1fcntl(2) System Calls Manual fcntl(2)
2
6 fcntl - manipulate file descriptor
7
9 Standard C library (libc, -lc)
10
12 #include <fcntl.h>
13 int fcntl(int fd, int cmd, ... /* arg */ );
15
17 fcntl() performs one of the operations described below on the open file
18 descriptor fd. The operation is determined by cmd.
19 fcntl() can take an optional third argument. Whether or not this argu‐
21 ment is required is determined by cmd. The required argument type is
22 indicated in parentheses after each cmd name (in most cases, the re‐
23 quired type is int, and we identify the argument using the name arg),
24 or void is specified if the argument is not required.
25 Certain of the operations below are supported only since a particular
27 Linux kernel version. The preferred method of checking whether the
28 host kernel supports a particular operation is to invoke fcntl() with
29 the desired cmd value and then test whether the call failed with EIN‐
30 VAL, indicating that the kernel does not recognize this value.
31 Duplicating a file descriptor
33 F_DUPFD (int)
34 Duplicate the file descriptor fd using the lowest-numbered
35 available file descriptor greater than or equal to arg. This is
36 different from dup2(2), which uses exactly the file descriptor
37 specified.
38 On success, the new file descriptor is returned.
40 See dup(2) for further details.
42 F_DUPFD_CLOEXEC (int; since Linux 2.6.24)
44 As for F_DUPFD, but additionally set the close-on-exec flag for
45 the duplicate file descriptor. Specifying this flag permits a
46 program to avoid an additional fcntl() F_SETFD operation to set
47 the FD_CLOEXEC flag. For an explanation of why this flag is
48 useful, see the description of O_CLOEXEC in open(2).
49 File descriptor flags
51 The following commands manipulate the flags associated with a file de‐
52 scriptor. Currently, only one such flag is defined: FD_CLOEXEC, the
53 close-on-exec flag. If the FD_CLOEXEC bit is set, the file descriptor
54 will automatically be closed during a successful execve(2). (If the
55 execve(2) fails, the file descriptor is left open.) If the FD_CLOEXEC
56 bit is not set, the file descriptor will remain open across an ex‐
57 ecve(2).
58 F_GETFD (void)
60 Return (as the function result) the file descriptor flags; arg
61 is ignored.
62 F_SETFD (int)
64 Set the file descriptor flags to the value specified by arg.
65 In multithreaded programs, using fcntl() F_SETFD to set the close-on-
67 exec flag at the same time as another thread performs a fork(2) plus
68 execve(2) is vulnerable to a race condition that may unintentionally
69 leak the file descriptor to the program executed in the child process.
70 See the discussion of the O_CLOEXEC flag in open(2) for details and a
71 remedy to the problem.
72 File status flags
74 Each open file description has certain associated status flags, ini‐
75 tialized by open(2) and possibly modified by fcntl(). Duplicated file
76 descriptors (made with dup(2), fcntl(F_DUPFD), fork(2), etc.) refer to
77 the same open file description, and thus share the same file status
78 flags.
79 The file status flags and their semantics are described in open(2).
81 F_GETFL (void)
83 Return (as the function result) the file access mode and the
84 file status flags; arg is ignored.
85 F_SETFL (int)
87 Set the file status flags to the value specified by arg. File
88 access mode (O_RDONLY, O_WRONLY, O_RDWR) and file creation flags
89 (i.e., O_CREAT, O_EXCL, O_NOCTTY, O_TRUNC) in arg are ignored.
90 On Linux, this command can change only the O_APPEND, O_ASYNC,
91 O_DIRECT, O_NOATIME, and O_NONBLOCK flags. It is not possible
92 to change the O_DSYNC and O_SYNC flags; see BUGS, below.
93 Advisory record locking
95 Linux implements traditional ("process-associated") UNIX record locks,
96 as standardized by POSIX. For a Linux-specific alternative with better
97 semantics, see the discussion of open file description locks below.
98 F_SETLK, F_SETLKW, and F_GETLK are used to acquire, release, and test
100 for the existence of record locks (also known as byte-range, file-seg‐
101 ment, or file-region locks). The third argument, lock, is a pointer to
102 a structure that has at least the following fields (in unspecified or‐
103 der).
104 struct flock {
106 ...
107 short l_type; /* Type of lock: F_RDLCK,
108 F_WRLCK, F_UNLCK */
109 short l_whence; /* How to interpret l_start:
110 SEEK_SET, SEEK_CUR, SEEK_END */
111 off_t l_start; /* Starting offset for lock */
112 off_t l_len; /* Number of bytes to lock */
113 pid_t l_pid; /* PID of process blocking our lock
114 (set by F_GETLK and F_OFD_GETLK) */
115 ...
116 };
117 The l_whence, l_start, and l_len fields of this structure specify the
119 range of bytes we wish to lock. Bytes past the end of the file may be
120 locked, but not bytes before the start of the file.
121 l_start is the starting offset for the lock, and is interpreted rela‐
123 tive to either: the start of the file (if l_whence is SEEK_SET); the
124 current file offset (if l_whence is SEEK_CUR); or the end of the file
125 (if l_whence is SEEK_END). In the final two cases, l_start can be a
126 negative number provided the offset does not lie before the start of
127 the file.
128 l_len specifies the number of bytes to be locked. If l_len is posi‐
130 tive, then the range to be locked covers bytes l_start up to and in‐
131 cluding l_start+l_len-1. Specifying 0 for l_len has the special mean‐
132 ing: lock all bytes starting at the location specified by l_whence and
133 l_start through to the end of file, no matter how large the file grows.
134 POSIX.1-2001 allows (but does not require) an implementation to support
136 a negative l_len value; if l_len is negative, the interval described by
137 lock covers bytes l_start+l_len up to and including l_start-1. This is
138 supported since Linux 2.4.21 and Linux 2.5.49.
139 The l_type field can be used to place a read (F_RDLCK) or a write
141 (F_WRLCK) lock on a file. Any number of processes may hold a read lock
142 (shared lock) on a file region, but only one process may hold a write
143 lock (exclusive lock). An exclusive lock excludes all other locks,
144 both shared and exclusive. A single process can hold only one type of
145 lock on a file region; if a new lock is applied to an already-locked
146 region, then the existing lock is converted to the new lock type.
147 (Such conversions may involve splitting, shrinking, or coalescing with
148 an existing lock if the byte range specified by the new lock does not
149 precisely coincide with the range of the existing lock.)
150 F_SETLK (struct flock *)
152 Acquire a lock (when l_type is F_RDLCK or F_WRLCK) or release a
153 lock (when l_type is F_UNLCK) on the bytes specified by the
154 l_whence, l_start, and l_len fields of lock. If a conflicting
155 lock is held by another process, this call returns -1 and sets
156 errno to EACCES or EAGAIN. (The error returned in this case
157 differs across implementations, so POSIX requires a portable ap‐
158 plication to check for both errors.)
159 F_SETLKW (struct flock *)
161 As for F_SETLK, but if a conflicting lock is held on the file,
162 then wait for that lock to be released. If a signal is caught
163 while waiting, then the call is interrupted and (after the sig‐
164 nal handler has returned) returns immediately (with return value
165 -1 and errno set to EINTR; see signal(7)).
166 F_GETLK (struct flock *)
168 On input to this call, lock describes a lock we would like to
169 place on the file. If the lock could be placed, fcntl() does
170 not actually place it, but returns F_UNLCK in the l_type field
171 of lock and leaves the other fields of the structure unchanged.
172 If one or more incompatible locks would prevent this lock being
174 placed, then fcntl() returns details about one of those locks in
175 the l_type, l_whence, l_start, and l_len fields of lock. If the
176 conflicting lock is a traditional (process-associated) record
177 lock, then the l_pid field is set to the PID of the process
178 holding that lock. If the conflicting lock is an open file de‐
179 scription lock, then l_pid is set to -1. Note that the returned
180 information may already be out of date by the time the caller
181 inspects it.
182 In order to place a read lock, fd must be open for reading. In order
184 to place a write lock, fd must be open for writing. To place both
185 types of lock, open a file read-write.
186 When placing locks with F_SETLKW, the kernel detects deadlocks, whereby
188 two or more processes have their lock requests mutually blocked by
189 locks held by the other processes. For example, suppose process A
190 holds a write lock on byte 100 of a file, and process B holds a write
191 lock on byte 200. If each process then attempts to lock the byte al‐
192 ready locked by the other process using F_SETLKW, then, without dead‐
193 lock detection, both processes would remain blocked indefinitely. When
194 the kernel detects such deadlocks, it causes one of the blocking lock
195 requests to immediately fail with the error EDEADLK; an application
196 that encounters such an error should release some of its locks to allow
197 other applications to proceed before attempting regain the locks that
198 it requires. Circular deadlocks involving more than two processes are
199 also detected. Note, however, that there are limitations to the ker‐
200 nel's deadlock-detection algorithm; see BUGS.
201 As well as being removed by an explicit F_UNLCK, record locks are auto‐
203 matically released when the process terminates.
204 Record locks are not inherited by a child created via fork(2), but are
206 preserved across an execve(2).
207 Because of the buffering performed by the stdio(3) library, the use of
209 record locking with routines in that package should be avoided; use
210 read(2) and write(2) instead.
211 The record locks described above are associated with the process (un‐
213 like the open file description locks described below). This has some
214 unfortunate consequences:
215 • If a process closes any file descriptor referring to a file, then
217 all of the process's locks on that file are released, regardless of
218 the file descriptor(s) on which the locks were obtained. This is
219 bad: it means that a process can lose its locks on a file such as
220 /etc/passwd or /etc/mtab when for some reason a library function de‐
221 cides to open, read, and close the same file.
222 • The threads in a process share locks. In other words, a multi‐
224 threaded program can't use record locking to ensure that threads
225 don't simultaneously access the same region of a file.
226 Open file description locks solve both of these problems.
228 Open file description locks (non-POSIX)
230 Open file description locks are advisory byte-range locks whose opera‐
231 tion is in most respects identical to the traditional record locks de‐
232 scribed above. This lock type is Linux-specific, and available since
233 Linux 3.15. (There is a proposal with the Austin Group to include this
234 lock type in the next revision of POSIX.1.) For an explanation of open
235 file descriptions, see open(2).
236 The principal difference between the two lock types is that whereas
238 traditional record locks are associated with a process, open file de‐
239 scription locks are associated with the open file description on which
240 they are acquired, much like locks acquired with flock(2). Conse‐
241 quently (and unlike traditional advisory record locks), open file de‐
242 scription locks are inherited across fork(2) (and clone(2) with
243 CLONE_FILES), and are only automatically released on the last close of
244 the open file description, instead of being released on any close of
245 the file.
246 Conflicting lock combinations (i.e., a read lock and a write lock or
248 two write locks) where one lock is an open file description lock and
249 the other is a traditional record lock conflict even when they are ac‐
250 quired by the same process on the same file descriptor.
251 Open file description locks placed via the same open file description
253 (i.e., via the same file descriptor, or via a duplicate of the file de‐
254 scriptor created by fork(2), dup(2), fcntl() F_DUPFD, and so on) are
255 always compatible: if a new lock is placed on an already locked region,
256 then the existing lock is converted to the new lock type. (Such con‐
257 versions may result in splitting, shrinking, or coalescing with an ex‐
258 isting lock as discussed above.)
259 On the other hand, open file description locks may conflict with each
261 other when they are acquired via different open file descriptions.
262 Thus, the threads in a multithreaded program can use open file descrip‐
263 tion locks to synchronize access to a file region by having each thread
264 perform its own open(2) on the file and applying locks via the result‐
265 ing file descriptor.
266 As with traditional advisory locks, the third argument to fcntl(),
268 lock, is a pointer to an flock structure. By contrast with traditional
269 record locks, the l_pid field of that structure must be set to zero
270 when using the commands described below.
271 The commands for working with open file description locks are analogous
273 to those used with traditional locks:
274 F_OFD_SETLK (struct flock *)
276 Acquire an open file description lock (when l_type is F_RDLCK or
277 F_WRLCK) or release an open file description lock (when l_type
278 is F_UNLCK) on the bytes specified by the l_whence, l_start, and
279 l_len fields of lock. If a conflicting lock is held by another
280 process, this call returns -1 and sets errno to EAGAIN.
281 F_OFD_SETLKW (struct flock *)
283 As for F_OFD_SETLK, but if a conflicting lock is held on the
284 file, then wait for that lock to be released. If a signal is
285 caught while waiting, then the call is interrupted and (after
286 the signal handler has returned) returns immediately (with re‐
287 turn value -1 and errno set to EINTR; see signal(7)).
288 F_OFD_GETLK (struct flock *)
290 On input to this call, lock describes an open file description
291 lock we would like to place on the file. If the lock could be
292 placed, fcntl() does not actually place it, but returns F_UNLCK
293 in the l_type field of lock and leaves the other fields of the
294 structure unchanged. If one or more incompatible locks would
295 prevent this lock being placed, then details about one of these
296 locks are returned via lock, as described above for F_GETLK.
297 In the current implementation, no deadlock detection is performed for
299 open file description locks. (This contrasts with process-associated
300 record locks, for which the kernel does perform deadlock detection.)
301 Mandatory locking
303 Warning: the Linux implementation of mandatory locking is unreliable.
304 See BUGS below. Because of these bugs, and the fact that the feature
305 is believed to be little used, since Linux 4.5, mandatory locking has
306 been made an optional feature, governed by a configuration option (CON‐
307 FIG_MANDATORY_FILE_LOCKING). This feature is no longer supported at
308 all in Linux 5.15 and above.
309 By default, both traditional (process-associated) and open file de‐
311 scription record locks are advisory. Advisory locks are not enforced
312 and are useful only between cooperating processes.
313 Both lock types can also be mandatory. Mandatory locks are enforced
315 for all processes. If a process tries to perform an incompatible ac‐
316 cess (e.g., read(2) or write(2)) on a file region that has an incompat‐
317 ible mandatory lock, then the result depends upon whether the O_NON‐
318 BLOCK flag is enabled for its open file description. If the O_NONBLOCK
319 flag is not enabled, then the system call is blocked until the lock is
320 removed or converted to a mode that is compatible with the access. If
321 the O_NONBLOCK flag is enabled, then the system call fails with the er‐
322 ror EAGAIN.
323 To make use of mandatory locks, mandatory locking must be enabled both
325 on the filesystem that contains the file to be locked, and on the file
326 itself. Mandatory locking is enabled on a filesystem using the "-o
327 mand" option to mount(8), or the MS_MANDLOCK flag for mount(2). Manda‐
328 tory locking is enabled on a file by disabling group execute permission
329 on the file and enabling the set-group-ID permission bit (see chmod(1)
330 and chmod(2)).
331 Mandatory locking is not specified by POSIX. Some other systems also
333 support mandatory locking, although the details of how to enable it
334 vary across systems.
335 Lost locks
337 When an advisory lock is obtained on a networked filesystem such as NFS
338 it is possible that the lock might get lost. This may happen due to
339 administrative action on the server, or due to a network partition
340 (i.e., loss of network connectivity with the server) which lasts long
341 enough for the server to assume that the client is no longer function‐
342 ing.
343 When the filesystem determines that a lock has been lost, future
345 read(2) or write(2) requests may fail with the error EIO. This error
346 will persist until the lock is removed or the file descriptor is
347 closed. Since Linux 3.12, this happens at least for NFSv4 (including
348 all minor versions).
349 Some versions of UNIX send a signal (SIGLOST) in this circumstance.
351 Linux does not define this signal, and does not provide any asynchro‐
352 nous notification of lost locks.
353 Managing signals
355 F_GETOWN, F_SETOWN, F_GETOWN_EX, F_SETOWN_EX, F_GETSIG, and F_SETSIG
356 are used to manage I/O availability signals:
357 F_GETOWN (void)
359 Return (as the function result) the process ID or process group
360 ID currently receiving SIGIO and SIGURG signals for events on
361 file descriptor fd. Process IDs are returned as positive val‐
362 ues; process group IDs are returned as negative values (but see
363 BUGS below). arg is ignored.
364 F_SETOWN (int)
366 Set the process ID or process group ID that will receive SIGIO
367 and SIGURG signals for events on the file descriptor fd. The
368 target process or process group ID is specified in arg. A
369 process ID is specified as a positive value; a process group ID
370 is specified as a negative value. Most commonly, the calling
371 process specifies itself as the owner (that is, arg is specified
372 as getpid(2)).
373 As well as setting the file descriptor owner, one must also en‐
375 able generation of signals on the file descriptor. This is done
376 by using the fcntl() F_SETFL command to set the O_ASYNC file
377 status flag on the file descriptor. Subsequently, a SIGIO sig‐
378 nal is sent whenever input or output becomes possible on the
379 file descriptor. The fcntl() F_SETSIG command can be used to
380 obtain delivery of a signal other than SIGIO.
381 Sending a signal to the owner process (group) specified by F_SE‐
383 TOWN is subject to the same permissions checks as are described
384 for kill(2), where the sending process is the one that employs
385 F_SETOWN (but see BUGS below). If this permission check fails,
386 then the signal is silently discarded. Note: The F_SETOWN oper‐
387 ation records the caller's credentials at the time of the fc‐
388 ntl() call, and it is these saved credentials that are used for
389 the permission checks.
390 If the file descriptor fd refers to a socket, F_SETOWN also se‐
392 lects the recipient of SIGURG signals that are delivered when
393 out-of-band data arrives on that socket. (SIGURG is sent in any
394 situation where select(2) would report the socket as having an
395 "exceptional condition".)
396 The following was true in Linux 2.6.x up to and including Linux
398 2.6.11:
399 If a nonzero value is given to F_SETSIG in a multi‐
401 threaded process running with a threading library that
402 supports thread groups (e.g., NPTL), then a positive
403 value given to F_SETOWN has a different meaning: instead
404 of being a process ID identifying a whole process, it is
405 a thread ID identifying a specific thread within a
406 process. Consequently, it may be necessary to pass F_SE‐
407 TOWN the result of gettid(2) instead of getpid(2) to get
408 sensible results when F_SETSIG is used. (In current
409 Linux threading implementations, a main thread's thread
410 ID is the same as its process ID. This means that a sin‐
411 gle-threaded program can equally use gettid(2) or get‐
412 pid(2) in this scenario.) Note, however, that the state‐
413 ments in this paragraph do not apply to the SIGURG signal
414 generated for out-of-band data on a socket: this signal
415 is always sent to either a process or a process group,
416 depending on the value given to F_SETOWN.
417 The above behavior was accidentally dropped in Linux 2.6.12, and
419 won't be restored. From Linux 2.6.32 onward, use F_SETOWN_EX to
420 target SIGIO and SIGURG signals at a particular thread.
421 F_GETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
423 Return the current file descriptor owner settings as defined by
424 a previous F_SETOWN_EX operation. The information is returned
425 in the structure pointed to by arg, which has the following
426 form:
427 struct f_owner_ex {
429 int type;
430 pid_t pid;
431 };
432 The type field will have one of the values F_OWNER_TID,
434 F_OWNER_PID, or F_OWNER_PGRP. The pid field is a positive inte‐
435 ger representing a thread ID, process ID, or process group ID.
436 See F_SETOWN_EX for more details.
437 F_SETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
439 This operation performs a similar task to F_SETOWN. It allows
440 the caller to direct I/O availability signals to a specific
441 thread, process, or process group. The caller specifies the
442 target of signals via arg, which is a pointer to a f_owner_ex
443 structure. The type field has one of the following values,
444 which define how pid is interpreted:
445 F_OWNER_TID
447 Send the signal to the thread whose thread ID (the value
448 returned by a call to clone(2) or gettid(2)) is specified
449 in pid.
450 F_OWNER_PID
452 Send the signal to the process whose ID is specified in
453 pid.
454 F_OWNER_PGRP
456 Send the signal to the process group whose ID is speci‐
457 fied in pid. (Note that, unlike with F_SETOWN, a process
458 group ID is specified as a positive value here.)
459 F_GETSIG (void)
461 Return (as the function result) the signal sent when input or
462 output becomes possible. A value of zero means SIGIO is sent.
463 Any other value (including SIGIO) is the signal sent instead,
464 and in this case additional info is available to the signal han‐
465 dler if installed with SA_SIGINFO. arg is ignored.
466 F_SETSIG (int)
468 Set the signal sent when input or output becomes possible to the
469 value given in arg. A value of zero means to send the default
470 SIGIO signal. Any other value (including SIGIO) is the signal
471 to send instead, and in this case additional info is available
472 to the signal handler if installed with SA_SIGINFO.
473 By using F_SETSIG with a nonzero value, and setting SA_SIGINFO
475 for the signal handler (see sigaction(2)), extra information
476 about I/O events is passed to the handler in a siginfo_t struc‐
477 ture. If the si_code field indicates the source is SI_SIGIO,
478 the si_fd field gives the file descriptor associated with the
479 event. Otherwise, there is no indication which file descriptors
480 are pending, and you should use the usual mechanisms (select(2),
481 poll(2), read(2) with O_NONBLOCK set etc.) to determine which
482 file descriptors are available for I/O.
483 Note that the file descriptor provided in si_fd is the one that
485 was specified during the F_SETSIG operation. This can lead to
486 an unusual corner case. If the file descriptor is duplicated
487 (dup(2) or similar), and the original file descriptor is closed,
488 then I/O events will continue to be generated, but the si_fd
489 field will contain the number of the now closed file descriptor.
490 By selecting a real time signal (value >= SIGRTMIN), multiple
492 I/O events may be queued using the same signal numbers. (Queu‐
493 ing is dependent on available memory.) Extra information is
494 available if SA_SIGINFO is set for the signal handler, as above.
495 Note that Linux imposes a limit on the number of real-time sig‐
497 nals that may be queued to a process (see getrlimit(2) and sig‐
498 nal(7)) and if this limit is reached, then the kernel reverts to
499 delivering SIGIO, and this signal is delivered to the entire
500 process rather than to a specific thread.
501 Using these mechanisms, a program can implement fully asynchronous I/O
503 without using select(2) or poll(2) most of the time.
504 The use of O_ASYNC is specific to BSD and Linux. The only use of
506 F_GETOWN and F_SETOWN specified in POSIX.1 is in conjunction with the
507 use of the SIGURG signal on sockets. (POSIX does not specify the SIGIO
508 signal.) F_GETOWN_EX, F_SETOWN_EX, F_GETSIG, and F_SETSIG are Linux-
509 specific. POSIX has asynchronous I/O and the aio_sigevent structure to
510 achieve similar things; these are also available in Linux as part of
511 the GNU C Library (glibc).
512 Leases
514 F_SETLEASE and F_GETLEASE (Linux 2.4 onward) are used to establish a
515 new lease, and retrieve the current lease, on the open file description
516 referred to by the file descriptor fd. A file lease provides a mecha‐
517 nism whereby the process holding the lease (the "lease holder") is no‐
518 tified (via delivery of a signal) when a process (the "lease breaker")
519 tries to open(2) or truncate(2) the file referred to by that file de‐
520 scriptor.
521 F_SETLEASE (int)
523 Set or remove a file lease according to which of the following
524 values is specified in the integer arg:
525 F_RDLCK
527 Take out a read lease. This will cause the calling
528 process to be notified when the file is opened for writ‐
529 ing or is truncated. A read lease can be placed only on
530 a file descriptor that is opened read-only.
531 F_WRLCK
533 Take out a write lease. This will cause the caller to be
534 notified when the file is opened for reading or writing
535 or is truncated. A write lease may be placed on a file
536 only if there are no other open file descriptors for the
537 file.
538 F_UNLCK
540 Remove our lease from the file.
541 Leases are associated with an open file description (see open(2)).
543 This means that duplicate file descriptors (created by, for example,
544 fork(2) or dup(2)) refer to the same lease, and this lease may be modi‐
545 fied or released using any of these descriptors. Furthermore, the
546 lease is released by either an explicit F_UNLCK operation on any of
547 these duplicate file descriptors, or when all such file descriptors
548 have been closed.
549 Leases may be taken out only on regular files. An unprivileged process
551 may take out a lease only on a file whose UID (owner) matches the
552 filesystem UID of the process. A process with the CAP_LEASE capability
553 may take out leases on arbitrary files.
554 F_GETLEASE (void)
556 Indicates what type of lease is associated with the file de‐
557 scriptor fd by returning either F_RDLCK, F_WRLCK, or F_UNLCK,
558 indicating, respectively, a read lease , a write lease, or no
559 lease. arg is ignored.
560 When a process (the "lease breaker") performs an open(2) or truncate(2)
562 that conflicts with a lease established via F_SETLEASE, the system call
563 is blocked by the kernel and the kernel notifies the lease holder by
564 sending it a signal (SIGIO by default). The lease holder should re‐
565 spond to receipt of this signal by doing whatever cleanup is required
566 in preparation for the file to be accessed by another process (e.g.,
567 flushing cached buffers) and then either remove or downgrade its lease.
568 A lease is removed by performing an F_SETLEASE command specifying arg
569 as F_UNLCK. If the lease holder currently holds a write lease on the
570 file, and the lease breaker is opening the file for reading, then it is
571 sufficient for the lease holder to downgrade the lease to a read lease.
572 This is done by performing an F_SETLEASE command specifying arg as
573 F_RDLCK.
574 If the lease holder fails to downgrade or remove the lease within the
576 number of seconds specified in /proc/sys/fs/lease-break-time, then the
577 kernel forcibly removes or downgrades the lease holder's lease.
578 Once a lease break has been initiated, F_GETLEASE returns the target
580 lease type (either F_RDLCK or F_UNLCK, depending on what would be com‐
581 patible with the lease breaker) until the lease holder voluntarily
582 downgrades or removes the lease or the kernel forcibly does so after
583 the lease break timer expires.
584 Once the lease has been voluntarily or forcibly removed or downgraded,
586 and assuming the lease breaker has not unblocked its system call, the
587 kernel permits the lease breaker's system call to proceed.
588 If the lease breaker's blocked open(2) or truncate(2) is interrupted by
590 a signal handler, then the system call fails with the error EINTR, but
591 the other steps still occur as described above. If the lease breaker
592 is killed by a signal while blocked in open(2) or truncate(2), then the
593 other steps still occur as described above. If the lease breaker spec‐
594 ifies the O_NONBLOCK flag when calling open(2), then the call immedi‐
595 ately fails with the error EWOULDBLOCK, but the other steps still occur
596 as described above.
597 The default signal used to notify the lease holder is SIGIO, but this
599 can be changed using the F_SETSIG command to fcntl(). If a F_SETSIG
600 command is performed (even one specifying SIGIO), and the signal han‐
601 dler is established using SA_SIGINFO, then the handler will receive a
602 siginfo_t structure as its second argument, and the si_fd field of this
603 argument will hold the file descriptor of the leased file that has been
604 accessed by another process. (This is useful if the caller holds
605 leases against multiple files.)
606 File and directory change notification (dnotify)
608 F_NOTIFY (int)
609 (Linux 2.4 onward) Provide notification when the directory re‐
610 ferred to by fd or any of the files that it contains is changed.
611 The events to be notified are specified in arg, which is a bit
612 mask specified by ORing together zero or more of the following
613 bits:
614 DN_ACCESS
616 A file was accessed (read(2), pread(2), readv(2), and
617 similar)
618 DN_MODIFY
619 A file was modified (write(2), pwrite(2), writev(2),
620 truncate(2), ftruncate(2), and similar).
621 DN_CREATE
622 A file was created (open(2), creat(2), mknod(2),
623 mkdir(2), link(2), symlink(2), rename(2) into this direc‐
624 tory).
625 DN_DELETE
626 A file was unlinked (unlink(2), rename(2) to another di‐
627 rectory, rmdir(2)).
628 DN_RENAME
629 A file was renamed within this directory (rename(2)).
630 DN_ATTRIB
631 The attributes of a file were changed (chown(2),
632 chmod(2), utime(2), utimensat(2), and similar).
633 (In order to obtain these definitions, the _GNU_SOURCE feature
635 test macro must be defined before including any header files.)
636 Directory notifications are normally "one-shot", and the appli‐
638 cation must reregister to receive further notifications. Alter‐
639 natively, if DN_MULTISHOT is included in arg, then notification
640 will remain in effect until explicitly removed.
641 A series of F_NOTIFY requests is cumulative, with the events in
643 arg being added to the set already monitored. To disable noti‐
644 fication of all events, make an F_NOTIFY call specifying arg as
645 0.
646 Notification occurs via delivery of a signal. The default sig‐
648 nal is SIGIO, but this can be changed using the F_SETSIG command
649 to fcntl(). (Note that SIGIO is one of the nonqueuing standard
650 signals; switching to the use of a real-time signal means that
651 multiple notifications can be queued to the process.) In the
652 latter case, the signal handler receives a siginfo_t structure
653 as its second argument (if the handler was established using
654 SA_SIGINFO) and the si_fd field of this structure contains the
655 file descriptor which generated the notification (useful when
656 establishing notification on multiple directories).
657 Especially when using DN_MULTISHOT, a real time signal should be
659 used for notification, so that multiple notifications can be
660 queued.
661 NOTE: New applications should use the inotify interface (avail‐
663 able since Linux 2.6.13), which provides a much superior inter‐
664 face for obtaining notifications of filesystem events. See ino‐
665 tify(7).
666 Changing the capacity of a pipe
668 F_SETPIPE_SZ (int; since Linux 2.6.35)
669 Change the capacity of the pipe referred to by fd to be at least
670 arg bytes. An unprivileged process can adjust the pipe capacity
671 to any value between the system page size and the limit defined
672 in /proc/sys/fs/pipe-max-size (see proc(5)). Attempts to set
673 the pipe capacity below the page size are silently rounded up to
674 the page size. Attempts by an unprivileged process to set the
675 pipe capacity above the limit in /proc/sys/fs/pipe-max-size
676 yield the error EPERM; a privileged process (CAP_SYS_RESOURCE)
677 can override the limit.
678 When allocating the buffer for the pipe, the kernel may use a
680 capacity larger than arg, if that is convenient for the imple‐
681 mentation. (In the current implementation, the allocation is
682 the next higher power-of-two page-size multiple of the requested
683 size.) The actual capacity (in bytes) that is set is returned
684 as the function result.
685 Attempting to set the pipe capacity smaller than the amount of
687 buffer space currently used to store data produces the error
688 EBUSY.
689 Note that because of the way the pages of the pipe buffer are
691 employed when data is written to the pipe, the number of bytes
692 that can be written may be less than the nominal size, depending
693 on the size of the writes.
694 F_GETPIPE_SZ (void; since Linux 2.6.35)
696 Return (as the function result) the capacity of the pipe re‐
697 ferred to by fd.
698 File Sealing
700 File seals limit the set of allowed operations on a given file. For
701 each seal that is set on a file, a specific set of operations will fail
702 with EPERM on this file from now on. The file is said to be sealed.
703 The default set of seals depends on the type of the underlying file and
704 filesystem. For an overview of file sealing, a discussion of its pur‐
705 pose, and some code examples, see memfd_create(2).
706 Currently, file seals can be applied only to a file descriptor returned
708 by memfd_create(2) (if the MFD_ALLOW_SEALING was employed). On other
709 filesystems, all fcntl() operations that operate on seals will return
710 EINVAL.
711 Seals are a property of an inode. Thus, all open file descriptors re‐
713 ferring to the same inode share the same set of seals. Furthermore,
714 seals can never be removed, only added.
715 F_ADD_SEALS (int; since Linux 3.17)
717 Add the seals given in the bit-mask argument arg to the set of
718 seals of the inode referred to by the file descriptor fd. Seals
719 cannot be removed again. Once this call succeeds, the seals are
720 enforced by the kernel immediately. If the current set of seals
721 includes F_SEAL_SEAL (see below), then this call will be re‐
722 jected with EPERM. Adding a seal that is already set is a no-
723 op, in case F_SEAL_SEAL is not set already. In order to place a
724 seal, the file descriptor fd must be writable.
725 F_GET_SEALS (void; since Linux 3.17)
727 Return (as the function result) the current set of seals of the
728 inode referred to by fd. If no seals are set, 0 is returned.
729 If the file does not support sealing, -1 is returned and errno
730 is set to EINVAL.
731 The following seals are available:
733 F_SEAL_SEAL
735 If this seal is set, any further call to fcntl() with
736 F_ADD_SEALS fails with the error EPERM. Therefore, this seal
737 prevents any modifications to the set of seals itself. If the
738 initial set of seals of a file includes F_SEAL_SEAL, then this
739 effectively causes the set of seals to be constant and locked.
740 F_SEAL_SHRINK
742 If this seal is set, the file in question cannot be reduced in
743 size. This affects open(2) with the O_TRUNC flag as well as
744 truncate(2) and ftruncate(2). Those calls fail with EPERM if
745 you try to shrink the file in question. Increasing the file
746 size is still possible.
747 F_SEAL_GROW
749 If this seal is set, the size of the file in question cannot be
750 increased. This affects write(2) beyond the end of the file,
751 truncate(2), ftruncate(2), and fallocate(2). These calls fail
752 with EPERM if you use them to increase the file size. If you
753 keep the size or shrink it, those calls still work as expected.
754 F_SEAL_WRITE
756 If this seal is set, you cannot modify the contents of the file.
757 Note that shrinking or growing the size of the file is still
758 possible and allowed. Thus, this seal is normally used in com‐
759 bination with one of the other seals. This seal affects
760 write(2) and fallocate(2) (only in combination with the FAL‐
761 LOC_FL_PUNCH_HOLE flag). Those calls fail with EPERM if this
762 seal is set. Furthermore, trying to create new shared, writable
763 memory-mappings via mmap(2) will also fail with EPERM.
764 Using the F_ADD_SEALS operation to set the F_SEAL_WRITE seal
766 fails with EBUSY if any writable, shared mapping exists. Such
767 mappings must be unmapped before you can add this seal. Fur‐
768 thermore, if there are any asynchronous I/O operations (io_sub‐
769 mit(2)) pending on the file, all outstanding writes will be dis‐
770 carded.
771 F_SEAL_FUTURE_WRITE (since Linux 5.1)
773 The effect of this seal is similar to F_SEAL_WRITE, but the con‐
774 tents of the file can still be modified via shared writable map‐
775 pings that were created prior to the seal being set. Any at‐
776 tempt to create a new writable mapping on the file via mmap(2)
777 will fail with EPERM. Likewise, an attempt to write to the file
778 via write(2) will fail with EPERM.
779 Using this seal, one process can create a memory buffer that it
781 can continue to modify while sharing that buffer on a "read-
782 only" basis with other processes.
783 File read/write hints
785 Write lifetime hints can be used to inform the kernel about the rela‐
786 tive expected lifetime of writes on a given inode or via a particular
787 open file description. (See open(2) for an explanation of open file
788 descriptions.) In this context, the term "write lifetime" means the
789 expected time the data will live on media, before being overwritten or
790 erased.
791 An application may use the different hint values specified below to
793 separate writes into different write classes, so that multiple users or
794 applications running on a single storage back-end can aggregate their
795 I/O patterns in a consistent manner. However, there are no functional
796 semantics implied by these flags, and different I/O classes can use the
797 write lifetime hints in arbitrary ways, so long as the hints are used
798 consistently.
799 The following operations can be applied to the file descriptor, fd:
801 F_GET_RW_HINT (uint64_t *; since Linux 4.13)
803 Returns the value of the read/write hint associated with the un‐
804 derlying inode referred to by fd.
805 F_SET_RW_HINT (uint64_t *; since Linux 4.13)
807 Sets the read/write hint value associated with the underlying
808 inode referred to by fd. This hint persists until either it is
809 explicitly modified or the underlying filesystem is unmounted.
810 F_GET_FILE_RW_HINT (uint64_t *; since Linux 4.13)
812 Returns the value of the read/write hint associated with the
813 open file description referred to by fd.
814 F_SET_FILE_RW_HINT (uint64_t *; since Linux 4.13)
816 Sets the read/write hint value associated with the open file de‐
817 scription referred to by fd.
818 If an open file description has not been assigned a read/write hint,
820 then it shall use the value assigned to the inode, if any.
821 The following read/write hints are valid since Linux 4.13:
823 RWH_WRITE_LIFE_NOT_SET
825 No specific hint has been set. This is the default value.
826 RWH_WRITE_LIFE_NONE
828 No specific write lifetime is associated with this file or in‐
829 ode.
830 RWH_WRITE_LIFE_SHORT
832 Data written to this inode or via this open file description is
833 expected to have a short lifetime.
834 RWH_WRITE_LIFE_MEDIUM
836 Data written to this inode or via this open file description is
837 expected to have a lifetime longer than data written with
838 RWH_WRITE_LIFE_SHORT.
839 RWH_WRITE_LIFE_LONG
841 Data written to this inode or via this open file description is
842 expected to have a lifetime longer than data written with
843 RWH_WRITE_LIFE_MEDIUM.
844 RWH_WRITE_LIFE_EXTREME
846 Data written to this inode or via this open file description is
847 expected to have a lifetime longer than data written with
848 RWH_WRITE_LIFE_LONG.
849 All the write-specific hints are relative to each other, and no indi‐
851 vidual absolute meaning should be attributed to them.
852
854 For a successful call, the return value depends on the operation:
855 F_DUPFD
857 The new file descriptor.
858 F_GETFD
860 Value of file descriptor flags.
861 F_GETFL
863 Value of file status flags.
864 F_GETLEASE
866 Type of lease held on file descriptor.
867 F_GETOWN
869 Value of file descriptor owner.
870 F_GETSIG
872 Value of signal sent when read or write becomes possible, or
873 zero for traditional SIGIO behavior.
874 F_GETPIPE_SZ, F_SETPIPE_SZ
876 The pipe capacity.
877 F_GET_SEALS
879 A bit mask identifying the seals that have been set for the in‐
880 ode referred to by fd.
881 All other commands
883 Zero.
884 On error, -1 is returned, and errno is set to indicate the error.
886
888 EACCES or EAGAIN
889 Operation is prohibited by locks held by other processes.
890 EAGAIN The operation is prohibited because the file has been memory-
892 mapped by another process.
893 EBADF fd is not an open file descriptor
895 EBADF cmd is F_SETLK or F_SETLKW and the file descriptor open mode
897 doesn't match with the type of lock requested.
898 EBUSY cmd is F_SETPIPE_SZ and the new pipe capacity specified in arg
900 is smaller than the amount of buffer space currently used to
901 store data in the pipe.
902 EBUSY cmd is F_ADD_SEALS, arg includes F_SEAL_WRITE, and there exists
904 a writable, shared mapping on the file referred to by fd.
905 EDEADLK
907 It was detected that the specified F_SETLKW command would cause
908 a deadlock.
909 EFAULT lock is outside your accessible address space.
911 EINTR cmd is F_SETLKW or F_OFD_SETLKW and the operation was inter‐
913 rupted by a signal; see signal(7).
914 EINTR cmd is F_GETLK, F_SETLK, F_OFD_GETLK, or F_OFD_SETLK, and the
916 operation was interrupted by a signal before the lock was
917 checked or acquired. Most likely when locking a remote file
918 (e.g., locking over NFS), but can sometimes happen locally.
919 EINVAL The value specified in cmd is not recognized by this kernel.
921 EINVAL cmd is F_ADD_SEALS and arg includes an unrecognized sealing bit.
923 EINVAL cmd is F_ADD_SEALS or F_GET_SEALS and the filesystem containing
925 the inode referred to by fd does not support sealing.
926 EINVAL cmd is F_DUPFD and arg is negative or is greater than the maxi‐
928 mum allowable value (see the discussion of RLIMIT_NOFILE in
929 getrlimit(2)).
930 EINVAL cmd is F_SETSIG and arg is not an allowable signal number.
932 EINVAL cmd is F_OFD_SETLK, F_OFD_SETLKW, or F_OFD_GETLK, and l_pid was
934 not specified as zero.
935 EMFILE cmd is F_DUPFD and the per-process limit on the number of open
937 file descriptors has been reached.
938 ENOLCK Too many segment locks open, lock table is full, or a remote
940 locking protocol failed (e.g., locking over NFS).
941 ENOTDIR
943 F_NOTIFY was specified in cmd, but fd does not refer to a direc‐
944 tory.
945 EPERM cmd is F_SETPIPE_SZ and the soft or hard user pipe limit has
947 been reached; see pipe(7).
948 EPERM Attempted to clear the O_APPEND flag on a file that has the ap‐
950 pend-only attribute set.
951 EPERM cmd was F_ADD_SEALS, but fd was not open for writing or the cur‐
953 rent set of seals on the file already includes F_SEAL_SEAL.
954
956 POSIX.1-2008.
957 F_GETOWN_EX, F_SETOWN_EX, F_SETPIPE_SZ, F_GETPIPE_SZ, F_GETSIG, F_SET‐
959 SIG, F_NOTIFY, F_GETLEASE, and F_SETLEASE are Linux-specific. (Define
960 the _GNU_SOURCE macro to obtain these definitions.)
961 F_OFD_SETLK, F_OFD_SETLKW, and F_OFD_GETLK are Linux-specific (and one
963 must define _GNU_SOURCE to obtain their definitions), but work is being
964 done to have them included in the next version of POSIX.1.
965 F_ADD_SEALS and F_GET_SEALS are Linux-specific.
967
969 SVr4, 4.3BSD, POSIX.1-2001.
970 Only the operations F_DUPFD, F_GETFD, F_SETFD, F_GETFL, F_SETFL,
972 F_GETLK, F_SETLK, and F_SETLKW are specified in POSIX.1-2001.
973 F_GETOWN and F_SETOWN are specified in POSIX.1-2001. (To get their
975 definitions, define either _XOPEN_SOURCE with the value 500 or greater,
976 or _POSIX_C_SOURCE with the value 200809L or greater.)
977 F_DUPFD_CLOEXEC is specified in POSIX.1-2008. (To get this definition,
979 define _POSIX_C_SOURCE with the value 200809L or greater, or
980 _XOPEN_SOURCE with the value 700 or greater.)
981
983 The errors returned by dup2(2) are different from those returned by
984 F_DUPFD.
985 File locking
987 The original Linux fcntl() system call was not designed to handle large
988 file offsets (in the flock structure). Consequently, an fcntl64() sys‐
989 tem call was added in Linux 2.4. The newer system call employs a dif‐
990 ferent structure for file locking, flock64, and corresponding commands,
991 F_GETLK64, F_SETLK64, and F_SETLKW64. However, these details can be
992 ignored by applications using glibc, whose fcntl() wrapper function
993 transparently employs the more recent system call where it is avail‐
994 able.
995 Record locks
997 Since Linux 2.0, there is no interaction between the types of lock
998 placed by flock(2) and fcntl().
999 Several systems have more fields in struct flock such as, for example,
1001 l_sysid (to identify the machine where the lock is held). Clearly,
1002 l_pid alone is not going to be very useful if the process holding the
1003 lock may live on a different machine; on Linux, while present on some
1004 architectures (such as MIPS32), this field is not used.
1005 The original Linux fcntl() system call was not designed to handle large
1007 file offsets (in the flock structure). Consequently, an fcntl64() sys‐
1008 tem call was added in Linux 2.4. The newer system call employs a dif‐
1009 ferent structure for file locking, flock64, and corresponding commands,
1010 F_GETLK64, F_SETLK64, and F_SETLKW64. However, these details can be
1011 ignored by applications using glibc, whose fcntl() wrapper function
1012 transparently employs the more recent system call where it is avail‐
1013 able.
1014 Record locking and NFS
1016 Before Linux 3.12, if an NFSv4 client loses contact with the server for
1017 a period of time (defined as more than 90 seconds with no communica‐
1018 tion), it might lose and regain a lock without ever being aware of the
1019 fact. (The period of time after which contact is assumed lost is known
1020 as the NFSv4 leasetime. On a Linux NFS server, this can be determined
1021 by looking at /proc/fs/nfsd/nfsv4leasetime, which expresses the period
1022 in seconds. The default value for this file is 90.) This scenario po‐
1023 tentially risks data corruption, since another process might acquire a
1024 lock in the intervening period and perform file I/O.
1025 Since Linux 3.12, if an NFSv4 client loses contact with the server, any
1027 I/O to the file by a process which "thinks" it holds a lock will fail
1028 until that process closes and reopens the file. A kernel parameter,
1029 nfs.recover_lost_locks, can be set to 1 to obtain the pre-3.12 behav‐
1030 ior, whereby the client will attempt to recover lost locks when contact
1031 is reestablished with the server. Because of the attendant risk of
1032 data corruption, this parameter defaults to 0 (disabled).
1033
1035 F_SETFL
1036 It is not possible to use F_SETFL to change the state of the O_DSYNC
1037 and O_SYNC flags. Attempts to change the state of these flags are
1038 silently ignored.
1039 F_GETOWN
1041 A limitation of the Linux system call conventions on some architectures
1042 (notably i386) means that if a (negative) process group ID to be re‐
1043 turned by F_GETOWN falls in the range -1 to -4095, then the return
1044 value is wrongly interpreted by glibc as an error in the system call;
1045 that is, the return value of fcntl() will be -1, and errno will contain
1046 the (positive) process group ID. The Linux-specific F_GETOWN_EX opera‐
1047 tion avoids this problem. Since glibc 2.11, glibc makes the kernel
1048 F_GETOWN problem invisible by implementing F_GETOWN using F_GETOWN_EX.
1049 F_SETOWN
1051 In Linux 2.4 and earlier, there is bug that can occur when an unprivi‐
1052 leged process uses F_SETOWN to specify the owner of a socket file de‐
1053 scriptor as a process (group) other than the caller. In this case, fc‐
1054 ntl() can return -1 with errno set to EPERM, even when the owner
1055 process (group) is one that the caller has permission to send signals
1056 to. Despite this error return, the file descriptor owner is set, and
1057 signals will be sent to the owner.
1058 Deadlock detection
1060 The deadlock-detection algorithm employed by the kernel when dealing
1061 with F_SETLKW requests can yield both false negatives (failures to de‐
1062 tect deadlocks, leaving a set of deadlocked processes blocked indefi‐
1063 nitely) and false positives (EDEADLK errors when there is no deadlock).
1064 For example, the kernel limits the lock depth of its dependency search
1065 to 10 steps, meaning that circular deadlock chains that exceed that
1066 size will not be detected. In addition, the kernel may falsely indi‐
1067 cate a deadlock when two or more processes created using the clone(2)
1068 CLONE_FILES flag place locks that appear (to the kernel) to conflict.
1069 Mandatory locking
1071 The Linux implementation of mandatory locking is subject to race condi‐
1072 tions which render it unreliable: a write(2) call that overlaps with a
1073 lock may modify data after the mandatory lock is acquired; a read(2)
1074 call that overlaps with a lock may detect changes to data that were
1075 made only after a write lock was acquired. Similar races exist between
1076 mandatory locks and mmap(2). It is therefore inadvisable to rely on
1077 mandatory locking.
1078
1080 dup2(2), flock(2), open(2), socket(2), lockf(3), capabilities(7), fea‐
1081 ture_test_macros(7), lslocks(8)
1082 locks.txt, mandatory-locking.txt, and dnotify.txt in the Linux kernel
1084 source directory Documentation/filesystems/ (on older kernels, these
1085 files are directly under the Documentation/ directory, and manda‐
1086 tory-locking.txt is called mandatory.txt)
1087Linux man-pages 6.04 2023-03-30 fcntl(2)