Linux loopback sometimes gets stuck, blocking squashfs based test containers

[jimklimov]

Sometimes Linux loopback gets stuck with a loop device that can not be managed - neither destroyed nor removed. This then blocks re-mounting of e.g. same squashfs image for different VMs or even reusing it for a single new one. Host has to be rebooted to clear that state. Different images can be used however.

CI-RESETVM-INFO: /root/ci-reset-virtual-machine.sh:
    Creating a new VM rootfs at '/srv/libvirt/rootfs/fty-upstream-devel'
CI-RESETVM-INFO: /root/ci-reset-virtual-machine.sh:
    Mount the common RO squashfs at
    '/srv/libvirt/snapshots/../overlays-ro/fty-devel-image-18.01.05-10.09.04+1_x86_64_fty-devel_x86_64_upstream.squashfs-ro'
mount: /dev/loop1 is already mounted or
    /var/scratch/_lxc/_srv_libvirt/overlays-ro/fty-devel-image-18.01.05-10.09.04+1_x86_64_fty-devel_x86_64_upstream.squashfs-ro busy
CI-RESETVM-FATAL: /root/ci-reset-virtual-machine.sh: Can't mount squashfs

This is documented quite a lot over the years, so is a system issue. Maybe ci-reset-virtual-machine.sh can do something to try to avoid it however - e.g. if it is discovered how to trigger the issue.