Re: [PATCH] dt-bindings: PCI: qcom: Fix register maps items and add 3.3V supply

[Rob Herring]

On Wed, Jun 05, 2024 at 11:19:28AM +0530, Manivannan Sadhasivam wrote:
> On Tue, Jun 04, 2024 at 05:58:06PM -0600, Rob Herring wrote:
> > On Tue, Jun 04, 2024 at 07:05:12PM +0300, Abel Vesa wrote:
> > > All PCIe controllers found on X1E80100 have MHI register region and
> > > VDDPE supplies. Add them to the schema as well.
> > > 
> > > Fixes: 692eadd51698 ("dt-bindings: PCI: qcom: Document the X1E80100 PCIe Controller")
> > > Signed-off-by: Abel Vesa <abel.vesa@xxxxxxxxxx>
> > > ---
> > > This patchset fixes the following warning:
> > > https://lore.kernel.org/all/171751454535.785265.18156799252281879515.robh@xxxxxxxxxx/
> > > 
> > > Also fixes a MHI reg region warning that will be triggered by the following patch:
> > > https://lore.kernel.org/all/20240604-x1e80100-dts-fixes-pcie6a-v2-1-0b4d8c6256e5@xxxxxxxxxx/
> > > ---
> > >  Documentation/devicetree/bindings/pci/qcom,pcie-x1e80100.yaml | 6 ++++--
> > >  1 file changed, 4 insertions(+), 2 deletions(-)
> > > 
> > > diff --git a/Documentation/devicetree/bindings/pci/qcom,pcie-x1e80100.yaml b/Documentation/devicetree/bindings/pci/qcom,pcie-x1e80100.yaml
> > > index 1074310a8e7a..7ceba32c4cf9 100644
> > > --- a/Documentation/devicetree/bindings/pci/qcom,pcie-x1e80100.yaml
> > > +++ b/Documentation/devicetree/bindings/pci/qcom,pcie-x1e80100.yaml
> > > @@ -19,11 +19,10 @@ properties:
> > >      const: qcom,pcie-x1e80100
> > >  
> > >    reg:
> > > -    minItems: 5
> > > +    minItems: 6
> > >      maxItems: 6
> > >  
> > >    reg-names:
> > > -    minItems: 5
> > >      items:
> > >        - const: parf # Qualcomm specific registers
> > >        - const: dbi # DesignWare PCIe registers
> > > @@ -71,6 +70,9 @@ properties:
> > >        - const: pci # PCIe core reset
> > >        - const: link_down # PCIe link down reset
> > >  
> > > +  vddpe-3v3-supply:
> > > +    description: A phandle to the PCIe endpoint power supply
> > 
> > TBC, this is a rail on the host side provided to a card? If so, we have 
> > standard properties for standard PCI voltage rails.
> 
> There is a 'vpcie3v3-supply' property and it should satisfy the requirement. But
> 'vddpe-3v3-supply' is already used on multiple Qcom SoCs. So changing the name
> in dts warrants the driver to support both supplies for backwards compatibility,
> but that should be fine.

Ideally, we would have a 'enable slot' or 'power on slot' function that 
turns on all the standard slot supplies and wiggles PERST# (and 
perhaps link state management) all following the timing defined in the 
PCI spec. Then the host drivers wouldn't have to do anything other than 
perhaps opt-in to use that.

> > It is also preferred that you put them in a root port node rather than the
> > host bridge.
> 
> Even though the resource split between root port and host bridge is not clear in
> Qcom SoCs, I think from logical stand point it makes sense.

Yeah, since it's typically 1:1 that's been blurred.

Rob