nexthop: Only emit a notification when nexthop is actually deleted

Currently, the in-kernel delete notification is emitted from the error
path of nexthop_add() and replace_nexthop(), which can be confusing to
in-kernel listeners as they are not familiar with the nexthop.

Instead, only emit the notification when the nexthop is actually
deleted. The following sub-cases are covered:

1. User space deletes the nexthop
2. The nexthop is deleted by the kernel due to a netdev event (e.g.,
   nexthop device going down)
3. A group is deleted because its last nexthop is being deleted
4. The network namespace of the nexthop device is deleted

Signed-off-by: Ido Schimmel <idosch@nvidia.com>
Reviewed-by: David Ahern <dsahern@gmail.com>
Signed-off-by: David S. Miller <davem@davemloft.net>

diff --git a/net/ipv4/nexthop.c b/net/ipv4/nexthop.c
index 13d9219..8c0f17c 100644 (file)
--- a/net/ipv4/nexthop.c
+++ b/net/ipv4/nexthop.c
@@ -870,8 +870,6 @@ static void __remove_nexthop_fib(struct net *net, struct nexthop *nh)
        bool do_flush = false;
        struct fib_info *fi;
 
-       call_nexthop_notifiers(net, NEXTHOP_EVENT_DEL, nh);
-
        list_for_each_entry(fi, &nh->fi_list, nh_list) {
                fi->fib_flags |= RTNH_F_DEAD;
                do_flush = true;
@@ -909,6 +907,8 @@ static void __remove_nexthop(struct net *net, struct nexthop *nh,
 static void remove_nexthop(struct net *net, struct nexthop *nh,
                           struct nl_info *nlinfo)
 {
+       call_nexthop_notifiers(net, NEXTHOP_EVENT_DEL, nh);
+
        /* remove from the tree */
        rb_erase(&nh->rb_node, &net->nexthop.rb_root);