Filesystems

A filesystem is how the kernel organizes data on a storage device. It defines the on-disk structure — inodes, directory entries, free space management, and metadata. Linux supports dozens of filesystems; the most common are ext4, xfs, and btrfs.

ext4 — The Workhorse

ext4 (Fourth Extended Filesystem) is the default on most Linux distributions. It’s a journaling filesystem with good performance, large filesystem support, and mature tooling.

# Create ext4 filesystem
mkfs.ext4 /dev/sdb1
mkfs.ext4 -L mydata /dev/sdb1          # with label
mkfs.ext4 -E lazy_itable_init=1 /dev/sdb1  # fast init (no zeroing)
 
# Tune ext4 at creation:
mkfs.ext4 -O has_journal,extent,dir_index,flex_bg /dev/sdb1
 
# Mount
mount -t ext4 /dev/sdb1 /mnt/data
 
# Key ext4 mount options:
mount -o noatime,nodiratime,errors=remount-ro /dev/sdb1 /mnt/data
# noatime: don't update atime on read (performance)
# nodiratime: don't update atime for directories
# errors=remount-ro: remount read-only on errors

ext4 Features

• Journaling: writes metadata changes to a journal first, replays on crash recovery
• Extent mapping: contiguous block allocation (vs block-mapped in ext2/3)
• Delayed allocation: writes buffered in memory before allocation (better clustering)
• Online defragmentation: e4defrag /mount/point
• Resize online: can grow filesystem while mounted
• inode size: configurable (default 256 bytes, can be 512 for extended attributes)

# Check filesystem
fsck.ext4 /dev/sdb1          # check (unmount first!)
dumpe2fs -h /dev/sdb1        # show superblock info
 
# Show inode usage
df -i /mnt/data
 
# Get inode count
tune2fs -l /dev/sdb1 | grep -i inode

xfs — Scalable and Fast

xfs is the default on RHEL/CentOS/Fedora. It’s optimized for large files and large filesystems (multi-TB), and performs well with concurrent I/O.

# Create xfs
mkfs.xfs /dev/sdb1
mkfs.xfs -L mydata /dev/sdb1
mkfs.xfs -f /dev/sdb1     # force (if existing filesystem)
 
# Mount
mount -t xfs /dev/sdb1 /mnt/data
 
# Key mount options:
mount -o noatime,nodiratime,logbufs=8,logdev=/dev/sdc1 /dev/sdb1 /mnt/data
# logdev: put journal on separate fast device (SSD)

xfs Features

• Journaled: external journal device option
• Grow online: xfs_growfs /mount/point
• Quota: project-based and user-based quotas
• Defragmentation: xfs_fsr (online)
• Free space management: UUIDs, labels, access time tracking

# Check xfs
xfs_check /dev/sdb1
xfs_info /mnt/data         # show xfs filesystem info
xfs_growfs /mnt/data        # grow (needs free space in LVM or partition)
 
# Repair (must unmount)
xfs_repair /dev/sdb1

btrfs — Copy-on-Write, Snapshots

btrfs is a modern CoW (copy-on-write) filesystem with built-in snapshots, compression, checksumming, and multi-device support. Good for containers and VMs.

# Create btrfs
mkfs.btrfs /dev/sdb1
mkfs.btrfs -L mydata /dev/sdb1
 
# Mount
mount -t btrfs /dev/sdb1 /mnt/data
 
# Compressed
mount -t btrfs -o compress=zstd /dev/sdb1 /mnt/data

btrfs Key Features

• Copy-on-Write (CoW): modified pages written to new blocks, original unchanged
• Snapshots: read-only snapshot of a subvolume at a point in time
• Subvolumes: independently mountable filesystem within a filesystem
• Checksumming: data integrity (CRC32c)
• Compression: zstd, zlib, lzo
• RAID: software RAID 0/1/10 (not RAID5/6 — unstable)
• Send/Receive: stream snapshots for replication

# Create subvolume
btrfs subvolume create /mnt/data/vol1
 
# Create snapshot
btrfs subvolume snapshot /mnt/data /mnt/data/snap1
 
# Show subvolumes
btrfs subvolume list /mnt/data
 
# Compressed mount
mount -o compress=zstd:3 /dev/sdb1 /mnt/data
 
# Usage
btrfs filesystem df /mnt/data
btrfs filesystem usage /mnt/data
 
# Balance (re-distribute data across devices)
btrfs balance start -dusage=0 /mnt/data
 
# Convert ext4 → btrfs (in-place, requires free space):
btrfs-convert /dev/sdb1

swap — Virtual Memory

# Create swap
mkswap /dev/sdb1
mkswap -L swap0 /dev/sdb1
 
# Enable
swapon /dev/sdb1
swapon -s                      # show active swaps
 
# Disable
swapoff /dev/sdb1
 
# Priority (higher = preferred)
swapon -p 100 /dev/sdb1
 
# In /etc/fstab:
# UUID=abc123 none swap sw,pri=100 0 0

vfat — FAT32 (USB, EFI)

mkfs.vfat /dev/sdb1
# or
mkfs.fat /dev/sdb1
 
mount -t vfat /dev/sdb1 /mnt/usb

Mount Options Reference

# Generic mount options (most filesystems):
defaults     # rw,suid,dev,exec,auto,nouser,async
ro           # read-only
rw           # read-write
noatime      # don't update access time (best for performance)
nodiratime   # don't update directory atimes
nosuid       # ignore suid bit
noexec       # no execution of binaries
nodev        # don't interpret device files
sync         # all writes synchronous (vs async)
async        # writes buffered (default)
suid         # honor suid bit (default)
user        # allow unprivileged user to mount
users       # allow any user to mount
auto         # mount at boot (via /etc/fstab)
noauto      # don't mount at boot

Comparing Filesystems

Feature	ext4	xfs	btrfs
Journal	Yes	Yes	Yes (CoW)
Max size	1EB	8EB	16EB
Max file	16TB	8EB	16EB
Copy-on-Write	No	No	Yes
Snapshots	No (via LVM)	No	Yes
Compression	No	No	Yes
Checksumming	No	No	Yes
Online grow	Yes	Yes	Yes
Online shrink	No	No	Yes
Default on	Ubuntu	RHEL	OpenSUSE