Hi,

On Mon, Aug 18, 2014 at 03:27:13PM +0200, Carlo Caione wrote:

On Sun, Aug 17, 2014 at 04:21:15PM +0200, Maxime Ripard wrote:

quoted

Hi Carlo,

Hi Maxime,
thank you for reviewing also these patches, I'll keep you in CC for the
next revisions if you are interested.

Yep, sure.

quoted

quoted

+static const char * const m6_common_board_compat[] = {
+	"amlogic,8726_mx",
+	"amlogic,8726_mxs",
+	"amlogic,8726_mxl",
+	"amlogic,meson6",

Why are all those compatibles needed? Usually, you add a single one
per SoC (which would be the last in your case I guess.

They are taken from the weird DTS in the original Amlogic sources but I
guess you are right. I am actually more inclined to just leave "8726_mx"
and "meson6" since online you can find equally both the versions for
exactly the same SoCs.

I don't really know what's the best option here, but you should really
choose one name and stick to it. Since the mach directory is called
meson, I guess meson6 would make more sense, but it's your call.

quoted

quoted

+	NULL,
+};
+
+DT_MACHINE_START(AML8726_MX, "Amlogic Meson6 platform")
+	.init_machine	= meson_init_machine_devicetree,

And since you don't need the init machine, you can just use the
generic machine support. I'm not sure what's been decided on this,
should we remove such empty machines?

I can get rid of the .init_machine but what about the .dt_compat field?

Technically, it would work. The only drawbacks are that you don't get
the machine name in /proc/cpuinfo, and that you'll probably have to
add this file at some point in the future anyway.

Maxime

-- 
Maxime Ripard, Free Electrons
Embedded Linux, Kernel and Android engineering
http://free-electrons.com