Re: [PATCH v4 7/8] media: qcom: iris: split firmware_data from raw platform data

[Dmitry Baryshkov]

On Fri, Mar 13, 2026 at 01:19:21PM +0530, Dikshita Agarwal wrote:

I'm sorry, I've refreshed the series before receiving this email. I will
send new iteration after settling the discussion here.

> On 3/13/2026 9:00 AM, Dmitry Baryshkov wrote:
> > Having firmware-related fields in platform data results in the tying
> > platform data to the HFI firmware data rather than the actual hardware.
> > For example, SM8450 uses Gen2 firmware, so currently its platform data
> > should be placed next to the other gen2 platforms, although it has the
> > VPU2.0 core, similar to the one found on SM8250 and SC7280 and so the
> > hardware-specific platform data is also close to those devices.
> > 
> > Split firmware data to a separate struct, separating hardware-related
> > data from the firmware interfaces.
> > 
> > Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@xxxxxxxxxxxxxxxx>
> > ---
> >  drivers/media/platform/qcom/iris/iris_buffer.c     |  84 +++----
> >  drivers/media/platform/qcom/iris/iris_core.h       |   1 +
> >  drivers/media/platform/qcom/iris/iris_ctrls.c      |   8 +-
> >  .../platform/qcom/iris/iris_hfi_gen1_command.c     |  10 +-
> >  .../platform/qcom/iris/iris_hfi_gen2_command.c     |  66 ++---
> >  .../platform/qcom/iris/iris_platform_common.h      |  79 +++---
> >  .../media/platform/qcom/iris/iris_platform_gen1.c  |  68 +++---
> >  .../media/platform/qcom/iris/iris_platform_gen2.c  | 268 +++++++--------------
> >  drivers/media/platform/qcom/iris/iris_probe.c      |   3 +-
> >  drivers/media/platform/qcom/iris/iris_vidc.c       |  10 +-
> >  10 files changed, 246 insertions(+), 351 deletions(-)
> > 
> 
> <snip>
> 
> > diff --git a/drivers/media/platform/qcom/iris/iris_platform_common.h b/drivers/media/platform/qcom/iris/iris_platform_common.h
> > index d1daef2d874b..1a870fec4f31 100644
> > --- a/drivers/media/platform/qcom/iris/iris_platform_common.h
> > +++ b/drivers/media/platform/qcom/iris/iris_platform_common.h
> > @@ -201,45 +201,16 @@ 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);
> 
> I still don't think it's right to keep vpu_buffer_size in firmware data as
> this would change mostly for every new VPU variant.
> 
> The buffer sizing logic depends on VPU generation (vpu2, vpu3, vpu33,
> vpu35) / SoC constraints, not on whether the HFI is Gen1 vs Gen2.

Okay, so how do we solve the SC7280 Gen1 vs Gen2 situation? I can keep
the function pointer in struct iris_platform_data for now, letting you
sort it out in your series.

> 
> <snip>
> 
> > diff --git a/drivers/media/platform/qcom/iris/iris_platform_gen2.c b/drivers/media/platform/qcom/iris/iris_platform_gen2.c
> > index 10a972f96cbe..a83f6910f8b7 100644
> > --- a/drivers/media/platform/qcom/iris/iris_platform_gen2.c
> > +++ b/drivers/media/platform/qcom/iris/iris_platform_gen2.c
> > @@ -906,41 +906,15 @@ static const u32 sm8550_enc_op_int_buf_tbl[] = {
> >  	BUF_SCRATCH_2,
> >  };
> >  
> > -const struct iris_platform_data sm8550_data = {
> > +const struct iris_firmware_data iris_hfi_gen2_data = {
> >  	.init_hfi_ops = iris_hfi_gen2_sys_ops_init,
> >  	.get_vpu_buffer_size = iris_vpu_buf_size,
> > -	.vpu_ops = &iris_vpu3_ops,
> > -	.icc_tbl = sm8550_icc_table,
> > -	.icc_tbl_size = ARRAY_SIZE(sm8550_icc_table),
> > -	.clk_rst_tbl = sm8550_clk_reset_table,
> > -	.clk_rst_tbl_size = ARRAY_SIZE(sm8550_clk_reset_table),
> > -	.bw_tbl_dec = sm8550_bw_table_dec,
> > -	.bw_tbl_dec_size = ARRAY_SIZE(sm8550_bw_table_dec),
> > -	.pmdomain_tbl = sm8550_pmdomain_table,
> > -	.pmdomain_tbl_size = ARRAY_SIZE(sm8550_pmdomain_table),
> > -	.opp_pd_tbl = sm8550_opp_pd_table,
> > -	.opp_pd_tbl_size = ARRAY_SIZE(sm8550_opp_pd_table),
> > -	.clk_tbl = sm8550_clk_table,
> > -	.clk_tbl_size = ARRAY_SIZE(sm8550_clk_table),
> > -	.opp_clk_tbl = sm8550_opp_clk_table,
> > -	/* Upper bound of DMA address range */
> > -	.dma_mask = 0xe0000000 - 1,
> > -	.fwname = "qcom/vpu/vpu30_p4.mbn",
> 
> Should fw_name be in firmware_data? as this can be change based on HFI
> versions?

That would fail because then each device will have to gain its own
struct iris_firmware_data.

But... Maybe we can do something as simple as:

struct iris_firmware_desc {
	const char *fwname;
	u32 (*get_vpu_buffer_size)(struct iris_inst *inst, enum iris_buffer_type buffer_type);
	bool (*checK_fw_match)(u8 *buf, size_t size);
	const struct iris_firmware_data *data;
};

and then

struct iris_platform_data {
	struct iris_firmware_desc *gen1, *gen2;
	// .. the rest as usual;
};


struct iris_core {
	u32 (*get_vpu_buffer_size)(struct iris_inst *inst, enum iris_buffer_type buffer_type);
	const struct iris_firmware_data *data;
	// ... the rest as expected
};

During first open the driver will try loading firmware from DT and
identifying it using the check_fw_match() callback. If DT doesn't have
firmware-name the driver will try loading gen2 and, if not found, gen1.
When firmware loading succeeds, it will set the pointer and the callback
in iris_core, settling the interface between the driver and the
firmware.

WDYT?

> > -const struct iris_platform_data sm8650_data = {
> > +const struct iris_firmware_data iris_hfi_gen2_vpu33_data = {
> 
> This proves my above point.
> 
> iris_hfi_gen2_data and iris_hfi_gen2_vpu33_data become identical except for
> get_vpu_buffer_size, which forces us to create multiple “firmware_data”
> variants just to carry a hardware-specific difference.
> 
> Also, it will scale poorly going forward. When we introduce vpu4 /
> vpu5–based platforms, we would need to add more copies of essentially the
> same HFI Gen2 firmware_data, differing only in the buffer sizing callback.

Yes. I'm not sure, if there is any difference between params / caps as
implremented by the firmware for those generations.

-- 
With best wishes
Dmitry