xref: /dragonfly/sys/netinet/tcp_var.h (revision 73e0051e)
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_rfc1644;
86 extern int tcp_do_rfc3390;
87 extern int tcp_do_sack;
88 extern int tcp_do_smartsack;
89 extern int tcp_aggregate_acks;
90 
91 /* TCP segment queue entry */
92 struct tseg_qent {
93 	LIST_ENTRY(tseg_qent) tqe_q;
94 	int	tqe_len;		/* TCP segment data length */
95 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
96 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
97 };
98 LIST_HEAD(tsegqe_head, tseg_qent);
99 extern int	tcp_reass_maxseg;
100 extern int	tcp_reass_qsize;
101 #ifdef MALLOC_DECLARE
102 MALLOC_DECLARE(M_TSEGQ);
103 #endif
104 
105 struct tcptemp {
106 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
107 	struct	tcphdr tt_t;
108 };
109 
110 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
111 
112 struct raw_sackblock {				/* covers [start, end) */
113 	tcp_seq rblk_start;
114 	tcp_seq rblk_end;
115 };
116 
117 /* maximum number of SACK blocks that will fit in the TCP option space */
118 #define	MAX_SACK_REPORT_BLOCKS	4
119 
120 TAILQ_HEAD(sackblock_list, sackblock);
121 
122 struct scoreboard {
123 	int nblocks;
124 	struct sackblock_list sackblocks;
125 	tcp_seq lostseq;			/* passed SACK lost test */
126 	struct sackblock *lastfound;		/* search hint */
127 };
128 
129 struct netmsg_tcp_timer;
130 
131 /*
132  * Tcp control block, one per tcp; fields:
133  * Organized for 16 byte cacheline efficiency.
134  */
135 struct tcpcb {
136 	struct	tsegqe_head t_segq;
137 	int	t_dupacks;		/* consecutive dup acks recd */
138 	int	tt_cpu;			/* sanity check the cpu */
139 
140 	struct	tcp_callout *tt_rexmt;	/* retransmit timer */
141 	struct	tcp_callout *tt_persist;/* retransmit persistence */
142 	struct	tcp_callout *tt_keep;	/* keepalive */
143 	struct	tcp_callout *tt_2msl;	/* 2*msl TIME_WAIT timer */
144 	struct	tcp_callout *tt_delack;	/* delayed ACK timer */
145 	struct	netmsg_tcp_timer *tt_msg; /* timer message */
146 
147 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
148 	int	t_state;		/* state of this connection */
149 	u_int	t_flags;
150 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
151 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
152 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
153 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
154 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
155 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
156 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
157 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
158 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
159 #define	TF_SACK_PERMITTED 0x00000200	/* other side said I could SACK */
160 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
161 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
162 #define	TF_NOPUSH	0x00001000	/* don't push */
163 #define	TF_REQ_CC	0x00002000	/* have/will request CC */
164 #define	TF_RCVD_CC	0x00004000	/* a CC was received in SYN */
165 #define	TF_SENDCCNEW	0x00008000	/* send CCnew instead of CC in SYN */
166 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
167 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
168 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
169 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
170 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
171 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
172 #define	TF_FIRSTACCACK	0x00400000	/* Look for 1st acceptable ACK. */
173 #define	TF_FASTREXMT	0x00800000	/* Did Fast Retransmit. */
174 #define	TF_EARLYREXMT	0x01000000	/* Did Early (Fast) Retransmit. */
175 #define	TF_FORCE	0x02000000	/* Set if forcing out a byte */
176 #define TF_ONOUTPUTQ	0x04000000	/* on t_outputq list */
177 #define TF_DUPSEG	0x08000000	/* last seg a duplicate */
178 #define TF_ENCLOSESEG	0x10000000	/* enclosing SACK block */
179 #define TF_SACKLEFT	0x20000000	/* send SACK blocks from left side */
180 #define TF_KEEPALIVE	0x40000000	/* temporary keepalive */
181 #define TF_RXRESIZED	0x80000000	/* rcvbuf was resized */
182 	tcp_seq	snd_up;			/* send urgent pointer */
183 
184 	tcp_seq	snd_una;		/* send unacknowledged */
185 	tcp_seq	snd_recover;		/* for use with NewReno Fast Recovery */
186 	tcp_seq	snd_max;		/* highest sequence number sent;
187 					 * used to recognize retransmits */
188 	tcp_seq	snd_nxt;		/* send next */
189 
190 	tcp_seq	snd_wl1;		/* window update seg seq number */
191 	tcp_seq	snd_wl2;		/* window update seg ack number */
192 	tcp_seq	iss;			/* initial send sequence number */
193 	tcp_seq	irs;			/* initial receive sequence number */
194 
195 	tcp_seq	rcv_nxt;		/* receive next */
196 	tcp_seq	rcv_adv;		/* advertised window */
197 	u_long	rcv_wnd;		/* receive window */
198 	tcp_seq	rcv_up;			/* receive urgent pointer */
199 
200 	u_long	snd_wnd;		/* send window */
201 	u_long	snd_cwnd;		/* congestion-controlled window */
202 	u_long	snd_wacked;		/* bytes acked in one send window */
203 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
204 					 * for slow start exponential to
205 					 * linear switch */
206 
207 	int	t_rxtcur;		/* current retransmit value (ticks) */
208 	u_int	t_maxseg;		/* maximum segment size */
209 	int	t_srtt;			/* smoothed round-trip time */
210 	int	t_rttvar;		/* variance in round-trip time */
211 
212 	u_int	t_maxopd;		/* mss plus options */
213 
214 	u_long	t_rcvtime;		/* inactivity time */
215 	u_long	t_starttime;		/* time connection was established */
216 	int	t_rtttime;		/* round trip time */
217 	tcp_seq	t_rtseq;		/* sequence number being timed */
218 
219 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
220 	u_int	t_rttmin;		/* minimum rtt allowed */
221 	u_int	t_rttbest;		/* best rtt we've seen */
222 	u_long	t_rttupdated;		/* number of times rtt sampled */
223 	u_long	max_sndwnd;		/* largest window peer has offered */
224 
225 	int	t_softerror;		/* possible error not yet reported */
226 /* out-of-band data */
227 	char	t_oobflags;		/* have some */
228 	char	t_iobc;			/* input character */
229 #define	TCPOOB_HAVEDATA	0x01
230 #define	TCPOOB_HADDATA	0x02
231 
232 /* RFC 1323 variables */
233 	u_char	snd_scale;		/* window scaling for send window */
234 	u_char	rcv_scale;		/* window scaling for recv window */
235 	u_char	request_r_scale;	/* pending window scaling */
236 	u_char	requested_s_scale;
237 	u_long	ts_recent;		/* timestamp echo data */
238 
239 	u_long	ts_recent_age;		/* when last updated */
240 	tcp_seq	last_ack_sent;
241 
242 /* RFC 1644 variables */
243 	tcp_cc	cc_send;		/* send connection count */
244 	tcp_cc	cc_recv;		/* receive connection count */
245 
246 /* experimental */
247 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
248 	u_long	snd_wacked_prev;	/* prior bytes acked in send window */
249 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
250 	tcp_seq snd_recover_prev;	/* snd_recover prior to retransmit */
251 	u_long	t_badrxtwin;		/* window for retransmit recovery */
252 	u_long	t_rexmtTS;		/* timestamp of last retransmit */
253 	u_char	snd_limited;		/* segments limited transmitted */
254 
255 	tcp_seq	rexmt_high;		/* highest seq # retransmitted + 1 */
256 	tcp_seq	snd_max_rexmt;		/* snd_max when rexmting snd_una */
257 	struct scoreboard scb;		/* sack scoreboard */
258 	struct raw_sackblock reportblk; /* incoming segment or D-SACK block */
259 	struct raw_sackblock encloseblk;
260 	int	nsackhistory;
261 	struct raw_sackblock sackhistory[MAX_SACK_REPORT_BLOCKS]; /* reported */
262 	TAILQ_ENTRY(tcpcb) t_outputq;	/* tcp_output needed list */
263 
264 	/* bandwith limitation */
265 	u_long	snd_bandwidth;		/* calculated bandwidth or 0 */
266 	u_long	snd_bwnd;		/* bandwidth-controlled window */
267 	int	t_bw_rtttime;		/* used for bandwidth calculation */
268 	tcp_seq	t_bw_rtseq;		/* used for bandwidth calculation */
269 
270 /* anti DoS counters */
271 	u_long	rcv_second;		/* start of interval second */
272 	u_long	rcv_pps;		/* received packets per second */
273 	u_long	rcv_byps;		/* received bytes per second */
274 
275 	u_int32_t	rfbuf_ts;	/* recv buffer autoscaling timestamp */
276 	int	rfbuf_cnt;		/* recv buffer autoscaling byte count */
277 };
278 
279 #define	IN_FASTRECOVERY(tp)	(tp->t_flags & TF_FASTRECOVERY)
280 #define	ENTER_FASTRECOVERY(tp)	tp->t_flags |= TF_FASTRECOVERY
281 #define	EXIT_FASTRECOVERY(tp)	tp->t_flags &= ~TF_FASTRECOVERY
282 
283 /*
284  * TCP statistics.
285  */
286 struct tcp_stats {
287 	u_long	tcps_connattempt;	/* connections initiated */
288 	u_long	tcps_accepts;		/* connections accepted */
289 	u_long	tcps_connects;		/* connections established */
290 	u_long	tcps_drops;		/* connections dropped */
291 	u_long	tcps_conndrops;		/* embryonic connections dropped */
292 	u_long	tcps_minmssdrops;	/* average minmss too low drops */
293 	u_long	tcps_closed;		/* conn. closed (includes drops) */
294 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
295 	u_long	tcps_rttupdated;	/* times we succeeded */
296 	u_long	tcps_delack;		/* delayed acks sent */
297 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
298 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
299 	u_long	tcps_persisttimeo;	/* persist timeouts */
300 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
301 	u_long	tcps_keepprobe;		/* keepalive probes sent */
302 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
303 
304 	u_long	tcps_sndtotal;		/* total packets sent */
305 	u_long	tcps_sndpack;		/* data packets sent */
306 	u_long	tcps_sndbyte;		/* data bytes sent */
307 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
308 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
309 	u_long	tcps_sndfastrexmit;	/* Fast Retransmissions */
310 	u_long	tcps_sndearlyrexmit;	/* early Fast Retransmissions */
311 	u_long	tcps_sndlimited;	/* Limited Transmit packets */
312 	u_long	tcps_sndrtobad;		/* spurious RTO retransmissions */
313 	u_long	tcps_sndfastrexmitbad;	/* spurious Fast Retransmissions */
314 	u_long	tcps_sndearlyrexmitbad;	/* spurious early Fast Retransmissions,
315 					   a subset of tcps_sndfastrexmitbad */
316 	u_long	tcps_eifeldetected;	/* Eifel-detected spurious rexmits */
317 	u_long	tcps_rttcantdetect;	/* Eifel but not 1/2 RTT-detectable */
318 	u_long	tcps_rttdetected;	/* RTT-detected spurious RTO rexmits */
319 	u_long	tcps_sndacks;		/* ack-only packets sent */
320 	u_long	tcps_sndprobe;		/* window probes sent */
321 	u_long	tcps_sndurg;		/* packets sent with URG only */
322 	u_long	tcps_sndwinup;		/* window update-only packets sent */
323 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
324 	u_long	tcps_sndsackpack;	/* packets sent by SACK recovery alg */
325 	u_long	tcps_sndsackbyte;	/* bytes sent by SACK recovery */
326 	u_long	tcps_snduna3;		/* re-retransmit snd_una on 3 new seg */
327 	u_long	tcps_snduna1;		/* re-retransmit snd_una on 1 new seg */
328 
329 	u_long	tcps_rcvtotal;		/* total packets received */
330 	u_long	tcps_rcvpack;		/* packets received in sequence */
331 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
332 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
333 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
334 	u_long	tcps_rcvmemdrop;	/* packets dropped for lack of memory */
335 	u_long	tcps_rcvshort;		/* packets received too short */
336 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
337 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
338 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
339 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
340 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
341 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
342 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
343 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
344 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
345 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
346 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
347 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
348 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
349 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
350 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
351 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
352 	u_long	tcps_predack;		/* times hdr predict ok for acks */
353 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
354 	u_long	tcps_pcbcachemiss;
355 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
356 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
357 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
358 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
359 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
360 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
361 	u_long	tcps_persistdrop;	/* timeout in persist state */
362 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
363 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
364 	u_long	tcps_listendrop;	/* listen queue overflows */
365 
366 	u_long	tcps_sc_added;		/* entry added to syncache */
367 	u_long	tcps_sc_retransmitted;	/* syncache entry was retransmitted */
368 	u_long	tcps_sc_dupsyn;		/* duplicate SYN packet */
369 	u_long	tcps_sc_dropped;	/* could not reply to packet */
370 	u_long	tcps_sc_completed;	/* successful extraction of entry */
371 	u_long	tcps_sc_bucketoverflow;	/* syncache per-bucket limit hit */
372 	u_long	tcps_sc_cacheoverflow;	/* syncache cache limit hit */
373 	u_long	tcps_sc_reset;		/* RST removed entry from syncache */
374 	u_long	tcps_sc_stale;		/* timed out or listen socket gone */
375 	u_long	tcps_sc_aborted;	/* syncache entry aborted */
376 	u_long	tcps_sc_badack;		/* removed due to bad ACK */
377 	u_long	tcps_sc_unreach;	/* ICMP unreachable received */
378 	u_long	tcps_sc_zonefail;	/* zalloc() failed */
379 	u_long	tcps_sc_sendcookie;	/* SYN cookie sent */
380 	u_long	tcps_sc_recvcookie;	/* SYN cookie received */
381 };
382 
383 #ifdef _KERNEL
384 
385 #if defined(SMP)
386 #define tcpstat	tcpstats_percpu[mycpuid]
387 #else
388 #define tcpstat	tcpstats_percpu[0]
389 #endif
390 
391 struct sockopt;
392 
393 extern struct tcp_stats		tcpstats_percpu[MAXCPU];
394 
395 static const int tcprexmtthresh = 3;
396 #endif
397 
398 /*
399  * Structure to hold TCP options that are only used during segment
400  * processing (in tcp_input), but not held in the tcpcb.
401  * It's basically used to reduce the number of parameters
402  * to tcp_dooptions.
403  */
404 struct tcpopt {
405 	u_long		to_flags;	/* which options are present */
406 #define	TOF_TS			0x0001		/* timestamp */
407 #define	TOF_CC			0x0002		/* CC and CCnew are exclusive */
408 #define	TOF_CCNEW		0x0004
409 #define	TOF_CCECHO		0x0008
410 #define	TOF_MSS			0x0010
411 #define	TOF_SCALE		0x0020
412 #define	TOF_SACK_PERMITTED	0x0040
413 #define	TOF_SACK		0x0080
414 	u_int32_t	to_tsval;
415 	u_int32_t	to_tsecr;
416 	tcp_cc		to_cc;		/* holds CC or CCnew */
417 	tcp_cc		to_ccecho;
418 	u_int16_t	to_mss;
419 	u_int8_t	to_requested_s_scale;
420 	u_int8_t	to_nsackblocks;
421 	struct raw_sackblock *to_sackblocks;
422 };
423 
424 struct syncache {
425 	inp_gen_t	sc_inp_gencnt;		/* pointer check */
426 	struct		tcpcb *sc_tp;		/* tcb for listening socket */
427 	struct		mbuf *sc_ipopts;	/* source route */
428 	struct		in_conninfo sc_inc;	/* addresses */
429 #define sc_route	sc_inc.inc_route
430 #define sc_route6	sc_inc.inc6_route
431 	u_int32_t	sc_tsrecent;
432 	tcp_cc		sc_cc_send;		/* holds CC or CCnew */
433 	tcp_cc		sc_cc_recv;
434 	tcp_seq		sc_irs;			/* seq from peer */
435 	tcp_seq		sc_iss;			/* our ISS */
436 	u_long		sc_rxttime;		/* retransmit time */
437 	u_int16_t	sc_rxtslot;		/* retransmit counter */
438 	u_int16_t	sc_peer_mss;		/* peer's MSS */
439 	u_int16_t	sc_wnd;			/* advertised window */
440 	u_int8_t	sc_requested_s_scale:4,
441 			sc_request_r_scale:4;
442 	u_int8_t	sc_flags;
443 #define SCF_NOOPT		0x01		/* no TCP options */
444 #define SCF_WINSCALE		0x02		/* negotiated window scaling */
445 #define SCF_TIMESTAMP		0x04		/* negotiated timestamps */
446 #define SCF_CC			0x08		/* negotiated CC */
447 #define SCF_UNREACH		0x10		/* icmp unreachable received */
448 #define	SCF_SACK_PERMITTED	0x20		/* saw SACK permitted option */
449 	TAILQ_ENTRY(syncache) sc_hash;
450 	TAILQ_ENTRY(syncache) sc_timerq;
451 };
452 
453 struct syncache_head {
454 	TAILQ_HEAD(, syncache)	sch_bucket;
455 	u_int		sch_length;
456 };
457 
458 /*
459  * The TAO cache entry which is stored in the protocol family specific
460  * portion of the route metrics.
461  */
462 struct rmxp_tao {
463 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
464 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
465 	u_short	tao_mssopt;		/* peer's cached MSS */
466 
467 #ifdef notyet
468 	u_short	tao_flags;		/* cache status flags */
469 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
470 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
471 #define	TAOF_UNDEF	0		/* we don't know yet */
472 #endif
473 };
474 
475 #define rmx_taop(rt)	((struct rmxp_tao *)(rt).rmx_filler)
476 
477 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
478 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
479 
480 /*
481  * The smoothed round-trip time and estimated variance
482  * are stored as fixed point numbers scaled by the values below.
483  * For convenience, these scales are also used in smoothing the average
484  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
485  * With these scales, srtt has 3 bits to the right of the binary point,
486  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
487  * binary point, and is smoothed with an ALPHA of 0.75.
488  */
489 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
490 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
491 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
492 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
493 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
494 
495 /*
496  * The initial retransmission should happen at rtt + 4 * rttvar.
497  * Because of the way we do the smoothing, srtt and rttvar
498  * will each average +1/2 tick of bias.  When we compute
499  * the retransmit timer, we want 1/2 tick of rounding and
500  * 1 extra tick because of +-1/2 tick uncertainty in the
501  * firing of the timer.  The bias will give us exactly the
502  * 1.5 tick we need.  But, because the bias is
503  * statistical, we have to test that we don't drop below
504  * the minimum feasible timer (which is 2 ticks).
505  * This version of the macro adapted from a paper by Lawrence
506  * Brakmo and Larry Peterson which outlines a problem caused
507  * by insufficient precision in the original implementation,
508  * which results in inappropriately large RTO values for very
509  * fast networks.
510  */
511 #define	TCP_REXMTVAL(tp) \
512 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
513 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
514 
515 /*
516  * TCB structure exported to user-land via sysctl(3).
517  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
518  * included.  Not all of our clients do.
519  */
520 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
521 struct	xtcpcb {
522 	size_t	xt_len;
523 	struct	inpcb	xt_inp;
524 	struct	tcpcb	xt_tp;
525 	struct	xsocket	xt_socket;
526 	u_quad_t	xt_alignment_hack;
527 };
528 #endif
529 
530 /*
531  * Names for TCP sysctl objects
532  */
533 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
534 #define	TCPCTL_DO_RFC1644	2	/* use RFC-1644 extensions */
535 #define	TCPCTL_MSSDFLT		3	/* MSS default */
536 #define TCPCTL_STATS		4	/* statistics (read-only) */
537 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
538 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
539 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
540 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
541 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
542 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
543 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
544 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
545 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
546 #define	TCPCTL_MAXID		14
547 
548 #define TCPCTL_NAMES { \
549 	{ 0, 0 }, \
550 	{ "rfc1323", CTLTYPE_INT }, \
551 	{ "rfc1644", CTLTYPE_INT }, \
552 	{ "mssdflt", CTLTYPE_INT }, \
553 	{ "stats", CTLTYPE_STRUCT }, \
554 	{ "rttdflt", CTLTYPE_INT }, \
555 	{ "keepidle", CTLTYPE_INT }, \
556 	{ "keepintvl", CTLTYPE_INT }, \
557 	{ "sendspace", CTLTYPE_INT }, \
558 	{ "recvspace", CTLTYPE_INT }, \
559 	{ "keepinit", CTLTYPE_INT }, \
560 	{ "pcblist", CTLTYPE_STRUCT }, \
561 	{ "delacktime", CTLTYPE_INT }, \
562 	{ "v6mssdflt", CTLTYPE_INT }, \
563 }
564 
565 #ifdef _KERNEL
566 #ifdef SYSCTL_DECL
567 SYSCTL_DECL(_net_inet_tcp);
568 #endif
569 
570 #define TCP_DO_SACK(tp)		((tp)->t_flags & TF_SACK_PERMITTED)
571 
572 TAILQ_HEAD(tcpcbackqhead,tcpcb);
573 
574 extern	struct inpcbinfo tcbinfo[];
575 extern	struct tcpcbackqhead tcpcbackq[];
576 
577 extern	int tcp_mssdflt;	/* XXX */
578 extern	int tcp_minmss;
579 extern	int tcp_delack_enabled;
580 extern	int path_mtu_discovery;
581 
582 int	 tcp_addrcpu(in_addr_t faddr, in_port_t fport,
583 	    in_addr_t laddr, in_port_t lport);
584 struct lwkt_port *
585 	tcp_addrport(in_addr_t faddr, in_port_t fport,
586 	    in_addr_t laddr, in_port_t lport);
587 struct lwkt_port *tcp_addrport0(void);
588 void	 tcp_canceltimers (struct tcpcb *);
589 struct tcpcb *
590 	 tcp_close (struct tcpcb *);
591 void	 tcpmsg_service_loop (void *);
592 void	 tcp_ctlinput (int, struct sockaddr *, void *);
593 int	 tcp_ctloutput (struct socket *, struct sockopt *);
594 struct lwkt_port *
595 	 tcp_cport(int cpu);
596 struct tcpcb *
597 	 tcp_drop (struct tcpcb *, int);
598 void	 tcp_drain (void);
599 void	 tcp_fasttimo (void);
600 struct rmxp_tao *
601 	 tcp_gettaocache (struct in_conninfo *);
602 void	 tcp_init (void);
603 void	 tcp_thread_init (void);
604 void	 tcp_input (struct mbuf *, ...);
605 void	 tcp_mss (struct tcpcb *, int);
606 int	 tcp_mssopt (struct tcpcb *);
607 void	 tcp_drop_syn_sent (struct inpcb *, int);
608 void	 tcp_mtudisc (struct inpcb *, int);
609 struct tcpcb *
610 	 tcp_newtcpcb (struct inpcb *);
611 int	 tcp_output (struct tcpcb *);
612 void	 tcp_quench (struct inpcb *, int);
613 void	 tcp_respond (struct tcpcb *, void *,
614 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
615 struct rtentry *
616 	 tcp_rtlookup (struct in_conninfo *);
617 int	 tcp_sack_bytes_below(struct scoreboard *scb, tcp_seq seq);
618 void	 tcp_sack_cleanup(struct scoreboard *scb);
619 int	 tcp_sack_ndsack_blocks(struct raw_sackblock *blocks,
620 	    const int numblocks, tcp_seq snd_una);
621 void	 tcp_sack_fill_report(struct tcpcb *tp, u_char *opt, u_int *plen);
622 boolean_t
623 	 tcp_sack_has_sacked(struct scoreboard *scb, u_int amount);
624 void	 tcp_sack_tcpcb_init(struct tcpcb *tp);
625 uint32_t tcp_sack_compute_pipe(struct tcpcb *tp);
626 boolean_t
627 	 tcp_sack_nextseg(struct tcpcb *tp, tcp_seq *nextrexmt, uint32_t *len,
628 			  boolean_t *losdup);
629 #ifdef later
630 void	 tcp_sack_revert_scoreboard(struct scoreboard *scb, tcp_seq snd_una,
631 				    u_int maxseg);
632 void	 tcp_sack_save_scoreboard(struct scoreboard *scb);
633 #endif
634 void	 tcp_sack_skip_sacked(struct scoreboard *scb, tcp_seq *prexmt);
635 void	 tcp_sack_update_scoreboard(struct tcpcb *tp, struct tcpopt *to);
636 void	 tcp_save_congestion_state(struct tcpcb *tp);
637 void	 tcp_revert_congestion_state(struct tcpcb *tp);
638 void	 tcp_setpersist (struct tcpcb *);
639 void	 tcp_slowtimo (void);
640 struct tcptemp *tcp_maketemplate (struct tcpcb *);
641 void	 tcp_freetemplate (struct tcptemp *);
642 void	 tcp_fillheaders (struct tcpcb *, void *, void *);
643 struct lwkt_port *
644 	 tcp_soport(struct socket *, struct sockaddr *,
645 	 	    struct mbuf **, int);
646 struct lwkt_port *
647 	 tcp_ctlport(int, struct sockaddr *, void *);
648 struct tcpcb *
649 	 tcp_timers (struct tcpcb *, int);
650 void	 tcp_trace (short, short, struct tcpcb *, void *, struct tcphdr *,
651 			int);
652 void	 tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
653 void	 syncache_init(void);
654 void	 syncache_unreach(struct in_conninfo *, struct tcphdr *);
655 int	 syncache_expand(struct in_conninfo *, struct tcphdr *,
656 	     struct socket **, struct mbuf *);
657 int	 syncache_add(struct in_conninfo *, struct tcpopt *,
658 	     struct tcphdr *, struct socket **, struct mbuf *);
659 void	 syncache_chkrst(struct in_conninfo *, struct tcphdr *);
660 void	 syncache_badack(struct in_conninfo *);
661 
662 extern	struct pr_usrreqs tcp_usrreqs;
663 extern	u_long tcp_sendspace;
664 extern	u_long tcp_recvspace;
665 tcp_seq tcp_new_isn (struct tcpcb *);
666 
667 #endif /* _KERNEL */
668 
669 #endif /* _NETINET_TCP_VAR_H_ */
670