[Intel-gfx] [PATCH] drm/i915: fix screen blank issue in TV detect

[Jesse Barnes]

On Mon, 21 Sep 2009 11:41:14 +0800
Zhenyu Wang <zhenyuw at linux.intel.com> wrote:

> On 2009.09.18 15:32:42 -0700, Jesse Barnes wrote:
> > On Mon, 14 Sep 2009 10:06:01 +0800
> > Zhenyu Wang <zhenyuw at linux.intel.com> wrote:
> > 
> > > On 2009.09.11 10:51:06 -0700, Eric Anholt wrote:
> > > > On Fri, 2009-09-11 at 13:19 +0800, Zhenyu Wang wrote:
> > > > > In load time, this one trys to detect TV by current encoder
> > > > > state instead of load-detect method. Because load-detect will
> > > > > set mode which disable VGA mode entirely, but couldn't
> > > > > restore back later. Destroy VGA mode which is the mode after
> > > > > system boot caused black screen problem, although load fbcon
> > > > > or X later could setup the mode on LVDS without problem.
> > > > > 
> > > > > That fixes screen blank issue seen on Macbook after boot and
> > > > > before fbcon or X kicks in, actually this affects all laptops
> > > > > with TV out, or without real port but our TV encoder probe
> > > > > fails to detect.
> > > > > 
> > > > > Signed-off-by: Zhenyu Wang <zhenyuw at linux.intel.com>
> > > > 
> > > > I think a black screen after boot and before fbcon loads sounds
> > > > like the right behavior.
> > > > 
> > > 
> > > Really? So this looks like you want to force fbcon loading for
> > > kernel, I also thought that way, but I don't know if we can do
> > > that in i915 module?
> > 
> > No, we don't need to force it.  It's up to the distro to load
> > things at the right time.
> > 
> 
> Well, yeah, you may prefer that way, but I don't think all distros do
> that automaticly. Debian doesn't do that for me after I build my own
> kernel and modules on MacBook...
> 
> We've already known that it'll be blank in the way, and we might be
> got more bug reports like
> http://bugs.freedesktop.org/show_bug.cgi?id=24021, why we can't just
> check the current hw status in load time instead of doing
> load_detect? Just realized that for older 8xx chips we also do
> load_detect for CRT, I can refresh my origin patch to not do any
> load_detect when loading.

Yeah, avoiding load detect where possible seems like a good thing.  I
just don't think we need to do it to avoid blanking the screen before
fbcon loads.

So I'm not opposed to the patch; would be cool if you can do it for
CRTs as well.

Thanks,
-- 
Jesse Barnes, Intel Open Source Technology Center