[PATCH 2/6] ARM: add Highbank core platform support

On Wed, Aug 24, 2011 at 09:45:28PM -0500, Rob Herring wrote:

Arnd,

On 08/20/2011 01:44 PM, Rob Herring wrote:

quoted

Arnd,

On 08/18/2011 10:34 AM, Arnd Bergmann wrote:

quoted

On Tuesday 16 August 2011, Rob Herring wrote:

quoted

+void __iomem *a9_base_addr = ((void __iomem *)(HB_MPIC_VIRT_BASE));
+void __iomem *sregs_base;
+
+static struct map_desc highbank_io_desc[] __initdata = {
+	{
+		.virtual	= HB_MPIC_VIRT_BASE,
+		.pfn		= 0, /* run-time */
+		.length		= SZ_4K,
+		.type		= MT_DEVICE,
+	},
+#ifdef CONFIG_DEBUG_LL
+	{
+		.virtual	= HB_DEBUG_LL_VIRT_BASE,
+		.pfn		= __phys_to_pfn(HB_DEBUG_LL_PHYS_BASE),
+		.length		= SZ_4K,
+		.type		= MT_DEVICE,
+	},
+#endif
+};
+
+static void __init highbank_map_io(void)
+{
+	unsigned long base;
+
+	/* Get SCU base */
+	asm("mrc p15, 4, %0, c15, c0, 0" : "=r" (base));
+
+	highbank_io_desc[0].pfn = __phys_to_pfn(base);
+	iotable_init(highbank_io_desc, ARRAY_SIZE(highbank_io_desc));
+}

I really liked the way that Barry moved the io_desc out to the
drivers using them, e.g arch/arm/mach-prima2/lluart.c.

Can you do the same thing with your lluart and with the a9_base_addr?
I guess it can live locally in platsmp.c.

Okay.

Looking at this some more, it doesn't work too well. platsmp.c depends
on CONFIG_SMP, but the SCU mapping is always needed even for !SMP
because the SCU has a power mode register for each core used by the
power controller. So putting it in platsmp.c would add ifdefs.

It sounds you have a well designed power controller which will shut
down different CA-9 power domains per SCU power mode settings, e.g.
it will shut off different set of CA-9 power domains for Dormant and
Shutdown modes.  Is that the case?

i.MX6Q is different from that case, so I can actually follow Arnd's
suggestion to keep scu_base locally in platsmp.c.

-- 
Regards,
Shawn