[U-Boot] ATAGS for Tegra, Sunxi, etc.

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

Hi,

On 17-12-15 10:21, Ian Campbell wrote:
> On Thu, 2015-12-17 at 07:40 +0100, Karsten Merker wrote:
>> On Thu, Dec 17, 2015 at 01:59:57PM +0900, Masahiro Yamada wrote:
>>> 2015-12-17 13:58 GMT+09:00 Masahiro Yamada <yamada.masahiro@socionext.c
>>> om>:
>>>> Hi,
>>>>
>>>> I noticed some well-maintained new SoC families still
>>>> define CONFIG_CMDLINE_TAG.
>>>>
>>>>
>>>> For example,
>> [...]
>>>> include/configs/sunxi-common.h
>>>>
>>>>   #define CONFIG_SETUP_MEMORY_TAGS
>>>>   #define CONFIG_CMDLINE_TAG
>>>>   #define CONFIG_INITRD_TAG
>>>>   #define CONFIG_SERIAL_TAG
>>
>>>> Do they still use ATAGS, not device tree?
>>
>> Sunxi uses devicetree for mainline kernels, but AFAIK ATAG
>> support is necessary to enable booting legacy vendor kernels.
>> There is still new sunxi-based hardware sold today that comes
>> with legacy 3.4-based kernels.
>
> That legacy kernel is FEX (allwinners own description blob) based, I don't
> know to what extent that involves ATAGs in some way though.
>
> There are also people who use the 3.4 based fork from linux-sunxi.org, but
> I don't know if that is DT or ATAGS or FEX.
>
> A dependency on CONFIG_OLD_SUNXI_KERNEL_COMPAT might be an option depending
> on what the kernels need, Hans probably knows better than I do.

The 3.4 based kernels use both ATAGS for things like memory size, and fex
for other hw config info.

I'm not in favor of wrapping things in CONFIG_OLD_SUNXI_KERNEL_COMPAT, because
recent 3.4 based kernels can boot without that, and I believe that removing
the ATAG support will break this, without really buying us much.

Regards,

Hans