Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
Date:   Fri, 14 Sep 2018 15:42:46 +0200
From:   Bernhard Frauendienst <kernel@nospam.obeliks.de>
To:     Rob Herring <robh+dt@kernel.org>
Cc:     David Woodhouse <dwmw2@infradead.org>,
        Brian Norris <computersforpeace@gmail.com>,
        Boris Brezillon <boris.brezillon@bootlin.com>,
        Marek =?utf-8?B?VmHFoXV0?= <marek.vasut@gmail.com>,
        Richard Weinberger <richard@nod.at>,
        Mark Rutland <mark.rutland@arm.com>,
        =?iso-8859-1?Q?Miqu=E8l?= Raynal <miquel.raynal@bootlin.com>,
        MTD Maling List <linux-mtd@lists.infradead.org>,
        devicetree@vger.kernel.org,
        "linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>
Subject: Re: [PATCH v3 3/3] mtd: mtdconcat: add dt driver for concat devices
Message-ID: <20180914134246.GB5913@dellux>
References: <20180908131345.8145-1-kernel@nospam.obeliks.de>
 <20180908131345.8145-4-kernel@nospam.obeliks.de>
 <CAL_JsqL+CMzkyeHLxa+6xf0FkM3mSPiGvoMrRoVV-FS1Q7SQsA@mail.gmail.com>
MIME-Version: 1.0
Content-Type: text/plain; charset=us-ascii
Content-Disposition: inline
In-Reply-To: <CAL_JsqL+CMzkyeHLxa+6xf0FkM3mSPiGvoMrRoVV-FS1Q7SQsA@mail.gmail.com>
User-Agent: Mutt/1.9.4 (2018-02-28)
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

Hi Rob,

sorry it took me so long to reply, I had to set up mutt first ;)

On Mon, Sep 10, 2018 at 04:47:22PM -0500, Rob Herring wrote:
> On Sat, Sep 8, 2018 at 8:20 AM Bernhard Frauendienst
> <kernel@nospam.obeliks.de> wrote:
> >
> > Some mtd drivers like physmap variants have support for concatenating
> > multiple mtd devices, but there is no generic way to define such a
> > concat device from within the device tree.
> 
> That is the generic way, but I suppose that only works for parallel
> memory mapped devices. So is it just spi-nor that you need to support?

I'm no expert in this domain, or regarding OpenWRT, but as far as I can
tell this only concerns spi-nor devices. It definitely does in my
specific use-case.

> Can we just make reg take multiple chip selects just like the existing
> support takes multiple reg entries?

That sounds like an interesting idea, but I'm afraid it's not a trivial
change, as far as I can tell from looking at spi.c. 

The 'reg' property is currently read in of_spi_parse_dt which gets
passed the spi device created in of_register_spi_device. The latter
would have to create multiple devices from a single OF node, and pass
the chipselect value to of_spi_parse_dt, which would have to parse the
same node multiple times, which feels a bit strange. To make things
worse, of_register_spi_device returns the created device, although that
value seems unused in the current tree (except for error checks).

Am I correct to assume that this change in method signature could be
accteptable since both functions are not exported?

Regards,
Bernhard