[U-Boot] [PATCH 1/2] sunxi: video: Do not use CONFIG_SYS_MEM_TOP_HIDE for the framebuffer

[Hans de Goede <hdegoede@redhat.com>]

Hi,

On 02/03/2015 02:20 PM, Ian Campbell wrote:
> On Mon, 2015-02-02 at 19:57 +0100, Hans de Goede wrote:
>> @@ -1237,6 +1243,20 @@ int sunxi_simplefb_setup(void *blob)
>>   		return 0; /* Keep older kernels working */
>>   	}
>>
>> +	/*
>> +	 * Do not report the framebuffer as free RAM to Linux, note we cannot
>> +	 * use fdt_add_mem_rsv() here, because then it is still seen as RAM,
>> +	 * and Linux refuses to iomap RAM on ARM, see:
>> +	 * linux/arch/arm/mm/ioremap.c around line 301.
>
> A little bit Linux centric ;-)
>
> I'd suggest s/Linux/OS for the first one and s/and Linux refuses/for
> example Linux refuses/.

OK, will fix.

>> +	 */
>> +	start = gd->bd->bi_dram[0].start;
>> +	size = gd->bd->bi_dram[0].size - CONFIG_SUNXI_MAX_FB_SIZE;
>> +	ret = fdt_fixup_memory_banks(blob, &start, &size, 1);
>
> I suppose this is currently safe when put up against arch_fixup_fdt(),
> but it might be a bit fragile against future changes?

We get called after arch_fixup_fdt(), so we should be able to simply
override what it has setup in the fdt.

> I'm not sure what else we can do though, directly adjusting
> gd->bd->bi_dram[0].size seems iffy.
>
> Perhaps a hook called from arch_fixup_fdt which is passed the arrays
> which are about to be passed to fdt_fixup_memory_banks so it can adjust
> them?

Then we need to keep track of if that callback succeeded, because without
the reservation we should not enable simplefb. I would prefer to just
keep this bit as is.

Regards,

Hans