xref: /netbsd-src/sys/netinet/tcp_var.h (revision aef5eb5f59cdfe8314f1b5f78ac04eb144e44010)
1 /*	$NetBSD: tcp_var.h,v 1.197 2022/09/20 07:19:14 ozaki-r 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  * TCP kernel structures and variables.
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_LOOKUP_SA() lookup to work.
161  */
162 #define	TCP_SIG_SPI	0x1000
163 #endif /* TCP_SIGNATURE */
164 
165 /*
166  * Tcp+ip header, after ip options removed.
167  */
168 struct tcpiphdr {
169 	struct ipovly ti_i;		/* overlaid ip structure */
170 	struct tcphdr ti_t;		/* tcp header */
171 };
172 #ifdef CTASSERT
173 CTASSERT(sizeof(struct tcpiphdr) == 40);
174 #endif
175 #define	ti_x1		ti_i.ih_x1
176 #define	ti_pr		ti_i.ih_pr
177 #define	ti_len		ti_i.ih_len
178 #define	ti_src		ti_i.ih_src
179 #define	ti_dst		ti_i.ih_dst
180 #define	ti_sport	ti_t.th_sport
181 #define	ti_dport	ti_t.th_dport
182 #define	ti_seq		ti_t.th_seq
183 #define	ti_ack		ti_t.th_ack
184 #define	ti_x2		ti_t.th_x2
185 #define	ti_off		ti_t.th_off
186 #define	ti_flags	ti_t.th_flags
187 #define	ti_win		ti_t.th_win
188 #define	ti_sum		ti_t.th_sum
189 #define	ti_urp		ti_t.th_urp
190 
191 /*
192  * SACK option block.
193  */
194 struct sackblk {
195 	tcp_seq left;		/* Left edge of sack block. */
196 	tcp_seq right;		/* Right edge of sack block. */
197 };
198 
199 TAILQ_HEAD(sackhead, sackhole);
200 struct sackhole {
201 	tcp_seq start;
202 	tcp_seq end;
203 	tcp_seq rxmit;
204 
205 	TAILQ_ENTRY(sackhole) sackhole_q;
206 };
207 
208 struct syn_cache;
209 
210 /*
211  * Tcp control block, one per tcp; fields:
212  */
213 struct tcpcb {
214 	int	t_family;		/* address family on the wire */
215 	struct ipqehead segq;		/* sequencing queue */
216 	int	t_segqlen;		/* length of the above */
217 	callout_t t_timer[TCPT_NTIMERS];/* tcp timers */
218 	short	t_state;		/* state of this connection */
219 	short	t_rxtshift;		/* log(2) of rexmt exp. backoff */
220 	uint32_t t_rxtcur;		/* current retransmit value */
221 	short	t_dupacks;		/* consecutive dup acks recd */
222 	/*
223 	 * t_partialacks:
224 	 *	<0	not in fast recovery.
225 	 *	==0	in fast recovery.  has not received partial acks
226 	 *	>0	in fast recovery.  has received partial acks
227 	 */
228 	short	t_partialacks;		/* partials acks during fast rexmit */
229 	u_short	t_peermss;		/* peer's maximum segment size */
230 	u_short	t_ourmss;		/* our's maximum segment size */
231 	u_short t_segsz;		/* current segment size in use */
232 	char	t_force;		/* 1 if forcing out a byte */
233 	u_int	t_flags;
234 #define	TF_ACKNOW	0x0001		/* ack peer immediately */
235 #define	TF_DELACK	0x0002		/* ack, but try to delay it */
236 #define	TF_NODELAY	0x0004		/* don't delay packets to coalesce */
237 #define	TF_NOOPT	0x0008		/* don't use tcp options */
238 #define	TF_REQ_SCALE	0x0020		/* have/will request window scaling */
239 #define	TF_RCVD_SCALE	0x0040		/* other side has requested scaling */
240 #define	TF_REQ_TSTMP	0x0080		/* have/will request timestamps */
241 #define	TF_RCVD_TSTMP	0x0100		/* a timestamp was received in SYN */
242 #define	TF_SACK_PERMIT	0x0200		/* other side said I could SACK */
243 #define	TF_SYN_REXMT	0x0400		/* rexmit timer fired on SYN */
244 #define	TF_WILL_SACK	0x0800		/* try to use SACK */
245 #define	TF_REASSEMBLING	0x1000		/* we're busy reassembling */
246 #define	TF_DEAD		0x2000		/* dead and to-be-released */
247 #define	TF_PMTUD_PEND	0x4000		/* Path MTU Discovery pending */
248 #define	TF_ECN_PERMIT	0x10000		/* other side said is ECN-ready */
249 #define	TF_ECN_SND_CWR	0x20000		/* ECN CWR in queue */
250 #define	TF_ECN_SND_ECE	0x40000		/* ECN ECE in queue */
251 #define	TF_SIGNATURE	0x400000	/* require MD5 digests (RFC2385) */
252 
253 
254 	struct	mbuf *t_template;	/* skeletal packet for transmit */
255 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
256 	struct	in6pcb *t_in6pcb;	/* back pointer to internet pcb */
257 	callout_t t_delack_ch;		/* delayed ACK callout */
258 /*
259  * The following fields are used as in the protocol specification.
260  * See RFC793, Dec. 1981, page 21.
261  */
262 /* send sequence variables */
263 	tcp_seq	snd_una;		/* send unacknowledged */
264 	tcp_seq	snd_nxt;		/* send next */
265 	tcp_seq	snd_up;			/* send urgent pointer */
266 	tcp_seq	snd_wl1;		/* window update seg seq number */
267 	tcp_seq	snd_wl2;		/* window update seg ack number */
268 	tcp_seq	iss;			/* initial send sequence number */
269 	u_long	snd_wnd;		/* send window */
270 /*
271  * snd_recover
272  * 	it's basically same as the "recover" variable in RFC 2852 (NewReno).
273  * 	when entering fast retransmit, it's set to snd_max.
274  * 	newreno uses this to detect partial ack.
275  * snd_high
276  * 	it's basically same as the "send_high" variable in RFC 2852 (NewReno).
277  * 	on each RTO, it's set to snd_max.
278  * 	newreno uses this to avoid false fast retransmits.
279  */
280 	tcp_seq snd_recover;
281 	tcp_seq	snd_high;
282 /* receive sequence variables */
283 	u_long	rcv_wnd;		/* receive window */
284 	tcp_seq	rcv_nxt;		/* receive next */
285 	tcp_seq	rcv_up;			/* receive urgent pointer */
286 	tcp_seq	irs;			/* initial receive sequence number */
287 /*
288  * Additional variables for this implementation.
289  */
290 /* receive variables */
291 	tcp_seq	rcv_adv;		/* advertised window */
292 
293 /*
294  * retransmit variables
295  *
296  * snd_max
297  * 	the highest sequence number we've ever sent.
298  *	used to recognize retransmits.
299  */
300 	tcp_seq	snd_max;
301 
302 /* congestion control (for slow start, source quench, retransmit after loss) */
303 	u_long	snd_cwnd;		/* congestion-controlled window */
304 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
305 					 * for slow start exponential to
306 					 * linear switch
307 					 */
308 /* auto-sizing variables */
309 	u_int rfbuf_cnt;		/* recv buffer autoscaling byte count */
310 	uint32_t rfbuf_ts;		/* recv buffer autoscaling timestamp */
311 
312 /*
313  * transmit timing stuff.  See below for scale of srtt and rttvar.
314  * "Variance" is actually smoothed difference.
315  */
316 	uint32_t t_rcvtime;		/* time last segment received */
317 	uint32_t t_rtttime;		/* time we started measuring rtt */
318 	tcp_seq	t_rtseq;		/* sequence number being timed */
319 	int32_t	t_srtt;			/* smoothed round-trip time */
320 	int32_t	t_rttvar;		/* variance in round-trip time */
321 	uint32_t t_rttmin;		/* minimum rtt allowed */
322 	u_long	max_sndwnd;		/* largest window peer has offered */
323 
324 /* out-of-band data */
325 	char	t_oobflags;		/* have some */
326 	char	t_iobc;			/* input character */
327 #define	TCPOOB_HAVEDATA	0x01
328 #define	TCPOOB_HADDATA	0x02
329 	short	t_softerror;		/* possible error not yet reported */
330 
331 /* RFC 1323 variables */
332 	u_char	snd_scale;		/* window scaling for send window */
333 	u_char	rcv_scale;		/* window scaling for recv window */
334 	u_char	request_r_scale;	/* pending window scaling */
335 	u_char	requested_s_scale;
336 	u_int32_t ts_recent;		/* timestamp echo data */
337 	u_int32_t ts_recent_age;	/* when last updated */
338 	u_int32_t ts_timebase;		/* our timebase */
339 	tcp_seq	last_ack_sent;
340 
341 /* RFC 3465 variables */
342 	u_long	t_bytes_acked;		/* ABC "bytes_acked" parameter */
343 
344 /* SACK stuff */
345 #define TCP_SACK_MAX 3
346 #define TCPSACK_NONE 0
347 #define TCPSACK_HAVED 1
348 	u_char rcv_sack_flags;		/* SACK flags. */
349 	struct sackblk rcv_dsack_block;	/* RX D-SACK block. */
350 	struct ipqehead timeq;		/* time sequenced queue. */
351 	struct sackhead snd_holes;	/* TX SACK holes. */
352 	int	snd_numholes;		/* Number of TX SACK holes. */
353 	tcp_seq rcv_lastsack;		/* last seq number(+1) sack'd by rcv'r*/
354 	tcp_seq sack_newdata;		/* New data xmitted in this recovery
355 					   episode starts at this seq number*/
356 	tcp_seq snd_fack;		/* FACK TCP.  Forward-most data held by
357 					   peer. */
358 
359 /* CUBIC variables */
360 	ulong snd_cubic_wmax;		/* W_max */
361 	ulong snd_cubic_wmax_last;	/* Used for fast convergence */
362 	ulong snd_cubic_ctime;		/* Last congestion time */
363 
364 /* pointer for syn cache entries*/
365 	LIST_HEAD(, syn_cache) t_sc;	/* list of entries by this tcb */
366 
367 /* prediction of next mbuf when using large window sizes */
368 	struct	mbuf *t_lastm;		/* last mbuf that data was sent from */
369 	int	t_inoff;		/* data offset in previous mbuf */
370 	int	t_lastoff;		/* last data address in mbuf chain */
371 	int	t_lastlen;		/* last length read from mbuf chain */
372 
373 /* Path-MTU discovery blackhole detection */
374 	int t_mtudisc;			/* perform mtudisc for this tcb */
375 /* Path-MTU Discovery Information */
376 	u_int	t_pmtud_mss_acked;	/* MSS acked, lower bound for MTU */
377 	u_int	t_pmtud_mtu_sent;	/* MTU used, upper bound for MTU */
378 	tcp_seq	t_pmtud_th_seq;		/* TCP SEQ from ICMP payload */
379 	u_int	t_pmtud_nextmtu;	/* Advertised Next-Hop MTU from ICMP */
380 	u_short	t_pmtud_ip_len;		/* IP length from ICMP payload */
381 	u_short	t_pmtud_ip_hl;		/* IP header length from ICMP payload */
382 
383 	uint8_t t_ecn_retries;		/* # of ECN setup retries */
384 
385 	const struct tcp_congctl *t_congctl;	/* per TCB congctl algorithm */
386 
387 	/* Keepalive per socket */
388 	u_int	t_keepinit;
389 	u_int	t_keepidle;
390 	u_int	t_keepintvl;
391 	u_int	t_keepcnt;
392 	u_int	t_maxidle;		/* t_keepcnt * t_keepintvl */
393 
394 	u_int	t_msl;			/* MSL to use for this connexion */
395 
396 	/* maintain a few stats per connection: */
397 	uint32_t t_rcvoopack;	 	/* out-of-order packets received */
398 	uint32_t t_sndrexmitpack; 	/* retransmit packets sent */
399 	uint32_t t_sndzerowin;		/* zero-window updates sent */
400 };
401 
402 /*
403  * Macros to aid ECN TCP.
404  */
405 #define TCP_ECN_ALLOWED(tp)	(tp->t_flags & TF_ECN_PERMIT)
406 
407 /*
408  * Macros to aid SACK/FACK TCP.
409  */
410 #define TCP_SACK_ENABLED(tp)	(tp->t_flags & TF_WILL_SACK)
411 #define TCP_FACK_FASTRECOV(tp)	\
412 	(TCP_SACK_ENABLED(tp) && \
413 	(SEQ_GT(tp->snd_fack, tp->snd_una + tcprexmtthresh * tp->t_segsz)))
414 
415 #ifdef _KERNEL
416 /*
417  * TCP reassembly queue locks.
418  */
419 static __inline int tcp_reass_lock_try (struct tcpcb *)
420 	__unused;
421 static __inline void tcp_reass_unlock (struct tcpcb *)
422 	__unused;
423 
424 static __inline int
425 tcp_reass_lock_try(struct tcpcb *tp)
426 {
427 	int s;
428 
429 	/*
430 	 * Use splvm() -- we're blocking things that would cause
431 	 * mbuf allocation.
432 	 */
433 	s = splvm();
434 	if (tp->t_flags & TF_REASSEMBLING) {
435 		splx(s);
436 		return (0);
437 	}
438 	tp->t_flags |= TF_REASSEMBLING;
439 	splx(s);
440 	return (1);
441 }
442 
443 static __inline void
444 tcp_reass_unlock(struct tcpcb *tp)
445 {
446 	int s;
447 
448 	s = splvm();
449 	KASSERT((tp->t_flags & TF_REASSEMBLING) != 0);
450 	tp->t_flags &= ~TF_REASSEMBLING;
451 	splx(s);
452 }
453 
454 #ifdef DIAGNOSTIC
455 #define	TCP_REASS_LOCK(tp)						\
456 do {									\
457 	if (tcp_reass_lock_try(tp) == 0) {				\
458 		printf("%s:%d: tcpcb %p reass already locked\n",	\
459 		    __FILE__, __LINE__, tp);				\
460 		panic("tcp_reass_lock");				\
461 	}								\
462 } while (/*CONSTCOND*/ 0)
463 #define	TCP_REASS_LOCK_CHECK(tp)					\
464 do {									\
465 	if (((tp)->t_flags & TF_REASSEMBLING) == 0) {			\
466 		printf("%s:%d: tcpcb %p reass lock not held\n",		\
467 		    __FILE__, __LINE__, tp);				\
468 		panic("tcp reass lock check");				\
469 	}								\
470 } while (/*CONSTCOND*/ 0)
471 #else
472 #define	TCP_REASS_LOCK(tp)	(void) tcp_reass_lock_try((tp))
473 #define	TCP_REASS_LOCK_CHECK(tp) /* nothing */
474 #endif
475 
476 #define	TCP_REASS_UNLOCK(tp)	tcp_reass_unlock((tp))
477 #endif /* _KERNEL */
478 
479 /*
480  * Queue for delayed ACK processing.
481  */
482 #ifdef _KERNEL
483 extern int tcp_delack_ticks;
484 void	tcp_delack(void *);
485 
486 #define TCP_RESTART_DELACK(tp)						\
487 	callout_reset(&(tp)->t_delack_ch, tcp_delack_ticks,		\
488 	    tcp_delack, tp)
489 
490 #define	TCP_SET_DELACK(tp)						\
491 do {									\
492 	if (((tp)->t_flags & TF_DELACK) == 0) {				\
493 		(tp)->t_flags |= TF_DELACK;				\
494 		TCP_RESTART_DELACK(tp);					\
495 	}								\
496 } while (/*CONSTCOND*/0)
497 
498 #define	TCP_CLEAR_DELACK(tp)						\
499 do {									\
500 	if ((tp)->t_flags & TF_DELACK) {				\
501 		(tp)->t_flags &= ~TF_DELACK;				\
502 		callout_stop(&(tp)->t_delack_ch);			\
503 	}								\
504 } while (/*CONSTCOND*/0)
505 #endif /* _KERNEL */
506 
507 /*
508  * Compute the current timestamp for a connection.
509  */
510 #define	TCP_TIMESTAMP(tp)	(tcp_now - (tp)->ts_timebase)
511 
512 /*
513  * Handy way of passing around TCP option info.
514  */
515 struct tcp_opt_info {
516 	int		ts_present;
517 	u_int32_t	ts_val;
518 	u_int32_t	ts_ecr;
519 	u_int16_t	maxseg;
520 };
521 
522 #define	TOF_SIGNATURE	0x0040		/* signature option present */
523 #define	TOF_SIGLEN	0x0080		/* sigature length valid (RFC2385) */
524 
525 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
526 #ifdef INET6
527 #define	in6totcpcb(ip)	((struct tcpcb *)(ip)->in6p_ppcb)
528 #endif
529 #ifndef INET6
530 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
531 #else
532 #define	sototcpcb(so)	(((so)->so_proto->pr_domain->dom_family == AF_INET) \
533 				? intotcpcb(sotoinpcb(so)) \
534 				: in6totcpcb(sotoin6pcb(so)))
535 #endif
536 
537 /*
538  * See RFC2988 for a discussion of RTO calculation; comments assume
539  * familiarity with that document.
540  *
541  * The smoothed round-trip time and estimated variance are stored as
542  * fixed point numbers.  Historically, srtt was scaled by
543  * TCP_RTT_SHIFT bits, and rttvar by TCP_RTTVAR_SHIFT bits.  Because
544  * the values coincide with the alpha and beta parameters suggested
545  * for RTO calculation (1/8 for srtt, 1/4 for rttvar), the combination
546  * of computing 1/8 of the new value and transforming it to the
547  * fixed-point representation required zero instructions.  However,
548  * the storage representations no longer coincide with the alpha/beta
549  * shifts; instead, more fractional bits are present.
550  *
551  * The storage representation of srtt is 1/32 slow ticks, or 1/64 s.
552  * (The assumption that a slow tick is 500 ms should not be present in
553  * the code.)
554  *
555  * The storage representation of rttvar is 1/16 slow ticks, or 1/32 s.
556  * There may be some confusion about this in the code.
557  *
558  * For historical reasons, these scales are also used in smoothing the
559  * average (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
560  * This results in alpha of 0.125 and beta of 0.25, following RFC2988
561  * section 2.3
562  *
563  * XXX Change SHIFT values to LGWEIGHT and REP_SHIFT, and adjust
564  * the code to use the correct ones.
565  */
566 #define	TCP_RTT_SHIFT		3	/* shift for srtt; 3 bits frac. */
567 #define	TCP_RTTVAR_SHIFT	2	/* multiplier for rttvar; 2 bits */
568 
569 /*
570  * Compute TCP retransmission timer, following RFC2988.
571  * This macro returns a value in slow timeout ticks.
572  *
573  * Section 2.2 requires that the RTO value be
574  *  srtt + max(G, 4*RTTVAR)
575  * where G is the clock granularity.
576  *
577  * This comment has not necessarily been updated for the new storage
578  * representation:
579  *
580  * Because of the way we do the smoothing, srtt and rttvar
581  * will each average +1/2 tick of bias.  When we compute
582  * the retransmit timer, we want 1/2 tick of rounding and
583  * 1 extra tick because of +-1/2 tick uncertainty in the
584  * firing of the timer.  The bias will give us exactly the
585  * 1.5 tick we need.  But, because the bias is
586  * statistical, we have to test that we don't drop below
587  * the minimum feasible timer (which is 2 ticks).
588  * This macro assumes that the value of 1<<TCP_RTTVAR_SHIFT
589  * is the same as the multiplier for rttvar.
590  *
591  * This macro appears to be wrong; it should be checking rttvar*4 in
592  * ticks and making sure we use 1 instead if rttvar*4 rounds to 0.  It
593  * appears to be treating srtt as being in the old storage
594  * representation, resulting in a factor of 4 extra.
595  */
596 #define	TCP_REXMTVAL(tp) \
597 	((((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar) >> 2)
598 
599 /*
600  * Compute the initial window for slow start.
601  */
602 #define	TCP_INITIAL_WINDOW(iw, segsz) \
603 	uimin((iw) * (segsz), uimax(2 * (segsz), tcp_init_win_max[(iw)]))
604 
605 /*
606  * TCP statistics.
607  * Each counter is an unsigned 64-bit value.
608  *
609  * Many of these should be kept per connection, but that's inconvenient
610  * at the moment.
611  */
612 #define	TCP_STAT_CONNATTEMPT	0	/* connections initiated */
613 #define	TCP_STAT_ACCEPTS	1	/* connections accepted */
614 #define	TCP_STAT_CONNECTS	2	/* connections established */
615 #define	TCP_STAT_DROPS		3	/* connections dropped */
616 #define	TCP_STAT_CONNDROPS	4	/* embryonic connections dropped */
617 #define	TCP_STAT_CLOSED		5	/* conn. closed (includes drops) */
618 #define	TCP_STAT_SEGSTIMED	6	/* segs where we tried to get rtt */
619 #define	TCP_STAT_RTTUPDATED	7	/* times we succeeded */
620 #define	TCP_STAT_DELACK		8	/* delayed ACKs sent */
621 #define	TCP_STAT_TIMEOUTDROP	9	/* conn. dropped in rxmt timeout */
622 #define	TCP_STAT_REXMTTIMEO	10	/* retransmit timeouts */
623 #define	TCP_STAT_PERSISTTIMEO	11	/* persist timeouts */
624 #define	TCP_STAT_KEEPTIMEO	12	/* keepalive timeouts */
625 #define	TCP_STAT_KEEPPROBE	13	/* keepalive probes sent */
626 #define	TCP_STAT_KEEPDROPS	14	/* connections dropped in keepalive */
627 #define	TCP_STAT_PERSISTDROPS	15	/* connections dropped in persist */
628 #define	TCP_STAT_CONNSDRAINED	16	/* connections drained due to memory
629 					   shortage */
630 #define	TCP_STAT_PMTUBLACKHOLE	17	/* PMTUD blackhole detected */
631 #define	TCP_STAT_SNDTOTAL	18	/* total packets sent */
632 #define	TCP_STAT_SNDPACK	19	/* data packlets sent */
633 #define	TCP_STAT_SNDBYTE	20	/* data bytes sent */
634 #define	TCP_STAT_SNDREXMITPACK	21	/* data packets retransmitted */
635 #define	TCP_STAT_SNDREXMITBYTE	22	/* data bytes retransmitted */
636 #define	TCP_STAT_SNDACKS	23	/* ACK-only packets sent */
637 #define	TCP_STAT_SNDPROBE	24	/* window probes sent */
638 #define	TCP_STAT_SNDURG		25	/* packets sent with URG only */
639 #define	TCP_STAT_SNDWINUP	26	/* window update-only packets sent */
640 #define	TCP_STAT_SNDCTRL	27	/* control (SYN|FIN|RST) packets sent */
641 #define	TCP_STAT_RCVTOTAL	28	/* total packets received */
642 #define	TCP_STAT_RCVPACK	29	/* packets received in sequence */
643 #define	TCP_STAT_RCVBYTE	30	/* bytes received in sequence */
644 #define	TCP_STAT_RCVBADSUM	31	/* packets received with cksum errs */
645 #define	TCP_STAT_RCVBADOFF	32	/* packets received with bad offset */
646 #define	TCP_STAT_RCVMEMDROP	33	/* packets dropped for lack of memory */
647 #define	TCP_STAT_RCVSHORT	34	/* packets received too short */
648 #define	TCP_STAT_RCVDUPPACK	35	/* duplicate-only packets received */
649 #define	TCP_STAT_RCVDUPBYTE	36	/* duplicate-only bytes received */
650 #define	TCP_STAT_RCVPARTDUPPACK	37	/* packets with some duplicate data */
651 #define	TCP_STAT_RCVPARTDUPBYTE	38	/* dup. bytes in part-dup. packets */
652 #define	TCP_STAT_RCVOOPACK	39	/* out-of-order packets received */
653 #define	TCP_STAT_RCVOOBYTE	40	/* out-of-order bytes received */
654 #define	TCP_STAT_RCVPACKAFTERWIN 41	/* packets with data after window */
655 #define	TCP_STAT_RCVBYTEAFTERWIN 42	/* bytes received after window */
656 #define	TCP_STAT_RCVAFTERCLOSE	43	/* packets received after "close" */
657 #define	TCP_STAT_RCVWINPROBE	44	/* rcvd window probe packets */
658 #define	TCP_STAT_RCVDUPACK	45	/* rcvd duplicate ACKs */
659 #define	TCP_STAT_RCVACKTOOMUCH	46	/* rcvd ACKs for unsent data */
660 #define	TCP_STAT_RCVACKPACK	47	/* rcvd ACK packets */
661 #define	TCP_STAT_RCVACKBYTE	48	/* bytes ACKed by rcvd ACKs */
662 #define	TCP_STAT_RCVWINUPD	49	/* rcvd window update packets */
663 #define	TCP_STAT_PAWSDROP	50	/* segments dropped due to PAWS */
664 #define	TCP_STAT_PREDACK	51	/* times hdr predict OK for ACKs */
665 #define	TCP_STAT_PREDDAT	52	/* times hdr predict OK for data pkts */
666 #define	TCP_STAT_PCBHASHMISS	53	/* input packets missing PCB hash */
667 #define	TCP_STAT_NOPORT		54	/* no socket on port */
668 #define	TCP_STAT_BADSYN		55	/* received ACK for which we have
669 					   no SYN in compressed state */
670 #define	TCP_STAT_DELAYED_FREE	56	/* delayed pool_put() of tcpcb */
671 #define	TCP_STAT_SC_ADDED	57	/* # of sc entries added */
672 #define	TCP_STAT_SC_COMPLETED	58	/* # of sc connections completed */
673 #define	TCP_STAT_SC_TIMED_OUT	59	/* # of sc entries timed out */
674 #define	TCP_STAT_SC_OVERFLOWED	60	/* # of sc drops due to overflow */
675 #define	TCP_STAT_SC_RESET	61	/* # of sc drops due to RST */
676 #define	TCP_STAT_SC_UNREACH	62	/* # of sc drops due to ICMP unreach */
677 #define	TCP_STAT_SC_BUCKETOVERFLOW 63	/* # of sc drops due to bucket ovflow */
678 #define	TCP_STAT_SC_ABORTED	64	/* # of sc entries aborted (no mem) */
679 #define	TCP_STAT_SC_DUPESYN	65	/* # of duplicate SYNs received */
680 #define	TCP_STAT_SC_DROPPED	66	/* # of SYNs dropped (no route/mem) */
681 #define	TCP_STAT_SC_COLLISIONS	67	/* # of sc hash collisions */
682 #define	TCP_STAT_SC_RETRANSMITTED 68	/* # of sc retransmissions */
683 #define	TCP_STAT_SC_DELAYED_FREE 69	/* # of delayed pool_put()s */
684 #define	TCP_STAT_SELFQUENCH	70	/* # of ENOBUFS we get on output */
685 #define	TCP_STAT_BADSIG		71	/* # of drops due to bad signature */
686 #define	TCP_STAT_GOODSIG	72	/* # of packets with good signature */
687 #define	TCP_STAT_ECN_SHS	73	/* # of successful ECN handshakes */
688 #define	TCP_STAT_ECN_CE		74	/* # of packets with CE bit */
689 #define	TCP_STAT_ECN_ECT	75	/* # of packets with ECT(0) bit */
690 
691 #define	TCP_NSTATS		76
692 
693 /*
694  * Names for TCP sysctl objects.
695  */
696 #define	TCPCTL_RFC1323		1	/* RFC1323 timestamps/scaling */
697 #define	TCPCTL_SENDSPACE	2	/* default send buffer */
698 #define	TCPCTL_RECVSPACE	3	/* default recv buffer */
699 #define	TCPCTL_MSSDFLT		4	/* default seg size */
700 #define	TCPCTL_SYN_CACHE_LIMIT	5	/* max size of comp. state engine */
701 #define	TCPCTL_SYN_BUCKET_LIMIT	6	/* max size of hash bucket */
702 #if 0	/*obsoleted*/
703 #define	TCPCTL_SYN_CACHE_INTER	7	/* interval of comp. state timer */
704 #endif
705 #define	TCPCTL_INIT_WIN		8	/* initial window */
706 #define	TCPCTL_MSS_IFMTU	9	/* mss from interface, not in_maxmtu */
707 #define	TCPCTL_SACK		10	/* RFC2018 selective acknowledgement */
708 #define	TCPCTL_WSCALE		11	/* RFC1323 window scaling */
709 #define	TCPCTL_TSTAMP		12	/* RFC1323 timestamps */
710 #if 0	/*obsoleted*/
711 #define	TCPCTL_COMPAT_42	13	/* 4.2BSD TCP bug work-arounds */
712 #endif
713 #define	TCPCTL_CWM		14	/* Congestion Window Monitoring */
714 #define	TCPCTL_CWM_BURSTSIZE	15	/* burst size allowed by CWM */
715 #define	TCPCTL_ACK_ON_PUSH	16	/* ACK immediately on PUSH */
716 #define	TCPCTL_KEEPIDLE		17	/* keepalive idle time */
717 #define	TCPCTL_KEEPINTVL	18	/* keepalive probe interval */
718 #define	TCPCTL_KEEPCNT		19	/* keepalive count */
719 #define	TCPCTL_SLOWHZ		20	/* PR_SLOWHZ (read-only) */
720 #define	TCPCTL_NEWRENO		21	/* NewReno Congestion Control */
721 #define TCPCTL_LOG_REFUSED	22	/* Log refused connections */
722 #if 0	/*obsoleted*/
723 #define	TCPCTL_RSTRATELIMIT	23	/* RST rate limit */
724 #endif
725 #define	TCPCTL_RSTPPSLIMIT	24	/* RST pps limit */
726 #define	TCPCTL_DELACK_TICKS	25	/* # ticks to delay ACK */
727 #define	TCPCTL_INIT_WIN_LOCAL	26	/* initial window for local nets */
728 #define	TCPCTL_IDENT		27	/* rfc 931 identd */
729 #define	TCPCTL_ACKDROPRATELIMIT	28	/* SYN/RST -> ACK rate limit */
730 #define	TCPCTL_LOOPBACKCKSUM	29	/* do TCP checksum on loopback */
731 #define	TCPCTL_STATS		30	/* TCP statistics */
732 #define	TCPCTL_DEBUG		31	/* TCP debug sockets */
733 #define	TCPCTL_DEBX		32	/* # of tcp debug sockets */
734 #define	TCPCTL_DROP		33	/* drop tcp connection */
735 #define	TCPCTL_MSL		34	/* Max Segment Life */
736 
737 #ifdef _KERNEL
738 
739 extern	struct inpcbtable tcbtable;	/* head of queue of active tcpcb's */
740 extern	const struct pr_usrreqs tcp_usrreqs;
741 
742 extern	u_int32_t tcp_now;	/* for RFC 1323 timestamps */
743 extern	int tcp_do_rfc1323;	/* enabled/disabled? */
744 extern	int tcp_do_sack;	/* SACK enabled/disabled? */
745 extern	int tcp_do_win_scale;	/* RFC1323 window scaling enabled/disabled? */
746 extern	int tcp_do_timestamps;	/* RFC1323 timestamps enabled/disabled? */
747 extern	int tcp_mssdflt;	/* default seg size */
748 extern	int tcp_minmss;		/* minimal seg size */
749 extern  int tcp_msl;		/* max segment life */
750 extern	int tcp_init_win;	/* initial window */
751 extern	int tcp_init_win_local;	/* initial window for local nets */
752 extern	int tcp_init_win_max[11];/* max sizes for values of tcp_init_win_* */
753 extern	int tcp_mss_ifmtu;	/* take MSS from interface, not in_maxmtu */
754 extern	int tcp_cwm;		/* enable Congestion Window Monitoring */
755 extern	int tcp_cwm_burstsize;	/* burst size allowed by CWM */
756 extern	int tcp_ack_on_push;	/* ACK immediately on PUSH */
757 extern	int tcp_log_refused;	/* log refused connections */
758 extern	int tcp_do_ecn;		/* TCP ECN enabled/disabled? */
759 extern	int tcp_ecn_maxretries;	/* Max ECN setup retries */
760 extern	int tcp_do_rfc1948;	/* ISS by cryptographic hash */
761 extern int tcp_sack_tp_maxholes;	/* Max holes per connection. */
762 extern int tcp_sack_globalmaxholes;	/* Max holes per system. */
763 extern int tcp_sack_globalholes;	/* Number of holes present. */
764 extern int tcp_do_abc;			/* RFC3465 ABC enabled/disabled? */
765 extern int tcp_abc_aggressive;		/* 1: L=2*SMSS  0: L=1*SMSS */
766 
767 extern int tcp_msl_enable;		/* enable TIME_WAIT truncation	*/
768 extern int tcp_msl_loop;		/* MSL for loopback		*/
769 extern int tcp_msl_local;		/* MSL for 'local'		*/
770 extern int tcp_msl_remote;		/* MSL otherwise		*/
771 extern int tcp_msl_remote_threshold;	/* RTT threshold		*/
772 extern int tcp_rttlocal;		/* Use RTT to decide who's 'local' */
773 extern int tcp4_vtw_enable;
774 extern int tcp6_vtw_enable;
775 extern int tcp_vtw_was_enabled;
776 extern int tcp_vtw_entries;
777 
778 extern	int tcp_rst_ppslim;
779 extern	int tcp_ackdrop_ppslim;
780 
781 #ifdef MBUFTRACE
782 extern	struct mowner tcp_rx_mowner;
783 extern	struct mowner tcp_tx_mowner;
784 extern	struct mowner tcp_reass_mowner;
785 extern	struct mowner tcp_sock_mowner;
786 extern	struct mowner tcp_sock_rx_mowner;
787 extern	struct mowner tcp_sock_tx_mowner;
788 extern	struct mowner tcp_mowner;
789 #endif
790 
791 extern int tcp_do_autorcvbuf;
792 extern int tcp_autorcvbuf_inc;
793 extern int tcp_autorcvbuf_max;
794 extern int tcp_do_autosndbuf;
795 extern int tcp_autosndbuf_inc;
796 extern int tcp_autosndbuf_max;
797 
798 struct secasvar;
799 
800 void	 tcp_canceltimers(struct tcpcb *);
801 struct tcpcb *
802 	 tcp_close(struct tcpcb *);
803 int	 tcp_isdead(struct tcpcb *);
804 #ifdef INET6
805 void	 *tcp6_ctlinput(int, const struct sockaddr *, void *);
806 #endif
807 void	 *tcp_ctlinput(int, const struct sockaddr *, void *);
808 int	 tcp_ctloutput(int, struct socket *, struct sockopt *);
809 struct tcpcb *
810 	 tcp_disconnect1(struct tcpcb *);
811 struct tcpcb *
812 	 tcp_drop(struct tcpcb *, int);
813 #ifdef TCP_SIGNATURE
814 int	 tcp_signature_apply(void *, void *, u_int);
815 struct secasvar *tcp_signature_getsav(struct mbuf *);
816 int	 tcp_signature(struct mbuf *, struct tcphdr *, int, struct secasvar *,
817 	    char *);
818 #endif
819 void	 tcp_drain(void);
820 void	 tcp_drainstub(void);
821 void	 tcp_established(struct tcpcb *);
822 void	 tcp_init(void);
823 void	 tcp_init_common(unsigned);
824 #ifdef INET6
825 int	 tcp6_input(struct mbuf **, int *, int);
826 #endif
827 void	 tcp_input(struct mbuf *, int, int);
828 u_int	 tcp_hdrsz(struct tcpcb *);
829 u_long	 tcp_mss_to_advertise(const struct ifnet *, int);
830 void	 tcp_mss_from_peer(struct tcpcb *, int);
831 void	 tcp_tcpcb_template(void);
832 struct tcpcb *
833 	 tcp_newtcpcb(int, void *);
834 void	 tcp_notify(struct inpcb *, int);
835 #ifdef INET6
836 void	 tcp6_notify(struct in6pcb *, int);
837 #endif
838 u_int	 tcp_optlen(struct tcpcb *);
839 int	 tcp_output(struct tcpcb *);
840 void	 tcp_pulloutofband(struct socket *,
841 	    struct tcphdr *, struct mbuf *, int);
842 void	 tcp_quench(struct inpcb *);
843 #ifdef INET6
844 void	 tcp6_quench(struct in6pcb *);
845 #endif
846 void	 tcp_mtudisc(struct inpcb *, int);
847 #ifdef INET6
848 void	 tcp6_mtudisc_callback(struct in6_addr *);
849 #endif
850 
851 void	tcpipqent_init(void);
852 struct ipqent *tcpipqent_alloc(void);
853 void	 tcpipqent_free(struct ipqent *);
854 
855 int	 tcp_respond(struct tcpcb *, struct mbuf *, struct mbuf *,
856 	    struct tcphdr *, tcp_seq, tcp_seq, int);
857 void	 tcp_rmx_rtt(struct tcpcb *);
858 void	 tcp_setpersist(struct tcpcb *);
859 #ifdef TCP_SIGNATURE
860 int	 tcp_signature_compute(struct mbuf *, struct tcphdr *, int, int,
861 	    int, u_char *, u_int);
862 #endif
863 void	 tcp_fasttimo(void);
864 struct mbuf *
865 	 tcp_template(struct tcpcb *);
866 void	 tcp_trace(short, short, struct tcpcb *, struct mbuf *, int);
867 struct tcpcb *
868 	 tcp_usrclosed(struct tcpcb *);
869 void	 tcp_usrreq_init(void);
870 void	 tcp_xmit_timer(struct tcpcb *, uint32_t);
871 tcp_seq	 tcp_new_iss(struct tcpcb *);
872 tcp_seq  tcp_new_iss1(void *, void *, u_int16_t, u_int16_t, size_t);
873 
874 void	 tcp_sack_init(void);
875 void	 tcp_new_dsack(struct tcpcb *, tcp_seq, u_int32_t);
876 void	 tcp_sack_option(struct tcpcb *, const struct tcphdr *,
877 	    const u_char *, int);
878 void	 tcp_del_sackholes(struct tcpcb *, const struct tcphdr *);
879 void	 tcp_free_sackholes(struct tcpcb *);
880 void	 tcp_sack_adjust(struct tcpcb *tp);
881 struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
882 int	 tcp_sack_numblks(const struct tcpcb *);
883 #define	TCP_SACK_OPTLEN(nblks)	((nblks) * 8 + 2 + 2)
884 
885 void	 tcp_statinc(u_int);
886 void	 tcp_statadd(u_int, uint64_t);
887 
888 int	 tcp_input_checksum(int, struct mbuf *, const struct tcphdr *, int, int,
889     int);
890 
891 int	tcp_dooptions(struct tcpcb *, const u_char *, int,
892 	    struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *);
893 #endif
894 
895 #endif /* !_NETINET_TCP_VAR_H_ */
896