1 .\" Copyright (c) 1983, 1991, 1993
2 .\" The Regents of the University of California. All rights reserved.
4 .\" Redistribution and use in source and binary forms, with or without
5 .\" modification, are permitted provided that the following conditions
7 .\" 1. Redistributions of source code must retain the above copyright
8 .\" notice, this list of conditions and the following disclaimer.
9 .\" 2. Redistributions in binary form must reproduce the above copyright
10 .\" notice, this list of conditions and the following disclaimer in the
11 .\" documentation and/or other materials provided with the distribution.
12 .\" 3. All advertising materials mentioning features or use of this software
13 .\" must display the following acknowledgement:
14 .\" This product includes software developed by the University of
15 .\" California, Berkeley and its contributors.
16 .\" 4. Neither the name of the University nor the names of its contributors
17 .\" may be used to endorse or promote products derived from this software
18 .\" without specific prior written permission.
20 .\" THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
21 .\" ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
22 .\" IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
23 .\" ARE DISCLAIMED. IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
24 .\" FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
25 .\" DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
26 .\" OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
27 .\" HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
28 .\" LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
29 .\" OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
32 .\" From: @(#)tcp.4 8.1 (Berkeley) 6/5/93
33 .\" $FreeBSD: src/share/man/man4/tcp.4,v 1.11.2.14 2002/12/29 16:35:38 schweikh Exp $
34 .\" $DragonFly: src/share/man/man4/tcp.4,v 1.8 2007/09/14 23:47:53 swildner Exp $
41 .Nd Internet Transmission Control Protocol
47 .Fn socket AF_INET SOCK_STREAM 0
51 protocol provides reliable, flow-controlled, two-way
52 transmission of data. It is a byte-stream protocol used to
55 abstraction. TCP uses the standard
56 Internet address format and, in addition, provides a per-host
59 Thus, each address is composed
60 of an Internet address specifying the host and network, with
63 port on the host identifying the peer entity.
65 Sockets utilizing the tcp protocol are either
69 Active sockets initiate connections to passive
72 sockets are created active; to create a
75 system call must be used
76 after binding the socket with the
79 passive sockets may use the
81 call to accept incoming connections. Only active sockets may
84 call to initiate connections.
86 also supports a more datagram-like mode, called Transaction
93 their location to match
94 incoming connection requests from multiple networks. This
96 .Dq wildcard addressing ,
98 server to provide service to clients on multiple networks.
99 To create a socket which listens on all networks, the Internet
104 port may still be specified
105 at this time; if the port is not specified the system will assign one.
106 Once a connection has been established the socket's address is
107 fixed by the peer entity's location. The address assigned the
108 socket is the address associated with the network interface
109 through which packets are being transmitted and received. Normally
110 this address corresponds to the peer entity's network.
113 supports a number of socket options which can be set with
117 .Bl -tag -width TCP_NODELAYx
119 Under most circumstances,
121 sends data when it is presented;
122 when outstanding data has not yet been acknowledged, it gathers
123 small amounts of output to be sent in a single packet once
124 an acknowledgement is received.
125 For a small number of clients, such as window systems
126 that send a stream of mouse events which receive no replies,
127 this packetization may cause significant delays.
130 defeats this algorithm.
132 By default, a sender\- and receiver-TCP
133 will negotiate among themselves to determine the maximum segment size
134 to be used for each connection. The
136 option allows the user to determine the result of this negotiation,
137 and to reduce it if desired.
140 usually sends a number of options in each packet, corresponding to
143 extensions which are provided in this implementation. The boolean
146 is provided to disable
148 option use on a per-connection basis.
150 By convention, the sender-TCP
153 bit and begin transmission immediately (if permitted) at the end of
160 option is provided to allow servers to easily make use of Transaction
163 When the option is set to a non-zero value,
165 will delay sending any data at all until either the socket is closed,
166 or the internal send buffer is filled.
169 The option level for the
171 call is the protocol number for
174 .Xr getprotobyname 3 ,
177 All options are declared in
182 transport level may be used with
186 Incoming connection requests that are source-routed are noted,
187 and the reverse source route is used in responding.
191 protocol implements a number of variables in the
196 .Bl -tag -width TCPCTL_DO_RFC1644
197 .It Dv TCPCTL_DO_RFC1323
199 Implement the window scaling and timestamp options of RFC 1323
201 .It Dv TCPCTL_DO_RFC1644
203 Implement Transaction
205 as described in RFC 1644.
206 .It Dv TCPCTL_MSSDFLT
208 The default value used for the maximum segment size
210 when no advice to the contrary is received from MSS negotiation.
211 .It Dv TCPCTL_SENDSPACE
213 Maximum TCP send window.
214 .It Dv TCPCTL_RECVSPACE
216 Maximum TCP receive window.
218 Log any connection attempts to ports where there is not a socket
219 accepting connections.
220 The value of 1 limits the logging to SYN (connection establishment)
222 That of 2 results in any TCP packets to closed ports being logged.
223 Any value unlisted above disables the logging
224 (default is 0, i.e., the logging is disabled).
226 The Maximum Segment Lifetime for a packet.
228 Timeout for new, non-established TCP connections.
230 Amount of time the connection should be idle before keepalive
231 probes (if enabled) are sent.
233 The interval between keepalive probes sent to remote machines.
236 (default 8) probes are sent, with no response, the connection is dropped.
237 .It tcp.always_keepalive
242 connections, the kernel will
243 periodically send a packet to the remote host to verify the connection
248 unreachable messages may abort connections in
254 reassembly queue if the system is low on mbufs.
256 If enabled, disable sending of RST when a connection is attempted
257 to a port where there is not a socket accepting connections.
261 Delay ACK to try and piggyback it onto a data packet.
263 Maximum amount of time before a delayed ACK is sent.
265 Enable TCP NewReno Fast Recovery algorithm,
266 as described in RFC 2582.
267 .It tcp.path_mtu_discovery
268 Enables Path MTU Discovery. PMTU Discovery is helpful for avoiding
269 IP fragmentation when tranferring lots of data to the same client.
270 For web servers, where most of the connections are short and to
271 different clients, PMTU Discovery actually hurts performance due
272 to unnecessary retransmissions. Turn this on only if most of your
273 TCP connections are long transfers or are repeatedly to the same
278 control-block hashtable
280 This may be tuned using the kernel option
283 .Va net.inet.tcp.tcbhashsize
287 Number of active process control blocks
290 Determines whether or not syn cookies should be generated for
291 outbound syn-ack packets. Syn cookies are a great help during
292 syn flood attacks, and are enabled by default.
293 .It tcp.isn_reseed_interval
294 The interval (in seconds) specifying how often the secret data used in
295 RFC 1948 initial sequence number calculations should be reseeded.
296 By default, this variable is set to zero, indicating that
297 no reseeding will occur.
298 Reseeding should not be necessary, and will break
300 recycling for a few minutes.
301 .It tcp.inet.tcp.rexmit_{min,slop}
302 Adjust the retransmit timer calculation for TCP. The slop is
303 typically added to the raw calculation to take into account
304 occasional variances that the SRTT (smoothed round trip time)
305 is unable to accommodate, while the minimum specifies an
306 absolute minimum. While a number of TCP RFCs suggest a 1
307 second minimum these RFCs tend to focus on streaming behavior
308 and fail to deal with the fact that a 1 second minimum has severe
309 detrimental effects over lossy interactive connections, such
310 as a 802.11b wireless link, and over very fast but lossy
311 connections for those cases not covered by the fast retransmit
312 code. For this reason we suggest changing the slop to 200ms and
313 setting the minimum to something out of the way, like 20ms,
314 which gives you an effective minimum of 200ms (similar to Linux).
315 .It tcp.inflight_enable
318 bandwidth delay product limiting. An attempt will be made to calculate
319 the bandwidth delay product for each individual TCP connection and limit
320 the amount of inflight data being transmitted to avoid building up
321 unnecessary packets in the network. This option is recommended if you
322 are serving a lot of data over connections with high bandwidth-delay
323 products, such as modems, GigE links, and fast long-haul WANs, and/or
324 you have configured your machine to accommodate large TCP windows. In such
325 situations, without this option, you may experience high interactive
326 latencies or packet loss due to the overloading of intermediate routers
327 and switches. Note that bandwidth delay product limiting only affects
328 the transmit side of a TCP connection.
329 .It tcp.inflight_debug
330 Enable debugging for the bandwidth delay product algorithm. This may
331 default to on (1) so if you enable the algorithm you should probably also
332 disable debugging by setting this variable to 0.
334 This puts an lower bound on the bandwidth delay product window, in bytes.
335 A value of 1024 is typically used for debugging. 6000-16000 is more typical
336 in a production installation. Setting this value too low may result in
337 slow ramp-up times for bursty connections. Setting this value too high
338 effectively disables the algorithm.
340 This puts an upper bound on the bandwidth delay product window, in bytes.
341 This value should not generally be modified but may be used to set a
342 global per-connection limit on queued data, potentially allowing you to
343 intentionally set a less then optimum limit to smooth data flow over a
344 network while still being able to specify huge internal TCP buffers.
345 .It tcp.inflight_stab
346 The bandwidth delay product algorithm requires a slightly larger window
347 then it otherwise calculates for stability. This parameter determines the
348 extra window in maximal packets / 10. The default value of 20 represents
349 2 maximal packets. Reducing this value is not recommended but you may
350 come across a situation with very slow links where the ping time
351 reduction of the default inflight code is not sufficient. If this case
352 occurs you should first try reducing tcp.inflight_min and, if that does not
353 work, reduce both tcp.inflight_min and tcp.inflight_stab, trying values of
354 15, 10, or 5 for the latter. Never use a value less then 5. Reducing
355 tcp.inflight_stab can lead to upwards of a 20% underutilization of the link
356 as well as reducing the algorithm's ability to adapt to changing
357 situations and should only be done as a last resort.
360 A socket operation may fail with one of the following errors returned:
363 when trying to establish a connection on a socket which
366 when the system runs out of memory for
367 an internal data structure;
369 when a connection was dropped
370 due to excessive retransmissions;
373 forces the connection to be closed;
374 .It Bq Er ECONNREFUSED
376 peer actively refuses connection establishment (usually because
377 no process is listening to the port);
380 is made to create a socket with a port which has already been
382 .It Bq Er EADDRNOTAVAIL
383 when an attempt is made to create a
384 socket with a network address for which no network interface
386 .It Bq Er EAFNOSUPPORT
387 when an attempt is made to bind or connect a socket to a multicast
403 .%T "TCP Extensions for High Performance"
408 .%T "T/TCP \- TCP Extensions for Transactions"
416 The RFC 1323 extensions for window scaling and timestamps were added