unix.7: Note the 108 bytes for sun_path is how things are done on Linux

[thirdparty/man-pages.git] / man7 / unix.7

.\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
.\" and Copyright (C) 2008-2014, Michael Kerrisk <mtk.manpages@gmail.com>
.\"
.\" %%%LICENSE_START(VERBATIM_ONE_PARA)
.\" and Copyright (C) 2008, 2012 Michael Kerrisk <mtk.manpages@gmail.com>
.\" 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
.\"
.\" Modified, 2003-12-02, Michael Kerrisk, <mtk.manpages@gmail.com>
.\" Modified, 2003-09-23, Adam Langley
.\" Modified, 2004-05-27, Michael Kerrisk, <mtk.manpages@gmail.com>
.\"     Added SOCK_SEQPACKET
.\" 2008-05-27, mtk, Provide a clear description of the three types of
.\"     address that can appear in the sockaddr_un structure: pathname,
.\"     unnamed, and abstract.
.\"
.TH UNIX  7 2015-02-21 "Linux" "Linux Programmer's Manual"
.SH NAME
unix \- sockets for local interprocess communication
.SH SYNOPSIS
.B #include <sys/socket.h>
.br
.B #include <sys/un.h>

.IB unix_socket " = socket(AF_UNIX, type, 0);"
.br
.IB error " = socketpair(AF_UNIX, type, 0, int *" sv ");"
.SH DESCRIPTION
The
.B AF_UNIX
(also known as
.BR AF_LOCAL )
socket family is used to communicate between processes on the same machine
efficiently.
Traditionally, UNIX domain sockets can be either unnamed,
or bound to a filesystem pathname (marked as being of type socket).
Linux also supports an abstract namespace which is independent of the
filesystem.

Valid socket types in the UNIX domain are:
.BR SOCK_STREAM ,
for a stream-oriented socket;
.BR SOCK_DGRAM ,
for a datagram-oriented socket that preserves message boundaries
(as on most UNIX implementations, UNIX domain datagram
sockets are always reliable and don't reorder datagrams);
and (since Linux 2.6.4)
.BR SOCK_SEQPACKET ,
for a connection-oriented socket that preserves message boundaries
and delivers messages in the order that they were sent.

UNIX domain sockets support passing file descriptors or process credentials
to other processes using ancillary data.
.SS Address format
A UNIX domain socket address is represented in the following structure:
.in +4n
.nf

.\" #define UNIX_PATH_MAX    108
.\" 
struct sockaddr_un {
    sa_family_t sun_family;               /* AF_UNIX */
    char        sun_path[108];            /* pathname */
};
.fi
.in
.PP
The
.I sun_family
field always contains
.BR AF_UNIX .
On Linux
.I sun_path
is 108 bytes in size; see also NOTES, below.

Various systems calls (for example,
.BR bind (2),
.BR connect (2),
and
.BR sendto (2))
take a
.I sockaddr_un
argument as input.
Some other system calls (for example,
.BR getsockname (2),
.BR getpeername (2),
.BR recvfrom (2),
and
.BR accept (2))
return an argument of this type.

Three types of address are distinguished in the
.I sockaddr_un
structure:
.IP * 3
.IR pathname :
a UNIX domain socket can be bound to a null-terminated
filesystem pathname using
.BR bind (2).
When the address of a pathname socket is returned
(by one of the system calls noted above),
its length is

    offsetof(struct sockaddr_un, sun_path) + strlen(sun_path) + 1

and
.I sun_path
contains the null-terminated pathname.
(On Linux, the above
.BR offsetof ()
expression equates to the same value as
.IR sizeof(sa_family_t) ,
but some other implementations include other fields before
.IR sun_path ,
so the
.BR offsetof ()
expression more portably describes the size of the address structure.)
.IP
For further details of pathname sockets, see below.
.IP *
.IR unnamed :
A stream socket that has not been bound to a pathname using
.BR bind (2)
has no name.
Likewise, the two sockets created by
.BR socketpair (2)
are unnamed.
When the address of an unnamed socket is returned,
its length is
.IR "sizeof(sa_family_t)" ,
and
.I sun_path
should not be inspected.
.\" There is quite some variation across implementations: FreeBSD
.\" says the length is 16 bytes, HP-UX 11 says it's zero bytes.
.IP *
.IR abstract :
an abstract socket address is distinguished (from a pathname socket)
by the fact that
.IR sun_path[0]
is a null byte (\(aq\\0\(aq).
The socket's address in this namespace is given by the additional
bytes in
.IR sun_path
that are covered by the specified length of the address structure.
(Null bytes in the name have no special significance.)
The name has no connection with filesystem pathnames.
When the address of an abstract socket is returned,
the returned
.I addrlen
is greater than
.IR "sizeof(sa_family_t)"
(i.e., greater than 2), and the name of the socket is contained in
the first
.IR "(addrlen \- sizeof(sa_family_t))"
bytes of
.IR sun_path .
The abstract socket namespace is a nonportable Linux extension.
.SS Pathname sockets
When binding a socket to a pathname, a few rules should be observed
for maximum portability and ease of coding:
.IP * 3
The pathname in
.I sun_path
should be null-terminated.
.IP *
The length of the pathname, including the terminating null byte,
should not exceed the size of
.IR sun_path .
.IP *
The
.I addrlen
argument that describes the enclosing
.I sockaddr_un
structure should have a value of at least:

.nf
    offsetof(struct sockaddr_un, sun_path)+strlen(addr.sun_path)+1
.fi
.IP
or, more simply,
.I addrlen
can be specified as
.IR "sizeof(struct sockaddr_un)" .
.PP
There is some variation in how implementations handle UNIX domain
socket addresses that do not follow the above rules.
For example, some (but not all) implementations
.\" Linux does this, including for the case where the supplied path
.\" is 108 bytes
append a null terminator if none is present in the supplied
.IR sun_path .

When coding portable applications,
keep in mind that some implementations
.\" HP-UX
have
.I sun_path
as short as 92 bytes.
.\" Modern BSDs generally have 104, Tru64 and AIX have 104,
.\" Solaris and Irix have 108

Various system calls
.RB ( accept (2),
.BR recvfrom (2),
.BR getsockname (2),
.BR getpeername (2))
return socket address structures.
When applied to UNIX domain sockets, the value-result
.I addrlen
argument supplied to the call should be initialized as above.
Upon return, the argument is set to indicate the
.I actual
size of the address structure.
The caller should check the value returned in this argument:
if the output value exceeds the input value,
then there is no guarantee that a null terminator is present in
.IR sun_path .
(See BUGS.)
.SS Socket options
For historical reasons, these socket options are specified with a
.B SOL_SOCKET
type even though they are
.B AF_UNIX
specific.
They can be set with
.BR setsockopt (2)
and read with
.BR getsockopt (2)
by specifying
.B SOL_SOCKET
as the socket family.
.TP
.B SO_PASSCRED
Enables the receiving of the credentials of the sending process in an
ancillary message.
When this option is set and the socket is not yet connected
a unique name in the abstract namespace will be generated automatically.
Expects an integer boolean flag.
.SS Autobind feature
If a
.BR bind (2)
call specifies
.I addrlen
as
.IR sizeof(sa_family_t) ,
.\" i.e., sizeof(short)
or the
.BR SO_PASSCRED
socket option was specified for a socket that was
not explicitly bound to an address,
then the socket is autobound to an abstract address.
The address consists of a null byte
followed by 5 bytes in the character set
.IR [0-9a-f] .
Thus, there is a limit of 2^20 autobind addresses.
(From Linux 2.1.15, when the autobind feature was added,
8 bytes were used, and the limit was thus 2^32 autobind addresses.
The change to 5 bytes came in Linux 2.3.15.)
.SS Sockets API
The following paragraphs describe domain-specific details and
unsupported features of the sockets API for UNIX domain sockets on Linux.

UNIX domain sockets do not support the transmission of
out-of-band data (the
.B MSG_OOB
flag for
.BR send (2)
and
.BR recv (2)).

The
.BR send (2)
.B MSG_MORE
flag is not supported by UNIX domain sockets.

The use of
.B MSG_TRUNC
in the
.I flags
argument of
.BR recv (2)
is not supported by UNIX domain sockets.

The
.B SO_SNDBUF
socket option does have an effect for UNIX domain sockets, but the
.B SO_RCVBUF
option does not.
For datagram sockets, the
.B SO_SNDBUF
value imposes an upper limit on the size of outgoing datagrams.
This limit is calculated as the doubled (see
.BR socket (7))
option value less 32 bytes used for overhead.
.SS Ancillary messages
Ancillary data is sent and received using
.BR sendmsg (2)
and
.BR recvmsg (2).
For historical reasons the ancillary message types listed below
are specified with a
.B SOL_SOCKET
type even though they are
.B AF_UNIX
specific.
To send them set the
.I cmsg_level
field of the struct
.I cmsghdr
to
.B SOL_SOCKET
and the
.I cmsg_type
field to the type.
For more information see
.BR cmsg (3).
.TP
.B SCM_RIGHTS
Send or receive a set of open file descriptors from another process.
The data portion contains an integer array of the file descriptors.
The passed file descriptors behave as though they have been created with
.BR dup (2).
.TP
.B SCM_CREDENTIALS
Send or receive UNIX credentials.
This can be used for authentication.
The credentials are passed as a
.I struct ucred
ancillary message.
Thus structure is defined in
.I <sys/socket.h>
as follows:

.in +4n
.nf
struct ucred {
    pid_t pid;    /* process ID of the sending process */
    uid_t uid;    /* user ID of the sending process */
    gid_t gid;    /* group ID of the sending process */
};
.fi
.in

Since glibc 2.8, the
.B _GNU_SOURCE
feature test macro must be defined (before including
.I any
header files) in order to obtain the definition
of this structure.

The credentials which the sender specifies are checked by the kernel.
A process with effective user ID 0 is allowed to specify values that do
not match its own.
The sender must specify its own process ID (unless it has the capability
.BR CAP_SYS_ADMIN ),
its user ID, effective user ID, or saved set-user-ID (unless it has
.BR CAP_SETUID ),
and its group ID, effective group ID, or saved set-group-ID
(unless it has
.BR CAP_SETGID ).
To receive a
.I struct ucred
message the
.B SO_PASSCRED
option must be enabled on the socket.
.SS Ioctls
The following
.BR ioctl (2)
calls return information in
.IR value .
The correct syntax is:
.PP
.RS
.nf
.BI int " value";
.IB error " = ioctl(" unix_socket ", " ioctl_type ", &" value ");"
.fi
.RE
.PP
.I ioctl_type
can be:
.TP
.B SIOCINQ
For
.B SOCK_STREAM
socket the function returns the amount of queued unread data in the receive buffer.
The socket must not be in LISTEN state, otherwise an error
.RB ( EINVAL )
is returned.
.B SIOCINQ
is defined in
.IR <linux/sockios.h> .
.\" FIXME . http://sources.redhat.com/bugzilla/show_bug.cgi?id=12002,
.\" filed 2010-09-10, may cause SIOCINQ to be defined in glibc headers
Alternatively,
you can use the synonymous
.BR FIONREAD ,
defined in
.IR <sys/ioctl.h> .
.\" SIOCOUTQ also has an effect for UNIX domain sockets, but not
.\" quite what userland might expect. It seems to return the number
.\" of bytes allocated for buffers containing pending output.
.\" That number is normally larger than the number of bytes of pending
.\" output. Since this info is, from userland's point of view, imprecise,
.\" and it may well change, probably best not to document this now.
For
.B SOCK_DGRAM
socket,
the returned value is the same as
for Internet domain datagram socket;
see
.BR udp (7).
.SH ERRORS
.TP
.B EADDRINUSE
The specified local address is already in use or the filesystem socket
object already exists.
.TP
.B ECONNREFUSED
The remote address specified by
.BR connect (2)
was not a listening socket.
This error can also occur if the target pathname is not a socket.
.TP
.B ECONNRESET
Remote socket was unexpectedly closed.
.TP
.B EFAULT
User memory address was not valid.
.TP
.B EINVAL
Invalid argument passed.
A common cause is that the value
.B AF_UNIX
was not specified in the
.I sun_type
field of passed addresses, or the socket was in an
invalid state for the applied operation.
.TP
.B EISCONN
.BR connect (2)
called on an already connected socket or a target address was
specified on a connected socket.
.TP
.B ENOENT
The pathname in the remote address specified to
.BR connect (2)
did not exist.
.TP
.B ENOMEM
Out of memory.
.TP
.B ENOTCONN
Socket operation needs a target address, but the socket is not connected.
.TP
.B EOPNOTSUPP
Stream operation called on non-stream oriented socket or tried to
use the out-of-band data option.
.TP
.B EPERM
The sender passed invalid credentials in the
.IR "struct ucred" .
.TP
.B EPIPE
Remote socket was closed on a stream socket.
If enabled, a
.B SIGPIPE
is sent as well.
This can be avoided by passing the
.B MSG_NOSIGNAL
flag to
.BR sendmsg (2)
or
.BR recvmsg (2).
.TP
.B EPROTONOSUPPORT
Passed protocol is not
.BR AF_UNIX .
.TP
.B EPROTOTYPE
Remote socket does not match the local socket type
.RB ( SOCK_DGRAM
versus
.BR SOCK_STREAM )
.TP
.B ESOCKTNOSUPPORT
Unknown socket type.
.PP
Other errors can be generated by the generic socket layer or
by the filesystem while generating a filesystem socket object.
See the appropriate manual pages for more information.
.SH VERSIONS
.B SCM_CREDENTIALS
and the abstract namespace were introduced with Linux 2.2 and should not
be used in portable programs.
(Some BSD-derived systems also support credential passing,
but the implementation details differ.)
.SH NOTES
In the Linux implementation, sockets which are visible in the
filesystem honor the permissions of the directory they are in.
Their owner, group, and permissions can be changed.
Creation of a new socket will fail if the process does not have write and
search (execute) permission on the directory the socket is created in.
Connecting to the socket object requires read/write permission.
This behavior differs from many BSD-derived systems which
ignore permissions for UNIX domain sockets.
Portable programs should not rely on
this feature for security.

Binding to a socket with a filename creates a socket
in the filesystem that must be deleted by the caller when it is no
longer needed (using
.BR unlink (2)).
The usual UNIX close-behind semantics apply; the socket can be unlinked
at any time and will be finally removed from the filesystem when the last
reference to it is closed.

To pass file descriptors or credentials over a
.BR SOCK_STREAM ,
you need
to send or receive at least one byte of nonancillary data in the same
.BR sendmsg (2)
or
.BR recvmsg (2)
call.

UNIX domain stream sockets do not support the notion of out-of-band data.
.\"
.SH BUGS
When binding a socket to an address,
Linux is one of the implementations that appends a null terminator
if none is supplied in
.IR sun_path .
In most cases this is unproblematic:
when the socket address is retrieved,
it will be one byte longer than that supplied when the socket was bound.
However, there is one case where confusing behavior can result:
if 108 non-null bytes are supplied when a socket is bound,
then the addition of the null terminator takes the length of
the pathname beyond
.IR sizeof(sun_path) .
Consequently, when retrieving the socket address
(for example, via
.BR accept (2)),
.\" The behavior on Solaris is quite similar.
if the input
.I addrlen
argument for the retrieving call is specified as
.IR "sizeof(struct sockaddr_un)" ,
then the returned address structure
.I won't
have a null terminator in
.IR sun_path .

In addition, some implementations
.\" i.e., traditional BSD
don't require a null terminator when binding a socket (the
.I addrlen
argument is used to determine the length of
.IR sun_path )
and when the socket address is retrieved on these implementations,
there is no null terminator in
.IR sun_path .

Applications that retrieve socket addresses can (portably) code
to handle the possibility that there is no null terminator in
.IR sun_path
by respecting the fact that the number of valid bytes in the pathname is:

    strnlen(addr.sun_path, addrlen \- offsetof(sockaddr_un, sun_path))
.\" The following patch to amend kernel behavior was rejected:
.\" http://thread.gmane.org/gmane.linux.kernel.api/2437
.\" Subject: [patch] Fix handling of overlength pathname in AF_UNIX sun_path
.\" 2012-04-17
.\" And there was a related discussion in the Austin list:
.\" http://thread.gmane.org/gmane.comp.standards.posix.austin.general/5735
.\" Subject: Having a sun_path with no null terminator
.\" 2012-04-18
.\"
.\" FIXME . Track http://austingroupbugs.net/view.php?id=561

Alternatively, an application can retrieve
the socket address by allocating a buffer of size
.I "sizeof(struct sockaddr_un)+1"
that is zeroed out before the retrieval.
The retrieving call can specify
.I addrlen
as
.IR "sizeof(struct sockaddr_un)" ,
and the extra zero byte ensures that there will be
a null terminator for the string returned in
.IR sun_path :

.nf
.in +3
void *addrp;

addrlen = sizeof(struct sockaddr_un);
addrp = malloc(addrlen + 1);
if (addrp == NULL)
    /* Handle error */ ;
memset(addrp, 0, addrlen + 1);

if (getsockname(sfd, (struct sockaddr *) addrp, &addrlen)) == \-1)
    /* handle error */ ;

printf("sun_path = %s\\n", ((struct sockaddr_un *) addrp)\->sun_path);
.in
.fi

This sort of messiness can be avoided if it is guaranteed
that the applications that
.I create
pathname sockets follow the rules outlined above under
.IR "Pathname sockets" .
.SH EXAMPLE
See
.BR bind (2).

For an example of the use of
.BR SCM_RIGHTS
see
.BR cmsg (3).
.SH SEE ALSO
.BR recvmsg (2),
.BR sendmsg (2),
.BR socket (2),
.BR socketpair (2),
.BR cmsg (3),
.BR capabilities (7),
.BR credentials (7),
.BR socket (7),
.BR udp (7)