Allow to remove pod sandbox on netns removal

saschagrunert · openshift-cherrypick-robot · commit 525b181a2058 · 2024-12-04T09:23:45.000Z
We should not error on sandbox removal if the netns path gets either
removed or invalidated. This path allows removing the sandbox and cleans
up the stale netns path.

Signed-off-by: Sascha Grunert &lt;sgrunert@redhat.com&gt;
diff --git a/internal/config/nsmgr/types_linux.go b/internal/config/nsmgr/types_linux.go
@@ -83,11 +83,17 @@ func (n *namespace) Remove() error {
 		return nil
 	}
 
-	// try to unmount, ignoring "not mounted" (EINVAL) error.
-	if err := unix.Unmount(fp, unix.MNT_DETACH); err != nil && err != unix.EINVAL {
-		return fmt.Errorf("unable to unmount %s: %w", fp, err)
+	// Don't run into unmount issues if the network namespace does not exist any more.
+	if _, err := os.Stat(fp); err == nil {
+		// try to unmount, ignoring "not mounted" (EINVAL) error.
+		if err := unix.Unmount(fp, unix.MNT_DETACH); err != nil && err != unix.EINVAL {
+			return fmt.Errorf("unable to unmount %s: %w", fp, err)
+		}
+
+		return os.RemoveAll(fp)
 	}
-	return os.Remove(fp)
+
+	return nil
 }
 
 // GetNamespace takes a path and type, checks if it is a namespace, and if so
diff --git a/server/sandbox_network.go b/server/sandbox_network.go
@@ -4,6 +4,7 @@ import (
 	"context"
 	"fmt"
 	"math"
+	"os"
 	"time"
 
 	cnitypes "github.com/containernetworking/cni/pkg/types"
@@ -187,7 +188,19 @@ func (s *Server) networkStop(ctx context.Context, sb *sandbox.Sandbox) error {
 		return err
 	}
 	if err := s.config.CNIPlugin().TearDownPodWithContext(stopCtx, podNetwork); err != nil {
-		return fmt.Errorf("failed to destroy network for pod sandbox %s(%s): %w", sb.Name(), sb.ID(), err)
+		retErr := fmt.Errorf("failed to destroy network for pod sandbox %s(%s): %w", sb.Name(), sb.ID(), err)
+
+		if _, statErr := os.Stat(podNetwork.NetNS); statErr != nil {
+			return fmt.Errorf("%w: stat netns path %q: %w", retErr, podNetwork.NetNS, statErr)
+		}
+
+		// The netns file may still exists, which means that it's likely
+		// corrupted. Remove it to allow cleanup of the network namespace:
+		if rmErr := os.RemoveAll(podNetwork.NetNS); rmErr != nil {
+			return fmt.Errorf("%w: failed to remove netns path: %w", retErr, rmErr)
+		}
+
+		log.Warnf(ctx, "Removed invalid netns path %s from pod sandbox %s(%s)", podNetwork.NetNS, sb.Name(), sb.ID())
 	}
 
 	return sb.SetNetworkStopped(ctx, true)
diff --git a/test/network.bats b/test/network.bats
@@ -166,3 +166,24 @@ function check_networking() {
 	# shellcheck disable=SC2010
 	[[ $(ls $CNI_RESULTS_DIR | grep "$POD") == "" ]]
 }
+
+@test "Clean up network if pod netns gets destroyed" {
+	start_crio
+
+	POD=$(crictl runp "$TESTDATA/sandbox_config.json")
+
+	# remove the network namespace
+	NETNS_PATH=/var/run/netns/
+	NS=$(crictl inspectp "$POD" |
+		jq -er '.info.runtimeSpec.linux.namespaces[] | select(.type == "network").path | sub("'$NETNS_PATH'"; "")')
+
+	# remove network namespace
+	ip netns del "$NS"
+
+	# fake invalid netns path
+	touch "$NETNS_PATH$NS"
+
+	# be able to remove the sandbox
+	crictl rmp -f "$POD"
+	grep -q "Removed invalid netns path $NETNS_PATH$NS from pod sandbox" "$CRIO_LOG"
+}
