1/* $NetBSD: tcp_var.h,v 1.199 2024/12/03 20:02:30 andvar 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 */
168struct tcpiphdr {
169 struct ipovly ti_i; /* overlaid ip structure */
170 struct tcphdr ti_t; /* tcp header */
171};
172#ifdef CTASSERT
173CTASSERT(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 */
194struct sackblk {
195 tcp_seq left; /* Left edge of sack block. */
196 tcp_seq right; /* Right edge of sack block. */
197};
198
199TAILQ_HEAD(sackhead, sackhole);
200struct sackhole {
201 tcp_seq start;
202 tcp_seq end;
203 tcp_seq rxmit;
204
205 TAILQ_ENTRY(sackhole) sackhole_q;
206};
207
208struct syn_cache;
209
210/*
211 * Tcp control block, one per tcp; fields:
212 */
213struct 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 callout_t t_delack_ch; /* delayed ACK callout */
257/*
258 * The following fields are used as in the protocol specification.
259 * See RFC793, Dec. 1981, page 21.
260 */
261/* send sequence variables */
262 tcp_seq snd_una; /* send unacknowledged */
263 tcp_seq snd_nxt; /* send next */
264 tcp_seq snd_up; /* send urgent pointer */
265 tcp_seq snd_wl1; /* window update seg seq number */
266 tcp_seq snd_wl2; /* window update seg ack number */
267 tcp_seq iss; /* initial send sequence number */
268 u_long snd_wnd; /* send window */
269/*
270 * snd_recover
271 * it's basically same as the "recover" variable in RFC 2852 (NewReno).
272 * when entering fast retransmit, it's set to snd_max.
273 * newreno uses this to detect partial ack.
274 * snd_high
275 * it's basically same as the "send_high" variable in RFC 2852 (NewReno).
276 * on each RTO, it's set to snd_max.
277 * newreno uses this to avoid false fast retransmits.
278 */
279 tcp_seq snd_recover;
280 tcp_seq snd_high;
281/* receive sequence variables */
282 u_long rcv_wnd; /* receive window */
283 tcp_seq rcv_nxt; /* receive next */
284 tcp_seq rcv_up; /* receive urgent pointer */
285 tcp_seq irs; /* initial receive sequence number */
286/*
287 * Additional variables for this implementation.
288 */
289/* receive variables */
290 tcp_seq rcv_adv; /* advertised window */
291
292/*
293 * retransmit variables
294 *
295 * snd_max
296 * the highest sequence number we've ever sent.
297 * used to recognize retransmits.
298 */
299 tcp_seq snd_max;
300
301/* congestion control (for slow start, source quench, retransmit after loss) */
302 u_long snd_cwnd; /* congestion-controlled window */
303 u_long snd_ssthresh; /* snd_cwnd size threshold for
304 * for slow start exponential to
305 * linear switch
306 */
307/* auto-sizing variables */
308 u_int rfbuf_cnt; /* recv buffer autoscaling byte count */
309 uint32_t rfbuf_ts; /* recv buffer autoscaling timestamp */
310
311/*
312 * transmit timing stuff. See below for scale of srtt and rttvar.
313 * "Variance" is actually smoothed difference.
314 */
315 uint32_t t_rcvtime; /* time last segment received */
316 uint32_t t_rtttime; /* time we started measuring rtt */
317 tcp_seq t_rtseq; /* sequence number being timed */
318 int32_t t_srtt; /* smoothed round-trip time */
319 int32_t t_rttvar; /* variance in round-trip time */
320 uint32_t t_rttmin; /* minimum rtt allowed */
321 u_long max_sndwnd; /* largest window peer has offered */
322
323/* out-of-band data */
324 char t_oobflags; /* have some */
325 char t_iobc; /* input character */
326#define TCPOOB_HAVEDATA 0x01
327#define TCPOOB_HADDATA 0x02
328 short t_softerror; /* possible error not yet reported */
329
330/* RFC 1323 variables */
331 u_char snd_scale; /* window scaling for send window */
332 u_char rcv_scale; /* window scaling for recv window */
333 u_char request_r_scale; /* pending window scaling */
334 u_char requested_s_scale;
335 u_int32_t ts_recent; /* timestamp echo data */
336 u_int32_t ts_recent_age; /* when last updated */
337 u_int32_t ts_timebase; /* our timebase */
338 tcp_seq last_ack_sent;
339
340/* RFC 3465 variables */
341 u_long t_bytes_acked; /* ABC "bytes_acked" parameter */
342
343/* SACK stuff */
344#define TCP_SACK_MAX 3
345#define TCPSACK_NONE 0
346#define TCPSACK_HAVED 1
347 u_char rcv_sack_flags; /* SACK flags. */
348 struct sackblk rcv_dsack_block; /* RX D-SACK block. */
349 struct ipqehead timeq; /* time sequenced queue. */
350 struct sackhead snd_holes; /* TX SACK holes. */
351 int snd_numholes; /* Number of TX SACK holes. */
352 tcp_seq rcv_lastsack; /* last seq number(+1) sack'd by rcv'r*/
353 tcp_seq sack_newdata; /* New data xmitted in this recovery
354 episode starts at this seq number*/
355 tcp_seq snd_fack; /* FACK TCP. Forward-most data held by
356 peer. */
357
358/* CUBIC variables */
359 ulong snd_cubic_wmax; /* W_max */
360 ulong snd_cubic_wmax_last; /* Used for fast convergence */
361 ulong snd_cubic_ctime; /* Last congestion time */
362
363/* pointer for syn cache entries*/
364 LIST_HEAD(, syn_cache) t_sc; /* list of entries by this tcb */
365
366/* prediction of next mbuf when using large window sizes */
367 struct mbuf *t_lastm; /* last mbuf that data was sent from */
368 int t_inoff; /* data offset in previous mbuf */
369 int t_lastoff; /* last data address in mbuf chain */
370 int t_lastlen; /* last length read from mbuf chain */
371
372/* Path-MTU discovery blackhole detection */
373 int t_mtudisc; /* perform mtudisc for this tcb */
374/* Path-MTU Discovery Information */
375 u_int t_pmtud_mss_acked; /* MSS acked, lower bound for MTU */
376 u_int t_pmtud_mtu_sent; /* MTU used, upper bound for MTU */
377 tcp_seq t_pmtud_th_seq; /* TCP SEQ from ICMP payload */
378 u_int t_pmtud_nextmtu; /* Advertised Next-Hop MTU from ICMP */
379 u_short t_pmtud_ip_len; /* IP length from ICMP payload */
380 u_short t_pmtud_ip_hl; /* IP header length from ICMP payload */
381
382 uint8_t t_ecn_retries; /* # of ECN setup retries */
383
384 const struct tcp_congctl *t_congctl; /* per TCB congctl algorithm */
385
386 /* Keepalive per socket */
387 u_int t_keepinit;
388 u_int t_keepidle;
389 u_int t_keepintvl;
390 u_int t_keepcnt;
391 u_int t_maxidle; /* t_keepcnt * t_keepintvl */
392
393 u_int t_msl; /* MSL to use for this connexion */
394
395 /* maintain a few stats per connection: */
396 uint32_t t_rcvoopack; /* out-of-order packets received */
397 uint32_t t_sndrexmitpack; /* retransmit packets sent */
398 uint32_t t_sndzerowin; /* zero-window updates sent */
399};
400
401/*
402 * Macros to aid ECN TCP.
403 */
404#define TCP_ECN_ALLOWED(tp) (tp->t_flags & TF_ECN_PERMIT)
405
406/*
407 * Macros to aid SACK/FACK TCP.
408 */
409#define TCP_SACK_ENABLED(tp) (tp->t_flags & TF_WILL_SACK)
410#define TCP_FACK_FASTRECOV(tp) \
411 (TCP_SACK_ENABLED(tp) && \
412 (SEQ_GT(tp->snd_fack, tp->snd_una + tcprexmtthresh * tp->t_segsz)))
413
414#ifdef _KERNEL
415/*
416 * TCP reassembly queue locks.
417 */
418static __inline int tcp_reass_lock_try (struct tcpcb *)
419 __unused;
420static __inline void tcp_reass_unlock (struct tcpcb *)
421 __unused;
422
423static __inline int
424tcp_reass_lock_try(struct tcpcb *tp)
425{
426 int s;
427
428 /*
429 * Use splvm() -- we're blocking things that would cause
430 * mbuf allocation.
431 */
432 s = splvm();
433 if (tp->t_flags & TF_REASSEMBLING) {
434 splx(s);
435 return (0);
436 }
437 tp->t_flags |= TF_REASSEMBLING;
438 splx(s);
439 return (1);
440}
441
442static __inline void
443tcp_reass_unlock(struct tcpcb *tp)
444{
445 int s;
446
447 s = splvm();
448 KASSERT((tp->t_flags & TF_REASSEMBLING) != 0);
449 tp->t_flags &= ~TF_REASSEMBLING;
450 splx(s);
451}
452
453#ifdef DIAGNOSTIC
454#define TCP_REASS_LOCK(tp) \
455do { \
456 if (tcp_reass_lock_try(tp) == 0) { \
457 printf("%s:%d: tcpcb %p reass already locked\n", \
458 __FILE__, __LINE__, tp); \
459 panic("tcp_reass_lock"); \
460 } \
461} while (/*CONSTCOND*/ 0)
462#define TCP_REASS_LOCK_CHECK(tp) \
463do { \
464 if (((tp)->t_flags & TF_REASSEMBLING) == 0) { \
465 printf("%s:%d: tcpcb %p reass lock not held\n", \
466 __FILE__, __LINE__, tp); \
467 panic("tcp reass lock check"); \
468 } \
469} while (/*CONSTCOND*/ 0)
470#else
471#define TCP_REASS_LOCK(tp) (void) tcp_reass_lock_try((tp))
472#define TCP_REASS_LOCK_CHECK(tp) /* nothing */
473#endif
474
475#define TCP_REASS_UNLOCK(tp) tcp_reass_unlock((tp))
476#endif /* _KERNEL */
477
478/*
479 * Queue for delayed ACK processing.
480 */
481#ifdef _KERNEL
482extern int tcp_delack_ticks;
483void tcp_delack(void *);
484
485#define TCP_RESTART_DELACK(tp) \
486 callout_reset(&(tp)->t_delack_ch, tcp_delack_ticks, \
487 tcp_delack, tp)
488
489#define TCP_SET_DELACK(tp) \
490do { \
491 if (((tp)->t_flags & TF_DELACK) == 0) { \
492 (tp)->t_flags |= TF_DELACK; \
493 TCP_RESTART_DELACK(tp); \
494 } \
495} while (/*CONSTCOND*/0)
496
497#define TCP_CLEAR_DELACK(tp) \
498do { \
499 if ((tp)->t_flags & TF_DELACK) { \
500 (tp)->t_flags &= ~TF_DELACK; \
501 callout_stop(&(tp)->t_delack_ch); \
502 } \
503} while (/*CONSTCOND*/0)
504#endif /* _KERNEL */
505
506/*
507 * Compute the current timestamp for a connection.
508 */
509#define TCP_TIMESTAMP(tp) (tcp_now - (tp)->ts_timebase)
510
511/*
512 * Handy way of passing around TCP option info.
513 */
514struct tcp_opt_info {
515 int ts_present;
516 u_int32_t ts_val;
517 u_int32_t ts_ecr;
518 u_int16_t maxseg;
519};
520
521#define TOF_SIGNATURE 0x0040 /* signature option present */
522#define TOF_SIGLEN 0x0080 /* sigature length valid (RFC2385) */
523
524#define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
525#define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
526
527/*
528 * See RFC2988 for a discussion of RTO calculation; comments assume
529 * familiarity with that document.
530 *
531 * The smoothed round-trip time and estimated variance are stored as
532 * fixed point numbers. Historically, srtt was scaled by
533 * TCP_RTT_SHIFT bits, and rttvar by TCP_RTTVAR_SHIFT bits. Because
534 * the values coincide with the alpha and beta parameters suggested
535 * for RTO calculation (1/8 for srtt, 1/4 for rttvar), the combination
536 * of computing 1/8 of the new value and transforming it to the
537 * fixed-point representation required zero instructions. However,
538 * the storage representations no longer coincide with the alpha/beta
539 * shifts; instead, more fractional bits are present.
540 *
541 * The storage representation of srtt is 1/32 slow ticks, or 1/64 s.
542 * (The assumption that a slow tick is 500 ms should not be present in
543 * the code.)
544 *
545 * The storage representation of rttvar is 1/16 slow ticks, or 1/32 s.
546 * There may be some confusion about this in the code.
547 *
548 * For historical reasons, these scales are also used in smoothing the
549 * average (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
550 * This results in alpha of 0.125 and beta of 0.25, following RFC2988
551 * section 2.3
552 *
553 * XXX Change SHIFT values to LGWEIGHT and REP_SHIFT, and adjust
554 * the code to use the correct ones.
555 */
556#define TCP_RTT_SHIFT 3 /* shift for srtt; 3 bits frac. */
557#define TCP_RTTVAR_SHIFT 2 /* multiplier for rttvar; 2 bits */
558
559/*
560 * Compute TCP retransmission timer, following RFC2988.
561 * This macro returns a value in slow timeout ticks.
562 *
563 * Section 2.2 requires that the RTO value be
564 * srtt + max(G, 4*RTTVAR)
565 * where G is the clock granularity.
566 *
567 * This comment has not necessarily been updated for the new storage
568 * representation:
569 *
570 * Because of the way we do the smoothing, srtt and rttvar
571 * will each average +1/2 tick of bias. When we compute
572 * the retransmit timer, we want 1/2 tick of rounding and
573 * 1 extra tick because of +-1/2 tick uncertainty in the
574 * firing of the timer. The bias will give us exactly the
575 * 1.5 tick we need. But, because the bias is
576 * statistical, we have to test that we don't drop below
577 * the minimum feasible timer (which is 2 ticks).
578 * This macro assumes that the value of 1<<TCP_RTTVAR_SHIFT
579 * is the same as the multiplier for rttvar.
580 *
581 * This macro appears to be wrong; it should be checking rttvar*4 in
582 * ticks and making sure we use 1 instead if rttvar*4 rounds to 0. It
583 * appears to be treating srtt as being in the old storage
584 * representation, resulting in a factor of 4 extra.
585 */
586#define TCP_REXMTVAL(tp) \
587 ((((tp)->t_srtt >> TCP_RTT_SHIFT) + (tp)->t_rttvar) >> 2)
588
589/*
590 * Compute the initial window for slow start.
591 */
592#define TCP_INITIAL_WINDOW(iw, segsz) \
593 uimin((iw) * (segsz), uimax(2 * (segsz), tcp_init_win_max[(iw)]))
594
595/*
596 * TCP statistics.
597 * Each counter is an unsigned 64-bit value.
598 *
599 * Many of these should be kept per connection, but that's inconvenient
600 * at the moment.
601 */
602#define TCP_STAT_CONNATTEMPT 0 /* connections initiated */
603#define TCP_STAT_ACCEPTS 1 /* connections accepted */
604#define TCP_STAT_CONNECTS 2 /* connections established */
605#define TCP_STAT_DROPS 3 /* connections dropped */
606#define TCP_STAT_CONNDROPS 4 /* embryonic connections dropped */
607#define TCP_STAT_CLOSED 5 /* conn. closed (includes drops) */
608#define TCP_STAT_SEGSTIMED 6 /* segs where we tried to get rtt */
609#define TCP_STAT_RTTUPDATED 7 /* times we succeeded */
610#define TCP_STAT_DELACK 8 /* delayed ACKs sent */
611#define TCP_STAT_TIMEOUTDROP 9 /* conn. dropped in rxmt timeout */
612#define TCP_STAT_REXMTTIMEO 10 /* retransmit timeouts */
613#define TCP_STAT_PERSISTTIMEO 11 /* persist timeouts */
614#define TCP_STAT_KEEPTIMEO 12 /* keepalive timeouts */
615#define TCP_STAT_KEEPPROBE 13 /* keepalive probes sent */
616#define TCP_STAT_KEEPDROPS 14 /* connections dropped in keepalive */
617#define TCP_STAT_PERSISTDROPS 15 /* connections dropped in persist */
618#define TCP_STAT_CONNSDRAINED 16 /* connections drained due to memory
619 shortage */
620#define TCP_STAT_PMTUBLACKHOLE 17 /* PMTUD blackhole detected */
621#define TCP_STAT_SNDTOTAL 18 /* total packets sent */
622#define TCP_STAT_SNDPACK 19 /* data packets sent */
623#define TCP_STAT_SNDBYTE 20 /* data bytes sent */
624#define TCP_STAT_SNDREXMITPACK 21 /* data packets retransmitted */
625#define TCP_STAT_SNDREXMITBYTE 22 /* data bytes retransmitted */
626#define TCP_STAT_SNDACKS 23 /* ACK-only packets sent */
627#define TCP_STAT_SNDPROBE 24 /* window probes sent */
628#define TCP_STAT_SNDURG 25 /* packets sent with URG only */
629#define TCP_STAT_SNDWINUP 26 /* window update-only packets sent */
630#define TCP_STAT_SNDCTRL 27 /* control (SYN|FIN|RST) packets sent */
631#define TCP_STAT_RCVTOTAL 28 /* total packets received */
632#define TCP_STAT_RCVPACK 29 /* packets received in sequence */
633#define TCP_STAT_RCVBYTE 30 /* bytes received in sequence */
634#define TCP_STAT_RCVBADSUM 31 /* packets received with cksum errs */
635#define TCP_STAT_RCVBADOFF 32 /* packets received with bad offset */
636#define TCP_STAT_RCVMEMDROP 33 /* packets dropped for lack of memory */
637#define TCP_STAT_RCVSHORT 34 /* packets received too short */
638#define TCP_STAT_RCVDUPPACK 35 /* duplicate-only packets received */
639#define TCP_STAT_RCVDUPBYTE 36 /* duplicate-only bytes received */
640#define TCP_STAT_RCVPARTDUPPACK 37 /* packets with some duplicate data */
641#define TCP_STAT_RCVPARTDUPBYTE 38 /* dup. bytes in part-dup. packets */
642#define TCP_STAT_RCVOOPACK 39 /* out-of-order packets received */
643#define TCP_STAT_RCVOOBYTE 40 /* out-of-order bytes received */
644#define TCP_STAT_RCVPACKAFTERWIN 41 /* packets with data after window */
645#define TCP_STAT_RCVBYTEAFTERWIN 42 /* bytes received after window */
646#define TCP_STAT_RCVAFTERCLOSE 43 /* packets received after "close" */
647#define TCP_STAT_RCVWINPROBE 44 /* rcvd window probe packets */
648#define TCP_STAT_RCVDUPACK 45 /* rcvd duplicate ACKs */
649#define TCP_STAT_RCVACKTOOMUCH 46 /* rcvd ACKs for unsent data */
650#define TCP_STAT_RCVACKPACK 47 /* rcvd ACK packets */
651#define TCP_STAT_RCVACKBYTE 48 /* bytes ACKed by rcvd ACKs */
652#define TCP_STAT_RCVWINUPD 49 /* rcvd window update packets */
653#define TCP_STAT_PAWSDROP 50 /* segments dropped due to PAWS */
654#define TCP_STAT_PREDACK 51 /* times hdr predict OK for ACKs */
655#define TCP_STAT_PREDDAT 52 /* times hdr predict OK for data pkts */
656#define TCP_STAT_PCBHASHMISS 53 /* input packets missing PCB hash */
657#define TCP_STAT_NOPORT 54 /* no socket on port */
658#define TCP_STAT_BADSYN 55 /* received ACK for which we have
659 no SYN in compressed state */
660#define TCP_STAT_DELAYED_FREE 56 /* delayed pool_put() of tcpcb */
661#define TCP_STAT_SC_ADDED 57 /* # of sc entries added */
662#define TCP_STAT_SC_COMPLETED 58 /* # of sc connections completed */
663#define TCP_STAT_SC_TIMED_OUT 59 /* # of sc entries timed out */
664#define TCP_STAT_SC_OVERFLOWED 60 /* # of sc drops due to overflow */
665#define TCP_STAT_SC_RESET 61 /* # of sc drops due to RST */
666#define TCP_STAT_SC_UNREACH 62 /* # of sc drops due to ICMP unreach */
667#define TCP_STAT_SC_BUCKETOVERFLOW 63 /* # of sc drops due to bucket ovflow */
668#define TCP_STAT_SC_ABORTED 64 /* # of sc entries aborted (no mem) */
669#define TCP_STAT_SC_DUPESYN 65 /* # of duplicate SYNs received */
670#define TCP_STAT_SC_DROPPED 66 /* # of SYNs dropped (no route/mem) */
671#define TCP_STAT_SC_COLLISIONS 67 /* # of sc hash collisions */
672#define TCP_STAT_SC_RETRANSMITTED 68 /* # of sc retransmissions */
673#define TCP_STAT_SC_DELAYED_FREE 69 /* # of delayed pool_put()s */
674#define TCP_STAT_SELFQUENCH 70 /* # of ENOBUFS we get on output */
675#define TCP_STAT_BADSIG 71 /* # of drops due to bad signature */
676#define TCP_STAT_GOODSIG 72 /* # of packets with good signature */
677#define TCP_STAT_ECN_SHS 73 /* # of successful ECN handshakes */
678#define TCP_STAT_ECN_CE 74 /* # of packets with CE bit */
679#define TCP_STAT_ECN_ECT 75 /* # of packets with ECT(0) bit */
680
681#define TCP_NSTATS 76
682
683/*
684 * Names for TCP sysctl objects.
685 */
686#define TCPCTL_RFC1323 1 /* RFC1323 timestamps/scaling */
687#define TCPCTL_SENDSPACE 2 /* default send buffer */
688#define TCPCTL_RECVSPACE 3 /* default recv buffer */
689#define TCPCTL_MSSDFLT 4 /* default seg size */
690#define TCPCTL_SYN_CACHE_LIMIT 5 /* max size of comp. state engine */
691#define TCPCTL_SYN_BUCKET_LIMIT 6 /* max size of hash bucket */
692#if 0 /*obsoleted*/
693#define TCPCTL_SYN_CACHE_INTER 7 /* interval of comp. state timer */
694#endif
695#define TCPCTL_INIT_WIN 8 /* initial window */
696#define TCPCTL_MSS_IFMTU 9 /* mss from interface, not in_maxmtu */
697#define TCPCTL_SACK 10 /* RFC2018 selective acknowledgement */
698#define TCPCTL_WSCALE 11 /* RFC1323 window scaling */
699#define TCPCTL_TSTAMP 12 /* RFC1323 timestamps */
700#if 0 /*obsoleted*/
701#define TCPCTL_COMPAT_42 13 /* 4.2BSD TCP bug work-arounds */
702#endif
703#define TCPCTL_CWM 14 /* Congestion Window Monitoring */
704#define TCPCTL_CWM_BURSTSIZE 15 /* burst size allowed by CWM */
705#define TCPCTL_ACK_ON_PUSH 16 /* ACK immediately on PUSH */
706#define TCPCTL_KEEPIDLE 17 /* keepalive idle time */
707#define TCPCTL_KEEPINTVL 18 /* keepalive probe interval */
708#define TCPCTL_KEEPCNT 19 /* keepalive count */
709#define TCPCTL_SLOWHZ 20 /* PR_SLOWHZ (read-only) */
710#define TCPCTL_NEWRENO 21 /* NewReno Congestion Control */
711#define TCPCTL_LOG_REFUSED 22 /* Log refused connections */
712#if 0 /*obsoleted*/
713#define TCPCTL_RSTRATELIMIT 23 /* RST rate limit */
714#endif
715#define TCPCTL_RSTPPSLIMIT 24 /* RST pps limit */
716#define TCPCTL_DELACK_TICKS 25 /* # ticks to delay ACK */
717#define TCPCTL_INIT_WIN_LOCAL 26 /* initial window for local nets */
718#define TCPCTL_IDENT 27 /* rfc 931 identd */
719#define TCPCTL_ACKDROPRATELIMIT 28 /* SYN/RST -> ACK rate limit */
720#define TCPCTL_LOOPBACKCKSUM 29 /* do TCP checksum on loopback */
721#define TCPCTL_STATS 30 /* TCP statistics */
722#define TCPCTL_DEBUG 31 /* TCP debug sockets */
723#define TCPCTL_DEBX 32 /* # of tcp debug sockets */
724#define TCPCTL_DROP 33 /* drop tcp connection */
725#define TCPCTL_MSL 34 /* Max Segment Life */
726
727#ifdef _KERNEL
728
729extern struct inpcbtable tcbtable; /* head of queue of active tcpcb's */
730extern const struct pr_usrreqs tcp_usrreqs;
731
732extern u_int32_t tcp_now; /* for RFC 1323 timestamps */
733extern int tcp_do_rfc1323; /* enabled/disabled? */
734extern int tcp_do_sack; /* SACK enabled/disabled? */
735extern int tcp_do_win_scale; /* RFC1323 window scaling enabled/disabled? */
736extern int tcp_do_timestamps; /* RFC1323 timestamps enabled/disabled? */
737extern int tcp_mssdflt; /* default seg size */
738extern int tcp_minmss; /* minimal seg size */
739extern int tcp_msl; /* max segment life */
740extern int tcp_init_win; /* initial window */
741extern int tcp_init_win_local; /* initial window for local nets */
742extern int tcp_init_win_max[11];/* max sizes for values of tcp_init_win_* */
743extern int tcp_mss_ifmtu; /* take MSS from interface, not in_maxmtu */
744extern int tcp_cwm; /* enable Congestion Window Monitoring */
745extern int tcp_cwm_burstsize; /* burst size allowed by CWM */
746extern int tcp_ack_on_push; /* ACK immediately on PUSH */
747extern int tcp_log_refused; /* log refused connections */
748extern int tcp_do_ecn; /* TCP ECN enabled/disabled? */
749extern int tcp_ecn_maxretries; /* Max ECN setup retries */
750extern int tcp_do_rfc1948; /* ISS by cryptographic hash */
751extern int tcp_sack_tp_maxholes; /* Max holes per connection. */
752extern int tcp_sack_globalmaxholes; /* Max holes per system. */
753extern int tcp_sack_globalholes; /* Number of holes present. */
754extern int tcp_do_abc; /* RFC3465 ABC enabled/disabled? */
755extern int tcp_abc_aggressive; /* 1: L=2*SMSS 0: L=1*SMSS */
756
757extern int tcp_msl_enable; /* enable TIME_WAIT truncation */
758extern int tcp_msl_loop; /* MSL for loopback */
759extern int tcp_msl_local; /* MSL for 'local' */
760extern int tcp_msl_remote; /* MSL otherwise */
761extern int tcp_msl_remote_threshold; /* RTT threshold */
762extern int tcp_rttlocal; /* Use RTT to decide who's 'local' */
763extern int tcp4_vtw_enable;
764extern int tcp6_vtw_enable;
765extern int tcp_vtw_was_enabled;
766extern int tcp_vtw_entries;
767
768extern int tcp_rst_ppslim;
769extern int tcp_ackdrop_ppslim;
770
771#ifdef MBUFTRACE
772extern struct mowner tcp_rx_mowner;
773extern struct mowner tcp_tx_mowner;
774extern struct mowner tcp_reass_mowner;
775extern struct mowner tcp_sock_mowner;
776extern struct mowner tcp_sock_rx_mowner;
777extern struct mowner tcp_sock_tx_mowner;
778extern struct mowner tcp_mowner;
779#endif
780
781extern int tcp_do_autorcvbuf;
782extern int tcp_autorcvbuf_inc;
783extern int tcp_autorcvbuf_max;
784extern int tcp_do_autosndbuf;
785extern int tcp_autosndbuf_inc;
786extern int tcp_autosndbuf_max;
787
788struct secasvar;
789
790void tcp_canceltimers(struct tcpcb *);
791struct tcpcb *
792 tcp_close(struct tcpcb *);
793int tcp_isdead(struct tcpcb *);
794#ifdef INET6
795void *tcp6_ctlinput(int, const struct sockaddr *, void *);
796#endif
797void *tcp_ctlinput(int, const struct sockaddr *, void *);
798int tcp_ctloutput(int, struct socket *, struct sockopt *);
799struct tcpcb *
800 tcp_disconnect1(struct tcpcb *);
801struct tcpcb *
802 tcp_drop(struct tcpcb *, int);
803#ifdef TCP_SIGNATURE
804int tcp_signature_apply(void *, void *, u_int);
805struct secasvar *tcp_signature_getsav(struct mbuf *);
806int tcp_signature(struct mbuf *, struct tcphdr *, int, struct secasvar *,
807 char *);
808#endif
809void tcp_drain(void);
810void tcp_drainstub(void);
811void tcp_established(struct tcpcb *);
812void tcp_init(void);
813void tcp_init_common(unsigned);
814#ifdef INET6
815int tcp6_input(struct mbuf **, int *, int);
816#endif
817void tcp_input(struct mbuf *, int, int);
818u_int tcp_hdrsz(struct tcpcb *);
819u_long tcp_mss_to_advertise(const struct ifnet *, int);
820void tcp_mss_from_peer(struct tcpcb *, int);
821void tcp_tcpcb_template(void);
822struct tcpcb *
823 tcp_newtcpcb(int, struct inpcb *);
824void tcp_notify(struct inpcb *, int);
825u_int tcp_optlen(struct tcpcb *);
826int tcp_output(struct tcpcb *);
827void tcp_pulloutofband(struct socket *,
828 struct tcphdr *, struct mbuf *, int);
829void tcp_quench(struct inpcb *);
830void tcp_mtudisc(struct inpcb *, int);
831#ifdef INET6
832void tcp6_mtudisc_callback(struct in6_addr *);
833#endif
834
835void tcpipqent_init(void);
836struct ipqent *tcpipqent_alloc(void);
837void tcpipqent_free(struct ipqent *);
838
839int tcp_respond(struct tcpcb *, struct mbuf *, struct mbuf *,
840 struct tcphdr *, tcp_seq, tcp_seq, int);
841void tcp_rmx_rtt(struct tcpcb *);
842void tcp_setpersist(struct tcpcb *);
843#ifdef TCP_SIGNATURE
844int tcp_signature_compute(struct mbuf *, struct tcphdr *, int, int,
845 int, u_char *, u_int);
846#endif
847void tcp_fasttimo(void);
848struct mbuf *
849 tcp_template(struct tcpcb *);
850void tcp_trace(short, short, struct tcpcb *, struct mbuf *, int);
851struct tcpcb *
852 tcp_usrclosed(struct tcpcb *);
853void tcp_usrreq_init(void);
854void tcp_xmit_timer(struct tcpcb *, uint32_t);
855tcp_seq tcp_new_iss(struct tcpcb *);
856tcp_seq tcp_new_iss1(void *, void *, u_int16_t, u_int16_t, size_t);
857
858void tcp_sack_init(void);
859void tcp_new_dsack(struct tcpcb *, tcp_seq, u_int32_t);
860void tcp_sack_option(struct tcpcb *, const struct tcphdr *,
861 const u_char *, int);
862void tcp_del_sackholes(struct tcpcb *, const struct tcphdr *);
863void tcp_free_sackholes(struct tcpcb *);
864void tcp_sack_adjust(struct tcpcb *tp);
865struct sackhole *tcp_sack_output(struct tcpcb *tp, int *sack_bytes_rexmt);
866int tcp_sack_numblks(const struct tcpcb *);
867#define TCP_SACK_OPTLEN(nblks) ((nblks) * 8 + 2 + 2)
868
869void tcp_statinc(u_int);
870void tcp_statadd(u_int, uint64_t);
871
872int tcp_input_checksum(int, struct mbuf *, const struct tcphdr *, int, int,
873 int);
874
875int tcp_dooptions(struct tcpcb *, const u_char *, int,
876 struct tcphdr *, struct mbuf *, int, struct tcp_opt_info *);
877#endif
878
879#endif /* !_NETINET_TCP_VAR_H_ */