dropbox/goebpf

Go eBPF

A nice and convenient way to work with eBPF programs / perf events from Go.

Requirements

• Go 1.11+
• Linux Kernel 4.15+

Supported eBPF features

• eBPF programs
  • SocketFilter
  • XDP
  • Kprobe / Kretprobe
  • tc-cls (tc-act is partially implemented, currently)
• Perf Events

Support for other program types / features can be added in future. Meanwhile your contributions are warmly welcomed.. :)

Installation

# Main library
go get github.com/dropbox/goebpf

# Mock version (if needed)
go get github.com/dropbox/goebpf/goebpf_mock

Quick start

Consider very simple example of Read / Load / Attach

    // In order to be simple this examples does not handle errors
    bpf := goebpf.NewDefaultEbpfSystem()
    // Read clang compiled binary
    bpf.LoadElf("test.elf")
    // Load XDP program into kernel (name matches function name in C)
    xdp := bpf.GetProgramByName("xdp_test")
    xdp.Load()
    // Attach to interface
    xdp.Attach("eth0")
    defer xdp.Detach()
    // Work with maps
    test := bpf.GetMapByName("test")
    value, _ := test.LookupInt(0)
    fmt.Printf("Value at index 0 of map 'test': %d\n", value)

Like it? Check our examples

Perf Events

Currently library has support for one, most popular use case of perf_events: where eBPF map key maps to cpu_id. So eBPF and go parts actually bind cpu_id to map index. It maybe as simple as:

    // Define special, perf_events map where key maps to CPU_ID
    BPF_MAP_DEF(perfmap) = {
        .map_type = BPF_MAP_TYPE_PERF_EVENT_ARRAY,
        .max_entries = 128,     // Max supported CPUs
    };
    BPF_MAP_ADD(perfmap);

    // ...

    // Emit perf event with "data" to map "perfmap" where index is current CPU_ID
    bpf_perf_event_output(ctx, &perfmap, BPF_F_CURRENT_CPU, &data, sizeof(data));

And the go part:

    perf, err := goebpf.NewPerfEvents("perfmap")
    // 4096 is ring buffer size
    perfEvents, err := perf.StartForAllProcessesAndCPUs(4096)
    defer perf.Stop()

    for {
        select {
            case data := <-perfEvents:
                fmt.Println(data)
        }
    }

Looks simple? Check our full XDP dump example

Kprobes

Library currently has support for kprobes and kretprobes. It can be as simple as:

    // kprobe handler function
    SEC("kprobe/guess_execve")
    int execve_entry(struct pt_regs *ctx) {
      // ...
      buf_perf_output(ctx);
      return 0;
    }

And the go part:

	// Cleanup old probes
	err := goebpf.CleanupProbes()

	// Attach all probe programs
	for _, prog := range bpf.GetPrograms() {
		err := prog.Attach(nil)
	}

	// Create perf events
	eventsMap := p.bpf.GetMapByName("events")
	p.pe, err = goebpf.NewPerfEvents(eventsMap)
	events, err := p.pe.StartForAllProcessesAndCPUs(4096)
	defer events.Stop()

	for {
		select {
		case data := <-events:
			fmt.Println(data) // kProbe event
		}
	}

Simple? Check exec dump example

Good readings

• XDP Tutorials
• Cilium BPF and XDP Reference Guide
• Prototype Kernel: XDP
• AF_XDP: Accelerating networking
• eBPF, part 1: Past, Present, and Future
• eBPF, part 2: Syscall and Map Types
• Oracle Blog: A Tour of eBPF Program Types
• Oracle Blog: eBPF Helper Functions
• Oracle Blog: Communicating with Userspace