Lesson 5 of 7

Files, Descriptors & Filesystems

On Unix, documents, , network sockets and more are all read and written through the same small file . You'll learn how files, descriptors and permissions work.

A filesystem maps names to data: directories are of name-to-inode entries, and an inode holds a file's metadata (size, owner, permissions, timestamps) plus to its data blocks on disk. The filename is not the file — several names can point to one inode (hard links), and a file deleted while open lives on until the last descriptor closes.

Bash

ls -li                    # -i shows inode numbers
stat README.md            # everything the inode knows

# Permissions: read/write/execute for user/group/other
# rwxr-xr--  =  owner: rwx, group: r-x, others: r--
chmod u+x script.sh       # let the owner execute
chmod 644 notes.txt       # rw-r--r-- in octal shorthand

# 'Everything is a file':
cat /dev/urandom | head -c 16 | xxd    # random bytes
echo hi > /dev/null                    # the bit bucket
cat /proc/cpuinfo | head -5            # kernel state as files

# File descriptors: small integers naming open files.
# Every process starts with three:
#   0 = stdin, 1 = stdout, 2 = stderr
import os
fd = os.open("data.txt", os.O_WRONLY | os.O_CREAT)
print(fd)                 # 3 — next free number
os.write(fd, b"hello\n")
os.close(fd)

# Shell redirection is just descriptor surgery:
#   ./prog > out.txt 2>&1
# means: point fd 1 at out.txt, then point fd 2
# wherever fd 1 points.

Finished reading?Mark it done to track your progress.

Tip: use the ← and → keys to move between lessons.