Re: [PATCH v4 2/4] ARM: dts: NSP: Add DT files for Meraki MX64 series

From: Arnd Bergmann <arnd@arndb.de>
To: Florian Fainelli <f.fainelli@gmail.com>
Cc: Matthew Hagan <mnhagan88@gmail.com>,
	Vladimir Oltean <olteanv@gmail.com>,
	Rob Herring <robh+dt@kernel.org>, Olof Johansson <olof@lixom.net>,
	SoC Team <soc@kernel.org>, Ray Jui <rjui@broadcom.com>,
	Scott Branden <sbranden@broadcom.com>,
	bcm-kernel-feedback-list <bcm-kernel-feedback-list@broadcom.com>,
	Viresh Kumar <viresh.kumar@linaro.org>,
	Sam Ravnborg <sam@ravnborg.org>,
	Geert Uytterhoeven <geert+renesas@glider.be>,
	DTML <devicetree@vger.kernel.org>,
	Linux Kernel Mailing List <linux-kernel@vger.kernel.org>,
	Linux ARM <linux-arm-kernel@lists.infradead.org>
Subject: Re: [PATCH v4 2/4] ARM: dts: NSP: Add DT files for Meraki MX64 series
Date: Fri, 25 Jun 2021 20:40:53 +0200	[thread overview]
Message-ID: <CAK8P3a3vHOyFVnBTBpJ9_N8QPtb_dhU+itBXixrL3gjEuGpKkA@mail.gmail.com> (raw)
Message-ID: <20210625184053.90LoYO-66Or60rS519Ryxnhx7a96RlHW653nqbF80Vs@z> (raw)
In-Reply-To: <21e895e9-0211-0c97-5e9e-781e6d2fe6dd@gmail.com>

On Fri, Jun 25, 2021 at 7:30 PM Florian Fainelli <f.fainelli@gmail.com> wrote:
> On 6/25/21 10:26 AM, Matthew Hagan wrote:
> > On 25/06/2021 10:59, Arnd Bergmann wrote:
> >
> > How about having separate bcm-nsp-ax and bcm-nsp-bx dtsi files with the
> > appropriate secondary-boot-reg and dma-coherent (or lack of)
> > properties, which then include bcm-nsp.dtsi. Thus we can also avoid use
> > of /delete-property/. Would this be preferable?

That sounds good to me.

> Is there any way that the Ax platforms could use a small shim between
> the boot loader and the kernel which could all of the necessary DT
> adaptation so the kernel only contains a single Device Tree source?
>
> Using something like this:
>
> https://github.com/zonque/pxa-impedance-matcher/
>
> could be useful.

I don't think that's necessary here, but I wouldn't object if someone
finds it useful and does the work. ;-)

> >> On the other hand, the /chosen, /aliases and  /memory nodes that you have
> >> in the .dtsi file should probably get moved into the .dts files, as these tend
> >> to be board specific settings, even if the examples you have are all
> >> the same.
> >
> > I did not come across any convention regarding this, though there are
> > plenty of cases where the /chosen, /aliases and /memory nodes are
> > defined in a .dtsi file and used by multiple similar boards. Also note
> > in this case /aliases is defined in bcm-nsp.dtsi, not by me. Would we
> > not prefer to avoid having 6x duplication?

We are not too consistent about this, and there are cases in which a
.dtsi file is used for a family of boards using different SoCs rather than
a particular SoC or SoC family.

In the bcm-nsp.dtsi example you mention, I would move the aliases into
the board files, mainly because there is no guarantee that each board
exposes both uarts and all three on-chip ethernet ports. Note that the
aliases are supposed to match whatever label you have on the board,
not what the numbers are in the chip.

      Arnd