This page collects detailed discussion and possible solutions to the problem described here.
Stupid: Not closing ANY block devices at all
One (really stupid) "solution" is simply not closing _ALL_ (below root-fs as well as non-root-fs) the block devices, by not calling the respective init-scripts in rc0 and rc6. While the data should still be safe (as all the filesystems are unmounted, which should be save) this is really ugly,.. as we wouldn't close devices, we actually could close (namely everything only below non-root-filesystems).
This would however solve the issue, that the user get ugly warnings, about not closable devices, used by the root-filesystem.
Stupid: Close all block devices BUT the ones used by the root-filesystem
- This is still unclean. Although the data should be safe.
- Security issues as dm-crypt keys are left in RAM.
- Ugly messages for block devices used by the root-filesystem, which couldn't be closed.
- Could we filter out the error messages for just those devices which _are_ somewhere below the root-fs? How?