Re: [RFC net-next 4/4] tcp: defer regular ACK while processing socket backlog

On Thu, 7 Sep 2023 20:05:29 +0200 Eric Dumazet wrote:

quoted

Interesting. Some folks at Meta were recently looking into parsing RPCs
in the kernel to avoid unnecessary wakeups. Poor man's KCM using BPF
sockmaps. Passing message size hints from the sender would solve so
many problems..  

Yes, RPC headers make things easier for sure.

(we internally have something similar named autolowat, where we parse
headers to set sk->sk_rcvlowat dynamically)

Could this be turned into a BPF hook, possibly?
That's basically what I suggested the BPF people did but they like 
to rewrite everything.