2 * Copyright (c) 1982, 1986, 1993, 1994, 1995
3 * The Regents of the University of California. All rights reserved.
5 * Redistribution and use in source and binary forms, with or without
6 * modification, are permitted provided that the following conditions
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.
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
33 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
34 * $FreeBSD: src/sys/netinet/tcp_var.h,v 1.56.2.13 2003/02/03 02:34:07 hsu Exp $
35 * $DragonFly: src/sys/netinet/tcp_var.h,v 1.2 2003/06/17 04:28:51 dillon Exp $
38 #ifndef _NETINET_TCP_VAR_H_
39 #define _NETINET_TCP_VAR_H_
41 #include <netinet/in_pcb.h> /* needed for in_conninfo, inp_gen_t */
44 * Kernel variables for tcp.
46 extern int tcp_do_rfc1323;
47 extern int tcp_do_rfc1644;
49 /* TCP segment queue entry */
51 LIST_ENTRY(tseg_qent) tqe_q;
52 int tqe_len; /* TCP segment data length */
53 struct tcphdr *tqe_th; /* a pointer to tcp header */
54 struct mbuf *tqe_m; /* mbuf contains packet */
56 LIST_HEAD(tsegqe_head, tseg_qent);
58 MALLOC_DECLARE(M_TSEGQ);
62 u_char tt_ipgen[40]; /* the size must be of max ip header, now IPv6 */
66 #define tcp6cb tcpcb /* for KAME src sync over BSD*'s */
69 * Tcp control block, one per tcp; fields:
70 * Organized for 16 byte cacheline efficiency.
73 struct tsegqe_head t_segq;
74 int t_dupacks; /* consecutive dup acks recd */
75 struct tcptemp *unused; /* unused */
77 struct callout *tt_rexmt; /* retransmit timer */
78 struct callout *tt_persist; /* retransmit persistence */
79 struct callout *tt_keep; /* keepalive */
80 struct callout *tt_2msl; /* 2*msl TIME_WAIT timer */
81 struct callout *tt_delack; /* delayed ACK timer */
83 struct inpcb *t_inpcb; /* back pointer to internet pcb */
84 int t_state; /* state of this connection */
86 #define TF_ACKNOW 0x00001 /* ack peer immediately */
87 #define TF_DELACK 0x00002 /* ack, but try to delay it */
88 #define TF_NODELAY 0x00004 /* don't delay packets to coalesce */
89 #define TF_NOOPT 0x00008 /* don't use tcp options */
90 #define TF_SENTFIN 0x00010 /* have sent FIN */
91 #define TF_REQ_SCALE 0x00020 /* have/will request window scaling */
92 #define TF_RCVD_SCALE 0x00040 /* other side has requested scaling */
93 #define TF_REQ_TSTMP 0x00080 /* have/will request timestamps */
94 #define TF_RCVD_TSTMP 0x00100 /* a timestamp was received in SYN */
95 #define TF_SACK_PERMIT 0x00200 /* other side said I could SACK */
96 #define TF_NEEDSYN 0x00400 /* send SYN (implicit state) */
97 #define TF_NEEDFIN 0x00800 /* send FIN (implicit state) */
98 #define TF_NOPUSH 0x01000 /* don't push */
99 #define TF_REQ_CC 0x02000 /* have/will request CC */
100 #define TF_RCVD_CC 0x04000 /* a CC was received in SYN */
101 #define TF_SENDCCNEW 0x08000 /* send CCnew instead of CC in SYN */
102 #define TF_MORETOCOME 0x10000 /* More data to be appended to sock */
103 #define TF_LQ_OVERFLOW 0x20000 /* listen queue overflow */
104 #define TF_LASTIDLE 0x40000 /* connection was previously idle */
105 #define TF_RXWIN0SENT 0x80000 /* sent a receiver win 0 in response */
106 int t_force; /* 1 if forcing out a byte */
108 tcp_seq snd_una; /* send unacknowledged */
109 tcp_seq snd_max; /* highest sequence number sent;
110 * used to recognize retransmits
112 tcp_seq snd_nxt; /* send next */
113 tcp_seq snd_up; /* send urgent pointer */
115 tcp_seq snd_wl1; /* window update seg seq number */
116 tcp_seq snd_wl2; /* window update seg ack number */
117 tcp_seq iss; /* initial send sequence number */
118 tcp_seq irs; /* initial receive sequence number */
120 tcp_seq rcv_nxt; /* receive next */
121 tcp_seq rcv_adv; /* advertised window */
122 u_long rcv_wnd; /* receive window */
123 tcp_seq rcv_up; /* receive urgent pointer */
125 u_long snd_wnd; /* send window */
126 u_long snd_cwnd; /* congestion-controlled window */
127 u_long snd_bwnd; /* bandwidth-controlled window */
128 u_long snd_ssthresh; /* snd_cwnd size threshold for
129 * for slow start exponential to
132 u_long snd_bandwidth; /* calculated bandwidth or 0 */
133 tcp_seq snd_recover; /* for use in NewReno fast recovery */
134 tcp_seq snd_high; /* for use in NewReno Fast Recovery */
136 u_int t_maxopd; /* mss plus options */
138 u_long t_rcvtime; /* inactivity time */
139 u_long t_starttime; /* time connection was established */
140 int t_rtttime; /* round trip time */
141 tcp_seq t_rtseq; /* sequence number being timed */
143 int t_bw_rtttime; /* used for bandwidth calculation */
144 tcp_seq t_bw_rtseq; /* used for bandwidth calculation */
146 int t_rxtcur; /* current retransmit value (ticks) */
147 u_int t_maxseg; /* maximum segment size */
148 int t_srtt; /* smoothed round-trip time */
149 int t_rttvar; /* variance in round-trip time */
151 int t_rxtshift; /* log(2) of rexmt exp. backoff */
152 u_int t_rttmin; /* minimum rtt allowed */
153 u_int t_rttbest; /* best rtt we've seen */
154 u_long t_rttupdated; /* number of times rtt sampled */
155 u_long max_sndwnd; /* largest window peer has offered */
157 int t_softerror; /* possible error not yet reported */
158 /* out-of-band data */
159 char t_oobflags; /* have some */
160 char t_iobc; /* input character */
161 #define TCPOOB_HAVEDATA 0x01
162 #define TCPOOB_HADDATA 0x02
163 /* RFC 1323 variables */
164 u_char snd_scale; /* window scaling for send window */
165 u_char rcv_scale; /* window scaling for recv window */
166 u_char request_r_scale; /* pending window scaling */
167 u_char requested_s_scale;
168 u_long ts_recent; /* timestamp echo data */
170 u_long ts_recent_age; /* when last updated */
171 tcp_seq last_ack_sent;
172 /* RFC 1644 variables */
173 tcp_cc cc_send; /* send connection count */
174 tcp_cc cc_recv; /* receive connection count */
176 u_long snd_cwnd_prev; /* cwnd prior to retransmit */
177 u_long snd_ssthresh_prev; /* ssthresh prior to retransmit */
178 tcp_seq snd_high_prev; /* snd_high prior to retransmit */
179 u_long t_badrxtwin; /* window for retransmit recovery */
183 * Structure to hold TCP options that are only used during segment
184 * processing (in tcp_input), but not held in the tcpcb.
185 * It's basically used to reduce the number of parameters
189 u_long to_flags; /* which options are present */
190 #define TOF_TS 0x0001 /* timestamp */
191 #define TOF_CC 0x0002 /* CC and CCnew are exclusive */
192 #define TOF_CCNEW 0x0004
193 #define TOF_CCECHO 0x0008
194 #define TOF_MSS 0x0010
195 #define TOF_SCALE 0x0020
198 tcp_cc to_cc; /* holds CC or CCnew */
201 u_int8_t to_requested_s_scale;
206 inp_gen_t sc_inp_gencnt; /* pointer check */
207 struct tcpcb *sc_tp; /* tcb for listening socket */
208 struct mbuf *sc_ipopts; /* source route */
209 struct in_conninfo sc_inc; /* addresses */
210 #define sc_route sc_inc.inc_route
211 #define sc_route6 sc_inc.inc6_route
212 u_int32_t sc_tsrecent;
213 tcp_cc sc_cc_send; /* holds CC or CCnew */
215 tcp_seq sc_irs; /* seq from peer */
216 tcp_seq sc_iss; /* our ISS */
217 u_long sc_rxttime; /* retransmit time */
218 u_int16_t sc_rxtslot; /* retransmit counter */
219 u_int16_t sc_peer_mss; /* peer's MSS */
220 u_int16_t sc_wnd; /* advertised window */
221 u_int8_t sc_requested_s_scale:4,
222 sc_request_r_scale:4;
224 #define SCF_NOOPT 0x01 /* no TCP options */
225 #define SCF_WINSCALE 0x02 /* negotiated window scaling */
226 #define SCF_TIMESTAMP 0x04 /* negotiated timestamps */
227 #define SCF_CC 0x08 /* negotiated CC */
228 #define SCF_UNREACH 0x10 /* icmp unreachable received */
229 #define SCF_KEEPROUTE 0x20 /* keep cloned route */
230 TAILQ_ENTRY(syncache) sc_hash;
231 TAILQ_ENTRY(syncache) sc_timerq;
234 struct syncache_head {
235 TAILQ_HEAD(, syncache) sch_bucket;
240 * The TAO cache entry which is stored in the protocol family specific
241 * portion of the route metrics.
244 tcp_cc tao_cc; /* latest CC in valid SYN */
245 tcp_cc tao_ccsent; /* latest CC sent to peer */
246 u_short tao_mssopt; /* peer's cached MSS */
248 u_short tao_flags; /* cache status flags */
249 #define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
250 #define TAOF_OK 0x0002 /* peer does understand rfc1644 */
251 #define TAOF_UNDEF 0 /* we don't know yet */
254 #define rmx_taop(r) ((struct rmxp_tao *)(r).rmx_filler)
256 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
257 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
260 * The smoothed round-trip time and estimated variance
261 * are stored as fixed point numbers scaled by the values below.
262 * For convenience, these scales are also used in smoothing the average
263 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
264 * With these scales, srtt has 3 bits to the right of the binary point,
265 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
266 * binary point, and is smoothed with an ALPHA of 0.75.
268 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
269 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
270 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
271 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
272 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
275 * The initial retransmission should happen at rtt + 4 * rttvar.
276 * Because of the way we do the smoothing, srtt and rttvar
277 * will each average +1/2 tick of bias. When we compute
278 * the retransmit timer, we want 1/2 tick of rounding and
279 * 1 extra tick because of +-1/2 tick uncertainty in the
280 * firing of the timer. The bias will give us exactly the
281 * 1.5 tick we need. But, because the bias is
282 * statistical, we have to test that we don't drop below
283 * the minimum feasible timer (which is 2 ticks).
284 * This version of the macro adapted from a paper by Lawrence
285 * Brakmo and Larry Peterson which outlines a problem caused
286 * by insufficient precision in the original implementation,
287 * which results in inappropriately large RTO values for very
290 #define TCP_REXMTVAL(tp) \
291 max((tp)->t_rttmin, (((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
292 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
296 * Many of these should be kept per connection,
297 * but that's inconvenient at the moment.
300 u_long tcps_connattempt; /* connections initiated */
301 u_long tcps_accepts; /* connections accepted */
302 u_long tcps_connects; /* connections established */
303 u_long tcps_drops; /* connections dropped */
304 u_long tcps_conndrops; /* embryonic connections dropped */
305 u_long tcps_closed; /* conn. closed (includes drops) */
306 u_long tcps_segstimed; /* segs where we tried to get rtt */
307 u_long tcps_rttupdated; /* times we succeeded */
308 u_long tcps_delack; /* delayed acks sent */
309 u_long tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
310 u_long tcps_rexmttimeo; /* retransmit timeouts */
311 u_long tcps_persisttimeo; /* persist timeouts */
312 u_long tcps_keeptimeo; /* keepalive timeouts */
313 u_long tcps_keepprobe; /* keepalive probes sent */
314 u_long tcps_keepdrops; /* connections dropped in keepalive */
316 u_long tcps_sndtotal; /* total packets sent */
317 u_long tcps_sndpack; /* data packets sent */
318 u_long tcps_sndbyte; /* data bytes sent */
319 u_long tcps_sndrexmitpack; /* data packets retransmitted */
320 u_long tcps_sndrexmitbyte; /* data bytes retransmitted */
321 u_long tcps_sndacks; /* ack-only packets sent */
322 u_long tcps_sndprobe; /* window probes sent */
323 u_long tcps_sndurg; /* packets sent with URG only */
324 u_long tcps_sndwinup; /* window update-only packets sent */
325 u_long tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
327 u_long tcps_rcvtotal; /* total packets received */
328 u_long tcps_rcvpack; /* packets received in sequence */
329 u_long tcps_rcvbyte; /* bytes received in sequence */
330 u_long tcps_rcvbadsum; /* packets received with ccksum errs */
331 u_long tcps_rcvbadoff; /* packets received with bad offset */
332 u_long tcps_rcvmemdrop; /* packets dropped for lack of memory */
333 u_long tcps_rcvshort; /* packets received too short */
334 u_long tcps_rcvduppack; /* duplicate-only packets received */
335 u_long tcps_rcvdupbyte; /* duplicate-only bytes received */
336 u_long tcps_rcvpartduppack; /* packets with some duplicate data */
337 u_long tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
338 u_long tcps_rcvoopack; /* out-of-order packets received */
339 u_long tcps_rcvoobyte; /* out-of-order bytes received */
340 u_long tcps_rcvpackafterwin; /* packets with data after window */
341 u_long tcps_rcvbyteafterwin; /* bytes rcvd after window */
342 u_long tcps_rcvafterclose; /* packets rcvd after "close" */
343 u_long tcps_rcvwinprobe; /* rcvd window probe packets */
344 u_long tcps_rcvdupack; /* rcvd duplicate acks */
345 u_long tcps_rcvacktoomuch; /* rcvd acks for unsent data */
346 u_long tcps_rcvackpack; /* rcvd ack packets */
347 u_long tcps_rcvackbyte; /* bytes acked by rcvd acks */
348 u_long tcps_rcvwinupd; /* rcvd window update packets */
349 u_long tcps_pawsdrop; /* segments dropped due to PAWS */
350 u_long tcps_predack; /* times hdr predict ok for acks */
351 u_long tcps_preddat; /* times hdr predict ok for data pkts */
352 u_long tcps_pcbcachemiss;
353 u_long tcps_cachedrtt; /* times cached RTT in route updated */
354 u_long tcps_cachedrttvar; /* times cached rttvar updated */
355 u_long tcps_cachedssthresh; /* times cached ssthresh updated */
356 u_long tcps_usedrtt; /* times RTT initialized from route */
357 u_long tcps_usedrttvar; /* times RTTVAR initialized from rt */
358 u_long tcps_usedssthresh; /* times ssthresh initialized from rt*/
359 u_long tcps_persistdrop; /* timeout in persist state */
360 u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */
361 u_long tcps_mturesent; /* resends due to MTU discovery */
362 u_long tcps_listendrop; /* listen queue overflows */
364 u_long tcps_sc_added; /* entry added to syncache */
365 u_long tcps_sc_retransmitted; /* syncache entry was retransmitted */
366 u_long tcps_sc_dupsyn; /* duplicate SYN packet */
367 u_long tcps_sc_dropped; /* could not reply to packet */
368 u_long tcps_sc_completed; /* successful extraction of entry */
369 u_long tcps_sc_bucketoverflow; /* syncache per-bucket limit hit */
370 u_long tcps_sc_cacheoverflow; /* syncache cache limit hit */
371 u_long tcps_sc_reset; /* RST removed entry from syncache */
372 u_long tcps_sc_stale; /* timed out or listen socket gone */
373 u_long tcps_sc_aborted; /* syncache entry aborted */
374 u_long tcps_sc_badack; /* removed due to bad ACK */
375 u_long tcps_sc_unreach; /* ICMP unreachable received */
376 u_long tcps_sc_zonefail; /* zalloc() failed */
377 u_long tcps_sc_sendcookie; /* SYN cookie sent */
378 u_long tcps_sc_recvcookie; /* SYN cookie received */
382 * TCB structure exported to user-land via sysctl(3).
383 * Evil hack: declare only if in_pcb.h and sys/socketvar.h have been
384 * included. Not all of our clients do.
386 #if defined(_NETINET_IN_PCB_H_) && defined(_SYS_SOCKETVAR_H_)
391 struct xsocket xt_socket;
392 u_quad_t xt_alignment_hack;
397 * Names for TCP sysctl objects
399 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
400 #define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
401 #define TCPCTL_MSSDFLT 3 /* MSS default */
402 #define TCPCTL_STATS 4 /* statistics (read-only) */
403 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
404 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
405 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
406 #define TCPCTL_SENDSPACE 8 /* send buffer space */
407 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
408 #define TCPCTL_KEEPINIT 10 /* timeout for establishing syn */
409 #define TCPCTL_PCBLIST 11 /* list of all outstanding PCBs */
410 #define TCPCTL_DELACKTIME 12 /* time before sending delayed ACK */
411 #define TCPCTL_V6MSSDFLT 13 /* MSS default for IPv6 */
412 #define TCPCTL_MAXID 14
414 #define TCPCTL_NAMES { \
416 { "rfc1323", CTLTYPE_INT }, \
417 { "rfc1644", CTLTYPE_INT }, \
418 { "mssdflt", CTLTYPE_INT }, \
419 { "stats", CTLTYPE_STRUCT }, \
420 { "rttdflt", CTLTYPE_INT }, \
421 { "keepidle", CTLTYPE_INT }, \
422 { "keepintvl", CTLTYPE_INT }, \
423 { "sendspace", CTLTYPE_INT }, \
424 { "recvspace", CTLTYPE_INT }, \
425 { "keepinit", CTLTYPE_INT }, \
426 { "pcblist", CTLTYPE_STRUCT }, \
427 { "delacktime", CTLTYPE_INT }, \
428 { "v6mssdflt", CTLTYPE_INT }, \
434 SYSCTL_DECL(_net_inet_tcp);
437 extern struct inpcbhead tcb; /* head of queue of active tcpcb's */
438 extern struct inpcbinfo tcbinfo;
439 extern struct tcpstat tcpstat; /* tcp statistics */
440 extern int tcp_mssdflt; /* XXX */
441 extern int tcp_delack_enabled;
442 extern int tcp_do_newreno;
443 extern int path_mtu_discovery;
445 extern int ss_fltsz_local;
447 void tcp_canceltimers __P((struct tcpcb *));
449 tcp_close __P((struct tcpcb *));
450 void tcp_ctlinput __P((int, struct sockaddr *, void *));
451 int tcp_ctloutput __P((struct socket *, struct sockopt *));
453 tcp_drop __P((struct tcpcb *, int));
454 void tcp_drain __P((void));
455 void tcp_fasttimo __P((void));
457 tcp_gettaocache __P((struct in_conninfo *));
458 void tcp_init __P((void));
459 void tcp_input __P((struct mbuf *, int, int));
460 void tcp_mss __P((struct tcpcb *, int));
461 int tcp_mssopt __P((struct tcpcb *));
462 void tcp_drop_syn_sent __P((struct inpcb *, int));
463 void tcp_mtudisc __P((struct inpcb *, int));
465 tcp_newtcpcb __P((struct inpcb *));
466 int tcp_output __P((struct tcpcb *));
467 void tcp_quench __P((struct inpcb *, int));
468 void tcp_respond __P((struct tcpcb *, void *,
469 struct tcphdr *, struct mbuf *, tcp_seq, tcp_seq, int));
471 tcp_rtlookup __P((struct in_conninfo *));
472 void tcp_setpersist __P((struct tcpcb *));
473 void tcp_slowtimo __P((void));
475 tcp_maketemplate __P((struct tcpcb *));
476 void tcp_fillheaders __P((struct tcpcb *, void *, void *));
478 tcp_timers __P((struct tcpcb *, int));
479 void tcp_trace __P((int, int, struct tcpcb *, void *, struct tcphdr *,
481 void tcp_xmit_bandwidth_limit(struct tcpcb *tp, tcp_seq ack_seq);
482 void syncache_init(void);
483 void syncache_unreach(struct in_conninfo *, struct tcphdr *);
484 int syncache_expand(struct in_conninfo *, struct tcphdr *,
485 struct socket **, struct mbuf *);
486 int syncache_add(struct in_conninfo *, struct tcpopt *,
487 struct tcphdr *, struct socket **, struct mbuf *);
488 void syncache_chkrst(struct in_conninfo *, struct tcphdr *);
489 void syncache_badack(struct in_conninfo *);
491 extern struct pr_usrreqs tcp_usrreqs;
492 extern u_long tcp_sendspace;
493 extern u_long tcp_recvspace;
494 tcp_seq tcp_new_isn __P((struct tcpcb *));
498 #endif /* _NETINET_TCP_VAR_H_ */