xref: /freebsd/sys/netinet/tcp_var.h (revision 41466b50c1d5bfd1cf6adaae547a579a75d7c04e)
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 /* TCP segment queue entry */
44 struct tseg_qent {
45 	LIST_ENTRY(tseg_qent) tqe_q;
46 	int	tqe_len;		/* TCP segment data length */
47 	struct	tcphdr *tqe_th;		/* a pointer to tcp header */
48 	struct	mbuf	*tqe_m;		/* mbuf contains packet */
49 };
50 LIST_HEAD(tsegqe_head, tseg_qent);
51 #ifdef MALLOC_DECLARE
52 MALLOC_DECLARE(M_TSEGQ);
53 #endif
54 
55 struct tcptemp {
56 	u_char	tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
57 	struct	tcphdr tt_t;
58 };
59 
60 #define tcp6cb		tcpcb  /* for KAME src sync over BSD*'s */
61 
62 /*
63  * Tcp control block, one per tcp; fields:
64  * Organized for 16 byte cacheline efficiency.
65  */
66 struct tcpcb {
67 	struct	tsegqe_head t_segq;
68 	int	t_dupacks;		/* consecutive dup acks recd */
69 	struct	tcptemp	*unused;	/* unused */
70 
71 	struct	callout *tt_rexmt;	/* retransmit timer */
72 	struct	callout *tt_persist;	/* retransmit persistence */
73 	struct	callout *tt_keep;	/* keepalive */
74 	struct	callout *tt_2msl;	/* 2*msl TIME_WAIT timer */
75 	struct	callout *tt_delack;	/* delayed ACK timer */
76 
77 	struct	inpcb *t_inpcb;		/* back pointer to internet pcb */
78 	int	t_state;		/* state of this connection */
79 	u_int	t_flags;
80 #define	TF_ACKNOW	0x00001		/* ack peer immediately */
81 #define	TF_DELACK	0x00002		/* ack, but try to delay it */
82 #define	TF_NODELAY	0x00004		/* don't delay packets to coalesce */
83 #define	TF_NOOPT	0x00008		/* don't use tcp options */
84 #define	TF_SENTFIN	0x00010		/* have sent FIN */
85 #define	TF_REQ_SCALE	0x00020		/* have/will request window scaling */
86 #define	TF_RCVD_SCALE	0x00040		/* other side has requested scaling */
87 #define	TF_REQ_TSTMP	0x00080		/* have/will request timestamps */
88 #define	TF_RCVD_TSTMP	0x00100		/* a timestamp was received in SYN */
89 #define	TF_SACK_PERMIT	0x00200		/* other side said I could SACK */
90 #define	TF_NEEDSYN	0x00400		/* send SYN (implicit state) */
91 #define	TF_NEEDFIN	0x00800		/* send FIN (implicit state) */
92 #define	TF_NOPUSH	0x01000		/* don't push */
93 #define	TF_REQ_CC	0x02000		/* have/will request CC */
94 #define	TF_RCVD_CC	0x04000		/* a CC was received in SYN */
95 #define	TF_SENDCCNEW	0x08000		/* send CCnew instead of CC in SYN */
96 #define	TF_MORETOCOME	0x10000		/* More data to be appended to sock */
97 #define	TF_LQ_OVERFLOW	0x20000		/* listen queue overflow */
98 #define	TF_LASTIDLE	0x40000		/* connection was previously idle */
99 	int	t_force;		/* 1 if forcing out a byte */
100 
101 	tcp_seq	snd_una;		/* send unacknowledged */
102 	tcp_seq	snd_max;		/* highest sequence number sent;
103 					 * used to recognize retransmits
104 					 */
105 	tcp_seq	snd_nxt;		/* send next */
106 	tcp_seq	snd_up;			/* send urgent pointer */
107 
108 	tcp_seq	snd_wl1;		/* window update seg seq number */
109 	tcp_seq	snd_wl2;		/* window update seg ack number */
110 	tcp_seq	iss;			/* initial send sequence number */
111 	tcp_seq	irs;			/* initial receive sequence number */
112 
113 	tcp_seq	rcv_nxt;		/* receive next */
114 	tcp_seq	rcv_adv;		/* advertised window */
115 	u_long	rcv_wnd;		/* receive window */
116 	tcp_seq	rcv_up;			/* receive urgent pointer */
117 
118 	u_long	snd_wnd;		/* send window */
119 	u_long	snd_cwnd;		/* congestion-controlled window */
120 	u_long	snd_ssthresh;		/* snd_cwnd size threshold for
121 					 * for slow start exponential to
122 					 * linear switch
123 					 */
124 	tcp_seq	snd_recover;		/* for use in fast recovery */
125 
126 	u_int	t_maxopd;		/* mss plus options */
127 
128 	u_long	t_rcvtime;		/* inactivity time */
129 	u_long	t_starttime;		/* time connection was established */
130 	int	t_rtttime;		/* round trip time */
131 	tcp_seq	t_rtseq;		/* sequence number being timed */
132 
133 	int	t_rxtcur;		/* current retransmit value (ticks) */
134 	u_int	t_maxseg;		/* maximum segment size */
135 	int	t_srtt;			/* smoothed round-trip time */
136 	int	t_rttvar;		/* variance in round-trip time */
137 
138 	int	t_rxtshift;		/* log(2) of rexmt exp. backoff */
139 	u_int	t_rttmin;		/* minimum rtt allowed */
140 	u_long	t_rttupdated;		/* number of times rtt sampled */
141 	u_long	max_sndwnd;		/* largest window peer has offered */
142 
143 	int	t_softerror;		/* possible error not yet reported */
144 /* out-of-band data */
145 	char	t_oobflags;		/* have some */
146 	char	t_iobc;			/* input character */
147 #define	TCPOOB_HAVEDATA	0x01
148 #define	TCPOOB_HADDATA	0x02
149 /* RFC 1323 variables */
150 	u_char	snd_scale;		/* window scaling for send window */
151 	u_char	rcv_scale;		/* window scaling for recv window */
152 	u_char	request_r_scale;	/* pending window scaling */
153 	u_char	requested_s_scale;
154 	u_long	ts_recent;		/* timestamp echo data */
155 
156 	u_long	ts_recent_age;		/* when last updated */
157 	tcp_seq	last_ack_sent;
158 /* RFC 1644 variables */
159 	tcp_cc	cc_send;		/* send connection count */
160 	tcp_cc	cc_recv;		/* receive connection count */
161 /* experimental */
162 	u_long	snd_cwnd_prev;		/* cwnd prior to retransmit */
163 	u_long	snd_ssthresh_prev;	/* ssthresh prior to retransmit */
164 	u_long	t_badrxtwin;		/* window for retransmit recovery */
165 };
166 
167 /*
168  * Structure to hold TCP options that are only used during segment
169  * processing (in tcp_input), but not held in the tcpcb.
170  * It's basically used to reduce the number of parameters
171  * to tcp_dooptions.
172  */
173 struct tcpopt {
174 	u_long	to_flag;		/* which options are present */
175 #define TOF_TS		0x0001		/* timestamp */
176 #define TOF_CC		0x0002		/* CC and CCnew are exclusive */
177 #define TOF_CCNEW	0x0004
178 #define	TOF_CCECHO	0x0008
179 	u_long	to_tsval;
180 	u_long	to_tsecr;
181 	tcp_cc	to_cc;		/* holds CC or CCnew */
182 	tcp_cc	to_ccecho;
183 };
184 
185 /*
186  * The TAO cache entry which is stored in the protocol family specific
187  * portion of the route metrics.
188  */
189 struct rmxp_tao {
190 	tcp_cc	tao_cc;			/* latest CC in valid SYN */
191 	tcp_cc	tao_ccsent;		/* latest CC sent to peer */
192 	u_short	tao_mssopt;		/* peer's cached MSS */
193 #ifdef notyet
194 	u_short	tao_flags;		/* cache status flags */
195 #define	TAOF_DONT	0x0001		/* peer doesn't understand rfc1644 */
196 #define	TAOF_OK		0x0002		/* peer does understand rfc1644 */
197 #define	TAOF_UNDEF	0		/* we don't know yet */
198 #endif /* notyet */
199 };
200 #define rmx_taop(r)	((struct rmxp_tao *)(r).rmx_filler)
201 
202 #define	intotcpcb(ip)	((struct tcpcb *)(ip)->inp_ppcb)
203 #define	sototcpcb(so)	(intotcpcb(sotoinpcb(so)))
204 
205 /*
206  * The smoothed round-trip time and estimated variance
207  * are stored as fixed point numbers scaled by the values below.
208  * For convenience, these scales are also used in smoothing the average
209  * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
210  * With these scales, srtt has 3 bits to the right of the binary point,
211  * and thus an "ALPHA" of 0.875.  rttvar has 2 bits to the right of the
212  * binary point, and is smoothed with an ALPHA of 0.75.
213  */
214 #define	TCP_RTT_SCALE		32	/* multiplier for srtt; 3 bits frac. */
215 #define	TCP_RTT_SHIFT		5	/* shift for srtt; 3 bits frac. */
216 #define	TCP_RTTVAR_SCALE	16	/* multiplier for rttvar; 2 bits */
217 #define	TCP_RTTVAR_SHIFT	4	/* shift for rttvar; 2 bits */
218 #define	TCP_DELTA_SHIFT		2	/* see tcp_input.c */
219 
220 /*
221  * The initial retransmission should happen at rtt + 4 * rttvar.
222  * Because of the way we do the smoothing, srtt and rttvar
223  * will each average +1/2 tick of bias.  When we compute
224  * the retransmit timer, we want 1/2 tick of rounding and
225  * 1 extra tick because of +-1/2 tick uncertainty in the
226  * firing of the timer.  The bias will give us exactly the
227  * 1.5 tick we need.  But, because the bias is
228  * statistical, we have to test that we don't drop below
229  * the minimum feasible timer (which is 2 ticks).
230  * This version of the macro adapted from a paper by Lawrence
231  * Brakmo and Larry Peterson which outlines a problem caused
232  * by insufficient precision in the original implementation,
233  * which results in inappropriately large RTO values for very
234  * fast networks.
235  */
236 #define	TCP_REXMTVAL(tp) \
237 	max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT))  \
238 	  + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
239 
240 /*
241  * TCP statistics.
242  * Many of these should be kept per connection,
243  * but that's inconvenient at the moment.
244  */
245 struct	tcpstat {
246 	u_long	tcps_connattempt;	/* connections initiated */
247 	u_long	tcps_accepts;		/* connections accepted */
248 	u_long	tcps_connects;		/* connections established */
249 	u_long	tcps_drops;		/* connections dropped */
250 	u_long	tcps_conndrops;		/* embryonic connections dropped */
251 	u_long	tcps_closed;		/* conn. closed (includes drops) */
252 	u_long	tcps_segstimed;		/* segs where we tried to get rtt */
253 	u_long	tcps_rttupdated;	/* times we succeeded */
254 	u_long	tcps_delack;		/* delayed acks sent */
255 	u_long	tcps_timeoutdrop;	/* conn. dropped in rxmt timeout */
256 	u_long	tcps_rexmttimeo;	/* retransmit timeouts */
257 	u_long	tcps_persisttimeo;	/* persist timeouts */
258 	u_long	tcps_keeptimeo;		/* keepalive timeouts */
259 	u_long	tcps_keepprobe;		/* keepalive probes sent */
260 	u_long	tcps_keepdrops;		/* connections dropped in keepalive */
261 
262 	u_long	tcps_sndtotal;		/* total packets sent */
263 	u_long	tcps_sndpack;		/* data packets sent */
264 	u_long	tcps_sndbyte;		/* data bytes sent */
265 	u_long	tcps_sndrexmitpack;	/* data packets retransmitted */
266 	u_long	tcps_sndrexmitbyte;	/* data bytes retransmitted */
267 	u_long	tcps_sndacks;		/* ack-only packets sent */
268 	u_long	tcps_sndprobe;		/* window probes sent */
269 	u_long	tcps_sndurg;		/* packets sent with URG only */
270 	u_long	tcps_sndwinup;		/* window update-only packets sent */
271 	u_long	tcps_sndctrl;		/* control (SYN|FIN|RST) packets sent */
272 
273 	u_long	tcps_rcvtotal;		/* total packets received */
274 	u_long	tcps_rcvpack;		/* packets received in sequence */
275 	u_long	tcps_rcvbyte;		/* bytes received in sequence */
276 	u_long	tcps_rcvbadsum;		/* packets received with ccksum errs */
277 	u_long	tcps_rcvbadoff;		/* packets received with bad offset */
278 	u_long	tcps_rcvmemdrop;	/* packets dropped for lack of memory */
279 	u_long	tcps_rcvshort;		/* packets received too short */
280 	u_long	tcps_rcvduppack;	/* duplicate-only packets received */
281 	u_long	tcps_rcvdupbyte;	/* duplicate-only bytes received */
282 	u_long	tcps_rcvpartduppack;	/* packets with some duplicate data */
283 	u_long	tcps_rcvpartdupbyte;	/* dup. bytes in part-dup. packets */
284 	u_long	tcps_rcvoopack;		/* out-of-order packets received */
285 	u_long	tcps_rcvoobyte;		/* out-of-order bytes received */
286 	u_long	tcps_rcvpackafterwin;	/* packets with data after window */
287 	u_long	tcps_rcvbyteafterwin;	/* bytes rcvd after window */
288 	u_long	tcps_rcvafterclose;	/* packets rcvd after "close" */
289 	u_long	tcps_rcvwinprobe;	/* rcvd window probe packets */
290 	u_long	tcps_rcvdupack;		/* rcvd duplicate acks */
291 	u_long	tcps_rcvacktoomuch;	/* rcvd acks for unsent data */
292 	u_long	tcps_rcvackpack;	/* rcvd ack packets */
293 	u_long	tcps_rcvackbyte;	/* bytes acked by rcvd acks */
294 	u_long	tcps_rcvwinupd;		/* rcvd window update packets */
295 	u_long	tcps_pawsdrop;		/* segments dropped due to PAWS */
296 	u_long	tcps_predack;		/* times hdr predict ok for acks */
297 	u_long	tcps_preddat;		/* times hdr predict ok for data pkts */
298 	u_long	tcps_pcbcachemiss;
299 	u_long	tcps_cachedrtt;		/* times cached RTT in route updated */
300 	u_long	tcps_cachedrttvar;	/* times cached rttvar updated */
301 	u_long	tcps_cachedssthresh;	/* times cached ssthresh updated */
302 	u_long	tcps_usedrtt;		/* times RTT initialized from route */
303 	u_long	tcps_usedrttvar;	/* times RTTVAR initialized from rt */
304 	u_long	tcps_usedssthresh;	/* times ssthresh initialized from rt*/
305 	u_long	tcps_persistdrop;	/* timeout in persist state */
306 	u_long	tcps_badsyn;		/* bogus SYN, e.g. premature ACK */
307 	u_long	tcps_mturesent;		/* resends due to MTU discovery */
308 	u_long	tcps_listendrop;	/* listen queue overflows */
309 };
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 
362 #ifdef _KERNEL
363 #ifdef SYSCTL_DECL
364 SYSCTL_DECL(_net_inet_tcp);
365 #endif
366 
367 extern	struct inpcbhead tcb;		/* head of queue of active tcpcb's */
368 extern	struct inpcbinfo tcbinfo;
369 extern	struct tcpstat tcpstat;	/* tcp statistics */
370 extern	int tcp_mssdflt;	/* XXX */
371 extern	int tcp_delack_enabled;
372 extern	int tcp_do_newreno;
373 extern	int ss_fltsz;
374 extern	int ss_fltsz_local;
375 
376 void	 tcp_canceltimers __P((struct tcpcb *));
377 struct tcpcb *
378 	 tcp_close __P((struct tcpcb *));
379 void	 tcp_ctlinput __P((int, struct sockaddr *, void *));
380 int	 tcp_ctloutput __P((struct socket *, struct sockopt *));
381 struct tcpcb *
382 	 tcp_drop __P((struct tcpcb *, int));
383 void	 tcp_drain __P((void));
384 void	 tcp_fasttimo __P((void));
385 struct rmxp_tao *
386 	 tcp_gettaocache __P((struct inpcb *));
387 void	 tcp_init __P((void));
388 void	 tcp_input __P((struct mbuf *, int));
389 void	 tcp_mss __P((struct tcpcb *, int));
390 int	 tcp_mssopt __P((struct tcpcb *));
391 void	 tcp_drop_syn_sent __P((struct inpcb *, int));
392 void	 tcp_mtudisc __P((struct inpcb *, int));
393 struct tcpcb *
394 	 tcp_newtcpcb __P((struct inpcb *));
395 int	 tcp_output __P((struct tcpcb *));
396 void	 tcp_quench __P((struct inpcb *, int));
397 void	 tcp_respond __P((struct tcpcb *, void *,
398 	    struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
399 struct rtentry *
400 	 tcp_rtlookup __P((struct inpcb *));
401 void	 tcp_setpersist __P((struct tcpcb *));
402 void	 tcp_slowtimo __P((void));
403 struct tcptemp *
404 	 tcp_maketemplate __P((struct tcpcb *));
405 void	 tcp_fillheaders __P((struct tcpcb *, void *, void *));
406 struct tcpcb *
407 	 tcp_timers __P((struct tcpcb *, int));
408 void	 tcp_trace __P((int, int, struct tcpcb *, void *, struct tcphdr *,
409 			int));
410 
411 extern	struct pr_usrreqs tcp_usrreqs;
412 extern	u_long tcp_sendspace;
413 extern	u_long tcp_recvspace;
414 tcp_seq tcp_new_isn __P((struct tcpcb *));
415 
416 #endif /* _KERNEL */
417 
418 #endif /* _NETINET_TCP_VAR_H_ */
419