1 /*!\file netinet/tcp_var.h
2 * Internal TCP structures and statistics.
6 * Copyright (c) 1982, 1986, 1993, 1994, 1995
7 * The Regents of the University of California. All rights reserved.
9 * Redistribution and use in source and binary forms, with or without
10 * modification, are permitted provided that the following conditions
12 * 1. Redistributions of source code must retain the above copyright
13 * notice, this list of conditions and the following disclaimer.
14 * 2. Redistributions in binary form must reproduce the above copyright
15 * notice, this list of conditions and the following disclaimer in the
16 * documentation and/or other materials provided with the distribution.
17 * 3. All advertising materials mentioning features or use of this software
18 * must display the following acknowledgement:
19 * This product includes software developed by the University of
20 * California, Berkeley and its contributors.
21 * 4. Neither the name of the University nor the names of its contributors
22 * may be used to endorse or promote products derived from this software
23 * without specific prior written permission.
25 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
26 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
27 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
28 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
29 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
30 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
31 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
32 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
33 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
34 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
37 * @(#)tcp_var.h 8.4 (Berkeley) 5/24/95
38 * $Id: tcp_var.h,v 1.36 1996/09/13 23:54:03 pst Exp $
41 #ifndef __NETINET_TCP_VAR_H
42 #define __NETINET_TCP_VAR_H
45 * Kernel variables for tcp.
49 /* AUTO_RCV_HITHRESH flush reassembly queue, drop incoming packets
51 #define AUTO_RCV_HITHRESH (u_long)(0.95 * NMBCLUSTERS)
53 /* AUTO_RCV_LOWTHRESH drop incoming packets
55 #define AUTO_RCV_LOWTHRESH (u_long)(0.9 * NMBCLUSTERS)
57 /* AUTO_SND_THRESH reduce so_snd.sb_hiwat by acked
59 #define AUTO_SND_THRESH (u_long)(0.50 * NMBCLUSTERS)
62 #if defined(TCP_SACK) || defined(TCP_FACK)
63 #include <netinet/tcp_scor.h>
72 * Tcp control block, one per tcp; fields:
75 struct tcpiphdr *seg_next; /* sequencing queue */
76 struct tcpiphdr *seg_prev;
77 /* !! struct ipqehead segq; */ /* sequencing queue */
78 int t_state; /* state of this connection */
79 int t_timer[TCPT_NTIMERS]; /* tcp timers */
80 int t_rxtshift; /* log(2) of rexmt exp. backoff */
81 int t_rxtcur; /* current retransmit value */
83 short t_padd; /* filler to preserve alignment */
85 short t_dupacks; /* consecutive dup acks recd */
87 u_int t_maxseg; /* maximum segment size */
88 u_int t_maxopd; /* mss plus options */
89 int t_force; /* 1 if forcing out a byte */
91 #define TF_ACKNOW 0x0001 /* ack peer immediately */
92 #define TF_DELACK 0x0002 /* ack, but try to delay it */
93 #define TF_NODELAY 0x0004 /* don't delay packets to coalesce */
94 #define TF_NOOPT 0x0008 /* don't use tcp options */
95 #define TF_SENTFIN 0x0010 /* have sent FIN */
96 #define TF_REQ_SCALE 0x0020 /* have/will request window scaling */
97 #define TF_RCVD_SCALE 0x0040 /* other side has requested scaling */
98 #define TF_REQ_TSTMP 0x0080 /* have/will request timestamps */
99 #define TF_RCVD_TSTMP 0x0100 /* a timestamp was received in SYN */
100 #define TF_SACK_PERMIT 0x0200 /* other side said I could SACK */
101 #define TF_NEEDSYN 0x0400 /* send SYN (implicit state) */
102 #define TF_NEEDFIN 0x0800 /* send FIN (implicit state) */
103 #define TF_NOPUSH 0x1000 /* don't push */
104 #define TF_REQ_CC 0x2000 /* have/will request CC */
105 #define TF_RCVD_CC 0x4000 /* a CC was received in SYN */
106 #define TF_SENDCCNEW 0x8000 /* send CCnew instead of CC in SYN */
108 struct tcpiphdr *t_template; /* skeletal packet for transmit */
109 struct inpcb *t_inpcb; /* back pointer to internet pcb */
111 * The following fields are used as in the protocol specification.
112 * See RFC783, Dec. 1981, page 21.
114 /* send sequence variables */
115 tcp_seq snd_una; /* send unacknowledged */
116 tcp_seq snd_nxt; /* send next */
117 tcp_seq snd_up; /* send urgent pointer */
118 tcp_seq snd_wl1; /* window update seg seq number */
119 tcp_seq snd_wl2; /* window update seg ack number */
120 tcp_seq iss; /* initial send sequence number */
121 u_long snd_wnd; /* send window */
122 /* receive sequence variables */
123 u_long rcv_wnd; /* receive window */
124 tcp_seq rcv_nxt; /* receive next */
125 tcp_seq rcv_up; /* receive urgent pointer */
126 tcp_seq irs; /* initial receive sequence number */
128 * Additional variables for this implementation.
130 /* receive variables */
131 tcp_seq rcv_adv; /* advertised window */
132 /* retransmit variables */
133 tcp_seq snd_max; /* highest sequence number sent;
134 * used to recognize retransmits
136 /* congestion control (for slow start, source quench, retransmit after loss) */
137 u_long snd_cwnd; /* congestion-controlled window */
138 u_long snd_ssthresh; /* snd_cwnd size threshold for
139 * for slow start exponential to
143 * transmit timing stuff. See below for scale of srtt and rttvar.
144 * "Variance" is actually smoothed difference.
146 u_int t_idle; /* inactivity time */
147 int t_rtt; /* round trip time */
148 tcp_seq t_rtseq; /* sequence number being timed */
149 int t_srtt; /* smoothed round-trip time */
150 int t_rttvar; /* variance in round-trip time */
151 u_int t_rttmin; /* minimum rtt allowed */
152 u_long max_sndwnd; /* largest window peer has offered */
154 /* out-of-band data */
155 char t_oobflags; /* have some */
156 char t_iobc; /* input character */
157 #define TCPOOB_HAVEDATA 0x01
158 #define TCPOOB_HADDATA 0x02
159 int t_softerror; /* possible error not yet reported */
161 /* RFC 1323 variables */
162 u_char snd_scale; /* window scaling for send window */
163 u_char rcv_scale; /* window scaling for recv window */
164 u_char request_r_scale; /* pending window scaling */
165 u_char requested_s_scale;
166 u_long ts_recent; /* timestamp echo data */
167 u_long ts_recent_age; /* when last updated */
168 tcp_seq last_ack_sent;
169 /* RFC 1644 variables */
170 tcp_cc cc_send; /* send connection count */
171 tcp_cc cc_recv; /* receive connection count */
172 u_long t_duration; /* connection duration */
175 caddr_t t_tuba_pcb; /* next level down pcb for TCP over z */
177 u_long t_rttupdated; /* number of times rtt sampled */
179 #if defined(TCP_FACK) || defined(TCP_AUTO)
180 u_short t_alt_flags; /* experimental flags */
181 #define TAF_SACK_SEEN 0x0001 /* other side can send SACKs */
182 #define TAF_RECOVERY 0x0002 /* We are recovering from a lost segment */
183 #define TAF_RATEHALF 0x0004 /* We are reducing the window during recovery */
184 #define TAF_REPAIRED 0x0008 /* We have retransmitted something */
185 #define TAF_WHOLD 0x0010 /* we are in the window hold state */
186 #define TAF_TOGGLE 0x0020 /* divide by 2 toggle */
187 #define TAF_AUTO_OFF 0x0040 /* autotuning is forced off */
190 #if defined(TCP_SACK) || defined(TCP_FACK)
191 #define SACK_LIST_LEN 10
193 /* Needed for SACK and FACK: */
194 struct sackblock sack_list[SACK_LIST_LEN]; /* Hack, keep 5 most recent SACKs */
195 struct scoreboard scrb;
204 #define TCP_FACK_REXMTTHRESH 3 /* number of SACKs before retransmitting a block */
205 u_long lothresh; /* recovery window floor */
206 u_long hithresh; /* maximum window following recovery */
211 * Structure to hold TCP options that are only used during segment
212 * processing (in tcp_input), but not held in the tcpcb.
213 * It's basically used to reduce the number of parameters
217 u_long to_flag; /* which options are present */
218 #define TOF_TS 0x0001 /* timestamp */
219 #define TOF_CC 0x0002 /* CC and CCnew are exclusive */
220 #define TOF_CCNEW 0x0004
221 #define TOF_CCECHO 0x0008
224 tcp_cc to_cc; /* holds CC or CCnew */
229 * The TAO cache entry which is stored in the protocol family specific
230 * portion of the route metrics.
233 tcp_cc tao_cc; /* latest CC in valid SYN */
234 tcp_cc tao_ccsent; /* latest CC sent to peer */
235 u_short tao_mssopt; /* peer's cached MSS */
237 u_short tao_flags; /* cache status flags */
238 #define TAOF_DONT 0x0001 /* peer doesn't understand rfc1644 */
239 #define TAOF_OK 0x0002 /* peer does understand rfc1644 */
240 #define TAOF_UNDEF 0 /* we don't know yet */
243 #define rmx_taop(r) ((struct rmxp_tao *)(r).rmx_filler)
245 #define intotcpcb(ip) ((struct tcpcb *)(ip)->inp_ppcb)
246 #define sototcpcb(so) (intotcpcb(sotoinpcb(so)))
249 * The smoothed round-trip time and estimated variance
250 * are stored as fixed point numbers scaled by the values below.
251 * For convenience, these scales are also used in smoothing the average
252 * (smoothed = (1/scale)sample + ((scale-1)/scale)smoothed).
253 * With these scales, srtt has 3 bits to the right of the binary point,
254 * and thus an "ALPHA" of 0.875. rttvar has 2 bits to the right of the
255 * binary point, and is smoothed with an ALPHA of 0.75.
257 #define TCP_RTT_SCALE 32 /* multiplier for srtt; 3 bits frac. */
258 #define TCP_RTT_SHIFT 5 /* shift for srtt; 3 bits frac. */
259 #define TCP_RTTVAR_SCALE 16 /* multiplier for rttvar; 2 bits */
260 #define TCP_RTTVAR_SHIFT 4 /* shift for rttvar; 2 bits */
261 #define TCP_DELTA_SHIFT 2 /* see tcp_input.c */
264 * The initial retransmission should happen at rtt + 4 * rttvar.
265 * Because of the way we do the smoothing, srtt and rttvar
266 * will each average +1/2 tick of bias. When we compute
267 * the retransmit timer, we want 1/2 tick of rounding and
268 * 1 extra tick because of +-1/2 tick uncertainty in the
269 * firing of the timer. The bias will give us exactly the
270 * 1.5 tick we need. But, because the bias is
271 * statistical, we have to test that we don't drop below
272 * the minimum feasible timer (which is 2 ticks).
273 * This version of the macro adapted from a paper by Lawrence
274 * Brakmo and Larry Peterson which outlines a problem caused
275 * by insufficient precision in the original implementation,
276 * which results in inappropriately large RTO values for very
279 #define TCP_REXMTVAL(tp) \
280 ((((tp)->t_srtt >> (TCP_RTT_SHIFT - TCP_DELTA_SHIFT)) \
281 + (tp)->t_rttvar) >> TCP_DELTA_SHIFT)
284 * We want to avoid doing m_pullup on incoming packets but that
285 * means avoiding dtom on the tcp reassembly code. That in turn means
286 * keeping an mbuf pointer in the reassembly queue (since we might
287 * have a cluster). As a quick hack, the source & destination
288 * port numbers (which are no longer needed once we've located the
289 * tcpcb) are overlayed with an mbuf pointer.
291 #define REASS_MBUF(ti) (*(struct mbuf **)&((ti)->ti_t))
295 * Many of these should be kept per connection,
296 * but that's inconvenient at the moment.
299 u_long tcps_connattempt; /* connections initiated */
300 u_long tcps_accepts; /* connections accepted */
301 u_long tcps_connects; /* connections established */
302 u_long tcps_drops; /* connections dropped */
303 u_long tcps_conndrops; /* embryonic connections dropped */
304 u_long tcps_closed; /* conn. closed (includes drops) */
305 u_long tcps_segstimed; /* segs where we tried to get rtt */
306 u_long tcps_rttupdated; /* times we succeeded */
307 u_long tcps_delack; /* delayed acks sent */
308 u_long tcps_timeoutdrop; /* conn. dropped in rxmt timeout */
309 u_long tcps_rexmttimeo; /* retransmit timeouts */
310 u_long tcps_persisttimeo; /* persist timeouts */
311 u_long tcps_keeptimeo; /* keepalive timeouts */
312 u_long tcps_keepprobe; /* keepalive probes sent */
313 u_long tcps_keepdrops; /* connections dropped in keepalive */
315 u_long tcps_sndtotal; /* total packets sent */
316 u_long tcps_sndpack; /* data packets sent */
317 u_long tcps_sndbyte; /* data bytes sent */
318 u_long tcps_sndrexmitpack; /* data packets retransmitted */
319 u_long tcps_sndrexmitbyte; /* data bytes retransmitted */
320 u_long tcps_sndacks; /* ack-only packets sent */
321 u_long tcps_sndprobe; /* window probes sent */
322 u_long tcps_sndurg; /* packets sent with URG only */
323 u_long tcps_sndwinup; /* window update-only packets sent */
324 u_long tcps_sndctrl; /* control (SYN|FIN|RST) packets sent */
326 u_long tcps_rcvtotal; /* total packets received */
327 u_long tcps_rcvpack; /* packets received in sequence */
328 u_long tcps_rcvbyte; /* bytes received in sequence */
329 u_long tcps_rcvbadsum; /* packets received with ccksum errs */
330 u_long tcps_rcvbadoff; /* packets received with bad offset */
331 u_long tcps_rcvshort; /* packets received too short */
332 u_long tcps_rcvduppack; /* duplicate-only packets received */
333 u_long tcps_rcvdupbyte; /* duplicate-only bytes received */
334 u_long tcps_rcvpartduppack; /* packets with some duplicate data */
335 u_long tcps_rcvpartdupbyte; /* dup. bytes in part-dup. packets */
336 u_long tcps_rcvoopack; /* out-of-order packets received */
337 u_long tcps_rcvoobyte; /* out-of-order bytes received */
338 u_long tcps_rcvpackafterwin; /* packets with data after window */
339 u_long tcps_rcvbyteafterwin; /* bytes rcvd after window */
340 u_long tcps_rcvafterclose; /* packets rcvd after "close" */
341 u_long tcps_rcvwinprobe; /* rcvd window probe packets */
342 u_long tcps_rcvdupack; /* rcvd duplicate acks */
343 u_long tcps_rcvacktoomuch; /* rcvd acks for unsent data */
344 u_long tcps_rcvackpack; /* rcvd ack packets */
345 u_long tcps_rcvackbyte; /* bytes acked by rcvd acks */
346 u_long tcps_rcvwinupd; /* rcvd window update packets */
347 u_long tcps_pawsdrop; /* segments dropped due to PAWS */
348 u_long tcps_predack; /* times hdr predict ok for acks */
349 u_long tcps_preddat; /* times hdr predict ok for data pkts */
350 u_long tcps_pcbcachemiss;
351 u_long tcps_cachedrtt; /* times cached RTT in route updated */
352 u_long tcps_cachedrttvar; /* times cached rttvar updated */
353 u_long tcps_cachedssthresh; /* times cached ssthresh updated */
354 u_long tcps_usedrtt; /* times RTT initialized from route */
355 u_long tcps_usedrttvar; /* times RTTVAR initialized from rt */
356 u_long tcps_usedssthresh; /* times ssthresh initialized from rt*/
357 u_long tcps_persistdrop; /* timeout in persist state */
358 u_long tcps_badsyn; /* bogus SYN, e.g. premature ACK */
359 u_long tcps_mturesent; /* resends due to MTU discovery */
360 u_long tcps_listendrop; /* listen queue overflows */
363 /* NOTE: This may break some programs that rely on this structure
364 * being fixed length.
366 u_long tcps_fack_recovery; /* fack: recovery episodes */
367 u_long tcps_fack_sndpack; /* fack: data packets sent */
368 u_long tcps_fack_sndbyte; /* fack: data bytes sent */
369 u_long tcps_fack_sndrexmitpack;/* fack: data packets retransmitted */
370 u_long tcps_fack_sndrexmitbyte;/* fack: data bytes retransmitted */
375 * Names for TCP sysctl objects
377 #define TCPCTL_DO_RFC1323 1 /* use RFC-1323 extensions */
378 #define TCPCTL_DO_RFC1644 2 /* use RFC-1644 extensions */
379 #define TCPCTL_MSSDFLT 3 /* MSS default */
380 #define TCPCTL_STATS 4 /* statistics (read-only) */
381 #define TCPCTL_RTTDFLT 5 /* default RTT estimate */
382 #define TCPCTL_KEEPIDLE 6 /* keepalive idle timer */
383 #define TCPCTL_KEEPINTVL 7 /* interval to send keepalives */
384 #define TCPCTL_SENDSPACE 8 /* send buffer space */
385 #define TCPCTL_RECVSPACE 9 /* receive buffer space */
386 #define TCPCTL_KEEPINIT 10 /* receive buffer space */
387 #define TCPCTL_MAXID 11
389 #define TCPCTL_NAMES { \
391 { "rfc1323", CTLTYPE_INT }, \
392 { "rfc1644", CTLTYPE_INT }, \
393 { "mssdflt", CTLTYPE_INT }, \
394 { "stats", CTLTYPE_STRUCT }, \
395 { "rttdflt", CTLTYPE_INT }, \
396 { "keepidle", CTLTYPE_INT }, \
397 { "keepintvl", CTLTYPE_INT }, \
398 { "sendspace", CTLTYPE_INT }, \
399 { "recvspace", CTLTYPE_INT }, \
400 { "keepinit", CTLTYPE_INT }, \