bpf: nodeport: wire up ext_err in revSNAT path

[julianwiedmann]

We've recently added error handling for the snat_*_rev_nat() calls from nodeport context. Now also wire up the reporting for any returned ext_err.

[julianwiedmann]

Note that I don't expect this to get any actual use today. We really shouldn't need to create CT entries from the revSNAT path :). But it's good future-proofing.

[julianwiedmann]

/test

[gentoo-root]

We really shouldn't need to create CT entries from the revSNAT path

That makes me think: maybe we should add an error counter and increase it if we happen to call ct_create from revSNAT?

[julianwiedmann]

We really shouldn't need to create CT entries from the revSNAT path

That makes me think: maybe we should add an error counter and increase it if we happen to call ct_create from revSNAT?

Yeah definitely good to have another look once your PR(s) are in. I'm not sure what this ct_create() would even do for us ... it's in the wrong direction, so would it still help to prevent GC for the revSNAT entry? (and why is there even a revSNAT entry, but no corresponding CT entry?!)

So at best we might even be able to just rip out the ct_create() for the revSNAT path. Which would be a nice reduction in instruction count I believe.