1 .\" Hey Emacs! This file is -*- nroff -*- source.
3 .\" This manpage is Copyright (C) 1992 Drew Eckhardt;
4 .\" 1993 Michael Haardt, Ian Jackson.
7 .\" Permission is granted to make and distribute verbatim copies of this
8 .\" manual provided the copyright notice and this permission notice are
9 .\" preserved on all copies.
11 .\" Permission is granted to copy and distribute modified versions of this
12 .\" manual under the conditions for verbatim copying, provided that the
13 .\" entire resulting derived work is distributed under the terms of a
14 .\" permission notice identical to this one.
16 .\" Since the Linux kernel and libraries are constantly changing, this
17 .\" manual page may be incorrect or out-of-date. The author(s) assume no
18 .\" responsibility for errors or omissions, or for damages resulting from
19 .\" the use of the information contained herein. The author(s) may not
20 .\" have taken the same level of care in the production of this manual,
21 .\" which is licensed free of charge, as they might when working
24 .\" Formatted or processed versions of this manual, if unaccompanied by
25 .\" the source, must acknowledge the copyright and authors of this work.
27 .\" Modified 1993-07-21 by Rik Faith <faith@cs.unc.edu>
28 .\" Modified 1994-08-21 by Michael Haardt
29 .\" Modified 1996-04-13 by Andries Brouwer <aeb@cwi.nl>
30 .\" Modified 1996-05-13 by Thomas Koenig
31 .\" Modified 1996-12-20 by Michael Haardt
32 .\" Modified 1999-02-19 by Andries Brouwer <aeb@cwi.nl>
33 .\" Modified 1998-11-28 by Joseph S. Myers <jsm28@hermes.cam.ac.uk>
34 .\" Modified 1999-06-03 by Michael Haardt
35 .\" Modified 2002-05-07 by Michael Kerrisk <mtk.manpages@gmail.com>
36 .\" Modified 2004-06-23 by Michael Kerrisk <mtk.manpages@gmail.com>
37 .\" 2004-12-08, mtk, reordered flags list alphabetically
38 .\" 2004-12-08, Martin Pool <mbp@sourcefrog.net> (& mtk), added O_NOATIME
39 .\" 2007-09-18, mtk, Added description of O_CLOEXEC + other minor edits
40 .\" 2008-01-03, mtk, with input from Trond Myklebust
41 .\" <trond.myklebust@fys.uio.no> and Timo Sirainen <tss@iki.fi>
42 .\" Rewrite description of O_EXCL.
43 .\" 2008-01-11, Greg Banks <gnb@melbourne.sgi.com>: add more detail
45 .\" 2008-02-26, Michael Haardt: Reorganized text for O_CREAT and mode
47 .\" FIXME . Apr 08: The next POSIX revision has O_EXEC, O_SEARCH, and
48 .\" O_TTYINIT. Eventually these may need to be documented. --mtk
49 .\" FIXME Linux 2.6.33 has O_DSYNC, and a hiddden __O_SYNC.
51 .TH OPEN 2 2009-09-20 "Linux" "Linux Programmer's Manual"
53 open, creat \- open and possibly create a file or device
56 .B #include <sys/types.h>
57 .B #include <sys/stat.h>
60 .BI "int open(const char *" pathname ", int " flags );
61 .BI "int open(const char *" pathname ", int " flags ", mode_t " mode );
63 .BI "int creat(const char *" pathname ", mode_t " mode );
70 returns a file descriptor, a small, nonnegative integer
71 for use in subsequent system calls
72 .RB ( read "(2), " write "(2), " lseek "(2), " fcntl "(2), etc.)."
73 The file descriptor returned by a successful call will be
74 the lowest-numbered file descriptor not currently open for the process.
76 By default, the new file descriptor is set to remain open across an
80 file descriptor flag described in
82 is initially disabled; the Linux-specific
84 flag, described below, can be used to change this default).
85 The file offset is set to the beginning of the file (see
91 .IR "open file description" ,
92 an entry in the system-wide table of open files.
93 This entry records the file offset and the file status flags
98 A file descriptor is a reference to one of these entries;
99 this reference is unaffected if
101 is subsequently removed or modified to refer to a different file.
102 The new open file description is initially not shared
103 with any other process,
104 but sharing may arise via
109 must include one of the following
111 .BR O_RDONLY ", " O_WRONLY ", or " O_RDWR .
112 These request opening the file read-only, write-only, or read/write,
115 In addition, zero or more file creation flags and file status flags
121 .I file creation flags
123 .BR O_CREAT ", " O_EXCL ", " O_NOCTTY ", and " O_TRUNC .
126 are all of the remaining flags listed below.
127 .\" FIXME . Actually is it true that the "file status flags" are all of the
128 .\" remaining flags listed below? SUSv4 divides the flags into:
132 .\" * Other (O_CLOEXEC, O_DIRECTORY, O_NOFOLLOW)
133 .\" though it's not clear what the difference between "other" and
134 .\" "File creation" flags is. (I've raised an Aardvark to see if this
135 .\" can be clarified in SUSv4; 10 Oct 2008.)
136 The distinction between these two groups of flags is that
137 the file status flags can be retrieved and (in some cases)
140 The full list of file creation flags and file status flags is as follows:
143 The file is opened in append mode.
146 the file offset is positioned at the end of the file,
150 may lead to corrupted files on NFS file systems if more than one process
151 appends data to a file at once.
152 .\" For more background, see
153 .\" http://bugs.debian.org/cgi-bin/bugreport.cgi?bug=453946
154 .\" http://nfs.sourceforge.net/
155 This is because NFS does not support
156 appending to a file, so the client kernel has to simulate it, which
157 can't be done without a race condition.
160 Enable signal-driven I/O:
163 by default, but this can be changed via
165 when input or output becomes possible on this file descriptor.
166 This feature is only available for terminals, pseudo-terminals,
167 sockets, and (since Linux 2.6) pipes and FIFOs.
172 .BR O_CLOEXEC " (Since Linux 2.6.23)"
173 Enable the close-on-exec flag for the new file descriptor.
174 Specifying this flag permits a program to avoid additional
177 operations to set the
181 use of this flag is essential in some multithreaded programs
182 since using a separate
187 flag does not suffice to avoid race conditions
188 where one thread opens a file descriptor at the same
189 time as another thread does a
193 .\" This flag fixes only one form of the race condition;
194 .\" The race can also occur with, for example, descriptors
195 .\" returned by accept(), pipe(), etc.
198 If the file does not exist it will be created.
199 The owner (user ID) of the file is set to the effective user ID
201 The group ownership (group ID) is set either to
202 the effective group ID of the process or to the group ID of the
203 parent directory (depending on file system type and mount options,
204 and the mode of the parent directory, see the mount options
210 .\" As at 2.6.25, bsdgroups is supported by ext2, ext3, ext4, and
211 .\" XFS (since 2.6.14).
215 specifies the permissions to use in case a new file is created.
216 This argument must be supplied when
222 is not specified, then
225 The effective permissions are modified by
228 in the usual way: The permissions of the created file are
229 .IR "(mode\ &\ ~umask)" .
230 Note that this mode only applies to future accesses of the
231 newly created file; the
233 call that creates a read-only file may well return a read/write
236 The following symbolic constants are provided for
240 00700 user (file owner) has read, write and execute permission
243 00400 user has read permission
246 00200 user has write permission
249 00100 user has execute permission
252 00070 group has read, write and execute permission
255 00040 group has read permission
258 00020 group has write permission
261 00010 group has execute permission
264 00007 others have read, write and execute permission
267 00004 others have read permission
270 00002 others have write permission
273 00001 others have execute permission
276 .BR O_DIRECT " (Since Linux 2.4.10)"
277 Try to minimize cache effects of the I/O to and from this file.
278 In general this will degrade performance, but it is useful in
279 special situations, such as when applications do their own caching.
280 File I/O is done directly to/from user space buffers.
283 flag on its own makes at an effort to transfer data synchronously,
284 but does not give the guarantees of the
286 that data and necessary metadata are transferred.
287 To guarantee synchronous I/O the
289 must be used in addition to
293 below for further discussion.
295 A semantically similar (but deprecated) interface for block devices
300 If \fIpathname\fP is not a directory, cause the open to fail.
301 .\" But see the following and its replies:
302 .\" http://marc.theaimsgroup.com/?t=112748702800001&r=1&w=2
303 .\" [PATCH] open: O_DIRECTORY and O_CREAT together should fail
304 .\" O_DIRECTORY | O_CREAT causes O_DIRECTORY to be ignored.
305 This flag is Linux-specific, and was added in kernel version 2.1.126, to
306 avoid denial-of-service problems if
309 FIFO or tape device, but should not be used outside of the
314 Ensure that this call creates the file:
315 if this flag is specified in conjunction with
328 When these two flags are specified, symbolic links are not followed:
329 .\" POSIX.1-2001 explicitly requires this behavior.
332 is a symbolic link, then
334 fails regardless of where the symbolic link points to.
337 is only supported on NFS when using NFSv3 or later on kernel 2.6 or later.
338 In environments where NFS
340 support is not provided, programs that rely on it
341 for performing locking tasks will contain a race condition.
342 Portable programs that want to perform atomic file locking using a lockfile,
343 and need to avoid reliance on NFS support for
345 can create a unique file on
346 the same file system (e.g., incorporating hostname and PID), and use
348 to make a link to the lockfile.
351 returns 0, the lock is successful.
354 on the unique file to check if its link count has increased to 2,
355 in which case the lock is also successful.
359 Allow files whose sizes cannot be represented in an
361 (but can be represented in an
365 .B _LARGEFILE64_SOURCE
366 macro must be defined in order to obtain this definition.
369 feature test macro to 64 (rather than using
371 is the preferred method of obtaining
372 method of accessing large files on 32-bit systems (see
373 .BR feature_test_macros (7)).
375 .BR O_NOATIME " (Since Linux 2.6.8)"
376 Do not update the file last access time (st_atime in the inode)
379 This flag is intended for use by indexing or backup programs,
380 where its use can significantly reduce the amount of disk activity.
381 This flag may not be effective on all file systems.
382 One example is NFS, where the server maintains the access time.
383 .\" The O_NOATIME flag also affects the treatment of st_atime
384 .\" by mmap() and readdir(2), MTK, Dec 04.
389 refers to a terminal device \(em see
391 \(em it will not become the process's controlling terminal even if the
392 process does not have one.
395 If \fIpathname\fP is a symbolic link, then the open fails.
396 This is a FreeBSD extension, which was added to Linux in version 2.1.126.
397 Symbolic links in earlier components of the pathname will still be
399 .\" The headers from glibc 2.0.100 and later include a
400 .\" definition of this flag; \fIkernels before 2.1.126 will ignore it if
403 .BR O_NONBLOCK " or " O_NDELAY
404 When possible, the file is opened in nonblocking mode.
407 nor any subsequent operations on the file descriptor which is
408 returned will cause the calling process to wait.
409 For the handling of FIFOs (named pipes), see also
411 For a discussion of the effect of
413 in conjunction with mandatory file locks and with file leases, see
417 The file is opened for synchronous I/O.
420 on the resulting file descriptor will block the calling process until
421 the data has been physically written to the underlying hardware.
422 .IR "But see NOTES below" .
425 If the file already exists and is a regular file and the open mode allows
430 it will be truncated to length 0.
431 If the file is a FIFO or terminal device file, the
434 Otherwise the effect of
438 Some of these optional flags can be altered using
440 after the file has been opened.
448 .BR O_CREAT|O_WRONLY|O_TRUNC .
453 return the new file descriptor, or \-1 if an error occurred
456 is set appropriately).
460 The requested access to the file is not allowed, or search permission
461 is denied for one of the directories in the path prefix of
463 or the file did not exist yet and write access to the parent directory
466 .BR path_resolution (7).)
471 .BR O_CREAT " and " O_EXCL
476 points outside your accessible address space.
483 While blocked waiting to complete an open of a slow device
486 the call was interrupted by a signal handler; see
491 refers to a directory and the access requested involved writing
499 Too many symbolic links were encountered in resolving
501 or \fBO_NOFOLLOW\fP was specified but
506 The process already has the maximum number of files open.
513 The system limit on the total number of open files has been reached.
517 refers to a device special file and no corresponding device exists.
518 (This is a Linux kernel bug; in this situation
524 is not set and the named file does not exist.
525 Or, a directory component in
527 does not exist or is a dangling symbolic link.
530 Insufficient kernel memory was available.
534 was to be created but the device containing
536 has no room for the new file.
539 A component used as a directory in
541 is not, in fact, a directory, or \fBO_DIRECTORY\fP was specified and
546 .BR O_NONBLOCK " | " O_WRONLY
547 is set, the named file is a FIFO and
548 no process has the file open for reading.
549 Or, the file is a device special file and no corresponding device exists.
553 refers to a regular file that is too large to be opened.
554 The usual scenario here is that an application compiled
555 on a 32-bit platform without
556 .I -D_FILE_OFFSET_BITS=64
557 tried to open a file whose size exceeds
563 This is the error specified by POSIX.1-2001;
564 in kernels before 2.6.24, Linux gave the error
567 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=7253
568 .\" "Open of a large file on 32-bit fails with EFBIG, should be EOVERFLOW"
569 .\" Reported 2006-10-03
574 flag was specified, but the effective user ID of the caller
575 .\" Strictly speaking, it's the file system UID... (MTK)
576 did not match the owner of the file and the caller was not privileged
581 refers to a file on a read-only file system and write access was
586 refers to an executable image which is currently being executed and
587 write access was requested.
592 flag was specified, and an incompatible lease was held on the file
596 SVr4, 4.3BSD, POSIX.1-2001.
602 flags are Linux-specific, and one may need to define
604 to obtain their definitions.
608 flag is not specified in POSIX.1-2001,
609 but is specified in POSIX.1-2008.
612 is not specified in POSIX; one has to define
614 to get its definition.
618 flag indicates that one wants to open
619 but does not necessarily have the intention to read or write.
620 This is typically used to open devices in order to get a file descriptor
624 Unlike the other values that can be specified in
629 .BR O_RDONLY ", " O_WRONLY ", and " O_RDWR ,
630 do not specify individual bits.
631 Rather, they define the low order two bits of
633 and are defined respectively as 0, 1, and 2.
634 In other words, the combination
635 .B "O_RDONLY | O_WRONLY"
636 is a logical error, and certainly does not have the same meaning as
638 Linux reserves the special, nonstandard access mode 3 (binary 11) in
641 check for read and write permission on the file and return a descriptor
642 that can't be used for reading or writing.
643 This nonstandard access mode is used by some Linux drivers to return a
644 descriptor that is only to be used for device-specific
647 .\" See for example util-linux's disk-utils/setfdprm.c
648 .\" For some background on access mode 3, see
649 .\" http://thread.gmane.org/gmane.linux.kernel/653123
650 .\" "[RFC] correct flags to f_mode conversion in __dentry_open"
651 .\" LKML, 12 Mar 2008
653 The (undefined) effect of
654 .B O_RDONLY | O_TRUNC
655 varies among implementations.
656 On many systems the file is actually truncated.
657 .\" Linux 2.0, 2.5: truncate
658 .\" Solaris 5.7, 5.8: truncate
659 .\" Irix 6.5: truncate
660 .\" Tru64 5.1B: truncate
661 .\" HP-UX 11.22: truncate
662 .\" FreeBSD 4.7: truncate
664 There are many infelicities in the protocol underlying NFS, affecting
666 .BR O_SYNC " and " O_NDELAY .
668 POSIX provides for three different variants of synchronized I/O,
669 corresponding to the flags
674 Currently (2.6.31), Linux only implements
680 to the same numerical value as
682 Most Linux filesystems don't actually implement the POSIX
684 semantics, which require all metadata updates of a write
685 to be on disk on returning to userspace, but only the
687 semantics, which require only actual file data and metadata necessary
688 to retrieve it to be on disk by the time the system call returns.
692 can open device special files, but
694 cannot create them; use
698 On NFS file systems with UID mapping enabled,
701 return a file descriptor but, for example,
705 This is because the client performs
708 permissions, but UID mapping is performed by the server upon
709 read and write requests.
711 If the file is newly created, its
716 (respectively, time of last access, time of last status change, and
717 time of last modification; see
720 to the current time, and so are the
726 Otherwise, if the file is modified because of the
728 flag, its st_ctime and st_mtime fields are set to the current time.
733 flag may impose alignment restrictions on the length and address
734 of userspace buffers and the file offset of I/Os.
736 restrictions vary by file system and kernel version and might be
738 However there is currently no file system\-independent
739 interface for an application to discover these restrictions for a given
741 Some file systems provide their own interfaces
742 for doing so, for example the
747 Under Linux 2.4, transfer sizes, and the alignment of the user buffer
748 and the file offset must all be multiples of the logical block size
750 Under Linux 2.6, alignment to 512-byte boundaries
755 flag was introduced in SGI IRIX, where it has alignment
756 restrictions similar to those of Linux 2.4.
759 call to query appropriate alignments, and sizes.
760 FreeBSD 4.x introduced
761 a flag of the same name, but without alignment restrictions.
764 support was added under Linux in kernel version 2.4.10.
765 Older Linux kernels simply ignore this flag.
766 Some file systems may not implement the flag and
772 Applications should avoid mixing
774 and normal I/O to the same file,
775 and especially to overlapping byte regions in the same file.
776 Even when the file system correctly handles the coherency issues in
777 this situation, overall I/O throughput is likely to be slower than
778 using either mode alone.
779 Likewise, applications should avoid mixing
781 of files with direct I/O to the same files.
785 with NFS will differ from local file systems.
787 kernels configured in certain ways, may not support this combination.
788 The NFS protocol does not support passing the flag to the server, so
790 I/O will only bypass the page cache on the client; the server may
792 The client asks the server to make the I/O
793 synchronous to preserve the synchronous semantics of
795 Some servers will perform poorly under these circumstances, especially
796 if the I/O size is small.
797 Some servers may also be configured to
798 lie to clients about the I/O having reached stable storage; this
799 will avoid the performance penalty at some risk to data integrity
800 in the event of server power failure.
801 The Linux NFS client places no alignment restrictions on
807 is a potentially powerful tool that should be used with caution.
808 It is recommended that applications treat use of
810 as a performance option which is disabled by default.
813 "The thing that has always disturbed me about O_DIRECT is that the whole
814 interface is just stupid, and was probably designed by a deranged monkey
815 on some serious mind-controlling substances." \(em Linus
818 Currently, it is not possible to enable signal-driven
826 .\" FIXME . Check bugzilla report on open(O_ASYNC)
827 .\" See http://bugzilla.kernel.org/show_bug.cgi?id=5993
847 .BR feature_test_macros (7),
849 .BR path_resolution (7),