Re: [PATCH v3 0/5] Introduce "non-pixel" sub node within iris video node

[Krzysztof Kozlowski <krzk@kernel.org>]

On 09/10/2025 11:52, Vikash Garodia wrote:
> 
> On 10/9/2025 6:34 AM, Krzysztof Kozlowski wrote:
>> On 09/10/2025 09:55, Bryan O'Donoghue wrote:
>>> On 09/10/2025 01:47, Krzysztof Kozlowski wrote:
>>>>> Maybe it would be possible to also use an inferred FUNCTION_ID somehow
>>>>> though TBH I think that's a work-around.
>>>> Three months ago I gave you the answer for that - it is inferred by
>>>> index on the list.
>>>
>>> But at least as I understand it, you can have multiple SID entries that 
>>> need to map to a FUNCTION_ID which means you need to encode that 
>>> inferred indexing in your driver.
>>
>> Yes.
>>
>>>
>>> So you can't have the iommu code just know what to do.. it has to be 
>>> driver specific.
>>
>> Yes.
>>
>>>
>>> The iommu description for this platform basically lacks the data that 
>>> _should_ be there -> FUNCTION_ID.
>>
>> No. The index tells that already.
> 
> Hardware1 can have 2 iommus entries for FUNCTION_ID_1, while Hardware2 can have
> 3 iommus entry. It would be good to keep this info in devicetree, as it is still
> hardware specific (function_id is the hardware identifier).
> Defaulting the index would imply, the common device driver would need to book

You don't "defaulting the index" but control the indices. All what you
wrote is just irrelevant.

> keep the iommus and hardware id in driver for every SOC. Any mismatch in order
> in DT and driver table would result in error too.

It's irrelevant. You write buggy code, you are responsible for it.

> 
> More importantly, this need to be agreed by IOMMU maintainer to extend the
> iommus property.
Luckily they are not hiding...

Best regards,
Krzysztof