Volumetric Distributed Denial of Service (DDoS) attacks, such as UDP reflection (NTP/DNS amplification) and TCP SYN floods, easily saturate standard Linux network stacks. Traditional user-space firewalls and even iptables / nftables fail because the kernel must allocate a full sk_buff (socket buffer) data structure for every incoming packet before evaluating rules.
At 14 million packets per second (Mpps), CPU cores become entirely saturated by packet allocation overhead, locking up servers before a single byte reaches your application.
eXpress Data Path (XDP) powered by eBPF (Extended Berkeley Packet Filter) revolutionizes packet processing by executing sandboxed bytecode directly inside the Network Interface Card (NIC) driver layer—before memory allocation takes place.
1. The Performance Gap: Traditional Stack vs. XDP
2. Core XDP Action Codes
When an eBPF program attached to an XDP hook evaluates a raw packet buffer (struct xdp_md), it returns one of five fast action verdicts:
XDP_DROP: Instantly discards the packet at the driver layer without passing it upstream.XDP_PASS: Hands the packet to the standard Linux network stack for normal processing.XDP_TX: Bounces the packet back out the same interface it arrived on (ideal for wire-speed Layer 4 load balancers like Meta's Katran).XDP_REDIRECT: Bypasses the local stack and forwards the packet to another NIC or virtual interface.XDP_ABORTED: Drops the packet and triggers an eBPF tracepoint for error inspection.
3. Production eBPF XDP Filter Implementation (C Code)
#include <linux/bpf.h>
#include <linux/if_ether.h>
#include <linux/ip.h>
#include <linux/udp.h>
#include <bpf/bpf_helpers.h>
// BPF Map to track and block malicious IP subnets dynamically
struct {
__uint(type, BPF_MAP_TYPE_HASH);
__uint(max_entries, 1000000);
__type(key, __u32); // IPv4 Address
__type(value, __u64); // Drop Counter
} blocked_ips SEC(".maps");
SEC("xdp_ddos_filter")
int xdp_prog(struct xdp_md ctx) {
void data_end = (void )(long)ctx->data_end;
void data = (void )(long)ctx->data;
struct ethhdr eth = data;
if ((void )(eth + 1) > data_end)
return XDP_PASS;
if (eth->h_proto != __constant_htons(ETH_P_IP))
return XDP_PASS;
struct iphdr iph = (void )(eth + 1);
if ((void )(iph + 1) > data_end)
return XDP_PASS;
__u32 src_ip = iph->saddr;
__u64 drop_count = bpf_map_lookup_elem(&blocked_ips, &src_ip);
if (drop_count) {
__sync_fetch_and_add(drop_count, 1);
return XDP_DROP; // Drop packet in 25 CPU cycles!
}
// Mitigate NTP Amplification: Drop UDP packets from port 123 with size > 400 bytes
if (iph->protocol == IPPROTO_UDP) {
struct udphdr udph = (void )iph + (iph->ihl 4);
if ((void *)(udph + 1) > data_end)
return XDP_PASS;
if (udph->source == __constant_htons(123) && (data_end - data) > 400) {
return XDP_DROP;
}
}
return XDP_PASS;
}
char _license[] SEC("license") = "GPL";
4. Attaching the XDP Program to the Network Driver
# Compile to eBPF ELF Object
clang -O2 -target bpf -c xdp_filter.c -o xdp_filter.o
# Attach in Native Driver Mode to interface eth0
ip link set dev eth0 xdpgeneric off
ip link set dev eth0 xdp obj xdp_filter.o sec xdp_ddos_filter
# Verify Active Hook
ip link show dev eth0
5. Frequently Asked Questions (FAQ)
What network cards support native XDP driver mode?
Most modern enterprise network cards support native XDP, including Intel (i40e, ixgbe, igb), Mellanox (mlx4, mlx5), Broadcom (bnxt), and AWS ENA.
Can XDP be used for Layer 4 Load Balancing as well as DDoS protection?
Yes. Prominent infrastructure providers like Meta (Facebook Katran) and Cloudflare use XDP to implement line-rate Maglev-based Layer 4 load balancers handling millions of connections with sub-millisecond latencies.