dantezy

Could you provide the kernel, os, and the bcc version?

same problem as `BPF program is too large. Processed 1000001 insn` The kernel version is our company's internal version ,but we didn't modify the verifier.c. ``` System OS: Linux 6.6.0-47.001-rc1-xxxxxx...

As the verifier log for the minimal repro: ``` returning from callee: frame1: R0_w=0 R1_w=0xc820 R2=map_key(map=AT_,ks=8,vs=8) R3=map_value(map=AT_,ks=8,vs=8) R4=fp[0]-8 R10=fp0 cb to caller at 21: R0=scalar() R1=map_ptr(map=AT_,ks=8,vs=8) R2=func() R3=fp-8 R4=0 R6=1...

Cool! Maybe it can also resolve the https://github.com/bpftrace/bpftrace/issues/3308 But also we need to consider the older version kernels.

I have written a `bpf_map_sum_elem_count` demo yesterday. And on x86_64 platform, I found the asm use `call -0x1` to invoke this kfunc, which is a indirect call. So I wonder...

> > @gustavogabaldon - Sorry, @aaliyahnl has already started work on this but please feel free to grab another task from the backlog. > > do you have any other...

@ajor > We might have an easier time fixing whatever the problems are if we rewrote it in C too. Do you mean we implement `strcontains` in a single bpf...

> > This could also use a few semantic analyser and runtime tests. > > Also a codegen unit test please! https://github.com/bpftrace/bpftrace/tree/master/tests/codegen Could I ask how to generate `call_path.ll`? I...

@jordalgo Hi, I have generate the call_path.ll with `BPFTRACE_UPDATE_TESTS=1 build/tests/bpftrace_test --gtest_filter="codegen.call_path"` I feel a little bit weird, since the code above generated the call_path.ll with `ast::CodegenLLVM.generate_ir()`, which has been modified...

> I think this PR still needs a semantic_analyser test or two. I will add them tomorrow.