Comment 0 for bug 1716465

After the move of livefs builds to lxd containers, cloud image builds will fail because loop devices are separate mount points set up by lxd and are not getting bind mounted correctly into the livecd-rootfs chroot, therefore they are not present at the right point for kpartx -d to run to free the devices.

This is fixed by mount --rbind of /dev instead of mount --bind, which then also requires changes for the unwinding.