On Sun, Dec 20, 2015 at 02:17:47PM +0100, Sven Eckelmann wrote:
@@ -278,8 +258,8 @@ static void batadv_neigh_node_free_rcu(struct rcu_head *rcu) neigh_node->addr); if (hardif_neigh) { /* batadv_hardif_neigh_get() increases refcount too */
batadv_hardif_neigh_free_now(hardif_neigh);
batadv_hardif_neigh_free_now(hardif_neigh);
batadv_hardif_neigh_free_ref(hardif_neigh);
}batadv_hardif_neigh_free_ref(hardif_neigh);
Hm, this trades one bug for an intermediate one, a nested call-rcu. Swapping this patch with patch number 9 to avoid this? (tried that an the diffs seem cleaner to me after swapping; or am I missing something?)
Despite from that, both patches 7 and 9 seem sane to me.