UNIX File Descriptors

UNIX File Descriptors
An Introduction

Prof. David Bernstein
James Madison University

Computer Science Department

bernstdh@jmu.edu

Review

Files in UNIX
- A file is a sequence of bytes
- All I/O devices are modeled as files
Universal Operations:
- Open - A statement of intent to use an I/O device
- Read/Write - Copy bytes from/to a file to/from memory
- Close - A statement of intent to stop using an I/O device

Opening a File

The "Statement":
- Made using a system call
The "Response":
- A nonnegative integer called a file descriptor that identifies the file
Information Maintained by the Kernel:
- Process-specific file descriptor tables
- System-wide open file table
- System-wide i-node table (called a v-node table in System V)

Contents of the Tables

File Descriptor Table:
- Flags controlling the operation of the file descriptor
- A pointer/reference to the open file table
Open File Table:
- The status flags specified when the file was opened
- The file access mode
- The current file offset (in bytes, from the start of the file)
- A pointer/reference to the i-node table
i-Node Table:
- File type
- The locks held on the file
- Properties of the file (e.g., size, timestamp, permissions, etc...)

Relationships Between Tables

Relationships Between Tables (cont.)

Pointers to the Open File Table:
- One process can have multiple file descriptors point to the same entry (e.g., as a result of a call to dup())
- Multiple processes (e.g., a parent and child) can have file descriptors that point to the same entry
Pointers to the i-Node Table:
- Multiple files can point to the same i-node table entry (e.g., open() is called multiple times for the same file)

Duplicating a File Descriptor: dup()

int dup(int oldfd);

Purpose:

Create a copy of a file descriptor using the lowest available number

Details:

`oldfd`	The existing file descriptor
Return	The new file descriptor; -1 on error

#include <unistd.h>

Duplicating a File Descriptor: dup2()

int dup2(int oldfd, int newfd);

Purpose:

Create a copy of a file descriptor

Details:

`oldfd`	The existing file descriptor
`newfd`	The new file descriptor (to silently close and re-) use
Return	The new file descriptor; -1 on error

#include <unistd.h>

Duplicating a File Descriptor: Some Values of errno

`EBADF`	`fd` isn't a valid open file descriptor
`EMFILE`	The per-process limit has been reached

An Example of the Relationships Between Tables

"Standard" File Descriptors

An Important Note:
- Each process created by a shell begins life with three open files: standard input (0), standard output (1) and standard error (2)
Portability:
- <unistd.h> defines STDIN_FILENO, STDOUT_FILENO and STDERR_FILENO

Closing a File

The "Announcement":
- Made using a system call
The "Response":
- The file descriptor is returned to the pool of available descriptors

Finding Open File Descriptors

/dev/fd:
- A (virtual) directory containing file names of the form /dev/fd/n where n is one of the process's open file descriptors
Opening these Files:
- Linux: Opening a file in /dev/fd is equivalent to reopening the oritinal file
- Other Systems: Opening a file in /dev/fd is equivalent to calling dup()
For Convenience:
- /dev/stdin, /dev/stdout, and /dev/stderr are symbolic links to /dev/fd/0, /dev/fd/1, and /dev/fd/2

Using <stdio> with File Descriptors

An Observation:
- The C I/O library functions use a FILE* not a file descriptor
What We Need:
- The ability to associate a FILE* (i.e., stream) with a file descriptor

Using <stdio>: fdopen()

FILE *fdopen(int fd, const char *mode)

Purpose:

Associate a FILE* (i.e., stream) with a valid file descriptor

Details:

`fd`	The descriptor
`mode`	"r", "w", "a", or "r+"/"w+" for read, write, append, read and write
Return	A pointer to a stream or NULL

#include <stdio.h>