At 01:07 PM 3/8/00 , Anders Magnusson wrote:
> > On Mon, 06 Mar 2000 19:26:33 -0800 
> >  Matt Thomas <matt@3am-software.com> wrote:
> > 
> >  > The problem is that the ncr53c9x MI code doesn't pass a ptr to the proc
> >  > so I can't pass it to bus_dmamap_load.  I didn't feel like changing the 
> >  > MI interface.
> > 
> > ...and, in fact, it's not appropriate to pass the proc * in, because the
> > code assumes kernel addresses, because physio() calls vmapbuf() :-)
> > 
>Yes, that is a problem here: All other drivers assume that
>vmapbuf/vunmapbuf do _not_ do anything, otherwise they do not work with
>raw I/O. 

I looked at modifying the ncr53c9x to pass the proc but that would involve
extensive changes (such more than i would want to make).  I might be able
to cheat and only do vmapbuf/vunmapbuf on KA46/KA48/KA49.

Also today I found out that the KA49 (and related designs) use a slightly
different mechanism for their SGMAP.  Since they have a SGEC, they only have
a 8192 entry SGMAP in dedicated SRAM.  So the vsbus_dma needs adapt to this
difference. 
-- 
Matt Thomas               Internet:   matt@3am-software.com
3am Software Foundry      WWW URL:    http://www.3am-software.com/bio/matt/
Cupertino, CA             Disclaimer: I avow all knowledge of this message