Re: [PATCH v2 02/16] dt-bindings: i2c: imx: update schema to align with original txt binding

From: Li Yang <leoyang.li@nxp.com>
To: Rob Herring <robh@kernel.org>
Cc: Shawn Guo <shawnguo@kernel.org>,
	"open list:OPEN FIRMWARE AND FLATTENED DEVICE TREE BINDINGS" 
	<devicetree@vger.kernel.org>,
	Oleksij Rempel <linux@rempel-privat.de>,
	linux-arm-kernel <linux-arm-kernel@lists.infradead.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v2 02/16] dt-bindings: i2c: imx: update schema to align with original txt binding
Date: Fri, 8 Oct 2021 22:08:34 -0500	[thread overview]
Message-ID: <CADRPPNSxLaG1WcxKY-KtBwJON51Rg5-euJNJnJObarjEKN2eSA@mail.gmail.com> (raw)
In-Reply-To: <YWDEESYAuQSIfOSQ@robh.at.kernel.org>

On Fri, Oct 8, 2021 at 5:20 PM Rob Herring <robh@kernel.org> wrote:
>
> On Fri, Oct 01, 2021 at 12:37:54PM -0500, Li Yang wrote:
> > On Fri, Oct 1, 2021 at 8:24 AM Rob Herring <robh+dt@kernel.org> wrote:
> > >
> > > On Thu, Sep 30, 2021 at 7:04 PM Li Yang <leoyang.li@nxp.com> wrote:
> > > >
> > > > When the binding was converted from txt to yaml, it actually added more
> > > > constrains than the original txt binding which was already used in many
> > > > in-tree DTSes.  Some of the newly added constrains are either not valid
> > > > or not neccessary.
> > >
> > > IMO, both of these should be fixed in the dts files.
> > >
> > > > Not all SoCs use ipg as the clock name for i2c.  There is no point in
> > > > having SoC integration information defined in i2c binding.  Remove the
> > > > clock name requirement in the schema.
> > >
> > > Any name you want is not fine. Your choices are remove clock-names,
> > > add all the names used, or change everyone to use 'ipg'.
> >
> > I understand that the name should be important as clocks are
> > referenced by name.  But since the i2c controller only has one clock ,
> > the name is never referenced in the driver.
>
> Then just remove 'clock-names' from the dts file.

Had thought the clock-names are mandatory, but it turns out not.
Removing it should be great.

>
> > If we really want to define the name, IMO, it should be from the
> > perspective of the i2c controller like "clkin" or "i2c" instead of the
> > "ipg" from the perspective of SoC integration which could be changing
> > with a different integration.  I can list both "i2c" and "ipg" for now
> > as a workaround though.
>
> $modulename for $foo-names always looks made up and pointless to me.
>
> >
> > >
> > > > The original txt binding didn't require the order of tx and rx for
> > > > dmas/dma-names.  Many in tree DTSes are already using the other order.
> > > > Both orders should just work fine.  Update the schema to allow both.
> > >
> > > Doesn't sound like a case where defining the order is challenging.
> >
> > No, it is not challenging.  But as dma channel is only referenced by
> > name instead of index.  I don't see too much benefit in enforcing the
> > order other than easier to create the schema.
>
> Easier is nice, and that's the 'DT way' is the other reason.

Ok.

Regards,
Leo