On 8/1/2018 2:44 PM, Chas Williams wrote:

On Wed, Aug 1, 2018 at 9:25 AM Radu Nicolau <radu.nicolau@intel.com 
<mailto:radu.nicolau@intel.com>> wrote:

    When a bonding port is stopped also stop and deactivate all slaves.
    Otherwise slaves will be still listed as active.


I have to think about this for a bit.  The last time I tried this I 
had a problem
because nothing activated the slaves again in 802.3ad mode because we
use an external state machine.

I did test it for mode 4, stop and start seems to be working as 
expected. From what is see, activate_slave will be either called 
directly or indirectly through the LSC when port is restarted.

    Fixes: 69bce062132b ("net/bonding: do not clear active slave count")
    Cc: stable@dpdk.org <mailto:stable@dpdk.org>

    Signed-off-by: Radu Nicolau <radu.nicolau@intel.com
    <mailto:radu.nicolau@intel.com>>
    ---
     drivers/net/bonding/rte_eth_bond_pmd.c | 11 +++++++----
     1 file changed, 7 insertions(+), 4 deletions(-)

    diff --git a/drivers/net/bonding/rte_eth_bond_pmd.c
    b/drivers/net/bonding/rte_eth_bond_pmd.c
    index 16105cb..960140c 100644
    --- a/drivers/net/bonding/rte_eth_bond_pmd.c
    +++ b/drivers/net/bonding/rte_eth_bond_pmd.c
    @@ -2229,12 +2229,15 @@ bond_ethdev_stop(struct rte_eth_dev *eth_dev)
    tlb_last_obytets[internals->active_slaves[i]] = 0;
            }

    -       internals->link_status_polling_enabled = 0;
    -       for (i = 0; i < internals->slave_count; i++)
    -               internals->slaves[i].last_link_status = 0;
    -
            eth_dev->data->dev_link.link_status = ETH_LINK_DOWN;
            eth_dev->data->dev_started = 0;
    +
    +       internals->link_status_polling_enabled = 0;
    +       for (i = 0; i < internals->slave_count; i++) {
    +               internals->slaves[i].last_link_status = 0;
    +  rte_eth_dev_stop(internals->slaves[i].port_id);
    +               deactivate_slave(eth_dev,
    internals->slaves[i].port_id);
    +       }
     }

     void
    -- 
    2.7.5