Loop Device Lifecycle Management

1. Creating the Image File

1.1 When Loop Storage Fits

Treat an ordinary file as a /dev/loopN block device: reads and writes go through the block device interface, while the backing file is still just a file on the host.

Mounting ISO and disk images
Local practice and stress testing
Encrypted volumes and ciphertext images
Simulating a cloud disk or a CSI block volume
Exposing files read-only
Temporary mounts

Compared with a real disk, loop device performance is affected by the filesystem hosting the backing file; the association is not persistent by default, so after a reboot you generally have to run losetup again or add it to /etc/fstab.

1.2 Creating the Image

Prepare a backing file on the host.

Sparse file, space allocated on demand

1
truncate -s 10G /data/tmp/disk.img

Pre-allocate the full size, suitable for stress testing

1
dd if=/dev/zero of=/data/tmp/disk.img bs=1M count=10240

Pre-allocate the full size, faster than dd

1
fallocate -l 10G /data/tmp/disk.img

Specify the block size, aligned with the target filesystem’s blocks

1
dd if=/dev/zero of=/data/tmp/disk.img bs=4k count=2621440

2. Associating the Loop Device

Attach the image file to /dev/loopN.

Automatically allocate a free loop device and associate it

1
losetup -f --show /data/tmp/disk.img

Associate a specific loop device

1
losetup /dev/loop0 /data/tmp/disk.img

Specify the sector size and read-only mode when associating

1
2
3
losetup /dev/loop0 /data/tmp/disk.img \
  --sector-size 4096 \
  --read-only

List all associated loop devices

1
losetup -a

Show the details of one loop device

1
losetup -l /dev/loop0

Look up the loop device for an image file

1
losetup -j /data/tmp/disk.img

3. Formatting

Create XFS on the whole disk

1
mkfs.xfs /dev/loop0

XFS read-only check

1
xfs_repair -n /dev/loop0

4. Mounting and Using

Mount the whole-disk loop device or the image file.

Create the mount point

1
mkdir -p /mnt/looptest

Ordinary mount

1
mount /dev/loop0 /mnt/looptest

Read-only mount

1
mount -o ro /dev/loop0 /mnt/looptest

Mount the image file directly

1
mount -t xfs -o loop /data/tmp/disk.img /mnt/looptest

mount will find a free loop device automatically.

Verify the mount

1
2
findmnt /mnt/looptest
mount | grep loop0

5. Growing

After growing the image, update the backing file, the loop device capacity, and the filesystem size in that order.

Grow the image file to 20G

1
fallocate -l 20G /data/tmp/disk.img

Change only the image size metadata

1
truncate -s 20G /data/tmp/disk.img

Tell the loop device to re-read the capacity

1
losetup -c /dev/loop0

XFS online growth

1
xfs_growfs /mnt/looptest

6. Unmounting and Detaching

Release in the order mount → loop association.

Unmount the filesystem

1
umount /mnt/looptest

Lazy unmount

1
umount -l /mnt/looptest

Kill the processes holding the mount point

1
fuser -km /mnt/looptest

Detach the loop device

1
losetup -d /dev/loop0

Detach all loop devices

1
losetup -D

7. Cleanup

Detach the loop association first, then delete the image file. Do it in the wrong order and losetup -a is left with a stale entry whose backing file carries (deleted).

Delete the image file

1
2
3
umount /mnt/looptest
losetup -d /dev/loop0
rm -f /data/tmp/disk.img

Confirm there is no leftover loop device

1
2
losetup -a
ls /dev/loop* 2>/dev/null

Clean up a (deleted) leftover

1
2
3
findmnt /dev/loop0
umount /mnt/looptest          # 若仍挂载
losetup -d /dev/loop0