Re: [PATCH v4 5/5] arm64: versal-net: Switch Versal NET to firmware clock interface

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] Device tree binding headers are incorrectly placed in arch/arm64/boot/dts/xilinx/ and duplicate existing bindings.
- [Medium] Use of #undef directives in device tree binding headers to remove inherited definitions.
--

--- Patch [5]: [PATCH v4 5/5] arm64: versal-net: Switch Versal NET to firmware clock interface ---
commit 768ce0b0d6c0eff83d929d6530cf30c0e1e4baaa
Author: Michal Simek <michal.simek@amd.com>

arm64: versal-net: Switch Versal NET to firmware clock interface

Switch Versal NET from using fixed clocks to the firmware-based clock
interface. Add DT macro headers for Versal NET and base Versal clocks,
power domains, and mandatory resets required by the clock dtsi.

> diff --git a/arch/arm64/boot/dts/xilinx/xlnx-versal-clk.h b/arch/arm64/boot/dts/xilinx/xlnx-versal-clk.h
> new file mode 100644
> index 0000000000000..d0c4abf78f307
> --- /dev/null
> +++ b/arch/arm64/boot/dts/xilinx/xlnx-versal-clk.h
> @@ -0,0 +1,123 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2019 - 2022, Xilinx, Inc.
> + * Copyright (C) 2022 - 2026, Advanced Micro Devices, Inc.
> + */
> +
> +#ifndef _XLNX_VERSAL_CLK_H
> +#define _XLNX_VERSAL_CLK_H
> +
> +#define PMC_PLL					1

[Severity: Medium]
Should these device tree binding headers be placed in include/dt-bindings/
instead of arch/arm64/boot/dts/xilinx/?

Placing them in the architecture directory bypasses the convention of having
binding definitions act as a shared contract between the device tree and
kernel drivers. This applies to xlnx-versal-clk.h, xlnx-versal-power.h,
and others introduced here.

[ ... ]

> diff --git a/arch/arm64/boot/dts/xilinx/xlnx-versal-net-clk.h b/arch/arm64/boot/dts/xilinx/xlnx-versal-net-clk.h
> new file mode 100644
> index 0000000000000..4a6add03c173e
> --- /dev/null
> +++ b/arch/arm64/boot/dts/xilinx/xlnx-versal-net-clk.h
[ ... ]
> +/* Remove Versal specific node IDs */
> +#undef APU_PLL
> +#undef RPU_PLL
> +#undef CPM_PLL

[Severity: Medium]
Will these undef directives work correctly with the device tree schema tools?

Since schema validation tools typically use simple regex-based parsers rather
than a full C preprocessor, they generally ignore undef statements. Could this
cause the validation schemas to incorrectly accept the removed IDs?

[ ... ]

> diff --git a/arch/arm64/boot/dts/xilinx/xlnx-versal-resets.h b/arch/arm64/boot/dts/xilinx/xlnx-versal-resets.h
> new file mode 100644
> index 0000000000000..fe00f4a0ba65b
> --- /dev/null
> +++ b/arch/arm64/boot/dts/xilinx/xlnx-versal-resets.h
> @@ -0,0 +1,106 @@
> +/* SPDX-License-Identifier: GPL-2.0 */
> +/*
> + * Copyright (C) 2020 - 2022, Xilinx, Inc.
> + * Copyright (C) 2022 - 2026, Advanced Micro Devices, Inc.
> + */
> +
> +#ifndef _XLNX_VERSAL_RESETS_H
> +#define _XLNX_VERSAL_RESETS_H
> +
> +#define VERSAL_RST_PMC_POR			(0xc30c001U)

[Severity: Medium]
Does this new file unnecessarily duplicate the already existing reset IDs
binding in include/dt-bindings/reset/xlnx-versal-resets.h?

Maintaining duplicate copies of these bindings introduces a risk of the
two files drifting out of sync.

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/cover.1783495122.git.michal.simek@amd.com?part=5