Hi Daniel,
On 8/29/25 6:21 PM, Daniel Stone wrote:
> Hi Cristian,
>
> On Mon, 25 Aug 2025 at 10:57, Cristian Ciocaltea
> <cristian.ciocal...@collabora.com> wrote:
>> @@ -1255,6 +1254,11 @@ struct dw_hdmi_qp *dw_hdmi_qp_bind(struct
>> platform_device *pdev,
>> return ERR_PTR(-ENODEV);
>> }
>>
>> + if (!plat_data->ref_clk_rate) {
>> + dev_err(dev, "Missing ref_clk rate\n");
>> + return ERR_PTR(-ENODEV);
>> + }
>
> This introduces another bisect cliff, as the Rockchip integration
> isn't added until patch 5/6, meaning together with the previous patch
> the driver isn't usable between patches 1-5. It would be most sensible
> I think to keep a default until the users have been fixed up. But
> maybe a better sequence for this series would be:
> * dev_err_probe() cleanup (easy, no dependencies)
> * add refclk to plat_data (populated but unused)
> * use refclk instead of hardcoded frequency in bridge driver, make it
> mandatory
> * add CEC IRQ to plat_data (populated but unused)
> * add CEC support to driver, probably make it not mandatory to provide
> CEC IRQ in DT since it doesn't seem required for correct operation?
> * enable CEC in defconfig
Yeah, this is pretty similar to how the initial series looked like. The
current sequence follows Heiko's suggestion, which I (still) think it's the
correct approach.
Both bisect issues are now fixed in v4:
https://lore.kernel.org/all/20250903-rk3588-hdmi-cec-v4-0-fa25163c4...@collabora.com/
Thanks for the review!
Regards,
Cristian
