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19 .\" manual page may be incorrect or out-of-date.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 2014-07-08 "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.
93 Certain of the operations below are supported only since a particular
95 The preferred method of checking whether the host kernel supports
96 a particular operation is to invoke
100 value and then test whether the call failed with
102 indicating that the kernel does not recognize this value.
103 .SS Duplicating a file descriptor
105 .BR F_DUPFD " (\fIint\fP)"
106 Find the lowest numbered available file descriptor
107 greater than or equal to
109 and make it be a copy of
111 This is different from
113 which uses exactly the descriptor specified.
115 On success, the new descriptor is returned.
121 .BR F_DUPFD_CLOEXEC " (\fIint\fP; since Linux 2.6.24)"
124 but additionally set the
125 close-on-exec flag for the duplicate descriptor.
126 Specifying this flag permits a program to avoid an additional
132 For an explanation of why this flag is useful,
133 see the description of
137 .SS File descriptor flags
138 The following commands manipulate the flags associated with
140 Currently, only one such flag is defined:
142 the close-on-exec flag.
145 bit is 0, the file descriptor will remain open across an
147 otherwise it will be closed.
149 .BR F_GETFD " (\fIvoid\fP)"
150 Read the file descriptor flags;
154 .BR F_SETFD " (\fIint\fP)"
155 Set the file descriptor flags to the value specified by
158 In multithreaded programs, using
161 to set the close-on-exec flag at the same time as another thread performs a
165 is vulnerable to a race condition that may unintentionally leak
166 the file descriptor to the program executed in the child process.
167 See the discussion of the
171 for details and a remedy to the problem.
172 .SS File status flags
173 Each open file description has certain associated status flags,
178 and possibly modified by
180 Duplicated file descriptors
185 etc.) refer to the same open file description, and thus
186 share the same file status flags.
188 The file status flags and their semantics are described in
191 .BR F_GETFL " (\fIvoid\fP)"
192 Get the file access mode and the file status flags;
196 .BR F_SETFL " (\fIint\fP)"
197 Set the file status flags to the value specified by
200 .RB ( O_RDONLY ", " O_WRONLY ", " O_RDWR )
201 and file creation flags
203 .BR O_CREAT ", " O_EXCL ", " O_NOCTTY ", " O_TRUNC )
207 On Linux this command can change only the
215 It is not possible to change the
219 flags; see BUGS, below.
220 .SS Advisory record locking
221 Linux implements traditional ("process-associated") UNIX record locks,
222 as standardized by POSIX.
223 For a Linux-specific alternative with better semantics,
224 see the discussion of open file description locks below.
230 are used to acquire, release, and test for the existence of record
231 locks (also known as byte-range, file-segment, or file-region locks).
234 is a pointer to a structure that has at least the following fields
235 (in unspecified order).
241 short l_type; /* Type of lock: F_RDLCK,
243 short l_whence; /* How to interpret l_start:
244 SEEK_SET, SEEK_CUR, SEEK_END */
245 off_t l_start; /* Starting offset for lock */
246 off_t l_len; /* Number of bytes to lock */
247 pid_t l_pid; /* PID of process blocking our lock
248 (set by F_GETLK and F_OFD_GETLK) */
255 .IR l_whence ", " l_start ", and " l_len
256 fields of this structure specify the range of bytes we wish to lock.
257 Bytes past the end of the file may be locked,
258 but not bytes before the start of the file.
261 is the starting offset for the lock, and is interpreted
263 the start of the file (if
267 the current file offset (if
271 or the end of the file (if
275 In the final two cases,
277 can be a negative number provided the
278 offset does not lie before the start of the file.
281 specifies the number of bytes to be locked.
284 is positive, then the range to be locked covers bytes
287 .IR l_start + l_len \-1.
290 has the special meaning: lock all bytes starting at the
291 location specified by
292 .IR l_whence " and " l_start
293 through to the end of file, no matter how large the file grows.
295 POSIX.1-2001 allows (but does not require)
296 an implementation to support a negative
300 is negative, the interval described by
306 This is supported by Linux since kernel versions 2.4.21 and 2.5.49.
310 field can be used to place a read
315 Any number of processes may hold a read lock (shared lock)
316 on a file region, but only one process may hold a write lock
318 An exclusive lock excludes all other locks,
319 both shared and exclusive.
320 A single process can hold only one type of lock on a file region;
321 if a new lock is applied to an already-locked region,
322 then the existing lock is converted to the new lock type.
323 (Such conversions may involve splitting, shrinking, or coalescing with
324 an existing lock if the byte range specified by the new lock does not
325 precisely coincide with the range of the existing lock.)
327 .BR F_SETLK " (\fIstruct flock *\fP)"
334 or release a lock (when
338 on the bytes specified by the
339 .IR l_whence ", " l_start ", and " l_len
342 If a conflicting lock is held by another process,
343 this call returns \-1 and sets
349 (The error returned in this case differs across implementations,
350 so POSIX requires a portable application to check for both errors.)
352 .BR F_SETLKW " (\fIstruct flock *\fP)"
355 but if a conflicting lock is held on the file, then wait for that
357 If a signal is caught while waiting, then the call is interrupted
358 and (after the signal handler has returned)
359 returns immediately (with return value \-1 and
366 .BR F_GETLK " (\fIstruct flock *\fP)"
367 On input to this call,
369 describes a lock we would like to place on the file.
370 If the lock could be placed,
372 does not actually place it, but returns
378 and leaves the other fields of the structure unchanged.
380 If one or more incompatible locks would prevent
381 this lock being placed, then
383 returns details about one of those locks in the
384 .IR l_type ", " l_whence ", " l_start ", and " l_len
387 If the conflicting lock is a traditional (process-associated) record lock,
390 field is set to the PID of the process holding that lock.
391 If the conflicting lock is an open file description lock, then
394 Note that the returned information
395 may already be out of date by the time the caller inspects it.
397 In order to place a read lock,
399 must be open for reading.
400 In order to place a write lock,
402 must be open for writing.
403 To place both types of lock, open a file read-write.
405 When placing locks with
409 whereby two or more processes have their
410 lock requests mutually blocked by locks held by the other processes.
411 For example, suppose process A holds a write lock on byte 100 of a file,
412 and process B holds a write lock on byte 200.
413 If each process then attempts to lock the byte already
414 locked by the other process using
416 then, without deadlock detection,
417 both processes would remain blocked indefinitely.
418 When the kernel detects such deadlocks,
419 it causes one of the blocking lock requests to immediately fail with the error
421 an application that encounters such an error should release
422 some of its locks to allow other applications to proceed before
423 attempting regain the locks that it requires.
424 Circular deadlocks involving more than two processes are also detected.
425 Note, however, that there are limitations to the kernel's
426 deadlock-detection algorithm; see BUGS.
428 As well as being removed by an explicit
430 record locks are automatically released when the process terminates.
432 Record locks are not inherited by a child created via
434 but are preserved across an
437 Because of the buffering performed by the
439 library, the use of record locking with routines in that package
440 should be avoided; use
446 The record locks described above are associated with the process
447 (unlike the open file description locks described below).
448 This has some unfortunate consequences:
452 file descriptor referring to a file,
453 then all of the process's locks on that file are released,
454 regardless of the file descriptor(s) on which the locks were obtained.
455 .\" (Additional file descriptors referring to the same file
456 .\" may have been obtained by calls to
457 .\" .BR open "(2), " dup "(2), " dup2 "(2), or " fcntl ().)
458 This is bad: it means that a process can lose its locks on
463 when for some reason a library function decides to open, read,
464 and close the same file.
466 The threads in a process share locks.
468 a multithreaded program can't use record locking to ensure
469 that threads don't simultaneously access the same region of a file.
471 Open file description locks solve both of these problems.
472 .SS Open file description locks (non-POSIX)
473 Open file description locks are advisory byte-range locks whose operation is
474 in most respects identical to the traditional record locks described above.
475 This lock type is Linux-specific,
476 and available since Linux 3.15.
477 For an explanation of open file descriptions, see
480 The principal difference between the two lock types
481 is that whereas traditional record locks
482 are associated with a process,
483 open file description locks are associated with the
484 open file description on which they are acquired,
485 much like locks acquired with
487 Consequently (and unlike traditional advisory record locks),
488 open file description locks are inherited across
494 and are only automatically released on the last close
495 of the open file description,
496 instead of being released on any close of the file.
498 Open file description locks always conflict with traditional record locks,
499 even when they are acquired by the same process on the same file descriptor.
501 Open file description locks placed via the same open file description
502 (i.e., via the same file descriptor,
503 or via a duplicate of the file descriptor created by
508 and so on) are always compatible:
509 if a new lock is placed on an already locked region,
510 then the existing lock is converted to the new lock type.
511 (Such conversions may result in splitting, shrinking, or coalescing with
512 an existing lock as discussed above.)
514 On the other hand, open file description locks may conflict with
515 each other when they are acquired via different open file descriptions.
516 Thus, the threads in a multithreaded program can use
517 open file description locks to synchronize access to a file region
518 by having each thread perform its own
520 on the file and applying locks via the resulting file descriptor.
522 As with traditional advisory locks, the third argument to
528 By contrast with traditional record locks, the
530 field of that structure must be set to zero
531 when using the commands described below.
533 The commands for working with open file description locks are analogous
534 to those used with traditional locks:
536 .BR F_OFD_SETLK " (\fIstruct flock *\fP)"
537 Acquire an open file description lock (when
543 or release an open file description lock (when
547 on the bytes specified by the
548 .IR l_whence ", " l_start ", and " l_len
551 If a conflicting lock is held by another process,
552 this call returns \-1 and sets
557 .BR F_OFD_SETLKW " (\fIstruct flock *\fP)"
560 but if a conflicting lock is held on the file, then wait for that lock to be
562 If a signal is caught while waiting, then the call is interrupted
563 and (after the signal handler has returned) returns immediately
564 (with return value \-1 and
571 .BR F_OFD_GETLK " (\fIstruct flock *\fP)"
572 On input to this call,
574 describes an open file description lock we would like to place on the file.
575 If the lock could be placed,
577 does not actually place it, but returns
583 and leaves the other fields of the structure unchanged.
584 If one or more incompatible locks would prevent this lock being placed,
585 then details about one of these locks are returned via
587 as described above for
590 In the current implementation,
591 .\" commit 57b65325fe34ec4c917bc4e555144b4a94d9e1f7
592 no deadlock detection is performed for open file description locks.
593 (This contrasts with process-associated record locks,
594 for which the kernel does perform deadlock detection.)
596 .SS Mandatory locking
598 the Linux implementation of mandatory locking is unreliable.
601 By default, both traditional (process-associated) and open file description
602 record locks are advisory.
603 Advisory locks are not enforced and are useful only between
604 cooperating processes.
606 Both lock types can also be mandatory.
607 Mandatory locks are enforced for all processes.
608 If a process tries to perform an incompatible access (e.g.,
612 on a file region that has an incompatible mandatory lock,
613 then the result depends upon whether the
615 flag is enabled for its open file description.
618 flag is not enabled, then
619 the system call is blocked until the lock is removed
620 or converted to a mode that is compatible with the access.
623 flag is enabled, then the system call fails with the error
626 To make use of mandatory locks, mandatory locking must be enabled
627 both on the filesystem that contains the file to be locked,
628 and on the file itself.
629 Mandatory locking is enabled on a filesystem
630 using the "\-o mand" option to
636 Mandatory locking is enabled on a file by disabling
637 group execute permission on the file and enabling the set-group-ID
643 Mandatory locking is not specified by POSIX.
644 Some other systems also support mandatory locking,
645 although the details of how to enable it vary across systems.
654 are used to manage I/O availability signals:
656 .BR F_GETOWN " (\fIvoid\fP)"
657 Return (as the function result)
658 the process ID or process group currently receiving
662 signals for events on file descriptor
664 Process IDs are returned as positive values;
665 process group IDs are returned as negative values (but see BUGS below).
669 .BR F_SETOWN " (\fIint\fP)"
670 Set the process ID or process group ID that will receive
674 signals for events on file descriptor
678 A process ID is specified as a positive value;
679 a process group ID is specified as a negative value.
680 Most commonly, the calling process specifies itself as the owner
689 status flag on a file descriptor by using the
695 signal is sent whenever input or output becomes possible
696 on that file descriptor.
698 can be used to obtain delivery of a signal other than
700 If this permission check fails, then the signal is
703 Sending a signal to the owner process (group) specified by
705 is subject to the same permissions checks as are described for
707 where the sending process is the one that employs
709 (but see BUGS below).
711 If the file descriptor
718 signals that are delivered when out-of-band
719 data arrives on that socket.
721 is sent in any situation where
723 would report the socket as having an "exceptional condition".)
724 .\" The following appears to be rubbish. It doesn't seem to
725 .\" be true according to the kernel source, and I can write
726 .\" a program that gets a terminal-generated SIGIO even though
727 .\" it is not the foreground process group of the terminal.
730 .\" If the file descriptor
732 .\" refers to a terminal device, then SIGIO
733 .\" signals are sent to the foreground process group of the terminal.
735 The following was true in 2.6.x kernels up to and including
739 If a nonzero value is given to
741 in a multithreaded process running with a threading library
742 that supports thread groups (e.g., NPTL),
743 then a positive value given to
745 has a different meaning:
746 .\" The relevant place in the (2.6) kernel source is the
747 .\" 'switch' in fs/fcntl.c::send_sigio_to_task() -- MTK, Apr 2005
748 instead of being a process ID identifying a whole process,
749 it is a thread ID identifying a specific thread within a process.
750 Consequently, it may be necessary to pass
756 to get sensible results when
759 (In current Linux threading implementations,
760 a main thread's thread ID is the same as its process ID.
761 This means that a single-threaded program can equally use
766 Note, however, that the statements in this paragraph do not apply
769 signal generated for out-of-band data on a socket:
770 this signal is always sent to either a process or a process group,
771 depending on the value given to
773 .\" send_sigurg()/send_sigurg_to_task() bypasses
774 .\" kill_fasync()/send_sigio()/send_sigio_to_task()
775 .\" to directly call send_group_sig_info()
776 .\" -- MTK, Apr 2005 (kernel 2.6.11)
779 The above behavior was accidentally dropped in Linux 2.6.12,
780 and won't be restored.
781 From Linux 2.6.32 onward, use
787 signals at a particular thread.
789 .BR F_GETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
790 Return the current file descriptor owner settings
791 as defined by a previous
794 The information is returned in the structure pointed to by
796 which has the following form:
809 field will have one of the values
816 field is a positive integer representing a thread ID, process ID,
822 .BR F_SETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
823 This operation performs a similar task to
825 It allows the caller to direct I/O availability signals
826 to a specific thread, process, or process group.
827 The caller specifies the target of signals via
829 which is a pointer to a
834 field has one of the following values, which define how
840 Send the signal to the thread whose thread ID
841 (the value returned by a call to
849 Send the signal to the process whose ID
854 Send the signal to the process group whose ID
857 (Note that, unlike with
859 a process group ID is specified as a positive value here.)
862 .BR F_GETSIG " (\fIvoid\fP)"
863 Return (as the function result)
864 the signal sent when input or output becomes possible.
865 A value of zero means
868 Any other value (including
871 signal sent instead, and in this case additional info is available to
872 the signal handler if installed with
877 .BR F_SETSIG " (\fIint\fP)"
878 Set the signal sent when input or output becomes possible
879 to the value given in
881 A value of zero means to send the default
884 Any other value (including
886 is the signal to send instead, and in this case additional info
887 is available to the signal handler if installed with
890 .\" The following was true only up until 2.6.11:
892 .\" Additionally, passing a nonzero value to
894 .\" changes the signal recipient from a whole process to a specific thread
895 .\" within a process.
896 .\" See the description of
898 .\" for more details.
902 with a nonzero value, and setting
907 extra information about I/O events is passed to
913 field indicates the source is
917 field gives the file descriptor associated with the event.
919 there is no indication which file descriptors are pending, and you
920 should use the usual mechanisms
926 set etc.) to determine which file descriptors are available for I/O.
928 By selecting a real time signal (value >=
930 multiple I/O events may be queued using the same signal numbers.
931 (Queuing is dependent on available memory).
932 Extra information is available
935 is set for the signal handler, as above.
937 Note that Linux imposes a limit on the
938 number of real-time signals that may be queued to a
943 and if this limit is reached, then the kernel reverts to
946 and this signal is delivered to the entire
947 process rather than to a specific thread.
948 .\" See fs/fcntl.c::send_sigio_to_task() (2.4/2.6) sources -- MTK, Apr 05
950 Using these mechanisms, a program can implement fully asynchronous I/O
959 is specific to BSD and Linux.
964 specified in POSIX.1 is in conjunction with the use of the
967 (POSIX does not specify the
976 POSIX has asynchronous I/O and the
978 structure to achieve similar things; these are also available
979 in Linux as part of the GNU C Library (Glibc).
984 (Linux 2.4 onward) are used (respectively) to establish a new lease,
985 and retrieve the current lease, on the open file description
986 referred to by the file descriptor
988 A file lease provides a mechanism whereby the process holding
989 the lease (the "lease holder") is notified (via delivery of a signal)
990 when a process (the "lease breaker") tries to
994 the file referred to by that file descriptor.
996 .BR F_SETLEASE " (\fIint\fP)"
997 Set or remove a file lease according to which of the following
998 values is specified in the integer
1003 Take out a read lease.
1004 This will cause the calling process to be notified when
1005 the file is opened for writing or is truncated.
1006 .\" The following became true in kernel 2.6.10:
1007 .\" See the man-pages-2.09 Changelog for further info.
1008 A read lease can be placed only on a file descriptor that
1009 is opened read-only.
1012 Take out a write lease.
1013 This will cause the caller to be notified when
1014 the file is opened for reading or writing or is truncated.
1015 A write lease may be placed on a file only if there are no
1016 other open file descriptors for the file.
1019 Remove our lease from the file.
1022 Leases are associated with an open file description (see
1024 This means that duplicate file descriptors (created by, for example,
1028 refer to the same lease, and this lease may be modified
1029 or released using any of these descriptors.
1030 Furthermore, the lease is released by either an explicit
1032 operation on any of these duplicate descriptors, or when all
1033 such descriptors have been closed.
1035 Leases may be taken out only on regular files.
1036 An unprivileged process may take out a lease only on a file whose
1037 UID (owner) matches the filesystem UID of the process.
1040 capability may take out leases on arbitrary files.
1042 .BR F_GETLEASE " (\fIvoid\fP)"
1043 Indicates what type of lease is associated with the file descriptor
1046 .BR F_RDLCK ", " F_WRLCK ", or " F_UNLCK ,
1047 indicating, respectively, a read lease , a write lease, or no lease.
1051 When a process (the "lease breaker") performs an
1055 that conflicts with a lease established via
1057 the system call is blocked by the kernel and
1058 the kernel notifies the lease holder by sending it a signal
1061 The lease holder should respond to receipt of this signal by doing
1062 whatever cleanup is required in preparation for the file to be
1063 accessed by another process (e.g., flushing cached buffers) and
1064 then either remove or downgrade its lease.
1065 A lease is removed by performing an
1071 If the lease holder currently holds a write lease on the file,
1072 and the lease breaker is opening the file for reading,
1073 then it is sufficient for the lease holder to downgrade
1074 the lease to a read lease.
1075 This is done by performing an
1082 If the lease holder fails to downgrade or remove the lease within
1083 the number of seconds specified in
1084 .IR /proc/sys/fs/lease-break-time ,
1085 then the kernel forcibly removes or downgrades the lease holder's lease.
1087 Once a lease break has been initiated,
1089 returns the target lease type (either
1093 depending on what would be compatible with the lease breaker)
1094 until the lease holder voluntarily downgrades or removes the lease or
1095 the kernel forcibly does so after the lease break timer expires.
1097 Once the lease has been voluntarily or forcibly removed or downgraded,
1098 and assuming the lease breaker has not unblocked its system call,
1099 the kernel permits the lease breaker's system call to proceed.
1101 If the lease breaker's blocked
1105 is interrupted by a signal handler,
1106 then the system call fails with the error
1108 but the other steps still occur as described above.
1109 If the lease breaker is killed by a signal while blocked in
1113 then the other steps still occur as described above.
1114 If the lease breaker specifies the
1118 then the call immediately fails with the error
1120 but the other steps still occur as described above.
1122 The default signal used to notify the lease holder is
1124 but this can be changed using the
1130 command is performed (even one specifying
1133 handler is established using
1135 then the handler will receive a
1137 structure as its second argument, and the
1139 field of this argument will hold the descriptor of the leased file
1140 that has been accessed by another process.
1141 (This is useful if the caller holds leases against multiple files).
1142 .SS File and directory change notification (dnotify)
1144 .BR F_NOTIFY " (\fIint\fP)"
1146 Provide notification when the directory referred to by
1148 or any of the files that it contains is changed.
1149 The events to be notified are specified in
1151 which is a bit mask specified by ORing together zero or more of
1182 into this directory).
1188 to another directory,
1192 A file was renamed within this directory
1196 The attributes of a file were changed
1205 (In order to obtain these definitions, the
1207 feature test macro must be defined before including
1211 Directory notifications are normally "one-shot", and the application
1212 must reregister to receive further notifications.
1217 then notification will remain in effect until explicitly removed.
1219 .\" The following does seem a poor API-design choice...
1222 requests is cumulative, with the events in
1224 being added to the set already monitored.
1225 To disable notification of all events, make an
1231 Notification occurs via delivery of a signal.
1232 The default signal is
1234 but this can be changed using the
1240 is one of the nonqueuing standard signals;
1241 switching to the use of a real-time signal means that
1242 multiple notifications can be queued to the process.)
1243 In the latter case, the signal handler receives a
1245 structure as its second argument (if the handler was
1250 field of this structure contains the file descriptor which
1251 generated the notification (useful when establishing notification
1252 on multiple directories).
1254 Especially when using
1256 a real time signal should be used for notification,
1257 so that multiple notifications can be queued.
1260 New applications should use the
1262 interface (available since kernel 2.6.13),
1263 which provides a much superior interface for obtaining notifications of
1267 .SS Changing the capacity of a pipe
1269 .BR F_SETPIPE_SZ " (\fIint\fP; since Linux 2.6.35)"
1270 Change the capacity of the pipe referred to by
1275 An unprivileged process can adjust the pipe capacity to any value
1276 between the system page size and the limit defined in
1277 .IR /proc/sys/fs/pipe-max-size
1280 Attempts to set the pipe capacity below the page size are silently
1281 rounded up to the page size.
1282 Attempts by an unprivileged process to set the pipe capacity above the limit in
1283 .IR /proc/sys/fs/pipe-max-size
1286 a privileged process
1287 .RB ( CAP_SYS_RESOURCE )
1288 can override the limit.
1289 When allocating the buffer for the pipe,
1290 the kernel may use a capacity larger than
1292 if that is convenient for the implementation.
1293 The actual capacity that is set is returned as the function result.
1294 Attempting to set the pipe capacity smaller than the amount
1295 of buffer space currently used to store data produces the error
1298 .BR F_GETPIPE_SZ " (\fIvoid\fP; since Linux 2.6.35)"
1299 Return (as the function result) the capacity of the pipe referred to by
1302 For a successful call, the return value depends on the operation:
1308 Value of file descriptor flags.
1311 Value of file status flags.
1314 Type of lease held on file descriptor.
1317 Value of descriptor owner.
1320 Value of signal sent when read or write becomes possible, or zero
1325 .BR F_GETPIPE_SZ ", " F_SETPIPE_SZ
1331 On error, \-1 is returned, and
1333 is set appropriately.
1336 .BR EACCES " or " EAGAIN
1337 Operation is prohibited by locks held by other processes.
1340 The operation is prohibited because the file has been memory-mapped by
1345 is not an open file descriptor, or the command was
1349 and the file descriptor open mode doesn't match with the
1350 type of lock requested.
1353 It was detected that the specified
1355 command would cause a deadlock.
1359 is outside your accessible address space.
1364 the command was interrupted by a signal; see
1367 .BR F_GETLK " and " F_SETLK ,
1368 the command was interrupted by a signal before the lock was checked or
1370 Most likely when locking a remote file (e.g., locking over
1371 NFS), but can sometimes happen locally.
1374 The value specified in
1376 is not recognized by this kernel.
1382 is negative or is greater than the maximum allowable value.
1386 is not an allowable signal number.
1397 was not specified as zero.
1402 the process already has the maximum number of file descriptors open.
1405 Too many segment locks open, lock table is full, or a remote locking
1406 protocol failed (e.g., locking over NFS).
1414 does not refer to a directory.
1417 Attempted to clear the
1419 flag on a file that has the append-only attribute set.
1421 SVr4, 4.3BSD, POSIX.1-2001.
1432 are specified in POSIX.1-2001.
1437 are specified in POSIX.1-2001.
1438 (To get their definitions, define either
1442 with the value 500 or greater, or
1444 with the value 200809L or greater.)
1447 is specified in POSIX.1-2008.
1448 (To get this definition, define
1450 with the value 200809L or greater, or
1452 with the value 700 or greater.)
1467 macro to obtain these definitions.)
1469 .\" SVr4 documents additional EIO, ENOLINK and EOVERFLOW error conditions.
1475 are Linux-specific (and one must define
1477 to obtain their definitions),
1478 but work is being done to have them included in the next version of POSIX.1.
1479 .\" FIXME Presumably, glibc will require _GNU_SOURCE
1481 The errors returned by
1483 are different from those returned by
1489 system call was not designed to handle large file offsets
1495 system call was added in Linux 2.4.
1496 The newer system call employs a different structure for file locking,
1498 and corresponding commands,
1503 However, these details can be ignored by applications using glibc, whose
1505 wrapper function transparently employs the more recent system call
1506 where it is available.
1508 The errors returned by
1510 are different from those returned by
1513 Since kernel 2.0, there is no interaction between the types of lock
1519 Several systems have more fields in
1521 such as, for example,
1523 .\" e.g., Solaris 8 documents this field in fcntl(2), and Irix 6.5
1524 .\" documents it in fcntl(5). mtk, May 2007
1525 .\" Also, FreeBSD documents it (Apr 2014).
1528 alone is not going to be very useful if the process holding the lock
1529 may live on a different machine.
1533 system call was not designed to handle large file offsets
1539 system call was added in Linux 2.4.
1540 The newer system call employs a different structure for file locking,
1542 and corresponding commands,
1547 However, these details can be ignored by applications using glibc, whose
1549 wrapper function transparently employs the more recent system call
1550 where it is available.
1551 .SS Record locking and NFS
1552 Before Linux 3.12, if an NFSv4 client
1553 loses contact with the server for a period of time
1554 (defined as more than 90 seconds with no communication),
1556 .\" Neil Brown: With NFSv3 the failure mode is the reverse. If
1557 .\" the server loses contact with a client then any lock stays in place
1558 .\" indefinitely ("why can't I read my mail"... I remember it well).
1560 it might lose and regain a lock without ever being aware of the fact.
1561 (The period of time after which contact is assumed lost is known as
1562 the NFSv4 leasetime.
1563 On a Linux NFS server, this can be determined by looking at
1564 .IR /proc/fs/nfsd/nfsv4leasetime ,
1565 which expresses the period in seconds.
1566 The default value for this file is 90.)
1569 .\" Note that this is not a firm timeout. The server runs a job
1570 .\" periodically to clean out expired stateful objects, and it's likely
1571 .\" that there is some time (maybe even up to another whole lease period)
1572 .\" between when the timeout expires and the job actually runs. If the
1573 .\" client gets a RENEW in there within that window, its lease will be
1574 .\" renewed and its state preserved.
1576 This scenario potentially risks data corruption,
1577 since another process might acquire a lock in the intervening period
1578 and perform file I/O.
1581 .\" commit ef1820f9be27b6ad158f433ab38002ab8131db4d
1582 if an NFSv4 client loses contact with the server,
1583 any I/O to the file by a process which "thinks" it holds
1584 a lock will fail until that process closes and reopens the file.
1586 .IR nfs.recover_lost_locks ,
1587 can be set to 1 to obtain the pre-3.12 behavior,
1588 whereby the client will attempt to recover lost locks
1589 when contact is reestablished with the server.
1590 Because of the attendant risk of data corruption,
1591 .\" commit f6de7a39c181dfb8a2c534661a53c73afb3081cd
1592 this parameter defaults to 0 (disabled).
1595 It is not possible to use
1597 to change the state of the
1602 .\" FIXME . According to POSIX.1-2001, O_SYNC should also be modifiable
1603 .\" via fcntl(2), but currently Linux does not permit this
1604 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=5994
1605 Attempts to change the state of these flags are silently ignored.
1607 A limitation of the Linux system call conventions on some
1608 architectures (notably i386) means that if a (negative)
1609 process group ID to be returned by
1611 falls in the range \-1 to \-4095, then the return value is wrongly
1612 interpreted by glibc as an error in the system call;
1613 .\" glibc source: sysdeps/unix/sysv/linux/i386/sysdep.h
1614 that is, the return value of
1618 will contain the (positive) process group ID.
1621 operation avoids this problem.
1622 .\" mtk, Dec 04: some limited testing on alpha and ia64 seems to
1623 .\" indicate that ANY negative PGID value will cause F_GETOWN
1624 .\" to misinterpret the return as an error. Some other architectures
1625 .\" seem to have the same range check as i386.
1626 Since glibc version 2.11, glibc makes the kernel
1628 problem invisible by implementing
1633 In Linux 2.4 and earlier, there is bug that can occur
1634 when an unprivileged process uses
1636 to specify the owner
1637 of a socket file descriptor
1638 as a process (group) other than the caller.
1645 even when the owner process (group) is one that the caller
1646 has permission to send signals to.
1647 Despite this error return, the file descriptor owner is set,
1648 and signals will be sent to the owner.
1650 .SS Deadlock detection
1651 The deadlock-detection algorithm employed by the kernel when dealing with
1653 requests can yield both
1654 false negatives (failures to detect deadlocks,
1655 leaving a set of deadlocked processes blocked indefinitely)
1658 errors when there is no deadlock).
1660 the kernel limits the lock depth of its dependency search to 10 steps,
1661 meaning that circular deadlock chains that exceed
1662 that size will not be detected.
1663 In addition, the kernel may falsely indicate a deadlock
1664 when two or more processes created using the
1667 flag place locks that appear (to the kernel) to conflict.
1669 .SS Mandatory locking
1670 The Linux implementation of mandatory locking
1671 is subject to race conditions which render it unreliable:
1672 .\" http://marc.info/?l=linux-kernel&m=119013491707153&w=2
1674 .\" Reconfirmed by Jeff Layton
1675 .\" From: Jeff Layton <jlayton <at> redhat.com>
1676 .\" Subject: Re: Status of fcntl() mandatory locking
1677 .\" Newsgroups: gmane.linux.file-systems
1678 .\" Date: 2014-04-28 10:07:57 GMT
1679 .\" http://thread.gmane.org/gmane.linux.file-systems/84481/focus=84518
1682 call that overlaps with a lock may modify data after the mandatory lock is
1686 call that overlaps with a lock may detect changes to data that were made
1687 only after a write lock was acquired.
1688 Similar races exist between mandatory locks and
1690 It is therefore inadvisable to rely on mandatory locking.
1697 .BR capabilities (7),
1698 .BR feature_test_macros (7)
1701 .IR mandatory-locking.txt ,
1704 in the Linux kernel source directory
1705 .IR Documentation/filesystems/
1706 (on older kernels, these files are directly under the
1709 .I mandatory-locking.txt