Hi Dan,
On 2018-03-12 22:29, Friedrich Daniel wrote:
Hello Jan,
let me quote my mail to Claudio, who asked more or less the same on March 7th(see mail below). Personally, hoped that someone from Intel/Altera has the guts to answer this question or at least the Maintainer of the socfpga. I'm by far not good at u-boot, so it is just hacked together what was needed for the project so far.
If any Maintainer reads this, I suggest removing the Arria10 from the current master branch as it is unusable(system does not even boot) for any one the way it is now.
Hello Claudio,
I tried it as well and failed. My personal guess is that Altera/Intel never pushed proper support into the official u-boot mainline. Even there master-branch (https://github.com/altera-opensource/u-boot-socfpga/tree/master) does not work. For "just a user" it is very frustrating getting a new board abut have to use old software. And it seems Intel does not care in providing documentation, up-to-date software or maintaining the released code. For example if you look closer into the source code in the socfpga_v2014.10_arria10_bringup-branch there is a module called cff which loads the rbf-file in early release mode. I could not find anything similar in the latest branches(neither from denx not altera). Also the images generated from the latest releases generates not the proper format needed to boot from an Arria10.
So for my needs I used (https://github.com/altera-opensource/u-boot-socfpga/tree/socfpga_v2014.10_ar...) which is the version used in the Altera/Intel tools But changed it to my needs (loading > rbf-file, kernel and devicetree from EXT, setting u-boot environment to my needs etc.) I compile everything with the toolchain from yocto V2.4. and so far it works good. My changes are made public here(https://github.com/nanosurf-ag/u-boot-socfpga) but this might still subject to change.
Hope this helps dan
Cheers Dan
I informed myself these days (ELC-NA...) about the situation from U-boot perspective. I'd still love to hear an official Intel statement as well.
For now I told our internal user that there is something to do (though apparently not a lot - with the right people on the code) in order to get a solution we can safely ship in a long living critical infrastructure system. The current situation is suboptimal, for a number of reasons, and that can easily disappoint our user about their hardware supplier.
Thanks, Jan