Re: [PATCH v1 2/2] arm64: dts: qcom: sa8775p: Add ep pcie1 controller node

[Mrinmay Sarkar]

On 11/8/2023 3:24 AM, Konrad Dybcio wrote:
> On 11/7/23 19:37, Andrew Halaney wrote:
> > On Tue, Nov 07, 2023 at 06:34:53PM +0530, Mrinmay Sarkar wrote:
> > > Add ep pcie dtsi node for pcie1 controller found on sa8775p platform.
> > > It supports gen4 and x4 link width. Limiting the speed to Gen3 due to
> > > stability issues.
> >
> > I wouldn't mind a bit more information on what "stability" issues
> > entails! I'm a sucker for details in a commit message.
> Yep, giving us a bit more than "doesnt work" may help us help you!
Actually if I enable gen4 some time I am getting link down and sometime
link getting establish at gen1. I am not getting stable gen4, may be we
need to program some register to get stable gen4 that I am checking.
> > > +
> > > +        interrupts = <GIC_SPI 518 IRQ_TYPE_LEVEL_HIGH>,
> > > +                     <GIC_SPI 152 IRQ_TYPE_LEVEL_HIGH>,
> > > +                     <GIC_SPI 474 IRQ_TYPE_LEVEL_HIGH>;
> Looks like the indentation is off?
>
> > > +
> > > +        interrupt-names = "global", "doorbell", "dma";
> > > +
> > > +        interconnects = <&pcie_anoc MASTER_PCIE_1 0 &mc_virt 
> > > SLAVE_EBI1 0>,
> > > +                <&gem_noc MASTER_APPSS_PROC 0 &config_noc 
> > > SLAVE_PCIE_1 0>;
> >
> > I keep seeing Konrad requesting that we use the #define instead of a raw
> > number 0, i.e. something like QCOM_ICC_TAG_ALWAYS (although if I'm
> > reading that correctly QCOM_ICC_TAG_ALWAYS doesn't evaluate to 0, so
> > make sure you pick the appropriate one).
> No it doesn't, but if you look at the code, tag being non-existent
> assigns QCOM_ICC_TAG_ALWAYS which is a workaround for DTBs from back
> when interconnect tags were not a thing
>
> > > +        interconnect-names = "pcie-mem", "cpu-pcie";
> >
> > This is nitpicky, but unless someone told you to do the whitespace
> > between some of these properties I'd get more consistent. i.e. reg and
> > reg-names has no newline between them, but clocks and clock-names does,
> > and then interconnects/interconnect-names does not.
> :)
I don't think there is any rule to add those white spaces, in next patch
I will align these white spaces with other pcie nodes.
> Konrad

Thanks,
Mrinmay