>the rnd driver also uses the clock cycle counter and has macros for
>accessing it.  Can we have only one set please?

That may mean ditching the rng macros, since they were (almost by
design) a poor fit for other would-be users, like high-res accurate
timekeeping in general, and NTP in particular.

Also, if we're going down this route, can we please, *PLEASE*
construct _nano_time counters, not microtime?  It doesn't require any
more bits, and microtime() on multi-GHz CPus looks a bit silly.