Re: [PATCH v2 1/3] dt-bindings: lgm-dma: Added intel,dma-sw-desc property.

[Krzysztof Kozlowski <krzk@kernel.org>]

On 15/08/2025 09:07, Yi xin Zhu wrote:
> Hi Krzysztof,
> 
> On 08/08/2025 14:10, Krzysztof wrote:
>>
>>
>> Where is the changelog?
>>
> Sorry for missing the changelog.  I will include it in V3 patch.
> 
>>> +  intel,dma-sw-desc:
>>> +    type: boolean
>>> +    description:
>>> +      Indicates that the DMA driver should operate in 
>>> + software-managed
>> mode.
>>> +      If this property is not present, it implies that DMA 
>>> + descriptors are
>> managed and generated by another hardware component that controls the 
>> DMA engine.
>>
>> Nothing improved:
>> 1. Still wrongly wrapped.
>> 2. Still SW property.
>>
>> I don't think you read the feedback. You need to describe hardware, not SW.
>>
>> Best regards,
>> Krzysztof
> 
> Could you please give me some advice with regarding to the device tree attribute?
> My understanding is if I can create a phandle link to describe the DMA descriptors are provided by some HW component.  
> That follows the comments. Something like this:
> 
> intel,dma-desc-provider = <&xxx_hw>;
> 
> Then it describes some HW component manages the DMA descriptors.


It is not sufficient to tell. Creating phandles representing fake
hardware is still not hardware description.

> Without it, it means SW manage the DMA descriptors.
> Please correct me if my understanding is wrong.


Yes, it is wrong, because you focus on syntax instead of on the hardware
problem or hardware characteristic.

> 
> However, in the existing driver,  HW manages descriptor mode is the default mode.
> I can't create such link in the device tree without breaking the devicetree ABI.
> So, is it okay I create an attribute like below ?
> 
> Intel,dma-desc-provider = "CPU";   // it means CPU manage DMA descriptors.
> 
> Intel,dma-desc-provider = "XXX_HW";  // it means XX_HW manage DMA descriptors.


And how is it not a OS policy? Bring reasoning and arguments, otherwise
I cannot help you. Your reply above has zero hardware-related arguments,
zero facts, zero hardware description.

Best regards,
Krzysztof