Hi Radu From: Radu Nicolau > Signed-off-by: Radu Nicolau <radu.nico...@intel.com> > --- > drivers/net/bonding/rte_eth_bond_8023ad.c | 9 +++++++-- > 1 file changed, 7 insertions(+), 2 deletions(-) > > diff --git a/drivers/net/bonding/rte_eth_bond_8023ad.c > b/drivers/net/bonding/rte_eth_bond_8023ad.c > index f8cea4b..730087f 100644 > --- a/drivers/net/bonding/rte_eth_bond_8023ad.c > +++ b/drivers/net/bonding/rte_eth_bond_8023ad.c > @@ -917,7 +917,7 @@ bond_mode_8023ad_activate_slave(struct rte_eth_dev > *bond_dev, > }; > > char mem_name[RTE_ETH_NAME_MAX_LEN]; > - int socket_id; > + int socket_id, ret; > unsigned element_size; > uint32_t total_tx_desc; > struct bond_tx_queue *bd_tx_q; > @@ -942,7 +942,12 @@ bond_mode_8023ad_activate_slave(struct > rte_eth_dev *bond_dev, > > /* use this port as agregator */ > port->aggregator_port_id = slave_id; > - rte_eth_promiscuous_enable(slave_id); > + > + /* try to enable multicast, if fail set promiscuous */ > + rte_eth_allmulticast_enable(slave_id); > + ret = rte_eth_allmulticast_get(slave_id); > + if (ret != 1) > + rte_eth_promiscuous_enable(slave_id);
It is still greedy to configure allmulticast for LACP, The user may not expect to get multicast traffic. Moreover, here each slave can be with different unexpected configuration, may be even worst from the application perspective. > > timer_cancel(&port->warning_timer); > > -- > 2.7.5
