DragonFly kernel List (threaded) for 2003-09
Re: Project UDI (Device Drivers)
I'm not particularly biased in favor of UDI. As a matter of fact I think the UDI
people always thought of it a secondary driver framework. I understand it should
basicly works on FreeBSD 4.x though, so I just wanted the idea floating around,
if it becomes important later on, great, if it definitely dies.. so be it..
I am wondering though.. will DragonFlyBSD keep newbus, perhaps as a userland API?
FreeBSD is making changes there (inheritance in kobj() ). What about busdma? I
guess you wouldn't want to break new drivers coming from FreeBSD.
Matthew Dillon wrote:
> IMHO I think we have to implement our own messaging framework first,
> because I think it would be too much of an interruption to try to
> integrate something like UDI from the get-go. But it sounds like once
> we do have our own framework that it *would* be possible, even practical,
> for those interested in UDI to port it as a user-level interface
> to begin with (since it appears to have abstractions sufficient to deal
> with a user-level interface), and once that is working it could be moved
> into the kernel. Development is a lot easier to do in userland.
> I do not personally believe I would have time to work on UDI myself, I
> have too many other bullet items to cover.