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. Neither the name of the University nor the names of its contributors
14 * may be used to endorse or promote products derived from this software
15 * without specific prior written permission.
17 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
18 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
19 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
20 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
21 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
22 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
23 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
24 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
25 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
26 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * @(#)tcp_input.c 8.5 (Berkeley) 4/10/94
30 * tcp_input.c,v 1.10 1994/10/13 18:36:32 wollman Exp
34 * Changes and additions relating to SLiRP
35 * Copyright (c) 1995 Danny Gasparovski.
37 * Please read the file COPYRIGHT for the
38 * terms and conditions of the copyright.
41 #include "qemu/osdep.h"
45 #define TCPREXMTTHRESH 3
47 #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
49 /* for modulo comparisons of timestamps */
50 #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
51 #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
54 * Insert segment ti into reassembly queue of tcp with
55 * control block tp. Return TH_FIN if reassembly now includes
56 * a segment with FIN. The macro form does the common case inline
57 * (segment is the next to be received on an established connection,
58 * and the queue is empty), avoiding linkage into and removal
59 * from the queue and repetition of various conversions.
60 * Set DELACK for segments received in order, but ack immediately
61 * when segments are out of order (so fast retransmit can work).
63 #define TCP_REASS(tp, ti, m, so, flags) { \
64 if ((ti)->ti_seq == (tp)->rcv_nxt && \
65 tcpfrag_list_empty(tp) && \
66 (tp)->t_state == TCPS_ESTABLISHED) { \
67 tp->t_flags |= TF_DELACK; \
68 (tp)->rcv_nxt += (ti)->ti_len; \
69 flags = (ti)->ti_flags & TH_FIN; \
71 if (tcp_emu((so),(m))) sbappend(so, (m)); \
73 sbappend((so), (m)); \
75 (flags) = tcp_reass((tp), (ti), (m)); \
76 tp->t_flags |= TF_ACKNOW; \
80 static void tcp_dooptions(struct tcpcb
*tp
, u_char
*cp
, int cnt
,
82 static void tcp_xmit_timer(register struct tcpcb
*tp
, int rtt
);
85 tcp_reass(register struct tcpcb
*tp
, register struct tcpiphdr
*ti
,
88 register struct tcpiphdr
*q
;
89 struct socket
*so
= tp
->t_socket
;
93 * Call with ti==NULL after become established to
94 * force pre-ESTABLISHED data up to user socket.
100 * Find a segment which begins after this one does.
102 for (q
= tcpfrag_list_first(tp
); !tcpfrag_list_end(q
, tp
);
103 q
= tcpiphdr_next(q
))
104 if (SEQ_GT(q
->ti_seq
, ti
->ti_seq
))
108 * If there is a preceding segment, it may provide some of
109 * our data already. If so, drop the data from the incoming
110 * segment. If it provides all of our data, drop us.
112 if (!tcpfrag_list_end(tcpiphdr_prev(q
), tp
)) {
114 q
= tcpiphdr_prev(q
);
115 /* conversion to int (in i) handles seq wraparound */
116 i
= q
->ti_seq
+ q
->ti_len
- ti
->ti_seq
;
118 if (i
>= ti
->ti_len
) {
121 * Try to present any queued data
122 * at the left window edge to the user.
123 * This is needed after the 3-WHS
126 goto present
; /* ??? */
132 q
= tcpiphdr_next(q
);
137 * While we overlap succeeding segments trim them or,
138 * if they are completely covered, dequeue them.
140 while (!tcpfrag_list_end(q
, tp
)) {
141 register int i
= (ti
->ti_seq
+ ti
->ti_len
) - q
->ti_seq
;
147 m_adj(q
->ti_mbuf
, i
);
150 q
= tcpiphdr_next(q
);
151 m
= tcpiphdr_prev(q
)->ti_mbuf
;
152 remque(tcpiphdr2qlink(tcpiphdr_prev(q
)));
157 * Stick new segment in its place.
159 insque(tcpiphdr2qlink(ti
), tcpiphdr2qlink(tcpiphdr_prev(q
)));
163 * Present data to user, advancing rcv_nxt through
164 * completed sequence space.
166 if (!TCPS_HAVEESTABLISHED(tp
->t_state
))
168 ti
= tcpfrag_list_first(tp
);
169 if (tcpfrag_list_end(ti
, tp
) || ti
->ti_seq
!= tp
->rcv_nxt
)
171 if (tp
->t_state
== TCPS_SYN_RECEIVED
&& ti
->ti_len
)
174 tp
->rcv_nxt
+= ti
->ti_len
;
175 flags
= ti
->ti_flags
& TH_FIN
;
176 remque(tcpiphdr2qlink(ti
));
178 ti
= tcpiphdr_next(ti
);
179 if (so
->so_state
& SS_FCANTSENDMORE
)
183 if (tcp_emu(so
,m
)) sbappend(so
, m
);
187 } while (ti
!= (struct tcpiphdr
*)tp
&& ti
->ti_seq
== tp
->rcv_nxt
);
192 * TCP input routine, follows pages 65-76 of the
193 * protocol specification dated September, 1981 very closely.
196 tcp_input(struct mbuf
*m
, int iphlen
, struct socket
*inso
, unsigned short af
)
198 struct ip save_ip
, *ip
;
199 struct ip6 save_ip6
, *ip6
;
200 register struct tcpiphdr
*ti
;
204 register struct tcpcb
*tp
= NULL
;
205 register int tiflags
;
206 struct socket
*so
= NULL
;
207 int todrop
, acked
, ourfinisacked
, needoutput
= 0;
211 struct sockaddr_storage lhost
, fhost
;
212 struct sockaddr_in
*lhost4
, *fhost4
;
213 struct sockaddr_in6
*lhost6
, *fhost6
;
214 struct gfwd_list
*ex_ptr
;
217 DEBUG_CALL("tcp_input");
218 DEBUG_ARG("m = %p iphlen = %2d inso = %p",
222 * If called with m == 0, then we're continuing the connect
228 /* Re-set a few variables */
234 tiflags
= ti
->ti_flags
;
240 ip
= mtod(m
, struct ip
*);
241 ip6
= mtod(m
, struct ip6
*);
245 if (iphlen
> sizeof(struct ip
)) {
246 ip_stripoptions(m
, (struct mbuf
*)0);
247 iphlen
= sizeof(struct ip
);
249 /* XXX Check if too short */
253 * Save a copy of the IP header in case we want restore it
254 * for sending an ICMP error message in response.
257 save_ip
.ip_len
+= iphlen
;
260 * Get IP and TCP header together in first mbuf.
261 * Note: IP leaves IP header in first mbuf.
263 m
->m_data
-= sizeof(struct tcpiphdr
) - sizeof(struct ip
)
264 - sizeof(struct tcphdr
);
265 m
->m_len
+= sizeof(struct tcpiphdr
) - sizeof(struct ip
)
266 - sizeof(struct tcphdr
);
267 ti
= mtod(m
, struct tcpiphdr
*);
270 * Checksum extended TCP header and data.
273 tcpiphdr2qlink(ti
)->next
= tcpiphdr2qlink(ti
)->prev
= NULL
;
274 memset(&ti
->ih_mbuf
, 0 , sizeof(struct mbuf_ptr
));
275 memset(&ti
->ti
, 0, sizeof(ti
->ti
));
277 ti
->ti_src
= save_ip
.ip_src
;
278 ti
->ti_dst
= save_ip
.ip_dst
;
279 ti
->ti_pr
= save_ip
.ip_p
;
280 ti
->ti_len
= htons((uint16_t)tlen
);
285 * Save a copy of the IP header in case we want restore it
286 * for sending an ICMP error message in response.
290 * Get IP and TCP header together in first mbuf.
291 * Note: IP leaves IP header in first mbuf.
293 m
->m_data
-= sizeof(struct tcpiphdr
) - (sizeof(struct ip6
)
294 + sizeof(struct tcphdr
));
295 m
->m_len
+= sizeof(struct tcpiphdr
) - (sizeof(struct ip6
)
296 + sizeof(struct tcphdr
));
297 ti
= mtod(m
, struct tcpiphdr
*);
300 tcpiphdr2qlink(ti
)->next
= tcpiphdr2qlink(ti
)->prev
= NULL
;
301 memset(&ti
->ih_mbuf
, 0 , sizeof(struct mbuf_ptr
));
302 memset(&ti
->ti
, 0, sizeof(ti
->ti
));
304 ti
->ti_src6
= save_ip6
.ip_src
;
305 ti
->ti_dst6
= save_ip6
.ip_dst
;
306 ti
->ti_nh6
= save_ip6
.ip_nh
;
307 ti
->ti_len
= htons((uint16_t)tlen
);
311 g_assert_not_reached();
314 len
= ((sizeof(struct tcpiphdr
) - sizeof(struct tcphdr
)) + tlen
);
320 * Check that TCP offset makes sense,
321 * pull out TCP options and adjust length. XXX
323 off
= ti
->ti_off
<< 2;
324 if (off
< sizeof (struct tcphdr
) || off
> tlen
) {
329 if (off
> sizeof (struct tcphdr
)) {
330 optlen
= off
- sizeof (struct tcphdr
);
331 optp
= mtod(m
, caddr_t
) + sizeof (struct tcpiphdr
);
333 tiflags
= ti
->ti_flags
;
336 * Convert TCP protocol specific fields to host format.
344 * Drop TCP, IP headers and TCP options.
346 m
->m_data
+= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
347 m
->m_len
-= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
350 * Locate pcb for segment.
353 lhost
.ss_family
= af
;
354 fhost
.ss_family
= af
;
357 lhost4
= (struct sockaddr_in
*) &lhost
;
358 lhost4
->sin_addr
= ti
->ti_src
;
359 lhost4
->sin_port
= ti
->ti_sport
;
360 fhost4
= (struct sockaddr_in
*) &fhost
;
361 fhost4
->sin_addr
= ti
->ti_dst
;
362 fhost4
->sin_port
= ti
->ti_dport
;
365 lhost6
= (struct sockaddr_in6
*) &lhost
;
366 lhost6
->sin6_addr
= ti
->ti_src6
;
367 lhost6
->sin6_port
= ti
->ti_sport
;
368 fhost6
= (struct sockaddr_in6
*) &fhost
;
369 fhost6
->sin6_addr
= ti
->ti_dst6
;
370 fhost6
->sin6_port
= ti
->ti_dport
;
373 g_assert_not_reached();
376 so
= solookup(&slirp
->tcp_last_so
, &slirp
->tcb
, &lhost
, &fhost
);
379 * If the state is CLOSED (i.e., TCB does not exist) then
380 * all data in the incoming segment is discarded.
381 * If the TCB exists but is in CLOSED state, it is embryonic,
382 * but should either do a listen or a connect soon.
384 * state == CLOSED means we've done socreate() but haven't
385 * attached it to a protocol yet...
387 * XXX If a TCB does not exist, and the TH_SYN flag is
388 * the only flag set, then create a session, mark it
389 * as if it was LISTENING, and continue...
392 if (slirp
->restricted
) {
393 /* Any hostfwds will have an existing socket, so we only get here
394 * for non-hostfwd connections. These should be dropped, unless it
395 * happens to be a guestfwd.
397 for (ex_ptr
= slirp
->guestfwd_list
; ex_ptr
; ex_ptr
= ex_ptr
->ex_next
) {
398 if (ex_ptr
->ex_fport
== ti
->ti_dport
&&
399 ti
->ti_dst
.s_addr
== ex_ptr
->ex_addr
.s_addr
) {
408 if ((tiflags
& (TH_SYN
|TH_FIN
|TH_RST
|TH_URG
|TH_ACK
)) != TH_SYN
)
411 so
= socreate(slirp
);
412 if (tcp_attach(so
) < 0) {
413 g_free(so
); /* Not sofree (if it failed, it's not insqued) */
417 sbreserve(&so
->so_snd
, TCP_SNDSPACE
);
418 sbreserve(&so
->so_rcv
, TCP_RCVSPACE
);
420 so
->lhost
.ss
= lhost
;
421 so
->fhost
.ss
= fhost
;
423 so
->so_iptos
= tcp_tos(so
);
424 if (so
->so_iptos
== 0) {
427 so
->so_iptos
= ((struct ip
*)ti
)->ip_tos
;
432 g_assert_not_reached();
437 tp
->t_state
= TCPS_LISTEN
;
441 * If this is a still-connecting socket, this probably
442 * a retransmit of the SYN. Whether it's a retransmit SYN
443 * or something else, we nuke it.
445 if (so
->so_state
& SS_ISFCONNECTING
)
450 /* XXX Should never fail */
453 if (tp
->t_state
== TCPS_CLOSED
)
459 * Segment received on connection.
460 * Reset idle time and keep-alive timer.
463 if (slirp_do_keepalive
)
464 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEPINTVL
;
466 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEP_IDLE
;
469 * Process options if not in LISTEN state,
470 * else do it below (after getting remote address).
472 if (optp
&& tp
->t_state
!= TCPS_LISTEN
)
473 tcp_dooptions(tp
, (u_char
*)optp
, optlen
, ti
);
476 * Header prediction: check for the two common cases
477 * of a uni-directional data xfer. If the packet has
478 * no control flags, is in-sequence, the window didn't
479 * change and we're not retransmitting, it's a
480 * candidate. If the length is zero and the ack moved
481 * forward, we're the sender side of the xfer. Just
482 * free the data acked & wake any higher level process
483 * that was blocked waiting for space. If the length
484 * is non-zero and the ack didn't move, we're the
485 * receiver side. If we're getting packets in-order
486 * (the reassembly queue is empty), add the data to
487 * the socket buffer and note that we need a delayed ack.
489 * XXX Some of these tests are not needed
490 * eg: the tiwin == tp->snd_wnd prevents many more
491 * predictions.. with no *real* advantage..
493 if (tp
->t_state
== TCPS_ESTABLISHED
&&
494 (tiflags
& (TH_SYN
|TH_FIN
|TH_RST
|TH_URG
|TH_ACK
)) == TH_ACK
&&
495 ti
->ti_seq
== tp
->rcv_nxt
&&
496 tiwin
&& tiwin
== tp
->snd_wnd
&&
497 tp
->snd_nxt
== tp
->snd_max
) {
498 if (ti
->ti_len
== 0) {
499 if (SEQ_GT(ti
->ti_ack
, tp
->snd_una
) &&
500 SEQ_LEQ(ti
->ti_ack
, tp
->snd_max
) &&
501 tp
->snd_cwnd
>= tp
->snd_wnd
) {
503 * this is a pure ack for outstanding data.
506 SEQ_GT(ti
->ti_ack
, tp
->t_rtseq
))
507 tcp_xmit_timer(tp
, tp
->t_rtt
);
508 acked
= ti
->ti_ack
- tp
->snd_una
;
509 sbdrop(&so
->so_snd
, acked
);
510 tp
->snd_una
= ti
->ti_ack
;
514 * If all outstanding data are acked, stop
515 * retransmit timer, otherwise restart timer
516 * using current (possibly backed-off) value.
517 * If process is waiting for space,
518 * wakeup/selwakeup/signal. If data
519 * are ready to send, let tcp_output
520 * decide between more output or persist.
522 if (tp
->snd_una
== tp
->snd_max
)
523 tp
->t_timer
[TCPT_REXMT
] = 0;
524 else if (tp
->t_timer
[TCPT_PERSIST
] == 0)
525 tp
->t_timer
[TCPT_REXMT
] = tp
->t_rxtcur
;
528 * This is called because sowwakeup might have
529 * put data into so_snd. Since we don't so sowwakeup,
530 * we don't need this.. XXX???
532 if (so
->so_snd
.sb_cc
)
533 (void) tcp_output(tp
);
537 } else if (ti
->ti_ack
== tp
->snd_una
&&
538 tcpfrag_list_empty(tp
) &&
539 ti
->ti_len
<= sbspace(&so
->so_rcv
)) {
541 * this is a pure, in-sequence data packet
542 * with nothing on the reassembly queue and
543 * we have enough buffer space to take it.
545 tp
->rcv_nxt
+= ti
->ti_len
;
547 * Add data to socket buffer.
550 if (tcp_emu(so
,m
)) sbappend(so
, m
);
555 * If this is a short packet, then ACK now - with Nagel
556 * congestion avoidance sender won't send more until
559 * It is better to not delay acks at all to maximize
560 * TCP throughput. See RFC 2581.
562 tp
->t_flags
|= TF_ACKNOW
;
566 } /* header prediction */
568 * Calculate amount of space in receive window,
569 * and then do TCP input processing.
570 * Receive window is amount of space in rcv queue,
571 * but not less than advertised window.
574 win
= sbspace(&so
->so_rcv
);
577 tp
->rcv_wnd
= MAX(win
, (int)(tp
->rcv_adv
- tp
->rcv_nxt
));
580 switch (tp
->t_state
) {
583 * If the state is LISTEN then ignore segment if it contains an RST.
584 * If the segment contains an ACK then it is bad and send a RST.
585 * If it does not contain a SYN then it is not interesting; drop it.
586 * Don't bother responding if the destination was a broadcast.
587 * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
588 * tp->iss, and send a segment:
589 * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
590 * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
591 * Fill in remote peer address fields if not previously specified.
592 * Enter SYN_RECEIVED state, and process any other fields of this
593 * segment in this state.
597 if (tiflags
& TH_RST
)
599 if (tiflags
& TH_ACK
)
601 if ((tiflags
& TH_SYN
) == 0)
605 * This has way too many gotos...
606 * But a bit of spaghetti code never hurt anybody :)
610 * If this is destined for the control address, then flag to
611 * tcp_ctl once connected, otherwise connect
614 (so
->so_faddr
.s_addr
& slirp
->vnetwork_mask
.s_addr
) ==
615 slirp
->vnetwork_addr
.s_addr
) {
616 if (so
->so_faddr
.s_addr
!= slirp
->vhost_addr
.s_addr
&&
617 so
->so_faddr
.s_addr
!= slirp
->vnameserver_addr
.s_addr
) {
618 /* May be an add exec */
619 for (ex_ptr
= slirp
->guestfwd_list
; ex_ptr
;
620 ex_ptr
= ex_ptr
->ex_next
) {
621 if(ex_ptr
->ex_fport
== so
->so_fport
&&
622 so
->so_faddr
.s_addr
== ex_ptr
->ex_addr
.s_addr
) {
623 so
->so_state
|= SS_CTL
;
627 if (so
->so_state
& SS_CTL
) {
631 /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
634 if (so
->so_emu
& EMU_NOCONNECT
) {
635 so
->so_emu
&= ~EMU_NOCONNECT
;
639 if ((tcp_fconnect(so
, so
->so_ffamily
) == -1) &&
641 (errno
!= EINPROGRESS
) && (errno
!= EWOULDBLOCK
)
644 DEBUG_MISC(" tcp fconnect errno = %d-%s", errno
, strerror(errno
));
645 if(errno
== ECONNREFUSED
) {
646 /* ACK the SYN, send RST to refuse the connection */
647 tcp_respond(tp
, ti
, m
, ti
->ti_seq
+ 1, (tcp_seq
) 0,
648 TH_RST
| TH_ACK
, af
);
652 code
= ICMP_UNREACH_NET
;
653 if (errno
== EHOSTUNREACH
) {
654 code
= ICMP_UNREACH_HOST
;
658 code
= ICMP6_UNREACH_NO_ROUTE
;
659 if (errno
== EHOSTUNREACH
) {
660 code
= ICMP6_UNREACH_ADDRESS
;
664 g_assert_not_reached();
666 HTONL(ti
->ti_seq
); /* restore tcp header */
670 m
->m_data
-= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
671 m
->m_len
+= sizeof(struct tcpiphdr
)+off
-sizeof(struct tcphdr
);
674 m
->m_data
+= sizeof(struct tcpiphdr
) - sizeof(struct ip
)
675 - sizeof(struct tcphdr
);
676 m
->m_len
-= sizeof(struct tcpiphdr
) - sizeof(struct ip
)
677 - sizeof(struct tcphdr
);
679 icmp_send_error(m
, ICMP_UNREACH
, code
, 0, strerror(errno
));
682 m
->m_data
+= sizeof(struct tcpiphdr
) - (sizeof(struct ip6
)
683 + sizeof(struct tcphdr
));
684 m
->m_len
-= sizeof(struct tcpiphdr
) - (sizeof(struct ip6
)
685 + sizeof(struct tcphdr
));
687 icmp6_send_error(m
, ICMP6_UNREACH
, code
);
690 g_assert_not_reached();
697 * Haven't connected yet, save the current mbuf
699 * XXX Some OS's don't tell us whether the connect()
700 * succeeded or not. So we must time it out.
704 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEP_INIT
;
705 tp
->t_state
= TCPS_SYN_RECEIVED
;
707 * Initialize receive sequence numbers now so that we can send a
708 * valid RST if the remote end rejects our connection.
710 tp
->irs
= ti
->ti_seq
;
718 * Check if the connect succeeded
720 if (so
->so_state
& SS_NOFDREF
) {
728 tcp_dooptions(tp
, (u_char
*)optp
, optlen
, ti
);
733 tp
->iss
= slirp
->tcp_iss
;
734 slirp
->tcp_iss
+= TCP_ISSINCR
/2;
735 tp
->irs
= ti
->ti_seq
;
738 tp
->t_flags
|= TF_ACKNOW
;
739 tp
->t_state
= TCPS_SYN_RECEIVED
;
740 tp
->t_timer
[TCPT_KEEP
] = TCPTV_KEEP_INIT
;
742 } /* case TCPS_LISTEN */
745 * If the state is SYN_SENT:
746 * if seg contains an ACK, but not for our SYN, drop the input.
747 * if seg contains a RST, then drop the connection.
748 * if seg does not contain SYN, then drop it.
749 * Otherwise this is an acceptable SYN segment
750 * initialize tp->rcv_nxt and tp->irs
751 * if seg contains ack then advance tp->snd_una
752 * if SYN has been acked change to ESTABLISHED else SYN_RCVD state
753 * arrange for segment to be acked (eventually)
754 * continue processing rest of data/controls, beginning with URG
757 if ((tiflags
& TH_ACK
) &&
758 (SEQ_LEQ(ti
->ti_ack
, tp
->iss
) ||
759 SEQ_GT(ti
->ti_ack
, tp
->snd_max
)))
762 if (tiflags
& TH_RST
) {
763 if (tiflags
& TH_ACK
) {
764 tcp_drop(tp
, 0); /* XXX Check t_softerror! */
769 if ((tiflags
& TH_SYN
) == 0)
771 if (tiflags
& TH_ACK
) {
772 tp
->snd_una
= ti
->ti_ack
;
773 if (SEQ_LT(tp
->snd_nxt
, tp
->snd_una
))
774 tp
->snd_nxt
= tp
->snd_una
;
777 tp
->t_timer
[TCPT_REXMT
] = 0;
778 tp
->irs
= ti
->ti_seq
;
780 tp
->t_flags
|= TF_ACKNOW
;
781 if (tiflags
& TH_ACK
&& SEQ_GT(tp
->snd_una
, tp
->iss
)) {
783 tp
->t_state
= TCPS_ESTABLISHED
;
785 (void) tcp_reass(tp
, (struct tcpiphdr
*)0,
788 * if we didn't have to retransmit the SYN,
789 * use its rtt as our initial srtt & rtt var.
792 tcp_xmit_timer(tp
, tp
->t_rtt
);
794 tp
->t_state
= TCPS_SYN_RECEIVED
;
798 * Advance ti->ti_seq to correspond to first data byte.
799 * If data, trim to stay within window,
800 * dropping FIN if necessary.
803 if (ti
->ti_len
> tp
->rcv_wnd
) {
804 todrop
= ti
->ti_len
- tp
->rcv_wnd
;
806 ti
->ti_len
= tp
->rcv_wnd
;
809 tp
->snd_wl1
= ti
->ti_seq
- 1;
810 tp
->rcv_up
= ti
->ti_seq
;
812 } /* switch tp->t_state */
814 * States other than LISTEN or SYN_SENT.
815 * Check that at least some bytes of segment are within
816 * receive window. If segment begins before rcv_nxt,
817 * drop leading data (and SYN); if nothing left, just ack.
819 todrop
= tp
->rcv_nxt
- ti
->ti_seq
;
821 if (tiflags
& TH_SYN
) {
831 * Following if statement from Stevens, vol. 2, p. 960.
833 if (todrop
> ti
->ti_len
834 || (todrop
== ti
->ti_len
&& (tiflags
& TH_FIN
) == 0)) {
836 * Any valid FIN must be to the left of the window.
837 * At this point the FIN must be a duplicate or out
838 * of sequence; drop it.
843 * Send an ACK to resynchronize and drop any data.
844 * But keep on processing for RST or ACK.
846 tp
->t_flags
|= TF_ACKNOW
;
850 ti
->ti_seq
+= todrop
;
851 ti
->ti_len
-= todrop
;
852 if (ti
->ti_urp
> todrop
)
853 ti
->ti_urp
-= todrop
;
860 * If new data are received on a connection after the
861 * user processes are gone, then RST the other end.
863 if ((so
->so_state
& SS_NOFDREF
) &&
864 tp
->t_state
> TCPS_CLOSE_WAIT
&& ti
->ti_len
) {
870 * If segment ends after window, drop trailing data
871 * (and PUSH and FIN); if nothing left, just ACK.
873 todrop
= (ti
->ti_seq
+ti
->ti_len
) - (tp
->rcv_nxt
+tp
->rcv_wnd
);
875 if (todrop
>= ti
->ti_len
) {
877 * If a new connection request is received
878 * while in TIME_WAIT, drop the old connection
879 * and start over if the sequence numbers
880 * are above the previous ones.
882 if (tiflags
& TH_SYN
&&
883 tp
->t_state
== TCPS_TIME_WAIT
&&
884 SEQ_GT(ti
->ti_seq
, tp
->rcv_nxt
)) {
885 iss
= tp
->rcv_nxt
+ TCP_ISSINCR
;
890 * If window is closed can only take segments at
891 * window edge, and have to drop data and PUSH from
892 * incoming segments. Continue processing, but
893 * remember to ack. Otherwise, drop segment
896 if (tp
->rcv_wnd
== 0 && ti
->ti_seq
== tp
->rcv_nxt
) {
897 tp
->t_flags
|= TF_ACKNOW
;
903 ti
->ti_len
-= todrop
;
904 tiflags
&= ~(TH_PUSH
|TH_FIN
);
908 * If the RST bit is set examine the state:
909 * SYN_RECEIVED STATE:
910 * If passive open, return to LISTEN state.
911 * If active open, inform user that connection was refused.
912 * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
913 * Inform user that connection was reset, and close tcb.
914 * CLOSING, LAST_ACK, TIME_WAIT STATES
917 if (tiflags
&TH_RST
) switch (tp
->t_state
) {
919 case TCPS_SYN_RECEIVED
:
920 case TCPS_ESTABLISHED
:
921 case TCPS_FIN_WAIT_1
:
922 case TCPS_FIN_WAIT_2
:
923 case TCPS_CLOSE_WAIT
:
924 tp
->t_state
= TCPS_CLOSED
;
936 * If a SYN is in the window, then this is an
937 * error and we send an RST and drop the connection.
939 if (tiflags
& TH_SYN
) {
945 * If the ACK bit is off we drop the segment and return.
947 if ((tiflags
& TH_ACK
) == 0) goto drop
;
952 switch (tp
->t_state
) {
954 * In SYN_RECEIVED state if the ack ACKs our SYN then enter
955 * ESTABLISHED state and continue processing, otherwise
956 * send an RST. una<=ack<=max
958 case TCPS_SYN_RECEIVED
:
960 if (SEQ_GT(tp
->snd_una
, ti
->ti_ack
) ||
961 SEQ_GT(ti
->ti_ack
, tp
->snd_max
))
963 tp
->t_state
= TCPS_ESTABLISHED
;
965 * The sent SYN is ack'ed with our sequence number +1
966 * The first data byte already in the buffer will get
967 * lost if no correction is made. This is only needed for
968 * SS_CTL since the buffer is empty otherwise.
971 tp
->snd_una
=ti
->ti_ack
;
972 if (so
->so_state
& SS_CTL
) {
973 /* So tcp_ctl reports the right state */
977 so
->so_state
&= ~SS_CTL
; /* success XXX */
978 } else if (ret
== 2) {
979 so
->so_state
&= SS_PERSISTENT_MASK
;
980 so
->so_state
|= SS_NOFDREF
; /* CTL_CMD */
983 tp
->t_state
= TCPS_FIN_WAIT_1
;
989 (void) tcp_reass(tp
, (struct tcpiphdr
*)0, (struct mbuf
*)0);
990 tp
->snd_wl1
= ti
->ti_seq
- 1;
991 /* Avoid ack processing; snd_una==ti_ack => dup ack */
996 * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
997 * ACKs. If the ack is in the range
998 * tp->snd_una < ti->ti_ack <= tp->snd_max
999 * then advance tp->snd_una to ti->ti_ack and drop
1000 * data from the retransmission queue. If this ACK reflects
1001 * more up to date window information we update our window information.
1003 case TCPS_ESTABLISHED
:
1004 case TCPS_FIN_WAIT_1
:
1005 case TCPS_FIN_WAIT_2
:
1006 case TCPS_CLOSE_WAIT
:
1009 case TCPS_TIME_WAIT
:
1011 if (SEQ_LEQ(ti
->ti_ack
, tp
->snd_una
)) {
1012 if (ti
->ti_len
== 0 && tiwin
== tp
->snd_wnd
) {
1013 DEBUG_MISC(" dup ack m = %p so = %p", m
, so
);
1015 * If we have outstanding data (other than
1016 * a window probe), this is a completely
1017 * duplicate ack (ie, window info didn't
1018 * change), the ack is the biggest we've
1019 * seen and we've seen exactly our rexmt
1020 * threshold of them, assume a packet
1021 * has been dropped and retransmit it.
1022 * Kludge snd_nxt & the congestion
1023 * window so we send only this one
1026 * We know we're losing at the current
1027 * window size so do congestion avoidance
1028 * (set ssthresh to half the current window
1029 * and pull our congestion window back to
1030 * the new ssthresh).
1032 * Dup acks mean that packets have left the
1033 * network (they're now cached at the receiver)
1034 * so bump cwnd by the amount in the receiver
1035 * to keep a constant cwnd packets in the
1038 if (tp
->t_timer
[TCPT_REXMT
] == 0 ||
1039 ti
->ti_ack
!= tp
->snd_una
)
1041 else if (++tp
->t_dupacks
== TCPREXMTTHRESH
) {
1042 tcp_seq onxt
= tp
->snd_nxt
;
1044 MIN(tp
->snd_wnd
, tp
->snd_cwnd
) /
1049 tp
->snd_ssthresh
= win
* tp
->t_maxseg
;
1050 tp
->t_timer
[TCPT_REXMT
] = 0;
1052 tp
->snd_nxt
= ti
->ti_ack
;
1053 tp
->snd_cwnd
= tp
->t_maxseg
;
1054 (void) tcp_output(tp
);
1055 tp
->snd_cwnd
= tp
->snd_ssthresh
+
1056 tp
->t_maxseg
* tp
->t_dupacks
;
1057 if (SEQ_GT(onxt
, tp
->snd_nxt
))
1060 } else if (tp
->t_dupacks
> TCPREXMTTHRESH
) {
1061 tp
->snd_cwnd
+= tp
->t_maxseg
;
1062 (void) tcp_output(tp
);
1071 * If the congestion window was inflated to account
1072 * for the other side's cached packets, retract it.
1074 if (tp
->t_dupacks
> TCPREXMTTHRESH
&&
1075 tp
->snd_cwnd
> tp
->snd_ssthresh
)
1076 tp
->snd_cwnd
= tp
->snd_ssthresh
;
1078 if (SEQ_GT(ti
->ti_ack
, tp
->snd_max
)) {
1081 acked
= ti
->ti_ack
- tp
->snd_una
;
1084 * If transmit timer is running and timed sequence
1085 * number was acked, update smoothed round trip time.
1086 * Since we now have an rtt measurement, cancel the
1087 * timer backoff (cf., Phil Karn's retransmit alg.).
1088 * Recompute the initial retransmit timer.
1090 if (tp
->t_rtt
&& SEQ_GT(ti
->ti_ack
, tp
->t_rtseq
))
1091 tcp_xmit_timer(tp
,tp
->t_rtt
);
1094 * If all outstanding data is acked, stop retransmit
1095 * timer and remember to restart (more output or persist).
1096 * If there is more data to be acked, restart retransmit
1097 * timer, using current (possibly backed-off) value.
1099 if (ti
->ti_ack
== tp
->snd_max
) {
1100 tp
->t_timer
[TCPT_REXMT
] = 0;
1102 } else if (tp
->t_timer
[TCPT_PERSIST
] == 0)
1103 tp
->t_timer
[TCPT_REXMT
] = tp
->t_rxtcur
;
1105 * When new data is acked, open the congestion window.
1106 * If the window gives us less than ssthresh packets
1107 * in flight, open exponentially (maxseg per packet).
1108 * Otherwise open linearly: maxseg per window
1109 * (maxseg^2 / cwnd per packet).
1112 register u_int cw
= tp
->snd_cwnd
;
1113 register u_int incr
= tp
->t_maxseg
;
1115 if (cw
> tp
->snd_ssthresh
)
1116 incr
= incr
* incr
/ cw
;
1117 tp
->snd_cwnd
= MIN(cw
+ incr
, TCP_MAXWIN
<< tp
->snd_scale
);
1119 if (acked
> so
->so_snd
.sb_cc
) {
1120 tp
->snd_wnd
-= so
->so_snd
.sb_cc
;
1121 sbdrop(&so
->so_snd
, (int )so
->so_snd
.sb_cc
);
1124 sbdrop(&so
->so_snd
, acked
);
1125 tp
->snd_wnd
-= acked
;
1128 tp
->snd_una
= ti
->ti_ack
;
1129 if (SEQ_LT(tp
->snd_nxt
, tp
->snd_una
))
1130 tp
->snd_nxt
= tp
->snd_una
;
1132 switch (tp
->t_state
) {
1135 * In FIN_WAIT_1 STATE in addition to the processing
1136 * for the ESTABLISHED state if our FIN is now acknowledged
1137 * then enter FIN_WAIT_2.
1139 case TCPS_FIN_WAIT_1
:
1140 if (ourfinisacked
) {
1142 * If we can't receive any more
1143 * data, then closing user can proceed.
1144 * Starting the timer is contrary to the
1145 * specification, but if we don't get a FIN
1146 * we'll hang forever.
1148 if (so
->so_state
& SS_FCANTRCVMORE
) {
1149 tp
->t_timer
[TCPT_2MSL
] = TCP_MAXIDLE
;
1151 tp
->t_state
= TCPS_FIN_WAIT_2
;
1156 * In CLOSING STATE in addition to the processing for
1157 * the ESTABLISHED state if the ACK acknowledges our FIN
1158 * then enter the TIME-WAIT state, otherwise ignore
1162 if (ourfinisacked
) {
1163 tp
->t_state
= TCPS_TIME_WAIT
;
1164 tcp_canceltimers(tp
);
1165 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1170 * In LAST_ACK, we may still be waiting for data to drain
1171 * and/or to be acked, as well as for the ack of our FIN.
1172 * If our FIN is now acknowledged, delete the TCB,
1173 * enter the closed state and return.
1176 if (ourfinisacked
) {
1183 * In TIME_WAIT state the only thing that should arrive
1184 * is a retransmission of the remote FIN. Acknowledge
1185 * it and restart the finack timer.
1187 case TCPS_TIME_WAIT
:
1188 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1191 } /* switch(tp->t_state) */
1195 * Update window information.
1196 * Don't look at window if no ACK: TAC's send garbage on first SYN.
1198 if ((tiflags
& TH_ACK
) &&
1199 (SEQ_LT(tp
->snd_wl1
, ti
->ti_seq
) ||
1200 (tp
->snd_wl1
== ti
->ti_seq
&& (SEQ_LT(tp
->snd_wl2
, ti
->ti_ack
) ||
1201 (tp
->snd_wl2
== ti
->ti_ack
&& tiwin
> tp
->snd_wnd
))))) {
1202 tp
->snd_wnd
= tiwin
;
1203 tp
->snd_wl1
= ti
->ti_seq
;
1204 tp
->snd_wl2
= ti
->ti_ack
;
1205 if (tp
->snd_wnd
> tp
->max_sndwnd
)
1206 tp
->max_sndwnd
= tp
->snd_wnd
;
1211 * Process segments with URG.
1213 if ((tiflags
& TH_URG
) && ti
->ti_urp
&&
1214 TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1216 * This is a kludge, but if we receive and accept
1217 * random urgent pointers, we'll crash in
1218 * soreceive. It's hard to imagine someone
1219 * actually wanting to send this much urgent data.
1221 if (ti
->ti_urp
+ so
->so_rcv
.sb_cc
> so
->so_rcv
.sb_datalen
) {
1227 * If this segment advances the known urgent pointer,
1228 * then mark the data stream. This should not happen
1229 * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1230 * a FIN has been received from the remote side.
1231 * In these states we ignore the URG.
1233 * According to RFC961 (Assigned Protocols),
1234 * the urgent pointer points to the last octet
1235 * of urgent data. We continue, however,
1236 * to consider it to indicate the first octet
1237 * of data past the urgent section as the original
1238 * spec states (in one of two places).
1240 if (SEQ_GT(ti
->ti_seq
+ti
->ti_urp
, tp
->rcv_up
)) {
1241 tp
->rcv_up
= ti
->ti_seq
+ ti
->ti_urp
;
1242 so
->so_urgc
= so
->so_rcv
.sb_cc
+
1243 (tp
->rcv_up
- tp
->rcv_nxt
); /* -1; */
1244 tp
->rcv_up
= ti
->ti_seq
+ ti
->ti_urp
;
1249 * If no out of band data is expected,
1250 * pull receive urgent pointer along
1251 * with the receive window.
1253 if (SEQ_GT(tp
->rcv_nxt
, tp
->rcv_up
))
1254 tp
->rcv_up
= tp
->rcv_nxt
;
1258 * If this is a small packet, then ACK now - with Nagel
1259 * congestion avoidance sender won't send more until
1262 if (ti
->ti_len
&& (unsigned)ti
->ti_len
<= 5 &&
1263 ((struct tcpiphdr_2
*)ti
)->first_char
== (char)27) {
1264 tp
->t_flags
|= TF_ACKNOW
;
1268 * Process the segment text, merging it into the TCP sequencing queue,
1269 * and arranging for acknowledgment of receipt if necessary.
1270 * This process logically involves adjusting tp->rcv_wnd as data
1271 * is presented to the user (this happens in tcp_usrreq.c,
1272 * case PRU_RCVD). If a FIN has already been received on this
1273 * connection then we just ignore the text.
1275 if ((ti
->ti_len
|| (tiflags
&TH_FIN
)) &&
1276 TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1277 TCP_REASS(tp
, ti
, m
, so
, tiflags
);
1284 * If FIN is received ACK the FIN and let the user know
1285 * that the connection is closing.
1287 if (tiflags
& TH_FIN
) {
1288 if (TCPS_HAVERCVDFIN(tp
->t_state
) == 0) {
1290 * If we receive a FIN we can't send more data,
1292 * Shutdown the socket if there is no rx data in the
1294 * soread() is called on completion of shutdown() and
1295 * will got to TCPS_LAST_ACK, and use tcp_output()
1300 tp
->t_flags
|= TF_ACKNOW
;
1303 switch (tp
->t_state
) {
1306 * In SYN_RECEIVED and ESTABLISHED STATES
1307 * enter the CLOSE_WAIT state.
1309 case TCPS_SYN_RECEIVED
:
1310 case TCPS_ESTABLISHED
:
1311 if(so
->so_emu
== EMU_CTL
) /* no shutdown on socket */
1312 tp
->t_state
= TCPS_LAST_ACK
;
1314 tp
->t_state
= TCPS_CLOSE_WAIT
;
1318 * If still in FIN_WAIT_1 STATE FIN has not been acked so
1319 * enter the CLOSING state.
1321 case TCPS_FIN_WAIT_1
:
1322 tp
->t_state
= TCPS_CLOSING
;
1326 * In FIN_WAIT_2 state enter the TIME_WAIT state,
1327 * starting the time-wait timer, turning off the other
1330 case TCPS_FIN_WAIT_2
:
1331 tp
->t_state
= TCPS_TIME_WAIT
;
1332 tcp_canceltimers(tp
);
1333 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1337 * In TIME_WAIT state restart the 2 MSL time_wait timer.
1339 case TCPS_TIME_WAIT
:
1340 tp
->t_timer
[TCPT_2MSL
] = 2 * TCPTV_MSL
;
1346 * Return any desired output.
1348 if (needoutput
|| (tp
->t_flags
& TF_ACKNOW
)) {
1349 (void) tcp_output(tp
);
1355 * Generate an ACK dropping incoming segment if it occupies
1356 * sequence space, where the ACK reflects our state.
1358 if (tiflags
& TH_RST
)
1361 tp
->t_flags
|= TF_ACKNOW
;
1362 (void) tcp_output(tp
);
1366 /* reuses m if m!=NULL, m_free() unnecessary */
1367 if (tiflags
& TH_ACK
)
1368 tcp_respond(tp
, ti
, m
, (tcp_seq
)0, ti
->ti_ack
, TH_RST
, af
);
1370 if (tiflags
& TH_SYN
) ti
->ti_len
++;
1371 tcp_respond(tp
, ti
, m
, ti
->ti_seq
+ ti
->ti_len
, (tcp_seq
) 0,
1372 TH_RST
| TH_ACK
, af
);
1379 * Drop space held by incoming segment and return.
1385 tcp_dooptions(struct tcpcb
*tp
, u_char
*cp
, int cnt
, struct tcpiphdr
*ti
)
1390 DEBUG_CALL("tcp_dooptions");
1391 DEBUG_ARG("tp = %p cnt=%i", tp
, cnt
);
1393 for (; cnt
> 0; cnt
-= optlen
, cp
+= optlen
) {
1395 if (opt
== TCPOPT_EOL
)
1397 if (opt
== TCPOPT_NOP
)
1410 if (optlen
!= TCPOLEN_MAXSEG
)
1412 if (!(ti
->ti_flags
& TH_SYN
))
1414 memcpy((char *) &mss
, (char *) cp
+ 2, sizeof(mss
));
1416 (void) tcp_mss(tp
, mss
); /* sets t_maxseg */
1423 * Collect new round-trip time estimate
1424 * and update averages and current timeout.
1428 tcp_xmit_timer(register struct tcpcb
*tp
, int rtt
)
1430 register short delta
;
1432 DEBUG_CALL("tcp_xmit_timer");
1433 DEBUG_ARG("tp = %p", tp
);
1434 DEBUG_ARG("rtt = %d", rtt
);
1436 if (tp
->t_srtt
!= 0) {
1438 * srtt is stored as fixed point with 3 bits after the
1439 * binary point (i.e., scaled by 8). The following magic
1440 * is equivalent to the smoothing algorithm in rfc793 with
1441 * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1442 * point). Adjust rtt to origin 0.
1444 delta
= rtt
- 1 - (tp
->t_srtt
>> TCP_RTT_SHIFT
);
1445 if ((tp
->t_srtt
+= delta
) <= 0)
1448 * We accumulate a smoothed rtt variance (actually, a
1449 * smoothed mean difference), then set the retransmit
1450 * timer to smoothed rtt + 4 times the smoothed variance.
1451 * rttvar is stored as fixed point with 2 bits after the
1452 * binary point (scaled by 4). The following is
1453 * equivalent to rfc793 smoothing with an alpha of .75
1454 * (rttvar = rttvar*3/4 + |delta| / 4). This replaces
1455 * rfc793's wired-in beta.
1459 delta
-= (tp
->t_rttvar
>> TCP_RTTVAR_SHIFT
);
1460 if ((tp
->t_rttvar
+= delta
) <= 0)
1464 * No rtt measurement yet - use the unsmoothed rtt.
1465 * Set the variance to half the rtt (so our first
1466 * retransmit happens at 3*rtt).
1468 tp
->t_srtt
= rtt
<< TCP_RTT_SHIFT
;
1469 tp
->t_rttvar
= rtt
<< (TCP_RTTVAR_SHIFT
- 1);
1475 * the retransmit should happen at rtt + 4 * rttvar.
1476 * Because of the way we do the smoothing, srtt and rttvar
1477 * will each average +1/2 tick of bias. When we compute
1478 * the retransmit timer, we want 1/2 tick of rounding and
1479 * 1 extra tick because of +-1/2 tick uncertainty in the
1480 * firing of the timer. The bias will give us exactly the
1481 * 1.5 tick we need. But, because the bias is
1482 * statistical, we have to test that we don't drop below
1483 * the minimum feasible timer (which is 2 ticks).
1485 TCPT_RANGESET(tp
->t_rxtcur
, TCP_REXMTVAL(tp
),
1486 (short)tp
->t_rttmin
, TCPTV_REXMTMAX
); /* XXX */
1489 * We received an ack for a packet that wasn't retransmitted;
1490 * it is probably safe to discard any error indications we've
1491 * received recently. This isn't quite right, but close enough
1492 * for now (a route might have failed after we sent a segment,
1493 * and the return path might not be symmetrical).
1495 tp
->t_softerror
= 0;
1499 * Determine a reasonable value for maxseg size.
1500 * If the route is known, check route for mtu.
1501 * If none, use an mss that can be handled on the outgoing
1502 * interface without forcing IP to fragment; if bigger than
1503 * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1504 * to utilize large mbufs. If no route is found, route has no mtu,
1505 * or the destination isn't local, use a default, hopefully conservative
1506 * size (usually 512 or the default IP max size, but no more than the mtu
1507 * of the interface), as we can't discover anything about intervening
1508 * gateways or networks. We also initialize the congestion/slow start
1509 * window to be a single segment if the destination isn't local.
1510 * While looking at the routing entry, we also initialize other path-dependent
1511 * parameters from pre-set or cached values in the routing entry.
1515 tcp_mss(struct tcpcb
*tp
, u_int offer
)
1517 struct socket
*so
= tp
->t_socket
;
1520 DEBUG_CALL("tcp_mss");
1521 DEBUG_ARG("tp = %p", tp
);
1522 DEBUG_ARG("offer = %d", offer
);
1524 switch (so
->so_ffamily
) {
1526 mss
= MIN(IF_MTU
, IF_MRU
) - sizeof(struct tcphdr
)
1527 - sizeof(struct ip
);
1530 mss
= MIN(IF_MTU
, IF_MRU
) - sizeof(struct tcphdr
)
1531 - sizeof(struct ip6
);
1534 g_assert_not_reached();
1538 mss
= MIN(mss
, offer
);
1540 if (mss
< tp
->t_maxseg
|| offer
!= 0)
1545 sbreserve(&so
->so_snd
, TCP_SNDSPACE
+ ((TCP_SNDSPACE
% mss
) ?
1546 (mss
- (TCP_SNDSPACE
% mss
)) :
1548 sbreserve(&so
->so_rcv
, TCP_RCVSPACE
+ ((TCP_RCVSPACE
% mss
) ?
1549 (mss
- (TCP_RCVSPACE
% mss
)) :
1552 DEBUG_MISC(" returning mss = %d", mss
);