Skip to main content

Creating and Exploring a Logical Disk

In this lab we’ll create a logical disk using a loopback device, format it with ext4, and explore the structures the filesystem lays down.

We’ll stick with ext4 here since it’s the default on many Linux distros and strikes a balance of performance and reliability. Other options include XFS, Btrfs, and ZFS—I’ve done comparisons of those in a separate article.

1. Setup​

Install required tools:

sudo apt-get update
sudo apt-get install -y e2fsprogs util-linux

2. Create a Logical Disk​

We’ll use a sparse file as the backing store. Sparse means the file appears to be 1 GiB, but the actual space is only allocated as we write into it.

mkdir -p ~/fs-lab && cd ~/fs-lab
truncate -s 1G disk.img   # create a 1 GiB sparse file

Check the difference between apparent vs actual size:

ls -lh disk.img   # shows "apparent size" (1.0G)
du -h disk.img    # shows "real allocated size" (33MB for me)

3. Attach as a Loopback Device​

Turn the file into a block device with losetup:

sudo losetup -fP disk.img
losetup -a   # list loop devices

You’ll see something like:

/dev/loop7: [2065]:1234567 (/home/user/fs-lab/disk.img)

Here /dev/loop7 is our logical disk.

4. Format with ext4​

sudo mkfs.ext4 -L FSLAB /dev/loop7
mkdir -p mnt
sudo mount /dev/loop7 mnt

Now mnt/ is an ext4 filesystem inside disk.img.

5. Inspect the Filesystem​

Immediately after formatting:

du -h .

Example output:

16K   ./mnt/lost+found
20K   ./mnt
33M   ./disk.img
33M   .

What’s going on?​

  • lost+found (16K) → auto-created by ext4; fsck uses it to store orphaned files recovered after crashes.

  • mnt/ (20K total) → just the root directory plus lost+found.

  • disk.img (33M) → although created as 1 GiB sparse, ext4 consumes ~33 MB immediately for metadata:

    • superblock + backup superblocks
    • group descriptors
    • inode tables (preallocated)
    • block & inode bitmaps
    • journal area

6. Logical vs Physical Space​

Key distinction:

  • Logical size: what the file pretends to be (ls -lh).
  • Physical size: how much real disk space is consumed (du -h).

Sparse allocation means you can over-provision. For example, you could create 5 × 400 GB loop disks on a 1 TB SSD. They’d all exist, but once you actually try to use more than 1 TB total, you’ll hit ENOSPC (“no space left on device”).

7. Safety Note​

  • Anything under /dev/loopX is safe to experiment with—it only touches disk.img.
  • Never run mkfs on your real disk (/dev/nvme0n1 or similar).

8. Next Steps​

From here, we can: