1/*-
2 * SPDX-License-Identifier: BSD-3-Clause
3 *
4 * Copyright (c) 1982, 1986, 1993
5 * The Regents of the University of California. All rights reserved.
6 *
7 * Redistribution and use in source and binary forms, with or without
8 * modification, are permitted provided that the following conditions
9 * are met:
10 * 1. Redistributions of source code must retain the above copyright
11 * notice, this list of conditions and the following disclaimer.
12 * 2. Redistributions in binary form must reproduce the above copyright
13 * notice, this list of conditions and the following disclaimer in the
14 * documentation and/or other materials provided with the distribution.
15 * 3. Neither the name of the University nor the names of its contributors
16 * may be used to endorse or promote products derived from this software
17 * without specific prior written permission.
18 *
19 * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
20 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
21 * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
22 * ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
23 * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
24 * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
25 * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
26 * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
27 * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
28 * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
29 * SUCH DAMAGE.
30 */
31
32#ifndef _NETINET_TCP_TIMER_H_
33#define _NETINET_TCP_TIMER_H_
34
35#ifdef _KERNEL
36
37/*
38 * The TCPT_REXMT timer is used to force retransmissions.
39 * The TCP has the TCPT_REXMT timer set whenever segments
40 * have been sent for which ACKs are expected but not yet
41 * received. If an ACK is received which advances tp->snd_una,
42 * then the retransmit timer is cleared (if there are no more
43 * outstanding segments) or reset to the base value (if there
44 * are more ACKs expected). Whenever the retransmit timer goes off,
45 * we retransmit one unacknowledged segment, and do a backoff
46 * on the retransmit timer.
47 *
48 * The TCPT_PERSIST timer is used to keep window size information
49 * flowing even if the window goes shut. If all previous transmissions
50 * have been acknowledged (so that there are no retransmissions in progress),
51 * and the window is too small to bother sending anything, then we start
52 * the TCPT_PERSIST timer. When it expires, if the window is nonzero,
53 * we go to transmit state. Otherwise, at intervals send a single byte
54 * into the peer's window to force him to update our window information.
55 * We do this at most as often as TCPT_PERSMIN time intervals,
56 * but no more frequently than the current estimate of round-trip
57 * packet time. The TCPT_PERSIST timer is cleared whenever we receive
58 * a window update from the peer.
59 *
60 * The TCPT_KEEP timer is used to keep connections alive. If an
61 * connection is idle (no segments received) for TCPTV_KEEP_INIT amount of time,
62 * but not yet established, then we drop the connection. Once the connection
63 * is established, if the connection is idle for TCPTV_KEEP_IDLE time
64 * (and keepalives have been enabled on the socket), we begin to probe
65 * the connection. We force the peer to send us a segment by sending:
66 * <SEQ=SND.UNA-1><ACK=RCV.NXT><CTL=ACK>
67 * This segment is (deliberately) outside the window, and should elicit
68 * an ack segment in response from the peer. If, despite the TCPT_KEEP
69 * initiated segments we cannot elicit a response from a peer in TCPT_MAXIDLE
70 * amount of time probing, then we drop the connection.
71 */
72
73/*
74 * Time constants.
75 */
76#define TCPTV_MSL MSEC_2_TICKS(30000) /* max seg lifetime (hah!) */
77#define TCPTV_MSL_LOCAL MSEC_2_TICKS(10) /* max seg lifetime for local comm */
78#define TCPTV_SRTTBASE 0 /* base roundtrip time;
79 if 0, no idea yet */
80#define TCPTV_RTOBASE MSEC_2_TICKS(1000) /* assumed RTO if no info */
81
82#define TCPTV_PERSMIN MSEC_2_TICKS(5000) /* minimum persist interval */
83#define TCPTV_PERSMAX MSEC_2_TICKS(60000) /* maximum persist interval */
84
85#define TCPTV_KEEP_INIT MSEC_2_TICKS(75000) /* initial connect keepalive */
86#define TCPTV_KEEP_IDLE MSEC_2_TICKS(120*60*1000) /* dflt time before probing */
87#define TCPTV_KEEPINTVL MSEC_2_TICKS(75000) /* default probe interval */
88#define TCPTV_KEEPCNT 8 /* max probes before drop */
89#define TCPTV_MAXUNACKTIME 0 /* max time without making progress */
90
91#define TCPTV_FINWAIT2_TIMEOUT MSEC_2_TICKS(60000) /* FIN_WAIT_2 timeout if no receiver */
92
93/*
94 * Minimum retransmit timer is 3 ticks, for algorithmic stability.
95 * TCPT_RANGESET() will add another TCPTV_CPU_VAR to deal with
96 * the expected worst-case processing variances by the kernels
97 * representing the end points. Such variances do not always show
98 * up in the srtt because the timestamp is often calculated at
99 * the interface rather then at the TCP layer. This value is
100 * typically 50ms. However, it is also possible that delayed
101 * acks (typically 100ms) could create issues so we set the slop
102 * to 200ms to try to cover it. Note that, properly speaking,
103 * delayed-acks should not create a major issue for interactive
104 * environments which 'P'ush the last segment, at least as
105 * long as implementations do the required 'at least one ack
106 * for every two packets' for the non-interactive streaming case.
107 * (maybe the RTO calculation should use 2*RTT instead of RTT
108 * to handle the ack-every-other-packet case).
109 *
110 * The prior minimum of 1*hz (1 second) badly breaks throughput on any
111 * networks faster then a modem that has minor (e.g. 1%) packet loss.
112 */
113#define TCPTV_MIN MSEC_2_TICKS(30) /* minimum allowable value */
114#define TCPTV_CPU_VAR MSEC_2_TICKS(200) /* cpu variance allowed (200ms) */
115#define TCPTV_REXMTMAX MSEC_2_TICKS(64000) /* max allowable REXMT value */
116
117#define TCP_MAXRXTSHIFT 12 /* maximum retransmits */
118
119#define TCPTV_DELACK MSEC_2_TICKS(40) /* 40ms timeout */
120
121/*
122 * If we exceed this number of retransmits for a single segment, we'll consider
123 * the current srtt measurement no longer valid and will recalculate from
124 * scratch starting with the next ACK.
125 */
126#define TCP_RTT_INVALIDATE (TCP_MAXRXTSHIFT / 4)
127
128/*
129 * Force a time value to be in a certain range.
130 */
131#define TCPT_RANGESET(tv, value, tvmin, tvmax) do { \
132 (tv) = (value) + tcp_rexmit_slop; \
133 if ((u_long)(tv) < (u_long)(tvmin)) \
134 (tv) = (tvmin); \
135 if ((u_long)(tv) > (u_long)(tvmax)) \
136 (tv) = (tvmax); \
137} while(0)
138
139#define TP_KEEPINIT(tp) ((tp)->t_keepinit ? (tp)->t_keepinit : tcp_keepinit)
140#define TP_KEEPIDLE(tp) ((tp)->t_keepidle ? (tp)->t_keepidle : tcp_keepidle)
141#define TP_KEEPINTVL(tp) ((tp)->t_keepintvl ? (tp)->t_keepintvl : tcp_keepintvl)
142#define TP_KEEPCNT(tp) ((tp)->t_keepcnt ? (tp)->t_keepcnt : tcp_keepcnt)
143#define TP_MAXIDLE(tp) (TP_KEEPCNT(tp) * TP_KEEPINTVL(tp))
144#define TP_MAXUNACKTIME(tp) \
145 ((tp)->t_maxunacktime ? (tp)->t_maxunacktime : tcp_maxunacktime)
146
147/*
148 * Obtain the time until the restransmit timer should fire.
149 * This macro ensures the restransmit timer fires at the earlier of the
150 * t_rxtcur value or the time the maxunacktime would be exceeded.
151 */
152#define TP_RXTCUR(tp) \
153 ((TP_MAXUNACKTIME(tp) == 0 || tp->t_acktime == 0) ? tp->t_rxtcur : \
154 max(1, min(tp->t_rxtcur, tp->t_acktime + TP_MAXUNACKTIME(tp) - ticks)))
155
156extern int tcp_persmin; /* minimum persist interval */
157extern int tcp_persmax; /* maximum persist interval */
158extern int tcp_keepinit; /* time to establish connection */
159extern int tcp_keepidle; /* time before keepalive probes begin */
160extern int tcp_keepintvl; /* time between keepalive probes */
161extern int tcp_keepcnt; /* number of keepalives */
162extern int tcp_delacktime; /* time before sending a delayed ACK */
163extern int tcp_maxunacktime; /* max time without making progress */
164extern int tcp_maxpersistidle;
165extern int tcp_rexmit_initial;
166extern int tcp_rexmit_min;
167extern int tcp_rexmit_max;
168extern int tcp_rexmit_slop;
169extern int tcp_ttl; /* time to live for TCP segs */
170extern int tcp_backoff[];
171extern int tcp_totbackoff;
172extern int tcp_rexmit_drop_options;
173
174extern int tcp_finwait2_timeout;
175extern int tcp_fast_finwait2_recycle;
176
177VNET_DECLARE(int, tcp_always_keepalive);
178#define V_tcp_always_keepalive VNET(tcp_always_keepalive)
179VNET_DECLARE(int, tcp_pmtud_blackhole_detect);
180#define V_tcp_pmtud_blackhole_detect VNET(tcp_pmtud_blackhole_detect)
181VNET_DECLARE(int, tcp_pmtud_blackhole_mss);
182#define V_tcp_pmtud_blackhole_mss VNET(tcp_pmtud_blackhole_mss)
183VNET_DECLARE(int, tcp_v6pmtud_blackhole_mss);
184#define V_tcp_v6pmtud_blackhole_mss VNET(tcp_v6pmtud_blackhole_mss)
185VNET_DECLARE(int, tcp_msl);
186#define V_tcp_msl VNET(tcp_msl)
187VNET_DECLARE(int, tcp_msl_local);
188#define V_tcp_msl_local VNET(tcp_msl_local)
189
190#endif /* _KERNEL */
191
192#endif /* !_NETINET_TCP_TIMER_H_ */