On 12/10/2016 11:15 PM, David Ahern wrote:
On 12/10/16 2:21 PM, Daniel Borkmann wrote:
Please name it bpf_prog_create() then, it would be consistent to
bpf_map_create() and shorter as well.
Sorry, lack of coffee, scratch that.
Can't the current bpf_prog_attach() stay as is, and you name the above new
functions bpf_prog_attach_fd() and bpf_prog_detach_fd()? I think that would
be better.
ok. no concerns about consistency with libbpf in the kernel repo?
Seems like making iproute2 and the kernel version the same will allow samples
and code to move between them much easier.
I think the lib/bpf.c code is quite different anyway, so I don't think it's
much of a concern or even requirement to look exactly the same as the samples
code (it was also never designed with such requirement). But besides that,
it's also trivial enough from reading the code due to the BPF_PROG_ATTACH
and BPF_PROG_DETACH anyway.
