do not prematurely release the main dev on link loss, should fix DHCP NAK handling

diff --git a/interface.c b/interface.c
index 4a42ad9a330fd4e5c4b4920411ba4bd00c43cf6f..245622766b4e14a33e22601b049df759adfb8af4 100644 (file)
--- a/interface.c
+++ b/interface.c
@@ -170,8 +170,6 @@ interface_event(struct interface *iface, enum interface_event ev)
 static void
 interface_flush_state(struct interface *iface)
 {
-       if (iface->main_dev.dev)
-               device_release(&iface->main_dev);
        if (iface->l3_dev.dev)
                device_release(&iface->l3_dev);
        interface_data_flush(iface);
@@ -404,6 +402,8 @@ interface_proto_cb(struct interface_proto_state *state, enum interface_proto_eve
 
                netifd_log_message(L_NOTICE, "Interface '%s' is now down\n", iface->name);
                mark_interface_down(iface);
+               if (iface->main_dev.dev)
+                       device_release(&iface->main_dev);
                interface_handle_config_change(iface);
                break;
        case IFPEV_LINK_LOST: