fcntl(2)


NAME
     fcntl - miscellaneous file descriptor control functions

SYNOPSIS
     #include <fcntl.h>

     int fcntl(int fd, int cmd, [data])

DESCRIPTION
     Fcntl()  performs  several  file  descriptor  related   functions,   like
     duplicating  a  file  descriptor,  setting the "close on exec" attribute,
     etc.  The fd argument is the file descriptor to operate on,  cmd  is  the
     command  code  of  the  operation  to  perform,  and  data is an optional
     argument to give or receive parameters.   The  command  codes  and  other
     symbols and types are declared in <fcntl.h>.  The commands are:

     fcntl(fd, F_DUPFD, int fd2)
          Returns a new file descriptor that is a duplicate of file descriptor
          fd.  It shares the same file pointer and the same file status flags,
          but has separate file descriptor flags that are initially off.   The
          value  of  the  duplicate  file  descriptor  is  the first free file
          descriptor greater than or equal to fd2.

     fcntl(fd, F_GETFD)
          Returns the file descriptor flags associated  with  file  descriptor
          fd.   The  flags  are the "close on exec" flag FD_CLOEXEC that, when
          set, causes the file  descriptor  to  be  closed  when  the  process
          executes  another  program.  The Minix-vmd specific FD_ASYNCHIO flag
          marks a file descriptor for asynchronous I/O operation.

     fcntl(fd, F_SETFD, int flags)
          Set the file descriptor flags of fd to flags.

     fcntl(fd, F_GETFL)
          Return the file status flags and file access modes  associated  with
          the  file associated with file descriptor fd.  The file status flags
          are O_NONBLOCK (non blocking I/O) and O_APPEND (append  mode).   The
          file  access  modes  are O_RDONLY (read-only), O_WRONLY (write-only)
          and O_RDWR (read-write).  These flags are also used  in  the  second
          argument of open(2).

     fcntl(fd, F_SETFL, int flags)
          Set the file status flags of the file referenced  by  fd  to  flags.
          Only  O_NONBLOCK and O_APPEND may be changed.  Access mode flags are
          ignored.

     The next four commands use a parameter  of  type  struct  flock  that  is
     defined in <fcntl.h> as:

          struct flock {
              short   l_type;     /* F_RDLCK, F_WRLCK, or F_UNLCK */
              short   l_whence;   /* SEEK_SET, SEEK_CUR, or SEEK_END */
              off_t   l_start;    /* byte offset to start of segment */
              off_t   l_len;      /* length of segment */
              pid_t   l_pid;      /* process id of the locks' owner */
          };

     This structure describes a  segment  of  a  file.   L_type  is  the  lock
     operation  performed  on  the  file segment:  F_RDLCK to set a read lock,
     F_WRLCK to set a write lock, and  F_UNLCK  to  remove  a  lock.   Several
     processes  may  have  a  read lock on a segment, but only one process can
     have a write  lock.   L_whence  tells  if  the  l_start  offset  must  be
     interpreted  from  the  start  of  the  file (SEEK_SET), the current file
     position (SEEK_CUR),  or  the  end  of  the  file  (SEEK_END).   This  is
     analogous  to  the third parameter of lseek(2).  These SEEK_* symbols are
     declared in <unistd.h>.  L_start is the starting offset of the segment of
     the  file.  L_end is the length of the segment.  If zero then the segment
     extends until end of file.   L_pid  is  the  process-id  of  the  process
     currently holding a lock on the segment.  It is returned by F_GETLK.

     fcntl(fd, F_GETLK, struct flock *lkp)
          Find out if some other process has a lock on a segment of  the  file
          associated  by  file  descriptor  fd  that overlaps with the segment
          described by the flock structure pointed to by lkp.  If the  segment
          is  not  locked  then  l_type is set to F_UNLCK.  Otherwise an flock
          structure is returned through lkp that describes the  lock  held  by
          the  other  process.   L_start  is  set relative to the start of the
          file.

     fcntl(fd, F_SETLK, struct flock *lkp)
          Register a lock on a  segment  of  the  file  associated  with  file
          descriptor  fd.   The  file segment is described by the struct flock
          pointed to by lkp.  This call returns an error if any  part  of  the
          segment is already locked.

     fcntl(fd, F_SETLKW, struct flock *lkp)
          Register a lock on a  segment  of  the  file  associated  with  file
          descriptor  fd.   The  file segment is described by the struct flock
          pointed to by lkp.  This call blocks waiting  for  the  lock  to  be
          released if any part of the segment is already locked.

     fcntl(fd, F_FREESP, struct flock *lkp)
          This Minix-vmd specific call frees a segment of disk space  occupied
          by  the  file  associated  with  file descriptor fd.  The segment is
          described by the struct flock  pointed  to  by  lkp.   The  file  is
          truncated  in  length  to  the byte position indicated by l_start if
          l_len is zero.  If l_len is nonzero then the file  keeps  its  size,
          but  the  freed  bytes  now  read as zeros.  (Other than sharing the
          flock structure, this call has nothing to do with  locking.)   (This
          call is common among UNIX(-like) systems.)

     fcntl(fd, F_SEEK, u64_t pos)
          This Minix-vmd specific call sets the  file  position  of  the  file
          associated  with  file descriptor fd to the byte offset indicated by
          the 64-bit number pos.  This is analogous to the call

               lseek(fd, pos, SEEK_SET)

          except that F_SEEK can be used on devices larger than 4 gigabyte.

SEE ALSO
     open(2), dup(2), lseek(2), ftruncate(3), int64(3).

DIAGNOSTICS
     Fcntl returns a file descriptor, flags, or 0  to  indicate  success.   On
     error  -1 is returned, with errno set to the appropriate error code.  The
     most notable errors are:

     EINTR
          If a blocked F_SETLKW operation is interrupted by a signal  that  is
          caught.

     EAGAIN
          By F_SETLK if a segment cannot be locked.

     EBADF
          A bad file descriptor in general, or an attempt  to  place  a  write
          lock on a file that is not open for writing, etc.

     ENOLCK
          No locks available, the file system code has  run  out  of  internal
          table space.

AUTHOR
     Kees J. Bot <kjb@cs.vu.nl>