bas smit
bas smit
as we're root anyway, maybe we can set the file limit to N*MAX_PROBES to prevent this too
right now: - each probe is 2 fds, one for the prog and one for the probe - each map is one fd - hand full for default fds, perf_event,...
i attached artifacts but I currently dont have an x86 machine to test with so if anyone could double check it would be great :)
I might be missing some info but what is the use of `get_field` why not use the regular struct field access `a.b` or `a->`?
ah yes why you added it makes sense now. The semantic analyser doesn't see it as a guard. I do agree that `get_field` is a lot easier to do but...
> may i ask you a question about #2216, > > i just add `offsetof(element, struct)`, and it's OK, but add `offsetof(struct, element)`, it's failed. > > in `src/parser.yy` add:...
Nice, what are your plans for this? Any reason it has to stay hidden behind a compile flag and not behind runtime detection (maybe a runtime feature flag tho)?
Would be nice to land this. Having feature detection + a flag (`--bpfringbuf`) sounds like a good way forward.
Did you try `#include ` or what ever the correct import is?
Which os/kernel are you on? `bpf_prog_info info’ has initializer but incomplete type` this type comes from bpf.h shipped with the kernel, is it available on your system? (`grep -ri bpf_prog_info...