1 files changed, 80 insertions, 13 deletions

diff --git a/net/ipv6/route.c b/net/ipv6/route.c
index affb51c11a25..4fbdc60b4e07 100644
--- a/net/ipv6/route.c
+++ b/net/ipv6/route.c
@@ -3757,6 +3757,7 @@ static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
 
 	if (rt->fib6_nsiblings && cfg->fc_delete_all_nh) {
 		struct fib6_info *sibling, *next_sibling;
+		struct fib6_node *fn;
 
 		/* prefer to send a single notification with all hops */
 		skb = nlmsg_new(rt6_nlmsg_size(rt), gfp_any());
@@ -3772,12 +3773,32 @@ static int __ip6_del_rt_siblings(struct fib6_info *rt, struct fib6_config *cfg)
 				info->skip_notify = 1;
 		}
 
+		/* 'rt' points to the first sibling route. If it is not the
+		 * leaf, then we do not need to send a notification. Otherwise,
+		 * we need to check if the last sibling has a next route or not
+		 * and emit a replace or delete notification, respectively.
+		 */
 		info->skip_notify_kernel = 1;
-		call_fib6_multipath_entry_notifiers(net,
-						    FIB_EVENT_ENTRY_DEL,
-						    rt,
-						    rt->fib6_nsiblings,
-						    NULL);
+		fn = rcu_dereference_protected(rt->fib6_node,
+					    lockdep_is_held(&table->tb6_lock));
+		if (rcu_access_pointer(fn->leaf) == rt) {
+			struct fib6_info *last_sibling, *replace_rt;
+
+			last_sibling = list_last_entry(&rt->fib6_siblings,
+						       struct fib6_info,
+						       fib6_siblings);
+			replace_rt = rcu_dereference_protected(
+					    last_sibling->fib6_next,
+					    lockdep_is_held(&table->tb6_lock));
+			if (replace_rt)
+				call_fib6_entry_notifiers_replace(net,
+								  replace_rt);
+			else
+				call_fib6_multipath_entry_notifiers(net,
+						       FIB_EVENT_ENTRY_DEL,
+						       rt, rt->fib6_nsiblings,
+						       NULL);
+		}
 		list_for_each_entry_safe(sibling, next_sibling,
 					 &rt->fib6_siblings,
 					 fib6_siblings) {
@@ -5025,12 +5046,37 @@ static void ip6_route_mpath_notify(struct fib6_info *rt,
 		inet6_rt_notify(RTM_NEWROUTE, rt, info, nlflags);
 }
 
+static bool ip6_route_mpath_should_notify(const struct fib6_info *rt)
+{
+	bool rt_can_ecmp = rt6_qualify_for_ecmp(rt);
+	bool should_notify = false;
+	struct fib6_info *leaf;
+	struct fib6_node *fn;
+
+	rcu_read_lock();
+	fn = rcu_dereference(rt->fib6_node);
+	if (!fn)
+		goto out;
+
+	leaf = rcu_dereference(fn->leaf);
+	if (!leaf)
+		goto out;
+
+	if (rt == leaf ||
+	    (rt_can_ecmp && rt->fib6_metric == leaf->fib6_metric &&
+	     rt6_qualify_for_ecmp(leaf)))
+		should_notify = true;
+out:
+	rcu_read_unlock();
+
+	return should_notify;
+}
+
 static int ip6_route_multipath_add(struct fib6_config *cfg,
 				   struct netlink_ext_ack *extack)
 {
 	struct fib6_info *rt_notif = NULL, *rt_last = NULL;
 	struct nl_info *info = &cfg->fc_nlinfo;
-	enum fib_event_type event_type;
 	struct fib6_config r_cfg;
 	struct rtnexthop *rtnh;
 	struct fib6_info *rt;
@@ -5155,13 +5201,27 @@ static int ip6_route_multipath_add(struct fib6_config *cfg,
 		nhn++;
 	}
 
-	event_type = replace ? FIB_EVENT_ENTRY_REPLACE : FIB_EVENT_ENTRY_ADD;
-	err = call_fib6_multipath_entry_notifiers(info->nl_net, event_type,
-						  rt_notif, nhn - 1, extack);
-	if (err) {
-		/* Delete all the siblings that were just added */
-		err_nh = NULL;
-		goto add_errout;
+	/* An in-kernel notification should only be sent in case the new
+	 * multipath route is added as the first route in the node, or if
+	 * it was appended to it. We pass 'rt_notif' since it is the first
+	 * sibling and might allow us to skip some checks in the replace case.
+	 */
+	if (ip6_route_mpath_should_notify(rt_notif)) {
+		enum fib_event_type fib_event;
+
+		if (rt_notif->fib6_nsiblings != nhn - 1)
+			fib_event = FIB_EVENT_ENTRY_APPEND;
+		else
+			fib_event = FIB_EVENT_ENTRY_REPLACE;
+
+		err = call_fib6_multipath_entry_notifiers(info->nl_net,
+							  fib_event, rt_notif,
+							  nhn - 1, extack);
+		if (err) {
+			/* Delete all the siblings that were just added */
+			err_nh = NULL;
+			goto add_errout;
+		}
 	}
 
 	/* success ... tell user about new route */
@@ -5516,6 +5576,13 @@ static int rt6_fill_node(struct net *net, struct sk_buff *skb,
 		expires -= jiffies;
 	}
 
+	if (!dst) {
+		if (rt->offload)
+			rtm->rtm_flags |= RTM_F_OFFLOAD;
+		if (rt->trap)
+			rtm->rtm_flags |= RTM_F_TRAP;
+	}
+
 	if (rtnl_put_cacheinfo(skb, dst, 0, expires, dst ? dst->error : 0) < 0)
 		goto nla_put_failure;