[PATCH 1/1] ARM: dts: don't make DP a consumer of DISP1 on Exynos5250

[Javier Martinez Canillas]

Hello,

On 03/16/2015 01:15 PM, Andreas Färber wrote:
> Am 14.03.2015 um 09:11 schrieb Javier Martinez Canillas:
>> By making the DP controller a consumer of DISP1, the PD is powered
>> off when the exynos-dp probe is deferred and powered on again when
>> the exynos-drm driver is probed.
>> 
>> But this causes the exynos-dp driver failing to obtain the stream
>> clock since the FIMD has been powered off with the DISP1 PD:
>> 
>> exynos-dp 145b0000.dp-controller: Timeout of video streamclk ok
>> exynos-dp 145b0000.dp-controller: unable to config video
>> 
>> The Exynos5250 documentation doesn't mention that the Display Port
>> Transmitter module is included in the DISP1 PD so the device should
>> not have a reference to this Power Domain.
>> 
>> This patch fixes video display on an Exynos5250 Snow Chromebook.
>> 
>> Fixes: 2d2c9a8d0a4f ("ARM: dts: add display power domain for exynos5250")
>> Signed-off-by: Javier Martinez Canillas <javier.martinez at collabora.co.uk>
>> ---
>> 
>> Hello Andrzej and Marek,
>> 
>> I need this patch to have display working on an Snow Chromebook
> 
> Tested-by: Andreas Färber <afaerber at suse.de>
> 
> This fixes the display on the Spring Chromebook as well!
>

Thanks for testing Andreas but this patch seems to only solve a symptom
rather than the root cause.

Since the error happens again when disabling and enabling the display:

# echo 1 > /sys/devices/platform/exynos-drm/graphics/fb0/blank
disp1-power-domain: Power-off latency exceeded, new value 225333 ns
# echo 0 > /sys/devices/platform/exynos-drm/graphics/fb0/blank
exynos-dp 145b0000.dp-controller: EDID data does not include any extensions.
exynos-dp 145b0000.dp-controller: EDID Read success!
exynos-dp 145b0000.dp-controller: Link Training Clock Recovery success
exynos-dp 145b0000.dp-controller: Link Training success!
exynos-dp 145b0000.dp-controller: Timeout of video streamclk ok
exynos-dp 145b0000.dp-controller: unable to config video

So what happens is that if the DISP1 power domain is powered off and
then powered on again, the display fails to be enabled.

Making the dp controller a consumer of the DISP1 pd only made this to
trigger on boot since the driver probe was deferred and the DISP1 pd
was turned off and on again on exynos-drm probe.

It seems that the kernel is not enabling everything that is needed for
the power domain and it is working on boot just because the bootloader
initialized everything properly.

This is similar to the problem we had in Exynos5420 and that was fixed
by Andrzej in the series "Fix power domains handling on exynos542x" [0].
So probably a similar solution is needed.

Andrzej,

I looked at the Exynos5250 manual and I didn't find anything obvious
that is missing in the DISP1 pd dev node but the asynchronous bridges
clocks needed on Exynos5420 was also not well documented so I don't
know if I'm missing something.

Do you know what could be missing here? Otherwise I think your patch
to add the DISP1 pd in the DT should be reverted to have display
working again on Exynos5250 boards.
 
> Thanks,
> Andreas
> 

Best regards,
Javier

[0]: https://lkml.org/lkml/2015/3/12/387