Hello,
It was very effective and one of the things I used to get to 10k+
routes. It's a really tiny patch.... bpf is universal... want it again?
I'll be glad to have a look, but I'm probably not going to merge it. If
anything deserves optimisation, it should be done in babeld itself, not by
modifying the traffic post hoc.
I understand the scenario as this:
1. Babeld sends a message to the kernel to change a route
2. The kernel does its thing to adjust the route
3. Since babeld is subscribed to an event feed containing route-updates,
this changed route is communicated to babeld again.
4. Babel will parse this information, use up CPU time but there is no
gain in this.
Just my two cents: I am with Dave on that one. Why bother with that
garbage inside babeld if we can cause the kernel to not send it back to
babeld in the first place?
Christof
_______________________________________________
Babel-users mailing list
[email protected]
https://alioth-lists.debian.net/cgi-bin/mailman/listinfo/babel-users
