Hi. I have experimented to make a csc for Khadas Vim2, similar to the one we already have for Vim1. It works pretty well. But I just wanted to check with the dev community before incorporating it to the build system, how do you guys feel about it.

As far as I understand, both options (kvim1\kvim2) require complete removal of the regular u-boot-2015 from eMMC to run. This is a dangerous approach with serious problems for users. U-boot-2015 AML has a serious difference from RK\AW - it uses a specific solution for writing \ updating firmware in eMMC, which is only available in u-boot-2015. This is the main way to work with firmware on AML. Its absence will cause a flurry of complaints and tantrums from users when they encounter serious problems and start blaming Armbian.

On 11/22/2020 at 8:22 AM, balbes150 said:

Its absence will cause a flurry of complaints and tantrums from users when they encounter serious problems and start blaming Armbian.

I see your point. You are used to deal with some profile of TV-box user, with high expectations and little technical knowledge. I personally would not call "dangerous" the standard approach of erasing eMMC in order to replace it for Armbian, it is what we do in many other boards. I have personally done it, the method is documented in official Khadas documentation, as well as the method for flashing the original Android again. But, again, I see your point, that it can be somehow "dangerous" for someone who has no clue of what he is doing.

 

So I think a sane approach is to keep the kvim2.csc the same way as we have been doing all along with kvim1.csc, which has not produced many complaints so far. The option in the build script is there, but we don't publish images among the supported boards, nor specify the method to install them. That way, only users with enough knowledge to use the build script will have those images. At least, for the time being, this may change in the future of course.

 

Besides that, I have developed a new function for the bash script, that grabs the meson-gxm boot blobs from the amlogic-boot-fip repo, as well as some patch to fix CPU scheduling in AML S912. These things can be useful if/when we start to support the Tartiflette.

I'm making a PR with all these changes.