[RFC] non-root X

[Tiago Vignatti]

On Thu, Jul 02, 2009 at 06:44:13AM +0200, Keith Packard wrote:
> On Wed, 2009-07-01 at 13:11 -0700, Jesse Barnes wrote:
> > or we can push
> > the xf86EnableIO call into drivers as needed).
> 
> I'd say drivers that need IO access should call the function; the server
> shouldn't do it for them. The big question is how to know ahead of time
> whether this is necessary. One could, in theory, have the init scripts
> probe the available video devices and chmod u+s /usr/bin/Xorg or some
> such, but that seems pretty ugly to me.

Yeah, I'd like to see a better and generic (non-KMS drivers) solution here as
well.

Maybe we could glue the enabling IO code together with some external tool that
we also need to do the int10 POST on (poor) chipsets that rely on this. 

E.g. Dave's posttool can be used for that. We can hook a HAL call inside
xf86CallDriverProbe asking for the "initialization" in posttool. Then
magically such tool returns all bits needed for the X server proceed its work.

Anyway, seems a little bit far from our reality :(


Cheers,

            Tiago