Currently during NIC profile disablement all VXLAN udp ports offloaded to the
HW are flushed and during its enablement the driver send notification to
the stack to inform the core that the entire UDP tunnel port state has been
lost, uplink representor doesn't have the same behavior which can cause
VXLAN udp ports offload to be in bad state while moving between modes while
VXLAN interface exist.
Fixed by aligning the uplink representor profile behavior to the NIC behavior.
Fixes: d81437200cd4 ("net/mlx5e: Move set vxlan nic info to profile init")
Signed-off-by: Maor Dickman <maord@nvidia.com>
Reviewed-by: Roi Dayan <roid@nvidia.com>
Signed-off-by: Saeed Mahameed <saeedm@nvidia.com>
#include "fs_core.h"
#include "lib/mlx5.h"
#include "lib/devcom.h"
+#include "lib/vxlan.h"
#define CREATE_TRACE_POINTS
#include "diag/en_rep_tracepoint.h"
#include "en_accel/ipsec.h"
rtnl_lock();
if (netif_running(netdev))
mlx5e_open(netdev);
+ udp_tunnel_nic_reset_ntf(priv->netdev);
netif_device_attach(netdev);
rtnl_unlock();
}
mlx5_notifier_unregister(mdev, &priv->events_nb);
mlx5e_rep_tc_disable(priv);
mlx5_lag_remove_netdev(mdev, priv->netdev);
+ mlx5_vxlan_reset_to_default(mdev->vxlan);
}
static MLX5E_DEFINE_STATS_GRP(sw_rep, 0);