table of contents
| ACCESS(2) | System Calls Manual | ACCESS(2) | 
NAME¶
access, eaccess,
    faccessat — check
    accessibility of a file
LIBRARY¶
Standard C Library (libc, -lc)
SYNOPSIS¶
#include
    <unistd.h>
int
  
  access(const
    char *path, int
    mode);
int
  
  eaccess(const
    char *path, int
    mode);
int
  
  faccessat(int
    fd, const char
    *path, int mode,
    int flag);
DESCRIPTION¶
The
    access()
    and eaccess() system calls check the accessibility
    of the file named by the path argument for the access
    permissions indicated by the mode argument. The value
    of mode is either the bitwise-inclusive OR of the
    access permissions to be checked (R_OK for read
    permission, W_OK for write permission, and
    X_OK for execute/search permission), or the
    existence test (F_OK).
For additional information, see the File Access Permission section of intro(2).
The
    eaccess()
    system call uses the effective user ID and the group access list to
    authorize the request; the access() system call uses
    the real user ID in place of the effective user ID, the real group ID in
    place of the effective group ID, and the rest of the group access list.
The
    faccessat()
    system call is equivalent to access() except in the
    case where path specifies a relative path. In this
    case the file whose accessibility is to be determined is located relative to
    the directory associated with the file descriptor fd
    instead of the current working directory. If
    faccessat() is passed the special value
    AT_FDCWD in the fd parameter,
    the current working directory is used and the behavior is identical to a
    call to access(). Values for
    flag are constructed by a bitwise-inclusive OR of
    flags from the following list, defined in
    <fcntl.h>:
- AT_EACCESS
- The checks for accessibility are performed using the effective user and
      group IDs instead of the real user and group ID as required in a call to
      access().
Even if a process's real or effective user has appropriate
    privileges and indicates success for X_OK, the file
    may not actually have execute permission bits set. Likewise for
    R_OK and W_OK.
RETURN VALUES¶
Upon successful completion, the value 0 is returned; otherwise the value -1 is returned and the global variable errno is set to indicate the error.
ERRORS¶
access(),
    eaccess(), or faccessat()
    will fail if:
- [EINVAL]
- The value of the mode argument is invalid.
- [ENOTDIR]
- A component of the path prefix is not a directory.
- [ENAMETOOLONG]
- A component of a pathname exceeded 255 characters, or an entire path name exceeded 1023 characters.
- [ENOENT]
- The named file does not exist.
- [ELOOP]
- Too many symbolic links were encountered in translating the pathname.
- [EROFS]
- Write access is requested for a file on a read-only file system.
- [ETXTBSY]
- Write access is requested for a pure procedure (shared text) file presently being executed.
- [EACCES]
- Permission bits of the file mode do not permit the requested access, or search permission is denied on a component of the path prefix.
- [EFAULT]
- The path argument points outside the process's allocated address space.
- [EIO]
- An I/O error occurred while reading from or writing to the file system.
Also, the faccessat() system call may fail
    if:
- [EBADF]
- The path argument does not specify an absolute path
      and the fd argument is neither
      AT_FDCWDnor a valid file descriptor.
- [EINVAL]
- The value of the flag argument is not valid.
- [ENOTDIR]
- The path argument is not an absolute path and
      fd is neither AT_FDCWDnor a file descriptor associated with a directory.
SEE ALSO¶
STANDARDS¶
The access() system call is expected to
    conform to IEEE Std 1003.1-1990
    (“POSIX.1”). The faccessat()
    system call follows The Open Group Extended API Set 2 specification.
HISTORY¶
The access() function appeared in
    Version 7 AT&T UNIX. The
    faccessat() system call appeared in
    FreeBSD 8.0.
SECURITY CONSIDERATIONS¶
The access() system call is a potential
    security hole due to race conditions and should never be used. Set-user-ID
    and set-group-ID applications should restore the effective user or group ID,
    and perform actions directly rather than use
    access() to simulate access checks for the real user
    or group ID. The eaccess() system call likewise may
    be subject to races if used inappropriately.
access() remains useful for providing
    clues to users as to whether operations make sense for particular filesystem
    objects (e.g. 'delete' menu item only highlighted in a writable folder ...
    avoiding interpretation of the st_mode bits that the application might not
    understand -- e.g. in the case of AFS). It also allows a cheaper file
    existence test than stat(2).
| September 15, 2014 | Debian |