On 10/14/26 15:45, [email protected] wrote: > - How the issue was discovered
Manual source review of the vnifilter netlink path, after noticing that vxlan_mdb.c range-validates its VNI attributes while vxlan_vnifilter.c does not. Prepared with LLM assistance, as the Assisted-by: tags say. > - Whether the issue was actually triggered, or is only theoretical Triggered and measured in QEMU, from a user namespace holding CAP_NET_ADMIN only inside it. There is no stack trace to quote: the symptoms are rtnl hold time and uncharged memory growth, not a crash. The changelogs carry the numbers. The cover letter states this is a policy and hardening change rather than a fix, targeting net-next with no Fixes tag, at Jakub's request: https://lore.kernel.org/netdev/[email protected]/
