Subject: Re: representation of persistent device status, was Re: devfs, was Re: ptyfs...
To: der Mouse <mouse@Rodents.Montreal.QC.CA>
From: Eric Haszlakiewicz <erh@jodi.nimenees.com>
List: tech-kern
Date: 11/18/2004 22:25:02
On Thu, Nov 18, 2004 at 10:15:37PM -0500, der Mouse wrote:
> > On Someday at Sometime, Someone Else wrote:
> > On the issue of namespaces, in one view of a "full devfs", even
> > having major,minor numbers as a namespace at all should go away, or
> > at least become a historical distraction.  The device *is* the vnode
> > in the devfs.
> 
> I think that would be a mistake.  I've too often had occasion to have
> device files in unusual places in the filesystem.  (Not all that often,
> but often enough - maybe a half-dozen times in the two-plus decades
> I've been a sysadmin.)  Some of those could be handled with multiple
> mounts of devfs, but some couldn't really.  I think it would be a major
> step backwards to lose the ability.

	I'll agree that completely losing the ability would be a bad thing.
However, I don't think it's a mistake to view the access through devfs
(i.e. through its vnodes) as the "official" device access paradigm.
	With the presence of devfs the majority of cases won't need a
mapping between device name and device major+minor numbers.  To continue
to allow access by on-disk device nodes I think it make more sense
to provide a comptibility mode where devices can be bound (by a userland
util even) to a particular major+minor.  Enable this option would
prebind common devices (like /dev/zero) to their historically expected
numbers, but a /dev/whatever that had never had a well defined major+minor
number could be assigned by the admin also.

eric