Re: [PATCH] arm64: dts: ti: k3-am69-sk: remove assigned-clock-parents for unused VP

[Jayesh Choudhary <j-choudhary@ti.com>]

Hello Tomi,

On 03/01/24 18:35, Tomi Valkeinen wrote:
> Hi,
> 
> On 21/12/2023 13:30, Jayesh Choudhary wrote:
>> VP2 and VP3 are unused video ports and VP3 share the same parent
>> clock as VP1 causing issue with pixel clock setting for HDMI (VP1).
>> So remove the parent clocks for unused VPs.
>>
>> Fixes: 6f8605fd7d11 ("arm64: dts: ti: k3-am69-sk: Add DP and HDMI 
>> support")
>> Reported-by: Nishanth Menon <nm@ti.com>
>> Closes: 
>> https://storage.kernelci.org/mainline/master/v6.7-rc6/arm64/defconfig/gcc-10/lab-ti/baseline-nfs-am69_sk-fs.txt
>> Signed-off-by: Jayesh Choudhary <j-choudhary@ti.com>
>> ---
>>
>> Local testing log for HDMI on AM69-SK:
>> <https://gist.github.com/Jayesh2000/517395cd85eb28d65b8ee4568cefb809>
>>
>>   arch/arm64/boot/dts/ti/k3-am69-sk.dts | 8 ++------
>>   1 file changed, 2 insertions(+), 6 deletions(-)
>>
>> diff --git a/arch/arm64/boot/dts/ti/k3-am69-sk.dts 
>> b/arch/arm64/boot/dts/ti/k3-am69-sk.dts
>> index 8da591579868..370980eb59b0 100644
>> --- a/arch/arm64/boot/dts/ti/k3-am69-sk.dts
>> +++ b/arch/arm64/boot/dts/ti/k3-am69-sk.dts
>> @@ -918,13 +918,9 @@ &dss {
>>       pinctrl-names = "default";
>>       pinctrl-0 = <&dss_vout0_pins_default>;
>>       assigned-clocks = <&k3_clks 218 2>,
>> -              <&k3_clks 218 5>,
>> -              <&k3_clks 218 14>,
>> -              <&k3_clks 218 18>;
>> +              <&k3_clks 218 5>;
>>       assigned-clock-parents = <&k3_clks 218 3>,
>> -                 <&k3_clks 218 7>,
>> -                 <&k3_clks 218 16>,
>> -                 <&k3_clks 218 22>;
>> +                 <&k3_clks 218 7>;
>>   };
>>   &serdes_wiz4 {
> 
> The SK has two outputs, using VP0 and VP1, so the above kind of makes 
> sense. Then again, setting up 4 clocks here really shouldn't break the 
> SK, should it? The AM69 has 4 available VPs. How does one configure the 
> clocks for a board that uses 4 VPs, or possibly a different selection of 
> VPs?

> 
> I think the patch desc should explain why this doesn't work. Afaik, the 
> dts is not wrong as such, but there's an underlying issue that breaks 
> the clocking if all four clocks are set up here.
> 

I discussed this with firmware team, there is an issue with sibling
child clocks. If parent clock is shared, DM cannot set its rate.
The determine_rate and set_rate query are behaving unexpectedly in
this case. determine_rate is returning 0 and set_rate is setting it
to 1.8G even when VP can support max 600M.

Jayesh

> So, with the desc updated, as this fixes an issue and is not wrong:
> 
> Reviewed-by: Tomi Valkeinen <tomi.valkeinen@ideasonboard.com>
> 
> But I also feel this is dodging a firmware (?) issue.
> 
>   Tomi
>