RE: io_uring: BPF controlled I/O
From: David Laight <hidden>
Date: 2021-06-14 07:54:11
Also in:
bpf, io-uring, lkml
From: Victor Stewart
Sent: 07 June 2021 19:51
...
coincidentally i'm tossing around in my mind at the moment an idea for offloading the PING/PONG of a QUIC server/client into the kernel via eBPF. problem being, being that QUIC is userspace run transport and that NAT-ed UDP mappings can't be expected to stay open longer than 30 seconds, QUIC applications bare a large cost of context switching wake-up to conduct connection lifetime maintenance... especially when managing a large number of mostly idle long lived connections. so offloading this maintenance service into the kernel would be a great efficiency boon. the main impediment is that access to the kernel crypto libraries isn't currently possible from eBPF. that said, connection wide crypto offload into the NIC is a frequently mentioned subject in QUIC circles, so one could argue better to allocate the time to NIC crypto offload and then simply conduct this PING/PONG offload in plain text.
Hmmmm... a good example of how not to type emails. Thought, does the UDP tx needed to keep the NAT tables active need to be encrypted? A single byte UDP packet would do the trick. You just need something the remote system is designed to ignore. David - Registered Address Lakeside, Bramley Road, Mount Farm, Milton Keynes, MK1 1PT, UK Registration No: 1397386 (Wales)