[PATCH V4 1/6] ARM: dts: imx6q: add common compatible name for reused modules

On Mon, Jan 13, 2014 at 04:16:23PM +0100, Sascha Hauer wrote:

On Mon, Jan 13, 2014 at 10:22:51PM +0800, Shawn Guo wrote:

quoted

On Mon, Jan 13, 2014 at 01:10:16PM +0100, Sascha Hauer wrote:

quoted

quoted

@@ -840,7 +840,7 @@
 			};
 
 			mmdc0: mmdc at 021b0000 { /* MMDC0 */
-				compatible = "fsl,imx6q-mmdc";
+				compatible = "fsl,imx6q-mmdc", "fsl,mmdc";

This is not nice. Here you introduce a fsl,mmdc compatible claiming all
mmdc are compatible to each other and in the driver code you have:

static const u32 imx6q_mmdc_io_dsm_offset[]
static const u32 imx6dl_mmdc_io_dsm_offset[]

which proves they are *not* compatible.

You do this just to share a

imx6_pm_get_base(&pm_info->mmdc_base, "fsl,mmdc");

across the different i.MX6 SoCs.

You can sanitize this by introducing a SoC struct which you populate
differently for the SoCs

static pm_soc_data imx6q_data {
	.mmdc_compat = "fsl,imx6q-mmdc",
};

And by putting cpu_type, mmdc_io_num and others in this struct you can
also remove the if(cpu_is_x()) else if (cpu_is_y()) else...

Good point.

Anson, the change below is a demonstration of Sascha's suggestion.
Sascha, correct me if I misunderstood your comment.

Looks good. That's exactly what I meant. Maybe the cpu_type field can
even be removed the way you did it.

Since the assembly function imx6_suspend() needs to check the cpu_type,
we have to keep it in imx6_cpu_pm_info.  So I assume you're suggesting
we remove the field from imx6_pm_socdata and determine the cpu_type by
comparing socdata pointer with particular imx6_pm_socdata like below?

	if (socdata == &imx6q_pm_data)
		pm_info->cpu_type = MXC_CPU_IMX6Q;
	else if (socdata == &imx6dl_pm_data)
		pm_info->cpu_type = MXC_CPU_IMX6DL;
	else if (socdata == &imx6sl_pm_data)
		pm_info->cpu_type = MXC_CPU_IMX6SL;

Looking at these if-clauses, I do not think we win too much from doing
thing in this way.

Shawn