From: Arnd Bergmann <arnd@arndb.de>
To: Alexandre Belloni <alexandre.belloni@free-electrons.com>
Cc: linux-arm-kernel@lists.infradead.org,
        Nicolas Ferre <nicolas.ferre@atmel.com>, linux-kernel@vger.kernel.org,
        =?utf-8?B?U3plbXrFkSBBbmRyw6Fz?= <sza@esh.hu>
Subject: Re: [PATCH v2 3/6] ARM: at91: Add armv7m support
Date: Thu, 20 Oct 2016 15:23:19 +0200
Message-ID: <19094397.gRAFyF5emk@wuerfel>
User-Agent: KMail/5.1.3 (Linux/4.4.0-34-generic; KDE/5.18.0; x86_64; ; )
In-Reply-To: <20161020102621.od7kkgpndooy25kz@piout.net>
References: <20161020094135.18221-1-alexandre.belloni@free-electrons.com> <5506721.HBUId2vnJb@wuerfel> <20161020102621.od7kkgpndooy25kz@piout.net>
MIME-Version: 1.0
Content-Transfer-Encoding: 7Bit
Content-Type: text/plain; charset="us-ascii"
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1673
Lines: 42

On Thursday, October 20, 2016 12:26:21 PM CEST Alexandre Belloni wrote:
> 
> On 20/10/2016 at 11:52:20 +0200, Arnd Bergmann wrote :
> > On Thursday, October 20, 2016 11:41:32 AM CEST Alexandre Belloni wrote:
> > > +
> > > +static void __init samx7_dt_device_init(void)
> > > +{
> > > +       struct soc_device *soc;
> > > +       struct device *soc_dev = NULL;
> > > +
> > > +       soc = at91_soc_init(samx7_socs);
> > > +       if (soc)
> > > +               soc_dev = soc_device_to_device(soc);
> > > +
> > > +       of_platform_populate(NULL, of_default_bus_match_table, NULL, soc_dev);
> > > +}
> > 
> > This was initially the idea for the soc_device, but we've stopped
> > using it as the parent for the on-chip devices a while ago.
> > 
> > Just register the device for identification here, and use
> > of_platform_default_populate with a NULL parent as most others do.
> > 
> > We should also investigate whether we can convert the three other
> > at91 variants to do the same without breaking expectations in user space.
> > 
> 
> My opinion is that we could just remove the whole at91_soc_init stuff
> but I think Nicolas still wants the two info lines to be printed for
> debugging/support purposes. I'm not sure how much this is used anyway
> and I don't find the sysfs attributes to be particularly useful.
> 
> Also, removing soc.c is a 10% reduction of the code in mach-at91 
> 

Having the soc_device driver is very valuable in order to have
an interface to be used from user space (and soon from the kernel)
to look up the exact SoC type in a generic way, so I'd definitely
want to keep that, though we may want to move that driver to
drivers/soc/.

	Arnd