Re: [PATCH bpf-next 5/8] bpf: Add link-based BPF program attachment to network namespace

On Thu, May 28, 2020 at 11:09:20AM -0700, Andrii Nakryiko wrote:

quoted

quoted

quoted

+       net = rcu_dereference(net_link->net);
+       if (!net || !check_net(net)) {
+               /* Link auto-detached or netns dying */
+               ret = -ENOLINK;

This is an interesting error code. If we are going to adopt this, we
should change it for similar cgroup link situation as well.

Credit goes to Lorenz for suggesting a different error code than EINVAL
for this situation so that user-space has a way to distinguish.

I'm happy to patch cgroup BPF link, if you support this change.

Yeah, I guess let's do that. I like "link was severed" message for
that errno :) But for me it's way more about consistency, than any
specific error code.

I like ENOLINK idea as well.
Let's fix it there and patch older kernels via separate patch.

Overall the set looks great. Looking forward to v2.