Re: [meta-arago] [PATCH] tisdk-uenv: Add instructions to load initramfs while booting linux

On 3/4/26 8:38 AM, Ryan Eatmon wrote:
> 
> 
> On 3/4/2026 12:32 AM, Moteen Shah wrote:
>>
>> On 03/03/26 22:49, Andrew Davis wrote:
>>> On 3/3/26 4:49 AM, Moteen Shah via lists.yoctoproject.org wrote:
>>>> Add instructions to load initramfs image and update the boot command
>>>> to use the image to boot the linux kernel.
>>>>
>>>> Signed-off-by: Moteen Shah <[email protected]>
>>>> ---
>>>>   .../recipes-tisdk/tisdk-uenv/tisdk-uenv/uEnv.txt  | 15 +++++++++++++++
>>>>   1 file changed, 15 insertions(+)
>>>>
>>>> diff --git a/meta-arago-distro/recipes-tisdk/tisdk-uenv/tisdk-uenv/uEnv.txt b/meta-arago-distro/recipes-tisdk/tisdk-uenv/tisdk-uenv/uEnv.txt
>>>> index 0af0eef3..5ac61b34 100644
>>>> --- a/meta-arago-distro/recipes-tisdk/tisdk-uenv/tisdk-uenv/uEnv.txt
>>>> +++ b/meta-arago-distro/recipes-tisdk/tisdk-uenv/tisdk-uenv/uEnv.txt
>>>> @@ -7,3 +7,18 @@
>>>>   #       default on all newer U-Boot images.  This also means that some
>>>>   #       variables such as bootdelay cannot be changed by this file since
>>>>   #       it is not evaluated until the bootcmd is run.
>>>> +
>>>> +# Initramfs configuration
>>>> +# Name of the initramfs file in the boot partition
>>>> +name_initramfs=ti-core-initramfs.cpio.xz
>>>> +
>>>> +# Command to load initramfs from boot partition (mmc device:partition)
>>>> +# Adjust the device number (0 or 1) based on your setup
>>>> +# mmcdev is set to 1 by default, partition 1 is the boot partition
>>>> +get_initramfs_mmc=fatload mmc ${mmcdev}:1 ${rdaddr} ${name_initramfs}
>>>> +
>>>
>>> This belongs in u-boot default env, it is basically the same as `get_fdt_mmc`
>>> and other helpers already there.
>>>
>>>> +# Load initramfs and set up variables
>>>> +load_initramfs=if run get_initramfs_mmc; then setenv _initramfs ${rdaddr}:${filesize}; setenv rd_spec ${_initramfs}; echo Initramfs loaded at ${rdaddr}, size ${filesize}; else echo WARNING: Initramfs not found, continuing without initramfs; setenv rd_spec -; fi
>>>> +
>>>
>>> This should be part of `mmcloados` in u-boot also.
>>
>>
>> mmcloados is only used by 'mmcboot' which is a differnt boot command other than bootcmd_ti_mmc?
>> We can consider adding the load_initramfs to 'get_kern_mmc' and we wont need to override the init_mmc with the change.
>>

Adding to `get_kern_mmc` wouldn't make sense, that is for loading the kernel as the
name suggests. If we don't go down the `mmcboot` path anymore then add it to whatever
path we do go down these days.

>>>
>>>> +# Custom boot command that loads initramfs before kernel
>>>> +init_mmc=run args_all args_mmc load_initramfs
>>>
>>> With the above added, you don't need to override `init_mmc` so
>>> this line goes away too.
>>>
>>> That leaves one line left, `name_initramfs`. How about just
>>> setting a sane default name in u-boot. We got away with something
>>> like this for K2 plats:
>>
>> We don't expect the name of the cpio to be variable, as the name of the cpio stays the same for all K3 platforms which is "ti-core-initramfs.cpio.xz", I can remove the variable altogether maybe and directly use the image name to load. Thoughts?
> 
> That is a very Yocto centric take.  But we also support armbian and other distros.  They may not want to name the file the same way we chose in meta-ti.  The question is, should we paint ourselves into a corner from the beginning or allow for some flexibility?
> 
> Plus, even in Yocto, a user could choose to not use ti-core-initramfs and instead make their own with their own name.  Making this more flexible feels more friendly in the long run.
> 

And you don't have to look too far, that encrypted boot series posted the other
day would have made use of a new named ti-encrypted-boot-initramfs.cpio.gz
(although the booting method would have been different)

> Just my two cents.
> 

Worth every penny :)

Andrew

> 
>> Regards,
>> Moteen
>>
>>>
>>> https://github.com/u-boot/u-boot/blob/master/board/ti/ks2_evm/k2e_evm.env#L12
>>>
>>> although I'd recommend a more generic filename.
>>>
>>> Andrew
>>>
>>>> \ No newline at end of file
>>>
>