File I/O and the Universal I/O Model.

• We are looking at chapters 4 and 5. READ them.
• As we have seen, file descriptior are small integers
  • Usually some are open when the program starts.
  • Others can be created with pipe.
  • We will soon look at open
• We will refer to this picture
  • 
• Note that each process has a File descriptor table
  • This points to the system wide Open File Table.
  • Which points to the system wide I-node Table
  • Which points to devices, file systems and other things.
• In this picture
  • For process A, fd1 and fd20 both point to the same open file, how did this happen?
  • For process A, fd2 and process B, fd2 both point to the same file, how did this happen?
• The Unix/Linux I/O system is called universal.
  • This means that everything that you can perform I/O on looks like a file.
  • Or at least uses the open, read, write, close, file descriptior semantics.
  • What are these files?
    • Caution, they are not text files, object files, libraries,...
    • These are all examples of regular files
    • Writing seven different types of regular files on a test is good for one point out of seven.
  • The seven types are
    • Regular files, what you think of as a file.
    • Directories, or what you might call a folder.
    • Symbolic links.
      • There is another type of link called a hard link, but this is really not a distinct file type, it is a regular file with multiple directory entries.
    • Named Pipes, a file system implementation of a pipe
    • Sockets, a file system implementation of a different communications device
    • Character special devices, the interface to hardware that transfers data a character at a time (keyboards)
    • Block special devices, the interface to large data transfer hardware (hard drives).
  • We should explore these
    • ls -l
      • type field.
      • Permission fields.
    • ln -s oldfile newfile
    • mkfifo pipename
    • /dev/
    • tty
• I have implemented a version of his copy program (page 71) here