> On 1997-08-21 at 06:58:20, David S. Miller wrote:
> > Oh no.
> >
> > There is already way too much crap under /proc as it is.
Its not the problem, that there is so much crap 'mounted' under /proc,
but that (nearly) all the code is in the fs/proc tree and the procfs-code
has become _very_ unreadable now.
> Ah, but the problem is that a file or a directory in /proc is way
> too easy to be created. :)
Yes ;-)
> Some people even dislike any new pseudo-fs detached from /proc,
> just because /proc is already there, why not use that...
Those pseudo-fs will be very procfs-like, so why not use /proc...
> > Make a new filesystem for each of the neat new toys we want to provide
> > a vfs interface to.
Well a real fs would be overkill, I think.
> Hmm.. but, to make things easier, probably the most inner core
> of procfs should be abstracted to be used by other "info" type
> filesystems; should that be possible and not be frowned upon? :)
Yes, this is probably the right way to go.
Just provide a simple-VFS filesystem, that provides a subset (like that
from proc) of the VFS features. It could handle single-file-mounts
(aka /proc/ksyms), or directory mounts (aka the /proc/<pid> stuff).
Itself would only manage the hierarchy of the tree below /proc.
This could clean up the code somewhat. I thougt about implementing this,
but I have no time at the moment :(
Richard.