tcp_output.c@ 50690

Last change on this file since 50690 was 44528, checked in by vboxsync, 12 years ago
header (C) fixes
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 23.1 KB

Line
1	/* $Id: tcp_output.c 44528 2013-02-04 14:27:54Z vboxsync $ */
2	/** @file
3	* NAT - TCP output.
4	*/
5
6	/*
7	* Copyright (C) 2006-2012 Oracle Corporation
8	*
9	* This file is part of VirtualBox Open Source Edition (OSE), as
10	* available from http://www.virtualbox.org. This file is free software;
11	* you can redistribute it and/or modify it under the terms of the GNU
12	* General Public License (GPL) as published by the Free Software
13	* Foundation, in version 2 as it comes in the "COPYING" file of the
14	* VirtualBox OSE distribution. VirtualBox OSE is distributed in the
15	* hope that it will be useful, but WITHOUT ANY WARRANTY of any kind.
16	*/
17
18	/*
19	* This code is based on:
20	*
21	* Copyright (c) 1982, 1986, 1988, 1990, 1993
22	* The Regents of the University of California. All rights reserved.
23	*
24	* Redistribution and use in source and binary forms, with or without
25	* modification, are permitted provided that the following conditions
26	* are met:
27	* 1. Redistributions of source code must retain the above copyright
28	* notice, this list of conditions and the following disclaimer.
29	* 2. Redistributions in binary form must reproduce the above copyright
30	* notice, this list of conditions and the following disclaimer in the
31	* documentation and/or other materials provided with the distribution.
32	* 3. All advertising materials mentioning features or use of this software
33	* must display the following acknowledgement:
34	* This product includes software developed by the University of
35	* California, Berkeley and its contributors.
36	* 4. Neither the name of the University nor the names of its contributors
37	* may be used to endorse or promote products derived from this software
38	* without specific prior written permission.
39	*
40	* THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
41	* ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
42	* IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
43	* ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
44	* FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
45	* DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
46	* OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
47	* HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
48	* LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
49	* OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
50	* SUCH DAMAGE.
51	*
52	* @(#)tcp_output.c 8.3 (Berkeley) 12/30/93
53	* tcp_output.c,v 1.3 1994/09/15 10:36:55 davidg Exp
54	*/
55
56	/*
57	* Changes and additions relating to SLiRP
58	* Copyright (c) 1995 Danny Gasparovski.
59	*
60	* Please read the file COPYRIGHT for the
61	* terms and conditions of the copyright.
62	*/
63
64	#include <slirp.h>
65
66	/*
67	* Since this is only used in "stats socket", we give meaning
68	* names instead of the REAL names
69	*/
70	const char * const tcpstates[] =
71	{
72	/* "CLOSED", "LISTEN", "SYN_SENT", "SYN_RCVD", */
73	"REDIRECT", "LISTEN", "SYN_SENT", "SYN_RCVD",
74	"ESTABLISHED", "CLOSE_WAIT", "FIN_WAIT_1", "CLOSING",
75	"LAST_ACK", "FIN_WAIT_2", "TIME_WAIT",
76	};
77
78	static const u_char tcp_outflags[TCP_NSTATES] =
79	{
80	TH_RST\|TH_ACK, 0, TH_SYN, TH_SYN\|TH_ACK,
81	TH_ACK, TH_ACK, TH_FIN\|TH_ACK, TH_FIN\|TH_ACK,
82	TH_FIN\|TH_ACK, TH_ACK, TH_ACK,
83	};
84
85
86	#define MAX_TCPOPTLEN 32 /* max # bytes that go in options */
87
88	/*
89	* Tcp output routine: figure out what should be sent and send it.
90	*/
91	int
92	tcp_output(PNATState pData, register struct tcpcb *tp)
93	{
94	register struct socket *so = tp->t_socket;
95	register long len, win;
96	int off, flags, error;
97	register struct mbuf *m = NULL;
98	register struct tcpiphdr *ti;
99	u_char opt[MAX_TCPOPTLEN];
100	unsigned optlen, hdrlen;
101	int idle, sendalot;
102	int size = 0;
103
104	LogFlowFunc(("ENTER: tcp_output: tp = %R[tcpcb793]\n", tp));
105
106	/*
107	* Determine length of data that should be transmitted,
108	* and flags that will be used.
109	* If there is some data or critical controls (SYN, RST)
110	* to send, then transmit; otherwise, investigate further.
111	*/
112	idle = (tp->snd_max == tp->snd_una);
113	if (idle && tp->t_idle >= tp->t_rxtcur)
114	/*
115	* We have been idle for "a while" and no acks are
116	* expected to clock out any data we send --
117	* slow start to get ack "clock" running again.
118	*/
119	tp->snd_cwnd = tp->t_maxseg;
120
121	again:
122	sendalot = 0;
123	off = tp->snd_nxt - tp->snd_una;
124	win = min(tp->snd_wnd, tp->snd_cwnd);
125
126	flags = tcp_outflags[tp->t_state];
127
128	Log2((" --- tcp_output flags = 0x%x\n", flags));
129
130	/*
131	* If in persist timeout with window of 0, send 1 byte.
132	* Otherwise, if window is small but nonzero
133	* and timer expired, we will send what we can
134	* and go to transmit state.
135	*/
136	if (tp->t_force)
137	{
138	if (win == 0)
139	{
140	/*
141	* If we still have some data to send, then
142	* clear the FIN bit. Usually this would
143	* happen below when it realizes that we
144	* aren't sending all the data. However,
145	* if we have exactly 1 byte of unset data,
146	* then it won't clear the FIN bit below,
147	* and if we are in persist state, we wind
148	* up sending the packet without recording
149	* that we sent the FIN bit.
150	*
151	* We can't just blindly clear the FIN bit,
152	* because if we don't have any more data
153	* to send then the probe will be the FIN
154	* itself.
155	*/
156	if (off < SBUF_LEN(&so->so_snd))
157	flags &= ~TH_FIN;
158	win = 1;
159	}
160	else
161	{
162	tp->t_timer[TCPT_PERSIST] = 0;
163	tp->t_rxtshift = 0;
164	}
165	}
166
167	len = min(SBUF_LEN(&so->so_snd), win) - off;
168	if (len < 0)
169	{
170	/*
171	* If FIN has been sent but not acked,
172	* but we haven't been called to retransmit,
173	* len will be -1. Otherwise, window shrank
174	* after we sent into it. If window shrank to 0,
175	* cancel pending retransmit and pull snd_nxt
176	* back to (closed) window. We will enter persist
177	* state below. If the window didn't close completely,
178	* just wait for an ACK.
179	*/
180	len = 0;
181	if (win == 0)
182	{
183	tp->t_timer[TCPT_REXMT] = 0;
184	tp->snd_nxt = tp->snd_una;
185	}
186	}
187	if (len > tp->t_maxseg)
188	{
189	len = tp->t_maxseg;
190	sendalot = 1;
191	}
192	if (SEQ_LT(tp->snd_nxt + len, tp->snd_una + SBUF_LEN(&so->so_snd)))
193	flags &= ~TH_FIN;
194
195	win = sbspace(&so->so_rcv);
196
197	/*
198	* Sender silly window avoidance. If connection is idle
199	* and can send all data, a maximum segment,
200	* at least a maximum default-size segment do it,
201	* or are forced, do it; otherwise don't bother.
202	* If peer's buffer is tiny, then send
203	* when window is at least half open.
204	* If retransmitting (possibly after persist timer forced us
205	* to send into a small window), then must resend.
206	*/
207	if (len)
208	{
209	if (len == tp->t_maxseg)
210	goto send;
211	if ((1 \|\| idle \|\| tp->t_flags & TF_NODELAY) &&
212	len + off >= SBUF_LEN(&so->so_snd))
213	goto send;
214	if (tp->t_force)
215	goto send;
216	if (len >= tp->max_sndwnd / 2 && tp->max_sndwnd > 0)
217	goto send;
218	if (SEQ_LT(tp->snd_nxt, tp->snd_max))
219	goto send;
220	}
221
222	/*
223	* Compare available window to amount of window
224	* known to peer (as advertised window less
225	* next expected input). If the difference is at least two
226	* max size segments, or at least 50% of the maximum possible
227	* window, then want to send a window update to peer.
228	*/
229	if (win > 0)
230	{
231	/*
232	* "adv" is the amount we can increase the window,
233	* taking into account that we are limited by
234	* TCP_MAXWIN << tp->rcv_scale.
235	*/
236	long adv = min(win,
237	(long)TCP_MAXWIN << tp->rcv_scale) -
238	(tp->rcv_adv - tp->rcv_nxt);
239
240	if (adv >= (long) (2 * tp->t_maxseg))
241	goto send;
242	if (2 * adv >= (long) SBUF_SIZE(&so->so_rcv))
243	goto send;
244	}
245
246	/*
247	* Send if we owe peer an ACK.
248	*/
249	if (tp->t_flags & TF_ACKNOW)
250	goto send;
251	if (flags & (TH_SYN\|TH_RST))
252	goto send;
253	if (SEQ_GT(tp->snd_up, tp->snd_una))
254	goto send;
255	/*
256	* If our state indicates that FIN should be sent
257	* and we have not yet done so, or we're retransmitting the FIN,
258	* then we need to send.
259	*/
260	if ( flags & TH_FIN
261	&& ((tp->t_flags & TF_SENTFIN) == 0 \|\| tp->snd_nxt == tp->snd_una))
262	goto send;
263
264	/*
265	* TCP window updates are not reliable, rather a polling protocol
266	* using ``persist'' packets is used to insure receipt of window
267	* updates. The three ``states'' for the output side are:
268	* idle not doing retransmits or persists
269	* persisting to move a small or zero window
270	* (re)transmitting and thereby not persisting
271	*
272	* tp->t_timer[TCPT_PERSIST]
273	* is set when we are in persist state.
274	* tp->t_force
275	* is set when we are called to send a persist packet.
276	* tp->t_timer[TCPT_REXMT]
277	* is set when we are retransmitting
278	* The output side is idle when both timers are zero.
279	*
280	* If send window is too small, there is data to transmit, and no
281	* retransmit or persist is pending, then go to persist state.
282	* If nothing happens soon, send when timer expires:
283	* if window is nonzero, transmit what we can,
284	* otherwise force out a byte.
285	*/
286	if ( SBUF_LEN(&so->so_snd)
287	&& tp->t_timer[TCPT_REXMT] == 0
288	&& tp->t_timer[TCPT_PERSIST] == 0)
289	{
290	tp->t_rxtshift = 0;
291	tcp_setpersist(tp);
292	}
293
294	/*
295	* No reason to send a segment, just return.
296	*/
297	tcpstat.tcps_didnuttin++;
298
299	LogFlowFuncLeave();
300	return (0);
301
302	send:
303	LogFlowFunc(("send\n"));
304	/*
305	* Before ESTABLISHED, force sending of initial options
306	* unless TCP set not to do any options.
307	* NOTE: we assume that the IP/TCP header plus TCP options
308	* always fit in a single mbuf, leaving room for a maximum
309	* link header, i.e.
310	* max_linkhdr + sizeof (struct tcpiphdr) + optlen <= MHLEN
311	*/
312	optlen = 0;
313	hdrlen = sizeof (struct tcpiphdr);
314	if (flags & TH_SYN)
315	{
316	tp->snd_nxt = tp->iss;
317	if ((tp->t_flags & TF_NOOPT) == 0)
318	{
319	u_int16_t mss;
320
321	opt[0] = TCPOPT_MAXSEG;
322	opt[1] = 4;
323	mss = RT_H2N_U16((u_int16_t) tcp_mss(pData, tp, 0));
324	memcpy((caddr_t)(opt + 2), (caddr_t)&mss, sizeof(mss));
325	optlen = 4;
326
327	#if 0
328	if ( (tp->t_flags & TF_REQ_SCALE)
329	&& ( (flags & TH_ACK) == 0
330	\|\| (tp->t_flags & TF_RCVD_SCALE)))
331	{
332	((u_int32_t ) (opt + optlen)) = RT_H2N_U32( TCPOPT_NOP << 24
333	\| TCPOPT_WINDOW << 16
334	\| TCPOLEN_WINDOW << 8
335	\| tp->request_r_scale);
336	optlen += 4;
337	}
338	#endif
339	}
340	}
341
342	/*
343	* Send a timestamp and echo-reply if this is a SYN and our side
344	* wants to use timestamps (TF_REQ_TSTMP is set) or both our side
345	* and our peer have sent timestamps in our SYN's.
346	*/
347	#if 0
348	if ( (tp->t_flags & (TF_REQ_TSTMP\|TF_NOOPT)) == TF_REQ_TSTMP
349	&& (flags & TH_RST) == 0
350	&& ( (flags & (TH_SYN\|TH_ACK)) == TH_SYN
351	\|\| (tp->t_flags & TF_RCVD_TSTMP)))
352	{
353	u_int32_t lp = (u_int32_t )(opt + optlen);
354
355	/* Form timestamp option as shown in appendix A of RFC 1323. */
356	*lp++ = RT_H2N_U32_C(TCPOPT_TSTAMP_HDR);
357	*lp++ = RT_H2N_U32(tcp_now);
358	*lp = RT_H2N_U32(tp->ts_recent);
359	optlen += TCPOLEN_TSTAMP_APPA;
360	}
361	#endif
362	hdrlen += optlen;
363
364	/*
365	* Adjust data length if insertion of options will
366	* bump the packet length beyond the t_maxseg length.
367	*/
368	if (len > tp->t_maxseg - optlen)
369	{
370	len = tp->t_maxseg - optlen;
371	sendalot = 1;
372	}
373
374	/*
375	* Grab a header mbuf, attaching a copy of data to
376	* be transmitted, and initialize the header from
377	* the template for sends on this connection.
378	*/
379	if (len)
380	{
381	if (tp->t_force && len == 1)
382	tcpstat.tcps_sndprobe++;
383	else if (SEQ_LT(tp->snd_nxt, tp->snd_max))
384	{
385	tcpstat.tcps_sndrexmitpack++;
386	tcpstat.tcps_sndrexmitbyte += len;
387	}
388	else
389	{
390	tcpstat.tcps_sndpack++;
391	tcpstat.tcps_sndbyte += len;
392	}
393
394	size = MCLBYTES;
395	if ((len + hdrlen + ETH_HLEN) < MSIZE)
396	size = MCLBYTES;
397	else if ((len + hdrlen + ETH_HLEN) < MCLBYTES)
398	size = MCLBYTES;
399	else if((len + hdrlen + ETH_HLEN) < MJUM9BYTES)
400	size = MJUM9BYTES;
401	else if ((len + hdrlen + ETH_HLEN) < MJUM16BYTES)
402	size = MJUM16BYTES;
403	else
404	AssertMsgFailed(("Unsupported size"));
405	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
406	if (m == NULL)
407	{
408	/* error = ENOBUFS; */
409	error = 1;
410	goto out;
411	}
412	m->m_data += if_maxlinkhdr;
413	m->m_pkthdr.header = mtod(m, void *);
414	m->m_len = hdrlen;
415
416	/*
417	* This will always succeed, since we make sure our mbufs
418	* are big enough to hold one MSS packet + header + ... etc.
419	*/
420	#if 0
421	if (len <= MHLEN - hdrlen - max_linkhdr)
422	{
423	#endif
424	#ifndef VBOX_WITH_SLIRP_BSD_SBUF
425	sbcopy(&so->so_snd, off, (int) len, mtod(m, caddr_t) + hdrlen);
426	m->m_len += len;
427	#else
428	m_copyback(pData, m, hdrlen, len, sbuf_data(&so->so_snd) + off);
429	#endif
430	#if 0
431	}
432	else
433	{
434	m->m_next = m_copy(so->so_snd.sb_mb, off, (int) len);
435	if (m->m_next == 0)
436	len = 0;
437	}
438	#endif
439	/*
440	* If we're sending everything we've got, set PUSH.
441	* (This will keep happy those implementations which only
442	* give data to the user when a buffer fills or
443	* a PUSH comes in.)
444	*/
445	if (off + len == SBUF_LEN(&so->so_snd))
446	flags \|= TH_PUSH;
447	}
448	else
449	{
450	bool fUninitiolizedTemplate = false;
451	if (tp->t_flags & TF_ACKNOW)
452	tcpstat.tcps_sndacks++;
453	else if (flags & (TH_SYN\|TH_FIN\|TH_RST))
454	tcpstat.tcps_sndctrl++;
455	else if (SEQ_GT(tp->snd_up, tp->snd_una))
456	tcpstat.tcps_sndurg++;
457	else
458	tcpstat.tcps_sndwinup++;
459
460	if ((hdrlen + ETH_HLEN) < MSIZE)
461	{
462	size = MCLBYTES;
463	}
464	else if ((hdrlen + ETH_HLEN) < MCLBYTES)
465	{
466	size = MCLBYTES;
467	}
468	else if((hdrlen + ETH_HLEN) < MJUM9BYTES)
469	{
470	size = MJUM9BYTES;
471	}
472	else if ((hdrlen + ETH_HLEN) < MJUM16BYTES)
473	{
474	size = MJUM16BYTES;
475	}
476	else
477	{
478	AssertMsgFailed(("Unsupported size"));
479	}
480	m = m_getjcl(pData, M_NOWAIT, MT_HEADER, M_PKTHDR, size);
481	if (m == NULL)
482	{
483	/* error = ENOBUFS; */
484	error = 1;
485	goto out;
486	}
487	m->m_data += if_maxlinkhdr;
488	m->m_pkthdr.header = mtod(m, void *);
489	m->m_len = hdrlen;
490	/*
491	* Uninitialized TCP template looks very suspicious at this processing state, thus why we have
492	* to workaround the problem till right fix. Warning appears once at release log.
493	*/
494	fUninitiolizedTemplate = RT_BOOL(( tp->t_template.ti_src.s_addr == INADDR_ANY
495	\|\| tp->t_template.ti_dst.s_addr == INADDR_ANY));
496	#ifndef DEBUG_vvl
497	if (fUninitiolizedTemplate)
498	{
499	static bool fWarn;
500	tcp_template(tp);
501	if(!fWarn)
502	{
503	LogRel(("NAT:TCP: TCP template was created forcely from socket information\n"));
504	fWarn = true;
505	}
506	}
507	#else
508	Assert((!fUninitiolizedTemplate));
509	#endif
510	}
511
512	ti = mtod(m, struct tcpiphdr *);
513
514	memcpy((caddr_t)ti, &tp->t_template, sizeof (struct tcpiphdr));
515
516	/*
517	* Fill in fields, remembering maximum advertised
518	* window for use in delaying messages about window sizes.
519	* If resending a FIN, be sure not to use a new sequence number.
520	*/
521	if ( flags & TH_FIN
522	&& tp->t_flags & TF_SENTFIN
523	&& tp->snd_nxt == tp->snd_max)
524	tp->snd_nxt--;
525	/*
526	* If we are doing retransmissions, then snd_nxt will
527	* not reflect the first unsent octet. For ACK only
528	* packets, we do not want the sequence number of the
529	* retransmitted packet, we want the sequence number
530	* of the next unsent octet. So, if there is no data
531	* (and no SYN or FIN), use snd_max instead of snd_nxt
532	* when filling in ti_seq. But if we are in persist
533	* state, snd_max might reflect one byte beyond the
534	* right edge of the window, so use snd_nxt in that
535	* case, since we know we aren't doing a retransmission.
536	* (retransmit and persist are mutually exclusive...)
537	*/
538	if (len \|\| (flags & (TH_SYN\|TH_FIN)) \|\| tp->t_timer[TCPT_PERSIST])
539	ti->ti_seq = RT_H2N_U32(tp->snd_nxt);
540	else
541	ti->ti_seq = RT_H2N_U32(tp->snd_max);
542	ti->ti_ack = RT_H2N_U32(tp->rcv_nxt);
543	if (optlen)
544	{
545	memcpy((caddr_t)(ti + 1), (caddr_t)opt, optlen);
546	ti->ti_off = (sizeof (struct tcphdr) + optlen) >> 2;
547	}
548	ti->ti_flags = flags;
549	/*
550	* Calculate receive window. Don't shrink window,
551	* but avoid silly window syndrome.
552	*/
553	if (win < (long)(SBUF_SIZE(&so->so_rcv) / 4) && win < (long)tp->t_maxseg)
554	win = 0;
555	if (win > (long)TCP_MAXWIN << tp->rcv_scale)
556	win = (long)TCP_MAXWIN << tp->rcv_scale;
557	if (win < (long)(tp->rcv_adv - tp->rcv_nxt))
558	win = (long)(tp->rcv_adv - tp->rcv_nxt);
559	ti->ti_win = RT_H2N_U16((u_int16_t) (win>>tp->rcv_scale));
560
561	#if 0
562	if (SEQ_GT(tp->snd_up, tp->snd_nxt))
563	{
564	ti->ti_urp = RT_H2N_U16((u_int16_t)(tp->snd_up - tp->snd_nxt));
565	#else
566	if (SEQ_GT(tp->snd_up, tp->snd_una))
567	{
568	ti->ti_urp = RT_H2N_U16((u_int16_t)(tp->snd_up - RT_N2H_U32(ti->ti_seq)));
569	#endif
570	ti->ti_flags \|= TH_URG;
571	}
572	else
573	/*
574	* If no urgent pointer to send, then we pull
575	* the urgent pointer to the left edge of the send window
576	* so that it doesn't drift into the send window on sequence
577	* number wraparound.
578	*/
579	tp->snd_up = tp->snd_una; /* drag it along */
580
581	/*
582	* Put TCP length in extended header, and then
583	* checksum extended header and data.
584	*/
585	if (len + optlen)
586	ti->ti_len = RT_H2N_U16((u_int16_t)(sizeof (struct tcphdr)
587	+ optlen + len));
588	ti->ti_sum = cksum(m, (int)(hdrlen + len));
589
590	/*
591	* In transmit state, time the transmission and arrange for
592	* the retransmit. In persist state, just set snd_max.
593	*/
594	if (tp->t_force == 0 \|\| tp->t_timer[TCPT_PERSIST] == 0)
595	{
596	tcp_seq startseq = tp->snd_nxt;
597
598	/*
599	* Advance snd_nxt over sequence space of this segment.
600	*/
601	if (flags & (TH_SYN\|TH_FIN))
602	{
603	if (flags & TH_SYN)
604	tp->snd_nxt++;
605	if (flags & TH_FIN)
606	{
607	tp->snd_nxt++;
608	tp->t_flags \|= TF_SENTFIN;
609	}
610	}
611	tp->snd_nxt += len;
612	if (SEQ_GT(tp->snd_nxt, tp->snd_max))
613	{
614	tp->snd_max = tp->snd_nxt;
615	/*
616	* Time this transmission if not a retransmission and
617	* not currently timing anything.
618	*/
619	if (tp->t_rtt == 0)
620	{
621	tp->t_rtt = 1;
622	tp->t_rtseq = startseq;
623	tcpstat.tcps_segstimed++;
624	}
625	}
626
627	/*
628	* Set retransmit timer if not currently set,
629	* and not doing an ack or a keep-alive probe.
630	* Initial value for retransmit timer is smoothed
631	* round-trip time + 2 * round-trip time variance.
632	* Initialize shift counter which is used for backoff
633	* of retransmit time.
634	*/
635	if ( tp->t_timer[TCPT_REXMT] == 0
636	&& tp->snd_nxt != tp->snd_una)
637	{
638	tp->t_timer[TCPT_REXMT] = tp->t_rxtcur;
639	if (tp->t_timer[TCPT_PERSIST])
640	{
641	tp->t_timer[TCPT_PERSIST] = 0;
642	tp->t_rxtshift = 0;
643	}
644	}
645	}
646	else
647	if (SEQ_GT(tp->snd_nxt + len, tp->snd_max))
648	tp->snd_max = tp->snd_nxt + len;
649
650	/*
651	* Fill in IP length and desired time to live and
652	* send to IP level. There should be a better way
653	* to handle ttl and tos; we could keep them in
654	* the template, but need a way to checksum without them.
655	*/
656	M_ASSERTPKTHDR(m);
657	m->m_pkthdr.header = mtod(m, void *);
658	m->m_len = hdrlen + len; /* XXX Needed? m_len should be correct */
659
660	{
661	((struct ip *)ti)->ip_len = m->m_len;
662	((struct ip *)ti)->ip_ttl = ip_defttl;
663	((struct ip *)ti)->ip_tos = so->so_iptos;
664
665	/* #if BSD >= 43 */
666	/* Don't do IP options... */
667	#if 0
668	error = ip_output(m, tp->t_inpcb->inp_options, &tp->t_inpcb->inp_route,
669	so->so_options & SO_DONTROUTE, 0);
670	#endif
671	error = ip_output(pData, so, m);
672
673	#if 0
674	/* #else */
675	error = ip_output(m, (struct mbuf *)0, &tp->t_inpcb->inp_route,
676	so->so_options & SO_DONTROUTE);
677	/* #endif */
678	#endif
679	}
680	if (error)
681	{
682	out:
683	#if 0
684	if (error == ENOBUFS)
685	{
686	tcp_quench(tp->t_inpcb, 0);
687	return (0);
688	}
689
690	if ( ( error == EHOSTUNREACH
691	\|\| error == ENETDOWN)
692	&& TCPS_HAVERCVDSYN(tp->t_state))
693	{
694	tp->t_softerror = error;
695	return (0);
696	}
697	#endif
698	if (m != NULL)
699	m_freem(pData, m);
700	return (error);
701	}
702	tcpstat.tcps_sndtotal++;
703
704	/*
705	* Data sent (as far as we can tell).
706	* If this advertises a larger window than any other segment,
707	* then remember the size of the advertised window.
708	* Any pending ACK has now been sent.
709	*/
710	if (win > 0 && SEQ_GT(tp->rcv_nxt+win, tp->rcv_adv))
711	tp->rcv_adv = tp->rcv_nxt + win;
712	tp->last_ack_sent = tp->rcv_nxt;
713	tp->t_flags &= ~(TF_ACKNOW\|TF_DELACK);
714	if (sendalot)
715	goto again;
716
717	return (0);
718	}
719
720	void
721	tcp_setpersist(struct tcpcb *tp)
722	{
723	int t = ((tp->t_srtt >> 2) + tp->t_rttvar) >> 1;
724
725	#if 0
726	if (tp->t_timer[TCPT_REXMT])
727	panic("tcp_output REXMT");
728	#endif
729	/*
730	* Start/restart persistence timer.
731	*/
732	TCPT_RANGESET(tp->t_timer[TCPT_PERSIST],
733	t * tcp_backoff[tp->t_rxtshift],
734	TCPTV_PERSMIN, TCPTV_PERSMAX);
735	if (tp->t_rxtshift < TCP_MAXRXTSHIFT)
736	tp->t_rxtshift++;
737	}

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/Devices/Network/slirp/tcp_output.c@ 50690

Download in other formats: