RFE: test kernel module loading using a module that is present by default

[infinisil]

Since 6.13, the arp_tables module is not available in the default kernel config anymore, so let's use a different module, because the test doesn't need any specific one.

Tested in Nix by adjusting NixOS/nixpkgs#465550

[LordGrimmauld]

I wonder whether we can instead locate a not-loaded-but-available kernel module automatically to prevent similar issues in the future? That of course risks a broken module being found by whatever heuristic introduced to get a loadable kernel module, so maybe just delaying this problem to some unknown future time is fine.

[rgbriggs]

I wonder whether we can instead locate a not-loaded-but-available kernel module automatically to prevent similar issues in the future? That of course risks a *broken* module being found by whatever heuristic introduced to get a loadable kernel module, so maybe just delaying this problem to some unknown future time is fine.

I've had the same thoughts and concerns since I first wrote the test. "sg" is as good as any, but still subject to uncertainty.

[pcmoore]

Perhaps the best option is to build and load our own dummy module? Look at what the selinux-testsuite does with the module_load test: https://github.com/SELinuxProject/selinux-testsuite/tree/main/tests/module_load