xref: /netbsd-src/sys/netinet/tcp_var.h (revision ba65fde2d7fefa7d39838fa5fa855e62bd606b5e)
1 /*	$NetBSD: tcp_var.h,v 1.169 2012/02/02 19:43:08 tls Exp $	*/
2 
3 /*
4  * Copyright (C) 1995, 1996, 1997, and 1998 WIDE Project.
5  * All rights reserved.
6  *
7  * Redistribution and use in source and binary forms, with or without
8  * modification, are permitted provided that the following conditions
9  * are met:
10  * 1. Redistributions of source code must retain the above copyright
11  *    notice, this list of conditions and the following disclaimer.
12  * 2. Redistributions in binary form must reproduce the above copyright
13  *    notice, this list of conditions and the following disclaimer in the
14  *    documentation and/or other materials provided with the distribution.
15  * 3. Neither the name of the project nor the names of its contributors
16  *    may be used to endorse or promote products derived from this software
17  *    without specific prior written permission.
18  *
19  * THIS SOFTWARE IS PROVIDED BY THE PROJECT AND CONTRIBUTORS ``AS IS'' AND
20  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22  * ARE DISCLAIMED.  IN NO EVENT SHALL THE PROJECT OR CONTRIBUTORS BE LIABLE
23  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29  * SUCH DAMAGE.
30  */
31 
32 /*
33  *      @(#)COPYRIGHT   1.1 (NRL) 17 January 1995
34  *
35  * NRL grants permission for redistribution and use in source and binary
36  * forms, with or without modification, of the software and documentation
37  * created at NRL provided that the following conditions are met:
38  *
39  * 1. Redistributions of source code must retain the above copyright
40  *    notice, this list of conditions and the following disclaimer.
41  * 2. Redistributions in binary form must reproduce the above copyright
42  *    notice, this list of conditions and the following disclaimer in the
43  *    documentation and/or other materials provided with the distribution.
44  * 3. All advertising materials mentioning features or use of this software
45  *    must display the following acknowledgements:
46  *      This product includes software developed by the University of
47  *      California, Berkeley and its contributors.
48  *      This product includes software developed at the Information
49  *      Technology Division, US Naval Research Laboratory.
50  * 4. Neither the name of the NRL nor the names of its contributors
51  *    may be used to endorse or promote products derived from this software
52  *    without specific prior written permission.
53  *
54  * THE SOFTWARE PROVIDED BY NRL IS PROVIDED BY NRL AND CONTRIBUTORS ``AS
55  * IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
56  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A
57  * PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL NRL OR
58  * CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
59  * EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
60  * PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
61  * PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
62  * LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
63  * NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
64  * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
65  *
66  * The views and conclusions contained in the software and documentation
67  * are those of the authors and should not be interpreted as representing
68  * official policies, either expressed or implied, of the US Naval
69  * Research Laboratory (NRL).
70  */
71 
72 /*-
73  * Copyright (c) 1997, 1998, 1999, 2001, 2005 The NetBSD Foundation, Inc.
74  * All rights reserved.
75  *
76  * This code is derived from software contributed to The NetBSD Foundation
77  * by Jason R. Thorpe of the Numerical Aerospace Simulation Facility,
78  * NASA Ames Research Center.
79  * This code is derived from software contributed to The NetBSD Foundation
80  * by Charles M. Hannum.
81  *
82  * Redistribution and use in source and binary forms, with or without
83  * modification, are permitted provided that the following conditions
84  * are met:
85  * 1. Redistributions of source code must retain the above copyright
86  *    notice, this list of conditions and the following disclaimer.
87  * 2. Redistributions in binary form must reproduce the above copyright
88  *    notice, this list of conditions and the following disclaimer in the
89  *    documentation and/or other materials provided with the distribution.
90  *
91  * THIS SOFTWARE IS PROVIDED BY THE NETBSD FOUNDATION, INC. AND CONTRIBUTORS
92  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED
93  * TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR
94  * PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS
95  * BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
96  * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
97  * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
98  * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
99  * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
100  * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
101  * POSSIBILITY OF SUCH DAMAGE.
102  */
103 
104 /*
105  * Copyright (c) 1982, 1986, 1993, 1994, 1995
106  *	The Regents of the University of California.  All rights reserved.
107  *
108  * Redistribution and use in source and binary forms, with or without
109  * modification, are permitted provided that the following conditions
110  * are met:
111  * 1. Redistributions of source code must retain the above copyright
112  *    notice, this list of conditions and the following disclaimer.
113  * 2. Redistributions in binary form must reproduce the above copyright
114  *    notice, this list of conditions and the following disclaimer in the
115  *    documentation and/or other materials provided with the distribution.
116  * 3. Neither the name of the University nor the names of its contributors
117  *    may be used to endorse or promote products derived from this software
118  *    without specific prior written permission.
119  *
120  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
121  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
122  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
123  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
124  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
125  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
126  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
127  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
128  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
129  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
130  * SUCH DAMAGE.
131  *
132  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
133  */
134 
135 #ifndef _NETINET_TCP_VAR_H_
136 #define _NETINET_TCP_VAR_H_
137 
138 #if defined(_KERNEL_OPT)
139 #include "opt_inet.h"
140 #include "opt_mbuftrace.h"
141 
142 #endif
143 
144 /*
145  * Kernel variables for tcp.
146  */
147 
148 #include <sys/callout.h>
149 
150 #ifdef TCP_SIGNATURE
151 /*
152  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
153  * for SADB verification and lookup.
154  */
155 #define	TCP_SIGLEN	16	/* length of computed digest in bytes */
156 #define	TCP_KEYLEN_MIN	1	/* minimum length of TCP-MD5 key */
157 #define	TCP_KEYLEN_MAX	80	/* maximum length of TCP-MD5 key */
158 /*
159  * Only a single SA per host may be specified at this time. An SPI is
160  * needed in order for the KEY_ALLOCSA() lookup to work.
161  */
162 #define	TCP_SIG_SPI	0x1000
163 #endif /* TCP_SIGNATURE */
164 
165 /*
166  * SACK option block.
167  */
168 struct sackblk {
169 	tcp_seq left;		/* Left edge of sack block. */
170 	tcp_seq right;		/* Right edge of sack block. */
171 };
172 
173 TAILQ_HEAD(sackhead, sackhole);
174 struct sackhole {
175 	tcp_seq start;
176 	tcp_seq end;
177 	tcp_seq rxmit;
178 
179 	TAILQ_ENTRY(sackhole) sackhole_q;
180 };
181 
182 /*
183  * Tcp control block, one per tcp; fields:
184  */
185 struct tcpcb {
186 	int	t_family;		/* address family on the wire */
187 	struct ipqehead segq;		/* sequencing queue */
188 	int	t_segqlen;		/* length of the above */
189 	callout_t t_timer[TCPT_NTIMERS];/* tcp timers */
190 	short	t_state;		/* state of this connection */
191 	short	t_rxtshift;		/* log(2) of rexmt exp. backoff */
192 	uint32_t t_rxtcur;		/* current retransmit value */
193 	short	t_dupacks;		/* consecutive dup acks recd */
194 	/*
195 	 * t_partialacks:
196 	 *	<0	not in fast recovery.
197 	 *	==0	in fast recovery.  has not received partial acks
198 	 *	>0	in fast recovery.  has received partial acks
199 	 */
200 	short	t_partialacks;		/* partials acks during fast rexmit */
201 	u_short	t_peermss;		/* peer's maximum segment size */
202 	u_short	t_ourmss;		/* our's maximum segment size */
203 	u_short t_segsz;		/* current segment size in use */
204 	char	t_force;		/* 1 if forcing out a byte */
205 	u_int	t_flags;
206 #define	TF_ACKNOW	0x0001		/* ack peer immediately */
207 #define	TF_DELACK	0x0002		/* ack, but try to delay it */
208 #define	TF_NODELAY	0x0004		/* don't delay packets to coalesce */
209 #define	TF_NOOPT	0x0008		/* don't use tcp options */
210 #define	TF_REQ_SCALE	0x0020		/* have/will request window scaling */
211 #define	TF_RCVD_SCALE	0x0040		/* other side has requested scaling */
212 #define	TF_REQ_TSTMP	0x0080		/* have/will request timestamps */
213 #define	TF_RCVD_TSTMP	0x0100		/* a timestamp was received in SYN */
214 #define	TF_SACK_PERMIT	0x0200		/* other side said I could SACK */
215 #define	TF_SYN_REXMT	0x0400		/* rexmit timer fired on SYN */
216 #define	TF_WILL_SACK	0x0800		/* try to use SACK */
217 #define	TF_REASSEMBLING	0x1000		/* we're busy reassembling */
218 #define	TF_DEAD		0x2000		/* dead and to-be-released */
219 #define	TF_PMTUD_PEND	0x4000		/* Path MTU Discovery pending */
220 #define	TF_ECN_PERMIT	0x10000		/* other side said is ECN-ready */
221 #define	TF_ECN_SND_CWR	0x20000		/* ECN CWR in queue */
222 #define	TF_ECN_SND_ECE	0x40000		/* ECN ECE in queue */
223 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
224 
225 
226 	struct	mbuf *t_template;	/* skeletal packet for transmit */
227 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
228 	struct	in6pcb *t_in6pcb;	/* back pointer to internet pcb */
229 	callout_t t_delack_ch;		/* delayed ACK callout */
230 /*
231  * The following fields are used as in the protocol specification.
232  * See RFC793, Dec. 1981, page 21.
233  */
234 /* send sequence variables */
235 	tcp_seq	snd_una;		/* send unacknowledged */
236 	tcp_seq	snd_nxt;		/* send next */
237 	tcp_seq	snd_up;			/* send urgent pointer */
238 	tcp_seq	snd_wl1;		/* window update seg seq number */
239 	tcp_seq	snd_wl2;		/* window update seg ack number */
240 	tcp_seq	iss;			/* initial send sequence number */
241 	u_long	snd_wnd;		/* send window */
242 /*
243  * snd_recover
244  * 	it's basically same as the "recover" variable in RFC 2852 (NewReno).
245  * 	when entering fast retransmit, it's set to snd_max.
246  * 	newreno uses this to detect partial ack.
247  * snd_high
248  * 	it's basically same as the "send_high" variable in RFC 2852 (NewReno).
249  * 	on each RTO, it's set to snd_max.
250  * 	newreno uses this to avoid false fast retransmits.
251  */
252 	tcp_seq snd_recover;
253 	tcp_seq	snd_high;
254 /* receive sequence variables */
255 	u_long	rcv_wnd;		/* receive window */
256 	tcp_seq	rcv_nxt;		/* receive next */
257 	tcp_seq	rcv_up;			/* receive urgent pointer */
258 	tcp_seq	irs;			/* initial receive sequence number */
259 /*
260  * Additional variables for this implementation.
261  */
262 /* receive variables */
263 	tcp_seq	rcv_adv;		/* advertised window */
264 
265 /*
266  * retransmit variables
267  *
268  * snd_max
269  * 	the highest sequence number we've ever sent.
270  *	used to recognize retransmits.
271  */
272 	tcp_seq	snd_max;
273 
274 /* congestion control (for slow start, source quench, retransmit after loss) */
275 	u_long	snd_cwnd;		/* congestion-controlled window */
276 	u_long	snd_ssthresh;		/* snd_cwnd size threshhold for
277 					 * for slow start exponential to
278 					 * linear switch
279 					 */
280 /* auto-sizing variables */
281 	u_int rfbuf_cnt;		/* recv buffer autoscaling byte count */
282 	uint32_t rfbuf_ts;		/* recv buffer autoscaling timestamp */
283 
284 /*
285  * transmit timing stuff.  See below for scale of srtt and rttvar.
286  * "Variance" is actually smoothed difference.
287  */
288 	uint32_t t_rcvtime;		/* time last segment received */
289 	uint32_t t_rtttime;		/* time we started measuring rtt */
290 	tcp_seq	t_rtseq;		/* sequence number being timed */
291 	int32_t	t_srtt;			/* smoothed round-trip time */
292 	int32_t	t_rttvar;		/* variance in round-trip time */
293 	uint32_t t_rttmin;		/* minimum rtt allowed */
294 	u_long	max_sndwnd;		/* largest window peer has offered */
295 
296 /* out-of-band data */
297 	char	t_oobflags;		/* have some */
298 	char	t_iobc;			/* input character */
299 #define	TCPOOB_HAVEDATA	0x01
300 #define	TCPOOB_HADDATA	0x02
301 	short	t_softerror;		/* possible error not yet reported */
302 
303 /* RFC 1323 variables */
304 	u_char	snd_scale;		/* window scaling for send window */
305 	u_char	rcv_scale;		/* window scaling for recv window */
306 	u_char	request_r_scale;	/* pending window scaling */
307 	u_char	requested_s_scale;
308 	u_int32_t ts_recent;		/* timestamp echo data */
309 	u_int32_t ts_recent_age;	/* when last updated */
310 	u_int32_t ts_timebase;		/* our timebase */
311 	tcp_seq	last_ack_sent;
312 
313 /* RFC 3465 variables */
314 	u_long	t_bytes_acked;		/* ABC "bytes_acked" parameter */
315 
316 /* SACK stuff */
317 #define TCP_SACK_MAX 3
318 #define TCPSACK_NONE 0
319 #define TCPSACK_HAVED 1
320 	u_char rcv_sack_flags;		/* SACK flags. */
321 	struct sackblk rcv_dsack_block;	/* RX D-SACK block. */
322 	struct ipqehead timeq;		/* time sequenced queue. */
323 	struct sackhead snd_holes;	/* TX SACK holes. */
324 	int	snd_numholes;		/* Number of TX SACK holes. */
325 	tcp_seq rcv_lastsack;		/* last seq number(+1) sack'd by rcv'r*/
326 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
327 					   episode starts at this seq number*/
328 	tcp_seq snd_fack;		/* FACK TCP.  Forward-most data held by
329 					   peer. */
330 
331 /* pointer for syn cache entries*/
332 	LIST_HEAD(, syn_cache) t_sc;	/* list of entries by this tcb */
333 
334 /* prediction of next mbuf when using large window sizes */
335 	struct	mbuf *t_lastm;		/* last mbuf that data was sent from */
336 	int	t_inoff;		/* data offset in previous mbuf */
337 	int	t_lastoff;		/* last data address in mbuf chain */
338 	int	t_lastlen;		/* last length read from mbuf chain */
339 
340 /* Path-MTU discovery blackhole detection */
341 	int t_mtudisc;			/* perform mtudisc for this tcb */
342 /* Path-MTU Discovery Information */
343 	u_int	t_pmtud_mss_acked;	/* MSS acked, lower bound for MTU */
344 	u_int	t_pmtud_mtu_sent;	/* MTU used, upper bound for MTU */
345 	tcp_seq	t_pmtud_th_seq;		/* TCP SEQ from ICMP payload */
346 	u_int	t_pmtud_nextmtu;	/* Advertised Next-Hop MTU from ICMP */
347 	u_short	t_pmtud_ip_len;		/* IP length from ICMP payload */
348 	u_short	t_pmtud_ip_hl;		/* IP header length from ICMP payload */
349 
350 	uint8_t t_ecn_retries;		/* # of ECN setup retries */
351 
352 	const struct tcp_congctl *t_congctl;	/* per TCB congctl algorithm */
353 
354 	/* Keepalive per socket */
355 	u_int	t_keepinit;
356 	u_int	t_keepidle;
357 	u_int	t_keepintvl;
358 	u_int	t_keepcnt;
359 	u_int	t_maxidle;		/* t_keepcnt * t_keepintvl */
360 
361 	u_int	t_msl;			/* MSL to use for this connexion */
362 };
363 
364 /*
365  * Macros to aid ECN TCP.
366  */
367 #define TCP_ECN_ALLOWED(tp)	(tp->t_flags & TF_ECN_PERMIT)
368 
369 /*
370  * Macros to aid SACK/FACK TCP.
371  */
372 #define TCP_SACK_ENABLED(tp)	(tp->t_flags & TF_WILL_SACK)
373 #define TCP_FACK_FASTRECOV(tp)	\
374 	(TCP_SACK_ENABLED(tp) && \
375 	(SEQ_GT(tp->snd_fack, tp->snd_una + tcprexmtthresh * tp->t_segsz)))
376 
377 #ifdef _KERNEL
378 /*
379  * TCP reassembly queue locks.
380  */
381 static __inline int tcp_reass_lock_try (struct tcpcb *)
382 	__unused;
383 static __inline void tcp_reass_unlock (struct tcpcb *)
384 	__unused;
385 
386 static __inline int
387 tcp_reass_lock_try(struct tcpcb *tp)
388 {
389 	int s;
390 
391 	/*
392 	 * Use splvm() -- we're blocking things that would cause
393 	 * mbuf allocation.
394 	 */
395 	s = splvm();
396 	if (tp->t_flags & TF_REASSEMBLING) {
397 		splx(s);
398 		return (0);
399 	}
400 	tp->t_flags |= TF_REASSEMBLING;
401 	splx(s);
402 	return (1);
403 }
404 
405 static __inline void
406 tcp_reass_unlock(struct tcpcb *tp)
407 {
408 	int s;
409 
410 	s = splvm();
411 	KASSERT((tp->t_flags & TF_REASSEMBLING) != 0);
412 	tp->t_flags &= ~TF_REASSEMBLING;
413 	splx(s);
414 }
415 
416 #ifdef DIAGNOSTIC
417 #define	TCP_REASS_LOCK(tp)						\
418 do {									\
419 	if (tcp_reass_lock_try(tp) == 0) {				\
420 		printf("%s:%d: tcpcb %p reass already locked\n",	\
421 		    __FILE__, __LINE__, tp);				\
422 		panic("tcp_reass_lock");				\
423 	}								\
424 } while (/*CONSTCOND*/ 0)
425 #define	TCP_REASS_LOCK_CHECK(tp)					\
426 do {									\
427 	if (((tp)->t_flags & TF_REASSEMBLING) == 0) {			\
428 		printf("%s:%d: tcpcb %p reass lock not held\n",		\
429 		    __FILE__, __LINE__, tp);				\
430 		panic("tcp reass lock check");				\
431 	}								\
432 } while (/*CONSTCOND*/ 0)
433 #else
434 #define	TCP_REASS_LOCK(tp)	(void) tcp_reass_lock_try((tp))
435 #define	TCP_REASS_LOCK_CHECK(tp) /* nothing */
436 #endif
437 
438 #define	TCP_REASS_UNLOCK(tp)	tcp_reass_unlock((tp))
439 #endif /* _KERNEL */
440 
441 /*
442  * Queue for delayed ACK processing.
443  */
444 #ifdef _KERNEL
445 extern int tcp_delack_ticks;
446 void	tcp_delack(void *);
447 
448 #define TCP_RESTART_DELACK(tp)						\
449 	callout_reset(&(tp)->t_delack_ch, tcp_delack_ticks,		\
450 	    tcp_delack, tp)
451 
452 #define	TCP_SET_DELACK(tp)						\
453 do {									\
454 	if (((tp)->t_flags & TF_DELACK) == 0) {				\
455 		(tp)->t_flags |= TF_DELACK;				\
456 		TCP_RESTART_DELACK(tp);					\
457 	}								\
458 } while (/*CONSTCOND*/0)
459 
460 #define	TCP_CLEAR_DELACK(tp)						\
461 do {									\
462 	if ((tp)->t_flags & TF_DELACK) {				\
463 		(tp)->t_flags &= ~TF_DELACK;				\
464 		callout_stop(&(tp)->t_delack_ch);			\
465 	}								\
466 } while (/*CONSTCOND*/0)
467 #endif /* _KERNEL */
468 
469 /*
470  * Compute the current timestamp for a connection.
471  */
472 #define	TCP_TIMESTAMP(tp)	(tcp_now - (tp)->ts_timebase)
473 
474 /*
475  * Handy way of passing around TCP option info.
476  */
477 struct tcp_opt_info {
478 	int		ts_present;
479 	u_int32_t	ts_val;
480 	u_int32_t	ts_ecr;
481 	u_int16_t	maxseg;
482 };
483 
484 #define	TOF_SIGNATURE	0x0040		/* signature option present */
485 #define	TOF_SIGLEN	0x0080		/* sigature length valid (RFC2385) */
486 
487 /*
488  * Data for the TCP compressed state engine.
489  */
490 union syn_cache_sa {
491 	struct sockaddr sa;
492 	struct sockaddr_in sin;
493 #if 1 /*def INET6*/
494 	struct sockaddr_in6 sin6;
495 #endif
496 };
497 
498 struct syn_cache {
499 	TAILQ_ENTRY(syn_cache) sc_bucketq;	/* link on bucket list */
500 	callout_t sc_timer;			/* rexmt timer */
501 	struct route sc_route;
502 	long sc_win;				/* advertised window */
503 	int sc_bucketidx;			/* our bucket index */
504 	u_int32_t sc_hash;
505 	u_int32_t sc_timestamp;			/* timestamp from SYN */
506 	u_int32_t sc_timebase;			/* our local timebase */
507 	union syn_cache_sa sc_src;
508 	union syn_cache_sa sc_dst;
509 	tcp_seq sc_irs;
510 	tcp_seq sc_iss;
511 	u_int sc_rxtcur;			/* current rxt timeout */
512 	u_int sc_rxttot;			/* total time spend on queues */
513 	u_short sc_rxtshift;			/* for computing backoff */
514 	u_short sc_flags;
515 
516 #define	SCF_UNREACH		0x0001		/* we've had an unreach error */
517 #define	SCF_TIMESTAMP		0x0002		/* peer will do timestamps */
518 #define	SCF_DEAD		0x0004		/* this entry to be released */
519 #define SCF_SACK_PERMIT		0x0008		/* peer will do SACK */
520 #define SCF_ECN_PERMIT		0x0010		/* peer will do ECN */
521 #define SCF_SIGNATURE	0x40			/* send MD5 digests */
522 
523 	struct mbuf *sc_ipopts;			/* IP options */
524 	u_int16_t sc_peermaxseg;
525 	u_int16_t sc_ourmaxseg;
526 	u_int8_t sc_request_r_scale	: 4,
527 		 sc_requested_s_scale	: 4;
528 
529 	struct tcpcb *sc_tp;			/* tcb for listening socket */
530 	LIST_ENTRY(syn_cache) sc_tpq;		/* list of entries by same tp */
531 };
532 
533 struct syn_cache_head {
534 	TAILQ_HEAD(, syn_cache) sch_bucket;	/* bucket entries */
535 	u_short sch_length;			/* # entries in bucket */
536 };
537 
538 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
539 #ifdef INET6
540 #define	in6totcpcb(ip)	((struct tcpcb *)(ip)->in6p_ppcb)
541 #endif
542 #ifndef INET6
543 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
544 #else
545 #define	sototcpcb(so)	(((so)->so_proto->pr_domain->dom_family == AF_INET) \
546 				? intotcpcb(sotoinpcb(so)) \
547 				: in6totcpcb(sotoin6pcb(so)))
548 #endif
549 
550 /*
551  * See RFC2988 for a discussion of RTO calculation; comments assume
552  * familiarity with that document.
553  *
554  * The smoothed round-trip time and estimated variance are stored as
555  * fixed point numbers.  Historically, srtt was scaled by
556  * TCP_RTT_SHIFT bits, and rttvar by TCP_RTTVAR_SHIFT bits.  Because
557  * the values coincide with the alpha and beta parameters suggested
558  * for RTO calculation (1/8 for srtt, 1/4 for rttvar), the combination
559  * of computing 1/8 of the new value and transforming it to the
560  * fixed-point representation required zero instructions.  However,
561  * the storage representations no longer coincide with the alpha/beta
562  * shifts; instead, more fractional bits are present.
563  *
564  * The storage representation of srtt is 1/32 slow ticks, or 1/64 s.
565  * (The assumption that a slow tick is 500 ms should not be present in
566  * the code.)
567  *
568  * The storage representation of rttvar is 1/16 slow ticks, or 1/32 s.
569  * There may be some confusion about this in the code.
570  *
571  * For historical reasons, these scales are also used in smoothing the
572  * average (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
573  * This results in alpha of 0.125 and beta of 0.25, following RFC2988
574  * section 2.3
575  *
576  * XXX Change SHIFT values to LGWEIGHT and REP_SHIFT, and adjust
577  * the code to use the correct ones.
578  */
579 #define	TCP_RTT_SHIFT		3	/* shift for srtt; 3 bits frac. */
580 #define	TCP_RTTVAR_SHIFT	2	/* multiplier for rttvar; 2 bits */
581 
582 /*
583  * Compute TCP retransmission timer, following RFC2988.
584  * This macro returns a value in slow timeout ticks.
585  *
586  * Section 2.2 requires that the RTO value be
587  *  srtt + max(G, 4*RTTVAR)
588  * where G is the clock granularity.
589  *
590  * This comment has not necessarily been updated for the new storage
591  * representation:
592  *
593  * Because of the way we do the smoothing, srtt and rttvar
594  * will each average +1/2 tick of bias.  When we compute
595  * the retransmit timer, we want 1/2 tick of rounding and
596  * 1 extra tick because of +-1/2 tick uncertainty in the
597  * firing of the timer.  The bias will give us exactly the
598  * 1.5 tick we need.  But, because the bias is
599  * statistical, we have to test that we don't drop below
600  * the minimum feasible timer (which is 2 ticks).
601  * This macro assumes that the value of 1<<TCP_RTTVAR_SHIFT
602  * is the same as the multiplier for rttvar.
603  *
604  * This macro appears to be wrong; it should be checking rttvar*4 in
605  * ticks and making sure we use 1 instead if rttvar*4 rounds to 0.  It
606  * appears to be treating srtt as being in the old storage
607  * representation, resulting in a factor of 4 extra.
608  */
609 #define	TCP_REXMTVAL(tp) \
610 	((((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar) >> 2)
611 
612 /*
613  * Compute the initial window for slow start.
614  */
615 #define	TCP_INITIAL_WINDOW(iw, segsz) \
616 	(((iw) == 0) ? (min(4 * (segsz), max(2 * (segsz), 4380))) : \
617 	 ((segsz) * (iw)))
618 
619 /*
620  * TCP statistics.
621  * Each counter is an unsigned 64-bit value.
622  *
623  * Many of these should be kept per connection, but that's inconvenient
624  * at the moment.
625  */
626 #define	TCP_STAT_CONNATTEMPT	0	/* connections initiated */
627 #define	TCP_STAT_ACCEPTS	1	/* connections accepted */
628 #define	TCP_STAT_CONNECTS	2	/* connections established */
629 #define	TCP_STAT_DROPS		3	/* connections dropped */
630 #define	TCP_STAT_CONNDROPS	4	/* embryonic connections dropped */
631 #define	TCP_STAT_CLOSED		5	/* conn. closed (includes drops) */
632 #define	TCP_STAT_SEGSTIMED	6	/* segs where we tried to get rtt */
633 #define	TCP_STAT_RTTUPDATED	7	/* times we succeeded */
634 #define	TCP_STAT_DELACK		8	/* delayed ACKs sent */
635 #define	TCP_STAT_TIMEOUTDROP	9	/* conn. dropped in rxmt timeout */
636 #define	TCP_STAT_REXMTTIMEO	10	/* retransmit timeouts */
637 #define	TCP_STAT_PERSISTTIMEO	11	/* persist timeouts */
638 #define	TCP_STAT_KEEPTIMEO	12	/* keepalive timeouts */
639 #define	TCP_STAT_KEEPPROBE	13	/* keepalive probes sent */
640 #define	TCP_STAT_KEEPDROPS	14	/* connections dropped in keepalive */
641 #define	TCP_STAT_PERSISTDROPS	15	/* connections dropped in persist */
642 #define	TCP_STAT_CONNSDRAINED	16	/* connections drained due to memory
643 					   shortage */
644 #define	TCP_STAT_PMTUBLACKHOLE	17	/* PMTUD blackhole detected */
645 #define	TCP_STAT_SNDTOTAL	18	/* total packets sent */
646 #define	TCP_STAT_SNDPACK	19	/* data packlets sent */
647 #define	TCP_STAT_SNDBYTE	20	/* data bytes sent */
648 #define	TCP_STAT_SNDREXMITPACK	21	/* data packets retransmitted */
649 #define	TCP_STAT_SNDREXMITBYTE	22	/* data bytes retransmitted */
650 #define	TCP_STAT_SNDACKS	23	/* ACK-only packets sent */
651 #define	TCP_STAT_SNDPROBE	24	/* window probes sent */
652 #define	TCP_STAT_SNDURG		25	/* packets sent with URG only */
653 #define	TCP_STAT_SNDWINUP	26	/* window update-only packets sent */
654 #define	TCP_STAT_SNDCTRL	27	/* control (SYN|FIN|RST) packets sent */
655 #define	TCP_STAT_RCVTOTAL	28	/* total packets received */
656 #define	TCP_STAT_RCVPACK	29	/* packets received in sequence */
657 #define	TCP_STAT_RCVBYTE	30	/* bytes received in sequence */
658 #define	TCP_STAT_RCVBADSUM	31	/* packets received with cksum errs */
659 #define	TCP_STAT_RCVBADOFF	32	/* packets received with bad offset */
660 #define	TCP_STAT_RCVMEMDROP	33	/* packets dropped for lack of memory */
661 #define	TCP_STAT_RCVSHORT	34	/* packets received too short */
662 #define	TCP_STAT_RCVDUPPACK	35	/* duplicate-only packets received */
663 #define	TCP_STAT_RCVDUPBYTE	36	/* duplicate-only bytes received */
664 #define	TCP_STAT_RCVPARTDUPPACK	37	/* packets with some duplicate data */
665 #define	TCP_STAT_RCVPARTDUPBYTE	38	/* dup. bytes in part-dup. packets */
666 #define	TCP_STAT_RCVOOPACK	39	/* out-of-order packets received */
667 #define	TCP_STAT_RCVOOBYTE	40	/* out-of-order bytes received */
668 #define	TCP_STAT_RCVPACKAFTERWIN 41	/* packets with data after window */
669 #define	TCP_STAT_RCVBYTEAFTERWIN 42	/* bytes received after window */
670 #define	TCP_STAT_RCVAFTERCLOSE	43	/* packets received after "close" */
671 #define	TCP_STAT_RCVWINPROBE	44	/* rcvd window probe packets */
672 #define	TCP_STAT_RCVDUPACK	45	/* rcvd duplicate ACKs */
673 #define	TCP_STAT_RCVACKTOOMUCH	46	/* rcvd ACKs for unsent data */
674 #define	TCP_STAT_RCVACKPACK	47	/* rcvd ACK packets */
675 #define	TCP_STAT_RCVACKBYTE	48	/* bytes ACKed by rcvd ACKs */
676 #define	TCP_STAT_RCVWINUPD	49	/* rcvd window update packets */
677 #define	TCP_STAT_PAWSDROP	50	/* segments dropped due to PAWS */
678 #define	TCP_STAT_PREDACK	51	/* times hdr predict OK for ACKs */
679 #define	TCP_STAT_PREDDAT	52	/* times hdr predict OK for data pkts */
680 #define	TCP_STAT_PCBHASHMISS	53	/* input packets missing PCB hash */
681 #define	TCP_STAT_NOPORT		54	/* no socket on port */
682 #define	TCP_STAT_BADSYN		55	/* received ACK for which we have
683 					   no SYN in compressed state */
684 #define	TCP_STAT_DELAYED_FREE	56	/* delayed pool_put() of tcpcb */
685 #define	TCP_STAT_SC_ADDED	57	/* # of sc entries added */
686 #define	TCP_STAT_SC_COMPLETED	58	/* # of sc connections completed */
687 #define	TCP_STAT_SC_TIMED_OUT	59	/* # of sc entries timed out */
688 #define	TCP_STAT_SC_OVERFLOWED	60	/* # of sc drops due to overflow */
689 #define	TCP_STAT_SC_RESET	61	/* # of sc drops due to RST */
690 #define	TCP_STAT_SC_UNREACH	62	/* # of sc drops due to ICMP unreach */
691 #define	TCP_STAT_SC_BUCKETOVERFLOW 63	/* # of sc drops due to bucket ovflow */
692 #define	TCP_STAT_SC_ABORTED	64	/* # of sc entries aborted (no mem) */
693 #define	TCP_STAT_SC_DUPESYN	65	/* # of duplicate SYNs received */
694 #define	TCP_STAT_SC_DROPPED	66	/* # of SYNs dropped (no route/mem) */
695 #define	TCP_STAT_SC_COLLISIONS	67	/* # of sc hash collisions */
696 #define	TCP_STAT_SC_RETRANSMITTED 68	/* # of sc retransmissions */
697 #define	TCP_STAT_SC_DELAYED_FREE 69	/* # of delayed pool_put()s */
698 #define	TCP_STAT_SELFQUENCH	70	/* # of ENOBUFS we get on output */
699 #define	TCP_STAT_BADSIG		71	/* # of drops due to bad signature */
700 #define	TCP_STAT_GOODSIG	72	/* # of packets with good signature */
701 #define	TCP_STAT_ECN_SHS	73	/* # of successful ECN handshakes */
702 #define	TCP_STAT_ECN_CE		74	/* # of packets with CE bit */
703 #define	TCP_STAT_ECN_ECT	75	/* # of packets with ECT(0) bit */
704 
705 #define	TCP_NSTATS		76
706 
707 /*
708  * Names for TCP sysctl objects.
709  */
710 #define	TCPCTL_RFC1323		1	/* RFC1323 timestamps/scaling */
711 #define	TCPCTL_SENDSPACE	2	/* default send buffer */
712 #define	TCPCTL_RECVSPACE	3	/* default recv buffer */
713 #define	TCPCTL_MSSDFLT		4	/* default seg size */
714 #define	TCPCTL_SYN_CACHE_LIMIT	5	/* max size of comp. state engine */
715 #define	TCPCTL_SYN_BUCKET_LIMIT	6	/* max size of hash bucket */
716 #if 0	/*obsoleted*/
717 #define	TCPCTL_SYN_CACHE_INTER	7	/* interval of comp. state timer */
718 #endif
719 #define	TCPCTL_INIT_WIN		8	/* initial window */
720 #define	TCPCTL_MSS_IFMTU	9	/* mss from interface, not in_maxmtu */
721 #define	TCPCTL_SACK		10	/* RFC2018 selective acknowledgement */
722 #define	TCPCTL_WSCALE		11	/* RFC1323 window scaling */
723 #define	TCPCTL_TSTAMP		12	/* RFC1323 timestamps */
724 #define	TCPCTL_COMPAT_42	13	/* 4.2BSD TCP bug work-arounds */
725 #define	TCPCTL_CWM		14	/* Congestion Window Monitoring */
726 #define	TCPCTL_CWM_BURSTSIZE	15	/* burst size allowed by CWM */
727 #define	TCPCTL_ACK_ON_PUSH	16	/* ACK immediately on PUSH */
728 #define	TCPCTL_KEEPIDLE		17	/* keepalive idle time */
729 #define	TCPCTL_KEEPINTVL	18	/* keepalive probe interval */
730 #define	TCPCTL_KEEPCNT		19	/* keepalive count */
731 #define	TCPCTL_SLOWHZ		20	/* PR_SLOWHZ (read-only) */
732 #define	TCPCTL_NEWRENO		21	/* NewReno Congestion Control */
733 #define TCPCTL_LOG_REFUSED	22	/* Log refused connections */
734 #if 0	/*obsoleted*/
735 #define	TCPCTL_RSTRATELIMIT	23	/* RST rate limit */
736 #endif
737 #define	TCPCTL_RSTPPSLIMIT	24	/* RST pps limit */
738 #define	TCPCTL_DELACK_TICKS	25	/* # ticks to delay ACK */
739 #define	TCPCTL_INIT_WIN_LOCAL	26	/* initial window for local nets */
740 #define	TCPCTL_IDENT		27	/* rfc 931 identd */
741 #define	TCPCTL_ACKDROPRATELIMIT	28	/* SYN/RST -> ACK rate limit */
742 #define	TCPCTL_LOOPBACKCKSUM	29	/* do TCP checksum on loopback */
743 #define	TCPCTL_STATS		30	/* TCP statistics */
744 #define	TCPCTL_DEBUG		31	/* TCP debug sockets */
745 #define	TCPCTL_DEBX		32	/* # of tcp debug sockets */
746 #define	TCPCTL_DROP		33	/* drop tcp connection */
747 #define	TCPCTL_MSL		34	/* Max Segment Life */
748 #define	TCPCTL_MAXID		35
749 
750 #define	TCPCTL_NAMES { \
751 	{ 0, 0 }, \
752 	{ "rfc1323",	CTLTYPE_INT }, \
753 	{ "sendspace",	CTLTYPE_INT }, \
754 	{ "recvspace",	CTLTYPE_INT }, \
755 	{ "mssdflt",	CTLTYPE_INT }, \
756 	{ "syn_cache_limit", CTLTYPE_INT }, \
757 	{ "syn_bucket_limit", CTLTYPE_INT }, \
758 	{ 0, 0 },\
759 	{ "init_win", CTLTYPE_INT }, \
760 	{ "mss_ifmtu", CTLTYPE_INT }, \
761 	{ "sack", CTLTYPE_INT }, \
762 	{ "win_scale", CTLTYPE_INT }, \
763 	{ "timestamps", CTLTYPE_INT }, \
764 	{ "compat_42", CTLTYPE_INT }, \
765 	{ "cwm", CTLTYPE_INT }, \
766 	{ "cwm_burstsize", CTLTYPE_INT }, \
767 	{ "ack_on_push", CTLTYPE_INT }, \
768 	{ "keepidle",	CTLTYPE_INT }, \
769 	{ "keepintvl",	CTLTYPE_INT }, \
770 	{ "keepcnt",	CTLTYPE_INT }, \
771 	{ "slowhz",	CTLTYPE_INT }, \
772 	{ 0, 0 }, \
773 	{ "log_refused",CTLTYPE_INT }, \
774 	{ 0, 0 }, \
775 	{ "rstppslimit", CTLTYPE_INT }, \
776 	{ "delack_ticks", CTLTYPE_INT }, \
777 	{ "init_win_local", CTLTYPE_INT }, \
778 	{ "ident", CTLTYPE_STRUCT }, \
779 	{ "ackdropppslimit", CTLTYPE_INT }, \
780 	{ "do_loopback_cksum", CTLTYPE_INT }, \
781 	{ "stats", CTLTYPE_STRUCT }, \
782 	{ "debug", CTLTYPE_STRUCT }, \
783 	{ "debx", CTLTYPE_INT }, \
784 	{ "drop", CTLTYPE_STRUCT }, \
785 	{ "msl", CTLTYPE_INT }, \
786 }
787 
788 #ifdef _KERNEL
789 extern	struct inpcbtable tcbtable;	/* head of queue of active tcpcb's */
790 extern	u_int32_t tcp_now;	/* for RFC 1323 timestamps */
791 extern	int tcp_do_rfc1323;	/* enabled/disabled? */
792 extern	int tcp_do_sack;	/* SACK enabled/disabled? */
793 extern	int tcp_do_win_scale;	/* RFC1323 window scaling enabled/disabled? */
794 extern	int tcp_do_timestamps;	/* RFC1323 timestamps enabled/disabled? */
795 extern	int tcp_mssdflt;	/* default seg size */
796 extern	int tcp_minmss;		/* minimal seg size */
797 extern  int tcp_msl;		/* max segment life */
798 extern	int tcp_init_win;	/* initial window */
799 extern	int tcp_init_win_local;	/* initial window for local nets */
800 extern	int tcp_mss_ifmtu;	/* take MSS from interface, not in_maxmtu */
801 extern	int tcp_compat_42;	/* work around ancient broken TCP peers */
802 extern	int tcp_cwm;		/* enable Congestion Window Monitoring */
803 extern	int tcp_cwm_burstsize;	/* burst size allowed by CWM */
804 extern	int tcp_ack_on_push;	/* ACK immediately on PUSH */
805 extern	int tcp_syn_cache_limit; /* max entries for compressed state engine */
806 extern	int tcp_syn_bucket_limit;/* max entries per hash bucket */
807 extern	int tcp_log_refused;	/* log refused connections */
808 extern	int tcp_do_ecn;		/* TCP ECN enabled/disabled? */
809 extern	int tcp_ecn_maxretries;	/* Max ECN setup retries */
810 extern	int tcp_do_rfc1948;	/* ISS by cryptographic hash */
811 extern int tcp_sack_tp_maxholes;	/* Max holes per connection. */
812 extern int tcp_sack_globalmaxholes;	/* Max holes per system. */
813 extern int tcp_sack_globalholes;	/* Number of holes present. */
814 extern int tcp_do_abc;			/* RFC3465 ABC enabled/disabled? */
815 extern int tcp_abc_aggressive;		/* 1: L=2*SMSS  0: L=1*SMSS */
816 
817 extern int tcp_msl_enable;		/* enable TIME_WAIT truncation	*/
818 extern int tcp_msl_loop;		/* MSL for loopback		*/
819 extern int tcp_msl_local;		/* MSL for 'local'		*/
820 extern int tcp_msl_remote;		/* MSL otherwise		*/
821 extern int tcp_msl_remote_threshold;	/* RTT threshold		*/
822 extern int tcp_rttlocal;		/* Use RTT to decide who's 'local' */
823 extern int tcp4_vtw_enable;
824 extern int tcp6_vtw_enable;
825 extern int tcp_vtw_was_enabled;
826 extern int tcp_vtw_entries;
827 
828 extern	int tcp_rst_ppslim;
829 extern	int tcp_ackdrop_ppslim;
830 
831 extern	int tcp_syn_cache_size;
832 extern	struct syn_cache_head tcp_syn_cache[];
833 extern	u_long syn_cache_count;
834 
835 #ifdef MBUFTRACE
836 extern	struct mowner tcp_rx_mowner;
837 extern	struct mowner tcp_tx_mowner;
838 extern	struct mowner tcp_reass_mowner;
839 extern	struct mowner tcp_sock_mowner;
840 extern	struct mowner tcp_sock_rx_mowner;
841 extern	struct mowner tcp_sock_tx_mowner;
842 extern	struct mowner tcp_mowner;
843 #endif
844 
845 extern int tcp_do_autorcvbuf;
846 extern int tcp_autorcvbuf_inc;
847 extern int tcp_autorcvbuf_max;
848 extern int tcp_do_autosndbuf;
849 extern int tcp_autosndbuf_inc;
850 extern int tcp_autosndbuf_max;
851 
852 
853 #define	TCPCTL_VARIABLES { \
854 	{ 0 },					\
855 	{ 1, 0, &tcp_do_rfc1323 },		\
856 	{ 1, 0, &tcp_sendspace },		\
857 	{ 1, 0, &tcp_recvspace },		\
858 	{ 1, 0, &tcp_mssdflt },			\
859 	{ 1, 0, &tcp_syn_cache_limit },		\
860 	{ 1, 0, &tcp_syn_bucket_limit },	\
861 	{ 0 },					\
862 	{ 1, 0, &tcp_init_win },		\
863 	{ 1, 0, &tcp_mss_ifmtu },		\
864 	{ 1, 0, &tcp_do_sack },			\
865 	{ 1, 0, &tcp_do_win_scale },		\
866 	{ 1, 0, &tcp_do_timestamps },		\
867 	{ 1, 0, &tcp_compat_42 },		\
868 	{ 1, 0, &tcp_cwm },			\
869 	{ 1, 0, &tcp_cwm_burstsize },		\
870 	{ 1, 0, &tcp_ack_on_push },		\
871 	{ 1, 0, &tcp_keepidle },		\
872 	{ 1, 0, &tcp_keepintvl },		\
873 	{ 1, 0, &tcp_keepcnt },			\
874 	{ 1, 1, 0, PR_SLOWHZ },			\
875 	{ 0 },					\
876 	{ 1, 0, &tcp_log_refused },		\
877 	{ 0 },					\
878 	{ 1, 0, &tcp_rst_ppslim },		\
879 	{ 1, 0, &tcp_delack_ticks },		\
880 	{ 1, 0, &tcp_init_win_local },		\
881 	{ 1, 0, &tcp_ackdrop_ppslim },		\
882 }
883 
884 struct secasvar;
885 
886 int	 tcp_attach(struct socket *);
887 void	 tcp_canceltimers(struct tcpcb *);
888 struct tcpcb *
889 	 tcp_close(struct tcpcb *);
890 int	 tcp_isdead(struct tcpcb *);
891 #ifdef INET6
892 void	 *tcp6_ctlinput(int, const struct sockaddr *, void *);
893 #endif
894 void	 *tcp_ctlinput(int, const struct sockaddr *, void *);
895 int	 tcp_ctloutput(int, struct socket *, struct sockopt *);
896 struct tcpcb *
897 	 tcp_disconnect(struct tcpcb *);
898 struct tcpcb *
899 	 tcp_drop(struct tcpcb *, int);
900 #ifdef TCP_SIGNATURE
901 int	 tcp_signature_apply(void *, void *, u_int);
902 struct secasvar *tcp_signature_getsav(struct mbuf *, struct tcphdr *);
903 int	 tcp_signature(struct mbuf *, struct tcphdr *, int, struct secasvar *,
904 	    char *);
905 #endif
906 void	 tcp_drain(void);
907 void	 tcp_drainstub(void);
908 void	 tcp_established(struct tcpcb *);
909 void	 tcp_init(void);
910 #ifdef INET6
911 int	 tcp6_input(struct mbuf **, int *, int);
912 #endif
913 void	 tcp_input(struct mbuf *, ...);
914 u_int	 tcp_hdrsz(struct tcpcb *);
915 u_long	 tcp_mss_to_advertise(const struct ifnet *, int);
916 void	 tcp_mss_from_peer(struct tcpcb *, int);
917 void	 tcp_tcpcb_template(void);
918 struct tcpcb *
919 	 tcp_newtcpcb(int, void *);
920 void	 tcp_notify(struct inpcb *, int);
921 #ifdef INET6
922 void	 tcp6_notify(struct in6pcb *, int);
923 #endif
924 u_int	 tcp_optlen(struct tcpcb *);
925 int	 tcp_output(struct tcpcb *);
926 void	 tcp_pulloutofband(struct socket *,
927 	    struct tcphdr *, struct mbuf *, int);
928 void	 tcp_quench(struct inpcb *, int);
929 #ifdef INET6
930 void	 tcp6_quench(struct in6pcb *, int);
931 #endif
932 void	 tcp_mtudisc(struct inpcb *, int);
933 
934 void	tcpipqent_init(void);
935 struct ipqent *tcpipqent_alloc(void);
936 void	 tcpipqent_free(struct ipqent *);
937 
938 int	 tcp_respond(struct tcpcb *, struct mbuf *, struct mbuf *,
939 	    struct tcphdr *, tcp_seq, tcp_seq, int);
940 void	 tcp_rmx_rtt(struct tcpcb *);
941 void	 tcp_setpersist(struct tcpcb *);
942 #ifdef TCP_SIGNATURE
943 int	 tcp_signature_compute(struct mbuf *, struct tcphdr *, int, int,
944 	    int, u_char *, u_int);
945 #endif
946 void	 tcp_slowtimo(void);
947 void	 tcp_fasttimo(void);
948 struct mbuf *
949 	 tcp_template(struct tcpcb *);
950 void	 tcp_trace(short, short, struct tcpcb *, struct mbuf *, int);
951 struct tcpcb *
952 	 tcp_usrclosed(struct tcpcb *);
953 void	 tcp_usrreq_init(void);
954 int	 tcp_usrreq(struct socket *,
955 	    int, struct mbuf *, struct mbuf *, struct mbuf *, struct lwp *);
956 void	 tcp_xmit_timer(struct tcpcb *, uint32_t);
957 tcp_seq	 tcp_new_iss(struct tcpcb *, tcp_seq);
958 tcp_seq  tcp_new_iss1(void *, void *, u_int16_t, u_int16_t, size_t,
959 	    tcp_seq);
960 
961 void	 tcp_sack_init(void);
962 void	 tcp_new_dsack(struct tcpcb *, tcp_seq, u_int32_t);
963 void	 tcp_sack_option(struct tcpcb *, const struct tcphdr *,
964 	    const u_char *, int);
965 void	 tcp_del_sackholes(struct tcpcb *, const struct tcphdr *);
966 void	 tcp_free_sackholes(struct tcpcb *);
967 void	 tcp_sack_adjust(struct tcpcb *tp);
968 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
969 void	 tcp_sack_newack(struct tcpcb *, const struct tcphdr *);
970 int	 tcp_sack_numblks(const struct tcpcb *);
971 #define	TCP_SACK_OPTLEN(nblks)	((nblks) * 8 + 2 + 2)
972 
973 void	 tcp_statinc(u_int);
974 void	 tcp_statadd(u_int, uint64_t);
975 
976 int	 syn_cache_add(struct sockaddr *, struct sockaddr *,
977 		struct tcphdr *, unsigned int, struct socket *,
978 		struct mbuf *, u_char *, int, struct tcp_opt_info *);
979 void	 syn_cache_unreach(const struct sockaddr *, const struct sockaddr *,
980 	   struct tcphdr *);
981 struct socket *syn_cache_get(struct sockaddr *, struct sockaddr *,
982 		struct tcphdr *, unsigned int, unsigned int,
983 		struct socket *so, struct mbuf *);
984 void	 syn_cache_init(void);
985 void	 syn_cache_insert(struct syn_cache *, struct tcpcb *);
986 struct syn_cache *syn_cache_lookup(const struct sockaddr *, const struct sockaddr *,
987 		struct syn_cache_head **);
988 void	 syn_cache_reset(struct sockaddr *, struct sockaddr *,
989 		struct tcphdr *);
990 int	 syn_cache_respond(struct syn_cache *, struct mbuf *);
991 void	 syn_cache_timer(void *);
992 void	 syn_cache_cleanup(struct tcpcb *);
993 
994 int	 tcp_input_checksum(int, struct mbuf *, const struct tcphdr *, int, int,
995     int);
996 #endif
997 
998 #endif /* !_NETINET_TCP_VAR_H_ */
999