On Mon, Mar 2, 2020 at 6:50 PM Alexei Starovoitov
[off-list ref] wrote:

On Fri, Feb 28, 2020 at 02:39:46PM -0800, Andrii Nakryiko wrote:

quoted

+int bpf_link_new_fd(struct bpf_link *link)
+{
+     return anon_inode_getfd("bpf-link", &bpf_link_fops, link, O_CLOEXEC);
+}

...

quoted

-     tr_fd = anon_inode_getfd("bpf-tracing-prog", &bpf_tracing_prog_fops,
-                              prog, O_CLOEXEC);
+     tr_fd = anon_inode_getfd("bpf-tracing-link", &bpf_link_fops,
+                              &link->link, O_CLOEXEC);

...

quoted

-     tp_fd = anon_inode_getfd("bpf-raw-tracepoint", &bpf_raw_tp_fops, raw_tp,
-                              O_CLOEXEC);
+     tp_fd = anon_inode_getfd("bpf-raw-tp-link", &bpf_link_fops,
+                              &raw_tp->link, O_CLOEXEC);

I don't think different names are strong enough reason to open code it.
I think bpf_link_new_fd() should be used in all cases.

Oh, this got simplified from initial implementation after few rounds
of refactorings and I didn't notice that now I can just use
bpf_link_new_fd() here. Will update.