[PATCH v1] ti: k3: abstract common fdt api for reserved mem fixups

[Tom Rini]

On Thu, May 22, 2025 at 08:39:40PM +0530, Anshul Dalal wrote:

> The usage of fdt_fixup_reserved is repeated for ATF and OP-TEE for
> multiple platforms, this patch creates a single fdt API for fixing up
> the reserved-memory node with added error handling.
> 
> All k3 platforms already share a common tispl template which ensures
> binaries are loaded as per the respective CONFIG_*_LOAD_ADDR. And the
> provided new_size for the fixup is overridden by the size from fdt node
> anyways. This allows for safe abstraction of the reserved memory fixups
> for all current platforms.
> 
> fdt_fixup_reserved now abstracts the ATF and OP-TEE fixups by calling
> the renamed static fdt_fixup_reserved_memory function with the required
> parameters.
> 
> Signed-off-by: Anshul Dalal <anshuld at ti.com>
[snip]

I think this shows the start of fixing up some problems, but that we
need to do more.

> diff --git a/arch/arm/mach-k3/am62ax/am62a7_fdt.c b/arch/arm/mach-k3/am62ax/am62a7_fdt.c
> index 7f764ab36b5..9a4599432ff 100644
> --- a/arch/arm/mach-k3/am62ax/am62a7_fdt.c
> +++ b/arch/arm/mach-k3/am62ax/am62a7_fdt.c
> @@ -10,8 +10,5 @@
>  
>  int ft_system_setup(void *blob, struct bd_info *bd)
>  {
> -	fdt_fixup_reserved(blob, "tfa", CONFIG_K3_ATF_LOAD_ADDR, 0x80000);
> -	fdt_fixup_reserved(blob, "optee", CONFIG_K3_OPTEE_LOAD_ADDR, 0x1800000);
> -
> -	return 0;
> +	return fdt_fixup_reserved(blob);
>  }

This is good.

> diff --git a/arch/arm/mach-k3/am62px/am62p5_fdt.c b/arch/arm/mach-k3/am62px/am62p5_fdt.c
> index 2c40fa5a594..9f4b1663864 100644
> --- a/arch/arm/mach-k3/am62px/am62p5_fdt.c
> +++ b/arch/arm/mach-k3/am62px/am62p5_fdt.c
> @@ -92,8 +92,5 @@ int ft_system_setup(void *blob, struct bd_info *bd)
>  	fdt_fixup_canfd_nodes_am62p(blob, k3_has_canfd());
>  	fdt_fixup_thermal_zone_nodes_am62p(blob, k3_get_max_temp());
>  	fdt_fixup_cpu_freq_nodes_am62p(blob, k3_get_a53_max_frequency());

Is it really OK for any of the above fixups to fail to happen and be
silently ignored? fdt_fixup_thermal_zone_nodes_am62p() for example looks
like it throws away error conditions from functions it calls. And other
boards are doing similar things.

[sniip]
> +int fdt_fixup_reserved(void *blob)
> +{
> +	int ret;
> +
> +	ret = fdt_fixup_reserved_memory(blob, "tfa", CONFIG_K3_ATF_LOAD_ADDR,
> +					0x80000);
> +	if (ret)
> +		return ret;

Good.

> +	ret = fdt_fixup_reserved_memory(blob, "optee",
> +					CONFIG_K3_OPTEE_LOAD_ADDR, 0x1800000);
> +	if (ret)
> +		return ret;
> +
> +	return 0;

But we can just return fdt_fixup_reserved_memory(...) here.

-- 
Tom
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 659 bytes
Desc: not available
URL: <https://lists.denx.de/pipermail/u-boot/attachments/20250604/cc4199d8/attachment.sig>