[meta-freescale] HDMI on SABRE Lite (was FlexCAN not working)

[Eric Nelson]

Hi Jose,

On 10/14/2013 02:40 AM, Jose Mª Ferreiro wrote:
>
> El 12/10/2013 0:11, Eric Nelson escribió:
>> On 10/11/2013 11:55 AM, Jose Mª Ferreiro wrote:
>> <snip>
>>>
>>> You are correct. I didn´t look well for it. It seems as it depends on a
>>> third GPIO signal to select one transceiver or the other. We will test
>>> it on monday.
>>>
>>> I would like to use linux-boundary, wich was the default, but we have to
>>> change to linux-imx due to video issues that did not work on
>>> linux-boundary: something about mfw_isink i think.
>>>
>>
>> Can you elaborate? I'm not aware of any board-specific code in
>> mfw_isink, and certainly not in the Nitrogen6x code.
>
> Hi and sorry,  Eric
>
> I was wrong. We were testing several things at the time. The problem was
> not with mfw_isink. The problem was with HDMI. We have a Samsung
> S22B300H monitor connected to our Sabrelite board through HDMI.
>
Okay, so we switched topics here.

> When linux-boundary is starting it shows many hdmi messages until it
> gets hanged when trying to start XServer. At both resolutions: 1920x1080
> and 1280x720 (attached startup log). With linux-imx, it starts without
> problems and the image is good.
>

 From the log, it appears that the HDMI driver is detecting a disconnect
and re-connect.

If you're not seeing this with the fslc kernel, the problem must stem
from this patch, which added support for DVI monitors into the
HDMI driver:
	https://github.com/boundarydevices/linux-imx6/commit/c0439e262bb6c23887d96466b2ab7916aa0488d9

> If we connect a DVI monitor to the HDMI SabreLite output we can work
> with linux-boundary at 1280x 720 but we have no image with 1920x1080.
>

That's odd. We put this patch in place to support DVI. On the older
fslc kernel, most DVI monitors would fail the detect logic as described
here:
	http://boundarydevices.com/dvi-support-on-i-mx6-boards/

Are you in a position to revert the patch above?

Do you see the Linux penguins during the boot process?

If so, I've been wanting to test a patch to ignore disconnect
events and this may be a good time to do so. Since many of our
customers are operating signage apps and the like, it's useful
to simply force detection and be done with the process.

Please advise,


Eric