From mboxrd@z Thu Jan  1 00:00:00 1970
From: arnd@arndb.de (Arnd Bergmann)
Date: Tue, 12 May 2015 22:49:34 +0200
Subject: [PATCH v3 1/4] ARM: pxa: add memory resource to RTC device
In-Reply-To: <20150512203613.GQ2067@n2100.arm.linux.org.uk>
References: <1431384089-28367-1-git-send-email-robh@kernel.org>
 <53644246.uSkhGz7C05@wuerfel> <20150512203613.GQ2067@n2100.arm.linux.org.uk>
Message-ID: <4925803.188r69qFM9@wuerfel>
To: linux-arm-kernel@lists.infradead.org
List-Id: linux-arm-kernel.lists.infradead.org

On Tuesday 12 May 2015 21:36:13 Russell King - ARM Linux wrote:
> On Tue, May 12, 2015 at 10:30:12PM +0200, Arnd Bergmann wrote:
> > On Tuesday 12 May 2015 22:24:49 Robert Jarzmik wrote:
> > > 
> > > I made the try, and Russell was right, this breaks the pxa architecture. The
> > > relevant extract of kernel message is in [1].
> > > 
> > > The consequences with your patches :
> > >  - pxa27x_init()
> > >    - platform_add_devices()
> > >      - the pxa_device_rtc fails (as resource is duplicated)
> > >      - all the devices are unregistered (rollback)
> > > And pxa27x fails.
> > > 
> > > Now I'm pondering about the right approach :
> > >  - either remove sa1100_device_rtc from pxas
> > >  - or remove pxa_device_rtc
> > >  - or both
> > >  - or something else
> > > 
> > > Let me think a bit about it.
> > > 
> > 
> > To solve the problem with the duplicate registration of one resource,
> > I'd suggest using platform_device_register_simple() for the registration,
> > which will copy the resource. You can then mark the resource as __initconst
> > and remove the device to save a little memory at runtime.
> 
> No, a better solution is to solve the problem which requires the
> duplication in the first place, which is a broken driver structure.

Yes, makes sense. Or possibly do both ;-)

	Arnd