Daniel Borkmann
Daniel Borkmann
(lifting blocker since this has not been hit in quite a while, I'm reworking L4LB CI in general in OSS main so I'll still take care of it, thus leaving...
Cc @aditighag
Related upstream discussion: https://lore.kernel.org/bpf/[email protected]/ EPERM is however a documented error code: https://man7.org/linux/man-pages/man2/connect.2.html > thus preventing the kernel from entering an infinite retry loop I still think this is a bug...
I think remap would be better fit, pushed here with an explanation: https://lore.kernel.org/bpf/72b5939b13efd4fde6e9c0f9fb00edd314f4bcce.1719392816.git.daniel@iogearbox.net/
Once this gets merged into net tree, and shipped to Linus within a ~week, then stable team will pick this up. I don't have experience with Longhorn, but I'd assume...
Patch got merged here: https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=626dfed5fa3bfb41e0dffd796032b555b69f9cde And is being shipped to Linus today, then stable will pick it some time after automatically.
I'm looking into it, it looks like there seems a kernel regression somewhere, we've seen this also when trying to upgrade bpf kernel on our CI where it's working before...
Here's a known fix to a regression which slipped in: https://patchwork.kernel.org/project/netdevbpf/patch/[email protected]/ If you have a chance to try locally to see if this solves it would be awesome.
It's probably easier to just clone the git tree from net fixes : https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/log/ The patch got applied there (https://git.kernel.org/pub/scm/linux/kernel/git/netdev/net.git/commit/?id=157f29152b61ca41809dd7ead29f5733adeced19). I'll be travelling next week but will check once back.