ioctl_userfaultfd.2: Update UFFDIO_API description

[man-pages.git] / man7 / socket.7

'\" t
.\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
.\" and copyright (c) 1999 Matthew Wilcox.
.\"
.\" %%%LICENSE_START(VERBATIM_ONE_PARA)
.\" Permission is granted to distribute possibly modified copies
.\" of this page provided the header is included verbatim,
.\" and in case of nontrivial modification author and date
.\" of the modification is added to the header.
.\" %%%LICENSE_END
.\"
.\" 2002-10-30, Michael Kerrisk, <mtk.manpages@gmail.com>
.\"     Added description of SO_ACCEPTCONN
.\" 2004-05-20, aeb, added SO_RCVTIMEO/SO_SNDTIMEO text.
.\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
.\"     Added notes on capability requirements
.\"     A few small grammar fixes
.\" 2010-06-13 Jan Engelhardt <jengelh@medozas.de>
.\"     Documented SO_DOMAIN and SO_PROTOCOL.
.\"
.\" FIXME
.\" The following are not yet documented:
.\"
.\" SO_PEERNAME (2.4?)
.\"     get only
.\"     Seems to do something similar to getpeername(), but then
.\"     why is it necessary / how does it differ?
.\"
.\" SO_TIMESTAMPNS (2.6.22)
.\"     Documentation/networking/timestamping.txt
.\"     commit 92f37fd2ee805aa77925c1e64fd56088b46094fc
.\"     Author: Eric Dumazet <dada1@cosmosbay.com>
.\"
.\" SO_TIMESTAMPING (2.6.30)
.\"     Documentation/networking/timestamping.txt
.\"     commit cb9eff097831007afb30d64373f29d99825d0068
.\"     Author: Patrick Ohly <patrick.ohly@intel.com>
.\"
.\"  SO_WIFI_STATUS (3.3)
.\"     commit 6e3e939f3b1bf8534b32ad09ff199d88800835a0
.\"     Author: Johannes Berg <johannes.berg@intel.com>
.\"     Also: SCM_WIFI_STATUS
.\"
.\" SO_NOFCS (3.4)
.\"     commit 3bdc0eba0b8b47797f4a76e377dd8360f317450f
.\"     Author: Ben Greear <greearb@candelatech.com>
.\"
.\"  SO_GET_FILTER (3.8)
.\"     commit a8fc92778080c845eaadc369a0ecf5699a03bef0
.\"     Author: Pavel Emelyanov <xemul@parallels.com>
.\"
.\" SO_SELECT_ERR_QUEUE (3.10)
.\"     commit 7d4c04fc170087119727119074e72445f2bb192b
.\"     Author: Keller, Jacob E <jacob.e.keller@intel.com>
.\"
.\" SO_MAX_PACING_RATE (3.13)
.\"     commit 62748f32d501f5d3712a7c372bbb92abc7c62bc7
.\"     Author: Eric Dumazet <edumazet@google.com>
.\"
.\" SO_BPF_EXTENSIONS (3.14)
.\"     commit ea02f9411d9faa3553ed09ce0ec9f00ceae9885e
.\"     Author: Michal Sekletar <msekleta@redhat.com>
.\"
.TH SOCKET 7 2016-10-08 Linux "Linux Programmer's Manual"
.SH NAME
socket \- Linux socket interface
.SH SYNOPSIS
.B #include <sys/socket.h>
.sp
.IB sockfd " = socket(int " socket_family ", int " socket_type ", int " protocol );
.SH DESCRIPTION
This manual page describes the Linux networking socket layer user
interface.
The BSD compatible sockets
are the uniform interface
between the user process and the network protocol stacks in the kernel.
The protocol modules are grouped into
.I protocol families
such as
.BR AF_INET ", " AF_IPX ", and " AF_PACKET ,
and
.I socket types
such as
.B SOCK_STREAM
or
.BR SOCK_DGRAM .
See
.BR socket (2)
for more information on families and types.
.SS Socket-layer functions
These functions are used by the user process to send or receive packets
and to do other socket operations.
For more information see their respective manual pages.

.BR socket (2)
creates a socket,
.BR connect (2)
connects a socket to a remote socket address,
the
.BR bind (2)
function binds a socket to a local socket address,
.BR listen (2)
tells the socket that new connections shall be accepted, and
.BR accept (2)
is used to get a new socket with a new incoming connection.
.BR socketpair (2)
returns two connected anonymous sockets (implemented only for a few
local families like
.BR AF_UNIX )
.PP
.BR send (2),
.BR sendto (2),
and
.BR sendmsg (2)
send data over a socket, and
.BR recv (2),
.BR recvfrom (2),
.BR recvmsg (2)
receive data from a socket.
.BR poll (2)
and
.BR select (2)
wait for arriving data or a readiness to send data.
In addition, the standard I/O operations like
.BR write (2),
.BR writev (2),
.BR sendfile (2),
.BR read (2),
and
.BR readv (2)
can be used to read and write data.
.PP
.BR getsockname (2)
returns the local socket address and
.BR getpeername (2)
returns the remote socket address.
.BR getsockopt (2)
and
.BR setsockopt (2)
are used to set or get socket layer or protocol options.
.BR ioctl (2)
can be used to set or read some other options.
.PP
.BR close (2)
is used to close a socket.
.BR shutdown (2)
closes parts of a full-duplex socket connection.
.PP
Seeking, or calling
.BR pread (2)
or
.BR pwrite (2)
with a nonzero position is not supported on sockets.
.PP
It is possible to do nonblocking I/O on sockets by setting the
.B O_NONBLOCK
flag on a socket file descriptor using
.BR fcntl (2).
Then all operations that would block will (usually)
return with
.B EAGAIN
(operation should be retried later);
.BR connect (2)
will return
.B EINPROGRESS
error.
The user can then wait for various events via
.BR poll (2)
or
.BR select (2).
.TS
tab(:) allbox;
c s s
l l l.
I/O events
Event:Poll flag:Occurrence
Read:POLLIN:T{
New data arrived.
T}
Read:POLLIN:T{
A connection setup has been completed
(for connection-oriented sockets)
T}
Read:POLLHUP:T{
A disconnection request has been initiated by the other end.
T}
Read:POLLHUP:T{
A connection is broken (only for connection-oriented protocols).
When the socket is written
.B SIGPIPE
is also sent.
T}
Write:POLLOUT:T{
Socket has enough send buffer space for writing new data.
T}
Read/Write:T{
POLLIN |
.br
POLLOUT
T}:T{
An outgoing
.BR connect (2)
finished.
T}
Read/Write:POLLERR:An asynchronous error occurred.
Read/Write:POLLHUP:The other end has shut down one direction.
Exception:POLLPRI:T{
Urgent data arrived.
.B SIGURG
is sent then.
T}
.\" FIXME . The following is not true currently:
.\" It is no I/O event when the connection
.\" is broken from the local end using
.\" .BR shutdown (2)
.\" or
.\" .BR close (2).
.TE
.PP
An alternative to
.BR poll (2)
and
.BR select (2)
is to let the kernel inform the application about events
via a
.B SIGIO
signal.
For that the
.B O_ASYNC
flag must be set on a socket file descriptor via
.BR fcntl (2)
and a valid signal handler for
.B SIGIO
must be installed via
.BR sigaction (2).
See the
.I Signals
discussion below.
.SS Socket address structures
Each socket domain has its own format for socket addresses,
with a domain-specific address structure.
Each of these structures begins with an
integer "family" field (typed as
.IR sa_family_t )
that indicates the type of the address structure.
This allows
the various system calls (e.g.,
.BR connect (2),
.BR bind (2),
.BR accept (2),
.BR getsockname (2),
.BR getpeername (2)),
which are generic to all socket domains,
to determine the domain of a particular socket address.

To allow any type of socket address to be passed to
interfaces in the sockets API,
the type
.IR "struct sockaddr"
is defined.
The purpose of this type is purely to allow casting of
domain-specific socket address types to a "generic" type,
so as to avoid compiler warnings about type mismatches in
calls to the sockets API.

In addition, the sockets API provides the data type
.IR "struct sockaddr_storage".
This type
is suitable to accommodate all supported domain-specific socket
address structures; it is large enough and is aligned properly.
(In particular, it is large enough to hold
IPv6 socket addresses.)
The structure includes the following field, which can be used to identify
the type of socket address actually stored in the structure:

.in +4n
.nf
    sa_family_t ss_family;
.fi
.in

The
.I sockaddr_storage
structure is useful in programs that must handle socket addresses
in a generic way
(e.g., programs that must deal with both IPv4 and IPv6 socket addresses).
.SS Socket options
The socket options listed below can be set by using
.BR setsockopt (2)
and read with
.BR getsockopt (2)
with the socket level set to
.B SOL_SOCKET
for all sockets.
Unless otherwise noted,
.I optval
is a pointer to an
.IR int .
.\" FIXME .
.\" In the list below, the text used to describe argument types
.\" for each socket option should be more consistent
.\"
.\" SO_ACCEPTCONN is in POSIX.1-2001, and its origin is explained in
.\" W R Stevens, UNPv1
.TP
.B SO_ACCEPTCONN
Returns a value indicating whether or not this socket has been marked
to accept connections with
.BR listen (2).
The value 0 indicates that this is not a listening socket,
the value 1 indicates that this is a listening socket.
This socket option is read-only.
.TP
.BR SO_ATTACH_FILTER " (since Linux 2.2), " SO_ATTACH_BPF " (since Linux 3.19)"
Attach a classic BPF
.RB ( SO_ATTACH_FILTER )
or an extended BPF
.RB ( SO_ATTACH_BPF )
program to the socket for use as a filter of incoming packets.
A packet will be dropped if the filter program returns zero.
If the filter program returns a
non-zero value which is less than the packet's data length,
the packet will be truncated to the length returned.
If the value returned by the filter is greater than or equal to the
packet's data length, the packet is allowed to proceed unmodified.

The argument for
.BR SO_ATTACH_FILTER
is a
.I sock_fprog
structure, defined in
.IR <linux/filter.h> :
.sp
.in +4n
.nf
struct sock_fprog {
    unsigned short      len;
    struct sock_filter *filter;
};
.fi
.in
.IP
The argument for
.BR SO_ATTACH_BPF
is a file descriptor returned by the
.BR bpf (2)
system call and must refer to a program of type
.BR BPF_PROG_TYPE_SOCKET_FILTER.

These options may be set multiple times for a given socket,
each time replacing the previous filter program.
The classic and extended versions may be called on the same socket,
but the previous filter will always be replaced such that a socket
never has more than one filter defined.

Both classic and extended BPF are explained in the kernel source file
.I Documentation/networking/filter.txt
.TP
.BR SO_ATTACH_REUSEPORT_CBPF ", " SO_ATTACH_REUSEPORT_EBPF
For use with the
.BR SO_REUSEPORT
option, these options allow the user to set a classic BPF
.RB ( SO_ATTACH_REUSEPORT_CBPF )
or an extended BPF
.RB ( SO_ATTACH_REUSEPORT_EBPF )
program which defines how packets are assigned to
the sockets in the reuseport group (that is, all sockets which have
.BR SO_REUSEPORT
set and are using the same local address to receive packets).

The BPF program must return an index between 0 and N\-1 representing
the socket which should receive the packet
(where N is the number of sockets in the group).
If the BPF program returns an invalid index,
socket selection will fall back to the plain
.BR SO_REUSEPORT
mechanism.

Sockets are numbered in the order in which they are added to the group
(that is, the order of
.BR bind (2)
calls for UDP sockets or the order of
.BR listen (2)
calls for TCP sockets).
New sockets added to a reuseport group will inherit the BPF program.
When a socket is removed from a reuseport group (via
.BR close (2)),
the last socket in the group will be moved into the closed socket's
position.

These options may be set repeatedly at any time on any socket in the group
to replace the current BPF program used by all sockets in the group.

.BR SO_ATTACH_REUSEPORT_CBPF
takes the same argument type as
.BR SO_ATTACH_FILTER
and
.BR SO_ATTACH_REUSEPORT_EBPF
takes the same argument type as
.BR SO_ATTACH_BPF.

UDP support for this feature is available since Linux 4.5;
TCP support is available since Linux 4.6.
.TP
.B SO_BINDTODEVICE
Bind this socket to a particular device like \(lqeth0\(rq,
as specified in the passed interface name.
If the
name is an empty string or the option length is zero, the socket device
binding is removed.
The passed option is a variable-length null-terminated
interface name string with the maximum size of
.BR IFNAMSIZ .
If a socket is bound to an interface,
only packets received from that particular interface are processed by the
socket.
Note that this works only for some socket types, particularly
.B AF_INET
sockets.
It is not supported for packet sockets (use normal
.BR bind (2)
there).

Before Linux 3.8,
this socket option could be set, but could not retrieved with
.BR getsockopt (2).
Since Linux 3.8, it is readable.
The
.I optlen
argument should contain the buffer size available
to receive the device name and is recommended to be
.BR IFNAMSZ
bytes.
The real device name length is reported back in the
.I optlen
argument.
.TP
.B SO_BROADCAST
Set or get the broadcast flag.
When enabled, datagram sockets are allowed to send
packets to a broadcast address.
This option has no effect on stream-oriented sockets.
.TP
.B SO_BSDCOMPAT
Enable BSD bug-to-bug compatibility.
This is used by the UDP protocol module in Linux 2.0 and 2.2.
If enabled, ICMP errors received for a UDP socket will not be passed
to the user program.
In later kernel versions, support for this option has been phased out:
Linux 2.4 silently ignores it, and Linux 2.6 generates a kernel warning
(printk()) if a program uses this option.
Linux 2.0 also enabled BSD bug-to-bug compatibility
options (random header changing, skipping of the broadcast flag) for raw
sockets with this option, but that was removed in Linux 2.2.
.TP
.B SO_DEBUG
Enable socket debugging.
Allowed only for processes with the
.B CAP_NET_ADMIN
capability or an effective user ID of 0.
.TP
.BR SO_DETACH_FILTER " (since Linux 2.2), " SO_DETACH_BPF " (since Linux 3.19)"
These two options, which are synonyms,
may be used to remove the classic or extended BPF
program attached to a socket with either
.BR SO_ATTACH_FILTER
or
.BR SO_ATTACH_BPF .
The option value is ignored.
.TP
.BR SO_DOMAIN " (since Linux 2.6.32)"
Retrieves the socket domain as an integer, returning a value such as
.BR AF_INET6 .
See
.BR socket (2)
for details.
This socket option is read-only.
.TP
.B SO_ERROR
Get and clear the pending socket error.
This socket option is read-only.
Expects an integer.
.TP
.B SO_DONTROUTE
Don't send via a gateway, send only to directly connected hosts.
The same effect can be achieved by setting the
.B MSG_DONTROUTE
flag on a socket
.BR send (2)
operation.
Expects an integer boolean flag.
.TP
.BR SO_INCOMING_CPU " (gettable since Linux 3.19, settable since Linux 4.4)"
.\" getsockopt 2c8c56e15df3d4c2af3d656e44feb18789f75837
.\" setsockopt 70da268b569d32a9fddeea85dc18043de9d89f89
Sets or gets the CPU affinity of a socket.
Expects an integer flag.

.in +4n
.nf
int cpu = 1;
socklen_t len = sizeof(cpu);
setsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);
.fi
.in

Because all of the packets for a single stream
(i.e., all packets for the same 4-tuple)
arrive on the single RX queue that is associated with a particular CPU,
the typical use case is to employ one listening process per RX queue,
with the incoming flow being handled by a listener
on the same CPU that is handling the RX queue.
This provides optimal NUMA behavior and keeps CPU caches hot.
.\"
.\" From an email conversation with Eric Dumazet:
.\" >> Note that setting the option is not supported if SO_REUSEPORT is used.
.\" >
.\" > Please define "not supported". Does this yield an API diagnostic?
.\" > If so, what is it?
.\" >
.\" >> Socket will be selected from an array, either by a hash or BPF program
.\" >> that has no access to this information.
.\" >
.\" > Sorry -- I'm lost here. How does this comment relate to the proposed
.\" > man page text above?
.\"  
.\" Simply that :
.\"  
.\" If an application uses both SO_INCOMING_CPU and SO_REUSEPORT, then
.\" SO_REUSEPORT logic, selecting the socket to receive the packet, ignores
.\" SO_INCOMING_CPU setting.
.TP
.B SO_KEEPALIVE
Enable sending of keep-alive messages on connection-oriented sockets.
Expects an integer boolean flag.
.TP
.B SO_LINGER
Sets or gets the
.B SO_LINGER
option.
The argument is a
.I linger
structure.
.sp
.in +4n
.nf
struct linger {
    int l_onoff;    /* linger active */
    int l_linger;   /* how many seconds to linger for */
};
.fi
.in
.IP
When enabled, a
.BR close (2)
or
.BR shutdown (2)
will not return until all queued messages for the socket have been
successfully sent or the linger timeout has been reached.
Otherwise,
the call returns immediately and the closing is done in the background.
When the socket is closed as part of
.BR exit (2),
it always lingers in the background.
.TP
.B SO_LOCK_FILTER
.\" commit d59577b6ffd313d0ab3be39cb1ab47e29bdc9182
When set, this option will prevent
changing the filters associated with the socket.
These filters include any set using the socket options
.BR SO_ATTACH_FILTER,
.BR SO_ATTACH_BPF,
.BR SO_ATTACH_REUSEPORT_CBPF
and
.BR SO_ATTACH_REUSEPORT_EPBF .

The typical use case is for a privileged process to set up a raw socket
(an operation that requires the
.BR CAP_NET_RAW
capability), apply a restrictive filter, set the
.BR SO_LOCK_FILTER
option,
and then either drop its privileges or pass the socket file descriptor
to an unprivileged process via a UNIX domain socket.

Once the
.BR SO_LOCK_FILTER
option has been enabled, attempts to change or remove the filter
attached to a socket, or to disable the
.BR SO_LOCK_FILTER
option will fail with the error
.BR EPERM .
.TP
.BR SO_MARK " (since Linux 2.6.25)"
.\" commit 4a19ec5800fc3bb64e2d87c4d9fdd9e636086fe0
.\" and    914a9ab386a288d0f22252fc268ecbc048cdcbd5
Set the mark for each packet sent through this socket
(similar to the netfilter MARK target but socket-based).
Changing the mark can be used for mark-based
routing without netfilter or for packet filtering.
Setting this option requires the
.B CAP_NET_ADMIN
capability.
.TP
.B SO_OOBINLINE
If this option is enabled,
out-of-band data is directly placed into the receive data stream.
Otherwise, out-of-band data is passed only when the
.B MSG_OOB
flag is set during receiving.
.\" don't document it because it can do too much harm.
.\".B SO_NO_CHECK
.\"     The kernel has support for the SO_NO_CHECK socket
.\"     option (boolean: 0 == default, calculate checksum on xmit,
.\"     1 == do not calculate checksum on xmit).
.\" Additional note from Andi Kleen on SO_NO_CHECK (2010-08-30)
.\"     On Linux UDP checksums are essentially free and there's no reason
.\"     to turn them off and it would disable another safety line.
.\"     That is why I didn't document the option.
.TP
.B SO_PASSCRED
Enable or disable the receiving of the
.B SCM_CREDENTIALS
control message.
For more information see
.BR unix (7).
.\" FIXME Document SO_PASSSEC, added in 2.6.18; there is some info
.\" in the 2.6.18 ChangeLog
.TP
.BR SO_PEEK_OFF " (since Linux 3.4)"
.\" commit ef64a54f6e558155b4f149bb10666b9e914b6c54
This option, which is currently supported only for
.BR unix (7)
sockets, sets the value of the "peek offset" for the
.BR recv (2)
system call when used with
.BR MSG_PEEK
flag.

When this option is set to a negative value
(it is set to \-1 for all new sockets),
traditional behavior is provided:
.BR recv (2)
with the
.BR MSG_PEEK
flag will peek data from the front of the queue.

When the option is set to a value greater than or equal to zero,
then the next peek at data queued in the socket will occur at
the byte offset specified by the option value.
At the same time, the "peek offset" will be
incremented by the number of bytes that were peeked from the queue,
so that a subsequent peek will return the next data in the queue.

If data is removed from the front of the queue via a call to
.BR recv (2)
(or similar) without the
.BR MSG_PEEK
flag, the "peek offset" will be decreased by the number of bytes removed.
In other words, receiving data without the
.B MSG_PEEK
flag will cause the "peek offset" to be adjusted to maintain
the correct relative position in the queued data,
so that a subsequent peek will retrieve the data that would have been
retrieved had the data not been removed.

For datagram sockets, if the "peek offset" points to the middle of a packet,
the data returned will be marked with the
.BR MSG_TRUNC
flag.

The following example serves to illustrate the use of
.BR SO_PEEK_OFF .
Suppose a stream socket has the following queued input data:

    aabbccddeeff
.IP
The following sequence of
.BR recv (2)
calls would have the effect noted in the comments:

.in +4n
.nf
int ov = 4;                  // Set peek offset to 4
setsockopt(fd, SOL_SOCKET, SO_PEEK_OFF, &ov, sizeof(ov));

recv(fd, buf, 2, MSG_PEEK);  // Peeks "cc"; offset set to 6
recv(fd, buf, 2, MSG_PEEK);  // Peeks "dd"; offset set to 8
recv(fd, buf, 2, 0);         // Reads "aa"; offset set to 6
recv(fd, buf, 2, MSG_PEEK);  // Peeks "ee"; offset set to 8
.fi
.in
.TP
.B SO_PEERCRED
Return the credentials of the foreign process connected to this socket.
This is possible only for connected
.B AF_UNIX
stream sockets and
.B AF_UNIX
stream and datagram socket pairs created using
.BR socketpair (2);
see
.BR unix (7).
The returned credentials are those that were in effect at the time
of the call to
.BR connect (2)
or
.BR socketpair (2).
The argument is a
.I ucred
structure; define the
.B _GNU_SOURCE
feature test macro to obtain the definition of that structure from
.IR <sys/socket.h> .
This socket option is read-only.
.TP
.B SO_PRIORITY
Set the protocol-defined priority for all packets to be sent on
this socket.
Linux uses this value to order the networking queues:
packets with a higher priority may be processed first depending
on the selected device queueing discipline.
.\" For
.\" .BR ip (7),
.\" this also sets the IP type-of-service (TOS) field for outgoing packets.
Setting a priority outside the range 0 to 6 requires the
.B CAP_NET_ADMIN
capability.
.TP
.BR SO_PROTOCOL " (since Linux 2.6.32)"
Retrieves the socket protocol as an integer, returning a value such as
.BR IPPROTO_SCTP .
See
.BR socket (2)
for details.
This socket option is read-only.
.TP
.B SO_RCVBUF
Sets or gets the maximum socket receive buffer in bytes.
The kernel doubles this value (to allow space for bookkeeping overhead)
when it is set using
.\" Most (all?) other implementations do not do this -- MTK, Dec 05
.BR setsockopt (2),
and this doubled value is returned by
.BR getsockopt (2).
.\" The following thread on LMKL is quite informative:
.\" getsockopt/setsockopt with SO_RCVBUF and SO_SNDBUF "non-standard" behavior
.\" 17 July 2012
.\" http://thread.gmane.org/gmane.linux.kernel/1328935
The default value is set by the
.I /proc/sys/net/core/rmem_default
file, and the maximum allowed value is set by the
.I /proc/sys/net/core/rmem_max
file.
The minimum (doubled) value for this option is 256.
.TP
.BR SO_RCVBUFFORCE " (since Linux 2.6.14)"
Using this socket option, a privileged
.RB ( CAP_NET_ADMIN )
process can perform the same task as
.BR SO_RCVBUF ,
but the
.I rmem_max
limit can be overridden.
.TP
.BR SO_RCVLOWAT " and " SO_SNDLOWAT
Specify the minimum number of bytes in the buffer until the socket layer
will pass the data to the protocol
.RB ( SO_SNDLOWAT )
or the user on receiving
.RB ( SO_RCVLOWAT ).
These two values are initialized to 1.
.B SO_SNDLOWAT
is not changeable on Linux
.RB ( setsockopt (2)
fails with the error
.BR ENOPROTOOPT ).
.B SO_RCVLOWAT
is changeable
only since Linux 2.4.
The
.BR select (2)
and
.BR poll (2)
system calls currently do not respect the
.B SO_RCVLOWAT
setting on Linux,
and mark a socket readable when even a single byte of data is available.
A subsequent read from the socket will block until
.B SO_RCVLOWAT
bytes are available.
.\" See http://marc.theaimsgroup.com/?l=linux-kernel&m=111049368106984&w=2
.\" Tested on kernel 2.6.14 -- mtk, 30 Nov 05
.TP
.BR SO_RCVTIMEO " and " SO_SNDTIMEO
.\" Not implemented in 2.0.
.\" Implemented in 2.1.11 for getsockopt: always return a zero struct.
.\" Implemented in 2.3.41 for setsockopt, and actually used.
Specify the receiving or sending timeouts until reporting an error.
The argument is a
.IR "struct timeval" .
If an input or output function blocks for this period of time, and
data has been sent or received, the return value of that function
will be the amount of data transferred; if no data has been transferred
and the timeout has been reached, then \-1 is returned with
.I errno
set to
.BR EAGAIN
or
.BR EWOULDBLOCK ,
.\" in fact to EAGAIN
or
.B EINPROGRESS
(for
.BR connect (2))
just as if the socket was specified to be nonblocking.
If the timeout is set to zero (the default),
then the operation will never timeout.
Timeouts only have effect for system calls that perform socket I/O (e.g.,
.BR read (2),
.BR recvmsg (2),
.BR send (2),
.BR sendmsg (2));
timeouts have no effect for
.BR select (2),
.BR poll (2),
.BR epoll_wait (2),
and so on.
.TP
.B SO_REUSEADDR
.\"     commit c617f398edd4db2b8567a28e899a88f8f574798d
.\"     https://lwn.net/Articles/542629/
Indicates that the rules used in validating addresses supplied in a
.BR bind (2)
call should allow reuse of local addresses.
For
.B AF_INET
sockets this
means that a socket may bind, except when there
is an active listening socket bound to the address.
When the listening socket is bound to
.B INADDR_ANY
with a specific port then it is not possible
to bind to this port for any local address.
Argument is an integer boolean flag.
.TP
.BR SO_REUSEPORT " (since Linux 3.9)"
Permits multiple
.B AF_INET
or
.B AF_INET6
sockets to be bound to an identical socket address.
This option must be set on each socket (including the first socket)
prior to calling
.BR bind (2)
on the socket.
To prevent port hijacking,
all of the processes binding to the same address must have the same
effective UID.
This option can be employed with both TCP and UDP sockets.

For TCP sockets, this option allows
.BR accept (2)
load distribution in a multi-threaded server to be improved by
using a distinct listener socket for each thread.
This provides improved load distribution as compared
to traditional techniques such using a single
.BR accept (2)ing
thread that distributes connections,
or having multiple threads that compete to
.BR accept (2)
from the same socket.

For UDP sockets,
the use of this option can provide better distribution
of incoming datagrams to multiple processes (or threads) as compared
to the traditional technique of having multiple processes
compete to receive datagrams on the same socket.
.TP
.BR SO_RXQ_OVFL " (since Linux 2.6.33)"
.\" commit 3b885787ea4112eaa80945999ea0901bf742707f
Indicates that an unsigned 32-bit value ancillary message (cmsg)
should be attached to received skbs indicating
the number of packets dropped by the socket between
the last received packet and this received packet.
.TP
.B SO_SNDBUF
Sets or gets the maximum socket send buffer in bytes.
The kernel doubles this value (to allow space for bookkeeping overhead)
when it is set using
.\" Most (all?) other implementations do not do this -- MTK, Dec 05
.\" See also the comment to SO_RCVBUF (17 Jul 2012 LKML mail)
.BR setsockopt (2),
and this doubled value is returned by
.BR getsockopt (2).
The default value is set by the
.I /proc/sys/net/core/wmem_default
file and the maximum allowed value is set by the
.I /proc/sys/net/core/wmem_max
file.
The minimum (doubled) value for this option is 2048.
.TP
.BR SO_SNDBUFFORCE " (since Linux 2.6.14)"
Using this socket option, a privileged
.RB ( CAP_NET_ADMIN )
process can perform the same task as
.BR SO_SNDBUF ,
but the
.I wmem_max
limit can be overridden.
.TP
.B SO_TIMESTAMP
Enable or disable the receiving of the
.B SO_TIMESTAMP
control message.
The timestamp control message is sent with level
.B SOL_SOCKET
and the
.I cmsg_data
field is a
.I "struct timeval"
indicating the
reception time of the last packet passed to the user in this call.
See
.BR cmsg (3)
for details on control messages.
.TP
.B SO_TYPE
Gets the socket type as an integer (e.g.,
.BR SOCK_STREAM ).
This socket option is read-only.
.TP
.BR SO_BUSY_POLL " (since Linux 3.11)"
Sets the approximate time in microseconds to busy poll on a blocking receive
when there is no data.
Increasing this value requires
.BR CAP_NET_ADMIN .
The default for this option is controlled by the
.I /proc/sys/net/core/busy_read
file.

The value in the
.I /proc/sys/net/core/busy_poll
file determines how long
.BR select (2)
and
.BR poll (2)
will busy poll when they operate on sockets with
.BR SO_BUSY_POLL
set and no events to report are found.

In both cases,
busy polling will only be done when the socket last received data
from a network device that supports this option.

While busy polling may improve latency of some applications,
care must be taken when using it since this will increase
both CPU utilization and power usage.
.SS Signals
When writing onto a connection-oriented socket that has been shut down
(by the local or the remote end)
.B SIGPIPE
is sent to the writing process and
.B EPIPE
is returned.
The signal is not sent when the write call
specified the
.B MSG_NOSIGNAL
flag.
.PP
When requested with the
.B FIOSETOWN
.BR fcntl (2)
or
.B SIOCSPGRP
.BR ioctl (2),
.B SIGIO
is sent when an I/O event occurs.
It is possible to use
.BR poll (2)
or
.BR select (2)
in the signal handler to find out which socket the event occurred on.
An alternative (in Linux 2.2) is to set a real-time signal using the
.B F_SETSIG
.BR fcntl (2);
the handler of the real time signal will be called with
the file descriptor in the
.I si_fd
field of its
.IR siginfo_t .
See
.BR fcntl (2)
for more information.
.PP
Under some circumstances (e.g., multiple processes accessing a
single socket), the condition that caused the
.B SIGIO
may have already disappeared when the process reacts to the signal.
If this happens, the process should wait again because Linux
will resend the signal later.
.\" .SS Ancillary messages
.SS /proc interfaces
The core socket networking parameters can be accessed
via files in the directory
.IR /proc/sys/net/core/ .
.TP
.I rmem_default
contains the default setting in bytes of the socket receive buffer.
.TP
.I rmem_max
contains the maximum socket receive buffer size in bytes which a user may
set by using the
.B SO_RCVBUF
socket option.
.TP
.I wmem_default
contains the default setting in bytes of the socket send buffer.
.TP
.I wmem_max
contains the maximum socket send buffer size in bytes which a user may
set by using the
.B SO_SNDBUF
socket option.
.TP
.IR message_cost " and " message_burst
configure the token bucket filter used to load limit warning messages
caused by external network events.
.TP
.I netdev_max_backlog
Maximum number of packets in the global input queue.
.TP
.I optmem_max
Maximum length of ancillary data and user control data like the iovecs
per socket.
.\" netdev_fastroute is not documented because it is experimental
.SS Ioctls
These operations can be accessed using
.BR ioctl (2):

.in +4n
.nf
.IB error " = ioctl(" ip_socket ", " ioctl_type ", " &value_result ");"
.fi
.in
.TP
.B SIOCGSTAMP
Return a
.I struct timeval
with the receive timestamp of the last packet passed to the user.
This is useful for accurate round trip time measurements.
See
.BR setitimer (2)
for a description of
.IR "struct timeval" .
.\"
This ioctl should be used only if the socket option
.B SO_TIMESTAMP
is not set on the socket.
Otherwise, it returns the timestamp of the
last packet that was received while
.B SO_TIMESTAMP
was not set, or it fails if no such packet has been received,
(i.e.,
.BR ioctl (2)
returns \-1 with
.I errno
set to
.BR ENOENT ).
.TP
.B SIOCSPGRP
Set the process or process group that is to receive
.B SIGIO
or
.B SIGURG
signals when I/O becomes possible or urgent data is available.
The argument is a pointer to a
.IR pid_t .
For further details, see the description of
.BR F_SETOWN
in
.BR fcntl (2).
.TP
.B FIOASYNC
Change the
.B O_ASYNC
flag to enable or disable asynchronous I/O mode of the socket.
Asynchronous I/O mode means that the
.B SIGIO
signal or the signal set with
.B F_SETSIG
is raised when a new I/O event occurs.
.IP
Argument is an integer boolean flag.
(This operation is synonymous with the use of
.BR fcntl (2)
to set the
.B O_ASYNC
flag.)
.\"
.TP
.B SIOCGPGRP
Get the current process or process group that receives
.B SIGIO
or
.B SIGURG
signals,
or 0
when none is set.
.PP
Valid
.BR fcntl (2)
operations:
.TP
.B FIOGETOWN
The same as the
.B SIOCGPGRP
.BR ioctl (2).
.TP
.B FIOSETOWN
The same as the
.B SIOCSPGRP
.BR ioctl (2).
.SH VERSIONS
.B SO_BINDTODEVICE
was introduced in Linux 2.0.30.
.B SO_PASSCRED
is new in Linux 2.2.
The
.I /proc
interfaces were introduced in Linux 2.2.
.B SO_RCVTIMEO
and
.B SO_SNDTIMEO
are supported since Linux 2.3.41.
Earlier, timeouts were fixed to
a protocol-specific setting, and could not be read or written.
.SH NOTES
Linux assumes that half of the send/receive buffer is used for internal
kernel structures; thus the values in the corresponding
.I /proc
files are twice what can be observed on the wire.

Linux will allow port reuse only with the
.B SO_REUSEADDR
option
when this option was set both in the previous program that performed a
.BR bind (2)
to the port and in the program that wants to reuse the port.
This differs from some implementations (e.g., FreeBSD)
where only the later program needs to set the
.B SO_REUSEADDR
option.
Typically this difference is invisible, since, for example, a server
program is designed to always set this option.
.\" .SH AUTHORS
.\" This man page was written by Andi Kleen.
.SH SEE ALSO
.BR wireshark (1),
.BR bpf (2),
.BR connect (2),
.BR getsockopt (2),
.BR setsockopt (2),
.BR socket (2),
.BR pcap (3),
.BR capabilities (7),
.BR ddp (7),
.BR ip (7),
.BR packet (7),
.BR tcp (7),
.BR udp (7),
.BR unix (7),
.BR tcpdump (8)