Re: [PATCH v3 07/10] riscv: boot: dts: r9a07g043: Add placeholder nodes

[Conor Dooley]

On Tue, Sep 20, 2022 at 02:17:50PM +0200, Geert Uytterhoeven wrote:
> Hi Conor,
> 
> On Fri, Sep 16, 2022 at 12:40 AM Conor Dooley <mail@xxxxxxxxxxx> wrote:
> > On 15/09/2022 23:26, Lad, Prabhakar wrote:
> > > On Thu, Sep 15, 2022 at 10:36 PM <Conor.Dooley@xxxxxxxxxxxxx> wrote:
> > >> On 15/09/2022 19:15, Prabhakar wrote:
> > >>> riscv: boot: dts: r9a07g043: Add placeholder nodes
> > >>> From: Lad Prabhakar <prabhakar.mahadev-lad.rj@xxxxxxxxxxxxxx>
> > >>> Add empty placeholder nodes to RZ/Five (R9A07G043) SoC DTSI.
> > >> Can you explain why do you need placeholder nodes for this and
> > >> cannot just directly include the other dtsis?
> > >>
> > > Since the RZ/G2UL SoC is ARM64 where it has a GIC and on RZ/Five SoC
> > > we have PLIC for interrupts. Also the interrupt numbering for RZ/Five
> > > SoC differs from RZ/G2UL SoC hence we are not directly using the
> > > RZ/G2UL SoC DTSI [0].
> > >
> > > [0] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/renesas/r9a07g043.dtsi?h=v6.0-rc5
> > >
> > > For the RZ/Five SMARC EVK I am re-using the below files [1] (SoM) and
> > > [2] (Carrier board) as the RZ/Five SMARC EVK is pin compatible.  Since
> > > I am re-using these when trying to compile the RZ/Five DTB I get
> > > compilation errors since the nodes dont exist (and there is no way
> > > currently we can delete the node reference when the actual node itself
> > > isn't present) hence these place holders.
> > >
> > > [1] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/renesas/rzg2ul-smarc-som.dtsi?h=v6.0-rc5
> > > [2] https://git.kernel.org/pub/scm/linux/kernel/git/torvalds/linux.git/tree/arch/arm64/boot/dts/renesas/rzg2ul-smarc.dtsi?h=v6.0-rc5
> >
> > If this method is acceptable to Geert, this explanation 100% needs to
> > go into the commit message.
> 
> We've been using these placeholders a lot in Renesas SoC-specific
> .dtsi files, as they allow us to introduce gradually support for a new SoC
> that is mounted on an existing PCB, and thus shares a board-specific
> .dtsi file.  Hence I'm fine with this.

Aye, if you're happy with it then I am too...
> 
> However, I think more properties can be dropped from the placeholders.
> There is no need to have e.g. 'reg-names' and 'status = "disabled"'
> (there is no compatible value, so no matching is done).

...and this makes a lot of sense. I'd still like a comment in the
commit message though explaining why placeholder nodes are needed as
opposed to just leaving it blank etc.

Thanks,
Conor.