[PATCH v2 08/11] clk: qcom: dispcc: Add support for display clock controller Kaanapali

Konrad Dybcio konrad.dybcio at oss.qualcomm.com
Thu Dec 4 01:34:58 PST 2025


On 12/4/25 8:14 AM, Taniya Das wrote:
> 
> 
> On 12/1/2025 6:50 PM, Konrad Dybcio wrote:
>> On 11/26/25 1:09 AM, Dmitry Baryshkov wrote:
>>> On Tue, Nov 25, 2025 at 11:15:17PM +0530, Taniya Das wrote:
>>>> Support the clock controller driver for Kaanapali to enable display SW to
>>>> be able to control the clocks.
>>>>
>>>> Signed-off-by: Taniya Das <taniya.das at oss.qualcomm.com>
>>>> ---
>>
>> [...]
>>
>>>> +/* 257.142858 MHz Configuration */
>>>
>>> This is a bit strange frequency for the boot config.
> 
> The lowest PLL configuration is used as boot config based on the
> MDP_CLK_SRC clock requirement. The PLLs on Kaanapali can support these
> lower frequencies as well.
> 
> 
>> The frequency map lists this odd cookie as the lowest predefined config,
>> perhaps it was pulled from there.
>>
> 
> Correct Konrad.
> 
>> More interestingly, the only consumer of this PLL (MDP_CLK_SRC) makes no
>> effort to use the m/n/d registers, instead relying on the PLL to re-clock
>> for its ratesetting with a fixed divider of 3 (and div1 @ XO rate).
>>
> 
> The m/n is not preferred in the cases where the PLL needs to slew to
> derive a new VCO frequency. That is the reason to keep the divider
> constant as much as possible to derive a particular frequency.

OK this is roughly what I expected, thanks for the explanation.

> 
>> 257.142858 * 3 = 771.428574 over-drives MDP_CLK_SRC, FWIW.
>>
> 
> The lowest frequency requirement is 85.7MHz and the frequency is derived
> using
> 257.142858 (PLL VCO) / 3 (RCG Div) = 85.714286 MHz
> 
> there is no over-drive at RCG of MDP.

Sorry, you're obviously right, I don't know how I got it backwards..

>> Taniya, we've seen something like this in camera too. Is there a reason
>> the frequency is being set this way?
>>
> 
> We start with the lowest frequency to configure the PLL and frequency
> requirements are decided based on usecases.

I meant the rate-change-through-PLL-reclocking, but you've answered
that above

Konrad



More information about the linux-arm-kernel mailing list