Hi Andrea,
On Sun, Oct 11, 2026 at 09:23:09PM +0200, Andrea Mayer wrote:
> Extend srv6_end_flavors_test.sh with tests for the SRv6 End behavior
> with both the NEXT-C-SID and PSP flavors.
>
> The tests use two SRv6 Policies with C-SID containers, where every
> router after the ingress one applies End with the NEXT-C-SID and PSP
> flavors. To check PSP, set net.ipv6.conf.all.seg6_enabled to 0 on the
> hosts in setup_hs(). The hosts then drop any packet that still carries
> an SRH, and a ping to an address on the loopback of a host succeeds
> only if the router connected to that host has removed the SRH. The tests
> fail when PSP is not applied together with NEXT-C-SID.
>
> Signed-off-by: Andrea Mayer <[email protected]>
> ---
> .../selftests/net/srv6_end_flavors_test.sh | 229 +++++++++++++++++-
> 1 file changed, 226 insertions(+), 3 deletions(-)
>
> diff --git a/tools/testing/selftests/net/srv6_end_flavors_test.sh
> b/tools/testing/selftests/net/srv6_end_flavors_test.sh
> index 30a939392eea..baa1b09a1aab 100755
> --- a/tools/testing/selftests/net/srv6_end_flavors_test.sh
> +++ b/tools/testing/selftests/net/srv6_end_flavors_test.sh
> @@ -439,6 +549,7 @@ setup_rt_local_sids()
> local devname
> local nsname
> local neigh
> + local csid
>
> nsname="$(get_rtname "${rt}")"
>
> @@ -452,6 +563,14 @@ setup_rt_local_sids()
> add "${LOCATOR_SERVICE}:${neigh}::/32" \
> table "${LOCALSID_TABLE_ID}" \
> via "${net_prefix}::${neigh}" dev "${devname}"
> +
> + # set underlay network routes to reach the C-SID containers
> + # that start with a C-SID of the neighbor, for any Function.
> + csid="$(build_csid "${neigh}" 00)"
> + ip -netns "${nsname}" -6 route \
> + add "${LCBLOCK_ADDR}:${csid}::/${LCBLOCK_NODE_BLEN}" \
> + table "${LOCALSID_TABLE_ID}" \
> + via "${net_prefix}::${neigh}" dev "${devname}"
> done
>
> # Local End behavior (note that "dev" is a dummy interface chosen for
> @@ -461,6 +580,17 @@ setup_rt_local_sids()
> table "${LOCALSID_TABLE_ID}" \
> encap seg6local action End dev "${DUMMY_DEVNAME}"
>
> + # Local End behavior with the NEXT-C-SID and PSP flavors, bound to the
> + # C-SID of the router (e.g. fcbb:0:04a1::/48 on rt-4). The /48 prefix
> + # covers every C-SID container whose first C-SID is this one, whatever
> + # the Argument that follows.
> + csid="$(build_csid "${rt}" "${END_CSID_PSP_FUNC}")"
> + ip -netns "${nsname}" -6 route \
> + add "${LCBLOCK_ADDR}:${csid}::/${LCBLOCK_NODEFUNC_BLEN}" \
> + table "${LOCALSID_TABLE_ID}" \
> + encap seg6local action End flavors next-csid,psp \
> + lblen "${LCBLOCK_BLEN}" nflen "${LCNODEFUNC_BLEN}" \
> + dev "${DUMMY_DEVNAME}"
This is a End + PSP specific encapsulation. If there is a new version patch,
can we move this block
> @@ -494,6 +634,7 @@ __setup_rt_policy()
> local in_nsname
> local rt_nsname
> local policy=''
> + local container
> local function
> local fullsid
> local op_type
> @@ -520,6 +661,15 @@ __setup_rt_policy()
> behavior_cfg="End flavors psp"
> ;;
>
> + "next-csid,psp")
> + # the behavior is already set up on every router by
> + # setup_rt_local_sids.
> + container="$(build_csid_container "${node}" \
> + "${END_CSID_PSP_FUNC}")"
> + policy="${policy}${container},"
> + continue
> + ;;
> +
to here? So we can align with the other SRv6 Endpoint behavior?
Because later in the End.X PSP support we also need to cleanup and
re-setup with END.X.
If this version is applied (which is also good), I can do the modification
in my patch set.
Others looks good to me.
Tested-by: Hangbin Liu <[email protected]>
Reviewed-by: Hangbin Liu <[email protected]>
Thanks
Hangbin
