On 17/01/2024 11:47, Siddharth Vadapalli wrote:
> Hello Krzysztof,
>
> On 17/01/24 16:04, Krzysztof Kozlowski wrote:
>> On 17/01/2024 11:25, Siddharth Vadapalli wrote:
>>> The existing implementation for validating the "num-lanes" property
>>> based on the compatible(s) doesn't enforce it. Fix it by updating the
>>> checks to handle both single-compatible and multi-compatible cases.
>>>
>>> Fixes: b3ba0f6e82cb ("dt-bindings: PCI: ti,j721e-pci-*: Add checks for num-lanes")
>>> Fixes: adc14d44d7cb ("dt-bindings: PCI: ti,j721e-pci-*: Add j784s4-pci-* compatible strings")
>>> Signed-off-by: Siddharth Vadapalli <s-vadapalli@xxxxxx>
>>> ---
>>> .../bindings/pci/ti,j721e-pci-ep.yaml | 26 ++++++++++++++-----
>>> .../bindings/pci/ti,j721e-pci-host.yaml | 26 ++++++++++++++-----
>>> 2 files changed, 38 insertions(+), 14 deletions(-)
>>>
>>> diff --git a/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml b/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml
>>> index 97f2579ea908..278e0892f8ac 100644
>>> --- a/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml
>>> +++ b/Documentation/devicetree/bindings/pci/ti,j721e-pci-ep.yaml
>>> @@ -68,8 +68,9 @@ allOf:
>>> - if:
>>> properties:
>>> compatible:
>>
>> Missing contains:, instead of your change.
>
> I did try the "contains" approach before determining that the implementation in
> this patch is more suitable. Please consider the following:
>
> For AM64 SoC the primary compatible is "ti,am64-pcie-ep" and fallback compatible
> is "ti,j721e-pcie-ep". For J7200 SoC the primary compatible is
> "ti,j7200-pcie-ep" while the fallback compatible is again "ti,j721e-pcie-ep".
>
> Therefore, the device-tree nodes for AM64 and J7200 look like:
>
> AM64:
> compatible = "ti,am64-pcie-ep", "ti,j721e-pcie-ep";
> ...
> num-lanes = 1;
>
> J7200:
> compatible = "ti,j7200-pcie-ep", "ti,j721e-pcie-ep";
> ...
> num-lanes = 4;
>
> This implies that when the check for "num-lanes" is performed on the device-tree
> node for PCIe in J7200, the fallback compatible of "ti,j721e-pcie-ep" within the
> AM64's "compatible: contains:" check will match the schema and it will check the
> existing "num-lanes" being described as "const: 1" against the value in J7200's
> PCIe node resulting in a warning.
What warning? What did you put to contains?
> Therefore, using "contains" will result in
> errors if the check has to be performed for device-tree nodes with fallback
> compatibles. The "items" based approach I have used in this patch ensures that
> the schema matches *only* when both the primary and fallback compatible are
> present in the device-tree node.
Long message, but I don't understand it. Why this binding is different
than all others which rely on contains?
>>> + - if:
>>> + properties:
>>> + compatible:
>>> + items:
>>> + - const: ti,j784s4-pcie-ep
>>
>> Why? Previous code was correct.
>
> Though I used "patience diff", for some reason the addition of
> "ti,j721e-pcie-ep" in the check has been treated as the removal of
> "ti,j784s4-pcie-ep" first followed by adding the same later for generating the
> diff in this patch. The diff above is equivalent to the addition of:
No, why do you change existing code? It is correct.
Best regards,
Krzysztof
