contrib/kind: enable XDP_TX from pod veth

Both veth in a pair require an XDP program installed for XDP_TX
to work. Since the host side veth created by kind doesn't have
an XDP program attached we can't run any tests in CI that require
XDP_TX.

The workaround itself is just an ip link set and ethtool away,
the problem is figuring out which interfaces we need to do the
magic to.

Use the approach used by kind-network-plugins and create our own
docker network with a specific name for the bridge device. We
can then iterate all children of the bridge and do our fixups.

We tell kind to use our own network by setting the (undocumented?)
KIND_EXPERIMENTAL_DOCKER_NETWORK environment variable.

See https://github.com/aojea/kind-networking-plugins

Signed-off-by: Lorenz Bauer <lmb@isovalent.com>

.github/workflows/conformance-datapath-v1.13.yaml

@@ -286,15 +286,15 @@ jobs:
           provision: 'false'
           cmd: |
             cd /host/
-            ./contrib/scripts/kind.sh "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
+            ./contrib/scripts/kind.sh --xdp "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
             ./cilium-cli install ${{ steps.vars.outputs.cilium_install_defaults }}
 
             ./cilium-cli status --wait
             ./cilium-cli connectivity test --datapath --collect-sysdump-on-failure \
               --sysdump-output-filename "cilium-sysdump-${{ matrix.name }}-<ts>"
             ./cilium-cli connectivity test --collect-sysdump-on-failure \
               --sysdump-output-filename "cilium-sysdump-${{ matrix.name }}-<ts>"
-            kind delete cluster
+            ./contrib/scripts/kind-down.sh --keep-registry
 
       - name: Run encryption tests
         if: ${{ matrix.encryption != 'disabled' }}
@@ -303,7 +303,7 @@ jobs:
           provision: 'false'
           cmd: |
             cd /host/
-            ./contrib/scripts/kind.sh "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
+            ./contrib/scripts/kind.sh --xdp "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
             ./cilium-cli install ${{ steps.vars.outputs.cilium_install_defaults }} \
               --encryption=${{ matrix.encryption }}
 

.github/workflows/conformance-datapath.yaml

@@ -379,7 +379,7 @@ jobs:
           provision: 'false'
           cmd: |
             cd /host/
-            ./contrib/scripts/kind.sh "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
+            ./contrib/scripts/kind.sh --xdp "" 3 "" "" "${{ matrix.kube-proxy }}" "dual"
             ./cilium-cli install ${{ steps.vars.outputs.cilium_install_defaults }}
 
             ./cilium-cli status --wait
@@ -389,7 +389,7 @@ jobs:
               --sysdump-output-filename "cilium-sysdump-${{ matrix.name }}-<ts>"
             ./cilium-cli connectivity test --collect-sysdump-on-failure \
               --sysdump-output-filename "cilium-sysdump-${{ matrix.name }}-<ts>"
-            kind delete cluster
+            ./contrib/scripts/kind-down.sh --keep-registry
 
       - name: Fetch artifacts
         if: ${{ !success() }}

.gitignore

@@ -115,3 +115,6 @@ images/*/Dockerfile.dockerignore
 
 # Clangd cache for indexed bpf code
 bpf/.cache
+
+# Include dummy bpf object necessary for XDP_TX
+!test/l4lb/bpf_xdp_veth_host.o

contrib/scripts/kind-down.sh

@@ -9,6 +9,10 @@ if ! have_kind; then
     echo "  https://kind.sigs.k8s.io/docs/user/quick-start/#installation"
 fi
 
+if [ "${1:-}" != "--keep-registry" ]; then
+    docker kill kind-registry && \
+        docker rm kind-registry
+fi
+
 kind delete clusters kind && \
-docker kill kind-registry && \
-docker rm kind-registry
+    docker network rm kind-cilium

contrib/scripts/kind.sh

@@ -14,8 +14,17 @@ default_pod_subnet=""
 default_service_subnet=""
 default_agent_port_prefix="234"
 default_operator_port_prefix="235"
+default_network="kind-cilium"
 
 PROG=${0}
+
+xdp=false
+if [ "${1:-}" = "--xdp" ]; then
+  xdp=true
+  shift
+fi
+readonly xdp
+
 controlplanes="${1:-${CONTROLPLANES:=${default_controlplanes}}}"
 workers="${2:-${WORKERS:=${default_workers}}}"
 cluster_name="${3:-${CLUSTER_NAME:=${default_cluster_name}}}"
@@ -27,10 +36,13 @@ pod_subnet="${PODSUBNET:=${default_pod_subnet}}"
 service_subnet="${SERVICESUBNET:=${default_service_subnet}}"
 agent_port_prefix="${AGENTPORTPREFIX:=${default_agent_port_prefix}}"
 operator_port_prefix="${OPERATORPORTPREFIX:=${default_operator_port_prefix}}"
+
+bridge_dev="br-${default_network}"
+v6_prefix="fc00:c111::/64"
 CILIUM_ROOT="$(git rev-parse --show-toplevel)"
 
 usage() {
-  echo "Usage: ${PROG} [control-plane node count] [worker node count] [cluster-name] [node image] [kube-proxy mode] [ip-family]"
+  echo "Usage: ${PROG} [--xdp] [control-plane node count] [worker node count] [cluster-name] [node image] [kube-proxy mode] [ip-family]"
 }
 
 have_kind() {
@@ -117,6 +129,16 @@ workers() {
   done
 }
 
+# create a custom network so we can control the name of the bridge device.
+# Inspired by https://github.com/kubernetes-sigs/kind/blob/6b58c9dfcbdb1b3a0d48754d043d59ca7073589b/pkg/cluster/internal/providers/docker/network.go#L149-L161
+docker network create -d=bridge \
+  -o "com.docker.network.bridge.enable_ip_masquerade=true" \
+  -o "com.docker.network.bridge.name=${bridge_dev}" \
+  --ipv6 --subnet "${v6_prefix}" \
+  "${default_network}"
+
+export KIND_EXPERIMENTAL_DOCKER_NETWORK="${default_network}"
+
 # create a cluster with the local registry enabled in containerd
 cat <<EOF | ${kind_cmd} --config=-
 kind: Cluster
@@ -144,7 +166,27 @@ containerdConfigPatches:
     endpoint = ["http://${reg_name}:${reg_port}"]
 EOF
 
-docker network connect "kind" "${reg_name}" || true
+if [ "${xdp}" = true ]; then
+  if ! [ -f "${CILIUM_ROOT}/test/l4lb/bpf_xdp_veth_host.o" ]; then
+    pushd "${CILIUM_ROOT}/test/l4lb/" > /dev/null
+    clang -O2 -Wall -target bpf -c bpf_xdp_veth_host.c -o bpf_xdp_veth_host.o
+    popd > /dev/null
+  fi
+
+  for ifc in /sys/class/net/"${bridge_dev}"/brif/*; do
+    ifc="${ifc#"/sys/class/net/${bridge_dev}/brif/"}"
+
+    # Attach a dummy XDP prog to the host side of the veth so that XDP_TX in the
+    # pod side works.
+    sudo ip link set dev "${ifc}" xdp obj "${CILIUM_ROOT}/test/l4lb/bpf_xdp_veth_host.o"
+
+    # Disable TX and RX csum offloading, as veth does not support it. Otherwise,
+    # the forwarded packets by the LB to the worker node will have invalid csums.
+    sudo ethtool -K "${ifc}" rx off tx off > /dev/null
+  done
+fi
+
+docker network connect "${default_network}" "${reg_name}" || true
 
 for node in $(kubectl get nodes --no-headers -o custom-columns=:.metadata.name); do
   kubectl annotate node "${node}" "kind.x-k8s.io/registry=localhost:${reg_port}";