>  > I've already gotten mail suggesting I switch to OpenBSD.  Well, I'm
>  > not currently looking at this solution, but why can't we have a
>  > machine dependent bounce buffer solution while we wait for a machine
>  > independent solution.  It's not like we don't have other machine
>  > dependent code.
> 
>     Umm, since no one seems to have pointed it out....  Just cuz it
> ain't in the master tree doesn't mean it doesn't exist.  ISA DMA has
> been available for NetBSD for some time; I was using it in my 32MB 486
> with an aha1542 without any difficulty for a couple of months.  I
> don't know if its actively being maintained at this point, but you can
> get it from ftp://bseis.eis.cs.tu-bs.de/pub/hannken/isadma-patch.


That's the patch directly out of OpenBSD.  Juergen Hannken-Illjes (the
author) worked with us OpenBSD folk to integrate that patch into
OpenBSD.  David Mazieres and I put some work into that too, to solve a
few rough edges.  I can't say how well it will work in NetBSD; it's
not trivial stuff.

But it does work very well, and performance is great, and a lot of
really valuable OpenBSD developers use aha controllers on their
development machines.

I predict the same discussion comes up next week about ATAPI cdrom.
And the answer will be the same.