[meta-freescale] No HDMI video on imx6qsabresd with dizzy

Andreas Müller schnitzeltony at googlemail.com
Thu Nov 13 07:48:11 PST 2014

On Thu, Nov 13, 2014 at 4:17 PM, Nikolay Dimitrov <picmaster at mail.bg> wrote:
> Hi Andreas,
> On 11/13/2014 11:29 AM, Andreas Müller wrote:
>> On Wed, Nov 12, 2014 at 11:58 PM, Nikolay Dimitrov <picmaster at mail.bg>
>> wrote:
>>> Hi guys,
>>> I'm trying to use unmodified dizzy build with imx6q sabresd, for
>>> testing video playback. U-Boot outputs HDMI image during boot, but in
>>> Linux there's no HDMI image (no console, no video playback).
>>> The image is "fsl-image-multimedia-full", kernel 3.10.17. The device
>>> files /dev/video* and /dev/fb* are available, but gstreamer playback
>>> doesn't output any image on the HDMI port (seems to be stuck instead of
>>> playing).
>>> The kernel cmdline doesn't specify any video settings, which typically
>>> is an issue, but when I add the video bootargs:
>>> ...video=mxcfb0:dev=hdmi,1024x768M at 60,if=RGB24...
>>> The board stops to print any boot messages after "Starting kernel..."
>>> and seems to be stuck. Removing the video bootargs allows again to boot
>>> normally.
>>> I must admit that I'm somewhat puzzled by this behavior, as I already
>>> have daizy running with 3.10.17 on a custom board, and the video is
>>> working there with almost the same bootargs. Do you guys have any ideas
>>> for how to configure imx6q sabresd for HDMI video?
>>> Thanks in advance. Regards,
>>> Nikolay
>>> --
>> Had same here.
>> I think there are two problems:
>> * 3.10.17 EDID decoding works only for monitors with EDID extended
>> blocks (patch 0001..)
>> * 1024x768 is not a CEA mode. If I read the code of 3.10.17 correct,
>> HDMI supports only modes which are found in
>> drivers/video/mxc/mxc_edid.c const struct fb_videomode
>> mxc_cea_mode[64]. I hacked up working solution (patch 0002..). Note
>> that this is just a hack - it might lead to incorrect colour space
>> conversion or wrong PixelRepetitionOutput (see mxc_hdmi_setup:
>> hdmi->vic is always 0 for my patches).
>> These patches were tested with two monitors (1280x1024 / 1650x<I
>> forgot>) and seem to work fine.
> Thanks for your detailed answer. I remember seeing discussions on the
> net about a safe fallback resolution for imx6 hdmi, so that's the only
> reason to choose 1024x768.
> Also, are you sure that the HDMI code doesn't use the CVT (Common VESA
> Table) resolutions? Need to check this. I'll also look a the code
> which you pointed me to, and try your patches.
> I'm not looking for big or pixel-exact resolutions on HDMI, just some
> reasonable resolution so I can test my gstreamer issues there :).
> Regards,
> Nikolay
After re reading this thread I saw that my patches are addressing a
different problem. In short: Whatever monitor I connected to HDMI (via
DVI) I got only 640x480 and complaints like

[    0.829299] mxc_hdmi 20e0000.hdmi_video: mxc_edid_read_internal
[    0.960509] mxc_hdmi 20e0000.hdmi_video: Read EDID again
[    0.960522] mxc_hdmi 20e0000.hdmi_video: mxc_edid_read_internal
[    1.091703] mxc_hdmi 20e0000.hdmi_video: No modes read from edid
[    1.091714] mxc_hdmi 20e0000.hdmi_video: create default modelist


More information about the meta-freescale mailing list