Re: [PATCH v1 6/9 RESEND] ARM: tegra: transformers: Add connector node for common trees

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [High] Migrating to the DRM bridge framework silently breaks HDMI CEC, audio, and AVI infoframes.
- [High] Strict removal of legacy `nvidia,hpd-gpio` and `nvidia,ddc-i2c-bus` breaks older kernels, while retaining them exposes a probe conflict.
--

commit fb3e359e4dd58d87afe7ba7cf548657a8457080f
Author: Svyatoslav Ryhel <clamor95@gmail.com>

ARM: tegra: transformers: Add connector node for common trees

This commit adds a micro-HDMI connector node to the ASUS Transformers
device trees to use the connector framework, now that Tegra HDMI has
bridge/connector support. It replaces the legacy hpd and ddc properties
with an endpoint connecting to the new hdmi-connector node.

> diff --git a/arch/arm/boot/dts/nvidia/tegra20-asus-transformer-common.dtsi b/arch/arm/boot/dts/nvidia/tegra20-asus-transformer-common.dtsi
> index b48f53c00efae..5b758aad8d150 100644
> --- a/arch/arm/boot/dts/nvidia/tegra20-asus-transformer-common.dtsi
> +++ b/arch/arm/boot/dts/nvidia/tegra20-asus-transformer-common.dtsi
> @@ -79,9 +79,11 @@ hdmi@54280000 {
>  			pll-supply = <&hdmi_pll_reg>;
>  			hdmi-supply = <&vdd_hdmi_en>;
>  
> -			nvidia,ddc-i2c-bus = <&hdmi_ddc>;
> -			nvidia,hpd-gpio = <&gpio TEGRA_GPIO(N, 7)
> -				GPIO_ACTIVE_HIGH>;

Does the removal of nvidia,hpd-gpio and nvidia,ddc-i2c-bus cause a regression
in device tree ABI backward compatibility for older kernels that lack
hdmi-connector bridge support?

If these legacy properties are retained to preserve compatibility, will
tegra_output_probe() unconditionally request the nvidia,hpd GPIO and conflict
with the display-connector driver requesting the same GPIO via hpd-gpios?

Since GPIOs are requested exclusively, this might result in an -EBUSY error
that fails the DRM probe entirely.

Does the DRM driver need to be patched to ignore legacy properties when a
bridge is present before applying this DT migration?

> +			port {
> +				hdmi_out: endpoint {
> +					remote-endpoint = <&hdmi_connector_in>;
> +				};
> +			};
>  		};
>  	};

Does migrating to the DRM bridge framework here cause a regression with HDMI
CEC, audio, and AVI infoframes?

When hdmi->output.bridge is used, tegra_hdmi_init() dynamically allocates a
connector, leaving the internal legacy hdmi->output.connector structure
zero-initialized.

Could this cause tegra_output_init() to skip CEC notifier registration because
connector_type evaluates to 0?

Additionally, does this cause tegra_output_is_hdmi() to return false (since
display_info.is_hdmi is 0), which would force DVI fallback mode in
tegra_hdmi_encoder_enable() by setting hdmi->dvi = true?

This fallback mode appears to bypass tegra_hdmi_setup_audio() and skip all
AVI and audio infoframe configurations.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260511074859.24930-1-clamor95@gmail.com?part=6