A new quadratic formula now maps cgroup v1 CPU shares to cgroup v2 CPU weight. Why? Because the old linear approach messed with CPU fairness; especially at low share values. This fix nails prioritization where it counts.
It lands at the OCI runtime layer, live in runc v1.3.2 and crun v1.23, so containers finally get CPU weights that reflect reality, not rounding errors.
Big picture: Kubernetes and cgroup v2 never quite agreed on CPU math. This update closes that gap, giving schedulers sharper control and workloads cleaner isolation.
