On 4/22/22 5:48 PM, Daniel Borkmann wrote:
On 4/20/22 6:52 PM, Eyal Birger wrote:
[...]
Ok, makes sense given for BPF_OK the dst->dev shouldn't change here (e.g. as opposed to BPF_REDIRECT). Applied, please also follow-up with a BPF selftest for test_progs so that this won't break in future when it's running as part of BPF CI.
(Coverage for lwt BPF flavor from test_progs is afaik non-existent aside from some section name tests. Would be great in general to have runtime tests asserting lwt behavior there if you have a chance.)
