[Buildroot] [PATCH v2 1/1] configs/stm32mp157*_dk*: bump TF-A to 2.10, Linux to 6.12.22 and U-Boot to 2025.04

[Raphaël Gallais-Pou]

Le 13/04/2025 à 12:37, Dario Binacchi a écrit :
> Hello Raphael,
> 
> On Fri, Apr 11, 2025 at 4:12 PM Raphaël Gallais-Pou
> <rgallaispou at gmail.com> wrote:
>>
>>
>>
>> Le 11/04/2025 à 10:58, Dario Binacchi a écrit :
>>> The patch bumps Arm Trusted Firmware (TF-A) to version 2.10 LTS the Linux
>>> kernel to version 6.12.22 and U-Boot to version 2025.04.
>>> The new version of U-Boot has added EFI Capsule support, which depends on
>>> GnuTLS, thus requiring the activation of BR2_TARGET_UBOOT_NEEDS_GNUTLS.
>>>
>>> Tested on STM32MP157C-DK2 Discovery Board.
>>>
>>> Signed-off-by: Dario Binacchi <dario.binacchi at amarulasolutions.com>
>>
>> Hi Dario,
>>
>> This patch tests fine on STM32MP15.  However it breaks the build on
>> STM32MP13.
> 
> As far as I know, my patch doesn't break the STM32MP13 configuration build.
> Instead, I noticed that this configuration shows hash errors for OP-TEE.
> I’ve sent a patch for that:
> 
> https://patchwork.ozlabs.org/project/buildroot/patch/20250413095900.1940373-1-dario.binacchi@amarulasolutions.com/
> 
> Could you let me know what kind of error you saw?

This is related to the "optee-os" and "optee-client" hash errors.

Actually, after further digging, your patch is not faulty at all.  The 
bug was introduced when upstreaming the new defconfig.  This option was 
implemented in between revision 2 and 3.  At the time optee was 
downloaded via the tarball, and then it was changed to the git method. 
Not long after the v4, the BR2_TARGET_OPTEE_OS_CUSTOM_VERSION option was 
added, on which I rebased for the v5, and forgot to update the hashes.

I guess it worked on my laptop even though the config was broken because 
I did not cleaned the dl/ folder.  Lesson learned as it turns out it 
was, in fact, my mistake.

> 
>> Now that STM32MP15 and STM32MP13 have been tied through this
>> patch[1] and the introduction of the later board [2], Linux and U-Boot
>> components need to be updated on both platforms at the same time.
> 
> As far as I know, I don't think it's necessary.
> Different sha256sums can be used within the same file.
Indeed you are right.  I read the documentation more thoroughly and 
Buildroot looks for "one or more" hashes.  The downloaded files that 
does not match the hashes get deleted.  So this means that for a shared 
hash file between multiple configs, several versions of the package can 
coexist.

> Anyway, I’ve also updated the STM32MP13 configuration regarding
> U-Boot and Linux as you requested.
> 
> https://patchwork.ozlabs.org/project/buildroot/patch/20250413095900.1940373-2-dario.binacchi@amarulasolutions.com/

Many thanks, I will review the serie promptly.

Regards,
Raphaël>
> Thanks and regards,
> Dario
> 
>>
>> Based on your changes, I tested TF-A 2.12 + U-Boot 2025.04 + Linux
>> 6.12.22 on STM32MP13 and it did alright.
>>
>> Can you update the STM32MP13 config ? Doing so will also remove the
>> second linux.hash and uboot.hash.
> 
>>
>> Thanks :)
>>
>> Regards,
>> Raphaël
>>
>>
>> [1]
>> https://git.buildroot.net/buildroot/commit/?id=f21db2bd83bf3073a50b8bcd79af6e7aed6fb000
>> [2]
>> https://git.buildroot.net/buildroot/commit/?id=9802fa0b5baa0b68fbc67bc07a2e4206012f7306
> 
> 
>