Re: [PATCH v2 01/11] dt-bindings: crypto: qcom,ice: Allow power-domain and iface clk

[Harshal Dev <harshal.dev@oss.qualcomm.com>]

Hi Krzysztof,

On 3/16/2026 4:39 PM, Krzysztof Kozlowski wrote:
> On 16/03/2026 11:56, Harshal Dev wrote:
>> Hello Krzysztof,
>>
>> On 3/13/2026 9:28 PM, Krzysztof Kozlowski wrote:
>>> On 13/03/2026 12:45, Harshal Dev wrote:
>>>>> Except new devices, like Eliza and Milos. And then this should go to
>>>>> current fixes.
>>>>
>>>> I'm not sure if I understand correctly, do you mean to say that except for Eliza
>>>> and Milos, new devices need to change their DT binding to 'required' with
>>>> corresponding DTS changes. And then, the patch updating the DT binding also needs
>>>> to be back-ported?
>>>
>>> No. All new devices must require this. You only preserve released ABI,
>>> so fix unreleased ABI (Eliza and Milos) now, before it gets released.
>>>
>>
>> I'm already being annoying, but I will disturb you one more time for clarification. :)
>>
>> By saying 'fix unreleased ABI now' do you mean to say that I should add another
>> trailing commit at the end of this patch series which marks these resources as
>> 'required' in the DT-binding without carrying the 'Fixes' tag? Specifically so that
>> Eliza and Milos carry this constrain.
> 
> Please post a v3 of this patch for crypto subsystem, doing what you did
> here plus requiring these clocks for Eliza and Milos, with explanation
> why this is a fix thus why this should go to current cycle.
>

Got it now, I will add additional bindings which make the clock and power-domain mandatory
for Eliza (unreleased ABI) while keeping them optional for others (released ABI).

But I feel I should add this as a separate commit which carries a Fixes tag for this:
https://lore.kernel.org/all/20260223-eliza-bindings-crypto-ice-v1-1-fc76c1a5adce@oss.qualcomm.com/

This commit with it's current Fixes tag is required to be back-ported on LTS branches because
they fail to boot for LeMans and Kodiak.

Let me know if that's not good, and I should add the additional binding in the same
commit.
 
> 
>>
>> From what I understood from Bjorn's comment, the DTS and ICE driver sources will reach
>> from different trees and either could be merged first. To maintain bisectability we
> 
> 
> They are applied to different trees. Period. This defines everything,
> you cannot fix it post factum, you cannot fix bisectability afterwards.

Ack.

> 
>> should first merge this patch series followed by a subsequent patch which marks these
>> resources as 'required' in the DT-binding along with accompanying ICE driver source
> 
> Then you have a released ABI and you cannot change it, so what does this
> achieve?

Ack.

> 
> Just look what is merged where and you will see the differences. I don't
> see Milos crypto engine in current cycle, do you?

I don't either. I believe only Eliza has been added till now, I'll re-base and add only
for Eliza right now. Subsequent patches for Milos will know where the entry needs to be
added.

> 
> And bisectability has nothing to do here. You need to fix ABI before it
> gets released.

Got it. Thanks!

Regards,
Harshal

> 
>> changes which fail probe when 'iface' clk isn't available. Of course, the subsequent
>> patch will not be back-ported as a fix.
> 
> 
> 
> 
> Best regards,
> Krzysztof