O_RDWR O_CREAT等open函数标志位在哪里定义？（格式还要编译，答案在最后一段）

man open
OPEN(2)                    Linux Programmer’s Manual                   OPEN(2)

NAME
       open, creat - open and possibly create a file or device

SYNOPSIS
       #include
       #include
       #include

       int open(const char *pathname, int flags);
       int open(const char *pathname, int flags, mode_t mode);

       int creat(const char *pathname, mode_t mode);

DESCRIPTION
       Given  a pathname for a file, open() returns a file descriptor, a small, non-negative integer for use in subse-quent system calls (read(2), write(2), lseek(2), fcntl(2), etc.).  The file descriptor returned by a successful call will be the lowest-numbered file descriptor not currently open for the process.

       By  default,  the  new  file  descriptor  is  set to remain open across an execve(2) (i.e., the FD_CLOEXEC file descriptor flag described in fcntl(2) is initially  disabled;  the  Linux-specific  O_CLOEXEC  flag,  described below,  can  be  used  to  change  this  default).   The  file  offset is set to the beginning of the file (see lseek(2)).

       A call to open() creates a new open file description, an entry in the system-wide table of  open  files.   This entry  records  the  file  offset and the file status flags (modifiable via the fcntl(2) F_SETFL operation).  A
       file descriptor is a reference to one of these entries; this reference is  unaffected  if  pathname  is  subse-
       quently  removed  or  modified  to  refer  to a different file.  The new open file description is initially not
       shared with any other process, but sharing may arise via fork(2).

       The argument flags must include one of the following  access  modes:  O_RDONLY,  O_WRONLY,  or  O_RDWR.   These
       request opening the file read-only, write-only, or read/write, respectively.

       In  addition,  zero  or  more file creation flags and file status flags can be bitwise-or’d in flags.  The file
       creation flags are O_CREAT, O_EXCL, O_NOCTTY, and O_TRUNC.  The file status flags  are  all  of  the  remaining
       flags  listed  below.   The  distinction between these two groups of flags is that the file status flags can be
       retrieved and (in some cases) modified using fcntl(2).  The full list of file creation flags  and  file  status
       flags is as follows:

       O_APPEND
              The  file  is  opened in append mode.  Before each write(2), the file offset is positioned at the end of
              the file, as if with lseek(2).  O_APPEND may lead to corrupted files on NFS file systems  if  more  than
              one  process  appends data to a file at once.  This is because NFS does not support appending to a file,
              so the client kernel has to simulate it, which can’t be done without a race condition.

       O_ASYNC
              Enable signal-driven I/O: generate a signal (SIGIO by default, but this can  be  changed  via  fcntl(2))
              when  input or output becomes possible on this file descriptor.  This feature is only available for ter-
              minals, pseudo-terminals, sockets, and (since Linux 2.6) pipes and  FIFOs.   See  fcntl(2)  for  further
              details.

       O_CLOEXEC (Since Linux 2.6.23)
              Enable  the  close-on-exec  flag for the new file descriptor.  Specifying this flag permits a program to
              avoid an additional fcntl(2) F_SETFD operation to set the FD_CLOEXEC flag.  Additionally,  use  of  this
              flag  is  essential  in some multithreaded programs since using a separate fcntl(2) F_SETFD operation to
              set the FD_CLOEXEC flag does not suffice to avoid race conditions where one thread opens a file descrip-
              tor at the same time as another thread does a fork(2) plus execve(2).

       O_CREAT
              If  the file does not exist it will be created.  The owner (user ID) of the file is set to the effective
              user ID of the process.  The group ownership (group ID) is set either to the effective group ID  of  the
              process or to the group ID of the parent directory (depending on file system type and mount options, and
              the mode of the parent directory, see the mount options bsdgroups and sysvgroups described in mount(8)).

              mode  specifies  the  permissions  to use in case a new file is created.  This argument must be supplied
              when O_CREAT is specified in flags; if O_CREAT is not specified, then mode is  ignored.   The  effective
              permissions  are  modified  by the process’s umask in the usual way: The permissions of the created file
              are (mode & ~umask).  Note that this mode only applies to future accesses of the newly created file; the
              open() call that creates a read-only file may well return a read/write file descriptor.

              The following symbolic constants are provided for mode:

              S_IRWXU  00700 user (file owner) has read, write and execute permission

              S_IRUSR  00400 user has read permission

              S_IWUSR  00200 user has write permission

              S_IXUSR  00100 user has execute permission

              S_IRWXG  00070 group has read, write and execute permission

              S_IRGRP  00040 group has read permission

              S_IWGRP  00020 group has write permission

              S_IXGRP  00010 group has execute permission

              S_IRWXO  00007 others have read, write and execute permission

              S_IROTH  00004 others have read permission

              S_IWOTH  00002 others have write permission

              S_IXOTH  00001 others have execute permission

       O_DIRECT (Since Linux 2.4.10)
              Try  to  minimize  cache effects of the I/O to and from this file.  In general this will degrade perfor-
              mance, but it is useful in special situations, such as when applications do their own caching.  File I/O
              is  done  directly  to/from user space buffers.  The I/O is synchronous, that is, at the completion of a
              read(2) or write(2), data is guaranteed to have been transferred.  See NOTES below for  further  discus-
              sion.

              A semantically similar (but deprecated) interface for block devices is described in raw(8).

       O_DIRECTORY
              If  pathname  is not a directory, cause the open to fail.  This flag is Linux-specific, and was added in
              kernel version 2.1.126, to avoid denial-of-service problems if opendir(3) is called on a  FIFO  or  tape
              device, but should not be used outside of the implementation of opendir(3).

       O_EXCL Ensure that this call creates the file: if this flag is specified in conjunction with O_CREAT, and path-
              name already exists, then open() will fail.  The behavior of O_EXCL is undefined if O_CREAT is not spec-
              ified.

              When  these  two  flags  are specified, symbolic links are not followed: if pathname is a symbolic link,
              then open() fails regardless of where the symbolic link points to.

              O_EXCL is only supported on NFS when using NFSv3 or later on kernel 2.6 or later.  In environments where
              NFS O_EXCL support is not provided, programs that rely on it for performing locking tasks will contain a
              race condition.  Portable programs that want to perform atomic file locking using a lockfile,  and  need
              to  avoid  reliance  on  NFS support for O_EXCL, can create a unique file on the same file system (e.g.,
              incorporating hostname and PID), and use link(2) to make a link to the lockfile.  If link(2) returns  0,
              the  lock  is  successful.   Otherwise,  use  stat(2)  on the unique file to check if its link count has
              increased to 2, in which case the lock is also successful.

       O_LARGEFILE
              (LFS) Allow files whose sizes cannot be represented in an off_t (but can be represented in  an  off64_t)
              to  be  opened.  The _LARGEFILE64_SOURCE macro must be defined in order to obtain this definition.  Set-
              ting the _FILE_OFFSET_BITS feature test macro to 64 (rather than using  O_LARGEFILE)  is  the  preferred
              method of obtaining method of accessing large files on 32-bit systems (see feature_test_macros(7)).

       O_NOATIME (Since Linux 2.6.8)
              Do  not update the file last access time (st_atime in the inode) when the file is read(2).  This flag is
              intended for use by indexing or backup programs, where its use can significantly reduce  the  amount  of
              disk  activity.   This  flag  may  not  be effective on all file systems.  One example is NFS, where the
              server maintains the access time.

       O_NOCTTY
              If pathname refers to a terminal device — see tty(4) — it will not become the process’s controlling ter-
              minal even if the process does not have one.

       O_NOFOLLOW
              If  pathname  is  a symbolic link, then the open fails.  This is a FreeBSD extension, which was added to
              Linux in version 2.1.126.  Symbolic links in earlier components of the pathname will still be  followed.

       O_NONBLOCK or O_NDELAY
              When  possible,  the  file is opened in non-blocking mode.  Neither the open() nor any subsequent opera-
              tions on the file descriptor which is returned will cause the calling process to wait.  For the handling
              of  FIFOs  (named pipes), see also fifo(7).  For a discussion of the effect of O_NONBLOCK in conjunction
              with mandatory file locks and with file leases, see fcntl(2).

       O_SYNC The file is opened for synchronous I/O.  Any write(2)s on the resulting file descriptor will  block  the
              calling  process  until  the data has been physically written to the underlying hardware.  But see NOTES
              below.

       O_TRUNC
              If the file already exists and is a regular file and the open mode allows writing (i.e.,  is  O_RDWR  or
              O_WRONLY)  it will be truncated to length 0.  If the file is a FIFO or terminal device file, the O_TRUNC
              flag is ignored.  Otherwise the effect of O_TRUNC is unspecified.

       Some of these optional flags can be altered using fcntl(2) after the file has been opened.

       creat() is equivalent to open() with flags equal to O_CREAT|O_WRONLY|O_TRUNC.

RETURN VALUE
       open() and creat() return the new file descriptor, or -1 if an error occurred (in  which  case,  errno  is  set
       appropriately).

ERRORS
       EACCES The  requested access to the file is not allowed, or search permission is denied for one of the directo-
              ries in the path prefix of pathname, or the file did not exist yet and write access to the parent direc-
              tory is not allowed.  (See also path_resolution(7).)

       EEXIST pathname already exists and O_CREAT and O_EXCL were used.

       EFAULT pathname points outside your accessible address space.

       EFBIG  pathname  refers to a regular file, too large to be opened; see O_LARGEFILE above.  (POSIX.1-2001 speci-
              fies the error EOVERFLOW for this case.)

       EINTR  While blocked waiting to complete an open of a slow device (e.g., a FIFO; see  fifo(7)),  the  call  was
              interrupted by a signal handler; see signal(7).

       EISDIR pathname refers to a directory and the access requested involved writing (that is, O_WRONLY or O_RDWR is
              set).

       ELOOP  Too many symbolic links were encountered in resolving pathname, or O_NOFOLLOW was specified but pathname
              was a symbolic link.

       EMFILE The process already has the maximum number of files open.

       ENAMETOOLONG
              pathname was too long.

       ENFILE The system limit on the total number of open files has been reached.

       ENODEV pathname  refers  to  a device special file and no corresponding device exists.  (This is a Linux kernel
              bug; in this situation ENXIO must be returned.)

       ENOENT O_CREAT is not set and the named file does not exist.  Or, a directory component in  pathname  does  not
              exist or is a dangling symbolic link.

       ENOMEM Insufficient kernel memory was available.

       ENOSPC pathname was to be created but the device containing pathname has no room for the new file.

       ENOTDIR
              A  component  used as a directory in pathname is not, in fact, a directory, or O_DIRECTORY was specified
              and pathname was not a directory.

       ENXIO  O_NONBLOCK | O_WRONLY is set, the named file is a FIFO and no process has the  file  open  for  reading.
              Or, the file is a device special file and no corresponding device exists.

       EPERM  The O_NOATIME flag was specified, but the effective user ID of the caller did not match the owner of the
              file and the caller was not privileged (CAP_FOWNER).

       EROFS  pathname refers to a file on a read-only file system and write access was requested.

       ETXTBSY
              pathname refers to an executable image which is currently being executed and write access was requested.

       EWOULDBLOCK
              The O_NONBLOCK flag was specified, and an incompatible lease was held on the file (see fcntl(2)).

CONFORMING TO
       SVr4,  4.3BSD,  POSIX.1-2001.  The O_DIRECTORY, O_NOATIME, and O_NOFOLLOW flags are Linux-specific, and one may
       need to define _GNU_SOURCE to obtain their definitions.

       The O_CLOEXEC flag is not specified in POSIX.1-2001, but is specified in POSIX.1-2001.

       O_DIRECT is not specified in POSIX; one has to define _GNU_SOURCE to get its definition.

NOTES
       Under Linux, the O_NONBLOCK flag indicates that one wants to open but does not necessarily have  the  intention
       to  read  or  write.   This  is  typically  used to open devices in order to get a file descriptor for use with
       ioctl(2).

       Unlike the other values that can be specified in flags, the access mode values O_RDONLY, O_WRONLY, and  O_RDWR,
       do  not  specify individual bits.  Rather, they define the low order two bits of flags, and are defined respec-
       tively as 0, 1, and 2.  In other words, the combination O_RDONLY | O_WRONLY is a logical error,  and  certainly
       does  not  have the same meaning as O_RDWR.  Linux reserves the special, non-standard access mode 3 (binary 11)
       in flags to mean: check for read and write permission on the file and return a descriptor that  can’t  be  used
       for  reading  or  writing.   This non-standard access mode is used by some Linux drivers to return a descriptor
       that is only to be used for device-specific ioctl(2) operations.

       The (undefined) effect of O_RDONLY | O_TRUNC varies among implementations.  On many systems the file  is  actu-
       ally truncated.

       There are many infelicities in the protocol underlying NFS, affecting amongst others O_SYNC and O_NDELAY.

       POSIX provides for three different variants of synchronized I/O, corresponding to the flags O_SYNC, O_DSYNC and
       O_RSYNC.  Currently (2.1.130) these are all synonymous under Linux.

       Note that open() can open device special files, but creat() cannot create them; use mknod(2) instead.

       On NFS file systems with UID mapping enabled, open() may return a file descriptor  but,  for  example,  read(2)
       requests  are  denied with EACCES.  This is because the client performs open() by checking the permissions, but
       UID mapping is performed by the server upon read and write requests.

       If the file is newly created, its st_atime, st_ctime, st_mtime fields (respectively, time of last access,  time
       of  last status change, and time of last modification; see stat(2)) are set to the current time, and so are the
       st_ctime and st_mtime fields of the parent directory.  Otherwise, if  the  file  is  modified  because  of  the
       O_TRUNC flag, its st_ctime and st_mtime fields are set to the current time.

   O_DIRECT
       The O_DIRECT flag may impose alignment restrictions on the length and address of userspace buffers and the file
       offset of I/Os.  In Linux alignment restrictions vary by file system and kernel version  and  might  be  absent
       entirely.  However there is currently no file system-independent interface for an application to discover these
       restrictions for a given file or file system.  Some file systems provide their own interfaces for doing so, for
       example the XFS_IOC_DIOINFO operation in xfsctl(3).

       Under  Linux 2.4, transfer sizes, and the alignment of user buffer and file offset must all be multiples of the
       logical block size of the file system.  Under Linux 2.6, alignment to 512-byte boundaries suffices.

       The O_DIRECT flag was introduced in SGI IRIX, where it has alignment restrictions similar  to  those  of  Linux
       2.4.   IRIX has also a fcntl(2) call to query appropriate alignments, and sizes.  FreeBSD 4.x introduced a flag
       of the same name, but without alignment restrictions.

       O_DIRECT support was added under Linux in kernel version 2.4.10.  Older Linux kernels simply ignore this  flag.
       Some file systems may not implement the flag and open() will fail with EINVAL if it is used.

       Applications  should  avoid mixing O_DIRECT and normal I/O to the same file, and especially to overlapping byte
       regions in the same file.  Even when the file system correctly handles the coherency issues in this  situation,
       overall  I/O  throughput  is  likely  to be slower than using either mode alone.  Likewise, applications should
       avoid mixing mmap(2) of files with direct I/O to the same files.

       The behaviour of O_DIRECT with NFS will differ from local file systems.  Older kernels, or  kernels  configured
       in  certain  ways, may not support this combination.  The NFS protocol does not support passing the flag to the
       server, so O_DIRECT I/O will only bypass the page cache on the client; the server may still cache the I/O.  The
       client  asks  the  server  to make the I/O synchronous to preserve the synchronous semantics of O_DIRECT.  Some
       servers will perform poorly under these circumstances, especially if the I/O size is small.  Some  servers  may
       also  be  configured to lie to clients about the I/O having reached stable storage; this will avoid the perfor-
       mance penalty at some risk to data integrity in the event of server power failure.  The Linux NFS client places
       no alignment restrictions on O_DIRECT I/O.

       In  summary,  O_DIRECT is a potentially powerful tool that should be used with caution.  It is recommended that
       applications treat use of O_DIRECT as a performance option which is disabled by default.

              "The thing that has always disturbed me about O_DIRECT is that the whole interface is just  stupid,  and
              was probably designed by a deranged monkey on some serious mind-controlling substances." — Linus

BUGS
       Currently,  it  is  not  possible  to  enable  signal-driven I/O by specifying O_ASYNC when calling open(); use
       fcntl(2) to enable this flag.

SEE ALSO
       chmod(2), chown(2), close(2), dup(2), fcntl(2), link(2),  lseek(2),  mknod(2),  mmap(2),  mount(2),  openat(2),
       read(2),   socket(2),  stat(2),  umask(2),  unlink(2),  write(2),  fopen(3),  feature_test_macros(7),  fifo(7),
       path_resolution(7), symlink(7)

COLOPHON
       This page is part of release 3.09 of the Linux man-pages project.  A description of the project,  and  informa-
       tion about reporting documentation bugs, can be found at http://www.kernel.org/doc/man-pages/.

Linux                             2008-08-21                           OPEN(2)

 

这个是open 的 man 文件，里面也没有提到O_RDWR的定义，但是在程序中确实看到了O_RDWR的身影，按么它是在哪里定义的呢？

 

我的fedora下是/usr/include/bits/fcntl.h ，里面有

/* open/fcntl - O_SYNC is only implemented on blocks devices and on files
   located on an ext2 file system */
#define O_ACCMODE       0003
#define O_RDONLY          00
#define O_WRONLY         01
#define O_RDWR             02
#define O_CREAT            0100 /* not fcntl */
#define O_EXCL              0200 /* not fcntl */
#define O_NOCTTY          0400 /* not fcntl */
#define O_TRUNC            01000 /* not fcntl */
#define O_APPEND          02000
#define O_NONBLOCK     04000
#define O_NDELAY           O_NONBLOCK
#define O_SYNC              010000
#define O_FSYNC             O_SYNC
#define O_ASYNC            020000