per-device sysctls

[M. Warner Losh]

In message: <20040226.130330.54448406.imp at bsdimp.com>
            "M. Warner Losh" <imp at bsdimp.com> writes:
: In message: <2468.1077825069 at critter.freebsd.dk>
:             "Poul-Henning Kamp" <phk at phk.freebsd.dk> writes:
: : In message <xzpbrnllnj6.fsf at dwp.des.no>, Dag-Erling =?iso-8859-1?q?Sm=F8rgrav?=
: :  writes:
: : 
: : >BTW, the dev tree on my laptop looks like this:
: : >
: : >des at des ~% sysctl dev
: : >dev.nexus0.npx0.%class: npx
: : >dev.nexus0.npx0.%desc: math processor
: : >dev.nexus0.npx0.%driver: npx
: : >dev.nexus0.acpi0.acpi_timer0.%class: acpi_timer
: : >dev.nexus0.acpi0.acpi_timer0.%desc: 24-bit timer at 3.579545MHz
: : >dev.nexus0.acpi0.acpi_timer0.%driver: acpi_timer
: : 
: : Just for the record:  Having based this on newbus it only documents
: : hardware-anchored device drivers.  I don't know if this is going to
: : be a limitation we will have to address or not, but I think you should
: : reserve the toplevel name "pseudo" or "sw" or similar as a placeholder
: : for non-hardware device drivers, just in case.
: 
: There's not really a tree for pseudo devices right now.  They just are
: kludged into the system any old way that they can be at the moment.
: They aren't part of the hardware device hierarchy.

Hate to reply to myself....

By impliciation we shouldn't kludge this mechanism to include them,
but consider having a pseudo0 attached to nexus0 that accumulates them
so they are less of a kludge to the whole system.  Also gives a more
uniform way of controlling them.

Warner