the current status of nullfs, unionfs

[Erez Zadok]

In message <76f962c6050310092461fc850 at mail.gmail.com>, Lou Kamenov writes:
> On Thu, 10 Mar 2005 09:19:10 -0500, Michael W. Lucas
> <mwlucas at blackhelicopters.org> wrote:
> > On Thu, Mar 10, 2005 at 12:38:43PM +0100, Jeremie Le Hen wrote:
> > But the mere existence of even a basic regression test would be a
> > start and would encourage people to not hose things further.
> [..]
> > Folks, don't let the fact that you're not a guru stop you from taking
> > a kiddie step and submitting a basic test!
> [..]
> 
> I do use unionfs on daily basis. Mostly to union $home/bin directories and
> such. For the last 1.5y I had it crash only 2 times. Of course trying
> to unmount /bin
> will turn into hell. I've used it successfully with pdumpfs from ports
> to restore
> old filespace view.  I surely think that a stable unionfs will be a
> good thing (tm).
> 
> Erez's unionfs has the same problem, the case there is that you wont be
> able to unmount it at all. (At least last time I tried with 1.0.3)

You should NEVER be allowed to unmount an underlying file system of a
stackable file system, if it's busy or in use, no more than you can remove a
/dev/sda drive while ffs is mounted on it.  Our approach in the Linux
unionfs is to prevent users from shooting themselves too much in the
foot. :-)

However, our unionfs does provide mechanisms such that read-only or non-busy
file system branches in the union, *can* indeed be removed safely.
Generally we can support arbitrary insertions and removals of branches
anywhere in the (fan-out) union; however, we found out that most of our
unionfs users rarely want or need that.

BTW, the latest version of our linux unionfs is 1.0.9.  There has been a lot
of work done on our unionfs recently, and it was deemed stable enough that
several LiveCD distros, including the just-released Knoppix 3.8, are using
it.

> Problem or not it could be easily solved with simple heuristics.
> Building a filespace
> with unioning shouldnt really  be that hard.
> 
> best,
> l
> _______________________________________________
> freebsd-fs at freebsd.org mailing list
> http://lists.freebsd.org/mailman/listinfo/freebsd-fs
> To unsubscribe, send any mail to "freebsd-fs-unsubscribe at freebsd.org"
> 



Erez.