xen-netfront: fix oops when disconnected from backend
authorDavid Vrabel <david.vrabel@citrix.com>
Wed, 18 Jun 2014 09:47:27 +0000 (10:47 +0100)
committerDavid S. Miller <davem@davemloft.net>
Sat, 21 Jun 2014 23:14:26 +0000 (16:14 -0700)
xennet_disconnect_backend() was not correctly iterating over all the
queues.

Signed-off-by: David Vrabel <david.vrabel@citrix.com>
Reviewed-by: Wei Liu <wei.liu2@citrix.com>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/xen-netfront.c

index 5a7872a..daaf1e5 100644 (file)
@@ -1287,7 +1287,7 @@ static irqreturn_t xennet_rx_interrupt(int irq, void *dev_id)
 
        if (likely(netif_carrier_ok(dev) &&
                   RING_HAS_UNCONSUMED_RESPONSES(&queue->rx)))
-                       napi_schedule(&queue->napi);
+               napi_schedule(&queue->napi);
 
        return IRQ_HANDLED;
 }
@@ -1437,10 +1437,11 @@ static void xennet_end_access(int ref, void *page)
 static void xennet_disconnect_backend(struct netfront_info *info)
 {
        unsigned int i = 0;
-       struct netfront_queue *queue = NULL;
        unsigned int num_queues = info->netdev->real_num_tx_queues;
 
        for (i = 0; i < num_queues; ++i) {
+               struct netfront_queue *queue = &info->queues[i];
+
                /* Stop old i/f to prevent errors whilst we rebuild the state. */
                spin_lock_bh(&queue->rx_lock);
                spin_lock_irq(&queue->tx_lock);