2016-07-08 15:22 GMT+03:00 Veli-Matti Lintu [off-list ref]:

2016-07-08 4:48 GMT+03:00 Jay Vosburgh [off-list ref]:

quoted

diff --git a/drivers/net/bonding/bond_3ad.c b/drivers/net/bonding/bond_3ad.c
index edc70ffad660..2da5be91226e 100644
--- a/drivers/net/bonding/bond_3ad.c
+++ b/drivers/net/bonding/bond_3ad.c

@@ -1713,18 +1713,12 @@ static void ad_agg_selection_logic(struct aggregator *agg,
                *update_slave_arr = true;
        }

-       /* if the selected aggregator is of join individuals
-        * (partner_system is NULL), enable their ports
-        */
        active = __get_active_agg(origin);

        if (active) {
-               if (!__agg_has_partner(active)) {
-                       for (port = active->lag_ports; port;
-                            port = port->next_port_in_aggregator) {
-                               __enable_port(port);
-                       }
-               }
+               for (port = active->lag_ports; port;
+                    port = port->next_port_in_aggregator)
+                       __enable_port(port);
        }

        rcu_read_unlock();


        Rather than adding a new loop as your original patch did, this

one repurposes the existing loop for individual links and does the
enable for all cases.

        I think this may still need a check for the port state in there,
as ports in the aggregator may be link up but still not suitable for
TX/RX if the LACP negotiation doesn't put the port into COLL_DIST state.

        Enabling a port that is already enabled is harmless, and if the
port is link down, the enable will do nothing, so I think this should
resolve things.

I can test any patches with additional checks, but this change seems
to fix it for us.

Do you think something else is needed is addition to this? This has
been working on my servers now for some time without problems. I
tested this also against mainline 4.8-rc1 that needed it for bonding
to work properly with the previous test case.

Veli-Matti