Unix Files

tags : Linux, Storage, Network Programming

• A file is an abstract resource from which a sequence of bytes is read from or written to.
• The source and destination of those is unspecified, those could be anything.

Everything in unix are files, hard drives, pipes, network devices, stdin/stdout etc. Let’s separate it into regular and special files.

Regular files §

These are disk files. See Filesystems

Directory files §

• See Directory layer
• File that contains mapping of file names with inode numbers
• We call the process of removing a dentry “unlinking”. If no names point to the inode, the last link is removed, then the file is considered removed and the data freed.

Symbolic link files §

• Returns only a single string of text representing the location of its target, the other file it points to. Like an alias.
• An improvement is fast symlink which is basically storing the filepath in the inode structure if the structure can take that in.

Hard link files §

• Hard links are simply directory entries.
• All files have at least one directory entry and thus all files must at least have one hard link.
• They are that association between a name and an inode.
• Multiple hard links = Multiple names to the same file
• Creating hard links to directories is forbidden (infinite loop)
• Max no. of hard links to a single file is limited by the size of the ref counter inside the inode

Named pipe files §

• Also called FIFO, File used for IPC.
• Traditional pipe is “unnamed” and disappear as soon as the process finishes, the named pipe is a file that can be reused. Allows for transfer of info in a pipeline manner through processes that don’t share the same parent.
• Created via mkfifo, can be deleted using rm
• Not so popular nowadays, use sockets instead.

Socket files §

• Unix domain sockets AF_UNIX
• Similar to named pipe
• While named pipe are unidirectional, sockets are full duplex.
• What can we send
  • Data
  • File Descriptors: Allows a process to manipulate a file it may not already have access to.

Device files §

• All devices, pseudo or not are implemented as device drivers in the kernel. (This statement is not verified)
• Device files have major and minor number stored as inode attributes. The meaning of a device file determined by these 2 numbers.
  • major number: The driver to use
  • minor number: The device that the driver controls
• Use: Communication w a piece of hardware, Accessing system resources(/dev/null, /dev/random )

Device Driver §

• Not all device files are handled by a file system driver.
• Device files are handled by specific drivers for every one of them. It’s the Device Driver, a program that controls some type of device attached to the computer, sw interface to hw devices.
• Device Driver Functions: HW device discovery, detecting state changes, sending events to udevd etc.

Character device §

• Unbuffered / Stream. Usually read/written sequentially.
• Also called raw devices sometimes.

Block device §

• Buffered. May be accessed randomly.
• Usually for things like disks, which behave like large, fixed-size files.
• Linux specific: To represent disk devices, Linux only uses block devices, it does give option to simulate character file opens at all.
• See arch wiki for naming conventions of different physical block devices.

Pseudo device §

• These are device nodes that do not have a backing physical device. They are either character or block devices.
• These are sometimes also called virtual devices. major number(device driver) is usually 1. (eg. Try stat /dev/null)
• These devices are mapped to certain paths in the file system. Mostly that’s /dev
• Can be created using mknod

/dev §

• Naming convention used by files in /dev is a complete mess.
• There will be entries in /dev for devices you do not have. (i.e They don’t always correspond to physical devices)
• /dev in modern system will have a devtmpfs fs mounted.
  • devtmpfs is a device manager in the form of a filesystem that displays each device as files.
  • It allows devices to be dynamically shown without storing on disk (it’s stored in memory).
• Some examples: /dev/null, /dev/zero, /dev/full, /dev/random, /dev/stdin

Resources §

• Special Files