xref: /dflybsd-src/sys/netinet/tcp_var.h (revision bc76a771df54af7e361532b257cecc26227736b4)
1 /*
2  * Copyright (c) 2003-2004 Jeffrey Hsu.  All rights reserved.
3  * Copyright (c) 1982, 1986, 1993, 1994, 1995
4  *	The Regents of the University of California.  All rights reserved.
5  *
6  * Redistribution and use in source and binary forms, with or without
7  * modification, are permitted provided that the following conditions
8  * are met:
9  * 1. Redistributions of source code must retain the above copyright
10  *    notice, this list of conditions and the following disclaimer.
11  * 2. Redistributions in binary form must reproduce the above copyright
12  *    notice, this list of conditions and the following disclaimer in the
13  *    documentation and/or other materials provided with the distribution.
14  * 3. All advertising materials mentioning features or use of this software
15  *    must display the following acknowledgement:
16  *	This product includes software developed by the University of
17  *	California, Berkeley and its contributors.
18  * 4. Neither the name of the University nor the names of its contributors
19  *    may be used to endorse or promote products derived from this software
20  *    without specific prior written permission.
21  *
22  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
23  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
24  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
25  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
26  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
27  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
28  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
29  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
30  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
31  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32  * SUCH DAMAGE.
33  *
34  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
35  * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.13 2003/02/03 02:34:07 hsu Exp $
36  * $DragonFly: src/sys/netinet/tcp_var.h,v 1.17 2004/04/10 00:10:42 hsu Exp $
37  */
38 
39 #ifndef _NETINET_TCP_VAR_H_
40 #define _NETINET_TCP_VAR_H_
41 
42 #include <netinet/in_pcb.h>		/* needed for in_conninfo, inp_gen_t */
43 
44 #ifndef _NETINET_TCP_STATS_H_
45 #include <netinet/tcp_stats.h>
46 #endif
47 
48 /*
49  * Kernel variables for tcp.
50  */
51 extern int	tcp_do_rfc1323;
52 extern int	tcp_do_rfc1644;
53 
54 /* TCP segment queue entry */
55 struct tseg_qent {
56 	LIST_ENTRY(tseg_qent) tqe_q;
57 	int	tqe_len;		/* TCP segment data length */
58 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
59 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
60 };
61 LIST_HEAD(tsegqe_head, tseg_qent);
62 extern int	tcp_reass_maxseg;
63 extern int	tcp_reass_qsize;
64 #ifdef MALLOC_DECLARE
65 MALLOC_DECLARE(M_TSEGQ);
66 #endif
67 
68 struct tcptemp {
69 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
70 	struct	tcphdr tt_t;
71 };
72 
73 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
74 
75 /*
76  * Tcp control block, one per tcp; fields:
77  * Organized for 16 byte cacheline efficiency.
78  */
79 struct tcpcb {
80 	struct	tsegqe_head t_segq;
81 	int	t_dupacks;		/* consecutive dup acks recd */
82 	struct	tcptemp	*unused;	/* unused */
83 
84 	struct	callout *tt_rexmt;	/* retransmit timer */
85 	struct	callout *tt_persist;	/* retransmit persistence */
86 	struct	callout *tt_keep;	/* keepalive */
87 	struct	callout *tt_2msl;	/* 2*msl TIME_WAIT timer */
88 	struct	callout *tt_delack;	/* delayed ACK timer */
89 
90 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
91 	int	t_state;		/* state of this connection */
92 	u_int	t_flags;
93 #define	TF_ACKNOW	0x00000001	/* ack peer immediately */
94 #define	TF_DELACK	0x00000002	/* ack, but try to delay it */
95 #define	TF_NODELAY	0x00000004	/* don't delay packets to coalesce */
96 #define	TF_NOOPT	0x00000008	/* don't use tcp options */
97 #define	TF_SENTFIN	0x00000010	/* have sent FIN */
98 #define	TF_REQ_SCALE	0x00000020	/* have/will request window scaling */
99 #define	TF_RCVD_SCALE	0x00000040	/* other side has requested scaling */
100 #define	TF_REQ_TSTMP	0x00000080	/* have/will request timestamps */
101 #define	TF_RCVD_TSTMP	0x00000100	/* a timestamp was received in SYN */
102 #define	TF_SACK_PERMIT	0x00000200	/* other side said I could SACK */
103 #define	TF_NEEDSYN	0x00000400	/* send SYN (implicit state) */
104 #define	TF_NEEDFIN	0x00000800	/* send FIN (implicit state) */
105 #define	TF_NOPUSH	0x00001000	/* don't push */
106 #define	TF_REQ_CC	0x00002000	/* have/will request CC */
107 #define	TF_RCVD_CC	0x00004000	/* a CC was received in SYN */
108 #define	TF_SENDCCNEW	0x00008000	/* send CCnew instead of CC in SYN */
109 #define	TF_MORETOCOME	0x00010000	/* More data to be appended to sock */
110 #define	TF_LQ_OVERFLOW	0x00020000	/* listen queue overflow */
111 #define	TF_LASTIDLE	0x00040000	/* connection was previously idle */
112 #define	TF_RXWIN0SENT	0x00080000	/* sent a receiver win 0 in response */
113 #define	TF_FASTRECOVERY	0x00100000	/* in NewReno Fast Recovery */
114 #define	TF_WASFRECOVERY	0x00200000	/* was in NewReno Fast Recovery */
115 #define	TF_FIRSTACCACK	0x00400000	/* Look for 1st acceptable ACK. */
116 #define	TF_FASTREXMT	0x00800000	/* Did Fast Retransmit. */
117 #define	TF_EARLYREXMT	0x01000000	/* Did Early (Fast) Retransmit. */
118 	int	t_force;		/* 1 if forcing out a byte */
119 
120 	tcp_seq	snd_una;		/* send unacknowledged */
121 	tcp_seq	snd_max;		/* highest sequence number sent;
122 					 * used to recognize retransmits
123 					 */
124 	tcp_seq	snd_nxt;		/* send next */
125 	tcp_seq	snd_up;			/* send urgent pointer */
126 
127 	tcp_seq	snd_wl1;		/* window update seg seq number */
128 	tcp_seq	snd_wl2;		/* window update seg ack number */
129 	tcp_seq	iss;			/* initial send sequence number */
130 	tcp_seq	irs;			/* initial receive sequence number */
131 
132 	tcp_seq	rcv_nxt;		/* receive next */
133 	tcp_seq	rcv_adv;		/* advertised window */
134 	u_long	rcv_wnd;		/* receive window */
135 	tcp_seq	rcv_up;			/* receive urgent pointer */
136 
137 	u_long	snd_wnd;		/* send window */
138 	u_long	snd_cwnd;		/* congestion-controlled window */
139 	u_long	snd_bwnd;		/* bandwidth-controlled window */
140 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
141 					 * for slow start exponential to
142 					 * linear switch
143 					 */
144 	u_long	snd_bandwidth;		/* calculated bandwidth or 0 */
145 	tcp_seq	snd_recover;		/* for use in NewReno fast recovery */
146 
147 	u_int	t_maxopd;		/* mss plus options */
148 
149 	u_long	t_rcvtime;		/* inactivity time */
150 	u_long	t_starttime;		/* time connection was established */
151 	int	t_rtttime;		/* round trip time */
152 	tcp_seq	t_rtseq;		/* sequence number being timed */
153 
154 	int	t_bw_rtttime;		/* used for bandwidth calculation */
155 	tcp_seq	t_bw_rtseq;		/* used for bandwidth calculation */
156 
157 	int	t_rxtcur;		/* current retransmit value (ticks) */
158 	u_int	t_maxseg;		/* maximum segment size */
159 	int	t_srtt;			/* smoothed round-trip time */
160 	int	t_rttvar;		/* variance in round-trip time */
161 
162 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
163 	u_int	t_rttmin;		/* minimum rtt allowed */
164 	u_int	t_rttbest;		/* best rtt we've seen */
165 	u_long	t_rttupdated;		/* number of times rtt sampled */
166 	u_long	max_sndwnd;		/* largest window peer has offered */
167 
168 	int	t_softerror;		/* possible error not yet reported */
169 /* out-of-band data */
170 	char	t_oobflags;		/* have some */
171 	char	t_iobc;			/* input character */
172 #define	TCPOOB_HAVEDATA	0x01
173 #define	TCPOOB_HADDATA	0x02
174 /* RFC 1323 variables */
175 	u_char	snd_scale;		/* window scaling for send window */
176 	u_char	rcv_scale;		/* window scaling for recv window */
177 	u_char	request_r_scale;	/* pending window scaling */
178 	u_char	requested_s_scale;
179 	u_long	ts_recent;		/* timestamp echo data */
180 
181 	u_long	ts_recent_age;		/* when last updated */
182 	tcp_seq	last_ack_sent;
183 /* RFC 1644 variables */
184 	tcp_cc	cc_send;		/* send connection count */
185 	tcp_cc	cc_recv;		/* receive connection count */
186 /* experimental */
187 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
188 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
189 	tcp_seq snd_recover_prev;	/* snd_recover prior to retransmit */
190 	u_long	t_badrxtwin;		/* window for retransmit recovery */
191 	u_long	t_rexmtTS;		/* timestamp of last retransmit */
192 	u_char	snd_limited;		/* segments limited transmitted */
193 };
194 
195 #define	IN_FASTRECOVERY(tp)	(tp->t_flags & TF_FASTRECOVERY)
196 #define	ENTER_FASTRECOVERY(tp)	tp->t_flags |= TF_FASTRECOVERY
197 #define	EXIT_FASTRECOVERY(tp)	tp->t_flags &= ~TF_FASTRECOVERY
198 
199 /*
200  * Structure to hold TCP options that are only used during segment
201  * processing (in tcp_input), but not held in the tcpcb.
202  * It's basically used to reduce the number of parameters
203  * to tcp_dooptions.
204  */
205 struct tcpopt {
206 	u_long		to_flags;	/* which options are present */
207 #define TOF_TS		0x0001		/* timestamp */
208 #define TOF_CC		0x0002		/* CC and CCnew are exclusive */
209 #define TOF_CCNEW	0x0004
210 #define	TOF_CCECHO	0x0008
211 #define	TOF_MSS		0x0010
212 #define	TOF_SCALE	0x0020
213 	u_int32_t	to_tsval;
214 	u_int32_t	to_tsecr;
215 	tcp_cc		to_cc;		/* holds CC or CCnew */
216 	tcp_cc		to_ccecho;
217 	u_int16_t	to_mss;
218 	u_int8_t 	to_requested_s_scale;
219 	u_int8_t 	to_pad;
220 };
221 
222 struct syncache {
223 	inp_gen_t	sc_inp_gencnt;		/* pointer check */
224 	struct 		tcpcb *sc_tp;		/* tcb for listening socket */
225 	struct		mbuf *sc_ipopts;	/* source route */
226 	struct 		in_conninfo sc_inc;	/* addresses */
227 #define sc_route	sc_inc.inc_route
228 #define sc_route6	sc_inc.inc6_route
229 	u_int32_t	sc_tsrecent;
230 	tcp_cc		sc_cc_send;		/* holds CC or CCnew */
231 	tcp_cc		sc_cc_recv;
232 	tcp_seq 	sc_irs;			/* seq from peer */
233 	tcp_seq 	sc_iss;			/* our ISS */
234 	u_long		sc_rxttime;		/* retransmit time */
235 	u_int16_t	sc_rxtslot; 		/* retransmit counter */
236 	u_int16_t	sc_peer_mss;		/* peer's MSS */
237 	u_int16_t	sc_wnd;			/* advertised window */
238 	u_int8_t 	sc_requested_s_scale:4,
239 			sc_request_r_scale:4;
240 	u_int8_t	sc_flags;
241 #define SCF_NOOPT	0x01			/* no TCP options */
242 #define SCF_WINSCALE	0x02			/* negotiated window scaling */
243 #define SCF_TIMESTAMP	0x04			/* negotiated timestamps */
244 #define SCF_CC		0x08			/* negotiated CC */
245 #define SCF_UNREACH	0x10			/* icmp unreachable received */
246 #define SCF_KEEPROUTE	0x20			/* keep cloned route */
247 	TAILQ_ENTRY(syncache)	sc_hash;
248 	TAILQ_ENTRY(syncache)	sc_timerq;
249 };
250 
251 struct syncache_head {
252 	TAILQ_HEAD(, syncache)	sch_bucket;
253 	u_int		sch_length;
254 };
255 
256 /*
257  * The TAO cache entry which is stored in the protocol family specific
258  * portion of the route metrics.
259  */
260 struct rmxp_tao {
261 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
262 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
263 	u_short	tao_mssopt;		/* peer's cached MSS */
264 #ifdef notyet
265 	u_short	tao_flags;		/* cache status flags */
266 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
267 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
268 #define	TAOF_UNDEF	0		/* we don't know yet */
269 #endif /* notyet */
270 };
271 #define rmx_taop(r)	((struct rmxp_tao *)(r).rmx_filler)
272 
273 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
274 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
275 
276 /*
277  * The smoothed round-trip time and estimated variance
278  * are stored as fixed point numbers scaled by the values below.
279  * For convenience, these scales are also used in smoothing the average
280  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
281  * With these scales, srtt has 3 bits to the right of the binary point,
282  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
283  * binary point, and is smoothed with an ALPHA of 0.75.
284  */
285 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
286 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
287 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
288 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
289 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
290 
291 /*
292  * The initial retransmission should happen at rtt + 4 * rttvar.
293  * Because of the way we do the smoothing, srtt and rttvar
294  * will each average +1/2 tick of bias.  When we compute
295  * the retransmit timer, we want 1/2 tick of rounding and
296  * 1 extra tick because of +-1/2 tick uncertainty in the
297  * firing of the timer.  The bias will give us exactly the
298  * 1.5 tick we need.  But, because the bias is
299  * statistical, we have to test that we don't drop below
300  * the minimum feasible timer (which is 2 ticks).
301  * This version of the macro adapted from a paper by Lawrence
302  * Brakmo and Larry Peterson which outlines a problem caused
303  * by insufficient precision in the original implementation,
304  * which results in inappropriately large RTO values for very
305  * fast networks.
306  */
307 #define	TCP_REXMTVAL(tp) \
308 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
309 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
310 
311 /*
312  * TCB structure exported to user-land via sysctl(3).
313  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
314  * included.  Not all of our clients do.
315  */
316 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
317 struct	xtcpcb {
318 	size_t	xt_len;
319 	struct	inpcb	xt_inp;
320 	struct	tcpcb	xt_tp;
321 	struct	xsocket	xt_socket;
322 	u_quad_t	xt_alignment_hack;
323 };
324 #endif
325 
326 /*
327  * Names for TCP sysctl objects
328  */
329 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
330 #define	TCPCTL_DO_RFC1644	2	/* use RFC-1644 extensions */
331 #define	TCPCTL_MSSDFLT		3	/* MSS default */
332 #define TCPCTL_STATS		4	/* statistics (read-only) */
333 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
334 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
335 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
336 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
337 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
338 #define	TCPCTL_KEEPINIT		10	/* timeout for establishing syn */
339 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
340 #define	TCPCTL_DELACKTIME	12	/* time before sending delayed ACK */
341 #define	TCPCTL_V6MSSDFLT	13	/* MSS default for IPv6 */
342 #define	TCPCTL_MAXID		14
343 
344 #define TCPCTL_NAMES { \
345 	{ 0, 0 }, \
346 	{ "rfc1323", CTLTYPE_INT }, \
347 	{ "rfc1644", CTLTYPE_INT }, \
348 	{ "mssdflt", CTLTYPE_INT }, \
349 	{ "stats", CTLTYPE_STRUCT }, \
350 	{ "rttdflt", CTLTYPE_INT }, \
351 	{ "keepidle", CTLTYPE_INT }, \
352 	{ "keepintvl", CTLTYPE_INT }, \
353 	{ "sendspace", CTLTYPE_INT }, \
354 	{ "recvspace", CTLTYPE_INT }, \
355 	{ "keepinit", CTLTYPE_INT }, \
356 	{ "pcblist", CTLTYPE_STRUCT }, \
357 	{ "delacktime", CTLTYPE_INT }, \
358 	{ "v6mssdflt", CTLTYPE_INT }, \
359 }
360 
361 #ifdef _KERNEL
362 #ifdef SYSCTL_DECL
363 SYSCTL_DECL(_net_inet_tcp);
364 #endif
365 
366 extern	struct inpcbinfo tcbinfo[];
367 extern	int tcp_mssdflt;	/* XXX */
368 extern	int tcp_delack_enabled;
369 extern	int tcp_do_newreno;
370 extern	int path_mtu_discovery;
371 extern	int ss_fltsz;
372 extern	int ss_fltsz_local;
373 
374 int	 tcp_addrcpu(in_addr_t faddr, in_port_t fport,
375 	    in_addr_t laddr, in_port_t lport);
376 struct lwkt_port *
377 	tcp_addrport(in_addr_t faddr, in_port_t fport,
378 	    in_addr_t laddr, in_port_t lport);
379 void	 tcp_canceltimers (struct tcpcb *);
380 struct tcpcb *
381 	 tcp_close (struct tcpcb *);
382 void	 tcp_ctlinput (int, struct sockaddr *, void *);
383 int	 tcp_ctloutput (struct socket *, struct sockopt *);
384 struct lwkt_port *
385 	 tcp_cport(int cpu);
386 struct tcpcb *
387 	 tcp_drop (struct tcpcb *, int);
388 void	 tcp_drain (void);
389 void	 tcp_fasttimo (void);
390 struct rmxp_tao *
391 	 tcp_gettaocache (struct in_conninfo *);
392 void	 tcp_init (void);
393 void	 tcp_thread_init (void);
394 void	 tcp_input (struct mbuf *, int, int);
395 void	 tcp_mss (struct tcpcb *, int);
396 int	 tcp_mssopt (struct tcpcb *);
397 void	 tcp_drop_syn_sent (struct inpcb *, int);
398 void	 tcp_mtudisc (struct inpcb *, int);
399 struct tcpcb *
400 	 tcp_newtcpcb (struct inpcb *);
401 int	 tcp_output (struct tcpcb *);
402 void	 tcp_quench (struct inpcb *, int);
403 void	 tcp_respond (struct tcpcb *, void *,
404 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int);
405 struct rtentry *
406 	 tcp_rtlookup (struct in_conninfo *);
407 void	 tcp_save_congestion_state(struct tcpcb *tp);
408 void	 tcp_revert_congestion_state(struct tcpcb *tp);
409 void	 tcp_setpersist (struct tcpcb *);
410 void	 tcp_slowtimo (void);
411 struct tcptemp *
412 	 tcp_maketemplate (struct tcpcb *);
413 void	 tcp_fillheaders (struct tcpcb *, void *, void *);
414 struct lwkt_port *
415 	 tcp_soport(struct socket *, struct sockaddr *nam);
416 struct tcpcb *
417 	 tcp_timers (struct tcpcb *, int);
418 void	 tcp_trace (int, int, struct tcpcb *, void *, struct tcphdr *,
419 			int);
420 void	 tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
421 void	 syncache_init(void);
422 void	 syncache_unreach(struct in_conninfo *, struct tcphdr *);
423 int	 syncache_expand(struct in_conninfo *, struct tcphdr *,
424 	     struct socket **, struct mbuf *);
425 int	 syncache_add(struct in_conninfo *, struct tcpopt *,
426 	     struct tcphdr *, struct socket **, struct mbuf *);
427 void	 syncache_chkrst(struct in_conninfo *, struct tcphdr *);
428 void	 syncache_badack(struct in_conninfo *);
429 
430 extern	struct pr_usrreqs tcp_usrreqs;
431 extern	u_long tcp_sendspace;
432 extern	u_long tcp_recvspace;
433 tcp_seq tcp_new_isn (struct tcpcb *);
434 
435 #endif /* _KERNEL */
436 
437 #endif /* _NETINET_TCP_VAR_H_ */
438