ibmveth: Fix leak when recycling skb and hypervisor returns error
authorAnton Blanchard <anton@samba.org>
Thu, 25 Aug 2011 00:56:15 +0000 (17:56 -0700)
committerDavid S. Miller <davem@davemloft.net>
Thu, 25 Aug 2011 00:56:15 +0000 (17:56 -0700)
If h_add_logical_lan_buffer returns an error we need to free
the skb.

Signed-off-by: Anton Blanchard <anton@samba.org>
Cc: stable <stable@kernel.org>
Signed-off-by: David S. Miller <davem@davemloft.net>
drivers/net/ibmveth.c

index ba99af05bf62744148a4d33d008c04ddece97a98..3e6679269400d95aa6c169443f7d55510844b6ce 100644 (file)
@@ -395,7 +395,7 @@ static inline struct sk_buff *ibmveth_rxq_get_buffer(struct ibmveth_adapter *ada
 }
 
 /* recycle the current buffer on the rx queue */
-static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
+static int ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
 {
        u32 q_index = adapter->rx_queue.index;
        u64 correlator = adapter->rx_queue.queue_addr[q_index].correlator;
@@ -403,6 +403,7 @@ static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
        unsigned int index = correlator & 0xffffffffUL;
        union ibmveth_buf_desc desc;
        unsigned long lpar_rc;
+       int ret = 1;
 
        BUG_ON(pool >= IBMVETH_NUM_BUFF_POOLS);
        BUG_ON(index >= adapter->rx_buff_pool[pool].size);
@@ -410,7 +411,7 @@ static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
        if (!adapter->rx_buff_pool[pool].active) {
                ibmveth_rxq_harvest_buffer(adapter);
                ibmveth_free_buffer_pool(adapter, &adapter->rx_buff_pool[pool]);
-               return;
+               goto out;
        }
 
        desc.fields.flags_len = IBMVETH_BUF_VALID |
@@ -423,12 +424,16 @@ static void ibmveth_rxq_recycle_buffer(struct ibmveth_adapter *adapter)
                netdev_dbg(adapter->netdev, "h_add_logical_lan_buffer failed "
                           "during recycle rc=%ld", lpar_rc);
                ibmveth_remove_buffer_from_pool(adapter, adapter->rx_queue.queue_addr[adapter->rx_queue.index].correlator);
+               ret = 0;
        }
 
        if (++adapter->rx_queue.index == adapter->rx_queue.num_slots) {
                adapter->rx_queue.index = 0;
                adapter->rx_queue.toggle = !adapter->rx_queue.toggle;
        }
+
+out:
+       return ret;
 }
 
 static void ibmveth_rxq_harvest_buffer(struct ibmveth_adapter *adapter)
@@ -1084,8 +1089,9 @@ restart_poll:
                                if (rx_flush)
                                        ibmveth_flush_buffer(skb->data,
                                                length + offset);
+                               if (!ibmveth_rxq_recycle_buffer(adapter))
+                                       kfree_skb(skb);
                                skb = new_skb;
-                               ibmveth_rxq_recycle_buffer(adapter);
                        } else {
                                ibmveth_rxq_harvest_buffer(adapter);
                                skb_reserve(skb, offset);