mlxsw: spectrum_router: Do not assume RTNL is taken when resolving underlay device
authorIdo Schimmel <idosch@mellanox.com>
Thu, 20 Feb 2020 07:07:55 +0000 (09:07 +0200)
committerDavid S. Miller <davem@davemloft.net>
Thu, 20 Feb 2020 18:04:34 +0000 (10:04 -0800)
The function that resolves the underlay device of the IPIP tunnel
assumes that RTNL is taken, but this will not be correct when RTNL is
removed from the route insertion path.

Convert the function to use dev_get_by_index_rcu() instead of
__dev_get_by_index() and make sure it is always called from an RCU
read-side critical section.

Signed-off-by: Ido Schimmel <idosch@mellanox.com>
Acked-by: Jiri Pirko <jiri@mellanox.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ethernet/mellanox/mlxsw/spectrum_router.c

index 21e8539727beb7cdf6a8358f0c0445014ff374c2..000aa68e7e316b2935c340b86612c9c3f2e5b356 100644 (file)
@@ -988,17 +988,23 @@ __mlxsw_sp_ipip_netdev_ul_dev_get(const struct net_device *ol_dev)
        struct ip_tunnel *tun = netdev_priv(ol_dev);
        struct net *net = dev_net(ol_dev);
 
-       return __dev_get_by_index(net, tun->parms.link);
+       return dev_get_by_index_rcu(net, tun->parms.link);
 }
 
 u32 mlxsw_sp_ipip_dev_ul_tb_id(const struct net_device *ol_dev)
 {
-       struct net_device *d = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
+       struct net_device *d;
+       u32 tb_id;
 
+       rcu_read_lock();
+       d = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
        if (d)
-               return l3mdev_fib_table(d) ? : RT_TABLE_MAIN;
+               tb_id = l3mdev_fib_table(d) ? : RT_TABLE_MAIN;
        else
-               return RT_TABLE_MAIN;
+               tb_id = RT_TABLE_MAIN;
+       rcu_read_unlock();
+
+       return tb_id;
 }
 
 static struct mlxsw_sp_rif *
@@ -1355,8 +1361,12 @@ mlxsw_sp_ipip_entry_find_by_ul_dev(const struct mlxsw_sp *mlxsw_sp,
                                        ipip_list_node);
        list_for_each_entry_continue(ipip_entry, &mlxsw_sp->router->ipip_list,
                                     ipip_list_node) {
-               struct net_device *ipip_ul_dev =
-                       __mlxsw_sp_ipip_netdev_ul_dev_get(ipip_entry->ol_dev);
+               struct net_device *ol_dev = ipip_entry->ol_dev;
+               struct net_device *ipip_ul_dev;
+
+               rcu_read_lock();
+               ipip_ul_dev = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
+               rcu_read_unlock();
 
                if (ipip_ul_dev == ul_dev)
                        return ipip_entry;
@@ -1722,9 +1732,12 @@ static void mlxsw_sp_ipip_demote_tunnel_by_ul_netdev(struct mlxsw_sp *mlxsw_sp,
 
        list_for_each_entry_safe(ipip_entry, tmp, &mlxsw_sp->router->ipip_list,
                                 ipip_list_node) {
-               struct net_device *ipip_ul_dev =
-                       __mlxsw_sp_ipip_netdev_ul_dev_get(ipip_entry->ol_dev);
+               struct net_device *ol_dev = ipip_entry->ol_dev;
+               struct net_device *ipip_ul_dev;
 
+               rcu_read_lock();
+               ipip_ul_dev = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
+               rcu_read_unlock();
                if (ipip_ul_dev == ul_dev)
                        mlxsw_sp_ipip_entry_demote_tunnel(mlxsw_sp, ipip_entry);
        }
@@ -3711,9 +3724,15 @@ static void mlxsw_sp_nexthop_neigh_fini(struct mlxsw_sp *mlxsw_sp,
 
 static bool mlxsw_sp_ipip_netdev_ul_up(struct net_device *ol_dev)
 {
-       struct net_device *ul_dev = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
+       struct net_device *ul_dev;
+       bool is_up;
+
+       rcu_read_lock();
+       ul_dev = __mlxsw_sp_ipip_netdev_ul_dev_get(ol_dev);
+       is_up = ul_dev ? (ul_dev->flags & IFF_UP) : true;
+       rcu_read_unlock();
 
-       return ul_dev ? (ul_dev->flags & IFF_UP) : true;
+       return is_up;
 }
 
 static void mlxsw_sp_nexthop_ipip_init(struct mlxsw_sp *mlxsw_sp,