[Intel-gfx] [PATCH 16/20] drm/i915: Ensure that while(INREG()) are bounded (v2)

[Chris Wilson]

On Wed, 22 Sep 2010 17:26:45 +0200, "Carlos R. Mafra" <crmafra2 at gmail.com> wrote:
> On Mi 22.Sep'10 at 15:29:56 +0100, Chris Wilson wrote:
> > So we detect a rogue SVideo connection and do not have a free crtc for the
> > valid VGA output. The implication is that the TV detection is not working,
> > can you add a msleep(30) to intel_tv.c::intel_tv_detect_type() after the
> > POSTING_READ:
> 
> I did this.
> Now there is an image in VGA output, but it detected the wrong resolutions:

Ok, this confirms that the TV DAC state change detection needs a little
more refinement. I'd like to be able to do away with the sleep there, but
there really may be no other method to ensure the PLLs are stable before
enabling the detector. Alas.

A slight change in behaviour is that if we already have two outputs lit,
we cannot perform a load detection on the TV connector and so we report it
as unknown (if forced). In theory this should allow the user to reconfigure
the outputs in fewer steps (and certainly not prevent valid changes).

However, it appears VGA EDID probing is still broken. The crt_ddc_pin fix
worked-for-me, however my Crestline box is still not happy, so I am not
yet able to confirm whether or not its VGA is working.
-Chris

-- 
Chris Wilson, Intel Open Source Technology Centre