Subject: Re: wsdisplay, early consoles etc.
To: None <>
From: Michael Lorenz <>
List: tech-kern
Date: 11/04/2006 13:54:02
Hash: SHA1


On Oct 26, 2006, at 13:36, Michael Lorenz wrote:

> I'd like to make the following addition to wsdisplay.c - a version of 
> wsdisplay_cnattach() that can be overridden exactly once. The reason 
> to do that is this - macppc has been hardcoded to use ofb as console 
> for ages but we do have a few accelerated framebuffer drivers now ( 
> like machfb, voodoofb, radeonfb, chipsfb ) which all work on macppc. 
> The problem is that macppc relies on having a rasops-based console 
> more or less from kernel startup.
> In my opinion it would be a good idea to have a generic framebuffer 
> console attach as early as possible and later allow some 
> hardware-specific driver to override it. This would nicely solve 
> several headaches - we wouldn't have to decide which driver to attach 
> on startup before autoconf. Specific drivers wouldn't have to 
> implement xxx_cnattach() methods. And drivers would be plug-compatible 
> between at least sparc64 and macppc ( sparc64 uses OF output until a 
> driver takes over ).
> The same could be useful on i386 and the like as well - cnattach vga 
> for early startup, then - maybe - let a framebuffer driver take over 
> for an accelerated high-resolution console.
> I have these changes in my local source tree for ages now, they seem 
> to be unproblematic.
> Besides that - we should come up with a generic interface to let a 
> driver find out if it's supposed to become the system console or not. 
> On OpenFirmware machines that's easy - just check /choosen/stdout - 
> but it's ugly to have OF-dependencies in drivers that reside in 
> sys/dev/pci. Maybe we should just have the MD autoconf code attach 
> properties for:
> - - console
> - - display mode
> - - font
> - - whatever I can't think of right now

I guess no protest means no objections?

have fun
Version: GnuPG v1.2.4 (Darwin)