On Fri, 12 Oct 2018 14:36:32 -0700, Douglas Anderson wrote:
> Digging through the "phy-qcom-qmp" showed me many inconsistencies
> between the bindings and the reality of the driver. Let's fix them
> all.
>
> * In commit 2d66eab18375 ("dt-bindings: phy: qmp: Add support for QMP
> phy in IPQ8074") we probably should have explicitly listed that
> there are no clocks for this PHY and also added the reset names in
> alphabetical order. You can see that there are no clocks in the
> driver where "clk_list" is NULL.
>
> * In commit 8587b220f05e ("dt-bindings: phy-qcom-qmp: Update bindings
> for QMP V3 USB PHY") we probably should have listed the resets for
> this new PHY and also removed the "(Optional)" marking for the "cfg"
> reset since PHYs that need "cfg" really do need it. It's just that
> not all PHYs need it.
>
> * In commit 7f0802074120 ("dt-bindings: phy-qcom-qmp: Update bindings
> for sdm845") we forgot to update one instance of the string
> "qcom,qmp-v3-usb3-phy" to be "qcom,sdm845-qmp-usb3-phy". Let's fix
> that. We should also have added "qcom,sdm845-qmp-usb3-uni-phy" to
> the clock-names and reset-names lists.
>
> * In commit 99c7c7364b71 ("dt-bindings: phy-qcom-qmp: Add UFS phy
> compatible string for sdm845") we should have added the set of
> clocks and resets for "qcom,sdm845-qmp-ufs-phy". These were taken
> from the driver.
>
> * Cleanup the wording for what properties child nodes have to make it
> more obvious which types of PHYs need clocks and resets. This was
> sorta implicit in the "-names" description but I found myself
> confused.
>
> * As per the code not all "pcie qmp phys" have resets. Specifically
> note that the "has_lane_rst" property in the driver is false for
> "ipq8074-qmp-pcie-phy". Thus make it clear exactly which PHYs need
> child nodes with resets.
>
> Signed-off-by: Douglas Anderson <dianders@xxxxxxxxxxxx>
> ---
>
> .../devicetree/bindings/phy/qcom-qmp-phy.txt | 31 ++++++++++++++-----
> 1 file changed, 23 insertions(+), 8 deletions(-)
>
Reviewed-by: Rob Herring <robh@xxxxxxxxxx>
