1. Advanced UNIX progamming
Fall 2002
Instructor: Ashok Srinivasan
Lecture 6
Acknowledgements: The syllabus and power point presentations are modified versions of those by T. Baker and X. Yuan

2. Announcements Reading assignment APUE Chapter 3 APUE Chapter 4
Pages 47-56, 56-62
APUE Chapter 4
Pages 77-81, 92-95
APUE Chapter 5

3. Review Portability Introduction to the UNIX API UNIX file system
Standards: ANSI, POSIX, etc
32 bit vs 64 bit
Byte order: Little endian vs big endian
Introduction to the UNIX API
Environment variables
Exit status
Process ID
User ID
UNIX file system
File system abstraction
Directories
File descriptors

4. Week 3 Topics UNIX file system
File system abstraction
Directories
File descriptors
Unix API Programming Examples and Techniques
Example with direct IO
open, close, fdopen, lseek, unlink
Variable argument list
HW1 hints

5. Week 3 Topics ... continued File I/O Process management Pipes
File descriptors
open, creat, close, dup, dup2
I/O redirection
Process management
fork, exit, wait, waitpid, execv
Pipes
Named and unnamed pipes
Implementing pipe in a shell

6. UNIX file system
File system abstraction
Directories
File descriptors

7. File system abstraction
File: a sequence of bytes of data
Filesystem: a space in which files can be stored
Link: a named logical connection from a directory to a file
Directory: a special kind of file, that can contain links to other files
Filename: the name of a link
Pathname: a chain of one or more filenames, separated by /'s

8. File system abstraction ... continued
inode: a segment of data in a filesystem that describes a file, including how to find the rest of the file in the system
File descriptor: a non-negative integer, with a per-process mapping to an open file description
Open file description: an OS internal data-structure, shareable between processes

9. Directories

10. Directories ... continued
Names belong to links, not to files
There may be multiple hard links to one file
Renaming only renames one link to that file
Unix allows both hard and soft links
A file will exist even after the last hard link to it has been removed, as long as there are references to it from open file descriptions
Soft links do not prevent deletion of the file
A directory may have multiple (hard) links to it
But this capability is usually restricted, to prevent creation of directory cycles

11. File Descriptors
Each open file is associated with an open file description
Each process has a (logical) array of references to open file descriptions
Logical indices into this array are file descriptors
These integer values are used to identify the files for I/O operations
The file descriptor 0 is reserved for standard input, the file descriptor 1 for standard output, and the file descriptor 2 for the standard error

12. File Descriptors ... continued

13. File Descriptors ... continued
The POSIX standard defines the following
File descriptor: A per-process, unique, nonnegative integer used to identify an open file for the purposes of file access
Open file description: A record of how a process or group of processes are currently accessing a file
Each file descriptor refers to exactly one open file description, but an open file description may be referred to by more than one file descriptor
A file offset, file status, and file access modes are attributes of an open file description
File access modes: Specification of whether the file can be read and written

14. File Descriptors ... continued
File offset: The byte position in the file where the next I/O operation through that open file description begins
Each open file description associated with a regular file, block special file, or directory has a file offset
There is no file offset specified for a pipe or FIFO (described later)
File status: Includes the following information
append mode or not
blocking/nonblocking
Etc

15. File Descriptors ... continued
FIFO special file: A type of file with the property that data written to such a file is read on a first-in-first-out basis
Pipe: An object accessed by one of the pair of file descriptors created by the pipe() function
Once created, the file descriptors can be used to manipulate the pipe, and it behaves identically to a FIFO special file when accessed this way
It has no name in the file hierarchy

16. File Descriptors ... continued
Important points
A file descriptor does not describe a file
It is just a number that is ephemerally associated with a particular open file description
An open file description describes a past "open" operation on a file; its does not describe the file
The description of the file is in the inode
There may be several different open file descriptors (or none) referring at it any given time

17. Unix API Programming Examples and Techniques
Examples with direct IO
open, close, fdopen, lseek, unlink
Variable argument list
Note:
Use man pages to get information on system calls
Look into the system header files
/usr/include/sys/types.h)

18. Direct I/O Using open()
The usual C-language stream-oriented I/O operations, like printf(), use buffers and process data character-by-character
They are implemented using the lower-level direct I/O operations read() and write()
In situations where we do not want to view the data as characters, where we want greater efficiency, or where the extra (stream) layer of buffering causes us problems with synchronization, it is better to use the direct I/O operations

19. Using man Look at the man page for open()
If there is more than one page on a given name, man will give you the one that is first in the chapter order of the Unix manual.
Shell commands are in Section 1, I/O and OS interface calls are in Section 2 and Section 3 respectively
Specification of section number varies
On Red Hat Linux, type man 2 open or man -S 2 open to see the page on open from Section 2 of the Unix manual
On Solaris, you can type man -s 2 open

20. man page for open #include <sys/types.h>
#include <sys/stat.h>
#include <fcntl.h>
Solaris 2.6 includes the following synopsis
int open(const char *path, int oflag, /* mode_t mode */ ...);
Red Hat Linux 6.2
int open(const char *pathname, int flags, mode_t mode);
1996 POSIX standard synopsis is as follows
int open(const char *path, int oflag, ...);
The latest official POSIX/Unix synopsis
No <sys/types.h>
What does the ... mean here? Will a compiler allow this in an actual program?

21. Variable Argument Lists
The ... indicates a variable number of arguments
Similar to that in printf()
For more on variable argument lists, look at the file /usr/include/stdarg.h
Functions with variable argument lists can be dangerous
It is difficult to check types, and the use of a correct number of arguments

22. example1.c illustrates a common programming error
Example Programs
example1.c illustrates a common programming error
Failure to provide the correct number of arguments to a vararg function
example2.c illustrates opening a file