re: PSA: Clock drift and pkgin

[matthew green <mrg%eterna23.net@localhost>]

> By the way - another comment/reflection on this. I don't really
> understand how a monotonic clock could ever go backwards.

in netbsd it usually means the timecounter reported a value
that was lesser than the most recent value it reported, but
without it overflowing properly, and what happens is that a
"u_int delta" becomes negative (aka huge positive) and this
is detected and the time went backwards message is printed.

you are 100% correct it is a driver bug.  however, it's quite
easy to trip on this bug in the code, while rarely seeing it
in the real world.

   http://mail-index.netbsd.org/source-changes/2018/01/12/msg091064.html

is my "mostly fix this for sparc SMP" hack for reference.
don't forget the missing final fix for this:

   http://mail-index.netbsd.org/source-changes/2018/01/12/msg091067.html



.mrg.