xref: /dragonfly/sys/netinet/tcp_var.h (revision 6e278935)
1 /*
2  * Copyright (c) 2003, 2004 Jeffrey M. Hsu.  All rights reserved.
3  * Copyright (c) 2003, 2004 The DragonFly Project.  All rights reserved.
4  *
5  * This code is derived from software contributed to The DragonFly Project
6  * by Jeffrey M. Hsu.
7  *
8  * Redistribution and use in source and binary forms, with or without
9  * modification, are permitted provided that the following conditions
10  * are met:
11  * 1. Redistributions of source code must retain the above copyright
12  *    notice, this list of conditions and the following disclaimer.
13  * 2. Redistributions in binary form must reproduce the above copyright
14  *    notice, this list of conditions and the following disclaimer in the
15  *    documentation and/or other materials provided with the distribution.
16  * 3. Neither the name of The DragonFly Project nor the names of its
17  *    contributors may be used to endorse or promote products derived
18  *    from this software without specific, prior written permission.
19  *
20  * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
21  * ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
22  * LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS
23  * FOR A PARTICULAR PURPOSE ARE DISCLAIMED.  IN NO EVENT SHALL THE
24  * COPYRIGHT HOLDERS OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT,
25  * INCIDENTAL, SPECIAL, EXEMPLARY OR CONSEQUENTIAL DAMAGES (INCLUDING,
26  * BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES;
27  * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED
28  * AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY,
29  * OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT
30  * OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  */
33 
34 /*
35  * Copyright (c) 1982, 1986, 1993, 1994, 1995
36  *	The Regents of the University of California.  All rights reserved.
37  *
38  * Redistribution and use in source and binary forms, with or without
39  * modification, are permitted provided that the following conditions
40  * are met:
41  * 1. Redistributions of source code must retain the above copyright
42  *    notice, this list of conditions and the following disclaimer.
43  * 2. Redistributions in binary form must reproduce the above copyright
44  *    notice, this list of conditions and the following disclaimer in the
45  *    documentation and/or other materials provided with the distribution.
46  * 3. All advertising materials mentioning features or use of this software
47  *    must display the following acknowledgement:
48  *	This product includes software developed by the University of
49  *	California, Berkeley and its contributors.
50  * 4. Neither the name of the University 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  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
55  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
56  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
57  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
58  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
59  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
60  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
61  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
62  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
63  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
64  * SUCH DAMAGE.
65  *
66  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
67  * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.13 2003/02/03 02:34:07 hsu Exp $
68  * $DragonFly: src/sys/netinet/tcp_var.h,v 1.42 2008/10/27 02:56:30 sephe Exp $
69  */
70 
71 #ifndef _NETINET_TCP_VAR_H_
72 #define _NETINET_TCP_VAR_H_
73 
74 #ifndef _NETINET_IN_PCB_H_
75 #include <netinet/in_pcb.h>		/* needed for in_conninfo, inp_gen_t */
76 #endif
77 #ifndef _NETINET_TCP_H_
78 #include <netinet/tcp.h>
79 #endif
80 
81 /*
82  * Kernel variables for tcp.
83  */
84 extern int tcp_do_rfc1323;
85 extern int tcp_do_rfc3390;
86 extern int tcp_do_sack;
87 extern int tcp_do_smartsack;
88 extern int tcp_aggregate_acks;
89 
90 /* TCP segment queue entry */
91 struct tseg_qent {
92 	LIST_ENTRY(tseg_qent) tqe_q;
93 	int	tqe_len;		/* TCP segment data length */
94 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
95 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
96 };
97 LIST_HEAD(tsegqe_head, tseg_qent);
98 extern int	tcp_reass_maxseg;
99 extern int	tcp_reass_qsize;
100 #ifdef MALLOC_DECLARE
101 MALLOC_DECLARE(M_TSEGQ);
102 #endif
103 
104 struct tcptemp {
105 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
106 	struct	tcphdr tt_t;
107 };
108 
109 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
110 
111 struct raw_sackblock {				/* covers [start, end) */
112 	tcp_seq rblk_start;
113 	tcp_seq rblk_end;
114 };
115 
116 /* maximum number of SACK blocks that will fit in the TCP option space */
117 #define	MAX_SACK_REPORT_BLOCKS	4
118 
119 TAILQ_HEAD(sackblock_list, sackblock);
120 
121 struct scoreboard {
122 	int nblocks;
123 	struct sackblock_list sackblocks;
124 	tcp_seq lostseq;			/* passed SACK lost test */
125 	struct sackblock *lastfound;		/* search hint */
126 };
127 
128 struct netmsg_tcp_timer;
129 
130 /*
131  * Tcp control block, one per tcp; fields:
132  * Organized for 16 byte cacheline efficiency.
133  */
134 struct tcpcb {
135 	struct	tsegqe_head t_segq;
136 	int	t_dupacks;		/* consecutive dup acks recd */
137 	int	tt_cpu;			/* sanity check the cpu */
138 
139 	struct	tcp_callout *tt_rexmt;	/* retransmit timer */
140 	struct	tcp_callout *tt_persist;/* retransmit persistence */
141 	struct	tcp_callout *tt_keep;	/* keepalive */
142 	struct	tcp_callout *tt_2msl;	/* 2*msl TIME_WAIT timer */
143 	struct	tcp_callout *tt_delack;	/* delayed ACK timer */
144 	struct	netmsg_tcp_timer *tt_msg; /* timer message */
145 
146 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
147 	int	t_state;		/* state of this connection */
148 	u_int	t_flags;
149 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
150 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
151 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
152 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
153 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
154 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
155 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
156 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
157 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
158 #define	TF_SACK_PERMITTED 0x00000200	/* other side said I could SACK */
159 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
160 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
161 #define	TF_NOPUSH	0x00001000	/* don't push */
162 #define TF_LISTEN	0x00002000	/* listen(2) has been called */
163 #define TF_SIGNATURE	0x00004000	/* require MD5 digests (RFC2385) */
164 #define TF_FASTKEEP	0x00008000	/* use a faster tcp_keepidle */
165 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
166 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
167 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
168 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
169 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
170 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
171 #define	TF_FIRSTACCACK	0x00400000	/* Look for 1st acceptable ACK. */
172 #define	TF_FASTREXMT	0x00800000	/* Did Fast Retransmit. */
173 #define	TF_EARLYREXMT	0x01000000	/* Did Early (Fast) Retransmit. */
174 #define	TF_FORCE	0x02000000	/* Set if forcing out a byte */
175 #define TF_ONOUTPUTQ	0x04000000	/* on t_outputq list */
176 #define TF_DUPSEG	0x08000000	/* last seg a duplicate */
177 #define TF_ENCLOSESEG	0x10000000	/* enclosing SACK block */
178 #define TF_SACKLEFT	0x20000000	/* send SACK blocks from left side */
179 #define TF_KEEPALIVE	0x40000000	/* temporary keepalive */
180 #define TF_RXRESIZED	0x80000000	/* rcvbuf was resized */
181 	tcp_seq	snd_up;			/* send urgent pointer */
182 
183 	tcp_seq	snd_una;		/* send unacknowledged */
184 	tcp_seq	snd_recover;		/* for use with NewReno Fast Recovery */
185 	tcp_seq	snd_max;		/* highest sequence number sent;
186 					 * used to recognize retransmits */
187 	tcp_seq	snd_nxt;		/* send next */
188 
189 	tcp_seq	snd_wl1;		/* window update seg seq number */
190 	tcp_seq	snd_wl2;		/* window update seg ack number */
191 	tcp_seq	iss;			/* initial send sequence number */
192 	tcp_seq	irs;			/* initial receive sequence number */
193 
194 	tcp_seq	rcv_nxt;		/* receive next */
195 	tcp_seq	rcv_adv;		/* advertised window */
196 	u_long	rcv_wnd;		/* receive window */
197 	tcp_seq	rcv_up;			/* receive urgent pointer */
198 
199 	u_long	snd_wnd;		/* send window */
200 	u_long	snd_cwnd;		/* congestion-controlled window */
201 	u_long	snd_wacked;		/* bytes acked in one send window */
202 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
203 					 * for slow start exponential to
204 					 * linear switch */
205 
206 	int	t_rxtcur;		/* current retransmit value (ticks) */
207 	u_int	t_maxseg;		/* maximum segment size */
208 	int	t_srtt;			/* smoothed round-trip time */
209 	int	t_rttvar;		/* variance in round-trip time */
210 
211 	u_int	t_maxopd;		/* mss plus options */
212 
213 	u_long	t_rcvtime;		/* inactivity time */
214 	u_long	t_starttime;		/* time connection was established */
215 	int	t_rtttime;		/* round trip time */
216 	tcp_seq	t_rtseq;		/* sequence number being timed */
217 
218 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
219 	u_int	t_rttmin;		/* minimum rtt allowed */
220 	u_int	t_rttbest;		/* best rtt we've seen */
221 	u_long	t_rttupdated;		/* number of times rtt sampled */
222 	u_long	max_sndwnd;		/* largest window peer has offered */
223 
224 	int	t_softerror;		/* possible error not yet reported */
225 /* out-of-band data */
226 	char	t_oobflags;		/* have some */
227 	char	t_iobc;			/* input character */
228 #define	TCPOOB_HAVEDATA	0x01
229 #define	TCPOOB_HADDATA	0x02
230 
231 /* RFC 1323 variables */
232 	u_char	snd_scale;		/* window scaling for send window */
233 	u_char	rcv_scale;		/* window scaling for recv window */
234 	u_char	request_r_scale;	/* pending window scaling */
235 	u_char	requested_s_scale;
236 	u_long	ts_recent;		/* timestamp echo data */
237 
238 	u_long	ts_recent_age;		/* when last updated */
239 	tcp_seq	last_ack_sent;
240 
241 /* experimental */
242 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
243 	u_long	snd_wacked_prev;	/* prior bytes acked in send window */
244 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
245 	tcp_seq snd_recover_prev;	/* snd_recover prior to retransmit */
246 	u_long	t_badrxtwin;		/* window for retransmit recovery */
247 	u_long	t_rexmtTS;		/* timestamp of last retransmit */
248 	u_char	snd_limited;		/* segments limited transmitted */
249 
250 	tcp_seq	rexmt_high;		/* highest seq # retransmitted + 1 */
251 	tcp_seq	snd_max_rexmt;		/* snd_max when rexmting snd_una */
252 	struct scoreboard scb;		/* sack scoreboard */
253 	struct raw_sackblock reportblk; /* incoming segment or D-SACK block */
254 	struct raw_sackblock encloseblk;
255 	int	nsackhistory;
256 	struct raw_sackblock sackhistory[MAX_SACK_REPORT_BLOCKS]; /* reported */
257 	TAILQ_ENTRY(tcpcb) t_outputq;	/* tcp_output needed list */
258 
259 	/* bandwith limitation */
260 	u_long	snd_bandwidth;		/* calculated bandwidth or 0 */
261 	u_long	snd_bwnd;		/* bandwidth-controlled window */
262 	int	t_bw_rtttime;		/* used for bandwidth calculation */
263 	tcp_seq	t_bw_rtseq;		/* used for bandwidth calculation */
264 
265 /* anti DoS counters */
266 	u_long	rcv_second;		/* start of interval second */
267 	u_long	rcv_pps;		/* received packets per second */
268 	u_long	rcv_byps;		/* received bytes per second */
269 
270 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
271 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
272 
273 	int	t_keepinit;		/* time to establish connection */
274 
275 	int	t_keepidle;		/* time before keepalive probes begin */
276 	int	t_keepintvl;		/* time between keepalive probes */
277 	int	t_keepcnt;		/* maximum number of keepalive probes */
278 	int	t_maxidle;		/* time to drop after starting probes */
279 };
280 
281 #define	IN_FASTRECOVERY(tp)	(tp->t_flags & TF_FASTRECOVERY)
282 #define	ENTER_FASTRECOVERY(tp)	tp->t_flags |= TF_FASTRECOVERY
283 #define	EXIT_FASTRECOVERY(tp)	tp->t_flags &= ~TF_FASTRECOVERY
284 
285 #ifdef TCP_SIGNATURE
286 /*
287  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
288  * for SADB verification and lookup.
289  */
290 #define TCP_SIGLEN      16      /* length of computed digest in bytes */
291 #define TCP_KEYLEN_MIN  1       /* minimum length of TCP-MD5 key */
292 #define TCP_KEYLEN_MAX  80      /* maximum length of TCP-MD5 key */
293 /*
294  * Only a single SA per host may be specified at this time. An SPI is
295  * needed in order for the KEY_ALLOCSA() lookup to work.
296  */
297 #define TCP_SIG_SPI     0x1000
298 #endif /* TCP_SIGNATURE */
299 
300 /*
301  * TCP statistics.
302  */
303 struct tcp_stats {
304 	u_long	tcps_connattempt;	/* connections initiated */
305 	u_long	tcps_accepts;		/* connections accepted */
306 	u_long	tcps_connects;		/* connections established */
307 	u_long	tcps_drops;		/* connections dropped */
308 	u_long	tcps_conndrops;		/* embryonic connections dropped */
309 	u_long	tcps_minmssdrops;	/* average minmss too low drops */
310 	u_long	tcps_closed;		/* conn. closed (includes drops) */
311 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
312 	u_long	tcps_rttupdated;	/* times we succeeded */
313 	u_long	tcps_delack;		/* delayed acks sent */
314 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
315 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
316 	u_long	tcps_persisttimeo;	/* persist timeouts */
317 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
318 	u_long	tcps_keepprobe;		/* keepalive probes sent */
319 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
320 
321 	u_long	tcps_sndtotal;		/* total packets sent */
322 	u_long	tcps_sndpack;		/* data packets sent */
323 	u_long	tcps_sndbyte;		/* data bytes sent */
324 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
325 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
326 	u_long	tcps_sndfastrexmit;	/* Fast Retransmissions */
327 	u_long	tcps_sndearlyrexmit;	/* early Fast Retransmissions */
328 	u_long	tcps_sndlimited;	/* Limited Transmit packets */
329 	u_long	tcps_sndrtobad;		/* spurious RTO retransmissions */
330 	u_long	tcps_sndfastrexmitbad;	/* spurious Fast Retransmissions */
331 	u_long	tcps_sndearlyrexmitbad;	/* spurious early Fast Retransmissions,
332 					   a subset of tcps_sndfastrexmitbad */
333 	u_long	tcps_eifeldetected;	/* Eifel-detected spurious rexmits */
334 	u_long	tcps_rttcantdetect;	/* Eifel but not 1/2 RTT-detectable */
335 	u_long	tcps_rttdetected;	/* RTT-detected spurious RTO rexmits */
336 	u_long	tcps_sndacks;		/* ack-only packets sent */
337 	u_long	tcps_sndprobe;		/* window probes sent */
338 	u_long	tcps_sndurg;		/* packets sent with URG only */
339 	u_long	tcps_sndwinup;		/* window update-only packets sent */
340 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
341 	u_long	tcps_sndsackpack;	/* packets sent by SACK recovery alg */
342 	u_long	tcps_sndsackbyte;	/* bytes sent by SACK recovery */
343 	u_long	tcps_snduna3;		/* re-retransmit snd_una on 3 new seg */
344 	u_long	tcps_snduna1;		/* re-retransmit snd_una on 1 new seg */
345 
346 	u_long	tcps_rcvtotal;		/* total packets received */
347 	u_long	tcps_rcvpack;		/* packets received in sequence */
348 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
349 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
350 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
351 	u_long	tcps_rcvmemdrop;	/* packets dropped for lack of memory */
352 	u_long	tcps_rcvshort;		/* packets received too short */
353 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
354 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
355 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
356 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
357 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
358 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
359 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
360 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
361 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
362 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
363 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
364 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
365 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
366 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
367 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
368 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
369 	u_long	tcps_predack;		/* times hdr predict ok for acks */
370 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
371 	u_long	tcps_pcbcachemiss;
372 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
373 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
374 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
375 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
376 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
377 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
378 	u_long	tcps_persistdrop;	/* timeout in persist state */
379 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
380 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
381 	u_long	tcps_listendrop;	/* listen queue overflows */
382 
383 	u_long	tcps_sc_added;		/* entry added to syncache */
384 	u_long	tcps_sc_retransmitted;	/* syncache entry was retransmitted */
385 	u_long	tcps_sc_dupsyn;		/* duplicate SYN packet */
386 	u_long	tcps_sc_dropped;	/* could not reply to packet */
387 	u_long	tcps_sc_completed;	/* successful extraction of entry */
388 	u_long	tcps_sc_bucketoverflow;	/* syncache per-bucket limit hit */
389 	u_long	tcps_sc_cacheoverflow;	/* syncache cache limit hit */
390 	u_long	tcps_sc_reset;		/* RST removed entry from syncache */
391 	u_long	tcps_sc_stale;		/* timed out or listen socket gone */
392 	u_long	tcps_sc_aborted;	/* syncache entry aborted */
393 	u_long	tcps_sc_badack;		/* removed due to bad ACK */
394 	u_long	tcps_sc_unreach;	/* ICMP unreachable received */
395 	u_long	tcps_sc_zonefail;	/* zalloc() failed */
396 	u_long	tcps_sc_sendcookie;	/* SYN cookie sent */
397 	u_long	tcps_sc_recvcookie;	/* SYN cookie received */
398 };
399 
400 #ifdef _KERNEL
401 
402 #if defined(SMP)
403 #define tcpstat	tcpstats_percpu[mycpuid]
404 #else
405 #define tcpstat	tcpstats_percpu[0]
406 #endif
407 
408 struct sockopt;
409 
410 extern struct tcp_stats		tcpstats_percpu[MAXCPU];
411 
412 static const int tcprexmtthresh = 3;
413 #endif
414 
415 /*
416  * Structure to hold TCP options that are only used during segment
417  * processing (in tcp_input), but not held in the tcpcb.
418  * It's basically used to reduce the number of parameters
419  * to tcp_dooptions.
420  */
421 struct tcpopt {
422 	u_long		to_flags;	/* which options are present */
423 #define	TOF_TS			0x0001		/* timestamp */
424 #define	TOF_MSS			0x0010
425 #define	TOF_SCALE		0x0020
426 #define	TOF_SACK_PERMITTED	0x0040
427 #define	TOF_SACK		0x0080
428 #define TOF_SIGNATURE   0x0100          /* signature option present */
429 #define TOF_SIGLEN      0x0200          /* sigature length valid (RFC2385) */
430 	u_int32_t	to_tsval;
431 	u_int32_t	to_tsecr;
432 	u_int16_t	to_mss;
433 	u_int8_t	to_requested_s_scale;
434 	u_int8_t	to_nsackblocks;
435 	struct raw_sackblock *to_sackblocks;
436 };
437 
438 struct syncache {
439 	inp_gen_t	sc_inp_gencnt;		/* pointer check */
440 	struct		tcpcb *sc_tp;		/* tcb for listening socket */
441 	struct		mbuf *sc_ipopts;	/* source route */
442 	struct		in_conninfo sc_inc;	/* addresses */
443 #define sc_route	sc_inc.inc_route
444 #define sc_route6	sc_inc.inc6_route
445 	u_int32_t	sc_tsrecent;
446 	tcp_seq		sc_irs;			/* seq from peer */
447 	tcp_seq		sc_iss;			/* our ISS */
448 	u_long		sc_rxttime;		/* retransmit time */
449 	u_int16_t	sc_rxtslot;		/* retransmit counter */
450 	u_int16_t	sc_peer_mss;		/* peer's MSS */
451 	u_int16_t	sc_wnd;			/* advertised window */
452 	u_int8_t	sc_requested_s_scale:4,
453 			sc_request_r_scale:4;
454 	u_int8_t	sc_flags;
455 #define SCF_NOOPT		0x01		/* no TCP options */
456 #define SCF_WINSCALE		0x02		/* negotiated window scaling */
457 #define SCF_TIMESTAMP		0x04		/* negotiated timestamps */
458 #define SCF_UNREACH		0x10		/* icmp unreachable received */
459 #define	SCF_SACK_PERMITTED	0x20		/* saw SACK permitted option */
460 #define SCF_SIGNATURE   0x40    /* send MD5 digests */
461 #define SCF_MARKER		0x80		/* not a real entry */
462 	TAILQ_ENTRY(syncache) sc_hash;
463 	TAILQ_ENTRY(syncache) sc_timerq;
464 };
465 
466 struct syncache_head {
467 	TAILQ_HEAD(, syncache)	sch_bucket;
468 	u_int		sch_length;
469 };
470 
471 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
472 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
473 
474 /*
475  * The smoothed round-trip time and estimated variance
476  * are stored as fixed point numbers scaled by the values below.
477  * For convenience, these scales are also used in smoothing the average
478  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
479  * With these scales, srtt has 3 bits to the right of the binary point,
480  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
481  * binary point, and is smoothed with an ALPHA of 0.75.
482  */
483 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
484 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
485 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
486 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
487 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
488 
489 /*
490  * The initial retransmission should happen at rtt + 4 * rttvar.
491  * Because of the way we do the smoothing, srtt and rttvar
492  * will each average +1/2 tick of bias.  When we compute
493  * the retransmit timer, we want 1/2 tick of rounding and
494  * 1 extra tick because of +-1/2 tick uncertainty in the
495  * firing of the timer.  The bias will give us exactly the
496  * 1.5 tick we need.  But, because the bias is
497  * statistical, we have to test that we don't drop below
498  * the minimum feasible timer (which is 2 ticks).
499  * This version of the macro adapted from a paper by Lawrence
500  * Brakmo and Larry Peterson which outlines a problem caused
501  * by insufficient precision in the original implementation,
502  * which results in inappropriately large RTO values for very
503  * fast networks.
504  */
505 #define	TCP_REXMTVAL(tp) \
506 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
507 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
508 
509 /*
510  * TCB structure exported to user-land via sysctl(3).
511  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
512  * included.  Not all of our clients do.
513  */
514 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
515 struct	xtcpcb {
516 	size_t	xt_len;
517 	struct	inpcb	xt_inp;
518 	struct	tcpcb	xt_tp;
519 	struct	xsocket	xt_socket;
520 	u_quad_t	xt_alignment_hack;
521 };
522 #endif
523 
524 /*
525  * Names for TCP sysctl objects
526  */
527 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
528 /* 2 was TCPCTL_DO_RFC1644 */
529 #define	TCPCTL_MSSDFLT		3	/* MSS default */
530 #define TCPCTL_STATS		4	/* statistics (read-only) */
531 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
532 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
533 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
534 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
535 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
536 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
537 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
538 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
539 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
540 #define	TCPCTL_MAXID		14
541 
542 #define TCPCTL_NAMES { \
543 	{ 0, 0 }, \
544 	{ "rfc1323", CTLTYPE_INT }, \
545 	  { "reserved", CTLTYPE_INT},	/* was rfc1644 */	\
546 	{ "mssdflt", CTLTYPE_INT }, \
547 	{ "stats", CTLTYPE_STRUCT }, \
548 	{ "rttdflt", CTLTYPE_INT }, \
549 	{ "keepidle", CTLTYPE_INT }, \
550 	{ "keepintvl", CTLTYPE_INT }, \
551 	{ "sendspace", CTLTYPE_INT }, \
552 	{ "recvspace", CTLTYPE_INT }, \
553 	{ "keepinit", CTLTYPE_INT }, \
554 	{ "pcblist", CTLTYPE_STRUCT }, \
555 	{ "delacktime", CTLTYPE_INT }, \
556 	{ "v6mssdflt", CTLTYPE_INT }, \
557 }
558 
559 #ifdef _KERNEL
560 #ifdef SYSCTL_DECL
561 SYSCTL_DECL(_net_inet_tcp);
562 #endif
563 
564 #define TCP_DO_SACK(tp)		((tp)->t_flags & TF_SACK_PERMITTED)
565 
566 TAILQ_HEAD(tcpcbackqhead,tcpcb);
567 
568 extern	struct inpcbinfo tcbinfo[];
569 extern	struct tcpcbackqhead tcpcbackq[];
570 
571 extern	int tcp_mssdflt;	/* XXX */
572 extern	int tcp_minmss;
573 extern	int tcp_delack_enabled;
574 extern	int path_mtu_discovery;
575 
576 union netmsg;
577 
578 int	 tcp_addrcpu(in_addr_t faddr, in_port_t fport,
579 	    in_addr_t laddr, in_port_t lport);
580 struct lwkt_port *
581 	tcp_addrport(in_addr_t faddr, in_port_t fport,
582 	    in_addr_t laddr, in_port_t lport);
583 struct lwkt_port *tcp_addrport0(void);
584 void	 tcp_canceltimers (struct tcpcb *);
585 struct tcpcb *
586 	 tcp_close (struct tcpcb *);
587 void	 tcp_ctlinput(union netmsg *);
588 void	 tcp_ctloutput(union netmsg *);
589 struct tcpcb *
590 	 tcp_drop (struct tcpcb *, int);
591 void	 tcp_drain (void);
592 void	 tcp_fasttimo (void);
593 void	 tcp_init (void);
594 void	 tcp_thread_init (void);
595 int	 tcp_input (struct mbuf **, int *, int);
596 void	 tcp_mss (struct tcpcb *, int);
597 int	 tcp_mssopt (struct tcpcb *);
598 void	 tcp_drop_syn_sent (struct inpcb *, int);
599 void	 tcp_mtudisc (struct inpcb *, int);
600 struct tcpcb *
601 	 tcp_newtcpcb (struct inpcb *);
602 int	 tcp_output (struct tcpcb *);
603 void	 tcp_quench (struct inpcb *, int);
604 void	 tcp_respond (struct tcpcb *, void *,
605 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
606 struct rtentry *
607 	 tcp_rtlookup (struct in_conninfo *);
608 int	 tcp_sack_bytes_below(struct scoreboard *scb, tcp_seq seq);
609 void	 tcp_sack_cleanup(struct scoreboard *scb);
610 int	 tcp_sack_ndsack_blocks(struct raw_sackblock *blocks,
611 	    const int numblocks, tcp_seq snd_una);
612 void	 tcp_sack_fill_report(struct tcpcb *tp, u_char *opt, u_int *plen);
613 boolean_t
614 	 tcp_sack_has_sacked(struct scoreboard *scb, u_int amount);
615 void	 tcp_sack_tcpcb_init(struct tcpcb *tp);
616 uint32_t tcp_sack_compute_pipe(struct tcpcb *tp);
617 boolean_t
618 	 tcp_sack_nextseg(struct tcpcb *tp, tcp_seq *nextrexmt, uint32_t *len,
619 			  boolean_t *losdup);
620 #ifdef later
621 void	 tcp_sack_revert_scoreboard(struct scoreboard *scb, tcp_seq snd_una,
622 				    u_int maxseg);
623 void	 tcp_sack_save_scoreboard(struct scoreboard *scb);
624 #endif
625 void	 tcp_sack_skip_sacked(struct scoreboard *scb, tcp_seq *prexmt);
626 void	 tcp_sack_update_scoreboard(struct tcpcb *tp, struct tcpopt *to);
627 void	 tcp_save_congestion_state(struct tcpcb *tp);
628 void	 tcp_revert_congestion_state(struct tcpcb *tp);
629 void	 tcp_setpersist (struct tcpcb *);
630 struct tcptemp *tcp_maketemplate (struct tcpcb *);
631 void	 tcp_freetemplate (struct tcptemp *);
632 void	 tcp_fillheaders (struct tcpcb *, void *, void *);
633 struct lwkt_port *
634 	 tcp_soport(struct socket *, struct sockaddr *, struct mbuf **);
635 struct lwkt_port *
636 	 tcp_ctlport(int, struct sockaddr *, void *);
637 struct tcpcb *
638 	 tcp_timers (struct tcpcb *, int);
639 void	 tcp_trace (short, short, struct tcpcb *, void *, struct tcphdr *,
640 			int);
641 void	 tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
642 void	 tcp_timer_keep_activity(struct tcpcb *tp, int thflags);
643 void	 syncache_init(void);
644 void	 syncache_unreach(struct in_conninfo *, struct tcphdr *);
645 int	 syncache_expand(struct in_conninfo *, struct tcphdr *,
646 	     struct socket **, struct mbuf *);
647 int	 syncache_add(struct in_conninfo *, struct tcpopt *,
648 	     struct tcphdr *, struct socket **, struct mbuf *);
649 void	 syncache_chkrst(struct in_conninfo *, struct tcphdr *);
650 void	 syncache_badack(struct in_conninfo *);
651 void	 syncache_destroy(struct tcpcb *tp);
652 
653 #ifdef TCP_SIGNATURE
654 int tcpsignature_apply(void *fstate, void *data, unsigned int len);
655 int tcpsignature_compute(struct mbuf *m, int len, int tcpoptlen,
656 		u_char *buf, u_int direction);
657 #endif /* TCP_SIGNATURE */
658 
659 extern	struct pr_usrreqs tcp_usrreqs;
660 extern	u_long tcp_sendspace;
661 extern	u_long tcp_recvspace;
662 tcp_seq tcp_new_isn (struct tcpcb *);
663 
664 #endif /* _KERNEL */
665 
666 #endif /* _NETINET_TCP_VAR_H_ */
667