On Saturday 15 February 2014 17:47:53 Linus Lüssing wrote:
+static void batadv_mcast_want_unsnoop_update(struct batadv_priv *bat_priv,
struct batadv_orig_node *orig,
uint8_t mcast_flags)
+{
- /* switched from flag unset to set */
- if (mcast_flags & BATADV_MCAST_WANT_ALL_UNSNOOPABLES &&
!(orig->mcast_flags & BATADV_MCAST_WANT_ALL_UNSNOOPABLES)) {
atomic_inc(&bat_priv->mcast.num_want_all_unsnoopables);
spin_lock_bh(&bat_priv->mcast.want_lists_lock);
hlist_add_head_rcu(&orig->mcast_want_all_unsnoopables_node,
&bat_priv->mcast.want_all_unsnoopables_list);
spin_unlock_bh(&bat_priv->mcast.want_lists_lock);
- /* switched from flag set to unset */
- } else if (!(mcast_flags & BATADV_MCAST_WANT_ALL_UNSNOOPABLES) &&
orig->mcast_flags & BATADV_MCAST_WANT_ALL_UNSNOOPABLES) {
atomic_dec(&bat_priv->mcast.num_want_all_unsnoopables);
spin_lock_bh(&bat_priv->mcast.want_lists_lock);
hlist_del_rcu(&orig->mcast_want_all_unsnoopables_node);
spin_unlock_bh(&bat_priv->mcast.want_lists_lock);
- }
Looks unsafe. How do you make sure that it is still in the list and not already already removed in the meantime? hlist_del_rcu does not re-initialize the pointers (only POISON them when CONFIG_DEBUG_LIST is activated). Thus calling hlist_del_rcu again on an object not in the list either kills other data structures or causes an Oops.
There are more *list_del* related problems of that type in your code. This is just one example I've picked.
See http://www.open-mesh.org/issues/217#note-5
Kind regards, Sven