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, ... /* arg */ );
13
15 fcntl() performs one of the operations described below on the open file
16 descriptor fd. The operation is determined by cmd.
17 fcntl() can take an optional third argument. Whether or not this argu‐
19 ment is required is determined by cmd. The required argument type is
20 indicated in parentheses after each cmd name (in most cases, the
21 required type is long, and we identify the argument using the name
22 arg), or void is specified if the argument is not required.
23 Duplicating a file descriptor
25 F_DUPFD (long)
26 Find the lowest numbered available file descriptor greater than
27 or equal to arg and make it be a copy of fd. This is different
28 from dup2(2), which uses exactly the descriptor specified.
29 On success, the new descriptor is returned.
31 See dup(2) for further details.
33 F_DUPFD_CLOEXEC (long; since Linux 2.6.24)
35 As for F_DUPFD, but additionally set the close-on-exec flag for
36 the duplicate descriptor. Specifying this flag permits a pro‐
37 gram to avoid an additional fcntl() F_SETFD operation to set the
38 FD_CLOEXEC flag. For an explanation of why this flag is useful,
39 see the description of O_CLOEXEC in open(2).
40 File descriptor flags
42 The following commands manipulate the flags associated with a file
43 descriptor. Currently, only one such flag is defined: FD_CLOEXEC, the
44 close-on-exec flag. If the FD_CLOEXEC bit is 0, the file descriptor
45 will remain open across an execve(2), otherwise it will be closed.
46 F_GETFD (void)
48 Read the file descriptor flags; arg is ignored.
49 F_SETFD (long)
51 Set the file descriptor flags to the value specified by arg.
52 File status flags
54 Each open file description has certain associated status flags, ini‐
55 tialized by open(2) and possibly modified by fcntl(). Duplicated file
56 descriptors (made with dup(2), fcntl(F_DUPFD), fork(2), etc.) refer to
57 the same open file description, and thus share the same file status
58 flags.
59 The file status flags and their semantics are described in open(2).
61 F_GETFL (void)
63 Read the file status flags; arg is ignored.
64 F_SETFL (long)
66 Set the file status flags to the value specified by arg. File
67 access mode (O_RDONLY, O_WRONLY, O_RDWR) and file creation flags
68 (i.e., O_CREAT, O_EXCL, O_NOCTTY, O_TRUNC) in arg are ignored.
69 On Linux this command can only change the O_APPEND, O_ASYNC,
70 O_DIRECT, O_NOATIME, and O_NONBLOCK flags.
71 Advisory locking
73 F_GETLK, F_SETLK and F_SETLKW are used to acquire, release, and test
74 for the existence of record locks (also known as file-segment or file-
75 region locks). The third argument, lock, is a pointer to a structure
76 that has at least the following fields (in unspecified order).
77 struct flock {
79 ...
80 short l_type; /* Type of lock: F_RDLCK,
81 F_WRLCK, F_UNLCK */
82 short l_whence; /* How to interpret l_start:
83 SEEK_SET, SEEK_CUR, SEEK_END */
84 off_t l_start; /* Starting offset for lock */
85 off_t l_len; /* Number of bytes to lock */
86 pid_t l_pid; /* PID of process blocking our lock
87 (F_GETLK only) */
88 ...
89 };
90 The l_whence, l_start, and l_len fields of this structure specify the
92 range of bytes we wish to lock. Bytes past the end of the file may be
93 locked, but not bytes before the start of the file.
94 l_start is the starting offset for the lock, and is interpreted rela‐
96 tive to either: the start of the file (if l_whence is SEEK_SET); the
97 current file offset (if l_whence is SEEK_CUR); or the end of the file
98 (if l_whence is SEEK_END). In the final two cases, l_start can be a
99 negative number provided the offset does not lie before the start of
100 the file.
101 l_len specifies the number of bytes to be locked. If l_len is posi‐
103 tive, then the range to be locked covers bytes l_start up to and
104 including l_start+l_len-1. Specifying 0 for l_len has the special
105 meaning: lock all bytes starting at the location specified by l_whence
106 and l_start through to the end of file, no matter how large the file
107 grows.
108 POSIX.1-2001 allows (but does not require) an implementation to support
110 a negative l_len value; if l_len is negative, the interval described by
111 lock covers bytes l_start+l_len up to and including l_start-1. This is
112 supported by Linux since kernel versions 2.4.21 and 2.5.49.
113 The l_type field can be used to place a read (F_RDLCK) or a write
115 (F_WRLCK) lock on a file. Any number of processes may hold a read lock
116 (shared lock) on a file region, but only one process may hold a write
117 lock (exclusive lock). An exclusive lock excludes all other locks,
118 both shared and exclusive. A single process can hold only one type of
119 lock on a file region; if a new lock is applied to an already-locked
120 region, then the existing lock is converted to the new lock type.
121 (Such conversions may involve splitting, shrinking, or coalescing with
122 an existing lock if the byte range specified by the new lock does not
123 precisely coincide with the range of the existing lock.)
124 F_SETLK (struct flock *)
126 Acquire a lock (when l_type is F_RDLCK or F_WRLCK) or release a
127 lock (when l_type is F_UNLCK) on the bytes specified by the
128 l_whence, l_start, and l_len fields of lock. If a conflicting
129 lock is held by another process, this call returns -1 and sets
130 errno to EACCES or EAGAIN.
131 F_SETLKW (struct flock *)
133 As for F_SETLK, but if a conflicting lock is held on the file,
134 then wait for that lock to be released. If a signal is caught
135 while waiting, then the call is interrupted and (after the sig‐
136 nal handler has returned) returns immediately (with return value
137 -1 and errno set to EINTR; see signal(7)).
138 F_GETLK (struct flock *)
140 On input to this call, lock describes a lock we would like to
141 place on the file. If the lock could be placed, fcntl() does
142 not actually place it, but returns F_UNLCK in the l_type field
143 of lock and leaves the other fields of the structure unchanged.
144 If one or more incompatible locks would prevent this lock being
145 placed, then fcntl() returns details about one of these locks in
146 the l_type, l_whence, l_start, and l_len fields of lock and sets
147 l_pid to be the PID of the process holding that lock.
148 In order to place a read lock, fd must be open for reading. In order
150 to place a write lock, fd must be open for writing. To place both
151 types of lock, open a file read-write.
152 As well as being removed by an explicit F_UNLCK, record locks are auto‐
154 matically released when the process terminates or if it closes any file
155 descriptor referring to a file on which locks are held. This is bad:
156 it means that a process can lose the locks on a file like /etc/passwd
157 or /etc/mtab when for some reason a library function decides to open,
158 read and close it.
159 Record locks are not inherited by a child created via fork(2), but are
161 preserved across an execve(2).
162 Because of the buffering performed by the stdio(3) library, the use of
164 record locking with routines in that package should be avoided; use
165 read(2) and write(2) instead.
166 Mandatory locking
168 (Non-POSIX.) The above record locks may be either advisory or manda‐
169 tory, and are advisory by default.
170 Advisory locks are not enforced and are useful only between cooperating
172 processes.
173 Mandatory locks are enforced for all processes. If a process tries to
175 perform an incompatible access (e.g., read(2) or write(2)) on a file
176 region that has an incompatible mandatory lock, then the result depends
177 upon whether the O_NONBLOCK flag is enabled for its open file descrip‐
178 tion. If the O_NONBLOCK flag is not enabled, then system call is
179 blocked until the lock is removed or converted to a mode that is com‐
180 patible with the access. If the O_NONBLOCK flag is enabled, then the
181 system call fails with the error EAGAIN.
182 To make use of mandatory locks, mandatory locking must be enabled both
184 on the file system that contains the file to be locked, and on the file
185 itself. Mandatory locking is enabled on a file system using the "-o
186 mand" option to mount(8), or the MS_MANDLOCK flag for mount(2). Manda‐
187 tory locking is enabled on a file by disabling group execute permission
188 on the file and enabling the set-group-ID permission bit (see chmod(1)
189 and chmod(2)).
190 The Linux implementation of mandatory locking is unreliable. See BUGS
192 below.
193 Managing signals
195 F_GETOWN, F_SETOWN, F_GETOWN_EX, F_SETOWN_EX, F_GETSIG and F_SETSIG are
196 used to manage I/O availability signals:
197 F_GETOWN (void)
199 Return (as the function result) the process ID or process group
200 currently receiving SIGIO and SIGURG signals for events on file
201 descriptor fd. Process IDs are returned as positive values;
202 process group IDs are returned as negative values (but see BUGS
203 below). arg is ignored.
204 F_SETOWN (long)
206 Set the process ID or process group ID that will receive SIGIO
207 and SIGURG signals for events on file descriptor fd to the ID
208 given in arg. A process ID is specified as a positive value; a
209 process group ID is specified as a negative value. Most com‐
210 monly, the calling process specifies itself as the owner (that
211 is, arg is specified as getpid(2)).
212 If you set the O_ASYNC status flag on a file descriptor by using
214 the F_SETFL command of fcntl(), a SIGIO signal is sent whenever
215 input or output becomes possible on that file descriptor.
216 F_SETSIG can be used to obtain delivery of a signal other than
217 SIGIO. If this permission check fails, then the signal is
218 silently discarded.
219 Sending a signal to the owner process (group) specified by
221 F_SETOWN is subject to the same permissions checks as are
222 described for kill(2), where the sending process is the one that
223 employs F_SETOWN (but see BUGS below).
224 If the file descriptor fd refers to a socket, F_SETOWN also
226 selects the recipient of SIGURG signals that are delivered when
227 out-of-band data arrives on that socket. (SIGURG is sent in any
228 situation where select(2) would report the socket as having an
229 "exceptional condition".)
230 The following was true in 2.6.x kernels up to and including ker‐
232 nel 2.6.11:
233 If a nonzero value is given to F_SETSIG in a multi‐
235 threaded process running with a threading library that
236 supports thread groups (e.g., NPTL), then a positive
237 value given to F_SETOWN has a different meaning: instead
238 of being a process ID identifying a whole process, it is
239 a thread ID identifying a specific thread within a
240 process. Consequently, it may be necessary to pass
241 F_SETOWN the result of gettid(2) instead of getpid(2) to
242 get sensible results when F_SETSIG is used. (In current
243 Linux threading implementations, a main thread's thread
244 ID is the same as its process ID. This means that a sin‐
245 gle-threaded program can equally use gettid(2) or get‐
246 pid(2) in this scenario.) Note, however, that the state‐
247 ments in this paragraph do not apply to the SIGURG signal
248 generated for out-of-band data on a socket: this signal
249 is always sent to either a process or a process group,
250 depending on the value given to F_SETOWN.
251 The above behavior was accidentally dropped in Linux 2.6.12, and
253 won't be restored. From Linux 2.6.32 onwards, use F_SETOWN_EX
254 to target SIGIO and SIGURG signals at a particular thread.
255 F_GETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
257 Return the current file descriptor owner settings as defined by
258 a previous F_SETOWN_EX operation. The information is returned
259 in the structure pointed to by arg, which has the following
260 form:
261 struct f_owner_ex {
263 int type;
264 pid_t pid;
265 };
266 The type field will have one of the values F_OWNER_TID,
268 F_OWNER_PID, or F_OWNER_PGRP. The pid field is a positive inte‐
269 ger representing a thread ID, process ID, or process group ID.
270 See F_SETOWN_EX for more details.
271 F_SETOWN_EX (struct f_owner_ex *) (since Linux 2.6.32)
273 This operation performs a similar task to F_SETOWN. It allows
274 the caller to direct I/O availability signals to a specific
275 thread, process, or process group. The caller specifies the
276 target of signals via arg, which is a pointer to a f_owner_ex
277 structure. The type field has one of the following values,
278 which define how pid is interpreted:
279 F_OWNER_TID
281 Send the signal to the thread whose thread ID (the value
282 returned by a call to clone(2) or gettid(2)) is specified
283 in pid.
284 F_OWNER_PID
286 Send the signal to the process whose ID is specified in
287 pid.
288 F_OWNER_PGRP
290 Send the signal to the process group whose ID is speci‐
291 fied in pid. (Note that, unlike with F_SETOWN, a process
292 group ID is specified as a positive value here.)
293 F_GETSIG (void)
295 Return (as the function result) the signal sent when input or
296 output becomes possible. A value of zero means SIGIO is sent.
297 Any other value (including SIGIO) is the signal sent instead,
298 and in this case additional info is available to the signal han‐
299 dler if installed with SA_SIGINFO. arg is ignored.
300 F_SETSIG (long)
302 Set the signal sent when input or output becomes possible to the
303 value given in arg. A value of zero means to send the default
304 SIGIO signal. Any other value (including SIGIO) is the signal
305 to send instead, and in this case additional info is available
306 to the signal handler if installed with SA_SIGINFO.
307 By using F_SETSIG with a nonzero value, and setting SA_SIGINFO
309 for the signal handler (see sigaction(2)), extra information
310 about I/O events is passed to the handler in a siginfo_t struc‐
311 ture. If the si_code field indicates the source is SI_SIGIO,
312 the si_fd field gives the file descriptor associated with the
313 event. Otherwise, there is no indication which file descriptors
314 are pending, and you should use the usual mechanisms (select(2),
315 poll(2), read(2) with O_NONBLOCK set etc.) to determine which
316 file descriptors are available for I/O.
317 By selecting a real time signal (value >= SIGRTMIN), multiple
319 I/O events may be queued using the same signal numbers. (Queu‐
320 ing is dependent on available memory). Extra information is
321 available if SA_SIGINFO is set for the signal handler, as above.
322 Note that Linux imposes a limit on the number of real-time sig‐
324 nals that may be queued to a process (see getrlimit(2) and sig‐
325 nal(7)) and if this limit is reached, then the kernel reverts to
326 delivering SIGIO, and this signal is delivered to the entire
327 process rather than to a specific thread.
328 Using these mechanisms, a program can implement fully asynchronous I/O
330 without using select(2) or poll(2) most of the time.
331 The use of O_ASYNC, F_GETOWN, F_SETOWN is specific to BSD and Linux.
333 F_GETOWN_EX, F_SETOWN_EX, F_GETSIG, and F_SETSIG are Linux-specific.
334 POSIX has asynchronous I/O and the aio_sigevent structure to achieve
335 similar things; these are also available in Linux as part of the GNU C
336 Library (Glibc).
337 Leases
339 F_SETLEASE and F_GETLEASE (Linux 2.4 onwards) are used (respectively)
340 to establish a new lease, and retrieve the current lease, on the open
341 file description referred to by the file descriptor fd. A file lease
342 provides a mechanism whereby the process holding the lease (the "lease
343 holder") is notified (via delivery of a signal) when a process (the
344 "lease breaker") tries to open(2) or truncate(2) the file referred to
345 by that file descriptor.
346 F_SETLEASE (long)
348 Set or remove a file lease according to which of the following
349 values is specified in the integer arg:
350 F_RDLCK
352 Take out a read lease. This will cause the calling
353 process to be notified when the file is opened for writ‐
354 ing or is truncated. A read lease can only be placed on
355 a file descriptor that is opened read-only.
356 F_WRLCK
358 Take out a write lease. This will cause the caller to be
359 notified when the file is opened for reading or writing
360 or is truncated. A write lease may be placed on a file
361 only if there are no other open file descriptors for the
362 file.
363 F_UNLCK
365 Remove our lease from the file.
366 Leases are associated with an open file description (see open(2)).
368 This means that duplicate file descriptors (created by, for example,
369 fork(2) or dup(2)) refer to the same lease, and this lease may be modi‐
370 fied or released using any of these descriptors. Furthermore, the
371 lease is released by either an explicit F_UNLCK operation on any of
372 these duplicate descriptors, or when all such descriptors have been
373 closed.
374 Leases may only be taken out on regular files. An unprivileged process
376 may only take out a lease on a file whose UID (owner) matches the file
377 system UID of the process. A process with the CAP_LEASE capability may
378 take out leases on arbitrary files.
379 F_GETLEASE (void)
381 Indicates what type of lease is associated with the file
382 descriptor fd by returning either F_RDLCK, F_WRLCK, or F_UNLCK,
383 indicating, respectively, a read lease , a write lease, or no
384 lease. arg is ignored.
385 When a process (the "lease breaker") performs an open(2) or truncate(2)
387 that conflicts with a lease established via F_SETLEASE, the system call
388 is blocked by the kernel and the kernel notifies the lease holder by
389 sending it a signal (SIGIO by default). The lease holder should
390 respond to receipt of this signal by doing whatever cleanup is required
391 in preparation for the file to be accessed by another process (e.g.,
392 flushing cached buffers) and then either remove or downgrade its lease.
393 A lease is removed by performing an F_SETLEASE command specifying arg
394 as F_UNLCK. If the lease holder currently holds a write lease on the
395 file, and the lease breaker is opening the file for reading, then it is
396 sufficient for the lease holder to downgrade the lease to a read lease.
397 This is done by performing an F_SETLEASE command specifying arg as
398 F_RDLCK.
399 If the lease holder fails to downgrade or remove the lease within the
401 number of seconds specified in /proc/sys/fs/lease-break-time then the
402 kernel forcibly removes or downgrades the lease holder's lease.
403 Once the lease has been voluntarily or forcibly removed or downgraded,
405 and assuming the lease breaker has not unblocked its system call, the
406 kernel permits the lease breaker's system call to proceed.
407 If the lease breaker's blocked open(2) or truncate(2) is interrupted by
409 a signal handler, then the system call fails with the error EINTR, but
410 the other steps still occur as described above. If the lease breaker
411 is killed by a signal while blocked in open(2) or truncate(2), then the
412 other steps still occur as described above. If the lease breaker spec‐
413 ifies the O_NONBLOCK flag when calling open(2), then the call immedi‐
414 ately fails with the error EWOULDBLOCK, but the other steps still occur
415 as described above.
416 The default signal used to notify the lease holder is SIGIO, but this
418 can be changed using the F_SETSIG command to fcntl(). If a F_SETSIG
419 command is performed (even one specifying SIGIO), and the signal han‐
420 dler is established using SA_SIGINFO, then the handler will receive a
421 siginfo_t structure as its second argument, and the si_fd field of this
422 argument will hold the descriptor of the leased file that has been
423 accessed by another process. (This is useful if the caller holds
424 leases against multiple files).
425 File and directory change notification (dnotify)
427 F_NOTIFY (long)
428 (Linux 2.4 onwards) Provide notification when the directory
429 referred to by fd or any of the files that it contains is
430 changed. The events to be notified are specified in arg, which
431 is a bit mask specified by ORing together zero or more of the
432 following bits:
433 DN_ACCESS A file was accessed (read, pread, readv)
435 DN_MODIFY A file was modified (write, pwrite, writev, trun‐
436 cate, ftruncate).
437 DN_CREATE A file was created (open, creat, mknod, mkdir, link,
438 symlink, rename).
439 DN_DELETE A file was unlinked (unlink, rename to another
440 directory, rmdir).
441 DN_RENAME A file was renamed within this directory (rename).
442 DN_ATTRIB The attributes of a file were changed (chown, chmod,
443 utime[s]).
444 (In order to obtain these definitions, the _GNU_SOURCE feature
446 test macro must be defined.)
447 Directory notifications are normally "one-shot", and the appli‐
449 cation must reregister to receive further notifications. Alter‐
450 natively, if DN_MULTISHOT is included in arg, then notification
451 will remain in effect until explicitly removed.
452 A series of F_NOTIFY requests is cumulative, with the events in
454 arg being added to the set already monitored. To disable noti‐
455 fication of all events, make an F_NOTIFY call specifying arg as
456 0.
457 Notification occurs via delivery of a signal. The default sig‐
459 nal is SIGIO, but this can be changed using the F_SETSIG command
460 to fcntl(). In the latter case, the signal handler receives a
461 siginfo_t structure as its second argument (if the handler was
462 established using SA_SIGINFO) and the si_fd field of this struc‐
463 ture contains the file descriptor which generated the notifica‐
464 tion (useful when establishing notification on multiple directo‐
465 ries).
466 Especially when using DN_MULTISHOT, a real time signal should be
468 used for notification, so that multiple notifications can be
469 queued.
470 NOTE: New applications should use the inotify interface (avail‐
472 able since kernel 2.6.13), which provides a much superior inter‐
473 face for obtaining notifications of file system events. See
474 inotify(7).
475 Changing the capacity of a pipe
477 F_SETPIPE_SZ (long; since Linux 2.6.35)
478 Change the capacity of the pipe referred to by fd to be at least
479 arg bytes. An unprivileged process can adjust the pipe capacity
480 to any value between the system page size and the limit defined
481 in /proc/sys/fs/pipe-size-max (see proc(5)). Attempts to set
482 the pipe capacity below the page size are silently rounded up to
483 the page size. Attempts by an unprivileged process to set the
484 pipe capacity above the limit in /proc/sys/fs/pipe-size-max
485 yield the error EPERM; a privileged process (CAP_SYS_RESOURCE)
486 can override the limit. When allocating the buffer for the
487 pipe, the kernel may use a capacity larger than arg, if that is
488 convenient for the implementation. The F_GETPIPE_SZ operation
489 returns the actual size used. Attempting to set the pipe capac‐
490 ity smaller than the amount of buffer space currently used to
491 store data produces the error EBUSY.
492 F_GETPIPE_SZ (void; since Linux 2.6.35)
494 Return (as the function result) the capacity of the pipe
495 referred to by fd.
496
498 For a successful call, the return value depends on the operation:
499 F_DUPFD The new descriptor.
501 F_GETFD Value of flags.
503 F_GETFL Value of flags.
505 F_GETLEASE
507 Type of lease held on file descriptor.
508 F_GETOWN Value of descriptor owner.
510 F_GETSIG Value of signal sent when read or write becomes possible, or
512 zero for traditional SIGIO behavior.
513 F_GETPIPE_SZ
515 The pipe capacity.
516 All other commands
518 Zero.
519 On error, -1 is returned, and errno is set appropriately.
521
523 EACCES or EAGAIN
524 Operation is prohibited by locks held by other processes.
525 EAGAIN The operation is prohibited because the file has been memory-
527 mapped by another process.
528 EBADF fd is not an open file descriptor, or the command was F_SETLK or
530 F_SETLKW and the file descriptor open mode doesn't match with
531 the type of lock requested.
532 EDEADLK
534 It was detected that the specified F_SETLKW command would cause
535 a deadlock.
536 EFAULT lock is outside your accessible address space.
538 EINTR For F_SETLKW, the command was interrupted by a signal; see sig‐
540 nal(7). For F_GETLK and F_SETLK, the command was interrupted by
541 a signal before the lock was checked or acquired. Most likely
542 when locking a remote file (e.g., locking over NFS), but can
543 sometimes happen locally.
544 EINVAL For F_DUPFD, arg is negative or is greater than the maximum
546 allowable value. For F_SETSIG, arg is not an allowable signal
547 number.
548 EMFILE For F_DUPFD, the process already has the maximum number of file
550 descriptors open.
551 ENOLCK Too many segment locks open, lock table is full, or a remote
553 locking protocol failed (e.g., locking over NFS).
554 EPERM Attempted to clear the O_APPEND flag on a file that has the
556 append-only attribute set.
557
559 SVr4, 4.3BSD, POSIX.1-2001. Only the operations F_DUPFD, F_GETFD,
560 F_SETFD, F_GETFL, F_SETFL, F_GETLK, F_SETLK, F_SETLKW, F_GETOWN, and
561 F_SETOWN are specified in POSIX.1-2001.
562 F_DUPFD_CLOEXEC is specified in POSIX.1-2008.
564 F_GETOWN_EX, F_SETOWN_EX, F_SETPIPE_SZ, F_GETPIPE_SZ, F_GETSIG, F_SET‐
566 SIG, F_NOTIFY, F_GETLEASE, and F_SETLEASE are Linux-specific. (Define
567 the _GNU_SOURCE macro to obtain these definitions.)
568
570 The errors returned by dup2(2) are different from those returned by
571 F_DUPFD.
572 Since kernel 2.0, there is no interaction between the types of lock
574 placed by flock(2) and fcntl().
575 Several systems have more fields in struct flock such as, for example,
577 l_sysid. Clearly, l_pid alone is not going to be very useful if the
578 process holding the lock may live on a different machine.
579
581 A limitation of the Linux system call conventions on some architectures
582 (notably i386) means that if a (negative) process group ID to be
583 returned by F_GETOWN falls in the range -1 to -4095, then the return
584 value is wrongly interpreted by glibc as an error in the system call;
585 that is, the return value of fcntl() will be -1, and errno will contain
586 the (positive) process group ID. The Linux-specific F_GETOWN_EX opera‐
587 tion avoids this problem. Since glibc version 2.11, glibc makes the
588 kernel F_GETOWN problem invisible by implementing F_GETOWN using
589 F_GETOWN_EX.
590 In Linux 2.4 and earlier, there is bug that can occur when an unprivi‐
592 leged process uses F_SETOWN to specify the owner of a socket file
593 descriptor as a process (group) other than the caller. In this case,
594 fcntl() can return -1 with errno set to EPERM, even when the owner
595 process (group) is one that the caller has permission to send signals
596 to. Despite this error return, the file descriptor owner is set, and
597 signals will be sent to the owner.
598 The implementation of mandatory locking in all known versions of Linux
600 is subject to race conditions which render it unreliable: a write(2)
601 call that overlaps with a lock may modify data after the mandatory lock
602 is acquired; a read(2) call that overlaps with a lock may detect
603 changes to data that were made only after a write lock was acquired.
604 Similar races exist between mandatory locks and mmap(2). It is there‐
605 fore inadvisable to rely on mandatory locking.
606
608 dup2(2), flock(2), open(2), socket(2), lockf(3), capabilities(7), fea‐
609 ture_test_macros(7)
610 See also locks.txt, mandatory-locking.txt, and dnotify.txt in the ker‐
612 nel source directory Documentation/filesystems/. (On older kernels,
613 these files are directly under the Documentation/ directory, and manda‐
614 tory-locking.txt is called mandatory.txt.)
615
617 This page is part of release 3.25 of the Linux man-pages project. A
618 description of the project, and information about reporting bugs, can
619 be found at http://www.kernel.org/doc/man-pages/.
620Linux 2010-06-19 FCNTL(2)