Service interruption via unreleased sockmap socket references

Memory leakage in the Linux kernel allows local users to exhaust system memory through BPF reuseport socket error paths. bpf_sk_select_reuseport() treats a non-null sk_reuseport_cb as proof that a sockmap result is non-refcounted, so two error paths fail to release an established TCP socket reference. Local access and permission to exercise BPF sockmap reuseport selection are required; repeated failures can interrupt service by exhausting kernel memory.