Re: [PATCH v2 10/11] media: qcom: iris: split firmware_data from raw platform data

[Dmitry Baryshkov]

On Wed, Mar 11, 2026 at 10:39:54AM +0530, Dikshita Agarwal wrote:
> 
> 
> On 3/11/2026 5:53 AM, Dmitry Baryshkov wrote:
> > On Fri, Mar 06, 2026 at 04:28:48PM +0530, Dikshita Agarwal wrote:
> >>
> >>
> >> On 2/28/2026 8:56 PM, Dmitry Baryshkov wrote:
> >>
> >>> diff --git a/drivers/media/platform/qcom/iris/iris_platform_common.h b/drivers/media/platform/qcom/iris/iris_platform_common.h
> >>> index a936ab2aedaf..be3449861993 100644
> >>> --- a/drivers/media/platform/qcom/iris/iris_platform_common.h
> >>> +++ b/drivers/media/platform/qcom/iris/iris_platform_common.h
> >>> @@ -201,9 +201,32 @@ enum platform_pm_domain_type {
> >>>  	IRIS_APV_HW_POWER_DOMAIN,
> >>>  };
> >>>  
> >>> -struct iris_platform_data {
> >>> +struct iris_firmware_data {
> >>>  	void (*init_hfi_ops)(struct iris_core *core);
> >>> +
> >>>  	u32 (*get_vpu_buffer_size)(struct iris_inst *inst, enum iris_buffer_type buffer_type);
> >>
> >> Makes sense to split firmware-related data out of iris_platform_data into a
> >> dedicated iris_firmware_data, one comment though, get_vpu_buffer_size
> >> should be in iris_platform_data since it describes the VPU’s buffer
> >> requirements independent of whether the SoC runs gen1/gen2 HFI firmware.
> > 
> > Not really... Your patchset to bringup Gen2 on Kodiak points out that
> > this callback depends on the firmware: we currently use
> > iris_vpu_buf_size() for Gen1, your patch have used iris_vpu33_buf_size()
> > with an explicit comment that it requires a bigger buffer.
> 
> Agreed. However, that case is somewhat unique because the firmware code
> changed significantly between the Gen1 and Gen2 releases. In general, the
> buffer size is determined primarily by the VPU architecture, as well as by
> the features being supported—especially when specific features require
> updated buffer allocations.

I'd still disagree. In my opinion in your patchset we should have a
single sc7280 platform data (because in the end it's a single platform)
and two firmware interfaces associated with it (one for Gen1, one for
Gen2). So, get_vpu_buffer_size() callback will have to go to the
firmware interface.

> 
> Thanks,
> Dikshita
> 
> > 
> > 

-- 
With best wishes
Dmitry