Files and the file system

What actually happens when you save a file, and why an inode is not the file name.

12 min · 36 XP

A file system organises blocks on a device into named files and directories. A directory is itself a file that maps names to inode numbers — the name is not the file; it is a label pointing at one.

An inode stores a file's metadata and where its data blocks live: size, permissions, timestamps, block pointers. It does not store the name. That is why one file can have several names (hard links) and none of them is "the real one".

Disk reads are thousands of times slower than memory, so the kernel keeps recently used file data in a page cache. A repeated read may never touch the disk at all.

# Путь — это просто имена каталогов, которые ядро проходит шаг за шагом.
# Конкретный ФАЙЛ живёт в inode, а не в имени.

$ ls -i /etc/hosts
42 /etc/hosts          # 42 — номер inode

$ ln /etc/hosts /tmp/alias     # жёсткая ссылка: +1 имя у ТОГО ЖЕ inode
$ ln -s /etc/hosts /tmp/soft    # симлинк: отдельный файл со ссылкой

# Место на диске занимают ДАННЫЕ, а не имя файла.

Practice

You write to a file and the call returns successfully. Is the data on disk?
Hints
- Where does the kernel put writes before they reach the disk?
- What is the point of fsync if writes already reached the disk?
12 XP
Put the steps of reading a file in order.
Hints
- A path is a sequence of names, each resolved in turn.
- Check the fast path before the slow one.
13 XP
Explain why deleting a file that another process has open does not necessarily free its disk space.
Hints
- What keeps the inode alive — the name, or something else?
- Count the references, not the directory entries.
13 XP