Re: [PATCH v2 13/13] ARM: dts: r9a06g032: Describe the QSPI controller

[Miquel Raynal]

On 16/01/2026 at 11:07:03 +01, Geert Uytterhoeven <geert@xxxxxxxxxxxxxx> wrote:

> Hi Miquel,
>
> On Fri, 16 Jan 2026 at 10:49, Miquel Raynal <miquel.raynal@xxxxxxxxxxx> wrote:
>> >> +               qspi0: spi@40005000 {
>> >> +                       compatible = "renesas,r9a06g032-qspi", "renesas,rzn1-qspi", "cdns,qspi-nor";
>> >> +                       reg = <0x40005000 0x1000>, <0x10000000 0x10000000>;
>> >> +                       interrupts = <GIC_SPI 64 IRQ_TYPE_LEVEL_HIGH>;
>> >> +                       clocks = <&sysctrl R9A06G032_CLK_QSPI0>, <&sysctrl R9A06G032_HCLK_QSPI0>,
>> >> +                                <&sysctrl R9A06G032_HCLK_QSPI0>;
>> >> +                       clock-names = "ref", "ahb", "apb";
>> >> +                       #address-cells = <1>;
>> >> +                       #size-cells = <0>;
>> >> +                       cdns,fifo-width = <4>;
>> >
>> > <4> is the default, right?
>>
>> It is the default in the bindings indeed, however the driver does not
>> imply that default and errors out if the property is missing. The
>> property is also marked required in the bindings, which is kind of
>> incorrect I guess. Also, all DTS explicitly set this value to 4.
>
> OK.
>
>>
>> However looking into the RM I found "Transmit and receive FIFOs are 16
>> bytes". I haven't tested that, I will.
>
> Oh, that bullet is not present in the docs on the CD I looked at.
> It is indeed documented in newer versions.
>
> There's also cdns,fifo-depth, which thus should be 4?

Ah, that's right. You mean fifo-width = 4 and fifo-depth = 4, right?
It is marked as being discoverable by reading CQSPI_REG_SRAMPARTITION,
but on my board this register returns 0. So I guess yes, I will go for
fifo-depth = 4 and we should be fine.

...

>> > The rest LGTM, ignoring my comments on the bindings:
>> > Reviewed-by: Geert Uytterhoeven <geert+renesas@xxxxxxxxx>
>>
>> Thanks for the review, but I guess if I end up changing the DTS snippet
>> I might drop it. Or would you like me to keep it anyway?
>
> Please keep it as long as you don't change the (SoC integration)
> things I typically focus on (address, interrupts, clocks), and don't make
> too wild changes ;-)

Thanks,
Miquèl