igb, ixgbe: netdev_tx_reset_queue incorrectly called from tx init path

igb and ixgbe incorrectly call netdev_tx_reset_queue() from
i{gb|xgbe}_clean_tx_ring() this sort of works in most cases except
when the number of real tx queues changes. When the number of real
tx queues changes netdev_tx_reset_queue() only gets called on the
new number of queues so when we reduce the number of queues we risk
triggering the watchdog timer and repeated device resets.

So this is not only a cosmetic issue but causes real bugs. For
example enabling/disabling DCB or FCoE in ixgbe will trigger this.

CC: Alexander Duyck <alexander.h.duyck@intel.com>
Signed-off-by: John Fastabend <john.r.fastabend@intel.com>
Tested-by: John Bishop <johnx.bishop@intel.com>
Tested-by: Aaron Brown <aaron.f.brown@intel.com>
Signed-off-by: Jeff Kirsher <jeffrey.t.kirsher@intel.com>

1 files changed, 2 insertions, 2 deletions

diff --git a/drivers/net/ethernet/intel/igb/igb_main.c b/drivers/net/ethernet/intel/igb/igb_main.c
index 5ec31598ee47..d22350055285 100644
--- a/drivers/net/ethernet/intel/igb/igb_main.c
+++ b/drivers/net/ethernet/intel/igb/igb_main.c
@@ -2771,8 +2771,6 @@ void igb_configure_tx_ring(struct igb_adapter *adapter,
 
         txdctl |= E1000_TXDCTL_QUEUE_ENABLE;
         wr32(E1000_TXDCTL(reg_idx), txdctl);
-
-        netdev_tx_reset_queue(txring_txq(ring));
 }
 
 /**
@@ -3282,6 +3280,8 @@ static void igb_clean_tx_ring(struct igb_ring *tx_ring)
                 igb_unmap_and_free_tx_resource(tx_ring, buffer_info);
         }
 
+        netdev_tx_reset_queue(txring_txq(tx_ring));
+
         size = sizeof(struct igb_tx_buffer) * tx_ring->count;
         memset(tx_ring->tx_buffer_info, 0, size);