[PATCH v8 06/17] dt-bindings: mfd: amd,pensando-elbasr: Add AMD Pensando Elba System Resource chip

[Geert Uytterhoeven]

Hi Brad,

On Thu, Nov 17, 2022 at 7:39 PM Larson, Bradley <Bradley.Larson at amd.com> wrote:
> > > > > --- /dev/null
> > > > > +++ b/Documentation/devicetree/bindings/mfd/amd,pensando-elbasr.yaml
> > > ...
> > > > > +
> > > > > +title: AMD Pensando Elba SoC Resource Controller
> > > > > +
> > > > > +description: |
> > > > > +  AMD Pensando Elba SoC Resource Controller functions are
> > > > > +  accessed with four chip-selects.  Reset control is on CS0.
> > > >
> > > > One device with 4 chip-selects? Then I'd expect 'reg = <0 1 2 3>;'
> > > >
> > > > Hard to say more because I don't have the whole thread nor remember what
> > > > exactly we discussed before. That was 100s of bindings ago...
> > >
> > > I agree and the example for v7 had all 4 chip-selects shown.
> >
> > That is not what I said. Look at 'reg' above again. You have 1 device,
> > but you have 4 nodes which looks like separate 4 devices. The
> > exception would be if what's on each chip select is independent from
> > each other.
> >
> > Describe what your h/w has/is/does so we can provide better guidance.
>
> This is the hardware design for every implementation with the Elba SoC
>
>   Elba <-- spi --> One FPGA or CPLD (4 chip-selects)
>
> where there are four functions in the FPGA accessed by userspace applications except for emmc reset control which is a bit in a CS0 register.  The IP at CS0 is a bunch of miscellaneous mgmt registers.  CS1 is a designware I2C master/slave.  CS2 is a Lattice dual I2C master.  CS3 is internal storage for the CPLD or FPGA depending on the hardware implementation.
>
> 'reg = <0 1 2 3>' in the dt fragment below is indicating the chip-select which is what other bindings appear to be doing.  Maybe one answer is delete this and add our compatible to spidev.c in the patchset we provide to customers.  Adding our compatible to spidev.c was nack'ed. Recommendation?

The fragment below does not have "reg = <0 1 2 3>", but individual
subnodes, each with their own "reg" property.

> Not including 'reg' results in this compile warning:
>
>   DTC     arch/arm64/boot/dts/amd/elba-asic.dtb
> arch/arm64/boot/dts/amd/elba-asic-common.dtsi:73.28-78.4: Warning (spi_bus_reg): /soc/spi at 2800/system-controller at 0: missing or empty reg property
> arch/arm64/boot/dts/amd/elba-asic-common.dtsi:80.22-84.4: Warning (spi_bus_reg): /soc/spi at 2800/system-controller at 1: missing or empty reg property
> arch/arm64/boot/dts/amd/elba-asic-common.dtsi:86.22-92.4: Warning (spi_bus_reg): /soc/spi at 2800/system-controller at 2: missing or empty reg property
> arch/arm64/boot/dts/amd/elba-asic-common.dtsi:94.22-98.4: Warning (spi_bus_reg): /soc/spi at 2800/system-controller at 3: missing or empty reg property
>   CALL    scripts/atomic/check-atomics.sh
>
> &emmc {
>         bus-width = <8>;
>         cap-mmc-hw-reset;
>         reset-names = "hw";
>         resets = <&rstc 0>;
>         status = "okay";
> };
>
> &spi0 {
>         #address-cells = <1>;
>         #size-cells = <0>;
>         num-cs = <4>;
>         cs-gpios = <0>, <0>, <&porta 1 GPIO_ACTIVE_LOW>,
>                    <&porta 7 GPIO_ACTIVE_LOW>;
>         status = "okay";
>
>         rstc: system-controller at 0 {
>                 compatible = "amd,pensando-elbasr";
>                 reg = <0>;
>                 spi-max-frequency = <12000000>;
>                 #reset-cells = <1>;
>         };
>
>         system-controller at 1 {
>                 compatible = "amd,pensando-elbasr";
>                 reg = <1>;
>                 spi-max-frequency = <12000000>;
>         };
>
>         system-controller at 2 {
>                 compatible = "amd,pensando-elbasr";
>                 reg = <2>;
>                 spi-max-frequency = <12000000>;
>                 interrupt-parent = <&porta>;
>                 interrupts = <0 IRQ_TYPE_LEVEL_LOW>;
>         };
>
>         system-controller at 3 {
>                 compatible = "amd,pensando-elbasr";
>                 reg = <3>;
>                 spi-max-frequency = <12000000>;
>         };

The above describes 4 individual but further identical (they have the
same compatible value) devices, while above you say they are not identical:

    The IP at CS0 is a bunch of miscellaneous mgmt registers.
    CS1 is a designware I2C master/slave.
    CS2 is a Lattice dual I2C master.
    CS3 is internal storage for the CPLD or FPGA depending on the
hardware implementation.

So either this should be modelled as a single subnode with 4 chip
selects[*]:

        system-controller at 0 {
                compatible = "amd,pensando-elbasr";
                reg = <0 1 2 3 4>;
                spi-max-frequency = <12000000>;
                #reset-cells = <1>;
        };

or as 4 separate subnodes, each using 4 different compatible values.
Giving the wildly different functionalities provided by each, you also need
4 binding documents.

[*] I'm not sure the Linux SPI core actually supports this yet.

> };

Gr{oetje,eeting}s,

                        Geert

--
Geert Uytterhoeven -- There's lots of Linux beyond ia32 -- geert at linux-m68k.org

In personal conversations with technical people, I call myself a hacker. But
when I'm talking to journalists I just say "programmer" or something like that.
                                -- Linus Torvalds