Subject: Re: union filesystem [was Re: debugging kernel core dumps ?]
To: None <current-users@NetBSD.ORG>
From: der Mouse <mouse@Collatz.McRCIM.McGill.EDU>
List: current-users
Date: 09/06/1994 10:00:22
>> mount -t fdesc fdesc /fdesc
>> mount -t union -o -b /fdesc /dev

> The way you're expected to mount fdesc is:

> mount -t fdesc -o union fdesc /dev

That is unsatisfactory, largely because /dev then becomes effectively
read-only - one has to unmount the fdesc layer to do anything, like
creating new device special files.

Hmm.  If I use nullfs to clone /dev somewhere else, _before_ I mount
fdesc -o union, perhaps that would let me modify /dev underneath the
union fdesc layer...and it would fix problems caused by root doing
something like "command producing lots of output > /dev/nulll" and the
like.  I'll have to try that - maybe this is a feature rather than a
bug.

But still, it seems to me that what I tried _ought_ to work.  (I
actually wanted -t union -o -b mounts for another purpose, and it
occurred to me to use it for /dev as well.)

> Regarding union fs in general, it's somewhat unreliable, but should
> be fixed when we (I?) import a new version, some time after the
> release.  (The changes are just too pervasive to do now.)

Oh, okay.  I'll wait, then.

					der Mouse

			    mouse@collatz.mcrcim.mcgill.edu

------------------------------------------------------------------------------