Re: Review/Test: Pseudo-device unit number management patch

From: M. Warner Losh <imp_at_bsdimp.com>
Date: Mon, 09 Feb 2004 01:03:18 -0700 (MST)
In message: <Pine.BSF.4.21.0402082112420.24800-100000_at_InterJet.elischer.org>
            Julian Elischer <julian_at_elischer.org> writes:
: 
: 
: On Mon, 9 Feb 2004, Poul-Henning Kamp wrote:
: 
: > In message <p06020404bc4c5fcbfe49_at_[128.113.24.47]>, Garance A Drosihn writes:
: > >At 9:11 AM +0100 2/8/04, Poul-Henning Kamp wrote:
: > >>Julian Elischer writes:
: > >>  >
: > >>  > Previously, nmdm started off with 4 devices visible in /dev now
: > >>  > I don't see that.. The old behaviour was intuitive... You saw
: > >>  > a device.
: > >>
: > >>This new behaviour is called a "cloning device" and it doesn't
: > >>allocate any resources until they are actually needed, which I
: > >>believe is the correct behaviour for pseudo-devices.
: > >
: > >Could there be some kind of fake "marker-device" sitting there,
: > >one which would use basically no resources, and which would
: > >change to the real-device when someone opens it? 
: > 
: > No, that is not possible with the model we have chosen.
: 
: s/we/I/

s/I/we/  Others have signed off on this model.  Maybe you could
explain how such a device would exist, and its exact semantics.  With
a cloning device that phk is talking about, you open /dev/foo, and
/dev/foo0, /dev/foo1, etc are automatically created.  This is how
things work on other systems for cloning devices.

One could argue the wisdom of making the tunnel device clonable (I
happen to agree with it, but I can see the oppsoing argument).
However, sensible cloning semantics have been well understood for
years.  BSD is very late to the party in not having these sorts of
devices before now.

Warner
Received on Sun Feb 08 2004 - 23:04:53 UTC

This archive was generated by hypermail 2.4.0 : Wed May 19 2021 - 11:37:42 UTC