1SOCKET(7)                  Linux Programmer's Manual                 SOCKET(7)
2
3
4

NAME

6       socket - Linux socket interface
7

SYNOPSIS

9       #include <sys/socket.h>
10
11       sockfd = socket(int socket_family, int socket_type, int protocol);
12

DESCRIPTION

14       This  manual  page  describes  the  Linux  networking socket layer user
15       interface.  The  BSD  compatible  sockets  are  the  uniform  interface
16       between the user process and the network protocol stacks in the kernel.
17       The protocol  modules  are  grouped  into  protocol  families  such  as
18       AF_INET, AF_IPX, and AF_PACKET, and socket types such as SOCK_STREAM or
19       SOCK_DGRAM.  See socket(2) for more information on families and types.
20
21   Socket-layer functions
22       These functions are used by the user process to send or receive packets
23       and  to  do  other  socket  operations.  For more information see their
24       respective manual pages.
25
26       socket(2) creates a socket, connect(2) connects a socket  to  a  remote
27       socket  address,  the bind(2) function binds a socket to a local socket
28       address, listen(2) tells the  socket  that  new  connections  shall  be
29       accepted, and accept(2) is used to get a new socket with a new incoming
30       connection.  socketpair(2)  returns  two  connected  anonymous  sockets
31       (implemented only for a few local families like AF_UNIX)
32
33       send(2),  sendto(2),  and  sendmsg(2)  send  data  over  a  socket, and
34       recv(2), recvfrom(2), recvmsg(2) receive data from a  socket.   poll(2)
35       and  select(2)  wait for arriving data or a readiness to send data.  In
36       addition, the standard I/O operations like write(2),  writev(2),  send‐
37       file(2), read(2), and readv(2) can be used to read and write data.
38
39       getsockname(2)  returns  the  local  socket  address and getpeername(2)
40       returns the remote socket address.  getsockopt(2) and setsockopt(2) are
41       used  to  set or get socket layer or protocol options.  ioctl(2) can be
42       used to set or read some other options.
43
44       close(2) is used to close a socket.   shutdown(2)  closes  parts  of  a
45       full-duplex socket connection.
46
47       Seeking,  or  calling  pread(2) or pwrite(2) with a nonzero position is
48       not supported on sockets.
49
50       It is possible to do nonblocking I/O on sockets by setting  the  O_NON‐
51       BLOCK flag on a socket file descriptor using fcntl(2).  Then all opera‐
52       tions that would block will (usually)  return  with  EAGAIN  (operation
53       should  be  retried  later);  connect(2) will return EINPROGRESS error.
54       The user can then wait for various events via poll(2) or select(2).
55
56       ┌────────────────────────────────────────────────────────────────────┐
57       │                            I/O events                              │
58       ├───────────┬───────────┬────────────────────────────────────────────┤
59       │Event      │ Poll flag │ Occurrence                                 │
60       ├───────────┼───────────┼────────────────────────────────────────────┤
61       │Read       │ POLLIN    │ New data arrived.                          │
62       ├───────────┼───────────┼────────────────────────────────────────────┤
63       │Read       │ POLLIN    │ A connection setup has been completed (for │
64       │           │           │ connection-oriented sockets)               │
65       ├───────────┼───────────┼────────────────────────────────────────────┤
66       │Read       │ POLLHUP   │ A disconnection request has been initiated │
67       │           │           │ by the other end.                          │
68       ├───────────┼───────────┼────────────────────────────────────────────┤
69       │Read       │ POLLHUP   │ A connection is broken (only  for  connec‐ │
70       │           │           │ tion-oriented protocols).  When the socket │
71       │           │           │ is written SIGPIPE is also sent.           │
72       ├───────────┼───────────┼────────────────────────────────────────────┤
73       │Write      │ POLLOUT   │ Socket has enough send  buffer  space  for │
74       │           │           │ writing new data.                          │
75       ├───────────┼───────────┼────────────────────────────────────────────┤
76       │Read/Write │ POLLIN |  │ An outgoing connect(2) finished.           │
77       │           │ POLLOUT   │                                            │
78       ├───────────┼───────────┼────────────────────────────────────────────┤
79       │Read/Write │ POLLERR   │ An asynchronous error occurred.            │
80       ├───────────┼───────────┼────────────────────────────────────────────┤
81       │Read/Write │ POLLHUP   │ The other end has shut down one direction. │
82       ├───────────┼───────────┼────────────────────────────────────────────┤
83       │Exception  │ POLLPRI   │ Urgent data arrived.  SIGURG is sent then. │
84       └───────────┴───────────┴────────────────────────────────────────────┘
85       An alternative to poll(2) and select(2) is to let the kernel inform the
86       application about events via a SIGIO signal.  For that the O_ASYNC flag
87       must be set on a socket file descriptor via fcntl(2) and a valid signal
88       handler for SIGIO must be installed via sigaction(2).  See the  Signals
89       discussion below.
90
91   Socket address structures
92       Each  socket  domain  has  its  own format for socket addresses, with a
93       domain-specific address structure.  Each  of  these  structures  begins
94       with  an  integer  "family" field (typed as sa_family_t) that indicates
95       the type of the address structure.   This  allows  the  various  system
96       calls  (e.g.,  connect(2), bind(2), accept(2), getsockname(2), getpeer‐
97       name(2)), which are generic to all socket  domains,  to  determine  the
98       domain of a particular socket address.
99
100       To  allow  any type of socket address to be passed to interfaces in the
101       sockets API, the type struct sockaddr is defined.  The purpose of  this
102       type is purely to allow casting of domain-specific socket address types
103       to a "generic" type, so as to avoid compiler warnings about  type  mis‐
104       matches in calls to the sockets API.
105
106       In  addition,  the  sockets  API  provides  the  data type struct sock‐
107       addr_storage.  This type  is  suitable  to  accommodate  all  supported
108       domain-specific  socket  address  structures; it is large enough and is
109       aligned properly.  (In particular, it is  large  enough  to  hold  IPv6
110       socket  addresses.)   The structure includes the following field, which
111       can be used to identify the type of socket address actually  stored  in
112       the structure:
113
114               sa_family_t ss_family;
115
116       The  sockaddr_storage  structure is useful in programs that must handle
117       socket addresses in a generic way (e.g., programs that must  deal  with
118       both IPv4 and IPv6 socket addresses).
119
120   Socket options
121       The  socket  options listed below can be set by using setsockopt(2) and
122       read with getsockopt(2) with the socket level set to SOL_SOCKET for all
123       sockets.  Unless otherwise noted, optval is a pointer to an int.
124
125       SO_ACCEPTCONN
126              Returns  a  value indicating whether or not this socket has been
127              marked to accept connections with listen(2).  The value 0  indi‐
128              cates that this is not a listening socket, the value 1 indicates
129              that this is a listening socket.  This socket  option  is  read-
130              only.
131
132       SO_ATTACH_FILTER (since Linux 2.2), SO_ATTACH_BPF (since Linux 3.19)
133              Attach  a  classic  BPF  (SO_ATTACH_FILTER)  or  an extended BPF
134              (SO_ATTACH_BPF) program to the socket for use  as  a  filter  of
135              incoming  packets.   A packet will be dropped if the filter pro‐
136              gram returns zero.  If the  filter  program  returns  a  nonzero
137              value  which  is  less than the packet's data length, the packet
138              will be truncated to the length returned.  If the value returned
139              by  the  filter  is  greater  than or equal to the packet's data
140              length, the packet is allowed to proceed unmodified.
141
142              The argument for SO_ATTACH_FILTER  is  a  sock_fprog  structure,
143              defined in <linux/filter.h>:
144
145                  struct sock_fprog {
146                      unsigned short      len;
147                      struct sock_filter *filter;
148                  };
149
150              The  argument for SO_ATTACH_BPF is a file descriptor returned by
151              the bpf(2) system call and must  refer  to  a  program  of  type
152              BPF_PROG_TYPE_SOCKET_FILTER.
153
154              These options may be set multiple times for a given socket, each
155              time replacing the previous filter  program.   The  classic  and
156              extended versions may be called on the same socket, but the pre‐
157              vious filter will always be replaced such that  a  socket  never
158              has more than one filter defined.
159
160              Both classic and extended BPF are explained in the kernel source
161              file Documentation/networking/filter.txt
162
163       SO_ATTACH_REUSEPORT_CBPF, SO_ATTACH_REUSEPORT_EBPF
164              For use with the SO_REUSEPORT option, these  options  allow  the
165              user  to  set  a  classic  BPF  (SO_ATTACH_REUSEPORT_CBPF) or an
166              extended BPF (SO_ATTACH_REUSEPORT_EBPF)  program  which  defines
167              how  packets  are assigned to the sockets in the reuseport group
168              (that is, all sockets which have SO_REUSEPORT set and are  using
169              the same local address to receive packets).
170
171              The  BPF  program  must return an index between 0 and N-1 repre‐
172              senting the socket which should receive the packet (where  N  is
173              the number of sockets in the group).  If the BPF program returns
174              an invalid index, socket selection will fall back to  the  plain
175              SO_REUSEPORT mechanism.
176
177              Sockets are numbered in the order in which they are added to the
178              group (that is, the order of bind(2) calls for  UDP  sockets  or
179              the  order  of  listen(2)  calls  for TCP sockets).  New sockets
180              added to a reuseport group will inherit the BPF program.  When a
181              socket  is  removed  from  a reuseport group (via close(2)), the
182              last socket in the group will be moved into the closed  socket's
183              position.
184
185              These options may be set repeatedly at any time on any socket in
186              the group to replace the current BPF program used by all sockets
187              in the group.
188
189              SO_ATTACH_REUSEPORT_CBPF   takes   the  same  argument  type  as
190              SO_ATTACH_FILTER and  SO_ATTACH_REUSEPORT_EBPF  takes  the  same
191              argument type as SO_ATTACH_BPF.
192
193              UDP  support  for this feature is available since Linux 4.5; TCP
194              support is available since Linux 4.6.
195
196       SO_BINDTODEVICE
197              Bind this socket to a particular device like “eth0”,  as  speci‐
198              fied  in  the  passed  interface  name.  If the name is an empty
199              string or the option length is zero, the socket  device  binding
200              is  removed.  The passed option is a variable-length null-termi‐
201              nated interface name string with the maximum size  of  IFNAMSIZ.
202              If a socket is bound to an interface, only packets received from
203              that particular interface are processed  by  the  socket.   Note
204              that this works only for some socket types, particularly AF_INET
205              sockets.  It is not supported for  packet  sockets  (use  normal
206              bind(2) there).
207
208              Before Linux 3.8, this socket option could be set, but could not
209              retrieved with getsockopt(2).  Since Linux 3.8, it is  readable.
210              The  optlen argument should contain the buffer size available to
211              receive the device name and is recommended to be IFNAMSIZ bytes.
212              The real device name length is reported back in the optlen argu‐
213              ment.
214
215       SO_BROADCAST
216              Set or get the broadcast flag.  When enabled,  datagram  sockets
217              are allowed to send packets to a broadcast address.  This option
218              has no effect on stream-oriented sockets.
219
220       SO_BSDCOMPAT
221              Enable BSD bug-to-bug compatibility.  This is used  by  the  UDP
222              protocol  module  in Linux 2.0 and 2.2.  If enabled, ICMP errors
223              received for a UDP socket will not be passed to  the  user  pro‐
224              gram.   In  later  kernel  versions, support for this option has
225              been phased out: Linux 2.4 silently ignores it,  and  Linux  2.6
226              generates  a  kernel  warning  (printk()) if a program uses this
227              option.  Linux 2.0 also  enabled  BSD  bug-to-bug  compatibility
228              options (random header changing, skipping of the broadcast flag)
229              for raw sockets with this option, but that was removed in  Linux
230              2.2.
231
232       SO_DEBUG
233              Enable  socket  debugging.   Allowed only for processes with the
234              CAP_NET_ADMIN capability or an effective user ID of 0.
235
236       SO_DETACH_FILTER (since Linux 2.2), SO_DETACH_BPF (since Linux 3.19)
237              These two options, which are synonyms, may be used to remove the
238              classic or extended BPF program attached to a socket with either
239              SO_ATTACH_FILTER or SO_ATTACH_BPF.  The option value is ignored.
240
241       SO_DOMAIN (since Linux 2.6.32)
242              Retrieves the socket domain as an  integer,  returning  a  value
243              such  as  AF_INET6.   See  socket(2)  for  details.  This socket
244              option is read-only.
245
246       SO_ERROR
247              Get and clear the pending socket error.  This socket  option  is
248              read-only.  Expects an integer.
249
250       SO_DONTROUTE
251              Don't send via a gateway, send only to directly connected hosts.
252              The same effect can be achieved  by  setting  the  MSG_DONTROUTE
253              flag  on a socket send(2) operation.  Expects an integer boolean
254              flag.
255
256       SO_INCOMING_CPU (gettable since Linux 3.19, settable since Linux 4.4)
257              Sets or gets the CPU affinity of a socket.  Expects  an  integer
258              flag.
259
260                  int cpu = 1;
261                  socklen_t len = sizeof(cpu);
262                  setsockopt(fd, SOL_SOCKET, SO_INCOMING_CPU, &cpu, &len);
263
264              Because  all of the packets for a single stream (i.e., all pack‐
265              ets for the same 4-tuple) arrive on the single RX queue that  is
266              associated  with  a  particular  CPU, the typical use case is to
267              employ one listening process per RX  queue,  with  the  incoming
268              flow  being  handled  by a listener on the same CPU that is han‐
269              dling the RX queue.  This provides  optimal  NUMA  behavior  and
270              keeps CPU caches hot.
271
272       SO_KEEPALIVE
273              Enable  sending  of  keep-alive  messages on connection-oriented
274              sockets.  Expects an integer boolean flag.
275
276       SO_LINGER
277              Sets or gets the SO_LINGER option.  The  argument  is  a  linger
278              structure.
279
280                  struct linger {
281                      int l_onoff;    /* linger active */
282                      int l_linger;   /* how many seconds to linger for */
283                  };
284
285              When  enabled,  a  close(2) or shutdown(2) will not return until
286              all queued messages for the socket have been  successfully  sent
287              or  the  linger  timeout  has been reached.  Otherwise, the call
288              returns immediately and the closing is done in  the  background.
289              When  the socket is closed as part of exit(2), it always lingers
290              in the background.
291
292       SO_LOCK_FILTER
293              When set, this option will prevent changing the filters  associ‐
294              ated  with  the socket.  These filters include any set using the
295              socket options SO_ATTACH_FILTER, SO_ATTACH_BPF, SO_ATTACH_REUSE‐
296              PORT_CBPF and SO_ATTACH_REUSEPORT_EPBF.
297
298              The typical use case is for a privileged process to set up a raw
299              socket (an operation that requires the CAP_NET_RAW  capability),
300              apply  a  restrictive filter, set the SO_LOCK_FILTER option, and
301              then either drop its privileges or pass the socket file descrip‐
302              tor to an unprivileged process via a UNIX domain socket.
303
304              Once  the  SO_LOCK_FILTER  option  has been enabled, attempts to
305              change or remove the filter attached to a socket, or to  disable
306              the SO_LOCK_FILTER option will fail with the error EPERM.
307
308       SO_MARK (since Linux 2.6.25)
309              Set  the  mark for each packet sent through this socket (similar
310              to the netfilter MARK target but  socket-based).   Changing  the
311              mark can be used for mark-based routing without netfilter or for
312              packet   filtering.    Setting   this   option   requires    the
313              CAP_NET_ADMIN capability.
314
315       SO_OOBINLINE
316              If  this  option is enabled, out-of-band data is directly placed
317              into the receive data stream.  Otherwise,  out-of-band  data  is
318              passed only when the MSG_OOB flag is set during receiving.
319
320       SO_PASSCRED
321              Enable  or  disable the receiving of the SCM_CREDENTIALS control
322              message.  For more information see unix(7).
323
324       SO_PEEK_OFF (since Linux 3.4)
325              This option, which is currently supported only for unix(7) sock‐
326              ets,  sets the value of the "peek offset" for the recv(2) system
327              call when used with MSG_PEEK flag.
328
329              When this option is set to a negative value (it is set to -1 for
330              all new sockets), traditional behavior is provided: recv(2) with
331              the MSG_PEEK flag will peek data from the front of the queue.
332
333              When the option is set to a value greater than or equal to zero,
334              then  the  next  peek at data queued in the socket will occur at
335              the byte offset specified by the  option  value.   At  the  same
336              time,  the  "peek  offset"  will be incremented by the number of
337              bytes that were peeked from the queue, so that a subsequent peek
338              will return the next data in the queue.
339
340              If  data  is  removed  from the front of the queue via a call to
341              recv(2) (or similar) without the MSG_PEEK flag, the  "peek  off‐
342              set" will be decreased by the number of bytes removed.  In other
343              words, receiving data without the MSG_PEEK flag will  cause  the
344              "peek  offset"  to  be adjusted to maintain the correct relative
345              position in the queued data, so  that  a  subsequent  peek  will
346              retrieve  the  data  that would have been retrieved had the data
347              not been removed.
348
349              For datagram sockets, if the "peek offset" points to the  middle
350              of a packet, the data returned will be marked with the MSG_TRUNC
351              flag.
352
353              The  following  example  serves  to  illustrate   the   use   of
354              SO_PEEK_OFF.   Suppose  a stream socket has the following queued
355              input data:
356
357                  aabbccddeeff
358
359              The following sequence of recv(2) calls would  have  the  effect
360              noted in the comments:
361
362                  int ov = 4;                  // Set peek offset to 4
363                  setsockopt(fd, SOL_SOCKET, SO_PEEK_OFF, &ov, sizeof(ov));
364
365                  recv(fd, buf, 2, MSG_PEEK);  // Peeks "cc"; offset set to 6
366                  recv(fd, buf, 2, MSG_PEEK);  // Peeks "dd"; offset set to 8
367                  recv(fd, buf, 2, 0);         // Reads "aa"; offset set to 6
368                  recv(fd, buf, 2, MSG_PEEK);  // Peeks "ee"; offset set to 8
369
370       SO_PEERCRED
371              Return  the credentials of the foreign process connected to this
372              socket.  This is possible  only  for  connected  AF_UNIX  stream
373              sockets  and  AF_UNIX  stream  and datagram socket pairs created
374              using socketpair(2); see unix(7).  The returned credentials  are
375              those  that were in effect at the time of the call to connect(2)
376              or socketpair(2).  The argument is a ucred structure; define the
377              _GNU_SOURCE  feature test macro to obtain the definition of that
378              structure from <sys/socket.h>.  This socket option is read-only.
379
380       SO_PRIORITY
381              Set the protocol-defined priority for all packets to be sent  on
382              this  socket.   Linux  uses  this  value to order the networking
383              queues: packets with a higher priority may  be  processed  first
384              depending on the selected device queueing discipline.  Setting a
385              priority outside the range 0 to  6  requires  the  CAP_NET_ADMIN
386              capability.
387
388       SO_PROTOCOL (since Linux 2.6.32)
389              Retrieves  the  socket protocol as an integer, returning a value
390              such as IPPROTO_SCTP.  See socket(2) for details.   This  socket
391              option is read-only.
392
393       SO_RCVBUF
394              Sets  or  gets  the maximum socket receive buffer in bytes.  The
395              kernel doubles this value (to allow space for bookkeeping  over‐
396              head) when it is set using setsockopt(2), and this doubled value
397              is returned by getsockopt(2).  The default value is set  by  the
398              /proc/sys/net/core/rmem_default  file,  and  the maximum allowed
399              value is set by the /proc/sys/net/core/rmem_max file.  The mini‐
400              mum (doubled) value for this option is 256.
401
402       SO_RCVBUFFORCE (since Linux 2.6.14)
403              Using  this  socket option, a privileged (CAP_NET_ADMIN) process
404              can perform the same task as SO_RCVBUF, but the  rmem_max  limit
405              can be overridden.
406
407       SO_RCVLOWAT and SO_SNDLOWAT
408              Specify  the  minimum  number  of  bytes in the buffer until the
409              socket layer will pass the data to the protocol (SO_SNDLOWAT) or
410              the  user on receiving (SO_RCVLOWAT).  These two values are ini‐
411              tialized to 1.  SO_SNDLOWAT is not changeable on Linux (setsock‐
412              opt(2)  fails  with  the  error  ENOPROTOOPT).   SO_RCVLOWAT  is
413              changeable only since Linux 2.4.  The select(2) and poll(2) sys‐
414              tem  calls  currently  do not respect the SO_RCVLOWAT setting on
415              Linux, and mark a socket readable when even  a  single  byte  of
416              data is available.  A subsequent read from the socket will block
417              until SO_RCVLOWAT bytes are available.
418
419       SO_RCVTIMEO and SO_SNDTIMEO
420              Specify the receiving or sending  timeouts  until  reporting  an
421              error.  The argument is a struct timeval.  If an input or output
422              function blocks for this period of time, and data has been  sent
423              or  received,  the  return  value  of  that function will be the
424              amount of data transferred; if no data has been transferred  and
425              the timeout has been reached, then -1 is returned with errno set
426              to EAGAIN or EWOULDBLOCK, or EINPROGRESS (for  connect(2))  just
427              as  if the socket was specified to be nonblocking.  If the time‐
428              out is set to zero (the default), then the operation will  never
429              timeout.   Timeouts  only have effect for system calls that per‐
430              form   socket   I/O   (e.g.,   read(2),   recvmsg(2),   send(2),
431              sendmsg(2));  timeouts  have  no  effect for select(2), poll(2),
432              epoll_wait(2), and so on.
433
434       SO_REUSEADDR
435              Indicates that the rules used in validating  addresses  supplied
436              in  a  bind(2)  call should allow reuse of local addresses.  For
437              AF_INET sockets this means that a socket may bind,  except  when
438              there  is an active listening socket bound to the address.  When
439              the listening socket is bound to INADDR_ANY with a specific port
440              then  it  is  not  possible  to  bind to this port for any local
441              address.  Argument is an integer boolean flag.
442
443       SO_REUSEPORT (since Linux 3.9)
444              Permits multiple AF_INET or AF_INET6 sockets to be bound  to  an
445              identical  socket  address.   This  option  must  be set on each
446              socket (including the first socket) prior to calling bind(2)  on
447              the  socket.   To  prevent  port hijacking, all of the processes
448              binding to the same address must have the  same  effective  UID.
449              This option can be employed with both TCP and UDP sockets.
450
451              For  TCP sockets, this option allows accept(2) load distribution
452              in a multi-threaded server to be improved by  using  a  distinct
453              listener  socket  for  each thread.  This provides improved load
454              distribution as compared to traditional techniques such using  a
455              single accept(2)ing thread that distributes connections, or hav‐
456              ing multiple threads that compete to  accept(2)  from  the  same
457              socket.
458
459              For  UDP sockets, the use of this option can provide better dis‐
460              tribution  of  incoming  datagrams  to  multiple  processes  (or
461              threads) as compared to the traditional technique of having mul‐
462              tiple processes compete to receive datagrams on the same socket.
463
464       SO_RXQ_OVFL (since Linux 2.6.33)
465              Indicates that an unsigned 32-bit value ancillary message (cmsg)
466              should  be  attached  to  received skbs indicating the number of
467              packets dropped by the socket since its creation.
468
469       SO_SNDBUF
470              Sets or gets the maximum socket send buffer in bytes.  The  ker‐
471              nel doubles this value (to allow space for bookkeeping overhead)
472              when it is set using setsockopt(2), and this  doubled  value  is
473              returned  by  getsockopt(2).   The  default  value is set by the
474              /proc/sys/net/core/wmem_default file  and  the  maximum  allowed
475              value is set by the /proc/sys/net/core/wmem_max file.  The mini‐
476              mum (doubled) value for this option is 2048.
477
478       SO_SNDBUFFORCE (since Linux 2.6.14)
479              Using this socket option, a privileged  (CAP_NET_ADMIN)  process
480              can  perform  the same task as SO_SNDBUF, but the wmem_max limit
481              can be overridden.
482
483       SO_TIMESTAMP
484              Enable or disable the receiving of the SO_TIMESTAMP control mes‐
485              sage.    The  timestamp  control  message  is  sent  with  level
486              SOL_SOCKET and the cmsg_data field is a struct timeval  indicat‐
487              ing  the reception time of the last packet passed to the user in
488              this call.  See cmsg(3) for details on control messages.
489
490       SO_TYPE
491              Gets the socket type as an integer  (e.g.,  SOCK_STREAM).   This
492              socket option is read-only.
493
494       SO_BUSY_POLL (since Linux 3.11)
495              Sets  the  approximate  time  in  microseconds to busy poll on a
496              blocking receive when there is no data.  Increasing  this  value
497              requires  CAP_NET_ADMIN.   The  default  for this option is con‐
498              trolled by the /proc/sys/net/core/busy_read file.
499
500              The value in the  /proc/sys/net/core/busy_poll  file  determines
501              how  long select(2) and poll(2) will busy poll when they operate
502              on sockets with SO_BUSY_POLL set and no  events  to  report  are
503              found.
504
505              In  both  cases,  busy polling will only be done when the socket
506              last received data from a  network  device  that  supports  this
507              option.
508
509              While  busy  polling  may  improve latency of some applications,
510              care must be taken when using it since this will  increase  both
511              CPU utilization and power usage.
512
513   Signals
514       When  writing onto a connection-oriented socket that has been shut down
515       (by the local or the remote end) SIGPIPE is sent to the writing process
516       and  EPIPE  is  returned.   The  signal is not sent when the write call
517       specified the MSG_NOSIGNAL flag.
518
519       When requested with the FIOSETOWN fcntl(2) or SIOCSPGRP ioctl(2), SIGIO
520       is  sent  when  an  I/O event occurs.  It is possible to use poll(2) or
521       select(2) in the signal handler to find  out  which  socket  the  event
522       occurred  on.  An alternative (in Linux 2.2) is to set a real-time sig‐
523       nal using the F_SETSIG fcntl(2); the handler of the  real  time  signal
524       will  be called with the file descriptor in the si_fd field of its sig‐
525       info_t.  See fcntl(2) for more information.
526
527       Under some circumstances (e.g., multiple processes accessing  a  single
528       socket),  the  condition  that caused the SIGIO may have already disap‐
529       peared when the process reacts to the signal.   If  this  happens,  the
530       process should wait again because Linux will resend the signal later.
531
532   /proc interfaces
533       The  core socket networking parameters can be accessed via files in the
534       directory /proc/sys/net/core/.
535
536       rmem_default
537              contains the default setting in bytes of the socket receive buf‐
538              fer.
539
540       rmem_max
541              contains the maximum socket receive buffer size in bytes which a
542              user may set by using the SO_RCVBUF socket option.
543
544       wmem_default
545              contains the default setting in bytes of the socket send buffer.
546
547       wmem_max
548              contains the maximum socket send buffer size in  bytes  which  a
549              user may set by using the SO_SNDBUF socket option.
550
551       message_cost and message_burst
552              configure  the  token  bucket  filter used to load limit warning
553              messages caused by external network events.
554
555       netdev_max_backlog
556              Maximum number of packets in the global input queue.
557
558       optmem_max
559              Maximum length of ancillary data and user control data like  the
560              iovecs per socket.
561
562   Ioctls
563       These operations can be accessed using ioctl(2):
564
565           error = ioctl(ip_socket, ioctl_type, &value_result);
566
567       SIOCGSTAMP
568              Return  a  struct timeval with the receive timestamp of the last
569              packet passed to the user.  This is useful  for  accurate  round
570              trip  time  measurements.  See setitimer(2) for a description of
571              struct timeval.  This ioctl should be used only  if  the  socket
572              option  SO_TIMESTAMP  is  not  set on the socket.  Otherwise, it
573              returns the timestamp of the last packet that was received while
574              SO_TIMESTAMP was not set, or it fails if no such packet has been
575              received, (i.e., ioctl(2) returns -1 with errno set to ENOENT).
576
577       SIOCSPGRP
578              Set the process or process group that is  to  receive  SIGIO  or
579              SIGURG  signals  when  I/O  becomes  possible  or urgent data is
580              available.  The argument is a pointer to a pid_t.   For  further
581              details, see the description of F_SETOWN in fcntl(2).
582
583       FIOASYNC
584              Change  the  O_ASYNC  flag to enable or disable asynchronous I/O
585              mode of the socket.  Asynchronous I/O mode means that the  SIGIO
586              signal  or the signal set with F_SETSIG is raised when a new I/O
587              event occurs.
588
589              Argument is an integer boolean flag.  (This operation is synony‐
590              mous with the use of fcntl(2) to set the O_ASYNC flag.)
591
592       SIOCGPGRP
593              Get  the current process or process group that receives SIGIO or
594              SIGURG signals, or 0 when none is set.
595
596       Valid fcntl(2) operations:
597
598       FIOGETOWN
599              The same as the SIOCGPGRP ioctl(2).
600
601       FIOSETOWN
602              The same as the SIOCSPGRP ioctl(2).
603

VERSIONS

605       SO_BINDTODEVICE was introduced in Linux 2.0.30.  SO_PASSCRED is new  in
606       Linux 2.2.  The /proc interfaces were introduced in Linux 2.2.  SO_RCV‐
607       TIMEO and SO_SNDTIMEO are supported since Linux 2.3.41.  Earlier, time‐
608       outs  were  fixed to a protocol-specific setting, and could not be read
609       or written.
610

NOTES

612       Linux assumes that half of the send/receive buffer is used for internal
613       kernel structures; thus the values in the corresponding /proc files are
614       twice what can be observed on the wire.
615
616       Linux will allow port reuse only with the SO_REUSEADDR option when this
617       option was set both in the previous program that performed a bind(2) to
618       the port and in the program that wants to reuse the port.  This differs
619       from  some implementations (e.g., FreeBSD) where only the later program
620       needs to set the SO_REUSEADDR option.   Typically  this  difference  is
621       invisible,  since,  for example, a server program is designed to always
622       set this option.
623

SEE ALSO

625       wireshark(1),   bpf(2),   connect(2),   getsockopt(2),   setsockopt(2),
626       socket(2),  pcap(3), capabilities(7), ddp(7), ip(7), packet(7), tcp(7),
627       udp(7), unix(7), tcpdump(8)
628

COLOPHON

630       This page is part of release 4.15 of the Linux  man-pages  project.   A
631       description  of  the project, information about reporting bugs, and the
632       latest    version    of    this    page,    can     be     found     at
633       https://www.kernel.org/doc/man-pages/.
634
635
636
637Linux                             2018-02-02                         SOCKET(7)
Impressum