1FCNTL(2) Linux Programmer's Manual FCNTL(2)
2
6 fcntl - manipulate file descriptor
7
9 #include <fcntl.h>
10 int fcntl(int fd, int cmd, ... /* arg */ );
12
14 fcntl() performs one of the operations described below on the open file
15 descriptor fd. The operation is determined by cmd.
16 fcntl() can take an optional third argument. Whether or not this argu‐
18 ment is required is determined by cmd. The required argument type is
19 indicated in parentheses after each cmd name (in most cases, the re‐
20 quired type is int, and we identify the argument using the name arg),
21 or void is specified if the argument is not required.
22 Certain of the operations below are supported only since a particular
24 Linux kernel version. The preferred method of checking whether the
25 host kernel supports a particular operation is to invoke fcntl() with
26 the desired cmd value and then test whether the call failed with EIN‐
27 VAL, indicating that the kernel does not recognize this value.
28 Duplicating a file descriptor
30 F_DUPFD (int)
31 Duplicate the file descriptor fd using the lowest-numbered
32 available file descriptor greater than or equal to arg. This is
33 different from dup2(2), which uses exactly the file descriptor
34 specified.
35 On success, the new file descriptor is returned.
37 See dup(2) for further details.
39 F_DUPFD_CLOEXEC (int; since Linux 2.6.24)
41 As for F_DUPFD, but additionally set the close-on-exec flag for
42 the duplicate file descriptor. Specifying this flag permits a
43 program to avoid an additional fcntl() F_SETFD operation to set
44 the FD_CLOEXEC flag. For an explanation of why this flag is
45 useful, see the description of O_CLOEXEC in open(2).
46 File descriptor flags
48 The following commands manipulate the flags associated with a file de‐
49 scriptor. Currently, only one such flag is defined: FD_CLOEXEC, the
50 close-on-exec flag. If the FD_CLOEXEC bit is set, the file descriptor
51 will automatically be closed during a successful execve(2). (If the
52 execve(2) fails, the file descriptor is left open.) If the FD_CLOEXEC
53 bit is not set, the file descriptor will remain open across an ex‐
54 ecve(2).
55 F_GETFD (void)
57 Return (as the function result) the file descriptor flags; arg
58 is ignored.
59 F_SETFD (int)
61 Set the file descriptor flags to the value specified by arg.
62 In multithreaded programs, using fcntl() F_SETFD to set the close-on-
64 exec flag at the same time as another thread performs a fork(2) plus
65 execve(2) is vulnerable to a race condition that may unintentionally
66 leak the file descriptor to the program executed in the child process.
67 See the discussion of the O_CLOEXEC flag in open(2) for details and a
68 remedy to the problem.
69 File status flags
71 Each open file description has certain associated status flags, ini‐
72 tialized by open(2) and possibly modified by fcntl(). Duplicated file
73 descriptors (made with dup(2), fcntl(F_DUPFD), fork(2), etc.) refer to
74 the same open file description, and thus share the same file status
75 flags.
76 The file status flags and their semantics are described in open(2).
78 F_GETFL (void)
80 Return (as the function result) the file access mode and the
81 file status flags; arg is ignored.
82 F_SETFL (int)
84 Set the file status flags to the value specified by arg. File
85 access mode (O_RDONLY, O_WRONLY, O_RDWR) and file creation flags
86 (i.e., O_CREAT, O_EXCL, O_NOCTTY, O_TRUNC) in arg are ignored.
87 On Linux, this command can change only the O_APPEND, O_ASYNC,
88 O_DIRECT, O_NOATIME, and O_NONBLOCK flags. It is not possible
89 to change the O_DSYNC and O_SYNC flags; see BUGS, below.
90 Advisory record locking
92 Linux implements traditional ("process-associated") UNIX record locks,
93 as standardized by POSIX. For a Linux-specific alternative with better
94 semantics, see the discussion of open file description locks below.
95 F_SETLK, F_SETLKW, and F_GETLK are used to acquire, release, and test
97 for the existence of record locks (also known as byte-range, file-seg‐
98 ment, or file-region locks). The third argument, lock, is a pointer to
99 a structure that has at least the following fields (in unspecified or‐
100 der).
101 struct flock {
103 ...
104 short l_type; /* Type of lock: F_RDLCK,
105 F_WRLCK, F_UNLCK */
106 short l_whence; /* How to interpret l_start:
107 SEEK_SET, SEEK_CUR, SEEK_END */
108 off_t l_start; /* Starting offset for lock */
109 off_t l_len; /* Number of bytes to lock */
110 pid_t l_pid; /* PID of process blocking our lock
111 (set by F_GETLK and F_OFD_GETLK) */
112 ...
113 };
114 The l_whence, l_start, and l_len fields of this structure specify the
116 range of bytes we wish to lock. Bytes past the end of the file may be
117 locked, but not bytes before the start of the file.
118 l_start is the starting offset for the lock, and is interpreted rela‐
120 tive to either: the start of the file (if l_whence is SEEK_SET); the
121 current file offset (if l_whence is SEEK_CUR); or the end of the file
122 (if l_whence is SEEK_END). In the final two cases, l_start can be a
123 negative number provided the offset does not lie before the start of
124 the file.
125 l_len specifies the number of bytes to be locked. If l_len is posi‐
127 tive, then the range to be locked covers bytes l_start up to and in‐
128 cluding l_start+l_len-1. Specifying 0 for l_len has the special mean‐
129 ing: lock all bytes starting at the location specified by l_whence and
130 l_start through to the end of file, no matter how large the file grows.
131 POSIX.1-2001 allows (but does not require) an implementation to support
133 a negative l_len value; if l_len is negative, the interval described by
134 lock covers bytes l_start+l_len up to and including l_start-1. This is
135 supported by Linux since kernel versions 2.4.21 and 2.5.49.
136 The l_type field can be used to place a read (F_RDLCK) or a write
138 (F_WRLCK) lock on a file. Any number of processes may hold a read lock
139 (shared lock) on a file region, but only one process may hold a write
140 lock (exclusive lock). An exclusive lock excludes all other locks,
141 both shared and exclusive. A single process can hold only one type of
142 lock on a file region; if a new lock is applied to an already-locked
143 region, then the existing lock is converted to the new lock type.
144 (Such conversions may involve splitting, shrinking, or coalescing with
145 an existing lock if the byte range specified by the new lock does not
146 precisely coincide with the range of the existing lock.)
147 F_SETLK (struct flock *)
149 Acquire a lock (when l_type is F_RDLCK or F_WRLCK) or release a
150 lock (when l_type is F_UNLCK) on the bytes specified by the
151 l_whence, l_start, and l_len fields of lock. If a conflicting
152 lock is held by another process, this call returns -1 and sets
153 errno to EACCES or EAGAIN. (The error returned in this case
154 differs across implementations, so POSIX requires a portable ap‐
155 plication to check for both errors.)
156 F_SETLKW (struct flock *)
158 As for F_SETLK, but if a conflicting lock is held on the file,
159 then wait for that lock to be released. If a signal is caught
160 while waiting, then the call is interrupted and (after the sig‐
161 nal handler has returned) returns immediately (with return value
162 -1 and errno set to EINTR; see signal(7)).
163 F_GETLK (struct flock *)
165 On input to this call, lock describes a lock we would like to
166 place on the file. If the lock could be placed, fcntl() does
167 not actually place it, but returns F_UNLCK in the l_type field
168 of lock and leaves the other fields of the structure unchanged.
169 If one or more incompatible locks would prevent this lock being
171 placed, then fcntl() returns details about one of those locks in
172 the l_type, l_whence, l_start, and l_len fields of lock. If the
173 conflicting lock is a traditional (process-associated) record
174 lock, then the l_pid field is set to the PID of the process
175 holding that lock. If the conflicting lock is an open file de‐
176 scription lock, then l_pid is set to -1. Note that the returned
177 information may already be out of date by the time the caller
178 inspects it.
179 In order to place a read lock, fd must be open for reading. In order
181 to place a write lock, fd must be open for writing. To place both
182 types of lock, open a file read-write.
183 When placing locks with F_SETLKW, the kernel detects deadlocks, whereby
185 two or more processes have their lock requests mutually blocked by
186 locks held by the other processes. For example, suppose process A
187 holds a write lock on byte 100 of a file, and process B holds a write
188 lock on byte 200. If each process then attempts to lock the byte al‐
189 ready locked by the other process using F_SETLKW, then, without dead‐
190 lock detection, both processes would remain blocked indefinitely. When
191 the kernel detects such deadlocks, it causes one of the blocking lock
192 requests to immediately fail with the error EDEADLK; an application
193 that encounters such an error should release some of its locks to allow
194 other applications to proceed before attempting regain the locks that
195 it requires. Circular deadlocks involving more than two processes are
196 also detected. Note, however, that there are limitations to the ker‐
197 nel's deadlock-detection algorithm; see BUGS.
198 As well as being removed by an explicit F_UNLCK, record locks are auto‐
200 matically released when the process terminates.
201 Record locks are not inherited by a child created via fork(2), but are
203 preserved across an execve(2).
204 Because of the buffering performed by the stdio(3) library, the use of
206 record locking with routines in that package should be avoided; use
207 read(2) and write(2) instead.
208 The record locks described above are associated with the process (un‐
210 like the open file description locks described below). This has some
211 unfortunate consequences:
212 * If a process closes any file descriptor referring to a file, then
214 all of the process's locks on that file are released, regardless of
215 the file descriptor(s) on which the locks were obtained. This is
216 bad: it means that a process can lose its locks on a file such as
217 /etc/passwd or /etc/mtab when for some reason a library function de‐
218 cides to open, read, and close the same file.
219 * The threads in a process share locks. In other words, a multi‐
221 threaded program can't use record locking to ensure that threads
222 don't simultaneously access the same region of a file.
223 Open file description locks solve both of these problems.
225 Open file description locks (non-POSIX)
227 Open file description locks are advisory byte-range locks whose opera‐
228 tion is in most respects identical to the traditional record locks de‐
229 scribed above. This lock type is Linux-specific, and available since
230 Linux 3.15. (There is a proposal with the Austin Group to include this
231 lock type in the next revision of POSIX.1.) For an explanation of open
232 file descriptions, see open(2).
233 The principal difference between the two lock types is that whereas
235 traditional record locks are associated with a process, open file de‐
236 scription locks are associated with the open file description on which
237 they are acquired, much like locks acquired with flock(2). Conse‐
238 quently (and unlike traditional advisory record locks), open file de‐
239 scription locks are inherited across fork(2) (and clone(2) with
240 CLONE_FILES), and are only automatically released on the last close of
241 the open file description, instead of being released on any close of
242 the file.
243 Conflicting lock combinations (i.e., a read lock and a write lock or
245 two write locks) where one lock is an open file description lock and
246 the other is a traditional record lock conflict even when they are ac‐
247 quired by the same process on the same file descriptor.
248 Open file description locks placed via the same open file description
250 (i.e., via the same file descriptor, or via a duplicate of the file de‐
251 scriptor created by fork(2), dup(2), fcntl() F_DUPFD, and so on) are
252 always compatible: if a new lock is placed on an already locked region,
253 then the existing lock is converted to the new lock type. (Such con‐
254 versions may result in splitting, shrinking, or coalescing with an ex‐
255 isting lock as discussed above.)
256 On the other hand, open file description locks may conflict with each
258 other when they are acquired via different open file descriptions.
259 Thus, the threads in a multithreaded program can use open file descrip‐
260 tion locks to synchronize access to a file region by having each thread
261 perform its own open(2) on the file and applying locks via the result‐
262 ing file descriptor.
263 As with traditional advisory locks, the third argument to fcntl(),
265 lock, is a pointer to an flock structure. By contrast with traditional
266 record locks, the l_pid field of that structure must be set to zero
267 when using the commands described below.
268 The commands for working with open file description locks are analogous
270 to those used with traditional locks:
271 F_OFD_SETLK (struct flock *)
273 Acquire an open file description lock (when l_type is F_RDLCK or
274 F_WRLCK) or release an open file description lock (when l_type
275 is F_UNLCK) on the bytes specified by the l_whence, l_start, and
276 l_len fields of lock. If a conflicting lock is held by another
277 process, this call returns -1 and sets errno to EAGAIN.
278 F_OFD_SETLKW (struct flock *)
280 As for F_OFD_SETLK, but if a conflicting lock is held on the
281 file, then wait for that lock to be released. If a signal is
282 caught while waiting, then the call is interrupted and (after
283 the signal handler has returned) returns immediately (with re‐
284 turn value -1 and errno set to EINTR; see signal(7)).
285 F_OFD_GETLK (struct flock *)
287 On input to this call, lock describes an open file description
288 lock we would like to place on the file. If the lock could be
289 placed, fcntl() does not actually place it, but returns F_UNLCK
290 in the l_type field of lock and leaves the other fields of the
291 structure unchanged. If one or more incompatible locks would
292 prevent this lock being placed, then details about one of these
293 locks are returned via lock, as described above for F_GETLK.
294 In the current implementation, no deadlock detection is performed for
296 open file description locks. (This contrasts with process-associated
297 record locks, for which the kernel does perform deadlock detection.)
298 Mandatory locking
300 Warning: the Linux implementation of mandatory locking is unreliable.
301 See BUGS below. Because of these bugs, and the fact that the feature
302 is believed to be little used, since Linux 4.5, mandatory locking has
303 been made an optional feature, governed by a configuration option (CON‐
304 FIG_MANDATORY_FILE_LOCKING). This is an initial step toward removing
305 this feature completely.
306 By default, both traditional (process-associated) and open file de‐
308 scription record locks are advisory. Advisory locks are not enforced
309 and are useful only between cooperating processes.
310 Both lock types can also be mandatory. Mandatory locks are enforced
312 for all processes. If a process tries to perform an incompatible ac‐
313 cess (e.g., read(2) or write(2)) on a file region that has an incompat‐
314 ible mandatory lock, then the result depends upon whether the O_NON‐
315 BLOCK flag is enabled for its open file description. If the O_NONBLOCK
316 flag is not enabled, then the system call is blocked until the lock is
317 removed or converted to a mode that is compatible with the access. If
318 the O_NONBLOCK flag is enabled, then the system call fails with the er‐
319 ror EAGAIN.
320 To make use of mandatory locks, mandatory locking must be enabled both
322 on the filesystem that contains the file to be locked, and on the file
323 itself. Mandatory locking is enabled on a filesystem using the "-o
324 mand" option to mount(8), or the MS_MANDLOCK flag for mount(2). Manda‐
325 tory locking is enabled on a file by disabling group execute permission
326 on the file and enabling the set-group-ID permission bit (see chmod(1)
327 and chmod(2)).
328 Mandatory locking is not specified by POSIX. Some other systems also
330 support mandatory locking, although the details of how to enable it
331 vary across systems.
332 Lost locks
334 When an advisory lock is obtained on a networked filesystem such as NFS
335 it is possible that the lock might get lost. This may happen due to
336 administrative action on the server, or due to a network partition
337 (i.e., loss of network connectivity with the server) which lasts long
338 enough for the server to assume that the client is no longer function‐
339 ing.
340 When the filesystem determines that a lock has been lost, future
342 read(2) or write(2) requests may fail with the error EIO. This error
343 will persist until the lock is removed or the file descriptor is
344 closed. Since Linux 3.12, this happens at least for NFSv4 (including
345 all minor versions).
346 Some versions of UNIX send a signal (SIGLOST) in this circumstance.
348 Linux does not define this signal, and does not provide any asynchro‐
349 nous notification of lost locks.
350 Managing signals
352 F_GETOWN, F_SETOWN, F_GETOWN_EX, F_SETOWN_EX, F_GETSIG, and F_SETSIG
353 are used to manage I/O availability signals:
354 F_GETOWN (void)
356 Return (as the function result) the process ID or process group
357 ID currently receiving SIGIO and SIGURG signals for events on
358 file descriptor fd. Process IDs are returned as positive val‐
359 ues; process group IDs are returned as negative values (but see
360 BUGS below). arg is ignored.
361 F_SETOWN (int)
363 Set the process ID or process group ID that will receive SIGIO
364 and SIGURG signals for events on the file descriptor fd. The
365 target process or process group ID is specified in arg. A
366 process ID is specified as a positive value; a process group ID
367 is specified as a negative value. Most commonly, the calling
368 process specifies itself as the owner (that is, arg is specified
369 as getpid(2)).
370 As well as setting the file descriptor owner, one must also en‐
372 able generation of signals on the file descriptor. This is done
373 by using the fcntl() F_SETFL command to set the O_ASYNC file
374 status flag on the file descriptor. Subsequently, a SIGIO sig‐
375 nal is sent whenever input or output becomes possible on the
376 file descriptor. The fcntl() F_SETSIG command can be used to
377 obtain delivery of a signal other than SIGIO.
378 Sending a signal to the owner process (group) specified by F_SE‐
380 TOWN is subject to the same permissions checks as are described
381 for kill(2), where the sending process is the one that employs
382 F_SETOWN (but see BUGS below). If this permission check fails,
383 then the signal is silently discarded. Note: The F_SETOWN oper‐
384 ation records the caller's credentials at the time of the fc‐
385 ntl() call, and it is these saved credentials that are used for
386 the permission checks.
387 If the file descriptor fd refers to a socket, F_SETOWN also se‐
389 lects the recipient of SIGURG signals that are delivered when
390 out-of-band data arrives on that socket. (SIGURG is sent in any
391 situation where select(2) would report the socket as having an
392 "exceptional condition".)
393 The following was true in 2.6.x kernels up to and including ker‐
395 nel 2.6.11:
396 If a nonzero value is given to F_SETSIG in a multi‐
398 threaded process running with a threading library that
399 supports thread groups (e.g., NPTL), then a positive
400 value given to F_SETOWN has a different meaning: instead
401 of being a process ID identifying a whole process, it is
402 a thread ID identifying a specific thread within a
403 process. Consequently, it may be necessary to pass F_SE‐
404 TOWN the result of gettid(2) instead of getpid(2) to get
405 sensible results when F_SETSIG is used. (In current
406 Linux threading implementations, a main thread's thread
407 ID is the same as its process ID. This means that a sin‐
408 gle-threaded program can equally use gettid(2) or get‐
409 pid(2) in this scenario.) Note, however, that the state‐
410 ments in this paragraph do not apply to the SIGURG signal
411 generated for out-of-band data on a socket: this signal
412 is always sent to either a process or a process group,
413 depending on the value given to F_SETOWN.
414 The above behavior was accidentally dropped in Linux 2.6.12, and
416 won't be restored. From Linux 2.6.32 onward, use F_SETOWN_EX to
417 target SIGIO and SIGURG signals at a particular thread.
418 F_GETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
420 Return the current file descriptor owner settings as defined by
421 a previous F_SETOWN_EX operation. The information is returned
422 in the structure pointed to by arg, which has the following
423 form:
424 struct f_owner_ex {
426 int type;
427 pid_t pid;
428 };
429 The type field will have one of the values F_OWNER_TID,
431 F_OWNER_PID, or F_OWNER_PGRP. The pid field is a positive inte‐
432 ger representing a thread ID, process ID, or process group ID.
433 See F_SETOWN_EX for more details.
434 F_SETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
436 This operation performs a similar task to F_SETOWN. It allows
437 the caller to direct I/O availability signals to a specific
438 thread, process, or process group. The caller specifies the
439 target of signals via arg, which is a pointer to a f_owner_ex
440 structure. The type field has one of the following values,
441 which define how pid is interpreted:
442 F_OWNER_TID
444 Send the signal to the thread whose thread ID (the value
445 returned by a call to clone(2) or gettid(2)) is specified
446 in pid.
447 F_OWNER_PID
449 Send the signal to the process whose ID is specified in
450 pid.
451 F_OWNER_PGRP
453 Send the signal to the process group whose ID is speci‐
454 fied in pid. (Note that, unlike with F_SETOWN, a process
455 group ID is specified as a positive value here.)
456 F_GETSIG (void)
458 Return (as the function result) the signal sent when input or
459 output becomes possible. A value of zero means SIGIO is sent.
460 Any other value (including SIGIO) is the signal sent instead,
461 and in this case additional info is available to the signal han‐
462 dler if installed with SA_SIGINFO. arg is ignored.
463 F_SETSIG (int)
465 Set the signal sent when input or output becomes possible to the
466 value given in arg. A value of zero means to send the default
467 SIGIO signal. Any other value (including SIGIO) is the signal
468 to send instead, and in this case additional info is available
469 to the signal handler if installed with SA_SIGINFO.
470 By using F_SETSIG with a nonzero value, and setting SA_SIGINFO
472 for the signal handler (see sigaction(2)), extra information
473 about I/O events is passed to the handler in a siginfo_t struc‐
474 ture. If the si_code field indicates the source is SI_SIGIO,
475 the si_fd field gives the file descriptor associated with the
476 event. Otherwise, there is no indication which file descriptors
477 are pending, and you should use the usual mechanisms (select(2),
478 poll(2), read(2) with O_NONBLOCK set etc.) to determine which
479 file descriptors are available for I/O.
480 Note that the file descriptor provided in si_fd is the one that
482 was specified during the F_SETSIG operation. This can lead to
483 an unusual corner case. If the file descriptor is duplicated
484 (dup(2) or similar), and the original file descriptor is closed,
485 then I/O events will continue to be generated, but the si_fd
486 field will contain the number of the now closed file descriptor.
487 By selecting a real time signal (value >= SIGRTMIN), multiple
489 I/O events may be queued using the same signal numbers. (Queu‐
490 ing is dependent on available memory.) Extra information is
491 available if SA_SIGINFO is set for the signal handler, as above.
492 Note that Linux imposes a limit on the number of real-time sig‐
494 nals that may be queued to a process (see getrlimit(2) and sig‐
495 nal(7)) and if this limit is reached, then the kernel reverts to
496 delivering SIGIO, and this signal is delivered to the entire
497 process rather than to a specific thread.
498 Using these mechanisms, a program can implement fully asynchronous I/O
500 without using select(2) or poll(2) most of the time.
501 The use of O_ASYNC is specific to BSD and Linux. The only use of
503 F_GETOWN and F_SETOWN specified in POSIX.1 is in conjunction with the
504 use of the SIGURG signal on sockets. (POSIX does not specify the SIGIO
505 signal.) F_GETOWN_EX, F_SETOWN_EX, F_GETSIG, and F_SETSIG are Linux-
506 specific. POSIX has asynchronous I/O and the aio_sigevent structure to
507 achieve similar things; these are also available in Linux as part of
508 the GNU C Library (Glibc).
509 Leases
511 F_SETLEASE and F_GETLEASE (Linux 2.4 onward) are used to establish a
512 new lease, and retrieve the current lease, on the open file description
513 referred to by the file descriptor fd. A file lease provides a mecha‐
514 nism whereby the process holding the lease (the "lease holder") is no‐
515 tified (via delivery of a signal) when a process (the "lease breaker")
516 tries to open(2) or truncate(2) the file referred to by that file de‐
517 scriptor.
518 F_SETLEASE (int)
520 Set or remove a file lease according to which of the following
521 values is specified in the integer arg:
522 F_RDLCK
524 Take out a read lease. This will cause the calling
525 process to be notified when the file is opened for writ‐
526 ing or is truncated. A read lease can be placed only on
527 a file descriptor that is opened read-only.
528 F_WRLCK
530 Take out a write lease. This will cause the caller to be
531 notified when the file is opened for reading or writing
532 or is truncated. A write lease may be placed on a file
533 only if there are no other open file descriptors for the
534 file.
535 F_UNLCK
537 Remove our lease from the file.
538 Leases are associated with an open file description (see open(2)).
540 This means that duplicate file descriptors (created by, for example,
541 fork(2) or dup(2)) refer to the same lease, and this lease may be modi‐
542 fied or released using any of these descriptors. Furthermore, the
543 lease is released by either an explicit F_UNLCK operation on any of
544 these duplicate file descriptors, or when all such file descriptors
545 have been closed.
546 Leases may be taken out only on regular files. An unprivileged process
548 may take out a lease only on a file whose UID (owner) matches the
549 filesystem UID of the process. A process with the CAP_LEASE capability
550 may take out leases on arbitrary files.
551 F_GETLEASE (void)
553 Indicates what type of lease is associated with the file de‐
554 scriptor fd by returning either F_RDLCK, F_WRLCK, or F_UNLCK,
555 indicating, respectively, a read lease , a write lease, or no
556 lease. arg is ignored.
557 When a process (the "lease breaker") performs an open(2) or truncate(2)
559 that conflicts with a lease established via F_SETLEASE, the system call
560 is blocked by the kernel and the kernel notifies the lease holder by
561 sending it a signal (SIGIO by default). The lease holder should re‐
562 spond to receipt of this signal by doing whatever cleanup is required
563 in preparation for the file to be accessed by another process (e.g.,
564 flushing cached buffers) and then either remove or downgrade its lease.
565 A lease is removed by performing an F_SETLEASE command specifying arg
566 as F_UNLCK. If the lease holder currently holds a write lease on the
567 file, and the lease breaker is opening the file for reading, then it is
568 sufficient for the lease holder to downgrade the lease to a read lease.
569 This is done by performing an F_SETLEASE command specifying arg as
570 F_RDLCK.
571 If the lease holder fails to downgrade or remove the lease within the
573 number of seconds specified in /proc/sys/fs/lease-break-time, then the
574 kernel forcibly removes or downgrades the lease holder's lease.
575 Once a lease break has been initiated, F_GETLEASE returns the target
577 lease type (either F_RDLCK or F_UNLCK, depending on what would be com‐
578 patible with the lease breaker) until the lease holder voluntarily
579 downgrades or removes the lease or the kernel forcibly does so after
580 the lease break timer expires.
581 Once the lease has been voluntarily or forcibly removed or downgraded,
583 and assuming the lease breaker has not unblocked its system call, the
584 kernel permits the lease breaker's system call to proceed.
585 If the lease breaker's blocked open(2) or truncate(2) is interrupted by
587 a signal handler, then the system call fails with the error EINTR, but
588 the other steps still occur as described above. If the lease breaker
589 is killed by a signal while blocked in open(2) or truncate(2), then the
590 other steps still occur as described above. If the lease breaker spec‐
591 ifies the O_NONBLOCK flag when calling open(2), then the call immedi‐
592 ately fails with the error EWOULDBLOCK, but the other steps still occur
593 as described above.
594 The default signal used to notify the lease holder is SIGIO, but this
596 can be changed using the F_SETSIG command to fcntl(). If a F_SETSIG
597 command is performed (even one specifying SIGIO), and the signal han‐
598 dler is established using SA_SIGINFO, then the handler will receive a
599 siginfo_t structure as its second argument, and the si_fd field of this
600 argument will hold the file descriptor of the leased file that has been
601 accessed by another process. (This is useful if the caller holds
602 leases against multiple files.)
603 File and directory change notification (dnotify)
605 F_NOTIFY (int)
606 (Linux 2.4 onward) Provide notification when the directory re‐
607 ferred to by fd or any of the files that it contains is changed.
608 The events to be notified are specified in arg, which is a bit
609 mask specified by ORing together zero or more of the following
610 bits:
611 DN_ACCESS
613 A file was accessed (read(2), pread(2), readv(2), and
614 similar)
615 DN_MODIFY
616 A file was modified (write(2), pwrite(2), writev(2),
617 truncate(2), ftruncate(2), and similar).
618 DN_CREATE
619 A file was created (open(2), creat(2), mknod(2),
620 mkdir(2), link(2), symlink(2), rename(2) into this direc‐
621 tory).
622 DN_DELETE
623 A file was unlinked (unlink(2), rename(2) to another di‐
624 rectory, rmdir(2)).
625 DN_RENAME
626 A file was renamed within this directory (rename(2)).
627 DN_ATTRIB
628 The attributes of a file were changed (chown(2),
629 chmod(2), utime(2), utimensat(2), and similar).
630 (In order to obtain these definitions, the _GNU_SOURCE feature
632 test macro must be defined before including any header files.)
633 Directory notifications are normally "one-shot", and the appli‐
635 cation must reregister to receive further notifications. Alter‐
636 natively, if DN_MULTISHOT is included in arg, then notification
637 will remain in effect until explicitly removed.
638 A series of F_NOTIFY requests is cumulative, with the events in
640 arg being added to the set already monitored. To disable noti‐
641 fication of all events, make an F_NOTIFY call specifying arg as
642 0.
643 Notification occurs via delivery of a signal. The default sig‐
645 nal is SIGIO, but this can be changed using the F_SETSIG command
646 to fcntl(). (Note that SIGIO is one of the nonqueuing standard
647 signals; switching to the use of a real-time signal means that
648 multiple notifications can be queued to the process.) In the
649 latter case, the signal handler receives a siginfo_t structure
650 as its second argument (if the handler was established using
651 SA_SIGINFO) and the si_fd field of this structure contains the
652 file descriptor which generated the notification (useful when
653 establishing notification on multiple directories).
654 Especially when using DN_MULTISHOT, a real time signal should be
656 used for notification, so that multiple notifications can be
657 queued.
658 NOTE: New applications should use the inotify interface (avail‐
660 able since kernel 2.6.13), which provides a much superior inter‐
661 face for obtaining notifications of filesystem events. See ino‐
662 tify(7).
663 Changing the capacity of a pipe
665 F_SETPIPE_SZ (int; since Linux 2.6.35)
666 Change the capacity of the pipe referred to by fd to be at least
667 arg bytes. An unprivileged process can adjust the pipe capacity
668 to any value between the system page size and the limit defined
669 in /proc/sys/fs/pipe-max-size (see proc(5)). Attempts to set
670 the pipe capacity below the page size are silently rounded up to
671 the page size. Attempts by an unprivileged process to set the
672 pipe capacity above the limit in /proc/sys/fs/pipe-max-size
673 yield the error EPERM; a privileged process (CAP_SYS_RESOURCE)
674 can override the limit.
675 When allocating the buffer for the pipe, the kernel may use a
677 capacity larger than arg, if that is convenient for the imple‐
678 mentation. (In the current implementation, the allocation is
679 the next higher power-of-two page-size multiple of the requested
680 size.) The actual capacity (in bytes) that is set is returned
681 as the function result.
682 Attempting to set the pipe capacity smaller than the amount of
684 buffer space currently used to store data produces the error
685 EBUSY.
686 Note that because of the way the pages of the pipe buffer are
688 employed when data is written to the pipe, the number of bytes
689 that can be written may be less than the nominal size, depending
690 on the size of the writes.
691 F_GETPIPE_SZ (void; since Linux 2.6.35)
693 Return (as the function result) the capacity of the pipe re‐
694 ferred to by fd.
695 File Sealing
697 File seals limit the set of allowed operations on a given file. For
698 each seal that is set on a file, a specific set of operations will fail
699 with EPERM on this file from now on. The file is said to be sealed.
700 The default set of seals depends on the type of the underlying file and
701 filesystem. For an overview of file sealing, a discussion of its pur‐
702 pose, and some code examples, see memfd_create(2).
703 Currently, file seals can be applied only to a file descriptor returned
705 by memfd_create(2) (if the MFD_ALLOW_SEALING was employed). On other
706 filesystems, all fcntl() operations that operate on seals will return
707 EINVAL.
708 Seals are a property of an inode. Thus, all open file descriptors re‐
710 ferring to the same inode share the same set of seals. Furthermore,
711 seals can never be removed, only added.
712 F_ADD_SEALS (int; since Linux 3.17)
714 Add the seals given in the bit-mask argument arg to the set of
715 seals of the inode referred to by the file descriptor fd. Seals
716 cannot be removed again. Once this call succeeds, the seals are
717 enforced by the kernel immediately. If the current set of seals
718 includes F_SEAL_SEAL (see below), then this call will be re‐
719 jected with EPERM. Adding a seal that is already set is a no-
720 op, in case F_SEAL_SEAL is not set already. In order to place a
721 seal, the file descriptor fd must be writable.
722 F_GET_SEALS (void; since Linux 3.17)
724 Return (as the function result) the current set of seals of the
725 inode referred to by fd. If no seals are set, 0 is returned.
726 If the file does not support sealing, -1 is returned and errno
727 is set to EINVAL.
728 The following seals are available:
730 F_SEAL_SEAL
732 If this seal is set, any further call to fcntl() with
733 F_ADD_SEALS fails with the error EPERM. Therefore, this seal
734 prevents any modifications to the set of seals itself. If the
735 initial set of seals of a file includes F_SEAL_SEAL, then this
736 effectively causes the set of seals to be constant and locked.
737 F_SEAL_SHRINK
739 If this seal is set, the file in question cannot be reduced in
740 size. This affects open(2) with the O_TRUNC flag as well as
741 truncate(2) and ftruncate(2). Those calls fail with EPERM if
742 you try to shrink the file in question. Increasing the file
743 size is still possible.
744 F_SEAL_GROW
746 If this seal is set, the size of the file in question cannot be
747 increased. This affects write(2) beyond the end of the file,
748 truncate(2), ftruncate(2), and fallocate(2). These calls fail
749 with EPERM if you use them to increase the file size. If you
750 keep the size or shrink it, those calls still work as expected.
751 F_SEAL_WRITE
753 If this seal is set, you cannot modify the contents of the file.
754 Note that shrinking or growing the size of the file is still
755 possible and allowed. Thus, this seal is normally used in com‐
756 bination with one of the other seals. This seal affects
757 write(2) and fallocate(2) (only in combination with the FAL‐
758 LOC_FL_PUNCH_HOLE flag). Those calls fail with EPERM if this
759 seal is set. Furthermore, trying to create new shared, writable
760 memory-mappings via mmap(2) will also fail with EPERM.
761 Using the F_ADD_SEALS operation to set the F_SEAL_WRITE seal
763 fails with EBUSY if any writable, shared mapping exists. Such
764 mappings must be unmapped before you can add this seal. Fur‐
765 thermore, if there are any asynchronous I/O operations (io_sub‐
766 mit(2)) pending on the file, all outstanding writes will be dis‐
767 carded.
768 F_SEAL_FUTURE_WRITE (since Linux 5.1)
770 The effect of this seal is similar to F_SEAL_WRITE, but the con‐
771 tents of the file can still be modified via shared writable map‐
772 pings that were created prior to the seal being set. Any at‐
773 tempt to create a new writable mapping on the file via mmap(2)
774 will fail with EPERM. Likewise, an attempt to write to the file
775 via write(2) will fail with EPERM.
776 Using this seal, one process can create a memory buffer that it
778 can continue to modify while sharing that buffer on a "read-
779 only" basis with other processes.
780 File read/write hints
782 Write lifetime hints can be used to inform the kernel about the rela‐
783 tive expected lifetime of writes on a given inode or via a particular
784 open file description. (See open(2) for an explanation of open file
785 descriptions.) In this context, the term "write lifetime" means the
786 expected time the data will live on media, before being overwritten or
787 erased.
788 An application may use the different hint values specified below to
790 separate writes into different write classes, so that multiple users or
791 applications running on a single storage back-end can aggregate their
792 I/O patterns in a consistent manner. However, there are no functional
793 semantics implied by these flags, and different I/O classes can use the
794 write lifetime hints in arbitrary ways, so long as the hints are used
795 consistently.
796 The following operations can be applied to the file descriptor, fd:
798 F_GET_RW_HINT (uint64_t *; since Linux 4.13)
800 Returns the value of the read/write hint associated with the un‐
801 derlying inode referred to by fd.
802 F_SET_RW_HINT (uint64_t *; since Linux 4.13)
804 Sets the read/write hint value associated with the underlying
805 inode referred to by fd. This hint persists until either it is
806 explicitly modified or the underlying filesystem is unmounted.
807 F_GET_FILE_RW_HINT (uint64_t *; since Linux 4.13)
809 Returns the value of the read/write hint associated with the
810 open file description referred to by fd.
811 F_SET_FILE_RW_HINT (uint64_t *; since Linux 4.13)
813 Sets the read/write hint value associated with the open file de‐
814 scription referred to by fd.
815 If an open file description has not been assigned a read/write hint,
817 then it shall use the value assigned to the inode, if any.
818 The following read/write hints are valid since Linux 4.13:
820 RWH_WRITE_LIFE_NOT_SET
822 No specific hint has been set. This is the default value.
823 RWH_WRITE_LIFE_NONE
825 No specific write lifetime is associated with this file or in‐
826 ode.
827 RWH_WRITE_LIFE_SHORT
829 Data written to this inode or via this open file description is
830 expected to have a short lifetime.
831 RWH_WRITE_LIFE_MEDIUM
833 Data written to this inode or via this open file description is
834 expected to have a lifetime longer than data written with
835 RWH_WRITE_LIFE_SHORT.
836 RWH_WRITE_LIFE_LONG
838 Data written to this inode or via this open file description is
839 expected to have a lifetime longer than data written with
840 RWH_WRITE_LIFE_MEDIUM.
841 RWH_WRITE_LIFE_EXTREME
843 Data written to this inode or via this open file description is
844 expected to have a lifetime longer than data written with
845 RWH_WRITE_LIFE_LONG.
846 All the write-specific hints are relative to each other, and no indi‐
848 vidual absolute meaning should be attributed to them.
849
851 For a successful call, the return value depends on the operation:
852 F_DUPFD
854 The new file descriptor.
855 F_GETFD
857 Value of file descriptor flags.
858 F_GETFL
860 Value of file status flags.
861 F_GETLEASE
863 Type of lease held on file descriptor.
864 F_GETOWN
866 Value of file descriptor owner.
867 F_GETSIG
869 Value of signal sent when read or write becomes possible, or
870 zero for traditional SIGIO behavior.
871 F_GETPIPE_SZ, F_SETPIPE_SZ
873 The pipe capacity.
874 F_GET_SEALS
876 A bit mask identifying the seals that have been set for the in‐
877 ode referred to by fd.
878 All other commands
880 Zero.
881 On error, -1 is returned, and errno is set to indicate the error.
883
885 EACCES or EAGAIN
886 Operation is prohibited by locks held by other processes.
887 EAGAIN The operation is prohibited because the file has been memory-
889 mapped by another process.
890 EBADF fd is not an open file descriptor
892 EBADF cmd is F_SETLK or F_SETLKW and the file descriptor open mode
894 doesn't match with the type of lock requested.
895 EBUSY cmd is F_SETPIPE_SZ and the new pipe capacity specified in arg
897 is smaller than the amount of buffer space currently used to
898 store data in the pipe.
899 EBUSY cmd is F_ADD_SEALS, arg includes F_SEAL_WRITE, and there exists
901 a writable, shared mapping on the file referred to by fd.
902 EDEADLK
904 It was detected that the specified F_SETLKW command would cause
905 a deadlock.
906 EFAULT lock is outside your accessible address space.
908 EINTR cmd is F_SETLKW or F_OFD_SETLKW and the operation was inter‐
910 rupted by a signal; see signal(7).
911 EINTR cmd is F_GETLK, F_SETLK, F_OFD_GETLK, or F_OFD_SETLK, and the
913 operation was interrupted by a signal before the lock was
914 checked or acquired. Most likely when locking a remote file
915 (e.g., locking over NFS), but can sometimes happen locally.
916 EINVAL The value specified in cmd is not recognized by this kernel.
918 EINVAL cmd is F_ADD_SEALS and arg includes an unrecognized sealing bit.
920 EINVAL cmd is F_ADD_SEALS or F_GET_SEALS and the filesystem containing
922 the inode referred to by fd does not support sealing.
923 EINVAL cmd is F_DUPFD and arg is negative or is greater than the maxi‐
925 mum allowable value (see the discussion of RLIMIT_NOFILE in
926 getrlimit(2)).
927 EINVAL cmd is F_SETSIG and arg is not an allowable signal number.
929 EINVAL cmd is F_OFD_SETLK, F_OFD_SETLKW, or F_OFD_GETLK, and l_pid was
931 not specified as zero.
932 EMFILE cmd is F_DUPFD and the per-process limit on the number of open
934 file descriptors has been reached.
935 ENOLCK Too many segment locks open, lock table is full, or a remote
937 locking protocol failed (e.g., locking over NFS).
938 ENOTDIR
940 F_NOTIFY was specified in cmd, but fd does not refer to a direc‐
941 tory.
942 EPERM cmd is F_SETPIPE_SZ and the soft or hard user pipe limit has
944 been reached; see pipe(7).
945 EPERM Attempted to clear the O_APPEND flag on a file that has the ap‐
947 pend-only attribute set.
948 EPERM cmd was F_ADD_SEALS, but fd was not open for writing or the cur‐
950 rent set of seals on the file already includes F_SEAL_SEAL.
951
953 SVr4, 4.3BSD, POSIX.1-2001. Only the operations F_DUPFD, F_GETFD,
954 F_SETFD, F_GETFL, F_SETFL, F_GETLK, F_SETLK, and F_SETLKW are specified
955 in POSIX.1-2001.
956 F_GETOWN and F_SETOWN are specified in POSIX.1-2001. (To get their
958 definitions, define either _XOPEN_SOURCE with the value 500 or greater,
959 or _POSIX_C_SOURCE with the value 200809L or greater.)
960 F_DUPFD_CLOEXEC is specified in POSIX.1-2008. (To get this definition,
962 define _POSIX_C_SOURCE with the value 200809L or greater, or
963 _XOPEN_SOURCE with the value 700 or greater.)
964 F_GETOWN_EX, F_SETOWN_EX, F_SETPIPE_SZ, F_GETPIPE_SZ, F_GETSIG, F_SET‐
966 SIG, F_NOTIFY, F_GETLEASE, and F_SETLEASE are Linux-specific. (Define
967 the _GNU_SOURCE macro to obtain these definitions.)
968 F_OFD_SETLK, F_OFD_SETLKW, and F_OFD_GETLK are Linux-specific (and one
970 must define _GNU_SOURCE to obtain their definitions), but work is being
971 done to have them included in the next version of POSIX.1.
972 F_ADD_SEALS and F_GET_SEALS are Linux-specific.
974
976 The errors returned by dup2(2) are different from those returned by
977 F_DUPFD.
978 File locking
980 The original Linux fcntl() system call was not designed to handle large
981 file offsets (in the flock structure). Consequently, an fcntl64() sys‐
982 tem call was added in Linux 2.4. The newer system call employs a dif‐
983 ferent structure for file locking, flock64, and corresponding commands,
984 F_GETLK64, F_SETLK64, and F_SETLKW64. However, these details can be
985 ignored by applications using glibc, whose fcntl() wrapper function
986 transparently employs the more recent system call where it is avail‐
987 able.
988 Record locks
990 Since kernel 2.0, there is no interaction between the types of lock
991 placed by flock(2) and fcntl().
992 Several systems have more fields in struct flock such as, for example,
994 l_sysid (to identify the machine where the lock is held). Clearly,
995 l_pid alone is not going to be very useful if the process holding the
996 lock may live on a different machine; on Linux, while present on some
997 architectures (such as MIPS32), this field is not used.
998 The original Linux fcntl() system call was not designed to handle large
1000 file offsets (in the flock structure). Consequently, an fcntl64() sys‐
1001 tem call was added in Linux 2.4. The newer system call employs a dif‐
1002 ferent structure for file locking, flock64, and corresponding commands,
1003 F_GETLK64, F_SETLK64, and F_SETLKW64. However, these details can be
1004 ignored by applications using glibc, whose fcntl() wrapper function
1005 transparently employs the more recent system call where it is avail‐
1006 able.
1007 Record locking and NFS
1009 Before Linux 3.12, if an NFSv4 client loses contact with the server for
1010 a period of time (defined as more than 90 seconds with no communica‐
1011 tion), it might lose and regain a lock without ever being aware of the
1012 fact. (The period of time after which contact is assumed lost is known
1013 as the NFSv4 leasetime. On a Linux NFS server, this can be determined
1014 by looking at /proc/fs/nfsd/nfsv4leasetime, which expresses the period
1015 in seconds. The default value for this file is 90.) This scenario po‐
1016 tentially risks data corruption, since another process might acquire a
1017 lock in the intervening period and perform file I/O.
1018 Since Linux 3.12, if an NFSv4 client loses contact with the server, any
1020 I/O to the file by a process which "thinks" it holds a lock will fail
1021 until that process closes and reopens the file. A kernel parameter,
1022 nfs.recover_lost_locks, can be set to 1 to obtain the pre-3.12 behav‐
1023 ior, whereby the client will attempt to recover lost locks when contact
1024 is reestablished with the server. Because of the attendant risk of
1025 data corruption, this parameter defaults to 0 (disabled).
1026
1028 F_SETFL
1029 It is not possible to use F_SETFL to change the state of the O_DSYNC
1030 and O_SYNC flags. Attempts to change the state of these flags are
1031 silently ignored.
1032 F_GETOWN
1034 A limitation of the Linux system call conventions on some architectures
1035 (notably i386) means that if a (negative) process group ID to be re‐
1036 turned by F_GETOWN falls in the range -1 to -4095, then the return
1037 value is wrongly interpreted by glibc as an error in the system call;
1038 that is, the return value of fcntl() will be -1, and errno will contain
1039 the (positive) process group ID. The Linux-specific F_GETOWN_EX opera‐
1040 tion avoids this problem. Since glibc version 2.11, glibc makes the
1041 kernel F_GETOWN problem invisible by implementing F_GETOWN using
1042 F_GETOWN_EX.
1043 F_SETOWN
1045 In Linux 2.4 and earlier, there is bug that can occur when an unprivi‐
1046 leged process uses F_SETOWN to specify the owner of a socket file de‐
1047 scriptor as a process (group) other than the caller. In this case, fc‐
1048 ntl() can return -1 with errno set to EPERM, even when the owner
1049 process (group) is one that the caller has permission to send signals
1050 to. Despite this error return, the file descriptor owner is set, and
1051 signals will be sent to the owner.
1052 Deadlock detection
1054 The deadlock-detection algorithm employed by the kernel when dealing
1055 with F_SETLKW requests can yield both false negatives (failures to de‐
1056 tect deadlocks, leaving a set of deadlocked processes blocked indefi‐
1057 nitely) and false positives (EDEADLK errors when there is no deadlock).
1058 For example, the kernel limits the lock depth of its dependency search
1059 to 10 steps, meaning that circular deadlock chains that exceed that
1060 size will not be detected. In addition, the kernel may falsely indi‐
1061 cate a deadlock when two or more processes created using the clone(2)
1062 CLONE_FILES flag place locks that appear (to the kernel) to conflict.
1063 Mandatory locking
1065 The Linux implementation of mandatory locking is subject to race condi‐
1066 tions which render it unreliable: a write(2) call that overlaps with a
1067 lock may modify data after the mandatory lock is acquired; a read(2)
1068 call that overlaps with a lock may detect changes to data that were
1069 made only after a write lock was acquired. Similar races exist between
1070 mandatory locks and mmap(2). It is therefore inadvisable to rely on
1071 mandatory locking.
1072
1074 dup2(2), flock(2), open(2), socket(2), lockf(3), capabilities(7), fea‐
1075 ture_test_macros(7), lslocks(8)
1076 locks.txt, mandatory-locking.txt, and dnotify.txt in the Linux kernel
1078 source directory Documentation/filesystems/ (on older kernels, these
1079 files are directly under the Documentation/ directory, and manda‐
1080 tory-locking.txt is called mandatory.txt)
1081
1083 This page is part of release 5.13 of the Linux man-pages project. A
1084 description of the project, information about reporting bugs, and the
1085 latest version of this page, can be found at
1086 https://www.kernel.org/doc/man-pages/.
1087Linux 2021-03-22 FCNTL(2)