xref: /dflybsd-src/sys/netinet/tcp_var.h (revision 56006bc89a9ac0796d957ae2ebe8823d77bf2a3b)
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_low_rtobase;
86 extern int tcp_do_sack;
87 extern int tcp_do_smartsack;
88 extern int tcp_do_rescuesack;
89 extern int tcp_aggressive_rescuesack;
90 extern int tcp_aggregate_acks;
91 extern int tcp_eifel_rtoinc;
92 
93 /* TCP segment queue entry */
94 struct tseg_qent {
95 	LIST_ENTRY(tseg_qent) tqe_q;
96 	int	tqe_len;		/* TCP segment data length */
97 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
98 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
99 };
100 LIST_HEAD(tsegqe_head, tseg_qent);
101 extern int	tcp_reass_maxseg;
102 extern int	tcp_reass_qsize;
103 #ifdef MALLOC_DECLARE
104 MALLOC_DECLARE(M_TSEGQ);
105 #endif
106 
107 struct tcptemp {
108 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
109 	struct	tcphdr tt_t;
110 };
111 
112 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
113 
114 struct raw_sackblock {				/* covers [start, end) */
115 	tcp_seq rblk_start;
116 	tcp_seq rblk_end;
117 };
118 
119 /* maximum number of SACK blocks that will fit in the TCP option space */
120 #define	MAX_SACK_REPORT_BLOCKS	4
121 
122 TAILQ_HEAD(sackblock_list, sackblock);
123 
124 struct scoreboard {
125 	int nblocks;
126 	struct sackblock_list sackblocks;
127 	tcp_seq lostseq;			/* passed SACK lost test */
128 	struct sackblock *lastfound;		/* search hint */
129 	struct sackblock *freecache;		/* one slot free block cache */
130 };
131 
132 struct netmsg_tcp_timer;
133 
134 /*
135  * Tcp control block, one per tcp; fields:
136  * Organized for 16 byte cacheline efficiency.
137  */
138 struct tcpcb {
139 	struct	tsegqe_head t_segq;
140 	int	t_dupacks;		/* consecutive dup acks recd */
141 	int	tt_cpu;			/* sanity check the cpu */
142 
143 	struct	tcp_callout *tt_rexmt;	/* retransmit timer */
144 	struct	tcp_callout *tt_persist;/* retransmit persistence */
145 	struct	tcp_callout *tt_keep;	/* keepalive */
146 	struct	tcp_callout *tt_2msl;	/* 2*msl TIME_WAIT timer */
147 	struct	tcp_callout *tt_delack;	/* delayed ACK timer */
148 	struct	netmsg_tcp_timer *tt_msg; /* timer message */
149 
150 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
151 	int	t_state;		/* state of this connection */
152 	u_int	t_flags;
153 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
154 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
155 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
156 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
157 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
158 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
159 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
160 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
161 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
162 #define	TF_SACK_PERMITTED 0x00000200	/* other side said I could SACK */
163 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
164 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
165 #define	TF_NOPUSH	0x00001000	/* don't push */
166 #define TF_LISTEN	0x00002000	/* listen(2) has been called */
167 #define TF_SIGNATURE	0x00004000	/* require MD5 digests (RFC2385) */
168 #define TF_SACKRESCUED	0x00008000	/* sent rescue SACK recovery data */
169 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
170 #define	TF_REBASERTO	0x00020000	/* Recalculate RTO based on new RTT */
171 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
172 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
173 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
174 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
175 #define	TF_FIRSTACCACK	0x00400000	/* Look for 1st acceptable ACK. */
176 #define	TF_FASTREXMT	0x00800000	/* Did Fast Retransmit. */
177 #define	TF_EARLYREXMT	0x01000000	/* Did Early (Fast) Retransmit. */
178 #define	TF_FORCE	0x02000000	/* Set if forcing out a byte */
179 #define TF_ONOUTPUTQ	0x04000000	/* on t_outputq list */
180 #define TF_DUPSEG	0x08000000	/* last seg a duplicate */
181 #define TF_ENCLOSESEG	0x10000000	/* enclosing SACK block */
182 #define TF_SACKLEFT	0x20000000	/* send SACK blocks from left side */
183 #define TF_KEEPALIVE	0x40000000	/* temporary keepalive */
184 #define TF_RXRESIZED	0x80000000	/* rcvbuf was resized */
185 	tcp_seq	snd_up;			/* send urgent pointer */
186 	u_long	snd_last;		/* time last data were sent */
187 
188 	tcp_seq	snd_una;		/* send unacknowledged */
189 	tcp_seq	snd_recover;		/* for use with NewReno Fast Recovery */
190 	tcp_seq	snd_max;		/* highest sequence number sent;
191 					 * used to recognize retransmits */
192 	tcp_seq	snd_nxt;		/* send next */
193 
194 	tcp_seq	snd_wl1;		/* window update seg seq number */
195 	tcp_seq	snd_wl2;		/* window update seg ack number */
196 	tcp_seq	iss;			/* initial send sequence number */
197 	tcp_seq	irs;			/* initial receive sequence number */
198 
199 	tcp_seq	rcv_nxt;		/* receive next */
200 	tcp_seq	rcv_adv;		/* advertised window */
201 	u_long	rcv_wnd;		/* receive window */
202 	tcp_seq	rcv_up;			/* receive urgent pointer */
203 
204 	u_long	snd_wnd;		/* send window */
205 	u_long	snd_cwnd;		/* congestion-controlled window */
206 	u_long	snd_wacked;		/* bytes acked in one send window */
207 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
208 					 * for slow start exponential to
209 					 * linear switch */
210 
211 	int	t_rxtcur;		/* current retransmit value (ticks) */
212 	u_int	t_maxseg;		/* maximum segment size */
213 	int	t_srtt;			/* smoothed round-trip time */
214 	int	t_rttvar;		/* variance in round-trip time */
215 
216 	u_int	t_maxopd;		/* mss plus options */
217 
218 	u_long	t_rcvtime;		/* reception inactivity time */
219 	u_long	t_starttime;		/* time connection was established */
220 	int	t_rtttime;		/* round trip time */
221 	tcp_seq	t_rtseq;		/* sequence number being timed */
222 
223 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
224 	u_int	t_rttmin;		/* minimum rtt allowed */
225 	u_int	t_rttbest;		/* best rtt we've seen */
226 	u_long	t_rttupdated;		/* number of times rtt sampled */
227 	u_long	max_sndwnd;		/* largest window peer has offered */
228 
229 	int	t_softerror;		/* possible error not yet reported */
230 /* out-of-band data */
231 	char	t_oobflags;		/* have some */
232 	char	t_iobc;			/* input character */
233 #define	TCPOOB_HAVEDATA	0x01
234 #define	TCPOOB_HADDATA	0x02
235 
236 /* RFC 1323 variables */
237 	u_char	snd_scale;		/* window scaling for send window */
238 	u_char	rcv_scale;		/* window scaling for recv window */
239 	u_char	request_r_scale;	/* pending window scaling */
240 	u_long	ts_recent;		/* timestamp echo data */
241 
242 	u_long	ts_recent_age;		/* when last updated */
243 	tcp_seq	last_ack_sent;
244 
245 /* experimental */
246 	int	t_srtt_prev;		/* adjusted SRTT prior to retransmit */
247 	int	t_rttvar_prev;		/* RTTVAR prior to retransmit */
248 	int	t_rxtcur_prev;		/* rexmt timeout prior to retransmit */
249 	tcp_seq	snd_max_prev;		/* SND_MAX prior to retransmit */
250 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
251 	u_long	snd_wacked_prev;	/* prior bytes acked in send window */
252 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
253 	tcp_seq snd_recover_prev;	/* snd_recover prior to retransmit */
254 	u_long	t_badrxtwin;		/* window for retransmit recovery */
255 	u_long	t_rexmtTS;		/* timestamp of last retransmit */
256 	u_char	snd_limited;		/* segments limited transmitted */
257 
258 	tcp_seq	rexmt_high;		/* highest seq # retransmitted + 1 */
259 	tcp_seq	rexmt_rescue;		/* rescue SACKED sequence number */
260 	tcp_seq	snd_max_rexmt;		/* snd_max when rexmting snd_una */
261 	struct scoreboard scb;		/* sack scoreboard */
262 	struct raw_sackblock reportblk; /* incoming segment or D-SACK block */
263 	struct raw_sackblock encloseblk;
264 	int	nsackhistory;
265 	struct raw_sackblock sackhistory[MAX_SACK_REPORT_BLOCKS]; /* reported */
266 	TAILQ_ENTRY(tcpcb) t_outputq;	/* tcp_output needed list */
267 
268 	/* bandwith limitation */
269 	u_long	snd_bandwidth;		/* calculated bandwidth or 0 */
270 	u_long	snd_bwnd;		/* bandwidth-controlled window */
271 	int	t_bw_rtttime;		/* used for bandwidth calculation */
272 	tcp_seq	t_bw_rtseq;		/* used for bandwidth calculation */
273 
274 /* anti DoS counters */
275 	u_long	rcv_second;		/* start of interval second */
276 	u_long	rcv_pps;		/* received packets per second */
277 	u_long	rcv_byps;		/* received bytes per second */
278 
279 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
280 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
281 
282 	int	t_keepinit;		/* time to establish connection */
283 
284 	int	t_keepidle;		/* time before keepalive probes begin */
285 	int	t_keepintvl;		/* time between keepalive probes */
286 	int	t_keepcnt;		/* maximum number of keepalive probes */
287 	int	t_maxidle;		/* time to drop after starting probes */
288 
289 	int	t_rxtsyn;		/* time spent in SYN or SYN|ACK rexmt */
290 };
291 
292 #define	IN_FASTRECOVERY(tp)	(tp->t_flags & TF_FASTRECOVERY)
293 #define	ENTER_FASTRECOVERY(tp)	tp->t_flags |= TF_FASTRECOVERY
294 #define	EXIT_FASTRECOVERY(tp)	tp->t_flags &= ~TF_FASTRECOVERY
295 
296 #ifdef TCP_SIGNATURE
297 /*
298  * Defines which are needed by the xform_tcp module and tcp_[in|out]put
299  * for SADB verification and lookup.
300  */
301 #define TCP_SIGLEN      16      /* length of computed digest in bytes */
302 #define TCP_KEYLEN_MIN  1       /* minimum length of TCP-MD5 key */
303 #define TCP_KEYLEN_MAX  80      /* maximum length of TCP-MD5 key */
304 /*
305  * Only a single SA per host may be specified at this time. An SPI is
306  * needed in order for the KEY_ALLOCSA() lookup to work.
307  */
308 #define TCP_SIG_SPI     0x1000
309 #endif /* TCP_SIGNATURE */
310 
311 /*
312  * TCP statistics.
313  */
314 struct tcp_stats {
315 	u_long	tcps_connattempt;	/* connections initiated */
316 	u_long	tcps_accepts;		/* connections accepted */
317 	u_long	tcps_connects;		/* connections established */
318 	u_long	tcps_drops;		/* connections dropped */
319 	u_long	tcps_conndrops;		/* embryonic connections dropped */
320 	u_long	tcps_minmssdrops;	/* average minmss too low drops */
321 	u_long	tcps_closed;		/* conn. closed (includes drops) */
322 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
323 	u_long	tcps_rttupdated;	/* times we succeeded */
324 	u_long	tcps_delack;		/* delayed acks sent */
325 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
326 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
327 	u_long	tcps_persisttimeo;	/* persist timeouts */
328 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
329 	u_long	tcps_keepprobe;		/* keepalive probes sent */
330 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
331 
332 	u_long	tcps_sndtotal;		/* total packets sent */
333 	u_long	tcps_sndpack;		/* data packets sent */
334 	u_long	tcps_sndbyte;		/* data bytes sent */
335 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
336 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
337 	u_long	tcps_sndsackrtopack;	/* packets sent by SACK after RTO */
338 	u_long	tcps_sndsackrtobyte;	/* bytes sent by SACK after RTO */
339 	u_long	tcps_sndfastrexmit;	/* Fast Retransmissions */
340 	u_long	tcps_sndearlyrexmit;	/* early Fast Retransmissions */
341 	u_long	tcps_sndlimited;	/* Limited Transmit packets */
342 	u_long	tcps_sndrtobad;		/* spurious RTO retransmissions */
343 	u_long	tcps_sndfastrexmitbad;	/* spurious Fast Retransmissions */
344 	u_long	tcps_sndearlyrexmitbad;	/* spurious early Fast Retransmissions,
345 					   a subset of tcps_sndfastrexmitbad */
346 	u_long	tcps_eifeldetected;	/* Eifel-detected spurious rexmits */
347 	u_long	tcps_rttcantdetect;	/* Eifel but not 1/2 RTT-detectable */
348 	u_long	tcps_rttdetected;	/* RTT-detected spurious RTO rexmits */
349 	u_long	tcps_sndacks;		/* ack-only packets sent */
350 	u_long	tcps_sndprobe;		/* window probes sent */
351 	u_long	tcps_sndurg;		/* packets sent with URG only */
352 	u_long	tcps_sndwinup;		/* window update-only packets sent */
353 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
354 	u_long	tcps_sndsackpack;	/* packets sent by SACK recovery alg */
355 	u_long	tcps_sndsackbyte;	/* bytes sent by SACK recovery */
356 	u_long	tcps_snduna3;		/* re-retransmit snd_una on 3 new seg */
357 	u_long	tcps_snduna1;		/* re-retransmit snd_una on 1 new seg */
358 	u_long	tcps_sndsackopt;	/* SACK options sent */
359 	u_long	tcps_snddsackopt;	/* D-SACK options sent */
360 	u_long	tcps_sndidle;		/* sending idle detected */
361 	u_long	tcps_sackrescue;	/* SACK rescue data packets sent */
362 	u_long	tcps_sackrescue_try;	/* SACK rescues attempted */
363 	u_long	tcps_eifelresponse;	/* Eifel responses */
364 
365 	u_long	tcps_rcvtotal;		/* total packets received */
366 	u_long	tcps_rcvpack;		/* packets received in sequence */
367 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
368 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
369 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
370 	u_long	tcps_rcvmemdrop;	/* packets dropped for lack of memory */
371 	u_long	tcps_rcvshort;		/* packets received too short */
372 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
373 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
374 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
375 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
376 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
377 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
378 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
379 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
380 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
381 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
382 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
383 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
384 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
385 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
386 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
387 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
388 	u_long	tcps_predack;		/* times hdr predict ok for acks */
389 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
390 	u_long	tcps_pcbcachemiss;
391 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
392 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
393 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
394 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
395 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
396 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
397 	u_long	tcps_persistdrop;	/* timeout in persist state */
398 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
399 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
400 	u_long	tcps_listendrop;	/* listen queue overflows */
401 	u_long	tcps_rcvbadsackopt;	/* rcvd illegal SACK options */
402 
403 	u_long	tcps_sacksbupdate;	/* times SACK scoreboard updated */
404 	u_long	tcps_sacksboverflow;	/* times SACK scoreboard overflowed */
405 	u_long	tcps_sacksbreused;	/* times SACK sb-block reused */
406 	u_long	tcps_sacksbfailed;	/* times SACK sb update failed */
407 	u_long	tcps_sacksbfast;	/* timee SACK sb-block uses cache */
408 
409 	u_long	tcps_sc_added;		/* entry added to syncache */
410 	u_long	tcps_sc_retransmitted;	/* syncache entry was retransmitted */
411 	u_long	tcps_sc_dupsyn;		/* duplicate SYN packet */
412 	u_long	tcps_sc_dropped;	/* could not reply to packet */
413 	u_long	tcps_sc_completed;	/* successful extraction of entry */
414 	u_long	tcps_sc_bucketoverflow;	/* syncache per-bucket limit hit */
415 	u_long	tcps_sc_cacheoverflow;	/* syncache cache limit hit */
416 	u_long	tcps_sc_reset;		/* RST removed entry from syncache */
417 	u_long	tcps_sc_stale;		/* timed out or listen socket gone */
418 	u_long	tcps_sc_aborted;	/* syncache entry aborted */
419 	u_long	tcps_sc_badack;		/* removed due to bad ACK */
420 	u_long	tcps_sc_unreach;	/* ICMP unreachable received */
421 	u_long	tcps_sc_zonefail;	/* zalloc() failed */
422 	u_long	tcps_sc_sendcookie;	/* SYN cookie sent */
423 	u_long	tcps_sc_recvcookie;	/* SYN cookie received */
424 };
425 
426 #ifdef _KERNEL
427 
428 #if defined(SMP)
429 #define tcpstat	tcpstats_percpu[mycpuid]
430 #else
431 #define tcpstat	tcpstats_percpu[0]
432 #endif
433 
434 struct sockopt;
435 
436 extern struct tcp_stats		tcpstats_percpu[MAXCPU];
437 
438 static const int tcprexmtthresh = 3;
439 #endif
440 
441 /*
442  * Structure to hold TCP options that are only used during segment
443  * processing (in tcp_input), but not held in the tcpcb.
444  * It's basically used to reduce the number of parameters
445  * to tcp_dooptions.
446  */
447 struct tcpopt {
448 	u_long		to_flags;	/* which options are present */
449 #define	TOF_TS			0x0001	/* timestamp */
450 #define	TOF_MSS			0x0010
451 #define	TOF_SCALE		0x0020
452 #define	TOF_SACK_PERMITTED	0x0040
453 #define	TOF_SACK		0x0080
454 #define TOF_SIGNATURE		0x0100	/* signature option present */
455 #define TOF_SIGLEN		0x0200	/* signature length valid (RFC2385) */
456 #define TOF_DSACK		0x0400	/* D-SACK */
457 #define TOF_SACK_REDUNDANT	0x0800	/* all SACK blocks are known */
458 	u_int32_t	to_tsval;
459 	u_int32_t	to_tsecr;
460 	u_int16_t	to_mss;
461 	u_int8_t	to_requested_s_scale;
462 	u_int8_t	to_nsackblocks;
463 	struct raw_sackblock *to_sackblocks;
464 };
465 
466 struct syncache {
467 	inp_gen_t	sc_inp_gencnt;		/* pointer check */
468 	struct		tcpcb *sc_tp;		/* tcb for listening socket */
469 	struct		mbuf *sc_ipopts;	/* source route */
470 	struct		in_conninfo sc_inc;	/* addresses */
471 #define sc_route	sc_inc.inc_route
472 #define sc_route6	sc_inc.inc6_route
473 	u_int32_t	sc_tsrecent;
474 	tcp_seq		sc_irs;			/* seq from peer */
475 	tcp_seq		sc_iss;			/* our ISS */
476 	u_long		sc_rxttime;		/* retransmit time */
477 	u_int16_t	sc_rxtslot;		/* retransmit counter */
478 	u_int16_t	sc_peer_mss;		/* peer's MSS */
479 	u_int16_t	sc_wnd;			/* advertised window */
480 	u_int8_t	sc_requested_s_scale:4,
481 			sc_request_r_scale:4;
482 	u_int8_t	sc_flags;
483 #define SCF_NOOPT		0x01		/* no TCP options */
484 #define SCF_WINSCALE		0x02		/* negotiated window scaling */
485 #define SCF_TIMESTAMP		0x04		/* negotiated timestamps */
486 #define SCF_UNUSED		0x08		/* unused */
487 #define SCF_UNREACH		0x10		/* icmp unreachable received */
488 #define	SCF_SACK_PERMITTED	0x20		/* saw SACK permitted option */
489 #define SCF_SIGNATURE		0x40		/* send MD5 digests */
490 #define SCF_MARKER		0x80		/* not a real entry */
491 	int		sc_rxtused;		/* time spent in SYN|ACK rxt */
492 	u_long		sc_sndwnd;		/* send window */
493 	TAILQ_ENTRY(syncache) sc_hash;
494 	TAILQ_ENTRY(syncache) sc_timerq;
495 };
496 
497 struct syncache_head {
498 	TAILQ_HEAD(, syncache)	sch_bucket;
499 	u_int		sch_length;
500 };
501 
502 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
503 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
504 
505 /*
506  * The smoothed round-trip time and estimated variance
507  * are stored as fixed point numbers scaled by the values below.
508  * For convenience, these scales are also used in smoothing the average
509  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
510  * With these scales, srtt has 3 bits to the right of the binary point,
511  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
512  * binary point, and is smoothed with an ALPHA of 0.75.
513  */
514 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
515 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
516 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
517 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
518 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
519 
520 /*
521  * The initial retransmission should happen at rtt + 4 * rttvar.
522  * Because of the way we do the smoothing, srtt and rttvar
523  * will each average +1/2 tick of bias.  When we compute
524  * the retransmit timer, we want 1/2 tick of rounding and
525  * 1 extra tick because of +-1/2 tick uncertainty in the
526  * firing of the timer.  The bias will give us exactly the
527  * 1.5 tick we need.  But, because the bias is
528  * statistical, we have to test that we don't drop below
529  * the minimum feasible timer (which is 2 ticks).
530  * This version of the macro adapted from a paper by Lawrence
531  * Brakmo and Larry Peterson which outlines a problem caused
532  * by insufficient precision in the original implementation,
533  * which results in inappropriately large RTO values for very
534  * fast networks.
535  */
536 #define	TCP_REXMTVAL(tp) \
537 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
538 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
539 
540 /*
541  * TCB structure exported to user-land via sysctl(3).
542  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
543  * included.  Not all of our clients do.
544  */
545 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
546 struct	xtcpcb {
547 	size_t	xt_len;
548 	struct	inpcb	xt_inp;
549 	struct	tcpcb	xt_tp;
550 	struct	xsocket	xt_socket;
551 	u_quad_t	xt_alignment_hack;
552 };
553 #endif
554 
555 /*
556  * Names for TCP sysctl objects
557  */
558 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
559 /* 2 was TCPCTL_DO_RFC1644 */
560 #define	TCPCTL_MSSDFLT		3	/* MSS default */
561 #define TCPCTL_STATS		4	/* statistics (read-only) */
562 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
563 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
564 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
565 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
566 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
567 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
568 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
569 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
570 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
571 #define	TCPCTL_MAXID		14
572 
573 #define TCPCTL_NAMES { \
574 	{ 0, 0 }, \
575 	{ "rfc1323", CTLTYPE_INT }, \
576 	  { "reserved", CTLTYPE_INT},	/* was rfc1644 */	\
577 	{ "mssdflt", CTLTYPE_INT }, \
578 	{ "stats", CTLTYPE_STRUCT }, \
579 	{ "rttdflt", CTLTYPE_INT }, \
580 	{ "keepidle", CTLTYPE_INT }, \
581 	{ "keepintvl", CTLTYPE_INT }, \
582 	{ "sendspace", CTLTYPE_INT }, \
583 	{ "recvspace", CTLTYPE_INT }, \
584 	{ "keepinit", CTLTYPE_INT }, \
585 	{ "pcblist", CTLTYPE_STRUCT }, \
586 	{ "delacktime", CTLTYPE_INT }, \
587 	{ "v6mssdflt", CTLTYPE_INT }, \
588 }
589 
590 #ifdef _KERNEL
591 #ifdef SYSCTL_DECL
592 SYSCTL_DECL(_net_inet_tcp);
593 #endif
594 
595 #define TCP_DO_SACK(tp)		((tp)->t_flags & TF_SACK_PERMITTED)
596 #define TCP_SACK_BLKEND(len, thflags) \
597 	((len) + (((thflags) & TH_FIN) != 0))
598 
599 TAILQ_HEAD(tcpcbackqhead,tcpcb);
600 
601 extern	struct inpcbinfo tcbinfo[];
602 extern	struct tcpcbackqhead tcpcbackq[];
603 
604 extern	int tcp_mssdflt;	/* XXX */
605 extern	int tcp_minmss;
606 extern	int tcp_delack_enabled;
607 extern	int path_mtu_discovery;
608 
609 union netmsg;
610 
611 int	 tcp_addrcpu(in_addr_t faddr, in_port_t fport,
612 	    in_addr_t laddr, in_port_t lport);
613 struct lwkt_port *
614 	tcp_addrport(in_addr_t faddr, in_port_t fport,
615 	    in_addr_t laddr, in_port_t lport);
616 struct lwkt_port *tcp_addrport0(void);
617 void	 tcp_canceltimers (struct tcpcb *);
618 struct tcpcb *
619 	 tcp_close (struct tcpcb *);
620 void	 tcp_ctlinput(union netmsg *);
621 void	 tcp_ctloutput(union netmsg *);
622 struct tcpcb *
623 	 tcp_drop (struct tcpcb *, int);
624 void	 tcp_drain (void);
625 void	 tcp_fasttimo (void);
626 void	 tcp_init (void);
627 void	 tcp_thread_init (void);
628 int	 tcp_input (struct mbuf **, int *, int);
629 void	 tcp_mss (struct tcpcb *, int);
630 int	 tcp_mssopt (struct tcpcb *);
631 void	 tcp_drop_syn_sent (struct inpcb *, int);
632 void	 tcp_mtudisc (struct inpcb *, int);
633 struct tcpcb *
634 	 tcp_newtcpcb (struct inpcb *);
635 int	 tcp_output (struct tcpcb *);
636 void	 tcp_quench (struct inpcb *, int);
637 void	 tcp_respond (struct tcpcb *, void *,
638 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
639 struct rtentry *
640 	 tcp_rtlookup (struct in_conninfo *);
641 int	 tcp_sack_bytes_below(struct scoreboard *scb, tcp_seq seq);
642 void	 tcp_sack_destroy(struct scoreboard *scb);
643 void	 tcp_sack_cleanup(struct scoreboard *scb);
644 void	 tcp_sack_report_cleanup(struct tcpcb *tp);
645 int	 tcp_sack_ndsack_blocks(struct raw_sackblock *blocks,
646 	    const int numblocks, tcp_seq snd_una);
647 void	 tcp_sack_fill_report(struct tcpcb *tp, u_char *opt, u_int *plen);
648 boolean_t
649 	 tcp_sack_has_sacked(struct scoreboard *scb, u_int amount);
650 void	 tcp_sack_tcpcb_init(struct tcpcb *tp);
651 uint32_t tcp_sack_compute_pipe(struct tcpcb *tp);
652 boolean_t
653 	 tcp_sack_nextseg(struct tcpcb *tp, tcp_seq *nextrexmt, uint32_t *len,
654 			  boolean_t *rescue);
655 #ifdef later
656 void	 tcp_sack_revert_scoreboard(struct scoreboard *scb, tcp_seq snd_una,
657 				    u_int maxseg);
658 void	 tcp_sack_save_scoreboard(struct scoreboard *scb);
659 #endif
660 void	 tcp_sack_skip_sacked(struct scoreboard *scb, tcp_seq *prexmt);
661 void	 tcp_sack_update_scoreboard(struct tcpcb *tp, struct tcpopt *to);
662 void	 tcp_save_congestion_state(struct tcpcb *tp);
663 void	 tcp_revert_congestion_state(struct tcpcb *tp);
664 void	 tcp_setpersist (struct tcpcb *);
665 struct tcptemp *tcp_maketemplate (struct tcpcb *);
666 void	 tcp_freetemplate (struct tcptemp *);
667 void	 tcp_fillheaders (struct tcpcb *, void *, void *);
668 struct lwkt_port *
669 	 tcp_soport(struct socket *, struct sockaddr *, struct mbuf **);
670 struct lwkt_port *
671 	 tcp_ctlport(int, struct sockaddr *, void *);
672 struct tcpcb *
673 	 tcp_timers (struct tcpcb *, int);
674 void	 tcp_trace (short, short, struct tcpcb *, void *, struct tcphdr *,
675 			int);
676 void	 tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
677 u_long	 tcp_initial_window(struct tcpcb *tp);
678 void	 tcp_timer_keep_activity(struct tcpcb *tp, int thflags);
679 void	 syncache_init(void);
680 void	 syncache_unreach(struct in_conninfo *, struct tcphdr *);
681 int	 syncache_expand(struct in_conninfo *, struct tcphdr *,
682 	     struct socket **, struct mbuf *);
683 int	 syncache_add(struct in_conninfo *, struct tcpopt *,
684 	     struct tcphdr *, struct socket *, struct mbuf *);
685 void	 syncache_chkrst(struct in_conninfo *, struct tcphdr *);
686 void	 syncache_badack(struct in_conninfo *);
687 void	 syncache_destroy(struct tcpcb *tp);
688 
689 #ifdef TCP_SIGNATURE
690 int tcpsignature_apply(void *fstate, void *data, unsigned int len);
691 int tcpsignature_compute(struct mbuf *m, int len, int tcpoptlen,
692 		u_char *buf, u_int direction);
693 #endif /* TCP_SIGNATURE */
694 
695 extern	struct pr_usrreqs tcp_usrreqs;
696 extern	u_long tcp_sendspace;
697 extern	u_long tcp_recvspace;
698 tcp_seq tcp_new_isn (struct tcpcb *);
699 
700 #endif /* _KERNEL */
701 
702 #endif /* _NETINET_TCP_VAR_H_ */
703