Le dimanche 18 mars 2012 à 19:24 -0400, Paul Gortmaker a écrit :

On Sun, Mar 18, 2012 at 5:55 PM, Eric Dumazet [off-list ref] wrote:

quoted

On Sun, 2012-03-18 at 17:39 -0400, Paul Gortmaker wrote:

quoted

The __netif_subqueue_stopped() just does the following:

        struct netdev_queue *txq = netdev_get_tx_queue(dev, queue_index);
        return netif_tx_queue_stopped(txq);

and since we already have the txq in scope, we can just call that
directly in this case.

Suggested-by: Eric Dumazet <redacted>
Signed-off-by: Paul Gortmaker <redacted>
---
 drivers/net/ethernet/freescale/gianfar.c |    2 +-
 1 files changed, 1 insertions(+), 1 deletions(-)

diff --git a/drivers/net/ethernet/freescale/gianfar.c b/drivers/net/ethernet/freescale/gianfar.c
index 6e66cc3..d9428f0 100644
--- a/drivers/net/ethernet/freescale/gianfar.c
+++ b/drivers/net/ethernet/freescale/gianfar.c

@@ -2565,7 +2565,7 @@ static int gfar_clean_tx_ring(struct gfar_priv_tx_q *tx_queue)
      }

      /* If we freed a buffer, we can restart transmission, if necessary */
-     if (__netif_subqueue_stopped(dev, tqi) && tx_queue->num_txbdfree)
+     if (netif_tx_queue_stopped(txq) && tx_queue->num_txbdfree)
              netif_wake_subqueue(dev, tqi);

      /* Update dirty indicators */

Please change netif_wake_subqueue() as well ;)

I looked at this earlier when I added patch #4 but I was concerned about
the different semantics.

The netif_wake_subqueue() just returns on a netpoll_trap but the other
netif_tx_wake_queue() actually calls netif_tx_start_queue() for the same
netpoll_trap instance.  Maybe that is OK, but I didn't want to be changing
the behaviour of subtleties in stuff where I am clearly still learning.

I see... commit 7b3d3e4fc68 added a small difference here...