LXD drivers fail to unmap non rootfs devices if deploy fails

[dann1]

Description
As seen in the forum in several entries LXD drivers fail to deploy containers because the context disk was detected already mounted prior to the process of setting up the container storage. This condition occurs when, for any reason, the container fails to start and the error handling code fails to detect the device corresponding to the container context mountpoint.

To Reproduce
The idea is to create a container that fails to start and then trying to start it again.

For example, taking advantege of #3058

• Create a container with an ovs nic
• poweroff the contaner
• start the container (this will fail and leave the context mounted)
• delete-recreate or retry (this will fail when detecting context already mounted)

Expected behavior
LXD drivers should revert all of the storage actions (maps and mounts) when a start fails for whatever reason

Details

• Affected Component: Virtualization Drivers
• Hypervisor: LXD
• Version: 5.8.0

Additional context
When LXD attempts to start a container with declared disks in it devices section it changes the mountpoint of block devices, ex. a context device mapping with the container running is /var/lib/lxd/devices/one-14/disk.context.context

Progress Status

• Branch created
• Code committed to development branch
• Testing - QA
• Documentation
• Release notes - resolved issues, compatibility, known issues
• Code committed to upstream release/hotfix branches
• Documentation committed to upstream release/hotfix branches