> From: greywolf@starwolf.starwolf.com
> Date: Tue, 11 Nov 1997 19:51:38 -0800
> 
> 38400 also worked in NetBSD 1.2.
> 
> [before 1.2.1, 1.2A, 1.2B...]

The fifo (or silo) overflows seen in the ZS driver are
caused by poor interrupt latency.  This is not the fault
of the ZS driver itself, but of other code in the sparc
kernel that keeps the ZS interrupts disabled too long.

I tried to explain this in my message on Sat, 25 Oct 1997.

No changes to the ZS driver can be expected to improve
this situation.  The problem is elsewhere.

Chuck Silvers did some work on tracking down the cause
of the poor ZS interrupt latency.  Any news there?