1FCNTL(2) Linux Programmer's Manual FCNTL(2)
2
6 fcntl - manipulate file descriptor
7
9 #include <unistd.h>
10 #include <fcntl.h>
11 int fcntl(int fd, int cmd);
13 int fcntl(int fd, int cmd, long arg);
14 int fcntl(int fd, int cmd, struct flock *lock);
15
17 fcntl() performs one of the operations described below on the open file
18 descriptor fd. The operation is determined by cmd.
19 Duplicating a file descriptor
21 F_DUPFD
22 Find the lowest numbered available file descriptor greater than
23 or equal to arg and make it be a copy of fd. This is different
24 from dup2(2) which uses exactly the descriptor specified.
25 On success, the new descriptor is returned.
27 See dup(2) for further details.
29 File descriptor flags
31 The following commands manipulate the flags associated with a file
32 descriptor. Currently, only one such flag is defined: FD_CLOEXEC, the
33 close-on-exec flag. If the FD_CLOEXEC bit is 0, the file descriptor
34 will remain open across an execve(2), otherwise it will be closed.
35 F_GETFD
37 Read the file descriptor flags.
38 F_SETFD
40 Set the file descriptor flags to the value specified by arg.
41 File status flags
43 Each open file description has certain associated status flags, ini‐
44 tialized by open(2) and possibly modified by fcntl(2). Duplicated file
45 descriptors (made with dup(), fcntl(F_DUPFD), fork(), etc.) refer to
46 the same open file description, and thus share the same file status
47 flags.
48 The file status flags and their semantics are described in open(2).
50 F_GETFL
52 Read the file status flags.
53 F_SETFL
55 Set the file status flags to the value specified by arg. File
56 access mode (O_RDONLY, O_WRONLY, O_RDWR) and file creation flags
57 (i.e., O_CREAT, O_EXCL, O_NOCTTY, O_TRUNC) in arg are ignored.
58 On Linux this command can only change the O_APPEND, O_ASYNC,
59 O_DIRECT, O_NOATIME, and O_NONBLOCK flags.
60 Advisory locking
62 F_GETLK, F_SETLK and F_SETLKW are used to acquire, release, and test
63 for the existence of record locks (also known as file-segment or file-
64 region locks). The third argument lock is a pointer to a structure
65 that has at least the following fields (in unspecified order).
66 struct flock {
68 ...
69 short l_type; /* Type of lock: F_RDLCK,
70 F_WRLCK, F_UNLCK */
71 short l_whence; /* How to interpret l_start:
72 SEEK_SET, SEEK_CUR, SEEK_END */
73 off_t l_start; /* Starting offset for lock */
74 off_t l_len; /* Number of bytes to lock */
75 pid_t l_pid; /* PID of process blocking our lock
76 (F_GETLK only) */
77 ...
78 };
79 The l_whence, l_start, and l_len fields of this structure specify the
81 range of bytes we wish to lock. l_start is the starting offset for the
82 lock, and is interpreted relative to either: the start of the file (if
83 l_whence is SEEK_SET); the current file offset (if l_whence is
84 SEEK_CUR); or the end of the file (if l_whence is SEEK_END). In the
85 final two cases, l_start can be a negative number provided the offset
86 does not lie before the start of the file. l_len is a non-negative
87 integer (but see the NOTES below) specifying the number of bytes to be
88 locked. Bytes past the end of the file may be locked, but not bytes
89 before the start of the file. Specifying 0 for l_len has the special
90 meaning: lock all bytes starting at the location specified by l_whence
91 and l_start through to the end of file, no matter how large the file
92 grows.
93 The l_type field can be used to place a read (F_RDLCK) or a write
95 (F_WRLCK) lock on a file. Any number of processes may hold a read lock
96 (shared lock) on a file region, but only one process may hold a write
97 lock (exclusive lock). An exclusive lock excludes all other locks, both
98 shared and exclusive. A single process can hold only one type of lock
99 on a file region; if a new lock is applied to an already-locked region,
100 then the existing lock is converted to the new lock type. (Such con‐
101 versions may involve splitting, shrinking, or coalescing with an exist‐
102 ing lock if the byte range specified by the new lock does not precisely
103 coincide with the range of the existing lock.)
104 F_SETLK
106 Acquire a lock (when l_type is F_RDLCK or F_WRLCK) or release a
107 lock (when l_type is F_UNLCK) on the bytes specified by the
108 l_whence, l_start, and l_len fields of lock. If a conflicting
109 lock is held by another process, this call returns -1 and sets
110 errno to EACCES or EAGAIN.
111 F_SETLKW
113 As for F_SETLK, but if a conflicting lock is held on the file,
114 then wait for that lock to be released. If a signal is caught
115 while waiting, then the call is interrupted and (after the sig‐
116 nal handler has returned) returns immediately (with return value
117 -1 and errno set to EINTR).
118 F_GETLK
120 On input to this call, lock describes a lock we would like to
121 place on the file. If the lock could be placed, fcntl() does
122 not actually place it, but returns F_UNLCK in the l_type field
123 of lock and leaves the other fields of the structure unchanged.
124 If one or more incompatible locks would prevent this lock being
125 placed, then fcntl() returns details about one of these locks in
126 the l_type, l_whence, l_start, and l_len fields of lock and sets
127 l_pid to be the PID of the process holding that lock.
128 In order to place a read lock, fd must be open for reading. In order
130 to place a write lock, fd must be open for writing. To place both
131 types of lock, open a file read-write.
132 As well as being removed by an explicit F_UNLCK, record locks are auto‐
134 matically released when the process terminates or if it closes any file
135 descriptor referring to a file on which locks are held. This is bad:
136 it means that a process can lose the locks on a file like /etc/passwd
137 or /etc/mtab when for some reason a library function decides to open,
138 read and close it.
139 Record locks are not inherited by a child created via fork(2), but are
141 preserved across an execve(2).
142 Because of the buffering performed by the stdio(3) library, the use of
144 record locking with routines in that package should be avoided; use
145 read(2) and write(2) instead.
146 Mandatory locking
148 (Non-POSIX.) The above record locks may be either advisory or manda‐
149 tory, and are advisory by default.
150 Advisory locks are not enforced and are useful only between cooperating
152 processes.
153 Mandatory locks are enforced for all processes. If a process tries to
155 perform an incompatible access (e.g., read(2) or write(2)) on a file
156 region that has an incompatible mandatory lock, then the result depends
157 upon whether the O_NONBLOCK flag is enabled for its open file descrip‐
158 tion. If the O_NONBLOCK flag is not enabled, then system call is
159 blocked until the lock is removed or converted to a mode that is com‐
160 patible with the access. If the O_NONBLOCK flag is enabled, then the
161 system call fails with the error EAGAIN or EWOULDBLOCK.
162 To make use of mandatory locks, mandatory locking must be enabled both
164 on the file system that contains the file to be locked, and on the file
165 itself. Mandatory locking is enabled on a file system using the "-o
166 mand" option to mount(8), or the MS_MANDLOCK flag for mount(2). Manda‐
167 tory locking is enabled on a file by disabling group execute permission
168 on the file and enabling the set-group-ID permission bit (see chmod(1)
169 and chmod(2)).
170 Managing signals
172 F_GETOWN, F_SETOWN, F_GETSIG and F_SETSIG are used to manage I/O avail‐
173 ability signals:
174 F_GETOWN
176 Get the process ID or process group currently receiving SIGIO
177 and SIGURG signals for events on file descriptor fd. Process
178 IDs are returned as positive values; process group IDs are
179 returned as negative values (but see BUGS below).
180 F_SETOWN
182 Set the process ID or process group ID that will receive SIGIO
183 and SIGURG signals for events on file descriptor fd. A process
184 ID is specified as a positive value; a process group ID is spec‐
185 ified as a negative value. Most commonly, the calling process
186 specifies itself as the owner (that is, arg is specified as get‐
187 pid()).
188 If you set the O_ASYNC status flag on a file descriptor (either
190 by providing this flag with the open(2) call, or by using the
191 F_SETFL command of fcntl()), a SIGIO signal is sent whenever
192 input or output becomes possible on that file descriptor.
193 F_SETSIG can be used to obtain delivery of a signal other than
194 SIGIO. If this permission check fails, then the signal is
195 silently discarded.
196 Sending a signal to the owner process (group) specified by
198 F_SETOWN is subject to the same permissions checks as are
199 described for kill(2), where the sending process is the one that
200 employs F_SETOWN (but see BUGS below).
201 If the file descriptor fd refers to a socket, F_SETOWN also
203 selects the recipient of SIGURG signals that are delivered when
204 out-of-band data arrives on that socket. (SIGURG is sent in any
205 situation where select(2) would report the socket as having an
206 "exceptional condition".)
207 If a non-zero value is given to F_SETSIG in a multi-threaded
209 process running with a threading library that supports thread
210 groups (e.g., NPTL), then a positive value given to F_SETOWN has
211 a different meaning: instead of being a process ID identifying a
212 whole process, it is a thread ID identifying a specific thread
213 within a process. Consequently, it may be necessary to pass
214 F_SETOWN the result of gettid() instead of getpid() to get sen‐
215 sible results when F_SETSIG is used. (In current Linux thread‐
216 ing implementations, a main thread's thread ID is the same as
217 its process ID. This means that a single-threaded program can
218 equally use gettid() or getpid() in this scenario.) Note, how‐
219 ever, that the statements in this paragraph do not apply to the
220 SIGURG signal generated for out-of-band data on a socket: this
221 signal is always sent to either a process or a process group,
222 depending on the value given to F_SETOWN. Note also that Linux
223 imposes a limit on the number of real-time signals that may be
224 queued to a process (see getrlimit(2) and signal(7)) and if this
225 limit is reached, then the kernel reverts to delivering SIGIO,
226 and this signal is delivered to the entire process rather than
227 to a specific thread.
228 F_GETSIG
230 Get the signal sent when input or output becomes possible. A
231 value of zero means SIGIO is sent. Any other value (including
232 SIGIO) is the signal sent instead, and in this case additional
233 info is available to the signal handler if installed with
234 SA_SIGINFO.
235 F_SETSIG
237 Sets the signal sent when input or output becomes possible. A
238 value of zero means to send the default SIGIO signal. Any other
239 value (including SIGIO) is the signal to send instead, and in
240 this case additional info is available to the signal handler if
241 installed with SA_SIGINFO.
242 Additionally, passing a non-zero value to F_SETSIG changes the
244 signal recipient from a whole process to a specific thread
245 within a process. See the description of F_SETOWN for more
246 details.
247 By using F_SETSIG with a non-zero value, and setting SA_SIGINFO
249 for the signal handler (see sigaction(2)), extra information
250 about I/O events is passed to the handler in a siginfo_t struc‐
251 ture. If the si_code field indicates the source is SI_SIGIO,
252 the si_fd field gives the file descriptor associated with the
253 event. Otherwise, there is no indication which file descriptors
254 are pending, and you should use the usual mechanisms (select(2),
255 poll(2), read(2) with O_NONBLOCK set etc.) to determine which
256 file descriptors are available for I/O.
257 By selecting a real time signal (value >= SIGRTMIN), multiple
259 I/O events may be queued using the same signal numbers. (Queu‐
260 ing is dependent on available memory). Extra information is
261 available if SA_SIGINFO is set for the signal handler, as above.
262 Using these mechanisms, a program can implement fully asynchronous I/O
264 without using select(2) or poll(2) most of the time.
265 The use of O_ASYNC, F_GETOWN, F_SETOWN is specific to BSD and Linux.
267 F_GETSIG and F_SETSIG are Linux specific. POSIX has asynchronous I/O
268 and the aio_sigevent structure to achieve similar things; these are
269 also available in Linux as part of the GNU C Library (Glibc).
270 Leases
272 F_SETLEASE and F_GETLEASE (Linux 2.4 onwards) are used (respectively)
273 to establish and retrieve the current setting of the calling process's
274 lease on the file referred to by fd. A file lease provides a mechanism
275 whereby the process holding the lease (the "lease holder") is notified
276 (via delivery of a signal) when a process (the "lease breaker") tries
277 to open(2) or truncate(2) that file.
278 F_SETLEASE
280 Set or remove a file lease according to which of the following
281 values is specified in the integer arg:
282 F_RDLCK
285 Take out a read lease. This will cause the calling
286 process to be notified when the file is opened for writ‐
287 ing or is truncated. A read lease can only be placed on
288 a file descriptor that is opened read-only.
289 F_WRLCK
291 Take out a write lease. This will cause the caller to be
292 notified when the file is opened for reading or writing
293 or is truncated. A write lease may be placed on a file
294 only if no other process currently has the file open.
295 F_UNLCK
297 Remove our lease from the file.
298 A process may hold only one type of lease on a file.
300 Leases may only be taken out on regular files. An unprivileged process
302 may only take out a lease on a file whose UID matches the file system
303 UID of the process. A process with the CAP_LEASE capability may take
304 out leases on arbitrary files.
305 F_GETLEASE
307 Indicates what type of lease we hold on the file referred to by
308 fd by returning either F_RDLCK, F_WRLCK, or F_UNLCK, indicating,
309 respectively, that the calling process holds a read, a write, or
310 no lease on the file. (The third argument to fcntl() is omit‐
311 ted.)
312 When a process (the "lease breaker") performs an open() or truncate()
314 that conflicts with a lease established via F_SETLEASE, the system call
315 is blocked by the kernel and the kernel notifies the lease holder by
316 sending it a signal (SIGIO by default). The lease holder should
317 respond to receipt of this signal by doing whatever cleanup is required
318 in preparation for the file to be accessed by another process (e.g.,
319 flushing cached buffers) and then either remove or downgrade its lease.
320 A lease is removed by performing an F_SETLEASE command specifying arg
321 as F_UNLCK. If we currently hold a write lease on the file, and the
322 lease breaker is opening the file for reading, then it is sufficient to
323 downgrade the lease to a read lease. This is done by performing an
324 F_SETLEASE command specifying arg as F_RDLCK.
325 If the lease holder fails to downgrade or remove the lease within the
327 number of seconds specified in /proc/sys/fs/lease-break-time then the
328 kernel forcibly removes or downgrades the lease holder's lease.
329 Once the lease has been voluntarily or forcibly removed or downgraded,
331 and assuming the lease breaker has not unblocked its system call, the
332 kernel permits the lease breaker's system call to proceed.
333 If the lease breaker's blocked open() or truncate() is interrupted by a
335 signal handler, then the system call fails with the error EINTR, but
336 the other steps still occur as described above. If the lease breaker
337 is killed by a signal while blocked in open() or truncate(), then the
338 other steps still occur as described above. If the lease breaker spec‐
339 ifies the O_NONBLOCK flag when calling open(), then the call immedi‐
340 ately fails with the error EWOULDBLOCK, but the other steps still occur
341 as described above.
342 The default signal used to notify the lease holder is SIGIO, but this
344 can be changed using the F_SETSIG command to fcntl(). If a F_SETSIG
345 command is performed (even one specifying SIGIO), and the signal han‐
346 dler is established using SA_SIGINFO, then the handler will receive a
347 siginfo_t structure as its second argument, and the si_fd field of this
348 argument will hold the descriptor of the leased file that has been
349 accessed by another process. (This is useful if the caller holds
350 leases against multiple files).
351 File and directory change notification (dnotify)
353 F_NOTIFY
354 (Linux 2.4 onwards) Provide notification when the directory
355 referred to by fd or any of the files that it contains is
356 changed. The events to be notified are specified in arg, which
357 is a bit mask specified by ORing together zero or more of the
358 following bits:
359 Bit Description (event in directory)
362 ─────────────────────────────────────────────────────────────
363 DN_ACCESS A file was accessed (read, pread, readv)
364 DN_MODIFY A file was modified (write, pwrite,
365 writev, truncate, ftruncate)
366 DN_CREATE A file was created (open, creat, mknod,
367 mkdir, link, symlink, rename)
368 DN_DELETE A file was unlinked (unlink, rename to
369 another directory, rmdir)
370 DN_RENAME A file was renamed within this
371 directory (rename)
372 DN_ATTRIB The attributes of a file were changed
373 (chown, chmod, utime[s])
374 (In order to obtain these definitions, the _GNU_SOURCE feature
376 test macro must be defined.)
377 Directory notifications are normally "one-shot", and the appli‐
379 cation must re-register to receive further notifications.
380 Alternatively, if DN_MULTISHOT is included in arg, then notifi‐
381 cation will remain in effect until explicitly removed.
382 A series of F_NOTIFY requests is cumulative, with the events in
384 arg being added to the set already monitored. To disable noti‐
385 fication of all events, make an F_NOTIFY call specifying arg as
386 0.
387 Notification occurs via delivery of a signal. The default sig‐
389 nal is SIGIO, but this can be changed using the F_SETSIG command
390 to fcntl(). In the latter case, the signal handler receives a
391 siginfo_t structure as its second argument (if the handler was
392 established using SA_SIGINFO) and the si_fd field of this struc‐
393 ture contains the file descriptor which generated the notifica‐
394 tion (useful when establishing notification on multiple directo‐
395 ries).
396 Especially when using DN_MULTISHOT, a real time signal should be
398 used for notification, so that multiple notifications can be
399 queued.
400 NOTE: New applications should consider using the inotify inter‐
402 face (available since kernel 2.6.13), which provides a superior
403 interface for obtaining notifications of file system events.
404 See inotify(7).
405
407 For a successful call, the return value depends on the operation:
408 F_DUPFD The new descriptor.
410 F_GETFD Value of flags.
412 F_GETFL Value of flags.
414 F_GETOWN Value of descriptor owner.
416 F_GETSIG Value of signal sent when read or write becomes possible, or
418 zero for traditional SIGIO behaviour.
419 All other commands
421 Zero.
422 On error, -1 is returned, and errno is set appropriately.
424
426 EACCES or EAGAIN
427 Operation is prohibited by locks held by other processes.
428 EAGAIN The operation is prohibited because the file has been memory-
430 mapped by another process.
431 EBADF fd is not an open file descriptor, or the command was F_SETLK or
433 F_SETLKW and the file descriptor open mode doesn't match with
434 the type of lock requested.
435 EDEADLK
437 It was detected that the specified F_SETLKW command would cause
438 a deadlock.
439 EFAULT lock is outside your accessible address space.
441 EINTR For F_SETLKW, the command was interrupted by a signal. For
443 F_GETLK and F_SETLK, the command was interrupted by a signal
444 before the lock was checked or acquired. Most likely when lock‐
445 ing a remote file (e.g. locking over NFS), but can sometimes
446 happen locally.
447 EINVAL For F_DUPFD, arg is negative or is greater than the maximum
449 allowable value. For F_SETSIG, arg is not an allowable signal
450 number.
451 EMFILE For F_DUPFD, the process already has the maximum number of file
453 descriptors open.
454 ENOLCK Too many segment locks open, lock table is full, or a remote
456 locking protocol failed (e.g. locking over NFS).
457 EPERM Attempted to clear the O_APPEND flag on a file that has the
459 append-only attribute set.
460
462 The errors returned by dup2() are different from those returned by
463 F_DUPFD.
464 Since kernel 2.0, there is no interaction between the types of lock
466 placed by flock(2) and fcntl(2).
467 POSIX.1-2001 allows l_len to be negative. (And if it is, the interval
469 described by the lock covers bytes l_start+l_len up to and including
470 l_start-1.) This is supported by Linux since Linux 2.4.21 and 2.5.49.
471 Several systems have more fields in struct flock such as e.g. l_sysid.
473 Clearly, l_pid alone is not going to be very useful if the process
474 holding the lock may live on a different machine.
475
477 A limitation of the Linux system call conventions on some architectures
478 (notably x86) means that if a (negative) process group ID to be
479 returned by F_GETOWN falls in the range -1 to -4095, then the return
480 value is wrongly interpreted by glibc as an error in the system call;
481 that is, the return value of fcntl() will be -1, and errno will contain
482 the (positive) process group ID.
483 In Linux 2.4 and earlier, there is bug that can occur when an unprivi‐
485 leged process uses F_SETOWN to specify the owner of a socket file
486 descriptor as a process (group) other than the caller. In this case,
487 fcntl() can return -1 with errno set to EPERM, even when the owner
488 process (group) is one that the caller has permission to send signals
489 to. Despite this error return, the file descriptor owner is set, and
490 signals will be sent to the owner.
491
493 SVr4, 4.3BSD, POSIX.1-2001. Only the operations F_DUPFD, F_GETFD,
494 F_SETFD, F_GETFL, F_SETFL, F_GETLK, F_SETLK, F_SETLKW, F_GETOWN, and
495 F_SETOWN are specified in POSIX.1-2001.
496 F_GETSIG, F_SETSIG, F_NOTIFY, F_GETLEASE, and F_SETLEASE are Linux spe‐
498 cific. (Define the _GNU_SOURCE macro to obtain these definitions.)
499
501 dup2(2), flock(2), open(2), socket(2), lockf(3), capabilities(7), fea‐
502 ture_test_macros(7)
503 See also locks.txt, mandatory.txt, and dnotify.txt in
505 /usr/src/linux/Documentation.
506Linux 2.6.14 2005-05-20 FCNTL(2)