nfp: flower: ensure dead neighbour entries are not offloaded

[ Upstream commit b809ec869b2cf2af053ffd99e5a46ab600e94aa2 ]

Previously only the neighbour state was checked to decide if an offloaded
entry should be removed. However, there can be situations when the entry
is dead but still marked as valid. This can lead to dead entries not
being removed from fw tables or even incorrect data being added.

Check the entry dead bit before deciding if it should be added to or
removed from fw neighbour tables.

Fixes: 8e6a9046b66a ("nfp: flower vxlan neighbour offload")
Signed-off-by: John Hurley <john.hurley@netronome.com>
Reviewed-by: Jakub Kicinski <jakub.kicinski@netronome.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
Signed-off-by: Greg Kroah-Hartman <gregkh@linuxfoundation.org>

diff --git a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
index ec524d97869d6fdf12160eda47aa4043aa4fedeb..5ef61132604ef8839e6813fad8d3517249212c2f 100644 (file)
--- a/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
+++ b/drivers/net/ethernet/netronome/nfp/flower/tunnel_conf.c
@@ -317,7 +317,7 @@ nfp_tun_write_neigh(struct net_device *netdev, struct nfp_app *app,
        payload.dst_ipv4 = flow->daddr;
 
        /* If entry has expired send dst IP with all other fields 0. */
-       if (!(neigh->nud_state & NUD_VALID)) {
+       if (!(neigh->nud_state & NUD_VALID) || neigh->dead) {
                nfp_tun_del_route_from_cache(app, payload.dst_ipv4);
                /* Trigger ARP to verify invalid neighbour state. */
                neigh_event_send(neigh, NULL);