xref: /freebsd/sys/netinet/tcp_var.h (revision 4a0f765fbf09711e612e86fce8bb09ec43f482d9)
1 /*
2  * Copyright (c) 1982, 1986, 1993, 1994, 1995
3  *	The Regents of the University of California.  All rights reserved.
4  *
5  * Redistribution and use in source and binary forms, with or without
6  * modification, are permitted provided that the following conditions
7  * are met:
8  * 1. Redistributions of source code must retain the above copyright
9  *    notice, this list of conditions and the following disclaimer.
10  * 2. Redistributions in binary form must reproduce the above copyright
11  *    notice, this list of conditions and the following disclaimer in the
12  *    documentation and/or other materials provided with the distribution.
13  * 3. All advertising materials mentioning features or use of this software
14  *    must display the following acknowledgement:
15  *	This product includes software developed by the University of
16  *	California, Berkeley and its contributors.
17  * 4. Neither the name of the University nor the names of its contributors
18  *    may be used to endorse or promote products derived from this software
19  *    without specific prior written permission.
20  *
21  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
22  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
23  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
24  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
25  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
26  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
27  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
28  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
29  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
30  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
31  * SUCH DAMAGE.
32  *
33  *	@(#)tcp_var.h	8.4 (Berkeley) 5/24/95
34  * 	$Id$
35  */
36 
37 #ifndef _NETINET_TCP_VAR_H_
38 #define _NETINET_TCP_VAR_H_
39 /*
40  * Kernel variables for tcp.
41  */
42 
43 /*
44  * Tcp control block, one per tcp; fields:
45  */
46 struct tcpcb {
47 	struct	tcpiphdr *seg_next;	/* sequencing queue */
48 	struct	tcpiphdr *seg_prev;
49 	int	t_state;		/* state of this connection */
50 	int	t_timer[TCPT_NTIMERS];	/* tcp timers */
51 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
52 	int	t_rxtcur;		/* current retransmit value */
53 	int	t_dupacks;		/* consecutive dup acks recd */
54 	u_int	t_maxseg;		/* maximum segment size */
55 	u_int	t_maxopd;		/* mss plus options */
56 	int	t_force;		/* 1 if forcing out a byte */
57 	u_int	t_flags;
58 #define	TF_ACKNOW	0x0001		/* ack peer immediately */
59 #define	TF_DELACK	0x0002		/* ack, but try to delay it */
60 #define	TF_NODELAY	0x0004		/* don't delay packets to coalesce */
61 #define	TF_NOOPT	0x0008		/* don't use tcp options */
62 #define	TF_SENTFIN	0x0010		/* have sent FIN */
63 #define	TF_REQ_SCALE	0x0020		/* have/will request window scaling */
64 #define	TF_RCVD_SCALE	0x0040		/* other side has requested scaling */
65 #define	TF_REQ_TSTMP	0x0080		/* have/will request timestamps */
66 #define	TF_RCVD_TSTMP	0x0100		/* a timestamp was received in SYN */
67 #define	TF_SACK_PERMIT	0x0200		/* other side said I could SACK */
68 #define TF_NEEDSYN	0x0400		/* send SYN (implicit state) */
69 #define TF_NEEDFIN	0x0800		/* send FIN (implicit state) */
70 #define TF_NOPUSH	0x1000		/* don't push */
71 #define TF_REQ_CC	0x2000		/* have/will request CC */
72 #define	TF_RCVD_CC	0x4000		/* a CC was received in SYN */
73 #define TF_SENDCCNEW	0x8000		/* send CCnew instead of CC in SYN */
74 
75 	struct	tcpiphdr *t_template;	/* skeletal packet for transmit */
76 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
77 /*
78  * The following fields are used as in the protocol specification.
79  * See RFC783, Dec. 1981, page 21.
80  */
81 /* send sequence variables */
82 	tcp_seq	snd_una;		/* send unacknowledged */
83 	tcp_seq	snd_nxt;		/* send next */
84 	tcp_seq	snd_up;			/* send urgent pointer */
85 	tcp_seq	snd_wl1;		/* window update seg seq number */
86 	tcp_seq	snd_wl2;		/* window update seg ack number */
87 	tcp_seq	iss;			/* initial send sequence number */
88 	u_long	snd_wnd;		/* send window */
89 /* receive sequence variables */
90 	u_long	rcv_wnd;		/* receive window */
91 	tcp_seq	rcv_nxt;		/* receive next */
92 	tcp_seq	rcv_up;			/* receive urgent pointer */
93 	tcp_seq	irs;			/* initial receive sequence number */
94 /*
95  * Additional variables for this implementation.
96  */
97 /* receive variables */
98 	tcp_seq	rcv_adv;		/* advertised window */
99 /* retransmit variables */
100 	tcp_seq	snd_max;		/* highest sequence number sent;
101 					 * used to recognize retransmits
102 					 */
103 /* congestion control (for slow start, source quench, retransmit after loss) */
104 	u_long	snd_cwnd;		/* congestion-controlled window */
105 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
106 					 * for slow start exponential to
107 					 * linear switch
108 					 */
109 /*
110  * transmit timing stuff.  See below for scale of srtt and rttvar.
111  * "Variance" is actually smoothed difference.
112  */
113 	u_int	t_idle;			/* inactivity time */
114 	int	t_rtt;			/* round trip time */
115 	tcp_seq	t_rtseq;		/* sequence number being timed */
116 	int	t_srtt;			/* smoothed round-trip time */
117 	int	t_rttvar;		/* variance in round-trip time */
118 	u_int	t_rttmin;		/* minimum rtt allowed */
119 	u_long	max_sndwnd;		/* largest window peer has offered */
120 
121 /* out-of-band data */
122 	char	t_oobflags;		/* have some */
123 	char	t_iobc;			/* input character */
124 #define	TCPOOB_HAVEDATA	0x01
125 #define	TCPOOB_HADDATA	0x02
126 	int	t_softerror;		/* possible error not yet reported */
127 
128 /* RFC 1323 variables */
129 	u_char	snd_scale;		/* window scaling for send window */
130 	u_char	rcv_scale;		/* window scaling for recv window */
131 	u_char	request_r_scale;	/* pending window scaling */
132 	u_char	requested_s_scale;
133 	u_long	ts_recent;		/* timestamp echo data */
134 	u_long	ts_recent_age;		/* when last updated */
135 	tcp_seq	last_ack_sent;
136 /* RFC 1644 variables */
137 	tcp_cc	cc_send;		/* send connection count */
138 	tcp_cc	cc_recv;		/* receive connection count */
139 	u_long	t_duration;		/* connection duration */
140 
141 /* TUBA stuff */
142 	caddr_t	t_tuba_pcb;		/* next level down pcb for TCP over z */
143 /* More RTT stuff */
144 	u_long	t_rttupdated;		/* number of times rtt sampled */
145 };
146 
147 /*
148  * Structure to hold TCP options that are only used during segment
149  * processing (in tcp_input), but not held in the tcpcb.
150  * It's basically used to reduce the number of parameters
151  * to tcp_dooptions.
152  */
153 struct tcpopt {
154 	u_long	to_flag;		/* which options are present */
155 #define TOF_TS		0x0001		/* timestamp */
156 #define TOF_CC		0x0002		/* CC and CCnew are exclusive */
157 #define TOF_CCNEW	0x0004
158 #define	TOF_CCECHO	0x0008
159 	u_long	to_tsval;
160 	u_long	to_tsecr;
161 	tcp_cc	to_cc;		/* holds CC or CCnew */
162 	tcp_cc	to_ccecho;
163 };
164 
165 /*
166  * The TAO cache entry which is stored in the protocol family specific
167  * portion of the route metrics.
168  */
169 struct rmxp_tao {
170 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
171 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
172 	u_short	tao_mssopt;		/* peer's cached MSS */
173 #ifdef notyet
174 	u_short	tao_flags;		/* cache status flags */
175 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
176 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
177 #define	TAOF_UNDEF	0		/* we don't know yet */
178 #endif /* notyet */
179 };
180 #define rmx_taop(r)	((struct rmxp_tao *)(r).rmx_filler)
181 
182 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
183 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
184 
185 /*
186  * The smoothed round-trip time and estimated variance
187  * are stored as fixed point numbers scaled by the values below.
188  * For convenience, these scales are also used in smoothing the average
189  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
190  * With these scales, srtt has 3 bits to the right of the binary point,
191  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
192  * binary point, and is smoothed with an ALPHA of 0.75.
193  */
194 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
195 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
196 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
197 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
198 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
199 
200 /*
201  * The initial retransmission should happen at rtt + 4 * rttvar.
202  * Because of the way we do the smoothing, srtt and rttvar
203  * will each average +1/2 tick of bias.  When we compute
204  * the retransmit timer, we want 1/2 tick of rounding and
205  * 1 extra tick because of +-1/2 tick uncertainty in the
206  * firing of the timer.  The bias will give us exactly the
207  * 1.5 tick we need.  But, because the bias is
208  * statistical, we have to test that we don't drop below
209  * the minimum feasible timer (which is 2 ticks).
210  * This version of the macro adapted from a paper by Lawrence
211  * Brakmo and Larry Peterson which outlines a problem caused
212  * by insufficient precision in the original implementation,
213  * which results in inappropriately large RTO values for very
214  * fast networks.
215  */
216 #define	TCP_REXMTVAL(tp) \
217 	((((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
218 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
219 
220 /* XXX
221  * We want to avoid doing m_pullup on incoming packets but that
222  * means avoiding dtom on the tcp reassembly code.  That in turn means
223  * keeping an mbuf pointer in the reassembly queue (since we might
224  * have a cluster).  As a quick hack, the source & destination
225  * port numbers (which are no longer needed once we've located the
226  * tcpcb) are overlayed with an mbuf pointer.
227  */
228 #define REASS_MBUF(ti) (*(struct mbuf **)&((ti)->ti_t))
229 
230 /*
231  * TCP statistics.
232  * Many of these should be kept per connection,
233  * but that's inconvenient at the moment.
234  */
235 struct	tcpstat {
236 	u_long	tcps_connattempt;	/* connections initiated */
237 	u_long	tcps_accepts;		/* connections accepted */
238 	u_long	tcps_connects;		/* connections established */
239 	u_long	tcps_drops;		/* connections dropped */
240 	u_long	tcps_conndrops;		/* embryonic connections dropped */
241 	u_long	tcps_closed;		/* conn. closed (includes drops) */
242 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
243 	u_long	tcps_rttupdated;	/* times we succeeded */
244 	u_long	tcps_delack;		/* delayed acks sent */
245 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
246 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
247 	u_long	tcps_persisttimeo;	/* persist timeouts */
248 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
249 	u_long	tcps_keepprobe;		/* keepalive probes sent */
250 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
251 
252 	u_long	tcps_sndtotal;		/* total packets sent */
253 	u_long	tcps_sndpack;		/* data packets sent */
254 	u_long	tcps_sndbyte;		/* data bytes sent */
255 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
256 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
257 	u_long	tcps_sndacks;		/* ack-only packets sent */
258 	u_long	tcps_sndprobe;		/* window probes sent */
259 	u_long	tcps_sndurg;		/* packets sent with URG only */
260 	u_long	tcps_sndwinup;		/* window update-only packets sent */
261 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
262 
263 	u_long	tcps_rcvtotal;		/* total packets received */
264 	u_long	tcps_rcvpack;		/* packets received in sequence */
265 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
266 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
267 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
268 	u_long	tcps_rcvshort;		/* packets received too short */
269 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
270 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
271 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
272 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
273 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
274 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
275 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
276 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
277 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
278 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
279 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
280 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
281 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
282 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
283 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
284 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
285 	u_long	tcps_predack;		/* times hdr predict ok for acks */
286 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
287 	u_long	tcps_pcbcachemiss;
288 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
289 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
290 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
291 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
292 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
293 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
294 	u_long	tcps_persistdrop;	/* timeout in persist state */
295 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
296 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
297 	u_long	tcps_listendrop;	/* listen queue overflows */
298 };
299 
300 /*
301  * Names for TCP sysctl objects
302  */
303 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
304 #define	TCPCTL_DO_RFC1644	2	/* use RFC-1644 extensions */
305 #define	TCPCTL_MSSDFLT		3	/* MSS default */
306 #define TCPCTL_STATS		4	/* statistics (read-only) */
307 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
308 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
309 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
310 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
311 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
312 #define	TCPCTL_KEEPINIT		10	/* receive buffer space */
313 #define TCPCTL_MAXID		11
314 
315 #define TCPCTL_NAMES { \
316 	{ 0, 0 }, \
317 	{ "rfc1323", CTLTYPE_INT }, \
318 	{ "rfc1644", CTLTYPE_INT }, \
319 	{ "mssdflt", CTLTYPE_INT }, \
320 	{ "stats", CTLTYPE_STRUCT }, \
321 	{ "rttdflt", CTLTYPE_INT }, \
322 	{ "keepidle", CTLTYPE_INT }, \
323 	{ "keepintvl", CTLTYPE_INT }, \
324 	{ "sendspace", CTLTYPE_INT }, \
325 	{ "recvspace", CTLTYPE_INT }, \
326 	{ "keepinit", CTLTYPE_INT }, \
327 }
328 
329 #ifdef KERNEL
330 extern	struct inpcbhead tcb;		/* head of queue of active tcpcb's */
331 extern	struct inpcbinfo tcbinfo;
332 extern	struct tcpstat tcpstat;	/* tcp statistics */
333 extern	int tcp_mssdflt;	/* XXX */
334 extern	u_long tcp_now;		/* for RFC 1323 timestamps */
335 
336 void	 tcp_canceltimers __P((struct tcpcb *));
337 struct tcpcb *
338 	 tcp_close __P((struct tcpcb *));
339 void	 tcp_ctlinput __P((int, struct sockaddr *, void *));
340 int	 tcp_ctloutput __P((int, struct socket *, int, int, struct mbuf **));
341 struct tcpcb *
342 	 tcp_drop __P((struct tcpcb *, int));
343 void	 tcp_drain __P((void));
344 void	 tcp_fasttimo __P((void));
345 struct rmxp_tao *
346 	 tcp_gettaocache __P((struct inpcb *));
347 void	 tcp_init __P((void));
348 void	 tcp_input __P((struct mbuf *, int));
349 void	 tcp_mss __P((struct tcpcb *, int));
350 int	 tcp_mssopt __P((struct tcpcb *));
351 void	 tcp_mtudisc __P((struct inpcb *, int));
352 struct tcpcb *
353 	 tcp_newtcpcb __P((struct inpcb *));
354 int	 tcp_output __P((struct tcpcb *));
355 void	 tcp_quench __P((struct inpcb *, int));
356 void	 tcp_respond __P((struct tcpcb *,
357 	    struct tcpiphdr *, struct mbuf *, u_long, u_long, int));
358 struct rtentry *
359 	 tcp_rtlookup __P((struct inpcb *));
360 void	 tcp_setpersist __P((struct tcpcb *));
361 void	 tcp_slowtimo __P((void));
362 struct tcpiphdr *
363 	 tcp_template __P((struct tcpcb *));
364 struct tcpcb *
365 	 tcp_timers __P((struct tcpcb *, int));
366 void	 tcp_trace __P((int, int, struct tcpcb *, struct tcpiphdr *, int));
367 
368 extern	struct pr_usrreqs tcp_usrreqs;
369 extern	u_long tcp_sendspace;
370 extern	u_long tcp_recvspace;
371 
372 #endif /* KERNEL */
373 
374 #endif /* _NETINET_TCP_VAR_H_ */
375