Systems Industries 9900 and Simacs

[Dave Martindale]

A word of warning about SI:  The 9800 controller has at least several
bugs in it.  The worst is incorrect emulation of the MBA byte count
register on writes which fail due to errors.  This renders the bad-sector
forwarding code in the 4.2BSD driver useless.  It is possible, but a pain,
to fix this.  At Waterloo, since we have real Massbus controllers as well
as a 9800 running the same code, the code tries to determine what type
of controller it has by looking at the drive serial numbers, which are
supposed to follow a special pattern for the SI controller.  They seem
to be correct after resetting the 9800, but sometime later they have
become wrong, so the code thinks it sees a real MBA out there.  I'll have
to figure out a way around this problem too.

The byte count bug was a known bug when the old 9400 controller was
around, and it wasn't fixed in the 9800.  Ask about its status in the 9900;
I'd be surprised if it was fixed.

And SI seems awfully paranoid about giving out circuit diagrams for
their stuff, an attitude which I do not like (though other manufacturers
have the same problem).