Re: [PATCH v3 5/6] dtc: dt-check-style: Do not sort top-level nodes in DTS by unit address

[sashiko-bot@kernel.org]

Thank you for your contribution! Sashiko AI review found 2 potential issue(s) to consider:
- [Medium] The checker script inadvertently disables unit address sorting validation for all immediate children of top-level reference nodes.
- [Medium] Top-level nodes with unit addresses are completely excluded from both address-based and name-based sorting checks.
--

commit 52ef4b5728a3b53330c7623fdea74a0e2eddd12d
Author: Krzysztof Kozlowski <krzysztof.kozlowski@oss.qualcomm.com>

dtc: dt-check-style: Do not sort top-level nodes in DTS by unit address

This commit modifies the dt-check-style script to stop enforcing unit
address sorting for top-level nodes in DTS files. It notes that top-level
nodes with unit addresses are typically sorted by node name instead.

> diff --git a/scripts/dtc/dt-check-style b/scripts/dtc/dt-check-style
> index d9080297bd4db..3c1aa9e28347e 100755
> --- a/scripts/dtc/dt-check-style
> +++ b/scripts/dtc/dt-check-style
> @@ -515,12 +515,18 @@ def _natural_sort_key(s):
>  
>  def check_child_address_order(ctx):
>      """Addressed siblings (foo@N) must appear in ascending address
> -    order within their parent node body."""
> +    order within their parent node body.
> +    Exception: Top-level in DTS follows name order, regardless of unit address
> +    in memory@N and soc@N nodes
> +    """
>      for children in _walk_bodies(ctx.lines):
>          addressed = []
>          for c in children:
>              if c.node_addr is None:
>                  continue
> +            if (ctx.file_type == 'dts') and (c.depth < 2):
> +                # Top-level does not use unit address sorting usually
> +                continue

[Severity: Medium]
Does this unintentionally disable address-order validation for the vast
majority of peripheral nodes added in board files?

When using top-level reference nodes (e.g., &i2c0 { sensor@50 { ... }; };)
in a .dts or .dtso file, the lexical depth of &i2c0 is 0. This makes the
depth of sensor@50 equal to 1. Since c.depth < 2 is true here, won't this
skip the unit address validation for all immediate children of top-level
references?

[Severity: Medium]
The commit message states that the DTS usually chooses sorting by node
name for the top-level part. However, does this change leave top-level
nodes with unit addresses completely unvalidated?

Looking at check_child_name_order() elsewhere in this script, it explicitly
skips all nodes with a unit address:

scripts/dtc/dt-check-style:check_child_name_order() {
    for children in _walk_bodies(ctx.lines):
        for c in children:
            if c.node_addr is not None:
                continue
...
}

Since check_child_name_order() bypasses nodes with unit addresses, and
this new logic skips them in check_child_address_order(), will nodes like
memory@0 and soc@0 bypass both sorting checks?

-- 
Sashiko AI review · https://sashiko.dev/#/patchset/20260706-dts-style-checker-v3-0-03ffacff9226@oss.qualcomm.com?part=5