Michael Hitch reports that changing the DELAY(100) in
lk201.c to DELAY(1000) fixes the mouse on a 3100;  in fact,
DELAY(150 works on a 3100.   Thanks, Michael.

I'd like to get this nailed and into the upcoming patch release.

Is there a brave soul who can try a new kernel on a 2100, and verify
that DELAY(150) also works on a 2100?  It should be, if the linear
fitting of the cpuspeed multipliers is accurate, but it'd be great to
get this nailed now rather than requesting another pullup.