On Fri, Jun 26, 2020 at 12:17:40PM -0500, Eric W. Biederman wrote:

quoted

I'm swamped with other stuff today and will test the set Sunday/Monday
with other patches that I'm working on.
I'm not sure why you want to rename the interface. Seems
pointless. But fine.

For maintainability I think the code very much benefits from a clear
separation between the user mode driver code from the user mode helper
code.

you mean different name gives that separation? makes sense.

quoted

As far as routing trees. Do you mind I'll take it via bpf-next ?
As I said countless times we're working on bpf_iter using fork_blob.
If you take this set via your tree we would need to wait the whole kernel release.
Which is 8+ weeks before we can use the interface (due to renaming and overall changes).
I'd really like to avoid this huge delay.
Unless you can land it into 5.8-rc2 or rc3.

I also want to build upon this code.

How about when the review is done I post a frozen branch based on
v5.8-rc1 that you can merge into the bpf-next tree, and I can merge into
my branch.  That way we both can build upon this code.  That is the way
conflicts like this are usually handled.

sure. that works too.

Further I will leave any further enhancements to the user mode driver
infrastructure that people have suggested to you.

ok