1 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>,
2 .\" Copyright (C) 2008-2014, Michael Kerrisk <mtk.manpages@gmail.com>,
3 .\" and Copyright (C) 2016, Heinrich Schuchardt <xypron.glpk@gmx.de>
5 .\" %%%LICENSE_START(VERBATIM_ONE_PARA)
6 .\" Permission is granted to distribute possibly modified copies
7 .\" of this page provided the header is included verbatim,
8 .\" and in case of nontrivial modification author and date
9 .\" of the modification is added to the header.
12 .\" Modified, 2003-12-02, Michael Kerrisk, <mtk.manpages@gmail.com>
13 .\" Modified, 2003-09-23, Adam Langley
14 .\" Modified, 2004-05-27, Michael Kerrisk, <mtk.manpages@gmail.com>
15 .\" Added SOCK_SEQPACKET
16 .\" 2008-05-27, mtk, Provide a clear description of the three types of
17 .\" address that can appear in the sockaddr_un structure: pathname,
18 .\" unnamed, and abstract.
20 .TH UNIX 7 2021-03-22 "Linux" "Linux Programmer's Manual"
22 unix \- sockets for local interprocess communication
25 .B #include <sys/socket.h>
26 .B #include <sys/un.h>
28 .IB unix_socket " = socket(AF_UNIX, type, 0);"
29 .IB error " = socketpair(AF_UNIX, type, 0, int *" sv ");"
36 socket family is used to communicate between processes on the same machine
38 Traditionally, UNIX domain sockets can be either unnamed,
39 or bound to a filesystem pathname (marked as being of type socket).
40 Linux also supports an abstract namespace which is independent of the
43 Valid socket types in the UNIX domain are:
45 for a stream-oriented socket;
47 for a datagram-oriented socket that preserves message boundaries
48 (as on most UNIX implementations, UNIX domain datagram
49 sockets are always reliable and don't reorder datagrams);
50 and (since Linux 2.6.4)
52 for a sequenced-packet socket that is connection-oriented,
53 preserves message boundaries,
54 and delivers messages in the order that they were sent.
56 UNIX domain sockets support passing file descriptors or process credentials
57 to other processes using ancillary data.
59 A UNIX domain socket address is represented in the following structure:
63 .\" #define UNIX_PATH_MAX 108
66 sa_family_t sun_family; /* AF_UNIX */
67 char sun_path[108]; /* Pathname */
78 is 108 bytes in size; see also NOTES, below.
80 Various systems calls (for example,
88 Some other system calls (for example,
94 return an argument of this type.
96 Three types of address are distinguished in the
101 a UNIX domain socket can be bound to a null-terminated
102 filesystem pathname using
104 When the address of a pathname socket is returned
105 (by one of the system calls noted above),
108 offsetof(struct sockaddr_un, sun_path) + strlen(sun_path) + 1
112 contains the null-terminated pathname.
115 expression equates to the same value as
116 .IR sizeof(sa_family_t) ,
117 but some other implementations include other fields before
121 expression more portably describes the size of the address structure.)
123 For further details of pathname sockets, see below.
126 A stream socket that has not been bound to a pathname using
129 Likewise, the two sockets created by
132 When the address of an unnamed socket is returned,
134 .IR "sizeof(sa_family_t)" ,
137 should not be inspected.
138 .\" There is quite some variation across implementations: FreeBSD
139 .\" says the length is 16 bytes, HP-UX 11 says it's zero bytes.
142 an abstract socket address is distinguished (from a pathname socket)
145 is a null byte (\(aq\e0\(aq).
146 The socket's address in this namespace is given by the additional
149 that are covered by the specified length of the address structure.
150 (Null bytes in the name have no special significance.)
151 The name has no connection with filesystem pathnames.
152 When the address of an abstract socket is returned,
156 .IR "sizeof(sa_family_t)"
157 (i.e., greater than 2), and the name of the socket is contained in
159 .IR "(addrlen \- sizeof(sa_family_t))"
163 When binding a socket to a pathname, a few rules should be observed
164 for maximum portability and ease of coding:
168 should be null-terminated.
170 The length of the pathname, including the terminating null byte,
171 should not exceed the size of
176 argument that describes the enclosing
178 structure should have a value of at least:
181 offsetof(struct sockaddr_un, sun_path)+strlen(addr.sun_path)+1
187 .IR "sizeof(struct sockaddr_un)" .
189 There is some variation in how implementations handle UNIX domain
190 socket addresses that do not follow the above rules.
191 For example, some (but not all) implementations
192 .\" Linux does this, including for the case where the supplied path
194 append a null terminator if none is present in the supplied
197 When coding portable applications,
198 keep in mind that some implementations
202 as short as 92 bytes.
203 .\" Modern BSDs generally have 104, Tru64 and AIX have 104,
204 .\" Solaris and Irix have 108
211 return socket address structures.
212 When applied to UNIX domain sockets, the value-result
214 argument supplied to the call should be initialized as above.
215 Upon return, the argument is set to indicate the
217 size of the address structure.
218 The caller should check the value returned in this argument:
219 if the output value exceeds the input value,
220 then there is no guarantee that a null terminator is present in
224 .SS Pathname socket ownership and permissions
225 In the Linux implementation,
226 pathname sockets honor the permissions of the directory they are in.
227 Creation of a new socket fails if the process does not have write and
228 search (execute) permission on the directory in which the socket is created.
231 connecting to a stream socket object requires write permission on that socket;
232 sending a datagram to a datagram socket likewise
233 requires write permission on that socket.
234 POSIX does not make any statement about the effect of the permissions
235 on a socket file, and on some systems (e.g., older BSDs),
236 the socket permissions are ignored.
237 Portable programs should not rely on
238 this feature for security.
240 When creating a new socket, the owner and group of the socket file
241 are set according to the usual rules.
242 The socket file has all permissions enabled,
243 other than those that are turned off by the process
246 The owner, group, and permissions of a pathname socket can be changed (using
250 .\" However, fchown() and fchmod() do not seem to have an effect
253 Socket permissions have no meaning for abstract sockets:
256 has no effect when binding an abstract socket,
257 and changing the ownership and permissions of the object (via
261 has no effect on the accessibility of the socket.
263 Abstract sockets automatically disappear when all open references
264 to the socket are closed.
266 The abstract socket namespace is a nonportable Linux extension.
269 For historical reasons, these socket options are specified with a
271 type even though they are
280 as the socket family.
283 Enabling this socket option causes receipt of the credentials of
284 the sending process in an
285 .B SCM_CREDENTIALS ancillary
286 message in each subsequently received message.
287 The returned credentials are those specified by the sender using
288 .BR SCM_CREDENTIALS ,
289 or a default that includes the sender's PID, real user ID, and real group ID,
290 if the sender did not specify
294 When this option is set and the socket is not yet connected,
295 a unique name in the abstract namespace will be generated automatically.
297 The value given as an argument to
299 and returned as the result of
301 is an integer boolean flag.
304 Enables receiving of the SELinux security label of the peer socket
305 in an ancillary message of type
309 The value given as an argument to
311 and returned as the result of
313 is an integer boolean flag.
317 option is supported for UNIX domain datagram sockets
318 .\" commit 877ce7c1b3afd69a9b1caeb1b9964c992641f52a
320 support for UNIX domain stream sockets was added
321 .\" commit 37a9a8df8ce9de6ea73349c9ac8bdf6ba4ec4f70
329 This read-only socket option returns the
330 credentials of the peer process connected to this socket.
331 The returned credentials are those that were in effect at the time
341 structure; define the
343 feature test macro to obtain the definition of that structure from
346 The use of this option is possible only for connected
348 stream sockets and for
350 stream and datagram socket pairs created using
354 This read-only socket option returns the
355 security context of the peer socket connected to this socket.
356 By default, this will be the same as the security context of
357 the process that created the peer socket unless overridden
358 by the policy or by a process with the required permissions.
362 is a pointer to a buffer of the specified length in bytes
363 into which the security context string will be copied.
364 If the buffer length is less than the length of the security
371 and returns the required length via
373 The caller should allocate at least
375 bytes for the buffer initially, although this is not guaranteed
377 Resizing the buffer to the returned length
378 and retrying may be necessary.
380 The security context string may include a terminating null character
381 in the returned length, but is not guaranteed to do so: a security
382 context "foo" might be represented as either {'f','o','o'} of length 3
383 or {'f','o','o','\\0'} of length 4, which are considered to be
385 The string is printable, does not contain non-terminating null characters,
386 and is in an unspecified encoding (in particular, it
387 is not guaranteed to be ASCII or UTF-8).
389 The use of this option for sockets in the
391 address family is supported since Linux 2.6.2 for connected stream sockets,
393 .\" commit 0b811db2cb2aabc910e53d34ebb95a15997c33e7
394 also for stream and datagram socket pairs created using
403 .IR sizeof(sa_family_t) ,
404 .\" i.e., sizeof(short)
407 socket option was specified for a socket that was
408 not explicitly bound to an address,
409 then the socket is autobound to an abstract address.
410 The address consists of a null byte
411 followed by 5 bytes in the character set
413 Thus, there is a limit of 2^20 autobind addresses.
414 (From Linux 2.1.15, when the autobind feature was added,
415 8 bytes were used, and the limit was thus 2^32 autobind addresses.
416 The change to 5 bytes came in Linux 2.3.15.)
418 The following paragraphs describe domain-specific details and
419 unsupported features of the sockets API for UNIX domain sockets on Linux.
421 UNIX domain sockets do not support the transmission of
422 out-of-band data (the
432 flag is not supported by UNIX domain sockets.
435 .\" commit 9f6f9af7694ede6314bed281eec74d588ba9474f
442 was not supported by UNIX domain sockets.
446 socket option does have an effect for UNIX domain sockets, but the
449 For datagram sockets, the
451 value imposes an upper limit on the size of outgoing datagrams.
452 This limit is calculated as the doubled (see
454 option value less 32 bytes used for overhead.
455 .SS Ancillary messages
456 Ancillary data is sent and received using
460 For historical reasons, the ancillary message types listed below
463 type even though they are
466 To send them, set the
475 For more information, see
479 Send or receive a set of open file descriptors from another process.
480 The data portion contains an integer array of the file descriptors.
482 Commonly, this operation is referred to as "passing a file descriptor"
484 However, more accurately,
485 what is being passed is a reference to an open file description (see
487 and in the receiving process it is likely that a different
488 file descriptor number will be used.
489 Semantically, this operation is equivalent to duplicating
491 a file descriptor into the file descriptor table of another process.
493 If the buffer used to receive the ancillary data containing
494 file descriptors is too small (or is absent),
495 then the ancillary data is truncated (or discarded)
496 and the excess file descriptors are automatically closed
497 in the receiving process.
499 If the number of file descriptors received in the ancillary data would
500 cause the process to exceed its
504 the excess file descriptors are automatically closed
505 in the receiving process.
509 defines a limit on the number of file descriptors in the array.
510 Attempting to send an array larger than this limit causes
512 to fail with the error
517 .\" commit bba14de98753cb6599a2dae0e520714b2153522d
521 Send or receive UNIX credentials.
522 This can be used for authentication.
523 The credentials are passed as a
526 This structure is defined in
533 pid_t pid; /* Process ID of the sending process */
534 uid_t uid; /* User ID of the sending process */
535 gid_t gid; /* Group ID of the sending process */
542 feature test macro must be defined (before including
544 header files) in order to obtain the definition
547 The credentials which the sender specifies are checked by the kernel.
548 A privileged process is allowed to specify values that do not match its own.
549 The sender must specify its own process ID (unless it has the capability
551 in which case the PID of any existing process may be specified),
552 its real user ID, effective user ID, or saved set-user-ID (unless it has
554 and its real group ID, effective group ID, or saved set-group-ID
562 option must be enabled on the socket.
565 Receive the SELinux security context (the security label)
567 The received ancillary data is a null-terminated string containing
568 the security context.
569 The receiver should allocate at least
571 bytes in the data portion of the ancillary message for this data.
573 To receive the security context, the
575 option must be enabled on the socket (see above).
577 When sending ancillary data with
579 only one item of each of the above types may be included in the sent message.
581 At least one byte of real data should be sent when sending ancillary data.
582 On Linux, this is required to successfully send ancillary data over
583 a UNIX domain stream socket.
584 When sending ancillary data over a UNIX domain datagram socket,
585 it is not necessary on Linux to send any accompanying real data.
586 However, portable applications should also include at least one byte
587 of real data when sending ancillary data over a datagram socket.
589 When receiving from a stream socket,
590 ancillary data forms a kind of barrier for the received data.
591 For example, suppose that the sender transmits as follows:
597 of four bytes, with no ancillary data.
600 of one byte, with ancillary data.
603 of four bytes, with no ancillary data.
607 Suppose that the receiver now performs
609 calls each with a buffer size of 20 bytes.
610 The first call will receive five bytes of data,
611 along with the ancillary data sent by the second
614 The next call will receive the remaining four bytes of data.
616 If the space allocated for receiving incoming ancillary data is too small
617 then the ancillary data is truncated to the number of headers
618 that will fit in the supplied buffer (or, in the case of an
620 file descriptor list, the list of file descriptors may be truncated).
621 If no buffer is provided for incoming ancillary data (i.e., the
625 structure supplied to
628 then the incoming ancillary data is discarded.
629 In both of these cases, the
631 flag will be set in the
639 calls return information in
641 The correct syntax is:
646 .IB error " = ioctl(" unix_socket ", " ioctl_type ", &" value ");"
656 sockets, this call returns the number of unread bytes in the receive buffer.
657 The socket must not be in LISTEN state, otherwise an error
662 .IR <linux/sockios.h> .
663 .\" FIXME . http://sources.redhat.com/bugzilla/show_bug.cgi?id=12002,
664 .\" filed 2010-09-10, may cause SIOCINQ to be defined in glibc headers
666 you can use the synonymous
670 .\" SIOCOUTQ also has an effect for UNIX domain sockets, but not
671 .\" quite what userland might expect. It seems to return the number
672 .\" of bytes allocated for buffers containing pending output.
673 .\" That number is normally larger than the number of bytes of pending
674 .\" output. Since this info is, from userland's point of view, imprecise,
675 .\" and it may well change, probably best not to document this now.
679 the returned value is the same as
680 for Internet domain datagram sockets;
686 The specified local address is already in use or the filesystem socket
687 object already exists.
690 This error can occur for
692 when sending a file descriptor as ancillary data over
693 a UNIX domain socket (see the description of
695 above), and indicates that the file descriptor number that
696 is being sent is not valid (e.g., it is not an open file descriptor).
699 The remote address specified by
701 was not a listening socket.
702 This error can also occur if the target pathname is not a socket.
705 Remote socket was unexpectedly closed.
708 User memory address was not valid.
711 Invalid argument passed.
712 A common cause is that the value
714 was not specified in the
716 field of passed addresses, or the socket was in an
717 invalid state for the applied operation.
721 called on an already connected socket or a target address was
722 specified on a connected socket.
725 The pathname in the remote address specified to
733 Socket operation needs a target address, but the socket is not connected.
736 Stream operation called on non-stream oriented socket or tried to
737 use the out-of-band data option.
740 The sender passed invalid credentials in the
744 Remote socket was closed on a stream socket.
748 This can be avoided by passing the
756 Passed protocol is not
760 Remote socket does not match the local socket type
769 While sending an ancillary message containing credentials
770 .RB ( SCM_CREDENTIALS ),
771 the caller specified a PID that does not match any existing process.
774 This error can occur for
776 when sending a file descriptor as ancillary data over
777 a UNIX domain socket (see the description of
780 It occurs if the number of "in-flight" file descriptors exceeds the
782 resource limit and the caller does not have the
785 An in-flight file descriptor is one that has been sent using
787 but has not yet been accepted in the recipient process using
790 This error is diagnosed since mainline Linux 4.5
791 (and in some earlier kernel versions where the fix has been backported).
792 .\" commit 712f4aad406bb1ed67f3f98d04c044191f0ff593
793 In earlier kernel versions,
794 it was possible to place an unlimited number of file descriptors in flight,
795 by sending each file descriptor with
797 and then closing the file descriptor so that it was not accounted against the
801 Other errors can be generated by the generic socket layer or
802 by the filesystem while generating a filesystem socket object.
803 See the appropriate manual pages for more information.
806 and the abstract namespace were introduced with Linux 2.2 and should not
807 be used in portable programs.
808 (Some BSD-derived systems also support credential passing,
809 but the implementation details differ.)
811 Binding to a socket with a filename creates a socket
812 in the filesystem that must be deleted by the caller when it is no
815 The usual UNIX close-behind semantics apply; the socket can be unlinked
816 at any time and will be finally removed from the filesystem when the last
817 reference to it is closed.
819 To pass file descriptors or credentials over a
822 to send or receive at least one byte of nonancillary data in the same
828 UNIX domain stream sockets do not support the notion of out-of-band data.
831 When binding a socket to an address,
832 Linux is one of the implementations that appends a null terminator
833 if none is supplied in
835 In most cases this is unproblematic:
836 when the socket address is retrieved,
837 it will be one byte longer than that supplied when the socket was bound.
838 However, there is one case where confusing behavior can result:
839 if 108 non-null bytes are supplied when a socket is bound,
840 then the addition of the null terminator takes the length of
842 .IR sizeof(sun_path) .
843 Consequently, when retrieving the socket address
846 .\" The behavior on Solaris is quite similar.
849 argument for the retrieving call is specified as
850 .IR "sizeof(struct sockaddr_un)" ,
851 then the returned address structure
853 have a null terminator in
856 In addition, some implementations
857 .\" i.e., traditional BSD
858 don't require a null terminator when binding a socket (the
860 argument is used to determine the length of
862 and when the socket address is retrieved on these implementations,
863 there is no null terminator in
866 Applications that retrieve socket addresses can (portably) code
867 to handle the possibility that there is no null terminator in
869 by respecting the fact that the number of valid bytes in the pathname is:
871 strnlen(addr.sun_path, addrlen \- offsetof(sockaddr_un, sun_path))
872 .\" The following patch to amend kernel behavior was rejected:
873 .\" http://thread.gmane.org/gmane.linux.kernel.api/2437
874 .\" Subject: [patch] Fix handling of overlength pathname in AF_UNIX sun_path
876 .\" And there was a related discussion in the Austin list:
877 .\" http://thread.gmane.org/gmane.comp.standards.posix.austin.general/5735
878 .\" Subject: Having a sun_path with no null terminator
881 .\" FIXME . Track http://austingroupbugs.net/view.php?id=561
883 Alternatively, an application can retrieve
884 the socket address by allocating a buffer of size
885 .I "sizeof(struct sockaddr_un)+1"
886 that is zeroed out before the retrieval.
887 The retrieving call can specify
890 .IR "sizeof(struct sockaddr_un)" ,
891 and the extra zero byte ensures that there will be
892 a null terminator for the string returned in
899 addrlen = sizeof(struct sockaddr_un);
900 addrp = malloc(addrlen + 1);
903 memset(addrp, 0, addrlen + 1);
905 if (getsockname(sfd, (struct sockaddr *) addrp, &addrlen)) == \-1)
908 printf("sun_path = %s\en", ((struct sockaddr_un *) addrp)\->sun_path);
912 This sort of messiness can be avoided if it is guaranteed
913 that the applications that
915 pathname sockets follow the rules outlined above under
916 .IR "Pathname sockets" .
918 The following code demonstrates the use of sequenced-packet
919 sockets for local interprocess communication.
920 It consists of two programs.
921 The server program waits for a connection from the client program.
922 The client sends each of its command-line arguments in separate messages.
923 The server treats the incoming messages as integers and adds them up.
924 The client sends the command string "END".
925 The server sends back a message containing the sum of the client's integers.
926 The client prints the sum and exits.
927 The server waits for the next client to connect.
928 To stop the server, the client is called with the command-line argument "DOWN".
930 The following output was recorded while running the server in the background
931 and repeatedly executing the client.
932 Execution of the server program ends when it receives the "DOWN" command.
940 $ \fB./client 11 \-5\fP
942 $ \fB./client DOWN\fP
955 #define SOCKET_NAME "/tmp/9Lq7BNBnBycd6nxy.socket"
956 #define BUFFER_SIZE 12
965 #include <sys/socket.h>
968 #include "connection.h"
971 main(int argc, char *argv[])
973 struct sockaddr_un name;
976 int connection_socket;
979 char buffer[BUFFER_SIZE];
981 /* Create local socket. */
983 connection_socket = socket(AF_UNIX, SOCK_SEQPACKET, 0);
984 if (connection_socket == \-1) {
990 * For portability clear the whole structure, since some
991 * implementations have additional (nonstandard) fields in
995 memset(&name, 0, sizeof(name));
997 /* Bind socket to socket name. */
999 name.sun_family = AF_UNIX;
1000 strncpy(name.sun_path, SOCKET_NAME, sizeof(name.sun_path) \- 1);
1002 ret = bind(connection_socket, (const struct sockaddr *) &name,
1010 * Prepare for accepting connections. The backlog size is set
1011 * to 20. So while one request is being processed other requests
1015 ret = listen(connection_socket, 20);
1021 /* This is the main loop for handling connections. */
1025 /* Wait for incoming connection. */
1027 data_socket = accept(connection_socket, NULL, NULL);
1028 if (data_socket == \-1) {
1036 /* Wait for next data packet. */
1038 ret = read(data_socket, buffer, sizeof(buffer));
1044 /* Ensure buffer is 0\-terminated. */
1046 buffer[sizeof(buffer) \- 1] = 0;
1048 /* Handle commands. */
1050 if (!strncmp(buffer, "DOWN", sizeof(buffer))) {
1055 if (!strncmp(buffer, "END", sizeof(buffer))) {
1059 /* Add received summand. */
1061 result += atoi(buffer);
1066 sprintf(buffer, "%d", result);
1067 ret = write(data_socket, buffer, sizeof(buffer));
1077 /* Quit on DOWN command. */
1084 close(connection_socket);
1086 /* Unlink the socket. */
1088 unlink(SOCKET_NAME);
1101 #include <sys/socket.h>
1104 #include "connection.h"
1107 main(int argc, char *argv[])
1109 struct sockaddr_un addr;
1112 char buffer[BUFFER_SIZE];
1114 /* Create local socket. */
1116 data_socket = socket(AF_UNIX, SOCK_SEQPACKET, 0);
1117 if (data_socket == \-1) {
1123 * For portability clear the whole structure, since some
1124 * implementations have additional (nonstandard) fields in
1128 memset(&addr, 0, sizeof(addr));
1130 /* Connect socket to socket address. */
1132 addr.sun_family = AF_UNIX;
1133 strncpy(addr.sun_path, SOCKET_NAME, sizeof(addr.sun_path) \- 1);
1135 ret = connect(data_socket, (const struct sockaddr *) &addr,
1138 fprintf(stderr, "The server is down.\en");
1142 /* Send arguments. */
1144 for (int i = 1; i < argc; ++i) {
1145 ret = write(data_socket, argv[i], strlen(argv[i]) + 1);
1152 /* Request result. */
1154 strcpy(buffer, "END");
1155 ret = write(data_socket, buffer, strlen(buffer) + 1);
1161 /* Receive result. */
1163 ret = read(data_socket, buffer, sizeof(buffer));
1169 /* Ensure buffer is 0\-terminated. */
1171 buffer[sizeof(buffer) \- 1] = 0;
1173 printf("Result = %s\en", buffer);
1183 For examples of the use of
1188 .BR seccomp_unotify (2).
1195 .BR capabilities (7),
1196 .BR credentials (7),