
Bypassing eBPF Limitations: How to Run XDP on Egress Traffic
It is widely believed that Linux's high-speed XDP mechanism only works on ingress, leaving you stuck with Traffic Control bottlenecks and sk_buff allocations for container egress. Yet the kernel does not classify packets by direction at all—it only checks whether a frame landed in an RX receive queue.
Loophole Labs engineers routed egress traffic through a standard veth pair, turned it into host ingress, and accelerated transmission 9.2 times faster than TC. Still, the moment you bypass the kernel, the very first switch silently drops your frames.






