BPF @ IETF 117 Follow Up

[David Vernet <void@manifault.com>]

Hello everyone,

We had our first official meeting of the BPF working group at IETF 117
on Monday in San Francisco. The meeting was very productive, so thank
you to all those who attended and participated. We're excited to
continue making progress on standardizing BPF with the IETF, and
hopefully we can carry forward the momentum of the conference as we
iterate on the many topics that were discussed.

The meeting can be viewed in its entirety on YouTube at [0], and the
meeting minutes can be accessed at [1]. A special thank you to those who
collected notes, and to the presenters.

[0]: https://www.youtube.com/watch?v=jTtPbJqfYwI
[1]: https://notes.ietf.org/notes-ietf-117-bpf

Meeting Recap
-------------

Let's go over some of the highlights from the meeting:

1. Issue Tracker

The WG expressed interest in using an issue tracker to ensure that any
points that have been raised for discussion are properly tracked. A few
different options were proposed, with the final rough consensus being
that the chairs would pick an issue tracker and workflow for the WG.
We'll get to work on this once the conference has concluded, and will
notify you when we have everything setup.

2. eBPF Instruction Set Verification I-D called for adoption

A call for adoption took place for Dave Thaler's eBPF Instruction Set
Verification I-D [2].

[2]: https://datatracker.ietf.org/doc/draft-thaler-bpf-isa/

Rough consensus was obtained at the meeting, and we'll follow up with a
formal call for adoption on the email list in the near future.

There are already some reviews and discussions following the meeting,
which is great to see. Please keep the reviews coming!

3. ISA Extension Policy

Dave Thaler led a discussion on what the policy should be for extending
the BPF Instruction Set Architecture (ISA) for future instructions which
are added after the initial ISA standard document is ratified. The
slides for this discussion can be found in [3].

[3]: https://datatracker.ietf.org/meeting/117/materials/slides-117-bpf-isa-extension-policy-03.pdf

We covered a lot of ground in this discussion, but we need to close the
loop on a few things such as whether the registry should be in the IANA
or the Linux kernel tree, what type of ISA registration policy to use,
etc.

4. All things ABI

We discussed ABI and BPF program interoperability in a number of
different contexts. Will Hawkins presented the slides in [4], and
indicated that he had begun work on an ABI document that he would send
to the bpf@ietf.org and bpf@vger.kernel.org lists sometime in the near
future.

[4]: https://datatracker.ietf.org/meeting/117/materials/slides-117-bpf-abi-00.pdf

Dave Thaler also presented on a topic related to ABI: the eBPF ELF
Profile Specification. The slides can be found in [5]. In Dave's
presentation, he pointed out that the btf.rst document [6] currently has
an "ELF File Format Interface" section that should likely be moved into
a separate document such as elf.rst. This is also the case for
instruction-set.rst, as mentioned in the thread in [7].

[5]: https://datatracker.ietf.org/meeting/117/materials/slides-117-bpf-elf-profile-specification-00.pdf
[6]: https://git.kernel.org/pub/scm/linux/kernel/git/bpf/bpf-next.git/tree/Documentation/bpf/btf.rst
[7]: https://mailarchive.ietf.org/arch/msg/bpf/-NrwjA_6EHQ8y83cMZZGzbT8-wc/

We agreed that the next step would be to move these ABI sections out of
any document that we're expecting to be a Proposed Standard, and into
one or more separate .rst files. While our immediate focus is on the ISA
document, getting a head-start on the ABI document(s) seems prudent if
folks have the bandwidth.

Closing Thoughts
----------------

Overall, the meeting seemed to go very well. We're still in the process
of learning how to effectively collaborate between the IETF and Linux
kernel communities, but it's encouraging to see the progress being made
on the ABI doc(s), and especially the ISA doc. Thank you everyone for
being patient as we navigate everything, and for setting a respectful
and collaborative tone for our WG. That said, if anyone is finding the
arrangement difficult or has feedback of any kind, please feel free to
reach out to us at bpf-chairs@ietf.org so we can help.

Lastly, we're excited to mention that the first ACM SIGCOMM workshop on
eBPF and Kernel Extensions will be taking place on September 10, 2023 at
Columbia University in New York, NY. The list of accepted papers [8]
looks very interesting, so consider attending if you'd like to learn
more about BPF, and see what kind of BPF-related research is taking
place.

[8]: https://conferences.sigcomm.org/sigcomm/2023/workshop-ebpf.html

Regards,
David and Suresh