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 Michael Kerrisk.
7 .\" %%%LICENSE_START(VERBATIM)
8 .\" Permission is granted to make and distribute verbatim copies of this
9 .\" manual provided the copyright notice and this permission notice are
10 .\" preserved on all copies.
12 .\" Permission is granted to copy and distribute modified versions of this
13 .\" manual under the conditions for verbatim copying, provided that the
14 .\" entire resulting derived work is distributed under the terms of a
15 .\" permission notice identical to this one.
17 .\" Since the Linux kernel and libraries are constantly changing, this
18 .\" manual page may be incorrect or out-of-date. The author(s) assume no
19 .\" responsibility for errors or omissions, or for damages resulting from
20 .\" the use of the information contained herein. The author(s) may not
21 .\" have taken the same level of care in the production of this manual,
22 .\" which is licensed free of charge, as they might when working
25 .\" Formatted or processed versions of this manual, if unaccompanied by
26 .\" the source, must acknowledge the copyright and authors of this work.
29 .\" Modified 1993-07-24 by Rik Faith <faith@cs.unc.edu>
30 .\" Modified 1995-09-26 by Andries Brouwer <aeb@cwi.nl>
31 .\" and again on 960413 and 980804 and 981223.
32 .\" Modified 1998-12-11 by Jamie Lokier <jamie@imbolc.ucc.ie>
33 .\" Applied correction by Christian Ehrhardt - aeb, 990712
34 .\" Modified 2002-04-23 by Michael Kerrisk <mtk.manpages@gmail.com>
35 .\" Added note on F_SETFL and O_DIRECT
36 .\" Complete rewrite + expansion of material on file locking
37 .\" Incorporated description of F_NOTIFY, drawing on
38 .\" Stephen Rothwell's notes in Documentation/dnotify.txt.
39 .\" Added description of F_SETLEASE and F_GETLEASE
40 .\" Corrected and polished, aeb, 020527.
41 .\" Modified 2004-03-03 by Michael Kerrisk <mtk.manpages@gmail.com>
42 .\" Modified description of file leases: fixed some errors of detail
43 .\" Replaced the term "lease contestant" by "lease breaker"
44 .\" Modified, 27 May 2004, Michael Kerrisk <mtk.manpages@gmail.com>
45 .\" Added notes on capability requirements
46 .\" Modified 2004-12-08, added O_NOATIME after note from Martin Pool
47 .\" 2004-12-10, mtk, noted F_GETOWN bug after suggestion from aeb.
48 .\" 2005-04-08 Jamie Lokier <jamie@shareable.org>, mtk
49 .\" Described behavior of F_SETOWN/F_SETSIG in
50 .\" multithreaded processes, and generally cleaned
51 .\" up the discussion of F_SETOWN.
52 .\" 2005-05-20, Johannes Nicolai <johannes.nicolai@hpi.uni-potsdam.de>,
53 .\" mtk: Noted F_SETOWN bug for socket file descriptor in Linux 2.4
54 .\" and earlier. Added text on permissions required to send signal.
55 .\" 2009-09-30, Michael Kerrisk
56 .\" Note obsolete F_SETOWN behavior with threads.
57 .\" Document F_SETOWN_EX and F_GETOWN_EX
58 .\" 2010-06-17, Michael Kerrisk
59 .\" Document F_SETPIPE_SZ and F_GETPIPE_SZ.
61 .TH FCNTL 2 2014-05-21 "Linux" "Linux Programmer's Manual"
63 fcntl \- manipulate file descriptor
66 .B #include <unistd.h>
69 .BI "int fcntl(int " fd ", int " cmd ", ... /* " arg " */ );"
73 performs one of the operations described below on the open file descriptor
75 The operation is determined by
79 can take an optional third argument.
80 Whether or not this argument is required is determined by
82 The required argument type is indicated in parentheses after each
84 name (in most cases, the required type is
86 and we identify the argument using the name
90 is specified if the argument is not required.
91 .SS Duplicating a file descriptor
93 .BR F_DUPFD " (\fIint\fP)"
94 Find the lowest numbered available file descriptor
95 greater than or equal to
97 and make it be a copy of
99 This is different from
101 which uses exactly the descriptor specified.
103 On success, the new descriptor is returned.
109 .BR F_DUPFD_CLOEXEC " (\fIint\fP; since Linux 2.6.24)"
112 but additionally set the
113 close-on-exec flag for the duplicate descriptor.
114 Specifying this flag permits a program to avoid an additional
120 For an explanation of why this flag is useful,
121 see the description of
125 .SS File descriptor flags
126 The following commands manipulate the flags associated with
128 Currently, only one such flag is defined:
130 the close-on-exec flag.
133 bit is 0, the file descriptor will remain open across an
135 otherwise it will be closed.
137 .BR F_GETFD " (\fIvoid\fP)"
138 Read the file descriptor flags;
142 .BR F_SETFD " (\fIint\fP)"
143 Set the file descriptor flags to the value specified by
146 In multithreaded programs, using
149 to set the close-on-exec flag at the same time as another thread performs a
153 is vulnerable to a race condition that may unintentionally leak
154 the file descriptor to the program executed in the child process.
155 See the discussion of the
159 for details and a remedy to the problem.
160 .SS File status flags
161 Each open file description has certain associated status flags,
166 and possibly modified by
168 Duplicated file descriptors
173 etc.) refer to the same open file description, and thus
174 share the same file status flags.
176 The file status flags and their semantics are described in
179 .BR F_GETFL " (\fIvoid\fP)"
180 Get the file access mode and the file status flags;
184 .BR F_SETFL " (\fIint\fP)"
185 Set the file status flags to the value specified by
188 .RB ( O_RDONLY ", " O_WRONLY ", " O_RDWR )
189 and file creation flags
191 .BR O_CREAT ", " O_EXCL ", " O_NOCTTY ", " O_TRUNC )
195 On Linux this command can change only the
203 It is not possible to change the
207 flags; see BUGS, below.
213 are used to acquire, release, and test for the existence of record
214 locks (also known as file-segment or file-region locks).
217 is a pointer to a structure that has at least the following fields
218 (in unspecified order).
224 short l_type; /* Type of lock: F_RDLCK,
226 short l_whence; /* How to interpret l_start:
227 SEEK_SET, SEEK_CUR, SEEK_END */
228 off_t l_start; /* Starting offset for lock */
229 off_t l_len; /* Number of bytes to lock */
230 pid_t l_pid; /* PID of process blocking our lock
238 .IR l_whence ", " l_start ", and " l_len
239 fields of this structure specify the range of bytes we wish to lock.
240 Bytes past the end of the file may be locked,
241 but not bytes before the start of the file.
244 is the starting offset for the lock, and is interpreted
246 the start of the file (if
250 the current file offset (if
254 or the end of the file (if
258 In the final two cases,
260 can be a negative number provided the
261 offset does not lie before the start of the file.
264 specifies the number of bytes to be locked.
267 is positive, then the range to be locked covers bytes
270 .IR l_start + l_len \-1.
273 has the special meaning: lock all bytes starting at the
274 location specified by
275 .IR l_whence " and " l_start
276 through to the end of file, no matter how large the file grows.
278 POSIX.1-2001 allows (but does not require)
279 an implementation to support a negative
283 is negative, the interval described by
289 This is supported by Linux since kernel versions 2.4.21 and 2.5.49.
293 field can be used to place a read
298 Any number of processes may hold a read lock (shared lock)
299 on a file region, but only one process may hold a write lock
301 An exclusive lock excludes all other locks,
302 both shared and exclusive.
303 A single process can hold only one type of lock on a file region;
304 if a new lock is applied to an already-locked region,
305 then the existing lock is converted to the new lock type.
306 (Such conversions may involve splitting, shrinking, or coalescing with
307 an existing lock if the byte range specified by the new lock does not
308 precisely coincide with the range of the existing lock.)
310 .BR F_SETLK " (\fIstruct flock *\fP)"
317 or release a lock (when
321 on the bytes specified by the
322 .IR l_whence ", " l_start ", and " l_len
325 If a conflicting lock is held by another process,
326 this call returns \-1 and sets
333 .BR F_SETLKW " (\fIstruct flock *\fP)"
336 but if a conflicting lock is held on the file, then wait for that
338 If a signal is caught while waiting, then the call is interrupted
339 and (after the signal handler has returned)
340 returns immediately (with return value \-1 and
347 .BR F_GETLK " (\fIstruct flock *\fP)"
348 On input to this call,
350 describes a lock we would like to place on the file.
351 If the lock could be placed,
353 does not actually place it, but returns
359 and leaves the other fields of the structure unchanged.
360 If one or more incompatible locks would prevent
361 this lock being placed, then
363 returns details about one of these locks in the
364 .IR l_type ", " l_whence ", " l_start ", and " l_len
369 to be the PID of the process holding that lock.
370 Note that the information returned by
372 may already be out of date by the time the caller inspects it.
374 In order to place a read lock,
376 must be open for reading.
377 In order to place a write lock,
379 must be open for writing.
380 To place both types of lock, open a file read-write.
382 As well as being removed by an explicit
384 record locks are automatically released when the process
385 terminates or if it closes
387 file descriptor referring to a file on which locks are held.
388 .\" (Additional file descriptors referring to the same file
389 .\" may have been obtained by calls to
390 .\" .BR open "(2), " dup "(2), " dup2 "(2), or " fcntl ().)
391 This is bad: it means that a process can lose the locks on
396 when for some reason a library function decides to open, read
399 Record locks are not inherited by a child created via
401 but are preserved across an
404 Because of the buffering performed by the
406 library, the use of record locking with routines in that package
407 should be avoided; use
412 .SS Mandatory locking
414 The above record locks may be either advisory or mandatory,
415 and are advisory by default.
417 Advisory locks are not enforced and are useful only between
418 cooperating processes.
420 Mandatory locks are enforced for all processes.
421 If a process tries to perform an incompatible access (e.g.,
425 on a file region that has an incompatible mandatory lock,
426 then the result depends upon whether the
428 flag is enabled for its open file description.
431 flag is not enabled, then
432 system call is blocked until the lock is removed
433 or converted to a mode that is compatible with the access.
436 flag is enabled, then the system call fails with the error
439 To make use of mandatory locks, mandatory locking must be enabled
440 both on the filesystem that contains the file to be locked,
441 and on the file itself.
442 Mandatory locking is enabled on a filesystem
443 using the "\-o mand" option to
449 Mandatory locking is enabled on a file by disabling
450 group execute permission on the file and enabling the set-group-ID
456 The Linux implementation of mandatory locking is unreliable.
466 are used to manage I/O availability signals:
468 .BR F_GETOWN " (\fIvoid\fP)"
469 Return (as the function result)
470 the process ID or process group currently receiving
474 signals for events on file descriptor
476 Process IDs are returned as positive values;
477 process group IDs are returned as negative values (but see BUGS below).
481 .BR F_SETOWN " (\fIint\fP)"
482 Set the process ID or process group ID that will receive
486 signals for events on file descriptor
490 A process ID is specified as a positive value;
491 a process group ID is specified as a negative value.
492 Most commonly, the calling process specifies itself as the owner
501 status flag on a file descriptor by using the
507 signal is sent whenever input or output becomes possible
508 on that file descriptor.
510 can be used to obtain delivery of a signal other than
512 If this permission check fails, then the signal is
515 Sending a signal to the owner process (group) specified by
517 is subject to the same permissions checks as are described for
519 where the sending process is the one that employs
521 (but see BUGS below).
523 If the file descriptor
530 signals that are delivered when out-of-band
531 data arrives on that socket.
533 is sent in any situation where
535 would report the socket as having an "exceptional condition".)
536 .\" The following appears to be rubbish. It doesn't seem to
537 .\" be true according to the kernel source, and I can write
538 .\" a program that gets a terminal-generated SIGIO even though
539 .\" it is not the foreground process group of the terminal.
542 .\" If the file descriptor
544 .\" refers to a terminal device, then SIGIO
545 .\" signals are sent to the foreground process group of the terminal.
547 The following was true in 2.6.x kernels up to and including
551 If a nonzero value is given to
553 in a multithreaded process running with a threading library
554 that supports thread groups (e.g., NPTL),
555 then a positive value given to
557 has a different meaning:
558 .\" The relevant place in the (2.6) kernel source is the
559 .\" 'switch' in fs/fcntl.c::send_sigio_to_task() -- MTK, Apr 2005
560 instead of being a process ID identifying a whole process,
561 it is a thread ID identifying a specific thread within a process.
562 Consequently, it may be necessary to pass
568 to get sensible results when
571 (In current Linux threading implementations,
572 a main thread's thread ID is the same as its process ID.
573 This means that a single-threaded program can equally use
578 Note, however, that the statements in this paragraph do not apply
581 signal generated for out-of-band data on a socket:
582 this signal is always sent to either a process or a process group,
583 depending on the value given to
585 .\" send_sigurg()/send_sigurg_to_task() bypasses
586 .\" kill_fasync()/send_sigio()/send_sigio_to_task()
587 .\" to directly call send_group_sig_info()
588 .\" -- MTK, Apr 2005 (kernel 2.6.11)
591 The above behavior was accidentally dropped in Linux 2.6.12,
592 and won't be restored.
593 From Linux 2.6.32 onward, use
599 signals at a particular thread.
601 .BR F_GETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
602 Return the current file descriptor owner settings
603 as defined by a previous
606 The information is returned in the structure pointed to by
608 which has the following form:
621 field will have one of the values
628 field is a positive integer representing a thread ID, process ID,
634 .BR F_SETOWN_EX " (struct f_owner_ex *) (since Linux 2.6.32)"
635 This operation performs a similar task to
637 It allows the caller to direct I/O availability signals
638 to a specific thread, process, or process group.
639 The caller specifies the target of signals via
641 which is a pointer to a
646 field has one of the following values, which define how
652 Send the signal to the thread whose thread ID
653 (the value returned by a call to
661 Send the signal to the process whose ID
666 Send the signal to the process group whose ID
669 (Note that, unlike with
671 a process group ID is specified as a positive value here.)
674 .BR F_GETSIG " (\fIvoid\fP)"
675 Return (as the function result)
676 the signal sent when input or output becomes possible.
677 A value of zero means
680 Any other value (including
683 signal sent instead, and in this case additional info is available to
684 the signal handler if installed with
689 .BR F_SETSIG " (\fIint\fP)"
690 Set the signal sent when input or output becomes possible
691 to the value given in
693 A value of zero means to send the default
696 Any other value (including
698 is the signal to send instead, and in this case additional info
699 is available to the signal handler if installed with
702 .\" The following was true only up until 2.6.11:
704 .\" Additionally, passing a nonzero value to
706 .\" changes the signal recipient from a whole process to a specific thread
707 .\" within a process.
708 .\" See the description of
710 .\" for more details.
714 with a nonzero value, and setting
719 extra information about I/O events is passed to
725 field indicates the source is
729 field gives the file descriptor associated with the event.
731 there is no indication which file descriptors are pending, and you
732 should use the usual mechanisms
738 set etc.) to determine which file descriptors are available for I/O.
740 By selecting a real time signal (value >=
742 multiple I/O events may be queued using the same signal numbers.
743 (Queuing is dependent on available memory).
744 Extra information is available
747 is set for the signal handler, as above.
749 Note that Linux imposes a limit on the
750 number of real-time signals that may be queued to a
755 and if this limit is reached, then the kernel reverts to
758 and this signal is delivered to the entire
759 process rather than to a specific thread.
760 .\" See fs/fcntl.c::send_sigio_to_task() (2.4/2.6) sources -- MTK, Apr 05
762 Using these mechanisms, a program can implement fully asynchronous I/O
771 is specific to BSD and Linux.
776 specified in POSIX.1 is in conjunction with the use of the
779 (POSIX does not specify the
788 POSIX has asynchronous I/O and the
790 structure to achieve similar things; these are also available
791 in Linux as part of the GNU C Library (Glibc).
796 (Linux 2.4 onward) are used (respectively) to establish a new lease,
797 and retrieve the current lease, on the open file description
798 referred to by the file descriptor
800 A file lease provides a mechanism whereby the process holding
801 the lease (the "lease holder") is notified (via delivery of a signal)
802 when a process (the "lease breaker") tries to
806 the file referred to by that file descriptor.
808 .BR F_SETLEASE " (\fIint\fP)"
809 Set or remove a file lease according to which of the following
810 values is specified in the integer
815 Take out a read lease.
816 This will cause the calling process to be notified when
817 the file is opened for writing or is truncated.
818 .\" The following became true in kernel 2.6.10:
819 .\" See the man-pages-2.09 Changelog for further info.
820 A read lease can be placed only on a file descriptor that
824 Take out a write lease.
825 This will cause the caller to be notified when
826 the file is opened for reading or writing or is truncated.
827 A write lease may be placed on a file only if there are no
828 other open file descriptors for the file.
831 Remove our lease from the file.
834 Leases are associated with an open file description (see
836 This means that duplicate file descriptors (created by, for example,
840 refer to the same lease, and this lease may be modified
841 or released using any of these descriptors.
842 Furthermore, the lease is released by either an explicit
844 operation on any of these duplicate descriptors, or when all
845 such descriptors have been closed.
847 Leases may be taken out only on regular files.
848 An unprivileged process may take out a lease only on a file whose
849 UID (owner) matches the filesystem UID of the process.
852 capability may take out leases on arbitrary files.
854 .BR F_GETLEASE " (\fIvoid\fP)"
855 Indicates what type of lease is associated with the file descriptor
858 .BR F_RDLCK ", " F_WRLCK ", or " F_UNLCK ,
859 indicating, respectively, a read lease , a write lease, or no lease.
863 When a process (the "lease breaker") performs an
867 that conflicts with a lease established via
869 the system call is blocked by the kernel and
870 the kernel notifies the lease holder by sending it a signal
873 The lease holder should respond to receipt of this signal by doing
874 whatever cleanup is required in preparation for the file to be
875 accessed by another process (e.g., flushing cached buffers) and
876 then either remove or downgrade its lease.
877 A lease is removed by performing an
883 If the lease holder currently holds a write lease on the file,
884 and the lease breaker is opening the file for reading,
885 then it is sufficient for the lease holder to downgrade
886 the lease to a read lease.
887 This is done by performing an
894 If the lease holder fails to downgrade or remove the lease within
895 the number of seconds specified in
896 .IR /proc/sys/fs/lease-break-time ,
897 then the kernel forcibly removes or downgrades the lease holder's lease.
899 Once a lease break has been initiated,
901 returns the target lease type (either
905 depending on what would be compatible with the lease breaker)
906 until the lease holder voluntarily downgrades or removes the lease or
907 the kernel forcibly does so after the lease break timer expires.
909 Once the lease has been voluntarily or forcibly removed or downgraded,
910 and assuming the lease breaker has not unblocked its system call,
911 the kernel permits the lease breaker's system call to proceed.
913 If the lease breaker's blocked
917 is interrupted by a signal handler,
918 then the system call fails with the error
920 but the other steps still occur as described above.
921 If the lease breaker is killed by a signal while blocked in
925 then the other steps still occur as described above.
926 If the lease breaker specifies the
930 then the call immediately fails with the error
932 but the other steps still occur as described above.
934 The default signal used to notify the lease holder is
936 but this can be changed using the
942 command is performed (even one specifying
945 handler is established using
947 then the handler will receive a
949 structure as its second argument, and the
951 field of this argument will hold the descriptor of the leased file
952 that has been accessed by another process.
953 (This is useful if the caller holds leases against multiple files).
954 .SS File and directory change notification (dnotify)
956 .BR F_NOTIFY " (\fIint\fP)"
958 Provide notification when the directory referred to by
960 or any of the files that it contains is changed.
961 The events to be notified are specified in
963 which is a bit mask specified by ORing together zero or more of
970 A file was accessed (read, pread, readv)
973 A file was modified (write, pwrite, writev, truncate, ftruncate).
976 A file was created (open, creat, mknod, mkdir, link, symlink, rename).
979 A file was unlinked (unlink, rename to another directory, rmdir).
982 A file was renamed within this directory (rename).
985 The attributes of a file were changed (chown, chmod, utime[s]).
989 (In order to obtain these definitions, the
991 feature test macro must be defined before including
995 Directory notifications are normally "one-shot", and the application
996 must reregister to receive further notifications.
1001 then notification will remain in effect until explicitly removed.
1003 .\" The following does seem a poor API-design choice...
1006 requests is cumulative, with the events in
1008 being added to the set already monitored.
1009 To disable notification of all events, make an
1015 Notification occurs via delivery of a signal.
1016 The default signal is
1018 but this can be changed using the
1022 In the latter case, the signal handler receives a
1024 structure as its second argument (if the handler was
1029 field of this structure contains the file descriptor which
1030 generated the notification (useful when establishing notification
1031 on multiple directories).
1033 Especially when using
1035 a real time signal should be used for notification,
1036 so that multiple notifications can be queued.
1039 New applications should use the
1041 interface (available since kernel 2.6.13),
1042 which provides a much superior interface for obtaining notifications of
1046 .SS Changing the capacity of a pipe
1048 .BR F_SETPIPE_SZ " (\fIint\fP; since Linux 2.6.35)"
1049 Change the capacity of the pipe referred to by
1054 An unprivileged process can adjust the pipe capacity to any value
1055 between the system page size and the limit defined in
1056 .IR /proc/sys/fs/pipe-max-size
1059 Attempts to set the pipe capacity below the page size are silently
1060 rounded up to the page size.
1061 Attempts by an unprivileged process to set the pipe capacity above the limit in
1062 .IR /proc/sys/fs/pipe-max-size
1065 a privileged process
1066 .RB ( CAP_SYS_RESOURCE )
1067 can override the limit.
1068 When allocating the buffer for the pipe,
1069 the kernel may use a capacity larger than
1071 if that is convenient for the implementation.
1074 operation returns the actual size used.
1075 Attempting to set the pipe capacity smaller than the amount
1076 of buffer space currently used to store data produces the error
1079 .BR F_GETPIPE_SZ " (\fIvoid\fP; since Linux 2.6.35)"
1080 Return (as the function result) the capacity of the pipe referred to by
1083 For a successful call, the return value depends on the operation:
1089 Value of file descriptor flags.
1092 Value of file status flags.
1095 Type of lease held on file descriptor.
1098 Value of descriptor owner.
1101 Value of signal sent when read or write becomes possible, or zero
1112 On error, \-1 is returned, and
1114 is set appropriately.
1117 .BR EACCES " or " EAGAIN
1118 Operation is prohibited by locks held by other processes.
1121 The operation is prohibited because the file has been memory-mapped by
1126 is not an open file descriptor, or the command was
1130 and the file descriptor open mode doesn't match with the
1131 type of lock requested.
1134 It was detected that the specified
1136 command would cause a deadlock.
1140 is outside your accessible address space.
1145 the command was interrupted by a signal; see
1148 .BR F_GETLK " and " F_SETLK ,
1149 the command was interrupted by a signal before the lock was checked or
1151 Most likely when locking a remote file (e.g., locking over
1152 NFS), but can sometimes happen locally.
1158 is negative or is greater than the maximum allowable value.
1162 is not an allowable signal number.
1167 the process already has the maximum number of file descriptors open.
1170 Too many segment locks open, lock table is full, or a remote locking
1171 protocol failed (e.g., locking over NFS).
1174 Attempted to clear the
1176 flag on a file that has the append-only attribute set.
1178 SVr4, 4.3BSD, POSIX.1-2001.
1189 are specified in POSIX.1-2001.
1194 are specified in POSIX.1-2001.
1195 (To get their definitions, define
1199 with the value 500 or greater, or define
1201 with the value 200809L or greater.)
1204 is specified in POSIX.1-2008.
1205 (To get this definition, define
1207 with the value 200809L or greater, or
1209 with the value 700 or greater.)
1224 macro to obtain these definitions.)
1226 .\" SVr4 documents additional EIO, ENOLINK and EOVERFLOW error conditions.
1228 The errors returned by
1230 are different from those returned by
1236 system call was not designed to handle large file offsets
1242 system call was added in Linux 2.4.
1243 The newer system call employs a different structure for file locking,
1245 and corresponding commands,
1250 However, these details can be ignored by applications using glibc, whose
1252 wrapper function transparently employs the more recent system call
1253 where it is available.
1255 Since kernel 2.0, there is no interaction between the types of lock
1261 Several systems have more fields in
1263 such as, for example,
1265 .\" e.g., Solaris 8 documents this field in fcntl(2), and Irix 6.5
1266 .\" documents it in fcntl(5). mtk, May 2007
1269 alone is not going to be very useful if the process holding the lock
1270 may live on a different machine.
1273 It is not possible to use
1275 to change the state of the
1280 .\" FIXME . According to POSIX.1-2001, O_SYNC should also be modifiable
1281 .\" via fcntl(2), but currently Linux does not permit this
1282 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=5994
1283 Attempts to change the state of these flags are silently ignored.
1285 A limitation of the Linux system call conventions on some
1286 architectures (notably i386) means that if a (negative)
1287 process group ID to be returned by
1289 falls in the range \-1 to \-4095, then the return value is wrongly
1290 interpreted by glibc as an error in the system call;
1291 .\" glibc source: sysdeps/unix/sysv/linux/i386/sysdep.h
1292 that is, the return value of
1296 will contain the (positive) process group ID.
1299 operation avoids this problem.
1300 .\" mtk, Dec 04: some limited testing on alpha and ia64 seems to
1301 .\" indicate that ANY negative PGID value will cause F_GETOWN
1302 .\" to misinterpret the return as an error. Some other architectures
1303 .\" seem to have the same range check as i386.
1304 Since glibc version 2.11, glibc makes the kernel
1306 problem invisible by implementing
1311 In Linux 2.4 and earlier, there is bug that can occur
1312 when an unprivileged process uses
1314 to specify the owner
1315 of a socket file descriptor
1316 as a process (group) other than the caller.
1323 even when the owner process (group) is one that the caller
1324 has permission to send signals to.
1325 Despite this error return, the file descriptor owner is set,
1326 and signals will be sent to the owner.
1327 .SS Mandatory locking
1328 The implementation of mandatory locking in all known versions of Linux
1329 is subject to race conditions which render it unreliable:
1330 .\" http://marc.info/?l=linux-kernel&m=119013491707153&w=2
1333 call that overlaps with a lock may modify data after the mandatory lock is
1337 call that overlaps with a lock may detect changes to data that were made
1338 only after a write lock was acquired.
1339 Similar races exist between mandatory locks and
1341 It is therefore inadvisable to rely on mandatory locking.
1348 .BR capabilities (7),
1349 .BR feature_test_macros (7)
1352 .IR mandatory-locking.txt ,
1355 in the Linux kernel source directory
1356 .IR Documentation/filesystems/
1357 (on older kernels, these files are directly under the
1360 .I mandatory-locking.txt
1364 This page is part of release 3.68 of the Linux
1367 A description of the project,
1368 information about reporting bugs,
1369 and the latest version of this page,
1371 \%http://www.kernel.org/doc/man\-pages/.