Subject: Re: aha0: DMA beyond end of ISA
To: None <current-users@NetBSD.ORG>
From: der Mouse <mouse@Holo.Rodents.Montreal.QC.CA>
Date: 11/17/1996 10:03:25
>> [ISA DMA past 16 megs, yet another round]
> It's the attitudes of those who tell the new NetBSD people that their
> hardware is the reason they can't run NetBSD that drive people away.
> I can handily name you one person who came to NetBSD before I did,
> and one who came after, who both run OpenBSD, simply because their
> 1542 will run in their machines.
> Don't let Michael VanLoon or whoever else would tell everyone that
> the solution is in the hardware drive people away from NetBSD before
> they've even come.
Here's my take on the whole thing. (I have no personal interest in
this; I don't run any port that has ISA busses.)
The problem is, at present much/all of the ISA code, including drivers
for ISA devices, is shared among multiple ports. Doing bounce buffers
the way "everybody else" does works for i386 folks, but it means either
(a) breaking the ISA code for other ports or (b) moving it back into MD
code, resulting in almost but not quite identical copies of drivers and
other ISA support code...or (c) doing it right, which takes a good deal
of effort and time.
core seems to consider neither (a) nor (b) acceptable. Given the
presence of the "unsupported" (meaning, supported only by other users,
not by NetBSD core) patches for people who just want it to work in i386
machines and don't care that the resulting source tree has at least one
of problems (a) or (b), I see nothing greatly wrong with this. And I
understand work is underway on (c)....
01 EE 31 F6 BB 0C 34 36 00 F3 7C 5A C1 A0 67 1D