[Intel-gfx] KMS breaks console

[Jesse Barnes]

On Tuesday, February 3, 2009 1:13 pm Paul Collins wrote:
> Jesse Barnes <jbarnes at virtuousgeek.org> writes:
> > On Tuesday, February 3, 2009 10:18 am Tino Keitel wrote:
> >> On Tue, Feb 03, 2009 at 10:11:54 -0800, Jesse Barnes wrote:
> >> > On Tuesday, February 3, 2009 9:51 am Tino Keitel wrote:
> >> > > Hi folks,
> >> > >
> >> > > I just enabled CONFIG_DRM_I915_KMS in 2.6.29-rc3 and now the text
> >> > > console is just black. This happens with the 2.6 branch on a Mac
> >> > > mini Core Duo with i945GM. Any hints how to solve this?
> >> >
> >> > You need to enable the framebuffer console and make sure it's loaded.
> >>
> >> It is:
> >>
> >> CONFIG_FRAMEBUFFER_CONSOLE=y
> >
> > Is your machine still running?  Can you capture kernel logs with drm
> > debugging enabled (just pass 'debug=1' when you modprobe drm)?
>
> I hit what may be the same problem on a Mac Mini Core 2 Duo with 945GM.
>
> After some poking around I saw there's some code in intel_lvds_init
> regarding weirdness with the Mac Mini reporting LVDS when there is none,
> but the workaround doesn't seem to be active, as the DRM_DEBUG is never
> logged (drivers/gpu/drm/i915/intel_lvds.c, line 516ish).
>
> Judging by the X log, the xorg driver has similar workaround:
>
>   (II) intel(0): Suspected Mac Mini, ignoring the LVDS
>
> I tried sticking a 'goto failed' near the the top of intel_lvds_init and
> then I did get a console, albeit at 800x600 rather than my screen's
> native 1680x1050.
>
> Here are two debug=1 logs, the first with vanilla code and the second
> with the crippled intel_lvds_init.

Ah right, the Mac mini quirk.  I don't have a machine to test with but if you 
want to try to fix up that code we'd be happy to take a patch.  You might 
just need to add some more ID checks.  For a real fix you'd have to figure 
out where they keep their mode timings (probably in OpenFirmware somewhere) 
and get them.

-- 
Jesse Barnes, Intel Open Source Technology Center