[LSF/MM/BPF ATTEND] BPF track

[LSF/MM/BPF ATTEND] BPF track

[Stanislav Fomichev]

Hi,

I'd like to participate in the BPF track of LSF/MM/BPF Summit 2020. I
work on BPF stuff at Google, I recently contributed the following
features to the upstream BPF subsystem:
* {set,get}sockopt hooks & BPF_SOCK_OPS_RTT_CB callback
* Optional socket storage cloning on accept (BPF_F_CLONE)
* Improvements to flow dissector BPF hook to make it work in driver
  (non-skb) context
* Various improvements to selftests / BPF_PROG_TEST_RUN framework

Topics I'd like to discuss:
* BTF/BPF meta-information: how to append enough info to BPF/BTF to
  recover build timestamp + commit sha1
* Global mode for cgroup storage as a way to have common/persistent
  per-cgroup scratch buffer
* Unit-testing and uBPF; uBPF integration with libbpf
* Average prog runtime; discuss any future plans to extend it to return a
  distribution, not average (with BPF?)

I don't think my topics deserve a dedicated 30-minutes session. I'm also
interested in general BPF direction and how we can apply future
potential features here at Google.

In particular, some of the proposed talks that are of particular
interest:
* BPF logging
* BPF unit testing