Re: [PATCH net] selftest/net: skip srv6_end_d[t/x][4/6]_*_test.sh if iproute2 too old

[Andrea Mayer]

On Thu, 27 Aug 2026 11:56:09 +0200
Alessio Faina <alessio.faina@xxxxxxxxxxxxx> wrote:

> [snip]
> 
> For the DX tests I was doing a local test but it's not in the original
> kernel, sorry, I must have confused 2 tasks with one, so disregard the
> DX4 and DX6 tests, they are not relevant to this issue; if we manage to
> agree on a solution for the DT4 and DT6 tests, I will discard those
> modification in a subsequent patch.

Ok, so this is about the dt4 and dt6 tests only.

> About the DT4 and DT6 tests, I thought to use the same method as the failure
> was practically the same as End.DT46.
> This is the version I'm currently using 
> ip utility, iproute2-ss200127 (should be 5.5.0 if I'm not mistaken)
> and I can confirm 
> [SEG6_LOCAL_ACTION_END_DT4]		= "End.DT4"
> [SEG6_LOCAL_ACTION_END_DT6]		= "End.DT6",
> is available.
> 
> What do you suggest would be the right way forward to check what is
> going on?

These tests need the vrftable attribute. The help does not list it
before v5.14.0, while ip parses it from v5.11.0, so a help check skips
tests that would run on v5.11.0, v5.12.0 and v5.13.0.
One possible way could be to try the route add in a throwaway netns and
skip the selftest if that fails. test_encap_lookup_supp_or_ksft_skip() in
srv6_encap_lookup_l3vpn_test.sh could provide the general idea, but here
the netns also needs the vrf device and the strict mode the test sets up.

That's more code than a help check, but it skips when the route really
cannot be installed. There is a trade-off to consider, and comments or
other ideas are very welcome.

Ciao,
Andrea