xref: /freebsd/sys/netinet/tcp_var.h (revision a1a4f1a0d87b594d3f17a97dc0127eec1417e6f6)
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  * $FreeBSD$
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	0x00001		/* ack peer immediately */
58 #define	TF_DELACK	0x00002		/* ack, but try to delay it */
59 #define	TF_NODELAY	0x00004		/* don't delay packets to coalesce */
60 #define	TF_NOOPT	0x00008		/* don't use tcp options */
61 #define	TF_SENTFIN	0x00010		/* have sent FIN */
62 #define	TF_REQ_SCALE	0x00020		/* have/will request window scaling */
63 #define	TF_RCVD_SCALE	0x00040		/* other side has requested scaling */
64 #define	TF_REQ_TSTMP	0x00080		/* have/will request timestamps */
65 #define	TF_RCVD_TSTMP	0x00100		/* a timestamp was received in SYN */
66 #define	TF_SACK_PERMIT	0x00200		/* other side said I could SACK */
67 #define	TF_NEEDSYN	0x00400		/* send SYN (implicit state) */
68 #define	TF_NEEDFIN	0x00800		/* send FIN (implicit state) */
69 #define	TF_NOPUSH	0x01000		/* don't push */
70 #define	TF_REQ_CC	0x02000		/* have/will request CC */
71 #define	TF_RCVD_CC	0x04000		/* a CC was received in SYN */
72 #define	TF_SENDCCNEW	0x08000		/* send CCnew instead of CC in SYN */
73 #define	TF_MORETOCOME	0x10000		/* More data to be appended to sock */
74 	int	t_force;		/* 1 if forcing out a byte */
75 
76 	tcp_seq	snd_una;		/* send unacknowledged */
77 	tcp_seq	snd_max;		/* highest sequence number sent;
78 					 * used to recognize retransmits
79 					 */
80 	tcp_seq	snd_nxt;		/* send next */
81 	tcp_seq	snd_up;			/* send urgent pointer */
82 
83 	tcp_seq	snd_wl1;		/* window update seg seq number */
84 	tcp_seq	snd_wl2;		/* window update seg ack number */
85 	tcp_seq	iss;			/* initial send sequence number */
86 	tcp_seq	irs;			/* initial receive sequence number */
87 
88 	tcp_seq	rcv_nxt;		/* receive next */
89 	tcp_seq	rcv_adv;		/* advertised window */
90 	u_long	rcv_wnd;		/* receive window */
91 	tcp_seq	rcv_up;			/* receive urgent pointer */
92 
93 	u_long	snd_wnd;		/* send window */
94 	u_long	snd_cwnd;		/* congestion-controlled window */
95 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
96 					 * for slow start exponential to
97 					 * linear switch
98 					 */
99 	u_int	t_maxopd;		/* mss plus options */
100 
101 	u_int	t_idle;			/* inactivity time */
102 	u_long	t_duration;		/* connection duration */
103 	int	t_rtt;			/* round trip time */
104 	tcp_seq	t_rtseq;		/* sequence number being timed */
105 
106 	int	t_rxtcur;		/* current retransmit value */
107 	u_int	t_maxseg;		/* maximum segment size */
108 	int	t_srtt;			/* smoothed round-trip time */
109 	int	t_rttvar;		/* variance in round-trip time */
110 
111 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
112 	u_int	t_rttmin;		/* minimum rtt allowed */
113 	u_long	t_rttupdated;		/* number of times rtt sampled */
114 	u_long	max_sndwnd;		/* largest window peer has offered */
115 
116 	int	t_softerror;		/* possible error not yet reported */
117 /* out-of-band data */
118 	char	t_oobflags;		/* have some */
119 	char	t_iobc;			/* input character */
120 #define	TCPOOB_HAVEDATA	0x01
121 #define	TCPOOB_HADDATA	0x02
122 /* RFC 1323 variables */
123 	u_char	snd_scale;		/* window scaling for send window */
124 	u_char	rcv_scale;		/* window scaling for recv window */
125 	u_char	request_r_scale;	/* pending window scaling */
126 	u_char	requested_s_scale;
127 	u_long	ts_recent;		/* timestamp echo data */
128 
129 	u_long	ts_recent_age;		/* when last updated */
130 	tcp_seq	last_ack_sent;
131 /* RFC 1644 variables */
132 	tcp_cc	cc_send;		/* send connection count */
133 	tcp_cc	cc_recv;		/* receive connection count */
134 };
135 
136 /*
137  * Structure to hold TCP options that are only used during segment
138  * processing (in tcp_input), but not held in the tcpcb.
139  * It's basically used to reduce the number of parameters
140  * to tcp_dooptions.
141  */
142 struct tcpopt {
143 	u_long	to_flag;		/* which options are present */
144 #define TOF_TS		0x0001		/* timestamp */
145 #define TOF_CC		0x0002		/* CC and CCnew are exclusive */
146 #define TOF_CCNEW	0x0004
147 #define	TOF_CCECHO	0x0008
148 	u_long	to_tsval;
149 	u_long	to_tsecr;
150 	tcp_cc	to_cc;		/* holds CC or CCnew */
151 	tcp_cc	to_ccecho;
152 };
153 
154 /*
155  * The TAO cache entry which is stored in the protocol family specific
156  * portion of the route metrics.
157  */
158 struct rmxp_tao {
159 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
160 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
161 	u_short	tao_mssopt;		/* peer's cached MSS */
162 #ifdef notyet
163 	u_short	tao_flags;		/* cache status flags */
164 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
165 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
166 #define	TAOF_UNDEF	0		/* we don't know yet */
167 #endif /* notyet */
168 };
169 #define rmx_taop(r)	((struct rmxp_tao *)(r).rmx_filler)
170 
171 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
172 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
173 
174 /*
175  * The smoothed round-trip time and estimated variance
176  * are stored as fixed point numbers scaled by the values below.
177  * For convenience, these scales are also used in smoothing the average
178  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
179  * With these scales, srtt has 3 bits to the right of the binary point,
180  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
181  * binary point, and is smoothed with an ALPHA of 0.75.
182  */
183 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
184 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
185 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
186 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
187 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
188 
189 /*
190  * The initial retransmission should happen at rtt + 4 * rttvar.
191  * Because of the way we do the smoothing, srtt and rttvar
192  * will each average +1/2 tick of bias.  When we compute
193  * the retransmit timer, we want 1/2 tick of rounding and
194  * 1 extra tick because of +-1/2 tick uncertainty in the
195  * firing of the timer.  The bias will give us exactly the
196  * 1.5 tick we need.  But, because the bias is
197  * statistical, we have to test that we don't drop below
198  * the minimum feasible timer (which is 2 ticks).
199  * This version of the macro adapted from a paper by Lawrence
200  * Brakmo and Larry Peterson which outlines a problem caused
201  * by insufficient precision in the original implementation,
202  * which results in inappropriately large RTO values for very
203  * fast networks.
204  */
205 #define	TCP_REXMTVAL(tp) \
206 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
207 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
208 
209 /*
210  * TCP statistics.
211  * Many of these should be kept per connection,
212  * but that's inconvenient at the moment.
213  */
214 struct	tcpstat {
215 	u_long	tcps_connattempt;	/* connections initiated */
216 	u_long	tcps_accepts;		/* connections accepted */
217 	u_long	tcps_connects;		/* connections established */
218 	u_long	tcps_drops;		/* connections dropped */
219 	u_long	tcps_conndrops;		/* embryonic connections dropped */
220 	u_long	tcps_closed;		/* conn. closed (includes drops) */
221 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
222 	u_long	tcps_rttupdated;	/* times we succeeded */
223 	u_long	tcps_delack;		/* delayed acks sent */
224 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
225 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
226 	u_long	tcps_persisttimeo;	/* persist timeouts */
227 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
228 	u_long	tcps_keepprobe;		/* keepalive probes sent */
229 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
230 
231 	u_long	tcps_sndtotal;		/* total packets sent */
232 	u_long	tcps_sndpack;		/* data packets sent */
233 	u_long	tcps_sndbyte;		/* data bytes sent */
234 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
235 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
236 	u_long	tcps_sndacks;		/* ack-only packets sent */
237 	u_long	tcps_sndprobe;		/* window probes sent */
238 	u_long	tcps_sndurg;		/* packets sent with URG only */
239 	u_long	tcps_sndwinup;		/* window update-only packets sent */
240 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
241 
242 	u_long	tcps_rcvtotal;		/* total packets received */
243 	u_long	tcps_rcvpack;		/* packets received in sequence */
244 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
245 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
246 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
247 	u_long	tcps_rcvshort;		/* packets received too short */
248 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
249 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
250 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
251 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
252 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
253 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
254 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
255 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
256 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
257 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
258 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
259 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
260 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
261 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
262 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
263 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
264 	u_long	tcps_predack;		/* times hdr predict ok for acks */
265 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
266 	u_long	tcps_pcbcachemiss;
267 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
268 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
269 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
270 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
271 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
272 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
273 	u_long	tcps_persistdrop;	/* timeout in persist state */
274 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
275 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
276 	u_long	tcps_listendrop;	/* listen queue overflows */
277 };
278 
279 /*
280  * TCB structure exported to user-land via sysctl(3).
281  * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
282  * included.  Not all of our clients do.
283  */
284 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
285 struct	xtcpcb {
286 	size_t	xt_len;
287 	struct	inpcb	xt_inp;
288 	struct	tcpcb	xt_tp;
289 	struct	xsocket	xt_socket;
290 	u_quad_t	xt_alignment_hack;
291 };
292 #endif
293 
294 /*
295  * Names for TCP sysctl objects
296  */
297 #define	TCPCTL_DO_RFC1323	1	/* use RFC-1323 extensions */
298 #define	TCPCTL_DO_RFC1644	2	/* use RFC-1644 extensions */
299 #define	TCPCTL_MSSDFLT		3	/* MSS default */
300 #define TCPCTL_STATS		4	/* statistics (read-only) */
301 #define	TCPCTL_RTTDFLT		5	/* default RTT estimate */
302 #define	TCPCTL_KEEPIDLE		6	/* keepalive idle timer */
303 #define	TCPCTL_KEEPINTVL	7	/* interval to send keepalives */
304 #define	TCPCTL_SENDSPACE	8	/* send buffer space */
305 #define	TCPCTL_RECVSPACE	9	/* receive buffer space */
306 #define	TCPCTL_KEEPINIT		10	/* receive buffer space */
307 #define	TCPCTL_PCBLIST		11	/* list of all outstanding PCBs */
308 #define TCPCTL_MAXID		12
309 
310 #define TCPCTL_NAMES { \
311 	{ 0, 0 }, \
312 	{ "rfc1323", CTLTYPE_INT }, \
313 	{ "rfc1644", CTLTYPE_INT }, \
314 	{ "mssdflt", CTLTYPE_INT }, \
315 	{ "stats", CTLTYPE_STRUCT }, \
316 	{ "rttdflt", CTLTYPE_INT }, \
317 	{ "keepidle", CTLTYPE_INT }, \
318 	{ "keepintvl", CTLTYPE_INT }, \
319 	{ "sendspace", CTLTYPE_INT }, \
320 	{ "recvspace", CTLTYPE_INT }, \
321 	{ "keepinit", CTLTYPE_INT }, \
322 	{ "pcblist", CTLTYPE_STRUCT }, \
323 }
324 
325 
326 #ifdef KERNEL
327 #ifdef SYSCTL_DECL
328 SYSCTL_DECL(_net_inet_tcp);
329 #endif
330 
331 extern	struct inpcbhead tcb;		/* head of queue of active tcpcb's */
332 extern	struct inpcbinfo tcbinfo;
333 extern	struct tcpstat tcpstat;	/* tcp statistics */
334 extern	int tcp_mssdflt;	/* XXX */
335 extern	u_long tcp_now;		/* for RFC 1323 timestamps */
336 extern	int tcp_delack_enabled;
337 
338 void	 tcp_canceltimers __P((struct tcpcb *));
339 struct tcpcb *
340 	 tcp_close __P((struct tcpcb *));
341 void	 tcp_ctlinput __P((int, struct sockaddr *, void *));
342 int	 tcp_ctloutput __P((struct socket *, struct sockopt *));
343 struct tcpcb *
344 	 tcp_drop __P((struct tcpcb *, int));
345 void	 tcp_drain __P((void));
346 void	 tcp_fasttimo __P((void));
347 struct rmxp_tao *
348 	 tcp_gettaocache __P((struct inpcb *));
349 void	 tcp_init __P((void));
350 void	 tcp_input __P((struct mbuf *, int));
351 void	 tcp_mss __P((struct tcpcb *, int));
352 int	 tcp_mssopt __P((struct tcpcb *));
353 void	 tcp_mtudisc __P((struct inpcb *, int));
354 struct tcpcb *
355 	 tcp_newtcpcb __P((struct inpcb *));
356 int	 tcp_output __P((struct tcpcb *));
357 void	 tcp_quench __P((struct inpcb *, int));
358 void	 tcp_respond __P((struct tcpcb *,
359 	    struct tcpiphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
360 struct rtentry *
361 	 tcp_rtlookup __P((struct inpcb *));
362 void	 tcp_setpersist __P((struct tcpcb *));
363 void	 tcp_slowtimo __P((void));
364 struct tcpiphdr *
365 	 tcp_template __P((struct tcpcb *));
366 struct tcpcb *
367 	 tcp_timers __P((struct tcpcb *, int));
368 void	 tcp_trace __P((int, int, struct tcpcb *, struct tcpiphdr *, int));
369 
370 extern	struct pr_usrreqs tcp_usrreqs;
371 extern	u_long tcp_sendspace;
372 extern	u_long tcp_recvspace;
373 
374 #endif /* KERNEL */
375 
376 #endif /* _NETINET_TCP_VAR_H_ */
377