Bug#353943: [Pkg-sysvinit-devel] Processed: wontfix

[Henrique de Moraes Holschuh]

On Tue, 28 Mar 2006, R.Ramkumar wrote:
> I guess I am mistaken. The feature request was to augment domount so
> that it could allow calls for bind mounts from init functions such
> as mountvirtfs. Correct me if I am wrong, but afaik these don't even
> need an entry in /etc/fstab (domount at best takes options from
> there if an entry exists).

You're correct, I did not understand the report fully.

> The currently distributed mountdevsubfs/mountvirtfs do not use bind
> mounts, but the local admin might need them in their custom init
> scripts - mount*fs are after all config files. The use case
> presented, as described in the original bug report, is to change the
> init script mountdevsubfs.sh, which mounts a tmpfs instance at
> /dev/shm to a shared instance between /dev/shm and /tmp, using bind
> mounts.

That would be the reply to Thomas's "why?" request.  Very well, looking at
where domount() is used, it looks sane to allow for some sort of messing
around with bind mounts.  Actally, with namespaces in general.

That's already supported in a kludgy way (see Thomas' reply).  We could
support a filesystem type of "none", which would bring support to non-kludge
status, as that's how mount(8) does it.

Thomas, what do you think?  mount will tell us if -t none is not what it
wants, and it is exactly what it wants for --bind, --move, --make-shared (or
their alternative fstab-friendly hacks of -obind, -omove...).

-- 
  "One disk to rule them all, One disk to find them. One disk to bring
  them all and in the darkness grind them. In the Land of Redmond
  where the shadows lie." -- The Silicon Valley Tarot
  Henrique Holschuh