A previous commit moved the ether_addr_copy() in i40e_set_mac() before
the mac filter del/add to avoid a race. However it wasn't taken into
account that this alters the mac address being handed to
i40e_del_mac_filter().

Also changed i40e_add_mac_filter() to operate on netdev->dev_addr,
hopefully that makes the code easier to read.

Fixes: 458867b2ca0c ("i40e: don't remove netdev->dev_addr when syncing uc list")

Signed-off-by: Stefan Assmann <sassm...@kpanic.de>
---
 drivers/net/ethernet/intel/i40e/i40e_main.c | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/drivers/net/ethernet/intel/i40e/i40e_main.c 
b/drivers/net/ethernet/intel/i40e/i40e_main.c
index 6d5b13f69dec..901ef799d2ad 100644
--- a/drivers/net/ethernet/intel/i40e/i40e_main.c
+++ b/drivers/net/ethernet/intel/i40e/i40e_main.c
@@ -1546,17 +1546,17 @@ static int i40e_set_mac(struct net_device *netdev, void 
*p)
                netdev_info(netdev, "set new mac address %pM\n", addr->sa_data);
 
        /* Copy the address first, so that we avoid a possible race with
-        * .set_rx_mode(). If we copy after changing the address in the filter
-        * list, we might open ourselves to a narrow race window where
-        * .set_rx_mode could delete our dev_addr filter and prevent traffic
-        * from passing.
+        * .set_rx_mode().
+        * - Remove old address from MAC filter
+        * - Copy new address
+        * - Add new address to MAC filter
         */
-       ether_addr_copy(netdev->dev_addr, addr->sa_data);
-
        spin_lock_bh(&vsi->mac_filter_hash_lock);
        i40e_del_mac_filter(vsi, netdev->dev_addr);
-       i40e_add_mac_filter(vsi, addr->sa_data);
+       ether_addr_copy(netdev->dev_addr, addr->sa_data);
+       i40e_add_mac_filter(vsi, netdev->dev_addr);
        spin_unlock_bh(&vsi->mac_filter_hash_lock);
+
        if (vsi->type == I40E_VSI_MAIN) {
                i40e_status ret;
 
-- 
2.19.2

Reply via email to