Re: [PATCH v20 08/13] media: qcom: camss: vfe: Add support for VFE 1080

[hangxiang . ma]

On 10/11/26 4:56 PM, Bryan O'Donoghue <bryan.odonoghue@xxxxxxxxxx> wrote:
> On 10/10/2026 17:01, Hangxiang Ma wrote:
> > +        if (!vfe->res->reg_update_after_csid_config)
> > +            ops->reg_update(vfe, line->id);
>
> So I'd prefer to have a discussion about if we can move all of the 
> updates to one method.
>
> i.e. why should any of the updates happen before csid ?
>
> Sure vfe 1080's documentation makes some sort of explicit statement 
> about an update that has been implied for some time.
>
> My question is why not gate all of the updates this way and make this 
> not a flag but a global behaviour ?
>
> ---
> bod

Yes Bryan. That's a fair point.

The reason I introduced the flag is that on VFE 1080 the documentation explicitly states that the register update should happen after the CSID configuration and it really follows the guidance, while older generations appear to have relied on the previous ordering implicitly.

That said, I agree that if there is no functional requirement for some updates to happen before CSID configuration, having a single update path would be cleaner and easier to maintain. The flag was mainly intended to minimize behavior changes for existing targets until we have a better understanding of the dependency.

I'll take another look at the update sequence and check whether other older platforms can safely move to the same post-CSID update flow. My colleagues who have such resource maybe can help me on this.  I can remove the platform-specific flag entirely and make this the common behavior if that holds.

In this way, I think I need to raise a new series to collect changes among other CSID genN part and VFE framework part. Could I mark it as the dependency as this KNP series? Or functional dependency is enough? Thanks.

---
Best Regards,
Hangxiang