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
7 .\" and Copyright (C) 2014 David Herrmann
9 .\" %%%LICENSE_START(VERBATIM)
10 .\" Permission is granted to make and distribute verbatim copies of this
11 .\" manual provided the copyright notice and this permission notice are
12 .\" preserved on all copies.
14 .\" Permission is granted to copy and distribute modified versions of this
15 .\" manual under the conditions for verbatim copying, provided that the
16 .\" entire resulting derived work is distributed under the terms of a
17 .\" permission notice identical to this one.
19 .\" Since the Linux kernel and libraries are constantly changing, this
20 .\" manual page may be incorrect or out-of-date. The author(s) assume no
21 .\" responsibility for errors or omissions, or for damages resulting from
22 .\" the use of the information contained herein. The author(s) may not
23 .\" have taken the same level of care in the production of this manual,
24 .\" which is licensed free of charge, as they might when working
27 .\" Formatted or processed versions of this manual, if unaccompanied by
28 .\" the source, must acknowledge the copyright and authors of this work.
31 .\" Modified 1993-07-24 by Rik Faith <faith@cs.unc.edu>
32 .\" Modified 1995-09-26 by Andries Brouwer <aeb@cwi.nl>
33 .\" and again on 960413 and 980804 and 981223.
34 .\" Modified 1998-12-11 by Jamie Lokier <jamie@imbolc.ucc.ie>
35 .\" Applied correction by Christian Ehrhardt - aeb, 990712
36 .\" Modified 2002-04-23 by Michael Kerrisk <mtk.manpages@gmail.com>
37 .\" Added note on F_SETFL and O_DIRECT
38 .\" Complete rewrite + expansion of material on file locking
39 .\" Incorporated description of F_NOTIFY, drawing on
40 .\" Stephen Rothwell's notes in Documentation/dnotify.txt.
41 .\" Added description of F_SETLEASE and F_GETLEASE
42 .\" Corrected and polished, aeb, 020527.
43 .\" Modified 2004-03-03 by Michael Kerrisk <mtk.manpages@gmail.com>
44 .\" Modified description of file leases: fixed some errors of detail
45 .\" Replaced the term "lease contestant" by "lease breaker"
46 .\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
47 .\" Added notes on capability requirements
48 .\" Modified 2004-12-08, added O_NOATIME after note from Martin Pool
49 .\" 2004-12-10, mtk, noted F_GETOWN bug after suggestion from aeb.
50 .\" 2005-04-08 Jamie Lokier <jamie@shareable.org>, mtk
51 .\" Described behavior of F_SETOWN/F_SETSIG in
52 .\" multithreaded processes, and generally cleaned
53 .\" up the discussion of F_SETOWN.
54 .\" 2005-05-20, Johannes Nicolai <johannes.nicolai@hpi.uni-potsdam.de>,
55 .\" mtk: Noted F_SETOWN bug for socket file descriptor in Linux 2.4
56 .\" and earlier. Added text on permissions required to send signal.
57 .\" 2009-09-30, Michael Kerrisk
58 .\" Note obsolete F_SETOWN behavior with threads.
59 .\" Document F_SETOWN_EX and F_GETOWN_EX
60 .\" 2010-06-17, Michael Kerrisk
61 .\" Document F_SETPIPE_SZ and F_GETPIPE_SZ.
62 .\" 2014-07-08, David Herrmann <dh.herrmann@gmail.com>
63 .\" Document F_ADD_SEALS and F_GET_SEALS
65 .TH FCNTL 2 2017-05-03 "Linux" "Linux Programmer's Manual"
67 fcntl \- manipulate file descriptor
70 .B #include <unistd.h>
73 .BI "int fcntl(int " fd ", int " cmd ", ... /* " arg " */ );"
77 performs one of the operations described below on the open file descriptor
79 The operation is determined by
83 can take an optional third argument.
84 Whether or not this argument is required is determined by
86 The required argument type is indicated in parentheses after each
88 name (in most cases, the required type is
90 and we identify the argument using the name
94 is specified if the argument is not required.
96 Certain of the operations below are supported only since a particular
98 The preferred method of checking whether the host kernel supports
99 a particular operation is to invoke
103 value and then test whether the call failed with
105 indicating that the kernel does not recognize this value.
106 .SS Duplicating a file descriptor
108 .BR F_DUPFD " (\fIint\fP)"
109 Duplicate the file descriptor
111 using the lowest-numbered available file descriptor greater than or equal to
113 This is different from
115 which uses exactly the file descriptor specified.
117 On success, the new file descriptor is returned.
123 .BR F_DUPFD_CLOEXEC " (\fIint\fP; since Linux 2.6.24)"
126 but additionally set the
127 close-on-exec flag for the duplicate file descriptor.
128 Specifying this flag permits a program to avoid an additional
134 For an explanation of why this flag is useful,
135 see the description of
139 .SS File descriptor flags
140 The following commands manipulate the flags associated with
142 Currently, only one such flag is defined:
144 the close-on-exec flag.
148 the file descriptor will automatically be closed during a successful
152 fails, the file descriptor is left open.)
155 bit is not set, the file descriptor will remain open across an
158 .BR F_GETFD " (\fIvoid\fP)"
159 Return (as the function result) the file descriptor flags;
163 .BR F_SETFD " (\fIint\fP)"
164 Set the file descriptor flags to the value specified by
167 In multithreaded programs, using
170 to set the close-on-exec flag at the same time as another thread performs a
174 is vulnerable to a race condition that may unintentionally leak
175 the file descriptor to the program executed in the child process.
176 See the discussion of the
180 for details and a remedy to the problem.
181 .SS File status flags
182 Each open file description has certain associated status flags,
187 and possibly modified by
189 Duplicated file descriptors
194 etc.) refer to the same open file description, and thus
195 share the same file status flags.
197 The file status flags and their semantics are described in
200 .BR F_GETFL " (\fIvoid\fP)"
201 Return (as the function result)
202 the file access mode and the file status flags;
206 .BR F_SETFL " (\fIint\fP)"
207 Set the file status flags to the value specified by
210 .RB ( O_RDONLY ", " O_WRONLY ", " O_RDWR )
211 and file creation flags
213 .BR O_CREAT ", " O_EXCL ", " O_NOCTTY ", " O_TRUNC )
217 On Linux, this command can change only the
225 It is not possible to change the
229 flags; see BUGS, below.
230 .SS Advisory record locking
231 Linux implements traditional ("process-associated") UNIX record locks,
232 as standardized by POSIX.
233 For a Linux-specific alternative with better semantics,
234 see the discussion of open file description locks below.
240 are used to acquire, release, and test for the existence of record
241 locks (also known as byte-range, file-segment, or file-region locks).
244 is a pointer to a structure that has at least the following fields
245 (in unspecified order).
251 short l_type; /* Type of lock: F_RDLCK,
253 short l_whence; /* How to interpret l_start:
254 SEEK_SET, SEEK_CUR, SEEK_END */
255 off_t l_start; /* Starting offset for lock */
256 off_t l_len; /* Number of bytes to lock */
257 pid_t l_pid; /* PID of process blocking our lock
258 (set by F_GETLK and F_OFD_GETLK) */
265 .IR l_whence ", " l_start ", and " l_len
266 fields of this structure specify the range of bytes we wish to lock.
267 Bytes past the end of the file may be locked,
268 but not bytes before the start of the file.
271 is the starting offset for the lock, and is interpreted
273 the start of the file (if
277 the current file offset (if
281 or the end of the file (if
285 In the final two cases,
287 can be a negative number provided the
288 offset does not lie before the start of the file.
291 specifies the number of bytes to be locked.
294 is positive, then the range to be locked covers bytes
297 .IR l_start + l_len \-1.
300 has the special meaning: lock all bytes starting at the
301 location specified by
302 .IR l_whence " and " l_start
303 through to the end of file, no matter how large the file grows.
305 POSIX.1-2001 allows (but does not require)
306 an implementation to support a negative
310 is negative, the interval described by
316 This is supported by Linux since kernel versions 2.4.21 and 2.5.49.
320 field can be used to place a read
325 Any number of processes may hold a read lock (shared lock)
326 on a file region, but only one process may hold a write lock
328 An exclusive lock excludes all other locks,
329 both shared and exclusive.
330 A single process can hold only one type of lock on a file region;
331 if a new lock is applied to an already-locked region,
332 then the existing lock is converted to the new lock type.
333 (Such conversions may involve splitting, shrinking, or coalescing with
334 an existing lock if the byte range specified by the new lock does not
335 precisely coincide with the range of the existing lock.)
337 .BR F_SETLK " (\fIstruct flock *\fP)"
344 or release a lock (when
348 on the bytes specified by the
349 .IR l_whence ", " l_start ", and " l_len
352 If a conflicting lock is held by another process,
353 this call returns \-1 and sets
359 (The error returned in this case differs across implementations,
360 so POSIX requires a portable application to check for both errors.)
362 .BR F_SETLKW " (\fIstruct flock *\fP)"
365 but if a conflicting lock is held on the file, then wait for that
367 If a signal is caught while waiting, then the call is interrupted
368 and (after the signal handler has returned)
369 returns immediately (with return value \-1 and
376 .BR F_GETLK " (\fIstruct flock *\fP)"
377 On input to this call,
379 describes a lock we would like to place on the file.
380 If the lock could be placed,
382 does not actually place it, but returns
388 and leaves the other fields of the structure unchanged.
390 If one or more incompatible locks would prevent
391 this lock being placed, then
393 returns details about one of those locks in the
394 .IR l_type ", " l_whence ", " l_start ", and " l_len
397 If the conflicting lock is a traditional (process-associated) record lock,
400 field is set to the PID of the process holding that lock.
401 If the conflicting lock is an open file description lock, then
404 Note that the returned information
405 may already be out of date by the time the caller inspects it.
407 In order to place a read lock,
409 must be open for reading.
410 In order to place a write lock,
412 must be open for writing.
413 To place both types of lock, open a file read-write.
415 When placing locks with
419 whereby two or more processes have their
420 lock requests mutually blocked by locks held by the other processes.
421 For example, suppose process A holds a write lock on byte 100 of a file,
422 and process B holds a write lock on byte 200.
423 If each process then attempts to lock the byte already
424 locked by the other process using
426 then, without deadlock detection,
427 both processes would remain blocked indefinitely.
428 When the kernel detects such deadlocks,
429 it causes one of the blocking lock requests to immediately fail with the error
431 an application that encounters such an error should release
432 some of its locks to allow other applications to proceed before
433 attempting regain the locks that it requires.
434 Circular deadlocks involving more than two processes are also detected.
435 Note, however, that there are limitations to the kernel's
436 deadlock-detection algorithm; see BUGS.
438 As well as being removed by an explicit
440 record locks are automatically released when the process terminates.
442 Record locks are not inherited by a child created via
444 but are preserved across an
447 Because of the buffering performed by the
449 library, the use of record locking with routines in that package
450 should be avoided; use
456 The record locks described above are associated with the process
457 (unlike the open file description locks described below).
458 This has some unfortunate consequences:
462 file descriptor referring to a file,
463 then all of the process's locks on that file are released,
464 regardless of the file descriptor(s) on which the locks were obtained.
465 .\" (Additional file descriptors referring to the same file
466 .\" may have been obtained by calls to
467 .\" .BR open "(2), " dup "(2), " dup2 "(2), or " fcntl ().)
468 This is bad: it means that a process can lose its locks on
473 when for some reason a library function decides to open, read,
474 and close the same file.
476 The threads in a process share locks.
478 a multithreaded program can't use record locking to ensure
479 that threads don't simultaneously access the same region of a file.
481 Open file description locks solve both of these problems.
482 .SS Open file description locks (non-POSIX)
483 Open file description locks are advisory byte-range locks whose operation is
484 in most respects identical to the traditional record locks described above.
485 This lock type is Linux-specific,
486 and available since Linux 3.15.
487 (There is a proposal with the Austin Group
488 .\" FIXME . Review progress into POSIX
489 .\" http://austingroupbugs.net/view.php?id=768
490 to include this lock type in the next revision of POSIX.1.)
491 For an explanation of open file descriptions, see
494 The principal difference between the two lock types
495 is that whereas traditional record locks
496 are associated with a process,
497 open file description locks are associated with the
498 open file description on which they are acquired,
499 much like locks acquired with
501 Consequently (and unlike traditional advisory record locks),
502 open file description locks are inherited across
508 and are only automatically released on the last close
509 of the open file description,
510 instead of being released on any close of the file.
512 Conflicting lock combinations
513 (i.e., a read lock and a write lock or two write locks)
514 where one lock is an open file description lock and the other
515 is a traditional record lock conflict
516 even when they are acquired by the same process on the same file descriptor.
518 Open file description locks placed via the same open file description
519 (i.e., via the same file descriptor,
520 or via a duplicate of the file descriptor created by
525 and so on) are always compatible:
526 if a new lock is placed on an already locked region,
527 then the existing lock is converted to the new lock type.
528 (Such conversions may result in splitting, shrinking, or coalescing with
529 an existing lock as discussed above.)
531 On the other hand, open file description locks may conflict with
532 each other when they are acquired via different open file descriptions.
533 Thus, the threads in a multithreaded program can use
534 open file description locks to synchronize access to a file region
535 by having each thread perform its own
537 on the file and applying locks via the resulting file descriptor.
539 As with traditional advisory locks, the third argument to
545 By contrast with traditional record locks, the
547 field of that structure must be set to zero
548 when using the commands described below.
550 The commands for working with open file description locks are analogous
551 to those used with traditional locks:
553 .BR F_OFD_SETLK " (\fIstruct flock *\fP)"
554 Acquire an open file description lock (when
560 or release an open file description lock (when
564 on the bytes specified by the
565 .IR l_whence ", " l_start ", and " l_len
568 If a conflicting lock is held by another process,
569 this call returns \-1 and sets
574 .BR F_OFD_SETLKW " (\fIstruct flock *\fP)"
577 but if a conflicting lock is held on the file, then wait for that lock to be
579 If a signal is caught while waiting, then the call is interrupted
580 and (after the signal handler has returned) returns immediately
581 (with return value \-1 and
588 .BR F_OFD_GETLK " (\fIstruct flock *\fP)"
589 On input to this call,
591 describes an open file description lock we would like to place on the file.
592 If the lock could be placed,
594 does not actually place it, but returns
600 and leaves the other fields of the structure unchanged.
601 If one or more incompatible locks would prevent this lock being placed,
602 then details about one of these locks are returned via
604 as described above for
607 In the current implementation,
608 .\" commit 57b65325fe34ec4c917bc4e555144b4a94d9e1f7
609 no deadlock detection is performed for open file description locks.
610 (This contrasts with process-associated record locks,
611 for which the kernel does perform deadlock detection.)
613 .SS Mandatory locking
615 the Linux implementation of mandatory locking is unreliable.
617 Because of these bugs,
618 and the fact that the feature is believed to be little used,
619 since Linux 4.5, mandatory locking has been made an optional feature,
620 governed by a configuration option
621 .RB ( CONFIG_MANDATORY_FILE_LOCKING ).
622 This is an initial step toward removing this feature completely.
624 By default, both traditional (process-associated) and open file description
625 record locks are advisory.
626 Advisory locks are not enforced and are useful only between
627 cooperating processes.
629 Both lock types can also be mandatory.
630 Mandatory locks are enforced for all processes.
631 If a process tries to perform an incompatible access (e.g.,
635 on a file region that has an incompatible mandatory lock,
636 then the result depends upon whether the
638 flag is enabled for its open file description.
641 flag is not enabled, then
642 the system call is blocked until the lock is removed
643 or converted to a mode that is compatible with the access.
646 flag is enabled, then the system call fails with the error
649 To make use of mandatory locks, mandatory locking must be enabled
650 both on the filesystem that contains the file to be locked,
651 and on the file itself.
652 Mandatory locking is enabled on a filesystem
653 using the "\-o mand" option to
659 Mandatory locking is enabled on a file by disabling
660 group execute permission on the file and enabling the set-group-ID
666 Mandatory locking is not specified by POSIX.
667 Some other systems also support mandatory locking,
668 although the details of how to enable it vary across systems.
677 are used to manage I/O availability signals:
679 .BR F_GETOWN " (\fIvoid\fP)"
680 Return (as the function result)
681 the process ID or process group currently receiving
685 signals for events on file descriptor
687 Process IDs are returned as positive values;
688 process group IDs are returned as negative values (but see BUGS below).
692 .BR F_SETOWN " (\fIint\fP)"
693 Set the process ID or process group ID that will receive
697 signals for events on the file descriptor
699 The target process or process group ID is specified in
701 A process ID is specified as a positive value;
702 a process group ID is specified as a negative value.
703 Most commonly, the calling process specifies itself as the owner
709 As well as setting the file descriptor owner,
710 one must also enable generation of signals on the file descriptor.
711 This is done by using the
716 file status flag on the file descriptor.
719 signal is sent whenever input or output becomes possible
720 on the file descriptor.
724 command can be used to obtain delivery of a signal other than
727 Sending a signal to the owner process (group) specified by
729 is subject to the same permissions checks as are described for
731 where the sending process is the one that employs
733 (but see BUGS below).
734 If this permission check fails, then the signal is
739 operation records the caller's credentials at the time of the
742 and it is these saved credentials that are used for the permission checks.
744 If the file descriptor
751 signals that are delivered when out-of-band
752 data arrives on that socket.
754 is sent in any situation where
756 would report the socket as having an "exceptional condition".)
757 .\" The following appears to be rubbish. It doesn't seem to
758 .\" be true according to the kernel source, and I can write
759 .\" a program that gets a terminal-generated SIGIO even though
760 .\" it is not the foreground process group of the terminal.
763 .\" If the file descriptor
765 .\" refers to a terminal device, then SIGIO
766 .\" signals are sent to the foreground process group of the terminal.
768 The following was true in 2.6.x kernels up to and including
772 If a nonzero value is given to
774 in a multithreaded process running with a threading library
775 that supports thread groups (e.g., NPTL),
776 then a positive value given to
778 has a different meaning:
779 .\" The relevant place in the (2.6) kernel source is the
780 .\" 'switch' in fs/fcntl.c::send_sigio_to_task() -- MTK, Apr 2005
781 instead of being a process ID identifying a whole process,
782 it is a thread ID identifying a specific thread within a process.
783 Consequently, it may be necessary to pass
789 to get sensible results when
792 (In current Linux threading implementations,
793 a main thread's thread ID is the same as its process ID.
794 This means that a single-threaded program can equally use
799 Note, however, that the statements in this paragraph do not apply
802 signal generated for out-of-band data on a socket:
803 this signal is always sent to either a process or a process group,
804 depending on the value given to
806 .\" send_sigurg()/send_sigurg_to_task() bypasses
807 .\" kill_fasync()/send_sigio()/send_sigio_to_task()
808 .\" to directly call send_group_sig_info()
809 .\" -- MTK, Apr 2005 (kernel 2.6.11)
812 The above behavior was accidentally dropped in Linux 2.6.12,
813 and won't be restored.
814 From Linux 2.6.32 onward, use
820 signals at a particular thread.
822 .BR F_GETOWN_EX " (\fIstruct f_owner_ex *\fP) (since Linux 2.6.32)"
823 Return the current file descriptor owner settings
824 as defined by a previous
827 The information is returned in the structure pointed to by
829 which has the following form:
842 field will have one of the values
849 field is a positive integer representing a thread ID, process ID,
855 .BR F_SETOWN_EX " (\fIstruct f_owner_ex *\fP) (since Linux 2.6.32)"
856 This operation performs a similar task to
858 It allows the caller to direct I/O availability signals
859 to a specific thread, process, or process group.
860 The caller specifies the target of signals via
862 which is a pointer to a
867 field has one of the following values, which define how
873 Send the signal to the thread whose thread ID
874 (the value returned by a call to
882 Send the signal to the process whose ID
887 Send the signal to the process group whose ID
890 (Note that, unlike with
892 a process group ID is specified as a positive value here.)
895 .BR F_GETSIG " (\fIvoid\fP)"
896 Return (as the function result)
897 the signal sent when input or output becomes possible.
898 A value of zero means
901 Any other value (including
904 signal sent instead, and in this case additional info is available to
905 the signal handler if installed with
910 .BR F_SETSIG " (\fIint\fP)"
911 Set the signal sent when input or output becomes possible
912 to the value given in
914 A value of zero means to send the default
917 Any other value (including
919 is the signal to send instead, and in this case additional info
920 is available to the signal handler if installed with
923 .\" The following was true only up until 2.6.11:
925 .\" Additionally, passing a nonzero value to
927 .\" changes the signal recipient from a whole process to a specific thread
928 .\" within a process.
929 .\" See the description of
931 .\" for more details.
935 with a nonzero value, and setting
940 extra information about I/O events is passed to
946 field indicates the source is
950 field gives the file descriptor associated with the event.
952 there is no indication which file descriptors are pending, and you
953 should use the usual mechanisms
959 set etc.) to determine which file descriptors are available for I/O.
961 Note that the file descriptor provided in
963 is the one that was specified during the
966 This can lead to an unusual corner case.
967 If the file descriptor is duplicated
969 or similar), and the original file descriptor is closed,
970 then I/O events will continue to be generated, but the
972 field will contain the number of the now closed file descriptor.
974 By selecting a real time signal (value >=
976 multiple I/O events may be queued using the same signal numbers.
977 (Queuing is dependent on available memory.)
978 Extra information is available
981 is set for the signal handler, as above.
983 Note that Linux imposes a limit on the
984 number of real-time signals that may be queued to a
989 and if this limit is reached, then the kernel reverts to
992 and this signal is delivered to the entire
993 process rather than to a specific thread.
994 .\" See fs/fcntl.c::send_sigio_to_task() (2.4/2.6) sources -- MTK, Apr 05
996 Using these mechanisms, a program can implement fully asynchronous I/O
1005 is specific to BSD and Linux.
1010 specified in POSIX.1 is in conjunction with the use of the
1013 (POSIX does not specify the
1022 POSIX has asynchronous I/O and the
1024 structure to achieve similar things; these are also available
1025 in Linux as part of the GNU C Library (Glibc).
1030 (Linux 2.4 onward) are used (respectively) to establish a new lease,
1031 and retrieve the current lease, on the open file description
1032 referred to by the file descriptor
1034 A file lease provides a mechanism whereby the process holding
1035 the lease (the "lease holder") is notified (via delivery of a signal)
1036 when a process (the "lease breaker") tries to
1040 the file referred to by that file descriptor.
1042 .BR F_SETLEASE " (\fIint\fP)"
1043 Set or remove a file lease according to which of the following
1044 values is specified in the integer
1049 Take out a read lease.
1050 This will cause the calling process to be notified when
1051 the file is opened for writing or is truncated.
1052 .\" The following became true in kernel 2.6.10:
1053 .\" See the man-pages-2.09 Changelog for further info.
1054 A read lease can be placed only on a file descriptor that
1055 is opened read-only.
1058 Take out a write lease.
1059 This will cause the caller to be notified when
1060 the file is opened for reading or writing or is truncated.
1061 A write lease may be placed on a file only if there are no
1062 other open file descriptors for the file.
1065 Remove our lease from the file.
1068 Leases are associated with an open file description (see
1070 This means that duplicate file descriptors (created by, for example,
1074 refer to the same lease, and this lease may be modified
1075 or released using any of these descriptors.
1076 Furthermore, the lease is released by either an explicit
1078 operation on any of these duplicate file descriptors, or when all
1079 such file descriptors have been closed.
1081 Leases may be taken out only on regular files.
1082 An unprivileged process may take out a lease only on a file whose
1083 UID (owner) matches the filesystem UID of the process.
1086 capability may take out leases on arbitrary files.
1088 .BR F_GETLEASE " (\fIvoid\fP)"
1089 Indicates what type of lease is associated with the file descriptor
1092 .BR F_RDLCK ", " F_WRLCK ", or " F_UNLCK ,
1093 indicating, respectively, a read lease , a write lease, or no lease.
1097 When a process (the "lease breaker") performs an
1101 that conflicts with a lease established via
1103 the system call is blocked by the kernel and
1104 the kernel notifies the lease holder by sending it a signal
1107 The lease holder should respond to receipt of this signal by doing
1108 whatever cleanup is required in preparation for the file to be
1109 accessed by another process (e.g., flushing cached buffers) and
1110 then either remove or downgrade its lease.
1111 A lease is removed by performing an
1117 If the lease holder currently holds a write lease on the file,
1118 and the lease breaker is opening the file for reading,
1119 then it is sufficient for the lease holder to downgrade
1120 the lease to a read lease.
1121 This is done by performing an
1128 If the lease holder fails to downgrade or remove the lease within
1129 the number of seconds specified in
1130 .IR /proc/sys/fs/lease-break-time ,
1131 then the kernel forcibly removes or downgrades the lease holder's lease.
1133 Once a lease break has been initiated,
1135 returns the target lease type (either
1139 depending on what would be compatible with the lease breaker)
1140 until the lease holder voluntarily downgrades or removes the lease or
1141 the kernel forcibly does so after the lease break timer expires.
1143 Once the lease has been voluntarily or forcibly removed or downgraded,
1144 and assuming the lease breaker has not unblocked its system call,
1145 the kernel permits the lease breaker's system call to proceed.
1147 If the lease breaker's blocked
1151 is interrupted by a signal handler,
1152 then the system call fails with the error
1154 but the other steps still occur as described above.
1155 If the lease breaker is killed by a signal while blocked in
1159 then the other steps still occur as described above.
1160 If the lease breaker specifies the
1164 then the call immediately fails with the error
1166 but the other steps still occur as described above.
1168 The default signal used to notify the lease holder is
1170 but this can be changed using the
1176 command is performed (even one specifying
1179 handler is established using
1181 then the handler will receive a
1183 structure as its second argument, and the
1185 field of this argument will hold the file descriptor of the leased file
1186 that has been accessed by another process.
1187 (This is useful if the caller holds leases against multiple files.)
1188 .SS File and directory change notification (dnotify)
1190 .BR F_NOTIFY " (\fIint\fP)"
1192 Provide notification when the directory referred to by
1194 or any of the files that it contains is changed.
1195 The events to be notified are specified in
1197 which is a bit mask specified by ORing together zero or more of
1228 into this directory).
1234 to another directory,
1238 A file was renamed within this directory
1242 The attributes of a file were changed
1251 (In order to obtain these definitions, the
1253 feature test macro must be defined before including
1257 Directory notifications are normally "one-shot", and the application
1258 must reregister to receive further notifications.
1263 then notification will remain in effect until explicitly removed.
1265 .\" The following does seem a poor API-design choice...
1268 requests is cumulative, with the events in
1270 being added to the set already monitored.
1271 To disable notification of all events, make an
1277 Notification occurs via delivery of a signal.
1278 The default signal is
1280 but this can be changed using the
1286 is one of the nonqueuing standard signals;
1287 switching to the use of a real-time signal means that
1288 multiple notifications can be queued to the process.)
1289 In the latter case, the signal handler receives a
1291 structure as its second argument (if the handler was
1296 field of this structure contains the file descriptor which
1297 generated the notification (useful when establishing notification
1298 on multiple directories).
1300 Especially when using
1302 a real time signal should be used for notification,
1303 so that multiple notifications can be queued.
1306 New applications should use the
1308 interface (available since kernel 2.6.13),
1309 which provides a much superior interface for obtaining notifications of
1313 .SS Changing the capacity of a pipe
1315 .BR F_SETPIPE_SZ " (\fIint\fP; since Linux 2.6.35)"
1316 Change the capacity of the pipe referred to by
1321 An unprivileged process can adjust the pipe capacity to any value
1322 between the system page size and the limit defined in
1323 .IR /proc/sys/fs/pipe-max-size
1326 Attempts to set the pipe capacity below the page size are silently
1327 rounded up to the page size.
1328 Attempts by an unprivileged process to set the pipe capacity above the limit in
1329 .IR /proc/sys/fs/pipe-max-size
1332 a privileged process
1333 .RB ( CAP_SYS_RESOURCE )
1334 can override the limit.
1336 When allocating the buffer for the pipe,
1337 the kernel may use a capacity larger than
1339 if that is convenient for the implementation.
1340 (In the current implementation,
1341 the allocation is the next higher power-of-two page-size multiple
1342 of the requested size.)
1343 The actual capacity (in bytes) that is set is returned as the function result.
1345 Attempting to set the pipe capacity smaller than the amount
1346 of buffer space currently used to store data produces the error
1349 .BR F_GETPIPE_SZ " (\fIvoid\fP; since Linux 2.6.35)"
1350 Return (as the function result) the capacity of the pipe referred to by
1354 File seals limit the set of allowed operations on a given file.
1355 For each seal that is set on a file,
1356 a specific set of operations will fail with
1358 on this file from now on.
1359 The file is said to be sealed.
1360 The default set of seals depends on the type of the underlying
1361 file and filesystem.
1362 For an overview of file sealing, a discussion of its purpose,
1363 and some code examples, see
1364 .BR memfd_create (2).
1367 file seals can be applied only to a file descriptor returned by
1368 .BR memfd_create (2)
1370 .B MFD_ALLOW_SEALING
1372 On other filesystems, all
1374 operations that operate on seals will return
1377 Seals are a property of an inode.
1378 Thus, all open file descriptors referring to the same inode share
1379 the same set of seals.
1380 Furthermore, seals can never be removed, only added.
1382 .BR F_ADD_SEALS " (\fIint\fP; since Linux 3.17)"
1383 Add the seals given in the bit-mask argument
1385 to the set of seals of the inode referred to by the file descriptor
1387 Seals cannot be removed again.
1388 Once this call succeeds, the seals are enforced by the kernel immediately.
1389 If the current set of seals includes
1391 (see below), then this call will be rejected with
1393 Adding a seal that is already set is a no-op, in case
1396 In order to place a seal, the file descriptor
1400 .BR F_GET_SEALS " (\fIvoid\fP; since Linux 3.17)"
1401 Return (as the function result) the current set of seals
1402 of the inode referred to by
1404 If no seals are set, 0 is returned.
1405 If the file does not support sealing, \-1 is returned and
1410 The following seals are available:
1413 If this seal is set, any further call to
1419 Therefore, this seal prevents any modifications to the set of seals itself.
1420 If the initial set of seals of a file includes
1422 then this effectively causes the set of seals to be constant and locked.
1425 If this seal is set, the file in question cannot be reduced in size.
1434 Those calls will fail with
1436 if you try to shrink the file in question.
1437 Increasing the file size is still possible.
1440 If this seal is set, the size of the file in question cannot be increased.
1443 beyond the end of the file,
1448 These calls will fail with
1450 if you use them to increase the file size.
1451 If you keep the size or shrink it, those calls still work as expected.
1454 If this seal is set, you cannot modify the contents of the file.
1455 Note that shrinking or growing the size of the file is
1456 still possible and allowed.
1457 .\" One or more other seals are typically used with F_SEAL_WRITE
1458 .\" because, given a file with the F_SEAL_WRITE seal set, then,
1459 .\" while it would no longer be possible to (say) write zeros into
1460 .\" the last 100 bytes of a file, it would still be possible
1461 .\" to (say) shrink the file by 100 bytes using ftruncate(), and
1462 .\" then increase the file size by 100 bytes, which would have
1463 .\" the effect of replacing the last hundred bytes by zeros.
1465 Thus, this seal is normally used in combination with one of the other seals.
1470 (only in combination with the
1471 .B FALLOC_FL_PUNCH_HOLE
1473 Those calls will fail with
1475 if this seal is set.
1476 Furthermore, trying to create new shared, writable memory-mappings via
1483 operation to set the
1487 if any writable, shared mapping exists.
1488 Such mappings must be unmapped before you can add this seal.
1489 Furthermore, if there are any asynchronous I/O operations
1490 .RB ( io_submit (2))
1491 pending on the file,
1492 all outstanding writes will be discarded.
1494 For a successful call, the return value depends on the operation:
1497 The new file descriptor.
1500 Value of file descriptor flags.
1503 Value of file status flags.
1506 Type of lease held on file descriptor.
1509 Value of file descriptor owner.
1512 Value of signal sent when read or write becomes possible, or zero
1517 .BR F_GETPIPE_SZ ", " F_SETPIPE_SZ
1521 A bit mask identifying the seals that have been set
1522 for the inode referred to by
1528 On error, \-1 is returned, and
1530 is set appropriately.
1533 .BR EACCES " or " EAGAIN
1534 Operation is prohibited by locks held by other processes.
1537 The operation is prohibited because the file has been memory-mapped by
1542 is not an open file descriptor
1550 and the file descriptor open mode doesn't match with the
1551 type of lock requested.
1557 and the new pipe capacity specified in
1559 is smaller than the amount of buffer space currently
1560 used to store data in the pipe.
1569 and there exists a writable, shared mapping on the file referred to by
1573 It was detected that the specified
1575 command would cause a deadlock.
1579 is outside your accessible address space.
1587 and the operation was interrupted by a signal; see
1598 and the operation was interrupted by a signal before the lock was checked or
1600 Most likely when locking a remote file (e.g., locking over
1601 NFS), but can sometimes happen locally.
1604 The value specified in
1606 is not recognized by this kernel.
1614 includes an unrecognized sealing bit.
1622 and the filesystem containing the inode referred to by
1624 does not support sealing.
1632 is negative or is greater than the maximum allowable value
1633 (see the discussion of
1644 is not an allowable signal number.
1655 was not specified as zero.
1661 and the per-process limit on the number of open file descriptors
1665 Too many segment locks open, lock table is full, or a remote locking
1666 protocol failed (e.g., locking over NFS).
1674 does not refer to a directory.
1680 and the soft or hard user pipe limit has been reached; see
1684 Attempted to clear the
1686 flag on a file that has the append-only attribute set.
1694 was not open for writing
1695 or the current set of seals on the file already includes
1698 SVr4, 4.3BSD, POSIX.1-2001.
1709 are specified in POSIX.1-2001.
1714 are specified in POSIX.1-2001.
1715 (To get their definitions, define either
1716 .\" .BR _BSD_SOURCE ,
1719 with the value 500 or greater, or
1721 with the value 200809L or greater.)
1724 is specified in POSIX.1-2008.
1725 (To get this definition, define
1727 with the value 200809L or greater, or
1729 with the value 700 or greater.)
1744 macro to obtain these definitions.)
1746 .\" SVr4 documents additional EIO, ENOLINK and EOVERFLOW error conditions.
1752 are Linux-specific (and one must define
1754 to obtain their definitions),
1755 but work is being done to have them included in the next version of POSIX.1.
1761 .\" FIXME . Once glibc adds support, add a note about FTM requirements
1763 The errors returned by
1765 are different from those returned by
1771 system call was not designed to handle large file offsets
1777 system call was added in Linux 2.4.
1778 The newer system call employs a different structure for file locking,
1780 and corresponding commands,
1785 However, these details can be ignored by applications using glibc, whose
1787 wrapper function transparently employs the more recent system call
1788 where it is available.
1791 Since kernel 2.0, there is no interaction between the types of lock
1797 Several systems have more fields in
1799 such as, for example,
1801 .\" e.g., Solaris 8 documents this field in fcntl(2), and Irix 6.5
1802 .\" documents it in fcntl(5). mtk, May 2007
1803 .\" Also, FreeBSD documents it (Apr 2014).
1806 alone is not going to be very useful if the process holding the lock
1807 may live on a different machine.
1811 system call was not designed to handle large file offsets
1817 system call was added in Linux 2.4.
1818 The newer system call employs a different structure for file locking,
1820 and corresponding commands,
1825 However, these details can be ignored by applications using glibc, whose
1827 wrapper function transparently employs the more recent system call
1828 where it is available.
1829 .SS Record locking and NFS
1830 Before Linux 3.12, if an NFSv4 client
1831 loses contact with the server for a period of time
1832 (defined as more than 90 seconds with no communication),
1834 .\" Neil Brown: With NFSv3 the failure mode is the reverse. If
1835 .\" the server loses contact with a client then any lock stays in place
1836 .\" indefinitely ("why can't I read my mail"... I remember it well).
1838 it might lose and regain a lock without ever being aware of the fact.
1839 (The period of time after which contact is assumed lost is known as
1840 the NFSv4 leasetime.
1841 On a Linux NFS server, this can be determined by looking at
1842 .IR /proc/fs/nfsd/nfsv4leasetime ,
1843 which expresses the period in seconds.
1844 The default value for this file is 90.)
1847 .\" Note that this is not a firm timeout. The server runs a job
1848 .\" periodically to clean out expired stateful objects, and it's likely
1849 .\" that there is some time (maybe even up to another whole lease period)
1850 .\" between when the timeout expires and the job actually runs. If the
1851 .\" client gets a RENEW in there within that window, its lease will be
1852 .\" renewed and its state preserved.
1854 This scenario potentially risks data corruption,
1855 since another process might acquire a lock in the intervening period
1856 and perform file I/O.
1859 .\" commit ef1820f9be27b6ad158f433ab38002ab8131db4d
1860 if an NFSv4 client loses contact with the server,
1861 any I/O to the file by a process which "thinks" it holds
1862 a lock will fail until that process closes and reopens the file.
1864 .IR nfs.recover_lost_locks ,
1865 can be set to 1 to obtain the pre-3.12 behavior,
1866 whereby the client will attempt to recover lost locks
1867 when contact is reestablished with the server.
1868 Because of the attendant risk of data corruption,
1869 .\" commit f6de7a39c181dfb8a2c534661a53c73afb3081cd
1870 this parameter defaults to 0 (disabled).
1873 It is not possible to use
1875 to change the state of the
1880 .\" FIXME . According to POSIX.1-2001, O_SYNC should also be modifiable
1881 .\" via fcntl(2), but currently Linux does not permit this
1882 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=5994
1883 Attempts to change the state of these flags are silently ignored.
1885 A limitation of the Linux system call conventions on some
1886 architectures (notably i386) means that if a (negative)
1887 process group ID to be returned by
1889 falls in the range \-1 to \-4095, then the return value is wrongly
1890 interpreted by glibc as an error in the system call;
1891 .\" glibc source: sysdeps/unix/sysv/linux/i386/sysdep.h
1892 that is, the return value of
1896 will contain the (positive) process group ID.
1899 operation avoids this problem.
1900 .\" mtk, Dec 04: some limited testing on alpha and ia64 seems to
1901 .\" indicate that ANY negative PGID value will cause F_GETOWN
1902 .\" to misinterpret the return as an error. Some other architectures
1903 .\" seem to have the same range check as i386.
1904 Since glibc version 2.11, glibc makes the kernel
1906 problem invisible by implementing
1911 In Linux 2.4 and earlier, there is bug that can occur
1912 when an unprivileged process uses
1914 to specify the owner
1915 of a socket file descriptor
1916 as a process (group) other than the caller.
1923 even when the owner process (group) is one that the caller
1924 has permission to send signals to.
1925 Despite this error return, the file descriptor owner is set,
1926 and signals will be sent to the owner.
1928 .SS Deadlock detection
1929 The deadlock-detection algorithm employed by the kernel when dealing with
1931 requests can yield both
1932 false negatives (failures to detect deadlocks,
1933 leaving a set of deadlocked processes blocked indefinitely)
1936 errors when there is no deadlock).
1938 the kernel limits the lock depth of its dependency search to 10 steps,
1939 meaning that circular deadlock chains that exceed
1940 that size will not be detected.
1941 In addition, the kernel may falsely indicate a deadlock
1942 when two or more processes created using the
1945 flag place locks that appear (to the kernel) to conflict.
1947 .SS Mandatory locking
1948 The Linux implementation of mandatory locking
1949 is subject to race conditions which render it unreliable:
1950 .\" http://marc.info/?l=linux-kernel&m=119013491707153&w=2
1952 .\" Reconfirmed by Jeff Layton
1953 .\" From: Jeff Layton <jlayton <at> redhat.com>
1954 .\" Subject: Re: Status of fcntl() mandatory locking
1955 .\" Newsgroups: gmane.linux.file-systems
1956 .\" Date: 2014-04-28 10:07:57 GMT
1957 .\" http://thread.gmane.org/gmane.linux.file-systems/84481/focus=84518
1960 call that overlaps with a lock may modify data after the mandatory lock is
1964 call that overlaps with a lock may detect changes to data that were made
1965 only after a write lock was acquired.
1966 Similar races exist between mandatory locks and
1968 It is therefore inadvisable to rely on mandatory locking.
1975 .BR capabilities (7),
1976 .BR feature_test_macros (7),
1980 .IR mandatory-locking.txt ,
1983 in the Linux kernel source directory
1984 .IR Documentation/filesystems/
1985 (on older kernels, these files are directly under the
1988 .I mandatory-locking.txt