Ploop/Mount helpers

Despite the fact that ploop is not a file system, a trick exist to make it look and feel like so — i.e. to use usual mount and umount commands rather than the ploop(8) tool. This article describes how it works and can be used.

Introduction

ploop is not a filesystem per se, but a kernel w:loop device driver, providing a way to represent a ploop image (or a set of stacked images) as a block device. On top of that block device provided by ploop, an ext4 file system is created and used, for example, for storing files of a specific container.

Therefore a complete ploop mount consists of two steps:

1. "mount" ploop image(s) to create a ploop device (/dev/ploopNNNNN)
2. mount a filesystem residing on this ploop device to a mount point

For simplicity, these two are usually wrapped to be done together in one step (say when you use ploop mount with -m option). In reality, step 1 is kernel assembling a device out of image(s), and step 2 is the real mount.

Mounting

You can use the following syntax to mount a ploop device and the filesystem inside it:

mount -t ploop [option ...] /path/to/DiskDescriptor.xml /mount/point

The following options are supported:

-r, -o ro

mount read-only

-v, --verbose

be more verbose

-f, --fake

do everything except for the actual mount

-n, --no-mtab, -s

these options are deliberately ignored

Unmounting

To unmount, DiskDescriptor can be specified:

umount /path/to/DiskDescriptor.xml

A mount point can be used as well:

umount /mount/point

Limitations

Note that umount can only work if:

1. /etc/mtab is a separate file (not a symlink to /proc/mounts);
2. mounting was done using mount (not ploop mount).

Otherwise, umount binary will not be able to find ploop as the "filesystem" field in /etc/mtab, and will not call umount.ploop helper. As a result, file system will be unmounted, but ploop device itself will stay mounted.

Alternative

ploop mount and ploop umount commands, as described in ploop(8) man page.

See also

• Ploop
• OpenVZ Bug #2817