On Tue, 03 Mar 2020 21:24:31 +0100 Toke Høiland-Jørgensen wrote: > I can see how the bpf_link abstraction helps by providing a single > abstraction for all the tracing-type attachments that are fd-based > anyway; but I think I agree with Daniel that maybe it makes more sense > to keep it to those? I.e., I'm not sure what bpf_link adds to XDP > program attachment? The expected_prev_fd field to replace a program > could just as well be provided by extending the existing netlink API? +1
