On Thu, Aug 18, 2022 at 12:46:33PM +0300, Krzysztof Kozlowski wrote:
> On 18/08/2022 12:22, Shawn Guo wrote:
> > On Thu, Aug 18, 2022 at 10:51:02AM +0300, Krzysztof Kozlowski wrote:
> >> On 18/08/2022 04:33, Shawn Guo wrote:
> >>> On Mon, Jul 04, 2022 at 11:12:09AM +0200, Krzysztof Kozlowski wrote:
> >>>>> diff --git a/Documentation/devicetree/bindings/net/fsl,fec.yaml b/Documentation/devicetree/bindings/net/fsl,fec.yaml
> >>>>> index daa2f79a294f..6642c246951b 100644
> >>>>> --- a/Documentation/devicetree/bindings/net/fsl,fec.yaml
> >>>>> +++ b/Documentation/devicetree/bindings/net/fsl,fec.yaml
> >>>>> @@ -40,6 +40,10 @@ properties:
> >>>>> - enum:
> >>>>> - fsl,imx7d-fec
> >>>>> - const: fsl,imx6sx-fec
> >>>>> + - items:
> >>>>> + - enum:
> >>>>> + - fsl,imx8ulp-fec
> >>>>> + - const: fsl,imx6ul-fec
> >>>>
> >>>> This is wrong. fsl,imx6ul-fec has to be followed by fsl,imx6q-fec. I
> >>>> think someone made similar mistakes earlier so this is a mess.
> >>>
> >>> Hmm, not sure I follow this. Supposing we want to have the following
> >>> compatible for i.MX8ULP FEC, why do we have to have "fsl,imx6q-fec"
> >>> here?
> >>>
> >>> fec: ethernet@29950000 {
> >>> compatible = "fsl,imx8ulp-fec", "fsl,imx6ul-fec";
> >>> ...
> >>> };
> >>
> >> Because a bit earlier this bindings is saying that fsl,imx6ul-fec must
> >> be followed by fsl,imx6q-fec.
> >
> > The FEC driver OF match table suggests that fsl,imx6ul-fec and fsl,imx6q-fec
> > are not really compatible.
> >
> > static const struct of_device_id fec_dt_ids[] = {
> > { .compatible = "fsl,imx25-fec", .data = &fec_devtype[IMX25_FEC], },
> > { .compatible = "fsl,imx27-fec", .data = &fec_devtype[IMX27_FEC], },
> > { .compatible = "fsl,imx28-fec", .data = &fec_devtype[IMX28_FEC], },
> > { .compatible = "fsl,imx6q-fec", .data = &fec_devtype[IMX6Q_FEC], },
> > { .compatible = "fsl,mvf600-fec", .data = &fec_devtype[MVF600_FEC], },
> > { .compatible = "fsl,imx6sx-fec", .data = &fec_devtype[IMX6SX_FEC], },
> > { .compatible = "fsl,imx6ul-fec", .data = &fec_devtype[IMX6UL_FEC], },
>
> I don't see here any incompatibility. Binding driver with different
> driver data is not a proof of incompatible devices.
To me, different driver data is a good sign of incompatibility. It
mostly means that software needs to program the hardware block
differently.
> Additionally, the
> binding describes the hardware, not the driver.
>
> > { .compatible = "fsl,imx8mq-fec", .data = &fec_devtype[IMX8MQ_FEC], },
> > { .compatible = "fsl,imx8qm-fec", .data = &fec_devtype[IMX8QM_FEC], },
> > { /* sentinel */ }
> > };
> > MODULE_DEVICE_TABLE(of, fec_dt_ids);
> >
> > Should we fix the binding doc?
>
> Maybe, I don't know. The binding describes the hardware, so based on it
> the devices are compatible. Changing this, except ABI impact, would be
> possible with proper reason, but not based on Linux driver code.
Well, if Linux driver code is written in the way that hardware requires,
I guess that's just based on hardware characteristics.
To me, having a device compatible to two devices that require different
programming model is unnecessary and confusing.
Shawn
