[Intel-gfx] [PATCH] LVDS: Fix output mode list construction.

[ykzhao]

On Wed, 2009-10-14 at 04:42 +0800, Eric Anholt wrote:
> On Wed, 2009-10-14 at 06:36 +1000, Dave Airlie wrote:
> > On Wed, Oct 14, 2009 at 5:18 AM, Adam Jackson <ajax at redhat.com> wrote:
> > > Construct a list of modes smaller than the LVDS panel size if either:
> > >
> > > - the panel does not have EDID or DisplayID
> > > - it has EDID/dID but does not claim to be continuous-frequency (in
> > >  which case the server would have done it for us already)
> > >
> > > The old method would attempt to do this by mangling the EDID block and
> > > parsed ranges, which was unreasonably fragile, and would fail to do so
> > > when EDID existed but claimed only a single mode.
> > 
> > This really should be done in the kernel from what I can see, so all KMS users
> > get a consistent list of modes for the LVDS device.
> 
> I'd been saying before that it's probably the job of userland to cook up
> a bunch of random modes for continuous-frequency outputs, if it wants to
> for its interface to users.  Some sort of server interface for "give me
> a collection of modes below this size at this refresh" (lvds) and "give
> me a collection of modes below this size at various refreshes" (vga)
> seems like it'd be nice for KMS drivers.
Agree with what Eric said.
It will be better that the userland adds a bunch of random modes for the
continuous-frequency output.
If LVDS output device has too many display mode before entering
X-environment, maybe it will select one display mode whose visible part
is less than LVDS panel.

Thanks.
   Yakui
>