return NETDEV_TX_OK;
 }
 
+static void ipgre_link_update(struct net_device *dev, bool set_mtu)
+{
+       struct ip_tunnel *tunnel = netdev_priv(dev);
+       int len;
+
+       len = tunnel->tun_hlen;
+       tunnel->tun_hlen = gre_calc_hlen(tunnel->parms.o_flags);
+       len = tunnel->tun_hlen - len;
+       tunnel->hlen = tunnel->hlen + len;
+
+       dev->needed_headroom = dev->needed_headroom + len;
+       if (set_mtu)
+               dev->mtu = max_t(int, dev->mtu - len, 68);
+
+       if (!(tunnel->parms.o_flags & TUNNEL_SEQ)) {
+               if (!(tunnel->parms.o_flags & TUNNEL_CSUM) ||
+                   tunnel->encap.type == TUNNEL_ENCAP_NONE) {
+                       dev->features |= NETIF_F_GSO_SOFTWARE;
+                       dev->hw_features |= NETIF_F_GSO_SOFTWARE;
+               }
+               dev->features |= NETIF_F_LLTX;
+       }
+}
+
 static int ipgre_tunnel_ioctl(struct net_device *dev,
                              struct ifreq *ifr, int cmd)
 {
                            struct netlink_ext_ack *extack)
 {
        struct ip_tunnel *t = netdev_priv(dev);
-       struct ip_tunnel_parm p;
        struct ip_tunnel_encap ipencap;
        __u32 fwmark = t->fwmark;
+       struct ip_tunnel_parm p;
        int err;
 
        if (ipgre_netlink_encap_parms(data, &ipencap)) {
        err = ipgre_netlink_parms(dev, data, tb, &p, &fwmark);
        if (err < 0)
                return err;
-       return ip_tunnel_changelink(dev, tb, &p, fwmark);
+
+       err = ip_tunnel_changelink(dev, tb, &p, fwmark);
+       if (err < 0)
+               return err;
+
+       t->parms.i_flags = p.i_flags;
+       t->parms.o_flags = p.o_flags;
+
+       if (strcmp(dev->rtnl_link_ops->kind, "erspan"))
+               ipgre_link_update(dev, !tb[IFLA_MTU]);
+
+       return 0;
 }
 
 static size_t ipgre_get_size(const struct net_device *dev)