linux-hardkernel: Init at 4.14.47-139

[qolii]

Motivation for this change

This introduces a build for the hardkernel vendor kernel tree, to support hardkernel hardware (e.g. the odroid series of SBCs).

Things done

I have built and tested this kernel on my Odroid HC1, using kernelPackages = pkgs.linuxPackages_hardkernel_latest. No problems encountered, but no further devices tested.

I just looked at existing linux kernel instances and duplicated the simplest I saw with the minimum novelty I could think of. I might definitely have missed something. Design advice is welcome.

• Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS)
• Built on platform(s)
  • NixOS
  • macOS
  • other Linux distributions
• Tested via one or more NixOS test(s) if existing and applicable for the change (look inside nixos/tests)
• Tested compilation of all pkgs that depend on this change using nix-shell -p nox --run "nox-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Fits CONTRIBUTING.md.

---

[andrew-d]

If you're interested, I had a similar-ish config that also automatically splits version numbers:

  # modDirVersion needs to be x.y.z.
  modDirVersion = head (stdenv.lib.splitString "-" version);

  # branchVersion needs to be x.y.
  extraMeta.branch = with stdenv.lib; concatStrings (intersperse "." (take 2 (splitString "." modDirVersion)));

And, thanks for submitting this PR! I've also been running NixOS on a ODROID HC2 for a couple months now with no trouble, but never got around to submitting a PR for kernel support - much appreciated 😀

[qolii]

@andrew-d, no problem! :)

I saw there are such tricks in the other kernel .nix files as well. But, is it a little bit overkill? It doesn't really provide much useful abstracting power in this case, and I think it does make something simple harder to read. But I guess it's only small. I'm happy to change it if feelings are strong.

[andrew-d]

I'm happy to change it if feelings are strong.

No strong feelings, no - just mentioning it if you were interested! 👍

[qolii]

Awesome, thanks!

[qolii]

Cc: @dezgeg, possibly.

[andrew-d]

Another thing I just noticed; do you want callPackages here, or callPackage (no s) like the other kernels?

[qolii]

Ugh, you're totally right. Thanks! Fixed.

[dezgeg]

This should use fetchFromGitHub to avoid possibly nondeterministic GitHub tarballs.

Looks good otherwise.

[qolii]

Ugh, is nothing sacred!! I assumed GitHub's tarballs at least would be deterministic.

Will fix. Thanks for looking!

[qolii]

Uhh, sorry about this, can you help me? I've changed it to

  src = fetchFromGitHub {
    owner = "hardkernel";
    repo = "linux";
    rev = version;
    sha256 = "1nxgd29y9wmifm11pjcdpd2y293p0dgi0x5ycis55miy97n0f5zy";
  };

which, as far as I can tell, is roughly the right way to do things. (The hash is wrong, but I was going to try to build it and then fix up the hash when it complained about it.)

However, I get the result that, when I run nixos-rebuild, I get:

these derivations will be built:
  /nix/store/5jcd3rxy9xmql16vvrx8vspinibqxmn2-source.drv
  /nix/store/hq0l73vjjy7py4g7cwrbd79mzzfnqvva-linux-config-4.14.47-139.drv
  /nix/store/kxwdk2f2lks3cannzqsw1iv634c75b8f-linux-4.14.47-139.drv
  /nix/store/phx3p9afjlkr1sl3pd194m4iq9j5n3q1-kernel-modules.drv
  /nix/store/sas9zlmpw0msri87mzx9qg1x9mh4vb3w-firmware.drv
  /nix/store/fa9dm9bbilinl9zkpw83yshvgrpsh7sz-kernel-modules-shrunk.drv
  /nix/store/02360f07416jlzgpkfqbn76zmrpl63z0-stage-1-init.sh.drv
  /nix/store/gigx079yqhiam6m9w7mxqkk1imwy2zwg-etc-nixos.conf.drv
  /nix/store/ic84x39phcwxxk439awdns9bymfyi05w-etc.drv
  /nix/store/lp5hwcwfks7f7iv5mfnjb9dhkswzl6kf-initrd.drv
  /nix/store/kaibkzq676y6alxs2nxwwmiwd3d9nncb-nixos-system-gospel-18.09.git.d6422c6.drv
building '/nix/store/5jcd3rxy9xmql16vvrx8vspinibqxmn2-source.drv'...

trying https://github.com/hardkernel/linux/archive/4.14.47-139.tar.gz
  % Total    % Received % Xferd  Average Speed   Time    Time     Time  Current
                                 Dload  Upload   Total   Spent    Left  Speed
100   129    0   129    0     0    390      0 --:--:-- --:--:-- --:--:--   390

at which point, it hangs forever. No error. That tar.gz path itself is fine:

# wget https://github.com/hardkernel/linux/archive/4.14.47-139.tar.gz
[successful noises]

works fine. However, I'm not sure if this is relevant, but:

# curl https://github.com/hardkernel/linux/archive/4.14.47-139.tar.gz
<html><body>You are being <a href="https://codeload.github.com/hardkernel/linux/tar.gz/4.14.47-139">redirected</a>.</body></html>

Anyway, I expect am making an obvious mistake! Any idea?

[qolii]

Alright, done. It wasn't hanging after all, just taking ages.

Seems to work for me.

[Mic92]

Please restrict the build to aarch64. It breaks my nix-review for kernel modules, because nix is trying to build it on x86_64.

[qolii]

@Mic92, sorry for the delay here. What is the most appropriate way to do this?

I see other kernel expressions using extraMeta.hydraPlatforms = [ ]; and extraMeta.hydraPlatforms = [ "armv7l-linux" ];. Would one of these do the trick?

[Mic92]

I assume:

extraMeta.hydraPlatforms = [ "armv7l-linux" ];

would do the trick then.

[Mic92]

Also something like extraMeta.platforms would be better because then also nix-review does not attempt to build it.

[Mic92]

done in 995d005