Skip site navigation (1)Skip section navigation (2)
Date:      Wed, 19 Apr 95 12:00:41 MDT
From:      terry@cs.weber.edu (Terry Lambert)
To:        simonm@dcs.gla.ac.uk (Dougal)
Cc:        hackers@freefall.cdrom.com
Subject:   Re: [DEVFS] your opinions sought!
Message-ID:  <9504191800.AA19187@cs.weber.edu>
In-Reply-To: <199504191528.IAA22087@freefall.cdrom.com> from "Dougal" at Apr 19, 95 03:28:00 pm

next in thread | previous in thread | raw e-mail | index | archive | help
> You know, I thought the major advantage of having a /dev directory is
> that the kernel doesn't have to know the names of all the various
> devices, because this mapping is specified by the filesystem.  The
> devfs is about to hardwire all this stuff in, at the expense of some
> kernel bloat.

Not true.  It is a device-install-time wiring that occurs; the only
think that will be wired is the relationship between node names and
actual devices.  This is erroneously missing from the /dev
implementation.  Please see my previous post on the merits of a
devfs.  I list at least 8 salient points in favor.

> Well, it's perfectly feasible (and adds less kernel bloat) to query
> the kernel at boot time for attached devices and build up the /dev
> directory based on the information.  This IMHO is a better solution
> than the devfs.

Trade the kernel bloat of devfs for the bloat removal of killing specfs.

> I hope this doesn't mean that some future things will actually
> *require* the devfs, please keep it optional.

I think remote boot and expanding minor numbers used as bit selectors
*require* a devfs to support remote boot from older (ie: Sun) machines.


					Terry Lambert
					terry@cs.weber.edu
---
Any opinions in this posting are my own and not those of my present
or previous employers.



Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?9504191800.AA19187>