Re: [RESEND] drm/i915/hdmi: Poll for 200 msec for TMDS_Scrambler_Status

["Ville Syrjälä" <ville.syrjala@linux.intel.com>]

On Thu, Mar 26, 2026 at 06:41:12AM +0100, Jerome Tollet wrote:
> From: Jerome Tollet <jerome.tollet@gmail.com>
> 
> Hi Ankit, Ville,
> 
> Following up on this patch.
> 
> As previously reported, the SCDC polling approach fixes the 4K@60Hz HDMI
> issue on my Alder Lake-N (N100) + Cisco Desk Pro setup, and I have not
> seen regressions in testing.
> 
> Ville, do you see any remaining concern with this approach, or is there
> anything else you would like me to test to help move this forward?

My main concern is having the synchronous polling appear to fix
something now, and if we later make it asynchronous the original
problem may re-appear.

That said, after perusing the spec a bit, it looks to me that sinks
should be able sync up after at most two video fields. I suppose
having effectively two extra vblank waits in a full modeset sequence
isn't the end of world. The extra complexity of asynchronous polling
might not even be worth the hassle.

But the commit message should still highlight that (AFAICS) synchronous
polling is not required by the spec, and thus the fact that this fixes
some monitors is sort of an accidental byproduct.

-- 
Ville Syrjälä
Intel