[PATCH] rockchip: rk3328: Fix Synchronous Abort on usb start

[Jonas Karlman]

Hi Alex,

On 2025-05-24 16:20, Alex Shumsky wrote:
> Remove incorrect assigned-clock-parents in rk3328-u-boot.dtsi.
> This incorrect parent has become a problem since recent commit with parent
> clock enablement.
> 
> Currently assigned-clock-parents is not a CLK node, but rk3328-usb2phy.
> It looks like a mistake in dts made a long time ago (~8 years ago).

No, this is very much correct, u2phy is a clk provider as can be seen
by the use of the '#clock-cells' prop so being able to use it as a
parent clk is allowed and the device tree is correct.

> 
>   => usb start
>   starting USB...
>   Bus usb at ff580000: "Synchronous Abort" handler, esr 0x96000210, far 0x4
> 
> Linux dts source reference:
> https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/commit/?id=c60c0373a5e85d8bd0bb026cd5440576249d2299
> 
> Fixes: ac30d90f3367 ("clk: Ensure the parent clocks are enabled while reparenting")
> Signed-off-by: Alex Shumsky <alexthreed at gmail.com>
> ---
> 
>  arch/arm/dts/rk3328-u-boot.dtsi | 4 ++++
>  1 file changed, 4 insertions(+)
> 
> diff --git a/arch/arm/dts/rk3328-u-boot.dtsi b/arch/arm/dts/rk3328-u-boot.dtsi
> index b0e50a973a..39dcc2313b 100644
> --- a/arch/arm/dts/rk3328-u-boot.dtsi
> +++ b/arch/arm/dts/rk3328-u-boot.dtsi
> @@ -150,6 +150,10 @@
>  	bootph-all;
>  };
>  
> +&u2phy {
> +	/delete-property/ assigned-clock-parents;
> +};

We should fix the driver code and not modify the device tree to fit the
code.

The u2phy ofnode is tied to multiple uclass drivers so is it possible
that the parent clock lookup from commit ac30d90f3367 ("clk: Ensure the
parent clocks are enabled while reparenting") is not getting the
UCLASS_CLK driver?

Regards,
Jonas

> +
>  #ifdef CONFIG_ROCKCHIP_SPI_IMAGE
>  &binman {
>  	simple-bin-spi {