* Extending sockmap and future of L7 BPF framework
@ 2020-02-13 10:12 Jakub Sitnicki
0 siblings, 0 replies; only message in thread
From: Jakub Sitnicki @ 2020-02-13 10:12 UTC (permalink / raw)
To: lsf-pc; +Cc: Lorenz Bauer, bpf
Extending sockmap and future of L7 BPF framework
At Plumbers ‘19 we presented our plan to make socket dispatch
programmable using BPF . A key component is the ability to look up
sockets from a map. Our proof-of-concept patch-set used the sockarray
map designed for reuseport BPF programs. Based on feedback from the
community we’re doing work to transition to sockmap and sockhash.
Specifically, we are adding support for TCP sockets in TCP_LISTEN state
and UDP sockets in general. Additionally, we’ll let the reuseport BPF
use sockmap and sockhash in addition to sockarray.
We propose giving an overview of this work, and the user-space and BPF
API changes it entails. This work brings up questions revolving around
the future of L7 BPF frameworks and how it should evolve that we would
like to discuss:
- Should sockmap be extended further to hold sockets in any state?
Would this make the user-experience better?
- How to make sockmap and sockarray gracefully share sk_user_data,
i.e. the link from the sock object to the map?
- Is there a potential for code sharing between sockmap and sockarray?
Would that make maintenance easier?
- How should API redirecting to UDP sockets with sockmap look like,
considering use of connected UDP sockets is not recommended?
- Can we have a common helper for map-based redirect that works with all
map types that can hold sockets, i.e. sockmap, sockhash, and
^ permalink raw reply [flat|nested] only message in thread
only message in thread, back to index
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2020-02-13 10:12 Extending sockmap and future of L7 BPF framework Jakub Sitnicki
BPF Archive on lore.kernel.org
Archives are clonable:
git clone --mirror https://lore.kernel.org/bpf/0 bpf/git/0.git
# If you have public-inbox 1.1+ installed, you may
# initialize and index your mirror using the following commands:
public-inbox-init -V2 bpf bpf/ https://lore.kernel.org/bpf \
Example config snippet for mirrors
Newsgroup available over NNTP:
AGPL code for this site: git clone https://public-inbox.org/public-inbox.git