xref: /freebsd/sys/netinet/tcp_var.h (revision a8445737e740901f5f2c8d24c12ef7fc8b00134e)
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: tcp_var.h,v 1.47 1998/08/23 03:07:15 wollman Exp $
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  * Organized for 16 byte cacheline efficiency.
46  */
47 struct tcpcb {
48 	struct	mbuf *t_segq;
49 	int	t_dupacks;		/* consecutive dup acks recd */
50 	struct	tcpiphdr *t_template;	/* skeletal packet for transmit */
51 
52 	int	t_timer[TCPT_NTIMERS];	/* tcp timers */
53 
54 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
55 	int	t_state;		/* state of this connection */
56 	u_int	t_flags;
57 #define	TF_ACKNOW	0x0001		/* ack peer immediately */
58 #define	TF_DELACK	0x0002		/* ack, but try to delay it */
59 #define	TF_NODELAY	0x0004		/* don't delay packets to coalesce */
60 #define	TF_NOOPT	0x0008		/* don't use tcp options */
61 #define	TF_SENTFIN	0x0010		/* have sent FIN */
62 #define	TF_REQ_SCALE	0x0020		/* have/will request window scaling */
63 #define	TF_RCVD_SCALE	0x0040		/* other side has requested scaling */
64 #define	TF_REQ_TSTMP	0x0080		/* have/will request timestamps */
65 #define	TF_RCVD_TSTMP	0x0100		/* a timestamp was received in SYN */
66 #define	TF_SACK_PERMIT	0x0200		/* other side said I could SACK */
67 #define TF_NEEDSYN	0x0400		/* send SYN (implicit state) */
68 #define TF_NEEDFIN	0x0800		/* send FIN (implicit state) */
69 #define TF_NOPUSH	0x1000		/* don't push */
70 #define TF_REQ_CC	0x2000		/* have/will request CC */
71 #define	TF_RCVD_CC	0x4000		/* a CC was received in SYN */
72 #define TF_SENDCCNEW	0x8000		/* send CCnew instead of CC in SYN */
73 	int	t_force;		/* 1 if forcing out a byte */
74 
75 	tcp_seq	snd_una;		/* send unacknowledged */
76 	tcp_seq	snd_max;		/* highest sequence number sent;
77 					 * used to recognize retransmits
78 					 */
79 	tcp_seq	snd_nxt;		/* send next */
80 	tcp_seq	snd_up;			/* send urgent pointer */
81 
82 	tcp_seq	snd_wl1;		/* window update seg seq number */
83 	tcp_seq	snd_wl2;		/* window update seg ack number */
84 	tcp_seq	iss;			/* initial send sequence number */
85 	tcp_seq	irs;			/* initial receive sequence number */
86 
87 	tcp_seq	rcv_nxt;		/* receive next */
88 	tcp_seq	rcv_adv;		/* advertised window */
89 	u_long	rcv_wnd;		/* receive window */
90 	tcp_seq	rcv_up;			/* receive urgent pointer */
91 
92 	u_long	snd_wnd;		/* send window */
93 	u_long	snd_cwnd;		/* congestion-controlled window */
94 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
95 					 * for slow start exponential to
96 					 * linear switch
97 					 */
98 	u_int	t_maxopd;		/* mss plus options */
99 
100 	u_int	t_idle;			/* inactivity time */
101 	u_long	t_duration;		/* connection duration */
102 	int	t_rtt;			/* round trip time */
103 	tcp_seq	t_rtseq;		/* sequence number being timed */
104 
105 	int	t_rxtcur;		/* current retransmit value */
106 	u_int	t_maxseg;		/* maximum segment size */
107 	int	t_srtt;			/* smoothed round-trip time */
108 	int	t_rttvar;		/* variance in round-trip time */
109 
110 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
111 	u_int	t_rttmin;		/* minimum rtt allowed */
112 	u_long	t_rttupdated;		/* number of times rtt sampled */
113 	u_long	max_sndwnd;		/* largest window peer has offered */
114 
115 	int	t_softerror;		/* possible error not yet reported */
116 /* out-of-band data */
117 	char	t_oobflags;		/* have some */
118 	char	t_iobc;			/* input character */
119 #define	TCPOOB_HAVEDATA	0x01
120 #define	TCPOOB_HADDATA	0x02
121 /* RFC 1323 variables */
122 	u_char	snd_scale;		/* window scaling for send window */
123 	u_char	rcv_scale;		/* window scaling for recv window */
124 	u_char	request_r_scale;	/* pending window scaling */
125 	u_char	requested_s_scale;
126 	u_long	ts_recent;		/* timestamp echo data */
127 
128 	u_long	ts_recent_age;		/* when last updated */
129 	tcp_seq	last_ack_sent;
130 /* RFC 1644 variables */
131 	tcp_cc	cc_send;		/* send connection count */
132 	tcp_cc	cc_recv;		/* receive connection count */
133 };
134 
135 /*
136  * Structure to hold TCP options that are only used during segment
137  * processing (in tcp_input), but not held in the tcpcb.
138  * It's basically used to reduce the number of parameters
139  * to tcp_dooptions.
140  */
141 struct tcpopt {
142 	u_long	to_flag;		/* which options are present */
143 #define TOF_TS		0x0001		/* timestamp */
144 #define TOF_CC		0x0002		/* CC and CCnew are exclusive */
145 #define TOF_CCNEW	0x0004
146 #define	TOF_CCECHO	0x0008
147 	u_long	to_tsval;
148 	u_long	to_tsecr;
149 	tcp_cc	to_cc;		/* holds CC or CCnew */
150 	tcp_cc	to_ccecho;
151 };
152 
153 /*
154  * The TAO cache entry which is stored in the protocol family specific
155  * portion of the route metrics.
156  */
157 struct rmxp_tao {
158 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
159 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
160 	u_short	tao_mssopt;		/* peer's cached MSS */
161 #ifdef notyet
162 	u_short	tao_flags;		/* cache status flags */
163 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
164 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
165 #define	TAOF_UNDEF	0		/* we don't know yet */
166 #endif /* notyet */
167 };
168 #define rmx_taop(r)	((struct rmxp_tao *)(r).rmx_filler)
169 
170 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
171 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
172 
173 /*
174  * The smoothed round-trip time and estimated variance
175  * are stored as fixed point numbers scaled by the values below.
176  * For convenience, these scales are also used in smoothing the average
177  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
178  * With these scales, srtt has 3 bits to the right of the binary point,
179  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
180  * binary point, and is smoothed with an ALPHA of 0.75.
181  */
182 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
183 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
184 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
185 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
186 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
187 
188 /*
189  * The initial retransmission should happen at rtt + 4 * rttvar.
190  * Because of the way we do the smoothing, srtt and rttvar
191  * will each average +1/2 tick of bias.  When we compute
192  * the retransmit timer, we want 1/2 tick of rounding and
193  * 1 extra tick because of +-1/2 tick uncertainty in the
194  * firing of the timer.  The bias will give us exactly the
195  * 1.5 tick we need.  But, because the bias is
196  * statistical, we have to test that we don't drop below
197  * the minimum feasible timer (which is 2 ticks).
198  * This version of the macro adapted from a paper by Lawrence
199  * Brakmo and Larry Peterson which outlines a problem caused
200  * by insufficient precision in the original implementation,
201  * which results in inappropriately large RTO values for very
202  * fast networks.
203  */
204 #define	TCP_REXMTVAL(tp) \
205 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
206 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
207 
208 /*
209  * TCP statistics.
210  * Many of these should be kept per connection,
211  * but that's inconvenient at the moment.
212  */
213 struct	tcpstat {
214 	u_long	tcps_connattempt;	/* connections initiated */
215 	u_long	tcps_accepts;		/* connections accepted */
216 	u_long	tcps_connects;		/* connections established */
217 	u_long	tcps_drops;		/* connections dropped */
218 	u_long	tcps_conndrops;		/* embryonic connections dropped */
219 	u_long	tcps_closed;		/* conn. closed (includes drops) */
220 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
221 	u_long	tcps_rttupdated;	/* times we succeeded */
222 	u_long	tcps_delack;		/* delayed acks sent */
223 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
224 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
225 	u_long	tcps_persisttimeo;	/* persist timeouts */
226 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
227 	u_long	tcps_keepprobe;		/* keepalive probes sent */
228 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
229 
230 	u_long	tcps_sndtotal;		/* total packets sent */
231 	u_long	tcps_sndpack;		/* data packets sent */
232 	u_long	tcps_sndbyte;		/* data bytes sent */
233 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
234 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
235 	u_long	tcps_sndacks;		/* ack-only packets sent */
236 	u_long	tcps_sndprobe;		/* window probes sent */
237 	u_long	tcps_sndurg;		/* packets sent with URG only */
238 	u_long	tcps_sndwinup;		/* window update-only packets sent */
239 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
240 
241 	u_long	tcps_rcvtotal;		/* total packets received */
242 	u_long	tcps_rcvpack;		/* packets received in sequence */
243 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
244 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
245 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
246 	u_long	tcps_rcvshort;		/* packets received too short */
247 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
248 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
249 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
250 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
251 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
252 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
253 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
254 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
255 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
256 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
257 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
258 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
259 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
260 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
261 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
262 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
263 	u_long	tcps_predack;		/* times hdr predict ok for acks */
264 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
265 	u_long	tcps_pcbcachemiss;
266 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
267 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
268 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
269 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
270 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
271 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
272 	u_long	tcps_persistdrop;	/* timeout in persist state */
273 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
274 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
275 	u_long	tcps_listendrop;	/* listen queue overflows */
276 };
277 
278 /*
279  * TCB structure exported to user-land via sysctl(3).
280  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
281  * included.  Not all of our clients do.
282  */
283 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
284 struct	xtcpcb {
285 	size_t	xt_len;
286 	struct	inpcb	xt_inp;
287 	struct	tcpcb	xt_tp;
288 	struct	xsocket	xt_socket;
289 	u_quad_t	xt_alignment_hack;
290 };
291 #endif
292 
293 /*
294  * Names for TCP sysctl objects
295  */
296 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
297 #define	TCPCTL_DO_RFC1644	2	/* use RFC-1644 extensions */
298 #define	TCPCTL_MSSDFLT		3	/* MSS default */
299 #define TCPCTL_STATS		4	/* statistics (read-only) */
300 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
301 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
302 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
303 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
304 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
305 #define	TCPCTL_KEEPINIT		10	/* receive buffer space */
306 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
307 #define TCPCTL_MAXID		12
308 
309 #define TCPCTL_NAMES { \
310 	{ 0, 0 }, \
311 	{ "rfc1323", CTLTYPE_INT }, \
312 	{ "rfc1644", CTLTYPE_INT }, \
313 	{ "mssdflt", CTLTYPE_INT }, \
314 	{ "stats", CTLTYPE_STRUCT }, \
315 	{ "rttdflt", CTLTYPE_INT }, \
316 	{ "keepidle", CTLTYPE_INT }, \
317 	{ "keepintvl", CTLTYPE_INT }, \
318 	{ "sendspace", CTLTYPE_INT }, \
319 	{ "recvspace", CTLTYPE_INT }, \
320 	{ "keepinit", CTLTYPE_INT }, \
321 	{ "pcblist", CTLTYPE_STRUCT }, \
322 }
323 
324 #ifdef KERNEL
325 extern	struct inpcbhead tcb;		/* head of queue of active tcpcb's */
326 extern	struct inpcbinfo tcbinfo;
327 extern	struct tcpstat tcpstat;	/* tcp statistics */
328 extern	int tcp_mssdflt;	/* XXX */
329 extern	u_long tcp_now;		/* for RFC 1323 timestamps */
330 extern	int tcp_delack_enabled;
331 
332 void	 tcp_canceltimers __P((struct tcpcb *));
333 struct tcpcb *
334 	 tcp_close __P((struct tcpcb *));
335 void	 tcp_ctlinput __P((int, struct sockaddr *, void *));
336 int	 tcp_ctloutput __P((struct socket *, struct sockopt *));
337 struct tcpcb *
338 	 tcp_drop __P((struct tcpcb *, int));
339 void	 tcp_drain __P((void));
340 void	 tcp_fasttimo __P((void));
341 struct rmxp_tao *
342 	 tcp_gettaocache __P((struct inpcb *));
343 void	 tcp_init __P((void));
344 void	 tcp_input __P((struct mbuf *, int));
345 void	 tcp_mss __P((struct tcpcb *, int));
346 int	 tcp_mssopt __P((struct tcpcb *));
347 void	 tcp_mtudisc __P((struct inpcb *, int));
348 struct tcpcb *
349 	 tcp_newtcpcb __P((struct inpcb *));
350 int	 tcp_output __P((struct tcpcb *));
351 void	 tcp_quench __P((struct inpcb *, int));
352 void	 tcp_respond __P((struct tcpcb *,
353 	    struct tcpiphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
354 struct rtentry *
355 	 tcp_rtlookup __P((struct inpcb *));
356 void	 tcp_setpersist __P((struct tcpcb *));
357 void	 tcp_slowtimo __P((void));
358 struct tcpiphdr *
359 	 tcp_template __P((struct tcpcb *));
360 struct tcpcb *
361 	 tcp_timers __P((struct tcpcb *, int));
362 void	 tcp_trace __P((int, int, struct tcpcb *, struct tcpiphdr *, int));
363 
364 extern	struct pr_usrreqs tcp_usrreqs;
365 extern	u_long tcp_sendspace;
366 extern	u_long tcp_recvspace;
367 
368 #endif /* KERNEL */
369 
370 #endif /* _NETINET_TCP_VAR_H_ */
371