Re: [RFC PATCH] zynqmp: Add binman description for SOM

[Michal Simek <michal.simek@amd.com>]

Hi Simon,

On 11/2/24 17:28, Simon Glass wrote:
> Hi Michal,
> 
> On Fri, 1 Nov 2024 at 14:52, Tom Rini <trini@konsulko.com> wrote:
>>
>> On Fri, Nov 01, 2024 at 02:09:38PM +0100, Michal Simek wrote:
>>> Hi Simon,
>>>
>>> On 11/1/24 09:27, Michal Simek wrote:
>>>>
>>>>
>>>> On 10/31/24 19:03, Simon Glass wrote:
>>>>> Hi Michal,
>>>>>
>>>>> On Wed, 9 Oct 2024 at 10:33, Michal Simek <michal.simek@amd.com> wrote:
>>>>>>
>>>>>> There is necessary to do some steps to compose boot images. These steps
>>>>>> were in scripts in layers for a while. That's why introduce description via
>>>>>> binman to simplify wiring and remove all scripting around.
>>>>>> This should make sure that everybody is up2date with the latest versions.
>>>>>>
>>>>>> The first step is to create fit image with DTBs with descriptions in
>>>>>> configuration node which is written as regular expression to match all SOM
>>>>>> versions.
>>>>>> Description is there for k24 and k26 in spite of low level psu_init
>>>>>> configuration is different. The reason is that it goes to u-boot.itb image
>>>>>> which is the same for k24 and k26.
>>>>>> u-boot.itb is another image which is generated. It is normally generated
>>>>>> via arch/arm/mach-zynqmp/mkimage_fit_atf.sh but this script is supposed to
>>>>>> be deprecated.
>>>>>> FIT image by purpose is using 64bit addresses to have default option to
>>>>>> move images to high DDR (above 4GB). TF-A and TEE are optional components
>>>>>> but in the most cases TF-A is present all the time and TEE(OP-TEE) is used
>>>>>> by some configurations too.
>>>>>>
>>>>>> 3rd generated image is boot.bin with updated user field which contains
>>>>>> version number. This image can be used with updated Image Selector
>>>>>> which supports A/B update mechanisms with rollback protection.
>>>>>>
>>>>>> 4th image is image.bin which binary file which contains boot.bin and
>>>>>> u-boot.itb together and can be programmed via origin Image Selector.
>>>>>> This image can be also used for creating one capsule which contains both
>>>>>> boot images (in SPL boot flow).
>>>>>>
>>>>>> Signed-off-by: Michal Simek <michal.simek@amd.com>
>>>>>> ---
>>>>>>
>>>>>> Currently I have this for testing purpose only to find missing bits and
>>>>>> pieces in binman for cases I want to support.
>>>>>>
>>>>>> This patch depends on
>>>>>> https://lore.kernel.org/r/ fbed0251437b61a2f7a85596d7403b5b9c8237c1.1728306322.git.michal.simek@amd.com
>>>>>>
>>>>>> ---
>>>>>>    arch/arm/dts/Makefile                |   1 +
>>>>>>    arch/arm/dts/zynqmp-som-binman.dts   | 224 +++++++++++++++++++++++++++
>>>>>>    arch/arm/mach-zynqmp/Kconfig         |  17 ++
>>>>>>    configs/xilinx_zynqmp_kria_defconfig |   2 +
>>>>>>    4 files changed, 244 insertions(+)
>>>>>>    create mode 100644 arch/arm/dts/zynqmp-som-binman.dts
>>>>>
>>>>> I'm pleased to see this. My only suggestion is to use '/bits/ 64'
>>>>> instead of the macros, for 64-bit values.
>>>>
>>>> When I was playing with it some time ago it didn't work but it works now
>>>> that's why no issue with it.
>>>
>>> One more thing on this one. I pretty much dislike that images are generated
>>> to u-boot root folder. Isn't there a way that they will be written to
>>> separate folder (deploy for example)?
>>
>> Or perhaps $(obj) ?
> 
> Yes $(obj) is where they end up today. I consider output files from
> binman to be on the same level as other files created by the build.
> Then again, it does end up a bit of a mess, when mixed with the output
> files.

Some cleanup would be good. I remember a lot of discussion about if people 
should be using u-boot or u-boot.elf. Answer was simple but if output images are 
in separate folder it would IMHO better to understand.

> I am not keen on 'deploy' as it nothing is being deployed - also it
> sounds like a satellite or military attack.

No problem with different name.

> 
> One long-standing issue is that intermediate files are created in the
> same directory. Perhaps we could put those in a binman-tmp directory?
> They are necessary when things go wrong, but don't need to be used all
> the time.

+1

But I think we should try to move also that images which users should be using 
to separate folder completely. Again it doesn't need to be done by default but a 
way to configure it would be good.

Thanks,
Michal