--- name: "unix" section: "4" platform: "freebsd" package: "base" version: "16.0-CURRENT-20261027-61c70dc16a79" release: "FreeBSD 16.0-CURRENT (2026-10-27)" platforms: ["linux", "freebsd", "openbsd", "macos"] url: "https://new.carta.tech/unix/4" --- # unix(4) UNIX-domain protocol family ## NAME **unix** – UNIX-domain protocol family ## SYNOPSIS **#include \** **#include \** ## DESCRIPTION The UNIX-domain protocol family is a collection of protocols that provides local (on-machine) interprocess communication through the normal [socket(2)](/socket/2.md) mechanisms. The UNIX-domain family supports the SOCK\_STREAM, SOCK\_SEQPACKET, and SOCK\_DGRAM socket types and uses file system pathnames for addressing. ## ADDRESSING UNIX-domain addresses are variable-length file system pathnames of at most 104 characters. The include file ⟨sys/un.h⟩ defines this address: ``` struct sockaddr_un { u_char sun_len; u_char sun_family; char sun_path[104]; }; ``` Binding a name to a UNIX-domain socket with [bind(2)](/bind/2.md) causes a socket file to be created in the file system. This file is *not* removed when the socket is closed — [unlink(2)](/unlink/2.md) must be used to remove the file. Prior to binding a socket, [fchmod(2)](/fchmod/2.md) can be used to set the permissions of the socket file. This avoids the race that would otherwise occur between creation of the file and a subsequent call to [chmod(2)](/chmod/2.md). Once the socket is bound to a file name, the permissions of the file can not be changed this way. Unlike with most other protocols, binding is not a prerequisite for [listen(2)](/listen/2.md). An unbound socket may listen, in which case it can only be reached by descriptor, as described in Naming a peer below. A socket may equally be bound after it listens, which lets a listener be published only once it is ready to accept, rather than leaving a window in which the socket file exists but connections to it are refused. The length of UNIX-domain address, required by [bind(2)](/bind/2.md) and [connect(2)](/connect/2.md), can be calculated by the macro **SUN\_LEN**() defined in ⟨sys/un.h⟩. The *sun\_path* field must be terminated by a NUL character to be used with **SUN\_LEN**(), but the terminating NUL is *not* part of the address. The UNIX-domain protocol family does not support broadcast addressing or any form of “wildcard” matching on incoming messages. All addresses are absolute- or relative-pathnames of other UNIX-domain sockets. Normal file system access-control mechanisms are also applied when referencing pathnames; e.g., the destination of a [connect(2)](/connect/2.md) or [sendto(2)](/sendto/2.md) must be writable. ### Naming a peer A peer can be named either by a pathname or, using [connectat(2)](/connectat/2.md), directly by a descriptor. Across these two forms a peer can be named five ways, all of which reach the same socket: | | | | | --- | --- | --- | | *Target* | *By descriptor* | *By pathname* | | The peer socket | the socket's fd | (none) | | A bound socket's file | an O\_PATH descriptor | an ordinary pathname | | A /dev/fd/N node | an O\_PATH descriptor | */dev/fd/N* | A pathname, accepted by both [connect(2)](/connect/2.md) and [connectat(2)](/connectat/2.md), names either a bound socket's file created by [bind(2)](/bind/2.md), or the */dev/fd/N* node that fdescfs(5) provides for descriptor *N*. In the latter case the */dev/fd/N* node must name the socket descriptor itself; the named descriptor is resolved a single level and not chased further. With a standard fdescfs(5) mount this means a node naming a traditional socket file, or another */dev/fd/N* node, does not resolve to a peer; a **nodup** mount, however, dereferences such a descriptor before the lookup completes. [connectat(2)](/connectat/2.md) also supports both of the above cases. The path is allowed to be empty — that is, a *namelen* equal to offsetof(struct sockaddr\_un, sun\_path) — in which case the given file descriptor will be used directly, like with most **\*at** system calls. (Note: */dev/fd/N* nodes are not recommended to be used for the [connectat(2)](/connectat/2.md), since the descriptor can just be named directly, and this is both more efficient and less userland code, but since [connect(2)](/connect/2.md) is implemented in terms of [connectat(2)](/connectat/2.md) this is provided for free.) Besides supporting cases that [connect(2)](/connect/2.md) also supports, [connectat(2)](/connectat/2.md) additionally supports directly naming the peer by the *fd* descriptor (and an empty path). The descriptor may be the peer socket itself. In all cases, the descriptor passed must carry the CAP\_CONNECTAT capability right, and in the */dev/fd/N* form the target descriptor must *also* carry that capability. A descriptor limited to CAP\_CONNECTAT alone is thus a pure “connect-to-me” token, usable as a connection target but not listened on, accepted from, or read. For the descriptor forms, possession of the descriptor is the authorization, as with descriptor passing over SCM\_RIGHTS; the file system access-control and [mac(4)](/mac/4.md) checks that apply to the pathname forms are not repeated. Because a descriptor alone suffices, a peer need never be bound at all: a datagram socket may connect to an unbound peer, and a listening socket may itself be unbound, neither of which a pathname could name. ## CONTROL MESSAGES The UNIX-domain sockets support the communication of UNIX file descriptors and process credentials through the use of the *msg\_control* field in the *msg* argument to [sendmsg(2)](/sendmsg/2.md) and [recvmsg(2)](/recvmsg/2.md). The items to be passed are described using a *struct cmsghdr* that is defined in the include file ⟨sys/socket.h⟩. To send file descriptors, the type of the message is SCM\_RIGHTS, and the data portion of the messages is an array of integers representing the file descriptors to be passed. The number of descriptors being passed is defined by the length field of the message; the length field is the sum of the size of the header plus the size of the array of file descriptors. The received descriptor is a *duplicate* of the sender's descriptor, as if it were created via dup(fd) or fcntl(fd, F\_DUPFD\_CLOEXEC, 0) depending on whether MSG\_CMSG\_CLOEXEC is passed in the [recvmsg(2)](/recvmsg/2.md) call. Descriptors that are awaiting delivery, or that are purposely not received, are automatically closed by the system when the destination socket is closed. The receiving socket can reject SCM\_RIGHTS at the [sendmsg(2)](/sendmsg/2.md) call with the socket option SO\_PASSRIGHTS, which can be set with [setsockopt(2)](/setsockopt/2.md) and tested with [getsockopt(2)](/getsockopt/2.md). Credentials of the sending process can be transmitted explicitly using a control message of type SCM\_CREDS with a data portion of type *struct cmsgcred*, defined in ⟨sys/socket.h⟩ as follows: ``` struct cmsgcred { pid_t cmcred_pid; /* PID of sending process */ uid_t cmcred_uid; /* real UID of sending process */ uid_t cmcred_euid; /* effective UID of sending process */ gid_t cmcred_gid; /* real GID of sending process */ short cmcred_ngroups; /* number of groups */ gid_t cmcred_groups[CMGROUP_MAX]; /* groups */ }; ``` The sender should pass a zeroed buffer which will be filled in by the system. The group list is truncated to at most CMGROUP\_MAX GIDs. The process ID *cmcred\_pid* should not be looked up (such as via the KERN\_PROC\_PID sysctl) for making security decisions. The sending process could have exited and its process ID already been reused for a new process. ## SOCKET OPTIONS UNIX domain sockets support a number of socket options for the options level SOL\_LOCAL, which can be set with [setsockopt(2)](/setsockopt/2.md) and tested with [getsockopt(2)](/getsockopt/2.md): - LOCAL\_CREDS This option may be enabled on SOCK\_DGRAM, SOCK\_SEQPACKET, or a SOCK\_STREAM socket. This option provides a mechanism for the receiver to receive the credentials of the process calling [write(2)](/write/2.md), [send(2)](/send/2.md), [sendto(2)](/sendto/2.md) or [sendmsg(2)](/sendmsg/2.md) as a [recvmsg(2)](/recvmsg/2.md) control message. The *msg\_control* field in the *msghdr* structure points to a buffer that contains a *cmsghdr* structure followed by a variable length *sockcred* structure, defined in ⟨sys/socket.h⟩ as follows: ``` struct sockcred { uid_t sc_uid; /* real user id */ uid_t sc_euid; /* effective user id */ gid_t sc_gid; /* real group id */ gid_t sc_egid; /* effective group id */ int sc_ngroups; /* number of supplemental groups */ gid_t sc_groups[1]; /* variable length */ }; ``` The current implementation truncates the group list to at most CMGROUP\_MAX groups. The **SOCKCREDSIZE**() macro computes the size of the *sockcred* structure for a specified number of groups. The *cmsghdr* fields have the following values: ``` cmsg_len = CMSG_LEN(SOCKCREDSIZE(ngroups)) cmsg_level = SOL_SOCKET cmsg_type = SCM_CREDS ``` On SOCK\_STREAM and SOCK\_SEQPACKET sockets credentials are passed only on the first read from a socket, then the system clears the option on the socket. This option and the above explicit *struct cmsgcred* both use the same value SCM\_CREDS but incompatible control messages. If this option is enabled and the sender attached a SCM\_CREDS control message with a *struct cmsgcred*, it will be discarded and a *struct sockcred* will be included. Many setuid programs will [write(2)](/write/2.md) data at least partially controlled by the invoker, such as error messages. Therefore, a message accompanied by a particular *sc\_euid* value should not be trusted as being from that user. - LOCAL\_CREDS\_PERSISTENT This option is similar to LOCAL\_CREDS, except that socket credentials are passed on every read from a SOCK\_STREAM or SOCK\_SEQPACKET socket, instead of just the first read. Additionally, the *msg\_control* field in the *msghdr* structure points to a buffer that contains a *cmsghdr* structure followed by a variable length *sockcred2* structure, defined in ⟨sys/socket.h⟩ as follows: ``` struct sockcred2 { int sc_version; /* version of this structure */ pid_t sc_pid; /* PID of sending process */ uid_t sc_uid; /* real user id */ uid_t sc_euid; /* effective user id */ gid_t sc_gid; /* real group id */ gid_t sc_egid; /* effective group id */ int sc_ngroups; /* number of supplemental groups */ gid_t sc_groups[1]; /* variable length */ }; ``` The current version is zero. The *cmsghdr* fields have the following values: ``` cmsg_len = CMSG_LEN(SOCKCRED2SIZE(ngroups)) cmsg_level = SOL_SOCKET cmsg_type = SCM_CREDS2 ``` The LOCAL\_CREDS and LOCAL\_CREDS\_PERSISTENT options are mutually exclusive. - LOCAL\_PEERCRED Requested via [getsockopt(2)](/getsockopt/2.md) on a SOCK\_STREAM or SOCK\_SEQPACKET socket returns credentials of the remote side. These will arrive in the form of a filled in *xucred* structure, defined in ⟨sys/ucred.h⟩ as follows: ``` struct xucred { u_int cr_version; /* structure layout version */ uid_t cr_uid; /* effective user id */ short cr_ngroups; /* number of groups */ gid_t cr_groups[XU_NGROUPS]; /* groups */ pid_t cr_pid; /* process id of the sending process */ }; ``` The *cr\_version* fields should be checked against XUCRED\_VERSION define. The credentials presented to the server (the [listen(2)](/listen/2.md) caller) are those of the client when it called [connect(2)](/connect/2.md); the credentials presented to the client (the [connect(2)](/connect/2.md) caller) are those of the server when it called [listen(2)](/listen/2.md). This mechanism is reliable; there is no way for either party to influence the credentials presented to its peer except by calling the appropriate system call (e.g., [connect(2)](/connect/2.md) or [listen(2)](/listen/2.md)) under different effective credentials. To reliably obtain peer credentials on a SOCK\_DGRAM socket refer to the LOCAL\_CREDS socket option. ## BUFFERING Due to the local nature of the UNIX-domain sockets, they do not implement send buffers. The [send(2)](/send/2.md) and [write(2)](/write/2.md) families of system calls attempt to write data to the receive buffer of the destination socket. The default buffer sizes for SOCK\_STREAM and SOCK\_SEQPACKET UNIX-domain sockets can be configured with *net.local.stream* and *net.local.seqpacket* branches of [sysctl(3)](/sysctl/3.md) MIB respectively. Note that setting the send buffer size (sendspace) affects only the maximum write size. The UNIX-domain sockets of type SOCK\_DGRAM are unreliable and always non-blocking for write operations. The default receive buffer can be configured with *net.local.dgram.recvspace*. The maximum allowed datagram size is limited by *net.local.dgram.maxdgram*. A SOCK\_DGRAM socket that has been bound with [bind(2)](/bind/2.md) can have multiple peers connected at the same time. The modern FreeBSD implementation will allocate *net.local.dgram.recvspace* sized private buffers in the receive buffer of the bound socket for every connected socket, preventing a situation when a single writer can exhaust all of buffer space. Messages coming from unconnected sends using [sendto(2)](/sendto/2.md) land on the shared buffer of the receiving socket, which has the same size limit. A side effect of the implementation is that it doesn't guarantee that writes from different senders will arrive at the receiver in the same chronological order they were sent. The order is preserved for writes coming through a particular connection. ## SEE ALSO [connect(2)](/connect/2.md), [connectat(2)](/connectat/2.md), [dup(2)](/dup/2.md), [fchmod(2)](/fchmod/2.md), [fcntl(2)](/fcntl/2.md), [getsockopt(2)](/getsockopt/2.md), [listen(2)](/listen/2.md), [recvmsg(2)](/recvmsg/2.md), [sendto(2)](/sendto/2.md), [setsockopt(2)](/setsockopt/2.md), [socket(2)](/socket/2.md), [CMSG\_DATA(3)](/CMSG_DATA/3.md), [intro(4)](/intro/4.md), [mac(4)](/mac/4.md), fdescfs(5), [sysctl(8)](/sysctl/8.md) “An Introductory 4.3 BSD Interprocess Communication Tutorial”, *PS1*, 7. “An Advanced 4.3 BSD Interprocess Communication Tutorial”, *PS1*, 8.