Re: [PATCH 2/2] x86: intel-mid: add watchdog platform code for Merrifield

[Guenter Roeck <linux@roeck-us.net>]

On Wed, Apr 09, 2014 at 06:10:11PM +0200, Alexander Stein wrote:
> On Wednesday 09 April 2014 08:18:22, Guenter Roeck wrote:
> > On Wed, Apr 09, 2014 at 04:03:14PM +0200, Alexander Stein wrote:
> > > On Wednesday 09 April 2014 14:58:48, One Thousand Gnomes wrote:
> > > > > I'm getting curious: How can I use device-tree on x86(_64)?
> > > > > Reading the dependencies from CONFIG_OF it can only be used on 32bit systems with some special hardware bases.
> > > > > So, how to use otherwise?
> > > > 
> > > > There isn't any fundamental thing tying device tree to a given
> > > > architecture or 32bitness, it's just that sane PC architectures use ACPI
> > > > to enumerate devices, and/or have a discoverable bus architecture.
> > > > 
> > > > Some of the phones don't so this now becoems a point of consideration. In
> > > > fact it's already also used on CE4100 (which is an embedded media SoC
> > > > found in some TV devices and set-top boxes) and on the OLPC (One laptop
> > > > per child). There is no intrinsic reason it couldn't be used in other x86
> > > > special cases.
> > > > 
> > > > If its PC shaped however it probably has ACPI and ACPI and DT are not a
> > > > 1:1 mapping. ACPI has method invocations, and various firmware provided
> > > > interfaces such as the EC, Device tree is better at some other bits.
> > > 
> > > Yep, like SPI devices on embedded x86 hardware.
> > > 
> > > > Converting the phones to embedded device tree rather than adding a
> > > > billion little platform files on the other hand seems to me like a
> > > > no-brainer.
> > > 
> > > I found http://thread.gmane.org/gmane.linux.drivers.devicetree/4475. Why didn't that get into mainline?
> > > 
> > 
> > Isn't that patch set in mainline ?
> 
> Oh, I see that this patch didn't get included 1:1 as shown in that link. I only checked USE_OF in arch/x86/Kconfig which didn't get included that way :(
> 
> > We are working on a patch set to make DT support on x86 more
> > widely available. Thierry Reding did some work on it a while ago.
> > Some of it is upstream, other parts are available in 
> > 	https://github.com/avionic-design/linux/commits/medatom/master
> > 
> > We'll see if and what will be accepted upstream.
> 
> As far as I can see, CONFIG_OF is only selected if a specific hardware is enabled.
> 
Correct, and I expect that won't change. Default on x86 platform is ACPI.

Guenter