2 .\" This manpage is Copyright (C) 1992 Drew Eckhardt;
3 .\" and Copyright (C) 1993 Michael Haardt, Ian Jackson;
4 .\" and Copyright (C) 1998 Jamie Lokier;
5 .\" and Copyright (C) 2002-2010, 2014 Michael Kerrisk;
6 .\" and Copyright (C) 2014 Jeff Layton
8 .\" %%%LICENSE_START(VERBATIM)
9 .\" Permission is granted to make and distribute verbatim copies of this
10 .\" manual provided the copyright notice and this permission notice are
11 .\" preserved on all copies.
13 .\" Permission is granted to copy and distribute modified versions of this
14 .\" manual under the conditions for verbatim copying, provided that the
15 .\" entire resulting derived work is distributed under the terms of a
16 .\" permission notice identical to this one.
18 .\" Since the Linux kernel and libraries are constantly changing, this
19 .\" manual page may be incorrect or out-of-date. The author(s) assume no
20 .\" responsibility for errors or omissions, or for damages resulting from
21 .\" the use of the information contained herein. The author(s) may not
22 .\" have taken the same level of care in the production of this manual,
23 .\" which is licensed free of charge, as they might when working
26 .\" Formatted or processed versions of this manual, if unaccompanied by
27 .\" the source, must acknowledge the copyright and authors of this work.
30 .\" Modified 1993-07-24 by Rik Faith <faith@cs.unc.edu>
31 .\" Modified 1995-09-26 by Andries Brouwer <aeb@cwi.nl>
32 .\" and again on 960413 and 980804 and 981223.
33 .\" Modified 1998-12-11 by Jamie Lokier <jamie@imbolc.ucc.ie>
34 .\" Applied correction by Christian Ehrhardt - aeb, 990712
35 .\" Modified 2002-04-23 by Michael Kerrisk <mtk.manpages@gmail.com>
36 .\" Added note on F_SETFL and O_DIRECT
37 .\" Complete rewrite + expansion of material on file locking
38 .\" Incorporated description of F_NOTIFY, drawing on
39 .\" Stephen Rothwell's notes in Documentation/dnotify.txt.
40 .\" Added description of F_SETLEASE and F_GETLEASE
41 .\" Corrected and polished, aeb, 020527.
42 .\" Modified 2004-03-03 by Michael Kerrisk <mtk.manpages@gmail.com>
43 .\" Modified description of file leases: fixed some errors of detail
44 .\" Replaced the term "lease contestant" by "lease breaker"
45 .\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
46 .\" Added notes on capability requirements
47 .\" Modified 2004-12-08, added O_NOATIME after note from Martin Pool
48 .\" 2004-12-10, mtk, noted F_GETOWN bug after suggestion from aeb.
49 .\" 2005-04-08 Jamie Lokier <jamie@shareable.org>, mtk
50 .\" Described behavior of F_SETOWN/F_SETSIG in
51 .\" multithreaded processes, and generally cleaned
52 .\" up the discussion of F_SETOWN.
53 .\" 2005-05-20, Johannes Nicolai <johannes.nicolai@hpi.uni-potsdam.de>,
54 .\" mtk: Noted F_SETOWN bug for socket file descriptor in Linux 2.4
55 .\" and earlier. Added text on permissions required to send signal.
56 .\" 2009-09-30, Michael Kerrisk
57 .\" Note obsolete F_SETOWN behavior with threads.
58 .\" Document F_SETOWN_EX and F_GETOWN_EX
59 .\" 2010-06-17, Michael Kerrisk
60 .\" Document F_SETPIPE_SZ and F_GETPIPE_SZ.
62 .TH FCNTL 2 2014-05-21 "Linux" "Linux Programmer's Manual"
64 fcntl \- manipulate file descriptor
67 .B #include <unistd.h>
70 .BI "int fcntl(int " fd ", int " cmd ", ... /* " arg " */ );"
74 performs one of the operations described below on the open file descriptor
76 The operation is determined by
80 can take an optional third argument.
81 Whether or not this argument is required is determined by
83 The required argument type is indicated in parentheses after each
85 name (in most cases, the required type is
87 and we identify the argument using the name
91 is specified if the argument is not required.
92 .SS Duplicating a file descriptor
94 .BR F_DUPFD " (\fIint\fP)"
95 Find the lowest numbered available file descriptor
96 greater than or equal to
98 and make it be a copy of
100 This is different from
102 which uses exactly the descriptor specified.
104 On success, the new descriptor is returned.
110 .BR F_DUPFD_CLOEXEC " (\fIint\fP; since Linux 2.6.24)"
113 but additionally set the
114 close-on-exec flag for the duplicate descriptor.
115 Specifying this flag permits a program to avoid an additional
121 For an explanation of why this flag is useful,
122 see the description of
126 .SS File descriptor flags
127 The following commands manipulate the flags associated with
129 Currently, only one such flag is defined:
131 the close-on-exec flag.
134 bit is 0, the file descriptor will remain open across an
136 otherwise it will be closed.
138 .BR F_GETFD " (\fIvoid\fP)"
139 Read the file descriptor flags;
143 .BR F_SETFD " (\fIint\fP)"
144 Set the file descriptor flags to the value specified by
147 In multithreaded programs, using
150 to set the close-on-exec flag at the same time as another thread performs a
154 is vulnerable to a race condition that may unintentionally leak
155 the file descriptor to the program executed in the child process.
156 See the discussion of the
160 for details and a remedy to the problem.
161 .SS File status flags
162 Each open file description has certain associated status flags,
167 and possibly modified by
169 Duplicated file descriptors
174 etc.) refer to the same open file description, and thus
175 share the same file status flags.
177 The file status flags and their semantics are described in
180 .BR F_GETFL " (\fIvoid\fP)"
181 Get the file access mode and the file status flags;
185 .BR F_SETFL " (\fIint\fP)"
186 Set the file status flags to the value specified by
189 .RB ( O_RDONLY ", " O_WRONLY ", " O_RDWR )
190 and file creation flags
192 .BR O_CREAT ", " O_EXCL ", " O_NOCTTY ", " O_TRUNC )
196 On Linux this command can change only the
204 It is not possible to change the
208 flags; see BUGS, below.
209 .SS Advisory record locking
210 Linux implements traditional ("process-associated") UNIX record locks,
211 as standardized by POSIX.
212 For a Linux-specific alternative with better semantics,
213 see the discussion of open file description locks below.
219 are used to acquire, release, and test for the existence of record
220 locks (also known as byte-range, file-segment, or file-region locks).
223 is a pointer to a structure that has at least the following fields
224 (in unspecified order).
230 short l_type; /* Type of lock: F_RDLCK,
232 short l_whence; /* How to interpret l_start:
233 SEEK_SET, SEEK_CUR, SEEK_END */
234 off_t l_start; /* Starting offset for lock */
235 off_t l_len; /* Number of bytes to lock */
236 pid_t l_pid; /* PID of process blocking our lock
237 (set by F_GETLK and F_OFD_GETLK) */
244 .IR l_whence ", " l_start ", and " l_len
245 fields of this structure specify the range of bytes we wish to lock.
246 Bytes past the end of the file may be locked,
247 but not bytes before the start of the file.
250 is the starting offset for the lock, and is interpreted
252 the start of the file (if
256 the current file offset (if
260 or the end of the file (if
264 In the final two cases,
266 can be a negative number provided the
267 offset does not lie before the start of the file.
270 specifies the number of bytes to be locked.
273 is positive, then the range to be locked covers bytes
276 .IR l_start + l_len \-1.
279 has the special meaning: lock all bytes starting at the
280 location specified by
281 .IR l_whence " and " l_start
282 through to the end of file, no matter how large the file grows.
284 POSIX.1-2001 allows (but does not require)
285 an implementation to support a negative
289 is negative, the interval described by
295 This is supported by Linux since kernel versions 2.4.21 and 2.5.49.
299 field can be used to place a read
304 Any number of processes may hold a read lock (shared lock)
305 on a file region, but only one process may hold a write lock
307 An exclusive lock excludes all other locks,
308 both shared and exclusive.
309 A single process can hold only one type of lock on a file region;
310 if a new lock is applied to an already-locked region,
311 then the existing lock is converted to the new lock type.
312 (Such conversions may involve splitting, shrinking, or coalescing with
313 an existing lock if the byte range specified by the new lock does not
314 precisely coincide with the range of the existing lock.)
316 .BR F_SETLK " (\fIstruct flock *\fP)"
323 or release a lock (when
327 on the bytes specified by the
328 .IR l_whence ", " l_start ", and " l_len
331 If a conflicting lock is held by another process,
332 this call returns \-1 and sets
339 .BR F_SETLKW " (\fIstruct flock *\fP)"
342 but if a conflicting lock is held on the file, then wait for that
344 If a signal is caught while waiting, then the call is interrupted
345 and (after the signal handler has returned)
346 returns immediately (with return value \-1 and
353 .BR F_GETLK " (\fIstruct flock *\fP)"
354 On input to this call,
356 describes a lock we would like to place on the file.
357 If the lock could be placed,
359 does not actually place it, but returns
365 and leaves the other fields of the structure unchanged.
367 If one or more incompatible locks would prevent
368 this lock being placed, then
370 returns details about one of those locks in the
371 .IR l_type ", " l_whence ", " l_start ", and " l_len
374 If the conflicting lock is a traditional (process-associated) record lock,
377 field is set to the PID of the process holding that lock.
378 If the conflicting lock is an open file description lock, then
381 Note that the returned information
382 may already be out of date by the time the caller inspects it.
384 In order to place a read lock,
386 must be open for reading.
387 In order to place a write lock,
389 must be open for writing.
390 To place both types of lock, open a file read-write.
392 As well as being removed by an explicit
394 record locks are automatically released when the process terminates.
396 Record locks are not inherited by a child created via
398 but are preserved across an
401 Because of the buffering performed by the
403 library, the use of record locking with routines in that package
404 should be avoided; use
410 The record locks described above are associated with the process
411 (unlike the open file description locks described below).
412 This has some unfortunate consequences:
414 If a process holding a lock on a file closes
416 file descriptor referring to the file,
417 then all of the process's locks on the file are released,
418 no matter which file descriptor they were obtained via.
419 .\" (Additional file descriptors referring to the same file
420 .\" may have been obtained by calls to
421 .\" .BR open "(2), " dup "(2), " dup2 "(2), or " fcntl ().)
422 This is bad: it means that a process can lose its locks on
427 when for some reason a library function decides to open, read,
428 and close the same file.
430 The threads in a process share locks.
432 a multithreaded program can't use record locking to ensure
433 that threads don't simultaneously access the same region of a file.
435 Open file description locks solve both of these problems.
436 .SS Open file description locks (non-POSIX)
437 Open file description locks are advisory byte-range locks whose operation is
438 in most respects identical to the traditional record locks described above.
439 This lock type is Linux-specific,
440 and available since Linux 3.15.
442 The principal difference between the two lock types
443 is that whereas traditional record locks
444 are associated with a process,
445 open file description locks are associated with the
446 open file description on which they are acquired,
447 much like locks acquired with
449 Consequently (and unlike traditional advisory record locks),
450 open file description locks are inherited across
456 and are only automatically released on the last close
457 of the open file description,
458 instead of being released on any close of the file.
460 Open file description locks always conflict with traditional record locks,
461 even when they are acquired by the same process on the same file descriptor.
463 Open file description locks placed via the same open file description
464 (i.e., via the same file descriptor,
465 or via a duplicate of the file descriptor created by
470 and so on) are always compatible:
471 if a new lock is placed on an already locked region,
472 then the existing lock is converted to the new lock type.
473 (Such conversions may result in splitting, shrinking, or coalescing with
474 an existing lock as discussed above.)
476 On the other hand, open file description locks may conflict with
477 each other when they are acquired via different open file descriptions.
478 Thus, the threads in a multithreaded program can use
479 open file description locks to synchronize access to a file region
480 by having each thread perform its own
482 on the file and applying locks via the resulting file descriptor.
484 As with traditional advisory locks, the third argument to
490 By contrast with traditional record locks, the
492 field of that structure must be set to zero
493 when using the commands described below.
495 The commands for working with open file description locks are analogous
496 to those used with traditional locks:
498 .BR F_OFD_SETLK " (\fIstruct flock *\fP)"
499 Acquire an open file description lock (when
505 or release an open file description lock (when
509 on the bytes specified by the
510 .IR l_whence ", " l_start ", and " l_len
513 If a conflicting lock is held by another process,
514 this call returns \-1 and sets
517 .\" FIXME: EACCESS should probably not be here
522 .BR F_OFD_SETLKW " (\fIstruct flock *\fP)"
525 but if a conflicting lock is held on the file, then wait for that lock to be
527 If a signal is caught while waiting, then the call is interrupted
528 and (after the signal handler has returned) returns immediately
529 (with return value \-1 and
536 .BR F_OFD_GETLK " (\fIstruct flock *\fP)"
537 On input to this call,
539 describes an open file description lock we would like to place on the file.
540 If the lock could be placed,
542 does not actually place it, but returns
548 and leaves the other fields of the structure unchanged.
549 If one or more incompatible locks would prevent this lock being placed,
550 then details about one of these locks are returned via
552 as described above for
554 .SS Mandatory locking
556 The above record locks may be either advisory or mandatory,
557 and are advisory by default.
559 Advisory locks are not enforced and are useful only between
560 cooperating processes.
562 Mandatory locks are enforced for all processes.
563 If a process tries to perform an incompatible access (e.g.,
567 on a file region that has an incompatible mandatory lock,
568 then the result depends upon whether the
570 flag is enabled for its open file description.
573 flag is not enabled, then
574 system call is blocked until the lock is removed
575 or converted to a mode that is compatible with the access.
578 flag is enabled, then the system call fails with the error
581 To make use of mandatory locks, mandatory locking must be enabled
582 both on the filesystem that contains the file to be locked,
583 and on the file itself.
584 Mandatory locking is enabled on a filesystem
585 using the "\-o mand" option to
591 Mandatory locking is enabled on a file by disabling
592 group execute permission on the file and enabling the set-group-ID
598 The Linux implementation of mandatory locking is unreliable.
608 are used to manage I/O availability signals:
610 .BR F_GETOWN " (\fIvoid\fP)"
611 Return (as the function result)
612 the process ID or process group currently receiving
616 signals for events on file descriptor
618 Process IDs are returned as positive values;
619 process group IDs are returned as negative values (but see BUGS below).
623 .BR F_SETOWN " (\fIint\fP)"
624 Set the process ID or process group ID that will receive
628 signals for events on file descriptor
632 A process ID is specified as a positive value;
633 a process group ID is specified as a negative value.
634 Most commonly, the calling process specifies itself as the owner
643 status flag on a file descriptor by using the
649 signal is sent whenever input or output becomes possible
650 on that file descriptor.
652 can be used to obtain delivery of a signal other than
654 If this permission check fails, then the signal is
657 Sending a signal to the owner process (group) specified by
659 is subject to the same permissions checks as are described for
661 where the sending process is the one that employs
663 (but see BUGS below).
665 If the file descriptor
672 signals that are delivered when out-of-band
673 data arrives on that socket.
675 is sent in any situation where
677 would report the socket as having an "exceptional condition".)
678 .\" The following appears to be rubbish. It doesn't seem to
679 .\" be true according to the kernel source, and I can write
680 .\" a program that gets a terminal-generated SIGIO even though
681 .\" it is not the foreground process group of the terminal.
684 .\" If the file descriptor
686 .\" refers to a terminal device, then SIGIO
687 .\" signals are sent to the foreground process group of the terminal.
689 The following was true in 2.6.x kernels up to and including
693 If a nonzero value is given to
695 in a multithreaded process running with a threading library
696 that supports thread groups (e.g., NPTL),
697 then a positive value given to
699 has a different meaning:
700 .\" The relevant place in the (2.6) kernel source is the
701 .\" 'switch' in fs/fcntl.c::send_sigio_to_task() -- MTK, Apr 2005
702 instead of being a process ID identifying a whole process,
703 it is a thread ID identifying a specific thread within a process.
704 Consequently, it may be necessary to pass
710 to get sensible results when
713 (In current Linux threading implementations,
714 a main thread's thread ID is the same as its process ID.
715 This means that a single-threaded program can equally use
720 Note, however, that the statements in this paragraph do not apply
723 signal generated for out-of-band data on a socket:
724 this signal is always sent to either a process or a process group,
725 depending on the value given to
727 .\" send_sigurg()/send_sigurg_to_task() bypasses
728 .\" kill_fasync()/send_sigio()/send_sigio_to_task()
729 .\" to directly call send_group_sig_info()
730 .\" -- MTK, Apr 2005 (kernel 2.6.11)
733 The above behavior was accidentally dropped in Linux 2.6.12,
734 and won't be restored.
735 From Linux 2.6.32 onward, use
741 signals at a particular thread.
743 .BR F_GETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
744 Return the current file descriptor owner settings
745 as defined by a previous
748 The information is returned in the structure pointed to by
750 which has the following form:
763 field will have one of the values
770 field is a positive integer representing a thread ID, process ID,
776 .BR F_SETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
777 This operation performs a similar task to
779 It allows the caller to direct I/O availability signals
780 to a specific thread, process, or process group.
781 The caller specifies the target of signals via
783 which is a pointer to a
788 field has one of the following values, which define how
794 Send the signal to the thread whose thread ID
795 (the value returned by a call to
803 Send the signal to the process whose ID
808 Send the signal to the process group whose ID
811 (Note that, unlike with
813 a process group ID is specified as a positive value here.)
816 .BR F_GETSIG " (\fIvoid\fP)"
817 Return (as the function result)
818 the signal sent when input or output becomes possible.
819 A value of zero means
822 Any other value (including
825 signal sent instead, and in this case additional info is available to
826 the signal handler if installed with
831 .BR F_SETSIG " (\fIint\fP)"
832 Set the signal sent when input or output becomes possible
833 to the value given in
835 A value of zero means to send the default
838 Any other value (including
840 is the signal to send instead, and in this case additional info
841 is available to the signal handler if installed with
844 .\" The following was true only up until 2.6.11:
846 .\" Additionally, passing a nonzero value to
848 .\" changes the signal recipient from a whole process to a specific thread
849 .\" within a process.
850 .\" See the description of
852 .\" for more details.
856 with a nonzero value, and setting
861 extra information about I/O events is passed to
867 field indicates the source is
871 field gives the file descriptor associated with the event.
873 there is no indication which file descriptors are pending, and you
874 should use the usual mechanisms
880 set etc.) to determine which file descriptors are available for I/O.
882 By selecting a real time signal (value >=
884 multiple I/O events may be queued using the same signal numbers.
885 (Queuing is dependent on available memory).
886 Extra information is available
889 is set for the signal handler, as above.
891 Note that Linux imposes a limit on the
892 number of real-time signals that may be queued to a
897 and if this limit is reached, then the kernel reverts to
900 and this signal is delivered to the entire
901 process rather than to a specific thread.
902 .\" See fs/fcntl.c::send_sigio_to_task() (2.4/2.6) sources -- MTK, Apr 05
904 Using these mechanisms, a program can implement fully asynchronous I/O
913 is specific to BSD and Linux.
918 specified in POSIX.1 is in conjunction with the use of the
921 (POSIX does not specify the
930 POSIX has asynchronous I/O and the
932 structure to achieve similar things; these are also available
933 in Linux as part of the GNU C Library (Glibc).
938 (Linux 2.4 onward) are used (respectively) to establish a new lease,
939 and retrieve the current lease, on the open file description
940 referred to by the file descriptor
942 A file lease provides a mechanism whereby the process holding
943 the lease (the "lease holder") is notified (via delivery of a signal)
944 when a process (the "lease breaker") tries to
948 the file referred to by that file descriptor.
950 .BR F_SETLEASE " (\fIint\fP)"
951 Set or remove a file lease according to which of the following
952 values is specified in the integer
957 Take out a read lease.
958 This will cause the calling process to be notified when
959 the file is opened for writing or is truncated.
960 .\" The following became true in kernel 2.6.10:
961 .\" See the man-pages-2.09 Changelog for further info.
962 A read lease can be placed only on a file descriptor that
966 Take out a write lease.
967 This will cause the caller to be notified when
968 the file is opened for reading or writing or is truncated.
969 A write lease may be placed on a file only if there are no
970 other open file descriptors for the file.
973 Remove our lease from the file.
976 Leases are associated with an open file description (see
978 This means that duplicate file descriptors (created by, for example,
982 refer to the same lease, and this lease may be modified
983 or released using any of these descriptors.
984 Furthermore, the lease is released by either an explicit
986 operation on any of these duplicate descriptors, or when all
987 such descriptors have been closed.
989 Leases may be taken out only on regular files.
990 An unprivileged process may take out a lease only on a file whose
991 UID (owner) matches the filesystem UID of the process.
994 capability may take out leases on arbitrary files.
996 .BR F_GETLEASE " (\fIvoid\fP)"
997 Indicates what type of lease is associated with the file descriptor
1000 .BR F_RDLCK ", " F_WRLCK ", or " F_UNLCK ,
1001 indicating, respectively, a read lease , a write lease, or no lease.
1005 When a process (the "lease breaker") performs an
1009 that conflicts with a lease established via
1011 the system call is blocked by the kernel and
1012 the kernel notifies the lease holder by sending it a signal
1015 The lease holder should respond to receipt of this signal by doing
1016 whatever cleanup is required in preparation for the file to be
1017 accessed by another process (e.g., flushing cached buffers) and
1018 then either remove or downgrade its lease.
1019 A lease is removed by performing an
1025 If the lease holder currently holds a write lease on the file,
1026 and the lease breaker is opening the file for reading,
1027 then it is sufficient for the lease holder to downgrade
1028 the lease to a read lease.
1029 This is done by performing an
1036 If the lease holder fails to downgrade or remove the lease within
1037 the number of seconds specified in
1038 .IR /proc/sys/fs/lease-break-time ,
1039 then the kernel forcibly removes or downgrades the lease holder's lease.
1041 Once a lease break has been initiated,
1043 returns the target lease type (either
1047 depending on what would be compatible with the lease breaker)
1048 until the lease holder voluntarily downgrades or removes the lease or
1049 the kernel forcibly does so after the lease break timer expires.
1051 Once the lease has been voluntarily or forcibly removed or downgraded,
1052 and assuming the lease breaker has not unblocked its system call,
1053 the kernel permits the lease breaker's system call to proceed.
1055 If the lease breaker's blocked
1059 is interrupted by a signal handler,
1060 then the system call fails with the error
1062 but the other steps still occur as described above.
1063 If the lease breaker is killed by a signal while blocked in
1067 then the other steps still occur as described above.
1068 If the lease breaker specifies the
1072 then the call immediately fails with the error
1074 but the other steps still occur as described above.
1076 The default signal used to notify the lease holder is
1078 but this can be changed using the
1084 command is performed (even one specifying
1087 handler is established using
1089 then the handler will receive a
1091 structure as its second argument, and the
1093 field of this argument will hold the descriptor of the leased file
1094 that has been accessed by another process.
1095 (This is useful if the caller holds leases against multiple files).
1096 .SS File and directory change notification (dnotify)
1098 .BR F_NOTIFY " (\fIint\fP)"
1100 Provide notification when the directory referred to by
1102 or any of the files that it contains is changed.
1103 The events to be notified are specified in
1105 which is a bit mask specified by ORing together zero or more of
1112 A file was accessed (read, pread, readv)
1115 A file was modified (write, pwrite, writev, truncate, ftruncate).
1118 A file was created (open, creat, mknod, mkdir, link, symlink, rename).
1121 A file was unlinked (unlink, rename to another directory, rmdir).
1124 A file was renamed within this directory (rename).
1127 The attributes of a file were changed (chown, chmod, utime[s]).
1131 (In order to obtain these definitions, the
1133 feature test macro must be defined before including
1137 Directory notifications are normally "one-shot", and the application
1138 must reregister to receive further notifications.
1143 then notification will remain in effect until explicitly removed.
1145 .\" The following does seem a poor API-design choice...
1148 requests is cumulative, with the events in
1150 being added to the set already monitored.
1151 To disable notification of all events, make an
1157 Notification occurs via delivery of a signal.
1158 The default signal is
1160 but this can be changed using the
1164 In the latter case, the signal handler receives a
1166 structure as its second argument (if the handler was
1171 field of this structure contains the file descriptor which
1172 generated the notification (useful when establishing notification
1173 on multiple directories).
1175 Especially when using
1177 a real time signal should be used for notification,
1178 so that multiple notifications can be queued.
1181 New applications should use the
1183 interface (available since kernel 2.6.13),
1184 which provides a much superior interface for obtaining notifications of
1188 .SS Changing the capacity of a pipe
1190 .BR F_SETPIPE_SZ " (\fIint\fP; since Linux 2.6.35)"
1191 Change the capacity of the pipe referred to by
1196 An unprivileged process can adjust the pipe capacity to any value
1197 between the system page size and the limit defined in
1198 .IR /proc/sys/fs/pipe-max-size
1201 Attempts to set the pipe capacity below the page size are silently
1202 rounded up to the page size.
1203 Attempts by an unprivileged process to set the pipe capacity above the limit in
1204 .IR /proc/sys/fs/pipe-max-size
1207 a privileged process
1208 .RB ( CAP_SYS_RESOURCE )
1209 can override the limit.
1210 When allocating the buffer for the pipe,
1211 the kernel may use a capacity larger than
1213 if that is convenient for the implementation.
1216 operation returns the actual size used.
1217 Attempting to set the pipe capacity smaller than the amount
1218 of buffer space currently used to store data produces the error
1221 .BR F_GETPIPE_SZ " (\fIvoid\fP; since Linux 2.6.35)"
1222 Return (as the function result) the capacity of the pipe referred to by
1225 For a successful call, the return value depends on the operation:
1231 Value of file descriptor flags.
1234 Value of file status flags.
1237 Type of lease held on file descriptor.
1240 Value of descriptor owner.
1243 Value of signal sent when read or write becomes possible, or zero
1254 On error, \-1 is returned, and
1256 is set appropriately.
1259 .BR EACCES " or " EAGAIN
1260 Operation is prohibited by locks held by other processes.
1263 The operation is prohibited because the file has been memory-mapped by
1268 is not an open file descriptor, or the command was
1272 and the file descriptor open mode doesn't match with the
1273 type of lock requested.
1276 It was detected that the specified
1278 command would cause a deadlock.
1282 is outside your accessible address space.
1287 the command was interrupted by a signal; see
1290 .BR F_GETLK " and " F_SETLK ,
1291 the command was interrupted by a signal before the lock was checked or
1293 Most likely when locking a remote file (e.g., locking over
1294 NFS), but can sometimes happen locally.
1300 is negative or is greater than the maximum allowable value.
1304 is not an allowable signal number.
1309 the process already has the maximum number of file descriptors open.
1312 Too many segment locks open, lock table is full, or a remote locking
1313 protocol failed (e.g., locking over NFS).
1316 Attempted to clear the
1318 flag on a file that has the append-only attribute set.
1320 SVr4, 4.3BSD, POSIX.1-2001.
1331 are specified in POSIX.1-2001.
1336 are specified in POSIX.1-2001.
1337 (To get their definitions, define
1341 with the value 500 or greater, or define
1343 with the value 200809L or greater.)
1346 is specified in POSIX.1-2008.
1347 (To get this definition, define
1349 with the value 200809L or greater, or
1351 with the value 700 or greater.)
1366 macro to obtain these definitions.)
1368 .\" SVr4 documents additional EIO, ENOLINK and EOVERFLOW error conditions.
1370 The errors returned by
1372 are different from those returned by
1378 system call was not designed to handle large file offsets
1384 system call was added in Linux 2.4.
1385 The newer system call employs a different structure for file locking,
1387 and corresponding commands,
1392 However, these details can be ignored by applications using glibc, whose
1394 wrapper function transparently employs the more recent system call
1395 where it is available.
1397 The errors returned by
1399 are different from those returned by
1402 Since kernel 2.0, there is no interaction between the types of lock
1408 Several systems have more fields in
1410 such as, for example,
1412 .\" e.g., Solaris 8 documents this field in fcntl(2), and Irix 6.5
1413 .\" documents it in fcntl(5). mtk, May 2007
1414 .\" Also, FreeBSD documents it (Apr 2014).
1417 alone is not going to be very useful if the process holding the lock
1418 may live on a different machine.
1422 system call was not designed to handle large file offsets
1428 system call was added in Linux 2.4.
1429 The newer system call employs a different structure for file locking,
1431 and corresponding commands,
1436 However, these details can be ignored by applications using glibc, whose
1438 wrapper function transparently employs the more recent system call
1439 where it is available.
1440 .SS Record locking and NFS
1441 Before Linux 3.12, if an NFSv4 client
1442 loses contact with the server for a period of time
1443 (defined as more than 90 seconds with no communication),
1445 .\" Neil Brown: With NFSv3 the failure mode is the reverse. If
1446 .\" the server loses contact with a client then any lock stays in place
1447 .\" indefinitely ("why can't I read my mail"... I remember it well).
1449 it might lose and regain a lock without ever being aware of the fact.
1450 (The period of time after which contact is assumed lost is known as
1451 the NFSv4 leasetime.
1452 On a Linux NFS server, this can be determined by looking at
1453 .IR /proc/fs/nfsd/nfsv4leasetime ,
1454 which expresses the period in seconds.
1455 The default value for this file is 90.)
1458 .\" Note that this is not a firm timeout. The server runs a job
1459 .\" periodically to clean out expired stateful objects, and it's likely
1460 .\" that there is some time (maybe even up to another whole lease period)
1461 .\" between when the timeout expires and the job actually runs. If the
1462 .\" client gets a RENEW in there within that window, its lease will be
1463 .\" renewed and its state preserved.
1465 This scenario potentially risks data corruption,
1466 since another process might acquire a lock in the intervening period
1467 and perform file I/O.
1470 .\" commit ef1820f9be27b6ad158f433ab38002ab8131db4d
1471 if an NFSv4 client loses contact with the server,
1472 any I/O to the file by a process which "thinks" it holds
1473 a lock will fail until that process closes and reopens the file.
1475 .IR nfs.recover_lost_locks ,
1476 can be set to 1 to obtain the pre-3.12 behavior,
1477 whereby the client will attempt to recover lost locks
1478 when contact is reestablished with the server.
1479 Because of the attendant risk of data corruption,
1480 .\" commit f6de7a39c181dfb8a2c534661a53c73afb3081cd
1481 this parameter defaults to 0 (disabled).
1484 It is not possible to use
1486 to change the state of the
1491 .\" FIXME . According to POSIX.1-2001, O_SYNC should also be modifiable
1492 .\" via fcntl(2), but currently Linux does not permit this
1493 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=5994
1494 Attempts to change the state of these flags are silently ignored.
1496 A limitation of the Linux system call conventions on some
1497 architectures (notably i386) means that if a (negative)
1498 process group ID to be returned by
1500 falls in the range \-1 to \-4095, then the return value is wrongly
1501 interpreted by glibc as an error in the system call;
1502 .\" glibc source: sysdeps/unix/sysv/linux/i386/sysdep.h
1503 that is, the return value of
1507 will contain the (positive) process group ID.
1510 operation avoids this problem.
1511 .\" mtk, Dec 04: some limited testing on alpha and ia64 seems to
1512 .\" indicate that ANY negative PGID value will cause F_GETOWN
1513 .\" to misinterpret the return as an error. Some other architectures
1514 .\" seem to have the same range check as i386.
1515 Since glibc version 2.11, glibc makes the kernel
1517 problem invisible by implementing
1522 In Linux 2.4 and earlier, there is bug that can occur
1523 when an unprivileged process uses
1525 to specify the owner
1526 of a socket file descriptor
1527 as a process (group) other than the caller.
1534 even when the owner process (group) is one that the caller
1535 has permission to send signals to.
1536 Despite this error return, the file descriptor owner is set,
1537 and signals will be sent to the owner.
1538 .SS Mandatory locking
1539 The Linux implementation of mandatory locking
1540 is subject to race conditions which render it unreliable:
1541 .\" http://marc.info/?l=linux-kernel&m=119013491707153&w=2
1543 .\" Reconfirmed by Jeff Layton
1544 .\" From: Jeff Layton <jlayton <at> redhat.com>
1545 .\" Subject: Re: Status of fcntl() mandatory locking
1546 .\" Newsgroups: gmane.linux.file-systems
1547 .\" Date: 2014-04-28 10:07:57 GMT
1548 .\" http://thread.gmane.org/gmane.linux.file-systems/84481/focus=84518
1550 call that overlaps with a lock may modify data after the mandatory lock is
1554 call that overlaps with a lock may detect changes to data that were made
1555 only after a write lock was acquired.
1556 Similar races exist between mandatory locks and
1558 It is therefore inadvisable to rely on mandatory locking.
1565 .BR capabilities (7),
1566 .BR feature_test_macros (7)
1569 .IR mandatory-locking.txt ,
1572 in the Linux kernel source directory
1573 .IR Documentation/filesystems/
1574 (on older kernels, these files are directly under the
1577 .I mandatory-locking.txt