1.\" Copyright (c) 1983, 1991, 1993 2.\" The Regents of the University of California. All rights reserved. 3.\" 4.\" Redistribution and use in source and binary forms, with or without 5.\" modification, are permitted provided that the following conditions 6.\" are met: 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. 19.\" 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 30.\" SUCH DAMAGE. 31.\" 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 1.9 2008/10/17 11:30:24 swildner Exp $ 35.\" 36.Dd February 14, 1995 37.Dt TCP 4 38.Os 39.Sh NAME 40.Nm tcp 41.Nd Internet Transmission Control Protocol 42.Sh SYNOPSIS 43.In sys/types.h 44.In sys/socket.h 45.In netinet/in.h 46.Ft int 47.Fn socket AF_INET SOCK_STREAM 0 48.Sh DESCRIPTION 49The 50.Tn TCP 51protocol provides reliable, flow-controlled, two-way 52transmission of data. It is a byte-stream protocol used to 53support the 54.Dv SOCK_STREAM 55abstraction. TCP uses the standard 56Internet address format and, in addition, provides a per-host 57collection of 58.Dq port addresses . 59Thus, each address is composed 60of an Internet address specifying the host and network, with 61a specific 62.Tn TCP 63port on the host identifying the peer entity. 64.Pp 65Sockets utilizing the tcp protocol are either 66.Dq active 67or 68.Dq passive . 69Active sockets initiate connections to passive 70sockets. By default 71.Tn TCP 72sockets are created active; to create a 73passive socket the 74.Xr listen 2 75system call must be used 76after binding the socket with the 77.Xr bind 2 78system call. Only 79passive sockets may use the 80.Xr accept 2 81call to accept incoming connections. Only active sockets may 82use the 83.Xr connect 2 84call to initiate connections. 85.Pp 86Passive sockets may 87.Dq underspecify 88their location to match 89incoming connection requests from multiple networks. This 90technique, termed 91.Dq wildcard addressing , 92allows a single 93server to provide service to clients on multiple networks. 94To create a socket which listens on all networks, the Internet 95address 96.Dv INADDR_ANY 97must be bound. The 98.Tn TCP 99port may still be specified 100at this time; if the port is not specified the system will assign one. 101Once a connection has been established the socket's address is 102fixed by the peer entity's location. The address assigned the 103socket is the address associated with the network interface 104through which packets are being transmitted and received. Normally 105this address corresponds to the peer entity's network. 106.Pp 107.Tn TCP 108supports a number of socket options which can be set with 109.Xr setsockopt 2 110and tested with 111.Xr getsockopt 2 : 112.Bl -tag -width TCP_NODELAYx 113.It Dv TCP_NODELAY 114Under most circumstances, 115.Tn TCP 116sends data when it is presented; 117when outstanding data has not yet been acknowledged, it gathers 118small amounts of output to be sent in a single packet once 119an acknowledgement is received. 120For a small number of clients, such as window systems 121that send a stream of mouse events which receive no replies, 122this packetization may cause significant delays. 123The boolean option 124.Dv TCP_NODELAY 125defeats this algorithm. 126.It Dv TCP_MAXSEG 127By default, a sender\- and receiver-TCP 128will negotiate among themselves to determine the maximum segment size 129to be used for each connection. The 130.Dv TCP_MAXSEG 131option allows the user to determine the result of this negotiation, 132and to reduce it if desired. 133.It Dv TCP_NOOPT 134.Tn TCP 135usually sends a number of options in each packet, corresponding to 136various 137.Tn TCP 138extensions which are provided in this implementation. The boolean 139option 140.Dv TCP_NOOPT 141is provided to disable 142.Tn TCP 143option use on a per-connection basis. 144.It Dv TCP_NOPUSH 145By convention, the sender-TCP 146will set the 147.Dq push 148bit and begin transmission immediately (if permitted) at the end of 149every user call to 150.Xr write 2 151or 152.Xr writev 2 . 153When the 154.Dv TCP_NOPUSH 155option is set to a non-zero value, 156.Tn TCP 157will delay sending any data at all until either the socket is closed, 158or the internal send buffer is filled. 159.El 160.Pp 161The option level for the 162.Xr setsockopt 2 163call is the protocol number for 164.Tn TCP , 165available from 166.Xr getprotobyname 3 , 167or 168.Dv IPPROTO_TCP . 169All options are declared in 170.In netinet/tcp.h . 171.Pp 172Options at the 173.Tn IP 174transport level may be used with 175.Tn TCP ; 176see 177.Xr ip 4 . 178Incoming connection requests that are source-routed are noted, 179and the reverse source route is used in responding. 180.Sh MIB VARIABLES 181The 182.Nm 183protocol implements a number of variables in the 184.Li net.inet 185branch of the 186.Xr sysctl 3 187MIB. 188.Bl -tag -width TCPCTL_DO_RFC1644 189.It Dv TCPCTL_DO_RFC1323 190.Pq tcp.rfc1323 191Implement the window scaling and timestamp options of RFC 1323 192(default true). 193.It Dv TCPCTL_MSSDFLT 194.Pq tcp.mssdflt 195The default value used for the maximum segment size 196.Pq Dq MSS 197when no advice to the contrary is received from MSS negotiation. 198.It Dv TCPCTL_SENDSPACE 199.Pq tcp.sendspace 200Maximum TCP send window. 201.It Dv TCPCTL_RECVSPACE 202.Pq tcp.recvspace 203Maximum TCP receive window. 204.It tcp.log_in_vain 205Log any connection attempts to ports where there is not a socket 206accepting connections. 207The value of 1 limits the logging to SYN (connection establishment) 208packets only. 209That of 2 results in any TCP packets to closed ports being logged. 210Any value unlisted above disables the logging 211(default is 0, i.e., the logging is disabled). 212.It tcp.msl 213The Maximum Segment Lifetime for a packet. 214.It tcp.keepinit 215Timeout for new, non-established TCP connections. 216.It tcp.keepidle 217Amount of time the connection should be idle before keepalive 218probes (if enabled) are sent. 219.It tcp.keepintvl 220The interval between keepalive probes sent to remote machines. 221After 222.Dv TCPTV_KEEPCNT 223(default 8) probes are sent, with no response, the connection is dropped. 224.It tcp.always_keepalive 225Assume that 226.Dv SO_KEEPALIVE 227is set on all 228.Tn TCP 229connections, the kernel will 230periodically send a packet to the remote host to verify the connection 231is still up. 232.It tcp.icmp_may_rst 233Certain 234.Tn ICMP 235unreachable messages may abort connections in 236.Tn SYN-SENT 237state. 238.It tcp.do_tcpdrain 239Flush packets in the 240.Tn TCP 241reassembly queue if the system is low on mbufs. 242.It tcp.blackhole 243If enabled, disable sending of RST when a connection is attempted 244to a port where there is not a socket accepting connections. 245See 246.Xr blackhole 4 . 247.It tcp.delayed_ack 248Delay ACK to try and piggyback it onto a data packet. 249.It tcp.delacktime 250Maximum amount of time before a delayed ACK is sent. 251.It tcp.newreno 252Enable TCP NewReno Fast Recovery algorithm, 253as described in RFC 2582. 254.It tcp.path_mtu_discovery 255Enables Path MTU Discovery. PMTU Discovery is helpful for avoiding 256IP fragmentation when tranferring lots of data to the same client. 257For web servers, where most of the connections are short and to 258different clients, PMTU Discovery actually hurts performance due 259to unnecessary retransmissions. Turn this on only if most of your 260TCP connections are long transfers or are repeatedly to the same 261set of clients. 262.It tcp.tcbhashsize 263Size of the 264.Tn TCP 265control-block hashtable 266(read-only). 267This may be tuned using the kernel option 268.Dv TCBHASHSIZE 269or by setting 270.Va net.inet.tcp.tcbhashsize 271in the 272.Xr loader 8 . 273.It tcp.pcbcount 274Number of active process control blocks 275(read-only). 276.It tcp.syncookies 277Determines whether or not syn cookies should be generated for 278outbound syn-ack packets. Syn cookies are a great help during 279syn flood attacks, and are enabled by default. 280.It tcp.isn_reseed_interval 281The interval (in seconds) specifying how often the secret data used in 282RFC 1948 initial sequence number calculations should be reseeded. 283By default, this variable is set to zero, indicating that 284no reseeding will occur. 285Reseeding should not be necessary, and will break 286.Dv TIME_WAIT 287recycling for a few minutes. 288.It tcp.inet.tcp.rexmit_{min,slop} 289Adjust the retransmit timer calculation for TCP. The slop is 290typically added to the raw calculation to take into account 291occasional variances that the SRTT (smoothed round trip time) 292is unable to accommodate, while the minimum specifies an 293absolute minimum. While a number of TCP RFCs suggest a 1 294second minimum these RFCs tend to focus on streaming behavior 295and fail to deal with the fact that a 1 second minimum has severe 296detrimental effects over lossy interactive connections, such 297as a 802.11b wireless link, and over very fast but lossy 298connections for those cases not covered by the fast retransmit 299code. For this reason we suggest changing the slop to 200ms and 300setting the minimum to something out of the way, like 20ms, 301which gives you an effective minimum of 200ms (similar to Linux). 302.It tcp.inflight_enable 303Enable 304.Tn TCP 305bandwidth delay product limiting. An attempt will be made to calculate 306the bandwidth delay product for each individual TCP connection and limit 307the amount of inflight data being transmitted to avoid building up 308unnecessary packets in the network. This option is recommended if you 309are serving a lot of data over connections with high bandwidth-delay 310products, such as modems, GigE links, and fast long-haul WANs, and/or 311you have configured your machine to accommodate large TCP windows. In such 312situations, without this option, you may experience high interactive 313latencies or packet loss due to the overloading of intermediate routers 314and switches. Note that bandwidth delay product limiting only affects 315the transmit side of a TCP connection. 316.It tcp.inflight_debug 317Enable debugging for the bandwidth delay product algorithm. This may 318default to on (1) so if you enable the algorithm you should probably also 319disable debugging by setting this variable to 0. 320.It tcp.inflight_min 321This puts an lower bound on the bandwidth delay product window, in bytes. 322A value of 1024 is typically used for debugging. 6000-16000 is more typical 323in a production installation. Setting this value too low may result in 324slow ramp-up times for bursty connections. Setting this value too high 325effectively disables the algorithm. 326.It tcp.inflight_max 327This puts an upper bound on the bandwidth delay product window, in bytes. 328This value should not generally be modified but may be used to set a 329global per-connection limit on queued data, potentially allowing you to 330intentionally set a less than optimum limit to smooth data flow over a 331network while still being able to specify huge internal TCP buffers. 332.It tcp.inflight_stab 333The bandwidth delay product algorithm requires a slightly larger window 334than it otherwise calculates for stability. This parameter determines the 335extra window in maximal packets / 10. The default value of 20 represents 3362 maximal packets. Reducing this value is not recommended but you may 337come across a situation with very slow links where the ping time 338reduction of the default inflight code is not sufficient. If this case 339occurs you should first try reducing tcp.inflight_min and, if that does not 340work, reduce both tcp.inflight_min and tcp.inflight_stab, trying values of 34115, 10, or 5 for the latter. Never use a value less than 5. Reducing 342tcp.inflight_stab can lead to upwards of a 20% underutilization of the link 343as well as reducing the algorithm's ability to adapt to changing 344situations and should only be done as a last resort. 345.El 346.Sh ERRORS 347A socket operation may fail with one of the following errors returned: 348.Bl -tag -width Er 349.It Bq Er EISCONN 350when trying to establish a connection on a socket which 351already has one; 352.It Bq Er ENOBUFS 353when the system runs out of memory for 354an internal data structure; 355.It Bq Er ETIMEDOUT 356when a connection was dropped 357due to excessive retransmissions; 358.It Bq Er ECONNRESET 359when the remote peer 360forces the connection to be closed; 361.It Bq Er ECONNREFUSED 362when the remote 363peer actively refuses connection establishment (usually because 364no process is listening to the port); 365.It Bq Er EADDRINUSE 366when an attempt 367is made to create a socket with a port which has already been 368allocated; 369.It Bq Er EADDRNOTAVAIL 370when an attempt is made to create a 371socket with a network address for which no network interface 372exists. 373.It Bq Er EAFNOSUPPORT 374when an attempt is made to bind or connect a socket to a multicast 375address. 376.El 377.Sh SEE ALSO 378.Xr getsockopt 2 , 379.Xr socket 2 , 380.Xr sysctl 3 , 381.Xr blackhole 4 , 382.Xr inet 4 , 383.Xr intro 4 , 384.Xr ip 4 385.Rs 386.%A V. Jacobson 387.%A R. Braden 388.%A D. Borman 389.%T "TCP Extensions for High Performance" 390.%O RFC 1323 391.Re 392.Sh HISTORY 393The 394.Nm 395protocol appeared in 396.Bx 4.2 . 397The RFC 1323 extensions for window scaling and timestamps were added 398in 399.Bx 4.4 . 400