vga arbiter (news)

[Tiago Vignatti]

Alan Cox escreveu:
>> [4] http://people.freedesktop.org/~vignatti/VGA.Notes
> 
> 
> This seems to imply that there is a single locked instance of the VGA
> legacy addressing. That isn't neccessarily the case. On some PA-RISC for
> example you have multiple PCI busses hanging off the system busses.
> 
> So the arbiter needs to know about PCI. Your kernel interface seems to
> but not your X one ?

yes, the architecture can set a hook to override enabling/disabling of a 
given device resources in the kernel. With the current userspace 
implementation (libvga) this is hide for the clients. So yes, the X 
doesn't know anything about this right now. It should?


> A second question this leads onto is basically about the interface. I
> would expect a sysfs interface to have made more sense with one node per
> candidate vga device.

This can be easily implemented. "Only" a massive modification on the 
userspace side :( But what would be the practical benefit of the sysfs 
interface that a char device doesn't has here?


thanks

-- 
Tiago Vignatti
C3SL - Centro de Computação Científica e Software Livre
www.c3sl.ufpr.br