd_ssthresh
+
1139 tp
->t_maxseg
* tp
->t_dupacks
;
1140 if (SEQ_GT(onxt
, tp
->snd_nxt
))
1143 } else if (tp
->t_dupacks
> tcprexmtthresh
) {
1144 tp
->snd_cwnd
+= tp
->t_maxseg
;
1145 (void) tcp_output(tp
);
1154 * If the congestion window was inflated to account
1155 * for the other side's cached packets, retract it.
1157 if (tp
->t_dupacks
> tcprexmtthresh
&&
1158 tp
->snd_cwnd
> tp
->snd_ssthresh
)
1159 tp
->snd_cwnd
= tp
->snd_ssthresh
;
1161 if (SEQ_GT(ti
->ti_ack
, tp
->snd_max
)) {
1162 tcpstat
.tcps_rcvacktoomuch
++;
1165 acked
= ti
->ti_ack
- tp
->snd_una
;
1166 tcpstat
.tcps_rcvackpack
++;
1167 tcpstat
.tcps_rcvackbyte
+= acked
;
1170 * If we have a timestamp reply, update smoothed
1171 * round trip time. If no timestamp is present but
1172 * transmit timer is running and timed sequence
1173 * number was acked, update smoothed round trip time.
1174 * Since we now have an rtt measurement, cancel the
1175 * timer backoff (cf., Phil Karn's retransmit alg.).
1176 * Recompute the initial retransmit timer.
1179 * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1182 if (tp
->t_rtt
&& SEQ_GT(ti
->ti_ack
, tp
->t_rtseq
))
1183 tcp_xmit_timer(tp
,tp
->t_rtt
);
1186 * If all outstanding data is acked, stop retransmit
1187 * timer and remember to restart (more output or persist).
1188 * If there is more data to be acked, restart retransmit
1189 * timer, using current (possibly backed-off) value.
1191 if (ti
->ti_ack
== tp
->snd_max
) {
1192 tp
->t_timer
[TCPT_REXMT
] = 0;
1194 } else if (tp
->t_timer
[TCPT_PERSIST
] == 0)
1195 tp
->t_timer
[TCPT_REXMT
] = tp
->t_rxtcur
;
1197 * When new data is acked, open the congestion window.
1198 * If the window gives us less than ssthresh packets
1199 * in flight, open exponentially (maxseg per packet).
1200 * Otherwise open linearly: maxseg per window
1201 * (maxseg^2 / cwnd per packet).
1204 register u_int cw
= tp
->snd_cwnd
;
1205 register u_int incr
= tp
->t_maxseg
;
1207 if (cw
> tp
->snd_ssthresh
)
1208 incr
= incr
* incr
/ cw
;
1209 tp
->snd_cwnd
= min(cw
+ incr
, TCP_MAXWIN
<<tp
->snd_scale
);
1211 if (acked
> so
->so_snd
.sb_cc
) {
1212 tp
->snd_wnd
-= so
->so_snd
.sb_cc
;
1213 sbdrop(&so
->so_snd
, (int )so
->so_snd
.sb_cc
);
1216 sbdrop(&so
->so_snd
, acked
);
1217 tp
->snd_wnd
-= acked
;
1221 * XXX sowwakup is called when data is acked and there's room for
1222 * for more data... it should read() the socket
1224 /* if (so->so_snd.sb_flags & SB_NOTIFY)
1227 tp
->snd_una
= ti
->ti_ack
;
1228 if (SEQ_LT(tp
->snd_nxt
, tp
->snd_una
))
1229 tp
->snd_nxt
= tp
->snd_una
;
1231 switch (tp
->t_state
) {
1234 * In FIN_WAIT_1 STATE in addition to the processing
1235 * for the ESTABLISHED state if our FIN is now acknowledged
1236 * then enter FIN_WAIT_2.
1238 case TCPS_FIN_WAIT_1
:
1239 if (ourfinisacked
) {
1241 * If we can't receive any more
1242 * data, then closing user can proceed.
1243 * Starting the timer is contrary to the
1244 * specification, but if we don't get a FIN
1245 * we'll hang forever.
1247 if (so
->so_state
& SS_FCANTRCVMORE
) {
1248 soisfdisconnected(so
);
1249 tp
->t_timer
[TCPT_2MSL
] = tcp_maxidle
;
1251 tp
->t_state
= TCPS_FIN_WAIT_2
;
1256 * In CLOSING STATE in addition to the processing for
1257 * the ESTABLISHED state if the ACK acknowledges our FIN
1258 * then enter the TIME-WAIT state, otherwise ignore
1262 if (ourfinisacked
) {
1263 tp
->t_state
= TCPS_TIME_WAIT
;
1264 tcp_canceltimers(tp
);
1265 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1266 soisfdisconnected(so
);
1271 * In LAST_ACK, we may still be waiting for data to drain
1272 * and/or to be acked, as well as for the ack of our FIN.
1273 * If our FIN is now acknowledged, delete the TCB,
1274 * enter the closed state and return.
1277 if (ourfinisacked
) {
1284 * In TIME_WAIT state the only thing that should arrive
1285 * is a retransmission of the remote FIN. Acknowledge
1286 * it and restart the finack timer.
1288 case TCPS_TIME_WAIT
:
1289 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1292 } /* switch(tp->t_state) */
1296 * Update window information.
1297 * Don't look at window if no ACK: TAC's send garbage on first SYN.
1299 if ((tiflags
& TH_ACK
) &&
1300 (SEQ_LT(tp
->snd_wl1
, ti
->ti_seq
) ||
1301 (tp
->snd_wl1
== ti
->ti_seq
&& (SEQ_LT(tp
->snd_wl2
, ti
->ti_ack
) ||
1302 (tp
->snd_wl2
== ti
->ti_ack
&& tiwin
> tp
->snd_wnd
))))) {
1303 /* keep track of pure window updates */
1304 if (ti
->ti_len
== 0 &&
1305 tp
->snd_wl2
== ti
->ti_ack
&& tiwin
> tp
->snd_wnd
)
1306 tcpstat
.tcps_rcvwinupd
++;
1307 tp
->snd_wnd
= tiwin
;
1308 tp
->snd_wl1
= ti
->ti_seq
;
1309 tp
->snd_wl2
= ti
->ti_ack
;
1310 if (tp
->snd_wnd
> tp
->max_sndwnd
)
1311 tp
->max_sndwnd
= tp
->snd_wnd
;
1316 * Process segments with URG.
1318 if ((tiflags
& TH_URG
) && ti
->ti_urp
&&
1319 TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1321 * This is a kludge, but if we receive and accept
1322 * random urgent pointers, we'll crash in
1323 * soreceive. It's hard to imagine someone
1324 * actually wanting to send this much urgent data.
1326 if (ti
->ti_urp
+ so
->so_rcv
.sb_cc
> so
->so_rcv
.sb_datalen
) {
1332 * If this segment advances the known urgent pointer,
1333 * then mark the data stream. This should not happen
1334 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1335 * a FIN has been received from the remote side.
1336 * In these states we ignore the URG.
1338 * According to RFC961 (Assigned Protocols),
1339 * the urgent pointer points to the last octet
1340 * of urgent data. We continue, however,
1341 * to consider it to indicate the first octet
1342 * of data past the urgent section as the original
1343 * spec states (in one of two places).
1345 if (SEQ_GT(ti
->ti_seq
+ti
->ti_urp
, tp
->rcv_up
)) {
1346 tp
->rcv_up
= ti
->ti_seq
+ ti
->ti_urp
;
1347 so
->so_urgc
= so
->so_rcv
.sb_cc
+
1348 (tp
->rcv_up
- tp
->rcv_nxt
); /* -1; */
1349 tp
->rcv_up
= ti
->ti_seq
+ ti
->ti_urp
;
1354 * If no out of band data is expected,
1355 * pull receive urgent pointer along
1356 * with the receive window.
1358 if (SEQ_GT(tp
->rcv_nxt
, tp
->rcv_up
))
1359 tp
->rcv_up
= tp
->rcv_nxt
;
1363 * Process the segment text, merging it into the TCP sequencing queue,
1364 * and arranging for acknowledgment of receipt if necessary.
1365 * This process logically involves adjusting tp->rcv_wnd as data
1366 * is presented to the user (this happens in tcp_usrreq.c,
1367 * case PRU_RCVD). If a FIN has already been received on this
1368 * connection then we just ignore the text.
1370 if ((ti
->ti_len
|| (tiflags
&TH_FIN
)) &&
1371 TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1372 TCP_REASS(tp
, ti
, m
, so
, tiflags
);
1374 * Note the amount of data that peer has sent into
1375 * our window, in order to estimate the sender's
1378 len
= so
->so_rcv
.sb_datalen
- (tp
->rcv_adv
- tp
->rcv_nxt
);
1385 * If FIN is received ACK the FIN and let the user know
1386 * that the connection is closing.
1388 if (tiflags
& TH_FIN
) {
1389 if (TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1391 * If we receive a FIN we can't send more data,
1393 * Shutdown the socket if there is no rx data in the
1395 * soread() is called on completion of shutdown() and
1396 * will got to TCPS_LAST_ACK, and use tcp_output()
1399 /* sofcantrcvmore(so); */
1402 tp
->t_flags
|= TF_ACKNOW
;
1405 switch (tp
->t_state
) {
1408 * In SYN_RECEIVED and ESTABLISHED STATES
1409 * enter the CLOSE_WAIT state.
1411 case TCPS_SYN_RECEIVED
:
1412 case TCPS_ESTABLISHED
:
1413 if(so
->so_emu
== EMU_CTL
) /* no shutdown on socket */
1414 tp
->t_state
= TCPS_LAST_ACK
;
1416 tp
->t_state
= TCPS_CLOSE_WAIT
;
1420 * If still in FIN_WAIT_1 STATE FIN has not been acked so
1421 * enter the CLOSING state.
1423 case TCPS_FIN_WAIT_1
:
1424 tp
->t_state
= TCPS_CLOSING
;
1428 * In FIN_WAIT_2 state enter the TIME_WAIT state,
1429 * starting the time-wait timer, turning off the other
1432 case TCPS_FIN_WAIT_2
:
1433 tp
->t_state
= TCPS_TIME_WAIT
;
1434 tcp_canceltimers(tp
);
1435 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1436 soisfdisconnected(so
);
1440 * In TIME_WAIT state restart the 2 MSL time_wait timer.
1442 case TCPS_TIME_WAIT
:
1443 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1449 * If this is a small packet, then ACK now - with Nagel
1450 * congestion avoidance sender won't send more until
1455 /* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1457 /* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1458 * (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
1459 * ((so->so_iptos & IPTOS_LOWDELAY) &&
1460 * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1462 if (ti
->ti_len
&& (unsigned)ti
->ti_len
<= 5 &&
1463 ((struct tcpiphdr_2
*)ti
)->first_char
== (char)27) {
1464 tp
->t_flags
|= TF_ACKNOW
;
1468 * Return any desired output.
1470 if (needoutput
|| (tp
->t_flags
& TF_ACKNOW
)) {
1471 (void) tcp_output(tp
);
1477 * Generate an ACK dropping incoming segment if it occupies
1478 * sequence space, where the ACK reflects our state.
1480 if (tiflags
& TH_RST
)
1483 tp
->t_flags
|= TF_ACKNOW
;
1484 (void) tcp_output(tp
);
1488 /* reuses m if m!=NULL, m_free() unnecessary */
1489 if (tiflags
& TH_ACK
)
1490 tcp_respond(tp
, ti
, m
, (tcp_seq
)0, ti
->ti_ack
, TH_RST
);
1492 if (tiflags
& TH_SYN
) ti
->ti_len
++;
1493 tcp_respond(tp
, ti
, m
, ti
->ti_seq
+ti
->ti_len
, (tcp_seq
)0,
1501 * Drop space held by incoming segment and return.
1508 /* , ts_present, ts_val, ts_ecr) */
1510 * u_int32_t *ts_val, *ts_ecr;
1513 tcp_dooptions(tp
, cp
, cnt
, ti
)
1517 struct tcpiphdr
*ti
;
1522 DEBUG_CALL("tcp_dooptions");
1523 DEBUG_ARGS((dfd
," tp = %lx cnt=%i \n", (long )tp
, cnt
));
1525 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1527 if (opt
== TCPOPT_EOL
)
1529 if (opt
== TCPOPT_NOP
)
1542 if (optlen
!= TCPOLEN_MAXSEG
)
1544 if (!(ti
->ti_flags
& TH_SYN
))
1546 memcpy((char *) &mss
, (char *) cp
+ 2, sizeof(mss
));
1548 (void) tcp_mss(tp
, mss
); /* sets t_maxseg */
1551 /* case TCPOPT_WINDOW:
1552 * if (optlen != TCPOLEN_WINDOW)
1554 * if (!(ti->ti_flags & TH_SYN))
1556 * tp->t_flags |= TF_RCVD_SCALE;
1557 * tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1560 /* case TCPOPT_TIMESTAMP:
1561 * if (optlen != TCPOLEN_TIMESTAMP)
1564 * memcpy((char *) ts_val, (char *)cp + 2, sizeof(*ts_val));
1566 * memcpy((char *) ts_ecr, (char *)cp + 6, sizeof(*ts_ecr));
1570 * * A timestamp received in a SYN makes
1571 * * it ok to send timestamp requests and replies.
1573 /* if (ti->ti_flags & TH_SYN) {
1574 * tp->t_flags |= TF_RCVD_TSTMP;
1575 * tp->ts_recent = *ts_val;
1576 * tp->ts_recent_age = tcp_now;
1585 * Pull out of band byte out of a segment so
1586 * it doesn't appear in the user's data queue.
1587 * It is still reflected in the segment length for
1588 * sequencing purposes.
1594 tcp_pulloutofband(so
, ti
, m
)
1596 struct tcpiphdr
*ti
;
1597 register struct mbuf
*m
;
1599 int cnt
= ti
->ti_urp
- 1;
1602 if (m
->m_len
> cnt
) {
1603 char *cp
= mtod(m
, caddr_t
) + cnt
;
1604 struct tcpcb
*tp
= sototcpcb(so
);
1607 tp
->t_oobflags
|= TCPOOB_HAVEDATA
;
1608 memcpy(sp
, cp
+1, (unsigned)(m
->m_len
- cnt
- 1));
1613 m
= m
->m_next
; /* XXX WRONG! Fix it! */
1617 panic("tcp_pulloutofband");
1623 * Collect new round-trip time estimate
1624 * and update averages and current timeout.
1628 tcp_xmit_timer(tp
, rtt
)
1629 register struct tcpcb
*tp
;
1632 register short delta
;
1634 DEBUG_CALL("tcp_xmit_timer");
1635 DEBUG_ARG("tp = %lx", (long)tp
);
1636 DEBUG_ARG("rtt = %d", rtt
);
1638 tcpstat
.tcps_rttupdated
++;
1639 if (tp
->t_srtt
!= 0) {
1641 * srtt is stored as fixed point with 3 bits after the
1642 * binary point (i.e., scaled by 8). The following magic
1643 * is equivalent to the smoothing algorithm in rfc793 with
1644 * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1645 * point). Adjust rtt to origin 0.
1647 delta
= rtt
- 1 - (tp
->t_srtt
>> TCP_RTT_SHIFT
);
1648 if ((tp
->t_srtt
+= delta
) <= 0)
1651 * We accumulate a smoothed rtt variance (actually, a
1652 * smoothed mean difference), then set the retransmit
1653 * timer to smoothed rtt + 4 times the smoothed variance.
1654 * rttvar is stored as fixed point with 2 bits after the
1655 * binary point (scaled by 4). The following is
1656 * equivalent to rfc793 smoothing with an alpha of .75
1657 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces
1658 * rfc793's wired-in beta.
1662 delta
-= (tp
->t_rttvar
>> TCP_RTTVAR_SHIFT
);
1663 if ((tp
->t_rttvar
+= delta
) <= 0)
1667 * No rtt measurement yet - use the unsmoothed rtt.
1668 * Set the variance to half the rtt (so our first
1669 * retransmit happens at 3*rtt).
1671 tp
->t_srtt
= rtt
<< TCP_RTT_SHIFT
;
1672 tp
->t_rttvar
= rtt
<< (TCP_RTTVAR_SHIFT
- 1);
1678 * the retransmit should happen at rtt + 4 * rttvar.
1679 * Because of the way we do the smoothing, srtt and rttvar
1680 * will each average +1/2 tick of bias. When we compute
1681 * the retransmit timer, we want 1/2 tick of rounding and
1682 * 1 extra tick because of +-1/2 tick uncertainty in the
1683 * firing of the timer. The bias will give us exactly the
1684 * 1.5 tick we need. But, because the bias is
1685 * statistical, we have to test that we don't drop below
1686 * the minimum feasible timer (which is 2 ticks).
1688 TCPT_RANGESET(tp
->t_rxtcur
, TCP_REXMTVAL(tp
),
1689 (short)tp
->t_rttmin
, TCPTV_REXMTMAX
); /* XXX */
1692 * We received an ack for a packet that wasn't retransmitted;
1693 * it is probably safe to discard any error indications we've
1694 * received recently. This isn't quite right, but close enough
1695 * for now (a route might have failed after we sent a segment,
1696 * and the return path might not be symmetrical).
1698 tp
->t_softerror
= 0;
1702 * Determine a reasonable value for maxseg size.
1703 * If the route is known, check route for mtu.
1704 * If none, use an mss that can be handled on the outgoing
1705 * interface without forcing IP to fragment; if bigger than
1706 * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1707 * to utilize large mbufs. If no route is found, route has no mtu,
1708 * or the destination isn't local, use a default, hopefully conservative
1709 * size (usually 512 or the default IP max size, but no more than the mtu
1710 * of the interface), as we can't discover anything about intervening
1711 * gateways or networks. We also initialize the congestion/slow start
1712 * window to be a single segment if the destination isn't local.
1713 * While looking at the routing entry, we also initialize other path-dependent
1714 * parameters from pre-set or cached values in the routing entry.
1719 register struct tcpcb
*tp
;
1722 struct socket
*so
= tp
->t_socket
;
1725 DEBUG_CALL("tcp_mss");
1726 DEBUG_ARG("tp = %lx", (long)tp
);
1727 DEBUG_ARG("offer = %d", offer
);
1729 mss
= min(if_mtu
, if_mru
) - sizeof(struct tcpiphdr
);
1731 mss
= min(mss
, offer
);
1733 if (mss
< tp
->t_maxseg
|| offer
!= 0)
1738 sbreserve(&so
->so_snd
, tcp_sndspace
+((tcp_sndspace
%mss
)?(mss
-(tcp_sndspace
%mss
)):0));
1739 sbreserve(&so
->so_rcv
, tcp_rcvspace
+((tcp_rcvspace
%mss
)?(mss
-(tcp_rcvspace
%mss
)):0));
1741 DEBUG_MISC((dfd
, " returning mss = %d\n", mss
));