TCPSection: Linux Programmer's Manual (7)
Index Return to Main Contents
NAMEtcp - TCP protocol
tcp_socket = socket(PF_INET, SOCK_STREAM, 0);
DESCRIPTIONThis is an implementation of the TCP protocol defined in RFC 793, RFC 1122 and RFC 2001 with the NewReno and SACK extensions. It provides a reliable, stream-oriented, full-duplex connection between two sockets on top of ip(7), for both v4 and v6 versions. TCP guarantees that the data arrives in order and retransmits lost packets. It generates and checks a per-packet checksum to catch transmission errors. TCP does not preserve record boundaries.
A newly created TCP socket has no remote or local address and is not fully specified. To create an outgoing TCP connection use connect(2) to establish a connection to another TCP socket. To receive new incoming connections, first bind(2) the socket to a local address and port and then call listen(2) to put the socket into the listening state. After that a new socket for each incoming connection can be accepted using accept(2). A socket which has had accept(2) or connect(2) successfully called on it is fully specified and may transmit data. Data cannot be transmitted on listening or not yet connected sockets.
Linux supports RFC 1323 TCP high performance extensions. These include Protection Against Wrapped Sequence Numbers (PAWS), Window Scaling and Timestamps. Window scaling allows the use of large (> 64K) TCP windows in order to support links with high latency or bandwidth. To make use of them, the send and receive buffer sizes must be increased. They can be set globally with the net.ipv4.tcp_wmem and net.ipv4.tcp_rmem sysctl variables, or on individual sockets by using the SO_SNDBUF and SO_RCVBUF socket options with the setsockopt(2) call.
The maximum sizes for socket buffers declared via the SO_SNDBUF and SO_RCVBUF mechanisms are limited by the global net.core.rmem_max and net.core.wmem_max sysctls. Note that TCP actually allocates twice the size of the buffer requested in the setsockopt(2) call, and so a succeeding getsockopt(2) call will not return the same size of buffer as requested in the setsockopt(2) call. TCP uses the extra space for administrative purposes and internal kernel structures, and the sysctl variables reflect the larger sizes compared to the actual TCP windows. On individual connections, the socket buffer size must be set prior to the listen(2) or connect(2) calls in order to have it take effect. See socket(7) for more information.
TCP supports urgent data. Urgent data is used to signal the receiver that some important message is part of the data stream and that it should be processed as soon as possible. To send urgent data specify the MSG_OOB option to send(2). When urgent data is received, the kernel sends a SIGURG signal to the process or process group that has been set as the socket "owner" using the SIOCSPGRP or FIOSETOWN ioctls (or the POSIX.1-2001-specified fcntl(2) F_SETOWN operation). When the SO_OOBINLINE socket option is enabled, urgent data is put into the normal data stream (a program can test for its location using the SIOCATMARK ioctl described below), otherwise it can be only received when the MSG_OOB flag is set for recv(2) or recvmsg(2).
Linux 2.4 introduced a number of changes for improved throughput and scaling, as well as enhanced functionality. Some of these features include support for zero-copy sendfile(2), Explicit Congestion Notification, new management of TIME_WAIT sockets, keep-alive socket options and support for Duplicate SACK extensions.
Address FormatsTCP is built on top of IP (see ip(7)). The address formats defined by ip(7) apply to TCP. TCP only supports point-to-point communication; broadcasting and multicasting are not supported.
SysctlsThese variables can be accessed by the /proc/sys/net/ipv4/* files or with the sysctl(2) interface. In addition, most IP sysctls also apply to TCP; see ip(7). Variables described as Boolean take an integer value, with a non-zero value ("true") meaning that the corresponding option is enabled, and a zero value ("false") meaning that the option is disabled.
- tcp_abort_on_overflow (Boolean; default: disabled)
- Enable resetting connections if the listening service is too slow and unable to keep up and accept them. It means that if overflow occurred due to a burst, the connection will recover. Enable this option only if you are really sure that the listening daemon cannot be tuned to accept connections faster. Enabling this option can harm the clients of your server.
- tcp_adv_win_scale (integer; default: 2)
Count buffering overhead as bytes/2^tcp_adv_win_scale
(if tcp_adv_win_scale > 0) or bytes-bytes/2^(-tcp_adv_win_scale),
if it is <= 0.
The socket receive buffer space is shared between the application and kernel. TCP maintains part of the buffer as the TCP window, this is the size of the receive window advertised to the other end. The rest of the space is used as the "application" buffer, used to isolate the network from scheduling and application latencies. The tcp_adv_win_scale default value of 2 implies that the space used for the application buffer is one fourth that of the total.
- tcp_app_win (integer; default: 31)
This variable defines how many
bytes of the TCP window are reserved for buffering
A maximum of (window/2^tcp_app_win, mss) bytes in the window are reserved for the application buffer. A value of 0 implies that no amount is reserved.
- tcp_bic (Boolean; default: disabled)
- Enable BIC TCP congestion control algorithm. BIC-TCP is a sender-side only change that ensures a linear RTT fairness under large windows while offering both scalability and bounded TCP-friendliness. The protocol combines two schemes called additive increase and binary search increase. When the congestion window is large, additive increase with a large increment ensures linear RTT fairness as well as good scalability. Under small congestion windows, binary search increase provides TCP friendliness.
- tcp_bic_low_window (integer; default: 14)
- Sets the threshold window (in packets) where BIC TCP starts to adjust the congestion window. Below this threshold BIC TCP behaves the same as the default TCP Reno.
- tcp_bic_fast_convergence (Boolean; default: enabled)
- Forces BIC TCP to more quickly respond to changes in congestion window. Allows two flows sharing the same connection to converge more rapidly.
- tcp_dsack (Boolean; default: enabled)
- Enable RFC 2883 TCP Duplicate SACK support.
- tcp_ecn (Boolean; default: disabled)
- Enable RFC 2884 Explicit Congestion Notification. When enabled, connectivity to some destinations could be affected due to older, misbehaving routers along the path causing connections to be dropped.
- tcp_fack (Boolean; default: enabled)
- Enable TCP Forward Acknowledgement support.
- tcp_fin_timeout (integer; default: 60)
- This specifies how many seconds to wait for a final FIN packet before the socket is forcibly closed. This is strictly a violation of the TCP specification, but required to prevent denial-of-service attacks. In Linux 2.2, the default value was 180.
- tcp_frto (Boolean; default: disabled)
- Enables F-RTO, an enhanced recovery algorithm for TCP retransmission timeouts. It is particularly beneficial in wireless environments where packet loss is typically due to random radio interference rather than intermediate router congestion.
- tcp_keepalive_intvl (integer; default: 75)
- The number of seconds between TCP keep-alive probes.
- tcp_keepalive_probes (integer; default: 9)
- The maximum number of TCP keep-alive probes to send before giving up and killing the connection if no response is obtained from the other end.
- tcp_keepalive_time (integer; default: 7200)
The number of seconds a connection needs to be idle
before TCP begins sending out keep-alive probes.
Keep-alives are only sent when the
socket option is enabled.
The default value is 7200 seconds (2 hours).
An idle connection is terminated after
approximately an additional 11 minutes (9 probes an interval
of 75 seconds apart) when keep-alive is enabled.
Note that underlying connection tracking mechanisms and application timeouts may be much shorter.
- tcp_low_latency (Boolean; default: disabled)
- If enabled, the TCP stack makes decisions that prefer lower latency as opposed to higher throughput. It this option is disabled, then higher throughput is preferred. An example of an application where this default should be changed would be a Beowulf compute cluster.
- tcp_max_orphans (integer; default: see below)
- The maximum number of orphaned (not attached to any user file handle) TCP sockets allowed in the system. When this number is exceeded, the orphaned connection is reset and a warning is printed. This limit exists only to prevent simple denial-of-service attacks. Lowering this limit is not recommended. Network conditions might require you to increase the number of orphans allowed, but note that each orphan can eat up to ~64K of unswappable memory. The default initial value is set equal to the kernel parameter NR_FILE. This initial default is adjusted depending on the memory in the system.
- tcp_max_syn_backlog (integer; default: see below)
- The maximum number of queued connection requests which have still not received an acknowledgement from the connecting client. If this number is exceeded, the kernel will begin dropping requests. The default value of 256 is increased to 1024 when the memory present in the system is adequate or greater (>= 128Mb), and reduced to 128 for those systems with very low memory (<= 32Mb). It is recommended that if this needs to be increased above 1024, TCP_SYNQ_HSIZE in include/net/tcp.h be modified to keep TCP_SYNQ_HSIZE*16<=tcp_max_syn_backlog, and the kernel be recompiled.
- tcp_max_tw_buckets (integer; default: see below)
- The maximum number of sockets in TIME_WAIT state allowed in the system. This limit exists only to prevent simple denial-of-service attacks. The default value of NR_FILE*2 is adjusted depending on the memory in the system. If this number is exceeded, the socket is closed and a warning is printed.
This is a vector of 3 integers: [low, pressure, high].
These bounds are used by TCP to track its memory usage.
defaults are calculated at boot time from the amount of
(TCP can only use
for this, which is limited to around 900 megabytes on 32-bit systems.
64-bit systems do not suffer this limitation.)
low - TCP doesn't regulate its memory allocation when the number of pages it has allocated globally is below this number.
pressure - when the amount of memory allocated by TCP exceeds this number of pages, TCP moderates its memory consumption. This memory pressure state is exited once the number of pages allocated falls below the low mark.
high - the maximum number of pages, globally, that TCP will allocate. This value overrides any other limits imposed by the kernel.
- tcp_orphan_retries (integer; default: 8)
- The maximum number of attempts made to probe the other end of a connection which has been closed by our end.
- tcp_reordering (integer; default: 3)
- The maximum a packet can be reordered in a TCP packet stream without TCP assuming packet loss and going into slow start. It is not advisable to change this number. This is a packet reordering detection metric designed to minimize unnecessary back off and retransmits provoked by reordering of packets on a connection.
- tcp_retrans_collapse (Boolean; default: enabled)
- Try to send full-sized packets during retransmit.
- tcp_retries1 (integer; default: 3)
- The number of times TCP will attempt to retransmit a packet on an established connection normally, without the extra effort of getting the network layers involved. Once we exceed this number of retransmits, we first have the network layer update the route if possible before each new retransmit. The default is the RFC specified minimum of 3.
- tcp_retries2 (integer; default: 15)
- The maximum number of times a TCP packet is retransmitted in established state before giving up. The default value is 15, which corresponds to a duration of approximately between 13 to 30 minutes, depending on the retransmission timeout. The RFC 1122 specified minimum limit of 100 seconds is typically deemed too short.
- tcp_rfc1337 (Boolean; default: disabled)
- Enable TCP behaviour conformant with RFC 1337. When disabled, if a RST is received in TIME_WAIT state, we close the socket immediately without waiting for the end of the TIME_WAIT period.
This is a vector of 3 integers: [min, default,
These parameters are used by TCP to regulate receive
TCP dynamically adjusts the size of the
receive buffer from the defaults listed below, in the range
of these sysctl variables, depending on memory available
in the system.
min - minimum size of the receive buffer used by each TCP socket. The default value is 4K, and is lowered to PAGE_SIZE bytes in low-memory systems. This value is used to ensure that in memory pressure mode, allocations below this size will still succeed. This is not used to bound the size of the receive buffer declared using SO_RCVBUF on a socket.
default - the default size of the receive buffer for a TCP socket. This value overwrites the initial default buffer size from the generic global net.core.rmem_default defined for all protocols. The default value is 87380 bytes, and is lowered to 43689 in low-memory systems. If larger receive buffer sizes are desired, this value should be increased (to affect all sockets). To employ large TCP windows, the net.ipv4.tcp_window_scaling must be enabled (default).
max - the maximum size of the receive buffer used by each TCP socket. This value does not override the global net.core.rmem_max. This is not used to limit the size of the receive buffer declared using SO_RCVBUF on a socket. The default value of 87380*2 bytes is lowered to 87380 in low-memory systems.
- tcp_sack (Boolean; default: enabled)
- Enable RFC 2018 TCP Selective Acknowledgements.
- tcp_stdurg (Boolean; default: disabled)
- If this option is enabled, then use the RFC 1122 interpretation of the TCP urgent-pointer field. According to this interpretation, the urgent pointer points to the last byte of urgent data. If this option is disabled, then use the BSD-compatible interpretation of the urgent pointer: the urgent pointer points to the first byte after the urgent data. Enabling this option may lead to interoperability problems.
- tcp_synack_retries (integer; default: 5)
- The maximum number of times a SYN/ACK segment for a passive TCP connection will be retransmitted. This number should not be higher than 255.
- tcp_syncookies (Boolean)
- Enable TCP syncookies. The kernel must be compiled with CONFIG_SYN_COOKIES. Send out syncookies when the syn backlog queue of a socket overflows. The syncookies feature attempts to protect a socket from a SYN flood attack. This should be used as a last resort, if at all. This is a violation of the TCP protocol, and conflicts with other areas of TCP such as TCP extensions. It can cause problems for clients and relays. It is not recommended as a tuning mechanism for heavily loaded servers to help with overloaded or misconfigured conditions. For recommended alternatives see tcp_max_syn_backlog, tcp_synack_retries, and tcp_abort_on_overflow.
- tcp_syn_retries (integer; default: 5)
- The maximum number of times initial SYNs for an active TCP connection attempt will be retransmitted. This value should not be higher than 255. The default value is 5, which corresponds to approximately 180 seconds.
- tcp_timestamps (Boolean; default: enabled)
- Enable RFC 1323 TCP timestamps.
- tcp_tw_recycle (Boolean; default: disabled)
- Enable fast recycling of TIME-WAIT sockets. Enabling this option is not recommended since this causes problems when working with NAT (Network Address Translation).
- tcp_tw_reuse (Boolean; default: disabled)
- Allow to reuse TIME-WAIT sockets for new connections when it is safe from protocol viewpoint. It should not be changed without advice/request of technical experts.
- tcp_window_scaling (Boolean; default: enabled)
- Enable RFC 1323 TCP window scaling. This feature allows the use of a large window (> 64K) on a TCP connection, should the other end support it. Normally, the 16 bit window length field in the TCP header limits the window size to less than 64K bytes. If larger windows are desired, applications can increase the size of their socket buffers and the window scaling option will be employed. If tcp_window_scaling is disabled, TCP will not negotiate the use of window scaling with the other end during connection setup.
- tcp_vegas_cong_avoid (Boolean; default: disabled)
- Enable TCP Vegas congestion avoidance algorithm. TCP Vegas is a sender-side only change to TCP that anticipates the onset of congestion by estimating the bandwidth. TCP Vegas adjusts the sending rate by modifying the congestion window. TCP Vegas should provide less packet loss, but it is not as aggressive as TCP Reno.
- tcp_westwood (Boolean; default: disabled)
- Enable TCP Westwood+ congestion control algorithm. TCP Westwood+ is a sender-side only modification of the TCP Reno protocol stack that optimizes the performance of TCP congestion control. It is based on end-to-end bandwidth estimation to set congestion window and slow start threshold after a congestion episode. Using this estimation, TCP Westwood+ adaptively sets a slow start threshold and a congestion window which takes into account the bandwidth used at the time congestion is experienced. TCP Westwood+ significantly increases fairness with respect to TCP Reno in wired networks and throughput over wireless links.
This is a vector of 3 integers: [min, default, max].
These parameters are used by TCP to regulate send buffer sizes.
TCP dynamically adjusts the size of the send buffer from the
default values listed below, in the range of these sysctl
variables, depending on memory available.
min - minimum size of the send buffer used by each TCP socket. The default value is 4K bytes. This value is used to ensure that in memory pressure mode, allocations below this size will still succeed. This is not used to bound the size of the send buffer declared using SO_SNDBUF on a socket.
default - the default size of the send buffer for a TCP socket. This value overwrites the initial default buffer size from the generic global net.core.wmem_default defined for all protocols. The default value is 16K bytes. If larger send buffer sizes are desired, this value should be increased (to affect all sockets). To employ large TCP windows, the sysctl variable net.ipv4.tcp_window_scaling must be enabled (default).
max - the maximum size of the send buffer used by each TCP socket. This value does not override the global net.core.wmem_max. This is not used to limit the size of the send buffer declared using SO_SNDBUF on a socket. The default value is 128K bytes. It is lowered to 64K depending on the memory available in the system.
Socket OptionsTo set or get a TCP socket option, call getsockopt(2) to read or setsockopt(2) to write the option with the option level argument set to IPPROTO_TCP. In addition, most IPPROTO_IP socket options are valid on TCP sockets. For more information see ip(7).
- If set, don't send out partial frames. All queued partial frames are sent when the option is cleared again. This is useful for prepending headers before calling sendfile(2), or for throughput optimization. As currently implemented, there is a 200 millisecond ceiling on the time for which output is corked by TCP_CORK. If this ceiling is reached, then queued data is automatically transmitted. This option can be combined with TCP_NODELAY only since Linux 2.5.71. This option should not be used in code intended to be portable.
- Allows a listener to be awakened only when data arrives on the socket. Takes an integer value (seconds), this can bound the maximum number of attempts TCP will make to complete the connection. This option should not be used in code intended to be portable.
- Used to collect information about this socket. The kernel returns a struct tcp_info as defined in the file /usr/include/linux/tcp.h. This option should not be used in code intended to be portable.
- The maximum number of keepalive probes TCP should send before dropping the connection. This option should not be used in code intended to be portable.
- The time (in seconds) the connection needs to remain idle before TCP starts sending keepalive probes, if the socket option SO_KEEPALIVE has been set on this socket. This option should not be used in code intended to be portable.
- The time (in seconds) between individual keepalive probes. This option should not be used in code intended to be portable.
- The lifetime of orphaned FIN_WAIT2 state sockets. This option can be used to override the system wide sysctl tcp_fin_timeout on this socket. This is not to be confused with the socket(7) level option SO_LINGER. This option should not be used in code intended to be portable.
- The maximum segment size for outgoing TCP packets. If this option is set before connection establishment, it also changes the MSS value announced to the other end in the initial packet. Values greater than the (eventual) interface MTU have no effect. TCP will also impose its minimum and maximum bounds over the value provided.
- If set, disable the Nagle algorithm. This means that segments are always sent as soon as possible, even if there is only a small amount of data. When not set, data is buffered until there is a sufficient amount to send out, thereby avoiding the frequent sending of small packets, which results in poor utilization of the network. This option is overridden by TCP_CORK; however, setting this option forces an explicit flush of pending output, even if TCP_CORK is currently set.
- Enable quickack mode if set or disable quickack mode if cleared. In quickack mode, acks are sent immediately, rather than delayed if needed in accordance to normal TCP operation. This flag is not permanent, it only enables a switch to or from quickack mode. Subsequent operation of the TCP protocol will once again enter/leave quickack mode depending on internal protocol processing and factors such as delayed ack timeouts occurring and data transfer. This option should not be used in code intended to be portable.
- Set the number of SYN retransmits that TCP should send before aborting the attempt to connect. It cannot exceed 255. This option should not be used in code intended to be portable.
- Bound the size of the advertised window to this value. The kernel imposes a minimum size of SOCK_MIN_RCVBUF/2. This option should not be used in code intended to be portable.
IoctlsThese following ioctl(2) calls return information in value. The correct syntax is:
int value; error = ioctl(tcp_socket, ioctl_type, &value);
ioctl_type is one of the following:
- Returns the amount of queued unread data in the receive buffer. The socket must not be in LISTEN state, otherwise an error (EINVAL) is returned.
Returns true (i.e.,
is non-zero) if the inbound data stream is at the urgent mark.
If the SO_OOBINLINE socket option is set, and SIOCATMARK returns true, then the next read from the socket will return the urgent data. If the SO_OOBINLINE socket option is not set, and SIOCATMARK returns true, then the next read from the socket will return the bytes following the urgent data (to actually read the urgent data requires the recv(MSG_OOB) flag).
Note that a read never reads across the urgent mark. If an application is informed of the presence of urgent data via select(2) (using the exceptfds argument) or through delivery of a SIGURG signal, then it can advance up to the mark using a loop which repeatedly tests SIOCATMARK and performs a read (requesting any number of bytes) as long as SIOCATMARK returns false.
- Returns the amount of unsent data in the socket send queue. The socket must not be in LISTEN state, otherwise an error (EINVAL) is returned.
Error HandlingWhen a network error occurs, TCP tries to resend the packet. If it doesn't succeed after some time, either ETIMEDOUT or the last received error on this connection is reported.
Some applications require a quicker error notification. This can be enabled with the IPPROTO_IP level IP_RECVERR socket option. When this option is enabled, all incoming errors are immediately passed to the user program. Use this option with care --- it makes TCP less tolerant to routing changes and other normal network conditions.
- The other end closed the socket unexpectedly or a read is executed on a shut down socket.
- The other end didn't acknowledge retransmitted data after some time.
- Passed socket address type in sin_family was not AF_INET.
Any errors defined for ip(7) or the generic socket layer may also be returned for TCP.
VERSIONSSupport for Explicit Congestion Notification, zero-copy sendfile(2), reordering support and some SACK extensions (DSACK) were introduced in 2.4. Support for forward acknowledgement (FACK), TIME_WAIT recycling, per connection keepalive socket options and sysctls were introduced in 2.3.
NOTESTCP has no real out-of-band data; it has urgent data. In Linux this means if the other end sends newer out-of-band data the older urgent data is inserted as normal data into the stream (even when SO_OOBINLINE is not set). This differs from BSD-based stacks.
Linux uses the BSD compatible interpretation of the urgent pointer field by default. This violates RFC 1122, but is required for interoperability with other stacks. It can be changed by the tcp_stdurg sysctl.
BUGSNot all errors are documented.
IPv6 is not described.
SEE ALSOaccept(2), bind(2), connect(2), getsockopt(2), listen(2), recvmsg(2), sendfile(2), sendmsg(2), socket(2), sysctl(2), ip(7), socket(7)
RFC 793 for the TCP specification.
RFC 1122 for the TCP requirements and a description of the Nagle algorithm.
RFC 1323 for TCP timestamp and window scaling options.
RFC 1644 for a description of TIME_WAIT assassination hazards.
RFC 3168 for a description of Explicit Congestion Notification.
RFC 2581 for TCP congestion control algorithms.
RFC 2018 and RFC 2883 for SACK and extensions to SACK.
- SEE ALSO