>I don't see why clockvar.h belongs there at all.  Alphas have multiple
>different system clock types, and so something like clockvar.h makes
>sense for them.  However, it does _not_ belong in a TC-specific
>directory for the Alpha, and i'm not sure it makes sense _at all_ for
>a pmax.

Chris,

I was following your own (outdated?) suggestion of using /sys/dev/tc
as a directory for DEC-specific drivers.  I was integrating some
patches that created verbatim copies of both clockvar.h and
mcclockvar.h.  Rather than commiting copies into the pmax tree, I put
them in /ssy/dev/tc.

The DECsystem 5400 and 5500 are Qbus-baed ``pmaxes'' with a VAX-style
interface, in many respects the same as the KA650- and 655-based Qbus
machines.  They have VAX-stlye interval timers and real-time clock
(they even use a VAX-style SCB, with console I/O via cnrint and cnxint
interrupts vectred through the SCB.)  There's discussion about adding
support for these systems after Jason Thorpe's rework of the pmax
autoconfig code gets committed.

I intend to use the Alpha port <clockvar.h> interface, unchanged, on
both Qbus and non-Qbus DECstations. The quotes around ``cpu-independent''
in clockvar.h can then go :).

I think /sys/dev/tc is the least-bad place in the tree for it at the
moment: it's the closest to a sys/dev/dec that we have.  If there's a
better place to put these, or you think we should not bother sharing
them at all, that's fine with me.  If you think it's time to create
/sys/dev/dec, that's fine with me too.  I'd prefer to make it a truly
MI interface, which would put it in /sys/dev/ or maye /sys/sys, but I
don't know if there are other ports that need to support different
clock hardware, and would use this interface.  Sparc, maybe?


I'm also hoping to properly emulaet the Digital RTC-as-TOY with the
RTC in either 1972 or 1973; you're welcome to share that, if the
change is worth it for any users who dual-boot Digital Unix and
NetbSD/alpha.