tc based bpf filter

Setup

1. Setup clang and llc and bpftool

$ sudo apt-get install make gcc libssl-dev bc libelf-dev libcap-dev \
	clang gcc-multilib llvm libncurses5-dev git pkg-config libmnl-dev \
	bison flex graph viz
$ sudo apt install linux-tools-generic

1. Setup golang -

$ export GOVERSION=1.15.8
$ mkdir /tmp/golang;
$ pushd; cd /tmp/golang;
$ wget https://dl.google.com/go/go$\{GOVERSION\}.linux-amd64.tar.gz
$ wget https://dl.google.com/go/go${GOVERSION}.linux-amd64.tar.gz
$ tar -xzf go${GOVERSION}.linux-amd64.tar.gz
$ mv go /usr/local/go${GOVERSION}
$ ln -sfn /usr/local/go${GOVERSION} /usr/local/go
	

1. Setup iproute2 -

$ git clone https://git.kernel.org/pub/scm/network/iproute2/iproute2.git
$ cd iproute2
$ ./configure --prefix=/usr
$ sudo make install
# Copy the `bpf_api.h` helpers file that lives under `./include`
# to your `/usr/include` directory.
#
# ps.: this could be anywhere - including your current source tree.
$ install -m 0644 ./include/bpf_api.h /usr/include/iproute2

Running

  $ make
  $ ./bin/main
  $ make drop-install

This repo is based on florianl/go-tc

Overview

After the eBPF code is loaded from ebpf/drop the eBPF program ingress_drop is loaded into the kernel. In a next step this PoC creates a dummy interface. So it does not alter existing configurations or network interfaces. Then a qdisc and filter are attached via the netlink interface of the kernel to this dummy interface. The file descriptor of the eBPF program ingress_drop is passed as argument of the filter to the kernel. With attaching the filter to the interface the eBPF program ingress_drop will run on every packet on the interface.

Privileges

This PoC uses the netlink and eBPF interface of the kernel and therefore it requires special privileges. You can provide this privileges by adjusting the CAP_NET_ADMIN capabilities.