byPassRT: A Lightweight User-Space Packet Processing Runtime

byPassRT is a DPDK-inspired, kernel-bypass networking runtime built for Linux using AF_XDP and eBPF. It shifts the burden of packet processing out of the Linux kernel and into user space, enabling ultra-fast, zero-copy packet I/O via lock-free ring buffers.

🚀 Project Overview

Traditional Linux networking (via standard sockets or even AF_PACKET) is relatively slow due to multiple memory copies, kernel-space to user-space context switches, and interrupt handling overhead.

byPassRT overcomes this by utilizing modern Linux AF_XDP sockets. The network card driver places packets directly into a shared memory region (UMEM) accessible by our user-space C application.

Key Features

• True Kernel Bypass: Uses eBPF/XDP for an ultra-fast data path.
• Lock-Free Queues: Utilizes single-producer/single-consumer (SPSC) lock-free rings for RX and TX logic.
• Mempools & Packet Buffers: Uses fixed-size, pre-allocated memory pools (mbuf-lite) to avoid expensive dynamic memory allocation (malloc) during the critical fast-path.
• Polling Engine: Processes packets in "bursts" using a tight user-space loop to eliminate interrupt overhead.

📁 File Structure & Roles

• src/main.c: The core application and user-space Polling Engine. It binds to a network interface, enters a tight infinite loop, and continuously reads, processes (e.g., swaps MAC addresses for loopback), and transmits packets in bulk.
• src/af_xdp.c & include/api.h: The I/O Engine. Manages the setup of the AF_XDP socket, shared memory (UMEM), and exposes simplified burst-processing APIs (lite_rx_burst, lite_tx_burst, lite_free) to the main application.
• src/mempool.c & include/mempool.h: The Memory Manager. Implementing DPDK-style packet buffer pools (mbuf-lite). Pre-allocates fixed memory frames and provides extremely fast alloc() and free() index operations decoupled from system calls.
• xdp/xdp_prog.c: Placeholder for custom eBPF packet-filtering programs (hardware-level DDOS dropping) for future expansions.
• benchmarks/run_benchmark.sh: A stress-testing script using flood-pings to validate high-throughput performance of our lock-free queues.
• tests/setup_veth.sh: A helper script to generate virtual ethernet interface pairs (veth) so that the application can be safely tested locally without needing a dedicated physical high-speed NIC.
• Makefile: The build system using gcc and linking against libbpf and libxdp.

🛠️ How to Build and Run

1. Dependencies

Ensure you have the latest eBPF and XDP development libraries mapping to your kernel:

sudo apt-get install libbpf-dev libxdp-dev

2. Build the Project

make clean
make

3. Testing via Virtual Interfaces (veth)

Testing AF_XDP requires attaching a packet filter to a network interface ring. To avoid locking yourself out of your actual network, use virtual pairs.

1. Setup the test interfaces:

chmod +x tests/setup_veth.sh
./tests/setup_veth.sh

2. Run the Runtime: Wait for byPassRT to bind to the test interface (Requires sudo to attach BPF programs).

sudo ./byPassRT veth-bypass

3. Generate Test Traffic (in another terminal): Send ping traffic or use packet generators like scapy over the peer interface.

ping -I veth-peer 10.0.0.1

You will see live metrics tracking packet reception and transmission bursts directly on the byPassRT output terminal!