Re: VAXstation 3100/30 KA420 SCSI DMA problem on netboot (and fix)

[Ken Wellsch <kwellsch%tampabay.rr.com@localhost>]

On Sun, Feb 18, 2024 at 12:28:18PM -0500, Ken Wellsch wrote:
> On Sun, Feb 18, 2024 at 11:59:30PM +0900, Izumi Tsutsui wrote:
> > 
> > Ok, this indicates:
> > - No valid register at 200C00E0 (KA420_STC_MODE for KA420 to switch DMA size)
> > - KA43_SC1_DADR/KA420_SCD_DADR (200C00A0),
> >   KA43_SC1_DCNT/KA420_SCD_DCNT (200C00C0), and
> >   KA43_SC1_DDIR/KA420_SCD_DDIR (200C00C4) are also available as KA420
> > - There is some register is at 200C00B0
> >   (no info at least in <machine/ka43.h>)
> 
> When I saw your reference to "200C00B0" I was curious and did a "Google"
> search on "VAX 200C00B0".
> 
> The top item that came up was a bitsavers.org reference to "KA46 System
> Board Specification" which pointed to this paragraph, for what it is
> worth...
> 
> 10.3.1 Register Definition
> 
>   The thirteen internal registers of the 53C94 SCSI Contoller chip are
>   accessed as bits<07:00> of thirteen sequential longwords. Writes to
>   the remaining bits of each longword are ignored, reads of those bits
>   return UNPREDICTABLE data. The address range used by the 53C94 chip
>   register set is 200C.0080 to 200C.00B0. In the following register
>   description, only bits<5:2> of the address will be referenced.

Oh, but wait. The document title made for another interesting search
item "ka43 system board specification" which lead me to a document
with this title:

"Engineering Specification for the KA43 Processor Module"

Looking at the discussion for the NCR chip an 80 series like the KA42,
it has an interesting reference to a "compatibility mode" with a "reduced
16K" DMA buffer to match what the VS410 (aka microVax 2000) uses...

Anyway, it looked rather tantalizing and someone with your knowledge
is more likely than me to determine whether it may impact your issue.