svn commit: r199002 - in head/sys: dev/fb dev/pci isa

[Robert Noland]

On Mon, 2009-11-09 at 13:39 -0500, Jung-uk Kim wrote:
> On Monday 09 November 2009 01:24 pm, Robert Noland wrote:
> > On Mon, 2009-11-09 at 13:22 -0500, Jung-uk Kim wrote:
> > > On Friday 06 November 2009 04:47 pm, John Baldwin wrote:
> > > > On Friday 06 November 2009 3:32:26 pm Jung-uk Kim wrote:
> > > > > Author: jkim
> > > > > Date: Fri Nov  6 20:32:26 2009
> > > > > New Revision: 199002
> > > > > URL: http://svn.freebsd.org/changeset/base/199002
> > > > >
> > > > > Log:
> > > > >   Remove duplicate suspend/resume code from vga_pci.c and let
> > > > > vga(4)
> > > >
> > > > register
> > > >
> > > > >   itself to an associated PCI device if it exists.  It is
> > > > > little bit hackish but it should fix build without frame
> > > > > buffer driver since r198964. Fix some style(9) nits in
> > > > > vga_isa.c while we are here.
> > > >
> > > > Hmm, did you consider having vga_isa use an identify routine to
> > > > attach itself as a child of vgapci0? The hack of knowing the
> > > > first thing in the softc is a pointer is really gross and I'd
> > > > rather avoid it.  Just creating a child of vgapci0 will
> > > > automatically cause suspend and resume to work w/o vgapci
> > > > having to have any special knowledge about vga_isa.
> > >
> > > In fact, I did consider it.  However, lack of my new bus foo
> > > prevented me from doing the right thing AND fixing the build
> > > breakage at the same.  I'll try little harder today.
> > >
> > > BTW, syscons <-> fb <-> vga <-> vgapci <-> etc. relationship is
> > > amazingly confusing. :-(
> >
> > Agreed, this is one of the things that has kept me from working on
> > kms...
> 
> Oh, are you actually thinking about implementing KMS?

It is going to become more and more difficult to maintain drm without
it.  Intel does at least somewhat support GEM w/o KMS, but radeon has
opted to simplify the approach and only support TTM/KMS as a single code
path.  They are technically seperate features, but we will mostly be on
our own if we do one without the other.  Nouveau also has KMS support
now, though I am not certain if they are supporting the non-KMS path as
well.

As more and more functionality is incorporated in KMS, the legacy paths
in Xorg will start to bit rot and we will be left in more of a
predicament than I am now...

robert.

> How about KGI?
> 
> http://www.kgi-project.org/
> 
> Whatever we do in this area, we will stall this newly resurrected 
> project again, unfortunately.  In fact, I am afraid I might have 
> already. :-(
> 
> Jung-uk Kim
-- 
Robert Noland <rnoland at FreeBSD.org>
FreeBSD