per-device sysctls

[M. Warner Losh]

In message: <3045.1077827869 at critter.freebsd.dk>
            "Poul-Henning Kamp" <phk at phk.freebsd.dk> writes:
: In message <20040226.125904.08946359.imp at bsdimp.com>, "M. Warner Losh" writes:
: >In message: <xzp8yipk4gh.fsf at dwp.des.no>
: >            des at des.no (Dag-Erling Smørgrav) writes:
: >: "Poul-Henning Kamp" <phk at phk.freebsd.dk> writes:
: >: > Dag-Erling =?iso-8859-1?q?Sm=F8rgrav?= writes:
: >: > > "M. Warner Losh" <imp at bsdimp.com> writes:
: >: > > > That is a good reason to transitioning to this, so long as we can come
: >: > > > up with a good way to represent detached nodes.
: >: > > As long as they have a device_t, it should be a piece of cake.
: >: > Having a device_t is a property of having hardware, not of being a
: >: > device driver.
: >: 
: >: I believe that by "detached nodes", Warner means hardware that does
: >: not have a driver.
: >
: >They have a device_t, however.  All nodes in the tree have a device_t
: >(kind of by definition). Not all nodes in the tree have a devclass
: >associated with their device_t (eg, not all devices are attached).
: >this is why you'll see lots of 'unknown' nodes in the devinfo output.
: 
: GEOM, NETGRAPH pty, tun, tap, nmdm and similar have no newbus
: infestation and there would have to really good reasons to infest them.

If they want to use this mechanism, they need to use the mechanism.
I'm cool with them not using this mechanism, but I don't think we
should kludge it to allow for them to use it.  That's what I'm
saying...  This is a newbus only mechanism and let's not get bogged
down in accomidatnig non-newbus things until we have the newbus parts
of it working.

Warner