xref: /freebsd/sys/netinet/tcp_timer.h (revision 47dd1d1b619cc035b82b49a91a25544309ff95ae)
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  *	@(#)tcp_timer.h	8.1 (Berkeley) 6/10/93
32  * $FreeBSD$
33  */
34 
35 #ifndef _NETINET_TCP_TIMER_H_
36 #define _NETINET_TCP_TIMER_H_
37 
38 /*
39  * The TCPT_REXMT timer is used to force retransmissions.
40  * The TCP has the TCPT_REXMT timer set whenever segments
41  * have been sent for which ACKs are expected but not yet
42  * received.  If an ACK is received which advances tp->snd_una,
43  * then the retransmit timer is cleared (if there are no more
44  * outstanding segments) or reset to the base value (if there
45  * are more ACKs expected).  Whenever the retransmit timer goes off,
46  * we retransmit one unacknowledged segment, and do a backoff
47  * on the retransmit timer.
48  *
49  * The TCPT_PERSIST timer is used to keep window size information
50  * flowing even if the window goes shut.  If all previous transmissions
51  * have been acknowledged (so that there are no retransmissions in progress),
52  * and the window is too small to bother sending anything, then we start
53  * the TCPT_PERSIST timer.  When it expires, if the window is nonzero,
54  * we go to transmit state.  Otherwise, at intervals send a single byte
55  * into the peer's window to force him to update our window information.
56  * We do this at most as often as TCPT_PERSMIN time intervals,
57  * but no more frequently than the current estimate of round-trip
58  * packet time.  The TCPT_PERSIST timer is cleared whenever we receive
59  * a window update from the peer.
60  *
61  * The TCPT_KEEP timer is used to keep connections alive.  If an
62  * connection is idle (no segments received) for TCPTV_KEEP_INIT amount of time,
63  * but not yet established, then we drop the connection.  Once the connection
64  * is established, if the connection is idle for TCPTV_KEEP_IDLE time
65  * (and keepalives have been enabled on the socket), we begin to probe
66  * the connection.  We force the peer to send us a segment by sending:
67  *	<SEQ=SND.UNA-1><ACK=RCV.NXT><CTL=ACK>
68  * This segment is (deliberately) outside the window, and should elicit
69  * an ack segment in response from the peer.  If, despite the TCPT_KEEP
70  * initiated segments we cannot elicit a response from a peer in TCPT_MAXIDLE
71  * amount of time probing, then we drop the connection.
72  */
73 
74 /*
75  * Time constants.
76  */
77 #define	TCPTV_MSL	( 30*hz)		/* max seg lifetime (hah!) */
78 #define	TCPTV_SRTTBASE	0			/* base roundtrip time;
79 						   if 0, no idea yet */
80 #define	TCPTV_RTOBASE	(  3*hz)		/* assumed RTO if no info */
81 
82 #define	TCPTV_PERSMIN	(  5*hz)		/* minimum persist interval */
83 #define	TCPTV_PERSMAX	( 60*hz)		/* maximum persist interval */
84 
85 #define	TCPTV_KEEP_INIT	( 75*hz)		/* initial connect keepalive */
86 #define	TCPTV_KEEP_IDLE	(120*60*hz)		/* dflt time before probing */
87 #define	TCPTV_KEEPINTVL	( 75*hz)		/* default probe interval */
88 #define	TCPTV_KEEPCNT	8			/* max probes before drop */
89 
90 #define TCPTV_FINWAIT2_TIMEOUT (60*hz)         /* FIN_WAIT_2 timeout if no receiver */
91 
92 /*
93  * Minimum retransmit timer is 3 ticks, for algorithmic stability.
94  * TCPT_RANGESET() will add another TCPTV_CPU_VAR to deal with
95  * the expected worst-case processing variances by the kernels
96  * representing the end points.  Such variances do not always show
97  * up in the srtt because the timestamp is often calculated at
98  * the interface rather then at the TCP layer.  This value is
99  * typically 50ms.  However, it is also possible that delayed
100  * acks (typically 100ms) could create issues so we set the slop
101  * to 200ms to try to cover it.  Note that, properly speaking,
102  * delayed-acks should not create a major issue for interactive
103  * environments which 'P'ush the last segment, at least as
104  * long as implementations do the required 'at least one ack
105  * for every two packets' for the non-interactive streaming case.
106  * (maybe the RTO calculation should use 2*RTT instead of RTT
107  * to handle the ack-every-other-packet case).
108  *
109  * The prior minimum of 1*hz (1 second) badly breaks throughput on any
110  * networks faster then a modem that has minor (e.g. 1%) packet loss.
111  */
112 #define	TCPTV_MIN	( hz/33 )		/* minimum allowable value */
113 #define TCPTV_CPU_VAR	( hz/5 )		/* cpu variance allowed (200ms) */
114 #define	TCPTV_REXMTMAX	( 64*hz)		/* max allowable REXMT value */
115 
116 #define TCPTV_TWTRUNC	8			/* RTO factor to truncate TW */
117 
118 #define	TCP_LINGERTIME	120			/* linger at most 2 minutes */
119 
120 #define	TCP_MAXRXTSHIFT	12			/* maximum retransmits */
121 
122 #define	TCPTV_DELACK	( hz/10 )		/* 100ms timeout */
123 
124 /*
125  * If we exceed this number of retransmits for a single segment, we'll consider
126  * the current srtt measurement no longer valid and will recalculate from
127  * scratch starting with the next ACK.
128  */
129 #define TCP_RTT_INVALIDATE (TCP_MAXRXTSHIFT / 4)
130 
131 #ifdef	TCPTIMERS
132 static const char *tcptimers[] =
133     { "REXMT", "PERSIST", "KEEP", "2MSL", "DELACK" };
134 #endif
135 
136 /*
137  * Force a time value to be in a certain range.
138  */
139 #define	TCPT_RANGESET(tv, value, tvmin, tvmax) do { \
140 	(tv) = (value) + tcp_rexmit_slop; \
141 	if ((u_long)(tv) < (u_long)(tvmin)) \
142 		(tv) = (tvmin); \
143 	if ((u_long)(tv) > (u_long)(tvmax)) \
144 		(tv) = (tvmax); \
145 } while(0)
146 
147 #ifdef _KERNEL
148 
149 struct xtcp_timer;
150 
151 struct tcp_timer {
152 	struct	callout tt_rexmt;	/* retransmit timer */
153 	struct	callout tt_persist;	/* retransmit persistence */
154 	struct	callout tt_keep;	/* keepalive */
155 	struct	callout tt_2msl;	/* 2*msl TIME_WAIT timer */
156 	struct	callout tt_delack;	/* delayed ACK timer */
157 	uint32_t	tt_flags;	/* Timers flags */
158 	uint32_t	tt_draincnt;	/* Count being drained */
159 };
160 
161 /*
162  * Flags for the tt_flags field.
163  */
164 #define TT_DELACK	0x0001
165 #define TT_REXMT	0x0002
166 #define TT_PERSIST	0x0004
167 #define TT_KEEP		0x0008
168 #define TT_2MSL		0x0010
169 #define TT_MASK		(TT_DELACK|TT_REXMT|TT_PERSIST|TT_KEEP|TT_2MSL)
170 
171 #define TT_DELACK_RST	0x0100
172 #define TT_REXMT_RST	0x0200
173 #define TT_PERSIST_RST	0x0400
174 #define TT_KEEP_RST	0x0800
175 #define TT_2MSL_RST	0x1000
176 
177 #define TT_STOPPED	0x00010000
178 
179 #define	TP_KEEPINIT(tp)	((tp)->t_keepinit ? (tp)->t_keepinit : tcp_keepinit)
180 #define	TP_KEEPIDLE(tp)	((tp)->t_keepidle ? (tp)->t_keepidle : tcp_keepidle)
181 #define	TP_KEEPINTVL(tp) ((tp)->t_keepintvl ? (tp)->t_keepintvl : tcp_keepintvl)
182 #define	TP_KEEPCNT(tp)	((tp)->t_keepcnt ? (tp)->t_keepcnt : tcp_keepcnt)
183 #define	TP_MAXIDLE(tp)	(TP_KEEPCNT(tp) * TP_KEEPINTVL(tp))
184 
185 extern int tcp_persmin;			/* minimum persist interval */
186 extern int tcp_persmax;			/* maximum persist interval */
187 extern int tcp_keepinit;		/* time to establish connection */
188 extern int tcp_keepidle;		/* time before keepalive probes begin */
189 extern int tcp_keepintvl;		/* time between keepalive probes */
190 extern int tcp_keepcnt;			/* number of keepalives */
191 extern int tcp_delacktime;		/* time before sending a delayed ACK */
192 extern int tcp_maxpersistidle;
193 extern int tcp_rexmit_min;
194 extern int tcp_rexmit_slop;
195 extern int tcp_msl;
196 extern int tcp_ttl;			/* time to live for TCP segs */
197 extern int tcp_backoff[];
198 extern int tcp_syn_backoff[];
199 
200 extern int tcp_always_keepalive;
201 extern int tcp_finwait2_timeout;
202 extern int tcp_fast_finwait2_recycle;
203 
204 VNET_DECLARE(int, tcp_pmtud_blackhole_detect);
205 #define V_tcp_pmtud_blackhole_detect	VNET(tcp_pmtud_blackhole_detect)
206 VNET_DECLARE(int, tcp_pmtud_blackhole_mss);
207 #define	V_tcp_pmtud_blackhole_mss	VNET(tcp_pmtud_blackhole_mss)
208 VNET_DECLARE(int, tcp_v6pmtud_blackhole_mss);
209 #define V_tcp_v6pmtud_blackhole_mss	VNET(tcp_v6pmtud_blackhole_mss)
210 
211 int tcp_inpinfo_lock_add(struct inpcb *inp);
212 void tcp_inpinfo_lock_del(struct inpcb *inp, struct tcpcb *tp);
213 
214 void	tcp_timer_init(void);
215 void	tcp_timer_2msl(void *xtp);
216 void	tcp_timer_discard(void *);
217 struct tcptw *
218 	tcp_tw_2msl_scan(int reuse);	/* XXX temporary? */
219 void	tcp_timer_keep(void *xtp);
220 void	tcp_timer_persist(void *xtp);
221 void	tcp_timer_rexmt(void *xtp);
222 void	tcp_timer_delack(void *xtp);
223 
224 #endif /* _KERNEL */
225 
226 #endif /* !_NETINET_TCP_TIMER_H_ */
227