1 .\" This man page is Copyright (C) 1999 Andi Kleen <ak@muc.de>.
2 .\" and Copyright (C) 2008-2014, Michael Kerrisk <mtk.manpages@gmail.com>
4 .\" %%%LICENSE_START(VERBATIM_ONE_PARA)
5 .\" and Copyright (C) 2008, 2012 Michael Kerrisk <mtk.manpages@gmail.com>
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 2014-12-31 "Linux" "Linux Programmer's Manual"
22 unix \- sockets for local interprocess communication
24 .B #include <sys/socket.h>
26 .B #include <sys/un.h>
28 .IB unix_socket " = socket(AF_UNIX, type, 0);"
30 .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 connection-oriented socket that preserves message boundaries
53 and delivers messages in the order that they were sent.
55 UNIX domain sockets support passing file descriptors or process credentials
56 to other processes using ancillary data.
58 A UNIX domain socket address is represented in the following structure:
62 #define UNIX_PATH_MAX 108
65 sa_family_t sun_family; /* AF_UNIX */
66 char sun_path[UNIX_PATH_MAX]; /* pathname */
76 Various systems calls (for example,
84 Some other system calls (for example,
90 return an argument of this type.
92 Three types of address are distinguished in the
97 a UNIX domain socket can be bound to a null-terminated
98 filesystem pathname using
100 When the address of a pathname socket is returned
101 (by one of the system calls noted above),
104 offsetof(struct sockaddr_un, sun_path) + strlen(sun_path) + 1
108 contains the null-terminated pathname.
111 expression equates to the same value as
112 .IR sizeof(sa_family_t) ,
113 but some other implementations include other fields before
117 expression more portably describes the size of the address structure.)
119 For further details of pathname sockets, see below.
122 A stream socket that has not been bound to a pathname using
125 Likewise, the two sockets created by
128 When the address of an unnamed socket is returned,
130 .IR "sizeof(sa_family_t)" ,
133 should not be inspected.
134 .\" There is quite some variation across implementations: FreeBSD
135 .\" says the length is 16 bytes, HP-UX 11 says it's zero bytes.
138 an abstract socket address is distinguished (from a pathname socket)
141 is a null byte (\(aq\\0\(aq).
142 The socket's address in this namespace is given by the additional
145 that are covered by the specified length of the address structure.
146 (Null bytes in the name have no special significance.)
147 The name has no connection with filesystem pathnames.
148 When the address of an abstract socket is returned,
152 .IR "sizeof(sa_family_t)"
153 (i.e., greater than 2), and the name of the socket is contained in
155 .IR "(addrlen \- sizeof(sa_family_t))"
158 The abstract socket namespace is a nonportable Linux extension.
160 When binding a socket to a pathname, a few rules should be observed
161 for maximum portability and ease of coding:
165 should be null-terminated.
167 The length of the pathname, including the terminating null byte,
168 should not exceed the size of
173 argument that describes the enclosing
175 structure should have a value of at least:
178 offsetof(struct sockaddr_un, sun_path)+strlen(addr.sun_path)+1
184 .IR "sizeof(struct sockaddr_un)" .
186 There is some variation in how implementations handle UNIX domain
187 socket addresses that do not follow the above rules.
188 For example, some (but not all) implementations
189 .\" Linux does this, including for the case where the supplied path
191 append a null terminator if none is present in the supplied
194 When coding portable applications,
195 keep in mind that some implementations
199 as short as 92 bytes.
200 .\" Modern BSDs generally have 104, Tru64 and AIX have 104,
201 .\" Solaris and Irix have 108
208 return socket address structures.
209 When applied to UNIX domain sockets, the value-result
211 argument supplied to the call should be initialized as above.
212 Upon return, the argument is set to indicate the
214 size of the address structure.
215 The caller should check the value returned in this argument:
216 if the output value exceeds the input value,
217 then there is no guarantee that a null terminator is present in
221 For historical reasons, these socket options are specified with a
223 type even though they are
232 as the socket family.
235 Enables the receiving of the credentials of the sending process in an
237 When this option is set and the socket is not yet connected
238 a unique name in the abstract namespace will be generated automatically.
239 Expects an integer boolean flag.
246 .IR sizeof(sa_family_t) ,
247 .\" i.e., sizeof(short)
250 socket option was specified for a socket that was
251 not explicitly bound to an address,
252 then the socket is autobound to an abstract address.
253 The address consists of a null byte
254 followed by 5 bytes in the character set
256 Thus, there is a limit of 2^20 autobind addresses.
257 (From Linux 2.1.15, when the autobind feature was added,
258 8 bytes were used, and the limit was thus 2^32 autobind addresses.
259 The change to 5 bytes came in Linux 2.3.15.)
261 The following paragraphs describe domain-specific details and
262 unsupported features of the sockets API for UNIX domain sockets on Linux.
264 UNIX domain sockets do not support the transmission of
265 out-of-band data (the
275 flag is not supported by UNIX domain sockets.
283 is not supported by UNIX domain sockets.
287 socket option does have an effect for UNIX domain sockets, but the
290 For datagram sockets, the
292 value imposes an upper limit on the size of outgoing datagrams.
293 This limit is calculated as the doubled (see
295 option value less 32 bytes used for overhead.
296 .SS Ancillary messages
297 Ancillary data is sent and received using
301 For historical reasons the ancillary message types listed below
304 type even though they are
316 For more information see
320 Send or receive a set of open file descriptors from another process.
321 The data portion contains an integer array of the file descriptors.
322 The passed file descriptors behave as though they have been created with
326 Send or receive UNIX credentials.
327 This can be used for authentication.
328 The credentials are passed as a
331 Thus structure is defined in
338 pid_t pid; /* process ID of the sending process */
339 uid_t uid; /* user ID of the sending process */
340 gid_t gid; /* group ID of the sending process */
347 feature test macro must be defined (before including
349 header files) in order to obtain the definition
352 The credentials which the sender specifies are checked by the kernel.
353 A process with effective user ID 0 is allowed to specify values that do
355 The sender must specify its own process ID (unless it has the capability
357 its user ID, effective user ID, or saved set-user-ID (unless it has
359 and its group ID, effective group ID, or saved set-group-ID
366 option must be enabled on the socket.
370 calls return information in
372 The correct syntax is:
377 .IB error " = ioctl(" unix_socket ", " ioctl_type ", &" value ");"
385 Returns the amount of queued unread data in the receive buffer.
386 The socket must not be in LISTEN state, otherwise an error
391 .IR <linux/sockios.h> .
392 .\" FIXME . http://sources.redhat.com/bugzilla/show_bug.cgi?id=12002,
393 .\" filed 2010-09-10, may cause SIOCINQ to be defined in glibc headers
395 you can use the synonymous
399 .\" SIOCOUTQ also has an effect for UNIX domain sockets, but not
400 .\" quite what userland might expect. It seems to return the number
401 .\" of bytes allocated for buffers containing pending output.
402 .\" That number is normally larger than the number of bytes of pending
403 .\" output. Since this info is, from userland's point of view, imprecise,
404 .\" and it may well change, probably best not to document this now.
408 The specified local address is already in use or the filesystem socket
409 object already exists.
412 The remote address specified by
414 was not a listening socket.
415 This error can also occur if the target pathname is not a socket.
418 Remote socket was unexpectedly closed.
421 User memory address was not valid.
424 Invalid argument passed.
425 A common cause is that the value
427 was not specified in the
429 field of passed addresses, or the socket was in an
430 invalid state for the applied operation.
434 called on an already connected socket or a target address was
435 specified on a connected socket.
438 The pathname in the remote address specified to
446 Socket operation needs a target address, but the socket is not connected.
449 Stream operation called on non-stream oriented socket or tried to
450 use the out-of-band data option.
453 The sender passed invalid credentials in the
457 Remote socket was closed on a stream socket.
461 This can be avoided by passing the
469 Passed protocol is not
473 Remote socket does not match the local socket type
481 Other errors can be generated by the generic socket layer or
482 by the filesystem while generating a filesystem socket object.
483 See the appropriate manual pages for more information.
486 and the abstract namespace were introduced with Linux 2.2 and should not
487 be used in portable programs.
488 (Some BSD-derived systems also support credential passing,
489 but the implementation details differ.)
491 In the Linux implementation, sockets which are visible in the
492 filesystem honor the permissions of the directory they are in.
493 Their owner, group, and permissions can be changed.
494 Creation of a new socket will fail if the process does not have write and
495 search (execute) permission on the directory the socket is created in.
496 Connecting to the socket object requires read/write permission.
497 This behavior differs from many BSD-derived systems which
498 ignore permissions for UNIX domain sockets.
499 Portable programs should not rely on
500 this feature for security.
502 Binding to a socket with a filename creates a socket
503 in the filesystem that must be deleted by the caller when it is no
506 The usual UNIX close-behind semantics apply; the socket can be unlinked
507 at any time and will be finally removed from the filesystem when the last
508 reference to it is closed.
510 To pass file descriptors or credentials over a
513 to send or receive at least one byte of nonancillary data in the same
519 UNIX domain stream sockets do not support the notion of out-of-band data.
522 When binding a socket to an address,
523 Linux is one of the implementations that appends a null terminator
524 if none is supplied in
526 In most cases this is unproblematic:
527 when the socket address is retrieved,
528 it will be one byte longer than that supplied when the socket was bound.
529 However, there is one case where confusing behavior can result:
530 if 108 non-null bytes are supplied when a socket is bound,
531 then the addition of the null terminator takes the length of
533 .IR sizeof(sun_path) .
534 Consequently, when retrieving the socket address
537 .\" The behavior on Solaris is quite similar.
540 argument for the retrieving call is specified as
541 .IR "sizeof(struct sockaddr_un)" ,
542 then the returned address structure
544 have a null terminator in
547 In addition, some implementations
548 .\" i.e., traditional BSD
549 don't require a null terminator when binding a socket (the
551 argument is used to determine the length of
553 and when the socket address is retrieved on these implementations,
554 there is no null terminator in
557 Applications that retrieve socket addresses can (portably) code
558 to handle the possibility that there is no null terminator in
560 by respecting the fact that the number of valid bytes in the pathname is:
562 strnlen(addr.sun_path, addrlen \- offsetof(sockaddr_un, sun_path))
563 .\" The following patch to amend kernel behavior was rejected:
564 .\" http://thread.gmane.org/gmane.linux.kernel.api/2437
565 .\" Subject: [patch] Fix handling of overlength pathname in AF_UNIX sun_path
567 .\" And there was a related discussion in the Austin list:
568 .\" http://thread.gmane.org/gmane.comp.standards.posix.austin.general/5735
569 .\" Subject: Having a sun_path with no null terminator
572 .\" FIXME . Track http://austingroupbugs.net/view.php?id=561
574 Alternatively, an application can retrieve
575 the socket address by allocating a buffer of size
576 .I "sizeof(struct sockaddr_un)+1"
577 that is zeroed out before the retrieval.
578 The retrieving call can specify
581 .IR "sizeof(struct sockaddr_un)" ,
582 and the extra zero byte ensures that there will be
583 a null terminator for the string returned in
590 addrlen = sizeof(struct sockaddr_un);
591 addrp = malloc(addrlen + 1);
594 memset(addrp, 0, addrlen + 1);
596 if (getsockname(sfd, (struct sockaddr *) addrp, &addrlen)) == \-1)
599 printf("sun_path = %s\\n", ((struct sockaddr_un *) addrp)\->sun_path);
603 This sort of messiness can be avoided if it is guaranteed
604 that the applications that
606 pathname sockets follow the rules outlined above under
607 .IR "Pathname sockets" .
612 For an example of the use of
622 .BR capabilities (7),
626 This page is part of release 3.79 of the Linux
629 A description of the project,
630 information about reporting bugs,
631 and the latest version of this page,
633 \%http://www.kernel.org/doc/man\-pages/.