MPTCP: Extending kernel functionality with eBPF and Netlink
summary
MPTCP is upstream and uses IPPROTO_MPTCP and SKB extensions; user space can manage paths via Netlink for path events (iprout2, ipmptcp, NetworkManager 1.40) but Netlink cannot block the kernel and floods are possible on busy servers. Per-packet scheduling moves into eBPF packet scheduler territory with prototype patches and BPF ops, yet faces performance overhead, missing atomic helpers (read_once/write_once), and API/versioning questions. Security concern: BPF access to connection tokens can enable subflow creation.
AI-written summary · about
chapters (4)
- 0:00Introduction to MPTCP
- 2:22Understanding MPTCP concepts
- 9:13Extending kernel with Netlink
- 16:30BPF for packet scheduling
related talks
54mHow to [not] implement a [not so] new net protocol
28mMultipath TCP hands-on on Fedora
47mMultipath TCP: Breaking Today's Networks with Tomorrow's Protocols
24mExtending packetdrill to test MPTCP
28mSafe sharing of the network with eBPF
30mA Beginner's Guide to eBPF Programming for Networking
32meBPF: BPF kernel infrastructure
14mMTP: Transport for In-Network Computing