http://blog.ghanmi.me/tags/linux/Recent content in Linux on MLogHugo -- gohugo.ioen-usTue, 24 Mar 2026 01:38:01 +0100http://blog.ghanmi.me/posts/building-k8s-cni-plugin-with-ebpf/Tue, 03 Sep 2024 00:00:00 +0000http://blog.ghanmi.me/posts/building-k8s-cni-plugin-with-ebpf/Prerequisites A working Kubernetes cluster for testing (a single-node kind or kubeadm cluster works perfectly) clang/llvm for compiling eBPF programs, bpftool What We’re Building Our plugin will handle a single, well-defined scope: pod-to-pod connectivity on a single node. That means every pod gets an IP address, and any pod can reach any other pod on the same machine. We won’t handle multi-node routing, services, or network policy. each of those is a project unto itself.http://blog.ghanmi.me/posts/building-an-ebpf-tracing-tool/Mon, 12 Aug 2024 00:00:00 +0000http://blog.ghanmi.me/posts/building-an-ebpf-tracing-tool/What Is eBPF, and Why Should You Care? eBPF (extended Berkeley Packet Filter) started life as a packet filtering mechanism, but it has evolved into a general-purpose in-kernel virtual machine. The key properties that make it so powerful: Safety. eBPF programs are verified before they run. The kernel’s built-in verifier performs static analysis to guarantee your program terminates, doesn’t access out-of-bounds memory, and doesn’t crash the kernel. You get kernel-level access with userspace-level safety guarantees.