Binman for RK35xx

[Jonas Karlman]

On 2023-04-03 15:33, Peter Robinson wrote:
> On Mon, Apr 3, 2023 at 2:30 PM Mark Kettenis <mark.kettenis at xs4all.nl> wrote:
>>
>>> From: <i.zafievsky at gmail.com>
>>> Date: Mon, 3 Apr 2023 12:27:36 +0300
>>>
>>> Hi!
>>>
>>>    Trying to build mainline U-Boot for Rockchip RK3568-based board (Forlinx
>>> RK3568), but I think it;s can be reproduced on all RK35 chips. The main
>>> problem is that BL32 (AKA OP-TEE) is provided by Roch-Chips as a
>>> pre-compiled blob and binman fails to implement that scenario. Used .dtsi
>>> file and binman verbose output provided in attachments. Do am I done
>>> something  wrong or it can be considered as bug? Can someone assist?
>>
>> What do you need OP_TEE for?  Many of us consider less blobs a good
>> thing and simply live without OP-TEE.
> 
> I think it's a moot point, OP_TEE is useful for a number of things,
> and it should be an option for those that wish to use it.

Your issue seem to be that your optee node is placed after the dynamic
@atf-SEQ node, try place the optee node before the @atf-SEQ node.

Trying to use the common rockchip-u-boot.dtsi and TEE env var result in
the following:

  binman: Node '/binman/simple-bin/fit': subnode 'images/@tee-SEQ': Failed to read ELF file: Magic number does not match

And I assume you are using a custom binman node because of such error?

Adding support for the rockchip bl31.bin blob format to binman and use
the common rockchip-u-boot.dtsi would be the proper way to fix this.

Regards,
Jonas