File I/O

1
File I/O

• Five basic UNIX functions
• open
• read
• write
• lseek
• close
• Unbuffered IO
• kernel calls

2
File Descriptors

• Files tracked by file descriptors
• non-negative integers (positive or 0)
• open, creat return FD
• read, write, lseek use FD
• shells use 0, 1, 2 as stdin, stdout, stderr
• not kernal function
• POSIX maps to STDIN_FILENO, STDOUT_FILENO,
  STDERR_FILENO
• max FDs system dependent (OPEN_MAX)

3
open Function

• Open or create files
• called with pathname, oflag, mode
• pathname is pathname of file to open/create
• oflag is options applied
• mode is file attribute set
• returns lowest numbered FD available, -1 on error

4
creat Function

• can be used to create new files
• takes pathname, mode
• creates file in write mode only
• outdated by new options to open
• returns FD or -1 on error

5
close Function

• closes open files
• takes FD
• releases record locks
• program termination closes all open files in UNIX
  
• returns 0 on success, -1 on error

6
lseek Function

• all open files have current file offset
• normally non-negative of bytes from beginning
  of file
• set to zero on file open, assuming no O_APPEND
• pointer to location in file for next read/write
• takes FD, offset, whence
• whence specifies option for how offset is updated
• SEEK_SET, relative to beginning of file
• SEEK_CUR, relative to current value
• SEEK_END, relative to end of file
• offset positive or negative
• returns new offset or -1 for error (check
  explicitly for -1)

7
read Function

• reads data from open file
• takes FD, buff, of bytes
• returns of bytes read, 0 on end of file, -1 on
  error
• can return less than of bytes requested

8
write Function

• writes to open file
• takes FD, buff, of bytes
• returns bytes written, -1 on error

9
File Sharing

• UNIX allows processes to share open files
• kernal uses three structures for each file
• process FD table
• file table
• v-node table

10
process FD table

• file descriptor flags
• pointer to file table entry

11
file table

• file status flags
• current file offset (lseek)
• pointer to v-node table entry

12
v-node table

• file type information
• functions that operate on file
• i-node information
• current file size

13
File Sharing Model

• process table entries map to file table entries
• file table entries map to v-node table entries
• processes sharing files have separate file table
  entries for each shared file
• relative to process attributes such as current
  file offset
• each open file has one v-node entry, regardless
  of how many processes are sharing

14
Atomic Operations

• Atomic operations solve the problem of
  overlapping operations by separate processes
• Atomic operations represent indivisible
  collections of separate functions, all performed
  together, or none performed at all

15
dup and dup2 Functions

• used to duplicate FDs
• dup takes FD
• dup2 takes FD, new FD
• dup returns next available FD, -1 on error
• dup2 returns specified new FD, -1 on error
• dup2 atomic

16
fcntl Function

• changes properties of open file
• dup existing FD
• get/set FD flags (process table)
• get/set file status flags (file table)
• get/set async I/O ownership
• get/set record locks
• takes FD, cmd, cmd dependent val
• return dependent on command, -1 on error

17
ioctl Function

• No single standard
• takes FD, request
• possible requests taken from device specific
  headers
• no single standard

18
/dev/fd

• Primarily intended to clean up shell pipelines
• opening the path /dev/fd/n equivalent to
  duplicating FD n
• shared file table entry

19
Write a program that will do the following

• Print to standard out the associated descriptive
  text of any given error number input as an
  argument, or a list of all possible error numbers
  and descriptions if given a particular flag
• Input will be either an error number, or a flag
  as follows
• help
• print a usage statement to standard out
• all
• print a full list of possible error numbers and
  their descriptions to standard out
• Any input NOT within the range of valid error
  numbers will generate an error message (including
  a list of defined numbers), and the usage message
  to standard out. Your program will need to
  determine all defined error numbers. For purposes
  of this project, you need not check for
  definition beyond the range 0-1000.