Third: The mount point (/tmp/btrfs) is never cleaned up in the Robertos
patch.
I do umount but NOT rmdir because if for some reason umount fails (expecially in combination with a bug that do not let the error code be returned, I saw in busybox) I am going to remove the filesystem content. Moreover, polluting the /tmp is usually not an issue.
Two toughts ecollided in my head:
1. do not run rm -rf on a dir that can contain a umounted filesystem
2. just in case do rmdir that will fail and do not harm the content
The result was quite curios: I the doubt I do not do anything but polluting /tmp!