I was experimenting with turning on the cache on the Sun3/260
(which appears to work!) but quickly discovered that the zs
driver would fail as soon as the CPU became much faster.

Interestingly, it fails with the same symptoms people saw when
trying to use it on the 3/160 (input interrupts don't work).
So, I became suspicious about speed-related issues in the zs
driver, and discovered (to my horror!) that there were several
places in the zs driver where the zs chip was accessed but
the recovery time requirement not satisfied!  That stupid chip
REALLY DOES NOT LIKE to be accessed within 1.6 uS. after any
previous access.  Evidently, it just goes comatose if you do!

Anyway, I suspect the ZS_DELAY fixes I just checked in might
also make the zs driver work on the Sun3/160 and similar.

Gordon