1OPEN(2) Linux Programmer's Manual OPEN(2)
2
6 open, creat - open and possibly create a file or device
7
9 #include <sys/types.h>
10 #include <sys/stat.h>
11 #include <fcntl.h>
12 int open(const char *pathname, int flags);
14 int open(const char *pathname, int flags, mode_t mode);
15 int creat(const char *pathname, mode_t mode);
17
19 Given a pathname for a file, open() returns a file descriptor, a small,
20 nonnegative integer for use in subsequent system calls (read(2),
21 write(2), lseek(2), fcntl(2), etc.). The file descriptor returned by a
22 successful call will be the lowest-numbered file descriptor not cur‐
23 rently open for the process.
24 By default, the new file descriptor is set to remain open across an
26 execve(2) (i.e., the FD_CLOEXEC file descriptor flag described in
27 fcntl(2) is initially disabled; the Linux-specific O_CLOEXEC flag,
28 described below, can be used to change this default). The file offset
29 is set to the beginning of the file (see lseek(2)).
30 A call to open() creates a new open file description, an entry in the
32 system-wide table of open files. This entry records the file offset
33 and the file status flags (modifiable via the fcntl(2) F_SETFL opera‐
34 tion). A file descriptor is a reference to one of these entries; this
35 reference is unaffected if pathname is subsequently removed or modified
36 to refer to a different file. The new open file description is ini‐
37 tially not shared with any other process, but sharing may arise via
38 fork(2).
39 The argument flags must include one of the following access modes:
41 O_RDONLY, O_WRONLY, or O_RDWR. These request opening the file read-
42 only, write-only, or read/write, respectively.
43 In addition, zero or more file creation flags and file status flags can
45 be bitwise-or'd in flags. The file creation flags are O_CREAT, O_EXCL,
46 O_NOCTTY, and O_TRUNC. The file status flags are all of the remaining
47 flags listed below. The distinction between these two groups of flags
48 is that the file status flags can be retrieved and (in some cases) mod‐
49 ified using fcntl(2). The full list of file creation flags and file
50 status flags is as follows:
51 O_APPEND
53 The file is opened in append mode. Before each write(2), the
54 file offset is positioned at the end of the file, as if with
55 lseek(2). O_APPEND may lead to corrupted files on NFS file sys‐
56 tems if more than one process appends data to a file at once.
57 This is because NFS does not support appending to a file, so the
58 client kernel has to simulate it, which can't be done without a
59 race condition.
60 O_ASYNC
62 Enable signal-driven I/O: generate a signal (SIGIO by default,
63 but this can be changed via fcntl(2)) when input or output
64 becomes possible on this file descriptor. This feature is only
65 available for terminals, pseudo-terminals, sockets, and (since
66 Linux 2.6) pipes and FIFOs. See fcntl(2) for further details.
67 O_CLOEXEC (Since Linux 2.6.23)
69 Enable the close-on-exec flag for the new file descriptor.
70 Specifying this flag permits a program to avoid additional
71 fcntl(2) F_SETFD operations to set the FD_CLOEXEC flag. Addi‐
72 tionally, use of this flag is essential in some multithreaded
73 programs since using a separate fcntl(2) F_SETFD operation to
74 set the FD_CLOEXEC flag does not suffice to avoid race condi‐
75 tions where one thread opens a file descriptor at the same time
76 as another thread does a fork(2) plus execve(2).
77 O_CREAT
79 If the file does not exist it will be created. The owner (user
80 ID) of the file is set to the effective user ID of the process.
81 The group ownership (group ID) is set either to the effective
82 group ID of the process or to the group ID of the parent direc‐
83 tory (depending on file system type and mount options, and the
84 mode of the parent directory, see the mount options bsdgroups
85 and sysvgroups described in mount(8)).
86 mode specifies the permissions to use in case a new file is cre‐
88 ated. This argument must be supplied when O_CREAT is specified
89 in flags; if O_CREAT is not specified, then mode is ignored.
90 The effective permissions are modified by the process's umask in
91 the usual way: The permissions of the created file are
92 (mode & ~umask). Note that this mode only applies to future
93 accesses of the newly created file; the open() call that creates
94 a read-only file may well return a read/write file descriptor.
95 The following symbolic constants are provided for mode:
97 S_IRWXU 00700 user (file owner) has read, write and execute
99 permission
100 S_IRUSR 00400 user has read permission
102 S_IWUSR 00200 user has write permission
104 S_IXUSR 00100 user has execute permission
106 S_IRWXG 00070 group has read, write and execute permission
108 S_IRGRP 00040 group has read permission
110 S_IWGRP 00020 group has write permission
112 S_IXGRP 00010 group has execute permission
114 S_IRWXO 00007 others have read, write and execute permission
116 S_IROTH 00004 others have read permission
118 S_IWOTH 00002 others have write permission
120 S_IXOTH 00001 others have execute permission
122 O_DIRECT (Since Linux 2.4.10)
124 Try to minimize cache effects of the I/O to and from this file.
125 In general this will degrade performance, but it is useful in
126 special situations, such as when applications do their own
127 caching. File I/O is done directly to/from user space buffers.
128 The O_DIRECT flag on its own makes at an effort to transfer data
129 synchronously, but does not give the guarantees of the O_SYNC
130 that data and necessary metadata are transferred. To guarantee
131 synchronous I/O the O_SYNC must be used in addition to O_DIRECT.
132 See NOTES below for further discussion.
133 A semantically similar (but deprecated) interface for block
135 devices is described in raw(8).
136 O_DIRECTORY
138 If pathname is not a directory, cause the open to fail. This
139 flag is Linux-specific, and was added in kernel version 2.1.126,
140 to avoid denial-of-service problems if opendir(3) is called on a
141 FIFO or tape device, but should not be used outside of the
142 implementation of opendir(3).
143 O_EXCL Ensure that this call creates the file: if this flag is speci‐
145 fied in conjunction with O_CREAT, and pathname already exists,
146 then open() will fail. The behavior of O_EXCL is undefined if
147 O_CREAT is not specified.
148 When these two flags are specified, symbolic links are not fol‐
150 lowed: if pathname is a symbolic link, then open() fails regard‐
151 less of where the symbolic link points to.
152 On NFS, O_EXCL is only supported when using NFSv3 or later on
154 kernel 2.6 or later. In NFS environments where O_EXCL support
155 is not provided, programs that rely on it for performing locking
156 tasks will contain a race condition. Portable programs that
157 want to perform atomic file locking using a lockfile, and need
158 to avoid reliance on NFS support for O_EXCL, can create a unique
159 file on the same file system (e.g., incorporating hostname and
160 PID), and use link(2) to make a link to the lockfile. If
161 link(2) returns 0, the lock is successful. Otherwise, use
162 stat(2) on the unique file to check if its link count has
163 increased to 2, in which case the lock is also successful.
164 O_LARGEFILE
166 (LFS) Allow files whose sizes cannot be represented in an off_t
167 (but can be represented in an off64_t) to be opened. The
168 _LARGEFILE64_SOURCE macro must be defined in order to obtain
169 this definition. Setting the _FILE_OFFSET_BITS feature test
170 macro to 64 (rather than using O_LARGEFILE) is the preferred
171 method of obtaining method of accessing large files on 32-bit
172 systems (see feature_test_macros(7)).
173 O_NOATIME (Since Linux 2.6.8)
175 Do not update the file last access time (st_atime in the inode)
176 when the file is read(2). This flag is intended for use by
177 indexing or backup programs, where its use can significantly
178 reduce the amount of disk activity. This flag may not be effec‐
179 tive on all file systems. One example is NFS, where the server
180 maintains the access time.
181 O_NOCTTY
183 If pathname refers to a terminal device — see tty(4) — it will
184 not become the process's controlling terminal even if the
185 process does not have one.
186 O_NOFOLLOW
188 If pathname is a symbolic link, then the open fails. This is a
189 FreeBSD extension, which was added to Linux in version 2.1.126.
190 Symbolic links in earlier components of the pathname will still
191 be followed.
192 O_NONBLOCK or O_NDELAY
194 When possible, the file is opened in nonblocking mode. Neither
195 the open() nor any subsequent operations on the file descriptor
196 which is returned will cause the calling process to wait. For
197 the handling of FIFOs (named pipes), see also fifo(7). For a
198 discussion of the effect of O_NONBLOCK in conjunction with
199 mandatory file locks and with file leases, see fcntl(2).
200 O_SYNC The file is opened for synchronous I/O. Any write(2)s on the
202 resulting file descriptor will block the calling process until
203 the data has been physically written to the underlying hardware.
204 But see NOTES below.
205 O_TRUNC
207 If the file already exists and is a regular file and the open
208 mode allows writing (i.e., is O_RDWR or O_WRONLY) it will be
209 truncated to length 0. If the file is a FIFO or terminal device
210 file, the O_TRUNC flag is ignored. Otherwise the effect of
211 O_TRUNC is unspecified.
212 Some of these optional flags can be altered using fcntl(2) after the
214 file has been opened.
215 creat() is equivalent to open() with flags equal to
217 O_CREAT|O_WRONLY|O_TRUNC.
218
220 open() and creat() return the new file descriptor, or -1 if an error
221 occurred (in which case, errno is set appropriately).
222
224 EACCES The requested access to the file is not allowed, or search per‐
225 mission is denied for one of the directories in the path prefix
226 of pathname, or the file did not exist yet and write access to
227 the parent directory is not allowed. (See also path_resolu‐
228 tion(7).)
229 EEXIST pathname already exists and O_CREAT and O_EXCL were used.
231 EFAULT pathname points outside your accessible address space.
233 EFBIG See EOVERFLOW.
235 EINTR While blocked waiting to complete an open of a slow device
237 (e.g., a FIFO; see fifo(7)), the call was interrupted by a sig‐
238 nal handler; see signal(7).
239 EISDIR pathname refers to a directory and the access requested involved
241 writing (that is, O_WRONLY or O_RDWR is set).
242 ELOOP Too many symbolic links were encountered in resolving pathname,
244 or O_NOFOLLOW was specified but pathname was a symbolic link.
245 EMFILE The process already has the maximum number of files open.
247 ENAMETOOLONG
249 pathname was too long.
250 ENFILE The system limit on the total number of open files has been
252 reached.
253 ENODEV pathname refers to a device special file and no corresponding
255 device exists. (This is a Linux kernel bug; in this situation
256 ENXIO must be returned.)
257 ENOENT O_CREAT is not set and the named file does not exist. Or, a
259 directory component in pathname does not exist or is a dangling
260 symbolic link.
261 ENOMEM Insufficient kernel memory was available.
263 ENOSPC pathname was to be created but the device containing pathname
265 has no room for the new file.
266 ENOTDIR
268 A component used as a directory in pathname is not, in fact, a
269 directory, or O_DIRECTORY was specified and pathname was not a
270 directory.
271 ENXIO O_NONBLOCK | O_WRONLY is set, the named file is a FIFO and no
273 process has the file open for reading. Or, the file is a device
274 special file and no corresponding device exists.
275 EOVERFLOW
277 pathname refers to a regular file that is too large to be
278 opened. The usual scenario here is that an application compiled
279 on a 32-bit platform without -D_FILE_OFFSET_BITS=64 tried to
280 open a file whose size exceeds (2<<31)-1 bits; see also O_LARGE‐
281 FILE above. This is the error specified by POSIX.1-2001; in
282 kernels before 2.6.24, Linux gave the error EFBIG for this case.
283 EPERM The O_NOATIME flag was specified, but the effective user ID of
285 the caller did not match the owner of the file and the caller
286 was not privileged (CAP_FOWNER).
287 EROFS pathname refers to a file on a read-only file system and write
289 access was requested.
290 ETXTBSY
292 pathname refers to an executable image which is currently being
293 executed and write access was requested.
294 EWOULDBLOCK
296 The O_NONBLOCK flag was specified, and an incompatible lease was
297 held on the file (see fcntl(2)).
298
300 SVr4, 4.3BSD, POSIX.1-2001. The O_DIRECTORY, O_NOATIME, and O_NOFOLLOW
301 flags are Linux-specific, and one may need to define _GNU_SOURCE to
302 obtain their definitions.
303 The O_CLOEXEC flag is not specified in POSIX.1-2001, but is specified
305 in POSIX.1-2008.
306 O_DIRECT is not specified in POSIX; one has to define _GNU_SOURCE to
308 get its definition.
309
311 Under Linux, the O_NONBLOCK flag indicates that one wants to open but
312 does not necessarily have the intention to read or write. This is typ‐
313 ically used to open devices in order to get a file descriptor for use
314 with ioctl(2).
315 Unlike the other values that can be specified in flags, the access mode
317 values O_RDONLY, O_WRONLY, and O_RDWR, do not specify individual bits.
318 Rather, they define the low order two bits of flags, and are defined
319 respectively as 0, 1, and 2. In other words, the combination O_RDONLY
320 | O_WRONLY is a logical error, and certainly does not have the same
321 meaning as O_RDWR. Linux reserves the special, nonstandard access mode
322 3 (binary 11) in flags to mean: check for read and write permission on
323 the file and return a descriptor that can't be used for reading or
324 writing. This nonstandard access mode is used by some Linux drivers to
325 return a descriptor that is only to be used for device-specific
326 ioctl(2) operations.
327 The (undefined) effect of O_RDONLY | O_TRUNC varies among implementa‐
329 tions. On many systems the file is actually truncated.
330 There are many infelicities in the protocol underlying NFS, affecting
332 amongst others O_SYNC and O_NDELAY.
333 POSIX provides for three different variants of synchronized I/O, corre‐
335 sponding to the flags O_SYNC, O_DSYNC, and O_RSYNC. Currently
336 (2.6.31), Linux only implements O_SYNC, but glibc maps O_DSYNC and
337 O_RSYNC to the same numerical value as O_SYNC Most Linux file systems
338 don't actually implement the POSIX O_SYNC semantics, which require all
339 metadata updates of a write to be on disk on returning to userspace,
340 but only the O_DSYNC semantics, which require only actual file data and
341 metadata necessary to retrieve it to be on disk by the time the system
342 call returns.
343 Note that open() can open device special files, but creat() cannot cre‐
345 ate them; use mknod(2) instead.
346 On NFS file systems with UID mapping enabled, open() may return a file
348 descriptor but, for example, read(2) requests are denied with EACCES.
349 This is because the client performs open() by checking the permissions,
350 but UID mapping is performed by the server upon read and write
351 requests.
352 If the file is newly created, its st_atime, st_ctime, st_mtime fields
354 (respectively, time of last access, time of last status change, and
355 time of last modification; see stat(2)) are set to the current time,
356 and so are the st_ctime and st_mtime fields of the parent directory.
357 Otherwise, if the file is modified because of the O_TRUNC flag, its
358 st_ctime and st_mtime fields are set to the current time.
359 O_DIRECT
361 The O_DIRECT flag may impose alignment restrictions on the length and
362 address of userspace buffers and the file offset of I/Os. In Linux
363 alignment restrictions vary by file system and kernel version and might
364 be absent entirely. However there is currently no file system-indepen‐
365 dent interface for an application to discover these restrictions for a
366 given file or file system. Some file systems provide their own inter‐
367 faces for doing so, for example the XFS_IOC_DIOINFO operation in
368 xfsctl(3).
369 Under Linux 2.4, transfer sizes, and the alignment of the user buffer
371 and the file offset must all be multiples of the logical block size of
372 the file system. Under Linux 2.6, alignment to 512-byte boundaries
373 suffices.
374 The O_DIRECT flag was introduced in SGI IRIX, where it has alignment
376 restrictions similar to those of Linux 2.4. IRIX has also a fcntl(2)
377 call to query appropriate alignments, and sizes. FreeBSD 4.x intro‐
378 duced a flag of the same name, but without alignment restrictions.
379 O_DIRECT support was added under Linux in kernel version 2.4.10. Older
381 Linux kernels simply ignore this flag. Some file systems may not
382 implement the flag and open() will fail with EINVAL if it is used.
383 Applications should avoid mixing O_DIRECT and normal I/O to the same
385 file, and especially to overlapping byte regions in the same file.
386 Even when the file system correctly handles the coherency issues in
387 this situation, overall I/O throughput is likely to be slower than
388 using either mode alone. Likewise, applications should avoid mixing
389 mmap(2) of files with direct I/O to the same files.
390 The behaviour of O_DIRECT with NFS will differ from local file systems.
392 Older kernels, or kernels configured in certain ways, may not support
393 this combination. The NFS protocol does not support passing the flag
394 to the server, so O_DIRECT I/O will only bypass the page cache on the
395 client; the server may still cache the I/O. The client asks the server
396 to make the I/O synchronous to preserve the synchronous semantics of
397 O_DIRECT. Some servers will perform poorly under these circumstances,
398 especially if the I/O size is small. Some servers may also be config‐
399 ured to lie to clients about the I/O having reached stable storage;
400 this will avoid the performance penalty at some risk to data integrity
401 in the event of server power failure. The Linux NFS client places no
402 alignment restrictions on O_DIRECT I/O.
403 In summary, O_DIRECT is a potentially powerful tool that should be used
405 with caution. It is recommended that applications treat use of
406 O_DIRECT as a performance option which is disabled by default.
407 "The thing that has always disturbed me about O_DIRECT is that
409 the whole interface is just stupid, and was probably designed by
410 a deranged monkey on some serious mind-controlling substances."
411 — Linus
412
414 Currently, it is not possible to enable signal-driven I/O by specifying
415 O_ASYNC when calling open(); use fcntl(2) to enable this flag.
416
418 chmod(2), chown(2), close(2), dup(2), fcntl(2), link(2), lseek(2),
419 mknod(2), mmap(2), mount(2), openat(2), read(2), socket(2), stat(2),
420 umask(2), unlink(2), write(2), fopen(3), feature_test_macros(7),
421 fifo(7), path_resolution(7), symlink(7)
422
424 This page is part of release 3.25 of the Linux man-pages project. A
425 description of the project, and information about reporting bugs, can
426 be found at http://www.kernel.org/doc/man-pages/.
427Linux 2010-06-14 OPEN(2)