Detect if the Linux Kernel is in lockdown mode

[grcevski]

This PR adds a Linux Kernel lockdown mode validation check for the two bpf programs which use bpf_probe_write_user. If the kernel is running in lockdown mode (typically auto-enabled with Secure Boot), we now put an explanatory error message instead of 'failed to load bpf program error'.

Relates to #290

[edeNFed]

I am not sure we should completely not load the instrumentation in case the kernel is in lockdown mode.
For example, process-internal context propagation would still work.

What do you think about producing spans and adding an error to them that specify something like context propagation failed please change kernel lockdown mode. I think this can be a good tradeoff between producing the best data that we can during kernel lockdown but also not confusing the users with a log message they will not find.

[grcevski]

Sure, I think I can do that. The only way I've been able to do this kind of thing is by using defines, compiling different versions of the BPF programs and loading selectively. I know about the bpf_core_enum_value_exists helper, but I've never been able to use it successfully, the verifier complains regardless of the if statement. Does this make sense, or do you have another approach in mind?

[MrAlias]

@edeNFed what do you think about doing that in a follow up PR?

[grcevski]

I'm happy to work on extending this PR and following with another that does the partial trace support. It will be easier for review if we are doing it incrementally, but I don't mind to extend this one if that's preferred.

[edeNFed]

@grcevski @MrAlias I am good with both options. Whatever easier for you.