On the FOSDEM side, I spent most of my time between the eBPF Track and the Distributions Track (where I was an organizer). The eBPF room was packed all day, and the content really shifted from “experiments” to “infrastructure readiness.”

Three talks stood out:

A Unified I/O Monitoring Framework

This proposed a really smart way to trace VFS functions to monitor I/O across different platforms (HPC, Cloud, K8s) and filesystems (LUSTRE, NFS) uniformly. It effectively decouples monitoring from the underlying storage layer, which is a huge win for observability in complex environments.

Aya — what’s new in Rust for eBPF?

This connected directly to my hallway discussions. The big news is the push to promote Rust’s eBPF targets to Tier 2, which would finally allow us to build eBPF programs on stable Rust without needing nightly toolchains.

String kfuncs — simplifying string handling

Handling strings in eBPF has always been painful, but this talk covered the new kfuncs added in kernel 6.17. It was great to see how tools like bpftrace have already adopted them to make string processing much more ergonomic and performant.