![[BACK]](/cvsweb/icons/back.gif){kind=icon}
Return to tcp_input.c CVS log ![[TXT]](/cvsweb/icons/text.gif){kind=icon}
![[DIR]](/cvsweb/icons/dir.gif){kind=icon}
Up to [Qemu by Fabrice Bellard] / qemu / slirp|
Return to tcp_input.c CVS log | Up to [Qemu by Fabrice Bellard] / qemu / slirp |
1.1 root 1: /*
2: * Copyright (c) 1982, 1986, 1988, 1990, 1993, 1994
3: * The Regents of the University of California. All rights reserved.
4: *
5: * Redistribution and use in source and binary forms, with or without
6: * modification, are permitted provided that the following conditions
7: * are met:
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.
20: *
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
31: * SUCH DAMAGE.
32: *
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
35: */
36:
37: /*
38: * Changes and additions relating to SLiRP
39: * Copyright (c) 1995 Danny Gasparovski.
40: *
41: * Please read the file COPYRIGHT for the
42: * terms and conditions of the copyright.
43: */
44:
45: #include <slirp.h>
46: #include "ip_icmp.h"
47:
48: struct socket tcb;
49:
50: int tcprexmtthresh = 3;
51: struct socket *tcp_last_so = &tcb;
52:
53: tcp_seq tcp_iss; /* tcp initial send seq # */
54:
55: #define TCP_PAWS_IDLE (24 * 24 * 60 * 60 * PR_SLOWHZ)
56:
57: /* for modulo comparisons of timestamps */
58: #define TSTMP_LT(a,b) ((int)((a)-(b)) < 0)
59: #define TSTMP_GEQ(a,b) ((int)((a)-(b)) >= 0)
60:
61: /*
62: * Insert segment ti into reassembly queue of tcp with
63: * control block tp. Return TH_FIN if reassembly now includes
64: * a segment with FIN. The macro form does the common case inline
65: * (segment is the next to be received on an established connection,
66: * and the queue is empty), avoiding linkage into and removal
67: * from the queue and repetition of various conversions.
68: * Set DELACK for segments received in order, but ack immediately
69: * when segments are out of order (so fast retransmit can work).
70: */
71: #ifdef TCP_ACK_HACK
72: #define TCP_REASS(tp, ti, m, so, flags) {\
73: if ((ti)->ti_seq == (tp)->rcv_nxt && \
74: (tp)->seg_next == (tcpiphdrp_32)(tp) && \
75: (tp)->t_state == TCPS_ESTABLISHED) {\
76: if (ti->ti_flags & TH_PUSH) \
77: tp->t_flags |= TF_ACKNOW; \
78: else \
79: tp->t_flags |= TF_DELACK; \
80: (tp)->rcv_nxt += (ti)->ti_len; \
81: flags = (ti)->ti_flags & TH_FIN; \
82: tcpstat.tcps_rcvpack++;\
83: tcpstat.tcps_rcvbyte += (ti)->ti_len;\
84: if (so->so_emu) { \
85: if (tcp_emu((so),(m))) sbappend((so), (m)); \
86: } else \
87: sbappend((so), (m)); \
88: /* sorwakeup(so); */ \
89: } else {\
90: (flags) = tcp_reass((tp), (ti), (m)); \
91: tp->t_flags |= TF_ACKNOW; \
92: } \
93: }
94: #else
95: #define TCP_REASS(tp, ti, m, so, flags) { \
96: if ((ti)->ti_seq == (tp)->rcv_nxt && \
97: (tp)->seg_next == (tcpiphdrp_32)(tp) && \
98: (tp)->t_state == TCPS_ESTABLISHED) { \
99: tp->t_flags |= TF_DELACK; \
100: (tp)->rcv_nxt += (ti)->ti_len; \
101: flags = (ti)->ti_flags & TH_FIN; \
102: tcpstat.tcps_rcvpack++;\
103: tcpstat.tcps_rcvbyte += (ti)->ti_len;\
104: if (so->so_emu) { \
105: if (tcp_emu((so),(m))) sbappend(so, (m)); \
106: } else \
107: sbappend((so), (m)); \
108: /* sorwakeup(so); */ \
109: } else { \
110: (flags) = tcp_reass((tp), (ti), (m)); \
111: tp->t_flags |= TF_ACKNOW; \
112: } \
113: }
114: #endif
115:
116: int
117: tcp_reass(tp, ti, m)
118: register struct tcpcb *tp;
119: register struct tcpiphdr *ti;
120: struct mbuf *m;
121: {
122: register struct tcpiphdr *q;
123: struct socket *so = tp->t_socket;
124: int flags;
125:
126: /*
127: * Call with ti==0 after become established to
128: * force pre-ESTABLISHED data up to user socket.
129: */
130: if (ti == 0)
131: goto present;
132:
133: /*
134: * Find a segment which begins after this one does.
135: */
136: for (q = (struct tcpiphdr *)tp->seg_next; q != (struct tcpiphdr *)tp;
137: q = (struct tcpiphdr *)q->ti_next)
138: if (SEQ_GT(q->ti_seq, ti->ti_seq))
139: break;
140:
141: /*
142: * If there is a preceding segment, it may provide some of
143: * our data already. If so, drop the data from the incoming
144: * segment. If it provides all of our data, drop us.
145: */
146: if ((struct tcpiphdr *)q->ti_prev != (struct tcpiphdr *)tp) {
147: register int i;
148: q = (struct tcpiphdr *)q->ti_prev;
149: /* conversion to int (in i) handles seq wraparound */
150: i = q->ti_seq + q->ti_len - ti->ti_seq;
151: if (i > 0) {
152: if (i >= ti->ti_len) {
153: tcpstat.tcps_rcvduppack++;
154: tcpstat.tcps_rcvdupbyte += ti->ti_len;
155: m_freem(m);
156: /*
157: * Try to present any queued data
158: * at the left window edge to the user.
159: * This is needed after the 3-WHS
160: * completes.
161: */
162: goto present; /* ??? */
163: }
164: m_adj(m, i);
165: ti->ti_len -= i;
166: ti->ti_seq += i;
167: }
168: q = (struct tcpiphdr *)(q->ti_next);
169: }
170: tcpstat.tcps_rcvoopack++;
171: tcpstat.tcps_rcvoobyte += ti->ti_len;
172: REASS_MBUF(ti) = (mbufp_32) m; /* XXX */
173:
174: /*
175: * While we overlap succeeding segments trim them or,
176: * if they are completely covered, dequeue them.
177: */
178: while (q != (struct tcpiphdr *)tp) {
179: register int i = (ti->ti_seq + ti->ti_len) - q->ti_seq;
180: if (i <= 0)
181: break;
182: if (i < q->ti_len) {
183: q->ti_seq += i;
184: q->ti_len -= i;
185: m_adj((struct mbuf *) REASS_MBUF(q), i);
186: break;
187: }
188: q = (struct tcpiphdr *)q->ti_next;
189: m = (struct mbuf *) REASS_MBUF((struct tcpiphdr *)q->ti_prev);
190: remque_32((void *)(q->ti_prev));
191: m_freem(m);
192: }
193:
194: /*
195: * Stick new segment in its place.
196: */
197: insque_32(ti, (void *)(q->ti_prev));
198:
199: present:
200: /*
201: * Present data to user, advancing rcv_nxt through
202: * completed sequence space.
203: */
204: if (!TCPS_HAVEESTABLISHED(tp->t_state))
205: return (0);
206: ti = (struct tcpiphdr *) tp->seg_next;
207: if (ti == (struct tcpiphdr *)tp || ti->ti_seq != tp->rcv_nxt)
208: return (0);
209: if (tp->t_state == TCPS_SYN_RECEIVED && ti->ti_len)
210: return (0);
211: do {
212: tp->rcv_nxt += ti->ti_len;
213: flags = ti->ti_flags & TH_FIN;
214: remque_32(ti);
215: m = (struct mbuf *) REASS_MBUF(ti); /* XXX */
216: ti = (struct tcpiphdr *)ti->ti_next;
217: /* if (so->so_state & SS_FCANTRCVMORE) */
218: if (so->so_state & SS_FCANTSENDMORE)
219: m_freem(m);
220: else {
221: if (so->so_emu) {
222: if (tcp_emu(so,m)) sbappend(so, m);
223: } else
224: sbappend(so, m);
225: }
226: } while (ti != (struct tcpiphdr *)tp && ti->ti_seq == tp->rcv_nxt);
227: /* sorwakeup(so); */
228: return (flags);
229: }
230:
231: /*
232: * TCP input routine, follows pages 65-76 of the
233: * protocol specification dated September, 1981 very closely.
234: */
235: void
236: tcp_input(m, iphlen, inso)
237: register struct mbuf *m;
238: int iphlen;
239: struct socket *inso;
240: {
241: struct ip save_ip, *ip;
242: register struct tcpiphdr *ti;
243: caddr_t optp = NULL;
244: int optlen = 0;
245: int len, tlen, off;
246: register struct tcpcb *tp = 0;
247: register int tiflags;
248: struct socket *so = 0;
249: int todrop, acked, ourfinisacked, needoutput = 0;
250: /* int dropsocket = 0; */
251: int iss = 0;
252: u_long tiwin;
253: int ret;
254: /* int ts_present = 0; */
255:
256: DEBUG_CALL("tcp_input");
257: DEBUG_ARGS((dfd," m = %8lx iphlen = %2d inso = %lx\n",
258: (long )m, iphlen, (long )inso ));
259:
260: /*
261: * If called with m == 0, then we're continuing the connect
262: */
263: if (m == NULL) {
264: so = inso;
265:
266: /* Re-set a few variables */
267: tp = sototcpcb(so);
268: m = so->so_m;
269: so->so_m = 0;
270: ti = so->so_ti;
271: tiwin = ti->ti_win;
272: tiflags = ti->ti_flags;
273:
274: goto cont_conn;
275: }
276:
277:
278: tcpstat.tcps_rcvtotal++;
279: /*
280: * Get IP and TCP header together in first mbuf.
281: * Note: IP leaves IP header in first mbuf.
282: */
283: ti = mtod(m, struct tcpiphdr *);
284: if (iphlen > sizeof(struct ip )) {
285: ip_stripoptions(m, (struct mbuf *)0);
286: iphlen=sizeof(struct ip );
287: }
288: /* XXX Check if too short */
289:
290:
291: /*
292: * Save a copy of the IP header in case we want restore it
293: * for sending an ICMP error message in response.
294: */
295: ip=mtod(m, struct ip *);
296: save_ip = *ip;
297: save_ip.ip_len+= iphlen;
298:
299: /*
300: * Checksum extended TCP header and data.
301: */
302: tlen = ((struct ip *)ti)->ip_len;
303: ti->ti_next = ti->ti_prev = 0;
304: ti->ti_x1 = 0;
305: ti->ti_len = htons((u_int16_t)tlen);
306: len = sizeof(struct ip ) + tlen;
307: /* keep checksum for ICMP reply
308: * ti->ti_sum = cksum(m, len);
309: * if (ti->ti_sum) { */
310: if(cksum(m, len)) {
311: tcpstat.tcps_rcvbadsum++;
312: goto drop;
313: }
314:
315: /*
316: * Check that TCP offset makes sense,
317: * pull out TCP options and adjust length. XXX
318: */
319: off = ti->ti_off << 2;
320: if (off < sizeof (struct tcphdr) || off > tlen) {
321: tcpstat.tcps_rcvbadoff++;
322: goto drop;
323: }
324: tlen -= off;
325: ti->ti_len = tlen;
326: if (off > sizeof (struct tcphdr)) {
327: optlen = off - sizeof (struct tcphdr);
328: optp = mtod(m, caddr_t) + sizeof (struct tcpiphdr);
329:
330: /*
331: * Do quick retrieval of timestamp options ("options
332: * prediction?"). If timestamp is the only option and it's
333: * formatted as recommended in RFC 1323 appendix A, we
334: * quickly get the values now and not bother calling
335: * tcp_dooptions(), etc.
336: */
337: /* if ((optlen == TCPOLEN_TSTAMP_APPA ||
338: * (optlen > TCPOLEN_TSTAMP_APPA &&
339: * optp[TCPOLEN_TSTAMP_APPA] == TCPOPT_EOL)) &&
340: * *(u_int32_t *)optp == htonl(TCPOPT_TSTAMP_HDR) &&
341: * (ti->ti_flags & TH_SYN) == 0) {
342: * ts_present = 1;
343: * ts_val = ntohl(*(u_int32_t *)(optp + 4));
344: * ts_ecr = ntohl(*(u_int32_t *)(optp + 8));
345: * optp = NULL; / * we've parsed the options * /
346: * }
347: */
348: }
349: tiflags = ti->ti_flags;
350:
351: /*
352: * Convert TCP protocol specific fields to host format.
353: */
354: NTOHL(ti->ti_seq);
355: NTOHL(ti->ti_ack);
356: NTOHS(ti->ti_win);
357: NTOHS(ti->ti_urp);
358:
359: /*
360: * Drop TCP, IP headers and TCP options.
361: */
362: m->m_data += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
363: m->m_len -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
364:
365: /*
366: * Locate pcb for segment.
367: */
368: findso:
369: so = tcp_last_so;
370: if (so->so_fport != ti->ti_dport ||
371: so->so_lport != ti->ti_sport ||
372: so->so_laddr.s_addr != ti->ti_src.s_addr ||
373: so->so_faddr.s_addr != ti->ti_dst.s_addr) {
374: so = solookup(&tcb, ti->ti_src, ti->ti_sport,
375: ti->ti_dst, ti->ti_dport);
376: if (so)
377: tcp_last_so = so;
378: ++tcpstat.tcps_socachemiss;
379: }
380:
381: /*
382: * If the state is CLOSED (i.e., TCB does not exist) then
383: * all data in the incoming segment is discarded.
384: * If the TCB exists but is in CLOSED state, it is embryonic,
385: * but should either do a listen or a connect soon.
386: *
387: * state == CLOSED means we've done socreate() but haven't
388: * attached it to a protocol yet...
389: *
390: * XXX If a TCB does not exist, and the TH_SYN flag is
391: * the only flag set, then create a session, mark it
392: * as if it was LISTENING, and continue...
393: */
394: if (so == 0) {
395: if ((tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) != TH_SYN)
396: goto dropwithreset;
397:
398: if ((so = socreate()) == NULL)
399: goto dropwithreset;
400: if (tcp_attach(so) < 0) {
401: free(so); /* Not sofree (if it failed, it's not insqued) */
402: goto dropwithreset;
403: }
404:
405: sbreserve(&so->so_snd, tcp_sndspace);
406: sbreserve(&so->so_rcv, tcp_rcvspace);
407:
408: /* tcp_last_so = so; */ /* XXX ? */
409: /* tp = sototcpcb(so); */
410:
411: so->so_laddr = ti->ti_src;
412: so->so_lport = ti->ti_sport;
413: so->so_faddr = ti->ti_dst;
414: so->so_fport = ti->ti_dport;
415:
416: if ((so->so_iptos = tcp_tos(so)) == 0)
417: so->so_iptos = ((struct ip *)ti)->ip_tos;
418:
419: tp = sototcpcb(so);
420: tp->t_state = TCPS_LISTEN;
421: }
422:
423: /*
424: * If this is a still-connecting socket, this probably
425: * a retransmit of the SYN. Whether it's a retransmit SYN
426: * or something else, we nuke it.
427: */
428: if (so->so_state & SS_ISFCONNECTING)
429: goto drop;
430:
431: tp = sototcpcb(so);
432:
433: /* XXX Should never fail */
434: if (tp == 0)
435: goto dropwithreset;
436: if (tp->t_state == TCPS_CLOSED)
437: goto drop;
438:
439: /* Unscale the window into a 32-bit value. */
440: /* if ((tiflags & TH_SYN) == 0)
441: * tiwin = ti->ti_win << tp->snd_scale;
442: * else
443: */
444: tiwin = ti->ti_win;
445:
446: /*
447: * Segment received on connection.
448: * Reset idle time and keep-alive timer.
449: */
450: tp->t_idle = 0;
451: if (so_options)
452: tp->t_timer[TCPT_KEEP] = tcp_keepintvl;
453: else
454: tp->t_timer[TCPT_KEEP] = tcp_keepidle;
455:
456: /*
457: * Process options if not in LISTEN state,
458: * else do it below (after getting remote address).
459: */
460: if (optp && tp->t_state != TCPS_LISTEN)
461: tcp_dooptions(tp, (u_char *)optp, optlen, ti);
462: /* , */
463: /* &ts_present, &ts_val, &ts_ecr); */
464:
465: /*
466: * Header prediction: check for the two common cases
467: * of a uni-directional data xfer. If the packet has
468: * no control flags, is in-sequence, the window didn't
469: * change and we're not retransmitting, it's a
470: * candidate. If the length is zero and the ack moved
471: * forward, we're the sender side of the xfer. Just
472: * free the data acked & wake any higher level process
473: * that was blocked waiting for space. If the length
474: * is non-zero and the ack didn't move, we're the
475: * receiver side. If we're getting packets in-order
476: * (the reassembly queue is empty), add the data to
477: * the socket buffer and note that we need a delayed ack.
478: *
479: * XXX Some of these tests are not needed
480: * eg: the tiwin == tp->snd_wnd prevents many more
481: * predictions.. with no *real* advantage..
482: */
483: if (tp->t_state == TCPS_ESTABLISHED &&
484: (tiflags & (TH_SYN|TH_FIN|TH_RST|TH_URG|TH_ACK)) == TH_ACK &&
485: /* (!ts_present || TSTMP_GEQ(ts_val, tp->ts_recent)) && */
486: ti->ti_seq == tp->rcv_nxt &&
487: tiwin && tiwin == tp->snd_wnd &&
488: tp->snd_nxt == tp->snd_max) {
489: /*
490: * If last ACK falls within this segment's sequence numbers,
491: * record the timestamp.
492: */
493: /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
494: * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len)) {
495: * tp->ts_recent_age = tcp_now;
496: * tp->ts_recent = ts_val;
497: * }
498: */
499: if (ti->ti_len == 0) {
500: if (SEQ_GT(ti->ti_ack, tp->snd_una) &&
501: SEQ_LEQ(ti->ti_ack, tp->snd_max) &&
502: tp->snd_cwnd >= tp->snd_wnd) {
503: /*
504: * this is a pure ack for outstanding data.
505: */
506: ++tcpstat.tcps_predack;
507: /* if (ts_present)
508: * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
509: * else
510: */ if (tp->t_rtt &&
511: SEQ_GT(ti->ti_ack, tp->t_rtseq))
512: tcp_xmit_timer(tp, tp->t_rtt);
513: acked = ti->ti_ack - tp->snd_una;
514: tcpstat.tcps_rcvackpack++;
515: tcpstat.tcps_rcvackbyte += acked;
516: sbdrop(&so->so_snd, acked);
517: tp->snd_una = ti->ti_ack;
518: m_freem(m);
519:
520: /*
521: * If all outstanding data are acked, stop
522: * retransmit timer, otherwise restart timer
523: * using current (possibly backed-off) value.
524: * If process is waiting for space,
525: * wakeup/selwakeup/signal. If data
526: * are ready to send, let tcp_output
527: * decide between more output or persist.
528: */
529: if (tp->snd_una == tp->snd_max)
530: tp->t_timer[TCPT_REXMT] = 0;
531: else if (tp->t_timer[TCPT_PERSIST] == 0)
532: tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
533:
534: /*
535: * There's room in so_snd, sowwakup will read()
536: * from the socket if we can
537: */
538: /* if (so->so_snd.sb_flags & SB_NOTIFY)
539: * sowwakeup(so);
540: */
541: /*
542: * This is called because sowwakeup might have
543: * put data into so_snd. Since we don't so sowwakeup,
544: * we don't need this.. XXX???
545: */
546: if (so->so_snd.sb_cc)
547: (void) tcp_output(tp);
548:
549: return;
550: }
551: } else if (ti->ti_ack == tp->snd_una &&
552: tp->seg_next == (tcpiphdrp_32)tp &&
553: ti->ti_len <= sbspace(&so->so_rcv)) {
554: /*
555: * this is a pure, in-sequence data packet
556: * with nothing on the reassembly queue and
557: * we have enough buffer space to take it.
558: */
559: ++tcpstat.tcps_preddat;
560: tp->rcv_nxt += ti->ti_len;
561: tcpstat.tcps_rcvpack++;
562: tcpstat.tcps_rcvbyte += ti->ti_len;
563: /*
564: * Add data to socket buffer.
565: */
566: if (so->so_emu) {
567: if (tcp_emu(so,m)) sbappend(so, m);
568: } else
569: sbappend(so, m);
570:
571: /*
572: * XXX This is called when data arrives. Later, check
573: * if we can actually write() to the socket
574: * XXX Need to check? It's be NON_BLOCKING
575: */
576: /* sorwakeup(so); */
577:
578: /*
579: * If this is a short packet, then ACK now - with Nagel
580: * congestion avoidance sender won't send more until
581: * he gets an ACK.
582: *
583: * Here are 3 interpretations of what should happen.
584: * The best (for me) is to delay-ack everything except
585: * if it's a one-byte packet containing an ESC
586: * (this means it's an arrow key (or similar) sent using
587: * Nagel, hence there will be no echo)
588: * The first of these is the original, the second is the
589: * middle ground between the other 2
590: */
591: /* if (((unsigned)ti->ti_len < tp->t_maxseg)) {
592: */
593: /* if (((unsigned)ti->ti_len < tp->t_maxseg &&
594: * (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
595: * ((so->so_iptos & IPTOS_LOWDELAY) &&
596: * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
597: */
598: if ((unsigned)ti->ti_len == 1 &&
599: ((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
600: tp->t_flags |= TF_ACKNOW;
601: tcp_output(tp);
602: } else {
603: tp->t_flags |= TF_DELACK;
604: }
605: return;
606: }
607: } /* header prediction */
608: /*
609: * Calculate amount of space in receive window,
610: * and then do TCP input processing.
611: * Receive window is amount of space in rcv queue,
612: * but not less than advertised window.
613: */
614: { int win;
615: win = sbspace(&so->so_rcv);
616: if (win < 0)
617: win = 0;
618: tp->rcv_wnd = max(win, (int)(tp->rcv_adv - tp->rcv_nxt));
619: }
620:
621: switch (tp->t_state) {
622:
623: /*
624: * If the state is LISTEN then ignore segment if it contains an RST.
625: * If the segment contains an ACK then it is bad and send a RST.
626: * If it does not contain a SYN then it is not interesting; drop it.
627: * Don't bother responding if the destination was a broadcast.
628: * Otherwise initialize tp->rcv_nxt, and tp->irs, select an initial
629: * tp->iss, and send a segment:
630: * <SEQ=ISS><ACK=RCV_NXT><CTL=SYN,ACK>
631: * Also initialize tp->snd_nxt to tp->iss+1 and tp->snd_una to tp->iss.
632: * Fill in remote peer address fields if not previously specified.
633: * Enter SYN_RECEIVED state, and process any other fields of this
634: * segment in this state.
635: */
636: case TCPS_LISTEN: {
637:
638: if (tiflags & TH_RST)
639: goto drop;
640: if (tiflags & TH_ACK)
641: goto dropwithreset;
642: if ((tiflags & TH_SYN) == 0)
643: goto drop;
644:
645: /*
646: * This has way too many gotos...
647: * But a bit of spaghetti code never hurt anybody :)
648: */
649:
650: /*
651: * If this is destined for the control address, then flag to
652: * tcp_ctl once connected, otherwise connect
653: */
654: if ((so->so_faddr.s_addr&htonl(0xffffff00)) == special_addr.s_addr) {
655: int lastbyte=ntohl(so->so_faddr.s_addr) & 0xff;
656: if (lastbyte!=CTL_ALIAS && lastbyte!=CTL_DNS) {
657: #if 0
658: if(lastbyte==CTL_CMD || lastbyte==CTL_EXEC) {
659: /* Command or exec adress */
660: so->so_state |= SS_CTL;
661: } else
662: #endif
663: {
664: /* May be an add exec */
665: struct ex_list *ex_ptr;
666: for(ex_ptr = exec_list; ex_ptr; ex_ptr = ex_ptr->ex_next) {
667: if(ex_ptr->ex_fport == so->so_fport &&
668: lastbyte == ex_ptr->ex_addr) {
669: so->so_state |= SS_CTL;
670: break;
671: }
672: }
673: }
674: if(so->so_state & SS_CTL) goto cont_input;
675: }
676: /* CTL_ALIAS: Do nothing, tcp_fconnect will be called on it */
677: }
678:
679: if (so->so_emu & EMU_NOCONNECT) {
680: so->so_emu &= ~EMU_NOCONNECT;
681: goto cont_input;
682: }
683:
684: if((tcp_fconnect(so) == -1) && (errno != EINPROGRESS) && (errno != EWOULDBLOCK)) {
685: u_char code=ICMP_UNREACH_NET;
686: DEBUG_MISC((dfd," tcp fconnect errno = %d-%s\n",
687: errno,strerror(errno)));
688: if(errno == ECONNREFUSED) {
689: /* ACK the SYN, send RST to refuse the connection */
690: tcp_respond(tp, ti, m, ti->ti_seq+1, (tcp_seq)0,
691: TH_RST|TH_ACK);
692: } else {
693: if(errno == EHOSTUNREACH) code=ICMP_UNREACH_HOST;
694: HTONL(ti->ti_seq); /* restore tcp header */
695: HTONL(ti->ti_ack);
696: HTONS(ti->ti_win);
697: HTONS(ti->ti_urp);
698: m->m_data -= sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
699: m->m_len += sizeof(struct tcpiphdr)+off-sizeof(struct tcphdr);
700: *ip=save_ip;
701: icmp_error(m, ICMP_UNREACH,code, 0,strerror(errno));
702: }
703: tp = tcp_close(tp);
704: m_free(m);
705: } else {
706: /*
707: * Haven't connected yet, save the current mbuf
708: * and ti, and return
709: * XXX Some OS's don't tell us whether the connect()
710: * succeeded or not. So we must time it out.
711: */
712: so->so_m = m;
713: so->so_ti = ti;
714: tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
715: tp->t_state = TCPS_SYN_RECEIVED;
716: }
717: return;
718:
719: cont_conn:
720: /* m==NULL
721: * Check if the connect succeeded
722: */
723: if (so->so_state & SS_NOFDREF) {
724: tp = tcp_close(tp);
725: goto dropwithreset;
726: }
727: cont_input:
728: tcp_template(tp);
729:
730: if (optp)
731: tcp_dooptions(tp, (u_char *)optp, optlen, ti);
732: /* , */
733: /* &ts_present, &ts_val, &ts_ecr); */
734:
735: if (iss)
736: tp->iss = iss;
737: else
738: tp->iss = tcp_iss;
739: tcp_iss += TCP_ISSINCR/2;
740: tp->irs = ti->ti_seq;
741: tcp_sendseqinit(tp);
742: tcp_rcvseqinit(tp);
743: tp->t_flags |= TF_ACKNOW;
744: tp->t_state = TCPS_SYN_RECEIVED;
745: tp->t_timer[TCPT_KEEP] = TCPTV_KEEP_INIT;
746: tcpstat.tcps_accepts++;
747: goto trimthenstep6;
748: } /* case TCPS_LISTEN */
749:
750: /*
751: * If the state is SYN_SENT:
752: * if seg contains an ACK, but not for our SYN, drop the input.
753: * if seg contains a RST, then drop the connection.
754: * if seg does not contain SYN, then drop it.
755: * Otherwise this is an acceptable SYN segment
756: * initialize tp->rcv_nxt and tp->irs
757: * if seg contains ack then advance tp->snd_una
758: * if SYN has been acked change to ESTABLISHED else SYN_RCVD state
759: * arrange for segment to be acked (eventually)
760: * continue processing rest of data/controls, beginning with URG
761: */
762: case TCPS_SYN_SENT:
763: if ((tiflags & TH_ACK) &&
764: (SEQ_LEQ(ti->ti_ack, tp->iss) ||
765: SEQ_GT(ti->ti_ack, tp->snd_max)))
766: goto dropwithreset;
767:
768: if (tiflags & TH_RST) {
769: if (tiflags & TH_ACK)
770: tp = tcp_drop(tp,0); /* XXX Check t_softerror! */
771: goto drop;
772: }
773:
774: if ((tiflags & TH_SYN) == 0)
775: goto drop;
776: if (tiflags & TH_ACK) {
777: tp->snd_una = ti->ti_ack;
778: if (SEQ_LT(tp->snd_nxt, tp->snd_una))
779: tp->snd_nxt = tp->snd_una;
780: }
781:
782: tp->t_timer[TCPT_REXMT] = 0;
783: tp->irs = ti->ti_seq;
784: tcp_rcvseqinit(tp);
785: tp->t_flags |= TF_ACKNOW;
786: if (tiflags & TH_ACK && SEQ_GT(tp->snd_una, tp->iss)) {
787: tcpstat.tcps_connects++;
788: soisfconnected(so);
789: tp->t_state = TCPS_ESTABLISHED;
790:
791: /* Do window scaling on this connection? */
792: /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
793: * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
794: * tp->snd_scale = tp->requested_s_scale;
795: * tp->rcv_scale = tp->request_r_scale;
796: * }
797: */
798: (void) tcp_reass(tp, (struct tcpiphdr *)0,
799: (struct mbuf *)0);
800: /*
801: * if we didn't have to retransmit the SYN,
802: * use its rtt as our initial srtt & rtt var.
803: */
804: if (tp->t_rtt)
805: tcp_xmit_timer(tp, tp->t_rtt);
806: } else
807: tp->t_state = TCPS_SYN_RECEIVED;
808:
809: trimthenstep6:
810: /*
811: * Advance ti->ti_seq to correspond to first data byte.
812: * If data, trim to stay within window,
813: * dropping FIN if necessary.
814: */
815: ti->ti_seq++;
816: if (ti->ti_len > tp->rcv_wnd) {
817: todrop = ti->ti_len - tp->rcv_wnd;
818: m_adj(m, -todrop);
819: ti->ti_len = tp->rcv_wnd;
820: tiflags &= ~TH_FIN;
821: tcpstat.tcps_rcvpackafterwin++;
822: tcpstat.tcps_rcvbyteafterwin += todrop;
823: }
824: tp->snd_wl1 = ti->ti_seq - 1;
825: tp->rcv_up = ti->ti_seq;
826: goto step6;
827: } /* switch tp->t_state */
828: /*
829: * States other than LISTEN or SYN_SENT.
830: * First check timestamp, if present.
831: * Then check that at least some bytes of segment are within
832: * receive window. If segment begins before rcv_nxt,
833: * drop leading data (and SYN); if nothing left, just ack.
834: *
835: * RFC 1323 PAWS: If we have a timestamp reply on this segment
836: * and it's less than ts_recent, drop it.
837: */
838: /* if (ts_present && (tiflags & TH_RST) == 0 && tp->ts_recent &&
839: * TSTMP_LT(ts_val, tp->ts_recent)) {
840: *
841: */ /* Check to see if ts_recent is over 24 days old. */
842: /* if ((int)(tcp_now - tp->ts_recent_age) > TCP_PAWS_IDLE) {
843: */ /*
844: * * Invalidate ts_recent. If this segment updates
845: * * ts_recent, the age will be reset later and ts_recent
846: * * will get a valid value. If it does not, setting
847: * * ts_recent to zero will at least satisfy the
848: * * requirement that zero be placed in the timestamp
849: * * echo reply when ts_recent isn't valid. The
850: * * age isn't reset until we get a valid ts_recent
851: * * because we don't want out-of-order segments to be
852: * * dropped when ts_recent is old.
853: * */
854: /* tp->ts_recent = 0;
855: * } else {
856: * tcpstat.tcps_rcvduppack++;
857: * tcpstat.tcps_rcvdupbyte += ti->ti_len;
858: * tcpstat.tcps_pawsdrop++;
859: * goto dropafterack;
860: * }
861: * }
862: */
863:
864: todrop = tp->rcv_nxt - ti->ti_seq;
865: if (todrop > 0) {
866: if (tiflags & TH_SYN) {
867: tiflags &= ~TH_SYN;
868: ti->ti_seq++;
869: if (ti->ti_urp > 1)
870: ti->ti_urp--;
871: else
872: tiflags &= ~TH_URG;
873: todrop--;
874: }
875: /*
876: * Following if statement from Stevens, vol. 2, p. 960.
877: */
878: if (todrop > ti->ti_len
879: || (todrop == ti->ti_len && (tiflags & TH_FIN) == 0)) {
880: /*
881: * Any valid FIN must be to the left of the window.
882: * At this point the FIN must be a duplicate or out
883: * of sequence; drop it.
884: */
885: tiflags &= ~TH_FIN;
886:
887: /*
888: * Send an ACK to resynchronize and drop any data.
889: * But keep on processing for RST or ACK.
890: */
891: tp->t_flags |= TF_ACKNOW;
892: todrop = ti->ti_len;
893: tcpstat.tcps_rcvduppack++;
894: tcpstat.tcps_rcvdupbyte += todrop;
895: } else {
896: tcpstat.tcps_rcvpartduppack++;
897: tcpstat.tcps_rcvpartdupbyte += todrop;
898: }
899: m_adj(m, todrop);
900: ti->ti_seq += todrop;
901: ti->ti_len -= todrop;
902: if (ti->ti_urp > todrop)
903: ti->ti_urp -= todrop;
904: else {
905: tiflags &= ~TH_URG;
906: ti->ti_urp = 0;
907: }
908: }
909: /*
910: * If new data are received on a connection after the
911: * user processes are gone, then RST the other end.
912: */
913: if ((so->so_state & SS_NOFDREF) &&
914: tp->t_state > TCPS_CLOSE_WAIT && ti->ti_len) {
915: tp = tcp_close(tp);
916: tcpstat.tcps_rcvafterclose++;
917: goto dropwithreset;
918: }
919:
920: /*
921: * If segment ends after window, drop trailing data
922: * (and PUSH and FIN); if nothing left, just ACK.
923: */
924: todrop = (ti->ti_seq+ti->ti_len) - (tp->rcv_nxt+tp->rcv_wnd);
925: if (todrop > 0) {
926: tcpstat.tcps_rcvpackafterwin++;
927: if (todrop >= ti->ti_len) {
928: tcpstat.tcps_rcvbyteafterwin += ti->ti_len;
929: /*
930: * If a new connection request is received
931: * while in TIME_WAIT, drop the old connection
932: * and start over if the sequence numbers
933: * are above the previous ones.
934: */
935: if (tiflags & TH_SYN &&
936: tp->t_state == TCPS_TIME_WAIT &&
937: SEQ_GT(ti->ti_seq, tp->rcv_nxt)) {
938: iss = tp->rcv_nxt + TCP_ISSINCR;
939: tp = tcp_close(tp);
940: goto findso;
941: }
942: /*
943: * If window is closed can only take segments at
944: * window edge, and have to drop data and PUSH from
945: * incoming segments. Continue processing, but
946: * remember to ack. Otherwise, drop segment
947: * and ack.
948: */
949: if (tp->rcv_wnd == 0 && ti->ti_seq == tp->rcv_nxt) {
950: tp->t_flags |= TF_ACKNOW;
951: tcpstat.tcps_rcvwinprobe++;
952: } else
953: goto dropafterack;
954: } else
955: tcpstat.tcps_rcvbyteafterwin += todrop;
956: m_adj(m, -todrop);
957: ti->ti_len -= todrop;
958: tiflags &= ~(TH_PUSH|TH_FIN);
959: }
960:
961: /*
962: * If last ACK falls within this segment's sequence numbers,
963: * record its timestamp.
964: */
965: /* if (ts_present && SEQ_LEQ(ti->ti_seq, tp->last_ack_sent) &&
966: * SEQ_LT(tp->last_ack_sent, ti->ti_seq + ti->ti_len +
967: * ((tiflags & (TH_SYN|TH_FIN)) != 0))) {
968: * tp->ts_recent_age = tcp_now;
969: * tp->ts_recent = ts_val;
970: * }
971: */
972:
973: /*
974: * If the RST bit is set examine the state:
975: * SYN_RECEIVED STATE:
976: * If passive open, return to LISTEN state.
977: * If active open, inform user that connection was refused.
978: * ESTABLISHED, FIN_WAIT_1, FIN_WAIT2, CLOSE_WAIT STATES:
979: * Inform user that connection was reset, and close tcb.
980: * CLOSING, LAST_ACK, TIME_WAIT STATES
981: * Close the tcb.
982: */
983: if (tiflags&TH_RST) switch (tp->t_state) {
984:
985: case TCPS_SYN_RECEIVED:
986: /* so->so_error = ECONNREFUSED; */
987: goto close;
988:
989: case TCPS_ESTABLISHED:
990: case TCPS_FIN_WAIT_1:
991: case TCPS_FIN_WAIT_2:
992: case TCPS_CLOSE_WAIT:
993: /* so->so_error = ECONNRESET; */
994: close:
995: tp->t_state = TCPS_CLOSED;
996: tcpstat.tcps_drops++;
997: tp = tcp_close(tp);
998: goto drop;
999:
1000: case TCPS_CLOSING:
1001: case TCPS_LAST_ACK:
1002: case TCPS_TIME_WAIT:
1003: tp = tcp_close(tp);
1004: goto drop;
1005: }
1006:
1007: /*
1008: * If a SYN is in the window, then this is an
1009: * error and we send an RST and drop the connection.
1010: */
1011: if (tiflags & TH_SYN) {
1012: tp = tcp_drop(tp,0);
1013: goto dropwithreset;
1014: }
1015:
1016: /*
1017: * If the ACK bit is off we drop the segment and return.
1018: */
1019: if ((tiflags & TH_ACK) == 0) goto drop;
1020:
1021: /*
1022: * Ack processing.
1023: */
1024: switch (tp->t_state) {
1025: /*
1026: * In SYN_RECEIVED state if the ack ACKs our SYN then enter
1027: * ESTABLISHED state and continue processing, otherwise
1028: * send an RST. una<=ack<=max
1029: */
1030: case TCPS_SYN_RECEIVED:
1031:
1032: if (SEQ_GT(tp->snd_una, ti->ti_ack) ||
1033: SEQ_GT(ti->ti_ack, tp->snd_max))
1034: goto dropwithreset;
1035: tcpstat.tcps_connects++;
1036: tp->t_state = TCPS_ESTABLISHED;
1037: /*
1038: * The sent SYN is ack'ed with our sequence number +1
1039: * The first data byte already in the buffer will get
1040: * lost if no correction is made. This is only needed for
1041: * SS_CTL since the buffer is empty otherwise.
1042: * tp->snd_una++; or:
1043: */
1044: tp->snd_una=ti->ti_ack;
1045: if (so->so_state & SS_CTL) {
1046: /* So tcp_ctl reports the right state */
1047: ret = tcp_ctl(so);
1048: if (ret == 1) {
1049: soisfconnected(so);
1050: so->so_state &= ~SS_CTL; /* success XXX */
1051: } else if (ret == 2) {
1052: so->so_state = SS_NOFDREF; /* CTL_CMD */
1053: } else {
1054: needoutput = 1;
1055: tp->t_state = TCPS_FIN_WAIT_1;
1056: }
1057: } else {
1058: soisfconnected(so);
1059: }
1060:
1061: /* Do window scaling? */
1062: /* if ((tp->t_flags & (TF_RCVD_SCALE|TF_REQ_SCALE)) ==
1063: * (TF_RCVD_SCALE|TF_REQ_SCALE)) {
1064: * tp->snd_scale = tp->requested_s_scale;
1065: * tp->rcv_scale = tp->request_r_scale;
1066: * }
1067: */
1068: (void) tcp_reass(tp, (struct tcpiphdr *)0, (struct mbuf *)0);
1069: tp->snd_wl1 = ti->ti_seq - 1;
1070: /* Avoid ack processing; snd_una==ti_ack => dup ack */
1071: goto synrx_to_est;
1072: /* fall into ... */
1073:
1074: /*
1075: * In ESTABLISHED state: drop duplicate ACKs; ACK out of range
1076: * ACKs. If the ack is in the range
1077: * tp->snd_una < ti->ti_ack <= tp->snd_max
1078: * then advance tp->snd_una to ti->ti_ack and drop
1079: * data from the retransmission queue. If this ACK reflects
1080: * more up to date window information we update our window information.
1081: */
1082: case TCPS_ESTABLISHED:
1083: case TCPS_FIN_WAIT_1:
1084: case TCPS_FIN_WAIT_2:
1085: case TCPS_CLOSE_WAIT:
1086: case TCPS_CLOSING:
1087: case TCPS_LAST_ACK:
1088: case TCPS_TIME_WAIT:
1089:
1090: if (SEQ_LEQ(ti->ti_ack, tp->snd_una)) {
1091: if (ti->ti_len == 0 && tiwin == tp->snd_wnd) {
1092: tcpstat.tcps_rcvdupack++;
1093: DEBUG_MISC((dfd," dup ack m = %lx so = %lx \n",
1094: (long )m, (long )so));
1095: /*
1096: * If we have outstanding data (other than
1097: * a window probe), this is a completely
1098: * duplicate ack (ie, window info didn't
1099: * change), the ack is the biggest we've
1100: * seen and we've seen exactly our rexmt
1101: * threshold of them, assume a packet
1102: * has been dropped and retransmit it.
1103: * Kludge snd_nxt & the congestion
1104: * window so we send only this one
1105: * packet.
1106: *
1107: * We know we're losing at the current
1108: * window size so do congestion avoidance
1109: * (set ssthresh to half the current window
1110: * and pull our congestion window back to
1111: * the new ssthresh).
1112: *
1113: * Dup acks mean that packets have left the
1114: * network (they're now cached at the receiver)
1115: * so bump cwnd by the amount in the receiver
1116: * to keep a constant cwnd packets in the
1117: * network.
1118: */
1119: if (tp->t_timer[TCPT_REXMT] == 0 ||
1120: ti->ti_ack != tp->snd_una)
1121: tp->t_dupacks = 0;
1122: else if (++tp->t_dupacks == tcprexmtthresh) {
1123: tcp_seq onxt = tp->snd_nxt;
1124: u_int win =
1125: min(tp->snd_wnd, tp->snd_cwnd) / 2 /
1126: tp->t_maxseg;
1127:
1128: if (win < 2)
1129: win = 2;
1130: tp->snd_ssthresh = win * tp->t_maxseg;
1131: tp->t_timer[TCPT_REXMT] = 0;
1132: tp->t_rtt = 0;
1133: tp->snd_nxt = ti->ti_ack;
1134: tp->snd_cwnd = tp->t_maxseg;
1135: (void) tcp_output(tp);
1136: tp->snd_cwnd = tp->snd_ssthresh +
1137: tp->t_maxseg * tp->t_dupacks;
1138: if (SEQ_GT(onxt, tp->snd_nxt))
1139: tp->snd_nxt = onxt;
1140: goto drop;
1141: } else if (tp->t_dupacks > tcprexmtthresh) {
1142: tp->snd_cwnd += tp->t_maxseg;
1143: (void) tcp_output(tp);
1144: goto drop;
1145: }
1146: } else
1147: tp->t_dupacks = 0;
1148: break;
1149: }
1150: synrx_to_est:
1151: /*
1152: * If the congestion window was inflated to account
1153: * for the other side's cached packets, retract it.
1154: */
1155: if (tp->t_dupacks > tcprexmtthresh &&
1156: tp->snd_cwnd > tp->snd_ssthresh)
1157: tp->snd_cwnd = tp->snd_ssthresh;
1158: tp->t_dupacks = 0;
1159: if (SEQ_GT(ti->ti_ack, tp->snd_max)) {
1160: tcpstat.tcps_rcvacktoomuch++;
1161: goto dropafterack;
1162: }
1163: acked = ti->ti_ack - tp->snd_una;
1164: tcpstat.tcps_rcvackpack++;
1165: tcpstat.tcps_rcvackbyte += acked;
1166:
1167: /*
1168: * If we have a timestamp reply, update smoothed
1169: * round trip time. If no timestamp is present but
1170: * transmit timer is running and timed sequence
1171: * number was acked, update smoothed round trip time.
1172: * Since we now have an rtt measurement, cancel the
1173: * timer backoff (cf., Phil Karn's retransmit alg.).
1174: * Recompute the initial retransmit timer.
1175: */
1176: /* if (ts_present)
1177: * tcp_xmit_timer(tp, tcp_now-ts_ecr+1);
1178: * else
1179: */
1180: if (tp->t_rtt && SEQ_GT(ti->ti_ack, tp->t_rtseq))
1181: tcp_xmit_timer(tp,tp->t_rtt);
1182:
1183: /*
1184: * If all outstanding data is acked, stop retransmit
1185: * timer and remember to restart (more output or persist).
1186: * If there is more data to be acked, restart retransmit
1187: * timer, using current (possibly backed-off) value.
1188: */
1189: if (ti->ti_ack == tp->snd_max) {
1190: tp->t_timer[TCPT_REXMT] = 0;
1191: needoutput = 1;
1192: } else if (tp->t_timer[TCPT_PERSIST] == 0)
1193: tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
1194: /*
1195: * When new data is acked, open the congestion window.
1196: * If the window gives us less than ssthresh packets
1197: * in flight, open exponentially (maxseg per packet).
1198: * Otherwise open linearly: maxseg per window
1199: * (maxseg^2 / cwnd per packet).
1200: */
1201: {
1202: register u_int cw = tp->snd_cwnd;
1203: register u_int incr = tp->t_maxseg;
1204:
1205: if (cw > tp->snd_ssthresh)
1206: incr = incr * incr / cw;
1207: tp->snd_cwnd = min(cw + incr, TCP_MAXWIN<<tp->snd_scale);
1208: }
1209: if (acked > so->so_snd.sb_cc) {
1210: tp->snd_wnd -= so->so_snd.sb_cc;
1211: sbdrop(&so->so_snd, (int )so->so_snd.sb_cc);
1212: ourfinisacked = 1;
1213: } else {
1214: sbdrop(&so->so_snd, acked);
1215: tp->snd_wnd -= acked;
1216: ourfinisacked = 0;
1217: }
1218: /*
1219: * XXX sowwakup is called when data is acked and there's room for
1220: * for more data... it should read() the socket
1221: */
1222: /* if (so->so_snd.sb_flags & SB_NOTIFY)
1223: * sowwakeup(so);
1224: */
1225: tp->snd_una = ti->ti_ack;
1226: if (SEQ_LT(tp->snd_nxt, tp->snd_una))
1227: tp->snd_nxt = tp->snd_una;
1228:
1229: switch (tp->t_state) {
1230:
1231: /*
1232: * In FIN_WAIT_1 STATE in addition to the processing
1233: * for the ESTABLISHED state if our FIN is now acknowledged
1234: * then enter FIN_WAIT_2.
1235: */
1236: case TCPS_FIN_WAIT_1:
1237: if (ourfinisacked) {
1238: /*
1239: * If we can't receive any more
1240: * data, then closing user can proceed.
1241: * Starting the timer is contrary to the
1242: * specification, but if we don't get a FIN
1243: * we'll hang forever.
1244: */
1245: if (so->so_state & SS_FCANTRCVMORE) {
1246: soisfdisconnected(so);
1247: tp->t_timer[TCPT_2MSL] = tcp_maxidle;
1248: }
1249: tp->t_state = TCPS_FIN_WAIT_2;
1250: }
1251: break;
1252:
1253: /*
1254: * In CLOSING STATE in addition to the processing for
1255: * the ESTABLISHED state if the ACK acknowledges our FIN
1256: * then enter the TIME-WAIT state, otherwise ignore
1257: * the segment.
1258: */
1259: case TCPS_CLOSING:
1260: if (ourfinisacked) {
1261: tp->t_state = TCPS_TIME_WAIT;
1262: tcp_canceltimers(tp);
1263: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1264: soisfdisconnected(so);
1265: }
1266: break;
1267:
1268: /*
1269: * In LAST_ACK, we may still be waiting for data to drain
1270: * and/or to be acked, as well as for the ack of our FIN.
1271: * If our FIN is now acknowledged, delete the TCB,
1272: * enter the closed state and return.
1273: */
1274: case TCPS_LAST_ACK:
1275: if (ourfinisacked) {
1276: tp = tcp_close(tp);
1277: goto drop;
1278: }
1279: break;
1280:
1281: /*
1282: * In TIME_WAIT state the only thing that should arrive
1283: * is a retransmission of the remote FIN. Acknowledge
1284: * it and restart the finack timer.
1285: */
1286: case TCPS_TIME_WAIT:
1287: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1288: goto dropafterack;
1289: }
1290: } /* switch(tp->t_state) */
1291:
1292: step6:
1293: /*
1294: * Update window information.
1295: * Don't look at window if no ACK: TAC's send garbage on first SYN.
1296: */
1297: if ((tiflags & TH_ACK) &&
1298: (SEQ_LT(tp->snd_wl1, ti->ti_seq) ||
1299: (tp->snd_wl1 == ti->ti_seq && (SEQ_LT(tp->snd_wl2, ti->ti_ack) ||
1300: (tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd))))) {
1301: /* keep track of pure window updates */
1302: if (ti->ti_len == 0 &&
1303: tp->snd_wl2 == ti->ti_ack && tiwin > tp->snd_wnd)
1304: tcpstat.tcps_rcvwinupd++;
1305: tp->snd_wnd = tiwin;
1306: tp->snd_wl1 = ti->ti_seq;
1307: tp->snd_wl2 = ti->ti_ack;
1308: if (tp->snd_wnd > tp->max_sndwnd)
1309: tp->max_sndwnd = tp->snd_wnd;
1310: needoutput = 1;
1311: }
1312:
1313: /*
1314: * Process segments with URG.
1315: */
1316: if ((tiflags & TH_URG) && ti->ti_urp &&
1317: TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1318: /*
1319: * This is a kludge, but if we receive and accept
1320: * random urgent pointers, we'll crash in
1321: * soreceive. It's hard to imagine someone
1322: * actually wanting to send this much urgent data.
1323: */
1324: if (ti->ti_urp + so->so_rcv.sb_cc > so->so_rcv.sb_datalen) {
1325: ti->ti_urp = 0;
1326: tiflags &= ~TH_URG;
1327: goto dodata;
1328: }
1329: /*
1330: * If this segment advances the known urgent pointer,
1331: * then mark the data stream. This should not happen
1332: * in CLOSE_WAIT, CLOSING, LAST_ACK or TIME_WAIT STATES since
1333: * a FIN has been received from the remote side.
1334: * In these states we ignore the URG.
1335: *
1336: * According to RFC961 (Assigned Protocols),
1337: * the urgent pointer points to the last octet
1338: * of urgent data. We continue, however,
1339: * to consider it to indicate the first octet
1340: * of data past the urgent section as the original
1341: * spec states (in one of two places).
1342: */
1343: if (SEQ_GT(ti->ti_seq+ti->ti_urp, tp->rcv_up)) {
1344: tp->rcv_up = ti->ti_seq + ti->ti_urp;
1345: so->so_urgc = so->so_rcv.sb_cc +
1346: (tp->rcv_up - tp->rcv_nxt); /* -1; */
1347: tp->rcv_up = ti->ti_seq + ti->ti_urp;
1348:
1349: }
1350: } else
1351: /*
1352: * If no out of band data is expected,
1353: * pull receive urgent pointer along
1354: * with the receive window.
1355: */
1356: if (SEQ_GT(tp->rcv_nxt, tp->rcv_up))
1357: tp->rcv_up = tp->rcv_nxt;
1358: dodata:
1359:
1360: /*
1361: * Process the segment text, merging it into the TCP sequencing queue,
1362: * and arranging for acknowledgment of receipt if necessary.
1363: * This process logically involves adjusting tp->rcv_wnd as data
1364: * is presented to the user (this happens in tcp_usrreq.c,
1365: * case PRU_RCVD). If a FIN has already been received on this
1366: * connection then we just ignore the text.
1367: */
1368: if ((ti->ti_len || (tiflags&TH_FIN)) &&
1369: TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1370: TCP_REASS(tp, ti, m, so, tiflags);
1371: /*
1372: * Note the amount of data that peer has sent into
1373: * our window, in order to estimate the sender's
1374: * buffer size.
1375: */
1376: len = so->so_rcv.sb_datalen - (tp->rcv_adv - tp->rcv_nxt);
1377: } else {
1378: m_free(m);
1379: tiflags &= ~TH_FIN;
1380: }
1381:
1382: /*
1383: * If FIN is received ACK the FIN and let the user know
1384: * that the connection is closing.
1385: */
1386: if (tiflags & TH_FIN) {
1387: if (TCPS_HAVERCVDFIN(tp->t_state) == 0) {
1388: /*
1389: * If we receive a FIN we can't send more data,
1390: * set it SS_FDRAIN
1391: * Shutdown the socket if there is no rx data in the
1392: * buffer.
1393: * soread() is called on completion of shutdown() and
1394: * will got to TCPS_LAST_ACK, and use tcp_output()
1395: * to send the FIN.
1396: */
1397: /* sofcantrcvmore(so); */
1398: sofwdrain(so);
1399:
1400: tp->t_flags |= TF_ACKNOW;
1401: tp->rcv_nxt++;
1402: }
1403: switch (tp->t_state) {
1404:
1405: /*
1406: * In SYN_RECEIVED and ESTABLISHED STATES
1407: * enter the CLOSE_WAIT state.
1408: */
1409: case TCPS_SYN_RECEIVED:
1410: case TCPS_ESTABLISHED:
1411: if(so->so_emu == EMU_CTL) /* no shutdown on socket */
1412: tp->t_state = TCPS_LAST_ACK;
1413: else
1414: tp->t_state = TCPS_CLOSE_WAIT;
1415: break;
1416:
1417: /*
1418: * If still in FIN_WAIT_1 STATE FIN has not been acked so
1419: * enter the CLOSING state.
1420: */
1421: case TCPS_FIN_WAIT_1:
1422: tp->t_state = TCPS_CLOSING;
1423: break;
1424:
1425: /*
1426: * In FIN_WAIT_2 state enter the TIME_WAIT state,
1427: * starting the time-wait timer, turning off the other
1428: * standard timers.
1429: */
1430: case TCPS_FIN_WAIT_2:
1431: tp->t_state = TCPS_TIME_WAIT;
1432: tcp_canceltimers(tp);
1433: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1434: soisfdisconnected(so);
1435: break;
1436:
1437: /*
1438: * In TIME_WAIT state restart the 2 MSL time_wait timer.
1439: */
1440: case TCPS_TIME_WAIT:
1441: tp->t_timer[TCPT_2MSL] = 2 * TCPTV_MSL;
1442: break;
1443: }
1444: }
1445:
1446: /*
1447: * If this is a small packet, then ACK now - with Nagel
1448: * congestion avoidance sender won't send more until
1449: * he gets an ACK.
1450: *
1451: * See above.
1452: */
1453: /* if (ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg) {
1454: */
1455: /* if ((ti->ti_len && (unsigned)ti->ti_len < tp->t_maxseg &&
1456: * (so->so_iptos & IPTOS_LOWDELAY) == 0) ||
1457: * ((so->so_iptos & IPTOS_LOWDELAY) &&
1458: * ((struct tcpiphdr_2 *)ti)->first_char == (char)27)) {
1459: */
1460: if (ti->ti_len && (unsigned)ti->ti_len <= 5 &&
1461: ((struct tcpiphdr_2 *)ti)->first_char == (char)27) {
1462: tp->t_flags |= TF_ACKNOW;
1463: }
1464:
1465: /*
1466: * Return any desired output.
1467: */
1468: if (needoutput || (tp->t_flags & TF_ACKNOW)) {
1469: (void) tcp_output(tp);
1470: }
1471: return;
1472:
1473: dropafterack:
1474: /*
1475: * Generate an ACK dropping incoming segment if it occupies
1476: * sequence space, where the ACK reflects our state.
1477: */
1478: if (tiflags & TH_RST)
1479: goto drop;
1480: m_freem(m);
1481: tp->t_flags |= TF_ACKNOW;
1482: (void) tcp_output(tp);
1483: return;
1484:
1485: dropwithreset:
1486: /* reuses m if m!=NULL, m_free() unnecessary */
1487: if (tiflags & TH_ACK)
1488: tcp_respond(tp, ti, m, (tcp_seq)0, ti->ti_ack, TH_RST);
1489: else {
1490: if (tiflags & TH_SYN) ti->ti_len++;
1491: tcp_respond(tp, ti, m, ti->ti_seq+ti->ti_len, (tcp_seq)0,
1492: TH_RST|TH_ACK);
1493: }
1494:
1495: return;
1496:
1497: drop:
1498: /*
1499: * Drop space held by incoming segment and return.
1500: */
1501: m_free(m);
1502:
1503: return;
1504: }
1505:
1506: /* , ts_present, ts_val, ts_ecr) */
1507: /* int *ts_present;
1508: * u_int32_t *ts_val, *ts_ecr;
1509: */
1510: void
1511: tcp_dooptions(tp, cp, cnt, ti)
1512: struct tcpcb *tp;
1513: u_char *cp;
1514: int cnt;
1515: struct tcpiphdr *ti;
1516: {
1517: u_int16_t mss;
1518: int opt, optlen;
1519:
1520: DEBUG_CALL("tcp_dooptions");
1521: DEBUG_ARGS((dfd," tp = %lx cnt=%i \n", (long )tp, cnt));
1522:
1523: for (; cnt > 0; cnt -= optlen, cp += optlen) {
1524: opt = cp[0];
1525: if (opt == TCPOPT_EOL)
1526: break;
1527: if (opt == TCPOPT_NOP)
1528: optlen = 1;
1529: else {
1530: optlen = cp[1];
1531: if (optlen <= 0)
1532: break;
1533: }
1534: switch (opt) {
1535:
1536: default:
1537: continue;
1538:
1539: case TCPOPT_MAXSEG:
1540: if (optlen != TCPOLEN_MAXSEG)
1541: continue;
1542: if (!(ti->ti_flags & TH_SYN))
1543: continue;
1544: memcpy((char *) &mss, (char *) cp + 2, sizeof(mss));
1545: NTOHS(mss);
1546: (void) tcp_mss(tp, mss); /* sets t_maxseg */
1547: break;
1548:
1549: /* case TCPOPT_WINDOW:
1550: * if (optlen != TCPOLEN_WINDOW)
1551: * continue;
1552: * if (!(ti->ti_flags & TH_SYN))
1553: * continue;
1554: * tp->t_flags |= TF_RCVD_SCALE;
1555: * tp->requested_s_scale = min(cp[2], TCP_MAX_WINSHIFT);
1556: * break;
1557: */
1558: /* case TCPOPT_TIMESTAMP:
1559: * if (optlen != TCPOLEN_TIMESTAMP)
1560: * continue;
1561: * *ts_present = 1;
1562: * memcpy((char *) ts_val, (char *)cp + 2, sizeof(*ts_val));
1563: * NTOHL(*ts_val);
1564: * memcpy((char *) ts_ecr, (char *)cp + 6, sizeof(*ts_ecr));
1565: * NTOHL(*ts_ecr);
1566: *
1567: */ /*
1568: * * A timestamp received in a SYN makes
1569: * * it ok to send timestamp requests and replies.
1570: * */
1571: /* if (ti->ti_flags & TH_SYN) {
1572: * tp->t_flags |= TF_RCVD_TSTMP;
1573: * tp->ts_recent = *ts_val;
1574: * tp->ts_recent_age = tcp_now;
1575: * }
1576: */ break;
1577: }
1578: }
1579: }
1580:
1581:
1582: /*
1583: * Pull out of band byte out of a segment so
1584: * it doesn't appear in the user's data queue.
1585: * It is still reflected in the segment length for
1586: * sequencing purposes.
1587: */
1588:
1589: #ifdef notdef
1590:
1591: void
1592: tcp_pulloutofband(so, ti, m)
1593: struct socket *so;
1594: struct tcpiphdr *ti;
1595: register struct mbuf *m;
1596: {
1597: int cnt = ti->ti_urp - 1;
1598:
1599: while (cnt >= 0) {
1600: if (m->m_len > cnt) {
1601: char *cp = mtod(m, caddr_t) + cnt;
1602: struct tcpcb *tp = sototcpcb(so);
1603:
1604: tp->t_iobc = *cp;
1605: tp->t_oobflags |= TCPOOB_HAVEDATA;
1606: memcpy(sp, cp+1, (unsigned)(m->m_len - cnt - 1));
1607: m->m_len--;
1608: return;
1609: }
1610: cnt -= m->m_len;
1611: m = m->m_next; /* XXX WRONG! Fix it! */
1612: if (m == 0)
1613: break;
1614: }
1615: panic("tcp_pulloutofband");
1616: }
1617:
1618: #endif /* notdef */
1619:
1620: /*
1621: * Collect new round-trip time estimate
1622: * and update averages and current timeout.
1623: */
1624:
1625: void
1626: tcp_xmit_timer(tp, rtt)
1627: register struct tcpcb *tp;
1628: int rtt;
1629: {
1630: register short delta;
1631:
1632: DEBUG_CALL("tcp_xmit_timer");
1633: DEBUG_ARG("tp = %lx", (long)tp);
1634: DEBUG_ARG("rtt = %d", rtt);
1635:
1636: tcpstat.tcps_rttupdated++;
1637: if (tp->t_srtt != 0) {
1638: /*
1639: * srtt is stored as fixed point with 3 bits after the
1640: * binary point (i.e., scaled by 8). The following magic
1641: * is equivalent to the smoothing algorithm in rfc793 with
1642: * an alpha of .875 (srtt = rtt/8 + srtt*7/8 in fixed
1643: * point). Adjust rtt to origin 0.
1644: */
1645: delta = rtt - 1 - (tp->t_srtt >> TCP_RTT_SHIFT);
1646: if ((tp->t_srtt += delta) <= 0)
1647: tp->t_srtt = 1;
1648: /*
1649: * We accumulate a smoothed rtt variance (actually, a
1650: * smoothed mean difference), then set the retransmit
1651: * timer to smoothed rtt + 4 times the smoothed variance.
1652: * rttvar is stored as fixed point with 2 bits after the
1653: * binary point (scaled by 4). The following is
1654: * equivalent to rfc793 smoothing with an alpha of .75
1655: * (rttvar = rttvar*3/4 + |delta| / 4). This replaces
1656: * rfc793's wired-in beta.
1657: */
1658: if (delta < 0)
1659: delta = -delta;
1660: delta -= (tp->t_rttvar >> TCP_RTTVAR_SHIFT);
1661: if ((tp->t_rttvar += delta) <= 0)
1662: tp->t_rttvar = 1;
1663: } else {
1664: /*
1665: * No rtt measurement yet - use the unsmoothed rtt.
1666: * Set the variance to half the rtt (so our first
1667: * retransmit happens at 3*rtt).
1668: */
1669: tp->t_srtt = rtt << TCP_RTT_SHIFT;
1670: tp->t_rttvar = rtt << (TCP_RTTVAR_SHIFT - 1);
1671: }
1672: tp->t_rtt = 0;
1673: tp->t_rxtshift = 0;
1674:
1675: /*
1676: * the retransmit should happen at rtt + 4 * rttvar.
1677: * Because of the way we do the smoothing, srtt and rttvar
1678: * will each average +1/2 tick of bias. When we compute
1679: * the retransmit timer, we want 1/2 tick of rounding and
1680: * 1 extra tick because of +-1/2 tick uncertainty in the
1681: * firing of the timer. The bias will give us exactly the
1682: * 1.5 tick we need. But, because the bias is
1683: * statistical, we have to test that we don't drop below
1684: * the minimum feasible timer (which is 2 ticks).
1685: */
1686: TCPT_RANGESET(tp->t_rxtcur, TCP_REXMTVAL(tp),
1687: (short)tp->t_rttmin, TCPTV_REXMTMAX); /* XXX */
1688:
1689: /*
1690: * We received an ack for a packet that wasn't retransmitted;
1691: * it is probably safe to discard any error indications we've
1692: * received recently. This isn't quite right, but close enough
1693: * for now (a route might have failed after we sent a segment,
1694: * and the return path might not be symmetrical).
1695: */
1696: tp->t_softerror = 0;
1697: }
1698:
1699: /*
1700: * Determine a reasonable value for maxseg size.
1701: * If the route is known, check route for mtu.
1702: * If none, use an mss that can be handled on the outgoing
1703: * interface without forcing IP to fragment; if bigger than
1704: * an mbuf cluster (MCLBYTES), round down to nearest multiple of MCLBYTES
1705: * to utilize large mbufs. If no route is found, route has no mtu,
1706: * or the destination isn't local, use a default, hopefully conservative
1707: * size (usually 512 or the default IP max size, but no more than the mtu
1708: * of the interface), as we can't discover anything about intervening
1709: * gateways or networks. We also initialize the congestion/slow start
1710: * window to be a single segment if the destination isn't local.
1711: * While looking at the routing entry, we also initialize other path-dependent
1712: * parameters from pre-set or cached values in the routing entry.
1713: */
1714:
1715: int
1716: tcp_mss(tp, offer)
1717: register struct tcpcb *tp;
1718: u_int offer;
1719: {
1720: struct socket *so = tp->t_socket;
1721: int mss;
1722:
1723: DEBUG_CALL("tcp_mss");
1724: DEBUG_ARG("tp = %lx", (long)tp);
1725: DEBUG_ARG("offer = %d", offer);
1726:
1727: mss = min(if_mtu, if_mru) - sizeof(struct tcpiphdr);
1728: if (offer)
1729: mss = min(mss, offer);
1730: mss = max(mss, 32);
1731: if (mss < tp->t_maxseg || offer != 0)
1732: tp->t_maxseg = mss;
1733:
1734: tp->snd_cwnd = mss;
1735:
1736: sbreserve(&so->so_snd, tcp_sndspace+((tcp_sndspace%mss)?(mss-(tcp_sndspace%mss)):0));
1737: sbreserve(&so->so_rcv, tcp_rcvspace+((tcp_rcvspace%mss)?(mss-(tcp_rcvspace%mss)):0));
1738:
1739: DEBUG_MISC((dfd, " returning mss = %d\n", mss));
1740:
1741: return mss;
1742: }
This archive runs on limited infrastructure. Preserving old code on modern bandwidth. Automated agents are requested to crawl responsibly.