zs[0-2]: silo overflow (Sun 3/280 problems?)

[Sundar Narasimhan]

Hi: We have a 25 MHz Sun 3-280 with a high-res monitor hooked upto another
VME through an HVE-2000 bus-to-bus adaptor. Our 16MHz earlier suns
(3-160's) have no problems under such a configuration. With the faster
boards though, any vmebus activity on the slave vme causes the sun to
freeze up for a while after which it emits one of these messages

	zs0: silo overflow 

Sometimes instead of zs0, other minor devices will be indicated (zs1 or
zs2 or zs3).

We have determined that no interrupt activity is actually present across
the bus adaptor, that it is not window system programs like suntools or X
that cause this problem, and that the serial line, mouse or keyboard have
nothing to do with this. We have also determined that we have the right
rev no./level of the 280 boards which reputedly had a lot of VME bus
problems initially.

One obvious hypothesis is that somehow the address mapping is all confused
and that the zs interfaces are responding to activity on the vme bus when
they shouldn't because they are supposed to be local (on-board the sun
cpu). Has anyone else seen this problem, and/or can someone explain how
the address mapping works to distinguish local from vme bus accesses? We
have tried the Sun customer support route and there wasn't anyone in Sun
who we talked to that could give us a clearer idea of what could be
happening..

-Sundar