libebpf

Features:

• Full featured eBPF intepreter, support all features listed on https://docs.kernel.org/bpf/standardization/instruction-set.html
• External helpers support
• Passes all unit tests of ubpf vm tests

How to use?

Add this directory into your CMake project, and links target libebpf. See libebpf*.h for details.

How to run vm tests?

• Build CMake target libebpf_test_runner
• cd vm-test && python3.8 -m venv env
• source ./env/bin/activate
• pip install -r requirements.txt
• pytest

How to run ebpf_state tests?

• Build CMake target ebpf_state_test
• Run ./build/execution-test/ebpf_state_test

Notes on ebpf_state

ebpf_vm is just a virtual machine, it doesn't have ability to manage or access maps. We provide ebpf_state for such operations. You may create a ebpf_state and create maps in it. Before running your ebpf programs using ebpf_vm, you should call ebpf_state__setup_internal_helpers to setup internal helpers (such as bpf_map_lookup_elem) to the vm, and should also point ebpf_state__thread_global_context to the context you want to use. Then you can run your ebpf programs with access to the context.

Notes on FFI

You may call ebpf_state__register_ffi_function to register a FFI function. FFI functions are type safe, which means that it always accept int64_t and returns int64_t at the bpf side, but it would convert the received argument to the proper types when calling the FFI function. You may use some helper macros for FFI callling in the BPF side. see libebpf_ffi.bpf.h for details. You may use LIBEBPF_EXPORT_FUNCTION_ARG[N] (where [N] is an integer range 0 to 6) to export a function and automatically registers when the ebpf_state was created. This macro is only applicatable in the target libebpf, since it needs a modified link process