bnx2: fixing a timout error due not refreshing TX timers correctly
authorBreno Leitao <leitao@linux.vnet.ibm.com>
Sat, 19 Dec 2009 04:35:34 +0000 (20:35 -0800)
committerDavid S. Miller <davem@davemloft.net>
Sat, 19 Dec 2009 04:35:34 +0000 (20:35 -0800)
When running the following script on an active bnx2 interface:

while(true); do ifconfig ethX mtu 9000; ifconfig ethX mtu 1500; done

A timeout error appears and dumps the following stack:

NETDEV WATCHDOG: eth4 (bnx2): transmit queue 0 timed out
------------[ cut here ]------------
Badness at net/sched/sch_generic.c:261
<snip>

This patch just fixes the way that ->trans_start is refreshed.

Signed-off-by: Breno Leitao <leitao@linux.vnet.ibm.com>
Acked-by: Michael Chan <mchan@broadcom.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/bnx2.c

index c3e2b279fca0731856464d53b58d717ce224fb9e..f38764ec02c3b7f62a548c758ca47a3ad63be78c 100644 (file)
@@ -654,9 +654,17 @@ bnx2_netif_stop(struct bnx2 *bp)
 {
        bnx2_cnic_stop(bp);
        if (netif_running(bp->dev)) {
+               int i;
+
                bnx2_napi_disable(bp);
                netif_tx_disable(bp->dev);
-               bp->dev->trans_start = jiffies; /* prevent tx timeout */
+               /* prevent tx timeout */
+               for (i = 0; i <  bp->dev->num_tx_queues; i++) {
+                       struct netdev_queue *txq;
+
+                       txq = netdev_get_tx_queue(bp->dev, i);
+                       txq->trans_start = jiffies;
+               }
        }
 }